Huddling remodels gut microbiota to reduce energy requirements in a small mammal species during cold exposure

1Social animals and humans have evolved cooperative behaviors to maximize their fitness in harsh environments. Among these behaviors, it has been hypothesized that huddling is one of the most acceptable benefits for species living in temperate and cold habitats to reduce thermoregulatory costs, especially in the cold season. Huddling mediates social thermoregulation by reducing the surface to volume ratio and, consequently, reduces heat loss and energy expenditure of the animal. Studies in some wild species such as the social degus (Octodon degus) and short-tailed field voles (Microtus agrestis) demonstrated that huddling reduced resting metabolic rate (RMR) and energy intake and that these metabolic advantages of huddling increased with lowered ambient temperature and increased group size.

2The coevolution of mammals with their microbial communities confers fitness benefits to both partners. Small herbivores depend mainly on caecal microbiota to digest cellulose and detoxify plant toxins. Microbial diversity has been reported to be influenced by host physiology and such factors as season, altitude, diet, photoperiod, and air temperature. Conversely, gut microbiota, by way of their metabolites and bacterial polysaccharides, can influence host energy metabolism, behavior, inflammation and immunity, homeostasis of enteric and central nervous systems, and aging. These functions of gut microbiota have been confirmed mainly by microbiota transplantation. Social rodent species increase the frequency of communal huddling in winter and, thus, are more likely to be coprophagous. Therefore, we hypothesize that huddling may be a strong force in shaping gut microbiota and, consequently, mediate host energetics and thermoregulation.

3Brandt’s voles (Lasiopodomys brandtii), a small non-hibernating herbivorous rodent species, are widely distributed in Inner Mongolian grasslands of China, Mongolia, and Southeast Baikal region of Russia. Its habitat is characterized by extremely cold dry winter and deep frozen soil. Brandt’s voles show seasonal variations in energy metabolism and thermoregulation to adapt to the seasonal environments. They have evolved group-living strategies such as making sounds to warn each other of predators, hoarding food collectively, and huddling to survive the cold winter. Group size in one colony increases from about 10 in summer to 20 individuals in winter. However, the benefits of huddling in energy metabolism in different seasons and the underlying physiological mechanisms remain largely unknown. Here, we investigated the changes in energy intake, thermogenesis, body temperature, and gut microbiota induced by huddling in Brandt’s voles in warm and cold environments, and we transplanted microbiota to examine the hypothesis that microbiota in huddling voles can reduce energy expenditure. We demonstrated that huddling alters gut microbiota to reduce energy intake and thermoregulatory responses in a small mammal species during cold exposure.

4Adult Brandt’s voles used in this study were from a laboratory breeding colony at the Institute of Zoology, Chinese Academy of Sciences (CAS). After weaning (at 3 weeks of age), the voles were housed with the same sex siblings at a light regime of 16-h light:8-h dark (lights on from 4:00 to 20:00) and room temperature of 23 ± 1 °C. The voles were fed a standard rabbit pellet chow (containing 18% protein, 3% fat, 12% fiber, and 47% carbohydrate, Beijing KeAo Bioscience Co.) and were provided with water ad libitum. The animal procedures were approved by the Animal Care and Use Committee of Institute of Zoology, CAS.

5A total of 124 female voles (aged 4 months) were placed in 31 cages (4 voles/cage) and were divided into 4 treatments: cold (at 4 ± 1 °C) huddling (CH, 7 cages), cold separated (CS, 7 cages), warm (at 23 ± 1 °C) huddling (WH, 8 cages), and warm separated (WS, 9 cages). Sibling voles were chosen preferentially and were acclimated to the cage (42 cm × 27 cm × 20 cm) for 3 weeks and then to the treatment conditions for another 3 weeks. The cages contained four equal compartments separated by stainless steel walls with small holes (6 mm in diameter). Openings (7 cm × 7 cm) in the walls allowed movement among the compartments. The openings were open for the huddling groups but closed for the separated groups, which had olfactory, visual, and vocal contacts. Individual voles were dyed on different positions of the body for identification.

6Voles were weighed at the beginning (day 0) and the end of (day 28) of the study using an electronic balance (Sartorius Model BL 1500, ± 0.1 g), and body weight change was calculated. Food was provided ad libitum and food intake and fecal output were measured for three consecutive days in the last week of acclimation. Uneaten food and feces were collected after the 3 days, oven dried at 60 °C until constant mass, and separated manually. Water content of the food offered was calculated by drying samples at 60 °C until constant mass, and its gross energy was measured by bomb calorimetry (Parr1281 Instrument, USA). Gross energy intake was calculated by subtracting the gross energy of the uneaten food from the gross energy of the food offered.

7RMR, measured as oxygen consumption, was determined in huddling and separated voles at their acclimated temperatures of 4 and 23 °C at the end of the experiment. An open-circuit respirometry system (TSE labmaster, Germany) with an air flow rate of 3 L/minute was used. Voles were maintained in a 5.8 L transparent metabolic chamber (type II for rats) with double-layered meshes for the separation of voles during the measurements. The animals were allowed 1 h for acclimation to the conditions, and then recordings of O2 uptake were taken for 2 h. An average of a minimum of three consecutive stable readings was taken as the RMR.

8NST was determined on individual animals at 25 °C on the day after RMR measurements. The volume of metabolic chamber was 2.7 L (type I for mice). The voles were injected subcutaneously with norepinephrine (NE) (Shanghai Harvest Pharmaceutical Co. LTD) at a dosage of 2.53Mb-0.4 (mg/kg; Mb, body mass). NST was estimated as the four highest consecutive stable readings of oxygen consumption during 1 h.

9In the last week of acclimation, surface body temperature (Tsurface) was read with an infrared camera (FLIR E60, UK) from a distance of 40 cm, and the data were analyzed by FLIR Tools software. Tsurface was averaged from seven to eight images for each cage. The perimeter temperature (Tp) of the voles was determined by fitting a polygon around the individual animal in the case of separated animals and around the entire group for huddling animals using the option “isotherm” of the software. The highest temperature in an image was selected as the maximum temperature (Tm). Furthermore, the contact temperature (Tc) between two animals in the huddling groups was determined by the “mobile” option.

10Core body temperature (Tcore) was recorded during the acclimation period through transponders (G2 E-Mitter, to ± 0.1 °C, STARR life sciences) implanted in the abdomen of one vole in each cage (n = 4–5 voles/group). The voles were anesthetized by intraperitoneal injection of pentobarbital sodium (50 mg/kg). After the abdominal skin was sterilized with iodophor, an incision of up to 1 cm in length was made below the diaphragm. The wound was closed with absorbable PGA surgical suture (Jinhuan Model R413, 4/0) and sterilized with iodophor again. The animals were allowed 10 days to recover from surgery. All receivers for collecting data were connected to a computer with the Vital View software.

11At the end of experiment, one vole from each cage (but not the one with E-Mitter) was sacrificed by CO2 asphyxiation. The digestive tract and caecal content were removed on a super-clean worktable, frozen immediately in liquid nitrogen, and stored at − 80 °C. Six SCFAs including acetic, propionic, butyric, isobutyric, valeric, and isovaleric acids were measured in caecal contents by high-performance gas chromatography (GC, Agilent 7890A; Agilent Technologies, Germany) with a GC autosampler and a FID system by the modified method. Caecal contents were extracted directly with water and did not require derivatization. Separations were performed in a 30 m × 0.25 mm × 0.25 μm DB-WAX column (Agilent Technologies) using 99.998% hydrogen as carrier gas at a flow rate of 1.0 mL/min. The system was operated at 250 °C. Injections were performed in the splitless mode at 230 °C, and 0.5 μL for each injection. The oven temperature was programmed from 60 °C (1 min) to 200 °C at 5 °C/min and then from 200 to 230 °C at 10 °C/min. The total running time of each sample was 32 min.

12DNA from caecal contents was extracted by 2 × CTAB (cetyltrimethyl ammonium bromide), phenol chloroform mixture (phenol:chloroform:isoamyl alcohol = 25:24:1). In the later steps of DNA isolation, we used the spin column from SanPrep Column DNA Gel Extraction Kit (Sangon Biotech, China) (based on a silica gel membrane) to purify and recover the DNA rapidly. DNA concentration was measured by fluorometry using the Qubit® dsDNA high-sensitivity assay kit and the Qubit® 2.0 fluorometer (Life Technologies, Carlsbad, CA, USA) as instructed by the manufacturer. DNA purity was also assessed by absorbance on a Nanodrop 2000 (Thermo Fisher Scientific, Carlsbad, CA, USA) by measuring the A260/A280 ratio. Only DNAs with an A260/A280 ratio of 1.8–2.0 were used for PCR amplification (n = 6 voles/group). Our target was the V3–V4 hyper-variable region of the bacterial 16S rRNA gene. PCR was started immediately after the DNA was extracted. The 16S rRNA V3–V4 amplicon was amplified using 2 × Taq PCR MasterMix (Tiangen, Beijing, China). Two universal bacterial 16S rRNA gene amplicon PCR primers (PAGE purified) were used: forward primer-341F (CCTACGGGNGGCWGCAG) and reverse primer-805R (GACTACHVGGGTATCTAATCC). To multiplex the samples during sequencing, barcodes were added to the 5′ termini of the forward primers (Additional file 1 Table S1). The PCR reaction was set up as follows: template DNA 2 μL, amplicon PCR forward primer (10 μM) 1 μL, amplicon PCR reverse primer (10 μM) 1 μL, and 2 × Taq PCR MasterMix 12.5 μL (total 25 μL). PCR was performed for each DNA sample in triplicate in the same thermal cycler (T100™ BIO-RAD) using the following program: 1 cycle of denaturing at 94 °C for 5 min, followed by 34 cycles of denaturing at 96 °C for 30 s, annealing at 52 °C for 30 s, elongation at 72 °C for 30 s, and a final extension at 72 °C for 5 min. The PCR products were checked using electrophoresis in 1% (w/v) agarose gels in TBE buffer (Tris, boric acid, EDTA) stained with ethidium bromide (EB) and visualized under UV light. PCR products were pooled and purified using Agencourt AMPure XP magnetic beads (Beckman) according to the manufacturer’s instructions. Then Sequencing was done on an Illumina HiSeq 2500.

13The 16S sequence paired-end data set was joined and quality filtered using the FLASH method described by Magoč and Salzberg. All sequences analysis was provided in the Quantitative Insights Into Microbial Ecology (QIIME, version 1.9.1) software suite, according to the Qiime tutorial (http://qiime.org/) with some modified methods. Chimeric sequences were removed using usearch61 with de novo models. Sequences were clustered against the 2013 Greengenes (13_5 release) ribosomal database’s 97% reference data set. Sequences that did not match any entries in this reference were subsequently clustered into de novo OTUs at 97% similarity with UCLUST. Taxonomy was assigned to all OTUs using the RDP classifier within QIIME and the Greengenes reference data set. Rarefaction and rank abundance curves were calculated from OTU tables using alpha diversity and rank abundance scripts within the QIIME pipeline. The hierarchical clustering based on population profiles of most common and abundant taxa was performed using UPGMA clustering (Unweighted Pair Group Method with Arithmetic Mean, also known as average linkage) on the distance matrix of OTU abundance. This resulted in a Newick formatted tree, which was obtained utilizing the QIIME package.

14To remove microbiota, the healthy 4-month-old male voles were housed in separate cages at a room temperature (23 ± 1 °C) and were offered water with fresh composite antibiotics (containing 100 μg/mL neomycin, 50 μg/mL streptomycin, 100 U/mL penicillin; Sigma, Germany) for 7 days as described previously. For microbiota transplantation, the caecal contents were collected from three donors each of CH, CS, and WS male voles, diluted (200 mg) in 0.9% sodium chloride injection (physiological saline, 2 mL), and then, a 200 μL suspension was delivered by intragastric gavage to each bacteria-restricted recipient vole (n = 6 voles/group). For the control group, saline (200 μL) was delivered by intragastric gavage to each animal. The recipients were still housed in separate cages at room temperature after CMT. Body weight and energy intake were measured during antibiotic treatment and within 1 week after CMT, and RMR and NST were determined 1 week after CMT. After the experiment, the voles were sacrificed and caecal contents were collected for 16S rRNA sequencing (n = 2–4/group) and SCFAs measurement.

15The software SPSS 17.0 was used for statistical analyses. Prior to statistical analyses, all the data were examined for assumptions of normality and homogeneity of variance by Kolmogorov–Smirnov and Levene tests, respectively. The data with abnormal distribution were transformed by natural logarithms to normalize them. The data for body weight gain were transformed by arcsine and analyzed by two-way ANOVA (cold and huddling). Energy intake, RMR and NST were analyzed by two-way ANCOVA with body weight as a covariate. Tsurface, Tcore, the concentration of SCFAs, and microbial composition were analyzed by two-way ANOVA. Significant group differences were further evaluated using Bonferroni post hoc tests. All values were presented as mean ± SEM (standard error of mean), and P < 0.05 was considered to be statistically significant.

16For microbiota data, to account for any bias caused by uneven sequencing depth, the least number of sequences present in any given sample from a sample category was selected randomly prior to calculating community-wide dissimilarity measures (α-diversity and β-diversity). We rarefied the OTU table to a sequencing depth of 12,000 per sample for both diversity analyses. All principal coordinate analyses (PCoA) were based on unweighted and weighted UniFrac distances using evenly sampled OTU abundances. Significance for PCoA (β-diversity) analyses was checked with multivariate permutation tests using the nonparametric method “ADONIS” (999 permutations) included in the package “vegan” of the QIIME-incorporated version of “R”. The boxplot representation of the α-diversity results was done with STAMP and the calculation of P values was done with Welch’s t test. The linear discriminant analysis (LDA) Effect Size (LEfSe) method was used to assess differences in microbial communities using a LDA score threshold of 2.

17Body weight at the end (day 28) of acclimation was not affected by huddling or cold (Additional file 1Table S2). The voles in cold air temperature (Ta) tended to gain less body weight than those in warm (F1,64 = 3.467, P = 0.067). Body weight gain was not affected by huddling (F1,64 = 1.120, P = 0.294) or the interaction between cold and huddling (F1,64 = 0.174, P = 0.678; Fig. 1a). Cold led to a significant increase in energy intake compared with warm condition (F1,22 = 46.259, P < 0.001; Fig. 1b). The huddling voles consumed less energy than separated voles in cold and warm conditions (F1,22 = 28.433, P < 0.001). Energy intake was not affected by the interaction between cold and huddling (F1,22 = 1.352, P > 0.05).

18At the end of experiment, huddling voles had lower RMR than separated voles (F1,26 = 27.041, P < 0.001; Fig. 1c). RMR was affected by cold (F1,26 = 47.378, P < 0.001), but was not affected by the interaction between cold and huddling (F1,26 = 2.550, P = 0.122). RMR of huddling voles was 37% lower than in separated ones, both in cold (P < 0.001) and warm Ta (P = 0.05) (Fig. 1c). The huddling voles also had lower NST than separated voles (F1,24 = 8.458, P = 0.008; Fig. 1d). NST was marginally affected by cold (F1,24 = 3.815, P = 0.063), but was not affected by the interaction between cold and huddling (F1,24 = 0.928, P = 0.345). NST of CH voles was 31% lower than CS voles (P = 0.01), but was not different from WH voles (P > 0.05) (Fig. 1d).

19The Tsurface including Tm and Tp was affected significantly by huddling (P < 0.001), cold (P < 0.001) and their interaction (P < 0.05; Fig. 2a, b). Huddling voles had higher Tm and Tp than separated voles both in warm (by 2 °C) and cold (by 3.1 °C for Tm and 4.4 °C for Tp) conditions. Cold exposure resulted in lower Tm (by 4.6 °C in huddling and by 5.8 °C in separated group) and Tp (by 8 °C in huddling and by 10.4 °C in separated group) than in warm Ta. The Tc was 7.5 °C lower in cold huddling than in warm huddling voles (t = 10.917, df = 13, P < 0.001).

20In the first 2 days of exposure to cold, Tcore dropped immediately both in huddling and separated voles. The separated voles in cold increased Tcore to the same level to the voles in warm, while the huddling voles continued to decrease Tcore (Fig. 2c). After 10 days of exposure, the voles kept relatively stable Tcore in cold Ta. Cold-exposed voles decreased their daily Tcore by 1 °C compared with voles in warm Ta (F1,52 = 994.295, P < 0.001) (Fig. 2d, e). Tcore was 0.3 °C lower in huddling voles than separated voles both in warm and cold conditions (F1,52 = 91.581, P < 0.001). The daily Tcore was not affected by the interaction between cold and huddling (F1,52 = 0.500, P > 0.05).

21Huddling voles had a higher phylogenetic diversity (α-diversity) than separated voles (P = 7.85e−4), whereas the effect of cold temperature on phylogenetic diversity was not significant (P = 0.938) (Fig. 3a, Additional file 1: Table S1, Additional file 1: Table S3). The PCoA of unweighted (Fig. 3b) and weighted (Fig. 3c) UniFrac distances (β-diversity) between each sample among four groups showed that the microbial communities were significantly separated by cold at the first principal coordinate (PC1 axis) except for the WH group and also by huddling at the second principal coordinate (PC2 axis) in the cold (Fig. 3b, Additional file 1: Table S4). There were significant differences in UniFrac distances within groups (F7,196 = 11.261, P < 0.001; Additional file 1: Figure S2). Intergroup UniFrac distances were markedly higher than intragroup distances (P < 0.001), indicating distinctive microbial community structures in each group. The intragroup distances were significantly affected by huddling (F1,56 = 4.903, P = 0.031) and the interaction of huddling and cold (F1,56 = 11.461, P = 0.001), but not affected by cold (F1,56 = 0.976, P = 0.328).

22Analysis at the phylum level showed that the dominant phyla (mean relative abundance > 1%) in voles included Firmicutes (76.9%) and Bacteroidetes (13.6%). Rare phyla included Spirochaetes (4.9%), Proteobacteria (3.2%), Cyanobacteria (0.6%), Actinobacteria (0.4%), and TM7 (0.1%) (Fig. 3d). Analysis at the family level showed that the dominant six families were Clostridiales (38.7%), Ruminococcaceae (23.1%), S24–7 (11.1%), Lachnospiraceae (10.9%), Spirochaetaceae (4.9%), and Lactobacillaceae (1.9%) (Fig. 3e).

23To assess differences in microbial communities affected by huddling and cold, we applied LEfSe method with LDA score > 2 (Fig. 3f, g). The results identified 2 discriminative features in the microbiota of CH voles, 9 in CS, 2 in WH, and 21 in WS voles. Huddling induced major shifts of the microbiota composition, with significant decreases in proportions of Cyanobacteria, Tenericutes and TM7 at the phylum level, and of RF39, Rickettsiales, Comamonadaceae, and Sinobacteraceae at the order or family level, but increases in Lachnospiraceae and Veillonellaceae when compared with the separated voles (Fig. 3f, h–m, Additional file 1: Table S5). The proportions of Tenericutes, RF39, Lactobacillaceae, Peptococcaceae, and Clostridiaceae were lower, but of Lachnospiraceae were higher in cold voles than in warm voles. The proportions of Bacteroidetes were affected by the interaction of cold and huddling and were higher in CH and WS groups than in CS and WH groups.

24The huddling voles increased acetic acid concentration in cold (F1,26 = 6.256, P = 0.019), but decreased isovaleric acid concentration in warm (F1,26 = 6.311, P = 0.019) when compared with separated voles (Fig. 4). Cold exposure induced increases in the concentrations of acetic acid in huddling voles (F1,26 = 4.649, P = 0.041) and butyric acid in separated voles (F1,26 = 4.588, P = 0.042) when compared with warm conditions. The concentrations of propionic acid, isobutyric acid, and valeric acid were not affected by cold or huddling (P > 0.05). Total SCFAs were higher in cold than those in warm (F1,26 = 5.293, P = 0.030) and in huddling than those in separated voles (F1,26 = 5.710, P = 0.024). The isovaleric acid concentration was affected by the interaction between cold and huddling (F1,26 = 5.094, P = 0.033).

25There was no difference in body weight among all groups during antibiotic treatment (F3,20 = 0.038, P = 0.990) or after CMT (F3,20 = 0.135, P > 0.05; Fig. 5a). Energy intake among groups did not differ during antibiotic treatment (F3,19 = 0.253, P = 0.871), but CMT from CS voles led to an increase in energy intake at day 3 (F3,19 = 10.253, P < 0.001) and day 6 (F3,19 = 8.914, P = 0.001; Fig. 5b) compared to the other groups. The voles transplanted with CS microbiota had higher RMR than the other three groups (F3,19 = 6.688, P = 0.003; Fig. 5c) and also had higher NST than the control group (F3,19 = 5.001, P = 0.010; Fig. 5d).

26CMT resulted in significant changes in the concentrations of acetic acid (F3,20 = 3.802, P = 0.026), butyric acid (F3,20 = 4.265, P = 0.018), isobutyric acid (F3,20 = 5.909, P = 0.005), and valeric acid (F3,20 = 7.910, P = 0.001) among groups (Fig. 5e, f). Voles transplanted with CS microbiota had higher concentrations of butyric acid and valeric acid than the other groups (P < 0.05; Fig. 5e, f).

27The data of 16S rRNA sequencing showed that antibiotic treatment and microbiota transplantation markedly altered microbial quantity as assessed by Shannon, chao1, observed OTUs, and PD whole tree (α-diversity, Additional file 1: Figure S3) and altered microbial community composition measured by unweighted UniFrac distances (β-diversity, Additional file 1: Figure S4 a–c). Group differences in microbial communities affected by antibiotic treatment and microbiota transplantation were found by LEfSe method with LDA score > 2 (Additional file 1: Figure S4 d). We also observed that the recipients (R) had the similar abundance of biomarkers as the donors (D) in the respective groups (Additional file 1: Figure S4 e–j).

28Huddling is a social thermoregulatory behavior in mitigating cold stress in endotherms. In the present study, we investigated the underlying coevolutionary and physiological mechanisms of huddling in regulating energy intake and thermogenesis from the new aspect of gut microbiota. We observed that huddling voles had higher Tsurface but lower Tcore than voles that could not huddle (separated). Huddling also resulted in decreases in energy intake, RMR, and NST in both warm and cold Ta. Both cold and huddling resulted in marked alterations in caecal bacterial composition at the phylum and family levels and also significant changes in microbial metabolites. Further, CMT indicated that huddling- or cold-induced variations in caecal microbiota regulate energy intake and metabolic rate of the host. These results suggest that the alteration in core temperature in the gut by huddling or cold may affect bacterial growth and activity and, hence, induce the observed changes in energy intake and thermoregulation.

29To maintain normal body temperature under cold Ta, mammals use many physiological mechanisms, including increased heat production, heat conservation, and energy intake. In small mammals, RMR and NST increased during cold exposure in Brandt’s voles, prairie voles (Microtus ochrogaster), and root voles (Microtus oeconomus). In the present study, although cold exposure induced an increase in RMR, huddling voles had a lower RMR than separated voles both at 4 and 23 °C. There is a lower RMR in huddling than in separated animals in many other wild mammals, such as social degus, short-tailed field voles, bank voles (Clethrionomys glareolus), townsend’s voles (Microtus townsendii), and African four-striped grass mice (Rhabdomys pumilio). Besides the lower RMR, the huddling voles did not increase NST in the cold indicating an effective social behavior for heat conservation in small mammals.

30Due to the lower energy expenditure, the huddling voles consumed less energy than separated voles in both cold and warm Ta. This was consistent with previous studies in mice (Mus musculus), golden mice (Ochrotomys nuttalli), furred Siberian hamsters (Phodopus sungorus), and social degus at low Ta. In contrast to warm Ta, however, the voles in cold would gain less body weight and/or body fat (data not shown) due to high metabolic costs, which was also found in the previous studies in voles. The thermoneutral zone of Brandt’s vole ranges between 27.5 and 32.5 °C and, therefore, the warm Ta of 23 °C used in this experiment was below the lower critical temperature. Therefore, huddling in warm group also had an energetic benefit.

31Maintenance of high and constant Tcore over a wide range of Ta is a high evolutionary feature of endotherms. Huddling animals reduce their thermal conductance in cold through the reduced surface to volume ratio. In addition, the contact increases skin thickness in the area of contact and thus heat transfers between voles with less loss to the environment. The decrease in thermal conductance in huddling animals may reduce temperature loss to the environment and thus result in higher Tsurface compared with separated animals. Further, the increased Tsurface as warm resources may attract voles to huddle, especially in the cold.

32We found that huddling voles had a lower Tcore than separated voles, and CH voles had the lowest Tcore among the four groups. A similar result of lower Tcore under huddling has been recorded in birds where the goslings of greater snow geese (Chen caerulescens atlantica) in cold arctic environments had a Tcore by 0.3 °C when huddling, and emperor penguin decreased Tcore by 0.9 °C and increased local Tsurface during huddling in pair period in Antarctic winters. We assumed that, during huddling, the voles were resting and reduced their vigilance and control system of internal heat balance, which also allowed a gradual decrease in Tcore to the possible minimum set-point of body temperature. Moreover, the decrease in Tcore in response to a low Ta supports the importance of sensors of the thermal environment in determining the set-point to which Tcore is regulated. Other explanations for decreased Tcore in huddling mostly came from the reduced metabolic rate in the present study and from previous studies. These data indicate that huddling contributes to the reduced set-point of Tcore and temperature gradient between Tcore and Tsurface to reduce individual internal heat loss and conserve energy.

33Gut microbial diversity is vulnerable to the environments of wild mammals. For example, seasonal variations in gut microbial diversity were reported in wild wood mice (Apodemus sylvaticus), ground squirrels (Ictidomys tridecemlineatus), and wild black howler monkeys (Alouatta pigra). Gut microbiota also changed its diversity with different altitudes and diet diversity in plateau pikas (Ochotona curzoniae), yaks (Bos grunniens), and Tibetan sheep (Ovis aries). We observed that huddling significantly increased the diversity of caecal microbiota compared with separated condition. Both cold and huddling could induce marked changes in caecal bacterial composition at the phylum and family levels, while huddling decreased the abundance of Cyanobacteria, Tenericutes, TM7, Comamonadaceae, and Sinobacteraceae. Most of these are pathogenic bacteria and have been found to be associated with host inflammatory mucosal diseases. Moreover, huddling increased the abundance of Lachnospiraceae and Veillonellaceae. The family Lachnospiraceae has been linked to protection from colon cancer mainly due to the production of butyric acid. The increased abundance of Veillonellaceae was previously found in the intestinal microbiota of breast-fed infants and in human type 2 diabetes following prebiotic fiber intake, implying a beneficial effect due to the production of acetate and propionate. Therefore, these data indicate that huddling may shape the hosts to develop a healthier gut microbial community.

34The shifts in caecal microbiota were associated with a reduction in Tcore during cold exposure and in huddling in Brandt’s voles. Hibernating mammals exhibit an annual temperature rhythm and alter the diversity and composition of gut microbiota over the circannual hibernation cycle. Most mammals such as mice, desert hamsters (Phodopus roborovskii), and even humans show daily body temperature rhythm and the diurnal variation of gut microbiota has been reported in mice. A recent study showed that a 2–3 °C increase in ambient temperature can cause a 34% loss in gut bacterial diversity in a vertebrate ectotherm, the common lizard (Zootoca vivipara). Although there is still no direct evidence about the relationship between the host body temperature and gut microbiota in mammals, it is possible that a reduced core temperature in the gut may change bacterial growth and activity and, hence, the observed changes in gut microbial community.

35Microbial metabolites, by interacting with enteric nervous system, provide the causal links between environment-induced alterations in gut microbiota and the physiological and behavioral responses of the host. We observed that the alteration in caecal microbiota was followed by changes in their metabolites. Cold exposure induced increases in concentrations of acetic acid in huddling voles and butyric acid in separated voles. A number of data indicated that the principal products of caecal fermentation of dietary fiber, such as acetic, butyric, and propionic acids, by acting on the free fatty acid receptors (FFAR2 and FFAR3, previously named G-protein-coupled receptors GPR43 and GPR41) contributed to regulating the release of anorexigenic hormones peptide YY (PYY) and glucagon-like peptide (GLP)-1 from the gut and leptin secretion from adipocytes, and controlled appetite, energy intake, and thermogenesis. Thus, the increased SCFAs, particularly acetic acid and butyric acid in cold and/or huddling animals, help maintain metabolic and thermal homeostasis.

36In the present study, the changes in energy intake, RMR and NST were accompanied by variations in caecal microbiota during cold exposure and during huddling. In addition, the voles transplanted with CH microbiota had a lower energy intake and RMR than voles transplanted with CS microbiota, confirming that gut microbiota regulated energy intake and metabolic rate of the host. Further support was provided by the findings that cold microbiota transplanted to germ-free mice led to an increase in white fat browning and energy expenditure. The concentrations of SCFAs also showed that CMT altered microbial metabolites, as CS microbiota increased the concentrations of acetic acid and butyric acid. These data provide further support for microbial functioning as a potential player in the shifted host metabolic physiology.

37For the first time, the current study presents new insight into the coevolutionary mechanism of host huddling and gut microbiota in thermoregulation and energy saving for winter survival in endotherms. Huddling voles reduced energy intake and conserved energy by reducing metabolic rate and set-point of Tcore. A healthier gut microbial community was detected in huddling than in separated voles which produced more SCFAs, particularly in cold Ta. Furthermore, CMT induced the alteration in energy intake, metabolic rate, microbial communities, and metabolites. These data demonstrate that the environment-associated reduction in host Tcore may change the gut microbial community, and the remodeling of gut microbiota and their metabolites mediate cold- and huddling-induced energy intake and thermoregulation and therefore orchestrates host metabolic and thermal homeostasis (Fig. 6).