Recently, small-molecule inhibitors of general transcriptional regulators such as BET proteins and the RNA-PolII–regulating kinase CDK7 have been shown to have efficacy in multiple solid and liquid tumors. An article in this issue of Cancer Discovery identifies a nongenetic mechanism of resistance related to deficiency of folate that leads, via increased S-adenosylhomocysteine and reduced repressive histone methylation, to reactivation of a transcriptional program which promotes AML cell survival under the pressure of BET inhibition.
See related article by Su et al., p. 1894.
An improved understanding of genome and epigenome alterations in cancer over the past two decades has highlighted a critical role for epigenetic regulators in the initiation and maintenance of malignancy. This observation has led to the development of a number of specifically and more generally targeted inhibitors of malignant transcription. In acute myeloid leukemia (AML), these inhibitors are showing promise in the clinic, and some, such as the IDH1 and 2 inhibitors ivosidenib and enasidenib, have already been licensed (1). More general transcriptional inhibitors include small-molecule inhibitors targeting the bromodomain and extra terminal (BET) proteins as well as inhibitors of the kinases that regulate RNA-PolII, CDK7, and CDK9. The BET family includes the ubiquitously expressed BRD2, BRD3, and BRD4, and the testes-specific BRDT (2). They are essential for cellular homeostasis, and regulate a plethora of cellular processes by binding acetylated lysine residues of histone (and non-histone) proteins, through their tandem N-terminal bromodomain. They can also mediate a number of effects including transcriptional activation via recruitment of other partner proteins, including the positive transcription elongation factor b (P-TEFb), which contains the RNA-PolII kinase CDK9. P-TEFb phosphorylates the C-terminal domain (CTD) of RNA-PolII at Serine 2 (S2), leading to transcriptional elongation. CDK7 is part of the general transcription factor TFIIH and phosphorylates S5 of the RNA-PolII CTD-regulating translation at its initiation (3).
Specific aspects of malignant transcription demonstrate a marked differential requirement for BET proteins and CDK7 and 9 function in malignant versus normal transcription, limiting the toxicity of small-molecule inhibitors and providing a therapeutic window in cancer. One such observation is that they have a role in regulating large or “super” enhancers that control a number of critical genes (4), including master regulators of cell fate and oncogenes pivotal for the maintenance of leukemia, such as BCL2 and MYC (5, 6). Their downregulation upon treatment with small-molecule inhibitors at least partially explains the effects of BET/CDK inhibition observed in hematologic and solid cancers. Since we and others first reported the preclinical efficacy of BET inhibitors in AML (5, 6), several clinical trials have opened to confirm the efficacy of these drugs in hematologic and solid malignancies (7), and trials are also currently ongoing with CDK7 and CDK9 inhibitors. However, although objective responses have been described in several tumor types, they are usually short-lived. As with all anticancer therapeutics, the development of resistance is to be expected rather than just feared. Genetic mechanisms of resistance to these targeted therapies have been demonstrated (8), and there is growing evidence that nongenetic causes of resistance, related to adaptive transcriptional plasticity, also occur (9). The important work performed by Su and colleagues (10), reported in this issue of Cancer Discovery, addresses resistance to this promising group of targeted therapies. Describing the impact of folate depletion on the therapeutic effects of BET and CDK7 inhibitors, the authors provide evidence that a priori metabolic and transcriptional adaptation can also be a cause of “innate” rather than acquired nongenetic resistance to these classes of drug.
Using gene set enrichment analysis (GSEA) of gene-expression datasets from two large AML patient cohorts, Su and colleagues identify that signatures associated with the folate cycle anticorrelate with active MYC signatures, suggesting a functional link between folate metabolism and the MYC transcriptional program. The authors then go on to demonstrate, in vitro and in vivo, that folic acid depletion reduced the efficacy of BET and CDK7 inhibitors. This relative resistance is mediated through deficiency of the enzyme methylenetetrahydrofolate reductase (MTHFR), whose impairment alters the response to both BET- and CDK7-targeted therapies, with the link phenotypically corroborated by knockdown of BRD4. Using primary AML samples and genome editing to generate isogenic AML cell lines, the authors further link the effects of folic acid deficiency with two common MTHFR homozygous polymorphisms (677 TT and 1,298 CC present in 10% of Caucasians) that cause significant reduction of MTHFR enzymic activity, with affected cells displaying similar resistance to BET inhibition.
Focusing on BET inhibition, Su and colleagues mechanistically link resistance with disruption of the folate cycle through the accumulation of S-adenosylhomocysteine (SAH), an intermediate in homocysteine synthesis, which is a potent inhibitor of S-adenosylmethionine (SAM)–dependent methylation reactions. In particular, the authors show that folic acid deprivation decreases deposition of the transcriptionally repressive H3K9me2 and H3K27me3 histone modification marks and, using an elegant CRISPR/Cas9 screen, identify EDD, a component of the PRC2 complex, and EHMT1 and SETDB1, two other SET-domain-containing H3K9 methyltransferases, as top hits in the screen. They then validate these candidates, demonstrating that their knockdown rescued cells from BET inhibition. Using transcriptomic analysis of AML cell lines, the authors further show that the reduced H3K9me2 and H3K27me3 methylation seen in the presence of folic acid withdrawal drives an adaptive SPI1-mediated transcriptional program, which promotes AML cell survival under the pressure of BET inhibition (Fig. 1).
By invoking metabolic rewiring as a novel mechanism of innate transcriptional resistance to BET/CDK7 inhibitors, the work published by Su and colleagues provides further evidence to inform our understanding of the deep connection between cellular metabolism and epigenetics in cancer and take this into the novel area of (relative) primary drug resistance. Moreover, by demonstrating a link between common MTHFR homozygous genetic variants and BET/CDK7 inhibitor resistance, the authors identify pharmacogenomics as a new and potentially important guide to direct targeted therapy in AML. Pharmacogenomic testing for TPMT and NUDT15 variants, to guide thiopurine dosage and avoid toxicity, in childhood ALL has become routine, and these studies suggest that AML will follow suit to tailor some therapies.
Moreover, beyond genetic variation and drug sensitivity, these findings have other translational implications. The demonstration that folate deficiency leads to a hyperhomocysteinemic state that confers resistance to targeted therapy has profound clinical implications, as not only MTHFR genotype but folate levels can be screened prospectively to predict response to MYC-targeted therapies. The finding that a 4-fold increase in 5-CH3 THF was sufficient to resensitize cells whose MTHFR expression was reduced below 50% to BETi suggests that simple dietary folate supplementation might become a relatively low-risk intervention to improve sensitivity to BET inhibitors; further work is however required to determine what folic acid concentration is needed in susceptible patients. Vitamin B12 and B6 deficiencies, also very prevalent, especially in the elderly population, can lead to hyperhomocysteinemia, as can chronic alcohol consumption and, rarely, homocystinuria. Therefore, there is also a rationale for checking homocysteine levels and, if necessary, replacement of the appropriate vitamin prior to the choice and start of MYC-targeted therapy (Fig. 1).
With regard to specific AML genotypes, Su and colleagues confirm the applicability of this paradigm not only to mixed lineage leukemia–translocated AML but also to the more commonly occurring core binding factor leukemias. The authors, interestingly, also demonstrate that the combination of the BET inhibitor OTX015 and the antimetabolite methotrexate, an inhibitor of DHFR, is antagonistic. This observation has additional clinical relevance in directing combination therapies rationally, especially in lymphomas and acute lymphoblastic leukemia, where BET, CDK7, and CDK9 inhibitors have all shown efficacy and where methotrexate is routinely used in combination regimens.
Overall, the work performed by Su and colleagues strengthens the link between metabolism and epigenetics in cancer therapy and suggests that further a priori knowledge of not only genomics, but also pharmacogenomics and the nutritional state of the patient will assist in the choices of optimal drugs and combinations for individual patients in this era of personalized cancer therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.