Regional and local environmental conditions do not shape the response to warming of a marine habitat-forming species

The differential response of marine populations to climate change remains poorly understood. Here, we combine common garden thermotolerance experiments in aquaria and population genetics to disentangle the factors driving the population response to thermal stress in a temperate habitat-forming species: the octocoral Paramuricea clavata. Using eight populations separated from tens of meters to hundreds of kilometers, which were differentially impacted by recent mortality events, we identify 25 °C as a critical thermal threshold. After one week of exposure at this temperature, seven of the eight populations were affected by tissue necrosis and after 30 days of exposure at this temperature, the mean % of affected colonies increased gradually from 3 to 97%. We then demonstrate the weak relation between the observed differential phenotypic responses and the local temperature regimes experienced by each population. A significant correlation was observed between these responses and the extent of genetic drift impacting each population. Local adaptation may thus be hindered by genetic drift, which seems to be the main driver of the differential response. Accordingly, conservation measures should promote connectivity and control density erosion in order to limit the impact of genetic drift on marine populations facing climate change.

. Two-way PERMANOVA results.Two-way PERMANOVA results from the common garden experiment considering the factors Locality and Depth and their interaction (Lo x De). Six populations were considered: Medes shallow and deep, Calanques 1 shallow and deep, Scandola shallow and deep. Df: degrees of freedom, SS: sum of squares, MS: mean squares, Pseudo-F: pseudo F statistic, P(perm): PERMANOVA p-value.

Factor
Factor levels Df SS MS Pseudo-F P(perm)    Supplementary Methods S1.Considerations on the time that colonies were placed in aquaria before starting the experiments and on energy reserves.

Time that colonies were placed in aquaria
Upon on the arrival of the samples were placed in the aquarium at 16-18ºC (temperature used in the control treatment) during 1 to several weeks. At our knowledge there is no information on the time required to adjust to the change in environment. However, since all samples were sampled and transported using identical treatment procedures we contend that the results were due to the populations' capacity to cope with thermal stress. During the experiment we conducted parallel tests to check the potential effect of acclimation period using different time from 1 day to 40days. In these settings the final outcomes in the response to the thermal stress were not significantly different. Besides other experiments carried with different red gorgonian Paramuricea clavata populations (see for instance 1-3 ) showed that 25ºC was the temperature threshold regardless the different timings used before and during the warming up period. Finally, the observed status of all control colonies at the end of the experiments (polyp activity and no tissue necrosis after feeding events) indicates that there was not effect of the time that colonies spent in aquaria before starting with the experiments.

Energy reserves
In order to determine whether or not the experimental conditions were suitable for the populations, we quantified the changes in energy reserves of the Control colonies over the experimentation as well as over sampled colonies in the field. As a proxy of energy reserves, we quantified the surface area (cm 2 /branch linear cm). The measure on surface areas was carried out using the NIH Image software. We compared the surface area (cm 2 /branch linear cm) of Control colonies at the beginning and at the end of the ladder experiment (see Table S4) conducted during the summer period, with that of colonies collected in the field during the fall (November 8). Overall there were significant differences in surface area between the three groups of colonies (F 2,27 =7.006, p-value=0.0035). The differences where due to a decrease in the area of the Control colonies over the summer period, between the beginning and the end of the ladder experiment (Scheffé Post Hoc test, p-value=0.014). The decrease in the area of the branches was about 25% which was similar to that observed in the field 4 . Accordingly, the area of the Control colonies at the end of the experiment did not differ from that of the branches collected in fall (Scheffé Post Hoc test, p-value=0.989).
The fact that the area of the colonies used in the experiments displayed a decrease in area during the summer period similar to that observed to occur in the field, indicates that energy reserves of the colonies in the aquaria performed rather similar to those in the field, suggesting that experimental conditions for Control colonies may be considered suitable for the red gorgonian Paramuricea clavata.

Supplementary Methods S2. Further information ofPERMANOVA analyses.
We chose the non-parametric method of PERMANOVA because of the non-normal distribution of the dependent variable 3 . This method uses Euclidean distance as the basis of the multivariate analysis (see 4 ) and relies on comparing the observed value of a test statistic (pseudo F-ratio) against a recalculated test statistic generated from random reordering (permutation) of the data 3 . Furthermore PERMANOVA was shown to be largely unaffected by heterogeneity of dispersions in the case of balanced designs 5 .
In order to simplify the analyses, we avoided to include the Aquarium nested factor in the model because previous one-way PERMANOVA tests did not show statistical differences between treatment Aquariums (p-value>0.05 for all pair-wise comparisons).
Accordingly, we pooled data from three replicates together to perform the analyses.
Instead of performing a repeated measure analysis of a single response variable, each time point was considered as a separate dependent variable (multivariate approach) because it completely avoided having to consider any notions of sphericity and interdependence of replicates 3 .

Supplementary Methods S3. DNA extraction and microsatellite genotyping.
Total genomic DNA extraction of the 240 colonies belonging to the 8 populations was conducted based on the salting out procedure adapted from Miller et al. 6 . The colonies were genotyped using six microsatellites (Parcla-09, Parcla-10, Parcla-12, Parcla-14, Parcla-17 and Par-d) following 7 . PCR products were analyzed on an ABI 3130 Genetic Analyzer with GeneScan 600 LIZ internal size standard (Applied Biosystems).
Genemapper version 3.0 (Applied Biosystems) was used to analyze the electrophoregrams. The occurrence of large allelic dropouts or scoring errors due to stuttering was checked with MICROCHECKER 8 . All the individuals that failed to amplify at more than 2 loci were deleted from the dataset. GIMLET 9 was used to remove repeated multilocus genotypes. Following analyses were conducted on 214 individuals. The frequencies of null alleles (r)were estimated for each population using the expectation maximization algorithm 10 in FREENA 11 . GENETIX 4.05 11