Anthropogenic drivers are restricting many species to small, genetically isolated populations. These are prone to inbreeding depression and are at an increased risk of extinction. Genetic rescue, the controlled introduction of genetic variation from another population, can alleviate inbreeding effects. A major conservation concern, restricting the use of this technique, is that such augmented gene flow may disrupt local adaptation crucial to a population’s persistence. Using populations of the red flour beetle (Tribolium castaneum) experimentally adapted to reproduce at higher temperatures, we assess whether genetic rescue attempts disrupt thermal adaptation. Rescuers, drawn from populations adapted to either 30°C or 38°C, were introduced into populations adapted to 38°C, which had been inbred for two generations. We recorded population productivity for three generations post-rescue, in the adapted 38°C environment. Rescuers with and without local adaptation significantly increased the productivity of recipient inbred populations but, importantly, those sharing local adaptation to reproduction at 38°C provided greater increases in productivity. For the first time, we show that co-adaptation between rescuing individuals and rescuee population maybe an essential aspect of achieving desired conservation outcomes.

Husbandry

T. castaneum populations were maintained on standard fodder (90% white organic flour, 10% brewer’s yeast and a layer of oats for traction) in a controlled environment of 30°C (unless otherwise stated) and 60% humidity with a 12:12 light-dark cycle. Populations were maintained following a standard cycle of virgin adults having seven days of mating and oviposition followed by the removal of adult beetle, using 2 mm and 850 µm sieves, so that only eggs remain in the fodder. Each generation is initiated with a number of adults (line dependent, see below) that are given 7 days to mate and lay eggs before being removed. The eggs are left for 35 days to develop into mature adults.

Tribolium castaneum lines

Krakow super strain (KSS): A combination of fourteen laboratory strains bred to maximise genetic diversity in one population maintained at a census size of 600 individuals (Laskowski et al., 2015). This line is highly productive at 30°C but has reduced fitness at 38°C. This was used as the non-adapted rescuer population.

Thermal lines: Ten independent lines (census size = 100 adults) founded from KSS, and  experimentally evolved for ~150 generations at an environmental temperature of 38°C (Dickinson, 2018), thus imposing selection for development and reproduction at this temperature, considerably above the thermal optimum for *T. castaneum *(Howe, 1962). All other conditions were as described above aside from a shorter development period of 27 days, reflecting accelerated development at 38°C. These were used as the thermally adapted rescuer populations.

Inbred lines: Ten inbred populations were created, one from each of the ten thermal lines described above. Adult beetles from each thermal line were housed individually for two weeks to ensure any fertilized eggs were laid. Single-pair matings were formed by housing together a previously isolated male and a female for 7 days of mating and oviposition, resulting in a single pair bottleneck for each thermal line. Full sibling offspring resulting from this pairing were again paired for a second bottleneck. The following generation was initiated with 10 male and 10 female full sibling offspring of full sibling pairs.  From the offspring of these groups, six inbred experimental populations (10 males and 10 females) were created from each of the 10 inbred thermal lines, to act as recipient populations for genetic rescue. One inbred population only produced four experimental populations, resulting in a total of 58 experimental populations split over two temporal blocks of 30 and 28. The two blocks were maintained one day apart for ease of handling but otherwise received identical treatment. Each population received a random ID number to blind the experiment and avoid bias. Experimental populations were initiated every generation using 10 males and 10 females sourced from the offspring of the previous generation to reduce density dependent effects (Duval et al., 1939; King and Dawson, 1972; Janus, 1989). All experimental inbred recipient populations were kept at 38°C in A.B. Newlife 75 Mk4 forced air egg incubators (A.B. Incubators, Suffolk, UK); all other conditions were kept as described above.

Genetic rescue protocol

Populations were kept in 125 ml tubs containing 70 ml of standard fodder. After seven days, adults were discarded and eggs were left to develop for ~21 days when 10 male and 10 female pupae were randomly selected to establish the next generation. Remaining individuals were maintained for 10 days following this and were then frozen and manually counted. Pupae taken at day ~21 were housed in plastic dishes containing 10ml standard fodder in single-sex groups until they matured into adults after 10±2 days, and the next generational cycle began with unmated adults, avoiding overlapping generations.

After a rest generation a single male from each inbred recipient population was removed and replaced with a single male rescuer to avoid demographic rescue effects (increased population fitness due to increased population size) (Bell et al., 2019; Ingvarsson, 2001). Three treatments were created: 1) control - 19 populations received no rescue (the male was not removed); 2) locally adapted rescue - 19 populations received a 38°C-adapted rescuer (a male from a different thermally adapted population); 3) non-locally adapted rescue - 20 populations received a non-thermally adapted rescuer (a KSS male, see above) (figure 1).

Population fitness was measured using productivity: the number of mature adult offspring the population produced each generation. Populations were maintained for three non-overlapping generations following rescue. Two replicates were lost after two generations due to human error, resulting in a third generation with 19 control, 18 thermally adapted rescue, and 19 non-thermally adapted rescue populations. Data for populations lost in generation three was included in the analysis of generations one and two.

Statistical analysis

R v.4.4.1 (R Core Team, 2024) was used with R Studio version 2024.04.2+764 (Posit team, 2024). Data management and exploration were performed with tidyverse (Wickham et al., 2019), stats (R Core Team, 2024), Rmisc (Hope, 2022) and googlesheets4 (Bryan, 2023). ggplot2 (Wickham, 2016) was used to visualise results. Data distribution was checked using the shapiro.test function (R Core Team, 2024). The glmmTMB package (Brooks et al., 2017) was used to fit generalised linear mixed models (GLMMs). DHARMa (Hartig, 2022) was used to check model fit and the check_collinearity function from the performance package (Lüdecke et al., 2021) to test variance inflation factor scores (VIF). No overdispersion or collinearity (VIF<3 for all variables) was found. R^2 ^was determined using the r.squaredGLMM function in MuMIn (Bartoń, 2024).

Counts of population productivity over all generations was analysed using a GLMM with a negative binomial errors and a log link function. Fixed explanatory variables were rescue treatment and generation as well as the interaction between these variables. To account for populations variance and relatedness between replicate populations a random factor was added nesting individual ID within the stock thermal line the population descended from. The control treatment was set as the baseline factor for comparison. The baseline was changed to non-adapted to compare between the two rescue treatments post-hoc.

GLMMs, constructed as described above, but excluding the generation variable, were then run post-hoc on each generation individually to test if there were significant differences between the treatments in each generation. The baseline was changed to non-adapted rescuers, to compare between the two rescue treatments post-hoc.