Competitive Exclusion of Phytopathogenic Serratia marcescens from Squash Bug Vectors by the Gut Endosymbiont Caballeronia

ABSTRACT

Many insects harbor microbial symbiotic partners that offer protection against pathogens, parasitoids, and other natural enemies. Mounting evidence suggests that these symbiotic microbes can play key roles in determining infection outcomes in insect vectors, making them important players in the quest to develop novel vector control strategies. Using the squash bug Anasa tristis, we investigated how the presence of Caballeronia symbionts affected the persistence and intensity of phytopathogenic Serratia marcescens within the insect vector. We reared insects aposymbiotically and with different Caballeronia isolates, infected them with S. marcescens, and then sampled the insects periodically to assess the intensity and persistence of pathogen infection. Squash bugs harboring Caballeronia consistently had much lower-intensity infections and cleared S. marcescens significantly faster than their aposymbiotic counterparts. These patterns held even when we reversed the timing of exposure to symbiont and pathogen. Taken together, these results indicate that Caballeronia symbionts play an essential role in S. marcescens infection outcomes in squash bugs and could be used to alter vector competence to enhance agricultural productivity in the future.

IMPORTANCE Insect-microbe symbioses have repeatedly been shown to profoundly impact an insect’s ability to vector pathogens to other hosts. The use of symbiotic microbes to control insect vector populations is of growing interest in agricultural settings. Our study examines how symbiotic microbes affect the dynamics of a plant pathogen infection within the squash bug vector Anasa tristis, a well-documented pest of squash and other cucurbit plants and a vector of Serratia marcescens, the causative agent of cucurbit yellow vine disease. We provide evidence that the symbiont Caballeronia prevents successful, long-term establishment of S. marcescens in the squash bug. These findings give us insight into symbiont-pathogen dynamics within the squash bug that could ultimately determine its ability to transmit pathogens and be leveraged to interrupt disease transmission in this system.

INTRODUCTION

Vector-borne diseases have posed significant threats to agricultural productivity as well as human and wildlife health for millennia (1, 2). Controlling populations of insect vectors has traditionally played a key role in human efforts to reduce the burden of vector-borne diseases, often to great effect (3). However, the negative environmental impacts associated with insecticides (4), the most common type of vector control, and rising incidences of pesticide resistance (5) have spurred a search for alternative control measures. The successful introduction of Wolbachia endosymbionts into Aedes aegypti mosquitoes for the control of dengue virus (6) highlights the potential for the development of other symbiotic microbes to control insect vector populations and the pathogens they transmit, a tactic known as symbiont-mediated vector control (7, 8).

While factors such as temperature (9, 10), host genetic background (11, 12), and innate immunity (13) have long been known to influence vector competence, i.e., an insect’s ability to acquire, maintain, and transmit pathogens (14), mounting evidence suggests that symbiotic microbes can play key roles in determining pathogen infection outcomes in insect vectors (15, 16). Interactions between insects and pathogens do not happen in isolation; many insects harbor microbial symbiotic partners that offer protection against insect pathogens, parasitoids, and other natural enemies (17–20). These microbial partners also interact with vectored pathogens in a myriad of ways that can either facilitate or inhibit infection. In some insects, for example, proteins produced by symbionts are essential for supporting parasite survival (21, 22), increasing vector competence. In other cases, symbionts decrease insects’ vector competence by interfering with the establishment of vectored pathogens (23, 24) or by promoting proper immune system development in their insect host and enabling a more robust response to subsequent pathogen infections (25). In fact, some vectored pathogens must actively disrupt the natural insect microbiota to successfully establish themselves (26).

A symbiont’s ability to coexist with or competitively exclude coinfecting pathogens within its insect host can dictate pathogen infection outcomes that can affect vector competence. Knowledge of these dynamics is an important early step in assessing the potential for symbiont-mediated vector control. Here, we explore how symbiotic microbes affect the dynamics of a phytopathogen infection in the squash bug (Anasa tristis DeGeer), an insect pest of agricultural importance. The squash bug is the primary vector of cucurbit yellow vine disease (CYVD) (27), caused by phytopathogenic lineages of Serratia marcescens (28, 29). Although other insects have been shown to acquire S. marcescens in artificial feeding systems, the squash bug is the only insect confirmed to transmit S. marcescens in the field (30). CYVD leads to significant yield losses in squash, pumpkin, and related crops (31). The phytopathogenic lineages of S. marcescens, unlike their entomopathogenic counterparts, have limited influence on insect fitness when ingested (see Fig. S1 in the supplemental material).

Like other stink bugs and their relatives, squash bugs form symbiotic associations with bacteria in the genus Caballeronia, formerly contained within the genus Burkholderia (32–34). Cabelleronia symbionts are acquired de novo from the environment at each host generation, allowing us a straightforward way to manipulate symbiont acquisition. Symbionts are housed in a specialized region of squash bugs’ posterior midgut, known as the crypts, and typically are acquired early in the insect life cycle (35). Previous work has established that successful host colonization by Caballeronia results in accelerated development and decreased mortality relative to aposymbiotic (symbiont-free) individuals (36). Leveraging the natural characteristics of this system, we investigated the impact of symbiont colonization on the intensity and persistence of S. marcescens infections in squash bugs and looked for evidence of priority effects. We provide evidence that the symbiont Caballeronia prevents successful, long-term establishment of S. marcescens in its A. tristis vector, regardless of whether it is the first to colonize the insect or not.

RESULTS

We conducted two experiments to determine how Caballeronia symbionts interacted with phytopathogenic S. marcescens. First, we determined whether the presence of Caballeronia symbionts influenced the outcome of S. marcescens infection in squash bugs. We then tested for priority effects by varying the order of exposure to pathogen and symbiont.

Symbionts reduce the persistence and intensity of pathogen infections in squash bugs.

We found a clear effect of Caballeronia symbiont colonization status on both the persistence and intensity of S. marcescens (strain Z01) infection (Fig. 1). Nearly all aposymbiotic bugs retained Z01 infection throughout the experiment, resulting in no temporal trend in pathogen prevalence among aposymbiotic bugs (generalized linear model [GLM] coefficient for days postacquisition [DPA] in aposymbiotic bugs of 0.02, standard error [SE] = 0.21, P = 0.91). There were, however, significant declines in the prevalence of infection among symbiont-positive bugs over time (GLM coefficient for DPA in symbiont positive bugs of −0.65, SE = 0.12, P = 4.09 × 10⁻⁸). Furthermore, prevalence of infection with Z01 declined significantly faster in symbiont-positive squash bugs relative to aposymbiotic bugs (difference in slope of −0.68, SE = 0.24, P = 5 × 10⁻³) (Fig. 1a).

FIG 1.

Symbiont status affects persistence and intensity of Serratia infection. (a) The proportion of individual bugs that tested positive for S. marcescens at a given time point across aposymbiotic and symbiont-positive treatments. For the aposymbiotic treatment, each point represents 15 squash bugs except for 9 DPA, which represents 10 individuals. For the symbiont-positive treatment, each point represents 20 squash bugs. (b) S. marcescens titer recovered from infected squash bugs over time. These data include only nonzero values. Points represent up to 15 individuals for the aposymbiotic treatment or up to 20 for the symbiont-positive treatment, with later time points typically representing fewer individuals, particularly in symbiont-positive bugs. Shading in both panels indicates 95% confidence intervals where possible to calculate.

Among squash bugs that were infected, those harboring Caballeronia symbionts had initial Z01 titers that were 1,000-fold lower than aposymbiotic bugs (difference in GLM intercept from aposymbiotic bugs of −4.13, SE = 0.36, P < 2 × 10⁻¹⁶). Differences in titer between the two groups persisted over time, with symbiont-positive bugs retaining significantly lower pathogen titers (difference in DPA coefficients of −0.74, SE = 0.09, P = 2.76 × 10⁻¹⁵). While symbiont-positive squash bugs experienced significant reductions in infection intensity over time (GLM coefficient for DPA in symbiont positive bugs of −0.67, SE = 0.07, P < 2 × 10⁻¹⁶), we did not identify a significant temporal trend in Z01 titer for aposymbiotic bugs (GLM coefficient for DPA in symbiont negative bugs of 0.07, SE = 0.07, P = 0.30) (Fig. 1b).

Accounting for symbiont strains and batches of aposymbiotic bugs showed similar patterns of Z01 clearance between aposymbiotic and symbiont-positive bugs, with no significant differences among different symbiont strains or among batches of aposymbiotic bugs (see Fig. S2a in the supplemental material). Variation was more apparent in our infection intensity data, particularly in initial titers. These differences largely disappeared over time; however, Apo3 remained somewhat anomalous within its treatment group. In the case of Apo3, the difference was driven by the fact that Apo3, which began at a lower titer than the other two aposymbiotic batches, rose in titer to a similar level during the course of the experiment. Individual differences in infection intensity between symbiont strains could be due to the drop in sampling power as bugs began clearing Z01 infections. However, it must be noted that, despite some statistically significant differences among Caballeronia strains, these differences were rather small, typically less than 10-fold, compared to the 1,000-fold differences between symbiont positive and aposymbiotic bugs (Fig. S2b).

Time and symbiont status were strong predictor variables in our model of persistence data (full model, R² = 0.62; time model, R² = 0.26; treatment model, R² = 0.34), whereas symbiont status was by far the strongest predictor variable in our data on infection intensity (full model, R² = 0.63; time model, R² = 0.04; treatment model, R² = 0.53). Despite not accounting for specific symbiont strains or batch-to-batch variation among squash bugs, our model with consolidated treatments still explained most of the variation in our Z01 infection intensity (R² = 0.63) and persistence (R² = 0.62) data.

Symbionts retain pathogen clearing properties regardless of order of exposure.

After establishing the effect of symbiont presence on Z01 infection intensity and persistence, we investigated whether these dynamics would persist if the order of exposure to symbiont and pathogen were reversed. Although initial Z01 infection prevalence was lower in our symbiont-first treatment than in either the pathogen only or pathogen first treatments, we found that both treatments that received symbionts, regardless of order, cleared Z01 faster than the pathogen only treatment in which bugs remained aposymbiotic (symbiont first, difference in DPA coefficient from aposymbiotic treatment of −0.58, SE = 0.24, P = 0.02; pathogen first, DPA coefficient from aposymbiotic treatment of −0.46, SE = 0.17, P = 6.6 × 10⁻³) (Fig. 2a). Despite differences in initial prevalence, we found that the rate of clearance did not differ significantly between symbiont-first and pathogen-first treatments (difference in DPA coefficient of 0.12, SE = 0.28, P = 0.67).

FIG 2.

Order of infection does not impact the effect of symbiont colonization on pathogen persistence and intensity. (a) The proportion of individual bugs that tested positive for S. marcescens at a given time point. Pathogen-only bugs remained aposymbiotic, but bugs in the pathogen-first and symbiont-first treatment were coinfected with symbiont and pathogen. Each point represents 10 individuals tested at that time point. (b) S. marcescens titer recovered from infected squash bugs over time. Only nonzero values were included in this analysis. Points represent up to 10 individuals in each treatment, with later time points typically representing fewer individuals. In both panels, shading indicates 95% confidence intervals where possible to calculate.

When we looked at Z01 infection intensity (Fig. 2b), we found that both treatments that received symbionts, regardless of order, differed significantly from the pathogen only treatment in initial titers (symbiont first, difference in intercept from aposymbiotic of −4.45, SE = 0.53, P < 2 × 10⁻¹⁶; pathogen first, difference in intercept from aposymbiotic of −2.03, SE = 0.39, P = 2.55 × 10⁻⁵). The symbiont-first treatment had the lowest initial titers, consistent with findings from our previous experiment. These treatments also differed significantly from the pathogen only treatment in their rate of pathogen loss over time (symbiont first, difference in DPA coefficient from aposymbiotic treatment of −1.20, SE = 0.23, P < 2.99 × 10⁻⁷; pathogen first, difference in DPA coefficient from aposymbiotic treatment of −0.59, SE = 0.07, P = 5.19 × 10⁻¹⁶). Despite a higher initial titer than the symbiont-first treatment, the pathogen-first treatment declined in titer at a similar rate (difference in DPA coefficient treatment of 0.61, SE = 0.3, P = 0.08).

Symbiont titer and persistence is unaffected by previous pathogen exposure.

Out of all 240 individuals exposed to Caballeronia, only 5 bugs were symbiont negative when tested, a 98% success rate for symbiont colonization regardless of treatment. We found no significant differences in symbiont titers in either of the pathogen-exposed treatments compared to symbiont only controls over the course of the experiment (symbiont first, difference in intercept from symbiont only of 0.08, SE = 0.05, P = 0.13; pathogen first, difference in intercept from symbiont only of 0.07, SE = 0.04, P = 0.11) (Fig. 3).

FIG 3.

Order of infection does not affect symbiont titers. Symbiont titers were recovered from squash bugs in each treatment. Only 2% of individuals across all treatments tested negative for the symbiont. Shading indicates 95% confidence intervals. No differences in symbiont infection intensity were detected between treatments.

DISCUSSION

We demonstrate that the presence of Caballeronia symbionts in squash bugs consistently results in the competitive exclusion of the plant pathogen S. marcescens resulting in lower intensity infections of shorter duration in symbiont-positive bugs compared to aposymbiotic individuals. Furthermore, we saw no evidence suggesting priority effects were in play for pathogen establishment. Patterns of competitive exclusion persisted regardless of whether insects were exposed to S. marcescens prior to symbiont establishment or after, and the ability of Caballeronia symbionts to successfully colonize their insect host was unaffected by previous pathogen exposure. Although there was some variability in their effects, we observed the same overall trends of S. marcescens persistence and infection intensity for all squash bug-associated Caballeronia. Differences observed among the symbiont strains were minimal compared to the much more striking differences between symbiotic and aposymbiotic insects. The magnitude of the differences we observed in S. marcescens-clearing abilities between aposymbiotic and symbiont-positive individuals is indicative of a significant within-host interaction between pathogen and symbiont.

The shorter persistence times of S. marcescens in symbiont-positive bugs relative to aposymbiotic bugs could be driven, at least in part, by their much lower initial titers. Although it is possible that symbionts alter insect feeding behavior, prompting less consumption of subsequently introduced bacteria, we think this unlikely given the failure of S. marcescens to maintain higher titers even when introduced prior to symbiont exposure. We think this pattern is more indicative of fast and aggressive action mediated by Caballeronia symbionts to prevent the establishment or proliferation of S. marcescens.

Examples of defensive symbioses, whereby symbiotic microbes aid in the clearance or tolerance of coinfecting microorganisms, are plentiful among insects (18, 37–40) and can result in patterns of competitive exclusion of pathogens similar to those we observed. In carrion beetles, for example, symbiotic microbes outcompete entomopathogenic bacteria in vivo and make their hosts resistant to deadly larval infections if present prior to pathogen exposure (20). The antiviral effects of Wolbachia against RNA viruses that infect Drosophila melanogaster are also well documented (18, 41). A general ecological framework of within-host microbial interactions can help us understand how symbionts confer protection to their hosts (42, 43).

There are three main, though not mutually exclusive, types of interactions that can result in symbiont-conferred protection within their host and symbionts, and coinfecting pathogens or parasites may (i) compete for a limiting resource (exploitative competition), (ii) deal each other direct damage (interference competition), or (iii) indirectly compete through a shared natural enemy, typically the host immune system (apparent competition) (44). Furthermore, we might expect the order and timing of organisms’ arrival to influence the outcomes of these interactions (i.e., priority effects). Priority effects can play a dominant role in competitive interactions in cases where interacting organisms have high niche overlap (45, 46). Examples of niche preemption whereby an early arriving species sequesters or uses up resources required for the successful establishment of later arrivals are abundant in community development, particularly in the assembly of the human gut microbiome (47). Immune priming, a prominent example of apparent competition in host-symbiont-pathogen interactions, may also require an established symbiont presence in order to ward off subsequent pathogen infections (25).

Because our study does not address the mechanism through which Caballeronia symbionts and S. marcescens interact within the squash bug, we cannot say with certainty that our results stem from one specific type of interaction or another. It is possible that pathogen and symbiont compete for space or other limiting resources within the insect gut after they are ingested. Other insects are known to rely on various physical barriers within the gut to prevent pathogens from disseminating into the hemocoel and becoming more permanently established (48). Gut-colonizing symbionts can enhance these barriers by forming biofilms that effectively block pathogens from passing through (24, 26). However, this type of inhibitory priority effect is not seen in the squash bug, as evidenced by S. marcescens’ failure to outperform Caballeronia when allowed to establish in insects beforehand. Work in a closely related stink bug species, also possessing Caballeronia symbionts, suggests that apparent competition through the insect immune system could explain the pathogen-clearing patterns we observe in the squash bug. In the bean bug Riptortus pedestris, symbiont colonization leads to differential immune gene expression within the insect gut (49). Immune activation is a common mechanism for the regulation of symbiont titer among insects and could adversely affect coinfecting pathogens (50, 51), regardless of order of exposure. However, more work is needed to determine whether this is the case in squash bugs. It is also possible that symbiont-positive squash bugs can mount a more robust immune response against S. marcescens given the general benefits conferred upon them by Caballeronia (36). If immune activation of the insect host is involved in the clearance of S. marcescens, the lack of priority effects observed suggests a very rapid immune response in individuals with pathogen exposure before symbiont establishment. Our results indicate that no matter the mechanism of interaction, Caballeronia symbionts maintain a competitive edge over S. marcescens within the squash bug.

Although primarily developed in the context of human health (52, 53), symbiont-mediated vector control tactics are now gaining traction among researchers hoping to curb the spread of agricultural pathogens (54, 55). The competitive exclusion of S. marcescens from squash bugs by their symbiotic bacteria is a promising result for further research in this area and could eventually lead to the development of novel methods for controlling cucurbit yellow vine disease (CYVD), which is transmitted by the squash bug and is of increasing concern in the United States (31, 56–58). The absence of priority effects in Caballeronia-S. marcescens interactions is a notable trait in the system. Both the symbionts’ ability to establish in insects and its ability to competitively exclude S. marcescens are maintained regardless of the insect’s order of exposure to pathogen and symbiont. This could vastly simplify the timing of a Caballeronia-based intervention, should it be developed and deployed in the field. However, it must be noted that although our results suggest that symbiont-positive squash bugs are less likely to be competent, long-term vectors of S. marcescens, we did not test vector competence directly from insects to plants. If Caballeronia symbionts are to be used to mitigate CYVD transmission, further work should seek to confirm whether symbiont-positive bugs are in fact less likely to transmit S. marcescens to plants.

Because squash bugs damage plants through a combination of pathogen vectoring and heavy feeding (31), squash bug control strategies must also consider the fact that symbiont-positive bugs develop faster and live longer than their aposymbiotic counterparts. An optimal control strategy must balance the risks of pathogen vectoring, where symbiont-positive bugs may be desirable, with the possibility of intense feeding damage if squash bug populations reach high densities, which is more likely if symbiont-positive bugs are more prevalent. Given the double threat posed by squash bugs, it is likely that potential symbiont-mediated vector control methods will need to be deployed in conjunction with other pest management tactics for optimal results.

MATERIALS AND METHODS

Insect rearing.

We reared all insects in an environmental chamber held at constant temperature (27°C) and 60% relative humidity under a long daylight cycle (16 h light, 8 h dark). We surface sterilized eggs collected from our existing Anasa tristis colony by alternately washing them in 70% ethanol and 10% bleach for 1 min each and then rinsing with 70% ethanol for 10 s. Once eggs hatched, we transferred first-instar nymphs to a sterile container and gave them slices of organic zucchini (Cucurbita pepo) fruit, surface sterilized with 70% ethanol and thinly wrapped in Parafilm. We maintained insects in sterile containers with regular fruit changes until they molted to second instars.

Symbiont effect on pathogen persistence. (i) Symbiont acquisition.

We collected batches of 1- to 3-day-old second instars (n = 150) into sterile rearing boxes for administration of treatments. All Caballeronia strains used were isolated from the crypts of bugs collected from fields in Georgia, Indiana, or North Carolina. We prepared liquid diets consisting of either sterile water and 100 μl of filter-sterilized blue dye (aposymbiotic treatments) or of Caballeronia culture (strain Sq4a, IN-SB1, NC-TM1, or GA-Ox1), sterile water, and 100 μl of blue dye (symbiont treatments) in 30-mm petri dishes. All bacterial cultures were grown in standard Luria-Bertani (LB) broth overnight in a shaking incubator at 28°C. We used optical density at 600 nm measurements to standardize all bacterial diets to 2 × 10⁷ CFU per ml in a total volume of 5 ml. The blue food dye allowed us to visually confirm the uptake of liquid diets in the insects. Once all liquid components were mixed in the petri dish, a sterile dental swab was placed in the open dish before the dish was sealed with Parafilm, leaving the dental swab protruding from one end so that the insects could feed freely. Insects had access to liquid diets for 24 h before being placed back on fruit. Due to constraints in rearing large numbers of insects synchronously, only two of the symbiont treatment groups (GA-Ox1 and Sq4a) had a corresponding aposymbiotic control group (Apo2 and Apo3, respectively) that was started on the same day from the same batch of second instars. All other treatment groups, i.e., two symbiont treatments (IN-SB1 and NC-TM1) and one aposymbiotic treatment (Apo1), were completed as independent batches. Insects matured to the third instar stage at various rates depending on symbiont status and the strain of symbiont received (36).

(ii) Pathogen acquisition.

Third instars (n = 80), which had molted from second instar 1 to 3 days before, were collected from the established treatment groups and fed a strain of phytopathogenic Serratia marcescens isolated from an infected zucchini and chromosomally labeled with red fluorescent protein (RFP), here referred to as Z01. Fluorescent labeling was achieved via triparental mating with Escherichia coli strains E1354 and E2072 carrying the pTNS3-asdEC and pmini-Tn7-gat-P1-rfp plasmids, respectively (59). This protocol consistently produces site-specific insertion of fluorescent proteins in the neutral attTn7 site (60, 61) 20 to 25 bp downstream of the glmS gene. The RFP-positive conjugant was species confirmed via 16S sequencing and displayed growth and morphology consistent with the parental strain. Feedings were done via vacuum-infused zucchini cubes by following established protocols (62). Briefly, liquid Z01 cultures were standardized to 2 × 10⁷ CFU per ml in approximately 25 ml of sterile water with 100 μl of blue dye, which was added to determine the success of the infusion. Slices of organic zucchini 6 mm thick were cut and quartered, placed in a vacuum flask, and then submerged in liquid Z01 culture. The vacuum was turned on, stoppered, and released repeatedly in 10-s intervals until zucchini slices were saturated. The slices were then removed from the flask, wrapped in a thin sheet of Parafilm, and given to the squash bugs. Squash bugs had access to Z01-infused squash for 48 h, after which they were placed back on regular zucchini fruit.

(iii) Sampling.

Once they had been fed Z01, insects were periodically sacrificed to determine the persistence (i.e., presence or absence) of S. marcescens and to quantify the intensity of infection (i.e., bacterial abundance among individuals that tested positive). Five individuals were sacrificed immediately after the 48-h Z01 acquisition period (day 0), every day for 3 days and then every 3 days after that.

At each time point, insects were surface sterilized for 5 min in 70% ethanol and allowed to dry. Whole insects were macerated with micropestles in 1.5-ml microcentrifuge tubes filled with 200 μl of 1× phosphate-buffered saline solution (PBS). Tenfold serial dilutions up to 10⁻⁶ were prepared in 96-well plates. Aliquots of 20 μl were plated on standard LB agar in quadruplicate. Plates were incubated at room temperature for 48 h before bacterial colonies were visualized and counted under a fluorescence microscope. The average number of colonies across replicates was then used to derive estimates of titer in number of CFU per bug.

Sampling times were variable across treatments, with some treatments regrettably cut short due to the cessation of nonessential research activities in response to the COVID-19 pandemic. Here, we focus on data collected up to the last major time point we were able to sample from most treatments. This leaves us with data for most treatments through day nine, with the exception of the aposymbiotic control group (Apo3) corresponding to a trial with Sq4a, which was only sampled through day six.

Order of exposure effect on pathogen persistence. (i) Bacterial acquisition.

We collected batches of second instars (n = 90), which had molted 1 to 3 days prior, into sterile rearing boxes assigned to one of four treatments: symbiont only, pathogen only, symbiont first, and pathogen first. For this experiment, we used Caballeronia sp. strain Sq4a labeled with green fluorescent protein (GFP) (36) as the symbiont and RFP-labeled Z01 as the pathogen. At the start, two treatments (symbiont only and symbiont first) received symbiont diets, and two treatments (pathogen only and pathogen first) received pathogen diets. All bacteria were grown and standardized to concentrations of 2 × 10⁷ CFU/ml as before. All bacterial or sterile water feedings were done via vacuum-infused zucchini cubes for consistency. After 24 h of access to their first bacterial diet, we placed insects on plain zucchini slices for 24 h. We sacrificed five individuals, chosen at random, from each treatment to confirm the presence of Sq4a or Z01 before proceeding to the next feeding. We then gave insects in the symbiont only and pathogen only treatments zucchini infused with sterile water, insects in the pathogen-first treatment symbiont diet, and insects in the symbiont-first treatment pathogen diet. After another 24-h period, insects were placed on plain zucchini fruit for the duration of the experiment. We chose to inoculate insects with both types of bacteria in the same life stage given the narrow window for successful symbiont establishment found in other bugs with Caballeronia symbionts (35).

(ii) Sampling.

Following the final bacterial or sterile water diet, insects were periodically sacrificed to determine the persistence and intensity of symbiont and pathogen infection. Ten individuals were sacrificed immediately after the final inoculation (day 0), every day for 3 days, at days five and seven, and then weekly until day 21. We used the same sampling methods as before, this time plating samples in triplicate. Treatments with both Sq4a and Z01 were plated on regular LB for pathogen detection and on LB with 80 μg/ml spectinomycin for symbiont detection. Symbiont only and pathogen only treatments were only plated on one type of plate, LB with spectinomycin or standard LB, respectively. Colonies were visualized as before.

Statistical analyses.

All statistical analyses were run in R version 3.6.1. We used the package “glmmTMB” (63) to fit generalized linear models (GLMs) accounting for both treatment and time, in this case days postacquisition (DPA) of Z01 or days after final inoculation, to our data on the presence and intensity of Z01 infections in squash bugs, using the binomial and negative binomial error distributions, respectively. All individuals were included in analyses of persistence data, but analyses of Z01 infection intensity were restricted to nonzero values, because, by definition, infection intensity represents the abundance of a parasite within infected individuals. We used the function r.squaredLR in the package “MuMIn” to derive pseudo-R² estimates for our models to facilitate comparisons (64).