Heat stress responses and population genetics of the kelp Laminaria digitata (Phaeophyceae) across latitudes reveal differentiation among North Atlantic populations

Abstract To understand the thermal plasticity of a coastal foundation species across its latitudinal distribution, we assess physiological responses to high temperature stress in the kelp Laminaria digitata in combination with population genetic characteristics and relate heat resilience to genetic features and phylogeography. We hypothesize that populations from Arctic and cold‐temperate locations are less heat resilient than populations from warm distributional edges. Using meristems of natural L. digitata populations from six locations ranging between Kongsfjorden, Spitsbergen (79°N), and Quiberon, France (47°N), we performed a common‐garden heat stress experiment applying 15°C to 23°C over eight days. We assessed growth, photosynthetic quantum yield, carbon and nitrogen storage, and xanthophyll pigment contents as response traits. Population connectivity and genetic diversity were analyzed with microsatellite markers. Results from the heat stress experiment suggest that the upper temperature limit of L. digitata is nearly identical across its distribution range, but subtle differences in growth and stress responses were revealed for three populations from the species’ ecological range margins. Two populations at the species’ warm distribution limit showed higher temperature tolerance compared to other populations in growth at 19°C and recovery from 21°C (Quiberon, France), and photosynthetic quantum yield and xanthophyll pigment responses at 23°C (Helgoland, Germany). In L. digitata from the northernmost population (Spitsbergen, Norway), quantum yield indicated the highest heat sensitivity. Microsatellite genotyping revealed all sampled populations to be genetically distinct, with a strong hierarchical structure between southern and northern clades. Genetic diversity was lowest in the isolated population of the North Sea island of Helgoland and highest in Roscoff in the English Channel. All together, these results support the hypothesis of moderate local differentiation across L. digitata's European distribution, whereas effects are likely too weak to ameliorate the species’ capacity to withstand ocean warming and marine heatwaves at the southern range edge.

capacity and strong spatial structuring (King, McKeown, et al., 2018;Miller et al., 2019). Studies on local adaptation in L. digitata suggest that differentiation between populations could have occurred due to their geographic position (range central and marginal as well as southern and northern). King et al. (2019) investigated the expression of genes coding for heat shock proteins (HSP) in response to an hour-long heat shock in L. digitata from Scotland (range center) and Southern England (trailing edge). Maximum HSP response was present at 4-8°C higher temperatures in the southern populations in this short-term study, despite comparably low genetic diversity . The reduced genetic diversity and altered reproductive strategy in a southern marginal population in Brittany, France, also suggests that local differentiation has taken place (Oppliger et al., 2014;Valero et al., 2011). Overall, research on integrative responses such as growth is lacking when assessing the intraspecific thermal variation of L. digitata. Additionally, few studies on thermal responses of kelps incorporate physiology and population genetics over large geographic scales, although they may help to better predict climate change effects (Nepper-Davidsen, Andersen, & Pedersen, 2019).
The main objective of this study was thus to assess differentiation in heat stress responses among populations of Laminaria digitata present along the entire Northeast Atlantic distribution zone through a mechanistic, common-garden experiment. We hypothesized that an increasing thermal selection pressure toward the southern distribution limit increased heat resilience of sporophytes from southern L. digitata populations. Because of high similarities of thermal characteristics across regions reported in previous comparative studies (Bolton & Lüning, 1982;tom Dieck, 1992), we expected local differentiation in response to heat to be of small extent and to occur mainly toward the upper temperature limit (see also King et al., 2019). We further expected phenotypic differentiation to occur more prominently in populations experiencing low amounts of gene flow, while we expected low genetic diversity to be associated with reduced heat resilience as a result of genetic drift and possible maladaptation, which we investigated by the use of neutral microsatellite markers.

| Sample collection and preparation
We collected 30-35 fertile L. digitata sporophytes (Figure 1a Entire sporophytes were stored in ambient seawater for up to two days before processing. At the sampling locations, clean material from the meristematic region was preserved in silica gel for microsatellite genotyping. For the heat stress experiment, six disks (Ø 20 mm) were cut from the meristematic region of each sporophyte (i.e., 180 disks per population) in a distance of 5-10 cm from the stipe-blade transition zone. Disks were stored moist in cool boxes (<15°C) and transported to the laboratory within 30 hr. All experiments were performed at the Alfred Wegener Institute in Bremerhaven, Germany.

| Experimental design
We designed the experiment ( Figure 2) as a mechanistic short-term exposure to heat stress around the upper survival temperature of L. digitata sporophytes (21°C for a two week exposure; tom Dieck, 1992). A temperature of 19°C was considered to be a sublethal treatment for all populations, 21°C a threshold treatment (lethal over a longer exposure time; tom Dieck, 1992;Wilson, Kay, Schmidt, & Lotze, 2015), and 23°C a critical stress treatment (Bolton & Lüning, 1982), which also surpassed mean daily maximum temperatures of all sampled populations in 2018 ( Figure 1c). We exposed all samples to the same temperatures, irrespective of the ecological significance for local populations, to investigate the thermal plasticity and potential of L. digitata across its entire distribution range.
The heat stress experiment was conducted in independent runs in common-garden conditions with material from Spitsbergen, Tromsø, Helgoland, Roscoff, and Quiberon. Due to logistic constraints, Bodø had to be excluded, and Spitsbergen material was only tested for growth and fluorescence characteristics and not for biochemistry and pigments.
For each population, five replicate pools each contained all meristem disks of six distinct sporophytes to prevent pseudoreplication.
Irradiance ranged between 30 and 35 µmol photons m −2 s −1 at the bottom of the beakers in a 16:8-hr light:dark (L:D) cycle (ProfiLux 3 with LED Mitras daylight 150, GHL Advanced Technology, Kaiserslautern, Germany). Beakers were aerated gently to ensure motion of disks and even light and nutrient availability.
To allow recovery from sampling stress, disks were cultivated at 10°C for two (Tromsø) or nine days (Spitsbergen due to logistic issues), or at 15°C for four (Roscoff, Quiberon) or three days (Helgoland) before the acclimation phase of the experiment. From each replicate pool, eight disks were then randomly assigned to one replicate 2 L glass beaker in each of the four temperature treatment groups (15, 19, 21, 23°C, n = 5). Six disks per replicate were marked by punching a small hole on the outer rim with a Pasteur pipette to be frozen for biochemical and pigment analysis during the experiment. The two unmarked disks were used for growth and fluorometric measurements over the course of the experiment.
At the beginning of the experiment, disks were acclimated at 15°C for five days to obtain a similar metabolic state (day −5 to day 0; Figure 2). Although the northern populations Spitsbergen and Tromsø do not usually experience temperatures this high ( Figure 1c), 15°C is a temperature within the growth optimum of L. digitata (Bolton & Lüning, 1982;tom Dieck, 1992), which is considered to be stable (Wiencke, Bartsch, Bischoff, Peters, & Breeman, 1994), even for the Spitsbergen population (Franke, 2019). Starting the heat stress treatment on day 0, temperature was increased by increments of 2°C day −1 until the desired temperature was reached. The maximum temperature 23°C was applied for five days, while 21°C and 19°C were applied for six and seven days, respectively, according to the acclimation scheme ( Figure 2). On day 8, temperature was set to 15°C for all treatment groups to initiate a recovery period of seven days. Measurements took place at the beginning of the experiment (day −5; Figure 2), F I G U R E 2 Timeline of the heat stress experiment of Laminaria digitata. Dotted lines separate experimental phases of acclimation at 15°C (days −5-0), heat treatment (days 0-8), and recovery at 15°C (days 8-15). Growth and F v /F m were measured on days −5, 0, 3, 6, 8, and 15. On days 0 and 8, rapid light curves were performed and samples were frozen for biochemical and pigment analyses the beginning of the heat treatment (day 0), before applying the maximum temperature 23°C (day 3), in the middle of the heat treatment (day 6), at the end of the heat treatment (day 8), and after the recovery period (day 15).

| Relative growth rates
Two disks per replicate were repeatedly measured for growth over the course of the experiment (n = 5). Disks were blotted dry and weighed for growth analyses. Relative growth rates (RGR) were calculated as where x 1 = weight (g) at time 1, x 2 = weight at time 2, t 1 = time 1 in days, and t 2 = time 2 in days.

| PAM Fluorometry
Fluorescence parameters were assessed to estimate photoacclimation reactions in response to temperature (Davison, Greene, & Podolak, 1991;Machalek, Davison, & Falkowski, 1996) and were all conducted using a PAM-2100 chlorophyll fluorometer (Walz, Effeltrich, Germany). Maximum quantum yield of photosystem II (F v /F m ) was repeatedly measured in two disks per replicate over the course of the experiment following 5 min dark acclimation (n = 5).
Before and after the heat treatment (day 0 and day 8), rapid light curves (RLC) were conducted after F v /F m measurements on one disk (n = 3). RLC irradiance steps ranged from 0 to 511 µmol photons m −2 s −1 . Based on the photon flux density (PFD) and the effective quantum yield, relative electron transport rates (rETR) in photosystem II were calculated following Hanelt (2018) as rETR was plotted against PFD, and the resulting curves were fitted following the model of Jassby and Platt (1976) to calculate the maximum relative electron transport rate rETR max , the saturation irradiance I k , and the photosynthetic efficiency α of each curve.
Nonphotochemical quenching was calculated following Serôdio and Lavaud (2011) as where F m = maximum fluorescence of a dark-adapted sample, and F m ′ = maximum fluorescence of a light-adapted sample.
NPQ versus irradiance curves were fitted following the model of Serôdio and Lavaud (2011) to calculate maximum nonphotochemical quenching NPQ max , the saturation irradiance E 50 , and the sigmoidicity coefficient n.

| Biochemistry
Biochemical and pigment analyses were conducted with material from Tromsø, Helgoland, Roscoff, and Quiberon. We assessed the early photosynthetic product mannitol, which is accumulated during summer (Schiener, Black, Stanley, & Green, 2015), and elemental carbon and nitrogen to estimate carbon assimilation and nutrient storage in response to temperature. In wild sporophytes, assimilated mannitol is metabolized into the long-term storage polysaccharide laminarin and translocated into the distal thallus (Gómez & Huovinen, 2012;Yamaguchi, Ikawa, & Nisizawa, 1966).
As the meristematic region only contains minimal amounts of laminarin in wild sporophytes (Black, 1954), and as maximum laminarin contents occur with a seasonal delay of 1-2 months in late autumn (Haug & Jensen, 1954;Schiener et al., 2015), we did not assess laminarin storage in our short-term experiment on isolated meristematic disks.
Before the start and at the end of the heat treatment (day 0
They were ground under dim light conditions, weighed to 50-80 mg, and extracted in 90% aqueous acetone in darkness for 24 hr at 7°C.

| Statistical analyses of physiological parameters
As we measured two disks per replicate, we calculated growth rates and F v /F m from mean values per replicate. One disk was removed from the Spitsbergen 23°C treatment due to bleaching during the heating ramp. Despite identification efforts in the field, almost none of the microsatellite markers amplified in two samples from Spitsbergen (see also 2.3.2). This led to the assumption that the two samples were of which is morphologically very similar to L. digitata (Dankworth, Heinrich, Fredriksen, & Bartsch, 2020;Longtin & Saunders, 2015).
One replicate pool probably containing meristem disks from both species was therefore removed from the experiment. Due to the mannitol extraction performed in triplicates, means of the three subsamples of each mannitol replicate were analyzed. In carbon and nitrogen analyses, four data points were deleted due to a measuring error on day 0.
In the xanthophyll pool and de-epoxidation analyses, one outlier was deleted due to implausibly high zeaxanthin contents about four times higher than the next highest value.
All analyses of the heat stress experiment were performed in the R statistical environment version 3.6.0 (R Core Team, 2019). We fitted generalized least squares models for all parameters and tested for significance using analyses of variance (ANOVA). All models were fitted using the "gls" function from the R package "nlme" (Pinheiro, Bates, DebRoy, & Sarkar, 2019) with weights arguments to counteract heterogeneity of variance of normalized model residuals (Zuur, Ieno, Walker, Saveliev, & Smith, 2009 were modeled as interactive fixed effects and a compound symmetry correlation structure was incorporated using a time covariate and replicate as grouping factor (Pekár & Brabec, 2016;Zuur et al., 2009).
Analyses of variance were then performed on the models with the "anova" function to assess the effects of the fixed effects temperature, population and exposure time, and their interactions. For all biochemical, pigment, and fluorometric analyses, initial contents at day 0 were incorporated in the models as covariates to account for baseline differences, and temperature and population were modeled as fixed effects.
Analyses of variance were performed to assess the effects of the initial value covariate and the fixed effects temperature and population, and their interaction. Pairwise comparisons were performed using the R package "emmeans" (Lenth, 2019) and using the "Satterthwaite" mode for calculation of degrees of freedom and Tukey adjustment of p-values for multiple comparisons between independent groups. For pairwise comparisons in the repeated measures analyses (growth and F v /F m ), the "df.error" mode for calculation of degrees of freedom was applied.
Because of the repeated measures design and because the "df.error" mode overestimates the degrees of freedom (Lenth, 2019), p-values were adjusted by means of the conservative Bonferroni correction for multiple testing to reduce the probability of type I errors. Correlation analyses (Kendall's rank correlation) were conducted between all parameters measured after the heat treatment (relative growth rates calculated between day 0 and day 8) using the "cor.test" function from the default R package "stats" (R Core Team, 2019).
Amplification was faulty for the population of Helgoland sampled in 2018, which could be linked to poor preservation or insufficient dehydration. Therefore, the dataset of the same population sampled at the same site in 2016 was used in the genetic analysis instead. In total, 190 individuals were initially genotyped for twelve microsatellite markers and 179 were retained.

| Genetic diversity
Prior to genetic analysis, the presence of null alleles was tested using the ENA method in FreeNa (Chapuis & Estoup, 2007). Single and multilocus estimates of genetic diversity were calculated for VAZ:Chl a ratio mg mg −1 Chl a = V + A + Z Chl a De − epoxidation ratio = Z + 0.5A V + A + Z each population as the mean number of alleles per locus (N a ), unbiased expected heterozygosity (H e , sensu Nei, 1978), observed heterozygosity (H o ), and number of private alleles (P a ) using GenAlEx 6.5 (Peakall & Smouse, 2006). In addition, allelic richness (AR) was computed using FSTAT 2.9.3 (Goudet, 2001) for each locus using the rarefaction method. Linkage disequilibrium between pairs of loci and single estimates of deviation from random mating (F IS ) was calculated according to Weir and Cockerham (1984), and statistical significance was computed using FSTAT based on 7920 permutations for linkage disequilibrium and 10 4 for F IS . To test the null hypothesis that populations did not differ in genetic diversity, a one-way ANOVA was performed for AR, P a, and H e in R (R Core

| Population structure
Population structure was investigated first by the analysis of the pairwise estimates of F ST (Weir & Cockerham, 1984), and their significance were computed using FSTAT (Goudet, 2001). Second, a Bayesian clustering method as implemented in Structure 2.3.4 (Pritchard, Stephens, & Donnelly, 2000) was used to determine the existence of differentiated genetic groups within L. digitata populations categorizing them into K subpopulations. A range of clusters (K) from one to six was tested with 100 iterations, a burn-in period of 100,000, and a Markov chain Monte Carlo of 500,000 (Gilbert et al., 2012). The most likely value of K was determined using Evanno ΔK (Evanno, Regnaut, & Goudet, 2005) obtained using Structure Harvester (Earl & vonHoldt, 2012). Replicates of Structure runs were combined using CLUMPP software (Jakobsson & Rosenberg, 2007).

| Heat stress experiment
The significant main effects of independent factors are only reported in the absence of significant interactive effects. Therefore, in the presence of significant interactive effects, the simultaneous effects of two or more independent variables on a given dependent variable are given more emphasis than significant main effects.

| Growth
The significant population × temperature × time interaction for relative growth rates ( Figure 3; Table 1) indicates that growth in the temperature treatments differed significantly between populations over exposure time. However, there were differences in general growth activity between populations already during acclimation at 15°C  Over the recovery period at 15°C (Figure 3c), specimens from all populations showed significantly decreased growth after exposure to 23°C compared to lower temperature treatments (Bonferroni tests, p < .05). Spitsbergen and Tromsø essentially ceased growth (RGR < 0.001 and 0.002 g g −1 day −1 , respectively), while Helgoland, Roscoff, and Quiberon maintained slow growth (0.006, 0.004, and 0.01 g g −1 day −1 , respectively). However, during recovery after exposure to 23°C, there were no significant differences between growth rates of the different populations (Bonferroni tests, p > .05).
Quiberon material recovered best, in that there were no significant differences between the 15 and 21°C treatments while disks in these treatments simultaneously grew significantly faster than those from the former 23°C treatment (Bonferroni tests, p < .01).
In the more detailed time course of growth rates ( Figure A1), it became evident that all populations showed a trend of recovery from 21°C as growth rates increased between day 8 and day 15 ( Figure A1), which was significant only for Quiberon (RM ANOVA;

| Photoacclimative responses
Maximum quantum yield of photosystem II (F v /F m ) in the temperature treatments differed between populations over time, which is represented by the significant population × temperature × time interaction ( Figure 4, Table 1). After acclimation, all samples showed no signs of stress with F v /F m ranging between 0.7 and 0.8 (Figure 4a).
At the end of the heat treatment ( Figure 4b), temperature effects on quantum yield contrasted between the two populations of At higher temporal resolution ( Figure A2), a general difference between southern and northern populations became more pronounced.
While the significant decrease in quantum yield at 23°C took place between day 6 and day 8 for Helgoland, Roscoff, and Quiberon (RM ANOVA; Table A1; Bonferroni tests, p < .05), this decrease already started between day 3 and 6 in Spitsbergen and Tromsø material (Bonferroni tests, p < .001). Only specimens from Spitsbergen, as the most susceptible population, significantly decreased quantum yield also at 21°C, between day 6 and day 8 (Bonferroni test, p < .01).
The stronger heat susceptibility of Spitsbergen material became evident also following the recovery period ( Figure 4c). While all other populations recovered from 23°C, in that there were no F I G U R E 3 Relative growth rates of Laminaria digitata disks over the experimental phases of (a) acclimation at 15°C, (b) heat treatment, and (c) recovery at 15°C. Mean values ± SD (n = 5, for Spitsbergen n = 4 Contrary to quantum yield, the photoacclimation parameters obtained from rapid light curves at the end of the heat treatment, maximum relative electron transport rate rETR max ( Figure A3a), saturation irradiance I k ( Figure A3b), and photosynthetic efficiency α ( Figure A3c) did not show significant effects or interactions of temperature and population (Table A2). In contrast, nonphotochemical quenching (NPQ) parameters showed no significant interaction effects, but significant effects of population on maximum nonphotochemical quenching NPQ max and saturation irradiance E 50 , and of temperature on the sigmoidicity coefficient n ( Figure A4; Table A3). Mean NPQ max ( Figure A4a

| Biochemistry
Tissue mannitol and carbon contents were not significantly affected by interactive effects of population and temperature ( Figure 5; Tukey tests, p < .05). Carbon contents were not affected by temperature, but differed significantly only between populations (Figure 5b; Table 2). As with mannitol, Tromsø material maintained a higher carbon content, in that the means were significantly (7%-9%) higher in Tromsø and Helgoland material than in Roscoff and Quiberon material ((TRO = HLG) > (ROS = QUI); Tukey tests, p < .001).
Nitrogen contents were significantly affected by interactive effects of population and temperature ( Figure 5c;

| Pigments
Chlorophyll a content was not significantly affected by interactive effects of population and temperature, but differed TA B L E 1 Results of generalized least squares models to examine variability of relative growth rates (RGR) and maximum quantum yield (F v /F m ) of Laminaria digitata disks in the heat stress experiment significantly between populations ( Figure 6a; Table 3 The mass ratio of xanthophyll pigments per chlorophyll a (VAZ : Chl a ratio) was affected significantly by initial values, and interactive effects of population and temperature ( Figure 6b; Table 3). De-epoxidation ratios of xanthophyll cycle pigments were affected significantly by initial values, and interactive effects of population and temperature ( Figure 6c, Table 3). The significant differences between populations in mean de-epoxidation ratios over Note: Molar mannitol content, carbon content, nitrogen content, and C:N ratio were tested against initial values as covariate and interactive effects of population and heat stress temperature treatment. n = 5, n = 4 for Quiberon in carbon, nitrogen, and C:N ratio. numDF, numerator degrees of freedom; denDF, denominator degrees of freedom. denDF = 59 for carbon, nitrogen, and C:N ratio. Statistically significant values are indicated in bold text.

F I G U R E 6 Pigment characteristics of
Laminaria digitata disks after acclimation (day 0, empty circles) and after the heat treatment (day 8, colored points). (a) Chlorophyll a contents, (b) mass ratio of xanthophyll pigments per Chlorophyll a (VAZ : Chl a ratio), (c) de-epoxidation ratio of xanthophyll pigments. Mean values ± SD (n = 5, n = 4 for Tromsø 23°C in VAZ : Chl a ratio and deepoxidation ratio). Significant differences between mean population responses are indicated by lowercase letters (Tukey tests, p < .05). Significant differences between temperature treatments within populations are indicated by dashed lines (Tukey tests, p < .05). Significance levels are given in the text all temperatures (Table 3)

| Microsatellite amplification
Null alleles were present in every population for at least two markers (Table A5). However, differences between F ST values in the pairwise comparison were never greater than 10 -3 (data not shown).
Therefore, we concluded that the frequency of null alleles was negligible and our dataset was analyzed without taking into account correction for null alleles. No significant linkage disequilibrium was observed in any of the populations (Table A6). We thus considered all of the markers as independent. The number of alleles per locus ranged from 2 to 22 (Lo454-27 and Ld371, respectively).

| Genetic diversity
Values of genetic diversity averaged over the 12 loci are provided in Year: year of the samples used for genetic analysis (except for Helgoland, the genotyped individuals are the same than those analyzed for the heat stress experiment); n, number of individuals for which at least 11 markers amplified; N a , mean number of observed alleles; AR, allelic richness standardized for equal sample size (21 individuals); P a , mean number of private alleles per locus; H e , expected heterozygosity; H o , observed heterozygosity; F IS , fixation index (inbreeding coefficient) of individuals with respect to local subpopulation. All parameters are expressed as means over all markers ± standard error. *, significant departure from random mating after correction for multiple testing (p < .0069, FSTAT). locus by locus see Table A7). Most quantities varied by a factor of 1.5 among populations; the lowest genetic diversity was always observed in Helgoland and the highest in Roscoff. Variation was the highest for the mean number of private alleles (P a ) which ranged from 0.083 to 0.583. The differences between populations were not significant when each parameter was tested independently (oneway ANOVA, data not shown). However, a Fisher test of pairwise differences between means revealed that AR and P a were significantly lower in Helgoland compared to Roscoff (data not shown). In addition, three of the twelve loci were monomorphic in Helgoland, compared to the other populations, in which a maximum of one monomorphic locus was observed (Table A7).

| Genetic structure
Genetic differentiation was significant for each pairwise population comparison (p = .003 for all pairs; FSTAT) with an average F ST value of 0.3795 (Table A8) (Table A7), where one allele is fixed for all southern populations.

| Reproductive system
L. digitata from Tromsø and Helgoland did not show any significant departure from random mating (F IS ). We identified F IS > 0.1 for Spitsbergen, Bodø, Roscoff, and Quiberon, (

| D ISCUSS I ON
We identified a uniform growth limit across European Laminaria digitata populations following a short-term application of 23°C, which conforms with previous studies (Bolton & Lüning, 1982;tom Dieck, 1992).

| Similarities in growth and biochemical responses along the latitudinal gradient
Growth responses among our tested populations suggest that the upper temperature tolerance limit of Laminaria digitata is uniform along its European latitudinal distribution. Growth is an integrative parameter of all metabolic processes and can thus be interpreted as a proxy for organismal stress response. We observed that growth almost completely ceased in the 23°C treatment for all populations (Figure 3), while all populations showed signs of recovery from 21°C when transferred to 15°C ( Figure A1). The populations of Tromsø and Spitzbergen showed significantly lower overall growth rates than the southern populations. The lower growth rates of the Arctic populations might be related to prevailing local environmental conditions during sampling (e.g., long day lengths, cold temperature) which may influence growth rates and circannual rhythmicity in kelps (Olischläger & Wiencke, 2013;Schaffelke & Lüning, 1994). Still, results of our study using vious studies using laboratory-cultivated whole juvenile L. digitata sporophytes, which also showed uniform upper temperature limits on both sides of the Atlantic and Spitsbergen (Bolton & Lüning, 1982;Franke, 2019;tom Dieck, 1992 In addition to the strong similarities in the upper thermal limits of growth in our study, carbon contents ( Figure 5b) and chlorophyll a contents ( Figure 6a) did not differ between temperature treatments at all. In contrast, the overall trend of increasing mannitol contents at high temperatures (Figure 5a) has been described for Saccharina latissima (Davison & Davison, 1987) and might be linked to the seasonal increase in kelp mannitol storage in summer during the period of slow growth (Haug & Jensen, 1954;Schiener et al., 2015), which, in wild sporophytes, is followed by a peak of the long-term storage compound laminarin in autumn (Haug & Jensen, 1954;Schiener et al., 2015).
The consistent responses of growth and biochemical contents across populations reported here indicate a strong acclimation potential of L. digitata's metabolism to high temperature. Acclimation to wide temperature ranges would reduce selective pressure of temperature in the wild and might explain the small magnitude of local differentiation observed in this study.

| Differences in growth and photosynthetic parameters among marginal populations
Despite the stability of the upper thermal growth limit, we observed subtle physiological differences in the common-garden heat stress experiment, mainly in the marginal populations of Spitsbergen, Helgoland, and Quiberon. Maximum quantum yield of photosystem II was most sensitive to thermal stress at 21°C and 23°C in Spitsbergen material (Figure 4; Figure A2). This is concordant with the subarctic to Arctic regional climate and provides first evidence for a loss of function in a leading-edge L. digitata population, but whether this represents an adaptive trait is yet unknown. Generally, very few cold-temperate algae occurring in the Arctic show true adaptations to the Arctic climate compared to their Atlantic populations (Bischoff & Wiencke, 1993;Wiencke et al., 1994), possibly because the Arctic did not provide a sufficiently stable environment for adaptive evolutionary processes to occur (Wiencke et al., 1994).
At the southern range edge, a slight advantage of Q uiberon material to grow at elevated temperatures became evident in the growth response at 19°C during the heat treatment, and in the full recovery from the 21°C treatment (Figure 3; Figure A1). In contrast, photoacclimative responses suggest that the marginal population on the island of Helgoland was most resistant to heat stress. Photosystem II of Helgoland material was minimally impaired by 23°C (Figure 4).
Additionally, reactions of xanthophyll pigments (Figure 6b,c) were significantly weaker in Helgoland material than other populations.
Increased xanthophyll contents may indicate a photoprotective acclimation reaction (Latowski, Kuczyńska, & Strzałka, 2011;Pfündel & Bilger, 1994;Uhrmacher et al., 1995), while the de-epoxididation ratio of xanthophyll cycle pigments represents the current capacity to quench excessive energy from the photosystem (Pfündel & Bilger, 1994). Helgoland material did not show a significant increase in xanthophyll pigments and presented significantly lower de-epoxidation ratios and therefore lower nonphotochemical quenching (NPQ max , Figure A4) than all other populations. Therefore, the two populations growing in the warmest of the tested locations, which may experience >4 week long periods of mean in situ temperatures of 18°C to 19°C in summer (Helgoland: Bartsch, Vogt, Pehlke, & Hanelt, 2013;Wiltshire et al., 2008;Quiberon: Oppliger et al., 2014;Valero, unpubl.), showed slight physiological advantages to shortterm heat exposure in growth and stress responses.
The southernmost populations of Quiberon and Roscoff were curiously the only populations with significantly reduced tissue nitrogen contents in the heat treatments ( Figure 5c). A variety of factors including temperature affects nutrient uptake and consequently tissue nitrogen contents, which could be species-specific (Roleda & Hurd, 2019). Therefore, published studies on the impacts of heat stress on nitrogen uptake and storage in kelps differ in their reports of decreased (Gerard, 1997), unaffected (Nepper-Davidsen et al., 2019, or increased nitrogen contents (Wilson et al., 2015).
Whether the underlying cause of reduced nitrogen during heat in our study is adaptive, maladaptive, or neutral toward heat resilience in the southern populations remains unclear until further investigation.

| Population genetics in relation to physiological thermal responses
Population genetics suggest that the slight phenotypic divergence of L. digitata might have been facilitated through phylogeographic separation into two clades and low genetic connectivity between populations. The hierarchical division into a northern and a southern clade in the Northeast Atlantic (Figure 7a) is likely due to postglacial recolonization by two distinct genetic groups located in refugia proposed for the Armorican/Celtic Sea (Brittany and South West UK) and a potential northern refugium at the west coast of Ireland and/or Scotland (Neiva et al., 2020; see also King et al., 2020).
Currently, the highest genetic diversity (H e ≥ 0.6) published for L.
digitata populations was observed in Scotland (King et al., 2019, Northwest Ireland (Neiva et al., 2020), and Northeast Ireland (Brennan et al., 2014), which all exceeded the genetic diversity of the populations investigated in this study. Due to a lack of data, it remains unclear whether a potential glacial refugium of L. digitata also corresponds to the well-described Southwest Ireland refugium proposed for many marine species (Kettle, Morales-Muñiz, Roselló-Izquierdo, Heinrich, & Vøllestad, 2011;Provan & Bennett, 2008).
Populations at the "leading edge" (high latitude) are said to be associated with low genetic diversity due to recolonization processes following the Last Glacial Maximum (Hampe & Petit, 2005; for marine seaweeds of the North Atlantic see Assis, Serrão, Claro, Perrin, & Pearson, 2014;Neiva et al., 2016;Provan & Maggs, 2012).
Therefore, effects of genetic drift (e.g., depleted genetic diversity, increased inbreeding) may be expected to reduce physiological function in these populations. Here, genetic diversity characteristics of L. digitata at its northern range limit (i.e., Spitsbergen) were not significantly lower compared to the other populations in this study and were similar to other Northern Norwegian populations (Neiva et al., 2020). A similar pattern was observed for another Arctic to cold-temperate kelp species, Saccharina latissima (Guzinski, Mauger, Cock, & Valero, 2016). Therefore, rather than effects of genetic drift, a lack of selection pressure in the Arctic might have led to a potential reduction of heat tolerance at the northern distribution limit (i.e., relaxed selection; Lahti et al., 2009;Zhen & Ungerer, 2008).
Probably due to the continuous rocky substrata along the Brittany coast, connectivity may be maintained between Q uiberon and neighboring populations, which may explain a certain level of gene flow between Roscoff and Quiberon via stepping stone habitats (Figure 7c). Low gene flow can reduce inbreeding depression and associated deleterious effects and may facilitate local adaptation at this southern range edge (Fitzpatrick & Reid, 2019;Sanford & Kelly, 2011). Genetic diversity characteristics for Brittany L. digitata populations in this study comply with previous reports (Oppliger et al., 2014;Robuchon et al., 2014). Compared to Roscoff, genetic diversity of L. digitata from the island of Helgoland was significantly lower. The population's reduced genetic diversity can be partly explained by genetic isolation due to habitat discontinuity as Helgoland is a rocky substrate surrounded by continuous sandy seafloor (Reichert, Buchholz, & Giménez, 2008). This may rather suggest maladaptation due to less effective selection (such as in Fucus serratus; Pearson et al., 2009). However, samples from Helgoland presented the weakest heat stress response in this study. Therefore, we can hypothesize either that historically greater diversity/connectivity was reduced via isolation and drift after resilience to local conditions was established, or that strong selective forces toward the upper thermal limit of L. digitata have counterbalanced the effect of genetic drift.
Significant departures from random mating were only observed for the populations of Bodø and Roscoff (F IS ; Table 4) and match the magnitude of recent descriptions for L. digitata populations Neiva et al., 2020). The higher F IS values in Roscoff L. digitata in our study compared to the nearby population of Santec  might be explained by the distance of >1 km between sites, which may already cause substantial variation in F IS (Billot, Engel, Rousvoal, Kloareg, & Valero, 2003). In contrast, the higher F IS values of Quiberon L. digitata in our study compared to Oppliger et al. (2014) who sampled at the same location (Pointe de Conguel North) may be an artifact of differing microsatellite markers or might indicate a change in the reproductive system over time (Oppliger et al., 2014;Valero et al., 2011). In all cases, in the absence of data on reproductive ecology, the underlying causes remain speculative.

| Outlook
The mechanistic temperature treatments applied in this study do not represent realistic temperature scenarios for all tested populations, especially not for the northern clade. However, during our sampling period in August 2018, acute heat spikes surpassed 20°C on twelve days on Helgoland, and on nine days in Quiberon in the shallow sublittoral (in situ data; Bartsch, unpubl.; Valero, unpubl.). Also in South England, L. digitata already encounters marine heatwaves reaching 20°C (Burdett, Wright, & Smale, 2019;Joint & Smale, 2017).
According to predictions of ocean warming (Müller et al., 2009) and marine heatwaves (Oliver et al., 2018), L. digitata will possibly encounter prolonged summer periods of 21°C-23°C at its warm distribution limit until the end of the century.

CO N FLI C T O F I NTE R E S T
All authors declare that they are free of competing interests.  Notes: Fresh weight relative growth rates and maximum quantum yield F v /F m over all time points (T-5 (only F v /F m ), T0, T3, T6, T8, T15) were tested against interactive effects of heat treatment and time for each population separately. Generalized least squares models were performed as described in the methods section, but without the fixed effect for population. Tested values are means of 2 per replicate (n = 5, n = 4 for Spitsbergen). numDF, numerator degrees of freedom; denDF, denominator degrees of freedom. Statistically significant values are indicated in bold text.

TA B L E A 2
Results of generalized least squares models to examine variability of photoacclimation parameters of Laminaria digitata disks obtained via rapid light curves in the heat stress experiment ( Figure A3) Notes: Maximum relative electron transport rate rETR max , saturation irradiance I k, and photosynthetic efficiency α were tested against initial values as covariate and interactive effects of population and heat stress temperature treatment. n = 3, n = 2 for Spitsbergen. numDF, numerator degrees of freedom; denDF, denominator degrees of freedom. Statistically significant values are indicated in bold text.

TA B L E A 3
Results of generalized least squares models to examine variability of nonphotochemical quenching parameters of Laminaria digitata disks obtained via rapid light curves in the heat stress experiment ( Figure A4) Notes: Maximum nonphotochemical quenching NPQ max , saturation irradiance E 50, and sigmoidicity coefficient n were tested against initial values as covariate and interactive effects of population and heat stress temperature treatment. n = 3, n = 2 for Spitsbergen. numDF, numerator degrees of freedom; denDF, denominator degrees of freedom. Statistically significant values are indicated in bold text.

TA B L E A 4
Correlation coefficients (Kendall's rank correlation tau) and p-values in parentheses between relative growth rates (RGR), maximum quantum yield (F v /F m ), biochemical, and pigment characteristics of Laminaria digitata during / after the heat treatment.   Billot et al., 1998) and Laminaria ochroleuca (Lo; Coelho et al., 2014). The p-value after multiple testing correction for 5% nominal level is 0.000126. No linkage disequilibrium is significant in the dataset.

TA B L E A 6 (Continued)
TA B L E A 7 Estimates of genetic diversity and deviation from random mating for each locus and each population of Laminaria digitata tested in this study.  Billot et al., 1998) and Laminaria ochroleuca (L o ; Coelho et al., 2014); n, number of individuals for which the marker amplified; N a , number of observed alleles; AR, allelic richness standardized for equal sample size (21 individuals); P a , number of private alleles per locus; H e , expected heterozygosity; H o , observed heterozygosity; F IS , fixation index (inbreeding coefficient) of individuals with respect to local subpopulation; #NV, no calculation of F IS in monomorphic loci. Note that in Helgoland, Roscoff and Quiberon, the locus Lo454-27 is fixed while it is polymorphic for Spitsbergen, Tromsø and Bodø. This explains why this locus was not included in the study of Robuchon et al. (2014). Notes: All p-values obtained with 300 permutations using FSTAT were 0.003 and therefore significant (the p-value corrected for multiple testing is .003).

F I G U R E A 1
Relative growth rates (RGR) of Laminaria digitata disks from (a) Spitsbergen, (b) Tromsø, (c) Helgoland, (d) Roscoff, and (e) Quiberon over the heat stress experiment. Points represent growth rates between subsequent measuring days. Mean values ± SD (n = 5, for Spitsbergen n = 4). Points at day 0 represent growth over acclimation at 15°C, the end of the heat treatment at day 8 is marked with a vertical dotted line, and zero growth is marked with a horizontal dotted line. For statistical analysis, see Table A1.

F I G U R E A 2
Maximum quantum yield (F v /F m ) of Laminaria digitata disks from (a) Spitsbergen, (b) Tromsø, (c) Helgoland, (d) Roscoff, and (e) Quiberon over the heat stress experiment. Mean values ± SD (n = 5, for Spitsbergen n = 4). End of the acclimation at 15°C and end of the heat treatment are marked with dotted lines. For statistical analysis, see Table A1.

F I G U R E A 3
Photoacclimation parameters of Laminaria digitata disks obtained via rapid light curves after acclimation at 15°C (day 0, empty circles) and after the heat treatment (day 8, colored points). (a) Maximum relative electron transport rate rETR max (relative unit), (b) saturation irradiance I k (µmol photons m −2 s −1 ), (c) photosynthetic efficiency α (rETR/µmol photons m −2 s −1 ). Mean values ± SD (n = 3, for Spitsbergen n = 2). Analyses of variance returned no significant differences between populations (indicated by lowercase letters) and temperatures (Table A2).

F I G U R E A 4
Nonphotochemical quenching parameters of Laminaria digitata disks obtained via rapid light curves after acclimation at 15°C (day 0, empty circles) and after the heat treatment (day 8, colored points). (a) Maximum nonphotochemical quenching NPQ max (relative unit), (b) saturation irradiance E 50 (µmol photons m −2 s −1 ), (c) sigmoidicity coefficient n (unitless). Mean values ± SD (n = 3, for Spitsbergen n = 2). Significant differences between mean population responses are indicated by lowercase letters (Table A3; Tukey tests, p < .05). (Evanno et al., 2005) plotted against K, associated with K = 2 to K = 5 obtained with Structure Harvester during the analysis of genetic structure of Laminaria digitata populations