Molecular basis of canalization in an ascidian species complex adapted to different thermal conditions

Canalization is a result of intrinsic developmental buffering that ensures phenotypic robustness under genetic variation and environmental perturbation. As a consequence, animal phenotypes are remarkably consistent within a species under a wide range of conditions, a property that seems contradictory to evolutionary change. Study of laboratory model species has uncovered several possible canalization mechanisms, however, we still do not understand how the level of buffering is controlled in natural populations. We exploit wild populations of the marine chordate Ciona intestinalis to show that levels of buffering are maternally inherited. Comparative transcriptomics show expression levels of genes encoding canonical chaperones such as Hsp70 and Hsp90 do not correlate with buffering. However the expression of genes encoding endoplasmic reticulum (ER) chaperones does correlate. We also show that ER chaperone genes are widely conserved amongst animals. Contrary to previous beliefs that expression level of Heat Shock Proteins (HSPs) can be used as a measurement of buffering levels, we propose that ER associated chaperones comprise a cellular basis for canalization. ER chaperones have been neglected by the fields of development, evolution and ecology, but their study will enhance understanding of both our evolutionary past and the impact of global environmental change.


Supplementary Table S1. Complete list of gene models found in the transcriptome analysis. (see separate PDF file)
Supplementary Table S2. The complete list of chaperone and other molecules found in the transcriptome analysis. Raw counts ('B count' and 'A count') were normalized by the RPKM method 46 . Those genes validated by qPCR are highlighted in blue. We did not evaluate genes with less than 10 counts in either species, and others were not validated as we were not able to develop effective qPCR primers. Gene models derive from the KH gene model set 15 . Note some genes are represented by more than one gene model, reflecting splice variants and/or uncertainty in the modeling process.  (Fig. 2). P-value below the Bonferroni-corrected significance threshold of 0.0045 are shown in green, and P < 0.01 is considered as marginally significant and shown in blue. *Only whole genome contig and/or incomplete gene prediction available. Orthology of hit verified by reciprocal BLAST.
Supplementary Fig. 1. Comparison of expression levels of dnajc3. a, b, show results from primer set 1, whereas c, d show primer set 2. a, c, Quantification of dnajc3 expression levels by qPCR in control (C) or heat shocked (H: 27ºC for 1h at neurula stage as in Fig. 1a) embryos. Heat shock did not significantly affect the expression of dnajc3, however expression in type A (AA) was significantly higher than in type B (BB) (P = 0.002). Reciprocal hybridization (AB/BA) showed strong correlation to the maternal genotype (P < 0.0001). Significances are shown by asterisks: **P <0.01; ***P <0.001. b, d, Correlation between the amount of dnajc3 transcripts and Normal proportion after heat shock in individual crosses. Analysis of Variance (ANOVA) rejected the null hypothesis and supported linear correlation of the amount of dnajc3 transcripts and normal proportion (P < 0.04 for primer sets 1 and P < 0.002 for primer sets 2). Heat shock did not significantly affect the expression of hspa83 (P = 0.013), but expression in type A (AA) was significantly higher than in type B (BB) (P = 0.0081). However, reciprocal hybridization (AB/BA) did not suggest this was linked to the maternal genotype (P > 0.85). Significances are shown by asterisks: ** P <0.01. b, d, Correlation between the amount of hspa83 transcripts and normal proportion after heat shock in individual crosses. Analysis of Variance (ANOVA) did not support linear correlation of the amount of hsp83 transcripts and Normal proportion (P < 0.008 for primer set 1; for primer set 2 P > 0.25).
Supplementary Fig. 4. Comparison of expression levels of hsp60. a, b, show results from primer set 1, whereas c, d show primer set 2. a, c, Quantification of hsp60 expression levels by qPCR in control (C) or heat shocked (H: 27ºC for 1h at neurula stage as in Fig. 1a) embryos. Heat shock significantly affected the expression of hsp60 only in type B (BB). Comparisons of expression in type A (AA) and in type B (BB) showed significant differences under control condition (P = 0.001). However, reciprocal hybridization (AB/BA) did not suggest this was linked to the maternal genotype (P = 0.2526). Significances are shown by asterisks: ** P <0.01; ***P <0.001. b, d, Correlation between the amount of hsp60 transcripts and Normal proportion after heat shock in individual crosses. Analysis of Variance (ANOVA) did not support linear correlation of the amount of hsp60 transcripts and normal proportion (P > 0.42 for both primer sets 1 and 2).
Supplementary Fig. 9. Comparison of expression levels of dnajb9. a, b, show results from primer set 1, whereas c, d show primer set 2. a, c, Quantification of dnajb9 expression levels by qPCR in control (C) or heat shocked (H: 27ºC for 1h at neurula stage as in Fig. 1a) embryos. Heat shock did not significantly affect the expression of dnajb9 (P < 0.4), and expression in type A (AA) did not show significant difference from expression in type B (BB). However, reciprocal hybridization (AB/BA) suggests this was linked to the maternal genotype (P = 0.0061). b, d, Correlation between the amount of dnajb9 transcripts and Normal proportion after heat shock in individual crosses. Analysis of Variance (ANOVA) did not support linear correlation of the amount of dnajb9 transcripts and normal proportion (P = 0.554 for primer set 1, and P = 0.335 for primer set 2).
Supplementary Fig. 10. Comparison of expression levels of hspa9b and hspa8. a, b, show results from hspa9b, whereas c, d show hspa8. a, c, Quantification of each gene expression levels by qPCR in control (C) or heat shocked (H: 27ºC for 1h at neurula stage as in Fig. 1a) embryos. Expression in type A (AA) did not show significant difference from expression in type B (BB) in both genes (for hspa9b P = 0.0366, for hspa8 P = 0.0882). Heat shock significantly affected the expression of hspa8 (P = 0.0021) but not in hspa9b (P = 0.048). Reciprocal hybridization (AB/BA) marginally supported this was linked to the maternal genotype in hspa9b (P = 0.0066) but not for hspa8 (P = 0.8298). Significances are shown by asterisks: **P <0.01. b, d, Correlation between the amount of each gene transcripts and Normal proportion after heat shock in individual crosses. Analysis of Variance (ANOVA) did not support linear correlation of the amount of transcripts and normal proportion in neither of these genes (P=0.1575 for hspa9b and P=0.2384 for hspa8).