Pain Perception in Mice Lacking the β3 Subunit of Voltage-activated Calcium Channels*

The importance of voltage-activated calcium channels in pain processing has been suggested by the spinal antinociceptive action of blockers of N- and P/Q-type calcium channels as well as by gene targeting of the α1B subunit (N-type). The accessory β3 subunits of calcium channels are preferentially associated with the α1B subunit in neurones. Here we show that deletion of the β3 subunit by gene targeting affects strongly the pain processing of mutant mice. We pinpoint this defect in the pain-related behavior and ascending pain pathways of the spinal cordin vivo and at the level of calcium channel currents and proteins in single dorsal root ganglion neurones in vitro. The pain induced by chemical inflammation is preferentially damped by deletion of β3 subunits, whereas responses to acute thermal and mechanical harmful stimuli are reduced moderately or not at all, respectively. The defect results in a weak wind-up of spinal cord activity during intense afferent nerve stimulation. The molecular mechanism responsible for the phenotype was traced to low expression of N-type calcium channels (α1B) and functional alterations of calcium channel currents in neurones projecting to the spinal cord.

The importance of voltage-activated calcium channels in pain processing has been suggested by the spinal antinociceptive action of blockers of N-and P/Q-type calcium channels as well as by gene targeting of the ␣1B subunit (N-type). The accessory ␤3 subunits of calcium channels are preferentially associated with the ␣1B subunit in neurones. Here we show that deletion of the ␤3 subunit by gene targeting affects strongly the pain processing of mutant mice. We pinpoint this defect in the pain-related behavior and ascending pain pathways of the spinal cord in vivo and at the level of calcium channel currents and proteins in single dorsal root ganglion neurones in vitro. The pain induced by chemical inflammation is preferentially damped by deletion of ␤3 subunits, whereas responses to acute thermal and mechanical harmful stimuli are reduced moderately or not at all, respectively. The defect results in a weak wind-up of spinal cord activity during intense afferent nerve stimulation. The molecular mechanism responsible for the phenotype was traced to low expression of N-type calcium channels (␣1B) and functional alterations of calcium channel currents in neurones projecting to the spinal cord.
Voltage-activated calcium channels represent a family of ionic channels that play a crucial role in the nervous system by controlling membrane excitability and neurotransmitter release. Neurones express L-, N-, P/Q-, and R-type calcium channels, which are composed of the pore-forming subunit ␣1 and the accessory subunits ␤, ␣2␦, (1, 2) and probably ␥ (3). The importance of ␣1D (L-type), ␣1B (N-type), ␣1A (P/Q-type), and ␣1E (probably R-type) for the function of the nervous system in vivo is underlined by recent studies with genetically engineered mice (4). By contrast, the in vivo function of the four known ␤ subunits (␤1, ␤2, ␤3, ␤4) of calcium channels is less understood. In heterologous expression systems, ␤ subunits are required for the functional expression of ␣1 subunits (5-7), enhance the current density, and shift the voltage dependence of the activation and inactivation of recombinant calcium chan-nels (8 -11). Additionally, ␤ subunits have been implicated in the modulation of calcium channels by G proteins (12)(13)(14). In vivo, the expression of ␤ subunits is modified in pathological states such as cardiac dysfunction (15) and diabetes (16). A natural occurring mutation of ␤4 is associated with ataxia and absence seizures in lethargic (lh/lh) mice, and this mutation modifies N-and P/Q-type calcium channels in the brain (17). The targeted disruption of ␤1 is lethal (18). In sympathetic neurones of ␤3-deficient mice, N-and L-type calcium channel currents are reduced, the activation of P/Q type calcium channels is altered, and no difference appears concerning inhibitory effects of norepinephrine on calcium channel currents (19). Such alterations of calcium channel currents in sympathetic neurones might be related to the cardiovascular phenotype of mice lacking the ␤3 subunit (20). Yet, the in vivo functions of the ␤3 subunit of voltage-activated calcium channels are not well established.
As the ␣1B subunit of N-type calcium channels is primarily associated with the ␤3 subunit in neurones (21,22), an altered neurotransmitter release at synaptic sites may be expected in the ␤3-deficient mice. Specifically, blockade of N-type calcium channels inhibits tachykinin release from afferent sensory nerves (23). Within the spinal cord, N-type calcium channels show the highest density in the superficial laminae (24,25). Accordingly, deletion of the ␣1B subunit (Ca v 2.2) (26 -28) and intrathecal injection of the N-type channel blocker, -conotoxin GVIA (29), reduce the behavior associated with pain in rodents. Furthermore, the ␤3 subunit might also associate with ␣1A subunits to form P/Q-type calcium channels, which are present in the deeper laminae of the spinal cord (30) and might be involved in nociceptive processing (29). In the present study, we examine the possible function of ␤3 subunits in nociception and sensory processing using a mouse model in which the ␤3 subunit has genetically been deleted. Because dorsal root ganglions (DRG) 1 contain the cell bodies of afferent sensory fibers and predominantly express N-type calcium channel currents (31), binding assays, and pharmacological testing on DRG neurones with the selective marker of N-type channels, -conotoxin GVIA, were used to determine the expression of ␣1B in DRG neurones. We have also explored for changes in the voltage dependence and G protein modulation of calcium channel currents in isolated DRG neurones. In vivo, we analyzed the pain-related behavior and the spinal cord response to an in-tense and persistent barrage of afferent nociceptive impulses (wind-up).

EXPERIMENTAL PROCEDURES
Generation of ␤3ϩ/Ϫ Mice-The organization of the ␤3 gene (32) and the targeting vector used to generate null mutations of the ␤3 gene are shown in Fig. 1. Linearized targeting constructs were electroporated into R1 embryonic stem (ES) cells, and the cells were selected with G418 and ganciclovir. We identified 3 of 470 ES cell clones with predicted genomic structures for the targeting vector IIMB. Selected ES cell clones were microinjected into C57BL/6 blastocysts and transferred into the uteri of pseudopregnant recipient females. Two independent ES cell clones were transmitted through the germ line. Mice were kept in essentially specific pathogen-free environment.
Construction and Purification of Glutathione-S-transferase (GST)-␤1b Fusion Protein and Generation of the Antibodies against ␤1b-Nucleotides 1269 -1857 (aa 403-597) of ␤1b (33) were amplified by PCR using the forward primer, 5Ј-cgg gat ccg agt act tgg aag cct act g, and the reverse primer, 5Ј-cgg aat tct cag cgg atg tag acg cct t. The amplified cDNA was ligated into the BamHI/EcoRI sites of pGEX-4T2 (Amersham Biosciences). The accuracy was confirmed by sequencing both strands. BL21(DE3) Escherichia coli were transformed with pGEX-4AT2-␤1b, and protein expression was carried out as previously described (34). A synthetic peptide ( 584 KNELEGWGQGVYIR 597 ) of ␤1b (33) was coupled to keyhole limpet hemocyanin for injection into rabbit 234. Antibody 234 recognizes the GST-␤1b fusion protein (M r ϳ59,000) and native ␤1b in brain and skeletal muscle (M r ϳ72,000). It does not recognize a GST-␤1a fusion protein (M r ϳ33,000) or native ␤1a (M r ϳ52,000), which is predominantly expressed in skeletal muscle. Other antibodies have been described previously (22). Immunoblots were repeated up to six times. Pooled tissue from up to three animals (brain microsomes) and up to 20 animals (DRGs) per genotype, respectively, were processed. The data shown (Fig. 1, D-E) represent independent, non-stripped, and non-reused blots.
Nociceptive Tests and Rotorod Test-The test for nociception were performed as described (35). The tail flick response was evoked by a light beam (irradiated heat) or by immersing two thirds of the tail in water at 52°C. In the hot plate test, mice were placed on a plate (56°C), and responses were counted when the mouse first licked the hind paws and jumped from the plate (cut-off time, 30 and 60s, respectively). Formalin (25 l; 2% paraformaldehyde in phosphate-buffered saline) was injected into the plantar surface of the left hind paw. The time spent licking or biting the injected paw was recorded as the nociceptive score. Inflammation was induced by subcutaneous injection of 10 l of complete Freund's adjuvant (CFA, 0.5 mg/ml Mycobacterium butyricum in paraffin oil and emulsifying agent) into the plantar surface of the left hind paw under short anesthesia with diethyl ether. Before and 60 h after the CFA injection, the animals were placed on a raised grid, and mechanical thresholds were measured using calibrated von Frey hairs. In the rotorod test, mice were trained on the rod for 1 min at rest and 5 min at 16 rpm, and 1 min at rest and 5 min at 24 rpm. For the test, mice were placed on the apparatus facing away from the experimenter and in the direction opposite to the rotating rod while it was moving at 32 rpm (cut off time 5 min). The tests were performed with adult male mice. The genotype of the animals was unknown to the investigator during the tests.
Extracellular Recordings in the Spinal Cord-Adult male mice (30 -40 g) of both genotypes were used for extracellular unit recordings. The anesthesia was induced with diethyl ether and subsequently with pentobarbital (intraperitoneal 0.5 mg/10 g of body weight). To control the deep of anesthesia, the tail vein was canulated, and pentobarbital was given in the form of intravenous bolus (0.05 mg/10 g of body weight) every 20 -25 min during the experiment. A laminectomy was performed to expose the spinal segments Th8-Th10, and the thoraxic vertebras were fixed with spinal hooks in a stereotaxic frame. The left sural nerve was exposed and placed on a bipolar electrode for stimulation (Grass). The experiments were started after a recovery period of 30 -45 min. Extracellular recordings were made with tungsten microelectrodes of 5 megohms resistance (WPI, Berlin, Germany) contralateral to the stimulated nerve (i.e. in the right spinal cord). Initially, the microelectrode was placed 350 -500 m lateral to the median dorsal spinal vein in the spinal segment Th8 and carefully inserted just under the surface. To search for responsive units, the electrode was driven from this zero point deep through the spinal cord and trains of 50-V pulses (5 ms long) were regularly delivered to the sural nerve. We used a frequency of 0.2 pulses per second (pps) in this searching phase to prevent wind-up. If the test pulses did not evoke spikes within driving distances of maxi-mally 250 m, the electrode was removed and the searching phase was reinitiated from a new zero point caudal to a previous one. When the test pulses evoked spikes, we first rule out the contribution of receptive fields by stimulating mechanically (brush, pinch) the skin of the back and hind paws on both ipsilateral and contralateral sides of recording. Because the sural nerve was cut distal to the bipolar electrode, responsive units were used for further analysis when no response to cutaneous stimulation was observed. The thresholds were determined with a series of 12 pulses of amplitudes between 10 V and 100 V (5 ms long, 0.2 pps). The wind-up was finally evoked with 50-V pulses (5 ms long) delivered for 16 s at 1 pps. Data were collected and analyzed using the Chart software for the PowerLab system (ADInstruments).
Isolation of DRG Neurones-DRGs were obtained from adult wild type and ␤3Ϫ/Ϫ mice, and neurons were dissociated by incubation in the presence of trypsin type I (0.67 mg/ml medium; Sigma) and collagenase type II (4 mg/ml medium; Biochrom, Berlin, Germany) for 20 min. The cell suspension was centrifuged for 5 min, the supernatant withdrawn, and Dulbecco's modified Eagle's medium supplemented by 10% fetal calf serum was added. Isolated neurones were plated onto poly-L-lysine (Sigma)coated glass cover slips and kept in an incubator. Electrophysiological experiments were performed within 4 -16 h after dissociation.
Whole Cell Patch Clamp-For electrophysiological recordings glass cover slips were placed into a recording chamber and mounted on an inverted microscope. Cells were perfused with extracellular solution, and electrophysiological recordings were started after 10 min. DRG neurons were identified based on their typical morphology using a 40ϫ phase contrast objective. Ba 2ϩ currents (I Ba ) from wild type and ␤3Ϫ/Ϫ DRG neurons were recorded using standard whole cell recording techniques (36). I Ba was recorded with an EPC-9 amplifier (HEKA). Upon establishment of the whole cell configuration, series resistance, fast and slow capacitance components were compensated using the internal circuit of the EPC-9 amplifier. The currents were leak-subtracted by applying n/4 correction mode. Data were sampled at 10 kHz and filtered at 3.3 kHz. Data were analyzed off line using the Pulse-Fit (HEKA) software package. For further analysis, current traces were imported into Sigmaplot (Jandel Scientific). For display current traces were transferred to CorelDraw (Corel Graphics). All patch clamp experiments were performed at room temperature. Pipettes of 2-3 megohms resistance were prepared on a DMZ Universal Puller from 1.5 mm borosilicate glass capillaries (Clark Electromedical Instruments). The composition of the recording solutions used was the following (in mM): intracellular solution, CsCl 95, CsSO 4 40, tetraethylammonium chloride 20, CaCl 2 1, EGTA 10, Mg-ATP 3, Hepes 10 pH 7.3 (CsOH); extracellular solution, tetraethylammonium hydroxide 135, BaCl 2 10, MgCl 2 1.2, Hepes 10, pH 7.4 (tetraethylammonium hydroxide).
Conotoxin Binding Assays-Whole dorsal root ganglions were isolated and homogenized in 5 volumes of 5 mM TRIS-HCl, pH 7.4, containing a protease inhibitor mix (0.1 mM phenylmethylsulfonyl fluoride, 1 mM phenanthroline, 1 mM iodoacetamide, 1 mM benzamidine, 1 M pepstatin A, 1 g/ml antipain, and 1 g/ml leupeptin). Equilibrium binding of 125 Ilabeled -conotoxin GVIA (Amersham Biosciences) in the absence and presence of 10 nM unlabeled -conotoxin GVIA was performed in duplicate at a total incubation volume of 500 l containing 10 g of protein.
Binding to isolated DRG neurones was performed as follows: intact neurones were incubated in the presence of 50 pM 125 I-labeled -conotoxin GVIA for 2 h and then washed three times with Dulbecco's modified Eagle's medium containing 0.1% bovine serum albumin. Nonspecific binding measured in the presence of 200 nM unlabeled -conotoxin GVIA was less than 20% of total binding. -Conotoxin GVIA binding has been quantified by non-linear regression analysis using the SPSS (Statistical Package for Social Sciences) software.

RESULTS
Targeted Disruption of the ␤3 Gene-The ␤3 gene (Fig. 1A) was mutated by replacing part of its exon 3 and the complete exon 4 (32) with a neomycin resistance gene (neo r ). Breeding of heterozygous mice generated ␤3 homozygous (␤3Ϫ/Ϫ) mice at the rate expected from the Mendelian frequency (160 ϩ/ϩ, 307 ϩ/Ϫ and 144 Ϫ/Ϫ). The ␤3Ϫ/Ϫ mice grew normally, lived longer than one year, were fertile, and had no obvious symptoms. Offspring were genotyped for the ␤3 mutation by Southern blot analysis. Wild type and mutant alleles were indicated by the presence of a 12and 8-kb EcoRI fragment, respectively (Fig. 1B). The deletion of the ␤3 gene was confirmed by Northern blot analysis (Fig. 1C) and immunoblotting of brain extracts (Fig. 1D). Additionally, we found no significant differences between wild type and ␤3Ϫ/Ϫ mice in the expression levels of ␤1, ␤2, and ␤4 proteins in whole brain (Fig. 1D), suggesting that there was no apparent compensatory increase of the expression of other ␤ subunits. Furthermore, it is likely that ␤3 subunits are also expressed in other parts of the nervous system. The DRGs contain cell bodies of primary afferent sensory fibers, including myelinated A fibers and unmyelinated C fibers, which project into the spinal cord. Thus, DRG neurones represent a central component of pain pathways, and additionally, isolated DRG neurones represent a well known cell system in studies of neuronal calcium channels (31). Because the expression pattern of ␤ subunits in DRG neurones is not known, we analyzed extracts of isolated DRGs using specific antibodies. Western blot analysis detected the ␤3 protein in DRGs from control animals but not from ␤3Ϫ/Ϫ mice (Fig. 1E). These results indicate that the gene targeting strategy used in the present study results in suppression of ␤3 protein expression in the DRGs of the adult animals that were selected for further experiments. The protein level of the ␤1, ␤2, and ␤4 subunits was very low but detectable (Fig. 1E). As in brain extracts, no significant change in the expression level of ␤1, ␤2, and ␤4 was apparent in extracts of DRGs isolated from ␤3Ϫ/Ϫ mice.
Nociception Is Altered in ␤3Ϫ/Ϫ Mice-Because ablation of ␤3 suppresses the expression of ␤3 subunits in the brain and sensory neurones (Fig. 1), it is likely that the deletion of the ␤3 subunit might have an impact on neuronal functioning in various regions of the nervous system and, thus, at different levels of pain processing pathways. It is known that blockers of Ntype calcium channels exert spinal antinociceptive actions (37)(38)(39). Additionally, P/Q-type calcium channels may be also involved in nociceptive processes at the spinal level (38,40). Because ␤3 subunits associate to form N-and P/Q-type calcium channels (21,22,41), the suppression of ␤3 subunit expression ( Fig. 1) likely produces major defects in the nociceptive pathway, and therefore, we analyzed the pain-related phenotype of the ␤3Ϫ/Ϫ mice in behavioral test for thermal, mechanical, and chemical nociception (42). Responses to acute thermal stimuli were studied using the tail flick ( Fig. 2A) and hot plate tests (Fig. 2B). Tail flick is primarily a spinal reflex, whereas a substantial supraspinal component, which involves lifting and licking of the hind paw, is required in the hot plate assay (35). Tail flick latencies were longer in the ␤3Ϫ/Ϫ mice both when the tail was exposed to heat by irradiation and by immersion in water at 52°C (Fig. 2A). In the hot plate test, the latencies for licking were also increased and the delay for escape jumping was almost doubled in the ␤3Ϫ/Ϫ mice (Fig. 2B). Thus, independent of whether spinal or supraspinal components are engaged in the nociceptive response, the thresholds for thermal stimuli appear to be higher in the ␤3Ϫ/Ϫ mice. By contrast, the responses to acute mechanical stimuli appear to be normal in the ␤3Ϫ/Ϫ mice since the mechanical thresholds measured with von Frey filaments were similar to the thresholds of wild type mice ( Fig. 2C; wild type, 8.12 Ϯ 1.6 mN, n ϭ 10; ␤3Ϫ/Ϫ, 6.15 Ϯ 0.7 mN, n ϭ 10). To study the role of the ␤3 subunit in persistent pain, we examine the behavior of ␤3Ϫ/Ϫ mice employing CFA and formalin. CFA was injected in the foot pad of the left hind paw and, within 24 -60 h, induced swelling in the injected paw in wild type mice (injected paw, 2.35 Ϯ 0.07 mm, contralateral side, 2.02 Ϯ 0.07 mm, p Ͻ 0.05, n ϭ 10) and ␤3Ϫ/Ϫ mice (injected paw, 2.47 Ϯ 0.09 mm, contralateral side, 2.17 Ϯ 0.06 mm, p Ͻ 0.05, n ϭ 10). The degree of swelling of the injected paw was similar in both groups of mice. After 60 h, the mechanical thresholds measured on the ipsilateral injected paw ( Fig. 2C; wild type, 1.65 Ϯ 0.4 mN, n ϭ 10; ␤3Ϫ/Ϫ, 1.40 Ϯ 0.4 mN, n ϭ 10) and on the contralateral paw (wild type, 5.21 Ϯ 0.5 mN, n ϭ 10; ␤3Ϫ/Ϫ, 3.87 Ϯ 0.6 mN, n ϭ 10) were significantly reduced (p Ͻ 0.05) in wild type and ␤3Ϫ/Ϫ mice when comparing them to the respective thresholds measured before the CFA injection ( Fig. 2C; wild type, 8.12 Ϯ 1.6 mN, n ϭ 10; ␤3Ϫ/Ϫ, 6.15 Ϯ 0.7 mN, n ϭ 10). Thus, CFA induced hyperalgesia on the ipsilateral and contralateral side of injection in both groups of animals. However, no difference between wild type and ␤3Ϫ/Ϫ mice was detected in the mechanical thresh-olds measured after CFA induced swelling. It appears, therefore, that ␤3 subunits are not critical for the development of hyperalgesia in the CFA-induced inflammatory pain. The neurochemical signature of the CFA-induced pain is the internalization of substance P receptors in laminae I, II-IV of the spinal cord (43). Considering that N-type calcium channels are primarily restricted to laminae I-II (24,25) and P type channels are localized in deeper laminae (30), it is likely that alterations of N-type channels in the ␤3Ϫ/Ϫ mice become more evident in specific assays of nociception, like the formalin-induced inflammation that is associated with internalization of substance P receptors in lamina I (43). Therefore, we examined the responses of the ␤3Ϫ/Ϫ mice to chemical stimuli in the formalin test, which is also a well established model to study central sensitization events at the spinal level after peripheral inflammatory states (35). Subcutaneous injection of formalin in the hind paw elicits biphasic pain responses in wild type and ␤3Ϫ/Ϫ mice (Fig. 3A). In phase 1, formalin stimulates the nociceptors generating acute pain, and in phase 2 the inflammation induced by formalin elicits persistent pain. Both, wild type and ␤3Ϫ/Ϫ mice behaved similarly in phase 1 but the nociceptive behavior during phase 2 was attenuated by 43% in ␤3Ϫ/Ϫ mice (Fig. 3, A and B). Because the tests for thermal, chemical, and mechanical nociception measure motor responses as end point, we examined possible motor/sedative defects in the ␤3Ϫ/Ϫ mice. In the rotorod test, the performance of wild type and ␤3Ϫ/Ϫ mice were indistinguishable (time on the rod: wild type, 64.62 Ϯ 17.0 s, n ϭ 8; ␤3Ϫ/Ϫ, 41.86 Ϯ 9.8 s, n ϭ 8), indicating that locomotor functions are not altered in the ␤3Ϫ/Ϫ mice, and thus, the behavioral tests used in the present study delineate the profile of nociception in the ␤3Ϫ/Ϫ mice. In summary, the ␤3Ϫ/Ϫ mice showed higher thresholds for acute thermal stimuli, but the responses to mechanical stimuli were not altered either before or after induction of inflammatory pain with CFA. The acute pain induced by formalin was not altered but the persistent pain observed in phase 2 of the formalin test was markedly reduced in the ␤3Ϫ/Ϫ mice.
Neuronal Activity in the Spinal Cord of ␤3-deficient Mice-

FIG. 2. Nociceptive responses of wild type (؉/؉) and ␤3-deficient mice (؊/؊) to thermal and mechanical stimuli.
A, tail flick latency to noxius heat (heat irradiation, n ϭ 13 for both groups; tail immersion at 52°C, n ϭ 14 for both groups). B, hot plate test at 56°C (n ϭ 14 for both groups). C, response to mechanical stimulation with von Frey hairs before and 60 h after injection of CFA (-CFA and ϩCFA, respectively; n ϭ 10 for both groups). All data are mean Ϯ S.E. and were analyzed by the Mann-Whitney U test (*, p Ͻ 0.05) Myelinated A fibers and small diameter, unmyelinated C fibers project to the spinal cord and form neuronal pathways that transmit nociceptive information from the peripheral site of injury to the brain. Among the multiple nociceptive pathways, the spinothalamic pathway originates primarily from neurones in the neck of the dorsal horn and terminates in the ventroposterior and ventrobasal thalamus (44). Within the spinal cord, L-type calcium channels are present in the dorsal horn with a density comparable to brain areas. N-type calcium channels show the highest densities in the superficial laminae of the spinal cord (24, 25, 30) and P-type calcium channels in the deeper laminae (30). Because the response to inflammatory pain appears to be preferentially damped by the deletion of ␤3 subunits, we examined in vivo neuronal responses in the spinal cord evoked by stimulation of peripheral nerves. The neuronal activity was recorded deep in the spinal segment Th8, where responsive units of the spinothalamic pathway are expected, and the stimuli were delivered to the sural nerve of the contralateral side (see "Experimental Procedures"). Under these conditions, trains of pulses with amplitudes up to 30 V, which were applied at a frequency of 0.2 pps, evoked a volley of spikes that lasted less than 100 ms (Fig. 4A). With pulse amplitudes of at least 50 V, we detected in wild type and ␤3Ϫ/Ϫ mice at least two types of responsive units that can be distinguished by the Voltage pulses with amplitudes between 10 and 100 V were delivered to the sural nerve at a frequency of 0.2 pps, and the evoked activity was recorded in the contralateral spinal segment Th8. A, representative recordings of spikes (upper panels) produced by responsive units in experiments with a wild type (ϩ/ϩ) and a ␤3Ϫ/Ϫ mouse (Ϫ/Ϫ). Voltage pulses with the indicated amplitudes were applied at the time points indicated by arrows. The histograms of spike latencies (lower panels) were compiled from the responses to trains of 12 pulses with an amplitude of 50 V. B, thresholds of responsive units. The spikes observed within 300 ms after each voltage pulse were counted and averaged for trains of 12 pulses (frequency, 0.2 pps; duration, 5 ms). To estimate the activity evoked by a given pulse amplitude, the averaged number of spikes of the corresponding train of pulses was normalized to the maximal value obtained with pulse amplitudes between 10 and 100 V. Each symbol represents a different experiment (wild type mice, upper panel, n ϭ 5; ␤3Ϫ/Ϫ mice, lower panel, n ϭ 4). Stimuli above 50 V produced maximal spinal cord activity in wild type and ␤3Ϫ/Ϫ mice. latencies of the evoked spikes (Fig. 4A, lower panels). The latencies shorter than 70 ms probably reflect the contribution of A fibers, whereas longer latencies likely reflect the activation of C fibers. To compare the thresholds for the activation of responsive units in wild type and ␤3Ϫ/Ϫ mice, we analyzed the spinal activity evoked by pulses with amplitudes between 10 and 100 V. Under the present experimental conditions, stimuli with amplitudes above 50 V evoked maximal spinal activity in wild type and ␤3Ϫ/Ϫ mice (Fig. 4B). The activity evoked in the spinal cord, however, depends on the frequency of stimulation. During trains of high frequency pulses, the neuronal activity increases spontaneously, a phenomenon that is observed in various nociceptive pathways and is known as wind-up (45). Thus, to examine the wind-up, we selected an amplitude of 50 V to evoke maximal responses (Fig. 4B) and applied trains of pulses with this amplitude at frequencies higher than 0.2 pps.
In several experiments with wild type mice, a moderate wind-up with a delay of 10 -15 s was observed with a frequency of 0.5 pps (not shown). Therefore, we compared the wind-up in wild type and ␤3Ϫ/Ϫ mice at a frequency of 1 pps (Fig. 5). In experiments with wild type mice, we consistently observed maximal activity after 5-7 s of repetitive stimulation, which were accounted for by the increase of spikes with latencies longer than 100 ms (Fig. 5A). Similar wind-up of the spinal cord activity was not observed in paired experiments with ␤3Ϫ/Ϫ mice (Fig. 5, A and B). Thus, this first characterization of the spinal cord activity in vivo suggests that the activation thresholds of responsive units appears not to be affected, but the wind-up is absent in the ␤3Ϫ/Ϫ mice.
Calcium Channel Currents in DRG Neurones-The previous results pinpoint the nociceptive defects resulting from the deletion of ␤3 subunits to the spinal level (Figs. 2-5). Because the expression of ␤1, ␤2, and ␤4 subunits appears unaltered in DRG neurones of ␤3-deficient mice (Fig. 1), the ␣1A, ␣1B, ␣1C, ␣1D, and ␣1E subunits, which have been detected in murine DRG neurones (46), might be associated either with ␤1, ␤2, or ␤4 subunits in the ␤3-deficient mice. To determine the functioning of neuronal calcium channels in the ␤3-deficient mice, we recorded whole cell barium currents (I Ba ) in isolated DRG neurones. Depolarizing voltage steps were applied from holding potentials of Ϫ50 and Ϫ110 mV, and difference currents were obtained by subtracting the records at Ϫ100 mV from the respective traces at Ϫ50 mV (Fig. 6). The large variability of DRG neurones concerning cell size and calcium channel expression (47,48) precluded, however, a quantitative analysis of current amplitudes. Another observation in this series of experiments was that the activation and inactivation time courses of compound and difference currents were similar in ␤3Ϫ/Ϫ DRG neurones and controls (Fig. 6A). Because the expression of ␤ subunits speeds up the activation and inactivation of recombinant calcium channels (e.g. 10) and antisense depletion of ␤ subunits produces the converse effects in DRG neurones (49), this observation indicates that ␤1, ␤2, or ␤4 are able to replace ␤3, at least, in what concerns activation and inactivation. Additionally, the coexpression of ␤ with ␣1 subunits shifts the membrane potential required for channel opening to more hyperpolarized potentials (1,2). We observed that the peak of the aggregate whole cell current measured at a holding potential of Ϫ110 mV was typically at Ϫ10 mV in control and at 0 mV in the ␤3Ϫ/Ϫ mutants (Fig. 6A). As can be seen in the normalized current-voltage relationships obtained at a holding potential of Ϫ70 mV (Fig. 6B), the whole cell current activation is shifted by ϳ8 mV to more depolarized potential in ␤3Ϫ/Ϫ neurones when compared with wild type. The activation midpoints in the normalized curves of Fig. 6B are Ϫ15.4 and Ϫ23.5 mV for ␤3Ϫ/Ϫ and wild type neurones, respectively. When the current voltage relationships of the individual neurones were fitted with a Boltzmann equation as described previously (50), the calculated potentials for halfactivation in ␤3Ϫ/Ϫ and wild type neurones were Ϫ9.1 Ϯ 1.1 mV (n ϭ 17) and Ϫ18.1 Ϯ 2.1 mV (n ϭ 11), respectively. A similar shift in the voltage dependence of calcium channel currents was observed in sensory neurones treated with generic anti-␤-antisense oligonucleotides (50). Because ␤ subunits have been implicated in the G protein-mediated inhibition and voltage-dependent facilitation of neuronal calcium currents (13,14), an abnormal modulation of whole cell calcium channel current by G proteins is expected in the ␤3-deficient mice. As previously reported (51,52), we found that activation of G proteins with GTP-␥-S slowed the activation kinetics of calcium channel currents and reduced the current amplitude in wild type DRG neurones (Fig. 7A). Depolarizing prepulses accelerated the activation of whole cell calcium channel currents, increased the current amplitude and shifted the current-voltage relationship to more negative potentials (Fig. 7A), in line with the voltage-dependent unblocking of G protein-mediated inhibition of neuronal calcium channels (53,54). The prepulse facilitation resulted in a voltage-dependent facilitation of calcium channel currents (Fig. 8A). By contrast, we found that depolarizing prepulses sped up the activation time course of whole cell calcium channel currents after dialysis of GTP-␥-S in the majority of ␤3Ϫ/Ϫ DRG neurones (9 of 13), but no increase of current amplitude was observed, and accordingly, there was no shift in the current-voltage relationship (Fig. 7B). Thus, the main difference between this group of ␤3Ϫ/Ϫ neurones and wild type neurones is observed in the degree of prepulse facilitation of whole cell calcium currents (Fig. 8, A and B). This observation is in line with previous studies, which showed that the degree of facilitation of recombinant calcium channels is strongly dependent on the expressed ␤ subunit, although the activation time course of whole cell currents facilitated by a prepulse is similar for all ␤ subunits (55). In the remaining ␤3Ϫ/Ϫ DRG neurones (4 of 13), GTP-␥-S had no effect at all and depolarizing prepulses produced an strong inactivation of calcium channel currents (Fig. 8B). All in all, this first electrophysiological characterization of DRG neurones of ␤3-deficient mice indicates that the voltage-dependent facilitation of calcium channel currents is altered and the voltage dependence of FIG. 6. Voltage dependence of calcium channel currents in DRG neurones. Whole cell currents (I Ba ) were evoked from various holding potentials (HP) with depolarizing pulses to potentials between Ϫ40 and ϩ40 mV in 10-mV increments. A, representative whole cell current families from DRG neurones of wild type (left panels, ϩ/ϩ) and ␤3Ϫ/Ϫ mice (right panels, Ϫ/Ϫ) elicited from the indicated holding potentials. Difference currents were obtained by subtracting current traces at HP ϭ Ϫ50 mV from the respective traces at HP ϭ Ϫ110 mV. In the wild type neurone, the I-V relationships obtained with holding potentials of Ϫ110 mV (continuous line) and Ϫ50 mV (dotted line) show peaks of I Ba amplitude at Ϫ10 mV and 0 mV, respectively. I Ba evoked in the ␤3Ϫ/Ϫ neurone from both holding potentials peaked at 0 mV. B, activation of whole cell currents in response to depolarizing pulses from a holding potential of Ϫ70 mV. Current amplitudes were normalized (filled circles, continuous line: wild type, n ϭ 11; open circles, dotted line: ␤3Ϫ/Ϫ neurones, n ϭ 17). The activation midpoints were Ϫ23.5 mV and Ϫ15.4 mV for wild type and ␤3Ϫ/Ϫ neurones, respectively. FIG. 7. Voltage-dependent facilitation of calcium channel currents in DRG neurones. After cell dialysis with GTP-␥-S (500 M), whole cell currents were elicited with test pulses from a holding potential of Ϫ70 mV to potentials between Ϫ30 mV and ϩ30 mV. The voltage-dependent facilitation was induced with a prespulse to ϩ80 mV. Superimposed are consecutive whole cell current traces obtained without (1) and with (2) prepulses at the test potentials indicated on the left (upper panels). The I-V relationships (bottom panels) show whole cell current amplitudes measured 10 ms after onset of test pulses without (1, dotted line) and with prepulse (2, continuous line). The representative example from a wild tpye DRG neurone (A) shows the slow activation of calcium channel currents after dialysis with GTP-␥-S (1) and the prepulse-dependent facilitation (2). Prepulses induced also a shift of the I-V relationship to hyperpolarizing potentials. As illustrated in the recordings from a ␤3Ϫ/Ϫ neurone (B), prepulses sped up the activation of whole cell currents in ␤3Ϫ/Ϫ DRG neurones but had almost no effect on the I-V relationship. the activation is shifted to more depolarized potentials, although the time course of activation and inactivation is apparently normal.
Expression of Calcium Channel ␣1B Subunits in Dorsal Root Ganglions-In murine DRG neurones, ␣1B, ␣1A, as well as ␣1C, ␣1D, and ␣1E have been detected with specific antibodies (46). Considering this wide expression of ␣1 subunits in DRG neurones, the deletion of the ␤3 subunit ( Fig. 1) could alter in principle N-, P/Q-, L-and R-type calcium channels. In sympathetic neurones of ␤3-deficient mice, N-type calcium channel current densities are reduced, while P/Q-type current densities appear to be normal (19). At the synaptic sites in the spinal cord, N-type calcium channels appear to mediate neurotransmitter release (56,57). As a consequence, ␤3 deletion could be important for nociception and sensory processing due to the reduction of N-type calcium channels in DRG neurones, as indicated by our whole cell current recordings (Fig. 6). Because the N-type component of calcium currents appears to be variable within the heterogeneous population of DRG neurones (47,48), we first analyzed the expression of ␣1B in DRG homogenates using binding assays with 125 I-labeled -conotoxin GVIA, a N-type channel blocker. The specific binding estimated with increasing concentrations of 125 Ilabeled -conotoxin GVIA was reduced by a factor of ϳ2 in homogenates of ␤3Ϫ/Ϫ DRG neurones (Fig. 9A). This result suggests that the expression of ␣1B is reduced in the mutant. Because ␤ subunits control the membrane targeting of ␣1 subunits, we repeated binding assays with intact neurones to estimate the expression of ␣1B in the membrane. As in the experiments with DRG homogenates (Fig. 9A), the specific binding at a saturating concentration of the N-type blocker was reduced in intact DRG neurones (Fig. 9B), indicating that the number of N-type calcium channels present in the membrane of DRG neurones is reduced in the ␤3-deficient mice. Because these observations suggest that the proportion of N-type currents is reduced in the mutant mice, we determine the inhibitory effects of -conotoxin GVIA on whole cell currents of DRG neurones. As reported previously for sympathetic neurones (19), we observed that the percentage of current inhibition by -conotoxin GVIA ranges from 37% to 70% in wild type and from 0% to 68% in ␤3Ϫ/Ϫ DRG neurones. The cumulative distributions indicate, however, that the percentage of inhibition is often less than 25% in ␤3Ϫ/Ϫ DRG neurones (Fig.  9C, lower panel). On average, the inhibitory effects of -conotoxin GVIA are less pronounced in ␤3Ϫ/Ϫ than in wild type DRG neurones (Fig. 9C, upper panel). Thus our binding studies and pharmacological experiments with -conotoxin GVIA imply that the number of functional N-type calcium channels and, consequently, the proportion of N-type calcium channel currents are reduced in DRG neurones of ␤3-deficient mice. DISCUSSION In the present study, we examined the role of the ␤3 subunit of calcium channels in the pain perception. The use of a genetargeting strategy to suppress the expression of ␤3 subunits allowed the analysis throughout various levels, including protein expression and calcium channel currents in single DRG neurones, functioning of neuronal circuits in the spinal cord of anesthetized animals, and nociceptive behavior in awake mice. Our results show that ␤3 subunits are critically involved in the function of nociceptive pathways and, thus, delineate the in vivo function of one of the so-called auxiliary subunits of voltagedependent calcium channels.
In behavioral assays, we studied the thermal, mechanical, and chemical nociception of the ␤3Ϫ/Ϫ mice. Responses to acute mechanical and thermal stimuli were not altered or only moderately increased in the ␤3Ϫ/Ϫ mice, respectively (Fig. 2). Similarly, the thresholds for activation of responsive units in the spinothalamic pathway were not apparently changed in the ␤3Ϫ/Ϫ mice (Fig. 4). However, the formalin test, which is based on the measurement of licking/biting time as a robust predictor for nociception (42), revealed that ␤3Ϫ/Ϫ mice develop weak hyperalgesia/allodynia after inflammation because the nociceptive behavior was strongly damped in the tonic, inflammatory phase 2 (Fig. 3). In this respect, the response of ␤3Ϫ/Ϫ mice compares to the behavior of mice lacking the ␣1B subunit, which also develop weak hyperalgesia/allodynia after formalin inflammation (26 -28). This parallel is in accordance with our finding that DRG neurones of ␤3Ϫ/Ϫ mice have fewer binding sites for -conotoxin GVIA, a blocker of N-type calcium channels (Fig. 9). Our binding and pharmacological studies with -conotoxin GVIA (Fig. 9), furthermore, also suggest that the deletion of the ␤3 subunit reduced the membrane expression of ␣1B. The contribution of L-type calcium channel currents to the development of formalin-induced hyperalgesia seems unlikely, although our results do not rule out a reduction of L-type current in DRG neurones of ␤3Ϫ/Ϫ mice. When L-type blockers are administered systemically, antinociceptive effects have been observed in rats using the formalin test (e.g. Ref. 58). However, direct application of L-type blockers to the surface of the spinal cord, which is equivalent to intrathecal injection, has no effect on the excitability of dorsal horn neurones produced by formalin inflammation in the rat (38). Accordingly, intrathecal injection of specific blockers of L-type calcium channels has no effect on the nociceptive behavior of rodents either during the early phase or during the late phase of the formalin test (29,39). Unfortunately, a comparison between the phenotypes of ␤3-, ␣1C-, and ␣1A-deficient mice is not possible because ␣1C (Ca v 1.2) and ␣1A (Ca v 2.1) deletion were embryonic lethal and generate strong neurological deficits, respectively (59,60). Nevertheless, application of the P-type blocker -agatoxin IVA to the surface of the spinal cord selectively blocks the excitability of dorsal horn neurones in phase 2 (38), and accordingly, intrathecal injection of the P-type blocker reduces the nociceptive behavior during the late phase but not during the acute phase of the formalin test in the rat (40). On the other hand, intrathecal injection of the N-type blocker -conotoxin GVIA reduces the nociceptive behavior in phase 1 and 2 of the formalin test in mice (39) and rats (29). In the formalin tests that were repeated independently in three laboratories of the authors, no clear differences were observed in the behavior of wild type and ␤3Ϫ/Ϫ mice during the early phase (Fig. 9). Mice deficient in ␣1B subunits also behave normally in the phase 1 (26 -28). Although the unaltered phase 1 in ␣1Band ␤3-deficient mice may reflect compensatory mechanisms, the possibility remains that defects of P-type calcium channels may account partially for the reduced nociceptive responses of ␤3Ϫ/Ϫ mice in the formalin test. We compared the pain-related behavior of the ␤3Ϫ/Ϫ mice in formalin and CFA tests, which are based on inflammatory processes that produce distinct neurochemical signatures in the spinal cord. The CFA-induced inflammation produces internalization of substance P receptors in laminae I, II-IV of the spinal cord, while the formalininduced inflammation appears to induced substance P receptor internalization in lamina I (43). Considering that N-type calcium channels are primarily restricted to laminae I-II (24,25) and that P type channels are localized in deeper laminae (30), our results (Figs. 2 and 3) favor the in vivo importance of a possible modification of spinal N-type channels by deletion of ␤3 subunits and underscore the importance of possible alterations of P-type calcium channels. Furthermore, deletion of ␤3 subunits appears to have not much effect on the expression but alters the voltage-dependent activation of P/Q-type calcium channels, at least in sympathetic neurones (19). In the bodies of ␤3Ϫ/Ϫ DRG neurones, we observed a depolarizing shift in the voltage dependence of the activation of compound calcium channels currents (Fig. 6). Thus, deletion of ␤3 alters the function of calcium channels that are expressed in the membrane, in way that makes them less sensitive to membrane depolarization. Additionally, the voltage-dependent facilitation of compound calcium channel currents ( Fig. 7 and 8) and, likely, the G protein modulation of calcium channels are altered in ␤3Ϫ/Ϫ DRG neurones. These defects in the activation and G protein modulation of the expressed calcium channels as well as the low expression of N-type calcium channels (Fig. 9) likely modify the pain perception at the spinal level. To support this conclusion, we studied responsive units in the spinothalamic pathway. In wild type but not in ␤3Ϫ/Ϫ mice, high frequency stimulation of the sural nerve induced wind-up, which probably represents sensitization at the spinal level (Fig. 5). This defect in the processing of pain perception at the spinal level likely reflects the reduced expression of N-type calcium channels together with altered calcium channel currents induced by the ablation of ␤3. By contrast, spinal and supraspinal nociceptive components account for the low sensitivity in the formalin test of mice lacking the ␣1E subunit (Ca v 2.3) (61).
The importance of calcium channel auxiliary subunits in vivo is underlined by the role of ␣2␦ and ␤ subunits in neuroplasticity after nerve injury (62) and in pathological states such as cardiac dysfunction (15) and diabetes (16), respectively. The results of the present study suggest an important in vivo function of ␤3 subunits in the nervous system, specifically, in the pain processing at the level of the spinal cord. It is not known whether the expression of ␤3 subunits is affected in pathological states, but the specific nociceptive action of the deletion of ␤3 subunits indicates that, like ␣1B (63) and ␣2␦ (64), the ␤3 A, specific CTX binding in homogenates of dorsal root ganglions from wild type (ϩ/ϩ, full circles) and ␤3Ϫ/Ϫ mice (Ϫ/Ϫ, open circles). Apparent B max values are significantly different (ϩ/ϩ: 25.08 fmol/mg protein (95% confidence interval 0.76 -1.27); Ϫ/Ϫ: 13.21 fmol/mg protein (95% confidence interval 0.43-0.65)), whereas K D values do not differ significantly (K D range 1.1 to 8.9 pM). Experiments have been performed in duplicate with DRGs isolated from 60 mice. B, specific CTX binding to intact DRG neurones (ϳ14,400 neurones per determination) isolated from wild type (filled box, n ϭ 7) and ␤3Ϫ/Ϫ mice (open box, n ϭ 6). C, inhibition of whole cell calcium channel currents by CTX (3 M) in DRG neurones. Whole cell currents were elicited by depolarizations (100 ms long) to Ϫ10 mV from a holding potential of Ϫ70 mV. The percentage of inhibition was calculated from peak current amplitudes obtained before and after (50 -100 s) bath application of CTX. Whole cell currents of ␤3Ϫ/Ϫ neurones (upper panel, open box, n ϭ 10) are less susceptible to inhibition by CTX than wild type neurones (upper panel, filled box, n ϭ 6). As seen in the cumulative probability plots for wild type (lower panel, continuous line, n ϭ 6) and ␤3Ϫ/Ϫ neurones (lower panel, dotted line, n ϭ 10), CTX rarely inhibits whole cell currents of ␤3Ϫ/Ϫ neurones by more than 25%. The values represent mean Ϯ S.E., the asterisks indicate p Ͻ 0.05, Student's t test.
subunits are important pharmacological targets for the modulation of pain in pathological states.