Central mechanisms mediating thrombospondin-4 induced pain states

Peripheral nerve injury induces increased expression of thrombospondin-4 (TSP4) in spinal cord and dorsal root ganglia (DRG) that contributes to neuropathic pain states through unknown mechanisms. Here, we test the hypothesis that TSP4 activates its receptor, the voltage-gated calcium channel Ca v α 2 δ 1 subunit (Ca v α 2 δ 1 ), on sensory afferent terminals in dorsal spinal cord to promote excitatory synaptogenesis and central sensitization that contribute to neuropathic pain states. We show that there is a direct molecular interaction between TSP4 and Ca v α 2 δ 1 in the spinal cord in vivo, and that TSP4/Ca v α 2 δ 1 dependent processes lead to increased behavioral sensitivities to stimuli. In dorsal spinal cord, TSP4/Ca v α 2 δ 1 dependent processes lead to increased frequency of miniature and amplitude of evoked excitatory-post-synaptic-currents in second order neurons, as well as increased VGlut 2 and PSD95 positive puncta, indicative of increased excitatory synapses. Author contributions: JP designed, performed, and analyzed experiments of in vitro binding, immunoprecipitation, protein expression and purification. YPY designed, performed, and analyzed experiments of confocal imaging and synaptogenesis. CYZ designed, performed, and analyzed experiments of electrophysiology. KWL, EC, DSK, BV, XZ and NG designed, performed, and analyzed experiments of transgenic animal models and behavioral pharmacology. IV and EPR designed and constructed the shRNA - AAV vectors for the behavioral pharmacology experiments. KS and OS designed, performed, and analyzed experiments of motor functions. DW and GF designed, generated, and analyzed the conditional knockout mice. CE and BB provided the TSP4 KO mice, and contributed to conception of the study. FZ designed and constructed the TSP4 expression vectors, assisted in protein purification and contributed to conception of the study. ZDL conceived and coordinated the study. ZDL, JP, YPY,


Introduction
Neuropathic pain due to peripheral nerve injury is associated with up-regulation of expression of thrombospondin-4 (TSP4) in spinal cord and dorsal root ganglia (DRG) that induces increased frequency of excitatory post-synaptic currents (mEPSC) in dorsal spinal cord and neuropathic pain states (1,2). Details about the mechanisms remain to be defined, however. TSP4 belongs to a five-member thrombospondin superfamily of oligomeric, extracellular matrix glycoproteins (TSP1-5) that can be subdivided into groups A (TSP1/2) and B (TSP3/4/5) based on structure and functional domain similarities (3). TSP proteins are important in mediating cell to cell, and cell to matrix interactions (3,4). TSP4 is expressed in multiple sites and its functions are not welldefined (5), although there is evidence that TSP4 promotes neurite outgrowth (6).
Collectively, these observations suggest the following intriguing hypothetical mechanistic model: 1) peripheral nerve injury upregulates Ca v α 2 δ 1 in peripheral sensory neurons and its central terminals; 2) peripheral nerve injury also triggers increased synthesis and release of TSP4 in spinal cord and DRG; 3) increased TSP4 interacts directly with its receptor Ca v α 2 δ 1 on the central terminals of sensory neurons to increase excitatory synaptogenesis and synaptic neurotransmission; 4) increased excitatory transmission in dorsal spinal cord contributes to central sensitization and other species cell lysates validated by Qiagen, Valencia, CA). The His-tagged proteins were purified using a Ni-NTA column based on the manufacturer's instructions (Invitrogen, Grand Island, NY), concentrated with Amicon Ultra-4 Centrifugal Filter Unit (50K Molecular weight cut off, Millipore, Billerica, MA), aliquoted and stored at -80 °C until use.
Immunoprecipitation: The spinal cord tissues from adult male mice and adult male Harlan Sprague-Dawley rats were collected by hydraulic extrusion from animals deeply anesthetized with isoflurane, and lysed in protein extraction buffer (50mM Tris buffer, pH 8.0, 0.1% Triton X-100, 150mM NaCl, 1mM EDTA, and protease inhibitors). The cell lysate was then incubated on ice for 15 min, centrifuged x 20,000 g, 20 min, at 4°C. The supernatant was incubated with anti-TSP4 polyclonal antibody (guinea pig, 1:750, validated previously (36)) over-night at 4 °C. Protein A/G-agarose beads (Thermo, Waltham, MA) were then added, incubated for 2 hrs at 4°C, and washed with protein extraction buffer. The antibody-captured proteins were eluted in nonreducing condition with low pH elution buffer (Thermo, Waltham, MA) at room-temperature (RT) and the same volume of control supernatant or immunocomplex samples was analyzed by Western blots under non-reducing conditions. Solid-phase binding: Briefly, FLAG-Ca v α 2 δ 1 cDNA was transiently transfected into the tsA-201 cell line stably expressing Cav2.2 and Cavβ3 (gift from Dr. D. Lipscombe from Brown University, (37) by Lipofectamine 2000 (Invitrogen, Grand Island, NY). The transfected cells were washed, extracted in protein extraction buffer (50 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.1% Triton-X, pH7.4) in two-three days. The cell lysate was incubated on ice for 15 min, then centrifuged x 13,000 g, at 4 °C for 20 min. The supernatant was rotating-incubated with anti-FLAG M2 agarose affinity resin (Sigma-Aldridge, St. Louis, MO) for 2 hrs at 4 °C, washed with protein extraction buffer. FLAG-Ca v α 2 δ 1 was eluted in elution buffer (0.1 M glycine, pH 3.5) and stored at -20 °C until use.
The reagents for solid-phase binding were from Invitrogen (Grand Island, NY). Recombinant TSP4 proteins (80 µg/mL) were immobilized onto a 96-well polystyrene plates (Thermo, Waltham, MA) overnight at 4 °C in coating buffer A. All further incubations were carried out at RT for 1 hr, and proteins or antibodies were diluted in assay buffer containing bovine serum albumin (BSA). After washing and blocking, the plates were incubated with FLAG-Ca v α 2 δ 1 , washed, then incubated with mouse monoclonal anti-FLAG antibodies (1:1000; Cat. #: F1804, validated against FLAG-fusion proteins by Sigma-Aldridge, St Louis, MO), followed by horseradish peroxidase (HRP)-conjugated secondary antibodies. The bound FLAG-Ca v α 2 δ 1 complexes were detected by measuring a color reaction at 450 nm after adding tetramethylbenzidine for 15 min followed by adding sulfuric acid to stop the reaction.
Surface Plasmon Resonance Binding (38): All experiments were carried out using BIAcore 3000 and CM5 Sensor Chip (GE Healthcare Sciences, Piscataway, NJ), and at 25°C. Ca v α 2 δ 1 antibody (mouse, Cat. #: D219, Sigma-Aldrich St. Louis, MO) was coupled to the dextran matrix of a CM5 sensor chip using Amine Coupling Kit as described (39). The antibody specificity for Ca v α 2 δ 1 had been confirmed with tissue samples from Ca v α 2 δ 1 knockout mice (Fig. 4C). Excess reactive esters were quenched by injection of 1.0 M ethanolamine-hydrochloride, pH 8.5. The binding assays were performed using HBS-P buffer (0.01M HEPES, pH 7.4, 0.15M NaCl, 0.005% and surfactant P20) as running buffer. Purified TSP4 proteins and Ca v α 2 δ 1 protein extracts from tsA-201 cells stably expressing Ca V 2.2e [Δ24a, 31a], Ca V β 3 , and Ca v α 2 δ 1 as described (40) were diluted in HBS-P buffer (GE Healthcare Sciences). Ca v α 2 δ 1 protein extracts were injected at a flow rate of 10 µL/min over the immobilized Ca v α 2 δ 1 antibody flow cells, followed by injection of purified TSP4 proteins at a flow rate of 20 µL/min. Non-specific binding of TSP4 to the flow cell without immobilized Ca v α 2 δ 1 antibody was subtracted from all binding curves using BIAevaluation software (version 3.0, GE Healthcare Sciences) and plotted using Graphpad Prism (Graphpad Software, San Diego, CA).
Spinal nerve ligation (SNL) (41): Briefly, the left L4 spinal nerve of mice, which is equivalent to L5 spinal nerve in rats (42), was exposed in isoflurane anesthetized animals, and tightly ligated between the DRG and their conjunction to form the sciatic nerve with a silk suture. Sham procedures were done in the same way without spinal nerve ligation. Behavioral tests were performed at designated time before collection of tissue samples, which were either processed immediately for biochemical studies or kept at -80 °C until use.
A cDNA encoding the complete coding sequence of the mouse Ca v α 2 δ 1 subunit was obtained from Open Biosystems (IMAGE: 40061614), then cloned into a mammalian expression vector (pYFP-C1, Clontech). Candidate shRNAs were designed using publicly available web tools (Invitrogen Block-it and Genscript). These shRNAs were imbedded in a mir30 backbone using opposing BsmBI sites to insert complementary oligonucleotides encoding the shRNAs without altering the mIR sequence (43,44). The shRNAs were cloned into a human H1 promoter amplified from pLVUTH (Addgene clone 11650). We engineered a novel scAAV using the pLVUTH backbone and the delta-ITR sequences described by McCarty and coworkers (45). This vector uses a CMV enhanced human synapsin-1 promoter (46) to drive the expression of mCherry (provided by Dr. Roger Y. Tsien, UCSD) that was modified by adding the C-terminal ER export signal (FCYENE) from Kir2.1 (47). Four shRNAs were screened for target knocking down after expression in HEK-293 cells. Knocking down efficiency was measured using qPCR with primers that encompassed the shRNA binding site (48). The shRNA with the highest knocking down efficacy (80%) relative to a scrambled control was AD1: AACTGGACAAGTGCCTTAGAT. CSRH1AD1 scAAV particles pseudotyped with serotype 8 were purified by the University of North Carolina Vector Core (titer 2 X 10 12 virus molecules/mL).
Intrathecal injection: Intrathecal injections between lumbar L5/6 regions for rats or L4/5 regions for mice were performed under light isoflurane anesthesia through a 30-gauge needle connected to a microinjector (Tritech Research, Inc., Los Angeles, CA). A total volume of 10 µL per rat or 5 µL per mouse was injected.
Behavioral test: Testing was performed in a blinded fashion. Behavioral test values between left and right hindpaws from the SNL groups were recorded separately and used for statistical analysis, and that from non-SNL groups were averaged and used for statistical analysis. Tactile allodynia: Hindpaw sensitivities to von Frey filament stimulation were tested for tactile allodynia as described previously (2,49,50). After acclimatization in wire mesh-floored transparent enclosures, the animals were accessed for the 50% paw withdrawal thresholds (PWT) to von Frey filament (Stoelting Wood Dale) stimulation using the up-down method (51). Briefly, the plantar surface of the hindpaw was contacted perpendicularly with the first filament (2.0 g for rats or 0.41 g for mice) until it was slightly bent. A positive response of paw lifting within 5s led to the use of the next lighter filament, and a negative response led to the use of the next heavier filament until a total of six measurements had been made, starting from the one before any change in the behavioral response. A score of 15 g for rats, or 3 g for mice was assigned if five consecutive negative responses had occurred or a score of 0.25 g for rats, or 0.01 g for mice was assigned if four consecutive positive responses had occurred. The data were then used to determine the 50% response threshold described previously (25). Thermal hyperalgesia: Reduced hindpaw withdrawal latency (PWL) to thermal stimuli was measured using a Hargreaves (hot box) apparatus (University of California San Diego, CA) (52) as the indication of thermal hyperalgesia. After acclimatization for at least 30 min on a glass floor maintained at 30 ± 0.1 °C in transparent enclosures, the hindpaw plantar surface of a freemoving animal was stimulated by radiant heat projecting from a high intensity light bulb through a small aperture below the glass surface. When the animal moved the paw away from the thermal stimulus, motion detectors on the apparatus turned off the heating light automatically. The paw withdrawal latency was recorded as the time between thermal stimulus application and hindpaw withdrawal. 20 s were set as the cut-off time to prevent thermal injury or skin sensitization. Two readings per paw were averaged for statistical analysis.
Mechanical hyperalgesia: After acclimatization for one-week to human holding and touch, rats were tested for mechanical hyperalgesia (Randall-Selitto Test, (53) using a Paw Pressure Analgesymeter (Ugo Basil North America). Briefly, a rat hindpaw was placed between a blunt pointer and a flat surface and subjected to an escalating force (16 grams/second) until paw withdrawal by the animal. The recorded hindpaw withdrawal inducing force was used as the paw pressure withdrawal thresholds (PPT). Locomotor function tests: After acclimatization daily for one-week to human handling and the open field test apparatus, mice were tested for locomotor function by a blinded observe using scores of 0 to 9 arranging from no ankle movement (0) to frequent or consistent coordinated plantar stepping, normal trunk stability and tail up position (9) as described by Basso et al. (54).
Western blots: Briefly, equal amounts of proteins were separated in 3-8% NuPAGE Tris-Acetate gels (Invitrogen, Grand Island, NY) by electrophoresis, then transferred to polyvinylidene difluoride membranes electrophoretically. After blocking non-specific binding sites with 5% low fat milk (in PBS-T containing 137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 1.4 mM KH 2 PO 4 , 0.1% Tween-20, pH 7.4), the membranes were cut into sections containing different target proteins, incubated with primary antibodies against: Ca v α 2 δ 1 (mouse, 1:1000, Cat. #: D219, Sigma-Aldrich, St. Louis, MO), TSP4 (rabbit, 1:1000, custom made and validated against purified TSP4 proteins, Genscript, Piscataway, NJ), β-actin (mouse, 1:10,000, Cat. #: MAB8929, validated against various β-actin expressing cell lines, Novus Biologicals, LLC, Littleton, CO) over-night at 4 °C, followed by horseradish peroxidaseconjugated secondary antibody (1:2000, Cell Signaling, Danvers, MA) for 1 hr at RT. After a brief incubation with chemiluminescent reagents (Thermo Scientific, Waltham, MA), the band densities were quantified by either imaging quantification (KODAK Image Station 2000MM) or densitometry within the linear range of the film sensitivity curve. The Ca v α 2 δ 1 band detected by the mouse Ca v α 2 δ 1 antibodies reflected the Ca v α 2 protein only (≈150 kDa) since the Ca v δ 1 peptide separates from the Ca v α 2 protein under reducing conditions in Western blots (55). For quantification, band density ratios for the protein of interest over that of β-actin (≈42 kDa) were calculated within each sample first for normalization of total protein loading before cross-sample comparisons. Band density variations for the proteins of interest in the contralateral (noninjury) side were determined by comparing each band density with the mean of that from at least two different control samples in the same Western blot after taking the ratios to β-actin band densities.
Spinal cord slice recording. a-amino-3hydroxyl-5-methylisoxazole-4-propionic acid (AMPA) receptor-mediated miniature excitatory post-synaptic currents (mEPSC) and evoked excitatory post-synaptic currents (eEPSC) were recorded from lumbar spinal cord transverse slices (300 µm). Briefly, spinal cord slices were prepared and transferred to the recording chamber as described previously (30,33). The patch electrode had a resistance of 5 -7 MΩ when filled with pipette solution that contained (mM): 135 potassium gluconate, 5 KCl, 5 EGTA, 0.5 CaCl 2 , 10 HEPES, 2 Mg-ATP, and 0.1 GTP (pH 7.2) with an osmolarity 300 mosmol/L. Superficial dorsal horn neurons were visualized with an upright microscope (Eclipse FN1, Nikon, Japan) and nearinfrared illumination based on the gelatinous (semi-transparent) appearance of lamina II (substantia gelatinosa). While neurons in superficial dorsal horn (including lamina I, lamina II outer or II o and lamina II inner or II i ) are heterogeneous (56,57), the boundaries of laminae I, II o and II i are ambiguous in live spinal cord slices so it was difficult to classify lamina specific neuron populations, a limitation of our sampling method. Thus, we pooled all recording data together for a general view of TSP4 effects in modulating synaptic transmission and regulation in superficial dorsal horn neurons. All recordings were performed at 32 ± 0.5 °C with MultiClamp 700B amplifiers (Axon Instruments, Molecular Devices, Union City, CA), Digidata 1440 analogto-digital converters (Axon Instruments) and pClamp 10.2 software (Axon Instruments). mEPSCs were recorded as described previously (2,30,33) in the presence of tetrodotoxin (TTX, 1 µM), strychnine (1 µM), bicuculline (10 µM), and 2-amino-5-phosphonopentanoic acid (AP5, 50 µM) to block TTX-sensitive Na + , glycinergic, GABAergic and N-methyl-D-aspartate (NMDA) currents, respectively. Membrane potential was held at -60 mV so that NMDA receptor-mediated currents were blocked (58), which only represented less than 10 percent mEPSCs in dorsal horn neurons and the remaining 90 percent mEPSCs can be blocked by AMPA receptor antagonist 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) (59). Series resistance was monitored without compensation throughout the experiment (Multiclamp 700B). Cells were excluded from analysis if the series resistance changed by more than 20% during the whole-cell recording. Signals were analyzed using clampfit 10.3 (Molecular Devices) after the traces were low-pass filtered at 2 kHz. Cumulative distribution of mEPSC frequency or amplitude of individual neurons from each experimental group was analyzed with Kolmogorov-Smirnov test (KS test). eEPSCs were similarly recorded from superficial dorsal horn neurons of L4 lumbar spinal cord slices upon stimulating (0.1 ms, 0.05 Hz) the attached dorsal roots or dorsal root entry zone with 0-500 µA stimulus intensity. At least six eEPSC events were recorded at each stimulus intensity. QX314 (5 mM) was added in intrapipete solution to prevent sodium channel activation.
Immunohistochemistry (2,28): Lumbar spinal cord and DRG samples were fixed in 4% paraformaldehyde overnight, cryoprotected in 30% sucrose, mounted in Optimum Cutting Temperature (O.C.T., Sakura Finetek, Torrance), and sectioned with a cryostat (CM1900, Leica Microsystems, Wetzlar, Germany) into 10 µm slices. Slices of spinal cord samples were pretreated with heat-based antigen retrieval (10nM Sodium Citrate, 0.05% Tween 20, pH 6.0, 5 minutes in pressure cooker), then incubated with primary antibodies against Ca v α 2 δ 1 (rabbit polyclonal, custom made by Thermo Fisher Scientific Inc., Waltham, MA, for spinal cord samples after antigen retrieval; or mouse monoclonal, Sigma-Aldrich for DRG samples without antigen retrieval. Both antibodies were validated with Ca v α 2 δ 1 knockout mice shown in Fig. 4A-D), VGlut 2 (guinea pig, Cat. #: 135404, validated previously (28,60-62), Synaptic Systems, Germany) and PSD95 (rabbit, Cat. #: MA1-045, validated previously (28,63,64), Thermo Fisher Scientific Inc.), followed by secondary antibodies with Alexafluor 488 or 594 (Invitrogen) against IgG of corresponding species of the primary antibodies. Sample sections from control and experimental groups (sides) within the same set of experiments were stained at the same time. Samples were mounted with Vectashield containing DAPI for cell nuclei staining (Vector Labs, Burlingame, California). Two images were taken from each superficial dorsal horn section randomly using a Zeiss LSM780 confocal microscope (UC Irvine Optical Biology Core) in 0.3 µm Z-stacks, and three consecutive Z stacks with the best signal were merged and used for data analysis with Volocity 6.0 (Perkin Elmer, Waltham, MA). Briefly, images from control and experimental groups within the same set of experiment were captured with the same setting. Volocity Find Object Using Percentage Intensity function was used to define background threshold, which was used for both contralateral and injury sides within each set of experiment. Fluorescent immunoreactivities above the background level were selected for analysis. From TSP4 or saline injected mouse samples, VGlut 2 /PSD95 co-stained samples (n = 36 over three animals, 100 µm apart) were analyzed to determine the numbers of total VGlut 2 + (green) PSD95 + (red) and VGlut 2 + /PSD95 + (yellow) puncta. Since the effect of intrathecal injection was bilateral, the ratio of VGlut 2 + /PSD95 + over VGlut 2 + /PSD95puncta from both sides was used to compare the differences between the TSP4 and saline treated groups.
Statistics. 1-way or 2-way ANOVA with posttests were performed for multi-group comparisons and unpaired Student's t tests were performed for pair-wise comparisons as indicated in figure legends. Significance was determined by a twotailed p value < 0.05.

Results
TSP4 interacts directly with Ca v α 2 δ 1 in rodent spinal cord in vivo. Data from in vitro immunoprecipitation and functional assays suggest that TSP4/Ca v α 2 δ 1 form a complex in mediating abnormal excitatory synapse formation in rat cerebral cortex (8). However, a direct binding or functional interaction between these proteins in spinal cord has not been shown. If interactions of these proteins lead to abnormal synaptogenesis and neuropathic pain states post injury, there should be direct molecular interactions between astrocyte-secreted TSP4 and Ca v α 2 δ 1 in the adult spinal cord in vivo. To test this, we first examined if TSP4 and Ca v α 2 δ 1 proteins were detectable in immunoprecipitation (IP) complexes from rodent spinal cord samples. Our results confirmed previous findings in rat cerebral cortex (8) and showed that Ca v α 2 δ 1 was detectable in TSP4immunoprecipitates from rat and mouse spinal cord (Fig. 1A), suggesting that these proteins may interact directly or indirectly in rodent spinal cord. The differences in the patterns of bands between the spinal cord lysates and IP samples are likely due to these factors. First, IP samples were more concentrated compared to the "Lys" control samples since we loaded an equal volume of each sample onto the same blot. Under non-reducing conditions, the IP complexes might contain more target proteins and other associated proteins. Second, it is possible, but needs to be confirmed, that TSP4 antibodies might have pulled down the extracellular Ca v α 2 domain (150 kDa) of the Ca v α 2 δ 1 subunit (175 kDa) without the Ca v δ 1 peptide (25 kDa).
To further assess whether TSP4 interacts with Ca v α 2 δ 1 directly, we examined interactions with solid phase binding and surface plasmon resonance spectroscopy assays. The solid phase binding assay confirmed that recombinant Ca v α 2 δ 1 proteins bind directly to immobilized TSP4 in a dose-dependent fashion (Fig. 1B). Conversely, surface plasmon resonance spectroscopy demonstrated dose-dependent binding of TSP4 to immobilized Ca v α 2 δ 1 on a BIAcore CM5 sensor chip with fast association and slow dissociation (Fig. 1C).
Interdependent interactions between TSP4 and Ca v α 2 δ 1 proteins contribute to pain states and dorsal horn neuron sensitization. If TSP4 induces neuropathic pain by interacting with Ca v α 2 δ 1 , then blocking this interaction pharmacologically should abrogate TSP4-induced pain.
One hour following gabapentin injection (300 µg/rat, i.t.), PWT increased to near the control level. The effect of gabapentin was reversible because PWT were again reduced to the pre-treatment level 24 hrs post gabapentin treatment (filled bars, Fig.  2A). Gabapentin treatment had a similar effect on TSP4-induced thermal hyperalgesia and mechanical hyperalgesia (data not shown).
We previously reported that TSP4-induced pain states correlated with increased mEPSC frequency, but not amplitude, in superficial dorsal horn neurons (2). If TSP4 mediates this effect via ongoing interaction with Ca v α 2 δ 1 , then blocking Ca v α 2 δ 1 with gabapentin should reduce mEPSC frequency. Consistent with our previous study (2), recordings from superficial dorsal horn neurons 3days after TSP4 injection revealed a significant over 100% increase in average mEPSC frequency without significant changes in its amplitude (Fig.  2B1, B2, B3). Following treatment with GBP (50 µM), elevated mEPSC frequency was dramatically decreased to the control level, but its basal level in the control group was not affected significantly (Fig. 2B1, B4). These changes were confirmed by Kolmogorov-Smirnov tests using cumulative distribution of mEPSC frequency or amplitude of individual neurons from each experimental group (bottom panels of Fig. 2 B2-B4). This gabapentin concentration is close to that in patients' cerebrospinal fluid after chronic gabapentin treatments. It has been reported that oral 900 mg/day gabapentin treatment for three months can reach peak plasmid concentration about 10 mg/L (53 µM) 3 hours after the last dose (65), and cerebrospinal fluid gabapentin concentration range is about 20-80% of that in the plasma after multiple dosing (66). Collectively, these results support the idea that TSP4/Ca v α 2 δ 1 dependent processes mediate both TSP4-induced pain states and the putative pathological underpinning (increased mEPSC frequency).
To further test whether blocking TSP4/Ca v α 2 δ 1 dependent processes by biochemical knocking down Ca v α 2 δ 1 could prevent TSP4-induced pain states, we investigated if TSP4-induced behavioral hypersensitivity could be prevented by preemptive knocking down of Ca v α 2 δ 1 with intrathecal treatment of anti-Ca v α 2 δ 1 small hairpin RNA in AAV vectors (Ca v α 2 δ 1 -AAV).
We previously showed that bolus intrathecal TSP4 injection induced tactile allodynia as evidenced by progressive reduction in PWT to a tactile stimulus (2). This effect was confirmed in the present study in rats that were pre-treated with intrathecal control AAV (10 6 units in 2 µL) 10-days prior to bolus TSP4 injection (45 µg/rat, i.t.) (Fig. 2C). In contrast, rats pre-treated with Ca v α 2 δ 1 -AAV (10 6 units in 2 µL), which diminished dorsal spinal cord Ca v α 2 δ 1 levels (Fig. 2C, insert), did not exhibit the progressive decreases in PWT that reflect tactile allodynia (Fig. 2C). These results support the conclusion that TSP4-induced allodynia does depend on Ca v α 2 δ 1 .
To further test this idea in the clinically relevant neuropathic model of spinal nerve ligation (SNL), we investigated whether knocking down Ca v α 2 δ 1 by intrathecal Ca v α 2 δ 1 -AAV could reverse SNL-induced behavioral hypersensitivity since our previous studies have implicated either Ca v α 2 δ 1 or TSP4 in neuropathic pain states (2,24,27,29,32). Bolus i.t Ca v α 2 δ 1 -AAV, but not the control vector, treatments (10 6 units in 2 µL) caused a time-dependent reversal of 2-week SNLinduced tactile allodynia without affecting the behavioral sensitivity in the contralateral (noninjury) side.
The anti-allodynic effects of Ca v α 2 δ 1 -AAV peaked about 10 days after the i.t. injection (Fig. 2D1), and lasted for over two weeks (data not shown). Western blot analysis of spinal cord samples from control and Ca v α 2 δ 1 -AAVtreated rats at the peak allodynia reversal time point (10 days post Ca v α 2 δ 1 -AAV treatment) confirmed Ca v α 2 δ 1 knockdown on both sides of dorsal spinal cord. Ca v α 2 δ 1 levels on the injury side were not significantly different from levels in the non-injury side of control vector-treated rats (Fig. 2D2).
The fact that Ca v α 2 δ 1 knockdown prevented allodynia development (Fig. 2C) and reversed established allodynia (Fig. 2D1) indicates that TSP-induced behavioral hypersensitivity requires ongoing interactions between TSP4 and Ca v α 2 δ 1 at the spinal cord level.
Next, we examined if blocking TSP4 could diminish pain states and putative pathological underpinning due to elevation of Ca v α 2 δ 1 . We have previously shown that transgenic (TG) mice with neuronal Ca v α 2 δ 1 overexpression have increased mEPSC frequency and pain states similar to mice with SNL injury (30,33,34,59,67). We first tested whether blocking TSP4 dependent processes with TSP4 antibodies would reverse Ca v α 2 δ 1 -overexpression-induced allodynia. Consistent with previous findings (34), TG mice with neuronal Ca v α 2 δ 1 overexpression have greatly reduced PWT to mechanical stimuli (the behavioral reflection of allodynia). Following treatment with the TSP4 antibody (10 µg/mouse, i.t., chicken polyclonal, validated previously (68)), PWT gradually increased to near the control level by 8 hours post-treatment, and then returned to the pre-treatment level by 24 hrs (Fig. 3A1). Antibody treatment did not affect baseline thresholds in age-and sex-matched wild type (WT) mice. Heat-denatured antibody was without effect (Fig. 3A1). Similar treatment with this antibody also reversed established neuropathic pain states in the more clinically relevant SNL model (2).
As an alternative approach, we tested if genetic ablation of TSP4 from the Ca v α 2 δ 1 TG mice would eliminate behavioral hypersensitivities previously reported in Ca v α 2 δ 1 TG mice (30,33,34,59). To test this, we crossed Ca v α 2 δ 1 TG mice that develop hypersensitivity to stimuli with TSP4 knockout mice to generate Ca v α 2 δ 1 TG/TSP4 KO mice with elevated neuronal Ca v α 2 δ 1 and TSP4 ablation. Similar to our previous findings (34), Ca v α 2 δ 1 overexpression in the TG mice resulted in behavioral hypersensitivities to mechanical (allodynia, Fig. 3A2, top) and thermal (thermal hyperalgesia, Fig. 3A2, middle) stimuli. Assessment of pain sensitivity in Ca v α 2 δ 1 TG/TSP4 KO mice revealed no increased sensitivity to either of these stimuli. Behavioral thresholds in mice with TSP4 knockout alone (TSP4 KO) were comparable to that in WT control mice (Fig. 3A2, top and middle). The locomotor function test scores of Basso Mouse Scale (BMS) in mice with these genetic modifications were similar to that in the control mice (Fig. 3A2, bottom). Thus, TSP4 basal level is not critical in maintaining basal sensory/motor functions, and differences in behavioral sensitivities to stimuli among these mouse groups are not due to changes in motor functions. Together, these findings support that TSP4/Ca v α 2 δ 1 dependent processes are also required for pain state processing induced by elevated Ca v α 2 δ 1 .
We also assessed whether the increase in mEPSC frequency reported previously in Ca v α 2 δ 1 TG mice (30,33,59) could be normalized by deleting TSP4 from the Ca v α 2 δ 1 TG/TSP4 KO mice. Similar to our previous findings (30,33,59), recordings from superficial dorsal horn neurons of Ca v α 2 δ 1 TG mice revealed significantly elevated frequency, but not amplitude, of mEPSC compared with that from control WT mice (Fig.  3B).
As another control, recordings from superficial dorsal horn neurons of TSP4 KO mice revealed mEPSC frequency/amplitude comparable to that seen in WT neurons (Fig. 3B). However, recordings from superficial dorsal horn neurons of Ca v α 2 δ 1 TG/TSP4 KO mice revealed that mEPSC frequency and amplitude were within the range of that in control mice (Fig. 3B). These findings were confirmed by Kolmogorov-Smirnov tests using cumulative distribution of mEPSC frequency or amplitude of individual neurons from each experimental group (bottom panels of Fig.  3B2). Collectively, these findings support that while basal level TSP4 is not required for maintaining a normal level of mEPSC, but TSP4/Ca v α 2 δ 1 dependent processes are required for Ca v α 2 δ 1 -induced increase of mEPSC frequency, an indication of enhanced pre-synaptic excitatory input.
We hypothesized that increased TSP4/Ca v α 2 δ 1 interactions in the spinal cord of the TG mice could lead to hyperexcitability of dorsal horn neurons to peripheral stimulation. To test this, we examined the amplitude of eEPSCs in L4 superficial dorsal horn neurons from the Ca v α 2 δ 1 TG mice in response to escalating intensities of stimulation to dorsal root entry zone. Compared with that from WT littermates, eEPSC amplitudes were increased at all levels of stimulus intensity tested (Fig. 3C), indicating hyper-responsiveness of the dorsal horn neurons to afferent activation as a result of Ca v α 2 δ 1 overexpression. Similar recordings in Ca v α 2 δ 1 TG/TSP4 KO mice revealed that eEPSC frequency was within the range of WT mice, as was also the case with TSP4 KO alone (Fig. 3C). These findings suggest that while basal level TSP4 is not required for maintaining normal dorsal horn neuron excitatory tone, but TSP4/Ca v α 2 δ 1 dependent processes are required for Ca v α 2 δ 1 -mediated dorsal horn neuron sensitization.
Interdependent interactions between TSP4 and Ca v α 2 δ 1 proteins promote aberrant excitatory synaptogenesis in animal models with pain states. Previous studies have shown that Ca v α 2 δ 1 proteins in sensory neurons are transported to central afferent terminals in the spinal dorsal horn under normal and post-injury conditions (27,29). Other studies have shown that TSP interactions with Ca v α 2 δ 1 promote excitatory synaptogenesis in vitro (8). Together, these findings raise the possibility that injury-induced TSP4 in spinal cord might interact with Ca v α 2 δ 1 at the pre-synaptic terminals of sensory afferents to promote aberrant excitatory synaptogenesis, which in turn would contribute to enhanced transmission along pain pathways.
To test this hypothesis, we took the first step to determine whether there was evidence for aberrant excitatory synaptogenesis with TSP4-induced pain states, if so, then to determine if genetic ablation of Ca v α 2 δ 1 could block TSP4-induced synaptogenesis.
We generated Ca v α 2 δ 1 conditional knockout (WT CKO) mice in which exon 6 of the Ca v α 2 δ 1 gene was floxed with loxP sites (Fig. 4A, B). Crossing these mice with Crerecombinase expressing mice resulted in deletion of Ca v α 2 δ 1 from Cre-expressing cells (Fig. 4C). The WT CKO mice were crossed with Advillin-Cre mice with Cre recombinase expression only in Advillin-positive sensory neurons that cover about 94% of all DRG neurons (35). Cre-driven Ca v α 2 δ 1 deletion from DRG neurons in homozygous CKO Adv-Cre+/mice was confirmed by immunostaining data showing that 2-week SNL could induce DRG Ca v α 2 δ 1 upregulation indicated as increased Ca v α 2 δ 1 antibody immunoreactivity in DRG neurons of control mice (WT CKO), as we have reported previously (32), but failed to do so in homozygous CKO Adv-Cre+/mice (Fig. 4D). We then immunostained sections of L4 lumbar spinal cord from TSP4 (5 µg/mouse, i.t.) injected mice 4days after the injection (peak pain states in control mice) with markers for excitatory synapses (VGlut 2 and post-synaptic density marker PSD95). Measurements of total number of synaptic puncta immunoreactive to both synaptic marker antibodies in spinal cord superficial dorsal horn (Fig. 4E) revealed substantial increases in excitatory synapses in TSP4 injected control WT CKO mice compared to vehicle injected WT CKO mice. These TSP4-induced changes were not seen with Ca v α 2 δ 1 ablation in the CKO Adv-Cre mice (Fig.  4F-H). Thus, TSP4-induced aberrant excitatory synaptogenesis in dorsal spinal cord requires TSP4/Ca v α 2 δ 1 dependent processes.
If TSP4-induced aberrant excitatory synaptogenesis in dorsal spinal cord contributes to pain processing, one would expect to see that absence of TSP4-induced excitatory synaptogenesis in the CKO Adv-Cre mice correlates with the absence of TSP4-induced pain states. We tested this by examining behavioral sensitivity of control and CKO Adv-Cre mice after TSP4 injections (5 µg/mouse, i.t.). Similar to previous findings in rats (2), TSP4 injection into control WT CKO mice led to tactile allodynia (Fig. 5A, B) and thermal hyperalgesia (Fig. 5C), which peaked approximately 4-days after TSP4 injections that correlated temporally with a significant increase of excitatory synapses in dorsal spinal cord of these mice (Fig. 4F-H). In contrast, similar TSP4 injection into CKO Adv-Cre mice failed to induce similar behavioral hypersensitivities (Fig. 5A-C). This supports that TSP4-induced aberrant excitatory synaptogenesis through TSP4/Ca v α 2 δ 1 dependent processes plays a role in transmitting nociceptive signals.
We next tested whether TSP4-induced synaptogenesis also plays a role in neuropathic pain development in the more clinically relevant SNL model. Our previous study has shown that SNL-induced allodynia is diminished in SNL TSP4 KO mice (2), suggesting a role of TSP4 in mediating neuropathic pain. Immunostaining for Ca v α 2 δ 1 revealed increases in Ca v α 2 δ 1 puncta in the superficial dorsal horn of the injury side in both WT and TSP4 KO mice 2-weeks post SNL (Fig.  6A, B), which correlated with severe allodynia in SNL WT, but not TSP4 KO, mice (2). Thus, elevated pre-synaptic Ca v α 2 δ 1 expression is not regulated by TSP4. Nerve injury also increased VGlut 2 puncta that mainly co-localized with at Univ of California -Irvine on May 11, 2016 http://www.jbc.org/

Downloaded from
Ca v α 2 δ 1 puncta in the WT mice (Fig. 6A, C, D), supporting that injury-induced Ca v α 2 δ 1 upregulation at the pre-synaptical terminals of injured afferents is associated with increased numbers of excitatory synapses in dorsal spinal card. Importantly, these changes were not seen in TSP4 KO mice with SNL (Fig. 6A, C, D). These data support that SNL does induce aberrant excitatory synaptogenesis that also requires TSP4/Ca v α 2 δ 1 dependent processes.

Discussion
Peripheral nerve injury induces upregulation of TSP4 in DRG/spinal cord that contributes to neuropathic pain states through mechanisms that were previously undefined (2,24,25,27,29,31). Here, we provided a large body of evidence to support that TSP4/Ca v α 2 δ 1 dependent processes are required in promoting central sensitization and pain states.
Our data confirm that there is a direct interaction between TSP4 and Ca v α 2 δ 1 in rodent spinal cord and in vitro. In addition, blocking or down-regulating Ca v α 2 δ 1 can block TSP4-induced pain states and increased mEPSC frequency in spinal cord neurons. Conversely, blocking or genetically deleting TSP4 can block Ca v α 2 δ 1 overexpression-induced behavioral hypersensitivity, increased mEPSC frequency and exaggerated eEPSC in spinal cord neurons. Furthermore, elevated TSP4 induces an increase of spinal excitatory synapses that correlates with heightened pain states, both of which can be normalized by Ca v α 2 δ 1 ablation from sensory neurons. Equally, TSP4 ablation blocks injuryinduced excitatory synaptogenesis associated with elevated Ca v α 2 δ 1 without affecting nerve injuryinduced Ca v α 2 δ 1 upregulation in the dorsal horn. Together, these findings support that elevated TSP4 in dorsal spinal cord can induce central sensitization by promoting exaggerated presynaptic excitatory input, excitatory synaptogenesis, and evoked excitability of dorsal horn neurons through activation of TSP4/Ca v α 2 δ 1 dependent processes.
In combination with previous findings from peripheral nerve injury models, our data reveal the importance of TSP4/Ca v α 2 δ 1 dependent processes in mediating central sensitization and chronic pain states post injury. Although TSP4 is upregulated within days after peripheral nerve injury in both DRG and dorsal spinal cord (2,69), there is a significant delay in peak Ca v α 2 δ 1 upregulation in the dorsal horn (weeks) (29), primary due to time required for initial translocation of elevated Ca v α 2 δ 1 from DRG to pre-synaptic central terminals of sensory afferents in the dorsal horn (27,29). The subsequent interaction of elevated TSP4 with excess pre-synaptic Ca v α 2 δ 1 in the dorsal horn then promotes aberrant excitatory synaptogenesis and dorsal horn neuron sensitization to maintain chronic pain states (Fig.  7). While dorsal horn neurons are heterogeneous and play distinct roles in transmitting modality specific information (56,57), our sampling method would not allow us to perform a sensory-neurontype specific dissection of these changes. Further studies, for example, using Cre-directed ablation of genes of interest from sub-populations of sensory neurons, are necessary to provide more deep mechanistic insights about the TSP4/Ca v α 2 δ 1 dependent processes in mediating modality specific nociception.
Interestingly, while TSP4-induced dorsal horn neuron sensitization and behavioral hypersensitivity is a slow process that requires four days to reach the peak effects, gabapentin at a clinically relevant concentration can block these effects within one hour. Even through we cannot rule out the possibility that a mechanism independent of its binging to Ca v α 2 δ 1 mediates the actions of gabapentin in reversing TSP4-induced dorsal spinal cord neuron sensitization and behavioral hyperalgesia, our data support that keeping the TSP4/Ca v α 2 δ 1 dependent processes in an active state is probably critical for the maintenance of TSP4-induced central sensitization and behavioral hypersensitivity. The fast antihyperalgesia actions of gabapentin may derive from interfering with the TSP4/Ca v α 2 δ 1 dependent processes that are likely key elements of pain genesis even after long-term pathological changes, such as aberrant excitatory pre-synaptic input and synaptogenesis, are established days post TSP4 injection or peripheral nerve injury. This may explain why gabapentinoids exhibit inhibitory effects on sensitized spinal neurons (33,70,71), neuropathic pain states in animal models (24,25,72-74) and patients (19)(20)(21)(75)(76)(77)(78)(79)(80)(81)(82), but do not affect baseline sensory neuron excitability and sensory thresholds in control animals (33, 70,71,73,83) and healthy volunteers (84,85), who should not have increased expression of Ca v α 2 δ 1 and/or TSP4 in the sensory pathway. This may also explain why gabapentin is only effective in some, but all, neuropathic pain patients with various etiologies (86) since it is less likely that all pain-inducing conditions are associated with increased expression of Ca v α 2 δ 1 and/or TSP4.
In summary, even though some details remained elusive, our findings provide a large body of multi-dimensional evidence to support that activation of the TSP4/Ca v α 2 δ 1 dependent processes is required for TSP4-induced central sensitization that leads to pain state development. Blocking this pathway may be a novel strategy for development of target-specific analgesics for chronic pain management.      induced tactile allodynia in WT CKO mice that was diminished in CKO Adv-Cre mice with Ca v α 2 δ 1 ablation from Advillin + DRG neurons. Means ± SEM. *p<0.05, ***p<0.001, ****p<0.0001 vs pre-treatment level by repeated measures 2-way ANOVA with Bonferroni post-tests. Peak allodynia (B) and thermal hyperalgesia (C) seen in WT CKO mice were blocked in CKO Adv-Cre mice. Means ± SEM from (n) indicated. ****p<0.0001 vs saline group by Student's t test. Fig. 6. TSP4 ablation blocked SNL-induced excitatory synaptogenesis. Co-immunostaining of Ca v α 2 δ 1 with synaptic markers was performed in thin sections of dorsal spinal cord samples from twoweek SNL WT or TSP4 KO mice when behavioral hypersensitivity occurred in the injury (Ipsi.) side of SNL WT mice, but not TSP4 KO mice (2). (A) Representative images showing Ca v α 2 δ 1 (red) and VGlut 2 (green) immunoreactivity and their colocalization (yellow) in superficial dorsal horn. Scale bar = 5 µm for all image panels. Summarized data of total Ca v α 2 δ 1 immunoreactivity intensity (B), VGlut 2 immunoreactivity surface area (C), and VGlut 2 + puncta with (yellow) or without (green) co-localization with Ca v α 2 δ 1 immunoreactivity (D) are presented as the means ± SEM collected from 60 images over three mice in each group. *p < 0.05, **p < 0.01, ***p < 0.001 compared with non-injury (contra.) side by paired Student's t test.