Rhythmic expression of cryptochrome induces the circadian clock of arrhythmic suprachiasmatic nuclei through arginine vasopressin signaling Mathew D. Edwardsa, Marco Brancaccioa, Johanna E. Cheshama, Elizabeth S. Maywooda, and Michael H. Hastingsa,1 a

Division of Neurobiology, Medical Research Council Laboratory of Molecular Biology, Cambridge CB2 0QH, United Kingdom

Circadian rhythms in mammals are coordinated by the suprachiasmatic nucleus (SCN). SCN neurons define circadian time using transcriptional/ posttranslational feedback loops (TTFL) in which expression of Cryptochrome (Cry) and Period (Per) genes is inhibited by their protein products. Loss of Cry1 and Cry2 stops the SCN clock, whereas individual deletions accelerate and decelerate it, respectively. At the circuit level, neuronal interactions synchronize cellular TTFLs, creating a spatiotemporal wave of gene expression across the SCN that is lost in Cry1/2deficient SCN. To interrogate the properties of CRY proteins required for circadian function, we expressed CRY in SCN of Cry-deficient mice using adeno-associated virus (AAV). Expression of CRY1::EGFP or CRY2:: EGFP under a minimal Cry1 promoter was circadian and rapidly induced PER2-dependent bioluminescence rhythms in previously arrhythmic Cry1/2-deficient SCN, with periods appropriate to each isoform. CRY1::EGFP appropriately lengthened the behavioral period in Cry1-deficient mice. Thus, determination of specific circadian periods reflects properties of the respective proteins, independently of their phase of expression. Phase of CRY1::EGFP expression was critical, however, because constitutive or phase-delayed promoters failed to sustain coherent rhythms. At the circuit level, CRY1::EGFP induced the spatiotemporal wave of PER2 expression in Cry1/2deficient SCN. This was dependent on the neuropeptide arginine vasopressin (AVP) because it was prevented by pharmacological blockade of AVP receptors. Thus, our genetic complementation assay reveals acute, protein-specific induction of cell-autonomous and network-level circadian rhythmicity in SCN never previously exposed to CRY. Specifically, Cry expression must be circadian and appropriately phased to support rhythms, and AVP receptor signaling is required to impose circuit-level circadian function. period

behavioral rhythms under constant conditions whereas Cry2-null mice have longer periods (2). These differences are also seen at the level of the SCN in vitro and are magnified in the presence of the Fbxl3Afh mutation that stabilizes CRY proteins (4). Whether these differences reflect intrinsic properties of the protein, differences in their phase of expression, or other features is unknown. Indeed, it is unclear whether rhythmic expression of the Cry genes is necessary for a fully functioning TTFL. Molecular rhythms can be induced in Cry1/2-null mouse embryonic fibroblasts (MEFs) when Cry is expressed rhythmically under a promoter containing E/E′-box, D-box, and RRE sequences (5, 6). On the other hand, constitutively available, cell-permeant CRY proteins can also induce rhythms in Cry1/2-null fibroblasts (7), and overexpression of CRY1 in WT MEFs does not abolish circadian rhythmicity (8), whereas constitutive overexpression of PER2 does, suggesting that circadian oscillations require rhythmic expression of Per2, but not Cry. The present study aimed to define critical properties of CRY proteins that sustain TTFL function in the SCN, the central circadian pacemaker. In contrast to mere loss-of-function studies, our approach focused on the specific circadian properties that are conferred on the SCN circuit by CRY proteins. We therefore developed a genetic restoration approach using adeno-associated virus (AAV) gene delivery to SCN of CRY-deficient mice in organotypic culture and in vivo. Restorative assays have been performed in fibroblast culture, but such cells lack the intercellular coupling present in the SCN and do not control behavior. Our approach would allow us to investigate the circadian Significance

| bioluminescence | arginine vasopressin | clock | oscillation

Daily cycles of behavior adapt us to the 24-h rhythm of day and night. These rhythms are coordinated by a central clock, the suprachiasmatic nucleus (SCN), in which self-sustaining transcriptional/posttranslational feedback loops define approximately 24-h (circadian) time. Cryptochrome1 (CRY1) and CRY2 are essential clock components, but how they sustain cell-autonomous and circuitlevel circadian timing in the SCN is not clear. To explore this, we developed a genetic complementation assay, using virally expressed, fluorescently tagged CRY proteins to test whether molecular pacemaking can be induced in arrhythmic Cry1/2-deficient SCN. We demonstrate protein-specific and robust induction of molecular pacemaking, the efficacy of which depends on the circadian pattern and phase of Cry expression and functional signaling via neuropeptide (AVP) receptors.

T

he suprachiasmatic nucleus (SCN) is a precise biological clock that coordinates circadian (∼24 h) rhythms in mammals. Composed of ∼10,000 neurons, the SCN sits atop the optic chiasm, through which it receives retinal innervation that entrains its cellular clockwork to the light/dark cycle. In turn, the SCN entrains peripheral circadian clocks distributed across the organism through neural pathways incorporating neuropeptidergic and GABA-ergic signaling (1). At the molecular level, circadian timing in SCN neurons revolves around transcriptional/posttranslational feedback loops (TTFL), in which the positive components BMAL1 and CLOCK activate E/E′box–mediated transcription of the negative components Period (Per1, Per2) and Cryptochrome (Cry1, Cry2). PER/CRY heterodimers then inhibit BMAL1/CLOCK activity and thus prevent their own transcription. Subsequent proteasomal degradation of PER and CRY proteins relieves this inhibition, allowing the molecular cycle to begin again. Genetic studies have established that CRY1 and CRY2 are essential for behavioral and molecular circadian rhythmicity, acting as core components of the TTFL (2, 3). The properties that allow them to fulfill this role are, however, unclear. For example, a surprising feature of the clock is the opposing circadian phenotypes in single Cry knockouts. Although sequence similarity between the CRY proteins is extremely high, Cry1-null mice have short period www.pnas.org/cgi/doi/10.1073/pnas.1519044113

Author contributions: M.D.E., M.B., E.S.M., and M.H.H. designed research; M.D.E., J.E.C., E.S.M., and M.H.H. performed research; M.B. contributed new reagents/analytic tools; M.D.E., J.E.C., and E.S.M. analyzed data; and M.D.E., M.B., E.S.M., and M.H.H. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1

To whom correspondence should be addressed. Email: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1519044113/-/DCSupplemental.

PNAS Early Edition | 1 of 6

NEUROSCIENCE

Edited by Joseph S. Takahashi, Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, TX, and approved January 26, 2016 (received for review September 25, 2015)

roles of CRY in both cell-autonomous and circuit-level pacemaking in behaviorally relevant neurons. Results Induction of Circadian Clock Function in Arrhythmic Cry1/2-Null SCN by AAV-Delivered Cry. Cry1/2-null SCN were transduced with AAVs

encoding the CRY1::EGFP fusion protein controlled by a minimal Cry1 promoter (pCry1) that exhibits circadian activation in the SCN (Fig. 1A) (9). Typically, ∼5 × 109 viral particles (1 μL of ∼5 × 1012 GC/mL) were dropped directly onto each slice 7 d after dissection or after bioluminescence recordings, as previously reported (10). Confocal microscopy confirmed nuclear expression of CRY1 (transduction efficiency of SCN cells, 88.0 ± 7.8%, n = 4) (Fig. 1A). Before AAV transduction, PER2::LUC bioluminescence was arrhythmic; however, ∼2 d after transduction with pCry1-Cry1:: EGFP, rhythms emerged (Fig. 1B) with a long period (26.9 ± 0.2 h, determined after medium change) appropriate to the Cry2-null, Cry1-proficient genotype of the transduced SCN (Fig. 1C). Thus, arrhythmic SCN previously unexposed to CRY proteins can rapidly express circadian molecular rhythms after virally mediated CRY1 expression. Cry1 cDNA was then replaced by Cry2 cDNA and AAVmediated nuclear expression of CRY2::EGFP in Cry1/2-null SCN confirmed by microscopy (83.2 ± 6.1%, n = 4) (Fig. S1A). CRY2:: EGFP induced strong bioluminescence rhythms in initially disorganized Cry1/2-null SCN (Fig. S2A). Importantly, these rhythms had a short period (22.0 ± 0.4 h) consistent with the Cry1-null, Cry2proficient genotype. Both CRY1- and CRY2-induced rhythms had periods significantly different from each other and from WT SCN (24.5 ± 0.1 h, P < 0.0001 for each) (Fig. 1C), although robustness of the oscillations, reported by the cycle-to-cycle variation of the relative amplitude error (RAE), was comparable to WT SCN (Fig. S2B). Thus, not only can pCry1-driven CRY1 and CRY2 induce coherent and stable molecular rhythms in Cry1/2-null SCN, but also can do so with isoform-specific periods, even when driven by the same promoter. Therefore, CRY1 and CRY2 are not simply permissive elements: the individual proteins confer specific temporal properties to the clockwork. To test whether virally expressed CRY can intermesh with ongoing molecular oscillations driven by endogenous CRY, rhythmic Cry1-null SCN were transduced with pCry1-Cry1::EGFP (Fig. 1D). The period of Cry1-null SCN before transduction was

short (22.3 ± 0.4 h), and pCry1-Cry1::EGFP expression stably lengthened period (26.4 ± 0.1 h, P < 0.001) (Fig. 1E) from 3 d after transduction (Fig. S2C). In addition, these rhythms had a lower RAE, as CRY1::EGFP conferred stability to the endogenous clockwork (Fig. S2D). In the converse test, long period Cry2-null SCN were transduced with pCry1-Cry2::EGFP (Fig. S2E), and this decreased period (P < 0.01) (Fig. S2 F and G), although RAE did not decrease following AAV transduction with Cry2 (Fig. S2H). Thus, CRY2 did not enhance the definition of the molecular cycle. Transgenic CRY1 and CRY2 can, therefore, interact with and modulate the period of the endogenous clockwork in single Cry-null SCN. To demonstrate the behavioral relevance of the slice data and test the effect of AAV-mediated CRY expression in vivo, Cry1-null animals received bilateral stereotaxic injections of pCry1-Cry1::EGFP AAV particles into the region of the SCN (Fig. S2I). Before treatment the free-running period of Cry1-null animals over 10 d in continuous darkness (DD) (following entrainment to a 12-h light:12-h dark cycle) was 23.3 ± 0.1 h (as measured by wheel-running activity, n = 8), consistent with the short period observed in Cry1-null SCN (Fig. 1 F and G). Following intracerebral injection of pCry1-Cry1:: EGFP, the free-running period of Cry1-null animals measured 2–4 wk post-administration lengthened to 24.7 ± 0.1 h (P < 0.001) (Fig. 1 F and G), whereas injection of the AAV-EGFP control vector had no effect on period (Fig. S2J). Virally expressed CRY1::EGFP in the SCN can therefore modulate the period of behavioral rhythms in Cry1-null animals consistent with results seen in SCN slices ex vivo, validating the behavioral relevance of effects observed in SCN slice culture. The effect of virally expressed CRY1::EGFP on the TTFL of individual SCN neurons was monitored by time-lapse CCD imaging (Fig. 1H and Movie S1). Nontransduced Cry1/2-null SCN neurons were arrhythmic, in contrast to SCN transduced with either Cry1 or Cry2 (Fig. 1I and Fig. S3A). These induced cellular oscillations had stable periods: the mean SD of the period of individual oscillators was comparable to that of WT SCN (WT = 0.13 ± 0.03 h, CRY1induced = 0.27 ± 0.12 h, P = 0.43 vs. WT; CRY2-induced = 0.17 ± 0.07 h, P = 0.62 vs. WT) (Fig. S3 B and C). Moreover, single oscillators were highly synchronized (Fig. S3D) as quantified by the length of the mean Rayleigh vector (r) of phase dispersal (Fig. S3E). Thus, virally expressed CRY proteins initiated cellular circadian oscillations in Cry1/2-null SCN, and these oscillations were also

Post-

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1519044113

Period (h)

J

****

****

G

****

Period (h)

Period (h)

I

Cells

Pre-

H

F

***

+ pCry1-Cry1

Biolum. (cps)

E

C

Days

D

B

Normalized signal

A

Biolum. (cps)

Fig. 1. Induction of molecular pacemaking in ar28 1500 rhythmic Cry1/2-null SCN. (A) EGFP signal from ITR pCry1 Cry::EGFP WPRE ITR * Cry1/2-null SCN transduced with AAV-pCry1-Cry1:: 26 1000 EGFP. ITR: inverted terminal repeats; WPRE: WHP 24 posttranscriptional response element. (Scale bar: 500 22 10 μm.) (B) PER2::LUC bioluminescence from a Cry1/2-null SCN before (gray) and after (green) 0 20 Merge DAPI CRY1::EGFP AAV: none Cry1 Cry2 0 4 8 12 16 20 24 transduction (arrow). Asterisk indicates medium genotype: WT Cry1/2-null Time (days) change. (C ) The period of Cry1- and Cry2- induced Free-running 0 SCN period rhythms compared with WT (n = 6, 4, 6). All error 26 period 2000 28 bars represent mean ± SEM; ****P < 0.0001 vs. WT. 10 26 25 1500 (D) Bioluminescence from a Cry1-null SCN before 24 (black) and after (green) transduction with pCry124 1000 20 Cry1::EGFP. (E ) Period of Cry1-null (n = 5) SCN be22 500 23 fore and after transduction with Cry AAVs; ***P < 20 30 0 22 0.001 vs. pretreatment, paired t-test. (F ) Actogram 0 2 4 6 8 10 12 14 Pre- PostPre- Post0 12 24 of Cry1-null mouse before and after stereotaxic Time (days) Hours injection of Cry1 AAVs into the SCN. (G) The freePhase 0h + 12 h Cry1/2-null 1 +Cry1 +Cry2 running period of Cry1-null animals in continuous of GFP 1 darkness before and after Cry1 AAV injection. 0 CRY1:: Cry1/2-null + Cry1 ***P < 0.001 vs. pretreatment, paired t test (n = 8). EGFP n.s. 0 (H) Time-lapse images of PER2::LUC bioluminesCT 0 CT 12 CRY2:: cence from a Cry1/2-null SCN before (Top) and after Cry1/2-null + Cry2 EGFP (Bottom) transduction with pCry1-Cry1::EGFP. CT: -1 circadian time. (Scale bar: 100 μm.) (I) Raster plots 0 2 4 6 0 24 48 72 0 24 48 72 0 12 24 of bioluminescence from single cells in H. (J) NorCT Time (h) Time (days) malized PER2::LUC bioluminescence (magenta) and EGFP fluorescence (green/red) over 3 d from Cry1/2-null SCN + pCry1-Cry1::EGFP (Left) or pCry1-Cry2::EGFP (Right). (K ) Peak phase of EGFP fluorescence (CT 12 defined as PER2::LUC peak) of CRY1::EGFP and CRY2::EGFP (n = 4, 3).

K

Edwards et al.

SCN Circadian Clock Requires Expression of Cry1 to Be Circadian and Correctly Phased. The isoform-appropriate period difference of

rhythms induced by Cry1 and Cry2, despite expression from the same Cry1 promoter, led us to question whether the temporal pattern of promoter activity has any bearing on the effects of CRY within the TTFL. We therefore tested the effects of constitutively expressed Cry1 on Cry1/2-null SCN function by replacing the Cry1 promoter with the neuronally specific human synapsin 1 promoter (pSyn1) (Fig. S1B). This is known to be noncircadian in the SCN (10), and constitutive activity was confirmed using time-lapse fluorescence imaging of WT SCN transduced with an AAV encoding an mCherry Cre recombinase fusion protein under the control of pSyn1 (Fig. S4A). Cry1/2-null SCN exhibited transduction with pSyn1-Cry1::EGFP comparable to pCry1-Cry1::EGFP (86.6 ± 10.6%, n = 3). Constitutive expression of Cry1 rapidly suppressed PER2:: LUC expression, accompanied by some weak rhythmicity in previously arrhythmic Cry1/2-null SCN (Fig. 2A). The period (24.3 ± 1.3 h) was not significantly different from that of rhythms induced by pCry1-Cry1::EGFP (P = 0.06) (Fig. 2B), but this reflects the large variance between individual SCN transduced with pSyn1-Cry1:: EGFP. Importantly, the cycle-to-cycle stability of the molecular rhythms within individual SCN was low, as indicated by significantly higher RAEs (pSyn1 = 0.087 ± 0.027 AU, P < 0.05 vs. pCry1) (Figs. 1

1500 1000 500 0

Edwards et al.

16

Biolum. (cps)

4 8 12 Time (days)

**

*

Normalized biolum.

Mean SD (h)

F

28 26 24 22 20

n.s. n.s.

C 0.5 0.4 0.3 0.2 0.1 0.0

* *

y1 n1 al1 y1 n1 al1 Cr Sy m Cr Sy m +p +p +pB +p +p +pB E + pSyn1+ pBmal135 + pCry1 2500 Cry1 Cry1 Cry1 * 2000 30 1500 25 1000 20 + pCry1 500 15 Cry2 0 0 2 4 6 0 2 4 6 8 10 0 2 4 6 8 10 0 4 8 12 16 Time (days) Cycle G +pBmal1-Cry1 H 1.0 + pCry1-Cry1 + pBmal1-Cry1 7 1 6 5 4 0.0 0 3 2 1 -1 0 -1.0 0 24 48 72 -1.0 0.0 1.0 -1.0 0.0 1.0 y1 yn1 al1 r m C S Normalized fluorescence Time (h) +p +p +pB 0

Normalized signal

Period variation (h)

D

B 30

*

RAE (AU)

Biolum. (cps)

2000

Period (h)

A

and 2C). To characterize further the quality of the induced rhythms, the period of individual cycles was measured over several days (Fig. 2D). When either Cry1::EGFP or Cry2::EGFP were driven by the Cry1 promoter, molecular rhythms had a stable peak-to-peak interval representative of the expected long (CRY1) or short (CRY2) period. Molecular rhythms induced by pSyn1-Cry1::EGFP, however, varied greatly from 30 h over successive cycles in individual SCN. Thus, constitutive expression of CRY1::EGFP could induce quasi-circadian cycles in Cry1/2-null SCN, but these were grossly imprecise compared with circadian, pCry1-driven expression. The efficacy of CRY1 in driving the SCN TTFL, therefore, is dependent on its rhythmic availability. To explore further the temporal properties necessary for CRY1 to direct SCN circadian pacemaking, we tested the effects of inappropriately phased expression of Cry1. In vivo the positive TTFL component Bmal1 is expressed ∼8 h after Per and Cry (11). This phase difference was confirmed using the luciferase reporters pCry1-Luc and pBmal1-Luc, transiently transfected into NIH 3T3 fibroblasts (12) (Fig. S4 B and C). The Cry1 promoter was replaced with the Bmal1 promoter (pBmal1) upstream of Cry1::EGFP. This directed nuclear CRY1::EGFP expression in Cry1/2-null SCN as effectively as pCry1 (transduction = 79.1 ± 5.6%, n = 3) (Fig. S1C). Before transduction, the Cry1/2-null SCN were arrhythmic, but pBmal1-driven expression of Cry1 was followed by cycles of PER2::LUC bioluminescence (Fig. 2E) with periods of 24.7 ± 1.5 h (Fig. 2B). Although this was not significantly different from that of rhythms induced by pCry1-Cry1::EGFP (P = 0.11 vs. pBmal1), the peak-to-peak periods varied widely, within and between slices, by as much as 12 h (Fig. 2D). Moreover, cycle-to-cycle signal reproducibility of the molecular rhythms was less consistent (RAE, pBmal1 = 0.294 ± 0.150 AU, P < 0.05 vs. pCry1) (Fig. 2C). Heterochronic expression of CRY1::EGFP was therefore as ineffective as constitutively expressed CRY1::EGFP in its capacity to induce circadian rhythms in Cry1/2-null SCN. To test the effect of phase-delayed expression of CRY1::EGFP on an intact TTFL, rhythmic Cry1-null SCN were transduced with pBmal1-Cry1::EGFP (Fig. S4D). In contrast to the period-lengthening effects of correctly phased CRY1::EGFP, the period did not change significantly (Fig. S4E). The amplitude of the oscillation was, however, markedly decreased, and RAE significantly increased

Fig. 2. Constitutive or phase-delayed expression of Cry1 compromises the SCN molecular clock. (A) Bioluminescence from a Cry1/2-null SCN before (gray) and after (dark green) transduction (arrow) with pSyn1-Cry1::EGFP. Asterisk indicates medium change. (B) Period of molecular rhythms induced by pSyn1Cry1::EGFP and pBmal1-Cry1::EGFP compared with pCry1-Cry1::EGFP (n = 5, 4, 6; data are from Fig. 1). (C) RAE of induced molecular rhythms. (D) The peakto-peak period of molecular rhythms in Cry1/2-null SCN induced by different Cry AAVs. Individual SCN represented by different lines. (E) As in A, with pBmal1-Cry1::EGFP (blue). (F) Mean SD of period of single cells in Cry1/2-null SCN transduced with either pSyn1-Cry1::EGFP or pBmal1-Cry1::EGFP (n = 4, 3 slices; n = 660, 1,548 cells) compared with pCry1-Cry1:: EGFP. (G) Normalized PER2::LUC bioluminescence (magenta) and EGFP fluorescence (blue) traces over 3 d from SCN transduced with pBmal1-Cry1::EGFP. (H) Mean normalized EGFP fluorescence plotted against normalized PER2::LUC bioluminescence through time for Cry1/2-null SCN transduced with pCry1-Cry1:: EGFP or pBmal1-Cry1::EGFP (n = 4, 3). *P < 0.05, **P < 0.01 vs. pCry1. All errors are SEM.

PNAS Early Edition | 3 of 6

NEUROSCIENCE

tightly synchronized. In the absence of any evident physical rearrangement of the SCN, this suggests that the neural architecture to support cellular synchrony preexists in the slice before transduction and thus does not require CRY proteins for its development. To examine circadian expression of CRY1::EFGP and CRY2:: EGFP, SCN were subjected to combined fluorescence and bioluminescence time-lapse imaging. The intensity of both CRY1:: EGFP and CRY2::EGFP signals was rhythmic, peaking shortly after PER2::LUC, conventionally defined as circadian time (CT) 12 (Fig. 1J). Following normalization to circadian time by correcting for period in solar time, no difference was observed in the phase of peak expression between the two CRY proteins (CRY1::EGFP = CT 17.0 ± 0.2 h vs. CRY2::EGFP = CT 17.3 ± 0.4 h, P = 0.52) (Fig. 1K). This is consistent with their expression from the same promoter and confirms that isoform-specific periods are determined by intrinsic properties of the proteins, not by their phase of expression.

Circuit-Level Organization of CRY-Dependent Molecular Rhythms in the SCN. Phase-dependent, rhythmic expression of Cry1 in Cry1/2-null

SCN supports a functioning TTFL within SCN neurons. We then tested whether Cry1 affected higher-order circuit-level behavior— specifically, the spatiotemporal wave of circadian gene expression that progresses dorsomedially to ventrolaterally across the SCN. The dynamics of bioluminescence can be represented by center-of-luminescence analysis (CoL), which tracks the geometric center of mass of the distributed bioluminescence signal across the SCN in X-Y coordinates. In PER2::LUC slices, this presents a specific trajectory that is sustained by, and is reflective of, interneuronal communication across the SCN (Fig. 3A). In contrast, bioluminescence in Cry1/2-null SCN did not exhibit a coherent spatiotemporal wave. The perimeter of CoL (an index of the wave) was significantly greater in WT than in Cry1/2-null slices (110 ± 13 μm vs. 21 ± 7 μm, P < 0.01) (Fig. 3B). Cry1/2-null SCN transduced with pCry1-Cry1::EGFP showed spatiotemporal dynamics comparable to WT SCN, rather than nontransduced Cry1/2-null SCN (Fig. 3A). The CoL perimeter of these induced waves was significantly higher than Cry1/2-null SCN (98 ± 10 μm, P < 0.01) and no different from WT slices (P = 0.56), demonstrating that the AAV-delivered Cry induced and sustained appropriate circuit-level organization of the SCN. CoL analysis was then used to assess network dynamics in Cry1/2-null SCN transduced with other Cry constructs. The CoL of SCN transduced with pCry1Cry2::EGFP was 63 ± 5 μm: greater than nontransduced Cry1/2-null SCN (P < 0.01) but significantly less than WT controls (P < 0.05) (Fig. 3 A and B). Similarly, oscillations induced by pSyn1-Cry1::EGFP had CoL (57 ± 9 μm) intermediate between those of Cry1/2-null (P < 0.05) and WT (P < 0.05) SCN (Fig. 3 A and B). The network 4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1519044113

A

B

Cry1/2-null

WT Y

Y

D L

M V

X

Cry1/2-null + pCry1-Cry1 Y

X

Cry1/2-null + pCry1-Cry2 Y

Perimeter of mean CoL (μm)

n.s. ††

150

*

100

††

50

*† ** n.s.

**

0 pSyn1- pBmal1pCry1AAV: none none Cry1 pCry1Cry2 Cry1 Cry1 genotype: WT Cry1/2-null X X Cry1/2-null Cry1/2-null + pSyn1-Cry1 + pBmal1-Cry1 Y Y

X

X

C

D 150

Y

Perimeter of CoL (μm)

(Fig. S4F). Before transduction the molecular rhythms had very stable peak-to-peak periods, but afterward they varied greatly within and between individual SCN by more than 15 h (Fig. S4G). Thus, incorrectly phased expression of Cry1 perturbed the ongoing rhythms in SCN with an otherwise competent clock. The molecular mechanism underpinning the TTFL is not fully induced when CRY1 is expressed either constitutively or in an incorrect phase. To explore this further, single-cell rhythms were examined by time-lapse imaging of Cry1/2-null SCN transduced with either pSyn1-Cry1::EGFP or pBmal1-Cry1::EGFP. Compared with rhythms induced by pCry1-Cry1::EGFP, individual oscillators in Cry1/ 2-null SCN transduced with pSyn1-Cry1::EGFP or pBmal1-Cry1:: EGFP displayed a much broader range of periods (Fig. S4 H and I) with a significantly higher mean SD of period across multiple slices (pSyn1 = 1.7 ± 0.5 h, P < 0.05 vs. pCry1, pBmal1 = 5.4 ± 0.8 h, P < 0.01 vs. pCry1) (Fig. 2F). In contrast to pCry1-Cry1::EGFP, constitutive or phase-delayed expression of Cry1 did not induce accurate (∼27 h) TTFL oscillations within individual SCN neurons. To examine the cause of this inadequacy, we examined the relative phases of expression of CRY1::EGFP and PER2::LUC using combined time-lapse imaging of Cry1/2-null SCN transduced with pBmal1-Cry1::EGFP. Although CRY1::EGFP fluorescence was rhythmic, it was less well defined than in SCN transduced with pCry1Cry1::EGFP, and a clear phase relationship between PER2::LUC bioluminescence and CRY1::EGFP fluorescence could not be calculated due to the high variability of both (Fig. 2G). To compare the temporal relationship between PER2::LUC bioluminescence and CRY1::EGFP fluorescence formally, the two measures from individual SCN were plotted against each other sequentially through time. In the case of oscillations driven by pCry1-Cry1::EGFP and pCry1-Cry2::EGFP, a clear circular trajectory was observed, reminiscent of a limit-cycle plot (Fig. 2H and Fig. S4J). The relationship between PER2::LUC and CRY1::EGFP in Cry1/2-null SCN transduced with pBmal1-Cry1::EGFP, however, failed to define such a trajectory, emphasizing failure to sustain a stable oscillation (Fig. 2H). Together, these data show that the molecular rhythms induced in Cry1/2-null SCN by constitutive or phasedelayed expression of CRY1::EGFP are less accurate and robust and that the internal structure of the TTFL is compromised, compared with when CRY1 and CRY2 are rhythmically driven in the correct phase by the Cry1 promoter.

100

X

*** **

50

0 -4 0 4 8 Cycle pre-/post- AAV transduction

Fig. 3. Circuit-level organization of molecular rhythms requires appropriately phased, circadian Cry1 expression. (A) Poincaré plots depicting progression of CoL in SCN over three circadian cycles from single SCN, with each cycle represented by a different color. (B) Comparison of the mean perimeter of CoL in different SCN; n = 5, 3, 4, 3, 4, 3 from left to right. *P < 0.05, **P < 0.01 vs. WT, †P < 0.05, ††P < 0.01 vs. Cry1/2-null. (C) Poincaré plot depicting CoL over representative cycles in a Cry1/2-null SCN before (gray) and after (green) transduction with pCry1-Cry1::EGFP. (D) Perimeter of CoL over each circadian day before and after transduction (day 0) (n = 4). **P < 0.01, ***P < 0.001 vs. pretransduction mean at cycle−1 (Bonferroni correction).

dynamics of bioluminescence in Cry1/2-null SCN transduced with the phase-delayed pBmal1-Cry1::EGFP construct were also disorganized, with a CoL of 23 ± 3 μm, not significantly different from that of nontransduced Cry1/2-null SCN (P = 0.76) and significantly lower than WT SCN (P < 0.01) (Fig. 3 A and B). Thus, to induce appropriate circuit-level SCN timekeeping, Cry expression has to be circadian and correctly phased. Finally, to explore the evolution of circuit-level organization of the induced rhythms, Cry1/2-null SCN were imaged for at least 3 d before and 10 cycles after transduction with pCry1-Cry1::EGFP. As expected, Cry1/2-null SCN were arrhythmic before transduction and lacked apparent spatiotemporal dynamics (Fig. 3C). After transduction with pCry1-Cry1::EGFP, however, stereotypical spatiotemporal dynamics emerged (Fig. 3C) with the perimeter of CoL over each circadian day increasing steadily and reaching a plateau after 1 wk (P < 0.001, one-way ANOVA) (Fig. 3D). These results show that the circuit-level behavior, stereotypical of WT SCN, progressively emerges in disorganized arrhythmic Cry1/2-null SCN following CRY1-mediated restoration of cellular TTFLs. Arginine Vasopressin Signaling Is Required for CRY-Dependent Induction of Circadian Timing in SCN. Induction of the spatiotem-

poral wave in Cry1/2-null SCN transduced with pCry1-Cry1::EGFP is indicative of interneuronal signaling. Arginine vasopressin (AVP) is a neuropeptide expressed in the SCN shell region thought to play a role in both SCN synchrony and output (13, 14). To test the potential role of AVP as a mediator of CRY-dependent circuitlevel timekeeping, Cry1/2-null SCN were treated with a mixture of AVP-receptor antagonists (AVPx, OPC-21268 and SSR 149415: V1a, V1b, respectively) (13). AVP-receptor blockade lengthened the period of WT SCN, and this was reversed on washout (Fig. S5 A–C). AVPx did not, however, affect the amplitude ratio or the RAE (Fig. S5 D and E). This lack of disruption of the ongoing circadian rhythm made it possible to test the effect of AVPx on CRY-mediated induction of the molecular clock in Cry1/2-null SCN. We recorded bioluminescence from Cry1/2-null SCN before application of AVPx or vehicle (Veh) in combination with pCry1Cry1::EGFP transduction. As anticipated, molecular rhythms appeared in Veh-treated SCN after ∼2 d (Fig. 4A). In slices treated with AVPx, however, rhythms expressed following transduction were of significantly lower amplitude (amplitude ratio vs. Edwards et al.

B

AVPx

Biolum. (cps)

1500

1500

1000

500 4

0

8 12 16 20 Time (days)

D

0.4

Y

0.3 0.2

*

0.1 0.0

1

Cells

F

X

Veh AVPx + pCry1-Cry1 + AVPx w/o

0 0

10 Time (days)

20

0

Pre AVPx w/o D L M V

4

8 12 16 20 Time (days)

E 120

* **

n.s.

80 40 0

Pr e AV Px w /o

Amplitude ratio

C

0

500

Perimeter of CoL (μm)

1000 0

Veh

AVPx

2000

+ pCry1-Cry1 + AVPx w/o 1 0 0

10 Time (days)

20

Fig. 4. AVP signaling is required for Cry-dependent induction of SCN circadian gene expression. (A) Bioluminescence from a Veh-treated (DMSO) Cry1/2-null SCN transduced with pCry1-Cry1::EGFP before application of AVPx. (B) Bioluminescence from a Cry1/2-null SCN transduced with pCry1-Cry1::EGFP in combination with AVPx application before AVPx washout and treatment with vehicle (Veh). (C) Amplitude of induced rhythms under AVPx or Veh, normalized to pretreatment baseline (n = 3, 4). (D) Poincaré plot depicting CoL in Cry1/2-null SCN before (gray) and after (black) transduction with pCry1-Cry1::EGFP and application of AVPx. Subsequent CoL following AVPx washout (w/o) is shown (green). (E) Comparison of the mean perimeter of CoL between conditions in D (n = 3). *P < 0.05, **P < 0.01, paired t tests. (F) Raster plot (Left) and single-cell trace (Right) from SCN in D.

pretreatment baseline, 0.25 ± 0.03 Veh vs. 0.08 ± 0.04 AVPx, P < 0.05) (Fig. 4 B and C). When Veh-treated controls were treated with AVPx, following induction of rhythms, the ongoing rhythms persisted (Fig. 4A), as in WT SCN treated with AVPx (Fig. S4). In slices treated with AVPx at the time of transduction, washout resulted in the emergence of molecular rhythms comparable to those of Veh-treated SCN. This demonstrates that competent AVP signaling is required for the induction, but not maintenance, of CRYdependent circadian rhythms in Cry1/2-null SCN (Fig. 4B). These results suggest that inhibition of AVP signaling prevents the circuitlevel couplings between SCN neurons that are required to boost rhythm amplitude and initiate the spatiotemporal wave. Time-lapse imaging was performed on Cry1/2-null SCN transduced with pCry1Cry1::EGFP in conjunction with AVPx. Pretreatment, Cry1/2-null SCN had no clear wave and a correspondingly low perimeter of CoL (34 ± 10 μm) (Fig. 4 D and E). Following coapplication of AVPx and pCry1-Cry1::EGFP, the perimeter of CoL was unchanged (34 ± 7 μm, P = 0.95). Following washout of the AVPx, however, a stereotypical wave emerged (102 ± 4 μm, P < 0.01 vs. AVPx) (Fig. 4 D and E). Similarly, single cell rhythms failed to be induced during AVPx treatment but did emerge on washout (Fig. 4F). Together, these data show that CRY requires competent AVP signaling to initiate network-level organization and induce a fully functioning clock in Cry1/ 2-null SCN. Discussion Using a gain-of-function genetic complementation approach, we show that both CRY1 and CRY2 can induce molecular circadian rhythms in arrhythmic Cry-deficient SCN. Thus, despite developing in the absence of CRY proteins, Cry1/2-null SCN were nevertheless able to express cell-autonomous and circuit-level circadian properties following appropriate virally mediated expression of Cry1 and Cry2. The rapidity of this effect (∼2 d) argues against any significant Edwards et al.

need for CRY to direct structural changes in postmitotic SCN neurons and circuits. Adult animals and SCN require CRY proteins for the expression of circadian rhythms of behavior and gene expression. Very early in postnatal development, however, welldefined, short period oscillations of Per gene expression are observed in ∼50% of Cry1/2-null SCN (15, 16), and so in the current study Cry1/2-null SCN were prescreened to confirm arrhythmicity before transduction. Given that cellular circadian rhythms may involve TTFLs acting in combination with intrinsically oscillatory cytosolic pathways (17), the rhythms in Cry1/2-null SCN may arise from such cytosolic mechanisms, sustained by neuropeptidergic and/or metabolic signaling. Early in development, therefore, the Cry1/2-null SCN is formed as a competent circadian oscillator, but CRY proteins later become necessary for expression of this latent capacity. When expressed under the same minimal Cry1 promoter in Cry1/ 2-null SCN, both CRY1 and CRY2 sustained robust molecular rhythms with isoform-specific periods, and in single Cry mutant SCN they lengthened and shortened the circadian period in an isoform-specific manner. The behavioral relevance of this effect was confirmed in Cry1-deficient mice, in which AAV-Cry1-CRY1:: EGFP expression in the SCN lengthened the period of circadian activity rhythms. The effectiveness of both CRY1 and CRY2 in the SCN contrasts with cell-based assays, in which CRY2 is insufficient to induce the molecular clock (6) and in dispersed SCN cultures, where loss of Cry1, but not loss of Cry2, compromises the clockwork; i.e., CRY2 alone cannot sustain rhythms (18). The powerful coupling of SCN neurons, which is not present in fibroblast cell culture and SCN dispersals, may therefore permit self-sustaining CRY2-mediated oscillations. Importantly, the sustained rhythms were induced with long (CRY1: 26.9 h) and short (CRY2: 22.0 h) periods, respectively, comparable to the periods of single Cry knockout SCN (26.2 and 22.3 h) (4). This protein-specific effect was also observed in single Cry knockout SCN, although in these cases the periods “over shot” the wild-type condition, likely due to an AAV copy number effect. This demonstrates that the presence of CRY is not merely permissive: rather than being a passive component of the molecular clock, CRY proteins actually confer temporal information via endogenous properties, such as stability and transcriptional repression, possibly arising from differential posttranslational modifications (19, 20) independently of their phase of expression. In cell-based assays, a delay in CRY-dependent feedback repression mediated by an evening-phased RRE sequence in the first intron of Cry1 (5) is necessary for accurate timekeeping. Limited AAV packaging capacity precluded inclusion of the RRE here, but its absence did not compromise effective control of SCN rhythms. This suggests that RRE-dependent delay is not necessary in the SCN, consistent with the close phasing of Cry1 and Per2 mRNA expression in the SCN, whereas in peripheral tissues and cells Cry1 is delayed due to RRE activity (21). The view that RRE does not influence the Cry1 phase in the SCN is supported by phase mapping of the minimal Cry1 promoter in SCN of Cry1-luc transgenic mice (10). Constitutive or phase-delayed expression of Cry1::EGFP using the Syn1 or Bmal1 promoters, respectively, did induce molecular rhythms, but they were unstable and less coherent than those induced by pCry1-Cry1::EGFP, demonstrating that phase-appropriate Cry1 expression is necessary to generate accurate rhythms within the SCN. When expressed under the pCry1 promoter, the association of CRY1::EGFP (and CRY2::EGFP) and PER2::LUC oscillations tracked a limit cycle, consistent with establishment of a self-sustained oscillator (22). In contrast, phase-delayed expression of Cry1 using pBmal1 collapsed this trajectory. Thus, appropriately phased rhythmic feedback repression by CRY1 is required for full circadian clock function (5, 6). This may appear to run counter to the detection of rhythms in Cry1/2-null fibroblasts treated with cellpermeant CRY proteins (7) and WT MEFs overexpressing CRY1 (8), but the contrast between intrinsically sloppy cell-based rhythms and the robust SCN oscillations makes comparison difficult. Sloppy rhythms were indeed also observed in SCN with constitutive CRY expression, but they did not approach the quality of rhythms supported by circadian expression of CRY. PNAS Early Edition | 5 of 6

NEUROSCIENCE

Veh

2000

Relative biolum.

Biolum. (cps)

A

Limit-cycle behavior defines temporal relationships, but SCN gene expression also follows a spatiotemporal wave that is dependent on interneuronal signaling and is thought to reflect differences in cellintrinsic periods (10, 23). Such dynamics were completely absent in Cry1/2-null SCN, but transduction with pCry1-Cry1::EGFP initiated and sustained waves of bioluminescence similar to those of WT SCN. Thus, the ability to express a spatiotemporal program of gene expression is prespecified in the SCN by CRY-independent mechanisms. The acute presence of CRY1 can induce its expression, as individual cells become competent to oscillate and progressively, over about 1 wk, the wave is established with appropriate phase relationships and an increasing trajectory, as an emergent property. This is consistent with reports based on quantitative spectral clustering of Cry1/2-null (24) where underlying spatial groupings can be identified, even though ensemble circadian behavior is disrupted. Circuit-level organization was not induced in Cry1/2-null SCN transduced with constitutive or phase-delayed Cry1, most likely because timing within the individual neurons was insufficiently stable for ensemble behavior to emerge. It remains to be seen how manipulating the dynamics of gene expression across the SCN, using different promoters to drive Cry expression, relates to circadian behavior in vivo, although proof-of-principle was provided by lengthening of behavioral rhythms in Cry1-null mice injected with AAV-pCry1-CRY1::EGFP. Moreover, recent evidence has shown that the spatiotemporal wave is modified by light exposure, with phase differences between SCN neurons increasing under longer day lengths (25). It is therefore likely that the phase relationship between SCN neurons, which results in the specific spatiotemporal wave of gene expression characteristic of the SCN, encodes timekeeping information in vivo. Network dynamics of gene expression are dependent on intercellular communication. For example, Vipr2-null SCN lacking VIPmediated signaling also lack spatial organization (24). Furthermore, rhythms of bioluminescence in Cry1/2-null SCN are induced by paracrine signaling from a WT SCN graft (15). Of these paracrine-signaling molecules, AVP is thought to play a role in SCN synchronization (26), with genetic loss of V1a and V1b receptors loosening coupling within the SCN, facilitating re-entrainment of molecular and behavioral rhythms (13). Additionally, when the

molecular clock is abolished in AVP-expressing cells, coupling between SCN neurons is weakened, and the period of behavioral rhythms is lengthened (14). Here we show that the restoration of molecular rhythms using transgenic Cry1 is compromised by pharmacological blockade of the AVP receptor and that AVP signaling is therefore required for the circuit-level organization of the SCN clock. Transcription of the AVP gene is dependent on canonical E-box sequences in the AVP promoter, subject to CLOCK/BMAL1 activation and PER/CRY-mediated inhibition (27). This provides a molecular link to explain how recombinant CRY can rhythmically regulate AVP expression and hence direct network organization in Cry1/2-null SCN. Further in vivo experiments in which SCN-specific CRY restoration with or without an AVP-signaling blockade would help determine the precise contribution of AVP signaling from the SCN neurons to overall behavioral rhythms. That is, can expression of Cry in a small population of neurons induce behavioral circadian rhythms in an arrhythmic animal, and if so, is this AVP-mediated? In a similar vein, it would be interesting to investigate the role of specific SCN neurons in rhythm generation. The development of Cre-loxP technology, for example, would allow targeted expression of Cry in SCN subpopulations both ex vivo and in vivo and would help address whether SCN neurons contribute equally and redundantly to circadian timekeeping or whether “pacemaker” cells exist that dictate time to the rest of the network.

1. Hastings MH, Reddy AB, Maywood ES (2003) A clockwork web: Circadian timing in brain and periphery, in health and disease. Nat Rev Neurosci 4(8):649–661. 2. van der Horst GT, et al. (1999) Mammalian Cry1 and Cry2 are essential for maintenance of circadian rhythms. Nature 398(6728):627–630. 3. Kume K, et al. (1999) mCRY1 and mCRY2 are essential components of the negative limb of the circadian clock feedback loop. Cell 98(2):193–205. 4. Anand SN, et al. (2013) Distinct and separable roles for endogenous CRY1 and CRY2 within the circadian molecular clockwork of the suprachiasmatic nucleus, as revealed by the Fbxl3(Afh) mutation. J Neurosci 33(17):7145–7153. 5. Ukai-Tadenuma M, et al. (2011) Delay in feedback repression by cryptochrome 1 is required for circadian clock function. Cell 144(2):268–281. 6. Khan SK, et al. (2012) Identification of a novel cryptochrome differentiating domain required for feedback repression in circadian clock function. J Biol Chem 287(31):25917–25926. 7. Fan Y, Hida A, Anderson DA, Izumo M, Johnson CH (2007) Cycling of CRYPTOCHROME proteins is not necessary for circadian-clock function in mammalian fibroblasts. Curr Biol 17(13):1091–1100. 8. Chen R, et al. (2009) Rhythmic PER abundance defines a critical nodal point for negative feedback within the circadian clock mechanism. Mol Cell 36(3):417–430. 9. Maywood ES, et al. (2013) Analysis of core circadian feedback loop in suprachiasmatic nucleus of mCry1-luc transgenic reporter mouse. Proc Natl Acad Sci USA 110(23):9547–9552. 10. Brancaccio M, Maywood ES, Chesham JE, Loudon AS, Hastings MH (2013) A Gq-Ca2+ axis controls circuit-level encoding of circadian time in the suprachiasmatic nucleus. Neuron 78(4):714–728. 11. Shearman LP, et al. (2000) Interacting molecular loops in the mammalian circadian clock. Science 288(5468):1013–1019. 12. O’Neill JS, Hastings MH (2008) Increased coherence of circadian rhythms in mature fibroblast cultures. J Biol Rhythms 23(6):483–488. 13. Yamaguchi Y, et al. (2013) Mice genetically deficient in vasopressin V1a and V1b receptors are resistant to jet lag. Science 342(6154):85–90. 14. Mieda M, et al. (2015) Cellular clocks in AVP neurons of the SCN are critical for interneuronal coupling regulating circadian behavior rhythm. Neuron 85(5):1103–1116. 15. Maywood ES, Chesham JE, O’Brien JA, Hastings MH (2011) A diversity of paracrine signals sustains molecular circadian cycling in suprachiasmatic nucleus circuits. Proc Natl Acad Sci USA 108(34):14306–14311.

16. Ono D, Honma S, Honma K (2013) Cryptochromes are critical for the development of coherent circadian rhythms in the mouse suprachiasmatic nucleus. Nat Commun 4:1666. 17. Hastings MH, Maywood ES, O’Neill JS (2008) Cellular circadian pacemaking and the role of cytosolic rhythms. Curr Biol 18(17):R805–R815. 18. Liu AC, et al. (2007) Intercellular coupling confers robustness against mutations in the SCN circadian clock network. Cell 129(3):605–616. 19. Gao P, et al. (2013) Phosphorylation of the cryptochrome 1 C-terminal tail regulates circadian period length. J Biol Chem 288(49):35277–35286. 20. Harada Y, Sakai M, Kurabayashi N, Hirota T, Fukada Y (2005) Ser-557-phosphorylated mCRY2 is degraded upon synergistic phosphorylation by glycogen synthase kinase-3 beta. J Biol Chem 280(36):31714–31721. 21. Etchegaray JP, Lee C, Wade PA, Reppert SM (2003) Rhythmic histone acetylation underlies transcription in the mammalian circadian clock. Nature 421(6919):177–182. 22. Leloup JC, Gonze D, Goldbeter A (1999) Limit cycle models for circadian rhythms based on transcriptional regulation in Drosophila and Neurospora. J Biol Rhythms 14(6):433–448. 23. Doi M, et al. (2011) Circadian regulation of intracellular G-protein signalling mediates intercellular synchrony and rhythmicity in the suprachiasmatic nucleus. Nat Commun 2:327. 24. Pauls S, et al. (2014) Differential contributions of intra-cellular and inter-cellular mechanisms to the spatial and temporal architecture of the suprachiasmatic nucleus circadian circuitry in wild-type, cryptochrome-null and vasoactive intestinal peptide receptor 2-null mutant mice. Eur J Neurosci 40(3):2528–2540. 25. Evans JA, Leise TL, Castanon-Cervantes O, Davidson AJ (2013) Dynamic interactions mediated by nonredundant signaling mechanisms couple circadian clock neurons. Neuron 80(4):973–983. 26. Li JD, Burton KJ, Zhang C, Hu SB, Zhou QY (2009) Vasopressin receptor V1a regulates circadian rhythms of locomotor activity and expression of clock-controlled genes in the suprachiasmatic nuclei. Am J Physiol Regul Integr Comp Physiol 296(3): R824–R830. 27. Jin X, et al. (1999) A molecular mechanism regulating rhythmic output from the suprachiasmatic circadian clock. Cell 96(1):57–68. 28. Hastings MH, Reddy AB, McMahon DG, Maywood ES (2005) Analysis of circadian mechanisms in the suprachiasmatic nucleus by transgenesis and biolistic transfection. Methods Enzymol 393:579–592.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1519044113

Materials and Methods Animal work was licensed under the UK Animals (Scientific Procedures) Act 1986 with Local Ethical Review. SCN organotypic slices from 7- to 10-d-old pups were prepared as previously described (28). Bioluminescence emission was recorded using photon multipliers (Hamamatsu), CCD cameras (Hamamatsu), or LV200 bioluminescence imaging systems (Olympus). Graphs were plotted, data analyses were performed, and statistical tests were calculated using Prism 5 (GraphPad). Unless otherwise stated, unpaired t tests were used for statistical comparisons. All recombinant AAVs were manufactured by Penn Vector Core (University of Pennsylvania). ACKNOWLEDGMENTS. We are grateful for excellent technical assistance provided by biomedical staff at the Medical Research Council (MRC) Ares Facility. This work was supported by the MRC of the United Kingdom, and the Electromagnetic Field Biological Research Trust.

Edwards et al.