Quantitative characterization of Tetraspanin 8 homointeractions in the plasma membrane

The spatial distribution of proteins in cell membranes is crucial for signal transduction, cell communication and membrane trafficking. Members of the Tetraspanin family organize functional protein clusters within the plasma membrane into so-called Tetraspanin-enriched microdomains (TEMs). Direct interactions between Tetraspanins are believed to be important for this organization. However, studies thus far have utilized mainly co-immunoprecipitation methods that cannot distinguish between direct and indirect, through common partners, interactions. Here we study Tetraspanin 8 homointeractions in living cells via quantitative fluorescence microscopy. We demonstrate that Tetraspanin 8 exists in a monomer-dimer equilibrium in the plasma membrane. Tetraspanin 8 dimerization is described by a high dissociation constant (Kd = 14700 ± 1100 Tspan/μm), one of the highest dissociation constants measured for membrane proteins in live cells. We propose that this high dissociation constant, and thus the short lifetime of the Tetraspanin 8 dimer, is critical for Tetraspanin 8 functioning as a master regulator of cell signaling.


Introduction
Tetraspanins (Tspans) are widely expressed membrane proteins of 200 -350 amino acids that cross the cell membrane four times.(1) They are involved in diverse cellular processes, such as cell proliferation, The functional units that Tspans form with their partner proteins in the membrane are called "Tetraspanin-enriched microdomains" (TEMs). TEMs were formerly described as large assemblies of multiple members of the Tspan family and their partner proteins. (5,14,(19)(20)(21)(22). Within these TEMs, different members of the Tspan family are proposed to be associating with each other, thereby bringing their partner proteins in close proximity to form large islands of functional signaling clusters.
Furthermore, Tspans are known to engage in a variety of Tspan-lipid interactions, e.g. with cholesterol or gangliosides, which are believed to be important for some biological processes, such as viral entry. (15)(16)(17)(18) Recent super-resolution and electron-microscopy studies have refined this picture, by revealing Tspan nanoclusters with only ~ 120nm diameters that contain a few copies of a single Tspan species. (23,24).
The unique ability of TEMs to control a broad range of biological functions makes them an interesting target to study. A classification of primary and secondary interactions has been proposed for Tspans. (5) Primary interactions involve a Tspan molecule and partner proteins and are believed to be mediated by contacts in the extracellular or intracellular domain of the Tspan protein. (6)(7)(8)(9) Common partner proteins include members of the integrin or immunoglobulin superfamilies. (9)(10)(11) Secondary interactions are weaker and occur between individual members of the Tspan family, and include Tspan homointeractions.(5) Tspan homointeractions have been proposed to be promiscuous, involving palmitoyl chains that modify cysteine residues in the juxtamembrane region. (12,13) Specific interactions between Tspans have also been suggested, but they are not well understood. (14) Here we sought to quantitatively characterize Tspan homointeractions, the 'molecular glue' of the Tspan web. Most of our current knowledge about Tspan interactions is based on non-quantitative studies using techniques such as co-immunoprecipitation, that do not directly report on interactions within the native membrane environment. (5,14,(19)(20)(21)(22) We use quantitative fluorescent microscopy and live cells to investigate homooligomerization of Tetraspanin 8 (Tspan8), the first Tspan molecule to be ever cloned (25). We determine both its oligomeric state, and the thermodynamics of association, which allows us to refine the current TEM model.
The plasmids were amplified in competent DH5α E. coli cells and purified using Qiagen's HiSpeed Plasmid Midi Kit (# 12643). All plasmids used in this work were verified by sequencing (Genewiz, Frederick, MD).

Cell culture and transfection
HEK293T cells, purchased from ATCC, were maintained at 37 °C, 5% CO 2 , in Dulbecco's modified eagle medium (Gibco, #31600034) supplemented with 10% fetal bovine serum (HyClone, #SH30070.03), 20 mM D-Glucose and 18 mM sodium bicarbonate. For imaging experiments, cells were seeded in 35 mm glass coverslip collagen coated Petri dishes (MatTek, P35GCOL-1.5-14-C) at a density of 2.5x10 5 cells per dish. Transient transfection was performed 24 hours after seeding using Lipofectamine 3000 (Invitrogen, #L3000008) according to the manufacturer's protocol. For FIF experiments, the cells were transfected with 1 -2 μg of DNA encoding Tspan8-eYFP. For FRET experiments, the cells were cotransfected with 5 -6 μg of mixtures of DNA encoding Tspan8-mTurquoise and Tspan8-eYFP or the mutants in varying ratios (1:3 -1:1). For the spectral unmixing of co-transfected cell images in FRET experiments, single-transfections were performed using 3 μg of plasmid DNA. 12 hours after transfection, the cells were rinsed twice with phenol-red free, serum free starvation media and were then serum starved for at least 12 hours.

Microscope imaging
Förster Resonance Energy Transfer (FRET) experiments were performed according to the published Fully Quantified Spectral Imaging-FRET (FSI-FRET) protocol. (26) Prior to imaging, the cells were subjected to osmotic stress with a pre-warmed (37 °C) 1:9 serum-free media:diH 2 O, 25 mM HEPES solution. This reversible osmotic swelling was necessary because the cell membrane is normally highly ruffled and its topology in microscope images is virtually unknown (26). The reversible osmotic stress eliminates these wrinkles and allows to convert effective 3D protein concentrations into 2D receptor concentrations in the plasma membrane (26). After swelling, the cells were allowed to stabilize for 10 min at room temperature. FSI-FRET imaging was performed using a two-photon microscope equipped with the OptiMis True Line Spectral Imaging system (Aurora Spectral Technologies, WI). The details of the Downloaded from http://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20210459/920646/bcj-2021-0459.pdf by guest on 18 September 2021 microscope have been described previously (27,28). For each cell, two microscope scans were acquired a direct acceptor emission scan using an excitation wavelength of λ=960 nm and a donor plus sensitized acceptor emission scan using an excitation wavelength of λ=840 nm. Each scan produced an image of 300x440 pixels, where every pixel contains a full fluorescence spectrum in the range of 420 -620 nm. Each imaging session did not exceed 2 hours.
Fluorescence Intensity Fluctuations Spectroscopy (FIF) experiments were performed with a TCS SP8 confocal microscope (Leica Biosystems, Wetzlar, Germany) equipped with a HyD hybrid detector. Images (1024x1024, 12bit) were acquired in photon counting mode with a scanning speed of 20 Hz and a 488 nm diode laser excitation. The emission spectra of eYFP were collected from 520-580nm.

Fluorescent protein standards
To convert measured intensities to concentration, the FSI method relies on calibrations with purified fluorescent protein solutions of known concentration.(26) Soluble monomeric eYFP and mTurquoise fluorescent proteins with an N-terminal 6x His tag were expressed and purified as described in (29). The

Fluorescence Intensity Fluctuations Spectroscopy (FIF) analysis
The basolateral membrane of HEK 293T cells was imaged with a Leica TCS SP8 confocal microscope. The images were analyzed using the FIF software described in (33). The basolateral membrane outlined by the researcher was segmented into 15x15 pixel segments by a moving squares method. The resulting segments were analyzed using the brightness and concentration calculator in the FIF software (33). The molecular brightness ε is related to the variance of fluorescence σ across segments and the average where σ is the variance of the noise of the detector. For a photon-counting detector, the brightness is

Calculation of dissociation constants
Tspan dimerization was described by a two-state thermodynamic model describing equilibrium between Tspans in the monomeric (m) and dimeric (d) states: Thus, the equilibrium dissociation constant, K d , is given by Since the total Tspan concentration is given by Using equations (5) and (6), one can calculate the fraction of dimeric membrane protein f d as a function of K d and [T]: The fraction of dimeric proteins can be linked to the FRET efficiency due to specific dimerization, E D , according to (26,30,31) = Ẽ x A in eq. 8 is the fraction of acceptors, which is measured for each cell (since both the donor and the acceptor concentration are measured in each cell). x A accounts for the fact that each cell has different expressions of donors and acceptors, and FRET depends on their relative expressions. It also accounts for the fact that only some of the dimers have both a donor and an acceptor and therefore contribute exclusively to the measured FRET efficiency.
Ẽ in eq. 8 is the "intrinsic FRET", the FRET efficiency in a dimer containing a donor and an acceptor fluorophore. It is a structural parameter, which depends only on the separation, orientation, and dynamics of the two fluorescent proteins in the dimeric complex and is constant for each dimer. Under the assumption that the fluorescent proteins can rotate freely, the intrinsic FRET E can be related to the inverse sixth power of the distance between the donor and acceptor fluorescent proteins d.
The assumption of free rotation can be justified since the fluorescent proteins are attached to the Tspan molecule via a flexible 15 amino acid (GGS) 5 linker. However, this assumption might not hold in all cases.
The measured FRET efficiency has contributions due to 1) specific dimerization of Tspans, E D , and 2) stochastic FRET. Stochastic FRET, which is significant within the two-dimensional plasma membrane, occurs when a donor and an acceptor approach each other by chance within close proximity (<100 Å).

Tspan8 homooligomerization within the plasma membrane
We  needed for quantification of the interactions. The high variability in receptor expression levels (Figure 2a and b), as well as in donor:acceptor ratios (Figure 2b), is due to transient transfection of HEK293T cells and is the strength of the FSI method to ultimately produce the best-fit binding curves. (26) Next, we sought to interpret the measured FRET binding curves and determine association constants by fitting the FRET data to a protein association model. However, the oligomer size must be known in order to perform the fit. In the literature, oligomer sizes ranging from dimers (hypothesized to occur via specific interactions) to higher-ordered oligomers (involving promiscuous interactions) have been proposed. (14,17,(35)(36)(37)(38) A limitation of FRET is that the oligomer size is not measured directly but is inferred from fits to different association models. Therefore, we performed fluorescence intensity fluctuation (FIF) experiments to directly assess the oligomer size (Figure 1 c). (33) In these experiments, the basolateral membrane of HEK293T cells expressing Tspan8 tagged with eYFP was imaged ( Figure   1c,i). The molecular brightness, which scales with the oligomer size, was calculated in small sections of the plasma membrane by constructing histograms of fluorescence intensities in these sections ( Figure   1c,ii). The brightness is calculated in each section as the ratio of the variance and average intensity. would be expected to 1) have a maximum position of n times the maximum position of LAT and 2) have the whole molecular brightness distribution shifted to values higher than E-cadherin. As we do not observe this, we conclude that Tspan8 does not form higher-order (n>2) oligomers in the plasma membrane but exists in either a monomeric or a dimeric form.
Based on this result, we analyzed the FRET data in Figure 2 a and b with a dimerization model (equation

3). Dimer formation is characterized by i) the dissociation constant K d and consequently the dimeric
fraction as a function of Tspan8 concentration and ii) the structural parameter E (or "intrinsic FRET"), which depends on the positioning of the fluorescent proteins in the Tspan8 dimers. However, the measured FRET data have two contributions: i) FRET due to specific Tspan8 dimerization and ii) FRET as a result of random approach (within 100 Å) of two Tspan8 molecules in the confined two-dimensional plasma membrane. The latter is also known as "stochastic" or "proximity" FRET and is well understood and can be corrected for. (32,42,43) Its contribution to the measured FRET efficiency is shown in Figure   2a with the solid line. The measured FRET efficiencies lie above the proximity FRET line, indicating that Tspan8 is self-associating in the plasma membrane, confirming the FIF results. Following a verified twostep fitting procedure, as outlined briefly in Materials and Methods, we fit the data to different dimerization models, varying K d and E .(32) The model with the smallest mean squared error best represents the experimental data. For Tspan8 homodimerization we find a dissociation constant of 14700 ± 1100 Tspan/μm 2 and an E -value of 0.60 ± 0.03. Another way to characterize dimer formation is through its dimerization curve (dimeric fraction as a function of Tspan8 concentration). The Tspan8 dimerization curve depends on K d and the Tspan8 concentration, and is calculated using equation 7. The best fit dimerization curve (solid blue line) to the data (blue circles) is shown in Figure 4a. The dashed lines indicate the 68% confidence intervals of the fit.
Tspans are posttranslationally modified via glycosylation of residues in the extracellular region and palmitoylation of intracellular juxtamembrane cysteine residues (13,44,45). It has been suggested that Tspan homointeractions are affected by the palmitoylation (12,46). To explore the significance of palmitoylation in the interactions that we measured, we mutated all juxtamembrane cysteine residues (C7, C74, C75, C83, and C232) to alanine (Tspan8-Δpalmitoylation) to completely block palmitoylation.
The FRET data as a function of acceptor concentration and the FIF data are shown in Figure 2c-d and Tspan homooligomers are believed to be fundamental building blocks of the Tspan web. So far, Tspan interactions have only been characterized by the strength of the detergent that they can withstand in co-immunoprecipitation experiments. (5,12,15,38) In this study, we have quantitatively investigated homoassociation of Tspan8 in the plasma membrane. The results are summarized in Table 1 Specific Tspan homodimerization could be the molecular mechanism that underlies Tspan function.
Tspan8 has one of the highest dissociation constants quantitatively measured in living cells, and thus it exhibits one of the weakest self-interaction propensity among membrane proteins. (26,40,(47)(48)(49)(50)(51)(52) Association rates are believed to be diffusion limited, and thus similar for all proteins, while dissociation rates reflect the strength of the interactions.(61) Thus, the measured high dissociation constant implies high dissociation rates and therefore short lifetimes of Tspan dimers. A short dimer lifetime may be an indispensable property of a protein that regulates signaling. Indeed, master organizers and regulators of the cell membrane, like Tspans, likely need to constantly adjust to cellular signals by remodeling their microenvironment.
It is generally believed that the strength of heterointeractions between Tspans and other membrane proteins are stronger that Tspan homointeractions (5,14). The weak homodimerization that we  Figure 5c). The latter is consistent with reports that TEMs contain not just one, but multiple Tspan species (63,64). It is also possible that the molecules in the patches bind more than one Tspan, while all Tspans are capable of homodimerization (Figure 5b)      . The fast dissociation rates allow Tspan8 to bring different partner proteins in proximity, but only small TEMs can form. b) Tspan8 partner protein (red and teal) can bind more than one Tspan8 molecule (blue). The formation of large TEMs is possible. c) Tspan 8 (blue) can associate with other Tspans (green), but only one Tspan can bind to its partner protein (red, grey). Large TEMs can form. Within TEMs Tspan8 explores a monomerdimer equilibrium with fast dissociation rates enabling dynamic remodeling of the TEMs.