Shared host genetic landscape of respiratory viral infection

Abstract

Respiratory viruses represent a major global health burden. Although these viruses have different life cycles, they may depend on common host genetic factors, which could be targeted by broad-spectrum host-directed therapies. We used genome-wide CRISPR screens and advanced data analytics to map a network of host genes that support infection by nine human respiratory viruses [influenza A virus, parainfluenza virus, human rhinovirus, respiratory syncytial virus, human coronavirus (HCoV)-229E, HCoV-NL63, HCoV-OC43, Middle East respiratory syndrome–related coronavirus, and severe acute respiratory syndrome–related coronavirus 2]. We explored shared pathways using knowledge graphs to inform on pharmacological targets. We selected and validated STT3A/B proteins of the N-oligosaccharyltransferase complex as host targets of broad-spectrum antiviral small molecules. Our work highlights the commonalities of viral host genetic dependencies and the feasibility of using this information to develop broad-spectrum antiviral agents.

Full text

PNAS 2025 Vol. 122 No. 20 e2414202122 https://doi.org/10.1073/pnas.2414202122 1 of 8
RESEARCH ARTICLE
|
Significance
There is a constant threat of
epidemics or pandemics of
respiratory viruses. The most
common treatment for
respiratory viral diseases is to
target viral proteins, but viruses
can evolve resistance quickly.
Host-directed therapy (HDT) can
decrease the potential of
emerging drug-resistance and can
increase the possibility of
developing a broad-spectrum
drug that targets multiple viruses,
including emerging new viruses.
To develop HDT against
panrespiratory viruses, we used
genome-wide CRISPR screens and
advanced data analytics to map a
network of host genes that
support infection by nine clinically
important human respiratory
viruses. We explored shared
pathways and found
pharmacological targets. The
development of eective and
broad-spectrum HDT could have a
dramatic impact on how we treat
and prevent infectious diseases.
This article is a PNAS Direct Submission.
Copyright © 2025 the Author(s). Published by PNAS.
This open access article is distributed under Creative
Commons Attribution- NonCommercial- NoDerivatives
License 4.0 (CC BY- NC- ND).
1L.B.S., D.R.B., I.B., and A.P. contributed equally to this work.
2Present address: Department of Pathology and
Immunology, Washington University School of Medicine,
Saint Louis, MO 63105.
3Present address: Department of Internal Medicine,
University of Texas Southwestern Medical Center, Dallas,
TX 75390.
4Present address: Broad Institute of Massachusetts
Institute of Technology and Harvard, Cambridge, MA
02142.
5To whom correspondence may be addressed. Email:
hwangseu@broadinstitute.org or atelenti@trailbiomed.
com.
This article contains supporting information online at
https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.
2414202122/- /DCSupplemental.
Published May 15, 2025.
MICROBIOLOGY
Shared host genetic landscape of respiratory viral infection
LeahB.Soriagaa,1 , DaleR.Balcea,1, IstvanBarthaa,1 , ArnoldParka,1, EmilyWonga, MichaelMcAllastera, ElizabethA.Muellera, OnaBarauskasa,
EstebanCarabajala, BeatrizKowalskia, SooyoungLeea, GaryLoa, TaraF.Mahoneya, MatteoMetruccioa, AnnaSahakyana, LogeshwaranSomasundarama,
TodSteinfelda, LishaWanga, LauraWedela, SamanthaS.Yima, LiYina, JiayiZhoua, ZachNewbya, WinstonTsea, JohannesGrossea,
HerbertW.Virgina,2,3 , SeungminHwanga,4,5 , and AmalioTelentia,5
Aliations are included on p. 7.
Edited by Karla Kirkegaard, Stanford University, Palo Alto, CA; received July 17, 2024; accepted March 1, 2025
Respiratory viruses represent a major global health burden. Although these viruses have
different life cycles, they may depend on common host genetic factors, which could be tar-
geted by broad- spectrum host- directed therapies. We used genome- wide CRISPR screens
and advanced data analytics to map a network of host genes that support infection by nine
human respiratory viruses [influenza A virus, parainfluenza virus, human rhinovirus, res-
piratory syncytial virus, human coronavirus (HCoV)- 229E, HCoV- NL63, HCoV- OC43,
Middle East respiratory syndrome–related coronavirus, and severe acute respiratory syn-
drome–related coronavirus 2]. We explored shared pathways using knowledge graphs to
inform on pharmacological targets. We selected and validated STT3A/B proteins of the
N- oligosaccharyltransferase complex as host targets of broad- spectrum antiviral small
molecules. Our work highlights the commonalities of viral host genetic dependencies
and the feasibility of using this information to develop broad- spectrum antiviral agents.
functional genomics | CRISPR | genome- wide | respiratory virus | host- directed therapy
Viral disease can be prevented and treated by vaccines, interferons and cytokines, antiviral
small molecules, and antibodies. All areas have made great progress over the years, most
recently through messenger ribonucleic acid vaccines, therapeutic monoclonal antibodies,
as well as new or improved small-molecule agents for HIV, hepatitis C virus, and severe
acute respiratory syndrome–related coronavirus 2 (SARS-CoV-2). Many of these
approaches are virus-specic; thus, there is ongoing interest in targeting additional viral
diseases and aspiration to develop broad-spectrum antiviral agents ( 1 ). One plausible path
to expand the antiviral armamentarium would be through targeting of host proteins that
support viral replication or antagonize one or multiple viral families ( 2 – 5 ).
In recent years, the most common approach to identify cellular requirements for viral
infection has been large scale screens using small interfering ribonucleic acids, CRISPR
screens, and proteomics ( 3 – 14 ). Such approaches generally interrogate individual viral
pathogens. e downstream analyses for such datasets include ranking of genes/proteins
associated with experimental readouts, higher-level proling of cellular pathways and
integrating with external knowledge. Finally, top ranked genes can be validated genetically,
using RNA interference, or chemically, using tool compounds.
e present work lays the foundation for broad-spectrum host-directed therapy of multiple
respiratory viral diseases with single medicines by interrogating multiple respiratory viruses
associated with a high healthcare burden and/or the danger of new pandemics ( 15 – 17 ): inu-
enza A virus (IAV), parainuenza virus (PIV), human rhinovirus (HRV), respiratory syncytial
virus (RSV), SARS-CoV-2, Middle East respiratory syndrome–related coronavirus
(MERS-CoV), and three human coronaviruses (HCoV-229E, HCoV-NL63, HCoV-OC43).
We implemented genome-wide screening and a data science analytical strategy to build a
comprehensive map and knowledge graph of host genetic dependencies for these major res-
piratory viruses. We optimized the screening conditions and adjusted selective pressure using
ow cytometry-based analyses to identify genes impacting viral entry and replication and cell
viability-based analyses to discover genes aecting the completion of viral replication cycle and
the cellular response to viral replication. Multiple cell lines from several animal species (mouse,
nonhuman primate, human) were used to support replication of the dierent viral families.
Results
Validation of Individual Screens. Twelve genome- wide screens using the CRISPR/Cas9
(clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR- associated
protein 9) system were performed across the nine respiratory viruses (multiple cell lines for
IAV; two variants of MERS- CoV) (SIAppendix, TableS1). For each screen, control and
OPEN ACCESS

2 of 8 https://doi.org/10.1073/pnas.2414202122 pnas.org
selection conditions were assessed by proling the guide coverage
(reads per guide, guides with zero counts), guide distribution
(Gini coecient), and fold- change between selection and control
conditions for control nontargeting guides. Genes for which
targeting guides were enriched under screen selection (e.g., lower
viral replication, better survival of infected cells) versus control
conditions, indicating loss- of- function, were deduced as proviral.
Conversely, genes for which targeting guides were depleted in
the same selection condition were deduced as antiviral. Genes
were scored for guide enrichment or depletion with hit calling at
FDR- adjusted P- value < 0.05. As few as 20 and as many as 442
genes were scored as proviral across screens (Fig.1 and Datasets
S1 and S2). e wide range in hit counts across the screens could
be due to variation in selection pressure as well as dierences
in host cell biology under screening. We validated the screen
selection conditions by conrming known viral receptors or viral
entry factors as the top hits in each screen (Fig.1). For example,
the known receptor ICAM1 (intercellular adhesion molecule
1) was scored as the top proviral hit for HRV (18); similarly,
ACE2 (angiotensin converting enzyme 2) for SARS- CoV2
and HCoV- NL63 (19, 20); DPP4 (dipeptidyl peptidase 4) for
MERS- CoV (21); ANPEP (alanyl aminopeptidase, membrane)
for HCoV- 229E (22). For IAV, SLC35A1 (solute carrier family
35 member A1) was the top proviral hit for each cell line screened,
which was identied to be essential for IAV receptor expression
and therefore entry (23). In addition, for each screen, gene-
set enrichment analysis identied several genes within known
functional protein complexes based on CORUM annotation
(a resource of manually annotated protein complexes from
mammalian organisms, https://mips.helmholtz- muenchen.de/
corum/). ese annotated complex subunits included top proviral
genes from the same screen, bolstering support for their shared
function as viral host factors (DatasetS3). Gene- set enrichment
analysis (DatasetS4) also pointed to known response pathways
to infection based on KEGG annotation (SIAppendix, Fig.S1).
To identify high-level similarity of proviral gene patterns across
viruses and screens, we applied principal component analysis
(PCA) on the normalized screen rank scores based on enrichment
(SI Appendix, Fig. S2 A and B ). We speculated that clustering of
viruses would reveal shared patterns of host gene dependencies.
In support of true biology, we observed tight colocalization of
score proles for the two dierent variants of MERS-CoV screened
under similar conditions in the same cell line in the top two PCs.
We also observed colocalization of the proles for common-cold
Fig. 1. Prole of individual CRISPR ko screen. Signicance
scores are presented as –log10(FDR) on the y- axis for genes
distributed along the x- axis (ordered by gene rank). Top
10 genes by FDR in each screen are labeled (ties method,
rst). Points colored as hit enrichment (red) for proviral
genes or depletion (blue) for antiviral genes at FDR < 0.05
threshold, no hit (gray), or known receptor hit (black). Note
that the previously identied viral receptor is in the top 10
hits for each screen when known (ACE2, ANPEP, DPP4, and
ICAM1). Panel legends correspond to Pathogen/Cell line/
screening endpoint/time after infection.

PNAS 2025 Vol. 122 No. 20 e2414202122 https://doi.org/10.1073/pnas.2414202122 3 of 8
coronaviruses (HCoV-229E, HCoV-NL63, and HCoV-OC43),
along with the phylogenetically distinct HRV along PC1 and PC2,
while the proles for the three IAV screens clustered in the center
of the opposite quadrant along PC1. Genes driving shared varia-
tion (SI Appendix, Fig. S2C ) among common-cold virus screens
included genes with roles in transcriptional elongation, such as
negative elongation factor complex member B (NELFB) and
NELFCD, and putative RNA helicases, such as DEAD-box hel-
icase 54 (DDX54) and DEAH-box helicase 33 (DHX33). On the
other hand, genes driving shared variation among IAV, PIV, RSV,
and SARS-CoV-2 screens were conserved Golgi complex genes,
such as component of oligomeric Golgi complex 2 (COG2),
COG3, COG6, and COG8 (see Integration of Screen Results
below), and SLC35A1, the sialic acid transporter proposed as
supporting IAV ( 23 ). With PC1 and PC2 accounting for ~32%
of the variation (SI Appendix, Fig. S2B ), this analysis suggests
additional virus, cell line, and screen-specic variations exist
within the remaining PCs.
In summary, individual screens were validated by the identi-
cation of known viral entry factors, and by the PCA that served
as a measure of quality control, by ruling out that dierences and
similarities across screens were due to type of phenotypic assay or
test cell line.
Integration of Screen Results. Our goal of this work was to
integrate individual screens into biological networks that would
reect shared mechanisms across viral families. In this analysis,
we considered n = 1,363 genes that scored as a hit in individual
screens for one or more viruses (one virus, n = 937; two viruses,
n = 261; three viruses, n = 106; four or more viruses, n = 59;
DatasetS5). We leveraged a functional interaction network built
on Reactome (24) to identify shared protein interaction hubs
across the various CRISPR screens. ese networks (graphs)
are representations of relational data, in which vertices model
genes and edges model relationships among genes. We rst built
the network on known biological processes and pathways as
supported by Reactome (SIAppendix, Fig.S3A) and a separate
network using solely CRISPR proles (SIAppendix, Fig.S3B).
e CRISPR- based network exhibited biologically informative
properties; thus, we merged both networks in a new network
shown in Fig.2A where we identied genes that scored as proviral
for multiple viruses and highlighted molecular systems that are
densely sampled.
Vacuolar ATPase (V-ATPase, Fig. 2B ) is a multisubunit enzyme
that mediates acidication of eukaryotic intracellular organelles,
which is crucial for the entry process of many viruses, like IAV ( 25 ).
Another system that we highlighted is the COG complex ( Fig. 2C ),
which is required for normal Golgi morphology and localization.
e COG complex is exploited by numerous viruses, such as
SARS-CoV-2, HIV, Chikungunya virus, Hepatitis C virus, Dengue
virus, and Orthopoxvirus ( 26 ). We also identied hypusination
( Fig. 2D ), a two-step enzymatic reaction involving deoxyhypusine
synthase (DHPS) and deoxyhypusine hydroxylase (DOHH) that
occurs during or shortly after the synthesis of eukaryotic translation
initiation factor 5A (EIF5A). Hypusinated EIF5A regulates protein
synthesis through its involvement in translation initiation, elonga-
tion, and termination ( 27 ). Hypusinated eIF5A is reported to play
a role in the replication of the Filoviruses Ebola and Marburg virus
as well as in modulating HIV replication ( 28 ). We also highlight
N-linked glycosylation ( Fig. 2E ), a process that occurs in the lumen
of the endoplasmic reticulum by the membrane-associated enzyme
complex oligosaccharyltransferase (OST). OST is reported to be a
host target for viral inhibition ( 29 – 32 ). Additional pathways that
are prominently associated with multiple pathogens and shown as
densely sampled areas of the network are described in SI Appendix,
Table S2 .
In summary, joint analysis of genome-wide CRISPR screens
for nine viruses resulted in a complex network of viral host
dependencies and highlighted densely sampled clusters of genes
that are revealing shared biology for respiratory viruses.
Validation of OST Complex as an Antiviral Target. As proof of
concept for developing broad- spectrum host- directed therapy
against multiple viral families based on the network of shared
mechanisms, we chose to explore the OST complex across
dierent viruses. Our prioritization was based on the availability
of validated chemical matter, NGI- 1, a cell- permeable small-
molecule inhibitor of the OST complex that was amenable for
rapid structure–activity relationship study and development of a
new potent antiviral drug (33, 34). Two catalytic isoforms [STT3
OST complex catalytic subunit A (STT3A) or B (STT3B)] are
assembled into two dierent OST complexes, and both complexes
can be inhibited by NGI- 1 (32, 35–37).
To validate the role of the OST complex and to probe genes
involved in the broader cellular glycosylation pathways, we gen-
erated a custom CRISPR ko subpool targeting rescreening
(Dataset S6 ). Subgenome level pooled CRISPR screens tend to
have higher sensitivity than whole-genome screens because the
average number of cells containing individual guides in the selec-
tion process increases statistical power. We focused on HRV (in
H1 Hela cells), IAV (in A549 and H1 Hela cells), and PIV (in
A549 cells). e aggregated results ( Fig. 3A ) revealed the impor-
tance of additional OST-A/B complex genes (including OSTC,
OST4, RPN1) in viral cellular infection aside from STT3A/B
(Dataset S6 ).
We then focused on the eect of genetically perturbing OST
complex catalytic subunits STT3A and STT3B on viral replication
using targeted approaches. Viral replication or survival of the
infected cells were measured in the cells transduced with single
guide RNAs (sgRNAs) targeting STT3A or STT3B without clonal
selection, compared to the control cells transduced with sgRNA
targeting intergenic regions (Nontarget/WT controls) or genes
known to be essential for viral infection (positive controls). sgRNA
guide activity was conrmed in target cell lines using
next-generation sequencing (NGS) (Dataset S7 ). Consistent with
the network graph analyses, genetic perturbation of STT3A and/
or STT3B reduced viral infection and/or increased survival of
infected cells in IAV, HRV, PIV, or HCoV-229E infection to
dierent degrees—but not RSV—further supporting the pertur-
bation of OST-complex function as a broad antiviral strategy
( Fig. 3B ).
We next assessed the antiviral activity of the tool compound
NGI-1, which uncompetitively inhibits the catalytic subunits of
OST, STT3A, and STT3B, by trapping the donor substrate
dolichyl-PP-GlcNAc2-Man9-Glc3 ( 38 ). After conrming that
NGI-1 indeed inhibits N-linked glycosylation in a cell-based
reporter assay ( Fig. 3C ), we evaluated the antiviral activity of the
compound against six viruses (PIV, HCoV-229E, SARS-CoV-2,
IAV, HRV, and RSV) in cell-based assays. For PIV, HCoV-229E,
and SARS-CoV-2, NGI-1 inhibited viral replication at concen-
trations similar to the eective concentration in the cell-based
glycosylation assay. ese antiviral eects occurred at concentra-
tions >100-fold lower than concentrations yielding toxicity in
uninfected cells, supportive of on-target antiviral activity ( Fig. 3C ).
We also saw evidence of antiviral activity at high concentrations
of NGI-1 (>1,000 nM) for IAV, though solubility constraints
prevented our ability to obtain a full dose–response curve and an
EC50 . Some antiviral activities were also observed for HRV and

4 of 8 https://doi.org/10.1073/pnas.2414202122 pnas.org
RSV. Altogether, these data suggested that chemical inhibition of
OST was sucient to inhibit the replication of multiple respira-
tory viruses in vitro. On this basis, we advanced OST inhibitors
with improved potency (EC50 1.0 nM) over NGI-1, which are
described in a separate publication ( 39 ).
Discussion
Network analysis of extensive functional genomics screening data
across multiple medically important respiratory viruses provided
a detailed view of the landscape of shared host factors necessary
for viral replication. Multiple genes of the same pathway emerged
as hits from the individual viral screens and contributed, via
knowledge graphs, to the identication of possible targets of
broad-spectrum antiviral agents.
Prominent targets identied in this study included the
V-ATPase complex, the conserved oligomeric Golgi complex, the
hypusination pathway, and the N-glycosylation pathway. ese
warrant campaigns to discover tool compounds and available
chemical molecules to examine the impact of perturbing such
complexes/pathways on viral replication. As a proof of principle,
we tested an OST inhibitor NGI-1 (N-linked Glycosylation
Inhibitor 1, 5 -[( dim eth yla min o)s ulf ony l]- N-( 5-m eth yl- 2-t hia zol yl)-
2- (1- pyr rol idi nyl )-b enz amide)), which inhibits the cat aly tic sub-
units of OST complex. NGI-1 inh ibited PI V, HCoV-2 29E , a nd
SARS-CoV- 2 a t l ow µM concentrati on; th is led to a drug develo
Fig. 2. Functional interaction network of protein hubs
across respiratory viral CRISPR screens. This analysis
used n = 1,363 genes that scored as positive hits for
one or more pathogens. (A) We leverage the Reactome
Functional Interactions gene–gene graph as our source
of truth for the underlying biological network. In addition,
we build a vector of numbers termed “CRISPR proles”
consisting of the minus log10 aggregated P- values and
the graph embeddings. We study the network localization
(hubs, pathways) and the similarities in common
response to pathogen perturbations by means of the
similarities of the CRISPR proles. For clarity, and with
few exceptions, we only label genes associated with four
or more pathogens. (B–E) Knowledge graphs of molecular
systems that are densely sampled in panel to provide
additional information (protein–protein interaction
between viral and host proteins, BioGRID data, genetic
associations, and paralogue information). (B) V- ATPase,
a multi- subunit enzyme that mediates acidication
of eukaryotic intracellular organelles. (C) Oligomeric
Golgi (COG) complex, a master regulator of membrane
tracking at the Golgi. (D) Hypusination, a two- step
enzyme- mediated post- translational modication of
the eukaryotic translation factor eIF5A. (E) N- linked
glycosylation, membrane- associated enzyme complex
OST. Additional densely sampled areas of the network
are discussed in SIAppendix, TableS2.

PNAS 2025 Vol. 122 No. 20 e2414202122 https://doi.org/10.1073/pnas.2414202122 5 of 8
pme nt campaign that identied more potent compounds eective
at nM concentrations, as described in a separate work ( 39 ). e
value of inhibiting OST as an antiviral strategy may extend from
respiratory viruses to Flavivirus, Herpes simplex, Lassa virus, and
HIV-1 ( 30 , 36 , 40 , 41 ). erefore, some of the ndings here may
reach beyond the diverse phylogenetic space targeted here.
Importantly, drugs targeting such pathways may be developed
clinically against common pathogens but may have future benet
in ghting pandemic pathogens phylogenetically unrelated to
those used for genetic assessment.
A limitation of the work presented here is that the screens and
validations were performed using cell lines due to the challenge
of scaling up primary cells for genome-wide screens. Utilizing
primary cells, more complex in vitro/ex vivo systems like
Fig. 3. Validation of STT3A and STT3B as targets of antiviral strategies. To conrm results from the network integration, we evaluated the eect on viral
replication of genetic perturbation of (A) the extended OST complex, and (B) the catalytic subunits STT3A and STT3B. Virus infection levels or survival in STT3A
or STT3B heterogenous knockout pools was compared to cells transduced with intergenic sgRNA controls (nontarget), sgRNAs targeting genes known to block
infection and additional negative controls where indicated. Perturbation of STT3A and/or STT3B increased survival and/or reduced infection against HRV, IAV,
PIV, and HCoV- 229E. Data are mean ± SEM pooled from three to four independent experiments normalized to the average eect on the nontarget controls.
*P < 0.05 compared to Nontarget #1 and Nontarget #2 determined by ANOVA with Tukey’s multiple comparison test. (C) NGI- 1 inhibits the replication of multiple
respiratory viruses. Dose–response curves of NGI- 1 activity and eect on cell viability for glycosylation reporter in H1 Hela cells, PIV infection model in A549 cells,
HCoV- 299E infection model in MRC- 5 cells, SARS- CoV2 infection model in A549 cells with ACE2/TMPRSS2 overexpression, IAV infection model in MDCK cells,
HRV infection model in H1HeLa cells, or RSV infection model in HEp- 2 cells. Mean values from at least two separate experiments were shown ± SD. Eective
concentration (EC50) and cytotoxic concentration (CC50) values were calculated from nonlinear t of dose–response curves (in the graphs). ND = not determined.

6 of 8 https://doi.org/10.1073/pnas.2414202122 pnas.org
organoids, and in vivo animal models could enhance the physio-
logical relevance of developing host-directed therapy against viral
diseases. Another technical limitation is the use of a pooled screen-
ing system, in which genetically perturbed cells were screened
together under the same selection pressure (i.e., viral infection).
e critical role of cell extrinsic factors (e.g., cytokine production
and eect) and the late stage of viral life cycle (e.g., virion assembly
and release) could be more completely sampled using an arrayed
screening system, in which genetically perturbed cells are individ-
ually screened. Of note, we did not explore the antiviral gene sets
of putative cellular restriction and defense mechanisms, which can
lead to the development of immune-based antiviral therapies, as
exemplied by use of agonists of the cGAS–STING cytosolic
DNA-sensing pathway to elicit antiviral activity against members
of multiple RNA viral families ( 42 ).
We explored public and proprietary collections of tool com-
pounds to pharmacologically validate our screen results. However,
the currently available gene-to-drug annotation is incomplete and
limits such validation attempts. Despite this limitation, our work
illustrates the feasibility of developing broad spectrum host-directed
antiviral against respiratory viruses. Host-directed therapy is
expected to have limited pathogen resistance and escape; however,
many host complexes and signaling pathways required by viruses
are also core cellular machinery that may not tolerate prolonged
perturbation. Further studies are warranted to discover host targets
with benecial therapeutic windows and to explore combinatorial
synergies between specic pathogen-directed and broad host-directed
therapies.
Materials and Methods
CRISPR Screens. Genome- wide CRISPR knockout (ko) libraries were screened
against cellular viral infection assays for IAV, PIV, HRV, human coronavirus 229E
(HCoV- 229E), human coronavirus OC43 (HCoV- OC43), human coronavirus NL63
(HCoV- NL63), Middle East respiratory syndrome coronavirus (MERS- CoV), severe
acute respiratory syndrome associated coronavirus (SARS- CoV- 2), and RSV. A
focused subpool library for in- depth analysis of ~500 genes, including those in
the OST complex and N- glycosylation pathway by CORUM and KEGG, was used for
the validation screen. CRISPR screening reagents including genome- wide knock-
out sgRNA libraries (human Brunello and mouse Brie CRISPR ko pooled libraries),
subpool knockout sgRNA libraries, and Cas9 expression vectors were obtained
through the Genetic Perturbation Platform (GPP, https://www.broadinstitute.org/
genetic- perturbation- platform) at the Broad Institute of Massachusetts Institute of
Technology (MIT) and Harvard. Generation of pooled CRISPR ko cell libraries was
performed as previously described (43). Cas9- expressing cell lines were transduced
with lentivirus of CRISPR sgRNA libraries, selected using puromycin and expanded
to sufficient numbers, targeting 500 to 1,000 cells per sgRNA for each experimental
sample. CRISPR library cells were infected with virus and selected for resistance to
infection as measured by resistance to virus- induced cytopathic effect and/or detect-
ing infection- dependent expression of viral proteins or reporter via flow cytometry.
Details of the infection assays including readouts, cell lines, and virus strains used
are summarized in SIAppendix, TableS1. Where possible, CRISPR screens were
performed for the same virus in multiple cell lines. Selected populations of interest
associated with reduced viral infection (surviving cells or high and low viral protein/
reporter expression as isolated by fluorescence- activated cell sorting) were collected
and NGS was performed to determine sgRNA sequence abundance as previously
described (43). NGS and sequencing data deconvolution were supported through
the GPP at the Broad Institute of MIT and Harvard.
Processing and Analysis of CRISPR Screens.
For each screen, each sequenced
sample was assessed for sgRNA coverage and representation by calculating
median reads per sgRNA and number of sgRNA with zero reads. For each screen,
MAGeCK RRA (44) was used to rank genes for enrichment and depletion of target-
ing guides under selection versus control conditions, sorting by positive selection
(mageck:0.5.9.2). Genes were scored for targeting guide enrichment (provi-
ral) and depletion (antiviral) with hits called at FDR- adjusted P- value < 0.05.
Toprofile the similarity of proviral gene rankings across screens using principal
component analysis (PCA), the gene ranks were first quantile normalized across
screens and then, the percentile ranks in each were transformed to normally
distributed scores (R:4.1.0). These normalized scores were also used to profile
the functional annotation of highly ranked proviral genes in each screen via gene
set enrichment analysis (R:4.1.0, fgsea:1.20.0) using the KEGG (msigdbr:7.5.1)
and CORUM annotation (28.11.2022 Corum 4.1 release).
Graph Network Analysis. We leveraged the Reactome (https://reactome.
org/) Functional Interactions gene–gene graph as our source of truth for the
underlying biological network. Each gene was mapped to a graph embedding
based on random walk and truncated singular value decomposition. For the
graph embedding, we follow the method described in ref. 45. Propathogen
P- values from multiple screens of the same pathogen had been aggregated
into a single per pathogen P- value by Fisher’s combination. We smoothed the
biological CRISPR signal on the network by first assigning each gene a binary
label per pathogen based on the aggregated per pathogen P- value and the
Benjamini–Hochberg procedure with a false discovery rate of 5%. For each
pathogen, we trained binary classifiers with the graph embeddings as features
and the threshold P- value label as target. We used threefold cross validation,
and from each held- out fold, we collected the inference output of the classifier.
We used these inference outputs as the smoothed biological signal. The chosen
binary classifier was random forest.
We selected a gene for further analysis and visualization if the gene’s per path-
ogen aggregated (nonsmoothed) P- value was below the Benjamini–Hochberg
threshold of 5% FDR or if the gene ranked within top 100 genes by its smoothed
signal. For a selected gene, we built a vector of numbers termed CRISPR profiles
consisting of the minus log10 aggregated (unsmoothed) P- values and the graph
embeddings. We studied the similarities in network localization (hubs, pathways)
and the similarities in common response to pathogen perturbations by means
of the similarities of the CRISPR profiles. In particular for visualization, we apply
K- means clustering and layout into two dimensions by UMAP.
To further understand the relationship within certain clusters, we inte-
grated the Reactome Functional Interaction data with knowledge graphs
using additional gene–gene and gene–viral relationships, including BioGRID
(46), CORUM (47), paralogues from Ensembl, human genetic associations
with SARS- CoV- 2 (Genomics England PanelApp) (48), and influenza infec-
tions (Open Targets Genetics, https://genetics.opentargets.org/), interactions
between human proteins and proteins of SARS- CoV- 2 (49, 50) and influenza
(5), and PrimeKG (51).
Individual ko Validation for STT3A and STT3B. Cas9- expressing cells were
transduced with lentivirus expressing sgRNAs targeting STT3A or STT3B result-
ing in a heterogenous knockout cell pool (ko cells). Two independent lentiviral
transductions were performed for each cell line and tested in single or dupli-
cate. sgRNA guide activity was performed on target cell lines and confirmed by
deep sequencing as previously described (52). Representative NGS results are
provided in DatasetS7. ko cells were tested in viral infection assays designed
to replicate conditions of the pooled CRISPR screens. Resistance to infection at
various multiplicities of infection (MOI)/virus dilutions and duration was meas-
ured by determining relative survival or relative infection in STT3A and STT3B ko
cells compared to cells transduced with nongene targeting intergenic controls.
Details for each assay are described in SIAppendix, TableS1. Briefly, IAV infec-
tion was measured by using imaging to determine the percentage of cells that
were positive for the influenza viral protein nucleoprotein (NP), PIV infection was
measured by using imaging determine the percentage of cells that were positive
for GFP expression, HCoV- 229E and RSV infection was measured by determining
the level of virus- induced cytopathic effect by measuring cell numbers in infected
wells using imaging of DAPI- stained cells, and RSV infection was measured by
determining the level of virus- induced cytopathic effect using the CellTiterGlo
(CTG) viability assay.
N- Linked Glycosylation Cell- Based Assay. H1 Hela cells were stably trans-
duced with a ERLuc reporter (34) construct containing an additional nanolucif-
erase sequence downstream of a P2A sequence. Cells were incubated in the
presence of NGI- 1 for 18 h before sequential addition of One- Glo or Nano- Glo
reagent to assess chemical inhibition of N- linked glycosylation or cell viability,
respectively.

PNAS 2025 Vol. 122 No. 20 e2414202122 https://doi.org/10.1073/pnas.2414202122 7 of 8
HCoV- 229E Cell- Based Assay. Inhibition of cell death due to HCoV- 229E infec-
tion and compound cytotoxicity was measured using CellTiterGlo luminescence
assays. MRC- 5 cells were infected with HCoV- 229E at an MOI yielding 70 to 80%
cell death. NGI- 1, controls, and virus were added to cells and incubated for 68 h.
Luminescence was measured, and EC50 values were calculated.
SARS- CoV- 2 Cell- Based Assay. A549 cells overexpressing ACE2 and TMPRSS2
were infected with SARS- CoV- 2- Nanoluciferase at MOI = 0.05 in the presence
of NGI- 1 or controls. After 24 h, NanoGlo reagent was added, and luminescence
was measured. Data were analyzed to determine compound efficacy.
PIV- GFP Cell- Based Assay. Inhibition of PIV3- GFP reporter expression and
compound cytotoxicity were measured by fluorescent imaging. A549 cells were
infected with PIV3- GFP at an MOI yielding 40 to 60% infection. After 44 h, cells
were fixed, stained with Hoechst, and imaged. GFP- positive nuclei and total nuclei
were quantified to determine compound efficacy and toxicity.
IAV Cell- Based Assay.
Assay was performed as previously described (53). Briefly,
MDCK- LN cells were infected with influenza A/PR8 at 70 FFU/well in the presence
of NGI- 1. After 24 h, cells were fixed, stained for influenza nucleoprotein, and
imaged. Positive cells were counted and normalized to controls to determine
IC50 values.
HRV Cell- Based Assay. Inhibition of cell death due to HRV infection and com-
pound cytotoxic effects was measured using CellTiterGlo luminescence assays. H1
HeLa cells were infected with HRV at an MOI yielding 70 to 80% cell death. NGI- 1,
controls, and virus were added to cells and incubated for 60 h. Luminescence was
measured, and EC50 values were calculated.
RSV Cell- Based Assay. Inhibition of RSV A2- GFP reporter expression and
compound cytotoxicity were measured in fluorescent imaging. Hep- 2 cells were
infected with RSV A2- GFP at an MOI yielding 70 to 90% infection. After 44 h, cells
were fixed, stained with Hoechst, and imaged. GFP- positive nuclei and total nuclei
were quantified to determine compound efficacy and toxicity.
Data, Materials, and Software Availability. Data generated in this study
have been deposited in BioGRID ORCS (https://wiki.thebiogrid.org/doku.php/
orcs:prepublication_datasets:soriaga2024). MAGeCK (44) and all other sources
of code for computational analyses are from previous publications and are cited
in the corresponding sections. Code availability including network integration
materials is provided through https://github.com/virbio/manuscript- landscape-
of- respiratory- viral- infection. All other data are included in the manuscript and/
or supporting information.
ACKNOWLEDGMENTS.
We thank John Doench, David E. Root, and Olivia Bare at
the Genetic Perturbation Platform (Broad Institute of Massachusetts Institute of
Technology and Harvard) for the CRISPR library construction, NGS and sequenc-
ing data deconvolution and Monica Sentmanat and Xiaoxia Cui at the Genome
Engineering and Stem Cell Center (Washington University in St. Louis School
of Medicine) for NGS of CRISPR ko cells. S.H. was supported by a grant of the
Korea- US Collaborative Research Fund, funded by the Ministry of Science and
Information and Communications Technology and Ministry of Health and Welfare,
Republic of Korea (grant number: RS- 2024- 00468455).
Author aliations: aVir Biotechnology Inc., San Francisco, CA 94158
Author contributions: L.B.S., D.R.B., I.B., A.P., E.W., M. McAllaster, E.A.M., B.K., S.L., T.F.M., M.
Metruccio, T.S., Z.N., W.T., J.G., H.W.V., S.H., and A.T. designed research; L.B.S., D.R.B., I.B.,
A.P., E.W., M. McAllaster, E.A.M., O.B., E.C., B.K., S.L., G.L., T.F.M., M. Metruccio, A.S., L.S., T.S.,
L. Wang, L. Wedel, S.S.Y., L.Y., J.Z., and Z.N. performed research; L.B.S., I.B., A.P., E.W., L.Y.,
and A.T. contributed new reagents/analytic tools; L.B.S., D.R.B., I.B., A.P., E.W., M. McAllaster,
E.A.M., O.B., E.C., B.K., S.L., G.L., T.F.M., M. Metruccio, A.S., L.S., T.S., L. Wang, L. Wedel, S.S.Y.,
L.Y., J.Z., Z.N., W.T., J.G., H.W.V., S.H., and A.T. analyzed data; and L.B.S., D.R.B., I.B., A.P., E.W.,
M. McAllaster, E.A.M., J.G., H.W.V., S.H., and A.T. wrote the paper.
Competing interest statement: All authors were or are employees of Vir Biotechnology
and may hold shares in Vir Biotechnology Inc. D.R.B., I.B., A.P., L.B.S., Z.N., W.T., J.G., S.H.,
A.T. have led a patent related to this work.
1. A. Ianevski et al., DrugVirus.info 2.0: An integrative data portal for broad- spectrum antivirals (BSA)
and BSA- containing drug combinations (BCCs). Nucleic Acids Res. 50, W272–W275 (2022).
2. R. S. Wallis, A. O’Garra, A. Sher, A. Wack, Host- directed immunotherapy of viral and bacterial
infections: Past, present and future. Nat. Rev. Immunol. 23, 121–133 (2023).
3. D. E. Gordon et al., Comparative host- coronavirus protein interaction networks reveal pan- viral
disease mechanisms. Science 370, eabe9403 (2020).
4. R. Wang et al., Genetic screens identify host factors for SARS- CoV- 2 and common cold coronaviruses.
Cell 184, 106–119.e14 (2021).
5. K. M. Haas et al., Proteomic and genetic analyses of influenza A viruses identify pan- viral host
targets. Nat. Commun. 14, 6030 (2023).
6. R. Konig et al., Human host factors required for influenza virus replication. Nature 463, 813–817
(2010).
7. B. Li et al., Genome- wide CRISPR screen identifies host dependency factors for influenza A virus
infection. Nat. Commun. 11, 164 (2020).
8. G. Randall et al., Cellular cofactors affecting hepatitis C virus infection and replication. Proc. Natl.
Acad. Sci. U.S.A. 104, 12884–12889 (2007).
9. L. Hao et al., Drosophila RNAi screen identifies host genes important for influenza virus replication.
Nature 454, 890–893 (2008).
10. M. N. Krishnan et al., RNA interference screen for human genes associated with West Nile virus
infection. Nature 455, 242–245 (2008).
11. A. J. Hirsch, The use of RNAi- based screens to identify host proteins involved in viral replication.
Future Microbiol. 5, 303–311 (2010).
12. R. C. Orchard et al., Discovery of a proteinaceous cellular receptor for a norovirus. Science 353,
933–936 (2016).
13. R. Zhang et al., Mxra8 is a receptor for multiple arthritogenic alphaviruses. Nature 557, 570–574
(2018).
14. S. B. Biering et al., Genome- wide bidirectional CRISPR screens identify mucins as host factors
modulating SARS- CoV- 2 infection. Nat. Genet. 54, 1078–1089 (2022).
15. R. K. Zimmerman et al., Population- based hospitalization burden estimates for respiratory viruses,
2015–2019. Influenza Other Respir. Viruses 16, 1133–1140 (2022).
16. R. Gupta, L. A. Purcell, D. Corti, H. W. Virgin, Pandemic preparedness strategies must go beyond
vaccines. Sci. Transl. Med. 15, eadd3055 (2023).
17. A. Telenti et al., After the pandemic: Perspectives on the future trajectory of COVID- 19. Nature 596,
495–504 (2021).
18. J. M. Greve et al., The major human rhinovirus receptor is ICAM- 1. Cell 56, 839–847 (1989).
19. M. Hoffmann et al., SARS- CoV- 2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a
clinically proven protease inhibitor. Cell 181, 271–280.e78 (2020).
20. H. Hofmann et al., Human coronavirus NL63 employs the severe acute respiratory syndrome
coronavirus receptor for cellular entry. Proc. Natl. Acad. Sci. U.S.A. 102, 7988–7993 (2005).
21. V. S. Raj et al., Dipeptidyl peptidase 4 is a functional receptor for the emerging human coronavirus-
EMC. Nature 495, 251–254 (2013).
22. C. L. Yeager et al., Human aminopeptidase N is a receptor for human coronavirus 229E. Nature 357,
420–422 (1992).
23. J. Han et al., Genome- wide CRISPR/Cas9 screen identifies host factors essential for influenza virus
replication. Cell Rep. 23, 596–607 (2018).
24. G. Wu, X. Feng, L. Stein, A human functional protein interaction network and its application to cancer
data analysis. Genome Biol. 11, R53 (2010).
25. R. Guinea, L. Carrasco, Requirement for vacuolar proton- ATPase activity during entry of
influenza- virus into cells. J. Virol. 69, 2306–2312 (1995).
26. S. Realegeno et al., Conserved Oligomeric Golgi (COG) complex proteins facilitate orthopoxvirus
entry, fusion and spread. Viruses 12, 707 (2020).
27. J. Zhou et al., Spermidine- mediated hypusination of translation factor EIF5A improves mitochondrial
fatty acid oxidation and prevents non- alcoholic steatohepatitis progression. Nat. Commun. 13, 5202
(2022).
28. M. E. Olsen, J. H. Connor, Hypusination of eIF5A as a target for antiviral therapy. DNA Cell Biol. 36,
198–201 (2017).
29. Y. J. Huang et al., Identification of oligosaccharyltransferase as a host target for inhibition of
SARS- CoV- 2 and its variants. Cell Discov. 7, 116 (2021).
30. S. Wang et al., Interferon- inducible guanylate- binding protein 5 inhibits replication of multiple
viruses by binding to the oligosaccharyltransferase complex and inhibiting glycoprotein maturation.
bioRxiv [Preprint] (2024). https://www.biorxiv.org/content/10.1101/2024.05.01.591800v1
(Accessed 12 December 2024).
31. A. Casas- Sanchez et al., Inhibition of protein N- glycosylation blocks SARS- CoV- 2 infection. mBio 13,
e0371821 (2021).
32. H. C. Huang et al., Targeting conserved N- glycosylation blocks SARS- CoV- 2 variant infection invitro.
EBioMedicine 74, 103712 (2021).
33. N. Rinis et al., Editing N- glycan site occupancy with small- molecule oligosaccharyltransferase
inhibitors. Cell Chem. Biol. 25, 1231–1241.e34 (2018).
34. C. Lopez- Sambrooks et al., Oligosaccharyltransferase inhibition induces senescence in RTK- driven
tumor cells. Nat. Chem. Biol. 12, 1023–1030 (2016).
35. C. D. Marceau et al., Genetic dissection of Flaviviridae host factors through genome- scale CRISPR
screens. Nature 535, 159–163 (2016).
36. A. S. Puschnik et al., A small- molecule oligosaccharyltransferase inhibitor with pan- flaviviral activity.
Cell Rep. 21, 3032–3039 (2017).
37. I. V. Alymova et al., Aberrant cellular glycosylation may increase the ability of influenza viruses to escape
host immune responses through modification of the viral glycome. mBio 13, e0298321 (2022).
38. B. L. Lampson et al., Positive selection CRISPR screens reveal a druggable pocket in an
oligosaccharyltransferase required for inflammatory signaling to NF- kappaB. Cell 187, 2209–2223.
e16 (2024).
39. J. E. Pero et al., Discovery of potent STT3A/B inhibitors and assessment of their multipathogen antiviral
potential and safety. J. Med. Chem. 67, 14586–14608 (2024), 10.1021/acs.jmedchem.4c01402.
40. H. Lu, N. A. Cherepanova, R. Gilmore, J. N. Contessa, M. A. Lehrman, Targeting STT3A-
oligosaccharyltransferase with NGI- 1 causes herpes simplex virus 1 dysfunction. FASEB J. 33,
6801–6812 (2019).
41. S. Zhu et al., Comprehensive interactome analysis reveals that STT3B is required for N- glycosylation
of Lassa virus glycoprotein. J. Virol. 93, e01443- 19 (2019).

8 of 8 https://doi.org/10.1073/pnas.2414202122 pnas.org
42. G. Garcia Jr. et al., Innate immune pathway modulator screen identifies STING pathway activation
as a strategy to inhibit multiple families of Arbo and respiratory viruses. Cell Rep. Med. 4, 101024
(2023).
43. J. G. Doench et al., Optimized sgRNA design to maximize activity and minimize off- target effects of
CRISPR- Cas9. Nat. Biotechnol. 34, 184–191 (2016).
44. S. Bodapati, T. P. Daley, X. Lin, J. Zou, L. S. Qi, A benchmark of algorithms for the analysis of pooled
CRISPR screens. Genome Biol. 21, 62 (2020).
45. H. Cho, B. Berger, J. Peng, Compact integration of multi- network topology for functional analysis of
genes. Cell Syst. 3, 540–548.e45 (2016).
46. R. Oughtred et al., The BioGRID database: A comprehensive biomedical resource of curated
protein, genetic, and chemical interactions. Protein Sci. 30, 187–200 (2021).
47. G. Tsitsiridis et al., CORUM: The comprehensive resource of mammalian protein complexes- 2022.
Nucleic Acids Res. 51, D539–D545 (2023).
48. M. E. Kars et al., A comprehensive knowledgebase of known and predicted human genetic variants
associated with COVID- 19 susceptibility and severity. medRxiv [Preprint] (2022). https://doi.
org/10.1101/2022.11.03.22281867 (Accessed 12 December 2024).
49. D. K. Kim et al., A proteome- scale map of the SARS- CoV- 2- human contactome. Nat. Biotechnol. 41,
140–149 (2023).
50. M. Bouhaddou et al., SARS- CoV- 2 variants evolve convergent strategies to remodel the host
response. Cell 186, 4597–4614.e26 (2023).
51. P. Chandak, K. Huang, M. Zitnik, Building a knowledge graph to enable precision medicine. Sci. Data
10, 67 (2023).
52. M. F. Sentmanat, S. T. Peters, C. P. Florian, J. P. Connelly, S. M. Pruett- Miller, A survey of validation
strategies for CRISPR- Cas9 editing. Sci. Rep. 8, 888 (2018).
53. C. Momont et al., A pan- influenza antibody inhibiting neuraminidase via receptor mimicry. Nature
618, 590–597 (2023).