Comparison of genomic landscapes at single cell resolution of pulmonary and salivary adenoid cystic carcinomas.

Abstract

e18152 Background: Primary pulmonary adenoid cystic carcinomas (PACCs) are rare salivary-gland-like neoplasms associated with poor long-term survival outcomes. How these cancers compare to primary salivary gland adenoid cystic carcinomas (SACCs) remains poorly understood. Recent attempts to elucidate biomarkers and actionable mutations of PACCs through bulk RNA sequencing studies have been met with some success. However, no study to date has compared composition of PACCS and SACCs on single cell RNA level. Methods: We identified two publicly available scRNA sequencing datasets in GEO processed on Illumina NovaSeq 6000 platform. GSE217084 dataset obtained from paraffin embedded samples derived from a patient with primary SACC and adjacent normal head and neck tissue, as well as a patient with 3 lung SACC metastases and adjacent normal lung tissue. GSE245170 dataset was derived from paraffin embedded tissue from a patient with primary PACC, adjacent normal lung tissue, and peripheral blood. Data analysis performed using R v.4.4.3 and Seurat v.5.2, and clustering was performed using Uniform Manifold Approximation and Projection (UMAP) approach. Each dataset analyzed separately, and then in combination with the blood sample removed. Harmony v.1.2.3 was used for batch correction of patient specific effects. Results: UMAP analysis of the primary and metastatic SACC with adjacent normal tissue resulted in clear visual overlap of SACC with related SACC lung metastases and normal head and neck tissue. A distinct clustering pattern from normal lung tissue appeared in close proximity to clusters of lung metastases as well as some cells segregated from the other three tissue types with similar cellular origins. PACC tissue from the GSE245170 dataset demonstrated a discrete pattern, unique from adjacent normal lung tissue and blood. Combined analysis of PACC, SACC and metastatic lung tissue scRNA from the two datasets with the exclusion of the blood sample and after batch correction resulted in a UMAP plot showing PACC cells cluster closely with lung metastases and normal lung tissue. Separate clusters are formed uniquely from primary SACC cells in close proximity to metastatic SACC cells, normal head and neck tissue, and PACC cells. Conclusions: Using UMAP approach to graphically plot scRNA sequencing data we identified varying degrees of overlap in molecular features across tissue samples from PACC, SACC with metastasis to lung, and normal lung, head and neck. ScRNA sequencing data from PACCs most closely resembled that from SACC lung metastases, with additional similarity to primary SACC and normal head and neck tissue. This raises the question of whether primary PACC diagnoses are instead oligometastatic SACCs without clear identification of primary head and neck disease either due to primary tumor regression, cramped head and neck anatomy, or aggressive nature of an early metastatic process.