Protein Arginylation Is Regulated during SARS-CoV-2 Infection

Background: In 2019, the world witnessed the onset of an unprecedented pandemic. By February 2022, the infection by SARS-CoV-2 has already been responsible for the death of more than 5 million people worldwide. Recently, we and other groups discovered that SARS-CoV-2 infection induces ER stress and activation of the unfolded protein response (UPR) pathway. Degradation of misfolded/unfolded proteins is an essential element of proteostasis and occurs mainly in lysosomes or proteasomes. The N-terminal arginylation of proteins is characterized as an inducer of ubiquitination and proteasomal degradation by the N-degron pathway. Results: The role of protein arginylation during SARS-CoV-2 infection was elucidated. Protein arginylation was studied in Vero CCL-81, macrophage-like THP1, and Calu-3 cells infected at different times. A reanalysis of in vivo and in vitro public omics data combined with immunoblotting was performed to measure levels of arginyl-tRNA-protein transferase (ATE1) and its substrates. Dysregulation of the N-degron pathway was specifically identified during coronavirus infections compared to other respiratory viruses. We demonstrated that during SARS-CoV-2 infection, there is an increase in ATE1 expression in Calu-3 and Vero CCL-81 cells. On the other hand, infected macrophages showed no enzyme regulation. ATE1 and protein arginylation was variant-dependent, as shown using P1 and P2 viral variants and HEK 293T cells transfection with the spike protein and receptor-binding domains (RBD). In addition, we report that ATE1 inhibitors, tannic acid and merbromine (MER) reduce viral load. This finding was confirmed in ATE1-silenced cells. Conclusions: We demonstrate that ATE1 is increased during SARS-CoV-2 infection and its inhibition has potential therapeutic value.


Introduction
In 2019, the world witnessed the onset of an unprecedented pandemic [1]. Patients in the capital and largest city in China's Hubei province, Wuhan, developed pneumonia associated with infection with a new type of coronavirus called severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [2,3]. The clinical symptoms presented by infected patients ranged from mild to severe and included nonspecific manifestations such as fever, cough, sore throat, respiratory failure, muscle damage, and death [2,[4][5][6]. Although there rule pathway and inhibits apoptosis, benefiting viral replication. In addition, arginylated peptides have already been identified in trypanosomatids and the putative protein ATE1 has been identified in P. falciparum, the etiologic agent of malaria. The enzyme has been shown to have a prokaryotic-like sequence, however, eukaryotic transferase specificity [50]. Modulation of the N-degron pathway has been shown to influence Bacillus anthracis infection, as the EF adenylate cyclase toxin is a substrate of the pathway and is critical for the progression of pathophysiology.
In this manuscript, we elucidate the role of arginylation during SARS-CoV-2 infection. We conducted a study on modulation of the N-degron pathway and protein arginylation in Vero CCL-81, macrophage-like THP1, and Calu-3 cells infected at different times. A reanalysis of public omics data combined with Western blotting was performed to measure the levels of ATE1 and arginylated proteins. Dysregulation of the N-degron pathway was specifically identified during coronavirus infection, not in other respiratory viruses. We demonstrated that during SARS-CoV-2 infection there is an increase in the expression of the ATE1 enzyme and its modified substrates in Calu-3, HEK 293T, and Vero CCL-81 cells. On the other hand, infected macrophages showed no ATE1 regulation. Moreover, ATE1 modulation was found to depend on the SARS-CoV-2 variant, as shown with P1 and P2 viral infections and transfection with the spike glycoprotein receptor-binding domain (RBD) of different variants. In addition, ATE1 inhibitors, tannic acid and merbromine (MER) reduced viral load, and this was confirmed in ATE1 silenced cells.

Bioinformatics Analysis
The tidyverse [68], biostrings, and seqinr [69] packages were used to map potentially arginylated proteins in the Homo sapiens and Chlorocebus sabaeus proteomes (downloaded in 1 May 2021, https://www.uniprot.org/, accessed on 1 May 2021). Signal peptide Viruses 2023, 15, 290 4 of 25 sequences were removed. Only proteins that have the potential to be arginylated at the N-terminus (NtE, NtD, NtC, NtN, NtQ) were retained. Caspase-generated fragments were not considered. It is important to mention that this list contains potential arginylated proteins that need to be confirmed experimentally and other targets might not be present in this list. The corrplot package was used to evaluate the correlation between proteins/genes, applying a Spearman test with a cut-off significance of p-value < 0.05. Protein subcellular locations were determined by UniProt release 12.4 (https://www.uniprot.org/news/20 07/10/23/release, accessed on 1 May 2021) and the pRoloc package [70]. The analysis of gene ontology (GO) was determined by the g:profile [71] and DAVID [72] tools. A q-value threshold of 0.05 was used, corrected by the Benjamini-Hochberg method [73]. InteractiVenn was used to build the Venn diagrams [74]. The String database v.11.5 was applied for protein network analysis (https://string-db.org/, accessed on 1 May 2021) with the following parameters: medium confidence score (0.400), text mining, coexpression, and neighborhood enabled.

Single-Cell RNA-seq Re-Analysis
Expression matrices were loaded into RStudio (v. 4.0.3) with the Seurat package [75]. A filter to remove cells with less than 200 expressed genes or more than 25% of mitochondrial transcripts was applied using the 'subset()' function in each sample. Then, cell counts were log-normalized by a size factor of 10,000 RNA counts and feature selection was performed by selecting the 2000 genes with the highest dispersion. Unsupervised identification of anchor correspondences between the canonical correlation analysis (CCA) space of each sample' normalized data was performed with the 'FindIntegrationAnchors()' function with 30 dimensions. After that, the data were integrated by 'IntegrateData()' function and scaled using 'ScaleData()'. Principal component analyses (PCA) and uniform approximation and projection dimension reduction (UMAP) with 30 principal components were applied. A nearest neighbor plot using 30 PCA reduction dimensions was calculated using 'Find-Neighbors()', followed by clustering using 'FindClusters()' with a resolution of 0.5. The Metaboanalyst platform [76] was used to evaluate differently regulated genes between cell clusters identified in the single-cell RNA-seq analysis.
All assays were performed in biological triplicates in a BSL-3 facility at the Institute of Biomedical Sciences, University of Sao Paulo, under the Laboratory biosafety guidance related to coronavirus disease (COVID-19): Interim guidance, 28 January 2021 (https://www.who.int/publications/i/item/WHO-WPE-GIH-2021.1, accessed on 1 May 2021).

Time-Course Evaluation of Protein Arginylation during Viral Infection
For comprehensive time course evaluation, Vero CCL-81 and Calu-3 cells were infected with SARS-CoV-2. Cell lysates were collected at 2, 6, 12, 24, and 48 h post infection (hpi) in 8 M urea supplemented with protease (cOmplete, Sigma-Aldrich, St. Louis, MO, USA) and phosphatase inhibitors (PhosStop, Sigma-Aldrich). Aliquots of cells and supernatants were collected at the different time points for virus RNA copy numbers quantification by reverse transcription-quantitative polymerase chain reaction (RT-qPCR), targeting the E gene as previously described [79].

siRNA-Directed Inhibition of ATE1
Predesigned siRNAs for the ATE1 transcript (siATE1) (hs.Ri.ATE1.13.3) were purchased from Integrated DNA Technologies (IDT, Coralville, IA). Calu-3 cells were transfected with 3 µL Lipofectamine 3000 reagent (Thermo Fischer Scientific, Waltham, MA, USA) alone (control) or with 30 pmol of siRNA-ATE1 in 12 well plates, according to the manufacturer's recommendation. After incubation for 2 h at 37 • C and 5% CO 2 , fresh cell culture medium supplemented with 5% FBS was added to each condition. At 48 h after transfection, viral and mock infections were carried out as described above. Cell lysates were collected in BE buffer containing protease (cOmplete, Sigma-Aldrich) and phosphatase inhibitors (PhosStop, Sigma-Aldrich).

Viral Quantification
Aliquots of supernatants from infected or mock-infected cells undergoing the above-mentioned treatments were collected at the different conditions for RNA extrac-  antimycotic (all from Life Technologies). The following day, FBS-supplemented DMEM was washed off and replaced by 2 mL fresh medium prior to polyethyleneimine (PEI)mediated transfection with a plasmid expressing either full-length spike protein (kindly provided by Dr. Jason S. McLellan, The University of Texas [80]) or with plasmids encoding the Spike protein receptor binding domain (RBD) amino acids 319 to 541 from the Wuhan strain (available at BEI Resources #NR-52309, https://www.beiresources. org/Catalog/BEIPlasmid Vectors/NR-52309.aspx, accessed on 1 May 2021) and from the beta, gamma (P1), and delta (synthesized by Genscript, Piscataway, NJ, USA). An empty vector was used as the control. One µg of each vector DNA was added to a final volume of 100 µL 150 mM NaCl solution containing 0.45 µg of PEI per µg of DNA. The mix was vortexed for 10 s, incubated for 10 min at room temperature, and evenly distributed in each well. Culture supernatants were removed 24, 48, 96, and 120 h after transfection, and 300 µL of BE buffer (HEPES 10 mM, SDS 1%, MgCl 2 .6H 2 O 1.5 mM, KCl 10 mM, DTT 1 mM, NP-40 0.1%) were added to each well to lyse the cells. The cell lysate was transferred to a 500 µL Eppendorf tube and frozen at −20 • C until use.

Western Blot
Proteins were extracted from cellular lysates and quantified using the Qubit Protein Assay Kit platform (Invitrogen) according to the manufacturer's instructions. A total of 15 µg of proteins were separated by SDS-PAGE and electro-transferred to PVDF membranes, which were directly incubated with blocking buffer (5% bovine serum albumin (BSA) in Tris-buffered saline (TBS) at 0.05% Tween-20 (TBST) for 1 h. Subsequently, the samples were incubated overnight with primary antibodies (Table 1) and washed three times with TBST. Then, the bands were incubated with the respective secondary antibodies for 1 h at room temperature. Immunoreactive bands were detected with the ChemiDoc XRS Imaging System equipment and protein quantification was performed using the ImageJ software. Graphs were plotted using GraphPad Prism version 8.1 software. Bands with statistically significant intensities among groups were evaluated by applying an Ordinary one-way ANOVA, with Tukey post hoc test (0.05 cut-off).

SARS-CoV-2 Infection Modulated the N-Degron Pathway and Increased ATE1 Enzyme Expression
To explore protein arginylation during SARS-CoV-2 infection, we performed an in silico multiomics data analysis and validated the findings in a time-course SARS-CoV-2 infection at the protein level by immunoblotting ( Figure 1A). Initially, the basal levels of enzymes involved in the N-degron pathway were evaluated in different uninfected cells ( Figure 1B). Enzymes involved in protein arginylation (ATE1), ubiquitination (UBR1, UBR2, UBR4, UBR5), arginine-tRNA ligase assembly (RARS2), deamidation (NTAN1), and N-terminal methionine removal (METAP1, METAP2) were identified in all uninfected cell models with no statistical difference among them. These findings indicated that enzymes involved in the protein arginylation pathway were not modulated based on the cell type or species in uninfected conditions. A total of 918 proteins ( Figure 1C) with potential to be arginylated at the N-terminus (NtE, NtD, NtC, NtN, NtQ), in agreement with the UniProt sequence, were identified in uninfected cell lines and showed a similar expression pattern ( Figure 1D), regardless of the organism (Green Monkey and Human).

SARS-CoV-2 Infection Modulated the N-Degron Pathway and Increased ATE1 Enzyme Expression
To explore protein arginylation during SARS-CoV-2 infection, we performed silico multiomics data analysis and validated the findings in a time-course SARS infection at the protein level by immunoblotting ( Figure 1A). Initially, the basal le enzymes involved in the N-degron pathway were evaluated in different uninfecte ( Figure 1B). Enzymes involved in protein arginylation (ATE1), ubiquitination ( UBR2, UBR4, UBR5), arginine-tRNA ligase assembly (RARS2), deamidation (NT and N-terminal methionine removal (METAP1, METAP2) were identified in al fected cell models with no statistical difference among them. These findings indicat enzymes involved in the protein arginylation pathway were not modulated based cell type or species in uninfected conditions. A total of 918 proteins ( Figure 1C) w tential to be arginylated at the N-terminus (NtE, NtD, NtC, NtN, NtQ), in agreemen the UniProt sequence, were identified in uninfected cell lines and showed a simi pression pattern ( Figure 1D), regardless of the organism (Green Monkey and Hum  Furthermore, we reanalyzed eight datasets covering transcriptomic and proteomic data of in vitro and in vivo SARS-CoV-2 infection of different biological systems [51][52][53][54][55][56][57] ( Figure 2A). ATE1 expression was higher during infection in most of the datasets, significantly upregulated at both transcript and protein levels [51,[54][55][56][57]. On the other hand, RARS1 and RARS2 protein expressions were opposite, with RARS1 being upregulated and RARS2 (mitochondrial) downregulated. UBR1, UBR2, and UBR5 ubiquitin-ligases (E3) expressions were increased in infection; however, UBR4 was regulated in a different direction at transcript (Wu et al.) and protein (Saccon et al.) levels. The expressions of proteins involved in the removal of the N-terminal methionine were variable among the different studies. However, protein expressions of the caspase family were upregulated, and especially CASP3 expression was statistically significant in four studies [51,53,54,56].
Western blot analysis was performed to measure the ATE1 level, which confirms the above findings ( Figure 2B). Consistent with the omics data, Calu-3 and Vero CCL-81 cells infected with SARS-CoV-2 had statistically higher ATE1 levels compared to the uninfected CTRL group ( Figure 2B). Time course data revealed an increase in ATE1 after 2 h of infection in Vero CCL-81 cells (p-value = 0.0294). On the other hand, statistical significance between groups was found after 48 h (p-value = 0.0259) in Calu-3 cells. It was found that the ubiquitin-conjugating E2 enzymes (UBE2G2, UBE2L3, UBE2D2, UBE2D3, UBE2K, UBE2D4, UBE2R2, UBA52, UBE2A, UBA3, UBE2W, UBE2L6, and UBE2E1) were also overexpressed in the infected groups (Supplementary File S1). To verify whether transfection with the spike protein of the Wuhan variant (WT) of SARS-CoV-2, instead of the whole virus, would be able to induce modulation of protein arginylation, HEK 293T cells were evaluated after 24, 48, and 96 h of infection ( Figure 2C). An increase in ATE1 was identified after 96 h of transfection compared to the CTRL empty vector group. Transient transfection with different receptor binding domains (RBD) can also modulate arginylation. We found that the DELTA variant has a more remarkable ability to induce ATE1 levels than the WT, BETA, and P1 variants after 96 h ( Figure 2D).

Increased ATE1 Expression in SARS-CoV-2 Infection Was Correlated with Events Linked to the Endoplasmic Reticulum (ER)
Once the increased abundance of ATE1 in the infection was confirmed, a multicorrelation analysis was performed using omic data to verify which proteins correlated with ATE1 ( Figure 3A), and which pathways could be associated with this increased abundance. Only differentially regulated proteins/genes were selected from six studies on SARS-CoV-2 [51,52,[54][55][56][57]. A total of 365 proteins/genes presented a significant correlation (p-value < 0.05) with ATE1 in at least two studies and 28 in at least three studies (Supplementary File S2). Analyzing the molecular functions (MF) of the 28 correlated proteins/genes, the enrichment of processes related to unfolded protein binding, protein-folding chaperone, and ubiquitin-protein ligase binding was found ( Figure 3B). Among the biological processes (BP), events related to ER and viral infection were enriched, such as protein target to ER, protein localization to ER, viral gene expression, and viral transcription ( Figure 3C). Pathways related to alterations in processes linked to RNA and coronavirus infection were also enriched ( Figure 3D). The GBP2 protein, involved in the cellular response to infections, was correlated with ATE1 in four studies. Due to the observed relationship between the processes linked to the ER ( Figure 3B,C), we monitored the direction of the correlation of HSPBP1 ( Figure 3E) and HSP90B1 ( Figure 3F) with ATE1. These proteins showed significant positive correlations, except for the negative correlation observed in lung tissue by Qiu et al. [54].
ubiquitin-ligases (E3) expressions were increased in infection; however, UB regulated in a different direction at transcript (Wu et al.) and protein (Saccon et al The expressions of proteins involved in the removal of the N-terminal methioni variable among the different studies. However, protein expressions of the caspas were upregulated, and especially CASP3 expression was statistically significan studies [51,53,54,56].   After identifying a relationship between ATE1 and ER-associated chaperones/processes during SARS-CoV-2 infection, the arginylation levels of proteins located in the ER, heat shock protein family A (Hsp70) member 5 (HSPA5, also known as BiP), calreticulin (CALR), and protein disulfide isomerase (PDI) were analyzed by Western blotting ( Figure  4). The BiP/HSPA5 arginylated protein level increased in both cell models over time with statistical significance 48 h after the onset of infection ( Figure 4A,B). Interestingly, the arginylated CALR protein level decreased in Calu-3 cells, while it increased in Vero CCL-81 ( Figure 4B) compared to the CTRL uninfected cells. PDI protein showed a significant increase 2 h after the onset of infection in Calu-3 cells, and was statistically more arginylated in infected Vero CCL-81 cells after 48 h ( Figure 4C). Arginylated proteins are targets of multiple pathways, including autophagy via binding to p62 and LC3B, as described by previous reports [34,81,82]. Based on this, we monitored and confirmed that Calu-3 and Vero CCL-81 cells present different modulation of the autophagy pathway when infected by SARS-CoV-2, especially of p92/SQSTM1 and LC3B proteins ( Figure S1). Thus, the substrates (CALR, BIP, and PDI) may present different behavior due to the modulation of the autophagy pathway. Furthermore, studies have shown that arginylated proteins can be relocated in the intracellular space [83] and be more accessible or inaccessible to degrada- After identifying a relationship between ATE1 and ER-associated chaperones/processes during SARS-CoV-2 infection, the arginylation levels of proteins located in the ER, heat shock protein family A (Hsp70) member 5 (HSPA5, also known as BiP), calreticulin (CALR), and protein disulfide isomerase (PDI) were analyzed by Western blotting (Figure 4). The BiP/HSPA5 arginylated protein level increased in both cell models over time with statistical significance 48 h after the onset of infection ( Figure 4A,B). Interestingly, the arginylated CALR protein level decreased in Calu-3 cells, while it increased in Vero CCL-81 ( Figure 4B) compared to the CTRL uninfected cells. PDI protein showed a significant increase 2 h after the onset of infection in Calu-3 cells, and was statistically more arginylated in infected Vero CCL-81 cells after 48 h ( Figure 4C). Arginylated proteins are targets of multiple pathways, including autophagy via binding to p62 and LC3B, as described by previous reports [34,81,82]. Based on this, we monitored and confirmed that Calu-3 and Vero CCL-81 cells present different modulation of the autophagy pathway when infected by SARS-CoV-2, especially of p92/SQSTM1 and LC3B proteins ( Figure S1). Thus, the substrates (CALR, BIP, and PDI) may present different behavior due to the modulation of the autophagy pathway. Furthermore, studies have shown that arginylated proteins can be relocated in the intracellular space [83] and be more accessible or inaccessible to degradation, resulting in a differential arginylation profile according to the substrate studied.  Searching for other organelles involved in arginylation during SARS-CoV-2 infection, we performed a subcellular localization analysis of proteins correlated with ATE1 in at least two studies ( Figure S2). These proteins mostly occupy complexes of chaperones, ribosomal, proteasome, cytoskeletal microtubules, and actin filament. Recently, Seo et al. [65] and Wong et al. [66] demonstrated experimentally that 152 were arginylated proteins, including mainly actins, chaperones, ribosomal components, and tubulins (Supplementary File S3), and nine proteins (VIM, HSPB1, PRDX4, ACTG1, ACTB, CALR, ATP5F1A, SPTAN1, and HSPA1B) overlapped in both studies. Bringing together arginylated proteins that were differentially regulated during SARS-CoV-2 infection and presented the same direction of regulation (upregulated or downregulated) in at least two studies ( Figure S2), we observed that tubulins and chaperones were increased in the infected group (INF); on the other hand, VIM and SPTAN1 proteins were downregulated. Collectively, data analysis of differentially regulated proteins pointed to an increased level of arginylated proteins in SARS-CoV-2 infection ( Figure S2B). Looking at the arginylated proteins evaluated here (Figure 4), we found that they are differentially regulated in distinct directions in the studies by Saccon et al. [51], Nie et al. [52], Wu et al. [56], and Leng et al. [53] ( Figure S2). The subcellular location of the 152 arginylated proteins was mainly in the cytoskeleton, cytoplasm, and nucleus ( Figure S2D). Since the ACTB protein was previously identified as arginylated by Seo et al. [65] and Wong et al. [66], Western blot analysis was performed to measure arginylated ACTB levels in infected Vero CCL-81 and Calu-3 cells ( Figure S2E

The Increase in ATE1 Levels Occurs Earlier with the Brazilian Variants P1 and P2 Compared to the Wuhan Variant (WT)
After verifying the modulation of arginylation resulting from infection by the Wuhan SARS-CoV-2 (WT) variant, Calu-3 cells were infected with the P1 and P2 variants, which were isolated for the first time in Brazil. In addition, 1 uM of the ER stress inducer thapsigargin was used ( Figure 5). The treatment with TAG induced the levels of ATE1, being possible to observe a significant increase in P1 in relation to the CTRL, WT, and P2 groups after 48 h of infection. The use of TAG induces stress even in the CTRL group. Untreated cells ( Figure 4B) show an increase or tendency to increase ATE1 only in infected groups (WT, P1, and P2). Furthermore, we verified that at 6 h (TAG-), it is possible to verify a significant increase in P2. The TAG-groups showed an increase in BiP/HSPA5 arginylation, as previously demonstrated ( Figure 4A,B). However, subjecting groups to TAG treatment, the observed behavior is the opposite. On the other hand, CALR showed increased levels of arginylation in both models (TAG+ and TAG-). Taken together, these data draw attention to a variant-dependent modulation of arginylation.

Transfection with Spike or RBD Highlights the Potential of SARS-CoV-2 to Induce Protein Arginylation
Transfection with Spike glycoprotein or RBD from different strains increased levels of BiP/HSPA5 arginylation ( Figure 6A,B). However, in the first 48 h after transfection with Spike, a decrease in BiP/HSPA5 arginylation was observed. The increase in levels occurred after 96 h (when there was an increase in ATE1). The DELTA variant RBD was able to induce R-BiP levels in the first 48 h and maintained this increase up to 120 h after infection. The CALR protein showed increased levels of arginylation in both transfections ( Figure 6A,B), as well as in the whole virus infection model (Figure 4). The transfections underscore the potential of SARS-CoV-2 or its viral particles to induce protein arginylation. Calu-3 cells infected with WT, P1, and P2 variants were evaluated at 6 hpi and 48 hpi when exposed to the stress inducer thapsigargin (A) or not exposed (B). Each point represents an independent experiment (n = 3). The significance level indicates: **** p < 0.0001; *** p < 0.001; ** p < 0.005; * p < 0.05.

Transfection with Spike or RBD Highlights the Potential of SARS-CoV-2 to Induce Protein Arginylation
Transfection with Spike glycoprotein or RBD from different strains increased levels of BiP/HSPA5 arginylation ( Figure 6A,B). However, in the first 48 h after transfection with Spike, a decrease in BiP/HSPA5 arginylation was observed. The increase in levels occurred after 96 h (when there was an increase in ATE1). The DELTA variant RBD was able to induce R-BiP levels in the first 48 h and maintained this increase up to 120 h after infection. The CALR protein showed increased levels of arginylation in both transfections ( Figure  6A,B), as well as in the whole virus infection model (Figure 4). The transfections underscore the potential of SARS-CoV-2 or its viral particles to induce protein arginylation. Calu-3 cells infected with WT, P1, and P2 variants were evaluated at 6 hpi and 48 hpi when exposed to the stress inducer thapsigargin (A) or not exposed (B). Each point represents an independent experiment (n = 3). The significance level indicates: **** p < 0.0001; *** p < 0.001; ** p < 0.005; * p < 0.05.

ATE1 Inhibition and Silencing Reduces SARS-CoV-2 Viral Release in Calu-3 Cells
In view of the close relationship between arginylation and SARS-CoV-2 infection demonstrated by previous data, enzyme inhibition assays by 1 µM tannic acid and 25 µM merbromine (MER) were performed on Calu-3 cells (Figure 7). These concentrations of tannic acid and MER did not affect cell viability. Notably, cells infected before any treatment (INF-24 h) had higher ATE1 expression than uninfected cells (CTRL), and treatment with tannic acid and MER significantly decreased the level of ATE1. Expression levels of ER proteins (CALR and BiP/HSPA5) decreased similarly to ATE1, but this was less pronounced. Of note, tannic acid and MER were able to reduce viral load or prevent virus entry into the cell. Such an effect was more relevant in the inhibition with MER ( Figure 7B).

ATE1 Inhibition and Silencing Reduces SARS-CoV-2 Viral Release in Calu-3 Cells
In view of the close relationship between arginylation and SARS-CoV-2 infection demonstrated by previous data, enzyme inhibition assays by 1 µM tannic acid and 25 µM merbromine (MER) were performed on Calu-3 cells (Figure 7). These concentrations of tannic acid and MER did not affect cell viability. Notably, cells infected before any treatment (INF-24 h) had higher ATE1 expression than uninfected cells (CTRL), and treatment with tannic acid and MER significantly decreased the level of ATE1. Expressio ER proteins (CALR and BiP/HSPA5) decreased similarly to ATE1, but this wa nounced. Of note, tannic acid and MER were able to reduce viral load or pre entry into the cell. Such an effect was more relevant in the inhibition with ME 7B). To confirm the reduction in viral load due to the decrease in ATE1 levels silencing assay was performed ( Figure 7C). It was possible to identify an incre abundance of ATE1 in the INF group in relation to the CTRL. In addition, there tion in ATE1 in the silenced groups, both infected and uninfected. Although efficiently silenced, the BiP/HSP5A protein increased arginylation levels in the I On the other hand, the CALR protein showed reduced arginylation. Viral load was confirmed ( Figure 7D) in the silenced infected group (INF-siATE1) in rela infected CTRL (INF-CTRL), confirming the results obtained with the ATE1 inhib finding strongly demonstrates that the reduction in ATE1 is correlated with the in viral release.
Reaffirming the previous findings, the anti-RBD antibody was used (Figu demonstrated lower band intensity in the INF-siATE1 group compared to the group. To confirm the reduction in viral load due to the decrease in ATE1 levels, an ATE1 silencing assay was performed ( Figure 7C). It was possible to identify an increase in the abundance of ATE1 in the INF group in relation to the CTRL. In addition, there is a reduction in ATE1 in the silenced groups, both infected and uninfected. Although ATE1 was efficiently silenced, the BiP/HSP5A protein increased arginylation levels in the INF group. On the other hand, the CALR protein showed reduced arginylation. Viral load reduction was confirmed ( Figure 7D) in the silenced infected group (INF-siATE1) in relation to the infected CTRL (INF-CTRL), confirming the results obtained with the ATE1 inhibitors. This finding strongly demonstrates that the reduction in ATE1 is correlated with the reduction in viral release.
Reaffirming the previous findings, the anti-RBD antibody was used ( Figure 7E) and demonstrated lower band intensity in the INF-siATE1 group compared to the INF-CTRL group.

Single Cell RNA-seq Data Showed That Macrophages and Epithelial Cells Express ATE1
After verifying the expression and subcellular location of proteins involved in the arginylation process, we investigated which cell types express ATE1. A reanalysis of single-cell RNASeq of nasopharyngeal/pharyngeal swabs samples published by Chua et al. [67] was conducted comparing the INF group consisting of critically ill patients, hospitalized for more than 20 days or who died from the progression of COVID-19 with the CTRL group of uninfected cases. A total of 17 cell clusters were identified ( Figure 8A). The gene sets of clusters 4, 7, and 8 presented a significant differential expression (p < 0.05) between the INF and CTRL groups ( Figure 8B). ATE1 was included in clusters 4, 9, 15, and 17 ( Figure S3A). The top five markers in cluster 4 were LYZ, SRGN, HLA-DPB1, CD74, and TYROBP, all markers of macrophages (Supplementary File S4). Cluster 4 also contained the macrophage markers MARCO [84], CD163 [85], MRC1 [86], and MSR1 [87], which reinforces the presence of macrophages in this cluster ( Figure S3). Based on these markers, the macrophage compartment was isolated in cluster 4, and the genes differentially regulated between the CTRL and INF groups were determined ( Figure S3B). The upregulated genes in cluster 4 were associated with interferon type I induction and signaling during SARS-CoV-2 infection, pulmonary fibrosis, proteasome degradation, and ferroptosis; the downregulated genes were related to peptide chain elongation, oxidative phosphorylation, and MHC class II complex ( Figure S4). However, the expressions of genes related to the modulation of arginylation: ATE1, CALR, ACTB, PDIA3, PDIA6, and PDIA4 did not show statistical significance between the groups ( Figure S3C); although, the BiP/HSPA5 gene expression was increased in the INF group.
We confirmed the absence of ATE1 modulation in macrophages at the protein level by Western blotting of macrophages infected with SARS-CoV-2 ( Figure 8C). These data suggested that the arginylation behavior in infected macrophages was different from that observed in Vero CCL-81 and Calu-3 cells. The inhibitors decreased ATE1 enzyme levels in macrophages at 48 h and 24 h after treatments. The expressions of ER chaperones, CALR and BIP, were significantly decreased in infected macrophages (48 h) compared to non-infected macrophages in Western blot assays ( Figure 8C). Looking at differentially regulated genes between the CTRL and INF groups in clusters 9, 15, and 17 ( Figure S5), we identified a statistically significant increase in ATE1 in the INF group in cluster 15, which was enriched mainly with epithelial cell markers confirming our data on Calu-3 cells (Supplementary File S4).

The N-Degron Pathway Was Regulated in SARS-CoV and MERS-CoV Infections but Not in H1N1 Influenza and Respiratory Syncytial Virus (RSV) Infections
We verified whether modulation of the N-degron pathway was recurrent in other respiratory viral infections or was a specific signature of SARS-CoV-2 ( Figure S6). The identification/regulation of proteins related to N-terminal methionine removal and ubiquitination was less recurrent in influenza, RSV, and human adenovirus infections. On the other hand, viruses of the Coronaviridae family, such as SARS-CoV and MERS-CoV, showed modulation of the proteins involved in these reactions. Convergently, ATE1 was not identified or was downregulated in in vitro models infected with H1N1 and RSV; in contrast, it was upregulated in cells infected with SARS-CoV and MERS-CoV. These data indicated an arginylation-dependent signature during infection with viruses from the Coronaviridae family.

Discussion
In this study, we described the modulation of the N-degron pathway and arginylation of proteins during SARS-CoV-2 infection by a combined in silico analysis of multiomic studies and data validation by Western blotting (Figure 1). We demonstrated an increase in ATE1 expression, a critical enzyme involved in arginylation, during in vitro SARS-CoV-2 infection (Figure 2). In fact, in human Calu-3 cells, a progressive increase in ATE1 expression was observed after 6 h of SARS-CoV-2 infection, when an increase in viral proteins has previously been demonstrated [55]. Interestingly, an increase in ATE1 expression was also observed in a monkey-derived cell line (Vero CCL-81), indicating that this modulation may occur independently of the species. The increase in ATE1 occurred when cells were infected with different variants, or transfected with the spike protein, or with RBDs. The P1 and P2 variants, first identified in Amazonas, Brazil [88], were considered to be of worldwide interest, and here we demonstrate that ATE1 activation occurred earlier in these variants than in WT (Wuhan). Moreover, the increase in ATE1 was demonstrated in MERS-CoV (24 h) and SARS-CoV (36 h) infections [60], but not in other respiratory virus infections such as RSV [63,89,90], and influenza [58,59], suggesting that involvement of the N-degron pathway may be a specific molecular signature for the Coronaviridae family. A recent study found lower levels of side chain arginylated peptides in the plasma collected from COVID-19 patients [91]. This apparent discrepancy compared to our study could be associated to the different biological systems under investigation, such as intracellular proteins in a cell line and mRNA levels of ATE1 in human broncholavage fluid compared to plasma circulating proteins. Moreover, our study focused on the N-terminal arginylation while the previous report analyzed side chain arginylation.
Previous studies by others and our group revealed an activation of the UPR pathway after 6 h of SARS-CoV-2 infection [17], corroborating ER stress enhancement [26] and UPR pathway activation during viral infection [18,19]. Based on these findings, we hypothesized that increased misfolded or unfolded proteins produced during SARS-CoV-2 infection may be tagged for degradation by arginylation in order to maintain cellular homeostasis. In fact, in silico multiomics analysis identified several ER related proteins associated with ATE1 expression (Figure 3). We showed that cells treated with the ER stress inducer, thapsigargin (TAG), kept ATE1 upregulation longer in the P1 variant, strengthening the relationship between arginylation and ER stress. As expected, BiP/HSPA5 expression increased 48 h after infection in both human and monkey cell lines. It has been shown that the viral spike glycoprotein plays a fundamental role in SARS-CoV-2 infection in the process of receptor recognition and cell membrane fusion [2], and it induced the transcriptional activation of Hsp90β member 1 and BiP/HSPA5 chaperones [14]. Increased expression of these chaperones has resulted in increased folding and processing of abundantly expressed proteins during SARS-CoV replication [20,92]. When cells were transfected with viral RBDs, we observed increased arginylation of BiP/HSPA5, especially in the DELTA variant. Recently, Zhang et al. showed that BiP/HSPA5 is an important therapeutic target in SARS-CoV-2 infection, as it can prevent viral binding and replication. In addition, four regions of the spike protein were predicted to bind BiP/HSPA5 [93]. Moreover, protein arginylation has also been induced by transient transfection of several dsDNAs, suggesting that its modulation may also be related to the detection of pathogenic dsDNA and activation of the immune system [94]. The CALR is a protein involved in the folding and maturation of glycoproteins [95,96]. CALR decrease has been described in other viral infections (influenza virus, SFV, or VSV) leading to accelerated maturation of cellular and viral glycoproteins, with a modest decrease in the folding efficiency [97]. We speculated that the progressive reduction in CALR levels in Calu-3 cells could be associated with an acceleration of coronavirus spike glycoprotein maturation. Based on previous observations, we speculate that the UPR pathway may be activated to restore ER homeostasis, and in the event of failure, apoptotic events would be induced [98]. Moreover, the virus may also hijack the host's ubiquitination machinery [99]; however, the biological functions of this action are still unknown.
Our multiomic analysis demonstrated that the arginylation-related proteins were mainly located in the ER, which is consistent with our in vitro results of the proteins involved in the UPR pathway. Additionally, the arginylation-related proteins were also located in the cytoskeleton, a key structure in the host-pathogen interaction [100]. Cytoskeleton proteins participate throughout the viral replication cycle, as SARS-CoV-2 enters into target cells using intermediate filament proteins, sequesters microtubules to transport itself to replication/assembly sites, and promotes the polymerization of actin filaments to exit the cell [101,102]. Moreover, cytoskeleton proteins were among the experimentally arginylated proteins identified in previous studies (Seo et al. [65] and Wong et al. [66]). We monitored the arginylation levels of one major cytoskeleton protein, ACTB, and observed its increase in the first 2 h after infection in Calu-3 cells, with a decrease at 48 h, when apoptosis-related proteins were activated [17]. The ACTB arginylation pattern was different in infected Vero CCL-81 cells, as was already observed for phosphorylation [103]. This observed difference stressed the importance of using multiple cell models to assess the cellular consequences of post-translational modifications in SARS-CoV-2 infection. Although arginylation of the ACTB protein has been explored [104][105][106], it has recently been shown that the N-terminal maturation of the protein is more complex than expected. It is believed that ACTB is terminated mainly by the enzyme NAA80, which acetylates the N-terminus of exposed Asp residues, and not by ATE1 [107].
Our analysis of single cell RNA-seq data highlighted macrophages as the cells within the clusters with differential gene expression levels in infected patients. In fact, the characterization of immune cells in bronchoalveolar lavage fluid has shown that pro-inflammatory monocyte-derived macrophages were more abundant in patients with severe SARS-Cov-2 infection than in those with moderate disease or healthy individuals. Furthermore, critically ill patients presented a lower proportion of myeloid dendritic cells, plasmacytoid dendritic cells, and T cells than patients with moderate infection [108]. We also demonstrated the expression of the ATE1 protein in macrophages; however, no difference in the abundance of proteins related to arginylation was detected comparing the CTRL and INF groups. In fact, a previous study has shown that SARS-CoV-2 was capable of infecting macrophages without causing any cytopathic effect, and the virus was also capable of inducing host immunoparalysis [109]. Moreover, we also identified a cellular compartment with epithelial cell markers presenting a significant increase in ATE1 expression, consistent with the previous observation of ATE1 protein expression in the main lung epithelial cells, ciliary cells type 1 and 2, infected by SARS-CoV-2 [110]. ATE1 expression in lung epithelial cells was higher in SARS-CoV-2 infected patients compared to controls.
Notably, we found that arginylation inhibitors, tannic acid and MER, decreased the viral load or prevented viral entry into the cell. Furthermore, our assays indicated a decrease in the abundance of ATE1. Tannic acid was recently described as a potent inhibitor of SARS-CoV-2 through the thermodynamically stable binding to the Mpro and TMPRSS2 proteins, crucial for the entry of the virus into the cell [111]. Here, we confirm the potential of tannic acid to reduce viral load, and furthermore, to modulate ATE1 levels during infection. In addition, we also demonstrated that, like tannic acid, the arginylation inhibitor MER was also capable of reducing both viral load and ATE1 level. Suramin treatment was shown to inhibit SARS-CoV-2 binding to the receptor, entry, and viral replication in Vero CCL-81 and Calu-3 cells [112]. Interestingly, suramin was also shown to inhibit ATE1 activity [113], confirming the results obtained in our study. It is important to mention that tannic acid and merbromin can modulate several intracellular signaling pathways and influence viral replication, independently of the levels of ATE1. Indeed, exposure to ATE1 inhibitors has been shown to activate/deactivate several pathways [111]. Therefore, to elucidate the direct influence of ATE1 on viral load reduction, we performed silencing of ATE1 in infected Calu-3 cells and confirmed the direct influence of the enzyme levels on viral replication, thus presenting the arginylation pathway as an important mechanism in SARS-CoV-2 infection.

Conclusions
Here, we elucidated the role of protein arginylation and modulation of the N-degron pathway during SARS-CoV-2 infection. Differential regulation of proteins involved in all reactions that make up the N-degron pathway was demonstrated, with emphasis on the upregulation of the ATE1 protein, evidenced by omics and Western blot data. Furthermore, the whole virus is able to promote this increase in ATE1 levels, but also transfection with the spike protein and RBD regions of the virus. We verified that variants declared as being of worldwide interest, P1 and DELTA, present higher levels compared to the classic variant (WT). We also showed that proteins that have their levels correlated with ATE1 perform biological functions linked to chaperone activity and binding to unfolded proteins. In fact, the ER stress inducer, TAG, increased ATE1 levels. Importantly, our findings revealed that modulation of the N-degron pathway differs between different types of infected cells, such as macrophages, Vero CCL-81, HEK 293T, and Calu-3 cells. Finally, we showed the importance of this process by reducing viral load using tannic acid and MER, known arginylation inhibitors. To strongly evidence the relationship between arginylation and SARS-CoV-2 infection, we showed that ATE1 silencing induces viral load reduction.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/v15020290/s1, Figure S1. Modulation of autophagy proteins. Representative Western blot images of p62 and LC3B proteins in Calu-3 and Vero CCL-81 cells after 2 h, 6 h, 12 h, 24 h and 48 h of infection (Wuhan strain). Figure S2. Subcellular localization of arginylation-related proteins. Figure S3. Expression of ATE1, RARS2, CD163, MARCO, MRC1, and MSR1 in cell clusters identified by reanalysis of single-cell RNA-seq data. Figure S4. Pathways and cellular components related to upregulated genes in cluster 4. Figure S5. Cell markers, ATE1 and polyubiquitins UBB or UBC expressions in cluster 9. Figure   Data Availability Statement: The datasets generated during and/or analyzed during the current study are available in the public repositories as described in the Materials and Methods section in the paragraph entitled: Data sources and curation.