Adipose tissue in COVID-19: detection of SARS-CoV-2 in adipocytes and activation of the interferon-alpha response

As shown in Table 1, three case groups have been investigated: (a) controls, i.e., subjects who died from acute causes other than infectious (trauma, sudden cardiac death; n = 12); (b) patients dying from COVID-19 (abdominal adipose tissue negative for SARS-CoV-2; n = 10); (b) patients dying from COVID-19 (abdominal adipose tissue positive for SARS-CoV-2; n = 13). Results refer to subjects of Caucasian ethnicity. Autopsies have been performed at the Unit of Forensic Medicine, Azienda USL Toscana Nordovest, Lucca, Italy. The recruitment area of the Institution includes four major City Hospitals: Lucca, Pisa and Massa-Carrara caring for approximately 1.3 million populations.

At autopsy, all cases were screened for the SARS-CoV-2 genome in the inferior lobe of both lungs as well as in the abdominal adipose tissue. As previously reported [14], alterations of lung parenchyma consistent with moderate to severe disease were detected in all COVID-19 cases, not in controls. Adipose tissue specimens of COVID-19 cases were compared to those of controls with regard to histopathological features (Fig. 1A and B), transcription levels of immune-related genes, infiltrating leukocyte markers, expression of SARS-CoV-2 antigens, and detection of SARS-CoV-2 genome by gene amplification. The study was approved by the local Ethical Committee (Comitato Etico Area Vasta Nord-Ovest, Italy No. 17327, 2020–05-14).

For each specimen, total RNA was extracted from four 10 µm-thick FFPE sections of abdominal WAT using the RNeasy FFPE kit (Qiagen, Hilden, Germany) and quantified by spectrophotometry (Trinean, Gentbrugge, Belgium). About 250 ng of RNA were used for each RT-PCR test to detect two genes of SARS-CoV-2 [viral nucleocapsid (N) and RNA-dependent RNA Polymerase (RdRp)] using the SARS-CoV-2 WE kit (Diatech Pharmacogenetics, Jesi, Italy). All samples were run in duplicate. Results were deemed positive when amplicons were obtained from at least one of the two genes. To evaluate the expression of viral nucleocapsid antigens and PKR by immunohistochemistry (IHC), 30 randomly selected high power (40x) fields were scanned, and the percentage of stained cells was calculated.

Gene expression analysis was performed by the nCounter technology (nanoString Technologies, Seattle, WA, USA) using two panels: (a) the Host Immune Response panel; (b) the Coronavirus panel plus that contains probes targeting the N and S ORFs of HCoV-229E, HCoV-HKU1, HCoV-NL63, HCoV-OC43, SARS-CoV, and the ACE2 virus receptor. Probes for the SARS-CoV-2 virus were designed according to the reference sequence, Wuhan-Hu-1 (NC_045512). For the assay, about 250 ng of RNA were hybridized with probes at 65 °C for 21 h. Procedures were performed following the manufacturer’s protocol.

Immunostaining for viral proteins was performed in all cases except for one case that was not assessed due insufficient tissue. Markers of immune cells have been determined in 11/12 controls, 7/10 COVID-19 cases virus-negative in WAT, 11/13 COVID-19 cases virus-positive in WAT. In some cases, the analysis could not be performed due to insufficient material. Three-micrometer-thick sections were stained with antibodies to SARS-CoV-2: rabbit polyclonal antibody to the nucleocapsid protein (NB100-56,683, Novus Biologicals, Centennial, CO, United States) and mouse monoclonal antibody to SARS-CoV-2 spike glycoprotein (GTX632304, GeneTex). Antibodies to immune cell markers (Roche Diagnostics, Ventana Medical Systems, Oro Valley, AZ, United States) comprised: CONFIRM anti-CD3 (2GV6) rabbit monoclonal, CONFIRM anti-CD20 (L26) mouse monoclonal, CONFIRM anti-CD45, LCA (RP2/18) mouse monoclonal, CONFIRM anti-CD68 (KP-1) mouse monoclonal, anti-CD57 (NK1) mouse monoclonal, CONFIRM anti-CD15 (MMA) mouse monoclonal. Staining for the human RNA-dependent protein kinase (PKR) was performed using a PKR rabbit polyclonal antibody (Invitrogen-ThermoFisher Scientific; 1:50 dilution, incubation at 36 °C for 48 min). Sections of normal human kidney were used as positive control [19]. Staining procedures were performed with an automated staining system (BenchMark ULTRA, Ventana Medical Systems). Antibody binding was revealed with the OptiView DAB IHC Detection Kit (Ventana Medical Systems). Slides were counterstained with Hematoxylin II and Bluing Reagent (Ventana Medical Systems). Counts of immune markers were performed independently by two pathologists (A.P. and F.B). Twenty randomly selected fields were scanned at 20 × magnification. IHC score: cell counts represent the mean numbers of cells per square mm.

Raw data of gene transcripts were normalized using the Advanced Analysis module of the nSolver v.4.0 (nanoString Technologies). Low count genes defined as those with raw expression level as low as 20 counts were omitted. Normalized gene expression levels were log2 transformed for further analysis. The principal component analysis (PCA) was performed by the procedures of the PCAtools Bioconductor package v.3.12 after removing 10% of genes with the lowest variance. Differentially expressed genes (DEG) were computed by a linear model adjusting for age, sex and BMI, and using the Advanced Analysis module of nSolver. In detail, control cases were used as baseline and three comparisons were done: all COVID-19 cases vs baseline; COVID-19 cases with no virus detected in WAT vs baseline; COVID-19 cases with SARS-CoV-2 detected in WAT vs baseline.

The IFNα score was computed by single sample Gene Set Enrichment Analysis (ssGSEA) using the procedures of the GSVA Bioconductor package v.1.38.2. The method described by Barbie [20]  was used. Gene Set Enrichment Analysis (GSEA) was run using the ranked gene list of differential expression analyses and following the procedures of the clusterProfiler Bioconductor package v.3.18.1. For analysis, ten was used as minimum gene set size cut-off, and the Hallmark collection of the Molecular Signatures Database (MSigDB) v.7.4 behaved as a reference [21] . Gene functions are reported according to the human gene databases GenCards and OMIM.

Survival times after initial symptoms were analyzed by Cox regression both in univariate and multivariate setting following the procedures of the survival R package v.3.2–13. Linear regression analysis was performed to evaluate the relationships among BMI and transcripts for SARS-CoV-2 entry molecules (ACE2, TMPRSS2, FURIN), as well as those between BMI and IFN alpha score. Pearson’s correlation coefficient was then calculated to quantify the associations of BMI and IFN-alpha score after adjusting for age and sex. Scatter plot were generated after partial regression. Analyses and plots were generated in R environment (https://​www.​r-project.​org/​, v.4.0.2, last accessed May 10, 2021) and with Prism (GraphPad Software, San Diego, CA).

The demographic and clinical data of the investigated autopsy groups are shown in Table 1. Compared to controls (sudden death from trauma or sudden cardiac arrest), overweight or obesity were more frequent among COVID-19 cases (p = 0.03). All COVID-19 patients were positive for the SARS-CoV-2 genome in lungs. Among COVID-19 cases, however, the virus genome was detected in abdominal WAT of 13/23 (56%) subjects. Uninfected controls and COVID-19 cases whose WAT was SARS-CoV-2-negative by PCR were not stained by antibodies to the virus nucleocapsid antigen. Conversely, 12/12 COVID-19 cases whose WAT was virus-positive by PCR showed granular cytoplasmic staining in 1–5% adipocytes. Due to insufficient material, one SARS-CoV-2-positive case by PCR could not be assessed by IHC (Fig. 1C). In the same specimens, the antibody to the SARS-CoV-2 spike glycoprotein produced diffused staining that did not allow the precise localization of the viral protein (data not shown).

The nCounter system—based on the direct hybridization of probes without amplification steps—could detect the virus genome in 7/13 (54%) WAT specimens of COVID-19 cases that were virus-positive by PCR in abdominal fat. In the latter cases, hybridization occurred with probes targeting SARS-CoV-2, not with probes specific for other coronaviruses (HCoV-229E, HCoV-HKU1, HCoV-NL63, HCoV-OC43, SARS-CoV). As expected, hybridization failed to detect coronaviruses in WAT of controls and of COVID-19 cases whose WAT was virus-negative by PCR.

Finally, it was evaluated whether SARS-CoV-2 positivity in subcutaneous WAT impacted survival times. Using a univariate setting, the association was not statistically significant (p = 0.13). However, upon adjustment for age, sex and BMI, detection of SARS-CoV-2 in subcutaneous WAT was associated with a reduced survival time from initial symptoms (HR 3.7, 95%CI 1.1–12.5, p = 0.03).

Inflammatory infiltrates were evaluated by staining for leukocyte markers. Compared to controls and to COVID-19 specimens that were negative for virus in WAT, virus-positive specimens showed significantly increased numbers of cells expressing CD45 pan-leukocyte marker (p < 0.01) (Fig. 1D), CD3 T-cells (p < 0.05) (Fig. 1E), CD57 natural killer cells (p < 0.05) (Fig. 1F), and CD68 macrophages (p < 0.05) (Fig. 1G) (Fig. 2). CD20 B-cell leukocytes could not be detected in any groups (Fig. 1H).

Expression of the interferon-induced PKR protein (whose activated form phosphorylates the Eukaryotic Translation Initiation Factor 2 Subunit Alpha (EIF2S1) which, in turn, inhibits protein synthesis) was evaluated by immunostaining [19]. Compared to the control group, WAT specimens of all COVID-19 cases showed significantly increased numbers of adipocytes expressing PKR, no matter if WAT specimens were virus-positive or virus-negative (Fig. 1I, L). In COVID-19 cases, each high-power field (40x) comprised 2–6 PKR-positive adipocytes. This was significantly different (P < 0.05) from what observed in controls which had less than 0.3 PKR-positive cells per high power field (data not shown).

Sections of normal human kidney were used as positive control for PKR staining. As expected, PKR was highly expressed in the cytoplasm of kidney tubular cells (Fig. 1 L, inset).

As compared to controls, COVID-19 cases showed no marked deregulations of immune gene transcripts. To highlight possible pathway deregulations in the presence of small fold changes, a GSEA analysis was performed Fig. 2. As compared to the control group, only the IFN-alpha pathway was significantly activated in virus-positive WAT specimens (FDR = 0.009, Fig. 3). SARS-CoV-2 utilizes the ACE2 (and possibly other factors such as BSG, NRP1 and HSPA5) as cell entry factors together with the priming proteases TMPRSS2, CTSL, FURIN that facilitate virus entry [22]. In subcutaneous WAT, only transcripts of ACE2 and FURIN genes could be detected. Transcription of ACE2 was upregulated in COVID-19 cases. Interestingly, the IFN-alpha score was directly related to ACE2 transcripts levels after adjustment for age and sex (partial r = 0.53, p = 0.0014) (Fig. 4). Comparable results were observed also after adjustment for BMI (partial r = 0.47, p = 0.006). Furthermore, transcription of the immunoglobulin alpha Fc receptor (FCAR)) gene was strongly downregulated, suggesting that SARS-CoV-2 plays some immunosuppressive role [23].