Horizontal and vertical gene transfer shape the plasmid host range in surface-associated microbial systems

Summary

Broad-host-range plasmids drive the spread of antibiotic resistance, particularly in surface-associated microbial systems prevalent in natural and host-associated environments. Predicting their realized host range is challenging because both transconjugant proliferation (vertical gene transfer, VGT) and conjugation (horizontal gene transfer, HGT) contribute to transconjugant diversity. Here, we hypothesized that the realized host range is determined by the interplay between VGT and HGT. We experimentally tested this hypothesis by analyzing transconjugant diversity under conditions that differ in their ability to support bacterial growth. Fast-growth conditions increased transconjugant abundance but reduced diversity, whereas slow-growth conditions supported fewer but more diverse transconjugants. We complemented these experiments with individual-based simulations that explicitly incorporated both VGT and HGT. Our results demonstrate that the realized host range is jointly governed by initial HGT events and subsequent VGT-driven expansion, highlighting the importance of integrating transfer and post-transfer dynamics when predicting plasmid-mediated antibiotic resistance spread.

Graphical abstract

Highlights

• •Vertical and horizontal gene transfer jointly shape plasmid host range
• •Individual-based simulations quantify the interplay between VGT and HGT
• •Trait heterogeneity among recipients predicts realized host range contraction
• •Growth and conjugation dynamics define plasmid host range beyond molecular limits

Molecular biology; Microbiology

Introduction

Plasmids are ubiquitous mobile genetic elements that support microbial adaptation by mediating gene transfer both vertically within and horizontally across microbial populations.^1,2 In vertical gene transfer (VGT), plasmids and their associated genes are transmitted to daughter cells during cell division via segregation control systems.^3,4 In horizontal gene transfer (HGT), plasmids move between cells via conjugation, transformation or transduction, where conjugation is typically the dominant mechanism for spreading context-dependent adaptive traits such as antibiotic resistance (AR).^2,5 Importantly, many conjugative plasmids can transfer across phylogenetic boundaries, with some exhibiting host ranges that extend beyond a single bacterial phylum.^6,7,8 This underpins their critical role in disseminating AR-encoding genes to otherwise sensitive microorganisms in diverse environments.^9,10,11,12

To persist within microbial communities, conjugative plasmids rely on both VGT and HGT. Although conjugation-mediated HGT can impose energetic burdens that may reduce host growth (VGT), the expression of this evolutionary tension is highly context-dependent and differs across taxa and environments.^4,12 In natural communities, recipient strains exhibit substantial heterogeneity in both growth capacity and conjugation propensity, and these traits can vary independently rather than conforming to a strict trade-off.¹³ As a result, plasmid persistence and spread emerge from the dynamic interplay between these heterogeneous and environment-dependent processes.^14,15,16

Surface-associated microbial systems, which drive all major biogeochemical processes, underlie many biotechnological processes,^17,18,19 and affect human health and disease,^20,21 provide especially conducive settings for plasmid transfer.²² In these spatially structured systems, cells are in a predominantly sessile state that enables them to achieve high local densities and prolonged physical contact with neighboring cells. These features facilitate the cell contact-dependent process of plasmid conjugation. Indeed, conjugation frequencies within surface-associated microbial systems are typically orders of magnitude higher than in planktonic systems.^23,24,25 Surface-associated microbial systems are therefore recognized as “hotspots” for the plasmid-mediated dissemination of AR genes in environmental settings.^{26,27,28,29,30} In such contexts, the properties of plasmids along with ecological factors such as population densities and interspecific interactions determine plasmid dynamics and the extents of VGT and HGT.

While our understanding of how surface association affects the frequency of plasmid transfer is improving, we do not yet understand how surface association affects plasmid host range. At the molecular-level, the fundamental host range of a plasmid (i.e., the set of microorganisms that a plasmid is able to persist in under ideal conditions) is determined by molecular features and constraints that control plasmid replication and stable inheritance.^31,32 Plasmid replication requires compatibility between the plasmid origin of replication and the host’s replication machinery. Stable inheritance requires high-fidelity segregation control systems that ensure the partitioning of plasmids to daughter cells during cell division, such as active partitioning systems and/or toxin-antitoxin modules that eliminate plasmid-free segregants.^33,34 Environmental factors, such as nutrient availability and environmental stress, can then restrict the fundamental plasmid host range (referred to as the realized plasmid host range) by imposing non-ideal conditions. Because plasmid persistence depends on both VGT and HGT, changes in growth conditions can shift the balance between these processes. Nutrient limitation or non-ideal temperature can repress the expression and functioning of conjugation systems and constrain the growth of newly formed transconjugants.^6,35,36,37 We therefore expect that environmental conditions that increase growth will enhance VGT in fast-growing lineages and narrow the realized plasmid host range. Therefore, to fully understand plasmid host range, we propose that it is essential to quantitatively describe plasmid behavior across diverse microorganisms while accounting for factors that can cause the fundamental and realized plasmid host ranges to deviate from each other.

Here, we propose that spatial self-organizing processes, which result in the spatial patterning of microorganisms across surfaces,^38,39,40,41 can cause deviations between the fundamental and realized plasmid host ranges. For example, self-organizing processes driven by positive metabolic interactions can promote the spatial intermixing of different microorganisms, consequently increasing the number of cell-cell contacts and fostering HGT between them.⁴² In contrast, self-organizing processes driven by negative interactions or by differences in intrinsic growth rates can drive the spatial segregation of different microorganisms, consequently reducing the number of cell-cell contacts and restricting HGT between them.⁴² Finally, self-organizing processes can create spatial patterns that prevent or promote VGT after cells acquire a new plasmid by positioning them at locations where resources are ample or limited.⁴²

While the role of spatial self-organizing processes in directing plasmid dynamics via their effects on spatial patterning is supported by experimental studies and theoretical considerations, we lack an understanding of how VGT, HGT, and their interplay determine the realized plasmid host range within such systems. The main issue is that conventional approaches for tracking the spread of antibiotic resistance within surface-associated microbial systems rely on monitoring plasmid-encoded fluorescent proteins, which can make it difficult to distinguish the contributions of VGT and HGT to fluorescence signals. Different microorganisms within these systems will have different conjugation probabilities, growth rates and inheritance stabilities,^13,43 which can lead to variations in VGT and HGT and obscure the realized plasmid host range. Here, we hypothesize that the realized plasmid host range is not governed solely by molecular compatibility, but also by the combined microorganism-specific contributions of VGT and HGT in a particular spatial context (Figure 1A). Our main expectation is that deviations between the realized and fundamental plasmid host ranges are determined by variations in VGT and HGT across different microorganisms within a particular system.

Figure 1.

Hypothesis and experiment system

(A) We hypothesize that vertical and horizontal gene transfer (VGT and HGT) are influenced by the characteristics of the potential recipient cell types and determine the proliferation and diversity of transconjugant cells. Because the potential recipient community comprises multiple cell types with varying growth traits and conjugation probabilities, we expect the resulting composition of transconjugant cells to be shaped by these cell type-specific traits.

(B) Our experimental system consists of E. coli MG1655 lacI^q-pLpp-mCherry as the plasmid donor strain and pB10 as the focal plasmid. pB10 donor cells express RFP from the chromosome and transconjugants express GFP from pB10.

To test this hypothesis, we experimentally quantified how VGT and HGT jointly affect the realized plasmid host range in a surface-associated context. We used a microbial community from the activated sludge basin of a wastewater treatment plant (WWTP) and the green fluorescent protein (GFP)-encoding broad-host-range plasmid pB10 as a model system. This is particularly relevant as WWTP effluents are recognized as a major source of AR genes entering the environment.^44,45 We performed surface-associated conjugation assays using an engineered strain of Escherichia coli as the pB10 donor and the WWTP microbial community as the set of potential recipients, which allowed us to distinguish plasmid donor, potential recipient and transconjugant cells in a spatially explicit manner (Figure 1B). We assessed the abundance, composition and diversity of transconjugants using a combination of flow cytometry (FC) in conjunction with fluorescence-activated cell sorting (FACS), 16S rRNA amplicon sequencing, and fluorescence microscopy. We further developed and applied an individual-based simulation model that tracks plasmid donor, potential recipient and transconjugant lineages during surface-associated growth. The model allows us to independently manipulate the conjugation probability and transconjugant growth rate to impose an interplay between VGT and HGT and to evaluate how these parameters direct plasmid dynamics. Our integration of experiments and simulations enables a quantitative understanding of the determinants of the realized plasmid host-range in surface-associated microbial systems and provides a mechanistic framework for understanding how the interplay between VGT and HGT shapes host-range estimates.

Results

Environmental conditions determine realized plasmid host range

We first tested how environmental conditions that affect growth rates affect the total abundance of transconjugant cells when the WWTP cells were the potential recipient cells. To vary environmental conditions, we used synthetic wastewater medium at normal strength (1×SWW) and at 10-fold concentration (10×SWW) as well as nutrient-rich LB. Their effects on growth are reported in the corresponding growth curves (Figure S1). We quantified the number of transconjugant cells as the number of cells exceeding a fluorescence threshold measured by FC-FACS, which corresponds to the acquisition of pB10 (Figure S2). We found that the proportion of WWTP cells that became transconjugant cells was higher for environmental conditions that support fast growth (Figure 2A) (two-way ANOVA with Holm correction, 1×SWW vs. 10×SWW, p = 5.4 × 10⁻⁵; 10×SWW vs. LB, p = 9.7 × 10⁻³; 1×SWW vs. LB, p = 9.7 × 10⁻³ [n = 3]). We note that these experiments do not isolate the influence of nutrient richness from temperature, as the 1xSWW assays were performed at 25 °C whereas the 10xSWW and LB assays were performed at 37 °C. Thus, we interpret the differences in transconjugant abundance and diversity as outcomes of multiple ecological filters rather than nutrient richness or temperature alone. To confirm that these differences reflected changes in the function of the conjugation machinery during the initial conjugation events rather than differences in early growth, we performed short-term surface mating assays under high-cell density conditions (Figure S3). We confirmed that the conjugation frequency itself was not significantly affected by the different environmental conditions (two-way ANOVA with Holm correction, 1×SWW vs. 10×SWW, p = 0.9; 10×SWW vs. LB, p = 0.1; 1×SWW vs. LB, p = 0.2 [n = 3]). Thus, environmental conditions primarily affect the proliferation of transconjugant cells after acquiring pB10 (VGT) rather than the probability of pB10 conjugation (HGT).

Figure 2.

Transconjugant proportions and diversities after surface-associated conjugation assays for different environmental conditions

(A) Proportion of transconjugant cells relative to total cells after surface-associated conjugation assays using the WWTP community as the potential recipient cell population. We conducted conjugation assays on 1×SWW, 10×SWW, or LB agar plates using E. coli MG1655 lacI^q-pLpp-mCherry as the pB10 donor strain.

(B) Relative abundances of bacterial class in the total potential recipient cell population (T) and the transconjugant cell population (TC) as identified by 16S rRNA gene sequencing. We separated and identified TC cells using FC-FACS-sorting of GFP-positive cells.

(C) Normalized Shannon index of the transconjugant populations after surface-associated conjugation assays on 1×SWW, 10×SWW, or LB agar plates. We normalized the Shannon index of the TC populations to their corresponding T populations.

(D) Principal coordinate analysis (PCoA) based on weighted UniFrac distances of T and TC populations after surface-associated conjugation assays on 1×SWW, 10×SWW, or LB agar plates.

(E) Phylogenetic tree of transconjugant ASVs detected after surface-associated conjugation assays on 1×SWW, 10×SWW, or LB agar plates. The outer colored box denotes the bacterial phylum of each ASV, corresponding to the phylum-level groupings shown in panel (B). The inner heatmap box aligned with each tip shows the log₁₀ fold-changes in ASV abundance (TC relative to T) across the three conditions. For (A and C), each point is an independent biological replicate (n = 3), horizontal bars are the means, error bars are ±1 standard deviation, and asterisks indicate statistically significant differences between the means based on two-way ANOVA with Holm correction (∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001, ns = not significant). For (D), each point is an independent biological replicate (n = 3).

We next used microscopy imaging to evaluate the morphology and sizes of the transconjugant cells for the different environmental conditions (Figure S4). The fluorescently labeled transconjugant cells exhibited diverse morphologies, including rod-shaped, coccoid and filamentous forms. This demonstrates that pB10 can indeed transfer to a wide variety of different microorganisms within the WWTP population under our experimental conditions.

We next analyzed the taxonomy of the transconjugant cells that emerged from the WWTP cells by sequencing the 16S rRNA genes of the total WWTP population and the transconjugant population as defined based on FC-FACS-sorted GFP-positive cells (Figures 2B–2E; Table S2; Figure S5). Although we detected a variety of taxonomic groups within both the WWTP and transconjugant populations, we found that both populations were predominantly composed of Gammaproteobacteria and Bacteroidia (Figure 2B). To independently validate that pB10 does indeed have a broad host range, we conducted surface-associated conjugation assays using pure cultures of multiple taxonomic classes as potential recipient strains (Figure S6). pB10 successfully transferred to E. coli, Flavobacterium sp., and Alcaligenes faecalis but not to Bacillus subtilis. These results support our conclusion that pB10 does indeed have a broad (but not universal) host range under our experimental conditions.

Despite the presence of multiple phylogenetic groups within the transconjugant population across the different environmental conditions, the transconjugant diversity varied significantly with environmental conditions. Transconjugant diversity as assessed by the normalized Shannon index was highest for 1xSWW and lowest for LB (two-way ANOVA with Holm correction, 1×SWW vs. 10×SWW, p = 0.06; 10×SWW vs. LB, p = 8.2 × 10⁻⁴; SW vs. LB, p = 1.2 × 10⁻⁵, n = 3) (Figure 2C), indicating that LB resulted in the dominance of fewer transconjugant lineages. We further performed weighted UniFrac-based principal coordinates analysis on both the total WWTP and transconjugant populations and observed clustering patterns that clearly reflected environmental condition-dependent shifts (Figure 2D). Both the total WWTP and the transconjugant populations clustered closely together for 1×SWW, indicating similar compositions. By contrast, the total WWTP and transconjugant populations formed separate clusters for 10×SWW or LB, and these were clearly distant from those for 1×SWW. Thus, while stable inheritance depends on reliable segregation control systems, the resulting community structure is governed primarily by intrinsic bacterial growth rates. 1×SWW resulted in slower growth that limited competitive exclusion, whereas 10×SWW and LB resulted in the rapid growth of fast-growing taxa that generated stronger shifts in community composition driven primarily by intrinsic growth differences rather than plasmid acquisition. This demonstrates that environmental conditions selected for distinct transconjugant populations. Additionally, the WWTP and transconjugant populations for 10×SWW or LB tended to cluster apart, indicating divergence between these two populations. Thus, environmental factors not only affect transconjugant proliferation but also drive shifts in the compositions of the transconjugant populations.

We next compared the taxonomic compositions of the total WWTP and transconjugant populations at the amplicon sequence variant (ASV) level for all the tested environmental conditions (Figure 2E). 1×SWW resulted in multiple ASVs with higher relative abundance in the transconjugant population than in the total WWTP population, suggesting broad host uptake across phylogenetically diverse microorganisms. In contrast, 10×SWW and LB resulted in transconjugant populations enriched in a narrower subset of ASVs, indicating that only a limited set of transconjugant lineages contributed to the observed transconjugant population. For LB, for example, the transconjugant population was primarily composed of the ASVs affiliated with genera such as Pseudomonas, and Aeromonas, whereas many other ASVs had minimal representation. Together, these findings demonstrate that environmental conditions that promote slower growth also facilitate broader observable pB10 dissemination across the WWTP community, while environmental conditions that promote faster growth result in the proliferation of a restricted subset of highly permissive transconjugant lineages that skew host-range estimates.

Effect of environmental conditions on plasmid host range is caused by differences in growth

Given that the proportions of the WWTP cells that became transconjugant cells were higher for environmental conditions that promote faster growth (Figure 2A), we hypothesized that VGT substantially contributed to transconjugant abundance during the surface-associated growth experiments. To directly observe VGT, we acquired sequential images of transconjugant cells during surface-associated growth using fluorescence microscopy (Figure 3A). Within a few hours of incubation, a few recipient cells began to exhibit green fluorescence, indicating successful transfer of pB10. Over time, the transconjugant cells continued to grow and formed microcolonies. However, not all transconjugant lineages exhibited robust growth; some had limited or no observable growth during this period. (Figure 3B). These observations indicate heterogeneity in transconjugant lineage growth during surface-associated growth, which may reflect lineage-specific differences in growth rate, conjugation frequency of pB10, and/or interactions with neighboring cells.

Figure 3.

Transconjugant growth during surface-associated conjugation assays for different environmental conditions

(A) Representative fluorescence microscopy images of transconjugant cells during surface-associated conjugation assays on LB agar plates. E. coli MG1655 lacI^q-pLpp-mCherry is the pB10 donor strain and show red fluorescence. Transconjugant cells are green. The time indicated in the images refers to the point at which transconjugant cells first became detectable.

(B) Normalized microcolony area (A/a₀) plotted as a function of time during the surface-associated conjugation assays on LB agar plates. A is the total microcolony area and a₀ is the initial transconjugant area. Connected data points are for individual colonies (n = 12).

(C) Microcolony area at the endpoint of the mating assay (t = 24 h) for different environmental conditions. The half-violin and scatterplots present the sample distribution and individual microcolony measurements for surface-associated conjugation assays on different medium (n_1xSWW = 880, n_10xSWW = 664, n_LB = 1,070, for microcolony number). We performed each experiment at least three independent experiments. Horizontal bars are the mean microcolony areas, error bars are the 99% confidence intervals, and asterisks indicate statistically significant differences between the means based on two-way ANOVA with Holm correction (∗∗p < 0.01, ∗∗∗∗p < 0.0001, ns = not significant).

To quantify the impact of environmental conditions on transconjugant proliferation, we measured the size of microcolonies at the end of the experiments for the three environmental conditions (Figure 3C). We observed transconjugant microcolonies for all conditions, confirming that post-transfer growth (i.e., VGT) contributed to overall transconjugant abundance. We also found that microcolony size increased with environmental conditions that support faster growth, with statistically significant differences across all conditions (1×SWW vs. 10×SWW, p = 7.9 × 10⁻⁴; 10×SWW vs. LB, p = 8.4 × 10⁻⁵; 1×SWW vs. LB, p = 2.6 × 10⁻⁵; two-way ANOVA tests with Holm corrections). Although the overall microcolony size distributions were modestly different (Kolmogorov-Smirnov tests; D < 0.5 and p = 0 for all pairwise comparisons), the use of LB exhibited the widest distribution of microcolony size, suggesting more variable transconjugant proliferation under the environmental conditions that support the fastest growth. We note that microcolony size may also reflect contributions from plasmid donor and potential recipient growth, but this does not alter our overall interpretation. These results collectively demonstrate that environmental conditions that promote faster growth not only enhance transconjugant proliferation (i.e., VGT) but also introduce greater heterogeneity in proliferation outcomes among lineages. This heterogeneity likely contributes to the variability in host-range estimates across different environmental conditions and underscores the importance of accounting for VGT when evaluating and predicting plasmid spread in surface-associated systems.

Growth and conjugation probability jointly determine transconjugant proliferation

To disentangle the relative contributions of VGT and HGT to transconjugant proliferation during surface-associated growth, we performed individual-based simulations with a computational model comprised of one plasmid donor cell type and one potential recipient cell type (Figures 4A, Video S1). This approach allows us to control growth rate and conjugation probability independently of other environmental factors (e.g., temperature, nutrient composition, and nutrient richness), which are difficult to experimentally separate in surface-associated conjugation assays. Thus, the simulations provide a mechanistic baseline for interpreting the experimentally observed patterns. In our framework, transconjugant cells arise through direct plasmid acquisition (HGT) and proliferate via growth (VGT). The model tracks the numbers of transconjugant cells, HGT events, and VGT events over time until reaching a fixed total number of cells of 20,000, which mimics the complete consumption of resources during surface-associated growth.

Figure 4.

Individual based simulations of plasmid dynamics with one potential recipient cell type

(A) Schematic of simulations with one potential recipient cell type capturing the dynamics of vertical gene transfer (VGT) and horizontal gene transfer (HGT). Plasmid donor (red) and potential recipient (white) cells grow at defined rates (u_d, and u_r respectively) and plasmid transfer occurs with a defined probability (P_c). After acquiring the plasmid, the resulting transconjugant cell (green) grows at a rate u_t. HGT events are counted when a plasmid is transferred from a plasmid donor cell to a potential recipient cell, whereas VGT events are counted when a transconjugant cell divides and passes the plasmid to its progeny through cell growth.

(B, and D) Representative simulation images at the endpoint of the simulations showing outcomes for different u_r/u_d with a fixed P_c of 0.05 (B) and different P_c with a fixed u_r/u_d of 1.0 (D). Red, white and green cells are plasmid donor, potential recipient, and transconjugant cells, respectively. Scale bars, 30 μm. (C, and E) Transconjugant proportion at the endpoint of the simulations as a function of u_r/u_d (C) and P_c.

(E). We quantified the transconjugant proportion by dividing the number of transconjugants by the total number of cells. Within the transconjugant population, we plotted the HGT-derived transconjugant cells and VGT-derived transconjugant cells separately. Each data point is an independent simulation (n = 3), horizontal bars are the means, and error bars are ±1 standard deviation.

(F and G) Heatmap for the transconjugant proportions and VGT/HGT ratio is a function of u_r/u_d and P_c. The VGT/HGT ratio is the number of VGT-derived transconjugants divided by the number of HGT-derived transconjugants at the simulation endpoint.

To evaluate how the growth of the potential recipient cell type affects transconjugant dynamics, we varied the growth rate of the potential recipient cell type (u_r) relative to that for the plasmid donor cell type (u_d) while keeping the conjugation probability (P_c) constant at 0.05. This parameter change altered the relative contributions of VGT and HGT. As u_r/u_d increased, the total number of transconjugant cells increased (Figures 4B and 4C; S7 A). Likewise, as u_r/u_d increased, the proportions of VGT- and HGT-derived transconjugant cells increased (Figures 4B and 4C). Together, these results demonstrate that faster-growing potential recipient cells are more likely to successfully acquire the plasmid via conjugation (HGT) and that the proliferation of transconjugant cells (VGT) becomes a dominant factor as the growth rate of the potential recipient cells increases. Notably, the number of transconjugant cells that emerged from VGT surpassed the number that emerged from HGT when u_r/u_d ≥ 0.5, marking a shift from transfer (HGT)-driven to growth (VGT)-driven dynamics. These findings indicate that, while initial transfer events are necessary to seed the transconjugant population, subsequent growth through VGT becomes the primary mechanism of transconjugant proliferation under favorable growth conditions.

We next assessed the effect of the conjugation efficiency on plasmid proliferation by varying P_c while keeping the growth rate of the potential recipient cell type fixed at u_r/u_d = 1.0. This again allowed us to alter the relative contributions of VGT and HGT and measure the consequences. As P_c increased, the total number of transconjugant cells increased, confirming that the conjugation probability contributes directly to the overall transconjugant population size (Figures 4D and 4E; S7B). However, the number of HGT-derived transconjugant cells was only weakly affected by P_c, while the number of VGT-derived transconjugant cells was strongly affected (Figures 4D and 4E). This indicates that increasing P_c indirectly stimulated the number of overall transconjugant cells primarily by forming more HGT-derived transconjugant cells at early times that subsequently proliferated via VGT. Despite increasing P_c, transconjugant proliferation remained modest, highlighting that transfer alone is insufficient to drive large-scale plasmid dissemination without accompanying VGT.

To evaluate the combined effects of growth and the conjugation probability, we performed a full parameter analysis by systematically varying u_r/u_d and P_c. We found that both factors positively affect the number of overall transconjugant cells (Figure 4F), but the transconjugant growth rate had a stronger overall effect than the conjugation probability. To quantify the relative contributions of VGT and HGT, we calculated the VGT/HGT ratio (Figure 4G) defined as the number of transconjugant cells generated by vertical proliferation divided by those generated by horizontal conjugation events. A VGT/HGT ratio >1 indicates that transconjugant proliferation is dominated by VGT, whereas a ratio <1 indicates that it is dominated by HGT. Notably, for scenarios where both P_c and u_r/u_d have high values, the system transitions from HGT-limited to VGT-driven proliferation. Together, these results demonstrate that transconjugant proliferation outcomes are jointly determined by the efficiency of conjugation and the growth capacity of recipient-derived transconjugants. While HGT is essential for initiating plasmid spread, VGT is the dominant mechanism of contributing to new transconjugant cells when growth is rapid. The proliferation of transconjugant cells via VGT continues until the total population reaches a predetermined saturation point, reflecting typical endpoint conditions of surface-associated conjugation assays.

Growth and conjugation probability jointly determine realized plasmid host range

Our experimental results indicated that environmental conditions that promote fast growth reduce the realized plasmid host range due to varying proliferation of a few transconjugant lineages (Figures 2 and 3). This leads to the expectation that variation in growth rate and conjugation probability among different potential recipient cell types will strongly influence the realized plasmid host range. To test this, we performed individual-based simulations incorporating one hundred distinct potential recipient cell types, where each cell type is assigned a set growth rate (u_r,i) and conjugation probability (P_c,ᵢ) (Figures 5A and 5B). To set the parameters and simulate heterogeneity, we assigned u_r,i and P_c,i values from four parameter distribution types: uniform (U), linear (L), small-skew log-scale (SL), and large-skew log-scale (LL) (Figure S8). We used log-scale distributions to introduce wider variation among the traits of the potential recipient cell types (Figure 5C). We generated potential recipient populations for all combinations of u_r,i and P_c,i distribution types, producing groups that varied in growth rate only, conjugation probability only, or in both traits simultaneously. We evaluated all combinations of distribution types for u_r,i and P_c,i and calculated the Shannon index and VGT/HGT ratio for the endpoint transconjugant populations (Figures 5D and 5E; S9 and S10). We positively rank-correlated u_r,i and P_c,i by independently sorting the sampled values in ascending order and pairing them by index (i.e., u_r,1 ≤ u_r,2 ≤ … ≤ u_r,100 and P_c,1 ≤ P_c,2 ≤ … ≤ P_c,100).

Figure 5.

Individual-based simulations of plasmid dynamics with one hundred recipient cell types

(A) Schematic of simulations with one hundred potential recipient cell types capturing the dynamics of vertical gene transfer (VGT) and horizontal gene transfer (HGT). We assigned each potential recipient cell type a unique growth rate (u_r,i) and conjugation probability (P_c,i), where the subscript i denotes the recipient cell type index.

(B) Representative simulation image at the endpoint of the simulation showing the outcome for uniform parameters (u_r,i = 0.5 and P_c,i = 0.01). Each color corresponds to a different recipient cell type that became a transconjugant cell. Plasmid donor and potential recipient cells are gray and white, respectively. Scale bars, 10 μm.

(C) Distribution histograms of the traits (u_r,i and P_c,i) of the potential recipient cell types that we used in the simulations. The traits followed either a small-skew log-scale (SL) or large-skew log-scale (LL) distribution.

(D and E) Simulation results with defined parameters for each of the one hundred different potential recipient cell types. We show the Shannon index (D) and VGT/HGT ratio (E) for each set of parameter distributions. We performed simulations for u_r,i and P_c,i sampled from a uniform (U), small-skew log-scale (SL), or large-skew log-scale (LL) distributions. Each plot is for an independent simulation replicate (n = 5). The middle horizontal line is the mean and the upper and lower lines are ±1 standard deviation.

(F and G) Simulation results with randomly assigned parameters for each of the one hundred different potential recipient cell types.

(F) Correlation between the Shannon index and the VGT/HGT ratio.

(G) Correlation between the Shannon index and the coefficient of variation (CV) of u_r,i and P_c,i. In these simulations, we randomly assigned each potential recipient cell type with a value of u_r,i and P_c,i sampled from the specified distributions. Each plot is the value from a single simulation with a different parameter assignment. Colors indicate the distribution that we used to obtain the parameters: yellow, we sampled both u_r,i and P_c,i from SL; purple, we sampled both parameters from LL. We conducted the simulations 100 times for each parameter distribution (100 yellow points and 100 purple points). Lines are linear regressions fit for each group in (F) and for all combined data in (G).

When we assigned potential recipient cell types with uniform parameters for both u_r,i and P_c,i, we observed the highest transconjugant diversity with the Shannon index remaining close to its maximum value. Introducing skew to either trait reduced diversity, and the reductions were largest for the large-skew log distributions (Figure 5D) (U_ur–U_pc vs. SL_ur–U_pc,ᵢ, p = 1.9 × 10⁻⁸; U_ur–U_pc vs. SL_ur–U_pc, p = 4.7 × 10⁻¹⁰; two-way ANOVA tests with Holm corrections, [n = 5]). Skew in P_c,i tended to reduce diversity more than skew in u_r,i, suggesting that variation in conjugation probability may have a somewhat larger role in concentrating plasmid spread into fewer potential recipient cell types (SL_ur–U_pc vs. U_ur– SL_pc, p = 6.0 × 10⁻⁴; LL_ur–U_pc vs. U_ur– LL_pc, p = 1.1 × 10⁻⁷; two-way ANOVA tests with Holm corrections, [n = 5]). When we sampled both traits from skewed distributions, the effects were synergistic; the large-skew combination produced the lowest diversity values. Although applying a linearly distributed trait alone (L) did not significantly change diversity for either u_r,i or P_c,i, combining a linear distribution for one trait with a log-skewed distribution for the other reduced diversity (Figure S11). These results indicate that strong heterogeneity in growth rate and conjugation probability shifts the balance toward post-transfer proliferation (VGT) of a limited set of highly permissive recipient types.

To further examine whether secondary transfer events alter these dynamics, we performed additional simulations in which transconjugants were allowed to act as plasmid donors. Allowing secondary conjugation produced only a modest decrease in the VGT/HGT ratio and did not significantly influence transconjugant diversity (Figure S12), indicating that primary donor-mediated plasmid transfer remained the dominant determinant of the realized host range. Furthermore, to examine whether morphological heterogeneity among potential recipients could affect our main outcomes, we performed simulations in which the initial cell length varied across potential recipient cell types, spanning a large log-scale distribution (Figure S13). We found that cell length diversity does not meaningfully alter the contraction of realized host range. These findings indicate that factors such as secondary conjugation and morphology exert minimal influence on the realized host range, highlighting that variations in growth rate and conjugation probability are the main determinants.

Based on our experimental findings, we concluded that VGT has a larger role in reducing diversity in surface-associated systems rather than HGT (Figures 2 and 3). However, in the defined-parameter simulations, the Shannon index does not associate with the VGT/HGT ratio (Figures 5D and 5E), likely because the fixed pairing of u_r,i and P_c,ᵢ limited variation in their combined effects across potential recipient cell types. To test whether greater heterogeneity in trait combinations would strengthen this relationship, we simulated systems in which each potential recipient cell type was randomly assigned u_r,i and P_c,ᵢ values from either SL or LL parameter distributions (i.e., u_r,i and P_c,ᵢ were independent of each other) (Figures 5F and 5G). For both the SL and LL simulations, the Shannon index was negatively correlated with VGT/HGT ratio with a stronger effect for the LL than the SL distributions (Figure 5F) (SL, r = −0.78; LL, r = −0.85, Pearson correlation coefficient [n = 100]). This indicates that in highly skewed communities, the balance between VGT and HGT exerts a stronger influence on the realized plasmid host range than in more evenly distributed communities. The VGT/HGT ratio was influenced by both growth rate and conjugation probability (Figure 4F), with the product of u_r,i and P_c,ᵢ positively correlated with VGT/HGT (Figure S14). These results highlight that VGT and HGT jointly determine estimates of plasmid host range in diverse systems, where HGT sets the initial breadth of potential recipient cell types that acquire the plasmid and VGT governs the extent to which those transconjugants proliferate.

Overall, heterogeneity in the combined u_r,i and P_c,ᵢ across potential recipient cell types is a primary determinant of transconjugant diversity and the realized plasmid host range. Consistent with this, the coefficient of variation (CV) of the product of u_r,i and P_c,ᵢ across recipient cell types was negatively correlated with the Shannon index for all parameter sets (Figure 5G) (all samples, r = −0.82, Pearson correlation coefficient, [n = 200]). In systems composed of multiple cell types with different growth rates and conjugation probabilities, the distributions of these traits, whether even or skewed, had a decisive role in shaping the final diversity of transconjugant cells. Evenly distributed traits favor broader representation of potential recipient cell types in the transconjugant population, whereas strong skew in either or both traits drives dominance by limited numbers of highly permissive and fast-growing lineages, thereby narrowing the realized plasmid host range.

Discussion

Broad-host-range plasmids are key vectors that enable microorganisms to adapt to environmental challenges and can lead to the rapid dissemination of AR within microbial systems.^7,46 In this study, we quantitatively evaluated the roles of VGT, HGT, and their interplay on transconjugant abundance and the realized plasmid host range. By integrating surface-associated conjugation experiments with individual-based simulations, we found that transconjugant abundance and diversity within surface-associated microbial systems are not solely determined by conjugation events (HGT) but are strongly influenced by the subsequent growth of transconjugant cells (VGT). This finding provides a framework to more accurately evaluate plasmid spread and to predict their consequences on human and environmental health.

Our results demonstrate that once a plasmid is horizontally acquired by new transconjugant cells (HGT), the subsequent growth and division of those transconjugant cells (VGT) is a defining factor determining plasmid spread and the realized plasmid host range. Environmental conditions that promote faster growth produced markedly larger transconjugant populations than those that restrict growth (Figure 2A), yet the larger transconjugant populations were less phylogenetically diverse. Because HGT probabilities were independent of environmental conditions (Figure S3), this demonstrates that conditions that promote faster growth also promote more vigorous VGT by a subset of lineages, thus reducing the apparent realized plasmid host range. Although conjugation machinery imposes an energetic burden, growth promoting conditions can mitigate these costs, enabling rapid proliferation of transconjugants, especially among taxa with inherently high growth rates or efficient plasmid maintenance capabilities. Individual-based simulations support this principle; when we held the conjugation probability constant while increasing the transconjugant growth rate, we observed a monotonic rise in transconjugant abundance (Figures 4B and 4C). Although both transconjugant growth rate and conjugation probability jointly shape outcomes, transconjugant growth rate is the dominant factor. These results align with theoretical considerations of plasmid dynamics, which predict that long-term plasmid persistence in the absence of positive selection depends on a balance between VGT and HGT.^47,48 Likewise, pB10 introduced at an initial frequency of ∼10⁻⁷ increased to ∼13% when nutrients were replenished, whereas no invasion occurred under nutrient limitation or in well-mixed liquid culture.⁴⁹ Collectively, our findings support a general model in which favorable growth conditions shift plasmid dissemination from an HGT-dominated to a VGT-dominated regime, allowing plasmids to persist even when HGT is infrequent.

Broad-host-range plasmids such as pB10 are known to transfer across taxonomically distant bacteria in environmental microbial communities.⁶ In our 16S rRNA amplicon analysis, pB10 was transferred to multiple phyla in growth repressing conditions, whereas the realized plasmid host range was narrower in growth promoting conditions. This narrowing likely reflects the greater post-transfer proliferation of a subset of lineages in surface-associated communities (Figure 3C). A broad fundamental plasmid host range does not guarantee that every permissive taxon contributes equally to the transconjugant pool,⁵⁰ and differences in conjugation probability may further influence the realized plasmid host range. We therefore infer that the realized plasmid host range is shaped not only by which microorganism can acquire the plasmid, but also by the distribution of microorganism-specific traits that determine their post-acquisition success. In our simulations, heterogeneity in growth rate and conjugation probability consistently reduced transconjugant diversity (Figure 5D), indicating that trait distributions in the recipient pool act as an ecological filter that narrows the fundamental plasmid host range to the realized plasmid host range. Variation in these traits within the potential recipient community can shift plasmid spread toward a smaller subset of highly permissive, fast-growing lineages. Communities with more even distributions of these traits are likely to maintain a broader representation of the potential plasmid recipient population in the transconjugant population, whereas strong skews favor dominance by a few lineages. Together, these insights highlight the importance of accounting for ecological trait variation when interpreting plasmid host range estimates and predicting plasmid spread in polymicrobial systems.

Our simulations further reveal that VGT dominance is governed jointly by the potential recipient growth rate and conjugation probability in surface-associated microbial systems. Although high cell density increases opportunities for cell-cell contacts, potential recipient cells with higher conjugation probabilities continue to seed more nascent transconjugant cells early, which then proliferate via growth and eventually dominate the population. In our model, the conjugation probability is the per-cell contact likelihood of plasmid transfer from a plasmid donor cell to a potential recipient cell and subsequent expression of GFP. From a biological perspective, this effective probability reflects both mating efficiency and the likelihood that the plasmid establishes and is retained long enough to replicate and be inherited. These processes are influenced by the host’s physiological state, with metabolically active cells more likely to maintain plasmids than dormant or slow-growing cells.⁵¹ This explains why taxa that combine rapid growth with high plasmid transfer capacity strongly influence outcomes. In our simulations, imposing skew in both the potential recipient growth rate and conjugation probability acted synergistically to reduce transconjugant diversity (Figure 5D). Moreover, in simulations with random parameter combinations, the sum of the joint traits (∑u_r,iP_c,i) in the potential recipient population positively correlated with VGT dominance (Figure S14), leading to the loss of diversity (Figure 5G). Together, these findings indicate that differences at the moment of plasmid acquisition are magnified through subsequent VGT-driven proliferation, making VGT a decisive force in contracting the realized plasmid host range in surface-associated microbial systems.

Our findings have important ecological and public health implications, particularly for the dissemination of AR genes. Microenvironments that have high cell densities and promote rapid growth, such as the human gut and activated sludge flocs, act as hotspots for conjugative plasmid transfer.^52,53 These conditions foster both cell growth and the physical contacts required for VGT and HGT, thereby accelerating the spread of AR genes. Effective control strategies must therefore address not only the inherent transferability of plasmids but also the environmental factors that enhance bacterial proliferation and cell-cell contacts. By integrating empirical observations with mechanistic simulations that consider the interplay between VGT and HGT, our study provides a mechanistic framework for evaluating the risks of plasmid spread in complex microbial systems.

Limitations of the study

Despite the complementary insights gained from our experimental and modeling approaches, several limitations should be acknowledged. Because detection relies on fluorescence, we may not capture all transconjugants, as only metabolically active transconjugants expressing GFP above a certain threshold are captured, possibly underestimating the host range of the plasmid.

Regarding the model formulation, although our simulations incorporate variation in growth rate, conjugation probability, and the availability of secondary plasmid donors, it remains challenging to fully reproduce experimental outcomes because growth and conjugation capabilities can differ widely among taxa and are strongly shaped by environmental conditions.

Furthermore, we varied nutrient composition and temperature together in our experimental design, and the observed patterns should therefore be interpreted as the combined outcome of multiple ecological filters. In principle, testing these traits for every community member would be required to parameterize a fully realistic model. In addition, our simulations do not currently include other important biological processes such as plasmid loss or cell motility, even though these processes may influence conjugation dynamics in natural systems.^41,54,55,56 Incorporating these features into future models is important for capturing the full complexity of plasmid-mediated gene flow in heterogeneous microbial communities.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Mamoru Oshiki (oshiki@eng.hokudai.ac.jp).

Materials availability

All chemicals were purchased from commercial resources and used as received. Plasmid DNA and cell strains constructed in this study are available from the lead contact upon request.

Data and code availability

• •Raw sequence data have been deposited in the DDBJ Sequence Read Archive under accession number PRJDB35968. Processed data supporting the findings of this study have been deposited in figshare (https://doi.org/10.6084/m9.figshare.30102145).
• •All analysis scripts used in this study have been deposited in figshare (https://doi.org/10.6084/m9.figshare.30102145).
• •Any additional information required to reanalyze the data reported in this article is available from the lead contact upon request.

Acknowledgments

We thank Dr. Yoshizawa for technical assistance of FACS analysis and Prof. Dr. Eva M. Top for kindly providing the pB10::gfp plasmid. This work was financially supported by Japanese Society for the Promotion of Science (JSPS) (24KJ0001 for K.T. and 23H02114 and 25K22850 for M.O.), Japan Science and Technology Agency FOREST Program (JPMJFR216Z for M.O.), Kurita Water and Environment Foundation (24H024 and 25K002 for K.T.), and Hokkaido University One Health Research grant.

Author contributions

Conceptualization, K.T., M.O., and D.R.J.; methodology, K.T., K.O., and K.H.; investigation, K.T., K.O., K.H., and T.O.; simulations, K.T. and T.O.; image analysis, K.T. and K.O.; visualization, K.T.; funding acquisition, K.T. and M.O.; supervision, M.O. and D.R.J.; writing – original draft, K.T., M.O., and D.R.J.; writing – review & editing, K.T., K.O., K.H., T.O., S.O., M.O., and D.R.J.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Escherichia coli MG1655 (wild type)	Lab stock; previous publications	MG1655
E. coli MG1655 lacIq-pLpp-mCherry	This study	MG1655 lacIq-pLpp-mCherry
E. coli DH5α (F−, lacZ m1, recA)	Takara Bio	Cat# 9057
E. coli β2163 (RP4 conjugation helper; ΔdapA)	Ref.57	β2163
E. coli β2163 lacIq-pLpp-mCherry	This study	β2163 lacIq-pLpp-mCherry
Alcaligenes faecalis NBRC13111	NBRC	NBRC13111
Flavobacterium sp. NBRC101303	NBRC	NBRC101303
Bacillus subtilis NBRC131719	NBRC	NBRC131719
Chemicals, peptides, and recombinant proteins
Lysogeny broth (LB Lennox)	Nacalai Tesque	Cat# 20066-95
Peptone, meat extract (for SWW media)	Nacalai Tesque	Cat# 26442-75
Critical commercial assays
MightyPrep Reagent for DNA	Takara Bio	Cat# 9182
In-Fusion HD Cloning Kit	Takara Bio	Cat# 639650
Mighty Mix DNA Ligation Kit	Takara Bio	Cat# 6023
Deposited data
Sequence data	This study	PRJDB35968
Custom image-analysis scripts	This study	https://doi.org/10.6084/m9.figshare.30102145
Custom CellModeller simulation scripts	This study	https://doi.org/10.6084/m9.figshare.30102145
Oligonucleotides
Primer: 515F: GTGCCAGCMGCCGCGGTAA	Ref.58	N/A
Primer: 806R: GGACTACHVGGGTWTCTAAT	Ref.58	N/A
Primer: lacI-F: GCATGCCTGCAGGTC GACGACACCATCGAATGGTGCAAAA	This study	N/A
Primer: lacI-R: TGTAGCGTTACAAGTA TAACACAAAGTTTTTTATGTTGAGAAT ATTTTTTTGATGGGAAGGCACTTATTT TCACTGCCCGCTTTCCAGTC	This study	N/A
Primer: mCherry-F: GTTATACTTGTAA CGCTACATCTAGAATTAAAGAGGAG AAATTAAGCATGGTGAGCAAGGGC GAGGAGGA	This study	N/A
Primer: mCherry-R: TTACTTGTACA GCTCGTCCATGCCG	This study	N/A
Primer: lacI-pLpp-mCherry-F (SmaI insertion): ATACCCGGGGCATGCC TGCAGGTCGACGA	This study	N/A
Primer: lacI-pLpp-mCherry-R: AATC TCGAGCGGCCAGTGAATTCGAGCTC	This study	N/A
Primer: lacI-pLpp-mCherry_Check-F: GATGACGGTTTGTCACATGGA	Ref.59	N/A
Primer: lacI-pLpp-mCherry_Check-R: GATGCTGGTGGCGAAGCTGT	Ref.59	N/A
Recombinant DNA, Plasmids
pB10:gfp (IncP-1β; TetR StrR SulR HgR; mini-Tn5 PA1/O4/O3:gfp)	Ref.60	pB10:gfp
pMAL expression vector (MBP-fusion; AmpR)	Addgene	Plasmid# 161783
MBP-mCherry expression plasmid (AmpR)	Addgene	Plasmid# 29747
pUC19-lacIq-pLpp-mCherry	This study	N/A
pGRG36 (Tn7 delivery vector; AmpR)	Addgene	Plasmid #16666
pGRG36-lacIq-pLpp-mCherry	This study	N/A
Software and algorithms
QIIME2 (v2021.4)	qiime2.org	https://qiime2.org
DADA2	Callahan et al.	https://benjjneb.github.io/dada2
SILVA 138 classifier	SILVA database	https://www.arb-silva.de
R (v4.4.1)	CRAN	https://cran.r-project.org
Python 3.9	Python Software Foundation	https://www.python.org
CellModeller v4.3	Grover/Jenkinson labs	https://cellmodeller.github.io

Experimental model and study participant details

Bacterial strains and plasmids

We used plasmid pB10gfp (referred to as pB10), which belongs to the IncP-1β subgroup, as the focal plasmid for all our experiments. This plasmid contains a Tn5-Km-P_A1-04/03:gfp cassette that enables us to repress GFP fluorescence in the pB10 donor strain via the lacI repressor gene.⁶⁰

We used a derivative of Escherichia coli MG1655 or β2163 that contains a chromosomally-integrated lacI repressor gene and mCherry (encodes for red fluorescent protein [RFP]) as the pB10 donor strain. Thus, the pB10 donor strain expresses RFP and lacI from the chromosome and represses GFP from pB10. Only transconjugants that lack the lacI repressor will express GFP. We describe the construction of this strain in detail in the below section (Construction of fluorescent donor strains). When plasmid maintenance was required, we supplemented the growth media with 100 μg mL⁻¹ ampicillin, 10 μg mL⁻¹ tetracycline, 50 μg mL⁻¹ kanamycin, and/or 100 μg mL⁻¹ diaminopimelic acid (DAP) as appropriate. For pure-culture conjugation assays, we used the potential recipient strains listed in key resources table. We routinely grew all strains on lysogeny broth (LB) agar plates or in well-mixed liquid LB at 37°C.

Potential recipient community

We used activated sludge as the potential recipient community (hereafter “WWTP community”), which we collected from the aeration tank of a municipal wastewater treatment plant in Hokkaido, Japan. We collected ∼50 mL of activated sludge in the tubes to serve as the potential recipient community.

Method details

Construction of fluorescent plasmid donor strains

Although E. coli MG1655 carries the native lacI gene, we observed GFP fluorescence from the PA1/04/03gfp reporter on pB10gfp due to incomplete repression by LacI alone (Figure 2C). To achieve tighter control of the reporter, we reconstructed an E. coli MG1655 strain carrying a chromosomally integrated lacI^q-pLpp-mCherry cassette following a previously described system⁶¹ with codon optimization of the pLpp promoter. We first amplified the lacI and mCherry genes by PCR using the plasmids pMAL and MBP-mCherry as templates and the primer sets lacI-F/lacI-R and mCherry-F/mCherry-R (key resources table), respectively. To insert the pLpp promoter between lacI and mCherry, we synthesized a gene fragment containing pLpp by PCR using the primer sets lacI-R/mCherry-F. We cloned the resulting amplicon into the pUC19 vector using the In-Fusion HD Cloning Kit (Takara Bio, Japan), yielding the plasmid pUC19-lacI^q-pLpp-mCherry.We introduced this plasmid into E. coli strain DH5α via electroporation.

To prepare for chromosomal insertion, we introduced SmaI and XhoI restriction sites flanking the gene cassette by PCR amplification using the primer set lacI-pLpp-mCherry-F/lacI-pLpp-mCherry-R. We then ligated the amplified fragment into the SmaI- and XhoI-digested pGRG36 vector using the Mighty Mix DNA Ligation Kit (Takara Bio, Japan) and introduced it back into strain DH5α. We then inserted the lacI^q-pLpp-mCherry cassette into the chromosomes of E. coli strains MG1655 and β2163 using the Tn7 transposon system.⁵⁹ We selected transposon-containing cells on LB agar plates supplemented with 100 μg mL⁻¹ ampicillin. To remove pGRG36, we incubated the selected transposon-containing cells at 42 °C overnight, exploiting the vector’s temperature-sensitive replication. We confirmed correct chromosomal integration by PCR using the primer set lacI-pLpp-mCherry_Check-F/lacI-pLpp-mCherry_Check-R. We list all the primers used in this study in key resources table.

Growth media and culture conditions

We used synthetic wastewater medium (1xSWW) containing 160 mg L⁻¹ peptone, 110 mg L⁻¹ meat extract, 30 mg L⁻¹ urea, 28 mg L⁻¹ K₂HPO₄, 7 mg L⁻¹ NaCl, 4 mg L⁻¹ CaCl₂·2H₂O, and 2 mg L⁻¹ MgSO₄·7H₂O. To test environmental conditions that support higher growth rates, we also used 10xSWW with 10-fold higher concentrations of all components and LB (Lennox).

We measured the growth rate of the pB10 donor strain when grown with the different media (environmental conditions). To measure growth rates, we inoculated exponential-phase cultures of the pB10 donor strain into fresh liquid 1xSWW, 10xSWW, or LB at an initial optical density at 660 nm (OD₆₆₀) of 0.1 in L-type test tubes. We then recorded the OD₆₆₀ every 5 min using an automated OD meter (TVS062CA, Advantec, Japan) and calculated the maximum growth rate from five consecutive data points during the period of most rapid growth for each culture.

Surface-associated conjugation assay

Conjugation with individual potential recipient strains

To quantify the transfer efficiency and host range of pB10 during surface-associated growth, we performed conjugation assays between the pB10 donor strain and the recipient strains listed in key resources table. For these experiments, we used the DAP auxotrophic strain E. coli β2163 lacI^q-pLpp-mCherry as the pB10 donor strain. This strain requires DAP for growth, which allowed us to specifically recover transconjugant cells by plating on DAP-free agar plates. We grew the pB10 donor and potential recipient strains individually to stationary phase in liquid LB medium and washed them with 0.9% NaCl. We then suspended the cells in 0.9% NaCl and adjusted the OD₆₀₀ to 10. We next mixed equal volumes of the pB10 donor and potential recipient suspensions, filtered them onto 0.2 μm PTFE membranes (Advantec Ltd., Japan) and placed the filters on 1.5% agar plates containing 1xSWW, 10xSWW or LB. We allowed conjugation to proceed on these agar plates for 10 min. After incubation, we suspended cells in phosphate buffered saline (PBS). We quantified all cells by plating the suspensions on LB agar plates containing DAP (100 μg mL⁻¹). We quantified transconjugant cells by plating the suspensions on LB agar plates supplemented with tetracycline and kanamycin but without DAP. We quantified the conjugation efficiency as the proportion of transconjugant colonies relative to the total number of colonies.

Conjugation with activated sludge microbial communities

We performed surface-associated conjugation assays with the WWTP community and the engineered plasmid donor strain E. coli MG1655 lacI^q-pLpp-mCherry. To disperse large sludge flocs, we sonicated the samples using an ultrasonic homogenizer (UH-50, SMT Co., Ltd., Japan). We then washed the sonicated samples twice with 0.9% NaCl to remove residual debris before use in conjugation assays. We next grew E. coli MG1655 lacI^q-pLpp-mCherry as the pB10 donor strain in liquid LB medium to stationary phase and washed the cells twice with 0.9% NaCl. We then adjusted the OD₆₀₀ of the WWTP cells and the pB10 donor cells to 1.0 with 0.9% NaCl and mixed the suspensions together at a volumetric ratio of 1:1. Finally, we conducted surface-associated conjugation assays using two different methods as described below, where one is based on FC-FACS and the other on microscopy.

For the FC-FACS analysis, we filtered 5 μL of the mixed suspension of the WWTP cells and pB10 donor cells onto 0.2 μm PTFE membranes and placed the membranes onto the surfaces of 1xSWW, 10xSWW, or LB agar plates. We then incubated the 1xSWW agar plates at 25 °C and the 10xSWW and LB agar plates at 37 °C for 24 h. We selected 25 °C for 1xSWW to reflect conditions common in temperate WWTP environments and 37 °C with 10xSWW or LB as a high-growth benchmark representative of nutrient-rich laboratory conditions. After incubation, we suspended cells in 3 mL of 0.9% NaCl solution and vortexed them for 3 min. We then quantified and sorted transconjugant cells using a JSAN DCS-380 cell sorter (Bay Bioscience, USA). We excited GFP with a 488-nm laser and detected emission using a 525/50-nm bandpass filter. For each sample, we measured the green fluorescence intensity for 10⁵ cells. We defined fluorescence thresholds based on control samples to distinguish GFP-positive transconjugant cells from background signals (Figure S2). For downstream 16S rRNA gene sequencing, we collected 15,000 events per sample that passed both the forward scatter (FSC) and GFP fluorescence thresholds. We performed data analysis using Python (version 3.9) with the FCSParser package (https://github.com/eyurtsev/fcsparser).

For microscopy imaging, we deposited 30 μL of the mixed suspension onto 1.5% agarose pads (1 cm², 1 mm thick) containing 1xSWW, 10xSWW, or LB medium. We then placed each agarose pad on a glass-bottom dish (D11131H, Matsunami Glass Ind., Ltd., Japan) and covered it with plastic wrap to prevent drying and maintain oxic conditions. We observed the cells while incubating them on the 1xSWW agarose pads at 25 °C or on the 10xSWW and LB agarose pads at 37 °C. We imaged the mixed suspensions deposited on the agarose pads using an inverted microscope (Axiovert 200, Zeiss, Germany) equipped with a 40×/NA0.75 or 100×/NA1.30 oil objective lens and a CCD camera (AxioCam MRc, Zeiss, Germany). We set the bandpass filter and long-pass filter to 450–490/515 nm for the excitation of GFP and to 540–552/590 nm for the excitation of RFP. We randomly selected fields where a single cell exhibited green fluorescence and took images sequentially. We analyzed images in Python (OpenCV version 4.12.0) by smoothing and binarizing with Otsu thresholding and quantified the area of each microcolony from a binary mask using connected-component/contour analysis. All code used in image process are publicly available in the data deposit (https://doi.org/10.6084/m9.figshare.30102145).

Sequencing and analysis

We extracted total (chromosomal and plasmid) DNA from the total surface-incubated cells and the FC-FACS-sorted GFP-positive cells using the MightyPrep Reagent for DNA (Takara Bio, Japan) following the manufacturer’s instructions. We PCR amplified 16S rRNA gene fragments (V4 region) from the extracted DNA using the 515F and 806R oligonucleotide primers (key resources table). We outsourced amplicon sequencing of the 16S rRNA gene to Fasmac Co., Ltd. or Bioengineering Lab. Co., Ltd. (both in Kanagawa, Japan), where libraries were prepared and run on an Illumina MiSeq to generate 2 × 250 bp (in Fasmac Co., Ltd., Japan.) or 2 × 300 bp (in Bioengineering Lab. Co., Ltd. , Japan) paired-end reads, respectively. We analyzed the raw paired-end reads using QIIME 2 (version 2021.4). We used the DADA2 plugin to denoise the data, trim 30 bases and truncate the lengths to 250 bp (forward) or 240 bp (reverse), remove chimeras, and generate representative amplicon sequence variant (ASV) sequences and feature tables. We performed taxonomic classification using a pretrained Naive Bayes classifier trained on the SILVA 138 database specific to the 515F–806R region.⁵⁸ The sequence reads of the 16S rRNA gene amplicons are available in the DDBJ nucleotide sequence database under accession number PRJDB35968.

To assess taxonomic structure, we first rarefied the feature table to the smallest number of sequences that passed all quality control checks across all the samples. We additionally evaluated whether our sequencing depth was sufficient by generating rarefaction curves for all sample (Figure S5). All curves reached clear plateaus, indicating that the sequencing depth was adequate to capture the majority of community richness. We then quantified within-sample diversity using the Shannon index (Equation 1);

$$H^{\prime }=-\sum _{i=1}^S{p}_i\ln p_i$$ (Equation 1)

where p_i is the relative abundance of ASV_i and S is the total number of ASVs observed in that sample.

We quantified taxonomic dissimilarity between samples using weighted UniFrac distances. Weighted UniFrac incorporates the phylogenetic relationships among taxa by comparing shared branch lengths in a phylogenetic tree weighted by relative abundance.⁶² We quantified both metrics using the GUniFrac package in R (version 4.4.1) based on the ASV count tables and an aligned phylogenetic tree constructed from the 16S rRNA sequences.

To visualize the taxonomic composition across samples, we computed the relative abundance of each taxon by normalizing the ASV numbers to the total number of reads per sample. We then generated a heatmap of these relative abundances using the R packages “phyloseq” to manage and transform taxonomic data and “pheatmap” to create the heatmap, thereby highlighting the distribution patterns of dominant taxa across experimental conditions.

Individual-based plasmid conjugation simulations (CellModeller)

We used CellModeller version 4.3,^63,64 which is an open-source Python-based individual-based modeling framework, to simulate the spatial dynamics of plasmid conjugation in microbial communities. In this framework, cells are modeled as rod-shaped capsules that grow in continuous space. In our implementation of the model, we initialized the cells to have a radius r = 0.5 μm and an initial length l₀ = 2 μm except when noted otherwise. Each cell grows by elongation, adds mass at a constant rate, and divides once it reaches a critical division length, which is drawn from a Gaussian distribution (Equation 2).

$$l_{\mathit{div}}=l_0+\mathrm{N}\left(\mu =l_0,\sigma =0.45\right)$$ (Equation 2)

In this equation, N(μ, σ) represents normally distributed random variation. Capsule mechanics prevent physical overlap and cell-cell overlaps are resolved through a biophysical engine that enforces volume exclusion.⁶⁴ We performed simulations across a two-dimensional plane by setting the z axis displacement to zero.

We simulated conjugation between neighboring cells as described elsewhere.⁴² Briefly, physical contacts between cells are tracked at each time step, which allows us to implement contact-dependent conjugation logic. In our implementation, we only allowed conjugation between plasmid donor and potential recipient cells and not between transconjugant and potential recipient cells except when noted otherwise. We set the probability of conjugation per cell-cell contact per timestep to either a fixed or a cell-type-specific value depending on the simulation setup as stated in the results section. If successful, we recorded the plasmid acquisition event by changing the potential recipient cell-type to a designated transconjugant cell-type.

We initialized the simulations by randomly placing 1,000 plasmid donor cells and 1,000 potential recipient cells (1:1 ratio) across a 100 × 100 μm² simulation space with minimum spacing to avoid physical overlap. Plasmid donor cells carry a plasmid that can transfer to potential recipient cells upon direct contact with a probability of conjugation per contact per simulation timestep defined as P_c. For example, P_c = 0.05 indicates a 5% chance of conjugation per timestep when a plasmid donor and a potential recipient cell are in contact. We set the growth rate of transconjugant cells to be 10% lower than that of potential recipient cells to simulate plasmid cost. We performed simulations until the total population size reached 20,000 cells. We performed all simulations with a uniform growth rate of the plasmid donor cell-type of u_d = 1.0. All parameter values are provided in Table S1.

Simulations with a single potential recipient cell-type

We first simulated conjugation between one plasmid donor cell-type and one potential recipient cell-type, where all potential recipient cells can proliferate and acquire the plasmid from plasmid donor cells (excluding spatial statistical constraints). For this model, we varied the growth rate of the potential recipient cell-type from 0.0 to 2.0 and the conjugation probability from 0.000 to 0.050. All other parameters followed the general setup described above and Table S1. For each simulation, we recorded the abundance of transconjugants and quantified their proliferation attributable to HGT (plasmid acquisition from plasmid donor cells) and VGT (cell division) for all cells in each simulation.

Simulations with multiple potential recipient cell-types

We simulated conjugation between a single plasmid donor cell-type and one hundred distinct potential recipient cell-types, where each potential recipient cell-type has a different growth rate and probability of conjugation. As with the simulations for a single potential recipient cell-type, we initialized the simulations with 1,000 potential recipient cells. In this case, however, the potential recipient cells consisted of ten individual cells for each of the 100 different potential recipient cell-types.

For simulations with defined parameters, we set each potential recipient cell-type i to have a specific growth rate (u_r,i) and conjugation probability (P_c,i) sampled from four different parameter distributions: uniform, linear, small-skew log-scale and large-skew log-scale (Figure S8). For the uniform distribution, we set u_r,i = 0.5 and P_c,i = 0.01 for all recipient types (i = 1 to 100). For the linear distribution, we calculated the parameter values for u_r,i and P_c,i assigned to the i-th potential recipient cell-type as follows:

$$u_{r,i}=u_{r,\mathit{min}}+\frac{i-1}{n-1}\left(u_{r,\mathit{max}}-u_{r,\mathit{min}}\right)$$ (Equation 3)

$$P_{c,i}=P_{c,\mathit{min}}+\frac{i-1}{n-1}\left(P_{c,\mathit{max}}-P_{c,\mathit{min}}\right)$$ (Equation 4)

For the small-skew and large-skew log-scale distributions

$$u_{r,i}=u_{r,\mathit{min}}^{1-{\left(\frac{i-1}{\mathrm{n}-1}\right)}^k}{u}_{r,\mathit{max}}^{{\left(\frac{i-1}{n-1}\right)}^k}$$ (Equation 5)

$$P_{c,i}=P_{c,\mathit{min}}^{1-{\left(\frac{i-1}{n-1}\right)}^k}{P}_{c,\mathit{max}}^{{\left(\frac{i-1}{n-1}\right)}^k}$$ (Equation 6)

Here, n is the total number of potential recipient cell-types and k is the skew factor. We set k = 3 for small-skew log-scale distributions and k = 13 for large-skew log-scale distributions. For u_r,i, we set the minimum and maximum values to 0.1 and 1.0, respectively. For P_c,i we set the minimum and maximum values to 0.0001 and 0.05. For all distributions, the parameter values increased with the potential recipient cell-type index (i) such that u_r,i and P_c,i were positively correlated across potential recipient cell-types (Figures 5D and 5E).

To evaluate the effect of secondary plasmid transfer events, we implemented an additional simulation scenario where transconjugant cells were permitted to act as plasmid donors. After a recipient cell acquired the plasmid and became a transconjugant, the transconjugant was assigned the same P_c,i as its corresponding initial plasmid recipient cell-type. Secondary plasmid donors could transfer plasmids to any remaining plasmid-free recipient cell upon physical contact following the same contact-dependent transfer rules used for primary plasmid donors.

To evaluate the effect of cell length variation among potential recipient cells on the realized plasmid host range, we implemented an additional simulation scenario where potential recipient cell-types differed in their initial cell length. For each of 100 potential recipient cell-types, the cell length (l_0,i) was sampled either from a uniform distribution (U; constant l_0,i = 2 μm for all i) or from a large-skew log-scale distribution (l_0,i = 1–10 μm) consistent with the trait distributions used in the main simulation framework.

To quantitatively describe the impact of parameter combinations on plasmid host range estimates, we simulated systems in which each potential recipient cell-type was randomly assigned u_r,i and P_c,i values (Figures 5F and 5G). We repeated each of these random parameter simulations 100 times to account for stochastic variation. Briefly, for the small-skew simulations, we assigned the potential recipient cell-types with u_r,i values ranging from 0.1 to 1.0 and P_c,i values ranging from 0.0001 to 0.05 while preserving a small-skew log-scale distribution for each parameter. For the large-skew simulations, we assigned values using the same ranges but with a large-skew log-scale distribution. Although we fixed the marginal distributions of u_r,i and P_c,i for each set of simulations, the distribution of their products (u_r,i × P_c,i) varied among simulations depending on the specific combinations assigned.

For each simulation, we recorded the abundance of transconjugant cells for each distinct potential recipient cell-type and quantified their proliferation via VGT and HGT. We summed the counts for all the potential recipient cell-types to obtain N_HGT and N_VGT and calculate the ratio VGT/HGT = N_VGT/N_HGT. At the simulation endpoint, we calculated the Shannon index from the abundances of distinct potential recipient cell-types using Equation 1.

Quantification and statistical analysis

We performed all statistical analyses in the Python environment (Python 3.9) using the SciPy package (https://scipy.org/). For comparisons involving multiple groups, we adjusted the p-values from the two-way ANOVA tests using the Holm correction. We used these dissimilarity scores for ordination analyses and group-level comparisons of taxonomic structure. We performed two-sample two-sided Kolmogorov-Smirnov tests to test for differences in the distributions of continuous variables. We reported the sample size (n) for each statistical test in the results section, where all sample sizes are the number of independent biological or simulation replicates. We do not draw any conclusions based on the absence of statistical significance.

Published: March 10, 2026

Supplemental information

Video S1. Representative video of simulations with one potential recipient cell type

Description: representative simulation with one potential recipient cell type where we randomly placed 1,000 plasmid donor cells (red) and 1,000 potential recipient cells (white) within a square area (100 μm × 100 μm). During the simulation, the plasmid donor and potential recipient cells grow, and a potential recipient cell becomes a transconjugant cell (green) upon contact with a plasmid donor cell at a defined probability. In this simulation, potential recipient cells grow at a rate of u_r = 1.0 and acquire plasmids with a conjugation probability of P_c = 0.05.

Video S2. Representative video of simulations with one hundred potential recipient cell types

Description: representative simulation with one hundred potential recipient cell types where we randomly placed 1,000 plasmid donor cells (red) and 1,000 potential recipient cells (10 of each cell type, all white) within a square area (100 μm × 100 μm). During the simulation, the plasmid donor and potential recipient cells grow, and a potential recipient cell becomes a transconjugant cell (with each color corresponding to a distinct recipient cell type) upon contact with a plasmid donor cell. In this simulation, all potential recipient cell types grow at a rate of u_r,i = 1.0 and acquire plasmids with a conjugation probability of P_c,i = 0.05.