In this study, we employed a two-step method to identify the differentially methylated regions and 4 regions were found to have different methylation levels in PFC between humans and rhesus macaques, which are informative candidates for further study of human brain evolution. However, there are also methodological limitations. The 150 candidate regions are all high CpG content regions because the NimbleGen method tends to find highly methylated regions with high CpG content. It was reported that high CpG content regions were of higher quality in MeDIP-Chip data . In studies of tissue specific DNA methylation regions, high CpG content regions such as the CpG islands showed less between individual variance and less differences between different tissues [28–30], and this may be the same case between different species. Though we cannot exclude the possibility that there are differentially methylated regions in low CpG content regions, the high CpG content regions, on which we focused in this study, are more likely to be functionally important [14, 15] and more likely to have definite between-species differences and less within-species variations, thus evolutionarily more significant. In addition, considering the multiple data filtering steps using different methods, the number of identified DMRs in this study was likely an underestimation of the between-species methylation divergence in the brain.
Repeat elements are important in DNA methylation. Farcas et al. reported a small region of ALU-Sg1 element that was differentially methylated between human and chimpanzee cerebral cortex. In this study, we found another case of SINE element differentially methylated between humans and rhesus macaques, implying the roles of repeat elements in the epigenetic evolution of primates. In the rhesus macaque genome, the MIR3 element located in DMR8 has a length of 98 bp and divergence of 34.4% to its consensus sequence. In DMR8, the human ortholog of this region have 6 substitutions and 1 deletion compared with rhesus macaque, and it is not annotated as repeat region by UCSC. MIR3 belongs to the MIR (Mammalian-wide interspersed repeats) family, which is one of the oldest tRNA-derived SINE (short interspersed element) elements, and they integrated into the host genomes before the radiation of mammals . In DMR8, three CpG sites (CpG12, CpG13 and CpG14) of rhesus macaques are located in MIR3. Interestingly, the three CpG sites are the only CpG sites that showed significant methylation differences between humans and rhesus macaques (Figure 3 and Table 4), suggesting that the MIR3 element has led to the differential methylation of DMR8 between humans and rhesus macaques.
Among the 4 validated DMRs, only one is located in the repeat region, suggesting that the differentially methylated regions between humans and nonhuman primates are not necessarily restricted to repeat regions and can be high CpG content regions, which have been previously thought to be less variable [28–30].
As DNA methylation in the promoter regions will repress gene expression, it would be informative to see whether the differential DNA methylations between humans and rhesus macaques would lead to expression differences. We analyzed the published gene expression data from human and rhesus macaque cerebral cortices . Among the four DMRs identified in our study, two DMR related genes (ICAM1andProSAPiP1) had eligible expression data and showed significant gene expression differences between human and rhesus macaque (P = 0.000757 and 0.00764 for ICAM1 and ProSAPiP1 respectively). However, for both genes, the species having higher DNA methylation levels also showed higher gene expression levels, which is inconsistent with the expected repression of gene expression by a higher level of DNA methylation in the promoter. However, these two DMRs also covered the first exons of the corresponding genes, and the gene-body methylation was reported to be positively correlated with gene expression in human cells . Hence, detailed analysis in the future is needed to reveal the influence of DMRs on gene expression.
It is difficult to address how the observed methylation differences have been formed during primate evolution. Trans-generational inheritance is crucial if epigenetic modification should play a role in evolution. Many studies have reported that epigenetic modifications can be inherited across generations [33–35]. But the inheritance of DNA methylation is not as stable as that of DNA sequence and the mechanisms underlying the trans-generational inheritance of epigenetic modification are not well understood.
Recently, it was reported that DNA methylation is correlated with DNA sequence variations [36–38] and non-coding RNAs [39–42], and these correlations might act as an indirect mechanism that explains the trans-generational inheritance of epigenetic modification. The DNA sequence substitutions during evolution may cause the emergence of new CpG sites and/or loss of the existing CpG sites, eventually leading to methylation divergence of specific genomic regions. It has been reported that the CpG-SNPs in the human genome have contributions to allele specific DNA methylation , implying that even at the population level, the methylation divergence can occur due to DNA sequence polymorphisms. Alternatively, DNA sequence differences could also result from DNA methylation differences between species [10–12].
Another possibility is that epigenetic modifications altered by environment can be transmitted to the next generation directly. Global epigenetic reprogramming including demethylation of DNA occurs in the mammalian primordial germ cells and in early embryos . Although global epigenetic reprogramming will restrict the trans-generational epigenetic inheritance, the erasure of DNA methylation modifications is not absolute [44, 45], suggesting the possibility of direct transmitting of DNA methylation modifications to the next generation. Additionally, at the whole genome level, the over-all similar DNA methylation patterns between humans and macaques cannot be explained by the similarity of living environment because they inhabit totally different environments, suggesting potentially vertical inheritance of methylation.
A lot of candidate DMRs obtained from the MeDIP-Chip analysis were not validated by MassARRAY and bisulfite sequencing. This discrepancy could be explained by two possible sources. First, for MeDIP-Chip analysis, we used different arrays for humans and rhesus macaques. Even though the validation for each species-specific array is good, the technical bias may still exist, as the array for rhesus macaques has not been rigorously tested. Second, besides detecting the genome-wide DNA methylation patterns, we focused on identifying differentially methylated regions between humans and non-human primates. Thus, a relaxed cut-off (nominal P smaller than 0.05 without multiple test correction) was applied for selecting the 150 candidates from the MeDIP-Chip data, likely resulting in a relatively high number of false-positives.