From chaos to clarity? decoding the inflammatory profiles in lower-risk MDS

Abstract Introduction Immune dysregulation is increasingly recognized as both a pathogenic driver and therapeutic target in Myelodysplastic Syndromes/Neoplasms (MDS). Despite improvements in prognostic stratification by the Molecular International Prognostic Scoring System (IPSS-M), MDS patients remain heterogeneous in their clinical behaviour, treatment responses and disease evolution. Inflammation impact seems to be preponderant in pathophysiology of lower risk (LR)-MDS. By assessing 36 analytes in LR-MDS serum at diagnosis with a screening panel, we stratified and characterized LR-MDS cases into biologically distinct subgroups with different clinical characteristics and treatment response to erythropoietin stimulating agents (ESAs). Methods We evaluated 95 MDS cases selected per lower risk IPSS-R (score < 3.5), and symptomatic anemia, analyzing 36 serum proteins (cytokines, alarmins, soluble receptors and growth factors). Complete clinical annotations (including IPSS-M scores) were collected. Serum protein levels were assessed using custom Luminex Screening Assay panels. For bioinformatic analysis, proteins were regrouped into simplified biologically relevant groups: pro-inflammatory (S100A8/A9, TNF-α, IL-1β, IL-18, IL-17E, IL-8, IFN-α/γ, CCL2/3/4, RANTES, CCL11, CXCL9/10/7/4; n=18), anti-inflammatory (IL-1RA, IL-10; n=2), bimodal (pro- and anti-inflammatory potential; IL-6, IL-4, IL-13; n=3), growth factors (VEGF, PDGF-BB, HGF, FGF-BASIC, M-CSF; n=5), receptors (TNFR1, CD25, GP130; n=3), apoptosis-related (TRAIL; n=1), and immune expansion (IL-2, IL-5, IL-7, IL-15; n=4). Following log-transformation and z-score normalization, we performed principal component analysis (PCA) and k-means clustering to identify distinct LR-MDS subgroups. Feature selection by random forest models identified the most discriminatory cytokines in LR-MDS. Subgroups were compared in terms of inflammatory state, clinical variables, and clinical endpoints including overall survival (OS) and ESA response duration. Results Median age at diagnosis was 74 years (range 47-91) (M:F ratio 1.6:1). Per WHO 2022 classification (n=69): MDS-LB = 44, MDS-SF3B1 = 21, and MDS-del(5q) = 4. IPSS-R scores (n=95) were: very low = 31, low = 55, intermediate = 9. IPSS-M scores (n=62): very low = 16, low = 32, moderate-low = 9, moderate-high = 4, and very high = 1. Through PCA we identified five distinct clusters based on serum protein levels. Cluster sizes were: C1 (n=18), C2 (n=17), C3 (n=12), C4 (n=22), and C5 (n=26). Inter-cluster comparisons revealed significant differences between all analytes annotated for biological function (adj. p < 0.001, except for receptors and apoptosis-related groups), especially for bimodal and pro-inflammatory proteins, and growth factors. Random forest analysis identified IL-4, IL-17E, IL-2, CCL3, IL-1β, IL-7, CCL4, IL-13, IFNγ and CXCL7 as most discriminatory serum proteins. Cluster-specific characteristics: C1 (72% low IPSS-R, median age 76 years) showed low pro-inflammatory levels (mean log10: 1.88); C2 had the highest MDS-del(5q) frequency (18%) and highest inflammatory burden (mean log10: 2.30, 3.2-fold higher than C3); C3 was the youngest (median age 58.6 vs 76.6 years overall, p=0.041), enriched for MDS with ring sideroblasts (MDS-RS, 33%), and displayed the lowest pro-inflammatory profile (mean log10: 1.80). There was a numerical difference in rate and duration of ESA response, with a trend to longer ESA response in C4 and C5 (log-rank p=0.089). Interestingly, clusters C1, C2 and C3 had higher rate of resistance or early relapse to ESA therapy (56%, 50%, and 67%, respectively) when compared to clusters C4 and C5 (18% and 23%, respectively). While overall survival did not significantly differ across clusters, cluster C2 showed higher rate of AML transformation (14% vs 6% non-C2). Conclusion By profiling serum analytes in 95 LR-MDS patients we identified five biologically distinct LR-MDS subgroups with differential inflammatory signatures correlating with ESA responses. Extreme pro-inflammatory profile (C2), and low inflammatory profiles (C1 and C3) identify LR-MDS cases refractory to ESAs. In C2 poor response may be correlated with a more severe disease course, with AML early progression; while C3 is enriched for young MDS cases with low-inflammation burden with MDS-RS. These results are relevant to support early therapeutic decision-making and prompt further studies in larger LR-MDS cohorts to dissect cytokine-mediated molecular pathways.