Artificial selection for diurnal preference
Flies showed a rapid and robust response to selection for phase preference. After 10 selection cycles, we obtained highly diurnal (D) and nocturnal (N) strains. The two control strains (C) showed intermediate (crepuscular) behaviour (Fig 1). To quantify diurnal preference, we defined the ND ratio, quantitatively comparing activity during a 12 h dark period and during a 12 h light period. As early as after one cycle of selection, the ND ratios of N and D flies, and as compared to the original (control) population, were significantly different (Fig 1A, B). After 10 generation of selection, the N and D populations were highly divergent (Fig 1B, Table S1).
The estimated heritability h2 was higher for diurnality (37.1%) than for nocturnality, (8.4%) reflecting the asymmetric response of the two populations (Table S1). We estimated heritability by parents-offspring regression (Fig 1C, D). The narrow-sense heritability was lower but significant (h2 = 14% p<0.05; Fig 1C). The heritability value was slightly higher when ND ratios of mothers and daughters were regressed (h2 = 16% p<0.05; Fig 1D) but was minute and insignificant in the case of father-son regression ( Fig S1; h2 = 2.5% NS).
Effects of Nocturnal/Diurnal phenotypes on fitness
A possible mechanism driving the observed asymmetric response to selection is unequal allele frequencies, whereby a slower response to selection is associated with increased fitness . We, therefore, tested whether our selection protocol asymmetrically affected the fitness of the N and D populations. After ~15 overlapping generations from the end of the selection, we tested viability, fitness and egg-to-adult developmental time of the selection and control populations. While the survivorship of males from the three populations was similar (χ2= 1.6, df=2, p=0.46; Fig 2A), we found significant differences in females. N females lived significantly longer than D females, while C females showed intermediate values (χ2=7.6, df=2, p<0.05; Fig 2A). The progeny number of N females was larger than that of D females, with C females showing intermediate values (♂F1,18 =5.12, p<0.05; ♀F1,18 =5.09, p<0.05; Fig 2B). Developmental time (egg-to-adult), another determinant of fitness, did not differ significantly between the nocturnal/diurnal populations (♂F2,27 =0.43, p=0.65, NS; ♀ F2,27 =0.27, p=0.76, NS; Fig 2C,D).
Effects on circadian behaviour
Since the circadian system is a conceivable target for genetic adaptations that underlie diurnal preference, we tested whether the circadian clock of the N and D strains were affected by the Nocturnal/Diurnal artificial selection. Accordingly, we recorded the locomotor activity of the selection lines following three generations after completion of the selection protocol, and measured various parameters of circadian rhythmicity.
The phase of activity peak in the morning (MP) and in the evening (EP) differed between the populations (Fig S2A). As expected, the MP of the N population was significantly advanced, as compared to that seen in both the C and D populations. The EP of the N population was significantly delayed, as compared to that noted in the two other populations. Concomitantly, the sleep pattern was also altered (S2B), with N flies sleeping much more during the day than did the other populations. While the D flies slept significantly more than C and N flies during the night, there was no difference between N and C flies.
In contrast to the striking differences seen between the selection lines in LD conditions, such differences were reduced in DD conditions (Fig 3, Fig S2C). The period of the free-run of activity (FRP) was longer in the C flies than in D flies, while the difference between the D and N groups was only marginally significant (Fig 3A). No significant difference in FRP was found between N and C flies. The phases (φ) of the three populations did not differ significantly (Fig 3B). We also tested the response of the flies to an early night (ZT15) light stimulus and found no significant differences between the delay responses (Fig 3C).
Circadian differences between Nocturnal/Diurnal isogenic strains
To facilitate genetic dissection of the nocturnal/diurnal preference, we generated nocturnal, diurnal and control isogenic strains (D*, N* and C*; one of each) from the selected populations. The strains were generated using a crossing scheme involving strains carrying balancer chromosomes. The ND ratios of the isogenic lines resembled those of the progenitor selection lines ( Fig S3A). The isogenic strains also differed in terms of their sleep pattern ( Fig S3B).
The circadian behaviour of the isogenic lines differed, with the N* line having a longer FRP than both the D* and C* lines (Fig S4A). The locomotory acrophase of the N* line was delayed by about 2 h, as compared to the D* line, and by 1.38 h, as compared to the C* line (F2,342 =6.01, p<0.01; Fig S4B). In contrast, circadian photosensitivity seemed to be similar among the lines, as their phase responses to a light pulse at ZT15 did not differ (F2,359:1.93, p =0.15, NS; Fig S4D). Since eclosion is regulated by the circadian clock [20, 21], we also compared the eclosion phase of the isogenic strains. Under LD, the eclosion phase of D* flies was delayed by ~2 h (becoming more diurnal), as compared to both N* and C* flies, whereas no difference between N* and C* flies was detected ( Fig S4C).
Diurnal preference is partly driven by masking
We reasoned that light masking (i.e., the clock-independent inhibitory or stimulatory effect of light on behaviour) could be instrumental in driving diurnal preference. We thus monitored fly behaviour in DD to assess the impact of light masking. We noticed that when N* flies were released in DD, their nocturnal activity was much reduced, whereas their activity during the subjective day increased (Fig 4). Indeed, the behaviour of N* and D* flies in DD became quite similar (Fig 4). Congruently, when we analysed the ND ratios of these flies in LD and DD, we found that that both N* and C* flies became significantly more “diurnal” when released into constant conditions (N*, p<0.0001; C*, p<0.001). In contrast, the ratios of D* flies did not significantly change in DD (p =0.94, NS). This result suggests that nocturnal behaviour is at least partially driven by a light-dependent repression of activity (i.e., a light masking effect).
Correlates of the molecular clock
To investigate whether differences in diurnal preference correlated with a similar shift in the molecular clock, we measured the intensity of nuclear PERIOD (PER) in key clock neurons (Fig 5). The peak of PER signals in ventral neurons (LNv: 5th-sLNv, sLNv, lLNv) was delayed in N* flies, as compared to the timing of such signals in D* fly 5th-sLNv, sLNv and lLNv neurons. In N* and D* flies, the phases of such peaks in dorsal neurons (DN, including the clusters LNd, DN1and DN2) were similar (Fig 5).
We also measured the expression of the Pigment Dispersing Factor (PDF) in LNv projections (Fig S5-S6). The signal measured in N* flies was lower than that measured in D* flies during the first part of the day (in particular at ZT3 and ZT7) yet increased during the day-night transition at ZT11 and ZT13. There were no differences seen during the rest of the night.
Global transcriptional differences between Nocturnal/Diurnal strains
To gain insight into the genetics of diurnal preference, we profiled gene expression in fly heads of individuals from the D*, C* and N* isogenic lines by RNAseq. We tested for differentially expressed genes (DEG) in all pairwise contrasts among the three strains at two time points. We found 34 DEGs at both ZT0 and ZT12 (Table S3). An additional 19 DEG were unique to ZT0 and 87 DEG were unique to ZT12 (Table S3). Functional annotation analysis (DAVID, https://david.ncifcrf.gov/ ) revealed similarly enriched categories at ZT0 and ZT12 (Fig S7). The predicted products of the DEGs were largely assigned to extracellular regions and presented secretory pathway signal peptides. DEG products identified only at ZT12 were related to the immune response, amidation and kinase activity. Given the intermediate phenotype exhibited by C* flies, we reanalysed the data, searching for DEGs where C* flies showed intermediate expression (D* > C* >N* or N* > C* > D*; Table S4). The list of DEGs consisted of 22 genes at ZT0 and 62 at ZT12. Amongst the different functions represented by these new lists were photoreception, circadian rhythm, sleep, Oxidation-reduction and mating behaviour were over-represented in both D* and N* flies. For example, Rhodopsin 3 (Rh3) was up-regulated in D* flies, while Rh2 and Photoreceptor dehydrogenase (Pdh) were down-regulated in N* flies. pastrel (pst), a gene involved in learning and memory, was up-regulated in D* flies, while genes involved in the immune response were up-regulated in N* flies. The only core clock gene that showed differential expression was Par Domain Protein 1 (Pdp1), which was up-regulated in D* flies. The clock output genes takeout (to) and pdf were up-regulated in N* flies. Overall, the results suggested that genes that are transcriptionally associated with diurnal preference are mostly found upstream (light input pathways), and downstream of genes comprising the circadian clock.
We investigated the contribution of various genes to nocturnal/diurnal behaviour using a modified version of the quantitative complementation test (QCT) . We tested the core circadian clock genes per and Clk, the circadian photoreceptor cry and the output gene Pdf and Pdfr, encoding its receptor (Fig 6, Table S5) . We also tested the ion channel-encoding narrow abdomen (na) gene, given its role in the circadian response to light and dark-light transition . The QCT revealed significant allele differences in per, Pdf, Pdfr, cry and na (Fig 6, Table S5), indicative of genetic variability in these genes contributing to the nocturnal/diurnal behaviour of the isogenic lines.
Since switching from nocturnal to diurnal behaviour in mice has been shown to be associated with metabolic regulation , we also tested Insulin-like peptide 6 (Ilp6), and chico, both of which are involved in the Drosophila insulin pathway. A significant effect was found in Ilp6 but not in chico (Table S5). Other genes that failed to complement were paralytic (para), encoding a sodium channel, and coracle (cora), involved in embryonic morphogenesis [27, 28].
We also tested genes that could affect the light input pathway, such as Arrestin2 (Arr2) and misshapen (msn) . There was a significant evidence for msn failing to complement but not for Arr2 (Table S5). Various biological processes are associated with msn, including glucose metabolism, as suggested by a recent study .