Integrative Transcription Start Site Analysis and Physiological Phenotyping Reveal Torpor-specific Expressions in Mouse Skeletal Muscle

Mice enter an active hypometabolic state, called daily torpor, when they experience a lowered caloric intake under cool ambient temperature (TA). During torpor, the oxygen consumption rate (VO2) drops to less than 30% of the normal rate without harming the body. This safe but severe reduction in metabolism is attractive for various clinical applications; however, the mechanism and molecules involved are unclear. Therefore, here we systematically analyzed the expression landscape of transcription start sites (TSS) in mouse skeletal muscles under various metabolic states to identify torpor-specific transcription patterns. We analyzed the soleus muscles from 38 mice in torpid, non-torpid, and torpor-deprived conditions, and identified 287 torpor-specific promoters. Furthermore, we found that the transcription factor ATF3 was highly expressed during torpor deprivation and that the ATF3-binding motif was enriched in torpor-specific promoters. Our results demonstrate that the mouse torpor has a distinct hereditary genetic background and its peripheral tissues are useful for studying active hypometabolism.


INTRODUCTION
1 Mammals in hibernation or in daily torpor reduce their metabolic rate to 1-30% of that of 2 euthermic states and enter a hypothermic condition without any obvious signs of tissue 3 injury (Bouma et al., 2012;Geiser, 2004). How mammals adapt to such a low metabolic rate 4 and low body temperature without damage remains as one of the central questions in 5 biology. Mammals maintain their TB within a certain range by producing heat. In cold, the 6 oxygen requirements for heat production increases, at a rate negatively proportional to the 7 body size (Heldmaier et al., 2004). Instead of paying the high cost for heat production, 8 some mammals are able to lower their metabolism by sacrificing body temperature 9 homeostasis. This condition, in which the animal reduces its metabolic rate followed by 10 whole-body hypothermia, is called active hypometabolism. As a result, the homeostatic 11 regulation of body temperature is suppressed, and the total energy usage is spared. This 12 hypometabolic condition is called hibernation when it lasts for a season, and daily torpor 13 when it occurs daily. 14 Four conditions have been proposed to occur in active hypometabolism in 15 mammals (Sunagawa and Takahashi, 2016): 1) resistance to low temperature, 2) 16 resistance to low oxygen supply, 3) suppression of body temperature homeostasis, and 4) 17 heat production ability under a low metabolic rate. Of these conditions, 1) and 2) were 18 found to be cell/tissue-specific or local functions, which prompted researchers to analyze 19 genome-wide molecular changes in various tissues of hibernators, including brain, liver, 20 heart, skeletal muscles, and adipose tissues. A major role of differential gene expression in 21 the molecular regulation of hibernation was first suggested by Srere with co-authors, who 22 The goal of this study was to analyze contribution of the genetic background to the 21 torpor phenotype by introducing the mouse as model for active hypometabolism, taking 22 advantage of the rich and powerful genetic technologies available for this animal. First, we 23 found that two genetically close mice inbred strains, C57BL/6J (B6J) and C57BL/6NJcl 24 (B6N), exhibit distinct torpor phenotypes; B6N has a higher metabolism during torpor and a 25 lower rate of torpor entry. To clarify the genetic link to the mouse torpor phenotype, we 26 performed Cap Analysis of Gene Expression (CAGE) in soleus muscles taken from 38 27 animals under various metabolic conditions. We found that entering torpor and restoring 28 activity were associated with distinct changes in the transcriptomic profile, including marked 29 Torpor Prevention at High TA Revealed Hypometabolism-associated Promoters. 1 Torpor can be induced by removing food for 24 hours only when the animal is placed in a 2 relatively low TA . We have shown that B6J mice enter torpor at a rate of 100% from TA = 3 12 °C up to TA = 24 °C (Sunagawa and Takahashi, 2016) and that some animals stop 4 entering torpor at TA = 28 °C ( Figures S3A and S1E). We further tested whether the animals 5 could enter torpor at TA = 32 °C ( Figure S3B). In this warm condition, even if the animals 6 were starved they did not enter torpor, possibly due to the lack of heat loss than at lower 7 TAs. Taking these two requirements into account, fasting and low TA, we designed two 8 torpor-preventive conditions and compared the expression in the muscles under these 9 conditions to that under the ideal torpor state ( Figure 3A). One is a high TA (HiT) 10 environment and the other is a non-fasted (Fed) condition. Both conditions prevented the 11 animals from inducing torpor, because the two essential requirements were lacking. We 12 then, compared the tissue from these conditions to the ideal torpid tissue, which was from 13 fasting animals at a low TA, and obtained the transcripts that were differentially expressed 14 from torpor in each non-torpor condition. The expression differences shared in these two 15 experiments would be those affected by both low TA and fasting, and therefore would be the 16 essential expressions for active hypometabolism, hereafter defined as hypometabolic 17

promoters. 18
We first compared the VO2 in the HiT and Fed groups against the Mid group 19 ( Figure 3B). Even though both groups had no animals entering torpor, the HiT group 20 showed a lower VO2 while the Fed group showed a higher metabolism. Next, we compared 21 the expression profile acquired from the CAGE analysis of tissues from both groups. The 22 MDS plot and hierarchical clustering showed that the Mid, Fed, and HiT groups consisted of 23 independent clusters ( Figures 3C and 3D). This finding indicated that the expressions 24 during torpor (Mid group) were distinct from those during starvation alone (HiT) or at low TA 25 alone (Fed). 26 To extract the hypometabolic promoters, we performed the DE analysis ( Figure  27 3E) between the HiT to Mid and the Fed to Mid. CAGE clusters up-regulated in both the HiT 28 to Mid and the Fed to Mid were those that were upregulated during torpor regardless of the 29 13 initial condition, i.e., warm TA or no fasting (green dots in Figure 3E). There were 330 of 1 these up-regulated hypometabolic promoters from the total 12,863. On the other hand, 2 CAGE clusters that were down-regulated in both the HiT to Mid and the Fed to Mid, were 3 promoters that were down-regulated regardless of the initial condition, and thus were the 4 down-regulated hypometabolic promoters (red dots in Figure 3E). The enrichment analyses 5 of GO terms and KEGG pathways were performed ( Figures 3F and 3G), and the motifs 6 enriched in the hypometabolic promoters were also analyzed ( Figures S3C and S3D). The 7 top five promoters that had annotated genes nearby are listed as up-and down-regulated 8 hypometabolic promoters in Figures 3H and 3I, respectively. 9 These results showed that considerable numbers of genes are involved in the 10 active hypometabolic process independent from the responses to both hunger and cold. 11 One of these genes, Ppargc1a, which was found at the top of the up-regulated 12 hypometabolic promoters, was also found at the top of up-regulated reversible promoters 13 ( Figure 2H). This is a good candidate for a torpor-specific gene, because it belongs to both 14 the reversible and the hypometabolic group in this study. Therefore, we next merged the 15 results of the reversible and the hypometabolic promoters to specify the torpor-specific  Our two independent analyses, which focused on two essential torpor characteristics, i.e., 21 reversibility and hypometabolism, revealed that the skeletal muscle of torpid mice has a 22 specific transcriptomic pattern. Combining these results, we obtained torpor-specific 23 promoters, defined as the intersection of the reversible and the hypometabolic promoters. 24 We found 226 up-regulated and 61 down-regulated torpor-specific promoters ( Figure 4A). 25 The top five promoters ordered according to the sum of the fold-change observed in the two 26 groups (reversible and hypometabolic promoters) are shown in Figures 4B and 4C. 27 Remarkably, "protein binding" in the molecular function category in the GO terms was listed 28 in the top ten enriched GO terms ( Figure S4A). This group includes various protein-binding 29 genes products, including transcription factors. To highlight the predominant transcriptional 1 pathway related to torpor, we ran an enrichment study of KEGG pathways with the torpor-2 specific promoters. We obtained 13 pathways that showed statistically significant 3 enrichment ( Figure 4D). In particular, the mTOR pathway, which includes various metabolic 4 processes related to both hibernation and starvation, was identified ( Figure S4B). conditions, we analyzed the promoter shape of each of the detected promoters in the 16 reversible, hypometabolic, and torpor-specific groups ( Figure 4E). In the torpor-specific 17 groups, the down-regulated promoters showed a significantly different shape when 18 compared to all muscle promoters ( Figure 4F), while the GC richness did not show a 19 difference ( Figure S4E). 20 The torpor-specific promoters we found may represent regulators both upstream 21 and downstream of the torpor transcriptional network. To further elucidate the early events 22 involved in torpor-specific metabolism in peripheral tissues, it was necessary to place the 23 animal in a condition where it had an unusually strong tendency to enter torpor, and to 24 compare the muscle gene expression with that of normal torpor entry. gently touching the animal. Even when the mouse was not allowed to enter torpor, the VO2 28 was close to that of Mid-torpor animals ( Figure S4F). Furthermore, the transcriptome profile 29 in the muscles from torpor-deprived animals did not show a clear difference from Mid-torpor 1 animals in MDS plots ( Figure S4G). When compared to Mid-torpor muscles, the torpor-2 deprived muscles had 45 up-and 27 down-regulated promoters ( Figure 4G). Among these 3 72 torpor-deprivation-specific promoters, one promoter starting at the minus strand of 4 chromosome 1: 191217941, namely the promoter of the activating transcription factor 3 5 (atf3) gene, was also found in the torpor-specific promoters ( Figure 4H). Surprisingly, the 6 binding site of ATF3 was one of the motifs enriched in the torpor-specific promoters (Figure  7 S4H). The Atf3 motif was found in 33 of 289 torpor-specific promoters, and the peak of the 8 motif probability was at 79 bp upstream of the TSS ( Figure 4I). 9 These results showed that tissues of torpid mice have a torpor-specific 10 transcription signature. We also found that one of the torpor-specific genes, encoding 11 transcription factor ATF3 was more highly expressed during torpor deprivation. 12 Furthermore, the ATF3-binding motif was found to be enriched in torpor-specific promoters. 13 These findings were indicative of a novel pathway of active hypometabolism in peripheral 14 tissues, possibly initiated by the torpor drive-correlated transcription factor ATF3. Finally, we 15 analyzed our promoter-based expression data with respect to the SNPs of B6J and B6N, to 16 find evidence that may explain the phenotypic difference between these two inbred strains. The classic laboratory mice B6J and B6N have very few genome differences, while they 20 show distinct torpor phenotypes ( Figure 1K). We discovered that the muscles in B6J mice 21 show torpor-specific expressions ( Figure 4A). Because our data were analyzed by CAGE-22 seq, the promoter information, which is usually non-coding sequences, is directly available. 23 Because most SNPs are found in non-coding regions, it is reasonable to analyze the SNP 24 enrichment at the promoter regions of the torpor-specific expressions to explain the 25 First, we tested whether the 13 torpor-specific pathways ( Figure 4D) were affected 27 by SNPs that are different between B6J and B6N (B6J/B6N). The SNPs located in the 28 promoter region of the genes included in the pathways were counted, and the enrichment 29 was compared to the baseline to test the significance ( Figure 5A). All 13 pathways showed 1 significant enrichment (p < 0.05) indicating that the SNPs in B6J/B6N are strongly involved 2 in the torpor-specific pathways. 3 Next, we tested the enrichment of SNPs at each promoter group. There were two 4 up-and four down-regulated promoters in the torpor-specific group that had at least one 5 SNP ( Figure 5B). Possibly due to the low number of SNPs in this dataset, we were not able 6 to confirm a significant enrichment in torpor-specific promoters. The detailed position of the 7 SNP in six promoters; Plin5 and Sik3 as up-regulated and Creb3l1, Bhlhe40, Rrad, and 8 Lrn1 as down-regulated promoters, are illustrated in Figures S5A and S5B. 9 Finally, we tested how the SNPs were distributed in the promoter region in each 10 group. We calculated the SNP density at a given position from the TSS ( Figure 5C). The 11 results indicated that SNPs tended to be enriched 5 kbp upstream from the TSS of torpor-12 specific down-regulated promoters. 13 These results collectively indicated that B6J/B6N SNPs may explain the torpor 14 phenotype difference in these two strains. In particular, the SNPs that were highly enriched 15 in the torpor-specific pathways designated the possible origin of the dissimilar torpor 16 phenotypes. 17

Mouse Torpor as a Model System for Active Hypometabolism 2
One goal of this study was to introduce mouse torpor as a study model for active 3 hypometabolism. Hibernation is the most extreme phenotype of active hypometabolism, 4 and there is a physiological distinction between hibernation and daily torpor (Ruf and 5 Geiser, 2015). We recently showed that mouse torpor shares a common thermoregulation 6 mechanism with hibernation in which the sensitivity of the thermoregulatory system is 7 reduced (Sunagawa and Takahashi, 2016). 8 In this study, we extended our previous work by evaluating another inbred strain 9 B6N. Despite the close genetic distance between B6N and B6J, we found that they had 10 distinct torpor phenotypes ( Figures 1F and S1E) due to a difference in heat production 11 sensitivity ( Figure S1F). Various inbred strains are reported to have distinct phenotypes, 12 indicating a genetic involvement in torpor phenotypes (Dikic et al., 2008). Our findings 13 strengthen this idea, because B6N and B6J have a very small genetic difference but a clear 14 difference in torpor phenotypes. We also showed that the inbred strain-specific torpor 15 phenotype is inheritable ( Figures 1K and S1J), further validating the link between genetic 16 background and torpor phenotype. 17 18

Torpor-specific Transcriptions Differ from those of Hibernation and Starvation 19
In this study, we identified 287 torpor-specific promoters in mouse skeletal muscle ( Figure  20 4A). Specificity was assured by including both reversible and hypometabolic promoters 21 (Figures 2A and 3A). The results enabled us to identify likely metabolic pathways that are 22 enriched during torpor ( Figure 4D). 23 Circadian rhythm was the most enriched KEGG pathway by torpor-specific 24 promoters ( Figure 4D). The circadian clock is important in organizing metabolism and 25 energy expenditure (Tahara and Shibata, 2013). In our study, the core circadian clock gene 26 down-regulation of igf1, which encodes IGF-1, and an activation of mtor, which encodes 6 mTOR, in torpor, which appear to be paradoxical to past studies. 7 The Insulin/Akt pathway also controls the phosphorylation and activation of the 8 FOXO1 transcription factor, a disuse atrophy signature that upregulates the muscle-specific 9 ubiquitin ligases trim63 (MuRF1) and fbxo30 (Atrogin-1). In our study, we found that 10 were up-regulated during torpor, indicating that atrophic changes is also progressed. 17 Furthermore, mTOR activation was found, which is a signature of muscle hypertrophy. 18 Thus, we can conclude that mouse torpor has a unique transcription profile, sharing 19 signatures with hibernation, starvation-induced atrophy, and muscle hypertrophy. 20 21

Dynamics of Torpor-specific Transcriptions 22
The deep CAGE technology enabled us to evaluate the dynamics of the torpor-specific 23 promoters. We found that down-regulated torpor-specific promoters were narrower than 24 other muscle promoters (Figures 4E and 4F). However, the GC content of the torpor 25 promoters was not significantly different from that of all muscle promoter regions ( Figure  26 S4E): approximately half of the TSSs were located in CpG islands, in which both AT-and 27 GC-rich motifs were overrepresented (Figures S4C and S4D). 28 To gain insight about the upstream network of torpor, we evaluated a torpor-29 deprived condition. Note that this dynamic state, the torpor-deprived condition, is very 1 difficult to induce in hibernators, because very little stimulation can cause them to halt 2 torpor induction. Taking advantage of this torpor-deprivation state in mice, we identified 3 transcription factor ATF3 as a candidate factor that is correlated with the need to enter study, Atf3 was identified not only as a stress-induced gene, but also as a torpor-drive 8 correlated factor. Torpor is an active-hypometabolic condition, which can be described as a 9 physiological ischemia. Although we lack direct evidence, we propose the hypothesis that 10 Atf3 may be a factor mediating the initiation of hypometabolism, and because of that, it is 11 expressed to protect the organs under stressful conditions such as ischemia. 12 Another advantage of deep CAGE is the rich information obtained about the 13 promoter region of the expression of interest. This study exploited our finding that B6J and 14 B6N have different torpor phenotypes. To identify which SNP was responsible for the 15 phenotype difference, we used all of the data acquired in this study. Although the down-16 regulated torpor-specific promoters tended to have more SNPs, we were unable to identify 17 specific SNPs related to the torpor phenotype from observation ( Figure 5B). Therefore, 18 further study is needed to test how the candidate SNPs ( Figures S5A and S5B) affect the 19 torpor phenotype by genetic intervention. 20 21

Fundamental Understanding of Active Hypometabolism for Medical Applications 22
The overall results of this study indicate that the mouse is an excellent animal for studying 23 the as-yet-unknown mechanisms of active hypometabolism. Understanding the core engine 24 of the hypometabolism in torpid tissues will be the key to enabling non-hibernating animals, 25 including humans, to hibernate. Inducing active hypometabolism in humans would be an 26 important breakthrough for many medical applications (Bouma et al., 2012). The benefits to 27 using mice are not limited to technological advances in genetics, but extend to the 28 enormous potential for in vitro studies using cell or tissue culture. In stem cell biology, 29 patient-derived stem cells represent a valuable resource for understanding diseases and 1 developing treatments, because the cells reflect the phenotype of the patient (Avior et al., 2 2016). We believe, similarly, that mouse-derived stem cells or tissues will provide a unique 3 platform for investigating strain-specific hypometabolic phenotypes in animals. Moreover, 4 because in vitro studies can be easily extended to experiments using human cells/tissue 5 derived from human induced pluripotent stem cells, active hypometabolism research in 6 mouse cells/tissues is an important step toward the realization of human hypometabolism.   hypometabolic promoters. Up-regulated torpor-specific promoters (n = 226), which were 3 CAGE clusters that were highly expressed exclusively during torpor, were at the 4 intersection of the up-regulated reversible (n = 589) and hypometabolic promoters (n = 5 330). Down-regulated torpor-specific promoters (n = 61), which were CAGE clusters that 6 were highly suppressed exclusively during torpor, were at the intersection of down-7 regulated reversible (n = 277) and hypometabolic promoters (n = 137).