Asymmetric regulation of diapause and development rate
The fact that a decision to enter diapause can be reversed by a changed photoperiod signal later in life than can a decision not to enter diapause is probably a result of strong selection against entering diapause unprepared (Friberg et al., 2011). Diapause tends to be a long-lasting and demanding state during which an insect is subject to adverse conditions such as extreme cold and drought, necessitating protective adaptations (Denlinger, 1991; Danks, 2000). Furthermore, a diapausing insect (especially a pupa) often has little or no access to food, and must therefore rely on resources gathered before diapause (Hahn & Denlinger, 2007, 2011). The physiological preparations required for successful diapause may take some time to establish (Koštál, 2006), and so it appears adaptive to resist a sudden switch to diapause development even given environmental cues signaling the end of the season, whereas cues signaling that the season is not close to ending (in this case, long days) can be more safely acted on by cancelling diapause and opting for an additional reproductive cycle within the same year.
Many insect species quantitatively regulate development rate in response to photoperiod (reviewed by Beck, 1980; examples in Shindo & Masaki, 1995; Gotthard, 1998; Gotthard et al., 1999; Strobbe & Stoks, 2004; Shama & Robinson, 2006), but the phenomenon has received less attention than the photoperiodic induction of diapause. A clear connection exists with time constraint: photoperiods that signal seasonal progression (shorter days in the summer and fall; longer days in the spring) tend to speed up development, preventing the insect’s life cycle from drifting out of sync with the changing environment or with conspecifics (Shindo & Masaki, 1995; Gotthard et al., 2000). In insects with the potential for more than one generation per year, these time horizons are more complicated, as a decision not to diapause imposes the additional time stress of fitting an additional generation into the remainder of the season (Roff, 1980; Kivelä et al., 2013). Accordingly, in P. aegeria , the long daylengths associated with nondiapause development are also associated with highly accelerated development and growth, and just as a change from short to long days was able to avert diapause, it also caused development to accelerate despite slow development earlier in life (Fig. 3).
While lengthening days always had the effect of speeding up development, the effect of shortening days was more complex. Shortening days in the fourth instar actually produced a slight increase in development rate (Fig. 3c), resulting in slightly earlier average pupation dates (Fig. 4). It is difficult to say whether this small boost is adaptive, or merely a physiological quirk. Responding to a drop in daylength at the end of the larval period by speeding up development may well improve fitness: larvae in this treatment are too far gone to switch to diapause development (Fig. 2), so if shortening days signal the approaching end of the season, adulthood should be attained fast instead, hence “making the best of a bad situation”. A similar boost was visible when days shortened in the third instar, suggesting that it is a general short-term effect (Fig. 3b), although the effect was later strongly reversed in those individuals that, as the short days continued, switched to diapause-track development and slowed down their development accordingly.
It should be noted that the photoperiods used here (21 versus 15 hours light) are extremes, that serve as unambiguous diapause/nondiapause signals for all three studied populations. Laboratory exposure ofP. aegeria larvae to constant, intermediate daylengths often produces individuals with a greatly extended larval period (up to three months) that nonetheless do not enter diapause at pupation, indicating that the development rate polyphenism and the diapause polyphenism are in fact at least semi-distinct on a physiological level, and have subtly different photoperiod thresholds (Nylin et al., 1989; Lindestad et al., 2019). But even if the two plastic switches operate semi-independently, the gradual change in daylength that will naturally occur across the season should differentially canalize the responses into distinct phenotypes. Larvae hatched late in the season will experience short daylengths and develop slowly, hence exposing them to even shorter daylengths later in life, and successfully inducing pupal diapause. Larvae hatched around the summer solstice will undergo sustained exposure to long days, leading to fast development to adulthood without diapause. For larvae hatched significantly before the solstice (which will be more common at lower latitudes), the effect will presumably be a synchronization of the non-diapausing cohort: early-hatched larvae may develop slowly at first, but gradually lengthening days will speed up development (and avert pupal diapause) to match larvae born later. These interactions between photoperiodic control of diapause and photoperiodic control of pre-diapause development rate, which have been reproduced in simulations (Lindestad et al., 2019), exemplify how developmental plasticity at different stages in an organism’s life can self-reinforce or modulate other plastic traits in a cascade fashion (West-Eberhart, 2003).
Although the overall pattern of development rates was more or less the same for all populations, there were differences in the degree to which development slowed down under short days (Fig. 3, Fig. 4). This is because the two daylengths used in the present experiment constitute “slices” through a much more complex photoperiodic reaction norm. Northern European P. aegeria larvae develop at their maximal rate at daylengths above a population-specific threshold; below this threshold, development slows down dramatically, then speeds up again in a more or less linear fashion as days continue to shorten (Nylin et al., 1989, 1995; Lindestad et al., 2019). While 21 hours of light is sufficiently far above the threshold to have a very similar effect on all populations, the 15-hour treatment intersected a somewhat different point on the linear phase of the reaction norm for each population (Fig. S3).