The recent finding that oxidation of 5mC to 5hmC by enzymes of the ten-eleven translocated (TET) family occurs in mammalian genomes has raised many questions regarding the role of this DNA modification in epigenetic regulation. Even though several studies have investigated the complex role of TET proteins and 5hmC in embryonic stem cell biology, the relevance of this mark in developing normal and adult tissues remained essentially unexplored.
Here, we developed a novel, robust immunohistochemical detection method for 5hmC and used this method to detect 5hmC in a large number of murine and human tissues. Interestingly, we found that hierarchically organized epithelia as well as hematopoietic cells in the bone marrow show a differentiation-dependent 5hmC distribution. Cells in the colonic crypt, basal cells of the prostate, as well as hematopoietic stem/progenitor cells exhibited greatly reduced 5hmC levels compared to more differentiated counterparts, suggesting that adult tissue stem/progenitor cells across a broad range of tissue types might be characterized by low 5hmC levels. Differentiation and maturation conversely appeared to be associated with an increase in 5hmC. Based on these data, we can hypothesize that accumulation of 5hmC in the genome is involved in differentiation of tissue stem/progenitor cells. This hypothesis is supported by recent reports showing that genetic disruption of TET2 in hematopoietic cells could lead to increased hematopoietic stem cell self-renewal, accumulation of hematopoietic stem/progenitor cells, and reduced differentiation of hematopoietic stem cells [39
This observation is somewhat in contrast to recent reports from murine embryonic stem cells, where the differentiation of embryonic stem cells appeared to be associated with a loss in 5hmC [24
]. These discrepancies could reflect differences in the biology between embryonic and tissue stem cells and could point to a differential role of 5hmC in very early development versus later development and adult tissue development/differentiation.
Recent reports on the detection of 5hmC in adult tissues have been somewhat conflicting [22
]. One explanation for these variable results is certainly the use of different detection methods. In this study, we noted that robust immunohistochemical detection of 5hmC from formalin fixed paraffin embedded tissue requires specific antigen retrieval. Omission of these antigen retrieval steps led to vastly different results (Supplementary Figure 1
) and, therefore, explained some of the prevailing discrepancies in the literature.
The functional role of 5hmC in regulating differentiation and epigenetic states of adult tissues remains unknown. It has been proposed that 5hmC cannot be bound by methyl-binding domain proteins such as MeCP2, MBD1, and MBD2 [41
], which are known to associate with 5mC and recruit the chromatin repression complex. Accumulation of 5hmC could therefore have a significant impact on gene expression states. Moreover, it was suggested that 5hmC is not recognized by the DNA methylation maintenance machinery, suggesting that the presence of 5hmC could lead to a passive loss of DNA methylation during cell division [44
]. Most interestingly however, the conversion of 5mC to 5hmC could also represent a mechanism for active demethylation. In a process that involves activation induced deaminase (AID) and base excision repair, 5hmC can be converted to cytosine [45
], providing a mechanism for the sequential, active conversion of 5mC to cytosine. Such a process provides an interesting mechanism for plasticity of DNA methylation marks.
Our observation that 5hmC levels are significantly reduced in three different types of human carcinoma suggests that the loss of 5hmC could be a general feature of carcinogenesis. Indeed, in several hematological malignancies including AML and MDS, reduced 5hmC levels have been associated with mutations in the TET
]. However, it is unlikely that missense mutation in the TET enzymes can explain the almost universal reduction in 5hmC levels in colorectal, prostate and breast carcinoma, since large scale sequencing efforts have not identified TET family members as frequently mutated in these tumors [49
]. Recent evidence suggests that a large number of oxidizing enzymes, including the TET family, can be inhibited by oncogenic metabolites, such as 2-hydroxyglutarate [53
]. It is, therefore, possible that cancer specific metabolic perturbations can influence 5hmC levels and, consequently, alter the epigenetic makeup of a cell.
In many solid tumors, cancer progression is associated with a progressive loss of 5mC marks resulting in a global hypomethylation phenotype [3
]. Since 5mC is required as a substrate for oxidation to generate 5hmC, reduced 5mC levels could explain, at least partly, the decrease of 5hmC observed in tumors. To address a possible correlation between 5hmC and 5mC loss we stained a series of tumor and normal tissues from prostate and colon with an antibody that specifically recognizes 5mC (Supplementary Figure 4
). Using this method, we observed only a modest reduction of global 5mC staining intensities between cancerous and normal tissue of the colon and prostate, which is inline with recent reports [38
]. Furthermore, we found no association between 5mC and 5hmC staining levels suggesting that the reduction in 5hmC can occur independently of reductions in 5mC.
In conclusion, our study identifies a hierarchical distribution of 5hmC levels in embryonic and adult tissues and provides evidence for a cancer-associated loss of 5hmC.