P-283 Computational Analysis from Single-cell Protein Data Reveals the Tuft Cell Differentiation Pathway in the Gut

    loading  Checking for direct PDF access through Ovid

Abstract

Background:

Tuft cells are a specialized epithelial cell type with chemosensory functions. Recent studies revealed their critical roles in modulating type 2 immune responses (Th2 and ILC2) in parasitic infections, suggesting their role in sensing the luminal microbiome. Tuft cell development has been a topic of debate in the field, with studies arguing for or against its secretory origin. Recent development of single-cell analysis techniques have enabled the multiplexed evaluation of all cell populations within a given tissue. In a renewing tissue such as the bone marrow and the gut, the majority of cellular transition states spanning stem to differentiated cells can be sampled. However, computational analysis tools that can robustly and statistically interrogate single-cell data to reveal biological insights regarding cell transitions are only emerging.

Methods:

We have developed a computational algorithm, p-Creode, that utilizes densities, graph theory, and hierarchical-tiered placement of data to infer temporal transition from single-cell data taken as a snap-shot in time. A Gromov–Hausdorff-based scoring metric was developed to statistically score the robustness of each result. We applied p-Creode to single-cell data generated by mass cytometry from the human bone marrow to benchmark our algorithm against known hematopoietic cell development. We generated multiplex immunofluorescence (MxIF) data on murine intestinal cell specification, with the capability to visualize and quantify >50 antibody stains per tissue section. p-Creode was applied to infer cell state transitions on this MxIF dataset.

Results:

p-Creode was able to reproducibly generate the correct cell-state transition map on hematopoiesis, compared to prior knowledge of hematopoietic cell differentiation. Our algorithm demonstrated dramatically improved robustness compared to a well-known algorithm in the field, as quantitatively determined by a statistical scoring metric. Application of p-Creode on single-cell intestinal data revealed that Tuft cells fall outside of the secretory specification lineage that is regulated by the transcription factor Atoh1, but their specification and behavior differ between the colon and small intestine. Follow-up experiments validated our computational results.

Conclusions:

We have developed and validated p-Creode, a novel and robust algorithm for interrogating single-cell data with respect to cell-state transitions. This algorithm presents significant improvement in reproducibility compared to previous methods. Application of p-Creode to intestinal cell specification revealed novel insights into Tuft cell specification and the heterogeneity of this process between the small intestine and colon.

Related Topics

    loading  Loading Related Articles