Cryptochrome proteins regulate the circadian intracellular behavior and localization of PER2 in mouse suprachiasmatic nucleus neurons

We first assessed the intracellular localization of PER2 (reported as P2V) in adult mouse SCN sections. As expression approached its peak in late circadian day (21), P2V was predominantly nuclear in cells of wild-type (WT) SCN. In contrast, it was significantly more cytoplasmic in SCN of Cry1, 2–deficient (CryDKO) mice (Fig. 1A). Correspondingly, the nucleus:cytoplasm (nuc:cyto) ratio of P2V fluorescence in individual WT SCN cells was ∼2, (>1 indicates predominantly nuclear localization) but was <1 and significantly lower in CryDKO SCN (Fig. 1B). The presence of a single copy of Cry2 was sufficient to sustain nuclear accumulation of P2V that was not significantly different from WT SCN (Fig. 1B). Importantly, this phenotype of increased cytoplasmic localization was observed for untagged WT PER2 protein, imaged through immunostaining (SI Appendix, Fig. S1 A and B↗), and thus not an artifact of the Venus tag. We next used fluorescence recovery after photobleaching (FRAP) in SCN organotypic slices to determine whether CRY proteins regulate PER2 mobility within and between nuclear and cytoplasmic compartments (Fig. 1 C–G and SI Appendix, Fig. S1 C–G↗ and Videos S1–S4↗). Curve fits identified two components with distinct “fast-” or “slow-” moving pools of P2V (PER2-fast; PER2-slow), as defined by their relative diffusion coefficients. A proportion of P2V fluorescence did not recover over the course of the experiments, indicative of immobile or very slowly moving PER2 molecules (Fig. 1 E and F and SI Appendix, Fig. S1 D and E↗). As previously demonstrated (21), even though overall abundance changed, there was no circadian day/night difference in PER2 mobility in WT SCN (SI Appendix, Fig. S1 D–F↗), and so all WT measures were combined to compare mobility in WT and CryDKO SCN slices in a time-independent manner (Fig. 1 C–G). Strikingly, the mobility of PER2-fast was significantly increased in the absence of CRYs, within both nucleus and cytoplasm, as well as for movement into the nucleus. A similar trend for increased nucleus-to-cytoplasm mobility was not significant (Fig. 1 C and D). There were no such changes observed for PER2-slow within or between compartments (Fig. 1 C and D), and the absence of CRYs did not affect the proportion of immobile PER2 in nucleus or cytoplasm (Fig. 1E). In WT SCN slices subject to whole-nucleus bleaching, thus measuring cytoplasm-to-nucleus mobility, fluorescence recovery was low. Consistent with this, >77% of PER2 was immobile (Fig. 1F). In CryDKO SCN slices, however, the equivalent immobile pool was significantly smaller (<54%) than WT (Fig. 1F; n > 3 for each group), and consequently, mobile pools made up a greater proportion of PER2 than in WT slices (Fig. 1G and SI Appendix, Fig. S1G↗). In summary, in WT SCN, PER2 is predominantly nuclear and consists of three mobility pools: fast, slow, and immobile, the last being dominant. In the absence of CRYs, PER2 is located more in the cytoplasm than the nucleus and is, overall, more mobile in terms of both the mobility of individual molecules in the PER2-fast pool and the lower proportion of immobile PER2 molecules. Thus, the intracellular localization and mobility of endogenous PER2 in the SCN are modulated by CRY proteins, likely because by associating with CRYs, PER2 is incorporated into larger, more slowly moving complexes that facilitate its nuclear retention.

To facilitate exploration of the mechanisms whereby CRYs modulate PER2 cellular behaviors and determine their consequences for SCN timekeeping, we designed adeno-associated viral vectors (AAVs) to express fluorescently tagged CRY1 (CRY1::mRuby3 or C1R) and CRY2 (CRY2::EGFP [enhanced green fluorescent protein] or C2G, and CRY2.T2A.mCherry or C2M), driven by minimal versions of their respective promoters (SI Appendix, Fig. S2A↗). Expression was confirmed through confocal imaging (Fig. 2A and SI Appendix, Fig. S2B↗) with transduction efficiency being >90% of cells, including almost all P2V-positive cells (SI Appendix, Fig. S2C↗; n > 3 SCN slices per group). In arrhythmic CryDKO SCN slices, all three AAVs initiated robust, high-amplitude rhythms of PER2::Luc (Fig. 2B and SI Appendix, Fig. S2 D and E↗). The C-terminal EGFP tag did not interfere with CRY2 circadian function and thus we pooled the data for the two CRY2 AAVs (C2G/M) (Fig. 2 C and D and SI Appendix, Fig. S2 D–F↗). The AAV-expressed CRYs were able to reset the period of SCN lacking one or other CRY: C1R expressed in Cry1^−/− SCN lengthened period from ∼22 to ∼26 h, and C2G/M shortened period of Cry2^−/− slices from ∼26 to ∼23 h (Fig. 2C and SI Appendix, Fig. S2 G–I↗; n > 4 per group). Finally, serial transduction of CryDKO SCN with C2G followed by C1R initiated rhythmicity and maintained a period of ∼25 h, confirming the appropriate interactions of the AAV-encoded CRY1 and CRY2 within the clock mechanism (SI Appendix, Fig. S2J↗). The slightly longer than WT period of 24 h is likely due to imprecise control of dose of each CRY protein, a point that we later address. In both CryDKO and single–Cry KO slices, the effects of AAV-CRY1 were more rapidly evident than AAV-CRY2 (Fig. 2D and SI Appendix, Fig. S2 K and L↗; n > 4 for each group). Nevertheless, although the changes in period effected by the two CRYs were comparable (SI Appendix, Fig. S2M↗; n > 4 for each group), AAV-CRY2-initiated SCN rhythms had higher amplitude compared to AAV-CRY1 (SI Appendix, Fig. S2M↗).

Having confirmed the efficacy of the CRY proteins, we next used confocal time-lapse imaging to visualize their behavior in relation to PER2. Only a few days after transduction of CryDKO slices, rhythmic C1R and C2G fluorescence signals were readily detected (Fig. 2E and SI Appendix, Fig. S3 A and B↗). Consistent with its more rapid circadian effects, C1R fluorescence increased faster than did C2G during the first 72 h after transduction (n = 3 per group) (SI Appendix, Fig. S3C↗), suggesting that the pCry1 promoter strength is a significant contributor to the more rapid initiation of PER2::Luc rhythms by AAV-CRY1 (although we do not rule out additional effects of the respective proteins). We then employed dual real-time imaging to explore the cellular-level spatiotemporal relationship between PER2 and CRY1. Robust oscillations of C1R were detected across the SCN 1 to 2 d after transduction, and this was accompanied by rhythmic P2V fluorescence. Surprisingly, in the steady state established after 4 to 5 d, C1R peaked at ∼CT (circadian time)18, that is ∼6 h after the peak of P2V (Fig. 2 E and F and SI Appendix, Fig. S3 D and E↗), indicative of temporally distinct roles for PER2 and CRY1. We then visualized the evolving temporal organization between P2V and C1R in the SCN by plotting P2V fluorescence with respect to C1R fluorescence through time (Fig. 2G). Initially, fluorescence was low for both proteins, but AAV-CRY1 initiated strong P2V oscillations by the second day, and by the third day, C1R was also robustly circadian. The evolving early relationship between P2V and C1R presented a circular trajectory of growing amplitude, illustrating the phase difference between the two oscillations, where the peak of PER2 preceded that of CRY1. To incorporate spatial information, we then used region of interest (ROI) analyses (ROIs provide a proxy for single cells) to ask whether transduction by AAV-C1R was also able to initiate the network-level, spatiotemporal patterning of PER2, stereotypical of WT SCN (Fig. 2H). Before transduction, P2V signal in CryDKO SCN did not exhibit any spatiotemporal organization, nor cellular synchrony. Following transduction with C1R, P2V rhythms within individual cells across the SCN network became well organized in relation to each other, with a characteristic spatiotemporal wave of PER2 (Fig. 2H) and with high levels of synchrony (Fig. 2I), as observed previously for PER2::Luc bioluminescence (11). Furthermore, C1R itself also exhibited spatiotemporal organization, comparable to that of P2V, suggesting that the initiated wave of PER2 is a direct consequence of CRY1 expression and that PER2 and CRY1 influence each other at a network level (SI Appendix, Fig. S3 F–H↗). In summary, real-time imaging demonstrated, first, the efficacy of AAV-CRY1 and AAV-CRY2 vectors. Second, it showed that circadian expression of CRY1 is substantially phase delayed relative to PER2, and third, that expression of CRY1 establishes spatiotemporal organization of PER2 in previously disorganized, arrhythmic CryDKO SCN.

We next asked whether CRY1 and CRY2 could control the intracellular localization of PER2 in SCN slices. As observed in adult SCN brain sections (Fig. 1 A and B), P2V was predominantly nuclear in WT SCN slices (nuc:cyto ratio > 1) and cytoplasmic in CryDKO SCN slices (nuc:cyto ratio <1; Fig. 3 A and B). Transduction with either C1R or C2M relocated PER2 into the nucleus, leading to substantially higher P2V fluorescence intensity (Fig. 3 A and B). This was reflected by a significant increase in the nuc:cyto ratio of P2V, to levels not significantly different from those of WT slices (Fig. 3C). To test whether changes to P2V cellular localization were a cell-autonomous effect of AAV-CRY, we compared P2V nuc:cyto ratio between CRY-positive (CRY+) and rare CRY-negative (CRY−) cells within the same SCN slice, using mRuby3 or mCherry as markers for CRY expression (Fig. 3D and SI Appendix, Fig. S4↗). This showed that the P2V nuc:cyto ratio was significantly higher in CRY+ cells than in the few CRY− cells for both CRY1 and CRY2. Interestingly, C2M-transduced slices showed a trend of greater difference between nuc:cyto ratio of C2M+ cells and C2M− cells compared with the C1R+ and C1R− cells in C1R-transduced slices (Fig. 3D and SI Appendix, Fig. S4C↗). Measuring nuc:cyto ratios of the CRYs themselves showed that C2G was more nuclear than C1R (SI Appendix, Fig. S4B↗), thus mirroring the corresponding trends in P2V localization in the presence of AAV-CRY2.

To acquire a more quantitative assessment of the effects of CRY proteins on PER2 intracellular behavior, we compared C1R fluorescence intensity and P2V nuc:cyto ratio on a single-cell basis. Across conditions, however, there was a binary rather than graded outcome. In the presence of AAV-CRY1, relocalization of PER2 to the nucleus appeared to be complete, regardless of the actual level of C1R fluorescence, suggesting that a threshold had been surpassed (Fig. 3E). To achieve finer, dose-dependent control of CRY1 at lower levels, we used a translational switching (ts) system, generating an AAV in which the C1R coding sequence contains an Amber stop codon substitution (tsC1R) (22) to make expression of AAV-CRY1 dependent on provision of the noncanonical amino acid, alkyne lysine (AlkK) (Fig. 3F and SI Appendix, Fig. S5 A and B↗). A transfer (t)RNA synthetase (with a blue fluorescent tag) capable of charging an orthogonal tRNA with AlkK was delivered in trans by a second AAV. We first confirmed that tsC1R fluorescence and initiation of PER2::Luc rhythms (in previously arrhythmic CryDKO SCN slices) were dependent on AlkK (Fig. 3 G and H and SI Appendix, Fig. S5 B and C↗). PER2::Luc rhythms were lost on removal of AlkK (Fig. 3G). As seen with nonswitchable C1R, the period of SCN rhythms initiated by tsC1R was ∼27 h, with comparable robustness (as assessed by goodness of fit) (SI Appendix, Fig. S5D↗). On application of increasing concentrations of AlkK, mRuby3 fluorescence increased accordingly in a dose-dependent manner (SI Appendix, Fig. S5E↗), although the intensity of fluorescence from tsC1R was lower than that of nonswitchable C1R (SI Appendix, Fig. S5F↗). Robust Cry1-Luc oscillations were initiated by AlkK at concentrations of 0.3 mM and above, whereas 0.1 mM AlkK was ineffective, thereby delineating a threshold between 0.1 and 0.3 mM (Fig. 3I). We found that this threshold represented ∼50% of endogenous CRY1 levels (at CT12) and approximately one-third of peak (CT18) endogenous CRY1 levels, as determined from a CRY1::mRuby3 knock-in mouse (23) (SI Appendix, Fig. S5 G and H↗). Correspondingly, there was a higher nuc:cyto ratio of tsC1R in cells treated with 0.3mM than with 0.1 mM (SI Appendix, Fig. S5I↗).

Having confirmed dose-dependent control of CRY1 expression and SCN function, we next asked whether the degree of PER2 relocalization was dose-dependent and, second, whether this correlated with dose-dependent control of SCN rhythms. Indeed, P2V nuc:cyto ratio significantly increased with AlkK dose (SI Appendix, Fig. S5J↗) and was correlated positively with the intensity of tsC1R fluorescence (Fig. 3J and SI Appendix, Fig. S5K↗). Moreover, this effect was cell autonomous, PER2 being localized to the nucleus of mRuby+ cells expressing tsCRY1 but not in the rare cells lacking tsCRY1 (SI Appendix, Fig. S5L↗). We next focused on the effects of tsCRY1 and PER2 localization on SCN rhythmicity. A characteristic feature of CRY-mediated initiation of SCN rhythms is an initial downward inflection that forms the first trough of the oscillation, resulting in a drop in Cry1-Luc baseline as transcriptional repression by newly translated CRY1 commences. This was observed at all effective concentrations of AlkK (Fig. 3I) and was positively correlated with the increased P2V nuc:cyto ratio (Fig. 3K). Equally, the amplitude of the tsCRY-induced rhythm was positively correlated with the degree of nuclear P2V. It should be noted, however, that relatively low levels of nuclear PER2 were sufficient to initiate and maintain clock function. At all effective concentrations of AlkK, the P2V nuc:cyto ratios were significantly lower than those of positive control (Cry2^+/−) SCN slices, as well as slices initiated with nonswitchable C1R (Fig. 3L and SI Appendix, Fig. S5J↗). This was clearly evident, qualitatively, in the corresponding confocal images, where appreciable amounts of cytoplasmic P2V were observed in SCN with sustained rhythms (Fig. 3M and SI Appendix, Fig. S5M↗). We then compared the P2V nuc:cyto ratio with the nuc:cyto ratio of tsCRY1 at different AlkK concentrations. This revealed positive correlations at AlkK concentrations around the threshold of rhythm initiation (0.1 and 0.3 mM AlkK), consistent with a dose-dependent nuclear shuttling of PER2 by low levels of tsCRY1. At concentrations of AlkK at or above 1 mM, this relationship was lost, indicative of a saturation of the system (SI Appendix, Fig. S5N↗). Thus, by exploiting variable expression of tsCRY1, we have revealed a “tuneable” relationship between the degree of CRY-mediated nuclear localization of PER2 and SCN circadian timekeeping and the unexpected efficacy of relatively low levels of nuclear CRY1 and PER2 in sustaining rhythms.

Having demonstrated the ability of CRY proteins to control PER behavior and SCN oscillation by using a gain-of-function approach, we then sought to test our model by exploring the effect of compromising the ability of CRY to enter the nucleus. To what extent does the nuclear localization of CRY1 influence PER2 behavior and SCN timekeeping? CRY1 has two predicted nuclear localization sequences (NLS): the first is well characterized, located in the Photolyase Homology Region (PHR); the second is poorly characterized and located in the C-terminal domain (CTD). In cell lines, deletion of a coiled-coil (CC) region in addition to the CTD caused complete relocalization of CRY1 to the cytoplasm and full exclusion from the nucleus (24). We therefore generated an AAV expressing C1R where the CC and CTD were deleted: CRY1Δtail::mRuby3 (C1RΔtail; Fig. 4A). We confirmed that the localization of C1RΔtail was indeed mostly cytoplasmic, where its nuc:cyto ratio was <1 in CryDKO and CRY1-deficient SCN slices, significantly lower than that of full-length C1R (Fig. 4C and SI Appendix, Fig. S6A↗). If CRY proteins associate with PER2 to affect its nuclear translocation, C1RΔtail should lose this capacity. Indeed, in the absence of any other CRYs in CryDKO SCN, transduction with C1RΔtail did not alter P2V localization (Fig. 4 D and E). It remained predominantly cytoplasmic, with nuc:cyto ratio < 1, and this ratio was not significantly different between mRuby3+ or mRuby3− cells within the same slices (Fig. 4E). In CryDKO SCN slices, transduction with full-length C1R initiated and sustained PER2::Luc rhythms (Fig. 4 F and G). In contrast, but consistent with C1RΔtail remaining in the cytoplasm and unable to drive PER2 into the nucleus, C1RΔtail failed to sustain or even initiate circadian rhythms in CryDKO SCN (Fig. 4 F and G and SI Appendix, Fig. S6B↗). These results highlight the importance of the CTD in the circadian function of CRY1 in the SCN and are consistent with C1RΔtail being a null allele. We then explored its effect in circadian-competent Cry1-null, Cry2^+/− SCN slices, in which PER2 is localized to the nucleus by CRY2 and which therefore oscillate with a stable but short circadian period. Transduction with C1RΔtail caused a modest but significant change in P2V localization, reducing levels in the nucleus and increasing cytoplasmic signal, with a consequent fall in its nuc:cyto ratio (Fig. 4 H and I). This effect was cell autonomous: evident in mRuby3+ cells but not mRuby3− cells (Fig. 4I). Moreover, the degree of P2V relocalization was dependent on the level of mRuby3+ expression (SI Appendix, Fig. S6C↗). Cells with higher mRuby3 fluorescence intensity tended to have less nuclear P2V compared those with lower mRuby3 signal and nontargeted cells from the same slice (SI Appendix, Fig. S6D↗). This was not reflected by corresponding increases in cytoplasmic P2V signal, perhaps indicative of greater degradation of PER2. Given the presence of WT CRY2 in these SCN, this suggests that C1RΔtail is not a null allele but rather displays dominant-negative behavior, attenuating CRY2-mediated nuclear localization of PER2. We therefore asked whether this altered mobility of PER2, as determined by CRY1Δtail, affected SCN circadian timekeeping. Cry1-null, Cry2^+/− SCN slices exhibited short-period (∼23 h) oscillations of PER2::Luc bioluminescence and, as previously shown, transduction with full-length C1R significantly lengthened period by ∼2.5 h, complementing the CRY1 deficiency but without affecting rhythm amplitude (Fig. 4 J and K). In contrast, C1RΔtail had two different effects on the SCN. First, it significantly shortened period by ∼1 h (Fig. 4K). Second, it reduced the amplitude of the rhythm of Cry1-null SCN slices by ∼40% (Fig. 4K and SI Appendix, Fig. S6E↗). Taken together, this suggests that the C-terminal tail is necessary for CRY1 to fulfill normal clock function, in part through nuclear entry for itself and for PER2 but that CRY1 can still interact with the clockwork, from the cytoplasm, in a way that is dominant to CRY2. In conclusion, our data show that CRY proteins are necessary for SCN circadian function through their role in regulating intracellular behavior of PER2, by ultimately promoting its nuclear retention.