Study species
The fairy shrimp B. wolfi is one of the dominant crustacean species in temporary aquatic ecosystems in Southern Africa, from large pans to small rock pools. In this study, we used populations from rock pools on the summit of Korannaberg mountain (Free State Province, South Africa), S 28° 51′13″, E 27° 13′51″ (Additional file 1), which have been the subject of several ecological studies (e.g. [12, 21,22,23,24]). Adapted to short-lived temporary waters, B. wolfi individuals generally hatch within the first days of a new inundation and reach maturity within just 6 days after pool filling. Subsequently, they reproduce sexually to produce between 20 and 30 dormant eggs per day [25, 26]. Like the propagules of many zooplankton species, the eggs are highly resistant to adverse conditions, including extreme temperatures and drought, and can remain viable in a dormant state for many years [25].
Sediment collection and egg production
Dry sediment with B. wolfi propagules was collected following the protocol by Vanschoenwinkel and colleagues [26] from nine rock pools representative of the whole cluster of 44 pools. Care was taken to select different types of pools including small, large, deep and shallow pool basins.
To start-up populations in the Laboratory of Aquatic Ecology, Evolution and Conservation, KU Leuven, 150 g of dry sediment from each population was transferred to an 8 L plastic aquarium and inundated with “Environmental Protection Agency” (EPA) medium with a conductivity of 50 µs cm−1 (distilled water with 0.00033 mol L−1 NaHCO3, 0.000098 mol L−1 CaSO4·H2O, 0.00014 mol L−1 MgSO4, 0.000015 mol L−1 KCL) [27]. Aquaria were aerated and kept under a 12 h light:dark cycle (white light, full spectrum, 4000 lx, lamp type Osram L 8W/640; Osram, Rotterdam, The Netherlands) in a temperature controlled incubator at 18 °C. The medium was replenished daily to maintain constant water levels and hatchlings were fed ad libitum with the unicellular alga Scenedesmus obliquus (± 106 cells mL−1). Before reaching maturity, all B. wolfi individuals were isolated from the rest of the zooplankton community, transferred to clean aquaria of the same volume and kept under identical rearing conditions. Subsequently, the populations were bred for two generations under these common garden conditions to minimize long lasting maternal effects resulting from differences in the environmental conditions in the pools of origin.
Dormant eggs were harvested using a glass pipette and transferred to petri-dishes in temperature-controlled incubators at 18 °C and under a 12 h light:dark regime for dehydration. B. wolfi eggs of two age classes were used as starting material for the long term experiment; ‘old eggs’, which had been stored for about 12 months and ‘young eggs’, which were 2 months old. Both batches of old and young eggs were composed of a random mixture selected in equal proportions from all nine laboratory populations.
Temperature cycles
Hourly temperature data for the Korannaberg region were collected between 01 January 2012 and 31 December 2014 (Centre for Environmental Management, University of the Free State, Bloemfontein, South Africa). Based on these measurements, we calculated the average temperature for each hour of the day, with separate values calculated for each month of the year. We converted these ambient air temperatures to the temperatures actually experienced in the sun-exposed sediment of a rock pool at a depth of 0.5 cm using an hourly conversion factor (Additional file 2). The hourly conversion factors were determined by calculating the ratio between the temperature that was measured at 0.5 cm depth in sun-exposed sediment of a rock pool and the air temperature at that moment.
Climate models predict a temperature increase of approximately 4 °C for the Korannaberg region by 2070 [28]. Based on this prediction and the ambient-sediment conversion factor, we reconstructed daily temperature cycles that are anticipated under climate change in 2070 (Additional file 2). In addition to current and cycles of expected future temperatures, we included a third treatment at a constant 18 °C, which is the temperature that results in optimal egg survival and subsequent hatching under laboratory conditions (unpublished data).
Incubation experiment, egg survival and hatching trials
A total of 2448 ‘old’ and 2448 ‘young’ viable B. wolfi eggs were divided randomly over the three temperature conditions. Since the laboratory experiment was initiated in the month of October, we opted to use the October temperature conditions as a starting-point (Additional file 2). Intact dry eggs (i.e. eggs with no external signs of degradation) were placed individually into the empty wells of a 24-well polystyrene multi-well plate. B. wolfi eggs rapidly disintegrate when the embryo is dead, so external features give a reasonable indication of viability. Each egg was assigned to a separate well to avoid any potential interference among eggs and to ensure statistical independence. Still, we chose to include ‘plate identity’ as a random factor in our analyses to correct for any potential plate-effects.
Plates were randomly positioned in temperature controlled incubators, which were programmed to maintain the desired temperature regimes, under a 12 h light:dark cycle and a constant relative humidity of 70%. To minimize any confounding effects of the different incubators, temperature regimes were re-divided across incubators and plates were randomly repositioned within incubators three times during the experiment. Furthermore, plates were randomly repositioned within each incubator on a weekly basis to exclude position effects.
We investigated the effects of the different temperature regimes on survival rates and hatching fractions of both old and young eggs after 8, 16, 24 and 36 weeks. Hatching fractions were established during common garden experiments under optimal hatching conditions (cf. [23]). During each hatching experiment, seven 24-well plates (i.e. 168 eggs) were taken from each of the six conditions and each well was inundated with 2 mL of EPA medium with a conductivity of 50 μS cm−1. The plates were randomly positioned within an incubator at 18 °C and under continuous light (white light, full spectrum, 4000 lx). Hatching was evaluated under a light microscope with a 40× magnification at 12 h intervals until no further hatching was observed for 36 h. At the end of each hatching experiment, the viability of every individual egg was checked according to the protocol of Pinceel and colleagues [24] by removing the egg shell with a fine pair of tweezers and evaluating the embryo under a light microscope. Based on this, the number of dead eggs was subtracted from the original number of eggs before calculating hatching fractions.
Statistical analyses
All analyses were performed in R v. 3.3.1 (R Development Core Team, Vienna, Austria, 2014) and the ‘lme4’ (version 1.1.10) and ‘multcomp’ (version 1.4.6) packages. We tested for effects of temperature regime on egg survival and hatching fraction using generalized linear mixed models (GLMM) with a binomial error distribution and corresponding logit link function since both egg survival (dead or alive) and hatching (no hatch or hatch) were measured as the binary response of individual eggs. In a first GLMM, ‘plate identity’ was included as a random factor, ‘temperature regime’ (present-day cycle, future cycle, constant 18 °C), ‘incubation period’ (8, 16, 24 or 36 weeks) and ‘egg age’ (old or new) as fixed categorical predictors and the binary variable ‘egg survival’ (dead or alive) as response variable. To test for effects of the temperature treatments on hatching, we used a second analogous GLMM with ‘egg hatching’ as a binary response variable. The models were built using the glmer function in the lme4 package for linear mixed effects models. We used likelihood ratio tests (LRT) to test the significance of the main effects. We did this by comparing models with one main effect (the one of interest) excluded to a model with all main effects and the random effect via the ‘drop1’ function. In addition, we performed Tukey post hoc tests using the ‘glht’ function in the ‘multcomp’ package to investigate pairwise differences in survival among resting eggs from the three different temperature regime treatments. Since different plates of eggs were removed from the incubators at each time step to test viability and hatching, there was no need for a repeated measures design with incubation period as a random factor.
Matrix population model
We estimated the impact of measured changes in survival of B. wolfi eggs on population demographics using a stochastic matrix population model [12]. This model was developed using realistic life-history parameters to simulate population growth rates and extinction risks across 1000 inundations from eight hydrological regimes (median inundation lengths from 5 to 12 days) that closely match those found in nature [12, 29]. The model comprises two life stages, one corresponding to two age classes to account for age-specific trait values: (1) eggs produced during the previous inundation, N0, (2) older eggs in the egg bank, N1, and (3) individuals from the active population in the water column, N2. Detailed information on the modelling procedure is included in Additional file 3 and in Pinceel et al. [12]. The selection of egg survival parameter values is motivated further in Additional file 4.