Development rate
Larval development rates were dependent on photoperiod regime, with larvae of all populations developing faster under constant long days than under constant short days (Fig. 3). Although subtle at first, the effect of daylength was detectable early in life: the molt to the third instar occurred on average two days later under short days than under long days (planned contrast: t186=7.79, p<0.001). This difference was magnified later during development, with the fourth instar typically taking nearly twice as long to complete for short-day control larvae than for long-day control larvae (planned contrast: t281=23.6; p<0.001). Development rate results were complex, largely owing to sex differences: In P. aegeria , larvae not headed for diapause are sexually dimorphic for development rate (Nylin et al., 1993), but this effect is only found in bivoltine populations (Aalberg Haugen & Gotthard, 2015). Hence, a three-way interaction was seen in the fourth instar (analysis of variance; sex × treatment × population F12=1.95; p=0.03). However, this contributed comparatively little to the overall models (Table S1); by far the largest amount of variation in development rate, especially in the fourth and final instar, was explained by the overall effect of photoperiod treatment.
Much like with diapause induction, when larvae experienced a change in daylength during development, the effects on development rate were asymmetric depending on the direction of change, and also depended greatly on the timing of the change. In the third instar (Fig. 3b), larvae that had recently been moved from short days to long days showed slightly increased development rates relative to larvae that remained in short days (planned contrast; t261=3.31, p=0.0011). A decrease in daylength, on the other hand, had little immediate effect on average; if anything, development was slightly faster than in the remaining long-day larvae (planned contrast; t261=2.60; p=0.0099). However, three individuals in this group instead showed drastically lowered development rates (two from Stockholm, one from Öland; all three later entered diapause). These three extreme outliers were excluded from the linear model for third-instar development rate, as they likely represent a biologically distinct response, but are shown as separate points in Figure 3b (complete raw data for this trait is shown in Fig. S1).
A similar but stronger short-term pattern was seen when the photoperiod change instead occurred in the fourth instar. Again, lengthening days in the fourth instar sped up development (Fig. 3c), leading to fourth-instar development rates intermediate between those for the long- and short-day controls groups (Tukey contrast; t281=10.81; p<0.001). A decrease in daylength did not result in a lower development rate, unlike what may be expected; on the contrary, a slight increase was seen relative to the long-day control group (Tukey contrast; t281=3.23; p=0.023). Finally, the most dramatic effects on fourth-instar development rate were observed in those larvae that had experienced a photoperiod switch in the third instar. Larvae that had experienced an increase in daylength had now fully adjusted their phenotype, and developed at a rate indistinguishable from that of the long-day control larvae. Meanwhile, larvae that had experienced a decrease in daylength showed a strongly bimodal response, which correlated closely with whether diapause occurred after pupation: those not headed for diapause developed fast, while those headed for diapause reversed their previous response and instead developed very slowly, mirroring their short-day control-group counterparts. In contrast, only a weak correlation between development rate and the eventual diapause decision could be observed for this treatment in the third instar (Fig. S1).
The overall outcomes of the regulation of development rate across the whole larval period are shown in Figure 4. Short-day control larvae pupated considerably later than long-day control larvae, with an average difference of 46% for Öland and 29% for Stockholm, respectively (no data for Skåne, as hatching dates were not recorded). Larvae switched from short to long days in the fourth instar ended up with intermediate pupation dates, i.e. partially compensating for slow early development, while larvae switched from long to short days in the fourth instar pupated at a very similar age to their long-day control counterparts (Fig. 4a-b). When the move from short to long days occurred as early as the third instar, larvae were much better able to adjust their phenotype (Fig. 4c-d). This was especially true for Stockholm, as this population showed a relatively small baseline difference between long-day and short-day development rates. Finally, larvae switched from long to short days in the third instar had very different outcomes depending on diapause decision: individuals headed for diapause pupated at times similar to the short-day control group, while those not headed for diapause pupated at times similar to the long-day control group.