A Rapid Diagnostic Assay to Detect a DNA Methylation Signature of Menin Inhibitor Response in Acute Myeloid Leukemia

Acute myeloid leukemia (AML) cells are arrested at early stages of myeloid development through disruption of epigenetic and transcriptional regulators of differentiation. In a subset of patients, epigenetic dysregulation occurs through aberrant activity of the Menin-KMT2A complex leading to increased expression of HOXA9, MEIS1 and several other genes that maintain an undifferentiated state. Altered Menin-KMT2A complex activity is associated with KMT2A rearrangements (KMT2Ar) as well as NPM1 mutations (NPM1^mut). Disruption of the Menin-KMT2A interaction reverses leukemogenic transcriptional programs leading to differentiation and elimination of tumor cells. Multiple small molecule menin inhibitors are in various stages of clinical development and show therapeutic benefits in many patients. Currently, patients are selected for menin inhibitor treatment primarily based on the presence of NPM1^mut or KMT2Ar.

We have previously shown that AML patients separate into 13 distinct categories based on genome-wide DNA methylation patterns termed ‘epitypes’. Most epitypes are highly enriched for recurrent mutations that establish epitype-specific DNA methylation patterns, and NPM1^mut and KMT2Ar are among the most common genetic events in 6/13 epitypes. Interestingly, patients lacking the cardinal epitype mutation but maintaining the associated epitype DNA methylation pattern display similar biological clinical features. Additionally, some patients bear NPM1^mut or KMT2A abnormalities but lack the associated epigenetic reprogramming. These findings highlight the importance of identifying patients that exhibit a DNA methylation pattern consistent with aberrant activity of the Menin-KMT2A complex who may benefit from menin inhibitors regardless of their genetic abnormalities.

Here we describe the development and validation of a rapid, inexpensive, targeted clinical grade assay to determine functional dysregulation of the Menin-KMT2A complex via the DNA methylation changes that result from aberrant activity of the complex. We analyzed genome-wide DNA methylation patterns using Illumina arrays from AML patients from prior studies and categorized patients into Menin-KMT2A complex-dependent (MEN-D) positive and negative based on epitype. We used recursive feature elimination to identify three CpG sites that were most significantly associated with MEN-D+ DNA methylation patterns. These 3 CpGs were located in the CELF2, MIRLET7BHG and CD34 loci. Cross-fold validation analysis showed that the methylation of the three CpGs demonstrated an accuracy of 97% in predicting MEN-D status. A pyrosequencing assay was designed to measure % methylation at these three CpGs. Evaluation of 103 blast-purified AML samples from untreated patients demonstrated high correlation with methylation array (R=0.94, 0.97 & 0.87 for CELF2, CD34 & MIRLET7BHG, respectively). A random forest classifier trained using pyrosequencing data predicted MEN-D status in 101/103 patients. By mixing cell lines with fully methylated and unmethylated loci and mixtures of synthetically methylated DNA, we demonstrated linearity of the pyrosequencing assay and accurate calls down to at least 10% lesional cells. Detection limits were also established by testing NPM1^mut specimens with varying blast content and by mixing purified myeloblasts from MEN-D+ patients with non-involved bone marrow samples. These studies established that the classifier successfully matches the microarray signature with 95-100% confidence down to at least 40% tumor content. Pyrosequencing-derived MEN-D+ signatures were concordant at diagnosis and relapse in 5 NPM1^mut patients.

In summary, we have developed a rapid and clinical grade DNA methylation-based assay that can detect the epigenetic signature associated with menin dependence and thus help select patients for treatment with menin inhibitors. This parsimonious and easy to perform and report assay avoids the complexity of detecting the various chromosome 11q rearrangements that may dysregulate KMT2A or imputing the functional significance of novel/uncommon KMT2A and NPM1 variants and other currently unknown genetic factors. Importantly, this approach expands the population of potential favorable responders by identifying functional dysregulation of the Menin-KMT2A complex that may occur via rare or cryptic genetic events, or potential non-genetic means.