Dynamic RNA Regulation in the Brain Underlies Physiological Plasticity in a Hibernating Mammal

Hibernation is a physiological and behavioral phenotype that minimizes energy expenditure. Hibernators cycle between profound depression and rapid hyperactivation of multiple physiological processes, challenging our concept of mammalian homeostasis. How the hibernator orchestrates and survives these extremes while maintaining cell to organismal viability is unknown. Here, we enhance the genome integrity and annotation of a model hibernator, the 13-lined ground squirrel. Our new assembly brings this genome to near chromosome-level contiguity and adds thousands of previously unannotated genes. These new genomic resources were used to identify 6,505 hibernation-related, differentially-expressed and processed transcripts using RNA-seq data from three brain regions in animals whose physiological status was precisely defined using body temperature telemetry. A software tool, squirrelBox, was developed to foster further data analyses and visualization. SquirrelBox includes a comprehensive toolset for rapid visualization of gene level and cluster group dynamics, sequence scanning of k-mer and domains, and interactive exploration of gene lists. Using these new tools and data, we deconvolute seasonal from temperature-dependent effects on the brain transcriptome during hibernation for the first time, highlighting the importance of carefully timed samples for studies of differential gene expression in hibernation. The identified genes include a regulatory network of RNA binding proteins that are dynamic in hibernation along with the composition of the RNA pool. In addition to passive effects of temperature, we provide evidence for regulated transcription and RNA turnover during hibernation. Significant alternative splicing, largely temperature dependent, also occurs during hibernation. These findings form a crucial first step and provide a roadmap for future work toward defining novel mechanisms of tissue protection and metabolic depression that may 1 day be applied toward improving human health.

Numbers report number of genes with reference to first of the two states (Figures 1, in each pair compared. DE, differentially expressed; q<0.001; Max FC, maximum fold change; GeneID unknown, no homolog or likely homolog to an identified human gene, although, as indicated in the text, the homologous genomic region may be apparent; NA, not applicable. Relates to Supplementary Material Tables 2-4 and Data Sheets 2-4.

Forebrain
SAvsSpD IBAvsSA IBAvsEnt EntvsLT  ArvsLT  IBAvsAr IBAvsSpD  total DE  2  411  790  31  12  1539  414  # increased  2  229  539  26  4  1026  224  # decreased  0  182  251  5  8  513  190  #≥2x increased  0  4 Figure 7A; dots represent DE genes in the pairwise comparison indicated below, each assigned to the state where it was increased; boxes delineate the interquartile range, horizontal lines mark the group median, and whiskers indicate the boundaries beyond which sample dots lying outside are outliers. Numbers above are the Wilcoxon test p values and horizontal gray lines represent median GC content of all detected genes in that region. B) AREscore for torpor-arousal cycle transitions, calculated using the web interface at http://arescore.dkfz.de/arescore.pl, to predict the destabilizing strength of type-II AU-rich elements for all 3'UTR sequences of coding genes with CDS length >200nt and 3'UTR length >200nt; plot and numbers are as described for panel A. C) Dots represent AREscores for genes in the indicated co-expression cluster. Horizontal gray line denotes the cutoff score of 8 used for the bias calculations in Figure  7B.  Figure 7C. E) Differential expression of mRNAs encoding ARE binding proteins (partially presented in Figure 7D).
Supplementary Figure 11. Alternative splicing events in hibernation. A) SRSF6, temperaturedependent intron retention; B) Pip4k2c, temperature-dependent exon skipping; C) Cttnbp2, temperature-dependent alternative 5' splice site; D) ACSM1_like, seasonal intron retention/exon skipping; and E) Sec31a, IBA-specific intron retention. Stacked bar graphs on the right show the relative abundance of each color-coded junction in each group. Splicegraphs are from MAJIQ analysis (Vaquero-Garcia et al., 2016) of hypothalamus RNA-seq data, but most of these events are common among all three brain regions. Mean alternative splicing patterns for all significant LSVs in three brain regions. dPSI > 0 indicates increased alternative splicing relative to SA (alternative junction becomes more common), while dPSI < 0 indicates decreased alternative splicing relative to to SA. Error bars show standard deviation. Venn diagrams quantify common and unique events among Forebrain (Fb), Medulla (Med) and Hypothalamus (Hy).
Supplementary Figure 13. Genes with alternative splice sites in forebrain. Heatmap of alternatively spliced LSVs clustered by pattern, each row represents one gene. For each LSV, the abundance of all junctions except the one most commonly observed in SA is plotted relative to SA. dPSI > 0 is increased alternative splicing relative to SA (alternative junction becomes more common), while dPSI < 0 indicates decreased alternative splicing relative to SA. Genes with multiple significant LSVs have numeric suffixes appended to their gene name following a colon.
Supplementary Figure 14. Genes with alternative splice sites in hypothalamus. Heatmap of alternatively spliced LSVs clustered by pattern, each row represents one gene. For each LSV, the abundance of all junctions except the one most commonly observed in SA is plotted relative to SA. dPSI > 0 is increased alternative splicing relative to SA (alternative junction becomes more common), while dPSI < 0 indicates decreased alternative splicing relative to SA. Genes with multiple significant LSVs have numeric suffixes appended to their gene name following a colon.
Supplementary Figure 15. Genes with alternative splice sites in medulla. Heatmap of alternatively spliced LSVs clustered by pattern, each row represents one gene. For each LSV, the abundance of all junctions except the one most commonly observed in SA is plotted relative to SA. dPSI > 0 is increased alternative splicing relative to SA (alternative junction becomes more common), while dPSI < 0 indicates decreased alternative splicing relative to SA. Genes with multiple significant LSVs have numeric suffixes appended to their gene name following a colon. Figure 16. Splice site strength of differentially retained introns in warm vs. cold states. Sequences surrounding three groups of introns were analyzed for splice site strength: A) H-bond score; B) 5'-maximum entropy score; and C) 3'-maxiumum entropy score. "ctrl" denotes all non-overlapping, non-differentially retained introns described by MAJIQ; "up_in_cold" denotes increased intron retention in the cold states vs. warm in at least one brain region, "down_in_cold" denotes introns less likely to be retained in the cold. Higher scores for both maxEnt (for 5' splice sites and 3' splice sites) and HBond (for 5' splice sites only) imply stronger and more canonical splice sites. Numbers indicate the p values from Wilcoxon test of control vs. labelled group. Mean intron retention patterns in three brain regions. dPSI > 0 indicates increased retention of the intron relative to SA, while dPSI < 0 indicates increased excision of the intron relative to SA. Error bars indicate standard deviation. Venn diagrams to the right of each cluster quantify common and unique events among Venn diagrams quantify common and unique events among Forebrain (Fb), Medulla (Med) and Hypothalamus (Hy).