Article

KDEL Receptors Assist Dengue Virus Exit from theEndoplasmic Reticulum

Graphical Abstract

Highlightsd Depletion of KDELR by siRNA reduces egress of DENV

progeny

d DENV1-3 structural protein prM interacts with KDELR

in the ER

d KDELR/prM interaction requires three positively charged

amino acids at N terminus of prM

d Disrupting this interaction inhibits DENV1 RSPs transport

from the ER to the Golgi

Authors

Ming Yuan Li, Marc Grandadam, ...,

Roberto Bruzzone, Pei Gang Wang

Correspondencebruzzone@hku.hk (R.B.),pgwang@163.com (P.G.W.)

In BriefViral receptors are key host factors for

virion entry; however, it is not known

whether trafficking and secretion of

progeny virus also requires host

intracellular receptors. Li et al. show that

dengue virus (DENV) interacts with host

KDEL receptors (KDELR) in the ER.

Depleting KDELR, disrupting DENV/

KDELR interaction or blocking KDELR

cycling between the ER and Golgi reduce

virus release, resulting in virus

accumulation in the ER. The authors

propose that KDELR functions as

intracellular receptors to assist in DENV

exit from the ER.

Li et al., 2015, Cell Reports 10, 1496–1507March 10, 2015 ª2015 The Authorshttp://dx.doi.org/10.1016/j.celrep.2015.02.021

Cell Reports

Article

KDEL Receptors Assist Dengue VirusExit from the Endoplasmic ReticulumMing Yuan Li,1 Marc Grandadam,2 Kevin Kwok,1 Thibault Lagache,3,4 Yu Lam Siu,1 Jing Shu Zhang,1 Kouxiong Sayteng,2

Mateusz Kudelko,1,5 Cheng Feng Qin,6 Jean-Christophe Olivo-Marin,3,4 Roberto Bruzzone,1,4,* and Pei Gang Wang1,7,*1HKU-Pasteur Research Pole and Centre of Influenza Research, School of Public Health, LKS Faculty of Medicine, The University of HongKong, Hong Kong SAR, China2Institut Pasteur du Laos, Vientiane, Lao PDR3Unite d’Analyse d’Images Biologiques, CNRS URA 2582, Department of Cell Biology and Infection, Institut Pasteur, 75015 Paris Cedex,France4Department of Cell Biology and Infection, Institut Pasteur, 75015 Paris Cedex, France5Department of Biochemistry, LKS Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China6Department of Virology, Beijing Institute of Microbiology and Epidemiology, Beijing 100071, PR China7Key Laboratory of Protein and Peptide Pharmaceuticals, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, PR China*Correspondence: bruzzone@hku.hk (R.B.), pgwang@163.com (P.G.W.)http://dx.doi.org/10.1016/j.celrep.2015.02.021This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

SUMMARY

Membrane receptors at the surface of target cells arekey host factors for virion entry; however, it is un-known whether trafficking and secretion of progenyvirus requires host intracellular receptors. In thisstudy, we demonstrate that dengue virus (DENV) in-teracts with KDEL receptors (KDELR), which cyclebetween the ER and Golgi apparatus, for vesiculartransport from ER to Golgi. Depletion of KDELR bysiRNA reduced egress of both DENV progeny and re-combinant subviral particles (RSPs). Coimmunopre-cipitation of KDELR with dengue structural proteinprM required three positively charged residues atthe N terminus, whose mutation disrupted proteininteraction and inhibited RSP transport from theER to the Golgi. Finally, siRNA depletion of class IIArfs, which results in KDELR accumulation in theGolgi, phenocopied results obtained with muta-genized prME and KDELR knockdown. Our resultshave uncovered a function for KDELR as an internalreceptor involved in DENV trafficking.

INTRODUCTION

Dengue, a mosquito-borne viral infection endemic in over100 countries, is caused by four serotypes of dengue virus(DENV1–4). In addition to a febrile, influenza-like illness, severedengue represents a public health concern in Asia and SouthAmerica where it is a major cause of death across all ages (Guz-man et al., 2010; Messina et al., 2014). Despite the global burdenof disease, there is no specific treatment and, therefore, amolec-ular understanding of host-pathogen interactions during thecellular life cycle is needed to guide the development of effectivedrugs (Guzman et al., 2010).

DENV has two structural glycoproteins: pre-membrane (prM)and envelope (E) (Kuhn et al., 2002); E mediates interactionwith cellular receptor(s) for viral attachment and entry (Chenet al., 1997), whereas prM assists E in its correct folding (Coura-geot et al., 2000) and protects it from pre-fusion in the acidicenvironment of the secretory pathway (Zhang et al., 2003). As-sembly of DENV occurs at the ER and requires interaction ofprM and E (Mukhopadhyay et al., 2005; Pryor et al., 2004).Nascent virions bud into the lumen of the ER, accumulating indilated cisternae oriented toward the cis-Golgi, and are translo-cated to the Golgi via trafficking vesicles (Welsch et al., 2009). Inthe trans-Golgi network (TGN), prM protein is cleaved by thecellular protease furin, resulting in the release of the pr peptideand formation of infectious DENV (Li et al., 2008; Yu et al.,2008). Besides mature virions, non-infectious recombinant sub-viral particles (RSP) can be produced by cells expressing DENVprME proteins (Mukhopadhyay et al., 2005). Dengue RSP trafficalong the same compartments as infectious DENV, and repre-sent a safe and convenient tool for the study of virus-host inter-actions during secretion (Wang et al., 2009).Although DENV egress has been studied for many years, most

cellular targets identified in high-throughput screens have notbeen mapped to the secretory pathway (Sessions et al., 2009).We recently identified two cellular factors, ADP-ribosylationfactor 4 and 5 (Arf4 and Arf5), which are involved in secretionof DENV progeny (Kudelko et al., 2012). Because Arfs play animportant role in the recruitment of coat proteins necessary forthe formation of trafficking vesicles (D’Souza-Schorey andChavrier, 2006), our results indicate that Arf4+5 are acting atan early step of DENV secretion (Kudelko et al., 2012). The spe-cific involvement of Arfs, which are dispensable factors for theconstitutive pathway, in DENV trafficking suggested that thevirus uses amore complex machinery and that other cellular fac-tors besides Arf4+5 might also assist to exit the infected cell.Sorting of cargo is dependent on molecular recognition, a

process equivalent to receptor-ligand interactions; however, itis not known whether newly formed DENV exploits host factors

1496 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

to move along the secretory pathway. Intriguingly, depletion ofArf4+5 has also been reported to inhibit the retrograde traf-ficking of KDEL receptor (KDELR) from Golgi to ER (Volpicelli-Daley et al., 2005). The three KDELR members (KDELR1–3)identified (Hsu et al., 1992; Lewis and Pelham, 1990, 1992b;Raykhel et al., 2007) are transmembrane proteins cycling be-tween ER and Golgi apparatus to prevent leakage of ER-resi-dent proteins, such as chaperones, and retrieve them back tothe ER (Lewis and Pelham, 1990). As KDELR binding to cargothrough a C terminus KDEL motif occurs only in the Golgi appa-ratus, we investigated their possible involvement in transloca-tion of DENV from assembly and budding sites in the ER tothe Golgi.We show here that KDELR1 and KDELR2 play crucial roles for

DENV1–3, but not DENV4 secretion. KDELR interacted withDENV through three positively charged amino acids at the N ter-minus of prM. DENV secretion could be blocked either by de-pletion of KDELR, arrest of KDELR cycle, or disruption of prM/KDELR interaction. Under these conditions, progeny DENVaccumulated in the ER and did not reach the Golgi apparatus.Our results demonstrate that KDELRs function as luminal recep-tors for DENV transport along the secretory pathway.

RESULTS

KDELR Interact with prM of DENV1We previously demonstrated that depletion of Arf4+5 inhibitedDENV1 and RSP release without disrupting constitutive secre-tion (Kudelko et al., 2012). To gain insight into the underlying mo-lecular mechanism, we investigated the role of KDELRs, whichaccumulate in a peri-nuclear region and are not recycled backto the ER following Arf4+5 depletion (Volpicelli-Daley et al.,2005). All three KDELR identified thus far were detected byRT-PCR in our cellular models (Figure 1A). We observed a redis-tribution of KDELR in cells stably transfected with prME (HeLa-prME-DENV1), with an apparent reduction of co-staining withthe cis-Golgi marker GM130 in comparison to parental HeLa(Figure 1B). Moreover, partial co-localization of E and KDELRwas observed in either HeLa-prME-DENV1 or Vero E6 cells in-fected with DENV1 (Figure 1C). Although a prominent aggrega-tion of E protein was seen in HeLa-prME-DENV1 (Figure 1C),this did not reflect different distribution, as prME co-localizedwith ER marker in both cell lines (data not shown). Similar resultswere obtained in cells co-transfected with DENV1 prME andKDELR1-RFP (Figure S1A). These observations suggested theparticipation of KDELRs in DENV1 life cycle.We next performed co-immunoprecipitation (coIP) using either

dengue patient serum (DPS), containing antibodies recognizingprME (Kudelko et al., 2012), or normal human serum (NHS) ascontrol. Pellets of coIPs were analyzed with western blotting(WB) with antibodies recognizing either prM and E, or the threeKDELR. prME glycoprotein was specifically pulled down byDPS, but not NHS (Figure 1D). Although expression levels werecomparable, KDELR were detected only in pellets from HeLa-prME-DENV1, but not parental cells (Figure 1D). Similarly,when coIP was performed with replication-competent DENV1,a strong signal for KDELRwas visible only in pellets from infectedVero E6 cells (Figure 1E), confirming a biochemical interaction

between KDELR and prME. KDELR were also precipitatedfrom lysates of stable cell lines producing RSP of DENV2 andDENV3, but not DENV4 (Figure 1F), suggesting a certain degreeof specificity between serotypes.To investigate which portion of the envelope glycoprotein, prM

or E, was responsible for interaction with KDELR, the 4G2mono-clonal antibody, recognizing E but not prM, was used in coIPassays. KDELR were not present in pellets obtained following in-cubation with the 4G2 monoclonal (Figure 1G), indicating that Ewas not responsible for interaction with KDELR. These observa-tions were corroborated by detecting KDELR in coIP pellets withthe prM-6.1 monoclonal antibody, which recognizes prM but notE protein (Figure 1G). The biochemical interaction between prMand KDELR was further validated by coIP in 293T cells co-trans-fected with prME and KDELR1-myc (Figure S1B). To conclu-sively define the role of prM and E in the interaction with KDELR,we used two complementary approaches. First, glutathioneS-transferase (GST)-fusion proteins with truncated prM frag-ments were incubated with lysates of cells stably transfectedwith KDELR1-myc. Pull-down assays showed that prM-DTM,the full-length protein without the transmembrane domain(130 amino acids), pr fragment (91 amino acids) and the first40 amino acids of prM sequence (pr40) could all interact withKDELR (Figure 1H), revealing that the amino-terminal domainof prM was sufficient to mediate interaction with KDELR.Second, mature and immature RSP were used as baits to pulldown KDELR from cell lysates. Immature RSP were producedin the presence of NH4Cl, an inhibitor of furin, and thus containedfull-length prM and E (Wang et al., 2009). Mature RSP were pro-duced from immature RSP after in vitro cleavage by furin, whichreleased the pr fragment and, therefore, contained only E andM.KDELR could only be detected when immature, but not matureRSP were incubated with HeLa cell lysates and then subjectedto immune-precipitation (Figure 1I). These results demonstratethat interaction with KDELR was dependent on the N-terminalpr fragment of prM, as its release from immature RSP preventedpull-down of KDELR.

KDELR Knockdown Reduces Secretion of DENV RSPTo investigate the impact of the interaction between KDELR andprM on DENV1 life cycle, we transfected HeLa-prME-DENV1cells that constitutively secrete RSP with siRNAs targeting allthree KDELR; this resulted in an 81% ± 4% (n = 3) reduction ofKDELR protein (Figure 2A). Silencing of KDELR did not affectcell viability, as determined by propidium iodide staining (Fig-ure S2A), or morphology (Figure S2B). Our results show thatdepletion of KDELR had no effect on intracellular E proteinexpression, but significantly reduced RSP secretion (Figure 2A).To test whether the effect of KDELR on RSP release was part

of a general mechanism that would interfere with the constitutivesecretory pathway, we analyzed ssHRP release (Bard et al.,2006; Kudelko et al., 2012). No difference was observed insecreted ssHRPor intracellular HRP activity after downregulationof KDELR or Arf4+5, when compared to controls (Figure 2B). Incontrast, secretion of ssHRP-KDEL occurred only in cells treatedwith KDELR or Arf4+5 siRNAs, confirming that both manoeuvershad interfered with retrieval of KDELR to the ER, resulting in aparallel reduction of HRP activity in cell lysates (Figure 2B). These

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1497

Figure 1. KDELR Interact with prM Glycoprotein of DENV1-3(A) Expression of all three KDELR (KDELR1–3) isoforms was detected by RT- PCR in the cell lines used in our experiments. GAPDH was used as control for the

amount of cDNA template.

(B) In contrast to parental HeLa, endogenous KDELR (red) did not accumulate in cis-Golgi (anti-GM130, green) in cells stably expressing prME of DENV1 (prME).

(C) Endogenous KDELR (red) and DENV1 prM and E proteins (revealed with an anti-prME polyclonal antibody, green) were partially co-localized in HeLa-prME-

DENV1 (prME, top) and DENV1 infected Vero E6 cells (MOI = 0.1, bottom); scale bar represents10 mm.

(legend continued on next page)

1498 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

results demonstrate that KDELR specifically assisted RSPrelease and their involvement was independent of perturbationof the constitutive pathway.To study the role of individual KDELR on DENV1 secretion,

siRNAs targeting individual KDELR were transfected into HeLa-prME-DENV1 in various combinations. Although individual knock-downs of KDELR did not induce significant changes, KDELR1+2and KDELR1+2+3 siRNAs drastically reduced RSP release(Figure 2C), indicating both a crucial role for KDELR1/KDELR2and functional compensation between these two isoforms.As KDELR depletion caused the release of KDEL-carrying pro-

teins (Figure 2B), the effect on DENV1 secretion might have been

(D) KDELRwere pulled downwith dengue patient serum (DPS) fromHeLa-prME-DENV1 (prME) but not parental HeLa cells. Normal human serum (NHS) was used

as control. Cell lysates (CL) or pellets following immunoprecipitation (IP) were analyzed withWB. InWB of IP pellets revealed with anti-E antibody, theweak bands

detected with NHS (middle) corresponded to IgG heavy chains of the IP antibody.

(E) CoIP with DPS pulled down KDELR from lysates of DENV1 infected Vero E6 (5 days postinfection; MOI = 0.01) but not uninfected cells.

(F) KDELR were precipitated by coIP from cells stably expressing prME of DENV1–3, but not DENV4. Cell lysates (CL) from HeLa-prME-DENV1–4 (labeled

DENV1–4 on top of the gel) were collected for coIP assay as described above. Whereas DPS could pull down comparable amounts of prM and E proteins of

DENV1–4, KDELRs were only detected in DENV1–3 IP pellets.

(G) KDELR were not pulled down by coIP using antibody 4G2, which recognizes E but not prM. In contrast, KDELRs were detected when the antibody prM-6.1,

which recognizes prM but not E, was used.

(H) GST-fusion proteins of prM fragments pulled down c-myc-tagged KDELR. Cell lysates (input) and pull-down pellets were revealed withWBwith anti-GST and

anti-c-myc antibodies.

(I) Immature (i) but not mature (m) RSP pull down KDELR. HeLa cells lysates were incubated with purified mature or immature RSP and then subjected to IP using

DPS or NHS.

Results are representative of at least three independent experiments.

Figure 2. KDELR Knockdown ReducesSecretion of DENV1 RSP(A) HeLa-prME-DENV1 cells were transfected with

siRNAs against all three KDELR isoforms, Arf4/Arf5

(Arf4+5), and non-targeting (NT). Knockdown effi-

ciency was determined with WB. E protein in su-

pernatants (SN) and cell lysates (CL) was depicted

by WB, quantified by densitometry and RSPs

secretion was expressed as the percentage of E

protein detected in SN relative to the total amount

(SN+CL). GAPDH was used as the loading control

across wells. Results are means ± SD (n = 3, right).

(B) KDELR downregulation did not inhibit consti-

tutive secretion. HeLa cells expressing a secreted

form of horseradish peroxidase (ssHRP) or ssHRP-

KDEL were transfected with siRNAs targeting

KDELR (gray) or Arf4+5 (black). Secreted and

intracellular ssHRP horseradish peroxidase activ-

ity was measured in both SN and CL. Results are

means ± SD (n = 3).

(C) HeLa-prME-DENV1 cells were transfected with

siRNAs targeting KDELR either individually or in

various combinations. SN and CL were analyzed

by WB to assess RSP secretion (anti-E), release of

chaperones (anti-Bip and anti-calreticulin) and

knockdown of KDELR (KR-10). GAPDH was used

as the loading control across wells. Results are

representative of three independent experiments.

*p < 0.05; **p < 0.005 versus control (NT) siRNAs.

simply due to shortage of chaperones,such as Bip and calreticulin, which arerequired for the assembly of DENV (Lim-jindaporn et al., 2009). This possibility

was excluded by showing that, although both proteins were de-tected in supernatants from cells treated with KDELR1+2 andKDELR1+2+3 siRNAs, their amount in cell lysates was notappreciably modified when compared to controls (Figure 2C).These experiments also show that only KDELR1 and KDELR2were necessary for chaperones retention in the ER (Figure 2C),suggesting that KDELR isoforms assisting DENV1 secretionand retrieving ER-resident proteins were the same. Since KR-10 antibody could not distinguish between the three KDELR (Fig-ure 2C), the efficacy of siRNA targeting each isoform was inde-pendently verified in cells expressing tagged KDELR1-3 (Fig-ure S2C) (Kudelko et al., 2012).

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1499

KDELR Knockdown Reduces DENV EgressWenext investigated the effect of KDELR knockdown on replica-tive DENV1 in Vero E6 cells. At 3 and 5 days post-infection, titersof progeny virus from KDELR-depleted cells were significantlylower than those of control cells (Figure 3A). A similar reductionwas also found in cells treated with Arf4+5 siRNA (Figure 3A; Ku-delko et al., 2012). Besides viral titers, a drastic reduction of Eprotein levels was detected in supernatants from cells treatedwith KDELR or Arf4+5 siRNAs (Figure 3A, inset). A significantdecrease of DENV1 egress was also observed in KDELR-depleted cells challenged with different MOI (Figure 3B). Infurther experiments, Vero E6 cells were co-transfected withKDELR and Arf4+5 siRNAs before being challenged withDENV1. Measurements of progeny virus titer and E proteinshowed a similar inhibition in comparison to control cells (Fig-ure 3C), suggesting that KDELR and Arf4+5 converged on thesame pathway to interfere with DENV1 secretion. The efficiencyof siRNA treatments was verified by WB (Figure S3A). We nexttested in parallel the impact of KDELR silencing on all fourDENV serotypes and observed a significant reduction of viralprogeny titer for both DENV2 and DENV3, but not DENV4 (Fig-ure 3D), consistent with the finding that only DENV1–3 wereable to interact with KDELRs (see Figure 1F). Control experi-ments confirmed that siRNA treatment did not affect cellmorphology and viability (Figures S3B and S3C). We then inves-tigated the impact of KDELRs on egress of West Nile Virus

(WNV), another flavivirus transmitted by mosquito vectors(Campbell et al., 2002) and found that viral progeny titer fromKDELR-depleted cells was not different from that measured incontrols (5.6 ± 2.5 3 109 versus 6.7 ± 1.6 3 109, respectively;mean ± SD of n = 6 from two independent experiments). Theseresults are in keeping with our previous findings that differentflaviviruses budding in the ER do not rely on the same cellularfactors for intracellular traffic (Kudelko et al., 2012).Finally, to exclude that the effect of KDELR knockdown on

DENV1 egress was the consequence of changes in the earlystages of the virus life cycle, we infected Vero E6 cells for18 hr, less than the minimum time required for newly formed vi-rions to be released from infected cells (Lindenbach and Rice,2001). Our experiments show that similar amounts of viral RNAwere measured in cells pre-treated with either KDELR or controlsiRNAs (Figure S3D), indicating that KDELR knockdown had noeffect on early stages of the viral life cycle.

prM/KDELR Interaction Occurs in the ERBecause KDELR are shuttling between ER and Golgi apparatus(Lewis and Pelham, 1992a; Raykhel et al., 2007), we designedexperiments to ascertain in which compartment the prM/KDELRinteraction occurred. We found that, in cells treated with Arf4+5siRNA-, almost all KDELR signal was co-localized with GM130,whereas in controls only a small fraction exhibited co-stainingwith the cis-Golgi marker (Figure 4A). Therefore, we took

Figure 3. KDELR Knockdown ReducesDENV Egress(A) KDELR knockdown reduced DENV1 virus

titer. Vero E6 cells were transfected with KDELR

siRNAs (gray) before being challenged with

DENV1 (MOI = 0.1). Non-targeting (NT, open) and

Arf4+5 (black) siRNAs served as negative and

positive controls, respectively. Viral titers were

measured at 1, 3, and 5 days postinfection (p.i.).

Results are means ± SD (n = 3). SN were

collected at 5 days p.i. and E protein was de-

tected by WB (inset).

(B) siRNA-treated Vero E6 cells were challenged

with DENV1 at increasing MOIs and viral titers

were measured 5 days p.i. Results are means ±SD

(six to eight observations from two independent

experiments). Inset showsWB of E protein in SN at

5 days p.i.

(C) The effect of simultaneous knockdown of

KDELR and Arf4+5 (K+A, hatched) on DENV1 titers

was measured at 1 and 5 days p.i. Results are

means ± SD (n = 3). E protein in SN was detected

by WB (inset).

(D) KDELR knockdown inhibited secretion of

DENV2 and DENV3, but not DENV4 progeny vi-

rus. Cells were treated as described in (A) and

infected at an MOI of 0.1. Progeny virus was

measured at 5 days p.i. Results are means ± SD

of three to four measurements from one of two

independent experiments with similar results.

E protein in SN was detected in parallel by WB

(inset). *p < 0.05; **p < 0.005 versus control (NT)

siRNAs.

1500 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

advantage of this observation, which identifies the cis-Golgi asthe peri-nuclear region where KDELRwas sequestered followingArf4+5 depletion (Volpicelli-Daley et al., 2005), to perform coIP inArf4+5 knockdown cells. Despite similar expression levels in allexperimental conditions, KDELR could be precipitated onlyfrom control, but not Arf4+5 depleted cells (Figure 4B), indicatingthat prM/KDELR interaction occurred in the ER.

H2, R19, and K21 Are Key Residues for prM/KDELRInteractionTo identify the putative region of prM interacting with KDELR, weanalyzed theN-terminal sequence of the pr fragment and noted ahigh proportion (seven of 26 residues) of positively chargedamino acids (Figure 5A) that were conserved in DENV1–3,whereas substitutions at residues H2 and H11 were present inDENV4 (Figure S4A). To test the role of this cluster of basic res-idues we generated prME-DENV1 constructs with neutral aminoacids substitutions (Figure 5A). Mutations did not affect expres-sion of KDELR, prM, and E, with the exception of R6S, whichreduced the levels of both viral proteins when compared towild-type prME (Figure 5B, upper; and Figure S4B). Cell lysateswere subjected to coIP and the ability of mutant prM to interactwith KDELR was determined by calculating the ratio betweenprecipitated KDELR and prM, which was then normalized tothat measured for wild-type prM (Figure 5B, middle). H2L, R6S,R19S, and K21T significantly reduced prM binding to KDELR(Figure 5B, lower) and release of RSP (Figure 5C), demonstratinga positive correlation between prM/KDELR interaction and RSPsecretion.prM/E interaction is critical for the formation of DENV (Coura-

geot et al., 2000): prM functions as the chaperone of E and its R6residue is predicted to be important for both interaction and viralassembly (Li et al., 2008). To test the effect of the mutated con-structs on prM/E interaction, we performed coIP assays on ly-sates of 293T cells and found that similar levels of prM proteincould be precipitated for all mutants with the exception of R6S(Figure S4B). These observations, while confirming the predictedinvolvement of R6 (Li et al., 2008), demonstrate that none of theother mutated residues was involved in the interaction betweenprM and E.Based on the results of coIP (Figure 5B) and RSP secretion

(Figure 5C), we generated a triple mutant (H2L/R19S/K21T)designated hereinafter as ‘‘Triple.’’ The Triple mutation did not

alter the ability to bind Arf4 and Arf5 (Figure S5), but abrogatedinteraction with KDELRs (Figure 5D) and blocked the secretionof RSP by 90% (Figure 5E), demonstrating the crucial role ofH2, R19, and K21 residues for prM interaction with KDELR andDENV1 secretion.

Triple Mutant prME Forms RSP and Is Translocatedwithin the ERTo exclude the possibility that disruption of prM/KDELR inter-action had compromised the assembly and formation of viralparticles, we performed freeze-and-thaw (F&T) experiments,which have been shown to release intracellular viruses fromhost cells (Burleson et al., 1992). Upon repeated cycles of F&T,cells expressing R6S released barely detectable levels of RSP(Figure 6A), confirming that disruption of prM/E interactioninhibited RSP formation. In contrast, similar amounts of RSPwere released from cells expressing either wild-type or TripleprME (Figure 6A), indicating that prM/KDELR interaction wasdispensable for the formation of viral particles and further sug-gesting that inhibition of DENV1 and RSP release was the conse-quence of a trafficking defect.Next, we studied the role of prM/KDELR interaction in RSP

traffic within the ER by immunofluorescence microscopy of cellsco-stained with anti-E as well as antibodies labeling ER, cis-Golgi and TGN. It has been observed that newly assembledDENV translocate within the ER to accumulate in dilatedcisternae oriented toward the cis-Golgi (Welsch et al., 2009)and, therefore, we used E protein aggregation as an index ofRSP trafficking within the ER. Aggregates were found next tothe Golgi apparatus in 50% of cells expressing either wild-typeor Triple mutant prME, but were hardly observed in cells stablytransfected with R6S (Figures 6B and 6C), suggesting that trans-port of newly assembled RSP inside ER lumen did not requireinteraction with KDELR. Similar aggregates were observed inDENV1-infected Vero E6 cells using either a polyclonal antibodyagainst prME or the anti-E monoclonal antibody 4G2 (Figure S6).

Triple Mutant RSP Cannot Exit from ERTo investigate whether interaction with KDELR was requiredfor DENV1 exit from ER, translocation of RSP to the cis-Golgiwas determined by measuring the percentage of total E proteinco-localized with GM130. We found that in cells expressingwild-type prME there was a significantly higher percentage of

Figure 4. prM/KDELR Interaction Occurs inthe ER(A) Depletion of Arf4+5 by siRNA blocked retrieval

of KDELR from Golgi. HeLa cells expressing

KDELR1-RFP (red) were transfected with Arf4+5

siRNA and stained with anti-GM130 antibody

(green). Non-targeting (NT) siRNA was used as

control. Scale bar represents 10 mm.

(B) KDELR sequestered in Golgi were not precipi-

tated by dengue patient serum (DPS). HeLa-prME-

DENV1 cells transfected with Arf4+5 or NT siRNAs

were subjected to coIP using DPS or normal human

serum (NHS). Cell lysates (CL) and pellets from

immunoprecipitation (IP) were analyzed by WB.

Results are representative of three independent

experiments.

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1501

E protein co-localized with GM-130 with respect to R6S, whichdoes not formRSP and served as the negative control, and Triplemutant, which behaved indistinguishably from R6S (Figures 6Band 6D). Furthermore, similar results were obtained when co-localization of E protein with a TGN marker was measured,with R6S and Triple mutant exhibiting a 50% reduction withrespect to wild-type prME (Figures 6B and 6E). These resultsindicate that RSP of Triple prMEwere not efficiently translocatedfrom the ER to the Golgi apparatus.

The involvement of KDELR in DENV1 exit from ER was furtherstudied by monitoring intracellular dimerization of E protein. Ithas been reported that DENV glycoprotein prME undergoes aconformational change in the Golgi apparatus, possibly causedby luminal acidification, which leads to the formation of E homo-dimers (Li et al., 2008; Yu et al., 2008). RSP released by F&Tshowed that the percentage of E/E homodimers was 3- to4-fold higher in cells expressing wild-type prME in compa-rison to Triple prME, which behaved in similar fashion to R6S(Figure 6F), corroborating immunofluorescence observations(Figures 6B–6D).

Finally, because translocation from ER to Golgi is known to bemediated by vesicular transport (Antonny and Schekman, 2001),we investigated whether KDELR/prME interaction also affectedthe formation of RSP-containing vesicles. Expression of thedifferent prME constructs did not change the distribution of ER-GIC53 (Figure 7A), which cycles between ER and Golgi (Schin-dler et al., 1993) and was used as marker of trafficking betweenthese two organelles. RSP-containing vesicles (puncta co-stained with anti-E and anti-ERGIC53) were rarely detected in

cells expressing Triplemutant or R6S prME, but readily observedwith wild-type prME (Figure 7B).

DISCUSSION

We provide here several lines of evidence to propose a role forKDELR in supporting the early steps of intracellular traffickingof both DENV1 and DENV1/RSP, namely their translocationfrom the ER to the cis-Golgi compartment. KDELR interactwith DENV1 through three positive charged amino acids at theN terminus of prM protein and DENV1/RSP egress is inhibitedeither by downregulation of KDELR, sequestration of KDELR inthe Golgi, or by disruption of prM/KDELR interaction. Our resultsfurther indicate that interaction with KDELR is important forDENV1/RSP to be licensed as cargo of trafficking vesicles leav-ing the ER. Because KDELR were not required for constitutivesecretion of soluble proteins, these findings demonstrate thatintracellular transport of DENV/RSP is regulated by interactionwith specific cellular factors and identify KDELR as an essentialcomponent of this process.The term receptor in virology refers to host plasma membrane

proteins that recognize viral structural components, triggeringreceptor-mediated endocytosis of the bound pathogen (Merceret al., 2010). The process of DENV transportation from ER toGolgi apparatus shares several similarities to viral entry. Inboth processes viruses can be viewed as cargo that is translo-cated from a neutral environment (extracellular milieu or ER) toan acidic compartment (endosome or Golgi), needs to overcomea lipid membrane barrier to reach its final destination (cytoplasm

Figure 5. prM/KDELR Interaction Is Depen-dent on Three Key Residues(A) Diagram highlighting positively charged amino

acids (aa) clustered at theN terminus of prM-DENV1.

(B) Substitutions at H2, R6, R19, and K21 signifi-

cantly reduced the ability of prM to bind KDELR in

transfected 293T cells; wild-type (WT) and empty

vector (ev) served as positive and negative controls,

respectively. Cell lysates (CL) were subjected to

coIP using dengue patient serum (DPS). prM and

KDELR were detected by WB in CL or IP pellets

(upper). Densitometric ratios of precipitated KDELR/

prM are shown as means ± SD (n = 3; lower).

(C) Substitutions at H2, R6, R19, and K21 of prM-

DENV1 reduced RSP secretion from 293T cells.

E protein in supernatants (SN) and CL was depicted

by WB, quantified by densitometry and RSPs

secretion was expressed as the percentage of

E protein detected in SN relative to total content

(SN+CL). Results are means ± SD of triplicate mea-

surements from three independent experiments.

(D) A triple prME mutant (H2L-R19S-K21T; Triple)

was completely devoid of interaction with KDELR

(no coIP with DPS).

(E) RSP were not secreted from cells transfected

with Triple mutant. RSPs release in the SN was

assessed by visualizing dengue E protein by WB

(left). Blots were quantified by densitometry and

RSPs secretion was expressed as the percentage

of E protein detected in the supernatant relative to

the total amount (SN+CL). Results are means ± SD

(n = 3, right). *p < 0.05; **p < 0.001 versus WT.

1502 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

for replication or extracellular milieu for another infection round),and is delivered in the form of trafficking vesicles (Humphries andWay, 2013; Modis, 2013; Rothman and Orci, 1992; Sun et al.,2013). Therefore, KDELR can be considered as intracellular re-ceptors for DENV/RSP trafficking, whose function is akin to therole played by cell surface proteins in mediating viral entry.

Many cellular factors have been shown to be crucial for DENVlife cycle (Sessions et al., 2009), but viral-host interactions thatassist in secretion of newly formed virions are still unclear andno host receptors mediating secretion of progeny virus areknown. Thus, although one study suggested an indirect roleof KDELR in early vaccinia virus biogenesis, by recruiting

Figure 6. RSP Formed by the Triple Mutant Do Not Exit the ER(A) Triple prME mutant (Triple), but not R6S-prME (R6S), assembled RSP. Cells were subjected to cycles of freeze-and-thaw (F&T) and E protein detected in cell

lysates (CL) or in supernatants after F&T was analyzed by WB (upper). The percentage of E released by F&T relative to the total amount (CL+F&T) was used as

index of RSP formation. Results are means ± SD (n = 3, lower).

(B) E protein localization in cells expressing wild-type (WT), Triple or R6S prME. Cells were co-stained with anti-E (4G2, green) and markers for various com-

partments (red): anti-calnexin (ER), anti-GM130 (cis-Golgi), anti-TGN46 (trans-Golgi network, TGN). Arrowheads indicate aggregates of E protein. Scale bar

represents 10 mm.

(C) The percentage of cells containing aggregates was quantified from at least three independent experiments. Results are means ± SD of the indicated number

of cells (WT, n = 3152; R6S, n = 2901; Triple, n = 2842).

(D and E) Quantification of E staining within Golgi apparatus. The region labeled with either anti-GM130 or anti-TGN46 was assigned to cis-Golgi and TGN,

respectively. The percentage of E protein in each Golgi sub-compartment was calculated as described in the Experimental Procedures. Results are means ± SD

from at least three independent experiments of the specified number of cells (for cis-Golgi: WT = 1238; R6S = 1159; Triple = 1188; for TGN: WT = 1268, R6S =

1209, Triple = 1214).

(F) Reduced dimerization of E protein in Triple mutant. RSP released after F&T were subjected to WB using anti-E antibody (left). The percentage of dimeric form

relative to total (dimeric + monomeric) E protein released by F&T is presented as means ± SD (n = 3, right). *p < 0.01; **p < 0.005 versus WT.

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1503

coatomer, their impact on intracellular transport was not investi-gated (Zhang et al., 2009). Similarly, the reduction of DENV2replication associated with KDELR downregulation had beenascribed to decreased cell surface expression of protein disul-phide isomerase, which has been proposed to function as anadditional DENV receptor (Wan et al., 2012). Our data define a re-ceptor role for KDELR in DENV egress, although it has to beacknowledged that depletion of KDELR in Vero E6 cells reducedviral titer by less than one order of magnitude. Clearly, additional,compensatory factors may assist trafficking of DENV from ER toGolgi, as indicated also by the lack of interaction betweenDENV4 and KDELRs. Further studies will be needed to identifyintracellular receptors for DENV4 as well as for other flaviviruses.

It has been demonstrated that, on immature DENV, prM sits ontop of E protein to protect it from the acidic environment alongthe secretory pathway (Zhang et al., 2003). This topology makesprM more accessible to interactions with host cellular proteins.However, the location of the key amino acid residues mutatedin our experiments is different, with R6 facing E protein, whileH2, R19, and K21 are positioned on the outside face (Li et al.,2008). This is consistent with our observation that mutation of

Figure 7. prM/KDELR Interaction AssistsVesicular Transport of DENV from ER toGolgi(A) Cells expressing wild-type (WT), Triple or R6S

prME were co-stained with anti-E (green) and anti-

ERGIC53 (red). Puncta labeled with anti-E and

anti-ERGIC53 (arrows) were rarely detected in

cells expressing Triple and R6S. Scale bar repre-

sents 10 mm.

(B) Quantification of puncta co-stained with anti-E

and anti-ERGIC53. Results are means ± SD of

the specified number of cells from at least three

independent experiments (WT, n = 1203; R6S, n =

1184; Triple, n = 1231). *p < 0.005 versus WT.

(C) Working model depicting the role of KDELR in

DENV transport. Newly formed virions assembled

in the ER exploit KDELR as luminal receptor to be

sorted as cargo of vesicles that reach the Golgi,

where they dissociate to allow KDELR to retrieve

ER resident proteins and become available for

more rounds of transport. The precise molecular

events regulating interactions of DENVwith Arf4+5

and their final trafficking remain to be elucidated.

R6 disrupted interaction with E and pre-vented assembly of RSP, whereas substi-tutions of H2, R19, and K21 abolishedinteraction with KDELR and perturbedRSP trafficking. Sequence alignmentshows that a similar cluster of positi-vely charged amino acid is present inDENV1–3 serotypes, whereas DENV4 ex-hibits non-basic substitutions at residuesH2 and H11. This difference correlateswith the inability of DENV4 to pull downKDELRs and be affected by KDELRdepletion. In keeping with our workingmodel, we found that secretion of WNV,

which lacks a high proportion of positively charged amino acidsat the N terminus of prM protein, was also unaffected by treat-ment with siRNAs targeting KDELR. As the acidic pH in the Golgiis closer to the pI of positively charged amino acids, it is temptingto speculate that this environment may facilitate dissociation ofprM/KDELR complexes by reducing their binding affinity. Incontrast, the canonical KDEL motif is more negatively charged(Wilson et al., 1993) and this difference may underlie the cargoswitching that allows retrieving of resident ER proteins fromGolgi.The finding that DENV4 was unaffected by KDELR depletion

was surprising. We had previously shown that release ofDENV4 was inhibited by knocking down Arf4+5 (Kudelko et al.,2012), which would result in sequestration of KDELR in the cis-Golgi and, therefore, reduce their availability for DENV4 traf-ficking. It is logical to postulate that DENV4 may interact in theER with an additional intracellular receptor in an Arf4+5-depen-dent manner and be able to translocate along the secretorypathway even when KDELRs were downregulated. Thus, traf-ficking of flaviviruses may require a specific complement of fac-tors for different viruses and/or strains. It should be pointed out

1504 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

that, although the Triple mutant was mainly localized in the ER, itwas still able to pull down both Arf4 and Arf5, confirming thatbinding to Arf4+5 and KDELR in the ER are independent events.Arf4+5 are localized at both Golgi and ER (Duijsings et al., 2009)and may play two crucial roles for DENV secretion, by beinginvolved in KDELR recycling and interacting with prM protein.Further experiments will be required to ascertain the preciselocation and role of class II Arf/prM interaction in DENVtrafficking.The function of prM in DENV biology is attracting more atten-

tion. Thus, prM has been recently shown to interact with the lightchain Tctex-1 of dynein and play a role in late stages of virusreplication (Brault et al., 2011). We demonstrate here that prMinteracts with KDELR during virus secretion. Our working hy-pothesis is that DENV1–3 use unoccupied KDELR, which arerecognized by a binding motif in the N terminus of prM, to exitthe ER as cargo of vesicles en route to the Golgi apparatus (Fig-ure 7C). We had previously characterized the function of class IIArf proteins in DENV/RSP egress (Kudelko et al., 2012). Simulta-neous depletion of Arf4+5 efficiently sequester intracellularKDELR in the Golgi and, therefore, it is logical to postulate thatboth factors converge on the same pathway to inhibit DENV/RSP secretion (Figure 7C). However, results with DENV4 andWNV suggest that additional host proteins are specificallyinvolved in sorting flaviviruses through late secretory compart-ments and assisting their release from infected cells. In recentyears, evidence has accumulated to suggest that, besides theirwell-established function in retrieving chaperones, KDELR canbe activated by cargo to trigger signaling pathways that regulateanterograde and retrograde traffic (Giannotta et al., 2012; Pulvir-enti et al., 2008). Specifically, it has been proposed that KDELRsrecognize chaperones that are carried by ER vesicles en route toGolgi (Cancino et al., 2013). It is tempting to speculate, therefore,that during DENV1–3 biogenesis, newly formed virions bind toKDELR to activate cell signaling pathways that facilitate theirtranslocation to the Golgi.

EXPERIMENTAL PROCEDURES

Cells, viruses, antibodies, and siRNA experiments are described in the Supple-

mental Experimental Procedures. Primers used for RT-PCR, GST pull-down,

and site-directed mutagenesis are shown in Tables S1, S2, and S3,

respectively.

Protein Analysis and RSP QuantificationGel electrophoresis and WB analysis were carried out as previously described

(Kudelko et al., 2012) To quantify RSP secretion, the area and mean lumines-

cence signals detected by WB in supernatants (SN) and cellular lysates (CL)

were measured by densitometry using Image Quant TL (Thermo Fisher). For

each condition, the relative amount of secreted RSP (E signal in SN) was calcu-

lated as the percentage of total signal (ESN/ESN+ECL).

Virus Infection ExperimentsViral stocks of DENV1-4 andWNV were titrated by determining the tissue cul-

ture infective dose 50% (TCID50/ml) in Vero E6 cells challenged with 10-fold

serial dilutions of infectious supernatants for 90 min at 37!C. Cells were

subsequently incubated in DMEM with 2.5% fetal calf serum. At 5–7 days

postinfection for DENV1–3 and 3–5 days postinfection for DENV4 and

WNV, culture supernatant was removed and cell monolayers were fixed in

4% formaldehyde. The percentage of cytopathic effects was used to calcu-

late the viral titer.

For measurements of progeny virus production, viral RNA was extracted

from culture supernatants and quantified by real-time RT-PCR (see the Sup-

plemental Experimental Procedures). The amount of viral RNA transcripts

was then calculated by generating a standard curve with 10-fold dilutions of

RNA isolated from a known amount of DENV1 stock and expressed as

TCID50/ml, as described above.

GST Pull-Down AssayFragments of the prM sequence of DENV1 were amplified by PCR (Table S2).

Ampliconswere subcloned in frame into the bacterial expression vector pGEX-

4T-1 to produce N-terminal tagged GST constructs (see the Supplemental

Experimental Procedures). Twenty micrograms of each purified protein bound

to sepharose 4B-glutathione beads was mixed with lysates of HeLa cells sta-

bly expressing cMyc-KDELR, incubated overnight at 4!C, and extensively

washed before eluting bound proteins, according to the manufacturer’s in-

structions, for WB analysis.

CoimmunoprecipitationSub-confluent monolayers of HeLa-prME-DENV1 or 293T cells transfected

with the specified constructs were lysed on ice for 30min with 1ml RIPA buffer,

supplemented with freshly added 1 mM PMSF and protease inhibitors cock-

tail. Cell debris were removed by centrifugation at 13,000 rpm for 15 min at

4!C and lysates were pre-cleared by incubation with 30 ml of 50% protein G

sepharose beads (Amersham Pharmacia) for 1 hr. Pre-cleared lysates

(400 ul) were then incubated for 2 hr at 4!Cwith additional 30 ml of 50% protein

G sepharose beads previously treatedwith either specific antibodies or control

IgGs. Beads were then pelleted by centrifugation at 13,000 rpm for 30 s at 4!C

and bound proteins were eluted by boiling in gel loading buffer, separated by

electrophoresis and analyzed with WB.

Freeze-and-Thaw AssayFor subcellular fractionation (Xu et al., 1997), sub-confluent HeLa cells stably

expressing either wild-type prME-DENV1 or the specified mutants were first

detached in PBS plus 5 mM EDTA at 37!C for 5 min and washed three times

on icewith PBS supplementedwith 1mMEGTA. Cells were then re-suspended

in a buffer containing 10% wieght/vol sucrose, 20 mM Tris HCl, 150 mM NaCl,

10 mM magnesium acetate, 1 mM EGTA (pH 7.6) supplemented with freshly

added 1 mM PMSF and protease inhibitors cocktail, and then subjected to

eight cycles of freeze (dry ice) and thaw (37!C water bath), 1 min each step.

Nuclei and cellular debris were removed by a short (5 s) spin atmaximumspeed

in a bench-top centrifuge at 4!C. Supernatants were collected and centrifuged

for 30 min at maximum speed at 4!C to pellet the membrane fraction. The final

supernatants, containing newly formed RSP, were analyzed with WB.

Fluorescence MicroscopyFor fluorescence microscopy, cells grown on glass coverslips were fixed, per-

meabilized, and incubated with primary antibodies (see the Supplemental

Experimental Procedures). Samples were then probed with appropriate sec-

ondary antibodies conjugated with fluorescein isothiocyanate or Texas Red

(both from Life Technologies). Nuclei were stained with DAPI and coverslips

were mounted on glass slides for image acquisition using either an Axio

Observer Z1 inverted microscope or an LSM 700 confocal microscope (Carl

Zeiss).

Quantitative Analysis of Fluorescent ImagesTo extract and quantify cells stained with the viral E protein, we developed

a specific protocol, ‘‘Stained cells,’’ in the ICY software (http://icy.

bioimageanalysis.org) (de Chaumont et al., 2012). To extract and quantify cells

that containedRSP aggregates, we developed a separate protocol, ‘‘Cells with

aggregates,’’ in the ICY software. Details of these protocols are provided in the

Supplemental Experimental Procedures.

To determine RSP localization in the Golgi apparatus, weighted co-localiza-

tion coefficients of E with Golgi markers were computed using the ZEN2011

co-localization coefficient software (Carl Zeiss). The sums of intensities of

pixels corresponding to anti-E (So) and to co-staining with anti-E and either

cis-Golgi or TGN marker (Sc) were computed and then weighted co-localiza-

tion coefficients, which are equal to the ratio of Sc to So, were used to

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1505

represent the percentage of RSP translocated to either cis-Golgi or TGN.

To determine the number of RSP-containing vesicles, we manually counted

(in blind) puncta co-labeled with anti-E and anti-ERGIC that were adjacent to

perinuclear E-staining. The total number of double-labeled puncta per field

was then calculated, divided by the number of cells expressing E protein

and displayed as the number of puncta per cell. Data sets for quantitative anal-

ysis were acquired from an average of 40–50 fields from four to five indepen-

dent experiments for each condition.

Statistical AnalysisResults are shown asmeans ±SD. Statistical significancewas analyzed by the

Student’s unpaired t test, with a confidence limit for significance set at 0.05 or

less.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures,

six figures, and three tables and can be found with this article online at

http://dx.doi.org/10.1016/j.celrep.2015.02.021.

AUTHOR CONTRIBUTIONS

M.Y.L., K.K., Y.L.S., and M.K. performed the experiments on prM/KDELR

interaction. M.Y.L., J.S.Z., and K.S. performed the experiments with replicable

dengue viruses. T.L. and J.-C.O.-M. designed the software for co-localization

analysis. C.F.Q. produced and purified the monoclonal antibody. M.G., J.-C.

O.-M., R.B., and P.G.W. analyzed the results. R.B. and P.G.W. prepared and

revised the manuscript.

ACKNOWLEDGMENTS

HKU-Pasteur Research Pole is a member of the Institut Pasteur International

Network. We thank V. Malhotra (Center for Genomic Regulation, Barcelona)

for the gift of soluble horseradish peroxidase construct; R. Gijsbers (Katholieke

Universiteit, Leuven) for the retroviral vector; P. Despres (Institut Pasteur,

Paris) for the anti-E antibody 4E11; P. Buchy (Institut Pasteur du Cambodge,

Phnom Penh) and M. Dupont-Rouzeyrol (Institut Pasteur de Nouvelle-Caledo-

nie, Noumea) formouse anti-prME antibody and dengue patients sera; S. Nop-

porn (Chiang Mai University, Thailand) for anti-prM antibody prM-6.1; P.H.M.

Leung (The Hong Kong Polytechnic University, Hong Kong SAR) for the anti-

prME polyclonal antibody; J. Guo and the Faculty Core Facility of the LKS

Faculty of Medicine of the University of Hong Kong for support with confocal

microscopy; P. Chavrier (Institut Curie, Paris), S. Sanyal (HKU-Pasteur

Research Pole), and C. Zurzolo (Institut Pasteur, Paris) for critical comments;

and members of HKU-Pasteur for useful discussion and technical advice.

This work was funded by RFCID (grant #10091312 to P.G.W.), Institut Pasteur

International Network (ACIP A-12-10 to P.G.W. andM.G.), BNP Paribas CIB (to

R.B.), and ANR-10-INBS-04-06 FranceBioImaging and ANR-10-LABX-62-

IBEID (to T.L. and J.-C.O.-M.). M.Y.L. was supported in part by The University

of Hong Kong through a postgraduate scholarship and by the LKS Faculty of

Medicine through a Research Postgraduate Student Exchange Scheme. T.L.

was funded by a Roux fellowship from Institut Pasteur.

Received: April 23, 2014

Revised: October 16, 2014

Accepted: February 4, 2015

Published: March 5, 2015

REFERENCES

Antonny, B., and Schekman, R. (2001). ER export: public transportation by the

COPII coach. Curr. Opin. Cell Biol. 13, 438–443.

Bard, F., Casano, L., Mallabiabarrena, A., Wallace, E., Saito, K., Kitayama, H.,

Guizzunti, G., Hu, Y., Wendler, F., Dasgupta, R., et al. (2006). Functional geno-

mics reveals genes involved in protein secretion and Golgi organization.

Nature 439, 604–607.

Brault, J.B., Kudelko, M., Vidalain, P.O., Tangy, F., Despres, P., and Pardigon,

N. (2011). The interaction of flavivirus M protein with light chain Tctex-1 of

human dynein plays a role in late stages of virus replication. Virology 417,

369–378.

Burleson, P.G., Chambers, T.M., and Wiedbrauk, D.L. (1992). Virology: A Lab-

oratory Manual (San Diego, CA: Academic Press).

Campbell, G.L., Marfin, A.A., Lanciotti, R.S., and Gubler, D.J. (2002). West Nile

virus. Lancet Infect. Dis. 2, 519–529.

Cancino, J., Jung, J.E., and Luini, A. (2013). Regulation of Golgi signaling and

trafficking by the KDEL receptor. Histochem. Cell Biol. 140, 395–405.

Chen, Y., Maguire, T., Hileman, R.E., Fromm, J.R., Esko, J.D., Linhardt, R.J.,

and Marks, R.M. (1997). Dengue virus infectivity depends on envelope protein

binding to target cell heparan sulfate. Nat. Med. 3, 866–871.

Courageot, M.P., Frenkiel, M.P., Dos Santos, C.D., Deubel, V., andDespres, P.

(2000). Alpha-glucosidase inhibitors reduce dengue virus production by

affecting the initial steps of virion morphogenesis in the endoplasmic reticu-

lum. J. Virol. 74, 564–572.

D’Souza-Schorey, C., and Chavrier, P. (2006). ARF proteins: roles in mem-

brane traffic and beyond. Nat. Rev. Mol. Cell Biol. 7, 347–358.

de Chaumont, F., Dallongeville, S., Chenouard, N., Herve, N., Pop, S., Pro-

voost, T., Meas-Yedid, V., Pankajakshan, P., Lecomte, T., Le Montagner, Y.,

et al. (2012). Icy: an open bioimage informatics platform for extended repro-

ducible research. Nat. Methods 9, 690–696.

Duijsings, D., Lanke, K.H.W., van Dooren, S.H.J., van Dommelen, M.M.T.,

Wetzels, R., de Mattia, F., Wessels, E., and van Kuppeveld, F.J.M. (2009). Dif-

ferential membrane association properties and regulation of class I and class II

Arfs. Traffic 10, 316–323.

Giannotta, M., Ruggiero, C., Grossi, M., Cancino, J., Capitani, M., Pulvirenti,

T., Consoli, G.M., Geraci, C., Fanelli, F., Luini, A., and Sallese, M. (2012).

The KDEL receptor couples to Gaq/11 to activate Src kinases and regulate

transport through the Golgi. EMBO J. 31, 2869–2881.

Guzman, M.G., Halstead, S.B., Artsob, H., Buchy, P., Farrar, J., Gubler, D.J.,

Hunsperger, E., Kroeger, A., Margolis, H.S., Martınez, E., et al. (2010). Dengue:

a continuing global threat. Nat. Rev. Microbiol. 8 (12, Suppl), S7–S16.

Hsu, V.W., Shah, N., and Klausner, R.D. (1992). A brefeldin A-like phenotype is

induced by the overexpression of a human ERD-2-like protein, ELP-1. Cell 69,

625–635.

Humphries, A.C., and Way, M. (2013). The non-canonical roles of clathrin and

actin in pathogen internalization, egress and spread. Nat. Rev. Microbiol. 11,

551–560.

Kudelko, M., Brault, J.B., Kwok, K., Li, M.Y., Pardigon, N., Peiris, J.S., Bruz-

zone, R., Despres, P., Nal, B., andWang, P.G. (2012). Class II ADP-ribosylation

factors are required for efficient secretion of dengue viruses. J. Biol. Chem.

287, 767–777.

Kuhn, R.J., Zhang,W., Rossmann, M.G., Pletnev, S.V., Corver, J., Lenches, E.,

Jones, C.T., Mukhopadhyay, S., Chipman, P.R., Strauss, E.G., et al. (2002).

Structure of dengue virus: implications for flavivirus organization, maturation,

and fusion. Cell 108, 717–725.

Lewis, M.J., and Pelham, H.R. (1990). A human homologue of the yeast HDEL

receptor. Nature 348, 162–163.

Lewis, M.J., and Pelham, H.R. (1992a). Ligand-induced redistribution of a hu-

man KDEL receptor from the Golgi complex to the endoplasmic reticulum. Cell

68, 353–364.

Lewis, M.J., and Pelham, H.R. (1992b). Sequence of a second human KDEL

receptor. J. Mol. Biol. 226, 913–916.

Li, L., Lok, S.M., Yu, I.M., Zhang, Y., Kuhn, R.J., Chen, J., and Rossmann, M.G.

(2008). The flavivirus precursor membrane-envelope protein complex: struc-

ture and maturation. Science 319, 1830–1834.

Limjindaporn, T., Wongwiwat, W., Noisakran, S., Srisawat, C., Netsawang, J.,

Puttikhunt, C., Kasinrerk, W., Avirutnan, P., Thiemmeca, S., Sriburi, R., et al.

(2009). Interaction of dengue virus envelope protein with endoplasmic reticu-

lum-resident chaperones facilitates dengue virus production. Biochem. Bio-

phys. Res. Commun. 379, 196–200.

1506 Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors

Lindenbach, B.D., and Rice, C.M. (2001). Flaviviridae: the viruses and their

replication. In Fields virology (Philadelphia, PA: Lippincott Williams & Wilkins),

pp. 991–1041.

Mercer, J., Schelhaas, M., and Helenius, A. (2010). Virus entry by endocytosis.

Annu. Rev. Biochem. 79, 803–833.

Messina, J.P., Brady, O.J., Scott, T.W., Zou, C., Pigott, D.M., Duda, K.A.,

Bhatt, S., Katzelnick, L., Howes, R.E., Battle, K.E., et al. (2014). Global spread

of dengue virus types: mapping the 70 year history. Trends Microbiol. 22,

138–146.

Modis, Y. (2013). Class II fusion proteins. Adv. Exp. Med. Biol. 790, 150–166.

Mukhopadhyay, S., Kuhn, R.J., and Rossmann, M.G. (2005). A structural

perspective of the flavivirus life cycle. Nat. Rev. Microbiol. 3, 13–22.

Pryor, M.J., Azzola, L., Wright, P.J., and Davidson, A.D. (2004). Histidine 39 in

the dengue virus type 2 M protein has an important role in virus assembly.

J. Gen. Virol. 85, 3627–3636.

Pulvirenti, T., Giannotta, M., Capestrano, M., Capitani, M., Pisanu, A., Polish-

chuk, R.S., San Pietro, E., Beznoussenko, G.V., Mironov, A.A., Turacchio, G.,

et al. (2008). A traffic-activated Golgi-based signalling circuit coordinates the

secretory pathway. Nat. Cell Biol. 10, 912–922.

Raykhel, I., Alanen, H., Salo, K., Jurvansuu, J., Nguyen, V.D., Latva-Ranta, M.,

and Ruddock, L. (2007). A molecular specificity code for the three mammalian

KDEL receptors. J. Cell Biol. 179, 1193–1204.

Rothman, J.E., and Orci, L. (1992). Molecular dissection of the secretory

pathway. Nature 355, 409–415.

Schindler, R., Itin, C., Zerial, M., Lottspeich, F., and Hauri, H.P. (1993). ERGIC-

53, a membrane protein of the ER-Golgi intermediate compartment, carries an

ER retention motif. Eur. J. Cell Biol. 61, 1–9.

Sessions, O.M., Barrows, N.J., Souza-Neto, J.A., Robinson, T.J., Hershey,

C.L., Rodgers, M.A., Ramirez, J.L., Dimopoulos, G., Yang, P.L., Pearson,

J.L., and Garcia-Blanco, M.A. (2009). Discovery of insect and human dengue

virus host factors. Nature 458, 1047–1050.

Sun, E., He, J., and Zhuang, X. (2013). Live cell imaging of viral entry. Curr.

Opin. Virol. 3, 34–43.

Volpicelli-Daley, L.A., Li, Y., Zhang, C.J., and Kahn, R.A. (2005). Isoform-selec-

tive effects of the depletion of ADP-ribosylation factors 1-5 on membrane

traffic. Mol. Biol. Cell 16, 4495–4508.

Wan, S.W., Lin, C.F., Lu, Y.T., Lei, H.Y., Anderson, R., and Lin, Y.S. (2012).

Endothelial cell surface expression of protein disulfide isomerase activates

b1 and b3 integrins and facilitates dengue virus infection. J. Cell. Biochem.

113, 1681–1691.

Wang, P.G., Kudelko, M., Lo, J., Siu, L.Y., Kwok, K.T., Sachse, M., Nicholls,

J.M., Bruzzone, R., Altmeyer, R.M., and Nal, B. (2009). Efficient assembly

and secretion of recombinant subviral particles of the four dengue serotypes

using native prM and E proteins. PLoS ONE 4, e8325.

Welsch, S., Miller, S., Romero-Brey, I., Merz, A., Bleck, C.K., Walther, P., Ful-

ler, S.D., Antony, C., Krijnse-Locker, J., and Bartenschlager, R. (2009).

Composition and three-dimensional architecture of the dengue virus replica-

tion and assembly sites. Cell Host Microbe 5, 365–375.

Wilson, D.W., Lewis, M.J., and Pelham, H.R. (1993). pH-dependent binding of

KDEL to its receptor in vitro. J. Biol. Chem. 268, 7465–7468.

Xu, Z., Bruss, V., and Yen, T.S. (1997). Formation of intracellular particles by

hepatitis B virus large surface protein. J. Virol. 71, 5487–5494.

Yu, I.M., Zhang, W., Holdaway, H.A., Li, L., Kostyuchenko, V.A., Chipman,

P.R., Kuhn, R.J., Rossmann, M.G., and Chen, J. (2008). Structure of the imma-

ture dengue virus at low pH primes proteolytic maturation. Science 319, 1834–

1837.

Zhang, Y., Corver, J., Chipman, P.R., Zhang, W., Pletnev, S.V., Sedlak, D.,

Baker, T.S., Strauss, J.H., Kuhn, R.J., and Rossmann, M.G. (2003). Structures

of immature flavivirus particles. EMBO J. 22, 2604–2613.

Zhang, L., Lee, S.Y., Beznoussenko, G.V., Peters, P.J., Yang, J.S., Gilbert,

H.Y., Brass, A.L., Elledge, S.J., Isaacs, S.N., Moss, B., et al. (2009). A role

for the host coatomer and KDEL receptor in early vaccinia biogenesis. Proc.

Natl. Acad. Sci. USA 106, 163–168.

Cell Reports 10, 1496–1507, March 10, 2015 ª2015 The Authors 1507

Cell Reports Supplemental Information

KDEL Receptors Assist Dengue Virus Exit from the Endoplasmic Reticulum Ming Yuan Li, Marc Grandadam, Kevin Kwok, Thibault Lagache, Yu Lam Siu, Jing Shu Zhang, Kouxiong Sayteng, Mateusz Kudelko, Cheng Feng Qin, Jean-Christophe Olivo-Marin, Roberto Bruzzone, and Pei Gang Wang

1

Supplemental Figures and Legends 1

 2

 3

Figure  S1.  KDELR  interact  with  DENV1  prM  glycoprotein,  Related  to  Figure  1.  A) 4

KDELR1  and DENV1‐E protein were partially  co‐localised  in HeLa  cells  co‐transfected 5

with prME and KDELR1‐RFP (red). Cells were stained with anti‐E 4G2 antibody (green) 6

and nuclei were visualised with DAPI (blue).  Images were acquired with a Zeiss Axio 7

Observer Z1 inverted microscope; bar: 10 µm. B) Co‐immunoprecipitation of prM with 8

KDELR1‐myc.  Lysates  from  293T  cells  transfected with  empty  vector  (ev),  prME  or 9

prME and KDELR1‐myc were subjected to  immunoprecipitation using rabbit either an 10

anti‐c‐myc antibody or normal rabbit IgG. Cell lysates (CL) or pellets were analysed by 11

Western blotting using anti‐myc or anti‐prME antibodies. prM/KDELR  interaction was 12

observed also when KDELR1 was used as the bait (data not shown). These results are 13

representative of two independent experiments. 14

2

 15

Figure S2. KDELR knockdown does not affect Hela‐prME‐DENV1 cell morphology and 16

viability, Related to Figure 2. A) siRNA‐treated Hela‐prME‐DENV1 cells (non‐targeting, 17

NT and KDELR) were re‐suspended  in PBS and stained by propidium  iodide (PI). Cells 18

subjected to osmotic  lysis  in ddH2O at 37˚C for 3h were used as the positive control. 19

Cell viability was analysed by flow cytometry and mean fluorescence intensity (MFI) of 20

siRNA‐treated cells was normalized to that of controls. Results are means ± SD of four 21

measurements  from  two  independent  experiments.  No  significant  difference  was 22

found  between NT  and  KDELR  siRNA  treated  cells. B) Hela‐prME‐DENV1  cells were 23

transfected with  siRNAs  targeting  all  KDELR  isoforms.  72h  post  transfection,  cells 24

were stained with anti‐KDELR monoclonal antibodies （red）to verify the knockdown 25

efficiency. NT siRNA was used as control. Nuclei were visualised with DAPI (blue). Cell 26

morphology was observed under bright  field  (BF) and  images were acquired with a 27

Carl Zeiss LSM 700 confocal microscope; bar:  10 µm. C) HeLa cells stably expressing 28

3

eGFP‐tagged  KDELR1‐3  were  transfected  with  either  NT  (open  symbols)  or  the 29

indicated siRNAs specific for  individual receptor  isoforms (grey‐tinted symbols). Cells 30

were fixed 72h post‐transfection and analysed by flow cytometry. MFI of cells treated 31

with  siRNAs  for  KDELR was  normalized  to  that  of NT  controls  (right‐hand  panels). 32

Results  are means ± SD of nine observations  from  three  independent experiments. 33

*P<0.005 vs lysed control cells (B) or NT (C) by the unpaired Student’s t‐test. 34

4

 35

Figure S3. KDELR knockdown does not affect early steps of DENV life cycle in Vero E6 36

cells, Related to Figure 3. A) Cells were transfected with siRNAs targeting KDELR or 37

Arf4+5, either alone or  in combination (K+A), before being challenged with DENV1 at 38

an MOI=0.1. Cellular  lysates were collected at 5 days post  infection and subjected to 39

Western blotting using relevant antibodies to verify the effect of protein knockdown 40

by  the  specified  siRNAs.  GAPDH was  used  as  the  loading  control  across wells.  B) 41

Morphology of Vero E6 after silencing KDELR by was similar to that of NT‐treated cells. 42

Cells were  stained with  anti‐KDELRs monoclonal  antibodies  to  confirm  knockdown 43

efficiency and  images were acquired with a Carl Zeiss LSM 700 confocal microscope; 44

bar:  10 µm. C) KDELR down  regulation did not  influence  cell  viability.  Experimental 45

conditions and  staining with propidium  iodide  (PI) were as described  in Fig. S2. Cell 46

5

viability  was  analysed  by  flow  cytometry  and  MFI  of  siRNA‐treated  cells  was 47

normalized  to  that of controls. Results are means ± SD of  four measurements  from 48

two  independent  experiments.  No  significant  difference  was  found  between  cells 49

treated  with  either  NT  or  KDELR  siRNA.  D)  siRNA‐treated  Vero  E6  cells  were 50

challenged with DENV1 at an MOI=1. Viral titres, measured by real‐time RT‐PCR at 18h 51

post  infection  (p.i.) were similar  to  those of control cells  treated with non‐targeting 52

(NT) siRNA. Results are means ± SD of three  independent experiments. *P<0.005 vs 53

control cells by the unpaired Student’s t‐test. 54

6

 55

Figure  S4.  The  R6S  point mutation  disrupts  the  interaction  between  prM  and  E 56

proteins, Related to Figure 5. A) Sequence alignment of the N‐terminal 40 amino acids 57

of DENV1‐4 prM.    Positively charged residues are showed in red. The three key amino 58

acids  for  prM/KDELRs  interaction  are  indicated  by  asterisks.  B)  Lysates  from  cells 59

transfected  with  either  wild‐type  or mutant  prME  (see  Fig.  4)  were  subjected  to 60

immunoprecipitation using the anti‐E monoclonal antibody 4E11. prM and E proteins in 61

cell  lysates  (CL)  or  pellets  (IP) were  then  analysed  by Western  blotting  using  the 62

indicated antibodies. These experiments showed that a drastic reduction of the prM/E 63

in the IP was observed only for the R6S mutation. These results are representative of 64

two independent experiments. 65

7

 66

Figure S5. A triple prME mutant (Triple) does not affect the interaction between prM 67

and Arf4/Arf5, Related to Figure 5. Lysates from cells transfected with either empty 68

vector  (ev), wild‐type  (WT), or Triple  (see also Experimental Procedures and Fig. 5) 69

were  subjected  to  immunoprecipitation  using dengue patient  serum  (DPS). Normal 70

human serum (NHS) was used the negative control. Arf4, Arf5 and prM proteins in cell 71

lysates (CL) or pellets (IP) were then analysed by Western blotting using the indicated 72

antibodies.  Similar  amounts  of  Arf4  and  Arf5  could  be  precipitated  from  cells 73

expressing either WT or the Triple prME mutant. These results are representative of 74

two independent experiments. 75

8

 76

Figure  S6.  Aggregates  of  E  protein  are  present  in  DENV1  infected  Vero  E6  cells, 77

Related to Figure 6. Cells were  infected at an MOI=0.1 and stained 5 days  later with 78

either  a  polyclonal  antibody  against  both  prM  and  E  (left  panel),  or  a monoclonal 79

antibody  recognizing  only  E  (right  panel).  Arrowheads  indicate  aggregates  of  E 80

protein, similarly to what observed in HeLa‐prME‐DENV1 cells (see Fig. 6B, WT panels). 81

Images were acquired with a Carl Zeiss LSM 700 confocal microscope; bar: 10 µm. 82

9

Supplemental Tables 83

Table S1. Primers used to identify KDELR by RT‐PCR, Related to Figure 1. 84

KDELR1‐f  5’TGTGGTGTTCACTGCCCGA3’ 

KDELR1‐r  5’CGGTAAACGCCTAGCGCAA3’ 

KDELR2‐f  5’AACATTTTCCGGCTGACT3’ 

KDELR2‐r  5’GACAAGATACAAAGCACGA3’ 

KDELR3‐f  5’AAGTGCTGCAAGGGCATCT3’ 

KDELR3‐r  5’CCAATGACTGGGACCAGAA3’ 

10

Table S2. Primers used to produce prM fragments used as baits for GST pull‐downs, 85

Related to Figure 1. 86

prM‐GST‐f  5’CGCGGATCCTTCCACCTGACCACCAGAG3’ 

pr‐GST‐r  5’CCGCTCGAGTCATCATCTCTTGTCTCTCCTGTGCT3’ 

prM‐ΔTM‐GST‐r  5’CCG CTCGAG TCA TCA CAG GGC CCA GGT CTC CAC CT3’ 

pr40‐GST‐r  5’CCG CTCGAG TCA TCA ATC CAT GGC GAT CAG GGT AC3’ 

11

Table S3. Primers used in overlapping PCR for site‐directed mutagenesis, Related to 87

Figure 5. 88

P1  5’CCATTGACGTCAATGGGAGT3’ 

P2‐H2L  5’GGTCAGGAGGAAGCAGTTCA3’ 

P3‐H2L  5’TGAACTGCTTCCTCCTGACC3’ 

P2‐R6S  5’TCGCCGCCGCTGGTGGTCA3’ 

P3‐R6S  5’TGACCACCAGCGGCGGCGA3’ 

P2‐H11L  5’ACACGATCATGAGGGGCTC3’ 

P3‐H11L  5’GAGCCCCTCATGATCGTGT3’ 

P2‐K16T  5’TCCTGCGTGGACACGATCA3’ 

P3‐K16T  5’TGATCGTGTCCACGCAGGA3’ 

P2‐R19S  5’AGGCTCTTGCCGCTCTCCT 3’ 

P3‐R19S  5’AGGAGAGCGGCAAGAGCCT3’ 

P2‐K21T  5’TGAACAGCAGGCTCGTGCC3’ 

P3‐K21T  5’GGCACGAGCCTGCTGTTCA3’ 

P2‐K26T  5’GCGCTGGTCGTGAACAGCA 3’ 

P3‐K26T  5’TGCTGTTCACGACCAGCGC 3’ 

P4  5’TCTAGACTCGAGCTAGCTAG3’ 

Bold‐faced letters designate the mutated base. 89

12

Supplemental Experimental Procedures, Related to Experimental Procedures. 90

Cells, Viruses and Antibodies 91

HeLa  and  human  embryonic  kidney  (293T)  cells  were  maintained  in  DMEM 92

supplemented with  10%  foetal bovine  serum  (FBS)  and  1% penicillin/streptomycin  at 93

37°C, with 5% CO2. The  stable cell  lines expressing prME‐DENV1 or  its mutants were 94

established using the retroviral vector pCHMWS‐IRES‐Hygromycin (kindly provided by 95

Dr.  Rik  Gijsbers,  Division  of  Molecular  Medicine,  Katholieke  Universiteit,  Leuven, 96

Belgium),  selected  following  a  2‐week  period  in  the  presence  of  500  µg/mL 97

hygromycin  and maintained  thereafter  the  same medium,  as  previously  described 98

(Wang et  al.,  2009). Vero E6  cells, which were derived  from African green monkey 99

kidney, were maintained  in DMEM supplemented with 5%  foetal calf serum  (FCS) at 100

37°C, with 5% CO2. All work with infectious DENV1 strain (DENV1 Hawaii), DENV2 strain 101

(Dengue 2 New Guinea), DENV3 strain (Dengue 3 strain H87), DENV 4 strain (Dengue 4 102

Jamaique 8343)  and WNV  strain  (West Nile B956,  lineage  II）  was performed  in  a 103

biosafety  level 2 plus  laboratory  (Institut Pasteur du Laos, Vientiane, Lao PDR) with 104

Vero  E6  cells.  For  biochemistry,  the  following  antibodies were  used: mouse  anti‐E 105

monoclonal  antibodies  (mAb)  4E11  (a  gift  of Dr.  Philippe Despres,  Institut  Pasteur, 106

Paris, France); mouse anti‐E mAb 4G2 was prepared using hybridoma cells D1‐4G2‐4‐15 107

from ATCC (Manassas, VA, USA); mouse anti‐prME antibody and dengue patients sera 108

(kindly  provided  by  Dr.  Philippe  Buchy,  Institut  Pasteur,  Cambodia  and  Myrielle 109

Dupont‐Rouzeyrol,  Institut Pasteur de Nouvelle‐Calédonie, Nouméa, New‐Caledonia) 110

13

and mouse anti‐prM mAb prM‐6.1  (a gift  from Dr. Sittisombut Nopporn, Chiang Mai 111

University,  Chiang Mai,  Thailand)  had  been  previously  used  (Junjhon  et  al.,  2008; 112

Kudelko et al., 2012); mouse anti‐KDELR mAb, rabbit anti‐GM130 mAb, rabbit anti‐Myc 113

tag  antibody  and  anti‐GAPDH  mAb  from  Abcam  (Cambridge,  MA,  USA);  rabbit 114

anti‐Arf4 from ProteinTech Group Inc (Chicago, IL, USA); mouse anti‐Arf5 mAb, rabbit 115

anti‐Bip  mAb  and  rabbit  anti‐calreticulin  antibody  from  Cell  Signaling  Technology 116

(Beverly, MA,  USA).  For  immunofluorescence,  antibodies  were  as  follows: mouse 117

anti‐E mAb 4G2, rabbit anti‐prME polyclonal antibody (kindly provided by Dr. Polly H.M. 118

Leung, The Hong Kong Polytechnic University, Hong Kong SAR), rabbit anti‐calnexin 119

mAb from Cell Signaling Technology, rabbit anti‐ERGIC53 from Sigma (Saint Louis, MO, 120

USA),  rabbit  anti‐GM130  mAb  from  Abcam,  rabbit  anti‐TGN46  from  Abnova 121

Corporation (Walnut, CA, USA). 122

 123

siRNA experiments 124

All  siRNAs  used  in  this  work,  including  non‐targeting  (NT)  siRNA  (D‐001206)  and 125

transfection  reagents  DharmaFECT  1  (T‐2001)  were  purchased  from  Dharmacon 126

Research  Inc.  (Lafayette, CO, USA). Arf4 siRNA  (L‐011582) and Arf5 siRNA  (L‐011584) 127

were provided as SMARTpool ON‐TARGET plus siRNAs, which are pools of four siRNAs 128

targeting various sites in a single gene. KDELR1 siRNA (J‐019136‐09/11), KDELR2 siRNA 129

(J‐0.12315‐05/07)  and  KDELR3  siRNA  (J‐012316‐05/06)  were  provided  as  individual 130

ON‐TARGET plus siRNAs targeting one site in a single gene. Briefly, siRNAs mixed with 131

14

DharmaFECT 1 reagents (Dharmacon) were added to 24‐well plates in DMEM medium 132

without  FBS  and  antibiotics.  Twenty minutes  later, 0.8 mL  cells  (75,000  cells/mL  in 133

DMEM supplemented with 2.5% FCS) were added to each well so that the final siRNA 134

concentration was 100nM for Arf4+5, and 150nM for both KDELR and NT. Cells were 135

incubated at 37°C for 48 hours. For RSPs experiments, medium was replaced with 0.3 136

mL of DMEM containing 2% FBS (without antibiotics) and, 14 hours later, supernatant 137

containing secreted RSPs was collected, cleared by centrifugation at 4,000 rpm for 15 138

min and analysed by Western blotting  (Wang et al., 2009). Cells were  lysed  in RIPA 139

buffer containing: 1% Triton X‐100, 150 mM NaCl, 50 mM Tris‐HCl, (pH 7.5), 1 mM EDTA, 140

0.5%  Na‐deoxycholate,  freshly  added  1  mM  PMSF  and  protease  inhibitors  cocktail 141

(Roche  Applied  Science,  Mannheim,  Germany),  for  15  min  on  ice,  with  frequent 142

vortexing.  For  dengue  virus  1  (DENV1)  infection  assay,  siRNA  treatment  was 143

performed 48 hours before viral infection using the same conditions described above. 144

Supernatant  containing  secreted  progeny  virus  was  collected,  cleared  by 145

centrifugation at 4000 rpm for 15 min and further processed to measured viral titre by 146

Real Time RT‐PCR. Cells were lysed in RIPA buffer as detailed above. 147

 148

Production, purification and quantification of GST‐fusion proteins 149

A  single  colony  of  BL21  bacteria  transformed  with  pGEX‐4T‐1  expression  vector 150

encoding  GST‐fused  prM  fragments  or  GST  protein  alone was  lifted  from  LB  agar 151

dishes containing 100µg/ml ampicillin and grown overnight at 37°C  in 5ml LB medium 152

15

supplemented with  the  same ampicillin  concentration. To  induce expression of GST 153

fusion  proteins,  isopropyl‐1thio‐b‐D‐galactopyanoside  (IPTG)  was  added  at  a  final 154

concentration of 1mM after cooling down on ice for 5min a 500 ml bacterial culture in 155

mid‐log phase (OD600 ~0.6‐0.7). After shaking for 2h at 25°C, bacteria were harvested 156

by centrifugation at 4,000 g for 10min at 4°C and re‐suspended in 10ml of ice‐cold lysis 157

buffer  I  [0.4M NaCL, 50 mM Tris‐HCL pH 7.5, 0.3%  (v/v) Triton X‐100], supplemented 158

with freshly added 2% (v/v) N‐lauroylsarcosine, 100 μg/ml lysozyme, 0.6 mM PMSF (all 159

from Sigma‐Aldrich, St. Louis, MO) and 1x protease inhibitors cocktail (Roche Applied). 160

The bacterial cell suspension was transferred to a 50ml Falcon tube, mixed well with 161

2% N‐lauroylsarcosine,  and  then homogenized on  ice by  sonication  (9x10sec pulses) 162

using  a  cell  disruptor  (Vibra  CellTM,  Sonics  and  Materials,  Inc.,  CT,  USA)  at  35% 163

amplitude. Cell debris were spun down at 12,000 g for 1 hr at 4°C. To purify GST fusion 164

proteins,  clarified  supernatants were  transferred  to  a  clean  Falcon  tube  containing 165

200μl  of  a  50%  slurry  of  pre‐washed  sepharose  glutathione  4B  beads  (Amersham 166

Biosciences, MA, USA). Following an overnight  incubation at 4°C on a rotating plate, 167

beads were  extensively washed with  800μl  of  ice‐cold  lysis  buffer  II  (0.2 M  NaCl, 168

50mM Tris‐HCl pH 7.5, 0.15% Triton X‐100,), freshly supplemented with 0.6mM PMSF, 169

1X Complete protease cocktail  inhibitor and 1% N‐lauroylsarcosine. Supernatants were 170

discarded  and  the  beads were  eventually mixed with  500μl  of  lysis  buffer  II.  Five 171

microlitres of  this mixture were  loaded on a  12% SDS‐PAGE,  stained with Coomassie 172

brilliant blue (G250; Sigma‐Aldrich) and quantified by densitometry using Image Quant 173

16

TL (Thermo Fisher Scientific Inc., Rockford, IL, USA) against a standard curve of bovine 174

serum albumin samples (2‐fold dilutions ranging from 54‐1,448ng/μl) ran in parallel. 175

 176

Mutagenesis 177

Site‐directed mutagenesis was performed by overlapping PCR using codon optimized 178

DENV1 prME as  the  template. Seven positive charged amino acids  in  the N‐terminal 179

portion of prM were substituted with neutral amino acid individually or in combination 180

(Table  S3).  prME  mutants  were  digested  with  BamHI/XhoI  and  subcloned  into 181

pcDNA3.1 (Invitrogen) or pCHMWS‐IRES‐Hygromycin vectors for further studies. 182

 183

Flow Cytometry 184

Stable HeLa cell  lines expressing eGFP tagged KDELR1, 2 or 3, were transfected with 185

siRNAs  targeting  individual  isoforms as  indicated. Cells were  re‐suspended and  fixed 186

72h  post‐transfection.  For  cell  viability  assays,  siRNA‐treated  HeLa‐prME‐DENV1  or 187

Vero E6 cells were re‐suspended in PBS containing 10µg/ml propidium iodide (PI) and 188

incubated on ice for 30min. Cells re‐suspended in ddH2O (3h at 37˚C before PI staining 189

as  above)  to  induce  osmotic  lysis  served  as  the  positive  controls.  For  both  sets  of 190

experiments,  after washing with  cold PBS,  cell  suspensions were  subjected  to  flow 191

cytometry using a FACSCalibur  (BD Biosciences, San  Jose, CA, USA), and more  than 192

20,000  singlet  living  cells  were  collected.  The  post‐acquisition  data  analysis  was 193

performed using FlowJo software (TreeStar, Ashland, OR, USA). 194

17

 195

RT‐PCR 196

Total RNA from HeLa, HeLa‐prME‐DENV1, 293T and VeroE6 cells was extracted using 197

RNeasy Mini  kit  (QIAGEN,  Valencia,  CA, USA)  and  eluted  in  50µl RNase‐free water, 198

according  to  the manufacturer’s  instructions.  cDNA  synthesis was  performed with 199

SuperScript  III First‐Strand Synthesis System  for RT‐PCR kit  (Invitrogen) using  1µg of 200

total RNA  in one  reaction mixture of 20µl  that  contained dNTP and Oligo  (dT)20, as 201

recommended.  Specific  individual  primers  (Table  S1)  designed  for  the  three  KDELR 202

isoforms  and  GAPDH were  then  used  in  the  following  amplification  reactions with 203

FastStart  Taq  DNA  Polymerase,  dNTPack  (Roche  Applied  Science).  PCR  cycling 204

conditions were as  follows: 95°C  for 4 min,  followed by 35 cycles of 95°C  for 30 sec, 205

54°C  for  30s  and  72°C  for  1  min.  PCR  products  were  analysed  by  agarose  gel 206

electrophoresis and detected by ethidium bromide staining. 207

 208

Real‐time RT‐PCR (TaqMan) 209

For measurements of progeny virus production, viral RNA was extracted from culture 210

supernatants  (150µl)  using  Nucleospin@  RNA  virus  kit  (MACHEREY‐NAGEL,  Düren, 211

Germany), following the instructions provided by the manufacturer, and eluted in 50µl 212

RNase‐free water.  The  same procedure was used  for  viral  replication  assay,  except 213

that 20µl of RNase‐Free proteinase K (20mg/ml stock solution; Roche Applied Science) 214

were added to the lysis mixture and DNA was removed by using RNase‐Free DNase kit 215

18

(QIAGEN, Chatsworth, CA, USA) in the washing steps, as recommended. The sequence 216

of DENV universal primers to detect all four serotypes, of West Nile Virus primers and 217

their specific fluorogenic probe (FAM/BHQ‐1) have been previously reported (Linke et 218

al., 2007; Warrilow et al., 2002). Briefly, for each reaction 5µl of RNA were combined 219

with 2µl of each primer (10µM) and 1µl of probe (10µM)  in a total mixture volume of 220

25µl by using  EXPRESS One‐Step  SuperScript qRT‐PCR Kits  (Invitrogen).  The  cycling 221

conditions used were: 45°C for 30 min and 95°C for 2 min, followed by 40 cycles as 95°C 222

for 15 sec and 60°C for 30 sec. Real‐time PCR was performed with the MiniOpticon™ 223

Real‐time PCR Detection system  (Bio‐Rad, CA, USA) and products were detected by 224

measuring the fluorescence signal from the FAM reporter. 225

 226

Quantitative analysis of fluorescent images 227

In protocol “Stained cells”, we first extracted signals of nuclei, stained by DAPI, and of 228

viral  E  protein,  stained  by  the  anti‐E  monoclonal  antibody  4G2,  which  were 229

significantly brighter than the cell background with a K‐mean algorithm (Forgy, 1965). 230

The Regions of  Interest  (ROIs) defined by  the extracted  signals were  then  filled  to 231

form  connected  regions,  and  the  boolean  intersection  between  nuclei  ROIs  and  E 232

protein ROIs was  computed  to extract  the nuclei  that  are  surrounded by E  signals. 233

Finally, by counting surrounded nuclei we obtained the number of stained cells in each 234

image. In “Cells with aggregates” protocol, we first extracted signal corresponding to 235

cell nuclei with a K‐mean algorithm and filled holes in nuclei ROIs to obtain connected 236

19

regions  as  described  previously.  In  parallel,  we  used  a  wavelet‐based  detection 237

method (Olivo‐Marin, 2002),  implemented as a plugin named “Spot detector”  in the 238

ICY  software  (scale=5,  threshold=100),  to  extract  the  large  spots  corresponding  to 239

RSPs aggregates, stained with  the anti‐E antibody, which were significantly brighter 240

than  the surrounding background. We  filtered  these detections with a size criterion 241

(spots>100  pixels),  and  a  compactness  criterion  (4π  Area/Perimeter2  >0.15;  Perfect 242

disks have a compactness equal to 1, whereas the compactness of filamentous spots 243

tends  to  0)  to  obtain  round  and  large  aggregate  spots. We  then  defined  nuclear 244

neighbourhoods by dilating nuclear ROIs by 15 pixels and filtered aggregate spots that 245

were inside these dilated nuclei ROIs to select aggregates spots that were sufficiently 246

close  to  cell  nuclei.  Finally,  by  counting  the  nuclei with  neighbourhood‐containing 247

aggregates, we obtained the number of cells with RSPs aggregates. 248

 249

Supplemental References 250

Forgy,  E.W.  (1965).  Cluster  analysis  of  multivariate  data:  efficiency  versus 251 interpretability of classifications. Biometrics 21, 768‐769. 252

Junjhon,  J., Lausumpao, M., Supasa, S., Noisakran, S., Songjaeng, A., Saraithong, P., 253 Chaichoun,  K.,  Utaipat,  U.,  Keelapang,  P.,  Kanjanahaluethai,  A.,  et  al.  (2008). 254 Differential modulation of prM cleavage, extracellular particle distribution, and virus 255 infectivity by conserved residues at nonfurin consensus positions of the dengue virus 256 pr‐M junction. J. Virol. 82, 10776‐10791. 257

Kudelko, M., Brault,  J.B.,  Kwok,  K.,  Li, M.Y.,  Pardigon, N.,  Peiris,  J.S., Bruzzone, R., 258 Despres,  P.,  Nal,  B.,  and  Wang,  P.G.  (2012).  Class  II  ADP‐ribosylation  Factors  Are 259 Required for Efficient Secretion of Dengue Viruses. J. Biol. Chem. 287, 767‐777. 260

Linke, S., Ellerbrok, H., Niedrig, M., Nitsche, A., and Pauli, G. (2007). Detection of West 261 Nile virus lineages 1 and 2 by real‐time PCR. J.Virol. Methods 146, 355‐358. 262

20

Olivo‐Marin,  J.C.  (2002).  Extraction  of  spots  in  biological  images  using  multiscale 263 products. Pattern. Recogn. 35, 1989‐1996. 264

Wang,  P.G.,  Kudelko,  M.,  Lo,  J.,  Siu,  L.Y.,  Kwok,  K.T.,  Sachse,  M.,  Nicholls,  J.M., 265 Bruzzone, R., Altmeyer, R.M., and Nal, B. (2009). Efficient assembly and secretion of 266 recombinant subviral particles of  the  four dengue serotypes using native prM and E 267 proteins. PLoS One 4, e8325. 268

Warrilow,  D.,  Northill,  J.A.,  Pyke,  A.,  and  Smith,  G.A.  (2002).  Single  rapid  TaqMan 269 fluorogenic  probe  based  PCR  assay  that  detects  all  four  dengue  serotypes.  J.Med. 270 Virol. 66, 524‐528. 271