(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Indirect Fitness Benefits Enable the Spread of Host Genes Promoting Costly Transfer of Beneficial Plasmids

['Tatiana Dimitriu', 'Institut National De La Santé Et De La Recherche Médicale', 'Université Paris Descartes', 'Sorbonne Paris Cité', 'Paris', 'Dusan Misevic', 'Chantal Lotton', 'Sam P. Brown', 'Georgia Institute Of Technology', 'School Of Biology']

Date: 2022-01 

Bacterial genes that confer crucial phenotypes, such as antibiotic resistance, can spread horizontally by residing on mobile genetic elements (MGEs). Although many mobile genes provide strong benefits to their hosts, the fitness consequences of the process of transfer itself are less clear. In previous studies, transfer has been interpreted as a parasitic trait of the MGEs because of its costs to the host but also as a trait benefiting host populations through the sharing of a common gene pool. Here, we show that costly donation is an altruistic act when it spreads beneficial MGEs favoured when it increases the inclusive fitness of donor ability alleles. We show mathematically that donor ability can be selected when relatedness at the locus modulating transfer is sufficiently high between donor and recipients, ensuring high frequency of transfer between cells sharing donor alleles. We further experimentally demonstrate that either population structure or discrimination in transfer can increase relatedness to a level selecting for chromosomal transfer alleles. Both mechanisms are likely to occur in natural environments. The simple process of strong dilution can create sufficient population structure to select for donor ability. Another mechanism observed in natural isolates, discrimination in transfer, can emerge through coselection of transfer and discrimination alleles. Our work shows that horizontal gene transfer in bacteria can be promoted by bacterial hosts themselves and not only by MGEs. In the longer term, the success of cells bearing beneficial MGEs combined with biased transfer leads to an association between high donor ability, discrimination, and mobile beneficial genes. However, in conditions that do not select for altruism, host bacteria promoting transfer are outcompeted by hosts with lower transfer rate, an aspect that could be relevant in the fight against the spread of antibiotic resistance.

In bacteria, genes can move between cells, sometimes with the donor host cell actively involved in the gene transfer mechanisms. This movement of genes is called horizontal gene transfer, and it increases the prevalence of mobile genes in bacterial populations. However, it is not clear if donor host cells benefit from gene spread, or are simply exploited by selfish genes. Here, we show with both modelling and experiments that for the donor host, investing in the transfer of beneficial genes—such as those conferring antibiotic resistance—can be understood as an altruistic behaviour. This behaviour is costly to the donor but beneficial to recipients and can be selected for if a sufficient proportion of recipient cells share the donors’ transfer allele. Preferential transfer from donors towards recipients that share this allele occurs when dispersal is limited or if discrimination mechanisms are present. Our work suggests that both processes are likely to be widespread in nature, promoting horizontal gene spread by host donor cells. As many antimicrobial resistance and virulence genes are mobile, our work further implies that the spread of harmful traits among human pathogens may be modulated by host bacteria in a direction that depends on the bacterial ability to transfer the traits specifically to their kind.

To quantitatively understand HGT, the selection acting on donor ability must be analysed in a social context, taking into account both the costs and benefits transfer bestows onto donor and recipient hosts. Here, we theoretically and experimentally analyse the evolution of host genes controlling plasmid transfer. We show that from the host side, transfer represents a form of altruism: actors pay a cost of investing in transmission and deliver a benefit to recipients of beneficial mobile elements. Altruistic donation of MGEs can be maintained when transfer is sufficiently biased towards cells sharing donation alleles, increasing the donor allele inclusive fitness. This bias can arise in structured populations or by an association between transfer and discrimination alleles. Fitness gains due to the transfer of mutualistic plasmids further select for genotypes where donor ability alleles, discrimination alleles, and mutualistic plasmids are associated.

On one side, horizontal transfer confers varied and often extreme costs onto the bacterial host. Phage mobility usually requires host cell lysis that leads to death, while plasmid transfer through conjugation renders host cells sensitive to male-specific phages [ 13 ] and decreases the host's growth rate and fitness [ 14 , 15 ]. Because of these costs, horizontal transfer has classically been considered as a selfish trait of parasitic MGEs, selected as it favours their spread [ 7 ]. Direct support for transfer being a purely costly trait to the host came from studies of plasmid–host coevolution, where host genes that decrease transfer were selected [ 16 ]. On the other side, it has also been suggested that HGT could benefit the host because of the transfer of accessory genes not directly involved in MGE maintenance and transfer. Indeed, MGEs are often thought to constitute a communal pool of genes [ 2 ], a flexible genome [ 17 ] that can be quickly shuffled by HGT in response to environmental changes, making host populations more robust [ 18 ]. In this view, HGT is beneficial to the host population because it allows cells to share beneficial traits and provides diversity at the population level. However, it is not clear that these proposed benefits are sufficient for HGT to be favoured by the host. Traits advantageous at the group level—here the maintenance of a communal pool of genes—are not necessarily selected for at the individual level, especially when individuals can benefit from others that invest in the trait while not paying the cost of investing themselves [ 19 ]. Indirect, population-wide benefits alone are not necessarily sufficient to explain the selection of host genes promoting costly transfer [ 20 ]. The ability to receive genes can clearly be directly selected for when these genes enhance individual fitness: for instance, CRISPR immunity against antibiotic resistance plasmids, a form of HGT resistance, was rapidly lost in the presence of antibiotics when receiving plasmids was beneficial to the host [ 21 ]. On the contrary, the ability to donate genes need not be selected, as the donor cell does not directly benefit from transferring genes to neighbouring recipients.

MGEs such as plasmids or phages are defined by their ability to undergo horizontal gene transfer (HGT) between bacterial hosts [ 1 ], and are widespread in nature. Genes present on MGEs often affect their hosts’ fitness in a specific environment [ 2 ]. Particularly, many mobile genes increase virulence or antibiotic resistance and thus have harmful consequences on human health. Antibiotic resistance genes are enriched on plasmids [ 3 ], leading to their fast spread among bacterial species via horizontal transfer [ 4 ]. Genes coding for secreted proteins, often involved in virulence, are also enriched on MGEs promoting cooperative secretion [ 5 , 6 ]. In order to better combat the medical issues arising from horizontal transfer, we must understand the selective pressures acting on gene mobility. The population dynamics and evolution of transfer have mostly been studied by focusing on MGEs themselves [ 7 ]; however, transfer is influenced not only by MGE genes, but also by genes of the bacterial host chromosome. Both donor [ 8 ] and recipient cells [ 9 , 10 ] can regulate transfer, with different donor and recipient genetic backgrounds resulting in as much as eight orders of magnitude variance in the transfer rates for the same plasmid [ 11 , 12 ]. Thus, to fully understand the evolution of horizontal gene spread and the natural variation in transfer rates among hosts, we must consider the selective pressures acting on hosts.

Results

Schematic Model We first perform a qualitative analysis to identify if and in which conditions a strain with high donor ability can be selected. We model the fitness of nonmobile host genes controlling donor ability for a given plasmid using a neighbour-modulated fitness approach that partitions fitness into the effects of an individual’s own genotype and those of social neighbours [22,23]. We consider a population of bacteria structured in an infinite number of patches [24] and model a simplified life cycle with nonoverlapping patch generations, in which the following processes occur successively [25,26]: founding, reproduction, transfer, selection, and dispersal (see S1 Text for details). A cell i in patch j is characterized by three traits: plasmid carriage p ij (p ij = 1 for plasmid-bearing cells and 0 for plasmid-free cells), donor ability q ij and recipient ability s ij . Successful transfer is controlled by three factors: the probability of contact between plasmid-bearing and plasmid-free cells, the donor ability of plasmid-bearing cells, and the recipient ability of plasmid-free cells. We assume that plasmid and host traits are distributed independently in the starting population so that the cell's donor ability q ij is independent from its initial plasmid content p ij . Initially uninfected cells become infected with a probability proportional to the patch level frequency of plasmid-bearing cells modulated by their own recipient ability and by the average patch donor ability. A cell i in patch j will thus be modified by transfer with the probability (1 − p ij ) p j q j s ij . Plasmid presence has an effect e p on the host cell, and we can express the plasmid effect on host fitness as e p p ij . The cost of donor ability is c q leading to an effect of transfer on host fitness that is proportional to donor ability, experienced only by cells bearing plasmids before transfer, and equal to −p ij c q q ij . Donor ability is costly independently of actual transfer efficiency, modelling the effect of expressing the transfer machinery (which happens even in the absence of successful transfer). The fitness of an individual founding cell i in patch j, measured over the patch life cycle, is noted by W ij . With W 0 being the basal host fitness, we obtain (see S1 Text): (1) To understand selection acting on donor ability q, we apply the Price equation [27,28] to Eq (1). We obtain the regression coefficient between fitness and donor ability, β(W ij ,q ij ), that describes the effect of donor ability on fitness (Eq 2). We provide the derivation of Eq (2) and a detailed analysis in S1 Text. (2) The E j [p j (1 − p j )] term describes the effect of patch composition on the efficiency of plasmid transfer: transfer events are more likely when both plasmid-bearing and plasmid-free cells are abundant within each patch. β(q j s ij ,q ij ) is a regression coefficient between individual donor ability q ij and the product of individual recipient ability with patch-level donor ability. It corresponds to the relatedness between plasmid donor and recipient cells, noted by R q , at the locus determining donor ability (see S1 Text for a detailed analysis): R q is higher when donor cells preferentially encounter recipients that share their donation allele and when transfer is more successful towards those cells. R q thus determines how much a donor cell transfers plasmids to individuals bearing the same donation allele because of population structure and specificity in transfer. Finally, p c q is the average cost of transfer for the donor genotype: high donor ability is costly to the proportion of cells that bear plasmids and express their transfer machinery. An increase in donor ability is selected for when it is correlated with increase in fitness, namely when β(W ij ,q ij ) > 0, which combined with Eq (2) leads to the following condition: (3) Eq 3 is a form of Hamilton’s rule [29], which postulates that a cooperative allele is selected for when its indirect benefits, weighted by relatedness among actors and recipients, outweigh its direct cost, maximizing its inclusive fitness (fitness inclusive of alleles present in other individuals). Applied here to donor ability, the indirect benefits are the benefits of plasmids to the recipient cells after transfer e p E j [p j (1 − p j )], and the direct cost is the cost of donor ability for cells bearing plasmids p c q . R q is the relatedness at the donor ability locus among donor and recipient cells of plasmid transfer. High, positive R q implies that most of transfer events from cells with high donor ability will be directed towards recipients sharing their donation allele. On the contrary, negative R q means that transfer will be biased towards cells with a different allele than the one carried by the donor. Thus, a high donor ability allele can be selected even when individually costly, when transfer maximizes its inclusive fitness through plasmid effects on recipient cells. We note that relatedness in bacteria can vary across loci [30], as it can be modified in a locus-specific way by mutation [31] or HGT [5,6]. Thus, unlike relatedness arising from genealogical kinship in sexually reproducing organisms, it will not necessarily tend to be the same across the genome. To underline this and avoid any potential semantic confusion, we follow nomenclature defined already in [30] and consider that cells that specifically share alleles at the locus of interest (plasmid donation) are cells of the same kind but not necessarily kin. R q will be positive when donors preferentially transfer plasmids to recipients of their kind. Positive relatedness generally arises through the combination of two processes: limited dispersal and discrimination mechanisms [29,32]. Here, limited dispersal is due to patch structure: the correlation between q j and q ij is governed by the initial repartition of genotypes among patches, with no migration before transfer occurs. Positive relatedness can arise from strong population bottlenecks leading to stochastic variations in founding cell frequencies among patches, followed by clonal reproduction [33]. Effective discrimination in transfer also leading to positive relatedness arises if s ij and q ij are positively correlated, with genotypes with high donor ability having higher recipient ability than average, or if donors have a way to direct transfer specifically to their kind (see S1 Text for discussion). Alternatively, negative relatedness can arise if s ij and q ij are negatively correlated, leading to preferential transfer to cells bearing a different donation allele. We can distinguish two scenarios for the effect and selection of transfer depending on the plasmid effects on the host cell. In the first case, the transferred plasmid is mutualistic with its host (e p > 0), for instance conferring antibiotic resistance: transfer is therefore an altruistic behaviour [29] with a direct cost of performing transfer and indirect benefits through the plasmid benefits in recipient cells. Transfer is selected if R q is positive and sufficiently high: R q > p c q / (e p E j [p j (1 − p j )]). In the second case, the transferred plasmid is parasitic (e p < 0): donor ability for parasitic plasmids, decreasing the fitness of recipient cells, is selected if R q is negative and sufficiently low: R q < p c q / (e p E j [p j (1 − p j )]). This would be a case of spiteful behaviour [32,34]. Specific population structure or discrimination processes are required to produce negative relatedness, and spite is thus thought to be less common than altruism [34]. We focus here on the transfer of mutualistic MGEs and more specifically on antibiotic resistance plasmids that allow their hosts to grow when antibiotics are present. The main prediction arising from our model is that donor ability for these mutualistic plasmids is an altruistic trait, counterselected if transfer occurs indiscriminately towards any cell, but selected for when plasmid donors and recipients share donation alleles. We present the model graphically in Fig 1, focusing on the three relevant scenarios affecting relatedness: random interactions between individuals (Fig 1A), discrimination in transfer (Fig 1B), and structured populations (Fig 1C). We next test the model’s predictions with both simulations and experiments, performing competition assays between strains with varying donor ability in order to investigate quantitatively if and how much selection favours donor ability in biologically realistic settings. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Graphical representation of different scenarios for the selection of transfer as an altruistic trait. In this simplified diagram, we follow a strain with high donor ability (red) in competition with another strain with no donor ability (white). Some cells of both strains bear an antibiotic resistance plasmid (black dots) that donors can transfer (red arrows) to a cell of either type, as long as it is plasmid-free. Our model predicts that donors are selected for when the red-framed equation is true (Eq 3, see main text). For clarity, we assume three sequential steps: (1) transfer, whose recipients depend on relatedness at the donor ability locus (R q ) and whose efficiency depends on plasmid frequency within patches p j ; (2) antibiotic selection, where only plasmid-bearing cells survive (e p > 0); and finally, (3) cell growth after selection, where donor cells experience a cost c q and grow more slowly. We describe three possible scenarios, depending on the properties of transfer and its effects on relatedness at the donation locus. A: In the absence of discrimination in transfer or population structure, relatedness among donors and recipients is null, and transfer occurs with the same efficiency towards all cells. B: In the presence of discrimination in transfer, good donors transfer plasmids specifically to their kind. C: In structured populations, good donors are surrounded by their kind, to which they preferentially transfer plasmids even in the absence of discrimination. In all scenarios, donor cells experience the cost of expressing the transfer machinery during growth. However, only in B and C does transfer bias lead to an enrichment of plasmids in the donor strain after transfer, which can compensate for donor ability cost when plasmids are selected for. https://doi.org/10.1371/journal.pbio.1002478.g001

Selection for Donor Ability through Discrimination in Transfer Discrimination in transfer occurs if during the encounters between a donor and potential recipients the plasmids are transferred to cells of the donor’s kind more often than would be expected based on its frequency in the population. Discrimination of plasmid recipients could be based on differences in the initial recognition between cells or differences in plasmid establishment in recipient cells. To search for evidence of discrimination, we analyse two available datasets [11,12] that quantify plasmid conjugation rates among different pairs of natural isolates. Both studies measured conjugation rates for the multiresistant R1 plasmid, among 10 strains from the ECOR collection [11] or 9 other natural Escherichia coli strains [12]. In each dataset, we compute normalized donor ability for each pair of donor and recipient strains (see Materials and Methods), which corrects for basal differences in donor ability between strains. We find that transfer to self occurs at rates higher than average in 18 out of 19 cases (Fig 2A). Additionally, in 8 out of 19 cases, the highest rate of transfer is from a strain to itself. Overall, transfer to self is 7.3 times higher than average donor ability over all tested isolates (two tailed t test for difference from 0 for normalized donor ability to self, p = 0.0003). In a mixed population with many different strains, the high rates of transfer to self we describe here would translate into a biased transfer between cells sharing donation alleles. This apparent discrimination does not imply that the same locus is responsible for high donor ability and for discrimination, as multiple genes could be involved in discrimination. However, the signal we observe in Fig 2A suggests that alleles for high donor ability and for discrimination in transfer are linked in natural isolates sufficiently to lead to an effective discrimination at the donor ability locus. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Selection for plasmid transfer through discrimination of recipients. A: Donor ability to self and others among natural isolates. A normalized donor ability for the R1 plasmid for couples of donor and recipient strains was obtained by correcting conjugation rates from [11] (left of the dashed line) and [12] (right of the dashed line) datasets by the average strain donor ability across all recipient strains tested. Points in red indicate donor ability measured from a strain to itself. B: Conjugation rates between E. coli K12 and B strains. Conjugation rates were measured in the same growth conditions as for Fig 2C competitions. Donors and recipients were mixed during exponential growth (optical density [OD] = 0.2), then donor (D), recipient (R), and transconjugant (T) densities were measured by plating. Conjugation rates were computed as (mL.cell-1.h_1), and are shown as geometric means ± standard error of the mean (SEM) (N ≥10). Red bars show conjugation data obtained with the wild type K12 strain, grey bars show conjugation using the K12ΔarcA mutant. C: Competition between K12 and B strains with antibiotic resistance transfer. The change in frequency of K12 strain is shown in competition with B strain in a single well-mixed population, in the absence (blue) or presence (yellow) of antibiotic selection at the end of competition, and for different initial proportions of the R1-19 plasmid, common to both strains. Pale colours show the outcome of competition with the K12ΔarcA strain. Results are shown as means ± SEM (N ≥10). Data are available from FigShare at http://dx.doi.org/10.6084/m9.figshare.3199252. https://doi.org/10.1371/journal.pbio.1002478.g002 We next experimentally investigate if discrimination may allow for the selection of host transfer genes. We use two widely studied E. coli strains, the K12 strain MG1655 [35] and the B strain REL606 [36], and the multiresistant R1-19 plasmid [37]. K12 and B strains bear different restriction-modification systems [9], potentially leading to discrimination in plasmid transfer [38,39]. We first measure the conjugation rate in a well-mixed environment between all four combinations of K12 and B as donor and recipient strains (Fig 2B, red bars). We find that K12 is generally a better donor, but also that K12 transfers R1-19 plasmid to itself at a 5-times higher rate than to B (Mann-Whitney Wilcoxon test, p = 0.003). Overall, the K12 strain is an example of good donor strain displaying discrimination for transfer, in comparison to the lower donor B strain. Moreover, R1-19 carriage leads to a 54% reduction in exponential growth rate for K12 strain, compared to a 1.6% reduction only for B strain (S1 Fig). To test if part of the costly effect R1-19 has on K12 is due to donor ability, we use a K12-derived strain with a deletion in the arcA gene, a gene known to affect transfer [40], and the repressed R1 plasmid, which transfers approximately 1,000-fold less than R1-19 [37,15]. The K12ΔarcA mutant transfers R1-19 plasmid at a strongly reduced rate to both itself and B (Fig 2B, grey bars), and R1-19 cost is reduced as well (6.4%, S1 Fig). Similarly, R1 plasmid imposes almost no cost to K12 growth (1.8%, S1 Fig). Both results suggest that most of R1-19 cost to K12 is due to its high transfer rate. We then test whether K12 discrimination in transfer can lead to biased transfer towards other K12 cells in a well-mixed population and subsequent selection of the better donor strain. We compete the K12 and B strains by mixing them equally in a well-mixed population, with a common proportion of cells from each strain initially bearing R1-19 plasmid. In the absence of antibiotic selection, the better donor K12 decreases in frequency (Fig 2C, dark blue), showing a lower basal fitness than B in those culture conditions. When antibiotic selection is applied at the end of the competition by plating the population on kanamycin (Kn)-containing medium, only Kn-resistant, plasmid-bearing cells grow. When all cells initially bear plasmids, selection does not favour the K12 strain (Fig 2C orange, 19% decrease in K12 frequency, two sided t test for difference from 0, p = 0.003). However, when only a fraction (2.5%) of both K12 and B cells initially bear R1-19 plasmid, providing opportunity for plasmid transfer, K12 is selected (19% increase in K12 frequency, two sided t test for difference from 0, p = 0.009). Finally, to confirm that this specific selection of donors is due to R1-19 transfer to K12 cells, we analyse the outcome of competition when the arcA gene is deleted from K12 and transfer is impaired. In the absence of antibiotic selection (Fig 2C, light blue), the arcA deletion does not affect K12 fitness when plasmids are absent or rare and increases K12 fitness when all cells bear plasmids (11% decrease for K12ΔarcA versus 25% decrease for K12, two-sided t test, p = 0.043), possibly because of the reduced plasmid cost for K12ΔarcA. With antibiotic selection, the specific selection of K12 when a fraction of cells bear plasmids disappears for K12ΔarcA (Fig 2C middle, yellow bars, two-sided t test, p = 2.10−5), demonstrating that K12 selection was due to plasmid transfer. Discrimination effectively biases antibiotic resistance transfer strongly enough so that the better donor K12 strain is selected for in the presence of antibiotics. Thus, when transferred plasmids are needed for growth, discrimination in transfer towards kind, at naturally appearing levels, can be sufficient to select for the better donor.

Selection for Donor Ability in Structured Populations A second possible reason for transfer bias is bacterial growth in structured populations, where donors interact preferentially with their kind. Next, we examine whether, in the absence of discrimination, structured populations can provide a sufficient bias in transfer to select for good donors. To analyse the effect of biased transfer in structured populations, we use a synthetic system with fluorescently tagged plasmids in which we can identify plasmid transfer between two strains. We adapted the system from the one we designed for an earlier study on interaction between conjugation and cooperation [6]. A helper plasmid F HR , that is nonmobile and thus behaves like a chromosomal allele, governs the host cell donor ability for a mobile plasmid C, which confers chloramphenicol (Cm) resistance. We compete (Fig 3A) two strains differing in their donor ability: the good donor D+ strain bears F HR that transfers C plasmids, and the nondonor D− strain does not (S2A Fig). After a transfer phase (t 0 to t 1 ), populations are grown with or without Cm during the selection phase (t 1 to t 2 ). We compare a single, well-mixed population (m), where D+ and D− are mixed in equal proportions, to a structured metapopulation (s), consisting of two subpopulations that grow separately during the transfer phase, s 1 and s 2 , founded respectively with a 10% and 90% proportion of D+ (leading to equal proportions of D+ and D− at the metapopulation level). In this setup, the changes in the good donor frequency can be followed both within and among populations to evaluate the effect of population structure on donor selection. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 3. Selection of donor ability in structured populations. A: Experimental setup. D+ (good donor, red) and D− (nondonor, blue) strains are competed. 2.5% of D+ and D− cells initially carry C plasmids (bright colours), while 97.5% do not (pale colours). The population m is a single well-mixed population; metapopulation s consists of two subpopulations, s 1 and s 2 , with initial D+/D− ratios of 1/9 and 9/1. After growth and transfer (t 0 to t 1 ), subpopulations from s are pooled and cells are grown to saturation with or without antibiotic (Cm) selection (t 1 to t 2 ). The proportions of different cell types are represented schematically and do not correspond to actual numbers. B: Selection of D+ strain. The frequency of the good donor D+ is shown for s (black) and m (green) populations, with (plain lines) or without (dashed lines) Cm antibiotic during the selection phase. Good donors are only selected for in the s metapopulation, in the presence of antibiotic. C: Plasmid dynamics. Plasmid frequency in each population is shown for the transfer phase (from t 0 to t 1 ) , in each of m, s 1 , and s 2 populations. Plasmids spread mostly in the s 2 subpopulation, enriched in the better donor, D+. D: Transfer bias. The proportion of C plasmids present in D+ strain, is shown as a function of time for s and m populations (same colour scheme as in B panel). C plasmids get enriched in the better donor D+ strain during the transfer phase, for the structured population s. All results are shown as means ± SEM. (N ≥ 6). Data are available from FigShare at http://dx.doi.org/10.6084/m9.figshare.3199252. https://doi.org/10.1371/journal.pbio.1002478.g003 D+ strain frequency does not change significantly in m or s populations during the transfer phase (Fig 3B left). D+ frequency then increases at t 2 only for the structured s population grown in the presence of Cm (Fig 3B right, 26% increase from t 0 , Mann-Whitney Wilcoxon test, p = 0.004). It decreases for m population with Cm (19% decrease, two-sided t test for difference from 0, p = 3.10−9) and stays constant in the absence of Cm. The dynamics generally follows our predictions: D+ selection requires both population structure and plasmid selection. However, the expected cost of D+ during the transfer phase is not present at the population level. We next investigate in more detail both this cost and the selection of the good donor strain. By looking at the dynamics of individual subpopulations during the transfer phase, we observed that D+ increases in frequency when prevalent (S3A Fig). We confirmed with an independent experiment that D+ fitness in competition with D− linearly increases with D+ frequency (S3B Fig). This positive frequency-dependence for donor fitness could be due to lethal zygosis, a phenomenon known to damage recipients at high donor cell frequencies [41], which could be aggravated by the absence of entry exclusion in our strains [42]. In natural systems, entry exclusion may protect new transconjugants but would also make initial plasmid-bearing donors immune to lethal zygosis, probably leading to a similar frequency-dependence of fitness when most donor cells initially bear plasmids. In our system, frequency-dependence leads to no observable cost for D+ at the metapopulation level. At low frequencies, donor ability still has a cost, which is also observed as a decrease in the strain’s growth rate when growing in isolation (D+ versus D−, S2C Fig). Interestingly, donor cells grow significantly more slowly when they bear C plasmids, which is not the case for nondonor cells (D+ C versus D− C , S2C Fig), suggesting that donor ability cost is enhanced by the presence of transferable plasmids in the cell. During the selection phase, good donors are selected only in the structured s population and only in the presence of Cm, meaning that donor selection requires both population structure during the transfer phase and subsequent antibiotic selection. We see that, as predicted by our model, biased transfer due to population structure promotes indirect selection of the donor strain. To better understand the factors affecting D+ selection, we proceed to analyse the dynamics of C plasmids (Fig 3C). During the transfer phase, plasmid frequency changes depend on the proportion of cells able to transfer. In the s 1 population where D+ cells are few, plasmid frequency declines slightly. It increases mostly in the s 2 population enriched in D+ strain. Increases are due to transfer, as the increase in plasmids present in D− strain is due to plasmids that originate from D+ (as identified by fluorescence markers, see Materials and Methods and S4 Fig). We then follow the proportion of C plasmids that are present in D+ cells, as plasmid localization controls survival in the presence of antibiotics. During the transfer phase, the proportion of C plasmids present in D+ cells compared to D− cells decreases in the well-mixed m population (13% decrease, Mann-Whitney Wilcoxon test, p = 0.004) but increases in the structured s population (28% increase, Mann-Whitney Wilcoxon test, p = 0.004) (Fig 3D). In the well-mixed population, the decrease is probably due to the strong fitness cost D+ cells incur specifically when they bear C plasmids (S2 Fig). The same cost also explains the subsequent decrease in D+ strain frequency under Cm selection in both populations. The enrichment of C plasmids in D+ cells depends on the population structure of the s population: total plasmid transfer is more prevalent in the s 2 subpopulation, effectively biasing transfer towards D+ at the metapopulation level. These results experimentally confirm our models prediction: in the absence of discrimination mechanisms, donor ability for antibiotic resistance plasmids can be selected when population structure ensures preferential transfer to cells sharing donation alleles.

[END]

[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.1002478

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/