Coping with cold: An integrative, multitissue analysis of the transcriptome of a poikilothermic vertebrate

Abstract

How do organisms respond adaptively to environmental stress? Although some gene-specific responses have been explored, others remain to be identified, and there is a very poor understanding of the system-wide integration of response, particularly in complex, multitissue animals. Here, we adopt a transcript screening approach to explore the mechanisms underpinning a major, whole-body phenotypic transition in a vertebrate animal that naturally experiences extreme environmental stress. Carp were exposed to increasing levels of cold, and responses across seven tissues were assessed by using a microarray composed of 13,440 cDNA probes. A large set of unique cDNAs (approximately 3,400) were affected by cold. These cDNAs included an expression signature common to all tissues of 252 up-regulated genes involved in RNA processing, translation initiation, mitochondrial metabolism, proteasomal function, and modification of higher-order structures of lipid membranes and chromosomes. Also identified were large numbers of transcripts with highly tissue-specific patterns of regulation. By unbiased profiling of gene ontologies, we have identified the distinctive functional features of each tissue's response and integrate them into a comprehensive view of the whole-body transition from one strongly adaptive phenotype to another. This approach revealed an expression signature suggestive of atrophy in cooled skeletal muscle. This environmental genomics approach by using a well studied but nongenomic species has identified a range of candidate genes endowing thermotolerance and reveals a previously unrecognized scale and complexity of responses that impacts at the level of cellular and tissue function.

Fig. 1.

Analysis of cold-induced gene expression. (a) Schematic diagram showing the cooling time course and sampling regime used. Warm-acclimated control fish (30°C) were sampled at three separate time points and compared with fish sampled more frequently over a 3-week time course of cooling to either 23°C, 17°C, or 10°C. (b) PCA of cold-induced tissue-expression profiles showing clear separation of the cooled and warm-acclimated samples. For PCA, the expression profile of each gene was summarized by two centroids, representing the average expression of each cDNA in the cooled fish compared with the warm-acclimated controls (arbitrarily set to 0). PCA used the entire set of cDNAs printed on the carp array and the axes represent the combinations of genes that explain most of the expression changes affected by cooling.

Fig. 2.

Identification of a common set of cold-regulated genes. cDNAs were grouped according to the GO biological processes in which they participate. (a) Nucleic acid processing (40 genes). (b) Transport (37 genes). (c) Protein catabolism (35 genes). (d) Cell stress or molecular chaperones (21 genes). (e) Metabolism (18 genes). (f) Signaling (13 genes). (g) Cell structure (12 genes). (h) Only eight cDNAs were globally repressed by cold. Representative or notable genes found in each class are indicated. The expression of each cDNA is presented as the ratio of transcript abundance relative to its mean abundance in the control warm-acclimated samples. Each row represents a different cDNA, and each column represents the expression of the corresponding transcript in the different tissues in the control animals (left columns) or after cooling to 23°C, 17°C, and 10°C on the basis of time (right columns). For kidney, only the 17°C cooling trajectory was collected. Red indicates a relative increase in transcript abundance with cooling, and green represents a decrease. (i) mRNA levels in gill of CIRBP, a representative gene in the common response showing that increased cooling resulted in a greater induction of transcript. (j), The cold-induced expression profile of putative yeast orthologues (1) to the induced genes in the carp common response.

Fig. 3.

Variation in the expression of 1,728 cDNAs in tissues of cooled fish. The expression of each cDNA is presented as the ratio of transcript abundance in each cooled sample relative to its mean abundance in the control warm-acclimated samples. (a) Cluster diagram showing the 23 K-means clusters of cDNAs that exhibited similar expression profiles. (b) GO-Matrix chart, a pseudocolor map of the significant over- or underrepresentation (red or blue, respectively) of genes within the common response and each of the K-means clusters. Saturated colors represent P values of <0.05. Detailed views of selected clusters are as follows: liver and intestinal mucosa-induced cDNAs (c), muscle-repressed transcripts (d), muscle-induced transcripts (e), brain-induced transcripts (f), and cDNAs induced in kidney (g). Representative or notable genes found in these clusters are indicated.