Manifold aspects of specificity in a nematode–bacterium mutualism

Authors


Jean-Baptiste Ferdy, Institut des Sciences de l’Évolution de Montpellier, UMR CNRS 5554, Université Montpellier 2, F-34095 Montpellier Cedex 05, France.
Tel.: +33 4 67 14 42 27; fax: +33 4 67 14 36 22; e-mail: ferdy@univ-montp2.fr

Abstract

Coevolution in mutualistic symbiosis can yield, because the interacting partners share common interests, to coadaptation: hosts perform better when associated with symbionts of their own locality than with others coming from more distant places. However, as the two partners of a symbiosis might also experience conflicts over part of their life cycle, coadaptation might not occur for all life-history traits. We investigated this issue in symbiotic systems where nematodes (Steinernema) and bacteria (Xenorhabdus) reproduce in insects they have both contributed to kill. Newborn infective juveniles (IJs) that carry bacteria in their intestine then disperse from the insect cadaver in search of a new host to infect. We ran experiments where nematodes coinfect insects with bacteria that differ from their native symbiont. In both Steinernema carpocapsae/Xenorhabdus nematophila and Steinernema feltiae/Xenorhabdus bovienii symbioses, we detected an overall specificity which favours the hypothesis of a fine-tuned co-adaptation process. However, we also found that the life-history traits involved in specificity strongly differ between the two model systems: when associated with strains that differ too much from their native symbionts, S. carpocapsae has low parasitic success, whereas S. feltiae has low survival in dispersal stage.

Introduction

Partners of symbiotic associations do coevolve because they interact for many important aspects of their life cycles. In pathogenic or parasitic associations, this coevolution can lead to increased host's defences and to new ways to by-pass those defences in parasites. In mutualistic associations, members of which share common interests, a simple prediction is that coevolution should favour the combinations of traits that are the most beneficial to both partners. Of course, this scenario cannot be taken as a general rule (e.g. Kaltz & Shykoff, 1998; Parker, 1999; Lively et al., 2004; Morgan et al., 2005), but if it holds we expect the benefits the two partners get from their interaction to be greater when they have coevolved for a long time than when they have recently met. In other words, under this simple scenario coevolution should, in mutualistic symbiosis, yield coadaptation.

Coadaptation in mutualistic symbiosis often comes under the form of very specific molecular recognition between the two interacting partners. This form of specificity in association is now well described in the case of the interaction between Rhizobia and legumes (Stacey et al., 2006), in the association between Vibrio bacteria and squids (Nyholm & McFall-Ngai, 2004) or between entomopathogenic nematodes and their bacterial symbionts (Cowles & Goodrich-Blair, 2008). What remains unclear, in all these systems, is whether or not these very specific recognition mechanisms correlate with coadaptation in other aspects of the interaction. For example, if a symbiont is not well recognized by the host, will it necessarily be a poor symbiotic partner if by chance it succeeds to associate? One prediction would be that hosts should specifically exclude symbionts with low performance. We then expect that symbionts that are not recognized are also those that perform poorly. Another prediction is that the two partners, although they share common interests, might be in conflict at least for some part of their life cycle. Coadaptation could then be detectable on some but not all life-history traits. Recognition and performances are determined by different sets of genes which can potentially be horizontally transferred between symbiotic lineages. They can therefore experience different selective pressures and the different possible sides of coadaptation might not perfectly match.

We can test part of these alternative predictions by depriving hosts from their symbionts, and combining the resulting aposymbiotic individuals with non-native partners. This approach requires that the symbiosis is not obligate, so that aposymbiotic individuals can be easily produced at least in experimental conditions. Once chimeric associations are obtained, their performances need to be measured and related to the phylogenetic distance of the experimentally associated symbiotic strain to the native symbiont. In fact, this approach does not really allows to test for coadaptation per se, as a single host strain is experimentally associated with many different symbiotic lineages, but the reciprocal recombinations are not performed. The poor performances of a symbiotic lineage in a given host might be interpreted in two different ways: (i) it might indicate that this lineage is maladapted to this host, but (ii) it might also mean that this symbiont provides few benefits whatever host it is associated with. In the following we will therefore use the term specificity, instead of coadaptation, which we think does not make any hypothesis about which process created the pattern we observe.

In this study we used Steinernema entomopathogenic nematodes symbiotically associated with Xenorhabdus bacteria as an experimental system. These symbioses are convenient experimental models because they have short life cycle, they are easy to maintain in laboratory conditions, the association is facultative for the nematodes and, thus, aposymbiotic individuals can be experimentally obtained (Sicard et al., 2003). The life cycle of Steinernema comprises two main stages: a free stage in the soil, where infective juveniles (IJs) carry their bacterial symbionts in their intestine (Bird & Akhurst, 1983), and a parasitic stage in the insect, where nematodes and bacteria multiply in parallel (see Fig. 1 for details). During the parasitic stage, nematodes benefit from the toxic activities of their bacteria towards insects and from the bio-conversion of insect cadaver into easily assimilable nutrients (Forst & Nealson, 1996). During the free stage, bacteria, which are unable to survive in the soil (Morgan et al., 1997), benefit from protection and transportation from host to host by the IJs. Nematode–bacterium symbioses thus match the classic definition of mutualism. However, recent works conducted on the Steinernema carpocapsae/Xenorhabdus nematophila symbiosis also demonstrated that bacteria are also costly to carry for nematodes during the free stage of their life cycle: IJs that carry bacteria survive less than those that do not (Mitani et al., 2004; Emelianoff et al., 2007). Both costs in free stage and benefits in parasitic stage correlate with the amount of symbionts nematodes carry, leading to a survival–reproduction trade-off in symbiotic nematodes (Emelianoff et al., 2008a).

Figure 1.

 Life cycle of SteinernemaXenorhabdus symbioses. Infective juveniles (IJs) are soil-dwelling third-stage larvae whose development has been arrested; they are nonfeeding and carry their bacterial symbionts in their intestine (Bird & Akhurst, 1983). After entering an insect host by natural openings, nematodes release their bacterial symbionts into the hemolymph. Both partners of the symbiosis secrete toxins against insect's immune system and the bacteria multiply, provoking septicaemia (Goodrich-Blair, 2007); all these mechanisms contribute to insect death within 48 h. Nematodes then feed on bacteria and metabolized insect tissues and reproduce into the dead insect for two or three generations. Newly formed IJs recruit one or two bacterial clone(s) in their intestine, leave the insect cadaver and disperse in the soil.

The costs and benefits of the symbiotic association between Steinernema and Xenorhabdus can vary from one stage of their life cycle to another. Similarly, the mechanisms and levels of specificity in the interaction between Steinernema and Xenorhabdus can vary along the symbioses life cycle. Indeed, two sides of specificity have already been described in Steinernema/Xenorhabdus symbioses (Sicard et al., 2004, 2005). The first type concerns bacterium–IJ re-association before dispersing from infected insect cadavers. As mentioned above, re-association implies that bacteria and nematodes can recognize each other and that bacteria colonize nematodes’ intestine. Both steps are involved in re-association specificity. The second type of specificity has been observed during the exploitation of the insect host: some bacterial traits involved in killing the insect or metabolizing its cadaver seem to be adapted to a single nematode strain. Clearly, other sides of specificity might be detectable in other aspects of the interaction between nematodes and their bacterial symbionts. This might be the case, for example, for traits involved in IJs survival during the free stage of the life cycle. Such sort of specificity has never been surveyed so far.

In this work, we study specificity patterns in two nematode species, S. carpocapsae and Steinernema feltiae, which have been shown to have different association modalities (Emelianoff et al., 2008b). Indeed, S. carpocapsae is associated with numerous bacteria cells and highly benefits from the symbiosis in terms of reproduction (Sicard et al., 2003). Moreover, its bacterial partner, X. nematophila, was only found in S. carpocapsae nematodes (Tailliez et al., 2006) and S. carpocapsae has been found not to be capable of carrying bacteria other than X. nematophila (Sicard et al., 2004). Conversely, S. feltiae is associated with less bacteria cells (Emelianoff et al., 2008b) and Xenorhabdus bovienii, its bacterial symbiont, is found in association with numerous other nematode species (Akhurst & Boemare, 1988; Mráček et al., 2006; Tailliez et al., 2006; Emelianoff et al., 2008b). Because of these differences, we suspect the two symbioses to display different specificity patterns.

In order to measure specificity in all of its aspects, we coinfected insect hosts with aposymbiotic IJs from one nematode species and one Xenorhabdus strain. This strain could be the native symbiotic strain, another strain from the same bacterial species than the native one, or a strain that belongs to another Xenorhabdus species. Different components of fitness (emergence rate, reproduction rate, death rate during the free stage and the number of bacteria carried) were measured on each of these experimental coinfections. We then quantified specificity by relating these traits to the phylogenetic distance between the non-native bacterial strain and the native symbiont.

Materials and methods

Biological material

In this work we used two nematode species: S. carpocapsae strain SK27 (Plougastel, Finistère, France), naturally associated with X. nematophila strain F1, and S. feltiae strain VIN (Aigues-Mortes, Gard, France), symbiotically associated with X. bovienii strain FR45. These nematode and bacterial species have been characterized on both phenotypical and molecular traits (Tailliez et al., 2006; Emelianoff et al., 2008b).

As they involve species that have contrasted ecological characteristics (Stuart et al., 2006), we expect differences in specificity between the two symbioses we study here. Part of these differences, though, might be due to the different histories of the strains we are using. The strain SK27 of S. carpocapsae has been kept in the laboratory for 25 years, by successive infestations of last instar Galleria mellonella larvae (Lepidoptera, Pyralidae). This strain is therefore well adapted to laboratory conditions. Conversely, strain VIN of S. feltiae was isolated from field in 2006 (Emelianoff et al., 2008b) and has therefore experienced a few cycles only of laboratory infestations.

For both species, aposymbiotic IJs were obtained by experimentally depriving symbiotic IJs of their bacteria as explained in Sicard et al. (2003). Aposymbiotic IJs from both species were kept alive for 3 months at 8 °C in the dark. Symbiotic bacteria of each nematode species were separately stored in 15% glycerol at −80 °C. In the following experiments, we used G. mellonella as an experimental insect host because it is highly permissive to entomopathogenic nematodes. This guarantees the fact that all bacterial strains will be capable of multiplying in the insect and that nematodes will be capable of reproducing, even without any symbiotic bacteria (Sicard et al., 2003).

Experimental design

After having produced aposymbiotic IJs for both S. carpocapsae and S. feltiae, we experimentally coinfected insects with these nematodes and various strains of Xenorhabdus. Steinernema feltiae has been combined with 14 X. bovienii strains (including FR45, the native symbiont of VIN nematode strain) and 14 strains from other Xenorhabdus species. Steinernema carpocapsae has been combined with 15 X. nematophila strains (including F1, the native symbiont of SK27 nematode strain) and 18 strains from other Xenorhabdus species (see Table 1). We thus have two native and 59 non-native experimental coinfections.

Table 1.   Bacterial strains of Xenorhabdus species used in experimental coinfections with each species of Steinernema.
Bacterial species and strainNative nematode partnerBacteria 16s rDNA GenBank accession no.Geographical originParasitic success of Steinernema carpocapsae at J28Number of bacteria per IJ in S. carpocapsaeParasitic success of Steinernema feltiae at J28Number of bacteria per IJ in S. feltiae
  1. For each experimental coinfection, the number of successful infections after 28 days over the total number of performed infections is given. When possible, i.e. when the infection was successful and produced enough infective juveniles (IJs), the average number of bacteria per IJ is also given.

X. bovienii FR45S. feltiaeEU190978France (Aigues-Mortes)0/50 44/5019.80
X. bovienii FR44S. feltiaeEU190978France (St Gély)  15/5016.13
X. bovienii FR10S. feltiaeAY521240France (Brest)  15/503.87
X. bovienii USAR01S. oregonenseFJ860885USA (Arizona)  40/508.13
X. bovienii CA04S. krausseiDQ205454Canada  0/50 
X. bovienii CS66S. krausseiDQ205451Czech Republic  15/481.37
X. bovienii USNY95S. krausseiDQ205453USA (New York)  49/490.07
X. bovienii CS03S. weiseriDQ205452Czech Republic  33/509.63
X. bovienii SiS. intermediumDQ205455Australia0/50 1/50 
X. bovienii F3S. affineDQ202311France  17/500.00
X. bovienii FR43S. affineEU190976France (Lansargues)  8/500.00
X. bovienii TB20S. sichuanenseDQ208305China (Tibet)0/50 2/50 
X. bovienii SA02S. ichnusaeEF219400Sardinia  15/492.75
X. bovienii T228S. feltiaeD78007Australia  48/504.91
X. nematophila F1S. carpocapsaeAY521241France77/10050.96  
X. nematophila ATCC19061S. carpocapsaeD78009USA (California)88/10029.8450/500.29
X. nematophila PL31S. carpocapsaeAY521242Poland38/500.99  
X. nematophila BE06S. carpocapsaeDQ211704Belgium29/500.34  
X. nematophila ES96S. carpocapsaeDQ211707Spain44/509.89  
X. nematophila ES98S. carpocapsaeFJ860888Spain41/5035.09  
X. nematophila A20S. carpocapsaeFJ860886Portugal40/5054.28  
X. nematophila A24S. carpocapsaeFJ860883Russia42/5043.05  
X. nematophila CBYS. carpocapsaeFJ860887China44/5041.60  
X. nematophila AN6S. carpocapsaeAY278674USA43/5058.23  
X. nematophila K97S. carpocapsaeFJ860884USA (California)39/5021.41  
X. nematophila DD136S. carpocapsaeFJ860889USA38/5018.04  
X. nematophila NC116S. carpocapsaeFJ860890USA (N Carolina)40/5034.55  
X. nematophila K102S. carpocapsaeFJ860891Mexico44/5079.56  
X. nematophila CA01S. carpocapsaeDQ211705Canada34/5051.91  
X. innexi DSM16336S. scapterisciAJ810292Uruguay0/50 5/500.04
X. stockiae TH01S. siamkayaiDQ202309Thailand1/50 31/500.11
X. hominicki KR01S. monticolumDQ205448Korea0/50   
X. hominicki KE01S. kariiDQ211719Kenya0/50 44/493.25
X. indica DSM17382S. thermophilumAM040494India0/50   
X. cabanillasii USTX62S. riobraveAY521244USA  46/500.00
X. budapestensis DSM16342S. bicornutumAJ810293Serbia0/50 22/500.62
X. szentirmaii DSM16338S. rarumAJ810295Argentina0/50 41/500.00
X. japonicus DSM16522S. kushidaiDQ202310Japan16/490.0012/500.00
X. griffiniae ID10S. hermaphroditumDQ211710Indonesia4/500.0542/500.00
X. ehlersii DSM16337S. longicaudumAJ810294China27/500.0040/500.00
X. poinarii CU01S. cubanumDQ211706Cuba2/50 31/500.00
X. poinarii G6S. glaseriD78010USA20/480.00  
X. kozodoii SaVS. arenariumDQ211716Russia32/500.0041/500.00
X. romanii PR06-AS. puertoricenseDQ211717Puerto Rico38/500.0028/500.00
X. doucetiae FRM16S. diaprepesiDQ211709Martinique1/50 38/500.00

For each experimental coinfection, which involves a given nematode species and a given bacterial strain, we infected 50 insects, collected emerging nematodes and measured various symbiotic life-history traits to infer overall specificity (both on free and parasitic stages). We also measured the capacity of non-native bacteria to re-associate with nematodes. Details on the methods used to perform coinfections and to estimate symbiotic life-history traits are given in the following sections.

Aposymbiotic nematodes of both species have served as a negative control in these experiments. For example, bacterial strains that reduce nematodes’ parasitic success below that of aposymbiotic nematodes can be considered as harmful. Similarly, aposymbiotic nematodes associated with the native bacterial symbiont will be used as a positive control. For example, bacterial strains that increase reproduction rate above that of aposymbiotic but below that of experimental coinfection with the native symbiont will be considered as beneficial but maladapted.

Experimental coinfections

We used an infection device consisting in a 1.5-mL microtube whose bottom has been cut off, encased in another intact microtube. The upper microtube contains a cone of filter paper which goes through the hole into the lower microtube. To realize a coinfection, aposymbiotic IJs were first deposited on the paper in the upper microtube. We used eight IJs for S. carpocapsae and 20 IJs for S. feltiae to take into account their different abilities to infect, following results of a pilot study. A G. mellonella larva was then introduced in each microtube. The nematodes and the insect were then incubated together at 24 °C for 24 h. After these 24 h, we injected 20 μL of a bacterial suspension prepared following Sicard et al. (2004) into the G. mellonella larvae. The insect was thus, at that time of the experiment, infected by both nematodes and bacteria. Preliminary studies have shown that the number of bacteria per IJ was much lower in S. feltiae (2–10 bacteria per IJ) than in S. carpocapsae (50–200 bacteria per IJ, Emelianoff et al., 2008a). In order to take this difference into account, we injected 1000 bacteria in each insect infected with eight S. carpocapsae IJs and 40 bacteria in each insect infected with 20 S. feltiae.

The number of bacteria injected was controlled a posteriori by streaking appropriate volumes of the bacterial suspension injected onto NBTA plates (31 g nutrient agar, 2.5 mg bromothymol blue, 1 L sterile H20) supplemented with 0.0025‰ (w/v) triphenyl tetrazolium chloride and counting colony forming units (CFU) after 48-h incubation at 28 °C.

Insects coinfected by nematodes and bacteria were then incubated at 24 °C for 24–48 h. After this incubation period, we added 1 mL of water in each lower microtube of the device, so that IJs newly emerged from the insect cadaver can migrate into the water. Emergence could thus easily be assessed by looking for nematodes in this small volume of water. ‘Each tube trap’ was observed several times after infection to check for the emergence of IJs. Newly emerged IJs were collected regularly to limit the time they spent at 24 °C in the trap, which could affect further measurements. All the nematodes collected from a single insect were gathered in a single 50-mL centrifugation tube and kept at 8 °C.

Estimation of phylogenetic distances

16S bacterial DNA sequences were extracted from GenBank database and aligned using BioEdit software (Ibis Biosciences, Carlsbad, CA, USA). A phylogenetic tree was then computed with RAxML software using a Gamma model with GTR substitution matrix (Stamatakis, 2006). Genetic distances were then calculated from the obtained tree. We will hereafter call bacteria distance the phylogenetic distance between a bacterial strain and the native strain of the studied nematode (FR45 for S. feltiae, F1 for S. carpocapsae).

Using RAxML, we also bootstrapped sequence data to compute confidence intervals around our estimates of bacteria distance. These calculations will not be presented here, but we verified that uncertainties in our estimates of phylogenetic distance do not affect significantly our conclusions.

Measurements of symbiotic life-history traits

In the following sections, we detail how life-history traits were measured. Emergence rate (rate at which nematode IJs emerge from coinfected insects), reproduction rate (number of IJs produced by a successful infection) and number of carried bacteria per IJ and IJs’ death rate during the free stage were estimated for each combination by averaging values obtained on replicate coinfections. The details on how these estimates were obtained are given in the following three sections. Using these per experimental coinfection estimates, we then studied how symbiotic life-history traits relate to the phylogenetic position of the bacterial strain. This was performed for each nematode species by using simple nonparametric correlations between the life-history trait and the bacteria distance described above. Comparison between the correlations observed for the two symbioses was then performed using a permutation technique. All these analyses were performed using R statistical software (R Development Core Team, 2008).

Nematodes emergence and reproduction rates

For each experimental coinfection, parasitic success is the proportion of successful infections, i.e. infections leading to emergence of 10 IJs at least. It was measured after 10, 12, 14, 21 and 28 days at 24 °C. For each successfully infected insect, data can be summarized as the day at which emerging nematodes were first observed. This sort of data can be analysed using survival models. Using such models, we can estimate an emergence rate which quantifies both how frequently nematodes succeed in emerging and how fast they do emerge. We used a parametric survival regression with a Weibull distribution for the time it takes after infestation for nematodes to emerge. The Weibull distribution allows emergence rate to vary over time. This is important in our situation because we expect, for a given coinfection, insects that are the least resistant to nematode infection to die first. This yields a decrease in emergence rate over time. We obtained estimates of this emergence rate by fitting such a model for each experimental coinfection.

In order to evaluate the reproduction rate, three successful infections on average were randomly chosen for each experimental coinfection. After all nematodes have emerged from these infections, the total number of IJs was estimated by counting IJs under a binocular microscope on a grid drawn on a Petri dish. The mean of the three samples was taken as a measurement of the mean reproduction rate of the experimental coinfection.

Re-association of nematodes and bacteria

In order to test whether IJs are colonized by non-native bacteria, and to quantify to which extent they are, 100 of newly emerged IJs were taken from the same three infections used to measure reproduction rate and mortality. These IJs were disinfected and rinsed following Sicard et al. (2003). They were then crushed together for 5 min with a microtube piston to liberate bacterial symbionts. Hundred-microlitre samples of appropriate dilutions were then streaked onto NBTA plates and incubated at 28 °C for 48 h. For each pool the number of bacteria per IJ was estimated by counting CFU on each dish and using the mean of the two to three counts.

For each infected insect, the number of carried bacteria was estimated several times, using different dilutions. Estimates of re-association for each distinct experimental coinfection were obtained by averaging independent counts over the three measured infections.

IJs’ death rate

For each experimental coinfection, death rate was measured on the same three successful infections we used to evaluate the reproduction rate and the number of carried bacteria. Dead and alive IJs were distinguished on the basis on their morphology. Dead IJs have a typically curved and granular appearance and do not respond to a tactile stimulus (Mitani et al., 2004; Emelianoff et al., 2007). For each infection, 1000 IJs were placed into a grid-drawn Petri dish 1 week after their collection. The first count of alive and dead IJs was made. Petri dishes were then placed at 24 °C and further counts on the same dish were made 8, 12 and 16 weeks after the initial measurement for S. carpocapsae, and after 2, 6, 10, 14, 18 and 22 weeks for S. feltiae. From these measures, the proportion of dead nematodes was calculated for each observation time.

Death rate was estimated for each experimental coinfection as the slope of a logistic regression of proportion of dead IJs over time with a quasi-binomial distribution, which takes over-dispersion into account, and a logit link function. It is possible, in the case of a low death rate that this slope is accidentally negative. The only biological signification of this situation is, of course, that the death rate is close to zero over the time period we considered.

Estimation of nematodes and bacteria overall fitnesses

We estimated nematodes overall fitness as the product of their parasitic success (proportion of successful infections), their reproduction rate (number of IJs emitted per successful infection) and the survival of newborn IJs.

Survival was estimated, from the calculation described in the previous section, as the predicted proportion of IJs that will survive on average after 100 days at 24 °C. For each experimental coinfection, we performed this calculation for the three infections we used to measure IJs’ death rate. We then multiplied the obtained prediction by the number of IJs these infections produced. Finally, we multiplied the average of the values we obtained by the proportion of infestations that yielded some alive IJs after 28 days. When the experimental coinfection never succeeded to produce living IJs, we fixed nematodes’ fitness to zero.

Bacteria fitness is estimated by multiplying nematodes’ fitness by the number of bacteria IJs retain. Again we used the number of bacteria IJs carry measured for the same three infections we used to measure death rate. We then averaged the three independent fitness estimates we obtained.

Results

Specificity in re-association

A first inspection of which bacteria IJs carry in S. carpocapsae shows that all the bacterial strains that belong to the species X. nematophila can re-associate with the nematode (see Table 1). Other strains that belong to other species of Xenorhabdus do not re-associate with the remarkable exception of strain ID10 of X. griffinae. The general picture is thus that re-association is a very specific process in S. carpocapsae.

In S. feltiae the situation is more complex. Outside the species X. bovienii, to which the native symbiont belongs, five strains are capable of re-associating with the nematode (see Table 1). This gives the general picture that re-association is less specific in S. feltiae than in S. carpocapsae. Another interesting finding is that although all X. nematophila strains do re-associate with S. carpocapsae, some X. bovienii strains do not re-associate with S. feltiae. These strains are not naturally associated with S. feltiae (see Table 1, strains F3 and FR43) but with S. affine.

In order to go beyond this simple inspection, we ran a logistic regression analysis on a binary variable which indicates whether or not a bacterial strain does re-associate with a nematode species (see Table 2). In this analysis, explanatory variables are the nematode species, the bacteria distance and the interaction between these two factors. We found a strong negative effect of bacteria distance but only a marginal interaction effect. This interaction term quantifies the difference between the two nematode species in how the probability that a bacterial strain does re-associate relates to bacteria distance. Our analysis therefore indicates that only a weak difference in re-association specificity exists between S. carpocapsae and S. feltiae.

Table 2.   sAnalysis of a binary variable which indicates whether a bacterial strain re-associate with a nematode species.
 d.f.DeviancePr (>χ2)
  1. The statistical model used here is a generalized linear model with quasi-binomial distribution and logit link function. Explanatory variables are the bacteria distance, the nematode species and the interaction between these two factors.

Nematode species12.0910.101
Bacteria distance117.3832.213E−06
Bacteria distance × nematode species14.4140.017

After having analysed which bacteria are retained, which probably reflects which bacteria are recognized by nematodes, we surveyed how many bacteria cells nematodes carry. In both S. carpocapsae and S. feltiae, the number of bacteria a single IJ carries sharply decreases when the distance between the bacterial strain and the native symbiont increases (see Fig. 2, Kendall correlation τ = − 0.64, P = 0.0018 and τ = − 0.55, P = 0.0045 respectively). A high number of bacteria carried indicates that: (i) bacteria are recognized by nematodes and (ii) that they can multiply in the nematode's intestine. Our results suggest that both aspects of re-association are specific to S. carpocapsae and S. feltiae.

Figure 2.

 Average number of carried bacteria per infective juvenile (IJ) as a function of the phylogenetic distance between the tested bacterial strain and the native symbiont. The vertical dashed line indicates the phylogenetic distance above which the bacterial strain used in the experimental coinfection does not belong to the same species than the native symbiont. The curve corresponds to the prediction of a generalized linear model with a Poisson distribution adjusted on raw data. Measurements have been made on coinfections that produced enough IJs (i.e. on 23 strains for Steinernema carpocapsae and on 25 strains for Steinernema feltiae). We found a significant negative correlation between bacteria distance and the number of bacteria retained per IJ in both species (Kendall correlation τ = − 0.64, P = 0.0018 in S. carpocapsae and τ = − 0.55, P = 0.0045 in S. feltiae) with no difference in this correlation between the two nematode species (permutation test, P = 0.378).

A nonparametric comparison of the two nematode species shows that S. feltiae carries less bacteria than S. carpocapsae (Wilcoxon test, W = 392.5, P = 0.0035). The same sort of statistical technique does not allow to detect any difference between the two nematode species in how the number of carried bacteria relates to bacteria distance: 378 random permutations out of 1000 yielded a squared difference between correlations that was greater or equal than the value of 0.019 we observed in our experiment. The correlations we measured in S. carpocapsae and S. feltiae are therefore not statistically different.

Specificity during insect host exploitation

Raw numbers of successful infestations 28 days after infection are given for each experimental coinfection in Table 1. Figure 3 presents the emergence rate for each nematode species as a function of the bacteria distance. The emergence rate measures both how frequently and how fast nematodes emerge from infected insects (for details, see the Materials and methods section).

Figure 3.

 Emergence rate as a function of the phylogenetic distance to the native bacteria. A value of zero corresponds to experimental coinfections where none of the 50 infested G. mellonella produced nematode infective juveniles (IJs). High values indicate coinfections where most infected insects rapidly yielded nematode IJs. Emergence rate was estimated by fitting, for each coinfection, a survival regression model with Weibull distribution on parasitic success data (see text for further details). The vertical dashed line indicates the phylogenetic distance above which the bacterial strain used in coinfection does not belong to the same species than the native bacterial strain. We found a negative correlation between bacteria distance and emergence rate in Steinernema carpocapsae (Kendall correlation, τ = − 0.39, P = 0.0018) but no relation in S. feltiae (Kendall correlation, τ = 0.09, P = 0.4768). The effect of bacteria distance on emergence rate statistically differs between the two species (permutation test, P = 0.0001).

In S. carpocapsae, the emergence rate sharply decreases with the bacteria distance (Kendall correlation, τ = − 0.39, P = 0.0018). When combined with some phylogenetically distant bacteria, S. carpocapsae often fails in exploiting insects or, if it does succeed, the process leading to nematodes emergence is much slower than with the native symbiont. In fact, nematodes perform better with no bacteria at all than with some of the phylogenetically distant bacterial strains (see Fig. 3). These bacteria have therefore a deleterious effect on nematodes.

In S. feltiae the situation is completely different. First, emergence rate is highly variable among strains that all belong to the same species than the native symbiont (see Fig. 3). In fact, even the strain FR44 yields a much lower emergence rate than the native symbiont, although it has the same 16S RNA sequence and its geographical origin is only a few kilometres away from that of the native symbiont. In addition, as in S. carpocapsae, there are strains that reduce S. feltiae’s emergence rate below that of aposymbiotic nematodes; but, contrary to S. carpocapsae, these deleterious strains all belong to the same species than the native symbiont (X. bovienii). At the same time, some phylogenetically distant bacteria increase S. feltiae’s emergence rate, sometimes even above the emergence rate observed with their native symbiont.

Overall, there is no clear relationship between the bacteria distance and the emergence rate in S. feltiae (Kendall correlation, τ = 0.09, P = 0.4768) and the relation between emergence rate and bacteria distance is significantly different between the two species (observed squared difference between the correlations is 0.24, with only one simulation out of 1000 yielding greater or equal squared difference).

Another aspect of how successful nematodes were in exploiting their host can be surveyed by analysing how many IJs emerge from an insect when infection is successful. These results are presented in Fig. 4. We found no correlation between bacteria distance and reproduction rate in both S. carpocapsae (Kendall's correlation, τ = 0.047, P = 0.7469) and S. feltiae (Kendall's correlation, τ = − 0.051, P = 0.7074).

Figure 4.

 Reproduction rate as a function of the phylogenetic distance to the native bacteria. This rate is the average number of infective juveniles (IJs) produced per infection. It has been estimated over three successful infections for each bacterial strain. Hence, bacterial strains which completely suppress nematode reproduction in the insect (nine strains of 33 for Steinernema carpocapsae, one strain of 28 for Steinernema feltiae) are not represented in this figure. We found no correlation between bacteria distance and reproduction rate in both species (Kendall's correlation, P = 0.7469 and 0.7074 respectively).

The other side of specificity: IJs’ death rate during free stage

Bacteria can impact IJs’ death rate in two different ways. First, bacteria determine in part how efficiently nematodes exploit the insect. They therefore determine somehow the quality of the parasitic environment in which IJs were born, which in turn has some consequences on how well they survive during their free stage. Second, when IJs carry them during their free stage, bacteria can represent a cost and therefore have a direct impact on survival (Mitani et al., 2004; Emelianoff et al., 2007, 2008a). In our experiment, these two effects are confounded when IJs carry bacteria.

We found that the death rate in S. carpocapsae tends to decrease with bacteria distance (see Fig. 5, Kendall correlation, τ = − 0.43, P = 0.0065). It is quite remarkable, in fact, that IJs die the most when they carry their native symbiont (see Fig. 5). Conversely, the lowest death rates are achieved with distant bacterial strains that nematodes do not retain. The direct cost of carrying bacteria seems therefore to prevail in S. carpocapsae, although the statistical signal is weak and the death rate is highly variable.

Figure 5.

 Death rate of newly emerged infective juveniles (IJs) as a function of the phylogenetic distance to native bacteria. Death rate has been estimated using a generalized linear model with quasi-binomial distribution and a logit link function where the proportion of dead nematodes is a function of time. The vertical dashed line indicates the phylogenetic distance above which the non-native and the native bacterial strains do not belong to the same species. Open symbols indicate bacterial strain that do not re-associate with nematodes; grey symbols indicate those that do re-associate. In Steinernema feltiae, nematodes which coinfected with strain FR44 survive as well as aposymbiotic nematodes. Measurements have been performed on the 21 strains for Steinernema carpocapsae and 26 for S. feltiae for which enough IJs were available. Death rate decreases with bacteria distance in S. carpocapsae (Kendall correlation, τ = − 0.43, P = 0.0065), whereas it increases in S. feltiae (Kendall correlation, τ = 0.36, P = 0.01) the difference between the two species being highly significant (permutation test, P < 0.001).

Conversely, in S. feltiae, death rate increases with bacteria distance (see Fig. 5, Kendall correlation, τ = 0.36, P = 0.01) this being significantly different from what we observed in S. carpocapsae (observed difference between correlations is 0.65, with no random permutation out of 1000 performed yielding such a difference). This indicates that the cost of carrying bacteria is probably lower in S. feltiae than in S. carpocapsae (which is consistent with the higher number of carried bacteria in the latter) and that the death rate is determined mostly by how well the IJs’ parents did exploit their insect hosts.

Getting the global picture: how nematodes and bacteria fitness relate to bacteria distance?

Nematodes’ fitness obviously depends on parasitic success, reproductive success and survival. We therefore combined these three measurements and related the resulting estimate of fitness to bacteria distance (see Fig. 6).

Figure 6.

 Top line: estimated nematodes’ fitness as a function of the phylogenetic distance to the native bacteria. Fitness is estimated as the product of parasitic success 28 days after infection, of the number of larvae emitted per successful infection and of the probability that larvae survive 100 days at 24 °C. In both species fitness negatively correlates with bacteria distance (Kendall correlation, τ = − 0.25, P = 0.051 for Steinernema carpocapsae, τ = − 0.25, P = 0.060 for Steinernema feltiae). We found no difference in the effect of bacteria distance on fitness between the two nematode species (permutation test, P = 0.989). Bottom line: estimated bacteria fitness. These estimates are obtained by multiplying nematodes’ fitness by the number of bacteria infective juveniles carry. Again, fitness negatively correlates with bacteria distance in both species (Kendall correlation, τ = − 0.52, P = 1.2 × 10−4 for S. carpocapsae, τ = − 0.50, P = 4.5 × 10−4 for S. feltiae) with no difference in correlation between the two species (permutation test, P = 0.725). In all graphs, the solid line represents a linear repression, whereas the vertical dashed line indicates the bacteria distance above which bacterial strains do not belong to the same species than the native symbiont.

For both S. carpocapsae and S. feltiae, although marginally, the estimated fitness negatively correlates with bacteria distance (Kendall correlation, τ = − 0.25, P = 0.051 for S. carpocapsae, τ = − 0.25, P = 0.060 for S. feltiae). We found no difference between these two correlations: observed squared difference between correlations is 3.6 × 10−6 and out of 1000 random permutations, 989 yielded a higher difference between the two correlation coefficients.

We also calculated bacteria fitness by multiplying the estimate we obtained for nematodes by the number of bacteria their IJs carry. Again, fitness negatively correlates with bacteria distance (Kendall correlation, τ = − 0.52, P = 1.2×10−4 for S. carpocapsae, τ = − 0.50, P = 4.5 × 10−4 for S. feltiae) with no difference in this correlation between the two nematode species (observed squared difference between correlations is P = 7.4 × 10−4, and 827 random permutations out of 1000 yielded comparable or greater differences between correlation coefficients).

Discussion

The case of Steinernema carpocapsae

In S. carpocapsae we found a clear negative correlation between bacteria phylogenetic distance in both re-association and emergence rate: bacteria that differ too much from the native symbiont do not re-associate with S. carpocapsae and are ‘toxic’ in the sense that they decrease its parasitic success below what it could achieve without any bacteria. These findings are in agreement with those reported by Sicard et al. (2004): with a different sampling, these authors also demonstrated specificity in both re-association and parasitic success; they also found that some phylogenetically distant bacteria were ‘toxic’ to nematodes. This general specificity pattern might indicate a long history of coevolution between X. nematophila and S. carpocapsae, which is probably allowed by a low frequency of horizontal transfer in this system.

There are exceptions to this general pattern, though. Some bacteria which do not re-associate with S. carpocapsae increase parasitic success compared with that of aposymbiotic nematodes. Conversely, some bacteria that provide a service comparable with that of the native symbiont do not re-associate well. Both host exploitation and re-association are specific, but the match between the specificity pattern in these two traits is not perfect. This indicates that the mechanisms involved in these two forms of specificity are different, those determining emergence rate being probably less specific than those involved in re-association.

The specificity of re-association indeed relies on molecular recognition between nematode and bacteria. Several studies have shown that bacterial genes coding surface molecules are necessary, if not sufficient, for X. nematophila to colonize S. carpocapsae IJs (Heungens et al., 2002; Cowles & Goodrich-Blair, 2005; Chandra et al., 2008). These genes are reported as being specific to X. nematophila. Our results confirm this view, with the remarkable exception of strain ID10 of X. griffinae which is capable of re-associating with S. carpocapsae. This finding would probably deserve a careful inspection and might improve our understanding of the mechanisms of re-association in the S. carpocapsae/X. nematophila symbiosis.

Conversely, the efficiency of host exploitation, which determines both emergence and reproduction rate, is probably multi-factorial. Several bacterial genes have already been shown to be involved in symbiosis during the parasitic stage (Heungens et al., 2002; Cowles & Goodrich-Blair, 2005; Banerjee et al., 2006; Cowles et al., 2007; Lanois et al., 2008). They determine two distinct types of trait in bacteria: (i) its pathogenicity towards the insect based on immuno-suppressive toxins secretion and (ii) its ability to digest and bio-convert insect cadaver into nutrients for nematodes. These two activities probably do not imply a precise molecular recognition from the nematode's side, contrary to what has been documented for nilABC involved in re-association (Cowles & Goodrich-Blair, 2004, 2008). A different mechanism could also make the symbiotic interaction during host exploitation specific: Xenorhabdus bacteria produce some toxins which are targeted against insects but might also have detrimental effects on nematodes (Han & Ehlers, 1999; Hu et al., 1999; Couillault & Ewbank, 2002; Sicard et al., 2007). It is possible that nematodes are co-adapted to their bacteria in such a way that they resist to the toxins produced by their symbionts but not to those produced by other Xenorhabdus species or strains.

Whichever mechanism applies, we found that emergence rate is specific, but reproduction rate is not. Whether nematodes emerge or not probably depends mostly on the very early stages of insect exploitation. Conversely, reproduction rate should depend more on how well nematodes multiply once these early stages have been successfully accomplished. Our analysis therefore indicates that the early stages of host exploitation are more specific than the later ones. This contradicts in part Sicard et al. (2004), who found specificity on both parasitic success and reproduction rate in S. carpocapsae, but similar findings have already been reported by Emelianoff et al. (2008a).

Finally, we also detected a negative correlation between S. carpocapsae IJs’ death rate and bacteria distance. This correlation is more complex to interpret, as IJs’ death rate depends on both the number of bacteria IJs carry and the parasitic environment in which they have been produced. Death rate can be high either because IJs carry many bacteria (Emelianoff et al., 2007, 2008a) or because their parents were not capable of exploiting efficiently the insect host. These two possible causes are confounding factors because they might both relate to bacteria distance. Still, a more detailed statistical analysis (data not shown here) indicates that the most prominent factor in determining S. carpocapsae IJs’ death rate is probably the number of bacteria they carry. The negative correlation between death rate and bacteria distance therefore mirrors the correlation between that same distance and the number of bacteria carried.

Steinernema feltiae: another type of host exploitation specificity

In S. feltiae, as in S. carpocapsae, we found a strong negative correlation between bacteria distance and the number of bacteria carried per IJ. Still, some phylogenetically distant bacteria are capable of re-associating with S. feltiae (among which one strain of X. nematophila). This indicates at least a qualitatively lower re-association specificity in S. feltiae compared with S. carpocapsae. A more quantitative approach, though, shows that the difference in re-association specificity between S. carpocapsae and S. feltiae is only weak.

Although S. feltiae exhibits specificity in re-association just as S. carpocapsae does, its emergence rate does not relate at all to bacteria distance. This is in part due to the fact that within the very species X. bovienii, to which the native symbiont belongs, the variance in emergence rate is huge. For example, ‘toxic’ bacteria, that reduce nematodes’ emergence rate below what they could do with no bacteria at all, exist in both S. feltiae and S. carpocapsae. However, in S. feltiae these strains all belong to the same species than the native symbiont, although in S. carpocapsae they are all phylogenetically distant from the native symbiont. In fact the best and worst bacterial strains, in terms of S. feltiae’s emergence rate, all belong to X. bovienii. Even the FR44 strain, which has been found associated with S. feltiae in a location 32 km away from that of the native symbiont, yields a lower emergence rate than the native symbiont. This high variance has to be contrasted with the situation in S. carpocapsae where bacterial strains belonging to the same species than the native symbiont yield comparable emergence rates (see Fig. 3). This striking difference can be interpreted in three different, but not exclusive, ways.

First, the difference might be due to higher frequency of horizontal transfer of X. bovienii. This is consistent with the lower specificity in re-association that we document here. This is also consistent with the fact that some X. bovienii strains are associated with nematodes that do not belong to S. feltiae. The bacteria phylogenetic distance would then be a bad predictor of co-adaptation between bacteria and nematode. This is unlikely as we did not find a clear difference in re-association specificity between S. carpocapsae and S. feltiae. This might reflect a lack of statistical power in our work, but still if a difference in specificity exists between the two nematode species it is probably too weak to explain the very different patterns observed in emergence rate.

A second, and somewhat similar, reason might be that the 16S RNA distance that we use in our study does not reproduce the phylogenetic history of the bacterial genes that determine emergence rate. This would mean that these genes can be transferred horizontally between bacteria that interact with S. feltiae. This is highly possible and the lack of congruence between 16S RNA phylogenies and phylogenies drawn from the sequences of the genes involved in symbiosis has been documented in other systems (e.g. Baily et al., 2007). However, as we found a correlation between 16S RNA bacteria distance and re-association capabilities, this would further imply that the rate of horizontal transfer is much higher in genes that are involved in host exploitation than in those that determine re-association. This would be possible if bacterial genes involved in host exploitation were carried, in X. bovienii, on mobile genetic elements such as bacterial plasmids. Sergeant et al. (2006) have shown that insecticidal toxin genes, forming a pathogenicity island, were integrated in the chromosome of both X. nematophila and X. bovienii, but that they were adjacent to phage-like structures in X. bovienii only. This clearly supports the idea that genes involved in host exploitation could be horizontally transferred at a higher rate in X. bovienii.

A third type of explanation would be that, as the ecology of S. feltiae differs a lot from that of S. carpocapsae (Stuart et al., 2006), the genes that determine the efficiency of insect host exploitation in these two nematode species do not experience the same selective pressures. This view is supported by the fact that, in S. feltiae, we did detect a positive correlation between bacteria phylogenetic distance and IJs’ death rate, whereas in S. carpocapsae bacteria phylogenetic distance correlates negatively with emergence rate. In fact, when we combine all our measurements in order to estimate nematodes’ fitness, we do not detect any difference in specificity between the two nematode species. Both symbiotic systems are therefore specific, but they differ in which life-history trait causes specificity. This can be rephrased by saying that ‘good’ bacteria in S. carpocapsae are those that increase parasitic success; ‘good’ bacteria in S. feltiae would be those that allow parent nematodes to produce IJs that survive longer. This is consistent with the fact that S. carpocapsae is an ambush predator that probably does not move much in the soil, whereas S. feltiae is more active in searching its preys (Stuart et al., 2006).

This latter scenario explains the positive correlation between bacteria distance and IJs’ death rate in S. feltiae; it can also explain the absence of any correlation between emergence rate and bacteria distance. It cannot explain, though, why some bacteria that belong to the same species than the native symbiont of S. feltiae are toxic. One possible explanation would be that these bacteria are toxic because they produce insecticidal toxins which the nematode strain we have used in our experiment is sensible to. The existence of such toxic effect has already been suggested (Hu & Webster, 1995; Han & Ehlers, 1999; Couillault & Ewbank, 2002; Sicard et al., 2007). Their role in the functioning of the symbiotic interaction remains to be studied.

Conclusion

We found that, for both nematodes and bacteria, fitness decreases when the phylogenetic distance to the native symbiont increases. This can be taken as a sign of coadaptation. A careful inspection revealed that different life-history traits are involved in specificity, and maybe in coadaptation, in S. carpocapsae/X. nematophila and S. feltiae/X. bovienii symbioses. These differences raise two remarks. First S. carpocapsae and S. feltiae, as they have followed distinct evolutionary paths, should be two models of interest in the study of the coevolution between Steinernema and Xenorhabdus. Second and more importantly, studying which life-history traits create specificity between hosts and symbionts can give us cues on how selection shapes mutualisms.

Acknowledgments

We are indebted to Angus Buckling and an anonymous referee who both helped improve this work. We are also grateful to Benoît Nabholz and Nicolas Galtier who both helped in analysing sequence data. JBF and EC have been funded by the Agence Nationale de la Recherche (grant ANR-JC EvolNemBact). This is publication ISEM 2009-086.

Ancillary