The gut microbiome, single nucleotide polymorphisms, and differentially expressed genes promote aggression in an ant

The gut microbiome, single nucleotide polymorphisms, and differentially expressed genes promote aggression in an ant | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The gut microbiome, single nucleotide polymorphisms, and differentially expressed genes promote aggression in an ant Patrick Krapf, Francesco Cicconardi, Martin Schilling, Gerhard Aigner, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8539228/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Animals frequently display aggressive behaviour, for example, when competing for food. Aggression is influenced by various extrinsic and intrinsic factors such as temperature, the microbiome, and genetics. However, we currently lack understanding what factors cause an animal to start aggression. Here, we use an ant species to test if chemical, microbiome, genomic, and/or transcriptomic traits correlate with the start of aggression and the reactions to it, that is, reacting aggressively or peacefully. We found nine bacterial operational taxonomic units, mutations in two genes, and eight differentially expressed genes, which were positively or negatively associated with the start of aggression or reactions to it. These traits are mainly linked to hormone signalling and neurological and synaptic functions. The results indicate that multiple traits, possibly acting in concert, affect the start of aggression and reactions to it. We speculate that such traits could promote aggression and could thus play important evolutionary roles. Biological sciences/Zoology/Animal behaviour Biological sciences/Genetics Biological sciences/Ecology/Behavioural ecology Whole-genome sequencing transcriptomics gut microbiome cuticular hydrocarbons behaviour start of aggression Tetramorium alpestre Figures Figure 1 Figure 2 Figure 2 Figure 3 Introduction Aggressive behaviour among individuals of the same species is a frequently observed behaviour in animals 1 . It is a vital aspect of animals’ fitness and survival and often context-dependent 2 , 3 . For example, it can occur during food or mate competition, territory defence, and offspring protection against predators 4 . Such adaptive aggression 3 , 5 can lead to increased fitness. For instance, winners of fights can consume more or higher-quality food or obtain mates for reproduction 1 . However, aggression can incur harms such as stress and energy or time costs. At its worst, it can also be deadly 6 by increasing the risk of injuries and/or exposure to predators 7 . Various extrinsic and intrinsic factors can lead to aggression. Extrinsic factors are, among others, higher ambient temperature and can lead to increased aggression in humans and animals 8 , 9 . Intrinsic factors such as experience (i.e., repeated stimuli such as winning aggressive encounters) 10 , neurochemical factors (i.e. changes in serotonin, dopamine, or octopamine) 3 , or differentially-expressed genes (DEGs) 3 influenced by the gut microbiome can also promote aggression 11 – 13 . Despite these promising insights, our understanding of the underlying mechanisms that lead to the start of aggression (i.e., when two individuals meet and one starts aggressive behaviours such as fighting) is limited. Nevertheless, some drivers are known: for example, individual experience 14 , 15 , previous experience in winning a fight 16 , or recognising another individual 17 can affect whether an individual starts aggression. In particular, animals such as insects use chemical cues 18 (cuticular hydrocarbons; CHCs) to recognise and attack enemies 19 . Besides experience and recognition, the microbiome 11–13,20−22 , genetic changes (e.g., mutations in genes 23 , 24 ), and/or DEGs (e.g., in neuronal or synaptic functions 17 ) may also affect whether individuals start aggression. Ants are known for their aggressive behaviour. For example, California harvester ants ( Pogonomyrmex californicus ) often fight for over 30 minutes, and such fights often result in fatal outcomes with one or both workers dying 25 . On the other end of this spectrum are ‘peaceful’ ants, which frequently refrain from fighting individuals from different colonies of the same species. Peacefull behaviour is less frequently observed, but is known from several species such as Lasius austriacus 26 , Lasius flavus 27 , and Tetramorium alpestre 9 . However, even in such predominantly peaceful species, aggression can be observed, leading to the unresolved question of what factors lead to the start of aggression 28 , 29 . Here, we used the high-elevation ant species T. alpestre to test whether chemical, microbiome, genomic, and/or transcriptomic traits correlate with the start of aggression in ants, specifically workers. This species displays a behavioural continuum ranging from aggression to peacefulness 9 , 30 , 31 . We collected workers from three colonies each from three previously described populations 9 , 30 , 31 . They either comprise single-queened and aggressive colonies (SQ-A), single-queened and non-aggressive colonies (SQ-N), or multiple-queened and non-aggressive colonies MQ-N (i.e., supercolonies consisting of multiple colonies connected over a large area 32 , N col = 9, Fig. 1 A-B, Tab. S1). We conducted recognition (own colony against alien colony) and aggression assays and selected individual worker ants that displayed either of the following behavioural states, started aggression , reacted aggressively , or reacted peacefully for chemical, microbiome, genomic, and transcriptomic analyses (Fig. 1 C). We then integrated results from these analyses in a final multinomial logistic regression to assess their joint impact on the behavioural states. Results Aggression tests, and selection of workers for whole-genome and -transcriptome sequencing To select workers for whole-genome and transcriptome sequencing that displayed either of the three behaviours, started aggression, reacted aggressively , and reacted peacefully , we conducted standardised one-on-one worker aggression tests 9 among all nine colonies. We analysed the behaviour of each individual worker and calculated a behaviour index. By conducting an Analysis of Variance (ANOVA), we found that the behaviour differed among the behavioural states (Fig. 2 A; ANOVA: df = 2, F-value = 78.02, p-value < 0.001). To confirm that peaceful behaviour has lower aggression values, we pairwise compared the behavioural states using a Tukey Honest Significant Test: Workers that started aggression and ones that reacted aggressively had significantly higher aggression values throughout the confrontations than workers that reacted peacefully ( started aggression vs reacted peacefully , p-value < 0.001; reacted aggressively vs reacted peacefully , p-value < 0.001). However, workers that started aggression and ones that reacted aggressively had similar aggression values ( started aggression vs reacted aggressively , p-value = 0.597). The within-colony behaviour (control; not shown) did not reveal any aggression. Additionally, workers preferred own odours over alien odours or a control (for details, see the section “ Recognition assays” in the Supplementary Results). Based on the aggression tests and ANOVA, we selected 85 and 109 workers for whole-transcriptome and whole-genome sequencing, respectively. Cuticular hydrocarbon (CHC) analysis The CHC bouquet did not differ starkly among colonies and populations. We found 78 compounds in the odour bouquets (hydrocarbon chain length C12 to C35; GC-MS analyses of CHC-extracts of five workers pooled per colony). From these, 63 compounds were present in all samples (Tab. S2). Visualised multidimensionally (PCA, Fig. 2 B), colonies of the single-queened and aggressive population SQ-A (colonies SQ-A2, SQ-A5 SQ-A6) overlapped with colonies of the single-queened and non-aggressive population SQ-N (SQ-N1, SQ-N4, SQ-N6) and of the multi-queened and non-aggressive population MQ-N (MQ-N1, MQ-N2, MQ-N5), but population MQ-N did so the most. Using the CHC compound data, we conducted a hierarchical cluster analysis and found that CHC extracts from SQ-N and MQ-N were more similar to each other and partially clustered together (Fig. S1 ). In contrast, samples from SQ-A5 were more similar to colonies from populations SQ-N and MQ-N than to SQ-A2 and SQ-A6 colonies. Whole-genome and whole-transcriptome analyses Observed heterozygosity and pairwise genomic differentiation were similar among samples, but relatedness was higher in multiple-queened and non-aggressive colonies. After quality checks and filtering, 184,145 and 69,191 Single Nucleotide Polymorphisms (SNPs) were kept in whole-genome and whole-transcriptome VCF files, respectively (109 and 83 samples, respectively). The mean observed heterozygosity for whole-genome samples was 0.30 (min = 0.23, max = 0.48) and for whole-transcriptome samples 0.15 (min = 0.001, max = 0.37). The pairwise genomic differentiation values (Weir-Cockerham F ST ) were very similar across populations with 0.005 for populations SQ-A:SQ-N, 0.004 for SQ-A:MQ-N, and 0.008 for SQ-N:MQ-N. Visualised multidimensionally (linkage disequilibrium-pruned PCA with DNA samples; Fig. 2 C), the multiple-queened and non-aggressive population MQ-N separated from the other two single-queened populations SQ-A and SQ-N. Samples from population SQ-N clustered together, while colonies from population SQ-A appeared well separated. Samples from population MQ-N clustered together regardless of colony identity. Mean within-colony relatedness was slightly higher in populations SQ-A and SQ-N than in population MQ-N (SQ-A: 0.57, SQ-N: 0.63; MQ-N: 0.33; Fig. 2 D, Tab. S3). The colony queen number (estimated using the relatedness values) was approximately one in all colonies of populations SQ-A and SQ-N and at least two in all colonies of population MQ-N (Tab. S3). We identified three SNPs associated with the behavioural states using a Genome-wide Efficient Mixed Model Association (GEMMA) analysis. We assessed associations between SNPs as well as Insertions/Deletions (‘InDels’) and three SNPs (henceforth SNP1-3; Fig. S2 ). SNP1 is in the sequence of the gene called “Mediator of RNA polymerase II transcription subunit 26” (located at Scaffold 11, site 2457277). SNP2 is in the sequence of an unknown gene (located at Scaffold 165; site 28,302). SNP3 is in the sequence of the gene “ gastrulation-defective ” ( gd ; located at Scaffold 185, site 110,741). The allelic states of the genomic SNPs differed between the behavioural states. We visualised the homozygous and heterozygous allelic states for the three behavioural states multidimensionally (PCA; Fig. 3 A). The behavioural state started aggression revealed a larger variation and had slightly different allelic states in the SNPs compared with the other two behavioural states. We assessed if the count of the reference or alternative alleles for each specific SNP was different across rows (Pearson’s Chi-squared test for count data with simulated p-value and 2000 Monte Carlo replicates) and subsequent post-hoc test. For SNP1, 36 out of 41 (88%) workers that started aggression , 35 out of 39 workers (90%) that reacted aggressively , and 26 out of 29 (90%) workers that reacted peacefully were homozygous for the reference allele (Fig. S3 ; Tab. S4). For SNP1, the allele counts were not different across behavioural states (Fig. S3 , one-SNP1; χ 2 = 0.09, p-value = 1.000). For SNP2, 18 out of 41 (43%) workers that started aggression were homozygous in the SNP state, while 2 out of 39 (5%) workers that reacted aggressively and 2 out of 29 (7%) workers that reacted peacefully were homozygous in the SNP state. More individuals that started aggression were homozygous for the reference allele (SNP2: χ 2 = 22.98, p-value < 0.001; see Tab. S4 for pairwise comparisons). For SNP3, 30 out of 41 (73%) workers that started aggression were heterozygous for the SNP state, 10 out of 39 (26%) workers that reacted aggressively , and 11 out of 29 (38%) workers that reacted peacefully were heterozygous for the behavioural state. More individuals that started aggression were heterozygous for the reference allele (SNP3: χ 2 = 19.381, p-value < 0.001). Differential gene-expression analyses We identified several differentially-expressed genes (DEGs) associated with the behavioural states. We pairwise compared all behavioural states (i.e., started aggression (N AntsSeq = 31), reacted aggressively (N AntsSeq = 28), and reacted peacefully (N AntsSeq = 23). In each comparison, we found ~ 17,000 DEGs or isoforms, of which roughly 100 were significant (for details on the comparisons, see the section “ Differentially-expressed genes ” in the Supplementary Results). In the comparison between ants that started aggression and reacted aggressively , we found 13 significantly up-regulated genes and 36 down-regulated genes in workers that started aggression (false-discovery rate (FDR)-corrected for multiple testing; Fig. S4 A, volcano plot - red dots). When comparing ants that started aggression with workers that reacted peacefully , we found 28 and 61 genes that were significantly up- and down-regulated, respectively. In the comparison between ants that reacted aggressively and reacted peacefully , no gene was significantly up- or down-regulated in workers that started aggression . We found 30 up-regulated and 28 down-regulated genes that were shared across all behavioural comparisons (< 0.05 -corrected genes with known functions were used; Fig. 3 B). For the 30 up-regulated genes, five genes ( CG3800, CG3902, CDase, Rhp , and Moe ; Tab. S5) were exclusively found in the comparison started aggression vs reacted aggressively , 22 genes ( CG34367, CG13625, CG3655, apolpp, CG14687, CG3655, Gat, Sur-8, mRpL9, CG9175, CG6656, Phm, Socs16D, Vav, CG3860, CG32225, CG9426, alph, CG16974, CG10483, AP-2alpha , and bchs ; Tab. S5) exclusively in the comparison started aggression vs reacted peacefully , and three ( CG3061, svr , and Syt4 ; Tab. S5) in both comparisons started aggression vs reacted aggressively and started aggression vs reacted peacefully . No gene was found to be differentially expressed in the comparison reacted aggressively vs reacted peacefully . For the 28 down-regulated genes, eight genes ( BicC, CG3238, Exo84, PlexA, Tret1-1, Taf5, Doa , and Vps35 ; Tab. S5; Fig. 3 B) were found in the comparison started aggression vs reacted aggressively , 19 in the comparison started aggression vs reacted peacefully ( CG3822, Rdl, RFC3, CG10431, Cdep, Dscam1, CG7492, CG6910, snRNP-U1-70K, CG31550, l(1)G0196, CG9346, CG32486, agt, Gcn5, baz, CG13366, U2af50 , and CG8108 ; Tab. S5), and one ( yellow-d2 ; Tab. S5) was found in both comparisons started aggression vs reacted aggressively and started aggression vs reacted peacefully . The log 2 fold changes of these genes ranged between − 3.06 and − 0.15 for started aggression vs reacted peacefully and between − 1.60 and − 0.15 for started aggression vs reacted peacefully (Tab. S5). Of these genes, two were highly expressed: gene CG3800 was highly up-regulated (log 2 fold change = 2.48) and gene BicC was highly down-regulated (log 2 fold change = -3.06). All the above-mentioned genes were used for the multinomial regression analyses (for details, see section below “ Analysing multiple data layers jointly ”). Analysis of high-throughput 16S rRNA gene sequencing data To assess whether the laboratory maintenance affected the microbiome and whether the microbiome (i.e., bacteria and archaea) is associated with the three behavioural states, 16S rRNA gene sequencing was conducted with 49 workers from the populations SQ-A and SQ-N, and 16 additional “control” samples from SQ-A and SQ-N (i.e., directly frozen in the field and not used in aggression tests; for details, see the Materials and Methods section). Subsequently, we conducted a Principal Coordinate Analysis (PCoA) with these samples. Laboratory maintenance did not change the bacterial operational taxonomic units (OTUs) composition in the ants (Fig. S5 A). Also, the behavioural states were mixed with control samples (Fig. S5 B). Only samples from one colony (SQ-N6) were separated from the other samples on the first axis. Four bacterial genera were frequently found across the data set. From an average of 116,096 ± 20,583 raw reads per sample, 79,844 ± 19,799 quality-filtered reads per sample remained, subsampled to an equal depth of 37,808 reads. After rarefaction, 22,215 unique OTUs were identified, including 264 archaea, 19,773 bacteria, and 2178 unknown OTUs. We excluded OTUs not classified at the genus level. Of the remaining OTUs, the genera Pseudomonas (8.7% relative abundance), Bacteroides (6.1%), Lactobacillus (5.2%), and Prevotella (4.4%) were found most frequently. Besides these four bacterial OTUs ( Pseudomonas , Bacteroides , Lactobacillus , and Prevotella ), we further calculated the relative frequency for four additional bacterial genera and one order, namely Acetobacter, Enterococcus, Fusobacterium, Megamonas , and the order Rhizobiales. These bacteria are also known to affect behaviour in humans 20 , 33 , dogs 12 , Drosophila 11 , and ants 13 . The genus Pseudomonas was most frequent (25.4%, Tab. S6) followed by Bacteroides (20.7%), Lactobacillus (18.6%), Prevotella (17.0%), Enterococcus (11.3%), Megamonas (5.4%), Acetobacter (0.7%), Fusobacterium (0.5%), and the order Rhizobiales (0.3%). Using these bacterial OTU genera, we selected OTUs that had a frequency of at least 100 across the behavioural states (N = 119), thus focusing on the most frequent OTUs. With these 119 OTUs, we conducted a sliding-window approach in a multinomial logistic regression to count how often they were significantly associated with the behaviour states. Across these models, the most frequent OTUs (N OTUs =58 with a frequency ≥ 10) included the genera Bacteroides (25% relative percentage across 58 models), Lactobacillus (9%), Prevotella (43%), Pseudomonas (17%), the order Rhizobiales (4%), and the genus Fusobacterium (1%). We further reduced the OTU number for downstream analyses yielding 18 OTUs (e.g., using OTUs with a higher or lower frequency than one across the counts of the behavioural states; for details see “ Analysis of 16S rRNA gene-sequencing data ” in the Materials and Methods). With these 18 OTUs, we assessed whether OTU counts differed among behavioural states by conducting a generalised linear model and pairwise comparison (“emmeans” package 34 ; Tukey corrected for multiple testing). Nine OTUs were significantly associated with the behavioural states, namely two Bacteroides spp., two Lactobacillus spp., three Prevotella spp., and two Pseudomonas spp. (Tab. S8, Fig. 3 C). In the two Bacteroides species, significantly more OTU counts occurred in the behavioural state started aggression and reacted aggressively than in reacted peacefully (OTU 1598) as well as fewer counts in started aggression than reacted aggressively or reacted peacefully (OTU 22324). For the genus Lactobacillus , significantly fewer and more OTU counts occurred in the behavioural state reacted peacefully than in started aggression or reacted aggressively in Lactobacillus mucosae and in Lactobacillus sp., respectively. In the three Prevotella and two Pseudomonas species, significantly more OTU counts occurred in the behavioural state started aggression than in the state reacted aggressively and reacted peacefully (OTUs Prevotella 377, 1887, 20448; Pseudomonas 366, 2442). For the three Prevotella species, also more OTU counts occurred in the behavioural state reacted peacefully than in reacted aggressively . Analysing multiple data layers jointly We integrated genomic, transcriptomic, chemical, and environmental data layers in 24 multinomial logistic regression models to assess if they were associated with the behavioural states. The site-specific environmental variables were calculated manually or extracted from the WorldClim dataset 35 (for details, see “ Environmental variables used in the multinomial regression analyses ” in the Materials and Methods section”). In more detail, we used SNPs, gene-expression counts, within-colony relatedness, site-specific air temperature, the first PC of the CHC analysis, soil nitrogen values, mean annual precipitation, precipitation of the warmest quarter, mean annual temperature, and maximum temperature of the warmest month as explanatory variables (for details, see Materials and Methods section “ Combining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression “). We used allelic states, normalized expression counts, and continuous environmental variables as predictors in a single model. We excluded microbiome data because they were not available for the multiple-queen population MQ-N. In the models, we used the behavioural state started aggression as the baseline and run an intercept-only model as reference. To find the best model explaining the data, we selected various combinations of explanatory variables resulting in 24 models: Models 1–8 used only the four genes that were found in both behavioural comparisons (genes CG3061, svr, Syt4, yellow-d2 ), and Models 9–16 and Models 17–24 included either all up-regulated or all down-regulated genes found in the differential gene expression analysis, respectively. We used log-likelihood ratio tests to the robustness of the variables. In the four best-fitting models, we found two SNPs and eight genes that were associated with the behavioural states. The models were Model 2, 4, 12, and 20, which included the following SNPs and DEGs: SNP2 and SNP3 ( gastrulation-defective ) and DEGs yellow-d2, BicC, Pif1, Exo84, PlexA, KaiR1d, Rdl , and RFC3 (for detailed results, see the supplementary results section “ Logistic multinomial regression analyses ” and Tab. S9-17). Although SNP1 was significant in Model 20, we excluded it because it was not significant in the SNPs-only model. We then combined the above-mentioned variables (i.e., SNP2, SNP3, yellow-d2, BicC, Pif1, Exo84, KaiR1d, RD, RFC3 , PlexA ) in a final multinomial logistic regression. This final model explained significantly more variance than the intercept-only model (Likelihood ratio test of multinomial models, likelihood ratio: 89.73, p-value < 0.001). A goodness-of-fit measure was calculated by comparing the fit of observed and expected values, and the model displayed a good fit (χ 2 = 74.89, df = 4; p-value < 0.001, Residual deviance = 89.23, AIC = 133.23). SNP2 and SNP3 significantly influenced the behavioural states, as well as genes yellow-d2, BicC, Pif1, Exo84, KaiR1d, RD , and RFC3 , but not gene PlexA (Tab. S18). In contrast, within-colony relatedness, CHCs, and the environmental variables were never associated with the start of aggression. The SNPs and DEGs contributed significantly to the behavioural associations, but the two SNPs and gene BicC contributed the most. Overall, the odds ratios of the SNPs and DEGs being associated with the behavioural states ranged from − 4.13 to 4.90 (Tab. S18; Fig. 3 D). The highest log n -values were found for SNP2, which were 4.90 times higher for reacted aggressively and 3.80 times higher for reacted peacefully compared with started aggression . The next highest values were in BicC , which were 0.6 times higher for reacted aggressively and reacted peacefully compared with started aggression . The odds ratios that the DEGs Exo84, yellow-d2, KaiR1d, Pif1, Rdl , and RFC3 were associated with the behavioural states were approximately 0.01 to 0.4 times higher for reacted aggressively and reacted peacefully compared with started aggression . For SNP3, values were − 4.1 and − 3.8 times lower for reacted aggressively and reacted peacefully compared with started aggression . The calculated pseudo-R 2 value “Nagelkerke” was 0.75. In the pairwise comparison of the behavioural states (“emmeans”), the mean of started aggression was lower than the means of reacted aggressively and reacted peacefully (means + confidence intervals started aggression 0.07 + 0.02–0.12, reacted aggressively 0.38 + 0.26–0.50, reacted peacefully , 0.55 + 0.42–0.69; contrasts estimate started aggression vs reacted aggressively : -0.31; df = 22, t-ratio = -5.04, p-value = 0.001; estimate started aggression vs reacted peacefully : -0.49; df = 22, t-ratio = -6.37, p-value < 0.001), but not for reacted aggressively vs reacted peacefully (estimate: -0.18; df = 22, t-ratio = -1.521, p-value = 0.303). Additionally, we conducted log-likelihood ratio tests to evaluate the significance of each focal variables by comparing a full model and a model that lacked the focal variable. Each focal variable contributed significantly to the respective model (Tab. S18; all models converged). Gene-enrichment analyses and gene-function predictions of the identified SNPs and DEGs We used the identified SNP and DEGs, namely SNP3, yellow-d2, BicC, Pif1, Exo84, KaiR1d, RD, RFC3 , and PlexA , in a gene-enrichment analysis and found that the identified are linked to depression restoration, synaptic and neurological functions, aggression, as well as plasticity. We conducted the gene-enrichment analysis in g:Profiler (using only annotated genes and FDR adjusted with p-values < 0.05) to test whether they were enriched for biological processes, molecular function, and/or cellular components. We used the fruitfly Drosophila melanogaster as a background gene set and conducted an unordered query to analyse whether certain biological pathways or gene sets were overrepresented. To increase the sample size of the gene-enrichment search, we used all genes regardless of whether they were up- or down-regulated or from different comparisons (for details, see the section “ Gene-enrichment analyses with known genes ” in the Supplementary Results). In total, six molecular functions, 12 biological processes, and eight cellular components were enriched (Tab. S19). The molecular functions can be broadly summarised into signal transduction and binding and enzymatic and catalytic functions. The biological processes can be summarised into neural signalling, ion transport dynamics, gene expression regulation, and DNA replication and elongation. The cellular components can be summarised into replication functions, vesicle transport, neuronal functions, and ion channel functions. Overall, we combined them into two categories, namely neurological and synaptic functions as well as DNA replication, repair, and genome stability functions (Fig. 3 E). The former included the genes BicC, Exo84, gd (i.e., SNP1 in the gene gd ), KaiR1d, PlexA, Rdl , and yellow-d2 , while the latter included Pif1 and RFC3 . Using a gene-prediction analysis, we also identified neurological and synaptic functions across the SNPs and DEGs. In detail, we used GeneMania 36 (FDR < 0.05; including gene PlexA because excluding PlexA only yielded non-significant results), which predicts gene function and searches for similar functional and related genes based on the initial gene list to find gene pathways or interactions. We included the same SNPs and DEGs in two queries, once with and once without gene gd as it contains a SNP. In the query with gd , we detected five biological processes and two molecular functions (Tab. S20). The biological processes can be summarised as regulation during cell division and the molecular functions as membrane transport functions. In the query without gd , we detected six biological processes and four molecular functions (Tab. S20). The biological processes can be summarised as neuronal signalling and regulation during cell division and the molecular functions as membrane transport functions, possibly linked to synaptic activity and signalling. Discussion Almost all animals display aggressive behaviour, but our understanding of the underlying mechanisms that promote the start of aggression is limited. Here, we integrated, for the first time, chemical, microbiome, genomic, transcriptomic, and environmental analyses and assessed whether these traits promote the start of aggression and reactions to it in the ant Tetramorium alpestre . We tested workers that displayed either of three behavioural states, namely started aggression, reacted aggressively , and reacted peacefully , identified in the aggression assays. Using the microbiome data, we discovered nine OTUs across four bacterial genera, Bacteroides, Lactobacillus, Prevotella , and Pseudomonas , that were associated with the behavioural states. We also identified three genes with a SNP each that were associated with the start of aggression (whole-genome data; GEMMA analysis), namely the gene mediator of RNA polymerase 2 transcription subunit 26 (SNP1), one unknown gene (SNP2), and the gene gastrulation-defective (SNP3). We also found significantly up-regulated (N = 30) and down-regulated genes (N = 28; FDR corrected for multiple testing) when comparing the state started aggression vs each state reacted aggressively or reacted peacefully . Finally, we integrated these SNPs, DEGs, as well as additionally collected colony and environmental variables (e.g., within-colony relatedness, CHCs, site-specific nitrogen and temperature values) in a multinomial logistic regression (multiple data layers jointly). We found that SNP2 and SNP3 (in the gene gd ) as well as the DEGs BicC, Exo84, KaiR1d, Pif1, PlexA, Rdl, RFC3 , and yellow-d2 are associated with the behavioural states, while CHC and colony and environmental variables were not. CHC compounds represent population structure but are not associated with the behavioural states Using the CHC compounds, genetic differentiation, and relatedness values, we corroborate the colony and population structure expected at the onset of this study. The PCA of the CHCs used 63 compounds, which is slightly more than found in a recent study on this species (N CHCs = 50) 9 . We expected the single-queened colonies to be separated from each other and the multiple-queened colonies to be mixed among colonies due to lower and higher relatedness, respectively. The combined analyses of the PCA of the CHC compounds, the CHC hierarchical cluster analysis, the genetic differentiation (pairwise F ST values of the genomic data), and relatedness values corroborated this expectation. In detail, workers of the colonies of population SQ-N are related to each other and likely have only one queen. This also explains the narrow distribution of the CHCs, as CHC bouquets are genetically determined and environmentally tuned 18 . Workers of the colonies of population SQ-A are not or little related and also likely have one queen. As a result, the distribution of the CHC bouquet in the PCA is wider. Workers from colonies of population MQ-N have a higher relatedness scattered across colonies. This indicated that they likely have multiple, possibly unrelated queens. Workers from different queens within the same colony have very different CHCs leading to a broader variety in PCA of the CHCs. While the microbiome, SNPs, and DEGs affected the behavioural states (discussed in the next three sections), the CHC bouquet did not. This is interesting because CHC differences can cause aggressive behaviour in ants 19 but not necessarily in every ant species 37 . Here and in a previous study using this species 9 , we did not find any association between CHC differences and aggression. This could be due to our study design: due to the small size of these animals, we were only able to use workers either for CHC extractions or for aggression assays (and subsequent genomic and transcriptomic analyses). By coincidence, the CHC bouquets of all fighting workers may have been more dissimilar from each other than the ones of the workers used for CHC analyses. Apart from this appearing unlikely as a pattern throughout, also a previous study on this ant 9 and other ant species 19 , 37 used different ants for CHC extractions and for aggression tests. They found a correlation 19 or not 37 suggesting that such a correlation could have been found if CHCs were important drivers of aggression in this ant. In contrast, this is the second study using this species indicating that CHC differences are not important for aggression in this species. Even though CHCs do not seem to elicit aggression in this species, we speculate that it still uses CHCs for nestmate recognition, but workers simply remain peaceful towards non-nestmates with different CHC bouquets. Gut bacteria are associated with the behavioural states We found nine gut bacteria across four genera, Bacteroides, Lactobacillus , Prevotella , and Pseudomonas , that were associated with the behavioural states. The genus Bacteroides is linked with the behavioural states. We found a higher frequency of one Bacteroides sp. (OTU 1598) in ants that were aggressive ( started aggression and reacted aggressively ) than in ants that reacted peacefully as well as a higher frequency of one Bacteroides sp. (OTU 22324) in ants that reacted aggressively and reacted peacefully than in ants that started aggression . Tillisch et al (2017) 20 found that women with a higher Bacteroides abundance had more dense white brain matter tracts, indicating altered sensory processing. This may indicate that Bacteroides bacteria affect how ants perceive other ants and react accordingly. Additionally, Lin et al. 2017 38 and Strandwitz et al. 2019 39 found that a reduced abundance of Bacteroides bacteria in the gut is possibly linked with depression in humans. In turn, this may indicate that ants with higher Bacteroides counts were positively stimulated and thus more proactive and reactive. We found a lower abundance of Lactobacillus mucosae (OTU 813) in ants that reacted peacefully , but a higher abundance of Lactobacillus sp. (OTU 21141), compared with ants that started aggression or reacted aggressively . Recent studies also found links between Lactobacillus and nestmate recognition in honey bees 40 as well as behavioural changes in dogs 12 , Drosophila flies 11 , and ants 13 . However, the results are partially contradictory: for example, one study found a higher Lactobacillus abundance in phobic dogs 12 , namely Lactobacillus plantarum , which has known psychobiotic properties 41 . However, another study on dogs and D. melanogaster males found that the genus Lactobacillus was more frequently present in aggressive dogs and D. melanogaster males 11 . While the underlying mechanisms remain unclear, a link between Lactobacillus and behavioural changes seems to appear. Also the bacteria Prevotella can affect behavioural states. We detected a higher frequency of three Prevotella spp. (OTU 377, 1887, 20448) in ants that started aggression than ants that reacted aggressively or reacted peacefully as well as in reacted peacefully than in reacted aggressively . A study in humans found that women with a higher abundance of Prevotella gut bacteria displayed higher negative affect when shown images with a negative emotional content, which was associated with both functional and structural differences in the hippocampus 20 . Speculatively, these associations of Prevotella with aggression may indicate an evolutionarily conserved pathway of these gut bacteria with negative stimuli, in humans and ants, and possibly other animals. Lastly, also the genus Pseudomonas is connected with behavioural states. We found two Pseudomonas spp. (OTUs 366, 2442) that had a higher frequency in ants that started aggression . To our best knowledge, no study has so far linked Pseudomonas to behaviours such as aggression. However, other studies have linked Pseudomonas strains with, for example, a potential insecticide resistance 42 or metabolising insecticides 43 . Notably, insecticide often affect neuronal or synaptic functions. It thus may be that Pseudomonas is connected to behavioural changes via such a metabolic pathway. While additional bacteria such as Acetobacter, Enterococcus, Fusobacterium , or Megamonas have been found to affect behaviour in humans 20 , 33 , dogs 12 , Drosophila 11 , and ants 13 , we did not find any association with these bacteria here. Our gut microbiome results are in line with other research indicating that there is accumulating evidence of microbiome effects on the recognition and behaviour of animals such as humans 20 , 22 , 38 , 39 , dogs 12 , 21 , Drosophila 11 , cockroaches, locusts, and termites 41 (and references therein), as well as ants 13 . For example, studies also suggests that the microbiome affects the behaviour via the gut-brain axis 12 , 44 : the authors suggest the gut microbiome ‘communicates’ with the central nervous system in various parallel ways such as the vagus nerve, signalling mechanisms, and the production of neuroactive chemicals (e.g., serotonin, gamma-amino butyric acid ‘GABA’) 12,44 . In turn, the central nervous system also communicates with the gut microbiome 12 , creating a feedback loop. While an overall association seems to emerge, the exact effects of specific bacteria on their hosts remain to be determined. Thus, further studies are needed to assess such bacteria-host interactions. SNPs and DEGs are also associated with behavioural changes We found one SNP each in two genes as well as eight DEGs associated with behavioural, neurological, and synaptic functions that may explain the observed behavioural states. SNP2 is at a site of an unknown gene and will not be discussed further, while SNP3 is at a site in the gene gd (gastrulation-defective). For this SNP in the gene gd , more ants that started aggression were heterozygous at the SNP site. Although we found no direct link between the gene gd and aggression, we speculate that there may be an indirect link. The activation of gd leads to the activation of the Toll pathway. The Toll pathway is conserved and is involved in the development of the dorsal-ventral embryonic axis 45 , but it also promotes the expression of the transcription factor nuclear factor kappa B , which has functional roles in neuroprotection and synaptic plasticity 24 . Gene gd may thus be associated with brain synaptic activity and thus possibly with the start of aggression. We further found six down-regulated genes associated with depression restoration, synaptic and neurological functions, aggression, as well as plasticity. These genes, BicC, Exo84, KaiR1d, Rdl , yellow-d2 , and PlexA , are down-regulated in workers that started aggression ( PlexA was kept here as it was driving significant results in the GeneMania analysis). In more detail, BicC is associated with depression restoration 46 , Exo84 with neurite differentiation 47 , KaiR1d with baseline synaptic transmission 48 , Rdl with neurotransmission and olfactory learning 49 , yellow-d2 with dopamine receptor signalling 50 , and PlexA as a receptor fir semaphorsin 51 (for more details on each gene, see the section “ Six down-regulated genes linked to synaptic functions ” in the Supplementary Discussion). The results indicate these SNPs and down-regulated genes could affect behavioural states in this ant species. Specifically, their associations with neurological and synaptic functions could indicate a potential direct link between them and the start of aggression and reactions to it. Further, they may affect the behaviour in this ant in a concerted way. We also found two up-regulated genes associated with DNA repair and replication. These genes, Pif1 and RFC3 , are not directly but indirectly associated with behaviour. For example, Gidron et al. (2006) 52 found that under specific conditions and repeated exposure to stressful situations, reactive oxygen species increased and can yield to DNA damage in animals. It may be that ants that started aggression were more sensible to stressful situations (e.g., sampling and laboratory maintenance) leading to an increase in oxidative stress scavenging mechanisms, which can reduce indices of oxidative stress 53 . In turn, this may explain the up-regulation of such genes. However, further studies need to shed light on such potential associations. We acknowledge the possibility that the identified bacteria, SNPs, and/or DEGs are false positives. However, we argue that this is unlikely because these bacteria, mutations, and DEGs were identified by using independent datasets, applying corrections for multiple comparisons to minimise retrieving false positives, and combining the data sets in a joint analysis (multinomial logistic regression). We further checked the robustness of the results by dropping focal variables using log-likelihood ratio tests. In contrast, we argue that the results of these three independent data sets represent three distinct lines of evidence suggesting that similar underlying mechanisms can contribute to the start of aggression or the reaction to it, for example via hormone and synaptic signalling. The observed behavioural states could thus be affected by these factors in a concerted way. At the same time, we stress that our results are correlative but not causal. Additionally, also other factors, such as epigenetic changes not tested here, may contribute to the start of aggression and reactions to it. Future studies should thus test whether the identified gut bacteria and genes are functionally relevant. This could be tested by conducting aggression tests with ants that have been fed with these bacteria or with ants in which these known genes are knocked out, impaired, or over-expressed. It would further be interesting to ascertain whether the same gut bacteria, genes, or gene homologs are important for the aggressive or peaceful behaviour in other (social) insects as well. Additionally, possible effects of epigenetic changes (e.g., DNA methylation, histone modifications) should be tested. Aggression and its possible positive effects can be adaptive 1 , 4 . This is especially true if starting aggression leads to increased fitness 1 . The start of aggression has been loosely associated with individual effects, such as changes in the gut microbiome 11–13,20−22 , SNPs in genes 23 , 24 or DEGs 17 . In this study, we integrated – for the first time to our best knowledge – gut microbiome data with chemical, genomics, transcriptomics, environmental, and behavioural assays, using the ant T. alpestre . We identified nine gut bacteria, two mutations, and eight DEGs that are associated with the three behavioural states started aggression , reacted aggressively , and reacted peacefully. In contrast, chemical and environmental factors were not associated with the behavioural states. The nine gut bacteria found are known to influence aggression and other behaviours in several organisms, for example, via hormone signalling 11 , 21 , 22 , 40 . The identified SNPs and DEGs were, among others, associated with neurological and synaptic functions. Based on these results, we speculate that these three traits can contribute the start of aggression, possibly in a synergistic mechanism. Materials & Methods Fieldwork and colony maintenance Between July 18th and 25th 2018, 500 workers were sampled from three colonies each in three populations (N colonies =9, Tab. S1). The populations were selected based on preliminary behavioural data (not shown): One population was located in South Tyrol, Italy, and comprised single-queened and aggressive colonies (“SQ-A”), one in Tyrol, Austria, comprising single-queened and non-aggressive colonies (“SQ-N”), and one in Carinthia, Austria, comprising multiple-queened and non-aggressive colonies (“MQ-N”; potentially supercolonial population). Of these 500 workers, 200 were immediately snap-frozen in the field using a dry shipper (CY50915D, Thermo-Fisher Scientific Inc., MA, USA) for CHC and molecular analyses. The remaining workers were transported alive to a laboratory at the University of Innsbruck and transferred to polypropylene boxes (10.5 × 10.5 cm; as of now “colony”) awaiting behavioural assays. These workers presumably included all polyethism stages 31 . To prevent workers from escaping, the walls of the boxes were Fluon-coated (GP1, De Monchy International BV, Rotterdam, Netherlands). Each box was equipped with soft tissue as a hiding place, two conical Eppendorf tubes filled with water or with diluted honey water and each plunged with cotton as a drinking aid, and a frozen Drosophila hydei fruit fly. The water, honey water, and fruit fly were refilled twice per week and present at all time ad libitum . The boxes were placed in a climate cabinet (MIR-254, Panasonic, Etten Leur, Netherlands) with constant dark conditions, a humidity of 50–70%, and at constant 18°C. Constant 18°C was selected to acclimatise workers that originated from slightly different elevations to a similar temperature. Before the various assays, the colonies were kept in the climate cabinets for two weeks. Pairwise geographic distances between populations SQ-N:MQ-N (Kuehtai – Mussen; Fig. 1 A) were 140 km, between SQ-A:MQ-N (Penser Joch – Mussen) 125 km, and between SQ-A:SQ-N (Penser Joch - Kuehtai) 40 km calculated using an online tool ( https://www.ibm.franken.de/gps03.html ). Recognition assays Between August 13th and 20th 2018, we conducted recognition assays to test if workers recognise and prefer their own colony odour over an alien colony or a control odour following Steiner et al. (2007) 26 . For these assays, we extracted cuticular hydrocarbons (CHCs) from workers of each colony separately using three different extraction solvents sequentially, starting in 100 µl hexane, then 100 µl ethyl acetate, and lastly 100 µl 96% ethanol (all three Merck, MA, USA) 26 . For the extraction, we transferred the workers into 1.1-ml conic glass vials (CZT, Kriftel, Germany). Workers remained in each extraction solvent for 90 s before being transferred to the next. The three solvents should extract as many CHCs as possible. We then transferred the workers to 96% ethanol, mixed the three solvents, and stored them at -20°C until further use. To account for body size differences, we used 15 workers for each colony from populations SQ-A and SQ-N and 22 workers for each colony from population MQ-N, which had smaller workers. In total, we generated nine solvents (one for each colony) to test if workers prefer their own, alien, or control (a mixture of the three extraction solvents without CHCs) odour. To do this, we created small filter paper disks (2 cm diameter; 75 g/qm, Altmann Analytik, München, DE) with three 120-degree sectors (own, alien, and control sector). Onto the sector “own”, we applied 1 µl of the extract of the colony to be tested, onto the sector “alien”, 1 µl of the extract of a different colony from the same population, and onto the sector “control”, 1 µl of the mixture of extraction solvents without CHCs. We transferred the solvents onto the paper disks using a 20-µl syringe (Hamilton, NV, USA). After transferring the solvents onto the papers, the solvents were left to evaporate for three minutes before we transferred the paper disks to the bottom of small glass vials (2 cm Ø). The bottom of each glass vial was covered with 1 µl paraffin oil as a keeper substance 26 . The walls of the glass vials were Fluon-coated to prevent workers from escaping. After evaporation, we transferred individual workers to the glass vials, which were covered with tin cans to simulate dark conditions. Workers were allowed to acclimatise for 15 minutes, after which we lifted the can for five seconds and noted the sector of the paper disk on which the worker was sitting. After each observation, we turned the vial 120 degrees and lightly tapped it thus forcing the worker to move. In each assay, we tested all colonies in a randomised order. Both conductors and evaluators were blind to the origin of colonies. We conducted the assays in an air-conditioned room with constant 18°C resembling the temperature in the climate cabinet. In one run, 36 workers were tested (four from each colony), and 15 runs were conducted. This procedure was replicated three times over three days resulting in 1,620 observations, which were analysed together using a multinomial Goodness-of-Fit test to test if workers recognise and prefer their own colony odour, an alien colony odour, or a control odour. Extraction and analysis of cuticular hydrocarbons (CHCs) We extracted CHCs from five workers per colony following Krapf et al. (2023) 9 . For the extraction, we transferred five workers, which had been immediately frozen after sampling, to 1.1-ml conic glass vials (CZT, Kriftel, Germany) and immediately added 100 µl n-pentane (Merck, MA, USA) using a 100-µl syringe (Hamilton, NV, USA). The CHCs were extracted for three minutes while the glass vials were being shaken at 450 rpm. After the extraction, we removed the workers from the vials and transferred them to Eppendorf tubes filled with 96% ethanol. The vials containing the CHC extracts were sealed until their analysis. For the analysis, a 7890 B Series gas chromatograph (Agilent, Waldbronn, Germany) equipped with a flame ionization detector (FID), a nonpolar DB-5 column (30m×0.25mminner diameter, J&W, Waldbronn, Germany), and hydrogen (2ml/min constant flow) as carrier gas was used. One µl of a sample was injected splitless at an initial oven temperature of 50°C. After 1 min, the splitting valve was opened and the temperature gradually increased by 10°C/min until it reached a final temperature of 310°C, which was kept constant for 50 min. To ensure the consistency of the analyses, GC runs were performed regularly with a synthetic alkane standard mixture. Structure elucidation of individual compounds was performed with an HP (Hewlett Packard) 6890 Series gas chromatograph connected to a mass selective detector (GC–MS; Quadrupole 5972, Agilent, Waldbronn, Germany). Helium was used as carrier gas (1.5 ml/min constant flow). The temperature program was the same as described above. The absolute and relative amounts of these compounds were determined by using Agilent ChemStation software (Agilent, Waldbronn, Germany). Structure assignments were carried out by comparison of mass spectra and retention times of natural products with corresponding data from synthetic reference samples using the NIST database and a database of the Institute of Evolutionary Ecology and Conservation Genomics at the University of Ulm, following previous work 54 , 55 . Peak identities across different runs were confirmed by GC-MS. To estimate relative proportions for further downstream analyses, we only used CHCs that were found in all samples. Further, we divided the absolute amounts of individual compounds by the sum of the absolute amounts of all compounds and multiplied by 100. With these CHC compounds, we created a PCA using the function “prcomp” (“ggfortify” package 56 ) to check if colonies and/or populations form distinct clusters. Further, we conducted a hierarchical cluster analysis with the CHC data using the function “agnes” and Ward’s minimum variance method (“cluster” package 57 ). We used the values of the first PCA in the multinomial regression. One-on-one aggression tests We conducted one-on-one aggression tests within each population on July 23rd 2018 to determine if the colonies displayed the expected behaviour (i.e., aggressive and non-aggressive). We conducted standardised aggression tests 9 in an air-conditioned room with constant 18°C. For each aggression test (i.e., one encounter), we randomly selected naïve single workers from different colonies from the same population and transferred to a small glass vial (1.4 cm inner diameter) with Fluon-coated walls preventing workers from escaping. Only workers actively running outside in the arena were selected, which likely were foragers 31 . We added a worker from one colony first and then the second worker. In the next encounter, we changed the order of workers introduced to prevent any effect of adding workers to the vial. We conducted five encounters for each colony combination to account for behavioural variation 58 . Each encounter lasted 180 s and was filmed using high-definition cameras (Handycam HDR-XR 155; HDRPJ810E, Sony, Tokyo, Japan). As workers might have been agitated after being transferred to the vials, the first 10 seconds of each encounter were regarded as an acclimatisation time and were thus excluded from further analyses 9 , 30 , 31 . The assay conductors were not blind to the colony’s origin. We further conducted one-on-one aggression tests between populations between July 25th and 28th 2018 following the approach described above. Within 10 minutes after the end of the aggression test, we separated the workers if fighting, transferred them individually to 1.5 ml tubes, and snap-froze them using liquid nitrogen. This procedure ensured that no early genes were expressed, which can start after 15 minutes 59 . At this point, the colony origin of the workers was unknown, but we later identified the colony identity using microsatellite analysis (see section below “ Microsatellite genotyping for reference workers ”). Additionally, we conducted within-colony aggression tests on July 27th 2018 to test if workers behaved peacefully, which was our expectation. One-on-one aggression analysis and worker selection for sequencing For an initial screening of the aggression test, we noted the behaviour of both workers every ten seconds as “aggressive”, “neutral”, or “peaceful” while conducting the aggression tests. Based on this initial screening, we selected 112 videos for a detailed analysis. From these videos, we examined the behaviour of each worker in slow-motion, and classified the behaviour of both workers second by second using the following scoring scale 60 : (− 4) trophallaxis, (− 3), allogrooming, (− 2) antennation, (− 1) being next to each other without contact, (0) ignoring, (1) avoiding, (2) mandible threatening, (3) fighting without gaster flexion, (4) fighting with gaster flexion, and (5) killing. The observer of the videos was blind to the origin of the colonies. Moreover, an aggression index AI 61 was calculated as detailed in Krapf et al. (2023) 9 . For AI , the duration of each behaviour was summed up and multiplied by its respective behaviour score (-4 to + 5). This value was divided by the total number of seconds with tactile interactions recorded. Lastly, the arithmetic mean of the five replicates was calculated. Using this detailed analysis, we defined three behavioural states: workers that ‘ started aggression’ , workers that ’reacted aggressively’ , or workers that ’reacted peacefully ’. For the aggressive states ( started aggression ; reacted aggressively ), we used workers that displayed a scoring value of 3 and higher to ensure that high aggression levels were used. Based on these three behavioural states, we selected 109 workers for whole-genome sequencing ( started aggression = 43 workers, reacted aggressively = 35 workers, reacted peacefully = 31 workers; see Tab. S1 for population and colony details) and, of those, we selected 85 workers for transcriptomic analyses ( started aggression = 31 workers, reacted aggressively = 29, reacted peacefully = 25). The additional 24 workers selected for whole-genome sequencing originated from the non-aggressive and polygynous population MQ-NS. They were used to account for multiple queens and reliably calculate within-colony relatedness and estimate queen numbers. DNA- and RNA-extractions and whole-genome and whole-transcriptome sequencing For whole-genome sequencing of samples, we cut off the mesosoma and abdomen from the head of each ant using sterile scalpels (Fig. P1). We used the mesosoma and abdomen for DNA extractions (N samples =109) and the head for RNA extractions (N samples =85). We extracted DNA using the QiAmp Micro DNA Kit (Qiagen, Hilden, Germany). For this, we transferred the mesosoma and abdomen of each worker to a sterile tube and submerged it into liquid nitrogen. We then ground the mesosoma and abdomen using disposable pestles. The extraction followed the manufacturer’s protocol except for the dilution, which was conducted twice, as follows: the first elution was done with 50 µl dH 2 0 for whole-genome sequencing and the second elution with 30 µl dH 2 0 for microsatellite genotyping to determine the colony affiliation (see section below “ Microsatellite genotyping to identify colony identity ”). We extracted RNA from the heads of 85 workers using the Nucleospin RNA Kit (Macherey-Nagel, Düren, Germany) following the manufacturer’s protocol. For this, we transferred the head of each worker to a sterile tube, submerged the tube into liquid nitrogen, and grinded the head using disposable pestles. The subsequent extraction followed the manufacturer’s protocol except for the dilution: RNA was eluted in 40 µl RNAse-free dH20 provided by the manufacturer. We conducted all DNA- and RNA-extraction steps under sterile conditions in a laminar flow hood. DNA and RNA extracts were stored at -70°C until being shipped for library preparation and whole-genome and -transcriptome sequencing outsourced to a commercial provider (IGATech, http://igatechnology.com/ ). Each worker was sequenced with 125-bp paired-end sequencing for both DNA- and RNA extractions on HiSeq2500. Microsatellite genotyping to identify colony identity We conducted microsatellite genotyping to assess the colony identity of workers used in aggression tests. First, we genotyped 12 reference workers from each colony (i.e., known colony identity) using eight microsatellite loci 9 , 30 . For this, we extracted DNA using the Sigma GenElute extraction kit following the manufacturer’s protocol, except for eluting in 50 µl. PCR for genotyping was done in 5 µL reaction volume with 0.5 µL template DNA, 2 × Rotorgene Master Mix (Qiagen, Hilden, Germany), 0.01 µM M13 tailed locus-specific forward primer, 0.1 µM fluorescent-labelled M13 primer, 0.1 µM untailed specific reverse primer, and 1.79 µL dH 2 O on a UnoCycler 1200 (VWR, Radnor, USA). Cycling conditions were 94°C for 5 min followed by 35 cycles at 94°C for 30 s, 60°C for 1 min, 72°C for 45 s, and a final extension at 68°C for 20 min. Fragment analysis was carried out on an ABI3730XL genetic analyser (Applied Biosystems, Foster City, USA) by a commercial provider (Comprehensive Cancer Center DNA Sequencing & Genotyping Facility, University of Chicago, USA). Microsatellites were genotyped using GeneMarker V.3.0.1 (SoftGenetics, State College, PA, USA). Following the same procedure, we genotyped workers from the aggression tests and reliably assigned the colony identity before shipping the samples to the commercial provider IGA for sequencing. Based on the genotypes of known colony identities, we calculated the probability of colony affiliations using the software GeneClass2 62 . GeneClass2 uses multilocus genotypes to select or exclude populations as origins of individuals. To find colony affiliations, we chose the Bayesian method by Rannala & Mountain (1997) 63 as the computation criteria, and the assignment threshold of the scores was 0.05. Further, we calculated within-colony relatedness based on the genotypes following Queller and Goodnight algorithm 64 and additionally, the number of queens following Pamilo (1991) 65 . Analysing whole-genome and whole-transcriptome sequences For both DNA and RNA files, we conducted the same analysis approach. Initial quality control of raw reads was conducted using FastQC ( https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ ) and MultiQC ( https://seqera.io/multiqc/ ). We trimmed adapters, duplicates, and contaminants using a “kraken” database and “bbduk” (“bbtools”, https://sourceforge.net/projects/bbmap/ ). We merged trimmed paired-end files into single files and mapped single files against the Tetramorium alpestre reference genome 66 using “bbmap” (bbtools) by applying quality trimming on both sides. For mapping, we indexed the files and quality-trimmed them using “bbmap” (minid = 0.9, k = 13). We called single nucleotide polymorphisms (SNPs) using the “callvariants” function from “bbtools” using the default settings except for ploidy = 2. For variant calling, we first called variants in an initial VCF file. Second, we calculated the true equality and then recalibrated them using the initial VCF file. Third, we created an unfiltered VCF file. In this unfiltered VCF file, we identified 1,249,705 and 312,297 SNPs for whole-genome sequences and whole-transcriptome sequences, respectively. We further filtered this unfiltered VCF file using a minimum coverage of 128, a minimum number of sequences of 4 with the alternative allele, a minimum mapping quality of 50, and including linkage-disequilibrium (LD) pruning. After filtering, 184,145 and 69,191 SNPs were kept in the final whole-genome and whole-transcriptome VCF file, respectively. Using the VCF file of the whole-genome data, we calculated the heterozygosity and Weir and Cockerham's F ST and created an LD-pruned PCA using VCFtools 67 . We further calculated the within-colony relatedness using the “relatedness” function in VCFtools 67 using the method of Manichaikul et al. (2010) 68 . We then comapred the within-colony relatedness from whole-genome data with the within-colony relatedness from microsatellite genotyping to assess concordance of values (Tab. S3). Genome-wide mixed-model association (GEMMA) analysis using whole-genome sequences We conducted a GEMMA 69 analysis using whole-genome sequence data to determine if the behavioural states were associated with SNPs in the VCF. Before the analysis, we excluded duplications in the VCF to reduce the bias of emphasising duplications. We used this VCF file without duplications to create a bimbam file using a custom-made Python script. After calculating the bimbam file, we calculated a centred relatedness matrix using “gemma”, which was used in the subsequent GEMMA analysis. In the GEMMA analysis, a phenotype list detailing the behavioural states of workers, a bam list, and an LD-covariance file were used. GEMMA results were visualised using Manhattan plots created in R using the function “Manhattan” (“qqman” package 70 ). We inspected genomic SNPs above the suggestive line by using them a PCA created with the function “prcomp” (“ggfortify” package 56 ) to check whether alleles clustered together. For this PCA, we dummy-coded individuals that were homozygous for the reference allele of the respective genes as 0/0 and individuals that were heterozygous for the reference alleles as 0/1. We did not find any individual that was homozygous for the alternative allele. Further, we conducted a Pearson’s Chi-squared test for count data with simulated p-value and 2000 Monte Carlo replicates to calculate the p-values. The count data represented the number of counts of all individuals for being homozygous or heterozygous for the reference allele for the three behavioural states. The idea was to check if individuals that were homozygous or heterozygous for the reference allele were more or less frequently observed in one of the three behavioural states. Differential gene expression and gene-enrichment analysis Differential gene expression The expression counts of each individual stemming from a newly created annotation (for details, see the section “ Tetramorium alpestre annotation ” in the Supplementary Materials and Methods) were merged using a customised R script. Using this merged data set, we analysed the expression counts of all individuals (“DESeq2” package 71 ). For this, we created a DESeqDataSet object to compare the expression of the behavioural states in a pairwise manner. The three behavioural comparisons were: started aggression vs reacted aggressively , started aggression vs reacted peacefully , and reacted aggressively vs reacted peacefully . As a pre-filtering step, we only kept rows that had at least 10 counts in total, thus excluding rows (i.e., genes) with fewer counts than 10. Next, we assessed the data quality of each sample using a pheatmap (“pheatmap“ package 72 ). Of the 85 samples, we excluded three due to low quality, yielding 82 samples for subsequent analysis. We conducted a differential gene expression analysis with these 82 samples based on the Negative Binomial (i.e., Gamma-Poisson) distribution and using the default settings. We created volcano plots for each behavioural comparison exported the results as table with a log fold change threshold of zero and using a of 0.05-Benjamini-Hochberg correction (“result” function; DESeq2 package; “false-discovery rate”, FDR). We created such result tables for all three comparisons and up-regulated as well as down-regulated genes separately and used them in subsequent analyses. Such tables included gene names, log 2 fold values, p-values, and FDR-corrected p-values for multiple testing. We further queried gene names in FlyBase (release FB2025_04) to obtain information on gene function. In subsequent gene-enrichment analyses and multinomial logistic regression analyses, we only used genes with a known (i.e., annotated) gene name. Gene-enrichment analyses For the three behavioural comparisons started aggression vs reacted aggressively , started aggression vs reacted peacefully , and reacted aggressively vs reacted peacefully , we conducted a gene enrichment analysis in g:Profiler ( https://biit.cs.ut.ee/gprofiler/gost ), a web server for functional gene-enrichment analysis. We only used known (i.e., annotated) genes with an FDR-adjusted p-value lower than 0.05 (Tab. S5). For each behavioural comparison, we conducted a query with an unordered list of genes based on the log 2 fold changes. We selected Drosophila melanogaster as the organism to match the query gene list. Further, we created Venn diagrams using the behavioural-comparison genes for the annotated and all genes in R using the function “ggvenn” (“ggvenn” package 73 ). This analysis allowed checking whether the same genes are up- or down-regulated in several comparisons. Microbiome DNA extraction and marker gene sequencing To test whether the microbiome influenced the three behavioural states, we conducted 16S rRNA gene sequencing. Due to a limited availability of samples, we used 49 workers from two populations: Specifically, we selected four workers each from two colonies of the single-queened and aggressive population SQ-A (colonies SQ-A5 and SQ-A6) and from two of the single-queened and non-aggressive population SQ-N (colonies SQ-N1 and SQ-N6; SQ-N6 with five workers) and from each behavioural state. This resulted in using 16 workers that started aggression , 16 that reacted aggressively , and 17 that reacted peacefully (Tab. S1). To test if the microbiome changed during laboratory maintenance, we selected 16 additional workers (4 workers each from the colonies SQ-A5, SQ-A6, SQ-N1, and SQ-N6) as control. These workers were immediately frozen after fieldwork and did not experience any laboratory maintenance. Before the extractions, we sterilised the surface of whole workers by transferring individual workers for 15 s into Eppendorf tubes filled with 100 µl 5% bleach and then for 15 s into Eppendorf tubes filled with 100 µl phosphate-buffered saline solution (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4) 74 . For the 16 control workers, we extracted DNA using the QIAamp DNA Mini kit (Qiagen, Hilden, Germany) and eluted twice each time with 30 µl of the elution buffer from the kit. For the remaining 49 workers, we dissected the heads from the mesosoma and gaster using a sterile scalpel. For microbiome analyses, we extracted DNA from the mesosoma and gaster using the QIAamp DNA Mini Kit. To determine colony affiliation using microsatellite genotyping (for details, see “Microsatellite genotyping of reference workers” above), we extracted DNA of the head using the DNEasy Blood and Tissue Kit (Qiagen, Hilden, Germany). We extracted DNA following the manufacturer’s protocol except for the elution: DNA was eluted twice each time with 30 µl of the elution buffer from the kit. We conducted all steps before and during the extraction under sterile conditions in a laminar flow hood. High-quality DNA extracts were sent to Novogene (Cambridge, United Kingdom) for marker gene sequencing on a NovaSeq6000 machine (Illumina, San Diego, CA, United States). The universal primer pair for bacteria 515F (5’-GTGCCAGCMGCCGCGGTAA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’) was used to target the V4 region of the 16S rRNA gene using a 2×250 bp approach. Analysis of 16S rRNA gene-sequencing data We merged the raw reads into contigs using flash” v.1.2.7 75 . We used Qiime v.1.7.0 for quality filtering following the standard operating procedures. We used SILVA v.138 as a reference database and to detect chimeric sequences by the UCHIME algorithm, which we removed from the data. Sequences were clustered into OTUs based on a ≥ 97% similarity threshold. We converted the raw data to a phyloseq object (“phyloseq” package 76 ) and rarefied to the smallest sample size, after removing the sample Nu_ctrl_153a as an outlier. We conducted principal coordinate analyses (PCoA) based on populations and behavioural states and visualized the data (“ampvis2” package) 77 . In total, we found 22,215 OTUs after rarefaction. We further calculated the frequency of the four most frequent bacterial genera as well as for four additional bacteria genera, Acetobacter, Enterococcus, Fusobacterium, Megamonas , and the orders Rhizobiales and Entomoplasmatales. Acetobacter , and Enterococcus have been associated with aggression in Drosophila melanogaster 11 , Fusobacterium and Megamonas have been associated with aggression and non-aggression in dogs 12 , 21 , and Rhizobiales and Entomoplasmatales have been associated with aggression in leaf-cutting ants 13 . Entomoplasmatales were not found in our data set. Using the bacterial OTU genera mentioned above, we selected OTUs that had a frequency of at least 100 across the behavioural states (N = 119), thus focusing on the most frequent OTUs and restricting the analysis to 119 OTUs. With these, we conducted a sliding-window approach with multinomial logistic regressions (function “multinom”, “nnet” package 78 ). A multinomial regression allows using more than one categorical variable as response variables (here, the three behavioural states). In the sliding-window approach, we created individual models that tested 20 OTUs simultaneously in a multinomial regression. Briefly, the first model used OTUs 1 to 20, the second model OTUs 2 to 21, etc. To evaluate the model fit and calculate p-values and log-likelihood tests, we conducted the same methods as described in the section “ Combining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression ”. In each model (N = 119), we used the behavioural state started aggression as the baseline. Manually checking 119 model fits and results was not efficient, so we created an R (version 4.3.0 79 ) script to extract model fits and model p-values for the different OTUs. The script also counted how often OTUs were significantly influencing the behavioural states and thus allowed checking if the same OTUs influenced the behavioural states more or less frequently. Our rationale was that if one or a few OTUs are present in many or all models, then these OTUs likely have a higher impact on the behavioural states than OTUs with a low frequency. If, however, OTUs are only counted a few times, they have likely arisen due to chance and may represent artefacts. From these models, we extracted the significant OTUs and counted their frequency across the models. Across these models, the most frequent OTUs (N OTUs =58 with a frequency ≥ 10) included the genera Bacteroides (relative percentage across the 58 models, 25%), Lactobacillus (9%), Prevotella (43%), Pseudomonas (17%), the order Rhizobiales (4%), and the genus Fusobacterium (1%). For these gut bacteria, we noted the OTU frequency in each behavioural state and the control. From the 58 OTUs, we excluded 40 OTUs (five because the frequency was significantly higher or lower than in the control, 16 OTUs because the count of the control was higher as the highest number of counts of one of the behaviours, six OTUs because the counts were evenly distributed across all behavioural states, five because the counts were less than 10 in one of the behavioural states, seven OTUs because the counts of the control was similar as the counts of the behavioural states, and one because the counts were not different between the control and the behavioural states) yielding 18 OTUs for further analyses, namely three Bacteroides spp., three Lactobacillus spp., nine Prevotella spp., three Pseudomonas spp., and one Rhizobiales sp. With this set of 18 OTUs, we assessed whether the counts differed between the behavioural states. For this, we conducted a generalised linear model with these count data (response = count; explanatory variable = behavioural states; Poisson-distributed) and assessed the pairwise comparisons (“emmeans” package; Tukey corrected for multiple testing). Nine OTUs revealed significant results and were further discussed, while others were non-significant or revealed inconsistent results ( i.e. , reacted aggressively higher than the other behavioural states). We could not analyse the microbiome data together with SNPs and DEGs because no samples for the population MQ-N were available for the microbiome analysis. Environmental variables used in the multinomial regression analyses For each colony, we estimated a standardised air temperature (TAS) as a rough measure of the colonies’ thermal niche 80 . Following the logic of Seifert and Pannier (2007) 81 , TAS was calculated for a sampling site as the mean air temperature of the period from May 1st to August 31st averaged over the years 1961 to 1990 of the nearest three meteorological stations (data provided by Klimaabteilung der Zentralanstalt für Meteorologie und Geodynamik (1996), Vienna, Austria). The data were corrected for an altitudinal decrease in temperature of 0.661°C per 100 m according to the equation of Seifert and Pannier (2007): TAS= -0.694×LAT + 0.078 ×LON-0.00661 ×ALT + 52.20, (1) where TAS is the predicted standardised air temperature in °C, LAT and LON denote the geographical latitude and longitude in decimal format, respectively, and ALT is the altitude above sea level in metres. From the WordlClim dataset 35 , we downloaded environmental variables from the years 1970 to 2000 and extracted site-specific values using the “extract” function (“raster” package 82 ). In particular, we selected data on mean annual precipitation, precipitation of the warmers quarter, mean annual temperature, and the maximum temperature of the warmest month both as temperature and precipitation affect the colonies’ environment and higher temperatures promote aggression in this species 9 . Further, we retrieved soil nitrogen values for each site from the European LUCAS topsoil dataset 83 . We used these variables in multinomial regression analyses (described in the next paragraph) to test if the environment is associated with the behavioural states. We recently found such an association in this ant, where higher temperature and nitrogen values were positively associated with aggression 9 . Combining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression In the multinomial logistic regression, we integrated principal component 1 of the CHC analysis, three SNPs, eight gene expression counts, and colony and environmental variables to assess whether they were associated with the three behavioural states ( started aggression , reacted aggressively , reacted peacefully ). Although CHCs, SNPs, gene expression counts, and environmental variables represent distinct biological and abiotic entities, they have high-dimensional features measured across the same samples and thus share a common statistical role. Moreover, a multinomial regression provides the opportunity to use a unified framework to quantify their joint contribution to categorical outcomes while preserving interpretability. In total, we tested 24 models (Tab. S9). Models 1–8 used expression counts of DEGs that were observed in both behavioural comparisons (Fig. 3 B; Tab. S3 highlighted cells). Models 9–16 used expression counts of DEGs that were up-regulated in the behavioural comparisons. Models 17–24 used expression counts of DEGs that were down-regulated in the behavioural comparisons. Fitting separate models with increasing number of input variables allowed us to assess if the input variables influence the behavioural states in combination or separately. For example, if some genes are up-regulated in workers that reacted aggressively but other genes are up-regulated in workers that reacted peacefully , using these genes in combination may lead to false conclusions. In detail, we tested the following models, namely “intercept-only” models (Models1, 9, 17), models with all three SNP states only (Models 2, 10, 18), models with DEGs that had a log 2 fold value of at least ± 0.5 (Models 3, 11, 19), models with all three SNP states and DEGs (log 2 fold of at least ± 0.5; Models 4, 12, 20), models with the within-colony relatedness values, standardised air temperature, the first PC from a CHC PCA, site-specific soil nitrogen values, mean annual precipitation, precipitation of the warmest quarter, mean annual temperature, and maximum temperature of the warmest month (“colony and environmental variables”; Models 5, 13, 21), models with all three SNP states and the colony and environmental variables (Models 6, 14, 22), models with DEGs (log 2 fold of at least ± 0.5) and colony and environmental variables (Models 7, 15, 23), and, lastly, models with all above-mentioned variables (Models 8, 16, 24). We compared the model fits using the “anova” function (basic stats package; “Chi-square test”). We further calculated the Akaike Information Criterion for small sample sizes (AICc) of the models (excluding the intercept-only model) using the “aictab” function (“AICcmodavg” package 84 ) and the models with the lowest ΔAICc (deltaAICc) values represented the best fitting models. Additionally, we calculated a goodness of fit measure for these models by comparing the fit of observed and expected values. To further test if the used variables are significantly influencing the behavioural states, we manually calculated the p-values using a two-tailed Wald Z test. We used the behavioural state started aggression as baseline in the logistic regression. To subsequently test if the behavioural states differ from each other, we compared their means in a pairwise manner. For this, we calculated the marginal means between the behavioural states using the functions "emmeans" and “contrast” (“emmeans” package). These post-hoc tests compared the behavioural states and allowed conducting hypothesis tests to determine whether the differences were statistically significant. We also calculated two pseudo coefficients of determination (R 2 , “Nagelkerke” and “McFadden”) to check how much of the variation is explained by the independent variables. As we used multinomial regressions, the pseudo-R 2 values were only approximated. We further assessed the significance of the independent variables individually using a likelihood ratio test “lrtest” (“lmtest” package 85 ). This test drops the focal variable in the model to assess its impact on the model (i.e., it compares a focal model with the same model by excluding the targeted independent variable). If the model differs significantly, the focal variable is a dominant variable in the model. Declarations Data Availability All code and datasets generated and/or analysed in the study will be made publicly available alongside the publication Acknowledgements We thank Philipp Andesner and Elisabeth Zangerl for help during lab work and Marlene Haider and Markus Möst for helpful discussions during statistical analysis. We thank the government of Carinthia for issuing a sampling permit for the protected area “Mussen”. The LUCAS topsoil dataset used was made available by the European Commission through the European Soil Data Centre managed by the Joint Research Centre (JRC; http://esdac.jrc.ec.europa.eu/). The computational results presented here have been achieved (in part) using the LEO HPC infrastructure of the University of Innsbruck and the MACH2 Interuniversity Shared Memory Supercomputer. This study was financially supported by the FWF (Austrian Science Fund) under Award Number P 30861 awarded to F.M.S. and by the European Union’s Horizon Europe programme under Marie Skłodowska-Curie Actions (MSCA) - Postdoctoral fellowship grant agreement no. 101204375 awarded to P.K. The maps were created using Stadia Maps (stadiamaps.com), Stamen design (stamen.com), and OpenStreetMap (openstreetmap.org/copyright). Author contributions Patrick Krapf: Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualisation. Francesco Cicconardi : Conceptualization, Methodology, Writing – review & editing. Martin Schilling: Methodology, Validation, Formal analysis, Investigation, Visualization, Writing – review & editing. Gerhard P. Aigner: Investigation, Formal analysis, Resources, Writing – review & editing. Thomas Klammsteiner: Methodology, Software, Formal analysis, Investigation, Writing – review & editing, Visualization. Manfred Ayasse: Conceptualization, Methodology, Investigation, Resources, Data curation, Writing – review & editing. Wolfgang Arthofer: Conceptualization, Methodology, Writing – review & editing. Alexander S. Mikheyev: Conceptualization, Methodology, Writing – review & editing. Florian M. Steiner: Conceptualization, Methodology, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition. Birgit C. Schlick-Steiner: Conceptualization, Methodology, Resources, Writing – review & editing, Supervision, Project administration, Funding acquisition. References Covington H, Newman IIIE, Leonard M, Miczek K (2019) Translational models of adaptive and excessive fighting: an emerging role for neural circuits in pathological aggression [version 1; peer review: 3 approved]. F1000Research 8, 18883 Marshall VT, Humfeld SC, Bee MA (2003) Plasticity of aggressive signalling and its evolution in male spring peepers, Pseudacris crucifer . Anim Behav 65:1223–1234 Natarajan D, Caramaschi D (2010) Animal violence demystified. Front Behav Neurosci 4:1–9 Craig W (1921) Why do animals fight? IJE 31, 264–278 Contala M-L, Krapf P, Steiner FM (2024) Schlick-Steiner, Foraging valor linked with aggression: selection against completely abandoning aggression in the high-elevation ant Tetramorium alpestre? . Insect Sci 31:953–970 Huntingford FA, Turner AK (1987) in Animal Conflict , F. A. Huntingford, A. K. Turner, EdsSpringer Netherlands, Dordrecht,. pp. 227–250 Guo X, Dukas R (2020) The cost of aggression in an animal without weapons. Ethology 126:24–31 Miles-Novelo A, Anderson CA (2022) Climate change and human behavior: Impacts of a rapidly changing climate on human aggression and violence. Cambridge University Press, Cambridge, p 68 Krapf P, Arthofer W, Ayasse M, Steiner FM (2023) Schlick-Steiner, Global change may make hostile – Higher ambient temperature and nitrogen availability increase ant aggression. Sci Total Environ 861:160443 Kudryavtseva NN, Bondar NP, Alekseyenko OV (2000) Behavioral correlates of learned aggression in male mice. Aggressive Behav 26:386–400 Jia Y et al (2021) Gut microbiome modulates Drosophila aggression through octopamine signaling. Nat Commun 12:2698 Mondo E et al (2020) Gut microbiome structure and adrenocortical activity in dogs with aggressive and phobic behavioral disorders. Heliyon 6:e03311 Teseo S et al (2019) The scent of symbiosis: gut bacteria may affect social interactions in leaf-cutting ants. Anim Behav 150:239–254 Van Wilgenburg E, Clémencet J, Tsutsui ND (2010) Experience influences aggressive behaviour in the Argentine ant. Biol Lett 6:152–155 Adamo SA, Hoy RR (1995) Agonistic behaviour in male and female field crickets, Gryllus bimaculatus , and how behavioural context influences its expression. Anim Behav 49:1491–1501 Margalida A, Bertran J, Boudet J, Heredia R (2004) Hatching asynchrony, sibling aggression and cannibalism in the Bearded Vulture Gypaetus barbatus . Ibis 146:386–393 Hsu Y, Earley RL, Wolf LL (2006) Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes. Biol Rev 81:33–74 Sprenger PP, Menzel F (2020) Cuticular hydrocarbons in ants (Hymenoptera: Formicidae) and other insects: how and why they differ among individuals, colonies, and species. Myrmecol News 30:1–26 Martin SJ, Vitikainen E, Drijfhout FP, Jackson D (2012) Conspecific ant aggression is correlated with chemical distance, but not with genetic or spatial distance. Behav Genet 42:323–331 Tillisch K et al (2017) Brain structure and response to emotional stimuli as related to gut microbial profiles in healthy women. Biopsychosoc Med 79:905–913 Kirchoff NS, Udell MAR, Sharpton TJ (2019) The gut microbiome correlates with conspecific aggression in a small population of rescued dogs ( Canis familiaris ). PeerJ 7, e6103 Sarkar A et al (2020) The role of the microbiome in the neurobiology of social behaviour. Biol Rev 95:1131–1166 Faraone SV et al (2005) Molecular genetics of Attention-Deficit/Hyperactivity Disorder. Biol Psychiatry 57:1313–1323 Engelmann C, Haenold R (2016) Transcriptional control of synaptic plasticity by transcription factor NF-κB. Neural Plasticity 7027949 (2016) De Vita J (1979) Mechanisms of interference and foraging among colonies of the harvester ant Pogonomyrmex californicus in the Mojave Desert. Ecology 60:729–737 Steiner FM et al (2007) Abandoning aggression but maintaining self-nonself discrimination as a first stage in ant supercolony formation. Curr Biol 17:1903–1907 Steinmeyer C, Pennings PS, Foitzik S (2012) Multicolonial population structure and nestmate recognition in an extremely dense population of the European ant Lasius flavus . Insectes Soc 59:499–510 Ortius D, Heinze J (1999) Fertility signaling in queens of a North American ant. Behav Ecol Sociobiol 45:151–159 Marinho CGS, Ribeiro MMR, Della Lucia TMC, Guedes RNC (2006) Aggressive response of pest ant species to b-eudesmol (Hymenoptera: Formicidae). Sociobiology 47:445–454 Krapf P et al (2018) An Alpine ant’s behavioural polymorphism: monogyny with and without internest aggression in Tetramorium alpestre . Ethol Ecol Evol 30:220–234 Krapf P, Sedfaoui K, Contala M-L, Steiner FM (2023) Schlick-Steiner, Short-time development of among-colony behaviour in a high-elevation ant. Behav Processes 208:104872 Helanterä H, Strassmann JE, Carrillo J, Queller DC (2009) Unicolonial ants: where do they come from, what are they and where are they going? Trends Ecol Evol 24:341–349 Kuziel GA, Rakoff-Nahoum S (2022) The gut microbiome. Curr Biol 32:R257–R264 R. V. Lenth, emmeans: estimated marginal means, aka least-squares means, R package version 1.11.2-8, (2025) ( https://CRAN.R-project.org/package=emmeans) Fick SE, Hijmans RJ (2017) WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. Int J Climatol 37:4302–4315 Warde-Farley D et al (2010) The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res 38:W214–W220 Errard C, Hefetz A, Jaisson P (2006) Social discrimination tuning in ants: template formation and chemical similarity. Behav Ecol Sociobiol 59:353–363 Lin P et al (2017) Prevotella and Klebsiella proportions in fecal microbial communities are potential characteristic parameters for patients with major depressive disorder. J Affect Disord 207:300–304 Strandwitz P et al (2019) GABA-modulating bacteria of the human gut microbiota. Nat Microbiol 4:396–403 Vernier CL et al (2020) The gut microbiome defines social group membership in honey bee colonies. Sci Adv 6:eabd3431 Ochoa-Repáraz J, Mielcarz DW, Begum- Haque S, Kasper LH (2011) Gut, bugs, and brain: Role of commensal bacteria in the control of central nervous system disease. Ann Neurol 69:240–247 d. Almeida LG, d. LAB, Moraes JR, Trigo C, Omoto FL, Cônsoli (2017) The gut microbiota of insecticide-resistant insects houses insecticide-degrading bacteria: A potential source for biotechnological exploitation. PLoS ONE 12:e0174754 Grant RJ, Daniell TJ, Betts WB (2002) Isolation and identification of synthetic pyrethroid-degrading bacteria. J Appl Micro 92:534–540 Dinan TG, Cryan JF (2017) The microbiome-gut-brain axis in health and disease. Gastroenterol Clin 46:77–89 Bao Y-Y et al (2014) Genomic insights into the serine protease gene family and expression profile analysis in the planthopper, Nilaparvata lugens . BMC Genomics 15:507 Ota KT, Andres W, Lewis DA, Stockmeier CA, Duman RS (2015) BICC1 expression is elevated in depressed subjects and contributes to depressive behavior in rodents. Neuropsychopharmacology 40:711–718 Tanaka T, Goto K, Iino M (2017) Diverse functions and signal transduction of the Exocyst complex in tumor cells. J Cell Physiol 232:939–957 Kiragasi B, Wondolowski J, Li Y, Dickman DK (2017) A presynaptic glutamate receptor subunit confers robustness to neurotransmission and homeostatic potentiation. Cell Rep 19:2694–2706 Liu X, Krause WC, Davis RL (2007) GABA A receptor RDL inhibits Drosophila olfactory associative learning. Neuron 56:1090–1102 Drapeau MD (2003) A novel hypothesis on the biochemical role of the Drosophila Yellow protein. Biochem Biophys Res Com 311:1–3 Tamagnone L et al (1999) Plexins are a large family of receptors for transmembrane, secreted, and gpi-anchored semaphorins in vertebrates. Cell 99:71–80 Gidron Y, Russ K, Tissarchondou H, Warner J (2006) The relation between psychological factors and DNA-damage: A critical review. Biol Psychol 72:291–304 Atanackovic D, Schulze J, Kröger H, Brunner-Weinzierl MC, Deter HC (2003) Acute psychological stress induces a prolonged suppression of the production of reactive oxygen species by phagocytes. J Neuroimmunol 142:159–165 Ayasse M, Paxton RJ, Tengo J (2001) Mating behavior and chemical communication in the order hymenoptera. Annu Rev Entomol 46:31–78 Conrad T, Paxton RJ, Barth FG, Francke W, Ayasse M (2010) Female choice in the red mason bee, Osmia rufa (L.) (Megachilidae). J Exp Biol 213:4065–4073 Tang Y, Horikoshi M, Li W (2016) ggfortify: Unified interface to visualize statistical results of popular R packages. ( https://doi.org/10.32614/RJ-2016-060 ) Maechler M, Rousseeuw P, Struyf A, Hubert M (2025) H. K, cluster: Cluster Analysis Basics and Extensions. ( https://doi.org/10.31234/osf.io/7awm6 ) Roulston TH, Buczkowski G, Silverman J (2003) Nestmate discrimination in ants: Effect of bioassay on aggressive behavior. Insectes Soc 50:151–159 Greenberg ME, Ziff EB (1984) Stimulation of 3T3 cells induces transcription of the c-fos proto-oncogene. Nature 311:433–438 Krapf P, Hochenegger N, Arthofer W, Schlick-Steiner BC, Steiner FM (2019) Comparing ant behaviour indices for fine-scale analyses. Sci Rep 9:6856 Errard C, Hefetz A (1997) Label familiarity and discriminatory ability of ants reared in mixed groups. Insectes Soc 44:189–198 Piry S et al (2004) GENECLASS2: A software for genetic assignment and first-generation migrant detection. J Hered 95:536–539 Rannala B, Mountain JL (1997) Detecting immigration by using multilocus genotypes. Proc Natl Acad Sci USA 94:9197–9201 Queller DC, Goodnight KF (1989) Estimating relatedness using genetic-markers. Evolution 43:258–275 Pamilo P (1991) Evolution of colony characteristics in social insects. II. Number of reproductive individuals. Am Nat 138:412–433 Cicconardi F et al (2020) Genomic signature of shifts in selection in a subalpine ant and its physiological adaptations. Mol Biol Evol 37:2211–2227 Danecek P et al (2011) The variant call format and VCFtools. Bioinformatics 27:2156–2158 Manichaikul A et al (2010) Robust relationship inference in genome-wide association studies. Bioinformatics 26:2867–2873 Zhou X, Stephens M (2012) Genome-wide efficient mixed-model analysis for association studies. Nat Gen 44:821–824 S. D. Turner, qqman: an R package for visualizing GWAS results using QQ and manhattan plots. JOSS 3, 005165 (2018) Love MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15:550 Kolde R pheatmap: Pretty Heatmaps, R package version 1.0.12, 2019 ( https://CRAN.R-project.org/package=pheatmap) Yan L ggvenn: Draw Venn Diagram by 'ggplot2', R package version 0.1.10, 2023 ( https://CRAN.R-project.org/package=ggvenn) Mendoza-Guido B et al (2023) Low diversity and host specificity in the gut microbiome community of Eciton army ants (Hymenoptera: Formicidae: Dorylinae) in a Costa Rican rainforest. Myrmecol News 33:19–34 Magoč T, Salzberg SL (2011) FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27:2957–2963 McMurdie PJ, Holmes S (2013) phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE 8:e61217 Andersen KS, Kirkegaard RH, Karst SM, Albertsen M (2018) ampvis2: an R package to analyse and visualise 16S rRNA amplicon data. bioRxiv . ( https://doi.org/10.1101/299537 ) Venables WN, Ripley BD (2002) Modern Applied Statistics with S, vol 4. Springer, New York, p 498 Core Team R (2022) R: A language and environment for statistical computing Wagner HC et al (2017) Light at the end of the tunnel: Integrative taxonomy delimits cryptic species in the Tetramorium caespitum complex (Hymenoptera: Formicidae). Myrmecol News 25:95–129 Seifert B, Pannier L (2007) A method for standardized description of soil temperatures in terrestrial ecosystems. Abh Ber Naturkundemus Görlitz 78:151–182 R. J. Hijmans, raster: Geographic Data Analysis and Modeling. R package version 3.1-5. (2020) Tóth G, Jones A, Montanarella L (2013) LUCAS Topsoil Survey: methodology, data and results . JRC Technical Reports (Publications Office of the European Union, EUR 26102 - Scientific and Technical Research series, Luxembourg Mazerolle M AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c), R package version 2.3.3, 2023 ( https://cran.r-project.org/package=AICcmodavg) Zeileis A, Hothorn T (2002) Diagnostic checking in regression relationships. R News 2:7–10 Additional Declarations There is NO Competing Interest. Supplementary Files FigS1HierarchClustAnalmod.pdf Figure S1 FigS2agg.pdf Figure S2 FigS3ManhattanDNA.pdf Figure S3 FigS4VolcanoPlot.pdf Figure S4 FigS5PCoA.pdf Figure S5 ForRhtseqallsamplecount.txt Dataset 2 MicrobSNPDEGAggression20260107Supp.docx Supplementary Info ForRresult.assoc.DNAagg.txt Dataset 3 ForRExcelFiles.zip Dataset 1 MicrobSNPDEGAggression.txt Dataset 5 indListpops.txt Dataset 4 Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8539228","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":570915738,"identity":"9a6e7523-eac4-4b56-bbee-5f232cf5e756","order_by":0,"name":"Patrick Krapf","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsElEQVRIiWNgGAWjYFACNhAhJwfhFBCvxdgYwjEgQUtiA9FazBnY0j783GGQvuFG8gPmAmK0WDawHZ7Ze8Ygd8ONNAPmGcRoMTjA3szA2/Ynd8OZAwbMPMRqYfzbZpBucOb4B2K1sB1m5m0zSDA43kOkLZbNbMnMsmcMDGce7yk4TJQWc/Y2Y8a3Owzk+Q6zb3zMU0GMw5iBBGMDhHOACA3QuINpGQWjYBSMglGAFQAAvl4wI/8Z0L8AAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-1625-4473","institution":"University of Amsterdam","correspondingAuthor":true,"prefix":"","firstName":"Patrick","middleName":"","lastName":"Krapf","suffix":""},{"id":570915739,"identity":"4d9e0df5-ad51-4178-a5f9-65c579428df0","order_by":1,"name":"Francesco Cicconardi","email":"","orcid":"","institution":"University of Bristol","correspondingAuthor":false,"prefix":"","firstName":"Francesco","middleName":"","lastName":"Cicconardi","suffix":""},{"id":570915740,"identity":"e7608341-33f1-494e-a3cc-57a43df823aa","order_by":2,"name":"Martin Schilling","email":"","orcid":"","institution":"Universität Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Martin","middleName":"","lastName":"Schilling","suffix":""},{"id":570915741,"identity":"7040a266-ed02-4cf6-a76a-7258c3d58f4f","order_by":3,"name":"Gerhard Aigner","email":"","orcid":"","institution":"Universität Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Gerhard","middleName":"","lastName":"Aigner","suffix":""},{"id":570915742,"identity":"e2c8fc4d-3b47-4f81-993d-2f46fa17adfd","order_by":4,"name":"Thomas Klammsteiner","email":"","orcid":"","institution":"Universität Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Klammsteiner","suffix":""},{"id":570915743,"identity":"21b8d972-2784-400a-bd8b-238dc017ba6e","order_by":5,"name":"Manfred Ayasse","email":"","orcid":"","institution":"University of Ulm","correspondingAuthor":false,"prefix":"","firstName":"Manfred","middleName":"","lastName":"Ayasse","suffix":""},{"id":570915744,"identity":"b5a4a2d5-93a6-42f6-b0f3-a3693c025eda","order_by":6,"name":"Wolfgang Arthofer","email":"","orcid":"","institution":"University of Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Wolfgang","middleName":"","lastName":"Arthofer","suffix":""},{"id":570915745,"identity":"32afb7c5-d8aa-4526-b2fd-49c28ec5b596","order_by":7,"name":"Alexander Mikheyev","email":"","orcid":"https://orcid.org/0000-0003-4369-1019","institution":"Australian National University","correspondingAuthor":false,"prefix":"","firstName":"Alexander","middleName":"","lastName":"Mikheyev","suffix":""},{"id":570915746,"identity":"897d9631-b11e-4913-9c33-9315114afb46","order_by":8,"name":"Birgit Schlick-Steiner","email":"","orcid":"","institution":"University of Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Birgit","middleName":"","lastName":"Schlick-Steiner","suffix":""},{"id":570915747,"identity":"ea091470-56a0-402b-a008-b5ce7fe73ca1","order_by":9,"name":"Florian Steiner","email":"","orcid":"","institution":"University of Innsbruck","correspondingAuthor":false,"prefix":"","firstName":"Florian","middleName":"","lastName":"Steiner","suffix":""}],"badges":[],"createdAt":"2026-01-07 09:15:40","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8539228/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8539228/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99862406,"identity":"ba423478-ace6-45bf-8d1a-916c9afdc503","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2115503,"visible":true,"origin":"","legend":"","description":"","filename":"MicrobSNPDEGAggression20260107Main.docx","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/a0d2866d81206ff5f2cb6622.docx"},{"id":100357504,"identity":"05636151-7e2d-4450-97ad-7d29aa38c21c","added_by":"auto","created_at":"2026-01-16 07:19:56","extension":"json","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10141,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS26000197.json","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/936b5158f9506ab61cc83742.json"},{"id":100357310,"identity":"4ea7ba4a-980e-4702-a4e2-d556a5950f47","added_by":"auto","created_at":"2026-01-16 07:19:38","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":23546,"visible":true,"origin":"","legend":"","description":"","filename":"FigS1HierarchClustAnalmod.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/2611eeba5fad5a0b15b913e6.pdf"},{"id":100358545,"identity":"16542298-aff1-4d38-b647-35fa3528887c","added_by":"auto","created_at":"2026-01-16 07:21:08","extension":"pdf","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11697560,"visible":true,"origin":"","legend":"","description":"","filename":"FigS2agg.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/624dee3bc1d6480fdceb7044.pdf"},{"id":100357700,"identity":"1b6a2919-41f3-4049-a3ab-0f9f7a814ab1","added_by":"auto","created_at":"2026-01-16 07:20:13","extension":"pdf","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53688,"visible":true,"origin":"","legend":"","description":"","filename":"FigS3ManhattanDNA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/7c1b1d29a0f84efe142d2c12.pdf"},{"id":100358513,"identity":"7cc9be2e-5bc4-420d-86a5-44359790e3bd","added_by":"auto","created_at":"2026-01-16 07:21:07","extension":"pdf","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3367702,"visible":true,"origin":"","legend":"","description":"","filename":"FigS4VolcanoPlot.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/09bb5312ec1871f687541a20.pdf"},{"id":99862422,"identity":"c6138013-83b8-45d2-9d68-53a40aa2bc77","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"pdf","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":111081,"visible":true,"origin":"","legend":"","description":"","filename":"FigS5PCoA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/e300b35f0305282d976b1f53.pdf"},{"id":100357598,"identity":"016ae970-eac9-4a75-82c5-8cc49e332586","added_by":"auto","created_at":"2026-01-16 07:20:06","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1010178,"visible":true,"origin":"","legend":"","description":"","filename":"MicrobSNPDEGAggression20260107Supp.docx","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/9bd9d6bd67b702438079a38b.docx"},{"id":99862431,"identity":"b0d1bc73-0dd8-4a23-92a8-2d4e1799d119","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":225338,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS260001970enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/21942d96b493dec5778fada1.xml"},{"id":99862427,"identity":"87f8590b-1d97-4014-8f0d-56fb176f05c7","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"pdf","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2244229,"visible":true,"origin":"","legend":"","description":"","filename":"Fig1EuropePW.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/6df2f54959f0e7bfc2e4e716.pdf"},{"id":99862429,"identity":"e3369443-25c0-46bb-85ea-f28fd159d3e4","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"pdf","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":402377,"visible":true,"origin":"","legend":"","description":"","filename":"Fig2CHCpcaAggFigPCARel.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/2d491bf666e1d3b67bcde675.pdf"},{"id":100357745,"identity":"f2c4a0f6-5159-445f-8470-8bf89f6666aa","added_by":"auto","created_at":"2026-01-16 07:20:16","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":414126,"visible":true,"origin":"","legend":"","description":"","filename":"Fig3PCASNPsManhattanRNAOTU.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/43ce89b8864d63c093c306c7.pdf"},{"id":100358233,"identity":"43f6b371-3251-419c-ba39-766fb24675a9","added_by":"auto","created_at":"2026-01-16 07:20:45","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1083066,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/564c2fb867370fba29d2901f.png"},{"id":100357563,"identity":"b72507ac-774d-488b-b26c-41ec9a7b05fd","added_by":"auto","created_at":"2026-01-16 07:20:04","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":330566,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/4c34083f07d10b12f1d17d8f.png"},{"id":100358158,"identity":"cdec8673-0d5d-4bb6-be7b-187e025a4793","added_by":"auto","created_at":"2026-01-16 07:20:40","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":468862,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/6b7fdf6da54f717f8b5cb8af.png"},{"id":99862417,"identity":"ef6a05ca-0105-48e4-96e2-18174b4f52e1","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":173723,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/d3a0053c0c460698f6274e80.png"},{"id":99862424,"identity":"971ca938-e753-4855-bec2-acb94d361e74","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":75640,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/7ecf37f3f10ffc9a290ef294.png"},{"id":99862428,"identity":"0ae98988-f04d-4e72-b026-322ae6c497c9","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":103141,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/6822c1d24b5703b3acf0e698.png"},{"id":100357904,"identity":"e3d6558f-f905-4edd-b938-78f78dd666f6","added_by":"auto","created_at":"2026-01-16 07:20:28","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":221280,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS260001970structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/f058f0a33d790c65fd52115d.xml"},{"id":99862432,"identity":"47d97843-046c-40ce-9502-8e4a6c446018","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":248627,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/80a3a2cb571b8afa1baa488b.html"},{"id":99862402,"identity":"d44f9367-a232-4457-bb92-4fb8b50b6eea","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":407729,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSampling map and schematic overview of the assays.\u003c/strong\u003e\u0026nbsp;A) Sampling area in Central Europe. The populations were defined based on preliminary data and aggression assays conducted in this study: Population SQ-N in green colours represents single-queened, non-aggressive colonies. Population SQ-A in orange represents single-queened, aggressive colonies. Population MQ-N in blue represents multiple-queened, non-aggressive, potentially supercolonial colonies. The inset in A) shows the three populations in closer detail and the linear distances between all three populations. B) Schematic overview of the assays, created in BioRender by Krapf, P. (2025), https://BioRender.com/ecq4a2x. Note: N\u003csub\u003e2\u003c/sub\u003e\u0026nbsp;= nitrogen; CHC = cuticular hydrocarbons; GC-MS = Gas-Chromatography Mass-Spectrometry\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/0940157b851d1d17d253ac81.png"},{"id":100357892,"identity":"3863119e-dbd8-46b0-8151-7a390d6599c8","added_by":"auto","created_at":"2026-01-16 07:20:27","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":284083,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResults of the aggression tests, PCA of the CHCs, PCA of the LD-pruned SNPs, and within- and between colony relatedness. A)\u003c/strong\u003e Combined boxplots, violin-, and scatter plots displaying the three behavioural states for the three populations along the behaviour index AI, which denotes aggressive (5-1), neutral (0), and peaceful (-4 to -1) behaviour. Workers from all three behavioural states \u003cem\u003estarted aggression\u003c/em\u003e, \u003cem\u003ereacted aggressively\u003c/em\u003e, and \u003cem\u003ereacted peacefully\u003c/em\u003e. Based on their behaviours, we selected workers for whole-genome and whole-transcriptome sequencing. Numbers above colony names represent the sample size for each group. Only workers used in the transcriptomic data are displayed. Note: SQ-N = single-queened and non-aggressive, MQ-N = multiple-queened and non-aggressive colonies. \u003cstrong\u003eB)\u003c/strong\u003e PCA using 63 cuticular hydrocarbon (CHC) compounds that were found in all analysed workers (N\u003csub\u003eworkers\u003c/sub\u003e=44). The three populations SQ-A, SQ-N, and MQ-N differ to some extent in their CHC bouquet on the first axis (45.1%). However, no cluster or complete separation is apparent. \u003cstrong\u003eC)\u003c/strong\u003e PCA of linkage-disequilibrium (LD)-pruned SNPs from whole-genome sequence data (N\u003csub\u003eworkers\u003c/sub\u003e=109). The first axis (4.0%) separates the supercolonial population MQ-N from the aggressive and non-aggressive populations SQ-A and SQ-N. The second axis (2.3%) separates the aggressive population SQ-A (bottom-left panel of the PCA) from the non-aggressive population SQ-N (upper-left panel of the PCA). \u003cstrong\u003eD) \u003c/strong\u003eThe within- and between-rest relatedness was calculated following the Manichaikul et al. (2010)\u003csup\u003e68\u003c/sup\u003e relatedness using genomic SNP data (N\u003csub\u003eworkers\u003c/sub\u003e=109). Each square represents a pairwise comparison between all samples. A dark red colour of a square indicates a close relatedness between two samples, while a dark blue colour indicates a distant/loose relatedness.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/57d3b5f32396aff4fa5b2b9c.png"},{"id":99901484,"identity":"5bba2259-fd42-4616-95de-c0afac1dcc26","added_by":"auto","created_at":"2026-01-09 15:52:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":284083,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eResults of the aggression tests, PCA of the CHCs, PCA of the LD-pruned SNPs, and within- and between colony relatedness. A)\u003c/strong\u003e Combined boxplots, violin-, and scatter plots displaying the three behavioural states for the three populations along the behaviour index AI, which denotes aggressive (5-1), neutral (0), and peaceful (-4 to -1) behaviour. Workers from all three behavioural states \u003cem\u003estarted aggression\u003c/em\u003e, \u003cem\u003ereacted aggressively\u003c/em\u003e, and \u003cem\u003ereacted peacefully\u003c/em\u003e. Based on their behaviours, we selected workers for whole-genome and whole-transcriptome sequencing. Numbers above colony names represent the sample size for each group. Only workers used in the transcriptomic data are displayed. Note: SQ-N = single-queened and non-aggressive, MQ-N = multiple-queened and non-aggressive colonies. \u003cstrong\u003eB)\u003c/strong\u003e PCA using 63 cuticular hydrocarbon (CHC) compounds that were found in all analysed workers (Nworkers=44). The three populations SQ-A, SQ-N, and MQ-N differ to some extent in their CHC bouquet on the first axis (45.1%). However, no cluster or complete separation is apparent. \u003cstrong\u003eC)\u003c/strong\u003e PCA of linkage-disequilibrium (LD)-pruned SNPs from whole-genome sequence data (Nworkers=109). The first axis (4.0%) separates the supercolonial population MQ-N from the aggressive and non-aggressive populations SQ-A and SQ-N. The second axis (2.3%) separates the aggressive population SQ-A (bottom-left panel of the PCA) from the non-aggressive population SQ-N (upper-left panel of the PCA). \u003cstrong\u003eD) \u003c/strong\u003eThe within- and between-rest relatedness was calculated following the Manichaikul et al. (2010)68 relatedness using genomic SNP data (Nworkers=109). Each square represents a pairwise comparison between all samples. A dark red colour of a square indicates a close relatedness between two samples, while a dark blue colour indicates a distant/loose relatedness.\u003c/p\u003e","description":"","filename":"21.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/ef1fe0bd6cc8f827c302aa06.png"},{"id":99862405,"identity":"62554fc8-5f6c-4367-9739-a873bdc074a9","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":283345,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePCA of the allelic states of the SNPs, Venn diagrams of the DEGs, bacterial OTU counts, results of the multinomial logistic regression, and log\u003c/strong\u003e\u003csub\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/sub\u003e\u003cstrong\u003efold change of the DEGs. A)\u003c/strong\u003e\u0026nbsp;Principal Component Analysis (PCA) of the genomic single nucleotide polymorphisms (SNP) states from the three genes that were above the suggestive line in the Genome-wide Efficient Mixed Model Association (GEMMA) analysis (see also Manhattan Plot, Fig. S4, (N\u003csub\u003eworkers\u003c/sub\u003e=109). In the PCA, individuals that are homozygous for the reference allele of the respective genes are represented as 0/0 and individuals that are heterozygous for the reference alleles are 0/1. The three behavioural states\u0026nbsp;\u003cem\u003estarted aggression\u003c/em\u003e,\u0026nbsp;\u003cem\u003ereacted aggressively\u003c/em\u003e, and\u0026nbsp;\u003cem\u003ereacted peacefully\u003c/em\u003e\u0026nbsp;are coloured in red, yellow, and purple, respectively. In the PCA, the behavioural state\u0026nbsp;\u003cem\u003estarted aggression\u003c/em\u003e\u0026nbsp;displays allelic combinations that are not observed in the other two behavioural states.\u0026nbsp;\u003cstrong\u003eB)\u003c/strong\u003e\u0026nbsp;Venn diagrams of 57 differentially-expressed genes (DEGs) which were significantly down- and up-regulated (FDR-corrected for multiple testing with \u0026lt;0.05) with known gene names based on Flybase (https://flybase.org/, retrieved 10.04.2025). Dark colours represent down-regulated genes and light colours up-regulated genes.\u0026nbsp;\u003cstrong\u003eC)\u003c/strong\u003e\u0026nbsp;Counts of nine operational taxonomic units (OTU) that were associated with the behavioural states in the multinomial logistic regression. Significant differences in the counts are represented with asterisks, *** = \u0026lt;0.001; ** = \u0026lt;0.01; * = \u0026lt;0.05.\u0026nbsp;\u003cstrong\u003eD)\u003c/strong\u003e\u0026nbsp;Results of the multinomial logistic regression displaying the logistic odds-ratio values for each SNPs and DEGs separately including the Chi-squared and p-values from the Likelihood Ratio Test (LRT) to assess the global significance of the SNPs or DEGs on the behaviour.\u0026nbsp;\u003cstrong\u003eE)\u003c/strong\u003e\u0026nbsp;Log\u003csub\u003e2\u003c/sub\u003efold change for the DEGs that we identified as significant in the multinomial logistic regression shown for the comparison of\u0026nbsp;\u003cem\u003estarted aggression\u0026nbsp;\u003c/em\u003evs\u0026nbsp;\u003cem\u003ereacted aggressively\u0026nbsp;\u003c/em\u003eand\u0026nbsp;\u003cem\u003estarted aggression\u0026nbsp;\u003c/em\u003evs\u0026nbsp;\u003cem\u003ereacted peacefully\u003c/em\u003e. The comparison\u0026nbsp;\u003cem\u003ereacted aggressively\u0026nbsp;\u003c/em\u003evs\u003cem\u003e\u0026nbsp;reacted peacefully\u0026nbsp;\u003c/em\u003edid not yield any significant up- or down-regulated genes and no data are now shown.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/b5de83e56d839ec3542f56dc.png"},{"id":100796279,"identity":"ffcf89ca-2332-428f-b8a5-3030c3b01a99","added_by":"auto","created_at":"2026-01-21 13:42:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2968937,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/d11bba41-f037-4061-859a-af319933a9f1.pdf"},{"id":99862403,"identity":"a5b574e5-15ef-42c2-a660-54b9e02c4384","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":23546,"visible":true,"origin":"","legend":"\u003cp\u003eFigure S1\u003c/p\u003e","description":"","filename":"FigS1HierarchClustAnalmod.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/84ccd0d082a839e3e364d897.pdf"},{"id":100358412,"identity":"5278669d-3292-4ecb-8caa-6732be9acf12","added_by":"auto","created_at":"2026-01-16 07:21:02","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":11697560,"visible":true,"origin":"","legend":"Figure S2","description":"","filename":"FigS2agg.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/9af62fcf2a70d6bf87526809.pdf"},{"id":99862409,"identity":"fd2313da-41b5-4a39-a532-36a51f2c9e7f","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":53688,"visible":true,"origin":"","legend":"Figure S3","description":"","filename":"FigS3ManhattanDNA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/02257848a026f8c34f2b00c9.pdf"},{"id":100357899,"identity":"a97484ce-912f-4c1a-a1ee-ff18d2c26b2d","added_by":"auto","created_at":"2026-01-16 07:20:28","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":3367702,"visible":true,"origin":"","legend":"Figure S4","description":"","filename":"FigS4VolcanoPlot.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/693c62452e1d1f694d0d250b.pdf"},{"id":99862407,"identity":"4ae3fcb9-3e97-4403-83f5-959a0bc33321","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":111081,"visible":true,"origin":"","legend":"Figure S5","description":"","filename":"FigS5PCoA.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/27cc352be8182a74119ec8ba.pdf"},{"id":99862419,"identity":"2f3124ea-9078-464d-949d-c43d8356f1e8","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"txt","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":5021327,"visible":true,"origin":"","legend":"Dataset 2","description":"","filename":"ForRhtseqallsamplecount.txt","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/39d2a5da65b6f2dc18db022e.txt"},{"id":100358126,"identity":"63870b7e-5d94-4a4d-aa97-6ed3b3e8a402","added_by":"auto","created_at":"2026-01-16 07:20:39","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":1010178,"visible":true,"origin":"","legend":"Supplementary Info","description":"","filename":"MicrobSNPDEGAggression20260107Supp.docx","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/b893bfcaace2ad48a1013de9.docx"},{"id":99862436,"identity":"6fe377b2-d4e5-413b-899a-8a2df8fba8ca","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"txt","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":33371856,"visible":true,"origin":"","legend":"Dataset 3","description":"","filename":"ForRresult.assoc.DNAagg.txt","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/9a8c14ab49b6c45d0382ead4.txt"},{"id":99862411,"identity":"37a5c2dc-9fb6-4dfe-9c5a-b03d55346148","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"zip","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":240565,"visible":true,"origin":"","legend":"Dataset 1","description":"","filename":"ForRExcelFiles.zip","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/a147e580062cae594f733498.zip"},{"id":99862425,"identity":"be531490-e8cb-45b2-b03c-6a3388a78abc","added_by":"auto","created_at":"2026-01-09 07:17:04","extension":"txt","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":136238,"visible":true,"origin":"","legend":"Dataset 5","description":"","filename":"MicrobSNPDEGAggression.txt","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/b99c3d6684ba5aa3915e0fbe.txt"},{"id":99862415,"identity":"babab11d-8b24-42be-8ac8-d66a26c924d0","added_by":"auto","created_at":"2026-01-09 07:17:03","extension":"txt","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":870,"visible":true,"origin":"","legend":"Dataset 4","description":"","filename":"indListpops.txt","url":"https://assets-eu.researchsquare.com/files/rs-8539228/v1/bc6fd6159877eac04bdb6ecd.txt"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"The gut microbiome, single nucleotide polymorphisms, and differentially expressed genes promote aggression in an ant","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAggressive behaviour among individuals of the same species is a frequently observed behaviour in animals\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. It is a vital aspect of animals\u0026rsquo; fitness and survival and often context-dependent\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. For example, it can occur during food or mate competition, territory defence, and offspring protection against predators\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Such adaptive aggression\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e can lead to increased fitness. For instance, winners of fights can consume more or higher-quality food or obtain mates for reproduction\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. However, aggression can incur harms such as stress and energy or time costs. At its worst, it can also be deadly\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e by increasing the risk of injuries and/or exposure to predators\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eVarious extrinsic and intrinsic factors can lead to aggression. Extrinsic factors are, among others, higher ambient temperature and can lead to increased aggression in humans and animals\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Intrinsic factors such as experience (i.e., repeated stimuli such as winning aggressive encounters)\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e, neurochemical factors (i.e. changes in serotonin, dopamine, or octopamine)\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, or differentially-expressed genes (DEGs)\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e influenced by the gut microbiome can also promote aggression\u003csup\u003e\u003cspan additionalcitationids=\"CR12\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eDespite these promising insights, our understanding of the underlying mechanisms that lead to the start of aggression (i.e., when two individuals meet and one starts aggressive behaviours such as fighting) is limited. Nevertheless, some drivers are known: for example, individual experience\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e,\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, previous experience in winning a fight\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e, or recognising another individual\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e can affect whether an individual starts aggression. In particular, animals such as insects use chemical cues\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e (cuticular hydrocarbons; CHCs) to recognise and attack enemies\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. Besides experience and recognition, the microbiome\u003csup\u003e11\u0026ndash;13,20\u0026minus;22\u003c/sup\u003e, genetic changes (e.g., mutations in genes\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e), and/or DEGs (e.g., in neuronal or synaptic functions\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e) may also affect whether individuals start aggression.\u003c/p\u003e \u003cp\u003eAnts are known for their aggressive behaviour. For example, California harvester ants (\u003cem\u003ePogonomyrmex californicus\u003c/em\u003e) often fight for over 30 minutes, and such fights often result in fatal outcomes with one or both workers dying\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. On the other end of this spectrum are \u0026lsquo;peaceful\u0026rsquo; ants, which frequently refrain from fighting individuals from different colonies of the same species. Peacefull behaviour is less frequently observed, but is known from several species such as \u003cem\u003eLasius austriacus\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eLasius flavus\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e, and \u003cem\u003eTetramorium alpestre\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. However, even in such predominantly peaceful species, aggression can be observed, leading to the unresolved question of what factors lead to the start of aggression\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eHere, we used the high-elevation ant species \u003cem\u003eT. alpestre\u003c/em\u003e to test whether chemical, microbiome, genomic, and/or transcriptomic traits correlate with the start of aggression in ants, specifically workers. This species displays a behavioural continuum ranging from aggression to peacefulness\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. We collected workers from three colonies each from three previously described populations\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. They either comprise single-queened and aggressive colonies (SQ-A), single-queened and non-aggressive colonies (SQ-N), or multiple-queened and non-aggressive colonies MQ-N (i.e., supercolonies consisting of multiple colonies connected over a large area\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e, N\u003csub\u003ecol\u003c/sub\u003e = 9, Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA-B, Tab. S1). We conducted recognition (own colony against alien colony) and aggression assays and selected individual worker ants that displayed either of the following behavioural states, \u003cem\u003estarted aggression\u003c/em\u003e, \u003cem\u003ereacted aggressively\u003c/em\u003e, or \u003cem\u003ereacted peacefully\u003c/em\u003e for chemical, microbiome, genomic, and transcriptomic analyses (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC). We then integrated results from these analyses in a final multinomial logistic regression to assess their joint impact on the behavioural states.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eAggression tests, and selection of workers for whole-genome and -transcriptome sequencing\u003c/h2\u003e \u003cp\u003eTo select workers for whole-genome and transcriptome sequencing that displayed either of the three behaviours, \u003cem\u003estarted aggression, reacted aggressively\u003c/em\u003e, and \u003cem\u003ereacted peacefully\u003c/em\u003e, we conducted standardised one-on-one worker aggression tests\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e among all nine colonies. We analysed the behaviour of each individual worker and calculated a behaviour index. By conducting an Analysis of Variance (ANOVA), we found that the behaviour differed among the behavioural states (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA; ANOVA: df\u0026thinsp;=\u0026thinsp;2, F-value\u0026thinsp;=\u0026thinsp;78.02, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001). To confirm that peaceful behaviour has lower aggression values, we pairwise compared the behavioural states using a Tukey Honest Significant Test: Workers that \u003cem\u003estarted aggression\u003c/em\u003e and ones that \u003cem\u003ereacted aggressively\u003c/em\u003e had significantly higher aggression values throughout the confrontations than workers that \u003cem\u003ereacted peacefully\u003c/em\u003e (\u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001; \u003cem\u003ereacted aggressively\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001). However, workers that \u003cem\u003estarted aggression\u003c/em\u003e and ones that \u003cem\u003ereacted aggressively\u003c/em\u003e had similar aggression values (\u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e, p-value\u0026thinsp;=\u0026thinsp;0.597). The within-colony behaviour (control; not shown) did not reveal any aggression. Additionally, workers preferred own odours over alien odours or a control (for details, see the section \u0026ldquo;\u003cem\u003eRecognition assays\u0026rdquo;\u003c/em\u003e in the Supplementary Results). Based on the aggression tests and ANOVA, we selected 85 and 109 workers for whole-transcriptome and whole-genome sequencing, respectively.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eCuticular hydrocarbon (CHC) analysis\u003c/h3\u003e\n\u003cp\u003eThe CHC bouquet did not differ starkly among colonies and populations. We found 78 compounds in the odour bouquets (hydrocarbon chain length C12 to C35; GC-MS analyses of CHC-extracts of five workers pooled per colony). From these, 63 compounds were present in all samples (Tab. S2). Visualised multidimensionally (PCA, Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB), colonies of the single-queened and aggressive population SQ-A (colonies SQ-A2, SQ-A5 SQ-A6) overlapped with colonies of the single-queened and non-aggressive population SQ-N (SQ-N1, SQ-N4, SQ-N6) and of the multi-queened and non-aggressive population MQ-N (MQ-N1, MQ-N2, MQ-N5), but population MQ-N did so the most. Using the CHC compound data, we conducted a hierarchical cluster analysis and found that CHC extracts from SQ-N and MQ-N were more similar to each other and partially clustered together (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). In contrast, samples from SQ-A5 were more similar to colonies from populations SQ-N and MQ-N than to SQ-A2 and SQ-A6 colonies.\u003c/p\u003e\n\u003ch3\u003eWhole-genome and whole-transcriptome analyses\u003c/h3\u003e\n\u003cp\u003eObserved heterozygosity and pairwise genomic differentiation were similar among samples, but relatedness was higher in multiple-queened and non-aggressive colonies. After quality checks and filtering, 184,145 and 69,191 Single Nucleotide Polymorphisms (SNPs) were kept in whole-genome and whole-transcriptome VCF files, respectively (109 and 83 samples, respectively). The mean observed heterozygosity for whole-genome samples was 0.30 (min\u0026thinsp;=\u0026thinsp;0.23, max\u0026thinsp;=\u0026thinsp;0.48) and for whole-transcriptome samples 0.15 (min\u0026thinsp;=\u0026thinsp;0.001, max\u0026thinsp;=\u0026thinsp;0.37). The pairwise genomic differentiation values (Weir-Cockerham F\u003csub\u003eST\u003c/sub\u003e) were very similar across populations with 0.005 for populations SQ-A:SQ-N, 0.004 for SQ-A:MQ-N, and 0.008 for SQ-N:MQ-N. Visualised multidimensionally (linkage disequilibrium-pruned PCA with DNA samples; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC), the multiple-queened and non-aggressive population MQ-N separated from the other two single-queened populations SQ-A and SQ-N. Samples from population SQ-N clustered together, while colonies from population SQ-A appeared well separated. Samples from population MQ-N clustered together regardless of colony identity. Mean within-colony relatedness was slightly higher in populations SQ-A and SQ-N than in population MQ-N (SQ-A: 0.57, SQ-N: 0.63; MQ-N: 0.33; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD, Tab. S3). The colony queen number (estimated using the relatedness values) was approximately one in all colonies of populations SQ-A and SQ-N and at least two in all colonies of population MQ-N (Tab. S3).\u003c/p\u003e \u003cp\u003eWe identified three SNPs associated with the behavioural states using a Genome-wide Efficient Mixed Model Association (GEMMA) analysis. We assessed associations between SNPs as well as Insertions/Deletions (\u0026lsquo;InDels\u0026rsquo;) and three SNPs (henceforth SNP1-3; Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). SNP1 is in the sequence of the gene called \u0026ldquo;Mediator of RNA polymerase II transcription subunit 26\u0026rdquo; (located at Scaffold 11, site 2457277). SNP2 is in the sequence of an unknown gene (located at Scaffold 165; site 28,302). SNP3 is in the sequence of the gene \u0026ldquo;\u003cem\u003egastrulation-defective\u003c/em\u003e\u0026rdquo; (\u003cem\u003egd\u003c/em\u003e; located at Scaffold 185, site 110,741).\u003c/p\u003e \u003cp\u003eThe allelic states of the genomic SNPs differed between the behavioural states. We visualised the homozygous and heterozygous allelic states for the three behavioural states multidimensionally (PCA; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). The behavioural state \u003cem\u003estarted aggression\u003c/em\u003e revealed a larger variation and had slightly different allelic states in the SNPs compared with the other two behavioural states. We assessed if the count of the reference or alternative alleles for each specific SNP was different across rows (Pearson\u0026rsquo;s Chi-squared test for count data with simulated p-value and 2000 Monte Carlo replicates) and subsequent post-hoc test. For SNP1, 36 out of 41 (88%) workers that \u003cem\u003estarted aggression\u003c/em\u003e, 35 out of 39 workers (90%) that \u003cem\u003ereacted aggressively\u003c/em\u003e, and 26 out of 29 (90%) workers that \u003cem\u003ereacted peacefully\u003c/em\u003e were homozygous for the reference allele (Fig. \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e; Tab. S4). For SNP1, the allele counts were not different across behavioural states (Fig. \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e, one-SNP1; χ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.09, p-value\u0026thinsp;=\u0026thinsp;1.000). For SNP2, 18 out of 41 (43%) workers that \u003cem\u003estarted aggression\u003c/em\u003e were homozygous in the SNP state, while 2 out of 39 (5%) workers that \u003cem\u003ereacted aggressively\u003c/em\u003e and 2 out of 29 (7%) workers that \u003cem\u003ereacted peacefully\u003c/em\u003e were homozygous in the SNP state. More individuals that \u003cem\u003estarted aggression\u003c/em\u003e were homozygous for the reference allele (SNP2: χ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;22.98, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001; see Tab. S4 for pairwise comparisons). For SNP3, 30 out of 41 (73%) workers that \u003cem\u003estarted aggression\u003c/em\u003e were heterozygous for the SNP state, 10 out of 39 (26%) workers that \u003cem\u003ereacted aggressively\u003c/em\u003e, and 11 out of 29 (38%) workers that \u003cem\u003ereacted peacefully\u003c/em\u003e were heterozygous for the behavioural state. More individuals that \u003cem\u003estarted aggression\u003c/em\u003e were heterozygous for the reference allele (SNP3: χ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;19.381, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eDifferential gene-expression analyses\u003c/h3\u003e\n\u003cp\u003eWe identified several differentially-expressed genes (DEGs) associated with the behavioural states. We pairwise compared all behavioural states (i.e., \u003cem\u003estarted aggression\u003c/em\u003e (N\u003csub\u003eAntsSeq\u003c/sub\u003e = 31), \u003cem\u003ereacted aggressively\u003c/em\u003e (N\u003csub\u003eAntsSeq\u003c/sub\u003e = 28), and \u003cem\u003ereacted peacefully\u003c/em\u003e (N\u003csub\u003eAntsSeq\u003c/sub\u003e = 23). In each comparison, we found\u0026thinsp;~\u0026thinsp;17,000 DEGs or isoforms, of which roughly 100 were significant (for details on the comparisons, see the section \u0026ldquo;\u003cem\u003eDifferentially-expressed genes\u003c/em\u003e\u0026rdquo; in the Supplementary Results). In the comparison between ants that \u003cem\u003estarted aggression\u003c/em\u003e and \u003cem\u003ereacted aggressively\u003c/em\u003e, we found 13 significantly up-regulated genes and 36 down-regulated genes in workers that \u003cem\u003estarted aggression\u003c/em\u003e (false-discovery rate (FDR)-corrected for multiple testing; Fig. \u003cspan refid=\"MOESM4\" class=\"InternalRef\"\u003eS4\u003c/span\u003eA, volcano plot - red dots). When comparing ants that \u003cem\u003estarted aggression\u003c/em\u003e with workers that \u003cem\u003ereacted peacefully\u003c/em\u003e, we found 28 and 61 genes that were significantly up- and down-regulated, respectively. In the comparison between ants that \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e, no gene was significantly up- or down-regulated in workers that \u003cem\u003estarted aggression\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eWe found 30 up-regulated and 28 down-regulated genes that were shared across all behavioural comparisons (\u0026lt;\u0026thinsp;0.05 -corrected genes with known functions were used; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). For the 30 up-regulated genes, five genes (\u003cem\u003eCG3800, CG3902, CDase, Rhp\u003c/em\u003e, and \u003cem\u003eMoe\u003c/em\u003e; Tab. S5) were exclusively found in the comparison \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e, 22 genes (\u003cem\u003eCG34367, CG13625, CG3655, apolpp, CG14687, CG3655, Gat, Sur-8, mRpL9, CG9175, CG6656, Phm, Socs16D, Vav, CG3860, CG32225, CG9426, alph, CG16974, CG10483, AP-2alpha\u003c/em\u003e, and \u003cem\u003ebchs\u003c/em\u003e; Tab. S5) exclusively in the comparison \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, and three (\u003cem\u003eCG3061, svr\u003c/em\u003e, and \u003cem\u003eSyt4\u003c/em\u003e; Tab. S5) in both comparisons \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e. No gene was found to be differentially expressed in the comparison \u003cem\u003ereacted aggressively\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eFor the 28 down-regulated genes, eight genes (\u003cem\u003eBicC, CG3238, Exo84, PlexA, Tret1-1, Taf5, Doa\u003c/em\u003e, and \u003cem\u003eVps35\u003c/em\u003e; Tab. S5; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB) were found in the comparison \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e, 19 in the comparison \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e (\u003cem\u003eCG3822, Rdl, RFC3, CG10431, Cdep, Dscam1, CG7492, CG6910, snRNP-U1-70K, CG31550, l(1)G0196, CG9346, CG32486, agt, Gcn5, baz, CG13366, U2af50\u003c/em\u003e, and \u003cem\u003eCG8108\u003c/em\u003e; Tab. S5), and one (\u003cem\u003eyellow-d2\u003c/em\u003e; Tab. S5) was found in both comparisons \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e. The log\u003csub\u003e2\u003c/sub\u003efold changes of these genes ranged between \u0026minus;\u0026thinsp;3.06 and \u0026minus;\u0026thinsp;0.15 for \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e and between \u0026minus;\u0026thinsp;1.60 and \u0026minus;\u0026thinsp;0.15 for \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e (Tab. S5). Of these genes, two were highly expressed: gene \u003cem\u003eCG3800\u003c/em\u003e was highly up-regulated (log\u003csub\u003e2\u003c/sub\u003efold change\u0026thinsp;=\u0026thinsp;2.48) and gene \u003cem\u003eBicC\u003c/em\u003e was highly down-regulated (log\u003csub\u003e2\u003c/sub\u003efold change = -3.06). All the above-mentioned genes were used for the multinomial regression analyses (for details, see section below \u0026ldquo;\u003cem\u003eAnalysing multiple data layers jointly\u003c/em\u003e\u0026rdquo;).\u003c/p\u003e\n\u003ch3\u003eAnalysis of high-throughput 16S rRNA gene sequencing data\u003c/h3\u003e\n\u003cp\u003eTo assess whether the laboratory maintenance affected the microbiome and whether the microbiome (i.e., bacteria and archaea) is associated with the three behavioural states, 16S rRNA gene sequencing was conducted with 49 workers from the populations SQ-A and SQ-N, and 16 additional \u0026ldquo;control\u0026rdquo; samples from SQ-A and SQ-N (i.e., directly frozen in the field and not used in aggression tests; for details, see the Materials and Methods section). Subsequently, we conducted a Principal Coordinate Analysis (PCoA) with these samples. Laboratory maintenance did not change the bacterial operational taxonomic units (OTUs) composition in the ants (Fig. \u003cspan refid=\"MOESM5\" class=\"InternalRef\"\u003eS5\u003c/span\u003eA). Also, the behavioural states were mixed with control samples (Fig. \u003cspan refid=\"MOESM5\" class=\"InternalRef\"\u003eS5\u003c/span\u003eB). Only samples from one colony (SQ-N6) were separated from the other samples on the first axis.\u003c/p\u003e \u003cp\u003eFour bacterial genera were frequently found across the data set. From an average of 116,096\u0026thinsp;\u0026plusmn;\u0026thinsp;20,583 raw reads per sample, 79,844\u0026thinsp;\u0026plusmn;\u0026thinsp;19,799 quality-filtered reads per sample remained, subsampled to an equal depth of 37,808 reads. After rarefaction, 22,215 unique OTUs were identified, including 264 archaea, 19,773 bacteria, and 2178 unknown OTUs. We excluded OTUs not classified at the genus level. Of the remaining OTUs, the genera \u003cem\u003ePseudomonas\u003c/em\u003e (8.7% relative abundance), \u003cem\u003eBacteroides\u003c/em\u003e (6.1%), \u003cem\u003eLactobacillus\u003c/em\u003e (5.2%), and \u003cem\u003ePrevotella\u003c/em\u003e (4.4%) were found most frequently.\u003c/p\u003e \u003cp\u003eBesides these four bacterial OTUs (\u003cem\u003ePseudomonas\u003c/em\u003e, \u003cem\u003eBacteroides\u003c/em\u003e, \u003cem\u003eLactobacillus\u003c/em\u003e, and \u003cem\u003ePrevotella\u003c/em\u003e), we further calculated the relative frequency for four additional bacterial genera and one order, namely \u003cem\u003eAcetobacter, Enterococcus, Fusobacterium, Megamonas\u003c/em\u003e, and the order Rhizobiales. These bacteria are also known to affect behaviour in humans\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eDrosophila\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, and ants\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. The genus \u003cem\u003ePseudomonas\u003c/em\u003e was most frequent (25.4%, Tab. S6) followed by \u003cem\u003eBacteroides\u003c/em\u003e (20.7%), \u003cem\u003eLactobacillus\u003c/em\u003e (18.6%), \u003cem\u003ePrevotella\u003c/em\u003e (17.0%), \u003cem\u003eEnterococcus\u003c/em\u003e (11.3%), \u003cem\u003eMegamonas\u003c/em\u003e (5.4%), \u003cem\u003eAcetobacter\u003c/em\u003e (0.7%), \u003cem\u003eFusobacterium\u003c/em\u003e (0.5%), and the order Rhizobiales (0.3%). Using these bacterial OTU genera, we selected OTUs that had a frequency of at least 100 across the behavioural states (N\u0026thinsp;=\u0026thinsp;119), thus focusing on the most frequent OTUs.\u003c/p\u003e \u003cp\u003eWith these 119 OTUs, we conducted a sliding-window approach in a multinomial logistic regression to count how often they were significantly associated with the behaviour states. Across these models, the most frequent OTUs (N\u003csub\u003eOTUs\u003c/sub\u003e=58 with a frequency\u0026thinsp;\u0026ge;\u0026thinsp;10) included the genera \u003cem\u003eBacteroides\u003c/em\u003e (25% relative percentage across 58 models), \u003cem\u003eLactobacillus\u003c/em\u003e (9%), \u003cem\u003ePrevotella\u003c/em\u003e (43%), \u003cem\u003ePseudomonas\u003c/em\u003e (17%), the order Rhizobiales (4%), and the genus \u003cem\u003eFusobacterium\u003c/em\u003e (1%). We further reduced the OTU number for downstream analyses yielding 18 OTUs (e.g., using OTUs with a higher or lower frequency than one across the counts of the behavioural states; for details see \u0026ldquo;\u003cem\u003eAnalysis of 16S rRNA gene-sequencing data\u003c/em\u003e\u0026rdquo; in the Materials and Methods). With these 18 OTUs, we assessed whether OTU counts differed among behavioural states by conducting a generalised linear model and pairwise comparison (\u0026ldquo;emmeans\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e; Tukey corrected for multiple testing).\u003c/p\u003e \u003cp\u003eNine OTUs were significantly associated with the behavioural states, namely two \u003cem\u003eBacteroides\u003c/em\u003e spp., two \u003cem\u003eLactobacillus\u003c/em\u003e spp., three \u003cem\u003ePrevotella\u003c/em\u003e spp., and two \u003cem\u003ePseudomonas\u003c/em\u003e spp. (Tab. S8, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). In the two \u003cem\u003eBacteroides\u003c/em\u003e species, significantly more OTU counts occurred in the behavioural state \u003cem\u003estarted aggression\u003c/em\u003e and \u003cem\u003ereacted aggressively\u003c/em\u003e than in \u003cem\u003ereacted peacefully\u003c/em\u003e (OTU 1598) as well as fewer counts in \u003cem\u003estarted aggression\u003c/em\u003e than \u003cem\u003ereacted aggressively\u003c/em\u003e or \u003cem\u003ereacted peacefully\u003c/em\u003e (OTU 22324). For the genus \u003cem\u003eLactobacillus\u003c/em\u003e, significantly fewer and more OTU counts occurred in the behavioural state \u003cem\u003ereacted peacefully\u003c/em\u003e than in \u003cem\u003estarted aggression\u003c/em\u003e or \u003cem\u003ereacted aggressively\u003c/em\u003e in \u003cem\u003eLactobacillus mucosae\u003c/em\u003e and in \u003cem\u003eLactobacillus\u003c/em\u003e sp., respectively. In the three \u003cem\u003ePrevotella\u003c/em\u003e and two \u003cem\u003ePseudomonas\u003c/em\u003e species, significantly more OTU counts occurred in the behavioural state \u003cem\u003estarted aggression\u003c/em\u003e than in the state \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e (OTUs \u003cem\u003ePrevotella\u003c/em\u003e 377, 1887, 20448; \u003cem\u003ePseudomonas\u003c/em\u003e 366, 2442). For the three \u003cem\u003ePrevotella\u003c/em\u003e species, also more OTU counts occurred in the behavioural state \u003cem\u003ereacted peacefully\u003c/em\u003e than in \u003cem\u003ereacted aggressively\u003c/em\u003e.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eAnalysing multiple data layers jointly\u003c/h2\u003e \u003cp\u003eWe integrated genomic, transcriptomic, chemical, and environmental data layers in 24 multinomial logistic regression models to assess if they were associated with the behavioural states. The site-specific environmental variables were calculated manually or extracted from the WorldClim dataset\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e (for details, see \u0026ldquo;\u003cem\u003eEnvironmental variables used in the multinomial regression analyses\u003c/em\u003e\u0026rdquo; in the Materials and Methods section\u0026rdquo;). In more detail, we used SNPs, gene-expression counts, within-colony relatedness, site-specific air temperature, the first PC of the CHC analysis, soil nitrogen values, mean annual precipitation, precipitation of the warmest quarter, mean annual temperature, and maximum temperature of the warmest month as explanatory variables (for details, see Materials and Methods section \u0026ldquo;\u003cem\u003eCombining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression\u003c/em\u003e\u0026ldquo;).\u003c/p\u003e \u003cp\u003eWe used allelic states, normalized expression counts, and continuous environmental variables as predictors in a single model. We excluded microbiome data because they were not available for the multiple-queen population MQ-N. In the models, we used the behavioural state \u003cem\u003estarted aggression\u003c/em\u003e as the baseline and run an intercept-only model as reference. To find the best model explaining the data, we selected various combinations of explanatory variables resulting in 24 models: Models 1\u0026ndash;8 used only the four genes that were found in both behavioural comparisons (genes \u003cem\u003eCG3061, svr, Syt4, yellow-d2\u003c/em\u003e), and Models 9\u0026ndash;16 and Models 17\u0026ndash;24 included either all up-regulated or all down-regulated genes found in the differential gene expression analysis, respectively. We used log-likelihood ratio tests to the robustness of the variables.\u003c/p\u003e \u003cp\u003eIn the four best-fitting models, we found two SNPs and eight genes that were associated with the behavioural states. The models were Model 2, 4, 12, and 20, which included the following SNPs and DEGs: SNP2 and SNP3 (\u003cem\u003egastrulation-defective\u003c/em\u003e) and DEGs \u003cem\u003eyellow-d2, BicC, Pif1, Exo84, PlexA, KaiR1d, Rdl\u003c/em\u003e, and \u003cem\u003eRFC3\u003c/em\u003e (for detailed results, see the supplementary results section \u0026ldquo;\u003cem\u003eLogistic multinomial regression analyses\u003c/em\u003e\u0026rdquo; and Tab. S9-17). Although SNP1 was significant in Model 20, we excluded it because it was not significant in the SNPs-only model. We then combined the above-mentioned variables (i.e., SNP2, SNP3, \u003cem\u003eyellow-d2, BicC, Pif1, Exo84, KaiR1d, RD, RFC3\u003c/em\u003e, \u003cem\u003ePlexA\u003c/em\u003e) in a final multinomial logistic regression. This final model explained significantly more variance than the intercept-only model (Likelihood ratio test of multinomial models, likelihood ratio: 89.73, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001). A goodness-of-fit measure was calculated by comparing the fit of observed and expected values, and the model displayed a good fit (χ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;74.89, df\u0026thinsp;=\u0026thinsp;4; p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001, Residual deviance\u0026thinsp;=\u0026thinsp;89.23, AIC\u0026thinsp;=\u0026thinsp;133.23). SNP2 and SNP3 significantly influenced the behavioural states, as well as genes \u003cem\u003eyellow-d2, BicC, Pif1, Exo84, KaiR1d, RD\u003c/em\u003e, and \u003cem\u003eRFC3\u003c/em\u003e, but not gene \u003cem\u003ePlexA\u003c/em\u003e (Tab. S18). In contrast, within-colony relatedness, CHCs, and the environmental variables were never associated with the start of aggression.\u003c/p\u003e \u003cp\u003eThe SNPs and DEGs contributed significantly to the behavioural associations, but the two SNPs and gene \u003cem\u003eBicC\u003c/em\u003e contributed the most. Overall, the odds ratios of the SNPs and DEGs being associated with the behavioural states ranged from \u0026minus;\u0026thinsp;4.13 to 4.90 (Tab. S18; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD). The highest log\u003csub\u003en\u003c/sub\u003e-values were found for SNP2, which were 4.90 times higher for \u003cem\u003ereacted aggressively\u003c/em\u003e and 3.80 times higher for \u003cem\u003ereacted peacefully\u003c/em\u003e compared with \u003cem\u003estarted aggression\u003c/em\u003e. The next highest values were in \u003cem\u003eBicC\u003c/em\u003e, which were 0.6 times higher for \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e compared with \u003cem\u003estarted aggression\u003c/em\u003e. The odds ratios that the DEGs \u003cem\u003eExo84, yellow-d2, KaiR1d, Pif1, Rdl\u003c/em\u003e, and \u003cem\u003eRFC3\u003c/em\u003e were associated with the behavioural states were approximately 0.01 to 0.4 times higher for \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e compared with \u003cem\u003estarted aggression\u003c/em\u003e. For SNP3, values were \u0026minus;\u0026thinsp;4.1 and \u0026minus;\u0026thinsp;3.8 times lower for \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e compared with \u003cem\u003estarted aggression\u003c/em\u003e. The calculated pseudo-R\u003csup\u003e2\u003c/sup\u003e value \u0026ldquo;Nagelkerke\u0026rdquo; was 0.75. In the pairwise comparison of the behavioural states (\u0026ldquo;emmeans\u0026rdquo;), the mean of \u003cem\u003estarted aggression\u003c/em\u003e was lower than the means of \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e (means\u0026thinsp;+\u0026thinsp;confidence intervals \u003cem\u003estarted aggression\u003c/em\u003e 0.07\u0026thinsp;+\u0026thinsp;0.02\u0026ndash;0.12, \u003cem\u003ereacted aggressively\u003c/em\u003e 0.38\u0026thinsp;+\u0026thinsp;0.26\u0026ndash;0.50, \u003cem\u003ereacted peacefully\u003c/em\u003e, 0.55\u0026thinsp;+\u0026thinsp;0.42\u0026ndash;0.69; contrasts estimate \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e: -0.31; df\u0026thinsp;=\u0026thinsp;22, t-ratio = -5.04, p-value\u0026thinsp;=\u0026thinsp;0.001; estimate \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e: -0.49; df\u0026thinsp;=\u0026thinsp;22, t-ratio = -6.37, p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001), but not for \u003cem\u003ereacted aggressively vs reacted peacefully\u003c/em\u003e (estimate: -0.18; df\u0026thinsp;=\u0026thinsp;22, t-ratio = -1.521, p-value\u0026thinsp;=\u0026thinsp;0.303). Additionally, we conducted log-likelihood ratio tests to evaluate the significance of each focal variables by comparing a full model and a model that lacked the focal variable. Each focal variable contributed significantly to the respective model (Tab. S18; all models converged).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eGene-enrichment analyses and gene-function predictions of the identified SNPs and DEGs\u003c/h3\u003e\n\u003cp\u003eWe used the identified SNP and DEGs, namely SNP3, \u003cem\u003eyellow-d2, BicC, Pif1, Exo84, KaiR1d, RD, RFC3\u003c/em\u003e, and \u003cem\u003ePlexA\u003c/em\u003e, in a gene-enrichment analysis and found that the identified are linked to depression restoration, synaptic and neurological functions, aggression, as well as plasticity. We conducted the gene-enrichment analysis in g:Profiler (using only annotated genes and FDR adjusted with p-values\u0026thinsp;\u0026lt;\u0026thinsp;0.05) to test whether they were enriched for biological processes, molecular function, and/or cellular components. We used the fruitfly \u003cem\u003eDrosophila melanogaster\u003c/em\u003e as a background gene set and conducted an unordered query to analyse whether certain biological pathways or gene sets were overrepresented. To increase the sample size of the gene-enrichment search, we used all genes regardless of whether they were up- or down-regulated or from different comparisons (for details, see the section \u0026ldquo;\u003cem\u003eGene-enrichment analyses with known genes\u003c/em\u003e\u0026rdquo; in the Supplementary Results). In total, six molecular functions, 12 biological processes, and eight cellular components were enriched (Tab. S19). The molecular functions can be broadly summarised into signal transduction and binding and enzymatic and catalytic functions. The biological processes can be summarised into neural signalling, ion transport dynamics, gene expression regulation, and DNA replication and elongation. The cellular components can be summarised into replication functions, vesicle transport, neuronal functions, and ion channel functions. Overall, we combined them into two categories, namely neurological and synaptic functions as well as DNA replication, repair, and genome stability functions (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE). The former included the genes \u003cem\u003eBicC, Exo84, gd\u003c/em\u003e (i.e., SNP1 in the gene \u003cem\u003egd\u003c/em\u003e), \u003cem\u003eKaiR1d, PlexA, Rdl\u003c/em\u003e, and \u003cem\u003eyellow-d2\u003c/em\u003e, while the latter included \u003cem\u003ePif1 and RFC3\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eUsing a gene-prediction analysis, we also identified neurological and synaptic functions across the SNPs and DEGs. In detail, we used GeneMania\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e (FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.05; including gene \u003cem\u003ePlexA\u003c/em\u003e because excluding \u003cem\u003ePlexA\u003c/em\u003e only yielded non-significant results), which predicts gene function and searches for similar functional and related genes based on the initial gene list to find gene pathways or interactions. We included the same SNPs and DEGs in two queries, once with and once without gene \u003cem\u003egd\u003c/em\u003e as it contains a SNP. In the query with \u003cem\u003egd\u003c/em\u003e, we detected five biological processes and two molecular functions (Tab. S20). The biological processes can be summarised as regulation during cell division and the molecular functions as membrane transport functions. In the query without \u003cem\u003egd\u003c/em\u003e, we detected six biological processes and four molecular functions (Tab. S20). The biological processes can be summarised as neuronal signalling and regulation during cell division and the molecular functions as membrane transport functions, possibly linked to synaptic activity and signalling.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAlmost all animals display aggressive behaviour, but our understanding of the underlying mechanisms that promote the start of aggression is limited. Here, we integrated, for the first time, chemical, microbiome, genomic, transcriptomic, and environmental analyses and assessed whether these traits promote the start of aggression and reactions to it in the ant \u003cem\u003eTetramorium alpestre\u003c/em\u003e. We tested workers that displayed either of three behavioural states, namely \u003cem\u003estarted aggression, reacted aggressively\u003c/em\u003e, and \u003cem\u003ereacted peacefully\u003c/em\u003e, identified in the aggression assays. Using the microbiome data, we discovered nine OTUs across four bacterial genera, \u003cem\u003eBacteroides, Lactobacillus, Prevotella\u003c/em\u003e, and \u003cem\u003ePseudomonas\u003c/em\u003e, that were associated with the behavioural states. We also identified three genes with a SNP each that were associated with the start of aggression (whole-genome data; GEMMA analysis), namely the gene \u003cem\u003emediator of RNA polymerase 2 transcription subunit 26\u003c/em\u003e (SNP1), one unknown gene (SNP2), and the gene \u003cem\u003egastrulation-defective\u003c/em\u003e (SNP3). We also found significantly up-regulated (N\u0026thinsp;=\u0026thinsp;30) and down-regulated genes (N\u0026thinsp;=\u0026thinsp;28; FDR corrected for multiple testing) when comparing the state \u003cem\u003estarted aggression\u003c/em\u003e vs each state \u003cem\u003ereacted aggressively\u003c/em\u003e or \u003cem\u003ereacted peacefully\u003c/em\u003e. Finally, we integrated these SNPs, DEGs, as well as additionally collected colony and environmental variables (e.g., within-colony relatedness, CHCs, site-specific nitrogen and temperature values) in a multinomial logistic regression (multiple data layers jointly). We found that SNP2 and SNP3 (in the gene \u003cem\u003egd\u003c/em\u003e) as well as the DEGs \u003cem\u003eBicC, Exo84, KaiR1d, Pif1, PlexA, Rdl, RFC3\u003c/em\u003e, and \u003cem\u003eyellow-d2\u003c/em\u003e are associated with the behavioural states, while CHC and colony and environmental variables were not.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eCHC compounds represent population structure but are not associated with the behavioural states\u003c/h2\u003e \u003cp\u003eUsing the CHC compounds, genetic differentiation, and relatedness values, we corroborate the colony and population structure expected at the onset of this study. The PCA of the CHCs used 63 compounds, which is slightly more than found in a recent study on this species (N\u003csub\u003eCHCs\u003c/sub\u003e = 50)\u003csup\u003e9\u003c/sup\u003e. We expected the single-queened colonies to be separated from each other and the multiple-queened colonies to be mixed among colonies due to lower and higher relatedness, respectively. The combined analyses of the PCA of the CHC compounds, the CHC hierarchical cluster analysis, the genetic differentiation (pairwise F\u003csub\u003eST\u003c/sub\u003e values of the genomic data), and relatedness values corroborated this expectation. In detail, workers of the colonies of population SQ-N are related to each other and likely have only one queen. This also explains the narrow distribution of the CHCs, as CHC bouquets are genetically determined and environmentally tuned\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Workers of the colonies of population SQ-A are not or little related and also likely have one queen. As a result, the distribution of the CHC bouquet in the PCA is wider. Workers from colonies of population MQ-N have a higher relatedness scattered across colonies. This indicated that they likely have multiple, possibly unrelated queens. Workers from different queens within the same colony have very different CHCs leading to a broader variety in PCA of the CHCs.\u003c/p\u003e \u003cp\u003eWhile the microbiome, SNPs, and DEGs affected the behavioural states (discussed in the next three sections), the CHC bouquet did not. This is interesting because CHC differences can cause aggressive behaviour in ants\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e but not necessarily in every ant species\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. Here and in a previous study using this species\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, we did not find any association between CHC differences and aggression. This could be due to our study design: due to the small size of these animals, we were only able to use workers either for CHC extractions or for aggression assays (and subsequent genomic and transcriptomic analyses). By coincidence, the CHC bouquets of all fighting workers may have been more dissimilar from each other than the ones of the workers used for CHC analyses. Apart from this appearing unlikely as a pattern throughout, also a previous study on this ant\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e and other ant species\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e used different ants for CHC extractions and for aggression tests. They found a correlation\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e or not\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e suggesting that such a correlation could have been found if CHCs were important drivers of aggression in this ant. In contrast, this is the second study using this species indicating that CHC differences are not important for aggression in this species. Even though CHCs do not seem to elicit aggression in this species, we speculate that it still uses CHCs for nestmate recognition, but workers simply remain peaceful towards non-nestmates with different CHC bouquets.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eGut bacteria are associated with the behavioural states\u003c/h2\u003e \u003cp\u003eWe found nine gut bacteria across four genera, \u003cem\u003eBacteroides, Lactobacillus\u003c/em\u003e, \u003cem\u003ePrevotella\u003c/em\u003e, and \u003cem\u003ePseudomonas\u003c/em\u003e, that were associated with the behavioural states. The genus \u003cem\u003eBacteroides\u003c/em\u003e is linked with the behavioural states. We found a higher frequency of one \u003cem\u003eBacteroides\u003c/em\u003e sp. (OTU 1598) in ants that were aggressive (\u003cem\u003estarted aggression\u003c/em\u003e and \u003cem\u003ereacted aggressively\u003c/em\u003e) than in ants that \u003cem\u003ereacted peacefully\u003c/em\u003e as well as a higher frequency of one \u003cem\u003eBacteroides\u003c/em\u003e sp. (OTU 22324) in ants that \u003cem\u003ereacted aggressively\u003c/em\u003e and \u003cem\u003ereacted peacefully\u003c/em\u003e than in ants that \u003cem\u003estarted aggression\u003c/em\u003e. Tillisch et al (2017)\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e found that women with a higher \u003cem\u003eBacteroides\u003c/em\u003e abundance had more dense white brain matter tracts, indicating altered sensory processing. This may indicate that \u003cem\u003eBacteroides\u003c/em\u003e bacteria affect how ants perceive other ants and react accordingly. Additionally, Lin et al. 2017\u003csup\u003e38\u003c/sup\u003e and Strandwitz et al. 2019\u003csup\u003e39\u003c/sup\u003e found that a reduced abundance of \u003cem\u003eBacteroides\u003c/em\u003e bacteria in the gut is possibly linked with depression in humans. In turn, this may indicate that ants with higher \u003cem\u003eBacteroides\u003c/em\u003e counts were positively stimulated and thus more proactive and reactive.\u003c/p\u003e \u003cp\u003eWe found a lower abundance of \u003cem\u003eLactobacillus mucosae\u003c/em\u003e (OTU 813) in ants that \u003cem\u003ereacted peacefully\u003c/em\u003e, but a higher abundance of \u003cem\u003eLactobacillus\u003c/em\u003e sp. (OTU 21141), compared with ants that \u003cem\u003estarted aggression\u003c/em\u003e or \u003cem\u003ereacted aggressively\u003c/em\u003e. Recent studies also found links between \u003cem\u003eLactobacillus\u003c/em\u003e and nestmate recognition in honey bees\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e as well as behavioural changes in dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eDrosophila\u003c/em\u003e flies\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, and ants\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. However, the results are partially contradictory: for example, one study found a higher \u003cem\u003eLactobacillus\u003c/em\u003e abundance in phobic dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, namely \u003cem\u003eLactobacillus plantarum\u003c/em\u003e, which has known psychobiotic properties\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. However, another study on dogs and \u003cem\u003eD. melanogaster\u003c/em\u003e males found that the genus \u003cem\u003eLactobacillus\u003c/em\u003e was more frequently present in aggressive dogs and \u003cem\u003eD. melanogaster\u003c/em\u003e males\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. While the underlying mechanisms remain unclear, a link between \u003cem\u003eLactobacillus\u003c/em\u003e and behavioural changes seems to appear.\u003c/p\u003e \u003cp\u003eAlso the bacteria \u003cem\u003ePrevotella\u003c/em\u003e can affect behavioural states. We detected a higher frequency of three \u003cem\u003ePrevotella\u003c/em\u003e spp. (OTU 377, 1887, 20448) in ants that \u003cem\u003estarted aggression\u003c/em\u003e than ants that \u003cem\u003ereacted aggressively\u003c/em\u003e or \u003cem\u003ereacted peacefully\u003c/em\u003e as well as in \u003cem\u003ereacted peacefully\u003c/em\u003e than in \u003cem\u003ereacted aggressively\u003c/em\u003e. A study in humans found that women with a higher abundance of \u003cem\u003ePrevotella\u003c/em\u003e gut bacteria displayed higher negative affect when shown images with a negative emotional content, which was associated with both functional and structural differences in the hippocampus\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. Speculatively, these associations of \u003cem\u003ePrevotella\u003c/em\u003e with aggression may indicate an evolutionarily conserved pathway of these gut bacteria with negative stimuli, in humans and ants, and possibly other animals.\u003c/p\u003e \u003cp\u003eLastly, also the genus \u003cem\u003ePseudomonas\u003c/em\u003e is connected with behavioural states. We found two \u003cem\u003ePseudomonas\u003c/em\u003e spp. (OTUs 366, 2442) that had a higher frequency in ants that \u003cem\u003estarted aggression\u003c/em\u003e. To our best knowledge, no study has so far linked \u003cem\u003ePseudomonas\u003c/em\u003e to behaviours such as aggression. However, other studies have linked \u003cem\u003ePseudomonas\u003c/em\u003e strains with, for example, a potential insecticide resistance\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e or metabolising insecticides\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. Notably, insecticide often affect neuronal or synaptic functions. It thus may be that \u003cem\u003ePseudomonas\u003c/em\u003e is connected to behavioural changes via such a metabolic pathway.\u003c/p\u003e \u003cp\u003eWhile additional bacteria such as \u003cem\u003eAcetobacter, Enterococcus, Fusobacterium\u003c/em\u003e, or \u003cem\u003eMegamonas\u003c/em\u003e have been found to affect behaviour in humans\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e, dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eDrosophila\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, and ants\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, we did not find any association with these bacteria here.\u003c/p\u003e \u003cp\u003eOur gut microbiome results are in line with other research indicating that there is accumulating evidence of microbiome effects on the recognition and behaviour of animals such as humans\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e, dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eDrosophila\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, cockroaches, locusts, and termites\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e (and references therein), as well as ants\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. For example, studies also suggests that the microbiome affects the behaviour via the gut-brain axis\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e: the authors suggest the gut microbiome \u0026lsquo;communicates\u0026rsquo; with the central nervous system in various parallel ways such as the vagus nerve, signalling mechanisms, and the production of neuroactive chemicals (e.g., serotonin, gamma-amino butyric acid \u0026lsquo;GABA\u0026rsquo;)\u003csup\u003e12,44\u003c/sup\u003e. In turn, the central nervous system also communicates with the gut microbiome\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e, creating a feedback loop. While an overall association seems to emerge, the exact effects of specific bacteria on their hosts remain to be determined. Thus, further studies are needed to assess such bacteria-host interactions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eSNPs and DEGs are also associated with behavioural changes\u003c/h2\u003e \u003cp\u003eWe found one SNP each in two genes as well as eight DEGs associated with behavioural, neurological, and synaptic functions that may explain the observed behavioural states. SNP2 is at a site of an unknown gene and will not be discussed further, while SNP3 is at a site in the gene \u003cem\u003egd (gastrulation-defective).\u003c/em\u003e For this SNP in the gene \u003cem\u003egd\u003c/em\u003e, more ants that \u003cem\u003estarted aggression\u003c/em\u003e were heterozygous at the SNP site. Although we found no direct link between the gene \u003cem\u003egd\u003c/em\u003e and aggression, we speculate that there may be an indirect link. The activation of \u003cem\u003egd\u003c/em\u003e leads to the activation of the \u003cem\u003eToll\u003c/em\u003e pathway. The \u003cem\u003eToll\u003c/em\u003e pathway is conserved and is involved in the development of the dorsal-ventral embryonic axis\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e, but it also promotes the expression of the \u003cem\u003etranscription factor nuclear factor kappa B\u003c/em\u003e, which has functional roles in neuroprotection and synaptic plasticity\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. Gene \u003cem\u003egd\u003c/em\u003e may thus be associated with brain synaptic activity and thus possibly with the start of aggression.\u003c/p\u003e \u003cp\u003eWe further found six down-regulated genes associated with depression restoration, synaptic and neurological functions, aggression, as well as plasticity. These genes, \u003cem\u003eBicC, Exo84, KaiR1d, Rdl\u003c/em\u003e, \u003cem\u003eyellow-d2\u003c/em\u003e, and \u003cem\u003ePlexA\u003c/em\u003e, are down-regulated in workers that \u003cem\u003estarted aggression\u003c/em\u003e (\u003cem\u003ePlexA\u003c/em\u003e was kept here as it was driving significant results in the GeneMania analysis). In more detail, \u003cem\u003eBicC\u003c/em\u003e is associated with depression restoration\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eExo84\u003c/em\u003e with neurite differentiation\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eKaiR1d\u003c/em\u003e with baseline synaptic transmission\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eRdl\u003c/em\u003e with neurotransmission and olfactory learning\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eyellow-d2\u003c/em\u003e with dopamine receptor signalling\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e, and \u003cem\u003ePlexA\u003c/em\u003e as a receptor fir semaphorsin\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e (for more details on each gene, see the section \u0026ldquo;\u003cem\u003eSix down-regulated genes linked to synaptic functions\u003c/em\u003e\u0026rdquo; in the Supplementary Discussion). The results indicate these SNPs and down-regulated genes could affect behavioural states in this ant species. Specifically, their associations with neurological and synaptic functions could indicate a potential direct link between them and the start of aggression and reactions to it. Further, they may affect the behaviour in this ant in a concerted way.\u003c/p\u003e \u003cp\u003eWe also found two up-regulated genes associated with DNA repair and replication. These genes, \u003cem\u003ePif1\u003c/em\u003e and \u003cem\u003eRFC3\u003c/em\u003e, are not directly but indirectly associated with behaviour. For example, Gidron et al. (2006)\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e found that under specific conditions and repeated exposure to stressful situations, reactive oxygen species increased and can yield to DNA damage in animals. It may be that ants that \u003cem\u003estarted aggression\u003c/em\u003e were more sensible to stressful situations (e.g., sampling and laboratory maintenance) leading to an increase in oxidative stress scavenging mechanisms, which can reduce indices of oxidative stress\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. In turn, this may explain the up-regulation of such genes. However, further studies need to shed light on such potential associations.\u003c/p\u003e \u003cp\u003eWe acknowledge the possibility that the identified bacteria, SNPs, and/or DEGs are false positives. However, we argue that this is unlikely because these bacteria, mutations, and DEGs were identified by using independent datasets, applying corrections for multiple comparisons to minimise retrieving false positives, and combining the data sets in a joint analysis (multinomial logistic regression). We further checked the robustness of the results by dropping focal variables using log-likelihood ratio tests. In contrast, we argue that the results of these three independent data sets represent three distinct lines of evidence suggesting that similar underlying mechanisms can contribute to the start of aggression or the reaction to it, for example via hormone and synaptic signalling. The observed behavioural states could thus be affected by these factors in a concerted way. At the same time, we stress that our results are correlative but not causal. Additionally, also other factors, such as epigenetic changes not tested here, may contribute to the start of aggression and reactions to it. Future studies should thus test whether the identified gut bacteria and genes are functionally relevant. This could be tested by conducting aggression tests with ants that have been fed with these bacteria or with ants in which these known genes are knocked out, impaired, or over-expressed. It would further be interesting to ascertain whether the same gut bacteria, genes, or gene homologs are important for the aggressive or peaceful behaviour in other (social) insects as well. Additionally, possible effects of epigenetic changes (e.g., DNA methylation, histone modifications) should be tested.\u003c/p\u003e \u003cp\u003eAggression and its possible positive effects can be adaptive\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. This is especially true if starting aggression leads to increased fitness\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. The start of aggression has been loosely associated with individual effects, such as changes in the gut microbiome\u003csup\u003e11\u0026ndash;13,20\u0026minus;22\u003c/sup\u003e, SNPs in genes\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e or DEGs\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. In this study, we integrated \u0026ndash; for the first time to our best knowledge \u0026ndash; gut microbiome data with chemical, genomics, transcriptomics, environmental, and behavioural assays, using the ant \u003cem\u003eT. alpestre\u003c/em\u003e. We identified nine gut bacteria, two mutations, and eight DEGs that are associated with the three behavioural states \u003cem\u003estarted aggression\u003c/em\u003e, \u003cem\u003ereacted aggressively\u003c/em\u003e, and \u003cem\u003ereacted peacefully.\u003c/em\u003e In contrast, chemical and environmental factors were not associated with the behavioural states. The nine gut bacteria found are known to influence aggression and other behaviours in several organisms, for example, via hormone signalling\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. The identified SNPs and DEGs were, among others, associated with neurological and synaptic functions. Based on these results, we speculate that these three traits can contribute the start of aggression, possibly in a synergistic mechanism.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e "},{"header":"Materials \u0026 Methods","content":"\u003ch2\u003eFieldwork and colony maintenance\u003c/h2\u003e \u003cp\u003eBetween July 18th and 25th 2018, 500 workers were sampled from three colonies each in three populations (N\u003csub\u003ecolonies\u003c/sub\u003e=9, Tab. S1). The populations were selected based on preliminary behavioural data (not shown): One population was located in South Tyrol, Italy, and comprised single-queened and aggressive colonies (\u0026ldquo;SQ-A\u0026rdquo;), one in Tyrol, Austria, comprising single-queened and non-aggressive colonies (\u0026ldquo;SQ-N\u0026rdquo;), and one in Carinthia, Austria, comprising multiple-queened and non-aggressive colonies (\u0026ldquo;MQ-N\u0026rdquo;; potentially supercolonial population). Of these 500 workers, 200 were immediately snap-frozen in the field using a dry shipper (CY50915D, Thermo-Fisher Scientific Inc., MA, USA) for CHC and molecular analyses. The remaining workers were transported alive to a laboratory at the University of Innsbruck and transferred to polypropylene boxes (10.5 \u0026times; 10.5 cm; as of now \u0026ldquo;colony\u0026rdquo;) awaiting behavioural assays. These workers presumably included all polyethism stages\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. To prevent workers from escaping, the walls of the boxes were Fluon-coated (GP1, De Monchy International BV, Rotterdam, Netherlands). Each box was equipped with soft tissue as a hiding place, two conical Eppendorf tubes filled with water or with diluted honey water and each plunged with cotton as a drinking aid, and a frozen \u003cem\u003eDrosophila hydei\u003c/em\u003e fruit fly. The water, honey water, and fruit fly were refilled twice per week and present at all time \u003cem\u003ead libitum\u003c/em\u003e. The boxes were placed in a climate cabinet (MIR-254, Panasonic, Etten Leur, Netherlands) with constant dark conditions, a humidity of 50\u0026ndash;70%, and at constant 18\u0026deg;C. Constant 18\u0026deg;C was selected to acclimatise workers that originated from slightly different elevations to a similar temperature. Before the various assays, the colonies were kept in the climate cabinets for two weeks. Pairwise geographic distances between populations SQ-N:MQ-N (Kuehtai \u0026ndash; Mussen; Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA) were 140 km, between SQ-A:MQ-N (Penser Joch \u0026ndash; Mussen) 125 km, and between SQ-A:SQ-N (Penser Joch - Kuehtai) 40 km calculated using an online tool (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ibm.franken.de/gps03.html\u003c/span\u003e\u003cspan address=\"https://www.ibm.franken.de/gps03.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eRecognition assays\u003c/h2\u003e \u003cp\u003eBetween August 13th and 20th 2018, we conducted recognition assays to test if workers recognise and prefer their own colony odour over an alien colony or a control odour following Steiner et al. (2007)\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. For these assays, we extracted cuticular hydrocarbons (CHCs) from workers of each colony separately using three different extraction solvents sequentially, starting in 100 \u0026micro;l hexane, then 100 \u0026micro;l ethyl acetate, and lastly 100 \u0026micro;l 96% ethanol (all three Merck, MA, USA)\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. For the extraction, we transferred the workers into 1.1-ml conic glass vials (CZT, Kriftel, Germany). Workers remained in each extraction solvent for 90 s before being transferred to the next. The three solvents should extract as many CHCs as possible. We then transferred the workers to 96% ethanol, mixed the three solvents, and stored them at -20\u0026deg;C until further use. To account for body size differences, we used 15 workers for each colony from populations SQ-A and SQ-N and 22 workers for each colony from population MQ-N, which had smaller workers. In total, we generated nine solvents (one for each colony) to test if workers prefer their own, alien, or control (a mixture of the three extraction solvents without CHCs) odour. To do this, we created small filter paper disks (2 cm diameter; 75 g/qm, Altmann Analytik, M\u0026uuml;nchen, DE) with three 120-degree sectors (own, alien, and control sector). Onto the sector \u0026ldquo;own\u0026rdquo;, we applied 1 \u0026micro;l of the extract of the colony to be tested, onto the sector \u0026ldquo;alien\u0026rdquo;, 1 \u0026micro;l of the extract of a different colony from the same population, and onto the sector \u0026ldquo;control\u0026rdquo;, 1 \u0026micro;l of the mixture of extraction solvents without CHCs. We transferred the solvents onto the paper disks using a 20-\u0026micro;l syringe (Hamilton, NV, USA). After transferring the solvents onto the papers, the solvents were left to evaporate for three minutes before we transferred the paper disks to the bottom of small glass vials (2 cm \u0026Oslash;). The bottom of each glass vial was covered with 1 \u0026micro;l paraffin oil as a keeper substance\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. The walls of the glass vials were Fluon-coated to prevent workers from escaping. After evaporation, we transferred individual workers to the glass vials, which were covered with tin cans to simulate dark conditions. Workers were allowed to acclimatise for 15 minutes, after which we lifted the can for five seconds and noted the sector of the paper disk on which the worker was sitting. After each observation, we turned the vial 120 degrees and lightly tapped it thus forcing the worker to move. In each assay, we tested all colonies in a randomised order. Both conductors and evaluators were blind to the origin of colonies. We conducted the assays in an air-conditioned room with constant 18\u0026deg;C resembling the temperature in the climate cabinet. In one run, 36 workers were tested (four from each colony), and 15 runs were conducted. This procedure was replicated three times over three days resulting in 1,620 observations, which were analysed together using a multinomial Goodness-of-Fit test to test if workers recognise and prefer their own colony odour, an alien colony odour, or a control odour.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eExtraction and analysis of cuticular hydrocarbons (CHCs)\u003c/h2\u003e \u003cp\u003eWe extracted CHCs from five workers per colony following Krapf et al. (2023)\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. For the extraction, we transferred five workers, which had been immediately frozen after sampling, to 1.1-ml conic glass vials (CZT, Kriftel, Germany) and immediately added 100 \u0026micro;l n-pentane (Merck, MA, USA) using a 100-\u0026micro;l syringe (Hamilton, NV, USA). The CHCs were extracted for three minutes while the glass vials were being shaken at 450 rpm. After the extraction, we removed the workers from the vials and transferred them to Eppendorf tubes filled with 96% ethanol. The vials containing the CHC extracts were sealed until their analysis. For the analysis, a 7890 B Series gas chromatograph (Agilent, Waldbronn, Germany) equipped with a flame ionization detector (FID), a nonpolar DB-5 column (30m\u0026times;0.25mminner diameter, J\u0026amp;W, Waldbronn, Germany), and hydrogen (2ml/min constant flow) as carrier gas was used. One \u0026micro;l of a sample was injected splitless at an initial oven temperature of 50\u0026deg;C. After 1 min, the splitting valve was opened and the temperature gradually increased by 10\u0026deg;C/min until it reached a final temperature of 310\u0026deg;C, which was kept constant for 50 min. To ensure the consistency of the analyses, GC runs were performed regularly with a synthetic alkane standard mixture. Structure elucidation of individual compounds was performed with an HP (Hewlett Packard) 6890 Series gas chromatograph connected to a mass selective detector (GC\u0026ndash;MS; Quadrupole 5972, Agilent, Waldbronn, Germany). Helium was used as carrier gas (1.5 ml/min constant flow). The temperature program was the same as described above. The absolute and relative amounts of these compounds were determined by using Agilent ChemStation software (Agilent, Waldbronn, Germany). Structure assignments were carried out by comparison of mass spectra and retention times of natural products with corresponding data from synthetic reference samples using the NIST database and a database of the Institute of Evolutionary Ecology and Conservation Genomics at the University of Ulm, following previous work\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e,\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e. Peak identities across different runs were confirmed by GC-MS.\u003c/p\u003e \u003cp\u003eTo estimate relative proportions for further downstream analyses, we only used CHCs that were found in all samples. Further, we divided the absolute amounts of individual compounds by the sum of the absolute amounts of all compounds and multiplied by 100. With these CHC compounds, we created a PCA using the function \u0026ldquo;prcomp\u0026rdquo; (\u0026ldquo;ggfortify\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e) to check if colonies and/or populations form distinct clusters. Further, we conducted a hierarchical cluster analysis with the CHC data using the function \u0026ldquo;agnes\u0026rdquo; and Ward\u0026rsquo;s minimum variance method (\u0026ldquo;cluster\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e). We used the values of the first PCA in the multinomial regression.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eOne-on-one aggression tests\u003c/h2\u003e \u003cp\u003eWe conducted one-on-one aggression tests within each population on July 23rd 2018 to determine if the colonies displayed the expected behaviour (i.e., aggressive and non-aggressive). We conducted standardised aggression tests\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e in an air-conditioned room with constant 18\u0026deg;C. For each aggression test (i.e., one encounter), we randomly selected na\u0026iuml;ve single workers from different colonies from the same population and transferred to a small glass vial (1.4 cm inner diameter) with Fluon-coated walls preventing workers from escaping. Only workers actively running outside in the arena were selected, which likely were foragers\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. We added a worker from one colony first and then the second worker. In the next encounter, we changed the order of workers introduced to prevent any effect of adding workers to the vial. We conducted five encounters for each colony combination to account for behavioural variation\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e. Each encounter lasted 180 s and was filmed using high-definition cameras (Handycam HDR-XR 155; HDRPJ810E, Sony, Tokyo, Japan). As workers might have been agitated after being transferred to the vials, the first 10 seconds of each encounter were regarded as an acclimatisation time and were thus excluded from further analyses\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. The assay conductors were not blind to the colony\u0026rsquo;s origin.\u003c/p\u003e \u003cp\u003eWe further conducted one-on-one aggression tests between populations between July 25th and 28th 2018 following the approach described above. Within 10 minutes after the end of the aggression test, we separated the workers if fighting, transferred them individually to 1.5 ml tubes, and snap-froze them using liquid nitrogen. This procedure ensured that no early genes were expressed, which can start after 15 minutes\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e. At this point, the colony origin of the workers was unknown, but we later identified the colony identity using microsatellite analysis (see section below \u0026ldquo;\u003cem\u003eMicrosatellite genotyping for reference workers\u003c/em\u003e\u0026rdquo;). Additionally, we conducted within-colony aggression tests on July 27th 2018 to test if workers behaved peacefully, which was our expectation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eOne-on-one aggression analysis and worker selection for sequencing\u003c/h2\u003e \u003cp\u003eFor an initial screening of the aggression test, we noted the behaviour of both workers every ten seconds as \u0026ldquo;aggressive\u0026rdquo;, \u0026ldquo;neutral\u0026rdquo;, or \u0026ldquo;peaceful\u0026rdquo; while conducting the aggression tests. Based on this initial screening, we selected 112 videos for a detailed analysis. From these videos, we examined the behaviour of each worker in slow-motion, and classified the behaviour of both workers second by second using the following scoring scale\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e: (\u0026minus;\u0026thinsp;4) trophallaxis, (\u0026minus;\u0026thinsp;3), allogrooming, (\u0026minus;\u0026thinsp;2) antennation, (\u0026minus;\u0026thinsp;1) being next to each other without contact, (0) ignoring, (1) avoiding, (2) mandible threatening, (3) fighting without gaster flexion, (4) fighting with gaster flexion, and (5) killing. The observer of the videos was blind to the origin of the colonies. Moreover, an aggression index \u003cem\u003eAI\u003c/em\u003e\u003csup\u003e61\u003c/sup\u003e was calculated as detailed in Krapf et al. (2023)\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. For \u003cem\u003eAI\u003c/em\u003e, the duration of each behaviour was summed up and multiplied by its respective behaviour score (-4 to +\u0026thinsp;5). This value was divided by the total number of seconds with tactile interactions recorded. Lastly, the arithmetic mean of the five replicates was calculated.\u003c/p\u003e \u003cp\u003eUsing this detailed analysis, we defined three behavioural states: workers that \u0026lsquo;\u003cem\u003estarted aggression\u0026rsquo;\u003c/em\u003e, workers that \u003cem\u003e\u0026rsquo;reacted aggressively\u0026rsquo;\u003c/em\u003e, or workers that \u003cem\u003e\u0026rsquo;reacted peacefully\u003c/em\u003e\u0026rsquo;. For the aggressive states (\u003cem\u003estarted aggression\u003c/em\u003e; \u003cem\u003ereacted aggressively\u003c/em\u003e), we used workers that displayed a scoring value of 3 and higher to ensure that high aggression levels were used. Based on these three behavioural states, we selected 109 workers for whole-genome sequencing (\u003cem\u003estarted aggression\u003c/em\u003e\u0026thinsp;=\u0026thinsp;43 workers, \u003cem\u003ereacted aggressively\u003c/em\u003e\u0026thinsp;=\u0026thinsp;35 workers, \u003cem\u003ereacted peacefully\u003c/em\u003e\u0026thinsp;=\u0026thinsp;31 workers; see Tab. S1 for population and colony details) and, of those, we selected 85 workers for transcriptomic analyses (\u003cem\u003estarted aggression\u003c/em\u003e\u0026thinsp;=\u0026thinsp;31 workers, \u003cem\u003ereacted aggressively\u003c/em\u003e\u0026thinsp;=\u0026thinsp;29, \u003cem\u003ereacted peacefully\u003c/em\u003e\u0026thinsp;=\u0026thinsp;25). The additional 24 workers selected for whole-genome sequencing originated from the non-aggressive and polygynous population MQ-NS. They were used to account for multiple queens and reliably calculate within-colony relatedness and estimate queen numbers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eDNA- and RNA-extractions and whole-genome and whole-transcriptome sequencing\u003c/h2\u003e \u003cp\u003eFor whole-genome sequencing of samples, we cut off the mesosoma and abdomen from the head of each ant using sterile scalpels (Fig. P1). We used the mesosoma and abdomen for DNA extractions (N\u003csub\u003esamples\u003c/sub\u003e=109) and the head for RNA extractions (N\u003csub\u003esamples\u003c/sub\u003e=85). We extracted DNA using the QiAmp Micro DNA Kit (Qiagen, Hilden, Germany). For this, we transferred the mesosoma and abdomen of each worker to a sterile tube and submerged it into liquid nitrogen. We then ground the mesosoma and abdomen using disposable pestles. The extraction followed the manufacturer\u0026rsquo;s protocol except for the dilution, which was conducted twice, as follows: the first elution was done with 50 \u0026micro;l dH\u003csub\u003e2\u003c/sub\u003e0 for whole-genome sequencing and the second elution with 30 \u0026micro;l dH\u003csub\u003e2\u003c/sub\u003e0 for microsatellite genotyping to determine the colony affiliation (see section below \u0026ldquo;\u003cem\u003eMicrosatellite genotyping to identify colony identity\u003c/em\u003e\u0026rdquo;).\u003c/p\u003e \u003cp\u003eWe extracted RNA from the heads of 85 workers using the Nucleospin RNA Kit (Macherey-Nagel, D\u0026uuml;ren, Germany) following the manufacturer\u0026rsquo;s protocol. For this, we transferred the head of each worker to a sterile tube, submerged the tube into liquid nitrogen, and grinded the head using disposable pestles. The subsequent extraction followed the manufacturer\u0026rsquo;s protocol except for the dilution: RNA was eluted in 40 \u0026micro;l RNAse-free dH20 provided by the manufacturer.\u003c/p\u003e \u003cp\u003eWe conducted all DNA- and RNA-extraction steps under sterile conditions in a laminar flow hood. DNA and RNA extracts were stored at -70\u0026deg;C until being shipped for library preparation and whole-genome and -transcriptome sequencing outsourced to a commercial provider (IGATech, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://igatechnology.com/\u003c/span\u003e\u003cspan address=\"http://igatechnology.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Each worker was sequenced with 125-bp paired-end sequencing for both DNA- and RNA extractions on HiSeq2500.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eMicrosatellite genotyping to identify colony identity\u003c/h2\u003e \u003cp\u003eWe conducted microsatellite genotyping to assess the colony identity of workers used in aggression tests. First, we genotyped 12 reference workers from each colony (i.e., known colony identity) using eight microsatellite loci\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. For this, we extracted DNA using the Sigma GenElute extraction kit following the manufacturer\u0026rsquo;s protocol, except for eluting in 50 \u0026micro;l. PCR for genotyping was done in 5 \u0026micro;L reaction volume with 0.5 \u0026micro;L template DNA, 2 \u0026times; Rotorgene Master Mix (Qiagen, Hilden, Germany), 0.01 \u0026micro;M M13 tailed locus-specific forward primer, 0.1 \u0026micro;M fluorescent-labelled M13 primer, 0.1 \u0026micro;M untailed specific reverse primer, and 1.79 \u0026micro;L dH\u003csub\u003e2\u003c/sub\u003eO on a UnoCycler 1200 (VWR, Radnor, USA). Cycling conditions were 94\u0026deg;C for 5 min followed by 35 cycles at 94\u0026deg;C for 30 s, 60\u0026deg;C for 1 min, 72\u0026deg;C for 45 s, and a final extension at 68\u0026deg;C for 20 min. Fragment analysis was carried out on an ABI3730XL genetic analyser (Applied Biosystems, Foster City, USA) by a commercial provider (Comprehensive Cancer Center DNA Sequencing \u0026amp; Genotyping Facility, University of Chicago, USA). Microsatellites were genotyped using GeneMarker V.3.0.1 (SoftGenetics, State College, PA, USA).\u003c/p\u003e \u003cp\u003eFollowing the same procedure, we genotyped workers from the aggression tests and reliably assigned the colony identity before shipping the samples to the commercial provider IGA for sequencing. Based on the genotypes of known colony identities, we calculated the probability of colony affiliations using the software GeneClass2\u003csup\u003e62\u003c/sup\u003e. GeneClass2 uses multilocus genotypes to select or exclude populations as origins of individuals. To find colony affiliations, we chose the Bayesian method by Rannala \u0026amp; Mountain (1997)\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e as the computation criteria, and the assignment threshold of the scores was 0.05. Further, we calculated within-colony relatedness based on the genotypes following Queller and Goodnight algorithm\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e and additionally, the number of queens following Pamilo (1991)\u003csup\u003e\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eAnalysing whole-genome and whole-transcriptome sequences\u003c/h2\u003e \u003cp\u003eFor both DNA and RNA files, we conducted the same analysis approach. Initial quality control of raw reads was conducted using FastQC (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.bioinformatics.babraham.ac.uk/projects/fastqc/\u003c/span\u003e\u003cspan address=\"https://www.bioinformatics.babraham.ac.uk/projects/fastqc/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and MultiQC (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://seqera.io/multiqc/\u003c/span\u003e\u003cspan address=\"https://seqera.io/multiqc/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). We trimmed adapters, duplicates, and contaminants using a \u0026ldquo;kraken\u0026rdquo; database and \u0026ldquo;bbduk\u0026rdquo; (\u0026ldquo;bbtools\u0026rdquo;, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://sourceforge.net/projects/bbmap/\u003c/span\u003e\u003cspan address=\"https://sourceforge.net/projects/bbmap/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). We merged trimmed paired-end files into single files and mapped single files against the \u003cem\u003eTetramorium alpestre\u003c/em\u003e reference genome\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e using \u0026ldquo;bbmap\u0026rdquo; (bbtools) by applying quality trimming on both sides. For mapping, we indexed the files and quality-trimmed them using \u0026ldquo;bbmap\u0026rdquo; (minid\u0026thinsp;=\u0026thinsp;0.9, k\u0026thinsp;=\u0026thinsp;13). We called single nucleotide polymorphisms (SNPs) using the \u0026ldquo;callvariants\u0026rdquo; function from \u0026ldquo;bbtools\u0026rdquo; using the default settings except for ploidy\u0026thinsp;=\u0026thinsp;2. For variant calling, we first called variants in an initial VCF file. Second, we calculated the true equality and then recalibrated them using the initial VCF file. Third, we created an unfiltered VCF file. In this unfiltered VCF file, we identified 1,249,705 and 312,297 SNPs for whole-genome sequences and whole-transcriptome sequences, respectively. We further filtered this unfiltered VCF file using a minimum coverage of 128, a minimum number of sequences of 4 with the alternative allele, a minimum mapping quality of 50, and including linkage-disequilibrium (LD) pruning. After filtering, 184,145 and 69,191 SNPs were kept in the final whole-genome and whole-transcriptome VCF file, respectively.\u003c/p\u003e \u003cp\u003eUsing the VCF file of the whole-genome data, we calculated the heterozygosity and Weir and Cockerham's F\u003csub\u003eST\u003c/sub\u003e and created an LD-pruned PCA using VCFtools\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e. We further calculated the within-colony relatedness using the \u0026ldquo;relatedness\u0026rdquo; function in VCFtools\u003csup\u003e\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u003c/sup\u003e using the method of Manichaikul et al. (2010)\u003csup\u003e\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e\u003c/sup\u003e. We then comapred the within-colony relatedness from whole-genome data with the within-colony relatedness from microsatellite genotyping to assess concordance of values (Tab. S3).\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eGenome-wide mixed-model association (GEMMA) analysis using whole-genome sequences\u003c/h2\u003e \u003cp\u003eWe conducted a GEMMA\u003csup\u003e\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e analysis using whole-genome sequence data to determine if the behavioural states were associated with SNPs in the VCF. Before the analysis, we excluded duplications in the VCF to reduce the bias of emphasising duplications. We used this VCF file without duplications to create a bimbam file using a custom-made Python script. After calculating the bimbam file, we calculated a centred relatedness matrix using \u0026ldquo;gemma\u0026rdquo;, which was used in the subsequent GEMMA analysis. In the GEMMA analysis, a phenotype list detailing the behavioural states of workers, a bam list, and an LD-covariance file were used. GEMMA results were visualised using Manhattan plots created in R using the function \u0026ldquo;Manhattan\u0026rdquo; (\u0026ldquo;qqman\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e). We inspected genomic SNPs above the suggestive line by using them a PCA created with the function \u0026ldquo;prcomp\u0026rdquo; (\u0026ldquo;ggfortify\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e) to check whether alleles clustered together. For this PCA, we dummy-coded individuals that were homozygous for the reference allele of the respective genes as 0/0 and individuals that were heterozygous for the reference alleles as 0/1. We did not find any individual that was homozygous for the alternative allele. Further, we conducted a Pearson\u0026rsquo;s Chi-squared test for count data with simulated p-value and 2000 Monte Carlo replicates to calculate the p-values. The count data represented the number of counts of all individuals for being homozygous or heterozygous for the reference allele for the three behavioural states. The idea was to check if individuals that were homozygous or heterozygous for the reference allele were more or less frequently observed in one of the three behavioural states.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eDifferential gene expression and gene-enrichment analysis\u003c/h2\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eDifferential gene expression\u003c/h2\u003e \u003cp\u003eThe expression counts of each individual stemming from a newly created annotation (for details, see the section \u0026ldquo;\u003cem\u003eTetramorium alpestre annotation\u003c/em\u003e\u0026rdquo; in the Supplementary Materials and Methods) were merged using a customised R script. Using this merged data set, we analysed the expression counts of all individuals (\u0026ldquo;DESeq2\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e\u003c/sup\u003e). For this, we created a DESeqDataSet object to compare the expression of the behavioural states in a pairwise manner. The three behavioural comparisons were: \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e, \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, and \u003cem\u003ereacted aggressively\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e. As a pre-filtering step, we only kept rows that had at least 10 counts in total, thus excluding rows (i.e., genes) with fewer counts than 10. Next, we assessed the data quality of each sample using a pheatmap (\u0026ldquo;pheatmap\u0026ldquo; package\u003csup\u003e\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e). Of the 85 samples, we excluded three due to low quality, yielding 82 samples for subsequent analysis. We conducted a differential gene expression analysis with these 82 samples based on the Negative Binomial (i.e., Gamma-Poisson) distribution and using the default settings. We created volcano plots for each behavioural comparison exported the results as table with a log fold change threshold of zero and using a of 0.05-Benjamini-Hochberg correction (\u0026ldquo;result\u0026rdquo; function; DESeq2 package; \u0026ldquo;false-discovery rate\u0026rdquo;, FDR). We created such result tables for all three comparisons and up-regulated as well as down-regulated genes separately and used them in subsequent analyses. Such tables included gene names, log\u003csub\u003e2\u003c/sub\u003efold values, p-values, and FDR-corrected p-values for multiple testing. We further queried gene names in FlyBase (release FB2025_04) to obtain information on gene function. In subsequent gene-enrichment analyses and multinomial logistic regression analyses, we only used genes with a known (i.e., annotated) gene name.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eGene-enrichment analyses\u003c/h2\u003e \u003cp\u003eFor the three behavioural comparisons \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted aggressively\u003c/em\u003e, \u003cem\u003estarted aggression\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, and \u003cem\u003ereacted aggressively\u003c/em\u003e vs \u003cem\u003ereacted peacefully\u003c/em\u003e, we conducted a gene enrichment analysis in g:Profiler (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://biit.cs.ut.ee/gprofiler/gost\u003c/span\u003e\u003cspan address=\"https://biit.cs.ut.ee/gprofiler/gost\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), a web server for functional gene-enrichment analysis. We only used known (i.e., annotated) genes with an FDR-adjusted p-value lower than 0.05 (Tab. S5). For each behavioural comparison, we conducted a query with an unordered list of genes based on the log\u003csub\u003e2\u003c/sub\u003efold changes. We selected \u003cem\u003eDrosophila melanogaster\u003c/em\u003e as the organism to match the query gene list. Further, we created Venn diagrams using the behavioural-comparison genes for the annotated and all genes in R using the function \u0026ldquo;ggvenn\u0026rdquo; (\u0026ldquo;ggvenn\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u003c/sup\u003e). This analysis allowed checking whether the same genes are up- or down-regulated in several comparisons.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eMicrobiome DNA extraction and marker gene sequencing\u003c/h2\u003e \u003cp\u003eTo test whether the microbiome influenced the three behavioural states, we conducted 16S rRNA gene sequencing. Due to a limited availability of samples, we used 49 workers from two populations: Specifically, we selected four workers each from two colonies of the single-queened and aggressive population SQ-A (colonies SQ-A5 and SQ-A6) and from two of the single-queened and non-aggressive population SQ-N (colonies SQ-N1 and SQ-N6; SQ-N6 with five workers) and from each behavioural state. This resulted in using 16 workers that \u003cem\u003estarted aggression\u003c/em\u003e, 16 that \u003cem\u003ereacted aggressively\u003c/em\u003e, and 17 that \u003cem\u003ereacted peacefully\u003c/em\u003e (Tab. S1). To test if the microbiome changed during laboratory maintenance, we selected 16 additional workers (4 workers each from the colonies SQ-A5, SQ-A6, SQ-N1, and SQ-N6) as control. These workers were immediately frozen after fieldwork and did not experience any laboratory maintenance.\u003c/p\u003e \u003cp\u003eBefore the extractions, we sterilised the surface of whole workers by transferring individual workers for 15 s into Eppendorf tubes filled with 100 \u0026micro;l 5% bleach and then for 15 s into Eppendorf tubes filled with 100 \u0026micro;l phosphate-buffered saline solution (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4)\u003csup\u003e\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eFor the 16 control workers, we extracted DNA using the QIAamp DNA Mini kit (Qiagen, Hilden, Germany) and eluted twice each time with 30 \u0026micro;l of the elution buffer from the kit. For the remaining 49 workers, we dissected the heads from the mesosoma and gaster using a sterile scalpel. For microbiome analyses, we extracted DNA from the mesosoma and gaster using the QIAamp DNA Mini Kit. To determine colony affiliation using microsatellite genotyping (for details, see \u0026ldquo;Microsatellite genotyping of reference workers\u0026rdquo; above), we extracted DNA of the head using the DNEasy Blood and Tissue Kit (Qiagen, Hilden, Germany). We extracted DNA following the manufacturer\u0026rsquo;s protocol except for the elution: DNA was eluted twice each time with 30 \u0026micro;l of the elution buffer from the kit. We conducted all steps before and during the extraction under sterile conditions in a laminar flow hood. High-quality DNA extracts were sent to Novogene (Cambridge, United Kingdom) for marker gene sequencing on a NovaSeq6000 machine (Illumina, San Diego, CA, United States). The universal primer pair for bacteria 515F (5\u0026rsquo;-GTGCCAGCMGCCGCGGTAA-3\u0026rsquo;) and 806R (5\u0026rsquo;-GGACTACHVGGGTWTCTAAT-3\u0026rsquo;) was used to target the V4 region of the 16S rRNA gene using a 2\u0026times;250 bp approach.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of 16S rRNA gene-sequencing data\u003c/h2\u003e \u003cp\u003eWe merged the raw reads into contigs using flash\u0026rdquo; v.1.2.7\u003csup\u003e75\u003c/sup\u003e. We used Qiime v.1.7.0 for quality filtering following the standard operating procedures. We used SILVA v.138 as a reference database and to detect chimeric sequences by the UCHIME algorithm, which we removed from the data. Sequences were clustered into OTUs based on a\u0026thinsp;\u0026ge;\u0026thinsp;97% similarity threshold. We converted the raw data to a phyloseq object (\u0026ldquo;phyloseq\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e\u003c/sup\u003e) and rarefied to the smallest sample size, after removing the sample Nu_ctrl_153a as an outlier. We conducted principal coordinate analyses (PCoA) based on populations and behavioural states and visualized the data (\u0026ldquo;ampvis2\u0026rdquo; package)\u003csup\u003e\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e. In total, we found 22,215 OTUs after rarefaction. We further calculated the frequency of the four most frequent bacterial genera as well as for four additional bacteria genera, \u003cem\u003eAcetobacter, Enterococcus, Fusobacterium, Megamonas\u003c/em\u003e, and the orders Rhizobiales and Entomoplasmatales. \u003cem\u003eAcetobacter\u003c/em\u003e, and \u003cem\u003eEnterococcus\u003c/em\u003e have been associated with aggression in \u003cem\u003eDrosophila melanogaster\u003c/em\u003e\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e, \u003cem\u003eFusobacterium\u003c/em\u003e and \u003cem\u003eMegamonas\u003c/em\u003e have been associated with aggression and non-aggression in dogs\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e, and Rhizobiales and Entomoplasmatales have been associated with aggression in leaf-cutting ants\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Entomoplasmatales were not found in our data set.\u003c/p\u003e \u003cp\u003eUsing the bacterial OTU genera mentioned above, we selected OTUs that had a frequency of at least 100 across the behavioural states (N\u0026thinsp;=\u0026thinsp;119), thus focusing on the most frequent OTUs and restricting the analysis to 119 OTUs. With these, we conducted a sliding-window approach with multinomial logistic regressions (function \u0026ldquo;multinom\u0026rdquo;, \u0026ldquo;nnet\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e). A multinomial regression allows using more than one categorical variable as response variables (here, the three behavioural states). In the sliding-window approach, we created individual models that tested 20 OTUs simultaneously in a multinomial regression. Briefly, the first model used OTUs 1 to 20, the second model OTUs 2 to 21, etc. To evaluate the model fit and calculate p-values and log-likelihood tests, we conducted the same methods as described in the section \u0026ldquo;\u003cem\u003eCombining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression\u003c/em\u003e\u0026rdquo;.\u003c/p\u003e \u003cp\u003eIn each model (N\u0026thinsp;=\u0026thinsp;119), we used the behavioural state \u003cem\u003estarted aggression\u003c/em\u003e as the baseline. Manually checking 119 model fits and results was not efficient, so we created an R (version 4.3.0\u003csup\u003e79\u003c/sup\u003e) script to extract model fits and model p-values for the different OTUs. The script also counted how often OTUs were significantly influencing the behavioural states and thus allowed checking if the same OTUs influenced the behavioural states more or less frequently. Our rationale was that if one or a few OTUs are present in many or all models, then these OTUs likely have a higher impact on the behavioural states than OTUs with a low frequency. If, however, OTUs are only counted a few times, they have likely arisen due to chance and may represent artefacts. From these models, we extracted the significant OTUs and counted their frequency across the models.\u003c/p\u003e \u003cp\u003eAcross these models, the most frequent OTUs (N\u003csub\u003eOTUs\u003c/sub\u003e=58 with a frequency\u0026thinsp;\u0026ge;\u0026thinsp;10) included the genera \u003cem\u003eBacteroides\u003c/em\u003e (relative percentage across the 58 models, 25%), \u003cem\u003eLactobacillus\u003c/em\u003e (9%), \u003cem\u003ePrevotella\u003c/em\u003e (43%), \u003cem\u003ePseudomonas\u003c/em\u003e (17%), the order Rhizobiales (4%), and the genus \u003cem\u003eFusobacterium\u003c/em\u003e (1%). For these gut bacteria, we noted the OTU frequency in each behavioural state and the control. From the 58 OTUs, we excluded 40 OTUs (five because the frequency was significantly higher or lower than in the control, 16 OTUs because the count of the control was higher as the highest number of counts of one of the behaviours, six OTUs because the counts were evenly distributed across all behavioural states, five because the counts were less than 10 in one of the behavioural states, seven OTUs because the counts of the control was similar as the counts of the behavioural states, and one because the counts were not different between the control and the behavioural states) yielding 18 OTUs for further analyses, namely three \u003cem\u003eBacteroides\u003c/em\u003e spp., three \u003cem\u003eLactobacillus\u003c/em\u003e spp., nine \u003cem\u003ePrevotella\u003c/em\u003e spp., three \u003cem\u003ePseudomonas\u003c/em\u003e spp., and one Rhizobiales sp.\u003c/p\u003e \u003cp\u003eWith this set of 18 OTUs, we assessed whether the counts differed between the behavioural states. For this, we conducted a generalised linear model with these count data (response\u0026thinsp;=\u0026thinsp;count; explanatory variable\u0026thinsp;=\u0026thinsp;behavioural states; Poisson-distributed) and assessed the pairwise comparisons (\u0026ldquo;emmeans\u0026rdquo; package; Tukey corrected for multiple testing). Nine OTUs revealed significant results and were further discussed, while others were non-significant or revealed inconsistent results (\u003cem\u003ei.e.\u003c/em\u003e, reacted aggressively higher than the other behavioural states). We could not analyse the microbiome data together with SNPs and DEGs because no samples for the population MQ-N were available for the microbiome analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003eEnvironmental variables used in the multinomial regression analyses\u003c/h2\u003e \u003cp\u003eFor each colony, we estimated a standardised air temperature (TAS) as a rough measure of the colonies\u0026rsquo; thermal niche\u003csup\u003e\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e. Following the logic of Seifert and Pannier (2007)\u003csup\u003e\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e, TAS was calculated for a sampling site as the mean air temperature of the period from May 1st to August 31st averaged over the years 1961 to 1990 of the nearest three meteorological stations (data provided by Klimaabteilung der Zentralanstalt f\u0026uuml;r Meteorologie und Geodynamik (1996), Vienna, Austria). The data were corrected for an altitudinal decrease in temperature of 0.661\u0026deg;C per 100 m according to the equation of Seifert and Pannier (2007):\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eTAS= -0.694×LAT + 0.078 ×LON-0.00661 ×ALT + 52.20, (1)\u003c/h3\u003e\n\u003cp\u003ewhere TAS is the predicted standardised air temperature in \u0026deg;C, LAT and LON denote the geographical latitude and longitude in decimal format, respectively, and ALT is the altitude above sea level in metres.\u003c/p\u003e \u003cp\u003eFrom the WordlClim dataset\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e, we downloaded environmental variables from the years 1970 to 2000 and extracted site-specific values using the \u0026ldquo;extract\u0026rdquo; function (\u0026ldquo;raster\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e\u003c/sup\u003e). In particular, we selected data on mean annual precipitation, precipitation of the warmers quarter, mean annual temperature, and the maximum temperature of the warmest month both as temperature and precipitation affect the colonies\u0026rsquo; environment and higher temperatures promote aggression in this species\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Further, we retrieved soil nitrogen values for each site from the European LUCAS topsoil dataset\u003csup\u003e\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e\u003c/sup\u003e. We used these variables in multinomial regression analyses (described in the next paragraph) to test if the environment is associated with the behavioural states. We recently found such an association in this ant, where higher temperature and nitrogen values were positively associated with aggression\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003eCombining SNPs, DEGs, CHCs, relatedness, and environmental variables counts in a multinomial regression\u003c/h2\u003e \u003cp\u003eIn the multinomial logistic regression, we integrated principal component 1 of the CHC analysis, three SNPs, eight gene expression counts, and colony and environmental variables to assess whether they were associated with the three behavioural states (\u003cem\u003estarted aggression\u003c/em\u003e, \u003cem\u003ereacted aggressively\u003c/em\u003e, \u003cem\u003ereacted peacefully\u003c/em\u003e). Although CHCs, SNPs, gene expression counts, and environmental variables represent distinct biological and abiotic entities, they have high-dimensional features measured across the same samples and thus share a common statistical role. Moreover, a multinomial regression provides the opportunity to use a unified framework to quantify their joint contribution to categorical outcomes while preserving interpretability.\u003c/p\u003e \u003cp\u003eIn total, we tested 24 models (Tab. S9). Models 1\u0026ndash;8 used expression counts of DEGs that were observed in both behavioural comparisons (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB; Tab. S3 highlighted cells). Models 9\u0026ndash;16 used expression counts of DEGs that were up-regulated in the behavioural comparisons. Models 17\u0026ndash;24 used expression counts of DEGs that were down-regulated in the behavioural comparisons. Fitting separate models with increasing number of input variables allowed us to assess if the input variables influence the behavioural states in combination or separately. For example, if some genes are up-regulated in workers that \u003cem\u003ereacted aggressively\u003c/em\u003e but other genes are up-regulated in workers that \u003cem\u003ereacted peacefully\u003c/em\u003e, using these genes in combination may lead to false conclusions.\u003c/p\u003e \u003cp\u003eIn detail, we tested the following models, namely \u0026ldquo;intercept-only\u0026rdquo; models (Models1, 9, 17), models with all three SNP states only (Models 2, 10, 18), models with DEGs that had a log\u003csub\u003e2\u003c/sub\u003efold value of at least\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 (Models 3, 11, 19), models with all three SNP states and DEGs (log\u003csub\u003e2\u003c/sub\u003efold of at least\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5; Models 4, 12, 20), models with the within-colony relatedness values, standardised air temperature, the first PC from a CHC PCA, site-specific soil nitrogen values, mean annual precipitation, precipitation of the warmest quarter, mean annual temperature, and maximum temperature of the warmest month (\u0026ldquo;colony and environmental variables\u0026rdquo;; Models 5, 13, 21), models with all three SNP states and the colony and environmental variables (Models 6, 14, 22), models with DEGs (log\u003csub\u003e2\u003c/sub\u003efold of at least\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5) and colony and environmental variables (Models 7, 15, 23), and, lastly, models with all above-mentioned variables (Models 8, 16, 24).\u003c/p\u003e \u003cp\u003eWe compared the model fits using the \u0026ldquo;anova\u0026rdquo; function (basic stats package; \u0026ldquo;Chi-square test\u0026rdquo;). We further calculated the Akaike Information Criterion for small sample sizes (AICc) of the models (excluding the intercept-only model) using the \u0026ldquo;aictab\u0026rdquo; function (\u0026ldquo;AICcmodavg\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e84\u003c/span\u003e\u003c/sup\u003e) and the models with the lowest ΔAICc (deltaAICc) values represented the best fitting models. Additionally, we calculated a goodness of fit measure for these models by comparing the fit of observed and expected values. To further test if the used variables are significantly influencing the behavioural states, we manually calculated the p-values using a two-tailed Wald Z test. We used the behavioural state \u003cem\u003estarted aggression\u003c/em\u003e as baseline in the logistic regression.\u003c/p\u003e \u003cp\u003eTo subsequently test if the behavioural states differ from each other, we compared their means in a pairwise manner. For this, we calculated the marginal means between the behavioural states using the functions \"emmeans\" and \u0026ldquo;contrast\u0026rdquo; (\u0026ldquo;emmeans\u0026rdquo; package). These post-hoc tests compared the behavioural states and allowed conducting hypothesis tests to determine whether the differences were statistically significant. We also calculated two pseudo coefficients of determination (R\u003csup\u003e2\u003c/sup\u003e, \u0026ldquo;Nagelkerke\u0026rdquo; and \u0026ldquo;McFadden\u0026rdquo;) to check how much of the variation is explained by the independent variables. As we used multinomial regressions, the pseudo-R\u003csup\u003e2\u003c/sup\u003e values were only approximated. We further assessed the significance of the independent variables individually using a likelihood ratio test \u0026ldquo;lrtest\u0026rdquo; (\u0026ldquo;lmtest\u0026rdquo; package\u003csup\u003e\u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e\u003c/sup\u003e). This test drops the focal variable in the model to assess its impact on the model (i.e., it compares a focal model with the same model by excluding the targeted independent variable). If the model differs significantly, the focal variable is a dominant variable in the model.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll code and datasets generated and/or analysed in the study will be made publicly available alongside the publication\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Philipp Andesner and Elisabeth Zangerl for help during lab work and Marlene Haider and Markus M\u0026ouml;st for helpful discussions during statistical analysis. We thank the government of Carinthia for issuing a sampling permit for the protected area \u0026ldquo;Mussen\u0026rdquo;. The LUCAS topsoil dataset used was made available by the European Commission through the European Soil Data Centre managed by the Joint Research Centre (JRC; http://esdac.jrc.ec.europa.eu/). The computational results presented here have been achieved (in part) using the LEO HPC infrastructure of the University of Innsbruck and the MACH2 Interuniversity Shared Memory Supercomputer. This study was financially supported by the FWF (Austrian Science Fund) under Award Number P 30861 awarded to F.M.S. and by the European Union\u0026rsquo;s Horizon Europe programme under Marie Skłodowska-Curie Actions (MSCA) - Postdoctoral fellowship grant agreement no. 101204375 awarded to P.K. The maps were created using Stadia Maps (stadiamaps.com), Stamen design (stamen.com), and OpenStreetMap (openstreetmap.org/copyright).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatrick Krapf:\u003c/strong\u003e Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing \u0026ndash; original draft, Writing \u0026ndash; review \u0026amp; editing, Visualisation. \u003cstrong\u003eFrancesco Cicconardi\u003c/strong\u003e: Conceptualization, Methodology, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eMartin Schilling:\u0026nbsp;\u003c/strong\u003eMethodology, Validation, Formal analysis, Investigation, Visualization, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eGerhard P. Aigner:\u003c/strong\u003e Investigation, Formal analysis, Resources, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eThomas Klammsteiner:\u003c/strong\u003e Methodology, Software, Formal analysis, Investigation, Writing \u0026ndash; review \u0026amp; editing, Visualization. \u003cstrong\u003eManfred Ayasse:\u003c/strong\u003e Conceptualization, Methodology, Investigation, Resources, Data curation, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eWolfgang Arthofer:\u003c/strong\u003e Conceptualization, Methodology, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eAlexander S. Mikheyev:\u0026nbsp;\u003c/strong\u003eConceptualization, Methodology, Writing \u0026ndash; review \u0026amp; editing. \u003cstrong\u003eFlorian M. Steiner:\u003c/strong\u003e Conceptualization, Methodology, Resources, Writing \u0026ndash; review \u0026amp; editing, Supervision, Project administration, Funding acquisition. \u003cstrong\u003eBirgit C. Schlick-Steiner:\u003c/strong\u003e Conceptualization, Methodology, Resources, Writing \u0026ndash; review \u0026amp; editing, Supervision, Project administration, Funding acquisition.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCovington H, Newman IIIE, Leonard M, Miczek K (2019) Translational models of adaptive and excessive fighting: an emerging role for neural circuits in pathological aggression [version 1; peer review: 3 approved]. \u003cem\u003eF1000Research\u003c/em\u003e 8, 18883\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarshall VT, Humfeld SC, Bee MA (2003) Plasticity of aggressive signalling and its evolution in male spring peepers, \u003cem\u003ePseudacris crucifer\u003c/em\u003e. Anim Behav 65:1223\u0026ndash;1234\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNatarajan D, Caramaschi D (2010) Animal violence demystified. Front Behav Neurosci 4:1\u0026ndash;9\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCraig W (1921) Why do animals fight? \u003cem\u003eIJE\u003c/em\u003e 31, 264\u0026ndash;278\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eContala M-L, Krapf P, Steiner FM (2024) Schlick-Steiner, Foraging valor linked with aggression: selection against completely abandoning aggression in the high-elevation ant \u003cem\u003eTetramorium alpestre?\u003c/em\u003e. Insect Sci 31:953\u0026ndash;970\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuntingford FA, Turner AK (1987) in \u003cem\u003eAnimal Conflict\u003c/em\u003e, F. A. Huntingford, A. K. Turner, EdsSpringer Netherlands, Dordrecht,. pp. 227\u0026ndash;250\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo X, Dukas R (2020) The cost of aggression in an animal without weapons. Ethology 126:24\u0026ndash;31\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiles-Novelo A, Anderson CA (2022) Climate change and human behavior: Impacts of a rapidly changing climate on human aggression and violence. Cambridge University Press, Cambridge, p 68\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrapf P, Arthofer W, Ayasse M, Steiner FM (2023) Schlick-Steiner, Global change may make hostile \u0026ndash; Higher ambient temperature and nitrogen availability increase ant aggression. Sci Total Environ 861:160443\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKudryavtseva NN, Bondar NP, Alekseyenko OV (2000) Behavioral correlates of learned aggression in male mice. Aggressive Behav 26:386\u0026ndash;400\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJia Y et al (2021) Gut microbiome modulates \u003cem\u003eDrosophila\u003c/em\u003e aggression through octopamine signaling. Nat Commun 12:2698\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMondo E et al (2020) Gut microbiome structure and adrenocortical activity in dogs with aggressive and phobic behavioral disorders. Heliyon 6:e03311\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeseo S et al (2019) The scent of symbiosis: gut bacteria may affect social interactions in leaf-cutting ants. Anim Behav 150:239\u0026ndash;254\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Wilgenburg E, Cl\u0026eacute;mencet J, Tsutsui ND (2010) Experience influences aggressive behaviour in the Argentine ant. Biol Lett 6:152\u0026ndash;155\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdamo SA, Hoy RR (1995) Agonistic behaviour in male and female field crickets, \u003cem\u003eGryllus bimaculatus\u003c/em\u003e, and how behavioural context influences its expression. Anim Behav 49:1491\u0026ndash;1501\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMargalida A, Bertran J, Boudet J, Heredia R (2004) Hatching asynchrony, sibling aggression and cannibalism in the Bearded Vulture \u003cem\u003eGypaetus barbatus\u003c/em\u003e. Ibis 146:386\u0026ndash;393\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHsu Y, Earley RL, Wolf LL (2006) Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes. Biol Rev 81:33\u0026ndash;74\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSprenger PP, Menzel F (2020) Cuticular hydrocarbons in ants (Hymenoptera: Formicidae) and other insects: how and why they differ among individuals, colonies, and species. Myrmecol News 30:1\u0026ndash;26\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartin SJ, Vitikainen E, Drijfhout FP, Jackson D (2012) Conspecific ant aggression is correlated with chemical distance, but not with genetic or spatial distance. Behav Genet 42:323\u0026ndash;331\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTillisch K et al (2017) Brain structure and response to emotional stimuli as related to gut microbial profiles in healthy women. Biopsychosoc Med 79:905\u0026ndash;913\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKirchoff NS, Udell MAR, Sharpton TJ (2019) The gut microbiome correlates with conspecific aggression in a small population of rescued dogs (\u003cem\u003eCanis familiaris\u003c/em\u003e). \u003cem\u003ePeerJ\u003c/em\u003e 7, e6103\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarkar A et al (2020) The role of the microbiome in the neurobiology of social behaviour. Biol Rev 95:1131\u0026ndash;1166\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFaraone SV et al (2005) Molecular genetics of Attention-Deficit/Hyperactivity Disorder. Biol Psychiatry 57:1313\u0026ndash;1323\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEngelmann C, Haenold R (2016) Transcriptional control of synaptic plasticity by transcription factor NF-κB. \u003cem\u003eNeural Plasticity\u003c/em\u003e 7027949 (2016)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Vita J (1979) Mechanisms of interference and foraging among colonies of the harvester ant \u003cem\u003ePogonomyrmex californicus\u003c/em\u003e in the Mojave Desert. Ecology 60:729\u0026ndash;737\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteiner FM et al (2007) Abandoning aggression but maintaining self-nonself discrimination as a first stage in ant supercolony formation. Curr Biol 17:1903\u0026ndash;1907\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteinmeyer C, Pennings PS, Foitzik S (2012) Multicolonial population structure and nestmate recognition in an extremely dense population of the European ant \u003cem\u003eLasius flavus\u003c/em\u003e. Insectes Soc 59:499\u0026ndash;510\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrtius D, Heinze J (1999) Fertility signaling in queens of a North American ant. Behav Ecol Sociobiol 45:151\u0026ndash;159\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarinho CGS, Ribeiro MMR, Della Lucia TMC, Guedes RNC (2006) Aggressive response of pest ant species to b-eudesmol (Hymenoptera: Formicidae). Sociobiology 47:445\u0026ndash;454\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrapf P et al (2018) An Alpine ant\u0026rsquo;s behavioural polymorphism: monogyny with and without internest aggression in \u003cem\u003eTetramorium alpestre\u003c/em\u003e. Ethol Ecol Evol 30:220\u0026ndash;234\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrapf P, Sedfaoui K, Contala M-L, Steiner FM (2023) Schlick-Steiner, Short-time development of among-colony behaviour in a high-elevation ant. Behav Processes 208:104872\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHelanter\u0026auml; H, Strassmann JE, Carrillo J, Queller DC (2009) Unicolonial ants: where do they come from, what are they and where are they going? Trends Ecol Evol 24:341\u0026ndash;349\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuziel GA, Rakoff-Nahoum S (2022) The gut microbiome. Curr Biol 32:R257\u0026ndash;R264\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR. V. Lenth, emmeans: estimated marginal means, aka least-squares means, R package version 1.11.2-8, (2025) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://CRAN.R-project.org/package=emmeans)\u003c/span\u003e\u003cspan address=\"https://CRAN.R-project.org/package=emmeans)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFick SE, Hijmans RJ (2017) WorldClim 2: new 1-km spatial resolution climate surfaces for global land areas. Int J Climatol 37:4302\u0026ndash;4315\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWarde-Farley D et al (2010) The GeneMANIA prediction server: biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res 38:W214\u0026ndash;W220\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErrard C, Hefetz A, Jaisson P (2006) Social discrimination tuning in ants: template formation and chemical similarity. Behav Ecol Sociobiol 59:353\u0026ndash;363\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin P et al (2017) \u003cem\u003ePrevotella\u003c/em\u003e and \u003cem\u003eKlebsiella\u003c/em\u003e proportions in fecal microbial communities are potential characteristic parameters for patients with major depressive disorder. J Affect Disord 207:300\u0026ndash;304\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStrandwitz P et al (2019) GABA-modulating bacteria of the human gut microbiota. Nat Microbiol 4:396\u0026ndash;403\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVernier CL et al (2020) The gut microbiome defines social group membership in honey bee colonies. Sci Adv 6:eabd3431\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOchoa-Rep\u0026aacute;raz J, Mielcarz DW, Begum- Haque S, Kasper LH (2011) Gut, bugs, and brain: Role of commensal bacteria in the control of central nervous system disease. Ann Neurol 69:240\u0026ndash;247\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ed. Almeida LG, d. LAB, Moraes JR, Trigo C, Omoto FL, C\u0026ocirc;nsoli (2017) The gut microbiota of insecticide-resistant insects houses insecticide-degrading bacteria: A potential source for biotechnological exploitation. PLoS ONE 12:e0174754\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrant RJ, Daniell TJ, Betts WB (2002) Isolation and identification of synthetic pyrethroid-degrading bacteria. J Appl Micro 92:534\u0026ndash;540\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDinan TG, Cryan JF (2017) The microbiome-gut-brain axis in health and disease. Gastroenterol Clin 46:77\u0026ndash;89\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBao Y-Y et al (2014) Genomic insights into the serine protease gene family and expression profile analysis in the planthopper, \u003cem\u003eNilaparvata lugens\u003c/em\u003e. BMC Genomics 15:507\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOta KT, Andres W, Lewis DA, Stockmeier CA, Duman RS (2015) BICC1 expression is elevated in depressed subjects and contributes to depressive behavior in rodents. Neuropsychopharmacology 40:711\u0026ndash;718\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTanaka T, Goto K, Iino M (2017) Diverse functions and signal transduction of the Exocyst complex in tumor cells. J Cell Physiol 232:939\u0026ndash;957\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKiragasi B, Wondolowski J, Li Y, Dickman DK (2017) A presynaptic glutamate receptor subunit confers robustness to neurotransmission and homeostatic potentiation. Cell Rep 19:2694\u0026ndash;2706\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu X, Krause WC, Davis RL (2007) GABA\u003csub\u003eA\u003c/sub\u003e receptor RDL inhibits \u003cem\u003eDrosophila\u003c/em\u003e olfactory associative learning. Neuron 56:1090\u0026ndash;1102\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDrapeau MD (2003) A novel hypothesis on the biochemical role of the \u003cem\u003eDrosophila\u003c/em\u003e Yellow protein. Biochem Biophys Res Com 311:1\u0026ndash;3\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTamagnone L et al (1999) Plexins are a large family of receptors for transmembrane, secreted, and gpi-anchored semaphorins in vertebrates. Cell 99:71\u0026ndash;80\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGidron Y, Russ K, Tissarchondou H, Warner J (2006) The relation between psychological factors and DNA-damage: A critical review. Biol Psychol 72:291\u0026ndash;304\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAtanackovic D, Schulze J, Kr\u0026ouml;ger H, Brunner-Weinzierl MC, Deter HC (2003) Acute psychological stress induces a prolonged suppression of the production of reactive oxygen species by phagocytes. J Neuroimmunol 142:159\u0026ndash;165\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAyasse M, Paxton RJ, Tengo J (2001) Mating behavior and chemical communication in the order hymenoptera. Annu Rev Entomol 46:31\u0026ndash;78\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eConrad T, Paxton RJ, Barth FG, Francke W, Ayasse M (2010) Female choice in the red mason bee, \u003cem\u003eOsmia rufa\u003c/em\u003e (L.) (Megachilidae). J Exp Biol 213:4065\u0026ndash;4073\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang Y, Horikoshi M, Li W (2016) ggfortify: Unified interface to visualize statistical results of popular R packages. (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.32614/RJ-2016-060\u003c/span\u003e\u003cspan address=\"10.32614/RJ-2016-060\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaechler M, Rousseeuw P, Struyf A, Hubert M (2025) H. K, cluster: Cluster Analysis Basics and Extensions. (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.31234/osf.io/7awm6\u003c/span\u003e\u003cspan address=\"10.31234/osf.io/7awm6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoulston TH, Buczkowski G, Silverman J (2003) Nestmate discrimination in ants: Effect of bioassay on aggressive behavior. Insectes Soc 50:151\u0026ndash;159\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreenberg ME, Ziff EB (1984) Stimulation of 3T3 cells induces transcription of the c-fos proto-oncogene. Nature 311:433\u0026ndash;438\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrapf P, Hochenegger N, Arthofer W, Schlick-Steiner BC, Steiner FM (2019) Comparing ant behaviour indices for fine-scale analyses. Sci Rep 9:6856\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErrard C, Hefetz A (1997) Label familiarity and discriminatory ability of ants reared in mixed groups. Insectes Soc 44:189\u0026ndash;198\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiry S et al (2004) GENECLASS2: A software for genetic assignment and first-generation migrant detection. J Hered 95:536\u0026ndash;539\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRannala B, Mountain JL (1997) Detecting immigration by using multilocus genotypes. Proc Natl Acad Sci USA 94:9197\u0026ndash;9201\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQueller DC, Goodnight KF (1989) Estimating relatedness using genetic-markers. Evolution 43:258\u0026ndash;275\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePamilo P (1991) Evolution of colony characteristics in social insects. II. Number of reproductive individuals. Am Nat 138:412\u0026ndash;433\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCicconardi F et al (2020) Genomic signature of shifts in selection in a subalpine ant and its physiological adaptations. Mol Biol Evol 37:2211\u0026ndash;2227\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDanecek P et al (2011) The variant call format and VCFtools. Bioinformatics 27:2156\u0026ndash;2158\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eManichaikul A et al (2010) Robust relationship inference in genome-wide association studies. Bioinformatics 26:2867\u0026ndash;2873\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou X, Stephens M (2012) Genome-wide efficient mixed-model analysis for association studies. Nat Gen 44:821\u0026ndash;824\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eS. D. Turner, qqman: an R package for visualizing GWAS results using QQ and manhattan plots. JOSS 3, 005165 (2018)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLove MI, Huber W, Anders S (2014) Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15:550\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKolde R pheatmap: Pretty Heatmaps, R package version 1.0.12, 2019 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://CRAN.R-project.org/package=pheatmap)\u003c/span\u003e\u003cspan address=\"https://CRAN.R-project.org/package=pheatmap)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYan L ggvenn: Draw Venn Diagram by 'ggplot2', R package version 0.1.10, 2023 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://CRAN.R-project.org/package=ggvenn)\u003c/span\u003e\u003cspan address=\"https://CRAN.R-project.org/package=ggvenn)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMendoza-Guido B et al (2023) Low diversity and host specificity in the gut microbiome community of \u003cem\u003eEciton\u003c/em\u003e army ants (Hymenoptera: Formicidae: Dorylinae) in a Costa Rican rainforest. Myrmecol News 33:19\u0026ndash;34\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMagoč T, Salzberg SL (2011) FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27:2957\u0026ndash;2963\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcMurdie PJ, Holmes S (2013) phyloseq: An R package for reproducible interactive analysis and graphics of microbiome census data. PLoS ONE 8:e61217\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndersen KS, Kirkegaard RH, Karst SM, Albertsen M (2018) ampvis2: an R package to analyse and visualise 16S rRNA amplicon data. \u003cem\u003ebioRxiv\u003c/em\u003e. (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1101/299537\u003c/span\u003e\u003cspan address=\"10.1101/299537\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVenables WN, Ripley BD (2002) Modern Applied Statistics with S, vol 4. Springer, New York, p 498\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCore Team R (2022) R: A language and environment for statistical computing\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWagner HC et al (2017) Light at the end of the tunnel: Integrative taxonomy delimits cryptic species in the \u003cem\u003eTetramorium caespitum\u003c/em\u003e complex (Hymenoptera: Formicidae). Myrmecol News 25:95\u0026ndash;129\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeifert B, Pannier L (2007) A method for standardized description of soil temperatures in terrestrial ecosystems. Abh Ber Naturkundemus G\u0026ouml;rlitz 78:151\u0026ndash;182\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR. J. Hijmans, raster: Geographic Data Analysis and Modeling. R package version 3.1-5. (2020)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eT\u0026oacute;th G, Jones A, Montanarella L (2013) \u003cem\u003eLUCAS Topsoil Survey: methodology, data and results\u003c/em\u003e. JRC Technical Reports (Publications Office of the European Union, EUR 26102 - Scientific and Technical Research series, Luxembourg\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMazerolle M AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c), R package version 2.3.3, 2023 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://cran.r-project.org/package=AICcmodavg)\u003c/span\u003e\u003cspan address=\"https://cran.r-project.org/package=AICcmodavg)\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeileis A, Hothorn T (2002) Diagnostic checking in regression relationships. R News 2:7\u0026ndash;10\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Whole-genome sequencing, transcriptomics, gut microbiome, cuticular hydrocarbons, behaviour, start of aggression, Tetramorium alpestre","lastPublishedDoi":"10.21203/rs.3.rs-8539228/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8539228/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAnimals frequently display aggressive behaviour, for example, when competing for food. Aggression is influenced by various extrinsic and intrinsic factors such as temperature, the microbiome, and genetics. However, we currently lack understanding what factors cause an animal to start aggression. Here, we use an ant species to test if chemical, microbiome, genomic, and/or transcriptomic traits correlate with the start of aggression and the reactions to it, that is, reacting aggressively or peacefully. We found nine bacterial operational taxonomic units, mutations in two genes, and eight differentially expressed genes, which were positively or negatively associated with the start of aggression or reactions to it. These traits are mainly linked to hormone signalling and neurological and synaptic functions. The results indicate that multiple traits, possibly acting in concert, affect the start of aggression and reactions to it. We speculate that such traits could promote aggression and could thus play important evolutionary roles.\u003c/p\u003e","manuscriptTitle":"The gut microbiome, single nucleotide polymorphisms, and differentially expressed genes promote aggression in an ant","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-09 07:16:58","doi":"10.21203/rs.3.rs-8539228/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"11af7556-975b-4ca3-b36c-56f67bcc13f4","owner":[],"postedDate":"January 9th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":60753525,"name":"Biological sciences/Zoology/Animal behaviour"},{"id":60753526,"name":"Biological sciences/Genetics"},{"id":60753527,"name":"Biological sciences/Ecology/Behavioural ecology"}],"tags":[],"updatedAt":"2026-03-09T14:55:51+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-09 07:16:58","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8539228","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8539228","identity":"rs-8539228","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}