CSF3R mutations and variants in myeloid neoplasms: associated phenotypes, co-mutations, and survival trends

Granulocyte colony-stimulating factor (GCSF) regulates cell proliferation, differentiation and survival through binding to the extracellular domain of the colony-stimulating factor 3 receptor (CSF3R).^1-3 Ligand binding induces formation of CSF3R homodimers, bringing the cytoplasmic domains into proximity.¹ This leads to the activation of signal transduction pathways such as JAK/STAT, MAPK/ERK and PI3K/AKT.^2,4

Studies have associated CSF3R mutations/variants with myeloid neoplasms.⁴ The CSF3R-T618I mutation has been widely reported in chronic neutrophilic leukemia (CNL).^5,6 This, along with other CSF3R mutations/variants, has also been sparsely reported in non-CNL myeloid neoplasms such as acute myeloid leukemia (AML), atypical chronic myeloid leukemia and chronic myelomonocytic leukemia (CMML).^7-9 The current study aims to elucidate the spectrum of CSF3R mutations/variants and associated disease phenotypes and genotypes in patients with myeloid neoplasms, and their impact on survival. The study was approved by the Mayo Clinic Institutional Review Board (IRB 12-003574).

We reviewed the Mayo Clinic database to extract records of patients with myeloid neoplasms who had CSF3R mutations/variants on next-generation sequencing performed between 2016 to 2024. Mutations/variants were classified as either pathogenic (CSF3R^{PAT H}) or variants of unknown significance (CSF3R^VUS) based on their characterization in the literature at the time of testing. Descriptive and analytical statistical methods were employed, including χ² tests, Kaplan-Meier estimates for overall survival (OS) and Cox proportional hazards regression.

Figure 1.Distribution of myeloid neoplasms and associated mutations/variants. Distribution of myeloid neoplasms in the subcohorts of patients with CSF3R pathogenic mutations (CSF3R^PATH) (top) and variants of unknown significance (CSF3R^VUS) (bottom). The pie charts (left) show the distribution of the myeloid neoplasms, while the bar charts (right) show the different mutations/ variants. AML: acute myeloid leukemia; CMML: chronic myelomonocytic leukemia; CNL: chronic neutrophilic leukemia; MDS: myelodysplastic syndrome; MDS-MPN: myelodysplastic/myeloproliferative neoplasm; MPN: myeloproliferative neoplasms (myelofibrosis, polycythemia vera, chronic myeloid leukemia, and essential thrombocythemia); PM: premalignant conditions; SM: systemic mastocytosis.

We identified 182 patients, including those with myeloid neoplasms (N=167, 91.8%) and premalignant conditions (N=15, 8.2%). The two premalignant conditions were clonal hematopoiesis of indeterminate potential (N=6, 3.3%) and clonal cytopenia of undetermined significance (N=9, 4.9%). The median age at CSF3R mutation detection for the entire cohort was 70 years (range, 17-93). Patients with premalignant conditions were younger, with a median age of 66 years (range, 30-92) compared to 71 years (range, 17-93) for patients with myeloid neoplasms.

A total of 52 distinct CSF3R mutations/variants were identified in the 182 patients (91 in each CSF3R^{PAT H} and CSF3R^VUS subcohort) (Online Supplementary Figure S1). CSF3R^PATH (N=21, 40.4%) included missense (N=5, 9.6%), nonsense (N=12, 23%) and frameshift mutations (N=4, 7.7%). CSF3R^VUS (N=31, 59.6%) included missense (N=27, 51.9%), nonsense (N=1, 1.9%), frameshift (N=1, 1.9%), synonymous (N=1, 1.9%) and splice site (N=1, 1.9%) variants. There were 12 patients with two CSF3R mutations/variants: seven had two CSF3R^PATH, three had one CSF3R^PATH and one CSF3R^VUS, and two had two CSF3R^VUS.

CSF3R^PATH was more prevalent in the cytoplasmic domain (N=17, 81%) than CSF3R^VUS (N=15, 48%; P=0.02). The overall median variant allele frequency (VAF) was 47% (interquartile range [IQR], 25-49). CSF3R^PATH had significantly lower VAF, with a median of 28% (IQR, 10-41) compared to CSF3R^VUS, with a median of 49% (IQR, 48-50; P<0.001). The latter observation suggested the possible inclusion of germline variants among patients with CSF3R^VUS. The most frequently identified CSF3R^PATH was T618I (N=53, 32%), observed in CNL (N=18, 34%), AML (N=15, 28%), and myelodysplastic/myeloproliferative neoplasm (N=8, 15%). CSF3R-E808K was the most frequent CSF3R^VUS (N=50, 30%) and was observed more commonly in AML (N=14, 8%), myelodysplastic syndrome (MDS) (N=17, 10%), and CMML (N=7, 4%). Figure 1 depicts details of CSF3R mutations/ variants and associated myeloid neoplasms.

Table 1.Distribution of co-mutations across different disease groups.^*

The median number of co-mutations was higher in patients with CSF3R^{PAT H}, at three (range, 0-6), with 50 patients (58%) having more than three co-mutations, than in patients with CSF3R^VUS, who had a median of one (range, 0-5) co-mutation, with 25 patients (28%) having more than three co-mutations (P<0.001). Patients with AML and CMML displayed higher numbers of co-mutations. In patients with CSF3R^{PAT H}, the most frequently observed co-mutations included ASXL1 (N=46, 53%), SETBP1 (N=19, 22%), and SRSF2 (N=18, 20%). In patients with CSF3R^VUS, the most frequently observed co-mutations involved ASXL1 (N=16, 18%), TET2 (N=15, 17%), DNMT3A (N=10, 11%), and TP53 (N=10, 11%). Table 1 and Online Supplementary Table S1 show the distribution of co-mutations in different diseases.

Figure 2.Overall survival in patients with myeloid neoplasms with CSF3R mutations/variations. Overall survival of patients with myeloid neoplasms with CSF3R pathogenic mutations (CSF3R^{PAT H} ) (top) and variants of unknown significance (CSF3R^VUS) (bottom). AML: acute myeloid leukemia; CNL: chronic neutrophilic leukemia; MDS-MPN: myelodysplastic/myeloproliferative neoplasm; PM: premalignant conditions; CMML: chronic myelomonocytic leukemia; MDS: myelodysplastic syndrome; MPN: myeloproliferative neoplasms (myelofibrosis, polycythemia vera, chronic myeloid leukemia, and essential thrombocythemia); SM: systemic mastocytosis; NR: not reached.

In AML, the median age at diagnosis was 69 years (range, 17-90). CSF3R-T618I (N=15, 28%) was the most prevalent CSF3R^{PAT H}, while CSF3R-E808K (N=14, 26%) was the most prevalent CSF3R^VUS. When compared to other myeloid neoplasms with CSF3R^{PAT H}, AML patients with CSF3R^{PAT H} were more likely to harbor co-mutations in CEBPA, DNMT3A, IDH1, IDH2, NRAS, RUNX1, and WT1. Among AML patients with CSF3R^VUS, FLT3 and KDM6A were more common. In MDS patients, the median age at CSF3R mutation detection was 74 years (range, 37-93). The most common CSF3R^PATH were T618I (N=3, 19%) and Q749* (N=4, 25%). Other frequent mutation/variants included truncating CSF3R mutations/variants (Y752*, W791*, Q776*, Q770*, Q743*, and L751P-fs). The most common CSF3R^VUS in MDS was E808K (N=18, 42%), and it was more frequent in MDS than in other myeloid neoplasms (N=32, 26%; P=0.047). Among patients with CSF3R^{PAT H}, co-mutations in U2AF1 were more frequent in those with MDS (N=5, 31%) than in those with other myeloid neoplasms (N=6, 9%; P=0.01). DDX41 mutations were more common in the CSF3R^VUS subcohort. Among the 20 patients with CNL, 18 (90%) harbored CSF3R-T618I and the median age at diagnosis was 70 years (range, 33-93). Compared to patients with other myeloid neoplasms with CSF3R mutations/variants, co-mutations were more frequent in CNL patients and involved ASXL1 (71% vs. 34%; P<0.01), SRSF2 (47% vs. 12%; P<0.01), and SETBP1 (41% vs. 11%; P<0.01).

The median follow-up time for the entire cohort was 43 months (95% confidence interval [95% CI]: 34-49 months). Overall survival data across different myeloid neoplasms with CSF3R mutations/variants are shown in Figure 2. Patients with MPN (including myelofibrosis, polycythemia vera, chronic myeloid leukemia, and essential thrombocythemia) had a significantly longer median OS of 78 months (95% CI: 59 months-not reached) in patients with CSF3R^{PAT H} and median OS not reached in those with CSF3R^VUS. In general, patients with CSF3R^PATH fared worse than those with CSF3R^VUS, in every disease group, with the distinct exception of AML, in which the opposite was true (Figure 2).

In AML patients, the median OS was 24 months (95% CI: 14-56 months) for those with CSF3R^{PAT H} compared to 9 months (95% CI: 5 months-not reached) for those with CSF3R^VUS (P=0.1). Comparing the most common CSF3R mutations/variants in both cohorts, AML patients with the E808K variant had a worse OS of 5 months (95% CI: 4 months-not reached) compared to those with the T618I mutation who had a median OS of 30 months (95% CI: 12 months-not reached; hazard ratio [HR]=4, 95% CI: 1.7-10; P<0.01); significance was sustained after adjustment for European LeukemiaNet 2022 risk stratification, age, and transplant therapy (HR=4, 95% CI: 1.5-10.7; P<0.01). The median OS for MDS patients with CSF3R^PATH was 21 months (95% CI: 12 months-not reached) compared to 29 months (95% CI: 10 months-not reached) for MDS patients with CSF3R^VUS (P=0.6). Unlike the case with AML, CSF3R-E808K in MDS did not confer an inferior prognosis (OS 27 months vs. 21 months in patients with and without the specific variant, respectively; P=0.4). The median OS for patients with CNL was 25 months (95% CI: 18 months-not reached). The median time to blast transformation in 15 patients who transformed to AML from other myeloid neoplasms was 14 months, with a 5-year blast transformation-free survival of 52% (95% CI: 30-90%) for patients with CSF3R^PATH compared to 89% (95% CI: 78-100%) for those with CSF3R^VUS (P<0.01).

The association between CSF3R mutations/variants and CNL has been well reported in the literature. The current study adds to this information in the context of a broader spectrum of CSF3R mutations/variants in myeloid neoplasms.^7-10 Our study is subject to limitations inherent to retrospective reviews, including the lack of functional studies, especially with regard to the distinction between germline variants and mutations. We were particularly intrigued by the frequent occurrence of the CSF3R-E808K variant and its prognostic relevance in AML.^11,12 Additional studies are needed to clarify the possible oncogenic/prognostic impact of CSF3R^VUS and recognize the fact that the characterizations regarding pathogenic versus VUS remain preliminary and are subject to change with future research.

Footnotes

• Received April 10, 2025
• Accepted June 23, 2025

Correspondence

A. TEFFERI - tefferi.ayalew@mayo.edu

Disclosures

NG has served on advisory boards for DISC Medicine and Agios.

Contributions