ASXL1/TET2 genotype-based risk stratification outperforms ASXL1 mutational impact and is independent of mutant variant allele fractions in chronic myelomonocytic leukemia

Truncating ASXL1 mutations are a high-risk feature in chronic myelomonocytic leukemia (CMML)¹ and are associated with inferior overall survival (OS) and acute myeloid leukemia-free survival (LFS).² Conversely, we previously showed that loss-of-function/hypomorphic mutations in TET2 are associated with better outcomes, with the ASXL1 wild-type TET2 mutant (ASXL1^wt/TET2^mut) genotype conferring a survival advantage independent of treatment.³ However, contemporary prognostic scoring systems - including the Groupe Francophone des Myelodysplasies (GFM),⁴ Mayo Molecular (Mayo-Mol),⁵ and CMML-specific prognostic scoring system molecular (CPSS-Mol)⁶ models - do not consider mutational variant allele fractions (VAF) or TET2 mutational status. Here, we expand upon our prior work by assessing mutation VAF, reconsidering the use of binary mutation status, and integrating TET2 into the prognostic models.

After Institutional Review Board approval, we cataloged CMML patients seen at two centers, Mayo Clinic (N=466, 52%) and MD Anderson Cancer Center (N=422, 48%). Next-generation sequencing (NGS) was carried out as described at CMML diagnosis.^{3 ,7} Variants were annotated against international normal allele and pathologic mutation databases, and variants of uncertain significance (VUS) were excluded from analysis. As TET2 mutations occur in multiple clonal states,^3,8 we considered the mutation with highest VAF when assessing the impact on outcomes. Copy number alterations and loss of heterozygosity data were only available for a small number of patients, as reported elsewhere,⁹ and thus were not considered for this analysis. Statistical analyses considered the parameters at the time of presentation to the respective institution. Categorical variables were compared by Fisher exact or Pearson χ² tests and continuous variables by Mann-Whitney U test or two-way ANOVA with Tukey P value correction for pairwise comparisons. Univariate and multivariate analyses were performed using Cox proportional hazards regression models. Models were compared using concordance indices (C-statistic), where higher values indicate a better fit, and receiver operator curve (ROC) analyses.¹⁰ Survival was assessed via the Kaplan-Meier method. P values <0.05 were considered significant. Calculations were performed using BlueSky Statistics (v10.3.1) or MedCalc (v22.016). The median age of the cohort (N=888) was 71 years (range, 20-94), 33% were female, 46% had proliferative CMML (pCMML), and 19% had CMML-2 by current criteria^1,11 (Table 1). The most frequently mutated genes were ASXL1 (45%), TET2 (44%), SRSF2 (41%), NRAS (15%), and RUNX1 (15%). The median number of mutations in ASXL1 was 1 (range, 1-3) and in TET2 was 1 (range, 1-5); however, multiple ASXL1 mutations were rare (3%) in comparison to multiple TET2 mutations (47%; Figure 1A). Most patients had ≥1 mutation in an epigenetic regulator (79%) or spliceosome gene (57%). RAS pathway mutations were observed in 37%. Transformation to AML occurred in 168 patients (19%) and there were 586 deaths (66%). The median OS (mOS) and mLFS of the cohort were 31.8 and 28.4 months, respectively, with a median follow-up of 63.1 months. Risk stratification according to the GFM, Mayo-Mol, and CPSS-Mol models is shown in Online Supplementary Figure S1A-C.

In order to evaluate the impact of ASXL1 and TET2 mutations on OS and LFS, the cohort was divided into four genotype-based subgroups: ASXL1^wt/TET2^wt (N=244, 28%), ASXL1^mut/TET2^wt (N=254, 29%), ASXL1^wt/TET2^mut (N=241, 26%), and ASXL1^mut/TET2^mut (N=149, 17%) (Table 1). Patients with ASXL1 mutations were more likely to be male (P=0.0135), have a higher white blood cell (WBC) count (P=0.0129), and harbor mutations in transcriptional and RAS pathways (P<0.0001). Patients with TET2 mutations were more likely to have a higher hemoglobin (P<0.0001) and a normal karyotype (P=0.0005). As previously documented,^3,12,13 those with isolated TET2 mutations had the longest mOS of 58 months whereas those with isolated ASXL1 mutations had the shortest mOS of 21 months (Figure 1B). Patients with the ASXL1^wt/TET2^wt and ASXL1^mut/TET2^mut genotypes fared similarly with mOS of 30 and 27 months, respectively (Online Supplementary Figure S1F). The same pattern was observed for LFS (Figure 1C). We hypothesized that the ASXL1 or TET2 mutation VAF would be more predictive of outcomes than a binary metric. The respective median VAF were 37% and 45% (Figure 1A). When treated as a continuous variable, there was no correlation between VAF and OS or LFS by either Pearson linear or Cox regression (P>0.39 for all correlations in both models; Figure 1D). Similarly, amongst patients with multiple ASXL1 or TET2 mutations, there was no association between the number of mutations and OS or LFS (P≥0.06). There was also no survival difference between those with 1 versus ≥2 mutations in either gene (P>0.05 for each). Although prior studies have inconsistently shown associations between the number of TET2 mutations and survival,^3,8 these results support the practice of considering ASXL1 and TET2 mutation status as binary metrics in prognostic models.

Unlike in the overall cohort, the ASXL1/TET2 genotypes did not accurately stratify patients with pCMML, CMML-2, or those considered high-risk by the prognostic models (Online Supplementary Figure S1J, K). In contrast, patients considered intermediate- and low-risk by the Mayo-Mol and CPSS-Mol models were further stratified by the ASXL1/TET2 genotypes. Therefore, we sought to improve the existing molecular models by incorporating TET2 mutation status as a favorable prognosticator. Given that TET2 mutations balanced detrimental ASXL1 mutations in the Kaplan-Meier analyses, TET2 mutation status was given equal weight as ASXL1 in the GFM (-2 points), Mayo-Mol (-1.5 points), and the genetic risk scoring of the CPSS-Mol models (-1 point) (Online Supplementary Table S1A). Sex-specific hemoglobin thresholds were used as a surrogate for transfusion dependency in the CPSS-Mol model.^6,14,15

Table 1.Characteristics of the four ASXL1/TET2 genotypes within the CMML cohort.

After adjusting the risk category cutoffs to accommodate TET2 scoring (Online Supplementary Table S1A), the number of patients downstaged was 122 (18.4%), 215 (25.3%), and 97 (14.6%) in the Mayo-Mol, CPSS-Mol, and GFM models, respectively (Figure 2). Although 2% of patients in the Mayo-Mol and 6% in the CPSS-Mol were upstaged, no patients with TET2 mutations were upstaged. With the addition of TET2 status, the intermediate-1 and intermediate-2 risk groups were not significantly different in the Mayo-Mol model (P=0.49), whereas the low and intermediate-1 risk groups were not significantly different in the CPSS-Mol model (P=0.084); thus, these were each combined into a single group, yielding a three-tiered stratification in both models. In the GFM with TET2 mutational status, this resulted in a mOS of 42, 21, and 14 months for the low-, intermediate-, and high-risk groups, respectively. In the Mayo-Mol with TET2, the mOS was 58, 31, and 15 months, respectively. In the CPSS-Mol with TET2, the mOS was 63, 30, and 16 months, respectively (Figure 1E-G). Similar results were obtained when patients in the Mayo Clinic subgroup (where hematopoietic cell transplantation data were available, N=18, 4%) were censored at the time of transplant. In all three models, the addition of TET2 mutation status improved prognostication compared to the parental model, as indicated by higher concordance indices for each model (Online Supplementary Table S1B). Likewise, the models with TET2 status performed similar to or better than the parental models in ROC analyses.

These findings were then validated in an external database from Moffitt Cancer Center (N=265, 31% female) with median age 71 years (range, 17-88 years) and 55% pCMML and 15% CMML-2 cases (Online Supplementary Table S2). The mOS and mLFS of the external cohort were 41 (95% confidence interval [CI]: 33-51) and 37 (95% CI: 28-46) months, respectively, with 55 (21%) blast transformation events and 136 (51%) deaths. The external cohort was grouped by ASXL1/TET2 genotype, providing: ASXL1^wt/TET2^wt (N=50, 19%), ASXL1^mut/TET2^wt (N=44, 17%), ASXL1^wt/TET2^mut (N=105, 40%), and ASXL1^mut/TET2^mut (N=66, 25%). As in the primary cohort, the ASXL1^wt/TET2^mut genotype conferred the longest mOS (61 months) and the ASXL1^mut/TET2^wt genotype the shortest mOS (22 months; Online Supplementary Figure S1L). The same trend was observed for LFS. Again, the pCMML (P=0.056) and CMML-2 (P=0.12) subgroups were not stratified by the genotypes, and there was no correlation between ASXL1 or TET2 VAF with either OS or LFS (P>0.25 for all comparisons). While patients were stratified by existing molecular models (as expected), the addition of TET2 mutation status to the Mayo-Mol and CPSS-Mol models again defined three risk groups (low, intermediate, and high) with respective mOS values of 77, 39, and 20 months for the Mayo-Mol (P<0.0001) and 77, 39, and 22 months for the CPSS-Mol model (P<0.0001; Figure 1I-J). The mOS with the GFM model incorporating TET2 status was 61, 31, and 15 months, respectively (P<0.0001; Figure 1H). Again, models incorporating TET2 mutation provided higher concordance indices and similar AUC values compared to parental models (Online Supplementary Table S1B).

Figure 1.Prognostic implications of ASXL1/TET2 co-mutation status in chronic myelomonocytic leukemia. (A) The upper panel (histogram) shows the distribution of cases with the indicated number of mutations in ASXL1 or TET2. The lower panel (violin plot) shows the distribution of mutational variant allele fractions (VAF) for ASXL1 in blue (median 37%, quartiles 28-45%) and TET2 in red (median 45%, quartiles 40-49%). (B) Median overall survival (mOS) and (C) acute myeloid leukemia-free survival (LFS) of the cohort stratified by ASXL1/TET2 genotype. (D) No correlation was observed between mutation VAF and either OS or LFS. P>0.39 for all correlations by both Pearson linear regression and Cox proportional hazard modeling. The mOS of the primary cohort stratified into low-, intermediate- (inter), and high-risk subgroups by the (E) Groupe Francophone des Myelodysplasies (GFM), (F) Mayo-Molecular (Mayo-Mol), and (G) CPSS-Molecular (CPSS-Mol) models each with the addition of TET2 mutation status The mOS of the external cohort stratified by the (H) GFM, (I) Mayo-Mol, and (J) CPSS-Mol models each with the addition of TET2 mutation status. In all Kaplan-Meier analyses, data are presented as median survival (95% confidence interval) in months with log-rank P values.

Figure 2.Risk restratification from existing chronic myelomonocytic leukemia prognostic models to models incorporating TET2 mutation status. Stacked colored bar plots show the restratification of existing prognostic models to the updated models incorporating TET2 mutation status for patients whom both scores could be calculated. Each bar corresponds to one existing risk category while colors represent the new risk categories with TET2 status. The gray bar plots represent the percentage of restratified patients within each contemporary model’s stratum. The pie charts depict the ASXL1/TET2 genotypes of the patients who were upstaged in each model. The addition of TET2 status to the Groupe Francophone des Myelodysplasies (GFM) model did not upstage any patients, and therefore no pie chart is depicted for this model. Inter: intermediate.

In summary, our data validates the positive prognostic impact of TET2 mutations in CMML, highlighting the importance of considering the ASXL1/TET2 co-mutational status for prognostication.^3,12,13 Expanding upon prior work, we further show that ASXL1 and TET2 mutational VAF does not impact prognostic outcomes, supporting the ongoing practice of binary assessments for molecularly-based CMML prognostication. Furthermore, in a large database and an external validation cohort, the addition of binary TET2 mutation status to existing molecularly-integrated CMML prognostic models simplified and refined risk stratification. Regardless of whether they are statistically superior, by downstaging some patients and harmonizing the models into three-tiered systems, these refined models may simplify risk stratification and clinical decision making. In this regard, the low-risk tiers of these models represent the lowest-risk patients whereas the intermediate- and high-risk tiers identify “higher-risk” patients. Finally, the favorable impact of TET2 mutations in hematological neoplasms is largely associated with CMML3 and biological studies understanding the underlying mechanism are needed.

Footnotes

• Received March 4, 2024
• Accepted June 13, 2024

Correspondence

^#EP, GM-B, and MMP contributed equally as senior authors

M.M. PATANIK - patnaik.mrinal@mayo.edu

Disclosures

MMP has received research funding from Kura Oncology, Stem Line, Epigenetix, Polaris and has served on the advisory board for CTI pharmaceuticals. All other authors have no conflicts of interest to disclose.