CSF3R Mutations Represent a Novel Therapeutic Target in Pediatric AML with a High Degree of Overlap with CEBPA Mutations: a Report from COG AAML0531 and COG/NCI Target AML Initiative

Activating mutations in Colony Stimulating Factor 3 Receptor (CSF3R, aka GCSFR) are present in ~80% of patients with Chronic Neutrophilic Leukemia (CNL) Despite the high frequency of these mutations in CNL, they are quite rare in adult acute myeloid leukemia (AML), in which only a single CSF3R mutated case was found in the TCGA AML analysis (0.5%).  We have previously demonstrated significant variation in genomic variants between pediatric and adult malignancies, thus prevalence of genomic variants identified in adults need be fully characterized in children. As part of COG/NCI TARGET AML initiative we interrogated the genomic makeup of 200 cases of childhood AML (discovery cohort) using whole genome sequencing and performed subsequent frequency determination of the variants in 787 unselected cases from COG AAML0531 (validation cohort).  Somatic variants in CSF3R were initially found to be recurrent in the discovery cohort and underwent frequency determination in the validation cohort to establish their prevalence and correlations with clinical characteristics and outcome.

Frequency determination of CSF3R mutation in 787 pediatric patients with available CSF3R data from AAML0531 identified 16 distinct CSF3R mutations in 28 patients (3.6%).  Somatic mutations in CSF3R identified in pediatric AML included known oncogenic variants mutations such as T618I and T615A, previously identified in adult CNL studies as well as novel truncations of the CSF3R cytoplasmic domain (Q749X, Y767fs, Y787X and P819/820fs), and missense mutations (E149D, A208V, R223Q, E405K, A431V, and Q516K).  Interestingly, although CSF3R truncations usually occur along with a T618I or T615A mutation in CNL/aCML, these two mutation categories were mutually exclusive in pediatric AML.

Initial correlation of all CSF3R variants with demographic and clinical/laboratory parameters determined that CSF3R variants were less prevalent in younger patients (age 0-2, p=0.039), with significantly higher association with t(8;21) (32% vs. 14%, p=0.012) and CEBPA mutations (35% vs. 5%, p<0.001).  Cumulatively, 18/28 patients with CSF3R mutations (64%) had either CBF translocations or CEBPA mutations and as a result, CSF3R mutation was highly associated with favorable risk (p=0.02) and inversely associated with standard risk disease (p=0.007).  Actuarial overall survival at 5 years for patients with and without CSF3R mutations was 91% vs. 64%, respectively (p<0.001).  In order to determine the oncogenic potential of the newly discovered variants, all untested variants (N=12) were cloned and expressed in Ba/F3 cells in order to determine whether these variants can bestow cytokine independence to these cells.  Of the 16 total CSF3R variants identified, 8 variants present in 18 patients exhibit oncogenic capacity (T615A, T618I, T640N, Q749X, Y767fs, S783fs, Y787X and F819fs). All of the novel variants that exhibited oncogenic potential were truncating mutations.

Compared to non-mutated cases, transforming CSF3R variants had a significant association with CEBPA mutations (44% vs. 5%, p<0.001), and led to significant association of CSF3R with favorable risk disease (67% vs. 39%, p=0.019). Actuarial overall survival at 5 years for those with and without transforming CSF3R mutations was 87% vs. 64%, (p=0.047). 

CSF3R mutations define a distinct molecular subset of pediatric AML, which could be therapeutically targeted in the future using kinase inhibitors such as ruxolitinib. The oncogenic CSF3R mutations found in pediatric AML are either the same point mutations or similar truncation mutations as seen in CNL, suggesting that other cooperating genomic alterations may be important in directing these distinct diseases. Interestingly, we found that the majority of pediatric AML patients with CSF3R mutation have either a core binding factor alteration (such as t(8;21)) or a mutation in CEBPA. The enrichment of CEBPA mutations with CSF3R mutations is particularly striking, as CEBPA mutations are ~9 fold more frequent in patients with transforming CSF3R mutations than those without. Understanding the role of cooperating genomic alteration in CSF3R-driven myeloid malignancies will be the subject of future work.

The authors would like to gratefully acknowledge the important contributions of the late Dr. Robert Arceci to the AML TARGET initiative.