On the origin and evolution of the dual oscillator model underlying the photoperiodic clockwork in the suprachiasmatic nucleus

Abstract

Decades have now passed since Colin Pittendrigh first proposed a model of a circadian clock composed of two coupled oscillators, individually responsive to the rising and setting sun, as a flexible solution to the challenge of behavioral and physiological adaptation to the changing seasons. The elegance and predictive power of this postulation has stimulated laboratories around the world in searches to identify and localize such hypothesized evening and morning oscillators, or sets of oscillators, in insects, rodents, and humans, with experimental designs and approaches keeping pace over the years with technological advances in biology and neuroscience. Here, we recount the conceptual origin and highlight the subsequent evolution of this dual oscillator model for the circadian clock in the mammalian suprachiasmatic nucleus; and how, despite our increasingly sophisticated view of this multicellular pacemaker, Pittendrigh’s binary conception has remained influential in our clock models and metaphors.

Introduction

Schematic diagrams—the ones that seek to provide visual representations of complex phenomena—are critical to scientific progress, not only for portraying known facts but also for helping to conceptualize our mental metaphors and future experimental designs (Perini 2005). In 1957, Colin Pittendrigh and Victor Bruce created a most consequential block diagram for a mechanism of the circadian clock (Pittendrigh and Bruce 1957; see Fig. 1A), featuring a tripartite organization with (a) a “clock” composed of one or more “endogenous self-sustaining oscillators” (ESSO’s) with natural inherent periods of ≈ 24 h, (b) input “black boxes” for entrainment to daily geophysical cues, e.g., the light–dark (LD) cycle, and (c) output “black boxes” for translation of the entrained timing signal to an observable rhythm with a period of exactly 24 h. Emphasized in the text is the mutual coupling of the ESSO’s:

Fig. 1.

A Pittendrigh and Bruce’s conception of the three elements of a biological clock. Modified and redrawn from the original in Pittendrigh and Bruce (1957). B Eskin’s simplified version of the system. Modified and redrawn from the original in Eskin (1979)

“Thus in our problem the stability and precision of the clock system which are two of its most challenging features could well be due to the mutual control exercised by a population of coupled ESSO’s within the cell, or between cells at the tissue level of organization.”

While the authors predicted that the basic mechanism of ESSO lies inside the cell, they also suggested the likelihood that metazoan clocks are composed of coupled neuronal populations:

“It is our view that probably all cells are autonomously capable of an ESSO with the long period of a solar day; but that in higher animals the function of daily time measurement is executed by a few that are specialized, in typical metazoan fashion, for this particular function.”

This schematic in turn served as the inspiration for Arnold Eskin’s even simpler conceptualization (Eskin 1979; see Fig. 1B), often referred to in tribute as the “eskinogram”. Eskin’s strictly linear version pointed to a strategy for localizing Pittendrigh and Bruce’s clock, by tracing the input forward and/or the output backward to it, but in so doing deleted some predictably important features (i.e., the multiple coupled ESSO’s within the clock and the “residual periodic variables” that might bypass the central clock mechanism). Now, over 65 years later, these influential models have helped to foster many discoveries, from the localization of the suprachiasmatic nucleus (SCN), to its multicellular organization, and on to how it promotes adaptive plasticity in behavior and physiology across the seasons. In this commentary, we look back and forward on the impact of a dual oscillator model for the SCN, with the view that its ongoing evolution can continue to inspire our experimental questions and designs.

A short history on the origin of mammalian E (evening) and M (morning) oscillators

Through the 1960’s, Pittendrigh argued for his concept of mutually coupled ESSO’s in part by presenting intriguing actograms that hinted at the presence of at least two dissociable oscillators underlying the circadian locomotor activity rhythm in some rodents (Pittendrigh 1960; 1967). In an Arctic ground squirrel (Spermophilus undulatus), a red backed vole (Clethrionomys rutilus), and a golden (Syrian) hamster (Mesocricetus auratus) maintained in constant ambient light (LL), the normally consolidated activity/rest interval was observed to separate into two activity bouts, each exhibiting a different free-running circadian period. This state appeared to be unstable; when one of the components with the relatively shorter or longer period essentially scanned the entire cycle, it reunited with the other component to re-form a reconstituted rhythm. Later, a more stable form of this phenomenon, now referred to as “splitting,” was demonstrated not only in the nocturnally active hamster but also in the diurnally active tree shrew (Tupaia belangeri): the two split bouts, after free-running with different periods for a number of cycles, re-synchronized with a common period but 180° in antiphase—12 circadian hours apart—and could remain in that state more or less indefinitely (Hoffmann 1971; Pittendrigh 1974).

Splitting became a centerpiece of Pittendrigh and Daan’s (1976) model of a rodent circadian clock composed of two mutually coupled circadian oscillators, which postulated an evening (E) oscillator—with an intrinsic period < 24 h, decelerated by light, and programming locomotor activity after dusk—and a morning (M) oscillator—with an intrinsic period > 24 h, accelerated by light, and programming locomotor activity before dawn. The model could account for several history-dependent features of the clock, and its inspiration had been provided by empirical characteristics of splitting in hamsters:

“The splitting phenomena and related behaviors...provide the first tangible evidence that circadian phenomena in animals do in fact involve oscillations which may be separately coupled to sunrise and sunset...”

And then the possible relevance to seasonal timekeeping was raised as:

“A pacemaker comprising two oscillators mutually interacting but coupled separately to sunrise and sunset enhances its competence to accommodate to seasonal change...”

Thus did the terms E and M come to refer both to the two split locomotor activity bouts in LL and to the two hypothesized oscillators controlling morning and evening components of non-split activity time, a language implicitly equating the oscillators that underlie splitting with those responsible for photoperiodic induction. While the experimental basis for the model proposed by Pittendrigh and Daan in 1976 relied on hamster locomotor activity recordings, LL-induced splitting also occurs in laboratory rats (Boulos and Terman 1979) and mice (Ohta et al. 2005), although perhaps less reliably.

As applied to the master mammalian circadian pacemaker in the SCN (see Hastings et al. 2018 for review), such a dual oscillator model could account for differential phase shifting of the onset and offset of rhythms regulated by the rodent SCN (e.g., locomotor activity (Honma et al. 1985; Meijer and De Vries 1995) and melatonin production (Illnerová and Vanĕček 1982); their common adjustment to transitions between long summer and short winter days (Elliott and Tamarkin 1994); and the contemporaneous splitting of hamster activity, drinking, body temperature, and endocrine rhythms (Shibuya et al. 1980; Pickard et al. 1984; Swann and Turek 1985). However, it has become clear that the E and M oscillators originally hypothesized to regulate both splitting and photoperiodic encoding are in fact not one and the same. Actually, Pittendrigh and Daan (1976) noted that in the majority of their hamsters they were unable to assign evening or morning origins to the split bouts, as also reported later by others (Earnest and Turek 1982; Morin and Cummings 1982; Swann and Turek 1982); the transition phenotype from the unsplit to split state is quite variable (Fig. 2). Notably, while unilateral SCN lesions could disrupt LL-induced split rhythms (Pickard and Turek 1982, but see Harrington et al. 1990), they did not block photoperiod-induced gonadal regression in male hamsters (Hastings et al. 1987). Moreover, theoretical analyses (Daan and Berde 1978) and experimental data (Lees et al. 1983; Boulos and Morin 1986; Meijer et al. 1990) indicated that the two split oscillators share circadian properties and must be fully competent pacemakers on their own. Daan and Berde’s own theoretical work had already prompted them to raise the possible functional significance of the SCN’s bilateral symmetry. Indeed, over 2 decades later, exploiting the rhythmic expression of the clock genes Per and Bmal1 as a tissue probe, it was finally demonstrated that the dual oscillators of the split state actually correspond to the left and right sides of the paired SCN reorganized and oscillating in antiphase (de la Iglesia et al. 2000; see also Ohta et al. 2005, Tavakoli-Nezhad and Schwartz 2005; Yan et al. 2005).

Fig. 2.

Double-plotted actograms of golden hamster wheel-running activity in LL. In the actogram format, the number of wheel revolutions over the course of each 24-h period is charted horizontally from left to right and succeeding days stacked vertically from top to bottom. Double-plotted actograms repeat this format for 48-h periods; that is, day n is followed by day n + 1 horizontally, succeeded by day n + 1 and n + 2 on the next line, then by day n + 2 and n + 3, and so on. In all cases, the arrow indicates the onset of the transition from a single, relatively consolidated, circa-24 h bout of activity in LL into two split components stably coupled 180° (circa-12 h) apart; the transition is often preceded by a change in free-running period or activity duration. While (A) represents the “classical” transition phenotype, with evening and morning components free-running with different periods for a number of cycles before re-synchronizing in antiphase, B–F show a variety of alternative patterns. In (B) and (C), the appearance of the split bouts is asymmetric; in (D) and (E), the split state is preceded by an interval of arrhythmicity; and in (F), the transition includes more than two components. From: de la Iglesia HO, Tavakoli-Nezhad M, Schwartz WJ, unpublished data

An evolving search for localized SCN E and M photoperiodic oscillators

The intrinsic SCN property that is most obviously altered by ambient photoperiod is its rhythmic waveform. In the long-day condition, the light phase and subjective day are characterized by expanded durations of elevated electrical activity in vivo and ex vivo (Mrugala et al. 2000; Schaap et al. 2003; Houben et al. 2009) and immediate-early and clock gene expression (see Johnston 2005 for multiple references to time-point sampling in a variety of species); in the short-day condition, the durations are compressed. Conversely, there are corresponding, reciprocal changes in the duration of the subjective night, as defined by c-fos photoinduction after timed light pulses (Sumová et al. 1995; Vuillez et al 1996). SCN photoperiod-dependent changes in electrical activity persist in constant darkness in vivo (VanderLeest et al. 2007; Houben et al. 2009) and exhibit bilaterally symmetrical expression of Per family genes (de la Iglesia et al. 2004b). The underlying basis for the plasticity in overall waveform has been clarified by analyses of murine rhythms at the cellular level, by recording electrical activity and imaging bioluminescent clock gene reporter expression. The data reveal that photoperiod is encoded by a reconfiguration of the phases of individual oscillating SCN cells—adopting a wide distribution across long days and a narrow distribution during short days—rather than by an expansion or compression of the activity rhythms of each cell (VanderLeest et al. 2007; Naito et al. 2008; Brown and Piggins 2009). Since SCN cells express a range of free-running circadian periods in culture (Welsh et al. 1995; Herzog et al. 1998; Honma et al. 1998), Beersma et al. (2008) performed computer simulations to show how a coupled system of similar cells with variable intrinsic periods could track dawn and dusk (with faster and slower cells synchronizing to dawn and dusk, respectively, relative to the ensemble), thus eliminating the need to postulate dedicated sets of E and M oscillators with special properties beyond variability in period.

Yet, in reality, SCN neurons are genetically and phenotypically heterogeneous (Fig. 3A). The phasing of electrical, cytosolic, and molecular rhythms varies across SCN neurons located in different compartments of the network in a manner that is stereotyped across individual samples (e.g., Nakamura et al. 2001; Yamaguchi et al. 2003; Evans et al. 2011; Enoki et al. 2012; Brancaccio et al. 2013; Yoshikawa et al. 2015, 2021; Hamnett et al. 2019). For instance, spatiotemporal maps of the peak phases of cellular PER2::luciferase rhythms in mouse SCN slices show a non-random but reproducible three-dimensional topography that is largely independent of the plane of section (Evans et al. 2011). Remarkably, when such slices are collected from mice entrained to long photoperiods, the map is reorganized into two distinct subpopulations, with a central region phase-leading the surround by a magnitude proportional to increasing day length (Fig. 3B, Evans et al. 2013). The pattern resembles the well-known anatomico-functional partition of a dorsomedial “shell” and ventrolateral “core” in coronal sections, originally described in the rat but generalized to the mouse, hamster, and other mammals (Moore et al. 2002). For most practical purposes, the two areas are identified by the differential presence of two peptide neurotransmitters, arginine vasopressin (AVP) in the shell and vasoactive intestinal polypeptide (VIP) in the core (Fig. 3A), although they also differ in their cytoarchitecture, other peptides, afferent and efferent connectivity, and photic responsiveness. Under special conditions, circadian rhythmicity in the two regions can be clearly dissociated: in rat organotypic slice cultures treated with antimitotic agents, rhythms of AVP and VIP in the culture medium express different circadian periods and phase shifts to N-methyl-D-aspartate (NMDA) application (Shinohara et al. 1995); and in rats maintained in an 11 h: 11 h LD cycle, SCN Per1 and Bmal1 rhythms are entrained to the LD cycle in the core but are unentrained with a period > 24 h in the shell (de la Iglesia et al. 2004a). Another intriguing feature of the dorsal/ventral design is the presence of a 24-h “wave” of cellular activation that generally travels across the SCN dorsally to ventrally (Yamaguchi et al. 2003; Foley et al. 2011; Enoki et al. 2012; Brancaccio et al. 2013; Maywood et al. 2013), with dorsomedial cells peaking a few hours before ventrolateral ones. While AVP cells with relatively short circadian periods (Koinuma et al. 2013; Mieda et al. 2016) might seem intuitive as instigators of this phenomenon, its mechanism remains unclear and likely complex. For instance, VIP and RGS16 signaling contribute to SCN patterning along its dorsoventral axis (Doi et al. 2011; Hamnett et al. 2019; Patton et al. 2020; Kim and McMahon 2021). Further, VIP and gamma-aminobutyric acid (GABA) signaling are involved in the dorsal/ventral rearrangement induced by photoperiod (Lucassen et al. 2012; Evans et al. 2013; Rohr et al. 2019), with GABA’s action as an inhibitory or excitatory neurotransmitter depending on the intracellular chloride concentration (Myung et al. 2015; Farajnia et al. 2014).

Fig. 3.

Cellular diversity among mouse SCN neurons. A Left panels: Genetic labeling of two SCN peptide classes that are located in different network compartments. Note: Only one SCN lobe is shown for each plane of section. Right panels: SCN neurons in different regions display different rhythmic properties, such as the phasing of peak PER2::LUC expression measured ex vivo. SCN slices for each experiment are represented in the coronal (top), horizontal (middle), and sagittal (bottom) planes. From: Rohr KE, Evans JA, unpublished data. B Schematic representation of how photoperiod modulates SCN organization. Under long days, SCN neurons in distinct network compartments display differences in the timing of peak clock gene expression and electrical activity. Under short days, SCN neurons throughout the network display similar peak times regardless of spatial location. Cartoons based on results of (Jagota et al. 2000; Inagaki et al. 2007; Evans et al. 2013; Buijink et al. 2016; Yoshikawa et al. 2017). AVP arginine vasopressin, VIP vasoactive intestinal polypeptide, aSCN anterior SCN, mSCN Middle SCN, pSCN posterior SCN, OC optic chiasm, 3V third ventricle

Importantly, there is also evidence for another, orthogonal photoperiod-induced SCN reorganization along the rostral-caudal axis, as measured by clock gene expression in Djungarian (Siberian) hamsters (Phodopus sungorus) and mice (Hazlerigg et al. 2005; Inagaki et al. 2007; Naito et al. 2008; Yan and Silver 2008; Buijink et al. 2016; Yoshikawa et al. 2017), with E and M posited to reside within regions rostrally and caudally, respectively (Fig. 3B). In addition, two discrete peaks of electrical activity can be recorded in the SCN of golden hamster tissue slices when cut in the horizontal (but not in the coronal) plane; one corresponds to projected dawn and the other to projected dusk, with differential phasing by antecedent photoperiod and ex vivo glutamate application (Jagota et al. 2000). While horizontal slices, unlike coronal ones, leave rostral-caudal connectivity intact, similar slices from rats and mice do not exhibit double peaks (Burgoon et al. 2004). The explanation for this interspecies difference has not been solved, but perhaps it relates to differences in SCN wiring and circuits that operate in mammalian species with different circadian phenotypes.

Looking forward to the future of SCN models

There is no doubt that conceptualizing dual clock components (e.g., E and M, shell and core, rhythm generating and light responsive) has advanced our understanding of the SCN’s functional anatomy. But does the concept of a binary architecture narrow our perspective and distract from a larger question: What are the functional units of the SCN network and where are they localized? Will our continued search to define the function of different clock components benefit from conceptualizing a larger complex? In this regard, it may even be time to revisit splitting—the original inspiration for the E and M concept—in addition to antiphase left–right oscillations, LL also causes inverse rhythms in cFOS and PER1 to manifest in the caudal versus rostral SCN of golden hamsters (Tavakoli-Nezhad and Schwartz 2005; Yan et al. 2005). After all, alternative networks need not be constrained to only one degree of freedom.

Like many fields, circadian biology is incorporating new techniques to identify and characterize cellular diversity in form and function (e.g., Shan et al. 2020; Wen et al. 2020; Todd et al. 2020; McManus et al. 2022). It is fair to predict that recent and ongoing advances in imaging, sequencing, and computational approaches will reveal novel SCN cell clusters and network properties. In this effort, genetic, molecular, and functional profiling of cellular activity will provide important clues to classes of SCN neurons that differ in their roles. Perhaps these new techniques will reveal SCN components acting in new capacities (e.g., as filters, buffers, resistors, amplifiers, switches, and/or adaptors), which may encourage us to revise our mechanistic metaphors. Moreover, there is growing reason to expect that these functions will not be fulfilled exclusively by neurons, with recent work revealing important roles played by SCN glia in circadian timekeeping (Brancaccio et al. 2017, 2019; Barca-Mayo et al. 2017; Tso et al. 2017). Functional profiling and cell-type-specific manipulation will likely be critical for understanding how these different types of cellular clock units are connected in time and space.

Comparative approaches imply that it may be difficult to strictly divide the SCN into dual E and M components. In Drosophila, E and M oscillators were initially localized to two cellular clusters (Grima et al. 2004; Stoleru et al. 2004), Stoleru et al. 2005), but further work has revealed a distributed clock system more dynamic and plastic in its functional organization (Yao and Shafer 2014; Yao et al. 2016; Shafer et al. 2022). Current lines of thought suggest that control of the fly clock system is not dominated by a single or two-part clock, but instead that multiple cell groups are reciprocally coupled in ways that depend on environmental context (e.g., Reinhard et al. 2022; Fujiwara et al. 2018; Liang et al. 2016, 2017; Nave et al. 2021). Are there parallels or analogs in the mammalian clock? In the SCN, VIP has long been recognized to be important for sustaining intrinsic rhythms, processing photic inputs, and encoding photoperiod (Vosko et al. 2007; Lucassen et al. 2012). Other neuropeptides and classes of SCN neurons have been shown to also contribute to network function in ways that are important for circadian and photoperiodic timekeeping (e.g., Shan et al. 2020, Mieda et al. 2016, Morris et al. 2021; Porcu et al. 2022, Joye et al. 2023; Xie et al. 2023). Further, studies using intersectional approaches suggest that different classes of SCN cells can interact to drive overt rhythms in different contexts (Smyllie et al. 2016; Brancaccio et al. 2019). Given this phenotypic and cellular complexity, it may prove difficult to divide the SCN into two exclusive, fixed E and M components.

Finally, it will be exciting to see how the field advances knowledge of how SCN components connect to their inputs and outputs. For inputs, different clock units could differ in their responses to afferents in quantitative and/or qualitative ways. For example, behavioral responses to light occur with a daily rhythm whereby the clock is reset in either the delay or advance direction in a time-dependent manner (i.e., the photic phase response curve). This overt rhythm in light responses reflects the composite responses of SCN neurons that process and transmit retinal inputs to the larger network. Photo-responsive SCN cells could differ in the sensitivity, magnitude, phase dependence, and/or the mechanistic pathways driving their photic responses. For outputs, comparative biology approaches could help map how different classes of SCN cells contribute to daily rhythms in behavior and physiology. Incorporating additional animal models using genome editing techniques is now more tractable, and this could provide models to leverage the unique strengths of different species (for example, diurnal animals and photoperiodic non-responsive Djungarian hamsters). Challenging these diverse clock systems with different environmental conditions could unmask additional clock units, their dynamic connections, and their larger contributions to the integrated function of the entire circadian system. Indeed, the dual oscillator model arose from observations of plasticity in a range of model species exposed to different lighting conditions. Thus, the use of comparative approaches may help to further the evolution of schematics shaping the future of the field.

Data availability

There is no data included in this manuscript, only schematics and representative images provided by the authors.