Inferring antibiotic resistance selection in the environment can be confounded by correlations between resistance genes and unrelated functional traits

11 Antimicrobial resistance (AMR) is a silent pandemic that is coupled with other crises such as 12 climate change in the polycrisis humanity is facing. One of the key questions is whether 13 antibiotic resistance genes (ARGs) are selected for at the low antibiotic concentrations typical 14 for most environments. Many studies have observed changes in the relative abundance of 15 ARGs from one environmental compartment to the next, e.g. from wastewater treatment 16 plant influent to effluent. Fewer studies have directly tested for selection by incubating 17 environmental samples in mesocosms or laboratory models at different concentrations of 18 antibiotics to infer minimal selective concentrations. We developed a mathematical model to 19 demonstrate that these studies can be confounded by shifts in the microbial community 20 composition that occur when a microbiome is transported from one environmental 21 compartment to another or when incubated under different conditions. Such community 22 shifts will confound tests of selection when there is an association between carriage of ARGs 23 and other functional traits . As an example, we show that there is a phylum-dependent 24 association between the number of ARGs and the number of ribosomal RNA genes, which are 25 both higher in fast growing, copiotrophic bacteria. We then show that specific growth rate or 26 nutrient concentration upshifts increased the proportion of copiotrophs in the community 27 and thus the relative abundance of ARGs. This result generalizes to community shifts for other 28 reasons if there is some association between ARGs and ecological niches. Therefore, most 29 studies of selection for ARGs in the environment are confounded. Solutions are proposed. 30 Highlights 31 • ARG copy numbers are correlated with 16S rRNA gene copy numbers 32 • High rRNA gene numbers are typical for copiotrophs, which have high growth rates 33 • Hence, copiotrophs tend to have higher ARG numbers 34 • Changes in environmental conditions that increase copiotrophs increase ARGs 35 • Thus, ARGs can increase without selection due to shifts in community composition 36

Antibiotic resistance; bacterial communities; batch reactor; chemostat; 37 copiotrophic bacteria; mathematical model; oligotrophic bacteria 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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 3

39 Antimicrobial resistance (AMR) is a silent pandemic , less noticed than pandemics of rapidly 40 spreading infectious diseases like COVID -19 and thus considered less urgent (Barron 2024). 41 Like other pandemics, it has been spreading globally, causing a high death toll. The Global 42 Burden of Disease (GBD) consortium study of 88 pathogen –drug combinations in 204 43 countries attributed 1·27 (95% UI 0·911–1·71) million deaths globally in 2019 to bacterial AMR 44 (Murray et al. 2022). From their follow -up study of the period 1990 -2021, they forecast an 45 increase in mortality to 1·91 million (1·56–2·26) deaths attributable to AMR in 2050 (Naghavi 46 et al. 2024). AMR, climate change, biodiversity loss and pollution are causally intertwined 47 (United Nations Environment Programme (UNEP) 2023) and linked with other crises in the 48 polycrisis humanity is currently facing (World Economic Forum (WEF) 2023, Lawrence et al. 49 2024). 50 The importance of including the environment in strategies to combat AMR has been 51 recognized in the One Health concept (Larsson et al. 2023). The key environmental processes 52 affecting the fate of resistant microbes are transport, growth and removal, selection and 53 horizontal gene transfer (HGT) of resistance genes. Here, we focus on selection, which like the 54 other processes is well understood in principle but not under the complexities of the 55 environment. With the main exceptions of hospital and antibiotic manufacturing wastewater, 56 concentrations of antibiotics in the environment are much lower than minimal inhibitory 57 concentrations (MICs) and mostly lower than predicted no -effect concentrations for 58 resistance selection (PNEC-Rs) derived from them (Bengtsson-Palme & Larsson 2016, Booth 59 et al. 2020). Recent, more elaborate approaches to deriving PNEC -Rs or related measures 60 (Emara et al. 2023, Kneis et al. 2025) produce wide distributions or confidence intervals that 61 make these PNEC-Rs difficult to apply in practice. As a result, the question whether such low 62 environmental concentrations of antibiotics are below the concentrations selecting for 63 resistance under the conditions in the environment remains open. 64 One way to answer the question is to determine minimal selective concentration s (MSCs), 65 which are concentrations where the growth rate of the resistant mutant (becoming higher 66 than the sensitive wildtype at higher concentrations of the selective agent) intersects with the 67 growth rate of the sensitive wildtype (higher in the absence of the selective agent due to a 68 presumed fitness cost of resistance). Laboratory studies of resistant mutants competing with 69 isogenic sensitive strains in rich media are highly sensitive and have shown that MSCs can be 70 about 2 orders of magnitude lower than the corresponding MICs (Gullberg et al. 2011, 2014). 71 In contrast, results from laboratory incubations of small environmental samples in rich media 72 are highly variable (due to small sample volumes) and MSCs tend to be hard to discern and 73 closer to the MICs (Murray et al. 2018, Stanton et al. 2020). More realistic studies using 74 mesocosms, where large environmental samples are incubated, or lab-scale reactors that can 75 mimic environmental conditions and maintain diversity to a larger degree, have found MSCs 76 to be higher than in pure culture lab experiments (Muñoz-Aguayo et al. 2007, Knapp et al. 77 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 4 2008, Lundström et al. 2016, Tian et al. 2020). The higher diversity of the microbial 78 communities in these systems has been shown to reduce selection for resistance (Klümper et 79 al. 2019, 2024, Fang et al. 2023). 80 Direct evidence for selection in situ is scarce and hard to obtain as it would require measuring 81 resistance levels in the same microbial community over time (even if flowing away) as well as 82 concentrations of potentially selective agents (which may fluctuate). Therefore, almost all 83 studies in the environment only provide indications that selection may happen. For example, 84 many studies have observed changes in the relative abundance of ARGs from one 85 environmental compartment to the next, e.g. from wastewater treatment plant influent to 86 effluent (Kim et al. 2010, Pallares-Vega et al. 2019, Honda et al. 2023, Park et al. 2024), and 87 some also measured antibiotic concentrations (Gao et al. 2012, Mao et al. 2015). If the 88 relative abundance of certain resistance genes increases, it is tempting to conclude that 89 selection for resistance is responsible for this increase, even more so when concentrations of 90 antibiotics were higher than in more pristine environments. For example, the relative 91 abundance of fluoroquinolone resistant bacteria increased in a WWTP in Haridwar at 92 fluoroquinolone concentrations of ~10 μg/L (Kurasam et al. 2022). 93 However, we argue here that such evidence is likely to be confounded by shifts in the microbial 94 community composition if there is an association between carriage of ARGs and ecological 95 characteristics, regardless of the reasons for such an association. Community shifts inevitably 96 occur when the community is transported from one compartment to another or when there 97 is an inflow of another community. Community shifts are also impossible to avoid in laboratory 98 studies where environmental samples containing a community of microbes are incubated 99 under conditions that cannot fully reflect the environment the sample is taken from and may 100 even be very far from environmental conditions, e.g. if a rich medium is added (Murray et al. 101 2018, Stanton et al. 2020). Community shifts are somewhat avoidable in mesocosms, 102 especially if they are placed in the environment and large and complex enough to maintain 103 the trophic structure and processes of the environment such as photosynthesis (Knapp et al. 104 2008). 105 Many studies have found changes in resistance levels and then erroneously taken an increase 106 in resistance levels as evidence for selection, while others also looked for changes in 107 community composition and some even recognizing the connection to shifts in resistance : 108 Ma et al. (2024) found that an increase in filamentous bacteria, causing bulking in activated 109 sludge, increased ARG levels because there are 1.5 times more ARGs per genome of 110 filamentous bacteria compared to activated sludge bacteria overall. Guo et al. (2024) treated 111 biofilms in tidal simulation mesocosms fed with water from the Yangtze estuary with silver 112 nanoparticles for 28 days and found, using machine learning, structural equation modelling 113 and metagenomic sequencing, that the nanoparticles reduced ARG levels by reducing 114 betaproteobacteria as most ARGs were hosted by this taxon. Luo et al. (2022) found that 115 addition of microplastics to lab-scale anaerobic digestors of activated sludge waste increased 116 the relative abundance of ARGs because the microplastics enriched biofilm-forming bacteria, 117 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 5 particularly Acinetobacter spp., which carry more ARGs. Forsberg et al. (2014) found that the 118 soil resistome correlates with phylogenetic structure as ARGs were mostly nonmobile and as 119 a result, nitrogen fertilisation shifted the community structure and ARG content. Zhao et al. 120 (2018) used qPCR and 16S amplicon sequencing of faeces from pigs receiving different 121 antibiotic feed additives in three large Chinese pig farms, finding that total ARG abundance 122 correlated with antibiotic use, abundance of bacteria and mobile genetic elements (MGEs); 123 they also found a correlation between specific ARGs and specific taxa, illustrating the 124 complexities of inferring selection in the real world. Several other studies found that changes 125 in community composition correlated with changes in the relative abundance of ARGs, e.g., 126 when pig farm wastewater enters a river (Jia et al. 2017), between different compartments of 127 WWTPs (Ju et al. 2019), or during thermophilic anaerobic digestion of microalgae -bacteria 128 aggregates (Ovis-Sánchez et al. 2025). Variation in microbial community composition was 129 the main reason for changes in ARG composition in an urban river polluted by point sources 130 (Zhou et al. 2017). Zhu et al. (2025) found ARG composition in activated sludge from 142 131 WWTPs to correlate with particular taxa that are uncommon in the human gut. Similarly, diet 132 changes causing shifts in community composition led to shifts in the gut resistome (Keskey et 133 al. 2025). Often, there will be differences in community composition that are associated with 134 differences in ARGs but drivers for those changes are unknown. Also, (partially) stochastic 135 community assembly can produce spurious correlations in a particular habitat and at a 136 particular time. 137 One common cause for community shifts when human or livestock wastes, full of AMR, enter 138 the aquatic or terrestrial environment are downshifts in organic matter concentration. Such 139 downshifts are expected to favour microbes adapted to resource-poor environments, known 140 as oligotrophs, versus the microbes adapted to resource -rich environments, known as 141 copiotrophs (Kuznetsov et al. 1979, Poindexter 1981, Koch 2001, Soler -Bistué et al. 2023). 142 Oligotrophs are common in oligotrophic environments such as oceans (Schut et al. 1997, 143 Button et al. 1998, Lauro et al. 2009) or bulk soils (Li et al. 2019, Dragone et al. 2024), though 144 resource hotspots like plant roots can enrich for copiotrophs even in such environments. 145 Copiotrophs will also be favoured when environmental samples are taken to the lab and 146 supplemented with rich media. Roller et al. (2016) found that the copy number of the 147 ribosomal RNA operon ( rrn, consisting of 5S, 16S and 23S rDNA) in bacterial genomes 148 increases with increasing maximal specific growth rate (log-log correlation), while the growth 149 yield decreases. Thus, rrn copy number can be used to distinguish oligotrophic from 150 copiotrophic bacteria, at least in a statistical sense. The copiotroph E. coli was found to be the 151 main host of ARGs in wastewater, supporting the link between copiotrophs and resistance 152 (Abdulkadir et al. 2024). 153 Here, we developed a simple mathematical model to demonstrate that observational and 154 experimental studies of selection for resistance under environmental conditions can be 155 confounded by shifts in the microbial community composition that inevitably occur when a 156 microbiome is transported from one environmental compartment to another or when an 157 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 6 environmental sample is incubated under different conditions. Since resource shifts are 158 particularly relevant to AMR , the model was based on competing oligotrophic with 159 copiotrophic bacteria. We hypothesized that copiotrophic bacteria such as E. coli are more 160 likely to harbour ARGs and more competitive in high resource environments such that 161 resource downshifts will lead to a shift from copiotrophic to oligotrophic bacteria , coupled 162 with a decline in ARGs. Using complete genomes from 11,559 prokaryotic species, we show 163 that there are indeed taxa -specific associations between ARG prevalence and rrn copy 164 number, which in turn is associated with a copiotrophic growth strategy. Using our 165 mathematical model, we then show that upshifts in specific growth rates or nutrient 166 concentrations increase d the proportion of copiotrophs in the community and thus the 167 relative abundance of ARGs. This result generalizes to other reasons for community shifts as 168 long as there is some association between ARGs and ecological niches. Therefore, most 169 studies of selection for ARGs in the environment are confounded. We propose potential 170 solutions for this problem. 171 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 7

& methods 172 Bioinformatic analysis of ARG copy number association with rRNA gene copy number 173 The RefSeq database (Goldfarb et al. 2025) was chosen to ob tain the widest high quality 174 genome collection. Using the metadata file from 01/2025, only complete prokaryotic 175 genomes were kept. Finally, genomes were chosen by using ETE (v3) (Huerta-Cepas et al. 176 2016) to obtain species names and randomly picking one unique entry per species, resulting 177 in a total of 11 ,559 genomes. Krona (Ondov et al. 2011) was employed to visualize the 178 taxonomic distribution of the selected 11,559 genomes. 179 All genomes were processed through gene calling using Prodigal (Hyatt et al. 2010) followed 180 by AMR annotation using DIAMOND (v2.0.9) with option --more-sensitive (Buchfink et al. 181 2021) against the CARD database (v3.1.4) (McArthur et al. 2013). We used a best hits strategy 182 but also filtered out any hit s with (i) E -values higher than 1e -5, (ii) percentage identities 183 smaller than 40% and (iii) alignment lengths smaller than 20% for either query or target. For 184 this analysis, efflux ARGs were ignored. 185 Each genome was also annotated for ribosomal DNA using Barrnap (v0.9--3) (Seemann 2018). 186 Whenever the number of identified 16S, 23S and 5S genes was not consistent, the average 187 was taken. 188 Model development and description 189 The ordinary differential equation (ODE) mode l was designed to simulate community shifts 190 due to changes in environmental conditions while implementing an imperfect association 191 between the ecological niches of the bacteria and their resistances. 192 Two environments were implemented in the model, a chemostat and a batch culture, as these 193 simplified, unstructured environments can be viewed as the two extremes of a continuum 194 along which most environments can be positioned, for a discussion of laboratory models of 195 the environment see Wimpenny (1988). In fact, the layouts of gastrointestinal tracts of 196 animals can be understood as different combinations of batch and chemostat reactors (Godon 197 et al. 2013). Chemostats are open systems with a continuous inflow of resources and 198 continuous removal of resources and biomass from the system, both at the same rate, the 199 dilution rate (Herbert et al. 1956). The supply of resources into the system and removal of 200 organisms capture fundamental characteristics of most environments and therefore 201 chemostats and their variations have been used as models of a range of environments 202 (Jannasch 1969, Wimpenny 1988, Finlay & Fenchel 1992, Jannasch & Egli 1993, Marsh 1995, 203 Bull 2010) . Due to negative feedback between growth and substrate concentration, 204 chemostats reach a steady state where the specific growth rate equals the dilution rate 205 (Herbert et al. 1956). Simply changing the dilution rate will change the system from an 206 oligotrophic (low specific growth rate at low substrate concentration ) to a copiotrophic 207 environment (high specific growth rate at high substrate concentration). It is this shift that we 208 are interested in here. 209 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 8 Batch cultures, on the other hand, are closed systems with an initial, often large, supply of 210 resources and without any removal of resources and biomass, common in laboratories but 211 also found in nature (e.g., dead organisms, fallen leaves, faeces). Here, we include batch 212 cultures to represent those experimental studies of selection where environmental samples 213 are taken to the laboratory to inoculate rich media under conditions far removed from the 214 conditions of the sampled environment (Gullberg et al. 2011, 2014, Murray et al. 2018, 2020, 215 Stanton et al. 2020). 216 Corresponding to the modelled chemostat at low dilution rates, we seeded our models with 217 three oligotrophic bacteria, which were adapted to this low growth rate and low resource 218 environment (e.g. clean river, ocean, forest soil) by having low maximal specific growth rates 219 but high substrate affinities. Corresponding to the chemostat at high dilution rates, we also 220 seeded our models with three copiotrophic bacteria, which were adapted to this high growth 221 rate and high resource environment by having high maximal specific growth rates but low 222 substrate affinities. These copiotrophs were also adapted to the batch culture. Oligotrophic 223 bacteria were assumed to be less likely to carry resistance genes than copiotrophic bacteria, 224 which we implemented by randomly assigning one out of three oligotrophic and two out of 225 three copiotrophic bacteria to carry resistance genes. There was no antibiotic and thus no 226 selection for resistance in these models. 227 The chemostat is a highly selective environment, with only one competitor for a shared, 228 limited resource able to persist in the steady state, which is why chemostats have been used 229 to understand this competitive exclusion principle (Armstrong & McGehee 1980) . However, 230 for our purposes, we needed to maintain our community and avoid competitive exclusion. 231 This was achieved by providing six substitutable substrates – as many different resources as 232 species. Additionally, the kinetics of the three oligotrophs were made distinct, yet similar, to 233 ensure they were not identical yet remained oligotrophs. To this end, the maximum specific 234 growth rates and substrate affinities for the different substrates were varied slightly. Likewise 235 for the copiotrophs (Table S1). 236 Some species might become extinct in the chemostat; this will happen if their specific growth 237 rate falls below the dilution rate at low concentrations of substrate (Tilman 1982). Once lost, 238 a species cannot recover even if the dilution rate shifts up, so a constant low rate of 239 immigration was included to continuously seed the system with each species. 240 We simulated two case studies. One where the chemostat environment changed from 241 oligotrophic to copiotrophic by increasing the dilution rate ten -fold (mimicking a situation 242 where bacteria are transported from one to another environmental compartment). This 243 growth rate upshift will be referred to as L2H for “Low to High growth rate shift ”. The other 244 case study simulated a different shift from oligotrophic to copiotrophic, switching from 245 chemostat to batch culture, by switching off the dilution rate and instead adding resources 246 once at the same time. This growth rate upshift will be referred to as E2L for “Environment to 247 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 9 Laboratory shift” as it corresponds to adding an environmental sample to rich medium in a 248 batch culture. 249 We used the standard set of equations for chemostat dynamics (without maintenance ) but 250 included an immigration rate. For the specific growth rates, we used the modified Monod 251 model for multiple substitutable substrates developed by Lendenmann & Egli (1998). All 252 equations and parameters are provided in the Supplementary text 1. Figure S2 visualizes the 253 growth kinetics of the six species for the growth rate upshift case L2H while Figure S5 visualizes 254 the slightly different kinetics of species for the environment to laboratory shift case E2L. 255 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 10

256 ARG numbers are associated with rRNA gene copy numbers in several phyla 257 Using the RefSeq database, we annotated AMR genes and ribosomal genes in complete 258 genomes of 11,559 unique prokaryotic species. As a consequence of research bias, visualized 259 using Krona (Figure S1), only 12 phyla contained more than 50 species with complete 260 genomes, with Pseudomonadota, Actinomycetota and Bacillota containing the largest 261 numbers of known genomes. The same phyla likely contain the largest fraction of copiotrophs. 262 While simply fitting the number of ARGs against the average number of ribosomal genes 263 across all genomes did detect a significant correlation, a much better fit is obtained by 264 assessing the correlation at the phylum level ( Figure 1). Several phyla showed a clear and 265 significant correlation, while others did not. It is likely that stronger correlations can be found 266 for some phyla at a finer taxonomic level. For our argument to be valid, it is not necessary that 267 all taxa show a correlation, only that changes in community composition have the potential 268 to cause changes in relative ARG abundance. Indeed, if all taxa had the same correlation, the 269 effects would cancel. 270 Figure 1: Ribosomal and AMR genes for several prokaryotic phyla showed a positive 271 correlation. The correlation tends to be stronger and more significant for phyla with a wider 272 range of ribosomal and AMR genes. Colour differentiates classes. 273 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 11 Effects of growth rate upshifts 274 In this shift from low to high growth rate case study (L2H), the chemostat dynamics were 275 simulated until the system had reached a steady state, which was characterised by low specific 276 growth rates and substrate concentrations. Then, the dilution rate was increased, and a new 277 steady state with higher specific growth rates (but only marginally higher substrate 278 concentrations) was reached, as typical for chemostats operating well below the critical 279 dilution rate (Herbert et al. 1956). In this new steady state, the copiotrophs were dominant 280 and as a result of the link between growth rate and resistance, the fraction of resistant 281 bacteria had increased ( Figure 2). In this case, also the absolute abundance of resistant 282 bacteria had increased (Figure S3). This is because in the chemostat, the population densities 283 depended primarily on the concentrations of substrates in the inflow (if the concentration in 284 the chemostat itself is negligible as in this case study, see Figure S4) times the (fixed) growth 285 yields. Since the steady states remain the same (they are attractors) regardless of the direction 286 or order of shifts, the reverse is also true, i.e., the resistance fraction would drop if the dilution 287 and thus specific growth rate were decreased. 288 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 12 Figure 2: An upshift in the chemostat’s dilution rate led to a shift from oligotrophic species 289 (blue colours) to copiotrophic species (red colours) and an increase in resistance. As two out 290 of three copiotrophic species and only one out of three oligotrophic species carried a 291 resistance gene, the relative abundance of the resistance gene (black line) increased following 292 the upshift. In this case, also the absolute abundance of the resistant bacteria increased 293 (Figure S 3). The chemostat dynamics was simulated long enough to reach a steady state 294 before and after the upshift, with low concentrations of the substrates before and after the 295 upshift (Figure S4). 296 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 13 Effects of substrate upshifts 297 The environment to laboratory case study (E2L) was modelled as a shift from a chemostat with 298 continuously low growth rates and low substrate concentrations, capturing characteristics of 299 most environments, to a batch reactor that supports fast growth for a short period of time . 300 The first part of this was essentially the same as the first case study (but not identical as kinetic 301 parameters were slightly adjusted). The chemostat was simulated until a steady state was 302 reached where the specific growth rates matched the dilution rate of the system. Due to low 303 specific growth rate s and low substrate concentrations, the oligotrophs with their higher 304 substrate affinity and lower specific growth rate potential were favoured and dominated the 305 community ( Figure 3). Copiotrophs were maintained in the chemostat only because of a 306 baseline immigration rate. 307 Once the steady state had been reached, the continuous supply of nutrients and removal of 308 biomass was turned off (the dilution rate was set to zero), and one shot of a large amount of 309 all substrates was given (equivalent to inoculating a batch culture with a sample from the 310 chemostat representing the environment). Despite the oligotrophs dominating the ‘inoculum’, 311 the temporarily high substrate concentrations in the batch reactor enabled the copiotrophs 312 to grow faster than the also growing oligotrophs until most of the substrate had been 313 depleted. While oligotrophs had a higher specific growth rate in this last phase of batch 314 growth, the small amount of substrate left then did not allow a large increase in biomass for 315 any of the competitors. Therefore, the copiotrophs increased more than the oligotrophs 316 (Figure 3), and as a result the absolute and relative abundance of resistant bacteria increased 317 (Figure S 6). Since our simple model does not contain any maintenance or death terms, a 318 ‘frozen’ steady state was reached where populations remained constant, in contrast to the 319 dynamic and stable steady state of the chemostat. 320 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 14 Figure 3: The fraction of resistant bacteria (black line) increased when simulating the E2L 321 case study. Here the shift was from a chemostat system with low growth rates and substrate 322 concentrations, typical for most environments, to a batch culture with initially high substrate 323 concentrations, typical for some environments and laboratory models. Since bacteria were 324 not removed in the batch phase, both oligotrophs and copiotrophs increased, albeit the latter 325 more strongly, which caused an increase in the relative abundance of the resistance gene. Also 326 in this case, the absolute abundance of the resistant bacteria increased (Figure S6). Since the 327 population in the batch culture reached a higher density than in the chemostat, the increase 328 in absolute abundance was above a proportional increase in relative abundance. The 329 chemostat dynamics was simulated long enough to reach a steady state before the shift and 330 then the batch culture was simulated until the substrates were consumed ( Figure S7) and 331 growth ceased. 332 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 15

333 Community shifts affect resistance levels 334 In both case studies, the switch favoured the copiotrophs and with their relative and absolute 335 increase in the community, the relative and absolute abundance of ARGs increased. The 336 reverse, switching from copiotrophic to oligotrophic environments, will lead to a drop in ARGs. 337 Thus, community shifts can confound observational studies of selection in the environment 338 where connected environmental compartments are compared, but also experimental studies 339 where resource levels or other conditions change, resulting in community shifts. Mesocosms 340 that better maintain environmental conditions should be the least confounded type of study. 341 Considering the flow of ARGs from human and animal origins through the environment, the 342 typical situation is a drop in resource levels, which should result in a shift from copiotrophs to 343 oligotrophs in these communities and therefore a drop in ARGs. Examples are the dilution of 344 human faeces entering sewers, followed by further drops in resource levels when the sewage 345 enters treatment plants and then rivers. Similarly for animal faeces entering slurry tanks or 346 manure heaps and then again upon spreading on soils. Indeed, levels of ARGs decrease along 347 those pathways (Lin et al. 2022, Martiny et al. 2022). 348 Absolute versus relative abundance 349 We have focussed so far on relative abundance since almost all studies report changes in 350 relative abundance of ARGs while not all report both relative and absolute abundances. If the 351 total community size does not change from one compartment to the next, changes in relative 352 and absolute abundance are equivalent. However, it is almost inescapable that the total 353 community size changes so it is important to consider how this will affect relative and absolute 354 abundances of ARGs. If total community size increases, absolute abundance obviously 355 increases if the relative abundance increases, but it can also increase if the relative abundance 356 decreases. Inversely, if total community size decreases, the absolute abundance obviously 357 decreases if the relative abundance decreases, but it can also decrease if the relative 358 abundance increases. 359 Associations between ARGs and ecological niches are not limited to copiotrophs 360 We have used the association between ARG prevalence and copiotrophic growth strategy as 361 an example, but any other association between resistance(s) and ecological niche(s) likely 362

in changes in the relative abundance of ARGs when the community composition 363 changes. For example, shifts in community composition due to changes in oxygen and 364 temperature along an urban river in the Yangtze watershed were coupled to decreasing ARG 365 abundance (Wu et al. 2023). In soil mesocosms, Li et al. (2025) have shown that higher plant 366 diversity, leading to higher resource diversity in root exudates, shifts the soil microbiome from 367 bacterial hosts that harbour abundant ARGs and MGEs to those that do not, leading to a 368 reduction in ARG and MGE abundance. 369 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 16 Colonization resistance can alter resistance levels without resistance selection 370 Whether due to directed transport of microbial communities from one compartment to 371 another, via wastewater pipeline or river or rainfall -runoff, or due to stochastic dispersal, 372 individual microbes or whole communities enter resident communities all the time. If the 373 resident community includes microbes that are ecologically similar, such as sensitive bacteria 374 that are relatives of invading resistant bacteria, a priority effect, known in gut microbiology as 375 colonization resistance, explains the hurdles faced by invaders (Freter et al. 1986, Brugiroux 376 et al. 2016, Eberl et al. 2021, Letten et al. 2021, Segura Munoz et al. 2022). A study of 377 immigration success of sewer communities from WWTP influents into lab -scale reactors 378 inoculated with activated sludge found both replacement of the indigenous strains or failure 379 of immigration, with the relative abundance of 11/15 ARGs in the activated sludge increasing 380 (Gibson et al. 2023). If the resident community has a higher biodiversity, more invaders will 381 be excluded, causing a barrier to invasion by resistant bacteria (Klümper et al. 2024). 382 Community shifts can be exploited to reduce AMR 383 Community shifts often occur naturally, through transport of microbiomes or changes in 384 environmental conditions in place, or they are engineered, such as in waste treatment 385 processes, biostimulation or faecal microbiota transplantation (FMT). While the natural shifts 386 often reduce AMR, engineered shifts could be specifically chosen or optimised to reduce AMR. 387 When freshwater enters estuaries and then the marine environment, community shifts are 388 inevitable and likely to result in changes in ARG levels. A global comparison of ARG prevalence 389 across different biomes found lower prevalence in marine water compared to freshwater as 390 would be expected from the many differences between marine and warm -blooded animal 391 habitats (Lin et al. 2022). The microbial community residing in the colon experiences its first 392 major shift upon defecation, when the environment changes from anaerobic to aerobic, as 393 well as drops in temperature and downshifts in organic material. Later, entering wastewater 394 treatment plants, further shifts can occur when the treatment process involves alternating 395 anaerobic, anoxic and aerobic compartments and these shifts have been found to reduce AMR 396 levels (Christgen et al. 2015). Treatment processes can also involve temperature shifts, e.g., 397 from mesophilic microalgae-bacterial systems to thermophilic anaerobic digestion of sludge, 398 leading to community shifts that result in a reduction of resistance (Ovis-Sánchez et al. 2025). 399 However, some findings are inconsistent and are best compared by systematic reviews and 400 metanalyses (Goulas et al. 2020). FMT partially replaces the gut microbiome of a recipient 401 with that of a donor. If the donor is chosen to lack, e.g., multidrug-resistant bacteria, FMT can 402 eliminate these bacteria through competition between closely related strains (Woodworth et 403 al. 2023). As a corollary, ecological boundaries, across which shifts in community composition 404 are expected to occur, have indeed been shown to affect the distribution of ARGs (Lin et al. 405 2022). 406 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 17 Confounding due to upstream selection 407 Another pitfall that can lead to wrongly concluding that resistance has been selected in a 408 particular environment is that the selection may have happened upstream of the environment 409 studied. For example, the study by Brown et al. (2024) compiled data on community antibiotic 410 prescription frequencies and combined this with an extensive timeseries of deep 411 metagenomes in the influent of a WWTP resulting from COVID surveillance, to find that ARGs 412 hosted by Enterobacteriaceae are associated with prescription frequency. However, selection 413 of these ARGs by antibiotics could have happened upstream, in the guts of patients, rather 414 than the sewers. This is in fact more likely as antibiotic concentrations in the patients are 415 higher than in sewers. ARGs hosted by sewer indigenous Pseudomonadaceae were associated 416 with prescription frequency after a delay of 1-3 months, which suggests more indirect effects 417 than selection or possibly spurious correlations resulting from seasonal changes in 418 prescription and seasonal changes in community composition that have independent causes 419 and are out of sync. Moreover, they also found negative correlations between prescription 420 frequency and relative abundance of ARGs, which could be an artefact of using relative 421 abundance of ARGs as an absolute increase can become a relative decline when the 422 community size increases while the composition shifts (Props et al. 2017, Vandeputte et al. 423 2017, Barlow et al. 2020, Ott et al. 2021). A more direct investigation of selection and HGT in 424 lab-scale sequencing batch reactors fed with wastewater by the same lab did not find a 425 consistent pattern of selection for resistance genes as most contigs with ARGs disappeared 426 while some ARGs persisted in some contigs but not others, suggesting that their fate was 427 coupled to the contig rather than selection (Brown, Maile-Moskowitz, et al. 2024). 428 How can confounding be avoided? 429 We have argued that community shifts and upstream selection can masquerade as selection 430 in the focal environment. But how can selection be evidenced directly? The most direct 431 evidence would come from measuring changes in populations of isogenic strains where one 432 strain is resistant while the other is sensitive, ideally the resistance gene would be the only 433 difference between the strains, as in the seminal work by Gullberg et al. (2011). As a 434 consequence, the strains would have nearly the same ecological niche, including competitors, 435 predators, phages etc. This approach has been successfully used, for example by Klümper et 436 al. (2019), finding that the presence of a pig faecal community increased the MSC of E. coli 437 due to increased fitness cost of the resistance and community protection of the sensitive 438 phenotype. Their following study showed that this ‘community effect’ is surprisingly robust 439 against perturbation of the community by another antibiotic (Fang et al. 2023). In a synthetic 440 community of isogenic sensitive and resistant E. coli with Bacillus subtilis, resistance selection 441 was slowed down not due to competition for the single carbon and energy source glucose but 442 due to the production of extracellular compounds by B. subtilis (Nair & Andersson 2023). 443 This approach has limitations since introducing such an isogenic pair into the environment 444 would only be acceptable if at least one of them had been isolated from the environment and 445 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 18 the other one had spontaneously arisen from resistance gene or plasmid loss or by de novo 446 resistance mutation rather than by genetic engineering (though a clean deletion could be 447 considered). However, such isogenic pairs may already exist in the environment studied. It is 448 also possible to incubate strains in the environment in some kind of ‘cage’ that combines 449 physical isolation with exposure to , and metabolite exchange with , the environment. For 450 example, Lessard & Sieburth (1983) exposed E. coli and enterococci populations from raw 451 sewage in a stirred, transparent cage immersed in an estuary and salt marsh to measure 452 ‘decay’ rates. Natural or engineered isogenic strains could be used in mesocosms that by and 453 large maintain the composition and diversity of an environmental community and reflect the 454 relevant physicochemical conditions. Such mesocosm studies allow experimental 455 manipulation of the concentrations of selective or co -selective compounds but have usually 456 not been used with isogenic pairs (Knapp et al. 2008, Quinlan et al. 2011, Baker et al. 2022). 457 In the absence of isogenic pairs of resistant and sensitive bacteria, there is another avenue to 458 evidence selection for resistance, which is also based on an isogenic background. Quintela -459 Baluja et al. (2021) designed qPCR primers to distinguish ARG carrying , from empty, clinical 460 class 1 integrons, finding that these integrons carried 10 times more ARGs in hospital 461 wastewater than in other compartments. Integrons in recycled activated sludge and the 462 receiving river had lost ARGs. Since the only difference was carriage of ARGs, this is direct and 463 clear evidence for selection for ARGs in hospital wastewater or ‘upstream’ in patients, and 464 lack of selection in activated sludge and receiving rivers. Similarly, selection could be tested 465 by tracking the fate of ARGs embedded in their (meta)genomic contexts while these contexts 466 are transported from one environmental compartment to another , such as from gut 467 microbiomes under antibiotic treatment to gut microbiomes not under treatment (Lee et al. 468 2023). 469 We are not aware of alternative avenues to directly test for resistance selection. Perhaps, 470 indirect evidence that is at risk of confounding could be used if sufficiently large studies 471 encompassing different and known levels of potentially confounding factors enable statistical 472 analysis that can control for these factors. 473 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 19

and recommendations 474 Our finding that microbial community shifts can result in changes of absolute and relative 475 abundance of ARGs and ARBs in the absence of selection (and HGT) is trivial in the sense that 476 it follows logically from the assumed coupling of resistance with ecology. This coupling is well 477 documented though complex and variable in its details. 478 Our point is that the consequences of this trivial result are far from trivial, as any observational 479 or experimental study of changes in ARG abundance cannot provide unequivocal evidence for 480 or against selection whenever there has also been a change in environmental conditions, 481 which is hard to avoid. Any incubation of a microbiome sampled from the environment in the 482 laboratory will undoubtedly change environmental conditions. Using larger mesocosms, 483 especially when placed in situ, will reduce this shift and maintain more of the diversity of the 484 source environment, but some change in conditions is unavoidable even in mesocosms. 485 While our example s were based on an association between ARGs and copiotrophs (fast -486 growing bacteria) , coupled to shifts from eutrophic conditions favouring copiotrophs to 487 oligotrophic conditions favouring oligotrophs, the conclusions are general. Any association 488 between ARGs and bacteria with particular ecological niches can lead to changes in relative 489 ARG abundance when environmental conditions change such that the community 490 composition changes (which is almost inevitable). 491 An association between ARGs and ecological niches may often be linked to an association 492 between phylogeny and ecology that has been observed on various levels of hierarchy. For 493 example, Pseudomonadota (Proteobacteria) phylum members tend to be fast -growing, 494 copiotrophic bacteria that also tend to harbour various plasmids including those with ARGs. 495 HGT of ARGs has the potential to break associations between ARGs and ecology, however, 496 associations between phylogeny and resistance plasmids on the one hand and phylogeny and 497 growth strategy on the other, mean that HGT within such taxa may not break the association. 498 The issue of selection being confounded by community shifts prompts the question of how to 499 obtain direct, unequivocal evidence of selection for or against resistance at certain 500 concentrations of antibiotics or other selective agents? 501 Crucial for any solution is to somehow eliminate ecological differences between sensitive and 502 resistant bacteria. Ideally, one would determine changes in relative abundance of a sensitive 503 bacterium competing with a resistant bacterium that is otherwise isogenic, such as a resistant 504 mutant of the same strain. This ought to guarantee that their ecology is very similar if not 505 identical, though the presence of a resistance gene or plasmid has the potential to perturb 506 gene expression and changes in ecological niche upon acquiring resistance have been 507 reported (Letten et al. 2021). Spiking such an isogenic pair into laboratory or mesocosm 508 incubations and tracking their abundance would then enable direct measurement of 509 selection. Keeping track of an isogenic pair in the environment is possible but will be harder. 510 For environmental observations there is fortunately an alternative, based on detecting the 511 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 20 gain or loss of resistance genes from mobile genetic elements in the same strain or genomic 512 context when the strain or genome is transported from one environmental compartment to 513 another, for an example see Quintela-Baluja et al. (2021). A tendency for a resistance gene to 514 get lost from the mobile genetic element, strain or genome in the downstream compartment 515 would indicate lack of selection for this resistance gene in the downstream compartment. 516 Both options rely on the use of an isogenic background to reduce confounding through 517 ecological differences. 518 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 21 Conflicts of Interest 519 The authors declare that the research was conducted in the absence of any commercial or 520 financial relationships that could be construed as a potential conflict of interest. 521 Author Contributions 522 Cansu Uluseker developed the model and code, simulated the model and made the 523 corresponding figures: Formal analysis , Investigation, Methodology, Software, Validation, 524 Visualization, Writing – original draft . Sébastien Raguideau did the bioinformatic and 525 statistical analysis and made the corresponding figures: Data curation , Formal analysis , 526 Investigation, Methodology, Validation, Visualization, Writing – original draft. Christopher 527 Quince guided the bioinformatic and statistical analysis: Conceptualization, Methodology, 528 Supervision. Jan-Ulrich Kreft came up with the idea for this study and guided the modelling 529 and drafting of the paper: Conceptualization, Funding acquisition , Project administration, 530 Supervision, Writing – original draft, Writing – review & editing. 531 Funding 532 C.U. and J.-U.K. acknowledge support through an international collaboration grant from the 533 Natural Environment Research Council (NERC) in the UK (NE/T013222/1) and the Department 534 of Biotechnology (DBT) in India (Computer No. 8981 || BT/IN/Indo-UK/AMR-Env/03/ST/2020-535 21 || AMRFlows) for the project “AMRflows: antimicrobials and resistance from 536 manufacturing flows to people: joined up experiments, mathematical modelling, and risk 537 analysis.” C.Q. and S.R. acknowledge the support of the Biotechnology and Biological Sciences 538 Research Council (BBSRC), part of UK Research and Innovation; Earlham Institute Strategic 539 Programme Grant (Decoding Biodiversity) BBX011089/1 and its constituent work package 540 BBS/E/ER/230002C; the Core Strategic Programme Grant (Genomes to Food Security) 541 BB/CSP1720/1 and its constituent work packages BBS/E/T/000PR9818 and 542 BBS/E/T/000PR9817; and the Core Capability Grant BB/CCG2220/1. 543 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 22

544 Abdulkadir N, Saraiva JP , Zhang J, Stolte S, Gillor O, Harms H, Rocha U (2024). Genome-545 centric analyses of 165 metagenomes show that mobile genetic elements are 546 crucial for the transmission of antimicrobial resistance genes to pathogens in 547 activated sludge and wastewater. Microbiology Spectrum 0: e02918-23 548 https://doi.org/10.1128/spectrum.02918-23 549 Armstrong RA, McGehee R (1980). Competitive exclusion. American Naturalist 115: 550 151–170 https://doi.org/10.1086/283553 551 Baker M, Williams AD, Hooton SPT, Helliwell R, King E, Dodsworth T, Baena-Nogueras 552 RM, Warry A, Ortori CA, Todman H, Gray-Hammerton CJ, Pritchard ACW, Iles E, 553 Cook R, Emes RD, Jones MA, Kypraios T, West H, Barrett DA, Ramsden SJ, Gomes 554 RL, Hudson C, Millard AD, Raman S, Morris C, Dodd CER, Kreft J-U, Hobman JL, 555 Stekel DJ (2022). Antimicrobial resistance in dairy slurry tanks: a critical point for 556 measurement and control. Environment International 169: 107516 557 https://doi.org/10.1016/j.envint.2022.107516 558 Barlow JT, Bogatyrev SR, Ismagilov RF (2020). A quantitative sequencing framework for 559 absolute abundance measurements of mucosal and lumenal microbial 560 communities. Nature Communications 11: 2590 561 https://doi.org/10.1038/s41467-020-16224-6 562 Barron M (2024). The Antimicrobial Resistance Pandemic: Breaking the Silence. 563 Microcosm 564 Bengtsson-Palme J, Larsson DGJ (2016). Concentrations of antibiotics predicted to 565 select for resistant bacteria: Proposed limits for environmental regulation. 566 Environment International 86: 140–149 567 https://doi.org/10.1016/j.envint.2015.10.015 568 Booth A, Aga DS, Wester AL (2020). Retrospective analysis of the global antibiotic 569 residues that exceed the predicted no effect concentration for antimicrobial 570 resistance in various environmental matrices. Environment International 141: 571 105796 https://doi.org/10.1016/j.envint.2020.105796 572 Brown CL, Maile-Moskowitz A, Lopatkin AJ, Xia K, Logan LK, Davis BC, Zhang L, 573 Vikesland PJ, Pruden A (2024). Selection and horizontal gene transfer underlie 574 microdiversity-level heterogeneity in resistance gene fate during wastewater 575 treatment. Nature Communications 15: 5412 https://doi.org/10.1038/s41467-576 024-49742-8 577 Brown CL, Rumi MA, McDaniel L, Maile-Moskowitz A, Sein J, Nguyen L, Choi M, Hindi F , 578 Mullet J, Emon M, Moumi NA, Blair MF , Davis BC, Rao J, Baffoe-Bonnie A, 579 Vikesland P , Pruden A, Zhang L (2024). Metagenomics disentangles 580 epidemiological and microbial ecological associations between community 581 antibiotic use and antibiotic resistance indicators measured in sewage. 582 2024.12.11.24318846 https://doi.org/10.1101/2024.12.11.24318846 583 Brugiroux S, Beutler M, Pfann C, Garzetti D, Ruscheweyh H-J, Ring D, Diehl M, Herp S, 584 Lötscher Y , Hussain S, Bunk B, Pukall R, Huson DH, Münch PC, McHardy AC, 585 McCoy KD, Macpherson AJ, Loy A, Clavel T, Berry D, Stecher B (2016). Genome-586 guided design of a defined mouse microbiota that confers colonization 587 resistance against Salmonella enterica serovar Typhimurium. Nature 588 Microbiology 2: 16215 https://doi.org/10.1038/nmicrobiol.2016.215 589 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 23 Buchfink B, Reuter K, Drost H-G (2021). Sensitive protein alignments at tree-of-life scale 590 using DIAMOND. Nature Methods 18: 366–368 https://doi.org/10.1038/s41592-591 021-01101-x 592 Bull AT (2010). The renaissance of continuous culture in the post-genomics age. Journal 593 of Industrial Microbiology & Biotechnology 37: 993–1021 594 Button DK, Robertson BR, Lepp PW, Schmidt TM (1998). A small, dilute-cytoplasm, 595 high-affinity, novel bacterium isolated by extinction culture and having kinetic 596 constants compatible with growth at ambient concentrations of dissolved 597 nutrients in seawater. Applied and Environmental Microbiology 64: 4467–4476 598 Christgen B, Yang Y , Ahammad SZ, Li B, Rodriquez DC, Zhang T, Graham DW (2015). 599 Metagenomics Shows That Low-Energy Anaerobic−Aerobic Treatment Reactors 600 Reduce Antibiotic Resistance Gene Levels from Domestic Wastewater. 601 Environmental Science & Technology 49: 2577–2584 602 https://doi.org/10.1021/es505521w 603 Dragone NB, Hoffert M, Strickland MS, Fierer N (2024). Taxonomic and genomic 604 attributes of oligotrophic soil bacteria. ISME Communications 4: ycae081 605 https://doi.org/10.1093/ismeco/ycae081 606 Eberl C, Weiss AS, Jochum LM, Raj ACD, Ring D, Hussain S, Herp S, Meng C, Kleigrewe 607 K, Gigl M, Basic M, Stecher B (2021). E. coli enhance colonization resistance 608 against Salmonella Typhimurium by competing for galactitol, a context-609 dependent limiting carbon source. Cell Host & Microbe 29: 1–13 610 https://doi.org/10.1016/j.chom.2021.09.004 611 Emara Y , Jolliet O, Finkbeiner M, Heß S, Kosnik M, Siegert M-W, Fantke P (2023). 612 Comparative selective pressure potential of antibiotics in the environment. 613 Environmental Pollution 318: 120873 614 https://doi.org/10.1016/j.envpol.2022.120873 615 Fang P , Elena AX, Kunath MA, Berendonk TU, Klümper U (2023). Reduced selection for 616 antibiotic resistance in community context is maintained despite pressure by 617 additional antibiotics. ISME Communications 3: 52 618 https://doi.org/10.1038/s43705-023-00262-4 619 Finlay BJ, Fenchel T (1992). An anaerobic ciliate as a natural chemostat for the growth of 620 endosymbiotic methanogens. European Journal of Protistology 28: 127–137 621 https://doi.org/10.1016/S0932-4739(11)80041-6 622 Forsberg KJ, Patel S, Gibson MK, Lauber CL, Knight R, Fierer N, Dantas G (2014). 623 Bacterial phylogeny structures soil resistomes across habitats. Nature 624 https://doi.org/10.1038/nature13377 625 Freter R, Brickner H, Temme S (1986). An understanding of colonization resistance of 626 the mammalian large intestine requires mathematical analysis. Microecology 627 and Therapy 16: 147–155 628 Gao P , Munir M, Xagoraraki I (2012). Correlation of tetracycline and sulfonamide 629 antibiotics with corresponding resistance genes and resistant bacteria in a 630 conventional municipal wastewater treatment plant. Science of The Total 631 Environment 421–422: 173–183 https://doi.org/10.1016/j.scitotenv.2012.01.061 632 Gibson C, Kraemer SA, Klimova N, Guo B, Frigon D (2023). Antibiotic resistance gene 633 sequencing is necessary to reveal the complex dynamics of immigration from 634 sewers to activated sludge. Frontiers in Microbiology 14: 635 https://doi.org/10.3389/fmicb.2023.1155956 636 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 24 Godon J-J, Arcemisbéhère L, Escudié R, Harmand J, Miambi E, Steyer J-P (2013). 637 Overview of the oldest existing set of substrate-optimized anaerobic processes: 638 digestive tracts. BioEnergy Research 6: 1063–1081 639 https://doi.org/10.1007/s12155-013-9339-y 640 Goldfarb T, Kodali VK, Pujar S, Brover V , Robbertse B, Farrell CM, Oh D-H, Astashyn A, 641 Ermolaeva O, Haddad D, Hlavina W, Hoffman J, Jackson JD, Joardar VS, 642 Kristensen D, Masterson P , McGarvey KM, McVeigh R, Mozes E, Murphy MR, 643 Schafer SS, Souvorov A, Spurrier B, Strope PK, Sun H, Vatsan AR, Wallin C, Webb 644 D, Brister JR, Hatcher E, Kimchi A, Klimke W, Marchler-Bauer A, Pruitt KD, 645 Thibaud-Nissen F , Murphy TD (2025). NCBI RefSeq: reference sequence 646 standards through 25 years of curation and annotation. Nucleic Acids Research 647 53: D243–D257 https://doi.org/10.1093/nar/gkae1038 648 Goulas A, Belhadi D, Descamps A, Andremont A, Benoit P , Courtois S, Dagot C, Grall N, 649 Makowski D, Nazaret S, Nélieu S, Patureau D, Petit F , Roose-Amsaleg C, Vittecoq 650 M, Livoreil B, Laouénan C (2020). How effective are strategies to control the 651 dissemination of antibiotic resistance in the environment? A systematic review. 652 Environmental Evidence 9: 4 https://doi.org/10.1186/s13750-020-0187-x 653 Gullberg E, Albrecht LM, Karlsson C, Sandegren L, Andersson DI (2014). Selection of a 654 Multidrug Resistance Plasmid by Sublethal Levels of Antibiotics and Heavy 655 Metals. mBio 5: e01918-14 https://doi.org/10.1128/mBio.01918-14 656 Gullberg E, Cao S, Berg OG, Ilbäck C, Sandegren L, Hughes D, Andersson DI (2011). 657 Selection of Resistant Bacteria at Very Low Antibiotic Concentrations. PLOS 658 Pathogens 7: e1002158 https://doi.org/10.1371/journal.ppat.1002158 659 Guo X, Chen X, Sidikjan N, Sha R (2024). Silver nanoparticles regulate antibiotic 660 resistance genes by shifting bacterial community and generating anti-silver 661 genes in estuarine biofilms. Aquatic Toxicology 276: 107131 662 https://doi.org/10.1016/j.aquatox.2024.107131 663 Herbert D, Elsworth R, Telling RC (1956). The continuous culture of bacteria; a 664 theoretical and experimental study. J Gen Microbiol 14: 601–622 665 Honda R, Matsuura N, Sorn S, Asakura S, Morinaga Y , Van Huy T, Sabar MA, Masakke Y , 666 Hara-Yamamura H, Watanabe T (2023). Transition of antimicrobial resistome in 667 wastewater treatment plants: impact of process configuration, geographical 668 location and season. npj Clean Water 6: 1–12 https://doi.org/10.1038/s41545-669 023-00261-x 670 Huerta-Cepas J, Serra F , Bork P (2016). ETE 3: Reconstruction, Analysis, and 671 Visualization of Phylogenomic Data. Molecular Biology and Evolution 33: 1635–672 1638 https://doi.org/10.1093/molbev/msw046 673 Hyatt D, Chen G-L, LoCascio PF , Land ML, Larimer FW, Hauser LJ (2010). Prodigal: 674 prokaryotic gene recognition and translation initiation site identification. BMC 675 Bioinformatics 11: 119 https://doi.org/10.1186/1471-2105-11-119 676 Jannasch HW (1969). Estimations of bacterial growth rates in natural waters. Journal of 677 Bacteriology 99: 156–160 678 Jannasch HW, Egli T (1993). Microbial growth kinetics: a historical perspective. Antonie 679 van Leeuwenhoek 63: 213–224 https://doi.org/10.1007/BF00871219 680 Jia S, Zhang X-X, Miao Y , Zhao Y , Ye L, Li B, Zhang T (2017). Fate of antibiotic resistance 681 genes and their associations with bacterial community in livestock breeding 682 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 25 wastewater and its receiving river water. Water Research 124: 259–268 683 https://doi.org/10.1016/j.watres.2017.07.061 684 Ju F , Beck K, Yin X, Maccagnan A, McArdell CS, Singer HP , Johnson DR, Zhang T, 685 Bürgmann H (2019). Wastewater treatment plant resistomes are shaped by 686 bacterial composition, genetic exchange, and upregulated expression in the 687 effluent microbiomes. The ISME Journal 13: 346–360 688 https://doi.org/10.1038/s41396-018-0277-8 689 Keskey R, Bluiminck S, Sangwan N, Meltzer R, Lam A, Thewissen R, Zaborin A, van Goor 690 H, Zaborina O, Alverdy J (2025). Dietary impact on the gut resistome: western diet 691 independently increases the prevalence of antibiotic resistance genes within the 692 gut microbiota. Microbiology Spectrum 0: e02766-24 693 https://doi.org/10.1128/spectrum.02766-24 694 Kim S, Park H, Chandran K (2010). Propensity of activated sludge to amplify or attenuate 695 tetracycline resistance genes and tetracycline resistant bacteria: A 696 mathematical modeling approach. Chemosphere 78: 1071–1077 697 https://doi.org/10.1016/j.chemosphere.2009.12.068 698 Klümper U, Gionchetta G, Catão E, Bellanger X, Dielacher I, Elena AX, Fang P , Galazka S, 699 Goryluk-Salmonowicz A, Kneis D, Okoroafor U, Radu E, Szadziul M, Szekeres E, 700 Teban-Man A, Coman C, Kreuzinger N, Popowska M, Vierheilig J, Walsh F , 701 Woegerbauer M, Bürgmann H, Merlin C, Berendonk TU (2024). Environmental 702 microbiome diversity and stability is a barrier to antimicrobial resistance gene 703 accumulation. Communications Biology 7: 1–13 https://doi.org/10.1038/s42003-704 024-06338-8 705 Klümper U, Recker M, Zhang L, Yin X, Zhang T, Buckling A, Gaze WH (2019). Selection for 706 antimicrobial resistance is reduced when embedded in a natural microbial 707 community. The ISME Journal 13: 2927–2937 https://doi.org/10.1038/s41396-708 019-0483-z 709 Knapp CW, Engemann CA, Hanson ML, Keen PL, Hall KJ, Graham DW (2008). Indirect 710 Evidence of Transposon-Mediated Selection of Antibiotic Resistance Genes in 711 Aquatic Systems at Low-Level Oxytetracycline Exposures. Environmental 712 Science & Technology 42: 5348–5353 https://doi.org/10.1021/es703199g 713 Kneis D, Barron M de la C, Konyali D, Westphal V , Schröder P , Westphal-Settele K, 714 Schönfeld J, Jungmann D, Berendonk TU, Klümper U (2025). Suppressing 715 selection for antibiotic resistance in the environment: A transparent, ecology-716 based approach to predicted no-effect concentrations. 2025.04.04.647007 717 https://doi.org/10.1101/2025.04.04.647007 718 Koch AL (2001). Oligotrophs versus copiotrophs. BioEssays 23: 657–661 719 https://doi.org/10.1002/bies.1091 720 Kurasam J, Mandal PK, Sarkar S (2022). Selective Proliferation of Antibiotic-Resistant 721 Bacteria in the Biological Treatment Process at a Municipal Wastewater 722 Treatment Plant in India. Journal of Environmental Engineering 148: 04022007 723 https://doi.org/10.1061/(ASCE)EE.1943-7870.0001980 724 Kuznetsov SI, Dubinina GA, Lapteva NA (1979). Biology of oligotrophic bacteria. Annual 725 Review of Microbiology 33: 377–387 726 https://doi.org/10.1146/annurev.mi.33.100179.002113 727 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 26 Larsson DGJ, Gaze WH, Laxminarayan R, Topp E (2023). AMR, One Health and the 728 environment. Nature Microbiology 1–2 https://doi.org/10.1038/s41564-023-729 01351-9 730 Lauro FM, McDougald D, Thomas T, Williams TJ, Egan S, Rice S, DeMaere MZ, Ting L, 731 Ertan H, Johnson J, Ferriera S, Lapidus A, Anderson I, Kyrpides N, Munk AC, 732 Detter C, Han CS, Brown MV , Robb FT, Kjelleberg S, Cavicchioli R (2009). The 733 genomic basis of trophic strategy in marine bacteria. Proceedings of the National 734 Academy of Sciences of the United States of America 106: 15527–15533 735 https://doi.org/10.1073/pnas.0903507106 736 Lawrence M, Homer-Dixon T, Janzwood S, Rockstöm J, Renn O, Donges JF (2024). Global 737 polycrisis: the causal mechanisms of crisis entanglement. Global Sustainability 738 7: e6 https://doi.org/10.1017/sus.2024.1 739 Lee K, Raguideau S, Sirén K, Asnicar F , Cumbo F , Hildebrand F , Segata N, Cha C-J, 740 Quince C (2023). Population-level impacts of antibiotic usage on the human gut 741 microbiome. Nature Communications 14: 1191 https://doi.org/10.1038/s41467-742 023-36633-7 743 Lendenmann U, Egli T (1998). Kinetic models for the growth of Escherichia coli with 744 mixtures of sugars under carbon-limited conditions. Biotechnology and 745 Bioengineering 59: 99–107 https://doi.org/10.1002/(sici)1097-746 0290(19980705)59:1%253C99::aid-bit13%253E3.0.co;2-y 747 Lessard EJ, Sieburth JMcN (1983). Survival of Natural Sewage Populations of Enteric 748 Bacteria in Diffusion and Batch Chambers in the Marine Environment. Applied 749 and Environmental Microbiology 45: 950–959 750 Letten AD, Hall AR, Levine JM (2021). Using ecological coexistence theory to understand 751 antibiotic resistance and microbial competition. Nature Ecology & Evolution 5: 752 431–441 https://doi.org/10.1038/s41559-020-01385-w 753 Li J, Mau RL, Dijkstra P , Koch BJ, Schwartz E, Liu X-JA, Morrissey EM, Blazewicz SJ, Pett-754 Ridge J, Stone BW, Hayer M, Hungate BA (2019). Predictive genomic traits for 755 bacterial growth in culture versus actual growth in soil. The ISME Journal 13: 756 2162–2172 https://doi.org/10.1038/s41396-019-0422-z 757 Li S, Zhou X, Liu L, Su Z, Zhao J, Zhang J, Cai Z, Peñuelas J, Huang X (2025). Plant 758 Diversity Reduces the Risk of Antibiotic Resistance Genes in Agroecosystems. 759 Advanced Science n/a: 2410990 https://doi.org/10.1002/advs.202410990 760 Lin Q, Xavier BB, Alako BTF , Mitchell AL, Rajakani SG, Glupczynski Y , Finn RD, Cochrane 761 G, Malhotra-Kumar S (2022). Screening of global microbiomes implies ecological 762 boundaries impacting the distribution and dissemination of clinically relevant 763 antimicrobial resistance genes. Communications Biology 5: 1–9 764 https://doi.org/10.1038/s42003-022-04187-x 765 Lundström SV , Östman M, Bengtsson-Palme J, Rutgersson C, Thoudal M, Sircar T, 766 Blanck H, Eriksson KM, Tysklind M, Flach C-F , Larsson DGJ (2016). Minimal 767 selective concentrations of tetracycline in complex aquatic bacterial biofilms. 768 Science of The Total Environment 553: 587–595 769 https://doi.org/10.1016/j.scitotenv.2016.02.103 770 Luo T, Dai X, Chen Z, Wu L, Wei W, Xu Q, Ni B-J (2022). Different microplastics 771 distinctively enriched the antibiotic resistance genes in anaerobic sludge 772 digestion through shifting specific hosts and promoting horizontal gene flow. 773 Water Research 228: 119356 https://doi.org/10.1016/j.watres.2022.119356 774 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 27 Ma Y , Qiao Y , Zhang X, Ye L (2024). Filamentous bacteria-induced sludge bulking can 775 alter antibiotic resistance gene profiles and increase potential risks in 776 wastewater treatment systems. Environment International 190: 108920 777 https://doi.org/10.1016/j.envint.2024.108920 778 Mao D, Yu S, Rysz M, Luo Y , Yang F , Li F , Hou J, Mu Q, Alvarez PJJ (2015). Prevalence and 779 proliferation of antibiotic resistance genes in two municipal wastewater 780 treatment plants. Water Research 85: 458–466 781 https://doi.org/10.1016/j.watres.2015.09.010 782 Marsh PD (1995). The role of continuous culture in modeling the human microflora. 783 Journal of Chemical Technology and Biotechnology 64: 1–9 784 Martiny H-M, Munk P , Brinch C, Aarestrup FM, Petersen TN (2022). A curated data 785 resource of 214K metagenomes for characterization of the global antimicrobial 786 resistome. PLOS Biology 20: e3001792 787 https://doi.org/10.1371/journal.pbio.3001792 788 McArthur AG, Waglechner N, Nizam F , Yan A, Azad MA, Baylay AJ, Bhullar K, Canova MJ, 789 De Pascale G, Ejim L, Kalan L, King AM, Koteva K, Morar M, Mulvey MR, O’Brien 790 JS, Pawlowski AC, Piddock LJV , Spanogiannopoulos P , Sutherland AD, Tang I, 791 Taylor PL, Thaker M, Wang W, Yan M, Yu T, Wright GD (2013). The Comprehensive 792 Antibiotic Resistance Database. Antimicrobial Agents and Chemotherapy 57: 793 3348–3357 https://doi.org/10.1128/aac.00419-13 794 Muñoz-Aguayo J, Lang KS, LaPara TM, González G, Singer RS (2007). Evaluating the 795 Effects of Chlortetracycline on the Proliferation of Antibiotic-Resistant Bacteria 796 in a Simulated River Water Ecosystem. Applied and Environmental Microbiology 797 73: 5421–5425 https://doi.org/10.1128/AEM.00708-07 798 Murray AK, Stanton IC, Wright J, Zhang L, Snape J, Gaze WH (2020). The ‘SELection End 799 points in Communities of bacTeria’ (SELECT) Method: A Novel Experimental 800 Assay to Facilitate Risk Assessment of Selection for Antimicrobial Resistance in 801 the Environment. Environmental Health Perspectives 128: 107007 802 https://doi.org/10.1289/EHP6635 803 Murray AK, Zhang L, Yin X, Zhang T, Buckling A, Snape J, Gaze WH (2018). Novel Insights 804 into Selection for Antibiotic Resistance in Complex Microbial Communities. 805 mBio 9: https://doi.org/10.1128/mBio.00969-18 806 Murray CJ, Ikuta KS, Sharara F , Swetschinski L, Aguilar GR, Gray A, et al. (2022). Global 807 burden of bacterial antimicrobial resistance in 2019: a systematic analysis. The 808 Lancet 399: 629–655 https://doi.org/10.1016/S0140-6736(21)02724-0 809 Naghavi M, Vollset SE, Ikuta KS, Swetschinski LR, Gray AP , Wool EE, et al. (2024). Global 810 burden of bacterial antimicrobial resistance 1990–2021: a systematic analysis 811 with forecasts to 2050. The Lancet 404: 1199–1226 812 https://doi.org/10.1016/S0140-6736(24)01867-1 813 Nair RR, Andersson DI (2023). Interspecies interaction reduces selection for antibiotic 814 resistance in Escherichia coli. Communications Biology 6: 1–9 815 https://doi.org/10.1038/s42003-023-04716-2 816 Ondov BD, Bergman NH, Phillippy AM (2011). Interactive metagenomic visualization in a 817 Web browser. BMC Bioinformatics 12: 385 https://doi.org/10.1186/1471-2105-818 12-385 819 Ott A, Quintela-Baluja M, Zealand AM, O’Donnell G, Haniffah MRM, Graham DW (2021). 820 Improved quantitative microbiome profiling for environmental antibiotic 821 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 28 resistance surveillance. Environmental Microbiome 16: 21 822 https://doi.org/10.1186/s40793-021-00391-0 823 Ovis-Sánchez JO, Vital-Jácome M, Buitrón G, Cervantes-Avilés P , Carrillo-Reyes J (2025). 824 Antibiotic resistance reduction mechanisms during thermophilic anaerobic 825 digestion of microalgae-bacteria aggregates. Bioresource Technology 419: 826 132037 https://doi.org/10.1016/j.biortech.2025.132037 827 Pallares-Vega R, Blaak H, van der Plaats R, de Roda Husman AM, Hernandez Leal L, van 828 Loosdrecht MCM, Weissbrodt DG, Schmitt H (2019). Determinants of presence 829 and removal of antibiotic resistance genes during WWTP treatment: A cross-830 sectional study. Water Research 161: 319–328 831 https://doi.org/10.1016/j.watres.2019.05.100 832 Park J-H, Bae K-S, Kang J, Park E-R, Yoon J-K (2024). Comprehensive Study of Antibiotic 833 Resistance in Enterococcus spp.: Comparison of Influents and Effluents of 834 Wastewater Treatment Plants. Antibiotics 13: 1072 835 https://doi.org/10.3390/antibiotics13111072 836 Poindexter JS (1981). Oligotrophy - Fast and Famine Existence. In: Alexander M (ed), 837 Advances in Microbial Ecology. Springer US, Boston, MA, pp. 63–89 838 Props R, Kerckhof F-M, Rubbens P , De Vrieze J, Hernandez Sanabria E, Waegeman W, 839 Monsieurs P , Hammes F , Boon N (2017). Absolute quantification of microbial 840 taxon abundances. The ISME Journal 11: 584–587 841 https://doi.org/10.1038/ismej.2016.117 842 Quinlan EL, Nietch CT, Blocksom K, Lazorchak JM, Batt AL, Griffiths R, Klemm DJ (2011). 843 Temporal Dynamics of Periphyton Exposed to Tetracycline in Stream 844 Mesocosms. Environmental Science & Technology 45: 10684–10690 845 https://doi.org/10.1021/es202004k 846 Quintela-Baluja M, Frigon D, Abouelnaga M, Jobling K, Romalde JL, Gomez Lopez M, 847 Graham DW (2021). Dynamics of integron structures across a wastewater 848 network – Implications to resistance gene transfer. Water Research 206: 117720 849 https://doi.org/10.1016/j.watres.2021.117720 850 Roller BRK, Stoddard SF , Schmidt TM (2016). Exploiting rRNA operon copy number to 851 investigate bacterial reproductive strategies. Nature Microbiology 1: 1–7 852 https://doi.org/10.1038/nmicrobiol.2016.160 853 Schut F , Prins RA, Gottschal JC (1997). Oligotrophy and pelagic marine bacteria: Facts 854 and fiction. Aquatic Microbial Ecology 12: 177–202 855 Seemann T (2018). Barrnap: BAsic Rapid Ribosomal RNA Predictor. 856 Segura Munoz RR, Mantz S, Martínez I, Li F , Schmaltz RJ, Pudlo NA, Urs K, Martens EC, 857 Walter J, Ramer-Tait AE (2022). Experimental evaluation of ecological principles 858 to understand and modulate the outcome of bacterial strain competition in gut 859 microbiomes. The ISME Journal 16: 1594–1604 https://doi.org/10.1038/s41396-860 022-01208-9 861 Soler-Bistué A, Couso LL, Sánchez IE (2023). The evolving copiotrophic/oligotrophic 862 dichotomy: From Winogradsky to physiology and genomics. Environmental 863 Microbiology 25: 1232–1237 https://doi.org/10.1111/1462-2920.16360 864 Stanton IC, Murray AK, Zhang L, Snape J, Gaze WH (2020). Evolution of antibiotic 865 resistance at low antibiotic concentrations including selection below the 866 minimal selective concentration. Communications Biology 3: 1–11 867 https://doi.org/10.1038/s42003-020-01176-w 868 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint Uluseker et al. (2025) Community shifts confound resistance selection 29 Tian Z, Palomo A, Zhang H, Luan X, Liu R, Awad M, Smets BF , Zhang Y , Yang M (2020). 869 Minimum influent concentrations of oxytetracycline, streptomycin and 870 spiramycin in selecting antibiotic resistance in biofilm type wastewater 871 treatment systems. Science of The Total Environment 720: 137531 872 https://doi.org/10.1016/j.scitotenv.2020.137531 873 Tilman D (1982). Resource Competition and Community Structure Princeton University 874 Press, New Jersey 875 United Nations Environment Programme (UNEP) (2023). Bracing for Superbugs: 876 Strengthening environmental action in the One Health response to antimicrobial 877 resistance Geneva 878 Vandeputte D, Kathagen G, D’hoe K, Vieira-Silva S, Valles-Colomer M, Sabino J, Wang J, 879 Tito RY , De Commer L, Darzi Y , Vermeire S, Falony G, Raes J (2017). Quantitative 880 microbiome profiling links gut community variation to microbial load. Nature 881 551: 507–511 https://doi.org/10.1038/nature24460 882 Wimpenny JWT (1988). CRC Handbook of Laboratory Model Systems for Microbial 883 Ecosystems CRC Press, Boca Raton, FL, USA 884 Woodworth MH, Conrad RE, Haldopoulos M, Pouch SM, Babiker A, Mehta AK, Sitchenko 885 KL, Wang CH, Strudwick A, Ingersoll JM, Philippe C, Lohsen S, Kocaman K, 886 Lindner BG, Hatt JK, Jones RM, Miller C, Neish AS, Friedman-Moraco R, 887 Karadkhele G, Liu KH, Jones DP , Mehta CC, Ziegler TR, Weiss DS, Larsen CP , 888 Konstantinidis KT, Kraft CS (2023). Fecal microbiota transplantation promotes 889 reduction of antimicrobial resistance by strain replacement. Science 890 Translational Medicine 15: eabo2750 891 https://doi.org/10.1126/scitranslmed.abo2750 892 World Economic Forum (WEF) (2023). Global Risks Report 2023 World Economic 893 Forum, Geneva 894 Wu D, Zhao J, Su Y , Yang M, Dolfing J, Graham DW, Yang K, Xie B (2023). Explaining the 895 resistomes in a megacity’s water supply catchment: Roles of microbial 896 assembly-dominant taxa, niched environments and pathogenic bacteria. Water 897 Research 228: 119359 https://doi.org/10.1016/j.watres.2022.119359 898 Zhao Y , Su J-Q, An X-L, Huang F-Y , Rensing C, Brandt KK, Zhu Y-G (2018). Feed additives 899 shift gut microbiota and enrich antibiotic resistance in swine gut. Science of The 900 Total Environment 621: 1224–1232 901 https://doi.org/10.1016/j.scitotenv.2017.10.106 902 Zhou Z-C, Zheng J, Wei Y-Y , Chen T, Dahlgren RA, Shang X, Chen H (2017). Antibiotic 903 resistance genes in an urban river as impacted by bacterial community and 904 physicochemical parameters. Environmental Science and Pollution Research 905 24: 23753–23762 https://doi.org/10.1007/s11356-017-0032-0 906 Zhu C, Wu L, Ning D, Tian R, Gao S, Zhang B, Zhao J, Zhang Ya, Xiao N, Wang Y , Brown 907 MR, Tu Q, Ju F , Wells GF , Guo J, He Z, Nielsen PH, Wang A, Zhang Yu, Chen T, He 908 Q, Criddle CS, Wagner M, Tiedje JM, Curtis TP , Wen X, Yang Y , Alvarez-Cohen L, 909 Stahl DA, Alvarez PJJ, Rittmann BE, Zhou J (2025). Global diversity and 910 distribution of antibiotic resistance genes in human wastewater treatment 911 systems. Nature Communications 16: 4006 https://doi.org/10.1038/s41467-025-912 59019-3 913 914 .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 October 12, 2025. ; https://doi.org/10.1101/2025.10.12.681873doi: bioRxiv preprint