Liver involvement in SARS-CoV-2 infection is an important clinical issue1,2. In general, severe liver injury and liver synthetic dysfunction is uncommon in patients with COVID-19 in the absence of underlying liver disease. However, we and others have shown that mortality in COVID-19 is significantly enhanced in those patients with advanced liver disease, with mortality rates of 90% in those with Child–Pugh C disease requiring respiratory support, and although the majority of deaths are due to pulmonary disease, liver decompensation is common during infection3. Furthermore, in people without underlying liver disease who contract COVID-19, abnormally high levels of liver enzymes are commonly observed and have been associated with an increase in subsequent intensive care unit admission, mechanical ventilation and death2,4,5. Whilst these observations may be confounded by enhanced liver monitoring and complications of systemic disease in sicker patients, deranged liver function tests at the time of hospital admission and temporal changes in these have also been linked to patient morbidity6. Together, these clinical observations have raised the important question as to whether SARS-CoV-2 directly infects the liver, causing biochemical derangements and poor clinical outcomes. Ultimately, a better understanding of multi-organ pathogenesis will give new biological insights into COVID-19 pathogenesis and may reveal new pathways for therapeutic targeting.

In this issue of Nature Metabolism, Wanner et al7. confirm previous observations that show that abnormal liver function tests are deranged in COVID-19 through the assessment of two large cohorts of patients hospitalized with COVID-196. They generated novel data through a detailed assessment of 42 liver autopsy specimens that were taken from patients who died from COVID-19 (Fig. 1). This is the first time that an analysis of this scale has been performed on liver autopsy specimens. Previously, the authors have shown that although the highest levels of SARS-CoV-2 are detected in the respiratory tract (by RT–PCR), lower levels may also be detected in the kidneys, liver, heart, brain and blood8,9. Viral RNA was detected in the kidney, and renal infection was associated with disease severity and death in the first 3 weeks of infection.

Fig. 1: Schematic diagram to show methods for detecting SARS-CoV-2 in autopsy liver samples.
figure 1

Evidence for SARS-CoV-2 replication and the intrahepatic host response through transcriptomic and proteomic analysis in patients who died with severe COVID-19. GI, gastrointestinal; sgRNA, single guide RNA.

The authors now show that SARS-Cov-2 may be detected using RT–PCR in 69% of autopsy liver specimen cases. However, using viral detection methods that extend beyond RT–PCR is critical, since SARS-CoV-2 RNA, in theory, may be contained within blood vessels supplying the organ of interest and would be detected by this method, and because SARS-CoV-2 detection is not proof of infection and replication. Convincingly, Wanner et al. show that SARS-CoV-2 RNA could also be identified in hepatic parenchymal cells directly, with high spatial resolution using in situ hybridization. A comparison between viral load in different organs showed higher median viral loads in lungs compared to liver though the range was similar. Owing to tissue degradation, the authors were not able to confirm the in situ hybridization data with electron microscopy, but in support of the direct infection hypothesis, SARS-CoV-2 spike protein and ACE2 expression using indirect immunofluorescence and confocal microscopy was observed in the livers. Most importantly, SARS-CoV-2 replication-competent virus was recovered from the liver in autopsy samples. Although not shown directly (as liver tissue contains multiple cell types), this replication data adds to the evidence that primary hepatocytes may support SARS-CoV-2 viral replication.

The functional consequences of SARS-CoV-2 liver tropism were also explored through an analysis of the proteomic and transcriptomic signatures in the autopsy liver specimens. A correlation of the signature with clinical outcomes was not possible, since all specimens were obtained in patients who died. Transcriptomic profiling confirmed the expression of known SARS-CoV-2 entry receptors and proteins that support infection including ACE2, TMPRSS2, CTSL and RAB7A. Intriguingly, a mismatch between expression of the ACE2 protein and location of the SARS-CoV-2 spike protein in the immunofluorescence experiments was observed, raising the possibility that other receptors (such as SR-B1, which has been shown to have a facilitatory role in SARS-CoV-2 viral entry with a wide distribution, including in the liver10) may have a role in liver tropism. A previous combined analysis of several single-cell RNA-sequencing datasets showed that the minority of hepatocytes co-express ACE-2 and TMPRS22, which are known to both be required for infection11. Together, this data suggests that further experiments using in vitro cellular systems and organoids expressing putative receptors may be required to fully elucidate extra pulmonary SARS-CoV-2 receptor pathways.

To explore the possible direct effects of SARS-CoV-2 infection, the autopsy liver samples from patients with COVID-19 who had virus detectable by RT–PCR in the liver were compared to those without detectable virus, and to control individuals without COVID-19. Transcriptomic analysis revealed a relative upregulation of type-I, -II and -III interferons, JAK/STAT and metabolic signalling in the RT–PCR-positive livers, which was confirmed by proteomic analysis. It was not possible to make a comparison of this signature to one that might be generated in the liver following acute infections with other hepatotropic or non-hepatotropic viruses, but a comparison of the gene expression signatures was made with liver specimens from patients with chronic hepatitis B virus (HBV), hepatitis C virus (HCV) and human immunodeficiency virus (HIV) infection. The transcriptomic signature in the livers from individuals with SARS-CoV-2 infection broadly overlapped with that found in these other infections and was most similar to livers from individuals with HCV infection. In this study, the authors found no evidence of a cellular inflammatory infiltrate or cytopathic changes in the COVID-19 livers, whereas a cellular infiltrate is commonly observed in chronic HCV infection, suggesting that different pathogenic mechanisms are likely to lead to a convergent transcriptomic profile that is dominated by interferon signalling. Clinical experience shows us that elevated liver biochemical derangement, a marker of hepatocyte damage and death, is very commonly observed in a wide range of clinical scenarios (including systemic inflammation, sepsis and after infection with pathogens that are not known to be hepatotropic) and this is thought to arise through collateral damage from cytokines and liver-infiltrating or activated immune cells12. This broad range of factors that have been associated with a liver inflammatory response highlight the fact that the drivers of the liver gene signatures in patients with critical illness must be interpreted with care. Nevertheless, the observation of a transcriptomic signature of interferon responsiveness, specifically in RT–PCR-positive liver, is evidence that SARS-CoV-2 infection may be directly enhancing intrahepatic innate immune responses.

The study of the liver as an immune organ remains an important and fascinating area of scientific enquiry in its own right. In many settings, the liver is thought to be immune tolerant, downregulating immune responses to gut endotoxins through direct exposure via the portal venous system. As a result, the liver has evolved features that are permissive to chronic hepatotropic infections like HBV and HCV, and may be exploited as a niche of replication by other pathogens such as Plasmodium falciparum. How SARS-Cov-2 enters the liver to infect hepatocytes is currently not known; infection directly from the blood stream seems unlikely, as infectious virus is generally not detected in the blood of patients with COVID-1913. Infection via the gut, a source of prolonged SARS-CoV-2 viral shedding during infection that communicates directly with the extensive portal venous system, seems more likely14. Recent intriguing data, generated using coxsackievirus as a model in mice, suggests that hepatocytes may function as a viral ‘filter’, removing viruses during infection at their own expense to protect the host against a variety of systemic infections15. The new data by Wanner et al. adds to a body of work that explores the role of hepatocytes in infection and represents a major step forward in understanding the interplay between SARS-CoV-2 and intrahepatic innate immunity, with a multiparametric analysis providing some of the first convincing evidence that SARS-CoV-2 may directly infect the liver in severe COVID-19.