The Construction of Mice with a Conditionally Active, Liver-Specific Rev-erbα Transgene
We wished to engineer a mouse strain with conditionally active circadian oscillators specifically in hepatocytes, in order to examine the contribution of local clocks and systemic Zeitgeber cues to rhythmic liver gene expression. As mentioned above, BMAL1 is a constituent of the molecular oscillator whose loss of function immediately results in the abolishment of all manifestations of circadian physiology and gene expression . We thus thought that the conditional expression of Bmal1 specifically in hepatocytes may provide such a model system. Bmal1 transcription follows a high-amplitude circadian cycle, owing to the circadian accumulation of REV-ERBα, a nuclear orphan receptor that strongly represses Bmal1 transcription (Figure 1A) . We thus exploited this regulatory mechanism to produce a mouse strain with a conditionally active Bmal1 gene in hepatocytes. First, a transgenic mouse strain was established in which transcription of an HA epitope-tagged REV-ERBα version (HA-REV-ERBα) is controlled by tetracycline-responsive elements (TREs). These mice were crossed with LAP-tTA mice, which express a tetracycline-dependent transactivator (tTA) specifically in hepatocytes . In LAP-tTA/TRE-Rev-erbα double transgenic mice, HA-REV-ERBα accumulated to constitutively high levels and suppressed Bmal1 expression throughout the day in the absence of the tetracycline analog doxycycline (Dox) (Figures 1B, 1C [left panel], and S1). LAP-tTA/TRE-Rev-erbα mice thus produced HA-REV-Erbα in a liver-specific and tetracycline-dependent fashion. In the presence of Dox, the TRE-Rev-erbα transgene remained silent, and circadian oscillator function was not perturbed in liver cells (Figures 1B, 1C [right panel], and S1). As shown in Figure 1B, the regulation of TRE-Rev-erbα transgene expression was exquisitely tight, since in Dox-fed mice, neither mRNA nor HA-REV-ERBα protein was detectable by Taqman RT-PCR assays and Western blot experiments, respectively. At least in part, the low levels of Bmal1 mRNA and protein observed in the liver of untreated animals may have been contributed by endothelial cells, bile duct cells, or Kupffer cells, which did not express the LAP-tTA transgene . In rat liver, nonparenchymal cells contribute about 35% of all cells and about 10% of the cellular hepatic volume [29,30], and we thus expect that around 10% of the liver RNA is contributed by cells that do not express the LAP-tTA transgene.
We have used mice homozygous for both the LAP-tTA and TRE-Rev-erbα transgenes in these experiments, in order to maximize transgene expression. Double homozygous mice were born at Mendelian ratios and did not show any phenotype with regard to morphology, vigor, litter size, or circadian behavior. We also determined the integration site for both transgenes (see Materials and Methods). LAP-tTA was found to be inserted in reverse orientation into the first intron of Zfp353 (Chromosome 8), more than 100 kilobases (kb) downstream of the first exon and about 200 kb upstream of the second exon. TRE-Rev-erbα was found to be inserted in reverse orientation into the first intron of Semaphorin3 (Chromosome 5), about 6 kb downstream of the first exon and more than 100 kb upstream of the second exon. We thus consider unlikely that the transgene integrations interfered with circadian clock function.
The Hepatic Expression of Putative BMAL1 Target Genes in Mice Fed with or without Dox
As would be expected for BMAL1 target genes, the expression of mPer1, Dbp, and endogenous Rev-erbα was low in the absence of Dox, when HA-REV-ERBα overexpression attenuated Bmal1 transcription. Unexpectedly, however, the circadian clock genes mCry1, mCry2, and mPer2 displayed milder expression differences in Dox-treated and untreated animals (Figure 2). Remarkably, the rhythmic expression of mPer2 mRNA and protein levels was almost unaffected by the down-regulation of Bmal1 expression. As reported previously , mCRY2 oscillated in abundance during the day despite nearly constant mCry2 mRNA levels. Conceivably, the association of mCRY2 with PER proteins—i.e., mPER2 in the absence of Dox—affected the metabolic stability of mCRY2 in a daytime-dependent manner.
The robust circadian expression of mPer2 in the liver of mice not receiving Dox is in stark contrast to the in situ hybridization experiments with coronal brain sections of Bmal1-deficient mice, which indicated that in the absence of BMAL1, mPer2 mRNA accumulates to insignificant levels throughout the day in SCN neurons . However, it is in keeping with the relatively high constitutive mPer2 mRNA concentrations observed in the liver of these Bmal1 knockout mice (J. S. Takahasi, unpublished data), assuming that in liver, mPer2 transcription depends less on BMAL1 than in the SCN. Nevertheless, our observation could be interpreted in two ways. Either, the residual BMAL1 levels in the liver of animals not treated with Dox were still sufficient to drive mPer2 transcription, or cyclic mPer2 expression was governed by oscillating systemic signals in these mice. In order to distinguish between these two scenarios, we wished to monitor temporal mPer2 expression in cultured liver explants, which obviously do not receive periodic signals from a master pacemaker. To this end, we crossed LAP-tTA/TRE-Rev-erbα mice with mPer2::luc knock-in mice , in which a luciferase open reading frame (ORF) was inserted by homologous recombination into the endogenous mPer2 locus. The mPER2::LUCIFERASE fusion protein encoded by this knock-in allele is fully functional, since it rescues all known rhythm phenotypes of mPer2 knockout mice . Tissue explants from the LAP-tTA/TRE-Rev-erbα transgenic mice carrying an mPer2::luc fusion allele were placed into culture medium containing luciferin, and bioluminescence was recorded in real time by photomultiplier tubes [2,31]. As shown in Figure 3A (top right panel), liver explants from these mice did not produce circadian luminescence cycles in normal culture medium, suggesting that overexpression of HA-REV-ERBα indeed arrested the hepatocyte clocks. However, when tissue pieces from the same livers were cultured in Dox-containing medium (Figure 3A, right center and bottom panels), circadian luminescence rhythms similar to those observed for explants of mPer2::luciferase mice not carrying the LAP-tTA and TRE-Rev-erbα transgenes (Figure 3A, left panels) could be observed. Interestingly, circadian luminescence cycles recorded from Dox-treated liver explants of LAP-tTA/TRE-Rev-erbα/mPer2::luc mice fed with normal chow (Figure 3A, right center panel), displayed a phase delay of approximately 6 h when compared to those obtained from liver explants of mPer2::luciferase mice (Figure 3A, left center panel). This phase delay probably reflected the time period required for the decay of HA-Rev-erbα mRNA and protein, and for the consecutive accumulation of BMAL1 to levels compatible with circadian rhythm generation. In keeping with this conjecture, no significant phase differences were observed between luminescence cycles monitored for liver explants from mPer2::luc and LAP-tTA/TRE-Rev-erbα/mPer2::luc mice pretreated with Dox by intraperitoneal injections 48 h and 24 h before being sacrificed, (Figure 3A, bottom panels). We have examined liver explants from five mice homozygous (Figure 3A, and unpublished data) and three mice heterozygous (Figure S2A) for the LAP-tTA/TRE-Rev-erbα transgenes, and in all cases, circadian mPer2::luc expression strictly depended upon the addition of Dox to the culture medium. As expected, lung explants from either homozygous (Figure 3B) or heterozygous (Figure S2B) LAP-tTA/TRE-Rev-erbα/mPer2::luc mice displayed circadian luminescence rhythms, irrespective of whether or not Dox has been added to the culture medium. Indeed, TRE-Rev-erbα transgene expression is not detectable in this tissue by quantitative TaqMan real-time RT-PCR (unpublished data).
Taken together, our observations made with LAP-tTA/TRE-Rev-erbα mice and tissue explants suggest that in liver, circadian mPer2 expression can be driven by systemic Zeitgeber cues in the absence of functional hepatocyte clocks as well as by hepatocyte oscillators in the absence of systemic Zeitgeber cues.
Genome-Wide Mapping of Circadian Transcripts in Liver Cells with Operative or Attenuated Circadian Oscillators
To discriminate between oscillator-dependent and -independent circadian gene expression in a genome-wide fashion, we compared the circadian liver transcriptomes of mice fed with or without Dox by Affymetrix (MOUSE 430 2.0) microarray hybridization (for details and data analysis, see Materials and Methods, the microarray data are available from the ArrayExpress repository [http://www.ebi.ac.uk/arrayexpress/] under accession number: E-MEXP-842). This analysis revealed 351 circadian transcripts (represented by 432 feature sets) for Dox-treated animals, including most mRNAs known to fluctuate with a robust daily amplitude (e.g., mPer1, mPer2, mPer3, mCry1, Rev-erbα, Rev-erbβ, Bmal1, Clock, Dbp, Tef, Nocturnin, Rorγ, E4bp4, Cyp7a1, or Alas1). In keeping with previous studies [23–27], many cyclically expressed genes are involved in various aspects of liver physiology such as xenobiotic detoxification (e.g., P450 oxidoreductase, Por, Cyp2b9, Cyp2b10, Cyp2g1, and Fmo5), carbohydrate and energy metabolism (e.g., Gk, and Pepck), or lipid and sterol homeostasis (e.g., Elovl3, Insig2, Lipin1, and Cyp7a1). Importantly, the cyclic expression of most rhythmically active genes appeared to depend on an intact hepatocyte oscillator, as the amplitude of circadian accumulation was greatly affected in animals not receiving Dox-supplemented food (Figure 4). Nevertheless, using the algorithms described in Materials and Methods, we identified 31 different transcripts (represented by 41 feature sets), whose circadian accumulation was not affected by the Dox treatment. These are listed in the phase maps of Figure 5A (compare left and right panels), and the circadian expression of some of these genes in the presence and absence of Dox has been validated by Northern blot hybridization (Figure 5B). As expected on the basis of the results displayed in Figure 2, mPer2 mRNA was included among the transcripts whose cyclic accumulation was controlled by systemic cues. Other genes whose transcripts accumulate with phase angles similar to that of mPer2 mRNA were the heat-shock protein genes Hspca (encoding HSP90), Hspa8 (encoding HSP70 isoform 8), Hspa1b (encoding HSP70 isoform 1A), Hsp105 (encoding HSP105), and Stip1 (encoding Stress-Induced Phoshoprotein 1, also known as Hsp70/Hsp90 organizing protein), and a tubulin gene (Tuba4). These genes were expressed in phase with mPer2, suggesting that their cyclic transcription was perhaps governed by similar systemic timing cues. Nocturnin (Ccrn4l), Fus, Chordc1, and Cirbp were additional genes whose circadian expression appeared to be system driven. However, the transcripts issued by these genes belonged to different phase clusters (Figure 5A and 5B), and their synthesis must thus have been regulated by mechanisms different from those governing rhythmic Hsp and/or mPer2 transcription. Particularly interesting was the diurnal expression of heat-shock protein genes and Cirbp, a gene encoding a cold-induced RNA-binding protein. While heat-shock protein mRNAs reached zenith levels at Zeitgeber times when body temperature was maximal, Cirbp mRNA levels peaked at Zeitgeber times when body temperature was minimal [5,32]. Hence temperature cycles oscillating by only a few degrees (35 °C to 38 °C) appeared to be translated into antiphasic Hsp and Cirbp expression cycles.
We also considered the possibility that some systemically regulated liver genes could display mRNA accumulation cycles with higher amplitudes in the absence of functional hepatocyte oscillators. For example, the accumulation of liver transcripts whose synthesis is influenced by local oscillators and systemic cues in an antiphasic manner may only be circadian in mice not containing hepatocyte clocks. However, our failure to identify such transcripts did not support such a regulatory mode (see Figure S3A and S3B, and corresponding figure legends), we thus feel that few if any genes produce robust daily mRNA accumulation cycles only in the absence of functional hepatocyte clocks.
Temperature-Dependence of mPer2 Expression
The expression of mPer2 and Hsp appears to be similar with respect to systemic regulation. We thus suspected that a common regulator might influence the transcription of these genes. Since Hsp transcription is governed primarily by heat-shock transcription factors (HSF) , we wondered whether mPer2 transcription was also inducible by elevated temperature. In order to examine this conjecture, we incubated cultured organ explants from LAP-tTA/TRE-Rev-erbα/mPer2::luc mice during 150 min at 40 °C (Figure 6) and recorded bioluminescence in real time. Although luciferase activity was somewhat decreased during the heat shock itself, presumably due to a general inhibition of translation , a subsequent strong increase in luciferase activity was observed in both liver and lung. Liver explants cultured in the absence of Dox showed a consistent 2-fold enhancement of luciferase activity, suggesting that the heat-dependent regulation of mPer2 did not require a circadian clock (Figure 6A). Lung explants, in which circadian oscillators are operative under these conditions (see Figures 3B and S2B), displayed a phase-specific induction of temperature-induced luciferase activity (Figure 6B). Thus, when the heat shock was performed at a circadian time at which luciferase activity was minimal, a strong induction was observed. On the other hand, a heat shock performed at a circadian time when luciferase activity was maximal did not result in a noteworthy increase in luciferase activity. Taken together, these results indicate that mPer2 is heat inducible and that the strength of this induction is gated by circadian time.
The minimal HSF binding sites (heat-shock elements [HSEs]) consist of two or more inverted or everted repeats of the pentameric sequence 5′-NGAAN-3′ (where N can be any nucleotide). Taking the two complementary DNA strands into consideration, the statistical frequency of HSEs is approximately 1/2,000 in random DNA, and it is thus impossible to identify functional HSEs solely by sequence inspection. Nevertheless, known functional HSEs are located within 5′-flanking regions of heat-shock protein genes , and the sequence analysis of mPer2 revealed a cluster of five HSEs within the 1,700 base pairs (bp) located upstream of the transcription initiation site (Figure S4). Of note, one of these elements (centered around −1,630) lies within a 22-bp sequence block that is 100% identical in mouse, rat, human, and dog (Figure S4). Whether this or any other HSEs displayed in Figure S4 are involved in the temperature-regulation of mPer2 will have to be examined by site-directed mutagenesis and chromatin immunoprecipitation experiments.