Early alveolar epithelial cell necrosis is a potential driver of COVID-19-induced acute respiratory distress syndrome

Summary Acute respiratory distress syndrome (ARDS) with COVID-19 is aggravated by hyperinflammatory responses even after the peak of the viral load has passed; however, its underlying mechanisms remain unclear. In the present study, analysis of the alveolar tissue injury markers and epithelial cell death markers in patients with COVID-19 revealed that COVID-19-induced ARDS was characterized by alveolar epithelial necrosis at an early disease stage. Serum levels of HMGB-1, one of the DAMPs released from necrotic cells, were also significantly elevated in these patients. Further analysis using a mouse model mimicking COVID-19-induced ARDS showed that the alveolar epithelial cell necrosis involved two forms of programmed necrosis, namely necroptosis, and pyroptosis. Finally, the neutralization of HMGB-1 attenuated alveolar tissue injury in the mouse model. Collectively, necrosis, including necroptosis and pyroptosis, is the predominant form of alveolar epithelial cell death at an early disease stage and subsequent release of DAMPs is a potential driver of COVID-19-induced ARDS.


INTRODUCTION
An infection with a novel strain of coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), causes coronavirus disease 2019 (COVID- 19) pneumonia. In the most severe cases, the disease progress to acute respiratory distress syndrome (ARDS), which is associated with severe alveolar tissue injury. 1,2 Interestingly, the disease severity is exacerbated by hyperinflammatory responses even after passing the peak of viral load. 3,4 However, the mechanisms that underlie disease aggravation in COVID-19-induced ARDS remain unclear. We and others have previously reported that alveolar epithelial injury at a very early disease stage is a hallmark of COVID-19-induced ARDS, 5,6 suggesting that alveolar epithelial injury may be a trigger of subsequent disease progression. Therefore, elucidating the detailed mechanisms by which alveolar epithelial injury occurs in COVID-19-induced ARDS may reveal a therapeutic target that prevents disease aggravation.
The alveolar epithelial injury in ARDS is characterized by cell death, which is divided into necrosis and apoptosis. Moreover, necrosis comprises not only accidental cell death but also several forms of programmed cell deaths. 7,8 Although previous studies have demonstrated that both alveolar epithelial necrosis and apoptosis are important for the pathogenesis of ARDS, 9 we have recently demonstrated that necrosis is the predominant form of alveolar epithelial cell death in lipopolysaccharide (LPS)-induced experimental ARDS. 10 In contrast to apoptosis, which does not elicit inflammation, necrosis causes the release of damage-associated molecular patterns (DAMPs) such as high-mobility group box (HMGB)-1 from dead cells. [11][12][13] Therefore, it is possible that the alveolar necrosis during early disease stages and subsequent release of DAMPs may drive disease progression in COVID-19-associated ARDS. [14][15][16][17]

OPEN ACCESS
We evaluated the circulating levels of three alveolar tissue injury markers: an alveolar epithelial injury marker (sRAGE) 19,20 and an endothelial injury marker (ANG-2), 21,22 along with an alveolar permeability indicator (SP-D). 23,24 All alveolar tissue injury markers levels after the admission were significantly higher in patients with ARDS versus healthy controls ( Figures 1A-1C). However, only sRAGE and SP-D levels of  Figure 1. Analysis of serum levels of alveolar tissue injury markers using enzyme-linked immunosorbent assays (ELISAs) (A-F) soluble receptors for advanced glycation end products (sRAGE), (B) angiopoietin (ANG)-2, and (C) surfactant protein (SP)-D levels in the serum of patients with COVID-19 with or without acute respiratory distress syndrome (ARDS) at admission (on the first or second hospital day), and healthy controls are shown. Bidaily temporal changes in (D) sRAGE, (E) ANG-2, and (F) SP-D in sera of COVID-19 patients with ARDS during the first 8 days after hospital admission are shown. In cases in which multiple values every 2 days were available, mean values were used. When only a single value was available, the value was used. (G) Days in which concentrations of each alveolar tissue injury marker peaked in COVID-19 patients with ARDS are shown. Values are presented as medians with interquartile ranges. *p < 0.05, **p < 0.01, ***p < 0.0001.

OPEN ACCESS
iScience 26, 105748, January 20, 2023 3 iScience Article patients with and without ARDS significantly differed ( Figures 1A-1C). In patients with ARDS, sRAGE levels were significantly elevated at admission and gradually decreased thereafter ( Figures 1D and 1G). Meanwhile, ANG-2 and SP-D levels peaked later ( Figures 1E-1G). Collectively, these results agree with prior work that demonstrated that severe alveolar epithelial cell injury at a very early disease stage is a hallmark of COVID-19-induced ARDS. 5,6 Epithelial necrosis markers and HMGB-1 are increased in COVID-19 acute respiratory distress syndrome Next, the levels of epithelial cell death markers were evaluated to elucidate the predominant form of alveolar epithelial cell death at the early disease stage of COVID-19-induced ARDS. The serum levels of CK18-M65 and -M30 antigens were measured to distinguish alveolar necrosis from apoptosis. CK18 is exclusively expressed in epithelial cells and is released upon cell death. The M65 antigen is an indicator of both epithelial cell necrosis and apoptosis. In contrast, the M30 antigen produced after caspase cleavage of CK18 is an indicator of apoptotic epithelial cell death. 25 Figure 2C). Moreover, we analyzed the CK18-M30/M65 ratio in BALF, which directly reflects the pulmonary pathology, from six patients with COVID-19 ARDS ( Figure S1). The characteristics of the included patients in the BALF analysis are shown in Table S1. The M30/M65 ratio in the BALF was 27.8% [IQR: 13.3-38.5] ( Figure S1), similar to that of serum samples. Collectively, these results indicate that alveolar epithelial cell death in COVID-19 ARDS is predominantly caused by necrosis. The serum levels of CK18-M65 ( Figure 2D), but not the CK18-M30 levels and the M30/M65 ratios ( Figures 2E and 2F), at 7 or 8 days after admission were significantly lower than those just after hospital admission.
Necrosis, unlike apoptosis, induces the release of DAMPs and augmentation of inflammation. HMGB-1, a most extensively studied DAMPs mediating organ injury during ARDS or sepsis, is increased in severe COVID-19. We also confirmed that serum levels of HMGB-1 were significantly elevated in patients with ARDS versus non-ARDS and healthy controls ( Figure 2G). Moreover, the analysis of correlations among these biomarkers in the serum of patients with COVID-19 demonstrated that HMGB-1 levels were most strongly correlated with levels of total epithelial cell death marker, M65 (correlation coefficient = 0.612, p < 0.0001, Figure S2).
Intratracheal instillation of SARS-CoV-2 spike proteins combined with poly (I:C) to mice induces lung injury mimicking COVID-19-induced acute respiratory distress syndrome The innate immune responses to components of SARS-CoV-2 are principal drivers of inflammation and alveolar tissue injury in COVID-19. [26][27][28] To elucidate mechanisms underlying alveolar epithelial cell death in COVID-19-induced ARDS, we established animal models of severe and mild COVID-19 by intratracheal instillation with the SARS-CoV-2 spike protein and poly (I:C), a synthetic analog of double-stranded RNA based on previous reports 18 and our preliminary experiments ( Figure S3). In the COVID-19 animal model, leukocytes infiltration ( Figure 3A), increased levels of protein, sRAGE, and ANG-2 in BALF ( Figures 3A-3D), and lung tissue injury ( Figure 3E) were observed. Moreover, levels of several chemokines and cytokines previously reported to be elevated in patients with COVID-19 29,30 were significantly increased in the BALF of animal models of severe COVID-19 versus controls ( Figure 3F).
We also performed bioinformatic analysis of lung tissue transcriptomes to determine whether the COVID-19 model using SARS-CoV-2 spike protein and poly (I:C) recapitulated the biological responses observed in previously reported mouse models infected with SARS-CoV-2. We identified 3,491 upregulated and 3,174 downregulated differentially expressed genes (DEGs) in our severe COVID-19 model compared with the control ( Figure S4). Furthermore, the gene set enrichment analysis (GSEA) 31 for REACTOME pathways revealed that immunological and inflammatory pathways were upregulated, while metabolic pathways were downregulated in the severe COVID-19 model ( Figure 4A). We found 6 publicly available lung tissue RNA-seq datasets of mice infected with SARS-CoV-2 from the NCBI Gene Expression Omnibus database (Table S2) (Table S3). However, reciprocal GSEA 31,38 revealed high concordance in upregulated gene sets and modest concordance in downregulated gene sets ( Figure 4B). Additionally, we compared the normalized enrichment scores (NESs) of REACTOME pathways among the datasets. The pathways significantly changed in at least one dataset were included in this analysis. The patterns of NESs showed high concordance among the groups ( Figure 4C), with a strong correlation between NES in our data and mean NES of the previously reported infection models (correlation   iScience Article relative to total epithelial cell death, decreased as lung injury increased in severity, as observed in patients with COVID-19 ( Figure 5C). Additionally, the HMGB-1 levels were significantly elevated in the severe COVID-19 animal model versus the other two groups ( Figure 5D). Taken together, these results demonstrated that the animal model of severe COVID-19 exhibited the same pattern of alveolar epithelial cell death as COVID-19 patients with ARDS. iScience Article Next, we determined whether PANoptosis (pyroptosis, apoptosis, and necroptosis), 39 inflammatory programmed cell death pathways, are involved in alveolar epithelial cell death in the animal models of COVID-19. In the lung tissues of the animal model of severe COVID-19, the levels of phospho-MLKL and cleaved GSDMD, executioners of necroptosis 40,41 and pyroptosis, 42 respectively, were significantly elevated compared with the other two groups ( Figures 5E, 5F, S7, and S8). Additionally, cleaved caspase-3, an executioner of apoptosis, was also significantly increased in the COVID-19 animal models ( Figure S9). Immunohistochemical analysis demonstrated that both phospho-MLKL and GSDMD are localized within alveolar walls ( Figure 5G). Collectively, these results indicate that PANoptosis, including necroptosis and pyroptosis, contributes to alveolar epithelial cell death in COVID-19-induced ARDS.

Anti-HMGB-1 antibody treatment attenuates alveolar tissue injury in animal models of severe COVID-19
Necrosis of alveolar epithelial cells seems to occur in the very early stages of the pathogenesis of COVID-19-induced ARDS, and it is difficult to prevent alveolar epithelial necrosis prior to hospital admission. Therefore, we assessed whether the inhibition of one of the DAMPs, HMGB-1, attenuated alveolar tissue injury in a severe COVID-19 animal model. Treatment with the anti-HMGB-1 neutralizing antibody 4 h after intratracheal instillation of poly (I:C) and the SARS-CoV-2 spike protein significantly decreased BALF levels of leukocyte infiltration, total protein, ANG-2, total CK18, and CK-18 M30 (Figures 6A-6F). On the other hand, the CK18-M30/total CK18 ratio was not affected by the anti-HMGB-1 treatment ( Figure 6G). These results suggest that DAMPs such as HMGB-1 are promising therapeutic targets that may be used to prevent the aggravation of COVID-19-induced ARDS after hospital admission.

DISCUSSION
In the present study, we demonstrated that necrosis is the predominant form of alveolar epithelial cell death in COVID-19-induced ARDS. Moreover, two forms of programmed necrosis, necroptosis, and Alveolar tissue injury in severe COVID-19 is aggravated after passing the peak viral load. 3,4 Therefore, the hyperinflammatory responses only against SARS-CoV-2 per se cannot fully explain the mechanisms underlying disease progression. Previous studies have reported that SARS-CoV-2-infected macrophages and monocytes undergo inflammasome activation and pyroptosis, potentially resulting in DAMPs release and excessive inflammation. [43][44][45] However, the pathological contribution of damages to non-immune cells remains nebulous. Recently, we have suggested that alveolar epithelial injury at a very early disease stage may trigger subsequent COVID-19 progression. 5 Herein, we show that the initial alveolar epithelial necrosis and subsequent release of DAMPs may also be a cause of excessive inflammation, which can progress even after viral loads have peaked. Not only cellular infection of SARS-CoV-2 but also inflammatory mediators, including TNF-a and IFN-g, can cause programmed necrosis. 39 It is possible that programmed necrosis of immune and alveolar epithelial cells synergistically augment inflammatory responses and cell death or each other.
Both necrosis and apoptosis are involved in alveolar epithelial cell injury in ARDS. 9 Since apoptosis can be easily assessed using TUNEL staining or caspase detection, the contribution of alveolar epithelial apoptosis in ARDS has been extensively studied. 46 However, we have previously demonstrated that necrosis is the dominant form of alveolar epithelial cell death in LPS-induced ARDS by the quantification of CK18-M30 and total CK18, which is equivalent to CK18-M65, in addition to cell labeling techniques. 10 In particular, quantification of CK18-M30 and M65 levels, using commercially available ELISA kit, can be applied for the evaluation of epithelial apoptosis and necrosis in the clinical setting. In fact, the patterns of epithelial cell death such as sepsis 47,48 and graft rejection after lung transplantation 49 have been previously analyzed. To the best of our knowledge, this is the first study to suggest that necrosis is the predominant form of alveolar epithelial cell death in human ARDS. Further studies are warranted to identify alveolar patterns of epithelial cell death in ARDS that is induced by disease etiologies other than COVID-19.
Necrosis has been previously thought to cause accidental cell death; however, some forms of necrosis, referred to as programmed necrosis, are regulated via molecular pathways. 7 Several animal studies have demonstrated that programmed necrosis is involved in alveolar epithelial cell death in ARDS. [50][51][52][53] Moreover, studies have suggested that SARS-CoV-2 activates intracellular necroptosis and pyroptosis pathways, 39,44,45,[54][55][56] and that the circulating level of receptor-interacting protein kinase 3, a kinase required for necroptosis, is elevated in critically ill patients with COVID-19. 57 In line with findings of these studies, our animal experiments suggest that necroptosis and pyroptosis are involved in alveolar epithelial cell death in COVID-19 ARDS. Programmed necrosis is a response to eliminate SARS-CoV-2 infection, however, it can also cause excessive inflammation and subsequent tissue damage.
The release of DAMPs to extracellular spaces is a characteristic of necrosis that distinguishes it from apoptosis. 58 Several previous studies have also reported that circulating levels of DAMPs, such as HMGB-1, 59-62 histone, 62,63 cell free-DNA, 62,63 mtDNA 64-66 and S100 proteins 67,68 are elevated in severe COVID-19. Alveolar epithelial necrosis likely occurs very early in COVID-19 progression 5,6 and potentially contributes to disease progression. Moreover, it is possible that these DAMPs can be released from necrosis of immune cells such as macrophages. 44,45 In both cases, the prevention of necrosis prior to the appearance of clinical symptoms is difficult; therefore, a strategy for preventing DAMPs-mediated disease aggravation is needed. HMGB-1 is the most extensively studied DAMPs that drives tissue injury during ARDS or sepsis; results from our animal study demonstrated that the inhibition of HMGB-1 efficiently attenuates disease progression. However, it remains unclear as to the type of DAMPs with the most contribution to disease progression. Further studies, including clinical trials, are warranted to investigate the clinical efficacy of DAMPs inhibition in patients with COVID-19 ARDS.
In the present study, an animal model mimicking COVID-19 was established by administering the SARS-CoV-2 spike protein combined with poly (I:C), similar to previous studies. 18,69,70 The animal models treated with infectious strains of SARS-CoV-2 are often ideal models for COVID-19; however, using infectious viruses in animal experiments can be difficult. First, SARS-CoV-2 infection does not occur in wild-type mice or rats. For infection to occur, expression of the human ACE receptor is needed. Second, appropriate ll OPEN ACCESS iScience 26, 105748, January 20, 2023 iScience Article facilities and equipment are needed to meet all the safety requirements when working with an infectious agent. Although the pathogenicity of SARS-CoV-2 is complex, stimulation of pathogen-associated pattern recognition receptors including toll-like receptors 26,27 and retinoic acid-inducible gene-I receptors 28 by viral components is the principal driver of lung inflammation and subsequent alveolar tissue damage. In the present study, the bioinformatic analysis of lung tissue transcriptomes demonstrated that the COVID-19 mimicking animal model with SARS-CoV-2 spike protein and poly (I:C) recapitulated key biological responses in inflammatory and cell death pathways induced by SARS-CoV-2. 32-37 Moreover, alveolar cell death patterns in our COVID-19 animal model were similar to those observed in human COVID-19. 29,30 Our results highlight the utility of investigating the pathophysiology and treatment of COVID-19 using animal models established with components of SARS-CoV-2.
Our data suggest that the plasma M30/M65 ratio (an indicator of apoptosis in relation to total levels of epithelial cell death) is a potential marker of COVID-19 severity. Our findings agreed with a previous study that showed M30/M65 ratios of hospitalized patients with COVID-19 were lower than those of non-hospitalized patients. 71 Additionally, different subtypes of COVID-19 respond differently to treatments. 72,73 The M30/M65 ratio may serve as a marker for selecting patients likely to benefit from anti-DAMPs treatment.
In summary, our data indicate that necrosis, including necroptosis and pyroptosis, is the predominant form of alveolar epithelial cell death in COVID-19-induced ARDS. The DAMPs released from necrotic alveolar epithelial cells are potential drivers of progressive alveolar tissue damage in COVID-19, and hence are promising targets for preventing the aggravation of ARDS in patients with COVID-19.

Limitations of the study
This study has some limitations. First, only patients admitted to a single center were included in the analysis due to the limited availability of clinical samples. Further studies with samples from multiple centers in different countries are warranted. Second, serum samples were used in most parts of the human study, as the bronchoalveolar lavage fluids were available only from a limited number of severe patients with COVID-19. However, severe organ injury was almost limited to lungs in the present cohort, tissue injury markers in serum samples might well reflect the pulmonary pathology. Third, our animal model was created by exposing mice to components of SARS-CoV-2, and not an infectious strain of SARS-CoV-2. Despite this, the observations of a COVID-19-like pathology by this study and previous reports 18,70 support the use of the animal model. Importantly, the use of a non-infectious model is convenient for laboratories that do not specialize in infectious disease research. Fourth, the efficacy of inhibiting only a single DAMP, HMGB-1, was evaluated. Several types of DAMPs are released from necrotic cells; therefore, whether the types of DAMPs are the primary therapeutic targets for COVID-19 remains to be determined.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Clinical study design
In this single-center, retrospective, prospective observational study, we analyzed serum samples of adult patients with COVID-19 who were admitted to Yokohama City University Hospital from January 2020 to January 2021 and healthy controls matched as closely as possible for age and sex. Inclusion criteria for COVID-19 patients were, as follows: 1) a diagnosis of COVID-19 based on a positive real-time PCR test, 2) age R18 years, and 3) available residual serum samples. Additionally, we analyzed BALF samples from patients with COVID-19-induced ARDS who were admitted to the Yokohama City University Hospital from April 2021 to January 2022. ARDS was diagnosed based on the Berlin definition. The study protocol was reviewed and approved by the institutional review board of Yokohama City University Hospital (B200700100, B200200048). The requirement for informed consent was waived due to the observational nature of the study. Some preliminary data from retrospectively collected samples have been previously published. 5

Clinical data collection
The following clinical data measured during the first 8 days of hospital admission were retrospectively collected from the medical charts of included patients: basal characteristics, vital signs, laboratory tests, and blood gas analysis findings. Th initial concentrations of these markers in ARDS and non-ARDS patients at admission (on the first or second hospital day) and healthy controls were compared. Further, temporal changes in levels of the markers were assessed in patients with ARDS throughout the 8-day period following hospital admission. In cases in which values were determined twice per day, mean values were used. In cases in which only a single value was available, the value was used.

Analysis of BALF samples from COVID-19-induced ARDS patients
BALF samples were obtained from six COVID-19-induced ARDS patients. A fiberoptic bronchoscope was wedged in a lateral or medial segmental bronchus of the right middle lobe, and lavage was performed using three aliquots of 50 mL of sterile isotonic sodium chloride solution. The collected BALF was centrifuged at 300 3 g for 5 min at 4 C, and the supernatant was stored at À80 C until analyses. The levels of CK18-M65 and CK18-M30 were quantified using ELISA, while the ratio of CK18-M30/M65 was calculated as described above.

Animal experiments
All animal experimental protocols were approved by the Animal Research Committee of the Yokohama City University. Male specific-pathogen-free C57BL/6J mice aged 8-10 weeks that were purchased from Japan SLC (Shizuoka, Japan) were used for all animal experiments. Mice were housed under a 12-h light/dark cycle with food and water available ad libitum.
A mouse model mimicking COVID-19 was established based on previous reports. 18,69 Intratracheal administration of polyinosinic:polycytidylic acid (poly (I:C)) (P1530, Sigma-Aldrich, St. Louis, MO, USA) and the SARS-CoV-2 spike protein (Z03481, Lot B2103045, GenScript, Piscataway, NJ) was performed via the exposed trachea through a small incision at the front of the neck. During the procedure, mice were placed ll OPEN ACCESS