I. INTRODUCTION

Top Previous Next References

The original notion that perturbations in central nervous system (CNS) functions produced by opioid drugs initiate homeostatic processes leading to the development of opioid dependence (191) stimulated attempts to explain the nature of the relevant adaptations in neurons responsible for opioid addiction. These studies have focused on identifying the biological basis of the core features of addiction to opioid drugs, particularly tolerance, the withdrawal syndrome, and compulsive use of the drug in the face of known harm. With repeated administration of opioid drugs, adaptive mechanisms are initiated that result in short-term as well as protracted changes in the functioning of opioid-sensitive neurons and neural networks. One such mechanism is the development of tolerance to opioid drugs, such that higher doses are required to gain the desired effect. Although associative or conditioned tolerance, where morphine treatment is always paired with a distinctive environment, plays an important role and is mediated by specific neural systems in behaving animals (e.g., Ref. 325), nonassociative or cellular tolerance is a process that has received considerable attention. There are two very general forms of nonassociative tolerance that develop in the CNS, isolated tissues, and cells: one is at the level of the opioid receptor, where effector coupling is reduced, and the second is at the cellular, synaptic, and network levels, where counteradaptive changes occur to bring about normal function despite the continued activity of the drug.

The mechanisms involved in the initiation of compulsive self-administration of many drugs seem to be seated in common central pathways, particularly those thought to mediate endogenous reward. With repeated administration of these drugs, adaptive mechanisms are initiated. One such mechanism is the development of tolerance. Another results from development of counteradaptations such that once the drug is removed a sequence of rebound signs and symptoms are manifested. This withdrawal syndrome has short-, long-, and very-long-term features that may include craving and relapse to drug use long after acute withdrawal has ended. Thus long-term adaptations induced by chronic opioid treatment are expressed in the absence of the triggering drug and indicate long-lasting change in the functioning of specific neural systems. It is intriguing that the mechanisms that seem to be responsible for these adaptive processes in neurons and synapses are reminiscent of mechanisms involved in "normal" plasticity, such as long-term potentiation (LTP) and long-term depression (LTD), which are thought to form the cellular basis of memory. Indeed, recent work suggests that these adaptive processes at the cellular, synaptic, and network levels downstream from the receptor may hold the keys to understanding of addiction.

Perhaps the most important adaptations that develop as a result of chronic opioid administration occur in neural systems responsible for the transition from casual to compulsive drug use. Although tolerance and withdrawal surely contribute to this process, mechanisms involved in the initiation of compulsive self-administration of opioids as well as other major drugs of abuse seem to be seated in common central neural systems. The mesolimbic dopaminergic system, thought to have a crucial role in the rewarding actions of drugs of abuse, is a prime candidate for mediating this process. There is growing evidence (reviewed in Refs. 398, 449) that mesolimbic dopaminergic neurons are involved in strengthening formation of associations between salient contextual stimuli and internal rewarding or aversive events. The common long-term adaptations produced by opioids and other drugs of abuse in this system could enhance these processes and thereby play a major role in initiation and maintenance of compulsive drug use.

This review focuses on the adaptive changes in cellular and synaptic function induced by chronic morphine treatment. Opioids are known to be intensely addictive and share some general actions with other addictive drugs including psychostimulants and nicotine. One advantage that studies of the opioid system have over other addictive drugs has resulted from the massive research effort to find an opioid that is effective for the treatment of pain but lacks addictive properties. A plethora of compounds are available with varying affinity and efficacy at all the known opioid receptor subtypes. Although compounds of extremely high potency have been produced, the problem of tolerance to and dependence on these agonists persists.

The initial steps of opioid action are mediated through the activation of G protein-linked receptors. As is true for all G protein receptors, opioid receptors activate and regulate multiple second messenger pathways associated with effector coupling, receptor trafficking, and nuclear signaling. These initial effects are critical for understanding the early events leading to tolerance and dependence in cells that have opioid receptors. Equally important are network changes that occur as a result of the altered synaptic regulation that may affect downstream neurons that may not have opioid receptors. Finally, opioids and other drugs of abuse have some common cellular and anatomical pathways. The characterization of common pathways particularly after chronic drug treatment is an important extension in the understanding of drug abuse.

	II. INITIAL STEPS OF OPIOID ACTION

Top Previous Next References

Multiple opioid receptors were initially predicted on the basis of the actions of various alkaloid agonists and antagonists in whole animal preparations (155, 315). Soon after the discovery of endogenous opioid peptides, multiple opioid receptors were confirmed functionally using isolated pharmacological preparations (286). From these studies three major receptor subtypes were identified: µ, , and  (161). Highly selective and potent ligands have been developed for each of the three general receptor subtypes (Fig. 1). At each of these receptors both agonists and antagonists exist that are >1,000-fold selective. As is always the case, however, the selectivity window of any ligand can be exceeded such that results obtained even with a highly selective agent may be misinterpreted. The issue of binding selectivity of ligands at the opioid receptor has been critically examined and reviewed by Goldstein (160).

View larger version (20K): [in this window] [in a new window]   Fig. 1. An illustration of the selectivity windows of some commonly used opioid agonists and antagonists, determined in an expression system (392). Top: compounds tht are selective for each of the opioid receptors. Note that although nor-BNI is highly selective, the inhibition constant (Ki) at µ-receptors is ~3 nM. Bottom: the selectivity of the endogenous opioids and other commonly used opioids. Again note that none of the endogenous opioids show a high degree of selectivity. DAMGO, [D-Ala2,N-Me-Phe4,Gly5-ol]-enkephalin; nor-BNI, norbinaltorphimine; CTAP, H-D-Phe-c[Cys-Tyr-D-Trp-Arg-Thr-Pen]-Thr-NH2; DPDPE, [D-Pen(2),(5)]-enkephalin.

Pharmacological studies have attempted to further divide opioid receptors in each of the three major subgroups. Although suggestive, pharmacologically defined subclasses of µ, , and  receptors are not well established. The cloning of each of the three major opioid receptors has done little to support further expansion of opioid receptor classification (392, 393). There is ~60% sequence homology between the µ, , and  receptors. Unless undiscovered opioid receptors are significantly different from receptors described to date, it appears that the opioid receptor family has been defined. There are reports of alternate splice variants, although it is not clear at what level they are expressed or if they can be distinguished pharmacologically (153, 256, 368, 421). The lack of molecular evidence for more than three opioid receptor subtypes indicates that further subclassification of receptors may result from mechanisms that may include posttranslational regulation, receptor dimerization, or even interactions with accessory proteins.

The use of ligands with differing efficacy in tissues having varying receptor reserve is one potential confounding problem in the pharmacological classification of multiple opioid receptors. The description of the epsilon opioid receptor in the rat vas deferens is one such example. In this preparation -endorphin decreased the muscle contraction evoked by electrically stimulating transmitter release from the nerves. Morphine was ineffective in this preparation. From this observation the -endorphin selective epsilon receptor was characterized (151, 419). Subsequent work showed that the receptor reserve of µ-receptors in this preparation was low enough that a partial agonist, such as morphine, acted as a pure antagonist (442, 427). The characterization of multiple receptors based on results obtained in more complex tissues using indirect assays are subject to the same difficulties in interpretation.

It now appears that many G protein-linked receptors exist as dimers (108). The most dramatic demonstration of dimerization of G protein-linked receptors is with the GABA_B receptor, where heterodimerization with two subtypes of the receptor are required for functional expression (228, 242, 271, 519). Both - and -opioid receptors have been reported to form homodimers. Recently, heterodimers of - and -opioid receptors were expressed in Chinese hamster ovary (CHO), HEK 293, and COS cells (229). The pharmacological profile of heterodimers was not completely characterized but differed from the homodimers of both - and -receptors. Heterodimerization of receptors in vivo could account for complex pharmacology even if there is only a single gene for each receptor.

The cellular and anatomical distribution of opioid receptors is important for the identification of neuronal systems and local networks involved in the initiation of drug action and the subsequent development of adaptations resulting in repeated drug use. Distinct distributions and developmental patterns of receptor and mRNA subtypes have been identified throughout the neuroaxis as well as in paracrine and exocrine tissues (14, 15, 307, 308, 547, 548) . The widespread occurrence of these receptors indicates that opioids have the potential for affecting multiple systems, both nervous and hormonal. The cellular distribution of µ- and -receptors seems to be largely along the plasma membrane, both at somata as well as dendrites and nerve terminals. Receptors are generally found in perisynaptic areas, rather than in subsynaptic sites (336). The -receptor differs in that it is most often found within cells associated with vesicles (547). The activity-dependent redistribution of both - (432) and -receptors (R. Elde, personal communication) from vesicles to the plasma membrane suggests that the localization of receptors is not static and may vary considerably with activity. The identity and distribution of receptor subtypes in local circuits will be discussed in specific sections on neuronal systems.

Activation of any of the three opioid receptor subtypes produces common cellular actions. Each receptor is coupled to pertussis toxin-sensitive G proteins, although some coupling to the pertussis toxin-insensitive G protein G_z has also been recognized (see Ref. 99 for review). The profile of coupling of the three opioid receptors to the spectrum of G proteins is similar, although subtle differences have been identified (99). The most commonly reported actions include inhibition of adenylyl cyclase, activation of a potassium conductance, inhibition of calcium conductance, and an inhibition of transmitter release (Fig. 2). More recent observations have extended the actions of opioids to include the activation of protein kinase C (PKC), the release of calcium from extracellular stores, the activation of the mitogen-activated protein kinase (MAPK) cascade, and the realization that receptor trafficking plays an important role in receptor function.

View larger version (27K): [in this window] [in a new window]   Fig. 2. An illustration of the best-characterized pathway of effector activation of opioids. Three primary classes of effectors include the inhibition of adenylyl cyclase, inhibition of vesicular release, and interactions with a number of ion channels. These effectors are affected by both the GTP-bound form of the -subunit as well as free /-subunits of pertussis toxin-sensitive G proteins. GIRK, G protein inwardly rectifying conductance.

Until recently, nothing was known of the physiological consequences of the acute inhibition of adenylyl cyclase by opioids. Two effects have now been identified: one is mediated by the modulation of a voltage-dependent current (I_h), which is also termed the pacemaker current (205, 471). This cation nonselective current is activated at hyperpolarized potentials to cause an inward current that depolarizes the membrane potential toward threshold. The voltage dependence of this current is regulated by cAMP, being activated at less negative potentials when cAMP levels are elevated (204). Opioids shift the voltage dependence to more negative potentials by decreasing intracellular cAMP. This inhibition was most easily observed after the activation of adenylyl cyclase with forskolin or PGE₂ (205) but has also been observed without prior activation (471). The consequence of this action of opioids was to decrease the amplitude of the inward current that drives spontaneous activity and thus decreases excitability. This action of opioids could have been predicted based on work done on the pacemaker current in sinoatrial nodal cells of the heart where the activation of M₂ muscarinic receptors shifted the voltage dependence of I_h through an inhibition of adenylyl cyclase (121). A family of these cation channels has now been cloned, some of which show the same cAMP-dependent changes in voltage dependence (152, 289, 412). A similar effect of opioids has also been observed on a tetrodotoxin-insensitive, cAMP-sensitive sodium current in cultured sensory neurons (158). Activation of adenylyl cyclase with prostaglandin E increased a sodium current that was depressed by [D-Ala²,N-Me-Phe⁴,Gly⁵-ol]-enkephalin (DAMGO). This effect, similar to the decrease in I_h, would be expected to reduce excitation caused by agents that are thought to mediate hyperalgesia.

The second consequence of the inhibition of adenylyl cyclase was an inhibition of transmitter release that was dependent on the activation of adenylyl cyclase (86, 203, 430). Previously there was no indication that the inhibition of adenylyl cyclase affected transmitter release. Under conditions where adenylyl cyclase was activated and caused an increase in transmitter release through activation of cAMP-dependent protein kinase (PKA), opioids decreased transmitter release via a PKA-dependent mechanism. This action of opioids was not observed at all opioid-sensitive synapses, which may suggest differential distribution of adenylyl cyclase isoforms at individual synapses (see sect. IIIB4).

An activation of adenylyl cyclase by opioids has been reported in both primary afferent neurons (105) and the olfactory bulb (362). Studies in the olfactory bulb indicate that the pA₂ for naloxone was ~8 [dissociation constant (K_d) = 10 nM], suggesting that this response was mediated through activation of -opioid receptors (362). The increase in adenylyl cyclase activity was not affected by pretreatment with cholera toxin and was blocked with pertussis toxin. More recently, the same group has found that the increase in adenylyl cyclase activity was mediated by the release of /-subunits from pertussis toxin-sensitive G proteins (360). A similar mechanism for the opioid activation of adenylyl cyclase was proposed in a study using a membrane preparation of longitudinal muscle-myenteric plexus from guinea pigs chronically treated with morphine (72). Thus it appears that the opioid regulation of adenylyl cyclase is dependent on the isoform under study and the absence or presence of coactivated G_s. In the olfactory bulb it appears that the conditions are such that acute administration of opioids can activate the cyclase, whereas in other tissues, this response is observed only after adaptations induced by chronic morphine treatment.

Opioids have been shown to activate at least three separate potassium conductances. The most commonly observed is the G protein-activated inwardly rectifying conductance (GIRK; Fig. 3, Table 1). All three opioid receptors have been shown to activate this conductance. The second messenger pathway is membrane delimited, mediated by a pertussis toxin-sensitive G protein (7), and it is presumed that the potassium conductance is activated by the /-subunits (212). Rapid application and washout of opioids allowed the determination of the kinetics of opioid action using acutely dissociated cells (202). The activation of the potassium conductance had a latency to onset of 50-100 ms and a time constant of activation of ~700 ms, which is similar to that observed for other receptors coupled to GIRK channels (202, 444, 445). The termination of the GIRK current was dependent on the agonist applied. The rate of recovery was slower when higher affinity agonists were used, suggesting that receptor unbinding may be the rate-limiting step for deactivation.

View larger version (19K): [in this window] [in a new window]   Fig. 3. The kinetics of opioid action on potassium and calcium conductances are similar. A: activation of opioid receptors on actuely isolated locus coeruleus (LC) neurons increases a potassium conductance with a time constant between 0.6 and 0.8 s (bottom trace; Ref. 202). The opioid activation of the potassium currents is much slower than the time course of solution exchange as illustrated in the top trace. B: time course of opioid action on potassium and calcium currents is comparable. Top traces show the voltage-clamp protocol. The traces below are superimposed current traces carried out in the absence and presence of [Met]enkephalin (ME) and showing an inward current at 60 and 80 mV and a decrease in the inward (calcium) current measured at 0 mV. The sharp inward current seen at the start of the step to 0 mV is a sodium current (no tetrodotoxin). The plot below shows the time course of these actions. The increase in potassium conductance and the inhibition of calcium current are mirror images. C: the onset of opioid inhibition of calcium current (ICa) measured in outside-out membrane patches from primary afferent neurons (521). With the rapid application of DAMGO (1 µM), there was a delay of 150 ms before any inhibition, and the time course of inhibition was fit by an exponential having a time constant of 1.3 s. [From Wilding et al. (521). Copyright 1995 by the Society for Neuroscience.]

The coupling of receptor to this potassium conductance was quite dependent on the experimental conditions. Whereas in brain slice experiments morphine produced an increase in potassium conductance that was equivalent to that induced by the peptide agonists DAMGO and [Met⁵]enkephalin, it was an antagonist when tested in isolated cells. Under the same conditions, however, morphine had an agonist action on the inhibition of calcium currents. Thus it appears that the coupling of opioid receptors to GIRKs may be less efficient than to other effectors (202).

Opioids have also been shown to activate a voltage-dependent potassium conductance in acutely dissociated cells from hippocampus (526) and in brain slices (295). The activation of a voltage-dependent potassium conductance was also suggested based on the blockade of opioid inhibition of transmitter release by 4-aminopyridine and dendrotoxin (see sect. II B4). In addition, opioids have been reported to activate the BK calcium-sensitive potassium conductance (490). This effect, coupled with recent reports of opioid-induced calcium release from internal stores (100, 217, 219, 475, 476), indicates the diversity of opioid action, which has been recognized to occur with other G protein-coupled receptors.

There are many examples of the inhibition of calcium currents by activation of all opioid receptor subtypes (Table 1). The inhibition of high-threshold calcium currents by opioids, in common with other receptors linked to pertussis toxin-sensitive G proteins, 1) is membrane delimited, 2) is mediated by the /-subunits of G proteins, 3) decreased the rate of current activation such that the inhibition was greater immediately after the voltage step, and 4) showed relief of inhibition following a depolarization to positive potentials (521). The kinetics of activation of this effect of opioids were similar to that reported for the activation of potassium conductance, having a latency of onset of ~150 ms and peaking in ~5 s (Fig. 3, Ref. 521).

The opioid inhibition of acetylcholine release in the guinea pig ileum and ATP release in the vas deferens has been used as pharmacological assays for many decades (187, 265, 371). In various peripheral preparations from different species, the activation of all three receptor subtypes has been found to cause inhibition of transmitter release (264, 318, 507). The activation of potassium conductance and/or the inhibition of calcium conductance and not the inhibition of adenylyl cyclase have been argued to account for this action (35, 355, 415), although recent work suggests that under some conditions the inhibition of adenylyl cyclase can also account for some of the decrease in transmitter release (see sect. III B). Direct inhibition of the release machinery, independent of potassium and calcium conductances, has also been reported (65).

Depending on the site, opioids inhibit release of excitatory and/or inhibitory transmitters (Table 1). Opioid inhibition of GABA release in local circuits, first observed in the hippocampus, has become a common observation that accounts for indirect excitatory, or disinhibitory, effects of opioids (351, 549). Opioids caused direct hyperpolarization of interneurons, thus decreasing excitability of these cells (296). In addition, spontaneous quantal release of GABA from terminals was decreased by opioids, suggesting that opioids also acted directly on axon terminals to decrease the probability of GABA release (92). A similar disinhibitory mechanism mediated by opioids acting on local circuits has now been described in brain regions where the local circuitry is not as well defined, such as the raphe magnus (370), ventral tegmental area (225), periaqueductal gray (PAG) (498, 499), and dorsal raphe (226, 227). This indirect action of opioids on output neurons from nuclei such as the VTA and PAG may be critical in the understanding of circuit adaptations in response to chronic morphine treatment.

Of equal significance is the fact that transmitter release is the result of a complex series of events with numerous protein-protein interactions such that there are multiple sites of potential regulation. Opioid receptors are one of a vast number of G protein-linked receptors that modify transmitter release. Given the potential interactions between these receptors and the effects of prior activity in any given terminal, the effects of opioids may vary considerably. Some of the consequences of the receptor interactions have been identified in the form of modulation of activity-dependent plasticity such as posttetanic potentiation, LTP, and LTD.

A long-term, selective augmentation of N-methyl-D-aspartate (NMDA)-mediated glutamate currents by activation of µ-opioid receptors was observed in brain slices of trigeminal nucleus (80). This augmentation was mimicked by phorbol esters and blocked by the peptide inhibitor of protein kinase C (PKC). It was concluded that opioids activated PKC, which then increased the conductance activated by NMDA receptor agonists. This was the first and remains the strongest evidence that opioids augment postsynaptic glutamate currents by a mechanism involving the activation of PKC. There have, however, been more recent studies showing augmented NMDA receptor-mediated excitatory postsynaptic currents (EPSCs) by µ-opioid agonists in both the nucleus accumbens and hippocampus (312-314, 384). This augmentation was not observed in the locus coeruleus (LC) (359), suggesting that it may be dependent on the makeup of NMDA receptor subunits and/or the isoforms of PKC present on any given cell type.

The activation of PKC by opioids appears to result from the activation of phospholipase C and/or phospholipase A₂, which is thought to result from an interaction of /-subunits of pertussis toxin-sensitive G protein and may require coactivation with the -subunits of pertussis toxin-insensitive G proteins (144, 342, 358, 441). The results suggest that in order for opioids to have a robust effect, coactivation with G_q subtype G proteins is required. A similar pathway is also thought to mediate the release calcium from inositol 1,4,5-trisphosphate (IP₃)-sensitive stores (see sect. IIB6).

Initial studies demonstrating a transient increase in intracellular calcium in NG108-15 cells were unexpected (217). Most of the rise in calcium found in both NG108-15 and ND8-47 cells was blocked with dihydropyridine calcium channel blockers or removal of extracellular calcium, suggesting that calcium entry across the plasma membrane was the primary source (217, 475). Further investigation in both NG108-15 and SH-SY5Y cells indicated that a component of the increase in calcium resulted from release from intracellular stores (100, 219). This effect of opioids was sensitive to pertussis toxin, depletion of stores by thapsigargin, and an inhibitor of phospholipase C, U73122 (219). Microinjection of a peptide that binds to /-subunits (QEHA) blocked the opioid-induced increase in calcium in NG108-15 cells (539). Injection of a peptide that blocked bradykinin-induced activation of G_q did not block the opioid-induced increase in calcium (539). Thus the interaction of /-subunits with phospholipase was not dependent on coactivation of G_q, although other potential -subunits were not excluded. In experiments on SH-SY5Y cells, the increase in intracellular calcium was dependent on coapplication of agonists (muscarinic) that activated receptors coupled to phospholipase C (100). Taken together, it appears that the opioid activation of phospholipase results from an interaction with /-subunits of pertussis toxin-sensitive G proteins and subsequent production of IP₃ and diacylglycerol (DAG), which release stores of calcium or activate PKC, respectively. This action of opioids is either significantly enhanced or completely dependent on the coactivation of receptors that are directly coupled to phospholipase. Although opioids have been shown to increase intracellular calcium in primary afferent neurons (476), the signaling pathway for this effect has not been identified.

With the cloning of opioid receptors has come a better understanding of the mechanisms that regulate the life cycle of receptors. It is clear that opioid receptors, as is the case with many G protein-linked receptors, are not static and cycle to and from the plasma membrane. Most opioid binding studies were primarily directed toward sites on the plasma membrane in both neuronal and nonneuronal tissues; however, opioid binding sites in intracellular compartments have been recognized for some time (32). These binding sites were considered to be newly synthesized or recycled receptors. On the basis of studies using antibodies directed at each of the opioid receptors, it was realized that significant immunoreactivity for both the - and -opioid receptors was associated with intracellular compartments, particularly in axonal projections (432, 547). From this observation, it was hypothesized that the intracellular receptors were associated with a regulated secretory pathway in terminals. -Opioid receptors were found on vesicles in nerve terminals of vasopressin-containing neurons and were translocated to the plasma membrane after a physiological stimulus (432). Interestingly, 1 h after the stimulus, the receptors disappeared from the plasma membrane and reappeared in the vesicular compartment. Thus receptors found in vesicular membranes could be both newly synthesized and/or recycled. Although it has not been determined if the receptors freshly inserted into the plasma membrane were functional, this observation suggests an interesting form of feedback inhibition that would be dependent on prior activity in any given terminal.

Receptor trafficking initiated by agonist binding and internalization through the endosomal pathway may be involved in desensitization and/or the initiation of nuclear signaling (see below). The COOH-terminal tail of opioid receptors, as other G protein-linked receptors, regulates the extent and efficiency of internalization. The events leading to internalization are based primarily on the -receptor model (Fig. 4). Upon agonist occupation, a receptor kinase (BARK2) phosphorylates the receptor, uncoupling the G protein and increasing the affinity of the receptor for arrestin (269). This triggers a series of events that carry the receptor complex to clathrin-coated pits and into an endosomal compartment.

View larger version (26K): [in this window] [in a new window]   Fig. 4. An illustration of the sequence of events leading to receptor internalization. Some but not all (notably morphine) opioid agonists can activate the pathway. The activated opioid receptor is phosphorylated by a G protein receptor kinase. The affinity of interaction between this complex and arrestin is increased. The arrestin-bound complex recruits c-Src adaptor proteins (AP-2 complex) that link arrestin and clathrin to promote endocytosis (reveiwed in Ref. 252).

Internalization of the -opioid receptor was suggested to depend on phosphorylation of the COOH terminal as indicated by experiments using either point mutations, at Ser-344 and Ser-363 or COOH-terminal truncations (484). Similar experiments were carried out using the two alternatively spliced isoforms of the µ-receptor, MOR1 and MOR1B (256). The MOR1B isoform is the shorter of the two and lacks one phosphorlyation site, Thr-394. The MOR1B receptor was more resistant to desensitization, more rapidly internalized, and recovered from desensitization more rapidly than MOR1. It was also suggested that MOR1B was internalized constitutively. Thus it appears that regulation of internalization is highly dependent on the COOH-terminal tail.

Internalization is also agonist dependent (245, 516). The amount of internalization induced by a series of agonists was compared with efficacy in standard signal transduction assays using the µ-opioid receptor transfected into HEK 293 cells (516). Morphine was among the agonists that did not cause internalization. Morphine is well known as a partial agonist so the lack of activity was not surprising (266). Methadone, another partial agonist like morphine, resulted in efficient internalization (516). In addition, morphine was capable of causing internalization under conditions where the coupling efficiency was increased by overexpression of GRK2 (547). Finally, with the use of a chimera of the µ-receptor combined with the COOH terminus of the -opioid receptor, morphine mediated internalization in both a cell line and in primary cultures of hippocampal cells, suggesting that this portion of the receptor was responsible for lack of morphine-induced internalization (516). Given this result, it would be interesting to examine the effect of morphine on the MOR1 and MOR1B splice variants. Internalization of -receptors in NG108-15 cells was facilitated by overexpressing arrestin, such that morphine was capable of causing internalization. It therefore appears that morphine is a differentially weak partial agonist at producing internalization. The quantification of relative efficacies of morphine and other opioids was incomplete in these experiments, relying only on maximal responses. It is therefore not yet clear how much the relative efficacies of morphine, methadone, and other opioids differed for different signal transduction processes. Similarly, differential rank orders of efficacy of DAMGO, methadone, L--acetyl methadol, morphine, and buprenorphine were previously reported for µ-receptor phosphorylation, potassium channel activation, and desensitization (541). The results of these studies suggest that the efficiency of agonists to couple with different effectors varies and that coupling to mechanisms responsible for phosphorylation and internalization is rather inefficient. It appears that all of the endogenous agonists and a number of alkaloid agonists (but not others) are potent activators of internalization regardless of their ability to induce G protein activation. Similar interpretations have been made for phosphorylation of µ-receptors by PKA (71) and efficiency of activation of different G protein -subunits (150). If this conclusion proves correct, then it implies that distinct µ-opioid receptor conformational states exist for coupling to different effectors (96), with the implication that agonists with selective conformational state profiles can be developed deliberately.

The inability for morphine to cause internalization under most circumstances may also be an important issue in the eventual understanding of the cellular, synaptic, and network effects of chronic morphine treatment. Given the idea that receptor internalization may be a mechanism that mediates receptor turnover and resensitization, the inability of morphine to mediate this basic response could result in a chronic receptor activation. The continuous stimulation of transduction pathways may recruit signaling pathways to cause downstream adaptations (counteradaptations) with repercussions unrelated to direct actions of opioids.

The regulation of cellular events through altered expression of proteins signaled by the chronic activation of opioid receptors is critical for understanding tolerance and dependence to opioids. The signaling pathways that lead to altered genetic expression are simply another effector system. Unlike membrane delimited effectors such as potassium and calcium channels, the extracellular signal-regulated kinase (ERK)/MAPK pathway that mediates nuclear signaling involves multiple protein-protein interactions, translocations, and phosphorylation events (Fig. 5). The activation of this pathway by opioids, like other opioid effectors, is sensitive to pertussis toxin. The kinetics of activation are longer than that for other effectors but occur over a period of several minutes to 1-2 h.

View larger version (37K): [in this window] [in a new window]   Fig. 5. Multiple pathways can lead to the activation of the extracellular signal-related kinase (ERK)/mitogen-activated protein kinase (MAPK) transduction cascade by opioids applied acutely or after chronic treatment. Three main transduction pathways can cause the activation of ERK/MAPK during acute application of opioid agonists. /-Subunit release could 1) stimulate phospholipase C and cause the release of calcium from internal stores and the production of diacylglycerol, which will in turn activate protein kinase C; 2) recruit to the membrane proteins such as Ras-GRF; and 3) activate the phosphatidylinositol 3-kinase (PI 3-kinase). The ERK/MAPK cascade can be blocked at the level of the MEK by pharmacological agents such as PD-98059 or by phosphatases. Activated ERK/MAPK has multiple targets, including nuclear transcription factors (such as CREB), cytoplasmic enzymes (including tyrosine hydroxylase), cytoskeletal proteins, and ion channels. Because morphine causes no or little internalization of µ-receptor, it is possible that the MAPK pathway is under permanent opioid stimulation during chronic drug treatment. After withdrawal, the adenylyl cyclase superactivation could now lead to the activation of the ERK/MAPK cascade through intracellular elevation of cAMP and activation of the protein kinase A (PKA). PIP2, phosphatidylinositol 4,5-bisphosphate; IP3, inositol 1,4,5-trisphosphate; PKC, protein kinase C.

It appears that there are at least three general pathways following the activation of G_i/G_o-linked receptors that eventually converge on the activation of MAPK. One pathway involved ß-subunit activation of phosphatidylinositol 3-kinase, which activates MAPK activity through a series of phosphorylation steps including the activation of c-Src (139, 382, 383). The second pathway involves phosphorylation of the receptor with a receptor kinase; the translocation and binding of arrestin to the receptor is followed by translocation of c-Src to the membrane before internalization of this receptor complex through clathrin-coated pits (198, 293). Once internalized, the complex activates ERK, which is translocated into the nucleus to affect gene regulation by any of a number of transcription factors (250). In the CNS, ERK/MAPK can be activated by PKA (200). The PKA-dependent activation of ERK/MAPK by opioids may not be important during acute opioid administration but may be facilitated by an upregulation of adenylyl cyclase activity with chronic treatment. Activated ERK/MAPK can phosphorylate multiple targets in the cytoplasm and in the nucleus (i.e., transcription factors such as CREB).

This second pathway is significant in that internalization is a necessary step. Morphine, an agonist that does not activate internalization, would therefore be ineffective at activating the MAPK pathway, at least through this mechanism. It is interesting to note that, with one exception (279), all reports on the activation of the MAPK pathway have used agonists, such as etorphine, DAMGO, [D-Pen(2),(5)]-enkephalin (DPDPE) and U69593, that effectively cause receptor internalization (30). In the study where morphine did activate MAPK activity, the activation was transient relative to that caused by etorphine, DAMGO, and PLO17 (279). In an in vivo study where phosphorylated MAPK was examined immunohistochemically after acute and chronic treatment of animals with morphine, the distribution of immunoreactivity was unchanged by acute morphine treatment (420). Withdrawal from chronic morphine treatment precipitated by injection of naloxone, however, produced a robust increase in phospho-MAPK immunoreactivity in specific brain regions, among them the LC. Taken together, these observations suggest that in many systems morphine alone may be ineffective at activating the MAPK pathway. After chronic morphine treatment, however, adaptive mechanisms may facilitate the activation of MAPK. For example, in many brain areas, MAPK can be increased by excitatory transmission (17). The increased release of glutamate in the LC during withdrawal could be a trigger that activates the MAPK pathway.

	III. CELLULAR ADAPTATIONS INDUCED BY CHRONIC MORPHINE TREATMENT

Top Previous Next References

A spectrum of cellular adaptations resulting from chronic exposure to opioids is responsible for nonassociative tolerance and physical dependence. Humans and experimental animals can develop profound tolerance to opioids over periods of several weeks of escalating chronic treatment. Thus hundreds of times the normal analgesic dose of morphine have been reported to produce only mild physiological effects in some addicts (211) and chronic pain patients (166). Tolerance development involves a number of distinct cellular and neural processes (see below). The desensitization/downregulation mechanisms involved in tolerance are necessarily passive and do not engage the rebound mechanisms that could underlie maintenance of drug dependence and the opioid withdrawal syndrome. The latter require development of counteradaptations as outlined as follows: 1) acute desensitization of opioid receptor to effector coupling and internalization that develops during and abates shortly (minutes to hours) after exposure to agonists; 2) long-term desensitization of receptor to effector coupling and downregulation of receptors that slowly develop and then persist for many hours to days after removal of agonists; 3) counteradaptations of intracellular signaling mechanisms in opioid sensitive neurons; and 4) counteradaptations in neuronal circuitry.

Two phases of development of and recovery from tolerance can be distinguished in humans and experimental animals. The onset of the rapid phase of tolerance occurs within minutes. It is difficult to quantify in whole animals but dissipates with a time course approximating the elimination of opioids (104). The slowly developing phase dissipates over several weeks regardless of the opioid used for induction (104). The rapid phase seems to predominantly involve acute desensitization, but counteradaptations can also play a role (see sect. III). The slow, persistent phase of tolerance involves the latter three mechanisms that are less fully understood than acute desensitization. The relative contributions of each of the above mechanisms to net tolerance development depends on the physiological system in question. However, the mechanisms involved have important implications in determining the extent of tolerance development, which varies greatly between different organ systems. The following will focus primarily on mechanisms identified for µ-receptor desensitization and downregulation because actions at µ-receptors are of major relevance for tolerance development. The involvement of - and -receptor desensitization and downregulation in tolerance development is unclear and has not been extensively studied.

A.  Desensitization, Internalization, and Downregulation of Receptor to Effector Coupling

1.  Acute desensitization

Acute desensitization may be homologous, heterologous, or can involve both mechanisms. Homologous desensitization is by definition restricted to the opioid receptor occupied by the agonist and its specific interactions with signal transduction cascades. In contrast, heterologous desensitization generalizes to other receptors present in the same cells, and/or other elements of the signal transduction cascade such as G protein and ion channel activity. As discussed below, homologous desensitization can potentially involve phosphorylation of occupied receptors, occupancy-dependent protein-protein interactions, and/or occupancy-dependent compartmentalization and internalization. Homologous desensitization of µ-receptors has been described in a variety of cellular models using different end points. Although a general consensus on the mechanisms involved is emerging, both qualitative and quantitative differences have been reported in various experimental systems. These differences may arise from expression of diverse signaling elements, e.g., G protein receptor kinases, the stoichiometry of signaling elements, and the end points measured.

Several mechanisms of homologous desensitization of µ-receptors have been recognized. The most thoroughly studied for µ-receptor desensitization and internalization involve G protein-coupled receptor kinase (GRK)-mediated receptor phosphorylation that promotes the binding of -arrestin proteins. This process not only uncouples opioid receptors from their cognate heterotrimeric G proteins, but also targets them for endocytosis. The processes by which this is thought to occur have been reviewed elsewhere (58, 277) and are outlined in Figure 4. Agonist binding to the receptor promotes a conformational change that results in G protein activation and dissociation from the receptor. Free G protein /-subunits facilitate translocation of GRKs to the membrane where they phosphorylate serine and threonine residues in the COOH-terminal region. The phosphorylated receptor binds with high affinity to the cytoplasmic protein arrestin, which prevents association of inactive G proteins with the receptor and initiates internalization. Many of the details of this scheme have been confirmed for the µ-receptor, but a number of contentious issues remain.

Phosphorylation of the µ-receptor by GRKs does not appear to be necessary in all cases for desensitization to occur in neurons and some test systems, implying that other mechanisms mediating desensitization exist. Truncation of the COOH-terminal tail of the µ-receptor, the likely region for interaction with and phosphorylation by GRKs, greatly attenuates desensitization. When reconstituted in Xenopus oocytes, homologous desensitization of µ-receptor-mediated coupling to potassium currents was dependent on GRK2 (268). However, the desensitization kinetics were much slower (tens of minutes vs. minutes) than observed in CNS neurons (364), suggesting that components of the signaling cascade or stoichiometry of signaling elements were inappropriate. Other studies in reconstituted systems have also suggested that GRK phosphorylation and activation of MAPK pathways are essential for µ-receptor desensitization to occur (382, 383). Inhibitors of GRK-mediated phosphorylation such as Zn and heparin should be able to address the issue of GRK requirement, but unfortunately, they disrupt many aspects of cell function. Nonetheless, these inhibitors have generally not been shown to reduce desensitization (27, 364). In LC neurons, rapid ( = 120 s), homologous desensitization of coupling between µ-receptors and inwardly rectifying potassium channels was not affected by heparin or staurosporine (180, 364), suggesting that GRK-dependent processes are not involved. Dominant negative GRK mutants (kinase activity deficient) should provide more conclusive results but have not yet been used to examine µ-receptor desensitization. Desensitization of some other GPCRs, e.g., adenosine A_2a and A_2b receptors, were greatly inhibited by dominant negative GRK2, but others such as prostanoid and somatostatin receptors were completely unaffected (27). GRK-/ inhibitory peptides also failed to inhibit somatostatin receptor-mediated desensitization (27). The possibility that subtype specific interactions of different G protein /-subunits with the MAPK signaling cascade (109) cannot adequately resolve these difficulties because receptor coupling through the same elements in single cells showed different results. Desensitization of m₂ muscarinic receptors required the presence of either authentic GRK2 or the dominant negative form (431), suggesting that binding of GRK2 to the receptor is sufficient to cause desensitization in the absence of phosphorylation and initiation of internalization processes. However, it was not established whether or not desensitization was homologous in those studies, so the importance of the mechanism is still unclear. Enhanced morphine antinociception in mice lacking -arrestin-2 (37) is suggestive of an involvement of GRK-dependent mechanisms in µ-receptor desensitization, but it has not been established that rapid desensitization is disrupted in these knockouts.

Other phosphorylation-dependent mechanisms of µ-receptor desensitization have also been examined. Although PKA-mediated desensitization occurs in some G protein-coupled receptor systems, this has not been established for the µ-receptor. Inhibitors of PKA signaling have generally had no effect on desensitization (180, 183, 333, 344, 364) and activators of PKA signaling have an inhibitory effect on desensitization (81, 180), but it is not clear whether this inhibition was homologous or heterologous. PKC-mediated phosphorylation of µ-receptors has also been reported to reduce sensitivity to opioids (494, 546), but the process is independent of occupancy by agonists and therefore is heterologous. Phosphorylation by calmodulin kinase II of Ser-261/Ser-266 in the third intracellular loop of the µ-receptor has also been implicated in increasing the rate of desensitization (255). The third intracellular loop of the µ-opioid receptor has also been implicated as a binding site for calmodulin, which when associated with the receptor reduces coupling to G proteins (506). However, the functional significance of the latter observations for signaling and desensitization has not been characterized. Various other kinases such as cGMP-dependent protein kinase (323) and casein kinase (481) have been shown to modulate the activity of other G protein-coupled receptors, but these have not been examined for the µ-receptor.

It is not yet clear if phosphorylation-independent mechanisms are involved in initial events of homologous µ-receptor desensitization. The temporal correlation between desensitization, measured by inhibition of adenylyl cyclase activity, and at least some types of receptor phosphorylation events is poor (129). As indicated above, it is possible that phosphorylation-independent interactions of GRKs with µ-receptors functionally uncouple G protein interactions. In LC neurons using coupling to activation of potassium currents as an end point, µ-receptor desensitization was largely homologous (135, 180, 363) but was not affected by GRK inhibitors such as heparin or serine/threonine kinase inhibitors including staurosporine (Fig. 6). The phosphatase inhibitors okadaic acid and microcystin had no effect on onset of desensitization but markedly slowed recovery. The time course of desensitization was rapid under these conditions with onset and offset time constants of ~3 min. This is comparable to desensitization rates measured by others using electrophysiological methods in neurons (e.g., Refs. 27, 333) but are markedly faster than time courses usually measured (when they have been) in reconstituted systems using electrophysiological techniques or biochemical measures such as GTPase activity and inhibition of adenylyl cyclase. The latter time courses often proceed over periods of several hours (e.g., Refs. 256, 266, 382, 383). Whether or not these differences reflect fundamentally different mechanisms or differences in stoichiometry of elements in native versus reconstituted systems has not been resolved.

View larger version (25K): [in this window] [in a new window]   Fig. 6. µ-Opioid receptor desensitization and tolerance in locus coeruleus neurons. Top trace is an example recording of the protocol used to measure acute opioid desensitization. This is a current record of a cell voltage-clamped at 60 mV. Superfusion of [Met]enkephalin (ME; 300 nM) caused an outward current. Superfusion of ME (30 µM) for 5 min resulted in a peak response followed by a decline to ~50% of the peak. Immediately after washing the high concentration of ME, the outward current caused by ME (300 nM) increased over a period of 15-25 min to a value close to that at the beginning of the experiment. Bottom left: a concentration-response curve to normorphine in control and immediately after a desensitizing treatment with ME (30 µM, 5 min). The dose-response curve is shifted to the right, and the peak response is reduced. [From Osborne and Williams (363).] Bottom right: a similar experiment done in brain slices taken from control animals (control) and animals that were treated chronically with morphine. As was observed with the acute desensitization, chronic morphine treatment shifted the concentration-response curve to the right and depressed the maximum response. [From Christie et al. (89).] In each experiment, the response to normorphine was normalized to the outward current induced by a maximal concentration of the 2-adrenoceptor agonist UK-14304. The conclusion of these experiments was that acute desensitization as well as more long-term tolerance results in a dramatic decline in µ-opioid receptor reserve. The acute desensitization recovers to a large degree within 30 min; however, in morphine-treated animals, there was little evidence for recovery over a period of 2-6 h.

Heterologous desensitization has been described in both model systems (267) and neurons (135, 353). By definition, heterologous desensitization generalizes to receptors and transduction mechanisms distinct from the one activated by the primary ligand and as such can represent at the cellular level the compensatory and incidental adaptive mechanisms discussed below. For example, heterologous desensitization that involves changes in the activity of kinases such as PKA and PKC can affect multiple elements of signal transduction cascades including diverse receptors, G proteins, and effectors such as ion channels.

It has long been recognized that internalization is not required for desensitization to occur, but when it does, the process necessarily affects sensitivity to agonists by removing surface receptors. Pak et al. (366) demonstrated that µ-receptor desensitization was associated with a loss of binding sites on the plasma membrane. As discussed above, agonist-evoked internalization has been observed for µ-receptors in model systems (246, 517) as well as neurons (246, 455). Internalization occurs predominantly via clathrin-coated pits. Receptor phosphorylation by GRKs appears to be a critical event in internalization (64), although others have suggested that MAPK activation is critical (382).

A number of groups have examined residues in the COOH-terminal region of the µ-receptor, known to be essential for internalization via the endosomal pathway. Pak et al. (367) found that Thr-394 was the primary recognition site for G protein-coupled receptor kinases, but Thr-383 was also required for complete agonist-induced desensitization. The specificity of Thr-394 as the primary initiation site appears to be dependent on the preceding acidic amino acid stretch, because a mutant in which glutamic acid residues at 388, 391, and 393 were replaced by glutamines the receptor was not internalized. Truncated µ-receptors suggested that Ser-356 and Ser-363 were important for agonist-induced internalization of the receptor but not phosphorylation (59).

Although GRKs and endosomal mechanisms have been established to play a major role, other internalization processes have not been ruled out (337). The importance of GRK-mediated processes for acute tolerance, usually studied using morphine (or heroin), is doubtful because morphine does not induce internalization in vivo. Speculations that differences in the abilities of different opioid agonists to induce internalization are related to their addiction liability or therapeutic value in opioid management of opioid dependence are intriguing but have yet to be fully established (516, 541). It is possible that agonists employed therapeutically in dependency management, such as methadone, l--acetyl methadol (LAAM), and buprenorphine, have beneficial effects on adaptive processes because they more readily induce internalization than morphine (and heroin). As discussed above, endosomal internalization mechanisms involve dephosphorylation, resensitization, and MAPK activation which initiates nuclear signal transduction cascades that can influence downstream adaptive processes and may regulate µ-receptor function. Any such mechanisms and potential therapeutic benefits must be purely fortuitous because methadone, LAAM, and buprenorphine were developed for opioid dependency management solely on the basis of their agonist efficacy, bioavailability, safety, and plasma half-lives.

Long-term adaptive processes play a major role in the capacity of animals and humans to tolerate profoundly escalating doses of opioids over periods of weeks to months, but the mechanisms involved have not been fully resolved. It is not yet clear whether or not the early events of desensitization and internalization are necessary antecedents or perhaps contribute directly to these longer term adaptive processes. Where they have been studied, however, evidence for a direct role is lacking (see below). Such long-term adaptations presumably involve chronic functional uncoupling of µ-receptors from signaling pathways, perhaps as a consequence of counteradaptations (see below) and/or downregulation of surface receptors. Early studies of µ-receptor density in whole brain or brain regions following chronic morphine treatment almost invariably found no reduction in total number of receptor binding sites (e.g., Ref. 514). Although some studies of surface receptors in cultured cells observed downregulation following chronic morphine and other agonists (386, 537, 544), the findings in brain appeared to indicate that adaptive processes must be targeted at intracellular domains of the receptor involved in coupling with G proteins and not the density of receptors at the plasma membrane surface. Radiolabeled opioid ligands used in early studies could not readily distinguish surface from intracellular receptors and ligand binding studies are also confounded by continued occupancy of receptors, during chronic treatment. Subsequent to the cloning of µ-receptors and the development of highly selective antibodies, more direct assessment of surface receptor densities has indicated that densities of µ-receptors are indeed downregulated in some brain regions after chronic treatment with morphine (34, 277, 477). In model systems expressing truncated µ-receptors, serine residues were identified in the COOH-terminal region (Ser-355 and Ser-363) that were necessary for opioid-induced (etorphine) downregulation (5, 59).

The mechanisms of µ-receptor downregulation do not appear to substantially involve GRK-dependent mechanisms. Studies in SH-SY5Y cells have suggested that µ-receptor downregulation is blocked by nonspecific serine-threonine kinase inhibitors but that a putative GRK2 inhibitor, suramin, had no effect (130). The absence of a role for GRK phosphorylation in downregulation is consistent with similar findings in GRK2 phosphorylation-deficient ₂-adrenoceptor mutants (172). The failure of morphine to engage endosomal internalization mechanisms is also consistent with a lack of involvement of GRK because morphine produces similar downregulation as other agonists (537). Although phosphorylation has been implicated in the process of downregulation, the nature of its role remains uncertain. It is possible that receptor cycling is affected in a GRK-independent fashion or that kinase activity affects synthesis or degradation rates.

Studies of µ-receptor coupling to proximal cellular effectors, such as GTPase activity, adenylyl cyclase inhibition, inwardly rectifying potassium channels, or calcium currents have also found functional uncoupling of receptors from effectors. ³⁵S-labeled guanosine 5'-O-(3-thiotriphosphate) (GTPS) binding reflects GTPase activity and is decreased following chronic morphine treatment in brain, including LC (423, 435) and in cultured cells (130). Similar results have been reported for coupling of µ-receptors to inwardly rectifying potassium channel currents (89) and calcium channel currents (97) in LC neurons, and calcium channel currents in SH-SY5Y cells (249). Where studied, the uncoupling processes were found to be homologous. Reduced coupling efficacy could have arisen in these studies from functional uncoupling of receptors from G proteins, or a loss of surface receptors. Although the former was generally assumed on the basis of negative results from ligand binding studies in brain (see above), a reduction in surface µ-receptor density actually was found in the only study that directly addressed the issue (130). It therefore remains uncertain whether or not the functional uncoupling of µ-receptors from proximal effectors widely observed in neurons and model systems arises solely from a reduction in the density of µ-receptors in the plasma membrane.

In summary, acute desensitization, internalization, and downregulation of µ-receptors all play roles in opioid tolerance measured at the cellular and synaptic level. Although these mechanisms can explain the development of nonassociative tolerance at the cellular level, adaptive mechanisms that occur with repeated and/or continuous morphine treatment to mediate associative tolerance are probably mediated by separate mechanisms (e.g., Ref. 325). The counteradaptive mechanisms not only mediate forms of tolerance to morphine but are also involved in opioid withdrawal and dependence. The relative contributions of each process to the extent and persistence of tolerance in different physiological systems in the behaving organism have not been elucidated. The extent of tolerance is usually rather small when examined in single cells (e.g., Ref. 89) compared with tolerance in whole animals (e.g., Ref. 104). Tolerance at systems levels must involve interaction of mechanisms of tolerance at molecular, cellular, and neural network levels throughout each system, but the details of such interactions are completely unknown.

Mechanisms subsequent to receptor activation that adapt to restore function in the presence of drug mediate a second form of tolerance (222). Tolerance produced by compensation, by definition, requires the presence of opioid agonists to maintain normal function. The adaptive responses observed at the cellular, synaptic, and network levels are therefore the core of acute aspects of opioid withdrawal. As with tolerance, different processes may mediate short- and long-term and protracted compensatory changes associated with chronic opioid treatment. Very-short-term counteradaptations can be observed after only several minutes of opioid application and abate just as rapidly (e.g., Refs. 221, 143). As discussed below, long-term compensatory changes have been most thoroughly studied.

The first and best-studied example of tolerance resulting from compensation used the inhibition of adenylyl cyclase as an assay (46, 425, 426). Acutely, opioids acting on -receptors inhibited adenylyl cyclase, but in the continued presence of morphine, there was an increase (upregulation) of adenylyl cyclase activity (46, 425, 426). When agonist was removed, the compensatory increase in adenylyl cyclase activity remained. The increased adenylyl cyclase activity was taken as an example of withdrawal at the cellular level.

Since these early studies, several isoforms of adenylyl cyclase have been identified and classified into three primary groups based on sequence similarities (102, 329, 330). All these enzymes are differentially regulated by a number of messenger pathways including calcium, G_i, G_s, G/, and PKC (102, 329, 330). In addition, each isoform has a distinct anatomical distribution. Three isoforms are primarily neuronal, ACI, ACII, and ACV. The distribution of cells expressing high levels of mRNA for each subtype has a distinct pattern. Type I is found in the dentate granule cells of hippocampus, cerebellar granule cells, and cortex; type V is found almost exclusively in striatum and nucleus accumbens, and type II is more diffusely located in cortex, hippocampus, cerebellar granule cells, and substantia nigra. At the subcellular level, the distribution seems to be highly localized to synapses (330). Subunit-selective antibodies have not been used extensively, limiting the interpretations of exact composition at many synapses, although suggestions have been made based on results from in situ hybridization experiments. The high density of immunoreactive substrates found both pre- and postsynaptically place this highly regulated molecule that is sensitive to a number of second messengers in an ideal position to mediate synaptic plasticity.

The sensitivity of various isoforms of adenylyl cyclase to opioids was examined in a series of studies done in COS-7 and CHO cells (18-20). Isoforms that were acutely inhibited by opioids (I, V, VI, and VIII) were upregulated or supersensitized by chronic treatment. Adenylyl cyclase I and V are expressed at high levels in the CNS, thus encouraging speculation of a similar upregulation in vivo (see below). The results of this series of experiments were similar in ways to the early experiments in NG108-15 cells with native opioid receptors and adenylyl cyclase. The upregulation varied between two- and fivefold, required 6-10 h of treatment, recovered within 2-3 h, and was sensitive to pertussis toxin and agents that scavenged free G/ subunits. There were two significant differences. First the kinetics of the upregulation were faster in the CHO and COS-7 cells than the NG108-15 cells. This may have resulted from different expression levels of receptor and/or adenylyl cyclase. The recovery after morphine treatment in the NG108-15 cells was complete after 24 h (426). The second potentially more significant difference was that the supersensitization in CHO cells was insensitive to cycloheximide and to a dominant negative ras mutant that blocked the activation of MAPK (18, 19). In NG108-15 cells, however, a large portion of the increased adenylyl cyclase activity was blocked by cycloheximide (426). The interpretation of these two observations is completely different. The result in CHO cells indicates that activity of adenylyl cyclase is increased, by an as yet uncharacterized mechanism, whereas the work in NG108-15 cells suggests that the upregulation is dependent on new protein, which may be but is not necessarily adenylyl cyclase itself. It is important to revisit the NG108-15 cell model, particularly the effects that inhibition of the MAPK pathway may have on the upregulation of adenylyl cyclase.

Studies examining the effects of chronic morphine on adenylyl cyclase in the brain and peripheral tissues have produced mixed results (71, 125, 479, 497). Acutely, opioids produced only a small inhibition in most areas (125, 479, 497) and increased activity in some areas (72, 105, 362). The upregulation of adenylyl cyclase activity induced by chronic morphine treatment, where it has been observed, was generally <0.5-fold. Preparation of brain and peripheral tissues has problems of heterogeneous cell types, multiple receptors, and adenylyl cyclase isoforms. Expression levels of each component can also reduce the signal-to-noise ratio and thus cloud interpretation of effects. Such difficulties are inherent in biochemical assays from complex tissues. Despite the fact that the upregulation of adenylyl cyclase in cell lines and expression systems is a robust and reliable measure, a similar approach measuring the bulk production of cAMP in tissues from animals treated with morphine has not brought new insights to this adaptive mechanism. That is not to say that it does not happen and is not important. In fact, there is building evidence that it may be critically important in the local area surrounding specific synapses. There may be additional adaptations resulting from the increased adenylyl cyclase activity mediated by changes in gene expression under the regulation of CREB (346).

One disappointing and recurring observation has been the lack of any change in the regulation of potassium or calcium conductances during acute morphine withdrawal. Probably the most thoroughly studied example of the lack of an adaptive process in response to chronic treatment is the potassium conductance in the LC. Aghajanian (6) was the first to record the firing rate of LC neurons in vivo from chronically morphine-treated rats. Morphine applied systemically caused an acute inhibition of firing. After 5 days of continuous morphine treatment, the spontaneous firing had returned to control values, indicating that LC neurons were tolerant to the levels of circulating morphine. An increase in firing rate above control levels upon application of naloxone by iontophoresis was taken as a cellular sign of withdrawal. More recent experiments in brain slices showed that opioids acutely activated an inwardly rectifying potassium conductance that caused a hyperpolarization to decrease the firing rate (374, 523). In addition, LC cells have a resting potassium conductance that is inwardly rectifying, suggesting that during withdrawal a reduction of this conductance could depolarize the cell and increase excitability (523). A decrease in inwardly rectifying potassium conductance in these and other cells by other G protein-linked receptors has been demonstrated, suggesting that opioid withdrawal could increase excitability by this mechanism (343, 500). When experiments were done with slices taken from morphine-treated animals, there was no evidence of a depression in the resting inwardly rectifying potassium conductance by naloxone (89). Subsequent experiments have indicated that the increase in firing observed in vivo was largely the result of an increase in glutamate release from the excitatory inputs to the LC (8).

Calcium currents have not been extensively characterized during withdrawal. Although biochemical studies suggest an increased calcium channel density after chronic morphine treatment, these results have not been confirmed in electrophysiological studies in isolated neurons. There was no change in the calcium channel current density or relative proportions of pharmacologically isolated calcium channel subtypes in acutely isolated LC neurons taken from morphine-treated animals (97). There was a modest reduction in efficacy of coupling between µ-opioid receptors and inhibition of calcium channel currents, but no rebound during naloxone precipitated withdrawal. Thus withdrawal from morphine treatment produced no compensatory action on either potassium or calcium conductances in LC neurons.

The acute inhibition of a cation conductance by opioids has been reported in several preparations (Table 1). In addition, an increase in a cation conductance has been reported to account for the withdrawal-induced increase in excitability in the PAG (84; E. E. Bagley and M. J. Christie, unpublished data) and proposed but not directly demonstrated in LC (10, 257), and indirectly implicated on GABAergic neurons in the vicinity of the dorsal raphe (227). This cation conductanc