SARS-CoV-2 infection alters mitochondrial and cytoskeletal function in human respiratory epithelial cells mediated by expression of spike protein

ABSTRACT Coronavirus disease 2019 (COVID-19) is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2, SCV2), which has resulted in higher morbidity and mortality rate than other respiratory viral infections, such as Influenza A virus (IAV) infection. Investigating the molecular mechanisms of SCV2-host infection vs IAV is vital in exploring antiviral drug targets against SCV2. We assessed differential gene expression in human nasal cells upon SCV2 or IAV infection using RNA sequencing. Compared to IAV, we observed alterations in both metabolic and cytoskeletal pathways suggestive of epithelial remodeling in the SCV2-infected cells, reminiscent of pathways activated as a response to chronic injury. We found that spike protein interaction with the epithelium was sufficient to instigate these epithelial responses using a SCV2 spike pseudovirus. Specifically, we found downregulation of the mitochondrial markers SIRT3 and TOMM22. Moreover, SCV2 spike infection increased extracellular acidification and decreased oxygen consumption rate in the epithelium. In addition, we observed cytoskeletal rearrangements with a reduction in the actin-severing protein cofilin-1 and an increase in polymerized actin, indicating epithelial cytoskeletal rearrangements. This study revealed distinct epithelial responses to SCV2 infection, with early mitochondrial dysfunction in the host cells and evidence of cytoskeletal remodeling that could contribute to the worsened outcome in COVID-19 patients compared to IAV patients. These changes in cell structure and energetics could contribute to cellular resilience early during infection, allowing for prolonged cell survival and potentially paving the way for more chronic symptoms. IMPORTANCE COVID-19 has caused a global pandemic affecting millions of people worldwide, resulting in a higher mortality rate and concerns of more persistent symptoms compared to influenza A. To study this, we compare lung epithelial responses to both viruses. Interestingly, we found that in response to SARS-CoV-2 infection, the cellular energetics changed and there were cell structural rearrangements. These changes in cell structure could lead to prolonged epithelial cell survival, even in the face of not working well, potentially contributing to the development of chronic symptoms. In summary, these findings represent strategies utilized by the cell to survive the infection but result in a fundamental shift in the epithelial phenotype, with potential long-term consequences, which could set the stage for the development of chronic lung disease or long COVID-19.

I nfection with coronavirus (SARS-CoV-2, SCV2) causes coronavirus disease 19  that has spread throughout the world. COVID-19 leads to a high inci dence of pneumonia, acute respiratory distress syndrome, and death (1,2). SCV2 entry into host cells requires binding between the angiotensin-converting enzyme-2 (ACE2) receptor and the viral receptor-binding domain of the spike protein. SCV2 entry can occur by one of two paths depending on the variant. Entry at the plasma membrane occurs through the cleavage of the spike protein by serine protease transmembrane serine protease 2 (TMPRSS2) or related serine proteases (3, 4) followed by the cells via endocytosis. Alternatively, virus particles can undergo early endocytosis with the spike protein proteolytically activated by cathepsin-like proteases in endosomes. By comparison, the seasonal influenza virus (influenza A and B virus, IAV, and IBV) enters the host cells via the binding of hemagglutinin to sialic acid residues, followed by internalization and low pH mediated triggering of hemagglutinin-mediated membrane fusion to allow for the transfer of viral ribonucleoproteins to the host cytoplasm (5,6). Despite these distinct mechanisms of cell entry and viral spread, both viruses infect similar cells in the lung epithelium. However, the mortality rate of COVID-19 is substan tially higher than influenza (7). Additionally, there is growing evidence for the increasing prevalence of "long COVID" and potentially persistent radiographic lung changes (8) raising the question of divergent long-term cellular responses (9). Therefore, we studied the epithelial responses to the SCV2 and IAV infections, including host immune response, tissue injury, and cellular metabolism (10)(11)(12).
Airway epithelium, the first point of contact for both IAV and SCV2, protects the lung against infection by maintaining a physical barrier impermeable to access to subepithelial tissues and an immunologic barrier triggering antiviral immune responses (13,14). Indeed, as both upper and lower airway epithelial cells are infected by both viruses, comparing IAV and SCV2 in both these epithelial compartments provides a more thorough understanding of the cellular responses triggered by these viruses, with the goal of identifying critical pathways that distinguish the effects of the two viruses and could explain the altered clinical presentations. In this study, we assessed differential gene expression of SCV2 vs IAV infection in human nasal epithelial cells (hNECs) to elucidate any distinct cellular responses between the two viruses that may lead to differences in clinical presentation. We found significantly less cytotoxicity and activation of cell death pathways in the SCV2-infected epithelia. However, the surviving epithelia demonstrated alterations in both cytoskeletal and metabolic pathways suggestive of epithelial remodeling and plasticity in the SCV2-infected cells, reminiscent of pathways activated in chronic injury responses. These strategies provide cellular resilience but could pave the way for more chronic symptoms by changing epithelial structure and function. Our study provides evidence of distinct cellular responses to SCV2 and IAV infection that could result in these divergent clinical outcomes and raises the possibility that the cell survival and resilience that were observed with SCV2 permits the develop ment of long covid.

Transcriptome analysis of human nasal cells with IAV and SCV2 infections
Nasal cells are the first site of viral contact harboring high levels of ACE2 and TMPRSS2 (15)(16)(17). Therefore, hNECs at the air-liquid interface (ALI) were infected with IAV and SCV2 for 12 and 24 h post infection (hpi) (MOI 1), and differentially expressed genes (DEGs) were identified. Eighteen RNA-seq libraries (three experimental replicates for each treatment and timepoint, including mock-infected controls) were prepared. We have previously confirmed that fully differentiated hNECs at ALI strongly correlate with cells freshly isolated from nasal brushes (18 (Fig. 1C) as shown in the heatmap (Fig. 1D). As gene expressions were differentially expressed between SCV2 and IAV at 24 hpi, pathway analysis was performed using KEGG Mapper (https:// www.genome.jp/kegg/mapper/). Compared to IAV, inflammation, signal transduction, cell adhesion, cell death, and metabolism were enriched in the downregulated genes in SCV2, but no enriched pathway was found in the upregulated genes in SCV2 (Fig.  1C). As COVID-19 patients exhibit lower cytokine levels than patients with influenza (19), in this study, we focused on the early changes in metabolism and cell adhesion pathways in SCV2 and IAV infection. In general, we found the suppression of cell death pathways with SCV2 infection, with decreased apoptosis ( Fig. 2A), necroptosis (Fig. 2B), and ferroptosis (Fig. 2C) in SCV2. Of note, general regulation of cellular transcription remained upregulated with SCV2 infection compared to IAV (Fig. 2D). These data indicate that the epithelium is more likely to survive SCV2 infection than IAV at this timepoint.
We found three genes that were commonly activated with infection with SCV2 at both 12 and 24 hpi (DUSP1, HES1, KLF2) ( Table 1) and some non-congruent genes at 12 hpi (CCL5, ZBP1, NRXN2, STAB1, SLC25A47, CXCL11) and 24 hpi (FOXA2, RHOB) ( Table 2). To assess the dependency on interaction with the spike protein, we infected normal human bronchial epithelial (NHBE) cells at the air-liquid interface with SCV2 spike pseudovirus. The vesicular stomatitis virus (VSV-G) pseudovirus is an enveloped nega tive-stranded RNA virus (20,21). Engineering the SCV2 spike protein into the VSV-G fused fluorescent protein DsRed allows us to analyze virus entry and determine the specific downstream mechanisms induced by the spike protein. Acquiring the pseudovirus model by infection of SCV2 spike pseudovirus (Spike) or control virus (VSV-G) to NHBE at ALI for 72 h, the infection efficiency was examined by red fluorescent protein (DsRed)tagged pVSV-G (Fig. 3A), which was calculated to be approximately 80%-90% of ciliated epithelial cells.   (Fig. 4G). The corresponding decrease in TOMM22 protein expression in Spike-infected NHBE cells was confirmed by immunofluorescence (Fig. 4H). Next, we utilized a differentiated human non-small-cell lung cancer cell line Calu3 at ALI, as this cell line has been demonstrated to be highly permissive to SCV2 with similar characteris tics as NHBE with high barrier integrity (23)(24)(25)(26)(27). Differentiated Calu3 cells were infected with the Spike or control virus and the infection efficiency was examined by

SCV2 alters epithelium persistently for chronic inflammatory changes
Airway epithelium is a frontline of defense against pathogen. Epithelial permeability is maintained by the apical junctional complexes that include tight junctions and adhesion junction and link with cellular cytoskeleton (30). In addition, focal adhesion and cytoske leton encompass a network to maintain the proper positioning of a cell's shape and positioning of its constituent organelles (31). SCV2-infected hNECs generally decreased genes involved in the tight junction (Fig. 6A), cell adhesion (Fig. 6B), focal adhesion ( Fig.  6C), as well as regulation of actin cytoskeleton (Fig. 6D). Notably, loss of cofilin-1, an actin-severing protein, occurs in chronically injured epithelium, as seen with recurrent insults that result in chronic obstructive pulmonary disease (COPD) (32). We have found that gene expression of cofilin-1, CFL1, was downregulated in both SCV2-infected hNECs (Fig. 6D) and the spike-infected NHBE cells (0.65-fold in Spike vs VSV-G, P = 0.0099) (Fig.  6E). The spike infection in NHBE cells caused the loss of cofilin-1 protein (0.65-fold in Spike vs VSV-G, P = 0.0134) (Fig. 6F). In our RNA sequencing data, we also observed that Ras homolog family member B (RHOB), which plays a vital role in cell-cell actin-based adherens junctions (33), was reversibly expressed between SCV2 and IAV2-infected hNECs (increased in SCV2 but decreased in IAV infections) ( Table 2). In the pseudovirusinfected NHBE cells, we found RHOB was induced in Spike by 1.35-fold in spike vs VSV-G (P = 0.0216) (Fig. 6G). As expected with a decrease in actin-severing, the localization of the decreased cofilin-1 and the increased F-actin (stained by phalloidin) was observed in pseudovirus-infected cells by immunofluorescence (Fig. 6H). Of note, the loss of cofilin-1 was associated with the decreased barrier integrity in NHBE (0.91-fold in Spike vs VSV-G, P = 0.0179) (Fig. 6I) and in Calu3 (0.78-fold in Spike vs VSV-G, P = 0.0015) (Fig. 6J). To confirm the increased F-actin in SCV2 infection, the immunofluorescent intensity of phalloidin was significantly higher in the lungs of COVID-19 patients compared to healthy individuals (Fig. 7) with xz-images showing the increase in phalloidin throughout the thickness of the tissue slice. The data imply that SCV2 infection dysregulates epithelial barrier integrity through the loss of cofilin-1 and increased F-actin.

DISCUSSION
Though respiratory viruses all initially engage with airway epithelial cells, the mecha nisms of the initial immune response to infection are distinct and can result in divergent clinical responses (34,35). Moreover, there is increasing recognition that SARS-CoV-2 infection can lead to prolonged symptoms or "long-COVID" and may even have persistent radiographic changes but the biological basis of which remains unknown.
Here, we compared the transcriptomic profile of the SCV2 vs IAV infection of the nasal epithelium to identify distinct viral responses. Surprisingly, we found that IAV resulted in much more injury to the epithelium at 24 h after infection compared to SCV2. In fact, the early epithelial responses to SCV2 were quite mild and very similar to that of the mock control. By 24 h after the IAV infection, cell death pathways were activated, which did not occur with the SCV2 infection. In contrast, we saw evidence of altered cell structure and metabolism, suggestive of cellular remodeling in response to the SCV2 infection. Infection with SCV2, moreover, caused a downregulation in the protein cofilin-1, and impaired cell-cell adhesion, a finding that is recapitulated with infection with the spike protein.
Although we did not focus on the inflammatory response, we did see an early upregulation of epithelial inflammatory processes. In both 12 and 24 hpi, we found three genes were specifically upregulated in SCV2-infected hNECs (DUSP1, HES1, KLF2). Dual specificity phosphatase (DUSP1) is a negative regulator of the p38 MAPK signaling and diminishes the inflammatory response (36,37). DUSP1 has been shown to be downregu lated at 24 h SCV2-infected human epithelial lung cell line A549 (MOI 0.2) (38), but we found early upregulation in the primary cells. HES, hes family BHLH transcription factor 1 (HES1), a basic helix-loop-helix transcriptional repressor, is a target gene of Notch signaling (39). As notch signaling positively regulates protease furin, protease ADAM Metallopeptidase Domain 17 (ADAM17), and angiotensin-converting enzyme 2 (ACE2) to promote SCV2 entry and infection to the host cells (40). Krüppel-like factor 2 (KLF2) is a transcriptional factor that suppresses inflammatory responses (41). Although we have found upregulation of KLF2 in the nasal at these early timepoints, the study of lung autopsies suggests downregulation in KLF2 in later stages of the disease in patients who Research Article mBio succumbed to COVID-19 (42). The role of KLF2 in early vs later disease can be the focus of future studies. Interestingly, we found the Ras homolog family member B (RHOB), a gene that belongs to the Rho GTPase family, was not congruently expressed between SCV2-and IAV infections at 24 hpi (Table 2). RHOB plays a vital role in cell-cell actin-based adherens junctions through ROCK/LIMK1 signaling pathway (33), a pathway highly implicated in actin cytoskeletal rearrangements. The LIM-kinase 1 (LIMK-1) induces phosphorylation and inactivation of cofilin leading to increased actin polymerization (43). Cofilin-1 is involved in many cellular processes, including actin cytoskeletal remodeling and cellular metabolism (44,45), and our group has found that loss of cofilin-1 contributes to the cytoskeletal remodeling that occurs with epithelial plasticity in response to chronic injury (32). Virus infection involves the remodeling of the actin cytoskeleton of the host cells (46,47), and COVID-19 hijacks the cytoskeleton and causes barrier dysfunction (48,49). In contrast, SCV2 infection in the nasal cells activates ROCK/LIMK signaling, leading to cofilin downregulation. Indeed, diminished inflammatory response in SCV2 infection than IAV has also suggested by the downregulation of cofilin as cofilin plays a key role in regulating the NLRP3 inflammasome (50). Previous studies have demonstrated that virus infections (transmissible gastroenteritis virus and porcine hemagglutinating encephalomyelitis) caused cofilin inactivation at the early stage of infection, leading to F-actin polymerization and rearrangement as well as further promoting virus entry (51,52).
Given the involvement of cofilin-1, it was not surprising that SCV2 infection also caused dysregulated host metabolism compared to IAV infection, with decreased mitochondrial and mitochondrial-related gene expression, and increased glycolysis. Others have shown evidence that SCV2 reprograms host metabolism (53,54). Both the ciliated epithelia and Type II alveolar epithelia are rich in mitochondria (55) with specialized activities (32,56) and mitochondrial dysfunction could significantly impact their cellular function. This mitochondrial dysfunction is consistent with differential metabolomic profiles seen in patients with acute respiratory distress syndrome due to COVID-19 versus H1N1 influenza (57). The shift to aerobic glycolysis could favor SCV2 replication and survival through initial infection (58). We hypothesize that these findings represent strategies utilized by the cell to survive the infection but result in a fundamen tal shift in the epithelial phenotype, with potential long-term consequences, which could set the stage for the development of chronic lung disease or long COVID (59).

Cell culture
Primary human nasal cells (hNECs) were obtained from patients with chronic rhinosinusi tis (CRS) who tested negative for COVID-19 and were approved through Johns Hopkins Institutional Review. All subjects signed informed consent (60). Primary human bronchial  Table S1. Briefly, the cells were cultured on the collagen I-coated transwell inserts with 0.4 µm pore (Corning) and differentiated for 4-6 weeks at the air-liquid interface (ALI) as described previously (32,61). Human lung epithelial cells Calu3 were purchased from ATCC and cultured in EMEM (ATCC) with 10% fetal bovine serum (ThermoFisher Scientific) and 1% penicillin streptomycin (ThermoFisher Scientific). Calu3 cells were seeded onto the transwell with 0.4 µm pore at 5 × 10 4 cells/cm 2 and cultured at the ALI for 1 week.

SARS-CoV-2 and influenza A infection in human nasal cells
Human nasal cells at ALI were infected with mock, SCV2 SARS-C0V-2/USA/HP7(27)/2020, and IAV A/Baltimore/R0243/2018 H3N2 at multiplicities of infection (MOI) of approxi mately 1.0 infectious unit per cell at the apical side for 120 min at 37°C followed by two PBS washes to eliminate unbound virus. The cells were incubated in a 37°C CO 2 incubator for 12 or 24 hpi and then lysed with TRIZOL (ThermoFisher Scientific) for RNA isolation. Infectious virus titers were determined by median tissue culture infectious dose (TCID50), as described previously (62,63).

RNA sequencing and analysis
Three replicates were prepared for each infection (mock, IAV, SCV2) at 12 and 24 hpi. Extraction, purification, and quality assessment of Total RNA from hNECs were performed Research Article mBio as previously described (64). Unique dual-index barcoded libraries for RNA-Seq were prepared from 125 ng total RNA using the Universal Plus Total RNA-Seq with NuQuant Library kit (Tecan Genomics), according to manufacturer's recommended protocol, with amplification performed for 16 cycles, as optimized by qPCR. Quality assessment and quantitation of libraries were performed as previously described (64)

Preparation of SCV2 pseudovirus and transduction
Full-length VSV glycoprotein (VSV-G) (Addgene, #12259) and pRP-Neo-EF1A-SARS-Cov2-Spike delta 18, fused with a DsRed tag, were kindly gifted from Vasudevan, Anand's lab. Plasmid for the SARS-CoV-2 spike delta 18 glycoprotein has a C-terminal deletion of the last 18 amino acids to improve binding to ACE2. Human NHBE and Calu3 cells were infected with the pseudovirus with VSV-G or Spike-VSV-G at MOI 1 with 10 mg/mL of polybrene (Sigma-Aldrich) followed by centrifugation at 1500 rpm at room temperature for 2h. Then, the infection media were removed, and the cells were washed with PBS twice. The infected cells were incubated at 48 and 72 hpi for Calu3 and NHBE, respec tively (>80% infection efficiency that was determined by DsRed fluorescent signal, data not shown). Cells were harvested at the designated time points.
For NHBE and Calu3 infected cells, at the desired time points, the cells were washed three times in phosphate-buffered saline (PBS) and then were fixed with 4% paraformal dehyde in PBS (Affymetrix Inc., OH, USA) for 15 min at room temperature. For human lungs, human lung was obtained from Johns Hopkins School of Medicine Pulmonary Department of Biorepository. Then, the lung tissues were embedded in paraffin and stained with hematoxylin and eosin (H&E) at Reference Histology Laboratory, Johns Hopkins Medical Institute-Pathology (Baltimore, USA). The patient demographics were shown in Table S1. The paraffin-embedded PCLS were deparaffinized in xylene followed by rehydration and antigen unmasking (Citrate Buffer pH:6.0).

Immunofluorescence confocal microscopy
Both cells and lung tissues were subjected to immunofluorescence staining as described previously (65). Briefly, the slides were permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) in PBS. Subsequently, they were blocked in 10% goat serum (Sigma-Aldrich) and 1% bovine serum albumin (Sigma-Aldrich) in 1× PBS for 1 h at room temperature. Next, they were incubated at 4°C overnight with primary antibodies rabbit against cofilin (D3F9) (1:200, Cell Signaling #5175), TOMM22 (1:200, Proteintech #11278-1-AP). After three washing with 1× PBS, the cells were incubated at the secondary antibody goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor 555 (1:200, Invitrogen # A-21428) and Alexa Fluor 647 Phalloidin (1:400, Invitrogen #A22287) for 2 h at room temperature. The cells were stained with 1 µg/mL of Hoechst (Thermo Fisher Scientific #62249) for visualization of the nuclei and mounted with ProLong Gold antifade (ThermoFisher Scientific). The slides were imaged using Zeiss LSM 700 Confocal or LSM880-AiryscanFAST with a 63 × oil objective. Images presented are a maximum projection of Z-stack images collected with LSM software unless the xz-plane was provided as with the tissue slices. Random fields were chosen for the imaging of cells, while all available sections with airways present were analyzed in the tissues.

Reverse transcription-quantitative polymerase chain reaction
RNA was isolated using the NucleoSpin TriPrep kit (TaKaRa Bio USA, Inc.) according to the manufacturer's instructions. One microgram of total RNA was reverse transcri bed using an ABI High-capacity cDNA Reverse Transcription kit (Foster City, CA, USA). The complementary DNA was subjected to QPCR using Power SYBR Green Master Mix (ThermoFisher Scientific) and performed in StepOne Plus (ABI). The relative gene expression changes were normalized by housekeeping gene GAPDH and calculated using the 2−ΔΔCt method (66). The qPCR primers are listed in Table S2.

Seahorse analysis
Cell Mitochondrial Stress (Agilent) was conducted to detect oxygen consumption rate and extracellular acidification rate. ATP production rate using Seahorse XFe24 Analyzer (Agilent) through the injection of 25 µg/mL oligomycin, 4 µM FCCP, 5 µM rotenone, and 10 µM antimycin A. Calu3 cells (1 × 10 5 cells/cm 2 ) were differentiated on the transwell with 0.4 µM pore for 1 week and infected with the pseudovirus. After 48 h infection, the cells were changed to Agilent Seahorse XF DMEM medium supplemented with 1 mM of sodium pyruvate, 2 mM of glutamine, and 1 mM of D-glucose and incubated at 37°C for 1 h. After washing with the assay buffer, OCR and ECAR were measured using Seahorse XF24 instrument and analyzed by Seahorse XF24 analyzer.

Statistical analysis
Results were expressed as the mean with the error bars representing means (±stand ard error of the mean[SEM]). Statistics were determined by Welch's test, with P < 0.05 considered statistically significant. All the data were analyzed and plotted by Prism 9.3 (Graphpad).