Affinity biosensors using recombinant native membrane proteins displayed on exosomes: application to botulinum neurotoxin B receptor

The development of simple molecular assays with membrane protein receptors in a native conformation still represents a challenging task. Exosomes are extracellular vesicles which, due to their stability and small size, are suited for analysis in various assay formats. Here, we describe a novel approach to sort recombinant fully native and functional membrane proteins to exosomes using a targeting peptide. Specific binding of high affinity ligands to the potassium channel Kv1.2, the G-protein coupled receptor CXCR4, and the botulinum neurotoxin type B (BoNT/B) receptor, indicated their correct assembly and outside out orientation in exosomes. We then developed, using a label-free optical biosensor, a new method to determine the kinetic constants of BoNT/B holotoxin binding to its receptor synaptotagmin2/GT1b ganglioside (kon = 2.3 ×105 M−1.s−1, koff = 1.3 10−4 s−1), yielding an affinity constant (KD = 0.6 nM) similar to values determined from native tissue. In addition, the recombinant binding domain of BoNT/B, a potential vector for neuronal delivery, bound quasi-irreversibly to synaptotagmin 2/GT1b exosomes. Engineered exosomes provide thus a novel means to study membrane proteins for biotechnology and clinical applications.


Results
Exosomal expression of membrane proteins. A patented technology which sorts chosen membrane protein to exosomes, was used to express 5 different transmembrane proteins: the BoNT/B receptors synaptotagmin 1 and 2, the seven transmembrane domain G-protein-coupled receptor CXCR4, the voltage-gated potassium channel Kv1.2, and the type 1 membrane protein Plasmodium falciparum apical membrane antigen (AMA1). A sequence encoding a peptide (DCTM), which promotes exosomal sorting 19 (patent no. WO2009115561), was fused to the 3′ end of each coding sequence cDNA and the resulting plasmids transfected in HEK cells. We established stable cell lines for synaptotagmins (SYTs) and transient expression for the other membrane proteins. Extracellular vesicles (exoKv1.2; exoCXCR4; exoAMA1; exoSYTs) were recovered from transfected HEK cell culture media and purified. The size of the membrane particles was evaluated by nanoparticle tracking analysis, yielding a mean diameter of 127 nm (Fig. 1A) for exoSYT2, consistent with the size of exosomes from HEK cells 20 .
Exogenously expressed proteins were detected in purified exosomal fractions by Western blotting (Fig. 1). All observed molecular weights were compatible with the predicted size of the mature full-length recombinant proteins (arrowheads, Fig. 1B,C). PNGase F treatment of exoSYT1/2 resulted in an increase in mobility, indicating N-glycosylation (Fig. 1C).
It has been previously shown that structurally complex proteins, such as G protein-coupled receptors, are addressed to exosomes when transiently transfected in cultured cells (Estelle 2005). We thus evaluated the relevance of our approach in the case of CXCR4 and compared its expression level, in the presence or absence of the DCTM tag. Exosomes purified from CXCR4 or CXCR4-DCTM transfected cells were coated in ELISA plates and CXCR4 expression was measured using anti-CXCR4 antibody. As shown in Fig. 1D, addition of the DCTM tag resulted in an approximately 10-fold (9 ± 0.8, n = 3 from 2 independent exosome preparations) increase in the expression of CXCR4.
Immunocharacterization of recombinant exosomes. To confirm the identity of the secreted membrane vesicles, we investigated the co-expression of exosomal surface markers in exoSYT2. Secreted membrane particles were captured on ELISA plates using specific antibodies and immobilized particles were probed with mAbs directed against exosomal surface markers. As shown in Fig. 2, immunocaptured SYT2 exosomes expressed three classical exosomal markers CD9, CD81 and CD63 21 . Similar results were obtained using exoCXCR4 (data not shown). A mAb directed against the N-terminal domain of SYT2, which is intraluminal in synaptic vesicles and extracellular at the neuronal plasma membrane recognized the immobilized particles in contrast to an absence of recognition by mAb1D12 that recognizes its cytoplasmic domain. These data indicated that as expected, the transmembrane orientation of SYT2 was identical to that in neuronal plasma membranes.
Surface plasmon resonance (SPR) was used to evaluate the expression level of recombinant SYT2 or CXCR4 in immunocaptured exosomes. Exosomes were specifically and stably captured on sensorchips and non-specific capture on control IgG was less than 10% of the total signal ( Supplementary Fig. 1). Anti-SYT2 antibody and a conformation-dependent anti-CXCR4 antibody were then injected at saturating concentrations. mAb binding responses (Rmax) reflected thus the total quantity of immobilized receptor. We found 0.07 RU (resonane units) of specific mAb/RU of immobilized exosomes for exoSYT2 and 0.1RU/RU for exoCXCR4, yielding a mean of 0.08 +/− 0.01 RU (n = 4) of bound antibody per RU of immobilized exosomes (not shown). Assuming that 1RU of mAb bound = 1 pg/mm 2 of molecule 22 on the exosome surface, a protein to lipid ratio of 2:1 (w:w) and that the molecular weight of SYT2/DCTM is 53 000 Da, one can calculate a content of 0.7-1.4 nmoles of SYT2/mg of protein. SYT2 would thus represent 3.6-7.2% of total proteins in SYT2 exosomes, depending on whether the mAb/ SYT2 interaction is monovalent or bivalent. A comparison of this value with our previous results obtained in similar conditions with synaptic vesicles 22 suggests that each exosome contains about 200 copies of exogenously expressed protein (see supplementary Methods).  Ligand binding to membrane proteins expressed in exosomes. We used the binding of specific ligands to address the proper assembly and/or conformation of the tetrameric Kv1.2 channel or CXCR4 at the surface of exosomal membranes. Radiolabeled 125 I-dendrotoxin, a specific ligand that binds to an extracellular site on the Kv1.2 channel, was incubated with exoKv1.2, in the presence or absence of an excess of unlabelled toxin. Exosomes were then filtered to separate free and bound toxin. While dendrotoxin did not bind to exocontrol, robust specific binding to exoKv1.2 was measured (Fig. 3A). Excess unlabelled dendrotoxin competitively inhibited binding, reaching the background signal determined using control exosomes (Fig. 3A).
SPR spectroscopy was used as a label-free sensing method to measure the conformationally-sensitive binding of SDF-1 to exoCXCR4 23 . To measure SDF-1 binding, sensor chips were functionalized independently on separate flowcells, with exoCXCR4 or exocontrol. SDF-1 binding to exoCXCR4 was quantified by automatically subtracting the non-specific signal obtained with exocontrol. As shown in Fig. 3B, SDF-1 (8 kDa) specifically bound to exoCXCR4 but binding was abolished if SDF-1 was injected in presence of the competitive antagonist AMD3100 (794 Da). These specific ligand-binding experiments suggest that recombinant proteins are expressed at the surface of exosomes with appropriate membrane topology and conformation.
ELISA detection of BoNT/B binding to exo SYT1/2. ELISA experiments showed that both GT1b and GD1a, the two main gangliosides that cooperate with SYT to form the high affinity BoNT/B binding site, were absent in exosomal preparations ( Supplementary Fig. 2). Accordingly, no specific BoNT/B binding was measured by ELISA upon direct coating of exoSYT1, exoSYT2 when compared to the negative control exoAMA1 (Fig. 4A, p > 0.05, n = 5 independent experiments). As GT1b can spontaneously insert in lipid membranes, exosomes were pre-incubated with GT1b before BoNT/B binding. Under these conditions, BoNT/B binding was detected at the surface of exoSYT1/GT1b and exoSYT2/GT1b (p < 0.05, 6 independent experiments, Wilcoxon matched pairs test). In accordance with the reported lower affinity of BoNT/B for SYT1 versus SYT2, BoNT/B binding to exoSYT1/GT1b was weaker, compared to exoSYT2/GT1b (Fig. 4B). The recombinant form of the C-terminal domain of BoNT/B (HcB) also bound to exoSYT2/GT1b (Fig. 4C). In contrast, BoNT/A did not bind neither to exoSYT2 nor to exoSYT1 in presence or absence of GT1b ( Supplementary Fig. 3). We checked if the ELISA detection of BoNT/B interaction with its receptor could be achieved in the presence of serum as the enzymatic activity of BoNT/B in vitro is largely inhibited under these conditions 16 . BoNT/B binding to SYT2/GT1b exosomes was readily revealed in the presence of 25% serum (Fig. 4D) and abrogated by pre-incubation of the toxin with neutralizing antibodies 16 , confirming binding specificity. Binding with serum gave higher signals than in its absence suggesting that unknown factors present in sera potentiate binding.
Biosensor determination of BoNT/B binding parameters in SYT2 exosomes. The binding affinity of BoNT/B was analyzed by SPR. ExoSYT2 and exoAMA1 were captured on sensor chips in separate flow cells, and GT1b was briefly loaded on immobilized exosomes before BoNT/B injection (Fig. 5A). Kinetic experiments were carried out by delivering increasing concentrations of BoNT/B, in a continuous flow, over both exoSYT2 and control exosomes. As BoNT/B binds the immobilized partner, the mass of material bound specifically to the surface increases with concentration as illustrated in Fig. 5B. Responses fit well to a 1:1 interaction model yielding a K D = 0.6 ± 0.1 nM (mean ± sem from 6 independent experiments) for BoNT/B holotoxin with an association rate constant (k on ) of 2.3 ± 1.10 5 M −1 s −1 and a dissociation rate constant (k off ) of 1.3 ± 0.5 10 −4 s −1 (Table 1). Similar Representative from 4 independent experiments. (B) SDF1 binding to CXCR4 exosomes. The CXCR4 ligand SDF-1 (50 nM) was injected over exoCXCR4 immobilized on a chip. SDF-1 binding was measured and its specificity confirmed using the CXCR4 antagonist AMD3100 (10 µM). Note that, as SPR monitors the mass of bound analyte, AMD3100 binding was not detected owing to it's low molecular mass relative to SDF-1. Representative SPR sensorgrams, after blank subtraction, from 3 independent experiments. values were obtained for BoNT/B (Table 1). Interestingly, while BoNT/B dissociated slowly (Fig. 5B), the isolated binding domain of BoNT/B (HcB) showed different kinetic characteristics (Fig. 5C). HcB bound very tightly to exoSYT2/GT1b with a k off < 10 −5 s −1 , precluding a precise affinity determination, as k off was under the limit of detection of the apparatus. Flat sensorgrams obtained upon injection of increasing concentrations of BoNT/A, BoNT/E holotoxins or HcA over exoSYT2, illustrate binding specificity ( Fig. 5B and C).
The equilibrium dissociation constant obtained from the kinetic parameters of BoNT/B binding to exoSYT2/ GT1b was very close to reported K D s measured at equilibrium using native nerve terminals 12,14 , thus validating the exosome biosensor approach. However it should be noted that binding assays require nanomolar toxin concentrations and are not adapted to the detection of extremely low BoNT/B concentrations for diagnostics and surveillance.

Discussion
There is an increasing demand for simple molecular assays to measure precise binding properties of ligands to membrane receptors that are major drug targets. However, successful reconstitution of functional membrane proteins is an empirical process, depending on many factors that must be optimized, including the formation and composition of the lipid environment 2 . We encountered this problem while developing cell-free biosensor assays to measure the affinity and kinetics of BoNT interactions with their membrane receptors. Reported binding assays, using full-length proteins or soluble domains of BoNT receptors, yielded affinities considerably lower than those found at synaptic terminals [24][25][26][27] . In addition, cellular models are cumbersome and less suitable for reproducible affinity measurements. Exosomes potentially provide an ideal mini-cell-like model, due to their stability, with a high cholesterol content, and a uniform native membrane receptor orientation, with post-translational modifications. Several reports indicated that exosomes can be modified to express soluble proteins 6,7 , however data concerning membrane protein targeting to exosomes are lacking. De Gassart and al. 19 , have demonstrated that the cytosolic domain of the transmembrane envelope (Env) protein of bovine leukemia virus contains motifs (DCTM) that potentially interact with the Endosomal Sorting Complexes Required for Transport (ESCRT) machinery, and can be used to address the ecto-domain of CD8 to exosomes. We therefore set out to develop a novel general approach to address membrane proteins to exosomes for simple assays, which can be automated and applied, in many laboratory settings. We confirm and extend the use of this exosomal targeting sequence (DCTM) to express the voltage-gated potassium channel Kv1.2, the G-protein coupled receptor CXCR4,  as well as the BoNT/B receptors SYT1 and SYT2. All recombinant proteins were properly oriented in vesicles which expressed typical tetraspanin exosomal markers. However, we did not investigate whether the transfected proteins were expressed in all exosomes or whether they were also present in other types of secreted vesicular particles. SPR chips were functionalized with specific antibodies, allowing exosomes to be captured and stably immobilized. As little as 500 pg of immobilized exosomes were enough to measure BoNT/B binding. We estimate that there are about 200 copies of SYT2 or CXCR4 per exosome. This high copy number was indirectly confirmed by SPR measurement of binding of the relatively small analyte SDF-1 (8 kDa) to CXCR4 exosomes. Our method clearly increased specific targeting to exosomes, compared to viral overexpression of a GPCR in the absence of a particular tag 28 .
Expression of the homo-tetrameric Kv1.2, CXCR4 and botulinum neurotoxin type B (BoNT/B) receptors in functional conformation was demonstrated in several binding assay formats: filtration assay, ELISA and SPR. The potential of this methodology was illustrated by measuring BoNT/B binding by ELISA and SPR to a double receptor consisting of SYT and the ganglioside GT1b. The ELISA method is effective in the presence of serum and could therefore be used to develop anti-BoNT/B antibodies, which are currently the only treatment for intoxication 29 and to screen for new antidotes acting on either the native receptor or the toxin. The exosome-based method, with a fully native double receptor, represents a flexible and more physiological context than minimal BoNT binding peptides. Furthermore, repeated injections of BoNT/B for clinical applications produce anti-toxin antibodies in a small number of patients 30 . The exosome assay could be used to determine whether these antibodies reduce therapeutic efficacy by perturbing BoNT/B binding to its receptor. It is however important to note that the exosome-based assay is not sensitive enough for direct detection of BoNT/B in the serum of patients with type B botulism, as toxin concentrations in this case are very low. Multi-step, sensitive (fM to pM) methods are commonly used to reveal, through the detection of its enzymatic activity, BoNT/B binding to native receptors 31 or recombinant peptide-based receptors 32,33 . A consecutive enzymatic assay should significantly increase the sensitivity of our system 16,34 . The native double receptor for BoNT/B i.e. synaptotagmin/GT1b was successfully reconstituted yielding K D s similar to those measured using native membranes. Studies using SYT peptides have always yielded lower affinities 24,26,27 and fail to reconstitute the double receptor. However, due to the native origin of exosomes, the copy number of BoNT receptors per exosome is significantly lower than peptide concentrations that can be coated on beads. Therefore, this test is neither appropriate for the detection of minute amounts of BoNT nor can it recapitulate all the BoNT intoxication steps. However it allows analysis of binding parameters independently from enzymatic activity. To date relatively cumbersome cell culture, ex-vivo methods and the mouse bioassay can evaluate all three functions necessary for toxicity (binding, translocation, endoprotease activity).
In this study, we used a SPR biosensor to establish a simplified automated assay for BoNT/B binding to exo-SYT2. Using two different BoNT/B sources K D values derived from kinetic constants were close to those using 125 I-BoNT/B at equilibrium with nerve terminals or SYT2/GT1b proteoliposomes, thus validating the accuracy of our approach 14,15 . Interestingly, the exosomal approach circumvents time-consuming detergent reconstitution procedures required to prepare proteoliposomes. Even though exosome purification involves multiple steps, a single exoSYT2 preparation provides sufficient material for more than 100 binding tests and can be frozen for long periods without degradation. Curiously, HcB, the BoNT/B domain that binds to SYT2, displayed a higher affinity than BoNT/B holotoxin, owing to an extremely slow dissociation phase. The difference in off-rate may reflect a reduced spatial restriction for HcB compared to BoNT/B, to accommodate the double SYT2/GT1b receptor 26 . Alternatively this might be explained by differences in affinity of BoNT subtypes, as it was reported that BoNT/B2 may be more active than BoNT/B1 29 . This finding could be of specific importance since HcB has potential for the development of protective recombinant vaccines against botulism, and to target specific drugs to nerve terminals 35,36 .
In conclusion, we have developed a new approach to express membrane proteins in exosomes providing detailed information on ligand-receptor interactions using biosensors. We focused on measuring the affinity of BoNT/B for its receptor, as precise determination of binding parameters can constitute a key factor in using BoNTs as therapeutic agents. Exosomes thus constitute a useful nanoscale system for manipulating membrane proteins in robust soluble structures for analysis of molecular interactions, vaccine production as well as screen for drugs that specifically target receptors embedded in exosomal membranes.
Exosome sizing. 0.1 μg of exosomal solution diluted in Milli-Q ® ultrapure water was sized using Nanoparticle Tracking Analysis (NanoSight NS300). Data were collected for 60 s and analyzed using NanoSight NTA 2.3 software.
Western blots. 3  conjugated antibodies were added and HRP activity measured at OD 450nm after addition of TMB and stopping the reaction by acidification.
ELISA detection of CXCR4. CXCR4 (ExoCXCR4) or DCTM-fused CXCR4 expressing exosomes (exo-CXCR4/DCTM) were produced and purified under the same conditions. Both exoCXCR4 and exoCXCR4/ DCTM (37,5 ng/well) were directly adsorbed in ELISA wells. Wells were blocked using Phosphate buffered Saline (PBS) containing 3% BSA, for 1 hour at 37 °C. Every subsequent step was preceded by washing using PBS supplemented with 0.05% Tween. Anti-CXCR4 (1 µg/ml) antibodies were incubated for 2 h at 37 °C in PBS and secondary HRP conjugated antibodies were added. HRP activity was measured at OD 450nm after addition of TMB and stopping the reaction by acidification.
ELISA detection of gangliosides. ExoSYT2 were captured and gangliosides detected using mAb anti GT1b (ascites diluted at 1/4000) and GD1a (2.5 µg/ml). Non-specific binding measured from control wells without exosomes was subtracted. Positive control for GT1b was performed by pre-incubating exosomes with exogenous GT1b 8 µM. HRP activity on secondary antibodies was measured at OD 450nm after addition of TMB and stopping the reaction by acidification.
Surface Plasmon Resonance (SPR) experiments. SPR measurements were performed at 25 °C using Biacore 3000 or Biacore T200 (GE Healthcare) apparatus. HBS (10 mM 4-(2-hydroxyethyl)-1-piperazine ethane sulfonic acid (HEPES)/NaOH pH 7.4, 150 mM NaCl) was used as running buffer and sensorchips were either CM5 from GE Healthcare or chip without dextran from Xantec (2D carboxymethyldextran surface CMDP). Sensorchips were functionalized by covalently coupling mAb antibodies recognizing CD81, CXCR4 or SYT2 (5B7a) using standard amine coupling chemistry (800-1200 RU on CMDP chip). Exocontrol or exoAMA1 were captured on flow cells functionalized with anti-exosomal marker CD81 antibody and used as negative control. ExoSYT2 were captured using 5B7a or anti-CD81 antibodies. Exosomes (0.3 mg/ml) were injected at a flow rate of 1 µl/min. The amount of bound exosomes varied between experiments and ranged from 500 to 2500 RU. For experiments involving exoSYT2, GT1b (1 µM) was added to pre-immobilized exosomes with a flow rate of 10 µl/min and the loading stopped upon reaching 1/3 of the exosome mass. Taking a mean molecular weight of 50 kDa for proteins and 2.2 kDa for GT1b, a protein to lipid ratio of 2 (w:w) similar to synaptic vesicles 42 , we calculated a molar ratio GT1b/total proteins of 10. Exosomes/GT1b were used for binding experiments at least 2 hours after GT1b loading. Five different concentrations of analytes, prepared by twofold dilutions ranging from 1 to 16 nM or 2 to 32 nM, were placed on the Biacore sample rack at 4 °C and assayed across the control and SYT2 exosomes surfaces at a flow rate of 30 µl/min. The data obtained from the control flow cell were automatically subtracted from experimental measurements to yield the specific signal. Data were analyzed using Biacore T200 Evaluation software 2.0 or BIAevaluation software 4.1 (GE Heathlcare). The single-cycle kinetic method was used in affinity measurements and K D values (k off /k on ) were calculated with the 1:1 titration kinetic binding model. Chips were regenerated using 100 mM octylglucoside as described 43 and used several times.

Statistical analysis.
Values presented are mean ± standard error of the mean (SEM) or mean ± standard deviation (SD). Statistical differences were assessed using the Wilcoxon matched-pair test. A p value < 0.05 was considered to be significant.