Co-expression network modeling identifies key long non-coding RNA and mRNA modules in altering molecular phenotype to develop stress-induced depression in rats

Abstract

Long non-coding RNAs (lncRNAs) have recently emerged as one of the critical epigenetic controllers, which participate in several biological functions by regulating gene transcription, mRNA splicing, protein interaction, etc. In a previous study, we reported that lncRNAs may play a role in developing depression pathophysiology. In the present study, we have examined how lncRNAs are co-expressed with gene transcripts and whether specific lncRNA/mRNA modules are associated with stress vulnerability or resiliency to develop depression. Differential regulation of lncRNAs and coding RNAs were determined in hippocampi of three group of rats comprising learned helplessness (LH, depression vulnerable), non-learned helplessness (NLH, depression resilient), and tested controls (TC) using a single-microarray-based platform. Weighted gene co-expression network analysis (WGCNA) was conducted to correlate the expression status of protein-coding transcripts with lncRNAs. The associated co-expression modules, hub genes, and biological functions were analyzed. We found signature co-expression networks as well as modules that underlie normal as well as aberrant response to stress. We also identified specific hub and driver genes associated with vulnerability and resilience to develop depression. Altogether, our study provides evidence that lncRNA associated complex trait-specific networks may play a crucial role in developing depression.

Fig. 1. Behavioral paradigm to induce learned helplessness in rat and measuring the effect of shock on their avoidance learning followed by adhering changes in coding genes (mRNA) and their expression-based module assignments to establish a module trait relationship in LH vs. TC rats.

a Schematic diagram of the timeline followed as part of the stress paradigm to induce LH behavior in rats. b Bar diagrams represent escape latencies in tested controls (TC), non-learned helpless (NLH), and learned helpless (LH) rats measured on days 2, 8, and 14, respectively. Data are the mean ± SEM. On day 2, the NLH rats did not show any significant (^ap = 0.15) difference in escape latency compared with the TC group. A significantly (^bp < 0.001) higher escape latency was observed for LH rats compared with TC on day 2. Similarly, LH rats showed significant difference (^cp < 0.001) in mean escape latency compared with the NLH group on day 2. On day 8, NLH rats did not show any significant (^dp = 0.74) difference in escape latency compared with the TC group. A significantly (^eP < 0.001) higher escape latency was noted for LH rats compared with TC rats. Similarly, LH rats showed significant difference (^fp < 0.001) in mean escape latency compared with the NLH group. On day 14, NLH rats did not show any significant (^gp = 0.19) difference in escape latency when compared with the TC group. Individual group comparison identified a significantly (^hp < 0.001) higher escape latency for LH rats compared with TC rats. Similarly, LH rats showed significant difference (ⁱp < 0.001) in mean escape latency compared with the NLH group. c The figure demonstrates the protein-coding gene-based cluster dendrogram analysis in LH vs. TC group. Five colors, which represent the modules, include blue, brown, green, turquoise, and yellow. Dynamic tree cut method was implemented for analysis. The degree of co-expression between the genes assigned by the same module was relatively higher. d The figure represents the correlation between mRNAs module eigengenes and phenotypic traits. Each row represents the module eigengene or ME (ME is the correlation matrix of module and sample, labeled by color) and each column represents a trait. Each square contains the Pearson’s correlation coefficient between the MEs and trait and their associated p values. The red and blue colors show a strong positive and negative correlation, respectively

Fig. 2. Identification of intramodular connectivity and gene significance supported by network analysis.

a The correlation analysis between intramodular connectivity and gene significance for five independent modules are individually represented with scatterplot. The intramodular connectivity of two modules containing turquoise and green was significantly correlated with gene significance. b The igraph generated network-based visualization demonstrates the hub genes from the “Turquoise module” associated with susceptibility phonotype. The red dots represent the hub genes which include Expi, Tas2r116, and Rnf29. c The figure represents the hub genes associated with “Green module”-based network. The two hub genes from this network, LOC690326 and Oprs1, are also represented with red dot

Fig. 3. LncRNA expression-based module assignments and module trait relationship in LH vs. TC groups.

a Using topological overlapping matrix dissimilarity, the cluster dendrogram was prepared to show four individual modules including LTCblue, LTCbrown, LTCturquoise, and LTCyellow. b The representative figure demonstrates the correlation analysis between the four modules and depression phenotype. Unlike the three other positively correlated modules (LTCbrown, LTCturquoise, and LTCyellow), the LTCblue module shows negative correlation with depression (LH) phenotype

Fig. 4. Correlation analysis for gene significance and scaled connectivity mapped with hub lncRNAs in the network.

a The scatterplots represent the relationship between gene significance and connectivity based on lncRNA associated expression changes in LH vs. TC groups. The correlation between gene significance and connectivity are associated with LTCbrown (p = 0.0017) and LTCturquoise (p = 0.047) modules. b The figure represents network associated with LTCbrown module with connectivity to lncRNA transcribing hub genes. c The figure represents the hub genes associated with “Turquoise module”-based network

Fig. 5. Venn diagram of overlapping mRNA and lncRNA derived from three group comparisons.

a Differentially expressed mRNAs are represented on this Venn diagram, showing either distinct or overlapping relationship with depression or resiliency phenotype. b The diagram represents the unique and overlapping phenotypic association of differentially expressed lncRNA with resiliency or susceptibility to develop depression