Transcriptional profiling of circadian patterns of mRNA expression in the chick retina.

Previous transcriptome analyses have identified candidate molecular components of the avian pineal clock, and herein we employ high density cDNA microarrays of pineal gland transcripts to determine oscillating transcripts in the chick retina under daily and constant darkness conditions. Subsequent comparative transcriptome analysis of the pineal and retinal oscillators distinguished several transcriptional similarities between the two as well as significant differences. Rhythmic retinal transcripts were classified according to functional categories including phototransductive elements, transcription/translation factors, carrier proteins, cell signaling molecules, and stress response genes. Candidate retinal clock transcripts were also organized relative to time of day mRNA abundance, revealing groups accumulating peak mRNA levels across the circadian day but primarily reaching peak values at subjective dawn or subjective dusk. Comparison of the chick retina transcriptome to the pineal transcriptome under constant conditions yields an interesting group of conserved genes. This group includes putative clock elements cry1 and per3 in addition to several previously unidentified and uninvestigated genes exhibiting profiles of mRNA abundance that varied markedly under daily and constant conditions. In contrast, many transcripts were differentially regulated, including those believed to be involved in both melatonin biosynthesis and circadian clock mechanisms. Our results indicate an intimate transcriptional relationship between the avian pineal and retina in addition to providing previously uncharacterized molecular elements that we hypothesize to be involved in circadian rhythm generation.

Circadian biological rhythms are a pervasive property of most multicellular organisms, most eukaryotic microorganisms, and at least some prokaryotic taxa. It has been known for some time that circadian rhythms are genetically determined (1). Circadian traits such as free-running periods () and phase-angle (⌿) to entraining light cycles can be selected for, and mutations for these traits have been identified, leading to the isolation and characterization of molecular clock compo-nents (2,3). In the last 10 -15 years, the mRNA and protein products of many of these "clock genes" have been shown to oscillate with a genotype-specific period, and their protein products have been shown to interact in vitro (4 -7).
These rhythmic profiles have suggested a generalized model for circadian rhythm generation that involves the transcription, translation, and feedback of these clock gene products on their own transcription. The proposed interlocking transcriptional/translational feedback loops are comprised of "negative elements" and "positive elements" which interact to drive overt rhythmicity (4 -7). Briefly, the negative elements in mammals are encoded by the period genes (per1-3), cryptochromes (cry1 and -2), and Rev-Erb␣. These are transcribed in response to the dimerization of the positive elements, which include clock (clock) and the Bmals (Bmal1, 2) and their subsequent binding to E-boxes in the promoter regions of several genes. The negative element transcripts are then translated, oligomerize, and re-enter the nucleus where they inhibit their own and other gene transcription, closing the feedback loop.
In vertebrates, circadian organization also relies on discrete neuroendocrine structures that drive and/or entrain downstream rhythms of biochemistry, physiology, and behavior. These structures include the hypothalamic suprachiasmatic nucleus (SCN), 1 the pineal gland, and the retina. In mammals, the SCN serves as a master pacemaker, responsible for coordinating all downstream oscillators and rhythmic processes. In birds and many other non-mammalian vertebrates, however, the pineal gland, retina, and the avian homologue of the SCN are equally important in regulating overt rhythmicity. The avian pineal gland and retina share several key processes; 1) both are directly photoreceptive, 2) both contain circadian oscillators, and 3) both rhythmically synthesize and release the indoleamine neurohormone melatonin.
At the systems level, the retina functions integrally with the pineal gland and the SCN such that these oscillatory structures feedback and influence each other (20 -23). Under several experimental circumstances loss of the retinal clock in turn disrupts and/or abolishes the clock of the organism or output processes (24 -27). The retina also is responsible for integrating and relaying photic information to the SCN via retinal ganglion cells and the retinohypothalamic tract. This tract serves as the major pathway for entrainment of the SCN in mammals (28,29) and has been an area of great focus in the search for the photopigment(s) existing in retinal ganglion cells that serves as the entraining photoreceptor of the system.
Our previous transcriptome analyses (30) identified candidate molecular components of the avian pineal clock. Here, we determine candidate retinal clock transcripts under light dark and constant darkness conditions to better understand the molecular organization of the avian system. Subsequently, a comparative transcriptome analysis of the two data sets determined candidate molecular machinery regulating the circadian system of chicks within both pacemakers and circadian oscillators.

EXPERIMENTAL PROCEDURES
Experimental Animals-Chicks were obtained from Hyline International (Bryan, TX) and housed under a lighting schedule of 12 h of light and 12 h of darkness (LD 12:12) for 7 days with food and water continuously available. For microarray hybridizations chicks (n ϭ 3/time point/experiment) were sacrificed by decapitation, and their retinas were dissected and pooled. Two separate retinal RNA sampling cycles were performed for array hybridizations. All DD dissections were performed using an infrared viewer. Animals were treated in accordance with Public Health Services guidelines; these procedures have been approved by the Texas A&M University Laboratory Animal Care Committee (AUP #2001-163).
Microarray Production and Hybridization-The cDNA microarrays were prepared as previously described (30) based upon ϳ4500 PCR products from each of the two pineal cDNA libraries constructed, for a total of ϳ9000 ESTs. Briefly, the cDNA in plasmids from each library were PCR-amplified using flanking primers (SK and T7), purified by ethanol precipitation using ammonium acetate, and placed in the wells of 96-well plates at a concentration of 50 g/ml in 3ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate). A GeneMachines OmniGrid microarrayer equipped with 8 Telechem SMP3 pins was used to spot the samples onto poly-L-lysine-coated slides (CEL Associates) in duplicate. Slides were stored desiccated at room temperature until their subsequent use. Before hybridization, the dried spots were hydrated gently over a steaming water bath and snap-dried, and the DNA was UVcross-linked using a Stratalinker (Stratagene).
Total RNA was harvested using Trizol Reagent (Invitrogen) every 4 h for 1 day in a light-dark cycle of 12 h light and 12 h darkness (12:12; lights on 0600 Central Standard Time; lights off 1800 Central Standard Time) and for 1 day in continuous darkness (DD). Times of sampling began 2 h after lights-n (zeitgeber time, ZT2) and continued every 4 h hence (ZT6, ZT10, ZT14, ZT18, ZT22). When lights would have normally turned on, the timer was disabled, and birds were placed in DD. They were then collected again every 4 h and designated circadian time (CT) CT2, CT6, CT10, CT14, CT18, and CT22. This sampling procedure was performed on two separate occasions. A total of four experimental microarray hybridizations were conducted for each time point, two from each respective biological sampling procedure.
The total RNA samples were then amplified to produce aRNA using a MessageAmp Kit (Ambion). Randomly primed fluorescent probes were produced from aRNA (amplified RNA) samples using a Genisphere 3DNA Array 900MPX expression array detection kit. The fluorescent dye on probes derived from the experimental aRNA was Cy5, whereas the dye on control probes was Cy3. In all experiments the control sample was derived from retina harvested at midnight (ZT18) under LD conditions.
Hybridizations and washes were conducted as suggested by Genisphere. The labeled arrays were scanned in an Affymetrix 428 array scanner. The images obtained were subsequently analyzed by GenePix-Pro software (Axon Instruments).
Data Analysis-Data files were then further analyzed by GeneSpring 6.1 (Silicon Genetics). GeneSpring allows the data to be normalized in a variety of ways, allows the assignment of parameters and interpretations, and then allows the data to be filtered to determine differential expression. The current analysis used intensity-dependent LOWESS normalization. Intensity-dependent normalization is just one technique used to eliminate dye-related artifacts in two-color experiments such as this. At each time point the results for each gene were reported as an average obtained from four slides. The data are reported as the normalized ratio of Cy5 (experimental) to Cy3 (control at ZT18). Thus, the data reported for each time point are relative to the abundance of the same gene at midnight (ZT18).
The criteria used to define rhythmic genes were those transcripts with a 2.0-fold or higher amplitude change in mRNA levels over a 24-h period relative to the gene respective mRNA abundance level at ZT18 in addition to exhibiting this oscillation in 3 of 4 hybridizations performed. Analysis at the 1.5-fold level was also performed, but not to the extent of examination as the 2.0-fold data set. Those gene sets that were determined to achieve the 1.5-and 2.0-fold criterions were analyzed statistically by one way analysis of variance, with p Ͻ 0.05 considered significant. Only statistically significant changes will be reported here. However, the entire dataset is available from the authors. 2

RESULTS
Transcriptome Analysis-Of the ϳ9000 cDNAs represented on the microarray, 546 classified transcripts oscillated with 1.5-fold amplitude, and 175 classified transcripts oscillated with 2-fold or greater amplitude in a light-dark cycle (Supplemental Tables 1 and 2). Upon examination of these transcripts under conditions of constant darkness, the number of oscillating transcripts observed was reduced to 383 at 1.5-fold and to only 49 at 2.0-fold or better (Supplemental Table 3 and 4). This percentage of rhythmic transcripts is similar to those found in the chick pineal gland (30) and other experimental analyses of rhythmic processes using cDNA and/or oligonucleotide arrays (31)(32)(33)(34). We restricted our most detailed analyses to those transcripts that varied by 2.0-fold or more.
Functional Categories of Rhythmically Expressed Transcripts-Oscillating retina transcripts in LD and DD were classified according to proposed function. Components of the melatonin generating enzymatic cascade were represented, as were many chick orthologs associated with mammalian circadian clock function. Furthermore, functional analysis revealed transcripts in several broad categories that were also rhythmically expressed. Functional categories were divided into 20 categories: intermediary metabolism (A), development (B), melatonin biosynthesis (C), neuronal associated (D), disease related (E), hormones/growth factors (F), cell cycle/cell death (G), carrier proteins/transport/circulation (H), cytoskeleton/microtubule-associated (I), cell signaling (J), cell adhesion (K), RNA synthesis/stability (L), circadian clock (M), photoreception/phototransductive/visual cycle elements (N), ribosomal proteins/ translation (O), stress response/host defense/chaperone (P), transcription factors (Q), protein modification (R), receptors (S), DNA synthesis/replication/binding (T), and miscellaneous functions that don't readily fit into a single category above (U). The stress response/chaperones and cytoskeletal classes under LD conditions had the highest number of oscillating transcripts, with 28 and 23, respectively. In DD the stress response/ chaperones again had the largest classification, with eight, followed by the cytoskeletal, carriers, phototransduction, and ribosomal clusters, with five members in each. These functional classifications are indicated in Fig. 1, A and B, and also Supplemental Tables 1-4.
Circadian Phase Analysis-Initial analysis of daily and circadian phase relationships revealed rhythmic transcripts that exhibited a diverse set of phases in the LD cycle. However, retinal transcripts rhythmically expressed in DD either accumulated primarily during the day or peaked during the night. These were clustered using a K-means clustering algorithm (GeneSpring) according to the phase of mRNA abundance. The clustering analysis revealed three primary clusters of peaking transcript abundance at CT22-2, CT6, and CT14 (Fig. 2). For each cluster a representative trace is shown indicating the average profile of all oscillating transcripts in each respective cluster. Cluster 1, which composed 20% of the unique DD transcript set, is indicative of transcripts peaking at approximately the onset of subjective dawn CT-22-CT-2; Cluster 2, 49% composed of the unique DD transcripts, at mid-subjective day CT-6; Cluster 3, composed 31% of the unique DD transcripts, at early subjective night CT-14. This distribution of phases is similar to previous findings of other groups (31)(32)(33)(34).
Melatonin Biosynthesis-As has been previously shown, the mRNAs encoding enzymes involved in melatonin biosynthesis were rhythmically expressed on a daily and circadian basis. Under LD conditions the data indicated a rhythm in tryptophan hydroxylase mRNA such that levels were highest in the early night (ZT14), and hydroxyindole-O-methyltransferase was highest at midnight (ZT18). AANAT mRNA exhibited a low amplitude rhythm that was highest during the night as well (Fig. 3A). These patterns corresponded favorably with Northern blot data employing radioactive probes derived from the microarray cDNAs corresponding to each respective gene (Fig. 3B). The rhythm in tryptophan hydroxylase persisted in DD with similar phase angles, whereas hydroxyindole-O-methyltransferase and AANAT mRNA were not significantly rhythmic in DD at the 2.0-fold criterion (Fig. 3A).
Clock Genes-Several mRNAs encoding putative orthologs of genes associated with circadian clock function were identified in the dataset. These include putative negative elements cryptochrome 1 (cry1), cryptochrome 2 (cry2), period 2 (per2), and period 3 (per3) (Fig. 4A). Under both LD and DD conditions cry1 and per3 exhibited rhythmic mRNA profiles, such that peak cry1 mRNA levels occurred during the midday (ZT6) in LD and mid-subjective day (CT6) in DD, and peak per3 mRNA levels occurred during the late (ZT22) night and subjective night (CT22). Although neither cry2 nor per2 mRNA levels were rhythmic in LD, per2 mRNA levels expressed a modest increase at CT10 in DD. These profiles and their phases were confirmed by Northern blot analyses (Fig. 4B). The putative positive elements clock, bmal1, and bmal2, as well as casein kinase 1 epsilon (ck1E) were examined. Analysis of the putative positive elements yielded a low amplitude rhythm in LD for bmal1 and bmal2, peaking in the early night, ZT14, whereas clock was not rhythmically expressed. In addition, no rhythm in ck1E mRNA abundance was observed. In DD, clock, bmal1, and bmal2 were rhythmically expressed, reaching peak mRNA abundance at CT10 -14 (Fig. 4C).
Retinal and Phototransduction Transcripts-Several components of the photoreceptive/phototransduction/visual cycle pathway were examined on the microarray, yielding many rhythmic profiles under LD and DD (Fig. 5). Because the microarray was produced from cDNA libraries derived from chick pineal glands, it is important to point out here that retinaspecific sequences (genes expressed in the retina but not the pineal gland), such as rhodopsin, are not represented here. Nonetheless, several genes associated with visual transduction were analyzed including melanopsin (Opn4) (35) and rod ␥-subunit phosphodiesterase 6 (Pde6) (36) as well as opsins and binding proteins associated with visual cycle function, such as peropsin (Rrh) (37), retinal G-protein-coupled receptor opsin (Rgr) (38), and retinal pigmented epithelium 65 (Rpe65) (39). The mRNA for all of these genes was expressed rhythmically under both LD and DD conditions with the exception of melanopsin. Although melanopsin was expressed predominantly in the late day through night in LD, there was no apparent circadian pattern of expression in DD. Additional potential elements of phototransduction and visual cycle function that were examined but found to be not significantly rhythmic under our criteria include transducin ␥-subunit (40) Tables 2 and 4. retinaldehyde-binding protein (CRALBP) (42), and retinal short chain dehydrogenase/reductase 2 (43).
Retinal Transport-The vertebrate visual cycle process involves at least three different retinal cell-types; that is, the photoreceptors, pigmented epithelial cells, and Muller glia. Thus, the problem of transported retinaldehyde and by-products from one compartment to another and from the periphery into the retina is likely solved by transport proteins. One of these is purpurin (44), whose mRNA is rhythmically expressed in the retina. Purpurin was expressed highly during the early day and late night in LD and early subjective day and late subjective night in DD (Fig. 6A). Another retinal transport molecule is transthyretin, which was found to be rhythmically expressed in the pineal gland (30), was not consistently rhythmic in the retina under constant conditions.
Translational Control and Protein Trafficking/Modification-A number of ribosomal and ribosome-associated proteins were rhythmically expressed in both LD and DD. The 40 S ribosomal proteins S3, S5, S18, and S27 as well as 60 S ribosomal proteins L10 and L18 were expressed predominantly during the day with a 2.0-fold rhythm. Of these, S27, L10, and L18 were expressed rhythmically with a 2.0-fold or higher amplitude such that expression was highest during subjective day in DD (Fig. 6B). In addition, chick elongation factor 2 (EF-2) and ribosomal protein L11 were expressed rhythmically in both LD and DD such that mRNA was highest in late day in LD and late subjective night in DD.
A large number of mRNA species associated with proteolytic proteasomal activity were rhythmically expressed in LD. These included several ubiquitin sequences as well as sequences associated with ubiquitin modification and several protein phosphatases and carboxypeptidase M (Supplemental Table 1). Of these, only carboxypeptidase M persisted in its rhythmicity in DD at the 2.0-fold criterion, peaking at CT14 in the early subjective night. However, most of the ubiquitin-associated sequences persisted in DD at the 1.5-fold criterion, peaking during the subjective day. Intermediary Metabolism-Many mRNA species associated with glycolytic, lactate fermentation, and electron transport processes were expressed rhythmically in LD at the 2.0-fold criterion (Supplemental Table 2). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), phosphoglucose isomerase, and lactate dehydrogenase were all expressed predominantly in the mid-to late day (ZT6 -10). Additionally, cytochrome b and NADH dehydrogenase subunit 4 were expressed highly at midday (ZT-6). ATP synthase ␤-subunit, ATP synthase Hϩ transporter, and NADH dehydrogenase (ubiquinone) 1␣ were expressed predominantly at early day (ZT2), and cytochrome c oxidase subunit II was expressed at early night (ZT14). In DD several of these oscillations failed to persist, although GAPDH and cytochrome c oxidase subunit II were expressed rhythmically at the 2.0-fold criterion, with the former expressed in the late subjective day and the latter expressed in the early to late subjective night (Fig. 6C). However, most of the electron transport system genes were expressed rhythmically in DD at the 1.5-fold criterion (Supplemental Table 3).
Stress Response/Host Defense/Chaperone-Many mRNA species associated with stress response were found to be rhythmic in LD cycles (Supplemental Tables 1 and 2). These included cathepsin and cystatin, cytochrome p450, the aryl hydrocarbon receptor, several heat shock proteins, and several genes associated with histocompatibility. Of these, cytochrome P450, 3 heat shock proteins (HSP8, HSP90, and a 70-kDa cognate heat shock protein ATPase), 3 antigen identification genes (␤2-microglobulin, BM88 antigen, major histocompatibility complex class II associated invariant chain proteins), and the aryl hydrocarbon receptor persisted in DD at the 2.0-fold criterion (Supplemental Table 4). DISCUSSION Both the retina and pineal gland have been identified as circadian pacemakers in the avian circadian system. Of course, both are photoreceptive structures, containing both common and tissue-specific photopigment and phototransductive cascades. For example, both tissues express melanopsin, peropsin, retinal G-protein-coupled receptor opsin, cryptochromes, transducins, and cGMP phosphodiesterase among other phototransductive sequences. Furthermore, photoreceptor cells in both tissues rhythmically synthesize and, at least in some species, release the indoleamine hormone melatonin, such that melatonin is synthesized during the night in LD cycles and during the subjective night in DD in both tissues. This circadian secretion of melatonin accounts for most if not all of the pineal and retina system level functions. However, the relative roles of the pineal and retina clock vary among avian species. For example, in oscine passerine birds such as the house sparrow, Passer domesticus, pinealectomy abolishes circadian patterns of locomotor activity and of physiological functions, such as body temperature (45,46), whereas complete enucleation has little effect on either entrainment or circadian patterns of activity (47). Conversely, pinealectomy of juvenile and adult domestic fowl has little effect on overt behavioral rhythms of activity and feeding (10), whereas enucleation abolishes locomotor rhythms A, retinal transport. Purpurin mRNA is rhythmically expressed in the retina with peak accumulation during the early day and late subjective night in DD. B, ribosomal proteins. S27, L10, and L18 in addition to chick elongation factor 2 (EF-2) and L11 were expressed rhythmically. C, intermediary metabolism. GAPDH and cytochrome c oxidase (COX) subunit II were also expressed rhythmically. (48). Even so, although pinealectomy has no effect on overt behavioral rhythms, the surgery abolishes circadian patterns of electroretinogram a-and b-wave function (10), and administration of exogenous melatonin affects both electroretinogram and visually evoked potentials in chicks and pigeons, Columba livia (10,49,50). In some species the retina and pineal contribute to overt rhythms equally. Neither pinealectomy nor enucleation alone abolishes locomotor rhythms in Japanese quail, Coturnix japonica, and pigeons, but the two surgeries combined abolish behavioral rhythms in both species (20,51). Thus, both the pineal gland and retina are photoreceptive, circadian oscillators that affect behavioral and physiological rhythms via the rhythmic production of melatonin. In chicks, therefore, they represent two sets of circadian pacemakers that reside in the same organism, which one would expect to employ identical molecular mechanisms to generate and regulate their rhythmic functions.
In the chick pineal gland tryptophan hydroxylase and AANAT mRNA are expressed rhythmically, such that levels are high during the night and low during the day in LD, whereas hydroxyindole-O-methyltransferase mRNA is expressed higher during the day than during the night. All three of these rhythms persist in DD (30,56).
In contrast, in the retina, tryptophan hydroxylase, AA-NAT, and hydroxyindole-O-methyltransferase mRNA are regulated on a daily basis such that tryptophan hydroxylase and hydroxyindole-O-methyltransferase are high at early to late night and AANAT is high at midnight (56,57). In DD, hydroxyindole-Omethyltransferase becomes arrhythmic, whereas tryptophan hydroxylase and AANAT mRNA continue to accumulate with similar phases as in the LD cycle, albeit with a lower amplitude (19,56). These patterns of expression are confirmed in the present study in both the microarray data and in Northern blot analyses of the same mRNA extracts (Fig. 3, A and B), demonstrating the validity of this microarray approach and punctuating the observation that, at the mRNA level, melatonin biosynthesis is regulated differentially in the chick pineal and retina.
The mechanism by which rhythmic melatonin biosynthesis is regulated has been assumed to include the transcription, trans-lation, and post-translational activation of chick orthologs of mammalian clock genes. Many of these orthologs have been identified and characterized, at least within the pineal gland (14,15,30,60,61). These include putative negative elements per2, per3, cry1, and cry2 and positive elements clock, bmal1, and bmal2 as well as ck1e. Furthermore, heterologous expression of positive element clock in COS7 cells has been shown to activate the chick AANAT expression and enzyme activity (62). In the present study retinal bmal1 and bmal2 are expressed rhythmically in LD and DD such that expression peaks at early night (ZT14/CT14), whereas clock does not appear rhythmically expressed. Of the putative negative elements in this tissue, only per3 and cry1 are expressed rhythmically, such that per3 is expressed predominantly in the late night/early day (ZT22-2) and late subjective night to early subjective day (CT22-6), whereas cry1 is expressed throughout the day in LD and subjective day in DD (Fig. 4A). The amplitudes of these rhythms are 2-3-fold. None of the other orthologs negative elements was found to be rhythmic, including per2 and cry2, nor was ck1e expressed rhythmically.
The temporal distributions of the mRNA for these clock proteins are interesting in several respects. First, there is no strict antiphase relationship of positive and negative elements that has been observed in Drosophila and mammalian models. Here, both bmals are expressed coincidentally with cry1, and of the rhythmic negative elements, cry1 and per3 are 180°in antiphase. This temporal pattern of clock gene mRNA abundance is similar but not identical to that found in the pineal gland (30). In the pineal gland, bmal1, bmal2, cry1, cry2, per2, and per3 are all expressed rhythmically with high amplitudes, and clock is expressed rhythmically in LD and DD, with a lower amplitude of 1.5-fold, all of which were confirmed by Northern analysis. The exception is the temporal distribution of retinal cry2, which is expressed bimodally in the pineal gland with peak mRNA values primarily during the late subjective day (CT10) and late subjective night (CT22). In the retina cry2 is not rhythmically expressed. Of course, we do not at this stage have any data concerning either the level or subcellular localization of the proteins that these genes encode. The similarities in temporal distribution between the chick retina and pineal gland and the differences between the temporal distributions of these mRNA species in these chick tissues and those of mammals and flies strongly suggest that other rhythmic mechanisms are in place. Furthermore, if clock gene expression regulates melatonin biosynthetic rhythms in these two tissues, one would predict that the downstream processes between clock function and its melatonin output differ between retina and pineal gland.
One might therefore ask, What are the other common elements between rhythmic retinal mRNA species in the present study and pineal rhythms in our previous work? First, in both tissues mRNAs encoding phototransduction and retinaldehyde transport are rhythmically expressed. This is not surprising, since both tissues are photoreceptive. Second, in both tissues, GAPDH and cytochrome c oxidase are expressed rhythmically such that GAPDH is expressed predominantly at CT10, and cytochrome oxidase is expressed highly from CT14 -18. Third, several ribosomal proteins and elongation factor 2 are expressed rhythmically in both tissues, suggesting protein synthesis and modification may play a role in regulating circadian function in these two tissues (Table I).
Several recent studies have pointed to rhythmic processes involving metabolic activity that are more or less independent of clock gene function. First, induction of clock gene rhythms does not necessarily induce metabolic rhythmicity in cultured cells (63). We have found that induction of robust clock gene rhythms in fibroblast lines by serum shock has no effect on glucose utilization in these cells even though rhythms of glucose utilization are accompanied by clock gene expression rhythms in SCN cells. Second, induction of metabolic rhythms in cultured astrocytes does not necessarily result in rhythmic clock gene expression, since glucose utilization rhythms can be induced in chick astrocytes with cycles of melatonin (64), but clock genes are not responsive to the hormone (65). Recent studies by McKnight and co-workers (66) have shown that the redox state in in vitro systems can determine the affinity of negative elements such as the cryptochromes for positive element activation of clock gene promoter regions. Both GAPDH and cytochrome oxidase have the potential of altering redox state via the reduction of NAD ϩ . Thus, it is possible that the rhythms in components of intermediary metabolic pathways we observe here in the retina and in the pineal gland elsewhere (30) may underlie a semi-independent oscillatory mechanism. Current research in our laboratory is focused on this issue.
It is interesting that several mRNA species involved in regulation of translation and protein trafficking are regulated on a circadian basis in the chick retina. Both 40 S and 60 S ribosomal proteins are regulated at the mRNA levels in this tissue in LD and DD as are several sequences associated with proteasomal proteolysis, suggesting a global regulation of many processes by the clock through the regulation of the translational machinery. This is a feature that is shared with the pineal gland as well (30). Although many researchers in the field have focused on rhythmic regulation of mRNA as a central mechanism in biological clock function (including the present study), translational control of clock components and their outputs that are independent of transcriptional regulation have been described in many model systems, including vertebrates. For example, in the filamentous fungus Neurospora crassa, circadian regulation of the clock gene frq and its protein FRQ can be separated by entrainment to long and short photoperiods, suggesting independent regulation of these two components (67). In Gonyaulax, circadian regulation of GAPDH levels and activity, a common clock-regulated protein (68), appears to be independent of mRNA levels (69), and whereas AANAT activity is rhythmically regulated in the sheep pineal gland, no rhythm in mRNA levels can be determined (54). This post-transcriptional regulation of key clock components and their outputs may be the result of circadian regulation of either rhythmic protein synthesis and/or proteolysis. Whether this post-transcriptional regulation is globally regulated or contains features that specifically target some protein regulation is not known at this stage.
In summary, the chick retina, an important component of the biological clock in birds, shares with the pineal gland many rhythmic mRNA species. These include components of the melatonin biosynthetic pathway as well as several clock genes. However, the details of their regulation are both similar and different, raising the reasonable possibility that the clocks in each of these tissues in the same organism are regulated dif-ferentially. Among the common elements between the rhythmic transcriptome in these tissues are key components of intermediary metabolism and protein synthesis and degradation, which should be considered candidates for clock regulation and its output.