DNA methyltransferase inhibitor exposure–response: Challenges and opportunities

Abstract

Although DNA methyltransferase inhibitors (DNMTis), such as azacitidine and decitabine, are used extensively in the treatment of myelodysplastic syndromes and acute myeloid leukemia, there remain unanswered questions about DNMTi's mechanism of action and predictors of clinical response. Because patients often remain on single‐agent DNMTis or DNMTi‐containing regimens for several months before knowing whether clinical benefit can be achieved, the development and clinical validation of response‐predictive biomarkers represents an important unmet need in oncology. In this review, we will summarize the clinical studies that led to the approval of azacitidine and decitabine, as well as the real‐world experience with these drugs. We will then focus on biomarker development for DNMTis—specifically, efforts at determining exposure–response relationships and challenges that remain impacting the broader clinical translation of these methods. We will highlight recent progress in liquid‐chromatography tandem mass spectrometry technology that has allowed for the simultaneous measurement of decitabine genomic incorporation and global DNA methylation, which has significant potential as a mechanism‐of‐action based biomarker in patients on DNMTis. Last, we will cover important research questions that need to be addressed in order to optimize this potential biomarker for clinical use.

INTRODUCTION

Myelodysplastic syndromes

Myelodysplastic syndromes (MDSs) are a heterogeneous group of clonal, acquired hematologic disorders characterized by progressive cytopenias and the potential for transformation to acute myeloid leukemia (AML). AML is defined by the World Health Organization as greater than 20% leukemic cells (blasts) in the bone marrow or blood, although this distinction is often blurred in the setting of high risk MDS, as defined by the Revised International Prognostic Scoring System (R‐IPSS), particularly in the case of MDS with excess blasts (MDS‐EBs). Allogeneic bone marrow transplantation (BMT) is the only known curative treatment for MDS; however, many patients with MDS are not candidates for BMT due to age and/or comorbidities. Unfortunately, despite significant research efforts, therapeutic options have remained limited to supportive care with transfusions, growth factors, and currently four US Food and Drug Administration (FDA)‐approved drugs: lenalidomide, luspatercept, and the DNA methyltransferase inhibitors (DNMTis) azacitidine and decitabine (the latter also approved in an oral combination with the cytidine deaminase inhibitor cedazuridine). ¹

In clinical trials in MDS, response to therapy is typically assessed by monitoring of symptoms, blood counts, and transfusion requirements, in addition to bone marrow examination after two to four cycles of therapy. The International Working Group (IWG) has proposed standardized MDS response criteria for complete remission (CR), partial remission (PR), hematologic improvement (HI), and stable disease. ² There are data, specifically in patients treated with DNMTis, that any objective response, from HI in any blood cell lineage to CR, is associated with both improved overall survival (OS) ³ , ⁴ and quality of life. ⁵

DNA methyltransferase inhibitors

Dysregulation in epigenetic processes as well as mutations in epigenetic modifiers are particularly common among myeloid malignancies. ⁶ Moreover, there is increasing evidence implicating epigenetic changes—particularly DNA hypermethylation and gene silencing—in progression and relapse in MDS and AML. ⁷ As such, there has been considerable effort over the past several decades focused on the development of therapeutics targeting these epigenetic changes. These efforts have resulted in the development of DNMTis, which, of note, are the only class of drugs shown to improve survival in high risk MDS and the related hematological neoplasm chronic myelomonocytic leukemia (CMML). ⁸ , ⁹ , ¹⁰ DNMTis, although not curative, represent a major therapeutic advance in the treatment of patients with high‐risk MDS and CMML as well as those with AML who are ineligible for more intensive treatments, such as induction chemotherapy and allogeneic BMT.

The DNMTis azacitidine and decitabine are analogs of the naturally occurring pyrimidine nucleoside cytidine. Their mechanism of action is concentration‐dependent, with induction of cytotoxicity at high doses ¹¹ and depletion of DNMTs at low doses resulting in DNA hypomethylation and alteration of the gene expression profile. ¹² The precise mechanism responsible for therapy response in myeloid malignancies remains controversial; the predominant theory has centered on DNA hypomethylation and subsequent re‐expression of cancer protective genes, although more recent evidence has invoked immune‐mediated effects via upregulation of endogenous retroviral transcripts. ¹³ Regardless, only half of the patients respond favorably to DNMTi treatment, and virtually all responders eventually relapse. ⁸

In addition to their modest response rates, a major limitation in the clinical management of patients on DNMTis is that several months of treatment are required before therapeutic efficacy can be assessed. For those that do not respond, there is unfortunately a significant opportunity cost, as they often miss out on novel therapies and instead experience a high rate of complications related to severe and prolonged cytopenias. Additionally, few alternative treatments exist for patients who fail to respond to DNMTis, and their prognosis is extremely poor. ¹⁴ Retrospective analysis of the outcomes of 435 patients with higher risk MDS who experienced azacitidine treatment failure demonstrated a median OS of 5.6 months. ¹⁵ A similar analysis of 87 patients with MDS or CMML after decitabine failure demonstrated a median OS of 4.3 months. ¹⁶ In an effort to improve upon the dismal outcomes associated with DNMTi failure, substantial research has focused on identifying specific biomarkers for predicting response and survival in DNMTi‐treated patients. A number of potential biomarkers have been theorized, but none have borne out in clinical studies. ¹⁷ There is a critical need for a response‐predictive biomarker, as this would allow clinicians to use DNMTis more judiciously in this particularly vulnerable patient population, maximizing therapeutic benefit and avoiding unnecessary toxicities and costs.

In this review, we will summarize the clinical studies that led to the approval of azacitidine and decitabine, as well as the real‐world experience with these drugs. We will then focus on progress in biomarker development for DNMTis. Specifically, we will discuss efforts at determining DNMTi exposure–response relationships and will highlight the challenges that remain impacting the broader clinical translation of these methods.

CLINICAL EXPERIENCE WITH DNMTIS

Azacitidine

In 1984, the Cancer and Leukemia Group B began a series of clinical trials evaluating azacitidine for MDS, utilizing a dose of 75 mg/m² s.c. daily for 7 days every 4 weeks. ¹⁸ , ¹⁹ , ²⁰ Results were encouraging, prompting a phase III study comparing azacitidine with supportive care in 191 patients with symptomatic MDS, using the same dosing regimen. In the azacitidine arm, the CR plus PR rate was 23%, compared with 0% in the supportive care arm, and an additional 37% of patients in the azacitidine arm experienced HI. The median time to response was 3 months, although responses were observed as late as cycle 17, and the median duration of response was 15 months. ²⁰ Azacitidine treatment was associated with superior overall survival, which was statistically significant in a 6‐month landmark analysis that controlled for crossover (18 vs. 11 months, p = 0.03). Based on these results, the FDA approved azacitidine in 2004 for the treatment of all MDS subtypes, including those with 20–30% blasts (which is now classified as AML).

In the pivotal phase III clinical trial AZA‐001, 358 patients with higher risk MDS were randomly assigned to either azacitidine (75 mg/m² s.c. daily for 7 days every 4 weeks) or conventional care, which included supportive care, low dose cytarabine, or intensive chemotherapy. ⁸ CR plus PR rates were 29% versus 12%, respectively, and HI rates 49% versus 29%, respectively, in the azacitidine and conventional care arms. Azacitidine also resulted in superior median OS (24.5 vs. 15 months, p = 0.0001). This survival benefit was seen in all prognostic subgroups and age groups, including those over 75 years of age. ²¹ In a secondary analysis of the AZA‐001 study, continued azacitidine treatment after initial response was associated with an objective improvement in response in 48% of patients. ²² Median OS was substantially shorter (13 months) in a subsequent multicenter study of 282 patients with higher risk MDS by Izykson et al. ³ ; however, these patients had, on average, higher risk features, including poor risk cytogenetics, poor performance status, and 28% secondary cases (vs. none in AZA‐001).

Given the above studies, the recommended starting dose for the first treatment cycle of azacitidine is 75 mg/m² daily for 7 days every 4 weeks. The dose can be adjusted in subsequent cycles based on hematologic response and toxicity. Alternative dosing regimens have been evaluated; however, because these studies have consisted primarily of patients with lower risk MDS, it remains unknown whether these regimens yield comparable efficacy as the approved schedule. ²³

Decitabine

The first clinical studies of decitabine in hematologic malignancies used doses of 1500 to 2500 mg/m² per course and noted promising activity. ²⁴ , ²⁵ However, these high‐dose regimens were not pursued because of severe and prolonged myelosuppression. ²⁵ Lower “epigenetic” doses (15 mg/m² 3 times daily for 3 days) demonstrated encouraging activity in MDS. ²⁶ The first phase I trial in leukemia evaluated decitabine doses ranging from 5 to 20 mg/m² daily for 10–20 days per cycle, repeated every 6 weeks. ²⁷ In this study, 15 mg/m² × 10 days appeared to be the optimal dose and schedule, with 11 of 17 patients (65%) achieving a response. A follow‐up phase II trial of decitabine in MDS randomized 95 patients to receive 10 mg/m² i.v. daily for 10 days, 20 mg/m² i.v. daily for 5 days, or 20 mg/m² s.q. for 5 days. ²⁸ Cycles were administered every 4 weeks and response evaluated after at least three cycles. Response rates were superior in the 20 mg/m² i.v. × 5 days arm, with an overall response rate (ORR; CR + PR + HI) of 73% and CR rate of 39%. Notably, this dose schedule also resulted in superior hypomethylation, measured by LINE‐1 bisulfite pyrosequencing, compared with other arms.

The encouraging phase II results prompted a multi‐institutional phase III trial in the United States, which randomized 170 patients with intermediate or high risk MDS or CMML to receive decitabine at 15 mg/m² every 8 h for 3 days repeated every 6 weeks, or best supportive care. ²⁹ Decitabine resulted in an CR plus PR rate of 17% and ORR of 30%, compared with 0% and 7%, respectively, in the supportive care arm. There was no difference in OS between the two arms in the primary analysis; however, survival was significantly improved in responders compared with nonresponders (23.5 vs. 13.7 months, p = 0.007). Based on these results, the FDA approved decitabine for the treatment of intermediate and high‐risk MDS and CMML in 2006. A follow‐up phase III multicenter trial comparing the same dose and schedule of decitabine with supportive care in patients greater than 60 years of age with high risk MDS also failed to show a statistically significant improvement in overall survival (10.1 vs. 8.5 months), although progression‐free survival and AML‐free survival were significantly prolonged with decitabine. ¹⁰

Subsequent studies attempted to optimize the dosing and schedule of decitabine for MDS and CMML in order to allow for outpatient administration akin to azacitidine. The phase II ADOPT study evaluated decitabine 20 mg/m² i.v. × 5 days given in 4‐week cycles in patients with MDS and IPSS score greater than or equal to 0.5. ⁹ The CR rate was 17%, ORR 51%, and median OS 19.4 months. This 5‐day regimen was subsequently approved by the FDA given similar efficacy. This regimen is occasionally extended to 10 days in patients with unfavorable cytogenetics and/or TP53 mutations, based on Welch et al. ³⁰ There remains a lack of consensus on optimal dose and schedule for pharmacodynamic effect and clinical response.

Oral DNMTi formulations

There has been much interest in the development of oral DNMTi formulations, which have the potential to decrease the burden of treatment and to thereby improve quality of life and even efficacy. The novel oral DNMTi ASTX727 consists of a fixed dose combination of decitabine at 35 mg and the cytidine deaminase (CDA) inhibitor cedazuridine at 100 mg, which limits CDA‐mediated degradation of decitabine in the gastrointestinal tract, increasing its bioavailability. Phase I–III studies with this agent have demonstrated equivalent systemic exposure as well as similar pharmacodynamic effects and clinical efficacy between ASTX727 and i.v. decitabine. ³¹ ASTX727 was thus approved by the FDA in 2020 for the treatment of higher risk MDS and CMML.

The oral azacitidine formulation CC‐486 has also been investigated in different clinical settings. In the phase III randomized QUAZAR AML‐001 trial, CC‐486 at a dose of 300 mg daily on days 1–14 of 28‐day cycles significantly improved OS and relapse‐free survival in older patients with AML who were in first remission after intensive chemotherapy and not candidates for allogeneic BMT. ³² It is currently FDA‐approved for this indication but not for patients with MDS. The same dose and schedule yielded encouraging results in a phase I–II study in patients with MDS and AML in remission after BMT ³³ ; this concept is being further evaluated in the phase III AMADEUS trial (NCT04173533). CC‐486 was also efficacious in a phase III study in low‐risk MDS with transfusion‐dependent anemia and thrombocytopenia. ³⁴ Oral azacitidine is currently under study in combination with cedazuridine, named ASTX030, in combined phase I–III trials modeled after those for ASTX727 (NCT04256317).

DNMTi‐based combinations

In order to improve the efficacy of DNMTis in higher risk MDS, many combinations with other classes of agents have been explored. Although several such combinations have appeared promising in phase II trials, none have demonstrated a statistically significant OS advantage in randomized trials in MDS. ³⁵ For older patients with AML considered unfit for intensive chemotherapy, however, the combination of the Bcl‐2 inhibitor venetoclax with either azacitidine or decitabine demonstrated response rates and OS substantially higher than historical expectations with DNMTi monotherapy (CR/CR with incomplete hematologic recovery [CRi] rate 67%; median OS 17.5 months) in a single‐arm study. ³⁶ This study led to the accelerated FDA approval of this combination for older/unfit patients with newly diagnosed AML in 2018. These data were eventually confirmed in the VIALE‐A study, which randomized patients with newly diagnosed AML who were ineligible for intensive induction to azacitidine plus either venetoclax or placebo. ³⁷ The combination was associated with significantly higher rates of CR/CRi (66.4% vs. 28.3%), which translated to significantly improved median OS (14.7 vs. 9.6 months, p < 0.001). For these reasons, the combination of DNMTi plus venetoclax has become standard of care for older/unfit newly diagnosed AML.

Combinations of DNMTis with targeted agents have also been evaluated in molecularly defined subsets of AML in recent years. In older patients with AML harboring mutations in FMS‐like tyrosine kinase 3 (FLT3) who are unfit for induction chemotherapy, the combination of azacitidine and sorafenib was shown to be efficacious with ORR of 78% and median OS of 8.3 months, and was well‐tolerated. ³⁸ An ongoing trial is evaluating the safety and tolerability of azacitidine combined with the second generation FLT3 inhibitor gilteritinib in the upfront setting (NCT02752035). In the recently published phase III AGILE study, ³⁹ azacitidine combined with the isocitrate dehydrogenase 1 (IDH1) inhibitor ivosidenib resulted in a significant OS advantage over azacitidine alone in newly diagnosed patients with IDH1‐mutated AML ineligible for induction chemotherapy (24.0 vs. 7.9 months, respectively, p = 0.001). Of note, these DNMTi‐based combinations have yet to be directly compared with DNMTi plus venetoclax in a randomized fashion.

DNMTis in solid tumors

Despite well‐demonstrated activity in myeloid malignancies, the use of azacitidine in solid tumors has been largely limited by toxicity and myelosuppression, as well as low CR and PR rates. ⁴⁰ , ⁴¹ In the 2000s, decitabine monotherapy also yielded dismal results for patients with solid tumors, ⁴² suggesting an as‐yet‐unexplained mechanism of primary resistance. A number of small clinical studies have since evaluated DNMTis in combination with various chemotherapeutic and immunotherapeutic drugs. In this context, DNMTis have been shown to restore chemosensitivity, particularly to platinums, ⁴³ as well as to reverse resistance to checkpoint blockade. ⁴⁴ However, these results await confirmation in larger, randomized studies before this therapeutic principle can be used.

MECHANISM OF ACTION

Azacitidine and decitabine are analogs of cytidine and deoxycytidine, respectively; in both, carbon 5 of the pyrimidine ring is replaced by a nitrogen atom. ⁴⁵ Cellular uptake of both DNMTis is mediated by human nucleoside transporters, as detailed below. Azacitidine is then phosphorylated to 5‐azacytidine monophosphate (aza‐CMP) by uridine‐cytidine kinase and then to 5‐azacytidine diphosphate (aza‐CDP) and 5‐azacytidine triphosphate (aza‐CTP) by UMP‐CMP kinase and nucleoside diphosphate kinase, respectively (Figure 1). ⁴⁵ Aza‐CTP is primarily incorporated into RNA. ⁴⁶ A small fraction, about 10%–20%, of the aza‐CDP intermediate, however, is converted to 5‐aza‐2′‐deoxycytidine diphosphate (aza‐dCDP) via the enzyme ribonucleotide reductase, followed by further phosphorylation to 5‐aza‐2′‐deoxycytidine triphosphate (aza‐dCTP). ⁴⁶ Decitabine is, after uptake in the cell, phosphorylated to 5‐aza‐2′‐deoxycytidine monophosphate (aza‐dCMP) by deoxycytidine kinase, and is then further phosphorylated to aza‐dCDP and aza‐dCTP by deoxycytidine monophosphokinase (dCMP kinase) and nucleoside diphosphate kinase, respectively (Figure 1). ¹¹ Both azacitidine and decitabine are rapidly eliminated by deamination by CDA, resulting in half‐lives of ~35–40 min. ⁴⁵

FIGURE 1.

Membrane transport, metabolism, and mechanism of action of azacitidine and decitabine. Azacitidine and decitabine enter the cells via concentrative nucleoside transporters (SLC 28A or 22A). The equilibrative nucleoside transporter (SLC 29A) mediate the entrance or export depending on the concentration gradient. Once inside the cell, azacitidine (5‐AZA) and decitabine (5‐AZA‐dC) are phosphorylated by a series of kinases, ultimately to 5‐azacytidine triphosphate (5‐AZA‐CTP) and 5‐aza‐2′‐deoxycytidine triphosphate (5‐AZA‐dCTP). Both 5‐AZA and 5‐AZA‐dC undergo deamination by cytidine deaminase (CDA). Ten to 20% of 5‐azacytidine diphosphate (5‐AZA‐CDP) is converted to 5‐aza‐2′‐deoxycytidine diphosphate (5‐AZA‐dCDP) by ribonucleotide reductase. 5‐AZA‐CTP is incorporated into RNA, leading to protein biosynthesis disruption and apoptosis. 5‐AZA‐dCTP is incorporated into DNA, leading to trapping and degradation of DNA methyltransferase (DNMT) enzymes, passive loss of DNA methylation, and ultimately differentiation and expression of tumor suppressive genes (TSGs). Particularly when DNA methyltransferase inhibitors (DNMTis) are given at higher doses, DNA incorporation and trapping of DNMT enzymes can lead to DNA damage and apoptosis. Created with BioRender.com.

The intracellularly formed triphosphates aza‐CTP and aza‐dCTP are responsible for the downstream pharmacological effects of azacitidine and decitabine. Aza‐CTP is incorporated into RNA, disrupting nuclear and cytoplasmic RNA metabolism, inhibiting protein synthesis, and ultimately leading to apoptosis. ⁴⁶ Incorporation of aza‐CTP into RNA also inhibits ribonucleotide reductase, leading to a reduced deoxyribonucleotide pool and thereby limiting DNA synthesis and repair. ⁴⁶ Aza‐dCTP competes with endogenous dCTP for DNA incorporation, leading to trapping and degradation of DNMT enzymes and passive loss of DNA methylation. ¹¹ , ¹⁴ DNMTis are also associated with cellular differentiation and cytotoxicity at higher doses. ⁴⁷ Although their mechanism of action is complex and certain elements have not been fully elucidated, particularly the biological consequences of azacitidine incorporation into RNA, ⁴⁸ DNMTis are thought to exert their antineoplastic activity predominantly via DNA incorporation, DNMT inhibition, and resultant DNA demethylation and re‐expression of cancer protective genes. ⁴⁷

Role of drug transporters in the disposition of DNMTis

Cellular transport of DNMTis across cell membranes is a crucial step prior to the purported mechanism of action involving incorporation into nucleic acids and is thus an important determinant of efficacy as well as resistance. Nucleoside analogs are known to be transported in the cells by two solute carrier families: SLC28, the concentrative nucleoside transporter (CNT) family, or SLC29, the equilibrative nucleoside transporter (ENT) family. ⁴⁹ , ⁵⁰ In one study, all human nucleoside transporters (NTs) were shown to facilitate the transport of azacitidine, whereas only ENT1/2 and CNT1/3 showed weak or poor transport of decitabine ⁵¹ ; the cytotoxicity of both azacitidine and decitabine is dependent on the presence of various NTs. ⁵¹ , ⁵² , ⁵³ Rius et al. identified azacitidine as a substrate of CNT1/2 and showed that expression of CNT1 in MDCK cells strongly enhances DNA demethylation induced by both azacitidine and decitabine. ⁴⁹ , ⁵⁴ In another study, decitabine was mainly transported by ENT1, and the uptake of decitabine was one of the key determinants of the activity of decitabine in the human colon cancer cell line HCT116. ⁵² In addition, knockdown of ENT1 in the leukemic cell line SKM‐1 weakened the hypomethylating effect of decitabine in vitro. ⁵⁵ In support of azacitidine transport by ENT1, a positive correlation was observed between ENT1 expression and the potency of azacitidine using an array of 60 human cancer cell lines. ⁵³ Furthermore, using an SLC‐focused CRISPR/Cas9 library in a haploid human cell line, loss of ENT1 induced resistance to both azacitidine and guadecitabine, a “next generation” dinucleotide decitabine analog. ⁵⁶ , ⁵⁷ Recently, the organic cation transporter SLC22A4 (OCTN1) was identified as a high‐affinity carrier of nucleoside analogs in vitro, including azacitidine and decitabine. ⁵⁸ Of note, the transporters responsible for efflux of DNMTis have been studied to a lesser extent, and we were unable to find studies indicating DNMTis as substrates of ATP‐binding cassette (ABC) transporters. ⁵⁹ However, studies have shown that modification of ABC transporters by DNMTis may enhance or reduce sensitivity to anticancer drugs. ⁶⁰ , ⁶¹ , ⁶² Taken together, these results indicate that cellular accumulation of DNMTis is likely primarily mediated by ENTs and CNTs, although other transporters may also be responsible for their movement across intracellular membranes.

Several studies have investigated the correlation between response to DNMTis and membrane transporter expression, with varying results. An ex vivo study in leukemic blasts obtained from 50 patients with AML at diagnosis demonstrated that increased ENT1 mRNA expression directly correlated with increased in vitro sensitivity to decitabine. ⁶³ In addition, high expression of ENT1 was associated with prolonged survival in higher‐risk patients with MDS treated with decitabine. ⁶⁴ Whereas, another group reported there was no significant difference in mRNA expression of ENT1, ENT2, or CNT3 between diagnosis and relapse in patients treated with decitabine. ⁶⁵ Moreover, no correlation between expression levels of ENT1 in responders versus nonresponders was found in a study evaluating mRNA from 57 patients with MDS treated with azacitidine. ⁶⁶ Further validation of the correlation between response to DNMTis and transporter expression is warranted.

RESPONSE PREDICTION

Given the limited number of patients who benefit from DNMTis, as well as the requirement for long‐term treatment on a regular and uninterrupted schedule for optimal outcomes, identifying biomarkers that can predict response versus resistance is essential. For patients with higher risk MDS, several clinical factors have been shown to be associated with superior response to DNMTis, including lack of prior therapy (cytarabine in particular), shorter MDS duration, bone marrow blast percentage less than 15%, normal karyotype, and platelet doubling time after the first cycle of azacitidine. ³ , ⁶⁷ , ⁶⁸ Izykson et al. ³ proposed a four‐category prognostic scoring system for OS in DNMTi treated patients, which assigned points for performance status, cytogenetics, circulating blasts, and transfusion dependence. In a multicenter cohort of 282 patients with higher risk MDS treated with azacitidine, patients' composite scores discriminated three groups with median OS times of 6, 15, and greater than 26 months. ³

Given the incorporation of molecular mutation analysis into MDS prognostic models, studies have increasingly focused on genomic biomarkers predictive of DNMTi response. Mutations in TET2, which facilitates cytosine demethylation, has been associated with a higher likelihood of response to DNMTis in MDS ³ , ⁶⁹ , ⁷⁰ and CMML. ⁷¹ Mutations in other DNA methylation regulators DNMT3A ⁷⁰ and IDH1/2 ⁷² may also predict response in both MDS and AML. The impact of TP53 mutations on response to DNMTis, particularly when dose intensive regimens are used, remains controversial. ³⁰ , ⁷³ , ⁷⁴ In a more recent study, Nazha et al. ⁷⁵ used machine learning to identify highly sensitive genomic associations which predicted resistance to DNMTis with high accuracy in an independent patient cohort enrolled on a randomized clinical trial. Importantly, these genomic biomarkers were only present in ~25% of patients, suggesting that other biologic mechanisms contribute to therapy response and resistance. ⁷⁵

Several studies have also evaluated the influence of DNA methylation patterns on treatment outcomes, with varied results. Shen et al. ⁷⁶ analyzed promotor methylation in a panel of 10 candidate genes in 317 patients with MDS treated with decitabine, and found no association between baseline methylation and treatment response. However, they did note a significant decrease in methylation in responders compared with nonresponders. ⁷⁶ A similar association between decrease in methylation of specific promoters and treatment response has been observed in other studies. ⁷⁷ , ⁷⁸ It is important to note that the decreases in methylation observed in these studies occurred after a minimum of one treatment cycle, and typically required several cycles. In a study by Mund et al., ⁷⁷ karyotype normalization preceded changes in promotor methylation in three patients with MDS treated with decitabine, indicating that methylation changes likely occurred in normal cells and thus may not have represented the pharmacodynamic effect of decitabine.

Additional studies have investigated the global effect of DNMTis on DNA methylation using array‐based approaches, ⁷⁹ , ⁸⁰ with one study ⁷⁹ noting a demethylating effect persisting 2 weeks after therapy cessation and returning to baseline prior to subsequent cycles. However, these methods have limited applicability in the clinical setting owing to long run times and low sensitivity. Newer mechanism‐of‐action based biomarkers with improved sensitivity and more robust clinical utility are thus desperately needed.

EXPOSURE–RESPONSE RELATIONSHIPS

Despite the convergent effects of DNMTis—namely, DNA incorporation, DNMT inhibition, DNA demethylation, and alteration in gene expression profile—there have been significant analytical challenges over the years in quantitatively demonstrating exposure–response relationships. ⁸¹ , ⁸² , ⁸³ In particular, bioanalytical methods for quantitating DNMTi pharmacokinetics—both systemically and intracellularly—have been impeded by chemical instability of the parent compounds requiring efficient pre‐analytical processing methods. ⁸² , ⁸³ , ⁸⁴ Quantitation of intracellular DNMTi metabolites as well as DNMTi DNA incorporation have been further complicated by the similar molecular weight of decitabine and the natural DNA nucleoside 2′‐deoxycytidine (2dC; 228.2 vs. 227.2), necessitating robust chromatographic resolution. ⁸⁵ Additionally, the detection of decitabine in the presence of abundant endogenous interferences requires a highly sensitive mass spectrometer. ⁸⁵ The sensitivity requirement is further increased in patients receiving low‐dose DNMTi regimens, which are essentially universal in the modern era.

Over the years, many groups have sought to develop bioanalytical methods that not only overcome the above‐mentioned technical challenges, but that are also clinically applicable. Earlier methods for quantitating DNMTis in plasma were limited in clinical applicability, largely owing to chemical instability. In more recent years, improvements in liquid‐chromatography tandem mass spectrometry (LC–MS/MS) technology have allowed for the quantitation of intracellular DNMTi metabolites, and ultimately DNMTi incorporation into DNA, which is the critical convergent step required for their pharmacodynamic effect. Here, we will review the progression of these methods, the challenges that have been overcome, and the challenges that still remain in elucidating DNMTi exposure–response.

Systemic pharmacokinetics

The earliest methods for quantitating DNMTis in plasma utilized microbiologic assays. ⁸⁶ These assays were limited by both instability of the parent drug and lack of specificity due to degradation products having residual cytostatic properties. These limitations ultimately led to a shift to chromatography‐based methods.

Azacitidine has also been quantitated in plasma using high‐performance liquid‐chromatography (HPLC)‐UV, ⁸² , ⁸⁷ , ⁸⁸ with assay sensitivity ranging from 0.05 μg/mL ⁸⁷ to 0.25 μg/mL. ⁸⁸ Rustum et al. ⁸⁸ noted that azacitidine was unstable in plasma with a 20% loss by 4.5 days when stored at −60°C and a 10% loss within 0.5 h at room temperature. Drug instability, high sample volume, lack of sensitivity, and long run times were the main limiting factors for the application of HPLC‐based methods in pharmacokinetic characterization.

In order to characterize the pharmacokinetics of azacitidine utilizing low‐dose, daily administration schedules, a more sensitive and specific LC–MS/MS method was developed and validated by our laboratory to quantitate azacitidine as low as 5 ng/mL. With this method, plasma concentration‐time profiles were obtained; however, there was significant interpatient variability in exposure, which was attributed, in part, to the instability of azacitidine in plasma. ⁸⁹ To account for variability as a result of potential degradation, the cytidine deaminase inhibitor tetrahydrouridine (THU) was added to plasma at 100 μM which increased freezer stability from 7 days to ~21 days. However, azacitidine remained unstable in plasma at room temperature despite the addition of THU, with 40% loss by 1 h. ⁸⁹

Our validated LC–MS/MS method was successfully used to characterize the pharmacokinetics of azacitidine administered in combination with the histone deacetylase inhibitor phenylbutyrate in two phase I trials at our institution. ⁹⁰ We have since further optimized the method, including streamlining sample extraction and shortening the run time. ⁹¹

Liu et al. ⁸³ reported an LC–MS/MS method for the quantitation of decitabine in plasma, and also found that the drug was highly unstable. In order to accurately quantitate decitabine, samples needed to be processed for extraction quickly and at 4°C. Liu et al.'s ⁹² method was applied in a phase I study of decitabine alone or in combination with valproic acid in AML, and the decitabine plasma level did not correlate with disease response. THU was found to increase the stability of decitabine, eliminating the need to carry out extractions at 4°C. ⁹³

Quantitation of phosphorylated DNMTi metabolites

Given the possibility of multiple upstream mechanisms of DNMTi resistance (decreased nucleoside transport via ENT downregulation and/or modified metabolism via dCK downregulation or CDA upregulation), several groups postulated that quantitation of intracellular aza‐CTP and aza‐dCTP by LC–MS/MS may correlate more predictably with pharmacological activity. ⁹⁴ , ⁹⁵ , ⁹⁶ Moreover, an understanding of intracellular pharmacokinetics was felt to be crucial for designing more optimal DNMTi dosing regimens.

Jansen et al. ⁹⁵ measured the intracellular aza‐dCMP, aza‐dCDP, and aza‐dCTP concentrations in peripheral blood mononuclear cells (PBMCs) obtained from three patients with MDS treated with decitabine 15 mg/m² administered as a 4‐h infusion, repeated every 8 h for 3 days with a cycle duration of 6 weeks. Although this was a small study, they found that the two patients who responded well to decitabine therapy had higher intracellular aza‐dCTP concentrations at the end of the treatment cycle than the one patient who did not respond.

Wang et al. measured intracellular aza‐dCTP concentrations in bone marrow cells and PBMCs from seven patients with AML who were treated with decitabine 20 mg/m² infused over 1 h for 10 days. ⁹⁶ Samples were collected post infusion on days 1 and 5. Their data suggested that intracellular decitabine‐TP concentration is relatively steady through the treatment course. Higher intracellular aza‐dCTP concentrations in both bone marrow and PBMCs also appeared to be associated with clinical response; however, a formal statistical analysis could not be performed given the small number of patients.

Last, Derissen et al. ⁹⁴ developed an assay to quantify aza‐CTP in PBMCs, which required high resolution mass spectrometry to distinguish aza‐CTP from endogenous nucleotides. The assay was applied to quantify aza‐CTP in PBMCs from two patients with MDS who were treated with azacitidine 75 mg/m² s.c. for 7 days, every 4 weeks. The aza‐CTP concentrations were significantly (2.5–55 times) higher than aza‐dCTP concentrations published by Jansen et al., ⁹⁵ which they felt was reflective of the known difference between endogenous ribonucleotide and deoxyribonucleotide concentrations.

Although the Jansen et al. and the Wang et al. ⁹⁵ , ⁹⁶ studies shed light on a possible relationship between intracellular aza‐dCTP concentrations and clinical response to decitabine therapy, these studies involved small numbers of patients, and there have not been any additional studies published that have further investigated this correlation – either for decitabine or azacitidine. Moreover, quantitation of phosphorylated DNMTi metabolites by LC–MS/MS is limited by low sensitivity necessitating prohibitively large sample volumes as well as interference from endogenous 2dC‐TP. ⁸¹ The intracellular aza‐CTP and aza‐dCTP concentrations required for optimal downstream effects, including DNA incorporation and demethylation, and ultimately clinical response, thus remains unknown.

Decitabine genomic incorporation

Given that nucleoside analog DNMTis share a mechanism for their epigenetic and antineoplastic effects that converges on incorporation of aza‐dCTP into DNA and resultant DNA demethylation, quantitative measurement of DNA incorporation could facilitate a deeper understanding of exposure–response parameters of these drugs. It has also been hypothesized that the degree of decitabine DNA incorporation could be correlated with drug efficacy and clinical response, and that decitabine DNA incorporation could therefore represent a mechanism‐based pharmacodynamic and response‐predictive biomarker. ⁴⁸ Such a biomarker may be even more useful than measurement of DNA demethylation globally or at specific genes, because this can be influenced by a variety of factors.

To date, bioanalytical methods for measuring decitabine genomic incorporation have been sparse, likely due to instability of the parent drug and endogenous molecules that cause interferences, as noted previously. ⁸² , ⁸³ , ⁸⁴ , ⁹⁷ Furthermore, given the polarity of both decitabine and 2dC, commonly used reverse phase chromatography methods have not proven sufficient to retain or separate the analytes. ⁸⁵ Last, adequate detection of decitabine DNA incorporation in the setting of low‐dose regimens and endogenous interferences requires a highly sensitive mass spectrometer, such as a triple quadrupole in addition to technically sophisticated methods of improving the signal to noise ratio such as with differential ion mobility separation or a high resolution mass spectrometer.

The first assay described explored incorporation of ³H decitabine in leukemia cell lines and patient samples as a potential phenotypic probe for therapeutic efficacy. ⁹⁸ The total amount of intracellular decitabine (including aza‐dCMP, ‐dCDP, and ‐dCTP) in cell lines and in primary bone marrow cells from patients with AML was measured and intracellular drug levels and DNA incorporation were correlated. However, their results suggested considerable interpatient variability and were limited in clinical translatability given the use of radiolabeling.

Our laboratory subsequently developed a robust pilot LC–MS/MS method that simultaneously measures decitabine incorporation into DNA (decitabine/2‐deoxycytidine) and global DNA methylation (5‐methyl‐deoxycytidine/2‐deoxycytidine). ⁸⁵ This method is significantly simpler pre‐analytically than the above‐referenced methods quantitating intracellular triphosphorylated DNMTi metabolites and is thus amenable to clinical translation. A Hypercarb porous graphite column provided adequate separation with minimal matrix effects, and the method was determined to have sufficient selectivity from endogenous interferences. The method was applied to show decitabine incorporation into genomic DNA in cancer cell lines and in tissue samples collected from mice treated with decitabine. ⁸⁵ However, this initial assay was somewhat limited in its analytical sensitivity, requiring in the range of 2–5 μg of genomic DNA to detect DNMTi incorporation at relevant doses, thus precluding implementation in a clinical setting. We have since further optimized the LC–MS/MS method, demonstrating 125‐fold net improvement in sensitivity and allowing for robust measurement of DNMTi incorporation from as little as 8 ng of genomic DNA after exposure to relevant doses of DNMTi in cell culture experiments.

Using similar LC–MS/MS methods, several groups have shown positive correlations between decitabine genomic incorporation and clinical response in patients with MDS and AML receiving single agent DNMTi therapy. ⁴⁸ , ⁸¹ , ⁹⁹ Unnikrishnan et al. ⁴⁸ analyzed bone marrow specimens from eight patients with MDS receiving their first cycle of treatment with azacitidine, and found increased decitabine incorporation at day 8 in responders compared to nonresponders. Roosendaal et al. ⁸¹ assessed the degree of decitabine incorporation and global methylation in DNA isolated from whole blood from four patients receiving guadecitabine, a novel nucleoside analog DNMTi under clinical evaluation. They found decitabine incorporation at day 6 of Cycle 1 but did not observe changes in methylation. More recently, Chilakala et al. ⁹⁹ analyzed peripheral blood mononuclear cells from five patients with MDS after 6 weeks of twice weekly decitabine, and observed both increased decitabine incorporation and DNA demethylation among responders.

Notably, the above‐mentioned studies have involved small numbers of patients, ranging from four to eight, and there have not been any additional studies published correlating decitabine genomic incorporation with clinical response, and no such studies in the clinical trial setting. Further validation of this correlation in a larger cohort of patients is required and represents a key knowledge gap in both the clinical pharmacology of DNMTis and, more broadly, in the treatment of hematologic malignancies.

FUTURE DIRECTIONS

Given the above referenced small clinical studies, ⁴⁸ , ⁸¹ , ⁹⁹ as well as preliminary (unpublished) preclinical and clinical data from our institution, we believe that simultaneous measurement of decitabine genomic incorporation and global DNA methylation by LC–MS/MS has significant potential as a mechanism‐of‐action based pharmacodynamic and predictive biomarker in patients on DNMTis. The availability of such a biomarker would have tremendous impact on the treatment of hematologic malignancies, where DNMTis are routinely used as standard‐of‐care despite a relatively low response rate and an evolving understanding of determinants of response versus resistance. ⁷⁵ , ¹⁰⁰ However, additional research is needed to advance the clinical translation of this biomarker; specifically, the following questions need to be addressed.

What is the optimal timing of clinical samples of patients on DNMTis for determination of genomic incorporation and global DNA methylation by LC–MS/MS?

Moving forward, it will be key to address the optimal timing of clinical samples of patients on DNMTis for determination of genomic incorporation and global DNA methylation by LC–MS/MS. The clinical studies by Unnikrishnan, ⁴⁸ Rosendaal, ⁸¹ and Chilakala ⁹⁹ used different dosing schema as well as different timepoints for sample collection, making it difficult to compare the dynamics of decitabine incorporation and DNA demethylation, as well as the clinical correlations.

We believe that additional in vitro research should include an analysis of the kinetics of decitabine genomic incorporation relative to alterations in global DNA methylation. For example, it may be that peak genomic incorporation occurs within 1–2 days of exposure but becomes undetectable after DNMTis are removed from the media after 1 week. However, it is possible that peak DNA demethylation requires 1 week or longer of continuous treatment. Simultaneously measuring genomic incorporation and DNA methylation over time in a variety of leukemia and MDS cell lines will provide invaluable insights into pharmacokinetic‐pharmacodynamic relationships and will guide more precise timing of sample collection, both in clinical trial and, eventually, in standard‐of‐care settings.

What is the impact of co‐administered drugs on decitabine genomic incorporation?

Another important research topic is the potential for co‐administered drugs to impact decitabine genomic incorporation. Given the limited activity of DNMTis as single agents in MDS and AML, there have been a multitude of clinical trials investigating various combinations, including DNMTis. The combination of DNMTis with the Bcl‐2 inhibitor venetoclax has been the most efficacious example to date, and has emerged as the standard of care in older and unfit patients with AML based upon results from the pivotal VIALE‐A clinical trial. ³⁷ Despite this, the precise mechanism of DNMTi and venetoclax synergy has not been fully elucidated, although there have been some theories postulated. ³⁷ In particular, it is unknown whether venetoclax or novel agents being studied in combination with DNMTis could potentially impact decitabine genomic incorporation, either through affecting the cell cycle or through other unanticipated mechanisms. This question can be addressed through in vitro experiments comparing decitabine genomic incorporation following treatment with DNMTis alone versus in various combination regimens.

Does decitabine genomic incorporation correlate with clinical response in patients on DNMTis?

Last, in order to establish decitabine genomic incorporation as a potential biomarker in patients receiving DNMTis, it will be crucial to validate the correlation between incorporation and clinical response in a larger cohort of patients. Ideally, this would be done in a clinical trial setting, because this would ensure a relatively homogenous patient population. Establishment of decitabine genomic incorporation as a pharmacodynamic and response‐predictive biomarker would allow clinicians to identify patients who are unlikely to respond to DNMTi containing regimens early in their treatment course, thereby minimizing unnecessary toxicities and costs. This mass spectrometry‐based biomarker thus has the potential to significantly improve the treatment of hematologic malignancies.

CONCLUSION

Despite their widespread use in myeloid malignancies, much is not understood about DNMTi exposure–response relationships or predictors of clinical response. A multitude of analytical methods have been developed over the years to quantitate these drugs, both systemically and intracellularly; however, the vast majority have not been clinically translatable. Our group and others have shown, both in preclinical experiments and in small clinical studies, that it is feasible to simultaneously measure both decitabine genomic incorporation and global DNA methylation using LC–MS/MS. Limited clinical data thus far indicates that this method represents a potential pharmacodynamic and response‐predictive biomarker; however, additional research is needed to ensure that it is optimized for clinical use. We are hopeful that this potential biomarker will improve our ability to individualize care delivery in patients with myeloid malignancies and ultimately in all patients receiving DNMTis.

FUNDING INFORMATION

This work was supported in part by the National Institute of Health grants: UM1CA186691 (M.A.R. and J.A.W.), U24CA247648 (M.A.R., N.M.A., and S.D.B.), P30CA006973 (M.A.R. and S.Y.), T32GM066691 (A.B.K. and D.A.G.), UL1TR003098 (M.A.R.), the OSU Comprehensive Cancer Center Pelotonia foundation (S.D.B.), and the Commonwealth Foundation (M.A.R. and S.Y.).

CONFLICT OF INTEREST STATEMENT

Nicole M. Anders, Srinivasan Yegnasubramanian, and Michelle A. Rudek are inventors on the patent for Quantitative Determination of Nucleoside Analogue Drugs in Genomic DNA or RNA (US Patent 11,035,850 B2; expiration April 10, 2037). Amanda B. Kagan, Dominique A. Garrison, Nicole M. Anders, Jonathan A. Webster, and Sharyn D. Baker declare no competing interests for this work.

Kagan AB, Garrison DA, Anders NM, et al. DNA methyltransferase inhibitor exposure–response: Challenges and opportunities. Clin Transl Sci. 2023;16:1309‐1322. doi: 10.1111/cts.13548