Genetics of gene expression responses to temperature stress in a sea urchin gene network

Abstract

Stress responses play an important role in shaping species distributions and robustness to climate change. We investigated how stress responses alter the contribution of additive genetic variation to gene expression during development of the purple sea urchin, Strongylocentrotus purpuratus, under increased temperatures that model realistic climate change scenarios. We first measured gene expression responses in the embryos by RNA-seq to characterize molecular signatures of mild, chronic temperature stress in an unbiased manner. We found that an increase from 12 to 18 °C caused widespread alterations in gene expression including in genes involved in protein folding, RNA processing and development. To understand the quantitative genetic architecture of this response, we then focused on a well-characterized gene network involved in endomesoderm and ectoderm specification. Using a breeding design with wild-caught individuals, we measured genetic and gene-environment interaction effects on 72 genes within this network. We found genetic or maternal effects in 33 of these genes and that the genetic effects were correlated in the network. Fourteen network genes also responded to higher temperatures, but we found no significant genotype-environment interactions in any of the genes. This absence may be owing to an effective buffering of the temperature perturbations within the network. In support of this hypothesis, perturbations to regulatory genes did not affect the expression of the genes that they regulate. Together, these results provide novel insights into the relationship between environmental change and developmental evolution and suggest that climate change may not expose large amounts of cryptic genetic variation to selection in this species.

Fig. 1

Growth temperature subtly alters embryonic development. (a) Gastrula-stage Strongylocentrotus purpuratus embryo. Cultures were sampled when the invaginating gut (archenteron) reached ~1/3 through the central region (blastocoel). Bars show measurements taken on voucher embryos: embryo length 1–3; embryo stage: ratio of 1–2:1–3. (b) Embryonic developmental rates increased greatly at higher temperatures, and caused subtle changes in embryonic morphology between families. The top row of numbers shows the time to gastrulation at each temperature. Despite these changes, embryos appeared generally healthy at all three temperatures. Example embryos sampled from each of three cultures at each of the three experimental temperatures are shown. (a, b) All scale bars are 50 lm. (c) Boxplots of embryo length. 18 and 15 °C embryos were shorter along the animal-vegetal axis than 12 °C embryos. (d) 18 °C cultures increased in both Hsp70 and Hsp90 expression, relative to 12 °C cultures. Expression is log₂FC from 12 to 18 °C relative to RBM8A. Error bars show two standard errors estimated from the model: log₂(exp)_ij = Temp_i + Female_j + e_ij.

Fig. 2

The expression of many genes was affected by higher temperatures. (a) M-A plot showing the average 12 °C expression level (in Reads per Kilobase per Million mapped reads, RPKM) of all 14 420 genes with an average of 10 RNA-seq counts per sample, and the log₂ fold change of each gene from 12 to 18 °C. Genes with a significant response at a False discovery rates (FDR) of 0.05 are coloured red. The horizontal blue lines show a ± 50% change in expression (log₂FC = ±0.58). The figure is truncated at RPKM −5.8 and log₂FC 6 for clarity, hiding 26 genes. (b) Genes that decreased at 18 °C relative to 12 °C tended to be high expressed at 12 °C. Kernel density curves of the expression (RPKM) those genes that increased (orange) decreased (green), or did not change (grey) at 18 °C relative to 12 °C at a FDR of 0.05.

Fig. 3

Estimates and credible intervals for the gene expression reaction norms to temperature. The magnitude in log₂ fold change (log₂FC) of the expression change from 12 to 15 °C (dotted lines) or 18 °C (red) for each of the 72 target genes. Bars cover the central 95% of posterior samples for each gene. Estimates are averaged across the two experimental replicates. The 15 °C response was generally similar to, but smaller than, the 18 °C response. 18 °C responses for gene names coloured black were significantly nonzero.

Fig. 4

Endomesoderm and ectoderm gene network response to higher temperatures. The sea urchin endomesoderm gene regulatory network controls early cell fate specification and morphological patterning in embryogenesis. This representation of the network is adapted from the BioTapestry model from the Davidson lab website: (http://www.its.caltech.edu/~mirsky/). The network becomes subdivided during embryogenesis into three major territories—the skeletogenic, the endomesodermal and ectodermal cells. All 72 network genes assayed are displayed, including the known regulatory relationships at gastrulation. Genes are placed in a territory where they are expressed at 27 h postfertilization (hpf) at 15 °C, corresponding to the sampling time in this study according to Bio-Tapestry. Arrows show positive regulatory events that are active at 27 hpf, and lines ending in bars show repressive regulatory events. All ectoderm regulatory events shown here (http://www.its.caltech.edu/~mirsky/) are assumed active. Genes are coloured according to the magnitude of response to 18 °C temperatures (more red = larger increase, more blue = larger decrease). Lines are coloured according to the product of the temperature responses of the two connected genes. Higher scores (more green for positive regulation and more purple for repression) indicate a greater similarity between the temperature response of a regulatory gene and the response of the gene it regulates. The mean correlation of interacting genes (r = 0.04, corrected for repressive regulators by multiplying by −1) was well within the range based on randomizing the temperature responses among genes, but preserving the topology of the network (P = 0.30, by randomization).

Fig. 5

Genetic effects tended to be larger than temperature effects, and female effects were larger than male effects among network genes. (a) The square root of additive genetic variation (estimated as four times the fitted male variance) and the absolute value of the difference in expression between 12 and 18 °C are plotted for each of the 72 genes. (b) The fitted female effect and male effect variance for each gene. (a, b) The diagonal line in each plot shows y = x.

Fig. 6

Male effect correlations were greater among directly interacting genes. Network diagram is as in Fig. 4. Here, genes are coloured according to the magnitude of the male effect variance, and edges are coloured according to the correlation between the male breeding values for the upstream and downstream genes. Overall, male effect correlations (corrected for repressive regulators by multiplying by −1) among interacting genes were higher than expected by chance (P = 0.0045).