Whole-genome single-cell copy number profiling from formalin-fixed paraffin-embedded samples

Abstract

A substantial proportion of tumors consist of genotypically distinct subpopulations of cancer cells. This intratumor genetic heterogeneity poses a substantial challenge for the implementation of precision medicine. Single-cell genomics constitutes a powerful approach to resolve complex mixtures of cancer cells by tracing cell lineages and discovering cryptic genetic variations that would otherwise be obscured in tumor bulk analyses. Because of the chemical alterations that result from formalin fixation, single-cell genomic approaches have largely remained limited to fresh or rapidly frozen specimens. Here we describe the development and validation of a robust and accurate methodology to perform whole-genome copy-number profiling of single nuclei obtained from formalin-fixed paraffin-embedded clinical tumor samples. We applied the single-cell sequencing approach described here to study the progression from in situ to invasive breast cancer, which revealed that ductal carcinomas in situ show intratumor genetic heterogeneity at diagnosis and that these lesions may progress to invasive breast cancer through a variety of evolutionary processes.

Figure 1. Schematic representation of the formalin-fixed paraffin-embedded (FFPE) single-cell sequencing platform

The procedure consists of 11 steps involving tissue microdissection, nuclei preparation and FACS analysis/sorting based on DAPI staining/DNA content (steps 1–5), single-nucleus DNA repair to correct for FFPE-induced DNA damage, whole-genome amplification (WGA), Illumina library preparation and multiplex sequencing (steps 6–10), and bioinformatics analysis (step 11). The multiplex PCR analysis is used to determine the quality of the DNA extracted from FFPE nuclei preparations, and is performed between steps 3 and 4.

Figure 2. Establishment of a whole-genome copy-number profiling method for single nuclei derived from FFPE samples

(a) Representative micrograph of the triple-negative invasive ductal carcinoma (Case 1) (hematoxylin-and-eosin; scale bar, 500 μm). (b) Ploidy flow cytometric profiles of nuclei extracted from FFPE/frozen tissue of Case 1 based on DAPI staining. (c) Multidimensional scaling of all FFPE and frozen nuclei sequenced. Each circle represents one cell (~3N FFPE (purple circles) and frozen (turquoise circles), n=36; 2N FFPE (purple circumference) and frozen (turquoise circumference), n=24). (d) Representative examples of genome-wide single-nucleus copy number (CN) profiles from FFPE (top) and frozen (middle) single nuclei derived from the ~3N (tumor) distribution. Bulk CN profile of 100,000 FFPE tumor cells is also shown (bottom). CN values of the bulk were rounded to the nearest integer. (e) Data bin-to-bin variance (expressed as the Sum of Square error - top panel) and percent concordance with bulk CN profiles (bottom panel) of FFPE repaired (n=36), FFPE unrepaired (n=24), frozen repaired (n=24) and frozen unrepaired (n=36) nuclei. Data are mean ± s.d. (f) Hierarchical clustering analysis of CN data from FFPE (n=33) and frozen (n=30) single cells from Case 1, using Manhattan distance and Ward’s method. (g) Close-up views of representative clonal CN alterations on chromosomes 5, 11 and 17 identified in FFPE (top) and frozen (bottom) from a minimum of 10 randomly selected single cells. Each colored line represents an individual single nucleus. Chr: chromosome, Sc: single cell.

Figure 3. Validation of the whole-genome CN method for FFPE-derived single nuclei using samples from Case 2, a synchronous DCIS and invasive breast cancer

(a) Representative micrograph of an estrogen receptor-positive/ERBB2 (HER2)-negative invasive breast cancer (mucinous type) with synchronous DCIS (hematoxylin-and-eosin; scale bar, 500 μm). (b) Ploidy flow cytometric profiles of nuclei extracted from DCIS/invasive and FFPE/frozen tissue of Case 2. (c) Representative examples of genome-wide single-cell copy number (CN) profiles from FFPE (top) and frozen (middle) single nuclei derived from the ~4N (tumor) distribution, and the bulk CN profile of the FFPE DCIS component (bottom). CN values of the bulk were rounded to the nearest integer. Red arrows indicate clonal CN events. (d) Bin-to-bin variance expressed as Sum of Square error of FFPE repaired (n=12), FFPE unrepaired (n=12) and frozen unrepaired (n=12) single cells (Methods). (e) Hierarchical clustering analysis of FFPE (n=120) and frozen (n=120) single cells from Case 2, using Manhattan distance and Ward’s method. Reference bars under the dendrogram indicate tumor component of origin (B1 and B2), type of tissue of origin (B3) and the clades (B4) defined by the conservative and arbitrary cut of the dendrogram (dotted line). The heatmap was constructed as described in the Methods. (f) Zoomed views of representative clonal CN alterations on chromosome 17 and chromosome 22 identified in FFPE (left) and frozen (right) single cells. (g) Zoomed views of representative subclonal CN alterations identified on chromosome 1 and chromosome 5 identified all FFPE (left) and frozen (right) nuclei sequenced. In the overlay images of (f) and (g) each colored line represents an individual single nucleus and a minimum of 10 randomly selected profiles are presented. Chr: chromosome, Sc: single cell.

Figure 4. Sequencing of FFPE single nuclei from samples of case 3 and phylogeny reconstruction of progression from DCIS to invasive breast cancer

(a) Representative micrograph of a diploid estrogen receptor-positive ERBB2 (HER2)-negative invasive ductal carcinoma (IDC) with synchronous DCIS (hematoxylin-and-eosin; scale bar, 500 μm). (b) Ploidy cytometry profiles of Case 3 FFPE DCIS and invasive (IDC) components. (c) Hierarchical clustering analysis of FFPE single nuclei (n=96) from Case 3, using Manhattan distance and Ward’s method. Reference bars: Bar 1 (B1), sample of origin from which the single cells were isolated (i.e. FFPE DCIS or FFPE IDC); Bar 2 (B2), classification of cells into normal, DCIS or IDC; Bar 3 (B3), subclonal populations identified in DCIS and IDC. Orange and black arrows (right side of heatmap) indicate clonal and subclonal alterations, respectively. Chr: chromosome, Sc: single cell. (d) Representative micrographs of fluorescence in situ hybridization analysis with probes for specific loci indicated in the Figure, confirming the existence of selected subclones present in the DCIS (DC-1 and DC-3) and IDC (IC-1 and IC-2) identified by single-cell sequencing (63× magnification). Percentages of the distinct clones as defined on the basis of quantification of FISH results as described in the Methods and Supplementary Table 2. (e) Phylogenetic trees of Case 3 based on single-cell CN data. Nodes in the trees correspond to observed subclones in DCIS (DC-1, DC-2, DC-3 and DC-4) and IDC (IC-1 and IC-2), which are color-coded to match the clusters in the dendrogram in panel c. The inferred most recent common ancestors (MRCAs) of subclones are indicated with a light-yellow circle with dashed outline. Values within circles indicate percentage of cells inferred as carrying distinct clonal/subclonal alterations. Connectors (thick black lines) are annotated with subclone specific CN alterations. The percentage of tumor cells (cancer cell fraction) harboring any given set of defining alterations is indicated for each subclone on the Y-axis of each DCIS and invasive phylogeny (i.e. the Y-axis coordinate of the top of each circle indicating a clone refers to its cancer cell fraction). For each phylogeny, branches were drawn to connect the clones for illustration purposes.

Figure 5. Whole-genome single-nucleus copy-number information for case 4, a suboptimal, overly damaged FFPE sample

(a) Representative micrograph of an estrogen receptor-positive/ERBB2 (HER2)-positive invasive ductal carcinoma (IDC) with DCIS. (b) Representative CN plots of single FFPE and frozen nuclei compared to the tumor bulk CN profile derived from whole exome sequencing data. (c) Zoomed in views of representative clonal alterations identified in single FFPE and frozen nuclei sequenced. Each colored line represents an individual single nucleus with a minimum of 10 randomly selected profiles presented. (d) Fluorescence in situ hybridization with probes for ERRB2 (HER2) highlighting the presence of clonal ERBB2 gene amplification, consistent with the single nuclei sequencing results. (e) Hierarchal clustering analysis of all single nuclei (FFPE, n=48 and frozen, n=48) sequenced from Case 4. Reference bars on the top indicate the tumor component of origin (B1) and type of tissue of origin (B2). Hierarchical clustering heatmap was constructed using CN profiles of individual nuclei using Manhattan and Ward as the distance function and clustering method, respectively (Methods). Chr: chromosome, Sc: single cell.