Small case series have evaluated severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) detection in formalin-fixed, paraffin-embedded tissue using reverse transcription–polymerase chain reaction, immunohistochemistry (IHC), and/or RNA in situ hybridization (RNAish).
To compare droplet digital polymerase chain reaction, IHC, and RNAish to detect SARS-CoV-2 in formalin-fixed, paraffin-embedded tissue in a large series of lung specimens from coronavirus disease 2019 (COVID-19) patients.
Droplet digital polymerase chain reaction and RNAish used commercially available probes; IHC used clone 1A9. Twenty-six autopsies of COVID-19 patients with formalin-fixed, paraffin-embedded tissue blocks of 62 lung specimens, 22 heart specimens, 2 brain specimens, and 1 liver, and 1 umbilical cord were included. Control cases included 9 autopsy lungs from patients with other infections/inflammation and virus-infected tissue or cell lines.
Droplet digital polymerase chain reaction had the highest sensitivity for SARS-CoV-2 (96%) when compared with IHC (31%) and RNAish (36%). All 3 tests had a specificity of 100%. Agreement between droplet digital polymerase chain reaction and IHC or RNAish was fair (κ = 0.23 and κ = 0.35, respectively). Agreement between IHC and in situ hybridization was substantial (κ = 0.75). Interobserver reliability was almost perfect for IHC (κ = 0.91) and fair to moderate for RNAish (κ = 0.38–0.59). Lung tissues from patients who died earlier after onset of symptoms revealed higher copy numbers by droplet digital polymerase chain reaction (P = .03, Pearson correlation = −0.65) and were more likely to be positive by RNAish (P = .02) than lungs from patients who died later. We identified SARS-CoV-2 in hyaline membranes, in pneumocytes, and rarely in respiratory epithelium. Droplet digital polymerase chain reaction showed low copy numbers in 7 autopsy hearts from ProteoGenex Inc. All other extrapulmonary tissues were negative.
Droplet digital polymerase chain reaction was the most sensitive and highly specific test to identify SARS-CoV-2 in lung specimens from COVID-19 patients.
Coronavirus disease 2019 (COVID-19), the disease that is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, can be life-threatening, largely because of development of acute respiratory distress syndrome. Indeed, acute respiratory distress syndrome has been found in 3.4% of patients with COVID-19 and 72% to 93% of patients with COVID-19 who died.1–3 Furthermore, we and others have shown that acute bronchopneumonia, diffuse alveolar damage (DAD), aspiration pneumonia, and thromboemboli are common findings in the lungs of these patients at autopsy.4–9 Although evidence suggests that COVID-19 also impacts other organs such as heart,9 liver,9 and kidneys,9,10 the virus has so far been identified with certainty only in the lungs and rarely the brain, heart, liver, kidney, placenta, and blood by electron microscopy, immunohistochemistry (IHC), in situ hybridization (ISH), and reverse transcription–polymerase chain reaction (RT-PCR).8,9,11–18
Infection with SARS-CoV-2 is usually identified by the detection of viral RNA using RT-PCR on nasopharyngeal or oropharyngeal swabs. Current or prior infection might also be recognized by a serum test for SARS-CoV-2 immunoglobulin G (IgG) antibodies.19 However, global vaccination efforts will likely lead to declining SARS-CoV-2 infections and eventually to reduced clinical testing for SARS-CoV-2. To this end, it is expected that patients with unrecognized SARS-CoV-2 infection who present with unexplained acute or acute on chronic respiratory failure will likely undergo lung biopsy. Furthermore, this same subset of patients with unrecognized COVID-19 may have an autopsy performed. Although DAD, acute bronchopneumonia, and thromboemboli in the lung are recognized by morphology, the etiology of those findings can be manifold and no COVID-19–specific histopathologic findings have been identified. Furthermore, viral cytopathic effects and inclusions observed in cytomegalovirus, herpes simplex virus, and adenovirus infections are not evident in SARS-CoV-2 infections. Therefore, a formalin-fixed, paraffin-embedded (FFPE) tissue–based test is needed to establish the diagnosis of COVID-19 and guide management of the patient.
Various methods for the identification of SARS-CoV-2 infection in FFPE tissue specimens have been published, including ISH, IHC, RT-PCR, and electron microscopy.8,9,11–13 However, electron microscopy to identify SARS-CoV-2 virus particles in the lung might not be helpful or feasible in routine clinical practice. Studies of the other modalities have largely included limited numbers of cases and lacked clinically validated sensitivities and specificities. In addition, the temporal variation of SARS-CoV-2 virus levels within organ systems during infection may impact the ability of the various assays to confirm the virus. Potential regional or temporal heterogeneity in viral genomic sequence due to genetic selection or drift may play a similar role, impacting the ability of existing targeted PCR assays to detect the virus in specimens from patients with COVID-19. Because formalin fixation can impact the quality and integrity of nucleic acid, sensitivity for the detection of microorganisms by PCR or other techniques may also be limited. Alternatively, newer and potentially more sensitive techniques such as reverse transcription droplet digital polymerase chain reaction (ddPCR) might enhance detection of SARS-CoV-2 in FFPE tissue or other matrices. Recently, ddPCR has been shown to have a significantly lower limit of detection of SARS-CoV-2 in nasopharyngeal swabs than RT-PCR.20 Although we recently showed that ddPCR did not detect SARS-CoV-2 infection in heart tissue from COVID-19 patients,21 this technique has not been thoroughly validated in FFPE specimens of other organs, specifically the lung, and has not been compared with other techniques in its capability to identify the virus in FFPE tissue.
We developed and compared assays for the detection of SARS-CoV-2 in FFPE tissue by RNA in situ hybridization (RNAish), IHC, and ddPCR using commercially available probes and antibodies. We prospectively stained lungs, hearts, liver, and brain autopsy tissue and an umbilical cord from patients with recently identified COVID-19 and a cell line infected with SARS-CoV-2. In addition, we assessed cell lines infected with other viruses to exclude potential cross-reactivity of our antibodies and probes. We compared our results with patients who died with other viral infections or acute lung injury due to other causes. Reproducibility in the evaluation of the IHC- and ISH-stained slides was also recorded.
MATERIALS AND METHODS
All autopsies performed at our institution (March 2020 through September 2020) on patients who tested positive for SARS-CoV-2 by antemortem nasopharyngeal or oropharyngeal swab were prospectively included. Almost all patients underwent a subsequent postmortem nasopharyngeal or oropharyngeal swab at time of autopsy. Viral swabs at autopsy were analyzed for SARS-CoV-2 based on RT-PCR technique as previously described.22 In 2 cases, postmortem serum was also analyzed for SARS-CoV-2 IgG antibodies by enzyme-linked immunosorbent assay as previously described.19 Clinical information was abstracted from medical records.
A subset of cases was obtained from ProteoGenex Inc (Inglewood, California). These patients had known COVID-19 and data on sex and age; no other clinical data were available. Another set of lung specimens from patients who underwent an autopsy and an umbilical cord was received from another institution; these patients had known COVID-19, and their age and sex was also known, but no other clinical information was available.
Control cases included patients who underwent autopsy and were found to have DAD or were diagnosed with influenza virus infection (February 2017–January 2020).
Legal next of kin consented for all autopsies that were performed at our institution, providing permissions that allowed for research and education. The study was approved by the institutional review board (20-003664).
At autopsy, multiple tissue blocks were retained from various organs, including but not limited to both lungs, heart, liver, and brain. Specifically, both lungs were individually perfused with formalin. After about 30 minutes to 1 hour, the lungs were serially sectioned and multiple sections were taken for microscopy from both lungs including areas of consolidation. All slides from the lung were reviewed by a thoracic pathologist, and blocks with the most pathologic findings, specifically acute lung injury, were selected for further testing.
In Situ Hybridization
The assay consisted of the nCoV-2019-S probe designed by Advanced Cell Diagnostics (Newark, California) containing 20 oligo pairs, each of which targets approximately 50 base pairs, within the region (21631–23303; accession NC_045512.2, Advanced Cell Diagnostics catalog 848569) of SARS-CoV-2. This sequence targets the S gene encoding the spike protein. The test was performed on the Ventana Discovery Ultra using RNAscope detection from Advanced Cell Diagnostics. Formalin-fixed, paraffin-embedded tissue blocks were cut at 4 μm. Staining protocol consisted of the following steps: (1) cell conditioning (Target Retrieval) for 16 minutes; (2) pretreatment with protease (pretreat 3 [mRNA pretreat B]) for 8 minutes; (3) slide incubation with target probe and hybridization for 2 hours at 43°C; (4) target signal amplification and detection with RNAscope VS Universal AP Reagent Kit (Red) (Advanced Cell Diagnostics) according to manufacturer's instructions. The final assay consisted of the nCoV-2019-S (target probe), Hs-UBC (positive RNA integrity control), and dapB (negative control) probes, each prepared on a separate slide, for each patient specimen. A positive control (Vero E6 cells were infected with SARS-CoV-2 USA WA1/2020 with multiplicity of infection = 0.01 and fixed with 10% formalin after 1 day postinfection) was floated on the bottom of each slide that was incubated with the target probe. The SARS-CoV-2 USA WA1/2020 was kindly provided by the World Reference Center for Emerging Viruses and Arboviruses at the University of Texas Medical Branch (Galveston). Inactivation and removal of SARS-CoV-2 samples from the biosafety level 3 facility was performed in accordance with standard operating protocols approved by our Institutional Biosafety Committee. All ISH-stained slides were reviewed and scored by 3 pathologists as positive, equivocal, or negative. Staining in areas of anthracotic pigment was regarded as nonspecific.
Formalin-fixed, paraffin-embedded tissue blocks were cut at 4 μm. Slides were stained with the mouse monoclonal SARS-CoV/SARS-CoV-2 (COVID-19) spike antibody (clone 1A9, GeneTex Inc, Irvine, California) at a 1/500 dilution. A positive control (same as used in RNAish experiments) was run with each set of experiments. All slides were reviewed by 2 pathologists and scored as positive, equivocal, or negative.
Reverse Transcriptase ddPCR
Tissue block(s) with the most significant morphologic findings of acute lung injury were selected for ddPCR. RNA, including both viral and human, was extracted from 3 to 5 FFPE unstained tissue scrolls cut at 10 μm by Qiagen RNeasy DSP FFPE Kit (Qiagen, Hilden, Germany). The RNA was quantified and ranged from 12 to 800 ng/μL. The assay was performed according to the manufacturer protocol, as outlined in the instructions for use for the Bio-Rad SARS-CoV-2 ddPCR Test (Bio-Rad Laboratories, Hercules, California), which has received Emergency Use Authorization from the US Food and Drug Administration. Briefly, 5.5 μL of RNA eluate (corresponding to a range of 66–800 ng of RNA; samples with high total RNA were also run at a dilution) was added to 16.5 μL of PCR master mix. The SARS-CoV-2 nucleocapsid (N1 and N2) target sequences, along with human RPP30 as a control, were amplified using Veriti thermocyclers (Applied Biosystems, Foster City, California). Droplets could be positive for one or multiple targets and were counted by the QX200 Droplet Reader (Bio-Rad). Signal data were analyzed using QuantaSoft Analysis Pro software version 1.0.596 (Bio-Rad). Positive cases were defined as those with more than 4 droplets of the N1 and/or N2 target. This cutoff was based on the manufacturer's recommendation for the Emergency Use Authorization version of the assay, which considers a sample positive when there are 2 or more N1- and/or N2-positive droplets, and our laboratory's validation studies. Note that the matrix of FFPE is significantly different from the matrix of the Emergency Use Authorization version of the assay. Therefore, the limit of detection was evaluated by spiking positive control N1 and N2 RNA into RNA extracted from SARS-CoV-2-negative RNA extracted from FFPE tissue. The presence of RPP30 in blocks containing human tissue was used to indicate successful RNA extraction. The test results were interpreted by a clinical molecular geneticist. Samples were run once, as per the clinical workflow. The mean value of the N1 and N2 SARS-CoV-2 copy number was used in subsequent analyses. For some patients, multiple tissue sections (blocks) were available. In those cases, the mean number of SARS-CoV-2 copies from the section with the greatest viral load was used in subsequent calculations.
Demographics and clinical features were summarized by medians and ranges. Interobserver reproducibility amongst the 2 (IHC) or 3 (RNAish) pathologists was assessed by using the weighted Cohen κ coefficient using the cohen.kappa function of the psych package (Revelle W . psych: Procedures for Psychological, Psychometric, and Personality Research. Northwestern University, Evanston, Illinois, R package version 2.0.12, https://CRAN.R-project.org/package=psych, in the R programming language [version 4.0.3] R Core Team . R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria) with default parameters. Interassay reproducibility between IHC, RNAish, and ddPCR was performed with the results of reviewer 1 by using the weighted Cohen κ coefficient using the cohen.kappa function of the psych package in R with default parameters. The interpretation of the κ values was 0 as no agreement, 0.01 to 0.2 as none to slight, 0.21 to 0.4 as fair, 0.41 to 0.6 as moderate, 0.61 to 0.8 as substantial, and 0.81 to 1 as almost perfect agreement. Correlation tests between continuous variables were performed using the Pearson product moment correlation coefficient, using the cor.test function in R with default parameters. Binomial 95% CIs around sensitivities and specificities were provided using the binom.test function in R with default parameters. Testing for independence of categorical variables was performed using the Fisher exact test for count data using the fisher.test function in R with default parameters.
Equivocal results were considered as negative in the statistical analysis rather than being eliminated or updated by discrepant resolution, according to guidelines of the US Food and Drug Administration.23
A P value of <.05 was considered significant.
The study included lungs from 26 autopsies of patients who had tested positive for SARS-CoV-2, including patients from our institution (n = 14), another institution (n = 7), and ProteoGenex Inc (n = 5). Only tissue blocks that had staining for the RNA integrity control (Hs-UBC) were included. These lungs resulted in 62 tissue blocks, as multiple tissue blocks were available from 14 autopsy lungs from our institution (median number of tissue blocks, 3; range, 2–9). From all other autopsy lungs a single tissue block per case was available. The demographics of the study population of lung autopsies and the results of COVID-19 testing at time of autopsy are summarized in Tables 1 and 2. All patients from our institution had tested positive for COVID-19 by nasopharyngeal swab within a median of 14.5 days (range, 0–67 days) before death. All patients from our institution but one were tested for COVID-19 again at time of autopsy; 10 of 13 (76.9%) tested positive. The median time between onset of COVID-19–related symptoms and death was 19.5 days (range, 3–100 days; unknown in 2 cases). The lung findings of 8 cases were reported previously.4
In addition, 22 specimens from 20 autopsy hearts (our institution, n = 12; ProteoGenex Inc, n = 10), 1 autopsy brain with 2 tissue blocks (our institution), 1 autopsy liver (our institution), and 1 umbilical cord (other institution) from COVID-19 patients were included.
Control cases included 9 autopsy lungs (our institution) from patients who died before the COVID-19 era and who were found to have DAD (n = 7), including patients with pneumonia associated with adenovirus, respiratory syncytial virus, or herpes simplex virus (n = 1 each) as identified at time of autopsy. The other 2 patients were diagnosed with influenza A or B (including the 1 patient who died in January 2020) and lungs showed either acute bronchopneumonia (n = 1) or pulmonary edema (n = 1). The demographics of the control group are summarized in Table 1. Controls of either tissue or cell lines infected with adenovirus, human herpesvirus 8, Epstein-Barr virus, JC virus, varicella-zoster virus, human papillomavirus 16, BK virus, enterovirus, cytomegalovirus, human papillomavirus 6 or 11, respiratory syncytial virus, herpes simplex virus I or II, polyomavirus, or influenza A virus and Raji cell line (n = 1 each) were also included as negative controls.
ddPCR Has Highest Sensitivity for SARS-CoV-2 Detection With High Specificity of ddPCR, IHC, and ISH
The results of ddPCR, IHC, and RNAish are summarized in Table 1 and detailed in Table 2 and the Supplemental Table (see supplemental digital content at https://meridian.allenpress.com/aplm in the July 2021 table of contents). Droplet digital polymerase chain reaction had the highest sensitivity for the detection of SARS-CoV-2 in lungs of patients who died with COVID-19. Droplet digital polymerase chain reaction was positive in 18 of 19 COVID-19 autopsy lungs tested (94.7%) whereas IHC was positive in 12 of 26 (46.1%) and RNAish was positive in 14 of 26 (53.8%) COVID-19 autopsy lungs.
As some patients had multiple samples collected and tests performed, we evaluated sensitivity and specificity of ddPCR, IHC, and RNAish in all available testing results. Using the original COVID-19 diagnosis of patients as the gold standard, ddPCR in autopsy lungs had a sensitivity of 96% (95% CI, 81%–100%) and a specificity of 100% (95% CI, 66%–100%) for the detection of SARS-CoV-2 in lungs of patients who died with COVID-19 (Table 1). In contrast, the sensitivity and specificity of IHC were 31% (95% CI, 19%–44%) and 100% (95% CI, 59%–100%), respectively, and of RNAish were 36% (95% CI, 24%–49%) and 100% (95% CI, 66%–100%), respectively. All control cases were negative by ddPCR, IHC, and RNAish.
Using ddPCR in autopsy lungs as the gold standard, the sensitivity and specificity of IHC and RNAish were 38% (95% CI, 19%–59%) and 100% (95% CI, 63%–100%) and 48% (95% CI, 28%–69%) and 100% (95% CI, 69%–100%), respectively.
The correlation between ddPCR and IHC and ddPCR and ISH using all available lung tissue blocks, including COVID-19 lungs and control lungs, was fair, with κ = 0.23 (n = 32) and κ = 0.35 (n = 35), respectively. The correlation between IHC and ISH was substantial, with κ = 0.75 (n = 66).
The interobserver variability for IHC between 2 reviewers was almost perfect, with κ = 0.91 (n = 63), and for RNAish between 3 reviewers was fair to moderate, with pairwise κ = 0.59, 0.38, and 0.43 (n = 55, 53, and 53).
Detection of SARS-CoV-2 in Lung Tissue at Autopsy Correlates With the Interval Between Onset of Symptoms and Death
The median time between onset of symptoms and death (available in 12 cases) was 19.5 days (range, 3–100 days). The number of days between onset of symptoms and death correlated with ddPCR levels in autopsy lung tissues (P = .03, Pearson correlation = −0.65) (Figure 1, A), with ddPCR copy numbers being higher in lungs of patients who died closer to onset of symptoms. Lungs that were positive by IHC or RNAish either trended to be or were significantly associated with being from patients who died sooner after onset of symptoms when compared with lungs with negative IHC or RNAish (P = .08 and P = .02, respectively) (Figure 1, B). The median time between first positive SARS-CoV-2 testing and death was 14.5 days (range, 0–67 days; n = 14). Similarly, the number of days between first positive SARS-CoV-2 testing and death trended to be lower with higher copy numbers of ddPCR in lungs (P = .06, Pearson correlation = −0.56). Patients with lungs that were positive by IHC or RNAish trended to have died earlier after first testing when compared with patients with IHC or RNAish negativity (P = .96 and P = .21, respectively). For instance, ddPCR, IHC, and RNAish were negative in the patient with onset of symptoms 100 days prior to death; the lung tissue from the 2 patients with onset of symptoms 47 and 55 days prior to death showed low copy numbers using ddPCR and was negative by IHC and RNAish. Notably, the patient who died at day of testing and 3 days after onset of symptoms showed very low copy numbers (30 and 36, respectively) using ddPCR, and IHC and RNAish were negative in multiple tissue blocks of the lung (n = 3).
Expression of SARS-CoV-2 Viral Protein and RNA is Patchy in Hyaline Membranes, Pneumocytes, and Respiratory Epithelium in Lungs of COVID-19 Patients
Expression of SARS-CoV-2 viral protein and RNA was very similar when compared between IHC and RNAish, although overall, staining was typically weaker with IHC. Expression patterns of viral protein and RNA for each patient are detailed in the Supplemental Table. Expression of SARS-CoV-2 viral protein and RNA was identified in hyaline membranes (Figure 2, A through C), in the cytoplasm of intra-alveolar cells (Figure 3, A through D), occasionally in the cytoplasm of pneumocytes (Figure 3, E), and rarely in the respiratory epithelium of bronchi or bronchioles (Figure 3, F). In some cases, staining was found in intra-alveolar cells that were associated with hemosiderosis (Figure 4, A through C). Whether the expression was strong or weak, the expression of SARS-CoV-2 viral protein and RNA appeared patchy or scattered in all cases (Figure 4, D and E). Therefore, sampling bias might play a role in tissue blocks and cases that appeared to be negative. There was a correlation between presence of hyaline membranes and expression of SARS-CoV-2 viral protein by IHC (P = .04; Table 3) but not viral RNA by RNAish (P = .17) or ddPCR (P = .94). There was no correlation between the presence of acute bronchopneumonia and ddPCR (P = .69), IHC (P = .78) or RNAish (P = .41) being positive for SARS-CoV-2 antigen or RNA, respectively. There was also no correlation between presence of organizing DAD and IHC (P = .74), RNAish (P = .53), or ddPCR (P = .53). We did note nonspecific staining in association with anthracotic pigment with IHC and RNAish. No definite expression of SARS-CoV-2 viral protein or RNA was identified in endothelial cells.
Results of SARS-CoV-2 Detection in Other Organs
In addition to lung specimens, 1 umbilical cord (other institution), 22 hearts (12 autopsy heart specimens from our institution, 1 of which has been previously reported,21 and 10 autopsy hearts from ProteoGenex Inc), 1 autopsy brain (2 tissue blocks, our institution), and 1 autopsy liver (our institution), all derived from COVID-19 patients, were also included. Three hearts from ProteoGenex Inc had to be excluded because of no or only weak RNA integrity. Droplet digital polymerase chain reaction showed a median viral SARS-CoV-2 copy number of 11, ranging from 5 to 28.5, in the 7 autopsy hearts from ProteoGenex Inc. All other cases, including the autopsy hearts from our institution, were negative for ddPCR and negative or equivocal for IHC and RNAish (IHC and RNAish reviewed by a single pathologist).
In our study of 26 autopsy cases resulting in 62 lung specimens of COVID-19 patients, we found that all 3 assays, ddPCR, IHC, and RNAish, were highly specific for the identification of SARS-CoV-2 in lung FFPE tissue. Furthermore, the sensitivity of ddPCR was superior to that of IHC and RNAish, both of which had similar sensitivities. We also confirmed that the expression of SARS-CoV-2 in lung tissue is highest within the first few weeks after infection. Expression of SARS-CoV-2 viral protein and RNA was patchy and occurred in hyaline membranes and the cytoplasm of pneumocytes and rarely respiratory epithelial cells in lungs. Expression of SARS-CoV-2 viral protein and RNA was also identified in intra-alveolar cells, but it was not clear whether those represented macrophages. Although high copy numbers of ddPCR were identified only in lung tissue of COVID-19 patients, low copy numbers were found in some postmortem hearts of COVID-19 patients from ProteoGenex Inc. Immunohistochemistry and RNAish identified SARS-CoV-2 infection only in lung tissue; all other tissues tested and derived from COVID-19 patients were negative. Our results suggest that ddPCR might be the most useful test for the identification of SARS-CoV-2 in FFPE tissue. Although ddPCR cannot identify the cell or morphologic structure that harbors the virus, it is the most sensitive and a highly specific test, and its results will be crucial for management of patients.
Only a few smaller studies have compared expression of SARS-CoV-2 viral protein and RNA by IHC and RNAish with PCR and, to our knowledge, none of the studies used ddPCR in FFPE tissue. For instance, Massoth et al12 compared RNAish and IHC with real-time quantitative RT-PCR, including 19 pulmonary and 39 extrapulmonary samples (heart, liver, kidney, small intestine, skin, adipose tissue, and bone marrow) from a total of 8 autopsies and control lungs from 37 autopsies performed before the COVID-19 pandemic. In contrast to our findings, their results in lung tissue from COVID-19 patients showed sensitivity of 86% and 87% for detection of the virus by IHC and RNAish, respectively, when compared with quantitative RT-PCR using a polyclonal anti-SARS nucleocapsid protein antibody and including equivocal staining as positive in 5 of 9 cases of IHC-tested lung specimens. When compared with ddPCR, the sensitivity of IHC and RNAish in our cases was only 38% and 48%, respectively. The lower sensitivity of these assays in our study most likely stemmed from the higher sensitivity of ddPCR in contrast to quantitative RT-PCR, which was used by Massoth et al.12 Potentially, the IHC clone 1A9 might be less sensitive than the anti-nucleocapsid protein polyclonal antibody that was used by Massoth et al.12 However, using that polyclonal nucleocapsid antibody, the specificity was reported as only 53.5%, which is much lower than the specificity of 100% in our study. In addition, we treated our equivocal results as negative, in contrast to Massoth et al,12 who treated those as positive. Furthermore, we noted that the expression of viral protein and viral RNA depended on the time between symptoms and death; in some of our cases that time was up to 100 days, and therefore our sensitivity might have been lower. It is not clear what the time between symptoms and death was in the series by Massoth et al.12
Although IHC for detection of the expression of SARS-CoV-2 viral protein had 100% specificity in our study, the sensitivity was only 31%. Several antibody clones have been proposed for the detection of SARS-CoV-2 antigen by IHC with various sensitivities and specificities for detection of the virus in FFPE tissue. For instance, Szabolcs et al24 showed that among 7 tested commercially available antibody clones, only clones 001 to nucleocapsid protein and 1A9 to S2 subunit spike protein were useful for detection of SARS-CoV-2 in FFPE tissue. Expression of SARS-CoV-2 viral RNA was confirmed by RNAish in that study. Carossino et al11 also tested various antibodies to spike and nucleocapsid protein of SARS-CoV-2, which showed equivalent cytoplasmic staining of infected Vero cells including clone 1A9; however, the nucleocapsid protein antibody clone 6F10 apparently revealed the most intense staining. Best Rocha et al17 identified SARS-CoV-2 expression in all 8 studied autopsy lungs from known COVID-19 patients by IHC using Bioss BSM-41411M anti–SARS-CoV-2 nucleocapsid protein antibody (His-Tag) (presumed clone 1C7), Bioss BSM-49131M anti-SARS nucleocapsid protein antibody (no tag) (presumed clone 8G8A), and ThermoFisher Scientific MAI-7404 anti-SARS nucleocapsid protein antibody preparation (presumed clone B46F).
After our own validation of multiple clones (data not shown), we chose clone 1A9 for IHC of SARS-CoV-2 because of the most pristine staining without background noise. Using this clone, we found viral expression in hyaline membranes and the cytoplasm of occasionally pneumocytes, and rarely in respiratory epithelium. We did not identify expression of viral protein in endothelial cells when using IHC or RNAish. In our experience, diagnostic sensitivity and the expression patterns of SARS-CoV-2 observed with IHC were very similar to those of RNAish. Indeed, in our study, the agreement of viral expression between IHC and RNAish was substantial. Interestingly, although the strength and extent of expression varied among cases, detection was focal or patchy in all cases, independent of method (IHC or RNAish). Szabolcs et al24 reported similar findings in 3 autopsy lungs with strong expression of SARS-CoV-2 viral protein by IHC in hyaline membranes, alveolar macrophages, and alveolar lining predominantly type II pneumocytes, with similar expression patterns reported with RNAish. Similar to our study, Massoth et al12 found extracellular viral protein and RNA, predominantly within hyaline membranes, in pneumocytes, and immune cells; the authors also described regional variation in the amount of detectable virus in lung samples. Borczuk et al8 performed IHC using clone 1A9 and RNAish, both of which were positive in tracheal epithelium, hyaline membranes, and atypical type II pneumocytes in 13 of 23 cases (56.5%). Bradley et al18 performed IHC using an antibody to spike protein (most likely clone 1A9) in 4 autopsy lungs of COVID-19 patients highlighting alveolar pneumocytes and in bronchioles with sloughed ciliated respiratory epithelium. In one patient, submucosal glands and lymphocytes stained as well. Only rare studies reported expression of SARS-CoV-2 protein or RNA in endothelial cells, including Li et al25 using RNAish and Bussani et al26 using IHC and RNAish. In contrast, we and others8,27 did not find definite viral expression in endothelial cells.
We found the interpretation of the SARS-CoV-2 RNAish can occasionally be challenging given the patchy and sometimes very focal expression of the viral RNA. Furthermore, nonspecific staining occurred in areas of anthracotic pigment. Not surprisingly, our interobserver reproducibility for RNAish was only moderate whereas for IHC reproducibility was almost perfect. Interestingly, in the study by Massoth et al12 the interobserver variability appeared to be better for RNAish ranging from moderate to almost perfect whereas that for IHC ranged from slight to moderate when evaluated by 4 independent reviewers. Overall, the more challenging reproducibility of IHC and RNAish might further support ddPCR as diagnostic test of choice for detection of SARS-CoV-2 in FFPE tissue.
Although in our study none of the extrapulmonary specimens including an umbilical cord showed expression of SARS-CoV-2 viral protein or RNA, except for low copy numbers by ddPCR of autopsy hearts from ProteoGenex, literature suggests that rare placentas can show expression of viral protein or RNA. For instance, Facchetti et al14 showed expression of SARS-CoV-2 protein and RNA in 1 of 15 placentas. In that placenta the virus was found in the cytoplasm of syncytiotrophoblasts using antibody clone 007 directed against spike protein and antibody directed against nucleocapsid protein by IHC and by RNAish. The nucleocapsid protein was also expressed by rare macrophages and Hofbauer cells in that study. In addition, there were reports of 4 cases of expression of SARS-CoV-2 protein and/or RNA as identified by IHC (n = 2) and ISH (n = 4) predominantly in syncytiotrophoblast cells of placentas of patients with COVID-19.15–17 However, another study did not reveal any expression of viral protein in 51 placentas using IHC (clone 1A9).28 The significance of the low copy numbers by ddPCR in the autopsy hearts from ProteoGenex in our study is not entirely clear, especially in light of multiple prior reports, including one from our institution, not revealing any virus in hearts from COVID-19 patients.18,21,24 Low-level contamination might be a possibility, as all these FFPE tissue blocks were derived from a single company. On the other hand, these results might represent true low levels of SARS-CoV-2 in these hearts. Bradley et al18 found IHC staining in the 2 kidneys tested that was patchy, granular cytoplasmic in renal tubular epithelial cells. In other studies, no expression of viral protein was found in kidneys.17 In the study by Bradley et al,18 quantitative RT-PCR detected virus in the lung, trachea, subcarinal lymph node, kidney, large intestine, and spleen of 3 cases tested; in 2 of these it was also detected in the liver, heart, and blood. In all 3 patients the lungs and trachea had the lowest cycle threshold values.
We found that the expression of SARS-CoV-2 RNA in lungs from patients with COVID-19 depended on the time between onset of symptoms and death, with only low copy numbers of the virus by ddPCR and no expression of the viral protein or RNA by IHC and RNAish found in cases beyond 46 days after onset of symptoms. Similarly, Borczuk et al8 found that most autopsy lungs showed viral RNA by ISH within the first 2 weeks of disease, but IHC and RNAish were less frequently positive beyond 2 weeks. The authors also reported that viral culture was positive in only 1 case after 26 days from clinical presentation. These observations might indicate that although the virus might be cleared from the lungs, the damage to the lungs is still ongoing.
Our study showed that ddPCR is the most sensitive and highly specific method to detect SARS-CoV-2 in FFPE tissue of lungs and is superior to IHC and RNAish. Furthermore, expression of the viral protein or RNA is primarily identified in lung FFPE tissue with this method; however, additional investigation and sampling of other tissues is warranted.
The SARS-CoV-2 USA WA1/2020 was kindly provided by the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA) at the University of Texas Medical Branch (UTMB).
Supplemental digital content is available for this article at https://meridian.allenpress.com/aplm in the July 2021 table of contents.
Robert Monroe is an employee and stockholder, Bio-Techne Corporation, Minneapolis, Minnesota. The other authors have no relevant financial interest in the products or companies described in this article.