The large extracellular loop of CD63 interacts with gp41 of HIV-1 and is essential for establishing the virological synapse

Human immunodeficiency virus type 1 (HIV-1) persists lifelong in infected individuals and has evolved unique strategies in order to evade the immune system. One of these strategies is the direct cell-to-cell spread of HIV-1. The formation of a virological synapse (VS) between donor and target cell is important for this process. Tetraspanins are cellular proteins that are actively involved in the formation of a VS. However, the molecular mechanisms of recruiting host proteins for the cell–cell transfer of particles to the VS remains unclear. Our study has mapped the binding site for the transmembrane envelope protein gp41 of HIV-1 within the large extracellular loop (LEL) of CD63 and showed that this interaction occurs predominantly at the VS between T cells where viral particles are transferred. Mutations within the highly conserved CCG motif of the tetraspanin superfamily abrogated recruiting of expressed HIV-1 GFP fused Gag core protein and CD63 to the VS. This demonstrates the biological significance of CD63 for enhanced formation of a VS. Since cell–cell spread of HIV-1 is a major route of persistent infection, these results highlight the central role of CD63 as a member of the tetraspanin superfamily during HIV-1 infection and pathogenesis.

an ideal docking station for other transmembrane proteins 23 . How CD63, one of these tetraspanins, and other transmembrane proteins are recruited to the VS and the role they play in the subsequent cell-to-cell transfer of HIV particles is not well understood 24 . There are several lines of evidence supporting a role of tetraspanins in the replication process of HIV, mainly in very early or late events. In this respect, it has been demonstrated that HIV-1 particles contain selectively incorporated CD63 as well as other tetraspanins [25][26][27] . Furthermore, this incorporation negatively regulates the infectivity of HIV particles 28 . CD63 and other tetraspanins, such as CD81, can be found at VS sites, and it has been shown that tetraspanins regulate cell-cell transfer of HIV-1 15 . In this report, we provide evidence that CD63 interacts with the transmembrane envelope protein gp41 of HIV-1 in the context of a VS. Moreover, our data identify the LEL of CD63 as the major interaction domain with the viral gp41 protein, highlighting the role of this tetraspanin during cell-to-cell transfer of HIV-1 particles. The findings provide novel insights concerning the mechanism how CD63 might act as a regulatory protein for cell-to-cell transfer.

Results
Mapping of the CD63 interaction domain. Using a membrane-based split ubiquitin screen, we have identified full-length CD63 from a Jurkat T cell cDNA library as a binding partner of gp41 29 . This yeast-based screening method is suitable to detect novel interactions between membrane proteins in their native conformation in vivo 30,31 . To analyze the interaction in more detail and to reveal the CD63 domains involved, pair-wise interaction testing of CD63 domains (Fig. 1A) with gp41 were performed. As bait for these experiments, a previously described vector named pBT3-SUC-gp41 was used 29 . It consists of a cDNA fragment encoding the ectodomain and transmembrane region of gp41 (HIV-1 Env aa 534-709) cloned into the pBT3-SUC vector upstream in-frame with the C-terminal part of yeast ubiquitin (C ub ) and the LexA-VP16 reporter cassette. We used the prey vector pPR3-N for cloning CD63 variants containing two transmembrane domains and pPR3-SUC for single pass membrane variants. In an initial experiment using positive and negative control prey-vectors, the general functionality of the bait vector was confirmed, as well as the membrane localization of the expressed bait in a cis position by using pCCW-AIg5 bait reference control (Fig. 1B). Co-expression of the bait along with the control prey Alg5-N ub I resulted in a reconstitution of split-ubiquitin through the strong affinity of wild-type N-terminal part of yeast ubiquitin (N ub I) for C ub and the concurrent activation of reporter genes. In contrast, when the bait is co-expressed together with the negative control prey pDL2-Alg5 reporter genes were not activated because the mutated N ub G (I13G) moiety had no affinity for C ub 29,30,32 . After generating a different set of CD63 mutant prey vectors, experiments were performed to analyze the interaction of CD63 variants with the bait pBT3-SUC-gp41 in order to identify the CD63 interacting domain. The interaction analysis was monitored by yeast growth and quantification of β-galactosidase. Transformed yeast colonies containing the full-length CD63 sequence TM1-4 (aa 1-238), sequence TM 3-4 (aa 74-238) and single membrane spanning mutant variant containing the TM4LEL (aa 108-238) showed a strong growth on selective plates SD W-L-H-A supplemented with 3-Amino-1,2,4-triazole (3-AT) (Fig. 1C). 3-AT competitively inhibits the HIS3 gene product, while 3-AT supplementation reduces the background growth (leakiness of the HIS3 reporter) due to HIS3 gene activity in the absence of a protein-protein interaction (PPI) 33,34 . Autoactivation of each construct was tested against prey or bait empty vectors, and background yeast growth was inhibited at a 3-AT concentration of 1 mM (Fig. 1C). A quantitative determination of the β-galactosidase activity confirmed that the strongest interaction took place with CD63 derived prey variants containing the LEL (Fig. 1D). These results indicate that the LEL residues of CD63 are important for the interaction with gp41. Point mutations introduced in the CCG motif of CD63 abrogate the interaction with gp41. The LEL of CD63 contains six cysteines 35 . In order to analyze the significance of cysteine loops for the interaction with gp41, we generated cysteine mutations (C→A) in the CD63LEL of the prey plasmid pPR3-SUC-TM4LEL ( Fig. 2A). Co-transformation of the bait pBT3-SUC-gp41 and prey pPR3-SUC-TM4LEL containing the wildtype CD63LEL sequence revealed the strongest yeast growth on selective SD W-L-H-A plate supplemented with 3-AT 5 mM when comparable with CD63LEL containing generated cysteine mutation sites C169A, C170A, C170A and C191A (Fig. 2B). However, the LEL with cysteine mutation sites C145A and C146A showed a strong defect in yeast growth on selective SD W-L-H-A plates supplemented with 3-AT 1 mM (Fig. 2B). At this concentration of 3-AT, the autoactivity from control yeast transformations with the prey TM4LEL showed no background growth (Fig. 1B). The quantification of the β-galactosidase activity confirmed that the interaction with cysteine mutations at positions C145A and C146A within the evolutionary highly-conserved tetraspanin CCG motif is very weak compared to other mutation sites (Fig. 2C). In contrast, the introduced cysteine point mutation sites C169A and C170A in the LEL showed a beta-galactosidase activity that was comparable to the wild-type LEL sequence. Interestingly, the point mutations site C191A lead to a reduced β-galactosidase activity, but no reduced yeast growth was observed. This observation can be explained by the failed formation of a disulfide bridge 36 of the first cysteine at aa position 145 from the CCG motif with the cysteine at aa position 191 (Fig. 1A). These results indicate that the LEL of CD63 formed by the first cysteine in the LEL within the CCG motif is important for the interaction with gp41 and sufficient for confirmation of the specific PPI found in the split-ubiquitin assay.
Fluorescence resonance energy transfer (FRET) analysis confirmed the pivotal role of the CCG motif within the LEL of CD63 in the interaction with gp41. Yeast-based interaction data have to be confirmed by alternative methods in order to exclude false positive interaction partners. In a first approach, a FRET method was used and acceptor photobleaching experiments were performed. HEK293T cells were transfected with plasmids encoding cyan fluorescent protein (CFP) as the energy donor and yellow fluorescent protein (YFP) as the acceptor fused to gp41 or CD63 sequences, respectively (Fig. 3A). This approach involves mon- www.nature.com/scientificreports/ itoring of the donor (SP-gp41-CFP) fluorescence signals in the presence (pre-bleach) and absence (post-bleach) of the acceptor (CD63-YFP). The donor construct contains a signal sequence (SP) from Ig kappa-chain V-J2-C 37 .
The fluorescence intensity profiles of the FRET bleaching experiments were examined 24 h post-transfection. We observed that, for the FRET pair SP-gp41-CFP/CD63-YFP, the CFP fluorescence intensity increased after acceptor photobleaching comparable to the positive control CFP-YFP at several time points (Fig. 3B,C). Calculated FRET efficiencies obtained from acceptor photobleaching experiments are summarized in Fig. 3D. These data show that the FRET efficiency calculated for the interaction pair SP-gp41-CFP/CD63-YFP can be compared to the positive control where CFP and YFP are expressed as a single fusion protein. This observation revealed that CD63-YFP/SP-gp41-CFP contained a bimolecular interaction characteristic. In contrast, the expressed FRET pair SP-gp41-CFP/CD63 C145A,C146A -YFP showed up to 58% reduced FRET efficiency values after acceptor photobleaching of YFP compared to when used CD63-YFP as acceptor. However, the FRET efficiency calculated for the expressed FRET pair SP-gp41-CFP/CD63 C145A,C146A -YFP were different from the pseudo-FRET efficiencies obtained from separately expressed CFP and YFP (Fig. 3D). In summary, these data confirm that CD63 and gp41 interact directly, and the CCG motif is important for this PPI.
Proximity ligation assays (PLAs) confirmed the interaction between CD63 and gp41. To evaluate the PPI found between CD63 and gp41 conditions without fluorescently labeled proteins, we performed www.nature.com/scientificreports/ quantitative in situ PLAs on HEK293T. PLA experiments require a pair of primary antibodies raised in different species, which is used to recognize the PPI of interest. The detection of bright red PLA dot signals is only possible when both interacting proteins are in close proximity and the conjugated DNA with protein specific antibodies can form a DNA circle. This circle is then ligated and hybridized with probes 38,39 . Expression vectors were used to generate CD63 fused with FLAG (pCMV-CD63-FLAG) and gp41 fused with V5 (pcDNA-SP-gp41-V5) antibody-binding epitopes. Pair-wise interaction PLA tests between the expressed CD63-FLAG and SP-gp41-V5 proteins were performed in HEK293T cells, and primary antibodies against the fused tag and against an epitope within the expressed proteins were used. First, PLA experiments were performed employing indirect PLA (Fig. 4A, images B1-B4) using primary antibodies against FLAG and V5 epitopes in combination with Duolink secondary antibodies conjugated with Duolink DNA oligonucleotides mouse PLUS (+) and goat MINUS (−). Second, using direct PLA, CD63 and gp41 (2F5) epitope binding antibodies were conjugated with PLA Duolink DNA oligonucleotides PLUS (+) and MINUS (−) (Fig. 4A, images B5-B8). Transfected cells expressing CD63-FLAG and SP-gp41-V5 showed several bright PLA signals. In contrast, in controls, only a few weak and scattered PLA signals were observed. In order to demonstrate the precision of the assay, cellular Na + , K + -ATPase was used as a reference protein 40 . This indirect PLA control experiment showed an absence of specific PLA signals, which means that SP-gp41-V5 is not in close proximity with the Na + , K + -ATPase. This confirmed that the PLA experiments performed specifically detected an interaction between two proteins in a pool of expressed proteins in the same compartment (Fig. 4B9). Quantification revealed at least a tenfold difference when cells were transfected with pCMV-CD63-FLAG/pcDNA-SP-gp41-V5 in the number of PLA signals per cell (Fig. 4C). Western blot analysis confirmed protein expression driven by transfected vectors, as well endogenous protein expression in HEK293T cells used for PLA experiments (Fig. 4D). Further, when single recognition PLA using anti-V5 or anti-FLAG polyclonal goat antibodies were performed, several PLA signals were detected on the surface of HEK293T cells, confirming localization of expressed CD63-FLAG and SP-gp41-V5 on the cell membrane (Fig. 4E). This PLA variant uses only one polyclonal primary antibody in order to detect close proximity of different epitopes within one protein species. In addition, applying an indirect immunostaining technique, it was shown that expressed CD63-FLAG and SP-gp41-V5 were predominantly expressed in the plasma membrane (Fig. 4F). These results corroborate and extend our previous finding indicating expression of CD63-FLAG and SP-gp41-V5 on the surface where both proteins are in close proximity, qualifying for a PPI.
Proximity of CD63 and gp41 at the VS. To explore where the PPI between CD63 and gp41 described above is formed under physiological relevant conditions, we attempted to answer this question with direct PLA  www.nature.com/scientificreports/ proteins. Especially the expression of gp120 that undergo a trimeric association with gp41 (pre-fusion confirmation) will enable to reach close physiological conditions. This HIV-1 clone is based on the NL4-3 isolate but contains the R5-tropic envelope of the JR-FL isolate in addition to a GFP gene inserted between the matrix and the capsid sequence flanked by the viral protease cleavage sites SQNYPIVQ 41   www.nature.com/scientificreports/ cells were prepared for PLA experiments. PLA signals were observed particularly at intercellular cell-cell contact sites consistent with VS structures but not on other cell sites (Fig. 5A1). At these contact sites, GFP signals were abundant, indicating significant formation of a VS for viral transfer. In contrast, in the absence of cell-to-cell contacts, PLA signals were rarely detected (Fig. 5A2). In parallel experiments with Jurkat cells that were not transfected with the molecular HIV JR-FL Gag-iGFP clone, PLA signals were not observed (Fig. 5A3). The fluorescence intensity for PLA signal along the VS showing that PLA signal appeared with increased GFP intensity (Fig. 5B). Quantification of PLA signals indicated high levels of PLA signals when a VS was established between Jurkat cells (Fig. 5C). When HEK293T cells were used we did not observe comparable amount of PLA signals at cell-cell contacts where GFP signals are polarized (Fig. 5D). PLA signal distribution having similar characteristics with Jurkat cells that are not in contact with other cells where GFP signals are not polarized. Taken together, these results suggest that CD63 interacts with the gp41 at cellular contact sites of CD4+ cells enriched with HIV Gag-iGFP signals indicating the virological synapse.
The CCG motif within the LEL of CD63 is essential for establishing of a VS and recruiting HIV-1 JR-FL Gag-iGFP formations to the VS. CD63 expression vectors fused C-terminally with the fluorescent protein mCherry or YFP, as well as a CD63 version tagged with mCherry containing C145A,C146A mutations in the LEL (Fig. 6A), were generated in order to analyze colocalization and to monitor the dynamics of these proteins at the VS by confocal microscopy. The Western blot analysis of cell lysates showed that the expression levels of the wild-type and mutant CD63-mCherry proteins were similar and that generated expression vectors are functional (Fig. 6B). CD63-mCherry versions containing the LEL show higher molecular weight (70 kDa) than the backbone CD63-mCherry (55 kDa) and is indicative for glycosylation of the LEL 42 . CD63 construct version with a deleted LEL was used in this case to show the expression without the glycosylated LEL. Difference of the protein location between CD63-mCherry and CD63 C145A,C146A -mCherry at the www.nature.com/scientificreports/ plasma membrane was not observed (Fig. 6C). Confocal images suggest that the intracellular localization of CD63 C145A,C146A -mCherry is slightly higher than of CD63-YFP or CD63-mCherry. However, the calculation of the Pearson correlation coefficient (PCC) indicates high degree of correlation between fluorescence signals emit- www.nature.com/scientificreports/ ted from CD63-YFP/CD63 C145A,C146A -mCherry (mean R = 0.83) and are comparable to the correlation between CD63-YFP/CD6-mCherry (mean R = 0.87) that was used as a wildtype reference (Fig. 6D). These observations indicate that the constructed vector expressing CD63 C145A,C146A -mCherry is an applicable control allowing to study the importance of CD63LEL at the VS. We transfected therefore Jurkat cells with the molecular clone HIV-1 JR-FL Gag-iGFP and pCMV-CD63-mCherry or pCMV-CD63 C145A,C146A -mCherry. In fixed samples, several cell-cell contacts were analyzed by confocal microscopy and revealed that fluorescence signals emitted from Gag-iGFP, endogenous CD63 and wild-type CD63 fused with mCherry are mainly detectable at the VS between infected and uninfected Jurkat cells. In contrast, CD63 with introduced cysteine mutations C145A, C146A showed a randomized distribution on the surface of Jurkat cells (Fig. 6E). The intensities of the fluorescence peaks for endogenous CD63 and CD63-mCherry correlated with the accumulation of HIV-1 JR-FL Gag-iGFP formations at the VS (Fig. 6F). Interestingly, the fluorescence intensities of HIV-1 JR-FL Gag-iGFP suggest that there is an absence of CD63 C145A,C146A -mCherry polarization at the VS (Fig. 6E,F). Furthermore, the recruitment of CD63 protein and HIV-1 JR-FL Gag-iGFP on several Jurkat cell-cell contacts was quantified. For this, the relative increase of the protein level in one cell-cell contact for the green (GFP) and the red channel (mCherry) recruitment factor ω was calculated by employing fluorescence intensity data. The value ω = 1 means that the fluorescence intensity of expressed proteins is evenly distributed over the cell surface, and ω > 1 means that the proteins are highly concentrated at the VS. These results indicate no significant recruitment to the VS for CD63 C145A,C146A -mCherry and a slightly decreased recruitment of the CD63-mCherry protein in comparison to endogenous expressed CD63 protein (Fig. 6G). A further calculation of the ω value for the GFP signals showed that the expression of CD63 C145A,C146A -mCherry significantly reduced the recruitment of HIV-1 JR-FL Gag-iGFP to the VS (Fig. 6H).

Discussion
Yeast-based split ubiquitin screen and pair-wise interaction analysis identified CD63 as an interaction partner of gp41 29 , and this finding was confirmed by the correlative FRET-PLA technique 40 . Our previously performed PPI analysis was appropriate to address the question of whether CD63 and gp41 can interact, but it did not provide details on where or under which circumstances the identified interaction happens. In this study, the yeast-based split ubiquitin interaction assay was utilized to map the interaction domain. We found that the LEL of CD63 is responsible for interaction with gp41. This lateral interaction position was demonstrated by testing the gp41 bait with control prey vectors where N ub and C ub only fused together when they are in a lateral position (Fig. 1B-D). This is consistent with reports that tetraspanins interact with other membrane proteins in a lateral position and that those interactions are mainly mediated by the LEL 43,44 . Members of the tetraspanin superfamily contain conserved cysteine residues within the LEL, among them is the CCG motif, which is highly conserved 45,46 . The logical step was to introduce C→A point mutations in the LEL in order to cause a major local change in the CD63 loop structure. Constructs containing mutations are powerful tools to specifically disrupt the interaction between proteins 47 and testing of the altered LEL sequence showed the specificity of the employed yeast-based split ubiquitin interaction assay and provides additional evidence for positive interactions. When a prey encoding the wild-type CD63 and a set of mutated prey constructs coding for CD63TM4LEL were used, it was observed that the CCG motif was essential for the interaction with the gp41 bait. However, it is unlikely that cysteines of the CCG would directly be involved in the interaction between CD63 and gp41 through residue-residue contacts, but they are important for the structural and interaction integrity of the LEL. Due to the important role of disulfide bonds in folding and stabilizing the tertiary structure of proteins 48 , their removal most likely causes a misfolding of the LEL. Misfolded regions can be used to determine important interaction regions within a protein. For CD81, a member of the tetraspanin superfamily, it has been shown that cysteines within the LEL are important for stabilizing the protein folding 49 . The same observation was made with FRET experiments where the point mutations C145A and C146A led to a decrease in PPI strength (Fig. 3). Performing interaction experiments in HEK293T cells, a close proximity between recombinant expressed SP-gp41-V5 and CD63-FLAG proteins was detected by indirect and direct PLA (Fig. 4B,C). In addition, the PLA results demonstrate that the employed PLA technique is very specific and able to detect the interaction between CD63 and gp41 in a pool of several expressed proteins, such as the reference protein Na + , K + -ATPase 40 . Further, PLA signals were observed exclusively at cell-cell contacts between Jurkat cells expressing the molecular clone HIV JR-FL Gag-iGFP (Fig. 5A). These cell contacts are termed VSs and represent multimolecular complexes at the interface between infected and uninfected cells 1,50 . At this point, our investigations were focused on the protein recruiting and dynamics at the VS, and it was shown that the wild-type CD63-mCherry was recruited to the VS together with HIV-1 JR-FL Gag-iGFP formations. Previous analysis of the VS performed by electron microscopy showed that small GFP puncta as we detected are indicative for the presence of HIV-1 particles in the target cell 12 . The overlap of Gag-iGFP fluorescence signals correlated strongly with increased CD63-mCherry signals (Fig. 6E). In contrast, when CD63-mCherry with C→A mutations within the CCG motif (CD63 C145A,C146A -mCherry) was used, a random distribution of fluorescence signals on the cell surface was observed (Fig. 6E). Interestingly, expression of mutated CD63-mCherry significantly reduced the recruitment of HIV-1 Gag-iGFP formations to the cell-cell contacts. Therefore, it can be assumed that endogenously expressed CD63 is displaced from the VS site because of overexpressed CD63-mCherry, which is accumulating at the cell surface and therefore cannot play an active role during cell-cell transfer. These observations taken together highly indicate that the LEL of CD63 is essential for establishing and sustaining the VS. Currently, it remains a significant challenge to explain how CD63 binds in detail to gp41. We used expressed gp41 in a post-fusion state without gp120 as well with gp120. The interaction was found predominantly at the VS when gp120 expression was driven by the HIV Gag-iGFP molecular clone and this implies that this interaction needs particular conformation of gp41 otherwise we would detect PLA signals not only exclusive at the VS. This observation indicates that gp41 interacts at a post-fusion conformation www.nature.com/scientificreports/ with an accessible ectodomain to CD63LEL. When we utilized HEK293T cells that are CD4 negative we do not observe such high agglomeration of PLA signals at cell-cell contacts (Fig. 5D). As the current model for cellto-cell transfer describes interaction of Env with CD4 at the VS this would explain that the gp41 ectodomain is accessible for the LEL of CD63 when using Jurkat cells for PLA interaction experiments. In contrast, when gp120 was not expressed we were able to detect PLA signals randomized on the cell surface ( Fig. 4B image B5). After identification of the CD63 domain interacting with p41 we analyzed where und under which circumstances the this PPI takes place under physiological conditions. We observed an impact of CD63 cysteine mutations within the conserved CCG motif at the VS. Further studies should determine the consequences using LEL cysteine mutations within the CCG motif of CD63 for cell-to-cell-transfer of HIV-1 particles. We assume that CD63 with its LEL binds laterally to gp41 in order to circumvent the induction of Env-mediated fusion between infected and uninfected cells. This assumption is based on the report that CD63 and CD9 block HIV-1-induced cell-cell fusion (syncytia) at the transition from hemifusion to pore opening 16 . It was also demonstrated that the formation of syncytia was repressed by overexpression of CD63 and CD9 51 . During infection, the formation of syncytia is not supportive for a sustained infection, because fused T cells will end up, after 48-72 h, in an apoptotic process 52,53 .
This might indicate that HIV-1 has evolved strategies to inhibit the formation of multinucleated cells in order to survive in an infected host. We suppose that gp41 reduces its fusogenic property 54 by binding to the LEL of CD63. This model of co-operativity between CD63 and gp41 at the VS can be strengthened by additional findings, for example by the fact that when CD63 was overexpressed in a cell-cell infection assay the cell-cell transfer of HIV-1 was reduced. Reduced cell-cell transmission was also observed when CD9 was overexpressed, but not in the case of CD81 expression 15 . Binding of the transmembrane envelope protein gp41 to CD63 at the VS would mean that the viral particles are transmitted in a controlled way to acceptor cells. Experiments analyzing the detailed mechanisms how HIV-1 utilizes CD63 in order to inhibit syncytia formation will be extremely interesting and undoubtedly important to understand the viral pathogenicity and cell-cell transfer of the virus more precisely.
In summary, the domain of interaction with gp41 was mapped to the LEL domain of CD63. Using the PLA technique, it was shown that CD63 and gp41 predominantly interact at the VS. We further found that the large extracellular loop is important for establishing the virological synapse. The findings provide novel insights concerning the mechanism how CD63 might act as a regulatory protein for cell-to-cell transfer.

Material and methods
Plasmids, molecular cloning and site directed mutagenesis. For the construction of prey plasmids containing a set of different CD63 domains (the amino acid [AA] sequence length is given in the Supplementary Table 1), we used the vectors pPR3-N and pPR3-SUC (Dualsystems Biotech AG, Schlieren, Switzerland). The indicated primers (Supplementary Table 2) with variable Sfil restriction sites 29 at the 5′ and 3′ regions for directional cloning were used to amplify CD63 sequence from the template vector pPR3-N-CD63 29 by using 1 unit of PfuUltra HF polymerase (Agilent Technologies, Böblingen, Germany). The amplicons were digested with the SfiI restriction enzyme (New England Biolabs, Frankfurt, Germany) overnight at 50 °C. The digested and purified fragments were ligated with T4 DNA ligase (New England Biolabs, Frankfurt, Germany) into the SfiI digested vector pPR3-N or pPR3-SUC (Dualsystems Biotech AG). To generate bait plasmids pPR3-SUC-TM4LEL containing different C→A point mutation sites within the LEL of CD63, we used site directed mutagenesis. PCR reaction of a total volume of 50 μl contained 5-50 ng of the plasmid pPR3-SUC-TM4LEL, 50-125 μM of each mutation introducing primer (Supplementary Table 2), 0.25 mM dNTP mix and 2.5 units of PfuUltra HF polymerase (Agilent Technologies, Böblingen, Germany). The PCR cycles were initiated at 94 °C for 2 min, followed by 18 amplification cycles. Each PCR cycle was performed at 94 °C for 0.5 min, and the annealing temperature was set 5 °C below the T m of the primers for 0.5 min, 72 °C for 1 min/kb. The PCR cycles were finished with an extension step at 72 °C for 4 min. The PCR products were treated with 20 units of DpnI at 37 °C for 3 h. PCR reaction was then transformed into chemocompetent E. coli strain XL1-Blue (Stratagene, Heidelberg, Germany). The expression construct pCMV-CD63-mCherry was generated by using Pfu-Ultra HF polymerase and the primers mCherry_for/mCherry_rev for amplification of the mCherry sequence from the vector template pmCherry-N1 (Clontech, Heidelberg, Germany). The amplified fragment and vector pCMV2B-CD63-FLAG 40 were digested with XhoI and ApaI enzymes (New England Biolabs, Frankfurt, Germany), then purified and ligated with T4 DNA ligase (New England Biolabs). In the same way as described above, we introduced C→A point mutation sites in the vector pCMV-CD63-mCherry (pCMV-CD63-mCherry) and generated the vector pCMV-CD63 C145A,C146A -mCherry. The control vector pCMV-CD63ΔLEL-mCherry was generated by cloning parts of the CD63 sequence into the vector pCMVtag2B in two steps. First, the transmembrane domain (TM) 1-3 of CD63 was amplified from the template vector pPR3-N-CD63 40 using the primers CD63_TM1-3_for/CD63_TM1-3_for (aa 1-110 GenBank accession no. KF998086) and introduced into the vector pCMV-Tag2B through the restriction sites BamHI/PstI; the sequence TM4 of CD63 (aa 201-238, Gen-Bank accession no. KF998086) was amplified with the primers CD63_TM4_for/CD63_TM4_rev and introduced using the restriction sites EcoRV/HindIII. The mCherry reporter sequence was subcloned from the vector pCMV-CD63-mCherry and ligated using the restriction sites XhoI/ApaI in the vector pCMV-CD63ΔLEL. In the constructed vector pCMV-CD63ΔLEL, the mCherry tag was introduced by ligation of XhoI and ApaI digested mCherry fragment from the donor vector pCMV-CD63-mCherry. The vector pCMV-CD63 C145A,C146A -YFP used for FRET experiments was generated by exchange of the mCherry sequence with the YFP sequence from the vector pCMV-CD63-YFP-FLAG 40 using XhoI and ApaI restriction sites, all other vectors used were previously described 40,55 . All sequences of cloning sites were verified by Sanger sequencing.
Yeast-based split ubiquitin interaction assay. All yeast experiments (yeast transformation, bait functionality control assay and yeast growth assay) were performed as previously described 29  www.nature.com/scientificreports/ was determined by yeast growth and quantification of β-galactosidase. For this, several yeast colonies transformed with bait and prey were collected and grown to an OD 600 of 0.8-1, and all samples were diluted to an OD 600 of 0.6; 10 µl were dropped on selective plates and incubated overnight at 30 °C. The chlorophenol-red β-d-galactopyranoside (CPRG) assay, in order to quantify the β-galactosidase gene activity, was done according the yeast protocols handbook published by Clontech.
Cell preparation for indirect immunofluorescence. Transfected HEK293T cells were grown in 6 wells, the cell culture was removed, and cells were fixed in 2% paraformaldehyde (PFA), diluted in PBS for 20 min at room temperature, washed 2 times with PBS and detached by pipetting. The cell suspension was dropped on poly-l-lysine treated slides within a hydrophobic marked circle area created by a pap pen (Sigma Aldrich), dried and permeabilized with 0.2% saponin (Carl Roth, Karlsruhe, Germany) in PBS for 30 min and blocked in 3% BSA (Carl Roth) in PBS for 2 h. Cells were then incubated with primary goat anti-FLAG (NB600-344, Novus-Bio, Littleton, USA) and mouse anti-V5 (MCA1360, Serotec, Düsseldorf, Germany) antibodies. Bound primary antibodies were visualized by incubation with fluorescently conjugated secondary antibodies for green channel anti-goat Dylight 488 (Abcam, Cambridge, UK) or red channel anti-mouse Dylight 594 (Abcam). Transfected Jurkat cells were transferred from 6 wells into a reaction tube, washed one time and fixed in 2% PFA for 20 min. Further steps were proceeded, as for the HEK293T cells. Endogenous CD63 expression was detected by incubation with mouse anti-CD63 primary antibodies (1:1 mixture of MEM259, ab8219 and NK1/C3, ab1318, Abcam), and bound primary antibodies were visualized by incubation with anti-mouse Dylight 594 (Abcam). All immunostaining procedures were performed by using 5 µg/ml of primary and 1 µg/ml of secondary antibodies diluted in 5% BSA.
Proximity ligation assay (PLA). Transfected HEK293T or Jurkat cells were prepared for PLA analysis as previously described 40,55 . We used the following antibodies for PLA experiments: mouse anti-V5 (MCA1360, Serotec), anti-FLAG (NB600-344, NovusBio) and anti-Na + , K + -ATPase (ab98787, Abcam). Confocal laser scanning microscopy (cLSM). All images were acquired using a Zeiss LSM 780 confocal laser scanning inverted microscope (Carl Zeiss, Oberkochen, Germany) and a 63× oil immersion objective. Fluorescence signals were detected by Zeiss ZEN smart setup instrument settings for 488 nm and 594 nm dyes. PLA images were obtained and quantification of PLA signals was performed as previously described 40,55 . Pearson correlation coefficient (PCC) was calculated using cLSM images processed by using ImageJ Fiji's plugin Coloc2 for co-localization analysis.
FRET experiments and FRET efficiency calculation. HEK293T cells (1 × 10 5 ) were seeded in an IBIDI µ-slide 8 well glass bottom. After 24 h, cells were transfected with 1 µg plasmid DNA of each indicated FRET vector using MetafectenePro (Biontex, Munich, Germany) according to the manufacturer's instructions, and cells were incubated 24 h before FRET images were obtained. ZEN smart setup instrument settings were done for 405 nm and 514 nm dyes (multitrack channel mode, linewise switch), and the ZEN bleaching option was configured. We defined bleached ROI (red frames), unbleached ROI (green frames) and bleached background ROI (blue frames). Before starting the bleaching procedure, three pre-bleaching and three post-bleaching images were recorded. The 514 nm laser line was set at 100% laser power intensity at the bleaching time point. We calculated FRET efficiency with Zeiss ZEN FRET implemented software to obtain values for each ROI.
Calculation of the recruiting factor ω. The values of the recruiting factor ω were calculated from fluorescence intensity data by using the following Eq. (1): where I is the fluorescence intensity, r 1 and r 2 are intervals without cell-cell contact and r 3 and r 4 are intervals at the VS. FindGraph (UNIPHIZ Lab) software was used for calculation of the definite integrals.