p38 MAPK and β-Arrestin 2 Mediate Functional Interactions between Endogenous μ-Opioid and α2A-Adrenergic Receptors in Neurons*

Formation of receptor complexes between μ-opioid and α2A-adrenergic receptors has been demonstrated in transfected cells. The functional significance and underlying mechanisms of such receptor interactions remain to be determined in neuronal systems. We examined functional interactions between endogenous μ and α2A receptors in mouse dorsal root ganglion neurons. Acute application of the μ agonist [d-Ala2,N-MePhe4, Gly-ol5]enkephalin (DAMGO) or the α2 agonist clonidine inhibited voltage-gated Ca2+ currents in these neurons. Prolonged treatment with either DAMGO or clonidine induced a mutual cross-desensitization between μ and α2A receptor-mediated current inhibition. The cross-desensitization was closely associated with simultaneous internalization of μ and α2A receptors. Morphine, a μ agonist triggering little μ receptor endocytosis, induced neither cross-desensitization nor internalization of α2A receptors. Furthermore, inhibition of p38 MAPK prevented the cross-desensitization as well as cointernalization of μ and α2A receptors. Changes in receptor trafficking profiles suggested that p38 MAPK activity was required for initiating μ receptor internalization and maintaining possible μ-α2A association during their cointernalization. Finally, the μ-α2A cross-desensitization was absent in dorsal root ganglion neurons lacking β-arrestin 2. These findings demonstrated p38 MAPK- and β-arrestin 2-dependent cross-regulation between neuronal μ and α2A receptors. By promoting receptor cross-desensitization and cointernalization, such functional interactions may serve as negative feedback mechanisms triggered by prolonged agonist exposure to modulate the signaling of functionally related G protein-coupled receptors.

G protein-coupled receptors (GPCRs) 2 interact with each other through formation of receptor complexes, including homo-or heterodimers and possibly higher order oligomers (1)(2)(3)(4). Heterooligomerization of GPCRs has been shown to enable cross-regulation between different receptor systems, resulting in various changes in receptor binding, signaling, and trafficking. For example, dimerization of -opioid and NK1 (neurokinin type 1) receptors in HEK-293 cells promotes agonist-induced cross-phosphorylation and cointernalization of the two receptors, whereas receptor binding and functional coupling are relatively unaffected (5). A similar pattern has been observed in cells expressing heterodimers of ␦-opioid and ␤ 2 -adrenergic receptors (6). Formation ofand ␦-opioid receptor complexes alters receptor properties, leading to synergistic enhancement of receptor binding and signaling by and ␦ ligands (7). The -␦ heterodimer may form as early as in the endoplasmic reticulum during receptor processing and allows co-trafficking of the two receptors (8). Controversial evidence exists, however, for agonist-induced, separate endocytosis of and ␦ receptors (9). Although these studies and many others have underscored the dynamic nature and divergent roles of receptor heterodimerization in GPCR modulation, the molecular basis and regulatory mechanisms for such interactions remain to be elucidated. In particular, most studies addressing this issue have been conducted in heterologous cells or in systems where receptors are overexpressed, which may lead to interactions nonexistent with endogenously expressed receptors. Further studies are necessary to identify and characterize interactions between naturally existing GPCRs in primary neurons.
We examined interactions between endogenous -opioid and ␣ 2A -adrenergic receptors in mouse dorsal root ganglion (DRG) neurons. Both and ␣ 2A receptors are coupled to G i and G o proteins and induce similar cellular responses, such as inhibition of voltage-gated Ca 2ϩ channels and activation of inwardly rectifying potassium channels. These cellular effects can lead to presynaptic inhibition of neurotransmitter release or hyperpolarization of postsynaptic neurons. Both are crucial mechanisms for opioid and adrenergic modulation of nociception. A functional synergy between the two systems has been demonstrated in vivo, evidenced by potentiation of morphine analgesia (10,11) and alleviation of opiate withdrawal (12) by the ␣ 2 -adrenergic agonist clonidine. Studies in transgenic mice lacking functional ␣ 2A receptors further indicate that the ␣ 2A receptor is the principal subtype mediating ␣ 2 agonist-induced analgesia at the spinal level (13) and responsible for the synergistic potentiation of morphine analgesia (14). The exact mech-anisms for this adrenergic-opioid synergy, however, remain unclear. Recently, the -␣ 2A receptor heterodimers have been detected in HEK-293 cells co-expressing both receptors (15,16). It is of great interest to further explore whether heterodimerization of and ␣ 2A receptors serves as a novel mechanism coordinating the function of both receptors in pain-processing pathways. Here we report that endogenous and ␣ 2A receptors interact in DRG sensory neurons via p38 MAPK-and ␤-arrestin 2-dependent mechanisms, which promote agonistselective cross-regulation of receptor signaling and internalization.
Electrophysiological Recordings-The voltage-gated Ca 2ϩ currents were recorded from DRG neurons with 15-30-m diameters under whole-cell voltage clamp conditions, as described (17). Cells were perfused with an external solution containing 10 mM CaCl 2 , 130 mM tetraethylammonium chloride, 5 mM HEPES, 25 mM D-glucose, and 0.25 M tetrodotoxin at pH 7.35. The patch electrode was filled with an internal solution composed of 105 mM CsCl, 40 mM HEPES, 5 mM D-glucose, 2.5 mM MgCl 2 , 10 mM EGTA, 2 mM Mg-ATP, and 0.5 mM GTP at pH 7.2. Ca 2ϩ currents were evoked every 10 s by 40-ms voltage steps from Ϫ80 to ϩ10 mV using an Axopatch 200A patch clamp amplifier. Capacitance and series resistance were corrected with the compensation circuitry on the amplifier. Series resistance was compensated by 80 -90%. Leak currents were subtracted using a P/6 protocol. Recorded signals were acquired and analyzed using Axon pCLAMP version 8.0 software (Axon Instruments). The amplitude of peak Ca 2ϩ currents was determined using the peak detect feature of the software.
Drug Application and Desensitization Protocols-The and ␣ 2 receptor ligands (Sigma) were prepared as stock solutions in water, diluted with external solution to the final concentration for acute bath application, or added into culture medium for pretreatment. Various kinase inhibitors (Sigma) were dissolved in DMSO and diluted with culture medium for pretreatment with a final DMSO concentration of 0.1%. Cells pretreated with the medium containing 0.1% DMSO served as vehicle controls in these experiments. In additional control experiments, DAMGO-or clonidine-induced current inhibition was compared between untreated cells and cells pretreated with 0.1% DMSO for 4 h. No significant differences were observed between the two treatments (data not shown).
During recording, the external solution was continuously applied at 2 ml/min through a 0.5-ml recording chamber carrying the culture coverslip. After establishing a stable base line, the or ␣ 2 agonist was applied for up to 1 min to observe the maximal change in Ca 2ϩ currents. Agonist-induced current inhibition was measured as the maximal reduction in the peak current amplitude during drug perfusion and expressed as percentage changes from the base-line level. The voltage dependence of agonist effect was assessed using a prepulse facilitation (PPF) protocol consisting of two normal test pulses (P1 and P2) and in between a strong depolarizing prepulse (Ϫ80 to ϩ80 mV, 40 ms) delivered 10 ms before P2. The PPF was expressed by the amplitude ratio of currents activated by the two test pulses (P2/P1). To induce chronic desensitization, DRG cultures were pretreated with either or ␣ 2 agonist for 4 h. After extensive washing, whole-cell recording was performed in pretreated cells, and the acute inhibitory effect of or ␣ 2 agonist on Ca 2ϩ currents was measured during a brief perfusion (0.5-1 min). The extent of desensitization was determined by the percentage reduction of the agonist effect in pretreated neurons relative to untreated neurons.
Immunocytochemical Analysis and Fluorescence Confocal Microscopy-Cellular distribution of and ␣ 2A receptors in DRG cultures was determined by immunocytochemical double labeling. After drug treatment, the cultures were fixed with 4% paraformaldehyde for 10 min, washed in phosphate-buffered saline (PBS), permeabilized, and preblocked with PBS containing 10% normal donkey serum and 0.1% Triton X-100 for 2 h at room temperature. The cultures were then incubated overnight at 4°C with the primary antibodies against the C-terminal sequence of the receptor (20) or of the ␣ 2A receptor (21) (sc-1478 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA); ab45871 (Abcam)). After washing, the cells were further incubated for 2 h with Alexa Fluor 488-labeled donkey anti-rabbit IgG for visualization of receptors and with biotinylated antigoat IgG (Calbiochem) and Cy3-conjugated streptavidin (Calbiochem) to detect ␣ 2A receptors. Control samples were prepared in the absence of the primary or secondary antibodies. The specificity of the anti-␣ 2A antibody was further confirmed by the absence of ␣ 2A receptor labeling in neurons from the ␣ 2A Ϫ/Ϫ mice (see Fig. 4B). The images of neurons with receptor labeling were acquired using a Leica TCS-SP confocal laserscanning microscope and processed either as a single scan or as maximum intensity projections of multiple scans taken at successive 1-m depths.
Flow Cytometric Measurement of Surface Receptors-Internalization of receptors was further accessed in living neurons by quantifying cell surface receptors with flow cytometry as described previously (22). DRG cultures were treated with clonidine, DAMGO, morphine, or control medium for 4 h at 37°C. Monensin (10 M) was added during the treatment to block recycling of internalized receptors in all experiments. DRG cells were then harvested in PBS containing 2 mM EDTA, spun at 300 ϫ g for 5 min, and resuspended with PBS containing 1% normal goat serum and 0.1% NaN 3 . Cell surface receptors were labeled at 4°C for 30 min with a polyclonal antibody against the third extracellular loop of the receptor (Chemicon International) (23) and visualized with Alexa Fluor 488-labeled donkey anti-rabbit IgG at 4°C for another 30 min. Cell surface immunofluorescence was measured with a FACScan flow cytometer at 5,000 -10,000 cells/sample. Data were acquired and analyzed using Cell Quest version 3.0.1 (BD Bioscience). The loss of cell surface receptors after agonist exposure was quantified by the reduction in the proportion of cells expressing detectable surface receptors (-positive cells) and in the density of surface receptors of -positive cells reflected by their mean fluorescence intensity. Nonspecific background fluorescence was determined in control samples processed without the primary antibody and was subtracted to obtain mean relative fluorescence intensity of experimental samples.
Measurement of Phospho-p38 MAPK-The cellular level of phospho-p38 MAPK was measured with flow cytometry (24), using a polyclonal antibody recognizing p38 MAPK dually phosphorylated at Thr 180 and Tyr 182 (Cell Signaling Technology) (25). After drug treatment, DRG cells were harvested from culture plates with PBS buffer containing 4 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 0.1 mM calyculin A, 1 g/ml leupeptin, and protease inhibitor mixture tablet (Sigma) (26). The cells were fixed in 2% formaldehyde for 10 min at room temperature, permeabilized in ice-cold 90% methanol for 30 min, incubated with the primary antibody for 60 min at room temperature, and labeled with an Alexa Fluor 488-conjugated second antibody for another 60 min. In between these steps, the cells were washed, centrifuged at 450 ϫ g for 5 min, and resuspended with fresh PBS. After the final wash, the fluorescence signal of the cells was analyzed with a FACScan flow cytometer at 5,000 cells/sample, and the data were processed with CellQuest software.
Statistical Analysis-All data are presented as means Ϯ S.E. One-way analysis of variance was applied for overall statistical significance across multiple group means, followed by the Bonferroni post hoc test for pairwise comparisons. Statistical significance was defined as p Ͻ 0.05.
Inhibition of neuronal Ca 2ϩ channels by GPCRs is mediated by both voltage-dependent and voltage-independent mechanisms. The voltage-dependent inhibition requires direct binding of G-protein ␤␥ subunits (G␤␥) to the Ca 2ϩ channel, a process reversible by strong depolarization. To determine whether this direct G␤␥-channel interaction was affected by clonidine, we examined voltage dependence of clonidine action using the PPF protocol ( Fig. 1, C and D). In cells acutely perfused with clonidine (10 M, 1 min), the depolarizing prepulse produced little relief of inhibition, with a P2/P1 ratio similar to that under basal conditions (1.06 Ϯ 0.04 versus 0.99 Ϯ 0.02, n ϭ 5 for each group, p Ͼ 0.05). Clonidine pretreatment for 4 h was also without effect on the P2/P1 ratio subsequently tested either in control perfusion medium (0.97 Ϯ 0.02, n ϭ 8) or during acute clonidine test (1.06 Ϯ 0.03, n ϭ 8). These results suggested that at the concentration we used, the effect of clonidine was primarily mediated by voltage-independent mechanisms. We then measured PPF induced by intracellular application of 0.1 mM GTP␥S that can directly activate G proteins and release G␤␥ subunits. GTP␥S-induced PPF was not significantly different between untreated control cells and clonidine-pretreated cells (1.68 Ϯ 0.23 versus 1.68 Ϯ 0.18 at 8 min after the application, n ϭ 12 for each group, p Ͼ 0.05) (Fig. 1, C and D). This confirmed that clonidine-induced desensitization was not associated with a diminished cell capacity for voltage-dependent G␤␥-Ca 2ϩ channel interactions.
Since clonidine is relatively nonselective to different subtypes of the ␣ 2 receptors, we examined the role of ␣ 2A receptors in the acute and chronic effects of clonidine. Acute application of 10 M clonidine induced negligible current inhibition in DRG neurons derived from the ␣ 2A Ϫ/Ϫ mice (2.5 Ϯ 1.1%, n ϭ 6, p Ͼ 0.05). In contrast to the wild-type neurons, the ␣ 2A Ϫ/Ϫ neurons also displayed no cross-desensitization to DAMGO following clonidine pretreatment (Fig. 4A). These results suggested that in our preparation, the ␣ 2A receptor was the major subtype responsible for clonidine-induced Ca 2ϩ current inhibition and its cross-desensitization to DAMGO.
Agonist-induced Cointernalization of and ␣ 2A Receptors-Next we determined whether the mutual cross-desensitization between and ␣ 2A receptors was associated with changes in agonist-induced receptor internalization. Immunohistochemi-cal double labeling and confocal microscopy showed that a substantial portion of and ␣ 2A receptors were colocalized on the cell membrane throughout the cell body and processes of cultured DRG neurons under basal conditions (Fig. 3A). Incubation with DAMGO for 30 min to 4 h induced simultaneous internalization of both receptors, which was blocked by coincubation with the antagonist CTAP (Fig. 3C). Clonidine also elicited internalization of both and ␣ 2A receptors, reversible by the ␣ 2 antagonist yohimbine (Fig. 3D). In contrast, morphine treatment up to 4 h failed to internalize either or ␣ 2A receptors (Fig. 3B). Furthermore, clonidine did not induce receptor endocytosis in ␣ 2A Ϫ/Ϫ neurons (Fig. 4B). Thus, similar to the cross-desensitization, the cointernalization was agonistselective and dependent upon the presence of functional and ␣ 2A receptors.
Fluorescence flow cytometry was then conducted in living DRG neurons to quantify agonist-induced receptor internalization. Surface receptors were detected in 77.7 Ϯ 3% of control cells. DAMGO treatment for 4 h reduced the portion of  5B). Furthermore, the intensity of surface receptor labeling in the -positive cells decreased by 43.1 Ϯ 7.8% (p Ͻ 0.05) following DAMGO treatment (Fig. 5A). These changes confirmed loss of surface receptors due to DAMGO-induced receptor internalization. Consistent with previous reports in heterologous cells (20,27), morphine was much less effective in triggering receptor internalization, causing little change in surface receptor expression after a 4-h treatment. Importantly, clonidine pretreatment reduced the portion of -positive cells by 16.7 Ϯ 5.7% and the level of surface receptors by 30.8 Ϯ 10.9% (p Ͻ 0.05 for both compared with the controls), and both effects were blocked by yohimbine (Fig. 5).

p38 MAPK and ␤-Arrestin 2 Mediate -␣ 2A Interactions
In a separate set of experiments, DRG neurons were pretreated with DAMGO (1 M, 4 h) in the presence of one of the kinase inhibitors and tested for acute responses to clonidine. Blocking p38 MAPK activity with PD169316 (5 and 20 M) during pretreatment attenuated DAMGO-induced cross-desensitization in a dose-dependent manner (Fig. 6B). The responses to subsequent clonidine test were significantly greater in cells co-treated with 20 M PD169316 compared with those treated with DAMGO alone (13.9 Ϯ 2.4% versus 5.7 Ϯ 0.8%, n ϭ 17 and 18, p Ͻ 0.01). Co-treatment with another specific p38 inhibitor SB203580 (20 M) (29) also significantly increased clonidine responses (Fig.  6B). Again, the inhibitors for phosphoinositide 3-kinase, ERK, protein kinase A, and protein kinase C were unable to prevent the cross-desensitization. Clonidine-induced current inhibition ranged from 4.0 to 7.5% in cells co-treated with one of these inhibitors, indistinguishable from those treated with DAMGO alone. None of the inhibitors examined above elicited significant changes in basal Ca 2ϩ currents when applied acutely in the absence of or ␣ 2 agonist (2.2-4.9% current inhibition, n ϭ 4 -6, p Ͼ 0.05). In addition, pretreatment with PD169316 (20 M) alone for 4 h did not affect the acute effect of DAMGO or clonidine (49.0 Ϯ 4.1 and 19.8 Ϯ 3.9%, n ϭ 11 and 9, p Ͼ 0.05 compared with vehicle-treated controls). Thus, blockade of the crossdesensitization by p38 MAPK inhibitors was not due to any direct effect of the inhibitors per se on Ca 2ϩ currents or to interference of acute DAMGO or clonidine action. Together, these results suggested that p38 MAPK activity was necessary for the -␣ 2 interaction leading to the mutual cross-desensitization.
Since the cross-desensitization was closely associated with cointernalization of and ␣ 2A receptors, we investigated whether reversal of the cross-desensitization by PD169316 was accompanied by blockade of receptor cointernalization. The p38 MAPK inhibitor was found to completely block internalization of both and ␣ 2A receptors in DAMGO-treated cells (Fig. 6C). Interestingly, PD169316 selectively blocked internalization of but not ␣ 2A receptors in clonidine-treated cells. This effect led to separation of the two receptors during clonidine treatment, with ␣ 2A receptors being internalized but receptors remaining on the plasma membrane of DRG neurons (Fig. 6, C and D). These results suggested that activation of p38 MAPK was necessary for receptor endocytosis as well as for supporting cointernalization of and ␣ 2A receptors.
Activation of p38 MAPK by DAMGO and Clonidine, but Not by Morphine-To obtain direct biochemical evidence for p38 activation by and ␣ 2A agonists, we measured p38 MAPK phosphorylation in DRG neurons. In a time course analysis, DAMGO treatment (1 M) induced rapid and sustained activa- In agreement with differential activation of p38 MAPK by the two opioid agonists, PD169316 selectively blocked DAMGObut not morphine-induced homologous desensitization (Fig.  7B). Acute DAMGO responses in cells co-treated with PD169316 (41.4 Ϯ 7.7%, n ϭ 9) were significantly greater than those pretreated with DAMGO alone (18.4 Ϯ 3.8%, n ϭ 7, p Ͻ 0.05) and comparable with untreated controls (49.4 Ϯ 7.5%, n ϭ 11, p Ͼ 0.05). In contrast, morphine-induced current inhibition was similar between cells treated with morphine alone and those co-treated with PD169316 (22.3 Ϯ 2.1% versus 28.3 Ϯ 2.9%, n ϭ 19 and 21, p Ͼ 0.05). Furthermore, although it blocked DAMGO-induced cross-desensitization to clonidine, PD169316 did not prevent homologously induced clonidine desensitization (Fig. 7C). This indicated that p38 MAPK activity was also differentially involved in the homologous versus heterologous desensitization of ␣ 2A receptors.

Cross-regulation of and ␣ 2A Receptor Signaling in Neurons-
Although formation of and ␣ 2A receptor complexes has been convincingly demonstrated in cell lines and transfected neurons (15), functional significance and regulatory mechanisms for the -␣ 2A interaction remain to be clarified. The initial

p38 MAPK and ␤-Arrestin 2 Mediate -␣ 2A Interactions
report showed that heterodimerization of and ␣ 2A receptors in HEK293 cells enhanced receptor signaling in response to acute application of morphine or clonidine (15). A recent study, however, demonstrates a direct conformational cross-talk between and ␣ 2A receptors within the heterocomplex that allows rapid inaction of one receptor by the other with subsecond kinetics (31). In locus coeruleus neurons naturally expressing high levels of ␣ 2A and receptors, analysis of the effect of co-applied and ␣ 2A agonists reveals no functional interactions between the two receptors (32). These findings raised the questions as to whether and how naturally existing and ␣ 2A receptors in neurons could interact with each other to regulate cellular function. The present study demonstrated strong functional interactions between endogenous and ␣ 2A receptors in sensory neurons. Such interactions required p38 MAPK activity and ␤-arrestin 2 and promoted receptor co-trafficking and cross desensitization.
Desensitization of GPCR signaling involves modifications at the receptor level and in downstream signal transduction pathways. We previously reported that chronic homologous DAMGO desensitization in DRG neurons was partially mediated by phosphoinositide 3-kinase-and ERK-mediated changes in voltage-dependent G␤␥-Ca 2ϩ channel interactions (17). In the present study, however, neither acute nor prolonged clonidine treatment altered PPF, a direct measure of voltagedependent G␤␥ effect on Ca 2ϩ channels. The -␣ 2A cross-desensitization was also unaffected by the selective inhibitors for phosphoinositide 3-kinase or ERK. These results suggest that modification of G␤␥-Ca 2ϩ channel interactions by these two kinases did not play a significant role in the cross-desensitization. Another scenario for the cross-desensitization to occur at the post-receptor level would be desensitization of signaling pathways shared by both receptors. Our results did not support this possibility either. p38 MAPK and ␤-arrestin 2 differentially regulated homologous and ␣ 2A desensitization, suggesting that divergent rather than common signaling pathways were engaged by and ␣ 2A receptors.
Receptor Cointernalization Contributes to -␣ 2A Crossdesensitization-An important feature of the cross-desensitization was its close association with agonist-induced cointernalization of and ␣ 2A receptors. It is well documented that both and ␣ 2A receptors undergo agonist-induced rapid endocytosis via clathrin coated-pits, a process regulated by receptor phosphorylation and binding with nonvisual ␤-arrestins (27, p38 MAPK and ␤-Arrestin 2 Mediate -␣ 2A Interactions MARCH 6, 2009 • VOLUME 284 • NUMBER 10 33, 34). How this event regulates GPCR signaling, especially in the case of opioid desensitization, has been a subject of intense investigation (35)(36)(37). Several studies in primary neurons and AtT20 cells indicate that receptor internalization does not contribute to rapid desensitization of receptors coupled to voltage-gated Ca 2ϩ channels (38,39) or inwardly rectifying potassium channels (40) when measured on the time scale of several seconds to minutes. Evidence exists, however, that chronic opioid desensitization developed in hours can be greatly affected by receptor internalization and recycling (41). Continued internalization of receptors during prolonged agonist treatment can attenuate opioid responses via physical removal of the receptor from cell surface, but it also can promote receptor dephosphorylation and recycling. When rapid recycling occurs, the internalization is considered an important means for receptor resensitization, effectively reducing the extent of apparent desensitization (41,42). Alternatively, if internalized receptors are trapped in endosomes, significant loss of surface receptors can occur, leading to enhancement of desensitization (43). Extended or repeated agonist exposure can also target internalized receptor to lysosomes for degradation, causing receptor down-regulation and more persistent signaling reduction (44,45). Therefore, the functional consequence of receptor internalization may vary considerably in different model systems, depending upon the rate and extent of endocytosis as well as its coupling with distinct intracellular sorting pathways that determine the postendocytic fate of the receptor.
Receptor oligomerization has added a new dynamic to the complex relationship between internalization and desensitization. Studies in HEK293 cells show that formation of heterocomplexes between receptors and other GPCRs often alters receptor trafficking profiles. In many cases, co-expressed receptors are both internalized in response to a single selective agonist (5,6,8), suggesting that ligand-activated receptor may "drag" another receptor in the same complex to the endocytic pathway (46). Similarly, we demonstrated that and ␣ 2A receptors colocalized on the plasma membrane of untreated DRG neurons and underwent simultaneous internalization when either receptor was activated. These findings were in agreement with the presence of -␣ 2A complexes. Nevertheless, a fraction of and ␣ 2A receptors were likely to exist as homodimers or monomers (4,8), as indicated by the less than complete colocalization of the two receptors in DRG neurons and a relatively low percentage of receptors cointernalized by clonidine measured with flow cytometry.
Importantly, the propensity of and ␣ 2A agonists to promote cross-desensitization was closely related to their ability to induce receptor cointernalization. Exposure to DAMGO or clonidine induced both cointernalization and cross-desensitization, whereas morphine treatment resulted in neither. Furthermore, blocking -␣ 2A cointernalization with p38 MAPK inhibitors effectively prevented the cross-desensitization. These findings strongly supported a crucial role of receptor cointernalization in -␣ 2A cross desensitization. It is conceivable that a substantial portion of and ␣ 2A receptors coexist on cell surface as receptor complexes and internalize together via p38 MAPK-dependent mechanisms. During chronic agonist treatment, receptor cointernalization may be coupled with delayed recycling, which reduces cell surface receptors and attenuates signaling. Indeed, receptors are known to traffic through both early and late sorting endosomes in DRG neurons, two sorting pathways differing significantly in the rate of recycling (39). If interactions with ␣ 2A receptors promote receptors to traffic through the slower late sorting pathway, p38 MAPK and ␤-Arrestin 2 Are Key Regulators of -␣ 2A Cross-regulation-We observed strong activation of p38 MAPK by DAMGO and clonidine. The enhanced p38 MAPK activity may play two different roles in the cross-regulation of and ␣ 2A receptors. First, activation of p38 MAPK is known to facilitate receptor internalization by enhancing the function of endocytic machinery regulated by the small GTPase Rab5 (47,48). By driving receptor endocytosis, p38 MAPK activity may trigger simultaneous internalization of those ␣ 2A receptors that are directly or indirectly associated with receptors. This scenario can well explain our findings that inhibition of p38 MAPK effectively blocked DAMGO-initiated internalization of both and ␣ 2A receptors. Second, our results indicated that p38 activity was not required for clonidine-initiated ␣ 2A receptor internalization but was necessary for co-trafficking of receptors with activated ␣ 2A receptors. Thus, activation of p38 MAPK may be essential for maintaining -␣ 2A association during the cointernalization.
␤-arrestins act as scaffold proteins to regulate spatial distribution and activity of MAPK cascades (49) and that activation of p38 MAPK by GPCRs requires the presence of ␤-arrestin isoforms (50 -52). Thus, ␤-arrestin 2 may regulate the cross-desensitization via its control over p38 MAPK. Such a serial pathway could well explain the requirement of both molecules for the cross-desensitization. Our data also showed that ␤-arrestin and p38 MAPK clearly differed in regulating the homologous desensitization. ␤-arrestin 2 but not p38 MAPK was required for clonidine-induced ␣ 2A desensitization, and their roles were reversed for DAMGO-induced desensitization. One possibility is that homologous desensitization was primarily mediated by endocytosis of or ␣ 2A homodimers and monomers, which have distinct requirement for p38 MAPK and ␤-arrestins as compared with -␣ 2A heterodimers.
In untreated ␤-arr2 Ϫ/Ϫ neurons, we observed less current inhibition by DAMGO or clonidine compared with the wildtype controls. The reduced agonist effect was reported previously in these neurons and explained by decreased constitutive internalization and recycling of receptors that are constitutively coupled with Ca 2ϩ channels. Such receptors remain on the cell membrane, reducing the pool of receptors available for ligand activation (30). A similar reduction in constitutive recycling of ␣ 2A receptors may be responsible for decreased clonidine action in ␤-arr2 Ϫ/Ϫ neurons. Such changes, however, would not account for the lack of the cross-desensitization in ␤-arr2 Ϫ/Ϫ neurons, since DAMGO-induced desensitization remained intact in these neurons.
The nonvisual ␤-arrestins (1 and 2) are multifunctional proteins regulating diverse cellular functions in addition to their best-recognized roles in initiating GPCR internalization. Heterodimerization of receptors with other GPCRs can alter receptor interaction with ␤-arrestin 2, leading to delayed recycling of cointernalized receptors (5) or a shift in the activation pattern of ERK pathways (53). Therefore, the absence of -␣ 2A cross-desensitization in ␤-arr2 Ϫ/Ϫ neurons could be the result of a lack of cointernalization, altered recycling, or activation of specific signaling cascades independent of internalization. A possible model congruent with our data and these recent findings is ␤-arrestin 2-dependent formation of a macromolecular signaling complex. The complex contains both and ␣ 2A receptors, either present as heterooligomers or indirectly associated with each other through binding with ␤-arrestins.
Receptor activation leads to recruitment of specific signaling pathways, such as p38 MAPK, to the complex in an agonistselective manner, which in turn modulates receptor trafficking and desensitization. Lack of ␤-arrestin 2 prevents the formation of the complex or destabilizes it, thus disrupting -␣ 2A crossregulation.
In summary, functional interactions between neuronal and ␣ 2A receptors can lead to mutual cross-desensitization and receptor cointernalization. p38 MAPK and ␤-arrestin 2 serve as two key regulators of such interactions. These findings provide new insight into the complex relationship between opioid and adrenergic systems in the pain processing pathways and functional significance of GPCR signaling complexes.