Abstract How gut symbionts contribute to host adaptation remains largely elusive. Studying co-diversified honeybees and gut bacteria across climates, we found cold-adapted species (Apis mellifera, A. cerana) exhibit enhanced genomic capacity for glucose, pyruvate, lipid and glucuronate production versus tropical species. Metagenomics revealed Gilliamella as the most enriched gut bacterium in cold-adapted bees. Germ-free honeybees inoculated with the Gilliamella from A. cerana showed increased activity, body temperature and fat storage upon cold exposure. Saccharide metabolomics demonstrated higher hindgut glucose levels in Gilliamella-colonized A. mellifera versus germ-free bees, and in A. cerana versus three sympatric tropical species. Although Gilliamella can hydrolyze β-glucan into glucose, cultural experiments suggest it preferentially degrades glucuronate to pyruvate. In turn, monocolonized bees upregulated hindgut glucose/pyruvate utilization while increasing glucuronate provision, suggesting nutritional complementarity. Gilliamella’s transporter genes predominantly target ascorbate (a glucuronate derivative), which is elevated in inoculated hindguts. Accordingly, Gilliamella converts ascorbate to D-xylulose-5P (promoting lipogenesis), while showing reduced growth on glucuronate/ascorbate versus glucose, potentially minimizing glucose competition with hosts. We revealed a highly coordinated host-symbiont metabolic synergy enhancing host energy acquisition for cold adaptation. Subject terms: Evolution, Microbiology Introduction Given the crucial roles of gut symbionts in host health^[46]1 and adaptation^[47]2, study of the underlying host-symbiont interactions has become a central focus in evolutionary and ecological research^[48]3,[49]4. These interactions are highly context-dependent^[50]5, necessitating further studies to identify overarching principles. While short-term host-symbiont interaction dynamics at individual and population levels have been extensively examined across taxa^[51]5, understanding their long-term development trajectories during host adaptation and speciation remains challenging. Nevertheless, environmental adaptation offers a powerful framework for examining how natural selection has shaped the host and its interactions with symbionts. Stable mutualistic relationships can enhance host fitness over generations, making them highly heritable^[52]6. Co-divergence and host-specificity are fundamental for interactions to maintain over generations^[53]5, potentially driving host speciation through adaptive divergence^[54]6,[55]7. In this context, Apis honeybees can serve as an excellent model for investigating the long-term development of host-gut symbiont interaction. Among extant Apis species (Fig. [56]1A), only the western (A. mellifera) and eastern (A. cerana) honeybees have successfully expanded their natural habitats to temperate regions (Fig. [57]1B), becoming essential pollinators worldwide^[58]8. Earlier diverged dwarf honeybees (A. andreniformis, A. florea) and giant honeybees (A. dorsata, A. laboriosa) are confined in (sub)tropical areas (Fig. [59]1B). In contrast to these tropical honeybees, A. mellifera and A. cerana both exhibit a higher capacity for heat production^[60]9 (Fig. [61]1A), yet the underlying physiological mechanisms remain unclear. In the context of cold adaptation, various animals^[62]10–[63]12 and gut bacteria^[64]13–[65]15 have shown genomic traits that might improve thermogenesis. However, existing research has primarily focused on either the host or the bacteria, while the combined contributions of this symbiotic system remain poorly characterized. Fig. 1. Phylogeny and distribution of Apis honeybees. [66]Fig. 1 [67]Open in a new tab A Schematic diagram of the honeybee phylogeny. The six species studied in the present work are shown with photos. Four of them have been examined previously^[68]9, characterized by varied capacities in heat production (numbers represent heat production per mass, W/kg). B Native distributions of honeybee species. The same honeybee species is marked in Panels (A, B) using the same color code. YN Yunnan Province, China; JL Jilin Province, China. Honeybees are well-established models for gut microbiome research, harboring core bacterial genera such as Gilliamella, Snodgrassella, Lactobacillus, and Bifidobacterium in the hindgut^[69]16. These bacteria have codiversified with honeybees for 80 million years^[70]17 and play vital roles in host health^[71]16. However, their contributions to honeybee adaptation remain underexplored. Previous investigations based on metagenomics have shown that the gut bacterial communities are host-specific among different honeybees^[72]17,[73]18. While Lactobacillus and Bifidobacterium are prevalent across varied honeybees (albeit differ at the abundance level), other bacteria show distinct host patterns: Snodgrassella is absent from A. dorsata and A. florea; Gilliamella is lacking in dwarf bees; and Dysgonomonas and Saezia are confined to A. dorsata. However, how variations in gut community structures and the enrichment of specific bacterial members affect the host remains unclear. Gut microbiota and metabolite changes have been associated with thermogenesis in mammals, such as Akkermansia in mice^[74]19 and propionate in gerbils^[75]20. We therefore asked whether some honeybee gut bacteria might have contributed to host cold adaptation during independent northward expansions of A. mellifera and A. cerana. By comparing both hosts and symbionts of cold-adapted honeybees to their closely related tropical relatives, we aim to understand how natural forces, such as temperature, act on the “holobiont”, including the host and its microbiota. The honeybee gut microbiota plays a crucial role in dietary saccharide digestion^[76]21–[77]23. Notably, the core bacterial members Gilliamella and Bifidobacterium hydrolyze pectin and hemicellulose. Therefore, we hypothesize that carbohydrate metabolic interplay is critical for energy utilization in heat production. To investigate whether the underlying cold adaptation mechanisms may involve host-symbiont interactions, we analyzed the gut microbiome, bacterial genomes, and host genomes for six Apis species, including two cold-adapted and four tropical species. Additionally, we conducted in vitro and in vivo experiments to assess the impact of gut bacteria on thermogenesis and identified beneficial metabolites exchanged between the symbiotic partners. Our results revealed that colonization by the gut bacterium Gilliamella significantly promotes honeybee lipogenesis and thermogenesis, uncovering a metabolic synergy between the host and the symbiont. Cold-resistant honeybees exhibit enhanced abilities for lipogenesis and the acquisition of glucose and pyruvate, while increasing the provisioning of glucuronate and ascorbate to Gilliamella as a carbon source. In turn, Gilliamella hydrolyzes pollen-derived β-glucan into glucose and converts glucuronate and galacturonate into pyruvate. Surprisingly, Gilliamella supplies the host with key energy substrates, glucose and pyruvate, while sustaining restrained growth via degrading glucuronate and ascorbate. By investigating host-symbiont interactions from the perspectives of both parties, we gain deeper insights into how this symbiosis facilitates adaptation to challenging environments. Results Gilliamella is most significantly enriched in cold-adapted honeybee guts We performed metagenome shotgun sequencing on the two cold-resistant honeybees (A. mellifera & A. cerana) and four tropical honeybees (dwarf honeybees A. andreniformis, A. florea, and giant honeybees A. dorsata, A. laboriosa). A total of 70 samples from 168 bee hindguts were obtained from Jilin and Yunnan Provinces, China (Fig. [78]1B, Fig. [79]2A, Fig. [80]S1a, Supplementary Data [81]1) and used to investigate the characteristic gut microbiota of cold-adapted honeybees. The gut microbiota composition confidently divided the six honeybee species into three distinct groups (partial least squares test: Q2Y = 0.783), in line with their phylogenetic relationships (Fig. [82]2A, Fig. [83]S1b, c). To test whether the gut microbiota composition is influenced by cold climate, independent of host phylogeny, we need to identify which bacterial members are associated with cold tolerance. Core bacteria, Gilliamella, Apibacter, Frischella, Bartonella, and Snodgrassella have contributed mainly to the partition and clustering of the cold-adapted bee group (Fig. [84]2A). Among these, Gilliamella was barely detected in dwarf bees (Fig. [85]S1b), whereas it was most significantly enriched (Fig. [86]2B) in the cold-resistant bees, despite typically occurring in the giant honeybees. In contrast, the gram-positive bacteria Lactobacillus and Bifidobacterium predominated in the gut of dwarf honeybees, while Gammaproteobacteria and Bacteroidota, such as Dysgonomonas and Apibacter, outcompeted others in giant honeybees. Fig. 2. Gilliamella is most significantly enriched in cold-adapted honeybee hindguts, promoting host thermogenesis. [87]Fig. 2 [88]Open in a new tab A Principal component analysis of gut bacteriome showing bacteria most strongly separating different bees. B Significantly enriched gut bacteria identified in the three honeybee subgenera. LDA: linear discriminant analysis. C Apis cerana workers with (Gc) or without (GF) Gilliamella colonization were subject to thermo imaging after cold exposure (10 °C) for 30 min. AV, average temperature for the framed area; HS and CS denote the highest and lowest temperatures for the entire area, respectively. D The same sets of bees were measured for abdominal temperature after 4 h cold exposure, using a thermocouple (the needle-like instrument touching the bee’s abdomen, as shown in the right photograph of the middle insert). Acer, A. cerana; Amel, A. mellifera; Ador, A. dorsata; Alab, A. laboriosa; Aand, A. andreniformis; Aflo, A. florea. Gilliamella colonization improves host heat production and cold tolerance To determine the potential impact of Gilliamella on honeybees’ cold resistance, we compared A. cerana workers monocolonized by Gilliamella to germ-free controls, measuring their cold tolerance at 10 °C. On day 8 post-inoculation, as detected by colony-forming unit (CFU) count, the Gilliamella strain B3835 stably colonized the bee hindguts at densities of 10^7–10^8 cells per gut (Fig. [89]S1d). When exposed to the cold condition for 30 min, Gilliamella-colonized groups remained active, while the germ-free groups became immobilized (Supplementary Movie [90]1). The bees sustained a higher body temperature when colonized with Gilliamella (as measured by thermal imaging, Fig. [91]2C). This difference remained significant after four hours of cold exposure as examined by the abdominal temperatures using a thermocouple (Fig. [92]2D). In contrast, the colonization of a Lactobacillus strain isolated from A. andreniformis did not improve A. mellifera thermogenesis (Fig. [93]S1e), indicating that not all honeybee gut bacteria can enhance host cold-tolerance. These results indicate that Gilliamella increased heat production in the host, thereby enhancing cold tolerance. Gilliamella excels in producing key energy substrates: pyruvate and glucose To evaluate the capacity of Gilliamella in its potential contribution to host thermogenesis, we inspected 149 Gilliamella genomes, as well as 17 Lactobacillus and 28 Bifidobacterium genomes (Fig. [94]3A, Fig. [95]S2, [96]3, Supplementary Data [97]1). Among all core gut bacteria derived from western honeybees, Gilliamella was experimentally shown to be superior in producing pyruvate and cross-feed Snodgrassella^[98]24. Hence, we explored the underlying metabolic pathways producing energy substrates required by thermogenesis, such as pyruvate^[99]25. Our results showed that besides glycolysis, all Gilliamella strains derived from cold-adapted honeybees can produce pyruvate via two dedicated metabolic modules that degrade glucuronate or galacturonate (Fig. [100]3A, B). No Lactobacillus or Bifidobacterium strain encodes the full glucuronate degradation module, and only a few Bifidobacterium strains possess the galacturonate degradation module (Fig. [101]S3), highlighting Gilliamella’s metabolic features in pyruvate production. Fig. 3. Characteristic carbohydrate utilization in Gilliamella. [102]Fig. 3 [103]Open in a new tab A Maximum-likelihood phylogeny of Gilliamella strains, annotated with the number of genes involved in carbohydrate metabolism: enzymes involved in degradation modules of glucuronate (M00061), galacturonate (M00631), and ascorbate (M00550); proteins transporting only glucose, glucose and maltose, or multiple sugars including glucose; primary hydrolases targeting polysaccharides. Circle and triangle sizes indicate gene copy number. B Schematic metabolic pathways inferred from Gilliamella and honeybee genomes. The Gilliamella-generated substrates can be potentially utilized by the host in lipogenesis. G3P glyceraldehyde 3-phosphate, LCFAs long-chain fatty acids, ELOVL6 elongation of very long chain fatty acids protein 6, ChREBP carbohydrate-responsive element binding protein, PRPP 5-phosphoribosyl diphosphate. C Gene clusters involved in polysaccharide deconstruction and further utilization generating pyruvate in the strain Gilliamella B3835. D Differential growth of Gilliamella on varied carbon sources demonstrates its preference for glucuronate, resulting in lower cell density. The Gilliamella B3835 strain was cultured in Brain Heart Infusion (BHI) broth (minus glucose) supplemented with 1, 10 or 20 mM of glucose, glucuronate, or ascorbate, respectively. Data are presented as means ± SD. Galacturonate is a backbone unit of pectin, where glucuronate is occasionally found as a side chain unit^[104]26,[105]27. Gilliamella can hydrolyze refractory polysaccharides, especially pectin from pollen walls, releasing sugars as energy sources^[106]26,[107]28. Our results unraveled prevalent gene clusters in the Gilliamella genome that contain modules responsible for polysaccharide degradation and subsequent monosaccharide transport and catabolism. For instance, in strain B3835 from A. cerana (Fig. [108]3C), genes for galacturonate and glucuronate degradation are located close to genes encoding pectin-degrading enzymes GH28^[109]29 and GH31^[110]30, while the transporters and catabolic enzymes for galactose are adjacent to β-D-galactofuranosidase GH43_3. Intriguingly, a pyruvate dehydrogenase complex repressor (pdhR) is encoded nearby, suggesting that downstream pyruvate catabolism can be suppressed. These enzymes are organized into carbohydrate-active enzyme gene clusters (CGCs)^[111]31, members of which are typically co-expressed^[112]32, enabling efficient stepwise conversion of polysaccharides into energy substrates. In addition to pectin, we found that Gilliamella can also hydrolyze β-glucan, which is another abundant polysaccharide component of the pollen coat (Fig. [113]3A). The backbone of hemicellulose β-glucan is glucose^[114]27, which could serve as a direct energy resource for Gilliamella. However, to utilize glucose as a routine energy supply, Gilliamella would require glucose-specific transporters, which we found surprisingly missing in their genomes. Almost all Gilliamella strains possess only one gene copy (mostly gene crr) to transport multiple sugars, including glucose (Fig. [115]3A, Supplementary Data [116]1). Alternatively, because bacteria secrete hydrolases for saccharification, Gilliamella is expected to release a considerable amount of glucose into the hindgut, which could become accessible to the host for lipogenesis^[117]33,[118]34 and thermogenesis^[119]34,[120]35. Complementary to their limited glucose uptake, Gilliamella strains possess numerous genes for utilizing alternative saccharides, such as galactitol, fructose, mannose, and β-glucoside (Fig. [121]S2). Particularly, most transporter genes target ascorbate. Gilliamella encodes an intact ascorbate degradation module generating D-xylulose-5P, which is lacking in most Lactobacillus strains and all Bifidobacterium strains (Fig. [122]3A, B, Fig. [123]S3). D-xylulose-5P can be converted to energy and 5-phosphoribosyl diphosphate (PRPP), a key nucleotide precursor, via the pentose phosphate pathway. Therefore, Gilliamella is capable of utilizing a diverse range of alternative saccharide substrates in the gut. To assess the growth performance of Gilliamella on different carbohydrate sources, we simultaneously cultured Gilliamella using a Brain Heart Infusion (BHI) broth without glucose, supplemented with glucuronate, ascorbate, or glucose (Fig. [124]3D). Compared to glucose and ascorbate, glucuronate-fed cultures evidently entered the log phase earlier, demonstrating a preference for glucuronate, thus validating our prediction based on genomics. Intriguingly, when supplied with equal molar amounts of carbon, glucuronate and ascorbate yielded cultures that plateaued at lower OD[600] values than glucose, indicating a lower proliferation on these substrates. These decreased cell densities are expected to consume less carbon sources, potentially alleviating competition with both the host and other gut microbes. Additionally, 10 mM glucuronate or ascorbate at 10 mM supported higher growth than 20 mM, indicating that Gilliamella thrives at moderate substrate concentrations. Microbiome analysis corroborates Gilliamella’s strength in energy generation To investigate the impact of Gilliamella enrichment on gut microbiome function, we quantified the relative abundance of carbohydrate metabolic pathways and their bacterial contribution in the gut metagenomes of six Apis honeybees. Our results showed that the abundance of genes encoding transporters capable of glucose uptake was negatively correlated with Gilliamella’s abundance, but positively correlated with that of Lactobacillus and Bifidobacterium (Fig. [125]4A). Our genome analyses (Fig. [126]3A, Supplementary Data [127]1) showed that most Gilliamella strains lack dedicated glucose transporters (e.g., glcU, ptsG) and instead carry only crr, which transports a broad range of sugars. Consistently, our metagenomic data reveal that glcU, ptsG, and the multiple sugar transporter msmX are rarely encoded by Gilliamella (Fig. [128]4A), with the maltose/glucose transporter malX found in strains derived from giant honeybees and a few western honeybees. In contrast, ptsG and msmX are common in Bifidobacterium, and especially abundant in Lactobacillus from dwarf honeybees, reflecting their robust glucose-uptake capacity. Hence, our findings from bacterial strain genomes and gut metagenomics both suggest that the enrichment of Gilliamella corresponds to a reduced overall capacity for glucose uptake within the gut microbiota of cold-adapted honeybees. Fig. 4. Gut microbiome comparison in carbohydrate metabolism across honeybees. [129]Fig. 4 [130]Open in a new tab A The gene abundance of glucose transporters in gut metagenomes contributed by varied bacteria. The y-axis is the gene abundance of transporters capable of glucose uptake. Pearson correlation coefficient (R) and P value (p) are annotated. Gilliamella rarely encodes glucose transporters, especially glucose-specific transporters (glucU and ptsG), across all metagenome samples (upper panel). And its abundance is negatively associated with the abundance of all glucose transporters in metagenomes (lower panel). The abundance comparison of genes encoded by three gut bacteria in the (B) degradation module of glucuronate, galacturonate and ascorbate, and (C) PRPP production from D-xylulose-5P shows Gilliamella’s superior degradation capacity in alternative carbon sources whereas lower consumption of D-xylulose-5P. ns: no significant difference, **P value < 0.01, ****P value < 0.0001 by Mann–Whitney U test (B–D). Each dot denotes one gut metagenome sample. D Glucose proportions across gut tissues in four sympatric honeybees, A. cerana (Acer), A. dorsata (Ador), A. andreniformis (Aand) and A. florea (Aflo), examined by saccharide metabolomics show significantly more glucose in Acer rectums. Two-way ANOVA and Tukey test were used for the comparison. Data are presented as means ± SD. With regards to carbohydrate metabolism, Gilliamella allocates more genes for pentose and glucuronate interconversions than Lactobacillus and Bifidobacterium (Fig. [131]S4a), specifically for the degradation modules of glucuronate, galacturonate, ascorbate and pectin (Fig. [132]4B, Fig. [133]S4b). This leads to enhanced production of pyruvate and D-xylulose-5P. Meanwhile, Gilliamella encodes fewer genes that convert D-xylulose-5P to D-ribose-5P and the downstream PRPP (Fig. [134]4C, Fig. [135]S4c). To test whether enriched Gilliamella in gut microbiome affects the availability of hindgut glucose for honeybee hosts, we applied saccharide metabolomics across the gastrointestinal tract of various honeybee species (Fig. [136]4D, Supplementary Data [137]2). In particular, four sympatric honeybees were included to alleviate the potential bias of sugar ingestion derived from diverged floral diversity. Among these, workers of three tropical honeybees, A. dorsata, A. florea and A. andreniformis, and the cold-adapted A. cerana, were collected in their natural range of close geographic proximity in Jinuo Co., Yunnan Province, China (Fig. [138]S1a). The diets of the four honeybee species showed no significant variations in glucose proportion, as demonstrated in beebread and honey samples retrieved from their colonies (Fig. [139]S5a, Supplementary Data [140]2). The glucose content also showed a similar level in both the midguts and ileums among honeybees, but was significantly elevated in the rectum of A. cerana (Fig. [141]4D). But, no evident difference was found between cold-adapted A. cerana and three tropical honeybees for other mono- or disaccharides (Fig. [142]S5b). Therefore, these results suggest that the enrichment of Gilliamella is associated with enhanced glucose availability in the hindgut. Gilliamella colonization assays confirm promotion of hindgut glucose and pyruvate access To further verify the influence of Gilliamella on hindgut glucose content, monocolonization was performed and compared to germ-free bees. Glucose content in the hindguts of Gilliamella-colonized bees is significantly higher at both 48 h and 7 days after inoculation (Fig. [143]5A). Host glucose absorption typically occurs in the midgut via a concentration-dependent process, where glycogen synthesized in gut epithelium cells creates concentration gradients, thus facilitating glucose uptake^[144]36,[145]37. However, little is understood about glucose absorption in the hindgut. We examined the ileum and rectum and found a considerable amount of glycogen in epithelial cells stained purple, as in the midgut (Fig. [146]5B), indicating potential for glucose absorption. Gilliamella inoculation up-regulated genes for glucose and pyruvate transport and catabolism (Fig. [147]5C), as well as pathways in carbohydrate metabolism (e.g., the tricarboxylic acid cycle) and lipid synthesis (Fig. [148]S6a, Supplementary Data [149]1). Hence, based on evidence from both cell morphology and gene expression, we infer that honeybee hosts can utilize glucose and pyruvate generated by Gilliamella in the hindgut. We further found that the genes involved in the synthesis and secretion of glucuronate are also up-regulated (Fig. [150]5C, Supplementary Data [151]1) in the Gilliamella-colonized group. These data suggest the host may secrete glucuronate to fuel Gilliamella, since the bacterium prefers the glucuronate degradation pathway to produce pyruvate. In turn, this cross-feeding would benefit the host by providing additional pyruvate. Fig. 5. Honeybee-Gilliamella coadaptation to coldness via enhanced lipid synthesis. [152]Fig. 5 [153]Open in a new tab A Glucose contents in A. mellifera hindguts colonized by Gilliamella (Gc) versus germ-free ones (GF), showing significant glucose increase following bacterial inoculation. B The glycogen synthesized in gut cells was stained purple using Periodic-Acid Schiff (PAS) staining, suggesting that the hindgut can absorb glucose like the midgut. C The up-regulated genes in Gc versus GF indicate potential cross-feeding between the host and Gilliamella. D Enzymes involved in glucose, pyruvate, lipid, and glucuronate productions contain more gene copies in cold-adapted honeybees. E Maximum-likelihood tree based on the ELOVL6 gene sequences with motifs annotated, suggesting historical gene duplication in cold-adapted honeybees and bumblebees. F Gilliamella inoculation resulted in significantly elevated abdominal triacylglycerol (TAG) on day 10 post-inoculation. G Histological comparison shows Gilliamella colonization affects host lipid storage. HE (hematoxylin and eosin), PAS, and ORO (Oil Red O) were used for cell, glycogen, and lipid staining, respectively. Black arrows indicate multilocular lipid droplets in Gc bees; red arrows indicate unilocular droplets in GF bees. H Hindgut ascorbate is significantly increased in A. mellifera and A. cerana 48 h after Gilliamella inoculation. Honeybee genomes and colonization assay reveal host-symbiont coadaptation via lipogenesis To explore potential mutualistic cross-feeding between host and symbiont and their synergy toward thermogenesis, we inspected the featured genomic contents of the cold-adapted honeybees, with a focus on functions involved in fuel synthesis for thermogenesis, including glucuronate, pyruvate, glucose, and lipid. The comparative genomic analyses (Fig. [154]5D, Supplementary Data [155]1) included genomes of six honeybees and two bumblebees that are also adapted to temperate climates. We found that A. mellifera and A. cerana generally possess more genes for synthesizing glucose, pyruvate, and lipids than their tropical relatives, albeit they differ in details. For example, A. mellifera possesses more genes for transporting trehalose, which can be hydrolyzed into two glucose molecules, while A. cerana has more genes for degrading maltose and sucrose into glucose. Similarly, A. mellifera possesses more genes for converting UDP-galactose to UDP-glucose, while A. cerana has more genes for converting serine and lactate to pyruvate. Notably, cold-adapted bees generally encode more copies of the long-chain fatty acid elongation protein ELOVL6 (K10203), a key protein involved in lipid synthesis, as seen with potential historical gene duplication in both A. mellifera and A. cerana, and the bumblebee B. terrestris (Fig. [156]5D, E). Different from those in the bumblebee, ELOVL6 homologs remain relatively conservative across all six honeybee species, with one identical homolog shared by all, and the other two vary at the sequence level but identical motifs (with one exception of a shortened protein in A. mellifera) (Fig. [157]5E). One of the derived ELOVL6 homologs was only shared between A. mellifera and A. cerana, showing small but consistent sequence differences between species. In contrast, no relevant protein showed higher gene copy numbers in the genomes of tropical honeybees when compared with cold-adapted honeybees. In mammals, glucose and pyruvate are required to synthesize lipids^[158]33, and these three energy substances mediate thermogenesis^[159]35,[160]38,[161]39. We hypothesize that honeybees and Gilliamella co-work to facilitate cold-tolerance through improving host lipogenesis. Triacylglycerol (TAG) stored in the fat body was measured to test the impact of Gilliamella colonization on host lipogenesis. Gilliamella-colonized bees demonstrated significantly higher TAG accumulation than the germ-free hosts (Fig. [162]5F). Gilliamella colonization also resulted in increased fat body cells (Fig. [163]S6b), forming a cell layer surrounding the inner side of the abdomen’s exoskeleton. On the contrary, only a few fat body cells were sporadically observed throughout the entire section of the germ-free abdomen. The presence of lipid droplets was confirmed using oil red O staining (Fig. [164]5G). In mammals, the white adipose tissue (WAT) stores fat as unilocular lipid droplets, while the brown adipose tissue (BAT) harbors multilocular lipid droplets characterized by highly condensed mitochondria for efficient thermogenesis^[165]25. Intriguingly, we observed multilocular lipid droplets in the Gilliamella-colonized bees resembling mammalian brown adipose tissue BAT, and unilocular lipid droplets similar to mammalian WAT in the germ-free bees. In addition, the fat cell nuclei of the Gilliamella-colonized bees were distorted and pressed by fat globules (Fig. [166]5G), indicating extensive fat accumulation^[167]40. Furthermore, the Gilliamella-colonized bees also contain more glycogen than the germ-free bees (Fig. [168]5G). Honeybees provide ascorbate when Gilliamella is inoculated Like the two bumblebees, A. mellifera, A. cerana and A. dorsata have more glucuronosyltransferase genes ([169]K00699, Fig. [170]5D) than other honeybees. Glucuronosyltransferase, shared by the glucuronate pathway and the ascorbate biosynthesis module, catalyzes glucuronate generation (Fig. [171]5D). Intriguingly, the accumulation of glucuronate is toxic to the honeybees^[172]41. Thus, its downstream transformation into ascorbate or D-xylulose-5P is necessary. Because it is difficult to distinguish glucuronate from galacturonate using regular gas chromatography-mass spectrometry gut metabolomics^[173]42, the significantly increased galacturonate reported previously in Gilliamella colonized guts^[174]43 might have included the glucuronate produced by the honeybee host. Thus, we alternatively detected its downstream metabolite ascorbate. We found that the hindgut ascorbate content is significantly higher in the Gilliamella-colonized bees than in the germ-free bees (Fig. [175]5H). Discussion In this study, we showed that gut symbiont Gilliamella assists host lipogenesis and thermogenesis and engages in a synergistic interaction with honeybees to orchestrate carbohydrate metabolism (Fig. [176]6). Concomitant with elevated lipogenesis and increased levels of glucose, pyruvate, and glucuronate in cold-adapted honeybees, Gilliamella efficiently converts pollen polysaccharides and host-derived glucuronate into glucose and pyruvate. Moreover, Gilliamella’s preference for glucuronate (over glucose) enables the bacterium to liberate extra glucose and pyruvate for its host. Additionally, honeybees and Gilliamella might exchange ascorbate and jointly elevate D-xylulose-5P levels to drive thermogenic fat synthesis^[177]44. Fig. 6. Schematic summary of the honeybee-Gilliamella crosstalk in carbohydrate metabolism, which collectively improves host cold adaptation. [178]Fig. 6 [179]Open in a new tab Codes 1–10 denote enzymes shown in Fig. [180]5D. Dashed lines present putative metabolic flows. Vc ascorbate, Xu5P D-xylulose-5P. Although gut microbiota has been suggested to be associated with cold tolerance in bumblebees and soldier flies^[181]15,[182]45, the underlying mechanism remains unclear. Our study uncovers that Gilliamella enrichment in cold-adapted honeybees is associated with increased glucose and pyruvate in hindguts. Strikingly, Gilliamella restrains its own growth by favoring non-glucose substrates, thereby lowering per-cell and total sugar consumption and reducing competition with the host. Specifically, Gilliamella genomes encode pectinase gene clusters and adjacent genes for glucuronate and galacturonate catabolism, highlighting their specialization for utilizing pollen-derived polysaccharides and their breakdown products. This gene organization suggests coordinated expression^[183]32, enabling efficient carbohydrate metabolism, and reinforcing niche specificity, which benefits both the symbiont and the host. Gilliamella also possesses a robust genetic capacity in degrading ascorbate into D-xylulose-5P, but lacks substantial genes converting D-xylulose-5P to PRPP, a nucleotide precursor^[184]46. This explains the limited growth of Gilliamella when feeding on ascorbate and its precursor, glucuronate. Together, these traits elevate D-xylulose-5P, activating the carbohydrate-responsive element binding protein (ChREBP) and further enhancing lipogenesis^[185]44. In summary, by providing fuels for host lipogenesis, Gilliamella promotes honeybee cold-tolerance via improving host carbohydrate metabolism. The synergistic interaction between Gilliamella and cold-adapted honeybees may have been favored by selection during host range expansion to temperate regions, resulting in elevated Gilliamella abundance in the host gut. Our study highlights the importance of glucose, pyruvate and lipids in honeybee cold resistance, echoing findings in mammals. Winter bees exhibit increased fat body cell numbers, lipid stores, and mitochondrial abundance^[186]47, which reflects a key feature of mammalian BAT^[187]47. Our findings in honeybee lipid morphology and the report on beige-like fat in cold-adapted chickens^[188]48 suggest that BAT thermogenesis may not be confined to mammals. Indeed, fat burning has been recently associated with thermogenesis in beetles, nematodes, and flies^[189]49, highlighting the importance of thermogenic fat in cold resistance for invertebrates. We also detected higher copy numbers of the ELOVL6 gene in both A. mellifera and A. cerana, which could boost the synthesis of unsaturated long-chain fatty acids. Unsaturated lipids confer greater cold tolerance than saturated lipids in mice^[190]50. These lipids are also commonly stockpiled by overwintering insects^[191]51,[192]52. ELOVL6 incorporates oleic acid into phospholipids and alleviates membrane rigidity under cold exposure^[193]53. Most importantly, ELOVL6 regulates the thermogenic capacity of BAT in mammals^[194]54. In addition to evidence in lipid morphology, we observed increased levels of glucose^[195]34,[196]35,[197]55 and pyruvate^[198]38, both of which are key prerequisite substrates for lipogenesis and thermogenesis, further supporting the presence of thermogenic fat in honeybees. Finally, these enhancements for lipogenesis and thermogenesis are found in both A. mellifera and A. cerana, suggesting convergent cold-adaptation strategies between these honeybees. The most striking finding of the present study is the evidence that suggests an elaborate host-gut bacterium synergy in carbohydrate metabolism. While honeybees provide pollen polysaccharides, glucuronate, and ascorbate to Gilliamella, whose degradation generates glucose, pyruvate, and D-xylulose-5P, and in turn benefits the host. In principle, Gilliamella could have relied on glucose derived from β-glucan as its energy source, but instead, it preferentially utilizes galacturonate from pectin and glucuronate from hosts. The glucuronate and galacturonate degradation modules share three enzymes; both are capable of producing pyruvate in only five steps, potentially enabling efficient pyruvate generation in Gilliamella. However, core gut bacteria Lactobacillus and Bifidobacterium lack this capability. This is likely why Gilliamella can produce the most pyruvate and cross-feed other symbionts^[199]24. While Gilliamella sacrifices glucose utilization to limit self-proliferation, hosts paradoxically sustain glucuronate production, despite its inherent toxicity to honeybees^[200]41. In the presence of Gilliamella, honeybees produce ascorbate from glucuronate and secrete it into the hindgut, where Gilliamella consumes glucuronate and ascorbate, thus preventing glucuronate accumulation. Although Gilliamella cannot synthesize ascorbate, it encodes most transporters for ascorbate, suggesting adaptation to host-derived metabolites. And its reduced growth yields on glucuronate or ascorbate imply that hosts may regulate symbiont density by providing particular metabolites^[201]56. Taken together, honeybees and Gilliamella coordinate nutrient exchange to achieve mutual benefits. These reciprocal metabolic exchanges balance cooperation and competition, reflecting a win-win strategy in host-symbiont interactions. By comparing hosts and their gut symbionts across closely related Apis species, we show how host-symbiont metabolic interaction has been potentially shaped by natural selection. A. mellifera and A. cerana have converged on a similar host-symbiont synergy when facing shared environmental challenges during independent expansion to temperate regions. Our study focuses on one core symbiont, but we acknowledge the broader ecological context, where Gilliamella’s function may intersect with other bacterial interactions and host pathways, as a critical avenue for future research. Our framework can be extended to study the long-term symbiosis by including various symbionts and bee hosts^[202]17,[203]18,[204]57. Bumblebee genomes encode more copies of ELOVL6 than honeybees, indicating a substantial lipogenesis enhancement on the host side. In contrast, bumblebee-derived Gilliamella^[205]22 and Lactobacillus^[206]58 lack extensive genes for saccharide utilization, highlighting a different host-microbe interaction that warrants investigation. We note that natural selection acted on honeybees and their gut bacteria within a pre-existing symbiosis, in which core bacterial taxa were already associated with bees before the Apis lineage emerged. Core bee gut bacteria (e.g., Gilliamella) were acquired early in the social bee evolution^[207]17 but developed subsequent adaptation apparently in a host-specific manner^[208]58. This evolutionary history indicates that the retention of Gilliamella cannot be attributed solely to cold adaptation. While our data reveal a highly synergistic carbohydrate metabolism between cold-adapted honeybees and their symbiotic Gilliamella, whether this reflects true co-evolution^[209]59 remains uncertain. Addressing this question would require testing whether Gilliamella has undergone host-specific evolutionary changes in response to cold selection pressure. Notably, many Gilliamella strains from bumblebees lack the degradation module for glucuronate and ascorbate, while most published genomes of honeybee-derived Gilliamella encode intact modules (Fig. [210]3A), indicating functional divergence associated with the social bee lineage. However, due to the limited isolates from the tropical honeybees, comparative analyses between tropical and cold-adapted strains remain insufficient. Future isolation and genome sequencing of Gilliamella strains for diverse Apis hosts will be essential to test for functional shifts associated with the host adaptation and to explore the potential for host-symbiont coevolution. Gilliamella is also prevalent in giant honeybees, yet its role in host adaptation remains to be investigated. We also note that the core bacterium Lactobacillus, which, although not enriched in cold-resistant honeybees, is one of the predominant bacteria, especially abundant in A. mellifera^[211]60,[212]61. Whether and to what extent different gut bacteria and strains assist bee hosts in lipogenesis and thermogenesis deserves further study. Another core symbiont, Bartonella, is also enriched in cold-adapted A. mellifera, becoming remarkably abundant in the overwintering phase^[213]62. Seasonal gut microbiota turnover likely reflects the drastic reduction in pollen consumption by winter honeybees^[214]62, where Bartonella potentially provides amino acids to the host^[215]63. Thus, distinct gut symbionts support honeybee hosts through diverse, complementary metabolic mechanisms, collectively enhancing holobiont resilience to environmental challenges. Our study highlighted the significance of host-symbiont interaction in the cold acclimation of the holobiont, providing insights into the long-term interaction development between honeybees and gut bacteria^[216]64. Methods Biogeographical data for honeybees’ native distribution were downloaded from the Global Biodiversity Information Facility ([217]https://www.gbif.org/). Sample collection For metagenome sequencing, we used 168 hindguts from six Apis honeybee species, including dwarf honeybees (A. andreniformis n = 39, A. florea n = 46) and giant honeybees (A. dorsata n = 24, A. laboriosa n = 36) from Yunnan Province, China, as well as eastern (A. cerana n = 14) and western (A. mellifera n = 9) honeybees from Yunnan and Jilin Province, China (Supplementary Data [218]1, Fig. [219]S1a, b). For saccharide metabolomics, different bee gut segments, honey and bee bread were collected (Supplementary Data [220]2) from four sympatric honeybee species, including cold-resistant eastern honeybees (A. cerana) and tropical species (A. andreniformis, A. florea, A. dorsata) at two sites (Mengla, Jinuo) in Yunnan (Fig. [221]S1a). To quantify saccharides in honeybee guts, midguts, ileums and rectums were dissected and stored at -80°C for examination. To assess dietary differences, 2–50 g bee bread and 10 mL honey were collected from each colony and stored at –20 °C. Gut metagenome analysis Total DNA from each hindgut sample was obtained using CTAB buffer and phenol/chloroform/isoamyl alcohol (25:24:1, pH 8.0) as previously described^[222]65. DNA from individual and pooled samples (Supplementary Data [223]1) was used for sequencing. A total of 70 samples were prepared for library construction (350 bp insert size) and total DNA shotgun sequencing (150 bp paired-end) using an Illumina NovaSeq 6000 platform, generating approximately 10 Gb of data per sample. Low-quality reads were filtered using fastp^[224]66 (version 0.13.1, -q 20 -u 10). Metagenomes were de novo assembled using MEGAHIT^[225]67 (version 1.1.2, -m 0.6 –k-list 31,51,71, –no-mercy). Contigs >500 bp were taxonomically assigned using DIAMOND^[226]68 (version 0.9.22, blastx -f 102 -k 1 -e 1e-3) against the NCBI nr database. Genes were predicted using MetaGeneMark^[227]69. Assemblies were assigned taxonomically when >70% of genes matched a bacterial taxon. Unclassified assemblies were further compared to NCBI bacterial genomes (blastn, -evalue 1e-5) as previously described^[228]65. Reads were mapped onto the assemblies and genes using SOAPaligner^[229]70 (version 2.21, -M 4 -l 30 -r 1 -v 6 -m 200), and the sequencing coverage and depth were summarized using the soap.coverage script (version 2.7.7, [230]https://github.com/aquaskyline/SOAPcoverage) to calculate the bacterial abundances via multiplication. Assemblies with >90% read coverage were retained. Gut community clustering was performed using R package vegan and ropls. Functional annotation of genes was conducted with KofamKOALA^[231]71 with the e-value < 0.001. KEGG (Kyoto Encyclopedia of Genes and Genomes) ortholog abundances were calculated as the sum of annotated gene abundances. LEfSe (29) was used to identify symbionts enriched in different bee groups. Pearson correlations between the abundance of bacteria and functions were evaluated and visualized using the R package ggpubr. Gene copy numbers of KEGG orthologs for specific metabolic pathways in Gilliamella, Lactobacillus and Bifidobacterium were compared using Mann–Whitney U test and visualized with ggplot2 and ggpubr. Saccharide metabolomics Beebread samples were dried at 37 °C and then homogenized with liquid nitrogen. Gut samples added with 100 μL of sterile pure water were homogenized using a handheld tissue grinder (TGrinder, TIANGEN, Beijing) with disposable plastic grinding pestles at 6000 rpm for about 5 s. The analysis of carbohydrates was performed using a Thermo Scientific Dionex ICS-5000+ ion chromatography system equipped with a CarboPac PA10 analytical column (4 × 250 mm) and a Dionex CarboPac PA10 guard column (4 × 50 mm). The eluents comprised phase A (ultrapure water) and phase D (200 mM NaOH). The gradient program was set as follows: 0–19 min (91% A + 9% D), 19–21 min linear gradient to 0% A + 100% D, 21–31 min isocratic 100% D, 31–32 min return to 91% A + 9% D, and 32–40 min equilibration under initial conditions (91% A + 9% D), with a constant flow rate of 1.0 mL/min. Prior to analysis, the column was equilibrated with a sodium acetate-containing solution (phase C) for 60 min to ensure system stability. The column temperature was maintained at 30 °C, and the injection volume was 10 μL. Detection was achieved using a pulsed amperometric detector (PAD) with a gold working electrode and an Ag/AgCl reference electrode. The applied “Carbo Quad” waveform included four potentials: +0.10 V (0.4 s, with integration period enabled in the first 0.2 s to enhance sensitivity), –2.00 V (0.02 s), +0.60 V (0.01 s), and –0.10 V (0.07 s). Data were acquired at 2 Hz with auto-zero enabled, and signals were recorded from both ED_1 (raw current) and ED_1_Total (integrated charge). Saccharide concentrations were converted to molarity for further relative abundance calculation of molecules and statistical analysis using the aov function in R. Genome analysis of gut bacteria The genomes of Gilliamella, Lactobacillus and Bifidobacterium derived from honeybees and bumblebees were retrieved from NCBI (Supplementary Data [232]1). Genes were predicted using MetaGeneMark^[233]69. Orthologs were identified using OrthoFinder^[234]72 for phylogenetic analysis. Functional annotation was performed with KEGG database (28) and dbCAN2 database^[235]73 for polysaccharide hydrolysis. Visualization was done with tvBOT^[236]74. Gilliamella growth assay Aliquots (OD[600] = 0.1) of Gilliamella strain B3835 were inoculated simultaneously in BHI broth without glucose, supplemented with glucose, glucuronate, or ascorbate at 1 mM, 10 mM, or 20 mM concentrations. They were cultured anaerobically in 24-well plates at 35 °C. OD[600] was measured every 4 h initially and every 12 h at the plateau phase. Bacterium transplantation and metabolite assays Eastern and western honeybee colonies for transplantation experiments were obtained from local beekeepers. Melanistic pupae were transferred from colony cells to sterile plastic cup cages and reared at 35 °C. Gilliamella B3835 was cultured in BHI medium at 35 °C within an incubator with 5% CO[2]. A Lactobacillus strain [237]B24606 was isolated from A. andreniformis using De Man-Rogosa-Sharpe (MRS) agar under the same incubation conditions and subsequently cultured for transplantation assays. For each monocolonization, bacteria were resuspended in (Phosphate Buffer Saline) PBS (OD[600] > 1) and blended with sterile pollen and sugar syrup (fructose: glucose, 6:4). A portion of 200 μL of inoculation syrup was pipetted onto bees’ bodies that could be ingested by companions immediately. The rest was provided to the bees ad libitum in cages for 24 h. PBS without bacteria was mixed with pollen and sugar syrup and used for the germ-free group. The remaining inoculation syrup was plated onto BHI, MRS agar for culture to examine alive inoculated bacteria and potential contamination from environmental bacteria. Sterile sugar syrup and bee pollen were used to feed honeybees after inoculation. To explore the metabolite exchange induced by Gilliamella inoculation, glucose and ascorbate were examined using kits (S0201S, Beyotime; AKVI005M-SY10S, boxbio) after 48 h. On day 8, hindguts were homogenized for colony-forming unit plating. Colony PCR of 16S rDNA (27 F/1492 R primers) and Sanger sequencing were performed to confirm the successful colonization of B3835. To assess the impact of Gilliamella on saccharide metabolism, hindguts were also dissected for saccharide metabolomics. To evaluate the impact of Gilliamella colonization on host lipogenesis, TAG stored in the abdomen (gastrointestinal tract being removed) was quantified using a commercial kit (A110-1-1, Nanjing Jiancheng Bioengineering Institute) on day 10. For histological analysis of lipids stored in the fat body and glycogen in gut cells, abdomens were fixed in 4% paraformaldehyde, dehydrated with ethanol, infiltrated with xylene, embedded in paraffin, and sectioned at 4 μm^[238]75. Sections were stained with hematoxylin and eosin (H&E) or Periodic-Acid Schiff (PAS). For the frozen section, 30% sucrose was used for dehydration and optimal cutting temperature compound for embedding. Frozen sections (10 μm) were stained with Oil Red O (ORO) stain. Cold resistance assay On day 10 post-inoculation, GF and monocolonized bees (A. cerana by Gilliamella B3835, A. mellifera by Lactobacillus [239]B24606) were exposed at 10 °C for 30 min, and their movements and temperatures were recorded using a thermal imager (testo 890) to assess their hardiness. After four hours of cold exposure, all bees could not move when abdominal temperatures were measured by a K-type thermocouple (KAIPUSEN) to assess thermogenesis. Transcriptome analysis To compare host gene expression with or without Gilliamella colonization, the Datasets S2 and S5 from our previous study^[240]76 were used to investigate whether the host provides metabolites to Gilliamella and benefits in turn from bacterial-generated glucose and pyruvate in the hindgut. Differentially expressed genes were subjected to functional investigation and pathway enrichment analysis using Fisher’s Exact Test in R package stats. Honeybee genome comparison Genomes of five honeybees and two bumblebees were retrieved from NCBI (GCF_003254395.2, GCA_001442555.1, GCA_000184785.2, GCA_000469605.1, GCA_014066325.1, GCF_910591885.1, GCF_024542735.1), and the A. andreniformis genome was newly generated (Xiao et al. under revision, BioProject PRJNA1178865). Protein sequences were functionally annotated using KofamScan^[241]71, and gene copy numbers for carbohydrate metabolism and lipid synthesis pathways were compared across species (Supplementary Data [242]1). Protein sequences of ELOVL6 were aligned by MUSCLE^[243]77, and further used for tree construction by FastTree^[244]78. Motif searching was performed by MEME ([245]https://meme-suite.org/meme/tools/meme). Ethics declarations No specific permits were required for this work or sampling. No endangered or protected species were involved in this study. Supplementary information [246]Supplementary information^ (9.6MB, pdf) [247]Supplementary data 1^ (3.7MB, xlsx) [248]Supplementary data 2^ (246.2KB, xlsx) [249]Supplementary movie 1^ (1.5MB, mp4) Acknowledgements