Somatic Gene Mutations Serve As Molecular Biomarkers Predictive for Response to Immunosuppressive Therapy (IST) in Myelodysplastic Syndromes (MDS)

Introduction

Immunosuppressive therapy (IST) yields durable hematologic improvement (HI) in a subset of patients (pts) with lower risk MDS. Age, HLA-DR15⁺, and duration of transfusion dependence are the strongest independent clinical variables predictive for response. We investigated the impact of somatic gene mutations on response to IST in lower risk MDS pts.

Methods

MDS pts who received ATG +/- CSA were identified at the Moffitt Cancer Center. The National Institutes of Health (NIH) IST response model was calculated for each pt. Next Generation sequencing (NGS) for somatic gene mutations was conducted using DNA extracted from archived BM prior to therapy. All pts underwent mutation analysis by a 49 myeloid gene panel. The library was generated with the ThunderBolt (RainDance Technologies, Billerica, MA) and sequenced on a MiSeq instrument (Illumina, San Diego, CA). Alignment and variant calling was performed with NextGene (Soft Genetics, State College, PA). 

Results

Sixty-six pts treated with ATG +/- CSA were identified. Median age was 61 and the majority of pts had IPSS lower risk disease with favorable risk karyotype. Median time to initiation of IST was 1 year. All pts received ATG (60% rabbit (r-ATG); 32% equine (e-ATG)), and CSA was used in 60% of pts. Overall frequency of HI was 42% with a trend favoring e-ATG vs. r-ATG (52% vs. 39%, p=0.09). Erythroid HI was evaluable in 30 pts with 60% responding, neutrophil improvement was evaluable in 15 pts and 39% responded, while platelet improvement was evaluable in 18 pts with 57% responding. Six of 18 (33%) pts with pancytopenia experienced trilineage response. Mean time from ATG to next therapy was 12 mo (median of 7.7 mo). Neither presence of an LGL clone, hypocellular BM or fibrosis, HLA DR15, trisomy 8, nor age influenced response to IST. Pts classified as IPSS-R Very high or high risk were unresponsive (n= 5), whereas 10 of 19 pts (53%) with intermediate risk responded. Poor risk IPSS karyotype was associated with a trend for lower response rate when compared to intermediate and good risk (25% vs. 41% vs. 44%; p=0.6). The response rate based on the NIH IST model was 38% for low response probability category pts and 45% for high probability category (p=0.5). Response rate to IST was higher if administered within 2 years from diagnosis, with an HI rate of 48% vs. 33% when treated after 2 years (p=0.04). Pts who received ATG as first line treatment or after lenalidomide had a trend for higher response rates than those treated after azacitidine (46%,75%, and 25% respectively). Addition of CSA significantly improved HI rate (51% vs. 27% for ATG alone, p=0.02). Transformation to AML occurred in 10 pts, 7% of responders and 24% of non-responders (p=0.08). Median OS was 67.2 mo with no significant difference based on IST response.

Among 40 pts evaluated by NGS, 20 (50%) had at least one demonstrable somatic mutation (SM) and 9 pts (22.5%) had two or more SM. SF3B1 was the most common SM detected (n=9, 22.5%), followed by ASXL-1 (n=7, 17.5%), TET-2 (n= 5, 12.5%), and STAG2, EZH-2 and ZRSR2 (2 pts each, 5%), and 1 pt each with IDH-1, KDM6A, SETBP1, RAD2, GNAS or GATA-2 . Absence of a SM was associated with a higher response to IST (70% vs. 40%, p=0.16), whereas number of SM (1 vs. 2+) did not influence response. The presence of an SF3B1 mutation was a significantly associated with IST nonresponse (1/9 SF3B1 SM, 11% vs. 21/31 WT, 68%; p=0.01). All pts with SF3B1 SM had ring sideroblasts >15% (RS) by morphology and the corresponding HI rate was 20% among pts with RS vs 50% for those without RS, p=0.09.  Median OS in pts with an SF3B1 SM was 111 mo vs. 54 mo in SF3B1 WT (p=0.016). The two pts with EZH-2 and the single pt with WT-1 S  achieved HI. Mean duration of response was 12 mo among pts with no SM vs. 9 mo in those harboring a SM (p=0.09). Rate of AML transformation among pts with a SM other than SF3B1was higher in pts without SM (4/11 pts, 36%vs. 1/20, 5%; p =0.023, with a corresponding reduced median OS (52 mo vs. 96 , p=0.24).

Conclusions

These findings support an improved response rate to ATG when administered in combination with CSA, and early in the disease course. The presence of an SF3B1 mutation adversely influences response to IST, suggesting a non-immunologic pathogenesis in this molecularly defined subset. The presence of non-SF3B1 somatic mutations adversely affects response duration and probability of AML transformation. SM should be considered in selection of IST in lower risk MDS patients.