Molecular Features Associated with Response to Enasidenib Plus Azacitidine in Newly Diagnosed IDH2-Mutated Acute Myeloid Leukemia

Molecular Features Associated with Response to Enasidenib Plus Azacitidine in Newly Diagnosed IDH2-Mutated Acute Myeloid Leukemia

Enasidenib is an oral, small-molecule inhibitor of mutant-IDH2 enzymes that suppresses 2-HG production and has been shown to reverse the leukemogenic blockade of cell differentiation [6].Enasidenib is approved in the United States for the treatment of patients with mutant-IDH2 relapsed or refractory (R/R) AML [7].Enasidenib monotherapy has demonstrated clinically meaningful improvements in morphologic response and event-free survival (EFS) compared with conventional salvage regimens in older patients with R/R AML [8] and clinical activity in newly diagnosed (ND) AML [9].Azacitidine, a hypomethylating agent (HMA), has been shown to improve overall survival (OS) in patients with ND-AML [10][11][12].
In this study, we investigated the molecular features associated with outcomes in the phase 1b/2 AG221-AML-005 trial [14].The objectives of this analysis were to evaluate changes in 2-HG and IDH2 variant allele frequency (VAF) levels from baseline; to characterize the baseline genetic landscape, including tumor mutation burden; and to evaluate changes from baseline in VAF for genes in the DNA methylation, receptor-tyrosine-kinase (RTK), and RAS canonical pathways, in association with clinical response to enasidenib + azacitidine and azacitidine only.

Materials & Methods
Eligibility criteria and study design have been previously reported [14].In the phase 2 portion of the trial, adult patients with mutant-IDH2 ND-AML were randomized 2:1 to receive enasidenib (100 mg/day) in combination with azacitidine (75 mg/m 2 /day ×7 days) or azacitidine only, in repeated 28-day treatment cycles.Review boards and ethics committees approved the trial protocol and amendments at all participating sites; patients provided informed consent prior to the study [14].

Measurement of 2-HG and IDH2 VAF
Total 2-HG was quantified by liquid chromatography-tandem mass spectrometry (Covance; Princeton, NJ), according to an analytically validated method.Peripheral blood plasma was isolated from patients at baseline, cycle (C)1-day (D)15, C2D1, C2D15, and on D1 of every other cycle from C3-C19.
Bone marrow mononuclear cells were isolated from patients at baseline, and on D1 of C2, C3, C5, C11, C17, and C23, and at end of treatment (EOT).IDH2 VAF was assessed in DNA by digital polymerase chain reaction (Sysmex; Baltimore, MD; lower limit of detection 0.02%-0.04%).
Extent of change in 2-HG and IDH2 VAF during treatment was evaluated in association with clinical response.Clinical response categories included complete remission (CR), less than CR (LTCR; non-CR response, including complete response with incomplete platelet recovery, complete response with incomplete blood count recovery, partial response, and morphologic leukemia-free state), per International Working Group (IWG) response criteria [15], and no response (NR; stable disease [IWG] or progressive disease).

Identification of co-occurring mutations
Bone marrow mononuclear cell DNA was isolated from patients at baseline and co-occurring gene mutations were identified by targeted next-generation sequencing using a 37-gene myeloid panel (ArcherDx; Boulder, CO) at a 1% level of detection (LOD).Manufacturer-recommended filter set was used to identify mutation variants.The baseline genetic landscape, including tumor mutational burden, was evaluated in association with clinical response (defined above).The baseline mutational burden was categorized as greater than or equal to or less than the median number of mutated genes for the entire population.

Gene pathway analysis
A gene pathway is considered mutated if there are mutations present in any of its genes.VAF aggregation was performed by taking the maximum value among all mutations in genes belonging to a given pathway.Gene VAFs was assessed at baseline and on D1 of C2, C3, C5, C11, and C17.Changes in gene VAFs during treatment were evaluated in association with clinical response (defined above).

Statistical analysis
Statistical analyses were performed using Prism version 8.0.0 (GraphPad; San Diego, CA) and the R survival (v. 3

2-HG and IDH2 VAF
Levels of 2-HG at baseline were similar between treatment arms [14] and in patients exhibiting R140 versus R172 IDH2 variants (Figure 1A).The relationship between baseline 2-HG levels and clinical response (as defined by best response) was evaluated in biomarker-evaluable patients with available bioanalytical data treated with enasidenib + azacitidine or azacitidine only.No statistically significant differences were observed in baseline 2-HG between clinical response categories of CR, LTCR, and NR in either treatment arm (enasidenib + azacitidine: P=0.3325, azacitidine only: P=0.8291; Figures 1B and 1C).
Associations between 2-HG and IDH2 VAF were evaluated using Spearman nonparametric correlation.A modest positive correlation was noted between baseline 2-HG and IDH2 VAF in patients with ND-AML with paired baseline 2-HG and IDH2 VAF data (Spearman r=0.4199,P<0.0001; Figure 3A).
The effect of treatment with enasidenib + azacitidine, or azacitidine only, on 2-HG and IDH2 VAF was assessed longitudinally in a subset of patients with ND-AML (2-HG: n=60 and n=28, respectively; IDH2 VAF: n=47 and n=21, respectively).In the combination arm, a significant (P ≤ 0.001) reduction in median 2-HG from baseline was observed by C1D15, and 2-HG levels remained decreased throughout treatment; decreases in 2-HG were statistically significant (P ≤ 0.01) through C9D1 when compared with baseline (Figure 1D).In the azacitidine only arm, 2-HG was not significantly reduced from baseline at any time (Figure 1E).Significant   (P ≤ 0.05) reductions from baseline were observed for IDH2 VAF by C3D1 through C17D1 in the combination arm (Figure 2E).Increasing IDH2 VAF was seen at C17D1 and beyond, with median EOT VAFs comparable to those at baseline.In contrast, IDH2 VAF was significantly reduced only at C11D1 with azacitidine only (P ≤ 0.05; Figure 2F).
There was no definitive correlation between minimum onstudy 2-HG and IDH2 VAF in the combination arm (r=0.2464,P=0.0671; Figure 3B), perhaps because most patients had robust reductions in 2-HG in response to treatment, whereas IDH2 VAF reductions were variable and required longer treatment duration.

IDH2 clonality
Clonal hierarchy and clonal evolution can potentially influence treatment outcomes [16][17][18][19].Thus, we evaluated the position of mutant-IDH2 in the clonal hierarchy by exploring baseline VAFs of co-occurring gene mutations relative to IDH2, and explored changes in the clonal landscape during treatment with enasidenib + azacitidine.IDH2 was defined as subclonal if any co-mutation VAF was greater than IDH2 VAF; otherwise, it was considered clonal.While IDH2 was primarily clonal to most other genes (but subclonal to DNMT3A) at baseline, IDH2 VAF reductions during combination treatment altered tumor clonality, with other driver genes (e.g., ASXL1, CEBPA, TET2) becoming clonal to IDH2 on-study (Figure 6).

Discussion
The current investigation characterizes molecular features associated with clinical response to combination therapy with enasidenib + azacitidine in patients with ND-AML.Our findings demonstrate a robust clinical validation of the synergistic effects of enasidenib + azacitidine seen in vitro [13].
Baseline levels of 2-HG or IDH2 VAFs were similar irrespective of IDH2 variant (R140 or R172) and were not predictive of best clinical response to enasidenib + azacitidine or azacitidine only.Longitudinal analysis revealed rapid reductions in 2-HG and IDH2 VAF during treatment with enasidenib + azacitidine; reductions were more rapid and robust with combination treatment versus azacitidine only.While higher levels of 2-HG were associated with higher IDH2 VAFs at baseline (possible indicator of disease burden), no correlation was observed between minimum 2-HG and IDH2 VAF in patients treated with enasidenib + azacitidine; levels of 2-HG were robustly reduced in most patients and IDH2 VAFs were more variable in response to treatment.
IDH2 VAF clearance was significantly more likely in patients with CR versus LTCR in both treatment arms, and more common in those with IDH2-R140 mutations than IDH2-R172.In both combination and azacitidine only treatment arms, 50% of those who achieved CR versus 4-8% who did not achieve CR had minimum on-treatment IDH2 VAFs <1% (P≤0.01).The observed association between IDH2 VAF clearance and CR suggests that clinical response is likely driven by clearance of IDH-mutated blasts regardless of the type of targeted therapy.However, it is important to note that the cohort size was limited.
As part of this analysis, we characterized the baseline genetic landscape of patients with IDH2-mutated ND-AML, including tumor mutation burden, and evaluated the association with clinical response.Consistent with other reports [20,21], IDH2-R140 and IDH2-R172 showed distinct co-mutational profiles at baseline, with SRSF2 preferentially co-mutated with IDH2-R140 and DNMT3A preferential for IDH2-R172.Mutational burden at baseline was similar in those with IDH2-R140 and IDH2-R172.CRs were attained in patients treated with enasidenib + azacitidine irrespective of co-occurring gene mutations, suggesting no significant association between comutations and response to therapy.
In patients with RAS-pathway mutations, combination treatment led to significant VAF reductions for those who attained CR and significantly prolonged EFS compared with azacitidine only.Mutations in RAS pathway genes (PTPN11, NRAS, KRAS), which occur in ~10-25% of patients with AML, are historically associated with poor outcomes and resistance to targeted therapies [6,[22][23][24][25], including primary resistance to enasidenib in R/R AML [6].In ND-AML, a study reported no association between the RAS-pathway mutations and decreased likelihood of response to enasidenib; however, the patient number was low [9].In this trial, the combination of enasidenib + azacitidine promoted morphologic response across all gene pathways, including RAS [14], and no individual gene in the RAS pathway was significantly associated with achievement of CR.As RAS-pathway mutations are known to induce resistance to enasidenib [6,25], but not azacitidine, the addition of azacitidine may reduce the risk of resistance to enasidenib in this setting.This hypothesis is further supported by the data from the AGILE study which showed that a subgroup of patients with IDH1-mutated AML plus RAS-pathway mutations were more likely to achieve CR to combination therapy with ivosidenib (a mutant IDH1 inhibitor) and azacitidine versus placebo and azacytidine [26].
Azacitidine has previously demonstrated efficacy in older patients with AML ineligible for intensive chemotherapy, and in combination with enasidenib in newly diagnosed patients and those with prior exposure to HMAs or enasidenib [10][11][12]14,27].However, to date, no long-term data have been published evaluating enasidenib + azacitidine with respect to extended durations of response, disease recurrence, and clonal selection in patients with IDH2-mutated AML (median follow-up of previous studies: 13.1-18.5months) [14,27].Selection pressure leading to acquired drug resistance has been reported for several targeted AML agents, including enasidenib and ivosidenib [28].In a subanalysis of the phase 1/2 AG221-C-001 trial (NCT01915498) assessing enasidenib monotherapy in patients with IDH2-mutated AML, it was shown that clonal selection or evolution of terminal or ancestral clones, as opposed to second-site mutations in the same IDH2 allele, resulted in resistance to enasidenib [29].In addition, several studies have demonstrated that resistance to azacitidine can also arise, yet the exact molecular mechanism is currently unknown [30].As such, doublet and triplet combination strategies that can target multiple mechanisms or clones driving AML disease progression are of particular interest [28][29][30][31].Recently, a triple combination of decitabine (an oral HMA) + venetoclax + enasidenib prolonged the duration of remission in a small number of patients who experienced disease progression with decitabine + venetoclax doublet therapy in a prospective phase 2 study (NCT03404193) [32], and is currently being evaluated in patients with R/R IDH2-mutated AML (NCT04774393).The development of alternative combination treatment options may help address complications relating to drug resistance in IDH2-mutated AML.Finally, further evaluation is needed to establish the long-term efficacy and safety of, and potential acquired resistance to, enasidenib + azacitidine in this patient population.

Conclusion
In conclusion, outcomes in this IDH2-mutated ND-AML cohort clinically validate the synergistic effects of enasidenib + azacitidine previously observed in vitro.One-half of patients who achieved CR with enasidenib + azacitidine and azacitidine only had IDH2 VAF reductions to <1%, suggesting that IDH2 VAF clearance may be associated with CR irrespective of therapy.Unlike prior reports of enasidenib or azacitidine monotherapy, combining these agents improved survival in patients with RAS mutations.Given the modest sample size in (28.4% Risueño A, See WL, Prebet T, DiNardo CD, Döhner H, Stein EM, et al.Molecular Features Associated with Response to Enasidenib Plus Azacitidine in Newly Diagnosed IDH2-Mutated Acute Myeloid Leukemia.J Clin Haematol.2023;4(1):20-34.

9 Figure 6 .
Figure 6.Clonal hierarchy of mutated IDH2 at screening and during treatment with enasidenib + azacitidine (A) or azacitidine-only (B).Blue text represents gene mutations that are predominantly clonal compared with IDH2 and red text represents gene mutations that are mostly sub-clonal to IDH2.IDH2: Isocitrate Dehydrogenase-2; pts: patients; VAF: Variant Allele Frequency.