Module-trait correlation analyses revealed two meta-modules presenting high positive/negative correlations with severe (GCL3, GCD2, E-IPI, SMI, and SEFI) or mild (GCL0-1 and GCD1) phenotypes. In the meta-module I the transcriptional modules are positively correlated with the severe phenotypes and negatively with the mild phenotypes. Inversely, in the meta-module II all modules are negatively correlated with the severe phenotypes and positively with the mild phenotypes (Fig. 2c), except the red module, the only module that is positively correlated with E-onset and severe memory impairment. These correlations clearly indicate two distinct CA3 transcriptome profiles: one for severe and another for mild phenotypes. In addition, the functional enrichment analysis showed that the red and blue modules had many genes over-represented in neuron/excitability processes, such as neuron, glia, axon, myelination, excitability, and synapse (Fig. 3). In order to get a better understanding of how these modules are associated with DG alterations, cognitive impairment, and disease onset, we focused on the hierarchical characterization and functional interpretation of the genes within each module.
The blue and yellow modules (Fig. 4) harbor HH genes positively associated with pathophysiological mechanisms in epilepsy - such as vascular remodeling (ITGB8, UCQRB), the glutamatergic excitatory system (TANC2, SCLC25A18) and reactive gliosis (SPOCK1) – and with putative compensatory mechanisms linked to hippocampal neuron survival (KIF1B, YAP1). In the turquoise module seven out of its nine HH genes are positively correlated with the mild phenotype GCL0-1 (Fig. 5). Two HH genes in this module, PIK3C3 and DCP1B, are respectively involved in CNS neuronal homeostasis and in regulating human memory performance, thus indicating protective and compensatory mechanisms. Significantly, all the genes mentioned above, excepting SPOCK1, are HHubs or eHubs i.e., genes with high intramodular and/or whole network connectivity. These genes probably are essential genes with a relevant role in defining the modules’ biological function23,25, namely, the association of blue and turquoise modules with severe and mild phenotypes, respectively.
The red module, occupying a separate branch in the eigengene dendrogram, is highly and positively correlated with E-onset, and positively correlated with severe memory impairment (Fig. 2c). Early MTLE onset with HS has a large impact on brain plasticity and on brain connectivity and memory78,79. Moreover, early seizure onset is also a predictive factor for pharmacoresistancy6,80, being associated with a more severe functional abnormality in the ictal hippocampus79. Interestingly, five out of the six HH genes in the red module are involved in processes compatible with early tissue damage and subsequent microstructural reorganization, specifically: axon sprouting (OLFM1); neuronal circuitry remodeling (ST8SIA3); hippocampal neurogenesis, (NCDN and ERBB3), and astrocyte morphology remodeling (SYNJ2). Interestingly, the two hypo-expressed HH genes in the red-module, ERBB3 and SYNJ2, are also related to cognitive abilities and memory. Therefore, the co-expression profile, functional role, and connective hierarchy of the HH genes in the red module (Fig. 5) appear to be compatible with the module’s phenotypic profile.
Contrariwise to hubs, that occupy a topologically central position in the modules and confer robustness to the co-expression network, HGS genes are usually at the network’s periphery and their expression shows significant variation across trait groups23,25,81. The HGS genes are significant for the traits, as described in the Results section and depicted in Fig. 6, and here we show how these genes may serve as biomarkers or therapeutic targets for RMTLE.
Let us consider first the six HGS genes correlated with GCL. We found that ISYNA1, whose brain expression is persistently elevated in a rat model of epilepsy64, is more expressed in GCL3 than in GCL2 and GCL0-1, confirming its proposed role as RMTLE biomarker and its potential as a therapeutic target64,82. CHSY1 is related to the maintenance of hippocampal volume in RMTLE patients and its expression increases from GCL3 to GCL0-1, so indicating a compensatory mechanism and the gene’s usefulness as biomarker. The genes AMFR and PTLP are respectively involved in neuronal survival and in neuroprotection, and both are necessary for memory and learning, but the expression of AMFR increases from GCL3 to GCL0-1 whereas that of PTLP decreases, thus indicating pathogenic and compensatory mechanisms. Finally, LGR4 and MGAT2, also HH genes and respectively involved in regulating Wnt/β catenin signaling and neuronal viability, have increased expression levels from GCL3 to GCL0-1, implying a compensatory mechanism and that these two genes may be meaningful biomarkers for RMTLE and hippocampal sclerosis.
Four HGS genes are correlated with GCD and two of them, ERBB4 and MDH2, directly involved in epileptic processes, have significantly higher expression in GCD3 and GCD2, being potential biomarker candidates. The other two genes, SMARCAD1 and DHX9, are involved in chromatin remodeling and maintenance of genomic stability, and their average expression levels varies across GCD grades. The HGS gene HYOU1 has a neuroprotective role and, as expectable, is relatively higher expressed in the absence of GCB. ASNSD1, a gene involved in biological processes related to memory, has a diminished expression in patients with severe memory impairment. Finally, three HGS genes, VLDR, FGFR2, and HBEGF, that are involved in synaptic plasticity and memory formation and belong to the red module - which is correlated with E-onset and memory impairment - have significant decreased expression in patients with early disease onset.
In conclusion, by adopting a systems biology approach and integrating clinical, histopathological and transcriptomic data, we were able to identify transcriptional modules highly correlated with DG alterations, cognitive dysfunctions, and disease onset in our RMTLE patients. The functional characterization of the high-hierarchy genes in each module allowed us to unveil the modules’ main biological functions, paving the way for further investigations on their roles in RMTLE pathophysiology. Moreover, we found several HGS genes which have the potential to become novel biomarkers and/or therapeutic targets in RMTLE. These results are relevant considering the urge for identifying the genomic mechanisms underlying RMTLE, what could lead to more effective therapeutic interventions.