Pennington Biomedical Research Center, 6400 Perkins Road, Baton Rouge, LA 70808. E-mail: firstname.lastname@example.org
Objective: The Agouti-related protein (AgRP), neuropeptide Y (NPY), proopiomelanocortin (POMC), cocaine and amphetamine-regulated transcript (CART), Orexin, melanin concentrating hormone (MCH), leptin, and its hypothalamic receptor (LR) are key regulators of food intake and energy homeostasis. In the present study, we examined the circadian expression profiles of these genes.
Research Methods and Procedures: We used quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) to measure mRNA levels, spectral analysis to evaluate periodicity, and correlation analysis to examine for associations with diurnal food intake.
Results: No gene in particular stood out as a strong candidate, but the overall circadian expression profiles of leptin and its hypothalamic receptor came close to statistically and graphically resembling the diurnal feeding behavior of mice. In mathematical terms, adrenal AgRP exhibited strong circadian expression and had the highest correlation with food intake, followed by leptin. Yet its highest point of expression occurred 8 hours after nocturnal food intake had peaked, suggesting that adrenal AgRP could not play a direct role in the initiation of nocturnal feeding; neither did hypothalamic AgRP, NPY, POMC, CART, Orexin, or MCH.
Discussion: These data show that ad libitum feeding in mice is influenced by complex central and peripheral circuits involving orexigenic and anorectic agents of which leptin and its hypothalamic receptor could play more prevalent roles.
The hypothalamus–pituitary–adrenal axis plays a key role in the regulation of food intake and energy homeostasis (1, 2). Agouti-related protein AgRP, 1neuropeptide Y (NPY), proopiomelanocortin (POMC), and cocaine- and amphetamine-regulated transcript (CART), Orexin, and melanin concentrating hormone (MCH) play essential roles in the regulation of food intake and energy balance (3). AgRP is expressed in the arcuate nucleus of the hypothalamus, the adrenal gland, and the testis (4, 5). The hormone leptin, which is secreted from fat in correlation to body fat mass, down-regulates the expression of AgRP and NPY (6) and increases that of POMC (7), thereby favoring melanocortin activation through α-melanocyte stimulating hormone. Leptin's action is mediated by binding to the long isoform of the leptin receptor (LR) present in the hypothalamus and activating signal cascades of different pathways like the AMP-kinase pathway (8), the JAK-STAT pathway (9), and the PI3K pathway (10). Orexin has been implicated in the regulation of sleep (11) and is likely involved in the coordination of feeding behavior, arousal, and emotion (12). The melanin-concentrating hormone (MCH) is also a significant regulator of food intake and a mediator of the leptin-deficient phenotype (12, 13). In the rat lateral hypothalamus, Orexin, MCH, and CART exhibit distinct expression patterns (14), but their diurnal expression during ad libitum feeding conditions in the mouse has not been investigated.
Alterations of the normal day/light cycle are one way to disrupt the regulatory balance via changes in diurnal gene expression, which may, in turn, alter food intake. Such a shifted circadian pattern as observed in night workers and night eaters, or due to depression and jet lag, can result in changes in metabolism and response to stress and can even result in chronic diseases and cancer (15, 16). The duration of a circadian rhythm normally is 24 hours. The phase of the central circadian pacemaker in the suprachiasmatic nucleus of the hypothalamus can be adjusted by external stimuli, such as the daily light cycle, and by internal stimuli reflecting the physiological and behavioral status of the organism, summarized as non-photic stimuli (17). Transcription factors such as Clock, Bmal, Per, and Cry are responsible for cellular events that can be organized in an oscillating daily loop (18). Defects in one of these key oscillators can have profound effects on daily food intake, as seen in Clock mutant mice, which have an attenuated diurnal feeding rhythm and show increased energy intake and body weight compared with wild-type mice (19).
Despite the importance of the suprachiasmatic nucleus in sensing environmental daily light changes, there exist also peripheral circadian clocks, building a complex network with cross-talk between them (20). A previous study measured AgRP in the hypothalamus of rats via in situ hybridization and revealed the existence of a circadian rhythm, which was similar to that of food intake (21). Depletion of corticosterone by adrenalectomy abolished this AgRP diurnal rhythm, which was restored by exogenous corticosterone replacement, highlighting the requirement of adrenal-secreted hormones to maintain the normal diurnal AgRP expression cycle (21).
The goal of the present study was to investigate, using the sensitive and quantitative real-time polymerase chain reaction (PCR) method, whether a circadian expression exists for the AgRP, NPY, POMC, CART, Orexin, MCH, leptin, and LR genes. Secondly, AgRP expression in two peripheral tissues (adrenal and testis) was investigated to examine whether this important anabolic regulator of food intake that infiltrates peripheral organs (22) exhibits circadian expression in the periphery. Using periodograms, we analyzed the expression profile for each gene and compared it with diurnal food intake to determine whether there was a correlation. This revealed that adrenal-expressed AgRP exhibited the most prominent circadian expression profile and also showed (along with leptin) the highest correlation with diurnal food intake. However, it was only the expression profiles of leptin and its hypothalamic receptor that exhibited the closest fit to diurnal food intake of mice.
Research Methods and Procedures
Seven-week-old male C57BL/6J (C57) mice (5 mice per group) were used for the study. Mice were fed regular chow diet (Labdiet 5001; Purina Mills, St. Louis, MO) after weaning. The mice were kept on a 12/12 hr dark/light cycle (lights on at 6:00 am and off at 6:00 pm). Mice were bred at the Pennington Biomedical Research Center and sacrificed at the indicated time-points by cervical translocation. The study protocol was approved by the Pennington Biomedical Research Center Institutional Animal Care and Use Committee. Adrenal glands, hypothalamus, and the right testis were removed over a period of 2 days to obtain samples of a 24-hour circadian pattern every 4 hours. With the exception of brains, excised tissues were immediately snap-frozen in liquid nitrogen and stored at −80 °C until RNA isolation. Whole brains were placed on dry ice for 10 minutes and then thawed for 2 minutes for the tissue to soften slightly, and hypothalami were excised with a razor blade by trimming out all of the brain structures including the midbrain, the cerebellum, and the cortex up to the corpus callosum. Data shown represent the mean ± standard error for 5 mice per time-point for the gene expression and the mean ± standard error for 12 mice per time-point for the food intake. Food intake was measured over a 48-hour period in metabolic chambers (Columbus Instruments, Columbus, OH), with each mouse being housed individually, and data are shown for every hour.
Real-time Reverse Transcriptase (RT)-PCR
Tissue was homogenized in TRIzol (Invitrogen, Carlsbad, CA) and then sheared through a 21-G needle, and total RNA was isolated using the commercial RNeasy kit (QIAGEN, Valencia, CA) according to the manufacturer's instructions. The integrity of the RNA was verified by EtBR staining of the 28S and 18S bands on a 1% gel. Quantitative RT-PCR was performed using the TaqMan one-step RT-PCR core reagents kit (Applied Biosystems) in MicroAmp Optic 384-well Reaction Plates (Applied Biosystems) on an ABI PRISM 7700 Sequence Detection system (Applied Biosystems) under standard conditions (30 minutes at 48 °C, 10 minutes at 95 °C, then 40 cycles of 15 seconds at 95 °C and 1 minute at 60 °C); 100 ng of total RNA for each sample was used in duplicate and each run included a standard curve with 5 serial dilutions and a non-template control. The level of gene expression for each gene was quantified relative to the level of the housekeeping gene cyclophilin using the standard curve method and is presented as relative expression units. These primers and probes were used to amplify the following mouse genes: AgRP forward: 5′-CTTTGGCGGAGGTGCTAGAT-3′, reverse: 5′-AGGACTCGTGCAGCCTTACAC-3′, and probe: 5′-6-FAM-CGAGTCTCGTTCTCCGCGTCGC-3′; NPY forward: 5′-CTCCGCTCTGCGACACTACA-3′, reverse: 5′-AATCAGTGTCTCAGGGCTGGA-3′, and probe: 5′-6FAM-CAATCTCATCACCAGACAGAGATATGGCAAGATBHQ1–3′; leptin forward: 5′-ATTTCACACACGCAGTCGGTAT-3′, reverse: 5′-AAGCCCAGGAATGAAGTCCA-3′, and probe: 5′-6-FAM-GCCAGTGACCCTCTGCTTGGCG-BHQ-1–3′; LR forward: 5′-TGTTCCTGGGCACAAGGACT-3′, reverse: 5′-TGATTCTGCGTGCTTGGTAAA-3′, and probe: 5′-6-FAM-AATTTCCAAAAGCCTGAAACATTTGAGCATCTTA-BHQ-1–3′; POMC forward: 5′-CAGTGCCAGGACCTCACCA-3′, reverse: 5′-AGCGAGAGGTCGAGTTTGCA-3′, and probe: 5′-6-FAM-AGAGCAACCTGCTGGCTTGCATCC-BHQ-1–3′; CART forward: 5′-AAGTCCCCATGTGTGACGCT −3′, reverse: 5′-GACAGTCACACAGCTTCCCGA-3′, and probe: 5′-6-FAM-CCCTTTCCTCACTGCGCACTGCTCT-BHQ-1–3′, Orexin forward: 5′-ATGAACTTTCCTTCTACAAAGGTTCC-3′, reverse: 5′-GGCAGCAGTAGCAGCAGCA-3′, and probe: 5′-6-FAM-CTGGGCCGCCGTGACGCT-BHQ-1–3, MCH forward: 5′-TTCAAAGAACACAGGCTCCAAA-3′, reverse: 5′-ACTCAGCATTCTGAACTCCATTCTC-3′, and probe: 5′-6-FAM-CTTACCTCGCTCTGAAAGGATCCGTAGCCBHQ-1–3, and Cyclophilin forward: 5′-TAGAGGGCATGGATGTGGTAC-3′, reverse: 5′-GCCGGAGTCGACAAGATG-3′, and probe: 5′-6-FAM-AGCCGGGACAAGCCACTGAAGGAT-BHQ-1–3′.
The Excel built-in tool for Student's t test and one-way ANOVA were used to perform pairwise and multiple comparisons.
For the spectral analysis, each mouse per time-point was considered to represent a separate day (e.g., 5 mice representing 5 days), although all 5 mice per time-point were killed at once to minimize variation due to climatic changes or aging or feeding behavior on a given day. Time-points represent the time of the day, which is a particular phase in circadian rhythm, but not necessarily a continuous calendar stretch.
Consider a series of expression levels for gene x with N samples of the form
This series can be converted from a time-domain, where each variable represents a measurement in time to a frequency domain using the Discrete Fourier Transform algorithm. Frequency domain representation of the series of experiments is also known as periodogram, which can be denoted by I(ω):
If a time series has a significant sinusoidal component with frequency ω ε [0, π], then the periodogram exhibits a peak at that frequency with a high probability. Conversely, if the time series is a purely random process (i.e., “white noise”), then the plot of the periodogram against the Fourier frequencies approaches a straight line (23).
Significance of the observed periodicity can be estimated by Fisher g-statistics (24), as recently recommended (25):
where I(ω) is a k-th peak of the periodogram. Large values of g indicate a non-random periodicity. To calculate the p value of the test under the null hypothesis, the exact distribution of g was used which is given by
where n = [N/2] and p is the largest integer <1/x.
This algorithm closely follows the guidelines recommended for analysis of periodicities in time-series microarray data (25), with the exception that we applied locally developed C++ code instead of R scripts.
Auto- and Cross-correlation
We also conducted auto-correlation analysis for a few selected genes to confirm the significance of the observed periodicity. For a given discrete time series Yt the auto-correlation is simply the correlation of the expression profile against itself with a time shift. If Yt is second order stationary with mean Ȳ, then this definition is
where E is the expected value and k is the time shift being considered (also referred to as the lag). This function has the attractive property of being in the range [−1,1], with 1 indicating perfect correlation (the signals exactly overlap when time shifts by k) and −1 indicating perfect anti-correlation.
Cross-correlation is very similar to auto-correlation except that here we correlated each time series expression profile x not to itself, but to a different profile y.
The phase shift k in this case can be used to identify the lag between two oscillating genes. Determination of the lag with this approach uses information of all time-points and, thus, is more reliable, although it may not coincide with the lag between individual peaks of the expression profiles.
To determine if genes involved in the regulation of food intake follow a circadian pattern under ad libitum feeding conditions, gene expression was analyzed for 24 hours, at 4-hour intervals (a total of 6 time-points). Various statistical approaches were used to examine the data from different angles. ANOVA and post hoc pairwise analysis were used to examine for differences in expression levels between time-points, while spectral analysis was used to examine for periodicity (i.e., circadian expression) and to determine the correlation between circadian expression and diurnal food intake. An important aspect of the latter approach is that it examines gene expression at various time-points as a continuous series of data rather than comparing between individual pairs of time-points.
Figure 1 shows the expression levels of all genes derived from real-time PCR analysis (solid line, left axis) along with the diurnal food intake data (dotted line, right axis). Table 1 shows a summary of the p values for the expression of each gene as determined by one-way ANOVA analysis, while Figure 1 shows the corresponding post hoc pairwise analysis (lowercase letters on solid lines). Adrenal AgRP and the hypothalamic LR showed highly significant variation (p < 0.01) in their expression profiles during the 24-hour period. The 2-pm time-point for adrenal AgRP represented its expression nadir and was statistically different from any other time-point. Hypothalamic LR showed its lowest expression at 6 pm, which was also significantly different from all other time-points. Hypothalamic and testicular AgRP, as well as hypothalamic NPY, also showed significant variation (p < 0.05), while anorectic genes, including leptin, showed no significant variations in their expression profiles.
Table 1. p Values (by ANOVA) determining the significance of the overall variation in gene expression between the six time points of the 24-hour cycle
Post hoc pairwise analysis is shown with the lowercase letters in the individual diagrams of Figure 1.
The food intake curve showed that mice ate the maximum of their food immediately after the lights went out (6 pm) and gradually decreased food consumption over the dark period. There were negligible amounts of food eaten during the light cycle, but food consumption began increasing 3 hours before lights out, perhaps in anticipation of the lights-out period. Spectral analysis showed that food consumption showed almost perfect consistency and explicit circadian periodicity (Table 2).
Table 2. Analysis of the circadian rhythm of gene expression and food intake
Spectral analysis was used to determine the periodicity of gene expression and, thus, evaluate circadian rhythm. Adrenal AgRP not only followed the strongest circadian pattern (Table 2) but also showed the highest correlation with food intake, at phase 5, of all genes studied (Table 3). Hypothalamic NPY circadian pattern of expression ranked second in terms of its p value (Table 2) but showed weak correlation with food intake (Table 3). Leptin did not exhibit a particularly strong circadian pattern (Table 2) but had the second strongest (after adrenal AgRP) correlation with food intake, also at phase 5 (Table 3). Yet leptin and LR circadian expression profiles matched well with the diurnal feeding of mice and their expected properties on food intake (i.e., the down-regulation of food intake). AgRP in the testis also exhibited circadian oscillation (Table 2) and the third highest correlation with food intake (Table 3). Hypothalamic AgRP, POMC, CART, Orexin, and MCH did not exhibit particularly strong circadian expression profiles (Figure 1; Table 2), nor did they correlate significantly with food intake (Table 2).
Table 3. Cross-correlation (R value) of circadian food intake with the expression profile of all genes
The number corresponding to the phase shift (0 to 5) that had the maximum correlation between gene expression and food intake is shown in parentheses.
Significant at p < 0.05 (the threshold value for R with 30 df is 0.3494)
Although these values were statistically significant (p < 0.05), they were not considered to be biologically important due to the low number (5) of periodic phases. A stringent threshold was applied to consider only R values higher than 0.5.
Food intake in laboratory rodents normally occurs during the dark cycle. A complex network of genes encoding orexigenic and anorectic neuropeptides, along with hormonal stimuli, are responsible for the initiation, duration, termination, and the frequency of meals. The present study set out to investigate whether these genes have circadian expression profiles under ad libitum food and water conditions and if their expression profiles in the hypothalamus are directly correlated to food intake. In addition, the circadian expression of AgRP was investigated in the adrenal gland and the testis to examine if this important regulator of energy balance exhibits circadian expression in peripheral tissues, given that it can infiltrate various organs and adipose and muscle tissues (22). Indeed, microarrays and spectral analyses have shown that up to 50% of genes follow daily oscillatory patterns in the liver while similar expression profiling is evident by genes in adipose tissue (26).
Spectral analysis, which examines for periodicity in an expression pattern, was used again and showed that adrenal AgRP had the strongest trend toward a circadian expression pattern compared with all other genes, including testicular and hypothalamic AgRP. Hypothalamic NPY also showed a robust circadian profile, but the hypothalamic LR, which had shown significant differences in its expression between time-points, did not adhere well to a 24-hour circadian profile. The results in Table 1 simply compare the differences of mRNA levels between time-points but do not analyze the issue of periodicity, which is addressed by the spectral analysis of circadian expression by comparing periodicity cycles (Table 2). Nonetheless, ANOVA and pairwise post hoc analysis showed that adrenal AgRP expression also exhibited the most significant changes during a 24-hour feeding cycle, further strengthening the results obtained from the spectral analysis.
A previous study in rats using the same light:dark cycles (lights out at 6 pm and lights on at 6 am) and in situ hybridization to measure AgRP mRNA, showed that AgRP peaked at 10 pm (21). In the same study, food intake showed two peaks: one at 10 pm, which coincided with AgRP mRNA, and a higher peak at 6 am, which was opposite to AgRP mRNA (21). There are distinct differences in the results between the rat study and our study, but the two studies not only used different species but also used different methodologies to evaluate gene expression levels and correlations with ad libitum feeding behavior.
Another study in rats using the same method that we used (i.e., real-time RT-PCR) but different light:dark cycles (lights out at 7 pm and lights on at 5 am) measured diurnal NPY, POMC, leptin, and LR (27). Again, there were no similarities with our data for any of these genes, but neither were there any similarities for POMC between the two studies performed in rats (21, 27). The differences in the data between our study and the one by Xu and colleagues could be attributed again to the differences in species used and the duration of the light:dark cycles. The study by Xu et al. (27) also applied food restriction to elegantly show that the diurnal expression profile of all of the genes they tested changed significantly despite the fact that the light:dark cycles remained the same (27).
The existence of circadian gene expression in the adrenal gland is well documented. First, c-fos expression in the rat adrenal was found to be constant at 8 pm, 12 am, and 4 am, but decreased to 52% at 8 am and to 18% at 12 pm, starting to rise again at 4 pm. This pattern closely followed ACTH secretion from the pituitary (28). Furthermore, constant light or constant dark (20 days) resulted in differential changes in rat adrenal NPY expression (29). AgRP expression is regulated in the adrenal gland by fasting in a similar fashion to that in the hypothalamus (30). Additionally, peripheral AgRP infiltrates various peripheral organs, including the adrenal gland and adipose tissue, in a fasting-dependent manner (22), that raises the possibility for autocrine/paracrine actions for AgRP (31).
The expression profile of leptin mRNA fits well with its expected property to down-regulate food intake. Indeed, the lowest mRNA levels for leptin were at 6 pm, which was also the point of the highest food consumption, while its highest mRNA levels coincided with the start of decline of food consumption. The leptin data fit well with those by Saladin et al. (32), who showed enhanced leptin expression 4 hours after the start of the dark phase (in our case, 10 pm), although they reported that serum leptin protein levels lagged behind its message (32). Food access restriction to the 12-hour light period reversed leptin rhythm, whereas a 5-hour sleep deprivation in the beginning of the light cycle failed to change leptin concentrations (33).
Even more than leptin, the circadian expression of its hypothalamic receptor showed significant variation between time-points and had its lowest mRNA levels at the point of the highest food intake (6 pm) and high mRNA levels at the lowest time-point of food intake (10 am). This expression profile again fits well with the potential role for the LR in the down-regulation of food intake. However, its circadian expression did not show significant periodicity, and it did not correlate with food intake as well as adrenal AgRP and leptin did. Yet, its expression profile befitted well its biological properties; and since this also coincided with the expression profile of leptin, we consider this gene (along with leptin) as a significant player in the regulation of diurnal food intake.
To the naked eye, the pattern of diurnal food intake was also closely associated with the expected suppression of food intake by POMC (7), but POMC did not exhibit a significant oscillatory pattern, nor did it correlate well with food intake. We are, therefore, reluctant to consider it as a key player in the regulation of ad libitum feeding behavior. CART expression resembled the expression profile of POMC, but, again, there was no significant oscillation. Others have shown that CART follows a circadian pattern both in blood and in different areas of the brain of rats, and that the circadian pattern was affected by fasting (34). The same has been shown for the two appetite effectors Orexin and MCH in rats (14), but we did not find statistically significant correlations with food intake or periodic expression for these two genes. Hypothalamic NPY showed the second highest circadian pattern after adrenal AgRP but did not correlate well with food intake. Circadian NPY gene expression has been studied extensively (35), revealing that NPY mRNA is increased in the basal hypothalamus of rats immediately preceding onset of the dark phase. In our study, hypothalamic NPY mRNA did not show a strong circadian profile, nor did it correlate well with food intake (similar to hypothalamic AgRP); we did not consider it as a key regulator for ad libitum feeding behavior.
In all, no gene in particular distinguished itself as a strong candidate for regulating diurnal food intake in mice. Only the circadian expression of leptin and its receptor came close to statistically and graphically resembling murine ad libitum food intake, also conforming to the biological properties of these genes. In mathematical terms, adrenal AgRP showed the best circadian expression profile and had the strongest correlation with food intake. However, its highest point of expression occurred 8 hours after food intake had peaked, suggesting that adrenal AgRP is not likely to play a direct role in the initiation of diurnal food intake, and neither is hypothalamic AgRP, NPY, POMC, CART, Orexin, or MCH. We conclude that ad libitum feeding in mice is influenced by a complex interplay of orexigenic and anorectic agents, of which leptin and its receptor could play more fundamental roles.
The authors thank Tammy Fairburn for excellent technical assistance. Supported by NIH Grant DK62156 (to G.A.) and by the Polish-U.S. Fulbright Commission, Warsaw, Poland (Grant PPLS/04/14 to J.S.).
The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.