Prognostic Relevance of Truncating ASXL1 Mutations and ASXL1-Integrated Molecular Models in Proliferative CMML

Introduction

Chronic myelomonocytic leukemia (CMML) can be classified into dysplastic (dCMML) and proliferative (pCMML) subtypes based on a white blood cell count (WBC) ≥ 13 x 10 ⁹/L for pCMML. Though this cutoff may seem arbitrary, there is evidence that pCMML is a unique biological entity with a distinct molecular profile and worse outcomes (Carr RM, et al. Nat Commun. 2021). We hypothesized that contemporary CMML prognostic scoring systems - such as the Mayo Molecular Model (MMM; Patnaik MM, et al. Leukemia. 2014) and the CPSS-Molecular Model (CMM; Elena C, et al. Blood. 2016) - would not adequately risk-stratify pCMML patients. Moreover, while truncating ASXL1 mutations are considered a high-risk feature in CMML, their prognostic relevance in pCMML is unknown.

Methods

After IRB approval, we compiled a large (n = 888), molecularly-annotated database of CMML patients seen at two US medical centers (Mayo Clinic and MD Anderson) via retrospective chart review. All statistical analyses considered the clinical and laboratory parameters obtained at the time of presentation. The cohort was divided into dCMML and pCMML subgroups, which were analyzed independently alongside the entire cohort. Categorical variables were compared by Fisher exact or Pearson χ ² tests and continuous variables by Mann-Whitney U tests. Univariate and multivariate analyses were performed using Cox proportional hazards regression models. Survival was assessed via the Kaplan-Meier method. P-values < 0.05 were considered significant. Calculations were performed using the BlueSky Statistics (v10.3.1) interface for R.

Results

We identified 474 (54%) dCMML cases and 410 (46%) pCMML cases with median age 71 years and 67% males. Patients with pCMML were more likely to have higher AMC (5.1 vs 1.5 x 10 ⁹/L, p < 0.0001), circulating immature myeloid cells (IMC; 65% vs 36%, p < 0.0001) and abnormal karyotypes (37% vs 30%, p = 0.0187). They were also overrepresented in higher-risk categories of the MMM and CMM compared to dCMML (p < 0.0001 for each model). The number of somatic mutations was higher in pCMML than dCMML (mean 3.0 vs 2.6, p = 0.0050). Mutations in ASXL1 (53% vs 39%), NRAS (21% vs 11%), CBL (18% vs 12%), SETBP1 (14% vs 7%), JAK2 (10% vs 3%), CEBPA (5% vs 2%), GATA2 (3% vs 0%), and FLT3 (3% vs 1%) were more frequent while mutations in TET2 (38% vs 48%), SF3B1 (2% vs 8%), ZRSR2 (2% vs 8%), and NPM1 (1% vs 3%) were less frequent in pCMML compared to dCMML, respectively (p < 0.05 for all).

Though the MMM and CMM effectively stratified the dCMML group, they did not provide meaningful risk stratification in the pCMML group (MMM p < 0.05 overall but could not distinguish between low and intermediate-1/2 or intermediate-1 and high risks; CMM p > 0.05). Likewise, there was no difference in the median overall survival (OS) or leukemia free survival (LFS) when pCMML patients were stratified by truncating ASXL1 mutations( Figure 1A).

We then sought to better risk stratify pCMML. Univariate analysis identified increasing age, male gender, abnormal karyotype, and elevated WBC, ANC, and AMC as adverse predictors for OS. In addition, the presence of IMC and peripheral blood (PB) blast % were adverse factors for LFS, while age was not significant for LFS. None of the aforementioned mutations were significant in the univariate model for OS, whereas GATA2 and NPM1 mutations were significant for LFS. In multivariate analysis for OS, only male gender and AMC ≥ 10 x 10 ⁹/L retained significance. In multivariate analysis for LFS, male gender, AMC ≥ 10 x 10 ⁹/L, and peripheral blasts ≥ 5% remained significant; GATA2 and NPM1 mutations were not significant. These variables risk-stratified pCMML patients based on the presence of 0, 1, or ≥ 2 risk factors ( Figure 1B). In comparison, OS was 40.7 months (95% CI 35.2 - 50.1 mo.) for dCMML patients, 20.1 months (95% CI 18.2 - 22.7 mo.) for pCMML patients, and 31.8 months (95% CI 28.4 - 33.5 mo.) for the entire cohort. At last follow up, 586 deaths (66%) and 168 AML transformations (19%) were recorded in the cohort. Deaths in dCMML vs pCMML were 294 (62%) vs 291 (71%; p = 0.0067), with AML transformations in 83 (18%) vs 84 (21%; p = 0.2055) patients, respectively.

Conclusions

Truncating ASXL1 mutations, along with ASXL1-mutation weighted CMML prognostic models failed to effectively risk stratify pCMML patients. In pCMML, male gender and AMC ≥ 10 x 10 ⁹/L, effectively risk stratified for OS, while these factors along with PB blasts ≥ 5% effectively risk stratified for LFS.