Differential upregulation in DRG neurons of an α2δ-1 splice variant with a lower affinity for gabapentin after peripheral sensory nerve injury

Summary A minor α2δ-1 splice-variant, (ΔA+BΔC), with reduced affinity for gabapentin, is differentially upregulated in DRG neurons following ligation. This could influence the efficacy of gabapentin.


a b s t r a c t
The a 2 d-1 protein is an auxiliary subunit of voltage-gated calcium channels, critical for neurotransmitter release. It is upregulated in dorsal root ganglion (DRG) neurons following sensory nerve injury, and is also the therapeutic target of the gabapentinoid drugs, which are efficacious in both experimental and human neuropathic pain conditions. a 2 d-1 has 3 spliced regions: A, B, and C. A and C are cassette exons, whereas B is introduced via an alternative 3 0 splice acceptor site. Here we have examined the presence of a 2 d-1 splice variants in DRG neurons, and have found that although the main a 2 d-1 splice variant in DRG is the same as that in brain (a 2 d-1 DA+B+C), there is also another a 2 d-1 splice variant (DA+BDC), which is expressed in DRG neurons and is differentially upregulated compared to the main DRG splice variant a 2 d-1 DA+B+C following spinal nerve ligation. Furthermore, this differential upregulation occurs preferentially in a small nonmyelinated DRG neuron fraction, obtained by density gradient separation. The a 2 d-1 DA+BDC splice variant supports Ca V 2 calcium currents with unaltered properties compared to a 2 d-1 DA+B+C, but shows a significantly reduced affinity for gabapentin. This variant could therefore play a role in determining the efficacy of gabapentin in neuropathic pain.
Ó 2013 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

Introduction
Voltage-gated Ca 2+ channels of the Ca V 1 and Ca V 2 families contain 3 subunits: the pore-forming a1 subunit, together with 2 auxiliary subunits, b and a 2 d, both of which increase the functional expression of the channels [4,11,13,20,21]. The a 2 d subunits are each the product of a single gene (CACNA2D1-4), encoding an a 2 d preprotein, which is posttranslationally processed into a 2 and d [22]. We have recently shown that a 2 d subunits can form glycosylphosphatidylinositol-anchored proteins [19], which are constitutively endocytosed and reinserted into the plasma membrane via the recycling endosomes [5,48]. We have identified that the mechanism whereby a 2 d-1 and -2 subunits enhance plasma-membrane expression of calcium channels involves the metal ion-dependent adhesion site (MIDAS) motif in their von Willebrand factor A (VWA) domain [12,31], and it is also important for the enhancement of presynaptic vesicular release [31].
The a 2 d-1 subunit has a widespread distribution; both mRNA and protein are found in neuronal tissue, heart, skeletal, and smooth muscle [3]. In dorsal root ganglion (DRG) neurons, a 2 d-1 is the main a 2 d subunit expressed [5,16,36]. The a 2 d-1 protein is upregulated following various types of peripheral nerve injury [5,39,49], and this upregulation is essential for the rapid development of the subsequent behavioural mechanical hypersensitivity seen in animal models [41]. By contrast, Ca V 2.2 mRNA and protein are not reported to be consistently upregulated following sensory nerve damage [1,36,49], suggesting that upregulated a 2 d-1 protein enhances Ca V 2.2 trafficking and presynaptic function.
The a 2 d-1 and a 2 d-2 subunits bind to the gabapentinoid drugs, gabapentin and pregabalin. These were developed as antiepileptic drugs and are also widely used in the treatment of various forms of neuropathic pain [46]. The a 2 d-1 subunit has been shown to represent the target for these drugs in the alleviation of hyperalgesia in experimental models of neuropathic pain [24]. We have found that gabapentin and pregabalin reduce the trafficking of a 2 d subunits both in vitro and in vivo [5,30,48], which is likely to represent one of their main mechanisms of action. Furthermore, a 2 d-1 has been reported to interact with thrombospondins, a family of extracellular matrix proteins, and this may influence its trafficking as well as the effect of gabapentinoid drugs [23].
Alternatively spliced isoforms of a 2 d-1 in different tissues have been observed previously [10,22]. It was reported that the mouse cacna2d1 gene has 3 alternatively spliced regions (A, B, and C), and 5 splice variants were identified [3] (Fig. 1A). In the present study we have examined the hypothesis that there may be differential upregulation of specific splice variants of a 2 d-1 following peripheral nerve damage, and that this might potentially contribute to either the state dependency [14,25] [17] was also included to identify transfected cells from which electrophysiological recordings were made. Transfection was performed as described previously [40].
In control experiments where a 2 d was omitted, the ratio was made up with empty vector.

Isolation of detergent-resistant membranes (DRMs)
This procedure was performed as previously described [19,32]. Briefly, confluent cells from 6 175-cm 2 flasks (72 hours after transfection), or pelleted homogenate from rat whole brain, were resuspended in 1.5 mL 2-(N-morpholino)ethanesulfonic acid (MES)-buffered saline containing 1% (2% for brain tissue) Triton X-100 (Perbio, Tattenhall, Cheshire, UK) and left on ice for 1 hour. After the addition of an equal volume of 90% (w/v) sucrose, the sample was overlaid with a 10-mL discontinuous sucrose gradient and centrifuged at 140,000 Â g for 18 hours at 4°C. One-mL fractions of the gradient were harvested from the top to the bottom of the tube, washed free of sucrose by dilution in MES-buffered saline and subsequent centrifugation. The detergent-resistant membranes (DRMs) were prepared from pooled fractions 4, 5, and 6. The resulting pellet was resuspended in 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) 10 mM pH 7.4 and complete protease inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany). The isolated DRM fractions were collected for Western blot analysis and used in the 3 H-gabapentin binding assay.

Immunoblotting
Western blotting was performed as described previously [40]. The following primary antibodies were used: anti-a 2 -1 (1:1000; mouse monoclonal, Sigma-Aldrich, St. Louis, MO, USA) and antiflotillin-1 (1:2000; mouse monoclonal, BD Biosciences, Franklin Lakes, NJ, USA). The secondary antibody used was goat anti-mouse coupled to horseradish peroxidase (Bio-Rad Laboratories, Inc., Hercules, CA, USA). Protein bands were quantified using Image J software (rsb.info.nih.gov). All a 2 d-1 splice variant expression lev-els were corrected for background and normalized to DA+B+C protein expression on the same gel.

3 H gabapentin binding assay
Binding of 3 H-gabapentin to DRM fractions from transfected tsA-201 cells was carried out in a final volume of 250 lL at room temperature for 45 minutes. DRMs (4 lg protein per tube) were incubated with various concentrations of 3 H-gabapentin [45] (30-50 Ci/mmol, American Radiolabeled Chemicals, St. Louis, MO, USA) in 10 mM HEPES/KOH pH 7.4, then rapidly filtered through glass fibre (B grade, Whatman) filters presoaked with 0.3% polyethyleneimine. Filters were washed 3 times with 3 mL ice-cold 100 mM Tris/HCl pH 7.4 and counted in a scintillation counter. Concentrations of 3 H-gabapentin >10 nM were achieved by adding nonradioactive gabapentin and correcting the specific binding by the dilution factor [12]. Nonspecific binding was determined in the presence of 100 lM nonradioactive gabapentin. Data points were determined in triplicate, and data were analysed by fitting specific binding to the equation for a rectangular hyperbola.

Spinal nerve ligation (SNL)
A total of 17 male Sprague-Dawley rats (Central Biological Services, University College London, London, UK) weighing 130-150 g at time of surgery were employed for this study. All experimental procedures were approved by the UK Home Office and followed the guidelines of the International Association for the Study of Pain [51]. Selective spinal nerve ligation (SNL) surgery was conducted as previously described [34]. Briefly, the left L5 and L6 spinal nerves were isolated and tightly ligated with 6-0 silk thread under isoflurane anaesthesia (50% O 2 : 50% N 2 O). Haemostasis was confirmed and the wound was sutured. After surgery the animals were allowed to recover and housed at a maximum of 5 per cage. Food and water were available ad libitum. The foot posture and general behaviour of the operated rats were monitored throughout the postoperative period, and the development of mechanical hypersensitivity was confirmed at 7 days post surgery in the affected limb ipsilateral to the ligation, as previously described [5].

Enrichment of small and large DRG neurons using density gradient centrifugation
In order to obtain sufficient material, both L5 and L6 DRG were extracted from the ipsilateral and contralateral sides of 2 rats, either naïve animals or 7 days after SNL. Nerve roots were trimmed and the isolated ganglia were incubated in 5 U/mL of collagenase (Sigma-Aldrich) in 2 mL of Hanks balanced saline solution in a shaking water bath at 37°C for 30 minutes, and then triturated. The cell suspension was sedimented at 4°C for 5 minutes at 200 Â g. The pellet was resuspended in culture medium composed of Dulbecco's Modified Eagle's Medium: Nutrient Mixture F-12 (Gibco, Invitrogen, Life Technologies, Grand Island, NY, USA), penicillin (100 U/mL), streptomycin (100 lg/mL), L-glutamine, (2 mM), nerve growth factor (100 ng/mL, Sigma-Aldrich), and 10% foetal calf serum. Four mL of DRG neurons in culture medium was filtered through a 100-lm cell strainer (BD Biosciences) and layered onto 5 mL of a solution of 40% Ficoll (Histopaque-1077, Sigma-Aldrich) and 60% phosphate-buffered saline containing 4.2 mM NaHCO 3 . The final pH was adjusted to 7.3-7.4 with HCl. The preparation was centrifuged at 100 Â g for 15 minutes at 4°C, and DRG neurons were separated according to their size into a low-density fraction (LDF) and a high-density fraction (HDF) [28]. The 2 fractions obtained were diluted to 10 mL with culture medium and then centrifuged at 200 Â g for 5 minutes at 4°C. The LDF and HDF pellets, enriched in viable small and large neurons, respectively, were . Leu-Glu-Ala, which is the end of a 2 -1) was submitted to Phyre2 [33] for structure prediction (http://www.sbg.bio.ic.ac.uk/phyre2). The amino acid numbering includes the N-terminal signal sequence of a 2 d-1. Region I ($amino acids 491-607) was predicted with $99% confidence, and modelled on a number of bacterial CSDs, including the extracellular domain of the Bacillus subtilis CSD (mmhk1s-z2) and the putative sensory box/ggdef protein from Vibrio parahaemolyticus. Region II ($amino acids 687-886) was modelled with a predicted $98% confidence on a separate subset of CSDs, including the C4-dicarboxylate transport sensor protein dctb, and the mcp_n and cache domains of methyl-2 accepting chemotaxis protein from Vibrio cholerae. The regions between domains I and II and beyond II are not modelled with high confidence. The approximate positions of the two spliced regions B and C are indicated. They are both likely to be in exposed loops, B being within the first chemosensory domain, and C being between the 2 domains. A is not present in this model. obtained and immediately used for RNA extraction, or plated on 13-mm-diameter poly-D-lysine-coated (10 mg/mL; Sigma-Aldrich) glass coverslips. Cells were incubated at 37°C (95% air, 5% CO 2 ), and observed in bright field after 2-5 hours to measure the density and diameter of the neurons in each fraction. Only phase-bright cells with a typical DRG morphology (large, clear nucleus with prominent nucleolus) were considered for statistical analysis. Larger cells were often slightly elliptical in cross-section, and the diameter was taken as the mean of the major and minor axes. Quantification of neurofilament-200 (NF-200) mRNA, a well-established marker for medium to large neurons [26,29] was employed to confirm the successful separation of neuronal fractions.

Quantification of splice variants and NF-200 by quantitative polymerase chain reaction (PCR) and capillary electrophoresis (CE)
RNA isolation and reverse-transcription polymerase chain reaction (RT-PCR) was carried out as follows. Total RNA was extracted from individual L4 or L5 pulverized frozen DRG 7 days after SNL or sham surgery, and pelleted LDF and HDF after density gradient centrifugation. RNA was isolated using the RNeasy Protect Mini Kit (Qiagen, GmbH, Hilden, Germany). RNA concentrations and purity were determined spectrophotometrically. RT was performed on 1 lg of total RNA using Superscript III reverse transcriptase (Invitrogen, Carlsbad CA, USA), using random hexamer primers (Promega, Madison WI, USA) and RNaseOUT (Invitrogen).
In order to quantify the a 2 d-1 splice variant expression pattern in rat tissue ( Primers were designed either manually or using the open-source software Pri-mer3. The techniques were optimized in a series of preliminary experiments for cycle number and starting quantity to ensure points of measurement were acquired during the linear phase of amplification ( Supplementary Fig. 1, and data not shown).
Capillary electrophoresis (CE) was carried out following dilution with deionized water of aliquots of RT-PCR products, and further analysed by laser-induced fluorescence (LIF). A detailed description and validation of the CE/LIF technique used for quantitative analy-sis of RT-PCR products was reported previously [43]. One lL of sample was diluted with 12 lL HiDi-Formamide (Applied Biosystems, Foster City CA, USA), and 0.5 lL of GeneScan 400HD ROX Size Standard (Applied Biosystems). Amplified PCR products were separated on an ABI 3100 Avant Genetic Analyzer (Applied Biosystems) running a 50-cm capillary with 3100 POP-6 polymer (Applied Biosystems). Each sample was injected in triplicate from separate wells on the plate. The amplified products were sized and quantified in GeneMapper v3.5 software using the Local Southern method. This analysis method was chosen for its reciprocal relationship between fragment size and mobility. The relative abundance of mRNA was determined as the ratio of integrated peak area for each PCR product relative to that of coamplified TBP in order to allow a direct comparison between different preparations. Then, data from the ipsilateral side were normalized to their respective contralateral side. The percentage of the total transcript represented by the minor DRG isoform (DA+BDC) was calculated as 100⁄ pa DA+BDC /(pa DA+B+C +pa DA+BDC ) [44]. Data were analysed using Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) and SPSS statistical software (SPSS Inc, Chicago IL, USA) or GraphPad Prism 4 (GraphPad Software, San Diego, CA, USA). Statistical significance was determined using the nonparametric Mann-Whitney U test and one-way analysis of variance with a suitable post hoc analysis. P values <0.05 were considered as statistically significant. All data are presented as mean ± SEM for the indicated number of experiments.

Distribution of different splice variants of a 2 d-1 in rat tissue
Splice variants of a 2 d-1 containing the 3 alternatively spliced regions (A, B, and C) were identified previously in mouse tissue [3]. In that study it was reported that regions A and B were part of the same exon, with alternative 3 0 splice acceptor sites, which spliced region A in or out [3]. However, the currently available human, mouse, and rat genomic sequences from Ensembl, ENSG00000153956 (human), ENSMUSG00000040118 (mouse), and ENSRNOG00000033531 (rat), indicate that regions A and B are in separate exons, with region A in rat being encoded by exon 18a, and B representing an alternative 3 0 splice acceptor site (start site) of exon 19. A diagram of potential splice variants is given in Fig. 1A.
The following cDNAs were assembled by standard molecular biological techniques in the rat a 2 d-1 backbone: +A+B+C, +A+BDC, DA+B+C, DADB+C, DADBDC, and DA+BDC. All constructs gave rise to products of the expected size with the PCR primers used in this study ( Fig. 1B and data not shown). We then examined their distribution in adult rat tissues. DADBDC was found to be the main splice variant in rat heart, but a number of other splice variants were also identified by the much more sensitive CE/LIF, demonstrating that this method can be used for identifying and quantifying a 2 d-1 splice variants (Fig. 1B, C). We found +A+BDC in skeletal muscle and DA+B+C in cerebral cortex (Fig. 1B), as shown previously for mouse tissues [3]. We also found the same brain splice variant, DA+B+C, in DRG (Fig. 1B).
3.2. Location of the spliced regions within a 2 d-1 The a 2 d subunits contain 2 domains with homology to bacterial extracellular chemosensory domains (CSDs) or Cache domains [2], which are downstream of the VWA domain. Bacterial extracellular CSDs bind a variety of nutrients and are generally associated with an intracellular histidine kinase signalling complex [15]. The plant ethylene receptor ETR1 is also a member of this family [15].
In a 2 d-1, the first of these CSDs is situated between amino acids $491-607 [21]. Fig. 1D gives the approximate locations of the spliced regions, and Fig. 1E [33] illustrates a homology model of the 2 CSDs, showing that regions B and C are situated within a loop in the first CSD of rat a 2 d-1 (I), and in the linker between CSD I and II, respectively. The model in Fig. 1E was generated using a sequence from which region A was absent, but A would be situated just upstream of region B. It is possible that within a 2 d subunits, these domains might be important for their ability to bind small ligands, including gabapentin [21].

Determination of 3 H-gabapentin binding affinity for a 2 d-1 splice variants
We then examined whether there were differences in 3 H-gabapentin binding affinity between the different a 2 d-1 splice variants ( Fig. 2). All determinations were performed on concentrated DRM fractions following expression in tsA-201 cells, since we have shown previously that there is a large increase in apparent affinity of both a 2 d-1 and a 2 d-2 for 3 H-gabapentin in DRMs [18,30].
To ensure that the radioligand binding experiments were not influenced by differential expression of a 2 d-1 protein between the different splice variants used, the relative expression level in DRMs of each splice variant was analysed by Western blot. All splice variants partitioned similarly into DRM fractions, and their level of expression was also found to be similar ( Fig. 2A, B). We found that both DADBDC (one of the splice variants found in heart) and DA+BDC showed significantly lower affinity for 3 H-gabapentin (K D values of 211 and 153 nM, respectively), compared to the major brain and DRG splice variant DA+B+C (Fig. 2C, D). Both these splice variants with lower affinity lacked region C, whereas all the splice variants containing region C, including DA+B+C, showed a higher affinity for 3 H-gabapentin, with K D values between 80 and 102 nM (Fig. 2C, D). Thus, there was a nearly 3-fold difference in affinity between the splice variants with the highest and lowest Gabapentinoid drugs are used to treat neuropathic pain, so we wondered whether there were changes in splice variant expression in DRG following SNL, which could potentially influence the effi-  The y-axis indicates the fluorescence signal peak height, which corresponds to the expression level of the respective splice variant. mRNA levels of DA+B+C and DA+BDC are higher in the side ipsilateral to SNL (B) than in the contralateral side (A). Note that the expression of the housekeeping gene (TATA-box binding protein) remains at a constant level on both sides. (C) SNL (7 days) leads to increased expression of the major a 2 d-1 (DA+B+C) splice variant in L5 DRG. Data are expressed as percentage of relative peak areas of the ipsilateral side normalized to the respective contralateral side. The relative increase is shown in L4 DRG from sham-operated rats (black bar, n = 6), L4 DRG from L5/L6 SNL rats (white bar, n = 11), L5 DRG from sham-operated rats (grey bar, n = 6), and L5 DRG after SNL (cross-hatched bar, n = 7). There was no significant difference in L4 DRG between SNL and sham-operated animals. Error bars represent SEM. Analysis of variance (ANOVA), F = 9.012, P < 0.0001, and Gabriel post hoc analysis: ⁄⁄ P < 0.01. (D) A pronounced upregulation of the a 2 d-1 splice variant DA+BDC was observed in ipsilateral L5 DRG 7 days after SNL. Data are expressed as percentage of relative peak areas of the ipsilateral side normalized to the respective contralateral side. The relative increase is shown in L4 DRG from sham-operated rats (black bar, n = 6), L4 DRG from SNL rats (white bar, n = 11), L5 DRG from sham-operated animals (grey bar, n = 6) and L5 DRG after SNL (cross-hatched bar, n = 7). Although DA+BDC mRNA levels are 2.4-fold higher in SNL L4 compared to L4 sham-operated animals, this was not statistically significant (P = 0.814). Error bars represent SEM. ANOVA, F = 11.12, P < 0.0001, and Gabriel post hoc analysis: ⁄⁄⁄ P = 0.001. (E) Comparison of the percentage of total transcript represented by DA+BDC splice variant in DRG ipsilateral to SNL, in L4 from sham-operated rats (black bar, n = 6), L4 DRG from SNL rats (white bar, n = 11), L5 DRG from sham-operated animals (grey bar, n = 4), and L5 DRG after SNL (cross-hatched bar, n = 8). The data show the pronounced shift in favour of the short splice variant in L5 after SNL. Data are expressed as percentage of relative peak areas for DA+BDC mRNA transcripts normalized to the sum of DA+B+C and DA+BDC peak areas. Error bars represent SEM. ANOVA, F = 9.39, P = 0.000, and Gabriel post hoc analysis: ⁄⁄ P < 0.01, ⁄⁄⁄ P < 0.001. cacy of these drugs. We chose 7 days after SNL as the time point for our analyses because we found previously that the increase in a 2 d-1 mRNA was not significantly different between 7 and 14 days after SNL, and we also noted less variability in the increase in a 2 d-1 mRNA at 7 than at 14 days [5].
In order to detect and quantify any minor a 2 d-1 splice variants in DRG, we used CE/LIF, and confirmed the presence of the major splice variant DA+B+C both in unaffected contralateral DRG (Fig. 3A), and in ligated DRG after SNL (Fig. 3B). The relative abundance of DA+B+C mRNA was quantified by CE/LIF after 27 cycles of RT-PCR (Fig. 3C). The DA+B+C-transcript level in ligated L5 showed a 3.6-fold increase, compared to the contralateral side. It was also significantly higher in L5 DRG following SNL, compared to L5 sham-operated animals, and compared to L4 SNL DRG and L4 DRG from sham-operated rats. There was no significant difference in a 2 d-1 levels in L4 DRG between SNL and sham-operated animals (Fig. 3C). Surprisingly, we also observed a novel splice variant in DRG neurons, DA+BDC a 2 d-1, which was particularly evident following the upregulation of a 2 d-1 mRNA that occurs after L5/L6 SNL (Fig. 3B).
DA+BDC was identified in DRG both by DNA sequencing following agarose gel separation (unpublished results), and by the size of the product on CE/LIF after RT-PCR (Fig. 3A, B). The mRNA level of the DA+BDC splice variant was quantified by CE/LIF after 29 cycles of RT-PCR (Fig. 3D). The expression level of DA+BDC was found to be $10 times higher (1083%) on the ipsilateral side, compared to the respective contralateral side in L5 DRG. Moreover, DA+BDC mRNA levels were significantly higher compared to L5 sham-operated animals, and L4 SNL and sham-operated rats. Although DA+BDC mRNA levels were 2.4-fold higher in SNL L4 compared to L4 sham-operated animals, this was not statistically significant (P = 0.814). Thus, the proportion of DA+B+C was significantly increased in L5, ipsilateral to SNL compared to sham-operated L5, but this was not the case for L4, which was not ligated. The DA+BDC splice variant represented 5.1 ± 0.5% of the total a 2 d-1 mRNA in shamoperated L5, and 8.4 ± 0.4% in SNL L5 DRG (Fig. 3E).
3.5. The relative abundance of DA+B+C and DA+BDC in different DRG classes DRG neurons are heterogeneous in morphology and the size of their somata relates to different functional subtypes. DRG with small and medium-sized somata form mainly nonmyelinated C fibres and myelinated Ad fibres, which conduct pain sensation, whereas nonnociceptive Ab fibres have larger somata [35]. a 2 d-1 is expressed in every sub-type of DRG neuron [5], but to date, the distribution of a 2 d-1 splice variants in DRG subpopulations is unknown. For the most prevalent DRG calcium channel a1 subunit, Ca V 2.2, it has been found that there is a differential distribution of a particular splice variant in small DRG neurons [1,6], and this splice variant is downregulated following SNL in rats [1]. Therefore, we wished to determine whether there was differential distribution in small and large DRG neurons of the a 2 d-1 DA+BDC splice variant, which showed reduced gabapentin binding affinity and a more pronounced upregulation following SNL, compared to the predominant a 2 d-1 splice variant DA+B+C.
To obtain populations enriched with small or large neurons, DRG neurons extracted from L5 and L6 of 2 SNL rats were separated according to their size, as described in Materials and Methods. The LDF was enriched in smaller neurons, while the HDF was enriched in larger neurons (Fig. 4A). The mean diameter for neurons in the LDF was 10.0 ± 0.2 lm (n = 418), with a pronounced peak at 5-10 lm. In the HDF, the smallest neurons (<15 lm in diameter) were almost absent, and the neuron diameter was up to 50 lm, with a peak at 21-25 lm. The mean diameter was 24.5 ± 0.7 lm (n = 260). As expected, NF-200 mRNA, a marker for large myelinated DRG neurons [42], measured by CE/LIF, was sig-nificantly higher in the HDF (13.9 ± 3.8 arbitrary units (AU)) compared to the LDF (1.8 ± 0.4 AU) (Fig. 4B), confirming that the HDF contains mainly larger neurons expressing NF-200.
We then used RT-PCR and CE/LIF to quantify the relative abundance of a 2 d-1 DA+B+C and DA+BDC in the small and large DRG neuron fractions after SNL. In agreement with the previous experiments, both DA+B+C (Fig. 4C) and DA+BDC (Fig. 4D) transcript levels were significantly increased on the ipsilateral side compared to the contralateral side after SNL. We further analysed whether the differential upregulation of DA+BDC compared to DA+B+C mRNA found in whole DRG (Fig. 3E) was present in both small and large neuron fractions. This was calculated as the percentage increase on the ipsilateral compared to the contralateral side, of the 2 transcripts in the small neuron fraction (white bars) compared to the large neuron fraction (grey bars). This analysis was performed to provide a direct comparison with the values found in whole DRG (Fig. 3C, D). We found that the increase of DA+BDC mRNA following SNL was significantly greater compared to the corresponding DA+B+C increase, only in the small neuron fraction, and not in the large neuron fraction (Fig. 4E). Thus, although upregulation following SNL of the 2 DRG transcripts DA+B+C and DA+BDC occurs in both small and large DRG neurons (Fig. 4C, D), the increased upregulation of DA+BDC compared to DA+B+C following SNL occurs preferentially in the small DRG neuron fraction containing nonmyelinated nociceptors (Fig. 4E).

Electrophysiological properties of the 2 splice variants present in DRG
It was important to determine whether a 2 d-1 DA+BDC was a functional splice variant, and we therefore compared the properties of Ca V 2.2 calcium channel currents co-expressed with b1b and either a 2 d-1 DA+BDC or DA+B+C. Ca V 2.2 was used because it is the main calcium channel in DRG neurons. We found that the properties of the currents formed from these combinations were very similar in terms of their ability to increase Ca V 2.2 current density compared to the absence of a 2 d (Fig. 5A-C), their ability to increase the inactivation rate of the currents (Fig. 5D, E), and their ability to hyperpolarize the steady-state inactivation of the currents (Fig. 5F). Since Ca V 2.1 is also present in DRG neurons, we also examined calcium currents formed by this channel. Similar results were obtained, in terms of the effect of the a 2 d-1 splice variants on current amplitude (Supplementary Fig. 2A-C), voltage-dependence of inactivation ( Supplementary Fig. 2D), and kinetics of inactivation ( Supplementary Fig. 2E).

Genomic arrangement of a 2 d-1 giving rise to splice variants
The cacna2d1 gene encoding mouse a 2 d-1 was found to contain 3 alternatively spliced regions within the a 2 moiety (A, B, and C) [3]. It was originally reported that regions A and B were part of the same exon, with alternative 3 0 splice acceptor site, which spliced region A in or out [3]. In this scenario, splice variants of a 2 d-1 should not exist in which region A is expressed without B.
However, our analysis of the mouse genomic sequence has found that A is encoded by exon 18a, and B is formed as a result of utilizing an alternative 3 0 splice acceptor site for exon 19. In confirmation of this, we have found splice variants in rat heart in which A is expressed without B (Fig. 1C).
In skeletal muscle, a 2 d-1 +A+BDC was the only splice variant detected [3]. We did not identify any minor splice variants in skeletal muscle in our study. We also found that the +A+BDC splice variant bound 3 H-gabapentin with high affinity, in agreement with previous findings of high-affinity binding of 3 H-gabapentin to native rat skeletal muscle a 2 d-1 [27]. Our results confirm that the lack of effect of gabapentin on skeletal muscle function is not a result of its inability to bind to the skeletal muscle a 2 d-1 isoform.
In mouse heart, a 2 d-1 DADB+C was previously found to be the most prevalent splice variant, although other splice variants were also observed, specifically DA+B+C, DADBDC, and DA+BDC [3]. In this study we found that DADBDC was the predominant isoform in rat heart tissue, but we also found 6 other splice variants from L5/L6 DRG, ipsilateral and contralateral to SNL. Data are expressed as DA+B+C or DA+BDC relative peak areas. Transcript levels in small (white bars) or large (grey bars) DRG neurons prepared from L5/L6 DRG, both contralateral (hatched bars) and ipsilateral (solid bars) to SNL. DA+B+C and DA+BDC transcript levels were significantly higher on the ipsilateral side compared to the contralateral side in both the small neurons (4.0 ± 0.6-fold for DA+B+C and 9.2 ± 2.4-fold for DA+BDC), and in the large neurons (4.0 ± 0.7-fold for DA+B+C and 7.4 ± 2.4-fold for DA+BDC). Error bars represent SEM. Statistical analyses were performed using 2 sample unpaired Student's t-test, for the data indicated. (E) Comparison of the upregulation in ipsilateral compared to contralateral L5 and L6 DRG of the 2 DRG a 2 d-1 transcripts, 7 days after SNL. In the small neuron fraction (white bars), the extent of upregulation of DA+BDC (right, n = 6) is significantly greater than that of DA+B+C (left, n = 7). In contrast, there is no significant difference between the upregulation of the 2 transcripts in the large neuron fraction (grey bars). Data are expressed as relative peak areas for the ipsilateral side normalized to the respective contralateral side. Error bars represent SEM. Statistical analyses were performed using unpaired Student's t-test. , or Ca V 2.2/b1b (j, n = 12). Current amplitude was normalized to whole-cell capacitance and plotted against membrane potential. Data are fitted using a modified Bolzmann function, as previously described [12]. (C) Peak current density was À103.56 ± 13.9 pA/pF in presence of DA+B+C (black bar) and À116.63 ± 13.9 pA/pF in presence of DA+BDC (grey bar). These values were both significantly higher than the current density measured in the absence of a 2 d-1, which was 9.6 ± 2 pA/pF (white bar). Error bars represent SEM. Statistical analysis was performed using 1-way analysis of variance (ANOVA) and Bonferroni post hoc analysis, ⁄⁄⁄ denotes P < 0.001). (D) Representative current traces in response to a long depolarizing voltage step (0.9 second) to 0 mV for Ca V 2.2/b1b, co-expressed in tsA-201 cells either without a 2 d (left), or with a 2 d-1 DA+B+C (middle) or a 2 d-1 DA+BDC (right). 1 mM Ba 2+ was used as a charge carrier. Holding potential was À90 mV. Traces are normalized to their peak. (E) a 2 d-1 DA+B+C and DA+BDC significantly accelerated the inactivation of currents compared to no a 2 d.
The decay phase of individual current traces at 0 mV was fitted with a single exponential function, and the mean time constant (s) of inactivation was 204.3 ± 10.3 ms for DA+B+C (black bar, n = 10), 186.4 ± 15.8 ms for DA+BDC (grey bar, n = 14), and 248.0 ± 23.6 ms for Ca V 2.2/b1b without a 2 d (white bar, n = 10). Statistical analyses were performed using 1-way ANOVA, Bonferroni post hoc analysis, ⁄ P < 0.05. (F) Steady-state inactivation curves for I Ba evoked by a test pulse to +20 mV after a 10-second conditioning prepulse of between À100 and 0 mV. Ca V 2.2/b1b/a 2 d-1 DA+B+C (d, n = 15), Ca V 2.2/ b1b/a 2 d-1 DA+BDC (D, n = 13), Ca V 2.2/ b1b (j, n = 10). Error bars represent SEM. Data were fitted with a single Boltzmann equation, and the mean voltages at which the channel is 50% inactivated were À61.3 ± 1.2 mV, À60.0 ± 1.9 mV, and À50.7 ± 1.9 mV, respectively. (Fig. 1C). It is of interest to note that the DADBDC splice variant has a significantly lower affinity for 3 H-gabapentin, compared to the major brain splice variant (Fig. 2). This result correlates with an early finding that the specific binding of 3 H-gabapentin to heart membranes is lower than that in brain membranes, although this may additionally represent a smaller number of binding sites [27]. In brain, the only splice variant found previously was DA+B+C [3], and its presence was confirmed in our study. We found that this was also the main splice variant in DRG neurons. Following SNL in rats, we found previously that the level of a 2 d-1 mRNA was increased >500% in L5/L6 DRG neurons, ipsilateral compared to contralateral to the ligation, and this was accompanied by a similar increase in a 2 d-1 protein [5]. We confirmed that result here, for the main splice variant DA+B+C, where a 3.6-fold increase was observed. However, by CE/LIF we were also able to observe the presence in DRG of a second, minor splice variant DA+BDC, whose level was increased over 10-fold following SNL. Therefore, this splice variant, which has a lower binding affinity for 3 H-gabapentin, was differentially upregulated in SNL. In the experiments described here, DA+BDC increased from 5.1% of the total transcript in sham-operated L5 DRG to 8.4% in L5 ipsilateral to SNL.

Involvement of spliced regions of a 2 d-1 in 3 H-gabapentin binding
To date, several sites in a 2 d-1 have been established to be important for its ability to bind gabapentin [9,50]. The triple arginine motif, situated just upstream of the VWA domain in a 2 d-1, is essential for high-affinity 3 H-gabapentin binding, and also for the ability of gabapentin to alleviate hyperalgesia [24], and to inhibit calcium currents when applied chronically [30]. One of the other regions identified by Wang et al. [50] is located just downstream of region B, and another is just upstream of region C.
In the present study, we found that both the DA+BDC and DADBDC splice variants possessed significantly lower affinity for 3 H-gabapentin, compared to the other splice variants tested. However, there was no significant difference in the gabapentin binding affinity to a 2 d-1 +A+B+C, containing region A, compared to its absence in DA+B+C. Similarly, for the presence or absence of region B, the 3 H-gabapentin binding affinity was similar for a 2 d-1 DA+B+C and DADB+C. These data indicate that A and B are not directly involved in the binding of gabapentin. This suggests that it is primarily the absence of region C in a 2 d-1 DA+BDC and DADBDC that is critical for the reduced 3 H-gabapentin binding affinity of these 2 splice variants. The sequence of region C in rat a 2 d-1 is SKKGKMK, which is positively charged, as is the triple arginine motif that is essential for 3 H-gabapentin binding.
It is notable that the affinity of gabapentin for a 2 d-1 is much greater (ie, a lower K D ) than the clinically relevant concentrations. One potential factor contributing to this discrepancy is that an endogenous ligand may occupy the gabapentin-binding site on a 2 d-1 in vivo, and compete with gabapentin. This also accounts for the finding that purification of the a 2 d proteins results in a marked increase in apparent affinity for gabapentin, as the endogenous ligand is removed during purification [8,18].

Efficacy of gabapentin following SNL
It has been demonstrated that the effect of gabapentin is state dependent, in that in most circumstances it is able to inhibit neuropathic pain responses in experimental animals, while having no effect on physiological nociception [14,24,25,41]. Alterations in gene expression occur following neuropathic nerve damage, which lead to changes in primary afferent inputs into the spinal cord, including the upregulation of a 2 d-1 in these terminals [5]. It has been demonstrated that gabapentin blocks calcium currents acutely in DRG from mice overexpressing a 2 d-1, but not in control animals [36], and this model may mimic the neuropathic condition. Our results indicate that the upregulation of specific splice variants of a 2 d-1 in SNL does not contribute to determining the state dependence of the efficacy of gabapentin, since no splice variants with an increased affinity for gabapentin were identified in DRG neurons following SNL. Thus, the state dependence of gabapentin is likely to depend on the overall elevation of all a 2 d-1 splice variants.

Biophysical properties of a 2 d-1 splice variants found in DRG
It is thought that increased presynaptic calcium currents, resulting from the upregulation of a 2 d-1, contribute to the hyperexcitability of DRG that underlies hyperalgesia and allodynia [36]. If the alternative splicing of a 2 d-1 were to modify the electrophysiological properties of a 2 d-1-containing channels, the consequences of altered function would be augmented following differential upregulation of the a 2 d-1 DA+BDC splice variant, as a consequence of nerve damage. However, we have demonstrated here that the ability of a 2 d-1 DA+BDC to elicit calcium currents is not affected by deletion of region C. a 2 d-1 DA+BDC enhances calcium currents resulting from the co-expression with either Ca V 2.2 or Ca V 2.1 to the same extent as a 2 d-1 DA+B+C. This result indicates that the differential upregulation of a 2 d-1 DA+BDC in DRG neurons does not contribute to triggering hyperexcitability to a greater extent than the main DRG splice variant.

Changes in calcium channel splicing following sensory nerve damage
It is of interest to compare our results with those for the exon 37a variant of Ca V 2.2, which is selectively expressed in DRG neurons, and conducts larger calcium currents than the exon 37b variant [1,6]. Despite being present overall at $7% of the main splice variant 37b of Ca V 2.2, it was found selectively in small DRG neurons also expressing VR1 [1,6]. Interestingly, the 37a splice variant was downregulated to $2% in DRG following SNL [1], whereas there was no change in the main Ca V 2.2 splice variant containing exon 37b.
In contrast, in our study there was a greater increase of a 2 d-1 DA+BDC compared to the main splice variant a 2 d-1 DA+B+C following SNL, and this differential increase occurred particularly in the nonmyelinated small DRG neuron fraction. This, together with our finding that DA+BDC had a significantly lower affinity for gabapentin, and an equivalent ability to enhance neuronal calcium currents, could be highly relevant to the response to gabapentin in patients with chronic neuropathic pain. The rat a 2 d-1 DA+BDC mRNA sequence used in this study has 95% amino acid homology with the human sequence, and C region is identical in both species. Therefore, it is reasonable to speculate that human a 2 d-1 DA+BDC (accession number P54289 isoform 5) is expressed in DRG neurons and undergoes similar upregulation in conditions leading to the development of chronic neuropathic pain. It is, further, tempting to speculate that differences in the extent of upregulation of a 2 d-1 DA+BDC in humans could potentially contribute to the variable efficacy of gabapentinoid drugs [37,38].

Conclusion
We have identified a novel a 2 d-1 splice variant DA+BDC, which is expressed in DRG neurons and is differentially upregulated com-pared to the main DRG splice variant a 2 d-1 DA+B+C, particularly in small nonmyelinated DRG neurons following SNL. This splice variant supports Ca V 2 calcium currents with unaltered properties, but shows a significantly reduced affinity for gabapentin. It could therefore play a role in determining the efficacy of gabapentin in different forms of neuropathic pain. Furthermore, targeting this splice variant for drug discovery could increase the therapeutic efficacy of gabapentinoid drugs in the future.