Divergent Roles of Clock Genes in Retinal and Suprachiasmatic Nucleus Circadian Oscillators

Numerous aspects of retinal physiology and function are under the control of an intrinsic retinal circadian clock, including rod disk shedding [1], melatonin release [2], [3], dopamine synthesis [4], [5], [6], gamma-aminobutyric acid (GABA) turnover rate and release [7], extracellular pH [8], electroretinogram (ERG) b-wave amplitude [9], and circadian clock gene expression [10], [11], [12]. The intrinsic retinal clock shapes retinal function into high acuity “day” and high sensitivity “night” states, in part through circadian release of dopamine which reconfigures retinal circuits [13]. In addition, the mammalian retinal clock and its outputs influence trophic processes in the eye including the susceptibility of photoreceptors to degeneration from light damage [14], photoreceptor survival in animal models of retinal degeneration [15], and the degree of refractive errors in primate models of myopia [16].

Mammalian tissues generate molecular circadian rhythms through self-sustaining transcription/translation feedback loops in which two transcription factors CLOCK and BMAL1 periodically drive the expression of three Period genes (Per1-3) and two Cryptochrome genes (Cry1-2), and the resulting PER and CRY protein complexes translocate back into the nucleus to suppress their own transcription [17]. Gene targeting studies have demonstrated that there are tissue-specific differences in the roles of clock genes in circadian rhythms generation. The central neural circadian clock (the suprachiasmatic nuclei, SCN) can more readily compensate for loss of individual clock genes compared to peripheral tissue circadian oscillators (e.g. liver or fibroblast), possibly because of strong inter-neural communication and the expression of Npas2, a Clock paralog [18], [19]. Thus, in the SCN the only single clock gene knockout (KO) able to ablate rhythmicity is Bmal1, whereas in peripheral tissue clocks Bmal1, Per1, Cry1 and Clock are all individually required for rhythms generation [9], [18], [20].

The core clock genes of the SCN are also expressed in the mammalian retina (for review, see [21]), where many show rhythmic variations in constant darkness (DD) [12]. The core clock gene Bmal1 is necessary for circadian rhythms of clock gene expression and of ERG b-wave amplitude in the mouse retina [9], but the dependence of the molecular retinal clock on the expression of other core clock genes has not been tested, nor has the clock-gene dependence of any neural circadian clocks outside the SCN been examined in detail. Here we have tested the clock gene dependence of the amplitude and period of molecular circadian rhythms generation in retinal explants from mice bearing bioluminescent circadian reporter transgenes and knockouts of each of the Period genes (Per1, 2, 3) [22], the Cryoptochrome genes (Cry 1, 2) [23] and the Clock gene [19]. Our findings indicate that the retina as a tissue exhibits a unique clock-gene dependence that is similar to the SCN central neural clock in the clock gene dependence of amplitude, but divergent in the gene dependence of period.

Our present study revealed that the molecular circadian rhythms expressed by the neural retina exhibit distinct dependence on individual core clock genes compared to those expressed by the SCN neural clock. There are striking similarities in the overall pattern of clock gene dependence of rhythmic amplitude in the retina and SCN neural clocks, but with individual gene knockouts having more severe effects in the retina. In retinal explants, Per1, Cry1, and Clock are each necessary for sustained molecular circadian rhythms, whereas in SCN explants they are not, although loss of each of these genes decreases the amplitude of SCN molecular rhythms. In contrast, the influence of individual clock genes on rhythmic period of these two neural oscillators is strikingly divergent, with all three Period genes and Clock having qualitatively different effects on the period of retinal rhythms versus the period of SCN rhythms.

In terms of the amplitude and sustainability of PER2::LUC rhythms, the effects of knocking out individual clock genes were qualitatively similar across the retina and SCN, but varied in degree. For example, knockout of Per1 rendered both retinal and SCN explants essentially arrhythmic for the first week in culture. The stimulus of a media change could then initiate sustained rhythms in SCN, but not in retinal explants. Effects of knocking out Cry1 were similar, with severe disruption of the amplitude and sustainability of both retinal and SCN rhythms upon initial culture, followed by more robust restoration of rhythms in SCN explants only following a media change. The increased severity in the rhythmic phenotype of single clock gene knockout in retina was most evident in Clock knockouts, in which retinal explants were rendered completely arrhythmic, whereas SCN explants showed a small and statistically non-significant reduction in rhythmic power, consistent with the preservation of SCN molecular and behavioral rhythms in Clock knockout mice [19]. Although for most gene knockouts the retinal effects were more severe, knockout of Per2 significantly decreased the rhythmic power of SCN explants (though they still produced sustained rhythms), but did not have a significant effect on retinal rhythms.

Whereas the pattern of gene dependence of rhythmic amplitude was qualitatively similar in retina and SCN, the gene dependence of circadian period was highly divergent across these two neural circadian pacemaker tissues. This was particularly apparent in the Per gene knockouts, in which loss of Per1 alleles shortened retinal period, but lengthened SCN period; loss of Per2 alleles had no effect on retinal period, but shortened SCN period; and loss of Per3 alleles, shortened retinal period, but had no effect on SCN period, consistent with previous reports of period effects of these gene knockouts in the in vitro SCN. Loss of Clock alleles also had opposing effects on period in the two pacemakers, lengthening in retina and shortening in SCN. In contrast, the period effects of Cry gene knockout were similar across the two oscillators, with Cry1 loss shortening and Cry2 loss lengthening, consistent with the previously described behavioral phenotypes [23], [28], [29]. There was no consistent correlation between amplitude effects and period effects either within or across retina and SCN.

Taken together, these results suggest that the Per genes and Clock have differing roles in the organization of the retina and SCN neural circadian clocks, whereas the Cry genes appear to play similar mechanistic roles in the two neural oscillators. Loss of either Per1 or Clock has a greater impact on the amplitude and sustainability of retinal molecular rhythms than on SCN rhythms and loss of each of the individual Per genes or of Clock has divergent effects on the period of the two oscillators. In contrast, loss of the Cry genes, individually or in combination, has similar effects on the amplitude and period of retina and SCN tissue pacemaking, with Cry1 being essential for sustained molecular rhythms in the retina.

Our study indicates that Clock is required for expression of PER2::LUC rhythms in the retinal clock, despite the fact that Npas2, a paralog of Clock that compensates for Clock loss in the SCN oscillator [30], is also expressed in the mouse retina by RT-PCR assay [31]. It was previously reported that other non-neural peripheral oscillators, including the liver and lung, are also dependent on Clock[20]. Therefore, one fundamental difference in clock gene dependence between the SCN clock versus the retina and peripheral tissue clocks is that the SCN clock is less dependent on Clock compared to peripheral clocks, which could be due to a higher levels of expression of Npas2 in SCN versus retina, or more robust intercellular coupling in the SCN clock.

Another difference between the retinal clock and the SCN clock is the role of Per3 in modulation of the circadian period. We found that disruption of Per3 greatly shortened the period of the retinal clock but not the period of the SCN clock. Peripheral tissue explants and fibroblasts from Per3−/− mice also displayed shorter periods than those from WT mice [18]. Thus, Per3 plays a greater role in the molecular clock of the retina and of peripheral tissues than in that of the SCN. The lack of circadian locomotor phenotype for Per3 KO mice has been a puzzle in light of the rich literature on Per3 gene mutations associated with disruption of sleep/wake cycle in humans [32], [33], [34], [35]. Interestingly, a recent report indicates that Per3 KO mice do have altered circadian rest/activity behaviors that are only revealed in a light-dependent manner, and thus may depend on a role for Per3 in the retina, rather than in the SCN [36].

Our data suggest that the retinal clock is more vulnerable to disruption by single gene mutations of Per1, Cry1, Clock and to modification by mutations in Per3, a much wider range of genes than the SCN central clock, in which Bmal1 is the only single clock gene knockout to result in complete arrhythmicity. One possible explanation for these data is that cellular oscillators in the SCN are tightly coupled via inter-neuronal communication and can maintain population synchrony in the tissue in the face of weakened individual cellular rhythms resulting from Per1 or Cry1 knockout [18]. In the retina, rhythmicity has been shown to be independent of many forms of chemical neurotransmission and of gap junctional neural communication [11] and therefore, individual oscillators may become more readily desynchronized if gene mutations weaken or degrade the precision of cellular oscillators. Lack of strong coupling in retinal oscillators may allow retinal rhythms to reset quickly in response to shifts in the light cycle - a function of the retina having direct access to the external light/dark cycle - whereas cellular coupling in the SCN may act to filter retinal input and to buffer this central clock from rapid shifts.

A likely functional consequence of the gene knockouts that render retinas molecularly arrhythmic would be loss of intrinsic physiological rhythms, such as has been shown for loss of the ERG rhythm in retina-specific Bmal1 knockout mice [9] and for ERG rhythms in Cry1/Cry2 double knockout mice [37]. In addition, given the role of retinal circadian rhythms in photoreceptor vulnerability and resilience [14], [15], clock gene mutations that disrupt retinal rhythms could impact retinal degeneration as well. Finally, intrinsic circadian rhythmicity is a widespread feature of sensory neural tissues, including Drosophila chemosensory antennae [38] and the mammalian olfactory bulb [39]. The results presented here suggest the possibility that these neural oscillators in sensory structures may operate via molecular mechanisms that are similar to, but have distinct features from central neural clocks.

A principal limitation of our study is that our current measurements lack cellular-level resolution to address issues such as which cell-types in the retina may be contributing to the rhythms we measure, and whether loss of rhythmic output by retinal explants is due to loss of cellular rhythms, or loss of synchrony among rhythmic cells. Previous work from our laboratory has established that the PER2::LUC bioluminescence rhythms we have measured from retinal explants emanate from all retinal layers, but particularly from the inner nuclear layer in the middle of the retina [11]. Neurons with nuclei in this layer include bipolar cells, horizontal cells, and amacrine cells which have been shown to express the core clock genes [12]. The rhythms measured here are likely due to the contribution of many cell types, but we have not yet established the means to reliably image the bioluminescence rhythms of individual cells within retinal explants. Thus, we also cannot differentiate the contributions of loss of cellular rhythms versus loss of cellular synchrony to the reductions in retinal rhythmic power seen with Per1, Cry1 and Clock gene knockouts, although in the SCN loss of Per1 or Cry1 results in weakened cellular rhythms [18].

As expression of Per1, Cry1 and Clock are each necessary for expression of molecular circadian rhythms by the retina, one possibility is that each of these genes is also necessary for rhythms generation at the cellular level in retinal cells. A similar requirement may exist for the expression of Bmal1[9]. In that case, the expression of these genes in individual retinal cells may be a marker of candidate cellular oscillators among retinal cell types, whereas expression of Per2, Per3 or Cry2, which are not required, would not necessarily identify retinal clock cell candidates.

In summary, we have studied tissue-autonomous real-time gene expression rhythms in retinal and SCN explants from mice with targeted disruption of Per1, Per2, Per3, Cry1, Cry2, or Clock and found both broad similarities and specific distinctions between the retinal and SCN clocks in the roles of these clock genes in the amplitude and period of circadian oscillations. Our results indicate that the Period genes and Clock play similar roles in supporting the amplitude of circadian oscillations in the retinal and SCN clocks, but divergent roles in regulating period in these two neural oscillators, while the Cry genes have similar roles in both dimensions in both neural clocks. This suggests that the roles of the Cry genes are preserved across these two neural circadian pacemaker tissues, while different roles of the Per genes and of Clock likely contribute to the differences in intrinsic period, entrained phase and damping rate between the autonomous retinal and SCN clocks [11]. The retina is unique among all circadian clock tissues in the mammal in that it contains both the capacity for rhythms generation and functional light entrainment pathways for its own rhythms as well as for the SCN. Future studies of this highly-ordered and well-characterized sensory organ and clock may further elucidate the molecular mechanisms and organization of circadian pacemaking.