Tumors smother their epigenome

ABSTRACT

DNA hypermethylation is pervasive in tumors, but the factors triggering this modification are largely unknown. We recently demonstrated that the activity of 10-11-translocation methylcytosine dioxygenases, initiators of DNA demethylation, is compromised in hypoxic tumors. The resultant accumulation of methylation inactivates associated genes, linking the tumor microenvironment to epigenetic changes in cancer cells.

Cytosine methylation (5mC) at the DNA of promoters regulates the expression of associated genes, with high levels of 5mC (hypermethylation) being associated with gene repression. Tumor cells are epigenetically distinct from their untransformed counterparts, displaying global DNA hypomethylation and focal hypermethylation.¹ The latter frequently affects tumor suppressor genes. Similar to inactivation through mutations, epigenetic inactivation of these genes confers selective growth advantages to tumor cells. How tumor cells acquire this hypermethylation is, however, mostly unknown. In a small subset of solid tumors, DNA hypermethylation is caused by mutations that either directly or indirectly inactivate 10-11-translocation methylcytosine dioxygenase (TET) enzymes, which are emerging as important tumor suppressors.² These TET enzymes normally catalyze DNA demethylation: 5mC is oxidized by TETs to 5-hydroxymethylcytosine (5hmC) and other cytosine derivatives, which, after base excision, are replaced by an unmodified cytosine to achieve DNA demethylation (Fig. 1).³ Loss of TET activity and DNA demethylation thus causes DNA hypermethylation.

Figure 1.

Hypoxia triggers DNA hypermethylation by inhibiting methylation turnover. DNA methyltransferases (DNMTs) methylate cytosines. Under normoxic conditions (left), 10-11-translocation methylcytosine dioxygenases (TETs) maintain active regulatory regions in an unmethylated state by oxidizing these methyl groups (me) to hydroxymethyl (hm) using α-ketoglutarate (αKG) and oxygen (O₂) as substrate and iron and vitamin C as cofactors. This leads to base-excision repair (BER) of these modified cytosines and removal of spurious methylation by active demethylation. Under hypoxic conditions (right) TET activity is compromised, leading to accumulation of methyl groups and causing DNA hypermethylation.

In a recent report,⁴ we demonstrated that purified TET enzymes require greater than 2% oxygen to function normally, and that TET activity is halved at oxygen levels of 0.5%. Because oxygen levels are below 0.5% in on average one-third of tumor areas, this finding implies that TET activity is often compromised in tumors. We confirmed this in vitro by measuring 5hmC levels in cell cultures exposed to 0.5% O₂, which showed global loss of 5hmC. Genome-wide analyses in MCF7 cells confirmed this hypoxia-associated loss of 5hmC, and demonstrated that it was linked to locus-specific gains in 5mC and to a reduction of gene expression. Comparison of 12 hypoxic and 12 normoxic tumors from lung cancer patients⁵ yielded similar results, with a hypoxia-associated loss of 5hmC and gain of 5mC at loci showing reduced 5hmC.

To assess the broader impact of hypoxia on DNA hypermethylation in patient tumors, we leveraged gene expression and DNA methylation data from 8 solid tumor types analyzed in The Cancer Genome Atlas (3,141 tumors).⁶ For each type, tumors were stratified into normoxic, intermediately hypoxic, and hypoxic groups based on an established hypoxia metagene expression signature. In the same tumors, we also quantified 2 independent measures of hypermethylation. On one hand, we determined average methylation levels per tumor at the 1,000 cytosines showing the strongest hypermethylation in tumor versus normal tissue. On the other hand, for each tumor we quantified the number of cytosines showing exceptionally high methylation levels (“hypermethylation events”). Importantly, both of these hypermethylation measures were increased in hypoxic versus normoxic tumors, confirming that tumor hypoxia causes DNA hypermethylation. Hypoxia-associated increases in hypermethylation events were moreover independent of changes in TET expression, frequent somatic mutations, proliferation, tumor cell percentage, tumor size, immune cell infiltration, or metastasis. Using statistical modeling, we estimated that, combined, these covariates predicted a significant fraction of hypermethylation events, with the largest fraction of this predictive power (33%) being ascribable to hypoxia. Hypoxia thus underlies a significant fraction of the hypermethylation that is present in any tumor. These findings highlight a strong connection between the tumor microenvironment and the epigenome of cancer cells, with the tumor vasculature shaping cancer cell-intrinsic properties. They also suggest that variability in the microenvironment can underlie epigenetic heterogeneity within tumors.

Importantly, hypermethylation events were functional as they were associated with a reduction in gene expression. Moreover, they frequently affected genes involved in cell cycle regulation, apoptosis, and DNA repair, functions typically linked to tumor suppressor genes. This was further confirmed in an analysis of breast tumor methylomes, in which a defined set of tumor suppressor genes, but not oncogenes, showed hypoxia-associated hypermethylation. To test whether this could also be recapitulated experimentally, we investigated a mouse model of spontaneous breast tumors. This model showed increases in methylation at promoters of tumor suppressor genes but not oncogenes, concomitant with oncogenic progression and the associated development of hypoxia in these tumors.⁷ Importantly, experimentally increasing hypoxia in this model by tumor blood vessel pruning accelerated this hypermethylation and was associated with a 25% reduction in 5hmC. Conversely, alleviating tumor hypoxia by blood vessel normalization increased 5hmC levels by 14% and rescued the hypermethylation at tumor suppressor gene promoters.

In conclusion, our findings highlight an important role for hypoxia-mediated reduction of TET activity in the acquisition of DNA hypermethylation and the resultant inactivation of tumor suppressor genes. They moreover indicate that bolstering TET activity can reduce the acquisition of this hypermethylation, suggesting novel avenues for cancer therapy. For example, antiangiogenic drugs currently available in the cancer clinic can promote blood vessel normalization and improve tumor oxygenation. Other mechanisms through which TET enzyme activity can be increased may similarly represent appealing targets: these include increases in TET expression, decreases in reactive oxygen species, ascorbate supplementation, and reducing levels of metabolites that compete with the TET cofactor α-ketoglutarate (Fig. 1).^2,8-10 Some drugs that interfere with these mechanisms are in development or already available, and it remains to be determined how the activity of this tumor suppressor can best be increased in a clinical setting. In addition, further research will establish whether boosting TET activity in tumors can also reverse the hypermethylation that was acquired prior to therapeutic interventions, and thus offer the exciting prospect of alleviating the epigenetic repression of tumor suppressor genes.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

ORCID

Bernard Thienpont http://orcid.org/0000-0002-8772-6845