Chromosomal capture of beneficial genes drives plasmids towards ecological redundancy

18 Plasmids are a ubiquitous feature of bacterial genomes, but the evoluAonary forces driving 19 genes to become associated with plasmids are poorly understood. To address this problem, 20 we compared the fitness effects of chromosomal and plasmid genes in the plant symbiont 21 Rhizobium leguminosarum. Here we show that plasmids are depleted in beneficial genes 22 compared to the chromosome, and this effect is stronger for ancient plasmids compared to 23 recently acquired plasmids. These findings support the hypothesis that evoluAon drives 24 beneficial genes to become localized to the bacterial chromosome, resulAng in a gradual 25 decay in the ecological value of plasmids. These findings quesAon the ecological importance 26 of plasmids and highlight the challenge of understanding how plasmids persist over the long 27 term. 28 29 Main text 30 Bacterial genomes are made up of chromosomes and plasmids that replicate independently 31 of the chromosome. Genes are conAnuously transferred between plasmids and 32 chromosomes, and uncovering the processes that drive genes and phenotypes to be 33 associated with plasmids as opposed to bacterial chromosomes is a fundamental challenge 34 in microbial ecology and evoluAon(1-8). 35 36 The dominant view in microbiology is that plasmids play a key role in bacterial adaptaAon 37 through the horizontal transfer of genes that are beneficial in defined ecological niches(9-38 13), such as genes associated with anAbioAc resistance, pathogen virulence, or novel 39 metabolic pathways(1, 8, 14-17). However, classic evoluAonary models that allow genes to 40 move between plasmids and the chromosome predict that beneficial genes should become 41 associated with chromosomes, as opposed to plasmids, quesAoning the role of plasmids in 42 bacterial adaptaAon (2, 18). It has been challenging to reconcile these two views of plasmids 43 (6, 8, 11, 12, 19, 20) because the relaAve ecological and evoluAonary importance of plasmid 44 genes remains poorly understood beyond the paradigmaAc examples highlighted above. 45 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 2 Here we address this problem by systemaAcally measuring the impact of plasmid and 1 chromosomal genes on bacterial fitness in the plant symbiont Rhizobium leguminosarum . 2 Wheatley et al(21) used transposon inserAons (22) to systemaAcally mutagenize the 3 genome of a strain of R. leguminosarum carrying a chromosome and 6 plasmids(17). 4 PopulaAons of pooled inserAon mutants were then assayed by deep sequencing under 5 condiAons that recapitulate the ecology of Rhizobium (23), including growth in the 6 rhizosphere, root colonisaAon, nodulaAon, and bacteroid formaAon (Figure 1 A,B). The use 7 of fitness assays under natural condiAons is a key feature of this data set, given that plasmids 8 are predicted to carry ecologically relevant genes whose effects may be missed in standard 9 lab-culture based measures of bacterial fitness. This experiment uncovered 603 unique 10 genes that were beneficial in either a single niche (specialist genes) or across mulAple niches 11 (generalist genes) (Figure 1 B,C). 12 13 14 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 3 1 Figure 1: Iden5fying beneficial genes by Tn-Seq. (A) SchemaAc of a transposon inserAon 2 sequencing experiment. First, a mutant library is constructed using transposon inserAon to 3 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 4 inacAvate genes on a genome-wide scale. When a selecAon pressure is applied to the 1 populaAon, mutants change in frequency in the populaAon depending on the contribuAon 2 of their mutated gene to fitness under that condiAon. The resulAng populaAons, including 3 the input library, are then sequenced and mapped back against the genome to determine 4 the posiAon of and frequency of transposon inserAon mutants across the genome. Mutants 5 of genes which are important for fitness will fall out of frequency of the populaAon, allowing 6 their idenAficaAon as beneficial genes. (B) Summary of the transposon inserAon sequencing 7 experiment previously conducted by Wheatley et al. (21) in which a R. leguminosarum 8 mutant library was assayed across mulAple stages of symbiosis: growth in the rhizosphere, 9 root colonisaAon, nodule formaAon, and bacteroid formaAon. The table indicates the 10 number of genes which were idenAfied to be beneficial across the corresponding niches. 11 Blue blocks indicate the number of genes beneficial in single niches, and purple blocks 12 indicate the number of genes beneficial in mulAple niches. This figure was made in 13 biorender. (C) The seven replicons of the Rhizobium genome are displayed on the outer 14 circle of this circos visualizaAon(24). Genes that were beneficial in a single niche (green inner 15 band) or across mulAple niches (generalist genes; blue inner band) and are marked for each 16 replicon. Ji\er was added along the y-axis (height) posiAon of the circles to aid visualizaAon 17 of genes in close proximity. 18 19 Plasmids are depleted in beneficial genes 20 21 To understand the benefits of plasmid and chromosomal genes, we calculated the fracAon of 22 plasmid and chromosomal genes that were beneficial in each niche. Crucially, the proporAon 23 of plasmid genes with beneficial effects on fitness was low relaAve to the chromosome in all 24 niches, challenging the ecological importance of plasmids (Figure 2A). 25 26 However, a limitaAon of this analysis is that it treats plasmid genes as a collecAve. If plasmids 27 are key drivers of niche adaptaAon, then individual plasmids might be associated with genes 28 involved in specializaAon on disAnct niches. Consistent with this idea, we found two clear 29 examples of niche-associated plasmids. Plasmid pRL10 carries genes that play important 30 roles in the establishment of symbioAc interacAons with legumes, including nitrogen fixaAon 31 (17, 21). As expected, this plasmid was associated with genes that were beneficial during 32 nodulaAon and bacteroid formaAon. Second, plasmid pRL7 was associated with genes that 33 were beneficial across all of the niches associated with plants, including root colonizaAon. 34 Although these examples highlight the associaAon between plasmids and niches, it is 35 important to emphasize that plasmids were not enriched in niche-adapAve genes compared 36 to the chromosome, except for a single case of genes involved in root specializaAon on 37 plasmid pRL7. 38 39 40 41 42 43 44 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 5 1 Figure 2: Plasmid are depleted in beneficial genes. Plots show comparisons of the 2 prevalence of beneficial genes (ie beneficial genes/total genes) between the chromosome 3 and all plasmid genes (A) and between individual replicons (B). We compared the 4 proporAons of beneficial genes on plasmids and the chromosomes using a normal 5 approximaAon to the binomial distribuAon. All comparisons between plasmids and the 6 chromosome in A were staAsAcally significant under a two-tailed null hypothesis with 7 P<1x10-10. In B we tested for an increased prevalence of beneficial genes on plasmids 8 compared to the chromosome. Plasmid pRL7 was enriched in root adapAve genes 9 compared to the chromosome (P one -tailed=.0017, **). 10 11 Plasmids are associated with niche specialist genes 12 13 If evoluAonary processes drive beneficial genes to become localized to the chromosome, 14 genes that are under strong selecAon should be more likely to be associated with the 15 chromosome compared to genes that are under weak selecAon(18). To test this predicAon, 16 we compared the distribuAon of genes that were beneficial in a single niche (specialist 17 genes) with those that were beneficial across mulAple niches (generalist genes). The 18 underlying assumpAon of this test is that genes that are beneficial in a single niche are under 19 weak selecAon compared to genes that are beneficial across mulAple niches when selecAon 20 is considered across the enAre life cycle of Rhizobium. 21 22 Overall, plasmids were not enriched in specialist genes compared to the chromosome 23 (Figure 3A). However, plasmids pRL10 and pRL7 were enriched in specialist genes, reflecAng 24 the roles that these plasmids play in interacAons between Rhizobium and plants (Figure 3C). 25 The overall lack of specialist genes on plasmids was driven by the fact that the remaining 26 plasmids were depleted in niche specialist genes, and this depleAon was most obvious for 27 plasmids pRL11 and pRL12. In contrast, generalist genes that were beneficial across mulAple 28 niches were strongly associated with the chromosome (Figure 3B). None of the plasmid 29 replicons were enriched in generalist genes, and the depleAon of generalist genes was 30 parAcularly clear for plasmids pRL9, pRL11 and pRL12 (Figure 3D). 31 32 To further test the hypothesis that genes under strong selecAon become associated with 33 chromosomes, we treated the number of niches where genes were beneficial as an ordinal 34 variable (i.e. 1-4 niches) as opposed to a binary variable (i.e. specialist or generalist). As 35 expected, plasmids were depleted in genes that were beneficial across mulAple niches 36 compared to the chromosome (Figure 3E). An alternaAve way to visualize this result is to 37 compare the prevalence of narrow range generalist genes that were beneficial in 2 niches 38 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 6 with broad range generalist genes that were beneficial in 3 or 4 niches (Figure 3F). Almost all 1 of the generalist genes carried by plasmids were narrow range, while broad and narrow 2 range generalist genes were equally represented on the chromosome. 3 4 5 6 7 8 Figure 3: Plasmids are associated with niche specialist beneficial genes. Bar charts show 9 the observed and expected number of specialist (A,C) and generalist (B,D) genes across the 10 genome. Panels A and B show a comparison of all plasmid genes with the chromosome, and 11 Panels C and D show individual plasmid replicons, with observed gene counts shown in 12 green and expected gene counts shown in dark blue. Expected gene numbers were 13 calculated based on the number of genes on each replicon under the null hypothesis that 14 the prevalence of beneficial genes is equal for all replicons. We tested for beneficial gene 15 enrichment using two-tailed binomial tests comparing all plasmids and the chromosome 16 (Panel A,C) or individual plasmid replicons (Panel B,D). StaAsAcal tests for individual 17 replicons were corrected for mulAple tesAng using the Bonferonni correcAon. Panel E shows 18 the proporAon of beneficial genes associated with plasmids as a funcAon of the number of 19 niches where the gene was beneficial. The number of plasmid associated beneficial genes 20 are shown and we tested the null hypothesis that beneficial genes are evenly distributed 21 across the genome using two-tailed binomial tests. Panel F shows the proporAon of 22 generalist genes that increased fitness in a narrow range (2 niches) or a broad range (3 or 4 23 niches) of niches for plasmids and the chromosome. We tested for a difference in the 24 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 7 proporAon of narrow and broad range generalist genes usign a normal approximaAon to the 1 binomial distribuAon. Significance: n.s: no significant enrichment; *, P<.05; **, P<.01; ***, 2 P<.001;. 3 4 Plasmids lose beneficial genes over 5me 5 If selecAon favours the movement of beneficial genes from plasmids to the chromosome, 6 then recently acquired plasmids should be rich in beneficial genes compared to ancient 7 plasmids. Plasmids pRL9, pRL11 and pRL12 lack moAlity systems and have a nucleoAde 8 composiAon that matches the chromosome, suggesAng that they were acquired by 9 Rhizobium in the distant past(17). The remaining plasmids (pRL7, pRL8, pRL10) have 10 divergent nucleoAde composiAon from the chromosome and plasmid mobilizaAon systems 11 (pRL7 and pRL8), implying that they have been more recently acquired. To test this 12 hypothesis, we compared the prevalence of all beneficial genes between recently acquired 13 and ancient plasmids (Figure 4). We did not disAnguish between specialist and generalist 14 genes in this analysis, due to the fact that plasmids carried few generalist genes that were 15 typically beneficial in only 2 niches (Figure 3E,F). Beneficial genes were over-represented on 16 recently acquired plasmids, whereas beneficial genes were strongly depleted from ancient 17 plasmids, suggesAng that plasmids become gradually depleted in beneficial genes over Ame. 18 19 20 21 22 Figure 4: Ancient plasmids are depleted in beneficial genes. Bar charts show the expected 23 and observed number of beneficial genes for recently acquired (pRL7,pRL8, pRL10) and 24 ancient (pRL9,pRL11,pRL12) plasmids. Expected gene numbers were calculated based on the 25 number of genes on each replicon under the null hypothesis that the prevalence of 26 beneficial genes is equal across plasmids. We tested for significant deviaAons from expected 27 gene counts using a two-tailed binomial test, and both P values were highly significant 28 (P<6x10-4). 29 30 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 8

1 Plasmids are a ubiquitous component of bacterial genomes, but their role in adaptaAon 2 remains unclear. Classic evoluAonary models allowing movement of genes between 3 chromosomes and plasmids predict that, over Ame, beneficial genes will become localized to 4 the chromosome(2). Consistent with this model, we found that plasmids were depleted in 5 beneficial genes compared to the chromosome (Figure 2A), because genes that were 6 beneficial across mulAple ecological niches were strongly localized to the chromosome 7 (Figure 3). If the chromosome effecAvely captures beneficial genes, then we would expect 8 plasmids to undergo a process of gradual ecological decay due to the loss of beneficial 9 genes. Consistent with this idea, we found that ancient plasmids were depleted in beneficial 10 genes compared to recently acquired plasmids (Figure 4). Our results suggest that plasmids 11 acquisiAon provides bacteria with beneficial genes, but the movement of beneficial genes to 12 the chromosome causes plasmids to degrade towards ecological redundancy, emphasizing 13 the challenge of understanding how plasmids can persist over the long term (5, 6, 25-29). 14 15 The paradigm that plasmids play a key role in adaptaAon by providing bacteria with genes 16 that are beneficial in specific ecological niches is deeply ingrained in microbiology(9-13). As 17 expected from this paradigm, we found that plasmids were associated with genes that 18 increased fitness in specific ecological niches (Figure 2B). One of the key insights from our 19 study is that this associaAon arises because evoluAon drives strongly beneficial genes, such 20 as those that increase fitness across mulAple niches, to become localized to the 21 chromosome, leaving plasmids associated with niche specialist genes (Figure 3). We argue 22 that this link between plasmid degeneraAon and niche specializaAon reconciles the 23 adaptaAonist view of plasmids that has emerged from empirical studies with evoluAonary 24 models that predict the degeneraAon of plasmids. 25 26 Many of the most important forms of anAbioAc resistance have been driven by the 27 acquisiAon of plasmids carrying anAbioAc resistance genes (14, 30). Our findings predict that 28 the strong selecAve pressures caused by the conAnued large-scale use of anAbioAcs will 29 stabilize resistance by acceleraAng the integraAon of resistance genes into the chromosomes 30 of pathogenic bacteria, as has already been observed for some resistance genes(31-33). 31 32

33 The supplementary data (Dataset S01 and Table S1-S8) was downloaded from Wheatley et al 34 (21) where a large-scale transposon inserAon sequencing experiment was conducted to 35 idenAfy genes required in R. leguminosarum bv. viciae 3841 (Rlv3841) to engage in symbiosis 36 with the legume host pea (Pisum saDvum). This dataset listed the R. leguminosarum genes 37 predicted to be required for fitness across four stages of symbiosis: (1) growth in the 38 rhizosphere, (2) root colonisaAon, (3) nodulaAon, and (4) bacteroid formaAon. In Wheatley 39 et al (21), a Hidden Markov Model was applied to classify genes into one of four state 40 classificaAons based on their read-mapping staAsAcs: essenAal (ES; no or very few inserAons, 41 i.e. inserAons are not tolerated), defecAve (DE; significantly fewer inserAon read counts 42 along a significant consecuAve stretch of inserAon sites, i.e. inserAon mutaAons impair 43 growth), advantaged (AD; significantly higher inserAon read counts along a significant 44 consecuAve stretch of inserAon sites, i.e. inserAon mutaAon enhances fitness), and neutral 45 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 9 (NE; within the boundaries of a mean parameter of inserAon read counts, inserAon mutaAon 1 has a neutral impact on fitness). To test the validity of their tn-seq experiment, Wheatley et 2 al tested the roles of 15 genes in follow-up experiments using independently constructed 3 mutants. Only a single one of these mutants did not recapitulate its predicted phenotype 4 inferred from the sequencing of pooled populaAons of transposon mutants. 5 6 For our analysis, we used the gene lists defined in Wheatley et al (21) as being required for 7 engaging in symbiosis (Figure 1B) which are composed of genes which were all identified 8 with a NE classification in the input library and either an ES or DE classification in at least 9 one of the symbiosis output libraries (rhizosphere growth, root colonisation, nodulation or 10 bacteroid formation). As such, we are analysing genes that can be considered beneficial 11 genes for plant-associated growth and symbiosis, as their mutation has a negative impact on 12 fitness. We used these previously defined lists (21) with the additional downstream 13 removal of genes with potential gene duplications in the Rlv3841 genome from the 14 analysis(17, 34). This was used to calculate a total number of genes on each replicon as the 15 denominator for the enrichment analysis, by subtracting potential gene duplications from 16 the input library from the total gene numbers on the replicons. This had minimal impact on 17 the output results. 18

19 1. A. E. Dewar et al., Plasmids do not consistently stabilize cooperaAon across bacteria 20 but may promote broad pathogen host-range. Nature Ecology & EvoluDon 5, 1624-+ 21 (2021). 22 2. C. Bergstrom, M. Lipsitch, B. Levin, Natural selecAon, infecAous transfer and the 23 existence condiAons for bacterial plasmids. GENETICS 155, 1505-1519 (2000). 24 3. F. Stewart, B.Levin, PopulaAon biology of bacterial plasmids – a priori condiAons 25 existence of conjugaAonally transmi\ed factors. GENETICS 87, 209-228 (1977). 26 4. J. Hall, D. Williams, S. Paterson, E. Harrison, M. Brockhurst, PosiAve selecAon inhibits 27 gene mobilizaAon and transfer in soil bacterial communiAes. NATURE ECOLOGY & 28 EVOLUTION 1, 1348-1353 (2017). 29 5. J. Hall, A. Wood, E. Harrison, M. Brockhurst, Source-sink plasmid transfer dynamics 30 maintain gene mobility in soil bacterial communiAes. PROCEEDINGS OF THE 31 NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA 113, 8260-32 8265 (2016). 33 6. M. Brockhurst, E. Harrison, Ecological and evoluAonary soluAons to the plasmid 34 paradox. TRENDS IN MICROBIOLOGY 30, 534-543 (2022). 35 7. J. Rodriguez-Beltran, J. DelaFuente, R. Leon-Sampedro, R. C. MacLean, A. San Millan, 36 Beyond horizontal gene transfer: the role of plasmids in bacterial evoluAon. Nature 37 Reviews Microbiology 19, 347-359 (2021). 38 8. M. Brockhurst et al., The Ecology and EvoluAon of Pangenomes. CURRENT BIOLOGY 39 29, R1094-R1103 (2019). 40 9. A. Norman, L. Hansen, S. Sorensen, ConjugaAve plasmids: vessels of the communal 41 gene pool. PHILOSOPHICAL TRANSACTIONS OF THE ROYAL SOCIETY B-BIOLOGICAL 42 SCIENCES 364, 2275-2289 (2009). 43 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 10 10. S. Sorensen, M. Bailey, L. Hansen, N. Kroer, S. Wuertz, Studying plasmid horizontal 1 transfer in situ: A criAcal review. NATURE REVIEWS MICROBIOLOGY 3, 700-710 2 (2005). 3 11. A. Beavan, M. Sananes, J. Mcinerney, ConAngency, repeatability, and predictability in 4 the evoluAon of a prokaryoAc pangenome. PROCEEDINGS OF THE NATIONAL 5 ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA 121, (2024). 6 12. F. Whelan, R. Hall, J. McInerney, Evidence for SelecAon in the Abundant Accessory 7 Gene Content of a Prokaryote Pangenome. MOLECULAR BIOLOGY AND EVOLUTION 8 38, 3697-3708 (2021). 9 13. A. McNally et al., Combined Analysis of VariaAon in Core, Accessory and Regulatory 10 Genome Regions Provides a Super-ResoluAon View into the EvoluAon of Bacterial 11 PopulaAons. Plos GeneDcs 12, (2016). 12 14. S. R. Partridge, S. M. Kwong, N. Firth, S. O. Jensen, Mobile GeneAc Elements 13 Associated with AnAmicrobial Resistance. Clinical Microbiology Reviews 31, (2018). 14 15. A. Greated, L. Lambertsen, P . Williams, C. Thomas, Complete sequence of the IncP-9 15 TOL plasmid pWW0 from Pseudomonas puDda . ENVIRONMENTAL MICROBIOLOGY 4, 16 856-871 (2002). 17 16. V. González et al., The mosaic structure of the symbioAc plasmid of Rhizobium etli 18 CFN42 and its relaAon to other symbioAc genome compartments -: art. no. R36. 19 GENOME BIOLOGY 4, (2003). 20 17. J. Young et al., The genome of Rhizobium leguminosarumhas recognizable core and 21 accessory components. GENOME BIOLOGY 7, (2006). 22 18. S. LehAnen, J. Huisman, S. Bonhoeffer, EvoluAonary mechanisms that determine 23 which bacterial genes are carried on plasmids. EVOLUTION LETTERS 5, 290-301 24 (2021). 25 19. A. Dewar, C. Hao, L. Belcher, M. Ghoul, S. West, Bacterial lifestyle shapes 26 pangenomes. PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE 27 UNITED STATES OF AMERICA 121, (2024). 28 20. J. McInerney, A. McNally, M. O'Connell, Why prokaryotes have pangenomes. NATURE 29 MICROBIOLOGY 2, (2017). 30 21. R. Wheatley et al., Lifestyle adaptaAons of Rhizobium from rhizosphere to symbiosis. 31 PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF 32 AMERICA 117, 23823-23834 (2020). 33 22. T. van Opijnen, K. Bodi, A. Camilli, Tn-seq: high-throughput parallel sequencing for 34 fitness and geneAc interacAon studies in microorganisms. NATURE METHODS 6, 767-35 U721 (2009). 36 23. G. Oldroyd, J. Murray, P . Poole, J. Downie, in ANNUAL REVIEW OF GENETICS, VOL 45. 37 (2011), vol. 45, pp. 119-144. 38 24. M. Krzywinski et al., Circos: An informaAon aestheAc for comparaAve genomics. 39 GENOME RESEARCH 19, 1639-1645 (2009). 40 25. W. Loqie-Eaton et al., Compensatory mutaAons improve general permissiveness to 41 anAbioAc resistance plasmids. NATURE ECOLOGY & EVOLUTION 1, 1354-1363 (2017). 42 26. A. San Millan et al., PosiAve selecAon and compensatory adaptaAon interact to 43 stabilize non-transmissible plasmids. NATURE COMMUNICATIONS 5, (2014). 44 27. J. Hall et al., Plasmid fitness costs are caused by specific geneAc conflicts enabling 45 resoluAon by compensatory mutaAon. PLOS BIOLOGY 19, (2021). 46 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 11 28. E. Harrison, D. Guymer, A. Spiers, S. Paterson, M. Brockhurst, Parallel Compensatory 1 EvoluAon Stabilizes Plasmids across the ParasiAsm-Mutualism ConAnuum. CURRENT 2 BIOLOGY 25, 2034-2039 (2015). 3 29. E. Harrison et al., Bacteriophages Limit the Existence CondiAons for ConjugaAve 4 Plasmids. Mbio 6, (2015). 5 30. R. C. MacLean, A. San Millan, The evoluAon of anAbioAc resistance. Science 365, 6 1082-1083 (2019). 7 31. C. Shen et al., Dynamics of mcr-1 prevalence and mcr-1-posiAve Escherichia coliaqer 8 the cessaAon of colisAn use as a feed addiAve for animals in China: a prospecAve 9 cross-secAonal and whole genome sequencing-based molecular epidemiological 10 study. LANCET MICROBE 1, E34-E43 (2020). 11 32. R. Li et al., GeneAc basis of chromosomally-encoded mcr-1 gene. INTERNATIONAL 12 JOURNAL OF ANTIMICROBIAL AGENTS 51, 578-585 (2018). 13 33. W. Matlock et al., E. coli phylogeny drives co-amoxiclav resistance through variable 14 expression of blaTEM-1. Pre-print bioRxiv doi:10.1101/2024.08.12.607562 (2024). 15 34. B. Perry, M. Akter, C. Yost, The Use of Transposon InserAon Sequencing to Interrogate 16 the Core FuncAonal Genome of the Legume Symbiont Rhizobium leguminosarum. 17 FRONTIERS IN MICROBIOLOGY 7, (2016). 18 19

20 We thank Professor Philip Poole for his comments on an early version of this data analysis, 21 and we thank Stu West, Liam Shaw, Michael Brockhurst and Alvaro San Millan for feedback 22 on a draq manuscript. Figures within this publicaAon were created using Biorender.com 23 (Figure 1A, Figure 1B). 24 25 Funding 26 R.M.W. is supported by a Vice-Chancellor's Illuminate Fellowship (Queen’s University Belfast) 27 and part of this research was conducted while visiAng the Okinawa InsAtute of Science and 28 Technology (OIST) through the TheoreAcal Sciences VisiAng Program (TSVP). 29 30 C.L. was supported by a Marie Skłodowska-Curie AcAons Postdoctoral Fellowship from the 31 UKRI Horizon Europe Guarantee program (grant agreement no. EP/Y029585/1) 32 33 R.C.M was supported by UKRI FronAers Grant (EP/Y031067/1). 34 35 Author contribu5ons: (CREDIT system) 36 Conceptualization: RCM, RMW 37 Methodology: RCM,RMW,CL 38 Investigation: RCM, RMW 39 Visualization: CL,RMW,RCM 40 Funding acquisition: RCM,RW,CL 41 Writing – original draft: RCM 42 Writing – review & editing: RCM, RW,CL 43 44 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint 12 Compe5ng interests: The authors declare no compeAng interests. 1 Data and materials availability: This study used publicly available datasets downloaded from 2 Wheatley et al (21) (Dataset S01 and Table S1-S8).Data sets used in this analysis are given in 3 Supplementary data file 1, and raw data for figures in supplementary data file 2. 4 5 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted January 24, 2025. ; https://doi.org/10.1101/2025.01.21.634075doi: bioRxiv preprint