Case 1 was a man in his 70 s with severe diabetes mellitus and renal failure who was diagnosed with severe COVID-19 infection and admitted to the hospital, and mechanical ventilation was started on admission. Sputum cytology was performed on the 30th day after illness onset. Case 2 was a woman in her 70s with a history of hypertension. She was diagnosed with a severe COVID-19 infection and admitted to the hospital. Oxygen was administered without the use of a ventilator. Pneumonia was thought to have become organizing pneumonia, but new consolidation on chest radiography was observed. Sputum cytology was performed on the 24th day after the onset of illness. In both cases, sputum culture did not show any evidence of secondary bacterial infection, but their symptoms worsened during the course of their hospitalization, and their SARS-CoV-2 PCR test results remained positive at the time of sputum cytology.
Sputum samples submitted to the pathology laboratory were processed in a Class II biological safety cabinet by technicians wearing full personal protective equipment, according to community-acquired pneumonia guidelines . Specimens were prepared by the rubbing method (gently rubbing two glass slides together). Two sputum samples, from patients without pulmonary infection clinically, and normal cytological findings, were used as clinical controls. One hundred cells with frosted nuclei, which are considered characteristic of virus-infected cells, were examined for the number of nuclei, presence of nuclear inclusion bodies, mutual pressure exclusion images of the nuclei, and presence of eosinophilic granules in the cytoplasm. If mutual pressure exclusion images were present, the nuclei were considered parallel to each other.
To confirm the origin of these cells and the infecting virus, a cell transcription assay was performed. After Papanicolaou staining, the samples were sealed with mounting medium. After immersion in xylene until the cover glass spontaneously peeled off, the cytoplasm was decolorized by passing through a further two baths of new xylene, followed by two baths of anhydrous alcohol (combined, about 30 min), and then rinsed under running water. Next, hematoxylin was decolorized with 1% hydrochloric acid alcohol (hydrochloric acid in 70% ethanol in a ratio of 1%) and then rinsed under running water. After immersion in 95% ethanol and a water-soluble aerosol fixative (Cytokeep II, Alfresa Pharma Corp., Osaka, Japan) for exfoliated cytology, the fixative solution was allowed to dry .
Cell immunostaining was performed using the same cells that were decolorized for Papanicolaou staining. Primary antibodies were against cytokeratin (AE1/AE3) (MilliporeSigma, Burlington, MA, USA, mouse monoclonal, clone MAB3412, dilution 1:200), HSV I + II (Dako, Glostrup, Denmark, rabbit polyclonal, 1:2000), VZV (MilliporeSigma, mouse monoclonal, clone MAB8612, 1:2000), CD68 (mouse monoclonal, clone M0876, 1:250), and SP-A (mouse monoclonal, clone M4501, 1:300). Each set of reagents was allowed to react at room temperature for 30 min. Secondary antibodies were Leica Bond polymer system (Leica Biosystems, Newcastle Upon Tyne, United Kingdom), IHC refine kit (DS9800), and 3,3′ diaminobenzidine (DAB), a chromogen. In addition, for double staining, a Vina Green™ Chromogen kit (Biocare Medical, Concord, CA, USA,) was used for blue coloration. Positively stained histopathology specimens were used as positive controls. The primary antibody to SARS-CoV-2 spike protein was mouse monoclonal antibody GTX632604 (GeneTex, Irvine, CA, USA) diluted 500 times and reacted 4 times overnight, and the secondary antibody was Histofine MAX-PO (M) (Nicherei Biosciences Inc., Tokyo, Japan), colored with DAB. As a methodological control, we confirmed that SARS-CoV-2 spike protein could be detected by immunohistochemistry after Papanicolaou staining and subsequent decolorization, using the SARS-CoV-2-infected VeroE6/TMPRSS2 cell line  cultured in DMEM, obtained from the Japanese Collection of Research Bioresources (JCRB) Cell Bank. VeroE6/TMPRSS2 cell line without SARS-CoV-2-infection was used as a negative control. In addition, as a negative control of clinical samples, we used cytologically normal sputum cells collected from two patients who were clinically free of respiratory infection. These controls were stained as same as described above (Supplemental Fig. 1).
In both cases, Papanicolaou staining revealed several cells with a frosted appearance that appeared to be infected with the virus. These cells were mononuclear or multinuclear with chromatin aggregation at the nuclear margins, eosinophilic intranuclear inclusion bodies, mutual pressure exclusion of nuclei, and eosinophilic granules in the cytoplasm (Fig. 1a–c). In a sample obtained from Case 1, there were 44 mononuclear cells, 56 multinucleated cells, 5 intranuclear inclusion bodies, and 86 acidophilic granules, and 56 multinuclear cells (100%) showed a mutual pressure exclusion image. Fungal cells (Aspergillus niger) were also observed. In a sample obtained from Case 2, 68 mononuclear cells, 32 multinuclear cells, 10 intranuclear inclusion bodies, and 86 eosinophilic granules and 31 of 32 multinuclear cells (96%) showed mutual nuclear exclusion (Table 1). Immunocytochemical staining revealed the expression of AE1/AE3 ( +) (Fig. 1d), HSV I + II ( +) (Fig. 2b), varicella-zoster virus (VZV) ( −) (Fig. 2d), SARS-CoV -2 spike protein ( +) (Fig. 3b), and CD68 ( −) cells (Fig. 4). Examination of the Papanicolaou-stained specimens revealed both mononuclear and multinuclear cells among the SARS-CoV-2 spike-protein-positive cells, including cells with large nuclear inclusion bodies. In Case 2, HSV I and II double staining of SARS-CoV-2 spike protein-positive cells was positive, suggesting that a single epithelial cell was heavily infected (Fig. 5a–c). The distribution of positive and negative epithelial cells on double staining is shown in Table 2. Of the 76 epithelial cells considered, 5 (7%) were double positive. Both mono- and multinucleated cells were included. Furthermore, infected cells with frosted nuclei in Case 2 were positive for SP-A, suggesting a type II alveolar pneumocyte or Clara cell (Fig. 5d).