Results
On average, ~12.5 million raw Illumina reads were obtained per sample. After quality trimming and removal of rRNA and algal endosymbiont components, an average of ~6.2 million paired reads were retained per sample. The transcriptome of purebred A. loripes contained ~291 k transcripts, and ~59 k transcripts were left after only retaining the longest isoforms and removal of small transcripts < 400 bp. See Table S3 for details of other transcriptomes used for preliminary analysis and evaluating treatment effect. For a total of ~35 k transcripts a match of coral origin was found in the NCBI nr database. Following the removal of duplicates and transcripts that consistently had zero or very low counts, 8800 transcripts were retained and used for downstream analyses.
Transcriptome-wide gene expression of the hybrids was similar to that of their maternal purebreds, yet distinct from their paternal purebreds and the reciprocal hybrids (Figures 1-3, S3). Principal component analyses (PCA) showed similar expression patterns of the hybrid LT with its maternal purebred LL under both ambient and elevated conditions (Figure 1). The only exception was one LL purebred sample which showed separation with the others in principle component two (Figure 1). Gene expression of the reciprocal hybrid TL also clustered with its maternal purebred TT (but note that n = 1 for TT), and was separated with hybrid LT and its paternal purebred LL under both treatment conditions (Figure 1). The four samples excluded from the main analyses due to small library sizes also supported the existence of maternal effects (Figure S4-S5). The amount of total RNA and cDNA input, as well as the number of raw reads of the samples showed no specific patterns in the PCA plots, suggesting that the observed maternal patterns of the offspring groups were not driven by these factors (Figure S4). Within an offspring group, gene expression did not differ between ambient and elevated conditions (Figure 1). Maternal patterns were not observed in the PCA plot and heatmap generated using mitochondrial genes only (Figure S6-S7).
Differential expression analysis resulted in only 40 DEGs between the maternal purebred LL and its hybrid LT (Figure 2). In contrast, almost 2000 DEGs were identified between the paternal purebred LL and its hybrid TL, as well as between the reciprocal hybrids LT and TL (Figure 2). Among these ~2000 DEGs, the hybrid LT and its maternal purebred LL shared 1343 genes that were differentially expressed from the hybrid TL (Figure 2). Maternal effects in gene expression were also evident in the heatmap of the 500 most variable genes across samples (Figure 3). The only exception was one purebred LL sample which clustered away from the other LL samples, and this was the same sample that showed separation in the PCA plot (Figure 1, 3). The heatmap that includes the four samples removed from the main analyses due to small library sizes confirmed the maternal effects observed in the smaller subset of samples (Figure S5).
Among the DEGs with the highest log-fold change (i.e., four DEGs for paternal purebred LL compared to its hybrid TL, and seven DEGs for hybrid LT compared to hybrid TL with LFC > 5), three were shared genes between the two pairs of comparison (Figure S3). Unfortunately, most of these DEGs were annotated as uncharacterized proteins and hence their potential functions were unknown (Table S4). Only one differentially expressed mitochondrial gene (TRINITY_DN76286_c6_g1_i1) was identified between hybrid TL and its paternal purebred LL (padj = 0.03). No differentially expressed mitochondrial genes were found in all other pairs of comparison.
For gene ontology (GO) analyses using GOseq, GO category “cytosol” (GO: 0005829) was underrepresented in both the comparisons between the paternal purebred LL with its hybrid TL and between the reciprocal hybrids LT and TL, with 90 and 96 DEGs respectively in this category (Table S5). Note that “cytosol” is a very broad GO category and it was comprised of 620 genes in this dataset. In addition, the GO category “membrane” (GO: 0016020) was also underrepresented in the comparison between the paternal purebred LL and its hybrid TL (Table S4). This was also a broad GO category with 255 genes in this dataset, 27 of which were DEGs. In contrast, GO analyses using the MWU test showed no significant GO category was over- or under-represented. However, note that the MWU test omits GO categories that are too broad (i.e., a GO category that contains a large proportion of the total number of genes). For this reason, it was unsurprising that the very broad GO categories “cytosol” and “membrane” that were identified as underrepresented using GOseq were not significant here.
For offspring groups that had different maternal parent species (i.e., between the hybrid TL and its paternal purebred LL, and between the reciprocal hybrids LT and TL), 84-88 DEGs were identified in GO categories with functions connected to the mitochondrion (Table S6). In contrast, no DEGs were found in GO categories linked to the mitochondrion when the offspring groups shared the same maternal parent species (i.e., between the hybrid LT and its maternal purebred LL). The proportion of DEGs over total number of genes was similar between genes in GO categories linked to the mitochondrion and genes in all GO categories (14.9-17.6%, Table S6).