Understanding the development and function of an organ requires the characterization of all of its cell types. Traditional methods for visualizing and isolating subpopulations of cells are based on messenger RNA or protein expression of only a few known marker genes. The unequivocal identification of a specific marker gene, however, poses a major challenge, particularly if this cell type is rare. Identifying rare cell types, such as stem cells, short-lived progenitors, cancer stem cells, or circulating tumour cells, is crucial to acquire a better understanding of normal or diseased tissue biology. To address this challenge we first sequenced the transcriptome of hundreds of randomly selected cells from mouse intestinal organoids, cultured self-organizing epithelial structures that contain all cell lineages of the mammalian intestine. Organoid buds, like intestinal crypts, harbour stem cells that continuously differentiate into a variety of cell types, occurring at widely different abundances. Since available computational methods can only resolve more abundant cell types, we developed RaceID, an algorithm for rare cell type identification in complex populations of single cells. We demonstrate that this algorithm can resolve cell types represented by only a single cell in a population of randomly sampled organoid cells. We use this algorithm to identify Reg4 as a novel marker for enteroendocrine cells, a rare population of hormone-producing intestinal cells. Next, we use Reg4 expression to enrich for these rare cells and investigate the heterogeneity within this population. RaceID confirmed the existence of known enteroendocrine lineages, and moreover discovered novel subtypes, which we subsequently validated in vivo. Having validated RaceID we then applied the algorithm to ex vivo-isolated Lgr5-positive stem cells and their direct progeny. We find that Lgr5-positive cells represent a homogenous abundant population of stem cells mixed with a rare population of Lgr5-positive secretory cells. We envision broad applicability of our method for discovering rare cell types and the corresponding marker genes in healthy and diseased organs.

Organoids are self‐organizing 3D structures grown from stem cells that recapitulate essential aspects of organ structure and function. Here, we describe a method to establish long‐term‐expanding human airway organoids from broncho‐alveolar resections or lavage material. The pseudostratified airway organoids consist of basal cells, functional multi‐ciliated cells, mucus‐producing secretory cells, and CC10‐secreting club cells. Airway organoids derived from cystic fibrosis (CF) patients allow assessment of CFTR function in an organoid swelling assay. Organoids established from lung cancer resections and metastasis biopsies retain tumor histopathology as well as cancer gene mutations and are amenable to drug screening. Respiratory syncytial virus (RSV) infection recapitulates central disease features, dramatically increases organoid cell motility via the non‐structural viral NS2 protein, and preferentially recruits neutrophils upon co‐culturing. We conclude that human airway organoids represent versatile models for the in vitro study of hereditary, malignant, and infectious pulmonary disease.

Lgr5 stem cells reside at small intestinal crypt bottoms, generating both the enterocyte and secretory lineage. Entry into the latter epithelial lineage requires silencing of Notch signaling. The Notch ligand Dll1 is strongly up-regulated in a small subset of immediate stem cell daughters. Lineage tracing utilizing a novel Dll1GFP-ires-CreERT2 knock-in mouse reveals that single Dll1high cells generate small, short-lived clones of all four secretory cell types. In culture, sorted Dll1high cells can form long-lived organoids when briefly exposed to Wnt3A. When Dll1 cells are genetically marked prior to tissue damage, significant numbers of stem cell tracing events occur. Lineage specification therefore occurs already in the earliest stem cell daughters through Notch lateral inhibition. Yet, specified secretory progenitors display plasticity and can regain stemness upon tissue damage.

SummaryAdult mitotic tissues like the intestine, skin, and blood undergo constant turnover throughout the life of an organism. Knowing the identity of the stem cell is crucial to understanding tissue homeostasis and its aberrations upon disease. Here we present a computational method for the derivation of a lineage tree from single-cell transcriptome data. By exploiting the tree topology and the transcriptome composition, we establish StemID, an algorithm for identifying stem cells among all detectable cell types within a population. We demonstrate that StemID recovers two known adult stem cell populations, Lgr5+ cells in the small intestine and hematopoietic stem cells in the bone marrow. We apply StemID to predict candidate multipotent cell populations in the human pancreas, a tissue with largely uncharacterized turnover dynamics. We hope that StemID will accelerate the search for novel stem cells by providing concrete markers for biological follow-up and validation.