Blog

By Ehab El-Awaad, Galyna Pryymachuk, Cora Fried, Jan Matthes, Jörg Isensee, Tim Hucho, Wolfram F. Neiss, Mats Paulsson, Stefan Herzig, Frank Zaucke, and Markus Pietsch

Excerpt from the article published in Scientific Reports 9, 16272 (2019), Published 07 November 2019 DOI: https://doi.org/10.1038/s41598-019-52655-y

Editor’s Highlights

• Thrombospondins (TSPs) form a family of five large oligomeric extracellular matrix glycoproteins that are expressed by numerous cell types.
• In neurons, astrocyte-secreted or recombinantly expressed TSP(s), particularly TSP-1, TSP-2 and TSP-4, were reported to promote the formation of excitatory synapses.
• The alpha2-detla (α₂δ) proteins (α₂δ‐1–4) are auxiliary subunits of voltage-gated calcium channels Ca_V1 and Ca_V2, and were found to be encoded by four different genes.
• The α₂δ-1 protein is functionally involved in TSP-induced synaptogenesis. This interaction is crucial for excitatory synaptogenesis and neuronal sensitisation, suggesting a potential role in the pathogenesis of neuropathic pain.
• The interaction of dysregulated TSP-4 with α₂δ-1 are partially independent of the role of the latter protein in regulating voltage-gated calcium channels’ trafficking and function.
• There is substantial evidence for a direct and specific Ca²⁺-insensitive TSP-4/α₂δ-1 interaction which is rather weak.

Abstract

The α₂δ‐1 subunit of voltage-gated calcium channels binds to gabapentin and pregabalin, mediating the analgesic action of these drugs against neuropathic pain. Extracellular matrix proteins from the thrombospondin (TSP) family have been identified as ligands of α₂δ‐1 in the CNS. This interaction was found to be crucial for excitatory synaptogenesis and neuronal sensitisation which in turn can be inhibited by gabapentin, suggesting a potential role in the pathogenesis of neuropathic pain. Here, we provide information on the biochemical properties of the direct TSP/α₂δ-1 interaction using an ELISA-style ligand binding assay. Our data reveal that full-length pentameric TSP-4, but neither TSP-5/COMP of the pentamer-forming subgroup B nor TSP-2 of the trimer-forming subgroup A directly interact with a soluble variant of α₂δ-1 (α₂δ-1_S). Interestingly, this interaction is not inhibited by gabapentin on a molecular level and is not detectable on the surface of HEK293-EBNA cells over-expressing α₂δ‐1 protein. These results provide biochemical evidence that supports a specific role of TSP-4 among the TSPs in mediating the binding to neuronal α₂δ‐1 and suggest that gabapentin does not directly target TSP/α₂δ-1 interaction to alleviate neuropathic pain.

Introduction

Thrombospondins (TSPs) form a family of five large oligomeric extracellular matrix glycoproteins that are expressed by numerous cell types, playing important roles in cellular migration, attachment and cytoskeletal dynamics^1,2. Several TSP isoforms have been shown to be involved in a variety of physiological and pathological processes, including regulation of angiogenesis, apoptosis and platelet aggregation^3,4,5. TSPs can be subdivided into subgroups A (TSP-1 and 2) and B (TSP-3–5, with TSP-5 also referred to as cartilage oligomeric matrix protein (COMP)) based on their oligomerisation state (trimeric or pentameric, respectively) and domain structure (Fig. 1A). In neurons, astrocyte-secreted or recombinantly expressed TSP(s), particularly TSP-1, TSP-2 and TSP-4, were reported to promote the formation of excitatory synapses both in vitro and in vivo through interaction with the voltage-gated calcium channel subunit α₂δ-1^6,7,8,9,10. The α₂δ proteins (α₂δ‐1–4) are auxiliary subunits of voltage-gated calcium channels Ca_V1 and Ca_V2, and were found to be encoded by four different genes^11,12,13. Functions of these auxiliary subunits include the modulation of trafficking, expression in the plasma membrane^14,15,16,17, and biophysical properties of the channels^15,17,18,19. Importantly, α₂δ-1 acts as a specific binding site for gabapentinoid drugs^20,21, mediating their analgesic effect in neuropathic pain^21,22,23. Furthermore, studies using different animal models of neuropathic pain indicated the involvement of α₂δ-1 in pain development, with nerve injuries leading to up-regulation of α₂δ‐1 in both dorsal root ganglion (DRGs) and spinal dorsal horn neurons^24,25,26,27 as well as to an increase of miniature excitatory post-synaptic current (mEPSC) frequency in the latter neurons^{22,25,26,28,29,30}. Similarly, injury-induced TSP-4 is reported to mediate central sensitisation and neuropathic pain states^{8,9,31,32,33,34,35,36,37}. This effect was recently shown to be mediated by activation of a TSP-4/α₂δ‐1-dependent pathway which requires a direct molecular interaction between the two proteins^9,34. Furthermore, the presence of TSP-4 was shown to modestly but significantly reduce the binding affinity of ³H-gabapentin (³H-GBP) towards α₂δ‐1 in membrane preparations from TSP-4/α₂δ‐1 co-transfected cells³⁸. Taken together, the TSP-4/α₂δ‐1 protein-protein interaction seems to be of potential translational importance and thus may serve as a novel target for developing a new class of analgesics against neuropathic pain.

The aim of the present study is to investigate the biochemical characteristics of the direct molecular interaction between TSPs and α₂δ‐1, addressing the question whether α₂δ‐1 binding is specific to TSP-4 or redundant among other TSPs. GBP has been shown so far to inhibit the interaction of α₂δ‐1 with a truncated form of TSP-2 in co-immunoprecipitation experiments⁷ as well as functionally by inhibiting synaptogenesis^7,8,10, neuron sensitisation and behavioural hypersensitivity induced by TSP-2, its truncated fragment and/or TSP-4^9,34,35. Thus, we examined whether the direct TSP/α₂δ‐1 interaction can be inhibited by GBP on a molecular level as well. We therefore generated purified recombinant forms of three full-length TSPs (TSP-2, TSP-4 and COMP) as well as soluble forms of α₂δ‐1 subunit (α₂δ‐1_S), that shows GBP binding affinity similar to that of wild-type α₂δ-1, and the α₂ peptide chain of α₂δ‐1 (Fig. 1). Both the interaction of these recombinant TSPs with α₂δ‐1 and the possible inhibition by GBP were examined in a solid-phase ELISA-style ligand binding assay, with the capability of soluble α₂δ‐1 to interact with GBP being proven by a newly developed surface plasmon resonance (SPR)-based binding assay. In order to demonstrate the characteristics of the direct TSP-4/α₂δ‐1 interaction in an environment similar to that of native cells, we attempted to visualise the interaction of fluorescently labelled TSP-4 with membrane-localised full-length (FL) α₂δ‐1 in a cell-based system.

Results

Biochemical characteristics of recombinant purified proteins expressed in HEK293-EBNA cells

To investigate the direct binding of different TSPs to α₂δ-1, three full-length recombinant TSPs, namely, the trimeric TSP-2, the pentameric proteins TSP-4 and COMP, and a soluble C-terminal deletion mutant of α₂δ-1 carrying an N-terminal (α₂δ-1_S NTST) or a C-terminal (α₂δ-1_S CTST) double strep II-tag (Fig. 1) were generated in a eukaryotic expression system. Coomassie stained gels and immunoblots of the purified proteins confirmed their purity, identity and integrity (Fig. 2; Supplementary Fig. S1). As expected for TSPs, the intact proteins showed high molecular weight bands with approximate apparent molecular weights (Mr) in the range ~500–670 kDa when compared to laminin-111 and thyroglobulin as marker proteins (Fig. 2A). The oligomerisation patterns of these high molecular weight TSPs (i.e. pentamers for TSP-4 and COMP and trimer for TSP-2) were confirmed by comparing immunoblots in the absence and presence of the reducing reagent DTT (Fig. 2B; Supplementary Fig. S1a). In the latter case, DTT reduces the interchain disulphide bonds within the oligomerisation domains of the analysed TSPs³⁹ and major bands of monomeric proteins with Mr of TSP-2 (~240 kDa), TSP-4 (~160 kDa), and COMP (~130 kDa) were observed (Fig. 2B, +DTT; Supplementary Fig. S1a, +DTT).

Similar analysis was performed for the generated α₂δ-1_S NTST and the respective C-terminally tagged α₂δ-1 variant, α₂δ-1_S CTST, showing single but smeared bands at approximate Mr ~200 kDa under non-reducing conditions (Fig. 2C, −DTT; Supplementary Fig. S1b, −DTT) and appear as distinct bands at approximate Mr ~180 kDa under reducing conditions (Fig. 2C, +DTT; Supplementary Fig. S1b, +DTT). So far, there is no comprehensive explanation for this gel band shift in the presence of DTT. However, it cannot be attributed to the reductive cleavage of the interchain disulphide bridge between α₂ and δ-1 followed by loss of the smaller δ-1 chain. This conclusion arises from the observation that α₂δ-1_S bearing the C-terminal double strep II-tag (α₂δ-1_S CTST) is still detectable in immunoblots probed with strep II-tag antibody following DTT treatment. In agreement with this result, mass spectra of α₂δ-1_S NTST recorded with and without DTT pre-treatment showed almost identical molecular ion peaks ([M + H]⁺; Table 1, Supplementary Fig. S2d,e). This observation confirms the results by Brown and Gee⁴⁰ who first described a similar soluble mutant of the porcine α₂δ-1 orthologue which retains high affinity for ³H-GBP. Although uncleaved α₂δ-1 may not represent a functional form as a subunit of the Ca_V channels and can inhibit native calcium currents in mammalian neurons⁴¹, the TSP/α₂δ-1 pathway is thought to be at least partially independent of the roles of α₂δ-1 as a Ca_V channel subunit^7,10. Therefore, the recombinant uncleaved α₂δ-1_S variant used in this study should be suitable for the purpose of investigating TSP binding biochemically. Notably, we observed a minor band in the immunoblots of α₂δ-1_S CTST at Mr ~25 kDa upon DTT treatment and detection with strep II-tag antibody (Fig. 2C, +DTT) which is most likely attributed to cleaved strep-tagged δ-1_Schain. This indicates the presence of a small fraction of the generated purified α₂δ-1_S in a proteolytically cleaved form.

Protein	m/zcalc [M + H]+	m/zexp [M + H]+
TSP-2 (monomer)	132,137	146,958
TSP-4 (monomer)	107,409	109,351
COMP (monomer)	84,714	88,463; 87,178
α2δ-1S NTST (−DTT)	123,000	152,236
α2δ-1S NTST (+DTT)	112,130*	152,258

In addition to the α₂δ-1_S variants generated, the α₂ peptide chain (α₂ NTST) was recombinantly produced in a similar way (Fig. 1B). Expression and purification of this fragment as well as analyses by SDS-PAGE and immunoblotting (Fig. 2D) were carried out as described above. Here, we observed the formation of a high molecular weight product under non-reducing conditions which dissociated into the monomeric form after DTT treatment (approximate Mr ~ 170 kDa, Fig. 2D) which points to the formation of interchain disulphide bonds in the absence of reducing agents (see also Discussion section below).

Notably, all recombinant proteins generated in this study that had been analysed by SDS-PAGE and Western Blot showed protein bands at remarkably higher Mr than expected from their amino acid sequences. It is known that the electrophoretic mobility of proteins can be greatly influenced by the extent of post-translational modifications (e.g. glycosylation) of the protein where the glycan chains do not bind SDS leading in many cases to decreased mobility, and increased Mr, of the glycoprotein analysed by SDS-PAGE⁴². In addition, sample treatment prior to loading onto the gel (e.g. heating at 95 °C with DTT) represents a possible source of abnormal protein migration on the gels through its impact on the structure of the analysed protein⁴³. Therefore, further analysis was carried out to determine the accurate molecular masses of the recombinant purified proteins using MALDI-TOF mass spectrometry. The results are shown in Table 1 and Supplementary Fig. S2. The molecular masses of TSP-4 and COMP, both in monomeric form, were found to be only slightly higher (1–3 kDa) than the theoretical masses calculated on the basis of each protein’s amino acid sequence, indicating minor post-translational modifications, e.g. glycosylation, of these proteins. In case of TSP-2 and α₂δ-1_S NTST, the experimentally determined masses were about 15 and 30 kDa, respectively, larger than the theoretical ones, which is likely attributed to heavy glycosylation, as shown previously for α₂δ-1 in the work of Kadurin et al.⁴⁴ (Table 1, Supplementary Fig. S2a,d,e).

Characterisation of the interaction of TSPs with α₂δ-1_S using an ELISA-style ligand binding assay

First, the purified recombinant full-length TSPs (soluble) were examined for their direct interaction with immobilised α₂δ-1_S NTST in an ELISA-style ligand binding assay, that was validated using the model interaction of COMP and matrilin-3 proteins⁴⁵ (Supplementary Fig. S4). Preliminary experiments with TSP concentrations up to ~500 nM had demonstrated that out of the three TSPs generated in this study, only TSP-4 was able to directly interact with α₂δ-1_S NTST (data not shown). To rule out the possibility of a very low binding affinity of TSP-2 and COMP towards α₂δ-1, we finally utilised a comparably high concentration (1,000 nM) of all TSP proteins in the same ELISA which confirmed our preliminary data (Fig. 3A). Therefore, we decided to focus on the interaction of TSP-4 with α₂δ-1_S and study it more in detail. Notably, we detected comparable binding signals of TSP-4 to either α₂δ-1_S NTST or α₂δ-1_S CTST variants in a preliminary experiment (data not shown) and hence we chose to proceed with one of the two variants, namely, α₂δ-1_S NTST in our binding assays. Titration of immobilised α₂δ-1_S NTST with increasing concentrations of TSP-4 (11-1,505 nM) showed saturable binding with an apparent K_D value of about 200 nM. Since Ca²⁺ binding is associated with major conformational changes and structural rearrangements of both TSPs^46,47 and the metal ion-dependent adhesion site (MIDAS) motif of α₂δ-1^17,48, we investigated the effect of Ca²⁺ on TSP-4/α₂δ-1_S NTST interaction. Our data show that the apparent K_D value was slightly, but not significantly, decreased in presence of 1 mM Ca²⁺(Table 2, Fig. 3B). Similarly, the maximum binding value (B_max, unitless) showed a small, statistically non-significant increase in presence of 1 mM Ca²⁺ (Table 2, Fig. 3B). These results indicate that the TSP-4/α₂δ-1_S NTST interaction is not sensitive to Ca²⁺ changes. Next, the ability of TSP-4 to directly interact with the α₂ fragment of α₂δ-1 was analysed. Here, we observed a two-fold, statistically significant increase in the binding signal of a single concentration of TSP-4 (1,000 nM) when using immobilised α₂ NTST instead of α₂δ-1_S NTST (Fig. 3C). These results indicate the localisation of the TSP-4 binding site(s) within the α₂region of α₂δ-1 in agreement with data obtained by Eroglu et al.⁷. The enhanced binding signal of TSP-4 to α₂ NTST is suggestive of a more favourable conformation of α₂ NTST which is more accessible for TSP-4 binding, as compared to the non-proteolytically cleaved α₂δ-1_S NTST.

Assay buffer	KD (nM)	Bmax
TBS	198 ± 49	0.579 ± 0.066
TBS + 1 mM Ca2+	153 ± 46	0.677 ± 0.054

To investigate the effect of the known α₂δ-1 ligand GBP²⁰ on the observed TSP-4/α₂δ-1_Sinteraction, we checked for the ability of the recombinantly expressed mutant α₂δ-1_S NTST to bind GBP with high affinity (K_D = 219 nM) using a label-free surface plasmon resonance (SPR) assay (Fig. 4A, Supplementary Fig. S3). Our results are in agreement with the binding data for ³H-GBP to membrane preparations of heterologously expressed human α₂δ-1 (K_D range of 140–175 nM)³⁸. Various GBP concentrations (0.05–1,000 µM) were used in the TSP-4 /α₂δ-1_SNTST binding assay to interfere with this protein-protein interaction but did not show inhibitory effects (Fig. 4B). To rule out the possibility that the GBP binding pocket might not be accessible after immobilisation of α₂δ-1_S NTST on the ELISA microplate, we had coated the wells with α₂δ-1_S NTST after pre-incubating the protein with GBP. In addition, GBP had been supplemented to the liquid phase during blocking and incubation with TSP-4 ensuring availability of a sufficient number of GBP molecules to α₂δ-1_S NTST and preventing dissociation of bound GBP during the experiment. In a further experiment, the binding of various concentrations of TSP-4 (12.5–1,000 nM) to α₂δ-1_S NTST was not affected by the highest GBP concentration (1,000 µM) investigated (Fig. 4C). Together, these data suggest that GBP alone is not sufficient to disrupt the interaction of TSP-4 with α₂δ-1 on a molecular level.

Fluorescent A555-TSP-4 does not bind to membrane-bound α₂δ-1 in a cell-based binding assay

To examine the interaction between TSP-4 and full-length membrane-bound α₂δ-1 in a cellular environment, HEK293-EBNA cells were transfected with either empty vector or vector encoding α₂δ-1 FL NTST. Immunocytochemical analysis of cells transfected with α₂δ-1 FL NTST revealed a marked increase in α₂δ-1 immunoreactivity, confirming the over-expression of heterologous α₂δ-1 (Fig. 5A, left column).

In cells immunostained without permeabilisation, a high degree of co-localisation of α₂δ-1 and wheat germ agglutinin (WGA) used for labelling glycoproteins or glycolipids of the outer leaflet of the plasma membrane⁴⁹ were observed, assuring the localisation of heterologous α₂δ-1 in the plasma membrane (Fig. 5A, upper left image). The transfected cells were incubated with increasing concentrations of fluorescently labelled A555-TSP-4 (9–909 nM), the binding of which to α₂δ-1_S NTST was found to be non-significantly lower than that of unlabelled TSP-4 in ELISA-style assay (Fig. 5B). Images of the stained cells were acquired using a High Content Screening microscope and the average A555-TSP-4 signal intensity was determined for each well. In addition, α₂δ-1 FL NTST was visualised by immunostaining to analyse the co-localisation of A555-TSP-4 with α₂δ-1 FL NTST. Results show low overall A555-TSP-4 signal intensity which increases upon increasing the concentrations of the fluorescent protein. However, no difference was observed when comparing α₂δ-1 FL NTST- and empty vector-transfected cells (Fig. 5C). Furthermore, no co-localisation of A555-TSP-4 and α₂δ-1 FL NTST signals was detected in these experiments (Supplementary Fig. S5). These results are in accordance with recently published data by Lana et al.³⁸ showing no TSP-4 co-localisation with α₂δ-1 on the cell surface of tsA-201 cells co-expressing both proteins or in mixed populations of cells transfected separately with either α₂δ-1 or TSP-4.

Discussion

All TSPs, i.e. TSPs 1–4 and COMP, were previously identified as synaptogenic proteins which, together with other astrocyte-derived factors, help to promote the formation of functional excitatory synapses in the CNS^6,7. The α₂δ-1 protein was demonstrated to be functionally involved in TSP-induced synaptogenesis by means of synaptic assays in retinal ganglion cells (RGCs)⁷, DRG/spinal cord primary neuron co-culture^8,9, purified cortical neurons¹⁰ as well as in dorsal spinal cord of mice³⁴. Biochemically, α₂δ-1 was reported to interact in co-immunoprecipitation experiments with TSP-1, TSP-2 and TSP-4 from rat cerebral cortex⁷ as well as with TSP-4 from rodent spinal cord³⁴. Similarly, a TSP-2 fragment containing all three EGF-like repeats, the calcium-binding repeats, and the C-terminal globular domain was co-purified with full-length α₂δ-1 or its protein-binding VWA domain after co-expression in HEK293 cells⁷. Recently, Park et al.^9,34 demonstrated for the first time a direct molecular interaction between α₂δ-1 and recombinant full-length TSP-4 or its fragments containing EGF-like or coiled coil domains. In the present study, we investigated the biochemical properties of the direct TSP-4/α₂δ-1 interaction. Furthermore, it was of importance to know whether other members of the TSP protein family are also able to directly bind to α₂δ-1 in an analogous manner to that of TSP-4. Our data demonstrated that only full-length TSP-4, but not TSP-2 or COMP, is able to directly interact with immobilised soluble α₂δ-1 variant (α₂δ-1_SNTST) in an ELISA-style ligand binding assay (Fig. 3A), indicating the specificity of this protein-protein interaction. This observation is in direct contrast to that of Eroglu et al.⁷ (see above). Nevertheless, TSP-4 is remarkably the only isoform of TSP proteins reported so far to be implicated in neuropathic and joint-mediated chronic pain in rodents along with neuronal α₂δ-1^8,9,31,34,35. During the processes resulting in such pain, both TSP-4 and α₂δ-1 are up-regulated on the protein level and temporally correlate with the development of behavioural hypersensitivity in the respective animal models (for review see ref.⁵⁰). In contrast, TSP-1/-2, though previously shown to be up-regulated after ischemic brain injury in rodents^4,51 and promoting the subsequent synaptic recovery⁵¹, are not dysregulated on the protein level in dorsal spinal cord after spinal nerve ligation in mice, even when behavioural hypersensitivity was evident in these animals. This observation led Kim et al.³¹ to rule out the possibility of the involvement of these two astrocyte-secreted proteins in mediating TSP/α₂δ-1-induced neuropathic pain. With regards to COMP, it has been reported to be expressed in several tissues including skeletal muscle, tendon, and cartilage. In the latter tissue, COMP is known to be mainly involved in chondrocyte differentiation, attachment, and cartilage extracellular matrix assembly^52,53,54,55, with mutations in COMP being associated with pseudoachondroplasia^56,57. COMP in skeletal muscle, tendons, and perichondrium can be theoretically in contact with nerve terminals containing α₂δ-1. In addition, COMP was shown to have synaptogenic potential in RGCs (as discussed) and shares a high degree of both overall sequence identity (~70%) and structural similarity (Fig. 1A) with TSP-4. That is why we investigated COMP as a potential interaction partner of α₂δ-1 in our binding studies. Nevertheless, COMP is, in contrast to TSP-4, neither abundant in neurons and astrocytes nor it is known to be dysregulated in neuropathic pain states. The fact that, beside TSP-4, all other TSPs were previously found to induce synaptogenesis through a mechanism involving neuronal α₂δ-1⁷ may refer to other cellular factors or scaffold proteins required to mediate their interaction with α₂δ-1 indirectly. Furthermore, it is tempting to speculate that TSP-induced synaptogenesis might be of little relevance to neuropathic pain development since, as previously mentioned, TSP-4 is the only member of TSP family found to be up-regulated in dorsal spinal cord following nerve injury. Indeed, a recent study shows that an enhanced presynaptic NMDA receptor activity, rather than synaptogenesis, is responsible for maintaining the increased synaptic excitatory transmission in dorsal spinal cord leading to chronic pain states following nerve injury in mice⁵⁸.

In further experiments, we observed a significantly increased binding of TSP-4 to α₂ NTST when compared to α₂δ-1_S NTST (Fig. 3C), confirming previous data by Eroglu et al.⁷demonstrating the TSP-4 binding site to be localised within the α₂ region of α₂δ-1 (VWA domain). The observed enhancement in binding towards α₂ might be attributed to more exposed TSP-4 binding motif(s) in the immobilised α₂ NTST compared to the non-proteolytically processed α₂δ-1_S NTST. This result is in agreement with recent findings from Lana et al.³⁸ where wild-type α₂δ-1 was very weakly co-immunoprecipitated with TSP-4, but no co-immunoprecipitation of the mutant α₂δ-1 (MIDAS^AAA) with TSP-4 was detected in lysates of co-transfected tsA-201 cells. In addition, the binding to α₂ NTST seems again to be TSP-4-specific since negligible binding signals were detected when equimolar concentrations of either COMP or a truncated TSP-4 fragment were utilized in a pilot experiment (data not shown). Although our results are supported by reported data, we cannot rule out the possibility that the enhanced TSP-4 binding signal is due to improperly expressed α₂ NTST since unpaired cysteine residues, normally involved in the formation of disulphide bridges with other cysteine residues in the deleted regions of wild-type α₂δ-1⁵⁹, become available. It is worth mentioning here that a very high tendency for multimerisation was observed for a recombinant VWA domain of α₂δ-1 generated in this study (data not shown) due to the formation of intermolecular disulphide bonds. Therefore, it might be appropriate to consider the expression of α₂ and VWA fragments in which the unpaired cysteines are replaced by other isosteric residues (e.g. serine) for future binding studies.

One of the reported small molecules capable of interfering with the TSP/α₂δ-1 interaction is GBP, an approved analgesic against neuropathic pain^60,61,62 and a known ligand of α₂δ-1²⁰. Biochemically, co-immunoprecipitation experiments showed that the interaction between a truncated TSP-2 fragment and α₂δ-1 FLAG in a co-culture of two populations of HEK293 cells was diminished in the presence of GBP⁷. In addition, as previously mentioned, TSP-4 modestly but significantly reduces the binding affinity of ³H-GBP to α₂δ-1, suggesting rather an allosteric than a pure competitive mode of inhibition³⁸. Furthermore, in vivo data revealed the ability of GBP to block TSP-4-induced neuronal sensitisation and behavioural hypersensitivity as well as changes in Ca²⁺ currents and intracellular Ca²⁺ transients after injuries to peripheral nerves or facet-joint in rodents^8,9,34,35,63. Similarly, several studies in neuropathic pain models demonstrated the ability of GBP to inhibit α₂δ-1-induced²⁶ or TSP-induced^7,8,34,35 synaptogenesis. Most recently, GBP was also shown to inhibit TSP-2-induced synapse formation in purified culture of cortical neurons¹⁰. Despite the multidimensional evidence of GBP interference with TSP/α₂δ-1 interaction, a direct GBP inhibition of this interaction on the molecular level has never been investigated before, to our knowledge. In the current study, we did not observe any inhibition of the direct TSP-4/α₂δ-1_SNTST interaction in the presence of increasing concentrations (up to 1 mM) of GBP (Fig. 4B). Furthermore, the highest GBP concentration used (1 mM) did not shift the TSP-4/α₂δ-1_SNTST binding curve (Fig. 4C). Although the utilised α₂δ-1_S NTST was mostly expressed as uncleaved form of the protein (in agreement with the original work describing a similar porcine α₂δ-1 mutant⁴⁰), we were able to demonstrate the ability of this α₂δ-1_S mutant to retain high affinity for GBP (Fig. 4A). For this purpose, a newly developed SPR-based binding assay suitable for detecting and quantifying the binding of small molecules to immobilized recombinant α₂δ-1_S was used. This SPR assay has the advantage of being radiolabel-free and can easily be used to determine the binding kinetics unlike the previously used ³H-GBP binding assay^38,40,64,65. Taken together, our data confirmed that the proteolytic cleavage of α₂δ-1 is not crucial for the formation of the GBP binding pocket⁴⁰. The complete lack of GBP inhibition towards the interaction of purified TSP-4 with α₂δ-1_S NTST raises questions regarding the exact mechanism by which GBP can block the above-mentioned TSP-induced changes. It is possible that other unknown factors in the cellular environment are essential for GBP to interfere with α₂δ-1/TSP-4 interaction and thereby mediating the known GBP inhibitory effects. Another possible explanation based on the recent findings of Chen et al.^58,66 is that the α₂δ-1/NMDA receptor complex, rather than α₂δ-1/TSP-4 binding, represents the molecular target of gabapentinoid drugs to alleviate neuropathic pain.

Our efforts were as well focused on the investigation of the TSP-4/α₂δ-1 interaction in a cellular system to get closer to the physiological/pathological situation in the CNS. We over-expressed full-length α₂δ-1 in HEK293-EBNA cells and demonstrated both its intracellular and plasma membrane localisation in transfected cells (Fig. 5A). Treatment with increasing concentrations (up to 909 nM) of fluorescently labelled A555-TSP-4, however, showed no differences in binding of the protein to α₂δ-1 overexpressing cells when compared to control cells (Fig. 5C, Supplementary Fig. S5). Furthermore, fluorescent signals of A555-TSP-4 did not co-localise with those of immunostained α₂δ-1 on the cells (Supplementary Fig. S5). The observed loss of binding cannot be attributed to an impairment of the interaction of the two proteins by the fluorescent label of TSP-4 since the fluorescent protein was generated with a minimal dye-to-protein molar ratio and showing substantial α₂δ-1_S NTST binding in the ELISA-style assay (Fig. 5B). A possible explanation could instead be arising from the weak interaction of the two proteins under the conditions of the cell-based assay, unlike the ELISA-style assay. This means that very high local concentrations of TSP-4 in the proximity of cell-surface α₂δ-1 would be required to enable the detection of their interaction by simulating the pathological situation (e.g. dramatic up-regulation following nerve injury). This could not, however, be achieved with the range of A555-TSP-4 concentrations (up to 909 nM) used in the assay. In our experiments, we over-expressed α₂δ-1 in HEK293-EBNA cells without co-expression of α₁ subunit, which interacts intracellularly with α₂δ-1 before trafficking of the complex to the cell surface⁶⁷. However, we assume that the lack of α₁ subunit did not hamper the putative binding of TSP-4 to α₂δ-1 on the cell surface. This assumption is based on recent data showing the ability of wild-type α₂δ-1, expressed in HEK293 cells without co-expression of α₁ subunits, to be efficiently transported to the cell surface and thereby become accessible to extracellular ligands like TSP¹⁰. Functionally, this α₂δ-1 over-expression construct alone was able to rescue synapses in cortical organotypic slices from α₂δ-1 knockout mice. This effect is found to be mediated through activation of the small Rho GTPase Ras-related C3 botulinum toxin substrate 1 (Rac1) and is independent of α₁ subunits of the postsynaptic L-type calcium channels Ca_V1.2 and Ca_V 1.3¹⁰.

The α₂δ subunits are thought to promote membrane trafficking of the pore subunits of voltage-gated calcium channels¹⁷ and α₂δ-1-driven allodynia in mice can be reversed by blockers of voltage-gated calcium channels like ω-conotoxin GVIA⁶⁸. However, other findings suggest that the maladaptive changes contributing to chronic pain in rodents following nerve injuries and resulting from the interaction of dysregulated TSP-4 with α₂δ-1 are partially independent of the role of the latter protein in regulating voltage-gated calcium channels’ trafficking and function⁵⁰.

As previously mentioned, our data align with those of Lana et al.³⁸ who reported no interaction of secreted TSP-4 with membrane-localised α₂δ-1 on tsA-201 cells when subjected to immunocytochemical analysis. On the other hand, the same study showed weak intracellular interaction of both proteins in co-immunoprecipitation experiments from cells over-expressing both proteins³⁸. It has therefore been postulated that the weak TSP-4/α₂δ-1 interaction may occur in an intracellular compartment rather than on the cell surface^38,69. This postulation is in contrast to the previous data showing synaptogenic effect of secreted TSPs which is mediated by neuronal α₂δ-1 thought to be located either pre-⁸ or post-^7,10synaptically. To our knowledge, there are no data so far showing the co-localisation of TSP(-4) and α₂δ-1 in neuronal cell cultures or spinal cord tissue where TSP-induced changes (e.g. synaptogenesis) were demonstrated. To reveal the exact cellular localisation of TSP-4/α₂δ-1 interaction it would be very helpful to simultaneously analyse co-immunostained TSP-4 and α₂δ-1 proteins in cultures of neurons utilized in in vitro synapse assays^6,7,10.

In summary, our results provide substantial in vitro biochemical evidence for a direct and specific Ca²⁺-insensitive TSP-4/α₂δ-1 interaction which is rather weak. Importantly, GBP does not inhibit this interaction on a molecular level, indicating the possible involvement of other unknown factors or targets in mediating GBP inhibitory effects in neuropathic pain.

We, therefore, need to understand the exact and complete molecular mechanism of the TSP/α₂δ‐1 interaction to really be able to design appropriate small molecule modulators – rather than being left to use and optimize the enigmatic properties of the serendipitously discovered gabapentinoid action.