Fitness and transcriptional plasticity of human breast cancer single-cell-derived clones

Abstract

Clonal fitness and plasticity drive cancer heterogeneity. We used expressed lentiviral-based cellular barcodes combined with single-cell RNA sequencing to associate single-cell profiles with in vivo clonal growth. This generated a significant resource of growth measurements from over 20,000 single-cell-derived clones in 110 xenografts from 26 patient-derived breast cancer xenograft models. 167,375 single-cell RNA profiles were obtained from 5 models and revealed that rare propagating clones display a highly conserved model-specific differentiation program with reproducible regeneration of the entire transcriptomic landscape of the original xenograft. In 2 models of basal breast cancer, propagating clones demonstrated remarkable transcriptional plasticity at single-cell resolution. Dichotomous cell populations with different clonal growth properties, signaling pathways, and metabolic programs were characterized. By directly linking clonal growth with single-cell transcriptomes, these findings provide a profound understanding of clonal fitness and plasticity with implications for cancer biology and therapy.

Keywords: CP: Cancer; breast cancer; cancer stem cells; cellular barcoding; clonal heterogeneity; clonal tracking; patient-derived tumor xenografts; plasticity; single-cell sequencing.

Graphical abstract

Figure 1

Landscape of 20,000 single-cell-derived clones reveals significant heterogeneity (A) Overall experimental schematic. (B) Overview of all clones detected in primary barcoded xenografts. Each circular diagram represents a single PDTX model, with each concentric circle representing the number and proportion representation of all clones detected in each replicate xenograft in order of highest (innermost circle) to lowest (outermost circle) cell dose implanted. Colors cannot be compared between xenograft model replicates, as none of the clones detected in individual xenografts are related.

Figure 2

Differences in clone-propagating activity revealed in secondary xenografts (A) Clones detected upon passaging of xenografts. Where a propagating clone was detected in multiple secondary xenograft replicates, the mean clone size is shown. (B) Number of clones detected in primary and secondary xenografts from multiplexed DNA amplicon sequencing. (C) CIC frequency for primary xenografts and frequency of propagating clones are shown for basal (black) and luminal B (gray) models from (A). Boxplots were computed using the median of the observations (center line). The first and third quartiles are shown as boxes, and the whiskers extend to the +/− 1.58 interquartile range divided by the square root of the sample size. Outliers are shown as dots. (D) The proportion of cells represented by propagating, transient, and emerging clones in primary (P) and secondary (S1–S3) xenograft replicates is shown. Same color legend as (A).

Figure 3

Clone in vivo doubling times differ by breast cancer subtype (A) Density distribution of in vivo doubling time calculated for all clones across 26 PDTX models (top). In silico simulations for models with 2 (middle-HCI010) and 3 (bottom, STG139) distributions are shown. Actual experimental data in green, mclust modeling in blue, and in silico simulation in red. (B) Summary of the characteristics of 3 distinct clone types defined by their in vivo doubling time. (C) Proportion of all cells from xenografts of each breast cancer subtype that belong to a clone with a fast, medium, or slow doubling time. Asterisks indicate statistical significance from chi-squared test (p < 2.2 × 10⁻¹⁶), indicating that the proportions of fast, medium, and slow clones are significantly different between subtypes. (D) Distribution of clones based on their in vivo doubling times, shown by subtype. Vertical black lines indicate the cutoffs between the Gaussian distributions for fast, medium, and slow clones as defined in (A) and (B).

Figure 4

Dominant propagating clones regenerate full PDTX model-specific transcriptional landscape (A) Two-dimensional plot of scRNA-seq data from 5 PDTX models (18 xenografts) showing distinct transcriptional cell states. (B) Ternary plot for each clone detected by scRNA-seq in primary xenografts shows the proportion of total unique molecular indexes (UMIs) for each clone that correspond to normal epithelial gene signatures for basal, LP, and/or ML cells. Vertical dashed line indicates the separation between clones from basal and luminal B models. (C) Distribution of cells per clone by model. Cell states represented by gray open circles. The largest transient clones are in black and the dominant propagating clones are colored by model. AB040 has a different dominant propagating clone in each secondary xenograft replicate, and these are distinctly colored. (D) Number and proportion of all clones detected by DNA amplicon sequencing for each PDTX model. The primary xenograft is represented by the innermost ring, followed by secondary xenograft replicates S1, S2, and S3 in concentric rings outward. The same colors between rings within each PDTX model indicate these are the same clones. The dominant propagating clones are colored the same as (C). (E) Correlation of the proportion contributed to each cell state between secondary xenograft replicates for each dominant propagating clone for STG139 and STG201. Blue lines show the linear correlations, and shaded area indicates the standard error. Adjusted R² is also provided for each correlation.

Figure 5

Dichotomous cell populations with differential signaling and metabolic responses in basal breast cancer distinguish between functional clone types (A) Epithelial and mesenchymal signature expression (from scRNA-seq) for all barcoded cells in STG139 color coded by clone type. Dashed lines indicate the threshold above and below which cells belonging to fraction 1 and fraction 2 are defined. Histogram density plots are also shown. (B) Density distribution of all barcoded cells from STG201 colored by clone type. The vertical dashed line defines the threshold above and below which cells belonging to fraction 1 and fraction 2 are defined, based on an epithelial signature score of 0.15, coinciding with the inflection point between the bimodal distribution in epithelial signature score for propagating clones analyzed in the primary xenograft. (C) Scatterplots compare signature expression levels for fraction 1 (cyan) and 2 (magenta) from STG139 related to (A). Density plots are also shown. (D) Same as (C) but for STG201.

Figure 6

Transcriptional similarity analysis of a dominant propagating clone reveals dynamic transcriptional plasticity (A) Transcriptional cell states represented in a two-dimensional plot colored by transcriptional proximity. (B) Gene-gene correlation plot where strong and highly variable expressed genes are clustered into 18 gene modules (GMs). (C) Fold-enrichment (y axis) over transcriptional distance. Error bars are standard error of mean. Dotted horizontal line indicates no enrichment. Top 5 enriched transcription factors and genes within the GM are indicated on the left and right of the plots, respectively. (D) Gene enrichment plots over transcriptional distance. Each data point represents a single cell state. Horizontal black bars indicate the mean log2 enrichment. Horizontal dashed line indicates no enrichment. All plots shown are for clone 1 from basal model STG139.