LRP1 influences trafficking of N-type calcium channels via interaction with the auxiliary α2δ-1 subunit

Voltage-gated Ca2+ (CaV) channels consist of a pore-forming α1 subunit, which determines the main functional and pharmacological attributes of the channel. The CaV1 and CaV2 channels are associated with auxiliary β- and α2δ-subunits. The molecular mechanisms involved in α2δ subunit trafficking, and the effect of α2δ subunits on trafficking calcium channel complexes remain poorly understood. Here we show that α2δ-1 is a ligand for the Low Density Lipoprotein (LDL) Receptor-related Protein-1 (LRP1), a multifunctional receptor which mediates trafficking of cargoes. This interaction with LRP1 is direct, and is modulated by the LRP chaperone, Receptor-Associated Protein (RAP). LRP1 regulates α2δ binding to gabapentin, and influences calcium channel trafficking and function. Whereas LRP1 alone reduces α2δ-1 trafficking to the cell-surface, the LRP1/RAP combination enhances mature glycosylation, proteolytic processing and cell-surface expression of α2δ-1, and also increase plasma-membrane expression and function of CaV2.2 when co-expressed with α2δ-1. Furthermore RAP alone produced a small increase in cell-surface expression of CaV2.2, α2δ-1 and the associated calcium currents. It is likely to be interacting with an endogenous member of the LDL receptor family to have these effects. Our findings now provide a key insight and new tools to investigate the trafficking of calcium channel α2δ subunits.

Low density lipoprotein (LDL) receptor-related protein-1 (LRP1) is a member of a large LDL receptor family 20 , which also includes LDL receptor, very low density lipoprotein receptor, apoE receptor 2, glycoprotein 330 (gp330/megalin/LRP2) and LRP1B 21 . In addition, this protein family comprises LRP5, and LRP6, which are involved in Wnt signalling 22 . LRP1 is multifunctional receptor, interacting with number of protein ligands, most of which bind to the four clusters of cysteine-rich and EGF-repeats, termed ligand binding domains (LBDs). It mediates both forward trafficking and endocytosis of protein cargoes 21 . Binding of LRP1 ligands at the plasma membrane can also promote its assembly into co-receptor systems and initiate intracellular cell signaling cascades 23 .
We surmised that LRP1 would be a good candidate for a role in trafficking the α 2 δ subunits, since it is involved in forward trafficking of a number of other proteins, including β 1-integrin 24 and GPI-anchored PrP 25 , and in clathrin-dependent endocytosis 21,26 . It is also involved in the trafficking of glucose transporter-4, which is rapidly inserted into the plasma membrane from sub-membrane storage vesicles in response to insulin 27 . In the nervous system, LRP1 is present in synapses 28 , as is α 2 δ -1 15 , and neuronal deletion of the LRP1 gene results in a phenotype including neurodegeneration and synapse loss, and development of behavioral and motor abnormalities 29,30 . The interaction of LRP1 with specific ligands has also been implicated in neurite outgrowth 31 . LRP1 has also been shown to bind to thrombospondins with high affinity 32 , which is implicated in synapse formation 19 , and is able to bind more than one ligand at different sites 33 .
LRP1 and related proteins interact with Receptor-Associated Protein (RAP), a 39 kDa specific chaperone protein resident in the ER, which prevents interaction of LRP1 with its ligands until they reach the Golgi 34 . The relatively acidic lumenal environment of the Golgi causes protonation of key histidine residues, and a structural rearrangement 'histidine switch' of RAP, which leads to its dissociation from LRP1 35 . RAP has been shown to be about 2-fold down-regulated in dorsal root ganglion (DRG) neurons following nerve injury 36 , suggesting that the interaction of LRP1 (and other family members) with their ligands might be dynamically regulated by the availability of RAP.
In the present study we show that α 2 δ -1 is a ligand for LRP1, and their interaction is modulated by RAP. This interaction regulates the trafficking, post-translational processing, and function of α 2 δ -1 as a calcium channel subunit. Our findings now provide a key insight and new tools to investigate the trafficking and function of the calcium channel α 2 δ subunits.

Results
Interaction between LRP1 and α 2 δ subunits. We first examined whether LRP1 is able to interact with α 2 δ -1 subunits. LRP1 is a 600 kDa pro-protein that is cleaved by the Golgi-associated protease furin into an ~85 kDa light-chain and a ~515 kDa heavy-chain; the two polypeptides remaining non-covalently-associated. The heavy-chain contains four homologous LBDs, which were identified to mediate binding to most of the LRP1 interacting proteins 37,38 . Each domain can be expressed independently, fused with the common 85 kDa transmembrane domain 39 . These are termed mini-receptors (LRP1-m1-m4), and have been shown to mimic the effect of full-length LRP1 38,39 . Most of the data obtained here involves experiments with a construct consisting of the fourth LBD of LRP1 (LRP1-m4).
Interaction between α 2 δ-1 and LRP1-m4 is reduced by RAP. The ER-resident chaperone RAP binds at multiple sites on LRP1, and has a dual role. It antagonises premature binding of LRP1 with its ligands in the ER and also promotes LRP1 folding and trafficking to the Golgi where ligand interaction occurs 41 . Premature binding to LRP1 in the ER was reported for other known LRP1 ligands and is considered to be a general property of LRP1 42 . The interaction of α 2 δ -1 with LRP1-m4 was significantly reduced, but not prevented, by co-expression of RAP (Fig. 1b, compare lane 1 with lanes 2 and 3 in the right upper panels, quantified in Fig. 1c). The amount of RAP used (1 μ g cDNA) was likely to be saturating, as doubling the amount of cDNA had no additional effect (Fig. 1b, compare lanes 2, and 3, quantification in Fig. 1c). Similar results were obtained with LRP1-m2 (data not shown), demonstrating that α 2 δ -1 binds indiscriminately to LBDs 2 and 4. As a negative control, instead of RAP we transfected the cells with GFP ( Fig. 1b; lane 1; Fig. 1c), which demonstrated that the effect of RAP on the interaction is specific and it is not due to reduced protein expression. Thus the continuing interaction of LRP1-m2 and -m4 with α 2 δ -1 when RAP is co-expressed is likely to represent an interaction that is relevant to α 2 δ -1 trafficking.
Interaction with LRP1 reduces the binding of gabapentin to α 2 δ subunits. We then examined the effect of LRP1-m4 co-expression on [ 3 H]-gabapentin binding to α 2 δ -1. LRP1-m4 and RAP were co-expressed with α 2 δ -1 in tsA-201 cells. Saturation binding experiments analysing the binding of [ 3 H]-gabapentin to membranes from cells expressing α 2 δ -1 and α 2 δ -1/LRP1-m4 demonstrate that LRP1-m4 does not alter the gabapentin binding affinity to α 2 δ -1 (Fig. 1d). The K D for α 2 δ -1 alone was 180.2 ± 23.0 nM, and for α 2 δ -1/LRP1-m4 it was 230.0 ± 53.0 nM (n = 6, P > 0.05). In contrast, the presence of LRP1-m4 resulted in a 38% decrease in the measured B max for [ 3 H]-gabapentin binding to α 2 δ -1 from 3.63 ± 0.56 pmol/mg protein, to 2.06 ± 0.52 pmol/mg protein when LRP1-m4 was co-expressed with α 2 δ -1 (n = 6, P = 0.0129). This suggests that when α 2 δ -1 interacts with LRP1-m4, the association may occlude the gabapentin binding site. α 2 δ-1 and LRP1-m4 bind in vitro. Further evidence for the interaction between α 2 δ -1 and LRP1-m4 was obtained from in vitro interaction experiments. Lysates from tsA-201 cells expressing HA-α 2 δ -1 were applied either to beads loaded with pre-bound Flag-LRP1-m4, or to control beads pre-incubated with WCL from untransfected cells under identical conditions (Fig. 2a). After washing, HA-α 2 δ -1 was found attached to LRP1-m4 containing beads, but not to control beads (Fig. 2b, lane 3). LRP1-m4 and α 2 δ-1 are associated on the cell surface in the presence of RAP. Since an interaction between α 2 δ -1 and LRP1-m4 is maintained in the presence of RAP (Fig. 1b,c), we hypothesised that the two proteins may remain associated on the cell surface. To test this, we applied Flag antibody to intact tsA-201 cells transfected with α 2 δ -1 and either Flag-LRP1-m4 or HA-LRP1-m4, with or without RAP; then incubated cells at 4 °C to prevent endocytosis, and immunoprecipitated Flag-LRP1-m4, following cell lysis. We found more immunoprecipitation of LRP1-m4 from the cell surface in the presence of co-transfected RAP, than in its absence . Western blots showing: left panels, WCL; right panels, immunoprecipitate (IP) with HA Ab. Top panels: Western blot against α 2 δ -1 (α 2 δ -1 mAb), bottom panels: western blot against HA for HA-LRP1-m2 or -m4. The IP western blot has been stripped from α 2 δ -1 mAb and reblotted against HA. Samples were not deglycosylated. (b) tsA-201 cells transfected with HA-α 2 δ -1 and Flag-LRP1-m4 (lanes 1-3), without RAP (lane 1) or co-transfected with 1 μ g or 2 μ g RAP (lanes 2 and 3). Ca V 2.2 and β 1b were co-transfected in all conditions. Samples were not deglycosylated. Left panels: Input WCL; right panels, co-IP with anti-Flag Ab. Top panels: HA-α 2 δ -1 (HA Ab), middle panels: Flag-LRP1-m4 (Flag Ab), lower panels: RAP reblot (anti-LRP1/RAP Ab), and anti-Akt loading control. (c) Quantification of co-IP measured as a ratio of the HA-α 2 δ -1 bands in the co-IP, relative to the input WCL (including experiment shown in (b)), in the presence of Flag-LRP1-m4, and in the absence (black bar) or presence of the two concentrations of RAP shown in (b) (blue and red bars). Mean ± SEM and individual data for n = 3 experiments, each normalised to the control in the absence of RAP. *** P < 0.001, 1-way ANOVA and Bonferroni's post-hoc test. (d) [ 3 H]-gabapentin binding curves, using membrane fractions from transfected tsA-201 cells, for binding to α 2 δ -1 (open squares, n = 6) and α 2 δ -1 + LRP1-m4 + RAP (black circles, n = 6). Mean ± SEM data were fit by the Hill equation: K D = 229.5 and 213.7 nM, respectively. B max = 4.02 and 1.76 pmol/mg protein, respectively. Full blots for all figure parts are shown in Supplementary information. . We also found that α 2 δ -1 co-immunoprecipitated with cell surface Flag-LRP1-m4, only in cells co-transfected with LRP1-m4 and RAP, and was not detectable without RAP (Fig. 2c, lower panel, lane 4 compared to lane 3). This also agrees with one of the main roles of RAP being to promote folding and trafficking of LRP1 and its ligands to the cell surface. LRP1 binds anchorless α 2 δ-1ΔC in vitro. We then utilised a soluble anchorless α 2 δ -1 construct (α 2 δ -1Δ C, Fig. 3a), which is truncated prior to the C-terminus so that it is not GPI-anchored, and has previously been characterized to be in part secreted as a soluble protein 43 . We found, using a dot blot technique previously employed to demonstrate interactions with LRP1 44 , that soluble α 2 δ -1Δ C-HA protein purified from conditioned medium ( Fig. 3b; arrow) and purified soluble LRP1-m2 LBD protein immobilised on nitrocellulose membrane (Fig. 3c), interact in vitro, as shown by the staining for α 2 δ -1Δ C-HA ( Fig. 3d; left panel). An identical amount of bovine serum albumin (BSA) protein immobilised on the same membrane adjacent to LRP1-m2 did not bind to α 2 δ -1Δ C-HA, demonstrating the specificity of the interaction ( Fig. 3d; left panel). As an additional control, membranes with immobilised BSA and LRP1-m2 LBD were treated under identical conditions except α 2 δ -1Δ C-HA was omitted, and the HA antibody signal was lost, showing that it is due to specific labelling of α 2 δ -1Δ C-HA by this antibody (Fig. 3d; right panel). Moreover, the interaction was dependent on structural features of LRP1-m2 LBD, because denaturation of the protein prior to its immobilisation on the membrane resulted in a decrease of the interaction with α 2 δ -1Δ C-HA (compare Fig. 3e with the left panel of Fig. 3d). Interaction between endogenous LRP1 and α 2 δ-1 in DRMs. Both α 2 δ proteins 7,8 and PrP 45 are GPI-anchored DRM-associated proteins, and both interact with LRP1. We therefore performed experiments using material from mouse brain, and found some LRP1 in DRMs from WT mouse brain, although most was in the soluble fractions (Fig. 4a, left top panel, white arrow). The same was true for RAP (Fig. 4a, bottom panels). Therefore, since most endogenous α 2 δ -1 is present in DRMs (Fig. 4a, left second panel, white arrow), we decided to use mouse brain DRMs as the input, to attempt to co-immunoprecipitate native α 2 δ -1 with LRP1. As controls, brains from α 2 δ -1 −/− mice were included (Fig. 4a, right panels), as well as control immunoprecipitation with an irrelevant IgG, and also co-immunoprecipitation of the known LRP1 ligand PrP. LRP1 was immunoprecipitated with LRP1 Ab ( Fig. 4b; top right panel, lanes 2 and 4).
The small amount of LRP1 in DRMs co-immunoprecipitated with a correspondingly small fraction of the total α 2 δ -1 in DRMs from α 2 δ -1 +/+ mice ( Fig. 4b; middle panel, lane 4). As a control, α 2 δ -1 was not co-immunoprecipitated with an irrelevant IgG ( Fig. 4b; middle panel, lane 3), or in identical experiments using α 2 δ -1 −/− brain DRMs ( Fig. 4b; middle panel, lanes 1 and 2). A similarly small fraction of PrP, relative to the total PrP that localised to DRMs, was co-immunoprecipited with LRP1 in parallel (  Fig. 5b). As shown previously 39 , the processing and surface expression of LRP1-m4 were increased in the presence of RAP, demonstrating its effects on the trafficking of LRP1 itself (Fig. 5a, compare lanes 2 and 3 in the bottom right panel). We also examined the cell-surface expression of a non-functional α 2 δ -1 construct with mutations in the metal ion-dependent adhesion site (MIDAS) motif of the VWA domain (α 2 δ -1-MIDAS AAA ) 4,10 . We have previously shown that this mutant demonstrates reduced cell surface expression, compared to WT α 2 δ -1 46 . We found here that plasma membrane expression of this mutant is not increased by co-expression of LRP1-m4 and RAP, unlike WT α 2 δ -1 (Fig. 5c, right panel; quantified in Fig. 5d). This suggests that during its trafficking α 2 δ -1 may require an intact VWA domain to be affected by LRP1.
Furthermore, in agreement with the cell surface biotinylation results, when tsA-201 cells were transfected with HA-α 2 δ -1 and LRP1-m4, with or without RAP, cell surface expression of HA-α 2 δ -1, detected immunocytochemically, was also increased by the presence of RAP (Fig. 5e, quantified in Fig. 5f).

RAP increases secretion of anchorless α 2 δ-1ΔC
in the presence of LRP1-m4. Secretion of C-terminally truncated soluble LRP1 constructs has been used previously to study the role of the interaction of RAP with LRP1 47 . In order to determine whether membrane anchoring of α 2 δ -1 was required for its interaction with LRP1, we utilised the anchorless α 2 δ -1 construct described above (α 2 δ -1Δ C, cartoon in Fig. 3a), which can be secreted as a soluble protein 43 . We studied the effect of LRP1-m4 and the role of RAP on the secretion of α 2 δ -1Δ C into the medium. When we co-expressed α 2 δ -1Δ C with increasing concentrations of RAP, we found that secretion of α 2 δ -1Δ C was increased compared to the control in the absence of RAP (Fig. 6a, lanes 3-5 compared to lane 2, quantified in Fig. 6b). Furthermore, RAP also antagonised the inhibitory effect of LRP1-m4 on α 2 δ -1Δ C secretion (Fig. 6a, lane 7 compared to lane 6, quantified in Fig. 6b). This result shows that RAP antagonised the interaction of α 2 δ -1Δ C with both exogenous LRP1-m4 and also endogenous LRP1 (or related endogenous members of the LDL-receptor family), to promote secretion of α 2 δ -1Δ C.
We have previously found that while most is secreted, some α 2 δ -1Δ C remains associated with the cell surface of tsA-201 cells and DRG neurons by non-covalent interactions, and we speculated that another molecular interactor mediates this cell-surface association, in the absence of GPI-anchoring 43 . In the light of the interaction with LRP1 identified here, and the modulatory role of RAP in this interaction, we decided to investigate the effect of RAP on the cell surface association of α 2 δ -1Δ C. Co-expression of RAP with α 2 δ -1Δ C in tsA-201 cells strikingly reduced the amount of α 2 δ -1Δ C on the cell surface, as measured by cell surface immunofluorescence, compared to the control without RAP (Fig. 6c,d), without significantly affecting the total expression of α 2 δ -1Δ C measured in permeabilised cells (data not shown). This suggests that the attachment of α 2 δ -1Δ C might be mediated by an endogenous molecule belonging to the LDL-receptor family, to which α 2 δ -1Δ C remains associated through the trafficking pathway to the cell surface, such that the binding can be antagonised by RAP, allowing increased secretion of α 2 δ -1Δ C. LRP1 and RAP affect the N-glycosylation and proteolytic processing of α 2 δ-1. Next we examined whether LRP1-m4 and RAP would affect the N-glycosylation pattern and proteolytic cleavage of α 2 δ -1, in order to provide insight into where their interaction occurs. The α 2 δ -1 protein goes through a series of post-translational processing steps in the ER and Golgi, including N-glycosylation at multiple sites, and proteolytic cleavage of the pro-form of α 2 δ -1 into α 2 and δ polypeptides (Fig. 7a) 48 . The core N-linked glycans, added co-translationally to the nascent proteins in the ER, are rich in mannose residues and termed high mannose-type or immature glycans. They are subsequently trimmed and modified in the Golgi to mature N-glycans 49 , although some membrane proteins by-pass this route 50 . The enzyme Endoglycosidase-H (Endo-H) discriminates between the two types of N-glycosylation, being able to strip the high mannose N-glycans from proteins, but unable to remove the mature N-glycans, formed by processing in the Golgi. The enzyme Peptide-N-Glycosidase-F (PNGase-F) is able to remove all types of N-glycosylation from proteins, without discrimination.
Since α 2 δ -1 is incompletely proteolytically cleaved when expressed in heterologous systems 43 , we investigated the N-glycosylation pattern of the cleaved α 2 -1 and uncleaved α 2 δ -1 proteins, as a marker of whether the mature cleaved α 2 -1 is derived from the ER or Golgi. We found that cleaved α 2 -1 is resistant to Endo-H (Fig. 7b; lane 1H, closed arrow) but sensitive to PNGase-F (Fig. 7b; lane 1F, closed arrow), showing that it contains mostly mature N-glycans, and suggesting that the proteolytic cleavage of α 2 δ -1 occurs after trimming of the glycans in the Golgi. In line with this result, the uncleaved pro-α 2 δ -1 form is mostly sensitive to both Endo-H (Fig. 7b; lane 1H, open arrow) and PNG-ase-F (Fig. 7b; lane 1F, open arrow), suggesting that this protein species derives mainly from ER, and has not yet been exported through the Golgi, where the N-glycans are modified. Furthermore, co-expression of LRP1-m4 resulted in a reduction of mature Golgi-type N-glycosylation on α 2 -1 (Fig. 7b, compare lanes 1H and 2H), and reduced α 2 δ -1 proteolytic cleavage (Fig. 7b, compare lanes 1F and 2F, closed arrow), suggesting that it retards forward trafficking of α 2 δ -1. In contrast, when RAP was co-expressed with LRP1-m4, it antagonised this effect, and increased both mature N-linked glycosylation (Fig. 7b, lane 3H) and the cleavage of α 2 δ -1 (Fig. 7b; compare lanes 3F and 2F). Quantification of proteolytic cleavage of α 2 δ -1 showed that the additional presence of RAP increased α 2 δ -1 cleavage ~2-fold compared to LRP1-m4 alone (Fig. 7c). This is in agreement with the hypothesis that in the absence of RAP, an interaction between LRP1 and α 2 δ -1 early after their translation results in ER retention; this is antagonised by RAP, allowing forward trafficking. It is likely that RAP, by increasing the folding of LRP1, allows trafficking of both LRP1 and α 2 δ -1 to the Golgi. Our results suggest that proteolytic cleavage of α 2 δ -1 occurs in Golgi or post-Golgi compartments. It is also likely that α 2 δ -1 is trafficked in a complex with LRP1-m4, as we still detected interaction of LRP1-m4 with α 2 δ -1 on the cell surface (see Fig. 2c). This provides further evidence for the dual role of RAP in antagonising premature interaction between LRP1 and its ligands in the ER, as well as in trafficking of both LRP1 and its ligands.
Effect of LRP1 and RAP on calcium currents and cell surface expression of Ca V 2.2 and α 2 δ-1.
We then examined whether the effects of LRP1-m4 and RAP on trafficking and proteolytic processing of α 2 δ -1 . White arrows on blot shows a minor fraction of the uncleaved pro-α 2 δ -1, which is Endo-H resistant. Note that Endo-H-resistant (Golgi-associated) species of α 2 δ -1 are reduced upon co-expression of Flag-LRP1-m4 (middle panel; lane 2H), which is reversed by co-expression of RAP (middle panel; lane 3H). PNGase-F treatment (right panel; lanes 1F-3F) caused removal of all types of N-glycans from both cleaved α 2 -1 and uncleaved α 2 δ -1 (arrows on right). Note that cleavage of α 2 -1 is reduced upon co-expression of Flag-LRP1-m4, and this is reversed by co-expression of RAP (α 2 -1 species; compare lanes 2F and 3F). Representative of n = 2 experiments. (c) Quantification of α 2 δ -1 proteolytic processing expressed as a ratio of cleaved α 2 -1 to total α 2 δ -1 (cleaved + uncleaved) in PNGase-F deglycosylated WCL, normalised to control conditions (open bar), in the presence of LRP1-m4 (black bar) or LRP1-m4 plus 1 μ g or 2 μ g RAP cDNA (blue and red bars respectively). Data are mean ± SEM with individual data points for n = 3 independent experiments. **P < 0.01, *P < 0.05, 1-way ANOVA and Bonferroni's post-hoc test. Full blots for all figure parts are shown in Supplementary information. would translate into effects on calcium currents. Ca V 2.2 was co-transfected with β 1b and α 2 δ -1, with either a control cDNA, LRP1-m4 alone or LRP1-m4 plus RAP. The peak I Ba current density in the presence of LRP1-m4 and RAP was 2.6-fold greater than in the absence of RAP, in agreement with the general hypothesis outlined above (Fig. 8a,b,c). Similar results were obtained using either an irrelevant protein vector or empty pcDNA3 vector in the control condition (data not shown).
Furthermore RAP alone produced a small increase in calcium currents, increasing the peak I Ba by 36% (Fig. 9a,b), and increased the amount of Ca V 2.2 on the cell surface, determined by α -bungarotoxin cell surface labelling of BBS-tagged Ca V 2.2 by 56.9% for 0.75 μ g RAP (Fig. 9c,d) 46 . Moreover, RAP alone also increased the cell surface expression of α 2 δ -1, determined by cell surface biotinylation, (Fig. 9e, compare lane 1 with lanes 3 and 4; quantified in Fig. 9f). The mean increase was 51.8% for 0.75 μ g RAP. These effects of RAP are likely to be mediated either via an influence on endogenous LRP1 or another member of this family, and suggest that endogenous RAP may be limiting in these circumstances.

Discussion
LRP proteins are widely recognised to be involved in trafficking and endocytosis of many protein ligands 21,51 , and loss of LRP1 in the central nervous system results in impaired synaptic function 29,30 . Furthermore, other LRP family members are also involved in a number of signalling pathways; for example LRP5/6 are co-receptors with the G-protein coupled receptor, frizzled, for Wnt signalling 22 . Here we show that the calcium channel auxiliary subunit α 2 δ -1 is a ligand for LRP1, which, when combined with its chaperone protein RAP, promotes the trafficking, cell surface expression, mature glycosylation and proteolytic cleavage of α 2 δ -1. We also provide evidence that LRP1 remains associated with α 2 δ -1 on the cell surface in the presence of RAP. The α 2 δ -1 subunit associates with Ca V 1 and Ca V 2 calcium channels, to promote their trafficking and voltage-dependent activation 6 . We find here that the LRP1/RAP combination also promotes functional expression of the neuronal (N-type) Ca V 2.2 channels, in terms of both cell surface expression and increased calcium currents. We have recently shown that proteolytic processing of α 2 δ into α 2 and δ is essential for this auxiliary subunit to enhance currents through Ca V 2.2 channels 48 , and we find here that proteolytic cleavage of α 2 δ -1 is increased by LRP1/RAP, which is likely to contribute to their effect on calcium currents. CACNA2D1, the gene encoding α 2 δ -1, is one of many genes whose expression is altered in sensory DRG neurons following peripheral nerve damage leading to neuropathic pain 15,52,53 . In contrast RAP is down-regulated in DRG neurons following sensory nerve injury 36 . The consequences of this for α 2 δ -1 trafficking and calcium channel function might be to retain some of the increased α 2 δ -1 in the ER because of premature interaction with LRP1. Relevant to this, we observed a substantial amount of α 2 δ -1 in the ER within rat DRG cell bodies following spinal nerve ligation 15 , and also observed uncleaved α 2 δ -1 in DRG somata 48 .
Of interest, the LRP1 gene has recently been identified as a susceptibility locus associated with common painful migraine 54 , and LRP1 has also been linked to the survival of Schwann cells and signalling in neuropathic pain 55 . Painful nerve injury also results in up-regulation of the extracellular matrix protein thrombospondin-4 56 . Thrombospondin family members are known LRP1 interacting proteins 32 . Indeed, we have previously confirmed this interaction, showing thrombospondin-4 to interact with the LRP1 mini-receptors used in the present study 40 . Thrombospondins have also been identified as binding partners for α 2 δ -1, by co-immunoprecipitation 19 . This interaction was found to be disrupted by gabapentin; and it was suggested that this might mediate a therapeutic action of gabapentin via inhibition of synaptogenesis 19 . In contrast, we found that there was no detectable interaction of the two proteins on the cell surface of transfected cells 40 . However the affinity of [ 3 H]-gabapentin binding to α 2 δ -1 was reduced by co-expression of thrombospondin-4; therefore we concluded there could be a weak interaction between co-expressed α 2 δ -1 and thrombospondin-4 in membranes derived from intracellular compartments. In the present study we have shown that LRP1 interacts with α 2 δ -1, both by immunoprecipitation and by pull-down with purified LRP1 mini-receptors. Therefore it is possible, since LRP1 has multiple LBDs, that LRP1 (or a related family member) could contribute to bridging an indirect interaction between α 2 δ -1 and thrombospondins, or indeed between PrP and α 2 δ -1 57 .
Although LRP1 is not primarily a DRM-associated protein, it has previously been observed in DRMs from some cell types 58 . Furthermore, it does interact with some DRM-associated proteins, for example with PrP 25 , which has been found to result in signalling in specific DRM domains 58,59 . LRP1 also affects PrP endocytosis in a RAP-dependent manner 25 . We find here that a small proportion of LRP1 is present in mouse brain DRMs, and here it co-immunoprecipitates a comparable fraction of α 2 δ -1 and PrP, both being established DRM residents. The relatively small co-immunoprecipitated fraction relative to the total α 2 δ -1 and PrP in DRMs is in line with the partial partitioning of LRP1 into DRMs.
LRP1 is known to be recruited to caveolae 59 , and to be involved in endocytosis 60 . Whether LRP1 and RAP also influence α 2 δ -1 endocytosis, in the same way as for PrP, and whether the same LRP1 molecule can bind both α 2 δ -1 and PrP ligands occupying different sites, remains to be determined in future studies.
LRP1 binds multiple ligands 37 , including von Willebrand factor 61 , and this contributes to the clearance of von Willebrand factor which is internalised from blood plasma by macrophages 62 . LRP1 also binds to the inserted VWA domain in α M β 2 integrin 51 . A poly-basic motif at the start of the VWA domain is implicated in this binding, rather than a direct interaction with the VWA domain MIDAS motif. It is of interest that one of the key motifs mediating gabapentin binding to α 2 δ -1 and α 2 δ -2 is a basic triple arginine sequence, just prior to its VWA domain 8,12,63 . In the present study we found that the number of gabapentin binding sites (B max ) associated with α 2 δ -1 was markedly reduced by co-expression of LRP1. It is thus possible that LRP1 binding to α 2 δ -1 involves the gabapentin binding pocket motif, thus occluding the gabapentin binding site.
Despite our evidence that endogenous LRP1 is a good candidate for interaction with α 2 δ -1, both during trafficking and on the cell surface, at this point we cannot exclude other endogenous members of LDL-receptor family of proteins, such as LRP5/6 which also interact with RAP 64 , and are involved in Wnt signalling 22 . It will be of interest in the future to determine whether these LRP family members interact with α 2 δ proteins.

Cell culture and transfection. The tsA-201 cells (European Collection of Authenticated Cell Cultures),
tested to be mycoplasma-free, were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal bovine serum (FBS), 1 unit/ml penicillin, 1 μ g/ml streptomycin and 1% GlutaMAX (Life Technologies, Waltham, MA). They were plated onto cell culture flasks, coverslips or glass-bottomed dishes (MatTek Corporation, Ashland, MA), coated with poly-L-lysine, and cultured in a 5% CO 2 incubator at 37 °C. Cells were transfected using Fugene6 (Promega, Fitchburg, WI), according to the manufacturer's protocol. For biochemical experiments cells were harvested 48 h after transfection, by washing cells in phosphate-buffered saline (PBS, pH 7.4, Sigma), detaching with a cell scraper into PBS, and pelleting by centrifugation at 1,000 × g.
For electrophysiology the DNA ratios for Ca V 2.2/β 1b/α 2 δ -1 were maintained constant within each set of experiments, usually at 3:2:2, and either empty vector or a non-functional membrane protein P2 × 2 K69A or Kir2.1 AAA were used in controls in place of LRP1-m4 or RAP, in order to maintain constant the total amount of cDNA transfected In some experiments β 1b-GFP was used in place of β 1b, to detect transfected cells. Site-directed mutagenesis was carried out using standard procedures, and all subcloning and mutations confirmed by sequencing.
Mice were housed in groups of no more than 5 on a 12 h:12 h light dark cycle; food and water were available ad libitum. Brains were obtained from male 10-week old α 2 δ -1 −/− mice and α 2 δ -1 +/+ littermates, obtained by breeding from heterozygotes, during the course of a previous study 67 , and frozen at − 80 °C until use. All experimental methods, procedures and protocols relating to breeding and obtaining tissue from these mice were approved by the UK Home Office and by the UCL local ethical committee. All methods were performed in accordance with the relevant guidelines and regulations.
Western blotting. Sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) was carried out as previously described 43 . The following secondary Abs were used for western blotting: goat anti-rabbit, goat anti-rat and goat anti-mouse Abs coupled to horseradish peroxidase (HRP) (Biorad). The signal was obtained by HRP reaction with fluorescent product (ECL 2; Thermo Scientific), and membranes were scanned on a Typhoon 9410 phosphorimager (GE Healthcare). Full blots for all figures are shown in Supplementary information.
Preparation of Triton X-100-insoluble DRMs and co-immunoprecipitation. The protocol was similar to that described previously 7,43 . All steps were performed on ice. Whole brains (without cerebellum) from WT and α 2 δ -1 −/− mice were used as the starting material. Tissue was homogenised using a Teflon homogeniser in Mes-buffered saline (MBS), containing 25 mM Mes (pH 6.5), 150 mM NaCl and protease inhibitors (PI, cOmplete, Roche; used according to manufacturer's instructions), containing 1% (v/v) Triton X-100 (Thermo Scientific), and left on ice for 1 h. The sample was then supplemented with 90% sucrose to 45% final concentration and overlaid with 10 ml of discontinuous sucrose gradient, consisting of 35% (w/v) sucrose in MBS (5 ml) and 5% (w/v) sucrose in MBS (5 ml), centrifuged at 138,000 × g for 18 h at 4 °C (Beckman SW40 rotor). Fractions (1 ml) were subsequently harvested from the top to the bottom of the tube, and DRM fractions from the gradient were washed free of sucrose by dilution in 25 volumes of ice-cold PBS, and ultracentrifugation (150,000 × g for 1 h at 4 °C) to pellet the DRMs to be used for co-immunoprecipitation as described below.
Cell surface biotinylation, cell lysis, deglycosylation and immunoblotting. The procedures were modified from those described previously 7 . Briefly, 48 h after transfection, tsA-201 cells were incubated for 30 min at room temperature with 0.5 mg/ml EZ-link Sulfo-NHS-LC-Biotin (Thermo Scientific) in PBS, and the reaction was quenched with 200 mM glycine. The cells were lysed for 45 min on ice in PBS, containing 1% Igepal; 0.1% SDS and PI. The WCL was then centrifuged at 20,000 × g at 4 °C for 20 min, and the pellet discarded. The supernatant was assayed for total protein (Bradford assay, Biorad). Immunoblot analysis was performed essentially as described previously 43 . Cleared WCL corresponding to 20-40 μ g total protein was diluted with Laemmli sample buffer 7 , supplemented with 100 mM dithiothreitol (DTT), incubated at 60 °C for 10 min and resolved by SDS-PAGE and western blotting. Biotinylated lysates (adjusted to between 0.5 and 1 mg/ml total protein concentration) were applied to 40 μ l prewashed streptavidin-agarose beads (Thermo Scientific) and rotated overnight at 4 °C. The beads were then washed 3 times with PBS containing 0.1% Igepal and, when required, the streptavidin beads were re-suspended in PNGase-F buffer (containing PBS, supplemented with 75 mM β -mercaptoethanol, 1% Igepal, 0.1% SDS, and PI) and deglycosylated for 3 h at 37 °C with PNGase-F (Roche Applied Science). When required, WCL were deglycosylated with PNGase-F under identical conditions described above or with Endo-H (NEB) following the manufacturer's instructions. The samples were then resuspended in an equal volume of 2 × Laemmli buffer with 100 mM DTT, followed by 10 min incubation at 60 °C, prior to analysis by SDS-PAGE and western blotting.
Co-immunoprecipitation. The assay was done as previously described with modifications 48 . Cell pellets were resuspended in co-immunoprecipitation buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Igepal, 0.5% Na deoxycholate, 0.1% SDS, 1 mM MgCl 2 , 1 mM CaCl 2, supplemented with EDTA-free PI), homogenised by 5 × passes through a 23 gauge syringe and 5 s sonication at 20,000 Hz, and then incubated for 1 h on ice. Insoluble material was removed by centrifugation at 20,000 × g for 30 min at 4 °C. The supernatant was incubated with 2 μ g/ml of anti-M2-Flag antibody overnight at 4 °C with constant rotation. 30 μ l prewashed A/G PLUS Agarose beads (Santa Cruz) were added to each tube and further rotated for 2 h at 4 °C. The beads were then pelleted by 500 × g centrifugation at 4 °C and washed twice with a co-immunoprecipitation wash high detergent buffer (1% Igepal, 20 mM Tris pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2, with EDTA-free PI), twice with co-immunoprecipitation wash high salt buffer (0.1% Igepal, 20 mM Tris pH 7.4, 500 mM NaCl, with EDTA-free PI), and twice with a low salt co-immunoprecipitation wash (0.1% Igepal, 20 mM Tris pH 7.4, with EDTA-free PI). Native LRP1 and α 2 δ were precipitated from whole brain DRMs concentrated as described above. DRMs were resuspended in a buffer consisting of 20 mM Tris pH 7.4, 150 mM NaCl, 1% 1-O-Octyl-β -D-glucopyranoside (OG; Sigma) and PI; then sonicated for 10 s at 20,000 Hz, rotated at 4 °C for 1 h and centrifuged at 20,000 × g at 4 °C. The supernatants were assayed for total protein, concentrations were equalised, and 20 μ g/ml of rabbit anti-LRP1 or 20 μ g/ml of rabbit non-specific IgG were added to the samples and rotated overnight at 4 °C. The samples were washed four times in the same buffer, but supplemented with 0.2% OG. The beads, with co-immunoprecipitated proteins, were resuspended in an equal volume of 2 × Laemmli buffer with 100 mM DTT, and analysed by SDS-PAGE and western blotting, together with aliquots of the initial WCL prior to co-immunoprecipitation.
Cell surface co-immunoprecipitation. The protocol was performed as described previously 25 , with modifications. 48 h after transfection, tsA-201 cells were rinsed twice with PBS and incubated with gentle rocking at 4 °C for 45 min with 4 μ g/ml monoclonal anti-M2-Flag antibody, diluted in a cold serum-free DMEM, supplemented with 10 mM HEPES pH 7.4. Cells were harvested and all subsequent procedures were performed on ice. The cell pellet was resuspended in co-immunoprecipitation buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Igepal, 0.5% Na deoxycholate, 0.1% SDS, 1 mM MgCl 2 , 1 mM CaCl 2 , with EDTA-free PI), homogenised as in co-immunoprecipitation, and incubated for 1 h on ice followed by centrifugation at 1000 × g for 5 min at 4 °C.
The supernatants were collected and assayed for total protein (Bradford assay) and the concentration was adjusted to 1 mg/ml with co-immunoprecipitation buffer. 40 μ l pre-washed A/G PLUS-Agarose beads (Santa Cruz) were added to each sample followed by rotation for 1 h at 4 °C. The beads were then pelleted by centrifugation (500 × g; 4 °C), washed twice with co-immunoprecipitation wash buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 0.5% Igepal, 1 mM MgCl 2 , 1 mM CaCl 2 , with PI) and the sample was resupended in an equal volume of 2 × Laemmli buffer with 100 mM DTT, and analysed by SDS-PAGE and western blotting.
Affinity purification of secreted α 2 δ-1ΔC-HA from conditioned media. tsA-201 cells were transfected with α 2 δ -1Δ C-HA and incubated for 48 h in a 5% CO 2 incubator at 37 °C in DMEM medium supplemented with 2% FBS, 1 unit/ml penicillin, 1 μ g/ml streptomycin and 1% GlutaMAX. The conditioned medium from 6 flasks was pooled, filtered (0.2 μ m membrane pore size) and concentrated (x 40 fold) on a 10 kDa Nominal Molecular Weight Limit (NMWL) Amicon column (Merck Millipore) following the manufacturer's instructions. The concentrated material was supplemented with 20 mM HEPES pH 7.4 and OG (Sigma) to 0.2% final concentration, and applied to a 1 ml agarose-anti-HA antibody column (Sigma), pre-equilibrated in wash buffer (20 mM HEPES pH 7.4, 200 mM NaCl, 0.2% OG), and rotated for 1 h at 20 °C. The column was washed extensively in wash buffer, and bound proteins were then eluted with 3 ml 100 mM glycine pH 2.5 in 0.5 ml fractions into tubes containing 50 μ l 1 M HEPES pH 7.4 for immediate neutralisation. The eluates were applied to 3 ml Slide-A-Lyzer ™ Dialysis Cassettes (3.5 K MWCO, Thermo Fisher Scientific) and dialysed overnight at 4 °C against 2 L of PBS, pH 7.4. Then aliquots of purified protein were analysed by SDS-PAGE and western blotting with anti-HA antibody, and on a silver stained gel, on which the concentration was estimated by running BSA standards of known concentration (not shown).
Dot blot binding assay. The assay was a modification of that described previously 44 . 0.2 μ g of purified Ligand binding cluster-II of LRP1 Recombinant Human protein (LRP1-m2 LBD; Thermo Fisher Scientific; dissolved to 0.1 mg/ml) or 0.2 μ g BSA as negative control were immobilised on dry nitrocellulose membranes (Biorad). Membranes were then blocked for 1 h at 20 °C with dot blot buffer (PBS supplemented with 0.5% Igepal, 3% BSA, 1 mM MgCl 2 , and 1 mM CaCl 2 ). Purified α 2 δ -1Δ C-HA was then diluted to 50 nM in dot blot buffer, and applied to the membranes for 1 h at 20 °C with gentle rocking. This was followed by three washes with dot blot wash buffer (dot blot buffer without BSA), and incubated for 1 h at 20 °C with anti-HA antibody diluted 1:1000 in dot blot buffer, three additional washes with dot blot wash buffer, and 30 min incubation at 20 °C with anti-rabbit-HRP diluted 1:3000 in dot blot buffer. After three final washes with dot blot wash buffer and a final wash with dot blot wash buffer without detergent, the proteins with bound antibodies were visualised with ECL 2.
In vitro binding technique. tsA-201 cells were either not transfected or transfected separately with cDNA encoding HA-α 2 δ -1 or Flag-LRP1-m4/RAP (cDNA ratio 10:1). Cell pellets were homogenised at 4 °C in binding buffer (PBS supplemented with 25 mM sucrose, 1.5 mM CaCl 2 , 1 mM MgCl 2 , EDTA-free PI, and 0.5% Igepal) as for co-immunoprecipitation, and incubated on ice for 1 h. The lysates were cleared by 20,000 × g centrifugation at 4 °C and supernatants were collected and assayed for total protein (Bradford assay). The lysates from Flag-LRP1-m4/RAP and untransfected cells were rotated overnight at 4 °C with 10 μ g/ml anti-M2-Flag antibody, then 50 μ l of beads (prewashed in binding buffer) were added, and the sample was rotated for 2 h at 4 °C followed by 3 × washes with wash buffer (binding buffer supplemented with 0.1% Igepal). HA-α 2 δ -1 lysate was then equilibrated to 20 °C, and applied to the beads, and the samples were incubated for 2 h at 20 °C followed by 3 × washes with wash buffer and elution by heating in 2 × Laemmli buffer with 100 mM DTT prior to analysis by SDS-PAGE and western blotting.
Imaging was performed on a Zeiss LSM 780 confocal microscope as described elsewhere 46 . Images were obtained at fixed microscope settings for all experimental conditions of each experiment. Images (1024 × 1024 pixels) of tsA-201 cells were obtained using a 40× or 63× objective and an optical section of 0.8-1 μ m. Every cell identified as transfected was included in the measurements, to ensure lack of bias. Images were analysed using imageJ (imagej.net) using a modification of the procedure described previously 7,43 . Surface labelling in non-permeabilised cells was measured using the freehand line tool and manually tracing the surface of the cell. The value of the mean pixel intensity in different channels was measured separately and background was subtracted by measuring the intensity of an imaged area without transfected cells. All data were then normalised to the appropriate positive control for each experiment before combining experiments.
[ 3 H] gabapentin binding assay. Binding of [ 3 H]-gabapentin to membranes from tsA-201 cells was carried out essentially as previously described 40 , in a final volume of 250 μ l at room temperature for 45 min. The cDNA for Kir2.1 AAA was used in control transfections to maintain the amount of total cDNA constant, and in order to have cells expressing similar amounts of protein under the experimental conditions used. Western blots for α 2 δ -1 expression were quantified for every experiment to ensure similar expression levels. Membrane fractions (60 μ g of protein per tube) were incubated with various concentrations of [ 3 H]-gabapentin (specific activity 36 Ci/mmol, American Radiolabeled Chemicals, St. Louis, MO, USA) in 10 mM HEPES/KOH pH 7.4, then rapidly filtered through GF/B filters, pre-soaked with 0.3% polyethyleneimine. Filters were washed three times with 3 ml ice-cold 50 mM Tris/HCl, pH 7.4 and counted on a scintillation counter. Concentrations of [ 3 H]-gabapentin greater than 50 nM were achieved by adding non-radioactive gabapentin and correcting the specific binding by the dilution factor 4 . Non-specific binding was determined in the presence of 20 μ M non-radioactive gabapentin. Data points were determined in triplicate, and data were analysed by fitting specific binding to the Hill equation 40 .
Data analysis. Data are given as mean ± SEM, with individual data points if < 30. Statistical comparisons were performed using either Student's t test, paired t test, Wilcoxon's matched pairs signed rank test, ANOVA with appropriate post-hoc test, as stated, using Graphpad Prism 5.