Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Jan 24;6(1):37-51.e9.
doi: 10.1016/j.cels.2017.10.012. Epub 2017 Nov 15.

Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut

Charles A Herring  1 , Amrita Banerjee  2 , Eliot T McKinley  3 , Alan J Simmons  2 , Jie Ping  4 , Joseph T Roland  5 , Jeffrey L Franklin  2 , Qi Liu  4 , Michael J Gerdes  6 , Robert J Coffey  7 , Ken S Lau  8
Affiliations

Unsupervised Trajectory Analysis of Single-Cell RNA-Seq and Imaging Data Reveals Alternative Tuft Cell Origins in the Gut

Charles A Herring et al. Cell Syst. .

Abstract

Modern single-cell technologies allow multiplexed sampling of cellular states within a tissue. However, computational tools that can infer developmental cell-state transitions reproducibly from such single-cell data are lacking. Here, we introduce p-Creode, an unsupervised algorithm that produces multi-branching graphs from single-cell data, compares graphs with differing topologies, and infers a statistically robust hierarchy of cell-state transitions that define developmental trajectories. We have applied p-Creode to mass cytometry, multiplex immunofluorescence, and single-cell RNA-seq data. As a test case, we validate cell-state-transition trajectories predicted by p-Creode for intestinal tuft cells, a rare, chemosensory cell type. We clarify that tuft cells are specified outside of the Atoh1-dependent secretory lineage in the small intestine. However, p-Creode also predicts, and we confirm, that tuft cells arise from an alternative, Atoh1-driven developmental program in the colon. These studies introduce p-Creode as a reliable method for analyzing large datasets that depict branching transition trajectories.

Keywords: cell-state transitions; differentiation hierachies; graph theory; intestine and colon; mass cytometry; pseudo-time analysis; single-cell RNA-seq; single-cell biology; trajectories; tuft cells.

PubMed Disclaimer

Figures

Figure 1
Figure 1. The p-Creode algorithm for analyzing single-cell data
(i) Synthetic dataset representing single cells in two-dimensional expression space with five end-states and three branch points. Overlay represents density of cells. (ii) Density-normalized representation of the original dataset from down-sampling. Overlay represents the density after down-sampling. (iii) Density-based k-nearest neighbor (d-kNN) network constructed from down-sampled data. Overlay represents the graph measure of closeness centrality derived from the d-kNN network, which is a surrogate for cell state (low – end-state, high – transition state). (iv) End-states identified by K-means clustering and silhouette scoring of cells with low closeness values (<mean). The number of end-state clusters is doubled to allow for rare cell types. End-state clusters are colored, and open circles represent the centroid per cluster. (v) Topology constructed with a hierarchical placement strategy of cells on path nodes between end-states (red), which allows for the placement of data points along an ancestral continuum. Overlay represents the original density of cells. (vi) Aligned topology (red) with maximal consensus though iterative assignment and repositioning of path nodes using neighborhood cell densities. (vii) Representative topology extracted using p-Creode scoring from an ensemble of N topologies. Node size in the output graph represents the original density of cells. See also Figure S1–S3.
Figure 2
Figure 2. p-Creode analysis of single-cell mass cytometry data identifies the hematopoietic differentiation hierarchy
(A) t-SNE analysis of a 13-marker panel mass cytometry dataset from Bendall et al. Cell types, as defined by clusters on the t-SNE map, were manually annotated. Overlay represents CD3 levels. (B) p-Creode analysis of the same dataset in A. The most representative graph over N=100 runs, as defined by the graph with the minimum p-Creode score when compared to all graphs in the analysis, is represented. Colored outlines indicate cell types defined in A, and overlay indicates CD3 levels. (C) Two random runs of the same dataset in A using SPADE (200 nodes), with the same color scheme for cell types and overlay in B. (D) Two random runs (N=100) of the same dataset in A using p-Creode, with the same color scheme for cell types and overlay in B. (E) Comparison of the robustness of p-Creode, SPADE run with 133 nodes, and SPADE run with 200 nodes. Each data point represents the mean p-Creode score calculated for each resulting graph (N=100). Boxes show the quartiles while whiskers show the minimum and maximum scores. See also Figure S4–S5.
Figure 3
Figure 3. p-Creode analysis of single-cell mass cytometry data generates topologies that reflect thymic T cell development
(A) p-Creode analysis of the first replicate 14-marker mass cytometry dataset from Setty et al. with PCA preprocessing, representative of N=100 runs. Cell populations were manually labeled. Overlay represents CD3 levels. (B, C) Marker trends along p-Creode trajectories with Diffusion Map preprocessing for CD8+ SP (B) and CD4+ SP (C) trajectories. Trends are similar to results obtained by Wishbone analysis and consistent with established stages of T cell differentiation. (D) t-SNE analysis of the Setty et al. dataset with manual annotation of clusters, including population X identified by p-Creode. (E) Diffusion Map of the dataset depicting T cell maturation. See also Figure S6–S7.
Figure 4
Figure 4. p-Creode analysis of single-cell multiplex immunofluorescence (MxIF) data reveals an alternate origin for tuft cells in small intestine versus colon
(A, B) MxIF images where quantitative single-cell data are derived by extracting segmented cell objects using a combined, “supermembrane” mask. Example staining for differentiated, transit-amplifying (TA), and stem cell markers in the small intestinal (A) and the colonic epithelium (B). (C, D) t-SNE analysis on 19-marker MxIF datasets of the small intestinal (C) and the colonic epithelium (D). Cell types, as defined by clusters on the t-SNE map, were manually annotated. Overlay represents DCLK1 levels. (E, F) p-Creode analysis of datasets in E and F with the most representative graphs over N=100 runs, for small intestine (E) and colon (F). Overlay represents DCLK1 levels. (G) Hierarchical clustering of major epithelial cell types by their response to in vivo stimulation by TNF. Clustering on all normalized signals (indicated by heat map) measured by DISSECT-CyTOF. See also Figure S8–S14.
Figure 5
Figure 5. Tuft cells have alternative specification requirements in small intestine versus the colon
(A) Control (Lrig1+/+;Atoh1fl/fl + tamoxifen) and (B) epithelial-specific Atoh1 ablated (Lrig1CreERT2/+;Atoh1fl/fl + tamoxifen) duodenum, with acute ablation of Atoh1 at 8 weeks of age and analysis performed 2 weeks later. Analysis of Paneth (Lysozyme+), goblet (CLCA1+), and tuft (DCLK1+; p-EGFR+) cells. Inset represents a multi-marker tuft cell signature of cells on the villi with certain markers (p-STAT6, p-EGFR) demonstrating an apical tuft staining pattern. (C) Control and (D) epithelial-specific Atoh1 ablated colon, analyzed the same way as A,B. (E,F) Quantitative analysis of DCLK1+ cells from images per crypt or villus in the small intestine (E) and colon (F). Error bars represent SEM from n=3 animals. **P<0.01, *P<0.05 by t-test. See also Figure S15.
Figure 6
Figure 6. Application of p-Creode on published scRNA-seq data reveals multi-branching topologies
(A) p-Creode analysis of the scRNA-seq dataset generated from alveolar cells by Treutlein et al. Cells collected over multiple developmental time points were mixed and analyzed together. Overlay represents developmental time that was recovered. (B) Overlay of selected transcripts depicting alveolar cell differentiation on the p-Creode topology generated in A. (C) p-Creode analysis of the scRNA-seq dataset generated from myeloid progenitor cells by Paul et al., most representative graphs over N=100 runs. Overlay represents Elane transcript levels. Inset represents an accepted model of myeloid differentiation. (D) Overlay of selected transcripts depicting myeloid cell differentiation on the p-Creode topology generated in C. Overlays represent ArcSinh-scaled gene expression data. See also Figure S16.
Figure 7
Figure 7. inDrop scRNA-seq reveals the developmental trajectory of Reg4+ secretory cells in the murine colon
(A) Human versus mouse β-actin transcript count by mapping to human and mouse reference genomes, respectively. Each data point represents a single cell. (B) t-SNE analysis of scRNA-seq data demonstrating the absence of segregation of data points from 2 replicates. (C) t-SNE analysis of murine colonic cells using scRNA-seq data. Cell types, as defined by clusters corresponding to specific cell type markers on the t-SNE map, were manually annotated. Overlay represents Krt8 transcript levels. (D) Overlay of selected transcripts depicting colonic cell lineages on the t-SNE map generated in C. (E) p-Creode analysis of scRNA-seq data generated by inDrop from colonic epithelial cells, most representative graph over N=100 runs. Overlay represents Muc2 transcript levels. (F) Overlay of selected transcripts depicting colonic epithelial cell differentiation on the p-Creode topology generated in E. Overlays represent ArcSinh-scaled gene expression data. See also Figure S17.
See this image and copyright information in PMC

Comment in

References

    1. Amir E-AD, Davis KL, Tadmor MD, Simonds EF, Levine JH, Bendall SC, Shenfeld DK, Krishnaswamy S, Nolan GP, Pe’er D. viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leukemia. Nat. Biotechnol. 2013;31:545–552. - PMC - PubMed
    1. Anchang B, Hart TDP, Bendall SC, Qiu P, Bjornson Z, Linderman M, Nolan GP, Plevritis SK. Visualization and cellular hierarchy inference of single-cell data using SPADE. Nat. Protoc. 2016;11:1264–1279. - PubMed
    1. Barker N, van Es JH, Kuipers J, Kujala P, van den Born M, Cozijnsen M, Haegebarth A, Korving J, Begthel H, Peters PJ, et al. Identification of stem cells in small intestine and colon by marker gene Lgr5. Nature. 2007;449:1003–1007. - PubMed
    1. Bendall SC, Simonds EF, Qiu P, Amir E-AD, Krutzik PO, Finck R, Bruggner RV, Melamed R, Trejo A, Ornatsky OI, et al. Single-cell mass cytometry of differential immune and drug responses across a human hematopoietic continuum. Science. 2011;332:687–696. - PMC - PubMed
    1. Bendall SC, Davis KL, Amir E-AD, Tadmor MD, Simonds EF, Chen TJ, Shenfeld DK, Nolan GP, Dana P. Single-Cell Trajectory Detection Uncovers Progression and Regulatory Coordination in Human B Cell Development. Cell. 2014;157:714–725. - PMC - PubMed

Publication types