(P 398) INTEGRATED MULTI-OMICS APPROACH REVEALED TUMOR-ASSOCIATED TERTIARY LYMPHOID STRUCTURE LANDSCAPE IN RETROPERITONEAL LIPOSARCOMA

Objective: Retroperitoneal liposarcoma (RPLS) is a rare type of mesenchymal tumor, which is characterized by challenging surgical treatment, immune-desert, dismal immunotherapy response and high local recurrence rate. The presence of tertiary lymphoid structure (TLS), like ectopic lymphoid aggregates discovered in tumor immune microenvironment (TIME), was found associated with favorable patient survival and response to immunotherapy. However, the mechanisms underpinning its clinical activity are unclear. Thus, we aimed to investigate the prognostic value and spatial distribution and functional involvement of TLS in RPLS.

Methods: 330 patients were retrospectively enrolled into five independent cohorts (Cohort-Primary (n=110), Cohort-Relapse (n=76), Cohort-Evolution (n=10, Same patient with well-differentiated liposarcoma (WDLPS) in primary tumor and dedifferentiated liposarcoma (DDLPS) in relapse tumor) Cohort-Validation (n=76) and Cohort-TCGA (n=58)) from the largest two retroperitoneal tumor research center of China and TCGA database. Sc-RNA seq (n=4) and spatial transcriptome seq (n=2) were performed for the estimation of TIME based on treatment-naive samples. Transcriptomic profiles and corresponding clinical information of 309 cases were enrolled from two public databases, including Cohort-TCGA (n=58), Cohort-GSE30929 (n=92), Cohort-GSE21050 (n=59), Cohort-GSE21122 (n=50) and Cohort-FD (n=50). TLS was quantified at three distinct anatomic subregions (intra-tumor, invasion margin and peri-tumor) and correlated with overall survival (OS) and disease free survival (DFS) by Cox regression and Kaplan-Meier analyses. Multiplex immunohistochemistry (mIHC) was performed to characterize and validate the spatial composition of TLS in another treatment-naive RPLS cohort (n=16) and neoadjuvant chemotherapy RPLS cohort (n=12).

Results: Quantitative and qualitative scoring system of TLS was established for the intra-tumor region (T score), invasion margin region (IM score) and peri-tumor region (P score), respectively. In Cohort-Primary, high T score was positively associated with favorable OS (p=0.003) and DFS (p < 0.001), whereas high P score predicted dismal OS (p=0.008) and DFS (p=0.001) (Figure 1). Additionally, the joint scoring system of T and P scores stratified RPLS into four immune classes with distinct TLS distribution patterns and prognoses (p < 0.001) (Figure 1). The C-index of the immune class was significantly higher than the TNM staging system (0.798 vs.0.62, p=0.005). These results were also validated in Cohort-Relapse and in concordance with the image-omic explanations in Cohort-TCGA. Intriguingly, the spatial distribution and abundance of TLS indicated a high degree of consistency between H&E evalution and transcriptomic level (Figure 2). Importantly, mIHC discovered that regulatory T cells (Tregs) were significantly increased in intra-tumoral TLS in DDLPS compared with those in WDLPS, displaying an immunosuppressive pattern (Figure 3). Interestingly, neoadjuvant chemotherapy could block this status of immunosuppressive, induced TLS formation and restored the antitumor immune balance in responsive patients, whereas the nonresponsive patients deteriorated into a more suppressive one (Figure 4). Strikingly, significant more CD38+ plasma cell could be found in responsive patients, compared with nonresponsive patients (Figure 4). sc-RNA seq and ST analysis further revealed significant intra-tumoral and inter-tumoral TIME heterogeneity and identified the underlying transcriptomic programs driving each phenotype.

Conclusion: Collectively, our study provides a high-resolution map of the TIME with treatment-naive and neoadjuvant chemotherapy RPLS, including the spatial distribution, composition and immunophenotype. We further uncovers the critical role for TLS-centered antitumor immunity. The results suggested that regulatory T cells might play an essential role in determining functional heterogeneity of TLS. Remarkably, the spatial distribution and abundance of TLS gradually decreased in intra-tumor area, whereas increased in peri-tumor area within the malignant progression from WDLPS to DDLPS. TLS located within distinct regions of TME presented opposite prognostic influences, and the immune classification was able to fuse these heterogeneous modalities to predict varying prognoses, even compared with TNM staging system. Efficient neoadjuvant chemotherapy can induce TLS formation and restore the antitumor immune balance in RPLS.