Insects were collected in northern Costa Rica at La Selva Biological Station (Organization for Tropical Studies), a 1500-hectare ecological reserve with a wet neotropical lowland climate (McDade et al., 1994). The station, located at the confluence of the Sarapiqui and Puerto Viejo rivers in the province of Heredia, Costa Rica (10°26′N, 83°59′W), covers ~6 sq miles, and is home to 30 pentatomid species (out of 170 species known to Costa Rica, James Lewis, Pers. Commun.). Collection and export permits were acquired through the Organization for Tropical Studies and the Costa Rican Ministry of the Environment and Energy (079-2013-SINAC). All pentatomids examined in this study (Figure 1) were collected using nets and plastic containers, within a ~1 km² area near the central laboratory, with the exception of Loxa, Euschistus, and some Mormidea specimens, which were collected ~5 km away. All were either preserved immediately for molecular and microscopic analysis, or kept alive for natural history observations, captive breeding, and rearing of nymphs. Specimens were routinely observed and collected on specific host plants, including Piper sancti-felicis for Sibaria englemani, legumes for Loxa sp., grasses within the Poaceae for Mormidea and Euschistus species, Handroanthus chrysanthus, and Pentaclethra macroloba for Edessa n. sp 1, and Guatteria amplifolia for Edessa n. sp 3. Antiteuchus costaricensis, Arvelius porrectispinus, the various other Edessa specimens were found in association with man-made structures such as UV light traps, window screens, and wooden walls. Most specimens were examined using a Leica S8APO stereomicroscope and photographed with a Nikon Coolpix P6000 digital camera. Identities of the host insects were determined by morphology (Fernandes, personal observation).

Adult specimens preserved immediately in 70% ethanol were dissected and DNA was extracted from the ceca of each individual using the Qiagen DNeasy kit, according to the manufacturers instructions (Qiagen, Valencia, CA). Whole specimens for 1st instar nymphs, and total abdominal tissue for 2nd–5th instar nymphs preserved in 70% ethanol were also extracted as described above. To assess whether insects hosted internal bacteria, a 1465-bp fragment of the bacterial 16S rRNA gene was PCR-amplified using bacteria-specific primers (27F and 1492R; Lane, 1991). Clone libraries of bacterial 16S rRNA amplicons were constructed from each insect, using the TOPO TA cloning kit (Invitrogen, Carlsbad, CA). Transformants were grown overnight in LB broth with 50 μg ml⁻¹ kanamycin, and screened directly for the presence of inserts using M13F and M13R vector primers and the thermal cycling conditions of an 8 min initial denaturation, followed by 30 s each of denaturation at 94°C, annealing at 54°C, elongation at 72°C (30 cycles), and a final 6 min of elongation at 72°C. In all cases, M13 amplicons were digested first with HaeIII (according to manufacturers instructions; New England Biolabs) in order to observe diversity and select unique samples for sequencing. In total, 19–48 clones were sequenced for each library, and revealed 78–100% dominance by a single bacterial OTU (97% similarity; Table 1). To selectively amplify the dominant bacteria within the Sibaria (Carpocarini) bacterial clade and the Edessa bacterial clade (Figure 2), diagnostic symbiont-specific primers were designed in silico (Sib_1F; 5′- GAAACTGCCC-GATGGAGG-3′ and Ed_1F; 5′- GGATCTACCTAGTGGAGGG-3′, respectively), and paired with 1492R to amplify a 16S rRNA product length of 1412-bp. This allowed for rapid screening of the presence of the dominant bacteria in eggs, nymphs, and field collected adults. Thermal cycling conditions included 60 s each of denaturation at 94°C, annealing at 54 and 52°C, for general and symbiont-specific 16SrRNA, respectively, elongation at 72°C (25 cycles), and a final extension at 72°C for 6 min. M13 amplicons, or 16S rRNA gene products sequenced directly, were cleaned prior to sequencing with MultiScreen HTS plates (Millipore Corporation, Bedford, MA). Samples were sequenced via ABI sequencing technology (Laragen, Inc., Culver City, CA). Sequences were assembled, edited, and aligned using Sequencher v4.10.1 (GeneCodes Corp.).

Closest relatives were acquired using the GenBank Basic Local Alignment Search Tool, BLASTn (Altschul et al., 1997) and identified using phylogenetic analysis. Our 16S rRNA sequences along with additional sequences obtained from GenBank were compiled in ARB, after initial alignment using the SILVA Aligner function, with subsequent manual refinements (Ludwig et al., 2004; Preusse et al., 2007). For near full-length representatives and closest relatives, maximum parsimony (MP) analysis was conducted. In some cases, partial sequences recovered in our study were added to the tree in ARB via parsimony insertion within a tree of longer sequences. NJ analysis was also performed with 2000 bootstrap replicates to assign confidence levels to nodes, shown in Figure 2 (PAUP4.0b10; Swofford, 1998). Bacterial 16S rRNA sequences obtained in this study have been deposited in the GenBank database under accession numbers KJ934755-KJ934778.

For examination by transmission (TEM) electron microscopy, samples (~1 mm³) were fixed in 3% glutaraldehyde buffered with 0.2 M phosphate (pH 7.2). Following a wash in 0.1 M sodium cacodylate containing 24% sucrose, samples were post-fixed with 1% OsO₄ in 0.1 M sodium cacodylate for 1 h, stained en bloc in 3% uranyl acetate in 0.1 M sodium acetate buffer for 1 h, dehydrated through an ethanol series, then infiltrated and embedded in Spurr's resin (Ted Pella, Redding, CA, US). Thick (0.4 μm) and thin (70 nm) sections were stained with methylene blue and lead citrate, respectively, then examined and photographed using a Zeiss Labrolux 12 light microscope and Zeiss EM109 TEM.

For fluorescence in situ hybridization (FISH), insects, initially preserved in 4°C 4% paraformaldehyde for 24 h, were rinsed twice with 1X phosphate-buffered saline (PBS), transferred to 70% ethanol and stored at −20°C. Adult cecal tissue, or whole nymphs, were embedded in Steedman's wax (one part cetyl alcohol was added to nine parts polyethylene glycol (400) distearate, mixed at 60°C; Steedman, 1957), and added to the sample in an ethanol:resin gradient of 3:1, 2:1, and 1:1, according to Pernthaler and Pernthaler (2005). Samples eventually embedded in full strength wax were allowed to solidify and were sectioned (5–10 mm thick) using a Leica RM2125 manual microtome, and placed onto Superfrost Plus slides (Fisher Scientific). Samples were de-waxed by three rinses in 100% ethanol (5 min each), followed by rehydration in 70% ethanol (5 min). Hybridization and wash buffers were made as described previously (Pernthaler and Pernthaler, 2005), using 35% formamide in the hybridization buffer and 450 mM NaCl in the wash solution. A universal bacterial probe set was used (EUB338I-III, Amann et al., 1990; Daims et al., 1999), along with symbiont-specific probes targeting the 16S rRNA gene of the Sibaria englemani symbiont (Pentatominae-symbiont clade, Tribe Carpocorini, Figure 2; Sib_A, 5′-CCTG-GGCAGTTTC-3′), and the Edessa specific clade (Edessinae-symbiont clade, Figure 2; Ed_A, 5′-CCCTCCACTAGGTAGATCC-3′). The Sib_A probe had 100% identical sequence match to related Erwinia/Pantoea-like bacteria in the GenBank database, and the Ed_A probe exhibited only one mismatch to other Heteropteran symbionts, with no other obvious hits. Both were designed to have a Tm of ~53°C, GC = 58–61%. Formamide stringency was determined empirically by testing each probe against its target, and reciprocally against its non-target, between 15–55% formamide. Signal of symbiont-specific probes was strongest at 35 and 25% formamide, for the Ed_A (Figure 3) and Sib_A (Figure 4) probes, respectively, without compromising stringency (Figure 5). A probe targeting Epsilonproteobacteria (EP404_AAAKGYGTCATCCTCCA; Macalady et al., 2006) was used as a negative control. All probes were labeled with Cy3. Hybridizations were conducted at 46°C for 2–8 h, followed by a 15 min wash at 48°C. Tissues were counter-stained with a dilute 4′6′-diamidino-2-phenylindole (DAPI) solution (5 mg ml⁻¹) for 1 min and examined under epifluorescence microscopy using a Nikon E80i epifluorescence research microscope with a Nikon DS-Qi1Mc high sensitivity monochrome digital camera.

In order to confirm the identity of egg masses and nymphs collected in the field, insect mitochondrial cytochrome c oxidase subunit I (COI) was amplified using one of two previously published sets of primers; insectCOIF (5′-TACAATTTATCGCCTAAACTTCAGCC-3′) and insectCOIR (5′-CCCGGT-AAAATTAAAATATAAACTTC-3′; Kuechler et al., 2011) or LCO1490 (5′-GGTCAACAAA-TCATAAAGATATTGG-3′) and HCO2198 (5′-TAAACTTCAGGGTGACCAAAAAATCA-3′; Folmer et al., 1994). Thermal cycling conditions included 60 s each of denaturation at 94°C, annealing at 50 and 48°C, respectively, elongation at 72°C (25 cycles), and a final extension at 72°C for 6 min. COI gene products were cleaned prior to direct sequencing with MultiScreen HTS plates (Millipore Corporation, Bedford, MA). Samples were sequenced via ABI sequencing technology (Laragen, Inc., Culver City, CA). Sequences were assembled, edited, and aligned using Sequencher v4.10.1 (GeneCodes Corp.). For Edessa n. sp. 1, all individuals, including eggs, nymphs, and adults, displayed less than 0.3% divergence in their COI sequences (3 out of 850 bp), confirming that they all belonged to the same species. Similarly, all specimens thought to be Sibaria englemani, including eggs, nymphs, and adults, had less than 0.4% divergence in their COI sequences (3 out of 730 bp). Insect cytochrome c oxidase (COI) sequences obtained in this study have been deposited in the GenBank database under accession numbers KM020416-KM020425.

Field-collected Sibaria englemani were housed in 14 × 14 × 24″ mesh enclosures under ambient conditions (24–32°C, 68–85% humidity, 12:12 h photoperiod) and supplied, every other day, with inflorescences of their native host plant, Piper sancti-felicis. Egg masses were removed from the enclosures within 12 h of oviposition and separated into one of several treatments to evaluate the mechanism of symbiont acquisition and impacts of symbiont deprivation (described below). Eggs and first instar nymphs (n = 58 egg masses from 190 females over 3.5 months) were incubated at 25–30°C in 3.5 cm petri dishes lined with filter paper, with wetted cotton for humidity. Second instar nymphs were then reared under the ambient conditions noted above. Populations of Mormidea aff. ypsilon, Edessa n. sp. 1, and Edessa bugabensis were also maintained on their respective host plants, to observe breeding and developmental progression of the nymphs (Figure 6).

Within ~6 h of hatching, Sibaria englemani cohorts (i.e., eggs from the same clutch; n = 4) were divided, with half of the 1st instar nymphs transferred to an identical petri dish to prohibit symbiont acquisition from the surface of the egg mass (chorion). First instar nymphs of the same cohort that were provided or denied access to the chorion were preserved ~24 h after hatching in 70% EtOH. Diagnostic PCR with symbiont-specific probes was conducted for both treatments to determine whether chorion probing facilitated symbiont acquisition.

To evaluate the impacts of symbiont deprivation on host development, Sibaria englemani eggs were surface sterilized upon hatching to prevent symbiont uptake by half of each cohort (n = 54 cohorts). Surface sterilization involved 5 min immersion in 10% bleach, modified from Prado et al. (2006). Symbiotic (control) eggs of each cohort were washed in sterile dH₂O. Previous studies suggest that washes of the chorion are not directly harmful to the insect itself (Prado et al., 2006; Prado and Almeida, 2009b). Upon hatching, all first instar nymphs were allowed to probe the chorion to attempt symbiont acquisition. Second instar nymphs were relocated to independent petri dishes (3.5 cm) and nymphs assigned the symbiont-deprivation treatment were provided with antibiotic-laden water (0.005% kanamycin/ampicillin; Kuriwada et al., 2010). Congruously reared symbiotic and symbiont-deprived siblings were compared within cohorts to determine the effects of symbiont knockdown.

Developmental duration, morphometrics, and survivorship were compared between symbiotic and symbiont-deprived Sibaria englemani nymphs to determine if bacterial symbionts play a role in nymph development. Stadia duration was evaluated using a series of nonparametric two-sample tests (Wilcoxon's rank tests) with a Bonferroni correction (p = 0.01 significance threshold) for each instar, due to the non-normal distribution of stadia duration data. Survivorship was compared using a contingency table with no fixed margins, and analyzed via a Pearson's chi- squared goodness of fit test (Figure 7). Size of the insect was defined as the product of pronotum width and body length (mm). This morphometric was plotted against the age of the insect (time from hatching), to obtain average growth curves for both symbiotic and symbiont-deprived nymph populations. The slopes of linearized (log-transformed) growth curves for each individual were compared for all nymphs that reached the third instar (n = 39), and evaluated using a two-sample t-test: data fulfilled assumptions of normality and variance for each treatment. Comparative analyses were performed using the Java Graphical User Interface for R (JGR v.1.7-16; GNU General Public License). Phylogenetic congruence between host and symbiont phylogeny was evaluated using the Icong index of topological similarity (de Vienne et al., 2007).

The bacteria-bearing organ, or ceca, was conspicuous in all specimens, with numerous tube- or sac-like convolutions (crypts) arranged in a coiled rosette (Figures 3, 4, 8). The ceca comprised 8–20% of the soft tissue biomass of Edessa specimens, whereas it comprised ~5% in Sibaria specimens. Crypts, which varied in size, color, and configuration between species, housed extracellular symbionts in the lumen (Figures 3, 4, 8). The cecal organ was morphologically distinct between host genera but not between closely related taxa within subfamilies. For example, crypts of the subfamily Edessinae were pale yellow in color with large and neatly aligned in two fused rows with few convolutions, in comparison to the translucent and elaborately folded crypt cavities of the subfamily Pentatominae (Figures 3, 4, 11). To determine the identity of microbes associated with each stinkbug species, 16S ribosomal RNA sequences were amplified from ceca DNA extracts, as well as extracts from whole eggs and nymphs (Figure 2). Results show that all Costa Rican pentatomids possessed a dominant gammaproteobacteria. Gut bacteria were polyphyletic with three primary clades generally congruous within subfamily designations of the stinkbug hosts. The Icong congruence index for testing topological similarity between trees revealed a P-value of 0.06 (de Vienne et al., 2007). For example, bacteria closely related to the cultured genus Erwinia were dominant within the subfamily Pentatominae, based on 98% similarity, including Sibaria englemani (adults, nymphs, and oviposited eggs), Mormidea collaris, M. aff. ypsilon, Euschistus sp., Arvelius porrectispinus, and Loxa sp., representing 78–98% of the recovered 16S rRNA sequences associated with each specimen (Figure 2, Table 1). Bacteria associated with these six stinkbug species were >97% similar to each other in 16S rRNA sequence. Further, similarity among bacterial associates was observed for the tribe Carpocorini, subfamily Pentatominae (only 1% divergence in 16S rRNA sequence; Figure 2), despite the very different diets among hosts, including grasses within the Poacea for Mormidea and Euschistus and Piper sancti-felicis for Sibaria englemani. The presence of a unique, dominant bacterial type within the Enterobacteriaceae (97–100% of recovered 16S rRNA sequences; Table 1), monophyletic and equidistant (based on ~90–94% similarity) between symbionts of the plataspid stinkbug Megacopta punctatissima (Candidatus Ishikawaella capsulata; Fukatsu and Hosokawa, 2002) and the brown marmorated stinkbug Halyomorpha halys, was also confirmed for six species within the subfamily Edessinae and Antiteuchus costaricensis, within the subfamily Discocephalinae (n = 1–4 individuals of each species; Figure 2, Table 1). Bacteria associated with the six Edessa species were >93% similar to each other in 16S rRNA sequence, while the ribotype associated with Antiteuchus was slightly less similar (91–93% of the 16S rRNA sequence). Overall, the two main bacterial clades associated with Sibaria and Edessa shared only ~88–92% similarity, based on 16S rRNA sequence.

For both Sibaria englemani and Edessa n sp. 1, the dominant bacterium was detected in all lifestages, including oviposited eggs and nymphs hatched in environmental isolation in the laboratory (Table 1). Dissection microscopy and FISH was utilized to locate and visualize symbionts in Sibaria englemani and Edessa n sp. 1. Transmission electron microscopy revealed the presence of bacteria-like cells of a single morphotype, within the ceca of both species (Figures 3, 4). The midgut symbionts within Edessa n sp. 1 were rod shaped, approx. 1.0–4.5 μm in length, and filled the entire lumen within each crypt (Figures 3C,D). The midgut symbionts within S. englemani were cocci shaped, approx. 1.0–1.5 μm in diameter (Figures 4C,D). Similarly, FISH microscopy with a specific probe revealed that a dominant gammaproteobacteria populated the lumen of crypts in both insects (Figures 3, 4). Bacterial colonization of specific midgut regions provided compelling evidence of a genuine symbiotic relationship. Fluorescence microscopy also revealed crypt-containing, symbiont-filled tissues, albeit much smaller than in adults, in a wild caught S. englemani 4th instar (Figure 9) and in 5th instar nymphs reared in the laboratory (Figures 10A–C). The ceca comprised ~1–2% of soft tissue biomass of Sibaria nymphs, as compared to ~5% in adults. Connections between the midgut main cavity and the symbiont-filled crypts were noted for both Sibaria and Edessa (Figures 3F, 4F), suggesting that the insects are able to excrete the symbiotic bacteria onto their eggs by surface contamination (Goodchild, 1963; Kikuchi et al., 2009).

Adult Sibaria englemani collected in June 2013 were maintained in ambient conditions, at 25–30°C, for ~55 days and reproduced successfully in captivity when provided with inflorescences of their native diet, Piper sancti-felicis. Females preferentially oviposited on mesh cage surfaces, typically producing 8 eggs per cohort (~1 mm in diameter; Figures 6, 11), with an incubation time of 7.3 ± 0.2 d and ~95% viability. Egg masses were turgid and shiny, adhering to leaf surfaces or mesh caging. Eggs developed visible eyespots and ruptor ovis approximately 2–3 days prior to emergence. Stadia durations averaged 4.5 ± 0.6, 7.9 ± 2.9, 7.3 ± 1.7, and 7.5 ± 1.4 d (mean ± standard error) for 1st through 4th instars, respectively. Upon hatching, teneral first instar nymphs aggregated on the eggs (Figures 11C,D), and were observed to probe the chorion for 1–6 h after hatching (Supplemental Video). Diagnostic PCR amplification using symbiont-specific primers demonstrated that intact eggs contained symbionts, whereas neither the ovaries of females nor eggs with the chorion removed did (data not shown). Further, diagnostic PCR amplification of the symbiont 16S rRNA gene indicated that only nymphs allowed to probe the chorion successfully acquired symbionts, whereas those that were removed from their egg masses within 10 min of hatching did not, supporting maternal transmission as the predominant mechanism of symbiont acquisition.

Complete symbiont-deprivation was not attained in all insects: ~50% of nymphs with surface sterilized chorions exhibited incomplete aposymbiosis (as revealed by PCR and QPCR amplification of symbiont specific 16S rRNA, data not shown). Several surface sterilization protocols, including immersion in 95% ethanol or 4% formalin prior to a wash in bleach also resulted in incomplete symbiont knockdown, despite reports of previous success in the literature (Bistolas, personal observation; Prado et al., 2006, 2009; Hosokawa et al., 2007). Overall, the process of surface sterilization of the chorion to knockdown symbionts did not impact total egg viability (95%). Survivorship of symbiont-deprived and symbiont-endowed nymphs was comparable (80 and 83%, respectively, p = 0.959, Pearson's Chi-squared-test), with both treatments experiencing high mortality at the second stadium when reared in captivity (Figure 7). Vulnerability to high mortality and failure to molt to the 3rd instar may be due to nutritional deficits or susceptibility to fluctuations in ambient conditions, as reported by studies on the Southern green stinkbug, Nezara viridula (Hokyo and Kiritani, 1963; Prado et al., 2009). Symbiont depression did not influence total developmental duration (34.0 ± 0.8 d vs. 37.0 ± 1.5 d, for symbiotic vs. symbiont-deprived nymphs, respectively). Stadia durations were equal for 1st and 3rd–5th stadia in both symbiotic and symbiont-deprived nymphs (p = 0.87, p = 0.26, p = 0.17, and p = 0.26, for first, third, fourth, and fifth instars respectively, Wilcoxon's rank). Standard length and pronotum width were comparable between symbiotic and symboint-deprived nymphs. Similarities in morphometrics and percent survivorship may be a reflection of incomplete symbiont knockdown.

Despite incomplete aposymbiosis, congruent rearing of symbiont-endowed and symbiont-depressed siblings provided insight into effects of the symbiont on host development. Symbiont-deprived insects experienced significantly elongated second stadia (9.1 ± 2.9 d for symbiont-deprived vs. 7.9 ± 2.9 d for symbiont-endowed nymphs, p = 0.0001, Wilcoxon's rank with Bonferroni correction; Figure 12) and significantly slower linearized growth rates (p = 0.005, Welch 2-sample t-test), requiring more time to achieve larger sizes (standard length x pronotum width) compared to symbiotic siblings (Figures 12B,C). Further, surface-sterilized S. englemani nymphs exhibited symbiont-deprived ceca with observable differences in cecal morphology compared to control nymphs, using FISH microscopy (Figures 10D–F). In one case, a 1st instar nymph allowed to probe the surface of it's surface-sterilized egg, acquired Burkholderia, most closely related to soil-dwelling B. cepacia (KF974366), in replacement of the native symbiont, both of which were confirmed via 16S rRNA amplification and sequencing. Burkholderia was also observed to be among the bacteria associated with non-manipulated egg surfaces (7% of the community; Table 1; data not shown). Thus, while results assist in the resolution of the status of the symbiosis, more vigorous surface sterilization protocols may inform future observations of the impacts of symbiont deprivation on nymph ontogeny and morphometry.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.