Post‐association barrier to host switching maintained despite strong selection in a novel mutualism

Abstract Following a host shift, repeated co‐passaging of a mutualistic pair is expected to increase fitness over time in one or both species. Without adaptation, a novel association may be evolutionarily short‐lived as it is likely to be outcompeted by native pairings. Here, we test whether experimental evolution can rescue a low‐fitness novel pairing between two sympatric species of Steinernema nematodes and their symbiotic Xenorhabdus bacteria. Despite low mean fitness in the novel association, considerable variation in nematode reproduction was observed across replicate populations. We selected the most productive infections, co‐passaging this novel mutualism nine times to determine whether selection could improve the fitness of either or both partners. We found that neither partner showed increased fitness over time. Our results suggest that the variation in association success was not heritable and that mutational input was insufficient to allow evolution to facilitate this host shift. Thus, post‐association costs of host switching may represent a formidable barrier to novel partnerships among sympatric mutualists.


| INTRODUC TI ON
Specialization, which is frequently observed in mutualistic interactions (Chomicki et al., 2020), can make host switching costly. These costs can manifest as lower growth, fecundity, or survival for hosts partnered with non-native symbionts (Chapuis et al., 2009;Ehinger et al., 2014;Parker, 1995;Sicard et al., 2005). Analogously, symbionts may be less competent in colonizing a novel host, and thus, less likely to be transmitted across hosts (Kwong et al., 2014;Sicard et al., 2005). When costs are severe, the partners must rapidly adapt in order for a new mutualism to be successful. Serial copassaging studies have shown that adaptation can occur very rapidly (Batstone et al., 2020;Robinson et al., 2018;Shapiro & Turner, 2018;Soto et al., 2012). However, even with strong selection, evolution could be constrained by low standing genetic variation or the relative strength of genetic drift in small populations (e.g., Castillo & Delph, 2016;Hoang et al., 2016;White et al., 2021). Additionally, if changes across multiple loci are required to adapt to a new host, increases in fitness may take longer to arise (Streicker et al., 2012), and may limit the evolutionary success of a new partnership in nature.
Here, we examine the response to strong selection following an experimental host shift between sympatric isolates of the Steinernema nematode-Xenorhabdus bacteria mutualism.
In the entomopathogenic mutualism between Steinernema nematodes and Xenorhabdus bacteria, nematodes transmit bacteria between insects, and bacteria help kill and digest the insect.
The bacteria also help to reduce competition by producing toxins against non-native nematodes and other co-infecting microbes (Bashey et al., 2013;Murfin et al., 2018). In a previous study, we paired nematodes with symbiotic bacteria isolated from other sympatric nematode species . This experiment focused on a single bacteria species, X. bovienii, isolated from each of three Steinernema nematode species. Symbionts faced no barriers to host switching within nematode species or between closely related nematode species. However, we observed strong barriers to host switching when nematode species were more distantly related. Specifically, S. kraussei paired with any strain of X. bovienii isolated from clade III nematodes (S. kraussei or S. texanum) showed no significant differences in fitness relative to the native pairing, while pairings across nematode clades were reciprocally unsuccessful. In fact, only the pairing used in this study (out of 10 attempted) had any infection success once associated . The newly associated partners were able to successfully infect insects and reproduce, albeit at a reduced probability and lower fecundity. This post-association barrier to host switching is predicted to limit the spread of this novel combination in nature.
Thus, although this host switch is possible, the post-association barriers result in partner fidelity feedback favoring the native pairing (Murfin et al., 2015;Sachs et al., 2004). Nevertheless, rapid evolution could rescue this otherwise unhopeful pairing allowing for a host shift.
In this study, we tested whether evolution could facilitate a host shift by experimentally passaging the novel pairing to see if it could respond to strong selection. In each of the nine passages through insects, we selected the most fecund of the novel pairings to propagate the mutualism. We assessed the fitness of the mutualism with three metrics: the proportion of successful infections, the mean number of nematodes emerging from successful infections, and the mean number of bacterial cells carried per nematode. We predicted that if the observed variation in fitness was heritable, the novel combination should evolve, exhibiting increased fitness across subsequent passages as the partners adapt to each other. Instead, we found that none of our metrics of fitness improved over time.
We suggest that low genetic variation, and repeated bottlenecks in the mutualism, may constrain the short-term response to selection.

| Pairings
We experimentally paired aposymbiotic (lacking symbionts) Steinernema kraussei nematodes with Xenorhabdus bovienii bacteria cultured from S. affine nematodes (Table 1)   . Here, we examine whether infection success and nematode emergence would respond to strong selection from repeated co-passaging. The experimental pairing was compared to the native nematode control group, S. kraussei, and the native bacteria control group, S. affine (Table 1). Both control groups carried their native bacteria, having never been cultured separately.

| Statistical analyses
All statistical analyses were performed using generalized linear models in R version 3.6.3 (R Core Team, 2020). We used the dplyr package to organize the data, and the ggplot2 package for all graphs (Wickham, 2016;Wickham et al., 2020) (Lenth, 2020) and the F statistics for each independent variable using the rstatix package (Kassambara, 2020). Note that each pairing was represented by one evolutionary lineage, so confidence intervals reflect variation across insect hosts. Mean trait values for F I G U R E 1 Schematic of the passaging protocol and fitness components measured for each nematode-bacteria pairing. Starting in passage 5 to lessen the effort needed to maintain the experimental lines, and to increase the selection on successful pairings, infection mixes were created by combining nematodes from the three best infections based on a visual inspection of the quality and abundance of emerging nematodes. In addition, only infection success was measured each pairing encompass both genetic and environmental changes with time. We focused on the interaction between pairing and passage in order to test whether the experimental pairing shows any evidence for adaptation in response to co-passaging. This approach uses the nematode control pairing as a benchmark for a well-adapted mutualism.

| Infection success
The experimentally associated pairing had significantly lower infection success than the native nematode control group (Figure 2a;

| DISCUSS ION
Post-association barriers to host switching can limit the reproductive success of novel pairings, especially when they face competition with native pairs. If partner fidelity affords significant fitness advantages, then novel pairings may be possible, but of little consequence (Murfin et al., 2015;Sachs et al., 2004). Evolution, however, has the potential to rescue (Bell, 2017) low fitness pairings. Here, we tested whether a post-association barrier to host switching could be overcome by strong selection. We expected that one or both partners would adapt, leading to increasing fitness over successive passages.
We found that none of the fitness measures exhibited a significant interaction between the mutualism pairing and passage number, indicating that the experimental pair did not change relative to the native pair(s) across the successive passages. Thus, post-association barriers were maintained despite repeated co-passaging. The lack of response to selection in our experiment suggests that host shifts in this system may be limited by a lack of additive genetic variation.
Because we passaged only the most fecund infections, and because unsuccessful infections have a fitness of zero for both partners, this experiment imposed very strong selection on both the nematode and on the bacteria. Despite the strength of selection, the experimental pair did not improve over the passages, suggesting low genetic variation in the nematode and bacteria. Large population sizes within the insect (nematodes >10 4 , bacteria >10 6 ) allow for mutational input to occur in both species. Indeed, previous studies have shown that some Xenorhabdus traits (e.g., growth rate and bacteriocin production) can evolve when passaged with nematodes in the lab (Bhattacharya et al., 2019;Morran et al., 2016). However, it is possible that adaptation to new partners requires multiple mutations, which would constrain the speed of evolutionary rescue.
Several experiments indicate that expanding host ranges relies on the accumulation of multiple mutations, which is less likely, and requires more time, than a single mutation (Hall et al., 2011;Longdon et al., 2014;Meyer et al., 2012;Quides et al., 2021;Soto et al., 2019;Streicker et al., 2012;Woolhouse et al., 2005 (Bashey & Lively, 2009;Stuart & Gaugler, 1996). In terms of the bacterial populations, we found that bacterial carriage in the experimental pairing was one-tenth that of the native pairs (Figure 2c), which would further limit the evolutionary potential of the novel mutualism. Thus, repeated, severe bottlenecking likely lowered the chance that a beneficial mutant was transferred to the next infection. Without new genetic input, variation in infection outcome would depend mostly on environmental variation across caterpillars. This could be due to demographic stochasticity affecting bacterial carriage and nematode survival within each insect host, as well as differences in the insect's nutrient composition, microbiome, and immune response.
So, does the maintenance of the post-association barriers we observed indicate that host switching is unlikely to be facilitated by evolutionary rescue? Not necessarily. Here, we attempted to evolve one lineage, but perhaps if we had passaged multiple lineages, one of the incipient pairings might have had the right mutation. Alternatively, a different source stock might have been more successful. However, the X. bovienii stocks isolated from S. affine are highly genetically similar and equally distant from the X. bovienii isolated from S. kraussei, so it is not obvious that another S. affine associate strain would be more likely to successfully shift hosts.