Bupivacaine Preferentially Blocks Ventral Root Axons in Rats

After we obtained approval of the protocol by the animal care committee, and as previously described, [20]we anesthetized 46 adult male Sprague-Dawley rats weighing 180–320 g with enflurane (1–2.5%) and nitrous oxide (70%) in oxygen. After mask induction and tracheotomy, anesthesia was delivered through a short polyethylene tracheal tube. End-tidal carbon dioxide and respiratory rate were measured continuously to monitor the depth of anesthesia. Fluid loss was replaced with lactated Ringer's solution at 4 ml [centered dot] kg sup -1 [centered dot] h sup -1. A thoracolumbar laminectomy extending from the T¹³thoracic vertebra to the L⁵lumbar vertebra followed by a long dural incision was made using microsurgical techniques. Single lumbar DR and VRs were cut proximally near their point of entry into the spinal cord and distally near their exit from the spinal canal and transferred to a Teflon perfusion/recording chamber. The roots were continuously superfused (at 4 ml/min) with artificial cerebrospinal fluid (150 mM Sodium sup +; 4 mM Potassium sup +; 127 mM Chlorine sup -; 2 mM Calcium sup ++; 1.3 mM Magnesium sup ++; 1.2 mM PO⁴sup -3; 26 mM HCO³sup -; 11 mM glucose) at 37 +/- 0.5 degrees Celsius and equilibrated with a 95% oxygen-5% carbon dioxide gas mixture to maintain a pH of 7.3–7.4.

The proximal end of the root (2–3 mm) was led into the recording chamber (Figure 1) through a separation in the partition, which was then sealed with silicone grease, and the artificial cerebrospinal fluid in that compartment was covered with mineral oil. The distal end of the root was attached to a silver wire suction electrode for stimulation. Single-fiber microdissection and recording techniques were used to isolate activity in individual spinal root axons.

Using a Grass S 48 stimulator (Grass Instruments, Quincy, MA), the distal end of the isolated root was constantly stimulated at 20% over threshold voltage at a frequency of 0.3 Hz throughout the experiment. Single-fiber action potentials were amplified and displayed on a digital storage oscilloscope (Tektronix 2221, Beavertown, OR) and recorded with a computer data-acquisition system (R. C. Electronics, Goleta, CA). Stimulus-evoked activity in individual axons was monitored, and the latency between stimulus and action potential was measured for 20–30 min (control period) to ensure stability of the preparation and to provide for elimination of the volatile anesthetics. Conduction velocity was calculated for each axon to measure conduction time (latency) and length of axon between stimulating and recording electrodes (CV [m/s]= conduction distance [mm] divided by the latency [ms]).

After control measurements, each root was exposed in a stepwise manner to increasing concentrations of bupivacaine (Astra Pharmaceuticals, Westborough, MA) using a calibrated infusion pump. By this means, the bupivacaine concentration in the perfusion chamber could be adjusted between 4 micro Meter and 88 micro Meter. To measure the minimum blocking concentration (C^m), bupivacaine was increased in steps of 2–6 micro Meter, allowing an equilibration time of 5–10 min to ensure steady-state conditions before increasing to the next step. The C^mwas the lowest concentration that resulted in conduction failure (no stimulus-evoked activity at 2x threshold voltage for 1 min). Conduction latency was measured at 1-min intervals until the unit failed. For each axon, the latency EC⁵⁰for bupivacaine was calculated from the slope of the line relating bupivacaine concentration to conduction latency at the point representing a 50% latency increase. Each spinal root was exposed to a sequence of increasing bupivacaine concentrations only once (Figure 2). Data were accepted only from those axons showing complete recovery (latency and amplitude within 5% of control values) after return to control artificial cerebrospinal fluid.

Dorsal root axons were divided into three groups based on their CV characteristics. Those DR axons with a CV greater than 3 m/s were presumed to be myelinated (A fiber), whereas those with a CV less than 1.5 m/s were presumed to be unmyelinated (C fiber). The group of DR axons with intermediate CVs (1.5–3 m/s) probably consisted of both large unmyelinated and small myelinated fibers (I fiber). Ventral root axons all had CVs greater than 5 m/s, comparable to the DR A-fiber group.

Analysis of the data using the Kolmogorov-Smirnov test revealed nongaussian distributions of all examined parameters. Thus all data were summarized as median and interquartile range (25–75%). Differences between groups were assessed using nonparametric tests. The Kruskal-Wallis test was used for multiple group comparisons, and the Mann-Whitney Rank test was used for two-group comparisons. Differences were considered significant at P < 0.05. Correlation analysis was performed using the Spearman rank test.

We studied 116 individual mammalian spinal root axons. Control CVs for 91 unmyelinated and myelinated mammalian DR axons ranged from 0.56 m/s to 30.55 m/s, whereas the CV for 25 VR axons ranged from 5.36 m/s to 27.94 m/s. Susceptibility to bupivacaine-induced conduction block was measured at stepwise increased local anesthetic concentrations, starting as low as 4 micro Meter. After exposure to each bupivacaine concentration, the conduction latency increased over the next 3–8 min until it stabilized (Figure 3). Incremental latency change was usually greatest just before conduction block.

The median minimum bupivacaine blocking concentration (C^m) for all DR axons (32.4 micro Meter) was 2.5 times greater than the median C^mfor all VR axons (13.8 micro Meter)(Figure 4). This difference was statistically significant (P < 0.0001). The interquartile ranges (25–75%) for DR axon C^m(26.2–43.2 micro Meter) and for VR axon C^m(7.1–25.6 micro Meter) did not overlap. Thus, at bupivacaine concentrations in which 75% of all VR fibers were blocked, fewer than 25% of all DR fibers were blocked. The distribution of C^ms for DR fibers was homogenous and ranged from 13.8 to 88.4 micro Meter, whereas the C^mrange for VR fibers (4.1–51 micro Meter) was distributed bimodally. Eighteen VR axons were blocked at bupivacaine concentrations less than 16 micro Meter, and six axons of the remaining seven VR axons required bupivacaine concentrations greater than 30.8 micro Meter.

Dorsal root axons were categorized into three groups depending on their CV. Axons with CVs less than 1.5 m/s were considered C fibers, axons with CVs greater than 3 m/s were considered A fibers, whereas those with CVs between 1.5 m/s and 3 m/s were designated intermediate fibers (I fibers). These three DR axon groups had similar bupivacaine C^ms (Figure 5). The median C^mwas 32.4 micro Meter (interquartile range, 29.5–37 micro Meter) for 17 C fibers, 32.4 micro Meter (23–40.7 micro Meter) for 15 I fibers, and 36.2 micro Meter (26.2–53 micro Meter) for 59 A fibers. Compared with VR axons, each DR axon group was significantly less sensitive to the conduction-blocking effects of bupivacaine (P < 0.0001 for VR fibers vs. C, I, or A fibers). Among the DR axons, C fibers showed the most homogenous distribution of C^ms, with the smallest range between the 10th and 90th percentiles (26.2–39.3 micro Meter). Although DR axons in the A-fiber group appeared to be less sensitive to bupivacaine, there was no significant difference in C^mamong the three DR axon groups. Within the A-fiber group there was a weak positive correlation (r^s= 0.36) between axon diameter (as indicated by control CV) and resistance to bupivacaine-induced conduction block (P < 0.0001 for the regression line). However, there were no significant correlations within the I-, C-, or VR-fiber groups. Spearman correlation coefficients (r^s) were -0.28 for C fibers, -0.09 for I fibers, and 0.18 for VR fibers.

The lengths of DR axons exposed to bupivacaine ranged from 15 mm to 23 mm, and the anesthetic-exposed lengths of VR axons ranged from 15 mm to 20 mm. As shown in Figure 6, there was no significant correlation between exposed length and the bupivacaine C^m, within the three DR axon groups (panels A-C), whereas the regression analysis for VR axons (panel D) revealed a positive correlation (r^s= 0.52; P < 0.01), with longer axons requiring higher concentrations of bupivacaine to block conduction.

The CV-slowing (latency increasing) effect is another measure of axon sensitivity to local anesthetic exposure, because CV decreases as the number of unblocked sodium channels decrease. [30,31]Compared with DR axons, VR axons were significantly more sensitive to this effect of bupivacaine (Figure 7), with a latency EC⁵⁰of 21.8 micro Meter in DR A fibers and 8.5 micro Meter in VR fibers. This difference was statistically significant (P < 0.0001). Latency EC⁵⁰s in the DR C-fiber (19.1 micro Meter) and I-fiber groups (18.5 micro Meter) was also significantly higher compared with VR axons (P < 0.002). Among DR axons, the latency EC⁵⁰for myelinated fibers was significantly higher than for unmyelinated fibers. Conduction velocities corresponding to conduction latencies at the time of bupivacaine-induced conduction block for all axons ranged from 37% to 93% of control CVs.

This study shows that in rats, VR axons are more sensitive than DR axons to the conduction-blocking effects of bupivacaine. Conduction block and decreased CV in VR axons occurred at significantly lower bupivacaine concentrations than in DR axons. This laboratory observation must be reconciled with clinical findings in which spinal and epidural bupivacaine in low concentrations appear to preferentially and persistently block small-diameter sensory nerves. [1,32]Extrinsic factors present in the clinical setting perhaps may mask intrinsic axonal differences in local anesthetic sensitivity. For example, Fink [33]hypothesized that differences in vascularity between DRs and VRs may contribute to the differential effects of epidural anesthetics. To test this hypothesis, he examined human spinal roots in a morphometric study, but he could not demonstrate any significant vascular differences. In contrast, laboratory studies of the intrinsic electrophysiologic properties of motor and sensory axons have revealed differences attributed to different Potassium sup + and Sodium sup + channel kinetics. Sensory and motor axons were found to differ in the kinetics of Potassium sup + inactivation and in individual Potassium sup + channel conductance. [34]These data suggest the existence of different types of Potassium sup + channel in sensory and motor nerve fibers. Other studies have showed that action potentials in motor axons last longer than do those in sensory axons, suggesting intrinsic differences in Sodium sup + channels. [35]More recently, these differences in Sodium sup + channel kinetics were confirmed in rat motor and sensory nerve fibers using a patch-clamp technique. [25]The unique presence of slowly inactivating Sodium sup + channels in rat motor axons may predispose these fibers to block by local anesthetics, which preferentially bind to Sodium sup + channels in the open activated state. [26]However, because these intrinsic electrophysiologic differences would exist in both clinical and laboratory settings, they cannot account for the clinically apparent resistance of motor nerves to bupivacaine.

Bupivacaine's unique relative specificity for sensory nerves has been shown in various clinical settings, including obstetrics, general surgery, and pain therapy, [36]whereas the mechanism of this differential sensory-motor block remains unknown. Clinically, low bupivacaine concentrations (0.0625–0.125%, equivalent to 2,000–4,000 micro Meter) produce analgesia with minimal motor impairment. This differential effect is preserved at higher concentrations, although it appears to diminish in magnitude over time. [6]McCrae and associates, [32]using 0.25% and 0.5% bupivacaine given by intermittent epidural injection to women during labor, recently confirmed the wide spread of sensory block compared with motor block. These observations of differential block correlated with previous results from Smedstad and Morison [37]using both intermittent and continuous infusion techniques with 0.25% bupivacaine. Contrary to clinical expectations, the results of this study show that bupivacaine directly applied to isolated rat spinal root nerves is significantly more effective at slowing and blocking conduction in VR (motor) axons than at slowing or blocking conduction in DR (sensory) axons. Although our observations may represent a species-specific effect of bupivacaine, we believe that is the least likely explanation.

In addition to differential sensory-motor block, clinical and laboratory studies have consistently reported that the various sensory modalities (such as light touch, temperature discrimination, and pain) are not blocked by local anesthetics at the same time or to the same dermatomal level during spinal and epidural anesthesia. It is well established that bupivacaine in particular can produce this differential sensory nerve block. [38]White and colleagues [5]were the first to measure the magnitude of differential sensory block under steady-state conditions during continuous epidural anesthesia with bupivacaine. They found that the limit of cold temperature sensation (T²-3) was in between the sensory levels for pinprick (C⁸) and dense anesthesia (T sub 7). Interestingly, previous investigators found this loss of temperature discrimination to be the most cephalad anesthetic effect. [2,3,39]After rejecting their original hypothesis that this discrepancy was due to characteristics of the local anesthetic itself (e.g., the negative logarithm of the acid ionization constant and lipid solubility), White and colleagues [5]speculated that it may be the result of the different methods used (spinal vs. epidural, addition of epinephrine, and the technique for evaluating cold sensation, for instance).

Laboratory studies of differential nerve block with bupivacaine were all conducted on peripheral nerves and typically involved measuring effects on the compound action potential amplitude. [9–11,40]Gissen and associates [9]could not show any differential anesthetic effects on sensory and motor fibers using bupivacaine (or etidocaine) in a desheathed rabbit vagus nerve preparation. In a subsequent study, they described a differential effect on the onset of conduction block when the perineural sheath was left intact. [10]In that study, bupivacaine, compared with etidocaine, had a slowly developing effect on fast-conducting A fibers. They suggested that this temporal difference, which is consistent with clinical observations, is the result of the lower lipid solubility and greater ionization of bupivacaine, which reduced diffusion across the permeability barriers present in fast fibers. The following year, Rosenberg and Heinonen [11]reported that bupivacaine caused a more rapid sensory block (greatest depression of amplitude in A delta- and C-fiber elevations of the compound action potential), and etidocaine caused more profound motor blockade (greatest depression of amplitude in A beta-fiber elevation) in rat vagus nerves with intact sheaths. In these studies, it is important to note that conduction block was assessed indirectly by measuring the reduction in the amplitude of the compound action potential and that this technique may overestimate anesthetic potency. Sub-blocking concentrations of local anesthetic may differentially slow CV and thereby disperse the individual action potentials that combine to create the compound action potential. Thus the amplitude of the compound action potential may be reduced without blocking conduction in any individual axons.

More recent studies on single nerve fibers in rabbit vagus nerve by Fink and Cairns [14,15,21]showed differential nerve block among different fiber types. Myelinated axons compared with unmyelinated axons had a greater sensitivity to lidocaine, measured by conduction block and by slowing of CV. However, vagus nerve axons of intermediate CV (1.2–4 m/s) were the most sensitive to these effects of lidocaine. No relation between fiber size (CV) and local anesthetic blocking concentration, latency increase (CV slowing), or conduction safety was evident within their unmyelinated, myelinated, or intermediate fiber groups. [12,16]Recently the first local anesthetic study using single spinal root axons was reported. [20]In that study the differential susceptibility of DR axon groups to lidocaine was measured, and myelinated axons (CV > 3 m/s) were more sensitive to the CV-slowing effects of lidocaine than were unmyelinated axons (CV < 1.4 m/s). Although these results are consistent with previous single-fiber studies in peripheral nerves, they do not correspond with the clinical observations of differential block. Interestingly, DR axons in the intermediate CV group were not the most sensitive to the CV-slowing effect of lidocaine, and there was no significant difference in the incidence of conduction block among the three DR axon groups.

The effects of bupivacaine at concentrations between 25 and 8,000 micro Meter have been described in previous studies based on peripheral nerve preparations. [9–11,40]Compared with these studies, every category of spinal root axon in the present study was more susceptible to the conduction blocking effects of bupivacaine than the corresponding category of axon in peripheral nerves. For example, Gissen and associates [9]calculated bupivacaine blocking concentrations (ED sub 50) for compound action potentials in rabbit vagus or sciatic nerves maintained in vitro. In that study, A fibers (30–60 m/s) were blocked at 48 +/- 23 micro Meter, B fibers (5–15 m/s) at 134 +/- 9 micro Meter, and C fibers (< 1 m/s) at 201 +/- 41 micro Meter, all substantially higher bupivacaine blocking concentrations than those we measured in the present study. In the only other study to measure local anesthetic susceptibility in spinal root axons, [20]the authors reported that rat DR axons were more sensitive to the conduction-blocking effects of lidocaine compared with axons in the vagus nerve. These findings correspond with the present study and suggest the existence of a fundamental pharmacologic difference between peripheral nerve axons and spinal root axons.

A second measure of local anesthetic sensitivity, latency EC sub 50 or the CV-slowing effect, also revealed significant differences between DR and VR axons, and among the three DR axon groups. Ventral root axons required significantly less bupivacaine than did DR axons to produce the same magnitude of CV slowing, indicating that VR axons are more sensitive to this local anesthetic effect. Among DR axons, C fibers were significantly more sensitive than A fibers. These results are consistent with a previous study [20]that showed that myelinated DR axons are more sensitive to the CV-slowing effect of lidocaine than are unmyelinated DR axons. Fink and Cairns [15]showed a difference in the CV-slowing effects of lidocaine on rabbit vagus nerve axons and reported that CV slowed more in myelinated axons before block than in unmyelinated axons. In both the present study and previous studies, within each fiber group local anesthetic-induced CV changes did not correlate with axon diameter (CV).

In contrast to studies using peripheral nerve axons, [22,23]increasing the length of DR or VR axons exposed to local anesthetic did not decrease the minimum concentration of bupivacaine required for conduction block. Within our limited range (15–23 mm), there was no significant correlation between exposed length and C^m. These results suggest that, at least for rat spinal root axons, local anesthetic exposure lengths greater than 15 mm exceed the minimum distance necessary for length-dependent conduction block and the effect of decremental conduction. These results do not preclude exposure length as an important factor in the clinical appearance of differential block. In the present study, only long lumbar spinal roots were tested, and differential block may only become apparent at thoracic and cervical dermatomes where the corresponding exposed nerve lengths in rats will be substantially less than 15 mm.

If we accept that the clinical observations on the onset and duration of specific sensory and motor blocks accurately reflect true differences in nerve fiber sensitivity at steady state, then, based on the results of the present study, the effect sites of epidural and spinal local anesthetics cannot be at the spinal root. The most probable alternative sites for local anesthetic effects are at the proximal end of the DR, or within the superficial layer of the dorsal horn where transmission at synaptic and interneuronal levels may be affected. The proximal portion of the dorsal root axon may be uniquely susceptible to local anesthetic effects. Each DR axon tapers to a smaller diameter before entering the dorsal horn, [17,40]and a transition in perineural cell population from Schwann cells to oligodendrocytes occurs. This anatomic transition zone may represent an area of decreased conduction safety and, therefore, a primary target for local anesthetic effects. Because sensory fibers become segregated by diameter in this zone, there is the potential to selectively affect sensory input from small-diameter afferents. For example, the medial segregation of unmyelinated axons away from the large myelinated axons may reduce the diffusion barriers for anesthetic access to these fibers. The superficial layer of the dorsal horn may represent another zone of decreased conduction safety for those axons that branch in that layer (e.g., small-diameter afferents), because branch points are areas of reduced conduction safety. There are no comparable areas of vulnerability for motor axons at the VR exit zone. Thus we can reconcile the difference between clinical and laboratory observations by hypothesizing that the primary site of local anesthetic effect on sensory axons is at the dorsal root entry zone or superficial layer of the dorsal horn and not along the principle portion of the dorsal root. Additional studies are needed to test this hypothesis.