Theory maintains that when future environment is predictable, parents should adjust the phenotype of their offspring to match the anticipated environment. The plausibility of positive anticipatory parental effects is hotly debated and the experimental evidence for the evolution of such effects is currently lacking. We experimentally investigated the evolution of anticipatory maternal effects in a range of environments that differ drastically in how predictable they are. Populations of the nematode Caenorhabditis remanei, adapted to 20°C, were exposed to a novel temperature (25°C) for 30 generations with either positive or zero correlation between parent and offspring environment. We found that populations evolving in novel environments that were predictable across generations evolved a positive anticipatory maternal effect, since they required maternal exposure to 25°C to achieve maximum reproduction in that temperature. In contrast, populations evolving under zero environmental correlation had lost this anticipatory maternal effect. Similar but weaker patterns were found if instead rate-sensitive population growth was used as a fitness measure. These findings demonstrate that anticipatory parental effects evolve in response to environmental change so that ill-fitting parental effects can be rapidly lost. Evolution of positive anticipatory parental effects can aid population viability in rapidly changing but predictable environments. 

Experimental evolution
As founder population, we used the wild-type SP8 strain of C. remanei, obtained from N. Timmermeyer at the Department of Biology, University of Tübingen, Germany. This strain was created by crossing three wild-type isolates of C. remanei (SB146, MY31, and PB206), harbour substantial standing genetic variation for life-history traits (Chen and Maklakov 2012; Lind et al. 2017), and has been lab-adapted to 20°C for 15 generations prior to setup of experimental evolution.

Experimental evolution was conducted for 30 generations in three climate cabinets; one set to 20°C, one to 25°C and one with a slowly increasing temperature (see below). Five experimental evolution regimes were used, and are summarized in Figure 1. Control 20°C was experiencing 20°C for 30 generations, and Warm 25°C was experiencing 25°C for 30 generations. Increased warming started in 20°C, the cabinet temperature was then raised by 0.1°C every 2.13 day (rounded to whole days) to reach 25°C the last day of experimental evolution. Slow temperature cycles spend their first 10 generations in 20°C, were then moved to the 25°C cabinet for 10 generations, to finish the last 10 generations in the 20°C cabinet. Finally, the Fast temperature cycles regime were moved between 20°C and 25°C every second generation, thus experiencing 14 temperature shifts.
Generation time in 20°C and 25°C was defined as the average difference in age between parents and offspring (Charlesworth 1994) and was calculated from the life-table of age-specific reproduction and survival following (McGraw and Caswell 1996) with trial data from the SP8 lines, and was 4.0 days in 20°C and 3.4 days in 25°C. This resulted 120 days of experimental evolution for Control 20°C, 114 days for Slow temperature cycles, 110 days for Increased warming and Fast temperature cycles and 101 days for Warm 25°C. For the two temperature cycle treatments, the worms spent shorter chronological time in 25°C than in 20°C, because of the faster generation time in 25°C. This ensured equal exposure to the two temperatures over biological time. Since we had no data for generation time in the intermediate temperatures between 20°C and 25°C, we did the simplifying assumption that the overall number of generations for the whole experiment would be similar in the Increased warming and Fast temperature cycle regimes. Therefore, the Increased warming regime was also run for 110 days, and the temperature increase was determined by the smallest temperature step the cabinet could be programmed (0.1°C).

Experimental evolution was conducted on 92 mm NGM-plates (Stiernagle 2006) and to combat infections the agar and bacterial LB also contained the antibiotics kanamycin and streptomycin, and the fungicide nystatin (Lionaki and Tavernarakis 2013; Lind et al. 2016). The plates were seeded with 2mL of the antibiotic resistant E. coli strain OP50-1 (pUC4K) obtained from J. Ewbank at the Centre d’Immunologie de Marseille- Luminy, France. To keep worm populations age-structured in overlapping generations, the populations were always kept in experimental growth face by cutting out a bit of agar containing 150 individuals of mixed ages and transferring this to freshly seeded plates. Transfer was conducted when needed (every 1-2 day), always before food was finished. Six independent replicate populations of each experimental evolution regime were set up, resulting in a total of 30 populations. All populations were expanded for two generations and frozen after 30 generations.

Phenotypic assays
Before assays, worms were recover from freezing and grown 2 generations in common garden, each generation synchronized by bleaching, a standard procedure that kills all life-stages but eggs (Stiernagle 2006). The common garden temperature was 20°C or 25°C (see below).

Phenotypic assays were performed to test for local adaptation to the experimental evolution regime and the evolution of adaptive maternal effects, and are summarized in figure 1. We therefore carried out three assays, by varying parental temperature (the 2 generations of common garden after defrosting the populations) and testing temperature for offspring. The 20°C - 20°C assay had both common garden and testing in 20°C. This is the environment the Control 20°C regime have experienced and tests for any cost of adaptation. Likewise, in the 25°C - 25°C assay both parents and testing worms experience 25°C, which is the selective environment for Warm 25°C and very close to the final environment for Increased warming. Finally, the 20°C - 25°C assay have 20°C as parental temperature, while the testing worms have their whole development in 25°C. This assay represents strong temperature fluctuations between generations, which is the selective environments for the Fast temperature cycle regime, and by comparing this assay to the 25°C - 25°C assay we can estimate the importance of maternal effects on fitness when adapting to a novel environment.

The assays were initiated by synchronised egg-laying in the testing temperature by 40 females of each population. After 5h, females were killed by bleaching, and setup of L4 larvae was initiated 39h later (in 25°C) or 50h later (in 20°C), due to temperature-specific development time. The setup consisted of eight testing females per plate, together with the same number of background males from the SP8 line. Sex ratio was kept 1:1 throughout the experiment by adjusting the number of males to match the number of females present. Age-specific fecundity was measured by each day allowing the females 3h of egg-laying on an empty plate, where after the females were returned to a new plate (together with the males) and the number of hatched offspring on the egg-laying plate were killed with chloroform and counted two days later. The exact time the females were added to and removed from each plate was noted, and the number of offspring was corrected by exact number of minutes available for egg laying, and the number of females alive. Thus, we did not collect individual level data on total reproduction, but daily snapshot, in order to increase the number of individuals assayed and improve the reproduction estimate of each population. Daily reproduction was collected until reproduction had ceased. Four replicate plates of each population was set up, and for the 20°C - 20°C and 20°C - 25°C assays the replicates were evenly split between two climate cabinets per temperature, in order to separate cabinet and temperature effects. However, for logistical reasons, the 25°C - 25°C assay was reduced. We excluded the Slow temperature cycle treatment from this assay, and unfortunately we lost two Warm 25°C populations during common garden (due to overcrowding and subsequent starving, which is known to induce epigenetic effects, Rechavi et al. 2014) and therefore these populations were excluded), leaving us with four replicate population of this treatment. This resulted in 30 replicate populations and 960 female worms for the 20°C - 20°C and 20°C - 25°C assays, and 23 replicate populations and 736 female worms for the 25°C - 25°C assay. A small number of plates were excluded from analyses, because of accidents during egg-laying or chloroforming: six plates were removed from the 20°C - 20°C assay, four plates from the 20°C - 25°C assay and six plates from the 25°C - 25°C assay.

Statistical analyses
The age-specific reproduction data was analysed as the total reproduction of each replicate plate, adjusted to the number of females present each day as well as rate-sensitive daily growth factor λ, which is equivalent to individual fitness (Brommer et al. 2002) but calculated per plate. λ encompasses the timing and number of offspring and is analogous to the intrinsic rate of population growth (Stearns 1992) and was calculated by constructing a population projection matrix for each plate, and then calculating the dominant eigenvalue from this matrix, following (McGraw and Caswell 1996). In the projection matrix, we set survival to 1 during the reproductive period, since any mortality during the early stages of the reproductive period (which is the most influential for λ) had non-natural causes (mainly worms climbing the wall of the plate where they dried out) and controlled for unequal number of worms by adjusting the reproductive value as described above. The first natural deaths in C. remanei are only observed long after the reproductive peak (Lind et al. 2016, 2017). Since population size and age-structure was kept constant during the experimental evolution, rate sensitive fitness is maximised during experimental evolution and λ is therefore the most appropriate fitness measure for this study (Mylius and Diekmann 1995).

All statistical analyses were done using R 3.6.1, and models were implemented using the lme4 package (Bates et al. 2015). Significance tests were performed using the lmerTest package (Kuznetsova et al. 2017), and contrasts were analysed using the emmeans package. Total reproduction and daily growth factor λ in 20°C was analysed in separate mixed effect models with experimental evolution regime as fixed effect and population and cabinet as random effects. Both response variables were log-transformed before analysis. In 25°C, λ and total reproduction was analysed with experimental evolution regime and parental temperature as crossed fixed effects, and replicate population as random effect. Models were constructed with a random slope (parental temperature) in addition to random intercept. Since the 25°C - 25°C assay was conducted in only one cabinet, and moreover the Slow temperature cycle treatment was not run, the random effect of cabinet was excluded from the parental temperature models, as was the Slow temperature cycle treatment.

The data is in three files, one for the 20C-20C assay (called Rawdata20), one for the 20C-25C assay (called Rawdata25) and finally one for the 25C-25C assay (called Rawdata2525).

*Selection: the experimental evolution regime

*Line: the replicate population number within the regime (use new.Line instead for unique ID)

*Plate: The replicate plate within the line (use new.Plate instead for unique ID)

*Day: the adult day (age) of the worms

*Cabinet: the climate cabinet the replicate plate was placed in (2 cabinets per temperature, except for the 25C-25C assay where only one cabinet was used)

*n.females: the number of females on the plate (started with 6 females, sometimes females are lost by climbing the walls and drying out between the daily assays)

*time start, time stop and timespan: the time females lay eggs on the plate (aimed for 3h, but differs sometimes a few minutes due to handling, this takes care of that.)

*n eggs: number of eggs scored on a plate, laid by n females

*n.eggs.3h: the avergae number of eggs per females over exactly 3h (this goes into analyses)

*new Line: unique line ID

*new Plate: unique plate ID