We identified a total of 469 DMRs, consisting of 267 neuronal and 202 nonneuronal DMRs. Among them, 67.4% (n = 180) and 41.1% (n = 83) were hypomethylated in neurons and nonneurons, respectively (Fig. 1b, Additional file 3: Tables S1–S4). DMRs were distributed throughout the mouse genome (Fig. 1c). Notably, in nonneurons, three DMRs detected in the upstream sequence of CamK2a showed hypermethylation in Tg mice compared to nonneurons of wild-type mice (Fig. 1a). In addition, one DMR detected in the Arsi showed hypermethylation in both cell types of Tg mice. However, these hypermethylation sites were attributed to changes in the methylation profile of the transgene construct consisting of a long upstream region of the Camk2a gene, which contains a part of a neighboring gene, Arsi. The fact that the promoter region of Camk2a on the transgene was densely methylated in only nonneurons suggested the successful neuron-specific regulation of the mutant Polg1 expression in Tg mice.
There were a few overlapped DMR-associated genes between neurons and nonneurons, and most of them showed cell-type-specific DNA methylation changes (Fig. 1b, Additional file 1: Discussion). Given that the deleted mtDNA primarily accumulates in neurons [10], the DMRs in nonneurons were considered to be secondarily induced by neuron-nonneuron interactions during development. The top hit list of DMR-associated genes in neurons included hypomethylation of Kif3b and hypermethylation of Sun5, Dclre1c, Bpifb2, Thoc3, and Nkain4. In nonneurons, hypomethylation of Zfp712, Gal3st1, and Ermp1, and hypermethylation of Kcnh5 and Gipc2 were included (Fig. 1d).
We performed Gene Ontology analysis using DMR-associated genes. In neurons, we found that cell cycle-, cell division-, and inhibition of peptidase activity-related genes were enriched, whereas cell junction, synapse-related genes, and GABA receptor-related genes were enriched in nonneurons (Fig. 1c and Additional file 3: Table S5).
We then compared the DMR-associated genes and differentially expressed genes (DEGs) that were previously obtained from the frontal cortices of mutant Polg1 Tg mice [13]. There were seven common genes between neuronal DMR-associated genes and DEGs (Bnip2, Mylip, Psmc4, Raf1, Trim2, Galnt2, and Lrpprc) (Fig. 1e and Additional file 1: Discussion). The relatively small number of overlapping may be because gene expression profiles were obtained from unsorted bulk cortical tissues. Among the overlapping genes, two genes (Trim2 in neurons and Lrpprc in nonneurons) showed a typical relationship between hypermethylation and downregulation of gene expression (Additional file 1: Discussion).
We then compared DMR-associated genes in mice and postmortem brains of patients with BD [15]. We found hypomethylation of two genes (Gli2 and Il1r2) in neurons and six genes (Ccr5, Crbn, Erlin2, Plch1, Prune2, and Sec16A) in nonneurons (Fig. 1e, Additional file 1: Discussion). We observed that DNA methylation changes in mutant Polg1 Tg mice shared several features of those identified in the postmortem brains of patients with BD. First, neurons showed more hypomethylation changes than hypermethylation changes in the mutant mice (binominal test, P < 0.0001) and also showed hypomethylation of molecular motor genes in neurons. Second, DMR-associated genes in nonneurons showed clearer enrichment in the synapse and neurotransmitter-related genes than in neurons.
In conclusion, we characterized epigenetic alterations in the neurons and nonneurons of mutant Polg1 Tg mice. There were some sex-specific behavioral alterations in mutant Polg1 Tg mice, such as the propensity for spontaneous recurrent hypoactivity in females [11]. Therefore, DMRs may be different between male and female Tg mice. Although further validation and functional assays as well as sex-specific analysis will be needed, the DMRs identified in this study, part of which were shared with BD patients, will contribute to the understanding of the pathophysiology of BD from an epigenetic perspective. Among the DMRs, those associated with gene expression changes, and those common to postmortem brains of BD will be particularly important candidates related to the pathophysiology of BD.