Dr D. S. Saunders, Institute of Cell, Animal and Population Biology, University of Edinburgh, West Mains Road, Edinburgh EH9 3JT, U.K. Tel: +44 (0) 131 6641819; e-mail: firstname.lastname@example.org
Abstract. Over several decades, formal experiments measuring diapause responses to variable light inputs have indicated that photoperiodic time measurement in insects is accomplished either by a nonoscillatory ‘hourglass-like’ mechanism or by oscillatory components of the circadian system. Although both are possible given the present state of our knowledge, a substantial body of evidence strongly suggests that night-length measurement is a function of the circadian system, and that ‘hourglass-like’ clocks are manifestations of damping circadian components. The two types of time measurement, ‘hourglass’ and circadian, are therefore parts of a spectrum of mechanisms differing in their damping coefficients. If this view is correct, it may follow that genes and proteins involved in circadian timing are also involved in photoperiodism, although additional genes, or known ‘clock’ genes used in novel ways, may also play a part. This review outlines the experimental evidence for the oscillator clock theory of photoperiodism and suggests ways in which further progress may be made.
In most areas of biological enquiry, formal experimentation precedes and provides guidance for subsequent investigations aiming to uncover a more fundamental, biochemical basis of the phenomenon. For example, the laws of Mendelian segregation and the behaviour of chromosomes provided a basis for enormous advances in molecular genetics, and formal investigations into circadian rhythmicity have provided, and continue to provide, an essential framework for molecular advances in that area. Indeed, one could go as far as to state that unless biochemical and genetic data explain formal properties of the system, they are wanting to some degree. Ultimately perhaps, computer-based models using the kinetics of biochemical components are required for a fully satisfactory explanation.
The same observations are almost certainly true for photoperiodism, a widespread phenomenon in which seasonal changes in day length or night length are responsible for directing metabolism down alternate developmental pathways − to diapause or continued development, or to different seasonally appropriate morphs (Saunders, 2002). Since its discovery in the 1920s, an enormous volume of work on insect photoperiodism has been carried out using a ‘black box’ approach in which the environmental input to the system (the light–dark cycle) has been systematically varied, and the output (diapause or development) then recorded, with very little understanding of the concrete events involved in time measurement (the so-called photoperiodic clock) itself.
Many of the formal approaches to photoperiodism have addressed the question of whether the time measurement characteristic of the system is accomplished by a nonoscillatory hourglass-like mechanism (Lees, 1973; Veerman, 2001), or by reference to the insect's circadian system (Bünning, 1936; Pittendrigh & Minis, 1964; Saunders, 2002). This distinction is important for a number of reasons. If photoperiodic time measurement is one of the important functions of the circadian system, genes and proteins now thought to be involved in circadian feedback loops may also be involved in seasonal timing. If it is not, then a completely different set of genes and proteins may be involved, or circadian ‘clock’ genes may be used in different, nonoscillatory, ways.
This review describes the known properties and characteristics of the photoperiodic and circadian clocks, and then considers the evidence both for and against a causal relationship between the two. It then goes on to consider ways of moving from formal descriptions and models to opening the ‘black box’.
The distribution and variety of insect photoperiodic responses
Photoperiodic regulation is widespread in terrestrial organisms including flowering plants, fungi, birds, mammals, molluscs and arthropods (Hastings, 2001). Among the insects, such seasonality has now been recorded in over 500 species from 17 orders (Nishizuka et al., 1998). This wide occurrence suggests that the phenomenon is even more common − perhaps almost universal − especially among insects from higher temperate regions with well-marked changes between the seasons. Most attention has been paid to the important phenomena of diapause induction and the appearance of seasonal morphs (as in aphids), although many other seasonally important strategies are known (e.g. aspects of cold tolerance, migration and growth rate) (Saunders, 2002). However, all photoperiodic responses have a feature in common, involving the apparent ‘measurement’ of day length (or night length) with a surprising degree of accuracy. If days are perceived to be greater than an often well-defined critical value, long-day (or short-night) responses appropriate for a summer pathway (e.g. ‘active’ development and reproduction) ensue. If, on the other hand, days are perceived to be less than the critical value, short-day (or long-night) responses appropriate for autumnal development and impending winter conditions are induced, leading, for example, to the appearance of an over-wintering diapause, or to an egg-laying morph.
In addition to distinguishing between a long and a short night, the photoperiodic system also accumulates successive long- or short-night cycles by the so-called photoperiodic ‘counter’ to some internal threshold at which diapause or nondiapause development is determined (Saunders, 1981a). Between known events on the input pathway (e.g. photoreception) to those on the output pathway (e.g. endocrine systems regulating diapause), there are unknown, presumably brain-centred, events in which night lengths are measured and accumulated by the photoperiodic ‘clock-counter’ mechanism.
However, these sweeping generalizations cover a wide variety of responses (Saunders, 2002). If observations are confined to diapause regulation, most is known about the control of over-wintering, with the onset of diapause occurring in the autumn as days shorten. This type of response is most commonly observed in temperate latitudes (e.g. 25–60°N) with a well-defined seasonality and a cold winter inimical to continued development and reproduction. However, other types of response also occur, including that of summer diapause (Masaki, 1980). This is often, but not always, in response to longer summer day lengths, inducing dormancy to assist survival during a long dry season.
In the case of over-wintering diapause, there is a characteristic response to photoperiod (Fig. 1) revealed by exposing groups of insects to a full range of experimental photoperiods, both ‘natural’ and ‘unnatural’. The central part of this curve (region c) is dominated by the ecologically important and latitudinally dependent critical day length (or night length) that operates the seasonally appropriate switch in metabolism from continuous development under long days to diapause under shorter days. Photoperiods in region b occur naturally, but may play little or no part in diapause regulation because they occur in the depth of winter when the insect may already be in diapause, or is unresponsive to photoperiod. Those photoperiods in regions a and d fall outside the natural range and are therefore only experienced in laboratory experiments. However, responses to these ultra-short and ultra-long regimes are important in defining properties of the time-measuring system involved, as will be discussed below.
Many of these photoperiodic responses are also temperature- dependent, with temperature affecting circadian entrainment, photoperiodic summation (by the ‘counter’) and aspects of general physiology involved in diapause induction (Saunders, 2002). These effects are largely outside the scope of the present review and will not be considered in detail.
Evidence for and against a circadian involvement in photoperiodic time measurement
Circadian rhythms and photoperiodism share functional dependency on the daily light–dark cycle, the former through entrainment to an adaptive phase relationship, and the latter by ‘measuring’ day length or night length to determine seasonal change. It was the German botanist, Erwin Bünning (1936) who made the first theoretical connection between the two. His idea, later supported by experimental data, suggested that light in photoperiodism had two roles: first, the entrainment of the circadian system and, second, the illumination (or not) of a particular light-sensitive phase as day length changed with the season (Fig. 2). This idea was later developed and expanded by Pittendrigh & Minis (1964) as the ‘external coincidence’ model, which preserved this dual role for light, or in an alternative ‘internal coincidence’ model (Pittendrigh, 1972) in which light had only one role (i.e. that of entrainment) but day length changes were detected in a multioscillator system by changing phase relationships between dusk and dawn entrained components. These models, and others, have recently been reviewed in detail by Vaz Nunes & Saunders (1999) and Saunders (2002).
An alternative to such circadian-based models for the photoperiodic clock was that of a dark-phase hourglass championed by A.D. Lees in his extensive work on seasonal morph determination in aphids (Lees, 1973) and by later investigators, particularly Veerman (2001) working on the spider mite, Tetranychus urticae. In this alternative, night length was ‘measured’ by a nonoscillatory (i.e. noncircadian) timer set in motion at dusk and used to distinguish short from long nights by whether the dawn transition arrived early or late with respect to the insect's critical night length. Today, both alternatives are advocated, generally in different species, but often supported by data from very similar experimental protocols. These experiments and their interpretation are reviewed in the next section.
However, before proceeding, it should be noted that Bünning regarded the up-and-down circadian leaf movements of the plants he studied as an overt expression of the unseen photoperiodic clock. He assumed that overt rhythms could be regarded as ‘hands of the clock’, both being manifestations of the same rhythmic system. However, in insects, although the eclosion rhythm of the flesh fly Sarcophaga argyrostoma provided a remarkable parallel for pupal diapause induction (Saunders, 1978, 1979, 1981b), significant differences between the two systems have emerged. In the pink boll worm moth Pectinophora gossypiella, Pittendrigh et al. (1970) showed that larvae could distinguish between 12 h of red light (640 nm) per day (98% diapause) and 14 h of red light per day (5% diapause), even though overt rhythms such as the egg hatch rhythm were ‘blind’ to wavelengths longer than approximately 520 nm (Bruce & Minis, 1969). More recently, studies on the blow fly Calliphora vicina have indicated that the photoperiodic mechanism possesses a free-running period close to 24 h, but is damping, whereas the overt rhythm of locomotor activity is fully self-sustained and free-runs with a period close to 22.5 h (Saunders, 1997, 2002). These differences show that the use of overt rhythms as ‘hands’ of the photoperiodic clock is, at best, limited, but in no way lead to the conclusion that the photoperiodic clock is noncircadian. The differences quoted above merely demonstrate that the two systems are ‘different’ (i.e. functionally distinct, or separate) with different circadian properties and perhaps different cellular substrates and input pathways.
‘Black box’ experiments using circadian entrainment theory
Formal or ‘black box’ experiments in which systematic variation of the light–dark input is assessed in terms of diapause incidence have been conducted using insect material for many decades (Saunders, 2002). The results of these experiments have been used to interpret photoperiodic time measurement either as a function of the insect's circadian system, or in terms of a noncircadian hourglass-like mechanism. It is the purpose of this review to suggest that there is very little difference between these two, apparently disparate, interpretations, and that the more parsimonious view (i.e. that Bünning's general hypothesis is largely correct) is one that will be the more useful to follow in attempting to unravel the concrete details of photoperiodic time measurement.
Formal experiments may conveniently be divided into: (i) those in which the length of the light–dark cycle (T) is 24 h or close to 24 h and (ii) those in which cycle duration T is greatly lengthened, generally for 2 or 3 days or more. In both procedures, the experimental light–dark cycle is repeated throughout the insect's photoperiodically sensitive period, and the response (proportion of the population entering diapause) is then determined at its conclusion.
Experiments in which T = 24 h, or T is close to 24 h
The most frequently conducted experiment is one where the dark period of an otherwise diapause-inductive long-night cycle (e.g. LD 12 : 12 h) is systematically interrupted with a short (e.g. 1 h) light pulse placed, in different experimental insect subgroups, at different times after the onset of darkness. These protocols have been called ‘night-interruption’ experiments. In most insects tested in this way, the scanning light pulse produces two points of diapause aversion (short-night effect), one (frequently called A) lying early in the night and a second (frequently called B) lying later in the night (Saunders, 2002). In many species (but not all), the diapause-averting effect of light falling at point B is greater than that at A. Point B generally falls at a position that delimits the critical night length for the species under investigation.
In the aphid Megoura viciae, Lees (1973) interpreted such a result in terms of a linked series of biochemical reactions distinguished on the basis of their responses to light breaks in the dark component of the cycle. The crucial point was his stage 3 leading up to the critical night length, 9.5 h after dusk. If the interrupting light pulse or the start of the new photophase (dawn) began at this position, equivalent to point B, short-night responses (virginopara production) occurred; on the other hand, if stage 3 fell in the dark, long-night responses (oviparae) were evident. Lees' model for time measurement in M. viciae therefore was a linear or nonoscillatory device measuring night length.
Using very similar night-interruption data obtained by Adkisson (1964) for the pink boll worm moth P. gossypiella, Pittendrigh & Minis (1964) provided a radically different interpretation for the bimodal response to light breaks. This was based on known peculiarities of the circadian entrainment phenomenon in the eclosion rhythm of the fruit fly Drosophila pseudoobscura (Pittendrigh, 1965, 1966) when exposed to similar night interruption photocycles. In doing so, they proposed a model for the photoperiodic clock (‘external coincidence’) in which light had a dual role: entrainment of the circadian system, and photic induction of nondiapause or diapause by light either coinciding or not, with a particular light-sensitive phase. This model was thus an extension of Bünning's general hypothesis. The photoinducible phase was identified as point B in night-interruption experiments (equivalent to stage 3 in Lees' hourglass-like timer).
By direct analogy with the eclosion rhythm in D. pseudoobscura, Pittendrigh & Minis (1964) suggested that the scanning light pulse caused phase delays in the early part of the night until, at point A, the photoinducible phase (at B) was delayed sufficiently to coincide with the dawn transition of the main photophase, thus causing short-night or diapause-averting responses. On the other hand, light pulses falling in the second part of the night caused phase advances until, at point B, a direct coincidence with light occurred, and diapause was again averted (Fig. 3).
The pupal eclosion rhythm of D. pseudoobscura could not be regarded as ‘hands’ of the photoperiodic clock in P. gossypiella because the two phenomena occurred in different species. Neither could the overt rhythms later studied in Pectinophora (Minis, 1965; Pittendrigh & Minis, 1971) for reasons of their differences outlined above. However, a subsequent study of pupal diapause induction in the flesh fly S. argyrostoma (Saunders, 1978, 1979) provided that opportunity because of the very close similarities between the proposed photoperiodic clock oscillator and that underlying eclosion. In other words, although the two phenomena (pupal diapause induction and eclosion rhythmicity) were clearly separate, both possessed properties indicating an involvement of the insect's circadian system and their similarities were sufficiently close to allow the use of the overt eclosion rhythm as an indicator of phase (‘hands of the clock’), as Bünning had originally proposed.
Using a computer programme based on phase response curves for the eclosion rhythm, and an assumption that the photoinducible phase fell at point B[Circadian time (Ct) 21.5 h], very close predictions for diapause incidence in S. argyrostoma were obtained (Saunders, 1978, 1979, 1981b). Such parallels were found not only for night interruptions in cycles where T = 24 h, but also in more elaborate cycles in which the hours of darkness before and after the scanning pulse were systematically varied, but the overall cycle remained within the primary range of entrainment. These latter experiments were almost exact replicas of those originally conducted by Lees (1971) for the aphid M. viciae but were here interpreted in terms of circadian entrainment instead of a nonoscillatory ‘hourglass’.
In other experiments using the same approach, the diapause induction effects of two short (1 h) pulses of light per cycle (a skeletal LD cycle), particularly within the characteristic circadian ‘zone of bistability’ (Saunders, 1975, 1978), and of a single 1-h pulse repeated in cycles with a period (T) varied between 21.5 and 30.5 h (Saunders, 1979), were found to be just as accurate as predictors of both circadian phase and diapause regulation. In every case, diapause incidence was low (short-night effects) when light coincided with the supposed photosensitive phase (Ct 21.5), but high when this phase fell in the dark. Of course, these experiments (and others described below) did not provide proof (in a Euclidean sense) of a causal relationship between circadian rhythmicity and the photoperiodic clock, but amounted to very strong circumstantial evidence that photoperiodic time measurement in S. argyrostoma was a function of the circadian system, and that the external coincidence model was a useful working hypothesis.
Experiments in which T = 48 h or longer
Formal experiments using light–dark cycles longer than 24 h fall into two main categories. The first is an extension of the night-interruption technique in which the greatly extended ‘nights’ of cycles as long as T = 48 or 72 h are so pulsed; these are commonly referred to as the Bünsow protocol after their originator (Bünsow, 1953). The second is the more widely used Nanda–Hamner experiment in which a short photophase (say 10 or 12 h) is coupled, in different experimental insect subsets, with a greatly variable scotophase to give cycle lengths ranging from perhaps 18–84 h (Nanda & Hamner, 1958). Such cycles are repeated throughout the insect's photoperiodic sensitive period, and the outcome (e.g. diapause incidence) noted, for each experimental group, at the end of the experiment. These experiments have been used in attempts to determine whether or not photoperiodic time measurement has a circadian component. For example, the scanning light pulse in Bünsow's protocol, or the main photophase in Nanda–Hamner experiments, ‘coming on’ at sequentially later times in the extended ‘night’ may produce results that reflect an on-going rhythm of light sensitivity (Vaz Nunes & Saunders, 1999; Saunders, 2002). Most experiments have been conducted using the Nanda–Hamner technique. Much of the present discussion will be limited to these, although Bünsow experiments will also be considered.
Results from Nanda–Hamner experiments commonly fall into two different categories. In one, sometimes called ‘negative’ responses, short-night effects (e.g. low diapause incidence, or virginopara production) occur in regimes with a scotophase of less than, say, 8 or 9 h. However, when the scotophase exceeds the critical night length, diapause incidence rapidly increases (or egg-laying aphid oviparae occur) until the resulting curve achieves a high nonrhythmic plateau. In so-called ‘positive’ responses, on the other hand, the curve after the initial upswing at the critical night length reveals peaks and troughs of diapause or nondiapause development with a circadian periodicity (Fig. 4). The period of this rhythm may be approximately 24 h, or as short as 19 h (Veerman & Vaz Nunes, 1980). It is generally held that this periodicity reflects an aspect of the circadian system in free-run, therefore showing the rhythm's value for its period (τ); computer analyses of circadian behaviour under these conditions certainly suggest this to be so (Saunders & Lewis, 1987b).
‘Negative’ responses to Nanda–Hamner and Bünsow experiments are frequently held to reflect nonoscillatory or hourglass-like responses, whereas ‘positive’ responses are thought to reflect, as outlined above, the participation of the organism's circadian system at some point in the response (Vaz Nunes & Saunders, 1999; Saunders, 2002). However, complications arise because some species show ‘positive’ responses in one set of conditions but ‘negative’ responses in others. For example, the flesh fly S. argyrostoma shows a clearly circadian response above about 20 °C, but an apparently nonoscillatory hourglass-like response at 16–18 °C (Fig. 4B) (Saunders, 1973, 1982). Furthermore, the beetle Pterostichus nigrita shows hourglass-like responses in strains from a higher latitude (Thiele, 1977), and the cabbage butterfly Pieris brassicae after a change in diet (Dumortier & Brunnarius, 1989).
In the cricket Pteronemobius nigrofasciatus, Masaki (1984) investigated three photoperiodic responses: egg diapause, wing form (microptery vs. macroptery) and the rate of nymphal development. In Bünsow experiments, the first two showed a clear circadian periodicity although nymphal development did not. However, statistical variance of mean nymphal development time showed a clear dependence on circadian ‘organization’, being relatively low when the scanning pulse fell approximately 24, 48 or 72 h after ‘dawn’, but high when it fell close to 36 or 60 h. Masaki suggested that photoperiodic responses could be due to the operation of an hourglass-like timer, but its outcome subjected to variance arising from circadian ‘interference’ to a downstream process. This may indeed be the case, particularly for nymphal development. However, the result with nymphal development is very similar to that described earlier for pupal eclosion in S. aryrostoma (Saunders, 1978), which was similarly attributed to interactions between the circadian system and out-of-phase disturbance from light pulses. Masaki's results clearly indicate the possibility of downstream effects on the circadian system, as in the Sarcophaga case, but do not provide persuasive evidence that the photoperiodic clock itself is noncircadian. No-one seriously doubts that eclosion is governed by a component of the circadian system, so why should Masaki's results point unequivocally to an hourglass?
Still other species, such as the spider mite T. urticae, may show clear signs of circadian involvement in some experiments, but results suggesting an hourglass timer in others. For example, Nanda–Hamner cycles reveal a clear circadian periodicity with an interpeak interval of approximately 19 h (Veerman & Vaz Nunes, 1980), whereas a T-experiment using one short light pulse per cycle provided no evidence for a circadian involvement (Vaz Nunes & Veerman, 1982), and results from so-called Veerman–Vaz Nunes experiments (Veerman & Vaz Nunes, 1987) were interpreted as strong evidence for a noncircadian or hourglass-like timer. Resolving these issues is clearly of importance in unravelling the mechanism of time measurement in photoperiodism.
The Veerman–Vaz Nunes protocol
Another experimental design using light–dark cycles longer than 24 h was devised by Veerman & Vaz Nunes (1987), ostensibly to differentiate between circadian and hourglass-like photoperiodic clocks. This design exposed organisms, during their sensitive period, to cycles of either LD 12 : 12 or LD 12 : 36 h. Because hourglass-like clocks, by definition, measure a long night only once, whereas a circadian based clock would reset itself in an extended dark phase, an hourglass would record the number of 12 or 36 h ‘nights’ equally, whereas a circadian-based clock would measure a 36 h ‘night’ twice. Diapause incidence with an hourglass-like timer should therefore be approximately one-half of that expected from a circadian-based clock. When applied to T. urticae, this was indeed the outcome, indicating the operation of an hourglass-like timer. However, application of this experimental design to the flies S. argyrostoma and C. vicina (Saunders & Lewis, 1988) produced results interpretable in terms of circadian involvement and, in the aphid, M. viciae, the archetypal ‘hourglass’, some results were consistent with an oscillator-based clock in which extended dark phases were seen as a sequence of long nights, measured modulo τ (Vaz Nunes & Hardie, 1993). This suggested that results from the Veerman and Vaz Nunes protocol were not always so easily interpreted.
Photoperiodic time measurement: hourglass or circadian?
The enormous volume of data generated by formal experimentation over several decades has led to a dichotomy of view as to the mechanism of night-length measurement in insect photoperiodism: hourglass or circadian. There are at least three possibilities.
First, organisms have two types of clock (hourglass and oscillator) with different mechanisms used to measure night length under different circumstances. Thus, in different species, described above, an hourglass-like clock may be used, for example, at lower temperature, higher latitude, with different diet, at different stages of development, or in different types of experiment. To the present authors, this seems unlikely. More probably, there is a single type of photoperiodic timer whose expression varies according to conditions. This first possibility is considered no further.
Second, night length is measured by a nonoscillatory hourglass timer in natural (24 h) cycles. The outcome of that time measurement may then be subjected to circadian influence on some downstream process during the sensitive period that is reflected in the Nanda–Hamner result. A ‘positive’ Nanda-Hamner profile with peaks and troughs of high and low diapause has nothing to do with the process of time measurement itself but may reflect operation of the ‘counter’ mechanism.
Third, night length in 24-h cycles is measured by an oscillatory (circadian) timer according to the rules of entrainment to such cycles. The ‘positive’ Nanda–Hamner profile represents the free-running activity of that oscillatory system through the sensitive period when the counter is operating, revealing among other things, its natural period (τ). The whole mechanism may be described in terms of entrainment to the train of light pulses making up the natural, or experimental, light–dark cycle. However, the photoperiodic oscillator(s) are separate (different) from those involved in the regulation of overt behavioural rhythms, presenting different properties of period and phase relationship to the light cycle; hence, the frequently encountered difficulties in using overt rhythms as ‘hands of the clock’.
The formal description of the photoperiodic clock as either hourglass-like or circadian is to some extent merely semantic if an hourglass-like timer is regarded as a damping circadian oscillator (Bünning, 1969; Lewis & Saunders, 1987; Saunders & Lewis, 1987a,b; Vaz Nunes & Hardie, 1987). In this view, the photoperiodic oscillator in some species persists (free-runs) in darkness for a few cycles with decreasing amplitude; in others damping is more rapid, leading to hourglass-like properties. Figure 5(A) shows such a model-generated oscillator, which damps rapidly below a threshold yet still maintains its period and phase in a suboptimal fashion before final extinction. Circadian oscillators involved in photoperiodic timing may thus show damping behaviour, and hourglass-like timers and more sustained, clearly oscillatory, timers may differ only in their damping coefficients. Experimental and theoretical evidence in support of this contention, and the operation of these damping oscillations as photoperiodic clocks, will be outlined below.
Damping circadian oscillator(s) as photoperiodic clocks
Organisms are replete with homeostatic mechanisms to maintain the constancy of the body through negative feedback control. Physiology is regulated by the comparison of some index of the state of the system with a required condition or inbuilt reference value. The difference or error between the reference value and the current condition determines the response. Homeostatic models based on these principles have been developed to explain the maintenance of the body (Cannon, 1932).
The role of feedback in homeostatic mechanisms within organisms is to eliminate high frequency oscillation in physiological variables. However, by deliberately introducing time delay in the feedback of information, it is possible to produce models of self-sustained oscillations at the circadian timescale. The period of the generated rhythms is typically about three to four times longer than the time delay, so to achieve circadian output the time delay needs to be 6–8 h. Under these circumstances, the control system is operating with out-of-date information and alternately overshoots and undershoots the reference value provided the system's response to the error is sufficiently robust. If the output of the system is feeble, the high amplitude oscillations cannot be maintained and the oscillations damp out after a few cycles. The external coincidence model for photoperiodic time measurement is based on a damped circadian clock of the orthopteran insect Hemideina thoracica (Gander & Lewis, 1979) in which the synthesis of proteins is regulated by the feedback system; this model had its origins in the Johnsson & Karlsson (1972) feedback model for circadian petal movements of the plant Kalanchoë blossfeldiana.
Biological oscillators are sensitive to external regulation by light and temperature pulses. According to the damped oscillator model, light destroys the protein. Under bright light, the protein concentration is held at a low level, the error is large and the protein's synthesis rate is maximal. When the light is removed, synthesis continues briefly at a high rate for the duration of the time delay and the concentration overshoots the reference value creating high amplitude oscillations prior to damping. Light cycles entrain the oscillator when the frequency and phasing of the external light cycle and internal biological rhythms are close to synchrony.
Details of the damping-circadian-oscillator model, as applied to photoperiodism, may be found in earlier papers (Lewis & Saunders, 1987; Saunders & Lewis, 1987a,b). The general properties of such oscillators, derived from a feedback control systems approach, are described in detail by Lewis (2002) and above. The essential feature of this model is the rhythmic production of a chemical (protein) c over time. If the synthesis rate (SR) of c is high, a self-sustained oscillation results; if SR is low, the oscillation damps below a threshold (Fig. 5A). A time delay in the oscillation regulates circadian period. The oscillator is also light sensitive since light destroys c at a rate proportional to its intensity. When light is removed, synthesis of c is then boosted to a higher amplitude. Trains of light pulses applied to the system reproduce all of the entrainment properties of a circadian system.
Although an ‘internal coincidence’ type of photoperiodic clock is possible, the model proposed by Lewis & Saunders (1987) suggests, as a working hypothesis, that long-night/short-night discrimination occurs according to the principles of ‘external coincidence’. A succession of long nights is then accumulated through the sensitive period to send development down the diapause pathway, whereas a succession of short nights leads to nondiapause development.
In 24-h light–dark cycles, the damped oscillator model happens to reveal essential similarities between a circadian and an hourglass-like device. As noted above, the responses to night-interruption experiments are easily described in terms of entrainment within the circadian system. However, since the oscillation is reset close to a constant phase (Ct 12) at the end of each photophase and then resumes its motion in darkness, it effectively measures night length in 24-h cycles as if it were an hourglass. The photophase serves to return the oscillation close to Ct 12 or, in terms of Lees' linear timer, to ‘turn the hourglass over’.
The characteristic fall in diapause incidence toward the left hand side of the photoperiodic response curve (Fig. 1) reflects the damping action of the oscillator(s) involved. In DD, the oscillators rapidly damp below threshold and no longer register ‘days in DD’ as either long or short; very short photophases (e.g. 2–5 h) are frequently too weak to maintain the oscillator's amplitude above threshold. All of these effects are closely reproduced by the damped-oscillator model (Saunders & Lewis, 1987a).
It is only when T is extended far beyond 24 h that the clearly circadian nature of the photoperiodic timer is revealed. Thus, Nanda–Hamner and Bünsow experiments reveal the free-running periodicity of the photoperiodic oscillator, with the interpeak interval showing the period τ for that part of the circadian system (Saunders & Lewis, 1987b). This is because, although the oscillator may damp below threshold in more prolonged dark phases, it may retain its phase and period. When light ‘comes on’ at the ‘correct’ phase (i.e. subjective day), the production of c is boosted above threshold more easily than when light falls at an ‘incorrect’ phase (i.e. subjective night) (Fig. 5B). As with T = 24 h cycles, all of these responses are reproduced by the damped oscillator model. In addition, since the diapause response is measured as a percentage within an experimental population, and within an individual insect may depend upon interactions between components of a multioscillator (multicellular) system, out-of-phase light pulses (e.g. those occurring at τ + ½τ h) may also cause a lack of synchrony between constituent parts of the timer mechanism, thereby contributing to a low incidence of diapause (Saunders, 1978; Masaki, 1984).
The damped-oscillator model suggests that the photoperiodic timer in the blow fly, C. vicina, has a period close to 24 h, and is damping, whereas that part of the circadian system regulating the locomotor activity rhythm has a period of approximately 22.5 h and is fully self-sustained (Saunders, 1997, 1998). Similar model predictions for the spider mite T. urticae suggest that the photoperiodic oscillator has a period of approximately 19 h and is rapidly damping, but a fairly high light sensitivity. This combination of factors would provide both the shape of the photoperiodic response curve in this species and the 19-h intervals between the peaks of high diapause in Nanda–Hamner experiments (Saunders & Lewis, 1987b).
T-experiments and symmetrical skeleton photoperiods in the ‘zone of bistability’ (see above) have failed to provide evidence, in T. urticae, for a circadian involvement in photoperiodism (Vaz Nunes & Veerman, 1997). However, this may have been due to the relatively weak (short) 1-h pulses of light used in these experiments. Although probably entraining the circadian oscillation thought here to be involved, 1-h pulses would only produce a narrow range of entrainment (Saunders, 2002) and such pulses falling at the extremes of this range might not boost the oscillation sufficiently. If longer or brighter pulses had been used, ‘positive’ results might have been obtained since a wider range of entrainment would have been engendered. This is certainly true of the T-experiment, since the left-hand side of the ‘positive’ Nanda–Hamner profile, using 8-h light pulses (Veerman & Vaz Nunes, 1980), is of the T-experiment type (Pittendrigh & Minis, 1971).
Results of the so-called Veerman–Vaz Nunes protocol (Veerman & Vaz Nunes, 1987) are also explicable in terms of a damping oscillator. In this experiment a series of LD 12 : 36 h cycles was found to be equivalent to a series of LD 12 : 12 h cycles, indicating that 36-h ‘nights’ were measured only once, similar to 12-h ‘nights’ (Veerman, 2001). This is, as the authors conclude, exactly what one would expect from an hourglass timer, or indeed from a circadian oscillator that damps below threshold in the latter part of a greatly extended dark phase.
Investigators who propose the action of an hourglass in photoperiodic time measurement are correct up to a point, but some fail to acknowledge that damped circadian oscillators may present the properties of an hourglass-like timer and therefore provide an alternative explanation for the phenomenon. Although it is not known for certain that photoperiodic time measurement is or is not a function of the circadian system until much more is known about the concrete events concerned, it is the authors' opinion that the wealth of experimental evidence may be interpreted to show that night-length hourglass-like timers are best regarded as damped circadian oscillators. This simple, single difference (in their damping coefficient) thus suggests that the two apparently different forms of time measurement are merely variants of a theme. Therefore, there is no need to invoke a completely different, noncircadian mechanism for night-length measurement, and we should seek resolution of the problem in the rapidly advancing studies on circadian molecular biology.
Circadian ‘clock’ genes
In the last two decades, enormous advances have been made in unravelling the molecular aspects of circadian rhythmicity (Hall, 1998; Dunlap, 1999; Rosato & Kyriacou, 2001), using established ideas of autoregulatory feedback loops and input–clock–output processes as models for their analysis. Here, some of the current views of these processes are outlined in order to introduce the major genes and proteins thought to participate in insect circadian rhythms. Having introduced these elements, it is then pertinent to ask if they are also involved in photoperiodic timing.
In the fruit fly Drosophila melanogaster, there are five recognized central ‘clock’ genes: period (per), timeless (tim), Clock (Clk), cycle (cyc) and doubletime (dbt) – with ‘fine-tuning’ provided by some others. Together with important genes operating on the input pathway [e.g. cryptochrome (cry)] and on output pathways [e.g. pigment dispersing factor (pdf), lark, takeout (to), etc.] these genes, and their proteins, provide the mechanisms thought to be required, at least in D. melanogaster, for generating the essentially negative feedback inherent in the circadian oscillator, the regulation of the circadian oscillator by light and the way, in turn, it controls rhythmic outputs. The current Drosophila model suggests that, during the photophase or the subjective day, the proteins CLOCK and CYCLE form CLK/CYC dimers that activate the transcription of per and tim. At the same time, CLK/CYC dimers inhibit transcription of the Clk gene. Transcription of per and tim gives rise to two proteins, PER and TIM, that form a complex in the cytoplasm. The formation of this dimer is delayed or accelerated by the action of DBT and TIM on PER stability. The PER/TIM complex then enters the nucleus during a specific window (Ct 19–21) to associate with the positive transcription factors encoded by Clk and cyc. Binding of CLK/CYC at this point removes transcriptional activation of per and tim and negative feedback is achieved. Light affects the system by degrading TIM through the agency of CRY (and possibly other routes).
Are circadian ‘clock’ genes involved in photoperiodic timing?
This review develops the idea that all insect photoperiodic clocks are based on a system of damping circadian oscillators, with the most highly damped examples presenting the properties of an hourglass-like timer. One may reasonably expect that known circadian ‘clock’ genes and their proteins could also play an important role in this form of time measurement. However, whether these elements function as in the currently accepted Drosophila model is a moot point because important differences seem to exist between D. melanogaster and other species (Sauman & Reppert, 1996) that may cast doubt on the general applicability of this model across a range of insects, let alone to photoperiodic time measurement. Nevertheless, these genes and proteins may be involved in photoperiodic timing, although ‘novel’ genes − or ‘known’ genes with novel uses − may also play a part.
There are at least three ways in which analysis of photoperiodic time measurement might proceed. These include: (i) determining the occurrence and kinetics of known ‘clock’ genes in species with robust photoperiodic responses; (ii) disruption of known ‘clock’ genes by genetic or molecular methods (gene ‘silencing’) to ascertain their possible role in photoperiodism; and (iii) molecular screening of the insect (i.e. Drosophila) genome using DNA microarrays, searching for other genes that might play a role.
Known ‘clock’ genes in photoperiodically important species
Two recent examples are considered: (i) flesh flies (Sarcophaga spp.) that show a robust pupal diapause induced by short days experienced by intrauterine embryos and the younger feeding larvae (Denlinger, 1971; Saunders, 1971); and (ii) the linden bug, Pyrrhocoris apterus, with a well-developed reproductive diapause most clearly expressed as an arrest of ovarian development in the adult female (Hodek, 1968; Saunders, 1983).
Working with S. crassipalpis, Goto & Denlinger (2002) examined expression patterns of the ‘clock’ genes period, timeless, cycle and cryptochrome in the heads of adult flies reared and maintained under long (LD 15 : 9 h) or short (LD 12 : 12 h) days. Daily rhythms of per and tim mRNAs were detected with peaks in the scotophase; cry and cyc mRNA abundance, on the other hand, remained fairly constant across the light–dark cycle. The duration of the photophase was also found to affect per and tim expression. Under short days (LD 12 : 12 h), the peak of per expression occurred at Zeitgeber time (Zt) 12–15, but was shifted to Zt 18 under LD 15 : 9 h. For tim mRNA, peak expression was also delayed under long days and, in addition, much reduced in amplitude when compared to that under LD 12 : 12 h.
In the linden bug Pyrrhocoris apterus, Hodkova et al. (2003) examined period gene expression under long-day (LD 18 : 6 h) and short-day (LD 12 : 12 h) photoperiods in both a wild-type and a nondiapausing strain. Unlike S. crassipalpis, no robust daily oscillations of per expression were observed in either photoperiod although levels of per mRNA were up to 10-fold higher under LD 12 : 12 h in the wild-type (diapausing) strain, suggesting a possible role of per in diapause expression.
These patterns are to some extent consistent with the molecular model outlined above, but must be regarded as preliminary. They were conducted with adult heads, including compound eyes, so that only a small proportion of the per activity detected may be attributable to diapause inductive events occurring in the brain. In addition, although adult heads of Sarcophaga are relevant to photoperiodic induction (Henrich & Denlinger, 1982), they are less so than brains of the embryonic stages and may therefore reflect adult locomotor rhythmicity more than diapause induction. Both studies also concerned just two photoperiods, and none in continuous darkness which might have revealed free-running and possibly damping behaviour. Nevertheless, in S. crassipalpis, the results give a tantalizing glimpse of a rhythm of per and tim expression that is phase set by the light cycle, probably also affected by light acting upon the TIM protein. In P. apterus, on the other hand, the lack of clear diurnal oscillation may reflect the indistinctly circadian response in Nanda–Hamner experiments or the unusual predominance of day length over night length in this species (Saunders, 1987a), but may suggest a possible additional role of PER in long-night summation. Further studies of this type, using the photoperiodically sensitive stages of insects with a robust diapause response are clearly needed.
Disrupting the action of known ‘clock’ genes
Diapause induction in period mutants of Drosophila melanogaster. The discovery of an ovarian diapause in some strains of D. melanogaster (Saunders et al., 1989, 1990) opened the possibility of using techniques in fly genetics to ascertain the possible role of circadian ‘clock’ genes in photoperiodic time measurement. Specifically, three mutant alleles of period, affecting both adult locomotor rhythms and pupal eclosion, were known by the late 1980s. These were perS with a shorter than normal period (τ approximately 19 h), perL. with a longer than normal period (τ approximately 29 h), and an apparently arrhythmic mutant, perO (Konopka & Benzer, 1971).
If these mutations were to affect the circadian oscillations thought to participate in photoperiodic time measurement in a similar period-altering fashion, marked effects on the photoperiodic clock would probably result. According to circadian theory (Pittendrigh, 1981; Saunders, 2002), when a biological oscillation (τ hours) becomes entrained to a light cycle (T hours), τ is ‘corrected’ to T every day by a phase shift equal to T − τ h. In doing so, the biological oscillation achieves a characteristic steady-state phase relationship to the light cycle: when τ is less than T, the oscillation phase-leads the light, and when τ is greater than T, it phase lags it. Such differences might have predictable effects on the critical photoperiod. In terms of a simple, external coincidence type of model, a short-period oscillation might produce a shorter critical night length − or a longer critical day length − whereas a long-period oscillation might produce the opposite. The effect of an arrhythmic mutation (perO) would be harder to predict, but there would surely be some impact.
When flies of the Canton-S strain carrying period mutations were examined for their diapause responses, females of perS and perL2 were found to show an identical critical day length to wild-type. Behaviourally arrhythmic females carrying a perO were also apparently able to distinguish short from long days, but with a critical day length a few hours shorter than wild-type. An even shorter critical day length was noted for females carrying a ‘double deletion’ of period (per–), which were both behaviourally arrhythmic and, as expected, producing no per mRNA. These results were taken as evidence that the period gene played no crucial role in photoperiodic time measurement (Saunders, 1990).
However, there are strong caveats to such a conclusion. The ovarian diapause in D. melanogaster is, at best, a very weak response to photoperiod, only observed in some strains at temperatures below approximately 14 °C (Saunders et al., 1989; Saunders & Gilbert, 1990). From studies on the locomotor activity rhythm of wild-type and per mutant flies (Konopka et al., 1989), it is known that the period of perL shortens with a fall in temperature (to 15 °C), whereas that of perS lengthens. Although studies on locomotor rhythmicity could probably not be performed at 12 °C (because of the flies' inactivity at this temperature), it is possible that the periods of perL and perS become indistinguishable from that of per+ at approximately 12 °C. This suggests that photoperiodic responses might also become indistinguishable at this temperature. In addition, phase response curves for locomotor activity rhythms in period mutants of D. melanogaster (Saunders et al., 1994) indicated that changes in τ may be attributed to shortened or lengthened phases during the subjective day, whereas the subjective night remained the same. Because the insect photoperiodic clock usually measures night length rather than day length (Saunders, 2002), these observations also suggest why critical night lengths remain the same in the period mutants. In short, despite the above results, the period gene and gene product may play an important role in photoperiodic timing.
However, the possibility that the period gene in D. melanogaster plays no essential role in diapause induction, has prompted consideration of other circadian components in this phenomenon. For example, Helfrich-Förster (2001) found evidence for a per-independent component in the adult activity rhythm. In the bimodal rhythm characteristic for this species, the evening peak was clearly regulated in part by per, because the peak attained appropriate phase-leading and phase-lagging phase relationships to the light cycle in perS. and perL., respectively. However, the morning peak did not; it also persisted, at least for a few cycles in some examples of ‘arrhythmic’perO. flies suggesting its ‘independence’ from period, possibly as a damping oscillator. Helfrich-Förster (2001) suggested that this per-independent oscillator might be involved in photoperiodic timing as well as the morning activity peak.
Double-stranded RNA interference
In a speculative essay on possible molecular mechanisms in insect photoperiodism, Tauber & Kyriacou (2001) drew attention to other ‘clock’ genes, particularly timeless and cryptochrome whose possible role could be investigated using tim and cry mutants. They also suggested that recent developments using double-stranded RNA interference (dsRNAi) technology, perhaps in insects with a more robust photoperiodic response than D. melanogaster, might reveal a role for such ‘clock’ genes in photoperiodism.
The dsRNAi technique has now been applied to timeless in the drosophilid Chymomyza costata (Pavelka et al., 2003). This species has a robust larval diapause induced by short days, and results from formal experiments indicate a role for the circadian system in photoperiodism (Yoshida & Kimura, 1995; Kostal et al., 2000). Pavelka et al. (2003) used two strains of C. costata: a wild-type in which short days induced diapause and was behaviourally rhythmic, and a nonphotoperiodic strain (npd), which lacked a photoperiodic response and was arrhythmic. When timeless expression was disrupted by injecting wild-type embryos with tim dsRNA, a proportion of the surviving larvae was found to be photoperiodically neutral, similar to npd mutants, failing to enter diapause under short days. However, a low rate of survival of injected embryos, a low rate of ‘positive’ responses and a relatively high rate of nonspecific responses to the buffer used as a control, indicated the need for caution. Nevertheless, this result suggests that timeless may indeed play an important role in photoperiodic time measurement.
Molecular screening of the insect genome
Three recent studies have used DNA microarrays to uncover genes exhibiting circadian properties in D. melanogaster (Claridge-Chang et al., 2001; McDonald & Rosbash, 2001; Ueda et al., 2002). These have revealed a large number (possibly up to 4% of the genome) of ‘oscillating’ genes including those already identified as bona fide‘clock’ genes (see above), but also a host of others involved in many biochemical and physiological processes, and still others with an unknown function. This suggests that circadian rhythmicity is a widespread and pervading phenomenon. However, screening for these genes was conducted by looking for those whose oscillatory output was self-sustained; perhaps one should also seek those whose output becomes damped after a short time in darkness? Genes should also be screened for those that are differentially expressed under long or short days, although this would undoubtedly reveal many ‘diapause-related’ genes (Denlinger, 2002) that are involved at many other layers of the complicated seasonal response.
A molecular model of the insect photoperiodic clock?
It is tempting at this point to suggest a molecular model for photoperiodic timing based on the action of Drosophila‘clock’ genes and the damped oscillator control systems model described above (Warman & Lewis, 2001; Lewis, 2002). In this view, the synthesis of protein c is seen as the formation and accumulation of the PER/TIM complex, and the time delay necessary for oscillation and its period, the result of the action of DBT on PER stability. According to the principles of external coincidence, light would have a dual role: (i) entrainment of the constituent oscillation(s) through the agency of cryptochrome and (ii) operating the long night–short night switch in metabolism using a carotenoid-based system of photoreceptors (Hardie & Vaz Nunes, 2001; Veerman, 2001).
The concept that circadian ‘clock’ genes could be involved in the control of a variety of rhythmic behaviours is not without precedent. Kyriacou & Hall (1980) showed that mutations in the per gene result not only in alterations to circadian locomotor activity and the gating of eclosion, but also influence the period of the ultradian mating song rhythm.
If circadian ‘clock’ genes are supplying the molecular machinery for a damped circadian oscillator that is distinct from the robust oscillator controlling locomotor activity, the most plausible way by which this could occur is via these genes acting in a different, damped manner to produce damping protein oscillations in a tissue that is anatomically distinct from the circadian pacemaker(s) controlling locomotor and eclosion behaviour.
With the increasing development of ‘antisense methodologies’ such as peptide nucleic acids (PNAs; Koppelhus et al., 2002), morpholinos (Heaseman, 2002) and double-stranded RNA interference (Martinek & Young, 2000), comes the ability to study the roles of specific genes in generating photoperiodic behaviour. However, trying to examine expression levels of a damped oscillator that has the same molecular components as more robust circadian oscillators, but in different tissues, is virtually impossible without a prior knowledge and isolation of the relevant tissue.
The ultimate test of this hypothesis therefore relies on tissue-specific alterations in the functioning of genes crucial to the generation of locomotor behaviour such as per or tim. This type of tissue specific ‘knock down’ of clock genes has been demonstrated in Drosophila (Martinek & Young, 2000) and has been used with great effect to examine the involvement of different photoreceptors and genes in controlling circadian behaviour in D. melanogaster by Helfrich-Förster (Foster & Helfrich-Förster, 2001). However, before this type of approach is employed, progress needs to be made in identifying potential target tissues for tissue-specific alteration of gene function.
This review develops the idea that photoperiodic responses of insects, as in other major taxa, are functions of the circadian system. In many species, the manifestations of photoperiodic induction in a wide range of simple and complex photocycles are so remarkably similar to those in circadian entrainment studies that the probability of this association cannot be ignored. However, many aspects of insect photoperiodism suggest that the constituent oscillator(s) are damping rather than fully self-sustained, and that species with more extremely damped oscillators show properties of an hourglass-like timer that fails to persist in longer periods of darkness without periodic (daily) boosts from the photophase. Simple computer models based on damping oscillators and the phenomenon of entrainment are sufficient to ‘explain’ such responses without recourse to a totally different, noncircadian, night-length timer.
If this view of photoperiodism is indeed true, seasonal regulation of diapause and the control of circadian rhythms may be based on feedback loops using the same, rather than different, sets of genes and their proteins. Just as progress in unravelling the molecular aspects of circadian rhythmicity was guided by the canonical properties of the circadian system (free-running behaviour, temperature compensation, entrainment to light, etc.), so investigations of the molecular basis of photoperiodism should pay full regard to the array of results from formal studies. These include the full range of photoperiods comprising the photoperiodic response curve, the evidence for damping in studies using extended light–dark cycles, and the abundant evidence for circadian oscillator entrainment in photoperiodic phenomena. Damped circadian oscillator models may thus assist future explorations of this important aspect of insect developmental biology.