Therapy-related myeloid neoplasms (t-MN) are a heterogeneous group of aggressive myeloid neoplasms that arise following exposure to cytotoxic therapy and/or ionizing radiation for treatment of prior non-myeloid malignancy or autoimmune disease.1 Leukemogenic agents that have been implicated in the development of t-MN include alkylating agents (e.g., cyclophosphamide, cisplatin, mitomycin C, etc.), ionizing radiation, topoisomerase II inhibitors (e.g., etoposide, doxorubicin, etc.), and other antimetabolite and antitubulin agents. Each therapeutic group has been associated with varying latency intervals from the time of therapy exposure to onset of t-MN, as well as certain recurrent genetic alterations. The focus of this review will be on the molecular genetic alterations that have been described in t-MNs, as well as recent updates regarding diagnostic classification.

The relative risk for t-MN after chemotherapy for primary malignancy ranges from 1.5 to greater than 10 for the general population, based on the Surveillance, Epidemiology and End Results (SEER) database.2,3 Any age group can be affected, but the risk of alkylating agent and ionizing radiation-associated t-MN generally increases with age and occurrence appears to correlate with intensity of chemotherapy.4 Even after allogenic hematopoietic stem cell transplant for treatment of t-MN, frequent high non-relapse mortality is noted, and median overall survival is 6-12 months.3,5, 6, 7, 8

The primary molecular mechanisms of t-MN ontogeny that have been proposed include: inherited cancer susceptibility, direct induction of an oncogene, induction of genetic instability, and selection of a pre-existing treatment-resistant hematopoietic stem cell clone that is susceptible to genetic instability.9 More recent studies favor the latter as the primary mechanism in t-MN pathogenesis.10, 11, 12

The revised 2017 WHO (4th ed.) classification unified therapy-related myelodysplastic syndrome (t-MDS), therapy-related myelodysplastic syndrome/myeloproliferative neoplasm (t-MDS/MPN), and therapy-related acute myeloid leukemia (t-AML), under the diagnostic umbrella of therapy-related myeloid neoplasms.1 The rationale for this broad diagnostic category is because of their similar pathogenesis, rapid progression from MDS to AML, and their equally poor prognosis. One study showed no significant difference in clinical outcome between patients with a diagnosis of t-MDS or t-AML, both of which showed uniformly poor outcome regardless of blast count.13 Much of the t-MN literature has focused on t-MDS and t-AML. In large population-based studies, patients diagnosed with t-AML account for 7-15% of AML patients. 5,14, 15, 16, 17 It is speculated that the rising incidence of t-MDS and t-AML is likely, in part, due to increasing numbers of cancer survivors.3 Progression to t-AML is typically ∼6 months after initial t-MDS diagnosis.9 Information on t-MDS/MPN is more scarce, considering they account for a minority of t-MN cases (∼10%) and reportedly have similar mutational profiles and outcomes as de novo MDS/MPN counterparts.18, 19, 20

Given the heterogeneity of myeloid neoplasms, efforts to classify diagnostic entities based on a more genetically defined system have been undertaken. Like patients diagnosed with de novo AML, it is recommended that patients with t-MN be treated according to their genetic risk profile.16,21,22 In order to minimize overlap between other diagnostic categories, prior leukemogenic therapy predisposing to development of t-MN, are now considered “qualifiers” to the diagnosis, rather than a specific diagnostic category in the most recent International Consensus Classification (ICC) and 2022 WHO (5th ed.) classification systems.22,23

The recent ICC system states that MDS and AML cases that are therapy-related should be qualified as such by using a “therapy-related” statement after the diagnosis, prioritizing morphologic and genetic characteristics.22,24 The ICC also recommends qualifying a diagnosis of clonal hematopoiesis of indeterminate potential (CHIP) and non-MDS clonal cytopenia of unknown significance (CCUS) as “therapy-related” if they arise following bone marrow exposure to chemotherapy or radiation therapy, given that both CHIP and CCUS can occur as precursors to t-MN.25 (Example: Acute myeloid leukemia with myelodysplasia-related cytogenetic abnormality, therapy-related).

The 2022 WHO (5th ed.) classification includes a newly segregated “secondary myeloid neoplasm” category that encompasses diseases that arise in the setting of predisposing factors, including prior exposure to cytotoxic therapy. In the 5th ed. WHO classification, “post cytotoxic therapy” should be used as a qualifier in the diagnosis of myeloid neoplasms (e.g., AML with KMT2A rearrangement, post cytotoxic therapy).23

Alkylating agents encompass a broad range of structurally diverse chemotherapeutic agents, including platinum-based agents and nitrogen mustards, that exert their anti-cancer affects by directly altering DNA. Damage occurs through the generation of DNA crosslinks, adducts, and strand breakage, resulting in advantageous therapeutic cytotoxicity.26,27 When DNA is damaged, the cell cycle is halted at specific control checkpoints in order to repair DNA damage, and prevent accumulation of deleterious alterations in subsequent cell cycles. If the damage cannot be repaired, programed cell death/apoptosis mechanisms are initiated to circumvent accumulation and propagation of genetic alterations. If hematopoietic stem cells are able to efficiently by-pass the checkpoints needed for DNA repair, therapeutic resistance may develop, enabling an abnormal clone to arise with a survival advantage. Alkylating agents are more often associated with unbalanced chromosomal aberrations (e.g., monosomy 5, monosomy 7, 17p losses) and complex karyotype, thought to be secondary to accumulation of genetic alterations. The latency interval from time of exposure of leukemogenic therapy to onset of t-MN is variable, but typically longer (5-10 years) than patients exposed to topoisomerase II inhibitors. However, latency interval and t-MN onset prediction can be difficult because patients are often exposed to various combinations of leukemogenic agents in modern combined modality therapeutic (CMT) regimens, who receive cytotoxic therapy and radiotherapy.2 Patients with t-MN following exposure to alkylating agents and/or ionizing radiation typically present with myelodysplasia and rapidly progress to overt t-AML.

Cytotoxicity following ionizing radiation exposure is primarily through induction of double-strand DNA breaks. Secondary effects include the generation of reactive oxygen species that oxidize proteins and lipids, induce single-strand DNA breaks, alter nucleic acids through sugar moiety modification, deamination of nitrogenous bases, and induction of apurinic and apyrimidinic sites.27 In hematopoietic stem cells that are resistant to ionizing radiation, these deleterious modifications can accumulate and may ultimately lead to leukemogenic mutations and cytogenetic abnormalities that can progress to t-MN. The incidence of t-MN secondary to the effects of modern limited-field radiation therapy, which significantly reduces bone marrow exposure, remains to be elucidated.28 Latency from time of ionizing radiation exposure to onset of t-MN is similar to that of patients exposed to alkylating agents (5-10 years). One study showed that patients with t-MN following treatment with radiation therapy-alone were genetically more similar to de novo MDS and AML compared to t-MN in patients who previously received chemotherapy-alone or CMT.28 Patients treated with radiation-only were significantly less likely to have chromosome 5 and chromosome 7 abnormalities compared to those who received chemotherapy or CMT.

Therapy-related myeloid neoplasms that develop following anthracycline (e.g., daunorubicin, idarubicin) and direct topoisomerase II inhibitors (e.g., etoposide) induce DNA damage through direct or indirect inhibition of topoisomerase II, which plays a key role in DNA replication, transcription, chromatin condensation, and chromatin segregation via its ability to induce and re-ligate double strand DNA breaks.27,29 Topoisomerase II inhibitors primarily induce DNA damage by interfering with the ligation step of topoisomerase II, after double-strand DNA breaks have been induced and stabilized. Risk is less clearly related to cumulative dose, but it is associated with dosing schedule.8,30 These patients typically do not present with a t-MDS phase, but rather with overt t-AML with balanced gene rearrangements (e.g., KMT2A::MLLT3, RUNX1::RUX1T1, CBFB::MYH11, etc.).31 These cases may be morphologically similar to de novo AML with similar cytogenetic abnormalities (e.g., AML with KMT2A rearrangement). Latency from exposure to topoisomerase II inhibitors to diagnosis of t-MN is typically shorter (2-3 years) than that of patients exposed to alkylating agents or ionizing radiation.

T-MN comprise a subset of “secondary” acute myeloid leukemia, which refers to a composite designation that alludes to AML that has arisen from prior myeloid malignancy (i.e., MDS, MPN, etc.) or is therapy-related.7 An 8-gene classifier that includes mutations involving chromatin remodeling and spliceosome machinery (SRSF2, U2AF1, SF3B1, ZRSR2, ASXL1, EZH2, BCOR, STAG2) was described by Lindsley et al. who detailed that the presence of at least one mutation in the aforementioned genes was >95% specific for secondary AML and was associated with inferior outcome.32 These mutations are sometimes referred to as “secondary-type” alterations. Additional accumulation of mutations in myeloid transcription factors (i.e., RUNX1, WT1, etc.), signaling pathways (i.e., FLT3, KRAS, etc.), and epigenetic regulators (i.e., IDH1, IDH2, TET2) also contribute to leukemogenesis.7 Frequent mutations involving ASXL1, SRSF2, and TET2 in t-CMML have been reported, but without significant differences in mutational profile compared to de novo CMML.33

Preleukemic clonal hematopoiesis (CH) is common in patients with t-MN at the time of their primary cancer diagnosis and the cumulative incidence of t-MN is significantly higher at 5 years and 10 years compared to patients without CH.25 Hematopoietic stem cells harboring TP53 and PPM1D (a regulator of p53) mutations at low variant allele frequency (VAF) have been identified in pre-therapy bone marrow and peripheral blood samples that preferentially expand following cytotoxic chemotherapy and ionizing radiation.10,11,25,34,35 In one study, 25% of patients with non-hematologic malignancy harbored CH-associated mutations in peripheral blood samples from the time of initial cancer diagnosis.11 Of those, DNMT3A, TET2, PPM1D, ASXL1, and TP53 mutations accounted for 60% of CH-associated mutations and were also associated with increased age, prior therapy, and tobacco use. Prior exposure to cytotoxic chemotherapy was specifically associated with TP53 and PPM1D mutations.11 Another study found that following therapeutic exposures, transformation of CH to t-MN is, in part, due to differential selection of clones harboring CH-associated mutations.12 Among patients where CH was previously detected, CH-associated mutations (e.g., DNMT3A, TET2, ASXL1, etc.) were persistent at t-MN diagnosis. Mutations in ASXL1 were enriched in smokers. Exposure to radiation, platinum-based agents, and topoisomerase II inhibitors preferentially selected for clones harboring mutations in DNA damage response genes like TP53, PPM1D, and CHEK2, which were able to outcompete other clones when exposed to certain therapies, however, these clones had lower competitive fitness relative to clones with non-DNA damage response gene (e.g., DNMT3A) mutations in the absence of cytotoxic or radiation therapy.12

Mutations in spliceosome genes (e.g., SRSF2 and SF3B1) were less common in one t-MN cohort, relative to other CH-associated mutations, but showed the strongest association with age.12 In patients with CH with mutations in epigenetic modifiers (DNMT3A, TET2), or splicing regulators (SRSF2, SF3B1, U2AF1), these clones were not strongly affected by exposure to therapy compared to clones harboring DNA damage response gene mutations (TP53, PPM1D, CHEK2), further supporting that the relative fitness of acquired mutations in CH is modulated by environmental factors that can progress to t-MN.12 Identification of treatment resistant clones may help facilitate screening and counseling of patients before initiating treatment of their primary disease.14,25

The most common chromosomal alterations in t-MN that have widely been described in both t-MDS and t-AML include monosomal and complex karyotypes that result from chromosomal deletions or losses. Complex karyotype is defined as at least 3 clonal cytogenetic abnormalities and is associated with chromosomal instability and alterations in TP53 and/or defective DNA repair mechanisms.24,36 Approximately 35% of t-AML patients have MDS-related cytogenetic abnormalities (e.g. complex karyotype and monosomal karyotypes).14 Deletion of the long arm of chromosome 5, and whole chromosome loss of or partial loss of chromosome 7, are common in t-MN, both of which are considered adverse cytogenetic risk.31,36, 37, 38 Chromosome 5 and 7 losses lead to haploinsufficiency in a number of genes implicated in malignancy, including RPS14, APC, EGR1, and CSNK1A1 on chromosome 5 and CUX1, KMT2C, LUC7L2, CUL1, and EZH2 on chromosome 7.39

Other common abnormalities include trisomy 8 and loss of chromosome 17p (TP53 locus). Trisomy 8 occurs in ∼15% of t-AML and is less frequent compared to de novo AML. TP53 mutations typically involve the DNA binding domain with missense variants being most common, but deletions, insertions, and nonsense variants can also occur.40 Deletion of the other wild-type TP53 allele, or alternative loss of heterozygosity can lead to a null TP53 state.7 Defective TP53 ultimately leads to unchecked cell cycle progression, cell proliferation and survival. These alterations are most strongly associated with exposure to alkylating agents and ionizing radiation exposure involving large fields of active bone marrow.10, 11, 12,25,34,35 (Figure 1. t-MN with TP53 alterations example case).

In the recent ICC classification, acute myeloid leukemia with TP53 mutations are defined as a distinctly aggressive AML category, whether they present de novo, as progression of MDS, or as a therapy-related myeloid neoplasm. While multi-hit TP53 mutation is required for diagnosis of “MDS with mutated TP53” by ICC classification, for diagnosis of “AML with mutated TP53” any pathogenic TP53 mutation VAF of >10% is sufficient. 36,41,42 Similarly, in the 2022 WHO (5th ed.) classification, the diagnostic category of “MDS with biallelic TP53 inactivation” requires two or more TP53 mutations, or 1 mutation with evidence of TP53 copy number loss or copy-neutral loss of heterozygosity.23 Complex karyotypes are usually also identified.

Balanced gene rearrangements are more common among patients with t-MN following topoisomerase II inhibitor therapy and patients typical present with AML, without a preceding MDS phase.8,27,29,31 Rearrangements involving KMT2A are most strongly associated with topoisomerase II inhibitors, however, other rearrangements can also occur.9 In one series, KMT2A::MLLT3 fusion accounted for 11% of t-AML cases.14 Patients with t-AML have similar frequency of recurrent gene fusions (RUNX1::RUNXT1, CBFB::MYH11, PML::RARA) as de novo AML, where these alterations are associated with favorable prognosis.9 One study by the Bone Marrow Pathology Group evaluated core-binding factor t-AML (RUNX1::RUNXT1 and CBFB::MYH11) and showed that much like de novo core-binding factor AML, t-AML with RUNX1::RUNXT1 or CBFB::MYH11 fusions had high response rates after intensive chemotherapy and similar achievement of complete remission as de novo AML.43 However, patients with core binding factor t-AML have shorter overall survival and more frequent relapse compared to de novo AML14,43, 44, 45, 46 Rare cases of therapy-related acute promyelocytic leukemia (t-APL) with PML::RARA fusion have been reported that are morphologically similar to de novo APL, with similar prognosis and response to all-trans retinoic acid (ATRA) therapy.47

t-MN with normal karyotype (NK) typically present as t-AML and share significant morphologic, immunophenotypic, and genetic overlap with de novo AML with normal karyotype, and also differ from “classic” t-MNs which are more often associated with complex karyotype with TP53 alterations.48, 49, 50 NK t-AML accounts for 20-25% of t-AML cases.9,14 A recent study by the Bone Marrow Pathology Group showed that there were significant differences in the frequency of FLT3, NPM1, KRAS, and GATA2 mutations in NK t-AML compared to de novo NK AML and the t-AML patients had inferior overall survival and relapse free survival.48 Multiple studies have shown frequent “secondary type” variants (SRSF2, U2AF1, SF3B1, ZRSR2, ASXL1, EZH2, BCOR, STAG2) and CH-associated variants (DNMT3A, ASXL1, TET2) in this group as well.48,50 Other studies have speculated that the inferior survival observed in NK t-AML may be secondary to older age at diagnosis and increased co-morbidity, compounded by complications from prior malignancy.49

Some authors have suggested that variants in genes involved in DNA repair, drug metabolism, and hematopoietic maturation are associated with increased t-MN susceptibility.30,51, 52, 53 One study by the German AML Study Group evaluated 2,835 patients treated for prior non-myeloid malignancy and found that 7% of patients developed t-AML after therapy (4-year median latency).14 In the same study, 3% of patients developed AML who never received therapy (5-year median latency) and postulated that some patients who developed t-AML may have been predisposed because of an inherited susceptibility rather than from therapeutic exposure. Germline variants in DNA damage response pathway genes have been associated with increased risk of t-AML (e.g., BRCA1, BRCA2, BARD1, TP53, RAD51, HLX1, BCL2L10, Fanconi genes), supporting an effect of cancer susceptibility to AML risk in some patients who received cytotoxic therapy. Single nucleotide polymorphism (SNP) array studies performed on nonmalignant DNA from patients diagnosed with t-MN have also identified significant SNP associations that suggest that prior cytotoxic therapy exposure is a potent modifier of t-MN susceptibility. Approximately 15 SNPs were significantly associated with t-MN, and SNPs involving TLE4, NRXN1, and ACCN were most significantly associated with loss of chromosome 5 and/or chromosome 7 in patients with prior alkylating agent exposure. However, these specific variant associations with t-MN clinical risk have not been extensively evaluated.30