Novel Features of Cryptochrome-Mediated Photoreception in the Brain Circadian Clock of Drosophila

CRY protein accumulates in clock neurons of dark-reared larvae but is absent in the DN2s

cry gene expression in the larval brain was studied with a cry-gal4 insertion (cry-gal4-39), selected for driving high expression of a UAS-gfp reporter (see Materials and Methods). As previously suggested by in situ hybridization (Emery et al., 2000b), cry was expressed in the larval LNs, which are the precursors to the small LN_vs (Fig. 1A-F). In the dorsal brain, cry expression was observed in the two larval DN1s but not in the DN2s. The selective absence of larval DN2 labeling was observed with two additional independent insertions of the same cry-gal4 construct (data not shown) and thus is probably not an artifact produced by positional effects on transgene expression.

To obtain direct evidence as to the expression pattern of the CRY protein, an anti-peptide antiserum was developed that worked on brain whole mounts (see Materials and Methods). It labeled the cell bodies of the LNs and DN1s, but not the DN2s, in wild-type larvae that were kept in permanent darkness (Fig. 1H,K,L). Staining levels were weak, however, as can be qualitatively appreciated from the comparison with CRY-overexpressing larvae raised under the same conditions (Fig. 1, compare H and I). Exposure of the latter larvae to intense (> 100 lux) white light for 1-2 hr completely abolished labeling (M.P., A.K., and F.R., unpublished data), consistent with the rapid light-induced degradation of CRY reported in head extracts (Emery et al., 1998) or transfected cells (Lin et al., 2001). In confocal microscopy, anti-CRY labeling of axonal and dendritic processes was clearly observed in the larval LNs (Fig. 1K,L). The labeling of neuritic endings relative to cell bodies was quite similar for CRY (Fig. 1H,I,L) and GFP (Fig. 1G,J).

Forced cry expression in the larval DN2s brings their PER oscillations into phase with the other clock neurons

The larval DN2s characteristically express PER in antiphase to the LNs and the DN1s (Kaneko et al., 1997) (Fig. 1A-F). We wondered whether this phase difference might be related to the absence of CRY in the DN2s. Overexpression of CRY protein was thus driven in all larval clock neurons, including the DN2s, with tim-gal4 (as documented in Fig. 1I). The controls had CRY overexpression restricted to either the LNs, using pdf-gal4 as a driver, or to the LNs and DN1s, with cry-gal4. No change was detected in the expression pattern of PER in either type of control (Fig. 2A-D, compare with Fig. 1B,E). In contrast, cry overexpression driven with tim-gal4 resulted in a complete phase reversal of PER cycling in the DN2s without any appreciable effect on the LNs or DN1s (Fig. 2E-G).

The cry gene is expressed in most if not all clock neuronal groups of the adult brain

In the adult brain, expression of a cry-gal4 transgene was clearly reported in the small LN_vs (s-LN_vs) and large LN_vs (l-LN_vs) and was suggested in the LN_ds (Emery et al., 2000b). Using PER colabeling, we demonstrate here that the same construct is indeed expressed in all of the LN_ds(Fig. 3A-C). In addition, there was clear PER and cry-gal4 coexpression in the two DN2s (conversely to the larval ones), in most of the DN1s, and in ∼25% of the DN3s (Fig. 3D-F). All of these groups were observed with two additional cry-gal4 insertions (data not shown). Anti-CRY antibodies labeled no adult clock neurons at ZT8 (in n = 10 hemispheres), although samples were systematically observed in confocal microscopy with tim-gal4-driven GFP to localize clock cells. Only very faint labeling was observed in only one s-LN_v in a single brain at ZT23 (n = 9). However, an average of 1.1 ± 0.1 (n = 10) s-LN_v was labeled at CT6 (i.e., after 18 hr in the dark), and all four s-LN_vs were systematically labeled after 3 d in DD (n = 10; data not shown) (Fig. 3G-I). These results are consistent with a very long half-life for CRY in the s-LN_vs in the dark. Under prolonged dark exposure, at least some neurons in each one of the l-LN_v, LN_d, and DN1 groups were also labeled but none of the DN2s and DN3s (Fig. 3G-L) (data not shown). When flies were reared in DD throughout, CRY immunoreactivity was also found in the dorsal projections of the s-LN_vs (Fig. 3J-L), just as it was in the larval LNs (Fig. 1K,L). In most cells, both nucleus and cytoplasm were similarly labeled. No labeling was observed in cry^b mutant brains reared in DD (data not shown).

Adult DN1s express the photoreceptor differentiation factor GLASS

The circadian clock of gl^60J cry^b double mutants has been reported to be completely blind. Surprisingly, the gl^60J mutation removes not only the visual system but also a subset of PER-expressing DN1 neurons (Helfrich-Förster et al., 2001). Double-labeling experiments showed that the GLASS and PER proteins are coexpressed in the DN1 neurons of wild-type flies (Fig. 4A,D). Colocalization was confirmed by confocal microscopy (data not shown). The number of colabeled cells approximately matched the number of DN1s that are missing in gl^60J brains (Fig. 4B,E; quantified in N). The DN2s were not affected in the mutant. The other two remaining cells (DN1^*) probably correspond to the two larval DN1s (Kaneko et al., 1997), which did not express GLASS and were present in gl^60J larvae as inferred from anti-PER labeling (data not shown). GLASS expression in the adult therefore distinguishes between the DN1s that were already present in larvae and those that appear during metamorphosis. Two additional cell groups, one comprising 40-50 cells and the other 7-8 cells, strongly expressed GLASS in the adult wild-type brain (Fig. 4G,K). None of these extra cells coexpressed PER, although the smaller group was quite close to the LN_ds, with which it was often interspersed (Fig. 4K).

GMR-hid flies are devoid of any of the known glass-dependent, rhodopsin-based photoreceptors, including the HB fiber (see Materials and Methods). However, staining of GMR-hid brains with an anti-PER antiserum showed the normal complement of PER-expressing neurons, including the DN1s (Fig. 4C,F,I,M,N). We did not observe PER or cry-gal4 expression in any of the GLASS-positive cells described in larval brains (Hassan et al., 2000; Malpel et al., 2004). These larval cells formed one small cluster (8.8 ± 0.3 cells; n = 14 hemispheres) and two large clusters (ventral, 30-50 cells; dorsal, 18.9 ± 0.3 cells; data not shown) similar in size and location to those found in the adult.

Absence of the adult DN1s correlates with a lower light sensitivity of the circadian clock

Because the adult DN1 neurons are the only clock cells absent in gl^60J brains and present in GMR-hid brains, we decided to test their putative role in circadian photoreception by comparing the light sensitivity of the two lines. Two different light regimes were used, LD entrainment and constant light. The norpA^p24 mutant, which abolishes rhodopsin-based photoreception (the norpA gene encodes a phospholipase component of the phototransduction cascade) and has normal numbers of DN1s, was also included in most comparisons. At high intensity, the wild-type, GMR-hid, norpA^p24, and gl^60J flies rapidly resynchronized after the LD schedule was delayed by 8 hr (Fig. 5). At a 1000-fold lower intensity, all mutants synchronized to the new schedule slower than the wild-type, but gl^60J flies synchronized much less efficiently than GMR-hid or norpA^p24 flies (Figs. 5, 6). As reported previously (Helfrich-Förster et al., 2001), the strongest gl^60J defect was in the resetting of the morning activity peak (Fig. 6). In contrast, the morning peak of GMR-hid flies synchronized as rapidly as the evening one, and both peaks synchronized faster than in the gl^60J flies. Entrained GMR-hid and gl^60J flies also had very different activity patterns (Fig. 6).

Constant light results in intensity-dependent arrhythmia in wild-type Drosophila (Konopka et al., 1989), whereas cry^b mutants stay strongly rhythmic (Emery et al., 2000a) with a lengthened period (Helfrich-Förster et al., 2001). We observed that gl^60J flies were much less severely affected than wild-type and rather comparable with cry^b (Table 1). Their period was significantly increased, even at the lowest light intensity tested. In contrast, both GMR-hid and norpA^p24 flies appeared at least as sensitive to LL as the wild type (Table 1). Together, these results show that the clock is more sensitive to light in GMR-hid and norpA^p24 than in gl^60J flies and strongly suggest that the adult-specific DN1s contribute to circadian photoreception.

Light responses of visual system-depleted flies depend on cryptochrome activation

Because cry is expressed in the DN1s, we tested whether the difference in light sensitivity between GMR-hid and gl^60J flies depends on cryptochrome. Like gl^60J cry^b flies, the GMR-hid; cry^b strain did not entrain to LD cycles of even the highest amplitude, although the activity pattern of the two strains differed markedly (Fig. 7). We also tested entrainment to red light/dark cycles. Under these conditions, CRY should not be activated at all (Suri et al., 1998; Ahmad et al., 2002). Wild-type flies synchronized to such cycles; however, neither GMR-hid nor gl^60J flies did (data not shown), which is consistent with CRY being the only pathway of entrainment by light for visual system-depleted flies.

The cry^b mutation very efficiently suppresses the arrhythmia of wild-type flies in LL (Emery et al., 2000a) (Table 1). It did the same for all visual system-defective strains (e.g., 93 ± 4% of GMR-hid; cry^b flies were rhythmic in LL; n = 3 experiments at 250-1000 lux). However, the period lengthening of cry^b flies in LL was suppressed in the gl^60J, GMR-hid, and norpA^p24 backgrounds, even at the highest light intensity (Table 2). This period lengthening thus requires a functional visual system.

PER oscillations in the DN1s require PDF-expressing LN_vs

Adult DN1 neurons display PER oscillations for at least 1 d in constant darkness in wild-type flies and in flies ablated for the PDF-expressing LN_vs (Blanchardon et al., 2001). Such PER oscillations have been proposed to play a role in the weak short-period rhythmicity found in mutant strains such as disco or pdf⁰¹ (Helfrich-Förster, 1998; Renn et al., 1999; Blanchardon et al., 2001). To test that hypothesis, we produced pdf⁰¹ gl^60J recombinants, which lack the adult DN1s, and compared them with DN1s-retaining GMR-hid; pdf⁰¹ flies. Both of these strains had rhythms quite similar to those of the pdf⁰¹ mutant (Table 3), suggesting that their rhythmicity does not involve the adult DN1s. Furthermore, oscillations of PER levels dampened rapidly in the DN1s of pdf⁰¹ flies after transfer to DD (Fig. 8B,D,E), whereas robust PER cycling persisted for at least 2.5 d in their s-LN_vs, l-LN_vs, LN_ds, and DN3s (data not shown). In wild-type flies, PER cycling persisted in all neuronal groups for at least 2 d in DD (Fig. 8A,C,E) (data not shown). We therefore conclude that the adult DN1 neurons play no part in the rhythmicity of the pdf⁰¹ flies but rather are the oscillators that are most dependent on PDF expression in the LN_vs.

Unexpected intercellular and intracellular distribution of CRY protein

Use of a novel anti-CRY antiserum showed expression of CRY protein in most clock neurons, with the exception of the larval DN2s and the adult DN2s and DN3s. Lack of cry expression in the larval DN2s is supported by the absence of cry-gal4-driven labeling of these cells and by the remarkable reversal of the phase of their PER cycling when they are forced to express cry (see below). Conversely, cry-gal4 expression in adult DN2s and DN3s strongly suggests that our failure to detect the CRY protein in these neurons is because of its low levels. Indeed, recent in situ hybridization data showed cry expression in several groups of adult dorsal neurons, including the DN1s and DN3s (Zhao et al., 2003). Moreover, the adult DN2s are the only brain neurons in which PER cycling is completely abolished by the cry^b mutation (Helfrich-Förster et al., 2001), suggesting that CRY plays a role in these neurons.

Under LD conditions, CRY levels remain very low in all brain neurons, consistent with light-dependent CRY degradation (Emery et al., 1998; Lin et al., 2001). We could detect CRY reliably in wild-type adult brains only after prolonged dark exposure. In addition, CRY protein accumulates to very different levels between neuronal groups, levels which do not generally correlate with the strength of cry gene expression as inferred from cry-gal4 (Emery et al., 2000b; Zhao et al., 2003; this study) and in situ hybridization data (Zhao et al., 2003). This reveals an unexpected level of cell-specific post-transcriptional control of CRY expression. Only the cells that express CRY already in larvae, four s-LN_vs and two of the DN1s, were reproducibly strongly labeled. This may in part reflect longer CRY accumulation. However, stronger labeling of the s-LN_vs also occurred when flies were transferred to constant darkness only after their eclosion as adults, suggesting higher stability of the CRY protein in these cells. The cry^b mutation differentially affects the amplitude of LD PER cycling in the different neuronal groups of the adult brain (Helfrich-Förster et al., 2001). Variations of CRY levels among brain neurons may therefore reflect functional differences.

Intracellularly, CRY clearly accumulates in both the nucleus and cytoplasm, including, at least in s-LN_vs, axonal and dendritic projections. The presence of CRY in neurites provides another potential level of CRY function and regulation and suggests that it may interact with proteins other than TIM.

Functional significance of the lack of CRY expression in larval DN2s

The developmental change in cry-gal4 expression in the DN2s correlates with the inversion of PER and TIM oscillations, which occurs in these cells during metamorphosis, bringing them into phase with the other brain clock cells (Kaneko et al., 1997, 2000). Driving CRY expression with a tim-gal4 driver conferred a normal phase to PER cycling in the larval DN2s without affecting it in other neurons. In contrast, when CRY overexpression was not extended to the DN2s, it did not alter PER cycling in the central brain, consistent with CRY acting autonomously in the larval DN2s to set their phase. Our results therefore indicate that CRY is sufficient to induce light-dependent degradation of TIM in brain neurons and to synchronize them accordingly. What then sets the (opposite) phase of wild-type larval DN2s? Their accumulation of TIM in the presence of light already suggested that they were intrinsically blind and thus had to rely on signals coming from light-sensitive neurons, which may include the PDF-expressing LNs that project nearby (Kaneko et al., 1997).

In contrast, the phase of PER and TIM cycling in the other larval brain neurons does not depend on CRY (Kaneko et al., 2000; Ivanchenko et al., 2001). This is consistent with these clock cells receiving inputs from the visual system, as shown for the LNs (Malpel et al., 2002).

Cryptochrome-based photoreceptive capabilities of the adult DN1s

Our results indicate that a structure present in GMR-hid flies and absent in gl^60J flies contributes in a CRY-dependent manner to the entrainment of the adult behavioral clock by LD cycles and to its sensitivity to constant light. The absence of the adult DN1s in the gl^60J mutant and the expression of CRY in these cells strongly point to their involvement in this light input pathway. A role for clock cells other than the LN_vs in CRY-dependent light responses was already suggested by targeted rescue experiments (Emery et al., 2000b). In addition to the LN_vs and the adult DN1s, the CRY expression profile suggests additional contributions from at least the LN_ds and the larva-originating DN1s, but their role in circadian photoreception remains unknown.

CRY-mediated photoreception is required for the entrainment of flies devoid of any visual structure, because GMR-hid; cry^b flies are circadianly blind. Conversely, the latter result also shows that CRY-independent pathways are confined to the visual system. Thus, the reported entrainment of norpA; cry^b flies by LD cycles (Stanewsky et al., 1998) must still rely on the visual system via either a residual function of the mutant no-receptor-potential A (NORPA) protein or a distinct phospholipase isozyme (Shortridge et al., 1991; Riesgo-Escovar et al., 1995).

Expression of GLASS in the adult DN1s indicates that the absence of these cells in the gl^60J mutant may be a cell-autonomous effect of the mutation. Indeed, the larval DN1s, which do not express GLASS, are not eliminated by the gl^60J mutation and persist in gl^60J adult brains. In the adult brain, GLASS and CRY coexpression occurs only in adult DN1s, encouraging further speculation as to a specific photoreceptive role for these cells. GLASS, which is required for the differentiation of rhodopsin-expressing photoreceptors, may control the expression in the adult DN1s of a rhodopsin-based CRY-independent photoreceptive pathway. However, the circadian blindness of GMR-hid; cry^b flies appears to exclude a CRY-independent contribution of nonvisual system cells to circadian photoreception.

Why are the DN1s preserved in GMR-hid brains? GLASS-expressing cells similarly survive in GMR-hid larval brains, whereas the larval photoreceptors disappear (Hassan et al., 2000; Malpel et al., 2004). In contrast, all GLASS-expressing clusters are absent from adult gl^60J mutant brains. The latter still express a GLASS-related epitope in ectopic locations, which may be attributable to the aberrant expression of either a truncated GLASS protein or a glass-related gene in gl^60J mutants. Such widespread perturbation of gene expression in gl^60J brains may be relevant to the surprising noncircadian light responses of gl^60J cry^b (Helfrich-Förster et al., 2001) but not GMR-hid; cry^b flies. Alternatively, a cry-independent function of the DN1s may prevent such noncircadian (or masking) effects to occur in GMR-hid; cry^b flies.

How do the DN1s contribute to circadian activity rhythms?

In Drosophila, an intensity-dependent increase of period was reported in LL between 0.1 and 10 lux, above which most individuals became arrhythmic (Konopka et al., 1989). Although no consistent trend in period lengthening with intensity was seen in our two control strains, arrhythmia did appear at approximately the expected intensity (between 0.5 and 5 lux). cry^b mutants remain strongly rhythmic in LL and display longer periods than in DD (Helfrich-Förster et al., 2001; this study). Helfrich-Förster et al. (2001) found that most gl^60J flies were arrhythmic in LL, whereas the rest displayed long periods. We found a majority of rhythmic gl^60J flies in LL with long period rhythms. The rhythmicity of gl^60J flies in LL contrasts with the strong arrhymicity of blind GMR-hid and norpA^p24 flies. Our data suggest that the absence of CRY-expressing adult DN1s may contribute to the persisting rhythms of both cry^b and gl^60J mutants in LL and to their period lengthening. However, the wild-type period of the double mutants gl^60J cry^b, GMR-hid; cry^b, and norpA^p24; cry^b in LL indicates that period lengthening in cry^b flies requires a functional visual system. Interactions between the visual system and CRY-expressing clock cells may modulate the electrical activity of the LNs, which was recently proposed to be an essential element of their molecular clock (Nitabach et al., 2002). The reported projections from the DN1s toward the s-LN_vs, and from the LN_ds to the DN1s and DN2s (Kaneko and Hall, 2000), indeed suggest that all of these PER-expressing cells form a functional network, which may include the visual system.

The strong rhythmicity of gl^60J flies in constant darkness argues against a major role of the adult DN1s in generating behavioral rhythms. PER coding sequences under glass control restored some long-period rhythmicity in a fraction of per⁰ flies (Vosshall and Young, 1995). The observed rescue may reflect PER expression in the adult CRY- and GLASS-expressing DN1s, but our results with the gl^60J pdf⁰¹ double mutant rule out any contribution of the adult DN1s to the weak (short-period) rhythmicity of pdf⁰¹ mutants and strains lacking LN_vs (Wheeler et al., 1993; Helfrich-Förster, 1998; Renn et al., 1999; Blanchardon et al., 2001). Indeed, in pdf⁰¹ brains, PER cycling defects were only observed in the DN1s, suggesting that the loss of molecular oscillations in these neurons may participate in the weakening of rhythmicity induced by the absence of PDF signaling from the LN_vs.