LETTERS

Platelet-derived growth factor-a receptor activationis required for human cytomegalovirus infectionLiliana Soroceanu1, Armin Akhavan1 & Charles S. Cobbs1,2

Human cytomegalovirus (HCMV) is a ubiquitous human herpes-virus that can cause life-threatening disease in the fetus and theimmunocompromised host1. Upon attachment to the cell, thevirus induces robust inflammatory, interferon- and growth-fac-tor-like signalling2–9. The mechanisms facilitating viral entry andgene expression are not clearly understood4. Here we show thatplatelet-derived growth factor-a receptor (PDGFR-a) is specif-ically phosphorylated by both laboratory and clinical isolates ofHCMV in various human cell types, resulting in activation of thephosphoinositide-3-kinase (PI(3)K) signalling pathway. Uponstimulation by HCMV, tyrosine-phosphorylated PDGFR-a associ-ated with the p85 regulatory subunit of PI(3)K and induced pro-tein kinase B (also known as Akt) phosphorylation, similar to thegenuine ligand, PDGF-AA. Cells in which PDGFR-a was genetic-ally deleted10 or functionally blocked were non-permissive toHCMV entry, viral gene expression or infectious virus production.Re-introducing human PDGFRA gene into knockout cells restoredsusceptibility to viral entry and essential viral gene expression.Blockade of receptor function with a humanized PDGFR-a block-ing antibody (IMC-3G3)11 or targeted inhibition of its kinase activ-ity with a small molecule (Gleevec)12 completely inhibited HCMVviral internalization and gene expression in human epithelial,endothelial and fibroblast cells. Viral entry in cells harbouringendogenous PDGFR-a was competitively inhibited by pretreat-ment with PDGF-AA. We further demonstrate that HCMV gly-coprotein B directly interacts with PDGFR-a, resulting in receptortyrosine phosphorylation, and that glycoprotein B neutralizingantibodies13 inhibit HCMV-induced PDGFR-a phosphorylation.Taken together, these data indicate that PDGFR-a is a criticalreceptor required for HCMV infection, and thus a target for novelanti-viral therapies.

HCMV, a b-herpesvirus, is the most common cause of congenitalinfection and an important pathogen in immunocompromised indi-viduals1. Viral attachment elicits a potent cellular interferon-like res-ponse2,5,6,9,14 which activates downstream growth-factor-like receptortyrosine kinase (RTK) and integrin signalling pathways4,15. HCMVmodulation of the PI(3)K/Akt pathway is an important mechanismof apoptotic inhibition, ensuring long-term virus survival16. Priorevidence suggested that HCMV activation of human epidermalgrowth factor receptor (EGFR), in conjunction with integrin co-receptors, facilitates activation of downstream signalling moleculessuch as PI(3)K/Akt, phospholipase Cc and focal adhesionkinase15,17,18. However, we and others demonstrated that EGFR isnot required for cellular expression of HCMV-essential genes, orfor virus-induced signalling19,20. Instead, we found that upon short-term infection of human cells, HCMV caused phosphorylation of anapproximately 180-kDa protein, distinct from EGFR, that could beco-immunoprecipitated with the p85 regulatory subunit of PI(3)K19.

Therefore, we hypothesized that another, as yet undiscovered, RTKmight be activated by HCMV and mediate viral entry and expression.

To identify this putative RTK, we used a human phospho-specificRTK antibody array to screen human embryonic lung fibroblasts(HELs) that were either mock- or HCMV (Towne strain)-infectedfor 10 min. Only PDGFR-awas highly tyrosine phosphorylated uponinfection with HCMV (Fig. 1a). Western blot analyses of these sameprotein lysates using a different phospho-specific antibody forPDGFR-a corroborated this observation (Fig. 1b). Independentquantitative enzyme-linked immunosorbent assays (ELISAs) con-firmed PDGFR-a phosphorylation by Towne, AD169 and TR21

strains (Fig. 1c). HCMV did not induce tyrosine phosphorylationof the related RTK, PDGFR-b (Supplementary Fig. 1). Ultraviolet-inactivated HCMV induced PDGFR-a phosphorylation, whereasvirus inactivated by heat did not (Supplementary Fig. 2).

To determine whether the HCMV-induced PDGFR-a phosphor-ylation was cell-type specific, we used U87 glioma (neuro-epithelialorigin), HEL (fibroblast) and human umbilical vein endothelial cells(HUVECs, mesenchymal origin). In all three cell types, infection withHCMV Towne induced PDGFR-a phosphorylation, similar to thegenuine ligand (Fig. 1d).

Based on these data, we hypothesized that PDGFR-a was theapproximately 180 kDa protein we previously identified associated withthe p85 regulatory subunit of PI(3)K, upon HCMV attachment19. Toconfirm this, we conducted co-immunoprecipitation experiments inHCMV or mock-infected cells. Immunoblotting of p85PI(3)K immuno-precipitated proteins and whole-cell lysates with antibodies specific forPDGFR-a and phosphotyrosine indicated that PDGFR-a was tyrosinephosphorylated and associated with p85PI(3)K upon HCMV short-termstimulation (Fig. 2a). Specificity of PDGFR-a phosphorylation by theTowne, AD169 and TR strains was examined in the presence or absenceof IMC-3G3, a humanized PDGFR-a blocking antibody (ImClone11).Pretreatment with IMC-3G3 significantly inhibited PDGFR-a phos-phorylation induced by all HCMV strains tested (Fig. 2b). To invest-igate HCMV-induced activation of the downstream PI(3)K/Aktsignalling pathway, we used both the IMC-3G3 antibody and aPDGFR-a kinase inhibitor, imatinib mesylate (Gleevec12), to blockeither PDGFR-abinding or its activation. Akt phosphorylation inducedby either HCMV or PDGF-AA was abolished by IMC-3G3 and Gleevec(Fig. 2c). An isotype-matched negative control antibody did not inhibitHCMV- or PDGF-induced PDGFR-a or Akt phosphorylation (notshown). Thus, blocking PDGFR-a or inhibiting its activity preventsHCMV-mediated activation of the PI(3)K/Akt signalling pathway, animportant pathway in the HCMV viral life cycle22.

To determine whether PDGFR-a is critical for viral internalizationand gene expression, we used a well-characterized viral entry assay tomeasure internalization of the pp65 viral tegument protein afterHCMV-treated cells were shifted from 4 uC to 37 uC23. We used

1Department of Neurosciences, California Pacific Medical Center Research Institute, Suite 220, 475 Brannan Street, San Francisco, California 94107, USA. 2Department ofNeurological Surgery, University of California, San Francisco, 787 Moffitt, 505 Parnassus Avenue, San Francisco, California 94143, USA.

Vol 455 | 18 September 2008 | doi:10.1038/nature07209

391

©2008 Macmillan Publishers Limited. All rights reserved

human and murine cells engineered to knockout, knockdown oroverexpress PDGFR-a. As shown in Fig. 3a, 60 min after shifting to37 uC, HEL cells demonstrate evidence of viral internalization, indi-cated by immunostaining of nuclear pp65 in over 65% of cells permicroscopic field (top row). Short interfering RNA (siRNA)-mediated knockdown of PDGFR-a (Fig. 3a, lowest two rows) causednear-complete blockade of viral internalization, compared with thenon-targeting, control siRNA-treated cells (P , 0.001, Fig. 3b).Similarly, murine fibroblasts obtained from embryos of Pdgfraknockout mice10 (embryonic lethal) showed no pp65-positive nuclei(Fig. 3a, second row) whereas in fibroblasts from parental strain(325S) over 25% of cells were pp65 positive (Fig. 3a, third row).Re-introducing human PDGFRA into the knockout cells restoredand augmented HCMV internalization in these cells (Fig. 3a, fourthrow, 80% of cells are pp65 positive) even compared with positivecontrol cells (Fig. 3b, P , 0.01, Student’s t-test).

To determine whether genetic ablation of PDGFR-a prevents cel-lular expression of essential HCMV gene products, we measured

expression of IE1 (UL123) in murine cells, as well as IE1 and pp65(UL83) in human cells after infection with HCMV. Viral geneexpression was undetectable in the PDGFR-a knockout murine cells(Fig. 3c) and in the human HEL cells pretreated with PDGFRAsiRNA(Fig. 3d), compared with controls. Prolonged activation ofhuman PDGFR-a by HCMV resulted in downregulation of receptorlevels, consistent with a recent report24 (Fig. 3d, upper panel, lane 2,in control siRNA-treated cells). IE1 expression was also undetectableafter infection of the PDGFR-a null mouse fibroblasts with theAD169 strain, indicating that this effect is not strain specific (datanot shown). Pretreatment with PDGFR-a blocking agents IMC-3G3antibody and Gleevec completely inhibited HCMV IE1 proteinexpression in human HEL and U87 glioma cells (Fig. 3e) as well asin HUVECs (Supplementary Fig. 3).

We next investigated whether PDGFR-a expression was requiredfor production of infectious virus using siRNA knockdown ofPDGFR-a in HEL cells and a Towne-green fluorescent protein(GFP)-expressing virus25 for visualization of infected cells and plaqueformation. Forty-eight to sixty hours after siRNA transfection, HELcells were exposed to HCMV for 1 h and monitored daily under afluorescence microscope. Duplicate cultures were used for IE1 stain-ing at 12 h after infection, whereas a third set of cultures was used tomeasure plaque formation (Supplementary Figs 4 and 5). Six daysafter infection, supernatants of these cells were used to infect naiveHEL cells and assess production of infectious virus (Fig. 3f andSupplementary Fig. 4). We found near-complete inhibition of both

0.25

0.20

0.10

0.05

00 0.5 1 2.5 5

0.15

Ab

sorb

ance

(arb

itrar

y un

its)

d

U87

MG

HE

LsH

UV

EC

s

HCMV: 10 min PDGF-AA: 10 min Mock: 10 min100 µm

kDa

10 25 50 100

190 -p-PDGFR-α

Moc

kPDGF-

AAHCM

V

PDGFR-α

a bMock HCMV

p-PDGFR-α

Towne

TR HCMVPDGF-AA

AD169

MOI

c

pp65CMVp-PDGFR-α

pp65CMVp-PDGFR-α

pp65CMVp-PDGFR-α

Figure 1 | HCMV induces tyrosine phosphorylation of human PDGFR-a.a, Lysates of mock-or HCMV-treated cells were hybridized to a humanphospho-RTK array. HCMV phosphorylates PDGFR-a (arrow). b, Westernblot of HELs stimulated with mock, HCMV, or PDGF-AA, with indicatedantibodies. c, Phospho-PDGFR-a-specific ELISA of HELs stimulated withindicated HCMV strains and PDGF-AA (dotted line corresponds to4,000 pg ml21 phospho-PDGFR-a). Mean absorbance values (n 5 6) 6 s.d.are shown. d, Immunofluorescence of pp65 and phospho-PDGFR-a inindicated cells stimulated with mock, HCMV or PDGF. Nuclei are stainedwith 4,6-diamidino-2-phenylindole (DAPI).

Western blot

PDGFR-α

p-Tyr

pp65HCMV

Moc

k

HCMV

PDGF-AA

Moc

k

HCMV

PDGF-AA

I.P.: p85PI(3)K Cell lysates

a

Ab

sorb

ance

194 -165 -194 -165 -

59 -

kDa

IMC-3

G3PD

GF-AA

IMC-3

G3+PD

GFTo

wne

IMC-3

G3+To

wneAD

169

IMC-3

G3+AD

169

TR H

CMV

IMC-3

G3+ T

R HCM

V

kDa

65 -

40 -

p-Akt

Akt

c

b

Moc

k

PDG

F-AA

HCMV

PDG

F-AA

HCMV

PDGF-

AAHCM

V

IMC-3G3 Gleevec

P < 0.001 P < 0.001 P < 0.005 P < 0.0050.25

0.20

0.15

0.10

0.05

0

Figure 2 | HCMV activates the PI(3)K/Akt pathway in a PDGFR-a-dependent manner. a, HEL cells stimulated with HCMV or PDGF-AA weresubjected to immunoprecipitation and western blot analyses with indicatedantibodies. b, Phospho-PDGFR-a ELISA of cells with or withoutpretreatment with IMC-3G3 (10mg ml21, 2 h) followed by HCMV(MOI 5 1) or PDGF (10 ng ml21) for 10 min. Mean values (n 5 6) 6 s.d. areshown. c, Western blot of HEL mock, HCMV or PDGF-AA stimulated(10 min), with or without IMC-3G3 (2 mg ml21) or Gleevec (100 nM). Thesame membrane was used for p-AktSer 473 and Akt.

LETTERS NATURE | Vol 455 | 18 September 2008

392

©2008 Macmillan Publishers Limited. All rights reserved

viral gene expression and infectious virus production in cells that didnot express PDGFR-a at the time of infection. Plaque formation inHEL cells was also completely blocked by PDGFR-a knockdown(Supplementary Fig. 5).

To determine the relative importance of PDGFR-a versus down-stream PI(3)K activation for viral gene expression and infectiousvirus production, we performed a series of experiments usingp110aPI(3)K (also known as PIK3CA) siRNA and non-targetingsiRNA-treated cells. Suppression of the PI(3)K pathway by p110aknockdown resulted in a delay in viral gene expression, yet allowedviral entry and infectious virus production, albeit at lower levels thancontrols, which is in agreement with previous studies using PI(3)Kinhibitors22 (Supplementary Fig. 6). These data indicate thatalthough PI(3)K activation is important for the HCMV life cycle,expression of a functional PDGFR-a is essential.

We next tested whether the authentic PDGFR-a ligand inhibits viralentry. Pretreatment with PDGF-AA significantly decreased HCMVentry in HEL cells, suggesting that PDGF-AA competes with anHCMV protein (Supplementary Fig. 7). Because HCMV envelope

glycoprotein B (UL55) mediates viral entry and cellular signalling26,27,we investigated whether glycoprotein B is the viral moiety directlyinteracting with PDGFR-a. Using a modified attachment assay, wefound that a purified glycoprotein B peptide was internalized in mousecells overexpressing human PDGFR-a, but not in PDGFR-a null cells(Fig. 4a). To demonstrate a direct interaction between PDGFR-a andglycoprotein B, we performed co-immunoprecipitation experimentsusing HEL cells that endogenously express PDGFR-a and purifiedrecombinant full-length glycoprotein B28. Reciprocal immunoblotanalyses (Fig. 4b, c) demonstrate that glycoprotein B and PDGFR-aco-immunoprecipitate, indicating direct association between PDGFR-a and glycoprotein B as a bona fide mechanism for attachment/inter-nalization of HCMV into the host cells. Using phosphor-PDGFR-aELISA and western blot approaches, we found that full-length gly-coprotein B induced PDGFR-a phosphorylation (Fig. 4d, e).Furthermore, two different glycoprotein B neutralizing antibodies13

significantly inhibited HCMV-induced PDGFR-a tyrosine phosphor-ylation (Fig. 4d, e), indicating that the PDGFR-a–glycoprotein B interaction is functionally relevant. Isotype control

0 min 15 min 30 min 60 min

HE

LsP

DG

FR-α

null

325S

PD

GFR

-α n

ull

+ hP

DG

FR-α

Co-

siR

NA

HE

LsP

DG

FRA

siR

NA

HE

Ls

100 µm

200 µm

a

Co-siR

NA0

20

40

60

80

100

HELs

PDGFR-αnu

ll32

5S

PDGFR-α

null

+ hPDGFR

-αPDGFR

A

siRNA

P < 0.001P < 0.001

P < 0.01

pp

65H

CM

V-p

ositi

ve c

ells

b

1 2 3 4 1 2 3 4IE1

Actin

Mock HCMV

72-

40-

kDa

c

1 2 1 2

Co-siRNA PDGFRA siRNA

200- 180-

78-

70-

40-

PDGFR-α

IE1

pp65

Actin

kDad

1 2 3 4 1 2 3 4kDa

80-40-

IE1Actin

HELs U87

e 0

25

50

75

100

P < 0.001 P < 0.001P < 0.001

Per

cent

age

IE1

pos

itive

cel

ls

f

Co-siR

NA

PDGFRA si

RNA

Co-siR

NA

PDGFRA si

RNA

Co-siR

NA

PDGFRA si

RNA

Primary infection

Secondary infection

TOW

NE

TOW

NE

AD

169

AD

169

TR TR

Figure 3 | Human PDGFR-a is required for HCMV entry, IE1 expression andinfectious virus production. a, pp65 immunofluorescence after HCMVtreatment (1 h, 4 uC) followed by shifting to 37 uC for indicated times. Rowsrepresent (top to bottom): HELs, PDGFR-a null fibroblasts, parentalfibroblasts, null fibroblasts overexpressing hPDGFR-a, HELs transfectedwith control or PDGFRA siRNA. Nuclei were stained with DAPI.b, Average 6 s.d. pp65 positive per 100 cells counted in triplicate froma. c, Mock- or HCMV-infected cells were analysed by western blotting withthe indicated antibodies. Lanes 1–4 indicate HELs, 325S, PDGFR-a null andPDGFR-a null overexpressing hPDGFR-a, respectively. d, HELs transfected

with control or PDGFRA siRNA were mock (lanes 1) or HCMV- treated(lanes 2) and subjected to western blots with indicated antibodies. e, Westernblot of HEL and U87 lysates after infection with mock (lanes 1), HCMV(lanes 2) or HCMV pretreated with IMC-3G3 (10 mg ml21, 12 h, lanes 3) orGleevec (100 nM, 1 h, lanes 4). f, PDGFRA siRNA-treated cells infected withindicated HCMV strains and immunostained for IE1 12 h after infection(primary infection). Six days after infection, supernatants were used to infectnaive HEL cells, followed by IE1 immunostaining and quantification(secondary infection). Average (n 5 6) values 6 s.d. are shown.

NATURE | Vol 455 | 18 September 2008 LETTERS

393

©2008 Macmillan Publishers Limited. All rights reserved

antibodies or neutralizing antibodies against other viral glycoproteins(glycoproteins H and N) did not prevent HCMV-induced receptoractivation (not shown).

Overall, data presented here indicate HCMV requires PDGFR-abinding and activation for viral internalization, expression of essentialviral genes, production of infectious virus and activation of down-stream PI(3)K/Akt signalling. These findings do not exclude a poten-tially important role for other co-receptors during HCMVinternalization and expression, such as integrin receptors. We demo-nstrate that viral interaction with PDGFR-a is facilitated by directbinding of the viral glycoprotein B to PDGFR-a. We further show thatblockade of the PDGFR-a receptor pathway with pharmaceuticalagents currently in human use may prove a powerful antiviral strategyfor the management of HCMV-related disease. Because both PDGFR-aand HCMV play important roles in the pathophysiology of humandevelopment, inflammation, vascular disease and cell-proliferative dis-orders, an increased understanding of their interaction may elucidatenovel molecular mechanisms underlying these conditions.

METHODS SUMMARYCells and viruses. PDGFR-aknockout and parental 325S mouse fibroblasts were a

gift from M. Tallquist (Southwestern University10). Although HCMV is a human-

specific virus and cannot be propagated in murine cells, HCMV internalizationand HCMV immediate early viral gene expression occur in murine cells29. cDNA

for human PDGFRA was obtained from C. Heldin (Upsala University). HCMV

strains AD169 and Towne (American Type Culture Collection, ATCC) were pro-

pagated in HELs for less than five passages. HELs were infected at a multiplicity of

infection (MOI) of 1 (in serum-free media) and cell supernatant was collected over

a period of 5–7 days after infection, when the cytopathic effect was about 100%.

Virus-containing media were first centrifuged (1,500g) to remove cell debris and

further concentrated using a sucrose gradient centrifugation (80,000g, 4 uC) as

described30. Virus stock aliquots were kept at 280 uC. Mock controls were gener-

ated in parallel by conditioning and processing uninfected cultures identically. TR

HCMV clinical isolate and GFP-CMV were obtained from W. Britt.

siRNA experiments. HELs were transfected with 100 nM of either the ‘smart

Pool’ siRNA to human PDGFRA, human p110aPI(3)K or the non-targeting siRNA

pool (Dharmacon) using standard Lipofectamine 2000 reagent protocol

(Invitrogen). Seventy-two hours after transfection, siRNA-transfected cells were

HCMV- or mock-infected. PDGFR-a or p110aPI(3)K expression levels were

measured using standard western blot and immunofluorescence analyses (anti-

bodies from Cell Signalling).

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 15 May; accepted 25 June 2008.Published online 13 August 2008.

1. Britt, W. J. & Alford, C. A. Fields Virology 3rd edn (Raven Press, 1996).2. Andreoni, K. A., Wang, X., Huang, S. M. & Huang, E. S. Human cytomegalovirus

hyperimmune globulin not only neutralizes HCMV infectivity, but also inhibitsHCMV-induced intracellular NF-kB, Sp1, and PI3-K signaling pathways. J. Med.Virol. 67, 33–40 (2002).

3. Boyle, K. A., Pietropaolo, R. L. & Compton, T. Engagement of the cellular receptorfor glycoprotein B of human cytomegalovirus activates the interferon-responsivepathway. Mol. Cell. Biol. 19, 3607–3613 (1999).

4. Compton, T. Receptors and immune sensors: the complex entry path of humancytomegalovirus. Trends Cell Biol. 14, 5–8 (2004).

5. Netterwald, J. R. et al. Postattachment events associated with viral entry arenecessary for induction of interferon-stimulated genes by humancytomegalovirus. J. Virol. 78, 6688–6691 (2004).

6. Ozato, K., Tailor, P. & Kubota, T. The interferon regulatory factor family in hostdefense: mechanism of action. J. Biol. Chem. 282, 20065–20069 (2007).

7. Simmen, K. A. et al. Global modulation of cellular transcription by humancytomegalovirus is initiated by viral glycoprotein B. Proc. Natl Acad. Sci. USA 98,7140–7145 (2001).

8. Yurochko, A. D. et al. Induction of the transcription factor Sp1 during humancytomegalovirus infection mediates upregulation of the p65 and p105/p50 NF-kB promoters. J. Virol. 71, 4638–4648 (1997).

9. Zhu, H., Cong, J. P. & Shenk, T. Use of differential display analysis to assess theeffect of human cytomegalovirus infection on the accumulation of cellular RNAs:induction of interferon-responsive RNAs. Proc. Natl Acad. Sci. USA 94,13985–13990 (1997).

10. Andrews, A. et al. Platelet-derived growth factor plays a key role in proliferativevitreoretinopathy. Invest. Ophthalmol. Vis. Sci. 40, 2683–2689 (1999).

11. Loizos, N. et al. Targeting the platelet-derived growth factor receptor alpha with aneutralizing human monoclonal antibody inhibits the growth of tumor xenografts:implications as a potential therapeutic target. Mol. Cancer Ther. 4, 369–379(2005).

12. Sandler, C. et al. Imatinib mesylate inhibits platelet derived growth factorstimulated proliferation of rheumatoid synovial fibroblasts. Biochem. Biophys. Res.Commun. 347, 31–35 (2006).

13. Britt, W. J. Neutralizing antibodies detect a disulfide-linked glycoprotein complexwithin the envelope of human cytomegalovirus. Virology 135, 369–378 (1984).

14. Waltenberger, J. et al. Different signal transduction properties of KDR and Flt1,two receptors for vascular endothelial growth factor. J. Biol. Chem. 269,26988–26995 (1994).

15. Feire, A. L., Koss, H. & Compton, T. Cellular integrins function as entry receptorsfor human cytomegalovirus via a highly conserved disintegrin-like domain. Proc.Natl Acad. Sci. USA 101, 15470–15475 (2004).

16. Cooray, S. The pivotal role of phosphatidylinositol 3-kinase-Akt signaltransduction in virus survival. J. Gen. Virol. 85, 1065–1076 (2004).

17. Wang, X., Huang, D. Y., Huong, S. M. & Huang, E. S. Integrin avb3 is a coreceptorfor human cytomegalovirus. Nature Med. 11, 515–521 (2005).

18. Wang, X. et al. Epidermal growth factor receptor is a cellular receptor for humancytomegalovirus. Nature 424, 456–461 (2003).

19. Cobbs, C. S. et al. Human cytomegalovirus induces cellular tyrosine kinasesignaling and promotes glioma cell invasiveness. J. Neurooncol. 85, 271–280(2007).

Max

imal

stim

ulat

ion

(%)

120100806040200

e

PDGF

gB Mab

758

+HCM

V

HCMV

Moc

k

pY754-PDGFR-αPDGFR-α

100 µm

IF: g

BIF

: PD

GFR

-α

PDGFR-α OE/gB PDGFR-α OE/mock PDGFR-α KO/gB

205150100

kDa Western blot PDGFR-α Western blot gB

1 2 3 4 5 6 7 8 9 10 11 12 13 14 Lysate

gBI.P. gB

I.P. PDGFR-α

+ + + + – – + + + + + – – –

b c

+ + – – + + – + + – – + + + + + + + + ++ + + + + +

PDGF

7–17

Neu

t gB

Mab

758

Neu

t gB

HCMV

HCMV+7–

17

HCMV+M

ab 7

58

PDGF+7–

17

PDGF+M

ab 7

58

gB (6

80)

P = 0.5P < 0.001

P < 0.001

d

a

Figure 4 | HCMV Glycoprotein B binds and activates PDGFR-a.a, Immunofluorescence detection of glycoprotein B (upper panels) andPDGFR-a (lower panels) in PDGFR-a knockout (KO) cells and PDGFR-aoverexpressing (OE) cells, after incubation with the glycoprotein B peptideor mock treatment. Reduced PDGFR-a surface staining in glycoprotein-B-treated cells is likely due to receptor internalization.b, c, Immunoprecipitation of PDGFR-a and glycoprotein B from HELlysates and full-length soluble glycoprotein B alone or pre-incubatedtogether. Immunoprecipitates were subjected to western blot with theindicated antibodies. d, ELISA measurements of human PDGFR-aphosphorylation after stimulation with HCMV (MOI 5 1), PDGF(10 ng ml21) or recombinant soluble glycoprotein B (30 mg ml21) in thepresence or absence of glycoprotein B neutralizing antibodies 7–17 and Mab758 (5mg ml21); bars, 6 s.d. e. Portions of the same lysates used in d wereanalysed by western blot with indicated antibodies.

LETTERS NATURE | Vol 455 | 18 September 2008

394

©2008 Macmillan Publishers Limited. All rights reserved

20. Isaacson, M. K., Feire, A. L. & Compton, T. Epidermal growth factor receptor is notrequired for human cytomegalovirus entry or signaling. J. Virol. 81, 6241–6247(2007).

21. Murphy, E. et al. Coding potential of laboratory and clinical strains of humancytomegalovirus. Proc. Natl Acad. Sci. USA 100, 14976–14981 (2003).

22. Johnson, R. A. et al. Human cytomegalovirus up-regulates the phosphatidyli-nositol 3-kinase (PI3-K) pathway: inhibition of PI3-K activity inhibits viralreplication and virus-induced signaling. J. Virol. 75, 6022–6032 (2001).

23. English, E. P., Chumanov, R. S., Gellman, S. H. & Compton, T. Rational developmentof beta-peptide inhibitors of human cytomegalovirus entry. J. Biol. Chem. 281,2661–2667 (2006).

24. Gredmark, S. et al. Human cytomegalovirus downregulates expression ofreceptors for platelet-derived growth factor by smooth muscle cells. J. Virol. 81,5112–5120 (2007).

25. Murphy, E. A., Streblow, D. N., Nelson, J. A. & Stinski, M. F. The humancytomegalovirus IE86 protein can block cell cycle progression after inducingtransition into the S phase of permissive cells. J. Virol. 74, 7108–7118 (2000).

26. Boyle, K. A. & Compton, T. Receptor-binding properties of a soluble form ofhuman cytomegalovirus glycoprotein B. J. Virol. 72, 1826–1833 (1998).

27. Carlson, C., Britt, W. J. & Compton, T. Expression, purification, andcharacterization of a soluble form of human cytomegalovirus glycoprotein B.Virology 239, 198–205 (1997).

28. Wang, Z. et al. Recombinant modified vaccinia virus Ankara expressing a solubleform of glycoprotein B causes durable immunity and neutralizing antibodies againstmultiple strains of human cytomegalovirus. J. Virol. 78, 3965–3976 (2004).

29. Thomas, B. Viruses and the Cellular Immune Response (Marcel Dekker, 1993).30. Huang, E. S., Chen, S. T. & Pagano, J. S. Human cytomegalovirus. I. Purification and

characterization of viral DNA. J. Virol. 12, 1473–1481 (1973).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank M. Tallquist for the PDGFR-a knockout mousefibroblasts, C. Heldin for the human PDGFR-a cDNA, D. Diamond for providing thesoluble glycoprotein B, and N. Loizos (ImClone) for the IMC-3G3 antibody. We aregrateful to W. Britt (University of Alabama at Birmingham) for viruses,glycoprotein B neutralizing antibodies and discussions. This study was supportedby an institutional grant from California Pacific Medical Center Research Instituteand by the Arthur Flaming Foundation.

Author Contributions L.S. and A.A. performed experiments; L.S., A.A. and C.S.C.designed experiments, analysed data and wrote the manuscript.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. Correspondence and requests for materials should beaddressed to C.C. (charles.cobbs@gmail.com).

NATURE | Vol 455 | 18 September 2008 LETTERS

395

©2008 Macmillan Publishers Limited. All rights reserved

METHODSCell culture, plasmids, transfection and additional HCMV strains. HELs, U87

glioma and HUVECs were obtained from ATCC and maintained in DMEM plus

10% FCS, except for HUVECs which were grown in endothelial cell media

(Cascade Biologicals), plus growth factors. Mouse PDGFR-a knockout cells were

transfected with human PDGFRA cDNA using Lipofectamine 2000 (Invitrogen)

according to the manufacturer’s instructions. Towne-GFP (from W. Britt,

University of Alabama at Birmingham) is a recombinant HCMV strain that

expresses GFP under the early promoter UL127, as previously described31.

Virus titres were determined by IE1 immunohistochemical staining, as previ-ously described32. Optimization of siRNA cell delivery was performed by co-

transfecting with the targeting siRNA pools (for example, PDGFRA siRNA) and

fluorescently labelled oligonucleotides (siGLO-RISC free, Dharmacon) that

localize to the nucleus and allow assessment of uptake into cells. Targeting

siRNA oligonucleotides and siGLO were mixed 1:1 (50 nM each) with different

amounts of Lipofectamine 2000. Twenty-four hours later, cells were fixed, coun-

terstained with DAPI and counted. The average number of green fluorescent cells

(a measure of transfection efficiency) was between 72% and 84% (from a total of

100% DAPI-positive nuclei).

siRNA sequences. The sequences are listed 59–39. PI(3)K p110a:

GCGAAAUUCUCACACUAUU; GUGGUAAAGUUCCCAGAUA; GCUUAGA

GUUGGAGUUUGA; GACCCUAGCCUUAGAUAAA. Human PDGFRA: CG

AGACUCCUGUAACCUUAUU; GAGCUUCACCUAUCAAGUUUU; GACAG

UGGCCAUUAUACUAUU; GAAUAGGGAUAGCUUCCUGUU. Non-targeting

sequences: UGGUUUACUAGUCGACUAA; UGGUUUACAUGUUUUCUGA;

UGGUUUACAUGUUGUGUGA; UGGUUUACAUGUUUUCCUA.

Western blot, immunoprecipitation and immunofluorescence analyses. For

stimulation experiments, cells were serum-starved for 24 h, followed by stimu-lation with HCMV (MOI 5 0.5), PDGF-AA (5–10 ng ml21, R&D Systems), or

mock, for 10 min. In some cases, cells were pretreated with IMC-3G3 (N. Loizos,

ImClone) at 10 mg ml21 (2 or 12 h) or Gleevec (100 nM, 1 h) before HCMV

exposure. Cell lysates were analysed by SDS–polyacrylamide gel electrophoresis

(SDS–PAGE). Antibodies used were as follows: polyclonal anti-PDGFR-a (1/

500), phosphor-tyrosine clone 4G10 (1/1,000) and monoclonal anti-p85 PI(3)K

(1/1,000) (Upstate Biotechnology), anti HCMV IE1 MAB810 (Chemicon;

1/1,000), anti HCMV pp65 (Novocastra, 1/1,000), anti-phosphor-PDGFR-a(1/500, pTyr 754), anti-phosphor-Akt (Ser 473, 1/1,000) and total Akt (1/

1,000; all from Cell Signalling). Anti-actin polyclonal control antibody was used

(1/500, Sigma). Immunoprecipitations were performed using protein G (Pierce)

according to the manufacturer’s instructions. For immunofluorescence, we used

monoclonal HCMV pp65 and phosphor-PDGFR-a (p-Tyr 754, Santa Cruz

Biotechnology), overnight at 4 uC, followed by incubation for 1 h with secondary

antibodies conjugated to Alexa 488, or Alexa 568 (1/5,000, Molecular Probes).

Nuclei were counterstained with DAPI. Co-immunoprecipitation experiments

were performed using full-length soluble purified recombinant glycoprotein B

(glycoprotein B 680, from D. Diamond, City of Hope, California) and detergent-soluble extracts of HEL cells generated in lysis buffer (1% NP-40, 75 mM NaCl

and 50 mM Tris-HCl). Protein (500mg) from total cell extract was incubated

with 25mg of glycoprotein B 680 in lysis buffer in the presence of an anti-

glycoprotein B (Virusys) or anti-PDGFR-a (R&D) antibody overnight at 4 uC.

Immune complexes were recovered with protein A–Sepharose beads (2 h

incubation), denatured and separated on SDS–PAGE for western blot with

anti-glycoprotein B (Virusys) or anti-PDGFR-a (Cell Signaling Technology)

antibodies.

Viral attachment and internalization assays. Mouse fibroblasts and HEL cells

(72 h after siRNA transfections) were incubated with HCMV (MOI 5 0.5) or

mock treated for 1 h at 4 uC, after which cells were returned to 37 uC for 15, 30 or

60 min. Cells were fixed using methanol (20 min) and processed for pp65 immu-

nofluorescence. Four low-power fields were counted for each condition and

pp65 immunoreactive cells were recorded for each 100 cells. Internalization

assays were repeated twice.

HCMV glycoprotein B (peptide) attachment assays. Glycoprotein B peptide

binding experiments were performed similar to viral attachment assays. After

incubation for 1 h with the glycoprotein B peptide (100 nM, 4 uC), cells were

returned to 37 uC for 60 min, washed, fixed and processed for double immuno-

fluorescence for glycoprotein B (1mg ml21, monoclonal antibody, Virusys) and

PDGFR-a (2mg ml21, Upstate). The glycoprotein B peptide (Ray Biotech) con-

tains amino acids 27–84 from the AD169 strain and 27–81 from the Towne

strain.

Human phospho-PDGFR-a ELISA. ELISA for human phospho-PDGFR-a was

performed with a kit (R&D, catalogue number DYC2114-2). HEL cells were

grown in 24-well plates (40,000 cells per ml) and serum-starved 48 h before

short-term (10 min) stimulation with various agents, as described in Figs 2b

and 4d. Lysis of cells was done as per kit instructions. The capture antibody

was a mouse anti-human PDGFR-a; anti-phosphotyrosine-HRP antibody was

used for detection. Recombinant human phosphorylated PDGFR-a was used as

a positive control. Reaction products were read using a microplate reader set at

450 nm. All samples were run in triplicate and each experiment was repeated at

least twice. For receptor-blocking experiments, cells were pretreated with IMC-

3G3 (10mg ml21, 12 h) or Gleevec (100 nM, 1 h). To test the effects of glycopro-

tein B neutralizing antibodies, HCMV (MOI 5 1) was pre-incubated in serum-

free media with antibodies specific for glycoprotein B 7–17, MAB 758 (ref. 33) or

control isotype-matched antibodies (5mg ml21) for 1 h before cell stimulation.

Additional antibodies tested included neutralizing antibody against HCMV gly-

coproteins N (ref. 34) and H (ref. 35).

Measurements of infectious virus production. HEL cells (where indicated

treated with siRNA) were infected with HCMV Towne or CMV-GFP for 1 h,

washed and grown for 6 days at 37 uC. A duplicate set of cultures was analysed by

immunofluorescence at 12 h after infection to assess the percentage of IE1 posi-

tive cells, as a measure of ‘primary infection’. Six days after infection, supernatant

from these cells were centrifuged to exclude cell debris and used to infect naive

HEL cultures as described above. Cells at 12 h after infection were stained for IE1,

and IE1-positive cells were counted among a total of 100 cells per low-magnifica-

tion microscopic field, four fields per condition. These counts were used to

determine the level of ‘secondary infection’: that is, infectious virus production

from the primary infected cells. Where indicated, CMV–GFP infected cells were

monitored daily under a fluorescence microscope. Each condition was assayed in

triplicate.

Plaque formation assays. Plaque formation was assayed as previously

described36. Briefly, confluent HEL cells in six-well cluster plates were incubated

with CMV-GFP (MOI 5 1, 1 h) in 0.5 ml growth media. Cells were washed and

returned to 37 uC (in complete growth media). Twenty-four hours later, super-

natant was harvested and used to infect naive HEL cultures, which were monit-

ored for plaque formation for 6–14 days. Plaque formation was photographed

daily using an inverted fluorescence microscope. At day 14, plaques in ten low-

magnification fields/conditions were counted (each experimental condition was

tested in six independent wells).

Statistical analyses. A two-tailed paired Student’s t-test was used to compare

data sets and obtain P values for all comparisons; P values are indicated on figure

panels.

31. Isomura, H. & Stinski, M. F. The human cytomegalovirus major immediate-earlyenhancer determines the efficiency of immediate-early gene transcription andviral replication in permissive cells at low multiplicity of infection. J. Virol. 77,3602–3614 (2003).

32. Chan, G., Stinski, M. F. & Guilbert, L. J. Human cytomegalovirus-inducedupregulation of intercellular cell adhesion molecule-1 on villoussyncytiotrophoblasts. Biol. Reprod. 71, 797–803 (2004).

33. Britt, W. J., Jarvis, M. A., Drummond, D. D. & Mach, M. Antigenic domain 1 isrequired for oligomerization of human cytomegalovirus glycoprotein B. J. Virol. 79,4066–4079 (2005).

34. Shimamura, M., Mach, M. & Britt, W. J. Human cytomegalovirus infection elicits aglycoprotein M (gM)/gN-specific virus-neutralizing antibody response. J. Virol.80, 4591–4600 (2006).

35. Li, L., Coelingh, K. L. & Britt, W. J. Human cytomegalovirus neutralizing antibody-resistant phenotype is associated with reduced expression of glycoprotein H. J.Virol. 69, 6047–6053 (1995).

36. Mar, E. C., Cheng, Y. C. & Huang, E. S. Effect of 9-(1,3-dihydroxy-2-propoxymethyl)guanine on human cytomegalovirus replication in vitro.Antimicrob. Agents Chemother. 24, 518–521 (1983).

doi:10.1038/nature07209

©2008 Macmillan Publishers Limited. All rights reserved