Evolution and maintenance of microbe-mediated protection under occasional pathogen attack

Every host is colonized by a variety of microbes, some of which can protect their hosts from pathogen infection. As microbe-mediated protection can be costly, it is hypothesized, that these costs only pay off in the presence of a pathogen. However, pathogen presence naturally varies over time and can expand once pathogen infection spreads to a new host, as during occasional spill-over from reservoir hosts. We experimentally coevolved populations of Caenorhabditis elegans worm hosts with bacteria possessing protective traits (Enterococcus faecalis), in treatments varying the infection frequency with pathogenic Staphylococcus aureus every host generation, alternating host generations, every fifth host generation or never. We additionally investigated the effect of initial pathogen presence at the formation of the defensive symbiosis. Our results show that enhanced microbe-mediated protection evolved during host-protective microbe coevolution when faced with spill-over infections by a human pathogen occurring rarely over evolutionary history. Initial pathogen presence had no effect on the evolutionary outcome of microbe-mediated protection. We also found that protection was only effective at preventing mortality during the time of pathogen infection. Overall, our results suggest that resident microbes can be a form of transgenerational immunity against occasional pathogen attack or spill-over events.

The net benefits of defensive mutualism are dependent upon the presence of pathogens King and Bonsall, 2017;Lively et al., 2005). Whilst hosts can benefit from microbemediated protection, defensive symbionts are often metabolically and physiologically costly in the absence of enemies (King, 2019). For example, in the interaction of aphids and the bacterium Hamiltonella defensa, the host tissue is harmed by defensive toxins that protect against infection from parasitoids (Vorburger and Gouskov, 2011). In some cases, the costs of possessing the protective microbe might be less energetically costly than investing in the host ś own immune system (Cayetano et al., 2014). From the perspective of the symbiont, it is most useful to its host under high pathogen prevalence, and thus can persist in the host population (Palmer et al., 2008).
Nevertheless, a stable symbiotic interaction is hypothesized to be evolved and maintained (Kwiatkowski and Vorburger, 2012) only when the host benefit of carrying defensive symbionts 4 outweighs any costs. The interactions of obligate symbionts and hosts can be stable for millions of years (Moran et al., 2005).
Not all environments are constantly pathogen rich which might shift the balance of costs and benefits during defensive mutualisms, particularly during coevolutionary interactions (King and Bonsall, 2017). Pathogen prevalence can be spatially (King et al., 2009) or temporally variable, the latter in the case of seasonal epidemics (e.g., flu peaks each winter in the northern hemisphere (Finkelman, 2007). Infection might occur even less often and be asynchronous with host generations due to host range expansion (Phillips et al., 2010) or pathogen spill-over from reservoir hosts (Lo Iacono et al., 2016). Pathogen spill-over can occur between unrelated species and be very harmful, such as zoonotic tuberculosis in humans caused by Mycobacterium bovis (Olea-Popelka et al., 2017). Ejeh et al observed that there were seasonal differences in zoonotic tuberculosis transmission (Ejeh et al., 2013). The impact of other temporally heterogeneous factors on the strength and direction of selection on species interactions have been explored (oxygen concentration (Dey et al., 2016), resource availability Hiltunen et al., 2012), environmental productivity (Harrison et al., 2013)). Whether the varied presence of pathogens can similarly alter selection for symbiotic interactions remains to be directly tested, but has been explored theoretically (Fenton et al., 2011), and in terms of symbiont spread through a host population (Oliver et al., 2014).
Here, we examined the impact of temporal variation in spill-over pathogen infection on the evolution of microbe-mediated protection. We used Caenorhabditis elegans as a worm host and allowed it to be colonised by a bacterium (Enterococcus faecalis) that protects against infection 5 by Staphylococcus aureus (King et al., 2016). Enterococcus faecalis has been shown to be protective across animal microbiomes (Kommineni et al., 2015;Martín-Vivaldi et al., 2010).
Pathogenic S. aureus is novel to C. elegans and is known to expand its range of host species, including different domestic animals, wild hosts, such as rodents, non-human primates and bats (Mrochen et al., 2018;Peton and Le Loir, 2014;Schaumburg et al., 2012). It was previously shown that E. faecalis can evolve to provide enhanced protection when residing in C. elegans hosts during constant pathogen infection (King et al., 2016;Rafaluk-Mohr et al., 2018). From this, we predict that variation in pathogen infection might limit the evolution of microbe-mediated protection. In the present study, we experimentally co-passaged C. elegans with protective E. faecalis and infected the host with evolutionary static pathogenic S. aureus at different intervals of host evolution. We also examined whether pathogen presence at the initial formation of the coevolving interaction is crucial to the evolution of protection. We show that enhanced microbe-mediated protection emerged out of coevolutionary host-microbe interactions and during pathogen infection, regardless of its temporal variability or the time point of first infection. Enhanced protection was only effective during pathogen infection. If hosts survived infection, they could recover and had the same longevity and reproductive output across treatments. These results thus suggest that even occasional pathogen attack or spill-over in the community can select for defensive mutualism, revealing the potential for this phenomenon to be widespread in nature.

Worm host and bacteria system
As a bacteriovore, Caenorhabditis elegans interacts constantly with a variety of bacteria either by feeding or hosting them (Cabreiro and Gems, 2013;Garsin et al., 2001;Schulenburg and Ewbank, 2004). Consequently, C. elegans is an established model for studying innate immunity (Gravato-Nobre and Hodgkin, 2005), as it can be infected with its natural (Jansson, 1994;Schulenburg and Ewbank, 2004) as well as opportunistic pathogens (Garsin et al., 2001;Tan et al., 1999). Most pathogens are taken up orally by the worm (Marsh and May, 2012), and some can proliferate and colonize the worm gut (King et al., 2016;Rafaluk-Mohr et al., 2018).
Naturally, C. elegans is a self-fertilising hermaphrodite (Brenner, 1974), but in this experiment obligate outcrossing worm populations (line EEVD00) with males and females (hermaphrodites that carry the fog-2(q71) mutation) were used (Theologidis et al., 2014). This lineage was generated by Henrique Teotonio (ENS Paris) and encompasses the genetic diversity of 16 natural worm isolates (Theologidis et al., 2014). Worms were kept on Nematode Growth Medium (NGM), inoculated with Escherichia coli OP50 (Brenner, 1974), hereafter referred to as food. Worms were infected with the pathogenic S. aureus (MSSA476) (Holden et al., 2004), which is virulent and kills worm hosts by lysing the intestinal cells lining the gut wall (Sifri et al., 2003). By using this human isolate of S. aureus (Holden et al., 2004), we can mimic pathogenic spill-over from a human background to a nematode infection. Worms were exposed to E. faecalis (OG1RF) (Garsin et al., 2001), which was isolated from the human digestive system, but was previously shown to colonize and proliferate in the worm´s gut (Ford et al., 2017;King et al., 2016;Rafaluk-Mohr et al., 2018), where it provides protection. 7 Experimental evolution -Design Six single clones of E. faecalis (one for each of the six replicate populations) and a single population of C. elegans were the ancestors (hereafter referred to as the Ancestor) for all evolving populations. To account for potential differences in virulence, a stock of four clones of S. aureus was used for pathogen infections. Both C. elegans and colonising E. faecalis were allowed to evolve in presence of each other, while S. aureus was kept evolutionarily static. Infection with S. aureus was varied over host evolutionary time (indicated by purple in Figure 1) to represent temporal heterogeneity in pathogen spill-over, including a range from always to every 2 nd generation, every 5 th generation, and never ( Figure 1). Moreover, we included differences in whether pathogens were present at the formation of the symbiotic interaction or later (2.1. vs. 2.2., and 5.1. vs. 5.2. in Figure 1). Controls for lab adaptation were maintained for the host (No Protective-Microbe control, NPM in Figure 1) and E. faecalis (No Host Control, NHC in    1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18  Experimental evolution -Culturing and passaging methods At the start of each generation, worms were bleached as described previously and left in M9 buffer overnight for larvae to hatch (Stiernagle, 2006). Simultaneously, E. faecalis clones were cultured overnight in Todd-Hewitt Broth (THB) in 600µl at 30ºC, while E. coli OP50 food was cultured overnight in LB broth. Subsequently, 9cm NGM plates inoculated with 300µl of each overnight culture. Plates with freshly inoculated bacteria were dried at room temperature before approximately 1000 L1 worms were added to each NGM plate. After these plates dried at room temperature, they were transferred to a 20ºC incubator and left for 48h. Simultaneously, a liquid culture of S. aureus was grown in THB from frozen stock, while a liquid culture of E. coli OP50 food was grown in LB, and both were incubated under shaking conditions at 30ºC. The following day, 100µl of each overnight culture were spread on 9cm plates, S. aureus on Tryptone Soy Broth agar (TSB) plates and E. coli OP50 food on NGM plates and incubated at 30ºC overnight. To transfer worms to the pathogen or food plates, nematodes were washed off the E. faecalis plates with M9 buffer and washed three times over small-pore filters to remove all externally attached bacteria, as previously described (Jansen et al., 2015;Papkou et al., 2019;Rafaluk-Mohr et al., 2018). Worms were infected with either S. aureus or exposed to food (Figure 1) and left at 25ºC for 24h. Worms were then washed off the plates with M9 buffer once more to plate them on NGM plates seeded with E. coli OP50 food for laying eggs. Roughly 10% of these worms were crushed and plated on E. faecalis selective medium (TSB + 100mg/ml Rifampicin). The remaining worms were left on food plates for 48h to allow for egg laying. To passage E. faecalis, roughly 100 E. faecalis colonies were picked and grown up shaking overnight in 600µl THB at 30ºC, while worms were bleached and left to hatch overnight. This cycle was repeated for 20 experimental host generations. 9 All passaged worms and E. faecalis samples were cryopreserved at -80 ºC. A proportion of the offspring of surviving worms were frozen in 40% DMSO, and 100µl of E. faecalis liquid culture was mixed with 100µl of glycerol before cryopreservation.
Host survival and fecundity assays All assays were conducted at the end of the evolution experiment on archived samples. Plates were randomized and fully encoded during each experiment to ensure the experimenter was blind to different treatments whilst collecting data. Basic procedures were adopted from the experimental evolution, but with the following alterations to keep the assays feasible with higher accuracy when scoring dead and alive worms: 400 L1 worms were exposed to 200µl of E. coli OP50 food and E. faecalis on 6cm NGM plates, while only 60µl S. aureus overnight culture was used to inoculate 6cm TSB plates.
To assess microbe-mediated protection of different combinations of worms and E. faecalis, 400 L1s were exposed to 50:50 mixtures of E. faecalis and E. coli OP50 food for 48h. Worms were then washed off these plates as described above and infected with S. aureus for 24h at 25°C. Survival in form of counting dead and alive worms was then scored.
To assess any long-term fitness consequences after protective microbe exposure and pathogen infection, long-term survival and fecundity were measured. Worms were exposed as described for the survival assays. Subsequently, five females and five males were picked onto 3cm E. coli OP50 food seeded NGM plates at 25°C and then transferred to new plates every 36h to avoid any confusion between offspring produced and original adults. At each time point, survival was scored. 10 To measure fecundity, the number of worm hosts eggs on the plates at 120h since bleaching were counted.

Statistical Analysis
Statistical analyses were carried out with RStudio (Version 1.1.463 for Mac), graphs created with the ggplot2 package (Version 2.1.0) and edited with Inkscape (Version 0.91). Host survival to E. coli OP50 food and ancestral E. faecalis and S. aureus were analysed with Wilcoxon Rank Tests with False Discovery Rate (FDR) correction for multiple testing. Host survival data across evolutionary treatments were analysed with nested binomial mixed effects models (R package lme4), followed by a Tukey multiple-comparison tests (R package multcomp). Life-span data were analysed with Kaplan Meier Log Rank test with FDR correction for multiple testing, and host fecundity was analysed with a nested linear mixed model (R package lme4). Filled symbols indicate those treatments being exposed to food in the later stage, while open symbols indicate those treatments being exposed to the pathogen S. aureus in the later stage.

Results
Confirming previous results, E. faecalis showed some host-protective potential against S. aureus.
Survival of worms raised on E. faecalis/food and then infected with S. aureus was greater than worms raised on food alone and then infected with S. aureus (Wilcoxon Rank Test, p=0.03; Figure   2, open symbols). These results demonstrate the protective effect of E. faecalis. Moreover, those worms raised on E. faecalis/food survived similarly to those worms raised on food alone   Figure 3A). Higher microbe-mediated protection occurred in all evolutionary histories involving infection at some point in the evolution experiment (Always, 2.1. and 5.1.). Host evolutionary history had a significant effect on host survival (Mixed Effects Model, X 2 =40.458, df=5, p<0.001; Figure 3B), but this difference was independent of the pathogen or protective 13 microbe, as it is also demonstrated in the absence of pathogen infection or the protective microbe (NPM control). These results suggest that worms have adapted under the procedure of the evolution experiment. No effect of bacteria evolutionary history alone on infected host survival was observed (Mixed Effects Model, X 2 =2.3891, df=5, p=0.7931; Figure 3C). Taken together, enhanced microbe-mediated protection evolved only as a product of coevolution and pathogen presence, the latter regardless of its heterogeneity.
As an additional form of pathogen heterogeneity, the impact of the timing of initial pathogen infection on the evolution of microbe-mediated protection was explored.  Figure S1).   Figure 4A). In addition, we did not find significant differences in fecundity among worm-E. faecalis pairs (Mixed Effects Model, X 2 =3.9163, df=4, p=0.4175, Figure 4B).

Discussion
It has been shown that hosts receive the greatest benefits from protective microbes under constant pathogen infection. We hypothesized that variation in novel pathogen presence over time would limit the evolution of microbe-mediated protection due to the reduced benefits to the host and bacterial symbiont. In our study, enhanced pathogen defence emerged out of host-symbiont coevolutionary interactions only when pathogens were present, independent of the interval of pathogen presence. Notably, the ultimate strength of microbe-mediated protection that evolved was not impacted by the number of host generations between pathogen infections, the proportion of generations infected, or the presence of the pathogen at the first host-microbe interaction. These results suggest that resident microbes can be a form of transgenerational immunity against occasional pathogen attack or spill-over events, even by novel pathogens.
We found that microbe-mediated protection is maintained even in the prolonged absence of pathogen, but that pathogen presence is necessary for microbe-mediated protection to evolve, as previously hypothesized by King & Bonsall (King and Bonsall, 2017). This result is unlike previous work showing that the scale of heterogeneity in abiotic conditions can affect the strength of selection for traits in some symbiotic interactions (Harrison et al., 2013). This discrepancy is potentially due to costs in our symbiotic system being ameliorated (at least in terms of host survival) in well-provisioned hosts, as hosts are provided with food alongside E. faecalis and are thus rescued from starvation (also see (Dasgupta et al., 2019)). Although protective symbionts can incur costs (e.g., Vorburger & Gouskov, 2011) for their hosts, with potential for impacts on coevolutionary interactions (King and Bonsall, 2017), it is possible that potential costs of bacterial colonisation might be only detectable when hosts are stressed (Lively, 2006). Higher protection 16 also does not always come with higher costs, as found in the black bean aphid-Hamiltonella defensa interaction (Cayetano et al., 2014). Thus, protective traits in an organism's commensal microbiota could be selected for under pathogen infection and easily maintained in subsequent uninfected generations.
Microbe-mediated protection was strongest between sympatric pairs when pathogens were present over evolutionary time, consistent with previous findings (Rafaluk-Mohr et al., 2018). In our study, protection emerged during coevolution after only 20 host generations, and not due to the independent evolution of either interacting species, but due to the coevolution of both species, which was already hypothesized by mathematical models in 2017 (King and Bonsall, 2017). The time-scale of these interactions is short compared to the longer shared evolutionary histories shared by other defensive mutualisms (Jousselin et al., 2003;Quek et al., 2004;Shoemaker et al., 2002).
Nevertheless, our findings reveal the potential for microbe-mediated protection to become enhanced during the formation of a coevolving host-microbiota relationship.
In conclusion, our results show that enhanced protection in host-microbe interactions can rapidly evolve and be maintained even under infrequent pathogen attack, suggesting that resident microbes can be a form of stable, transgenerational immunity. The protective benefit of an organism's microbiota might remain undetected for several host generations until pathogens re-emerge. Future research on the failure of pathogens to spill-over or to transmit within host populations should consider the contribution of the protective microbiota to prevent disease spread.