Late-onset NPM1 mutation in a MYC-amplified relapsed/refractory acute myeloid leukemia patient treated with gemtuzumab ozogamicin and glasdegib

In March 2020, a 72-year-old male patient with no relevant medical history presented with 4 weeks of asthenia and adynamia, 10-kg weight loss and petechias on both hands. Examination of the bone marrow (BM) revealed 80% infiltration by immature blastoid cells with the following immune phenotype: low-intermediate CD45 expression, CD34^+/-, CD117⁺, HLA-DR⁺, CD38⁺, CD13^+/-, CD33^-/+, myeloperoxidase (MPO)^-/+, CD15^-/+, nuTdt^-/+, CD123^-/+, CD7⁺. In the next-generation sequencing (NGS) acute myelogenous leukemia (AML) panel only one mutation in TET2 p.Y1245Lfs*22 was reported. Fluorescence in situ hybridization (FISH) identified a MYC amplification, characterized by double minutes (dmin) involving chromosome band 8q24 in 71% of uncultured BM cells. The disease was classified as AML not otherwise specified (AML-NOS) according to Eurpoean Leukemia Net (ELN) 2017 criteria.¹ Intensive treatment within the “GnG” trial was initiated after obtaining the patient´s written informed consent comprising gemtuzumab-ozogamicin and “7+3” during induction and the hedgehog inhibitor glasdegib during consolidation and maintenance therapy.² The course of events is described in detail in Figure 1. Glasdegib maintenance therapy was prematurely discontinued after three cycles due to persistent dysgeusia. For this specific type of AML, no targeted maintenance therapy had been approved in the European Union at that time. However, cytologic relapse occurred 3 months after the suspension of glasdegib with 60% BM blasts (Figure 1). NGS revealed several new molecular mutations including a typical NPM1 4bp insertion (Table 1). FISH analysis showed the initial MYC amplification in 1% of uncultured BM cells, however a new trisomy 8 involving the MYC locus (8q24) occurred. We initiated salvage therapy with azacitidine and venetoclax (aza-ven)³ and prepared the patient for allogeneic stem cell transplantation. However, BM aspiration after the second cycle revealed refractory disease with 60% blasts and persistent NPM1 mutational load. Another salvage therapy with high-dose cytarabine and mitoxantrone (HAM) was initiated, but the disease was refractory with novel extramedullary skin lesions. Considering the poor prognosis of this chemo-refractory disease, as well as the patient’s age and wish a best supportive care therapy was initiated. The patient died 2 months after the appearance of the skin lesions.

Figure 1.Timeline of therapy, clinical response, and sampling time points. The patient was randomized to receive gemtuzumab ozogamicin (GO) in addition to “7+3” during induction therapy as follows: cytarabine 200  mg/m² administered via continuous intravenous (IV) infusion for a total of 7 days and daunorubicin 60 mg/m² days 1, 2, and 3; in addition GO 3  mg/m² IV over 1 hour (h), on days 1, 4, and 7. During the consolidation phase, the patient received 2 cycles of cytarabine (1.0  g/m²) administered by IV infusion every 12  h on days 1, 2, and 3. Glasdegib 100  mg was administered from day 1 to 28 of each consolidation cycle. Maintenance therapy with glasdegib 100 mg was started after the end of the second consolidation therapy cycle for a total of 3 additional cycles of 28 days each^.2 The patient achieved complete hematological remission with incomplete hematological recovery (CRi) after induction. Minimal residual disease (MRD) assessed by flow cytometry remained positive. After the first consolidation therapy, CR with MRD negativity was achieved. At this point, an allogeneic transplantation was disclaimed by the patient. Thus, a second cycle of consolidation therapy was administered. Maintenance therapy with glasdegib was initiated 5 months after induction therapy. Glasdegib was prematurely discontinued after 3 cycles of maintenance therapy due to the patient’s wish because of adverse effects (dysgeusia and muscle cramps grade II). Cytologic relapse occurred 3 months after the suspension of glasdegib. Bone marrow (BM) examination revealed 60% blasts. Flow cytometry analysis showed a leukemia population with the following characteristics: CD34^-, CD117^-, HLA-DR⁺, CD38⁺, CD33⁺, CD13^-/+, CD64⁺, NG2⁺, CD15⁺ and a second population displaying the initial immunophenotype. Next-generation sequencing revealed molecular aberrations not detected at initial diagnosis. The NPM1-mut/ABL ratio quantified by real-time polymerase chain reaction was 520%. Salvage therapy with azacitidine and venetoclax (aza-ven) was initiated. After 1 cycle, blast percentage dropped to 9%. A second cycle of aza-ven was administered and the patient was prepared for allogeneic transplantation. BM aspiration after the second cycle of aza-ven revealed refractory disease with 60% blasts and an NPM1-mut/ABL ratio of 416%. Another salvage therapy with high-dose cytarabine and mitoxantrone (HAM) was initiated. Day +15 BM aspiration revealed only 1% blasts but MRD positivity for NPM1. The patient did not achieve hematological recovery, and on day 35, the patient presented with 23% BM blasts. In addition, he developed progressive skin papules suspicious of chloroma covering the entire body, and skin biopsy confirmed AML cell infiltration. Best supportive care therapy was initiated. The patient died 2 months later. CRi: complete remission with incomplete hematological recovery, ID: initial diagnosis; CRi-1: complete remission with incomplete hematological recovery one; CRi-2: complete remission with incomplete hematological recovery two; Rel-1: relapse one; PD: progressive disease; RD: refractory disease.

In order to gain a deeper understanding of the clonal composition and evolution of this unusual course of the disease, we performed single-cell DNA sequencing (scSeq) at four time points indicated in Figure 1 and Table 1 using the Mission Bio Tapestri single-cell DNA and protein sequencing platform investigating genes covered in Mission Bio’s myeloid panel⁴ with v2 of the protocol. Additionally, we used antibodies against four surface proteins to simultaneously generate a single-cell protein library. Libraries were sequenced to a DNA/protein read depth of 324/31 M (ID), 267/31 M (CRi-1), 292/37 M (CRi-2) and 225/37 M (Rel-1) on an Illumina NextSeq2000 sequencer (P3, 2x150bp). Sequencing details are available upon request. Using scSeq, we were able to confirm the MYC amplification at initial diagnosis in a large fraction of CD34⁺ blasts.

In order to discern MYC-amplified from wild-type cells, we used CloneTracer to assign clonal probabilities to each cell.⁵ Those cells being assigned to the MYC clone were termed MYC-amplified (MYCamp) (Figure 2A). Notably, healthy T cells were reliably identified as non-MYCamp as expected (Figure 2A). We identified an EZH2 missense mutation p.T683A in CD34⁺ cells with aberrant CD7 expression, which seemed to be present in the MYCamp and also non-MYCamp CD34⁺CD7⁺ cells, but not in healthy CD3⁺CD34^- T cells. This suggested that the EZH2 mutation occurred in an ancestral committed progenitor cell giving rise to the MYCamp leukemic clone (Figure 2B, C). Moreover, an ASXL1 p.E635Rfs mutation was detected in around 1% of the cells not harboring MYCamp (Figure 2B, C).

In the remission sample post induction (CRi-1), which was considered measurable residual disease (MRD)-negative by flow cytometry, the EZH2 mutation was present in the majority of CD3^- BM cells, but MYCamp was detectable in a minority of cells (Figure 2D). However, we identified a JAK2 V617F mutation typically found in myeloproliferative neoplasms (MPN) and AML,⁶ in 9% of the cells, and the ASXL1 mutation had increased in frequency. Notably, we observed a high allelic dropout rate of EZH2, suggesting that many cells called as EZH2 wild-type were actually mutant, and that the overall fraction of EZH2 mutant myeloid cells was very high. Considering also that the false-positive rate of the applied technology is very low,^7,8 the abundant co-occurrence of the ASXL1 and EZH2 mutations in the first remission time point (Figure 2D) strongly suggested that ASXL1 occurred downstream of the EZH2 mutation.

Table 1.Clonal composition and evolution of the disease.

Figure 2.Single-cell DNA sequencing analysis. (A) Dot plots showing total counts on the x-axis and MYC counts on the y-axis, as well as highlighted in color the cell type (left panels: red: T cells, gray: other cells) or mutational status (right panels: blue: wild-type, yellow: MYC-amplified) at the indicated time points initial diagnosis (top) and relapse (bottom). (B) Dot plots showing total counts on the x-axis and MYC counts on the y-axis, as well as highlighted in color the mutational status (blue: wild-type, yellow: mutant) at the indicated time points initial diagnosis (top) and relapse (bottom). (C-E) Heatmaps showing the mutation-al status for the indicated genes at first diagnosis (C), remission 1 (D), and at relapse (E). Cells with dropout in any of the genes were removed, and the cells were additionally annotated with the expression of surface proteins (CD34, CD7, CD3, CD14) measured with cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq). (F) Schematic of the clonal evolution as detected through single-cell DNA sequencing.

In the second remission sample post consolidation two (CRi2) MYCamp was detectable at a very low rate by scSeq, while FISH was still negative. The JAK2 clone increased in abundance (25%), continued to co-occur with the EZH2 p.T683A variant, but did not overlap with MYCamp cells (Table 1). In addition, a new PHF6 mutation (p.Y103stop) was detected in 10% of cells. These cells shared the EZH2 p.T683A, but they were devoid of the JAK2 mutation and MYCamp. Of note, blasts were below 5% at this stage and flow cytometry did not detect the leukemia-associated immunophenotype from initial diagnosis, so that the patient was considered flow MRD-negative.

In the cytologic relapse sample, which was acquired 6 months later, MYCamp was detectable again by scSeq (12%; Figure 2E, F). The JAK2 clone had decreased from 25% to 7%, while the PHF6-mutated cells represented the second largest clone at relapse downstream of the EZH2 mutation (22%). An NPM1 TCTG-insertion was found in 15% of cells. Interestingly, the NPM1-mutated cells harbored the ASXL1 mutation, which was already present at diagnosis independently of the other three clones (Figure 2E, F).

Several observations are remarkable in this scSeq study: an EZH2 missense mutation p.T683A was detectable already at initial diagnosis and was shared by the initial MYCamp clone, but also by two distinct JAK2- and PHF6-mutated clones emerging during cytologic remission. Based on the in-depth scSeq information this AML would have been classified as high-risk AML with myelodysplasia-related gene mutations (AML-MR) according to the new World Health Organization (WHO) and ELN 2022 classification, so that a therapy with CPX-351 would have been recommended.¹

While the JAK2-mutated clone decreased from remission to relapse, the PHF6-mutated clone became the second largest subclone downstream of the EZH2 mutation at relapse potentially pointing towards clonal selection during the maintenance therapy with glasdegib. PHF6 mutations in AML are often frameshift or nonsense mutations.⁹ As PHF6 is an X-linked gene, these alterations are predominantly found in males as in our patient. PHF6 mutations are associated with immature AML (FAB subtypes M0–M2).⁹ Moreover, PHF6 along with other epigenetic regulators such as DNMT3A, TET2, and ASXL1, is frequently altered in patients with clonal hematopoiesis.⁹ Genomic amplifications, such as dmin, homogeneously staining regions (hsr), are rare in leukemia, accounting for less than 1% of cytogenetically abnormal hematological malignancies and are associated with elderly patients and poor prognosis.¹⁰ While we found MYCamp in form of dmin at initial diagnosis by FISH, the relapse sample revealed a novel trisomy 8 involving the MYC locus confirming the hypothesis that different MYCamp mechanisms can co-exist within the same leukemic cell population as previously described.¹¹ scSeq was able to detect MYCamp in both conditions, and the fraction of cells highly corresponded to the fractions detected by FISH or cytogenetics. MYC mutations have been reported to coincide with other mutations in AML, such as FLT3, NPM1, and DNMT3A mutations.¹² The interesting observation in this case is, that the additional clones developing during therapy did not occur in the MYCamp clone as expected. This might suggest, that the MYCamp clone facilitated development of the other AML clones by paracrine mechanisms, e.g., via BM niche remodeling.

Another striking observation in this case is the occurrence of a therapy-associated NPM1 mutation. Approximately 15% of therapy-related AML cases carry NPM1 mutations, frequently accompanied by a normal karyotype and DNMT3A mutations, and rarely associated with chromosome aberration.¹³ Cases of NPM1-mutated AML following chemotherapy for previous lymphoid malignancies appear to arise from a background of DNMT3A- or TET2-driven clonal hematopoiesis (CH) rather than being a direct result of cytotoxic therapy.¹⁴ It is therefore remarkable, that the NPM1 mutation in this case occurred in cells harboring mutations in two other epigenetic regulators, ASXL1 and EZH2, suggesting a common mechanism. The ASXL1-mutated background and the co-existence of NPM1 wild-type clones might explain the non-response to aza-ven in this case.

In summary we present a very unusual AML case with emergence of multiple parallel genetic clones during remission after treatment with intensive chemotherapy and targeted drugs including the hedgehog inhibitor glasdegib and the CD33-directed antibody “GO”. The complex clonal evolution observed in this case resulted in a very aggressive and multi-refractory disease, which raises the question, whether and how the agents applied might have contributed to clonal selection or exit from dormancy. These findings underscore the importance of employing scSeq to dissect individual cases, which can offer valuable mechanistic insights and help optimize therapeutic strategies.

Footnotes

• Received January 10, 2024
• Accepted May 20, 2024

Correspondence

C. PABST - Caroline.Pabst@med.uni-heidelberg.de

Disclosures

No conflicts of interest to disclose.

Funding

SDS@hd