Intact Cell Dissociation of Formalin-Fixed Paraffin-Embedded Colon Tissue Enhances Cell Recovery and Gene Detection in Single-Cell RNA Sequencing

Graphical abstract

Single-cell RNA sequencing (scRNA-seq) has transformed the understanding of tissue heterogeneity at a cellular resolution.¹ However, its application to archived human specimens, particularly formalin-fixed paraffin-embedded (FFPE) tissue, remains limited despite the widespread clinical use of FFPE preservation. Recent developments in single-cell sequencing technologies for FFPE tissue are largely limited to single-nucleus RNA sequencing applications, as chemical crosslinking and RNA fragmentation under FFPE conditions hinder recovery of intact whole cells.² How these effects alter cellular composition and bias gene expression profiles, particularly in fragile epithelial and stromal populations that are tightly coupled to the extracellular matrix and tissue architecture, remains incompletely understood.3, 4, 5

Here, we advance a tissue dissociation strategy for FFPE scRNA-seq that is based on FFPE-DISSECT (Disaggregation for Intracellular Signaling in Single Epithelial Cells from Tissue), originally developed for signaling protein analysis using cytometry by time-of-flight.^6,7 Our single-cell approach preserves intact cells and, therefore, cytoplasmic RNA, enabling transcriptomic profiling using the 10x Genomics Chromium Fixed RNA Profiling assay (Figure 1A). We benchmarked our approach against a conventional nuclei isolation protocol on 2 contrasting tissue types: colon and thymus. We also incorporated imaging cytometry into our workflow using the Thermo Fisher Attune CytPix, enabling high-throughput analysis to quantitatively validate cell morphology integrity prior to single-cell sequencing.

Figure 1.

FFPE-DISSECTenables whole-cell dissociation in colonic and thymic epithelium. (A) Schematic overview of the FFPE-DISSECT workflow for generating single whole cells, with downstream applications including imaging flow cytometry and single-cell RNA sequencing (scRNA-seq). (B) Representative images from Attune CytPix imaging flow cytometer for colonic tissue. Left: Nuclei (standard preparation); and Right: Whole-cell (FFPE-DISSECT preparation). (C) Top: Distribution of cell circularity (%) measured using the Attune CytPix imaging flow cytometer in colonic tissue. Bottom: Distribution of cell major diameter measured using the Attune CytPix imaging flow cytometer in colonic tissue. (D) Quantification of cell circularity and major diameter length in colonic tissue (N = 5; 1-tailed t test). (E) Representative images from Attune CytPix imaging flow cytometer for thymic tissue. Left: Nuclei (standard preparation); and Right: Whole-cell (FFPE-DISSECT preparation). (F) Top: Distribution of cell circularity (%) measured using the Attune CytPix imaging flow cytometer in thymic tissue. Bottom: Distribution of cell major diameter measured using the Attune CytPix imaging flow cytometer in thymic tissue. (G) Quantification of cell circularity and major diameter length in thymic tissue (N = 3; 1-tailed t test). (B–G) Blue: Standard preparation; pink: FFPE-DISSECT preparation. (D and G) ns, not significant (P > .05); ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. DISSECT, disaggregation for intracellular signaling in single epithelial cells from tissue; FFPE, formalin-fixed paraffin-embedded; FLEX, fixed RNA profiling.

Intact cells were visualized in suspensions prepared using FFPE-DISSECT, in contrast to nuclei from standard nuclei dissociation of human colonic tissue (Figure 1B; Supplementary Figure 1A). To quantify whether intact cells were preserved in large numbers, we used imaging flow cytometry to assess 2 morphologic features: circularity percentage and major diameter for each image event. Preparations from FFPE-DISSECT on human colonic tissue showed a significant decrease in circularity and a significant increase in major diameter compared with standard nuclei dissociation (Figure 1C and D; Supplementary Figure 1B–D), consistent with the preservation of canonical columnar epithelial cell shape. Conversely, thymic tissue showed mostly circular events from both approaches but a significant increase in major diameter in FFPE-DISSECT (Figure 1E–G; Supplementary Figure 1E–H). These findings are consistent with the thymus being composed mainly of small, round lymphocytes while accounting for their larger profiles reflected by whole cells recovered by FFPE-DISSECT. These data support that FFPE-DISSECT effectively dissociates cells, instead of nuclei, from FFPE tissues while retaining cytoplasmic contents.

To determine whether intact cell dissociation yields high data quality, we generated 28 human colon scRNA-seq datasets across FFPE-DISSECT and standard nuclei dissociation methods, with all datasets integrated in a common data space and clustered by cell type (Figure 2A; Supplementary Figure 2A–D). FFPE-DISSECT significantly reduced ambient RNA contamination, as quantified by Ambiquant (Supplementary Figure 3),⁸ while increasing the number of detected genes and unique molecular identifiers, suggesting improved overall data quality (Figure 2B). In support of this, FFPE-DISSECT colonic samples exhibited a higher number of marker genes when differentially expressed gene (DEG) analysis was performed, regardless of whether cell numbers or gene counts were downsampled to equal those of standard dissociation (Supplementary Figure 2E). Comparison of the top 50 DEGs per cell type revealed substantial overlap between standard and FFPE-DISSECT methods, indicating consistent cell identity classification (Figure 2C). However, FFPE-DISSECT identified more cell-type-representative markers, including MZB1 and JCHAIN in plasma cells, which were not detected in standard preparations (Figure 2D). Additionally, several genes showed improved ranking in FFPE-DISSECT, such as AQP3 (rank 83 to 2) and ATOH1 (rank 27 to 8) for goblet cells or PLVAP (rank 15 to 6) for endothelial cells, reflecting enhanced marker resolution (Figure 2D; Supplementary Figure 2F). FFPE-DISSECT also resulted in a significant increase in the number of high-quality cells retained after filtering (Figure 2B). Analysis of 7 paired colon samples (Supplementary Figure 2G and H) showed that FFPE-DISSECT preferentially preserves fragile epithelial and stromal cell populations, such as fibroblasts, which are frequently under-represented in standard scRNA-seq workflows due to dissociation-induced loss (Figure 2E–G; Supplementary Figure 2I–K).^3,4,9 Overall, FFPE-DISSECT enhances scRNA-seq performance in archived colonic tissue by increasing cell recovery and gene detection across cell types when compared with standard nuclei dissociation.

Figure 2.

FFPE-DISSECTscRNA-seqin colonic tissue yields increased high-quality cells and improved cell-type gene representation. (A) Uniform Manifold Approximation and Projection (UMAP) of cell type annotations for colonic tissue samples (N = 28). (B) Quantification of quality control metrics for colon scRNA-seq datasets (Mann–Whitney U test). (C) Top 50 differentially expressed genes (DEGs) for each annotated cell type in colon samples comparing standard and FFPE-DISSECT preparations. (D) Top 50 FFPE-DISSECT–enriched gene ranks for goblet, endothelial, and plasma cells. (E) Cell type composition in paired colon samples (n = 14; 7 standard preparations, 7 FFPE-DISSECT preparations). Vertical error bars: Standard error of the mean (SEM) in FFPE-DISSECT samples. Horizontal error bars: SEM in standard samples. (F) Composition of fibroblast and endothelial cell populations in paired colon samples (paired t test). (G) Raw cell class counts in paired samples grouped by lineage: immune, stromal, and epithelial. Immune: B, T, mast, plasma, macrophage, and monocyte cells. Stromal: myofibroblast, fibroblast, pericyte, and endothelial cells. Epithelial: colonocyte, neuroendocrine, enterocyte, goblet, and tumor cells. (H) UMAP of cell type annotations for thymic tissue samples (N = 16). (I) Quantification of quality control metrics for thymus scRNA-seq datasets (Mann–Whitney U test). (J) Top 50 DEGs for each annotated cell type in thymus samples comparing standard and FFPE-DISSECT preparations. (K) Raw cell class counts in paired thymus samples (n = 4; 2 standard preparations, 2 FFPE-DISSECT preparations) grouped by lineage: stromal, epithelial, myeloid, and lymphocyte. Stromal: fibroblast and endothelial cells. Epithelial: cTEC and mTEC cells. Myeloid: macrophage, monocyte, aDC, cDC1, pDC, and erythrocyte cells. Lymphocyte: B, CD8aa, DN, DP(P), DP(Q), SP(CD4+), SP(CD8+), Treg, αβT, and NK cells. (B and I) From left to right: ambient RNA contamination score (lower is better), mitochondrial read fraction, total genes detected, total unique molecular identifiers (UMIs), and total cell count. (B–G, I–K) Blue: standard preparation; pink: FFPE-DISSECT preparation; purple: overlap/share between preparations. (B, F and I) ns, not significant (P > .05); ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. DISSECT, disaggregation for intracellular signaling in single epithelial cells from tissue; FFPE, formalin-fixed paraffin-embedded; scRNA-seq, single-cell RNA sequencing.

The same scRNA-seq comparison was performed on human thymus FFPE specimens, yielding 16 newly generated datasets (Figure 2H; Supplementary Figure 4A–C). In contrast to the colon, FFPE-DISSECT did not reduce ambient RNA contamination and instead showed an increase in mitochondrial reads (Figure 2I). Although FFPE-DISSECT significantly increased the number of detected genes and unique molecular identifiers (Figure 2I), this did not translate into improved DEGs, with many cell types having more DEGs in standard dissociation conditions (Supplementary Figure 4D), and reduced overlap in the top 50 DEGs between standard and FFPE-DISSECT preps across cell types compared with the colon (Figure 2J). Importantly, FFPE-DISSECT did not result in a higher absolute number of high-quality cells after filtering (Figure 2I and K), whereas overall cell type composition remained comparable between methods (Supplementary Figure 4E). FFPE-DISSECT does not improve scRNA-seq performance in the thymus, consistent with cellular morphologic analysis of this lymphocyte-rich tissue (Supplementary Figure 4E), where the low cytoplasmic RNA content means that nuclei capture alone yields sufficient transcriptome quality.

In this study, we demonstrate that the recovery of intact whole cells from FFPE archival blocks substantially improves the scRNA-seq data quality for colonic tissues. Large and cytoplasm-rich cells, such as columnar colonocytes, exhibit reduced ambient RNA contamination and increased transcript detection, translating into improved cell-type–specific gene representation and potentially more robust downstream analysis. This result is consistent with the expected loss of information when the cytoplasm, an RNA-rich compartment, is excluded, as in single-nucleus RNA sequencing, where only nuclei are analyzed. This phenomenon has been previously documented when nuclei, rather than whole cells, are used for sequencing in fresh samples.¹⁰

In addition to improving the epithelial cell quality, FFPE-DISSECT also enhanced the recovery of fragile cell populations, including fibroblasts, which are often under-represented and require careful dissociation strategies.^5,9 Our study thus establishes cellular integrity and cytoplasmic content as key determinants of data quality in scRNA-seq of FFPE archival tissue blocks and identifies dissociation strategy as a source of cell-type representation bias. More broadly, our dissociation strategy unlocks the untapped potential of FFPE tissue archives for high-resolution, probe-based single-cell transcriptomic studies, providing a foundation for additional probe-based molecular profiling assays, with relevance to retrospective investigations of gastrointestinal diseases.