Spatial Transcriptomic Profiling from over 300 Leiomyosarcoma Samples.

11508 Background: Leiomyosarcoma (LMS) is a smooth muscle-derived tumor with significant heterogeneity and limited treatment options for recurrent/metastatic disease. A lack of targetable driver mutations and prognostic and predictive biomarkers have hampered the care of patients with LMS. There is a great need to better understand the biology of LMS and develop novel therapeutics. Advances in single cell RNA sequencing (scRNA-seq) have allowed for better understanding of intratumoral heterogeneity in diverse cancer subtypes. However, tissue dissociation during this process leads to loss of spatial context. Spatial gene expression analysis builds upon scRNA-seq and has the potential to yield information about tissue organization, cell-cell interactions, niches, and cell states. To date there have been limited application of spatial transcriptomics to sarcoma. Methods: We performed single nucleus multiome (snRNA-seq and snATAC-seq) on a cohort of 16 primary, untreated LMS samples including 12 soft tissue (STLMS) and 4 uterine (ULMS) tumors. We then designed a custom 480-gene panel using the differentially expressed genes from clusters identified in snRNA-seq data to be able to identify spatial relationships between these clusters and to assess these clusters on a larger scale. We utilized the 10x Genomics Xenium platform. This was applied to LMS tissue microarrays (TMAs) comprising a total of 326 tissue cores from 127 unique patients. Matched primary and metastatic samples from the same patient were available for 33 patients. Results: Analysis of scRNAseq data identified 2 distinct subtypes: a dedifferentiated subtype with mesenchymal features (MES) and a differentiated subtype with enrichment of smooth muscle cell markers (SMC). Integration of chromatin accessibility data from snATACseq showed enrichment of nuclear factor I (NFI) transcription factor (TF) motifs in the MES and AP-1 motifs in the SMC group. Whole genome sequencing did not reveal an obvious genomic etiology for these subtypes. Spatial transcriptomics was able to identify these 2 subtypes in a larger cohort of tumors. Consistent with snRNAseq data, we find that most tumors had almost exclusively either MES or SMC cells. We assessed spatial relationships between these subtypes and infiltrating immune cells. This revealed an enrichment in immunosuppressive macrophages and exhausted T cells in MES tumors compared to SMC tumors. Analysis of matched primary and metastatic tumors demonstrated that the subtype (MES or SMC) generally remains consistent between different sites of disease. Conclusions: We identify 2 novel LMS subtypes (MES and SMC) driven by distinct TFs. Spatial transcriptomic analysis confirmed the presence of these 2 subtypes in a larger cohort and demonstrated that MES tumors are associated with a more immunosuppressive tumor microenvironment.