Variation in chronotype is associated with migratory timing in a songbird

Abstract

Like many organisms, birds exhibit daily (circadian) and seasonal biological rhythms, and within populations both daily and seasonal timing often vary among individuals. Because photoperiod interacts with the circadian rhythms of many organisms to induce seasonal changes in behaviour and physiology, it is hypothesized that differences in daily timing, called chronotypes, underpin differences among individuals in the timing of seasonal events. For seasonal events stimulated by increasing daylength, this hypothesis predicts a positive relationship between the timing of daily and seasonal activities of individuals, with advanced chronotypes expressing events earlier in the year. The few previous tests of this hypothesis have focused on seasonal reproductive timing in birds. However, the hypothesis predicts that this relationship should extend to other photoinduced seasonal events. Therefore, we tested whether variation in chronotype was associated with variation in spring migratory timing in a captive songbird model, the pine siskin (Spinus pinus). We found that pine siskins expressing migratory restlessness exhibited repeatable chronotypes in their timing of nocturnal activity. Further, chronotype was significantly associated with the onset date of migratory behaviour, consistent with the hypothesized relationship between chronotype and seasonal timing.

1. Introduction

Many organisms respond to the procession of day and night with daily rhythms of behaviour and physiology. These rhythms are driven by circadian clocks—endogenous pacemakers with periods that approximate 24 h which are entrained by temporal cues in the environment [1]. That is, the clock advances or delays to assume a stable phase relationship with temporal cues (i.e. zeitgebers)—typically light, although other cues can also act as zeitgebers [2–4]. The phase of entrainment between circadian clock and zeitgeber can vary between individuals, leading to individuals expressing circadian rhythms that are consistently advanced or delayed compared with the solar day [5]. These differences in circadian phenotype are described as ‘chronotypes’. While the physiological mechanisms underlying chronotype remain poorly understood [6,7], variation in chronotype within populations is described in humans, rodents, insects, fish and birds [5,7–14].

Beyond coordinating daily rhythms, circadian clocks also provide a reference for detecting changes in photoperiod across the year and, consequently, a means for anticipating seasonal changes [15]. Diverse organisms employ circadian clocks in this twofold role [16–18], including birds [19–21]. For many birds, changes in photoperiod stimulate seasonal events (e.g. spring migration and reproduction) [22–24]. Like chronotype, seasonal timing can also both vary within populations [25–27] and be a repeatable trait of individuals [25,26,28].

The dual role of circadian clocks in coordinating behaviour within days and years has led to the hypothesis that variation among individuals in the intrinsic period of their circadian clocks (i.e. the period of a circadian rhythm in the absence of zeitgebers) contributes to differences in seasonal timing of events stimulated by increasing daylengths (hereafter, ‘spring’ events) [12,13]. Although chronotype has been used previously as an indicator of intrinsic period, chronotype itself, as an expression of an organism's entrained circadian clock under natural conditions, may be of greater biological relevance [29]. Therefore, we offer a modified hypothesis: variation in the entrainment of circadian clocks, indicated by chronotype, contributes to variation in seasonal timing of spring events. This hypothesis posits that different chronotypes will experience light during different phases of their circadian rhythm and will consequently perceive different daylengths (e.g. advanced chronotypes will perceive sunset to occur later in the day, indicating a longer day). If this hypothesis is correct, then different chronotypes will sense photoinductive daylengths on different dates, such that advanced and delayed chronotypes will express spring behaviours earlier and later in the year, respectively. Thus far, this prediction has been tested in two species by examining the relationship between daily timing and reproductive timing. A study of captive great tits (Parus major) found no significant relationship between egg lay date and chronotype [12]. But a study of free-living great tits and dark-eyed juncos (Junco hyemalis) found that chronotype was significantly associated with egg laying date in both species [13].

Although prior tests of the hypothesis have focused on seasonal reproduction, it applies to other photoperiod-dependent seasonal events. Therefore, this study tested whether chronotype is associated with variation in the timing of spring migratory behaviour in the pine siskin (Spinus pinus, hereafter, ‘siskin’). Although siskins have spatially and temporally flexible migratory patterns [30,31], they exhibit a period of spring nomadism stimulated, at least in part, by increasing photoperiod [32]. During spring nomadism, captive siskins express nocturnal migratory restlessness (i.e. Zugunruhe) [33] similar to that described in predictable seasonal migrants. This behaviour is characterized by quiescence in the early night followed by elevated locomotor activity [33]. Here, we used nocturnal migratory restlessness to quantify both individual chronotype and seasonal timing of migratory behaviour. Although previous studies have used the onset time of morning activity to characterize chronotype, the onset of morning activity is influenced by the intensity of light experienced at or preceding dawn [13,14]. Therefore, we instead used the onset time of nocturnal activity to measure chronotype independent of variation in light intensity. To develop this measure of chronotype, we assessed whether individuals exhibited consistent (i.e. repeatable) onset times of nocturnal activity. We then examined whether chronotype was associated with variation in the onset date of migratory restlessness. By testing this previously unconsidered prediction, we can begin to assess the generality of the hypothesized relationship between chronotype and the timing of seasonal spring behaviours.

2. Material and methods

We provide a more detailed description of the methods in the electronic supplementary material.

(a). Animals

Data for this study came from three previous experiments investigating the use of environmental cues to time spring migration: the ‘photoperiod’ and ‘timing’ experiments from Robart et al. [32] and the ‘males-only’ experiment from Robart and Watts (A. R. Robarts, H. E. Watts 2017, unpublished data). In all experiments, wild-caught siskins were held in individual cages on photoperiods mimicking naturally increasing daylengths at a north-temperate latitude (table 1). Birds were fed ad libitum and housed such that they could hear, but not see, conspecifics.

Table 1.

Details of experimental groups of pine siskins.

experimental group	sample size	dates of data collection	photoperiod (latitude mimicked)	nights of activity data per bird	reference
photoperiod	11 (3 females, 8 males)	21 Dec 2015–2 Jun 2016	34°N	$\overline{x}=68.2$ (range: 9–150)	[32]
timing	6 (3 females, 3 males)	21 Feb 2017–23 Apr 2017	42°N	$\overline{x}=19.7$ (range: 9–29)	[32]
males-only	6 (6 males)	21 Feb 2017–11 May 2017	42°N	$\overline{x}=44.7$ (range: 12–62)	Robart & Watts unpublished data

(b). Activity data

All experiments began in winter and continued over the period when spring nomadic migratory transitions typically occur [32,33], though the timing of each experiment varied (table 1). Activity of individual birds was recorded continuously by a system of passive infrared motion sensors mounted on each cage and connected to a VitalView Data Acquisition System (Starr Life Sciences, Oakmont, PA); movements were summed in 10 min intervals.

To determine when each bird began to express spring migratory restlessness (seasonal timing), we examined activity data recorded between 23.00 and 3.00, when siskins typically express nocturnal migratory restlessness [33]. Based on prior studies [32,33], we considered a bird to have expressed migratory restlessness on a given night if at least seven 10 min intervals contained greater than or equal to 10 movements. The onset date of migratory restlessness was defined as the first night of a greater than or equal to 5-consecutive-night period of migratory restlessness. This measure conservatively excluded predawn activity and isolated nights of elevated activity. Only birds that met this criterion were included in analysis, resulting in a sample of 16 males and seven females.

To characterize chronotype (daily timing), we used the onset time of nocturnal activity relative to lights-off for each night after a bird began expressing migratory restlessness. For this measure, we aimed to capture activity broadly across the night. For each night, we examined activity data between the onset of quiescence (defined in [33]) and 30 min prior to lights-on. We defined the onset of nocturnal activity as the first time during this period when activity levels increased to an average of greater than or equal to 10 movements per 10 min interval for at least 30 min.

(c). Statistical analyses

Our general approach was to use linear mixed models implemented in R 3.5.0 [34] with the lme4 package [35]. We visually inspected plots of residuals to check for deviations from normality and homoscedasticity and estimated random effects for deviations from normality. We tested for model effects using Satterthwaite's method in the lmerTest package [36], and report estimated coefficients (β) and standard errors for fixed effects.

We calculated repeatability for the onset time of nocturnal activity in a linear mixed model with onset time of nocturnal activity as the dependent variable, experiment and days since winter solstice (hereafter, ‘date’) as fixed effects, and bird identity as a random effect (with random slopes with respect to date). We also tested for the effect of sex and the interaction of date and experiment; however, neither had a significant effect (all p-values ≥ 0.19) and they were removed from the model. Repeatability was calculated with the rptR package [37], using 1000 bootstrap samples to obtain a 95% confidence interval and p-value.

We tested for an association between chronotype and the onset date of migratory restlessness using a linear mixed model with onset time of nocturnal activity—our measure of chronotype—as the dependent variable, the onset date of migratory restlessness, experiment and date as a fixed effects, and bird identity as a random effect (with random slopes with respect to date). Again, we tested for the effect of sex and the interaction of date and experiment, but we excluded them from the final model as they had no significant effect (all p-values ≥ 0.20).

3. Results

The onset time of nocturnal activity was repeatable (repeatability = 0.56, 95% CI = 0.38, 0.67, p < 0.0001; figure 1). Nocturnal activity began earlier in the night relative to lights-off as the year progressed (date: β = −53.42, s.e. = 8.22, t_15.35 = −6.50, p < 0.0001). Timing did not vary significantly across experiments; compared with the birds from the photoperiod experiment, there was a trend for birds in the males-only experiment to become active earlier at night (β = −49.43, s.e. = 25.49, t_19.99 = −1.94, p = 0.067), but no such trend existed for the birds in the timing experiment (β = 8.10, s.e. = 27.96, t_22.81 = 0.29, p = 0.77).

Figure 1.

Nightly onset times of nocturnal activity (squares) for individual birds, ordered by conditional means of nocturnal activity onset (triangles). Conditional means describe the relative timing of individual birds compared with the rest of the sample population while accounting for the effects of experiment and date.

We found a significant, positive relationship between the onset date of migratory restlessness and the onset time of nocturnal activity (figure 2; β = 35.42, s.e. = 8.50, t_14.87 = 4.17, p = 0.0008), indicating that individuals initiating migratory restlessness later in the year also became active later at night. As before, we found that nocturnal activity began earlier in the night as the year progressed (date: β = −55.08, s.e. = 8.20, t_16.01 = −6.72, p < 0.0001). Birds in the males-only experiment became active significantly earlier than birds in the photoperiod experiment (β = −51.71, s.e. = 18.65, t_17.75 = −2.77, p = 0.013), while the onset timing of birds in the timing experiment was not significantly different from the photoperiod experiment (β = −3.39, s.e. = 22.79, t_26.63 =−0.149, p = 0.88).

Figure 2.

Relationship between the onset time of nocturnal activity and the onset date of migratory restlessness. For visualization, individual conditional means of onset time are plotted against onset dates with a linear regression. Symbols indicate experimental group: photoperiod (triangles), timing (squares) and males-only (circles).

4. Discussion

We found that siskins exhibit repeatable chronotypes in their timing of nocturnal activity and that these chronotypes are significantly associated with the onset date of migratory restlessness. Birds that commenced activity earlier in the night also expressed migratory restlessness earlier in the year, consistent with the hypothesis that differences in entrained circadian rhythms, indicated by chronotype, are connected to differences in the timing of spring events. The patterns of nocturnal activity shown by captive siskins in this study are consistent with data from migrating birds under more naturalistic conditions. Individual consistency in the onset time of nightly activity has also been shown in common redstarts (Phoenicurus phoenicurus) held in temporary captivity during migratory stopovers [38]. Further, our finding that siskins began their activity earlier relative to lights-off as the year progressed matches the behaviour of free-living birds, which depart on migratory flights earlier in the night as spring progresses [39]. It remains to be determined how nocturnal chronotype characterized here relates to diurnal chronotype; there is evidence that diurnal and nocturnal activity may be controlled by different circadian oscillators with different intrinsic periods [40].

The relationship we found between chronotype and migratory timing echoes the association between reproductive timing and chronotype observed by Graham et al. [13]. Together, these studies suggest a robust relationship between chronotype and the timing of avian seasonal events, which holds across different spring events and songbird species. Another previous study found no relationship between either chronotype or free-running circadian period and egg lay date [12]. However, that study had limited data with which to characterize chronotype [12], and free-running rhythm is only one contributing factor to the entrained rhythm of an organism [6]. Additional evidence for a link between variation in circadian rhythms and seasonal timing comes from studies of the circadian gene Clock, which indicate that both polymorphism and methylation at this locus can, in some cases, predict variation in seasonal timing [41–43] (cf. [44]). However, one important next step in this research will be to establish whether differences in chronotype are causally linked to variation in seasonal timing.

Variation in chronotype and seasonal timing can both impact fitness [29,45–49]. The relationship shown here between these traits further suggests that selection on one trait may affect the other. However, studies of insects and rodents indicate that chronotype and seasonal timing can evolve independently [50–53]. As a rapidly changing climate makes timing mismatches between seasonal behaviours and environment more likely [54,55], elucidation of the physiological mechanisms underlying the relationship between the circadian clock and the timing of seasonal events will be essential to understanding how organisms might respond to this change.

Ethics

Birds were collected under scientific permits from the US Fish and Wildlife Service, and the states of California, Oregon, Washington and Wyoming. All procedures were approved by the Loyola Marymount University Institutional Animal Use and Care Committee (protocol: LMU IACUC 2014 FA 02).

Data accessibility

Data available from Dryad Digital Repository at: https://doi.org/10.5061/dryad.p226m8r [56].

Authors' contributions

H.E.W. conceived the study. A.R.R. and J.L.R. collected the data. J.L.R. and H.E.W analysed the data. J.L.R. wrote the first draft of the manuscript. All authors contributed to revising the manuscript, approved the final version, and accept responsibility for its content.

Funding

This work was supported by NSF IOS: 1456954/1756976; Santa Monica Bay Audubon Society; Pasadena Audubon Society; WSU College of Arts and Sciences.

Competing interests

We declare we have no competing interests.