Dispersal behaviour after antibiotic treatment to remove endosymbionts
We compared the dispersal behaviour of wild-caught individuals (males and females) in which we had manipulated endosymbiont infections through treatment with the antibiotics tetracycline (T) or penicillin-G (P). Individuals within a control group were treated with water (C). Wolbachia, Rickettsia and Spiroplasma are all sensitive to tetracycline; Cardinium is sensitive to both tetracycline and penicillin. All four types of endosymbiont were detected by polymerase chain reaction (PCR, see Methods) in our control (C) population (n = 27 females: Cardinium 70%, Rickettsia 44%, Spiroplasma 11%, Wolbachia 0%; n = 19 males: Cardinium 58%, Rickettsia 47%, Spiroplasma 16%, Wolbachia 5%).
Bacterial transmission rates from females to eggs were used as a measure of antibiotic treatment efficiency given that tetracycline is bacteriostatic and inactive bacteria might remain in antibiotic-treated adults. Lower transmission rates of Rickettsia and Cardinium from infected females to egg sacs following tetracycline treatment suggested that these bacteria had been successfully targeted in this sex (see Methods). The proportion of adult males infected with Cardinium, Spiroplasma and Wolbachia conform to those expected if treatment has been successful (% infected C > T and P > T) but the proportions carrying Rickettsia did not (% infected C < T and P < T), thus in the absence of an alternative method of assessment we cannot rule out the possibility that our treatment of males was less effective.
We compared the behaviour of T, P and C individuals in an artificial wind tunnel, designed specifically to simulate conditions suitable for dispersal. We used the adoption of a tiptoeing posture, which is exclusively associated with aeronautic dispersal and which is essential for becoming airborne, as our ballooning indicator. T females adopted this posture more readily than either C or P females (Figure 1). Multivariate GLM (MANOVA) analysis showed a significant effect of treatment on ballooning posture driven by significant effects on latency (F
2,48 = 5.007; P = 0.011) and total time spent adopting the posture (F
2,48 = 7.139; P = 0.002). Overall, significantly more T females (53%) adopted the ballooning posture than did those in C (26%, χ
2 = 5.81; P = 0.007) or P groups (23%, χ
2 = 5.81; P = 0.019). A similar analysis of male behaviour detected no significant effect of treatment on male ballooning posture. T and P group males had a lower latency (C vs. T P = 0.004, C vs. P P = 0.019) but the difference between T and P was not significant. There were no significant differences amongst treatment groups in the proportions of males exhibiting tiptoeing behaviour (C 26%, P 30%, T 42%, all P > 0.4).
Treatment with tetracycline is a widely employed method of curing arthropods of Rickettsia and Wolbachia infections but the potential side-effects of such a broad-spectrum antibiotic are not well characterised. We found no significant effect of treatment on the total number of eggs produced by females in our experiment (F
2,48 = 0.138; P = 0.872) or on the manner in which they laid their egg sacs (females typically clump some of their egg sacs together but leave others individually spaced; the proportions of spaced egg sacs did not differ significantly amongst treatment groups, C 63%, P 74%, T 82%). We found no significant effect of treatment on general levels of activity, as measured by the amount of time individuals spent in a stationary position (F
2,93 = 2.74; P = 0.07). Side effects of treatment seem unlikely to offer the full explanation for the observed difference in ballooning behaviour given that other traits such as reproductive output and levels of general mobility appear unchanged in response to the same treatment.
Dispersal behaviour and endosymbiont infections in laboratory-reared animals
We subsequently studied the lifetime reproductive success of captive-bred E. atra females and their behaviour under wind tunnel conditions (as in the experiment above) and then identified bacterial infections in these individuals by PCR [11]. Rickettsia (25%), Spiroplasma (6%), Cardinium (21%) but no Wolbachia were detected. We tested for associations between two types of dispersal behaviour that require silk to be produced: long-distance (ballooning) and short-distance (rigging, where anchored silk is used to swing the spider from one point to another close by). Given the results of our first experiment (where tetracycline but not penicillin treatment was associated with a behavioural difference) we tested for an association between Rickettsia infection and dispersal (Figure 2). Rickettsia-infected individuals had a reduced long-distance dispersal probability (F
1,53 = 6.34; P = 0.015) and frequency (F
1,53 = 5.22; P = 0.026). Other behaviours did not differ between Rickettsia-infected and uninfected females: probability of rigging, F
1,56 = 0.09, P = 0.769; rigging frequency, F
1,53 = 1.15, P = 0.337; rigging ability (= length of the produced silk thread), F
1,39.5 = 0.34, P = 0.565; time spent web-building, F
1,52 = 0.29, P = 0.59. No evidence was found that the infection state of mothers affected the sex-ratio of offspring (Wald χ
2 = 0.02; df = 12; P = 0.874) or their fecundity (F
1,12 = 0.01; P = 0.931). Rickettsia-infected females, however, lived an average of 18.68 ± 8.45 days shorter than uninfected females (uninfected lifespan 50.98 ± 7.45 days; -36.64%; F
1,39.6 = 4.89; P = 0.03). Similarly, the total number of egg sacs decreased from 7.45 ± 0.44 in uninfected to 5.14 ± 0.61 in infected females (F
1,52 = 7.48; P = 0.0085) although we could detect no effect on overall lifetime fecundity (F
1,52.9 = 2.11; P = 0.152). The latter is due to marginally lower numbers of eggs in the last egg sacs produced by uninfected individuals. Our data thus indicate a specific association between Rickettsia and dispersal, but only that of the long-distance kind. Subsequent tests established no similar effect of Spiroplasma or Cardinium on either type of dispersal movement.
Dispersal behaviour and endosymbiont infections in natural populations
We studied levels of endosymbiont infections in a large number of individuals from a range of natural populations (Figure 3). All individuals were captured using d-vac sampling. We found several endosymbionts to be present (Rickettsia 772 individuals tested, mean proportion infected 0.67, S.E. ± 0.038; Wolbachia 796 individuals tested, mean proportion infected 0.05, S.E. ± 0.018). Rickettsia was found in a proportion of individuals from all 25 populations tested, but only 14 of these also carried Wolbachia. The mean proportion of individuals found to be infected with Rickettsia did not differ between populations where Wolbachia was also detected and those where it was not detected (mean proportion infected with Rickettsia where Wolbachia also detected = 0.65, S.E. ± 0.054; where Wolbachia not detected = 0.75, S.E. ± 0.046, t = 1.34, df = 25, P = 0.19).
We collected spiders from aerial [17] traps from two locations (wheat and grass fields respectively) near the Seale-Hayne Field Station, Devon, UK and compared endosymbiont infection frequencies in these spiders with those obtained from suction samples of the ground population at the same locations. Comparisons between aerial and ground samples must be interpreted carefully given the potentially larger action radius of aerial movers when compared with ground movers. Nonetheless, it is interesting that an analysis of the combined data indicates a significantly larger proportion of Rickettsia-infected females in ground samples (air 8%, ground 25%, exact test, P = 0.044). This is driven by a highly significant difference at the first site (air 11%, ground 83%, exact test, P = 0.002) with a larger (although not significantly so) proportion at the second site (air 7%, ground 15%, P = 0.32).
We detected no significant differences between aerial and ground samples for any of the other bacteria (Wolbachia, Spiroplasma, Cardinium) and no differences between aerial and ground males at either site for Rickettsia or any of the microbes. We observed that the distribution of individuals between aerial and ground samples varied significantly between the sexes. Males were almost twice as likely as females to be found in aerial traps (77% of the males at site 1 were found in aerial traps vs. only 43% of the females, χ
2 = 6.4, P = 0.011; at site 2 this was 72% of males vs. 29% of females, χ
2 = 30.16, P < 0.0001).