Circadian clock properties of fruit flies Drosophila melanogaster exhibiting early and late emergence chronotypes.

The role of circadian clocks in timing daily behaviors is widely acknowledged, and while empirical evidence suggests that clock period is correlated with the preferred phase of a rhythmic behavior (chronotype), other clock properties have also been hypothesized to underlie chronotype variation. Here, we report that fruit fly Drosophila melanogaster populations exhibiting evening emergence chronotype (late) are characterized by higher incidence of behavioral arrhythmicity in constant dim light, wider range of entrainment, reduced rates of re-entrainment to simulated jet-lag and higher amplitude of both entrained and free-running rhythms as compared to those exhibiting morning emergence chronotype (early). Our results thus highlight the role of circadian clock properties such as zeitgeber sensitivity, amplitude and coupling in driving chronotype variation.


Introduction
Entrainment of circadian rhythms to environmental cycles (zeitgebers) is characterized by establishment of a stable and reproducible phase relationship (ψ ent ) with the zeitgebers, and chronotypes are characterized by the variation in ψ ent which is widely observed across a variety of species (Daan and Aschoff, 1975;Duffy et al., 2001;Rémi et al., 2010). Clock period (τ) is correlated with ψ ent such that individuals with shorter τ exhibit an advanced ψ ent ("early" chronotypes) and those with longer τ exhibit a delayed ψ ent ("late" chronotypes) (Aschoff, 1965;Aschoff and Pohl, 1978;Rémi et al., 2010;Roenneberg, 2012;Wright et al., 2005). However, attempts to explain this association using classical models of entrainment (Aschoff, 1979;Pittendrigh, 1981b) have only been partially successful, thus invoking a possible role of other clock properties. Furthermore, even though several physiological and behavioral differences have been found to be associated with morningness-eveningness (chronotypes) in humans (Dijk and Lockley, 2002), unlike the assessment of ψ ent in other species which rely primarily on controlled experiments, the same in humans is highly variable due to a plethora of reasons, and thus correlations of chronotypes with other circadian clock properties are generally weak (reviewed in Levandovski et al., 2013), thereby rendering them less reliable for asserting circadian regulation of ψ ent .
Previously, laboratory selection and latitudinal cline studies have assessed strains of Drosophila pseudoobscura (Pittendrigh, 1967), Pectinophora gossypiella (Pittendrigh and Minis, 1971), Drosophila auraria (Pittendrigh and Takamura, 1987) and Drosophila subobscura (Lankinen, 1993) differing in ψ ent (reviewed in Vaze & Sharma, 2013). These studies have greatly motivated conceptualization of circadian clocks as a network of coupled oscillators (reviewed in Bell-Pederson et al., 2005), and have been instrumental in exploring properties of circadian clocks associated with ψ ent variation, some of which are discussed below.
Among the properties known to influence ψ ent are the strength or amplitude of the zeitgeber (A z ) and intrinsic amplitude of the clock (A o ) (Brown et al., 2008;Pittendrigh, 1981a;Pittendrigh et al., 1991;Vitaterna et al., 2006), such that higher A z /A o leads to larger phase-shift incurred by the clock, and consequently a wider entrainment range. In other words, low-amplitude circadian oscillators would have higher zeitgeber sensitivity. Additionally, magnitude of coupling between the constituent oscillators of circadian clock also influence ψ ent (Abraham et al., 2010;Granada et al., 2013;Pittendrigh, 1981a;Pittendrigh et al., 1991). "Coupling" refers to the interactions between individual neurons or oscillators that form a network referred to as the circadian clock, and not to that between clock and zeitgeber, input or output pathways. Another clock property studied in this regard is the relaxation rate, which refers to the propensity of the clock or oscillator to converge to its intrinsic limit cycle amplitude (A o ) following perturbations. Oscillators with higher relaxation rates recover faster, and are therefore termed rigid, while those with lower relaxation rates take longer to relapse to their intrinsic limit cycle, and are termed weak oscillators (Abraham et al., 2010;Granada et al., 2013;Granada and Herzel, 2009). Furthermore, mathematical models based on generic Poincaré, Hopf and Becker-Weimann-Bernard oscillators proposed that relaxation rate differences among oscillators can influence amplitude, rates of re-entrainment, entrainment range and ψ ent (Abraham et al., 2010;Becker-Weimann et al., 2004;Bernard et al., 2007;Bordyugov et al., 2011;Granada et al., 2013;Guckenheimer and Holmes, 1983).
In this study, we used fruit fly D. melanogaster populations which are products of a long-term (over 14-years) laboratory selection study for morning (early) and evening (late) adult emergence to explore the circadian clock properties underlying ψ ent . In response to the selection imposed, the early populations evolved shorter τ (of emergence rhythm) and advanced ψ ent , while the late populations evolved longer τ and delayed ψ ent and also shorter and longer τ of activity-rest rhythm, respectively (Kumar et al., 2007a), implying that the diverged ψ ent stems from a common central circadian clock governing the two rhythms. Based on studies under different photoperiods, we had reported that the early and the late populations might have evolved dominant morning (M) and evening (E) oscillators, respectively (Kumar et al., 2007b). A closer look at the same data revealed that across photoperiods ranging from light-dark (LD) 8:16 to LD16:8, ψ ent of the early populations changed by 2 h, while that of the late populations by 4 h (twice the magnitude), indicating phase-sensitivity differences, as also substantiated by~42% greater standard deviation in ψ ent across photoperiods in the late as compared to the early populations. In another study, we observed that when T = 24 h, and therefore for a period mismatch of 0.4 h for the early (τ = 23.6 h) and 0.2 h for the late (τ = 24.2 h) populations (Kumar et al., 2007a), the difference in ψ ent between the two varied considerably from 1.71 h to 5.8 h depending on zeitgeber conditions (Nikhil et al., 2014). Even though it is well known that small variation in τ can lead to large variation in ψ ent (Merrow et al., 1999;Ouyang et al., 1998;Ralph and Menaker, 1998), high phase sensitivity and large variation in the difference of ψ ent between populations for same τ-T mismatch was intriguing, and suggested differential entrainment, as also highlighted by Vaze et al. (2012). Therefore, in light of the propositions by Pittendrigh (1981a) regarding association of clock properties with ψ ent variation, and that from the above-discussed observations by Kumar et al. (2007aKumar et al. ( , 2007b, Vaze et al. (2012) and Nikhil et al. (2014), we hypothesized that zeitgeber sensitivity and amplitude of circadian oscillators might be associated with ψ ent differences between the early and the late populations, in addition to differences in τ.
In this regard, we initially tested if the early and the late populations have evolved differences in light sensitivity (the populations were selected under LD12:12 at constant temperature and humidity, rendering light as the only zeitgeber) by assaying activity-rest rhythm under constant light (LL), and further explored the possible involvement of other above-mentioned clock properties as well, the results of which will be discussed later.

Experimental populations
We used D. melanogaster populations comprising three sets of large (~1200 individuals; male:female ratio~1), out-breeding populations (early, control and late) with four replicates (henceforth referred to as blocks) each, thus totaling 12 populations. Each of the early and the late blocks (early i and late j , where i = j = blocks 1...4) was initiated from a control k (k = 1...4) block sharing the same subscript, and therefore every block of the early and the late population originated from a common control block. All populations were maintained on a 21-day non-overlapping generation cycle. Every generation, the parental populations were provided with yeast paste for 3 days following which eggs were collected for next generation and seeded into culture vials containing~6 ml Banana-Jaggery media (~300 eggs per vial). From the initiation of emergence, flies that emerged between zeitgeber time (ZT) 21-01 (ZT00 represents the time of lights-ON under 12:12 h lightdark cycles -LD12:12) in the morning were collected for four consecutive days to form the early populations and those emerging in the evening (ZT09-13) for the same four consecutive days formed the late populations. The control populations constituted flies emerging across 4 days and thus did not experience any selection. All populations were selected and maintained in LD12:12 (with light intensity of~0.4 Wm −2 ) at 25 ± 0.5°C temperature and 75 ± 5% relative humidity. The populations had experienced selection for over 220 generations (>14 years) when the study was performed. All experiments were performed on the progeny of standardized populations, a description of which along with other relevant details can be found elsewhere (Kumar et al., 2007a).

Activity-rest recording
All activity-rest recordings were performed on 4to 5-day-old virgin males in activity tubes with sucrose-yeast media using Drosophila Activity Monitors (DAM) (Trikinetics, Waltham, MA, USA). With the exception of light regimes, environmental conditions for all experiments were same as described for population maintenance.

Recording in constant light (LL)
Recording in LL was performed under light intensities of 4 × 10 −3 Wm −2 and 4 × 10 −4 Wm −2 . Flies were initially subjected to 4 days of LD12:12 with light intensity during the day same as that under the LL regime, and then recorded in LL for 12 days. Block-wise percentage of flies exhibiting free-running rhythm (presence of a single statistically significant period), complex rhythm (presence of two or more statistically significant periods) and arrhythmicity (absence of any statistically significant period) in LL were then calculated. Statistical analyses of percentage values were implemented using a randomized block design mixed-model analysis of variance (ANOVA) with "population" as fixed and "block" as random factors following Shapiro-Wilk test (Shapiro and Wilk, 1965) for normality. Unless otherwise specified, block-wise averages for all measures used henceforth were analyzed by ANOVA with population and light regime or phase (whichever applicable) as fixed factors and block as random factor. Post-hoc multiple comparisons were performed using Tukey's HSD method (Tukey, 1949) at α of 0.05 in Statistica (StatSoft, Tulsa, OK, USA).

Recording in T-cycles
Since pre-adult rearing and virgin collection was in LD12:12, flies were initially exposed to T-cycles (T18: 9 h light and 9 h darkness; T20, T28 and T30, with light intensity during the light phase being 4 × 10 −4 Wm −2 ) for 7 days, loaded into activity tubes and recorded for seven cycles in the respective T-cycle followed by 5 days in constant darkness (DD) to assess phase control during entrainment. Activity-rest behavior in T-cycles was categorized as (1) free-run: if neither morning nor evening-activity components were phase-locked to the LD cycles, (2) weakly entrained: if one of the two components was phase-locked to the LD cycles while the other exhibited free-run and (3) entrained: if both activity components were phase-locked to the LD cycles (Supplementary Figure S1). For individuals that did not survive till recording in DD, phases of activity onsets and offsets were assessed, and the rhythm was considered entrained only if phase markers were either advanced or delayed with respect to ZT00. Flies with activity onsets coinciding with ZT00 were not considered as it might be masked to light. The percentage of free-running, weakly entrained and entrained flies were not found to qualify the assumptions of ANOVA, and hence were analyzed by the non-parametric Kruskal-Wallis test (Kruskal and Wallis, 1952) followed by post-hoc multiple comparisons using "multiple comparisons of mean ranks for all groups" test. This test is based on the Mann-Whitney test and is detailed in Siegel and Castellan (1988). All the above-mentioned tests were implemented in Statistica (Statsoft).

Adult emergence assay
Adult emergence assay was performed in LD12:12 in high (4 × 10 −1 Wm −2 ) and low (4 × 10 −4 Wm −2 ) light intensities in 10 replicate vials (~300 eggs per vial) for every block. Following egg collection, vials were transferred to respective light regimes for development, and upon initiation of emergence, numbers of flies emerging from every vial was recorded at 2 h intervals for five consecutive days. Only those vials that exhibited rhythmic emergence (with a minimum of 15 flies in a day) for at least three consecutive days were considered, and the number of flies at each time-point was averaged over multiple days independently for each block. Due to differences in modality and gate width of emergence (Kumar et al., 2007a) between populations, a single reference point could not be used to measure the rhythm amplitude. Instead, the net difference in number of flies emerging across the day between the two regimes (number of flies at low-high light intensities) was used as a measure.

Assessing rhythm amplitude
Due to lack of a standard procedure to assess the intrinsic amplitude of the circadian clock directly, we measured the amplitude of activity-rest rhythm as a proxy.

Amplitude in LD (entrained amplitude: A ent )
The activity profile of D. melanogaster in LD12:12 is bimodal comprising morning (M) and evening (E) activity peaks. To estimate amplitude of entrainment (A ent ), activity-rest data for at least 7 days recorded under LD12:12 (light intensity = 4 × 10 −4 Wm −2 ) was used to plot activity profile at 1 h bins (activity counts/h) using Microsoft Excel (Microsoft, USA). From this profile, total activity in the morning (ZT22-02) and evening (ZT10-14) was calculated and served as the amplitude of morning and evening peaks, respectively.

Amplitude in DD (intrinsic amplitude: A o )
Data for activity-rest rhythm recorded in DD for 10 days were binned into 1 h intervals and used to plot the activity profile as activity counts/h using Table Curve-2D (Systat Software Inc., San Jose, CA, USA). From the profiles thus obtained, the highest activity count/h was identified and used to calculate activity count/min. Since activity/min values were obtained from data binned at 1 h intervals, highest value indicates that the individual exhibited highest activity for the day in that 1 h duration representing the peak of activity, and thus served as a measure of intrinsic amplitude (A o ). The day-wise highest activity count/min values were first averaged across days for a given individual, and then across replicate individuals of a given block. We observed dramatic reduction in amplitude for the first 3 days following transfer from LD (in which the flies were initially reared) to DD, and therefore data from the first 3 days in DD were excluded from analysis. Normalized actograms (modulo τ) were used to estimate activity onsets and offsets using clocklab (Actimetrics, Wilmette, IL, USA) from which the total activity during the subjective day and night was calculated.

Re-entrainment to phase-shifted LD cycles
Activity-rest was recorded in LD12:12 for the first 5 days following which the flies were subjected to either a 9 h advance or 9 h delay on day 6 and recorded for next 10 days. The days taken to reentrain (transients) were estimated by a method similar to that in Sharma and Daan (2002). The phases of activity offsets from day 6 were regressed over the next 10 days to estimate the slope and R 2 of the regressed line. R 2 , also known as the coefficient of determination, is a measure of goodnessof-fit of the regressed line to the data and indicates percentage variation in the data that can be explained by the regression line. For instance, R 2 = 0.9 indicates that 90% of variability in the data can be explained by the regressed line, and therefore higher R 2 value would indicate better fit. After the first regression, data from day 6 were removed and those from day 7 were regressed over next 9 days, followed by 8 days and so on. This was continued till slope and R 2 values of the regressed line were closest to zero (complete lack of regression), indicating steady-state attainment, and therefore complete re-entrainment to new LD cycles. The number of days taken to reach steady state was considered as transients.
Since re-entrainment rate is known to be a function of the phase of the circadian rhythm (limit cycle) at the time of phase shift (Granada and Herzel, 2009), we estimated block-wise reentrainment rates as follows: (ψ before-shift -ψ aftershift )/transients, and the block means were then log transformed and analyzed by ANOVA.

Estimating photic dose response curve (DRC)
The activity of 3-day-old flies was recorded in LD12:12 (light intensity = 16 × 10 −2 Wm −2 ) for 6 days and transferred to DD on day 7. During the first cycle in DD, different sets of flies were subjected to 5-min light pulses of intensities 0, 4 × 10 −3 , 4 × 10 −2 , 4 × 10 −1 or 4 Wm −2 (0-1000 lux) at circadian time (CT) 14 and CT22 which represent the phases of maximum phase delay and phase advance, respectively (Dunlap et al., 2004). Recording was then continued in DD for the next 12 days. During exposure to light pulse, the activity monitors had to be displaced from the recording incubator to the light chamber. Therefore, to account for phase-shifts due to physical disturbance during light pulse, additional sets of "disturbance controls" were maintained, and each set was only physically disturbed at the respective phases similar to the experimental flies but were not exposed to light. Activity offset of each cycle was marked, a regression line drawn through the activity offsets for days 1-6, and extrapolated to predict the phase of offset on day 7 (day of light pulse). Similarly, another regression line was drawn through offsets from days 10 to 16 and extrapolated backward to identify the phase of offset on day 7. The time difference between the phases extrapolated from both the regression lines indicates the magnitude of phase-shift. All "disturbance controls" were also analyzed in the same way, and the phase-shift differences between the experimental and disturbance controls were considered as the phase-shift due to light pulse alone. This procedure was implemented on data from each fly and averaged across flies of a given block to obtain block average.
Since bright light saturates Drosophila photoreceptors (Juusola and Hardie, 2001), we reasoned that the light intensity required to elicit maximum phase-shift might not serve as a reliable measure of photosensitivity, and thus IPS50 was chosen; however, one may also choose IPS25 or IPS75 which represents the first and third quartiles of the phase-shift distribution. To estimate the light intensity eliciting 50% of maximum phase-shift (IPS50median of the underlying phase-shift distribution), the maximal phase-shift value for a given block was set to 1, and all other lower phaseshift values for that block were expressed as the proportion of the maximum-phase shift between 0 and 100%, with 100% being the maximum phaseshift. A nonlinear dose response equation of the form Y = 100/(1 + 10^((logEC50-X)×hill slope)) was fit to the phase-shift data using simple leastsquares method in Prism-5 (GraphPad, La Jolla, CA, USA) with phase-shift and log (light intensity) as ordinate and abscissa, respectively. From this plot, the abscissa corresponding to the ordinate value of 50% of maximal phase-shift was estimated and served as the IPS50. This procedure was implemented separately for each of the four blocks at two phases (CT14 and CT22). Details of hillslope calculations can be found in Prism-5.
Estimating area under photic phase response curve (PRC) Flies were recorded for 6 days in LD12:12 and on the first day in DD (day 7), they were subjected to a brief 5-min light pulse of intensity 16 × 10 −2 Wm −2 (70 lux) every 4 h starting at CT02, after which they were recorded in DD for 12 days. Phase-shift estimation for PRC analysis was same as that used for DRC.
Several nonlinear polynomials were fit to the phase-shift data using Table Curve (Systat, USA), and from the list of polynomials thus obtained, the polynomial with a combination of least number of coefficients and highest R 2 (explained earlier) was considered (Supplementary Table S1), and one such polynomial was obtained for every block. Since we were interested in estimating the area under the curve (AUC) and the experimental data were obtained at 4 h intervals which is a poor resolution to calculate AUC, we used the best-fit polynomial to obtain predicted phaseshifts at intermediate CT with a resolution of 0.04 h by interpolation. In other words, from the best-fit polynomial of the form y = f(x) we calculated the ordinate (y = phase-shift) for inrease in abscissa (x = CT) by 0.04 h, and the interpolated phase-shift values were then used to estimate AUC by simple integral under the curve method. To ensure that the polynomials accurately reflected the experimental data, we estimated the sum-ofsquares (SS) values between the phase-shift values obtained from the experiment at every 4 h intervals with those of the interpolated data from the polynomial. We found that the SS values were very low and did not differ significantly from 0 (t-test at α = 0.01; Supplementary Figure S2), suggesting that the polynomials effectively represented the experimental data. Block-wise AUC values thus obtained were subjected to ANOVA and posthoc comparisons.

Late populations exhibit higher arrhythmicity under dim constant light (LL)
To test if the early and the late populations evolved differential light sensitivity, we assayed activity-rest behavior under dim LL which is known to render the behavior arrhythmic (Konopka et al., 1989).
Under dim LL of 4 × 10 −4 Wm −2 , significantly higher percentage of flies from the late populations (58.48%) were arrhythmic as compared to the early (24.98%) and the control (21.83%) populations (F 2,6 = 22.34, p < 0.001; Figure 1b), while the latter two did not differ from each other. Percentage of individuals exhibiting complex rhythms did not differ significantly across early (42.22%), control (43.18%) and late (33.46%) populations (F 2,6 = 1.19, p > 0.05; Figure 1b), but a significantly lower percentage of the late populations (8.05%) exhibited freerunning rhythm as compared to the early (32.75%) and the control (34.98%; F 2,6 = 74.82,  Figure 1. Percentage of early, control and late flies exhibiting arrhythmic (row-1), complex (row-2) and rhythmic (row-3) activity-rest behavior under constant light (LL) of high (4 × 10 −3 Wm −2 ; left) and low (4 × 10 −4 Wm −2 ; right) light intensities. The actograms were analyzed by Chi-square periodogram and the behavior was categorized as follows. Free-running rhythm: presence of a single statistically significant period; complex rhythm: presence of two or more statistically significant periods; and arrhythmicity: absence of any statistically significant period. Error bars represent 95% CI, with asterisks indicating significant differences between populations (p < 0.05). A significantly higher proportion of late populations exhibit behavioral arrhythmicity under dim LL, suggesting that they might have evolved higher light sensitivity.
p < 0.01; Figure 1b), while the latter two did not differ from each other. Interestingly, proportion of flies from the late populations exhibiting arrhythmic behavior did not differ across the two light intensities but that for the early and the control increased significantly at high light intensity. This suggests that besides enhanced light sensitivity the late populations might have also evolved an oscillator network marked by weak coupling and will be discussed in detail later.

Late populations exhibit high amplitude of entrainment (A ent )
To further confirm if the late populations indeed exhibit enhanced light sensitivity, we also assessed activity-rest rhythm in LD12:12, the rationale being that higher light sensitivity would promote robust entrainment with higher amplitude and power of rhythm.
ANOVA on activity/h revealed a statistically significant effect of "population × phase" interaction (F 46,138 = 11.74, p < 0.01), and activity levels of the late populations were significantly higher at most phases ( Figure 2a). As an additional measure of A ent , we calculated total activity around the M-and E-peaks as these are known to be clockcontrolled (Stoleru et al., 2004). M-activity was significantly higher for the late populations as compared to the other two populations (F 2,6 = 9.65, p < 0.05) while the latter two did not differ (Figure 2a and b). Activity around the E-peak was also significantly higher for the late populations (F 2,6 = 5.89, p < 0.05) as compared to the early populations (Figure 2a and b). We also calculated the ratio of daytime/nighttime activity and found it to be significantly higher for the late (2.15) populations as compared to that for the early (1.72) and the control (1.93) populations (F 2,6 = 20, p < 0.01; Figure 2c). Further, power of the rhythm as estimated by amplitude of the Chi- square periodogram was also significantly higher for the late (211.64) populations followed by the control (202.76) and the early (183.34) populations (F 2,6 = 5.89, p < 0.05; Figure 2d). These indicators highlight a relatively robust entrainment in the late populations even under low intensity LD cycles, further suggesting that these populations might have evolved higher light sensitivity.

Emergence rhythm in the late populations is highly light sensitive
We further wished to test if the observed effects are restricted to activity-rest rhythm alone or if it is manifested in emergence rhythm as well. Observing similar effects in two independent rhythms would strengthen the idea that such differences might stem from a common central oscillator governing both the rhythms, and therefore we assayed adult emergence of all populations under high and low light intensity LD12:12. We used LD cycles instead of LL because emergence rhythm is a populational rhythm therefore it is not meaningful to analyze incidence of arrhythmic and complex phenotypes unlike activity rhythms which are studied at the individual level.
In addition, anticipation index (AI) for the late populations increased by 0.33 units with decreasing light intensity, significantly higher than the early (0.0002) populations which and low (4 × 10 −4 Wm −2 ) light intensities during the day, (b) Difference in entrainment amplitude (A ent ) calculated as number of flies at low-high light intensities and (c) AI calculated as (AI at low light intensity -AI at high light intensity) of emergence under high and low light intensity LD12:12 for the three populations. AI was calculated as the ratio "number of flies emerged 2 h prior to ZT00/ number of flies emerged 4 h before lights-on." Error bars represent 95% CI, with asterisks indicating significant differences between populations (p < 0.05). The late populations exhibit significantly higher responses to changes in light intensity as can be observed in A ent and AI differences. exhibited almost no response, while the control populations responded intermediately (0.10) (Figure 3c). Similar trends in the differences between population for activity and emergence rhythms indicate that the observed phenotypes stem from components pertaining to a common central clock. Therefore, we restricted further studies to only the activity-rest rhythm.

Late populations exhibit high intrinsic amplitude (A o )
We also reasoned that higher amplitude of entrainment in the late populations might not necessarily be due to higher light sensitivity but also due to higher intrinsic clock amplitude (A o ). Additionally, previous studies (see Introduction) have also reported that the ψ ent is correlated with A o . To test for these possibilities, we estimated the amplitude of free-running rhythm (A o ) in all the three populations.
A o was found to be significantly higher for the late populations (3.02 activity counts/min) as compared to the early (2.42 activity counts/min) and the control (2.41 activity counts/min) populations (F 2,6 = 24.81, p < 0.01; Figure 4a). Also, the ratio of activity in the subjective day to that of the subjective night was significantly higher in the late (5.91) populations as opposed to that for the early (4.05) and the control (3.77) populations (F 2,6 = 13.06, p < 0.01; Figure 4b), while the early and the control populations did not differ between each other for both the measures (Figure 4a and b). Taken together, these indicators suggest that the late populations have evolved high amplitude circadian oscillations.
We further calculated difference in amplitudes of entrained and free-running rhythms (amplitude expansion = A ent − A o ). Amplitude expansion can be influenced by both light sensitivity and nature of inter-oscillator coupling and therefore might reveal interesting properties of clocks in these populations.
To facilitate comparison of amplitudes of entrained and free-running rhythms, we computed the activity/min values in LD similar to that in DD. As expected, activity/min was found to be highest at the two phases corresponding to the M-and E-peaks. Concordant with earlier results, ANOVA on average activity/min in LD revealed a statistically significant effect of population (F 1,3 = 31.9, p < 0.001) but not of "peak" or "population × peak" interaction, indicating that the M-and E-peaks did not differ for a given population, and therefore, we averaged the activity/min values across M-and E-peaks (Supplementary Figure S3). The averaged values were then considered as measures of amplitude under LD to calculate amplitude expansion.
While there was a trend of increase in amplitude expansion from early to late, ANOVA on "amplitude expansion" values did not reveal any significant effect of population (F 2,6 = 2.88, p = 0.13), suggesting that the three populations exhibit similar amplitude expansion (early = 2.44 activity/ min; control = 2.67 activity/min; late = 2.77 activity/min; Figure 4c). Therefore, it appears that high amplitude of entrainment in the late populations is probably driven by high A o and not necessarily higher light sensitivity.

Late populations exhibit slower rate of re-entrainment
A o is inversely related to the phase-resetting ability of the clock (see Introduction), and therefore we tested this proposition by estimating the re-entrainment rates of all populations to 9 h phase advance or delay in LD cycles with the rationale that higher A o in the late populations would reduce their re-entrainment rate.
ANOVA on re-entrainment rates revealed a statistically significant effect of population for both advance (F 2,6 = 10.03, p < 0.05) and delay (F 2,6 = 13.03, p < 0.01) phase-shifts. The reentrainment rate of the late populations was significantly lower for both advance (2.5 h/day) and delay phase-shifts (2.5 h/day) as compared to those for the early (advance = 4.50 h/day; delay = 6.71 h/day) and the control (advance = 2.80 h/ day; delay = 6.30 h/day) populations (Figure 5; Figure S4).  Figure 5. Representative actograms depicting re-entrainment of early, control and late populations to 9 h phase-advanced (column-1) and delayed (column-2) 12:12 h light-dark (LD12:12) cycles. The shaded regions represent night phase. Bottom panel depicts the re-entrainment rates to 9-h phase-advanced (left) and delayed (right) LD12:12 cycles. Re-entrainment rates were calculated as 9 h (magnitude of phase-shift)/number of transients. Error bars represent 95% CI, with asterisks indicating significant differences between populations (p < 0.05). The figures indicate that the late populations take significantly longer to re-entrain to both advance and delay phase-shifts.
Had the late populations evolved enhanced light sensitivity as suggested by some of the earlier results, they would be expected to exhibit high re-entrainment rate as well, which is contrary to what we observed. Therefore, it appears that reduced rates of re-entrainment in the late populations might be driven by their reduced ability to undergo large phase-shifts due to high A o value.

Late populations exhibit wider entrainment range
Additionally, we also assessed the entrainment range to test if high A o in the late populations would curb its phase-resetting ability, and consequently, restrict their entrainability to non-24 h T-cycles.
In T18, a significantly higher percentage of the early (59.85%) populations failed to entrain compared to the late (20.15%) (Kruskal-Wallis H = 8, df = 2, p < 0.05), whereas the control (30.08%) populations did not differ from either ( Figure 6row-1). The percentage of individuals exhibiting entrainment was almost negligible while significantly higher percentage (77.57%) of the late populations weakly entrained as opposed to 37.86% in the early and 69.82% in the control populations ( Figure 6-row-1).
In T20, the percentage of individuals which failed to entrain (free-running) was low (early = 8.25%, control = 29.33%, late = 10.01%; Figure 6row-2) and the populations did not differ significantly (Kruskal-Wallis H = 2.59, df = 2, p > 0.05), and the same was observed for weakly entrained individuals as well ( Figure 6-row-2).  Figure 6. Percentage of early, control and late populations exhibiting free-run (column-1), weakly entrained (column-2) and entrained (column-3) activity-rest rhythm in T18 (row-1), T20 (row-2), T28 (row-3) and T30 (row-4) regimes with light intensity during the day being 4 × 10 −4 Wm −2 . All actograms were visually analyzed and the entrained behavior was categorized based on the following criteria. Free-run: if neither morning nor evening activity components were phase-locked to the LD cycles; weakly entrained: if one of the two components was phase-locked to the LD cycles while the other exhibited free-run; entrained: if both activity components were phase-locked to the LD cycles (Supplementary Figure S1). Error bars represent SEM across blocks, with asterisks indicating significant differences between populations (p < 0.05). The late populations appeared to exhibit a wider entrainment range as opposed to the other two populations in which lower proportion of individuals entrained to extreme T-cycles.
Surprisingly, contrary to our expectation of reduced entrainment range in late populations, while both control and late populations entrained to all T-cycles, the late populations consistently showed higher incidence of entrainment probably driven by high zeitgeber sensitivity.

Dose response curves (DRCs) do not indicate difference in light-induced phase-shifts
In light of the seemingly contradicting results from re-entrainment rate and entrainment range assays, we further decided to confirm if the late populations actually differ in phase-resetting ability, and therefore generated DRCs by measuring phase shifts elicited by light pulses of increasing intensities at CT14 and CT22.
Therefore, even though high A o can partly account for reduced re-entrainment rate, lack of difference in the DRCs further complicate the interpretability of lower re-entrainment rate and wider entrainment range in the late populations. . Procedure for estimating IPS50 is detailed in materials and methods. Briefly, the phase-shift data were used to fit a nonlinear DRC of the form Y = 100/(1 + 10^((LogEC50-X)×hill slope)) from which the abscissa value corresponding to 50% of the maximal phase-shift was estimated and served as the IPS50 value. Error bars represent 95% CI. DRC and IPS50 analysis indicates that the three populations do not differ in magnitude of instantaneous phase-shifts even across multiple light intensities.
Therefore, we speculated that the observed differences probably arise from mechanisms apart from mere light-induced phase-resetting.

Area under phase response curves (PRCs) suggests continuous effects of light
Based on the results from previous section, we speculated that wide entrainment range might be driven by continuous or tonic effect of light, and not by phasic effects as assessed by the DRC. To test this, we estimated AUC for photic-PRCs of all populations which is a measure of overall phase shift accumulated over a longer duration. ANOVA on phase-shift values reported a statistically significant effect of "population × phase" interaction (F 10,30 = 2.58, p < 0.05). The late populations exhibited larger phase-shifts at CT18 as compared to the early and the control populations ( Figure 8a). ANOVA on AUC values reported a statistically significant effect of "population" for delay phase-shifts (F 2,6 = 6.22, p < 0.05) and was marginal for advance phase-shifts (F 2,6 = 4.48, p = 0.06). Post-hoc comparisons revealed that area under the advance zone was considerably smaller for the late (19.47 h 2 ) populations as compared to the early (33.41 h 2 ) and the control (169.98 h 2 ) but did not differ statistically (Figure 8b), whereas area under the delay zone was significantly greater for the late (943.62 h 2 ) populations as compared to the early (735.76 h 2 ) and the control populations (689.66 h 2 ; Figure 8c).
These results partly suggest that the populations might have evolved differences in their ability to integrate light over prolonged durations.

Discussion
Motivated by the results from our previous studies, in conjunction with those from others as discussed in the introduction, we employed populations of D. melanogaster selected for early and late emergence to explore the association of circadian clock (network) properties with the ψ ent .
We first assessed if the early and the late populations might have evolved light-sensitivity differences. Significantly higher proportion of the late populations were observed to exhibit behavioral arrhythmicity under dim LL as compared to the other two populations which exhibited free-run (Figure 1b), suggesting that the late populations might have evolved higher light sensitivity. This was further substantiated by robust highamplitude entrainment of both emergence and activity-rest rhythms in the late populations under low intensity LD cycles (Figures 2 and 3).  Figure 8. (a) Phase response curves depicting magnitude of phase-shift elicited by 5 min light pulse of intensity 16 × 10 −2 Wm −2 across different circadian times of the day for early, control and late populations, (b) and (c) represent the total area under the curve (AUC) for advance and delay zones, respectively, in the three populations. To estimate AUC, the polynomial that best fits to the experimental PRC data was used to interpolate phase-shift values at intermediate circadian times every 0.04 h. The interpolated phase-shift values were then used to estimate the area under the advance and delay zones by integral under the curve method. Error bars represent SEM across blocks, with asterisks indicating significant differences between populations (p < 0.05). It can be observed that the AUC under the delay zone was significantly higher for the late populations, suggesting a possible role of tonic effects of light on entrainment.
Alternatively, we reasoned that the high amplitude of entrainment in the late populations ( Figure 2) may not necessarily be due to enhanced light sensitivity but can also be driven by a high-amplitude circadian oscillation, or in other words, highamplitude circadian clocks, which turned out to be the case when we observed that the late populations exhibited high-amplitude activity-rest rhythm in DD. As discussed earlier, sensitivity of circadian clocks to zeitgeber is inversely proportional to its amplitude. Therefore, based on the observed higher arrhythmicity in the late populations under dim LL, one would expect circadian clock amplitude in these populations to be lower than that of the other populations which is clearly not the case, thus implying that higher arrhythmicity in the late populations might not entirely be due to light-sensitivity differences but may involve other mechanisms as will be discussed later. Nevertheless, higher A o in the late populations clarifies their reduced re-entrainment rates to both advance and delay jet-lag ( Figure 5). In other words, higher A o in the late populations would lead to lower A z /A o , thus resulting in small-magnitude phase-shifts, and consequently reduced re-entrainment rate which is in accordance with the observed results.
Having observed divergent clock amplitudes and re-entrainment rates to simulated jet-lag, we further assessed if such properties were intertwined with phase-resetting ability of the underlying clocks to gain further insights into how entrainment differences might drive early and late emergence chronotypes. Intriguingly, contrary to the observations by Pittendrigh (1967Pittendrigh ( , 1981a and Takamura (1987, 1989), despite A o differences, the early and the late populations did not differ in their phase-resetting ability even across light intensities spanning orders of 10 4 (Figures 7 and 8); whereas the late populations exhibited wider entrainment range ( Figure 6) which appears counterintuitive under the realms of the discrete entrainment model (Pittendrigh, 1960). However, the underlying assumption of this model is that light instantaneously shifts the clock phase, which even though successfully tested has been reconsidered multiple times motivating the proposal of a modified discrete entrainment model involving continuous (tonic) effect of light . Therefore, even though the late populations do not differ in instantaneous phase-shifts, the wider entrainment range in these populations might additionally be facilitated by tonic effects of light over longer durations. Previously, Vaze et al. (2012) reported that when entrained to skeleton photoperiod comprising 15 min light pulses in the morning and evening, none of the populations exhibited their respective ψ ent as observed in LD12:12, but when entrained to asymmetric skeleton photoperiods, the late populations required longer duration of light (6 h) in the morning while the early populations required longer light duration in the evening to exhibit their respective ψ ent , indicating that these populations integrated light information over prolonged durations. This proposition is further supported by the observation that the accumulated phase-shifts over longer duration (as estimated by the AUC) in the delay zone of the PRC of late populations is significantly greater as compared to the other populations, thereby suggesting the role of both phasic and tonic effects of light in the late populations.
Previously, we had reported that the early and the late populations might have evolved dominant M-and E-oscillators or neurons (Kumar et al., 2007b), whose coupling and consequently dominance is known to change with photoperiod, thus driving seasonal adaptations (Grima et al., 2004;Stoleru et al., 2004Stoleru et al., , 2007. As a further extension of Kumar et al. (2007b), most if not all of the results of our study can be explained in the framework of dominant E-neurons in the late populations. E-neurons primarily contribute to delay phase-shifts (Stoleru et al., 2005), and thus can account for larger AUC in delay zone of the late populations. Alternatively, enhanced AUC in the delay zone might also be facilitated by higher CRY expression in the E-neurons. This proposition also explains several of our observations in the late populations. CRY in E-neurons considerably reduces the ability of the M-neurons to dominate over the E-neurons , thus rendering the latter relatively independent. Also, E-neurons alone have been implicated to maintain rhythmicity in LL (Stoleru et al., 2007). Therefore, higher CRY levels in the E-neurons of the late populations would render them more sensitive to to explore how these properties influence other aspects of entrainment such as rhythm stability and adaptation to seasonal changes in nature.