The herpesvirus replication cycle comprises maturation processes in the nucleus and cytoplasm of the infected cells. After their nuclear assembly viral capsids translocate via primary envelopment towards the cytoplasm. This event is mediated by the nuclear envelopment complex, which is composed by two conserved viral proteins belonging to the UL34 and UL31 protein families. Here, we generated recombinant viruses, which express affinity-tagged pM50 and/or pM53, the pUL34 and pUL31 homologues of the murine cytomegalovirus. We extracted pM50- and pM53-associated protein complexes from infected cells and analysed their composition after affinity purification by mass spectrometry. We observed reported interaction partners and identified new putative protein–protein interactions for both proteins. Endophilin-A2 was observed as the most prominent cellular partner of pM50. We found that endophilin-A2 binds to pM50 directly, and this interaction seems to be conserved in the pUL34 family.
Herpesvirus maturation is controlled by the action of specific protein complexes of viral and host cell origin and takes its course in both the nucleoplasm and the cytoplasm (Johnson and Baines, 2011). After their nuclear assembly viral capsids incorporate single full-length genomes and migrate into the cytosol to acquire a tegument and their final envelope (McVoy et al., 2000; Mettenleiter, 2006). For this, however, the nucleocapsids have to traverse the nuclear envelope (NE). This structure represents a physical barrier which is composed of an outer and an inner nuclear membrane (ONM, INM) and is closely associated with a protein meshwork of intermediate filaments, the so-called nuclear lamina, on its interior (Foisner, 2001). Although many pore complexes cross the NE and allow translocation of biomolecules between nucleus and cytoplasm, herpesvirus particles exceed the size limit of the nuclear pores and are unable to traverse (Pante and Kann, 2002). Instead, fully assembled nucleocapsids contact the INM and bud into the lumen of the NE. Hereby they transiently acquire a primary envelope that eventually fuses with the ONM or the endoplasmic reticulum (ER) respectively (Stackpole, 1969; Lee and Chen, 2010). This primary envelopment seems to be a common feature of all herpesviruses and shows mechanistic similarities to a recently described nuclear pore independent cellular RNP export mechanism (Speese et al., 2012). In herpesviruses the nuclear egress is managed by two conserved gene products belonging to the UL34 and UL31 protein families (Muranyi et al., 2002; Lake and Hutt-Fletcher, 2004; Reynolds et al., 2004).
The pUL34 family members are type II integral membrane proteins which localize to the NE and the ER, while the pUL31 homologues are nuclear matrix-associated phosphoproteins (Reynolds et al., 2001; Muranyi et al., 2002; Gonnella et al., 2005). The proteins associate at the INM and form the so-called nuclear egress complex (NEC) whose function is to induce a partial disassembly of the nuclear lamina. The UL34 and UL31 protein homologues interact with lamin A/C and lamin B (Dal Monte et al., 2002; Muranyi et al., 2002; Reynolds et al., 2004; Gonnella et al., 2005; Milbradt et al., 2007; Mou et al., 2008) and recruit viral (Marschall et al., 2005; Lee et al., 2008; Mou et al., 2008) and cellular (Muranyi et al., 2002; Park and Baines, 2006; Milbradt et al., 2009; Leach and Roller, 2010) kinases like protein kinase C. Phosphorylation events disrupt the interaction between lamins and stabilizing integral INM proteins like lamin B receptor (LBR) and emerin by which the nuclear lamina eventually becomes disintegrated (Scott and O'Hare, 2001; Leach et al., 2007). Moreover, the nuclear egress proteins have been shown to be associated with perinuclear virions from which they become stripped off at the ONM during the budding process (Fuchs et al., 2002; Reynolds et al., 2002; Padula et al., 2009). Although this is a conserved process, the details and the number of proteins involved may still be slightly different within the herpesvirus subfamilies.
In this study we sought to identify cell proteins associated with the murine cytomegalovirus (MCMV) NEC which is composed of pM50 and pM53, members of the UL34 and UL31 protein family respectively. By improved galK-mediated bacterial artificial chromosome (BAC) mutagenesis, we generated recombinant MCMVs expressing tagged pM50 and/or pM53 from their native loci. The affinity-tagged viruses expressed the NEC proteins with wild type (wt) characteristics and allowed large-scale purification of the NEC and associated proteins from infected murine cell lines. Therefore, protein complexes associated with tagged pM50 and/or pM53 in their natural composition during viral infection became accessible to LC-MS/MS (liquid chromatography-tandem mass spectrometry) analysis. Several proteins were retrieved; some of them such as pm14 (Fossum et al., 2009), as well as the cellular proteins lamin A/C, lamin B and C1QBP/p32, are already known interaction partners of the NEC proteins of different herpesviruses (Dal Monte et al., 2002; Muranyi et al., 2002; Reynolds et al., 2004; Gonnella et al., 2005; Milbradt et al., 2007; Mou et al., 2008). As a novel direct interaction partner of pM50 and pUL34 of herpes simplex virus 1 (HSV-1), we identified endophilin-A2.
Construction and characterization of MCMV expressing pFLAGM53 and/or pM50HA
The MCMV NEC is based on the interaction of the viral proteins pM50 and pM53 during viral infection. For studying associated proteins galK-based BAC recombineering was used to design recombinant viruses which encode for epitope-tagged pM50 and/or pM53. In previous studies HA and FLAG affinity tags demonstrated their applicability for coimmunoprecipitation (Co-IP) without affecting the functionality of the NEC proteins (Rupp et al., 2007; Popa et al., 2010). Here, we tagged the M50 open reading frame (ORF) with a HA epitope at its C-terminus and the M53 ORF with a FLAG epitope at the N-terminus. We constructed MCMV-BACs expressing pFLAGM53 (pSM3fr-FLAGM53) and pM50HA (pSM3fr-pM50HA) (Fig. S1B and C). We also introduced the genetically tagged M50 gene sequence into the pSM3fr-FLAGM53 resulting in pSM3fr-M50HA-FLAGM53 expressing both tagged NEC proteins. All three BACs were reconstituted successfully after transfection of mouse embryonic fibroblasts (MEFs) with kinetics resembling those of the wt pSM3fr. The resulting recombinant viruses MCMV-M50HA, MCMV-FLAGM53 and MCMV-M50HA-FLAGM53 were passaged four times on M2-10B4 cells and high titre stocks were prepared.
Next, we analysed whether the tagging had an influence on the functionality of pM50 and pM53. For this purpose we infected NIH/3T3 cells for 36 h either with wt MCMV or with the recombinant viruses MCMV-M50HA, MCMV-FLAGM53 and MCMV-M50HA-FLAGM53 and compared the expression patterns of pM50 and pM53 in all viruses via Western blot analysis. To ensure that the recombinant NEC proteins expressed their respective affinity tags, the blots were additionally incubated with anti-FLAG and anti-HA antibody. Staining with serum against the immediate early protein 1 and with an antibody specific to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as viral and cellular loading control respectively. All protein lysates exhibited comparable protein contents and the three viral mutants expressed the tagged NEC proteins similar to the wt virus (Fig. 1a). Additionally, we determined the propagation kinetics of the recombinant viruses by performing multiple-step growth curves in NIH/3T3 cells (Fig. 1b). All recombinant viruses produced infectious particles comparable to wt MCMV, demonstrating that the nuclear egress proteins maintained their essential function upon tagging.
Mass spectrometry analysis of the MCMV NEC identified cellular endophilin-A2 as a novel interaction partner
Viruses expressing tagged gene products were used to identify interaction partners of pM50 and/or pM53 during viral infection via LC-MS/MS. After infecting NIH/3T3 cells at a multiplicity of infection (MOI) of 1 for 36 h, we achieved optimal protein extraction by lysis with 1% Triton X-100 and 400 mM NaCl (data not shown). The extracted protein complexes were affinity-purified on either anti-FLAG or anti-HA matrices, typically digested after elution and subsequently analysed by mass spectrometry. NIH/3T3 infected with wt MCMV served to control the specificity of the immune precipitations. Only proteins which appeared in Co-IPs via the HA and/or the FLAG affinity tag, but not in the respective control samples, were of interest. In eight independent experiments several proteins of viral (Table 1) and cellular (Table 2) origin were found to be associated with pM50HA and/or pFLAGM53.
Table 1. MCMV proteins identified as putative interaction partners of the NEC proteins by LC-MS/MS analysis
Among the viral proteins precipitating together with pM50HA and/or pFLAGM53 (Table 1), we observed pM36. This protein carries out anti-apoptotic functions and is the MCMV homologue of the human cytomegalovirus (HCMV) protein pUL36 (Skaletskaya et al., 2001; Menard et al., 2003). It is required for replication in macrophages (Menard et al., 2003) and is essential for in vivo infection (Cicin-Sain et al., 2005). Specific interactions with pM50 and pM53 were also observed with the serine-alanine-rich glycoprotein pM116, whose function is unknown yet (Rawlinson et al., 1996). The last identified proteins are products of the US22 genes M139, M140 and M141 and have been shown to regulate MCMV replication in macrophages. All three polypeptides can interact with each other and assemble into stable protein complexes (Menard et al., 2003; Karabekian et al., 2005).
Among the cellular proteins (Table 2) we identified proteins implicated in clathrin-coated endocytosis and vesicular trafficking, namely endophilin-A2 (Giachino et al., 1997; Ringstad et al., 1997) and sorting nexin 9 (SNX9) (Worby and Dixon, 2002; Lundmark and Carlsson, 2003) as well as the chaperone proteins GRP94 (Taipale et al., 2010), hypoxia upregulated protein 1 precursor (Takeuchi, 2006) and serpin H1 (Gettins, 2002). Besides cytoskeletal components like tubulin beta 2c (Wang et al., 1986), spectrin alpha and beta chain (Gallagher and Forget, 1993), filamin-B (Stossel et al., 2001), microtubule-associated protein 1B (Pedrotti and Islam, 1995) and talin-1 (Tanentzapf and Brown, 2006), we detected 14 additional proteins with miscellaneous cellular functions (Table 2).
Endophilin-A2 coimmunoprecipitated most regularly with pM50HA and pFLAGM53 under various extraction conditions (150 mM to 1 M NaCl concentration). Therefore, this protein was analysed in more detail. To validate the LC-MS/MS results, NIH/3T3 cells were mock-treated, and infected at a MOI of 1 with MCMV-FLAGM53 and MCMV-M50HA-FLAGM53 for subsequent HA-IP or with MCMV wt and MCMV-FLAGM53 for FLAG-IP. After 36 h the cells were lysed and the protein extracts were subjected to HA- or FLAG-IP assays. The protein extracts of mock and MCMV-FLAGM53-infected cells served as controls to distinguish specific and unspecific protein interactions. Eluted protein was separated by SDS-PAGE on a 12% polyacrylamide gel and either silver-stained to visualize the total protein in each sample (Fig. 2a) or blotted onto a PVDF membrane and reacted with endophilin-A2 antibody plus donkey anti-goat HRP-coupled antibody (Fig. 2b). Notably, also in the controls a considerable amount of protein was binding unspecifically to the affinity gel (Fig. 2a). Especially during FLAG-IP we frequently observed a high degree of unspecific protein binding. In the immunoblot (Fig. 2b), however, endophilin-A2 could not be detected in any of the controls, but coimmunoprecipitated specifically with the NEC proteins after HA- and FLAG-IP respectively. In both Co-IP settings endophilin-A2 was identified by a protein band at 42 kDa (Fig. 2b).
Endophilin-A2 interacts with the NEC via pM50
Endophilin-A2 coprecipitated via both pM50HA and pFLAGM53 (Table 2, Fig. 2b). First, we wished to determine whether endophilin-A2 interacts with either pM50HA or pFLAGM53 independently from other viral proteins Therefore, we analysed the endophilin-A2–NEC interactions in transfection. M2-10B4 cells were transfected with an empty expression plasmid as mock control (Fig. 2c, lane 1), with plasmids expressing pM50HA and pFLAGM53 alone and in combination (Fig. 2c, lanes 2–4) or with an expression vector for the protein mutant M50HA-ΔP, which lacks a functionally important proline-rich sequence motif (Fig. 2c, lane 5) (Rupp et al., 2007). Twenty-four hours post transfection, proteins were extracted and immunoprecipitated using either HA-, FLAG- or endophilin-A2-specific antibody (Fig. 2c; Input: upper, middle and lower rows respectively). Both input material (Fig. 2c; Input) and IP eluates (Fig. 2c; IP) were immunoblotted and reacted with specific antibodies as indicated. An equal load of sample material was verified by GAPDH staining (Fig. 2c; Input: upper row). In HA-specific IP endophilin-A2 precipitated with pM50HA in absence or presence of pFLAGM53 (Fig. 2c; IP: upper row). Interestingly, endophilin-A2 did not coimmunoprecipitate with pM50HA-ΔP (Fig. 2c; IP: upper row, lane 5), although latter protein was present in higher amounts compared to pM50HA in lanes 2 and 4. When the cell extracts were subjected to FLAG-IP, endophilin-A2 precipitated only in presence of pM50HA, but not with pFLAGM53 alone (Fig. 2c; IP: middle row). Conversely, if endophilin-A2 was pulled down from cell extracts it was able to co-purify pM50HA, but not pM50HA-ΔP (Fig. 2c; IP: lower row). These data suggested that the association of endophilin-A2 with the nuclear egress proteins is based on the interaction with pM50. Notably, endophilin-A2 harbours a SH3 domain and pM50 has proline-rich sequence motif. We hypothesized that both proteins might interact with each other via these domains. Interestingly, pM50HA-ΔP and endophilin-A2 were not able to coimmunoprecipitate each other (Fig. 2c; IP: upper and lower rows, lane 5), which implies that the proline-rich stretch is necessary for this interaction to happen.
Endophilin-A2 interacts directly with pM50 and pUL34 of HSV-1
Our data indicated that the endophilin-A2–pM50 interaction relied on the proline-rich sequence within pM50 (Fig. 2c; IP: lane 5). To further pinpoint the interaction sites within both proteins we analysed deletion mutants of endophilin-A2 and pM50 in the yeast-two-hybrid (Y2H) interaction assay (Fig. 3b). In addition, also pUL34 was included, the pM50 homologue of the prototypic alpha-herpesvirus HSV-1. For this Y2H assay we used the human endophilin-A2 which shares 94% sequence identity with its mouse homologue. Here, also pM50 interacted with human endophilin-A2 and with the deletion construct endophilin-A2300–368, largely being comprised of the endophilin-A2 SH3 domain (Table 3, Fig. 3b). Also pUL34 bound to full-length endophilin-A2, but not to endophilin-A2300–368. No interaction was detectable between pM50 or pUL34 and endophilin-A21–300 lacking the SH3 domain (Table 3, Fig. 3b). The deletion mutant pM50172–288 contained the proline-rich domain (PRD) and showed strong binding to full-length endophilin-A2 and the SH3 domain harbouring endophilin-A2300–368, but not to endophilin-A21–300 lacking the SH3 domain (Table 3, Fig. 3b). Three constructs of pM50 lacking the PRD (pM501–170, pM501–191 and pM50172–288Δ179–201) were also tested (Fig. 3b). The protein mutant M501–170, however, allowed yeast growth in all prey construct combinations and had therefore to be considered as auto-activating mutant (data not shown). The polypeptides pM501–191 and pM50172–288Δ179–201 did not show any cooperation with the endophilin-A2 preys (Table 3, Fig. 3b). In summary, both pUL34 and pM50 interacted with endophilin-A2. The PRD of pM50 was required to bind to the endophilin-A2 SH3 domain. As pUL34 does not contain a definitive PRD, its interaction with endophilin-A2 has to be mediated by other motifs or structural elements (Fig. 3a).
Table 3. Interaction intensitiesa of pUL34, pM50 and pM50 mutants with endophilin-A2 constructs in a Y2H assay
Prey – pGADT7 harbouring:
Bait – pGBKT7 harbouring:
aInteraction intensities are represented as follows: +++, strong; ++, medium; +, weak; −, no interaction.
Subcellular localization of endophilin-A2 and herpesvirus proteins pUL34 and pM50 in infected cells
It was of interest to determine the intracellular compartment in which the interaction between endophilin-A2 and pM50 as well as pUL34 occurs. The subcellular localization of endophilin-A2 was studied by immunofluorescence using either specific antibodies or transiently expressed mCherry–endophilin-A2 fusion protein. The cells were counterstained by the nucleic acid stain TO-PRO-3 iodide to indicate the nucleoplasm (Fig. 4a ). The specific antibody detected some endophilin-A2 in the nucleus, but the bulk of the signal was found in the cytoplasm of uninfected NIH/3T3 cells, arranged in a filamentous reticular configuration (Fig. 4b). The same localization was confirmed with transiently expressed mCherry-tagged endophilin-A2 (data not shown). In infected cells nucleoplasm and cytoplasm are labelled with N and C respectively (Fig. 4f, i, j, p, r). Upon infection with MCMV or HSV-1 cells rounded off, with the vast majority of endophilin-A2 being present in the cytoplasm (Fig. 4e, m, r; red signal). Compared to uninfected cells (Fig. 4b), endophilin-A2-positive vesicular structures greatly increased in number and size during infection (compare Fig. 4b and 4e as well as 4o) and were embedded in perinuclear notches tightly surrounded by membrane structures loaded with pM50 or pUL34 respectively (Fig. 4f, j, k, p). As expected for type II membrane proteins, pM50 and pUL34 were resident in structures typical for the ER and they showed the expected nuclear rim staining (Fig. 4d, m, n; green signal).
Moreover, pM50 was detectable in membrane structures within the nucleoplasm (Fig. 4d, g; green signal), in ring-like structures originating from the INM (Fig. 4j; arrow) as well as in nuclear membrane duplications similar to those caused by the pUL34 homologue BFRF1 of Epstein–Barr virus (Gonnella et al., 2005). Although an almost complete colocalization of pM50 and endophilin-A2 at the nuclear rim was also observed in a small proportion of MCMV-infected cells (Fig. 4g, h, i), usually both endophilin-A2 and the pUL34 family members colocalized in vesicular structures contiguous to the nuclear rim (Fig. 4j) and within the cytoplasm (Fig. 4f, m, p). In Fig. 4 kareas of strong colocalization are highlighted in purple and display mainly roundish structures. Statistics on the colocalization of endophilin-A2 and the viral proteins pUL34 and pM50 were calculated using the BioImageXD analysis software (Kankaanpää et al., 2006). 2D scatterplots were computed for comparison of the distribution and intensity of red and green signals within the images. Both pM50 and pUL34 apparently colocalized with endophilin-A2 in infected cells as indicated by the calculated P-values. In the enlarged section in Fig. 4j the colocalization parameters Pearson's correlation coefficient/M1/M2 of pM50 and endophilin-A2 were determined with 0.7567/0.8326/0.9387 respectively. In Fig. 4n the colocalization parameters Pearson's correlation coefficient/M1/M2 of pUL34 versus endophilin-A2 were calculated with 0.6803/0.9955/0.9533. The corresponding 2D scatterplots are shown in Fig. 4l (belonging to Fig. 4j) and in Fig. 4q (belonging to Fig. 4p). MCMV- and also HSV-1-infected cells underwent a structural polarization, in which the often kidney-shaped nucleus localized to one side of the infected cell, whereas endophilin-A2-positive vesicular speckles were situated to the other side (Fig. 4e, m, r; red signal). Also pM50 (Fig. 4d; green signal) and pUL34 (Fig. 4m; green signal) primarily showed their typical localization at the nuclear rim in addition to a less pronounced ER signal. The vesicular endophilin-A2 staining colocalized with the trans-Golgi network (TGN) marker TGN38 (Fig. 4r; arrow). This implied that the numerous cytoplasmic endophilin-A2-containing formations are TGN-derived.
Endophilin-A2 is important for HSV-1 propagation
Next, having seen this interaction of endophilin-A2 with pUL34 members, we wished to test whether endophilin-A2 is necessary for virus replication. For this purpose endophilin-A2-specific siRNAs were introduced into NIH/3T3 and HeLa cells by transfection to downregulate endophilin-A2. Unfortunately, in NIH/3T3 cells this knockdown was lethal (data not shown). The viability of the HeLa cells was not affected during the course of the experiments by the endophilin-A2 knockdown as shown by the Trypan blue exclusion assay (data not shown). As HeLa cells cannot be infected with MCMV, no conclusion could be drawn about the effect of endophilin-A2 knockdown on MCMV infection. To determine the impact of endophilin-A2 knockdown on HSV-1 infection, HeLa cells were infected with HSV-1 48 h after siRNA transfection. The number of released infectious particles was followed over time by collecting aliquots of the culture supernatants and performing plaque assay on Vero cells (Fig. 5). After one viral replication cycle (18 h), the number of progeny viruses was reduced threefold compared to the control. However, the effect became more pronounced 45 h post infection with a more than 200-fold reduction in virus titre. Altogether, these data showed that endophilin-A2-specific RNAi reduces the productivity of herpesviral infection. This points to a functional role of endophilin-A2 during herpesvirus replication, although a direct connection between this function and the interaction with pUL34 homologues remains to be shown.
Here, we identified interaction partners of the NEC proteins pM50 and pM53 during MCMV infection. To this end, the genes encoding the NEC proteins were tagged in their authentic genomic position. The resulting virus mutants expressed functional tagged NEC proteins and grew to titres comparable to the viruses from which they derived from. The affinity tags allowed us to isolate proteins associated with pM50 and pM53 from MCMV-infected cells. The protein complexes were extracted and immunoprecipitated under varying salt concentrations to identify both weak and strong interactions. After LC-MS/MS protein hits were predicted using the Mascot search engine (Matrix Science). Mascot output files were imported into the Scaffold 3 Proteome Software (Proteome Software), which allowed comparing controls and specific samples. Proteins also present in the controls were marked as unspecific. Only proteins to which a minimum of two peptides could be assigned were considered. Both peptides and proteins had to be identified with a probability of at least 95%. In repeated experiments we detected a total of 16 viral and 174 cellular proteins, from which only seven viral and 25 cellular proteins precipitated specifically with the tagged NEC proteins (Data S1–S4).
In earlier Co-IP and Y2H screens based on transfected viral genes, several viral interaction partners of pM50 and pM53 have been listed by Fossum et al. (2009). pm14, pM55 and pm119 were detected in our approach as well, but either appeared also in control samples or were identified with a too low protein identification probability. Cellular proteins were of main interest, as it has been shown that the expression of pUL31 and pUL34 family members can trigger budding events even in the absence of any other viral protein (Klupp et al., 2007). Several proteins have earlier been described as potential interaction partners of the NEC, e.g. lamins (Dal Monte et al., 2002; Muranyi et al., 2002; Reynolds et al., 2004; Gonnella et al., 2005; Milbradt et al., 2007). In our assay, prelamin A/C was present in three experiments after HA and FLAG pull-down and lamin B1 after protein extraction in presence of 1 M NaCl and subsequent HA pull-down. However, both proteins were not enlisted in Table 2, as lamin B1 did not meet our stringent protein identification criteria and prelamin A/C was also found in the control. Another cellular protein, C1QBP/p32, has been established as an interaction partner of pUL50, the HCMV homologue of pM50 (Milbradt et al., 2007). We also identified this protein as NEC binding protein. C1QBP/p32 interacts with the LBR which resides in the INM and serves as anchor for the nuclear lamina and heterochromatin (Ye and Worman, 1996; Foisner, 2001). Remarkably, with our approach we could not confirm some already described interactions between the NEC and cellular proteins. Interactions between LBR and the nuclear egress proteins of HCMV have been established in transfection experiments (Milbradt et al., 2009). It has also been shown that LBR is reorganized upon HSV-1 infection (Scott and O'Hare, 2001). Protein kinase C, which has been shown to be recruited to the nuclear rim by the nuclear egress proteins of alpha- (Park and Baines, 2006) and beta-herpesviruses (Muranyi et al., 2002; Milbradt et al., 2007), was also not identified in our assay. We could not find any evidence of an interaction between pM50 or pM53 with torsin A, which was recently described to be involved in the nuclear egress of HSV-1 capsids (Maric et al., 2011). There are several explanations for the difference to previous reports. First, transient or weak protein–protein interactions may not be sufficiently robust to withstand the extraction procedures, in particular if the interaction is indirect and involves other proteins as well. Second, during natural infection proteins may also engage in additional interactions and may be expressed only in minor amount in comparison to gene transfection. Third, the protein detection by mass spectrometry is less sensitive than certain antibody-based techniques and some proteins might simply escape identification.
Yet, we identified a number of NEC putative partners, some of them novel. Among the novel cellular partners endophilin-A2 was constantly specifically immunoprecipitated and was identified with a probability of 100% at all times. The designation endophilin is derived from the well-documented specific interaction of this protein family with proteins involved in clathrin-mediated endocytosis (Micheva et al., 1997). Endophilins bind via their SH3 domains to PRD within interacting proteins (Micheva et al., 1997). In a comparable manner endophilin-A2 apparently binds to the PRD of pM50 of MCMV, whereas the binding site in the HSV-1 pUL34 needs to be defined. Altogether, we believe that this protein interaction represents a conserved feature within the pUL34 protein family. Notably, also other viruses exploit endophilin-A2 functions, e.g. by interaction with the M2 protein of MHV-68 (Herskowitz et al., 2008) and the Gag protein of Moloney murine leukaemia virus (Wang et al., 2003). Endophilin-A2 has been reported to be a cytoplasmic protein (Shang et al., 2007), although also nucleo-cytoplasmic shuttling has been observed (Cheung et al., 2004). In our experiments endophilin-A2 localized in a filamentous reticular staining within the cytoplasm of uninfected NIH/3T3 cells (Fig. 4b and e). In neuronal cells, HeLa and 293, a similar reticular appearance was recorded for endophilin-A3 (Hughes et al., 2004), which shares a high sequence homology with endophilin-A2 (Giachino et al., 1997). Upon infection endophilin-A2 and pM50 partially colocalized at the perinuclear cytosolic vesicular structures in close proximity to the nucleus.
A much more difficult question is how endophilin-A2 might contribute to viral morphogenesis. Endophilin-A2 and also the pUL34 family proteins such as pUL34 and pM50 are membrane-associated proteins. Both pUL34 and pM50 are type II membrane protein with a C-terminal single transmembrane anchor and decorate membranes of the nucleus as well as of the ER (Muranyi et al., 2002). The N-terminal protein moiety, which contains the relatively conserved constant region of the UL34 protein family (Fig. 3a) and the PRD domain necessary for endophilin-A2 binding, is perpendicular to the membrane bilayer and might therefore be accessible to interaction partners such as endophilin-A2 (Bubeck et al., 2004). There is evidence that in addition to the pUL34/pUL31 interaction, which mediates NE targeting, pUL34 and pUL31 family members control another steps in nuclear egress subsequent to targeting to the NE, namely the potential to induce membrane curvatures (Roller et al., 2010). Notably, presence of only these two NEC proteins in cells already suffices for the generation of membrane vesicles (Klupp et al., 2007). Therefore, if endophilin-A2 was present in the nucleoplasm, it might, at least at first glance, support membrane bending and vesiculation. It is well known that upon dimerization endophilin-A2 senses and even generates membrane curvature by association of its banana-shaped N-BAR domain with lipid bilayers (Gallop et al., 2006). Unfortunately, this type of bending would rather support vesiculation into but not out of the nucleus. Therefore, the budding of nucleocapsids, at least as we understand it so far, can hardly be explained by a nuclear function of endophilin-A2.
Yet, there may be a cytoplasmic function of endophilin-A2 in cooperation with pUL34 homologues. pUL34 proteins are synthesized in the ER and enter the nucleus through contiguous membranes to become arrested in the nuclear membrane apparently via their interaction with pUL31 family proteins. Free pUL34 residing or returning to the ER, however, might engage in other functions. This is highlighted by a most recent observation of Haugo et al. (2011), who showed that pUL34 of HSV-1 is required for proper localization of gE and perhaps other glycoproteins and thus fulfils additional post-nuclear functions during viral maturation within the cytoplasm. It was already shown that the structural integrity of the NE becomes altered upon herpesvirus infection (Scott and O'Hare, 2001; Muranyi et al., 2002; Gonnella et al., 2005; Leach et al., 2007; Camozzi et al., 2008; Lee et al., 2008). One might speculate that endophilin-A2 contributes to this process by its membrane-deforming capacity which creates membrane environments for cytosolic budding processes (Morgan et al., 1973; Buser et al., 2007). Indeed, colocalization of endophilin-A2 and pM50 (and pUL34) was evident within the cytoplasm. Especially near the nucleus, bright endophilin-A2-enriched vesicles grew in number and size during infection. Interestingly, Das et al. described a structure in the cytoplasm of HCMV-infected cells which is highly vacuolated (Das et al., 2007). This structure was referred to as cytoplasmic assembly complex (AC) which develops at the microtubule organizing centre and comes along with dramatic structural rearrangements of the nucleus and the cytoplasmic endomembrane system (ER, Golgi, TGN, endosomes) (Sanchez et al., 2000; Das et al., 2007; Buchkovich et al., 2010). Here, the nucleus and cytoplasmic compartments are brought into spatial proximity and merge into a single entity. A microtubular network pervades the AC and provides ideal highways for the transport of vesicles (Hunter and Wordeman, 2000; Sanchez et al., 2000). Similar to this model we noticed nuclei taking a kidney-like shape in MCMV-infected cells. From one side of the nucleus microtubules emanate that partially colocalized with pM50. Also endophilin-A2 has been shown to interact with microtubules (Cheung et al., 2004). Endophilin-A2 is especially suited to play a role in such a cytoplasmic AC. Endophilins can bind to proteins controlling vesicular trafficking (McMahon and Gallop, 2005; Gallop et al., 2006) and A-type endophilins in particular have been shown to be involved in tubulovesicular membrane and protein trafficking events between different cellular compartments (ER, Golgi, multi-vesicular bodies, endosomes) (Schmidt et al., 1999; Farsad et al., 2001; Farsad and De Camilli, 2003; McMahon and Gallop, 2005; Gallop et al., 2006). Accordingly, in infected cells a certain proportion of the observed cytoplasmic endophilin-A2-positive vesicles colocalized with the trans-Golgi marker protein TGN-38 (Fig. 4r). Endophilin-A2 might aid to shape TGN-derived vesicles or tubules, into which cytoplasmic capsids bud for secondary envelopment (Mettenleiter, 2006). This is further indicated by the fact that SNX9 was also found in several Co-IPs only via pM50HA (Table 2). SNX9 is involved in similar processes as A-type endophilins (Worby and Dixon, 2002) and is required for protein trafficking to the TGN (McMahon and Gallop, 2005). In summary, both pM50 and endophilin-A2 might be functionally involved in membrane tubulation and vesiculation in the AC.
Altogether, we describe for the first time that endophilin-A2 is important for herpesvirus replication. The interaction with this protein seems to be a conserved feature among pUL34 family members. Both proteins induce membrane dynamics which may support the maturation of herpesvirus particles in the nucleus and in the cytoplasm.
Primary MEFs prepared from BALB/c mice (Brune et al., 2001) and TCMK-1 epithelial cells (ATCC CCL-139) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (GIBCO). HeLa cell (ATCC CCL-2) growth medium was composed of DMEM/F12 plus 10% fetal bovine serum and 2 mM L-glutamine. NIH/3T3 fibroblasts (ATCC CRL-1658) were cultivated in DMEM supplemented with 5% newborn calf serum (GIBCO). The bone marrow fibroblasts M2-10B4 (ATCC CRL-1972) were propagated in RPMI-1640 medium plus 10% fetal bovine serum. All media were supplemented with 1% penicillin-streptomycin and cells were cultured at 37°C in 5% CO2.
All MCMVs used in this study derived from BACs based on pSM3fr (Wagner et al., 1999) and were reconstituted by transfection of MEF cells as previously described (Menard et al., 2003). To remove the BAC cassette from the viral genome, the reconstituted viruses were passaged four times on M2-10B4 cells. The pSM3fr-derived virus was defined as wt MCMV. The HSV-1 strain F was used as wt HSV (ATCC VR-733). Stocks of wt and recombinant MCMV were prepared in large-scale format on M2-10B4 (3.5 × 108 cells) as described earlier (Brune et al., 2001). Virus titres were determined by standard plaque assay (Reddehase et al., 1985) on MEF for MCMV, or on Vero cells for HSV-1. Replication dynamics of recombinant MCMVs were determined on NIH/3T3 in 24-well plates by performing growth curves using multistep conditions. At 10–15% confluence cells were infected with recombinant and wt virus in parallel at a MOI of 0.1 for 1 h. Unbound viral particles were removed by phosphate-buffered saline (PBS) wash and fresh medium was added. At each day cell culture supernatants were harvested in triplicates and virus titres were determined by standard plaque assay (Reddehase et al., 1985).
In the MCMV genome the 3′ end of M50 and the gene-start of M49 feature a 26 bp overlap (Fig. S1B). By BAC mutagenesis both ORFs were separated, whereby the 26 bp overlap sequence was duplicated. However, to preserve the functionality of the M49 gene and concomitantly eliminate undesired homologous recombination between these duplications, the codon usage of the overlap sequence within the M49 gene was modified (Fig. S1B). Into the overlap sequence within the M50 ORF two point mutations were introduced, one of which changed the original M49 start codon ATG into a GTG preventing the initiation of translation from this site. A HA tag coding sequence was attached to the end of the M50 gene via a six-nucleotide linker sequence (Fig. S1B). The latter encodes for two serines which serve as flexible linker between the M50 protein and the C-terminal HA tag. For this cloning the oligo nucleotides ieHA1 and ieHA2 (Table S1) were inserted into pOriR6K-zeo-ie-M50 giving rise to pOriR6K-zeo-ie-M50HA. Primers M50mut-for and M50mut-rev (Table S1) were used to run a PCR on the pSM3fr wt BAC as template. This 3′ sequence modification of M50 was inserted by NotI and XhoI into pOriR6K-zeo-ie-M50HA resulting in pOriR6K-zeo-ie-M50HAmut. To alter the codon usage of the second 26 bp overlap copy, the M49 gene sequence was amplified by PCR on pSM3fr as template with the primers NEW-M49-for and NEW-M49-rev (Table S1) and the resulting amplicon was inserted into pOriR6K-zeo-ie-M50HAmut by digestion with ApaI resulting in pOriR6K-zeo-ie-M50HAmut-M49.
The genomic 5′ regulatory region of the M53 gene was amplified by primers NATfor and NATrev (Table S1). This PCR product was inserted into pOriR6K-zeo-ie-FLAGM53 (Popa et al., 2010) using KpnI and NheI. The resulting plasmid was designated pOriR6K-zeo-ie-NAT-FLAGM53. The modified M50 and M53 ORFs were subsequently introduced into pSM3fr by a modified version of the galK-based BAC recombineering (Fig. S1A) (Warming et al., 2005). For the replacement of the endogenous M50 and M53 coding sequences, the respective galK-kn targeting cassettes (Dölken et al., 2010) were amplified using both a pgalK-kn plasmid as template (Pubmed accession: FR832405) and primers which provide 50 bp homologies either to the M50 (H3-M50c-gk and H5-M50c-gk) or to the M53 (H5-FLAG-M53 and H3-FLAG-M53) locus (Table S1). The amplicons were introduced into pSM3fr by homologous recombination using induced SW102 carrying the wt BAC as described (Warming et al., 2005). Successful primary targeting conferred resistance to kanamycin alongside with the chloramphenicol resistance encoded by the BAC cassette. Double-resistant single colonies were picked and analysed for galK functionality as described. The M50HA and FLAGM53 coding sequences were released from pOriR6K-zeo-ie-M50HAmut-M49 and pOriR6K-zeo-ie-NAT-FLAGM53, respectively, by digestion with PvuII and Eco47III. The released fragments were introduced into the galK-kn-labelled intermediate BACs by a second round of homologous recombination. Here the correct recombinants were selected by loss of the galK gene by plating the transformants on M63 minimal plates containing 2% 2-deoxy-galactose. Successful replacement of the galK-kn cassette resulted in either pSM3fr-FLAGM53 or pSM3fr-M50HA (Fig. S1B and C). The M50 mutagenesis was repeated on the pSM3fr-FLAGM53 BAC exactly as described above resulting in the double tagged MCMV BAC pSM3fr-FLAGM53-M50HA. At every targeting level BAC sequences were verified by restriction pattern analysis and sequencing of the target sites.
The following primary antibodies and antisera were used in this study for immunoblotting and immunofluorescence: anti-pM50 and anti-pM53 polyclonal antisera (Muranyi et al., 2002), pUL34-specific antibody (provided by Susanne M. Bailer, Ludwig-Maximilians Universität München, Germany), GAPDH monoclonal antibody (clone EPR1977Y, Epitomics), CROMA-101 monoclonal antibody directed against the immediate-early 1 (ie1) protein of MCMV (supplied by Stipan Jonjic, University of Rijeka, Croatia), endophilin-A2 antibody (clone S20, Santa Cruz), TGN38 antibody (clone C15, Santa Cruz), anti-FLAG monoclonal antibody (M2, Sigma Aldrich), anti-HA-High Affinity monoclonal antibody (clone 3F10, Roche), beta-tubulin antibody (Santa Cruz). HRP-coupled anti-goat and anti-rabbit produced in donkey as well as anti-mouse and anti-rat raised in goat (Dianova) were used as secondary reagents for immunoblots. Polyclonal donkey anti-rabbit Alexa Fluor 488, anti-goat Alexa Fluor 488, anti-rat Alexa Fluor 488, anti-goat Alexa Fluor 555 as well as the DNA counterstains TO-PRO-3 iodide and YOYO-1 iodide were all purchased from Invitrogen and served for visualization in immunofluorescence.
Affinity purification of the tagged MCMV NEC proteins
For mass spectrometry 3 × 108 NIH/3T3, M2-10B4 or TCMK-1 cells were either mock-treated or infected with the recombinant viruses MCMV-M50HA, MCMV-FLAGM53, MCMV-M50HA-FLAGM53 or wt MCMV at a MOI of 1. Thirty-six hours post infection, cells were harvested using a cell scraper and washed twice with PBS. All following steps were carried out at 4°C or below in presence of protease inhibitors (Halt Protease Inhibitor Single-Use Cocktail, EDTA-free, Thermo Fisher Scientific). Cells were taken up in homogenization buffer (0.25 M sucrose, 25 mM KCl, 5 mM MgCl2, 20 mM Tricine-KOH, pH 7.8), disrupted by applying 12 strokes in a motor-driven Potter-Elvehjem homogenizer (Schuett-Biotec GmbH) with 150 μm clearance at 1700 r.p.m. and centrifuged at 800 g for 10 min. The crude pellet was washed another time in homogenization buffer, spun down and subsequently resuspended in 7 ml of lysis buffer (1% Triton X-100, 150–1000 mM NaCl, 5 mM EDTA, 20 mM Tris-HCl, pH 7.5) supplemented with 650 units of Benzonase nuclease (Novagen). The standard NaCl concentration for protein extraction in this study was 400 mM. After 2.5 h in a rolling incubator the nuclear lysate was centrifuged for 30 min at 22 000 g at 4°C to remove insoluble matter. Dependent on the protein tag 160 μl of HA (EZview Red Anti-HA Affinity Gel, Sigma Aldrich) or FLAG matrix (ANTI-FLAG M2 Affinity Agarose Gel, Sigma Aldrich) were added to each cleared lysate and incubated overnight on a rolling mixer. The material bound to the matrix was washed three times extensively with lysis buffer and two times with washing buffer (1% Triton X-100, 150 mM NaCl, 5 mM EDTA, 20 mM Tris-HCl, pH 7.5) and once with 20 mM Tris-HCl, pH 7.5. Captured protein complexes were eluted from the affinity beads under denaturing conditions with 100 μl of elution buffer (8 M urea, 50 mM Tris-HCl pH 8.0) for subsequent in-solution digestion or with 50 μl of SDS-PAGE loading buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 6 M urea, 5% β-mercaptoethanol, 0.01% bromophenolblue, 0.01% phenol red) for later in-gel digestion.
For immunoblotting analysis NIH/3T3 cells on 6 cm or 10 cm culture dishes were infected at a MOI of 1. Alternatively, M2-10B4 cells were transfected with the expression vectors pOriR6K-zeo-ie (as mock control), pOriR6K-zeo-ie-M50HA, pOriR6K-zeo-ie-FLAGM53 and/or pOriR6K-zeo-ie-M50HA-ΔP using Lipofectamine LTX with PLUS Reagent (Invitrogen) according to the manufacturer's instructions. Buffer volumes and materials for affinity purification were used as mentioned above and scaled down according to the respective number of cells. For Co-IP via endophilin-A2 Protein G Sepharose (GE Healthcare Life Sciences) coupled with anti-endophilin-A2 antibody (clone S20) was utilized. Bound protein complexes were released from the sepharose matrix by elution buffer (2% SDS, 10 mM β-mercaptoethanol, 100 mM Tris-HCl, pH 6.7).
Protein electrophoresis and immunoblotting
Protein samples were resuspended in SDS-PAGE loading buffer and were incubated for 30 min at 37°C to reduce membrane protein aggregation and then separated on 12% bis-Tris polyacrylamide gels. Samples determined for mass spectrometric analysis were allowed to migrate into the separating gel over a distance of only 1 cm. Lane sections comprising the protein content at large were cut and subjected to in-gel digestion. Gels were either silver-stained according to Blum et al. (1987) or used for protein transfer onto Hybond-P PVDF membranes (GE Healthcare). Blocking and antibody incubations were carried out in Tris-buffered saline (TBS-T) with 0.05% Tween-20 plus 5% dry milk for 1 h at room temperature (RT) or overnight at 4°C. Protein bands were visualized with ECL Plus Western Blotting Detection System (GE Healthcare).
Tryptic digest and mass spectrometry
The preparation of peptides for LC-MS/MS analysis was either performed as in-solution or in-gel digestion (Rosenfeld et al., 1992). For in-gel digestion gel slices (see above) were chopped and destained with water and 40 mM ammonium bicarbonate. The gel pieces were dehydrated using acetonitrile and subsequently submerged in DTT buffer (10 mM DTT, 40 mM ammonium bicarbonate) for 1 h to break disulfide bonds. For alkylation of free thiol groups the samples were incubated for another 30 min in iodoacetamide buffer (55 mM iodoacetamide, 40 mM ammonium bicarbonate) in the dark. After washing in 40 mM ammonium bicarbonate the gel pieces were again dehydrated using acetonitrile and soaked in 40 mM ammonium bicarbonate containing sequencing grade modified trypsin (Promega). The samples were incubated overnight at 37°C to digest the proteins and the resulting peptides were extracted by 5% formic acid. For in-solution digestion the proteins eluted with 8 M urea directly from affinity beads were reduced in 50 mM DTT for 1 h at RT and acylated in 10 mM iodoacetamide for 30 min in the dark. After quenching with 50 mM DTT the samples were diluted with 40 mM ammonium bicarbonate to final urea concentration of 1 M. The proteins were digested overnight at 37°C with trypsin under gentle rocking. Peptides derived from both preparation methods were dried overnight using a SpeedVac concentrator, resuspended in 15 μl of 0.1% formic acid, and analysed in a nano-ESI-LC-MS/MS. Individual samples were first separated on a C18 reversed phase column (75 μm i.d. × 15 cm, packed with C18 PepMap, 3 μm, 100 Å; LC Packings) via a linear acetonitrile gradient (UltiMate 3000 system, Dionex) before MS and MS/MS spectra were acquired on an Orbitrap mass spectrometer (Thermo Scientific). Recorded spectra were analysed via the Mascot™ Software (Matrix Science) using the NCBInr and a MCMV protein database. Scaffold (version Scaffold_3.3.1, Proteome Software, Portland, OR, USA) was used to validate MS/MS-based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 95% probability as specified by the Peptide Prophet algorithm (Keller et al., 2002). Protein identifications were accepted if they could be established at greater than 95% probability and contained at least two identified peptides. Proteins were assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.
Confocal laser scanning microscopy
For immunofluorescence experiments NIH/3T3 or HSV-1 cells were seeded into chamber slides (Lab-Tek II – Chamber Slide System, Nunc) to a confluence of 50%. After 24 h the cells were processed in two ways: (i) cells were directly infected at a MOI of 1 for 36 h in case of MCMV and 10 h in case of HSV-1; or (ii) cells were first transfected with a plasmid encoding for endophilin-A2–mCherry and were infected 24 h post transfection at a MOI of 1 for 36 h. The samples were rinsed in PBS and fixed in 4% paraformaldehyde in PBS for 15 min at 37°C. Next, cells were washed with PBS and incubated in FcR blocking reagent (Miltenyi Biotec) according to the manufacturer's instructions to minimize the background caused by virally encoded Fc receptors. After two additional washes in PBS, cells were incubated in blocking solution for 1 h at RT in PBS containing 0.3% Triton X-100 and 5% normal donkey serum (Millipore). Primary antibodies were diluted between 1:100 and 1:600 and secondary antibodies between 1:100 and 1:300 in blocking solution. Primary antibodies were applied overnight at 4°C and Alexa dye-coupled secondary antibodies and TO-PRO-3 iodide DNA stain for 1 h at RT in the dark. Subsequently, cells were extensively washed in high-salt PBS (400 mM NaCl) and PBS. Cells were mounted in confocal matrix (Micro-Tech-Lab). Images were acquired on Zeiss LSM 510 or Leica TCS SP2 confocal laser scanning microscopes. The BioImageXD analysis software (Kankaanpää et al., 2006) was used to process images and to analyse colocalization using the Costes algorithm (Costes et al., 2004) with automatic thresholding. The probability of colocalization is determined by calculating inverted P-values as described by Costes et al., whereby a value of 1.0 indicates true colocalization (Costes et al., 2004; Kankaanpää et al., 2006). Although often the Pearson's correlation coefficient is used to quantitatively estimate colocalization (Manders et al., 1992), the colocalization coefficients M1 and M2 were also computed (Manders et al., 1993). In this case only signals from the red and the green channels being above a calculated threshold value are considered as specific. The proportion of red and green signals within these specific areas is then defined as colocalization coefficients (Costes et al., 2004). Additionally, 2D scatterplots were computed in which the intensities of red and green signals in an image are plotted against each other.
Yeast-two-hybrid interaction analysis
For the cloning of Y2H vectors the M50 and endophilin-A2 gene sequences were amplified by nested PCR under the conditions described in Fossum et al. (2009) using pDONR207-M50wt or pDONR223-endophilin-A2 as template. All clones were verified by DNA sequencing. M50 constructs (M50 full-length, M501–170, M501–191, M50172–288, M50172–288Δ179–201) and the pUL34 (UL341–246) coding sequence (Fossum et al., 2009) were inserted into the bait vector pGBKT7-DEST, and endophilin-A2 constructs (endophilin-A2 full-length, endophilin-A21–300, endophilin-A2300–368) were cloned into the prey plasmid pGADT7-DEST by LR-clonase. To generate pGBKT7-M50172–288Δ179–201 lacking the proline-rich sequence, the region encoding residues 179–201 was excised from pGBKT7-M50172–288 by BamHI digestion, blunt-ended using DNA polymerase I, and religated after a second restriction digest with SmaI. For Y2H analysis haploid yeast strains AH109 and Y187 were transformed using 1 μg of prey (pGADT7-DEST) or bait (pGBKT7-DEST) plasmid DNA, respectively, and grown on SD medium (3% agar) lacking either leucine (-leu) or tryptophane (-trp). Prey- and bait-expressing yeast cells were mated on YPD medium for 5 h at RT. Subsequently, these cells were streaked on SD-leu-trp plates for diploid selection and grown for 2 days at 30°C. To detect reporter gene activation diploids were transferred to SD-leu-trp-his plates containing various amounts of 3 aminotriazole (3-AT).
RNA interference-directed knockdown
An endophilin-A2-specific SMARTpool, consisting of the four small interfering RNAs (siRNAs) 5′-GCAAGGCGCUGUACGACUU-3′, 5′-GAUCGCAGCUUCAUCGUCU-3′, 5′-ACAUCGAGGUCAAGCAGAA-3′ and 5′-GCAUGAUCCGCCACGGGAA-3′, was used to silence the gene expression of endophilin-A2 in HeLa cells. siRNAs 5′-UUCUCCGAACGUGUCACGUTT-3′ and 5′-TTAAGAGGCUUGCACAGUGCA-3′ were used as control (Qiagen). Endophilin-A2-specific and control siRNAs were transfected in duplicates into HeLa cells using Dharmafect 1 (DF1; Dharmacon, Thermo Fisher Scientific). Forty-eight hours after siRNA transfection the cells were infected with HSV-1 at a MOI of 0.1. Supernatants of infected cells were harvested various times after infection and the release of infectious particles was analysed by plaque assay on Vero cells. The siRNA SMARTpool was analysed for its effect on cell viability by counting uncoloured cells in triplicates in a 4% trypan blue solution.
We are grateful to Sigrid Seelmeir and Simone Boos for their excellent technical assistance. We thank Hannah Striebinger for her support during the Y2H assays and Harald Wodrich for providing the mCherry cloning vector. This work was supported by the DFG priority research programme SPP1175.