rnai screening reveals proteasome- and cullin3-dependent stages in vaccinia virus infection

Please cite this article in press as: Mercer et al., RNAi Screening Reveals Proteasome- and Cullin3-Dependent Stages in Vaccinia Virus Infection, CellReports (2012), http://dx.doi.org/10.1016/j.celrep.2012.09.003

Cell Reports

Resource

RNAi Screening Reveals Proteasome- andCullin3-Dependent Stages in Vaccinia Virus InfectionJason Mercer,1,5 Berend Snijder,3,5 Raphael Sacher,2 Christine Burkard,1 Christopher Karl Ernst Bleck,4

Henning Stahlberg,4 Lucas Pelkmans,3,* and Ari Helenius1,*1Institute of Biochemistry2Institute of Molecular Systems Biology

ETH Zurich, 8093 Zurich, Switzerland3Institute of Molecular Life Sciences, University of Zurich, 8057 Zurich, Switzerland4Center for Cellular Imaging and NanoAnalytics (C-CINA), Structural Biology and Biophysics, University Basel, Biozentrum, 4058 Basel,

Switzerland5These authors contributed equally to this work*Correspondence: [email protected] (L.P.), [email protected] (A.H.)

http://dx.doi.org/10.1016/j.celrep.2012.09.003

SUMMARY

A two-step, automated, high-throughput RNAisilencing screen was used to identify host cellfactors required during vaccinia virus infection.Validation and analysis of clustered hits revealedpreviously unknown processes during virus entry,including a mechanism for genome uncoating. Viralcore proteins were found to be already ubiquiti-nated during virus assembly. After entering thecytosol of an uninfected cell, the viral DNA wasreleased from the core through the activity of thecell’s proteasomes. Next, a Cullin3-based ubiquitinligase mediated a further round of ubiquitinationand proteasome action. This was needed in orderto initiate viral DNA replication. The results accen-tuate the value of large-scale RNAi screens inproviding directions for detailed cell biological in-vestigation of complex pathways. The list of cellfunctions required during poxvirus infection will,moreover, provide a resource for future virus-hostcell interaction studies and for the discovery ofantivirals.

INTRODUCTION

Poxviruses are enveloped DNA viruses characterized by their

large size, intricate structure, and a complex cytoplasmic repli-

cation cycle (Moss et al., 2007). Vaccinia virus (VACV), the

prototypic poxvirus used in this study, served as a vaccine

during the eradication of smallpox, one of the most devas-

tating diseases of the 20th century (Moss et al., 2007). The

potential deployment of the smallpox virus as a biological

weapon, risks associated with resumed vaccination, and recent

monkeypox outbreaks warrant continued research toward new

antipoxvirus agents (Di Giulio and Eckburg, 2004; Harrison

et al., 2004).

The VACV lifecycle begins with macropinocytic internalization

of the virus into host cells (Huang et al., 2008; Mercer and Helen-

ius, 2008). This is followed by low pH-dependent membrane

fusion to release the viral core into the cytosol (Townsley et al.,

2006). Postpenetration steps include loss of core-associated

lateral bodies, core expansion, and transcription of early viral

genes within the expanded core (Dales, 1963; Kates and Bee-

son, 1970; Moss, 1990; Pedersen et al., 2000). Early messenger

RNAs (mRNAs) generate about 100 different viral proteins,

including unidentified factors required for DNA uncoating (Joklik,

1964b; Magee and Miller, 1968). During the subsequent

uncoating process, cores disappear as visible structures, the

viral DNA is released and replicated, intermediate and late genes

are expressed, and cytosolic virus factories are formed. While

many of the cellular factors and functions required for initial entry

have been described (Huang et al., 2008; Laliberte et al., 2011;

Mercer and Helenius, 2008; Schmidt et al., 2012; Townsley

et al., 2006), the cellular factors that participate in viral genome

uncoating remain largely undefined.

In the last few years, large-scale RNAi screening has become

a powerful tool for the analysis of pathogen-host interactions.

This approach has been previously applied to a variety of

pathogens (reviewed in Cherry, 2009). For VACV, a Drosophila

kinome screen was used to identify and describe a role for

AMP-activated protein kinase (AMPK) during infection (Moser

et al., 2010).

Our goal was to elucidate the host factors in human tissue

culture cells needed in the VACV life cycle. Using a large-

scale RNAi screen, we identified critical cellular processes that

the virus exploits during infection. By focusing our validation

studies on protein clusters involved in ubiquitin-mediated

proteasome degradation, which has been previously implicated

in poxvirus DNA replication (Satheshkumar et al., 2009;

Teale et al., 2009), we uncovered additional roles for protea-

somes and ubiquitination in virus assembly and genome

uncoating. In addition, we identified a Cullin3-based ubiquitin

ligase, as required for VACV genome replication. Ultimately,

this information could facilitate the development of novel

antiviral agents that would target the host rather than the virus

directly.

Cell Reports 2, 1–12, October 25, 2012 ª2012 The Authors 1

mailto:[email protected]

mailto:[email protected]


https://www.researchgate.net/publication/26646794_What_have_RNAi_screens_taught_us_about_viral-host_interactions?el=1_x_8&enrichId=rgreq-66a3d4e1-ae8a-4dab-9392-74e6d618d7e4&enrichSource=Y292ZXJQYWdlOzIzMjUzMDMxOTtBUzoxMDM4NzU5NDgzMTg3MjNAMTQwMTc3NzM2NDczOA==

https://www.researchgate.net/publication/221961347_Poxviridae_The_Viruses_and_Their_Replication?el=1_x_8&enrichId=rgreq-66a3d4e1-ae8a-4dab-9392-74e6d618d7e4&enrichSource=Y292ZXJQYWdlOzIzMjUzMDMxOTtBUzoxMDM4NzU5NDgzMTg3MjNAMTQwMTc3NzM2NDczOA==

https://www.researchgate.net/publication/221961347_Poxviridae_The_Viruses_and_Their_Replication?el=1_x_8&enrichId=rgreq-66a3d4e1-ae8a-4dab-9392-74e6d618d7e4&enrichSource=Y292ZXJQYWdlOzIzMjUzMDMxOTtBUzoxMDM4NzU5NDgzMTg3MjNAMTQwMTc3NzM2NDczOA==

A

B

CYTOPLASM

EXTRACELLULAR SPACE

NUCLEUS

NUP98

NUP62

KPNB1 KPNA2

NUP153

LBR TXN2

TCF7L2

GTPBP4 WTAP

LAMA1

IL26

LAMC2

SELPLG

DDX48

SNRPD3

SNRPA1

PRPF31

NARG1

HTATIP TCF1

PA2G4

NEK11 CAD RPA1

FHL2

ETS2

NFATC3

ESRRA

POLE

POLL

POLR2C

POLA

SNRP70

XAB2

KIF17

VN1R1

PFN1

TNFRSF13C

IL1RAPL1

IL17RE

IL17RA

PAK1

MS4A7

SLC35B2

LCK

RAC1

ATP6V1B2

RACGAP1

EGFR

HCK

FES

GP1BARXFP4

CXCL13

XCL1

FPR1

CCL21

SUCNR1

P2RY1

KCNF1

GPR20HTR4HTR3A

TRAR3

PEAK1

OR2H2

GPR132

OR4D1

RAB34

LYST COPB2

NACA

RPL7A RPL27

RPL36 RPL38

TAOK3

WNK4

WNK1

PLIN3

RIPK3 MAST4

SGK2

TRPC3

PPP2R1A

LATS2 MAP3K7IP1

HES5

HK1

PIK4CB

MAOB

TESSP1

TMPRSS9

PSMA6

FRAP1

PRKAR1B

PRKAR1A

PSMC4

CUL3

RBX1

USP11

HCRTR2 TUBB3OGFR

GPR73 GPR83

NEU1

SLC40A1

CCT4

PTPN23

PSMB4

PSMD7

PSMA3PSMB3

PSMD2

PSMA1

PSMC3

BUB1B

MAFK

PSMC5

TSG101

HRKFASTK

NGFRAP1 DPF1

MYLIP

STR

ING

inte

ract

ion

scor

e

0.35

0.68

1

SPLICEOSOME

NUCLEAR LUMENDNA REPLICATION

POSITIVEREGULATION OFTRANSCRIPTION

RNASPLICING

REGULATION OF ACTIN CYTOSKELETON

INTEGRAL TO MEMBRANE

TYROSINEPROTEIN

KINASE

CHEMOTAXISION CHANNELACTIVITY GPCR

PROTEIN TRANSPORT

RIBOSOME

SERINE/THREONINE-PROTEIN KINASE

PROTEIN KINASE, C-TERMINAL

INDUCTION OF APOPTOSIS

REGULATION OF KINASE ACTIVITY

MITOCHONDRION OUTER MEMBRANE

PEPTIDASEACTIVITY

PROTEASOMEUBIQUITIN-DEPENDENT PROTEIN CATABOLIC PROCESS

PEPTIDE RECEPTOR ACTIVITY

VESICLE

(REGULATION OF)CELL MIGRATION

NUCLEAR PORE

PROTEIN MODIFICATIONBY SMALL PROTEIN

CONJUGATION

CHEMOKINE ACTIVITY

Seco

ndar

y sc

reen

& an

alys

is 276 genes (Ambion)

Analysis as before

3 siRNAs / gene, screened 3x

DAPI~200

GFP Features

Anal

ysis Automated fluorescence imaging

9 images per well, 550•106 cellsAutomated image analysisSupervised machine learning

Data

inte

grat

ion

Interactionnetwork

Functionalannotations

210 xPrim

ary

scre

en 7,000 genes (Qiagen)3 siRNAs / gene, screened 3xReverse transfection of siRNAs (72h)Infection with VACV-EGFP-MV (8h)

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner 12 x

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner

ABC

J

LMNO

ED

IHGF

K

P

1 232221201918171615141312111098765432 24

grei

ner

FUNCTIONAL MODULES188 CONFIRMED ‘HITS’276 PRELIMINARY ‘HITS’

2 Cell Reports 2, 1–12, October 25, 2012 ª2012 The Authors



RESULTS

RNAi Screening Identifies 188 Cellular Factors Requiredfor VACV InfectionWe used an automated, high-content, high-throughput RNAi

screen in human tissue culture cells (HeLa) to identify host cell

factors and processes required during the VACV life cycle. Using

the mature virion (MV) form of a recombinant VACV that

expresses enhanced green fluorescent protein (EGFP) under

a synthetic early/late viral promoter (VACV-EGFP) (Mercer and

Helenius, 2008), we could readily distinguish infected from

noninfected cells based on EGFP expression 8 hr postinfection

(hpi). Automated fluorescence microscopy and image analysis

was employed to score for cell factors required during the repli-

cation cycle, up to and including translation of late viral genes,

without distinguishing between early and late defects.

To avoid some of the pitfalls encountered in previous virus

infection screens (Cherry, 2009; Mohr et al., 2010), we chose

to screen a set of 7,000 genes selected for their potential to be

inhibited by small compounds (the 7,000 druggable genome

library; QIAGEN). Although covering only one-third of human

genes, these represent a well-studied and well-annotated

cross-section of the full genome, and for these factors, inhibitors

and other reagents are more readily available for follow-up

studies. In addition, we improved image analysis, hit visualiza-

tion, and hit validation (Extended Experimental Procedures;

Figure S1; Bray et al., 2012; Snijder et al., 2012).

To increase reliability, we introduced a fully automated

computational pipeline for high-content image analysis

(Carpenter et al., 2006; Snijder et al., 2009). Over 200 quantitative

features were extracted from each of 624 million individual cells

from 1.45 million images (Carpenter et al., 2006). Using these

features as a basis, several iterations of supervised machine

learning were applied to identify virus infection, mitosis,

apoptosis, and technical phenotypes (Ramo et al., 2009; Snijder

et al., 2009) (see Extended Experimental Procedures for details).

Cell number and density (population-context) effects were

normalized to improve both reproducibility of screening results

in different cell lines and consistency of phenotypes obtained

by different small interfering RNAs (siRNAs) targeting the same

gene (Snijder et al., 2012).

The RNAi screen was carried out in two steps, as detailed in

Figure 1A and in experimental procedures. Three nonoverlap-

ping siRNAs against each gene were used in a primary screen

and three additional siRNAs in a secondary follow-up screen.

Both screens were repeated independently three times. The

primary screen using the 7,000 druggable genomes led to the

identification of 276 genes, in which two or three of the three

siRNAs reduced VACV infection relative to controls (i.e., for the

majority, the median over the triplicates of at least two siRNAs

Figure 1. VACV RNAi Screen and Functional Map of Host Factors Req

(A) Outline of screening and bioinformatic procedures.

(B) Host factors in larger functional clusters that interact or share functional ann

functional annotation cluster in which they were present (Figure S2). Functional clu

white boxes with dashed outlines. Yellow lines between genes of different cluste

genes within the same functional cluster include lower confidence (>0.35) STRIN

are indicated as inverted arrowheads. Functional gene clusters are pictured in th

had a median absolute deviation less than �1.5 over the whole

screen). To provide a resource for future studies of poxvirus–

host cell interaction, the full set of images and analysis results

have been made available (see Web Resources).

The secondary screen directed against the genes identified as

hits in the primary screen was performed using siRNAs from

a different vendor (Ambion). It confirmed 68% of the hits from

the primary screen with a correlation coefficient of 0.34 between

the validated RNAi phenotypes of the two screens (p = 2.8 3

10�6). A stringent, control-based analysis of nontargeting control

siRNAs gave a false positive rate of 0.06% in the secondary

screen, giving further confidence in the confirmed host factors.

The final hit list shown in Table S1 had 188 genes, representing

a broad range of cell functions. Among them, several proteins

have been previously implicated in VACV infection: EGFR,

RAC1, PAK1, laminin (LAMA1 and 2), the proteasome, Tsg101,

profilin, and RIPK3 (Chiu et al., 2007; Cho et al., 2009; Eppstein

et al., 1985; Honeychurch et al., 2007; Li et al., 2008; Locker

et al., 2000; Mercer and Helenius, 2008; Mercer et al., 2010;

Satheshkumar et al., 2009; Teale et al., 2009; Villa et al., 2010).

In the only other high-throughput RNAi screen directed against

VACV, Moser et al. screened 440 unique Drosophila genes

and identified seven hits, of which the human orthologs are

PRKAA2, PRKAG2, PRKAB1, PIKfyve, PIK3C2A, STAM, and

PTPN23 (Moser et al., 2010). Of these, we identified PRKAB1

and PTPN23 (Figure 1B; Table S1). While no exact match for

PIK3C2A was found, PI4KB and FRAP1 were hits, further high-

lighting the importance of phosphoinositide-kinase activity for

VACV infection (McNulty et al., 2010; Mercer and Helenius,

2008; Moser et al., 2010). Other hits, such as ribosomal proteins,

were expected for any virus. However, themajority of hits had no

prior connection to VACV infection.

To extract information regarding critical processes in the

infection cycle, the list was submitted to rigorous bioinformatic

analysis. For this, we developed an algorithm that combined

functional- and interaction-based information. This algorithm

has been made available (see Supplemental Network Visualiza-

tion Algorithm) and is applicable for the visualization and analysis

of any gene list. The hits were first assigned to ‘‘functional anno-

tation clusters,’’ representing sets of proteins that share

common annotations within public databases [DAVID; (Huang

da et al., 2009)]. As shown in Table S2 and Figure S2, the genes

were enriched within 13 clusters and several highly connected

functional annotation networks.

The function-based information was combined with data on

protein interactions between individual host factors (STRING;

Szklarczyk et al., 2011). The result of the combined analysis is

shown for 126 of the 188 host factors in Figure 1B. Automatic

annotation was used to visualize the interaction networks (yellow

to red lines) both within and between functional annotation

uired for Infection

otations are depicted. Individual genes were assigned to the highest enriched

sters identified by DAVID with simplified annotations (Table S2) are pictured as

rs indicate high-confidence (>0.9) STRING interactions, while those between

G interactions. Selected genes not found within a functional annotation cluster

eir approximate cellular location.


Figure 2. Proteasome but not E1-Activating Enzyme Function Is Required for VACV Genome Uncoating

(A) RNAi-mediated silencing of proteasome subunits PSMA1 and PSMB3 impair VACV infection. Representative images from the screen are displayed with nuclei

in red and infected cells in green.

(B) Proteasome (MG132) and E1-activating enzyme (UBEI-41) inhibitors block VACV production (MOI 1; 24 hpi; 10 mM AraC, 10 mg/ml CHX, 25 mMMG132, and

50 mM UBEI-41 used throughout).

(C) MG132 and UBEI-41 allow for early but not late VACV gene expression. Cells were infected with VACV (MOI 1), expressing EGFP from either an early (Early) or

a late (Late) promoter. Infectionswere performed in the presence of the indicated inhibitors. Cells were harvested 6 hpi and analyzed for the number of early or late

EGFP-expressing cells, respectively. Cells infected in the presence of CHX or AraC were used as controls.

(D) MG132 and UBEI-41 inhibit VACV infection after internalization. VACV-EGFP MVs (MOI 1) were bound to cells for 1 hr in the presence of inhibitors. Cells were

then washed, treated with pH 7.4 or pH 5.0 media for 5 min, washed, and media containing the inhibitors added back. Cells were analyzed by flow cytometry for

EGFP at 6 hpi.




clusters (gray dashed boxes). Without going into detail, it was

evident that, although replication and assembly of progeny virus

occurs in the cytoplasm of host cells, VACV depends on cyto-

plasmic as well as nuclear factors. Since the nucleus and its

many functions, such as splicing and nuclear pore complex

function, have been largely ignored with regard to the infectious

lifecycle of poxviruses, these findings are likely to open new

avenues in VACV research. The cytoplasmic functions included

membrane trafficking, signaling, proteolysis, and ion transport.

Factors in the tyrosine kinase and actin clusters have already

been shown to be necessary for phosphatidylserine-mediated

macropinocytosis of the virus during entry (Mercer and Helenius,

2008; Mercer et al., 2010).

VACV Core Breakdown Requires Proteasome Activity,but Is Independent of New UbiquitinationIn the follow-up, we focused on two prominent clusters: the pro-

teasome and ubiquitination (Figure 1B). Both have been previ-

ously implicated in VACV genome replication (Satheshkumar

et al., 2009; Teale et al., 2009). SiRNA-mediated depletion of

proteasome subunits caused a nearly complete block in late viral

gene expression (Figure 2A). For confirmation, we determined

the virus yield in the presence of MG132, a proteasome inhibitor,

and UBEI-41, an inhibitor of the cellular E1 ubiquitin-conjugating

enzyme UBA1. Both decreased viral yield by �3 logs, similar to

cycloheximide (CHX), a protein synthesis inhibitor (Figure 2B).

When infection with VACV strains that express EGFP specifically

from an early (E-EGFP-VACV) or a late (L-EGFP-VACV) promoter

was analyzed by flow cytometry, we found that the inhibitors

blocked late but not early viral gene expression (Figure 2C).

This suggested that early steps in the virus lifecycle, such as

virus binding and endocytosis, were not impacted by these

inhibitors. To confirm this, we tested whether the effects of

MG132 and UBEI-41 could be circumvented by forcing fusion

of virions at the plasma membrane with low pH. This treatment

results in deposition of viral cores into the cytoplasm, effectively

bypassing the need for endocytosis (Mercer and Helenius,

2008). Consistent with a postpenetration block, the inhibitors

could not be bypassed upon low pH treatment (Figure 2D). Viral

yield, late gene expression, and inhibition after bypass were also

affected when using the more specific proteasome inhibitor,

Velcade (Figure S3). Thus, as reported by Teale et al. and

Satheshkumar et al. (Satheshkumar et al., 2009; Teale et al.,

2009), we found that proteasomes and UBA1 had no role in virus

endocytosis and penetration.

Next, we addressed genome uncoating. The efficient release

of viral DNA from internalized cores is of interest because

poxvirus particles and cores are extremely stable when exposed

to desiccation, denaturants, proteases, extremes of tempera-

(E) VACV core degradation is proteasome-, but not E1-activating enzyme-depen

MG132, or UBEI-41. At 4 hpi, cells were fixed and intact virions identified by immu

intact virions aremarked by closed arrowheads and free green cytoplasmic cores

number of cores per cell under each condition displayed as mean ± SEM (right)

(F) Electron microscopy (EM) shows that MG132 and UBEI-41 inhibit distinct stag

EM. Intact viral cores are indicated by arrowheads. Blowup of virion (MG132) ind

500 nm. All experiments have been performed in triplicate and displayed as the

See also Figures S3, S4, S5, and S6.

ture, etc. (Essbauer et al., 2007; MacCallum and McDonald,

1957; Malkin et al., 2003).

Using a microscopy-based core stabilization assay that relied

on immunofluorescence staining of released cores, Satheshku-

mar and coworkers reported that inhibition of proteasome

activity had only a minor impact on genome uncoating (Sathesh-

kumar et al., 2009). Although over a 2-fold increase in core stabi-

lization was observed in the presence of MG132, they concluded

that inhibition of proteasome activity may delay or reduce

uncoating, but the effect was not sufficient to account for the

reduction in infectivity.

In light of these previous findings and the observation that

VACV cores are rapidly and efficiently disassembled during

entry into a new host cell (Joklik, 1964b; Magee and Miller,

1968), we found it conceivable that proteasome-directed core

degradation was directly connected to DNA release. To further

investigate this possibility, we followed the fate of individual

viruses and cores by fluorescence microscopy using a virus in

which the core protein A5 was tagged with EGFP (Mercer and

Helenius, 2008) and by immunofluorescence staining of the

viral membrane protein L1 (Schmidt et al., 2011). HeLa cells

were incubated at a multiplicity of infection (MOI) of 10 for 4 hr,

a time sufficient for virions to enter and undergo genome

uncoating. Intact virions that still contained the viral membrane

were yellow (closed arrowheads), and internalized, released

viral cores devoid of membrane were green (open arrowheads).

When the cores in control cells were quantified (Figure S4),

few were observed (25–50/cell; green) (Figure 2E; untreated

[UNTR]). However, in the presence of CHX, which prevents

core breakdown and genome uncoating (Joklik, 1964b), intact

cytoplasmic cores numbered 275–320 per cell (Figure 2E;

CHX). The same was observed when the proteasome was

inhibited with MG132 (250–300/cell) (Figure 2E; MG132). UBEI-

41 had virtually no impact on core breakdown (50–60 cores/

cell) (Figure 2E; UBEI-41). This suggested that protein synthesis

and the proteasome were needed for uncoating, but UBA1

was not.

When the state of cytoplasmic cores after entry in the pres-

ence of MG132 was investigated using electron microscopy,

we could confirm that viral cores remained intact (Figure 2F,

MG132; Figure S5). While core expansion occurred (Figure S6),

it was evident that the genome (i.e., the electron dense material

within the cores) was not released (Figure 2F; blow-up). Electron

microscopy also showed that, while UBEI-41 did not inhibit core

breakdown, it prevented formation of viral DNA replication

factories (Figure 2F; UBEI-41). This was consistent with previous

reports that ubiquitination is needed for DNA replication (Sa-

theshkumar et al., 2009; Teale et al., 2009). Taken together, the

results suggested, somewhat paradoxically, that proteasomes

dent. Cells were infected with WR-EGFP-A5 (MOI 10) in the presence of CHX,

nofluorescence against the viral membrane protein L1R. Representative yellow

by open arrowheads. Experimentswere performed in triplicate, and the average

(Figure S4). Scale bars, 10 mm.

es of the VACV lifecycle. Cells treated as in (F) (MOI 10; 4 hpi) were analyzed by

icates that the electron dense viral genome is still within the core. Scale bars,

mean ± SD (B)–(D) or representative images displayed (A), (E), and (F).


Figure 3. Polyubiquitination of VACV Cores

Facilitates Their Degradation and Subse-

quent Genome Uncoating

(A) MG132 prevents VACV genome release, and

UBEI-41 blocks viral DNA replication. Cells were

infected (MOI 10) in the absence of inhibitors

(UNTR) or in the presence of MG132 or UBEI-41.

Cells were fixed at 4 hpi and immunofluorescence

against the viral DNA binding protein, I3, per-

formed, followed by staining of nuclei with Draq5.

In all images, nuclei are colored red and I3 staining

is in green. Cells treated with CHX, which prevents

genome release, or AraC, which allows for release

of parental genomes, but not their replication,

served as controls for the assay. Scale bars,

10 mm. The number of cells positive for parental

DNA staining (I3 spot formation) under each

condition is displayed as the mean ± SD (right).

(B) VACV core proteins are packaged in a poly-

ubiquitinated state. Immunoblots directed against

total (mono and poly) ubiquitin were performed on

1.0, 2.5, and 5.0 mg of purified VACV MVs (left), on

2.5 mg of whole (W) VACV MVs, or MVs separated

into core (C) andmembrane (M) fractions (right). As

a control for the detection of ubiquitin, 100 mg of

cell lysate (Cells) was used. Experiments were

performed in triplicate and representative blots

shown.

(C) Viral core proteins are positive for K48-, but

not K63-linked ubiquitin chains. Immunoblots

directed against K48- or K63-specific ubiquitin

chains were performed on whole (W), core (C), and

membrane (M) fractions of VACV MVs. One

hundred micrograms of cell lysate was used as

a control for each antibody (Cells).

(D) Late addition of MG132 or UBEI-41 does not

impede late viral gene expression. Cells were

infected (MOI 1) with VACV-EGFP-LATE. Either

10 mM AraC or 50 mM UBEI-41 were added at the

indicated times. Infection was allowed to proceed

for a total of 12 hr before cells were harvested and

analyzed by flow cytometry for EGFP expression.

Values are displayed as the percentage of cells

expressing late genes relative to untreated in-

fected control cells at 12 hpi.

(E) Inhibition of E1-ligase function at 6 hpi impedes

production of infectious virus. Cells were infected

with VACV MVs (MOI 1) and AraC or UBEI-41

added at either the time of infection (0) or after

6 hr (6). Cells were harvested 24 hpi, and the

progeny virus was purified by banding (Figure S8), followed by titration for plaque-forming units/ml. (A)–(C) Experiments were performed in triplicate and

representative images shown. (D) and (E) Experiments were performed in triplicate and displayed as the mean ± SD.

See also Figures S7 and S8.


were needed for breakdown of the core, whereas ubiquitination,

which normally serves to prepare substrates for degradation,

was required for a subsequent step, one preceding DNA

replication.

The microscopy-based uncoating assays used by Satheshku-

mar and coworkers and here by us rely on core destabilization as

an indicator of genome uncoating. However, the release of

parental genomes from incoming viral cores is not analyzed.

Thus, it was possible that the viral DNA was released from viral

cores in the presence of MG132, but the cores themselves

were not degraded in the absence of proteasome activity. To


directly monitor release of incoming viral DNA as a consequence

of core breakdown, we used immunofluorescence staining for

the viral protein I3L that is bound to the viral DNA (Welsch

et al., 2003). That antibodies directed against this protein can

be used to detect uncoated parental VACV DNA in the absence

of viral replication has been demonstrated (Domi and Beaud,

2000; Welsch et al., 2003). While unable to access the antigen

in intact cores (Figure S7), antibodies to I3L stained cytoplasmic

viral DNA after release from incoming cores (Figure 3A). Later in

infection, it also stained the I3L complexed with viral DNA in

the large perinuclear virus factories. When control cells were


infected for 4 hr at a MOI of 10 and the viral DNA visualized by

anti-I3L, replication factories and cytoplasmic I3 staining were

seen (green) (Figure 3A; UNTR). With CHX present to prevent

early gene expression and genome release, only 5% of cells

had I3L-positive spots in the cytoplasm. In the presence of

cytosine arabinoside (AraC), an inhibitor of viral DNA replication,

78% of cells contained cytoplasmic I3L spots, indicating that

incoming viral DNA had been released, but not replicated. Cells

infected in the presence of MG132 were positive for diffuse cyto-

plasmic I3L staining, but did not contain I3L-positive viral DNA

spots. This showed that, even though expression of I3L (an early

protein) had occurred, viral genomes had not been released

(Figure 3A; MG132). In contrast, UBEI-41-treated cells displayed

diffuse cytoplasmic I3L staining and I3L-positive spots, as seen

in AraC-treated cells, confirming that viral genomes were

released but not replicated. Collectively, these results showed

that the proteasome and UBA1 acted in distinct steps in the early

infection cycle: proteasomes in core breakdown and DNA

release and UBA1 (and thus ubiquitination) in replication of

released viral DNA.

VACV Core Proteins Are Packaged in a K48-Ubiquitinated State to Mediate Proteasome-DependentGenome UncoatingThat core degradation was not dependent on UBA1 suggested

that ubiquitin must have been added to core proteins prior to

virus entry. Proteomic analysis of isolated VACV virions has indi-

cated that ubiquitin accounts for 1–3 mole percent of total

protein (Chung et al., 2006). However, the location of this ubiqui-

tin within the viral particle and the potential role of this host

protein in the virus lifecycle have not been investigated.

Immunoblot analysis on highly purified VACV virions

confirmed that they contained a striking amount of ubiquitin (Fig-

ure 3B, left). The molecular weight of the detected bands indi-

cated that these were ubiquitin-conjugated proteins and not

free ubiquitin. To assess the location of the ubiquitin-conjugated

proteins, virions were separated into membrane and core frac-

tions (Mercer and Traktman, 2003). Immunoblot analysis on

these fractions indicated that the vast majority of the ubiquiti-

nated factors were components of the viral core (Figure 3B,

right). Using lysine-48- and lysine-63-linked polyubiquitin

specific antibodies, we showed that only lysine-48-linked ubiqui-

tinated proteins were present (Figure 3C). Since lysine-48-linked

ubiquitin chains mark proteins for proteasomal degradation

(Clague and Urbe, 2010; Komander, 2009), it was evident that

this was the mechanism that triggered proteasome-mediated

genome uncoating.

Since ubiquitination of core proteins must occur during virion

production in infected cells, we asked whether infectious virus

could be produced if ubiquitination was inhibited late in infection.

Addition of AraC at 6 hr had little effect on late viral gene expres-

sion (Figure 3D, white) and only marginally reduced the 24 hr

virus yield (Figure 3E, white). As late genes are expressed from

newly replicated genomes, this indicated that, by 6 hpi, viral

DNA replication and late gene expression was sufficient for

significant virus production. When UBEI-41 was added at

6 hpi, again there was no impact on late viral gene expression

(Figure 3D, gray). However, unlike AraC, which had little effect

on virus yield, addition of UBEI-41 after 6 hr prevented the

production of infectious virus (Figure 3E, gray). We were, in

fact, unable to isolate virus particles from infected cells treated

with UBEI-41 from 6 hpi (Figure S8). This indicated that there

are at least two steps in the VACV infectious cycle that require

ubiquitination: not only viral DNA replication after genome

uncoating, but also assembly of progeny virus.

A Cullin-3 Ubiquitin Ligase Complex Is Requiredfor VACV Genome ReplicationIt has been reported that poxvirus DNA replication is inhibited in

MG132- and Velcade-treated cells (Satheshkumar et al., 2009;

Teale et al., 2009). Could this observation be explained by the

upstream inhibition of uncoating, or was there a second step in

the early infection program that required ubiquitination and the

proteasome? After confirming that both MG132 and UBEI-41

prevented VACV replication site formation (Figure 4A), we used

an AraC wash-out assay to test for direct effects on replication.

AraC has no impact on the uncoating of viral genomes, but

prevents their replication. Cells were infected in the presence

of AraC for 6 hr to allow uncoating, but to stop replication of

incoming genomes. AraC was then washed out or replaced

with either UBEI-41 or MG132. Infection was allowed to proceed

for an additional 6 hr, after which cells were analyzed for viral

replication sites (Figure 4B). In control cells without a second

inhibitor (No Inh.), replication site formation was normal.

However, in the presence of UBEI-41 or MG132, the formation

of viral replication sites did not occur. These results confirmed

that the initiation of DNA replication requires both a ubiquitination

step and proteasome-mediated degradation.

Reasoning that a cellular ubiquitin ligase was likely to be

involved, we turned to the hits in the RNAi screen. The ubiquitin

functional annotation cluster (Figure 1B) contained two compo-

nents of the BCR E3 ubiquitin-ligase complex: Cullin3 (Cul3)

and RING-box protein 1 (Rbx1). Images from the screen indi-

cated that depletion of these factors resulted in the absence of

late gene expression (Figure 4C). A requirement for Cul3 and

Rbx1 was validated with three independent siRNAs each. Deple-

tion of either factor decreased the 24 hr viral yield by two logs

(Figure 4D; left). Silencing of Cullin-1 (Cul1), a component of

the SCF E3 ubiquitin-ligase complex that was not a hit in the

screen, had no impact (Figure 4C and 4D; left). The knockdown

efficiency of these siRNAs was confirmed by Western blot (Fig-

ure 4D; right).

Additional experiments were performed using the siRNA with

the strongest knockdown efficiency (Cul3-1, Rbx1-3, and Cul1-

2). Depletion of Cul3 or Rbx1 was found to significantly decrease

late but not early gene expression (Figure 4E), mirroring the effect

of AraC and confirming a defect in DNA replication. In line with

this, Cul3 and Rbx1 knockdown inhibited the formation of cyto-

plasmic DNA replication factories (Figure 4F). We concluded that

UBA1, a Cul3 ubiquitin ligase, and the proteasomewere required

to initiate the replication of uncoated VACV DNA.

DISCUSSION

Our results demonstrated that large-scale RNAi screening

can, despite inherent problems and pitfalls (off-target effects,


Figure 4. A Cullin3 E3-Ligase Complex Is Required for VACV DNA Replication

(A) MG132 and UBEI-41 block VACV DNA replication site formation. Cells were infected with VACV MVs (MOI 1) in the presence of AraC, MG132, or UBEI-41.

Cells were fixed and stained for DNA at 4 hpi. Representative images (left) scale bars, 2 mm; mean ± SD of triplicate experiments (right).

(B) MG132 and UBEI-41 block VACV DNA replication after uncoating. Cells were infected with VACV MVs (MOI 1) in the presence of AraC. At 6 hpi, cells were

washed and released from AraC (No Inh.) or shifted into MG132 or UBEI-41. Cells were fixed at 12 hpi and assessed for the presence of cytoplasmic DNA

replication sites.

(C) RNAi against Cul3 or Rbx1, but not Cul1, impairs VACV infection. Representative images from the screen are displayed with nuclei in red and infected cells in

green.

(D) Silencing of Cul3 or Rbx1, but not Cul1, reduces virus production. Cells were reverse transfected with 20 nM (final concentration) of three independent siRNAs

directed against either Cul3, Rbx1, or Cul1. After 72 hr, cells were infected with VACV MVs (MOI 1). Twenty-four hours later, cells were harvested and lysates

titered for plaque-forming units/ml (left). Immunoblot analysis of Cul3, Rbx1, and Cul1 after siRNA mediated silencing for 72 hr. Immunoblot analysis directed

against actin was used as a loading control (right; representative blots shown).




insufficient knockdown, cell population context effects, incom-

plete genome annotation, etc. [Cherry, 2009; Mohr et al.,

2010]) serve as a valuable tool in the study of pathogen/host

interactions. The positive outcome relied in this case on a robust

image-based assay, on supervised classification of cellular

phenotypes, a separation of indirect and direct effects of

siRNAs, and on improved bioinformatics analysis to identify

gene clusters (Snijder et al., 2012). That the screening data, anal-

ysis, and bioinformatic tools used for the purpose of this study

are publically available will be of general usefulness in the anal-

ysis of future screening data.

Among the 188 hits, we identified numerous clusters in which

multiple hits occurred in the same pathway or complex.

Regarding the validity of the final hit list, we are encouraged by

how many of the hits corresponded to factors identified earlier

as essential genes in VACV infection. Validation studies with

two of the clusters as a starting point allowed us to extend

previous findings regarding a role for ubiquitination in DNA repli-

cation (Satheshkumar et al., 2009; Teale et al., 2009) and

discover a new role for ubiquitination and the proteasome in virus

assembly and uncoating (Figure 4G). We could thus confirm the

value of the hit list as a resource for functional studies of poxvirus

host cell interactions.

Our results showed that ubiquitination comes into play as an

essential process already during the assembly of VACV particles

in virus factories located in the cytoplasm of producer cells. Core

proteins undergo extensive K48-linked polyubiquitination. This is

consistent with the accumulation of ubiquitin observed in

poxvirus replication sites and the detection of ubiquitin within

purified VACV particles (Chung et al., 2006; Nerenberg et al.,

2005). Polyubiquitination is probably an essential step in virus

assembly, because inhibition of the E1 ubiquitin-activating

enzyme prevented the generation of viral particles. VACV

encodes one ubiquitin ligase, p28, which localizes to VACV repli-

cation sites (Nerenberg et al., 2005). However, as this protein is

not required for virus production in tissue culture, it is unlikely to

be the ligase responsible for the ubiquitination of viral core

proteins. Future work is needed to identify the core components

that are ubiquitinated as well as the ubiquitin ligase(s) respon-

sible. During egress from the viral factory, the viral membrane

may protect the ubiquitinated core proteins from the degradation

machinery.

After macropinocytic internalization of VACV into a new host

cell and low pH-dependent membrane fusion, polyubiquitinated

cores are released into the cytosol. They immediately undergo

activation marked by major expansion in size, a step that does

not require early gene expression (Dales, 1963; Ichihashi et al.,

(E) Silencing of Cul3 and Rbx1, but not Cul1, reduces late but not early viral gene e

by infection with VACV (MOI 1), expressing EGFP specifically from an early or a l

expressing early and late EGFP, respectively. Cells infected in the presence of C

(F) Cul3 and Rbx1 are required for VACV DNA replication site formation. Cells in w

with VACVMVs (MOI 1). At 4 hpi, cells were fixed, stained with Draq5, and analyze

triplicate and displayed as mean ± SD (A, B, and D–F).

(G) Model of VACV core ubiquitination, core degradation/genome uncoating, an

ubiquitinated in a K48-linked fashion. Upon infection of naive cells, fusion of viral a

A first round of cellular proteasome action directs the degradation of the ubiquitin

second round of proteasome action serves to initiate replication of the releas

accessible (green); proteasomes (gray bullets); cullin-based ubiquitin ligase com

1984; Pedersen et al., 2000). The expanded cores serve as the

site for early gene transcription (Joklik, 1964a; Magee andMiller,

1968). That they are not yet substrates for proteasome-mediated

uncoating suggests that the ubiquitin chains are not accessible.

For degradation of the core and uncoating of the DNA, early

genes have to be expressed. One or more of the early proteins

are likely responsible for making the ubiquitinated core proteins

accessible as substrates for the proteasome. The core particle is

destroyed and some of the core components, such as the A5

protein, degraded.

As a result of the proteolysis, the viral DNA genome is released

and the core itself is no longer recognizable as a structure in

the cytosol. That UBA1 activity is not required indicates that

the ubiquitin present in the core is sufficient to mediate genome

uncoating. Although proteasome function has been implicated in

the regulation of viral trafficking, replication, egress, and immune

evasion (reviewed in Banks et al., 2003), a direct role for protea-

somes in the uncoating of viruses or viral capsids has not, to our

knowledge, been observed before.

It was previously reported that viral DNA replication requires

ubiquitination and proteasome activity (Satheshkumar et al.,

2009; Teale et al., 2009). Our results extended these findings

to show that replication of the viral DNA depends on Cul3 and

Rbx1, two components of an important, multifunctional E3-

ligase family (Petroski and Deshaies, 2005). As recently reviewed

(Barry et al., 2010), VACV encodes at least four Cul3 substrate

adapters: A55R, C2L, C5L, and F3L (Beard et al., 2006; Froggatt

et al., 2007; Pires de Miranda et al., 2003). Although each of

these is required for VACV virulence, they are not essential for

replication in tissue culture. This implies that they are not the

adaptor proteins used by Cul3 to facilitate VACV DNA replica-

tion. While the substrates are yet to be identified, one possibility

is that they represent remaining DNA-associated proteins to be

removed before replication.

The proteasome thus plays a central role in at least two steps

in the replication cycle of VACV. Collectively, the RNAi screen

and follow-up studies, like those described here, are likely to

provide a starting point for detailed cell and molecular biology

analysis of poxvirus-host cell interactions, and theymay facilitate

the development of novel antipoxvirus agents that target the host

cell rather than viral factors.

EXPERIMENTAL PROCEDURES

RNAi Screen

HeLa MZ cells from Marino Zerial (MPI-CBG, Dresden) were maintained at

37�C and 5% CO2 in Dulbecco’s modified Eagle’s medium (DMEM; GIBCO

xpression. Cells were subjected to siRNA-mediated silencing as in (D) followed

ate promoter. Cells were harvested 6 hpi and analyzed for the number of cells

HX or AraC were used as controls.

hich Cul3, Rbx1, or Cul1 had been silenced (20 nM siRNA; 72 hr) were infected

d for the presence of viral DNA replication sites. Experiments were performed in

d genome replication. During assembly of VACV MVs, viral core proteins are

nd cellular membranes releases the ubiquitinated viral core into the cytoplasm.

ated core and concomitant genome release. A Cul3-based ubiquitin ligase and

ed viral genomes. Ubiquitination events: new (red); nonaccessible (yellow);

plex (gray crescents).



BRL) supplemented with 10% fetal calf serum (FCS) and glutamax. All liquid-

handling steps, including preparing cell plates, virus infection, and cell fixation,

were performed on a Freedom Evo200 liquid-handling robot (Tecan) set up

specifically for this purpose. Seeding, fixing, and DAPI staining of the cells

was performed using a Matrix Wellmate Microplate Dispenser (Thermo

Scientific).

RNA interference against the 6,979 gene targets was achieved with

three independent siRNAs targeting each (siRNAs #1 and #2 from QIAGEN

druggable genome version 2 and siRNA #3 from QIAGEN druggable

genome version 3) or three independent siRNAs from Ambion for sec-

ondary screening (custom ordered). Each siRNA was tested in triplicate.

Experiments were conducted in a 384 well plate format, amounting to 210

plates for the primary screen (see Extended Experimental Procedures for plate

layout).

All images were acquired on automated wide field cellWoRx microscopes

(Applied Precision) with a 10X objective, and 23 2 binning per pixel. Multiwell

plates were loaded onto the cellWoRx microscopes using Freedom Evo

robotics from Tecan. Three by three directly adjacent images were taken per

well, covering over �90% of each well surface. An image-based autofocus

was performed on the DAPI signal for each imaged site. The images were

recorded with 12-bit charge-coupled device (CCD) cameras and stored as

individual 16-bit uncompressed tagged image file format (TIFF) files per

imaged site and per channel.

Computational Image Analysis, Supervised Classification of Cellular

Phenotypes, Hit Scoring, DAVID Annotation Clustering Network, and

Combined DAVID/STRING Functional Gene View

Please refer to the Extended Experimental Procedures for details on bio-

informatics methodology.

Cells, Viruses, and Reagents

HeLa ATCC and HeLa MZ cells were maintained at 37�C and 5% CO2 in

DMEM (GIBCO BRL) supplemented with 10% FCS and glutamax. Wild-type

vaccinia virus (strain WR), (EGFP-VACV) containing the EGFP gene driven by

a synthetic VACV early-late promoter inserted into the thymidine kinase (tk)

locus, and vaccinia virus containing EGFP-tagged versions of the core protein

A5 (WR-EGFP-CORE) were generated as previously described (Mercer and

Helenius, 2008). AraC, CHX, and MG132 were purchased from Sigma.

UBEI-41 (PYR-41) was purchased from Calbiochem.

Flow Cytometry Analysis and Perturbant Analysis

To analyze temporal viral gene expression, cells were infected with recombi-

nant VACV viruses that express EGFP from the J2R early viral promoter

(E-EGFP-VACV) or the F18R late viral promoter (L-EGFP-VACV). Cells were

infected at an MOI of 1. For all experiments, cells were harvested at 6 hpi

and prepared as previously described (Mercer and Helenius, 2008). Cells ex-

pressing early or late EGFP were gated based on uninfected controls. CHX

and AraC served as specific controls for inhibition of early and late gene

expression, respectively. Flow cytometry was performed on a BD Bioscience

Calibur System. For each sample, 10,000 cells were analyzed. For all experi-

ments, cells were infected with E- or L-EGFP-VACV in the presence of drug

at an MOI of 1, unless otherwise indicated.

Viral DNA Uncoating

HeLa cells were infected in the presence of CHX (25 mg/ml), AraC (10 mM),

MG132 (25 mM), or UBEI-41 (50 mM). Cells were then infected at an MOI of

10. Four hpi cells were fixed and stainedwith antisera directed against vaccinia

I3L protein (1:500) (a generous gift of Jacomine Krijnse-Locker; University of

Heidelberg, Germany) followed by Alexa594 secondary antibody (1:1000)

and draq5 (1:5000). Images were acquired on a Zeiss LSM510 confocal micro-

scope at 100X oil immersion objective.

Virion Ubiquitin Immunoblot Analysis

One to five micrograms of band-purified wild-type (WT) MVs were left

untreated or were fractionated into membrane and core components as previ-

ously described (Mercer and Traktman, 2003). Samples were separated on

SDS-PAGE, transferred to nitrocellulose, and immunoblot analysis was per-


formed for total mono- and polyubiquitination (clone FK2, Enzo Life sciences),

k48-linked ubiquitin chains (clone Apu2; Millipore), or K63-linked ubiquitin

chains (clone Apu3; Millipore).

siRNA Transfection

For siRNA transfection, 25,000 HeLa ATCC cells were seeded in a 24 well dish

1 day prior to transfection. A 20 nM final concentration of siRNA was used to

transfect cells 72 hr prior to infection. siRNAs were purchased from QIAGEN:

Cullin1: (1) Hs_CUL1_6; (2) Hs_CUL1_5; (3) Hs_CUL1_3, Cullin3: (1)

Hs_CUL3_5; (2) Hs_CUL3_8; (3) Hs_CUL3_10, and Rbx1: (1) Hs_RBX1_6; (2)

Hs_RBX1_5; (3) Hs_RBX1_10.

siRNA Depletion Confirmation

Immunoblot analyses to confirm the depletion of target proteins by siRNAwere

performed on siRNA-treated cells. Antibodies were directed against Cullin1

(1:250; Zymed Laboratories), Skp1 (1:1000; Cell Signaling Technology), or

Rbx1 (1:250; Sigma-Aldrich). Cul3 antibody (1:1000) was a generous gift of

Prof. M. Peter (Institute of Biochemistry, ETH Zurich).

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures, eight

figures, two tables, and a supplemental network visualization algorithm and

can be found with this article online at http://dx.doi.org/10.1016/j.celrep.

2012.09.003.

LICENSING INFORMATION

This is an open-access article distributed under the terms of the Creative

Commons Attribution-Noncommercial-No Derivative Works 3.0 Unported

License (CC-BY-NC-ND; http://creativecommons.org/licenses/by-nc-nd/3.0/

legalcode).

ACKNOWLEDGMENTS

We thank T. Steiger and O. Byrde for data storage and cluster computing,

Pauli Ramo for statistical and image analysis, Karin Mench for technical assis-

tance, and Jacomine Krijnse-Locker for anti-I3L. J.M. is supported by an SNF

Ambizione (PZ00P3_131988), L.P. by the University of Zurich, SNSF, Sys-

temsX.ch (http://SystemsX.ch) RTD project InfectX, and the European Union,

C.K.E.B by the SNF Sinergia CRSII3_125110 and the SystemsX.ch RTD

project C-CINA, and A.H. by ETH Zurich, InfectX, and an ERC advanced inves-

tigator grant.

Received: July 26, 2012

Revised: August 30, 2012

Accepted: September 7, 2012

Published online: October 18, 2012

WEB RESOURCES

The URLs for data presented herein are as follows:

CellClassifier, http://www.cellclassifier.ethz.ch

CellProfiler, http://www.cellprofiler.org/

Full set of images and analysis results, http://www.infectome.org

STRING databse, http://www.string-db.org

REFERENCES

Banks, L., Pim, D., and Thomas, M. (2003). Viruses and the 26S proteasome:

hacking into destruction. Trends Biochem. Sci. 28, 452–459.

Barry, M., van Buuren, N., Burles, K., Mottet, K., Wang, Q., and Teale, A.

(2010). Poxvirus exploitation of the ubiquitin-proteasome system. Viruses 2,

2356–2380.



http://creativecommons.org/licenses/by-nc-nd/3.0/legalcode

http://creativecommons.org/licenses/by-nc-nd/3.0/legalcode

http://SystemsX.ch

http://www.cellclassifier.ethz.ch

http://www.cellprofiler.org/

http://www.infectome.org

http://www.string-db.org


Beard, P.M., Froggatt, G.C., and Smith, G.L. (2006). Vaccinia virus kelch

protein A55 is a 64 kDa intracellular factor that affects virus-induced cytopathic

effect and the outcome of infection in amurine intradermal model. J. Gen. Virol.

87, 1521–1529.

Bray, M.A., Fraser, A.N., Hasaka, T.P., and Carpenter, A.E. (2012). Workflow

and metrics for image quality control in large-scale high-content screens. J.

Biomol. Screen. 17, 266–274.

Carpenter, A.E., Jones, T.R., Lamprecht, M.R., Clarke, C., Kang, I.H., Friman,

O., Guertin, D.A., Chang, J.H., Lindquist, R.A., Moffat, J., et al. (2006).

CellProfiler: image analysis software for identifying and quantifying cell pheno-

types. Genome Biol. 7, R100.

Cherry, S. (2009). What have RNAi screens taught us about viral-host interac-

tions? Curr. Opin. Microbiol. 12, 446–452.

Chiu, W.L., Lin, C.L., Yang, M.H., Tzou, D.L., and Chang, W. (2007). Vaccinia

virus 4c (A26L) protein on intracellular mature virus binds to the extracellular

cellular matrix laminin. J. Virol. 81, 2149–2157.

Cho, Y.S., Challa, S., Moquin, D., Genga, R., Ray, T.D., Guildford, M., and

Chan, F.K. (2009). Phosphorylation-driven assembly of the RIP1-RIP3

complex regulates programmed necrosis and virus-induced inflammation.

Cell 137, 1112–1123.

Chung, C.S., Chen, C.H., Ho, M.Y., Huang, C.Y., Liao, C.L., and Chang, W.

(2006). Vaccinia virus proteome: identification of proteins in vaccinia virus

intracellular mature virion particles. J. Virol. 80, 2127–2140.

Clague, M.J., and Urbe, S. (2010). Ubiquitin: same molecule, different degra-

dation pathways. Cell 143, 682–685.

Dales, S. (1963). The uptake and development of vaccinia virus in strain L cells

followed with labeled viral deoxyribonucleic acid. J. Cell Biol. 18, 51–72.

Di Giulio, D.B., and Eckburg, P.B. (2004). Human monkeypox: an emerging

zoonosis. Lancet Infect. Dis. 4, 15–25.

Domi, A., and Beaud, G. (2000). The punctate sites of accumulation of vaccinia

virus early proteins are precursors of sites of viral DNA synthesis. J. Gen. Virol.

81, 1231–1235.

Eppstein, D.A., Marsh, Y.V., Schreiber, A.B., Newman, S.R., Todaro, G.J., and

Nestor, J.J., Jr. (1985). Epidermal growth factor receptor occupancy inhibits

vaccinia virus infection. Nature 318, 663–665.

Essbauer, S., Meyer, H., Porsch-Ozcurumez, M., and Pfeffer, M. (2007). Long-

lasting stability of vaccinia virus (orthopoxvirus) in food and environmental

samples. Zoonoses Public Health 54, 118–124.

Froggatt, G.C., Smith, G.L., and Beard, P.M. (2007). Vaccinia virus gene F3L

encodes an intracellular protein that affects the innate immune response. J.

Gen. Virol. 88, 1917–1921.

Harrison, S.C., Alberts, B., Ehrenfeld, E., Enquist, L., Fineberg, H., McKnight,

S.L., Moss, B., O’Donnell, M., Ploegh, H., Schmid, S.L., et al. (2004). Discovery

of antivirals against smallpox. Proc. Natl. Acad. Sci. USA 101, 11178–11192.

Honeychurch, K.M., Yang, G., Jordan, R., and Hruby, D.E. (2007). The vaccinia

virus F13L YPPL motif is required for efficient release of extracellular envel-

oped virus. J. Virol. 81, 7310–7315.

Huang, C.Y., Lu, T.Y., Bair, C.H., Chang, Y.S., Jwo, J.K., andChang,W. (2008).

A novel cellular protein, VPEF, facilitates vaccinia virus penetration into HeLa

cells through fluid phase endocytosis. J. Virol. 82, 7988–7999.

Huang da, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and

integrative analysis of large gene lists using DAVID bioinformatics resources.

Nat. Protoc. 4, 44–57.

Ichihashi, Y., Oie, M., and Tsuruhara, T. (1984). Location of DNA-binding

proteins and disulfide-linked proteins in vaccinia virus structural elements. J.

Virol. 50, 929–938.

Joklik, W.K. (1964a). The Intracellular Uncoating of Poxvirus DNA. I. The Fate

of Radioactively-Labeled Rabbitpox Virus. J. Mol. Biol. 8, 263–276.

Joklik, W.K. (1964b). The Intracellular Uncoating of Poxvirus DNA. Ii. The

Molecular Basis of the Uncoating Process. J. Mol. Biol. 8, 277–288.

Kates, J., and Beeson, J. (1970). Ribonucleic acid synthesis in vaccinia virus. I.

Themechanism of synthesis and release of RNA in vaccinia cores. J. Mol. Biol.

50, 1–18.

Komander, D. (2009). The emerging complexity of protein ubiquitination.

Biochem. Soc. Trans. 37, 937–953.

Laliberte, J.P., Weisberg, A.S., and Moss, B. (2011). The membrane fusion

step of vaccinia virus entry is cooperatively mediated by multiple viral proteins

and host cell components. PLoS Pathog. 7, e1002446.

Li, Y., Grenklo, S., Higgins, T., and Karlsson, R. (2008). The profilin:actin

complex localizes to sites of dynamic actin polymerization at the leading

edge of migrating cells and pathogen-induced actin tails. Eur. J. Cell Biol.

87, 893–904.

Locker, J.K., Kuehn, A., Schleich, S., Rutter, G., Hohenberg, H., Wepf, R., and

Griffiths, G. (2000). Entry of the two infectious forms of vaccinia virus at the

plasma membane is signaling-dependent for the IMV but not the EEV. Mol.

Biol. Cell 11, 2497–2511.

MacCallum, F.O., and McDonald, J.R. (1957). Effect of temperatures of up to

45 degrees C on survival of variola virus in human material in relation to

laboratory diagnosis. Bull. World Health Organ. 16, 441–443.

Magee,W.E., andMiller, O.V. (1968). Initiation of vaccinia virus infection in acti-

nomycin D-pretreated cells. J. Virol. 2, 678–685.

Malkin, A.J., McPherson, A., and Gershon, P.D. (2003). Structure of intracel-

lular mature vaccinia virus visualized by in situ atomic force microscopy.

J. Virol. 77, 6332–6340.

McNulty, S., Bornmann, W., Schriewer, J., Werner, C., Smith, S.K., Olson,

V.A., Damon, I.K., Buller, R.M., Heuser, J., and Kalman, D. (2010). Multiple

phosphatidylinositol 3-kinases regulate vaccinia virus morphogenesis. PLoS

ONE 5, e10884.

Mercer, J., and Traktman, P. (2003). Investigation of structural and functional

motifs within the vaccinia virus A14 phosphoprotein, an essential component

of the virion membrane. J. Virol. 77, 8857–8871.

Mercer, J., and Helenius, A. (2008). Vaccinia virus uses macropinocytosis and

apoptotic mimicry to enter host cells. Science 320, 531–535.

Mercer, J., Knebel, S., Schmidt, F.I., Crouse, J., Burkard, C., and Helenius, A.

(2010). Vaccinia virus strains use distinct forms of macropinocytosis for host-

cell entry. Proc. Natl. Acad. Sci. USA 107, 9346–9351.

Mohr, S., Bakal, C., and Perrimon, N. (2010). Genomic screening with RNAi:

results and challenges. Annu. Rev. Biochem. 79, 37–64.

Moser, T.S., Jones, R.G., Thompson, C.B., Coyne, C.B., and Cherry, S. (2010).

A kinome RNAi screen identified AMPK as promoting poxvirus entry through

the control of actin dynamics. PLoS Pathog. 6, e1000954.

Moss, B. (1990). Regulation of vaccinia virus transcription. Annu. Rev. Bio-

chem. 59, 661–688.

Moss, B., Knipe, D.M., and Howley, P.M. (2007). Poxviridae: The viruses and

their replication. In Fields Virology (Philadelphia: Lippincott-Raven).

Nerenberg, B.T., Taylor, J., Bartee, E., Gouveia, K., Barry, M., and Fruh, K.

(2005). The poxviral RING protein p28 is a ubiquitin ligase that targets ubiquitin

to viral replication factories. J. Virol. 79, 597–601.

Pedersen, K., Snijder, E.J., Schleich, S., Roos, N., Griffiths, G., and Locker,

J.K. (2000). Characterization of vaccinia virus intracellular cores: implications

for viral uncoating and core structure. J. Virol. 74, 3525–3536.

Petroski, M.D., and Deshaies, R.J. (2005). Function and regulation of cullin-

RING ubiquitin ligases. Nat. Rev. Mol. Cell Biol. 6, 9–20.

Pires de Miranda, M., Reading, P.C., Tscharke, D.C., Murphy, B.J., and Smith,

G.L. (2003). The vaccinia virus kelch-like protein C2L affects calcium-indepen-

dent adhesion to the extracellular matrix and inflammation in a murine intra-

dermal model. J. Gen. Virol. 84, 2459–2471.

Ramo, P., Sacher, R., Snijder, B., Begemann, B., and Pelkmans, L. (2009).

CellClassifier: supervised learning of cellular phenotypes. Bioinformatics 25,

3028–3030.



Satheshkumar, P.S., Anton, L.C., Sanz, P., and Moss, B. (2009). Inhibition of

the ubiquitin-proteasome system prevents vaccinia virus DNA replication

and expression of intermediate and late genes. J. Virol. 83, 2469–2479.

Schmidt, F.I., Bleck, C.K., Helenius, A., and Mercer, J. (2011). Vaccinia extra-

cellular virions enter cells by macropinocytosis and acid-activated membrane

rupture. EMBO J. 30, 3647–3661.

Schmidt, F.I., Bleck, C.K., and Mercer, J. (2012). Poxvirus host cell entry. Curr.

Opin. Virol. 2, 20–27.

Snijder, B., Sacher, R., Ramo, P., Damm, E.M., Liberali, P., and Pelkmans, L.

(2009). Population context determines cell-to-cell variability in endocytosis

and virus infection. Nature 461, 520–523.

Snijder, B., Sacher, R., Ramo, P., Liberali, P., Mench, K.,Wolfrum, N., Burleigh,

L., Scott, C.C., Verheije, M.H., Mercer, J., et al. (2012). Single-cell analysis of

population context advances RNAi screening at multiple levels. Mol. Syst.

Biol. 8, 579.

Szklarczyk, D., Franceschini, A., Kuhn, M., Simonovic, M., Roth, A., Minguez,

P., Doerks, T., Stark, M., Muller, J., Bork, P., et al. (2011). The STRING data-


base in 2011: functional interaction networks of proteins, globally integrated

and scored. Nucleic Acids Res. 39(Database issue), D561–D568.

Teale, A., Campbell, S., Van Buuren, N., Magee, W.C., Watmough, K.,

Couturier, B., Shipclark, R., and Barry, M. (2009). Orthopoxviruses require

a functional ubiquitin-proteasome system for productive replication. J. Virol.

83, 2099–2108.

Townsley, A.C., Weisberg, A.S., Wagenaar, T.R., and Moss, B. (2006).

Vaccinia virus entry into cells via a low-pH-dependent endosomal pathway.

J. Virol. 80, 8899–8908.

Villa, N.Y., Bartee, E., Mohamed, M.R., Rahman, M.M., Barrett, J.W., and

McFadden, G. (2010). Myxoma and vaccinia viruses exploit different mecha-

nisms to enter and infect human cancer cells. Virology 401, 266–279.

Welsch, S., Doglio, L., Schleich, S., and Krijnse Locker, J. (2003). The vaccinia

virus I3L gene product is localized to a complex endoplasmic reticulum-asso-

ciated structure that contains the viral parental DNA. J. Virol. 77, 6014–6028.

Supplemental Information

EXTENDED EXPERIMENTAL PROCEDURES

siRNA Screen Reverse TransfectionReverse transfection of siRNAwas performed by seeding the cells into a well containing 20ml of transfectionmix. The transfectionmix

for onewell contained 0.1ml LP2000 (QIAGEN), 14.9ml Optimemand 5ml of 1mMsiRNA (diluted in RNase free ddH2O). In eachwell, 900

cells were seeded in 50 ml/well of complete medium (10% FCS and glutamax). Cells were then left at 37�C for 72h prior to infection.

InfectionA single sucrose purified VACV stock (see Experimental Procedures) was utilized for the entire screening procedure. Amean infection

index of �10% was chosen to allow for a dynamic range of possible upregulating and downregulating hits. Infection index is

a combined function of the amount of virus added and the properties of the host cells (Snijder et al., 2009). Therefore, the necessary

amount of virus needed to achieve 10% infection was determined using ‘‘checkerboard’’ assays (see Figure S1).

Fixation and Staining

All assays were fixed by adding formaldehyde to a final concentration of 4% at 8hpi. Cells were incubated for 30min, followed by two

PBS washes. To assure a homogeneous DAPI stain, cells were permeabilized for 3 min with PBS containing 0.1% Triton X-100. After

one PBS wash, the cells were incubated with PBS containing 1mg/ml DAPI for 10 min. Before scanning, a final wash with ddH2O was

performed to minimize contaminants when imaging with automated microscopes.

ImagingAll imageswere acquired on 2 automated wide field cellWoRxmicroscopes (Applied Precision) with a 103 objective, and 2x2 binning

per pixel. Multiwell plates were loaded onto the cellWoRx microscopes using Freedom Evo robotics from Tecan. 33 3 directly adja-

cent images were taken per well, covering over�90% of each well surface. An image-based auto-focus was performed on the DAPI

signal for each imaged site. The images were recorded with 12-bit CCD cameras, and stored as individual 16-bit uncompressed TIFF

files per imaged site and per channel.

Computational Image AnalysisFor the primary screen 1,451,520 images (combining both DAPI andGFP channel) were analyzed, amounting to 1.2TB of image data.

The images were analyzed with open-source image analysis software (CellProfiler; Carpenter et al., 2006), combined with previously

published software for supervised classification of cellular phenotypes (Ramo et al., 2009), and image analysis methods described

previously (Snijder et al., 2009, 2012). All computational image analysis was run on the high-performance computing cluster Brutus,

and data storage on the central IT services, both from the ETH Zurich. The analysis resulted in 540GB of MATLAB analysis result files,

describing �200 measurements for each of 6.24x108 cells.

CellProfiler Image Analysis

The CellProfiler (Carpenter et al., 2006) image analysis pipeline was as follows: first, nuclei were identified based on the DAPI stain.

Next, cell boundaries were estimated using nuclear expansion. Standard CellProfiler texture, intensity, size, and shape features were

extracted from nucleus and cell regions, as well as from complete images, for all available channels. In total, around 200 features

were measured per individual cell.

Supervised Classification of Cellular Phenotypes

We applied supervised machine learning using the open source support vector machine (SVM) learning tool CellClassifier (Ramo

et al., 2009), to identify biologically and technically relevant cellular phenotypes. Using relevant subsets of the �200 raw single-

cell features we classified all cells in each screen into 5 distinct binary phenotype classes (infected/non-infected, interphase/non-

interphase, mitotic/non-mitotic, apoptotic/non-apoptotic, blob/non-blob). Classification was performed as described (Snijder

et al., 2009, 2012). In total, 6.24x108 cells were analyzed, of which 15% were discarded by the SVM based selection (cell segmen-

tation errors, out-of-focus image regions, and experimental contaminants), leaving 5.44x108 high-quality cells in the final analysis.

After SVM cleanup, the median correlation coefficient for infection over replicate plates was 0.45, and for the total number of cells

per well it was 0.89, indicating a high reproducibility of the applied screening pipeline.

Hit Scoring

Genes were scored according their changed level of infection after several steps of data normalization. Cytotoxic RNAi knockdown

phenotypes were filtered from the hit scoring by excluding wells in which fewer than 300 cells passed the SVM-based quality control.

Fold-change of infection was calculated for the remaining wells as the log2 transform of the measured infection divided over the pre-

dicted level of infection for that well. Well infection predictions were calculated given the population-context (See Figure S1). This

model-predicted value represents the expected infection level in the absence of any direct effect of the RNAi perturbation. The

fold change of infection over this prediction therefore represents the RNAi perturbation effect that is independent of changes in

the population context. We show this normalization to improve screening results in general and for VACV infection specifically in

(Snijder et al., 2012). These fold-changes of infection were next corrected for column and row-wise plate effects, and plate-wise

z-score normalized (subtract plate mean, then divide by plate stdev). The median of the triplicate z-scores was determined for

each individual siRNA. As three siRNAs per gene were used, we next calculated the median of the three siRNAs scores correspond-

ing to each gene. The genes were then ranked from lowest to highest based on these median-of-median scores. Selection of the 276

Cell Reports 2, 1–12, October 25, 2012 ª2012 The Authors S1

genes that were followed up was done by manually selecting the genes with lowest ranking scores taken from two analyses that

differed only in their use of the population-context normalization step.

The gene phenotypes for the secondary screen, performed in experimental triplicates with three independent siRNAs from

a different vendor (Ambion) compared to the primary screen, were calculated in the same way as described above, except that

the population-context dependent model was built per plate from cells derived from negative control wells (containing either non-

targeting siRNA or only the RNAi transfection reagent). These infection scores were compared against a set of negative control wells

with a two tailed t test. Those genes with a probability p < 0.01 for belonging to the control wells distribution were scored as validated

hits, resulting in 188 genes of the original 276 being validated (68%) (see below).

DAVID Annotation Clustering Network

We performed functional annotation clustering and enrichment scoring of the 188 validated hits for VACV infection using DAVID

(http://david.abcc.ncifcrf.gov) (Huang da et al., 2009) version 6.7, with ‘‘high’’ classification stringency settings. We included all stan-

dard annotation categories except UP_SEQ_FEATURE. This yielded 76 functional annotation clusters, of which 15 highly enriched

(EASE score above 1) against the background list of all genes tested in the druggable genome. However, the clusters showed signif-

icant overlap regarding the genes they described, with many genes being associated in multiple clusters (see below).

We therefore next visualized the overlap between these functional annotation clusters by drawing clusters as nodes in a network,

with edges representing the fraction of genes shared between two nodes as a function of the cluster with fewest genes, displayed in

Figure S2. All edges above 75% gene-identity (of the smaller cluster) were depicted. We further added edges of 33% or higher for

otherwise unconnected annotation cluster nodes. The code used to generate the visualization was implemented in Matlab, and

requires Cytoscape for visualization. This code has been tested for use on both PC and Mac and can be found with a short manual

in the Supplementary Materials (draw_david_network.m).

Combined DAVID and STRING Functional Gene View

The functional annotation clustering was combined with the functional interactions described in the STRING database for the 188

validated host factors (version 9) (Szklarczyk et al., 2011). To create the combined network viewwe assigned each gene to the cluster

with the highest enrichment, in which the genewas present in the highest fraction of individual annotations. For visualization purposes

the ATP/GTP-binding cluster was removed from analysis. Genes not clustered in any of the DAVID clusters were left unconnected,

and selected unclustered genes were placed back in the visualization as inverted arrowheads. We next added all STRING interac-

tions (yellow solid lines) with a confidence score of 0.9 or higher, and all interactions of 0.35 or higher between genes within the same

functional annotation cluster. The functional annotation clusters were manually placed in their approximate cellular locations in the

final image (see Figure 1B). Some smaller annotation clusters and unconnected genes (52 genes in total) were left out of the visual-

ization due to space limitation. As above, the code used to generate the visualization was implemented in Matlab, and requires Cyto-

scape for visualization. This code has been tested for use on both PC and Mac and can be found in the Supplementary Materials

(draw_string_david_combination_network.m).

24 Hr Viral Yield

HeLa cells were infected with WT WR (MOI = 1) in the presence of inhibitors and harvested 24 hr post infection (hpi). Cells were

washed with PBS, resuspended in 1 mM Tris-HCl pH 9.0 and disrupted by three consecutive freeze/thaw cycles and the virus

concentration determined by serial titration on BSC-40 cells.

Analysis of DNA Replication Site Formation

HeLa cells were infected with EGFP-VACV (MOI 1) in the presence of the indicated inhibitors and fixed 5 hpi with 4% formaldehyde.

Draq5 (Biostatus, UK) was used to visualize cellular and viral DNA. One hundred cells per condition were scored for the presence of

cytoplasmic replication sites.

Quantification of Core Breakdown

Preliminary studies indicated that 4 hr post infection in the absence of virion pre-binding was the optimal condition to assess core

breakdown in untreated cells and stabilized cores in cyclohexamide control cells. Using this experimental setup cells were infected

with WR-EGFP-CORE virions with 400 virions per cell in the presence of the indicated inhibitors. Four hpi cells were fixed and

membrane-containing virions distinguished using an antibody against L1R (1:10,000; a gift of A. L. Schmaljohn, University of Mary-

land School of Medicine, Baltimore, MD) followed by Alexa594 secondary antibody. Imaging was performed on a Zeiss LSM 510

Meta confocal microscope. Images were taken using the appropriate excitation and emission settings with a 100x objective. Ten

confocal Z-sections were collected for each cell and displayed as a maximum projection. To quantify the intracellular cores L1R/

EGFP positive spots representing external virions were subtracted from the EGFP images. Images representing intracellular cores

were then inverted, subjected to auto-thresholding and cores quantified with the ImageJ particle analysis plugin (Size pixel>.05-5;

Circularity 0.0-1.0). 20 cells per condition from 3 independent experiments were analyzed.

Electron Microscopy

For conventional electron microscopy, WT VACV infected cells were fixed in 2.5% glutaraldehyde (0.05 M sodium cacodylate

adjusted to pH 7.2, 50 mM KCl, 2.5 mM CaCl2) for 60 min at RT. After several washes with 50 mM sodium cacodylate buffer,

post-fixed in reduced OsO4 (2% OsO4, 1.5% potassium ferrocyanide in water) for 1 hr on ice, followed by block staining in 0.5%

aqueous uranyl acetate over night. The specimens were then serially dehydrated in ethanol and propylene oxide, followed by embed-

ding in Epon. Ultrathin sections (50-60 nm) were obtained using a FC7/UC7-ultramicrotome (Leica, Vienna, Austria). Sections were

S2 Cell Reports 2, 1–12, October 25, 2012 ª2012 The Authors

http://david.abcc.ncifcrf.gov

acquired with a CM10 Philips transmission electron microscope with an Olympus ‘‘Veleta’’ 2k x 2k side-mounted TEM CCD camera.

Raw images were adjusted for brightness and contrast. Images are representative of 3 independent experiments.

SUPPLEMENTAL REFERENCES

Carpenter, A.E., Jones, T.R., Lamprecht, M.R., Clarke, C., Kang, I.H., Friman, O., Guertin, D.A., Chang, J.H., Lindquist, R.A., Moffat, J., et al. (2006). CellProfiler:

image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 7, R100.

Griffiths, G., and Hoppeler, H. (1986). Quantitation in immunocytochemistry: correlation of immunogold labeling to absolute number of membrane antigens. J.

Histochem. Cytochem. 34, 1389–1398.

Huang da, W., Sherman, B.T., and Lempicki, R.A. (2009). Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Pro-

toc. 4, 44–57.

Ramo, P., Sacher, R., Snijder, B., Begemann, B., and Pelkmans, L. (2009). CellClassifier: supervised learning of cellular phenotypes. Bioinformatics 25, 3028–

3030.

Snijder, B., Sacher, R., Ramo, P., Damm, E.M., Liberali, P., and Pelkmans, L. (2009). Population context determines cell-to-cell variability in endocytosis and virus

infection. Nature 461, 520–523.

Snijder, B., Sacher, R., Ramo, P., Liberali, P., Mench, K., Wolfrum, N., Burleigh, L., Scott, C.C., Verheije, M.H., Mercer, J., et al. (2012). Single-cell analysis of

population context advances RNAi screening at multiple levels. Mol. Syst. Biol. 8, 579.

Szklarczyk, D., Franceschini, A., Kuhn, M., Simonovic, M., Roth, A., Minguez, P., Doerks, T., Stark, M., Muller, J., Bork, P., et al. (2011). The STRING database in

2011: functional interaction networks of proteins, globally integrated and scored. Nucleic Acids Res. 39(Database issue), D561–D568.


Figure S1. Extended Experimental Screening Procedures, Related to Figures 1, 2, 3, and 4

(A) siRNA screen plate layout. Each plate of the primary screen had 37 mock (no RNAi), three transfection controls (RNAi against PLK1, see total cell number;

right), four AllStars scrambled siRNA controls, and 16 no-virus controls.

(B) Checkerboard assay for determination of optimal virus amount. The optimal virus amount needed for �10% infection, at 8 hr postinfection (hpi), over a wide

range of cell numbers was determined by seeding a gradient of HeLa MZ cells in the columns of a 384-well plate and infecting with a gradient of virus in the rows.

Results for the selected final virus dilution are shown on the right; the local mean (black line) and local standard deviation (outer gray area) of 144 wells are shown.

Note that the primary screen is performed with an average total cell number of �7,000 cells per well.

(C) Typical population context dependent VACV infection observed for amultiwell plate in the primary screen. Well infection predictions were calculated given the

population context (population size, local cell density, cell-islet edge, cell size) dependency of VACV infection measured over the entire 384-well plate (a typical

example from the primary screen and the measured population-context of individual cells for that well are displayed.

(D) Number of times a gene appears in different functional annotation clusters. Shown are genes present in 4 or more clusters based on functional annotation

clustering and enrichment scoring of the 188 validated hits for VACV infection using DAVID with ‘‘high’’ classification stringency settings.


Figure S2. Network View of 73 Functional Annotation Clusters Identified by DAVID 6.7 Within the 188 Host Factors Required for VACV

Infection, Related to Figure 1 and Table S2

Network nodes represent individual functional annotation clusters. Node sizes reflect the number of factors in each cluster (see legend). Node color represents

the enrichment relative to the total genes screened (see legend and supplementary information). Lines between the functional clusters indicate shared genes. The

line thickness defines the percentage (33%–100%) of genes in the smaller cluster that are shared with the larger cluster. Solid lines represent all edges of at least

75% similarity, while dashed lines represent the single strongest edge of 33% or higher for otherwise unconnected clusters. Individual networks are outlined in

dashed gray lines. The largest network (1) consists of 98 genes, occurring 233 times in 25 functional annotation clusters, and is centered around both ATP/GTP

binding genes and proteasomal genes. The second largest network (2) consists of 93 genes occurring 191 times in 18 functional annotation clusters and contains

transcriptional regulators and nuclear lumen annotated genes involved in mRNA and protein transport. Smaller networks include: (3) Membrane intrinsic genes

including cation channels and G protein coupled receptors; (4) Ion binding genes with (Ring-type) zinc-finger domains, EGF binding domains, and DNA poly-

merases; (5) positive and negative regulators of kinase activity; and (6) apoptosis. Highly enriched clusters not in larger networks included the spliceosome,

nuclear mRNA splicing, and the ribosome. Less enriched functional clusters included chemotaxis, cell-migration, mitosis, axon-guidance, laminin G, lipid

degradation, heparin binding, and vesicle (membrane) annotated genes.


Figure S3. Proteasome Inhibition by Velcade Blocks VACV Infection Akin to MG132, Related to Figures 2 and 4

(A) Proteasome inhibitor (Velcade) blocks VACV production (MOI 1; 24 hpi; 10-100 nM Velcade).

(B) Velcade has no impact on early, but inhibits late VACV gene expression (MOI 1; 6 hpi). Cells infected in the presence of CHX or ARAC served as controls for

gating early and late gene expressing cells.

(C) Velcade inhibits VACV infection after internalization. VACV-EGFP (MOI 1) was bound to cells for 1 hr in the presence of 100 nMVelcade. Cells werewashed and

treated with pH 7.4 or pH 5.0 media for 5 min, after which media containing 100 nM Velcade was added back. Cells were analyzed by flow cytometry for EGFP at

6 hpi.

(D) Velcade blocks VACV DNA replication site formation. Cells were infected with VACVMVs (MOI 1) in the presence of 10 mMAraC or 100 nMVelcade. Cells were

fixed and stained for DNA at 4 hpi. Representative images (left); Mean ± SD of triplicate experiments (right).

(E) Velcade stabilizes cytoplasmic VACV cores (assay performed in parallel to Figure 2E).


Figure S4. Quantification of internalized cores, Related to Figure 2

Ten confocal Z-sections were collected for each cell and displayed as a maximum projection. In the merged image virus cores are (green), the virus membrane

(red), and intact virions (yellow). To quantify the stabilized intracellular cores (Total cores), the virus membrane signal (Membrane) was subtracted from core

images. Thus both the red signal and overlapping red and green signals were removed from the core channel, leaving only intracellular cores (Cores with no

membrane). These images were inverted and subjected to auto-thresholding. Cores were then quantified with the ImageJ particle analysis plugin with settings

(Size pixel^.05-5; Circularity 0.0-1.0), which assured that only individual cores were counted with 94% accuracy. 20 cells per condition from 3 independent

experiments were analyzed.


Figure S5. Quantification of internalized cores visualized by EM, Related to Figure 2

(A) The number of stabilized cores in control cells and cells infected in the presence of CHX,MG132, or UBEI-41 inmicrographs of 0.1 mm thin epoxy sections were

quantified by point counting. For each sample condition, 25 micrographs were acquired at a magnification of 5800x on a 2k x 2k camera (pixel size = 8.36 nm). A

square lattice grid (2 cm grid spacing) was placed on printed micrographs to cover eachmicrograph completely. The number of cores per grid point in the cytosol

were counted and analyzed as published by Griffiths and Hoppeler, 1986.

(B) Cores are stabilized in the absence of proteasome, but not UBEI-41 function. The average number of cores observed per cell and per unit volume in untreated,

CHX, MG132, and UBEI-41 EM samples.


Figure S6. MG132 Does not Impact Core Expansion, Related to Figure 2

Cores within intact external virions (black bars) and internalized cytoplasmic cores in the presence of cycloheximide (CHX; light gray bars) or MG132 (dark gray

bars) were measured for both length and width, as illustrated at the lower right. EM micrographs of untreated and treated infected cells were used for

measurements. When compared to cores within virions, the length and width of cores in the cytoplasm were found to be significantly different (**** upper right

table; p < 0.5). However, the length and width of cores within the cytoplasm of cells treated MG132 were no different from those found in CHX-treated samples.

This indicated that MG132 had no impact on core expansion. A minimum of 40 cytoplasmic cores and 80 intact virions from each condition were measured.


Figure S7. The I3 Protein Is not Accessible in Intact VACV MVs, Related to Figure 3

VACV MVs, which package the A5 core protein as a mCherry fusion, were left untreated (UNTR) or were treated with 1% NP40/50 mM DTT to remove the viral

membrane. These virions were bound to a coverslip and stained with anti-I3L antibody, followed by secondary staining with a 488-coupled secondary antibody.

The I3 protein could only be detected in samples that had been treated with NP40/DTT.


Figure S8. VACV Virions Cannot Be Isolated in the Presence of UBEI-41, Related to Figure 3

HeLa cells were infected at an MOI of 1 with VACV-EGFP-A5. At 6 hpi, UBEI-41 (50 mM) was added to one set of cells, a time when late gene expression was

determined to be unaffected (Figure 3D). Infections were allowed to proceed for a total of 24 hr before cells were harvested and progeny virions were collected

and purified over a 25%–40% sucrose gradient. Virions can be visualized as a white light-scattering band. Virions could not be obtained from samples treated

with UBEI-41. Virions from untreated (UNTR) cells were purified in parallel for comparison (white band). One ml was collected from the same region of both

gradients and titered to determine the number of infectious particles produced. For the untreated sample�13 109 infectious particles were recovered. When the

same region of the UBEI-41 treated gradient was measured by plaque titration �1 3 106 infectious particles were recovered.


rnai screening reveals proteasome- and cullin3-dependent stages in vaccinia virus infection

Documents