Deep divergence and genomic diversification of gut symbionts of neotropical stingless bees

Social bees harbor conserved gut microbiota that may have been acquired in a common ancestor of social bees and subsequently co-diversified with their hosts. However, most of this knowledge is based on studies on the gut microbiota of honey bees and bumble bees. Much less is known about the gut microbiota of the third and most diverse group of social bees, the stingless bees. Specifically, the absence of genomic data from their microbiota presents an important knowledge gap in understanding the evolution and functional diversity of the social bee microbiota. Here we combined community profiling with culturing and genome sequencing of gut bacteria from six neotropical stingless bee species from Brazil. Phylogenomic analyses show that most stingless bee gut isolates form deep-branching sister clades of core members of the honey bee and bumble bee gut microbiota with conserved functional capabilities, confirming the common ancestry and ecology of their microbiota. However, our bacterial phylogenies were not congruent with those of the host indicating that the evolution of the social bee gut microbiota was not driven by strict co-diversification, but included host switches and independent symbiont gain and losses. Finally, as reported for the honey bee and bumble bee microbiota, we find substantial genomic divergence among strains of stingless bee gut bacteria suggesting adaptation to different host species and glycan niches. Our study offers first insights into the genomic diversity of the stingless bee microbiota, and highlights the need for broader samplings to understand the evolution of the social bee gut microbiota. Importance Stingless bees are the most diverse group of the corbiculate bees and represent important pollinator species throughout the tropics and subtropics. They harbor specialized microbial communities in their gut that are related to those found in honey bees and bumble bees and that are likely important for bee health. Few bacteria have been cultured from the gut of stingless bees which has prevented characterization of their genomic diversity and functional potential. Here, we established cultures of major community members of the gut microbiota of six stingless bee species and sequenced their genomes. We find that most stingless bee isolates belong to novel bacterial species distantly related to those found in honey bees and bumble bees and encoding similar functional capabilities. Our study offers a new perspective on the evolution of the social bee gut microbiota and presents the basis to characterize the symbiotic relationships between gut bacteria and stingless bees.


25
Social bees harbor conserved gut microbiota that may have been acquired in a common 26 ancestor of social bees and subsequently co-diversified with their hosts. However, most of this 27 knowledge is based on studies on the gut microbiota of honey bees and bumble bees. Much 28 less is known about the gut microbiota of the third and most diverse group of social bees, the 29 stingless bees. Specifically, the absence of genomic data from their microbiota presents an 30 important knowledge gap in understanding the evolution and functional diversity of the social 31 bee microbiota. Here we combined community profiling with culturing and genome 32 sequencing of gut bacteria from six neotropical stingless bee species from Brazil. 33 Phylogenomic analyses show that most stingless bee gut isolates form deep-branching sister 34 clades of core members of the honey bee and bumble bee gut microbiota with conserved 35 functional capabilities, confirming the common ancestry and ecology of their microbiota. 36 However, our bacterial phylogenies were not congruent with those of the host indicating that 37 the evolution of the social bee gut microbiota was not driven by strict co-diversification, but 38 included host switches and independent symbiont gain and losses. Finally, as reported for the 39 honey bee and bumble bee microbiota, we find substantial genomic divergence among strains 40 of stingless bee gut bacteria suggesting adaptation to different host species and glycan niches. 41 Our study offers first insights into the genomic diversity of the stingless bee microbiota, and 42 highlights the need for broader samplings to understand the evolution of the social bee gut 43 microbiota. 44 Introduction of the host (16, 33). Therefore, it has been suggested that the core members of the microbiota 83 were acquired in a common ancestor of the social bees (20) and possibly co-diversified with 84 the host (16, 33). In addition, studies in the Western honey bee (Apis mellifera) have shown 85 that the diversification of the bee gut microbiota was also driven by adaptation to different 86 spatial and metabolic niches within the gut (27, 30, 33-35). For example, strains of closely 87 related sub-lineages of Lactobacillus Firm-5 and Bifidobacterium can coexist in individual 88 bees. They carry distinct gene sets for the breakdown and utilization of pollen-derived 89 carbohydrates which allows them to partition the available dietary glycan niches in the gut (27, 90 30, 34). 91 In contrast to the microbiota of honey bees and bumble bees, much less is known about the gut 92 microbiota of the third group of social bees, the stingless bees (Meliponini). Previous studies 93 have focused on determining the taxonomic composition of the gut microbiota of these bees 94 using 16S rRNA gene sequencing (15,17,20,(37)(38)(39). However, only a few bacteria have been 95 cultured from stingless bees (40-42) and, except for two strains of Bombilactobacillus Firm-4 96 recently isolated from bees from Australia (43), no genomic data is currently available for core 97 members of the gut microbiota of stingless bees. 98 With >500 described species, stingless bees present the largest and most diverse group of the 99 social bees (44, 45). They are naturally distributed throughout the tropical and subtropical 100 regions of Africa, Asia, Australia, and the Americas and exhibit great variation in morphology, 101 diet, foraging range, social structure, and nesting habits (44, 45). As host phylogeny and 102 ecology are both key determinants of gut microbiota composition (46-51), we hypothesize that 103 genomic studies on bacterial isolates will help us to understand the functional diversity of gut 104 bacteria of stingless bees and provide novel insights into the evolution of these bacteria across 105 social bees, specifically in respect of the possible co-diversification with the host. 106 To address these questions, we looked at the gut microbiota of six neotropical species of 107 stingless bees from Brazil: Frieseomelitta varia (Fv), Scaptotrigona polysticta (Sp), Melipona 108 fuliginosa (Mf), Melipona interrupta (Mi), Melipona seminigra (Ms), and Melipona lateralis 109 (Ml). We determined the composition of the gut microbiota of these bees using 16S rRNA gene 110 sequencing, established a comprehensive culture collection of bacterial isolates, and conducted 111 genome sequencing and comparative genomics to determine the phylogenetic placement, 112 genomic diversity, and functional capabilities of these bacteria relative to those previously 113 isolated from honey bees and bumble bees. 114

115
Six neotropical stingless bee species from Brazil harbor distinct gut microbiota 116 dominated by nine bacterial families. 117 We sampled three colonies of six stingless bee species (Fv, Sp, Mf, Mi, Ms, Ml) from a 118 meliponary located in the Amazonian rainforest near Manaus (Supplementary Table 1 Figure 1C). 163 However, the shared ASVs belonged to the nine predominant bacterial families and represented 164 a large fraction of the total number of reads per sample (53.3-99.7 %; except for the samples 165 of Fv and P. helleri: 10.6-15 % of the reads) (Supplementary Figure 1D). In particular, bees 166 sampled in the same country or belonging to the same bee genus shared the same ASVs, 167 explaining the clustering of these samples in the NMDS analysis. Together, these results show 168 that despite the large variability observed, the gut microbiota of most stingless bee species is 169 dominated by a few bacterial families and that bee species of the same genus, or with 170 overlapping geographic distribution, have similar community profiles at the 16S rRNA gene 171 level. 172 173

Establishment of a strain collection of gut bacteria isolated from stingless bees. 174
To enable genomic and experimental analyses of stingless bee gut bacteria, we established a 175 culture collection of bacteria isolated from Fv, Sp, Mf, Mi, Ms, and Ml. We plated 176 homogenized gut samples of the six bee species on eight different semi-solid media and under 177 three different atmospheres (microaerobic and anaerobic). This resulted in the cultivation of 98 178 distinct bacterial isolates (i.e. different 16S rRNA genotype or isolated from a different bee 179 species or colony) from 11 bacterial families (Figure 2A, Supplementary Table 4). Most 180 bacteria grew under both microaerobic or anaerobic conditions on generic growth media and 181 formed colonies after 2-4 days of growth. The 16S rRNA genotypes of the isolated strains 182 matched to 32 ASVs, accounting for 16 -87 % of the overall community of the six stingless 183 bee species and including many shared ASVs (Figure 2A and B). BLASTN  fruits. The percent identity of many of the blast hits was relatively low (<98%) suggesting that 192 the isolated strains potentially correspond to new bacterial species (Figure 2A Table 4). Genome comparisons with other bacteria, 205 including strains isolated from honey bees and bumble bees, showed that most stingless bee 206 gut bacteria had 80% average nucleotide identity with previously sequenced strains indicating 207 that we have isolated strains of novel bacterial species or genera ( Figure 2D). 208 Accordingly, genome-wide phylogenies based on single-copy orthologs showed that most 209 isolates formed deep-branching, stingless bee-specific lineages, exclusive of any previously 210 sequenced strain. However, consistent with the results of the 16s rRNA gene analysis, several 211 of these lineages were related to major phylotypes of the honey bees and bumble bee gut 212 microbiota ( Figure 3A Gilliamella, and Lactobacillus Firm-5, the stingless bee-specific lineages formed a 215 monophyletic clade with lineages of honey bee and bumble bee isolates ( Figure 3A-C and 3F). 216 Notably, in all three cases, the bacteria from stingless bees presented the earliest branching 217 lineages, i.e. the honey bee and bumble bee gut bacteria diverged after the split from the 218 stingless bee gut bacteria. While these results suggest that these bacteria have derived from a 219 common ancestor that was already adapted to social bees, the bacterial phylogenies were 220 incongruent with current phylogenies of the host, which show that the honey bees (Apini) 221 diverged before the split of stingless bees (Meliponini) and bumble bees (Bombini) (Figure  222 3E). A different pattern was observed for Bifidobacterium. In this case, strains isolated from 223 stingless bees, honey bees, and bumble bees were not monophyletic. In fact, the stingless bee 224 isolates belonged to a different clade than the honey bee isolates, while the strains isolated from 225 bumble bees belonged to either of them ( Figure 3D and 3G). Similarly, the two 226 Acetobacteraceae strains (ESL0695 and ESL0709) were not monophyletic with the honey bee 227 isolates of the genus Bombella, although they belonged to the same Hymenoptera-associated 228  two Streptococcaceae strains, which both also fell below the species-level ANI cut-off. 257 Notably, some Bifidobacterium and Lactobacillus Firm-5 strains that were isolated from the 258 same sample belonged to different ANI clusters, indicating that divergent bacterial species can 259 co-occur in the same host species and colony. Inversely, strains belonging to the same ANI 260 cluster were sometimes isolated from different bee species, suggesting that these bacterial 261 species clusters are not necessarily host-specific ( Figure 4B and C). 262 263

Core microbiota members in stingless bees have similar functional capabilities as in 264 honey bees and bumble bees. 265
To assess the functional potential of stingless bee gut bacteria, we determined the genomic 266 completeness of different metabolic pathways and functions in the genomes of the sequenced 267 strains and compared it to related bacteria isolated from honey bees, bumble bees, or from 268 elsewhere, and which were included in our phylogenomic analysis. We specifically looked at 269 energy and carbon metabolism, amino acid, co-factor, and nucleoside biosynthesis, as well as 270 secretion, motility, and adhesion ( Figure 5). Another example is GH13 which includes neopullulanases and α-amylases for the breakdown 286 of plant-derived starch. Together, these results suggest that stingless bee gut isolates of breakdown pollen or nectar-derived glycans, similar as previously reported for the 289 corresponding bacteria in the gut of honey bees or bumble bees. Notably, there was substantial 290 variation in the number and type of glycoside hydrolase family genes between divergent 291 strains, which goes in line with the extensive genomic diversity detected between stingless bee 292 gut isolates of these three phylotypes. A complete TCA cycle was only found in the genomes 293 of Neisseriaceae sp. ESL0693, Acinetobacter ESL0695, the Enterobacteriaceae, and the 294 Erwiniaceae. The same strains also harbored the most complete gene sets for oxidative 295 phosphorylation. Notably, Neisseriaceae sp. ESL0693 also lacked key genes in the EMP, PPP, 296 and the ED pathways and contained very few GH family genes ( Figure 5). This suggests that 297 this bacterium cannot utilize sugars and obtains energy via aerobic respiration, as previously 298 found for Snodgrassella isolates of honey bees and bumble bees (23). 299 300 Amino acid, nucleoside, and co-factor biosynthesis: Differences between stingless bee gut 301 isolates of different taxonomic groups were also found in terms of their biosynthetic potential. 302 Strains belonging to the Lactobacillus Firm-5 clade were auxotrophic for the production of 303 most amino acids (i.e. all except for Lys, Gln, and Asn) as well as purine and several co-factors 304 (e.g. heme, vitamin B6 and B12) ( Figure 5). Isolates of the Bifidobacteria, Streptococcaceae, 305 and Leuconostocaceae were also auxotrophic for many co-factors, but for much fewer amino 306 acids than Lactobacillus Firm-5. Interestingly, there was variation in auxotrophies among the 307 Bifidobacterial strains, especially for the production of purine, NAD+, Thr, Lys, Arg, Gly, and 308 chorismate. Other strains (such as those of Gilliamella, Snodgrassella, Acetobacteraceae, 309 Enterobacteriaceae, and Erwiniaceae) had fewer auxotrophies. Similar biosynthetic capability 310 profiles were found in related strains included in our phylogenies, which suggests that these 311 functional profiles are not specific to stingless bee gut bacteria but rather conserved across the 312 entire phylotype (Supplementary Figures 2-6). 313 Secretion, adhesion, motility: Secretion systems, pili, and flagella were mostly restricted to 315 the gram-negative bacteria of the isolated strains. Type 1, Type 5, and Type 6 secretion systems 316 were prevalent across these bacteria, whereas Type 2 and Type 4 secretion systems were only 317 present in a few strains ( Figure 5). Flagella were detected in the two Acetobacteraceae, all 318 Erwiniaceae, and Enterobacteriaceae strain ESL0689. Tad pili were not detected in any of the 319 analyzed bacteria, while Type IV pili components were mostly found in Neisseriaceae sp. 320 ESL0693 and Acinetobacter ESL0695, and to some extent also in Orbaceae, Erwiniaceae, and 321 Enterobacteriaceae ESL0689. Similar gene sets for secretion, adhesion, and motility were also 322 found in related gut bacteria from honey bees or bumble bees, as shown by the functional 323 profiles of all strains included in our phylogenies (Supplementary Figures 2-6). 324 325 Altogether, this first assessment of the gene content of the stingless bee gut bacteria show that 326 they have similar functional potential as isolates from honey bees and bumble bees suggesting 327 that they occupy similar ecological niches in the gut across social bees. 328

329
Previous findings suggested that the core members of the bee gut microbiota have been 330 acquired in a common ancestor of the social bees (20) and possibly co-diversified with their 331 hosts over millions of years (16, 33). However, the lack of genomic data from gut bacteria of 332 stingless bees has limited our view on the evolution of these specialized microbial 333 communities. With the establishment of isolate genomes of diverse stingless bee gut bacteria 334 our study fills an important knowledge gap and provides new lines of evidence that rule out 335 strict co-diversification between the core microbiota members and social bees. Another piece of evidence indicating that the microbiota across social bees may be more 373 variable than previously assumed comes from the observation that some of the designated core 374 members were not always detected in the sampled bees. For example, while Lactobacillus and 375 Bifidobacterium were prevalent across all six stingless bee species sampled in our study, they 376 were absent from the gut microbial communities of some of the previously sampled bee species 377 (15, 17, 37, 57). Likewise, stingless bees of the genus Melipona were shown to systematically 378 lack the two core members Snodgrassella and Gilliamella (37). Both taxa were also rare across 379 the four Melipona species analyzed in our study. However, three out of 12 analyzed colonies 380 had high abundances of Snodgrassella. This suggests that this bacterium is not completely 381 absent from this bee genus, but may be occasionally acquired from other bee species, varies in 382 prevalence depending on season, bee age, or development, or is restricted to only the Melipona 383 species analyzed in our study. 384 385 Finally, representatives of core members of the social bee gut microbiota were recently also 386 found in bees of the distant genus Xylocopa (Carpenter bees) (58, 59), suggesting that these 387 bacteria may have been associated with bees before the emergence of sociality, or that they 388 have a broader and less-specific distribution across bees than previously suspected. 389 390 In summary, our results together with previous findings indicate a rather dynamic evolutionary 391 background of the core members of the social bee gut microbiota. Rather than having strictly 392 co-diversified with their hosts extended periods of host-restricted evolution (and likely co-393 diversification in some lineages) seem to have been interrupted by host switches, and 394 independent symbiont gains and losses. Our observation that the stingless bee isolates 395 repeatedly form the most a sister group to bumble bee and honey bee isolates (for Gilliamella, 396 Snodgrassella, and Lactobacillus Firm-5) is intriguing given the host phylogeny. It may 397 suggest that these core members have an origin in stingless bees and then spread to the other 398 two groups, especially for Lactobacillus where two stingless bee isolates clades split before 399 the split between bumble bee and honey bee isolates. However, given the large diversity of 400 social and solitary bees, it is clear that the currently available datasets are insufficient to explain 401 the distribution and phylogenetic relationships of these gut symbionts across hosts. Broader 402 samplings of stingless bees, honey bees, and bumble bees, combined with genome-resolved 403 approaches, are needed to fully understand the diversity, distribution, and evolutionary 404 trajectories of social bee gut bacteria and to accurately reconstruct the ancestral bee 405 microbiome composition. Formal analysis should be applied to test for co-diversification. 406

407
The importance of sampling biases when assessing patterns of co-diversification between hosts 408 and their gut bacteria is highlighted by the analyses of Bacteroidaceae gut symbionts of 409 hominids. While originally reported to have co-diversified with their hosts (60), re-examination 410 with increased sampling disrupted the co-diversification pattern observed earlier (61). In 411 contrast, a recent study identified strong signals of parallel evolutionary history between seven 412 (out of 56 tested) gut bacterial taxa and human populations (62), and phylogenetic congruency 413 has also been found for certain stinkbug insects and their primary gut symbiont (63). This 414 demonstrates that co-diversification has occurred between certain gut bacteria and their hosts. 415 416 Besides offering new insights into the evolution of the social bee gut microbiota, our genomic 417 analysis also revealed the functional potential of major gut symbionts of the analyzed stingless 418 bee species. All isolates of Lactobacillus Firm-5, Bifidobacterium, and Gilliamella carried 419 genes for the saccharolytic fermentation of diet-derived carbohydrates. In contrast, 420 Snodgrassella ESL0689 lacked such functions in its genome, but instead harbored genes for 421 aerobic respiration. These results are consistent with findings from honey bees and bumble 422 bees (22, 23, 26-28, 30) and hence suggest that the core microbiota members occupy similar 423 ecological niches across the three groups of social bees. 424 425 Another parallel to findings from honey bees and bumble bees was the extensive genomic 426 divergence present among strains of the core members Lactobacillus Firm-5, Gilliamella, and 427 Bifidobacterium, even when isolated from the same host species. Moreover, we found genomic 428 variation in carbohydrate breakdown and amino acid and nucleoside biosynthesis functions 429 among these strains. This suggests that the diversification of these bacteria has not only been 430 driven by isolation into different host species but also by the adaptation to different ecological 431 niches in the gut, similar as shown for bumble bees and honey bees (28, 33, 35, 64). These 432 parallels may not be surprising as the dietary preferences of the analyzed stingless bee species 433 are similar to those of honey bees and bumble bees. In the future, it will be interesting to look without NaCl, BHIA (brain heart infusion agar) and SDA (Sabouraud dextrose agar). Plates 560 were incubated in two different conditions: in a microaerobic 5% CO2-enriched atmosphere 561 and in an anaerobic chamber (72% N2, 8% H2, 20% CO2), both at 34°C. After 2-7 days of 562 incubation, colonies of different size and appearance were picked and re-grown on the same 563 media and culturing conditions. Cryo-stocks of bacterial strains of interest were prepared by 564 harvesting bacterial biomass in liquid media corresponding to the solid growth media and 565 supplemented with 20% glycerol. For DNA isolation, bacteria were grown from the stocks, 566 and a single colony was picked and re-grown on fresh media before harvesting bacterial 567 biomass. 568

569
Genotyping 570 All colonies that were selected for culturing were genotyped by PCR and Sanger sequencing 571 of a 16S rRNA gene fragment. To this end, a small amount of bacterial material was transferred 572 to a lysis buffer (Tris-HCl 1M pH=7.5, EDTA 0.5M, SDS 10%) containing 2.5 μl lysozyme 573 (20mg/mL) and 2.5 μl proteinase K (20mg/mL), and incubated for 10 min at 37°C, 20 min at 574 55°C, and 10 min at 95°C. PCR was performed with universal bacterial primers that amplify 575 the V1-V5 region of the 16S rRNA gene (27F -AGRGTTYGATYMTGGCTCAG and 907R-576 CCGTCAATTCMTTTRAGTTT) using the following reagents and thermocycler program: 577 initial denaturing at 94°C for 5 min, followed by 32 cycles of denaturing at 94°C for 30 sec, 578 annealing at 56°C for 30 sec, and extension 72°C for 1 min, and a final extension at 72°C for 579 7 min. PCR results were checked on a 1% agarose gel. PCR reactions selected for Sanger 580 sequencing were purified using ExoSAP-IT TM (1μl ExoSAP 5x, 4μl ddH2O) with the 581 thermocycler program: 30 min at 37°C followed by 15 min at 80°C. Purified samples were 582 then sent to Eurofins® for sequencing. Sanger sequences were analyzed with Geneious suite 583 (Geneious®) and compared to GenBank at NCBI using BLAST tools (66). DNA isolation for Oxford Nanopore Technologies (ONT) sequencing was carried out using a 600 custom DNA extraction protocol for Gram-positive bacteria. Tubes were prepared with glass 601 beads and 160 μl of buffer P1 (Qiagen). Then bacteria were harvested and resuspended in these 602 tubes by intensive vortexing. Lysozyme (20 μl, 100 mg/ml) was added and after gentle mixing, 603 tubes were incubated at 56°C with shaking at 600 rpm for 30 min. Then 4 μl RNase A (100 604 mg/ml) were added to the tubes followed by 150 μl of buffer AL (lysis buffer, Qiagen). After 605 mixing by vortexing, tubes were incubated in a thermomixer (37°C, 900 rpm) for 20 min. Tubes 606 were centrifuged 10 min at 14'000 rpm to pellet the beads, the supernatant was transferred to 607 new tubes with 35 μl Na-acetate and 270 μl isopropanol and mixed by inverting. Following an 608 incubation of 1 h at 4°C, DNA was pelleted by centrifugation at 14'000 rpm for 10 min at 609 25°C. The supernatant was discarded, and the pellets were washed with 1 ml EtOH 80%. After 610 a second centrifugation (14'000 rpm, 10 min at 25°C) the ethanol was removed, and the pellet 611 left to dry at room temperature. DNA pellets were solubilized with 50 μl TER (10 mM Tris-612 HCl, 1 M EDTA, pH 8.0, 2 mg/ml RNase A) and tubes incubated at 37°C for 15 min. The 613 solution was then transferred to PCR tubes. 40 μl NGClean beads were added, and the solution 614