Abstract
Introduction: Whole genome sequencing has demonstrated tremendous heterogeneity in the mutations and chromosomal translocations associated with acute myeloid leukemia (AML), and there are several correlates with prognosis, yet we remain quite limited in our ability to predict specific chemotherapy drug sensitivity based on genomics with the exception of a few selected mutations or translocations, such as FLT3 -ITD or PML-RARA. One third of new diagnosis patients and over half of relapsed patients will not respond to initial chemotherapy regimens that incur appreciable toxicity and result in prolonged hospitalization. We therefore seek to define molecular information that might better predict response to conventional or novel therapies.
Methods: MyAML™ uses next generation sequencing (NGS) to analyze the 3' and 5' UTR and exonic regions of 194 genes and potential genomic breakpoints within known somatic gene fusion breakpoints known to be associated with AML. Fragmented genomic DNA (~3.4Mb) is captured with a customized probe design, and sequenced with 300bp paired end reads on an Illumina MiSeq instrument to an average depth of coverage >1000x. Using a custom bioinformatics pipeline, MyInformatics™, single nucleotide variants (SNVs), insertion/deletions (indels), inversions and translocations are identified, annotated, characterized, and allelic frequencies calculated. Commonly associated variants in dbSNP and 1000 genomes may be eliminated, as well as variants with allele frequencies less than 5%. High throughput drug sensitivity testing was performed against a panel of 160 drugs, of which 56 are FDA approved and 104 are investigational. De-identified samples from 12 patients with de novo AML and 12 patients with relapsed AML were analyzed. For 2 patient samples, Duplex Sequencing was also performed to detect sub-clonal mutations below the detection limit of conventional NGS. Pearson and Spearman correlations were performed between all possible pairs of genes containing missense or indel mutations and the in vitro cytotoxicity response across the same set of 24 patients.
Results: From the 24 patient samples analyzed to date, an average of 129 missense mutations were identified in each sample with an allelic frequency >5%. Of these, an average of over 21 missense variants were observed in COSMIC and less than 3 were novel (not in dbSNP). These samples also contained an average of over 12 coding indels (~5 frameshift and 7 inframe indels per sample). In addition, MyAML™ identified 3 samples with inv(16) and 6 samples with translocations, including the cryptic NUP98-NSD1 t(5;11) that was not detected by karyotyping. For 2 of the samples, Duplex Sequencing was performed at a depth of at least 6000X, and an accuracy of 10-7, and showed concordance of some of the mutations, with each method identifying additional mutations not observed by the other, an expected finding, as each method targeted distinct regions, and Duplex Sequencing had a greater depth of coverage. Fourteen genes were observed to exhibit at least one indel with a frameshift at frequency greater than 5% in more than one patient. In order to identify significantly associated drugs and genes containing indel mutations, we computed Pearson and Spearman correlations between drugs and these 14 genes across 24 patients. The correlation analyses revealed significant associations (p= 0.006 to 0.04) between indel mutations in three genes and chemosensitivity to drugs commonly used in AML such as cladribine, clofarabine, cytarabine, daunorubicin, etoposide, fludarabine and mitoxantrone. Similarly, significant associations (p<0.05) were identified between missense mutations in 5 genes and chemosensitivity to these drugs.
Conclusion: Personalized data derived from a targeted genomic assay and in vitro chemotherapy sensitivity testing of individual patient AML samples will likely lead to innovation in treatment, identification of novel targeted agents, and improved outcomes in AML.
Disclosures
Xie: Invivoscribe: Employment. Carson:Genection: Employment. Patay:Genection: Employment.
Genomic markers of midostaurin drug sensitivity in FLT3 mutated and FLT3 wild-type acute myeloid leukemia patients