Single cell RNA sequencing analysis did not predict hepatocyte infection by SARS-CoV-2.

To the Editor:

It was with great interest that we read the research article by Wang et al. In their manuscript, the presence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) particles in hepatocytes is shown with additional arguments for viral replication and cytopathy in infected hepatocytes, which provides a partial explanation for the high prevalence of abnormal liver laboratory tests in patients with COVID-19.

It has already been proposed that SARS-CoV-2 might directly infect liver cells. It certainly seemed that a risk stratification for direct liver cell infection could be predicted based on single cell transcriptomic data for the SARS-CoV-2 entry factors ACE2 (SARS-CoV-2 cellular entry) and TMPRSS2 (spike protein priming). Many researchers have analyzed single cell RNA sequencing (scRNASeq) datasets. The general conclusion was that the liver is a low-risk organ for SARS-CoV-2 infection, though cholangiocytes proved to be the cells with the highest ACE2 expression. , This led to the hypothesis that cholangiocytes are the most likely target of a direct SARS-CoV-2 infection in the liver.

We did note, however, that these scRNASeq analysis manuscripts analyzed only 1 liver dataset, did not focus on liver tissue specifically or reported only on ACE2 expression without considering TMPRSS2 expression. Additionally, since substantial clinical data on COVID-19 infection and chronic liver disease (CLD) is currently limited, we thought it would be informative to evaluate the potential vulnerability of individual cell types in patients with CLD. Enhanced ACE2 and TMPRSS2 expression in cirrhotic livers would leave these patients potentially more susceptible to liver infection with possible worse disease outcomes.

To address these issues, we analyzed 3 publicly available human liver datasets published by Aizarani et al., Macparland et al. and Ramachandran et al. The latter includes cirrhotic livers from patients undergoing orthotopic liver transplantation, caused by non-alcoholic liver disease, alcohol-related liver disease and primary biliary cholangitis. Using these 3 datasets, we verified ACE2 and TMPRSS2 expression in healthy and diseased human livers (Fig. 1 A-B).

Fig. 1

scRNAseq analysis of ACE2 and TMPRSS2 expression in livers.

(A) Schematic of liver cell types represented in the 3 publicly available scRNAseq data sets of human liver tissue. Cellular markers used for cluster identification are shown. (B) ACE2 and TMPRSS2 RNA (co)expression in liver cells. All raw cell counts were normalized, scaled and clustered using principle component analysis. Cell types were identified using the same markers as in the original papers (Fig. S1). Surface of dots represents percentage of cells with greater-than-zero RNA expression (per cell type). Color intensity represents expression value of genes.

Cholangiocytes are among the highest expressors of ACE2 in all datasets, which is in line with previous scRNAseq reports. However, only a low percentage of cholangiocytes express ACE2 RNA, except for the MacParland dataset (14.29% vs. 0.99 and 0.82 %). ACE2 expression in hepatocytes from the dataset by Ramachandran et al. shows a higher frequency of ACE2 positive cells (10.2%) compared to the other datasets. However, seeing as they are KRT7, EPCAM and ALB positive, these hepatocytes presumably represent cells undergoing a ductular reaction (Fig. S1-2). Hepatocytes from the datasets by Aizarani (0.73% ACE2+) and Macparland (0.26% ACE2+) do not show KRT7 or EPCAM expression. TMPRSS2 is expressed by a higher percentage of hepatocytes and cholangiocytes in all datasets, suggesting that this is not the limiting factor for cellular entry, as is also the case for other tissues. Other cell types such as immune, endothelial and mesenchymal cells express limited to no ACE2 or TMPRSS2. Analysis of double positive cells of healthy individuals shows that only 0.04% and 0.03% of the hepatocytes co-express ACE2 and TMPRSS in the Aizarani and Macparland studies. Respectively, 0.45% and 2.52% of cholangiocytes co-express ACE2 and TRMPRSS2 in the Ramachandran and Macparland studies (no co-expression in Aizarani study). Furthermore, when comparing cells from healthy and diseased livers, we do not see any increase in ACE2 expression nor in ACE2-TMPRSS2 co-expression in cholangiocytes.

In conclusion, scRNAseq analyis does not point towards hepatocytes as a likely point of entry for SARS-CoV-2 infection. The low expression of ACE2 seen in this data presumably represents technical limitations of the scRNAseq technique, rather than an absolute absence of ACE2 in these cells, leading to an underestimation of ACE2 expressing hepatocytes. Indeed, even in alveolar epithelial type II cells, the cell type playing a crucial role in SARS-CoV pathogenesis, ACE2 expression levels were reported to be low in single cell analysis. Interestingly, while the percentage of ACE-TMPRSS2 co-expressing hepatocytes is extremely low in the 2 datasets containing representative hepatocytes, , it is not zero. Since the human liver is estimated to contain tens of billions of hepatocytes, this very low percentage could still leave millions of hepatocytes at risk. However, this does not explain the absence of SARS-CoV-2 viral particles in cholangiocytes, which leaves the possibility of alternate cellular entry receptors or requirements for co-receptors, as hypothesized by Wang et al., to explain the seemingly hepatocyte-specific tropism of SARS-CoV-2 in the liver. Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITE-seq) using antibodies against ACE2 and TMPRSS2 could help to gain more insight into the identity of cell types at risk of SARS-Cov-2 infection. Finally, despite great insights into cellular identities across the entire human body in health and disease, the findings by Wang et al. highlight a need for caution when interpreting analyses of scRNAseq data for cell susceptibility to SARS-CoV-2 viral infection.

Financial support

Vincent De Smet: Fonds Wetenschappelijk Onderzoek (BE) 1192920N.

Leo A van Grunsven: Vrije Universiteit Brussel (BE) and Fonds Wetenschappelijk Onderzoek (BE)

Authors' contributions

Vincent De Smet: Conceptualization, Writing – original draft, Writing – review & editing. Stefaan Verhulst: Formal analysis, Methodology, Visualization, Contributed equally to first author. Leo A van Grunsven: Conceptualization, Funding acquisition, Supervision, Writing – review & editing.

Conflict of interest

The authors declare no conflicts of interest that pertain to this work.

Please refer to the accompanying ICMJE disclosure forms for further details.