Co-mapping clonal and transcriptional heterogeneity in somatic evolution via got-multi.

Abstract

Abstract Somatic evolution drives cancer progression and therapy resistance. Underlying phenotypic progressions is the development of marked clonal heterogeneity in most cancers. However, deciphering the downstream molecular effects of clonal heterogeneity in primary cancer has been limited due to the inability to isolate subclones by standard methods. While single-cell RNA-sequencing (scRNA-seq) provides high-resolution mapping of heterogeneous cell states, it cannot link transcriptional programs to cancer-driving mutations in these individual cells. We and others developed methods to detect genotypes and their transcriptional outputs in single cells (e.g., Genotyping of Transcriptomes; Nature, 2019). However, these approaches have technical limitations to broadly profile numerous mutations, which is nonetheless required for clonally complex neoplasms. The existing approaches are also restricted to fresh or frozen samples, limiting the use of the widely available pathology formalin fixed paraffin embedded (FFPE) tissues. To address these challenges, we developed Genotyping of Transcriptomes for Multiple Targets and Sample Types, GoT-Multi, a next generation single-cell multi-omics method compatible with FFPE tissues that integrates multiplexed genotyping with transcriptomic profiling. GoT-Multi adapts a probe-based scRNA-seq method (Flex Fixed RNA Profiling, 10x Genomics) by including custom mutation-specific probes along with the standard transcriptional probes. Furthermore, we developed an ensemble-based machine learning pipeline, GoT-Multi-ML, that optimizes the genotyping calls by denoising technical artifacts. GoT-Multi-ML leverages multiple machine learning models to learn the non-linear patterns of the genotyping data features (e.g., reads per genotyping transcript, GC content of probes), agnostic to the whole transcriptomic data. We validated GoT-Multi via a cell line-mixing experiment using SK-BR-3 and MCF-7 cell lines. GoT-Multi detected the targeted mutations (HIST1H1C A24T and S100A10 A77V) in the expected cell line (MCF-7) with an average genotyping rate of 72% and accuracy of 97%.To define the impact of clonal evolution on phenotypic progression, we focused on a prototype of cancer evolution, the progression of chronic lymphocytic leukemia (CLL) to therapy-resistant large B-cell lymphoma (LBCL) called Richter Transformation (RT). We applied GoT-Multi to a unique cohort of primary nodal cryopreserved samples containing both CLL and LBCL components (n = 5). We targeted 18 mutations (1-6 targets per sample) with a 98% median accuracy. We observed heterogeneous cancer cell states, including proliferating, stress response and inflammatory (n = 51,465 cells). LBCL was enriched in proliferating and cyclin D2-elevated states, whereas CLL was associated with a stress response signature. GoT-Multi enabled us to reconstruct the clonal architectures, up to 4 distinct subclones per lymphoma, and their associated cell states. Differential gene expression between mutated and wildtype cells revealed that genetically heterogenous subclones, including those with SRRM2 and JUNB mutations and therapy resistance-associated mutations in PLCG2 and BTK, displayed enrichment in inflammatory cell states, upregulating TNF and interferon signaling genes while suppressing cell cycle-related genes. Conversely, other subclones with mutations in, e.g., IRF8, POU2F2, PRKDC were associated with MYC activation and/or enhanced proliferation. These findings suggested that heterogeneous genotypes may converge on similar downstream transcriptional cell states. In addition, TNF itself was overexpressed in the proliferative subclones, suggesting that TNF production from these subclones may enhance inflammatory signaling of the TNF-responsive subclones, indicative of crosstalk between subclones. Finally, we demonstrated the applicability of GoT-Multi to pathology archived FFPE tissues. We profiled 9 mutations with an accuracy of 98% in an FFPE RT sample (n = 2,140 nuclei), allowing us to detect 5 subclones and their associated cell states. In summary, co-mapping of clonal and cell state heterogeneity at single cell resolution through GoT-Multi suggested that heterogeneous subclonal genotypes may converge on similar downstream oncogenic pathways to enhance overall tumor fitness. We envision that the broad applicability of GoT-Multi may help uncover the molecular underpinnings of cancer progression and therapy-resistance across oncology.