Page 61 - scheppingen

CODING AND SMALL NON-CODING TRANSCRIPTIONAL LANDSCAPE OF TSC
Of the 269 genes over-expressed in the TSC cortical tubers, 23 were specific to microg- lia, 3 to oligodendrocytes, 5 to neurons, and 8 to astrocytes (Supplementary Table 1). Amongst the 169 genes under-expressed in the TSC cortical tubers, 6 were specific to neurons, 2 to astrocytes, 1 to microglia and 1 to oligodendrocytes (Supplementary Table 1). A Fisher’s exact test revealed that amongst the significantly over-expressed genes there was significant enrichment for microglia specific (p-value<2.2e-16) and astrocyte specific (p-value<0.002) genes, amongst the significantly under-expressed genes there was a suggestive enrichment of neuron specific genes (p-value<0.05). The 32 genes spe- cific to microglia and astrocytes that were over-expressed in the TSC subjects included the complement system related genes, C1QA, C1QB, C1QC and C4B (Figure 1c). We did not observe gene expression differences between important subgroups, that is individu- als with TSC2 versus TSC1 mutations, or mild versus severe intellectual disability.
The small non-coding RNA landscape of tuberous sclerosis complex brain tissue
In order to further explore the brain transcriptome of TSC subjects relative to control subjects we performed small RNA-seq analysis on the same set of RNA samples (Table 1). Each sequencing run produced ~9 million paired-end reads for each sample. After quality assessment and filtering, ~5 million paired-end reads remained for each sample, of which ~82% were mapped to the reference genome (GRCh38). Differential expression analysis of the aligned small RNA transcripts revealed a total of 991 significantly altered tran- scripts, 59 were elevated and 932 were decreased in TSC cortical tubers compared to controls (Fig. 2a). The differentially expressed small RNAs were not only miRNAs but also other classes (Fig. 2b). The largest class of altered small non-coding RNA was the small nuclear RNA (snRNA). Other classes of altered small RNAs in TSC relative to control patients (in decreasing order) were the C/D box small nucleolar RNAs (snoRNAs), miR- NAs, H/ACA box snoRNAs, orphan snoRNA and the small Cajal body RNAs (scaRNAs). Interestingly, the majority of snRNAs, snoRNAs and scaRNAs were under-expressed in TSC cortical tubers compared to control cortex (Fig. 2b). Highly expressed miRNAs in TSC subjects included, miR34a (3.1-fold), miR34b (2.6-fold), miR34c (2.5-fold), miR302a (2.2-fold), miR577 (4-fold) and miR21 (2.9-fold) (Fig. 2c), all members of the miR34 family were validated using RT-qPCR (data not shown).
Previously reports of age dependent miRNA expression patterns in the brain and cardiac tissue46, 47, notably of miR34a48, coupled with the variability of age in our study cohort motivated us to evaluate the association of age to expression patterns of miR34a and the other members of the miR34 family members. We found no significant correla- tion (Pearson’s correlation) between expression patterns of miR34 family members and age (r<0.41, p-value>0.05, data not shown).
Gene co-expression network modules and miRNA targets
To better understand the organization of the protein coding and small non-coding RNA transcriptome in TSC and control subject brain tissue we applied an unsuper- vised weighted gene co-expression network approach (WGCNA)49, 50. On the basis of a Spearman’s correlation matrix, a weighted network of RNA transcripts was constructed that ensured scale-free topology (see Methods). Unsupervised hierarchical clustering uncovered 21 modules of highly inter-correlating transcripts, each harboring more than
59
three