The lungs and selected extrapulmonary organs (heart, kidney, liver, and spleen) of 13 deceased COVID-19 patients were biopsied postmortem under computed tomography (CT) guidance at the Jessa Hospital, Hasselt, Belgium during the period of April 15th-June 30th 2020. Additionally, plasma was collected from aortic blood and fractionated by ultracentrifugation. Patient demographics and clinical information are summarized in Table S1.
Viral loads in the lungs and RNAemia are higher when patients succumb rapidly to infection
Viral RNA was detected in the lungs and plasma of 9/13 and 8/13 patients, respectively. Cases were then stratified based on duration of disease (short <20 days; long >20 days). Patients who succumbed within 20 days following onset of symptoms generally had a higher viral RNA load in the lungs and plasma when compared to those that had lived longer (P=0.002; F=12.589; df=23; Fig. 1A). Our results support two phases of fatal disease evolution, including (i) short-lived disease with high viral loads in lungs and plasma, associated with a histological pattern of acute exudative alveolar damage in the lungs, and (ii) long-lived disease with low (or undetectable) viral loads in lungs and plasma, associated with a chronic pattern of lung injury (Fig. 1A and B, Fig. S1). These results agree with findings of another autopsy study9. Similarly, intra- and extracellular presence of SARS-CoV-2 nucleocapsid protein (NP) was more frequently identified in the lungs of cases with short-lived disease (4 out of 5), compared to those with long-lived disease (1 out of 8). Over all cases, bronchial epithelial cells and alveolar epithelial cells were the dominant cell type expressing SARS-CoV-2 NP, but we also identified viral NP in alveolar macrophages (Fig. 1B), as described by others8,9,11. Patient 13, who was under rituximab treatment for B cell lymphoma at the time of infection, did not follow the trend and had an exceptionally high viral RNA load in the lungs (106.5 copies/40 ng RNA) and plasma (106 copies/mL) after 88 days of disease. SARS-CoV-2 NP was ubiquitously and abundantly found intracellularly and extracellularly in hyaline membranes in the latter patient’s lungs, which showed a “remodeling pattern” with interstitial fibrosis and consolidation of airspace (Fig. 1B).
In immunocompromised patients replication-competent SARS-CoV-2 spreads systemically and disseminates to extrapulmonary organs
The frequent detection of SARS-CoV-2 RNAemia (9/13) in this cohort indicates that systemic dissemination of viral components is quite common in severe COVID-19 cases, as described previously12. However, viremia (i.e., circulation of infectious virions) was found only in one case (patient 13). In this case, VeroE6 cells showed cytopathogenic effects upon inoculation with the plasma pellet, but not after incubation with the other 8 RNAemic plasma pellets (Fig. S2A). Further, intact virions were present in patient 13’s plasma pellet, as observed with transmission electron microscopy (Fig. 1C and S2C). These results do not exclude the possibility of viremia in other cases, as current isolation methods might not suffice to isolate virions when viral RNA loads in plasma are below a certain threshold.
Next, we identified two distinct types of disease progression based on viral RNA spread to extrapulmonary organs (i.e. heart, kidney, liver, and/or spleen), with intra-organ spread only occurring in 3 out of 13 cases (Fig. 1D). Digital droplet PCR was run on all biopsies of these three patients to quantify absolute SARS-CoV-2 copy numbers and confirmed spread of viral RNA to multiple organs (Fig. S2B). Viral dissemination to multiple organs was strongly associated with profound immune suppression (chronic high dose corticosteroid and/or rituximab treatment) at the time of infection (Fisher exact P=0.014; Table S1). We hypothesize that inadequate immune responses during the early phase of SARS-CoV-2 infection resulted in enhanced viral replication and spread to extrapulmonary organs. Chronic high dose corticosteroid treatment dampens viral-induced danger signals of the host immune response, resulting in impaired release of critical antiviral components (e.g., interferons)13,14. Second, rituximab induces lysis and apoptosis of normal and malignant human B lymphocytes, essential for the production of virus-specific antibodies15. These findings point out the importance of patient management in severely immunocompromised COVID-19 patients.
Positive SARS-CoV-2 nucleocapsid (NP) staining was found in all organs of case 13, in renal and splenic tissue of case 07 and in splenic tissue of case 06. Viral NP was observed in cardiomyocytes and interstitial cells (heart), podocytes and tubular epithelial cells (kidney), hepatocytes and Kupffer cells (liver) and myeloid cells (spleen) (Fig. 1E). Viral RNA and proteins have been observed multiple times in myeloid cells, tubular cells and podocytes, but this is the first unambiguous evidence of hepatocytes and cardiomyocytes being in vivo SARS-CoV-2 targets8,16,17. Interestingly, SARS-CoV-2 NP was detected in cell types expressing both the SARS-CoV-2 main receptor and co-receptor (i.e. angiotensin-converting enzyme type 2 [ACE2] and transmembrane serine protease 2 [TMPRSS2], respectively) across all organs examined (Fig. S3), confirming the in vivo relevance of ACE2 and TMPRSS2 in SARS-CoV-2 cell infection.
Further, we isolated virus from extrapulmonary organs that could replicate on VeroE6 cells. Infectious SARS-CoV-2 was isolated from the heart and kidney of case 07 and from all organs of case 13. These progeny viruses were subjected to full-length sequencing to confirm SARS-CoV-2 presence. The fact that we were unable to isolate infectious virus from SARS-CoV-2 RNA- and NP-positive splenic tissue in two out of three cases (case 06 and 07) might indicate that the signal in these tissues derived from phagocytosed virions (and thus, viral RNA and proteins) in immune cells, rather than active viral replication in splenic cells. Viral loads in cardiac tissue of case 06 likely were too low for successful virus isolation. Of note, presence of SARS-CoV-2 in extrapulmonary organs was rarely associated with pathological alterations in the respective organs, except for local cytolysis of cardiomyocytes in the heart. This cytolysis was likely induced by viral replication, and splenic lymphocyte depletion in case 13, which most likely was the result of the rituximab treatment (Fig. 1E, upper right HE images).
SARS-CoV-2 nucleocapsid protein predominates in epithelial cells as well as cells from the myeloid lineage
Although immunohistochemistry analysis can, to a limited extent, identify SARS-CoV-2 target cells, it does not allow marker co-localization at the cellular level. To identify SARS-CoV-2 tissue-specific target cell types in cases with intra-organ viral RNA dissemination (case 06, 07 and 13) we used double immunofluorescence staining and confocal microscopy. As shown in Fig. 2, the majority of SARS-CoV-2 NP-positive cells resided in the lungs (48.73% on a total of 516 positive cells), followed by the kidneys (29.69%), the spleen (10.98%), the liver (9.80%), and the heart (1.68%). SARS-CoV-2 NP was predominantly found in cytokeratin-positive (epithelial) cells in the lungs (67.90%), liver (66.63%), and kidneys (90.90%), while it was more commonly observed in CD14-positive (myeloid) cells in the heart (50.00%) and spleen (57.63%). ICAM-positive (endothelial) cells expressing SARS-CoV-2 NP occasionally were detected in the lungs (7.14%), kidney (3.00%), and spleen (10.02%). In general, SARS-CoV-2 NP was found only in ACE2-positive cells.
Organ-specific SARS-CoV-2 evolution in an immunocompromised patient
We hypothesized that SARS-CoV-2 replication in multiple anatomical compartments would result in the emergence of specific variants in distinct organs, as described for other RNA viruses including poliovirus and HIV18,19. A recent study showed SARS-CoV-2 sequence diversity between respiratory and gastro-intestinal tract swabs from three COVID-19 patients20. In general, the SARS-CoV-2 genome changes at a steady mutational rate of 0.0008 substitutions per site per year. As a result, acute respiratory viral infections have low intra-host diversity21,22. However, there is compelling evidence that SARS-CoV-2 evolution is accelerated in the respiratory tract of persistently infected immunocompromised hosts, reflecting reduced selective immune pressure6,23–26. Therefore, we compared viral genome sequences from different organs in a patient with profound systemic and intra-organ viral spread (case 13). The complete clinical history and disease course in case 13 is summarized in Fig. S4.
Phylogenetic analysis confirmed that all SARS-CoV-2 genomes isolated from distinct anatomical compartments of case 13 descended from a common ancestor derived from clade 20B (Fig. 3A). The data show that different populations of viral genomes were found in multiple organs. Consensus viral genomes retrieved from plasma and spleen were phylogenetically most closely related to the founder virus, followed by those from lungs, heart, liver, and kidneys. A more detailed comparative analysis identified 50 (sub-)consensus single nucleotide variations (SNVs) (18 synonymous and 32 non-synonymous mutations), 1 small and 4 large deletions in viral genomes derived from different organs or plasma, as compared to the clade 20B consensus genome (Fig. 3B, Table S2 and S3). These mutations were distributed in the 5’ and 3’ UTR and across 7 out of 10 protein-coding genes, including ORF1ab, S, E, ORF7a, ORF8, N, and ORF10. Three SNVs were fixed in all variants isolated from different compartments (frequencies >94%) and were therefore most likely present in the founder virus. In contrast, all other SNVs and deletions were detected at variable frequencies ranking between 1.11% and 98% depending on tissue origin, illustrating within-host organ-specific evolution of SARS-CoV-2.
Interestingly, several organs harbored viral populations distinct from all other compartments. For instance, four additional SNVs (T7247G, C7279T, and A8387G in ORF1a, and A27574T in ORF7a) were present at frequencies above 80% in the kidneys (Fig. 3B and Table S2). In addition, six SNVs distributed across ORF1ab (A13433G, C16092T, T18024C, T18750C, C18979T) and ORF10 (C29592T) were almost uniquely retrieved from kidneys. Six out of these ten SNVs were non-synonymous inducing amino acid substitutions in viral proteins including NSP3, NSP14, ORF7a, and ORF10. Still, viral infection capacity was not reduced by the majority of these SNVs (5/6), as these mutations were also identified in the viral progeny of VeroE6 cells inoculated with kidney-derived viruses (Table S2, in bold). NSP3, NSP14 and ORF7a are involved in viral protein processing, viral release, genome replication and immune evasion, while the in vivo role of ORF10 is still under debate27–30. Variation in these proteins likely arose during extensive viral replication and spread in the kidneys, as evidenced by the large number of SARS-CoV-2 NP-positive cells in the kidney, and may have favored infection of the kidney following bottleneck events and viral adaptation to local environments. Besides multiple SNVs, the dominant SARS-CoV-2 genotype in renal tissue displayed a 91 bp deletion in ORF8 and a 1927 bp deletion in the N-terminal tail of S protein comprising the receptor-binding domain (RBD) (Fig. S5 and Table S3). The latter deletion in S, as well as a 1947 bp deletion in the same S region was found in almost one quarter of the splenic viral population. Further, a 422 bp deletion in S a few base pairs upstream of RBD was observed in the majority of liver-derived viral genomes. These deletion mutations did not grow on VeroE6 cells, questioning their in vivo infectivity. Still, how these deletion mutants accumulate in multiple organs remains to be elucidated. In this context, similar variants with deletions in ORF8, but not S, have been detected in patients from different countries and have been associated with milder infection31–33. Interestingly, other coronaviruses have been shown to shift tissue tropism due to deletions in S protein34. Alternatively, it is possible that SARS-CoV-2 defective genomes might modulate viral replication or serve as immune decoys, thereby promoting viral persistence, as described for other RNA viruses35. We speculate that S deletion mutants may be involved in viral occupation of the kidney, spleen and liver, but not in viral propagation in VeroE6 cells.
Besides the above-described deletions in S protein, the viral population in the spleen was characterized by several unique SNVs that did not affect viral infectivity. For instance, we identified mutations in ORF1ab (C12513T [T4083M amino acid substitution in NSP8] and C14937T [no amino acid substitution in RNA-dependent RNA polymerase]), E (C26351T [A36V amino acid substitution in E protein]) and 3’ UTR (G29744A) with frequencies ranging between 33.61 and 64.31% (Table S2). Specific alterations in NSP8 and E protein may favor viral infection or propagation in splenic tissue, as these viral proteins are involved in viral replication and budding36.
Interestingly, in the liver, viral genomes with a 422 bp deletion in S consistently displayed the A23063T (N501Y amino acid substitution in S protein) alteration, a key mutation found in three important variant strains (South African B.1.351, UK B.1.1.7, and Brazil P1 strain) that promotes viral binding, infectivity and virulence4,37,38. In addition, this mutation is associated with adaptation to rodents37. The same SNV was also present in genomes derived from other organs, but in lower frequencies and independent from the 422 bp deletion. Still, this mutation remained present in viruses propagated in VeroE6 cells from different tissues, highlighting the infection capacity of mutant N501Y viruses.
Similar to the N501Y mutation, we also identified the C24642T (T1027I amino acid substitution) mutation in S, present in current strains of the B.1.1.248 lineage, at peaking concentrations of 50% in lungs and plasma, as well as in their viral offspring in VeroE6 cells4. In addition, a high SNV variability was detected in viral S genes derived from the lungs, which included mutation A22920T at a frequency of 52.52% leading to Y453F amino acid substitution in RBD of S protein. Remarkably, this mutation has been suggested to be the hallmark of the “mink variant”. It is believed to increase viral binding to mink ACE2, and presumably also human ACE210. However, since genomes with this SNV did not replicate in VeroE6 cells, binding and entry in monkey cells may be reduced. Our results suggest that viral evolution in the respiratory tract, but also in extrapulmonary organs of immunocompromised COVID-19 patients may prompt the emergence of more virulent and contagious SARS-CoV-2 variants with the capacity to infect other species. These results highlight the utmost importance of hygienic and preventive measures to avoid viral spread from and to immune suppressed patients.