Host and microbiome features of secondary infections in lethal covid-19

Summary Secondary infections contribute significantly to covid-19 mortality but driving factors remain poorly understood. Autopsies of 20 covid-19 cases and 14 controls from the first pandemic wave complemented with microbial cultivation and RNA-seq from lung tissues enabled description of major organ pathologies and specification of secondary infections. Lethal covid-19 segregated into two main death causes with either dominant diffuse alveolar damage (DAD) or secondary pneumonias. The lung microbiome in covid-19 showed a reduced biodiversity and increased prototypical bacterial and fungal pathogens in cases of secondary pneumonias. RNA-seq distinctly mirrored death causes and stratified DAD cases into subgroups with differing cellular compositions identifying myeloid cells, macrophages and complement C1q as strong separating factors suggesting a pathophysiological link. Together with a prominent induction of inhibitory immune-checkpoints our study highlights profound alterations of the lung immunity in covid-19 wherein a reduced antimicrobial defense likely drives development of secondary infections on top of SARS-CoV-2 infection.


INTRODUCTION
Covid-19 originates from infection of the upper respiratory tract with SARS-CoV-2, which can progress into severe acute lung injury (ALI). Based on the tissue-typic expression of the viral host-entry receptor ACE2 and certain proteases (e.g. TMPRSS2) facilitating cellular uptake, also other organs like the kidney could be directly infected (Hou et al., 2020b). In addition, severe disturbances of immune and coagulation systems during covid-19 lead to a multifaceted disease with variable multi-organ damages (Ramlall et al., 2020). A consistent finding in severe covid-19 is initial immune hyperactivation (called ''cytokine storm'') leading to subsequent immune exhaustion, a phenomenon also known in other severe infections (Blanco-Melo et al., 2020;Lowery et al., 2021;Roquilly et al., 2020;Wang et al., 2021). Consequently, secondary infections which develop on top of SARS-CoV-2 infection contribute significantly to covid-19 mortality similar to severe influenza (Buehler et al., 2021). Curiously, the pathophysiology leading to the development of secondary lung infections is generally poorly understood. We performed an autopsy study of 20 consecutive covid-19 patients who died during the first pandemic wave. Full autopsies were performed and various specimen types were collected for tissue-based investigations, molecular measures including deep sequencing and cultivation of virus and other microbes. Integrating all information gained from this ''holistic'' autopsy approach allowed us to gain a deeper understanding of host and microbial factors contributing to secondary infections as a major sequel of lethal covid-19. iScience Article within the same time period were included for comparisons (Tables S1 and S2). Patients were tested for SARS-CoV-2 tissue distributions by quantitative RT-PCR (target: nucleocapsid-gene) and most positive samples with the highest viral loads originated from the respiratory tract, followed by myocardium, liver, kidney and pleural effusions. Other tissues and body liquids were positive only in single cases or tested overall negative ( Figure 1A). Notably, deep RNA-seq generated from lung tissues revealed SARS-CoV-2 transcripts in each covid-19 case, including the four qRT-PCR negative ones, showing increased sensitivity of deep transcriptomic analysis (127 G 29 million reads were generated per sample on average; Figure 1B). The viral genome was entirely captured by RNA-seq yielding more plus-strand reads (mean: 37.89 reads per million; range: 0.02-131,165.41) than minus-strand reads (mean: 1.81 reads per million; range: 0-484.81; Figure 1C). In addition, 11 SARS-CoV-2 strains could be cultivated from post-mortem lung tissues using Vero cells (Table S3). Successful virus cultivation significantly correlated with abundance of SARS-CoV-2 reads ( Figures 1D and S2).
SARS-CoV-2 genotyping facilitated by PCR and sequencing directly from autopsy specimens yielded 14 complete viral genomes (Table S4). Nine different sequence variants were detected showing up to 12 nucleotide changes compared to the reference (SARS-CoV-2 Wuhan-Hu-1; total genome size 29,903 bp; Table S5). Strains corresponded to the pangolin lineages B. 1.22, B1.5, B and B.1.160.1, respectively (clades 19A and 20A), representing the dominant genotypes of the first pandemic wave ( Figure 1E). Twelve strains harbored a D614G mutation in the spike (S) protein, which leads to increased viral transmissibility and, therefore, this genotype superseded the wild-type strain already early in the pandemic (Hou et al., 2020a). We identified also 2 viral clusters in our cohort, cluster 1 (case 3, 4, 6, 8, and 9) and cluster 2 (case 1, 2, and 5), respectively ( Figure 1F). Notably, cases 6, 8 and 9 from cluster 1 originated from the same residential care home and all cases from cluster 2 stayed in the same hospital ward before covid-19. Thus, it is very likely that these individuals were infected from the same sources and/or transmission occurred.

Major organ pathologies and death causes
Lungs showed the dominant pathologies in relation to covid-19, only one case (#1) presented with acute myocardial infarction as the ascribed death cause. Diffuse alveolar damage (DAD), the histopathological representation of ALI, in a patchy distribution and often prevalent in multiple lung segments was the major finding in 11 cases. Early exudative stages and later organizing stages of DAD were found within the same patient together, often adjacent to nearly normal or less affected parenchyma indicating ongoing tissue damage (Figures 2A and S3-S5). Also, a significant positive correlation of SARS-CoV-2 loads from nasopharyngeal tissues to lungs was found ( Figure 2B) likely suggesting active seeding of infectious particles from the upper respiratory tract via micro-aspiration (Hou et al., 2020b). We extensively assessed microscopic lung features (see STAR Methods for details of histopathological scoring) to specify and grade the severity of lesions and also to capture the heterogeneity of different lung pathologies. Features greatly varied between cases and the majority of patterns did not correlate with disease duration (defined as the interval between the first SARS-CoV-2 positive PCR and death) or viral loads ( Figure 2C). Although early DAD features (intra-alveolar edema, hyaline membranes) correlated positively with shorter disease duration and late features (fibrosis) increased with disease duration, early and late features were often intermixed showing no inverse correlation ( Figure 2D) corroborating findings also by other studies (Borczuk et al., 2020). Of importance, the clearest discriminating feature of cases was the presence of neutrophilic granulocytes, indicative of secondary infections (i.e., ''pneumonia''), in comparison to DAD. Three cases showed DAD superimposed with acute inflammation and 5 cases showed mainly pneumonia as the dominant pathology, wherein DAD was only focally visible overlaid with dense inflammation. Altogether, 16 cases  iScience Article showed neutrophilic granulocytes present in bronchi, bronchioli, or alveoli suggestive of secondary infections. Thus, lung histopathology in lethal covid-19 could be stratified into DAD, DAD superimposed with pneumonia and dominating pneumonia ( Figures 2E and S3). Pneumonia cases showed increased IL-6 levels compared with pure DAD cases, whereas CRP levels were weaker discriminators ( Figure 2F). It is noteworthy that neither disease duration nor viral loads correlated with the presence of neutrophils, nor did any other clinical parameter (Figures 2G and S6). Other organs showed features of preexisting comorbidities including arteriosclerosis, hypertension and diabetes, especially in the kidneys, wherein SARS-CoV-2 could be detected in tubular epithelia by positive immunohistochemistry ( Figure S7). Heart and liver specimens revealed no clear evidence of direct SARS-CoV-2 carriage or features of myocarditis or hepatitis (Figures S8 and S9). A detailed summary of organ histopathologies is given in the supplementary material (Table S6).

Lung microbiome alterations and secondary infections in lethal covid-19
The lung microbiome is altered in DAD and thought to be a relevant factor for the development of secondary infections (Dickson et al., 2016;Luyt et al., 2020). RNA-seq from lung tissues was screened for microbial sequences and bacterial (16S rRNA gene) and fungal (internal transcribed spacer, ITS) marker genes were amplified to additionally specify microbial changes. On average 6573.33 G 2552.32 (MWGSD) reads per million (rpm) per sample were not human in RNA-seq and likely of microbial origin, of those 2.02 G 4.00% and 0.03 G 0.05% could be clearly annotated to specific microbes with different microbial annotation pipelines ( Figure 3A). Excluding SARS-CoV-2 reads, which were the dominant microbial component in several cases (range: 0.01-131218.36 rpm), bacterial sequences were dominant, significantly increased in covid-19 cases with pneumonia compared to DAD and controls. Fungal and viral sequences other than SARS-CoV-2 were also significantly enriched in covid-19 cases with pneumonia ( Figure 3B). Number of bacterial reads significantly correlated with neutrophil scores suggesting that their presence is a sign of secondary infections, however, the post-mortem interval did not, precluding a strong influence of post-mortal bacterial overgrowth in our investigation ( Figure 3C). Bacteria are assumed to be the dominant microbiome component in lungs (Huffnagle et al., 2017). Analysis based on the bacterial 16S rRNA gene marker showed that richness was significantly decreased in the DAD and pneumonia cases of covid-19 compared to controls indicating an overall reduced biodiversity ( Figure 3D). In contrast, evenness was significantly decreased in the pneumonia group of covid-19 only, suggesting a dominance of certain taxa, possibly representing the agents of secondary infections (pairwise Kruskal-Wallis; *p < 0.05, **p < 0.005). Principal component analysis (PCA) clearly separated controls from covid-19 cases with DAD and cases with pneumonia indicating significantly different bacterial community compositions ( Figure 3E). Lung tissues were also cultured for bacteria and fungi and both-covid-19 cases and controls-yielded cultivable microorganisms but in different quantities and taxonomic constellations (Table S7).
Finally, we integrated RNA-seq, 16S, ITS, and culture data to define dominant pathogens, most likely representing the agents of secondary infections and to account for the different samples used for microbial identifications in the light of the patchy disease representations likely impacting the microbial repertoires ( Figure S10 and Table S8). Dominant pathogens were defined if they were dominant in the RNA and/or DNA data (representing >10% of microbial reads excluding SARS-CoV-2) and if they also yielded a reasonable culture growth (R10 4 cfu/mL). Dominant taxa were typical agents of pulmonary secondary infections like Staphylococcus aureus, Enterococcus faecium, or Klebsiella pneumoniae, as well as fungi like Candida spp. or the mold Rhizopus microsporus identified in one case (#18; (Zurl et al., 2021)). Often multiple pathogens were found simultaneously indicating polymicrobial infections (e.g., in case #16 wherein K. pneumonia, S. aureus and Candida glabrata were cultivated in reasonable amounts and were also captured by RNA-seq). In addition, 5 covid-19 cases yielded transcripts of Epstein-Barr virus (EBV), which were also iScience Article detectable by RNA in-situ hybridization of lung tissues but not in controls ( Figure 3H). EBV often emerges because of endogenous reactivation in the context of impaired immunity (Tangye et al., 2017). Control cases yielded microbial sequences and cultivable microbes in lower quantity and they often belonged to known contaminants like Lactobacillus sp. or Propionibacterium sp. (Table S8). In summary, the lung microbiome in covid-19 shows a reduced taxonomic richness but harbors a diverse spectrum of bacterial and fungal pathogens typically associated with secondary lung infections. Prominent pathogens like S. aureus, Klebsiella or Candida spp. are also known agents of secondary infection in influenza, SARS, and MERS (Klein et al., 2016;Morens et al., 2008). Notably, secondary infections were rarely detected ante-mortem in our cohort (Table S8). The presence of poly-microbial infections and the relatively high proportion of EBV positivity suggest an overall impaired immunity in covid-19 lungs.

The lung metatranscriptome mirrors the major death categories DAD and pneumonia
Deep RNA-seq of lung tissue revealed 4,547 differentially expressed genes between covid-19 cases and controls (adj. p < 0.05). Hierarchical clustering indicated depleted (cluster 1) or enriched (cluster 2) genes in covid-19 compared to controls ( Figure 4A). Pathway analysis indicated impaired central cellular functions within mRNA metabolism, post-translational protein modification, the respiratory chain, VEGFA signaling and extracellular matrix organization in covid-19. Enriched pathways consisted mainly of innate and adaptive immune functions, neutrophil degranulation, cytokine signaling as well as complement activation ( Figure 4B). Overall, these data confirm profound and complex transcriptional alterations in covid-19 lung tissue (Delorey et al., 2021;Liao et al., 2020;Wu et al., 2020). Unsupervised principal components analysis (PCA) of differentially expressed genes clearly separated covid-19 samples on principal component 1 (PC1) from controls but also clearly separated pneumonia samples from pure DAD cases ( Figure 4C). Comparison of differentially expressed genes between these major death categories indicated that the major discriminator from controls was DAD showing 3862 unique differentially expressed genes (adj. P < 0.05 and abs. LFC >= 0.58) followed by pneumonia with 1673 unique differentially expressed genes ( Figure 4D). DAD and pneumonia differed by only 226 differentially expressed genes. Notably, among the top 50 differential expressed genes enriched in pneumonia cases several macrophage markers were evident, including the receptor ADGRE1 (murine homolog F4/80) as top-hit, the interleukin-1 receptor-associated kinase-like 2 (IRAK2) or PSTPIP2, which is involved in macrophage polarization ( Figure 4E). Thus, macrophages seem to be implicated in covid-19 secondary infections. In summary, deep transcriptomic analyses specified multiple dysregulated processes in covid-19, including vascular and coagulation systems, connective tissue remodeling as well as activated immunity and complement (Nie et al., 2021). Similar to histopathology, the major discriminator from controls based on gene expression was DAD followed by pneumonia likely mirroring the development of secondary infections on top of ALI caused by the virus.

Cellular deconvolution subgroups covid-19 lung pathology
Cellular compositions were inferred from RNA-seq by using xCell (Aran et al., 2017). Hierarchical clustering based on cellular compositions clearly separated samples into four distinct groups. Group 1 (''control'') consisted only of control cases and was related to group 2 (''DAD1'') consisting of covid-19 cases with pure DAD (in addition to one control case #22). Group 3 (''DAD2'') also contained DAD cases, including one sample with the histological category DAD and secondary pneumonia. This group was related to group 4 (''pneumonia'') composed of all pneumonia cases, in addition to 3 DAD cases and the two remaining DAD cases with secondary pneumonia ( Figure 5A). It is noteworthy that neither disease duration nor early versus late DAD histological features or SARS-CoV-2 loads significantly correlated with a specific grouping ( Figure S11). Cell types discriminating these groups showed a specific assembly ( Figure 5B).  iScience Article Cluster 1 consisted mainly of vascular and stromal cell types like endothelial cells, pericytes and fibroblasts, enriched in ''DAD1'' and ''DAD2''. Top enriched genes in this cluster were certain collagen genes, the vascular transcription factor sox 18, the basement membrane protein ladinin-1 or the endothelial protein stabilin-1 ( Figure 5C). Cell types of cluster 2 consisted mainly of structural and stromal cells, in addition to certain immune and blood cell types and they were overall reduced in covid-19. Top down-regulated genes included the extracellular matrix proteins sparc-like 1, multimerin-1 and plastin-3 or the cell adhesion molecule p-selectin important for the recruitment of leukocytes typically prevalent on activated endothelial cells and platelets ( Figure 5C). Together these alterations highlight the vascular and connective tissue changes emerging during DAD development ( Figure 2B) (Ackermann et al., 2020;Hughes and Beasley, 2017). Cell types of cluster 3, which were dominantly induced in ''DAD2'' and to a lesser extent in ''pneumonia'' showed enrichment of myeloid (cluster 3a) and epithelial cell types (cluster 3b). Top induced genes in cluster 3a were the myeloid cell specific genes siglec-1, CD11c and complement factor C1q. Top induced genes in cluster 3b consisted of keratin 6A, collagen XVII, the tyrosine kinase signaling protein cbl-c and beta-1,3-N-acetylglucosaminyltransferase 3 (b3gnt3), typically expressed in epithelia and also involved in lymphocyte trafficking and homing. Cell types in cluster 4, strongly increased in ''pneumonia'' consisted of different leukocyte classes including B-, T-cells and (neutrophilic) granulocytes. Top induced genes consisted of the interleukin 8 receptor genes cxcr1 and cxcr2, the chemokine receptor type 2 (CCR2), CEA-CAM3, CD22 (B cell marker), and the cell surface receptors TREML2 and FCGR3B.
In summary, cellular deconvolution clearly sub-stratified the major categories DAD and pneumonia of covid-19 lung pathology. Noteworthy, DAD subclustered into two different groups, one showing mainly induction of vascular and stromal cell elements (''DAD1''), the other dominant induction of genes related to myeloid and epithelial cells (''DAD2''), and this subgroup showed more commonalities with the pneumonia group.

Macrophages complement c1q and immune impairment in covid-19 lungs
Myeloid cells including macrophages play a central role in the pathogenesis of DAD Fan and Fan, 2018;Huang et al., 2018), and bronchalveolar lavage fluids (BALFs) of patients with severe covid-19 reveal high proportions of macrophages Wang et al., 2020). We confirmed significantly increased macrophages in covid-19 lungs by CD163 immunohistochemistry, depicting a M2-type macrophage marker ( Figure 6A), corroborating a recent proteomic study wherein CD163 was found among the most induced proteins in lungs and spleens derived from covid-19 autopsies (Nie et al., 2021). Deconvolution indicated both M1and M2-type macrophages significantly enriched predominantly in ''DAD2'' whereas monocytes were mainly induced in the ''pneumonia'' group ( Figure 6B). Increased CD163 positive macrophages gathering around virus positive cells were recently shown also in a macaque model of SARS-CoV infection, indicating that infected pneumocytes may lead to macrophage recruitment in coronavirus infections (Liu et al., 2019).
Among the most discriminative genes between DAD subtypes we found complement factor C1q dominantly induced in ''DAD2'' ( Figure 5C). Complement activation is implicated in DAD pathogenesis and linked to severe covid-19 (Holter et al., 2020;Java et al., 2020;Perico et al., 2021). Other complement factors showed no discriminative expression pattern between pathological subgroups in our cohort, except certain complement receptors and properdin mainly induced in pneumonia cases ( Figure 6C). C1q levels did not correlate with survival times of patients ( Figure S12). Western blots generated from extracts of lungs confirmed significantly increased C1q protein ( Figure 6D). A major source of C1q are macrophages corroborated also by a recent single-cell transcriptomic analysis of covid-19 lungs ( Figure S13) (Lu et al., 2008;Xu et al., 2020) suggesting a strong connection between macrophages and complement C1q in covid-19 (Carvelli et al., 2020). Immunohistochemical analysis of lung tissues with a C1q specific antibody showed staining of the vasculature, the interstitial and alveolar space but also of alveolar cells including macrophages and pneumocytes, indicating a multifaceted deposition of C1q in the context of covid-19 in our series ( Figure 6E).
C1q is the initiating component of the classical complement cascade but exhibits also immune regulatory functions. It induces the development of pro-resolving M2 type macrophages and is involved in the iScience Article clearance of apoptotic and necrotic cells, which are highly increased in covid-19 lungs (Bohlson et al., 2014;Li et al., 2020b;Mulay et al., 2021). In this process C1q binds to cellular break-down products and is subsequently recognized by phagocyte receptors like the leukocyte-associated immunoglobulin-like receptor 1 (LAIR-1; syn.: CD305) conferring uptake and triggering a tolerogenic state in the phagocyte (Son et al., 2015;Thielens et al., 2017). As shown by a recent single-cell transcriptomics analysis of covid-19 lungs, LAIR-1 is mainly present in macrophages ( Figure S14) (Delorey et al., 2021). LAIR-1 together with LILRB4 (leukocyte immunoglobulin-like receptor subfamily B member 4; syn.: ILT3) belong to immunoglobulinlike receptors recognizing collagen domains such as present in C1q, thereby inhibiting immune activation (Lebbink et al., 2006;Son et al., 2012). RNA-seq confirmed significant induction of LAIR-1 and LILRB4 dominantly in ''DAD2'' followed by ''pneumonia'' ( Figure 7A). Expressions of all 3 C1q polypeptide chains (A, B, & C) significantly correlated with LAIR-1 and LILRB4 expression but not with induced collagens (Figures 7B and S15). This might suggest a functional link between C1q and the immune inhibitory receptors LAIR-1 and LILRB4.
Because the development of secondary infections is likely driven by local immune-impairment we screened for other anti-inflammatory markers. TGF-b1 is a key factor in the development and healing response of ALI and also implicated in covid-19 lung pathology (Peters et al., 2014;Vaz dePaula et al., 2021). TGF-b1 transcription was significantly increased in covid-19, showing a huge variability, however, TGF-b1 protein measured by immunohistochemistry and western blotting showed no significant induction compared to controls ( Figures 7C and S16). Because several immune and non-immune cell types are able to produce TGF-b1, the observed variability in expression might reflect the temporal heterogeneity of lung pathologies in our cohort. Induction of certain inhibitory immune-checkpoints is reported in covid-19 (Bobcakova et al., 2021;Diao et al., 2020;Files et al., 2021;Jeannet et al., 2020;Li et al., 2020a). We confirmed transcriptional induction of several inhibitory immune-checkpoints in covid-19 lungs (LAG3, PDCD1, ADORA2A, VSIR, CTLA4, SIRPA, LAIR1, SIGLEC9, LILRB4, SIGLEC7, and HAVCR2), whereas some showed no induction (CD276, SP140, IDO1, KLRG1, and CD274) or were even reduced (GGT1 and CD80) compared with controls ( Figures 7D and S17). Of note, the top induced inhibitory immune-checkpoint was found to be LAG3 (lymphocyte-activation gene 3; syn.: CD223), dominantly induced in ''DAD2'' and ''pneumonia'' based on RNA-seq, which was also confirmed by immunohistochemistry wherein mainly lymphocytes showed strong staining signals ( Figures 7E and S18). LAG3 was recently described as a major increased factor in a plasma proteomic study of severe covid-19 (Filbin et al., 2021). During immune exhaustion multiple inhibitory receptors act often in synergy amplifying immune impairment, like LAG3 and PD-1 co-induced during chronic viral infections (Blackburn et al., 2009). We confirmed synergistic induction of several inhibitory immune-checkpoints in covid-19 lungs, which showed a different costimulatory pattern compared to controls ( Figure 7F). Of interest, co-expression patterns discriminated cases with high viral loads (KLRG1, CTLA4, SP140, CD274, IDO1, LAG3, PDCD1, HAVCR2, and CD80) from samples with pneumonia (CD276, TGFB1, ADORA2A, LILRB4, LAIR1, VSIR, SIRPA, SIGLEC7, and SIGLEC9) suggesting a divergent pattern of induction of inhibitory immune-checkpoints during the course of covid-19 lung pathology ( Figure 7G). In summary, these data highlight that multiple pillars of immune impairment act in severe covid-19, leading to a reduced antimicrobial defense in lungs driving the development of secondary infections. The molecular dissection of cell types and immune inhibitory signals might enable the development of specific measures counteracting this potentially lethal complication.

DISCUSSION
We performed a systematic autopsy study of 20 consecutive covid-19 cases and 14 controls to gain unbiased information about lethal disease courses from the early pandemic. Integration of autopsy, cultivation and deep sequencing provided important clues about host and microbial factors involved in the development of secondary infections as a major sequel of lethal covid-19. Thus, our study might serve iScience Article as a blue-print for a ''holistic'' autopsy approach tempting to gain relevant pathophysiological insights from a newly emerging disease (Layne et al., 2022). Viral genotyping facilitated directly from autopsy material provided epidemiologic clues about transmission and captured already early events of viral genetic adaptation. Of note, a significant proportion of corpses yielded cultivable SARS-CoV-2 indicating that autopsy might facilitate virus spread and that special safety requirements should be applied during post-mortem examinations of covid-19 patients (Loibner et al., 2021). Our investigation showed that covid-19 lung pathology is multifaceted and that a major discriminator of lethal courses is DAD and the presence of secondary infections. This was evident by histology but also mirrored by the deep transcriptomic analysis and microbiology. Secondary infections are reported to develop in up to 42% of patients with covid-19 (Buehler et al., 2021). Notably, DAD caused by the virus itself and secondary infections are chronologically divergent and provoke overtly different host reactions. It is also noteworthy that SARS-CoV-2 infection alone might not trigger prominent neutrophil recruitment to the lung at all and neutrophil signatures found in recent covid-19 studies might likely already indicate secondary infections Melms et al., 2021;Nie et al., 2021;Wauters et al., 2021;Xu et al., 2020). Thus, it is important to seek for a proper pre-classification of tissue samples based on histology to omit wrong conclusions in molecular down-stream analyses.
The resident lung microbiome is a relevant factor in the pathogenesis of lung infections and reported to be altered in sepsis and DAD (Dickson et al., 2016;Luyt et al., 2020). Secondary lung infections are also complicating influenza, SARS and MERS, wherein bacteria like S. aureus or Klebsiella spp. and fungi such as Candida or Aspergillus spp. are found (Klein et al., 2016;Morens et al., 2008). Such microbial agents were also present in our cohort. Curiously, the mechanistic understanding why secondary infections develop on top of viral infections is still limited. We could not identify any associated clinical parameter clearly correlated with secondary infections but showed that lung immunity is impaired in covid-19, which might drive these infections. This finding was also underscored by the presence of polymicrobial infections and EBV indicative for a general decreased immunity (Tangye et al., 2017). Typical comorbidities of covid-19, like chronic kidney disease or diabetes, are already signified by a lowered immunity increasing the infection risk (Carey et al., 2018; Syed-Ahmed and Narayanan, 2019). Moreover, ICU admission and mechanical ventilation are established risk factors for the development of pneumonias (Wu et al., 2019). Thus, secondary lung infections in covid-19 could originate from different sources including direct immune challenge by SARS-CoV-2, the underlying condition and medical interventions (Callender et al., 2020). Curiously, the majority of our patients received antibiotics, which failed to effectively protect against secondary infections. This circumstance might have been influenced by underdeveloped antibiotic regimens or ineffective substances (e.g., chloroquine; (Axfors et al., 2021)) given during the early pandemic but also from concomitant therapeutic immunosuppression (e.g., corticosteroids) altogether modulating the infection risk on an individual base. Notably, ICU admission and intubation, however, were generally associated with lower neutrophil scores. Because all intubated patients received antibiotics in our series, this therapy might have delayed secondary pneumonia development.
Severe cases of covid-19 typically present with high inflammatory markers complicating the distinction between severe courses with pure DAD or bacterial or fungal secondary infections. As in our cohort, hospitalized covid-19 patients are empirically treated with broad-spectrum antibiotics but efficacy of this therapy is still debated (Sieswerda et al., 2021). Because patients frequently need prolonged hospitalizations and respiratory support, unnecessary antibiotic therapy also likely increases the risk of hospital-acquired pneumonias caused by resistant bacteria. On a population level increased antibiotic usage likely leads to rising antimicrobial resistance further complicating management of the pandemic (Chong et al., 2021;Sopirala, 2021). Noteworthy, antibiotic challenge of the resident lung microbiome might also impact secondary infection development. In analogy to the GI tract wherein depletion of the resident microbiome potentially provokes overgrowth of pathogens (e.g., in C. difficile colitis; (Chang et al., 2008)), similar mechanisms might happen in the respiratory tract. Notably, we detected a significant reduced biodiversity in  iScience Article lungs of our covid-19 patients which was likely influenced by antibiotics. Single reports indicate a rise of secondary infections in the second pandemic wave compared to the initial phase indicating a fluctuating picture of secondary infections (Fortarezza et al., 2022;Hedberg et al., 2022). This development might have been influenced by several factors including a changed epidemiology of patients (different comorbidities and ages), seasonal effects, which seem to be particularly important for the development of fungal pneumonias but also changes in clinical practice (Ayzac et al., 2016;Group et al., 2021).
Immune exhaustion seems to follow the systemic immune hyperactivation in severe covid-19 and myeloid cells, which are important for the recognition of virus infected cells, are key to initiate the proinflammatory response (McGonagle et al., 2020;Merad and Martin, 2020). Recent single-cell transcriptomic studies of covid-19 patients identified myeloid cells as a major induced cell type in BALF specimens with high proportions of proinflammatory macrophages Melms et al., 2021). Generally, M1-type macrophages dominate early DAD, whereas later DAD stages show increased M2-types involved in tissue repair with immunosuppressive features (Huang et al., 2018). Thus, later (organizing) phases of DAD might be specifically prone to acquire secondary infections. Respiratory failure in covid-19 is linked to strong complement activation (Holter et al., 2020;Java et al., 2020;Messner et al., 2020), which likely occurs when the disease progresses (Nienhold et al., 2020). Extensive deposition of complement factors, including C1q, in vessels and epithelial cells of lungs and skin was reported in covid-19 (Macor et al., 2021;Magro et al., 2020). Notably, the SARS-CoV-2 spike-protein might directly activate complement via the alternative pathway (Yu et al., 2020). Complement in covid-19 is currently discussed mainly in the context of endothelial injury and fibrin-clot formation (Perico et al., 2021). Our study suggests another pathophysiological role, wherein C1q and macrophages might perpetuate immune impairment. Immune complexes formed by viral antigens and antibodies can activate factor C1 as shown in SARS-CoV infection (Yang et al., 2005). C1q is involved in the clearance of apoptotic and necrotic cells by phagocytes, a process termed efferocytosis (Doran et al., 2020). Apoptosis and necrosis are prominent in covid-19 lungs (Filbin et al., 2021;Li et al., 2020b;Mulay et al., 2021). During efferocytosis suppression of overwhelming inflammation is important and phagocytes involved in this process are producing anti-inflammatory cytokines. Therefore, C1q binds to molecules released from apoptotic and necrotic cells (e.g., phosphatidylserine, nucleic acids, etc.) and these complexes are recognized by receptors present on phagocytes, like LAIR-1, conferring uptake and inducing a tolerogenic state (Bohlson et al., 2014;Lu et al., 2008;Son et al., 2015;Thielens et al., 2017). Noteworthy, binding of C1q to LAIR-1 on plasmacytoid DCs restricts the production of type I interferons impairing antiviral defense, which also occurs in covid-19 (Son et al., 2012(Son et al., , 2015. The ''DAD2'' subtype in our study shows increased macrophages, C1q and LAIR-1 and might therefore represent cases with a lowered immune tone prone for the development of secondary infections. Overall the progression of early (exudative) DAD into late (fibrotic) DAD indicates healing of ALI characterized by significant connective tissue remodeling and a reduced inflammatory tone (Margaroli et al., 2021;Matthay et al., 2019). Immune suppressive factors such as TGF-b1 are known to be involved in this process and also LAIR-1 and LILRB4 recognizing collagens or collagen-like proteins might act anti-inflammatory during this disease phase (Fernandez and Eickelberg, 2012;Fouet et al., 2021;Paavola et al., 2021;Son et al., 2012;Tomic et al., 2021). Moreover, the synergistic induction of several tolerogenic factors including inhibitory immune-checkpoints (Bobcakova et al., 2021;Diao et al., 2020;Filbin et al., 2021;Files et al., 2021;Hadjadj et al., 2020;Jeannet et al., 2020;Li et al., 2020a) and increased (apoptotic) cell death of immune-cells (Cizmecioglu et al., 2021;Feng et al., 2020) altogether perpetuate immune failure in covid-19.

Limitations of the study
The limitations of our descriptive study are that causalities cannot be directly inferred and that the relatively small cohort cannot reveal the entire picture of severe covid-19 and associated secondary infections.
Varying clinical courses and different comorbidities might also have influenced our findings. In addition, treatment of covid-19 has changed since the early pandemic, thus, current severe courses and developing sequels might also have changed. We also cannot be sure whether the two described forms of DAD might represent just a spectrum of pathophysiological states or are specific pathotypes. Moreover, post-mortem effects like RNA degradation might have introduced additional noise in our investigation. Nevertheless, we found autopsy complemented with microbiology and molecular measures as a powerful tool to gain relevant clues about covid-19 pathophysiology. Of importance, there exists an obvious knowledge gap in the understanding of the molecular mechanisms driving the development of secondary infections on top of in viral lung diseases. This should initiate further studies to understand the molecular pathways in more detail and to unravel chronological phases of immuno-suppression which could also lead to development of

RNA extraction
Samples consisted of swabs (eSwab, Copan), tissues and body fluids, the latter were collected with sterile syringes. Fresh tissues were sampled directly into Magna Lyser Green Beads tubes (Roche) pre-filled with 400mL lysis buffer.

SARS-CoV-2 quantitative RT-PCR
qRT-PCR for detection and quantification of SARS-CoV-2 in autopsy samples were performed as described (Corman et al., 2020). Briefly, primers, probes and 5 mL of RNA were added to 10 mL of SuperScript III One- Step RT-PCR System with Platinum Taq High Fidelity DNA Polymerase mastermix (ThermoFisher). PCR was performed on a Quantstudio 7 instrument (ThermoFisher) with the following cycling conditions: 55 C for 15 min, 95 C for 3 min; 45 cycles consisting of 95 C for 15 sec and 58 C for 30 sec. Amplification data was downloaded and processed using the qpcR package of the R project (https://www.r-project.org/). Amplification efficiency plots were visually inspected and Cp2D (cycle peak of second derivative) values were calculated for samples with valid amplification curves. Plots were generated with R using the reshape, tidyverse and ggplot packages. qRT-PCR of virus cultures employed primer sets recommended by the CDC detecting three different regions of the viral nucleocapsid and human RNAseP or GAPDH as control (https:// www.cdc.gov/coronavirus/2019-ncov/lab/rt-pcr-panel-primer-probes.html). PCR was performed with the SYBR Green PCR Mastermix (Applied Biosystems) on a Quantstudio 7 instrument (ThermoFisher) with the following cycling conditions: 25 C for 2 min, 50 C for 15 min, 95 C for 10 min, 45 cycles consisting of 95 C for 3 sec and 55 C for 30 sec.

Viral genome sequencing
PCR primers spanning the whole genome of SARS-CoV-2 were designed yielding in about 2kb amplicons (Table S9). 2.5 mL of RNA were used in three separate RT-PCR reactions as described above with oligonucleotide primers at 400 nM concentration with the following cycling conditions: 55 C for 15 min, 95 C for 3 min; 35 cycles consisting of 95 C for 15 sec and 57 C for 3 min; final extension at 72 C for 10 min. PCR products were combined and purified by incubation with 1.8X Ampure XP beads (Beckman Coulter) followed by two washes with 75% ethanol and elution in 30 mL water. Amplicons were fragmented to 150-250 bp length and Ion Torrent barcode and sequencing adapters were ligated to the fragments using the NEBNext Fast DNA Fragmentation & Library Prep Set for Ion Torren kit (New England Biolabs) according to the manufacturer's recommendations. Libraries were sequenced on an Ion Torrent S5XL instrument using a 540 Chip Kit and the 200 bp sequencing kit (ThermoFisher). Sequences were aligned to the SARS-CoV-2 reference genome (acc. no.:NC_045512.2) using TMAP (v5.10.11) and variants were called with the Torrent Variant Caller (v5.10-12). All called variants were visually inspected and consensus sequences of the viral genomes were generated with bcftools (v1.3.1). Consensus sequences were aligned using clustalw (v2.1) (Larkin et al., 2007), guide trees were visualized in figtree (v1.4.4) and final adjustments were made with Incscape (v0.92). SARS-CoV-2 genomes from our study were uploaded and analyzed with the GISAID SARS-CoV-2 (hCoV-19) databasewhich can be accessed via https://www.gisaid.org/ epiflu-applications/next-hcov-19-app/ (Hadfield et al., 2018;Sagulenko et al., 2018).

RNA sequencing
Libraries for RNA sequencing (RNA-seq) from lung tissues ( iScience Article loading control to determine protein abundance and band density was quantified and compared by using ImageJ.

QUANTIFICATION AND STATISTICAL ANALYSIS
GraphPad Prism and R were used for data analysis and imaging. All data are represented as means G SD if not otherwise specified. Statistical significance testing employed the Mann-Whitney and Kruskal-Wallis tests and p-values <0.05 were considered statistically significant. Correlation analyses employed Spearman and Pearson correlation. For differentially gene expression and gene enrichment analyses a False Discovery Rate (FDR) <0.05 was used. Permanova was used for statistical determination in the Principal Component Analysis (PCA). The n number is specific for the number of human subjects.