Role of PDZ-binding motif from West Nile virus NS5 protein on viral replication

West Nile virus (WNV) is a Flavivirus, which can cause febrile illness in humans that may progress to encephalitis. Like any other obligate intracellular pathogens, Flaviviruses hijack cellular protein functions as a strategy for sustaining their life cycle. Many cellular proteins display globular domain known as PDZ domain that interacts with PDZ-Binding Motifs (PBM) identified in many viral proteins. Thus, cellular PDZ-containing proteins are common targets during viral infection. The non-structural protein 5 (NS5) from WNV provides both RNA cap methyltransferase and RNA polymerase activities and is involved in viral replication but its interactions with host proteins remain poorly known. In this study, we demonstrate that the C-terminal PBM of WNV NS5 recognizes several human PDZ-containing proteins using both in vitro and in cellulo high-throughput methods. Furthermore, we constructed and assayed in cell culture WNV replicons where the PBM within NS5 was mutated. Our results demonstrate that the PBM of WNV NS5 is important in WNV replication. Moreover, we show that knockdown of the PDZ-containing proteins TJP1, PARD3, ARHGAP21 or SHANK2 results in the decrease of WNV replication in cells. Altogether, our data reveal that interactions between the PBM of NS5 and PDZ-containing proteins affect West Nile virus replication.


Results
The PBM within NS5 impacts West Nile virus replication. The NS5 protein of some Flaviviruses are known to contain an internal and/or C-terminus PBM. The WNV NS5 protein contains a −TVL COOH motif at the C-terminal extremity, corresponding to a PBM of class I with the motif S/T-X-Φ COOH where X is any amino-acid, and Φ is a hydrophobic amino-acid (Fig. 1A). As NS5 of Flaviviruses provides RNA cap MTase and RdRp activities, we focused our efforts to determine whether the PBM of WNV NS5 affects virus replication. To evaluate the contribution of the PBM to WNV replication, we constructed a DNA based sub-genomic WNV replicon expressing the 5′ and 3′ UTR, the genomic region encoding all the non-structural proteins of WNV as well as the reporter gene GLuc, thus generating a luciferase expressing replicon (Rep-IS98-Gluc-wt) (Fig. 1A). We also constructed a negative control replicon coding for an inactive RdRp with a single mutation in the replication "pocket" on WNV NS5 (Rep-IS98-Gluc-GVD) and a replicon with a deletion of the three residues -TVL identified as the PBM motif (Rep-IS98-Gluc-∆PBM) (Fig. 1A). The replicons were transfected into BHK-21 cells and their replication rates were monitored using luciferase luminescence at different time points over 72 h post-transfection (Fig. 1B). The deletion of the nucleotidic sequence coding for the PBM (∆PBM) resulted in a drastic effect with an absence of replication of the WNV replicon (Fig. 1B). In order to confirm the role of the PBM sequence in the replication of WNV, a mutagenized PBM in the replicon was also designed (Rep-IS98-Gluc-TVM: mPBM). We introduced suitable mutations into the replicon taking care not to alter the RNA structure. Indeed, the genome of WNV is an RNA molecule displaying a single ORF flanked by untranslated regions (UTRs). These UTRs are structural conserved parts in the viral genome and play important roles in viral translation and replication. Optimal viral RNA structure at the 3′ UTR is required for efficient polymerase activity during WNV replication leading to viral production in cultured cells 13 . Introducing a mutation in the RNA sequence encoding a PBM located at the hinge between translated and untranslated regions has proved challenging since cis-acting RNA structures in the 3′-genome must be considered in WNV replication in cells in addition to RdRp activity. We did identify amino-acids and nucleotides required to maintain replication competence and replication efficiency as some PBM mutations in replicons abrogated replication, due to modifications of secondary structures of the 3′UTR. A functional replicon that conserved a proper predicted RNA structure while altering the PBM was thereby obtained using the mutation Leu to Met at the C-terminus of NS5. Thus, we introduced a mutation in the PBM of NS5 that alters the PBM motif and hence potentially affecting the affinities and the pattern of cellular PDZ-containing proteins that could be recognized by the NS5 protein during the infection. Interestingly, the substitution of -TVL COOH by -TVM COOH (mPBM) resulted in a significant delay in replication and also in a significantly lower replication with a ten-fold decrease compared to the wt at 48 and 72 h (Fig. 1B) and a five-fold decrease of band intensity for mPBM NS5 compared to wt NS5 as monitored by western blot experiment (Fig. 1C). Collectively, these results show that WNV replication is affected by an amino-acid substitution that alters the PBM within NS5.
Then, we have attempted to look at the knock-down on viral replication of wt WNV and mPBM WNV. We observed viral replication quantified by RTqCR (Suppl. Fig. 2A) and by FFU (Suppl. Fig. 2B). As we did not see any difference for both quantitation, we wondered if any reversions were occurring, permitting this replication. Indeed, after viral amplification, the last residue of the mPBM (Met, ATG) virus had reverted to a Leu (TTG) as the wild-type virus which has a Leu (CTG) (Suppl. Fig. 2C,D). These observations suggest that the Leu in the PBM motif if important for virus amplification.
West Nile virus NS5 recognizes several cellular PDZ-containing proteins. In order to quantify the binding activity of the NS5 PBM we used an in vitro automated high-throughput chromatography assay called holdup ( Fig. 2A) 14 . A 13-mer peptide was synthetized encompassing the C-terminal PBM sequence of WNV NS5 protein linked to a biotinyl group. This peptide was used as bait to quantify the interaction between the WNV NS5 PBM peptide and a full human PDZ domain library expressed in Escherichia coli. The updated PDZ library called PDZome V.2 covers 97.3% of the human "PDZome" (259 over 266 PDZ domains identified in the human genome) 13 . The holdup approach displays a high sensitivity for low-to-medium affinity PDZ/PBM  ARHGAP21, HTRA1, PARD3_3, PDZD7_3, SCRIB_1, ARHGEF12,  DLG2_1, DLG4_3, TJP3_2, DLG1_2, SNTB1, SYNJ2BP and NHERF2_1) ( Fig. 2A).
In parallel, we monitored full-length NS5-PDZ domain interactions in a cellular context using the Split-Nanoluciferase Protein Complementation Assay (N2H) 15 . We used the full-length (FL) NS5 protein (wt PBM) instead of NS5 PBM peptide used for holdup assay. We constructed a FL NS5ΔPBM protein which contains -TVL base deletions (∆PBM), used as control. The FL NS5 proteins and the library of PDZ domains (257 over 266 PDZ   (Fig. 2B). Thus, wt PBM in the context of full length NS5 protein is functional in cellulo, comforting our in vitro results with the holdup assay. Together, these two high-throughput methods confirmed the functionality of WNV NS5 PBM to bind PDZcontaining proteins. To go further, we started assembling from the human ORFeome collection a subset of ORFs www.nature.com/scientificreports/ encoding full-length proteins containing PDZ domains to be transferred in the N2H system. To date, our bank covers 60.5% of the human PDZome with 91 full-length proteins out of the 152 human proteins known to contain PDZ domains (Table 1). 23 full-length proteins interacting with full-length NS5 protein displayed an NLR > 1.35 NLR (Suppl. Fig. 1A). We concluded that PBM of NS5 and these PDZ domains were unambiguously functional in this context of full-length proteins. The three lists of PDZ proteins obtained by high-throughput holdup and N2H methods (PDZ domains and full-length PDZ proteins) were compared using a Venn diagram (Suppl. Fig. 1). Eleven proteins were identified in common by holdup assay and N2H method (PDZ domains), five were identified by holdup assay and N2H method (full length proteins) and four were identified by the two N2H methods (PDZ domains and full-length proteins).
To determine whether a particular ontology class or pathway could be established among the PDZ-containing proteins identified as WNV NS5 binders by our experimental approaches, we chose to use the protein annotation program PANTHER (protein annotation through evolutionary relationships). This overrepresentation test led us to analyze our large-scale experimental data against current annotated gene data set. PANTHER analysis was performed with the two lists of PDZ-containing proteins obtained from holdup (29 PDZ NS5 binders; 27 proteins) and from N2H (28 PDZ NS5 binders; 25 proteins) assays. Protein IDs were imported into the Gene Ontology (GO) enrichment analysis tool based on protein analysis through the evolutionary relationships classification system. The cellular component categories of GO having P values < 0.05 were considered to be statistically significant and then selected. PANTHER overrepresentation test based on the GO of molecular functions, cellular component, biological process and protein class (Suppl. Fig. 1B-D and Suppl. Table 1) showed that the majority of the 41 proteins have binding function, are involved in cellular process and more particularly the top five were annotated to membrane, cell periphery, plasma membrane, cell junction and cell projection. From these 41 proteins, some were already known to interact with viruses, and more particularly with Flaviviruses. Considering the results obtained with our high-throughput approaches, PANTHER analysis and the literature, we selected 10 PDZ-containing proteins for further analysis, which were at least in two lists of proteins identified (Suppl. Fig. 1) and which have at least been documented in interaction with another viral protein: ARHGAP21, CNKSR3, DLG1, HTRA1, MAGI1, PARD3, RIMS2, SHANK2, SNX27 and TJP1 ( Table 1).

Depletion of cellular PDZ-containing proteins affects West Nile virus replication. To investigate
whether the selected cellular PDZ proteins we identified to interact with the WNV NS5 PBM could have a functional role in viral replication, we decided to test depletion of these proteins from HEK-293T cells using siRNAs (Fig. 3). We first compared the replication rate of wt and mPBM replicons with the non-targeted control siRNA using a luciferase assay at different time point over 48 h post-transfection of HEK-293T cells (Fig. 3A). We obtained results similar to that presented in Fig. 1B using BHK-21 cells with a delay in replication for the mPBM (Fig. 3A) and a significant ten-fold decrease compared to wt at 24-and 48-h post-transfection (Fig. 3B). Then, we tested for depletion of the 10 selected cellular PDZ proteins, namely ARHGAP21, CNKSR3, DLG1, HTRA1, MAGI1, PARD3, RIMS2, SHANK2, SNX27 and TJP1 ( Table 1). Depletion of the corresponding transcripts was assessed by real-time quantitative PCR and transcripts with a significant decrease in quantity are presented in Fig. 4A. The level of protein expression was checked by Western Blot after siRNA treatment (Fig. 4B). We obtained both a significant decrease in transcripts and a decrease in protein level ranging from 52.5 to 91.6% in HEK-293T cells at 48 h post-transfection (Fig. 4A,B).
We then monitored the Gluc activity of both wt and mPBM replicons over 48 h post-transfection upon siRNA treatment. Depletion of RIMS2 and SNX27 did not impact replication of either replicon in our conditions at 24-and 48 h post-transfection compared to control siRNA treatment (non-targeted siRNA) (Fig. 5A). RIMS2 and SNX27, although able to interact with WNV NS5 PBM through their PDZ domains in vitro and in cell, do not seem to play a role in the virus replication in our assay. Depletion of CNKSR3, DLG1, HTRA1 and MAGI1 affected replication of both wt and mPBM replicons in our conditions at 24-and 48 h post-transfection compared to control siRNA treatment, suggesting that these proteins impact WNV replication independently from the substitution we incorporated in the PBM sequence (Fig. 5B). In contrast, depletion of PARD3, ARHGAP21 and SHANK2 resulted in a significant 5.3-, 7.2-and 5.1-fold decrease respectively at 24 h and 2.7-, 19.7-and 14.0fold decrease respectively at 48 h for the wt replicons post-transfection but not for mPBM replicons (Fig. 5C). Similarly, depletion of TJP1 resulted in a significant two-fold decrease at 48 h of replication for the wt replicons but not for the mPBM replicon. Thus, these results reveal a significant PBM-dependent effect of NS5 on WNV replication involving at least the four host PDZ-containing proteins PARD3, ARHGAP21, SHANK2 and TJP1.
N2H method with full-length proteins (full-length NS5 and full-length human ORF protein) confirmed the interaction between WNV NS5 and four PDZ-containing proteins PARD3, ARHGAP21, SHANK2 and TJP1. Of note, TJP1, PARD3, SNX27 and DLG1 are not present in this bank. We performed a protein-protein interaction in cellular context using N2H method with this new bank and found strong interactions for ARHGAP21 and SHANK2 (NLRs values of 4.41 and 4.38; respectively (Table 1, second part)), in the top 4 over the 91 NLR obtained, we also found a weaker interaction with CNKSR3 (NLR of 1.9). By contrast, RIMS2, HTRA1, MAGI1 displayed low NLR values (Table 1, second part). Thus, we confirmed in cell the significant interaction of full length PDZ-containing proteins ARHGAP21 and SHANK2 with the FL NS5 protein.

Discussion
The viral NS5 protein is the largest and most conserved Flavivirus proteins. It plays a major role in viral replication through its MTase and RdRP domains and in IFN signaling 11,17,18 . Interactions between NS5 PBM and PDZ-containing cellular proteins have been associated with interferon antagonism, viral replication and disease pathogenesis 19 www.nature.com/scientificreports/ regarding viral replication [9][10][11] . In this study, we showed that the PBM (-TVL COOH ) motif at the C-terminus of WNV NS5 protein interacts with human PDZ domains in vitro but is also capable of interacting with cellular PDZ-containing proteins in a cellular context. We established in this study the promiscuity of 41 different PDZcontaining proteins both in vitro and in cellulo (listed in Table 1) with the NS5 PBM of WNV by screening a PDZ bank which covers more than 96% of the human "PDZome". Melik and co-workers previously observed such a broad repertoire of potential host PDZ proteins recognized by the NS5 PBM of WNV using a qualitative PDZ array partially covering the human PDZome 10 . We noted that the lists of partners selected from our two high-throughput screenings only partially overlapped. Several reasons may explain this observation; by contrast with their cellular expression in the N2H assay, the concentrations of partners are strictly controlled in vitro holdup assay. While the interactions involved viral PBM peptide and the PDZome in holdup assays, the NH2 screening involves full-length NS5 as bait, the PDZome but also full length PDZ-containing proteins in the case of ORFome screening. Thus, the expression environment and the type of constructs used could also have a strong impact on PPI detection, justifying our choice to combine results from holdup and N2H assays.
Our study supports a short-list of 10 PDZ-containing proteins selected from our crossed high-throughput screenings and for their functions in the cells and during viral infection. We tested the physiological relevance of these 10 interactors on WNV RNA replication using the siRNA technology and the replicon containing an altered PBM with the mutation Leu to Met at the C-terminus of NS5 allowing to conserve a proper predicted RNA structure. Eight of the ten selected PDZ-containing proteins significantly impacted WNV replicon replication: CNKSR3, DLG1, HTRA1, MAGI1, PARD3, ARHGAP21, SHANK2 and TJP1. We show that PARD3, ARH-GAP21, SHANK2 and TJP1 displayed a significant PBM-dependent effect on WNV replication using the L-to-M PBM mutation as a gauge of the NS5 PBM influence. In contrast, CNKSR3, DLG1, HTRA1 and MAGI1 affected WNV replication regardless of the presence or absence of mutation in the PBM. We cannot exclude the presence of an unidentified internal PBM in NS5 that might interfere as other PBM motifs located internally rather than at the C-terminus of NS5 were shown to play a role the replication of TBEV and DV 10 . How CNKSR3, DLG1, HTRA1 and MAGI1 proteins are directly or indirectly connected to WNV replication should be documented in further studies. Finally, we identified RIMS2 and SNX27 as WNV NS5 PBM binders. They did not interfere with the viral replication assay, but they could participate in other steps of the viral cycle. Interestingly, RIMS2 has been shown to interact with TBEV NS5, potentially affecting the neuronal signalling of this neurotropic Flavivirus 9 .
Importantly, all our selected PDZ-containing proteins, except CNKSR3 and SHANK2, have been previously reported as host proteins targeted by viruses. MAGI1, PARD3, RIMS2 and TJP1 are targets of Flavivirus proteins (WNV, TBEV and DV) [9][10][11] . Others cellular PDZ proteins were also identified as target for other virus including DLG1, MAGI1 and ARHGAP21 for Influenza virus 16,[22][23][24] ; HTRA1, PARD3, SNX27, DLG1, MAGI1 for Human Papillomavirus 11,25-28 . The large number of cell proteins targeted by WNV could be related to the broad host repertoire of Flaviviruses replicating in very different cells and species necessary for its complex zoonotic www.nature.com/scientificreports/ transmission cycle. Our PANTHER analysis shows that these ten proteins were involved in a global perturbation of cellular homeostasis, architecture of the cells (membrane trafficking, cell junction, cellular polarity…), innate immune response (IFN mediated-JAK-STAT signalling) and neuronal cells (synapse). We report that NS5 of WNV can interact with TJP1 (also known as ZO-1), ARHGAP21 and PARD3 proteins through PBM-PDZ interactions. These three proteins are involved in cell polarity, trafficking, cell adhesion and cell junction. ARHGAP21 protein is involved in trafficking through the control of CDC42 activity 29 . Taye et al. performed a microarray analysis and obtained an overexpression of ARHGAP21 which negatively regulates the transport of Influenza A virus (H1N1) neuraminidase to the cell surface 22,30 . TJP1 was reported to interact with other Flavivirus NS5 9,10 . Interestingly, we found both TJP1 and TJP3 as interactors of WNV NS5 PBM in our study ( Fig. 2A and B and Table 1). TJP1, TJP2 and TJP3 are members of the membrane-associated guanylate kinase (MAGuK) family of proteins 31 that play a particular role in tight-junctions (TJs) organization 32 . TJP1 can bind directly to TJP3 to form a TJP1/TJP3 complex that acts directly on the actin skeleton. Interestingly, it was demonstrated that actin filaments participate in WNV maturation process 33 . Moreover, TJs assembly and permeability are highly regulated by signalling pathways 34 . It was shown that TBEV NS5 interacts with GIPC and TJP2 through PDZ/PBM complex. The MAGuK TJP2 protein is an interactor of Scribble, a PDZ-containing protein involved in cell polarity and neuronal function. The complex NS5-TJP2-hScrib may shuttle the polymerase to the plasma membrane, regulating the flaviviral replication that occurs at ER membrane-associated complexes 11 . As TJP1, PARD3 displayed a significant NS5 PBM-dependent effect on WNV replication. Chen et al. showed that suppression of PARD3 expression by RNA interference (RNAi) caused a dramatic disruption of TJ assembly 35 . Of note, we also detected PARD3B, a human PARD3 homologue ( Fig. 2A,B and Table 1), which also localizes at the TJs 36 . Moreover, PARD3 and TJP1 are recruited by JAM-2 to cell-cell contacts 37 . TJP1 is also associated with JAM-1 through PDZ domain 3 38 . The association of TJP1 with these different proteins in TJs serves to cluster the integral membrane proteins at TJs and could be partially disrupted by WNV NS5 during infection resulting in a progressive disruption of TJ, which leads to leakage between adjacent cells, loss of barrier function, and infiltration of WNV virions 37 . Suppression of host innate immunity is one of the multiple crucial functions of Flavivirus NS5 protein. The type I IFN signalling pathway included in the first lines of defence against Flavivirus infection of mammals. TBEV and WNV NS5 inhibit IFN-I signalling by suppressing surface expression of interferon alpha and beta receptor subunit 1 (IFNAR1) 39 . Indeed, TBEV may manipulate the IFN pathways through interaction with TJP1 as TBEV-NS5 (and DV-NS5) was shown to be present in the nucleus at the same moment as TJP1 9 . To note, in WNV-infected cells, a high proportion of NS5 resides in the nucleus 40 but it is also accurate for other Flavivirus NS5 (DV 41 ; Yellow Fever 42 ). A consequence of the NS5/PDZ-containing protein interactions could be  Asterisks indicate significant differences between replicon activities, *p < 0.05, **p < 0.01 and ***p < 0.005 obtained with a Mann-Whitney unpaired t-test. www.nature.com/scientificreports/ the relevance of our findings as WNV is indeed a neurotropic virus. The nature of the assay we used, restricted the characterization of the functional significance of such PDZ dependent WNV-host interactions to viral replication. Further characterization of the mechanisms by which PBM/PDZ interactions may alter neuronal functions is therefore essential.
Production of recombinant WNV. We first produced a full-length infectious clone for wt WNV and mPBM WNV as previously described by Alsaleh et al. 44  Holdup assay with human PDZ domains. All PDZ domains (259/266 PDZ) were expressed following the high-throughput protocol previously described 45 . The peptides biotinylated WNV NS5 PBM (peptide sequence-RYEDTTLVEDTVL) was synthesized in solid phase using Fmoc strategy (Proteogenix) and resuspended in H 2 O. The holdup assay was carried out against the biotinylated WNV NS5 PBM in quadruplicates as previously described 14,45 with minor modifications. We measured WNV NS5 PBM interactions against 259 human PDZ domains. The minimal BI threshold value is 0.2 to define a significant interaction as previously reported 14 .
Cloning all PDZ domains and ORFs into gateway-compatible expression plasmids for split-nanoluciferase protein complementation assay (N2H). All PDZ domains (257/266 PDZ) and ORFs containing PDZ domains (91 full-length proteins out of the 152 human proteins) were introduced into Gateway destination vector pYN2H-LEU-N1 45 whereas NS5 and NS5∆PBM (NS5 with a deletion of the three residues -TVL-) were introduced into Gateway destination vector pYN2H-TRP-N2, via LR clonase-mediated Gateway reaction (Life Technologies), following the protocol previously described 46 . Briefly, LR reaction products were subsequently transformed into DH5α competent bacterial cells and grown for 24 h at 37 °C at 900 rpm in carbenicillin-containing TB liquid medium. Plasmid DNA was extracted using a NucleoSpin 96 Plasmid kit from Macherey-Nagel, following manufacturer's instructions. All DNA were sequenced using plasmid-specific primers.
Split-nanoluciferase protein complementation assay (N2H) into HEK cells. HEK-293T cells were seeded at 5.4 × 10 4 cells per well in 96-well, flat-bottom, cell culture microplates (Greiner Bio-One, #655083), and cultured in Dulbeccoʼs modified Eagleʼs medium (DMEM) supplemented with 10% fetal calf serum at 37 °C and 5% CO 2 . Twenty-four hour later, cells were transfected with 100 ng of pYN2H-LEU-N bait and 100 ng of pYN2H-TRP-N2 prey using linear polyethylenimine (PEI) to co-express the peptide/protein pairs or the protein/protein pairs fused with complementary NanoLuc fragments, N1 and N2, as previously described by Choi et al. 46 . Twenty four hours after DNA transfection, the cell culture medium was removed and 50 µL of mix (100 mM MES pH 6.0, 1 mM CDTA, 0.5% (v/v) Tergitol, 0.05% (v/v), Mazu DF 204, 150 mM KCl, 1 mM DTT, and 35 mM thiourea; Furimazine 20 µg/ml) were added to each well, plates were incubated for 3 min at room temperature, and luminescence output was measured using a Centro XS LB 960 luminometer (Berthold; 2 s integration time). All binary PPI experiments were independently performed three times. Control experiments were performed similarly where NS5/NS5∆PBM protein fused to the NanoLuc fragment N1 was co-expressed with the matching NanoLuc fragment alone. For each peptide/protein pair -N1/-N2, we calculated a normalized luminescence ratio (NLR) corresponding to the raw luminescence value of the tested pair (-N1/-N2) divided by the maximum luminescence value from the control, as previously described by Cassonnet et al. 15
Cells were lysed at 4-, 24-, 48-and 72 h post electroporation with Renilla Luciferase Assay Lysis Buffer according to the manufacturer's instruction (Promega). Mean values of Renilla-normalized firefly Luciferase expression were determined from quadruplicate wells. At least three independent transfections in cells were performed for each experiment.
RTqPCR. Total RNA from infected or transfected HEK-293T cells was extracted using Nucleospin RNA kit (Machery Nagel) according to the manufacturer's protocol and reverse transcribed into cDNA using Moloney Murine Leukemia Virus Reverse Transcriptase (Invitrogen, Life Technologies). A SYBR Green-based real-time PCR assay (QuantiTect SYBR Green kit, Qiagen) for relative quantification of transcripts was performed on 384well plate QuantStudio 6 Flex real-time PCR system (Applied Biosystems) according to protocol as previously described 19,20 . The relative quantities were determined by using raw Cp values as input for qBase, a flexible and open source program for qPCR data management and analysis 48 ; the primers used in this manuscript were listed in Suppl. Table 2.
Western Blot. Protein lysates were prepared by cell lysis in radioimmunoprecipitation assay (RIPA) buffer (BioBasic Canada) containing protease inhibitors (Roche), then the cells were frozen at −20 °C. Five μl of NuPAGE LDS sample Buffer (Thermo Fisher Scientific) was added to 20 μl of lysate and incubated at 95 °C for 5 min. Samples were separated on NuPAGE 4-12% Bis-Tris Protein Gels and transferred to a polyvinylidene difluoride (PVDF) membrane (Life Technologies). The membrane was washed for 1 h at room temperature in PBS-Tween (PBS-T) plus 5% milk, then incubated overnight at 4 °C with the different antibodies (Suppl. Table 2). After washing in PBS-T the membrane was incubated for 1 h at room temperature with HRP-conjugated antibody (BioRad). After washes in PBS-T, the membrane was developed with enhanced chemiluminescence reagent (Pierce) and exposed to film.