Diet breadth and exploitation of exotic plants shift the core microbiome of Cephaloleia, a group of tropical herbivorous beetles

Specimen collection

Adult beetles (n = 38; Fig. 1) were collected in 2014–2016 at La Selva Biological Station, a 1,500 ha ecological reserve in a lowland tropical rainforest site in northern Costa Rica (10°26′N, 83°59′W; current Costa Rican Ministry of the Environment and Energy permit #R-026-2015-OT-CONAGEBIO). Adults were located by searching for rolled Zingiberales leaves that were then unrolled and beetles collected with forceps. Beetles with excised exoskeletons were preserved in 70% ethanol, at 4 °C, for later molecular analysis. Cephaloleia eggs (n = 5) were collected adhered to plastic after mating pairs were kept for a brief time in captivity in bags. Usually the adults laid eggs on the plastic, which was remote from where they spent most of their time (in a rolled leaf near the bottom of the bag, or in the zipper seals of the bag). The piece of plastic with attached egg was then surface sterilized and kept in ethanol until analysis. Any bacterial presence in the eggs was therefore attributed to vertical transmission by adults, and not due to adult contamination of the eggs. For bacterial cultivation, the digestive systems of the beetles were dissected in sterile 1× phosphate buffered saline (PBS) and homogenized using a ground glass mortar and pestle. The resulting homogenate was spread onto Trypticase Soy Agar (TSA) and incubated at ambient temperatures. Resulting bacterial colonies were selected and stored at −80 °C in 30% glycerol (in 1× PBS). All samples were transported back to the US for further processing.

Specimen identification

Adult beetles were identified based on diagnostic morphology and host plant (the latter mainly for specialist beetles that feed on only a single plant type; McKenna & Farrell, 2005; García-Robledo et al., 2013). In one cased we employed DNA analysis for identification; C. dilaticollis currently encompasses two cryptic species, one of which is a specialist and the other a generalist. Total genomic DNA was extracted using the Qiagen DNeasy Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. The cytochrome c oxidase I (COI) gene was amplified via the polymerase chain reaction (PCR) using the insect COI primers 1718F (5′-GGAGGATTTGGAAATTGATTAGTTCC-3′) and 3661R (5′-CCACAAATTTCTGAACATTGACCA-3′) according to McKenna & Farrell (2005). Successful PCR reactions, determined via electrophoresis, were cleaned using MultiScreen HTS plates (Millipore Corporation, Bedford, MA, USA) and sequenced via Laragen, Inc. (Los Angeles, CA, USA). Beetle species identification was confirmed based on COI sequences, upon consultation with Dr. Carlos García-Robledo (University of Connecticut).

Molecular microbiome analysis

Adult Cephaloleia beetles from six species found on native plants (n = 29) and three species found on invasive plants (n = 9) were examined for microbiome composition via 16S rRNA gene barcode sequencing (Table 1). As described above, total genomic DNA was extracted using the Qiagen DNeasy Kit (Qiagen, Valencia, CA, USA) according the manufacturer’s instructions. All extractions were from individual beetles, with the exception of C. dorsalis (Baly) which due to its small size, two to three individuals were pooled to achieve positive PCR results. PCR amplification of the bacterial 16S rRNA gene was performed using the specific primers, 515F (5′-GTGCCAG-CMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′; Caporaso et al., 2011). The thermal cycling profile used was as follows: an initial denaturation at 94 °C, then 45 s at 94 °C, 1 min at 50 °C, and 90 s at 72 °C, for 29 cycles, followed by 10 min at 72 °C. Successful PCR amplifications, assessed via electrophoresis, were pooled, in duplicate, and barcodes were added according to the Earth Microbiome Project (EMP; Caporaso et al., 2011); 5 μl of the amplicon product from PCR#1 was used as template in a five cycle, 25 μl reconditioning reaction with the same EMP-recommended conditions and the full EMP primers (515f_barcode: AATGATACGGCGAC-CACCGAGATC-TACACTATGGTAATTGTGTGCCAGCMGCCGCGGTAA; 806r_barcode: CAAGCAGAA-GACGGCATACGAGAT-X-AGTCAGTCAGCCGGACTACHVGGGTWTCTAAT), where X indicates a unique 12 bp barcode. Adding the barcode indices at the second step minimizes PCR bias that would result from employing long primers over many cycles (Berry et al., 2011). Further, the use of the “reconditioning” PCR for barcoding, as well as the pooling of duplicate amplifications ahead of barcoding, was an attempt to minimize PCR errors and bias, respectively (Kennedy et al., 2014). Samples were mixed together in equimolar amounts and purified in bulk through a Qiagen PCR Purification kit. At all PCR steps, amplification success and purity was checked by gel electrophoresis. Paired-end sequences (2x 250 bp) were generated from barcoded amplicon products at Laragen, Inc., on an Illumina MiSeq platform. At Laragen, the raw data was passed through a filter which demultiplexed the library into individual samples and removed any sequences which had >1 bp mismatch on the 12 bp barcode sequence, and assigned quality scores to each basepair call on every sequence. At the same time, adapter, barcode, and primer sequences were removed.

Sequence processing was performed in QIIME 1.8.0 (Quantitative Insights Into Microbial Ecology; Caporaso et al., 2010). Sequences were clustered at 99% similarity and a representative sequence from each cluster was assigned a taxonomic identification using parameter -m in QIIME and the Silva115 database. Via barcode amplicon sequencing, 13829–68837 sequences were recovered from each specimen (Table 1; Table S1). To avoid artifacts of sequencing depth, the number of sequence reads was standardized to 13829 sequences per specimen, based on the lowest sequence number for specimen “Cp_inv1” (C. placida on invasive white ginger; Table 1). The dataset was further cut off at 1% (i.e., the number of sequence hits for a single bacterial OTU across all 38 specimens must have been greater than 138 to be included). After this cutoff, sequences ranged from 9899 to 13324 per specimen (Table 1). Wolbachia was observed in 18 specimens, among three species (range of 0.1–56.6% for all sequences; C. belti avg 22% ± 18%, C. reventazonica avg 41% ± 17%, C. fenestrata avg 10% ± 13%,), but was removed from subsequent analysis based on its known prevalence in insects as a reproductive pathogen (Table S2). Sequences corresponding to chloroplasts and mitochondria were also removed from the data set. NMDS, ANOSIM, and SIMPER analyses were completed in Primer-E after square-root transforming the dataset and calculating Bray–Curtis similarities (Clarke & Warwick, 2001). An ANOSIM R value close to “1.0” suggests dissimilarity between groups. Close environmental and cultured relatives were chosen using top hits based on BLAST (www.ncbi.nlm.nih.gov). RStudio was used to perform ANOVA calculations using a script available at https://sites.oxy.edu/sgoffredi/Symbiosis_Lab/LabScripts.html.

Cephaloleia belti phenotypes and diagnostic PCR

Cephaloleia belti individuals (n = 45) were photographed and sized (length and width at pronotum) using imageJ (Schneider, Rasband & Eliceiri, 2012). The eight largest and eight smallest beetles were dissected for molecular analysis according to the methods described above. Total DNA of the body of the beetles was extracted using the Qiagen DNeasy Kit (Qiagen, Valencia, CA, USA) according the manufacturer’s instructions. For these 16 beetles, a diagnostic PCR using two different sets of pathogen-specific PCR primers was performed specifically for the bacterial genera Rickettsia (Rsp-F 5′-CGCAACCCTCATTCTTATTTGC-3′, Rsp-R 5′-CCTCTGTAAACACCAT-TGTAGCA-3′; Giulieri et al., 2012) and Spiroplasma (Spiro_16SF 5′-GGTCTTCGGATTGTAAAGGTCTG-3′, Spiro_16SR 5′-GGTGTGTACAAGACCCGAGAA-3′; Haselkorn, Markow & Moran, 2009) with the following thermal protocol: an initial 5 min denaturation at 94 °C, then 1 min at 94 °C, 1 min at 56 °C, and 1 min at 72 °C, for 29 cycles, and a final 5 min extension at 72 °C. Successful PCR amplification was determined via electrophoresis and confirmed to be Rickettsia or Spiroplasma via direct Sanger sequencing (Laragen, Inc., Los Angeles, CA, USA).

Bacterial cultivation

Initial bacterial suspensions in 30% glycerol stocks were re-grown on TSA plates at 30 °C. Growth was checked for morphological purity before being suspended in 40 μl of alkaline PEG (60 g of PEG 200 with 0.93 ml of 2 M KOH and 39 ml of water). This suspension was then heated to 96 °C for 20 min in order to lyse the bacterial cells and liberate the DNA. The 16S rRNA gene was then amplified directly using the general PCR primers 27F and 1492R (Lane, 1991) and the following thermal protocol: an initial 5 min denaturation at 94 °C, followed by 94 °C for 45 s, 54 °C for 1 min, and 72 °C for 90 s, for 29 cycles, and a final 72 °C extension for 10 min. Successful amplifications were checked via electrophoresis, cleaned using MultiScreen HTS plates (Millipore Corporation, Bedford, MA, USA), and sequenced at Laragen, Inc. Sequences were compared with the NCBI BLAST database to determine bacterial identity. Bacteria were propagated on TSA plates to ensure proper activity prior to metabolic testing. The ability to digest lactose/glucose, xylose, mannitol, and pectin was determined using phenol red agar (HiMedia, with 10% of each substrate). Protein digestion was determined using Litmus Milk tubes purchased from Carolina Biological Supply Company (Burlington, NC, USA). The ability to breakdown lipids was tested using an APIZYM analysis (bioMerieux, Inc., Durham, NC, USA), according to the manufacturer’s instructions.

Data availability

The raw barcode sequence data are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.5fj6t. Raw sequences were aligned and quality control for unidentified base pairs and chimeras was performed according to the specifics noted at https://sites.oxy.edu/sgoffredi/Symbiosis_Lab/LabScripts.html. The QIIME processed data are also available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.5fj6t. 16S rRNA sequences for bacterial isolates are available from GenBank under accession numbers MF776885–MF776899.

The limited core microbiome of Cephaloleia beetles

Using barcode 16S rRNA analysis, the microbiome of adults of six species of Cephaloleia, foraging on native plant diets, was characterized taxonomically (Table 1). Collectively, 168 bacterial OTU’s (“operational taxonomic unit;” defined as 99% sequence similarity) were recovered from all adult specimens examined (n = 29), while individual beetles generally associated with 47–152 OTUs (100 ± 31). Greater than 60% of Cephaloleia specimens contained a core group of eight bacterial OTUs, including three members of the Acinetobacter, two undefined Enterobacteriacea, Pseudomonas, Lactococcus, and a Comamonas (Figs. 2 and 3; Table S1). These eight bacterial OTUs comprised the majority of 16S rRNA sequences recovered from each Cephaloleia individual (up to 88%).

A single bacterial family, Moraxellaceae, dominated the microbiomes of all 29 specimens feeding on native plants, combined (representing 24% of the total recovered sequences). The genus Acinetobacter, in particular, comprised the vast majority of the Moraxellaceae sequences and accounted for 23% of the 16S rRNA sequences recovered overall. Of the 19 different Acinetobacter OTUs, three were responsible for 60% of the total Acinetobacter diversity and were each present in >75% of beetles (OTUs-131476, 21817, and 28305; Fig. 3, shown in purple; Table S1), suggesting them to be members of the core Cephaloleia microbiome. Acinetobacter OTU131476 was found in 22 of 29 specimens, and was 11% abundant (on average, for all sequences recovered in each of 29 beetles found on native plants; Fig. 3). This OTU was 98% similar to bacteria associated with both leaf cutter ants and fig wasps (GenBank accession #’s LN564930, HQ639556). Acinetobacter OTU21817 was also found in 22 of 29 specimens, represented 6% average abundance (Fig. 3), and was 100% similar to bacteria found in the midgut of a leafworm moth (GenBank accession # KU841476). Acinetobacter OTU28305 was found within 23 of 29 specimens, totaling 6% average abundance (Fig. 3), and was 100% identical to Acinetobacter baylyi (GenBank accession # NR115042), and others found in the rhizosphere.

Unidentified Enterobacteriaceae were also dominant in beetles found on native plants, representing 12% of the total recovered sequences. Of the 10 OTUs that comprised the Enterobacteriaceae within Cephaloleia beetles, a single OTU was responsible for 65% of the total Enterobacteriaceae diversity. This dominant Enterobacteriaceae OTU-79811 was present in >93% of beetle specimens, was 8% abundant on average (for all 29 beetles found on native plants; Fig. 3, shown in blue), and was 100% similar to Enterobacter/Klebsiella bacteria recovered from scarab beetles, sand flies, and pill bugs. Two additional OTUs (OTU127346 and OTU79806) each accounted for ∼15% of the remaining Enterobacteriaceae and were present in >15 of 29 specimens (Fig. 3). These OTUs were related to Citrobacter and Raoultella OTUs found in the microbiome of numerous insects, including honeybees (KR269812), scarab beetles (KT956239), and fruit flies (KX997073).

Three additional OTUs were highly prevalent in Cephaloleia microbiomes (present in ∼64% of individual beetles), including a Lactococcus OTU-157643 representing an average abundance of ∼4% (Fig. 3), a Pseudomonas OTU-126400 with an average abundance of ∼2% (related to bacteria recovered from mosquitoes and sand flies; KY041526; Li et al., 2016), and a Comamonas OTU-104860, also with an average abundance of ∼2% (most closely related to bacteria found in association with fruit flies; KX994588; Fig. 3; Table S1).

A large number of cultured isolates recovered from the digestive systems of Cephaloleia (81% of 37 isolated bacterial colonies) were members of the Moraxellaceae, Enterobacteriaceae, and Pseudomonadaceae, based on 16S rRNA gene sequencing. Several isolates had 16S rRNA sequences identical to the dominant bacteria identified via barcode 16S rRNA sequencing, including Acineto3 (=Acinetobacter OTU-28305), Entero4 (=Enterobacteriaceae OTU-127346), and Pseudo2 (=Pseudomonas OTU-126400). The Enterobacteriaceae were found to utilize plant-based compounds, including xylose and pectin (seven of nine isolates), mannitol (eight isolates), and lactose/glucose (all nine isolates). In contrast, none of the four Acinetobacter isolates in this study were able to digest these compounds, but instead uniquely displayed esterase C4, lipase C8, and lipase C14 capabilities (Fig. 4).

Diet breadth influences the microbiome of Cephaloleia beetles

Overall, the microbiome of specialist beetle species was significantly higher in diversity than generalists (2.6 ± 0.5 versus 1.9 ± 0.5, respectively; p = 0.0006, one-way ANOVA, Fig. 5). Measures of bacterial diversity (via the Shannon diversity index) were 0.8–2.5 for C. belti, 1.1–1.9 for C. reventazonica, 2.1–3.3 C. fenestrata, 2.0–3.0 for C. dorsalis, 2.4–2.9 for C. dilaticollis, and 1.5–3.0 for C. placida (Table 1). Further, NMDS ordination revealed the microbial assemblages of Cephaloleia to be strongly differentiated by diet breadth (i.e., generalist versus specialist; R = 0.74, p = 0.001, analysis of similarity (ANOSIM); Fig. 6A). SIMPER analysis implicated several bacterial families associated with this difference. For example, the Moraxellaceae, Enterobacteriaceae, and Pseudomonadaceae comprised a significantly higher percentage of the bacterial community in beetle species categorized as specialists (34%, 17%, and 6% of recovered sequences on average for specialist individuals, respectively) versus generalists (4%, 3%, 0.3%, respectively; all p < 0.0013, one-way ANOVA; Figs. 2 and 5). In contrast, results indicate that generalist Cephaloleia beetles were colonized by bacteria traditionally thought of as pathogens, including Rickettsia and Rickettsiella (discussed in more detail in supplemental information results). The Rickettsiaceae comprised a significantly higher percentage of the bacterial community in beetle species categorized as generalists, C. belti and C. reventazonica (28% of recovered sequences on average) versus specialists (only 0.8%; p < 0.0001, one-way ANOVA; Fig. 4; Table S1). For example, two Rickettsia OTUs (OTU7980 and OTU153335; Rickettsiaceae) were collectively present in only 11 out of 29 specimens, but comprised 11–26% abundance on average (for all 29 beetles found on native plants; Fig. 3). These OTU’s were 99% similar to Rickettsia found in leafhoppers (KR709154) and ticks (MF002591), to name a few. Similarly, a Spiroplasma OTU (Spiroplasmataceae) was present in 12 of 29 specimens, represented an average abundance of ∼12%, and was primarily observed in four individuals, two C. dorsalis and two C. belti (Figs. 2 and 3). This OTU57080 was 100% similar to the Spiroplasma associated with Drosophila. The incidence of Rickettsia and Spiroplasma was determined, via diagnostic PCR, to be highest in the smallest individuals of C. belti (81% prevalence, n = 8) versus the largest (56% prevalence, n = 8). Finally, a single OTU of Rickettsiella (Coxielliaceae) was only found in three C. belti individuals, but was highly abundant (∼31% on average; Fig. 3). This OTU43304 was 99% similar to Rickettsiella bacteria, also found in sand flies and ticks.

Bacteria associated with the eggs of Cephaloleia beetles found on native plants

Similar to the adults, Cephaloleia eggs were examined for microbiome composition via 16S rRNA gene barcode sequencing (Table 1). Four of the most common bacterial OTUs associated with adult beetles (Acinetobacter, Enterobacteriaceae, Comamonas, and Pseudomonas) were consistently observed in eggs (Fig. 7; Table 1), suggesting likely pseudovertical transmission from mother to offspring. NMDS analysis revealed a general overlap of the microbial communities associated with eggs and adults (ANOSIM R = 0.19, p = 0.060; Fig. 6B). The egg-associated microbiome of the generalist C. belti was significantly different from the adults (ANOSIM R = 0.95, p = 0.022), based mainly on a near absence of Rickettsia in the eggs (only 0.01% abundance; Fig. 7; Table 1).

Cephaloleia beetles foraging on invasive plants have distinct microbiomes

The specialist Cephaloleia species collected on invasive plants exhibited an apparent dysbiosis in their microbiome. NMDS ordination revealed a distinct bacterial community structure between the specialist beetles collected from native plants, compared to those on invasive plants, including C. placida and C. dilaticollis both on white ginger (Hedychium coronarium; ANOSIM R = 0.97, p = 0.001; Fig. 6C). The specialist species found on invasive plants possessed a lower abundance of both Moraxellaceae (21% average 16S rRNA abundance when on invasive plants compared to 41% for native feeders; p = 0.0472, one-way ANOVA) and Enterobacteriaceae (6% average abundance in beetles feeding on invasive plants, as opposed to 18% for those feeding on native plants; p = 0.0357; Fig. 2; Table S1). This decrease in typical microbiome membership may relate directly to a concomitant increase in microbiome members such as Brevinemataceae, which was significantly more abundant in both specialist species on invasive plants (21% average abundance; a single OTU43892, ∼99% similar to Brevinema found in insect larvae and other invertebrates, represented 97% of all Spirochaete sequences), compared to those on native plants (2%; p = 0.012; Fig. 2). The four specimens of the generalist species C. belti collected on pink banana (Musa velutina) also showed a slight shift in microbiome (ANOSIM R = 0.27, p = 0.05; Fig. 6B; Table S1), with significant increases in Spiroplasmataceae and Enterobacteriaceae abundance (ANOVA p = 0.05 for both; Fig. 2).