Reversible insulin resistance helps Bactrian camels survive fasting

Camels have hunger tolerance and can adapt to the severe environment of the desert. Through the comparison of insulin signalling pathway genes in different tissues in different eating periods (feeding, fasting and recovery feeding), it was found that IRS1, PIK3CB, PIK3R1 and SLC2A4 expression was significantly downregulated in the fore hump and hind hump during the fasting period. In addition, there was no difference in serum insulin levels among the three stages. However, the serum leptin and adiponectin levels decreased significantly during fasting. Additionally, insulin tolerance tests during the three stages showed that camels were insensitive to insulin during fasting. Further study of the serum metabolites showed that serum branched-chain and aromatic amino acid levels increased during the fasting period. Finally, analysis of microbial diversity in camel faeces at different stages showed that during the fasting period, the proportion of Firmicutes and Actinobacteria increased, while that of Bacteroides and the butyrate-producing bacterium Roseburia decreased. The results of this study show that fasting is accompanied by changes in the activation of insulin pathways in various camel tissues, normal insulin levels, and increased lipolysis and insulin resistance, which return to normal after eating.

www.nature.com/scientificreports/ fully recover, the recovery period lasted until the 38th day, when the body weight was basically the same as that on the 0th day ( Fig. 1).
Insulin pathway gene expression. The results (Fig. 2) showed that in fore hump and hind hump, insulin receptor substrate-1 (IRS1), PIK3CB, PIK3R1 and solute carrier family 2 member 4 (SLC2A4) expression was upregulated in feeding period and recovery feeding period but downregulated in fasting period. At the same time, in subcutaneous fat, at fasting period, IRS1 expression was upregulated, but PIK3R1 expression was downregulated. It is worth noting that in biceps femoris muscle, in fasting period, INSR, IRS1, PIK3CB, MTOR, AKT3, GSK3A, PTEN, TSC2 and SLC2A4 expression was upregulated, and only PIK3R1 expression was downregulated.
Serum phenotypes and insulin resistance test. There was no difference in serum insulin levels among the three periods (Fig. 3A). However, the blood glucose (Fig. 3B) and serum leptin (Fig. 3C) and adiponectin ( Fig. 3D) levels in fasting period were significantly lower than those in feeding and recovery feeding period.
Black, red and grey asterisks indicate significant differences compared to a, b and c, respectively. All error bars represent ± SEM. www.nature.com/scientificreports/ period. Finally, the insulin tolerance test (ITT) results in the three stages showed that camels were insensitive to insulin in stage b (Fig. 3G).
Metabolomics analysis of serum. First, the serum metabolites in three stages were analysed by principal component analysis (PCA). The results showed that serum in fasting period was significantly separated from serum in feeding and recovery feeding period (Fig. 4A). Furthermore, to investigate the changes in serum metabolites before and after fasting, we compared and analysed the differences. A total of 445 differentially expressed metabolites (DEMs) were identified in serum in feeding vs. fasting period, of which 298 and 147 metabolites in serum in feeding and fasting period were enriched, respectively (Table 1, Fig. 4B). In addition, a total of 532 DEMs were identified from serum in fasting vs. recovery feeding period, of which 183 and 349 metabolites in serum in fasting and recovery feeding period were enriched, respectively (Table 1, Fig. 4C). The above two groups of DEMs were analysed by Venn analysis. A total of 365 overlapping metabolites were obtained, which were involved in the whole process of stage a to stage b and stage b to stage c changes in serum (Fig. 4D). Furthermore, the KEGG pathways of these 365 DEMs were analysed, and 32 KEGG pathways were identified (P adjusted < 0.05, Supplementary Table 1, 2). These pathways are mainly involved in amino acid metabolism, biosynthesis of other secondary metabolites, carbohydrate metabolism, membrane transport, metabolism of cofactors and vitamins, endocrine and metabolic diseases, lipid metabolism and other biological processes. The top 5 KEGG pathways were secondary bile acid biosynthesis, central carbon metabolism in cancer, protein digestion and absorption, phenylalanine metabolism and aminoacyl-tRNA biosynthesis. In addition, the analysis of branched-chain amino acids (BCAAs) and aromatic amino acids (AAAs) in serum showed that glycine, glutamic acid, alanine, isoleucine, tyrosine, valine, leucine, tryptophan, proline and phenylalanine levels were significantly upregulated in serum in fasting period (Fig. 4E).
Analysis of microbial diversity of faeces. Across all 24 faecal samples, 1,453,555 high-quality sequences were classified as bacteria, with an average length of 433 bp (Supplementary Table 3). The rarefaction curve (average curve of each group) results showed that most of the microbial diversity was fully captured (Fig. 5A). The statistical estimates of α-diversity showed that the Chao1 diversity index, Shannon estimator and Sobs diversity index of faeces in fasting period were significantly lower than those in feeding and recovery feeding period ( Fig. 5B-D).
The results of the nonmetric multidimensional scaling (NMDS, Fig. 5E) ordination plot and principal coordinates analysis (PCoA, Fig. 5F) showed that the bacterial communities in faeces were obviously distinct in the feeding, fasting and recovery feeding period. To further examine the bacterial groups in different dietary stages, the bacterial groups with a relative abundance of more than 1% were classified and analysed. At the phylum level, Firmicutes and Bacteroidetes were the most predominant phyla in the faeces (Fig. 5G). Furthermore, the Kruskal-Wallis H test was used to analyse the community abundance data of each faecal sample. The results showed that in the faeces in fasting period, the proportions of Firmicutes, Proteobacteria, Actinobacteria and Lentisphaerae increased, while the proportions of Bacteroidetes and Tenericutes decreased ( Fig. 6, P < 0.01).
To explore the function of microorganisms in the samples, we used PICRUSt for functional analysis. Comparing the predicted clusters of orthologous groups of proteins (COG) and KEGG functions of the faeces in fasting period with those of the faeces in feeding and recovery feeding period, we found that the faeces in fasting period had strong functions of carbohydrate metabolism, amino acid metabolism, lipid transport and metabolism, biosynthesis and catabolism of secondary metabolites, inorganic ion transport and metabolism ( Fig. 7, P < 0.01). However, the G protein-coupled receptors, apoptosis, ether lipid metabolism, nucleotide metabolism, and cell division energy metabolism pathways the faeces in fasting period were significantly downregulated (Supplementary Table 4, P < 0.01).

Discussion
Although all animals experience feeding and fasting periods, the species that are most interesting are those that can survive fasting for a long period of time. How to fast for a long time by relying only on endogenous resources has been a hot topic for researchers for a long time 7 . At present, there are few studies on camel fasting and hunger tolerance, and the mechanism remains to be further clarified. In this study, qRT-PCR, 16S rDNA sequencing and metabonomics techniques, combined with phenotypic tests, were used to reveal the strategies that Bactrian camels use to cope with fasting.
IRS1 plays an important role in controlling the dynamic balance of growth and nutrition and acts as an on/ off switch to transduce insulin action 8,9 . In addition, IRS1 is closely related to PI3K (PIK3CB and PIK3R1) and AKT (AKT1, AKT2 and AKT3) activation. Loss of IRS1 can cause PI3K inactivation and eventually lead to insulin resistance 10 . Insulin resistance is also associated with decreased expression of the insulin-sensitive glucose transporter 4 protein (GLUT4) encoded by SLC2A4 11 . GLUT4 is the main glucose transporter subtype in insulin-responsive tissues. Genetically engineered mice overexpressing the exogenous GLUT4 gene in skeletal muscle or adipose tissue exhibit increased insulin responsiveness and peripheral glucose utilization 12 13 . Therefore, we speculate that in fasting period, the insulin signalling pathway is inhibited in fore hump and hind hump, while it is enhanced in biceps femoris muscle, and its function in  www.nature.com/scientificreports/ subcutaneous fat remains to be clarified. According to reports, the adipose tissue and muscles of brown bears exhibit insulin resistance during hibernation, which helps them survive the long winter 14 . Interestingly, the camel fore hump and hind hump exhibit insulin resistance during the fasting period, but the insulin sensitivity of the muscle seems to be enhanced. We speculate that the differences in strategies for coping with famine may be related to their respective living environments. Camels living in the Gobi Desert face food shortages not only in winter but also in other seasons. Therefore, camels do not rely on sleep to reduce the energy consumption of the whole body, as the brown bear does, but instead use the internal coordination of the body; camels therefore first guarantee the energy supply of more critical parts, such as muscles.
We initially assumed that prolonged fasting would induce hypoinsulinaemia and lead to loss of activation of the insulin pathway 15 . This may be the reason for the downregulation of the expression of the main insulin pathway genes of fore hump and hind hump in fasting period. Contrary to our expectations, insulin levels of all camels did not change at all three stages (Fig. 3A). This result is consistent with hibernating animals during the winter fasting period 16 . This further indicates that during fasting, the insulin sensitivity of hump tissue decreases, while that of muscle tissue increases. The level of blood glucose in stage b decreased significantly, which was closely related to fasting (Fig. 3B). Fasting reduces circulating leptin levels, while eating or obesity increases leptin levels [16][17][18] , which is consistent with our results (Fig. 3C). Adiponectin can enhance peripheral insulin sensitivity, which is negatively correlated with insulin resistance [19][20][21][22][23][24] . In addition, elevated serum NEFA levels can induce insulin resistance 25,26 . Therefore, serum adiponectin (Fig. 3D) and NEFA levels (Fig. 3E) support the decreased insulin sensitivity of camels in stage b.
In addition to controlling blood glucose, insulin also effectively inhibits lipolysis of adipose tissue 27 , which is a multistep process that promotes the triglycerides into glycerol and free fatty acids. However, the results showed that the serum triglyceride content of camels in fasting period was significantly increased (Fig. 3F), indicating active lipolysis. It has been reported that a high lipolysis rate is associated with reduced human insulin sensitivity 28 . Therefore, camels may have insulin resistance in stage b. For insulinaemia, the different activation of the insulin pathway at different stages indicates that the camel's response to insulin is different. To solve this problem directly, we conducted an ITT. The results showed insulin resistance in stage b camels (Fig. 3G). This finding also indicates that the downregulation of the expression of insulin pathway genes in F and H is positively correlated with insulin resistance. Collectively, these data suggest that Bactrian camels modulate states of reversible insulin responsiveness.
In view of the change in insulin sensitivity of camels caused by fasting, we then investigated its effects on metabolites in camel serum. PCA shows (Fig. 4A) that the differences between data sets are largely due to fasting, which is consistent with the trend of insulin pathway genes. In recent years, scientists have found that BCAAs [29][30][31][32][33][34][35] and AAAs [36][37][38] in blood are positively correlated with insulin resistance. This finding is consistent with our results (Fig. 4E). In stage b, serum BCAA (leucine, isoleucine and valine) and AAA (tyrosine, phenylalanine and tryptophan) levels increased significantly, which further indicated that camels may be in a state of insulin resistance during fasting.
One of the characteristics of the serum metabolites of individuals with insulin resistance is an increased serum BCAA level, which is related to the intestinal microflora and has rich potential for BCAA biosynthesis 39 . The increase of branched chain amino acids in insulin resistance may be related to the changes of peripheral amino acid metabolism, but intestinal microflora has been proved to be very important for the supply of leucine, isoleucine and valine (BCAA) in mammalian hosts. In addition, increasing evidence shows that there is a link between the intestinal microflora and metabolic health [40][41][42][43] . The insulin resistance phenotype is transferable through faecal microbiota transplantation (FMT) 44,45 . Therefore, we analysed the faecal microbial diversity of camels in different stages.
It has been reported that hydrolysis and fermentation of dietary polysaccharides by the intestinal microflora will produce a large number of monosaccharides and short-chain fatty acids (SCFAs) 46 , which can be absorbed and utilized as energy by the host, and obesity-associated microflora generate energy from the diet more effectively 47,48 . Through the study of humans and animals, it has been found that a higher ratio of Firmicutes to Bacteroidetes in the intestinal tract can result in more effective energy acquisition from the diet and may lead to obesity 49,50 , and at the same time, it will promote the increase of BCAA 51 . It was also reported that insulinresistant subjects showed an increase in the relative abundances of Firmicutes and Actinobacteria and a decrease in the relative abundance of Bacteroides 52 . This finding is consistent with our results (Figs. 5G, 6), indicating that camels may exhibit insulin resistance during the fasting period, and the camel intestinal microflora is more efficient at obtaining energy from gastrointestinal contents to better survive through the fasting period.
Studies have shown that FMT from lean male donors significantly improved insulin sensitivity in male patients with metabolic syndrome and increased intestinal microbial diversity, including a significant increase in the abundance of butyrate-producing strains 46 . Compared with that in the intestinal flora of healthy people, Table 1. Comparison of different metabolites in serum (false discovery rate (FDR) < 0.05, variable importance in the projection (VIP) > 1, fold change (FC) > 2, or FC < 0.5, differentially expressed metabolites (DEMs), serum (E), a: feeding period, b: fasting period, c: recovery feeding period).  that is, the abundance of Roseburia decreased 41,42 . Our research shows that the Roseburia abundance of camels decreased significantly during the fasting period (Fig. 5H). Studies have shown that the increase of Proteobacteria will promote the synthesis of BCAA, which is consistent with our research (Fig. 6) 53 .Taken together, these results indicate that the intestinal flora of camels during the fasting period is similar to that of obese or diabetic patients, which may lead to insulin resistance. It should be pointed out that the investigation of the genes involved in insulin signaling has importance, but does not define the real insulin signaling pathway, which depends on tyrosine kinase activity of insulin receptor, downstream tyrosine phosphorylation and serine phosphorylations. This is the limitation of this study, and will be explored in follow-up studies.
The results of this study showed that during the fasting period, the insulin sensitivity of fore hump and hind hump decreased, while that of muscle increased, but overall, the camels were in a state of insulin resistance, and lipolysis increased. Additionally, serum BCAA and AAA levels, as well as the abundances of Firmicutes, Actinobacteria, Bacteroides and Roseburia, can be used as candidate markers in camels for the diagnosis of insulin resistance; thus, these results provide directions for future research. Our findings reveal the molecular mechanism of camel tolerance to hunger and may provide a new animal model for human research on metabolic diseases.

Materials and methods
Animals and sample collection. We used 6 male domestic Bactrian camels (4 years old) that were born in the same captive environment and trained to meet the needs of the experiment. These camels were kept in individual pens. Camels ate and drank freely for a long time, mainly including alfalfa and corn (dry matter www.nature.com/scientificreports/ 28.75 g/100 g, protein 4.46 g/100 g, calcium 0.45 g/100 g, phosphorus 0.07 g/100 g, carotene 4.4 mg/100 g), which lasted until day 0. Before day 0 was the feeding period (a), 1 to 14 days was the fasting period (b), no water and food, and 14 to 38 days was the recovery feeding period (c), the diet was the same as that of stage a. Experiments and sample collection were carried out on the 0th day, 14th day and 38th day (Fig. 8). The following samples were collected: fore hump (F), hind hump (H), subcutaneous fat (S), biceps femoris muscle (M), serum (E) and faeces (D). Tissue samples were biopsied using a 6 mm circular punch. Each area was shaved and aseptically prepared for biopsy. Subcutaneous fat was obtained from the hindlimb region over the gluteus muscles. Muscle samples were obtained with the 6 mm punch via a short stab incision through the dermis over the lateral aspect of either the left or right biceps femoris muscle. Tissue samples were placed in a 2 ml freezing tube, immediately immersed in liquid nitrogen, and then stored at -80 °C until the subsequent test. After camels defecated, faecal samples were quickly collected and immediately preserved in liquid nitrogen and then stored at − 80 °C until DNA extraction. Whole blood was collected from the jugular vein into vacutainers without anticoagulant. For clinical chemistry panels (CCPs), blood without anticoagulant was centrifuged at 3000× g at 4 °C for 15 min to obtain serum. The separated serum was immediately analysed by an automatic biochemical analyser (Hitachi   Insulin tolerance tests and challenges. Intravenous injection of 0.4 U/kg 55 insulin (Novolin, Novo Nordisk) was conducted on 6 camels. Then, a blood glucose metre (Accu-Chek Active, Roche Diagnostics, Basel, Switzerland) was used to measure blood glucose levels. The separated serum was packed in a 200 µl sterile tube and stored at − 80 °C for later verification of blood glucose metre readings. When the blood glucose concentration was lower than 2.5 mmol/l, a 50% glucose solution was immediately injected intravenously, and the tolerance test was completed. This experiment was carried out on days 0, 15 and 38, respectively. 16S rRNA sequencing. Microbial genomic DNA was extracted from faecal samples (D) using a Mag-Bind Soil Kit (Omega, M5635). After the genomic DNA extraction was completed, 1% agarose gel electrophoresis was used to detect the quality of the extracted DNA, and a NanoDrop2000 was used to determine the concentration and purity of the DNA. The V3-V4 hypervariable region of the 16S rRNA gene was amplified by PCR with 338F (5′-ACT CCT ACG GGA GGC AGC AG-3′) and 806R (5′-GGA CTA CHVGGG TWT CTAAT-3′). After mixing the PCR products of the same sample, a 2% agarose gel was used to recover PCR products, an AxyPrep DNAGel Extraction Kit (Axygen Biosciences, Union City, CA, USA) was used to purify the recovered products, 2% agarose gel electrophoretic detection was used, and a QuantusTM Fluorometer (Promega, USA) was used to detect and determine the amount of the recovered products. A NEXTFLEX Rapid DNA-Seq Kit was used to build the library. The MiSeq PE300 platform (Illumina) was used for sequencing.

Metabolomics analysis.
Statistical analysis. GraphPad Prism version 8.0.2 was used to analyse the serum biochemical indexes and qRT-PCR data. MetaboAnalyst4.0 (http:// www. metab oanal yst. ca) software was used to analyse the differences in and enrichment of metabolites. Microbial diversity was analysed using the R 3.6.3 software package (https:// www.r-proje ct. org). The sequencing data from this study were deposited in the NCBI Sequence Read Archive (SRA) under accession number PRJNA714601.
Ethics statement. All experimental design and procedures of this study was carried out in compliance with the ARRIVE guidelines (https:// arriv eguid elines. org). The procedures and protocols were approved by the animal care committee of the Camel Protection Association of Inner Mongolia.