Background: Tyrosine kinase inhibitors (TKI) revolutionized treatment of chronic myeloid leukemia (CML). While response rates are high, a subset of patients does not show the expected decrease or eradication of BCR-ABL1 . According to ELN guidelines (Baccarani et al. 2013) BCR-ABL1/ABL1 ratio>1% after one year of TKI is considered failure. Mechanisms of failure include the acquisition of additional genetic changes, e.g. cytogenetic aberrations and mutations in the ABL1 kinase domain but also in other genes. Here, we ask the question, if specific mutations are not only acquired but may already be present at initial diagnosis of patients with slow or no TKI response (primary failure according to ELN).

Aim: To screen CML patients at diagnosis, who failed ELN remission criteria after one year, for mutations in genes, which are frequently mutated in myeloid malignancies.

Patients and Methods: In a backtracking approach, we selected diagnostic samples of 48 CML patients (25 males, 23 females) with a median age of 64 (range: 23-84) years, who failed first year molecular remission criteria. Median BCR-ABL1/ABL1 ratio was 3.603 [1.158-58.318] after 12 (11-16) months of imatinib (n=42), second generation TKI (n=5) treatment or both (n=1). Patients had a p210 BCR-ABL1 fusion (e13a2 [n=24], e14a2 [n=18], both [n=6]) and no additional cytogenetic aberrations at diagnosis. We sequenced DNA of bone marrow (n=28) or peripheral blood (n=20) samples drawn at diagnosis on NextSeq or MiSeq sequencers after TruSeq library preparation (Illumina, San Diego, CA). We used SeqNext 4.3 for data analysis (JSI, Kippenheim, Germany). The panel covered 48 genes including the epigenetic regulators: ASXL1, ASXL2, BCOR, DNMT3A, EZH2, IDH1, IDH2 and TET2 .

Results: We identified mutations in 13/48 samples at initial diagnosis (27%). Only one patient had two mutations (total, n=14 mutations). There was no correlation of mutations with types of transcript, age, gender, hemoglobin level, WBC count, percentage of Ph+ cells by FISH, BCR-ABL1/ABL1 ratio at diagnosis or with BCR-ABL1/ABL1 ratio after one year of TKI. Twelve of the 14 (86%) mutations were found in genes involved in chromatin modification or DNA methylation. We screened four splicing machinery genes (SF3B1, SRSF2, U2AF1, ZRSR2), which are frequently mutated in other myeloid malignancies, but no mutation was found. In 8/48 (17%) patients, we found an ASXL1 mutation (57% of all mutations). Other mutated genes were ASXL2, BCOR, CALR, NRAS (n=1 each) and DNMT3A (n=2).

The comparison of ASXL1 mutation types in this CML cohort to other myeloid malignancies revealed a significant shift in mutation types: e.g. in myelodysplastic syndrome (Haferlach et al. 2014) and chronic myelomonocytic leukemia (Meggendorfer et al . 2012), c.1934dup and c.1900_1922del (ENST00000375687) constitute the majority of mutations (64 [42%] and 23 [15%] of 152, respectively), whereas there was only one c.1900_1922del and no c.1934dup in our CML cohort (p=0.023).

Further, we used follow-up samples after one year of TKI to evaluate if mutations are derived from the CML clones. In all patients with available data for ASXL1 (n=6), mutation burdens showed a quantitative course which paralleled BCR-ABL1 ratio indicating that ASXL1 mutations are part of the CML clone present at diagnosis. In the patient with a CALR mutation, the CALR burden remained unchanged despite BCR-ABL1 reduction, indicating a BCR-ABL1 gain in a preexisting CALR mutated myeloproliferative neoplasm.

In comparison to ASXL1 mutation rates (4/74 [5%]) in a published cohort with optimal responders (Kim et. al. 2017), we found a significantly higher number of ASXL1 mutations (8/48 [17%]) at diagnosis of our CML patients, who failed first year molecular remission criteria (p=0.041).

Conclusions: (1) At diagnosis of CML, mutations were detected (esp. in ASXL1) in 27% of patients with TKI failure according to ELN criteria after one year. Compared to other myeloid malignancies, patients showed a different pattern of ASXL1 mutations. (2) No mutations in spliceosome genes were found. (3) Stringent BCR-ABL1 monitoring allows early adjustment of therapy in case of suboptimal response, but analysis of additional mutation patterns could help understanding the biology of failure and would allow up-front adjusted therapy decisions or different monitoring strategies. (4) Mutations such as ASXL1 at diagnosis may contribute to suboptimal TKI response.


Baer: MLL Munich Leukemia Laboratory: Employment. Kern: MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach: MLL Munich Leukemia Laboratory: Employment, Equity Ownership. Haferlach: MLL Munich Leukemia Laboratory: Employment, Equity Ownership.

Author notes


Asterisk with author names denotes non-ASH members.