Comparative meta-analysis of transcriptomic studies in spinal muscular atrophy: comparison between tissues and mouse models

Abstract

Background: Spinal Muscular Atrophy (SMA), a neuromuscular disorder that leads to weakness in the muscles due to degeneration of motor neurons. Mutations in the survival motor neuron 1 (SMN1) gene leads to the deficiency of SMN protein that causes SMA. The molecular alterations associated with SMA extends across the transcriptome and proteome. Although several studies have examined the transcriptomic profile of SMA, the difference in experimental settings across these studies highlight the need for a comparative meta-analysis to better understand these differences.

Methods and data: We conducted a systematic comparative meta-analysis of publicly available gene expression data from six selected studies to elucidate variations in the transcriptomic landscape across different experimental conditions, including tissue types and mouse models. We used both microarray and RNA-seq datasets, retrieved from Gene Expression Omnibus (GEO) and ArrayExpress (AE). Methods included normalization, differential expression analysis, gene-set enrichment analysis (GSEA), network reconstruction and co-expression analysis.

Results: Differential expression analysis revealed varying numbers of differentially expressed genes ranging between zero and 1,655 across the selected studies. Notably, the Metallothionein gene Mt2 was common in several of the eight comparisons. This highlights its role in oxidative stress and detoxification. Additionally, genes such as Hspb1, St14 and Sult1a1 were among the top ten differentially expressed genes in more than one comparison. The Snrpa1 gene, involved in pre-mRNA splicing, was upregulated in the spinal cord and has a strong correlation with other differentially expressed genes from other comparisons in our network reconstruction analysis. Gene-set enrichment analysis identified significant GO terms such as contractile fibers and myosin complexes in more than one comparison which highlights its significant role in SMA.

Conclusions: Our comparative meta-analysis identified only few genes and pathways that were consistently dysregulated in SMA across different tissues and experimental settings. Conversely, many genes and pathways appeared to play a tissue-specific role in SMA. In comparison with the original studies, reproducibility was rather weak.

Keywords: Differential expression analysis; Meta-analysis; Network reconstruction; Spinal muscular atrophy; Transcriptomics; WGCNA.

Fig. 1

Upset plots [12] of common DEGs among all the group comparisons based on either the selection criterion with an FDR-adjusted p-values < 0.05 and an absolute logFC > 2 (left) or on the criterion of being among the top500 (right). Only intersections among at least three comparisons are shown

Fig. 2

Upset plot (left) showing the number of GO terms related to biological processes (BP) that occur among the top 100 in at least four of the eight comparisons, and corresponding table (right) listing the 14 GO terms highlighted in the upset plot. The comparison number in the table corresponds to the order (from top to bottom) in the upset plot

Fig. 3

Networks constructed using the expression profiles from the GSE10224 data, SMA versus control, for the top 5 DEGs from all the datasets that were merged creating a gene set for each study group, where the top plot corresponds to SMA group and the plot at the bottom corresponds to control group. Nodes represents the genes and edges the interaction between them. The size and colour of the node shows the interaction size of the genes. The thickness of edge represents the correlation between the genes

Fig. 4

Co-expression modules and Topological Overlap Matrix (TOM) Plot associated with SMA (a, b) and Control (c, d) group of data set GSE102204. Plot a & c) WGCNA dendogram where the coloured bars towards the x-axis represent the module colours and y-axis represents the similarity between genes based on the height of the branches in the dendogram. Higher the value, greater the similarity between genes. SMA group has 9 modules and Control group has 7 modules excluding the grey module which correspond to the less correlated gene group. Plot b & d) TOM plot showing the topological overlap of the genes from data set GSE102204. Rows and columns represent the genes, darker the red colour higher the correlation between the genes. Their corresponding gene dendogram are towards the top and along the left of the plot

Fig. 5

Heat map representing the amount of common genes in percentage between modules in the SMA and control group of data set GSE102204. Rows denote the control group and columns denote the SMA group