key: cord-0827364-8htn36c5
authors: Sandoval, Daniel R.; Clausen, Thomas Mandel; Nora, Chelsea; Cribbs, Adam P.; Denardo, Andrea; Clark, Alex E.; Garretson, Aaron F.; Coker, Joanna K.C.; Narayanan, Anoop; Majowicz, Sydney A.; Philpott, Martin; Johansson, Catrine; Dunford, James E.; Spliid, Charlotte B.; Golden, Gregory J.; Payne, N. Connor; Tye, Mark A.; Nowell, Cameron J.; Griffis, Eric R.; Piermatteo, Ann; Grunddal, Kaare V.; Alle, Thibault; Magida, Jason A.; Hauser, Blake M.; Feldman, Jared; Caradonna, Timothy M.; Pu, Yuan; Yin, Xin; McVicar, Rachael N.; Kwong, Elizabeth M.; Weiss, Ryan J.; Downes, Michael; Tsimikas, Sotirios; Smidt, Aaron G.; Ballatore, Carlo; Zengler, Karsten; Evans, Ron M.; Chanda, Sumit K.; Croker, Ben A.; Leibel, Sandra L.; Jose, Joyce; Mazitschek, Ralph; Oppermann, Udo; Esko, Jeffrey D.; Carlin, Aaron F.; Gordts, Philip L.S.M.
title: The Prolyl-tRNA Synthetase Inhibitor Halofuginone Inhibits SARS-CoV-2 Infection
date: 2021-03-26
journal: bioRxiv
DOI: 10.1101/2021.03.22.436522
sha: 4a63fcf7d182d395957a9eb01aa106a5bec893c2
doc_id: 827364
cord_uid: 8htn36c5

Summary Paragraph We identify the prolyl-tRNA synthetase (PRS) inhibitor halofuginone1, a compound in clinical trials for anti-fibrotic and anti-inflammatory applications2, as a potent inhibitor of SARS-CoV-2 infection and replication. The interaction of SARS-CoV-2 spike protein with cell surface heparan sulfate (HS) promotes viral entry3. We find that halofuginone reduces HS biosynthesis, thereby reducing spike protein binding, SARS-CoV-2 pseudotyped virus, and authentic SARS-CoV-2 infection. Halofuginone also potently suppresses SARS-CoV-2 replication post-entry and is 1,000-fold more potent than Remdesivir4. Inhibition of HS biosynthesis and SARS-CoV-2 infection depends on specific inhibition of PRS, possibly due to translational suppression of proline-rich proteins. We find that pp1a and pp1ab polyproteins of SARS-CoV-2, as well as several HS proteoglycans, are proline-rich, which may make them particularly vulnerable to halofuginone’s translational suppression. Halofuginone is orally bioavailable, has been evaluated in a phase I clinical trial in humans and distributes to SARS-CoV-2 target organs, including the lung, making it a near-term clinical trial candidate for the treatment of COVID-19.

halofuginone inhibited authentic SARS-CoV-2 replication post-entry. Inhibition of prolyl-tRNA synthetase (PRS) activity was responsible for both the HS-dependent and the HS-independent antiviral properties of halofuginone. These findings identify halofuginone as a translational regulator of cell surface HS biosynthesis and a potentially potent, orally bioavailable, therapeutic for the treatment and prevention of COVID-19.

To identify potential agents that attenuate SARS-CoV-2 spike protein binding, we screened a library of small-molecule antagonists of host epigenetic regulators and protein translation elements of the prolyl-tRNA synthetase complex, along with their chemically matched inactive analogs [16] [17] [18] . The compound library has been successfully used to investigate targets mediating anti-inflammatory or anti-proliferative effects in a variety of biological contexts 16 (Fig. 1a) . As a positive control cells were treated with a mixture of heparin lyases, which digest HS chains 3 . Class 1 histone deacetylase inhibitors, such as Romidepsin (1 µM) or Belinostat (5 µM), decreased SARS-CoV-2 RBD binding by ~50% compared to excipient control (DMSO; Fig 1a) . Halofuginone (1 µM) reduced binding by 84%, a level of reduction similar to the impact of heparin lyase treatment (Fig 1a) . Halofuginone targets the prolyl-tRNA synthetase (PRS) active site of the human glutamyl-prolyl tRNA synthetase (EPRS) and has been used in clinical and preclinical studies to treat fibrotic disease and to attenuate hyperinflammation 1, 19 . We then treated Vero E6 African green monkey kidney epithelial cells, Caco-2 human epithelial colorectal adenocarcinoma cells, and Calu-3 human epithelial lung adenocarcinoma cells with halofuginone. Halofuginone reduced RBD protein binding to Hep3B, Calu-3 and Caco-2 but showed very modest effects in Vero E6 cells (Fig. 1b-e) . The degree of inhibition was greatest in the Calu-3 line (5-fold reduction in RBD binding; Fig. 1e ). To test the effect of halofuginone on the binding of spike protein in a more native presentation, we examined the ability of halofuginone to inhibit infection of SARS-CoV-2 spike protein pseudotyped VSV in Hep3B cells. Halofuginone inhibited infection by up to ~30-fold in a dose-dependent manner (Fig.   1f ). Consistent with the inability of halofuginone to reduce RBD binding to Vero E6 cells (Fig. 1c) , no inhibition of infection was seen in these cells (Fig. 1g) . We next tested if halofuginone inhibits the infection of authentic SARS-CoV-2 in Hep3B. Cells were treated with halofuginone at different doses for 24 h prior to and during infection (MOI of 0.1) (Fig. 1h) 20 

The binding of SARS-CoV-2 spike protein to cells is HS-dependent 3 . HS is a linear polysaccharide attached to serine residues in HS proteoglycans (HSPGs) 21 . The HS polysaccharides consist of alternating residues of N-acetylated or N-sulfated glucosamine (GlcNAc or GlcNS) and either glucuronic acid (GlcA) or iduronic acid (IdoA) (Fig. 2a) . Hence, we tested if halofuginone reduces spike protein binding by reducing HS presentation at the cell surface. Hep3B and Vero E6 cells were treated for 18 h with halofuginone and the effects on cellular HS were evaluated using the monoclonal antibody (mAb) (10E4) that recognizes a common epitope in HS ( Fig. 2b & Extended data Fig. 3) . Halofuginone dose-dependently reduced 10E4 binding in Hep3B (Fig. 2c) , whereas 10E4 binding increased in Vero E6 upon halofuginone treatment (Fig. 2d) . This directly correlates with the level of spike RBD binding and S protein pseudotyped virus infection in these cells (Fig. 1b-c) . This finding suggests that halofuginone inhibits spike protein binding and SARS-CoV-2 viral attachment by altering cell surface HS content. Analysis of Hep3B cells showed that 0.5 µM halofuginone reduced cell surface HS in Hep3B cells by ~4-fold (Fig. 2e) . No difference was seen in the disaccharide composition following halofuginone treatment, suggesting that halofuginone affects total HS synthesis but does not alter HS specific sulfation (Fig. 2f) . To test if the observed decrease in total HS was due to a decrease in availability of HSPG core proteins to carry HS, we lysed treated Hep3B cells, ran the lysates on SDS-PAGE and stained using an mAb (3G10) that recognizes a neo-epitope remaining on the proteoglycans (Fig. 2b,g) 22 . The analysis revealed that halofuginone reduced the expression of core HSPGs in a dose-dependent manner ( Fig. 2g & Fig. 4) . These data suggest that halofuginone inhibits HS-mediated binding of the SARS-CoV-2 spike protein to cells by inhibiting the expression of HSPGs.

To broadly examine the effects of halofuginone treatment, Hep3B cells were treated with vehicle or halofuginone at 200 or 500 nM for 6 h and 18 h and processed for RNA-sequencing (RNA-Seq). Halofuginone profoundly impacted the transcriptome (~2700 differentially expressed genes), which increased with longer incubation time (Extended data Fig. 5a-b) . Principal component analysis clearly segregated the response with respect to the duration of treatment (Extended data Fig. 5a ). Analysis of halofuginone upregulated genes did not identify common pathways associated with antiviral or inflammatory responses (Extended data Fig. 5c-d) 23 .

Additionally, no significant difference in expression of host factors exploited by SARS-CoV-2, such as TMPRSS2 and ACE2, were observed (Extended data Fig. 5f-g) . Gene annotation analysis of downregulated genes at 6 h showed that 200 nM halofuginone altered the expression of genes involved in glycoprotein biosynthesis and proteoglycan metabolic processes (Extended data Fig.   5e ) 23 . A select group of core HSPGs were downregulated at the mRNA level (Extended data Fig.   5f -g). In particular, HSPGs GPC2 and SDC1 were significantly downregulated in conjunction with the HS biosynthetic enzymes B3GAT3 and EXTL3 (Extended data Fig. 5f-g & 6) . Taken together, the data demonstrates that halofuginone suppresses the expression of proteins involved in HS and HSPG production.

Aminoacyl-tRNA-synthetases (AARS) catalyze the ATP-dependent synthesis of aminoacylated tRNAs via a two-step reaction involving an aminoacyl-adenylate intermediate with subsequent transfer to the cognate tRNAs 24 . Halofuginone and chemically unrelated compounds like prolyl-sulfamoyl adenosine (ProSA), specifically and potently inhibit human prolyl-tRNA synthetase (PRS) by blocking distinct portions of their ligand binding pockets (Fig. 3a-c) 15, 25 . PRS inhibition can specifically suppress translation of proteins, such as collagens, that are enriched in prolines while having minimal effects on general protein synthesis 15 . However, PRS inhibition can also lead to GCN2-mediated activation of the Integrated Stress Response (ISR). GCN2 (gene symbol EIF2AK4) senses uncharged tRNAs and phosphorylates the eukaryotic transcription initiation factor 2a (eIF2a), leading to a general reduction in 5'cap-mediated RNA translation and selective translation of the eukaryotic transcription factor ATF4 and its target genes (Fig. 3a) Interestingly, halofuginone induced a general ATF4-mediated ISR but did not activate the unfolded protein-induced ER stress response, as reported previously (Extended data Fig. 5d ) 27 .

To date, alterations in HS biosynthesis have not been identified as a hallmark of the ISR. To better understand the relationship between PRS inhibition and HS expression we tested related PRS inhibitor analogs to deconstruct the PRS pathway in relation to HS biosynthesis and spike RBD binding. Treatment with the non-cleavable and highly selective prolyl-AMP substrate analog ProSA at 5 µM prevented HS presentation and spike RBD binding to similar levels as treatment with halofuginone (500 nM), as illustrated by mAb 10E4 stain ( Fig. 3c -d) 28 . Additionally, halofuginol (HFol), a halofuginone derivative that inhibits PRS, significantly decreased 10E4 and spike RBD binding while the MAZ1310 negative control compound, had no effect (Fig. 3d-e) 15, 25 .

Collectively, the data demonstrates that PRS inhibition is sufficient to reduce HS biosynthesis and spike RBD binding (Fig. 3d-e) . Halofuginone competes with proline for the PRS active site 15 .

Addition of excess proline to the media (4mM) of Hep3B cells prevented the ability of halofuginone to inhibit RBD protein binding to the cells (Fig. 3f ).

To determine if halofuginone suppresses HS biosynthesis by activating the ISR we coincubated halofuginone in the presence of a general inhibitor of the ISR, ISRIB (Integrated Stress Response inhibitor), selective eIF2a kinase inhibitors GCN2-IN-1 (GCN2i) targeting GCN2 (general control nonderepressible 2), or GSK2606414 (PERKi) targeting eIF2a kinase 3 (eIF2AK3), also known as protein kinase R-like endoplasmic reticulum kinase (PERK) (Fig.   3A ) 26, 29 . Neither PERKi nor ISRIB affected 10E4 or RBD binding, and neither had an effect on halofuginone inhibition (Fig. 3g-h) . In contrast, GCN2i reduced 10E4 and RBD binding almost to the same extent as halofuginone activity (Fig. 3g-h) , suggesting a role for GCN2 in the regulation of HS biosynthesis independent of the ISR. Neither GCN2-IN-1, GSK2606414, nor ISRIB reversed the halofuginone induced reduction in 10E4 or spike protein binding, suggesting that halofuginone does not suppress HS biosynthesis and spike binding by activating the ISR (Fig. 3g- h).

PRS inhibitors can selectively modulate the translational efficiency of proline-rich proteins, such as collagens (Extended data Fig. 7) . Interestingly, HSPGs, such as agrin, perlecan, collagen 18 and syndecans 1 and 3, are relatively proline-rich compared to other proteins, such as TMPRSS2, ACE2 and lysosomal or cholesterol biosynthetic proteins ( Fig. 3i and Extended data Fig. 7-9 ; Extended data Table 1 ). These observations suggest that inhibition of prolyl-tRNA charging could impair production of key membrane and extracellular matrix HSPGs ( Fig. 3i and   2g ). Together, these data suggest that PRS inhibitors, such as halofuginone, inhibit HS and proteoglycan expression both at the translational and transcriptional level.

To determine if halofuginone inhibition of SARS-CoV-2 virus production was due to reduced viral entry or subsequent intra-host replication, Huh 7.5 cells were treated with 100 nM halofuginone or vehicle, either before, after, or before and after infection with SARS-CoV-2.

Similarly, as in Hep3B halofuginone prevented productive infection of Huh 7.5 without affecting cell viability ( Fig. 4a -b, and Extended data Fig. 10 ). Pretreatment with halofuginone alone significantly reduced SARS-CoV-2 infection by ~30-fold (Fig. 4a) . However, halofuginone added after viral infection reduced the amount of secreted infectious virions by nearly 1000-fold (Fig. 4a) .

Intracellular viral RNA did not change when the cells were only treated before infection as expected, but viral RNA levels dramatically decreased 10-to 100-fold when halofuginone was present after infection (Fig. 4b) . These observations suggest that halofuginone inhibits SARS-CoV-2 viral replication in addition to suppressing HS-dependent infection.

Halofuginone did not decrease cellular HS or the binding of recombinant RBD protein in Vero E6 cells. However, given the effects of halofuginone treatment on viral replication we examined the effect of halofuginone on infection by authentic SARS-CoV-2 in Vero E6 cells as well. Halofuginone completely inhibited SARS-CoV-2 infectivity at 50 nM with an IC50 of 13 nM as measured by immunofluorescent (IF) detection of the nucleocapsid protein and plaque assays ( Fig. 4c-d & Extended data Fig. 11 ). In comparison, Remdesivir had a calculated IC50 of 8 µM.

Thus, halofuginone showed a ~1,000-fold more potent inhibition of infection as compared to Remdesivir in this experimental setup ( Fig. 4c-d) . Similarly, halofuginone was 100-fold more potent compared to chloroquine (IC50 1.9 µM) (Extended data Fig. 12) . Moreover, halofuginone treatment reduced SARS-CoV-2 spike intracellular mRNA levels more than 20,000-fold with an IC50 of 34.9 nM (Fig. 4e) .

We next examined if the impact of halofuginone on viral replication was dependent on PRS inhibition. Halofuginone competes with proline for the PRS active site 15 . Addition of excess proline to the media (4mM) of SARS-CoV-2 infected Vero E6 cells increased the IC50 of halofuginone from 12.5 nM to 210 nM (Fig 4f) . Commercial halofuginone is a racemate of two, 2S,3R-(-) and 2R,3S-(+) enantiomers 30 . Hence, we evaluated if viral inhibition was due to ontarget effects of the PRS targeting 2R,3S-(+) enantiomer (Fig. 4g) 30 . The 2R,3S-(+) inhibited SARS-CoV-2 infectivity with an IC50 of 12 nM compare to an IC50 of 28 nM for the racemic mixture.

No inhibition was observed when using the non-targeting 2S,3R-(-) (Fig. 4g) . Analogous, only the active 2R,3S-(+) enantiomer inhibited HS production and authentic SARS-CoV-2 spike RBD protein binding (Extended data Fig. 13-14) .

CoV-2 infection of Vero E6 cells, although at significantly higher doses (Fig. 4h) . The negative control compound, MAZ1310, did not (Fig. 4h) . These results demonstrate that inhibition of PRS activity suppresses SARS-CoV-2 infection (Fig. 4h) . Inhibitors of the ISR, as well as inhibitors of eIF2-alpha kinases PERK and GCN2 were unable to attenuate the halofuginone-mediated inhibition of SARS-CoV-2 infectivity (Fig. 4i) . Hence, the observed antiviral effect is not dependent on ISR activation, consistent with the effects seen on the binding of recombinant RBD spike protein and mAb 10E4 binding ( Fig. 3c-d) . 31,32 Upon cell entry, SARS-CoV-2 genomic RNA is translated into two large polyproteins, pp1a (>400 kDa) and pp1ab (>700 kDa) that undergo proteolytic processing into 11 or 16 non-structural proteins, respectively, many of which are required for RNA replication. Both pp1a (R1A) and pp1ab (R1AB), and to a lesser degree spike protein, have high proline contents comparable to collagens ( Fig. 4j & Extended data Fig. 15 ;

Extended data table 1). To probe the importance of the viral protein proline content versus general mRNA translation inhibition we evaluated two other AARS inhibitors, borrelidin and serylsulfamoyl adenosine (SerSA) that target the threonyl-and seryl-tRNA-synthetase, respectively.

Niether of these AARS inhibitors were able to significantly attenuate viral replication ( There is a desperate need for a potent, orally bioavailable antiviral for the treatment of COVID-19 that can be administered early in the disease to prevent hospitalization and the development of severe pulmonary disease. Here, we demonstrate that halofuginone is a potent Expressed as area under the curve, halofuginone exposure was more than 87-fold higher in the lung compared to plasma after a single intravenous injection in mice 36 . This suggests that although it may be difficult to obtain halofuginone plasma levels significantly above the IC50 values determined in this study, doses tested in phase I trials may be sufficient to achieve significant anti-SARS-CoV-2 activity in the lung and other organs infected by SARS-CoV-2 37-40 .

Excessive inflammation can contribute to inflammatory organ injury during severe COVID-

and low dose dexamethasone recently demonstrated lower 28-day mortality in individuals receiving respiratory support 41 . In addition to the antiviral activity against SARS-CoV-2 that is described here, halofuginone has anti-inflammatory and anti-fibrotic activity that may provide yet another additive benefit to individuals with COVID-19 pneumonia 1,2,15,42-44 .

We demonstrate that other PRS inhibitors, ProSA and halofuginol, also inhibit infection by authentic SARS-CoV-2. Our data suggests that PRS inhibition may be particularly effective at suppressing the production of SARS-CoV-2 polyproteins pp1a and pp1ab that are long and proline-rich. Other positive-sense ssRNA viruses produce long polyproteins that may be similarly sensitive to PRS or other AARS inhibitors. Consistent with this idea, halofuginone demonstrates antiviral activity against future deleterious coronaviruses and other positive ssRNA viruses, including Chikungunya virus, Dengue virus, Zika virus, and Sindbis virus (Extended data Fig. 17 ;

Extended data table 1) 45 .

In addition to halofuginone, other PRS inhibitors are being evaluated in humans, including DWN12088 that is in phase I clinical trials in Australia and anticipated to be used for the treatment of interstitial pulmonary fibrosis. Thus, evaluation of PRS inhibitors, halofuginone analogs, and other AARS inhibitors may lead to the identification of additional broad-spectrum antiviral agents.

Viruses require host cell resources for replication and thus can be inhibited by therapeutics that target essential host factors. This approach could provide broad activity against diverse viruses while decreasing the risk of emerging viral resistance as the therapy is not directed against a specific virally encoded product. On the other hand, targeting host factors can lead to cytostatic and cytotoxicity, and we did observe the reported anti-proliferative effect of halofuginone in our assays (Extended data Fig. 17 ) 2 . Although inhibition of global translation could lead to excessive toxicity, inhibiting AARSs could provide a more precise way of inhibiting viral proteins and thus limit toxicity. It appears that normal cells are relatively tolerant of decreased AARS levels with minimal effects on global translation. Individuals who are heterozygous carriers for recessive inactivating AARS mutations linked to hypomyelinating leukodystrophy do not display disease phenotypes 46 .

In conclusion, we identified halofuginone as an antiviral with potent inhibitory activity against SARS-CoV-2 infection in multiple human cell types. We show that halofuginone reduces both HS and HSPG biosynthesis, that are required for viral adhesion. In addition, we find that the inhibitory capacity of halofuginone on SARS-CoV-2 infection is not limited to HS reduction, but also shows potent inhibition in the viral replication stage. Mechanistically, our data suggests that these effects are caused by PRS inhibition and its effect on production of proline-rich proteins and not dependent on ISR activation. This observation is in agreement with previous studies reporting that coronaviruses overcome the inhibitory effects of eIF2a phosphorylation on viral mRNA translation 31,32 . Furthermore, many viruses, including coronaviruses (SARS1 and SARS2), shutdown host translation using nonstructural protein 1 (Nsp1) and activate eIF2a associated PERK dependent stress responses that benefit the virus 47 . Halofuginone may prevent SARS-CoV-2 from circumventing these mechanisms of host protein translational shutdown.

Halofuginone oral administration has been evaluated in a phase I clinical trial in humans and based on pharmacological studies in mice is distributed to SARS-CoV-2 target organs, including the lung. In conclusion, we show that halofuginone is a potent inhibitor of SARS-CoV-2 infection which emphasizes its potential as an effective treatment for COVID-19 in the clinic.

Beyond its antiviral activity, halofuginone has potent anti-inflammatory and anti-fibrotic properties that may provide additive benefit in the treatment of COVID-19 pneumonia. Based on this in vitro preclinical data we believe halofuginone could be an effective antiviral and antifibrotic agent for the treatment of individuals with COVID-19. and Vero E6 cells were from ATCC. The Hep3B cells carrying mutations in HS biosynthetic enzymes was derived from the parent ATCC Hep3B stock, and have been described previously streptomycin sulfate and grown under an atmosphere of 5% CO2 and 95% air. Cells were passaged before 80% confluence was reached and seeded as explained for the individual assays. Cell viability was measured using LDH leakage (Promega) or alamarBleu (Thermo Fisher). 

For binding studies, recombinant SARS-CoV-2 Spike RBD protein (RBD, His Tag) was conjugated with EZ-Link TM Sulfo-NHS-Biotin (1:3 molar ratio; Thermo Fisher) in Dulbecco's PBS at room temperature for 30 min. Glycine (0.1 M) was added to quench the reaction and the buffer was exchanged for PBS using a Zeba spin column (Thermo Fisher).

Cells at 50-70% confluence were treated with halofuginone or other inhibitors for 18-24 hrs. The 

Fresh Hep3B cells were was washed in PBS, lifted with trypsin, and centrifuged for 5 min a 500

x g. The trypsin supernatants were saved for GAG analysis, and the cell pellets were washed in incubated for 2 hr at 37°C, shaking. Finally, the GAGs were purified over a DEAE column and precipitated by PD-10 chromatography 3 .

For HS quantification and disaccharide analysis, purified GAGs were digested with a mixture of heparin lyases I-III (2 mU each) for 2 hr at 37 °C in lyase buffer (40 mM ammonium acetate and 

Cells were lysed using RIPA buffer, and protein was quantified using a BCA assay. Protein was analyzed by SDS-PAGE on 4-12% Bis-Tris gradient gels (NuPage; Invitrogen) with an equal amount of protein loading. Proteins were visualized after transfer to Immobilon-FL PVDF membrane (Millipore). Membranes were blocked with Odyssey blocking buffer (LI-COR Biosciences) for 30 min and incubated overnight at 4°C with 3G10 antibodies. Mouse antibodies were incubated with secondary Odyssey IR dye antibodies (1:14,000) and visualized with an Odyssey IR imaging system (LI-COR Biosciences).

Cells were lysed in Trizol and total RNA was extracted using the Direct-zol kit ( 

A computational pipeline was written calling scripts from the CGAT toolkit to analyse the RNA sequencing data (https://github.com/cgat-developers/cgat-flow) 4,5 . Briefly, FASTQ files were generated and assessed for quality using FASTQC, aligned to GRCh38 (hg38) and then aligned to the transcriptome using hisat2 v2.1.0 6 . To count mapped reads to individual genes, featurecounts v1.4.6, part of the subreads package 7 , was used. Differential gene expression analysis was performed using DESeq2 using treatment and time as factors in the model. Genes were considered to be differentially expressed based on log2 fold change and p-value < 0.05. R scripts used to analyze the transcriptomic data are available through GitHub (https://github.com/Acribbs/deseq2_report). Motif enrichment was performed using homer as described before 8, 9 . 

Sindbis virus (SINV) with mCherry-tagged NSP3 was propagated on BHK-15 cells 11 . Uganda 

Protein amino acid sequences were downloaded from UniProtKB 13 . Proline distribution was analyzed using custom code in R (v3.6.0) 14 , including commands from the packages dplyr 15 tidyr 16 , and stringr 17 . All plots were generated in ggplot2 18 . For individual protein plots, histograms were constructed with geom_histogram(binwidth = 1) and kernel density estimations (KDEs) were constructed with geom_density(kernel = "gaussian") and added as custom annotations (ggpubr package 19 ). The bandwidth of KDEs for individual plots was assessed separately for each protein distribution using maximum likelihood cross-validation 20,21 with the h.mlcv command from the kedd package 22 . The proline distribution score was calculated using the following formula:

where BP is the number of 10-amino acid blocks in a protein that contain one or more prolines (a protein with length 100 amino acids has 10 blocks); BT is the total number of 10-amino acid blocks in a protein; and P is the total number of prolines in a protein. BP values were obtained using custom code in R. 

Highly permissive cell lines for subgenomic and genomic hepatitis C virus RNA replication

Corrigendum to: Triglyceride-rich lipoprotein binding and uptake by heparan sulfate proteoglycan receptors in a CRISPR/Cas9 library of Hep3B mutants

Current protocols in molecular biology

CGAT-core: a python framework for building scalable, reproducible computational biology workflows

CGAT: computational genomics analysis toolkit

Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype

The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads

Simple combinations of lineage-determining transcription factors prime cisregulatory elements required for macrophage and B cell identities

Generation of VSV pseudotypes using recombinant DeltaG-VSV for studies on virus entry, identification of entry inhibitors, and immune responses to vaccines

Spatial and Temporal Analysis of Alphavirus Replication and Assembly in Mammalian and Mosquito Cells

Rescue of the 1947 Zika Virus Prototype Strain with a Cytomegalovirus Promoter-Driven cDNA Clone

UniProt: a worldwide hub of protein knowledge

R: A language and environment for statistical computing

dplyr: A Grammar of Data Manipulation

Tidy Messy Data

Simple, Consistent Wrappers for Common String Operations

Elegant Graphics for Data Analysis

ggpubr: 'ggplot2' Based Publication Ready Plots

A stepwise discrimination analysis program using density estimation

On the choice of smoothing parameters of Parzen estimators of probability density functions

Kernel estimator and bandwidth selection for density and its derivatives