Patients with AML and an IDH2-R172 mutation exhibit a unique initial response to intensive chemotherapy induction Open Access

Consensus guidelines for acute myeloid leukemia (AML) treatment¹ recommend that patients who receive intensive chemotherapy induction undergo a midcycle bone marrow biopsy (BMB) on days 14 to 21. Reinduction is recommended for all patients with evidence of residual disease, under the assumption that these patients are more chemotherapy resistant than those with midcycle blast clearance, and escalation of their treatment may increase response rates. The utility of this practice is controversial and not endorsed at all centers.^2-4 However, even when contested, there is little consideration as to the possibility of different response dynamics among genetically defined subgroups. Similarly, it is expected that patients have count recovery within ∼4 weeks from intensive chemotherapy induction, at which point they will frequently be considered refractory if their bone marrow still has a blast count >5%. There are surprisingly little data on variations in response kinetics among different AML subgroups.

Mutations in isocitrate dehydrogenase 2 (IDH2) are reported in 10% to 20% of patients with AML,⁵ of which close to 20% are mutated at the R172K locus (R172-m) and most of the remaining are in the R140Q residue (R140-m).⁶ Wildtype IDH1/2 produce α-ketoglutarate and reduced NAD phosphate (NADPH) from isocitrate and NADP⁺ and participate in cell regulation. Mutated IDH1/2 converts α-ketoglutarate to the oncometabolite, 2-hydroxyglutarate (2HG).^7,8 It has been found that in IDH-mutated AML, 2HG accumulation alters gene expression, inhibits differentiation, and promotes leukemogenesis.^9,10 Specifically, elevated 2HG has been associated with hypermethylation, DNA repair inhibition, and genomic instability.¹¹ Consequently, myeloid cells become enriched for stem cell markers and differentiation is inhibited.^9,12 

There have been conflicting reports regarding prognosis among patients with AML and IDH mutations.^13-16 Particularly in the case of R172-m, some have suggested that it confers adverse risk, whereas others the opposite. Papaemmanuil et al, for example,¹⁷ provisionally defined R172-m AML as a genomically characterized group, noting that mutations in NPM1 and IDH2-R172 are mutually exclusive but both groups have comparably favorable survival. Although it has been established that the accumulation of 2HG contributes to leukemogenesis in AML with IDH1/2 mutations, there appear to be distinct mechanisms in each of the 3 isoforms (IDH1-R132/R140-m/R172-m), presumably related to wildtype activity dependence, cellular compartment location of the mutated enzyme, and different levels of 2HG. In cells with R172-m, levels of 2HG have been reported to be at least 3 times higher than in R140-m, positively correlating with degree of differentiation arrest.¹⁸ 

Clinical observations prompted an investigation into kinetics of response to intensive chemotherapy among patients with R172-m, after noting high midcycle blast counts and what appeared to be slow achievement of complete remission (CR) in these patients.

A retrospective single-center analysis was conducted among adult patients with newly diagnosed AML with IDH2 mutations, who received intensive chemotherapy induction from March 2012 to December 2023 at Memorial Sloan Kettering Cancer Center, New York, United States. Patients with concurrent IDH1 and IDH2 mutations were excluded, including 1 patient with a mutation in the R377C codon and 2 patients with copy number variations.

Patient charts were manually reviewed for AML diagnosis, IDH2 mutation status, co-occurring mutations, and outcomes. AML diagnosis and risk category were reported based on European LeukemiaNet (ELN) 2022 guidelines, regardless of year of diagnosis. Response criteria as described elsewhere.¹⁹ BMB results were serially documented from time of induction to remission or relapse.

Patients with R140-m were used for comparison to patients with R172-m. Dynamics of blast reduction were compared and correlated with outcome. Statistical analysis was performed in R. Continuous demographic variables were compared using Wilcoxon rank sum test, and categorical demographic variables were compared with χ² and Fisher exact tests. Midcycle BMB blasts were compared with Welch t test. Kaplan-Meier curves were constructed to estimate time to blast clearance (defined as days from induction to first BMB with ≤5% blasts, censored at time of second treatment or death). Fisher exact and log-rank tests were used for associations.

This project was approved by the Institutional Review Board of Memorial Sloan Kettering Cancer Center and was conducted in accordance with the Declaration of Helsinki.

A total of 52 patients were identified, 33 with R140-m and 19 with R172-m. Median age was 63 years and 54% were female. All patients were treated with a “7+3” induction (anthracycline and cytarabine) or CPX-351 (liposomal encapsulation of "7+3"; daunorubicin and cytarabine). Furthermore, 10% received a FLT3 inhibitor on days 8 to 21, and 21% of each group had participated in a clinical trial with the addition of the IDH2 inhibitor enasidenib to “7+3” (see Table 1 for patient characteristics).

There were no differences between the groups regarding age, sex, race, blast percentage at diagnosis, or induction regimen. Consistent with previous studies, the mutational landscape of both groups was different.¹⁴ No NPM1 mutations were found among the R172-m group, compared with 39% among the R140-m. There was a similar occurrence between the 2 groups of RUNX1, ASXL1, and DNMT3A mutations. The R140-m group had significantly more co-occurring SRSF2 mutations. Patients with R172-m had more BCOR mutations, which reached statistical significance. Among the R140-m, there were a number of patients with FLT3 and NRAS mutations, which did not occur among the R172-m in the current cohort. There were no patients in either group with complex karyotypes or TP53 mutations, and there was a similar distribution of adverse risk patients.

A total of 15 patients with R172-m and 26 patients with R140-m had evaluable midcycle BMBs. Most (76%) were performed on days 14 to 15 from induction. A subset of patients with FLT3 mutations who received a FLT3 inhibitor on days 8 to 21 from induction underwent a midcycle BMB circa completion of treatment (days 21-22), according to the RATIFY trial protocol.²¹ There was 1 case of a BMB performed on day 23, which was not considered midcycle because the patient had already recovered counts, and with blast clearance, was considered to be in CR. All midcycle BMBs ranged from days 13 to 22.

A significant difference was found between the 2 groups in midcycle BMB cellularity and in blast count. Among the patients with R172-m, median blast count was 70% (range, 25-87). Among those with R140-m, it was 5% (range, 0-81; P < .001; Figure 1A), comparable to a median of 5% to 10% reported in the general AML population (see Table 2 for comparison).^22,23 Midcycle BMB results revealed a trend toward increased cellularity among patients with R172-m (P = .045; Figure 1B).

Among patients with R140-m, 58% of those with midcycle BMB available for evaluation had ≤5% blasts. This is comparable to blast clearance rates reported in large prospective trials including all AML subtypes of between 45% and 69% meeting similar definitions^3,22-25 (see Table 2 for comparison).

Strikingly, none of the midcycle BMBs of patients with R172-m had blast clearance.

Treatment decisions based on midcycle BMB results varied, depending on patient status and provider preference. None of the 15 patients with R140-m who had blast clearance on midcycle marrow received reinduction. A total of 26 patients had blast counts above 5%; all 15 of those with R172-m who had midcycle marrow evaluations (100%) and 11 of those with R140-m (42%). Among these, a total of 7 patients received reinduction based on their midcycle BMB, 4 patients with R140-m and 3 with R172-m. The regimens used were standard salvage protocols, including FLAG-IDA (fludarabine, cytarabine, and idarubicin), MEC (mitoxantrone, etoposide, and cytarabine), and high- or intermediate-dose cytarabine (HiDAC/IDAC). Two patients received the low-intensity regimen of azacitidine and venetoclax as second line. Of the remaining 19 patients, the decision was to wait, either with repeating BMBs with short intervals, or with a plan to repeat a BMB on count recovery. The reasons cited for refraining from reinduction were as follows (in order of frequency): ongoing infection and/or patient comorbidities, evidence of significant blast reduction compared with diagnosis, or for technical or logistical reasons. Continued treatment depended on outcomes. Nine patients achieved CR without additional therapy and 10 received second-line therapy based on subsequent biopsies. See supplemental Figure 1 for a detailed breakdown of continued treatment.

Among the total of 32 patients who achieved CR after induction alone, there were 8 with R172-m. Of these, half exhibited a pattern of delayed achievement of CR, defined as CR after day 40, with evidence of gradual blast reduction (CR on days 42, 43, 45, and 53 from single induction, with no additional therapy). Of the 24 patients with R140-m who achieved CR after a single induction, there was 1 such case of delayed CR (with CR on day 42). The blast reduction dynamics among the “delayed CR” patients was documented with short-interval repeat BMB and coincided with gradual count recovery. In several of these cases, providers had refrained from reinduction due to speculation regarding the possibility of delayed CR. Count recovery generally coincided with blast reduction. See Table 3 for details on patients with delayed and gradual blast reduction (gradual blast reduction includes patients with gradual reduction of blasts and CR before day 40 or in 1 case, after second-line treatment). Figure 2 reveals a graphical depiction of blast count reduction among those who achieved CR after a single cycle of induction.

Overall, it took significantly longer for patients with R172-m to achieve blast clearance after a single induction, compared with those with R140-m (median days to blast clearance 44 vs 24; P = .017; Figure 3A).

Regarding cellularity of the BMB, 2 of the delayed CR patients had initially hypercellular BMB that became less cellular with blast reduction and the remaining had hypocellular marrows that for the most part became normocellular with count recovery (Table 3).

Despite midcycle marrow blasts and longer time to blast clearance among patients with R172-m, there was no significant difference in overall survival or outcomes compared to patients with R140-m. The 24-month overall survival was 78%; 74% among patients with R140-m (95% confidence interval, 60-91) and 89% among those with R172-m (95% confidence interval, 76-100) (Figure 3B). Furthermore, 79% of patients with R172-m achieved CR after 1 or 2 cycles of treatment vs 85% among those with R140-m. A higher proportion of patients with R172-m underwent allogeneic stem cell transplant (79% vs 61% among R140-m), and there were more relapses among patients with R140-m (40% vs 18% among R172-m). Among patients with R140-m, 54% (18) achieved long-term CR (defined as CR for more than a year), compared with 68% (13) of patients with R172-m (see Table 4 for outcomes).

There was no effect of induction intensity on midcycle blast counts or days to blast clearance, when comparing patients who had received induction with daunorubicin 90 mg/m² vs those who received 60 mg/m² (supplemental Figure 2).

Among the patients with R140-m, there were 13 (39%) who were considered favorable risk, according to ELN risk stratification¹⁹ (11 with NPM1 mutations, 2 with core binding factor translocations). None of the patients with R172-m were classified as favorable risk. In a sensitivity analysis with removal of ELN favorable risk patients, there remained a significant difference between the 2 groups in midcycle BMB blast count, with a median of 6% blasts among the R140-m, and 70% among the R172-m (supplemental Figure 3). With regard to days to blast clearance, there was a small difference between the groups, but this did not reach statistical significance (supplemental Figure 4).

The patients who received enasidenib in addition to standard intensive chemotherapy induction as part of a clinical trial were distributed evenly among the 2 groups, 21% of either group, a total of 11 patients (7 with R140-m and 4 with R172-m). These patients had results similar to those found in the entire cohort, as demonstrated in supplemental Table 1. There was only 1 patient who had a clinically and morphologically suspected episode of differentiation syndrome (DS; this was a patient with R172-m, who continued with second-line therapy until achieving CR on day 86). In a sensitivity analysis with removal of the patients who had received enasidenib in addition to 7+3 induction, there remained a significant difference between the 2 groups both in midcycle BMB blast count (supplemental Figure 5) and in days to blast clearance (supplemental Figure 6).

Taken together, although not a single patient with R172-m had midcycle blast clearance, and despite significantly slower responses to intensive chemotherapy induction compared with the R140-m group, there were no differences in outcomes, which were overall favorable.

One can hypothesize as to why patients with R172-m may have slower responses to intensive chemotherapy, particularly when compared with patients with R140-m.

Significantly higher levels of 2HG¹⁸ and fewer comutations among patients with R172-m^14,17 suggest that 2HG plays a crucial role in leukemogenesis. Knock-in of R172-m to murine hematopoietic cells resulted in a more severe myeloid disease phenotype compared with R140-m.¹² 

Mason et al²⁶ reported on patients with IDH-mutated AML who were treated on the aforementioned trial investigating the addition of IDH inhibitors to a 7+3 induction (most of whom were not part of the current cohort). Their series included patients with IDH2 and IDH1 mutations; 18% with IDH2-R172 mutations. Investigators noted what they termed a differentiation-like response in approximately a third of the patients overall and in two thirds of the R172-m patients. The differentiation pattern was defined as midcycle BMB blasts ≥5%, cellularity ≥10%, and morphologic and/or flow cytometry midcycle findings of what could be considered “monocytic differentiation” or “granulocytic differentiation.” Included in this category were only the patients who later achieved CR after 1 or 2 cycles of chemotherapy. Clinical signs of differentiation syndrome (DS) were not required and did not occur in most of these patients. Mason et al assumed the differentiation-like response to be mediated by the addition of IDH inhibitors to intensive chemotherapy induction. However, when evaluating response patterns of the current patient cohort, most of whom did not receive the addition of IDH inhibitors to intensive induction, it appears that many of the patients, particularly those with R172-m, exhibited response patterns similar to those described as differentiation by Mason et al (Table 3).

Intensive chemotherapy alone is generally not assumed to cause differentiation; however, it has been noted in the past. Thus, for example, in 1985, Kantarjian et al²⁷ described 5 patients with acute promyelocytic leukemia (APL) who received intensive chemotherapy induction, in the pre-all-trans retinoic acid/arsenic era. These patients exhibited a slow and gradual blast reduction in sequential BMBs until achieving delayed CR. Their description did not include any clinical characteristics, aside from gradual improvement in signs of coagulopathy and gradual count recovery. Investigators then presented a retrospective analysis including 32 patients with APL and 134 patients with AML, who all achieved CR after induction with intensive chemotherapy alone. Among the patients with APL, 70% had hypercellular midcycle BMBs with >35% blasts and promyelocytes, whereas this occurred in only 15% of the non-APL patients with AML who achieved CR. Kantarjian et al hypothesized that differentiation in response to intensive chemotherapy may be a mechanism of remission induction in certain subtypes of AML, such as APL, possibly related to intrinsic cell biology. They concluded that recognition of this phenomenon can improve management of APL.

We hereby propose a similar hypothesis regarding response kinetics of patients with R172-m. If the primary cause of differentiation arrest in R172-m patients is elevated levels of 2HG, it is possible that cytotoxic elimination of leukemia cells causes reduction in extracellular 2HG levels, which can theoretically explain a differentiation-like response occurring with the release of the differentiation block. It seems that this pattern of response does not correspond to the clinical inflammatory-like syndrome associated with DS or with classic DS morphology findings of granulocytic differentiation and a hypercellular marrow.^28,29 There was only a single case in the current study of clinically suspected DS, and this occurred in a patient who had received enasidenib in addition to 7+3 induction. The hypercellularity of the BMB does not appear to be as salient a feature as in patients with APL, as many of the patients with R172-m had a hypoplastic midcycle BMB, although their marrows were overall slightly more cellular than those with R140-m. The current phenomenon may be more accurately described as delayed or gradual achievement of CR rather than differentiation per se.

Future studies should evaluate the correlation between 2HG levels and blast reduction, to better understand the role of 2HG in differentiation arrest and its reversal and identify predictors of pending response vs refractory patients. It is also unclear whether a delayed CR response affects measurable residual disease measurement postinduction and its interpretation.

In this study, there were 2 patients with R140-m who exhibited gradual blast reduction or delayed CR (one of whom had been treated with concurrent enasidenib). If this response is mediated by initially high 2HG levels, one can consider the possibility that it occasionally occurs in patients with R140-m or IDH1 mutations.

In subgroup analyses, there was no effect of induction type or addition of enasidenib to current findings. As mentioned previously, and in line with previous studies,^14,17 the mutational landscape was different between the 2 groups. A sensitivity analysis demonstrated that the early blast clearance among patients with R140-m was partially driven by patients with ELN favorable risk, although there was still a significant difference in midcycle BMB blast count between the 2 groups. Moreover, as demonstrated in Table 2, results among the R140-m were similar to those found in the general AML population in terms of median blast count in midcycle marrows, suggesting that responses found among the R172-m were the anomaly. Late blast clearance has been shown to be adversely correlated with outcomes, independent of ELN risk status.³⁰ The late blast clearance currently demonstrated among patients with R172-m highlights the uniqueness of their response, as there was no difference in overall survival compared with the R140-m.

Limitations of this study are its retrospective nature and small sample size, as R172-m is a rare AML subtype. One of the practical questions that arises is what harm would come from withholding additional treatment in these patients until a determination can be made whether they are refractory or having delayed CR. Current study results suggest that waiting was not detrimental, but this must be validated, as most patients received reinduction when thought to have an inadequate response to intensive chemotherapy. Double induction has been associated with significantly higher morbidity compared with single induction, in large clinical trials.²⁴ The risks of receiving more chemotherapy than is needed for CR achievement in patients with R172-m should be weighed against the risks of waiting, preferably prospectively evaluated.

To conclude, reported here is a unique and yet undescribed response pattern to intensive chemotherapy induction in patients with AML and R172-m. The pathophysiology and extent of this phenomenon should be further characterized and validated. These findings suggest caution in interpretation of midcycle BMBs in these patients and challenge the prevailing understanding of response dynamics to myeloablative chemotherapy. Clinicians should consider repeating BMBs with short intervals to assess for gradual blast clearance. Importantly, awareness of the possibility for slower-than-expected blast reduction in patients with R172-m can potentially minimize unnecessary additional cycles of chemotherapy.

The authors thank Emily F. Mason and Michael R. Savona for sharing information.

Contribution: M.Y.S. and E.M.S. designed the study; M.Y.S. and A.R.T. wrote the manuscript; M.Y.S., A.R.T., K.-K.C., Y.K.V., and L.B. performed the research and analyzed the data; A.D. and D.N. performed the statistical analysis; T.S. and W.X. revised pathology; C.F. assisted with data collection; J.C., J.M.R., M.S.T., and E.M.S. provided expert advice and manuscript revision; and all authors participated in the editing of the manuscript and approved the final version.

Conflict-of-interest disclosure: M.Y.S. received consultancy fees from Intellisphere, LLC and Sobi. Y.K.V. received consultancy fees from EastRx. J.M.R. received consultancy fees from BioSight. M.S.T. served on advisory board for Foghorn Therapeutics, SDK Therapeutics, HOVON DSMB, and Moleculin Biotech. E.M.S. received consultancy fees from Bristol Myers Squibb, Servier, and Agios. The remaining authors declare no competing financial interests.

Correspondence: Eytan M. Stein, Leukemia Service, Department of Medicine, Memorial Sloan Kettering Cancer Center, 530 E 74th St, New York, NY 10021; email: steine@mskcc.org.