Pervasive effects of Wolbachia on host activity

Abstract

Heritable symbionts have diverse effects on the physiology, reproduction and fitness of their hosts. Maternally transmitted Wolbachia are one of the most common endosymbionts in nature, infecting about half of all insect species. We test the hypothesis that Wolbachia alter host behaviour by assessing the effects of 14 different Wolbachia strains on the locomotor activity of nine Drosophila host species. We find that Wolbachia alter the activity of six different host genotypes, including all hosts in our assay infected with wRi-like Wolbachia strains (wRi, wSuz and wAur), which have rapidly spread among Drosophila species in about the last 14 000 years. While Wolbachia effects on host activity were common, the direction of these effects varied unpredictably and sometimes depended on host sex. We hypothesize that the prominent effects of wRi-like Wolbachia may be explained by patterns of Wolbachia titre and localization within host somatic tissues, particularly in the central nervous system. Our findings support the view that Wolbachia have wide-ranging effects on host behaviour. The fitness consequences of these behavioural modifications are important for understanding the evolution of host–symbiont interactions, including how Wolbachia spread within host populations.

1. Introduction

Insects harbour microorganisms that have wide-ranging effects on their performance and fitness [1–3], including manipulating reproduction [4–7], provisioning nutrients [1,8,9], modifying thermotolerance [10,11] and defending against pathogens [12–15]. Microbes may also alter host behaviour [16–21]. In extreme instances, parasitic microbes can induce behaviours that increase the likelihood of transmission—for example, by directing hosts to habitats that promote transmission [22–28]. Infected hosts may also change their own behaviour as an immune strategy against infection, including seeking warm temperatures to induce a ‘behavioural fever' [29,30] or reducing activity and increasing sleep time [19,31–34]. Such behavioural modifications have important implications for microbe spread and host fitness.

Maternally transmitted Wolbachia are the most common endosymbionts in nature, infecting many arthropods [5,35,36] and some filarial nematodes [37,38]. Discordant Wolbachia and host phylogenies indicate that many hosts have recently acquired Wolbachia via introgressive and horizontal transfer [39–44]. Wolbachia are primarily transmitted vertically by female hosts, so natural selection favours beneficial effects on host fitness that promote spread [45–48]. Maternal transmission occurs via the host reproductive system, but Wolbachia are also found in host somatic tissues, including nervous, digestive and metabolic tissues [49–52]. Still, the behavioural and physiological consequences of somatic infections are poorly understood [19,52].

Prior work indicates Wolbachia influence several host behaviours [19,53,54], including sleep [55–57] and temperature preference [20,58,59]. We broadly test for Wolbachia effects on the locomotor activity of Drosophila hosts infected with A-group Wolbachia (N = 11), B-group Wolbachia (N = 1) and an A- and B-group co-infection (N = 1). Our analysis includes two prominent A-group clades that recently spread among Drosophila: wMel-like Wolbachia (wMelCS, two wMel variants, wYak, wSan and wTei) and wRi-like Wolbachia (wRi, wSuz and wAur) [43,44]. We find that Wolbachia effects on host activity are common, particularly for wRi-like Wolbachia, a ‘super-spreader’ strain that rapidly spread among Drosophila species in approximately the last 14 000 years [43].

2. Methods

(a) . Fly lines

We evaluated 13 different Wolbachia-infected host genotypes (figure 1; electronic supplementary material, table S1), consisting of nine Drosophila species infected with 14 different A- and B-group Wolbachia that diverged up to 46 Ma [60]. For two of the host species, D. melanogaster and D. simulans, we tested multiple Wolbachia-infected genotypes. This included a D. simulans host co-infected with A-group wHa and B-group wNo [61–64]. We used tetracycline treatment as previously described [20] to generate uninfected genotypes to pair with each infected genotype, while taking care to avoid detrimental effects of the antibiotic treatment on mitochondrial function [65] (see electronic supplementary material, Methods).

Figure 1.

The activity of uninfected and infected flies for each sex of each genotype. Activity is measured as mean ADS. Significance was evaluated using linear models (electronic supplementary material, tables S2 and S3).

(b) . Host locomotor activity assays

We reared flies at 25°C under a 12 h L : 12 h D cycle (Percival model I-36LL) on a standard food diet [20]. Each day, we collected a batch of female and male virgins for one pair of uninfected and infected genotypes. The four treatment groups (uninfected females, infected females, uninfected males and infected males) were maintained in isolation until they were 3 to 5 days old. We then measured the locomotor activity of the batch of flies using a 16-chamber flow-through respirometry and data acquisition system (MAVEn, Sable Systems International). The MAVEn has sixteen 2.4 ml volume polycarbonate animal chambers and an activity board that uses infrared light (invisible to flies) to monitor animal activity in each chamber, sampled at 1 Hz (electronic supplementary material, figure S1). Individual flies were aspirated into a randomly assigned chamber and allowed to adjust to the new environment for 0.5 h. Activity measurements were then recorded over a 3 h period between the hours of 09.00 and 16.00.

The raw outputs from the activity sensors were transformed into the activity index absolute difference sums (ADS). We calculated ADS by first calculating the cumulative sum of the absolute difference between consecutive activity readings and then calculating the slope of cumulative activity versus time [66,67]. We used mean ADS over the 3 h period as our estimate of locomotor activity for each fly; however, our analyses were robust regardless of how we quantified activity (see electronic supplementary material, Methods). We found that the mean ADS activity data required a transformation for statistical analysis; however, a single data transformation was not suitable for all host species. We used a log transformation of mean ADS for D. simulans, D. suzukii, D. auraria, D. mauritiana and D. sechellia, and a square root transformation for D. melanogaster, D. yakuba, D. santomea and D. teissieri. We present a full statistical analysis of all datasets in the electronic supplementary material, tables S2 and S3, respectively.

We used the log- and square root-transformed mean ADS data as dependent variables in linear models. We included infection status, sex and an infection-by-sex interaction effect as independent variables, as well as additional independent variables to account for other potential sources of activity variation: randomly assigned animal chamber (1–16), experimental start time, mean water vapour (ppt), mean relative humidity (%), mean temperature (°C) and mean light intensity (lux) [66,67]. We evaluated the significance of individual effects using F tests and type III sum of squares using the ‘Anova' function in the car R package [68,69].

(c) . Wolbachia phylogenomic analysis

We used publicly available Wolbachia genome assemblies [70–76], and new Illumina sequencing, to generate a Bayesian phylogram [20] (see electronic supplementary material, Methods). Wolbachia effects on host activity were especially common for wRi-like Wolbachia, so we used the phylogram to test whether Wolbachia effects on hosts exhibit phylogenetic signal. First, we treated Wolbachia effects on host locomotor activity as a binary trait and tested for a phylogenetic signal using the D statistic [77], implemented in the caper R package [78]. Second, we treated Wolbachia effects on activity as a continuous trait and tested for phylogenetic signal using Pagel's lambda (λ) [79]. Here, we analysed each sex separately, because we found significant infection-by-sex interaction effects on activity (electronic supplementary material, tables S2 and S3). For each sex, we extracted the least-square (LS) mean ADS for infected and uninfected flies from the linear models (electronic supplementary material, tables S2 and S3) and used the change in LS mean activity as a continuous character to calculate the maximum-likelihood value of Pagel's λ [79,80]. We used a likelihood ratio test to compare our fitted value of λ to a model assuming no phylogenetic signal (λ = 0) using the ‘phylosig' function in the R package phytools [81].

3. Results

(a) . Wolbachia infections modify host locomotor activity

We assayed the locomotor activity of 3104 flies (figure 1). Wolbachia had a significant effect on the activity of six host genotypes, including hosts infected with both A- and B-group Wolbachia. Interestingly, the direction of Wolbachia effects on host activity varied by genotype and sex (figure 2). We found a significant Wolbachia infection-by-sex interaction effect for the wMelCS-D. melanogaster genotype that increased male activity (F = 4.566, p = 0.033; electronic supplementary material, table S3). We also found a significant infection-by-sex effect for the wRi-D. simulans genotype, but Wolbachia increased female activity (F = 8.150, p = 0.005; electronic supplementary material, table S2). The two other closely related wRi-like Wolbachia, wSuz and wAur, also had significant effects on host activity. The wSuz-D. suzukii genotype had a significant main effect on Wolbachia that reduced host activity (F = 11.311, p < 0.001; electronic supplementary material, table S2), and the wAur-D. auraria genotype had a significant infection-by-sex interaction that reduced female activity (F = 6.584, p = 0.011; electronic supplementary material, table S2). The wHa-D. simulans genotype had a significant main effect of Wolbachia that increased host activity (F = 7.764, p = 0.006; electronic supplementary material, table S2). Lastly, we found the wHa-wNo co-infected D. simulans genotype had a significant infection-by-sex interaction effect that reduced male activity (F = 7.076, p = 0.008; electronic supplementary material, table S2). Because this genotype is co-infected, we do not know the relative contributions of wHa and wNo to variation in host activity.

Figure 2.

(a) Estimated Bayesian phylogram for A- and B-group Wolbachia strains. The divergence estimate for A- and B-groups is superimposed from Meany et al. [60]. All nodes have Bayesian posterior probabilities of 1. (b) Wolbachia effects on host activity scored as a binary trait: Wolbachia significantly altered host activity (black circle) or had no effect (white circle). (c) Wolbachia effects on activity scored as a continuous trait: the change in LS mean log-transformed activity (ADS) for each sex. LS means were generated from linear models (electronic supplementary material, table S2). LS mean square root-transformed ADS data are shown in the electronic supplementary material, figure S3.

(b) . Limited evidence for phylogenetic signal

We estimated a Bayesian phylogram of A- and B-group Wolbachia using 211 single-copy genes of identical length in all Wolbachia genomes, spanning 178 569 bp (figure 2). We then tested whether closely related Wolbachia have similar effects on host activity. When treating Wolbachia effects on activity as a binary trait, our estimate of D = 0.322 was low, but not statistically different from a model of D = 1 assuming phylogenetic randomness (p = 0.101) or a model of D = 0 with strong phylogenetic signal (p = 0.198). Simulations of similar phylogenies with an increasing number of Wolbachia strains suggest that at least N = 50 strains are required to differentiate our estimated value of D = 0.322 from a model of phylogenetic randomness (D = 1) (electronic supplementary material, figures S4 and S5). Thus, Wolbachia effects on host activity may exhibit phylogenetic signal, but many more Wolbachia strains are required to test this hypothesis. Unfortunately, N = 50 strains are not presently available in culture. We also treated Wolbachia changes to host activity as a continuous trait; however, we found that maximum-likelihood fitted λ values were extremely low, indicative of no phylogenetic signal. λ values generated from the LS mean log-transformed ADS data were not statistically different from zero for females (λ < 0.001, p = 1) or males (λ < 0.001, p = 1). This was also true when we repeated the analyses for the LS mean square root-transformed ADS data for females (λ < 0.001, p = 1) and males (λ < 0.001, p = 1).

4. Discussion

Our analyses suggest that Wolbachia commonly alter host locomotor activity, which may affect host fitness. Locomotion is a basic host activity underlying many ecologically important behaviours, including foraging, thermoregulation and mate seeking. In combination with our recent work demonstrating pervasive effects of A- and B-group Wolbachia on host temperature preference [20], we posit that Wolbachia infections may often alter host behaviour.

The wRi-like Wolbachia strains in our study (wRi, wSuz and wAur) consistently altered host activity. We found a low, but non-significant D value of 0.322, suggesting effects on host activity may exhibit phylogenetic signal; although an excessive number of Wolbachia strains are required to test this hypothesis. Our findings are consistent with prior experiments demonstrating that wRi increased female D. simulans activity in response to olfactory cues [82,83]. We hypothesize that the prominence of wRi-like Wolbachia effects on host activity relative to other strains may be due to variation in Wolbachia tissue localization [50,53]. wRi occurs at high titre in adult D. simulans brains and localizes to specific regions, whereas wMel shows a relatively even distribution in D. melanogaster [53]. wRi also occurs at a higher titre in the ventral nerve cord, which is a major neural circuit centre for motor activities such as walking [53,84–86]. Future experiments should compare Wolbachia titre and localization in adult brains for wRi-like variants and strains that do not alter locomotor activity.

We also found considerable variation in the direction and sex-bias of Wolbachia effects on locomotor activity (figure 2). Wolbachia decreased activity for wSuz, wAur and the wNo-wHa co-infection, whereas wMelCS, wRi and wHa increased activity. These effects were female-biased for wRi and wAur, but male-biased for wMelCS and wNo-wHa. This variation had no relationship to the Wolbachia phylogeny, because we found no evidence for phylogenetic signal when measuring Wolbachia effects on females and males as a continuous trait (λ < 0.001). Specific Wolbachia effects on host activity may depend on interactions with the host background. For example, our work and others' suggests that identical wMelCS variants have different effects on D. melanogaster temperature preference depending on the host background [20,58,59]. Host genomes also modify Wolbachia titre [87], Wolbachia maternal transmission [88], components of host fitness [89–92] and the strength of cytoplasmic incompatibility [93–95].

Changes to host activity could underlie Wolbachia-induced behaviours that promote infection spread. For example, wMel-infected D. melanogaster have higher field recapture rates than uninfected flies [96], and long distance dispersal of the spider Erigone atra is altered by Rickettsia, an endosymbiont closely related to Wolbachia [97]. Our own work suggests Wolbachia may alter host temperature preference to promote Wolbachia replication within host bodies [20]. Other experiments suggest that wMel and wRi may influence male mating rate [98,99]. Alternatively, hosts may be modifying their own behaviour as a response to Wolbachia infection. Several studies indicate that wMel alters circadian activity and sleep patterns of D. melanogaster [53,55–57]. For example, Bi et al. [56] report that wMel increases sleep time, which could represent a host immune response to infection [19]. Ultimately, these effects on host behaviour factor into how Wolbachia influence host fitness, which determines the spread and persistence of Wolbachia in host populations [4,100–103]. Because locomotor activity is such a fundamental host behaviour, our results suggest Wolbachia may have complex and variable effects on many components of host fitness.

Data accessibility

All data and code are available on Dryad (https://doi.org/10.5061/dryad.6t1g1jwxv) [104]. Newly reported genome assemblies are available on GenBank (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA717279).

Authors' contributions

M.T.J.H. and B.S.C. conceived of the study. M.T.J.H. conducted the experiments, analysed the data, and drafted the manuscript. M.T.J.H., H.A.W. and B.S.C. designed the experiments, edited the manuscript, and approved the final version. B.S.C. coordinated the study. All authors agree to be held accountable for the content therein.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by the NIH under award R35GM124701 to B.S.C.