Transcriptome profiling

35 days after stratification, rosette tissue of all plants were harvested and flash frozen in liquid nitrogen. Random samples from each replicate experiment for both temperatures were taken for eight accessions to profile the transcriptome with RNA-sequencing. The eight accessions were selected to represent the climatic variation in the full panel (Figure 3—figure supplement 1, Figure 1—source data 1). Total RNA was extracted using the KingFisher Duo Prime System (Thermo Fisher Scientific) together with a high-performance RNA bead isolation kit (Molecular Biology Service, VBC Core Facilities, Vienna). To determine the quantity of RNA, we used Fluorometer Qubit 4 (Invitrogen) and Qubit RNA BR Kit (Invitrogen). For each sample, 1 µg of total RNA was treated with the poly(A) RNA Selection Kit (Lexogen) and eluted in 12 µl of Nuclease-Free Water. Libraries were prepared according to the manufacturer’s protocol in NEBNext Ultra II Directional RNA Library Prep Kit (New England Biolabs) and individually indexed with NEBNext Multiplex Oligos for Illumina (New England Biolabs). The quantity and quality of each amplified library were analyzed by using Fragment Analyzer (Agilent) and HS NGS Fragment Kit (Agilent). Libraries were sequenced with an Illumina HiSeq2500 in paired-end mode with read length of 125 bp. Sequencing was performed by the Next Generation Sequencing Facility at Vienna BioCenter Core Facilities (VBCF), member of the Vienna BioCenter (VBC), Austria. Samples were distributed over four independent libraries. This was due to failed samples that needed replacement. Detailed info on which samples belong to which library is listed in SRA (https://www.ncbi.nlm.nih.gov/sra/PRJNA807069). Gene expression was quantified by using quasi-mapping in salmon (1.2.1; Patro et al., 2017). The salmon indices were built separately for each accession as we incorporated the SNP variation from the 1001 Genomes Consortium, 2016 into the reference transcriptome. The heatmap was built using pheatmap (1.0.12; Kolde, 2019) in R (4.0.3; R Development Core Team, 2017). The clustering was done by complete clustering on Euclidean distances for both rows and columns. The gene clusters were defined by cutting the dendrogram in seven groups. With a -test, we tested for overrepresentation of a temperature effect on the expression of the 251 selected cold-acclimation genes (df=1) compared to the remaining 18,784 background genes. We used the chisq.test function in R (4.0.3; R Development Core Team, 2017). Differential expression analysis was conducted with the DESeq2 package (1.30.0; Love et al., 2014) in R (4.0.3; R Development Core Team, 2017). A full model was used, with expression depending on replicate+accession+temperature+replicate:temperature+accession:temperature , after which significance of each model coefficient was defined with a negative binomial Wald test. Differential expression for each accession was then extracted by specifying the respective contrasts using the lfcShrink function in DESeq2 with the adaptive shrinkage estimator (Stephens, 2017). Genes were considered differentially expressed when the adjusted p-value was <0.05.