Epigenetics modulates the complexity of the response to Immune Checkpoint Blockade

The emergence of Immune Checkpoint Blockade (ICB) therapy has unleashed an abundance of promising cancer treatment alternatives that focus on the re-activation of an immunosuppressed tumour environment. These strategies, initially applied to advanced stages of cancer with durable responses and acceptable toxicity, are also transforming the clinical practice in the adjuvant and neoadjuvant settings and are indicated in more than 15 types of tumours. However, the frequent occurrence of innate or acquired resistance, of up to 85% of the patients, has turned the focus on precision medicine to decipher which patients will benefit from the treatment, and on combination therapies with multiple targets that can holistically combat the disease [1]. In this context, despite a nascent instrument in clinical practice, epigenetic marks stand as putative biomarkers informative both of response to immunotherapy and of the disease or therapy-induced dynamic landscape. In addition, such mouldable nature of the epigenetic system is the cornerstone of the development of several epigenetic drugs that are currently combined with Checkpoint Inhibitors in Phase I and II Clinical Trials [2]. Amongst the plethora of receptors, ligands, cytokines and immune cells that interact in the tumour microenvironment to promote or counteract immunosuppression, Lymphocyte- associated gene 3 (LAG3) stands as a promising checkpoint target for the promotion of tumour immunosurveillance when combined with other checkpoint molecules [3]. Given the increasing evidence on the insufficient immune-reactivation as one of the main reasons for low response rates in ICB in cancer, and the emerging relevance of epigenetic regulation of T lymphocytes after antigen presentation [4], in this article of EBioMedicine, Fröhlich et al. [5] address the evaluation of LAG3 methylation as a prognostic factor and predictor of response to anti-PD-1 and anti-CTLA4 ICB, also in combination with anti-IDO immunotherapy in melanoma. Remarkably, they identify a pattern of differential methylation in the promoter region of LAG3 with respect to the gene body and 3′ UTR regions, and trace its association to mRNA expression, a “hot” tumour signature of pro-inflammatory immune cells and cytokines, as well as to improved overall survival in melanoma and progression-free survival in the ICB treated cohort. In addition, they demonstrate in vitro that LAG3 is also expressed to a certain extent in melanoma cells, and, relevantly, that expression is inducible under exposure to 5-azacytidine.

Until recently, LAG3 epigenetic control had been restricted to histone modifications and miRNA inhibition; however, earlier this year, Klümper et al. reported for the first time the association of LAG3 methylation with gene expression in clear cell renal cell carcinoma [6]. Fröhlich et al. [5] extend this association to TCGA bulk melanoma samples and delineate region specific correlations between DNA methylation and mRNA expression, associating the hypomethylation of the promoter region of LAG3, which contains a CpG island (CGI), and the hypermethylation of the gene body and 3′UTR region outside the CGI, with expression. Interestingly, they are able to assign this specific methylation pattern to the tumour immune infiltrate by correlating it with the leukocyte fraction, lymphocyte score and Interferon gamma, and the opposite pattern to the tumour cell fraction by means of association with the tumour content and purity. Given that bulk, normally formalin-fixed paraffin-embedded tissues, are the most abundant deposited clinical samples, LAG3 methylation could be of use as a surrogate marker of immune infiltration. In addition, this hypothesis is reinforced by the identification of similar patterns of deconvoluted TCGA methylation in pro-inflammatory and activated immune cells, as well as in regulatory T lymphocytes. Moreover, these correlations are also validated in a separate melanoma cohort with bulk tissues where they associate the N-shore, first exon, promoter CpG hypomethylation with the immune infiltrate, as well as the respective hypermethylation with the tumour fraction.

While it is known that gene body methylation can be a mark of transcriptional activity, gene body regions and enhancers of actively transcribed genes can also be enriched in DNA hydroxymethylation [7]. Moreover, DNA hydroxymethylation is typical of intermediate density CpG regions. According to that, and given that DNA hydroxymethylation cannot be distinguished from DNA methylation with the methods used in this work, one could also speculate that the hypermethylation observed in the non-CGI gene body and 3′UTR regions of LAG3 is actually increased DNA hydroxymethylation, which is why the correlation with expression is positive in those regions. The identification of a CCCTC-binding factor (CTCF) binding site in the 3′UTR region of LAG3, led the authors to hypothesize that DNA methylation in this region could affect the binding of this dual Transcription Factor. Noteworthy, CTCF binding sites have also been associated to DNA hydroxymethylation, so the question arises whether LAG3 upregulation in the context of the T lymphocyte exhaustion phenotype induced by persistent tumour antigen presentation could be associated with the re-patterning of both DNA methylation and hydroxymethylation. Interestingly, the regions that tend to have higher variation in DNA methylation are the shores and shelves, and indeed Fröhlich et al. [5], when comparing LAG3 methylation across different cell types (peripheral blood mononuclear cells (PBMCs), melanocytes and melanoma cell lines), unmask DNA methylation variability between melanoma cell lines and melanocytes at precisely CpG sites located in LAG3 regions of non high CpG density.

In addition to the cell type specific dissection of the correlations between LAG3 methylation and expression, the current work by Fröhlich et al. [5] includes another major asset that is the novel prognostic and response predictive value of LAG3 methylation in the context of anti-PD-1 and anti-CTLA-4 ICB, being hypomethylation associated with better outcomes. This is particularly relevant given that only in few genes, methylation has been previously associated to drug response to ICB (PD-L1, CTLA-4, FOXP3) [2], and evidences the myriad of players in the regulation of the dynamics of the immunosuppression in the tumour microenvironment that can influence the clinical benefit of this kind of therapies.

Finally, another important finding in Fröhlich et al.´s work [5] is the detection of a low basal but epigenetically inducible LAG3 expression in melanoma cell lines, opening a new avenue of investigation on the putative interference of the tumour co-expression of LAG3 and its ligands MHC-II in general, and FGL1 o LSectin in melanoma cells [3], in the response to ICB, as it occurs with the tumour co-expression of PD-1 and anti-PD-L1, where they mediate the resistance to anti-PD-1 and anti-PD-L1 treatments by activation of AKT and ERK1/2 [8].

In summary, the work of Fröhlich et al. [5] introduces several findings relevant to the field of ICB, where melanoma LAG3 region-specific methylation deserves further attention with respect to the prediction of prognosis and the response to ICB, and the intrinsic tumour LAG3 expression inducible by pharmacological de-methylation raises the need to decipher its interaction with ICB and with epigenetic drugs that are currently being tested in combination with ICB.

Funding sources

The work in Isabel Barragan's laboratory is funded by Instituto de Salud Carlos III through the project “PI18/01592” (Co-funded by European Regional Development Fund/European Social Fund “A way to make Europe”/“Investing in your future”), Sistema Andaluz de Salud, Nicolas Monardes Program, through the project “SA 0263/2017”, Spanish Group of Melanoma (Award to Best Research Project 2020), Fundación Bancaria Unicaja through the project “C19048”, and Andalusia-Roche Network Mixed Alliance in Precision Medical Oncology.

Declaration of Competing interest

The author declares no conflicts of interest.