A diversity of paracrine signals sustains molecularcircadian cycling in suprachiasmatic nucleus circuitsElizabeth S. Maywood1, Johanna E. Chesham, John A. O’Brien, and Michael H. Hastings1

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

Edited by Joseph S. Takahashi, Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, TX, and approved June 27, 2011(received for review February 3, 2011)

The suprachiasmatic nucleus (SCN) is the principal circadian pace-maker of mammals, coordinating daily rhythms of behavior andmetabolism. Circadian timekeeping in SCN neurons revolvesaround transcriptional/posttranslational feedback loops, in whichPeriod (Per) and Cryptochrome (Cry) genes are negatively regulatedby their protein products. Recent studies have revealed, however,that these “core loops” also rely upon cytosolic and circuit-levelproperties for sustained oscillation. To characterize interneuronalsignals responsible for robust pacemaking in SCN cells and circuits,we have developed a unique coculture technique using wild-type(WT) “graft” SCN to drive pacemaking (reported by PER2::LUCIFER-ASE bioluminescence) in “host” SCN deficient either in elements ofneuropeptidergic signaling or in elements of the core feedbackloop. We demonstrate that paracrine signaling is sufficient to re-store cellular synchrony and amplitude of pacemaking in SCN cir-cuits lacking vasoactive intestinal peptide (VIP). By using graftswith mutant circadian periods we show that pacemaking in thehost SCN is specified by the genotype of the graft, confirminggraft-derived factors as determinants of the host rhythm. By com-bining pharmacological with genetic manipulations, we show thata hierarchy of neuropeptidergic signals underpins this paracrineregulation, with a preeminent role for VIP augmented by contri-butions from arginine vasopressin (AVP) and gastrin-releasingpeptide (GRP). Finally, we show that interneuronal signaling is suf-ficiently powerful to maintain circadian pacemaking in arrhythmicCry-null SCN, deficient in essential elements of the transcriptionalnegative feedback loops. Thus, a hierarchy of paracrine neuropep-tidergic signals determines cell- and circuit-level circadian pacemak-ing in the SCN.

organotypic slice | VPAC2 | BB2r

The circadian pacemaker of the suprachiasmatic nucleus (SCN)directs, on a daily basis, the most fundamental processes of

physiology and metabolism, sleep and wakefulness (1). The es-tablished molecular model of the SCN oscillator involves inter-locked transcriptional/posttranslational negative feedback loops,centered on rhythmic expression of the transcriptional inhibitorsPeriod (Per) and Cryptochrome (Cry), driven by the positiveactivators Clock and Bmal1. The period of the daily oscillation isdetermined by the transcriptional activity of Clock (1) and by thestability of Per and Cry proteins (2, 3). More recently the model ofSCN neurons as autonomous transcriptional/translational feed-back oscillators has been reappraised. Rhythmic cytosolic sig-naling pathways, particularly cAMP and Ca2+, not only are drivenby the core feedback loops, but also support them in a mutuallydependent fashion (4, 5). Furthermore, within SCN circuits, lossof signaling by the neuropeptide vasoactive intestinal peptide(VIP) through its VPAC2 receptor (a positive regulator of cAMP)both desynchronizes and damps cellular transcriptional pace-making (6, 7). Conversely, intercellular coupling maintains ro-bustness within SCN cells deficient in components of the corefeedback loop (8, 9). Finally, in Bmal1-null SCN, devoid of tran-scriptional activation of the core loop, intercellular coupling of anunidentified nature is sufficient to drive stochastic transcriptional/translational rhythms with quasi-circadian periods (10).

Coherent pacemaking in the SCN is reliant, therefore, uponmutually dependent intra- and intercellular processes. IsolatedSCN neurons are less reliable oscillators (11) and “neuronal net-work properties are integral to normal function of the SCN” (ref. 9,p. 551). Identifying how circadian information is communicatedacross SCN circuits is therefore an important goal for circadianneurobiology. Various mechanisms, including gap junctions, syn-aptic signaling, and GABA-ergic and neuropeptidergic cues (re-viewed in ref. 5) have been proposed as synchronizers, althoughthese factors may act over different temporal and spatial frame-works. For example, the SCN is divided into core and shell sub-regions (12), and both peptidergic and GABA-ergic cues havebeen implicated in core-shell signaling. Analysis of circadiancommunication between SCN neurons is thus best studied in anorganotypic slice preparation where local and regional circuitry ispreserved. We therefore developed a unique coculture procedurein which a circadian deficient “host” SCN carrying a biolumi-nescent reporter (PER2::LUC) receives a second nonreporter“graft” SCN. By varying the genetic condition of host and graft, incombination with pharmacological manipulations, this prepara-tion provides a unique opportunity to examine how both pre-synaptic (graft) and postsynaptic (host) mechanisms coordinateSCN cellular and circuit pacemaking.

ResultsParacrine Signaling Restores Circadian Pacemaking to VIP-DeficientSCN. Wild-type (WT), PER2::LUC SCN slices exhibited stableand precise bioluminescent rhythms for >20 d without mediumchange (Fig. S1A). Although utilization of the luciferin substratecaused a progressive decline in peak amplitude, the precision ofthe oscillation remained high, as evidenced by the low relativeamplitude error (RAE) in Brass analysis (days 1–10, RAE= 0.05 ±0.01; days 11–20, RAE = 0.02 ± 0.002; n = 4, mean ± SEM).In contrast, the bioluminescence rhythm of Vip−/− SCN slicesdamped out after ∼10 d (Fig. 1A) (RAE = 0.14 ± 0.01, n = 27).Placement of a freshly prepared, WT nonreporter graft SCNonto the VIP-null host immediately (<2 d) and reliably (26/27)restored and maintained high-amplitude, persistent (≥10 d)bioluminescence rhythms in the mutant SCNs (Fig. 1 A and B).The restored cycles were significantly (P < 0.01) more precisethan before grafting (RAE postgraft = 0.04 ± 0.01, n = 26).Neonatal and adult SCN slices were equally effective in theirability to restore rhythms in the VIP-null SCN (Fig. S1B). Incontrast, cerebral cortex or hypothalamic paraventricular nu-cleus (PVN) tissues did not restore coherent rhythmic expression

Author contributions: E.S.M. and M.H.H. designed research; E.S.M., J.E.C., J.A.O., andM.H.H. performed research; E.S.M. and M.H.H. analyzed data; and E.S.M. and M.H.H.wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

See Commentary on page 13883.1To whom correspondence may be addressed. E-mail: emaywood@mrc-lmb.cam.ac.uk ormha@mrc-lmb.cam.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1101767108/-/DCSupplemental.

14306–14311 | PNAS | August 23, 2011 | vol. 108 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1101767108

Dow

nloa

ded

by g

uest

on

July

20,

202

1

to the VIP-null host, indicating a specific action of SCN-derivedfactors (Fig. 1B and Fig. S1C).Restoration of circadian gene expression does not, of itself,

prove that the observed rhythm is dependent upon rhythmic cuesfrom the graft. The graft may simply provide a limiting factor that,once replenished, activates the intrinsic pacemaker of the host.To exclude this, VIP-null hosts received grafts from either shortperiod (20 h, CSK1eTau/Tau) or long period (28 h, Fbxl3Afh/Afh)Vip+/+ nonreporter mice. Both types of mutant SCN restoredbioluminescence rhythms to their hosts. Importantly, the periodof the restored rhythms was determined by the genotype-specificperiod of the grafts (Fig. 1C). Thus, Tau mutant grafts drovecoherent short period rhythms whereas the Afh mutants restoredcoherent long period rhythms (RAE mean + SEM, WT graftbefore = 0.13 + 0.02, after = 0.03 + 0.01; Tau graft before =0.15 + 0.03, after = 0.06 + 0.01; Afh graft before = 0.19 + 0.04,after = 0.05 + 0.01). Genotypically specific restoration of periodconfirms that rhythmic signals emanating from the graft conferspecific circadian information to the host SCN.In WT slices, CCD imaging of individual neurons showed

that bioluminescence rhythms were highly synchronized, as evi-denced by the mean vector of Rayleigh plots (0.96 ± 0.01, n = 3)(Fig. 1D). In contrast, damping of the rhythm in VIP-null hostSCN arose from both a decline in the amplitude of peak bio-luminescence in cells and a progressive loss of synchrony betweencells, evidenced as a reduction in the Rayleigh mean vector (Fig.1E andMovie S1). Following addition of the graft, cellular PER2::LUC expression levels were increased and the oscillations of in-dividual neurons were rapidly (within two cycles) brought backinto synchrony (Movie S2), such that synchrony among graftedVIP-null SCN was comparable to that of WT slices (mean vectorpregraft, 0.40 ± 0.20; 3 d postgraft, 0.99 ± 0.01, n = 3) (Fig. 1F).The rapid (<6 h) activation of bioluminescence and neuronal

resynchronization (evident by 24 h and complete within 48 h)occurred well before precise synaptic communication could beestablished between graft and host. The spatial orientation of thegraft and host did not compromise effective communication ofcircadian time. Placing the mutant SCN above the wild-type SCN

(Fig. S2 A and B) or rotating the dorso-ventral alignment be-tween graft and host did not affect the restoration of rhythmicity.SCN grafts deliberately aligned to the host (i.e., ventral to ven-tral) or 180° out of alignment generated rhythms with RAEs of0.03 ± 0.01 and 0.04 ± 0.01, respectively (RAEs of hosts beforegrafting: 0.11 ± 0.02 and 0.11 ± 0.01, respectively, n = 6 and 6).There was, therefore, no evident requirement for point-to-pointcommunication for restoration of host rhythms, and a graft canreadily drive a target ∼500 μm, possibly 1 mm, distant. To un-derstand better the neural basis of circadian communication,seven cocultures in which circadian bioluminescence rhythmshad been restored for >10 d were reconstructed by confocalmicroscopic imaging. This process confirmed the absence ofVIP-immunoreactive (−ir) cell bodies from the host SCN, whichwas identified by arginine vasopressin (AVP)-ir cells in spatialregister with the bioluminescence signal (Fig. 2A and Fig. S3 A andB). VIP-ir and AVP-ir cell bodies identified the graft SCN, andVIP-ir varicosities could be seen sporadically across host tissue.The restorative effects of the grafts were rapid and inde-

pendent of anatomical position, indicating that graft-derivedVIP may act in a paracrine fashion before synaptic communi-cation could be established. To demonstrate definitively a non-synaptic, paracrine effect, a semipermeable molecular weightcutoff (MWCO) membrane was placed onto the Vip−/− hostimmediately before coculture. This action did not affect mea-surement of bioluminescence emission (Fig. S3C). The molecu-lar mass of VIP is 3.3 kDa and with a 10-kDa cutoff membranebetween SCN, the graft rapidly reinstated circadian biolumi-nescence rhythms in the host (Fig. 2B). When the graft andmembrane were inverted after 4 d, to bring the slices into im-mediate contact, the amplitude and coherence of the biolumi-nescence rhythm were increased further [RAE (mean ± SEM)before = 0.12 ± 0.03, after = 0.05 ± 0.02, n= 3]. Thus, paracrinesignals were effective in restoring circadian function to the VIP-null SCN. When VIP-null host and WT graft were separated bya 2-kDa MWCO for 3 d, the graft failed to reinstate rhythms inthe host (RAE before = 0.17 ± 0.03; after, no detectable rhythm;n = 3). When the graft and membrane were inverted to allow

Fig. 1. Restoration of SCN circadian gene expression bycoculture. (A) Representative bioluminescence rhythmfrom VIP-null SCN grafted with nonreporter WT SCN after10 d. Note damping followed by immediate restoration ofrhythms by graft. (B) Relative amplitude (percentage ofinitial peak value) of successive peaks of circadian bio-luminescence of VIP-null SCN recorded for 10 d and thengiven grafts of SCN, PVN, or cerebral cortex (plotted asmean ± SEM, n ≥ 5). (C) Period of bioluminescence rhythmof VIP-null SCN (mean ± SEM) grafted with WT, Tau, or Afhmutant SCN (n ≥ 4). **P = 0.01 vs. WT treatment (ANOVA,F = 163.8). (D) Representative raster plot of circadian geneexpression in 50 cells from WT SCN with accompanyingRaleigh plot. (E and F) Representative raster plot from VIP-null SCN before (E) and after (F) grafting with WT SCN.Progressive loss of synchrony and its restoration arereflected in Rayleigh plots on days −9, −1, and +3 relativeto grafting.

Maywood et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14307

NEU

ROSC

IENCE

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

July

20,

202

1

unimpeded graft-to-host communication, the host rapidly resumedcoherent rhythmicity (<2 d, RAE = 0.06 ± 0.02) (Fig. 2B). Thisresult confirms that a nonsynaptic, paracrine factor(s) between∼2 and 10 kDa, most likely VIP, is responsible for rapid (<2 d)activation of the VIP-deficient SCN. To confirm the essentialrole of VIP, VIP-null hosts received VIP-null SCN grafts. In theabsence of VIP from the graft, circadian bioluminescencerhythms were not restored in the host SCN (Fig. S4 A and B).

Restoration of Circadian Pacemaking in VIP Receptor-Deficient SCN.To test whether signaling pathways other than VIP can confercircadian organization to SCN circuits, the coculture techniquewas applied to PER2::LUC slices lacking the VPAC2 receptorfor VIP (Vip2r−/−). As with VIP-null slices, the aggregate bio-luminescence rhythm of VPAC2-null SCN damped rapidly (Fig.3 A and B and Movie S3) due to a decline in cellular emissionand phase dispersal of the individual cellular oscillators (Fig. 3C)(RAE = 0.19 ± 0.02, n = 35). The initial response to a WT graftwas less pronounced than in VIP-null slices, and it took severaldays for a bioluminescence rhythm to develop, reaching peakamplitude after ∼7 d (Fig. 3B, Fig. S5A, and Movie S4) (RAE =0.05 ± 0.01, n = 30/35 grafts measurable). The resynchronizationof neurons within the host SCN was evident in the Rayleighmean vector (Fig. 3C and Fig. S5B).The response of VPAC2-null slices differed from that of VIP-

null in another important way. Whereas Vip−/− SCN could ex-press rhythms over a wide range of periods when driven by cir-cadian mutant grafts, VPAC2-null SCN were unable to respondeffectively to mutant SCN (Tau n = 5/7, Afh 4/5). Hence, whenrhythms were reestablished, the period of the restored rhythmwas essentially WT and not significantly different (ANOVA: F =2.6, not significant) between the grafts of different genotypes(Fig. S5C); i.e., the period of the host did not reflect the geno-type of the graft. Moreover, where the rhythms were detected inVip2r−/− SCN with circadian mutant grafts, they were highlyvariable in their interpeak period and their quality (RAE) waslow. Whereas both Tau and Afh mutant SCN significantly re-duced the RAE error of VIP-null slices, this was not the case forVPAC2-null slices (Fig. S5 D and E). Only WT grafts were ableto reduce RAE measures in VPAC2-null slices. Together, thedata with VPAC2-null hosts support several conclusions. First,

factors other than VIP/VPAC2 signaling can reliably and sus-tainably synchronize SCN cellular pacemakers. Second, suchfactors are less effective as mediators between circuits that donot share a common intrinsic period. Third, interneuronal sig-naling via VIP/VPAC2 appears to make a preeminent contri-bution to circadian coordination in the SCN and may havea particular role in setting pacemaker period.

A Hierarchy of Factors Drives SCN Paracrine Circadian Coordination.Having established that paracrine factors other than VIP canrestore bioluminescence rhythms, we tested specifically the rolesof two other SCN neuropeptides, gastrin-releasing peptide(GRP) and AVP. GRP receptors (BB2r) are expressed in thedorsal SCN but, compared with vehicle-treated slices, adminis-tration of the BB2r antagonist PD176252 (5 μM; n = 3) had noeffect on the rate of damping or the amplitude of rhythmicityover 10 d (Fig. S6). BB2r signaling is not, therefore, essential tothe WT SCN clockwork. Equally, administration of 5 μMPD176252 to Vip−/− SCN slices at the time of grafting with WTSCN had no effect on the ability of the graft to drive rhythms inthe host (Fig. 4A). In contrast, concomitant administration of theBB2r antagonist at the time of grafting onto Vip2r−/− SCN slicessignificantly suppressed the induced rhythms (Fig. 4 C and E).Although the graft appeared to be initially effective, after aboutthree cycles the amplitude of the restored bioluminescencerhythms decreased in comparison with vehicle-treated, graftedVPAC2-null slices. Thus, in the absence of VIP-mediated sig-naling, GRP/BB2r becomes a limiting component of the para-crine signals driving the SCN pacemaker.

Fig. 2. Paracrine signaling of circadian cues between SCN. (A) Confocalreconstruction of representative VIP-null SCN with WT graft. (Left) Lowpower (10×) phase and bioluminescence images identifying boxed areasexamined at higher power. (Scale bar, 1 mm.) (Right) Confocal images (60×)for AVP- and VIP-ir in graft and host tissues. (Scale bar, 50 μm.) (B) Repre-sentative bioluminescence rhythm from VIP-null SCN grafted after 10 d withWT SCN separated by 2-kDa (red) or 10-kDa (blue) cutoff membranes. After3 d the grafts were placed directly on the host.

Fig. 3. Restoration of SCN circadian gene expression by signals other thanVIP. (A) Representative bioluminescence emission from VPAC2-null SCNgrafted with nonreporter WT SCN after 10 d. Note damping of rhythm be-fore grafting, followed by progressive restoration of rhythm. (B) Relativeamplitude (percentage of initial peak value) of successive peaks of circadianbioluminescence of WT SCN (green, n = 4) recorded for 20 d (mean ± SEM) orof VPAC2-null (red, n = 6) or VIP-null (blue, n = 6) SCN recorded for 10 d andthen given grafts of WT SCN. (C) Representative raster plot of circadianbioluminescence in 50 cells from VPAC2-null SCN with accompanying Raleighplots from days −9, −1, +3, and +6 relative to grafting. Note progressive lossof synchrony and its restoration.

14308 | www.pnas.org/cgi/doi/10.1073/pnas.1101767108 Maywood et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

Both AVP1a and -1b receptors are expressed in the SCN, andso a combination of AVP1a (SR49059; 10 nM) and AVP1b re-ceptor antagonists (SSR149415; 10 nM) was applied to WT slices(n = 3). This action had no effect on the rate of damping or theamplitude of rhythms compared with vehicle (Fig. S6). WhenVip−/− SCN received both AVP receptor antagonists at the timeof grafting, the graft initially restored bioluminescence levels butthe amplitude of the restored cycles decreased over time (Fig.4B). In the absence of VIP, therefore, AVP-mediated signalingbecame an essential feature of the circuit. The contribution ofAVP-mediated signals was further emphasized in Vip2r−/− SCNslices, where treatment with the AVP antagonists completelyprevented the restoration of bioluminescence rhythms (Fig. 4D and F). Thus, synchronized cellular pacemaking in the SCN isprimarily dependent upon VIP-ergic signaling. Nevertheless, inits absence, cues delivered by AVP and to a lesser degree GRPcan contribute.

Cellular Pacemaking in Cry1/Cry2-Deficient SCN Slices and ItsCoordination by Paracrine Signaling. Thus, paracrine signalingcan drive circadian pacemaking in SCN circuits defective in VIPsignaling. In such slices, however, the apparatus of the corecellular feedback loops remains intact. It has been suggested thatinterneuronal, circuit-level signaling can defend the SCN clock-work from perturbation of the core feedback loop arising fromdeletion of Cry1 or Per1 genes (9). To test whether paracrinesignals are sufficiently powerful to sustain molecular pacemakingin the absence of multiple core feedback elements, Cry1−/− andCry2−/− double-null slices were used as hosts. At the behaviorallevel, Cry-null, Per2Luc mice became arrhythmic immediately on

transfer to continuous dim red light (Fig. S7A). Consistent withthis result, a number of Cry-null SCN slices recorded by photo-multiplier (PMT) failed to exhibit circadian bioluminescencerhythms in culture (Fig. S7B). Remarkably, and contrary toexpectations however, the majority (23/39, 59%) of Cry-null sli-ces continued to oscillate for at least four cycles and somecontinued for up to eight cycles (Fig. 5A). The period of theserhythmic Cry-null slices (19.5 ± 0.9 h, n = 23) was significantly(P < 0.01) shorter than that of WT and single mutant Cry1- andCry2-null SCN (24.2 ± 0.1 h, 22.58 ± 0.09 h, and 26.16 ± 0.12 h,respectively, n = 5–8). The rhythms were also less well defined,with a higher relative amplitude error (RAE: WT, 0.04 ± 0.01;Cry double-null, 0.20 ± 0.02) and lower overall amplitude. CCDimaging revealed that individual cells in nonrhythmic, Cry-nullSCN slices continued to exhibit bioluminescence rhythms (Fig.5B and Movie S5). The frequency of rhythmic cells was lowerand more variable than in WT slices (100% vs. 67.6 ± 18.1%, 50cells analyzed in three and seven SCN, respectively) and theirrhythms were of lower amplitude (1,211 ± 276 vs. 175 ± 51) andhigher RAE (0.08 ± 0.01 vs. 0.47 ± 0.08). Nevertheless, thesedata show that in the absence of Cry-mediated transcriptionalfeedback, SCN neurons retain some competence as circadianpacemakers, consistent with the view that interneuronal andcytosolic events outside the core feedback loops contribute topacemaking.Once Cry-null SCN had become arrhythmic, as monitored by

PMT for >10 d, they received a WT SCN. The immediate res-toration of rhythmicity seen in grafted VIP-null slices was notseen, however. Instead, for 2–3 d there was a progressive in-crease in bioluminescence from the host, and only then werespontaneous cycles of bioluminescence initiated (Fig. 5 C andD). These rhythms attained their peak amplitude significantly(P < 0.01) later than did VIP-null SCN (Fig. S5A). Nevertheless,once established, circadian rhythms in the Cry-null SCN weresustained for the remainder of the recording (>10 d). As with theVIP-null slices, CCD imaging showed that restored pacemakingarose from changes across the population of individual neurons(Fig. 5D): Not only did the level of bioluminescence increase,but also the cellular rhythms became better defined (RAE pre-graft = 0.47 ± 0.08, postgraft = 0.20 ± 0.01) and the cells pro-

Fig. 4. Absence of VIP/VPAC2 signaling reveals dependence of SCN circa-dian gene expression on AVP- and GRP-mediated signaling. (A) Relativeamplitude (percentage of initial peak value) of successive peaks of circadianbioluminescence of VIP-null SCN grafted with WT SCN and treated withvehicle (n = 3) or GRP receptor antagonist (n = 4, mean ± SEM). (B) As in Abut with AVP V1a and V1b antagonists (vehicle, n = 4; antagonists, n = 5). (C)Representative bioluminescence recordings of VPAC2-null SCN grafted withWT SCN and treated with vehicle or GRP antagonist. (D) As in C but treat-ment with vehicle or AVP V1a and V1b receptor antagonists. (E) Relativeamplitude (percentage of initial peak value) of successive peaks of circadianbioluminescence of VPAC2-null SCN grafted with WT SCN and treated withvehicle (n = 5) or GRP receptor blocker (n = 6, mean ± SEM). (F) As in E butwith vehicle (n = 5) or AVP V1a and V1b receptor antagonists (n = 7).

Fig. 5. Residual circadian pacemaking in Cry-null SCN slices and cells andits coordination by interneuronal signaling. (A) Representative coherentcircadian bioluminescence rhythms of Cry-null SCN. (B) Aggregate bio-luminescence emission of representative nonrhythmic Cry-null SCN and in-dividual raster plots from 50 cells. (C) Representative bioluminescenceemission from (nonrhythmic) Cry-null (black) and VIP-null (blue) SCN graftedwith WT SCN. Note immediate restoration of rhythms in VIP-null and pro-gressive effect in Cry-null SCN. (D) Raster and Rayleigh plots (with accom-panying mean vector) of bioluminescence of 50 cells in representative Cry-null SCN that received a WT graft.

Maywood et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14309

NEU

ROSC

IENCE

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

July

20,

202

1

gressively became synchronized (Fig. 5D and Movie S6), as evi-denced by an increase in the mean vector (Fig. S7C). Thus, eventhough the effect takes longer to become apparent, paracrinesignals delivered from the graft were sufficiently powerful tosustain rhythmic gene expression in the absence of essentialelements of the core transcriptional feedback loop.

DiscussionCoordination of behavioral and physiological rhythms dependsupon synchronization of the ∼10,000 neurons within the SCNcircuit (5). By developing a unique coculture technique, com-bined with genetic and pharmacological manipulations, we haveshown how a diversity of neuropeptidergic, paracrine cues sus-tains amplitude, period, and synchrony of cellular circadianpacemaking in the intact SCN. Preeminent is the VIP/VPAC2signaling cascade, but in its absence AVP- and GRP-mediatedsignals can be effective, although the efficiency of these non-VPAC2 mechanisms, in terms of rate of restoration and controlover circadian period, is less. These paracrine cues are suffi-ciently potent to overcome genetic deficiencies within the coretranscriptional feedback loop of Cry-null SCN. These findingsprovide a unique perspective on the relative contributions ofintra- and intercellular, circuit-level mechanisms in coordinatingthe circadian pacemaker.Previous interest in circadian paracrine/humoral signaling has

focused on SCN outputs, rather than how SCN neurons mightcommunicate with each other. For example, SCN grafts can re-store genotypically specific in vivo rest activity and (some)physiological cycles in SCN-ablated adult hosts (13–15). More-over, SCN grafts encapsulated in semipermeable membrane toprevent neuronal outgrowth restore behavioral, but not other,circadian rhythms (16), whereas in vitro, SCN cells (17) andslices (18) can restore gene oscillations in fibroblasts andrhythmic electrical activity in the PVN (19). As for intra-SCNcommunication, unknown paracrine signals have been inferredin early fetal development, when synaptogenesis is incomplete(20). The MWCO barriers demonstrate that paracrine signalsare potent stimuli within the adult SCN. In particular, the ∼2-kDa membrane, which would exclude access to the host of graft-derived VIP and GRP1-27, blocked circadian cues: consistentwith the genetic and pharmacological evidence of the circadianroles of these peptides. The current study, therefore, uniquelyreveals the potency of diverse paracrine signals within and be-tween intact SCN circuits. Moreover, it shows explicitly that thispotency is a property of cues derived from the graft SCN andthat such cues can impose long-term circadian function overgenetically deficient SCN circuits. Indeed, the host SCN behavesmuch like a driven oscillator. Of interest, therefore, is the degreeto which particular factors (AVP, VIP, and GRP) and secondmessenger pathways (cAMP and Ca2+) contribute to signalingboth within the SCN and between the SCN and its targets. Itshould be noted, however, that both VIP and AVP are expressedin tissues (cerebral cortex and PVN) that were unable to restorecircadian cycles in mutant host SCN. Thus, the SCN circuitry aswell as its neurochemistry confers specific pacemaking properties.VIP/VPAC2 signals have a primary role in communicating

circadian information across the SCN circuit, conferring period,amplitude, and synchronization. These properties are sensitiveto signaling by cAMP (21), the biochemical mediator of VIP/VPAC2 in the SCN (22). The downstream consequences ofVPAC2/cAMP activation in SCN neurons will ultimately involveinduction of Per genes via their cAMP/calcium-response ele-ments (23). Indeed, acute treatment of VIP-null SCN with ex-ogenous VIP rapidly induces Per-driven bioluminescence. Thisactivation may involve both direct (cAMP/PKA) and indirectbiochemical pathways as well as altered patterns of electric firingactivity (24). Remarkably, these network properties of VIP areechoed in Drosophila, where the neuropeptide pigment-dispersingfactor (PDF) synchronizes fly clock neurons via the fly homologof the VPAC2 receptor (25). Thus, VIP/PDF/VPAC2 signaling,

and its associated regulation of cAMP (21, 25), may be a con-served mechanism in circadian circuits.Pharmacological blockade of GRP receptors had no effect

on WT SCN slices, consistent with the normal behavioral andSCN gene expression rhythms in BB2r-null mice (26). Similarly,although loss of the AVP V1a receptor may disrupt behavioralrhythms to a variable extent (27), Per gene expression in the SCNis unaffected in AVP V1a-null mice and blockade of AVP-Gq/11-coupled signaling with specific antagonists to both the V1aand V1b subtypes did not affect bioluminescence rhythms in WTSCN. Critically, however, V1a, V1b, and BB2r blockade revealedhow both peptidergic pathways are necessary in SCN lackingVIP/VPAC2. This VIP/VPAC2-independent signaling took lon-ger to take effect, however, and failed to drive host SCN at atyp-ical circadian periods. Moreover, there were differences betweenAVP- and GRP-mediated effects. In VIP-null SCN, blockade ofV1a/V1b prevented the long-term restoration of rhythms whereasblockade of BB2r receptors had no effect (perhaps due to com-petent AVP signals). Equally, in VPAC2-null SCN hosts, V1a/V1b blockade was immediately effective, whereas BB2r blockadetook several days to become effective, indicating a more potentrole for AVP over GRP. Thus, within the intact SCN circuit,distinct neurochemical pathways may convey specific circadianproperties. What is being communicated (amplitude, period, andphase) may vary between intercellular contacts and neurochem-ical pathways, with a VIP > AVP > GRP hierarchy of effect. Inaddition to this peptidergic system, GABA appears to modulatepeak firing rate and precision of the oscillations whereas addi-tional factors signaling via PTX-sensitive G proteins may con-tribute to synchronization (28).Of particular interest is the interdependence between the two

principal subregions, the AVP-rich shell and the VIP- and GRP-rich, retinorecipient core. Their surgical separation desynch-ronizes the dorsomedial shell, presumably due to loss of VIP-and GRP-ergic signals (29), and a simple model is that core andshell contain reciprocally supportive oscillators coupled by pep-tidergic signals. Thus, the shell receives a dense innervation byVIP and GRP (30) and expresses VPAC2 (31) and BB2r (26)whereas the V1a receptor is widely distributed across SCN (27).This model complements the conclusion that there is no uniquelyspecialized or anatomically localized class of cell-autonomouspacemakers (11): rather, AVP, VIP, and other SCN neurons areintrinsic but unstable circadian oscillators that rely on networkinteractions to stabilize their otherwise noisy cycling. It has beenargued that intrinsic instability is an adaptive property of cellssuch that it renders them more responsive to extracellular cues.In the context of synchronization, this result has two aspects: thesynchronization of the circuit’s spontaneous oscillation and theresynchronization associated with entrainment by environmentalcues, e.g., light. The present study reveals a role for VIP, AVP,and GRP in the former, and VIP and GRP also mediate reset-ting by light. The underlying cellular processes may therefore becommon regardless of the context in which synchronization occurs.An unexpected finding was that cells in Cry-null SCN retain

oscillatory properties and that ensemble circadian gene expres-sion persisted for several cycles in the majority of SCN. Thisresult contrasts with the immediate loss of circadian behavior inCry-null mice placed into continuous darkness and the absenceof spontaneous circadian rhythms of ensemble electrical activityin acute SCN slices (32). Thus, the Cry-null SCN may continue tooscillate in vivo but may be deficient in peptidergic factors, e.g.,prokineticin (33) necessary to coordinate behavior and/or theelectrical patterns required to release them at postsynaptic tar-gets. Thus, although Cry1 and Cry2 have been viewed as essentialcomponents of the negative limb of the circadian clock feedbackloop (34), our results show that oscillatory properties exist in SCNneurons in the absence of Cry. A parallel to this was recentlyreported from SCN slices lacking the positive circadian regulatorBmal1 (10), which exhibited long-term stochastic oscillations inthe circadian domain, dependent on intercellular communicationand cAMP signaling. Cry proteins have recently been identified

14310 | www.pnas.org/cgi/doi/10.1073/pnas.1101767108 Maywood et al.

Dow

nloa

ded

by g

uest

on

July

20,

202

1

as negative regulators of cAMP signaling (35) and so an up-regulation of cAMP-mediated cues might be anticipated in Cry-null SCN cells and possibly adequate to sustain Per gene ex-pression. The results on Bmal1- and Cry-null SCN are consistentwith the view that oscillatory mechanisms involving cytoplasmicsignaling and membrane electrical activity are important com-ponents of the cellular pacemaker (4, 36). This effect may alsounderlie the (in vivo) restoration of behavioral rhythms in Per2mutant mice exposed to continuous bright light that would acti-vate both glutamatergic and peptidergic signaling within theSCN (37). In conclusion, intercellular signals emitted by WT graftSCN were able to restore competent circadian gene expressionrhythms in Cry-null SCN. Therefore, loss of intracellular factors(Cry) or intercellular factors (VIP) can confound the SCN circuit,but suitable restoration of intercellular signals can overcomethese deficits.

Materials and MethodsExperiments were conducted under the 1986 Home Office Animals (ScientificProcedures) Act (United Kingdom). PER2::LUC, Vip−/−, Vipr2−/−, Cry1−/−:Cry2−/−

CSK1eTau/Tau, and Fbxl3Afh/Afh mice were generated by Joe Takahashi (Uni-versity of Texas Southwestern Medical Center, Dallas), Chris Colwell (Uni-versity of California, Los Angeles), Tony Harmar (University of Edinburgh,Edinburgh), G. van der Horst (Erasmus University, Rotterdam), AndrewLoudon (University of Manchester, Manchester, UK), and Pat Nolan (MedicalResearch Council, Harwell, UK) and subsequently bred at the Medical Re-search Council, Cambridge, UK, on C57Bl6 backgrounds. Mice were housedunder 12 h:12 h light:dim red light (LD) with ad libitum food and water.

Brains were removed from adults or pups of at least 10 d of age andsectioned as reported previously (7). Bioluminescent emissions from PER2::

LUC SCN slices were recorded by PMTs and CCD cameras (Hamamatsu) asdescribed previously (7). After at least 10 d of recording from the host SCN,a second, freshly prepared nonreporter organotypic SCN slice (graft) fromeither WT or circadian period mutant mice (10–14 d old) was placed directlyon top of the host PER2::LUC SCN. Rhythms in bioluminescence wererecorded for a further 10 d. To assess the specificity of the SCN graft, somePER2::LUC, Vip−/− slices received a graft of either hypothalamic PVN or ce-rebral neocortex from nonreporter mice. Waveforms of rhythmic bio-luminescence emission from whole slices and individual cells were analyzedin BRASS software (A. Millar, University of Edinburgh, Edinburgh, andM. Straume, University of Virginia, Charlottesville, VA).

Following recording, slices were fixed in 4% paraformaldehyde for 30 minat room temperature, washed in 0.1M PBS, and then incubated for 1 h in 20%donkey serum in 0.01M PBS containing 0.3% BSA and 0.1% Triton. They werethen incubated overnight at room temperature with primary antisera againstAVP (rabbit, 1:50; Abcam), VIP (guinea-pig, 1:5,000; Peninsula Laboratories),and GFAP (chicken, 1:5,000; Abcam); washed; and then incubated for 2 h insecondary antibody (AVP, anti-rabbit Alexa Fluor 488; VIP, anti-guinea pigAlexa Fluor 647; GFAP, anti-chicken Alexa Fluor 568). Slices were washed,mounted on slides using Vectashield-DAPI mounting medium, and imagedwith a Bio-Rad Radiance 2100 confocal microscope. BB2r antagonist PD176252and AVP1a antagonist SR49059 were from Tocris, and AVP1b antagonistSAR149415 was a gift from C. Serradeil Le-Gal, Sanofi, Toulouse, France. Datawere compared by two- and one-way ANOVA in GraphPad Prism with a posthoc Bonnferroni t test or Student’s t test where applicable.

ACKNOWLEDGMENTS. We thank Mrs. Tracey Butcher for expert technicalassistance. This work was supported by the Medical Research Council (UnitedKingdom) and the Biotechnology and Biological Sciences Research Council(United Kingdom) (Grants EO232231 and EO225531).

1. Takahashi JS, Hong HK, Ko CH, McDearmon EL (2008) The genetics of mammaliancircadian order and disorder: Implications for physiology and disease. Nat Rev Genet9:764–775.

2. Meng QJ, et al. (2008) Setting clock speed in mammals: The CK1 epsilon tau mutationin mice accelerates circadian pacemakers by selectively destabilizing PERIOD proteins.Neuron 58:78–88.

3. Godinho SI, et al. (2007) The after-hours mutant reveals a role for Fbxl3 in de-termining mammalian circadian period. Science 316:897–900.

4. Hastings MH, Maywood ES, O’Neill JS (2008) Cellular circadian pacemaking and therole of cytosolic rhythms. Curr Biol 18:R805–R815.

5. Mohawk JA, Takahashi JS (2011) Cell autonomy and synchrony of suprachiasmaticnucleus circadian oscillators. Trends Neurosci 34:349–358.

6. Harmar AJ, et al. (2002) The VPAC(2) receptor is essential for circadian function in themouse suprachiasmatic nuclei. Cell 109:497–508.

7. Maywood ES, et al. (2006) Synchronization and maintenance of timekeeping insuprachiasmatic circadian clock cells by neuropeptidergic signaling. Curr Biol 16:599–605.

8. Liu AC, et al. (2007) Intercellular coupling confers robustness against mutations in theSCN circadian clock network. Cell 129:605–616.

9. Welsh DK, Takahashi JS, Kay SA (2010) Suprachiasmatic nucleus: Cell autonomy andnetwork properties. Annu Rev Physiol 72:551–577.

10. Ko CH, et al. (2010) Emergence of noise-induced oscillations in the central circadianpacemaker. PLoS Biol 8:e1000513.

11. Webb AB, Angelo N, Huettner JE, Herzog ED (2009) Intrinsic, nondeterministic cir-cadian rhythm generation in identified mammalian neurons. Proc Natl Acad Sci USA106:16493–16498.

12. Abrahamson EE, Moore RY (2001) Suprachiasmatic nucleus in the mouse: retinal in-nervation, intrinsic organization and efferent projections. Brain Res 916:172–191.

13. King VM, et al. (2003) A hVIPR transgene as a novel tool for the analysis of circadianfunction in the mouse suprachiasmatic nucleus. Eur J Neurosci 17:822–832.

14. Sujino M, et al. (2003) Suprachiasmatic nucleus grafts restore circadian behavioralrhythms of genetically arrhythmic mice. Curr Biol 13:664–668.

15. Guo H, Brewer JM, Champhekar A, Harris RB, Bittman EL (2005) Differential control ofperipheral circadian rhythms by suprachiasmatic-dependent neural signals. Proc NatlAcad Sci USA 102:3111–3116.

16. Silver R, LeSauter J, Tresco P, Lehman MN (1996) A diffusible coupling signal from thetransplanted SCN controlling circadian locomotor rhythms. Nature 382:810–813.

17. Allen GC, Farnell Y, Bell-Pedersen D, Cassone VM, Earnest DJ (2004) Effects of alteredclock gene expression on the pacemaker properties of SCN2.2 cells and oscillatoryproperties of NIH/3T3 cells. Neuroscience 127:989–999.

18. Li N, et al. (2008) Suprachiasmatic nucleus slices induce molecular oscillations in fi-broblasts. Biochem Biophys Res Commun 377:1179–1184.

19. Tousson E, Meissl H (2004) Suprachiasmatic nuclei grafts restore the circadian rhythmin the paraventricular nucleus of the hypothalamus. J Neurosci 24:2983–2988.

20. Moore RY, Bernstein ME (1989) Synaptogenesis in the rat suprachiasmatic nucleusdemonstrated by electron microscopy and synapsin I immunoreactivity. J Neurosci 9:2151–2162.

21. O’Neill JS, Maywood ES, Chesham JE, Takahashi JS, Hastings MH (2008) cAMP-

dependent signaling as a core component of the mammalian circadian pacemaker.

Science 320:949–953.22. Rea MA (1990) VIP-stimulated cyclic AMP accumulation in the suprachiasmatic hy-

pothalamus. Brain Res Bull 25:843–847.23. Nielsen HS, Hannibal J, Fahrenkrug J (2002) Vasoactive intestinal polypeptide induces

per1 and per2 gene expression in the rat suprachiasmatic nucleus late at night. Eur J

Neurosci 15:570–574.24. Irwin RP, Allen CN (2010) Neuropeptide-mediated calcium signaling in the supra-

chiasmatic nucleus network. Eur J Neurosci 32:1497–1506.25. Shafer OT, et al. (2008) Widespread receptivity to neuropeptide PDF throughout the

neuronal circadian clock network of Drosophila revealed by real-time cyclic AMP

imaging. Neuron 58:223–237.26. Aida R, et al. (2002) Gastrin-releasing peptide mediates photic entrainable signals to

dorsal subsets of suprachiasmatic nucleus via induction of Period gene in mice. Mol

Pharmacol 61:26–34.27. 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:R824–R830.28. Aton SJ, Huettner JE, Straume M, Herzog ED (2006) GABA and Gi/o differentially

control circadian rhythms and synchrony in clock neurons. Proc Natl Acad Sci USA 103:

19188–19193.29. Yamaguchi S, et al. (2003) Synchronization of cellular clocks in the suprachiasmatic

nucleus. Science 302:1408–1412.30. Drouyer E, LeSauter J, Hernandez AL, Silver R (2010) Specializations of gastrin-

releasing peptide cells of the mouse suprachiasmatic nucleus. J Comp Neurol 518:

1249–1263.31. Kalló I, et al. (2004) Transgenic approach reveals expression of the VPAC2 receptor in

phenotypically defined neurons in the mouse suprachiasmatic nucleus and in its ef-

ferent target sites. Eur J Neurosci 19:2201–2211.32. Albus H, et al. (2002) Cryptochrome-deficient mice lack circadian electrical activity in

the suprachiasmatic nuclei. Curr Biol 12:1130–1133.33. Prosser HM, et al. (2007) Prokineticin receptor 2 (Prokr2) is essential for the regulation

of circadian behavior by the suprachiasmatic nuclei. Proc Natl Acad Sci USA 104:

648–653.34. Kume K, et al. (1999) mCRY1 and mCRY2 are essential components of the negative

limb of the circadian clock feedback loop. Cell 98:193–205.35. Zhang EE, et al. (2010) Cryptochrome mediates circadian regulation of cAMP signal-

ing and hepatic gluconeogenesis. Nat Med 16:1152–1156.36. Nitabach MN, Holmes TC, Blau J (2005) Membranes, ions, and clocks: Testing the Njus-

Sulzman-Hastings model of the circadian oscillator. Methods Enzymol 393:682–693.37. Spoelstra K, Daan S (2008) Effects of constant light on circadian rhythmicity in mice

lacking functional cry genes: Dissimilar from per mutants. J Comp Physiol A Neuro-

ethol Sens Neural Behav Physiol 194:235–242.

Maywood et al. PNAS | August 23, 2011 | vol. 108 | no. 34 | 14311

NEU

ROSC

IENCE

SEECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

July

20,

202

1