The specificity of homomeric clustering of CD81 is mediated by its δ‐loop

Tetraspanins are cell membrane‐scaffolding proteins interacting with one another and a repertoire of interaction partners. Through these interactions, they form extended molecular networks as tetraspanin webs or tetraspanin‐enriched microdomains. Microscopic data suggest that these networks contain tetraspanin clusters, with poor overlap between clusters formed by different tetraspanins. Here, we investigate the possibility of targeting tetraspanins CD9 or CD151 to clusters formed by the tetraspanin CD81. We find that the δ‐loop from the large extracellular domain of CD81 is sufficient for targeting of CD9/CD151 to CD81 clusters. Moreover, in a pull‐down assay, CD9 coprecipitates more CD81 when it carries the CD81 δ‐loop. In conclusion, the information for forming homomeric CD81 clusters is encoded in the δ‐loop.

Tetraspanins are cell membrane-scaffolding proteins interacting with one another and a repertoire of interaction partners. Through these interactions, they form extended molecular networks as tetraspanin webs or tetraspanin-enriched microdomains. Microscopic data suggest that these networks contain tetraspanin clusters, with poor overlap between clusters formed by different tetraspanins. Here, we investigate the possibility of targeting tetraspanins CD9 or CD151 to clusters formed by the tetraspanin CD81. We find that the d-loop from the large extracellular domain of CD81 is sufficient for targeting of CD9/CD151 to CD81 clusters. Moreover, in a pull-down assay, CD9 coprecipitates more CD81 when it carries the CD81 d-loop. In conclusion, the information for forming homomeric CD81 clusters is encoded in the d-loop.
Tetraspanins are a family of small membrane proteins expressed in animals, plants and fungi, with 33 members in humans. They act as scaffolding proteins in the cell membrane, forming large interaction networks with one another and other molecules. Tetraspanin interactions partners are diverse, including integrins/ other adhesion molecules, members of the immunoglobulin superfamily, signalling receptors and gangliosides [1,2]. This wide repertoire of factors explains why tetraspanins play so many roles in physiological as well as in pathophysiological processes, as cell proliferation, signal-transduction, vesicle trafficking, cell-cell fusion, adhesion, spreading, migration, cancer, infectious diseases and host-pathogen interactions [3][4][5][6][7]. They are also considered to be 'master organizers' of the plasma membrane.
Commonly, tetraspanin interactions are identified and classified into primary and secondary interactions employing immunoprecipitation with detergents of variable strength. Robust and direct interactions formed between a tetraspanin and a specific partner are primary complexes, while interactions only stable under milder solubilization conditions are secondary, as the many tetraspanin-tetraspanin associations that crosslink primary complexes to large tetraspanin webs [8]. Secondary interactions are further stabilized by a third level of weak and indirect interactions [6]. These are based on hydrophobic palmitates attached through palmitoylation to tetraspanin residues close to the inner plasmalemmal leaflet [9,10].
Physical dimensions of TEMs studied by electron microscopy reveal that CD63-and CD9-enriched TEMs occupy an area of 0.2 lm 2 , although size and shape vary considerably [11]. The majority of CD63 and CD9 form distal clusters. Super-resolution light microscopy, focussing on the tetraspanins CD37, CD53, CD81 and CD82, determined a tetraspanin cluster size in the range of 100-150 nm [12,13]. As CD63-/CD9-clusters in electron microscopy, also in light microscopy individual members of the tetraspanin family form clusters largely devoid from other tetraspanins [13]. Tetraspanin cluster segregation in microscopy likely reflects evidence from biochemical cross-linking showing a higher level of tetraspanin homo-dimers/higher homo-oligomers when compared to hetero-dimers [14].
There are also parallels between biochemistry and microscopy regarding the interaction strength between two partners. For instance, CD81 and its primary interaction partner EWI-2 overlap stronger with each other than CD81 with its secondary interaction partner CD9 [15]. Moreover, CD53 and CD81 are in closer proximity to their primary interaction partners than to other tetraspanins [13].
Tetraspanins share a common structure composed of four transmembrane domains, small intracellular segments and two extracellular loops. The structure of the small extracellular loop is unknown, but the large extracellular loop (LEL) can be subdivided into five helical segments [16,17]. The structure and the conformational flexibility of the LEL of CD81 have been studied [17][18][19]. Based on the crystal structure of the CD81 large extracellular loop it was proposed that CD81 dimerizes via a conserved hydrophobic interface [17]. However, due to the antiparallel orientation of the two dimerizing proteins this mechanism can only account for dimers forming on opposed or extremely curved membranes. Deleting subdomains within the LEL it was shown that a small segment comprising the d-loop is required for directing CD81 molecules into CD81 clusters, possibly involving a CD81 dimerization step [15]. However, it is not clear whether the d-loop is required for cluster stabilization or mediates the specificity in CD81 oligomerization.
Here, we report that the d-loop determines the specificity in clustering showing that the tetraspanins CD9 or CD151 are directed into CD81 clusters when they carry the CD81 d-loop.

Constructs
Plasmids encoding for CD81-GFP and CD9-RFP were previously described ( [15]; GFP is a monomeric variant from EGFP; RFP is a monomeric variant carrying at its C terminus a myc-tag). CD9 Chim -RFP is based on CD9-RFP and was generated by fusion PCR, substituting the CD9 d-loop located between cysteine residues 167 and 181 with the CD81 d-loop region (between cysteine residues 175-190). For the CD151-RFP construct, the CD151 sequence was amplified from a plasmid described in Ref. [20], C-terminally fused to RFP (see above), and cloned into the pEGFP_C1 vector (6084-1; Clonetech, Mountain View, CA, USA). CD151 Chim -RFP was generated from CD151-RFP via fusion PCR, by substituting the CD151 d-loop between the cysteine residues 192 and 208 with the CD81 d-loop (see above). All constructs were verified by sequencing, using as reference the respective homo sapiens sequences (CD81, NP_004347.1; CD9, NP_001760.1; CD151, NP_001034579.1).

Cell culture and membrane sheets
Jurkat E6.1 and HepG2 cells were maintained and transfected as described previously [15]. For each construct, 30 lg of plasmids was used for transfections. From transfected Jurkat E6.1 or HepG2 cells membrane sheets were generated in ice-cold sonication buffer after 1 day or 2 days, respectively, as previously described [15].

Microscopy
Confocal scanning microscopy was done on an Olympus FluorView 1000 microscope (Olympus, Tokyo, Japan) used previously [15]. In brief, intact cells coexpressing CD81-GFP in combination with CD9-RFP or CD9 Chim -RFP were adhered on poly-L-lysine-coated glass coverslips as described previously [15]. Cells were directly fixed for 30 min at RT with 4% paraformaldehyde (PFA) in PBS (137 mM NaCl, 2.7 mM KCl, 8.1 mM Na 2 HPO 4 , pH 7.4), treated with NH 4 Cl in PBS for quenching of PFA, and imaged in PBS. For imaging, the glass-adhered plasma membrane was positioned into the focal plane. GFP-, RFPand DIC images were recorded with a pixel size of 103 nm, scanning a 300 pixel 9 300 pixel area. GFP and RFP were excited by 488 nm and 543 nm lasers, respectively.
Epifluorescence microscopy was performed using the same microscopic equipment and settings described previously [15]. In brief, membrane sheets from double-transfected cells were generated by a 100-ms sonication pulse, fixed with PFA and treated with 50 mM NH 4 Cl in PBS for quenching of PFA. Imaging was performed in PBS containing TMA-DPH [1-(4-tri-methyl-ammonium-phenyl)-6-phenyl-1,3,5-hexatriene-p-toluenesulfonate (T204; Thermofisher, Waltham, MA, USA)] for visualizing the membranes. In addition we added tetraspeck beads (Invitrogen, Carlsbad, CA, USA) enabling to correct for lateral shifts. Images were recorded in the green (CD81-GFP), red (CD9-RFP, CD9 Chim -RFP, CD151-RFP or CD151 Chim -RFP) and blue (TMA-DPH) channels. In the green channel, exposure times for Jurkat T and HepG2 membrane sheets were 1s and 100 ms, in the red channel exposure times were 2s and 1s, respectively.
The similarity between the signal distribution in the green and the red channels was quantified by calculation of the Pearson correlation coefficient (PCC) as previously described [15]. In brief, to avoid any bias in analysis, the regions of interest (ROIs) were positioned in the TMA-DPH or DIC image that illustrate the membrane or cellular shape (without knowing the distribution of the green and the red signals), respectively. After transferring the ROIs to the green and red channels, it was checked whether any highly fluorescent structures as bright artefacts, tetraspeck beads, organelles or membrane remnants are located in the ROIs. If so, the ROI was slightly moved to a different location or if this was not possible, the membrane sheet or cell was excluded from the analysis. Images are shown at arbitrary scalings.  ChromoTek, Planegg-Martinsried, Germany). The solutions were incubated under rotation at 4°C for 1 h. Afterwards, beads were collected by a quick centrifugation step (2 min at 2500 g at 4°C) and washed twice with HEPES buffer supplemented with protease inhibitor cocktail and 10 lM PMSF. The pulled down proteins were then subjected to western blot analysis under nonreducing conditions. We subsequently immunoblotted for CD81 and RFP-labelled constructs using mouse monoclonal anti-CD81 (1.3.3.22) antibody overnight at 4°C and rabbit polyclonal anti-RFP antibody for 1 h at RT. Detection of the respective primary antibodies was performed using HRP-coupled secondary antibodies that were visualized on autoradiography films using a Luminol chemiluminescence kit (sc-2048; Santa Cruz). Bands were quantified on scanned radioactive films. For quantification of the coimmunoprecipitated CD81, background values obtained from the Lifeact-RFP control were subtracted. Western blot parts from the same membrane are shown at the same scaling.

Results and Discussion
We previously set-up an assay for analysing the microscopic overlap of tetraspanin constructs in the cell membrane of Jurkat T cells [15]. In brief, Jurkat T cells coexpressing GFP-and RFP-labelled tetraspanins were adhered to a glass-surface and then exposed to a brief ultrasound pulse, which removes the upper cellular part, leaving behind the intact glass-adhered native membrane. In this preparation, the mobility behaviours of membrane proteins has been shown to be unchanged when compared to intact cells [22,23]. As the preparation is two-dimensional, imaging can be performed by simple epi-fluorescence microscopy with a CCD camera at high signal-to-noise ratio, thus avoiding confocal microscopy-based optical sectioning techniques (in combination with a photomultiplier) with a reduced signal-to-noise ratio. This becomes important when less potent fluorophores as RFP are recorded.
Quantifying the overlap between the pairs CD81-RFP/CD81-GFP and CD9-RFP/CD81-GFP yielded Pearson correlation coefficient (PCC) values of 0.6 and 0.35, respectively [15]. Due to instrument noise and other factors, even double-tagged proteins do not yield a PCC even close to 'one' (indicating identical images), but, for example, 0.63 [24]. On the other hand it should be considered, that due to the meso-scale organization of protein clusters in multiprotein assemblies [25] also unrelated proteins show a PCC higher than 'zero', typically in the range below 0.1. Within the theoretical dynamic range between % 0.1 and % 0.6, the above PCC values for CD81-RFP/CD81-GFP and CD9-RFP/CD81-GFP suggest that the overlap between these constructs is perfect and weak, respectively. Comparing CD81-RFP to either CD81-GFP or CD81-Dd-GFP (here a small 11 aa segment from the LEL was removed) we found that overlap is lost, suggesting targeting of CD81 into its clusters requires this region [15]. However, successful targeting can have several reasons, for example, we cannot differentiate between a stabilizing effect or a preceding interaction mediating the specificity of the interaction. For clarification, we tested whether a 'wrong' tetraspanin could be targeted to CD81 clusters solely by carrying the CD81s' d-loop. Using the tetraspanins CD9 and CD151, chimeras were generated replacing their dloops by the one of CD81 (Fig. 1), in the following termed CD9 Chim and CD151 Chim . We first compared CD81/CD9 to CD81/CD9 Chim asking whether we could increase the low overlap between CD81 and CD9 by using the CD9 Chim construct. We performed the experiment both in paraformaldehyde fixed cells (Fig. 2) and membrane sheets (Fig. 3). As previously observed [15] the PCC for CD81/CD9 was low both in cells (Fig. 2) and membrane sheets (Fig. 3), but when using CD9 Chim it increases by % 0.1 in cells and more than 0.2 in membrane sheets. The PCCs in cells are lower due to the lower signal-to-noise ratio imaging in confocal microscopy that can be appreciated when comparing the images. The confocal GFP-channel resembles a noisy and poorly resolved cluster pattern which is better detected with the CCD camera on membrane sheets (compare CD81-GFP signal shown in insets from Fig. 2 with the magnified views of CD81-GFP in Fig. 3). In case of the less potent RFP fluorophore, in the confocal microscope the signal does not even display discrete clusters, but only noisy areas with increased or decreased fluorescence intensity (see insets in the RFP channel in Fig. 2). Therefore, on intact cells it is not possible to obtain high PCC values by confocal microscopy. Nevertheless, the experiment shows that the improved targeting of CD9 Chim into CD81 clusters is not a phenomenon associated with membrane sheets.
The membrane sheet signal intensities vary with the expression levels of the GFP and the RFP constructs and these expression levels also influence the signal-tonoise ratio. To exclude that the lower PCC for CD81/ CD9 is due to the imaging of dimmer membrane sheets, we compared the expression levels of CD81-GFP/CD9-RFP and CD81-GFP/CD9 Chim -RFP from all membrane sheets included into the analysis. As shown in Fig. 3D, the expression levels were in the same range. Independent from the type of fluorescent label, the PCCs tended to be higher for the pair CD81-GFP/ CD9 Chim -RFP. Hence, the differences in the PCCs are not due to differences in expression levels.
We also aimed for confirming the d-loop effect applying IP after mild solubilization with CHAPS, which preserves the interactions between different individual tetraspanins. Jurkat T cells, cells were lysed in 1% CHAPS and after a short centrifugation step the RFP tag was precipitated and the pulled down endogenous CD81 was quantified. Compared to CD9, CD9 Chim was more efficient in pulling down endogenous CD81 (Fig. 4), albeit the d-loop-mediated increase in immunoprecipitation is not as convincing as the produced increase in microscopic overlap.
When comparing the pairs CD81/CD151 and CD81/ CD151 Chim , a similarly strong gain in overlap was observed, although starting from an overall lower level (Fig. 5).
We next turned to a different cellular system, namely HepG2 cells. These cells differ from Jurkat T cells as they have no endogenous CD81, and they are adherent cells, making the adhesion step prior to sonication redundant. We analysed the same construct pairs as in Jurkat T cells. As shown in Figs 6 and 7, the same basic finding is obtained, although the d-loop-mediated gain in overlap was below 0.2, which is slightly less when compared to Jurkat T cells. This might be due to the stronger expression levels of the constructs in HepG2 cells leading to less distinct cluster patterns (compare Figs 3 and 5 to Figs 6 and 7).
In conclusion, analysis of microscopic overlap in two different cell systems shows that CD9 and CD151 do not overlap to a large extent with CD81, which is in line with earlier reports on the segregation of clusters formed by different tetraspanins [13,15]. Importantly, overlap increases strongly when CD9 and CD151 carry the d-loop of CD81.
The data show the d-loop plays a pivotal role in homophilic clustering and carries the information for targeting CD81 into CD81 clusters. As previously suggested by the cluster phase model, clusters are composed of transiently forming binary complexes (in this case homo-dimers and/or CD81-primary partners) that exchange molecules with each other [15]. In this model, the specificity of the d-loop would allow only CD81 molecules to enter the cluster phase that likely also contains other, but not tetraspanin, molecules. However, the model would be also in line with a cluster composed exclusively of CD81.
It should be noted that the d-loop is not alone responsible for cluster building and maintenance, as de-palmitoylation of tetraspanins leads to smaller clusters (from 140 to 97 nm; [12]) and a small diminishment in CD81 cluster targeting efficiency [15].
The d-loop is located within the variable domain of the LEL, which has been suggested to be important for partner selection and initial contact establishment [18,19,26,27]. In line with the tetraspanin-partner pair model [1], a CD81-partner complex would be connected via a specific d-loop-mediated dimerization step to another CD81-partner complex, of identical or different composition. Alternatively, an already formed CD81 dimer may bind to primary partners. As has been pointed out, by this mechanism, tetraspanins, or more precisely functional dimers [14], have a specific role in partner selection and a nonspecific role in regulating the stoichiomertry of the interactions that define the tetraspanin web. This model is compatible with the cluster phase model outlined above, with the only exception that in the cluster phase model at any moment only binary complexes are present, that in the next moment may dissociate in order to form new combinations of binary complexes.