Which human cell types can SARS-CoV-2 actually infect?
COVID-19 presented with respiratory failure but also with cardiac, metabolic, gastrointestinal, ocular and neurological manifestations, and outcomes were linked to diabetes. Whether those manifestations reflect infection of the corresponding tissues was unresolved, and the models available could not settle it. Most laboratory work used African green monkey cells, human cancer lines carrying tumour-associated mutations and in some cases defects in innate sensing, or mice engineered to express human ACE2, none of which represents the diversity of human cell types the virus might reach.
The question also has a viral side. Where a cell is permissive, the efficiency with which the virus enters it is a property of Spike, and by mid-2020 a Spike variant carrying glycine rather than aspartate at position 614 had become globally dominant. Because that change sits in linkage disequilibrium with an ORF1b P314L variant in the same clade, population genetics could not attribute any phenotype to Spike alone.
Yang 2020 carries collaborative and was led by the Chen, Evans and Schwartz groups. The tenOever laboratory contribution was the authentic SARS-CoV-2 infection work and the transcriptional comparison against human lung autopsy material, performed by Nilsson-Payant, Hoagland and tenOever. Eight derivatives of human pluripotent stem cells spanning all three germ layers were stained for ACE2 and screened for entry, and the positives were challenged with authentic virus. 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 protein in several of them. Adult primary human islets and adult hepatocyte and cholangiocyte organoids reproduced the pancreatic and hepatic results. Infected endocrine cells and liver organoids induced a chemokine-dominated program while tissue-identity metabolic programs fell, and the chemokine signature was compared against the COVID-19 lung autopsy dataset from Blanco-Melo 2020.
Eriksen 2021 carries co-led, with tenOever and Blenkinsop joint on conceptualization, supervision and correspondence. Ocular surface tissue from three post-mortem donors who had tested positive carried spike protein in the epithelium, co-localizing extensively with ACE2. Cells cultured from six regions of healthy donor globes showed spike protein in cornea, limbus, sclera and retinal pigment epithelium but not iris or choroid, with ACE2 and TMPRSS2 highest in the limbus, where viral reads in genetically matched material were also highest. A serine protease inhibitor nearly abolished infection of limbal cells. Stem cell derived whole-eye cultures supported productive replication, and single-cell profiling separated what a cell does because it contains virus from what it does because its neighbours do, with NF-kappa B and TNF-associated genes highest in the most heavily infected limbal cluster and interferon-stimulated genes raised across the culture but comparatively lower in that same cluster.
Daniloski 2021 in eLife carries co-led with the Sanjana laboratory, the tenOever contribution being the replication-competent virus work. The D614G substitution was built into a cloned Spike so that the only difference between two arms was a single codon. Pseudotyped lentiviral particles bearing G614 transduced more cells at every dose in all four human lines tested, and a trans-complementation assay, in which ACE2 and one Spike variant are supplied to otherwise poorly infectable cells, reproduced the increase with replication-competent virus. Bio-layer interferometry found comparable binding of the S1 subunit to ACE2 for both variants, while G614 Spike was more resistant to proteolytic cleavage in cells and on particles without a difference in incorporation.
Three lines converge on a single negative result and one positive method. The negative result is that receptor expression does not predict permissiveness. Yang 2020 found several ACE2-positive lineages non-permissive and read this as pointing to additional host factors, naming TMPRSS2 as an example without testing it. Eriksen 2021 arrived at the same conclusion by a different route, with permissiveness in the eye tracking a serine protease profile and TMPRSS4 elevated roughly elevenfold in limbal over corneal cells, offered as a candidate explanation and not tested by loss of function. The positive method is the trans-complementation assay in Daniloski 2021, which allows a Spike variant to be tested against replication-competent virus without a reverse genetics system.
The evidentiary weight across these studies is uneven and the pseudotyped reporter is the recurring caution. In Yang 2020 most permissiveness calls at the screening stage rest on a replication-incompetent vesicular stomatitis virus particle bearing Spike, which reports entry only in a non-coronavirus context, and the in vivo kidney capsule xenograft used that reporter rather than authentic virus in an immunodeficient host, so entry in vivo is established and replication in vivo is not. Infections were at a single 24 hour point at low multiplicity, derivatives are developmentally immature, the platform has no immune component, and the response comparison uses lung autopsy tissue as the reference for pancreatic and hepatic models because pancreatic patient samples were unobtainable. In Eriksen 2021 the three infected post-mortem donors cannot establish whether the eye was the portal of entry or was seeded from an airway infection, the whole-eye cultures are immature and the infected and comparison cultures were differentiated for different lengths of time, and the protease requirement rests on a broad inhibitor that does not identify which protease matters. In Daniloski 2021 all work is in transformed lines, two of which overexpress ACE2 and in which the largest effects were seen, binding was measured on S1 monomers rather than trimers, cleavage was measured on transfected cells and lentiviral particles rather than authentic virions with two isogenic virus studies from other groups finding no such difference, and nothing addresses transmissibility or severity.
The host side of the same entry step belongs to host factors and druggable signaling, where the genome-scale screen in Daniloski 2021 in Cell found endosomal machinery required and RAB7A loss reducing surface ACE2. Whether receptor availability and Spike cleavage state interact is a question the two Daniloski papers raise together and neither answers. The response signatures reported in both tissue studies are read against the imbalanced host response, with Blanco-Melo 2020 supplying the standard. Yang 2020 also belongs to models for pandemic virology, since its panel is a model-building contribution as much as a tropism result, and the observation in Eriksen 2021 that interferon-stimulated genes rise in uninfected neighbours anticipates the non-cell-autonomous effects pursued under post-acute sequelae by Zazhytska 2022.
Human ocular surface tissue carries SARS-CoV-2 entry machinery and supports productive replication, with the limbus most permissive in both adult donor cells and stem cell derived whole-eye cultures, where infection drives NF-kB chemokine induction and blunted interferon signaling.
Isolating the SARS-CoV-2 Spike D614G substitution from the ORF1b P314L variant it travels with, through pseudotyped lentiviral particles and a trans-complementation assay with replication-competent virus, shows that the single Spike change raises entry efficiency across human lung, liver and colon cell lines without altering S1 affinity for ACE2.
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.