key: cord-0317662-9cxqa0aq
authors: Poston, Daniel; Weisblum, Yiska; Hobbs, Alvaro; Bieniasz, Paul D.
title: VPS29 exerts opposing effects on endocytic viral entry
date: 2021-08-08
journal: bioRxiv
DOI: 10.1101/2021.08.06.455441
sha: 62c55134a99a3ff10a7bc6aee69141c213d6fd35
doc_id: 317662
cord_uid: 9cxqa0aq

Emerging zoonotic viral pathogens threaten global health and there is an urgent need to discover host and viral determinants influencing infection. We performed a loss-of-function genome-wide CRISPR screen in a human lung cell line using HCoV-OC43, a human betacoronavirus. One candidate gene, VPS29, was required for infection by HCoV-OC43, SARS-CoV-2, other endemic and pandemic threat coronaviruses as well as ebolavirus. However, VPS29 deficiency had no effect on certain other viruses that enter cells via endosomes and had an opposing, enhancing effect on influenza A virus infection. VPS29 deficiency caused changes endosome morphology, and acidity and attenuated the activity of endosomal proteases. These changes in endosome properties caused incoming coronavirus, but not influenza virus particles, to become entrapped therein. Overall, these data show how host regulation of endosome characteristics can influence viral susceptibility and identify a host pathway that could serve as a pharmaceutical target for intervention in zoonotic viral diseases.

confirmed knock-out (KO), both by sequencing target loci and by western blot analyses ( Figure  134 S1A). Importantly, KO of these genes did not affect cellular viability or proliferation. Because 135 viral dependency factors identified via CRISPR screening might be required in a cell-type 136 specific manner, we evaluated the requirement of these genes for infection in multiple cell lines 137 expressing ACE2 (the receptor for both SARS-CoV-2 and HCoV-NL63): A549-ACE2, HT1080-138 ACE2, and 293T-ACE2. 139 140

Given their function in endosomal trafficking, we hypothesized that these hits would most likely 141 affect viral entry. We therefore performed single-cycle infection assays and quantified infected 142 cells via flow cytometry. There was strong requirement for VPS29/CCC complex as well as 143 WDR81/91 in A549 cells for all CoVs tested (Figure 2A-D) . However, the was no requirement 144 these factors in for IAV, adenovirus, or RSV infection of A549 cells ( Figure 2E -G). In all other 145 cell lines tested, there was a strong requirement for VPS29 for all coronaviruses but no 146 dependency on VPS29 or the other candidate proteins was found for adenovirus and RSV 147

( Figure We found that VSV infection was unaffected by VPS29 KO ( Figure 2U ). Because the sole 158 difference between rVSV/SARS-CoV-2 and VSV itself is that rVSV/SARS-CoV-2 enters cells 159

using the SARS-CoV-2 spike protein in lieu of VSV-G, these data suggest that it is the entry 160 pathway that imposes the requirement for VPS29. Given the strong requirement for VPS29 by all 161

tested HCoVs, in all cell lines tested, we sought to further confirm the relevance of VPS29 to 162

HCoV infection. To do so, we used CRISPR/Cas9 to KO VPS29 in normal human bronchial 163 epithelial (NHBE) primary lung cells. Loss of VPS29 strongly inhibited HCoV-OC43 infection 164

in NHBE cells ( Figure 2V ), suggesting that VPS29 is important for HCoV infection of 165 physiologically relevant cells. 166 167

In contrast to effects on coronavirus infection, we observed precisely the opposite effect of 168 VPS29 or CCC complex deficiency on IAV infection in HT1080-ACE2 and 293T-ACE2 cells. 169 That is, KO of VPS29 or CCC complex components enhanced IAV infection ( Figure 2M ,S) 170 while WDR81/91 KO had no effect. To confirm the phenotype observed using the IAV strain 171

A/WSN/33, we analyzed two separate strains of 2009 pandemic H1N1 IAV; 172

A/Netherlands/602/2009 (H1N1)pdm09 (H1N12009 Netherlands) and A/California/04/2009 173 (H1N1)pdm09 (H1N12009 California). We found that the ability of VPS29 KO to enhance IAV entry 174 was conserved in the pandemic IAV strains ( Figure 2W ,X). That the same set of endocytic 175

factors could promote infection of coronaviruses while antagonizing IAV infection indicates 176

endosome-based viral entry pathways are influenced by specific sets of host proteins that can 177 facilitate or restrict viral entry. 178 179 HCoV-229E, and rVSV/SARS-CoV-2 infection ( Figure 3A -D). These data strongly suggest that 189 the participation of VPS29 in the Retromer complex (VPS26A/VPS29/VPS35), which is 190 recruited to endosomes via Rab7A, is the means by which it facilitates CoV infection (Rojas et 191 al., 2008 Figure 4A ,B and S2). Importantly, there was a return to normal 236 endosome phenotype after reconstitution with wildtype VPS29, confirming that this effect is due 237

to VPS29 KO ( Figure 4C and S2). The appearance of enlarged, deacidified vesicles was 238 maintained in VPS29 KO cells reconstituted with VPS29I91D or VPS29L152E ( Figure 4D ,E and 239 S2), suggesting that this phenotype is due to retromer disfunction. Quantification of the pH-240

sensitive Dextran signal from these images revealed a 3.7-fold decrease in fluorescence intensity 241 in VPS29 KO cells ( Figure 4F ) that is rescued upon reconstitution with WT VPS29, but not with 242 VPS29I91D or VPS29L152E. Importantly, the enlarged endosomes in VPS29 KO cells exhibited 243 equivalent fluorescent intensity to endosomes in normal cells when cells were incubated with 244 pH-insensitive AF-488 Dextran, indicating that while they were deacidified, they were not 245 impaired in cargo loading ( Figure 5A , B and S3). 246 247 VPS29 KO results in entrapment of rVSV/SARS-CoV-2 in endosomes 248

Given the above findings, we hypothesized that CoV infection is impaired in VPS29 KO cells 249 due to impediment in spike dependent egress from endosomes. To test this idea, we generated 250 rVSV/SARS-CoV-2NG-P, a replication-competent chimeric VSV expressing SARS-CoV-2 Spike 251 protein in lieu of VSV-G, and containing the VSV structural protein P fused to mNeonGreen 252 (NG-P), thus enabling the direct observation of entering viral particles (Schott et al., 2005) . 253 254

At 60 minutes post infection of parental HT1080 cells few NG-P punctae were evident within 255 2xFYVE-mSCAR labeled endosomes, suggesting successful egress of most rVSV/SARS-CoV-256 2NG-P particles ( Figure 6A and S4A) and minimal accumulation therein. However, in VPS29 KO 257 cells, enlarged endosomes contained many rVSV/SARS-CoV-2NG-P punctae at 60 min after 258 infection. Likewise, when cells were infected in the presence of labeled Dextran and imaged 60 259 minutes post infection, we observed a similar phenotype with rVSV/SARS-CoV-2 particles 260 accumulated in enlarged, Dextran-containing vesicles in VPS29 KO cells ( Figure 6B and S4B). 261

Overall, these data indicate that the major inhibitory effect of VPS29 KO on CoV infection is the 262 result of failed egress from endosomes. 263 264

Similar experiments in which incoming IAV virions were detected by immunofluorescence 60 265 min after ( Figure 6C and S5) revealed that IAV particles did not accumulate in the enlarged 266 2xFYVE-mSCAR labeled endosomes in VPS29 KO cells. Thus, the effect of VPS29 KO on 267 rVSV/SARS-CoV-2 was indeed specific. In fact, there was significantly greater association 268 between incoming IAV and 2xFYVE-labeled endosomes in parental HT1080 cells as compared 269 to VPS29 KO cells ( Figure 6C and S5), mirroring the opposing effects of VPS KO on HCoV and 270 IAV infection. 271 272 We hypothesized that such effect might be due to VPS29-dependent trafficking of antiviral 273 proteins with activity against IAV to endosomes, such as IFITM3. We observed that IFITM3 274 knockdown enhanced IAV infection of parental HT1080 cells ( Figure S6A ), in agreement with 275 previous reports (Feeley et al., 2011) . However, IFITM3 knockdown augmented IAV infection 276 in VPS29 KO cells ( Figure S6A ), suggesting that the enhancement of IAV infection in VPS29 277

KO cells was not the result of loss of IFITM3 activity. Concordantly, IFITM3 localized to 278 2xFYVE-labeled endosomes in both WT and VPS29 KO cells, and there was no clear difference 279

in localization ( Figure S6B ). Overall. these finding suggest that enhanced IAV infection in 280

VPS29 KO cells is due to increased egress from endosomes but is not due to altered localization 281 and/or impaired activity of IFITM3. 282 283

cathepsin activity 285

The aforementioned findings indicate that the reduced susceptibility to HCoV infection in 286

VPS29 KO cells is spike-specific and is the consequence of failed egress from endosomes. We 287

hypothesized that this effect could be due to impaired spike processing by endosomal proteases 288 during entry. We used HIV-1-based pseudotyped viruses to test the susceptibility of various CoV 289 spikes to VPS29 KO and cathepsin inhibition using the drug E64d. As rVSV/SARS-CoV-2 bears 290 a point mutation, R683G, which ablates the polybasic furin cleavage site, we tested pseudotypes 291

bearing WT or R683G mutant spike proteins, as well as spike proteins from SARS-CoV and 292 SARS-like CoV from bats and pangolins, which also do not contain polybasic cleavage sites 293 (Coutard et al., 2020) . 294 295

Pseudotypes bearing both the WT and R683G mutant SARS-CoV-2 spike proteins were sensitive 296 to VPS29 KO and cathepsin inhibition. However, cathepsin inhibition did not further decrease 297 infection of VPS29 KO cells ( Figure 7A ). The SARS-CoV-2R683G (Figure7B), SARS-CoV 298 (Figure7C), and the SARS-like bat ( Figure 7D ) and pangolin viruses ( Figure 7E ,F) that lack furin 299 cleavage sites were more impacted by VPS29 KO and cathepsin inhibition than WT SARS-CoV-300 2. Indeed, in several instances, VPS29 KO and/or cathepsin inhibition resulted in undetectable 301 infection by SARS-CoV-2R683G, SARS-CoV, and the SARS-like bat/pangolin CoVs. Similarly, 302

infectivity assays utilizing rVSV/SARS-CoV-2 also revealed a dose-dependent inhibition of 303 infectivity upon cathepsin inhibition in parental HT1080, but no impairment of infection upon 304 cathepsin inhibition in VPS29 KO cells ( Figure 7G ). 305 306

That there was no further effect of cathepsin inhibition on CoV infection in VPS29 KO cells 307

suggests that the effect of these two manipulations converge on a common pathway in promoting 308 egress from endosomes. We thus hypothesized that VPS29 KO impedes CoV infection by 309

impairing proper processing of spike by cathepsins. If this were indeed the case, then VPS29 KO 310 should impair infection mediated by ebolavirus (EBOV) glycoprotein (GP), which is known to 311 require processing by endosomal cathepsins (Schornberg et al., 2006) . To test this, we performed 312 infectivity assays in WT and VPS29 KO cells using a recombinant VSV expressing EBOV GP in 313 lieu of VSV-G (rVSV/EBOV-GP) (Mulherkar et al., 2011) . Indeed, we observed a strong 314 inhibition of rVSV/EBOV-GP with both cathepsin inhibition and loss of VPS29 ( Figure 7H ). 315

This result suggests that the susceptibility of VPS29 KO is mediated by impaired cathepsin 316 activity. 317 318

Consistent with the above conclusion, when parental HT1080 cells were treated with the 319 cathepsin inhibitor E64d, infected with rVSV/SARS-CoV-2 NG-P and examined microscopically, 320

we observed a phenotype similar to that seen in VPS29 KO cells (see Figure 6A ). Specifically, 321

substantially more rVSV/SARS-CoV-2NG-P punctae were evident within endosomes, and the 322 endosomes appear enlarged with similar appearance and morphology to those observed in 323 VPS29 KO cells ( Figure 7I and S7). To directly test whether VPS29 KO results in impaired 324 endosomal cathepsin activity, we measured endosomal cathepsin activity in WT and VPS29 KO 325

HT1080 cells using a substrate that generates a fluorescent signal upon cleavage by cathepsin L. 326

Indeed, in WT cells, we observed a strong red fluorescence signal in vesicular structures, 327

indicating high levels of cathepsin activity. However, in VPS29 KO cells the red fluorescence 328 signal was nearly absent, indicative of impaired cathepsin activity in VPS29 KO cells ( Figure  329 7J). To determine if the loss of cathepsin L activity was the result of failed trafficking of 330 cathepsins to the endolysosomal system, we performed immunofluorescence studies utilizing 331 tagged cathepsin L in cells with endosomes labeled with 2xFYVE-mSCAR. There was no 332 change in cathepsin L localization to 2xFYVE-mSCAR-positive endosomes in VPS29 KO cells 333

( Figure 7K ). These data suggest that the loss of cathepsin activity in VPS29 KO cells is not a 334 result of impaired trafficking of cathepsin itself to endosomes, but rather change endosomal 335 conditions in VPS29 KO cells, such as increased pH, reduces cathepsin activity therein. 336 337 DISCUSSION 338

While the advent of robust, high-throughput screening modalities has generated a wealth of 339 information regarding host-viral interactions, the underlying mechanism of action for many host 340 proteins implicated by these screens remain incompletely understood. Here, utilizing HCoV-341 OC43 as a model HCoV, we employed a genome-wide loss-of-function CRISPR screen to 342 identify and characterize factors required for efficient CoV infection. In particular, we show that 343 the retromer subunit protein VPS29 is required for productive infection by diverse CoVs in a 344 variety of cell types. Other genome-wide screens using SARS-CoV-2 have also suggested a role 345

for VPS29 and the CCC as well as RAB7A, which recruits retromer to endosomes ( in C, as well as the additional representative images in Supplemental Figure S5 ). 498

Error bars indicate SD. Statistical test: Student's T test. 499 representation (the resulting library contained 0.0% undetected guides and a skew ratio of the top 566

10% represented guides to the bottom 10% represented guides was 3.94, well below the 567 recommended cutoff of 10 for an "ideal" library (Joung et al., 2017) ). To generate lentiviral 568 preparations of the Brunello library, 293T cells (6 x 10 6 cells per 10 cm dish) were transfected 569 with 6µg lentiCRISPRv2-Brunello, 6µg NL-gagpol, and 1.2 µg VSV-G using PEI. 48 hours post 570 transfection, supernatants were pooled and concentrated using Amicon Ultra Centrifugal Filters. 571

Concentrated lentiviral preps were stored at -80°C and titrated on A549 cells based on 572 puromycin resistance. Briefly, 10-fold serial dilutions (from 10 -1 to 10 -6 ) were used to transduce 573 40,000 A549 cells in a 24 well plate format. 48 hours post transduction, cells were trypsinized 574 and moved up to 6 well plates in the presence of 1.25 µg/mL puromycin. 9 days post 575 transduction, cells were fixed, stained with Crystal Violet, and stained foci were counted to 576 measure the number of cells surviving selection (i.e. those that were transduced with 577 lentiCRISPRv2 harboring a puromycin resistance cassette). To perform the screen, 1.3 x 10 8 578 A549 cells were transduced with lentiCRISPRv2-Brunello at an MOI of 0.3 in order to generate 579 a population of single KO cells at high (>500X) coverage. Two days post transduction, cells 580

were placed in selection with 1.25 µg/mL puromycin and passaged for 7 days, until there were 581 no untransduced cells remaining. Thereafter, in triplicates with 8x10 6 cells per flask, A549-582

Brunello cells were infected or not with HCoV-OC43 at an MOI of 0. Pathway Analysis of screen hits 592

All 34 candidate genes were searched using the STRING database (https://string-db.org) for 593 functional enrichment of protein-protein interactions using default settings, except the minimum 594 required interaction score was changed from medium confidence (0.400) to high confidence 595 (0.700). Subsequently, genes were annotated with UniProt keywords (https://uniprot.org) 596 597

Validation of CRISPR hits 598

Individual sgRNAs targeting hits of interest were cloned into lentiCRISPRv2 via linearization 599

with BsmBI followed by ligation of annealed oligos with compatible sticky ends using primers: 

10% Transmission for the DAPI channel. For co-Dextran-treated cells, images were 708 acquired on a DeltaVision OMX SR imaging system using a 60X Widefield oil immersion 709 objective (Olympus) with an exposure time of 25ms, 10.0% Transmission for the AF-488 710 channel, an exposure time of 50ms, 10% Transmission for the A568 channel

Microscopy of rVSV/SARS-CoV-2 infected cells 714 Cells were plated in a Nunc Lab-Tek II Chamber Slide (Thermo) at 5x10 3 cells per well. The 715 next day, cells were transduced with 2xFYVE-mSCAR to label endosomes

48 hours post transduction cells were treated with 5µM E64d (Sigma Aldrich 717 E8640-250UG) for 30 minutes, followed by inoculation with rVSV/SARS-CoV-2NG-P at an MOI 718 of 2. 60 minutes post infection, cells were washed 3x with PBS and fixed in 4% PFA

For cells with 2x-FYVE labeled endosomes, 722 images were acquired on a DeltaVision OMX SR imaging system using a 60X Widefield oil 723 immersion objective (Olympus) with an exposure time of 50ms, 10% Transmission for the AF-724 488 channel, an exposure time of 100ms, 10% Transmission for the A568 channel, and an 725 exposure time of 150ms, 10% Transmission for the DAPI channel. For cells with Dextran Red 726 labeled endosomes, images were acquired on a DeltaVision OMX SR imaging system using a 727 60X Widefield oil immersion objective (Olympus) with an exposure time of 50ms, 10% 728 Transmission for the AF-488 channel, an exposure time of 50ms, 10% Transmission for the 729 A568 channel, and an exposure time of 200ms, 10% Transmission for the DAPI channel. 730 731 Influenza virus immunofluorescence 732 Cells were plated in a Nunc Lab-Tek II Chamber Slide (Thermo) at 5x10 3 cells per well. The 733 next day, cells were transduced with a construct expressing 2xFYVE-mSCAR to label 734 endosomes. 48 hours post transduction, cells were infected with IAV at an MOI of ~10. 60 735 minutes post infection, cells were washed with PBS

Images were acquired on a DeltaVision OMX SR imaging system using a 60X Widefield oil 739 immersion objective (Olympus) with an exposure time of 50ms, 5.0% Transmission for the AF-740 488 channel, an exposure time of 100ms, 10% Transmission for the A568 channel

Quantification of fluorescence microscopy 744 For each cell, Regions Of Interest (ROIs) corresponding to labeled endosomes were defined 745 using the freehand selection tool in Fiji. Quantification of mean fluorescence intensity inside 746 each ROI was determined using the Measure command. For punctae quantification, the number 747 of punctae inside each ROI was counted and summed to give the total

ROIs for each cell. Additionally, the total number of punctae outside of ROIs in each cell was 749 measured. The reported % of virus

VPS29, a tweak tool of endosomal 790 recycling

The 792 hemagglutinin: a determinant of pathogenicity

Retromer function in endosome

Golgi retrograde transport is regulated by the yeast Vps34 PtdIns 3-kinase

Vps29 has a 797 phosphoesterase fold that acts as a protein interaction scaffold for retromer assembly

The 800 species Severe acute respiratory syndrome-related coronavirus: classifying 2019-nCoV and 801 naming it SARS-CoV-2

The 803 spike glycoprotein of the new coronavirus 2019-nCoV contains a furin-like cleavage site absent 804 in CoV of the same clade

Retromer has a selective function in cargo sorting via endosome transport 807 carriers

Identification of Required Host Factors for 810 SARS-CoV-2 Infection in Human Cells

APEX2-mediated RAB proximity labeling 813 identifies a role for RAB21 in clathrin-independent cargo sorting

Optimized sgRNA design to maximize activity 816 and minimize off-target effects of CRISPR-Cas9

IFITM3 inhibits influenza A virus infection by 819 preventing cytosolic entry

Role of heparan sulfate in the Zika virus entry, replication, and cell death

Localization of phosphatidylinositol 3-phosphate in yeast and 824 mammalian cells

The cell biology of receptor-mediated virus entry

The cargo-selective retromer complex is a recruiting hub for protein complexes that 829 regulate endosomal tubule dynamics

Functional 832 interrogation of a SARS-CoV-2 host protein interactome identifies unique and shared 833 coronavirus host factors

A stabilized respiratory syncytial virus reverse genetics system amenable to 836 recombination-mediated mutagenesis

pH-induced conformational transitions 838 of the propeptide of human cathepsin L. A role for a molten globule state in zymogen activation

Structural and mechanistic insights into regulation of the 842 retromer coat by TBC1d5

VSV-Displayed HIV-1 Envelope Identifies Broadly 845 Neutralizing Antibodies Class-Switched to IgG and IgA

Genome-scale CRISPR-Cas9 knockout and transcriptional 848 activation screening

MX2 is an interferon-induced 851 inhibitor of HIV-1 infection

Human coronavirus circulation in the United States

Host cell 856 entry of respiratory syncytial virus involves macropinocytosis followed by proteolytic activation 857 of the F protein

Endocytosis of influenza viruses. Microbes 859 Infect

Airway proteases: an emerging drug target for influenza 861 and other respiratory virus infections

MAGeCK enables robust identification of essential genes from genome-864 scale CRISPR/Cas9 knockout screens

Haploid Genetic Screen Identifies Heparan Sulfate-Associated Genes and the 868

Macropinocytosis Modulator TMED10 as Factors Supporting Vaccinia Virus Infection

Virus entry: open sesame

Retriever is a multiprotein complex for retromer-873 independent endosomal cargo recycling

Adenovirus endocytosis

Human 876 coronavirus NL63 utilizes heparan sulfate proteoglycans for attachment to target cells

Host cell proteases: Critical determinants of 879 coronavirus tropism and pathogenesis

The Ebola 881 virus glycoprotein mediates entry via a non-classical dynamin-dependent macropinocytic 882 pathway

A genome-wide 885 CRISPR screen identifies a restricted set of HIV host dependency factors

COMMD1 is linked to the WASH 889 complex and regulates endosomal trafficking of the copper transporter ATP7A

Direct binding of retromer to human papillomavirus type 16 893 minor capsid protein L2 mediates endosome exit during viral infection

Regulation of retromer 897 recruitment to endosomes by sequential action of Rab5 and Rab7

Optimized libraries for CRISPR-Cas9 900 genetic screens with multiple modalities

Measuring SARS-CoV-2 903 neutralizing antibody activity using pseudotyped and chimeric viruses

Genome-scale 906 identification of SARS-CoV-2 and pan-coronavirus host factor networks

Role of 908 endosomal cathepsins in entry mediated by the Ebola virus glycoprotein

An antiviral role for the 911 RNA interference machinery in Caenorhabditis elegans

Sialic acids as receptor determinants for 914 coronaviruses

Retromer-mediated endosomal 916 protein sorting: all WASHed up!

Endosomal PI(3)P regulation by the 919 COMMD/CCDC22/CCDC93 (CCC) complex controls membrane protein recycling

STRING v11: protein-protein association 923 networks with increased coverage, supporting functional discovery in genome-wide experimental 924 datasets

UniProt: a worldwide hub of protein knowledge

Proteolytic processing of filovirus glycoproteins

Genetic Screens 932 Identify Host Factors for SARS-CoV-2 and Common Cold Coronaviruses

Identification of a minimal size requirement 935 for termination of vesicular stomatitis virus mRNA: implications for the mechanism of 936 transcription

Fusion of enveloped viruses in endosomes

The polybasic cleavage site in the SARS-CoV-2 941 spike modulates viral sensitivity to Type I interferon and IFITM2

A genome-wide CRISPR screen identifies host factors that regulate SARS-CoV-2 entry

lentiCRISPRv2, which does not harbor an sgRNA cassette, was used. Lentiviral preparations 607were obtained by transfecting 1x10 6 293Ts with 1µg lentiCRISPRv2, 1µg NL-gagpol, and .2µg 608 VSV-G using PEI. 2 days post transfection, supernatants were collected, filtered, and used to 609 transduce 5x10 4 A549-ACE2, HT1080-ACE2, 293T-ACE2, or NHBE cells. 2 days post 610 transduction, cells were trypsinized, placed in selection with 1.25µg/mL puromycin, and 611passaged until there were no remaining viable untransduced cells. CRISPR Raw FASTQ files were aligned to the Brunello library and scored using the MAGeCK statistical 782package. All flow cytometry data were analyzed using FlowJo software, version 10.6.1. All 783 graphs were generated using GraphPad Prism, version 8. Error bars correspond to the standard 784 deviation. All images were generated by maximum intensity projection using Fiji (https://fiji.sc/).