Physiological Genomics


There is currently much interest in clinical applications of therapeutic hypothermia. Hypothermia can be a consequence of hypometabolism. We have recently established a procedure for the induction of a reversible deep hypometabolic state in mice using 5′-adenosine monophosphate (5′-AMP) in conjunction with moderate ambient temperature. The current study aims at investigating the impact of this technology at the gene expression level in a major metabolic organ, the liver. Our findings reveal that expression levels of the majority of genes in liver are not significantly altered by deep hypometabolism. However, among those affected by hypometabolism, more genes are differentially upregulated than downregulated both in a deep hypometabolic state and in the early arousal state. These altered gene expression levels during 5′-AMP induced hypometabolism are largely restored to normal levels within 2 days of the treatment. Our data also suggest that temporal control of circadian genes is largely stalled during deep hypometabolism.

  • hypothermia
  • circadian
  • microarray
  • 5′-adenosine monophosphate

the effectiveness of hibernation for energy conservation has inspired scientists to contemplate the possibility of creating similar states in nonhibernating mammals including humans. There have been a few reported methods that use pharmaceutical agents to induce hypometabolic states in nonhibernating mammals. Methods include the use of subtoxic levels of metabolic inhibitors such as hydrogen sulfide (7), deoxyglucose (12), and 3-iodothyronamine (37), while we have used the naturally occurring metabolite 5′-adenosine monophosphate (5′-AMP) (11, 57). These technologies for hypometabolism induction are all in different stages of development. Whether any of them share common mechanistic pathways remains unclear.

We have shown that mice given a sufficient dose of 5′-AMP undergo a reversible and deep hypometabolic state when kept in a 15°C ambient temperature (Ta) (57). We call this state AMP-induced hypometabolism (AIHM), which typically results in a drop of animal core body temperature (Tb) to 1–2°C above Ta. We found that AIHM-driven hypothermia was successful in all species of animals that have been tested (11). AIHM-treated animals do not shiver and appear relaxed during the cool-down of body temperature (56). Mice injected with 5′-AMP enter a state of depressed metabolism within minutes with much reduced physical movement. At a Ta of 14–15°C, mice enter a deep hypometabolic state after ∼90 min. Mice in such a deep hypometabolic state usually consume <10% of the normal amount of oxygen and exhibit a Tb as low as 15–16°C. Animals can remain in this deep hypometabolic state for 4–9 h at a Ta of ∼15°C, until they spontaneously awake. Mice can safely undergo AIHM treatment repetitively if the procedures are spaced at least 24 h apart. Given the dramatic change in physiology between euthermia and the deep hypometabolic state, we hypothesized that molecular changes in gene expression levels must be significant between these vastly different metabolic states. Therefore in the present study, we undertook an investigation of the impact of AIHM on gene expression in liver, a key metabolic organ, using microarray technology. Microarrays containing ∼45,000 unique sequences were tested for differential expression. Our studies suggest that during deep hypometabolism, the expression levels of >99% of genes in the liver were not changed significantly from euthermia. Fewer than 40 genes were differentially expressed compared with controls during AIHM, while ∼200 genes are differentially expressed at the onset of arousal. In addition, our data suggest that AIHM interrupts the temporal progression of the circadian clock and the expression of clock control genes during deep hypometabolism. Thus, our studies reveal that during deep hypometabolism, key cellular activities such as changes in gene expression and circadian regulation are either very slow or halted.



Mice (C57BL/6) were purchased from Harlan Laboratories (Indianapolis, IN). Mice were housed in a standard animal husbandry facility under a 12-h/12-h light/dark cycle with Ta between 22 and 24°C. Female mice aged between 12–14 wk were used in this study. All mouse studies were carried out under institutionally approved animal research protocols HSC-AWC-07-102 (UTHealth).

AIHM procedure.

The AIHM procedure is as previously described (11). Briefly, each mouse was injected intraperitoneally (ip) with 0.5 mg/g body wt of freshly prepared 5′-AMP (Sigma catalog #A1752-25G) dissolved in phosphate-buffered saline. Within minutes, these mice enter a state of depressed metabolism with much reduced physical movement. Mice given 5′-AMP and maintained at a Ta of ∼15°C entered into deep hypometabolism after ∼90 min. Standard water and food supplies were provided ad libitum at all times. The heart and breathing rates were <1/10 of the normal euthermic rates. The reverse-flip behavior, when the animal regains the ability to right itself on its feet, signifies the early arousal stage and is used as an indicator for the arousal process in this study. After the animals exhibit reverse-flip behavior, they gradually rewarm and resume normal behavior within 1–2 h.

RNA sample collection.

Control livers for total RNA extraction were collected from three untreated mice at the equivalent zeitgeber time zero (ZT0). For untreated ZT groups, livers for total RNA extraction were collected from three untreated mice for each of the ZT6 and ZT18 groups, while four mice were used for the ZT12 group. For AIHM, arousal, 24 h, and 48 h groups, 0.5 mg per gram body wt 5′-AMP was administered at ZT0. For the AIHM group, livers were collected in the middle of AIHM (n = 4), ∼6 h post-AMP injection. Livers from the arousal group were collected when the mice exhibited reverse-flip behavior (n = 4), ∼12 h post-AMP administration. Livers for the 24 h (n = 4) and 48 h (n = 3) groups were harvested 24 h and 48 h following 5′-AMP administration, respectively. The liver tissues were immediately frozen in liquid nitrogen upon harvesting and the total RNA was extracted from each tissue sample using the guanidium/CsCl method (28).

Microarray analysis.

In brief, microarray analysis was performed using Illumina Sentrix Beadchip Array Mouse WG-6 v2.0 Beadchips containing 45,281 probe sequences that span the murine transcriptome. We amplified and purified 200 ng of total RNA using the Illumina TotalPrep RNA Amplification Kit (Ambion cat. #IL1791) following the manufacturer's instructions. The first-strand cDNA was synthesized by incubating RNA with T7 oligo(dT) primer and reverse transcriptase mix at 42°C for 2 h. RNase H and DNA polymerase master mix were immediately added to the reaction following reverse transcription, and samples were incubated for 2 h at 16°C to synthesize second-strand cDNA. In vitro transcription was performed and biotinylated cRNA was synthesized by a 16 h amplification with dNTP mix containing biotin-dUTP and T7 RNA polymerase. Amplified cRNA was subsequently purified and the concentration was measured using a NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies). An aliquot of 1.5 μg of amplified products was loaded onto Illumina Murine WG-6 Beadchips and hybridized at 58°C for 17 h, washed, and incubated with streptavidin-Cy3 to detect biotin-labeled cRNA on the arrays. Arrays were dried and scanned with a BeadArray Reader (Illumina). Data were analyzed using GenomeStudio v1.03 software (Illumina). Preprocessing utilized quantile normalization and background subtraction. Differential expression was determined at P < 0.01 with multiple testing correction using Benjamini-Hochberg to reduce the false discovery rate (FDR) (6, 55). Clustering and pathway analyses were performed with a Multiple Array Viewer v. 4.3.02 (36) and Ingenuity Pathfinder Analysis (IPA) v8.6. Raw and normalized data sets for all samples involved have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus (GEO) for repository under the accession number GSE23975.

Network analysis.

Associations of the differentially expressed genes with broadly defined molecular networks were carried out using IPA (Ingenuity Systems, tools. Molecular networks are predicted based on the direct and indirect relationships of the differentially expressed molecules to members of the networks in the IPA database. When using IPA, we processed the analyses of upregulated and downregulated genes for each time point group as separate groups. For all time point groups, 5′-AMP was administered at ZT0, and all experimental data were compared with data from untreated samples collected at ZT0 as control. Only values that were significant at P < 0.01 with correction for FDR and with at least a twofold change in mRNA level were included in the analysis.

Suprachiamastic nucleus immunofluorescence staining.

For the AIHM group, 5′-AMP was administered ip at ZT0, and arousals took place ∼9 h after 5′-AMP administration. Fresh brains obtained at each indicated time point were fixed with 4% PFA and embedded in paraffin. Coronal sections containing the suprachiamastic nucleus (SCN) were 8 μm in thickness. Antigen retrieval was performed in citrate buffer (pH 6.0). Specimens were blocked with 1% BSA/5% goat serum in PBS for 30 min. Specimens were then treated with anti-PER2 antibody (59) for 48 h at 4°C. Primary antibody was visualized using anti-rabbit IgG Alexa 555 antibody (Invitrogen).

Quantitative polymerase chain reaction.

The intron-spanning primer pairs used for quantitative polymerase chain reaction (qPCR) of all genes tested were selected from the primer bank (, Table 1). First-strand cDNA was synthesized using SuperScript III Reverse Transcriptase from Invitrogen (cat. no. 18080-044) with 2 μg of total RNA for each sample in a 40 μl volume. Each qPCR contained 1 μl of first-strand cDNA reaction mix and the respective primer set using the Maxima SYBR Green qPCR reaction kit (Fermentas Life Sciences, cat. no K0251). A Stratagene MX3000p was used for running 96 sample qPCRs simultaneously, and MxPro software was used for data acquisition, tracking, and preliminary analysis. Two quantization control references, L7 ribosomal protein and GAPDH, were included in every qPCR reaction. The relative amount of each transcript was determined using the ΔΔCt method (26). For the AIHM group, the average of the normalized AIHM samples was compared with the average of the ZT0 control group samples to calculate the differential expression fold value for each gene examined. The same data analysis steps were used for the data sets of the arousal group, 24 h group, and 48 h group.

View this table:
Table 1.

qPCR primers


Transcript profiling was carried out for liver samples from untreated control mice, mice during AIHM, and at the beginning of the recovery when the mice exhibited reverse-flip behavior: early arousal and 24 h and 48 h post-AMP injection.

Alteration in gene expression profile during the AIHM state.

Out of a total of 45,281 unique sequences examined, 33 unique transcripts were significantly (P < 0.01) upregulated by at least twofold (Table 2), including 31 annotated genes. A total of six genes were significantly downregulated by at least twofold, and five of these were annotated. The expression levels of >99% of the genes after 6 h of AIHM were similar to those of the untreated euthermic control at ZT0.

View this table:
Table 2.

Differentially expressed genes

Fifteen of the known upregulated genes are involved in a broadly defined IPA network of behavior, gene expression, nervous system development and function (Table 3 and Fig. 1A; Supplemental Figs. a–c).1 Thirteen of the upregulated genes could be associated with the broadly defined IPA network of cellular development, cellular growth and proliferation, and cell death (Fig. 1B; Supplemental Figs. d and e). However, many of the molecules in the broad network, including some of the molecules connecting multiple pathways, were not significantly upregulated. Thus, we reasoned that the impact of the differentially upregulated genes associated with AIHM in these various networks could be relatively moderate. Given that half of the network-eligible molecules (PER1, SOCS3, ATF3, ZFP36, GADD45G, PLK3, EGR1, CSRNP1, JUNB, FOSB, FOS, JUN, GADD45A, DUSP1, BHLHE40, ADRB2) are factors that are directly involved in regulation of gene expression, upregulation of these genes in the liver during the AIHM state may be linked to their role in switching on and off relevant target genes. Interestingly, only one upregulated gene is associated with the lipid metabolism pathway.

View this table:
Table 3.

IPA Networks impacted by differentially expressed genes

Fig. 1.

IPA networks that are affected by the differentially expressed genes. A and B: IPA networks affected by the upregulated genes in AMP-induced hypometabolism (AIHM) state. C–G: the IPA networks affected by the upregulated genes in the early arousal stage. H: the IPA network affected by the downregulated gene in the early arousal stage. I: the IPA network affected by the upregulated genes at 24 h. J: the IPA network affected by the downregulated genes at 24 h. K: the IPA network affected by the downregulated genes at 48 h.

Of the six downregulated genes, none is IPA network eligible, and their functions are largely unknown. Only one gene, OSGIN1 (OKL38), has been characterized in earlier studies. It is a p53 target gene with proapoptotic regulatory activities (53), though the effect of its downregulation in AIHM is currently unclear.

Differential gene expression in the arousal stage.

In the early arousal phase, a total of 159 unique transcripts were upregulated (Table 2), including 140 annotated genes and 19 genes of unknown function. The outcome of IPA is shown in Table 3 and Fig. 1, C–G. Although these genes seem to affect many functional pathways and IPA networks through direct and indirect interactions, it appears that their most critical function is the regulation of gene expression, as more than one-third of the 82 network eligible molecules (28 molecules including SOCS3, PER1, GADD45G, LGALS9, VEGFA, KLF9, JUN, GADD45A, THRAP3, ITGAV, BLM, TNFRSF1B, NFKBIB, ADRB2, IL10, TXNIP, EGR1, CSRNP1, CEBPB, NFKBIZ, MAFF, FGF1, FOS, DUSP1, BHLHE40, CDKN1A, SKIL, CXCL2) are known to be directly involved in regulation of gene expression. Since the liver is a vital metabolic organ, it is conceivable that the expression of the relevant genes is linked to restoration of the euthermic state during the arousal phase. Surprisingly, only three upregulated genes are associated with carbohydrate, lipid, or amino acid metabolic pathways. We concluded that most of the genes encoding the required metabolic regulators are unaltered by AIMH and they remain available during arousal to serve their normal function. Notably, 49 genes that were upregulated have no known described function.

Among the 51 transcripts that are downregulated (Table 2), 44 have been annotated. Five of the downregulated genes can be associated with a broadly defined network of infectious disease, infection mechanism, and inflammatory disease (Table 3). As shown in Fig. 1H, the vast majority of the molecules in the network, including most of the nodal molecules that bridge many pathways, show no detectable change. Therefore, the impact of these downregulated molecules on the network may be limited. Twenty-four of the downregulated genes have never been characterized or described in the literature.

AIHM's effect on circadian clock-regulated genes.

Previous studies have shown that the transcription levels of ∼10% of liver genes oscillate and follow a circadian rhythm (39). Therefore, it is important to examine whether the rhythmicity of the circadian clock is maintained when the animals were in AIHM. AIHM and arousal liver samples were collected at time points equivalent to ZT6 and ZT12, respectively. A comparison of transcript levels in liver tissue from untreated mice at ZT0, ZT6, ZT12, and ZT18 to their expression level in both AIHM and arousal phase reveals that cyclic genes observed in untreated mice have lost rhythmicity in the animals in AIHM and arousal states. In fact the expression levels of these cyclic genes most closely resemble their expression levels at ZT0, when 5′-AMP was originally administered. Figure 2 illustrates the data from 45 such genes, including many key circadian genes. Our analysis of these circadian genes is illustrated by heat maps. These particular genes were selected because their level of expression falls within the expression range of the few well-described circadian genes and thus can be displayed most clearly in a heat map. Other circadian-controlled genes whose expression was higher or lower than the genes illustrated behaved in a similar manner. These data suggest that AIHM treatment puts the circadian clock on hold in the liver.

Fig. 2.

Heat map of representative circadian oscillating genes. As control groups, the data in zeitgeber time (ZT) 0, ZT6, ZT12, and ZT18 groups are from untreated mice. The expression levels of AIHM and arousal groups closely resemble values for the ZT0 time point of the control group, although the sampling times for AIHM and arousal group are close to ZT6 and ZT12, respectively. The color scale reflects the absolute value of the expression level of each gene. *Expression levels were log2-transformed before analysis.

The liver data suggested to us that the peripheral circadian clock had been interrupted by the AIHM treatment. We next investigated whether the central circadian clock of the SCN was also stalled by AIHM. The basic clock mechanism is based on a transcription and translation feedback loop, and Per2 is one of the key regulators in the feedback loop (21). The cyclic expression of PER2 in the SCN is a widely used indicator for central circadian clock regulation. Therefore, to investigate whether central clock regulation in the SCN has been interrupted, PER2-specific immunohistochemical staining studies were carried out in SCN of mice treated with AIHM over a 24 h period at 4 h intervals (Fig. 3). Mice in the AIHM group were administered 5′-AMP at ZT0, and they became aroused ∼9 h after 5′-AMP administration. In the untreated controls the peak PER2 expression was between ZT8 and ZT12, consistent with previous studies (52). In contrast, PER2 expression in AIHM treated mice peaked much later, between ZT16 and ZT20, suggesting that the circadian clock resumed its function after the mice were aroused. Thus, similar to liver, the central clock in the SCN was also stalled during the deep hypometabolic state.

Fig. 3.

Immunofluorescence staining of PER2 in the suprachiamastic nuclei (SCN) of untreated mice (top lane) and AIHM-treated mice (bottom lane) with antibodies to mPER2. For the treated group, 5′-adenosine monophosphate (5′-AMP) was administered ip at ZT0, and arousals occurred about 9 h after AMP administration. Note, the ZT0 slide in the bottom panel shows the SCN at 24 h after 5′-AMP administration. The blue DAPI signals indicate nuclei.

Differential gene expression at 24 and 48 h posttreatment.

The AIHM state in mice at 14–15°C typically lasts 4–9 h before the animals begin to arouse. The mice usually resume normal behavior ∼2 h after initiation of arousal. To assess the recovery process, gene expression profiles at 24 h and 48 h after 5′-AMP ip injection were undertaken. At 24 h post-AMP injection, there were 8 upregulated and 25 downregulated unique genes compared with the control group (Tables 2 and 3; Fig. 1, I and J). Since AIHM stalls the circadian clock until recovery, the majority of these network eligible upregulated genes are in a network with other well-known circadian genes, such as Bmal1 (ARNTL), Per1, Per2, Clock, Cry1, and Cry2 (Table 3, Fig. 1I). These observations suggest that by 24 h post-AMP administration, the circadian clock has not fully reset to its normal rhythm.

By 48 h post-AMP injection, there were only five upregulated genes and three downregulated genes (Tables 2 and 3, Fig. 1K). None of the differentially expressed genes overlaps with the differentially expressed genes in the AIHM and aroused groups. Four of the eight differentially expressed genes are not annotated. Therefore by 48 h posttreatment, mice aroused from AIHM appear mostly recovered at the gene expression level. Consistent with our behavioral observations, the effects of AIHM treatment at the gene expression level in the liver appear to be short lived.

Genes differentially expressed in more than one state.

To assess the dynamics of the changes of transcriptional expression from AIHM state to 48 h posttreatment, a Venn gram analysis was carried out to reveal the genes that are upregulated or downregulated continuously from one state to the next (Fig. 4, Table 4). The majority of the differentially expressed genes in AIHM remain differentially expressed in a similar manner in the arousal state. In addition, there was a large increase in the number of the genes that are both differentially upregulated and downregulated in the arousal state.

Fig. 4.

The dynamics of the changes of transcriptional expression from AIHM stage to 48 h posttreatment stage. The Venn grams reveal the genes that are upregulated (A) or downregulated (B) continuously from 1 stage to the next. Each oval represent each of the 4 stages: AIHM in blue; arousal in yellow, 24 h in green, and 48 h in red. The numbers in the overlapping regions of the oval indicate the numbers of the genes expressed in the stages represented by the participating ovals.

View this table:
Table 4.

Genes that are differentially expressed in more than one stage

No genes that are differentially expressed in the AIHM state, and few in the arousal state, overlapped with genes differentially expressed at 24 h posttreatment. One transcript that was upregulated in both arousal and 24 h posttreatment was Slc7a2, but it was no longer upregulated by 48 h. Slc7a2 is also called the low-affinity cationic amino acid transporter-2, involved in regulating l-arginine availability (33). Two genes, ankyrin repeat and SOCS box-containing 2 (Asb2), T-cell receptor beta, variable 13 (Tcrb-V13), are both downregulated in both the arousal state and 24 h posttreatment state. There is limited information on the function of these genes, and Abs2 appears to target selected proteins for degradation. No genes that were differentially expressed in either the AIHM or arousal states are still differentially expressed at 48 h posttreatment, indicating expression of all AIHM-induced transcripts has returned to the basal level by 48 h.

At both 24 h and 48 h posttreatment, two downregulated genes, erythroid differentiation factor 1 (Erdr1) and insulin-like growth factor-binding protein 1 (Igfbp1), overlapped, but there was no overlap of any upregulated gene. There has been little description of Erdr1 in the available literature. Igfbp1 has been described as a prosurvival factor (25) that protects liver cells from apoptosis. Binding of Igfbp1 to insulin-like growth factors extends the life of the target proteins and modifies their interaction with cell surface receptors. Igfbp1 expression can be induced by stresses, such as exercise (4) and hypoxia (41).

qPCR verification of the microarray method.

To substantiate the microarray results, we carried out a qPCR analysis of several transcripts using the same total RNA that was used for the microarray studies (Fig. 5). The representative transcripts were selected with two criteria: transcripts that are changed in both AIHM and arousal states and transcripts expressed at levels within 2 orders of magnitude in range (100–10,000 relative fluorescent units). Two housekeeping genes commonly used as qPCR internal controls, GAPDH and ribosomal protein L7, were used simultaneously to calculate the fold change of each transcript. We found that the fold changes from qPCR for these transcripts closely resembled those obtained by microarray analysis, which validates the conclusions drawn from the microarray analysis in this study.

Fig. 5.

qPCR and microarray comparison: the correlation of array data with qPCR data using L7 as the reference is 0.99, and that of array data using GAPDH as the reference is 0.97 for the AIHM state (n = 4). The correlation of array data and qPCR data using L7 as the reference is 0.99, and that of array data with GAPDH as the reference is 0.98 for the arousal state (n = 4).


All mammals are endothermic with the ability to maintain their body temperature at varying environmental temperatures. Nonhibernating mammals maintain a relatively constant temperature of ∼37°C throughout their healthy adult life. Their basal level of metabolism is adjusted to maintain a homeothermic state in vastly different environments. Hibernating mammals have the physiological ability to enter into a hypometabolic state in response to seasonal and environmental cues, allowing their core body temperature to drop significantly toward the environmental temperature, and the physiological ability to exit hibernation and return to their homeothermal state. While the hibernation of many different species is well described, the core regulatory mechanism is still very much a subject of investigation. Although we have established an AIHM protocol that allows us to induce a deep hypothermal state in several nonhibernators, the molecular changes that accompany this induced hypometabolism remain largely unknown. Here we investigated the AIHM state at the gene expression level in the liver, a key metabolic organ.

We found using microarray technology that more genes are differentially expressed at arousal than in the AIHM state. The fact that 99.9% of transcripts analyzed did not change significantly during AIHM suggests that regulation of gene expression was suppressed when core body temperature was at ∼15°C. This observation is consistent with reduction of life processes to a minimum during AIHM, with only a minimum number of necessary genes upregulated to meet the change in physiological need. In contrast, in the recovery stage, the body is “resetting” gene expression patterns to assist the resumption of normal function by upregulating transcripts of necessary genes while downregulating others. Interestingly, more genes are upregulated (33 transcripts) than downregulated (6 transcripts) in the AIHM state (33 vs. 6), as well as in the arousal state (159 vs. 51). In addition, many differentially expressed transcripts have not been characterized: 5 of the 6 AIHM downregulated transcripts (83%), 49 of the 159 arousal upregulated transcripts (31%) and 24 of the 51 arousal downregulated transcripts (47%). This may be an indication that the AIHM induction and recovery processes involve physiological functions that remain poorly understood.

Six of the top 10 upregulated transcripts in the AIHM state are also among the top 10 upregulated genes in the arousal state. These are ATF3, FOS, FOSB, JUN, JUNB, and GDF15. In addition, EGR1, GADD45γ and DUSP1 transcripts are also upregulated in both the AIHM and the arousal states. A common characteristic of these strongly upregulated genes is that they are immediate early genes, that are transcribed in response to stimuli independent of protein synthesis. They are typically activated by a “cellular stress” signal and usually play a protective role. The stress induced ATF3 variant has been shown to prevent c-Jun N-terminal kinase (JNK)-induced neuron death (31) and affect cell fate by regulating alternative splicing and apoptosis (18). FOS, FOSB, JUN, JUNB are leucine-zipper proteins that can give rise to homo- and heterodimers to form the activation protein-1 (AP-1) transcription factor. AP-1 regulates many target genes and is implicated in regulating stress responses, as well as various cell growth and differentiation processes that depend on the specific cellular context (13, 48). GDF15, an immediate early gene associated with cellular stress (60), is also known as a factor that is expressed at high levels during ineffective erythropoiesis (43). Our recent study demonstrated that depressed oxygenation of erythrocytes is a crucial aspect of the mechanism for AIHM (11). Thus upregulation of GDF15 in response to AIHM would be consistent with our observation of reduced erythrocyte oxygenation during AIHM. EGR1 is among the most commonly described early stress-responsive genes that also function as a tumor suppressor (1, 23, 24). Gadd45γ is a member of the family of stressor sensors (17) that also have been described as tumor suppressors (20, 54). Interestingly, AP-1 and Gadd45γ have been described as nitric oxide-induced genes (19, 45). Historically, nitric oxide has both pathogenic and therapeutic applications (50). DUSP1 was initially characterized as a MAPK phosphatase that deactivates MAPK1 (15). Many stimuli are known to elicit the expression of immediate early genes through MAPK signaling pathways. Accumulated evidence suggests that DUSP1 activation helps to dampen the MAPK signaling cascade and balance the immediate early gene activation responses to stimuli (30). DUSP1 is also known to be a critical player for regulating the innate immune response and suppressing endotoxic shock (58). Upregulation of DUSP1 has also been suggested to be protective in transplant reperfusion (8).

Immediate early genes, such as c-fos, have also been used as biomarkers for active neurons in mapping neuronal functional activities, since active neurons tend to express immediate early genes (22). Another study showed c-fos, fosB, fosl2, junB, egr-1 were upregulated in various regions of the brain in sleep-deprived mice, while fosl2 was also upregulated in a different brain region in the recovery phase (44). Thus, immediate early genes are active in many physiological conditions that may require adjustment and adaptation.

A recent study demonstrated that both Arrdc4 and Txnip inhibit glucose uptake and lactate output in cultured skin fibroblasts (34). These two transcripts may play similar roles in other tissues and regulate glucose and lactate levels. We have observed an increased level of glucose and reduced level of lactate in serum of mice in both AIHM and arousal states (11), consistent with the observation that these genes are up regulated in both states. (TXNIP is upregulated 1.9-fold in the AIHM state, thus was not included in Table 2). In the liver, lactate can be salvaged to form pyruvate, a central metabolite that feeds into many key metabolic processes including the Kreb's cycle.

That the “gene expression” function was enhanced during both AIHM and arousal phases as revealed by IPA is unmistakable, since gene expression patterns were changed in both stages. However, other subnetworks linked to the gene expression network in the networks broadly defined by IPA were less compelling to us. The IPA database is largely based on keyword search of publications for indirect links between molecules and functions. Thus, it reveals many relationships that are suggestive and may provide clues for possible novel interactions. Users are cautioned against using IPA to define molecular interactions. On the other hand, we have not been able to detect any physiological or behavioral impairment in mice caused by regular AIHM treatments, even when the animals are treated every other day for 4 wk (data not shown). We recognize that under this relatively transient deep hypometabolic state, life processes at the organ, cellular, and molecular level could have been challenged in a way that is unparalleled by other physiological processes commonly encountered and described. Therefore, we present these data without trying to overinterpret them.

Initially data were analyzed in pairs such as AIHM vs. ZT6 and arousal vs. ZT12. Most of the differentially expressed transcripts were from circadian oscillating genes, similar to ZT0 vs. ZT12 and ZT0 vs. ZT6, respectively. When the expressions of hundreds of circadian oscillating genes from AIHM, arousal, and the four ZT control groups were plotted in one heat map, it became clear that the expression levels of the circadian oscillating genes in the AIHM and arousal groups are not significantly different from their expression levels at ZT0 when the 5′-AMP was initially administered but were different from other ZT groups. Therefore, the ZT0 group was chosen as the reference. The present study suggests that clock-controlled genes no longer display temporal control in the AIHM state. Given the severe metabolic suppression, it is not surprising that many normal physiological and cellular processes are suspended temporarily. In addition, the gene expression profile of the majority of nonclock control genes also displayed minimal changes from the untreated control. Interestingly, it has been reported that the circadian clock stopped ticking during hibernation in the hamster (35). Thus, our finding may be a general consequence of extremely low metabolic activity on gene expression. The majority of the transcripts that were differentially regulated in AIHM and arousal stages are no longer differentially regulated after 24 h. By 48 h, these circadian genes are mostly back to their normal rhythm. These observations indicate that the effects of AIHM on gene expression are relatively short term.

One well-known clock gene, Per1, seems to behave differently from most other circadian genes: it is upregulated in both AIHM and arousal states. While one can argue that the circadian clock is not completely stalled, an alternative explanation is that Per1 is capable of behaving similar to immediate early genes that are activated almost immediately in response to certain signals. Earlier studies have demonstrated that Per1 behaved like an immediate early gene, such as c-Fos, in response to light stimulation (2, 5). AIHM treatment could have similarly induced an immediate early response from Per1.

A recent study based on new analyses of previously published data proposed that torpor-arousal (TA) behavior could be controlled by a TA clock for animals under prolonged torpor and hibernation (27). The author suggested that the TA clock could be a peripheral circadian clock, distinct from the SCN circadian clock. They seem to assume that the period of the central circadian clock in the SCN remains ∼24 h during hibernation, while the proposed TA clock that controls the TA cycle is temperature dependent and possibly regulated by an epigenetic mechanism. Our experiments demonstrate that the post-AIHM arousal behavior is based on a temperature-sensitive mechanism. We observed that both the circadian clock in the SCN and peripheral (liver) tissue are drastically slowed during the AIHM. This finding is similar to the previously reported finding that the SCN of hibernating European hamster ceased to output rhythmic signals during hibernation (46). Studies by van Breukelen and Martin (47) have also shown that both transcription and translation activities are practically ceased during deep torpor. Our findings of a stalled circadian clock in the deep hypometabolic state of AIHM are consistent with these previous observations.

Our characterization of AIHM was recently published (11). Recent independent studies have also shown that administration of 5′-AMP to mice and rats resulted in the lowering of the body temperature. Swoap and colleagues (42) suggested that AIHM is mediated through adenosine receptor signaling pathways and is thus different from torpor. They proposed that AIHM is mediated through a bradycardia effect of adenosine receptor pathways in the heart. However, our studies showed that genetic deficiencies in adenosine receptors (A1, A2A, A2B, and A3) did not affect the ability of mice to undergo AIHM. Furthermore, mice deficient in Ecto-nucleotidase (CD73) required much less 5′-AMP for AIHM. Ecto-nucleotidase (CD73) is the extracellular enzyme that converts 5′-AMP into adenosine. It has been previously shown that adenosine cannot induce bradycardia in A1 receptor-deficient mice (10). These findings suggest that adenosine signaling does not play a major role in AIHM. Similar to independent findings (56), our studies also showed that mice in AIHM have slower heart rates, hypotension, much higher blood glucose (hyperglycemia) levels, and low blood lactate levels. It should be noted that hamsters undergoing natural daily torpor also display increase levels of blood glucose in the initial phase, suggesting that daily torpor unlike hibernation preferentially uses carbohydrate as an energy source (16). The decrease in blood pressure (hypotension) reflects vasodilation of the peripheral vessels. This is a physiological response that enhanced the loss of body heat.

The complex and well-coordinated seasonal molecular reprogramming that enables hibernators to transition safely between the summer euthermic state to the winter hibernating state has been studied using transcriptional profiling (9, 29, 40, 49, 51), proteomics (14, 38), and metabolomics (32). Although molecular evidence consistent with well-established observations, such as energy sources and metabolic rates switch, is shown from these different approaches, each approach also contributes distinctive details. We observed that 99.9% of transcripts analyzed did not change significantly during AIHM. This suggests that regulation of the physiological changes observed during AIHM may occur at the posttranslational level.

It appears that the AIHM gene expression profile bears limited resemblance to that of hibernation, where genes involved in metabolism are differentially regulated (9, 40, 49). Knowing the reprogramming for hibernation is a gradual seasonal process, and AIHM is a pharmacologically induced acute hypometabolic state, the lack of parallel at the transcriptional level is not surprising. However, a recent study of torpor behavior in ground squirrels showed that in the liver, c-Jun and c-Fos transcripts were strongly upregulated in the early arousal phase from torpor (51). This overlap to our findings suggests that global gene expression regulators, such as AP-1, can be important players in protection and recovery processes of animals recovering from a hypometabolic state. It appears that pharmacologically induced hypometabolism has some similarities to natural torpor but differs in other aspects (3).

In summary, the current study demonstrates that the process of AIHM changes the expression level of a small number of transcripts, some of which are immediate early genes. The expression levels of additional genes, though still a small population, are changed in the early arousal state. Within 2 days after AIHM treatment, the gene expression pattern is almost the same as that of the pretreatment state. Thus, AIHM's impact on gene expression seems to be temporary and is, by and large, a reversible, induced physiological and cellular process.


This work is supported by National Institutes of Health Director Pioneer award NIH/NDPA 5 DP1 OD000895-03 (to C. C. Lee).


No conflicts of interest, financial or otherwise, are declared by the author(s).


The microarray assay was carried out at the microarray core laboratory under Dr. David S. Loose at UTHealth, Medical School. We thank Dr. J. Lever for helpful comments and reading of the manuscript.

Author contributions: Z. Zhao was responsible for experimental design, animal handling, sample collection, qPCR primer selection, and data analysis; A. Van Oort-Jansen was responsible for total RNA extraction and characterization; T. Miki was responsible for the SCN immunofluorescence analysis; D. S. Loose was in charge of microarray analysis; T. Matsumoto was responsible for performing qPCR; C. C. Lee directed the studies; Z. Zhao and C. C. Lee wrote the manuscript.


  • 1 The online version of this article contains supplemental material.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
View Abstract