Depletion of Calcium Stores in Injured Sensory Neurons: Anatomic and Functional Correlates

All methods and use of animals were approved by the Medical College of Wisconsin (Milwaukee, Wisconsin) Institutional Animal Care and Use Committee.

Male Sprague-Dawley rats weighing 160 to 180 g (Taconic Farms Inc., Hudson, NY) were subjected to spinal nerve ligation in a manner derived from the original technique.22Rats were anesthetized with 2% isoflurane in oxygen, and the right paravertebral region was exposed. After removal of the sixth lumbar (L6) transverse process, the L5 and L6 spinal nerves were ligated with 6-0 silk suture and transected distal to the ligature. The fascia was closed with 4–0 resorbable polyglactin suture, and the skin was closed with staples. Control animals received anesthesia, skin incision, and stapling only. After surgery, the rats were returned to their cages and kept under normal housing conditions with access to pellet food and water ad libitum .

Rats underwent sensory testing for hyperalgesic behavior on three different days between 10 days and 17 days after surgery, as previously described.2,20Briefly, right plantar skin was mechanically stimulated with a 22-gauge spinal needle with adequate pressure to indent, but not penetrate, the skin. Whereas control animals respond with only a brief reflexive withdrawal, rats after SNL may display a complex hyperalgesia response that incorporates sustained licking, chewing, grooming, and elevation of the paw. The frequency of hyperalgesia responses was tabulated for each rat.

The L4 and L5 dorsal DRGs were excised after isoflurane anesthesia and decapitation 21 to 27 days after SNL or skin sham surgery, at which time hyperalgesia is fully developed.20DRGs were incubated in 0.0625% trypsin (Sigma Aldrich, St. Louis, MO), 0.0125% DNAse (Invitrogen, Carlsbad, CA) and 0.01% blendzyme 2 (Roche Diagnostics, Indianapolis, IN) in Dulbecco modified Eagle's medium/F12 with glutaMAX (Invitrogen) for 1.5 h, centrifuged, and triturated with fire-polished pipettes in culture medium containing Neural Basal Media A with B27 supplement (Invitrogen), 0.5 mm glutamine, 100 ng/ml nerve growth factor 7S (Alomone Labs, Jerusalem, Israel), and 0.02 mg/ml gentamicin (Invitrogen). Dissociated neurons were plated onto poly-l-lysine–coated glass cover slips (Deutsches Spiegelglas; Carolina, Burlington, NC) and maintained at 37°C in humidified 95% air and 5% CO²for 2 h and were studied no later than 6 h after harvest.

Unless otherwise specified, the bath contained Tyrode solution (in mm): NaCl 140, KCl 4, CaCl²2, Glucose 10, MgCl²2, and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) 10. Agents were obtained as follows: dantrolene and dimethylsulfoxide from Sigma Aldrich, BODIPY FL-X Thapsigargin (BODIPY-TG), ER-Tracker Blue-White DPX, Fura-2-AM, ionomycin, and Pluronic F-127 from Invitrogen. Stock solutions of ionomycin, dantrolene, and Fura-2-AM were dissolved in dimethylsulfoxide and subsequently diluted in the relevant bath solution such that final bath concentration of dimethylsulfoxide was 0.1% or less, which has no effect on [Ca²⁺]^c(n = 20, data not shown). The recording chamber was superfused by a Tyrode solution (3 ml/min), and agents were delivered by directed microperfusion through a 500-μm diameter hollow quartz fiber 300 μm upstream from the neurons.

Similar to our previously described technique,2,3coverslips carrying the plated neurons were dye-loaded in Tyrode solution containing Fura-2-AM (5 μm) and Pluronic F-127 (0.04%) for 30 min and left in regular Tyrode for deesterification for an additional 30 min, after which coverslips were mounted onto the recording chamber. Images were collected at 510 nm by using a cooled 12-bit digital camera (Coolsnap fx, Photometrics, Tucson, AZ), and excitation wavelengths were 340 and 380 nm. Recordings from each neuron were obtained as separate regions of interest (MetaFluor; Molecular Devices, Downingtown, PA). The [Ca²⁺]^cwas estimated by the formula [Ca²⁺]^c= K^d·β· (R–R^min)/(R^max–R) where β= (I^380max)/(I^380min) and R = (I³⁴⁰)/(I³⁸⁰). Values of R^min, R^max, and β were determined by in situ  calibrations and were 0.38, 8.49, and 9.54, whereas 224 nm was used as K^d.23 

Live, dissociated neurons were exposed to 1 μm ER-Tracker Blue-White DPX in Tyrode for 30 min. After washing, images were acquired by using an inverted microscope (Diaphot 200; Nikon Instruments, Melville, NY) with a 20× objective lens and a 12-bit digital camera (Coolsnap fx; Photometrics). Excitation was achieved with a 150-W Xenon lamp (Lambda DG-4; Sutter, Novato, CA) at 380 nm, and images were collected at 510 nm.

After 2 h in culture, dissociated neurons were fixed in buffer containing 4% formaldehyde, 4% sucrose, and 0.2% in phosphate-buffered saline for 5 min and were permeabilized in methanol at −20°C for 10 min. After three washes in phosphate-buffered saline with 0.2% glycine, neurons were incubated in gelatin diphosphate buffer, which contained 2% gelatin, 20% Triton X-100, 0.2 m phosphate buffer and 5 m NaCl in dH²O, with 8% normal goat serum for 1 h. The coverslips were then incubated with a polyclonal rabbit calreticulin antibody (Abcam, Cambridge, MA) at 1:100 and 4% normal goat serum in gelatin diphosphate buffer overnight at 4°C. After washing with phosphate-buffered saline, they were incubated with a Texas-red goat antirabbit secondary antibody (Jackson ImmunoResearch Laboratories, West Grove, PA) overnight at 4°C. After three rinses with phosphate-buffered saline and mounting on slides, they were imaged at 600× on an Eclipse confocal microscope (Nikon Instruments, Melville, NY). For intensity analysis, a representative region of the cytoplasm, excluding the nucleus and the plasmalemma, was selected in each neuron for intensity measurement (MetaMorph, Molecular Devices). Neurons that showed sign of lysis or crenulation or were overlapped by glia were excluded by brightfield examination.

Dissociated neurons were stained with BODIPY-TG while still in culture. Coverslips were incubated with 1 μm of the dye in Tyrode for 5 min. The medium was changed to dye-free Tyrode, and slips were immediately imaged on an inverted microscope (Diaphot 200, Nikon). Excitation was achieved with a 150-W Xenon lamp (Lambda DG-4; Sutter) at 495 nm by using a fluorescein isothiocyanate filter cube (Nikon), and images were acquired at 510 nm with a 12-bit digital camera (Coolsnap fx, Photometrics) and a 20× objective lens. Intensities were measured by using a procedure similar to the one described above.

DRGs were harvested from control (L5 DRG from both sides, 19 days after skin incision surgery) and SNL rats (L4 and L5 DRGs from two rats, 19 days and 26 days after SNL), fixed for 1 h in 2.5% glutaraldehyde in 0.1 m cacodylate buffer and postfixed with 1% OsO⁴in buffer for 1 h. Specimens were then washed with distilled water (2 × 5 min); dehydrated in methanol (50%, 70%, 85%, 95%, and 3× 100%) followed by 2 × 10-min acetonitrile; they were then infiltrated with Epon 812 resin (Shell Chemical, Houston, TX) in acetonitrile 1:1 for 1 h followed by 100% Epon for 3 h and moved to fresh 100% Epon and polymerized at 70°C overnight. Sections 70-nm-thick were cut and stained with saturated uranyl acetate in 50% ethanol and Reynold lead citrate. Images were generated from six cells in each DRG (three larger and three smaller than 1,000 μm²) by using a transmission electron microscope (JEM2100; Japanese Electron Optics Limited, Tokyo, Japan) with a high-resolution charge-coupled device digital camera (Ultrascan 1000; Gatan Inc., Pleasanton, CA). Each cell had two perinuclear and two peri-plasmalemmal fields (5.4 × 5.4 μm, area of 29.2 μm²) imaged at 4,000× magnification, chosen to be distributed in four separate quadrants of the neuron to provide an unbiased sample of cytoplasmic contents. In each of the 144 images, the length of each ER profile was measured in a blinded fashion by using an image analysis program (ImageJ; NIH, Bethesda, MD) to trace its running length. Only those profiles that had a typical tubular pattern with clear walls and a terminus indicating a closed compartment were included. The Golgi apparatus was easily distinguished by its characteristic curved series of flattened saccules that are expanded at their ends and were not included in measurements. Thereafter, the perinuclear data for each neuron were pooled, as were the peri-plasmalemmal data.

DRGs were harvested from control rats weighing 200–250 g during anesthesia with isoflurane (2–3% in oxygen). A laminectomy was performed up to the sixth thoracic level while the surgical field was perfused with artificial cerebrospinal fluid (in mm: NaCl 128, KCl 3.5, MgCl²1.2, CaCl²2.3, NaH²PO⁴1.2, NaHCO³24.0, glucose 11.0) bubbled with 5% CO²and 95% O²to maintain a pH of 7.35. The L4 and L5 DRGs and attached dorsal roots were removed, and the connective tissue capsule was dissected away from the ganglia under 20× magnification. Ganglia were transferred to a glass-bottomed recording chamber and perfused with 35°C artificial cerebrospinal fluid. The proximal cut end of dorsal roots was placed on a pair of platinum wire stimulating electrodes. DRG neurons were viewed by using an upright microscope equipped with differential interference contrast optics and infrared illumination. Neuronal soma diameter was determined with the focal plane adjusted to reveal the maximum somatic area, which was measured by using a calibrated video image.

Intracellular recordings were performed with microelectrodes fashioned from borosilicate glass (OD 1 mm, ID 0.5 mm, with Omega fiber; FHC Inc., Bowdoin, ME) using a programmable micropipette puller (P-97; Sutter Instrument Co., Novato, CA). Microelectrode resistances were 80–120 MΩ when filled with 2 m potassium acetate. DRG neurons were impaled under direct vision, and data were acquired after stable recordings were achieved. Membrane potential was recorded by using an active bridge amplifier (Axoclamp 2B; Axon Instruments, Foster City, CA), except in protocols in which depolarizing current was passed through the recording electrode, for which we used discontinuous current clamp recording mode, during which the switching frequency was 2 kHz, and full settling of the electrode charge was confirmed. Currents were filtered at 1 kHz (discontinuous mode) and 10 kHz (bridge mode), and then digitized at 10 kHz (discontinuous mode) or 40 kHz (bridge mode; Digidata 1322A and Axograph 4.9; Axon Instruments) for data acquisition and analysis. Somatic action potentials (APs) were elicited by natural conduction from the site of dorsal root stimulation with square-wave pulses of up to 90 mA lasting 0.06 ms. To determine the rheobase current threshold for initiating an AP and to identify the firing pattern during sustained membrane depolarization, current was injected directly into the neuronal soma through the recording electrode.

Criteria for inclusion of data were a resting membrane potential (RMP) negative to –50 mV, and an AP amplitude greater than 40 mV. Using techniques described previously,19AP measures, including RMP, conduction velocity (CV), AP amplitude and duration, and afterhyperpolarization amplitude, duration, and area were obtained from single traces after consistency of dimensions was confirmed by comparison to ten sequential APs. Input resistance was calculated from the shift of RMP generated by a 100-ms injection of a hyperpolarizing current through the recording electrode by using a current amplitude (0.2–0.5 nA) that failed to show any time-dependent rectification.24Rheobase was determined as the minimum current able to elicit an AP during incremental depolarizing current injection of 0.5–10 nA for 100 mS. The pattern of impulse generation during sustained depolarization was determined during incremental depolarizing current steps to twice rheobase, at which neurons either continued to produce single APs or fired repetitively. The influence of dantrolene upon AP firing pattern was determined at the depolarizing voltage that first produced a drug-induced difference in the number of APs generated, which is a sensitive indicator of drug effect upon excitability.

The following frequency is the fastest rate at which the neuronal soma may be excited by APs conducted along the axon, and it represents the degree of afferent signal filtering by the impedance mismatch where the distal and proximal axons meet the T-branch that leads to the soma.25Trains of 20 axonal stimuli were presented with 3-s intervals between trains, each train having progressively greater stimulation frequency. The maximal at which each stimulus in the train produced a full somatic AP was considered the somatic following frequency.

Classification of neurons24used CV, measured by dividing the distance between stimulation and recording sites by the conduction latency. Neurons with dorsal root CV less than 1.5 m/s were considered C-type, neurons with CV greater than 15 m/s were considered Aα/β-type, which are putative low-threshold receptor modality.26Neurons with CV greater than 1.5 m/s but CV less than 10 m/s were considered AΔ-type, which are putative nociceptor modality. For neurons with CV between 10 and 15 m/s, long AP duration was used to categorize the neurons as AΔ-types.24 

Dantrolene was dissolved in bath solution and was delivered by microperfusion from a pipette with a 4-μm-diameter tip positioned 100 μm from the impaled neuron. Preliminary experiments indicated an effective fourfold dilution of pipette solution into the bath at the neuronal membrane, so the dantrolene concentration was approximately 25 μm at the DRG surface. Preliminary experiments showed a full onset of action within 3 min.

Statistical analyses were performed with Statistica (StatSoft, Tulsa, OK). Two-tailed Student t  tests were used to compare means. One-way ANOVA with Bonferroni post hoc  testing was employed to detect the influence of injury group on measured parameters. Quantifications of features in the analysis of electron microscopy (EM) images were treated as nonparametric data; therefore, the Kruskal-Wallis test followed by multiple comparisons of mean ranks was used. Correlation between two observers was determined by linear regression. Paired, one-tailed t  tests were used in the analyses of the intracellular recordings because we only compared between injury and control conditions and a hypothesis for the direction of the changes already existed. Results are reported as mean ± SD. A P < 0.05 was considered significant.

Upon needle stimulation, all SNL animals (n = 14) displayed a hyperalgesia response rate that was greater than 25% and averaged 42 ± 19%, whereas control animals (n = 27) showed a hyperalgesic response 0 ± 0% (P < 0.001).

We first characterized the distribution of ER with the selective dye ER-Tracker Blue-White DPX, which has the property of localizing to the membrane of ER through an unexplained mechanism. Dissociated, live neurons stained with ER-Tracker (1 μm, 30 min, fig. 1A) showed distribution of the dye throughout the cytoplasm without nuclear staining without any differences between injured and control neurons.

To gain more consistent quantification of the ER, an immunocytochemical approach was used to target the Ca²⁺-binding protein calreticulin, which functions as the dominant binding protein regulating the mass of stored Ca²⁺. It is largely excluded from the cytoplasm, but it does appear in the cell membrane in addition to the ER,27so we avoided including this binding in our measure by quantifying signal only in cytoplasmic areas of confocal images. Staining with the calreticulin antibody (fig. 1B) was distributed comparably to ER-Tracker and demonstrated no quantitative differences between groups (fig. 1C).

The specificity of the SERCA blocker TG may be used for fluorescent quantification of binding sites by conjugation to the fluorescein-like dye BODIPY FL-X. Staining of live neurons (fig. 1, D and E) was distributed comparably to ER-Tracker and showed fluorescence intensities that were decreased by 34% in L4 neurons and by 40% in L5 neurons compared to control neurons (fig. 1F). This finding is compatible with either a lower density of SERCA in the ER membrane of injured neurons or alternatively less total ER.

A possible explanation for both diminished BODIPY-TG staining and loss of releasable Ca²⁺stores after neuronal injury4is loss of the structure that contains the stored Ca²⁺, which is predominantly ER.8,17Accordingly, we examined dimensions of the ER in EM images of injured and control neurons categorized into large (area 1438 ± 295 μm², n = 16) and small (area 495 ± 185 μm², n = 16) types. Injury in other neurons has been reported to lead to preferential ER displacement or loss in the perinuclear region28; we therefore examined perinuclear and peri-plasmalemmal areas separately. EM images displayed profiles that were readily identified as ER (fig. 2). There was a preponderance of smooth variety,28,29with only relatively small segments endowed with ribosomes, as has been described in normal DRG neurons before.10,30Quantification of linear dimensions (table 1) showed that the summated total length of ER segments is decreased in axotomized small neurons, with changes evident in both perinuclear and peri-plasmalemmal regions. In contrast, total ER increased in L4 neurons after SNL in both small and large neurons. This was attributable to increases in both cellular subregions in the large neurons, but selectively to an increase in the peri-plasmalemmal area in small neurons.

Initial examination of the EM images suggested that ER profiles in axotomized SNL L5 neurons were typically shorter, as has been reported.11,31Therefore, we also evaluated the effect of injury on the lengths of each segment of ER represented by a discontinuous profile in the different groups in a blinded fashion (table 1). Although median segment lengths were unaffected in large neurons, small neurons developed shorter ER profiles (fig. 2) after axotomy in both subcellular regions, and SNL L4 neurons showed longer profiles in the peri-plasmalemmal region.

The neuronal ER system has been shown to be organized into functionally and anatomically distinct elements, and cisternae that stack as plates may play a special role in Ca²⁺management.32,33We therefore quantified these specialized structures by grading each micrographic field as having no such cisternal stacks (score 0), minimal stack formation (one group with least four parallel ER profiles, score 1), clear stack formation with multiple groups (score 2), or extensive formation of stacks that fill at least one-quarter of the field (score 3). Scores assigned separately by two blinded observers correlated at R = 0.75 (P < 0.001). Although others have observed preferential development of these structures in the peri-plasmalemmal region,34we did not identify any preferential subcellular localization in control neurons, so data were combined for perinuclear and peri-plasmalemmal regions. Analysis of large and small neurons separately by Kruskall-Wallis ANOVA by ranks showed no effect of injury in large neurons (fig. 3A), but near total elimination of cisternal stacks in axotomized small neurons in the L5 DRG after SNL (figs. 2 and 3A), which was significantly different from both control and L4 neurons (multiple comparisons of mean ranks).

The assembly of ribosomes into rosette-like polyribosomes (polysomes) takes place upon the initiation of peptide assembly, and loss of polysomes is a marker of interrupted protein synthesis. As a comparative index of protein synthesis,35we characterized the relative abundance of polysomes in each EM field as having minimal or no polysomes (score 0), an approximately even presentation of ribosomes as free or assembled into rosettes (score 1), or predominantly polysomal rosettes (score 2). Scores assigned separately by two blinded observers correlated at R = 0.55 (P < 0.001). There was no influence of neuronal size or location within the cell. However, axotomy reduced the level of aggregation in both large (fig. 3B) and small neurons (figs. 2 and 3B) in the L5 group compared to both the control and L4 neurons (Kruskall-Wallis ANOVA by ranks, multiple comparisons of mean ranks).

A substantial portion of the transient rise in [Ca²⁺]^ctriggered by neuronal activation is the result of influx through voltage-gated Ca²⁺channels. However, additional Ca²⁺is released during neuronal depolarization through the activation of CICR. We quantified this component by blocking CICR with dantrolene,36which acts directly and selectively on the RyR to decrease its sensitivity of Ca²⁺and thereby inhibit channel activation.37We first confirmed blockade of RyRs in sensory neurons by showing that prior application of dantrolene (10 μm, 10 min) reduced caffeine-induced transient area to 6% (1016 ± 819 nm · s, n = 40) of control transients. Application of dantrolene (10 μm) for 10 min (fig. 4A) reduced K⁺-evoked transient area by 39% compared to a previous baseline transient (fig. 4B). In SNL L4 neurons, the area decreased by 26% only; in SNL L5 neurons, the effect of dantrolene was eliminated (−1%). Additional determination of transient amplitude confirmed a diminished inhibition by dantrolene in L5 neurons after SNL (−19 ± 15%, n = 21; P < 0.001 vs.  C and L4) compared to SNL L4 (−36 ± 11%, n = 12) and control neurons (−41 ± 9%, n = 16). Transient slope also showed diminished inhibition by dantrolene in L5 neurons after SNL (−17 ± 24%, n = 21; P < 0.05 vs.  C) compared control neurons (−37 ± 18%, n = 16), with no effect on SNL L4 (−34 ± 15%, n = 12). Together, these findings point to a complete elimination of CICR in axotomized neurons.

To determine the functional consequences of loss of CICR during neuronal activity, we investigated the effects of application of dantrolene during electrical activation of neurons in intact DRGs for neurons of different CV categories. Dantrolene, at an approximate concentration of 25 μm at the DRG surface, decreased afterhyperpolarization duration in both Aα/β and AΔ neuronal populations and decreased the afterhyperpolarization area in the AΔ neurons (fig. 5A, table 2). Because the late afterhyperpolarization is particularly dependent on Ca²⁺released by CICR,38,39we measured the afterhyperpolarization amplitude at 350 and 1,350 ms after the onset of the AP. In Aα/β neurons, the late afterhyperpolarization amplitude was not affected by dantrolene, whereas the late afterhyperpolarization was diminished in AΔ neurons. Dantrolene also increased the firing rate during membrane depolarization in both neuron types (fig. 5B, table 3), and accelerated the following frequency in AΔ neurons (fig. 5C). At baseline, 6/13 Aα/β and 5/13 AΔ neurons generated repetitive APs during sustained depolarization beyond rheobase. Separate consideration of these nonaccommodating neurons showed that administration of dantrolene shifted the relationship between firing rate and degree of depolarization in all of these neurons such that repetitive firing started at a lower depolarization voltage and/or the firing rate increased at each voltage. An additional Aα/β neuron that fired only singly at baseline also developed repetitive firing during dantrolene. At the depolarizing voltage that first produced a dantrolene-induced difference in the number of APs, the firing rate among nonaccommodating neurons was increased by dantrolene in both Aα/β neurons (from 2.7 ± 4.1 100 ms⁻¹at baseline to 4.9 ± 3.9 100 ms⁻¹during dantrolene, n = 7; P < 0.01) and AΔ neurons (from 1.4 ± 0.9 100 ms⁻¹at baseline to 3.2 ± 1.3 100 ms⁻¹during dantrolene, n = 5; P < 0.01).

Similar recordings were made from nonmyelinated C-type neurons that were recognized by their CV of less than 1.5 m/s. In this population, application of dantrolene at a concentration of approximately 25 μm significantly (P < 0.001) depolarized the RMP by 12.6 ± 9.5 mV, in contrast to the lack of effect on RMP in Aα/β (P < 0.001) and AΔ (P < 0.01) neurons. Full AP generation was lost concurrent with this depolarization in 7 of 9 neurons, which was a result of the effect of depolarization inactivating voltage-gated Na⁺channels, as repolarization by injection of conditioning current through the recording pipette resulted in return of the AP (fig. 6). This return of the AP with repolarization alone is consistent with a selective action of dantrolene on ryanodine receptors rather than voltage-gated conductances.40Because of the greater sensitivity of C-type neurons to dantrolene, we decreased the applied concentration to approximately 12 μm. In this group of experiments, depolarization was less (7.7 ± 8.1 mV). For further analysis, the influence of dantrolene on excitability was isolated from its effect on RMP by using conditioning current to repolarize RMP in 4 C-type neurons.

Dantrolene prolonged the AP duration in C-type neurons (table 3), possibly as a result of the loss of Ca²⁺-activated K⁺currents that contribute to repolarization.41Dantrolene also decreased afterhyperpolarization duration and area (table 2). The late afterhyperpolarization amplitude at baseline was greater in C-type neurons than in both the Aα/β and AΔ neurons, and the inhibition of the late afterhyperpolarization by dantrolene was greater than in the other neuronal groups, indicating a greater contribution of CICR in the C-type neurons. At baseline, only 1 of 12 C-type neurons generated repetitive APs during sustained depolarization beyond rheobase; during application of dantrolene, but 6 additional neurons fired repetitively during depolarization. At the depolarizing voltage that first produced a dantrolene-induced difference in the number of APs, the firing rate of these nonaccommodating neurons was increased by dantrolene (from 1.3 ± 0.8 100 ms⁻¹at baseline to 2.7 ± 1.1 100 ms⁻¹during dantrolene, n = 7; P < 0.001). Following frequency was not tested in C-type neurons because this category of neuron is capable of following only very slow and highly variable frequencies.

Although Ca²⁺signaling regulates nearly all critical neuronal functions, there has been only minimal exploration of the influence of painful nerve injury on Ca²⁺stores in sensory neurons. The investigations reported here extend our findings, reported in the accompanying paper,4of depleted Ca²⁺stores and depressed [Ca²⁺]^Lafter axotomy. In the current study, we have found that injury diminishes the anatomical extent of ER and reduces levels of SERCA expression, both of which may contribute to diminished stores. Our data also demonstrate substantial functional impairment attributable to this loss of Ca²⁺stores. Specifically, we have observed direct evidence of injury-induced loss of CICR that regulates neuronal excitability and indirect evidence of disruption of protein synthesis.

Calreticulin is the dominant buffer within the ER27,42and is distributed uniformly throughout neuronal ER.43Its role in regulating the size of the Ca²⁺pool is demonstrated by expansion of releasable Ca²⁺stores in cells overexpressing calreticulin.43Each calreticulin molecule possesses up to 50 binding sites for Ca²⁺, accommodating a large amount of Ca²⁺, and its low affinity permits the generation of high [Ca²⁺]^Lthat are necessary for rapid Ca²⁺release and facilitates longitudinal transport of Ca²⁺within the ER lumen.17We examined calreticulin levels by semiquantitative immunocytochemistry rather than Western blotting to avoid contamination by the numerous glial cells in the DRG. The lack of difference in injured neurons that we observed probably indicates that calreticulin is not the pathway through which injury influences Ca²⁺stores.

The most direct approach for measuring the anatomic extent of the ER is morphometric analysis at the ultrastructural level by EM. By this means, we have identified a reduction of total ER length in small neurons after axotomy, which may in part account for reduced ability to store Ca²⁺. A potential link between stored Ca²⁺mass and ER dimensions is supported by two observations. First, a deficit in both of these was found selectively only in small neurons. Second, the extent of these deficits is comparable, with 53% loss in total ER length per neuron and a loss of releasable Ca²⁺ranging from 25% to 59% for the various techniques used in the accompanying paper.4In addition to the total ER length per neuron, the length of individual profiles of ER segments is also diminished in axotomized small neurons. This may represent fragmentation of the ER tubules into disconnected segments. Alternatively, this appearance may result from increased tortuosity of the ER components, which then enter and leave the plane of the section, producing shorter ER profiles. Our data cannot resolve between these two possibilities, but fragmentation would restrict the coordinated function and signal distribution role of the ER.18The finding of undiminished total cellular calreticulin is consistent with an unchanged ER volume, or the calreticulin is concentrated in a smaller total ER volume.

Although the most prominent findings are in the axotomized neurons of the L5 DRG after SNL, certain distinct features appear in the adjacent L4 neurons. Total ER length is increased in both large and small L4 neurons, in contrast to the decrease seen in small L5 neurons. This may explain our finding in the accompanying paper of a greater capacity for expanding the Ca²⁺stores by neuronal activation in L4 neurons than in control neurons, whereas the opposite occurs in L5 neurons.4 

A striking contrast is evident between axotomized and control small neurons in the organization of the ER elements. After axotomy, there is almost complete elimination of the formation of ER into stacks of cisternae. Calcium mapping by electron energy loss spectroscopy has identified these cisternae as regions in the ER system with particularly high levels of Ca²⁺.44Thus, selective elimination after injury could disproportionately depress levels of releasable Ca²⁺.

In addition to dissolution of ER elements, our findings show depression of BODIPY-TG binding after axotomy. This agent identifies SERCA and indicates a decrease in the cellular density of SERCA pumps that generate Ca²⁺influx into the ER from the cytoplasm. This finding supports our conclusion in the accompanying paper that compromised SERCA function contributes to depressed [Ca²⁺]^Lafter axotomy.4However, the concordance of the extent of loss of BODIPY-TG binding sites (40% in axotomized neurons compared to control) and the loss of ER (50%) suggests that the loss of SERCA may be simply the result of net loss of ER membrane. This proportionality would predict that the density of SERCA in the ER membrane is only minimally changed by injury, in which case depressed [Ca²⁺]^Lafter axotomy cannot be attributed to loss of SERCA. Our data are not adequate to resolve this mechanistic difference, and we also note that BODIPY-TG staining decreases in L4 neurons, whereas [Ca²⁺]^Ldoes not.4Although we cannot therefore explain the decrease in [Ca²⁺]^Lafter axotomy, our data point to this feature plus the loss of anatomic ER as combined causes of decreased Ca²⁺stores after axotomy.

Our study also examined key functional consequences that might follow from a loss of Ca²⁺stores in neurons. Although published evidence is not conclusive for all neuronal types, there is a consensus that activation of peripheral sensory neurons is accompanied by CICR.17It is expected that injury-induced loss of stores should compromise the ability of CICR to amplify the Ca²⁺signal initiated by neuronal activation. Our findings confirm that this is the case. The amount of CICR in dissociated neurons was gauged by sensitivity to blockade with the selective RyR blocker dantrolene, which revealed elimination of CICR after axotomy. This complete loss of functional CICR despite an only partial loss of Ca²⁺stores and ER total length per neuron may indicate a disproportionate dependence of CICR on Ca²⁺residing in the cisternal stack pool, which is also nearly eliminated by injury. The close stacking of adjacent ER components in cisternal stacks is thought to provide cooperative interactions that trigger regenerative Ca²⁺release,42which may be particularly important for CICR.

CICR amplification of the Ca²⁺signal during firing of APs contributes to triggering currents that underlie the afterhyperpolarization45,46; we therefore suspected that the loss of CICR in injured neurons might contribute to their hyperactivity, as has been shown in other tissues.39,47We modeled this in intact DRGs by blocking CICR in control neurons with dantrolene, and we observed decreased afterhyperpolarization and increased neuronal excitability. The greater sensitivity of C-type presumed nociceptive neurons is likely a result of their greater expression of the low-conductance SK isoform of Ca²⁺-activated K⁺channels,48which are particularly dependent on Ca²⁺derived from CICR for activation.38,49The dominant component of cytoplasmic Ca²⁺transient that accompanies an AP is derived from entry through voltage-gated Ca²⁺channels, as is demonstrated by the persistence of the majority of the transient after RyR blockade with dantrolene. However, the loss of the CICR component clearly increases neuronal excitability during dantrolene application; it is likely that a similar loss of CICR in axotomized L5 neurons, particularly in nociceptors, may lead to excess neuronal excitability after injury as well. By this means, loss of Ca²⁺stores in sensory neurons may contribute to the generation of pain after peripheral nerve injury. This speculation is supported by the ability of dantrolene to replicate injury effects,19specifically reduction of afterhyperpolarization dimensions and increased repetitive firing.

Apart from providing a source for Ca²⁺release, it is now recognized that maintenance of normal Ca²⁺levels in the ER lumen is required for normal synthesis and maturation of proteins, including the processes of translational initiation, glycosylation, and protein folding.50It is particularly the RyR-sensative Ca²⁺pool that regulates these functions in neurons.51Initiation of translation and peptide assembly is associated with aggregation of ribosomes into clusters (rosettes), whereas disaggregation is a hallmark of impaired translational initiation and arrest of protein synthesis.52Others have also observed ribosomal disaggregation as an early response to axotomy in other neuronal tissues,10,11,53and we now report the phenomenon after spinal nerve axotomy. Various proteins show decreased expression after peripheral sensory neuron axotomy.54,55Substantial selective modulation of gene expression after injury occurs at the transcription level56–58; therefore, it is likely that additional translational and posttranslational influences result from depletion of Ca²⁺in the ER.

Our data do not offer direct insights into mechanisms that may produce the ultrastructural findings we have reported. However, these changes show morphological similarities with the process of apoptosis, which occurs in adult DRG neurons within 1 week of axotomy, peaks in 2 months, and continues as late as 6 months.59,60Comparable to our findings in DRG neurons, apoptosis in cultured sympathetic neurons diminishes the extent of ER.61In forebrain neurons, various apoptotic models, as well as natural apoptosis during brain development, produce fragmentation of the ER membrane and disaggregation of polysomes.62The cause for apoptosis after axotomy is not known, but several possibilities are suggested by previously published findings. Although dependence of DRG neurons on neurotrophic factors for survival diminishes with postnatal age, axotomy-induced apoptosis is inhibited by glial-derived neurotrophic factor,63and glial-derived neurotrophic factor reverses sensory changes.64Alternatively, activation of glia and production of cytokines after nerve injury may initiate DRG neuron apoptosis,65and depletion of Ca²⁺stores may lead directly to apoptosis in neurons.66The contribution of apoptosis to generation of chronic pain after peripheral nerve injury is shown by prevention of delayed pain through suppression of DRG neuron apoptosis with the hematopoietic cytokine erythropoietin.67 

Prevention or correction of ER Ca²⁺depletion may be a novel target for treatment for chronic pain after nerve injury. For instance, various proteins such as Bcl-2, grp78, and grp94 prevent ER Ca²⁺depletion and subsequent cytotoxicity,68–70and they upregulate SERCA function.71Molecular techniques may become available to overexpress these proteins in DRG neurons as a new therapy for chronic neuropathic pain.

The authors thank Clive Wells, M.I.Biol., C.Biol., Electron Microscopy Program Manager, Department of Cell Biology, Medical College of Wisconsin, Milwaukee, Wisconsin, for his expert assistance.