A Human Pluripotent Stem Cell-based Platform to Study SARS-CoV-2 Tropism and Model Virus Infection in Human Cells and Organoids
A panel of eight human pluripotent stem cell derivatives spanning all three germ layers, together with adult primary islets and liver organoids, identifies pancreatic alpha and beta cells, hepatocytes, cholangiocytes, cardiomyocytes and dopaminergic neurons as permissive to SARS-CoV-2 and shows that permissiveness does not track ACE2 expression alone.
- Senior authors
- Shuibing Chen
- Correspondence
- Fong Cheng Pan; David D. Ho; Benjamin R. tenOever; Todd Evans; Robert E. Schwartz; Shuibing Chen
Research areas & themes
Citation
Yang L, Han Y, Nilsson-Payant BE, Gupta V, Wang P, Duan X, Tang X, Zhu J, Zhao Z, Jaffré F, Zhang T, Kim TW, Harschnitz O, Redmond D, Houghton S, Liu C, Naji A, Ciceri G, Guttikonda S, Bram Y, Nguyen DHT, Cioffi M, Chandar V, Hoagland DA, Huang Y, Xiang J, Wang H, Lyden D, Borczuk A, Chen HJ, Studer L, Pan FC, Ho DD, tenOever BR, Evans T, Schwartz RE, Chen S. A Human Pluripotent Stem Cell-based Platform to Study SARS-CoV-2 Tropism and Model Virus Infection in Human Cells and Organoids. Cell Stem Cell. 2020. Volume 27, issue 1, pages 125-136.e7. DOI 10.1016/j.stem.2020.06.015. PMID 32579880. PMCID PMC7303620.
One-sentence contribution
A panel of eight human pluripotent stem cell derivatives spanning all three germ layers, together with adult primary islets and liver organoids, identifies pancreatic alpha and beta cells, hepatocytes, cholangiocytes, cardiomyocytes and dopaminergic neurons as permissive to SARS-CoV-2 and shows that permissiveness does not track ACE2 expression alone.
Executive summary
COVID-19 presents with respiratory failure but also with cardiac, metabolic, gastrointestinal and neurological manifestations, and clinical reports linked poor outcomes to diabetes. At the time of the study most laboratory work used African green monkey Vero cells, human cancer lines carrying tumor-associated mutations and in some cases defects in innate immune sensing, or mice engineered to express human ACE2, none of which represent the diversity of human cell types the virus might reach.
The authors built a platform by directed differentiation of human pluripotent stem cells into eight cell types and organoids representing endoderm, mesoderm and ectoderm, stained each for ACE2, and measured entry with a vesicular stomatitis virus particle pseudotyped with SARS-CoV-2 Spike. Cell types scoring positive were then tested with authentic SARS-CoV-2, and the key findings were checked in adult primary human islets, in adult hepatocyte and cholangiocyte organoids, and in a xenograft of stem cell derived pancreatic endocrine cells under the kidney capsule of immunodeficient mice.
Pancreatic alpha and beta cells, liver organoids, cardiomyocytes and dopaminergic neurons supported entry and authentic virus replication, while endothelial cells, macrophages, microglia and cortical neurons did not, despite ACE2 protein being detectable in several of the non-permissive types. Transcriptional profiling of infected endocrine cells and liver organoids showed strong chemokine induction alongside loss of tissue-specific metabolic programs, with a chemokine signature comparable to that in COVID-19 lung autopsy tissue.
Scientific context
Respiratory failure dominates severe COVID-19, but retrospective series had reported in-hospital cardiac injury in nearly 20 percent of patients, gastrointestinal manifestations in about 25 percent, and an association between poor outcomes and type 2 diabetes, with suggestions that infection can precipitate new-onset diabetes. ACE2 had been identified as the entry receptor and TMPRSS2 as a priming protease. The paper states the gap in terms of models. Vero cells are not human, the human lines in wide use are cancer derived, carry mutations such as in P53 that has been reported to regulate SARS coronavirus replication, include lines such as Huh7.5 with a known RIG-I defect that would obscure antiviral responses, and are proliferative and often unpolarized. Human pluripotent stem cell derivatives had already been used to study Zika virus neuropathology and to screen antivirals in work cited from several groups, which supplied the precedent for applying the same approach to SARS-CoV-2.
Central question
Which human cell types are permissive to SARS-CoV-2, how well does permissiveness correspond to ACE2 expression, and can human pluripotent stem cell derivatives and adult organoids serve as physiologically relevant models of infection and of the cellular response to it?
Experimental strategy
The design is a screen followed by orthogonal validation. Directed differentiation generates eight derivatives from one starting material, which keeps genetic background constant across cell types and avoids the confounds of comparing unrelated cancer lines. Entry is measured first with a replication-incompetent vesicular stomatitis virus particle bearing SARS-CoV-2 Spike and encoding luciferase, which reads out receptor engagement and membrane fusion alone and can be handled without high containment, with both a plate-level luciferase measurement and single-cell imaging so that permissiveness can be assigned to marker-defined cell types within a mixed culture.
Cell types positive in that screen were then challenged with authentic SARS-CoV-2 in the tenOever laboratory, with subgenomic N RNA as a replication readout that distinguishes replication from input genome, plus Spike protein staining. Because a criticism of stem cell derivatives is developmental immaturity, the pancreatic and hepatic results were repeated in adult primary human islets and in organoids derived from adult liver, and pancreatic endocrine cells were transplanted under the kidney capsule to test entry in vivo. Single-cell RNA sequencing of adult islets provides an independent, unbiased map of ACE2 and TMPRSS2 across islet cell types. Finally, transcriptional responses were compared against lung autopsy material from COVID-19 patients and healthy donors to ask whether the in vitro response resembles the human disease.
Key findings
-
ACE2 protein was detected in stem cell derived alpha and beta cells but not delta cells, in albumin positive hepatocytes, in endothelial cells, cardiomyocytes, microglia, macrophages and dopaminergic neurons, and only at low levels in cortical neurons, and was not detected in undifferentiated stem cells (Figure 1B).
-
Pseudo-entry virus produced high luciferase in pancreatic endocrine cells, liver organoids, cardiomyocytes and dopaminergic neurons, and low or absent signal in endothelial cells, microglia, macrophages and cortical neurons, with the same pattern at 48 hours as at 24 hours (Figure 1C and Supplementary Figure 2E). Within mixed cultures, luciferase localized to alpha and beta cells but not delta cells, to albumin positive hepatocytes, to cardiomyocytes and to dopaminergic neurons (Figure 1D to 1L).
-
Single-cell RNA sequencing of adult human islets resolved nine cell types and placed ACE2 and TMPRSS2 expression in acinar, ductal, beta, alpha, mesenchymal and endothelial cells (Figure 2A to 2D), with ACE2 protein in primary beta and alpha cells confirmed by imaging (Figure 2E).
-
Authentic SARS-CoV-2 infected primary human islets, with Spike protein detected in both insulin positive and glucagon positive cells at 24 hours (Figure 2F).
-
Stem cell derived pancreatic endocrine cells transplanted under the kidney capsule retained ACE2 expression after two months and took up pseudo-entry virus in vivo, with luciferase detected in insulin positive and glucagon positive cells of the xenograft (Figure 2G to 2J). This establishes entry in vivo. It does not establish authentic virus replication in vivo, which was not tested.
-
Authentic SARS-CoV-2 replicated in stem cell derived pancreatic endocrine cells in a dose-dependent manner by subgenomic N RNA, with Spike protein in both endocrine subtypes and read coverage across the viral genome (Figure 3A to 3D).
-
Infected endocrine cells separated from mock by principal component analysis, with enrichment of viral infection pathways and of the insulin resistance pathway, and downregulation of calcium signaling, glucagon signaling and metabolic pathways (Figure 3E and 3F). Caspase 3 positive fractions rose in both alpha and beta cells after infection, and the authors interpret the loss of identity-associated programs as arising mainly from increased apoptosis rather than from dedifferentiation. That attribution is an interpretation supported by the caspase staining and apoptosis gene signature and is not established against the alternative.
-
Chemokines including CCL2, CXCL5 and CXCL6 were upregulated in COVID-19 lung autopsy tissue relative to healthy donor lung, and comparable chemokine and cytokine induction was seen in infected endocrine cells (Figure 3G and 3H). The resemblance is a correlation between two systems and is presented as such.
-
Adult primary hepatocyte organoids and cholangiocyte organoids supported both pseudo-entry virus and authentic SARS-CoV-2, with high subgenomic N RNA, Spike protein in a significant fraction of cells, and genome-wide read coverage (Figure 4A to 4H and Supplementary Figure 4).
-
Infected hepatocyte organoids induced CXCL1, CXCL3, CXCL5, CXCL6 and CCL20 and downregulated hepatocyte metabolic markers CYP7A1, CYP2A6, CYP1A2 and CYP2D6, with enrichment of cytokine-cytokine receptor interaction, IL-17, chemokine, TNF and NF-kappa B signaling (Figure 4I to 4K). Cholangiocyte organoids induced CXCL1, CXCL2, CXCL3 and CCL2 with similar pathway enrichment (Figure 4L to 4N).
-
Authentic virus subgenomic RNA was high in cardiomyocytes and dopaminergic neurons and low or absent in cortical neurons, microglia and macrophages, matching the pseudo-entry virus results (Supplementary Figure 4I to 4M).
-
Several ACE2-positive cell types, including endothelium, macrophages and cortical neurons, showed little or no permissiveness. The authors read this as evidence that factors beyond ACE2, such as TMPRSS2, govern entry. The non-correspondence is the observation. The specific role of TMPRSS2 in these cell types was not tested here.
Mechanistic model
The study does not establish a mechanism for cell-type permissiveness and does not claim to. It establishes a map. Entry-competent cell types were identified with a Spike-pseudotyped particle and confirmed with authentic virus, and the central negative observation is that ACE2 protein detection does not predict permissiveness, which the authors take as pointing to additional required host factors, naming TMPRSS2 as an example. No host factor was manipulated, so the requirement is inferred rather than demonstrated.
For the cellular response, what the data support is that infection of pancreatic endocrine cells and of hepatic organoids induces a chemokine-dominated program while tissue-identity metabolic programs fall, and that apoptosis increases in infected endocrine cultures. Whether infection of beta cells contributes to the clinical association between COVID-19 and diabetes is raised as a question by the authors and is not addressed by these experiments. Whether the cell types identified here are major sites of infection in patients is explicitly left open in the paper's own limitations statement.
Conceptual or technical advance
The work supplies a standing panel of genetically matched, non-transformed human cell types and organoids in which tropism questions can be asked side by side, and it demonstrates the value of that design by producing a result a cancer line panel could not, namely a systematic mismatch between receptor expression and permissiveness across lineages. It also makes pancreatic endocrine cells and hepatic and biliary organoids available as infection models, and adds a xenograft configuration in which human endocrine tissue can be challenged in an animal. The authors propose the platform for antiviral drug screening, which is a stated prospect.
Relationship to the broader research program
The tenOever laboratory contribution here is the authentic SARS-CoV-2 infection work and the transcriptional comparison to human lung autopsy material, performed by Nilsson-Payant, Hoagland and tenOever according to the author contributions statement. The COVID-19 lung autopsy dataset and the chemokine-dominated response signature come from the laboratory's own work published as Blanco-Melo and colleagues 2020, which is cited for that comparison. Category 3 synthesis, visible when this paper is set beside that one and the laboratory's other 2020 SARS-CoV-2 work, is that the laboratory supplied a consistent transcriptional benchmark, an induced chemokine program with a muted interferon response, against which many collaborating groups calibrated their own model systems during that period.
Related publications
- Blanco-Melo and colleagues 2020. Methodological foundation and companion. Source of the COVID-19 lung autopsy transcriptional data and of the response signature used for comparison here.
- Han and colleagues 2020, on candidate COVID-19 therapeutics using stem cell derived lung organoids. Companion. Overlapping author group applying the same platform logic to the airway, cited here for consistency of the inflammatory signature.
- Zhao and colleagues 2020, on SARS-CoV-2 infection of liver ductal organoids. Companion. Independent report from another group consistent with the biliary findings here.
Limitations and boundaries
The authors state their own boundaries. Whether any of these cell types are major targets in patients cannot be settled without analysis of primary patient material, and the work focuses on entry and to some extent replication, leaving viral release and secondary spread across lineages unexamined. The platform is simplified relative to intact organ systems and contains no immune component. Beyond that, most permissiveness calls at the screening stage rest on a pseudotyped particle that reports entry only, in a vesicular stomatitis virus context rather than a coronavirus one. Infections were analyzed at a single time point of 24 hours at low multiplicity, with n equal to three biological replicates for the quantitative comparisons. Stem cell derivatives are developmentally immature, which the authors mitigate for the pancreas and liver by using adult primary material but not for cardiomyocytes, neurons or the myeloid lineages. The in vivo xenograft experiment used pseudo-entry virus, not authentic SARS-CoV-2, and an immunodeficient host. The comparison to COVID-19 patients uses lung autopsy tissue as the reference for responses measured in pancreas and liver models, since pancreatic samples from patients were not obtainable. The assignment of downregulated identity programs to apoptosis rather than to loss of cell identity rests on caspase 3 staining and a gene signature. One authentic virus isolate, USA-WA1/2020, was used throughout.
Audience summaries
25 words
Stem cell derived human tissues and adult organoids show that SARS-CoV-2 infects pancreatic insulin and glucagon cells, liver cells, heart muscle and dopaminergic neurons.
75 words
Laboratory study of SARS-CoV-2 largely relied on monkey and cancer cell lines. Differentiating human pluripotent stem cells into eight cell types and organoids allowed side-by-side comparison of which human tissues the virus can enter. Pancreatic alpha and beta cells, liver organoids, cardiomyocytes and dopaminergic neurons were permissive, while endothelial cells, macrophages, microglia and cortical neurons were not, even where the ACE2 receptor was present. Infected cells produced strong chemokine responses and lost tissue-specific metabolic gene expression.
150 words
COVID-19 involves organs beyond the lung, and available infection models were monkey cells, mutation-bearing human cancer lines or ACE2 transgenic mice. Eight derivatives of human pluripotent stem cells spanning all three germ layers were stained for ACE2 and challenged with a Spike-pseudotyped vesicular stomatitis virus reporter, then with authentic SARS-CoV-2. Pancreatic alpha and beta cells, liver organoids, cardiomyocytes and dopaminergic neurons supported entry and replication, while endothelial cells, macrophages, microglia and cortical neurons did not, despite detectable ACE2 in several of them, indicating that additional host factors govern entry. Adult primary human islets and adult hepatocyte and cholangiocyte organoids confirmed the pancreatic and hepatic results, and endocrine cells grafted under the mouse kidney capsule took up pseudo-entry virus in vivo. Infected endocrine cells and liver organoids induced chemokines resembling those in COVID-19 lung autopsy tissue, while losing hormone signaling and cytochrome P450 metabolic gene expression.
Documented publication relationships
- Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19 — methodological foundation.