key: cord-0427856-elkk1su4
authors: Stoy, Kayla; Chavez, Joselyne; De Las Casas, Valeria; Talla, Venkat; Berasategui, Aileen; Morran, Levi; Gerardo, Nicole
title: Evaluating the role of coevolution in a horizontally transmitted mutualism
date: 2021-12-06
journal: bioRxiv
DOI: 10.1101/2021.12.04.471243
sha: 24f7e83f878d97a47982c04f4872e9f5f2c54138
doc_id: 427856
cord_uid: elkk1su4

Mutualism depends on the alignment of host and symbiont fitness. Horizontal transmission can readily decouple fitness interests, yet horizontally transmitted mutualisms are common in nature. We hypothesized that pairwise coevolution and specialization in host-symbiont interactions underlies the maintenance of cooperation in a horizontally transmitted mutualism. Alternatively, we predicted selection by multiple host species may select for cooperative traits in a generalist symbiont through diffuse coevolution. We tested for signatures of pairwise coevolutionary change between the squash bug Anasa tristis and its horizontally acquired bacterial symbiont Caballeronia spp. by measuring local adaptation. We found no evidence for local adaptation between sympatric combinations of A. tristis squash bugs and Caballeronia spp. across their native geographic range. To test for diffuse coevolution, we performed reciprocal inoculations to test for specialization between three Anasa host species and Caballeronia spp. symbionts isolated from conspecific hosts. We observed no evidence of specialization across host species. Our results demonstrate generalist dynamics underlie the interaction between Anasa insect hosts and their Caballeronia spp. symbionts. Specifically, diffuse coevolution between multiple host species with a shared generalist symbiont may maintain cooperative traits despite horizontal transmission.

Hosts across all domains of life form essential mutualistic interactions with microbial symbionts. 28 Mutualistic microbes fulfill a variety of integral functions for host development and survival, 29 including augmenting nutrition, providing defense against pathogens, and facilitating 30 development (Pais et al. 2008; Boucias et al. 2012; Salem et al. 2013; Masson et al. 2015; 31 Vorburger and Perlman 2018; Gerardo et al. 2020; Kaltenpoth and Flórez 2020) . Likewise, 32 symbionts may depend on hosts for nutrition, replication without competition, and transmission 33 (Lee & Ruby, 1994; Prell et al., 2009; Macdonald et al., 2012; Wollenberg & Ruby, 2012; but 34 see also Garcia & Gerardo, 2014) . The persistence of cooperative interactions relies on the Therefore, we hypothesize that coevolution may also underlie the long-term maintenance of 70 horizontally transmitted mutualisms. 71 Horizontal transmission often presents greater inherent risks for hosts than their microbial 72 symbionts. Hosts that depend on microbial symbionts but fail to acquire symbionts or acquire symbionts that reciprocate selection on local host lineages to provide fitness benefits may evolve 82 greater reliance on these hosts, increasing their investment in the interaction. Specialization can 83 align partner fitness by increasing the rewards available through the interaction (Douglas 1998; 84 Schwartz and Hoeksema 1998). As a result, pairwise coevolution that produces specialized 85 interactions across spatially structured populations may reduce the costs of horizontal 86 transmission by facilitating fitness alignment. 87 Despite its potential benefits, specialization may also limit the pool of potential partners with 88 which hosts can interact. Specialization may then result in an evolutionary dead end for hosts if 89 the environment changes and decreases the prevalence of compatible symbiont genotypes 90 (Futuyma and Moreno 1988; Douglas 1998; Sachs and Simms 2006) . Meanwhile, hosts that 91 interact with a range of generalist symbionts are more likely to find compatible partners. 92 Moreover, long-term interactions between hosts and generalist symbionts may promote diffuse 93 coevolution, a potential pathway by which the costs of the horizontal transmission of generalist 94 symbionts may be limited. Diffuse coevolution results through reciprocal selection between a 95 range of mutualistic partners with their shared common partner (Hougen-Eitzman and Rausher 96 1994; Iwao and Rausher 1997) . In mutualism, for example, diffuse coevolution may occur 97 through selection between a common generalist symbiont with a range of host species. Through 98 diffuse coevolutionary interactions, the exchange of benefits is maintained, and exploitation is 99 limited, via reciprocal selection for cooperative traits between the symbiont and its range of 100 hosts. Diffuse coevolution may align fitness interests because symbionts interacting with a range 101 of hosts are under constant selection for cooperative traits, and selection from multiple hosts 102 limits opportunities for symbiont to become highly adapted to exploit any single host. Symbionts 103 may also benefit from diffuse coevolution with a range of hosts by increasing their own 104 transmission potential while avoiding exploitation and sequestration by a host. As a result, 105 diffuse coevolution between hosts and generalist symbionts has potential to align host and 106 symbiont fitness and limit the costs of horizontal transmission, especially in variable or unstable 107 environments where specificity can reduce the probability of finding a compatible partner.

Patterns of host-symbiont specialization (or lack thereof) can provide evidence of whether 109 pairwise or diffuse coevolution contribute to mutualistic interactions. Host-symbiont 110 specialization is a signature of pairwise coevolution (Thompson 1994). Pairwise coevolution 111 promotes specialization because hosts and symbionts achieve the greatest fitness advantages by 112 imposing constant strong selection on specific partners. Specialization between hosts and 113 symbionts can be observed across different levels of host and symbiont interactions, including 114 within and across species. For mutualistic interactions, specialization results when specific host 115 and symbiont combinations produce higher fitness interactions than those attained through 116 alternative host-symbiont combinations. Alternatively, generalist interactions occur when fitness 117 benefits do not vary across specific host-symbiont combinations. Diffuse coevolution results 118 when the strength and direction of selection on a common partner is driven by selection from 119 multiple mutualistic partners and/or genetic correlations across these interactions links their 120 evolutionary trajectories. (Hougen-Eitzman and Rausher 1994; Iwao and Rausher 1997) . Under 121 diffuse coevolution, phenotypic variation across pairwise interactions will not be observed 122 (Hougen-Eitzman and Rausher 1994; Iwao and Rausher 1997; Inouye and Stinchcombe 2011). 123 Therefore, observing beneficial interactions that lack specificity across pairwise interactions 124 between a range of hosts with a shared generalist symbiont indicates diffuse coevolution may 125 underlie the interactions.

Reciprocal inoculations can be used to provide evidence of pairwise and diffuse coevolution.

Between host and symbiont lineages (i.e. within species), reciprocal inoculations can be used to 128 provide evidence of host-symbiont specificity by testing for local adaptation, where local 129 genotypes outperform foreign genotypes in their home environments (Kawecki and Ebert 2004; 130 Hereford 2009). Observing local adaptation demonstrates evidence of reciprocal evolutionary 131 change between both host and symbiont, consistent with pairwise coevolution. Studies of host-132 parasite interactions frequently measure local adaptation to assess evidence for coevolution 133 (Lively 1989; Ebert 1994; Thrall et al. 2002; Fischer and Foitzik 2004; Lively et al. 2004; Little 134 et al. 2006; Greischar and Koskella 2007; Hoeksema et al. 2008; King et al. 2009 For specialization assays across host species, we collected A. tristis, A. andresii, and A. 197 scorubtica hosts from four sites in Gainesville, Florida separated by a distance of 1.6 to 14.5 km.

Insects were returned to the lab, scanned for ectopic parasites, and allocated for use in either isolates used for experiments and additional isolates from the same populations in three ways. 209 We first used traditional sanger sequencing of 16s rRNA to identify Caballeronia symbionts 210 isolated from field collected squash bugs. We then used whole genome sequencing to gain 211 greater insight into the genetic variation present in the symbiont populations. Finally, we used 212 high throughput 16s rRNA sequencing of whole crypt communities to gain greater insight into 213 the distribution of Caballeronia variants across individuals and populations.

To isolate individual symbionts for local adaptation assays, we dissected the crypts of one to ten Experimentally assessing host-symbiont specificity by testing for local adaptation 295 We established squash bug lineages for local adaptation assays, as described above. Field for Eastern A. tristis and the F2 progeny for Western A. tristis. 302 We selected a single symbiont strain from each collection site for reciprocal inoculations at the 303 smallest geographic scale (GAWG2-4, GACF4, GAFFF3, and GAOX1). We selected a single 

Prior to inoculations, frozen Caballeronia spp. samples were revived as described above. Liquid 310 feeding solutions for symbiont inoculations were prepared as described in Acevedo et al. 2021. 311 Specifically, liquid cultures were prepared and incubated overnight at 28 °C with shaking.

Overnight cultures were diluted 1:5 in LB and incubated at 28 °C with shaking for two hours.

Bacterial feeding solutions (10mL) were prepared by diluting the two-hour liquid cultures with 314 sterile molecular water to ~2x10 7 cells/mL. Blue dye (1%) was added to each solution to allow 315 for visual confirmation of the feeding solution in squash bug guts. Feeding solutions were poured 316 into 35mm Petri dishes, and a cotton dental swab was placed in each dish for squash bugs to feed 317 from. Plates were wrapped in parafilm to prevent spilling or squash bug drowning.

For reciprocal inoculations at the small geographic scale, squash bug eggs from each mating pair 319 for a given collection site were pooled. Eggs were surface sterilized with ethanol (70%) and 320 bleach (10%) and returned to environmental chambers. Emerging first instar nymphs were fed 321 surface sterilized zucchini or squash. At second instar, nymphs were starved for seven to nine 322 hours. Symbiont inoculations were performed such that squash bugs from each collection site 323 were provided with bacterial feeding solutions from each collection site (i.e. one sympatric and 324 three allopatric combinations per host population). After 24 hours, feeding solutions were 325 removed. Thirty nymphs from each treatment were divided into plastic vented containers 326 housing five bugs each and fed zucchini or squash. Due to limitations in egg production, 327 inoculations using each symbiont strain occurred separately over four consecutive weeks. This 328 procedure was repeated until all symbionts were fed to hosts from each population. In total, hosts 329 from each collection site were inoculated with bacteria from each collection site for a total of 330 four sympatric (n = 30 bugs/treatment x 4 sympatric treatments = total of 120 bugs receiving 331 sympatric bacteria) and twelve allopatric inoculations (n = 30 bugs/treatment x 12 allopatric 332 treatments = total 360 bugs receiving allopatric bacteria). Aposymbiotic controls were 333 established by feeding nymphs water according to the protocol above. Twenty-five hosts from Reciprocal inoculations at the large geographic scale were performed using A. tristis collected 380 from Arizona and Georgia. Egg collection, sterilization, and feeding solutions were prepared as 381 described previously. Bugs from each state were provided bacteria from each state for a total of 382 two sympatric (n = 30 bugs/treatment x 2 sympatric treatments = total 60 total bugs receiving 383 sympatric bacteria) and two allopatric treatments (n = 30 bugs/treatment x 2 sympatric 384 treatments = total 60 total bugs receiving allopatric bacteria). Inoculated bugs were placed in 385 groups of five in plastic vented containers and fed zucchini or yellow crookneck squash. Host 386 survival and development rate were recorded as described previously. Symbiont fitness was 387 measured at this geographic scale as described for the small geographic scale.

Testing for specialization between host species and Caballeronia spp. strains 389 The native geographic range of A. tristis overlaps with two closely Caballeornia-harboring sister 390 species, A. andresii and A. scorbutica (Acevedo et al. 2021 ). In the field, these species are 391 observed inhabiting the same plants, indicating the potential for symbiont sharing between 392 species. We tested for specialization between these overlapping species with their respective 393 Caballeronia spp. strains by performing reciprocal inoculations.

Field collected squash bugs were used to establish laboratory colonies and bacterial strains for 395 reciprocal inoculations, as described previously. We randomly selected a single bacterial strain 396 isolated from each host species for reciprocal inoculations: AAF181, ATF2731, and ASM285. receiving conspecific-derived symbionts + 40 bugs receiving heterospecific-derived symbionts = 407 60 total bugs). We prepared bacterial feeding solutions and measured host and symbiont fitness 408 as described for the local adaptation assays.

For reciprocal inoculations within species, we performed used cox proportional hazard models to 411 test whether host survival varied based on the following main effects included in the model: host 412 origin, symbiont origin, or in response to an interaction between host and symbiont (Table 1) .

Death was considered an event, and bugs were censored once reaching adult. Occasionally bugs 414 were accidentally killed while collecting data, and these bugs were censored. For the 415 intermediate scale, we repeated reciprocal inoculations two times, and the effect of inoculation 416 block was included as a fixed effect for both survival and development rate analyses. We also 417 performed cox proportional hazard models using the survival package in R to test whether host 418 development rate to adult varied in response to the following main effects: host origin, symbiont 419 origin, or an interaction between host and symbiont origin (Table 1) . We chose to perform 420 analysis for the rate to adult because variation across treatments was greatest to this stage, and 421 bugs reach reproductive maturity at adult, making the rate to adult the best measure of host 422 fitness. We considered reaching adulthood as an event, and bugs that died before reaching 423 adulthood were censored. If we detected an interaction between host and symbiont origin for 424 survival or development rate to adult, we performed contrasts using the emmeans package in R 425 to test for an effect of sympatric versus allopatric combinations of host and symbiont.

For reciprocal inoculations across species, we used a cox proportional hazard analysis to test 427 whether host survival varied in response to symbiont origin, which was included as a main effect 428 (Table 2) . We considered death an event, and bugs that reached adult were censored. We used a 429 cox proportional hazard model to test whether host rate of development to adult varied in 430 response to symbiont origin, which was included as a main effect (Table 2) . Analyses for each 431 host species were performed separately because difference in the rate of survival and 432 development are likely intrinsic to each separate species. We considered reaching adulthood as 433 an event, and bugs that died before reaching adulthood were censored.

For analysis both within and across species, we further assessed host survival using a Chi-435 squared test to assess the proportion of hosts surviving to adulthood in sympatric versus 436 allopatric combinations of host and symbiont. We used a quasipoisson distributed generalized 437 linear model to test whether symbiont fitness (logCFU/crypt) varied in response to the following 438 main effects: host origin, symbiont origin, or an interaction between host and symbiont origins.

Water was excluded from models comparing the effect of host and symbiont origin on partner 440 fitness, and we compared the effect of receiving a symbiont versus water treatment on host 441 development rate and survival using the same statistical tests mentioned above. Models were fit 442 using R, version 4.1.0.

Geographic genetic variation is observed across A. tristis host and symbiont populations 445 We performed genetic analysis to test for spatial structure and underlying genetic variation identity, suggesting specialization between strain and host species as an unlikely outcome. 480 We tested for specificity consistent with pairwise coevolution by measuring local adaptation At the small geographic scale, we detected a significant effect of the geographic origin of host 494 (χ 2 = 11.55, df = 3, p = 0.009; Table 1 ) and symbiont (χ 2 = 20.70, df = 3, p = 0.0001; Table 1 (Figures 5, S5 ). We detected a significant effect of host (F3,169 = 3.0913, p = 508 0.02878) and symbiont (F3,166 = 20.2003 , p < 0.001) ( Figure S5 ). However, we did not observe a 509 significant interaction between geographic origin of host and symbiont for CFUs per squash bug 510 crypt (F9,157 = 10.04, p > 0.11) (Figures 5, S5) . Taken together, these results indicate that despite 511 phenotypic variation in the effect of the geographic origin of host and symbiont for host fitness, 512 local adaptation has not evolved at this geographic scale.

Because reciprocal inoculations were not performed synchronously, we predicted the observed 514 effects of host and symbiont may reflect variation across replicate rather than variation across 515 strains. To further assess whether geographic origin of host and symbiont affects host survival 516 and development rate, we repeated reciprocal inoculations for two host populations (CF and 517 FFF), such that each symbiont strain was synchronously provided to each host population. When 518 inoculations were performed synchronously, we did not detect an effect of geographic origin of 519 host nor symbiont for host survival (Table S1; Figure S6 ). However, we did observe an 520 interaction between host and symbiont geographic origin for host survival (χ 2 = 8.92, df = 3, p = 521 0.03). We contrasted the survival of hosts paired with sympatric versus allopatric symbionts to 522 test whether this interaction resulted from local adaptation. We observed no difference in host 523 survival when paired with a sympatric versus an allopatric symbiont (p = 0.90), providing no 524 support for local adaptation between host and symbiont. When inoculations were performed 525 synchronously, we observed no effect of symbiont geographic origin, host geographic origin, nor 526 an interaction between host and symbiont geographic origin, for rate of development to adult 527 (Table S1; Figure S6 ). These results indicate that previous variation in the effect of symbiont for 528 host fitness likely resulted from variation across replicate rather than as a result of variation 529 across sites. Moreover, these results are consistent with those obtained previously and provide 530 further support for a lack of local adaptation between host and symbiont at this small geographic (Table 1; Figures 6, S7) . We also observed no interaction 539 between geographic origin of host and symbiont for host survival (Table 1; Figures 6, S7) . 540 Furthermore, the proportion of hosts surviving to adult did not differ between sympatric and 541 allopatric combinations of host and symbiont (Table 1; Figure 6 ). For host survival to adult, we 542 observed a significant effect of host geographic origin (χ 2 = 12.89, df = 2, p = 0.002; Table 1 , 543 Figures 6, S7) . However, we observed no effect of geographic origin of symbiont nor an 544 interaction between the geographic origins of host and symbiont (Table 1; Figures 6, S7) . Taken Figure 7 ). We observed a significant effect of host origin for development rate to adult (χ 2 = 559 10.49, df = 1, p = 0.001; Table 1; Figure 7) , such that hosts from the western United States 560 developed slower than those from the eastern United States. We observed no effect of symbiont 561 geographic origin nor a significant interaction between host and symbiont geographic origin for 562 host development rate (Table 1; Figure 7) . For symbiont fitness, we measured the number of 563 CFUs per crypt of surviving adult squash bugs. We detected a significant effect of symbiont 564 origin (F1,15 = 8.52, p = 0.01), but we did not observe a significant effect of host origin nor an 565 interaction between host and symbiont. Overall, we did not find support for our prediction that 566 pairwise coevolution underlies the maintenance of this horizontally transmitted mutualisms. Specialization is not observed between host species and associated symbiont strains 572 We tested for specialization between symbionts and the host species from which they originated.

If interactions were specialized, we predicted we would observe higher fitness interactions 574 between hosts and conspecific-derived symbionts relative to symbionts derived from a 575 heterospecific host. We did not detect a significant effect of symbiont origin for host survival 576 across A. tristis, A. andresii, or A. scorbutica hosts (Table 2; Figures 8, S8 ). For all three host 577 species, the proportion of hosts surviving to adult did not vary in response to receiving a 578 symbiont from a conspecific host versus a heterospecific-derived symbiont (Figure 8 ). Similarly, 579 Table 1 . Statistics for host survival and development rate for local adaptation reciprocal inoculations. We performed cox proportional hazard models to test for effects of geographic origin of host and symbiont and an interaction between host and symbiont geographic origin on host time to death and time to adult. If a significant interaction was observed, we performed a linear contrast to test whether host fitness varied in response to sympatric versus allopatric symbionts. In this study, we tested the hypothesis that coevolution underlies the maintenance of horizontally 587 transmitted mutualisms. We assessed the role of pairwise coevolution by measuring host-588 symbiont specificity across three geographic scales. We observed no specificity between 589 lineages. We then tested for specialization between three host species, A. tristis, A. scorbutica, 590 and A. andresii with their associated Caballeronia symbionts but observed no evidence for 591 specialization. Our results strongly demonstrate a lack of host-symbiont specificity in these 592 interactions and fail to provide support for our hypothesis that pairwise coevolution underlies 593 this horizontally transmitted mutualism. Instead, we observe evidence of generalist, beneficial 594 symbionts likely under selection from a range of hosts for fixed phenotypic traits. These to capture variation by picking genetically distant strains; however, because symbiont selection 658 was dependent on variation in the 16s gene, it may not represent genetic variation relevant to the 659 mutualistic interaction. In the future, our research will focus on characterizing genes underlying 660 the interaction, variation at these loci, and assessing whether variation at specific loci 661 corresponds to phenotypic variation across symbiont strains.

There has been a long recognized need to assess the coevolutionary dynamics between 663 mutualists within a community context (Thompson 1994; Bronstein et al. 2003; Thrall et al. 664 2007 . Heat map showing the genetic distance across symbionts strains isolated from A. tristis (AT; pink), A. andresii (AA; dark purple), and A. scorbutica (AS; light purple). All hosts and their associated symbionts were collected in Florida, USA. Genetic distance was calculated using average nucleotide identity (ANI), which varied from 0.85 to 0.99. We observed little variation across strains isolated from heterospecific hosts, with only one strain (ATM282) exhibiting substantial variation from other strains. Far right, evolutionary relationships between strains were assessed using phylogenomic analysis. Strains ASM285, AAF181, and ATF2731, asterisks, were used for reciprocal inoculations to test for specialization between host species and symbiont strain. * * * * * * Figure S1 . Pangenome analysis of symbiont strains isolated from IN (blue), GA (purple), and NC (green). Dark regions show the presence of gene clusters, and faded regions indicate the absence of gene clusters. The box to the right shows the relative number of gene clusters across genomes, relative number of genes present in only one genome (singleton genes), relative number of genes per kbp, redundancy, genome completion, relative GCcontent, and total length of the sequence. Variation can be observed across the genomes of symbionts isolated from across their geographic range. Figure S2 . Pangenome analysis of symbiont strains isolated from three different host species A. tristis (AT; yellow), A. andresii (AA; orange), and A. scorbutica (AA; pink). Dark regions represent the presence of gene clusters and faded regions represent the absence of gene clusters. The box to the right shows the relative number of gene clusters across genomes, relative number of genes present in only one genome (singleton genes), relative number of genes per kbp, redundancy, genome completion, relative GC-content, and total length of the sequence. Variation can be observed across the genomes, particularly by the presence of gene clusters in A. scorbutica that are not present in A. tristis or A. andresii strains. Overall, little variation exists outside of these regions of the genome.

Num genes per kbp Singleton gene clusters Num gene clusters Figure 5A shows host development rate across all developmental life stages. We performed a cox proportion hazard model to determine whether host rate of development to adult varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. We observed a significant effect of symbiont (p < 0.001) and an interaction between host and symbiont (p = 0.03), which was not driven by differences in the effect of sympatric versus allopatric symbionts on host fitness (p = 0.23). Hosts receiving water developed slower than those that received a symbiont across all life stages. See figure S4 for host development rate across all pairwise combinations of host and symbiont. Figure 5B shows the proportion of bugs surviving to adult across experimental treatments. The proportion of bugs surviving to adult did not vary across sympatric and allopatric treatments (p = 0.75) but was significantly lower for bugs receiving water versus those receiving a symbiont (p < 0.001). Figure 5C shows the survival curves across symbiont treatments (see figure S4 for survival curves across all pairwise combinations of host and symbiont). We performed a cox proportional hazard model to determine whether host survival varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. We observed and effect of symbiont (p = 0.0001) and host (p = 0.009) for host survival. Figure 5D shows the effect of treatment on symbiont fitness (logCFU/crypt) (see figure S6 for each pairwise combination of host and symbiont). We observed no effect of host on symbiont fitness. C.

D.

E. Figure S4 . Host development rate and survival when paired with symbionts isolated from each site (CF, FFF, Ox, and WG) at the small geographic scale. Figures S4A-D show host development rate across all development stages. We performed a cox proportional hazard model to test for an effect of host origin, symbiont origin, and an interaction between host and symbiont origin on host development rate to adult. We observed a significant effect of symbiont origin (p < 0.001) and an interaction between host and symbiont origin (p = 0.03) for host development rate to adult. This interaction was not driven by differences in the effect of sympatric versus allopatric symbionts on host fitness (p = 0.23), as shown in Figures S4A-D. Figures S4E-H show host survival. We performed a cox proportion hazard mode to test for an effect of host origin, symbiont origin, and an interaction between host and symbiont origin on host survival. We observed a significant effect of symbiont origin (p = 0.0001) and host origin (p = 0.009) on host survival. Overall, we find no evidence for genetic specificity between host and symbiont. 729 730 Figure S5 . Symbiont fitness (logCFU/crypt) for the small geographic scale. We performed a quasipoisson distributed generalized linear model to determine whether symbiont fitness varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. Plots A-D demonstrate the fitness for each individual symbiont when paired with each host. Symbiont titer within crypts was measured by dissecting adult squash bugs. Symbiont fitness did not vary in response to host origin, providing no evidence for local adaptation. Figure S6 . Host fitness at the small geographic scale for the FFF and CF host populations when inoculations with symbionts from each population were performed synchronously. Plots A and B show host development rate for the FFF and CF populations. We performed a cox proportion hazard model to determine whether host rate of development to adult varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. Development rate did not vary in response to the origin of host, symbiont, nor an interaction between the origin of host and symbiont. Plot C shows overall development rates for sympatric versus allopatric combinations of host and symbiont. Plots D and E show host survival for the FFF and CF populations. We performed a cox proportional hazard model to determine whether host survival varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. We observed no effect of host nor symbiont origin on host survival. However, we did observe an interaction between host and symbiont origin (p = 0.03), but this was not driven by differences in the effect of sympatric versus allopatric symbionts on host fitness (p = 0.90). Plot F shows the overall survival for sympatric versus allopatric combinations of host and symbiont. . We performed a cox proportional hazard model to determine whether host development rate to adult varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. We observed a significant effect of host geographic origin for rate of development to adult (p = 0.002). Bugs receiving water developed slower than those receiving a symbiont. Plot B shows the proportion of bugs surviving to adult across experimental treatments. We performed a cox proportional hazard model to determine whether host survival varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. Survival to adult for bugs receiving water was significantly reduced compared to those receiving a symbiont (p < 0.0001) but did not differ between sympatric and allopatric treatments. Survival over time did not vary across bugs receiving water versus those receiving a symbiont (C). This trend was driven by bugs that became developmentally "stuck" at a juvenile life stage but took a long time to die. We did not observe a significant effect of host origin, symbiont origin, nor an interaction between host and symbiont origin for host survival. We were unable to collect symbiont fitness data due to lab closures resulting from the COVID-19 pandemic.

A.

C.

B. Figure S7 . Host development rate and survival for all pairwise combinations of hosts and symbionts isolated across the three states (GA, IN, and NC) at the intermediate geographic scale. Plots A-C show host development rate across all developmental life stages. We performed a cox proportional hazard model to test for an effect of host origin, symbiont origin, or an interaction between host and symbiont origin for rate of development to adult. We detected a significant effect of host geographic origin on host development rate. We did not observe an effect of symbiont nor an interaction between host and symbiont geographic origin. Plots D-F show host survival. We performed a cox proportional hazard model to determine whether survival varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. Host survival did not vary in response to geographic origin of host, symbiont, nor an interaction between host and symbiont geographic origin. host Figure 7 . Host fitness at the large geographic scale. Plot A shows host development rate across all life stages when paired with a sympatric versus an allopatric symbiont. We performed a cox proportional hazard model to determine whether rate of development to adult varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. Plot B shows the proportion of bugs surviving to adult across experimental treatments. We detected no difference in survival to adult between sympatric and allopatric treatments (χ 2 = 0.25, df = 1, p = 0.62).

No bugs receiving water survived to adult, so they are not depicted here. Plot C shows survival across all experimental treatments. We performed a cox proportional hazard model to determine whether survival varied in response to host origin, symbiont origin, or an interaction between host and symbiont origin. We detected a significant effect of host origin (p = 0.001) on host fitness. Survival over time did not vary across bugs receiving water versus those receiving a symbiont. This trend was driven by bugs that became developmentally "stuck" at a juvenile life stage but took a long time to die. We did not observe a significant effect of host origin, symbiont origin, nor an interaction between host and symbiont origin for host survival. Plot D shows symbiont fitness for sympatric and allopatric treatments. We observed a significant effect of symbiont (F1,15 = 8.52, p = 0.01) on symbiont fitness but no effect of host nor an interaction between host and symbiont fitness.

A. B.

C. D.

Host fitness data for reciprocal inoculations to test for specialization across host species with symbiont strain. Plots A-C show host development rate when paired with symbionts derived from a conspecific versus heterospecific host across all developmental stages for A. tristis (A), A. scorbutica (B), and A. andresii (C). We performed a cox proportional hazard model to determine whether symbiont origin had an effect on host rate of development to adult. We observed no effect of symbiont on host development rate to adult (see figure S8 for pairwise combinations of host species and symbiont strain). Plots D-F show host survival for A. tristis (A), A. scorbutica (B), and A. andresii (C) when paired with symbionts derived from a conspecific versus heterospecific host. We performed a cox proportional hazard model to determine whether host survival varied in response to symbiont origin. We observed no effect of symbiont origin on host survival (see figure S8 for pairwise combinations of host species and symbiont strain). Plots G-I show the proportion of bugs surviving to adult across all experimental treatments. We observed no difference in host survival when paired with symbionts isolated from a conspecific versus a heterospecific host. . We performed a cox proportional hazard model to assess the rate of development to adult across treatments. We did not observe an effect of symbiont origin on host development rate. Plots D-F show host survival for each host species: A. tristis (D), A. scorbutica (E), and A. andresii (F). We performed a cox proportional hazard model to assess the effect of symbiont origin on host survival, and we did not observe a significant effect of symbiont origin on survival for any host species.

Symbiont fitness (logCFU/crypt) when paired with hosts that are conspecific or heterospecific to the hosts from which they were derived. Symbiont fitness was measured by dissecting the crypts of bugs that survived to adult during host fitness assays and counting the number of CFUs per crypt for each host species: A. tristis (A), A. scorbutica (B), A. andresii (C)

Crypts of squash bugs that survived to adult were dissected, crushed, and plated to assess the number of symbiont CFUs per crypt. The fitness of symbionts isolated from each host species is shown when symbionts were paired with A. tristis (A), A. scorbutica (B), and A. andresii (C)

 Figure S3 . Relative abundance of OTUs within the Burkholderiaceae (Bacterial family containing Caballeronia spp.) across the crypts isolated from 39 A. tristis samples originating in Georgia (GA), Indiana (IN), and North Carolina (NC). Analysis was limited to the top twenty OTUs. Each bar represents the variation in Burkholderiaceae OTUs for a single sample. OTU analysis was performed using a sampling depth of 34,993 reads, which included 48% of all reads across 58 Burkholderiacea taxa. Analysis of the relative abundance of OTUs within the Burkholderiaceae across samples was randomized and limited to the top 20 OTUs, demonstrated here. The analysis indicates co-colonization of Burkholderiacea taxa commonly occurs within the crypts, despite observing little variation across cultivable strains. Moreover, this analysis indicates variation in crypt composition across individual bugs.

GACF2 GACF5  INDec1  INDec5  INDec7  NCF1  NCF2  GAFFF3  GAFFF4  GAFFF5  INFOX1  INFOX4  INFOX6  INFOX7  INML3  INML5  GAOx1  GAOx2  GAOx3  NCRB1  NCRB3  NCRB4  INSB1  INSB4  INSB5  INSB6  NCTM3  NCTM4  NCTM5  NCWD1  GAWG1-1P  GAWG1-