Insulinoma-associated protein (IA)-2 and IA-2β are transmembrane proteins involved in neurotransmitter secretion. Mice with targeted disruption of both IA-2 and IA-2β (double-knockout, or DKO mice) have numerous endocrine and physiological disruptions, including disruption of circadian and diurnal rhythms. In the present study, we have assessed the impact of disruption of IA-2 and IA-2β on molecular rhythms in the brain and peripheral oscillators. We used in situ hybridization to assess molecular rhythms in the hypothalamic suprachiasmatic nuclei (SCN) of wild-type (WT) and DKO mice. The results indicate significant disruption of molecular rhythmicity in the SCN, which serves as the central pacemaker regulating circadian behavior. We also used quantitative PCR to assess gene expression rhythms in peripheral tissues of DKO, single-knockout, and WT mice. The results indicate significant attenuation of gene expression rhythms in several peripheral tissues of DKO mice but not in either single knockout. To distinguish whether this reduction in rhythmicity reflects defective oscillatory function in peripheral tissues or lack of entrainment of peripheral tissues, animals were injected with dexamethasone daily for 15 days, and then molecular rhythms were assessed throughout the day after discontinuation of injections. Dexamethasone injections improved gene expression rhythms in liver and heart of DKO mice. These results are consistent with the hypothesis that peripheral tissues of DKO mice have a functioning circadian clockwork, but rhythmicity is greatly reduced in the absence of robust, rhythmic physiological signals originating from the SCN. Thus, IA-2 and IA-2β play an important role in the regulation of circadian rhythms, likely through their participation in neurochemical communication among SCN neurons.
- suprachiasmatic nucleus
- vasoactive intestinal peptide
insulinoma-associated protein 2 (IA-2; also called islet antigen 2, ICA 512, and Prprn) and IA-2β (also called phogrin and Ptprn2) are transmembrane proteins with homology to protein tyrosine phosphatase but which lack phosphatase activity (29, 35, 36; for review, see Ref. 51). IA-2 and IA-2β are components of dense core vesicle membranes and play important roles in vesicle loading and release (11, 14, 17, 29, 44, 49). Targeted disruption of these genes in mice leads to impaired secretion of hormones and neurotransmitters, which are more severe in mice lacking functional alleles of both genes (16, 22, 24–26, 42, 48). These proteins are also of interest at the clinical level, since the majority of newly diagnosed diabetic patients have antibodies against these proteins, and antibodies appear years before disease development, identifying them as major autoantigens predictive of type 1 diabetes (30, 36, 49).
A recent study to assess the cardiovascular impact of targeted disruption of IA-2 and IA-2β revealed that mice lacking these proteins did not have daily rhythms in blood pressure, heart rate, core body temperature, or spontaneous locomotor activity (21). Electrophysiological studies on the suprachiasmatic nuclei (SCN), the site of the brain's master circadian oscillator, showed that disruption of both IA-2 and IA-2β results in significant alterations in neuronal firing rhythms in the SCN (21). Thus, disruption of IA-2 and IA-2β has a profound effect on the generation and expression of circadian rhythms.
One hypothesis to explain the disruption of electrophysiological, physiological, and behavioral rhythms in double-knockout (DKO) mice lacking IA-2 and IA-2β is that these proteins are critical for neurochemical interactions that couple SCN neurons to each other (21), as occurs in mice lacking vasoactive intestinal peptide (VIP) or the VPAC2 receptor (3, 5, 7, 8, 13, 15, 40; for review, see Ref. 52). Another plausible hypothesis is that the molecular oscillator within the SCN functions properly, but the ability of the SCN to express this rhythmicity as a rhythm in firing rate, and thus to convey it to downstream targets, is defective (27, 46). In this latter case, neurons within the SCN would be expected to retain molecular rhythmicity in DKO mice, while in the former, arrhythmicity would be expected.
Molecular and physiological rhythms exist in many organs and tissues, in addition to the SCN. Circadian rhythms are regulated by a transcriptional-translational feedback loop. In the SCN and in other tissues, the products of “clock genes” participate in a feedback loop with circa-24-h cycle length that is the basis for cellular circadian rhythmicity (for reviews, see Refs. 53 and 55). The positive drive to the circadian feedback loop is provided by CLOCK (or NPAS2) and BMAL1, basic helix-loop-helix PAS-containing transcription factors that form heterodimers capable of activating gene expression. Among the genes regulated by these heterodimers are two Period genes (Per1 and Per2) and two Cryptochrome genes (Cry1 and Cry2). Products of the Per and Cry genes form protein complexes that inhibit CLOCK/NPAS2:BMAL1-mediated transcription, closing the feedback loop.
Because of the widespread expression of clock genes, many tissues manifest circadian rhythms in gene expression. Normally, SCN-regulated physiological rhythms lead to entrainment of peripheral oscillators, coordinating the cells within a tissue and resulting in rhythmic gene expression throughout the body (50, 53). Even when the SCN is dysfunctional, manipulations, including glucocorticoid treatment, temporally restricted food intake, and forced exercise can entrain rhythmic gene expression rhythms in peripheral tissues (4, 39, 45, 47, 50).
Peripheral oscillators of animals failing to express robust behavioral and physiological rhythmicity would be expected to lack robust rhythmicity, due to the absence of rhythmic entraining signals (1, 34, 47). Thus, we predict that DKO mice would have severely compromised rhythmicity in peripheral tissues. Attenuation of rhythmic gene expression in peripheral tissues could reflect inability of the peripheral organs to oscillate, or just the lack of synchronizing inputs. These alternatives can be explored by artificially imposing rhythmicity in vivo.
The present work describes experiments to test these hypotheses, by assessing the impact of disruption of IA-2 and IA-2β on molecular rhythms in the SCN and peripheral tissues and assessing the impact of glucocorticoid treatment on molecular rhythms in peripheral tissues of DKO mice.
MATERIALS AND METHODS
Animals and Study Protocols
Targeted disruption of the mouse IA-2 and IA-2β genes has been described previously (24, 25, 48). The targeted alleles were backcrossed to the C57BL/6J genetic background for eight (IA-2) and four (IA-2β) generations, and heterozygotes were crossed to give double heterozygotes. Double heterozygotes were then interbred to generate lines of wild-type (WT) (IA-2+/+/IA-2β+/+) mice, two lines of single knockouts [IA-2-KO (IA-2−/−/IA-2β+/+) and IA-2β-KO (IA-2+/+/IA-2β−/−)], three-allele mutants (IA-2+/−/IA-2β−/−), and DKO mice (IA-2−/−/IA-2β−/−). Study mice were generated by breeding animals of the same genotype within each line, except that male IA-2−/−/IA-2β−/− mice were bred to female three-allele female (IA-2+/−/IA-2β−/−) mice to generate DKO mice because female DKO mice are infertile (25). All mice were genotyped before and after use in the study by PCR, as previously described (24, 48). The colonies were maintained in a 12:12-h light-dark lighting schedule, with lights on at 0600 and maintained under specific pathogen-free conditions with food and water available ad libitum. By convention, Zeitgeber Time (ZT) refers to time relative to the onset of light, so the light phase is ZT0–ZT12, and the dark phase is ZT12–ZT24. Circadian Time (CT) is used to refer to corresponding clock times, for animals housed in constant darkness (DD).
For evaluating gene expression in the SCN, brains were collected from mice euthanized by CO2 asphyxiation at 4-h intervals over 24 h. To avoid a direct influence of light on SCN gene expression, samples were collected on the first day in DD, starting at CT2 (14 h after the last light exposure) through CT22. Additional mice of each genotype (4–6) were exposed to room light for 1 h, from CT17 to CT18, before brain collection. Brains were removed, frozen in cooled 2-methylbutane, and stored at −80°C until sectioning (see In Situ Hybridization). Only WT and DKO animals were examined in this study.
For analysis of peripheral gene expression rhythms, organs (heart, kidney, liver, and skeletal muscle) were collected from WT and DKO mice at 4-h intervals over 24 h from animals housed in a light-dark cycle. Animals were euthanized by CO2 asphyxiation at each time point. Animals from time points ZT4, -8, -12, and -24 were euthanized in the light, and mice from time points ZT16 and ZT20 were euthanized in the dark. The genes examined in this study were several core clock genes (Per1, Per2, Bmal1, Cry1, and Cry2) and two genes that are widely and rhythmically expressed “clock-controlled genes” (Dbp and Pai-1).
An independent study of similar design was performed to compare peripheral gene expression in single-KO lines, relative to WT controls and DKO mice, with all animals housed in a light-dark cycle. The four genes showing the most robust rhythmicity across tissues in the previous experiment (Per1, Per2, Dbp, and Bmal1) were included in this experiment.
To assess whether the gene expression rhythms in peripheral tissues of DKO mice were directly driven by the light-dark cycle, liver tissue was collected from WT and DKO mice on the first day in DD. Expression of four genes (Per1, Per2, Dbp, and Bmal1) was examined.
Experiments 2–4 indicated disruption of rhythms in peripheral tissues of DKO mice. We therefore examined whether daily dexamethasone injections could restore gene expression rhythms in peripheral tissues of DKO mice. Dexamethasone 21-phosphate (DEXA, 2 mg/kg ip; Sigma D-1159) was injected daily at 5–6 P.M. (ZT11–ZT12) for 15 days. A single injection of DEXA at this time of day is expected to produce a phase delay of molecular oscillations in WT mice (4, 47). Groups of DKO and WT animals were injected with DEXA, whereas parallel groups of DKO and WT mice received the solvent (PBS with 0.15% ethanol) for 15 days. After the 15th injection, mice were transferred to DD. The next day, starting at CT6 (18 h after the last injection and last light exposure), animals were euthanized by CO2 asphyxiation at 6-h intervals, and organs were harvested. Discontinuation of the lighting cycle and of DEXA injection during the collection period allow distinction between possible acute effects of DEXA or light on gene expression profiles vs. DEXA-induced rhythmicity in peripheral oscillators.
Animals used in these studies were matched for age and gender across time points within each experiment. All experimental procedures were conducted at the National Institutes of Health (NIH), in accordance with the Principles of Laboratory Animal Care, National Institute of Health guidelines, and with the approval of the NIH Institutional Animal Care and Use Committee.
In Situ Hybridization
Radioactively labeled cRNA probes were generated by in vitro transcription in the presence of [35S]UTP to generate antisense and sense (control) probes as previously described (38). The plasmids used to generate probes to detect Per1, Per2, Dbp, and Rev-erb-α have been described previously (9).
Brains were sectioned in a cryostat at 15-μm thickness, with sections collected to slides as a 1-in-8 series so that each slide contained sections at 120-μm intervals through the SCN region. Slides were prehybridized, probe was applied (10 million counts·min−1·ml−1), and slides were then hybridized overnight at 53°C (9). High-stringency washes were performed the following day. All slides hybridized with a probe were processed in a single run. The details of the procedures have been described previously (9). After dehydration, slides were apposed to BioMax film. Image analysis was performed on the section representing the mid-SCN region on each slide. The optical density of the SCN was determined using a computer-based system and NIH Image software. The system was calibrated to absolute optical densities using Kodak photographic step tablet no. 2. Values are expressed as means ± SE for four to eight brains per genotype per time point, for each gene. Samples from WT mice collected at CT22 were lost.
Analysis of In Situ Hybridization Data
All SCN gene expression data were analyzed using GraphPad Prism (GraphPad Software, La Jolla, CA). Circadian time course data were initially analyzed by two-way ANOVA, using data from CT2 to CT18. Comparison across genotypes at each circadian time was achieved using Bonferroni's t-test. Subsequently, each genotype was analyzed separately by one-way ANOVA to determine if significant differences existed among the time points within each genotype. A significant variation among the time points, identified by ANOVA, was taken as evidence of rhythmicity; conversely, when there was no significant main effect of time, data sets are referred to as “arrhythmic.”
To assess the impact of light at night, Per1 and Per2 gene expression levels in SCN were compared by two-way ANOVA. Preplanned comparisons of the effect of light within each genotype were conducted using Student's t-tests.
For all comparisons, statistical significance was determined as P < 0.05.
RNA Extraction and cDNA Synthesis
Tissues were homogenized in 1 ml of Trizol (Invitrogen), and total RNA was then extracted according to the manufacturer's instructions. Total RNA (5 μg) was then treated with 2 units of RNase-free DNase (Sigma-Aldrich) for 30 min at 37°C to remove genomic DNA. DNase-treated RNA samples were tested for DNA contamination by PCR using primers to β-actin. DNase-treated RNA was then reverse transcribed with Moloney murine leukemia virus reverse transcriptase using random primers (Invitrogen). The β-actin primers were used to assess cDNA quality after reverse transcription.
Quantitative Real-Time Reverse-Transcription PCR
Primers were designed for the genes Per1, Per2, Cry1, Cry2, Bmal1, Dbp, Pai-1, and Gapdh (Table 1). Primers were designed using Primer 3 software (Tm = 60°C) and were synthesized by Integrated DNA Technologies (Coralville, IA). Real-time PCR [quantitative real-time reverse-transcription PCR (qRT-PCR)] was done using SYBR Green I Master mix using an ABI PRISM 7700 Sequence Detection System. For each gene, the primers were tested for amplification of a single product by melting curve analysis and visualizing products on agarose gels. Primers resulting in multiple peaks in the melting curve analysis or multiple bands on agarose gels were redesigned. Duplicate reactions from each sample were processed together and averaged. Data for each transcript were normalized to Gapdh as a control using the 2−ΔΔCT method.
Statistical Analysis of qRT-PCR Data
Gene expression values from qRT-PCR are often not normally distributed (S. Baker, personal communication), making parametric statistical analysis inappropriate. Because of the sample sizes used within each gene-tissue-treatment combination, it was not possible to assess the distribution within individual groups. It was possible, however, to assess the extent to which the residuals (difference between each value in a cell and the mean of that cell) were normally distributed. SPSS was used to perform this analysis. In the vast majority of gene-tissue combinations, the residuals were more normally distributed after transformation of the values by taking the natural log (Ln), relative to using the original data. Thus, all qRT-PCR data were analyzed using the Ln of the original values. Mean and SE values calculated in Ln units by SPSS were converted to linear units using Microsoft Excel, as follows: Mean = EXP(MeanLn), Mean − SE = EXP(MeanLn − SELn), Mean + SE = EXP(MeanLn + SELn). This method of analysis results in plus and minus standard errors that are not equal when plotted in linear units. All qPCR data (relative gene expression levels and rhythm amplitude) are plotted in linear units, using values derived from the Ln data.
Peripheral tissue gene expression data were first analyzed as “relative gene expression level.” For each gene in each tissue, expression levels were normalized to the lowest mean value in the experiment, which was set to one. Because all values from all genotypes are divided by the same value for a given gene-tissue combination, expression levels are directly comparable across genotypes. (Data were then converted to Ln values for statistical analysis, as discussed above.)
Statistical analysis of relative gene expression data began with two-way ANOVA, to assess the main effect of time, genotype, and interactions. Data were analyzed with a general linear model (version 19; IBM SPSS Statistics). Tukey's Honest Significant Differences (HSD) test was used for multiple-comparison testing (SPSS); data from all genotypes and times were analyzed together, and significant differences among preplanned comparisons were identified. For reporting purposes, comparisons were limited to assessing all possible differences between six time points within a genotype (15 comparisons), and comparisons across genotype at each of six time points, for each gene in each tissue studied.
Most peripheral gene expression profiles retained significant differences among the time points when data were expressed as relative gene expression level. However, many rhythms in DKO mice were characterized by a reduction in peak expression level and elevation of nadir levels. Analysis of the relative gene expression as described above seemed to underestimate the effect of IA-2/IA-2β gene disruption. Data were thus also expressed as “rhythm amplitude,” following the procedure described by Loh et al. (34) in their analysis of similar data (except that our analysis utilized Ln conversion of the data). Within each genotype, rhythm amplitude in Ln units was calculated by subtracting each value at the time of peak expression by the average of the values at the time of lowest expression. (This is equivalent to division of the peak values by the mean of the nadir when working in linear units.) Amplitude thus represents the fold change in gene expression (peak-to-trough ratio), calculated separately for each genotype, tissue, and gene. Amplitude values were compared across genotypes using GraphPad Prism, using Student's t-test (experiments 2 and 4), one-way ANOVA followed by Dunnett's t-test (to compare each mutant genotype with the WT group in experiment 3), or one-way ANOVA followed by Tukey's HSD test (experiment 5).
SCN Gene Expression Rhythms are Disrupted in DKO Mice
In situ hybridization was used to examine gene expression profiles in the SCN of WT and DKO mice (Fig. 1). Two-way ANOVA revealed a significant effect of time for each of the four genes examined (Per1, Per2, Dbp, and Rev-erb-α). With the exception of Per2, the main effect of genotype or the genotype × time interaction was also significant for each gene, and the gene expression levels differed significantly between the genotypes at one or more time points (Fig. 1). Conducting one-way ANOVA within each genotype revealed robust rhythms of Per1, Per2, Dbp, and Rev-erb-α expression in the SCN of WT mice (each P ≤ 0.0001). In contrast, expression of each of these four genes was not rhythmic in the SCN of DKO mice (Per1, F4,25 = 1.02, P = 0.4187; Per2, F4,26 = 1.00, P = 0.426; Dbp, F4,25 = 0.66, P = 0.625; Reverb-α, F4,26 = 2.08, P = 0.057). Notably, in DKO mice, SCN expression levels for each gene were near the average of peak and trough values observed in the WT mice. Thus the altered patterns in DKO mice are not due to a loss of expression levels; they are due to a loss of rhythmic gene expression.
Preserved Retinohypothalamic Signaling in DKO Mice
One mechanism by which the absence of IA-2 and IA-2β could disrupt SCN rhythmicity involves alterations in signaling from the retina. Exposure of mice to low-intensity constant light lengthens the circadian period in rodents while exposure to brighter light leads to arrhythmicity. Alterations in retinohypothalamic signaling, perhaps through altered release of glutamate, pituitary adenylyl cyclase-activating peptide, or little SAAS, could contribute to disrupted SCN function (2, 12). To test the functional integrity of the retinohypothalamic tract and SCN responsiveness to lighting information, groups of animals of each genotype were exposed to light for 1 h beginning at CT17, and brains were collected at CT18. These brains were collected at the same time as those described above, and Per1 and Per2 gene expression levels were assessed in the same in situ run as the circadian time course study. Thus, CT18 values from Fig. 1 are replotted in Fig. 2 as the dark-housed controls. Exposure to 1 h of light at night was expected to increase Per1 and Per2 gene expression levels in the SCN of WT mice (57).
Two-way ANOVA of the Per1 data revealed a significant main effect of light exposure (P = 0.0009) but no genotype effect or interaction (P = 0.541 and P = 0.420, respectively). Preplanned comparisons between dark- and light-exposed groups revealed that light exposure caused a significant increase in Per1 expression levels in both WT (P = 0.0192) and DKO (P = 0.0263) mice.
Two-way ANOVA of the Per2 data revealed a significant main effect of light (P < 0.0001) and genotype (P = 0.0033) and a significant interaction (P = 0.0105). Preplanned comparisons revealed that light exposure significantly increased Per2 in SCN of WT mice (P < 0.0172) but did not significantly increase Per2 levels in DKO mice (P = 0.1183). This lack of significant response could be due to the higher baseline level of Per2 in the DKOs and to the fact that Per2 typically peaks 90–120 min after light onset (57). With a 60-min exposure period, the response of Per2 would not be maximal.
Peripheral Gene Expression Rhythms are Blunted in DKO Mice
In experiment 2, qRT-PCR was used to assess gene expression levels at 4-h intervals throughout the 24-h light-dark cycle in WT and DKO mice (Fig. 3). The results of two-way ANOVAs on these data appear in the appendix as Table A1. Two-way ANOVAs revealed a significant main effect of time for all gene-tissue combinations studied except Cry2 in heart and skeletal muscle (Table A1). Post hoc analysis, comparing across time points within each genotype, indicated that significant differences among the time points were more prevalent in WT tissues than in DKO tissues (Table A1), suggesting a difference in rhythm amplitude. Direct comparison between WT and DKO gene expression levels revealed a relatively low number of significant differences (Fig. 3), but, in each case, the differences reflected a decrease in rhythm amplitude (reduction in the high point or elevation of the low point) in DKO mice compared with WT mice.
To confirm the impression of reduced rhythm amplitude, gene expression was analyzed as the peak-to-trough ratio (rhythm amplitude, or fold increase in gene expression over basal levels) for each gene, separately for each genotype (Fig. 4). The amplitude of rhythmicity was reduced significantly in DKO mice, relative to WT controls, for most genes in each of the tissues studied (Fig. 4). Rhythm amplitude for Cry1, Cry2, and Pai-1 was modest in the tissues studied, even in WT mice, so these genes were not analyzed in subsequent studies.
Peripheral Gene Expression Rhythms are Relatively Unaffected in Single-KO Mice
Several lines of evidence indicate functional redundancy between IA-2 and IA-2β. Most notably in this context, loss of behavioral and physiological rhythms occurs only in mice with disruption of both genes (21). The single-KO mice have, at most, a subtle circadian phenotype (21). We predicted that the maintenance of physiological rhythmicity in each single-KO line would maintain peripheral gene expression rhythms. Experiment 3 was thus conducted to assess the contribution of IA-2 and IA-2β, individually, to peripheral gene expression profiles (Figs. 5 and 6).
Samples from single-KO mice were processed in parallel with a set of tissue samples from WT and DKO mice that were independent of the samples processed in experiment 2 and assessed for gene expression in three peripheral tissues (heart, kidney, and liver). Two-way ANOVAs on expression levels of each of four genes revealed significant effects of time and a significant interaction, in most analyses (see appendix, Table A2). Significant effects of genotype were observed in one-half of the comparisons, but notably in each case there was a significant interaction term as well. Multiple-comparison tests were used to identify the location of significant differences. As in the previous study, reduced rhythm amplitude in DKO mice was suggested by the reduced number of significant comparisons among times within genotype, relative to WT. In addition, multiple-comparison testing revealed that values from DKO mice differed significantly (P < 0.05) from those of WT mice in several comparisons. More specifically, WT mice differed from DKO mice in 17 comparisons (heart: Per1, Per2, and Dbp at ZT4, and Per2 at ZT24; kidney: Per2 at ZT16 and ZT24, Dbp at ZT12 and ZT24, Bmal1 at ZT8, ZT12, and ZT24; liver: Per1 at ZT4 and ZT24; Per2 at ZT4; Dbp at ZT4 and ZT24; Bmal1 at ZT12). There were only two other comparisons among the entire data set where values differed significantly between WT and a mutant line: Bmal1 expression differed between IA2β-KO and WT mice in both heart and liver at ZT8.
Analysis of rhythm amplitude provided clear evidence that gene expression rhythms were blunted in the DKO mice but not in the single-KO lines (Fig. 6). ANOVA revealed significant differences in rhythm amplitude among the four genotypes for each gene and tissue examined. Dunnett's t-test was used to compare rhythm amplitude for each of the three mutant genotypes with WT controls (Fig. 6). Single KOs (IA2-KO or IA2β-KO mice) differed significantly from WT mice in only 2 of 12 comparisons for each genotype. In contrast, rhythm amplitude was reduced significantly in 10 of 12 comparisons between DKO mice and WT controls, consistent with the results from experiment 2 (Fig. 4). Thus, neither single-KO line recapitulates the generalized reduction of rhythm amplitude observed in the DKO animals. This suggests that IA-2 and IA-2β are partially redundant in function, as previously observed for other physiological endpoints, including glucose tolerance, insulin secretion in vitro, density of dense core vesicles, fertility and behavior (6, 24, 25, 26, 41, 48), as well as for behavioral circadian rhythms (21).
Disruption of Peripheral Rhythms in DKO Mice Housed in DD
The presence of rhythmicity in DKO mice when housed in a light-dark cycle could reflect rhythmicity imposed or amplified by the environmental cyclicity. To more clearly assess the extent to which circadian rhythmicity is disrupted in DKO mice, we conducted a preliminary experiment comparing WT and DKO mice on the first day in DD (Fig. 7). Two-way ANOVAs revealed significant main effects of genotype and time, and significant interactions, for each of the four genes analyzed (Fig. 7). Pairwise comparisons revealed several significant differences between the genotypes (Fig. 7). Rhythm amplitude for each of the four genes examined (Per1, Per2, Dbp, and Bmal1) was reduced significantly in DKO mice relative to WT mice (Fig. 7, bottom). While direct quantitative comparison across the experiments is not possible, the extent of disruption of gene expression rhythms appears greater in DKO mice on the first day in DD (Fig. 7) relative to when DKO mice are housed in a light-dark cycle (Figs. 3 and 5).
Peripheral Gene Expression Rhythms in DKO Mice: Effects of DEXA Injections
The previous experiments indicate that DKO mice have defects in both SCN and peripheral tissue rhythmicity. The circadian system is hierarchical, with the SCN generating numerous physiological rhythms that impact peripheral tissues, synchronizing their rhythmicity. In the absence of a functional SCN, peripheral gene expression rhythms are disrupted. Importantly, DEXA treatment can restore rhythmicity in peripheral tissues of mice with SCN lesions (47). Experiment 5 was thus performed to test whether it is possible to bypass the SCN to entrain peripheral tissues in DKO mice. More specifically, we assessed whether daily DEXA injections could restore gene expression rhythms in liver and heart of DKO mice. Groups of DKO and WT animals housed in a 12:12-h light-dark cycle were injected with DEXA or vehicle control in the last hour of the daily light phase (ZT11–ZT12) for 15 days. Mice were then transferred to DD, and tissues were collected at 6-h intervals starting at CT6 (18 h after the last injection and last light exposure). The lighting cycle and DEXA injection were discontinued during the collection period to allow assessment of rhythms in the absence of acute responses to these stimuli.
Three-way ANOVA was used to assess the effects of circadian time, genotype, and treatment (DEXA vs. saline). For both tissues and for all genes studied, there was a significant effect of time (P < 0.001). In liver, significant effects of treatment were observed for Per1, Per2, and Bmal1, and a significant effect of genotype was observed for Per2. Notably, however, for each gene, several of the interaction terms were also significant. Similarly, in heart, significant effects of treatment were observed for Per1, Per2, and Bmal1, a significant effect of genotype was apparent for Dbp, and numerous interaction terms were highly significant. Because of the numerous comparisons possible, we focused our attention on assessing whether gene expression levels differed between DEXA- and vehicle-treated animals of each genotype at each circadian time (Fig. 8), and on assessment of rhythm amplitude (Fig. 9).
DEXA treatment affected relative gene expression levels in both WT and DKO mice, in both tissues (Fig. 8).
In vehicle-treated animals studied on the first day in DD, rhythm amplitude of all four genes differed between WT and DKO mice in liver (Fig. 9), consistent with results in Fig. 7. Per1 and Per2 rhythm amplitude also differed between the genotypes in heart.
DEXA treatment had gene- and tissue-specific effects in DKO mice. In all cases where it had a significant effect in DKO mice, its effect was to increase rhythm amplitude (Fig. 9). More specifically, DEXA significantly increased rhythm amplitude for Per1 and Per2 in liver and for Per2, Dbp, and Bmal1 in heart. DEXA was without significant effect on rhythm amplitude for Per1 in heart and for Dbp and Bmal1 in liver.
DEXA treatment also significantly affected rhythm amplitude in WT mice. Rhythm amplitude of Per1, Per2, and Dbp was unexpectedly decreased by DEXA treatment in WT liver. Bmal1 rhythm amplitude was increased by DEXA in both liver and heart. Rhythm amplitude for Per1, Per2, and Dbp was not affected significantly by DEXA in heart of WT mice (Fig. 9). Based on prior literature, one would expect a phase shift of peripheral rhythmicity following DEXA administration in WT mice, rather than the effects on rhythm amplitude noted here (4, 23).
To summarize, gene expression rhythms in DKO mice collected on the first day in DD were generally reduced in amplitude, relative to WT mice, and DEXA treatment increased rhythm amplitude for several genes in liver and heart of DKO mice.
A previous study of physiological rhythms in DKO mice revealed significant disruption of rhythms of body temperature, heart rate, and blood pressure and altered electrical activity rhythms in the SCN (21). Results presented here indicate a severe defect in molecular rhythms in the SCN, despite preservation of the functional integrity of the retinohypothalamic tract that conveys light information from the retina to the SCN. Perhaps not unexpectedly in view of the hierarchical nature of the mammalian circadian timing system, DKO mice have blunted rhythms in gene expression in peripheral tissues. Neither single-KO line has as severe a disruption of rhythms as the DKO line. DEXA treatment generally improved rhythms in peripheral tissues of DKO mice.
Our qRT-PCR studies show that the amplitude and peak levels of gene expression are reduced in several peripheral tissues of DKO mice. Peripheral tissues have a small number of dense core and synaptic vesicles, relative to their mass, making it very unlikely that the loss of IA-2 and IA-2β within the peripheral tissues is solely responsible for the molecular alterations measured in those tissues. This instead suggests that changes in peripheral gene expression rhythms in DKO mice do not arise locally but instead result from the lack of signaling from the SCN. DEXA improves rhythms in DKO mice, most notably for the clock genes Per1 and Per2. The suppression of Per1, Per2, and Dbp rhythm amplitude in liver of WT mice was unexpected. These results strongly suggest that peripheral oscillators of DKO mice contain competent oscillators; the disruption of peripheral gene expression rhythms in DKO mice appears secondary to the loss of entraining signals emanating from the SCN. In this regard, DKO mice may resemble mice in which the SCN has been disrupted by electrolytic lesion; arrhythmicity of peripheral gene expression is rescued in SCN-lesioned mice by DEXA injection (47). The absence of glucocorticoid receptors in SCN, the requirement for hepatic glucocorticoid receptors for DEXA-induced shifting of hepatic gene expression rhythms, and the SCN-independent effects of DEXA all indicate that DEXA acts directly on peripheral tissues rather than through the SCN (4, 47). Mice lacking VIP have a phenotype similar to DKO mice, with attenuation of peripheral circadian rhythms that appears secondary to failure of the SCN to provide a strong entraining signal (34). It is interesting to note that peripheral gene expression rhythms are not completely absent in DKO mice, especially when housed in a light-dark cycle. This suggests that the SCN (or other neural sites) retain the capacity to transmit information to peripheral sites. Whether this reflects a form of masking by environmental stimuli, leading to modest synchronization of oscillators in peripheral tissues, is not clear.
How does loss of IA-2 and IA-2β disrupt SCN function? Our assessment of SCN gene expression in DKO mice indicates that these synaptic vesicle proteins are not essential for expression of key circadian clock genes. Rhythmic gene expression is lost in DKO mice, but gene expression levels are near the average observed over the course of the day in WT mice. This could reflect either arrest of the circadian cycle within each neuron or failure of coordination among a population of rhythmic SCN neurons. The mechanisms of disrupted SCN function in DKO mice could be addressed through imaging Per2::luciferase rhythms of individual cells plated at low density [to assess the functional autonomy of individual SCN neurons or fibroblasts (32)] and by examining the ability of SCN slices from DKO mice to send and respond to paracrine signals that can synchronize SCN neuronal populations (38). In addition, confirmation of the presumed deficits in release of neuropeptides (and classical transmitters, including γ-aminobutyric acid) would be informative.
Recent studies have emphasized the importance of communication among SCN neurons for rhythmicity at the tissue level. Liu et al. (32) showed that SCN neurons can compensate for each others' genetic shortcomings: the effects of “clock gene” disruption are much more severe when assessed in single SCN neurons than when examined in slices of SCN tissue. The network properties of the SCN allow a small population of functional oscillators to synchronize the larger population of SCN neurons, greatly amplifying their molecular and neurochemical outputs. Attempts to identify the population of functional oscillators suggests that there is no single population and that instead SCN neurons are unstable oscillators that rely on network interactions to function as the central oscillator (54). Neuropeptidergic interactions, through effects on downstream signaling pathways, play a key role in these interactions (18, 38). Coupling among SCN neurons thus emerges as a key feature important for SCN function and secondarily for synchronizing peripheral oscillators.
Indeed, circadian rhythms can be disrupted by interfering with the ability of SCN neurons to communicate with each other and with their downstream targets. Several neuropeptide-related genes play key roles in generating robust rhythmicity at the behavioral level. Mice with targeted disruption of the Prokineticin 2 (PK2) gene or the PK2R2 receptor have reduced amplitude of sleep-wake, activity, and temperature rhythms (31, 46). Notably, SCN rhythmicity, as assessed by in vitro Per2::luciferase rhythmicity and in situ hybridization for clock genes, is unaffected (31, 46). Thus, PK2 and PK2R2 are not necessary for SCN oscillator function but are required for the SCN to communicate temporal signals to effectors within the brain, leading to physiological rhythms. Our in situ hybridization results showing loss of rhythmic SCN gene expression in DKO mice indicates that IA-2 and IA-2β contribute to circadian function at the level of the oscillator rather than simply in the output pathway.
In marked contrast to the maintenance of SCN function in mice with disruption of the PK2 system, and more similar to DKO mice, is the profound disruption of physiological rhythms and SCN function in mice lacking VIP or the VPAC2 receptor for VIP (3, 5, 7, 8, 13, 15, 33, 40, 45). In VPAC2-deficient mice, rhythmic gene expression in the SCN is reduced markedly in amplitude, and analysis of individual cells indicates a loss of rhythmicity in many neurons (3, 40). Thus, VIP is thought to be important for maintaining circadian rhythmicity in a subset of SCN neurons and furthermore plays a key role in maintaining coordinated rhythmicity of the ensemble of SCN neurons.
IA-2 and IA-2β are transmembrane proteins located in both dense core vesicles and synaptic vesicles (11, 42, 49). These genes play important roles in synaptic and dense core vesicle regulation, and their disruption leads to reduced insulin secretion and dense core vesicle number in pancreatic β-cells (6, 14, 16). Although DKO mice have impaired first-phase insulin secretion and modest hyperglycemia in response to glucose challenge, they do not have elevated nonfasting glucose and do not have insulin resistance (26); they are thus not overtly diabetic. High-dose streptozotocin treatment induces severe hyperglycemia and diabetes and alters gene expression rhythms in mouse liver (but not SCN) (28, 43), but this is not likely relevant to the circadian phenotype of DKO mice.
In addition, the IA-2 and IA-2β proteins are important for neurotransmitter secretion (41, 42), and their disruption leads to behavioral abnormalities and loss of female fertility (25, 42). Disruption of both IA-2 and IA-2β genes, which are expressed throughout the SCN (21), likely leads to disruption of neurotransmitter release critical for coordination of SCN neurons. Mutations affecting other components of the vesicle recycling machinery further implicate a role for vesicle trafficking in circadian rhythms. Mice lacking chromogranin A have disrupted diurnal rhythmicity in blood pressure (37). The Earlybird mutation of Rab3a (an abundant Ras-associated binding protein that regulates synaptic vesicle trafficking) has a modest short-period phenotype in DD (19). The circadian phenotype of mice bearing the mutant Rab3a allele is more severe than the phenotype of mice homozygous for a null allele, however, indicating the mutant gene product may have dominant-negative activity; additional members of the Rab family may compensate for the loss of Rab3a in the KO mice and may attenuate the impact of the Earlybird mutation.
Using a proteomics approach, Deery et al. (10) identified rhythmic expression in the SCN of several proteins involved in secretory vesicle trafficking. To test the functional importance of vesicle recycling, organotypic SCN slices were subjected to long-term treatment with agents that disrupt the endocytosis/exocytosis vesicle cycle. These treatments reduced the amplitude of reporter gene rhythms while video microscopy revealed reduced amplitude of rhythmicity within individual SCN neurons and loss of phase coherence of oscillations between neurons (10). Thus, disruption of vesicle trafficking compromised SCN circadian pacemaking. Recent studies involving manipulation of synaptic vesicle cycling in Drosophila circadian clock neurons also indicate an important role in circadian timekeeping (20, 56).
Our work with DKO mice reveals functional redundancy between IA-2 and IA-2β in the maintenance of circadian rhythms (Ref. 21 and the present results). This likely stems from the overlapping function of these proteins in maintaining dense core vesicles; disruption of either gene alone affects neurotransmitter and insulin release, but the most profound deficits are found in mice with disruption of both genes (6). Remarkably, our studies and those of others reveal that alterations in secretory vesicle proteins, which are not canonical clock proteins, can have a profound effect on circadian rhythms. Examination of additional genes involved in regulation of neurosecretory and dense core vesicles for a role in mammalian circadian rhythms is warranted.
This work was supported by the intramural research program of the National Institute of Dental and Craniofacial Research, National Institutes of Health (NIH), and by NIH Grant NS-056125 (to D. R. Weaver). K. K. Rumery was supported in part by a Summer Undergraduate Research Program fellowship supported under NIH Grant 5R25HL-092610-03 and by the Office of Research of the University of Massachusetts Medical School while an undergraduate at St. Olaf College, Northfield, MN.
The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors declare no conflicts of interest.
S.P., A.L.N., and D.R.W. did the conception and design of the research; S.P., K.K.R., E.A.Y., C.M.L., and D.R.W. performed the experiments; S.P., K.K.R., C.M.L., and D.R.W. analyzed the data; S.P., K.K.R., A.L.N., and D.R.W. interpreted the results of the experiments; S.P., K.K.R., and D.R.W. prepared the figures; S.P., K.K.R., E.A.Y., C.M.L., A.L.N., and D.R.W. edited and revised the manuscript; S.P., K.K.R., E.A.Y., C.M.L., A.L.N., and D.R.W. approved the final version of the manuscript; D.R.W. drafted the manuscript.
We are deeply indebted to Stephen P. Baker for extensive consultations regarding statistical analysis of the data; his effort is supported by the University of Massachusetts Center for Clinical and Translational Science Award No. NIH 8UL1TR000161.
Present addresses: S. Punia, Departments of Internal Medicine and Genetics, Yale School of Medicine, New Haven, CT 06520-8029; and K. K. Rumery, Minneapolis Heart Institute Foundation, 920 East 28th St., Ste. 620, Minneapolis, MN 55407.