Selective Postsynaptic Inhibition of Tonic-firing Neurons in Substantia Gelatinosa by μ-Opioid Agonist

Tight-seal recordings were done using 200- or 300-μm coronal slices prepared from the lumbar enlargement of the spinal cord of 2- to 7-week-old Wistar rats.28The animals were killed in accordance with the national guidelines (Direcção Geral de Veterinária, Ministério da Agricultura, Lisboa, Portugal). After anesthesia by intraperitoneal injection of Na⁺pentobarbital (30 mg/kg), the vertebral column was quickly cut out and immersed in ice-cold oxygenated artificial cerebrospinal fluid. The segment of the lumbar enlargement was dissected and glued to the stage of the tissue slicer. Slices were prepared and incubated for 40–60 min in oxygenated artificial cerebrospinal fluid at 33°C. For recording, the slices were transferred into a 0.7-ml chamber and continuously perfused at a rate of 8 ml/min. All recordings were done at 22°–24°C. SG (lamina II) was identified as a translucent band in the dorsal horn. Each neuron was localized during recording according to the position of the pipette tip on the video image of SG.

Artificial cerebrospinal fluid contained 115 mm NaCl, 5.6 mm KCl, 2 mm CaCl², 1 mm MgCl², 11 mm glucose, 1 mm NaH²PO⁴, and 25 mm NaHCO³(pH 7.4 when bubbled with a 95%–5% mixture of oxygen–carbon dioxide). Standard pipette solution contained 6 mm NaCl, 128 mm KCl, 2 mm MgCl², 10 mm EGTA, and 10 mm HEPES. The pH value was adjusted to 7.3 with KOH (final [K⁺] was 160.5 mm). In 36 experiments, guanosine 5′-triphosphate (100 μm) and adenosine 5′-triphosphate (2 mm) were added to the pipette solution. All chemicals were purchased from Sigma-Aldrich (Sintra, Portugal). DAMGO, baclofen, and naloxone (antagonist of μ-opioid receptors) were dissolved in distilled water and stored in aliquots of 4, 1.5, and 0.125 mm, respectively, at −20°C. Fresh dilutions were made with artificial cerebrospinal fluid just before the experiment. Adenosine 5′-triphosphate and guanosine 5′-triphosphate were dissolved in internal solution at final concentrations and kept in aliquots at −20°C until the experiment.

The patch pipettes had a resistance of 3–5 MΩ after fire polishing. An EPC-9 amplifier (HEKA, Lambrecht, Germany) was used in all experiments. The effective corner frequency of the low-pass filter was 3 kHz, and traces were digitized at 10 kHz. For measurements of transient K⁺(K^A) currents, a standard P/n protocol was used for transients and leakage subtraction. Offset and liquid junction potentials were corrected for in all experiments. In neurons subjected to detailed analysis, the series resistance was 6–20 MΩ and was compensated by 60%. Fast current clamp mode of the EPC-9 amplifier was used for voltage recording. Input resistance (R^IN) of SG neurons was measured in current clamp mode using negative 500-ms current pulses of 5–10 pA.

Special precautions were taken to correctly measure the resting potential in SG neurons. In current clamp mode, most commercially available patch clamp amplifiers inject into the cell a small uncompensated current of 5–20 pA, which can vary from day to day and cannot be compensated by users in a simple way (our personal observations with several EPC7, EPC9, and Axopatch 200B [Axon Instruments, Union City, CA] amplifiers). Such a current can depolarize by tens of millivolts the cells with R^INin GΩ range. Therefore, before each experiment, we measured in current clamp mode a voltage decrease on a 0.5-GΩ resistor and determined the current needed to bring it to an expected 0 mV. In all following current clamp recordings, this sustained current was applied to the neuron and was considered zero current level. Under these conditions, the resting potential measured in current clamp mode was equal to the potential at which zero absolute current was recorded in voltage clamp mode. This correction may explain, at least in part, more negative resting potential values obtained here than in other patch clamp studies of dorsal horn neurons.27,29 

All numbers in the text and figures are given as mean ± SEM; numbers in table 1are given as mean ± SD. The parameters were compared using a paired or independent Student t  test. The current study is based on recordings from 149 SG neurons.

Based on several criteria, SG neurons (n = 146) were separated into three groups (fig. 1): tonic-firing neurons (TFNs), adapting-firing neurons (AFNs), and delayed-firing neurons (DFNs). Three additional neurons could not be clearly classified and therefore were not subjected to study.

Tonic-firing neurons (n = 53) were able to support firing during 500-ms depolarization induced by a sustained current injection (fig. 1, A1 ). They had a low firing threshold (−51.2 ± 1.7 mV, n = 11), and 10- to 20-pA current pulses were sufficient to evoke firing (fig. 1, A2 ). Under the voltage clamp condition, the current–voltage relations in all TFNs showed pronounced inward rectification (fig. 1, A3 ). A characteristic feature of AFNs (n = 46) was a burst-like firing of only two to six spikes at the beginning of depolarization (fig. 1, B1 ). At any stimulus intensity, AFN discharges could not continue during the whole pulse. The firing threshold of −50.7 ± 1.4 mV (n = 10) was not significantly different from that in TFNs (P > 0.75, independent Student t  test). The membrane properties of AFNs (fig. 1, B2  and B3 ) were fairly similar to those of TFNs. In both TFNs and AFNs, a K^Acurrent was not seen at voltage steps from −80 to −60 mV (fig. 1, A3  and B3 , insets ).

A principal difference of DFNs (n = 47) was a presence of large K^Acurrent (137.9 ± 10.5 pA at voltage step from −80 to −60 mV, n = 10; P < 0.001, independent Student t  test, for comparison with either TFN or AFN group; fig. 1, C3 , inset ), which substantially influenced the firing pattern (fig. 1, C1  and C2 ). The spike threshold in DFNs (−36.4 ± 1.6 mV, n = 10) was considerably higher than in TFNs (P < 0.001, independent Student t  test) or AFNs (P < 0.001, independent Student t  test) and was reached at stimulation as strong as 50–70 pA because a large portion of injected current was compensated by activating K^Acurrent. In DFNs, the first spikes typically occurred with a considerable time delay at the end of the pulse and moved to its beginning as the stimulation increased. R^INin DFNs was approximately half of that in the other cell types (P < 0.001 for TFNs and P < 0.001 for AFNs, independent Student t  tests), but a more negative resting potential (table 1; P < 0.05 for either TFN or AFN group, independent Student t  tests) was closer to K⁺equilibrium potential of −84 mV, indicating a presence of larger resting K⁺conductance. An inward rectification was less pronounced in DFNs (fig. 1, C3 ). However, there were some minor variations in discharge patterns of DFNs at strong stimulation. Some cells discharged regularly during the whole pulse (fig. 1, C1 ), whereas others belonging to this group showed interrupted bursts of even single spikes (not shown).

In all three types of neurons, membrane response to hyperpolarizing voltages consisted of dominating fast inward-rectifier current. A slow hyperpolarization-activated I^H-like current carried by both K⁺and Na⁺ions24,30,31was negligible in majority of SG neurons studied here using both normal and adenosine 5′-triphosphate/guanosine 5′-triphosphate–containing (n = 36) pipette solutions.

Each group of neurons had its specific distribution pattern within SG (figs. 2A and B). TFNs were relatively homogeneously distributed over the whole length of SG. Few AFNs were found in medial and intermediate region, but most of them were localized in a lateral SG, especially its ventral part. Although DFNs could be found in all SG regions, their density in the medial and lateral parts was higher. The sensitivity to DAMGO was tested for the neurons of each type located in different SG regions (indicated by crosses).

Only one neuron per slice was tested for its sensitivity to DAMGO, and the slice was not exposed to the drug before beginning of the recording to avoid a desensitization of opioid responses.16,32 

In each of 53 TFNs tested, a 30-s application of 1 μm DAMGO evoked a robust hyperpolarization ranging from 5 to 13 mV, which was accompanied by a considerable decrease in R^IN(fig. 3, A1 , and table 1; P < 0.001, paired Student t  test). Membrane currents were reversibly increased in such a way that the current–voltage curves in control and 1 μm DAMGO crossed each other at a potential close to the K⁺equilibrium potential (fig. 3, A2 ), indicating an activation of K⁺selective current. Although the DAMGO-activated hyperpolarization lasted several minutes (up to 18 min) without remarkable inactivation, most TFNs responded to the drug only once, and only some cells responded to several applications. Presence of 2 mm adenosine 5′-triphosphate and 100 μm guanosine 5′-triphosphate in the pipette solution16in 9 of 53 tested TFNs did not improve their responsiveness to repetitive DAMGO applications. The hyperpolarization in TFNs was prevented if DAMGO was applied together with 100 nm naloxone (n = 7).

Neither AFNs (n = 30) nor DFNs (n = 31) responded to 1 μm DAMGO, and their R^INdid not change (figs. 3B and Cand table 1; P > 0.85 for AFNs and P > 0.8 for DFNs, paired Student t  tests). Increasing the duration of DAMGO application to 1–3 min in all AFNs and DFNs or inclusion of 2 mm adenosine 5′-triphosphate and 100 μm guanosine 5′-triphosphate to the pipette solution in 15 of 30 AFNs and 12 of 31 DFNs did not affect their sensitivity to DAMGO.

Therefore, only TFNs and not AFNs or DFNs were hyperpolarized by DAMGO, which activated K⁺-selective conductance. The sensitivity of a neuron to DAMGO depended on its type rather than its location within the SG (fig. 2B).

In experiments shown in figure 4, the effects of Ba²⁺, Cs⁺, and baclofen were studied. In seven TFNs, membrane currents were first activated by 1 μm DAMGO, and then 500 μm Ba²⁺was added. In all these neurons, Ba²⁺blocked the DAMGO-activated current (fig. 4, A1 ). Direct application of 500 μm Ba²⁺(without DAMGO) blocked the resting membrane conductance in all cell types in a voltage-independent manner. At −120 mV, the inward current was blocked to 38.8 ± 4.3% in TFNs (n = 4, not shown; P < 0.01, paired Student t  test), 32.4 ± 3.8% in AFNs (n = 8; fig. 4B1; P < 0.001, paired Student t  test), and 30.9 ± 3.1% in DFNs (n = 11; fig. 4, C1 ; P < 0.001, paired Student t  test). Application of 500 μm Cs⁺also blocked the resting K⁺conductance in all groups of neurons with stronger effects seen at more negative potentials (fig. 4, A2 , B2 , and C2 ). At −120 mV, an inward current was suppressed to 42.0 ± 3.9% in TFNs (n = 6; P < 0.001, paired Student t  test), 43.9 ± 5.4% in AFNs (n = 8; P < 0.001, paired Student t  test), and 30.4 ± 3.4% in DFNs (n = 5; P < 0.001, paired Student t  test). All tested TFNs (n = 21), AFNs (n = 12), and DFNs (n = 12) responded to the GABA^Breceptor agonist baclofen (table 1), which, similar to DAMGO, was applied to one neuron per slice. In all three groups, 50 μm baclofen reversibly induced K⁺current (fig. 4, A3 , B3 , and C3 ). At −120 mV, the baclofen-induced current was −90.7 ± 7.1 pA in TFNs (n = 21), −73.4 ± 10.1 pA in AFNs (n = 12), and −72.9 ± 7.1 pA in DFNs (n = 12). In contrast to DAMGO, all neurons responded to several applications of baclofen. Therefore, all groups of SG neurons possessed GIRK conductance, which could be blocked by Ba²⁺and Cs⁺but activated by baclofen via  the GABA^Breceptor pathway.

To study the functional consequences of postsynaptic effect of opioids, the firing patterns before and after 30-s application of 1 μm DAMGO were compared in all 53 TFNs. In 44 of them (83%), the DAMGO application reversibly induced a spike frequency adaptation in such a way that the neurons were not able to support tonic firing at any stimulation strength (fig. 5A). In addition, much stronger current injection was needed to reach the firing threshold of the first spike (fig. 5A). The discharge pattern modified by DAMGO was not improved when the membrane potential was returned to −70 mV by injecting persistent current through the recording pipette (fig. 5A, 5 min after DAMGO application). In the other 9 TFNs (17%), the tonic discharge pattern remained after DAMGO application. However, the firing characteristic constructed for those 9 TFNs was considerably shifted to the right (fig. 5B).

The current results have shown that the μ-opioid agonist DAMGO selectively inhibits TFNs in SG, whereas it had no effect on AFNs and DFNs. Our finding implies that TFNs, which receive direct inputs from Aδ- and C-type afferents,27represent a primary postsynaptic target for both administered opioids and endogenous enkephalins in the spinal cord. Cell classification on the basis of firing patterns was important for this study. Our data support several patch clamp investigations showing that SG is formed by neurons with diverse intrinsic firing properties23,27,33,34but conflict with Ruscheweyh and Sandkuhler,29who have found only adapting discharge patterns in neurons from rat lamina II (SG). Our classification is also similar to those suggested on the basis of recordings with a sharp electrode.25,26However, it cannot be excluded that some TFNs from the current study would appear as AFNs if recorded with a sharp electrode introducing larger somatic shunt.

Each group of neurons had its specific distribution pattern within SG. The intermediate SG was dominated by TFNs, the medial region contained both TFNs and DFNs, and the lateral zone contained all three types of cells with a high percentage of AFNs. This is in good agreement with Melnick et al. ,23who showed that in parasagittal spinal cord slices, including mostly the intermediate region of SG, TFNs represent 70% of the total neuronal population. The sensitivity of a given neuron to DAMGO depended on its firing properties rather than its location within the SG. Therefore, our results suggest that the percentage of neurons responding to opioids might vary along the medial–lateral axis of SG, being highest in the intermediate zone.

Agonists binding to μ-opioid receptors hyperpolarize membrane through the activation of GIRK (also known as Kir3) channels functioning as effectors.19,35,36It seems that all three types of SG neurons possess GIRK conductance, which could be blocked by Ba²⁺and Cs⁺as well as activated by baclofen via  the GABA^Breceptor pathway.18Selective inhibition of TFNs by DAMGO could therefore be explained by a specific targeting of postsynaptic μ-opioid receptors to TFNs rather than AFNs and DFNs. Alternatively, it is possible that AFNs and DFNs also express μ-opioid receptors but that the G protein coupling them to GIRK channels is not present. Because the majority of SG neurons expressing μ-opioid receptors do not contain GABA or glycine and therefore are excitatory interneurons,14we suggest that TFNs can function as excitatory interneurons. The intrinsic firing properties would allow them to convert stronger synaptic inputs into higher discharge frequencies in a broad range of synaptic stimulation.

The current results show that the mechanism of postsynaptic action of opioids on SG neurons is more complex and effective than it was assumed so far. Membrane hyperpolarization due to activation of GIRK conductance resulted in increased stimulation intensity needed to reach the firing threshold. In addition, most TFNs showed a transition from tonic to adapting firing mode. Therefore, the input–output characteristics of the neurons were modified in such a way that stronger synaptic input could no longer be converted into increasing numbers of generated spikes. It should be also noted that the opioid-induced plasticity of firing behavior reported here can be a complex phenomenon, which, in addition to modulation of GIRK channels,37may also involve G protein–dependent regulation of some other ion channel systems.38–40 

DAMGO-sensitive interneurons from SG send some of their axons to lamina I and V,22where most projection neurons that target supraspinal regions are located.41,42Therefore, μ-opioid agonists can inhibit SG neurons that relay primary nociceptive afferent inputs to ascending projection neurons. Suppression of excitatory interneurons with a tonic-firing pattern by enkephalins may be an important mechanism of endogenous pain control because μ-opioid receptor–expressing dorsal horn neurons were shown to be located in close proximity to enkephalinergic terminals.7,43Under physiologic conditions, they may be involved in an endogenous, i.e. , stress-induced, analgesia.

Furthermore, TFNs possessing μ-opioid receptors are likely to participate in analgesic effects of administered opioids. High sensitivity of TFNs to μ-opioid agonist implies their involvement in analgesic effects of spinal, epidural, and systemic opioids, like morphine.9–11,44At low doses, opioids do not block voltage-gated Na⁺and K⁺channels,45–47and therefore, their specific action via  opioid receptors in tonic-firing spinal sensory neurons may contribute to profound and prolonged relief of pain with virtually no motor blockade.

The authors thank Helena Pereira (Technician, Departamento de Neurobiologia Básica e Clínica, Instituto de Biologia Molecular e Celular, Porto, Portugal) for technical assistance.