Paper Details

ABSTRACT

Introduction

Mutations (mut) in RUNX1 encoding the RUNX1 transcription factor occur in ~10% of cases of acute myeloid leukaemia (AML). RUNX1^mut is considered an adverse prognostic marker in the context of standard intensive induction chemotherapy (IC), but does not portend a worse prognosis, and may even have a favorable impact, following treatment with venetoclax-azacitidine (ven-aza). RUNX1^mut are heterogeneous, including frameshift and missense variants. The mechanistic underpinnings of ven-aza responsiveness in RUNX1^mut AML remains unknown, thus, we characterized the impact of different RUNX1 variants on clinical outcomes following IC vs. ven-aza.

Methods

For correlative analyses, RNA-sequencing of primary CD34+ AML cells was used for differential gene expression. Doxycycline inducible HA-RUNX1 frameshift proteins were expressed in OCI-AML5^+/+ (CRISPR-Cas9 repaired) cells. Mutant-protein stability and integrated stress response were assessed with immunoblotting. Unfolded protein was measured by TPE-MI fluorescence and oxidative phosphorylation by seahorse assay. Venetoclax dose-response were assessed in both RUNX1-mutated primary cells and the OCI-AML5 models.

To evaluate clinical outcomes, we assembled a cohort of patients with RUNX1^mut AML (n=510) treated with ven-aza (n=134; University of Colorado and PRISM cohorts) or IC (n=376; University of Colorado and UK-NCRI cohorts- Tazi et al. 2022). Included patients were classified into one of 4 biological RUNX1 variant subclasses: missense (n=202), truncating-frameshift (TF; n=113), extending-frameshift (EF; n=81), or other (nonsense, splice-site, or indeterminate; n=114). In 77 patients, more than one RUNX1 mutation was identified. These cases were classified by their dominant phenotype from the functional data or classified as ‘other’ if a single dominant class was not evident in (n=34) (retained within 'other'). Clinical endpoints included overall response rate (ORR = CR+CRi+MLFS by ELN 2022) and overall survival (OS). OS was calculated from treatment start for the ven-aza cohort and diagnosis for the IC cohort to match the publicly available OS origins. Comparisons made by Fisher exact test, Kaplan-Meier with early-weighted Gehan-Wilcoxon test (to account for larger N follow-up at earlier timepoints), and Cox regression for hazard ratios (HR) as indicated.

Results

RNA-sequencing of primary AML cells with TF mutations showed enrichment of a mitochondrial gene signature and, in isogenic TF mutated OCI-AML5 cell models, increased spare oxidative-phosphorylation capacity compared to OCI-AML5^+/+ cells. TF mutations encoded highly unstable proteins that drove PERK-dependent activation of the integrated stress response and proteasomal degradation; venetoclax further overloaded TF mutated cells with unfolded protein. This produced selective venetoclax sensitivity: primary TF mutated blasts were more sensitive than missense or non-RUNX1 mutated cells (mean IC₅₀ 137 vs. 595 vs. >3000 nM; 85.5% vs. 68.9% vs. 39.9% blast kill at 2 µM, approximating peak plasma levels), predicting that TF (but not EF or missense) mutations would sensitize patients to venetoclax.

Consistent with these results, patients with TF mutations had the highest response rate to ven-aza: ORR 92% vs. 68% for missense (Fisher p=0.024) with EF intermediate (80%). TF mutations were associated with longer OS than missense (median 20.4 vs. 9.0 months, Gehan-Wilcoxon p=0.026; 12-month OS 70% vs. 47%). At 1 year, TF mutants carried less than half the hazard of death of missense mutants (HR 0.44, 95% CI 0.20-0.96; p=0.039). By contrast, OS did not differ significantly by mutation class under IC (missense vs. TF Gehan-Wilcoxon p=0.23; 1-year HR 0.91, 95% CI 0.59-1.40, p=0.67), indicating the TF mutant advantage was specific to venetoclax-based therapy. Among ven-aza responders, relapse rates were similar across classes (43-46%; TF vs. missense Fisher p=1).

Conclusions

RUNX1 truncating-frameshift mutations, shown functionally to yield unstable proteins, proteostatic stress, and induce a mitochondrial-adaptation response, were associated with the improved response and survival on venetoclax-based therapy, most clearly relative to missense mutations. Mutation class was not prognostic following intensive chemotherapy, consistent with a venetoclax-specific effect. These findings support prospective evaluation of ven-aza vs. intensive chemotherapy stratified by RUNX1 mutation class.