Mixed Wolbachia infections resolve rapidly during in vitro evolution

The intracellular symbiont Wolbachia pipientis evolved after the divergence of arthropods and nematodes, but it reached high prevalence in many of these taxa through its abilities to infect new hosts and their germlines. Some strains exhibit long-term patterns of co-evolution with their hosts, while other strains are capable of switching hosts. This makes strain selection an important factor in symbiont-based biological control. However, little is known about the ecological and evolutionary interactions that occur when a promiscuous strain colonizes an infected host. Here, we study what occurs when two strains come into contact in host cells following horizontal transmission and infection. We focus on the faithful wMel strain from Drosophila melanogaster and the promiscuous wRi strain from Drosophila simulans using an in vitro cell culture system with multiple host cell types and combinatorial infection states. Mixing D. melanogaster cell lines stably infected with wMel and wRi revealed that wMel outcompetes wRi quickly and reproducibly. Furthermore, wMel was able to competitively exclude wRi even from minuscule starting quantities, indicating that this is a nearly deterministic outcome, independent of the starting infection frequency. This competitive advantage was not exclusive to wMel’s native D. melanogaster cell background, as wMel also outgrew wRi in D. simulans cells. Overall, wRi is less adept at in vitro growth and survival than wMel and its in vivo state, revealing differences between cellular and humoral regulation. These attributes may underlie the observed low rate of mixed infections in nature and the relatively rare rate of host-switching in most strains. Our in vitro experimental framework for estimating cellular growth dynamics of Wolbachia strains in different host species, tissues, and cell types provides the first strategy for parameterizing endosymbiont and host cell biology at high resolution. This toolset will be crucial to our application of these bacteria as biological control agents in novel hosts and ecosystems.


Abstract:
The intracellular symbiont Wolbachia pipientis evolved after the divergence of arthropods and nematodes, but it reached high prevalence in many of these taxa through its abilities to infect new hosts and their germlines.Some strains exhibit long-term patterns of co-evolution with their hosts, while other strains are capable of switching hosts.This makes strain selection an important factor in symbiont-based biological control.However, little is known about the ecological and evolutionary interactions that occur when a promiscuous strain colonizes an infected host.Here, we study what occurs when two strains come into contact in host cells following horizontal transmission and infection.We focus on the faithful w Mel strain from Drosophila melanogaster and the promiscuous w Ri strain from Drosophila simulans using an in vitro cell culture system with multiple host cell types and combinatorial infection states.Mixing D. melanogaster cell lines stably infected with w Mel and w Ri revealed that wMel outcompetes w Ri quickly and reproducibly.Furthermore, w Mel was able to competitively exclude w Ri even from minuscule starting quantities, indicating that this is a nearly deterministic outcome, independent of the starting infection frequency.This competitive advantage was not exclusive to wM el's native D. melanogaster cell background, as w Mel also outgrew w Ri in D. simulans cells.Overall, w Ri is less adept at i n vitro growth and survival than w Mel and its in vivo state, revealing differences between cellular and humoral regulation.These attributes may underlie the observed low rate of mixed infections in nature and the relatively rare rate of host-switching in most strains.Our in vitro experimental framework for estimating cellular growth dynamics of Wolbachia strains in different host species, tissues, and cell types provides the first strategy for parameterizing endosymbiont and host cell biology at high resolution.This toolset will be crucial to our application of these bacteria as biological control agents in novel hosts and ecosystems.

Author Summary:
Wolbachia pipientis is one of the most common bacterial endosymbionts due to its ability to manipulate host reproduction, and it has become a useful biological control tool for mosquito populations.Wolbachia is passed from mother to offspring, however the bacterium can also "jump" to new hosts via horizontal transmission.When a Wolbachia strain successfully infects a new host, it often encounters a resident strain that it must either replace or co-exist with as a superinfection.Here, we use a Drosophila melanogaster cell culture system to study the dynamics of mixed Wolbachia infections consisting of the high-fidelity w Mel and promiscuous w Ri strains.The w Mel strain consistently outcompetes the w Ri strain, regardless of w Mel's initial frequency in D. melanogaster cells.This competitive advantage is independent of host species.While both strains significantly impede host cell division, only the w Mel strain is able to rapidly expand into uninfected cells.Our results suggest that the w Ri strain is pathogenic in nature and a poor cellular symbiont, and it is retained in natural infections because cell lineages are not expendable or replaceable in development.These findings provide insights into mixed infection outcomes, which are crucial for the use of the bacteria in biological control.

Introduction
The alphaproteobacterium Wolbachia pipientis became a widespread intracellular symbiont of arthropods and nematodes through its ability to infect novel hosts and establish germline transmission.Hundreds of millions of years after the divergence of Arthropoda and Nematoda (ca.500 mya [1,2] , Wolbachia endosymbionts evolved (ca.100-200 mya [3] ) and spread to infect a high proportion of these hosts [4][5][6] .Following horizontal transmission to a new host and establishment of a stable infection, Wolbachia targets the host germline to achieve vertical transmission from one host generation to the next [4,7,8] .Thus, at least two core mechanisms have contributed to the rise of Wolbachia in ecdysozoan hosts: high infectivity and targeted germline transmission.These two traits appear primed for conflict, as natural selection for infectivity is often linked to pathogenicity, which could interfere with normal host development.However, they have harmonized in Wolbachia to produce the planet's largest pandemic [9] .Significant variation exists among closely related Wolbachia strains in their ability to infect new hosts.While all strains examined undergo vertical transmission through the host germline [10] , some strains are also adept at colonizing new hosts through horizontal transmission and novel infection establishment.Promiscuous Wolbachia strains, such as the w Ri strain from Drosophila [11] and w Jho from butterflies [12] , are found in unrelated hosts or multiple hosts ( i.e. , superinfections, see S1 Fig) .These strains often exhibit strong reproductive manipulations, such as cytoplasmic incompatibility (CI), that drive Wolbachia infections to high frequencies in host populations from low starting frequencies [13] .Indeed, recent biological control applications using Wolbachia infections rely on strong and predictable CI in non-native hosts for their spread across targeted populations [14] .Selection for beneficial host-symbiont emergent functions and phenotypes may also be sufficient to increase and maintain infection frequencies in strains lacking reproductive manipulations [4,15] .Successful host-switches are the culmination of a successful horizontal transmission event, stable host colonization and propagation, co-option of germline transmission, and establishment across individuals in a population (Supplemental Figure S1 and reviewed in [4] ).Attempts to model Wolbachia infection distributions based on an average turnover process produce estimates that explain global infection frequencies, but that fail to explain strain-to-strain variation in horizontal transmission ability and novel infection establishment [16] .A major challenge involves parameterizing the infrequent, but vital events in the process.Based upon the low rates of mixed infections in infected hosts and the novel infections in uninfected hosts [17][18][19] , joint rates of horizontal transmission and successful proliferation in a new host are exceedingly low.However, it is unclear whether both rates are low, or if horizontal transmission rates are high, but exceedingly few bacteria persist and colonize host tissues.Furthermore, it is unknown how divergent strains ecologically interact within a single host, especially if one strain is more promiscuous than the other.
To study the finescale ecological events that occur among endosymbionts and hosts in novel host infections, we developed an in vitro Drosophila cell line system infected with faithful and promiscuous strains of Wolbachia .We leveraged two different Drosophila melanogaster somatic cell types infected with the native w Mel strain and the non-native, promiscuous w Ri strain from Drosophila simulans to study what occurs when a promiscuous strain infects a host with a stable endosymbiont.Then, we use a novel D. simulans cell line immortalized for this study to explore the reciprocal mixed infection in one of w Ri's native hosts.Lastly, we measure infection expansion into uninfected host cells to parameterize a model of endosymbiont in vitro growth, host cell segregation, and cell-to-cell transfer.Overall, this work reveals that closely related strains have significantly different capacities for cellular proliferation that are counterintuitive based on their distributions among hosts.Furthermore, we show that mixed infections resolve rapidly and predictably across cell types and hosts, shedding light on the rarity of mixed infections in nature.These results significantly increase our understanding of what occurs when novel strains interact within host cells and tissues.This knowledge is critical to ensuring the safety of biological applications that release hosts infected with non-native Wolbachia strains into natural ecosystems.

Results and Discussion
In vitro Wolbachia infections in D. melanogaster cell culture are stable over time We successfully established and maintained in vitro w Mel and w Ri infections in two D. melanogaster cell lines, the neuroblast-like JW18 cell line [20] and the macrophage-like S2 cell line [21] (diagramed in Fig S2).Fluorescence in situ hybridization (FISH) with 16S rRNA probes visually confirmed the presence of Wolbachia in infected cells (Supplemental Fig S1B-F) and its absence from doxycycline-cured (DOX) cells (Supplemental Fig S3A-F).We used whole genome sequencing (WGS) and reference genome mapping to confirm infection strain identities, estimate the genomic titer of each symbiont infected cell line, and observe fluctuations in titers over time.We consistently observe wMel at a higher titer (~10-30) than w Ri (~0.1-3) (Fig S3G).
The wMel strain outcompetes the wRi strain from equal starting ratios The w Mel strain of Wolbachia outcompetes the w Ri strain in D. melanogaster in vitro infections.To recapitulate the conditions of a mixed Wolbachia strain infection in vitro , we mixed w Mel and w Ri infected cells at approximately equal genomic titers (Fig 1A).This equal starting ratio was selected to not advantage either strain and study the differences in infected host cell and strain growth rates.Each mixed culture was split into triplicate, and passaged every seven days, with a sample collected for sequencing at each passage.We estimated the abundance of each symbiont by calculating the proportion of total coverage contributed by that symbiont, which is the average coverage of the symbiont divided by the sum of the average coverages of both symbionts and the host.In the first 3 weeks immediately following the initial mixing of the two strains, both w Mel and w Ri increased in frequency.However, after this phase of initial expansion, only w Mel continued to increase in frequency.By week 5, w Mel accounted for an average proportion of total coverage of 90% (Fig 1F).During this timeframe, the JW18 neuroblast-like D. melanogaster cell culture cells exhibited adherence defects that suggested the cells were under stressful conditions, whereas the S2 cells maintained their normal phenotypes (Fig 1B-E).
We used a simple haploid model of relative fitness (see Methods) to estimate the selection coefficient ( ω ) of w Mel in the mixed infection experiments.Because we observed that w Mel replaces w Ri within five to seven weeks post mixing, we constrained our selection coefficient estimates to six weeks post-mixing in order to capture the early dynamics of selection acting on the two strains (Fig 1G).We estimated selection coefficients ranging from 2.81-3.06 in the JW18 cell line, and 1.80-2.36 in the S2 cell line (Table S1).These values indicate that w Mel is far fitter in D. melanogaster cells than w Ri.However, this selective advantage may have been influenced by wMel's high starting concentration.Next, we explore whether w Mel outcompetes w Ri when it is a minority constituent in two-strain mixtures.genomes out of the total coverage mapped to all Wolbachia and D. melanogaster host genomes, plotted by replicate and host cell type (S2, dashed or JW18, solid).G) Relative growth rates of wMel compared to wRi over the first seven weeks of wMel exponential growth for the cell lines and replicates in (F).The slope of these plots was used to calculate the selection coefficients in Table S1.

Deterministic growth: wMel's selective advantage is independent of starting infection frequency
The w Mel strain outcompetes w Ri when it is the minority strain in host cell culture cells, indicating that w Mel is a deterministic competitor whose selective advantages are not dependent on starting infection frequency.To assess whether w Mel's competitive advantage is frequency-dependent or deterministic, we mixed w Mel-infected and w Ri-infected cells at approximately 1:100 and 1:1000 ratios based on the relative genomic titers of the respective strain in the stable-infected cell lines.The w Ri strain is at lower titer than the w Mel strain in both S2 and JW18 cells, limiting the titer mixtures to this value (~0.3-4.1).Relative titers were measured by Illumina whole genome sequencing each week over 11 weeks.Similar to the equal titer mixtures, we observed a rapid increase of the frequency of w Mel within five to seven weeks post-mixing in both the 1:100 and 1:1000 mixtures across both cell lines and all replicates (Figure 2A,B).However, in contrast to the 1:1 mixtures, the w Mel strain required more time to become fixed, only reaching an average proportion of total coverage of 86% by week 10 in both the 1:100 and 1:1000 mixtures.
The frequency of w Mel relative to w Ri increased continually over the 11 week experiment in both cell lines in the 1:100 mixtures, allowing us to estimate the strength of selection acting on w Mel over the total length of the experiment (Figure 2C).However, in the 1:1000 mixtures wMel was undetectable in week 0, highlighting the extreme disadvantage in initial frequency when compared to wRi.Therefore, we estimated selection coefficients for these mixtures from week 1 onwards (Figure 2D).In the 1:100 mixtures, selection coefficients ( ω) for w Mel ranged between 2.62-2.92and 2.31-2.33 in the JW18 and S2 cell lines, respectively.In the 1:1000 mixtures ω ranged from 3.36-3.63 in the JW18 cell line, and 2.76-2.87 in the S2 cell line (Fig 2F ,G).Interestingly, we found that in the 1:1000 mixtures, the wMel strain grows significantly faster than wRi and exhibits higher selection coefficients than in the 1:100 mixtures across both cell lines (Fig S4).This suggests that wMel is able to modulate its growth rate to more efficiently populate host cells when starting at a lower initial frequency.Given Wolbachia's propensity for recombination [22][23][24] , we tested for the presence of recombinant haplotypes between the competing strain genomes in the 1:1, 1:100, and 1:1000 wMel:wRi mixed infection experiments (Supplemental Fig S5).Recombinant haplotypes were detected by their chimeric alignments to both the w Mel and w Ri genomes in regions of high mappability.While more of these alignments were found in cultures with more equal mixtures of the two strains, they were still quite rare: at most 1 in 500,000 alignments were recombinant.The highest recombinant fractions occurred when strains co-occurred the longest, in the 1:1000 S2 mixtures.These results make intuitive sense, as recombination mediated through passive processes such as homology-directed repair with divergent strain eDNA requires high concentrations (equal strain mixtures) and many chances (long co-culture times).
The competitive dynamics between w Mel and w Ri in our in vitro experiments offer insight into the mechanisms that might limit the frequency and stability of mixed infections in vivo, in nature.The quick and reproducible competitive exclusion of w Ri by w Mel in two D. melanogaster cell types across a range of starting frequencies suggests that mixed infections resolve reliably and quickly, consistent with theoretical predictions [25] .This potentially explains why unstable mixed infections (opposed to stable superinfections) are rarely observed in nature [17][18][19] .The selection coefficients estimated for w Mel demonstrate a strong relative fitness compared to w Ri across both D. melanogaster cell lines.However, w Mel is natively associated with D. melanogaster, therefore this competitive advantage may reflect host-specific adaptations [26] .To explore whether the relative superiority of w Mel as a cellular symbiont is specific to its native host, we immortalized a D. simulans cell line to repeat these investigations in w Ri's native host background.The slopes from F and G were used to calculate the selection coefficients (ω) overlaid in the plots, also in Table S1.

Reciprocal infections: The wMel strain maintains its competitive advantage in wRi's native host D. simulans
To assess the contribution of host-specific adaptations to the competitive advantage of wMel in Drosophila melanogaster , we immortalized a new D. simulans cell line from the white eye fly stock infected with the Riv84 w Ri strain named Dsim6B.Initially, these cells were heterogeneous and infected with w Ri (Fig 3A).Often the wRi-infected cells exhibited aberrant cellular morphologies.As the Dsim6B cell line stabilized and became more clonal, the infection was lost (Fig 3A , S5).Despite high w Ri titers in D. simulans in vivo fly tissues (4.5x average genomic titer, Figure 3B) [6] , repeated attempts to reinfect the cells with w Ri from the stably infected D. melanogaster cell lines via the shell vial technique failed (Figure 3C, S6).In contrast, infections of Dsim6B cells with the w Mel strain were very successful (Fig S6 ), and the rate of titer increased to stable levels of 1-2x genomic titer depending on the initial input concentration (Figure 3D).
The differential success of w Ri and w Mel infections observed in our D. simulans cell line suggests that host developmental programs may enable the persistence of costly Wolbachia infections.Cell culture conditions are distinguished from in vivo conditions primarily by their simplicity of cell and organism types (sterile monoculture for both host and symbiont), which wRi may be poorly evolved to handle, despite its close relationship to wMel (99.91% identical across the 1.3-1.4Mb genomes).Alternatively, wRi may be a better "developmental symbiont" than a "cellular symbiont".The w Ri strain's in vivo high titers and promiscuity across fly species suggests that its persistence may be heavily reliant on a developmentally-constrained system in which the maintenance of specific host cell lineages is crucial for organismal survival.In a cell culture system, cells can replicate freely because they are free of the limitations placed on cell proliferation in a developing host.Consequently, if the growth of uninfected cells outpaces infected cells, then the infection will be lost.Given that we were able to establish and maintain w Mel infections in both D. melanogaster and D. simulans cell lines, w Mel may not rely as heavily on the developmental context of the host as w Ri.To explore this idea further, we characterize the growth dynamics of each strain into uninfected host cells over time.A) The Dsim6B cell line was immortalized from D. simulans [w-] embryos infected with the Riv84 strain of w Ri.The primary and early immortalized cell line was infected with wRi, but the bacteria were gradually lost as the cells increased in growth rate and clonality.By nine months post-infection the Dsim6B cell line had cured itself of its wRi infection.Repeated attempts to reinfect the Dsim6B cell line with wRi were unsuccessful.B) Bar plots of stable wMel (blue) and wRi (red) titers in D. melanogaster and D. simulans cells and flies (three bars indicated with fly icons).C) FISH widefield images of Dsim6B cell lines uninfected (0:1 titer), infected with the wMel strain (2:1 titer), and after attempts to reinfect with the wRi strain (~0:1).D) Titer increase over time in the Dsim6B cell line infected with wMel via the shell vial technique at 1/10x the concentration in JW18 cells and at 1x, compared to stable Dsim6BwMel cell line infections (maintained for more than three months).

Infection expansion into uninfected host cells recapitulates w Mel's spread into w Ri-containing cells
Successful Wolbachia cellular infections require healthy host cell growth, in addition to some rate of bacterial segregation during host cell division and cell-to-cell transfer to uninfected host cells.The weights of these three parameters are interdependent: if infections impact host cell growth, then cell-to-cell transfer rates need to be high to enable the infection of faster-growing uninfected cells.Otherwise, the infection will be lost due to uninfected cell overgrowth.Similarly, cell-to-cell transfer rates can only be negligible if the infection has minimal cost on host cell growth rates and segregation is efficient.
To understand the cellular basis for w Mel's competitive advantage over w Ri in vitro , we studied the expansion of these Wolbachia strains into uninfected host cells, revealing that w Ri fails to establish when fewer than 50% of host cells are infected.We mixed JW18 and S2 cells infected with the wMel or the wRi strain of Wolbachia and uninfected at approximately equal quantities.Infection growth curves following the addition of 1:1 uninfected host cells to w Mel-infected cell lines revealed a similar pattern of expansion as in the w Melw Ri competition experiments: across both the JW18 and S2 cell lines and all three replicates, w Mel genomic titer increased rapidly in the first five weeks, and remained at a relatively stable frequency throughout the rest of the experiment (Fig 4A , S7).On average, w Mel titer increased by 17% and 16% per week in the JW18 and S2 cell lines, respectively (Fig 4E, Table S2).Conversely, in the wRi-DOX mixtures, we observed the continuous decline of w Ri genomic titer in the JW18 cell line, with an average rate of 14% per week.Similarly, in the S2 cell line w Ri genomic titer declined on average by 10% per week, despite the initial increase in the first week post co-culture (Fig 4B ,E, Table S2).Overall, the observed patterns of wMel's growth in the wMel-DOX mixtures illustrate that the symbiont can effectively establish and expand an infection within the cell lines, and suggests horizontal transmission as a mechanism for infection establishment.To assess the impact of Wolbachia infection on host cell dynamics, we next compared the growth rates of infected and uninfected D. melanogaster cells.
Measuring the growth rate of D. melanogaster cells with and without Wolbachia infections revealed that both strains slow host cell division, suggesting that successful establishment requires cell-to-cell transfer.Both JW18 and S2 cell lines divide significantly faster when uninfected than when infected with either the wMel or wRi strain of Wolbachia ( p <0.01 Wilcoxon rank sum test; Fig 4C) .When uninfected, JW18 cells double in 2.09 +/-0.31days (a growth rate of 1.45x cells per day), whereas wMel-infected JW18 cells require 3.75 +/-0.34 days to double (0.385x cells per day) and wRi-infected JW18 cells require a massive 25.0 +/-29.3days to double (0.16x cells per day).Similarly, when uninfected, S2 cells double in 2.02 +/-0.37 days (1.73x cells per day).When infected with the wMel strain, S2 cells require 3.71 +/-1.06 days to double (0.46x cells per day) and infected with wRi, they require 3.21 +/-0.63 days to double (0.57x cells per day).Interestingly, wMel Wolbachia infection has minimal impact on the D. simulans Dsim6B cell line (3.11+/-1.26vs 3.23 +/-0.95days to double and 0.86x and 0.64x cells per day, respectively).This may be due to the Dsim6B cell line's lower growth rate: this cell line is highly adherent and fails to grow well at the ⅙ starting dilution that the uninfected S2 and JW18 D. melanogaster cell lines thrive with.
The negative impact of Wolbachia infection on host cell growth combined with wMel's ability to rapidly increase in titer upon exposure to uninfected host cells indicates that cell-to-cell transfer is essential to the colonization process.D. melanogaster host cells require nearly twice as long to divide when infected with Wolbachia than when uninfected (Fig 4C)   S2).F) Cell growth rates measured by hemocytometer cell counts, quantified as the proportional growth per day from the starting cell count at 23°C (purple) and 26 ° C (orange).Wilcoxon rank sum p-values **p<=0.01 and ***p<=0.001.

Conclusion
Wolbachia pipientis is an obligate intracellular alphaproteobacterium that infects a diverse range of arthropods, many of which are disease agents, vectors, and agricultural pests [10] .Composed of genetically distinct strains spanning 16 lineage groups [27] , Wolbachia demonstrate a variety of interactions with their hosts, ranging from mutualism to reproductive parasitism [28] .The widespread prevalence of Wolbachia is largely due to its ability to rapidly shift to new and diverse hosts [4,29] , but little is known about the microevolutionary events that occur immediately after a strain infects a novel host.Here, we used a Wolbachia-infected Drosophila melanogaster cell line system to investigate the outcomes of mixed and novel infections in vitro .
Our findings provide valuable insight into the ability of an invading Wolbachia strain to establish an infection in a host already infected by a different, resident Wolbachia strain.We show that wMel consistently emerges as the dominant strain, quickly and effectively supplanting wRi in mixed infections, independent of starting frequency.These results confirm predictions made by Keeling et al. in 2003, that one strain is always driven extinct in homogeneous mixed infections.However, the strain that wins is not determined by founder effects in the wMel-vs-wRi case, but the differential intrinsic abilities to propagate and colonize new host cells.These quick and reproducible resolutions of mixed infections in our cell culture system suggest an explanation for the paucity of observations from nature: mixed infections resolve quickly by competitive exclusion, before they can be sampled.
In addition to providing insight into Wolbachia infection establishment and mixed infection dynamics, this work highlights the potential role host development may play in determining the success or failure to establish an infection.Despite the promiscuous w Ri's strain's relatively high titer in whole-fly extracts (4.5x vs. wMel at 0.79x, Fig 3B ) and tissues [6,30] , it occurs at titers an order of magnitude lower than w Mel in D. melanogaster cell lines (Fig S2C ,F) and fails to persist in cell lines derived from its native D. simulans host (Fig 3C).This suggests that w Ri is costly at the cellular level and in vivo development offers a mechanism of protection from loss because most cell lineages are required for normal development.Similarly, wMel's higher titer in vitro than in vivo suggests that host development and non-cell autonomous mechanisms are involved in their regulation in nature.Thus, in a developmentally constrained system, w Ri's high cellular cost and failure to transmit to uninfected cells (Fig 4B ,C) do not prevent its persistence like they do in vitro .
The future of Wolbachia-mediated host biological control applications rely on understanding the mechanisms of novel Wolbachia infection and persistence in non-native hosts.From understanding which cell types and developmental time points different strains have affinities for, to predicting the outcome of rare mixed infections in unintended hosts, this work offers a powerful platform to disentangling bacterial-vs-host and cellular-vs-organismal driven phenotypes.Given that rare horizontal transmission events can produce mixed infections in novel hosts that may persist, generate recombinant Wolbachia strain genomes [5,31,32] , and have unintended ecosystem-level consequences, these results are vital to future safe applications of Wolbachia in the field.

Cell Culture Maintenance and Cell Line Generation
All Drosophila cells were maintained on either Shields and Sang M3 Insect Medium (MilliporeSigma S3652) or Schneider's Insect Medium (MilliporeSigma S0146) supplemented with 10% v/v Fetal Bovine Serum (FBS, ThermoFisher A3160502).Cells were maintained in 4 mL of media in plug-seal T-25 flasks (Corning 430168) in a refrigerated incubator at either 25-27°C or 22.5-23.5°C,as indicated in the text.We performed weekly cell splits at a 1:6 dilution for uninfected cell lines and 1:2 or 1:3 dilutions for Wolbachia -infected cell lines, following visual inspections of cell growth and contamination.Adherent cells were removed by scraping with sterile, bent glass pipettes.Transitions between media types were performed in 25% intervals, requiring four weeks to transition to 100% Shields and Sang or Schneider's Medium.
Drosophila melanogaster JW18 cells [20] and S2 cells (Thermo Fisher and [21] ) were derived from a primary culture of 1-15 hr and 20-24 hr-old embryos, respectively.JW18 cells are naturally infected with the w Mel strain of Wolbachia (from the in vivo infection in the fly line the cells were derived from) and S2 cells are naturally uninfected.We found that incubation temperature exerts an observable effect on Wolbachia density within these cell lines, supporting the well established relationship between Wolbachia density and temperature [33] .Specifically, cells cultured at 26°C in 2021 exhibited higher symbiont titers compared to the same cultures incubated at 23°C in 2023.Importantly, the relative differences between wMel and wRi titer are consistent between these temperature regimes: wMel is always at an order of magnitude higher titer than wRi.The difference in incubator temperatures was necessitated by the last author's starting her new lab and buying a new incubator capable of maintaining 23°C.To generate uninfected JW18 cells, we treated JW18 wMel-infected cells with 10 µg/mL doxycycline in supplemented Shields and Sang media.
We generated the Drosophila simulans Dsim[w-]6B cell line from a w[-] (white eye) fly line previously infected with the Riverside 1984 strain of w Ri Wolbachia [34,35] according to the method described in [36] .Briefly, 1-20 hr old embryos laid on grape-agar plates by Wolbachia -infected flies were collected, surface sterilized, homogenized, and plated in flasks on rich media containing 20% FBS and Wolbachia-resistant antibiotics, 60 and 100 μg/ml penicillin-streptomycin and 50 μg/ml gentamicin.During the next six months of maintenance, two of the initial twenty seed flasks converted into immortal tissue culture lines.The Dsim[w-]6B cell line was selected for further pursuit due to its planar growth pattern and ability to hold a Wolbachia infection.The native wRi infection is unstable in Drosophila in vitro culture systems long-term, as described in the Results and Discussion sections, and the natural infection was lost naturally over the course of the first year of culture.
Wolbachia infections were introduced by adding 1.2 µm-filtered infected cell lysate to uninfected D. melanogaster JW18 and S2 and D. simulans Dsim[w-]6B cells.Infected cell lysate was either obtained from Wolbachia-infected cell cultures or fly embryos (collected on grape plates, as described in [36] .Infected cells were serially passed through 5 µM and 1.2 µM syringe filters to produce Wolbachia -containing cell lysate.The w Mel strain was applied directly (in 3 mL lysate) to uninfected S2 cells to produce the S2wMel cell line in 2017.To produce the wRi-infected cell lines and the Dsim[w-]6B cell lines, we applied 0.5-1 mL of Wolbachia -containing cell lysate to a monolayer of uninfected host cells in a flat-bottom shell vial and centrifuged the bacterial cells down onto the cell surface in a swinging bucket centrifuge at 2500 x g for 1 hr at 15°C ( i.e. , the shell vial technique [37] ).We transferred these cells to T-12 flasks in a final volume of 2 mL for five days before scraping and transferring the cells to a T-25 flask with 2 mL of fresh media.These lines were maintained by weekly 1:2 "soft splits", which removed no media.
All cell lines were validated with DNA-based probes and whole genome sequencing after construction and continuously during maintenance and experimentation.Cell line infection status was continuously monitored by PCR and fluorescence in situ hybridization (FISH) of Wolbachia -specific markers.Primers for the Wolbachia Surface Protein (WSP) gene [35] were used to confirm the presence and absence of wMel and wRi strains in infected and uninfected cell lines, respectively.Sanger sequencing of the WSP amplicons was performed by Azenta to confirm the strain-specific amplicon sequences.Oligonucleotide DNA probes complementary to the Wolbachia 16S rRNA sequence were used in FISH experiments following the protocol in [38] to confirm infections and estimate per-cell Wolbachia titer.Whole genome shotgun sequencing was performed with Illumina sequencing (see below) to confirm host species and Wolbachia strain identities and test for contamination.

Mixed cell line experiments
We used average genomic titer measurements from sequencing stable cell line infections to calculate how many cells of each infection and host cell type to mix for the desired mixture ratios.Host cell concentrations were quantified with a hemocytometer manually or with a Millicell® Digital Cell Imager.
For strains A and B at titers of Y A and Y B symbiont cells/host cell, within host cells growing at densities of X A and X B cells/mL, mixed at a ratio of A/B, in a final volume of 4 mL per cell culture flask: Volume of host cell culture infected with strain A = V A = 4 mL/(X A /X B * Y A /Y B * 1/(A/B) + 1) Volume of host cell culture infected with strain B = V B = 4 mL -V A Samples were collected prior to mixing, immediately after mixing, and weekly when splitting infected cell cultures into new flasks at 1⁄2 dilutions.For each culture at each timepoint, one mL of scraped and mixed cell-containing media was transferred to a 1.5 mL Eppendorf tube, the cells were centrifuged at 16,000xg at 4-10°C for 10 min, supernatant was discarded, and the cell pellet was snap-frozen and stored at -80°C until DNA extraction.Pellets were processed for library prep within one month of sample collection.

Shell vial experiments
To monitor how Wolbachia infections spread across uninfected host cells following introduction with the shell vial technique [37] , we performed shell vial infections as described above for the creation of novel cell lines.Given the limited material at the start of these protocols (~2 mL per experiment and < 1 million cells), we waited until the transfer step to T-25 flasks to take the first sample for Illumina sequencing and genomic titer quantification.Host cell-free w Mel Wolbachia lysate was either added at the full concentration derived from host cell lysis or a 1/10 dilution to approximate the lower concentrations exhibited by w Ri infections.

Cell growth rate experiments
Cell line cells were quantified upon splitting and seeding into new flasks and after a week's incubation with a hemocytometer and Millicell® Digital Cell Imager.While handling the cell lines as described above for "Cell Culture Maintenance", we added one extra mL of fresh media to each flask so that one mL could be removed for sampling cell concentration and relative Wolbachia genomic titer (as a final step in the splits).These one mL samples were then quantified by counting cells in a 10 uL volume ( X number of cells (>100) measured per Y number of boxes (>1 if <100 cells/box) * W dilution factor (2 if diluted by ½) * 10,000 mL -1 = Z number of cells/mL).The rest of the cell suspension was pelleted by centrifugation (as described above), snap-frozen, and stored at -20-80°C until DNA extraction.This process was repeated one week later, except cells were resuspended by scraping prior to media removal so that the week's worth of growth could be quantified.Cells were then diluted as described above for normal maintenance.This modified step was repeated every other week for six weeks, at most frequent.

Cell Imaging and Image Analysis
Cell lines and experiments were continuously monitored with a tissue culture (TC) microscope and imaged with a monochromatic digital camera.Weekly, stable line and experiment cell splits were imaged on Zeiss Primovert TC microscope or a Leica DMi8 inverted microscope for confluency and contamination.
Infections were confirmed by fluorescence in situ hybridization (FISH) using DNA oligonucleotide probes complementary to the Wolbachia 16S ribosomal RNA sequence, following the protocol in White et al. 2017.Briefly, for each cell type, infection state, or experimental replicate, 1 mL of confluent cells were pipetted from a T-25 flask into a 6-well dish (Corning) one to three days before fixation.Upon confluency in the dish, cells were fixed in 8% paraformaldehyde in 1x phosphate-buffered saline (PBS) for 15 min at room temperature (RT).Following two washes with 1x PBS, cells were treated with prehybridization buffer, consisting of 50% deionized formamide by volume, 4x saline sodium citrate (SSC), 0.5x Denhardt's solution, 0.1 M dithiothreitol (DTT), and 0.1% Tween 20 in deionized water, for one and a half hours.Following prehybridization, cells were incubated in hybridization buffer (prehybridization buffer without Tween 20) containing 500 nM Wolbachia W2 fluorescent DNA probe (5-CTTCTGTGAGTACCGTCATTATC-3) (Bioresearch Technologies) [39] at 37°C overnight.Wet kimwipes were added to the dish to prevent dehydration.The next day, cells were washed three times with 1x SSC with 0.1% Tween 20 at RT quickly, at RT for 15 min, and at 42°C for 30 min.Next, the cells were washed with 0.5x SSC at RT quickly, at 42°C for 30 min, and at RT for 15 min.These stringent washes aimed to remove unbound FISH probes from the cells.Finally, cells were washed three times with 1x PBS at RT before either staining with 3uM DAPI (4',6-diamidino-2-phenylindole) in 1x PBS for 10 min or mounting in Vectashield fluorescent mounting medium with DAPI (Vector Laboratories).
FISH experiments were imaged on a Leica DM5500B widefield microscope or an inverted DMi8 equipped with LEDs for epifluorescence imaging.Raw Leica images were processed in Fiji [40] and analyzed in R [41] .

DNA Extraction
Cell pellets were lysed and digested using lysis buffer (100mM NaCl, 50mM Tris-HCl pH 8, 1mM EDTA pH 8, 0.5% sodium dodecyl sulfate) and Proteinase K (NEB).Reactions were incubated at room temperature overnight.Genomic DNA was purified from cell lysates using SPRI beads and quantified using a Thermo Fisher Qubit fluorometer and Qubit dsDNA Broad Range assay kit.

Tn5 Library Prep
We generated short-read sequence libraries using a custom tagmentation protocol adapted from [42] .The full protocol is available on https://www.protocols.io/[43] .Briefly, Tn5 Tagmentation reactions were prepared as follows: 10ng gDNA, 1uL Tn5-AR, 1uL Tn5-BR, 4uL TAPS-PEG 8000 and nuclease free water to final volume of 20uL.See Table S3 for Tn5-A, -B, and -R oligo sequences.Reactions were incubated at 55C for 8 minutes then killed by transferring to ice and adding 5uL 0.2% sodium dodecyl sulfate.Tagmentation product was amplified using the KAPA Biosystems HiFi polymerase kit and unique indexed primers.Pooled libraries were size selected using the Zymo Select-a-Size DNA Clean & Concentrator Kit and NEB Monarch Gel Extraction Kit.Library pools were then quantified using the Qubit dsDNA HS Assay Kit and the Agilent TapeStation.

Data processing
We developed a Snakemake [44] workflow to estimate symbiont titers from the raw sequencing data ( https://github.com/cademirch/wolb-cov-workflow).First, we generated a composite reference genome consisting of the host and symbiont genomes (Table S4).We then calculated per-base mappability scores across the merged genome using genmap [45] with the parameters "-k 150 -e 0" .Next, reads are trimmed of sequencing adapters and filtered for quality using fastp [46] .We aligned the filtered reads to the composite reference genome using bwa mem [47] .The resulting alignments were then filtered using samtools [48] , keeping only unique alignments with a mapping quality greater than 20.Additionally, we used sambamba [49] to mark optical duplicates in the filtered alignments.Next, we calculated the mean depth for each mappable (mappability == 1) position in the merged genome using mosdepth [50] .Using mean depth statistics, we estimated symbiont genomic titer using the following equation: Symbiont titer = mean depth of symbiont / mean depth of host Note, we only considered the 5 autosomal chromosomes (2L, 2R, 3L, 3R, and 4) of the host Drosophila genome for our titer calculations.

Selection coefficient calculation
We leveraged population genomics theory on selection between competing strains in a chemostat [51] to model selection in Wolbachia -infected Drosophila cell culture.Weekly splits with removal and disposal of the overlying media approximate a chemostat, as the number of cells is kept within a tolerance range and the physical and chemical resources are kept plentiful.
In a bacterial chemostat (extracellular or intracellular), the frequencies of strains A and a under selection at time t can be shown to be p t =(p t -1) ω 11 / ὼ and q t =(q t -1) ω 22 / ὼ , respectively.Let the selection parameter ω = ω 11 / ω 22 .
Measuring strain A's fitness as a fraction of strain a's fitness from one generation to the next is described by the equation p t /q t =(p t-1 /q t-1 ) ω .Solving for any generation gives the formula, p t /q t =(p 0 /q 0 ) ω t .The plot of ln(p t /q t ) should be linear with a slope equal to ln ω: ln(p t /q t ) = ln(p 0 /q 0 ) + tln ω.
Thus, the selection coefficient ( ω ) for strain A versus strain a is given by the e raised to the slope of the line fit to the plot of relative strain frequency over time.In our calculations, p = frequency of w Mel and q=(1-p) = frequency of w Ri.Thus ω reflects w Mel's selection coefficient relative to w Ri.We fit linear regressions to the w Mel growth curves using R v4.1.2[41] and ggplot2 v3.4.1 [52] .
In order to understand the effects of cell line and starting infection ratio on relative strain frequency over time, and consequently selection coefficients, we used a linear mixed-effects model with autoregressive moving average using nlme v3.1-146 [53] .Then, we plotted the observed points, fitted lines, and 95% confidence intervals using ggplot2 v.3.4.1.

Recombinant haplotype detection with Illumina sequencing
To detect potential recombinant haplotypes stemming from recombination between wMel and wRi, we used a custom Python script to select paired-end alignments where one end aligned with wMel and the other wRi.Then, using samtools, we filtered these chimeric alignments, keeping only alignments that overlapped a "mappable" region within the respective genome of that alignment.By filtering the alignments this way, each end of alignment is anchored to a sequence that is unique to each symbiont genome, suggesting that the alignment is truly chimeric, likely due to recombination.Normalized recombinant read counts were calculated by dividing the number of chimeric reads by the total number of sequence reads for a given sample.RCD: study conception and design and writing.SLR: study conception and design, data production and analysis, and writing.

Fig S1. Schematic overview of steps required for successful horizontal transmission.
Host-switching of an endosymbiont requires successful horizontal transmission, intracellular proliferation, germline targeting for vertical transmission, and a mechanism for population establishment.Here, we use an in vitro Wolbachia -infected cell culture system to study the early stages in this process (#1 and 2 in bold) that are often lost to chance.By focusing on closely related strains with promiscuous and stable host-associations, we can understand how cell identities, divergent hosts, and resident strains impact novel infection events.The S2 and JW18 D. melanogaster cell lines were derived previously from fly embryos of unknown infection status and infected with wMel, respectively.The Dsim6B cell line was derived in this work, from embryos from the D. simulans white eye fly line infected with the Riv84 wRi strain (see methods panel through embryo homogenization).Uninfected cell lines were obtained by treatment with 10 µg/mL doxycycline (DOX) in the cell culture media for nine weeks, followed by at least two months recovery from antibiotic treatment mitochondrial effects.Wolbachia strains were swapped among cell lines with the shell vial technique (see methods panel through shell vial technique).The wMel strain is consistently at ~10x higher titer than the wRi strain in D. melanogaster cells.Titers measured in 2021 were from cells maintained at 25-26°C, whereas titers measured in 2023 were from cells maintained at 23°C.Temperature has a similar impact on both strains titers, with both exhibiting proportionately lower titers at 23°C than 25-26°C.

Fig S4. Mixed Effects Regression analysis of relative strain frequency
To assess how cell line and initial infection ratios influenced wMel's competitive advantage over wRi, we utilized a linear mixed-effects model incorporating these variables as fixed effects.Prediction lines and 95% confidence intervals from the model and observed points for the two cell lines A) JW18 and B) S2 at starting ratios 1:100 (red) and 1:1000 (blue).

Fig S8 . Log-linear regression analysis of infected-uninfected mixtures
Log-linear regression analysis for A) wMel and B) wRi in 1:1 mixtures with uninfected cells.Regression summary statistics are annotated in Table S2.

Table S1. Selection coefficients estimated in competition experiments.
Table S2.Regression statistics from log-linear regression analysis of wMel:DOX and wRi:DOX experiments.Table S3.Oligonucleotide sequences used for Tn5 based library preps.Table S4.NCBI RefSeq genome accessions for reference genomes used in bioinformatics analyses.

Fig 1 .
Fig 1.The w Mel strain consistently outcompetes the w Ri strain in D. melanogaster cell culture.A) Schematic overview of the 1:1 w Ri: w Mel mixed infected cell line experiment.B-E) Tissue culture micrographs of the mixed cell lines at B,D) week 3 and C,E) week 4. B,C at 40x and D,E at 20x magnification.F) Proportion of Illumina whole genome sequencing coverage mapped to the w Mel (blue) and w Ri (red)genomes out of the total coverage mapped to all Wolbachia and D. melanogaster host genomes, plotted by replicate and host cell type (S2, dashed or JW18, solid).G) Relative growth rates of wMel compared to wRi over the first seven weeks of wMel exponential growth for the cell lines and replicates in (F).The slope of these plots was used to calculate the selection coefficients in TableS1.

Fig 2 .
Fig 2. The w Mel strain deterministically outcompetes the w Ri strain in mixed infections, even when starting at only 1/100th or 1/1000th the frequency of w Ri.A) Schematic overview of the 1:100 and 1:1000 wRi:wMel mixed infected cell line experiments.B,C) Representative epifluorescence FISH images of week two of the 1:1000 wMel:wRi mixture (replicate A).B) JW18 cell line and C) S2 cell line at 20x; scale bar = 50 µm, DAPI=blue, Jupiter-GFP=green (JW18 only), and Wolbachia 16S rRNA=red.D,E) Proportion of w Mel (blue) and w Ri (red) genome coverage out of the total coverage of all Wolbachia strains and D. melanogaster host genomes, plotted by replicate and host cell type (S2, dashed or JW18, solid) in mixed infections started at wMel:wRi ratios of D) 1:100 and E) 1:1000.F,G) Relative growth rates of w Mel compared to w Ri in mixed infections started at F) 1:100 and G) 1:1000 ratios.The slopes from F and G were used to calculate the selection coefficients (ω) overlaid in the plots, also in TableS1.
. The loss of wRi from the 1:1 DOX-wRi mixtures is consistent with the replacement of infected cells with faster-growing uninfected cells over the ten weeks of co-culture (Fig 4B).In contrast, wMel's increase in frequency over time after 1:1 mixture with DOX-cured host cells (Fig 4A), despite their inhibition of host cell division rates (Fig 4C), is consistent with efficient cell-to-cell transfer to uninfected host cells.This transfer process not only increases wMel frequency in the culture, but also prevents uninfected host cells from remaining uninfected and out-growing the infected cell population.

Fig 4 .
Fig 4. The wMel strain is able to efficiently spread to uninfected cells through faithful segregation and cell-to-cell transfer, whereas the wRi strain cannot .A) Schematic overview of the 1:1 wMel:DOX and wRi:DOX mixed infected cell line experiments.C,D) Genomic titers for C) wMel (blue) and D) wRi (red) over time in 1:1 mixtures with uninfected JW18 (solid line) and S2 (dashed line) cells.E) Fold change in symbiont titer per week in each mixture.Fold change was calculated by log-linear regression (Fig S8, TableS2).F) Cell growth rates measured by hemocytometer cell counts, quantified as the proportional growth per day from the starting cell count at 23°C (purple) and 26 ° C (orange).Wilcoxon rank sum p-values **p<=0.01 and ***p<=0.001.

Fig S2 .
Fig S2 .Overview of Drosophila in vivo and in vitro resources.The S2 and JW18 D. melanogaster cell lines were derived previously from fly embryos of unknown infection status and infected with wMel, respectively.The Dsim6B cell line was derived in this work, from embryos from the D. simulans white eye fly line infected with the Riv84 wRi strain (see methods panel through embryo homogenization).Uninfected cell lines were obtained by treatment with 10 µg/mL doxycycline (DOX) in the cell culture media for nine weeks, followed by at least two months recovery from antibiotic treatment mitochondrial effects.Wolbachia strains were swapped among cell lines with the shell vial technique (see methods panel through shell vial technique).

Fig S3 .
Fig S3.Natural and introduced Wolbachia infections in D. melanogaster cell lines are stable over time.The wMel strain is consistently at ~10x higher titer than the wRi strain in D. melanogaster cells.Titers measured in 2021 were from cells maintained at 25-26°C, whereas titers measured in 2023 were from cells maintained at 23°C.Temperature has a similar impact on both strains titers, with both exhibiting proportionately lower titers at 23°C than 25-26°C.

Fig S5 .
Fig S5 .Recombinants detected between the wMel and wRi strains in D. melanogaster cell culture.Recombinant alignments were detected by extracting the reads chimerically mapped to the wMel and wRi genomes in regions of high mappability (containing SNPs, indels, or structural variation).