Sign and strength of pairwise interactions in natural isolates depend on environment type

20 Bacterial interactions- whether positive or negative – are crucial for the functioning of microbial 21 communities. Though bacterial interactions are mainly expected to be negative, the sign and 22 strength of interactions are predicted to be context dependent, with interactions typically being 23 more positive in more stressful and nutrient-poor conditions. However, systematic studies 24 investigating how the environment affects interactions between multiple taxa are lacking. 25 Here, we determine if interactions between a panel of natural soil isolates change in response to 26 the environment in which they are grown, with two different artificial media used (one simple and 27 one complex) and a more ecologically relevant soil wash. To maximise natural variation in 28 interactions, we collected multiple isolates from multiple sites: co-occurring (sympatric) isolates 29 were predicted to show more negative interactions than allopatric isolates because of greater 30 overlap in resource use. Pairwise interactions were in general negative, but more negative when 31 grown in a complex lab-derived medium (Tryptic Soy Broth). Mutually beneficial interactions were 32 most common in a simple resource medium (M9 minimal media) and exploitative interactions were 33 most frequent in a soil broth. These patterns were independent of whether species originated from 34 the same or a different site. The study supports the prediction that nutrient rich environments 35 promote more negative interactions, and that measuring interactions of soil isolates in standard lab 36 media is likely to misrepresent interactions occurring in natural environments. 37 Key Words 38 Bacterial Interactions, resources, phylogenetic diversity, natural resources. 39 40

41 Bacterial communities underpin many key ecological processes and live in diverse environments 42 that vary in their resource complexity and availability. Interactions between co-occurring species 43 can range from cooperative and mutually beneficial (+/+), through exploitative (+/-) to competitive (-44 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 3 /-). The prevailing interaction type can shape community composition and ecosystem services (1-45 3), and influence overall stability, with competitive interactions predicted to increase the latter (4, 46 5). Understanding interspecific bacterial interactions is therefore key to predicting and maintaining 47 bacterial community function in both natural (6) and applied contexts (7). Though similar 48 interactions can occur within-species, these primarily involve different mechanisms (e.g. kin-49 selection (8)) that are not the focus of the current study. 50 Resource characteristics, including complexity, availability, and overall abundance, are key factors 51 influencing the type and strength of bacterial interactions (9, 10). For example, environments which 52 are rich and diverse in nutrients will support high bacterial densities, which in turn increase the 53 magnitude of competition for space and resources (11), leading to competitive interactions (-/-) (12-54 14). More complex nutrient-rich environments can allow bacterial investment both into growth and 55 competitor suppression (e.g. through production of toxins), as observed in Bacillus sp.(15). 56 Alternatively, environments containing resources which are less readily available can lead to 57 bacterial species alleviating competition through niche complementarity, for instance as species 58 break down recalcitrant resources, or as they rely on conspecifics for enzymatic breakdown. (9, 59 10). 60 Both mutually beneficial (+/+) and exploitative (+/-) interactions are equally dependent on 61 environmental resources, and on the resulting flow of metabolites between species. Metabolic 62 cross feeding, where species (co) consume metabolic byproducts of others (16), can lead to 63 mutual benefits for interacting species, as in seen between methanogenic archaea and sulphate-64 reducing bacteria (17, 18), where by-products produced by the reducing bacteria are used by 65 methanogens, preventing harmful waste buildup that would inhibit bacterial growth; whilst 66 adaptation to utilise waste products has also been shown to decrease competitive interactions 67 amongst bacterial isolates in beech tea medium (19). Such metabolic dependencies are sensitive 68 to changes in the resource environment and interaction types can alter in response to changes in 69 external nutrient supply (20, 21). For example, the mutualistic interaction between E. coli and S. 70 enterica shifts to exploitative (22) when adding limiting metabolites: while S. enterica still benefits 71 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 4 from the acetate produced by E. coli, the latter gains no return and competes for shared nutrients. 72 A similar pattern was observed in pairwise combinations of 4 bacterial species relative to 73 monoculture, where the addition of external nutrients led to increased competition, relative to the 74 mutually beneficial interactions seen in a harsher environment (23). 75 The way in which species interact with one another not only depends on the resource environment 76 but also on whether these originate from the same (sympatric) or a different (allopatric) 77 environment. Sympatric bacteria are more likely to overlap in resource use and niche space (i.e. 78 due to ecological filtering (24)) and are often more closely related, which tends to favour 79 competitive interactions (13, 25-28). The higher relatedness of sympatric bacteria (29), could 80 however lead to more positive interactions as a result of complementary cell surface adhesion and 81 quorum sensing communication, although the reverse has also been found (30). 82 Laboratory growth media vary widely in resource complexity, from minimal media containing a 83 single carbon source (e.g., M9) to rich, undefined media such as tryptic soy broth. These 84 differences in resource composition are likely to favour different interaction outcomes by shaping 85 resource use and niche overlap among competing strains. Additionally, pairwise competition 86 experiments are often conducted in such laboratory media, designed to maximize bacterial cell 87 densities. Under these conditions, selection for rapid growth—combined with ecological filtering for 88 bacteria that share similar physiological traits (31, 32) —is likely to bias observed interactions 89 toward more negative outcomes. 90 Crucially, laboratory media is likely to not be representative of a natural environment and may not 91 allow the quantification of interactions which naturally occur. Therefore, systematic investigations of 92 how bacterial interactions vary across different types of laboratory media, and relative to more 93 natural conditions - whilst considering metabolic similarity driven by spatial co-occurrence - are 94 critical. 95 Here, we tested whether interactions between bacterial pairs change between a semi-natural (soil 96 wash) and two artificial growth media (Tryptic soy broth and minimal media), that differ in 97 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 5 complexity. Furthermore, we tested the importance of geographic distance in driving the nature of 98 species interactions. 99 We hypothesised that a rich and complex laboratory medium (Tryptic soy broth: TSB) would lead to 100 high growth rates in monoculture and that this may lead to antagonistic interactions in coculture 101 due to the ability to invest into growth and interference mechanisms (15), and could result in the 102 loss of one competing isolate. Considering the more simplified resource of the minimal media, it 103 was predicted that pairs may alleviate competition to synthesise the limiting resource (9, 10, 33) 104 and result in less negative interactions relative to those seen in TSB. Finally, we predicted that soil 105 wash would be an intermediate between the other two media and would allow the quantification of 106 interactions more representative of those occurring in the natural environment. We predicted that 107 sympatric bacteria (isolated from same soil sample) would experience more resource overlap and 108 interactions would be largely competitive when compared to allopatric bacteria – particularly in the 109 soil wash media, where the resource profile will match that of the species’ natural environment 110 (34). 111 112 113 114 115 116

117 Bacterial isolation 118 Bacteria were sampled from four different sites, all contained within the gardens of the Penryn 119 Campus, Penryn, Cornwall (50.170, 5.125). At each site, five soil samples were taken by fully 120 inserting a sterile pipette tip (40mm) into the soil, after which tips were individually stored in 15 mL 121 of minimal salts buffer. Soil suspensions were then plated onto King’s Medium B (KB) agar 122 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 6 containing Nystatin (20/i1mg/L) to hamper fungal growth and incubated for 24 hours at 28°C. For 123 each of the 20 plated soil samples, five morphologically dissimilar colonies were picked, yielding a 124 total 100 bacterial isolates. Each isolate was regrown for 24 hours in TSB before being frozen in 125 glycerol at a final concentration of 25% at -70°C. 126 All 100 isolates were tested for antibiotic resistance by selective plating onto KB agar 127 supplemented with Gentamicin (20µg/L;10µg/L) and Tetracycline (10µg/L). To distinguish paired 128 isolates that were visually similar, clones were paired such that one was antibiotic resistant and 129 one susceptible. This allowed for stamp plating of coculture plates onto selective agar, and 130 quantification of colony numbers of each isolate in pairwise assays. We note that selective plating 131 was only used for three pairs as isolates within a pair were morphologically distinct on KB agar. 132 133 Figure 1: Bacterial isolation method showing how the three locality categories used relate to 134 geographic proximity. Highest spatial overlap is the local treatment, where isolates were picked 135 from the same plated community, originating from the same tip. General locality are isolates which 136 are picked from different plated communities from different tips, but which were taken at the same 137 site. The treatment different has the least spatial overlap, with isolates picked from different plated 138 communities originating from different sites. 139 d .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 7 18 pairs of isolates were selected based on their antibiotic sensitivity (ensuring one resistant and 140 one susceptible), with 6 pairs each taken from general, local and different environments (Figure 1). 141 These three different localities are classed based on geographic proximity of pairs: ‘Local’ denotes 142 isolates picked from the same plated community, originating from a single pipette tip; ‘General’ 143 denotes isolates picked from different plated communities (different pipette tips), but which 144 originated from the same site and ‘Different’ denotes isolates from different plated communities and 145 sites. 146 Values missing from subsequent plots were due to insufficient growth in either plated inoculum 147 densities or following 48 hours of growth in media (isolates 21,22 and 35 removed from the study 148 and 1 rep missing in isolates 16,33 and 34; SI). 149 Experimental Procedure 150 To quantify how the resource environment affects bacterial growth and species interactions, we 151 used three different media types: (1)TSB (Sigma-Aldrich), (2) minimal media supplemented with 152 glucose (final volume 1 L: 20mL glucose solution, 2mL MgSO4 (1M), 0.1mL CaCL2 solution (1M), 153 100mL 10*M9 salts (SI), 878mL dH20) and (3) soil wash. The latter was made by collecting 200g 154 of soil from the same general area the bacteria were isolated from and suspending this in 1L of 155 water for 24 hours, after which the soil wash was filtered to remove larger particles and autoclaved 156 twice with a 48-hour gap between both runs. It is important to note that the same soil wash was 157 used for all replicates, and the collected soil used in the wash was not from any of the sampling 158 locations but taken from a distinct soil environment within the same location. 159 To set up the experiment, single colonies of all selected isolates were grown statically in 6mL of 160 TSB for 24 hours. To minimise resource transfer, 1mL of each culture was centrifuged for three 161 minutes at 14000 rpm, after which we removed the supernatant, and resuspended the pellet into 162 1mL M9 buffer to remove access resources. These washed cultures were then used to inoculate 163 the experimental media, and frozen in 50% glycerol at -70°C to determine inoculum densities. 164 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 8 For monocultures, 60µL of resuspended culture was added to 6mL of relevant media in glass vials, 165 for cocultures 30 µL of each member was added, keeping total inoculation density constant. The 166 coculture treatment was repeated three times for each pair per medium. These were incubated 167 statically at 28°C for two days after which all vials were vortexed and glycerol freezer stocks made. 168 All freezer stocks were plated onto KB agar and colony counts were performed. 169 Colony PCR and species identification 170 To understand the phylogenetic relationship of competing pairs, the 16S rRNA gene of each of the 171 26 isolates was sequenced, using 27F and 1492R primers and Sanger sequencing (Source 172 Bioscience). 173 Data Analyses 174 For all analyses R Version 4.3.1 was used. Model behaviour was tested using the ‘DHARMa’ 175 package (35). Post-Hoc analyses were carried out using the ‘emmeans’ package (36), with p-176 values adjusted to correct for multiple testing, and all plots were made using ‘ggplot2’ (37), 177 additional packages used are cited below. To determine the significance of variables, models were 178 simplified by sequentially removing variables and compared via F, χ2 or likelihood ratio tests (38), 179 where appropriate. 180 To assess how the different environments affected isolate growth, we carried out a linear model, 181 testing differences in monoculture density change in response to media type. 182 Monoculture density change = (ln/g3014/g3042/g3041/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3005/g3032/g3041/g3046/g3036/g3047/g3052 /g3010/g3041/g3042/g3030/g3048/g3039/g3048/g3040 /g3005/g3032/g3041/g3046/g3036/g3047/g3052 ). 183 Where ‘Inoculum density’ is the inoculated colony counts of each isolate and ‘Monoculture density’ 184 is the final colony counts following 48 hours growth in each media type. 185 For co-cultures, we quantified species interaction signs by calculating the Relative Intensity Index 186 (RII) using mono- and co-culture colony counts for each isolate following 48 hrs of growth: 187 Relative Intensity Index (RII) = /g4666/g3004/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3030/g3042/g3039/g3042/g3041/g3052 /g3030/g3042/g3048/g3041/g3047/g3046/g2879/g3014/g3042/g3041/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3030/g3042/g3039/g3042/g3041/g3052 /g3030/g3042/g3048/g3041/g3047/g3046/g4667 /g4666/g3004/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3030/g3042/g3039/g3042/g3041/g3052 /g3030/g3042/g3048/g3041/g3047/g3046/g2878/g3014/g3042/g3041/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3030/g3042/g3039/g3042/g3041/g3052 /g3030/g3042/g3048/g3041/g3047/g3046/g4667 188 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 9 This proxy reduces the influence of extreme outliers arising in pairs where single isolates were 189 outcompeted (39), with interaction indices for isolates bound between -1 and 1, such that positive 190 values represent an increase in coculture relative to monoculture density and negative values 191 reflect a decrease. 192 To test the effect of media type and locality on RII, a linear mixed effects model (Lmer) was fitted 193 using the ‘lme4’ package (40) , with a random intercept fitted for each clone and each pair to 194 account for multiple observations. We tested for the main effects of media, locality and their two-195 way interaction – however this interaction did not significantly improve model behaviour and was 196 subsequently removed from the final model. 197 To determine whether individual growth explains performance in pairwise co-culture, we carried out 198 a linear model, testing the effect of media and monoculture growth, as well as their two-way 199 interaction, on average RII, where Average RII = (RIIrep1 + RIIrep2 + RIIrep3/3). A random intercept for 200 ‘pairs’ was included, however the variance explained approach zero and its removal greatly 201 improved model behaviour. The interaction fitted between media and monoculture growth did not 202 significantly improve model behaviour and was subsequently removed from the final model. 203 Individual RII scores of paired isolates were then used to class interactions for each pair into 204 mutually beneficial (+/+), exploitative (+/-) or competitive (-/-). For example, if isolate 1 RII = 0.1 205 and isolate 2 RII = -0.2, the pairwise interaction is classed as exploitative (+/-), with isolate 1 206 growing better in coculture relative to monoculture and the reverse being true for isolate 2. This 207 was carried out for each individual replicate. 208 The likelihood of isolates having either mutually beneficial or exploitative interactions was 209 assessed through a multinomial logistic regression, using the ‘nnet’ package (41), with the 210 competitive interactions as base category, as this was the most frequent interaction seen across all 211 isolates. Plotting and confidence intervals were carried out using ‘ggeffects’ (42). 212 Sanger sequences were analysed, using sangeranalyseR (1.20.0), where forward and reverse 213 reads were trimmed, processed and aligned (43). Isolate 4 was unable to be sequenced, resulting 214 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 10 in 35 sequenced isolates. For these 35 isolates, a pairwise distance matrix was created using the 215 dist.dna function in the ‘ape’ package, and a phylogenetic tree constructed (44). Taxonomies 216 shown on the phylogenetic tree were assigned using dada2, utilising the RDP Naïve Baysesian 217 Classifier (45), taxonomies were unable to be assigned for 5 isolates. Chromatograms for each 218 reaction and overall quality checks are available in the SI. 219

220 Monoculture growth as a function of resource environment 221 To understand how the resource environment affected individual growth, we calculated the relative 222 density change of each isolate following 48 hrs of incubation (Figure 2). Growth varied significantly 223 across environments (linear model (lm), F= 64.557, P<0.001, Pairwise contrasts p<0.001 for all), 224 with isolates growing fastest in TSB (estimated mean [95% confidence intervals], 3.47, 225 [2.84,4.10]). The minimal media decreased densities relative to the inoculum (-1.69, [-2.32,-1.05]) 226 and the soil wash on average kept densities similar to the inoculum (0.57, [-0.05,1.19]). 227 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 11 228 229 Figure 2: Plot showing individual growth (n = 36 isolates) as a function of resource environment. 230 Growth was expressed as (ln /g3014/g3042/g3041/g3042/g3030/g3048/g3039/g3047/g3048/g3045/g3032 /g3005/g3032/g3041/g3046/g3036/g3047/g3052 /g3010/g3041/g3041/g3042/g3030/g3048/g3039/g3048/g3040 /g3005/g3032/g3041/g3046/g3036/g3047/g3052 ). Grey lines link performance of individual isolates 231 across environments, with the black line showing the mean growth. Blue dots show growth in 232 minimal media, orange in soil wash and grey in Tryptic Soy Broth (TSB). Values above zero 233 indicate isolates reached higher densities following 2 days incubation compared to that of the 234 inoculum. 235 236 Interactions vary across media types, independent of geographic proximity 237 Overall, negative interactions dominated across all treatments, with 195 negative interactions 238 observed against 100 positive ones (i.e. uni-directional RIIs of individual isolates). 239 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 12 Despite the prevalence of negative interactions, and within pair variation (which accounted for 25% 240 of variation within the model: S1), resource environment significantly affected the nature of species 241 interaction (Lmer on RII, media main effect, χ2=11.065, p=0.004, Figure 3, S4). 242 243 Figure 3: Boxplot showing the RII index of individual isolates as a function of resource 244 environment (minimal media, TSB, Soil Wash) and geographic proximity (local, general, different) 245 averaged across triplicates (n = 6 pairs and 12 isolates per unique treatment combination). Size of 246 the data points depicts standard deviation around mean per isolate. 247 Isolates grown in TSB had the most negative interaction sign (estimated mean, [confidence 248 intervals], -0.44, [-0.60, -0.27]), whilst those grown in minimal media were the least negative, with 249 interactions on average being neutral (-0.07 [-0.24, 0.09]). These resource environments differed 250 most in terms of their resource abundance and complexity, which was reflected in a greater 251 difference in RII index (post-hoc comparison: t=3.383, p=0.005, S2). This is in accordance with our 252 predictions of more restricted resources favouring more positive interactions, and complex media 253 allowing clones to invest in interference mechanisms and/or higher densities, thereby increasing 254 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 13 competition. We therefore tested whether isolates with a greater propensity for growth in 255 monoculture showed more negative RII indices in pairwise co-culture, and found this indeed to be 256 the case (linear model, density change main effect, estimated effect [Standard error], -0.056 257 [0.013], F1,94=19.794, P<0.001, S3) independent of the resource environment (resource x density 258 effect, F = 0.5868, P = 0.558). 259 As predicted, interactions in soil wash tended to be less negative relative to TSB (-0.20 [-0.37, -260 0.04], soil wash-TSB post-hoc contrast: t=-2.184, p=0.088), and were not as positive as those seen 261 in minimal media, though this difference was not significant (soil wash-minimal media post-hoc 262 contrast, t=1.231, p=0.443). 263 We predicted sympatric isolates to interact more competitively than allopatric isolates due to 264 similarities in resource use resulting from phylogenetic similarities. However, geographic proximity 265 did not significantly affect bacterial interactions (Lmer on RII index: effect of media x locality 266 interaction: χ2=3.827, p=0.430, locality main effect, (χ2=0.245 p=0.363)). Subsequent phylogenetic 267 analysis showed that this was most likely due to most competing isolates within a pair belonging to 268 different genera – even in pairs isolated in the same community (Figure 4) – such that 269 phylogenetic diversity was similar across localities. 270 The absence of a phylogenetic signal in response to location may reflect our sampling approach, 271 which had limited ecological differentiation among sites. This may have promoted phylogenetic 272 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 14 clustering through habitat filtering, thereby obscuring location-specific patterns.273 274 Figure 4 Phylogenetic tree showing genus level identification via 16S Sanger sequencing (see 275 Methods) for each isolate. Pairs are made up in a stepwise fashion (e.g. pair 1: isolates 1 and 2; 276 pair 2: isolates 3 and 4 etc). Some isolates could not be identified at the genus level, and for these, 277 distances were calculated by aligning sequences. Isolate 4 is missing due to low quality reads. 278 Reciprocal interactions depend on resource environment with exploitative interactions dominating 279 in soil wash 280 We next classified reciprocal interactions by pairing the RII index of paired isolates as: cooperative 281 (+/+), exploitative (+/-) or competitive (-/-) (see methods). Resource environment significantly 282 affected the likelihood of different interactions (Likelihood ratio test, LR=12.411, p=0.015) 283 Competitive interactions dominated in TSB (60% of total). However, there was a shift in reciprocal 284 interactions in minimal media and soil wash towards mutualistic (+/+) interactions and exploitative 285 interactions (+/-), respectively. (Figure 5). 286 287 288 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 15 289 Figure 5: Plot showing the extracted probabilities predicted by the multinomial analysis with 290 confidence intervals displayed above and below. The plot is facetted by interaction type and media 291 shown on the x-axis. 292 To assess how species interaction strength was affected by resource environment, the absolute 293 interaction index values (Figure 6) were used. This showed significant differences across 294 environments in absolute isolate RII indices (LMER, media main effect, χ2=7.954, p=0.019, Figure 295 6). Interactions were strongest in TSB, where competitive (-/-) interactions prevailed, whilst 296 interactions in soil wash were weakest overall. 297 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 16 298 299 Figure 6: Boxplot showing the absolute values for the mean isolate RII index, given by the 300 average isolate interaction value across the three experimental repetitions. All three levels of 301 locality are combined so that the full data is plotted. Larger values represent stronger interactions. 302

303 We here tested the effect of environment type and geographic proximity on the nature of species 304 interactions using a panel of environmental bacterial taxa isolated from different soil localities. 305 Although we found competitive interactions to dominate, environments significantly affected the 306 nature of species interactions: The most rich and complex environment (TSB) caused interactions 307 to be significantly stronger and more negative, the few mutualistic interactions observed occurred 308 primarily in the minimal medium, and the soil wash led to primarily exploitative and weaker 309 interactions. In contrast to our expectations, interactions were independent of whether paired 310 isolates originated from the same or different community/environment, which was likely due a lack 311 of phylogenetic signal of locality, with phylogenetic diversity of isolates being equal across our 312 locality treatments. 313 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 17 The lack of ecological differentiation among the sampled sites likely led to phylogenetic clustering 314 because of habitat filtering and the reduced overall diversity seen in the sequenced isolates, with 315 the majority being identified as Pseudomonas sp. Though isolates were not paired based on 316 morphology, some sampling bias may have led to phylogenetically similar species as a result of 317 picking colonies which looked morphologically distinct. In addition, the plating of soil communities 318 on King’s Broth (KB) agar plates will have imposed some selective bias on all picked isolates. 319 Consistent with our predictions (11-13), competitive interactions dominated in TSB while the 320 highest occurrence of mutualistic interactions occurred in minimal media. This is likely because the 321 growth of some isolates was significantly constrained in minimal medium, leading to isolates 322 engaging in cross-feeding behaviours when grown in coculture, utilising metabolic waste products 323 of others (16). More generally, we found that an increase in individual growth was associated with 324 more negative RII values (S3). In contrast with previous work (15), our findings suggest that 325 increased growth, rather than interference investment, was the main driver underpinning 326 competitive interactions in our study - with isolate specific trade-offs between investment in 327 monoculture growth and survival in coculture potentially driving the trend seen, similar to that seen 328 in cocultures of S.marcescens and N.capsulatum (21). These findings highlight that species 329 interactions are highly context dependent, and that resource richness and diversity should be 330 considered when studying species interactions in bacterial communities. 331 The soil wash was used to best reflect the isolates’ natural conditions. Consistent with previous 332 work on natural isolates the highest proportion of exploitative interactions (+/-) occurred in this 333 environment (46). Cross-feeding interactions in diverse soil communities are likely to consist of 334 numerous uni-directional relationships, in which individuals attain higher densities through the 335 utilization of by-product metabolites produced by neighbouring taxa. In addition, the type of cross-336 feeding interactions occurring in diverse soil communities are likely multiple uni-directional 337 relationships, whereby individuals may reach higher densities because of by-product metabolite 338 production from another individual (16). The exploitative interactions observed in soil wash 339 communities may therefore arise from such cross-feeding dynamics, where accidental 340 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 18 overproduction or metabolite leakage generates an asymmetric interaction in which one individual 341 benefits from metabolic waste and potentially outcompetes the producing strain. The fact 342 exploitative interactions in soil wash media were the weakest interactions overall is consistent with 343 theory that exploitative interactions may increase stability in bacterial communities by reducing 344 dependency between cooperators, and that weak interactions generally are more likely in natural 345 communities (4, 5, 9) 346 Mutualistic interactions are predicted to be more common between isolates from different localities 347 as inter-site resource differences increase (13, 30, 46, 47). Surprisingly, geographic proximity did 348 not affect either the variation in interaction sign or the likelihood of a certain interaction being 349 present in our study. This can in part be explained by the observed lack in phylogenetic-geographic 350 relationship – there were few closely related pairs in our ‘local isolation’ category (and vice versa 351 with a lack of distantly related pairs in pairs from different sites). Therefore, the similar phylogenetic 352 diversity identified within and between sites likely masked any location effects, and although the 353 sampled scale may ignore some natural complexity, small spatial differences have previously been 354 shown to shape bacteria-phage local adaptation in soil environments (48). 355

356 This study looked at the effect of three different environments of different resource complexity and 357 richness on bacterial pairwise interactions across a panel of diverse soil bacteria. In addition, 358 bacteria were isolated from different sites to test if geographic proximity affected interactions due to 359 resource use overlap. We found that interactions were negative overall and varied across 360 resources environments, but not with isolate locality. 361 Overall, this study demonstrated how a rich and complex laboratory medium leads to high growth 362 rates and strong negative interactions, whilst showing how more simple resources lead to weaker 363 interactions and in cases can lead to cooperation, though rare. Notably, the medium designed to 364 mimic soil conditions was characterized by weak, predominantly exploitative interactions. 365 366 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 19 367 1. Seymour JR, Amin SA, Raina J-B, Stocker R. Zooming in on the phycosphere: the 368 ecological interface for phytoplankton–bacteria relationships. Nature Microbiology. 369 2017;2(7):17065. 370 2. Hansen SK, Rainey PB, Haagensen JAJ, Molin S. Evolution of species interactions 371 in a biofilm community. Nature. 2007;445(7127):533-6. 372 3. Romdhane S, Spor A, Aubert J, Bru D, Breuil M-C, Hallin S, et al. Unraveling 373 negative biotic interactions determining soil microbial community assembly and 374 functioning. The ISME Journal. 2022;16(1):296-306. 375 4. Palmer JD, Foster KR. Bacterial species rarely work together. Science. 376 2022;376(6593):581-2. 377 5. Coyte KZ, Schluter J, Foster KR. The ecology of the microbiome: networks, 378 competition, and stability. Science. 2015;350(6261):663-6. 379 6. Cavaliere M, Feng S, Soyer OS, Jiménez JI. Cooperation in microbial communities 380 and their biotechnological applications. Environmental microbiology. 2017;19(8):2949-63. 381 7. Tan C, Kalhoro MT, Faqir Y, Ma J, Osei MD, Khaliq G. Climate-resilient microbial 382 biotechnology: A perspective on sustainable agriculture. Sustainability. 2022;14(9):5574. 383 8. Smith P , Schuster M. Public goods and cheating in microbes. Current Biology. 384 2019;29(11):R442-R7. 385 9. Rivett DW, Scheuerl T, Culbert CT, Mombrikotb SB, Johnstone E, Barraclough TG, 386 et al. Resource-dependent attenuation of species interactions during bacterial succession. 387 The ISME Journal. 2016;10(9):2259-68. 388 10. Deng Y-J, Wang SY . Synergistic growth in bacteria depends on substrate 389 complexity. Journal of Microbiology. 2016;54(1):23-30. 390 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 20 11. Hibbing ME, Fuqua C, Parsek MR, Peterson SB. Bacterial competition: surviving 391 and thriving in the microbial jungle. Nature reviews microbiology. 2010;8(1):15-25. 392 12. Qiao Y , Huang Q, Guo H, Qi M, Zhang H, Xu Q, et al. Nutrient status changes 393 bacterial interactions in a synthetic community. Applied and Environmental Microbiology. 394 2024;90(1):e01566-23. 395 13. Freilich S, Zarecki R, Eilam O, Segal ES, Henry CS, Kupiec M, et al. Competitive 396 and cooperative metabolic interactions in bacterial communities. Nature Communications. 397 2011;2(1):589. 398 14. Weiss AS, Burrichter AG, Durai Raj AC, von Strempel A, Meng C, Kleigrewe K, et 399 al. In vitro interaction network of a synthetic gut bacterial community. The ISME journal. 400 2022;16(4):1095-109. 401 15. Yang C, Dong Y , Friman VP , Jousset A, Wei Z, Xu Y , et al. Carbon resource 402 richness shapes bacterial competitive interactions by alleviating growth‐ antibiosis trade‐ off. 403 Functional Ecology. 2019;33(5):868-75. 404 16. D'Souza G, Shitut S, Preussger D, Yousif G, Waschina S, Kost C. Ecology and 405 evolution of metabolic cross-feeding interactions in bacteria. Natural product reports. 406 2018;35(5):455-88. 407 17. Fiege K, Asbun AA, Boeren S, Engelmann JC, Villanueva L. Membrane changes 408 during syntrophic interactions of an archaeal-bacterial coculture. BMC microbiology. 2025. 409 18. Sela-Adler M, Ronen Z, Herut B, Antler G, Vigderovich H, Eckert W, et al. Co-410 existence of methanogenesis and sulfate reduction with common substrates in sulfate-rich 411 estuarine sediments. Frontiers in microbiology. 2017;8:766. 412 19. Lawrence D, Fiegna F, Behrends V, Bundy JG, Phillimore AB, Bell T, et al. Species 413 interactions alter evolutionary responses to a novel environment. PLoS biology. 414 2012;10(5):e1001330. 415 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 21 20. Nathaniel Holland J, DeAngelis DL. Consumer ‐ resource theory predicts dynamic 416 transitions between outcomes of interspecific interactions. Ecology Letters. 417 2009;12(12):1357-66. 418 21. Pekkonen M, Ketola T, Laakso JT. Resource availability and competition shape the 419 evolution of survival and growth ability in a bacterial community. PLoS One. 420 2013;8(9):e76471. 421 22. Hammarlund SP, Chacón JM, Harcombe WR. A shared limiting resource leads to 422 competitive exclusion in a cross‐ feeding system. Environmental Microbiology. 423 2019;21(2):759-71. 424 23. Piccardi P, Vessman B, Mitri S. Toxicity drives facilitation between 4 bacterial 425 species. Proceedings of the National Academy of Sciences. 2019;116(32):15979-84. 426 24. Goberna M, García C, Verdú M. A role for biotic filtering in driving phylogenetic 427 clustering in soil bacterial communities. Wiley Online Library; 2014. p. 1346-55. 428 25. Tóth AB, Alroy J, Lyons SK, Allen AP. Effects of phylogenetic distance, niche 429 overlap and habitat alteration on spatial co-occurrence patterns in Neotropical bats and 430 birds. Proceedings of the Royal Society B: Biological Sciences. 2025;292(2051). 431 26. Redford AJ, Bowers RM, Knight R, Linhart Y , Fierer N. The ecology of the 432 phyllosphere: geographic and phylogenetic variability in the distribution of bacteria on tree 433 leaves. Environmental microbiology. 2010;12(11):2885-93. 434 27. Horner-Devine MC, Bohannan BJ. Phylogenetic clustering and overdispersion in 435 bacterial communities. Ecology. 2006;87(sp7):S100-S8. 436 28. Fahimipour AK, Gross T. Mapping the bacterial metabolic niche space. Nature 437 communications. 2020;11(1):4887. 438 29. Butait ė E, Kramer J, Kümmerli R. Local adaptation, geographical distance and 439 phylogenetic relatedness: Assessing the drivers of siderophore‐ mediated social 440 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 22 interactions in natural bacterial communities. Journal of Evolutionary Biology. 441 2021;34(8):1266-78. 442 30. Giri S, Oña L, Waschina S, Shitut S, Yousif G, Kaleta C, et al. Metabolic dissimilarity 443 determines the establishment of cross-feeding interactions in bacteria. Current Biology. 444 2021;31(24):5547-57. e6. 445 31. Hahn MW, Koll U, Schmidt J. Isolation and Cultivation of Bacteria. In: Hurst CJ, 446 editor. The Structure and Function of Aquatic Microbial Communities. Cham: Springer 447 International Publishing; 2019. p. 313-51. 448 32. Overmann J, Abt B, Sikorski J. Present and Future of Culturing Bacteria. Annual 449 Review of Microbiology. 2017;71(Volume 71, 2017):711-30. 450 33. Mori AS, Furukawa T, Sasaki T. Response diversity determines the resilience of 451 ecosystems to environmental change. Biological reviews. 2013;88(2):349-64. 452 34. Russel J, Røder HL, Madsen JS, Burmølle M, Sørensen SJ. Antagonism correlates 453 with metabolic similarity in diverse bacteria. Proceedings of the National Academy of 454 Sciences. 2017;114(40):10684-8. 455 35. F H. DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) 456 Regression Models. R Packag version 0202018. 457 36. Lenth RV. Least-Squares Means: The R Package lsmeans. Journal of Statistical 458 Software. 2016;69(1):1 - 33. 459 37. Wickham H. Ggplot2 : elegant graphics for data analysis. New York, NY: Springer 460 Science+Business Media, LLC; 2016. xvi, 260 pages p. 461 38. Venables WN, Ripley BD. Modern applied statistics with S: Springer Science & 462 Business Media; 2013. 463 39. Meroz N, Livny T, Friedman J. Quantifying microbial interactions: concepts, caveats, 464 and applications. Current Opinion in Microbiology. 2024;80:102511. 465 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint 23 40. Bates D, Maechler M, Bolker B, Walker S, Christensen RHB, Singmann H, et al. 466 Package ‘lme4’. convergence. 2015;12(1):2. 467 41. Ripley B, Venables W, Ripley MB. Package ‘nnet’. R package version. 2016;7(3-468 12):700. 469 42. Lüdecke D. ggeffects: Tidy data frames of marginal effects from regression models. 470 Journal of open source software. 2018;3(26):772. 471 43. Chao K-H, Barton K, Palmer S, Lanfear R. sangeranalyseR: simple and interactive 472 processing of sanger sequencing data in R. Genome Biology and Evolution. 473 2021;13(3):evab028. 474 44. Paradis E, Schliep K. ape 5.0: an environment for modern phylogenetics and 475 evolutionary analyses in R. Bioinformatics. 2019;35(3):526-8. 476 45. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJA, Holmes SP. DADA2: 477 High-resolution sample inference from Illumina amplicon data. Nature methods. 478 2016;13(7):581-3. 479 46. Foster KR, Bell T . Competition, not cooperation, dominates interactions among 480 culturable microbial species. Current biology. 2012;22(19):1845-50. 481 47. Schlechter RO, Kear EJ, Bernach M, Remus DM, Remus-Emsermann MN. 482 Metabolic resource overlap impacts competition among phyllosphere bacteria. The ISME 483 Journal. 2023;17(9):1445-54. 484 48. Vos M, Birkett PJ, Birch E, Griffiths RI, Buckling A. Local adaptation of 485 bacteriophages to their bacterial hosts in soil. Science. 2009;325(5942):833-. 486 487 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 31, 2026. ; https://doi.org/10.64898/2026.03.31.715556doi: bioRxiv preprint