A baseline transcriptomic signature predictive of clinical response to menin inhibitors in AML : The PRE-men retrosprospective Study Free

Menin inhbitors (MENi) have been developed mainly in KMT2-r and NPM1cmut aute myeloid leukemia (AML) and showed encouraging results in relapse/refractory (R/R) settings. Overall response rate (ORR) of MENi monotherapy in AML is usually around 50% with a median duration of response of 4-5 months. However, there are currently no predictive markers of response to MENi. In this study, we assessed the transcriptomic profiles of R/R AML patients at baseline and during early MENi therapy to: (1) develop a gene expression signature predictive of treatment response, and (2) gain insight into the molecular features distinguishing sensitive from primary resistant patients.

Methods In this monocentric retrosprospective study, patients treated with MENi monotherapy (revumenib (n=15), ziftomenib (n=3), bleximinib (n=1), enzomenib (n=1)) either in French revumenib compassionate use program or in dedicated clinical trials (2020-004104-34, 2023-510509-17, 2023-505584-36, 2022-502741-10-00) were included. ORR was defined as patients achieving CR, CRi, CRh or MLFS at any time. Patient samples were collected at baseline (prior to MENi initiation), at early time points (day+15 and/or day+30) and whenever possible during follow-up up to disease progression/relapse. We performed targeted RNA sequencing on patient-derived mononuclear cells using a custom panel of 1219 genes involved in cancer or hematological disorders. Differential gene expression analysis was performed using DESEQ2 with a false discovery rate (FDR) cutoff of 0.05. Over-representation analysis was performed using KEGG and GO-BP signature databases.

Results Overall, 20 patients were included. The median age was 58.5 (range: 6-75) years. Eleven patients had NPM1c mutation, 6 had KMT2A rearrangement (-r) and 3 had NUP98-r. The median number of prior treatment lines was 2 (range: 1-3). After a median follow-up of 2.3 months (range: 0.9-10.2), ORR was 50% with a median time to response of 1.4 months.

Differential gene expression analysis at baseline (D0) between overall responders (ORR) and non-responders revealed 50 significantly differentially expressed genes (DEG), with an equal distribution between up (n=25) and down (n=25) regulated genes. Over-Representation Analysis (ORA) revealed that most of the differentially expressed genes were involved in adaptive (mainly TH1, TH2, TH17 and T-cell receptor) and innate (NK-kappa B, NK-cell cytotoxicity and JAK-stat) immune signaling pathways, with also an over-representation of oncogenic (PI3K-Akt and Ras) signaling pathways. A gene-expression signature at baseline based on these 50 genes showed a good discrimination of responders (AUROC [95%CI] 0.97 [0.9-1]). This signature was progressively lost at day+30 (AUROC 0.63 [0.37-0.89]), with only 5 DEG conserved between D0 and D30. As previously reported, we observed downregulation of targeted genes such as HOXA3, HOXA5, HOXA9 and MEIS1 associated with an CD14 upregulation in responding patients. However, baseline levels of HOX/MEIS cluster was not predictive for response.

When focusing on patients with NPM1^mut, differential gene-expression analysis at baseline between responders and non-responders found 18 DEG, all overexpressed in the responder group and mostly involved in the PI3K-Akt and JAK-STAT signaling pathways. HIF1A was the most overexpressed gene in responding patients (log2FC 10.72, adjusted p-value <0.0001).

Regarding KMT2A-r patients, differential gene-expression analysis at baseline between responders and non-responders found 157 DEG, including 52 upregulated and 105 downregulated, mostly involved in adaptive and innate/inflammatory immune signaling pathways. Interestingly, only 1 DEG (DKK2) was conserved across the NPM1c^mut and KMT2-r patients.

Overall, 9/10 responding patients ultimately relapsed after a median time of 6.43 months. Only one case of menin mutations (M327I (27%) and G331D (28%) detected on the same allele) was found at relapse (D+160). At baseline, both mutations were undetectable with (500X sequencing depth).

Conclusion These results show thatdifferent transcriptomic profiles can be distinguished between responders and non-responders to MENi at baseline. Most of the DEG found were involved in immune signaling pathways. A transcriptomic signature predictive of response to MENi therapy is currently evaluated in an independent external validation cohort and will be presented at the meeting.