Common endosymbionts affect host fitness and sex allocation via egg size provisioning

Abstract

It is hard to overemphasize the importance of endosymbionts in arthropod biology, ecology and evolution. Some endosymbionts can complement host metabolic function or provide defence against pathogens; others, such as ubiquitous Wolbachia and Cardinium, have evolved strategies to manipulate host reproduction. A common reproductive manipulation strategy is cytoplasmic incompatibility (CI) between differently infected individuals which can result in female mortality or male development of fertilized eggs in haplodiploid hosts. Recently, an additional role of endosymbionts has been recognized in the modification of sex allocation in sexually reproducing haplodiploids. This was theoretically expected due to the maternal inheritance of endosymbionts and natural selection for them to increase infected female production, yet the underlying mechanism remained unknown. Here, we tested whether and how Cardinium and Wolbachia causing different CI types interact to increase female production in a haplodiploid thrips species where sex allocation depends on both maternal condition and egg size provisioning. We found that Cardinium augmented female production by increasing maternal fitness and egg size, thereby boosting fertilization rate and offspring fitness. Wolbachia, in contrast, reduced the beneficial effects of Cardinium. Our results demonstrate different invasion strategies and antagonistic effects of endosymbiotic bacteria on host fitness and evolution of sex allocation.

1. Introduction

Maternally inherited endosymbionts are common and play pivotal roles in arthropod biology, ecology and evolution [1–3]; several manipulate host reproduction and behaviour, and can drive the evolution of mating systems [4,5]. Wolbachia and Cardinium occur in 50% and 13% of all arthropod species [5] and can influence host fitness and life-history traits in diverse ways, depending on their own and the hosts' genotypes, as well as ecological and environmental factors [6]. Endosymbiont effects in hosts can vary from deleterious to beneficial and are particularly well studied for Wolbachia [5]. While positive fitness effects drive the spread of beneficial endosymbionts in host populations [5], some maternally inherited endosymbionts, such as Wolbachia and Cardinium, have also evolved a variety of other strategies to promote their invasion of host populations, by causing infected females to produce more (infected) daughters than uninfected females, even when in conjunction with detrimental host fitness effects [6–8].

Since Wolbachia and Cardinium are maternally transmitted within host eggs, their spread in and across host populations requires an increased production of infected over uninfected females [1,9,10]. This is generally achieved by manipulation of host reproduction [11,12], with cytoplasmic incompatibility (CI) being the most common reproductive manipulation [13,14]. In its simplest form, CI occurs when eggs of an uninfected female are fertilized with sperm of an infected male [1]. In diplodiploid hosts, CI leads to increased mortality of embryos irrespective of their sex [12], whereas in haplodiploids, CI can result in fertilized eggs experiencing either female mortality (FM-CI) or conversion of fertilized embryos to viable haploid males, also known as ‘male development’ CI (MD-CI) [15,16]. Due to endosymbiont-induced CI, infected females have a fitness advantage over uninfected females because they can produce female offspring upon mating with both infected and uninfected males. This also increases endosymbiont prevalence in populations, and, therefore, CI is an effective mechanism for endosymbionts to spread.

Models evaluating endosymbiont invasion dynamics suggest that CI is a less efficient drive mechanism in haplodiploid than in diplodiploid hosts, because of the production of uninfected males by incompatible matings [17]. Therefore, in haplodiploids, a higher proportion of infected females is required for an endosymbiont to rise above the unstable equilibrium threshold in order to spread [17,18]. This is predicted to be even more crucial for MD-CI because the higher production of uninfected sons reduces the likelihood of incompatible mating events and, therefore, the frequency of infected females. Thus, endosymbionts, in particular the ones inducing MD-CI (with a higher unstable equilibrium threshold for invasion), may have evolved complementary strategies, like provision of fitness benefits, to boost their invasion success and persistence in haplodiploid populations. For example, Wolbachia can increase fitness of hosts by conferring resistance to viral infections [19] or increasing reproductive lifespan and fecundity [5].

Another adaptive strategy of endosymbionts may be to bias the host's sex allocation towards daughters even in compatible matings. A few studies have shown that Wolbachia, in diplodiploids, can bias the primary sex ratio towards females by impacting sex chromosome inheritance [20] and, in haplodiploids, by increasing the fertilization rate [17]. For example, Wolbachia-infected females produce more female offspring than uninfected females in the spider mite Tetranychus urticae [21] and the parasitoid wasps Leptopilina heterotoma [18] and Nasonia vitripennis [22]. Similarly, Rickettsia endosymbionts can promote host fitness and increase daughter production in the whitefly Bemisia tabaci [23]. Furthermore, Hamiltonella and Arsenophonus endosymbionts also bias offspring sex ratio towards females in whitefly species [24,25]. Therefore, increasing the fertilization rate as an invasion strategy may have evolved independently in diverse maternally inherited endosymbionts, yet the underlying mechanisms by which endosymbionts achieve this remain unknown.

Remarkably, females of haplodiploid species can control their offspring sex ratio by regulating sperm access to eggs, with fertilization resulting in diploid females while haploid males develop from unfertilized eggs [26]. It has been shown that females assign fertilization and female development to larger eggs where fitness increments associated with size have greater investment returns to mothers [27,28]. Thus, given that female-destined eggs typically require more nutrients for an increased fertilization rate and juvenile survival [28], we hypothesized that endosymbionts may affect sex allocation of haplodiploids by influencing resource investment in eggs thereby ensuring a higher fertilization rate.

We have previously established Kelly's citrus thrips, Pezothrips kellyanus, as a haplodiploid study system to investigate the evolution of sex allocation and, here, investigate interactions of its allocation system with endosymbionts. This species has an egg size-mediated fertilization mechanism, i.e. larger eggs are more likely to be fertilized and develop into fitter offspring [28]. Furthermore, maternal condition can influence reproductive strategies, as smaller females keep all or most of their eggs unfertilized, producing less costly sons, despite carrying viable spermatozoa in their spermathecae [29]. This results in a bimodal sex allocation pattern in mated females, with larger females producing strongly female-biased broods (F broods), and smaller females producing strongly male-biased broods (M broods) [29]. Beyond this, P. kellyanus is naturally co-infected with both Cardinium and Wolbachia [30] which independently cause different modes of CI, i.e. moderate MD-CI observed in Cardinium and FM-CI observed in Wolbachia [31]. Furthermore, these bacteria appear to increase the egg fertilization rate in P. kellyanus, since mated females co-infected with both bacteria are more likely than uninfected females to produce F broods [32]. However, it remains to be investigated whether these two endosymbionts interactively facilitate higher fertilization rates in haplodiploid hosts. This requires disentangling how the two endosymbionts affect host reproduction, sex allocation and fitness. Here, we tested whether Cardinium and Wolbachia affect sex allocation in P. kellyanus by increasing maternal and/or paternal fitness and thereby altering the fertilization rate. Moreover, we analysed the interactions between the two endosymbionts in affecting host reproduction and offspring fitness traits.

2. Methods

(a) . Establishment of thrips lines and experimental cohorts

A laboratory population infected with both Cardinium and Wolbachia (I_CW) was established with P. kellyanus collected in Kulnura, New South Wales (Australia) in 2017 [28]. An uninfected line (U) and a Cardinium-only line (I_C) were obtained by antibiotic treatment as previously described [31]. Our attempts to obtain a Wolbachia-only line (I_W) were unsuccessful (electronic supplementary material, figure S1). The fitness assays were conducted seven to eight generations after antibiotic treatment. The host genetic background was homogenized among the three lines by crossing virgin females of I_CW and I_C with U males for two successive generations. While this might not have completely homogenized the host genetic background, we also emphasize that all experimental lines were derived from the same inbred laboratory population (electronic supplementary material). All lines were maintained on organic lemons sprinkled with Typha sp. pollen under controlled conditions (20 ± 1°C, 70 ± 2% relative humidity and 16 : 8 h light : dark). Two generations before the experiment, population densities and sex ratios were controlled at the beginning of each generation [N = approximately 300; 1 : 1 (♂ : ♀)] by removal of excess individuals. The infection status was confirmed by diagnostic PCR (electronic supplementary material, table S1) for 20 individuals per line before the experiment, and for all individuals of the fitness assays at the end of the experiment. To ensure sufficient numbers of replicates, the experiment was run in three cohorts commencing on three consecutive days (electronic supplementary material, figure S2). From each cohort, we established equally aged females and males for fitness assays as previously described [28]. All individuals were reared at the same larval density to ensure similar developmental conditions for individuals used in the fitness assays (electronic supplementary material).

(b) . Fitness assay

For each line some females (1–2 days old) were randomly selected and kept virgin; others were individually placed into small Petri dishes, each with 10 virgin males (2–3 days old) of the same infection type (U × U, I_C × I_C, I_CW × I_CW). We controlled for parental relatedness by pairing males and females produced by different mothers (electronic supplementary material, figure S3). Females were mated once and copulation duration was recorded. Females were immediately removed to avoid further matings. Each mated male was preserved in a 1.5 ml microcentrifuge tube and frozen at −20°C for subsequent size measurement and infection screening. The average pre-oviposition period of P. kellyanus was 6 days, independent of mating and infection status (electronic supplementary material). Therefore, to induce simultaneous oviposition, the virgin and once-mated females were isolated individually in small Petri dishes for at least 6 days to ensure simultaneous oviposition. During this period, females were supplied with Typha sp. pollen and honey water (50% w/v) every second day. Then, females were allowed to oviposit on an agarose plate for 24 h [28]. Females of each subsequent cohort were allowed to lay eggs with 1 day delay compared to the previous cohort (electronic supplementary material, figure S2). Then the egg number, egg size and sex ratio of the first clutch (early reproductive effort) of each female were assessed together with other maternal fitness traits (see below). Females tested for each line were categorized into three types depending on their mating status and offspring sex ratio: virgins (V); mated females producing M broods [M(♂)]; and mated females producing F broods [M(♀)]. Accordingly, their offspring were categorized into four types depending on their sex and mother type (i.e. one female offspring type and three male offspring types).

(c) . Sperm detection in spermathecae

Ten newly emerged virgin females were randomly selected from each line and individually placed into small Petri dishes with virgin males of the same line. Mating was observed. Then, mated females were individually isolated in small Petri dishes with pollen and honey water for at least 6 days. Subsequently, females were individually kept in small containers containing a lemon sprinkled with pollen for oviposition for 24 h. Subsequently, females were removed, immobilized on ice for 10 s and their spermathecae dissected, transferred to microscope slides with a drop of Ringer's solution and topped with a coverslip. The presence of motile sperm cells was verified using a phase-contrast light microscope under a 100 X/1.25 oil-immersion objective (Zeiss Axio Scope A1, Germany). The containers with the lemons were kept for offspring development and subsequent sex ratio assessment.

(d) . Effect of endosymbionts on egg size and fecundity

Five to eight eggs per female (approx. 50% of eggs oviposited) were randomly sampled. Egg volume was measured and calculated following Katlav et al. [28]. Subsequently, the egg volume was compared among all three lines (electronic supplementary material). To assess the early reproductive effort of mothers, all egg size measurements were included in the analysis regardless of offspring survival and sex.

(e) . Endosymbionts and egg size-mediated sex allocation

After size measurement, each egg was immediately placed on an agarose gel plug in a 0.5 ml microcentrifuge tube and incubated until it hatched (electronic supplementary material, figure S4A). Once hatched, neonate larvae were individually transferred to a 1.5 ml microcentrifuge tube supplied with a lemon leaf disc and pollen, both renewed every second day (electronic supplementary material, figure S4B). Individual offspring were monitored daily until adulthood. Offspring sex was determined at the pupal stage, based on a distinctive sexual size dimorphism [28]. All virgins produced only male offspring, confirming arrhenotoky in P. kellyanus [31]. The two mated mother types [M(♂) and M(♀)] were distinguished according to their offspring sex ratio. The proportion of M(♂) and M(♀), as well as the population-level offspring sex ratio of mated females, were compared among the three lines.

(f) . Effect of endosymbionts on offspring fitness traits

Fitness traits of all offspring types were assessed among the lines. For this, larval development and survival from egg to adulthood were recorded daily. After emergence, adults were transferred into other 1.5 ml tubes supplied with a lemon leaf disc and pollen, both renewed every second day, but without honey water (imposing dehydration stress), and monitored daily for survival. Upon adult death, forewing length, a reliable proxy of body size [28,29], was measured.

(g) . Comparison of resource allocation between co-infected and naturally uninfected mothers

To ensure that recorded parameters in resource allocation of uninfected individuals of the U line (which was obtained by antibiotic treatment of the I_CW line) is not an outcome of potential long-term effects of microbiome disruption due to antibiotic treatment, another experiment was conducted with individuals of a naturally uninfected line (NU; obtained from the same field population as I_CW) and I_CW to compare their egg number and size (electronic supplementary material).

(h) . Statistical analyses

All statistical analyses were performed using R (v. 3.5.1) [33] and SPSS (v. 25, IBM Corporation, USA). The distribution of variables was checked for normality using the Shapiro–Wilk test and the homogeneity of variance using Levene's test (p < 0.05). Data violating normality and homogeneity assumptions were log10-transformed. A chi-square test was used to examine the effect of infection on the fertilization rate. A Kruskal–Wallis test was used to evaluate the effect of infection on the population-wide sex ratio for each infection line. Analysis of egg volume and number among different treatments was performed using a general linear model (three-way ANOVA) using the lme4 package with three fixed factors (infection type, mother type/offspring type, experimental cohort) and mother ID as a random factor to control for the variation between individual mothers. An analysis with a similar model structure was conducted for the offspring fitness data (development time, juvenile survival rate, adult forewing length and longevity) with the same fixed and random factors. In addition, mother forewing and egg size (depending on the response variable) were incorporated in the model as a covariate (ANCOVA) to determine the net effects of fixed factors. Following a significant effect of factors or their interactions, additional post hoc tests for multiple comparisons were conducted with Tukey's HSD test (p < 0.05) to determine which treatment differed within the fixed factor of interest. A three-way ANOVA table for the treatments' structure was constructed. In cases of no cohort effects and no interactions with other factors, the respective figures are represented with data of all experimental cohorts combined.

3. Results

(a) . Endosymbionts affect egg fertilization rate and host sex allocation

Comparison of the sex ratio among the three infection types revealed that the proportion of F brood production was higher for infected females (figure 1). Further comparison between mated I_C and I_CW females revealed that this effect was due to Cardinium and not Wolbachia. Mated I_C and I_CW females had a higher proportion of M(♀) when compared to mated U females (electronic supplementary material table S2; figure 2a). Given that the proportion of M(♂) in I_CW did not differ significantly from that in I_C (electronic supplementary material, table S2), it can be concluded that the increased fertilization rate is due to Cardinium only. Further analysis of the sex ratio at the population level revealed higher female-biased population sex ratios in mated I_CW (approx. 39%) and mated I_C (approx. 36%) when compared to mated U (approx. 61%) (electronic supplementary material, table S2).

Figure 1.

Proportion of mated females with female-only or highly female-biased (F) broods [M(♀)] versus mated females with male-only or highly male-biased (M) broods [M(♂)] across pairs with different infection types. (Online version in colour.)

Figure 2.

(a) Mean (±s.e.) egg volume across different infection (U, I_CW, I_C) and mother (virgin V, mated M) types. (b) Egg size range variation (±s.e.) across different infection and mother types. (c) Mean (±s.e.) egg volume across different infection and offspring (son S, daughter D) types. Different lowercase letters denote statistically significant differences between mother and offspring types within the corresponding infection type (p < 0.05). Asterisks denote statistically significant differences between infection and mother types (***p < 0.001; **p < 0.01; n = number of eggs examined). (Online version in colour.)

A separate analysis of the offspring sex ratio of M(♀) revealed a significant effect of infection type (electronic supplementary material, table S3). I_C M(♀) females had higher female production than U M(♀) females. However, this effect was removed by Wolbachia in I_CW M(♀) females because their offspring sex ratio was not significantly different from U M(♀) females (electronic supplementary material, figure S5).

(b) . Endosymbionts affect maternal fitness but not paternal fitness and sperm transfer

The copulation time of individual pairs of the same infection type was similar among the three infection types and the two mated mother types (electronic supplementary material, figure S6a), indicating that copulation time does not drive the observed differences in fertilization rates. We also did not detect an effect of mother or father body size on copulation time (electronic supplementary material, table S4). However, we detected an effect of mother size, but not father size on sex allocation (electronic supplementary material, table S4), because M(♀) females were larger than M(♂) females across all infection types (electronic supplementary material, figure S6b). Moreover, infected mothers were larger than uninfected mothers, yet I_C and I_CW mothers were similar in size (electronic supplementary material, figure S6a; electronic supplementary material, table S4); this effect was not seen for fathers (electronic supplementary material, figure S6c and table S4). Motile sperm was detected in the spermathecae of all mated females irrespective of infection type and offspring sex ratio. Therefore, the higher proportion of M(♂) females among U pairs is unlikely to be due to a sperm transfer failure.

(c) . Endosymbionts affect egg size and fecundity

(i) . Egg volume among mother types

Mother type significantly affected egg size because M(♀) females produced larger eggs than M(♂) females (suggesting an egg size effect on fertilization rate) and V females (suggesting a mating effect on egg size) (electronic supplementary material, table S3). In addition, egg size increased with endosymbiont infection (figure 2a) as eggs of I_C females were, on average, 7.5% larger than those of I_CW females, and 11% larger than those of U females. These results indicate a positive effect of Cardinium on egg size, which is reduced by Wolbachia in females infected with both endosymbionts (electronic supplementary material, table S3). Incorporation of mother forewing length as a covariate in the model did not remove the effect of infection type on average egg size (electronic supplementary material, table S4), suggesting that egg size is not only a function of mother size, but also directly influenced by endosymbiont infection.

There were no significant differences in egg size range (egg size_max − egg size_min) variation across mother and infection types (figure 2b; electronic supplementary material, table S4), suggesting consistent effects of mating and endosymbionts. These consistent effects were also evident in the analysis of minimum and maximum egg size, with mother type and infection type affecting both parameters (electronic supplementary material, table S4), albeit the minimum egg size difference between I_CW and I_C pairs was marginally insignificant (F_2,123 = 2.29; p = 0.058).

(ii) . Egg volume among offspring types

Eggs in M(♀) females showed sexual size dimorphism, irrespective of infection type (F_2,440 = 0.54; p = 0.578), as larger eggs developed into female offspring, whereas smaller eggs developed into male offspring (F_1,440 = 81.76; p < 0.001). In addition, egg size was very similar among all male offspring types (F_2,606 = 3.34; p = 0.071) and independent of infection type (F_2,606 = 0.015; p = 0.998) (figure 2c). Moreover, infection types affected egg volume of all offspring types, with I_C offspring emerging from the largest eggs, and I_CW offspring emerging from larger eggs than U offspring (figure 2c; electronic supplementary material, table S3).

(iii) . Egg number (clutch size)

U females produced larger first egg clutches than infected ones and this pattern was consistent across all mother types (figure 3; electronic supplementary material, table S3). However, this was not consistent across all experimental cohorts, with the third cohort showing no significant clutch size difference among infection lines (F_2,62 = 1.56; p = 0.219) (figure 3). When the average egg size was included as a covariate in the model, there was no effect of infection type on the clutch size (electronic supplementary material, table S3), indicating that the endosymbiont effect on egg size is traded-off with clutch size. Moreover, we found that independent of infection type (F_2,183 = 1.56; p = 0.289), clutch size and average egg size were negatively correlated (F_1,183 = 11.13; p = 0.0011, r² = 0.11) with a similar slope (electronic supplementary material, figure S7). This confirms that the egg size versus clutch size trade-off does not depend on infection type (average egg size × Infection type: F_2,183 = 0.28; p = 0.75).

Figure 3.

Mean (±s.e.) clutch size (number of eggs) across three cohorts with different infection types (U, I_CW, I_C; n = number of females examined). Different lowercase letters denote statistically significant differences between infection types within the corresponding experimental cohort (p < 0.05; pooled data of the two mated mother types).

(d) . Endosymbionts affect offspring fitness traits

(i) . Offspring development time

Across all infection types, female offspring developed slower than male offspring (figure 4a). Cardinium did not affect development time compared to uninfected individuals; however, co-infection with Wolbachia significantly extended development time (figure 4a; electronic supplementary material, table S3), albeit this was marginally not significant in the second experimental cohort (F_1,222 = 3.78; p = 0.053).

Figure 4.

Offspring fitness in response to different infection types. (a) Mean (±s.e.) offspring development time (days). (b) Mean (±s.e.) juvenile offspring survival rate. (c) Mean (±s.e.) adult offspring forewing length. (d) Mean (±s.e.) adult offspring longevity (day) under dehydration stress across different infection (U, I_CW, I_C) and offspring types (son S, daughter D) of virgin (V) or mated (M) mothers with F or M broods. Different lowercase letters denote statistically significant differences between infection types (p < 0.05) in each offspring type (n = number females examined). (Online version in colour.)

(ii) . Juvenile survival rate

Compared to U, survival from egg to adulthood was 15% higher in I_C, independent of the mother type. This Cardinium effect was not changed by co-infection with Wolbachia in mated I_CW females (figure 4b). Moreover, we detected a significant effect of egg size (covariate) in the model, suggesting that the effect of endosymbionts on offspring survivorship is partially (if not entirely) exerted through egg size (electronic supplementary material, table S3).

(iii) . Offspring adult size

The effect of Cardinium on egg size persisted until offspring developed into adults irrespective of mother type, with larger eggs in I_C developing into larger adults (across all offspring types) when compared to U. However, the negative effect of Wolbachia on egg size detected in I_CW was still evident in adults, with adult I_CW offspring being significantly smaller than adult I_C offspring, yet still significantly larger than those of U (figure 4c). When egg size was included as a covariate, the effect of the endosymbionts was reduced, yet remained significant (electronic supplementary material, table S3), suggesting that effects of both endosymbionts on adult size are beyond their impact on egg size.

(iv) . Adult offspring longevity under dehydration stress

Irrespective of infection type, female offspring lived on average 0.71 day longer than male offspring. Cardinium increased adult offspring longevity by nearly 1 day, but this fitness effect did not significantly change in individuals co-infected with Wolbachia (figure 4d). This endosymbiont effect was uniform across all offspring types. However, offspring of the first cohort lived 0.37 days longer than those of the two subsequent cohorts. Incorporation of adult size as a covariate in the analysis revealed that this fitness effect is partly, if not completely, an endosymbiont effect on adult size (electronic supplementary material, table S3).

(e) . Higher reproductive effort in co-infected than naturally uninfected mothers

Consistent with the previous comparisons of the antibiotic-treated uninfected (U) and the naturally co-infected (I_CW) lines, naturally uninfected (NU) females produced significantly smaller eggs than I_CW females. Mating also significantly increased egg size, irrespective of infection type (electronic supplementary material, figure S8).

4. Discussion

Our study demonstrated that the maternally inherited endosymbionts Cardinium and Wolbachia can affect their hosts' fitness and sex allocation differently, and antagonistically when co-occurring in the same host. In P. kellyanus, Cardinium increased maternal investment in egg size, which enhanced fertilization rate and female-biased brood production. Furthermore, Cardinium increased host offspring size and survival. Such endosymbiont-moderated sex allocation and fitness benefits likely facilitate the spread and maintenance of Cardinium in host populations on top of its capacity to induce CI [31]. By contrast, the presence of Wolbachia in co-infected individuals was antagonistic to the beneficial effects of Cardinium on host fitness. Furthermore, the fitness advantage conferred by Cardinium may compensate for the lower invasion efficiency observed in this host of Cardinium-induced MD-CI than Wolbachia-induced FM-CI because MD-CI can compromise endosymbiont invasion success due to the increased production of uninfected males [17].

(a) . Cardinium-infected females produce more daughters due to increased fitness and resource investment

In haplodiploids, highly male-biased offspring sex ratios of mated females can be due to a lack of sperm in females, such as failure in mating or sperm transfer [34]. However, we confirmed the presence of motile sperm in all dissected mated females, including those that had produced only male offspring. Our study also found neither an effect of endosymbiont infection on copulation duration nor an effect of copulation duration on the brood sex ratio. Therefore, decreased daughter production by uninfected mated females cannot be due to failure of mating or sperm transfer. Moreover, unlike in several parasitoid wasp species where paternal size influences offspring sex allocation [35,36], male size in P. kellyanus did not contribute to the offspring sex ratio. In some insect species, male ejaculates contain substances that promote vitellogenesis and fertilization [37], but this (and a role of endosymbionts therein) has yet to be investigated in P. kellyanus.

Previous studies have demonstrated that, in P. kellyanus and possibly in many other haplodiploid hosts, egg size influences the fertilization rate [27,28], and this is positively correlated with maternal condition [29]. We showed that females of a line lacking endosymbionts produced smaller eggs than females of the lines infected with Cardinium. The size reduction was noticeable for all eggs and was also evident in virgin mothers. This indicates lower resource availability and egg provisioning in uninfected females compared to those with Cardinium. Lower fitness and reduced resource allocation was also seen in NU thrips, suggesting that our findings are not a potential long-term effect of microbiome disruption due to antibiotic treatment. Therefore, our study indicates that Cardinium promotes fertilization in mated females by increasing maternal resource investment towards egg size. This is consistent with several studies suggesting that maternally inherited symbionts can nutritionally provision hosts [38,39]. For example, in whiteflies, Hamiltonella and Arsenophonus supply B vitamins which probably support oogenesis and fertilization [25]. Furthermore, increased provisioning to egg size may be an adaptive response of the host to a higher resource availability provided by an endosymbiont, which then, in turn, ensures higher female sex allocation [32,40]. However, this remains to be elucidated in combination with the assessment of the contribution of host genetic variation and environmental factors in the plasticity of egg size.

(b) . Interactive effects of Cardinium and Wolbachia on host fitness

Disentangling interactions between endosymbionts in their hosts is important for understanding the evolutionary history and the population dynamics of host-endosymbiont associations [2,41]. Our study demonstrated interactive effects of Cardinium and Wolbachia on reproductive strategies and fitness traits of P. kellyanus. We showed that the effect of Cardinium on investment into egg size was higher when Wolbachia was absent, indicating a likely cost of Wolbachia. Therefore, the fitness benefit of Cardinium may partly be used by the host to counter the increased metabolic cost of bearing Wolbachia, resulting in fewer resources for egg provisioning. It has previously been shown that in P. kellyanus the Cardinium density is independent of Wolbachia infection as it remained unchanged after the removal of Wolbachia by antibiotics [34]. Therefore, it is unlikely that the Wolbachia cost derives from direct competitive interaction between the two endosymbionts.

We also found that the Wolbachia cost did not influence the offspring sex ratio. This may indicate that there is an egg size threshold for fertilization [27] that is still achieved by mated I_CW females. Reproductive fitness benefits of Cardinium have also been shown for a predatory mite, Metaseiulus occidentalis, but only in terms of a fecundity increase in females [42]. Conversely, in pharaoh ants, Wolbachia causes a higher reproductive investment and results in more female-biased sex ratios [43]. Our study demonstrates this also for Cardinium. However, we are not aware of other studies showing that Cardinium facilitates fertilization and increases daughter production by increasing host investment in egg size, yet this could be a common strategy.

Increased egg size investment by Cardinium-infected females came at the expense of a smaller clutch size, suggesting a trade-off between egg size and number [28,44]. This can be related to maternal condition that limits how much a mother can invest into her offspring [27,45], and thus, an increase in egg size must be compensated by a decrease in egg number. Our finding shows that uninfected mothers have lower fitness (smaller bodies) and, therefore, are selected to produce a larger number of smaller eggs [29]. Without egg size measurement, we would have erroneously concluded that Cardinium imposed a cost in terms of lower reproductive output. Importantly, egg number alone cannot be an accurate proxy of endosymbiont effects on reproductive outputs. The fecundity pattern in uninfected mothers was inconsistent across experimental cohorts, with a smaller clutch size in the last cohort, possibly because of the 2 days delay in oviposition associated with our experimental design. Other studies in parasitoid wasps have attributed a lower fecundity due to an oviposition delay to resorption of eggs as a female strategy of biological maintenance [46]. However, clutch size did not change with the oviposition delay in neither I_CW nor I_C, again indicating a beneficial effect of Cardinium.

Consistent with a previous study on P. kellyanus, which did not investigate endosymbiont effects [28], offspring fitness increased with egg size. Egg size increase due to Cardinium resulted in offspring with a higher survival rate, larger adult size and increased longevity under dehydration stress. However, we also found direct beneficial effects of Cardinium irrespective of egg size. Body size is an important indicator of the reserved energy budget in animals [47], with larger individuals of both sexes often enjoying higher reproductive success [48,49] and survival [50]. Furthermore, body size is an important indicator of mating success in the context of sexual selection [51]: males may prefer to mate with larger females as they are more fecund [52], and larger males may be better competitors in acquiring mates [53]. Several studies have shown positive Wolbachia effects on adult body size and survival [54,55]. However, we only found such benefits for Cardinium. By contrast, analysis of co-infected hosts suggests that the effects of Wolbachia were either costly or neutral for the host. Therefore, it is possible that Wolbachia, other than its CI induction, may profit from the beneficial fitness effects of Cardinium to invade host populations. To our knowledge, studies demonstrating this are scarce [56–58]. One example involved the parasitoid wasp Encarsia inaron, in which Cardinium alone can increase adult survival, while co-infections with Wolbachia can mask this effect [57].

Additionally, we found that Wolbachia prolonged development time from egg to adult emergence in all offspring types. Such a Wolbachia-caused developmental delay has also been shown in mosquitoes [59,60]. In our study, slower development might be an adaptive response of Wolbachia-infected thrips to obtain more food during development, compensating for a cost at the embryonic stage, and maintaining the fitness advantage conferred by Cardinium on survival and adult longevity. Conversely, slower development could constitute a negative fitness aspect, delaying maturation and reducing population growth rate [61]. A prolonged development time may also increase the mortality rate due to risks of predation and parasitism in field populations [62]. Therefore, I_CW thrips may be less competitive than I_C.

Finally, the more beneficial role of Cardinium may be an evolutionary compensatory response to excessive male production due to MD-CI induction in P. kellyanus [31] which would reduce its invasion efficiency [17]. Such a compensatory role may be less important for Wolbachia invasion of P. kellyanus populations, given that compared to Cardinium, Wolbachia induces a more invasion-efficient CI type (FM-CI) in this host [31]. Moreover, it has also been hypothesized that a costly symbiont can hitchhike with a more beneficial partner and thereby invade a host population more efficiently [63]. Our study suggests that Wolbachia may fall into this category, justifying its considerable prevalence (although lower than Cardinium) in P. kellyanus populations [30]. This finding may also explain why Wolbachia-only infected P. kellyanus are almost absent in nature [30], despite the expression of complete CI and high transmission fidelity in laboratory experiments [31]. A similar scenario of a parasitic endosymbiont supported by a beneficial endosymbiont has been demonstrated for the cost-inflicting X-type endosymbiont, Fukatsuia symbiotica [64], which can only persist in pea aphid populations in co-infection with the beneficial endosymbiont Hamiltonella [65].

(c) . Lower prevalence of Wolbachia than Cardinium in field populations of Pezothrips kellyanus

The factors contributing to a lower prevalence of Wolbachia (60–90%) than Cardinium (approx. 100%) in several native (Australian) field populations of P. kellyanus and the complete absence of Wolbachia in independently colonised invasive ranges (New Zealand and Mediterranean Region) are still unclear [30]. However, the findings of this study may attribute the lower prevalence of Wolbachia at least partly to its fitness costs revealed here. Together with a likely severe population bottleneck and/or the yet unknown identity of the source populations within Australia, the revealed Wolbachia fitness cost may have contributed to its absence in the invasive ranges of P. kellyanus. Future investigations of endosymbiont prevalence and host genetic diversity should address this further.

5. Conclusion

Our study revealed that Cardinium and Wolbachia affect host fitness and sex allocation, and this may contribute to their invasion success beyond the induction of CI which had been suggested earlier in haplodiploid hosts [17]. In P. kellyanus, Cardinium biases host sex allocation towards females by affecting the egg size-mediated fertilization system. It achieves this by increasing host fitness, leading to greater maternal resource allocation towards egg size, which subsequently affects offspring fitness. Wolbachia appears to act antagonistically on the fitness benefit provided by Cardinium, which may be linked to why it has a lower prevalence in native populations, has not been found as a single infection and is absent from some invasive populations of P. kellyanus [30]. Finally, egg size-mediated fertilization might be an outcome of endosymbiont–host genetic conflicts over the sex ratio [66]. Therefore, our findings direct future studies towards understanding the evolutionary forces driving egg size and sex allocation in haplodiploids which constitute a large proportion of biodiversity infected by endosymbionts.

Data accessibility

The data supporting the findings of this study are available within the electronic supplementary material [67] and at https://doi.org/10.6084/m9.figshare.17032025.

Authors' contributions

A.K.: conceptualization, data curation, software, validation, visualization, writing—original draft and writing—review and editing; J.M.C.: conceptualization, supervision and writing—review and editing; M.R.: conceptualization, funding acquisition, project administration, supervision and writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

The authors declare no conflict of interest.

Funding

This project was supported by a Western Sydney University Postgraduate Research Award and a F.G. Swain Award of the Hawkesbury Foundation to A.K., and an Australia and Pacific Science Foundation research grant (grant no. APSF16/5) to M.R. and J.M.C.