11067 Background: GEIS 30 was a phase II study showing moderate activity of pazopanib in well-differentiated/dedifferentiated LPS (cohort A:37 pts, Progression free Survival (PFS) at 12 weeks (w) 43.2%) and no activity in Myxoid/round cell LPS (cohort B: 15 pts, PFS at 12w 13.3%). The present study aims to identify tumor and plasma biomarkers that are differently expressed in long responders (LRs) PFS > 24 w. Methods: Serum samples and paraffin-blocks at diagnosis were collected for 28 pts in cohort A and 11 in B. A total of 13 pts were LRs. Serum samples were obtained at baseline, after 3 w of treatment and at progression. A panel of 15 cytokines and growth factors were evaluated using Luminex technology (Millipore). Paraffin-blocks were evaluated for microvessel density (MVD) by CD31 immunostaining and Image Pro-Plus 7.0 Image Analysis System (Media-Cybernetics). For RNA expression analysis, the Oncology Biomarker Panel with probes for 2559 transcripts was used (HTG Molecular Diagnostics). Results: Serum cytokines showed that expression of angiopoietin (AGO) and BMP9 decreases during treatment, whereas GCSF increases. In addition, AGO and BMP overexpression at baseline was associated with worse prognosis in arm A (p = 0.09) and B (p = 0.01) respectively. No differences between study arms and outcome were appreciated by MVD. Gene expression revealed differences in some immunomodulators: PDL-1 was correlated with serum cytokines VEGFD (p = 0.04) and LEPTIN (p = 0.02); and PD-1 with VEGFD (p = 0.04), LEPTIN (p = 0.02), VEGFA (p = 0.02) and ANGIO (p = 0.03). PD1, in addition,was overexpressed in LRs (p = 0.05). Interestingly, three groups were identified in cohort A according to gene expression: Cluster 1, characterized by short-responder patients with the shortest overall survival, whereas cluster 3 (40% LRs) and 2 (63.6% LRs) showed longer survival rates (14.3%, 45.5% and 30% respectively) (p = 0.001). Conclusions: Gene expression profiling unveils three WD/DD-LPS biotypes according to their response to pazopanib, which could be potentially used to select pts for treatment