Cancer Hallmarks Define a Continuum of Plastic Cell States between Small Cell Lung Cancer Archetypes

Abstract

Small Cell Lung Cancer (SCLC) tumors are heterogeneous mixtures of transcriptional subtypes. Understanding subtype dynamics could be key to explaining the aggressive properties that make SCLC a recalcitrant cancer. Applying archetype analysis and evolutionary theory to bulk and single-cell transcriptomics, we show that SCLC cells reside within a cell-state continuum rather than in discrete subtype clusters. Gene expression signatures and ontologies indicate each vertex of the continuum corresponds to a functional phenotype optimized for a cancer hallmark task: three neuroendocrine archetypes specialize in proliferation/survival, inflammation and immune evasion, and two non-neuroendocrine archetypes in angiogenesis and metabolic dysregulation. Single cells can trade-off between these defined tasks to increase fitness and survival. SCLC cells can easily transition from specialists that optimize a single task to generalists that fall within the continuum, suggesting that phenotypic plasticity may be a mechanism by which SCLC cells become recalcitrant to treatment and adaptable to diverse microenvironments. We show that plasticity is uncoupled from the phenotype of single cells using a novel RNA-velocity-based metric, suggesting both specialist and generalist cells have the capability of becoming destabilized and transitioning to other phenotypes. We use network simulations to identify transcription factors such as MYC that promote plasticity and resistance to treatment. Our analysis pipeline is suitable to elucidate the role of phenotypic plasticity in any cancer type, and positions SCLC as a prime candidate for treatments that target plasticity.

Spatial epitope barcoding reveals subclonal tumor patch behaviors

“…The NCI-H82 xenografts analyzed in this study had heterogeneity demarcated within distinct regions of the tumor. From a global non-spatially defined perspective, this heterogeneity is in line with previous studies indicating that NCI-H82 cells exist in a rather undefined region of the low-dimensional space from the continuum of functional states of several SCLC cell lines (Groves et al 2021). In our dataset, we identified three major states of NCI-H82 cells: NE cells, NE cells lacking H3K27ac, and non-NE cells.…”

“…Similar to other subtypes of SCLC, SCLC-N tumors are heterogeneous and harbor cancer cells with neuroendocrine (NE) and non-neuroendocrine (non-NE) features (Ireland et al 2020; Stewart et al 2020). Single-cell RT-qPCR analyses suggested that cells with NE and non-NE features co-exist in NCI-H82 cells in culture (Lim et al 2017; Groves et al 2021). Accumulating evidence indicates that the interplay between NE and non-NE cells is critical for tumor growth and evolution in SCLC (reviewed in (Shue, Lim, and Sage 2018)).…”

Computational analysis reveals similarities and differences between SCLC subtypes

“…Thus, we plotted the hierarchical clustering dendrogram to visualize five gene clusters prevalent in CCLE and GSE73160 datasets (Fig 2A, i; 2B, i). The gene-set clusters identified using these dendrograms had a high degree of similarity with the composition of groups reported earlier (Groves et al, 2021 S3). The extent of overlap between the gene clusters identified in our study and the ones previously reported is represented as an overlap matrix.…”

“…These studies, and others, led to a classification of SCLC into four consensus phenotypes defined by enrichment of TFs: ASCL1 (SCLC-A), NEUROD1 (SCLC-N), YAP1 (SCLC-Y) and POU2F3 (SCLC-P) (Rudin et al, 2019). Another subtype -SCLC-A2 -driven by ASCL1 but different from SCLC-A has been proposed (Wooten et al, 2019), also supported by archetypal analysis of SCLC phenotypic landscape which proposed a 105-gene signature to classify phenotypes into five categories (Groves et al, 2021). Further, another SCLC subtype -marked not by enrichment of canonical TFs, but by enrichment of inflammation and immune-related genes (SCLC-I) -was proposed (Gay et al, 2021), further complicating ongoing attempts to arrive at a consensus of SCLC phenotypes and their distinct molecular and functional footprints.…”

“…The 105 gene list comprised of markers for the 5 subtypes (SCLC-A, SCLC-A2, SCLC-N, SCLC-P and SCLC-Y) identified using archetypal analysis (Groves et al, 2021). Thus, we plotted the hierarchical clustering dendrogram to visualize five gene clusters prevalent in CCLE and GSE73160 datasets (Fig 2A, i; 2B, i).…”