Fitness costs and benefits vary for two facultative Burkholderia symbionts of the social amoeba, Dictyostelium discoideum

Abstract Hosts and their associated microbes can enter into different relationships, which can range from mutualism, where both partners benefit, to exploitation, where one partner benefits at the expense of the other. Many host–microbe relationships have been presumed to be mutualistic, but frequently only benefits to the host, and not the microbial symbiont, have been considered. Here, we address this issue by looking at the effect of host association on the fitness of two facultative members of the Dictyostelium discoideum microbiome (Burkholderia agricolaris and Burkholderia hayleyella). Using two indicators of bacterial fitness, growth rate and abundance, we determined the effect of D. discoideum on Burkholderia fitness. In liquid culture, we found that D. discoideum amoebas lowered the growth rate of both Burkholderia species. In soil microcosms, we tracked the abundance of Burkholderia grown with and without D. discoideum over a month and found that B. hayleyella had larger populations when associating with D. discoideum while B. agricolaris was not significantly affected. Overall, we find that both B. agricolaris and B. hayleyella pay a cost to associate with D. discoideum, but B. hayleyella can also benefit under some conditions. Understanding how fitness varies in facultative symbionts will help us understand the persistence of host–symbiont relationships. OPEN RESEARCH BADGES This article has earned an Open Data Badge for making publicly available the digitally‐shareable data necessary to reproduce the reported results. The data is available at https://openscholarship.wustl.edu/data/15/

independent fitness interests that often conflict with those of their host (Douglas & Smith, 1989;Garcia & Gerardo, 2014;Mushegian & Ebert, 2016). Understanding these symbiont fitness interests and accounting for the costs and benefits of symbiosis in microbes have largely been neglected. However, symbiont fitness is important for understanding how and under what conditions symbiotic host-microbe interactions persist.
There are currently multiple evolutionary explanations for the persistence of host-microbe interactions. Pathogenic infections are a common host-microbe interaction, for which hosts have evolved to resist or escape, but the interaction can persist because the microbe benefits. Long-term persistence is more likely, though, when both the host and microbe benefit, that is, when their interaction is mutualistic. If both the host and the microbial symbiont have higher fitness when in association, then selection will act upon both partners to maintain the interaction, even though conflict persists. This has largely been assumed to operate within symbiotic host-microbe interactions in which the host benefits from the microbe. There is evidence supporting this mutualistic explanation, showing that host association is responsible for larger populations of facultative symbionts (Kuykendall, 1989;Lee & Ruby, 1994;Storelli et al., 2018).
For example, living in the gut of Drosophila leads to a larger overall population of the bacterium Lactobacillus plantarum compared to just living in the environment (Storelli et al., 2018). An alternative explanation for the maintenance of host-microbe symbioses is exploitative host control, interactions in which the host benefits from a symbiont but the symbiont does not benefit [also called extortion or imprisonment (Garcia & Gerardo, 2014)]. Although mutualistic interactions commonly impose costs on symbionts, exploitative host control differs in that the costs are not offset by the benefits, when averaged across all conditions. The role of exploitative host control in stabilizing symbioses has been modeled (Frean & Abraham, 2004;Hilbe, Nowak, & Sigmund, 2013), but only a few studies have shown that it operates in natural systems (Johnson, Oldach, Delwiche, & Stoecker, 2007;Lowe, Minter, Cameron, & Brockhurst, 2016).
One reason that the effect of symbiosis on microbial symbionts has been neglected is that it can be difficult to quantify symbiont fitness and especially difficult to compare symbiont fitness in different environments (Garcia & Gerardo, 2014). Focusing on horizontal or environmentally acquired symbionts resolves some of these issues because many can be grown in culture or in nonhost environments in the laboratory (Takeshita & Kikuchi, 2017). However, it is sometimes unclear how symbiont fitness should be measured. The number or biomass of nodules on legume roots is frequently used as a rhizobial fitness proxy, but it can be problematic because the relationship between nodule size and symbiont abundance varies between rhizobia isolates (Ratcliff, Underbakke, & Denison, 2012), and nodules can contain more than one isolate (Denison & Kiers, 2011). Fungal symbionts such as arbuscular mycorrhizae and lichen fungi grow filamentously, making it hard to count individual cells (Pringle & Taylor, 2002). They can also have complex lifecycles that can include sexual and asexual reproduction. In addition, hosts can provide benefits such as dispersal (Nazir, Tazetdinova, & Elsas, 2014) or protection from predators that only have an effect on symbiont fitness in certain environments.
Here, we use a simple growth-based approach to determine the effect of host association on bacterial fitness. Although it seems intuitive, comparing bacterial abundance in host and nonhost environments (e.g., comparing symbionts in hosts and soil) can be misleading because it is difficult to determine an equivalent sampling area between the two habitats. Instead, we compare bacterial abundance in an environment with and without hosts (e.g., comparing symbionts in soil with and without hosts; Figure 1d). For the no-host treatment, bacteria are inoculated directly into the nonhost environment and quantified after a period of growth. For the host treatment, bacteria and hosts are added to the nonhost environment and then bacteria both within and outside the host cells are quantified. This is advantageous because it accounts for effects the host has on the bacteria within its cells and in the environment. We also emphasize the use of a natural substrate for the nonhost environment instead of artificial growth media so that we can understand the host-microbe interaction in an ecologically relevant context.
The social amoeba host Dictyostelium discoideum and its minimicrobiome are an ideal symbiotic system in which to quantify the effect of host association on bacterial fitness. D. discoideum live in soil or feces as single-celled, vegetative amoebas. When their bacterial prey is depleted, thousands of D. discoideum amoebas aggregate to form a multicellular slug that disperses through the soil to form a fruiting body that holds spores aloft for further dispersal (Smith, Queller, & Strassmann, 2014;Figure 1a). In the wild, D. discoideum contains a small microbiome of edible and inedible bacteria . Of these bacteria, three Burkholderia species confer upon D. discoideum the ability to carry bacterial prey during dispersal (Figure 1b; Brock, Douglas, Queller, & Strassmann, 2011;Brock, Hubert, et al., 2018;DiSalvo et al., 2015;. This provides a fitness advantage to D. discoideum when there is not an acceptable food source in the new location (Figure 1b; Brock et al., 2011;DiSalvo et al., 2015), but decreases slug migration distance (Brock, Jones, Queller, & Strassmann, 2015) and is costly when a Burkholderia infection is newly established (Shu, Brock, et al., 2018) or when food is abundant (Brock et al., 2011). Burkholderia symbionts provide a further competitive advantage to their hosts by suppressing the growth of nearby D. discoideum uninfected with Burkholderia through the release of small molecules (Brock, Read, Bozhchenko, Queller, & Strassmann, 2013 Shu, Brock, et al., 2018), which makes it likely that they could have an increased dispersal capability, as D. discoideum do (Smith et al., 2014), from sitting atop a stalk. Burkholderia also remain viable throughout the D. discoideum lifecycle and can exit from postdispersal spores and reproduce, further supporting a dispersal advantage. It is unclear what other advantages or disadvantages might apply, but potential benefits include nutrient acquisition and protection from predators or pathogens in the soil. Here, we use growth and abundance as fitness measures as a baseline determination of the costs and benefits to B. agricolaris and B. hayleyella. F I G U R E 1 Overview of the effect of Burkholderia symbionts on the lifecycle of Dictyostelium discoideum and of the experiments done in this study. (a) D. discoideum fruiting bodies showing the sorus, a mass of spores and extracellular matrix, that is held aloft by the stalk. Picture taken by Tyler Larsen. (b) When D. discoideum is not colonized by Burkholderia, (1a) vegetative amoebas feed on bacteria (Klebsiella pneumoniae in our experiments) until they are depleted. The amoebas then aggregate to form multicellular slugs that disperse and eventually form a fruiting body for further dispersal. (1b) Spores in the fruiting body are devoid of prey bacteria. (1c) If the spores are dispersed to a location with sparse or poor quality prey, the amoebas quickly aggregate and produce few spores (Brock et al., 2011). (2a) When D. discoideum is colonized with Burkholderia, some prey bacteria remain and are carried with Burkholderia throughout the aggregation and dispersal of D. discoideum. (2b) As a result, sori are colonized by Burkholderia and prey bacteria. (2c) If the spores are dispersed to a location without prey, D. discoideum can grow and eat the descendants of the prey that were carried through dispersal. Once colonized, D. discoideum can carry Burkholderia and prey bacteria for many generations (DiSalvo et al., 2015). (c) In experiment 1, we compare the growth rates of Burkholderia in liquid culture alone to Burkholderia in liquid coculture with D. discoideum amoebas uninfected with Burkholderia (see Methods for further detail). (d) In experiment 2, we used soil microcosms to measure the abundance of Burkholderia added to the soil as nonsymbiotic cells or in symbiosis with D. discoideum. An equivalent number of Burkholderia were added in both treatments, and Burkholderia abundance was measured at four timepoints  Francis and Eisenberg (1993) Abbreviations: H Arb, Houston Arboretum; MLBS, Mountain Lake Biological Station; L Falls, Linville Falls; Lake I, Lake Itasca; LBG, Little Butts Gap.
a Inedible indicates D. discoideum uninfected by Burkholderia is unable to grow with that Burkholderia isolate as its sole food source.
b Toxicity of Burkholderia cells to D. discoideum uninfected by Burkholderia is indicated by decreased D. discoideum spore production when grown on 5% or 0.25% of the Burkholderia compared to growth only with food bacteria as reported in Haselkorn et al. (2019). Toxicity of Burkholderia supernatant to D. discoideum uncolonized by Burkholderia is indicated by decreased D. discoideum spore production when grown on a filter infiltrated with cell-free supernatant from stationary-phase Burkholderia at a concentration of OD 600 = 1.5 compared to a filter infiltrated with starvation buffer as reported in Brock et al. (2013). In this study, we use the D. discoideum-Burkholderia system to investigate the effect of symbiosis on the fitness of facultative symbionts ( Figure 1). We focused on two of the three Burkholderia symbiont species (B. agricolaris and B. hayleyella) because they encompass most of the phenotypic and genotypic diversity found in the D. discoideum symbionts . In experiment 1, we measured the effect of D. discoideum on the growth rate of independent, nonsymbiotic Burkholderia cells in liquid culture (Figure 1c). Although a less realistic environment, liquid culture allowed us to take the multiple measurements necessary to accurately determine growth rate. We used D. discoideum that were cured of Burkholderia in this experiment so that established Burkholderia symbionts would not confound the host effect. In experiment 2, we measured the abun- is likely to be more adapted to associating with D. discoideum, while B. agricolaris may be less adapted or more likely to be taken advantage of by the host. This is unsurprising since B. hayleyella has a reduced genome consistent with strong host dependence. We discuss the implications these findings have for the evolution of host-microbe interactions along the mutualism-parasitism continuum and the role symbiont fitness has in maintaining symbiotic interactions.

| Culturing and maintenance of D. discoideum clones and Burkholderia isolates
All of the D. discoideum clones used in this study were collected as an effort to establish wild-caught, non-laboratory-adapted D. discoideum clones as experimental models. They were collected from a variety of locations over a period of years, mostly by our laboratory group [but see (Francis & Eisenberg, 1993);

| Experiment 1: Burkholderia growth rate in liquid culture with and without D. discoideum
We measured the maximum specific growth rate of each Burkholderia isolate alone and in coculture with D. discoideum in a Tecan Infinite M200 PRO microplate reader ( Figure 1c). The D. discoideum clones used had been treated to remove pre-existing Burkholderia. We did this by treating the D. discoideum with selective antibiotics and confirming the absence of Burkholderia (Brock et al., 2011;DiSalvo et al., 2015). Cured D. discoideum clones were used in order to isolate the host's effect on Burkholderia and remove any effect that Burkholderia already within D. discoideum may have on the interaction. Each Burkholderia isolate was partnered with the cured version of its native D. discoideum host (i.e., the D. discoideum clone that hosted the Burkholderia isolate when they were isolated from natural soil).
To prepare D. discoideum amoebas for the growth assay, we plated 4 × 10 4 spores from frozen stock on SM/5 with ~2.5 × 10 8 cells of

| Experiment 2: Burkholderia abundance in soil microcosms with and without D. discoideum
We measured the abundance of Burkholderia with and without D. discoideum by inoculating sterilized soil microcosms with D. discoideum spores infected with their native Burkholderia symbiont or with Burkholderia alone (Figure 1d). We tested five isolates of B. agricolaris and four isolates of B. hayleyella (Table 1)

| qPCR assays
Burkholderia agricolaris and B. hayleyella were quantified in experiment 2 using species-specific quantitative PCR (qPCR) assays. We designed primers specific to each Burkholderia species using the genomes of B. agricolaris isolates B317s and B1045 and B. hayleyella isolates B11 and B69 as a guide. We used AlleleID (PREMIER Biosoft) to design primers for genes that were identified as unique to each Burkholderia species Assays were considered specific to one species if spiking with a nontarget sample changed the C q of the target sample by 1.5% or less. We also determined the melting temperature of each assay and included melt curves in every qPCR run to ensure nonspecific products were not amplified.
We experimentally determined the optimal annealing temperature for each assay by running a dilution series on a temperature gradient.
The annealing temperature that produced the best efficiency was used. We made standard curves for each assay by purifying PCR am-

| Statistical analyses
The residuals for experiment 1 were not normally distributed, so we analyzed the data with a Kruskal-Wallis rank sum test with continuity correction using the kruskal.test command in R. We in-

| D ISCUSS I ON
Facultative and environmentally acquired microbes are important components of many organisms� microbiomes, but it is not well understood when it is favorable for microbes to live in hosts when other habitats are available. It has been presumed that microbes benefit from host association, but the fitness of microbes in hosts and in nonhost environments has rarely been measured and compared. Here, we show that D. discoideum provides a fitness benefit to one facultative Burkholderia symbiont species by generating a larger population, while a second symbiont species does not benefit in this way. The fitness differences between these two symbiont species suggest their relationships with D. discoideum fall on different points of the antagonism-mutualism spectrum or reflect different levels of host adaptation.
Dictyostelium discoideum and the Burkholderia symbionts are both facultative partners, indicating they are likely to have a high potential for partner-switching and forming new symbioses. This potential is supported by the high susceptibility of naïve or uninfected D. discoideum to Burkholderia symbionts  and the close proximity of Burkholderia-infected and Burkholderia-uninfected D. discoideum clones in nature .
Partner-switching and infecting new partners require adaptations in the host or the symbiont for finding and contacting potential partners. Previous work has shown that B. agricolaris and B. hayleyella use molecules secreted by D. discoideum to locate and move toward the amoeba hosts for colonization (Shu, Zhang, et al., 2018).
In this study, we found that uninfected D. discoideum clones suppress the growth of B. agricolaris and B. hayleyella in liquid coculture. This is likely due to a secreted molecule as opposed to competition for nutrients, because wild, bactivorous clones of D. discoideum exhibit very little pinocytosis (ingestion of extracellular fluid) in liquid media (Bloomfield & Kay, 2016;Kessin, 2001). However, it seems counterintuitive that Burkholderia symbionts would be attracted to F I G U R E 2 In liquid culture, Burkholderia have lower growth rates in coculture with Dictyostelium discoideum than in monoculture. Maximum specific growth rate is equal to the natural log of 2 divided by the doubling time and is determined from the maximum slope of the growth curve. Circles are replicate growth rate measurements, and lines are the median of the replicates. Points are jittered along the x-axis for visibility F I G U R E 3 In liquid culture, Dictyostelium discoideum lowers the growth rate of Burkholderia agricolaris and B. hayleyella. D. discoideum significantly lowered the growth rate of B. agricolaris and B. hayleyella (Wilcoxon rank sum test, W = 22, p = 2.5 × 10 −10 ), but there was no significant difference between Burkholderia species (Wilcoxon rank sum test, W = 553, p = 0.1294). Data are the same as in Figure 2, pooled by Burkholderia species. Circles are the mean maximum growth rate (n = 3) for each Burkholderia isolate, and each line is the median of all isolates within each Burkholderia species. Points are jittered along the x-axis for visibility a host that can suppress their growth, but two possibilities could account for this scenario. First, the Burkholderia symbionts, especially B. hayleyella (Khojandi et al., 2019), may be largely pathogenic to D. discoideum, so growth inhibition may be an attempt by D. discoideum to resist colonization. Burkholderia symbionts are known to be beneficial only when D. discoideum have dispersed to a location without adequate or high-quality prey, since B. hayleyella and B. agricolaris facilitate transport of other edible bacteria (Brock et al., 2011;DiSalvo et al., 2015;Khojandi et al., 2019). It is unknown how often this occurs, which means it is possible that Burkholderia symbionts are more akin to a pathogen with context-dependent benefits.
Alternatively, growth suppression may be part of the symbiont selection and acquisition process. Many hosts employ a set of harsh, sometimes lethal, conditions that a symbiont must survive in order to colonize (Bright & Bulgheresi, 2010). For example, Vibrio fischeri, the bioluminescent symbiont of the Hawaiian bobtail squid, must travel through an acidic mucus matrix on the surface of the squid into ducts laden with antimicrobials, reactive oxygen species, and immune cells in order to colonize the light organ in the squid's mantle (Schwartzman & Ruby, 2016 Hosts, due to their larger size and lower potential to evolve in response to symbionts, have generally been predicted to be the partner exerting control over the formation and maintenance of mutualisms (Douglas, 2010;Sachs, Mueller, Wilcox, & Bull, 2004), especially in environmentally acquired mutualisms (Kiers, Rousseau, West, & Denison, 2003). For example, some hosts produce symbiont-controlling antimicrobials in their symbiont organs (Login et al., 2011;Mergaert, Kikuchi, Shigenobu, & Nowack, 2017;Park et al., 2018;Wang, Wu, Yang, & Aksoy, 2009), impose sanctions on or expel underperforming symbionts (Baghdasarian & Muscatine, 2000;Kiers et al., 2003;Sachs et al., 2010), or direct symbiont development toward metabolite production at the cost of symbiont reproduction (Kereszt, Mergaert, & Kondorosi, 2011 (Khojandi et al., 2019;Shu, Brock, et al., 2018) suggests a fraction of B. hayleyella cells may be able to leave the host to live in the soil. Understanding whether B. hayleyella benefits from D. discoideum by remaining in the host or by taking host resources and moving into the soil is key to understanding whether the B. hayleyella-D. discoideum relationship is antagonistic or mutualistic.
In this study, we found costs and benefits to symbiont growth, but B. agricolaris and B. hayleyella could benefit from D. discoideum in other ways, such as an increased dispersal capability. D. discoideum forms fruiting bodies at the end of its social lifecycle to aid in its own dispersal (Smith et al., 2014). Both B. agricolaris and B. hayleyella are present in the spores (Khojandi et al., 2019;Shu, Brock, et al., 2018) and are viable after dispersal (Brock et al., 2011;DiSalvo et al., 2015), and so are likely viably dispersed with D. discoideum. D. discoideum spores can potentially be moved over a range of kilometers by animal vectors, which is likely much further than Burkholderia symbiont cells can move on their own. This potential benefit should be further evaluated in the future as it is likely a factor in the co-occurrence and interaction between D. discoideum and Burkholderia symbionts.
Taken together, our results indicate that associating with D. discoideum is costly to B. agricolaris growth, while D. discoideum has both costs and benefits to B. hayleyella growth. It has been hypothesized that host-microbe interactions evolve toward more benefit [at least for the host; (Ishikawa et al., 2016)] and more fitness alignment between the partners, but that does not always seem to be the case (Chong & Moran, 2016;Keeling & McCutcheon, 2017;Minter et al., 2018). The relationship between D. discoideum and the two Burkholderia symbiont species then seems to be evidence of ongoing power struggles between partners. While B. hayleyella benefits more than B. agricolaris from D. discoideum, the opposite is true from the host perspective-D. discoideum benefits more from B. agricolaris than from B. hayleyella (Brock et al., 2011;DiSalvo et al., 2015;Khojandi et al., 2019). It is unclear whether this fitness misalignment/ conflict is due to recency of establishment. Phylogenetic placement suggests B. hayleyella has been evolving in isolation from related Burkholderia for a long time (Brock, Hubert, et al However, the relationship between evolutionary history and the fitness consequences of interactions needs to be further investigated in this symbiosis and in general. Some facultative symbionts, such as rhizobia in terminally differentiated root nodules (Mergaert et al., 2006) and algal cells subjected to kleptoplasty (Pierce & Curtis, 2012), seem to suffer unequivocally from associating with certain hosts. Most facultative symbionts, however, are more complicated. There is plenty of evidence that hosts transfer nutrients to symbionts (Ceh et al., 2013;Denison & Kiers, 2011;Graf & Ruby, 1998;Yellowlees, Rees, & Leggat, 2008), but few studies have examined whether these benefits are parlayed into higher fitness via larger total populations. Of those that have, there has been evidence both for and against increased symbiont fitness from host association.
Associating with Drosophila leads to a larger total population (environmental and host-associated populations combined) for the facultative gut symbiont, Lactobacillus plantarum. However, the "environmental" substrate tested in that study was yeast media, which may limit the conclusion's applicability to natural populations (Storelli et al., 2018).
In one of the most convincing studies of increased symbiont fitness, release of rhizobia from senescing soybean nodules led to populations of rhizobia in the soil five times larger than in fields where plants were removed or in fields that did not include plants (Kuykendall, 1989).
Overall, associating with bobtail squid hosts leads to an increase in Vibrio fischeri symbionts within ocean environments that contain squid (Jones, Maruyama, Ouverney, & Nishiguchi, 2007;Lee & Ruby, 1994), but further investigation has indicated there is strain-level differentiation that has led to a trade-off between growth in the squid host and growth in seawater (Wollenberg & Ruby, 2012). These data suggest that free-living and symbiotic bacteria may actually be two specialized subpopulations of one bacterial species, but it is not clear how widespread this phenomenon is in facultative symbionts (Denison & Kiers, 2004 (Lowe et al., 2016). Overall, there may not be a general rule for host effects on the fitness of facultative symbionts and each interaction will need to be evaluated on a case-by-case basis.
Here, we present evidence that hosts can provide fitness benefits to facultative symbionts. This is consistent with the "reciprocated benefits" hypothesis of interspecies mutualism, where each partner pays a cost to interact with the other, but overall the interaction results in a net benefit to the symbionts. However, we found a growth benefit accrued only to one symbiont species, while the second symbiont species only had costs to growth from host association. In addition, we used growth rate and abundance as fitness measures in this study, but there are other aspects of the interaction, such as the likely increased capacity for dispersal, that could shed more light on this interaction and remain to be tested. These different interactions may be the result of varying degrees of host adaptation, but further work is necessary to fully understand the evolutionary history between D. discoideum and Burkholderia symbionts.

ACK N OWLED G M ENTS
We would like to thank members of the Queller-Strassmann laboratory for comments that improved this manuscript and Michael

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interests. wustl.edu/data/15/).

O RCI D
Justine R. Garcia https://orcid.org/0000-0002-4183-1404 F I G U R E 5 Burkholderia hayleyella has a larger total population size with Dictyostelium discoideum, while B. agricolaris does not. We measured the total abundance of Burkholderia inoculated into soil microcosms as nonsymbiotic cells (Burkholderia treatment) or symbiotic within D. discoideum (Burkholderia + Dictyostelium treatment). We determined the abundance of each Burkholderia species in the entire soil microcosm, which included Burkholderia within D. discoideum and in the soil.

O PE N R E S E A RCH BA D G E S
This article has earned an Open Data Badge for making publicly available the digitally-shareable data necessary to reproduce the reported results. The data is available at https ://opens chola rship. wustl.edu/data/15/