Designing Myeloid Gene Panels: Evaluation of the Association for Molecular Pathology–Recommended Core 34-Gene Panel for Next-Generation Sequencing in Chronic Myeloid Neoplasms

We conducted a retrospective review of patients who underwent NGS testing at the Cleveland Clinic between June 2018 and December 2018 using a 62-gene panel that included all 34 AMP core genes and an additional 28 genes. Patients receiving a final diagnosis using current WHO criteria¹  of a CMN, clonal hematopoiesis of indeterminate potential (CHIP), or clonal cytopenia(s) of undetermined significance (CCUS) based on pathology review at the Cleveland Clinic were selected for inclusion. The Cleveland Clinic institutional review board approved this retrospective study.

DNA extracted from peripheral blood or bone marrow samples was subjected to nested multiplex polymerase chain reaction–based target enrichment. Targeted regions of 62 genes (Table 1) were amplified and sequenced on an Illumina instrument (San Diego, California) with paired-end 150 × 2 cycle reads. A customized bioinformatic analytical pipeline was used to map reads to the reference human genome (Genomic Build GRCh37/hg19). The limit of detection for this assay has been established to be 1% variant allele fraction for the JAK2 V617F point mutation and 5% variant allele fraction for other single-nucleotide variants and insertion/deletion mutations. Variants are classified according to the AMP/American Society of Clinical Oncology/College of American Pathologists guidelines.¹³  Reported results include variants of strong or potential clinical significance and variants of unknown clinical significance (VUSs).

Categorical variables were compared by χ² analysis, and continuous variables were compared by Student t test using Prism version 7.0 software (GraphPad Software, San Diego, California).

Between June 2018 and December 2018, we identified 185 patients with a median age of 67 years (range, 20–92 years). Eighty-five patients were female and 100 were male. Diagnoses included 5 CHIP (3%), 11 CCUS (6%), 62 MDS (33%), 77 MPN (42%), 26 myelodysplastic/myeloproliferative neoplasm (MDS/MPN) (14%), and 4 SM (2%). Of note, 1 patient had a diagnosis of both chronic myeloid leukemia and SM. Across the cohort, variants of strong, potential, or unknown clinical significance were identified in 176 cases (95%). Numbers of variants with strong/potential clinical significance and numbers of variants with unknown significance for each of the 62 genes tested are shown in Figures 1 and 2, respectively.

Overall, the AMP 34-gene core had a diagnostic yield of 85.4% (158 of 185 patients) to detect at least 1 variant with strong/potential clinical significance and 57.8% (107 of 185 patients) to detect at least 2 strong/potential clinically significant variants (Table 2; Figure 3). All 5 CHIP patients harbored at least 1 variant with strong/potential clinical significance in the AMP 34 genes. Among 11 CCUS patients, 10 (90.9%) had at least 1 strong/potential clinically significant variant, whereas the remaining patient showed an isolated deletion of chromosome 20q by cytogenetics. For patients with overt myeloid neoplasms (ie, excluding CHIP and CCUS) the diagnostic yield was 84.6% (143 of 169 patients) and 58.62% (99 of 169) for at least 1 or at least 2 variants of strong/potential clinical significance, respectively. Although differences in yield between diagnostic categories were not statistically significant (P = .16, χ²), the yield in patients with overt CMNs was highest in patients with MDS/MPN, with 25 of 26 (96.2%) for the detection of at least 1 strong/potential clinically significant variant and 24 of 26 (92.3%) for the detection of at least 2 variants. Among 185 patients, 49 (26.5%) were found to have at least 1 VUS in the AMP 34 genes. Genes containing VUSs in 2 or more patients included DNMT3A (n = 8), EZH2 (n = 6), ASXL1 (n = 5), BCORL1 (n = 5), TET2 (n = 5), ZRSR2 (n = 4), BCOR (n = 3), RUNX1 (n = 3), CEBPA (n = 2), and NF1 (n = 2) (detailed in Supplemental Table 1; see supplemental digital content at https://meridian.allenpress.com/aplm in the August 2022 table of contents).

The expanded 62-gene NGS panel had an overall diagnostic yield of 86.5% (160 of 185 patients) to detect at least 1 variant with strong/potential clinical significance and 60.5% (112 of 185 patients) to detect at least 2 strong/potential clinically significant variants (Table 2; Figure 3). For patients with overt myeloid neoplasms (ie, excluding CHIP and CCUS) the diagnostic yields were 85.8% (145 of 169 patients) and 61.5% (104 of 169) for at least 1 or at least 2 variants of strong/potential clinical significance, respectively. Variants of strong or potential clinical significance were detected in at least 2 cases in CUX1 (n = 9), ETNK1 (n = 4), RIT1 (n = 4), SUZ12 (n = 3), DDX41 (n = 2), and SH2B3 (n = 2) (Table 3). There were 2 patients, 1 with MDS and 1 with chronic myelomonocytic leukemia, for whom no mutation was detected by the AMP 34-gene set, whereas at least 1 variant with strong/potential clinical significance was detected by the expanded 62-gene panel. The MDS patient harbored a potentially clinically significant variant in WT1 (p.R370Pfs*16, c.1105_1108dupCGAC; variant allele fraction 18.2%), and the patient with chronic myelomonocytic leukemia had a potentially clinically significant variant in KDM6A (p.Q191*, c.571C>T; variant allele fraction 25.3%). Among patients with overt CMNs, the diagnostic yield was highest in patients with MDS/MPN, with 100% for the detection of at least 1 strong/potential clinically significant variant and 92.3% for the detection of at least 2 variants (P = .06 across diagnostic categories, χ²). Among all 185 patients, 76 (41.1%) were found to have at least 1 VUS with the complete 62-gene set. Genes from the additional 28-gene set containing VUSs in 2 or more patients included KMT2A (n = 10), DDX41 (n = 7), CUX1 (n = 3), SH2B3 (n = 3), WT1 (n = 3), and SUZ12 (n = 2) (detailed in Supplemental Table 2). Twenty-five of the 185 patients (13.5%) showed wild-type results with all 62 genes, including 1 of 11 CCUS (9.1%), 9 of 62 MDS (14.5%), 13 of 77 MPN (7.3%), and 2 of 4 SM (50.0%).

The total number of detected variants per case was compared using the core 34-gene set and the expanded 62-gene set (Table 4). The mean number of mutations found per case increased with the larger gene set (2.2 ± 0.12 versus 2.3 ± 0.13, P < .001, Student t test). The median number of variants was similar in both panels, with a median of 2 variants per case and a range from 0 to 8 variants per case.

The CMNs are a heterogeneous group of hematopoietic conditions that include MDSs, MPNs, MDS/MPNs, and SM. In the 2016 WHO classification, most diagnostic categories for CMNs lack a single variant as a common driver for the group; in contrast, each neoplasm typically contains 1 or more driver mutations.^1,12  Furthermore, it has been discovered that similar variants can be present in hematopoietic cells of individuals showing no clinical features of a myeloid neoplasms. This phenomenon is known as CHIP (in the setting of normal peripheral counts) and CCUS (in the setting of unexplained cytopenias). Both have been recognized as premalignant conditions and are associated with an increased risk of a subsequent overt hematopoietic neoplasm.^14–18  Because of the complexity of mutation profiles in CMNs and their premalignant conditions, sequencing studies have been increasingly used in routine clinical practice and have demonstrated their advantage in efficiently establishing clonality as well as monitoring clonal evolution in patients with CMNs.^19–21 

Currently, there is significant variability in the genes included in targeted NGS panels from different laboratories. It remains challenging for laboratories to determine how many genes and which genes need to be included in NGS panels that are intended to evaluate patients with CMNs. The AMP CMN Working Group recommended 34 genes as a minimum list to provide relevant critical information for the management of most CMNs.¹²  Our study demonstrated that the AMP-recommended 34-gene set showed a high diagnostic yield for CMNs, with at least 1 variant with strong/potential clinical significance detected in 85.4% of patients. Particularly for patients with MDS/MPN, the yield is as high as 96.2%. Cases of MDS/MPN are often diagnostically challenging, especially when the karyotype is normal and they are evaluated at an early stage of disease presentation. Although mutational results in isolation are not diagnostic of an overt myeloid neoplasm, the current WHO classification includes the demonstration of acquired somatic mutations as a diagnostic criterion for MDS/MPN, including chronic myelomonocytic leukemia and juvenile myelomonocytic leukemia.¹  Therefore, NGS panels that include these 34 core genes can provide critical information to assist in the diagnosis of such cases.

We also examined the impact of expanded gene targets beyond the AMP-recommended 34-gene set. The expanded 62-gene panel slightly increased the overall diagnostic yield by 1.1% in comparison with the AMP 34-gene set, with 86.5% of cases containing at least 1 variant of strong/potential clinical significance. For the detection of at least 2 such variants, the overall diagnostic yield was increased by 2.7% to 60.5%. Among the additional 28 genes, CUX1, ETNK1, RIT1, and SUZ12 were the most frequently mutated genes (Table 3).

The CUX1 gene is located on chromosome 7q22.1 and contains 4 DNA-binding motifs. It has been shown that inactivating mutations of CUX1 or loss of chromosome 7q/monosomy 7 resulting in haploinsufficiency of CUX1 is important in some patients with MDS and AML.²²  In our cohort, 6 of 9 CUX1 mutant cases harbored diagnosis of MDS, with a wide spectrum of mutations including missense, nonsense, and frameshift mutations. The ETNK1 gene is located on chromosome 12p12.1 and encodes an ethanolamine kinase within the phosphatidylethanolamine synthesis pathway. Previous studies showed that ETNK1 mutations were recurrent in BCR/ABL1-negative atypical chronic myeloid leukemia and chronic myelomonocytic leukemia.^23,24  In our cohort, 4 ETNK1 mutant cases presented as 4 different pathologic diagnoses. However, 3 cases harbored the same variant (p.N244S), which is located within the kinase domain and represents the most common recurrent ETNK1 mutation.^24,25  The RIT1 gene on chromosome 1q21 encodes a member of the RAS family of GTPases. RIT1 mutations were found primarily in MDS/MPN overlap disorders in this series, in keeping with prior studies describing RIT1 mutations in a variety of myeloid neoplasms, with a highest incidence in chronic myelomonocytic leukemia.²⁶  The SUZ12 gene is located on chromosome 17q11.2 and encodes the polycomb protein Suz12, a member of the polycomb repressive complex 2 (PRC2), which also includes EED and EZH2. Suz12 has been shown to be required for hematopoietic stem cell function and lymphopoiesis.²⁷  Mutations in SUZ12 have been detected in acute myeloid leukemia, MDS/MPNs, and myelofibrosis.^28,29  In our cohort, 2 SUZ12 frameshift mutations were detected in patients diagnosed with MDS and primary myelofibrosis, and 1 splice-site variant was found in an MDS/MPN. Based upon our findings and the prior literature regarding the biologic effects of these mutations in CMN, we recommend inclusion of CUX1, ETNK1, RIT1, and SUZ12 in addition to the AMP 34-gene core set for routine evaluation of CMN. Although none of these genes define a diagnostic category in the current WHO classification, the demonstration of mutations in these genes can assist in classification of challenging cases, as they are more common in some myeloid neoplasms than others, and they may add prognostic information either individually or collectively with other variants tested. As described in a separate recent study,³⁰  we also recommend the inclusion of DDX41 for identification of CMNs associated with germline predisposition.

Although the expanded 62-gene panel slightly increased the overall diagnostic yield by 1.1% in comparison with the AMP 34-gene set, the detection rate of VUSs markedly increased from 26.5% with the AMP 34-gene set to 41.1% using the expanded 62-gene panel. This phenomenon has been reported by other researchers when evaluating clinical genetic testing in pediatric cardiomyopathy.³¹  Therefore, laboratories need to consider the burden of variant interpretation from an increasing detection rate of VUSs when contemplating a larger panel for the evaluation of CMNs. For example, VUSs were also detected in this cohort in DDX41, CUX1, and SUZ12. Our recommended expansion of routine testing to include these 3 additional genes therefore also has potential to increase the rate of VUSs. In our view, the advantages of including these targets as described above outweigh the risk of VUS detection.

Our study has certain limitations. This was a retrospective study performed in a single institution. The results of our study were based on the performance of our own NGS panel; therefore, they may not apply to NGS panels from other laboratories, and additional genes not included in this study may further alter the rate of abnormal results. Moreover, our patient cohort included relatively limited numbers of cases for certain pathologic diagnoses. Finally, this study did not attempt to address issues of reimbursement for clinical testing, as insurance coverage and reimbursement may vary among geographic areas.

In conclusion, the AMP core 34-gene set has a high diagnostic yield for CMNs, especially for MDS/MPN cases. Laboratories should consider including the core 34 genes when designing an NGS panel with intent to evaluate CMN cases. Selected additional genes, including CUX1, ETNK1, RIT1, SUZ1, and DDX41, may provide additional clinical utility. However, laboratories should be aware that inclusion of additional genes beyond the core 34 genes will likely also further increase the detection rate of VUSs, which would potentially increase the burden of variant interpretation.