Longitudinal Clonal Tracking Reveals That Early and Sensitive Detection of Blood Cancer-Related Gene Variants in Patients with Chronic Myeloid Leukemia Predicts Treatment Failure

Background

Treatment failure in CML is associated with additional variants in genes that drive acute myeloid and lymphoid leukemia. Some variants are detected at diagnosis where ASXL1 is the most commonly mutated gene (~9% of patients [pts]), whereas others are acquired during therapy. Notably, multiple studies have shown a high frequency of concurrence of BCR::ABL1 kinase domain (KD) mutations and blood cancer-related gene variants (herein named cancer variants). Little is known about the dynamics of cancer variants over time and whether they precede the acquisition of KD mutations.

Aim

To interrogate longitudinal samples to assess the dynamics and time differential between cancer variant detection, KD mutations and treatment failure.

Method

A sensitive RNA-based hybridization capture next-generation sequencing method was used to target 53 myeloid and lymphoid blood cancer genes. Single nucleotide variants, small insertions/deletions, RNA splice altering variants, and specific gene fusions and deletions were detectable. Serial samples of 35 pts with treatment failure (2020 European LeukemiaNet criteria) were tested. Samples sequenced had BCR::ABL1 ≥1% IS by qRT-PCR. Only variants that met strict criteria for pathogenicity were retained.

Results

Pts were treated with frontline BCR::ABL1 inhibitors and received treatment intervention for lack of milestone molecular responses. The first failure criterion for the 35 pts occurred at a median of 6 months after commencing therapy, range 6-70 months: 22 pts failed milestone molecular responses; 9 lost response with a BCR::ABL1 rise above 1% IS; and 4 developed sudden blast phase. At last follow up, 8 pts had developed lymphoid blast phase, 10 myeloid blast phase, 2 accelerated phase, 3 received allogeneic transplants in chronic phase, 2 died in chronic phase and 10 maintained chronic phase.

At diagnosis, 32 cancer variants were detected in 18 of 35 pts (51%). ASXL1 variants were enriched at diagnosis: 11 of 35 pts (31%). However, only 2 of these 11 pts later developed blast phase compared to 16 of the remaining 24 pts.

The emergence of cancer variants over time was assessed. Pts had a median of 4 tests, range 2-18. BCR::ABL1 fusion transcripts were detected in all samples using the RNA-based gene panel. Overall, 31 of 35 pts (89%) with treatment failure had cancer variants detected in the follow up samples. In total, 72 cancer variants were gained after commencing therapy. Additionally, 29 BCR::ABL1 KD mutations were gained in 15 of 35 pts (43%). Every pt with a KD mutation also had 1-8 additional cancer variants. Notably, of the 15 pts who gained KD mutations, 9 had cancer variants at diagnosis and 2 other pts gained cancer variants prior to KD mutation detection. Before the first failure criterion occurred in the 35 pts, cancer variants were detected in 19 pts at a median of 6 months before failure, range 2-43 months.

All 18 pts in blast phase had cancer variants. Gene fusions and deletions were a common feature of blast phase and were detected in 13 of 18 pts (72%). RUNX1 variants were also enriched in blast phase, 8 of 18 pts (44%), but emerged in only 1 of the other 17 pts with treatment failure. The time differential between the onset of blast phase and the acquisition of cancer variants was examined. Deletions affecting IKZF1, RB1, CDKN2A, BTG1 and SETD2 were detected in 8 pts and coincided with the onset of blast phase in 6. The onset of blast phase occurred within 4 months of first detection for 6 of 7 gene fusions involving KMT2A, RUNX1, ETV6, CBFB and MECOM. The latency of blast phase for the 8 pts with RUNX1 variants was variable: median 5.7 months after first detection of the RUNX1 variant, range 0-48.

Cancer variant dynamics over time revealed: 1) several cases of convergent evolution where multiple different variants in the same gene were detected in an individual pt; 2) new variants that emerged with a change of BCR::ABL1 inhibitor; 3) clonal competition where some variants dominated in blast phase; 4) specific variants associated with the onset of blast phase; and 5) cancer variant detection prior to KD mutations in most pts.

Conclusion

Our study highlights the potential for sensitive genomic testing to enhance prediction of relapse and transformation. Cancer variants were frequently detected prior to treatment failure and KD mutations. The detection of specific variants associated with progression could warrant early consideration of allograft options.