Introduction

Many insects are associated with microbial partners in host-beneficial symbioses [1,2,3,4]. Nourishing the symbiont in specialized organs and ensuring transmission to the next generation in exchange for essential nutrients allows the host to thrive on challenging and nutritionally imbalanced diets, like plant sap, wood, or vertebrate blood [5]. Such stable symbiotic associations commonly experience co-evolutionary dynamics, including co-adaptation and co-speciation [6, 7]. The continued isolation from environmental bacteria, as well as strong population bottlenecks during symbiont transmission, genetic drift, symbiont and host-level selection result in rapid genomic changes that lead to metabolic specialization of the nutritional symbiont [8, 9]. The outcome is a drastically reduced symbiont genome encoding besides general genetic information processing including DNA replication and repair, transcription, and translation, to sustain the symbiont’s metabolism under extensive host provisioning, only specific biosynthetic pathways to supplement essential nutrients that complement the host’s metabolism [10]. Symbionts can also be lost or replaced when they are no longer needed or capable of sufficiently supporting their host’s metabolism—as genome erosion can lead to reduced efficiency in the symbionts [11,12,13,14].

An alternative fate of obligate symbioses is the acquisition of a second symbiont that takes over part of the original symbiont’s function or provides additional metabolic capacities to the host. This can result in multipartite symbioses with co-obligate symbionts that exhibit complementary metabolisms, for example dividing pathways for the essential amino acids between two symbionts [15,16,17] or specializing on either essential amino acids or vitamin biosynthesis, respectively [18,19,20]. One example is the glassy-winged sharpshooter Homalodisca vitripennis (Hemiptera: Cicadellidae), where the β-proteobacterial symbiont Baumannia retains pathways for vitamins needed by the host, while the Bacteroidota symbiont Sulcia muelleri retains genes for the production of most essential amino acids, resulting in metabolic complementarity [20].

Co-obligate symbioses have so far only been functionally characterized across several lineages of Hemiptera [16, 21,22,23,24]. Recently, however, a dual symbiosis with two Bacteroidota bacteria was described for beetles of the family Bostrichidae (Coleoptera) [25]. The family Bostrichidae (Latreille, 1802) evolved between 170 [26] and 155 [27] Mya and consists of phytophagous beetles. While most species are xylophagous and live and develop within twigs, branches or trunks of dead or dying trees, some species are economically important pests of wood products or stored foods, including staple roots and cereals [28, 29]. Their endosymbionts are closely related to intracellular symbionts of cockroaches and Auchenorrhyncha, Blattabacterium spp. and Sulcia muelleri, respectively, and particularly to Candidatus Shikimatogenerans silvanidophilus (for simplicity we will omit the Candidatus nomenclature for endosmbionts from here on), the endosymbiont of the sawtoothed grain beetle Oryzaephilus surinamensis [25, 30]. The latter provisions tyrosine precursors to the beetle that complement the tyrosine-deficient diet of stored grain products [31].

Supplementation of precursors for tyrosine synthesis has been found to be important for cuticle melanisation and sclerotisation, as all of the cuticular crosslinking agents are derived from the aromatic amino acid tyrosine [32, 33]. Tyrosine-supplementing symbionts can inhabit the gut as in turtle ants of the genus Cephalotes [34, 35] or the bean weevil Callosobruchus maculatus [36], but most are located within bacteriomes, like Westeberhardia cardiocondylae (Enterobacteriaceae) in the tramp ant Cardiocondyla obscurior [37], the γ-proteobacterial symbionts Nardonella dryophthoridicola and Sodalis pierantonius in weevils [38, 39], as well as the Bacteroidota endosymbiont S. silvanidophilus in O. surinamensis [31]. The widespread occurrence of such symbioses provides evidence for this aromatic amino acid being a key nutrient for many insects to produce their strongly sclerotised and melanised exoskeleton, thereby increasing desiccation resistance and protection from predators and pathogens [31, 38,39,40,41].

Here we set out to functionally characterize the microbial symbionts of Bostrichidae as the first dual symbiosis of insects outside of the Hemiptera. We collected 28 beetle species and used available datasets of four additional species across five subfamilies and performed metagenome sequencing and fluorescence in situ hybridization. Based on the symbiont genomes and host mitochondrial genomes and nuclear markers, we reconstructed the molecular phylogenies of host and Bacteroidota symbionts. We demonstrate that (i) most bostrichids are associated with Shikimatogenerans whose genome is highly degraded, retaining almost only the shikimate pathway for tyrosine precursor provisioning besides general genetic information processing including DNA replication and repair, transcription, and translation; (ii) beetles of the genera Lyctus and Dinoderus are associated with a second Bacteroidota symbiont that encodes the capacity for lysine biosynthesis and nitrogen recycling from urea; and (iii) host and symbiont phylogenies exhibited a high degree of co-cladogenesis, indicating an ancient association that resulted in obligate mutual dependence and co-diversification. Our results shed light on the evolutionary dynamics of multipartite symbioses beyond the well-studied Hemiptera and highlight the importance of tyrosine provisioning and nitrogen recycling for the ecology of xylophagous beetles.

Results

We collected and sequenced the metagenomes of 28 beetle species of the family Bostrichidae and supplemented our dataset with four publicly available datasets from NCBI (Prostephanus truncatus 16 S rRNA [MF183960], Apatides fortis mitochondrial genome [FJ613421], Sinoxylon sp. SIN01 mitochondrial genome [JX412742], and Xylobiops basilaris transcriptome [SRR2083737]) (Supplement Tables 1 and 2). The resulting 32 species covered nine tribes within five of the nine subfamilies of Bostrichidae. For thirteen of the 32 species, we were able to assemble the full and circularized genome of the Bacteroidota endosymbiont Shikimatogenerans bostrichidophilus, with the longest closed symbiont genome being 200,377 bp and the shortest 172,971 bp in length, and an average GC content of 15.1% (Supplement Table 2). For ten additional species, we assembled draft genomes of Shikimatogenerans based on multiple contigs extracted from the metagenome assemblies via taxonomic classification, GC content filtering, as well as by manually searching for tRNAs and ribosomal protein genes as well as enzymes of the shikimate pathway of Bacteroidota bacteria. For two species (Dinoderus bifoveolatus & Prostephanus truncatus), we only retrieved the 16 S rRNA sequence, and for another one (Xylobiops basilaris) only the 16 S rRNA and aroA gene sequences of the endosymbiont. In the metagenome data of the remaining four species, we were not able to detect any sequence of Bacteroidota bacteria and the two final datasets contained only mitochondrial genomes.

As expected from a previous study [25], we found the genome of a second Bacteroidota endosymbiont, which we named Bostrichicola ureolyticus (see below), in some species of the subfamilies Lyctinae and Dinoderinae. In particular, we detected this co-obligate symbiont in all three Lyctus and two out of three Dinoderus species, but not in other members of the Lyctinae (Trogoxylon impressum) or Dinoderinae (Rhyzopertha dominica). The genomes of Bostrichicola were on average 337,500 bp in length and had an average GC content of 22.4%.

The metagenomic datasets were used to reconstruct the phylogeny of the host species (Fig. 1 and Supplement Fig. 1) based on either the assembled mitochondrial genomes (Fig. 1), or on 22 aligned and concatenated Benchmarking Universal Single-Copy Ortholog (BUSCO) genes [42] found across all species (Supplementary Fig. 2). Both phylogenies yielded very similar results, revealing two main clades of the Bostrichidae beetles, separating the Lyctinae and Dinoderinae from the Euderinae, Apatinae and Bostrichinae. However, the two phylogenies differed in the placement of Micrapate scabrata, which either grouped within the Sinoxylonini as a sister clade to the Xyloperthini (BUSCO genes) or as an outgroup to the Sinoxylonini and Xyloperthini (based on the mitochondrial genomes; Supplementary Fig. 2).

Fig. 1: Comparison of phylogenetic relationships between Bostrichidae beetle hosts and their Bacteroidota endosymbionts.
figure 1

Left: Bayesian phylogeny of 30 Bostrichidae beetle species inferred from concatenated nucleotide alignment of 13 mitochondrial genes. Right: Bayesian phylogeny of Bacteroidota symbionts of Bostrichidae beetles inferred from concatenated nucleotide alignment of 350 genes. Node numbers represent posterior probabilities of Bayesian analyses. Different colors denote the different host subfamilies with Bostrichicola (blue/cold colors) and Shikimatogenerans (red/warm colors). Host-symbiont associations are highlighted by connecting trapezoids between the phylogenies. Inferred, most parsimonious gain and loss events of both symbionts are indicated by circles (Shikimatogenerans) and triangles (Bostrichicola) on the host phylogeny. ° = Bostrichicola and/or Shikimatogenerans 16S rRNA was detected by FISH, PCR or metagenome sequencing, but no genomes could be assembled * = Shikimatogenerans could not be detected by any method.

Full size image

For the endosymbionts, the metagenome assemblies were used to generate a whole genome-based phylogeny (Fig. 1). This phylogenetic reconstruction based on 350 conserved genes confirmed the monophyly of the Shikimatogenerans endosymbionts of Bostrichidae and Silvanidae beetles and their close relationship to other insect-associated Bacteroidota bacteria, specifically to Blattabacterium spp. and Sulcia muelleri, which had been previously reported based on 16S rRNA gene phylogenies [25, 30, 43]. The Bostrichicola symbionts in the Lyctus as well as Dinoderus species clustered in a distinct, more basally branching monophyletic clade within the Bacteroidota. The two clades of Bostrichidae endosymbionts were separated by the Bacteroidota symbiont Shikimatogenerans silvanidophilus OSUR of the sawtoothed grain beetle Oryzaephilus surinamensis (Silvanidae) as well as the clade of Sulcia muelleri endosymbionts of the Auchenorrhyncha (Hemiptera). A second phylogeny of the endosymbionts based on the 16S rRNA sequences that allowed us to include more taxa revealed a highly similar distribution within the Bacteroidota (Supplementary Fig. 2). The main difference between both phylogenies (Supplementary Fig. 2) was the placement of the endosymbiont of Calopertha truncatula, which formed an outgroup to endosymbionts of the Sinoxylonini and of Xyloperthini in the 16S rRNA gene-based phylogeny, while it was placed basally in the endosymbiont clade of Sinoxylonini within the whole genome-based phylogeny.

A comparison between the endosymbiont phylogeny based on the 350 conserved genes and the host mitochondrial phylogeny showed a high degree of co-cladogenesis (Fig. 1). For S. bostrichidophilus, the only incongruence concerned the placement of Micrapate scabrata as an outgroup for the Xyloperthini and Sinoxylonini in the host phylogeny, but a grouping within the Sinoxylonini as a sister clade to the Xyloperthini in the symbiont phylogeny. However, the latter placement was also found in the BUSCO-based host phylogeny, strongly suggesting an incorrect placement in the mitochondrial phylogeny rather than a discrepancy between host and symbiont phylogenies (Supplementary Fig. 2). For B. ureolyticus, host and symbiont phylogenies were congruent on the host genus level, but the relationship of the three Lyctus species differed from that of their endosymbionts.

The genomes of both Bostrichidae endosymbionts were highly reduced and showed clear signs of genome erosion. Both endosymbionts retained genes involved in the cellular core processes of genetic information processing including DNA replication and repair, transcription, and translation (Supplement Fig. 3). In addition, Shikimatogenerans encoded all of the genes of the shikimate pathway except a shikimate dehydrogenase (aroE [EC:1.1.1.25]) (Fig. 2). Also, these genomes encoded the bifunctional aroG/pheA gene (phospho-2-dehydro-3-deoxyheptonate aldolase/chorismate mutase [EC:2.5.1.54 5.4.99.5]), capable of catalysing the Claisen rearrangement of chorismate to prephenate and the decarboxylation/dehydration of prephenate to phenylpyruvate in Escherichia coli [44].

Fig. 2: Reconstructed metabolism of Bacteroidota endosymbionts in Bostrichid beetles.
figure 2

A Reconstructed metabolism of the two D. porcellus endosymbionts S. bostrichidophilus DPOR and B. ureolyticus DPOR, inferred from genomic data. Enzymes and arrows in grey were missing in the genome annotation. Dashed arrows indicate transport processes without annotated transporters. B Comparison of the functional gene repertoires of Bacteroidota symbionts of Bostrichid beetles that could be assembled into full continuous genomes. Coloured boxes indicate the presence in the symbiont genomes. Box colours are based on KEGG’s categories (see legend for depicted categories).

Full size image

The genome of Bostrichicola encoded both urease α and γ subunits (ureC [EC:3.5.1.5]) to recycle nitrogen from urea, as well as a glutamate dehydrogenase (gdhA [EC:1.4.1.4]) that enables the integration of the resulting ammonium into the amino acid metabolism via glutamate (Fig. 2A). In addition, they encoded for an aspartate aminotransferase (aspB [EC:2.6.1.14]) to transfer the amino group from glutamate to oxaloacetate, as well as an almost complete diaminopimelate pathway to synthesize the essential amino acid lysine from aspartate. They also retained a methionine synthase to convert L-homoserine to L-methionine, a menaquinone biosynthesis pathway, and a fragmented folate biosynthesis pathway. By contrast, we did not find a single gene of the shikimate pathway to synthesize aromatic amino acids. However, the Bostrichicola genomes encoded for a complete fatty acid and peptidoglycan biosynthesis, albeit other cell envelope components apparently cannot be synthesized. The genomic data revealed no transporters, so it remains unknown how the symbionts exchange metabolites with the host and with each other (Fig. 2A). In addition, genes encoding signal transduction, cell surface structures, and motility were absent (Fig. 2B).

We also compared the set of genes that are not encoded in all genomes (Supplement Fig. 4). For Shikimatogenerans, it was particularly noticeable that mutL (DNA mismatch repair protein) was still present in the Dinoderinae+Lyctinae symbionts but had been lost in the Euderinae+Apatinae+Bostrichinae. For Bostrichicola, all genes for peptidoglycan biosynthesis (murA, murB, murC, murD, murE, murF, murG, mraY and mrcA) were still encoded in the Dinoderinae, whereas the symbiont of L. brunneus lost murC, murD and murG, and the symbiont of L. cavicollis lost all of the peptidoglycan biosynthesis genes. The comparison between the full genomes of all Shikimatogenerans and Bostrichicola strains showed a high degree of synteny within, but not between, the two symbiont genera (Fig. 3).

Fig. 3: Gene synteny of  Bacteroidota endosymbionts in Bostrichid beetles.
figure 3

Gene order comparison between S. bostrichidophilus and B. ureolyticus genomes that could be assembled into full contiguous genomes, showing a high degree of synteny within, but not between, the two symbiont genera. Grey shades show the percentage of identity between homologous proteins from different genomes (based on amino acid sequences). The phylogenetic tree on the left is based on the symbiont phylogeny displayed in Fig. 1.

Full size image

Based on the close phylogenetic relationship to Shikimatogenerans silvanidophilus OSUR and the presence of the shikimate pathway in the highly eroded genome, we propose the name ‘Candidatus Shikimatogenerans bostrichidophilus’ for the endosymbiont of Bostrichidae beetles. The genus name Shikimatogenerans refers to its ability to perform the shikimate pathway. Previous studies have shown that closely related Bacteroidota bacteria are also associated with other beetle families such as the Silvanidae and the Nosodendridae [25, 45]. Thus, we propose bostrichidophilus as a species epithet to indicate that this symbiont clade is associated with beetles of the family Bostrichidae. As all the symbionts encode highly similar genomes, we propose to add a four-letter abbreviation of the host species to denote the strain (first letter of the host genus and first three letters of the host species epithet), for example S. bostrichidophilus RDOM for the endosymbiont of Rhyzopertha dominica. For the second co-obligate endosymbiont found in Bostrichidae beetles of the subfamily Dinoderinae and Lyctinae, we propose the name ‘Candidatus Bostrichicola ureolyticus’. Its genus name refers to its association with Bostrichid beetles, while ureolyticus refers to its metabolic potential to recycle nitrogen from urea as inferred from the genomic data. In analogy to Shikimatogenerans, we propose to add a four-letter abbreviation of the host species to identify the strains, for example B. ureolyticus LBRU for the Bostrichicola endosymbiont of Lyctus brunneus.

The 16 S rRNA fluorescence in situ hybridisation with eight species, localized the bacterial symbionts intracellularly in bacteriomes in the abdomen of the host (Fig. 4). All bacteriomes are comprised of multiple syncytial bacteriocytes. The bacteriomes are located between the gut, fat body and reproductive organs, but without direct connection to any of these tissues. Bostrichid beetles of all subfamilies harbored one paired bacteriome with symbionts stained by a probe specific for members of the Shikimatogenerans symbiont clade [25, 31] (Fig. 4A, E–H). In addition, species of the genera Dinoderus and Lyctus contained a second pair of bacteriomes stained by a probe specific to B. ureolyticus [25] (Fig. 4B–D). The Bostrichicola-harboring bacteriomes were distinct in ultrastructure, but closely adjacent to the ones containing Shikimatogenerans, sometimes with direct physical contact (Fig. 4B).

Fig. 4: Tissue localization of Bacteroidota endosymbionts in Bostrichid beetles.
figure 4

Fluorescence in situ hybridisation micrographs of S. bostrichidophilus and B. ureolyticus in sections of (A) T. impressum, (B) L. cavicollis, (C) D. minutus, (D) D. porcellus, (E) R. dominica, (F) P. truncatus, (G) X. picea and (H) S. anale. Sections are stained with a Shikimatogenerans specific probe (magenta), a Bostrichicola specific probe (yellow), and DAPI targeting DNA (white). Scale bars represent 20 µm.

Full size image

Discussion

In this study, we characterized the intracellular bacterial symbionts across 32 species of auger or powderpost beetles (Coleoptera: Bostrichidae) and assessed their functional potential and co-speciation with their hosts based on comparative genomics. The functional characterization of this multipartite symbiosis enhances our understanding of the ecological relevance of microbial symbionts for a beetle family containing important wood and stored grain pest species and provides first insights into the evolutionary history and dynamics of co-obligate symbioses beyond the well-studied Hemiptera.

Based on a set of BUSCO as well as mitochondrial genomes, we reconstructed the first molecular phylogeny of Bostrichidae, after the morphological phylogeny by Liu & Schönitzer [46]. Both molecular datasets resulted in well supported phylogenies that were highly congruent and separated the Bostrichidae into two main clades: The Lyctinae and Dinoderinae grouped together, as did the Euderinae, Apatinae and Bostrichinae. The main difference was the placement of Micrapate scabrata, which clustered within the tribe Sinoxylonini in the BUSCO-based phylogeny, but as an outgroup to Sinoxylonini + Xyloperthini in the mitochondrial phylogeny (Supplement Fig. 1). Overall, our molecular phylogenies supported the earlier morphological work, with one major exception: Liu & Schönitzer placed the Euderiinae with a single monotypic genus as a basal branch of the Bostrichidae and suggested to even place them in a separate family. In our analyses, Euderia squamosa robustly grouped within the Bostrichidae in a separate branch between the Lyctinae/Dinoderinae and Apatinae/Bostrichinae, confirming the affiliation of the Euderiinae to the Bostrichidae.

Given their economic importance as pests of wood or stored grains, several species of Bostrichidae beetles were intensively studied almost a century ago and found to harbor intracellular symbionts [47,48,49,50], which were recently identified as members of the insect-associated Bacteroidota (Flavobacteriaceae) clade [25, 30]. In our broad phylogenetic survey, we were able to detect Shikimatogenerans bostrichidophilus in 26 out of 32 examined Bostrichidae species. For the few species where we could not detect any symbiont (Dinapate wrightii, Amphicerus bicaudus, Heterobostrychus aequalis, and Sinoxylon japanicum), three scenarios are possible. First, it is known for several beetle taxa that the endosymbiont is degraded or lost in male beetles after metamorphosis [38, 51,52,53]. In cases where symbiont-provided benefits are only relevant during larval development and/or metamorphosis, males can benefit from recycling their symbionts and symbiotic organs, given that they do not transmit the symbionts to their offspring [38, 54]. Thus, as individual beetle specimens in our study may have been males, the lack of symbionts in some species may represent the absence of symbionts only in adult males rather than in all individuals. Second, in some species, aposymbiotic individuals and even populations occur in the field, due to elevated sensitivity of the symbionts to environmental stressors like heat [49, 55], or possibly due to the application of certain agrochemicals like glyphosate that can eliminate symbionts encoding a sensitive aroA gene in the shikimate pathway [31]. Third, Shikimatogenerans may have truly been lost within these species. As we had only single specimens available for the species in which we failed to detect symbionts, we cannot confidently reject any of these hypotheses. However, based on the widespread occurrence of S. bostrichidophilus across Bostrichidae, we can conclude that this symbiont originated from a single acquisition event at the origin of the Bostrichidae and was retained by most if not all species.

The co-symbiont B. ureolyticus appears to be confined to the genera Lyctus and Dinoderus. It is unlikely that we missed the co-symbiont in other species of these two subfamilies, T. impressum and R. dominica. We had multiple individuals and life stages of R. dominica available as well as four specimens of T. impressum and never found any indication for a second bacteriome-localised symbiont, neither within our genomic datasets nor during FISH. Concordantly, previous studies [25, 30] did not report on a second symbiont in R. dominica, while it could already be discerned based on morphologically differentiated bacteriomes in Lyctus and Dinoderus species [25, 49].

Phylogenetic reconstructions based on either the symbiont 16S rRNA gene or 350 conserved genes resulted in highly congruent phylogenies that only differed in some of the deeper splits, which were better supported in the multi-gene phylogeny than the 16S rRNA gene phylogeny. However, both agreed well with previous phylogenetic analyses based on the 16 s rRNA gene alone regarding relations with better sampled other symbiotic Bacteroidota bacteria [25]. The resolution within S. bostrichidophilus benefitted here from the broader sampling, was better supported and separated symbionts of Lyctinae and Dinoderinae in different clades opposed to an intermingled clade. Consequences of the exclusive host associated lifestyle including long isolation within the different host lineages with restricted genetic exchange, different evolutionary rates, selection bias resulting in extremely low GC content, a limited number of sequenced genome and biased sampling of possible extant symbioses might cause phylogenetic artefacts [19]. However, the tight association of these obligate symbionts makes it also unlikely to miss major evolutionary events. While evolutionary relationships between symbiont species or clades have to be considered with care, the relationships within the respective species are likely relatively robust. The endosymbiont and the host phylogenies showed a high degree of co-cladogenesis, strongly supporting single acquisition events for each symbiont and subsequent co-cladogenesis with the host, as has been found for many obligate symbionts as well as some host-specific parasites [56, 57].

Within the Bostrichoidea, the Bostrichidae split from the Ptinidae between 170 [26] and 155 Mya [27] in the Jurassic period, which roughly coincides with the inferred age of S. bostrichidophilus (274-158 Mya) based on a bacterial phylogeny calibrated with estimated origins of symbiotic associations with insects [25]. As closely related families within the Bostrichoidea (Dermestidae, Ptinidae) [27] are not known to harbor bacterial endosymbionts but in some cases (Anobiinae) associate with yeast-like endosymbionts (reviewed in [58]), S. bostrichidophilus was likely acquired by the ancestor of the Bostrichidae. The symbiosis with Bostrichicola is of somewhat more recent origin, around 100 Mya [25], and occurred in the ancestor of the Dinoderinae and Lyctinae (plus possibly Psoinae and Polycaoninae, for which we were unable to obtain specimens), and was then at least lost in Rhyzopertha, Prostephanus and Trogoxylon. Although the alternative scenario of two independent acquisitions in the Dinoderinae and Lyctinae seems possible, the high degree of genome synteny between the symbionts of both genera renders a single acquisition much more likely.

Both endosymbionts are characterized by extremely small, heavily eroded and A + T-biased genomes with very limited biosynthetic capabilities, akin to other strictly vertically transmitted symbionts [59] as well as other intracellular genetic elements [60]. The genome of S. bostrichidophilus encodes for the shikimate pathway to synthesize precursors of aromatic amino acids. Of the seven canonical genes in the shikimate pathway (aroG, aroB, aroD, aroE, aroK, aroA, aroC), the S. bostrichidophilus genome only lacks the gene for shikimate dehydrogenase (aroE), which catalyses the reversible reduction of 3-dehydroshikimate to shikimate. However, the shikimate pathways of Nardonella EPO, the endosymbiont of the sweetpotato weevil Euscepes postfasciatus (Curculionidae: Cryptorhynchinae) [39, 61] and Carsonella ruddii, the endosymbiont of the gall-forming psyllid Pachypsylla venusta (Aphalaridae: Pachypsyllinae) [62], as well as the closely related S. silvanidophilus OSUR in the sawtoothed grain beetle O. surinamensis also lack aroE, but remain functional [31, 39], indicating that the function of aroE is taken over by other enzymes of either host or endosymbiont origin. Hence, S. bostrichidophilus is inferred to transform phosphoenolpyruvate (PEP) and erythrose-4-phosphate (E4P) to prephenate/chorismate via the shikimate pathway [63, 64], which can then be converted by the host to the aromatic amino acid tyrosine [65]. As all of the cuticular crosslinking agents as well as precursors for black and brown pigments (melanins) are derived from tyrosine [32, 33], it constitutes the key metabolite in cuticle synthesis, melanisation and sclerotisation, thereby strongly affecting the physicochemical properties of the cuticle [66]. Concordantly, symbiont-mediated supplementation of tyrosine precursors enables the production of a thicker, stronger, and darker cuticle and thereby enhances protection against desiccation, predation, and infection by entomopathogens [25, 39, 40, 67].

The reduced genome of the co-obligate symbiont B. ureolyticus encodes the genes for urea recycling and the diaminopimelate pathway to synthesize lysine. In addition, B. ureolyticus retained partial pathways to convert intermediates of the lysine biosynthesis into methionine, folate and menaquinone [5, 68], and it can synthesize some components of the cell envelope (fatty acids and peptidoglycan). However, several of these pathways exhibit differential erosion between the Lyctus and Dinoderus symbionts, but even between the symbionts of one of the genera (Supplemental Fig. S4). Especially the pathways for synthesis of the cell envelopes are intriguing, as cell wall synthesis has been demonstrated to be complemented and controlled by host encoded genes and inter-symbiont exchange of metabolites in hemipteran symbionts [24, 69, 70].

Nitrogen recycling is well documented within some Bacteroidota endosymbionts [71,72,73] and can be an important benefit for insects developing in nitrogen-limited diets [28, 29, 74,75,76]. In Bostrichid beetles, the recycling of urea as a source of amino groups is likely important for the formation of tyrosine, but also other amino acids and amino acid-derived components of the cuticle like N-acetyl-glucosamine, the monomer of chitin [35]. However, why B. ureolyticus retained a lysine biosynthesis pathway is less clear. Lysine constitutes an important amino acid of cuticular proteins, as its ε-amino group represents an anchor point for cross linking [77]. In addition, grain diets, but also staple roots are specifically limited in lysine [78, 79], so symbiont-mediated lysine supplementation could be an important benefit for the stored product pest beetles of the genus Dinoderus, but also other species of the Dinoderinae subfamily. However, a deeper understanding of why certain genera of Bostrichidae benefit from such provisioning and thus retain Bostrichicola while others do not, is currently hampered by the scarcity of information on the ecology of most Bostrichidae [46].

Based on our phylogenetic analyses, Shikimatogenerans and Bostrichicola are derived from the same ancestor as Blattabacterium spp., Walczuchella monophlebidarum and Uzinura diaspidicola, but then diverged and evolved different functional specializations [25, 31, 45, 80]. S. silvanidophilus OSUR [31] – the sister taxon of S. bostrichidophilus – retained the shikimate pathway as well as two urease subunits putatively involved in nitrogen recycling [31] - and there is evidence for nitrogen recycling in Blattabacterium [71, 73, 81] and Walczuchella [72], so urea catabolism seems to be a widespread and possibly ancestral benefit provided by Bacteroidota symbionts of insects.

Beneficial associations with two metabolically complementary symbionts synthesizing essential amino acids and vitamins have thus far only been functionally characterized for multiple different lineages of plant sap-feeding Hemiptera [16, 18, 21, 68, 82,83,84,85]. In these cases, however, the co-obligate symbionts usually originated from different classes or phyla, with the exception of some associations between Sternorrhyncha and two co-obligate γ-proteobacterial symbionts [86]. The secondary symbiont is in these cases largely considered to replace eroded metabolic capabilities of the primary symbiont [20]. To our knowledge, the Bostrichidae are thus far unique in containing species that harbor two Bacteroidota endosymbionts, which are closely related but diverged to metabolically complementary symbionts. Further, based on the metabolic capabilities of different Shikimatogenerans species and the timepoint of acquisition, it is unlikely that Bostrichicola was acquired to complement lost capabilities of Shikimatogenerans, both rahter provided from the beginning complementary nutrients. Whether the two symbionts’ contributions provide independent benefits or even interact synergistically remains to be determined. This finding highlights the versatile nature of symbioses with Bacteroidota bacteria that are emerging as widespread beneficial symbionts across at least four beetle families (Silvanidae: [25, 43], Coccinellidae: [87, 88], Nosodendridae [45], and Bostrichidae: this study and [25]) as well as at least two other insect orders [20, 25, 89]. The repeated independent acquisitions of these specific clades of Bacteroidota symbionts suggest that these bacteria were once specialized in establishing lasting infections in insects, akin to Wolbachia [90] or Sodalis [9]. Considering that a basal clade of Bacteroidota endosymbionts are male-killing endosymbionts in different ladybird beetles [87, 88, 91], it is conceivable that the ancestral success of Bacteroidota symbionts relied on reproductive manipulation to spread in insect populations. Later, some of these originally parasitic associations may have evolved partially mutualistic phenotypes, but retained the ability to readily infect novel hosts when transfer by vectors like parasitoids occurred. Similar phenotypes can currently be observed by above mentioned Wolbachia and Sodalis bacteria. These diverged lines might represent the source of the extant, highly diverged, beneficial symbioses in the different insect orders based on the symbionts’ capacity to provision varying limiting nutrients. An analogous convergence can be seen in the vitamin-provisioning Wolbachia symbionts of bedbugs that also evolved a bacteriome localization [92]. Given the sometimes severe fitness consequences of reproductive manipulation on the host and the ensuing evolutionary arms race between host and parasite [93, 94], the parasitic interactions may have been evolutionarily labile. By contrast, interactions that evolved towards mutualism likely remained long-term stable, which may explain the bias towards beneficial interactions observed in extant insect-associated Bacteroidota.

Material and methods

Insect collection

Specimens of 28 species were collected or provided by experts in the field from Germany, the Czech Republic, Yemen, the United Arabic Emirates, the United States of America, Japan, and New Zealand (Supplement Table 1), in compliance with the Nagoya protocol. In addition, four publicly available data sets of Bostrichid beetles were retrieved from NCBI (SRR2083737, MF183960, FJ613421 and JX412742).

Symbiont genome sequencing, assembly, and annotation

Total DNA was isolated using the Epicentre MasterPure Complete DNA and RNA Purification Kit (Illumina Inc., Madison, WI, USA) including RNase digestion, or the QIAGEN Genomic-tip kit using 20/G columns (Qiagen, Hilden, Germany). Short-read library preparation and sequencing were performed at the Max-Planck-Genome-Centre Cologne, Germany (SRR19201352 - SRR19201388) on a HiSeq3000 Sequencing System (Illumina Inc., Madison, WI, USA), or at CeGaT on a HiSeq2500 Sequencing System (Tübingen, Germany) or a MiSeq (Illumina Inc., Madison, WI, USA) of AIST Japan (DRR414867). Adaptor and quality trimming was performed with Trimmomatic [95].

Long-read sequencing for D. porcellus (SRR19201386 and SRR19201352) and L. brunneus (SRR19201357) was performed on a MinION Mk1B Sequencing System (Oxford Nanopore Technologies (ONT), Oxford, UK). Upon receipt of flowcells, and again immediately before sequencing, the number of active pores on flowcells was measured using the MinKNOW software (v18.12.9 and 19.05.0, ONT, Oxford, UK). Flowcells were replaced into their packaging, sealed with parafilm and tape, and stored at 4 °C until use. Library preparation was performed with the Ligation Sequencing Kit (SQK-LSK109, ONT, Oxford, UK) and completed libraries were loaded on a flowcell (FLO-MIN106D, ONT, Oxford, UK) following the manufacturer’s instructions. PacBio long-read sequencing of D. porcellus (SRR19201385) was performed at the Max-Planck-Genome-Centre Cologne, Germany on a Sequel II system (PacBio, Menlo Park, CA, USA).

Quality-controlled long reads were taxonomy-filtered using a custom-made kraken2 database [96, 97] containing the publicly available genomes of Bacteroidota bacteria to extract beetle-associated Bacteroidota sequences using the supercomputer Mogon of the Johannes Gutenberg-University (Mainz, Germany). Assembly of Illumina reads was performed using SPAdes (v3.15.0) with default settings [98]. Additional hybrid assemblies with long-read libraries were performed using Masurca (v) [99] using the app MaSuRCA Assembler - v3.2.9 in KBase [100]. The resulting contigs were binned using BusyBee Web [101], and screened by GC content, coverage and taxonomic identity for Bacteroidota bacteria. Assemblies were in addition manually screened by mapping contigs to Shikimatogenerans and Bostrichicola genomes and by mapping representative pathway gene sequences of both symbionts onto entire libraries. The extracted contigs were de novo assembled in Geneious Prime 2019 (v2019.1.3, https://www.geneious.com). The resulting contigs were then automatically annotated with PROKKA [102] using the app Annotate Assembly and Re-annotate Genomes (v1.14.5) in KBase [100]. Completeness was assessed with checkM [103] and by comparing number of ribosomal proteins and tRNAs with genomes of other Bacteroidota bacteria with reduced genomes. Synteny analysis of complete endosymbiont genomes was performed using Clinker with default settings [104].

Fluorescence in situ hybridisation

Endosymbionts of D. minutus, D. porcellus, L. cavicollis, P. truncatus, R. dominica, S. anale, T. impressum and X. picea were localised by fluorescence in situ hybridisation (FISH) on semi-thin sections of adult beetles, targeting the 16S rRNA sequence. Adult beetles were fixed in 80% tertiary butanol (Roth, Karlsruhe, Germany), 3.7% paraformaldehyde (Roth, Karlsruhe, Germany) and 3.7% glacial acetic acid (Sigma-Aldrich, Taufkirchen, Germany) for 2 hours, followed by post-fixation in alcoholic formaldehyde (3.7% paraformaldehyde and 80% tertiary butanol). After dehydration, the specimen were embedded in Technovit 8100 (Kulzer, Germany)100 and cut into 8 µm sagittal sections using a Leica HistoCore AUTOCUT R microtome (Leica, Wetzlar, Germany) equipped with glass knives. The obtained sections were mounted on silanised glass slides. For FISH, each slide was covered with 100 µL of hybridization mix, consisting of hybridization buffer (0.9 M NaCl, 0.02 M Tris/HCl pH 8.0, 0.01% SDS; Roth, Germany) and 0.5 µM of the Shikimatogenerans bostrichidophilus-specific probe (5′-CTTCCTACACGCGAAATAG-3′ [25]) labelled with Cy5, as well as the Bostrichicola ureolyticus-specific probe (5′-TACTCGATGGCAATTAACAAC-3′ [25]) labelled with Cy3. DAPI (0.5 µg/mL) was included as a general counterstain for DNA. Slides were covered with glass cover slips and incubated in a humid chamber at 50 °C overnight. After washing and incubating them for 20 minutes at 50 °C in wash buffer (0.1 M NaCl, 0.02 M Tris/HCl, 5 mM EDTA, 0.01% SDS), they were washed in deionized water for 20 minutes, dried and mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA). The sections were observed under a Zeiss AxioImager.Z2 equipped with an Apotome.2 (Zeiss, Jena, Germany) and illuminated by a SOLA Light Engine (Lumencor, Beaverton, OR, USA).

Phylogenetic analyses

We generated phylogenetic trees based on the metagenome data generated from our Bostrichid taxa (SRR19201352 - SRR19201388) as well as three Bostrichidae shotgun sequencing and one 16S rRNA datasets available on NCBI (SRR2083737, MF183960, FJ613421 and JX412742).

A phylogenetic tree of the mitochondrial genes of the hosts was reconstructed by assembling the mitochondrial genome using NOVOPlasty [105] and MitoZ [106] and afterwards annotating them with Mitos [107] (http://mitos.bioinf.uni-leipzig.de/index.py). Subsequently, 13 mitochondrial genes were translated and aligned using MUSCLE [108] (v3.8.425) as implemented in Geneious Prime 2019 (v2019.1.3, https://www.geneious.com). Additionally, we generated a second (codon-based) nucleotide alignment based on Benchmarking Universal Single-Copy Orthologs (BUSCO) using a custom pipeline [42] to extract the genes from the metagenome datasets. BUSCO analysis was performed for each dataset using the insecta_odb10 database (1,658 genes) to extract BUSCO genes that were found across all species [109]. The corresponding nucleotide sequences were then extracted and aligned with MAFFT [110] with --auto and default options. Gaps in the resulting alignment were then trimmed from the alignment using trimAl (v1.2), accepting 5% gaps for each position [111]. Afterwards, the aligned nucleotide sequences for each taxon were concatenated.

For the phylogenetic analyses of the intracellular symbionts of Bostrichid beetles, coding sequences were extracted from the genomes, aligned based on the nucleotide sequence with MAFFT [110], and concatenated in Geneious Prime 2019 (v2019.1.3, https://www.geneious.com). Additionally, beetle symbiont 16 S rRNA sequences were aligned to representative Bacteroidota 16 S rRNA sequences obtained from the NCBI database, using the SILVA algorithm [25, 112, 113]. Since complete genomes were not available for some of the species, the 16 S rRNA alignment allowed us to incorporate a larger number of species in the phylogenetic analysis, albeit at a lower resolution due to the limited amount of information contained in this single gene.

Phylogenetic reconstructions for all alignments were done by Bayesian inference applying a GTR + G + I model using MrBayes (v3.2.7) [114,115,116,117]. The analysis ran for 10,000,000 generations with a “Burnin” of 25% and tree sampling every 1,000 generations. We confirmed that the standard deviation of split frequencies converged to <0.01. The obtained trees were visualized using FigTree (v1.4.4, http://tree.bio.ed.ac.uk/software/figtree/).