We designed an experiment using sibling clonal gametophyte isolates from the L. digitata populations of Helgoland (North Sea) and Spitsbergen (Arctic), because they represent genetically distinct populations thriving under contrasting thermal conditions along the latitudinal distribution gradient (Liesner et al., 2020a). We induced fertilization of gametophytes, creating clonal lineages of inbred selfings and outbred reciprocal crosses, and defined the upper thermal tolerance and differential survival capacity of these microscopic sporophytes over time (experiment 1). At a later stage, three-month-old and 4–7 cm long sporophytes of the same selfings and crosses were subjected to control (10°C) and sublethal temperature (19 and 20.5°C; based on the results of experiment 1) for 18 days. We measured the physiological traits of growth and photosynthetic optimum quantum yield, and compared transcriptomic responses to the highest sublethal heat stress at 20.5°C among selfings and outbred crosses (experiment 2).

We used sibling clonal gametophyte cultures each derived from a single gametophyte isolate in 2015 of one Laminaria digitata sporophyte from Kongsfjorden, Spitsbergen, Norway (AWI seaweed culture collection number: ♀ 3472, ♂ 3471) and one sporophyte from Helgoland, Germany (AWI seaweed culture collection number: ♀ 3436, ♂ 3435). Spitsbergen material had been verified as L. digitata morphologically and by DNA barcoding of the sporophyte (ID 78 in Dankworth et al., 2020) to avoid confusion with the morphologically similar Arctic kelp Hedophyllum nigripes (Franke et al., 2021). Prior to the start of the experiment, cultures were maintained vegetatively under red light (approx. 3 μmol photons m^–2 s^–1; ProfiLux 3 with LED Mitras daylight 150, GHL Advanced Technology, Kaiserslautern, Germany) in a 16:8 h light:dark cycle at 15°C in a temperature-controlled cooling chamber (error ± 1°C) in sterile Provasoli-enriched seawater (PES; Provasoli, 1968; modifications: HEPES-buffer instead of TRIS, double concentration of Na₂glycerophosphate; iodine enrichment following Tatewaki, 1966).

To perform fertilization of selfings and outcrosses, stock suspensions of each gametophyte culture were prepared by gently fragmenting gametophyte material using a mortar and pestle. Suspensions were sieved to obtain a fraction of filaments of length 50–100 μm. Gametophytes were added to petri dishes (Ø 5 cm) containing four glass cover slips and filled with 12 mL half-strength (½) PES. This created the “lineage” treatments (n = 4) of H × H (Helgoland female × Helgoland male), H × S (Helgoland female × Spitsbergen male), S × H (Spitsbergen female × Helgoland male), S × S (Spitsbergen female × Spitsbergen male). Despite efforts to sow gametophytes at identical densities, Spitsbergen gametophytes (f: 227 ± 22 gametophytes cm^–2 14 days after sowing; m: 195 ± 10 gametophytes cm^–2; mean over all replicates ± SD, n = 8) were about twice as dense as Helgoland gametophytes (f: 120 ± 11 gametophytes cm^–2; m: 95 ± 15 gametophytes cm^–2). Gametogenesis was induced at 10°C and 15–18 μmol photons m^–2 s^–1 white light, which is optimal for both populations (tom Dieck, 1992; Martins et al., 2020). After 28 days, means of 75–80% (S × S, S × H, H × S) and 90% (H × H) of female gametophytes had sexually reproduced to generate sporophytes [ANCOVA, F_(3,11) = 8.13, p = 0.0039].

Following sporophyte production, the four cover slips per replicate were divided into four plastic dishes (Coria, Ø 6 cm, polystyrol, VKF Renzel, Germany) to conduct an experiment on the upper thermal limit of microscopic sporophytes (experiment 1). The remaining sporophytes in the petri dishes were pooled by lineage and were reared for 73 more days first in glass dishes (Ø 9 cm) at the same conditions. Macroscopic sporophytes were subsequently cultivated in 1 L glass beakers and 5 L bottles with gentle aeration at 10°C under increased irradiance of 30–35 μmol photons m^–2 s^–1 with weekly changes of ½ PES and were then used in the heat stress experiment on cm-long juvenile sporophytes (experiment 2). Throughout the experiments, 10°C served as the control temperature because it has been described as an optimal growth temperature for temperate and Arctic sporophytes of L. digitata (Bolton and Lüning, 1982; tom Dieck, 1992; Franke et al., 2021).

The four cover slips of one gametogenesis replicate were divided and each assigned to one experimental replicate in each temperature treatment to define the upper thermal limit of microscopic sporophytes (20, 21, 22, 23°C ± 0.1°C). Replicate dishes each containing one coverslip and 100 mL ½ PES were acclimated at 14°C for 11 days, followed by 18°C for two days before reaching the experimental temperatures in water baths controlled by thermostats (Huber Variostat CC + Pilot ONE, Peter Huber Kältemaschinen GmbH, Offenburg, Germany). The number of healthy (fully pigmented), unhealthy (partial bleaching) and dead (fully bleached) sporophytes was counted at the start of the experiment (day 0), after seven and 14 days at the temperature treatments, and after a post-cultivation period of 14 days at the sublethal and growth-suppressing temperature of 20°C (Bolton and Lüning, 1982; tom Dieck, 1992) to identify any delayed treatment effects. In each replicate, ∼300 sporophytes were randomly counted. Medium was changed at the start of the experiment and at the end of the 14-day temperature treatment. Mid-parent heterosis (MPH; hybrid vigour) of survival in the reciprocal crosses was calculated by subtracting the average relative survival over all replicates of both inbred lineages (H × H and S × S) from each single replicate of an outcrossed lineage (either H × S or S × H, respectively), resulting in the difference in percentage of survival for each replicate of the crosses, following Hochholdinger and Hoecker (2007):

where F₁ is the trait of a cross (one replicate of either H × S or S × H, respectively), and P₁ and P₂ are average traits of the parental inbred lineages (all replicates of H × H and S × S).

In the second experiment, 4–7 cm long sporophytes of the same inbred and outcrossed lineages were subjected to temperatures of 10°C (control), 19 and 20.5°C to assess heat stress responses among lineages. These temperatures were chosen based on the results of experiment 1, where the majority of sporophytes survived for 14 days at 20°C and only the Spitsbergen selfing showed reduced tolerance at 21°C (Figure 1), which corresponds to the published upper thermal limit of L. digitata (Bolton and Lüning, 1982; tom Dieck, 1992; Franke et al., 2021). Seven sporophytes were assigned to one replicate plastic container (n = 4; Wide neck containers series 310 PETG, 2000 mL, Kautex GmbH & Co. KG, Bonn-Holzlar, Germany) filled with 1.8 L of ½ PES. Medium was exchanged every 3–4 days. Samples were acclimated for one day at 13.5°C and one day at 17°C before reaching the experimental temperatures of 19 and 20.5°C on (nominal) day 0 of the experiment, while the control treatments remained at 10°C. Two sporophytes per replicate were marked to be used for repeated measurements of growth and quantum yield throughout the experiment by punching small holes in the distal thallus with a Pasteur pipette. Of the unmarked five sporophytes per replicate, three were frozen in liquid nitrogen throughout the experiment (before acclimation, day 1 of temperature treatment, day 18 of temperature treatment). Samples frozen before acclimation and after 18 days of temperature treatment were used for transcriptomic analysis, stored at -80°C and processed within three weeks. The remaining two sporophytes served as backup.

Physiological measurements were conducted on days -2 (before acclimation), 0 (beginning of experiment), 3, 7, 10, 14 and 17 on two sporophytes per replicate. Average values per replicate were used for statistical analyses. Fluorometric measurements were conducted with an Imaging-PAM (M-Series, MAXI version, Heinz Walz GmbH, Effeltrich, Germany) following 10 min dark acclimation. Maximum quantum yield (F_v/F_m) was measured in the meristematic region of the sporophytes. Fresh weight was quantified after patting dry sporophytes with paper wipes. Relative growth rates were calculated as

where x₁ and x₂ are fresh weights at the successive time points t₁ and t₂, respectively.

All analyses were conducted in R 4.0.4 (R Core Team, 2021). To analyse survival of microscopic sporophytes in experiment 1, we calculated fractions of unbleached (defined as healthy) sporophytes in each replicate to produce survival curves. We then modelled percentages of healthy sporophytes after 14 days of exposure against a density covariate and the fixed factor lineage for each experimental temperature using linear models (function “lm”). Relative growth rates and F_v/F_m in experiment 2 were modelled as a function of the fixed factor lineage for each temperature separately on day 3 (all lineages) and day 17 (S × S only included at 10°C because of mortality at 19 and 20.5°C) using linear models. Normal distribution and variance homogeneity of standardised model residuals were verified using Shapiro-Wilk and Levene’s test, respectively. Factor significance was assessed via analyses of variance (ANOVA) and p-values were corrected for multiple testing using the false discovery rate approach (FDR; Benjamini and Hochberg, 1995). Significant differences in MPH were assessed using t-tests.

For RNA extraction from frozen sporophyte material, we applied the protocol of Heinrich et al. (2012a) with modifications described by Heinrich (2018). RNA purity was inspected by NanoDrop ND-1000 Spectrophotometer (NanoDrop Technologies LLC, Wilmington, United States) and RNA integrity was confirmed by capillary electrophoresis (Agilent 2100 Bioanalyzer, Agilent Technologies, Santa Clara, United States). The RNA was enriched for polyA+ species and converted to a cDNA sequencing library using a TruSeq Stranded mRNA kit (Illumina, San Diego, United States) according to the manufacturer’s protocol. The cDNA was sequenced as 75 bp paired end libraries on an Illumina HiSeq 4000 at Cologne Center for Genomics, Cologne, Germany. Raw reads were quality trimmed with Trimmomatic (Bolger et al., 2014) and deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession PRJNA665130.

The L. digitata transcriptome was assembled de novo from 33 libraries (n = 3) sampled on day 0 (10°C, N = 12) and day 18 (10°C, N = 12, and 20.5°C, N = 9) of experiment 2 using rnaSPAdes (Bushmanova et al., 2019) with kmer length 37 on a Linux server (40 cores and 250Gb RAM). Quality assessment of the assembly was performed using Transrate v1.0.3 (Smith-Unna et al., 2016) with read-mapping metrics (SNAP; Zaharia et al., 2011; Salmon; Patro et al., 2015). A reference set of transcripts was built by predicting open reading frames (ORFs) from the high-quality contigs returned by Transrate using FragGeneScan (Rho et al., 2010). Subsequent clustering of predicted ORFs was performed at 97% nucleotide identity using VSEARCH (Rognes et al., 2016). Clustered ORFs were then screened for biological evidence against Stramenopile proteins in the NCBI nr database using local blastx, retaining only those with top hits against Phaeophyceae to produce the final reference transcriptome. The completeness of this reference was evaluated with Benchmarking Universal Single-Copy Orthologs (BUSCO) v2/3 against the OrthoDB v9 eukaryote database (gVolante server¹; Nishimura et al., 2017). The reference was annotated by local blastx against the Uniref90 and Swissprot databases² and by blastp against the KEGG database (Kyoto Encyclopedia of Genes and Genomes; KAAS server³). Accessions for hits to Uniref90 were linked to Gene Ontology, InterPro and Pfam functional terms.

Read pairs for each library were mapped onto the indexed reference transcripts with Bowtie2 (Langmead and Salzberg, 2012) using RSEM (Li and Dewey, 2011), which returned expected count data for each contig. Count data were used for differential gene expression analysis using edgeR (Robinson et al., 2009; McCarthy et al., 2012), following the pipeline described by Smyth et al. (2018) with the requirements for differentially expressed transcripts of log₂-fold change (log₂FC) > 1 and adjusted p-value < 0.01 (FDR). The analysis was repeated on KEGG orthology (KO) terms, after aggregating read counts where mulitple transcripts were assigned to the same KO entry. Comparisons were made only for samples taken on day 18 to identify (a) the general heat stress response across lineages, (b) differences in gene regulation among lineages at the control temperature of 10°C, and (c) differences in the heat stress response among lineages at 20.5°C. Due to mortality of the Spitsbergen selfed sporophytes at 20.5°C, this treatment could not be included in the comparisons, and one further replicate of the Spitsbergen selfing at 10°C was removed as an outlier (S × S 10 c in Supplementary Figure 1).

Figure 1 shows the percentage of fully pigmented (“healthy”) microscopic sporophytes for each lineage over 14 days of heat treatment and subsequently 14 days of post-cultivation at 20°C in four panels corresponding to the temperature treatments. While almost all sporophytes survived at 20°C for 14 days, 23°C was lethal for most sporophytes across lineages. At 20°C, there was no significant difference among lineages as 99–100% of sporophytes remained healthy across lineages over the 14-day treatment (p = 0.4164, Table 1), but in the Spitsbergen selfing the fraction of healthy sporophytes decreased to 83% during the subsequent 14 days of post-cultivation at 20°C. Major temperature effects were visible in S × S at 21°C, which significantly decreased the fraction of healthy sporophytes to 68% after 14 days compared to all other lineages, which remained 93–99% healthy (p < 0.0001, Table 1; Tukey tests, p < 0.0001). Again, health of S × S sporophytes further decreased to 28% during post-cultivation at 20°C. At 22°C, health of S × S sporophytes decreased drastically to 6%, while the other crosses remained at 70–80%, and the Helgoland selfing retained 90% healthy sporophytes [p < 0.0001, Table 1; Tukey tests, H × H > (S × H = H × S) > S × S, p < 0.05]. At 23°C, fractions of healthy sporophytes averaged 0% (S × S), 0.08% (S × H), 0.58% (H × S) and 2.96% (H × H), but were not significantly different among lineages (p = 0.0600, Table 1). Only in the Helgoland selfing, 4.5% of healthy sporophytes persisted during post-cultivation. These results led to the decision to perform experiment 2 with a control temperature of 10°C, 19°C as a moderate heat stress treatment and 20.5°C as a critical but sublethal heat stress treatment.

The survival of the reciprocal crosses at 21 and 22°C can be related to the survival of the inbred lineages by quantifying mid-parent heterosis (MPH). This parameter describes whether the performance of hybrid offspring is significantly better than the average value of the two parental inbred lines (Hochholdinger and Hoecker, 2007). Mid-parent heterosis was evident in both outcrossed lineages after 14 days of exposure to both 21 and 22°C. Namely, in H × S, relative survival at 21°C was higher by 9.7% ± 1.2% (mean ± SD, n = 4, p = 0.0017, t-test) and by 23.3% ± 3.5% at 22°C (p = 0.0028) compared to the average response of both selfings. In S × H, relative survival at 21°C was higher by 14.1% ± 0.7% (p = 0.0002) and by 31.4% ± 2.7% at 22°C (p = 0.0008) compared to the average selfing response. Contrary to the relative survival rates, MPH was significantly higher in S × H compared to H × S at 21°C (p = 0.0034) and 22°C (p = 0.0209).

In experiment 2, we aimed to identify differences among lineages in response to each sublethal temperature treatment to assess whether response patterns differed from those of microscopic sporophytes, and to link physiological responses to gene expression. Most strikingly, all sporophytes of the Spitsbergen selfing bleached at 19 and 20.5°C following day 3. Thus, we analysed differences among lineages within each temperature treatment on day 3 (including the Spitsbergen selfing), and on day 17 (excluding the Spitsbergen selfing at 19 and 20.5°C) to accommodate for the unbalanced design. Relative growth rates of fresh weight (RGR) decreased over time for all lineages at 19 and 20.5°C (Figure 2). We show RGR over the acclimation period (days -2–0) and over 17 days of heat treatment. Each point refers to growth rates over the time period starting from the previous sampling date. Over the first three days at 10°C, growth rates among lineages were not significantly different (p = 0.9521, Table 2) and all achieved > 0.12 g g^–1 d^–1. Until day 3 at 19°C, growth of the Spitsbergen selfing was significantly reduced to 0.06 g g^–1 d^–1 compared to all other lineages at 19°C, which ranged around 0.1 g g^–1 d^–1 (p < 0.0001, Table 2; Tukey tests, p < 0.001). Likewise, at 20.5°C, growth of the Spitsbergen selfing was significantly reduced at < 0.05 g g^–1 d^–1 compared to the remaining lineages (p = 0.0008, Table 2; Tukey tests, p < 0.01), which showed similar growth rates around 0.07 g g^–1 d^–1. Between days 14 and 17, growth rates differed significantly among lineages at 10°C (p = 0.0154, Table 2). H × S sporophytes grew significantly faster at 0.09 g g^–1 d^–1 than S × S at 0.07 g g^–1 d^–1 [Tukey tests, (H × S > S × S) = (S × H = H × H), p < 0.01]. At 19°C, growth rates among the three remaining lineages did not differ significantly (p = 0.3885, Table 2). At 20.5°C, growth rates differed significantly among lineages (p = 0.0296, Table 2), in that H × S retained slightly higher growth at 0.02 g g^–1 d^–1 compared to the Helgoland selfing at 0.01 g g^–1 d^–1 [Tukey tests, (H × S > H × H) = S × H, p < 0.05].

Optimum quantum yield F_v/F_m was relatively stable across lineages and temperatures and was mostly in a healthy range of 0.6–0.7 relative units (Figure 3). On day 3, F_v/F_m differed significantly among lineages at all temperatures. At 10°C, F_v/F_m was at 0.64 in the Spitsbergen selfing and significantly lower than in the other lineages, which ranged from 0.66 to 0.67 (p = 0.0006, Table 3; Tukey tests, p ≤ 0.01). At 19°C, F_v/F_m was highest in the Helgoland selfing at 0.70 and lowest in the Spitsbergen selfing at 0.67, while the reciprocal crosses were intermediate at 0.68 [p = 0.0004, Table 3; Tukey tests, H × H > H × S = (S × H > S × S), p < 0.05]. At 20.5°C, F_v/F_m was at 0.65 in the Spitsbergen selfing and significantly lower than in the Helgoland selfing for which F_v/F_m was 0.70 [p = 0.0414, Table 3; Tukey tests, (H × H > S × S) = (S × H = H × S), p < 0.05]. On day 17, F_v/F_m did not differ significantly among lineages at 10°C (p = 0.0629, Table 3) and ranged between 0.67 and 0.69. At 19°C, F_v/F_m differed significantly among lineages (p = 0.0218, Table 3), in that F_v/F_m was higher in S × H at 0.72 than in H × S and the Helgoland selfing at 0.70 [Tukey tests, S × H > (H × S = H × H), p < 0.05]. At 20.5°C, F_v/F_m ranged between 0.68 and 0.71 and again did not differ significantly among lineages (p = 0.4549, Table 3).

The number of reads per library ranged between 28 million and 44 million with an average of 33.5 million reads per library. The final reference transcriptome based on Phaeophyceae hits contained 34,337 contigs with an average contig size of 928 bp (N50 = 1,230 bp). BUSCO analysis of our reference against the core eukaryotic gene set indicated that 78.22% were present and complete (96.04% complete and partial), with 3.96% missing core genes. The duplication rate was 1.18 orthologs per core gene, indicating a low level of redundancy. A total of 11,519 contigs (33.55%) were annotated using the KEGG database, yielding 3,705 unique KO accessions for which the contig count data were aggregated. Following the removal of lowly expressed KEGG genes, 3,398 were retained to identify differentially expressed genes (DEGs) among the four lineages at 10 and 20.5°C on day 18 of experiment 2.

Out of the 3,398 KEGG genes, we extracted 1,111 (32.7%) DEGs among all comparisons in the differential expression analysis. Multidimensional scaling (MDS, Figure 4) of DEG expression revealed that the replicates of the same lineage and temperature treatment clustered closely. Distances between samples can be interpreted as the root-mean-square average (RMS) of the log₂-fold-changes for all 1,111 DEGs between each pair of samples (Ritchie et al., 2015). Distribution of data along dimension 1 (66.1% of variance explained) clearly separated the two temperature treatments of 10 and 20.5°C, while dimension 2 (12.5% of variance explained) separated samples based on origin of the parental gametophytes. At 10°C, the Helgoland and Spitsbergen selfings each formed clearly separated clusters while the reciprocal crosses grouped together between the selfings. At 20.5°C, the reciprocal crosses separated slightly from each other. The placement of the reciprocal crosses between both selfings verifies that outcrossing was successful for the sequenced sporophytes.

We assessed general heat responses by analysing differential gene expression between 20.5 and 10°C within each lineage (Figures 5A,B). Gene expression changes in response to 20.5°C was lowest in H × S, where 162 genes were up-regulated and 223 genes were down-regulated compared to the 10°C control. In S × H at 20.5°C, 371 genes were up-regulated and 276 genes were down-regulated, which is in a comparable range to the 388 up-regulated and 355 down-regulated genes in the Helgoland selfing at 20.5°C. Among these, 121 DEGs were constitutively up-regulated across all three lineages, and 131 were constitutively down-regulated (Figure 5B).

To identify general gene expression patterns among genotypes of the different locations, we conducted comparisons among all lineages at 10°C (Figures 5C–E). In the Spitsbergen selfing, regulation was strongest as 118 genes were up-regulated and 38 genes were down-regulated compared to the Helgoland selfing (Figure 5C). In contrast, both reciprocal crosses showed reduced gene regulation compared to either selfing, indicating a more generalist response intermediate to the selfings. Compared to the Helgoland selfing at 10°C, H × S up-regulated 9 genes and down-regulated 13 genes at 10°C, whereas S × H up-regulated 5 genes and down-regulated 6 genes at 10°C. When compared to the Spitsbergen selfing at 10°C, H × S up-regulated 11 genes and down-regulated 8 genes, whereas S × H up-regulated 7 genes and down-regulated 5 genes. No genes were differentially expressed among the two reciprocal crosses at 10°C. To investigate potential population-specific regulatory heat responses among the reciprocal crosses, we compared all surviving lineages at 20.5°C (Figures 5F,G). In comparison with the Helgoland selfing, H × S up-regulated 95 genes while 69 genes were down-regulated (Figure 5F). Similarly, in S × H, 72 genes were up-regulated while 70 genes were down-regulated. Among these, both reciprocal crosses shared the up-regulation of 34 KEGG genes and down-regulation of 48 KEGG genes (Figure 5G). In a direct comparison of H × S against S × H at 20.5°C, 7 genes were significantly up-regulated and 2 genes were significantly down-regulated.

In experiment 1, we identified an upper survival temperature of 21–22°C for microscopic L. digitata sporophytes over 14 days, while we note that a very low percentage of sporophytes survived even at 23°C in all but the Arctic inbred lineage (Figure 1). This is concordant with the published limits for macroscopic, cultivated L. digitata sporophytes (Bolton and Lüning, 1982; tom Dieck, 1992; Martins et al., 2019; Franke et al., 2021). In contrast, the macroscopic sporophytes of the Spitsbergen selfing were even more sensitive to temperatures ≥ 19°C than the reciprocal crosses and the Helgoland selfing. A possible explanation for the reduced tolerance of the Spitsbergen L. digitata sporophytes is inbreeding depression through self-fertilization (Raimondi et al., 2004; Camus et al., 2021). In support of this, the bleaching of all sporophytes of the Spitsbergen selfing after 3 days at 19 and 20.5°C occurred only in experiment 2 and not at comparable temperatures in experiment 1, which complies with the theory that inbreeding depression tends to be more pronounced in later life history phases such as growth and reproduction in routinely selfing species like kelp (Husband and Schemske, 1996; Raimondi et al., 2004). In such species, it is plausible that deleterious alleles affecting the earliest life stages have already been purged by selection (Husband and Schemske, 1996; Raimondi et al., 2004; Camus et al., 2021). For instance, in Macrocystis pyrifera, inbreeding correlated with a decrease of sorus area in fertile sporophytes (Raimondi et al., 2004). In comparison, juvenile macroscopic L. digitata sporophytes obtained by fertilization of several different gametophyte isolates from Spitsbergen survived 22°C over two weeks (Franke et al., 2021), and the upper thermal limit of mature sporophytes along L. digitata’s distribution was recently described as uniform (Liesner et al., 2020a). It has been assumed that genotypic diversity correlates with increased response plasticity among individuals and therefore increased population resilience (Wernberg et al., 2018; Liesner et al., 2020b; Alsuwaiyan et al., 2021), but a potential correlation of intra-individual genetic diversity (i.e., heterozygosity) and environmental tolerance remains to be investigated. An alternative explanation for the reduced thermal tolerance of microscopic and macroscopic S × S sporophytes could be contamination with pathogens (e.g., oomycetes; Gachon et al., 2010; Murúa et al., 2020). However, staining of the affected sporophytes and gametophyte stocks using calcofluor white and DAPI yielded no evidence for the presence of eukaryotic pathogens (C. Gachon, pers. comm.; platform ‘‘My Seaweed Looks Weird’’⁴). Regardless of the cause of the reduced thermal tolerance of Spitsbergen sporophytes in our experiment, outbreeding with a distantly related genotype (see Liesner et al., 2020a) restored the tolerance of outcrossed sporophytes.

In our study, Arctic L. digitata showed increased gene regulation compared with the cold-temperate Helgoland material at presumed optimum conditions of 10°C (Figure 5C). In contrast, gene regulation in Saccharina latissima from the cold-temperate population of Roscoff was double that of Arctic Spitsbergen samples following gametogenesis, sporophyte recruitment and growth at 8°C for three months (Monteiro et al., 2019b), which indicates different regulation patterns across species. Alternatively, if we interpret the magnitude of gene expression itself as a stress response associated with metabolic costs (Clarke, 2003; Dekel and Alon, 2005; de Nadal et al., 2011), the pattern of increased gene expression in the Spitsbergen selfing might indicate a general disadvantage of this lineage already under control conditions. Local adaptation to cold temperature seems an unlikely explanation, as the growth optimum of juvenile Arctic L. digitata sporophytes was reported at 15°C (Franke et al., 2021). In Arabidopsis lyrata, inbreeding correlated with generally increased gene expression, whereas specifically regulated functions differed between two tested populations (Menzel et al., 2015). A study on Drosophila melanogaster related increased gene expression in inbred individuals to protective responses compensating for inbreeding effects (García et al., 2013), but there are no comparable studies in kelps as yet.

Especially compared to sporophytes from the Spitsbergen selfing, which potentially suffered from inbreeding depression, a benefit of outbreeding may be quantified by applying the concept of mid-parent heterosis on the relatively high survival of the reciprocal crosses in experiment 1 (Figure 1). In experiment 2, the Helgoland selfing and both reciprocal crosses behaved very similarly apart from small, but significant differences in growth rates and optimum quantum yield F_v/F_m (Figures 2, 3). Heterosis has been shown to occur in offspring of closely related kelp species (tom Dieck and de Oliveira, 1993; Martins et al., 2019) and may manifest in different ways. For instance, kelp hybrids of L. digitata and L. pallida showed mid-parent heterosis in optimal growth at 12°C (Martins et al., 2019), but had a 2–3°C higher thermal tolerance than either inbred parent (best-parent heterosis sensu Hochholdinger and Hoecker, 2007). Within a species, outbreeding across populations may alleviate the negative effects of genetic drift and inbreeding depression (Charlesworth and Charlesworth, 1987; Raimondi et al., 2004). Outbreeding among isolates of Saccharina latissima from Helgoland and a less heat tolerant strain from Nova Scotia, Canada, also resulted in offspring with a heat tolerance comparable to the Helgoland strain (Lüning et al., 1978). Beneficial effects of outbreeding have also been shown for the giant kelp Macrocystis pyrifera. For instance, best-parent heterosis significantly increased growth of several lineages of 10-week-old sporophytes produced by crossing gametophyte isolates collected along 250 km of the Chilean coast (Westermeier et al., 2010). However, the potential for outbreeding among natural populations is likely limited in seaweeds due to their low dispersal capacity (Norton, 1992; Billot et al., 2003; King et al., 2018). Still, crossing of few isolates from the same Arctic population of L. digitata as the isolates used here produced offspring sporophytes with a thermal tolerance of 22°C over two weeks (Franke et al., 2021), which supports the theory that for kelps, it is primarily self-fertilization, the most extreme form of inbreeding, that may dramatically lower fitness (Raimondi et al., 2004).

To investigate how the different lineages maintained a similar phenotype in experiment 2, we examined the genes commonly regulated within all lineages in response to heat stress at 20.5°C compared to 10°C. The functionally annotated genes (Table 4) reflect many known pathways in kelp stress responses (see, e.g., Heinrich et al., 2012b; Salavarría et al., 2018; Li et al., 2019; Monteiro et al., 2019b; Zhang X. et al., 2021). The differential expression of chaperone-related genes such as heat-shock proteins, co-chaperones (GrpE, HscB) and functionally similar proteins (prefoldin alpha, FK506-binding protein) demonstrates that the sporophytes were experiencing stress (e.g., Silberg et al., 1998; Siegert et al., 2000). In contrast, down-regulation of genes with potential heat shock protein or co-chaperone function (DnaJ; Qiu et al., 2006) might be due to a general reduction of HSP expression in longer term temperature acclimation (Heinrich et al., 2015; Monteiro et al., 2019b). The up-regulation of genes coding for two proteins involved in quenching of the reactive oxygen species hydrogen peroxide (H₂O₂) further supports the expression of cellular stress responses (Collén and Davison, 2001; Rhee et al., 2012). Destructive effects of oxidative stress might have also increased the investment in maintenance and repair processes. For instance, an increased turnover of proteins and mitochondria is indicated by the up-regulation of genes involved in ribosome biogenesis (Supplementary Table 1; Kültz, 2005; Bischof and Rautenberger, 2012) and mitochondrial biogenesis (Supplementary Table 1; Zhang et al., 2009).

Gene expression patterns showed a general reduction of metabolic activity related to carbohydrate metabolism in response to 20.5°C (Table 4). The down-regulation of key genes participating in the citric acid cycle (fumarate hydratase; Nast and Müller-Röber, 1996), and glycolysis and gluconeogenesis (PEP carboxykinase, phosphoglycerate kinase, triosephosphate isomerase; Boldt et al., 1992; Gómez and Huovinen, 2012) indicates that both catabolic and anabolic carbohydrate pathways were down-regulated. Investment in cell wall glycan production was reduced (mannan polymerase II complex MNN10 subunit; Jungmann et al., 1999), as was mobilization of the long-term storage compound laminarin (endo-1,3-beta-glucanase; Chesters and Bull, 1963; Martín-Cuadrado et al., 2008). In combination with the reduction of assimilative processes indicated by the down-regulation of carbonic anhydrase (Raven, 1991; Badger and Price, 1994) and genes related to nitrogen uptake (ferredoxin-nitrite reductase, Lomas and Glibert, 2000; nitrate/nitrite transporter, Fukuda et al., 2015) these results support a regulation of the metabolism in favour of cell maintenance rather than growth (Kültz, 2005), as is supported by the reduced growth rates in experiment 2 (Figure 2) and by the reduced expression of several genes related to mitosis (Table 4).

The regulation of genes involved in lipid metabolism indicates that, in addition to reduction of anabolism, cell and plastid membrane properties were altered in response to heat (Table 4). Neutral lipid contents (i.e., storage lipids) in vegetative kelp sporophyte tissue are generally low, but may still be mobilised for metabolic processes in the meristem (Haug and Jensen, 1954; Velimirov, 1979). The up-regulation of acetyl-CoA C-acetyltransferase and two genes involved in CoA synthesis might indicate an increase of fatty acid degradation (but note the metabolic versatility of CoA; Blanco and Blanco, 2017), while lipid synthesis was strongly down-regulated, which further indicates a reduction of anabolism. As polyunsaturated fatty acids (PUFAs) and high temperature increase cell membrane fluidity, decreasing the expression of a lipid desaturase (linoleoyl-CoA desaturase) might maintain cell membrane rigidity in response to stress (Neidleman, 1987; Tatsuzawa et al., 1996; Maulucci et al., 2016). Modifications to sulfolipid (sulfoquinovosyltransferase) and glycolipid metabolism (1,3-propanediol dehydrogenase) indicate the maintenance of photosynthetic membranes in the chloroplasts (Boudière et al., 2014; Zhan et al., 2019).

Regulation of chlorophyll biosynthesis is a known response to several stressors such as hyposalinity, temperature, PAR and UV radiation in the seaweeds Saccharina latissima (Heinrich et al., 2015; Monteiro et al., 2019b) and Desmarestia anceps (Iñiguez et al., 2017). Here, a reduction of chlorophyll biosynthesis was indicated in response to heat by the down-regulation of, among others, genes coding for the key enzymes magnesium chelatase (Pontier et al., 2007; Stenbaek and Jensen, 2010) and protochlorophyllide reductase (Begley and Young, 1989). Whereas growth at sublethal temperature may increase chlorophyll biosynthesis in kelps (Li et al., 2019), the reduction in our study may be interpreted as a heat stress-induced impairment of chlorophyll synthesis (Tewari and Tripathy, 1998). The thermal optimum of photosynthesis is often higher than that of growth (Davison, 1987; Graiff et al., 2015), so the fact that key processes of photosynthesis were down-regulated in our study indicates that this was not a response to increased temperature, but a general stress response, as has been described for plants and kelps (e.g., Mickelbart et al., 2015; Monteiro et al., 2019b). Still, the stability of optimum quantum yield at 20.5°C in this study suggests that regulation of the above genes may have maintained photosynthetic efficiency even under high thermal stress (Figure 3).

KEGG genes linked to a variety of pathways were differentially regulated between the reciprocal crosses at 20.5°C (Figure 5F; Table 5). In combination, these differential regulations may in part be responsible for the subtly increased heat tolerance of only the H × S lineage versus the H × H lineage after 17 days at 20.5°C (Figure 2C). This became evident only under heat stress and not at the control temperature of 10°C (i.e., 0 DEGs among the reciprocal crosses at 10°C; Figure 5C). The increased expression of heat shock protein 20 in H × S (and H × H; Supplementary File 1) compared to S × H at 20.5°C may be related to population-specific responses in HSP expression previously described for HSP70 among British L. digitata populations (King et al., 2019). One of the many functions described for aldehyde dehydrogenase (ALDH) is scavenging of toxic aldehydes produced by interactions of ROS with lipids, proteins and nucleic acids (Sunkar et al., 2003). Reduced expression of ALDH genes in H × S might therefore indicate either low levels of oxidative stress or a high tolerance capacity. In contrast, increased terpene biosynthesis in H × S, as indicated by expression of transcripts coding for hydroxymethylglutaryl-CoA synthase, might indicate defense reactions or increased antioxidant capacity (Fisch et al., 2003; de Oliveira et al., 2017). As protein-serine/threonine kinases are involved in stress-related signaling cascades (Graves et al., 1995; Bischof et al., 2019), the increased expression of related genes in H × S indicates a stronger response to heat stress than in S × H. The increased expression of other KEGG genes in H × S points toward measures of metabolic maintenance. NADH:ubiquinone reductase is involved in cyclic electron transport in photosystem I (Shikanai and Aro, 2016; Zhao et al., 2018), while the expression of RRP5 indicates increased ribosome biogenesis (Missbach et al., 2013). Malate dehydrogenase is a key enzyme in the citric acid cycle (Musrati et al., 1998), the maintenance of which has been shown to be correlated to heat tolerance in land plants (Xu and Huang, 2010). The dynein assembly factor DAW1 has been previously described as sex-biased during fertility of male gametophytes and was linked to flagella production in sperm (Pearson et al., 2019). In H × S at 20.5°C, the higher expression of DAW1 stands in contrast to the reduced expression of DMC1 coding for a meiotic recombination protein, which is reduced under heat stress during fertility in Arabidopsis (Ning et al., 2021). However, we conducted the experiment with juvenile sporophytes which do not produce meiospores, leaving the specific function of these two transcripts unclear.

Female and male gametophytes may contribute differently to sporophyte traits. Martins et al. (2019) suggested a female-specific component to thermal inheritance in kelps, because growth rates of sporophytes from interspecific crosses of L. digitata and the warm-temperate L. pallida in part responded more closely to the female parent at sublethal temperatures. Here, such an effect may explain the subtle differences in physiological and molecular traits between the reciprocal crosses. Sex-biased gene expression has been described in the haploid, sexual gametophytes of the kelp S. latissima (Monteiro et al., 2019a; Pearson et al., 2019). It remains hypothetical whether a female-specific inheritance pattern in diploid, asexual sporophytes may be related to a potential cyto-nuclear compatibility, in which the optimal function of mitochondria and plastids inherited via eggs depends on female-associated nuclear genes.

Our results imply that outbreeding among differentiated kelp lineages and populations may mitigate effects of genetic drift and may increase stress tolerance in comparison to fully inbred lineages. Interestingly, this became already apparent by combining just two clonal lineages from diverged populations. As shown by the subtly increased performance of the H × S lineage at thermal stress, performance during heat stress may even improve by outbreeding with an Arctic population. This adds to the extensive knowledge from terrestrial agriculture and recently also kelp mariculture, that inter-cultivar crosses may produce stable, healthy descendants with superior characteristics compared to their inbred parental lineages (Li et al., 2007; Westermeier et al., 2010; Fu et al., 2014). However, to reduce response variation, produce stable phenotypes and maximise heterosis, homozygous lines should be inbred over several generations before application in mariculture (Robinson et al., 2013), and investigations on the optimal genetic distance should be conducted (e.g., intra- vs. inter-population crosses; Li et al., 2008). Meanwhile, natural kelp populations, including Laminaria digitata, are threatened by ocean warming especially at the warm range edges (Raybaud et al., 2013; Voerman et al., 2013; Wernberg et al., 2016; Assis et al., 2018; Smale et al., 2019). In addition to gradual ocean warming, increasingly intense and frequent marine heatwaves (Hobday et al., 2016; Oliver et al., 2018) may reduce genetic diversity of persistent populations (Coleman et al., 2020a; Gurgel et al., 2020). An ecologically cautious method for conservation of threatened natural populations may therefore be an introduction of intraspecific cultivars from distant populations in an attempt of assisted evolution (Filbee-Dexter and Smajdor, 2019; Coleman et al., 2020b).

An important caveat regarding assisted evolution is the potential for outbreeding depression in crosses of genetically diverged, locally adapted populations. In contrast to inbreeding depression, outbreeding depression describes the reduced performance of offspring of genetically diverged lineages due to the disruption of interactive gene complexes or local adaptation (McKay et al., 2005; Aitken and Whitlock, 2013). As yet, outbreeding depression could not be identified among distant populations of the fucoid seaweed Hormosira banksii (McKenzie and Bellgrove, 2006), and was also not obvious in performance of the reciprocal crosses here. This suggests that genetic divergence among L. digitata populations is low enough to maintain genetic compatibility, despite the strong spatial structuring and thermal gradient which acts along the species’ distribution gradient (Liesner et al., 2020a). Alternatively, outbreeding depression may manifest later in the life cycle especially with respect to reproduction (Schierup and Christiansen, 1996; McKenzie and Bellgrove, 2006). Therefore, careful assessments of the performance and viability of numerous outbred lineages covering a gradient of genetic differentiation are necessary across several generations before considering an application in mariculture or conservation.