The main goal of this study was to examine the spectrum of autopsy findings in COVID-19 patients after long-term medical treatment, with a main focus on microbiological findings and particular emphasis on mycotic infections. Compared with an average autopsy rate of less than 4% in Germany , we reached an autopsy frequency of 47% in COVID-19 patients. This was obviously mainly the direct consequence of the bigger interest of the clinicians to perform autopsies; but it could also—at least partly—be attributed to the high willingness of relatives to agreement for performing autopsies to gain additional knowledge of this novel disease. Infections among the autopsy staff did not occur, and all performed antigen tests were negative, given the implementation of special safety precautions. All deceased patients suffered from non-severe comorbidities like hypertension and obesity. One severe previously known disease was liver cirrhosis in one patient (P4). This was clinically attributed to persistent alcohol abuse. Autoptically, however, we detected alpha-1-antitrypsin deficiency in this patient, whose course presumably was unfavorably influenced by the elevated alcohol consumption. This diagnosis alone underlines the general importance of autopsies, as this diagnosis will naturally have impact on the relatives.
Compared with other autopsy studies (mean: 72–79.2 years old [2, 8, 11, 14]), our patients were younger, with a median age of 62 years. In addition, they showed a longer hospitalization time (median 33.6 days) compared to other autopsy studies (6–16 days) [2, 8, 11, 14]. The younger age of our patients is presumably a consequence of the strict ECMO-criteria at the University Hospital of Regensburg. As a result of long-term maximum therapy (including mechanical ventilation partly combined with a prone position, dialysis, ECMO, and convalescent plasma) and the younger age, patients survived for a longer time. This gave us the opportunity to investigate a group of long-term treated patients with possibly a different spectrum of diseases and findings that have not yet been systemically investigated so far. In line with the hypothesis of time-related changes of alterations in autopsy findings, we found a relatively low rate and generally minor manifestation of the pathologic alterations that have been described as being typical for the early phase of the disease, such as endothelitis, thrombosis, or capillaritis . However, as a result of the pulmonary infection, we noticed severe damage and fibrosis, and remodeling of the lung parenchyma corresponding to different stages of DAD in 7 of 8 cases. Corresponding to the long-term ventilation, most of the patients showed DAD already in early fibrotic stage. Only one patient showed a severe pulmonary fibrosis with dens collagene deposits (Fig. 3e), leading to death in respiratory failure. Virus-infected cells similar to those observed in the course of cytomegalovirus or adenovirus infection with virus inclusions were not detected.
Importantly, structures that may represent autolytic virus particles were identified only in 3 of 35 electron microscopy samples. Actually, the potential of identifying coronavirus by electron microscopy remains a matter of debate , particularly in autopsy samples. The fact that in most of the analyzed specimens no or at best a small number of “suspicious” structures was found underlines the insensitivity and the lack of specifity of electron microscopy for this scope, at least in the post mortem context. Therefore, confirmatory immune electron microscopic examinations, which, however, are still pending, are mandatory to better understand the infectious route of the virus in different human tissues. Currently, molecular methods such as PCR seem to be better suited for virus detection in tissues, although they cannot locate the virus in a cell-specific manner.
PCR analyses of viral nucleid acids showed positive results in different organs (lung 6/8, trachea 2/3, heart 2/8, liver 3/8, kidney 3/8, heart 2/6, gut 1/1) in this study, with the highest detectable virus copy numbers in the lung (maximum 7*105 cop./ml). To further investigate whether the virus is still fully capable of replication requires the isolation and cultivation of SARS-CoV-2 via cell culture in future studies.
The clinically most important finding of this examination was the detection of invasive mycosis in inner organs in all but two cases. Moreover, 50% of the patients died of the consequences of mycotic sepsis. In general, aspergillus infections occur on average in about 10–12% of patients with DAD of any reason [6, 18]. Of note, specific mycosis was never diagnosed in our cohort clinically or microbiologically antemortem, despite prophylactic calculated antimycotic treatment (caspofungin or isavuconazole), which, on average, was given for 9.5 days. In addition, despite clear-cut histopathological evidence of angioinvasive pulmonary mycosis, subsequent molecular analyses were unable to detect fungal DNA in related FFPE samples. The lack of microbiological detection of fungi in these samples might be the known consequence of formalin fixation and the resulting DNA degradation , which hampers the results of microbiological methods that are designed to analyze fresh samples.
Particularly interesting was the finding of invasive infections with mucormycetes in P3 and P4. PCR testing of various samples obtained antemortem and postmortem were positive for Rhizopus microsporus/azygosporus DNA, indicating systemic mucormycosis originating from the gastrointestinal (P3) and respiratory tract (P4), respectively. Rhizopus DNA could also be detected in a retrospectively analyzed EDTA blood sample of P4, but all serum samples were negative. However, other studies reported better results when testing serum samples of patients at risk of mucormycosis . Patients with severe, community-acquired viral pneumonia, i.e., influenza, or receiving long-term ventilation are as yet not known to be at risk of superinfections with mucorales . Mucormycosis is typically a rare infection described in severely immunocompromised patients, suffering from hematological malignancies, solid tumors, ketoacidosis due to diabetes mellitus, transplantation, treatment with steroids and myelosuppressive agents, iron chelator therapy, and traumatic skin and soft tissue infections . P3 and P4 received hydrocortisone treatment, equivalent to 60 mg prednisolone/day, which is in the range of the recommended Dexamethasone dose for severe COVID 19 patients, before the first evidence of fungal infection was detected. However, this does not exclude that the infection preceded the administration of steroids. Steroids might have been a co-factor in the development of mucormycosis, although in general this is not a very frequent complication of steroid therapy. Additional factors may have contributed. For instance, P3 did not produce SARS-CoV2 antibodies during the disease. Therefore, an unknown immunodeficiency might be present in this patient. Furthermore, the immune modulation described in severe COVID-19  might specifically support the development of mucormycosis in long-term ventilated SARS-CoV-2 patients. There is one additional study describing superinfections of COVID-19 patients with mucormycetes. The retrospective study included 257 patients in the Chinese province of Jiangsu from January 22 to February 2 and reported infections with Mucor spp. in 6 cases . However, these patients suffered from mild to moderate disease and there was no histologic confirmation. In addition, one other autopsy study showed mucor infection in 1/10 investigated cases. This patient was intubated for 22 days but other clinical information could not be obtained from this report. Nevertheless, this shows that mycotic and specifically mucor infections are not a local phenomenon and may particular gain importance during the course of the disease .
Histopathology of two patients (P5, P7) revealed locally restricted fungal infections, which might be superinfections due to previous damage to the mucosal or skin barrier. In P5, PCR found molds (Aspergillus fumigatus and Mucor species) consistent with the microscopic detection of hyphae. Yeast-like elements in the skin sample of P7 could not be specified with fungal PCRs; however, the patient was colonized with Candida parapsilosis in the respiratory and urogenital tract. Candida parapsilosis is prone to colonize the human skin ; therefore, this yeast might have superinfected the preexisting skin ulcer.
Histopathological and mycological findings in P1, P2, and P6 indicated invasive pulmonary aspergillosis, which is an infectious complication in critically ill patients and known as superinfection in ventilated influenza patients [15, 26], and has already been described in patients with severe SARS-CoV-2 pneumonia .
P8 had fungal bloodstream infection with Candida albicans grown in blood cultures obtained 1 day before the patient died. Candidemia is a well-known nosocomial infection in intensive care patients,  often due to intravascular catheters infection or damaged mucous membranes of the gastrointestinal tract.
Systemic fungal infections in COVID-19 patients have been described in individual cases [11, 24], but we were able to detect them in 6 of 8 investigated patients in this study. In line with Gangneux , we recommend considering invasive mycosis including mucormycosis as a complication in long-term ventilated SARS-CoV-2 patients.
Additionally, we could find signs of a hemophagocytic lymphohistiocytosis (HLH) with increased phagocytosis of bone marrow cells by macrophages in bone marrow specimens of all but one patient (Table 1), similar to the findings of Hanley et al. . In P6 and P7, this finding was in line with the clinical presentation as HLH was already evident at life time (fulfillment of 5/6 tested criteria for HLH, according to the HLH-2004 guidelines ). In P3, P4, and P5, the clinical presentation was not suspicious for HLH, while in P1, P2, and P8, HLH was possible (3/5, 2/5, and 4/5 positive HLH-2004-guideline criteria). The latter group was designated as “maybe HLH” in Table 1, because additional testing of the remaining three criteria could have allowed for fulfillment of the criteria and the clinical presentation was compatible with HLH. The higher incidence of macrophage activation syndrome in patients with COVID-19-infection has already been described [5, 8, 18]. It is regarded as being a consequence of the hyperinflammatory state  of the patients which is confirmed by our study as 7/8 patients showed hyperinflammation according to the IL-6 levels (Table 1).
Conclusively, the results of this study underline the importance of performing autopsies, especially in gaining better insights into unknown or poorly defined diseases like SARS-COV2. Therefore, the close interaction between the clinicians and the pathologists as well as microbiologists is essential for the accurate assessment of the findings obtained during the autopsy and the possible resulting adjustment of diagnostics and therapy in the future. All our patients had comorbidities, but only one patient was ill enough that she could possibly have died even without the SARS-CoV-2 infection as a result chronic liver failure with acute liver decompensation in the background of liver cirrhosis. The findings of macrophage activation syndrome together with fungal superinfections suggest that long-term treated COVID-19 patients suffer from severe immunopathologies that favor fungal (super)-infections, particularly Mucor and Aspergillus species. Finally, we have demonstrated that clinically undetected fungal infections are a major cause of death in COVID-19 patients after long-term treatment. We believe that this important finding warrants alterations in current microbiological screening strategies that should also include the implementation of suitable diagnostic methods already in local hospitals, to allow for an early diagnosis and subsequent specific and thus hopefully successful antifungal treatment.