Selectivity of δ- and κ-Opioid Ligands Depends on the Route of Central Administration in Mice

A variety of G protein-coupled receptors (GPCRs) are known to associate as heterodimers in cultured cells (Milligan, 2004; Terrilon and Bouvier, 2004; Bulenger et al., 2005; Prinster et al., 2005). The effect of heterodimerization may be manifested in a number of ways, including changes in trafficking, function, binding, and pharmacological selectivity. Opioid receptors are among those receptors in the rhodopsin family of GCPRs that have been reported to undergo heterodimerization (Prinster et al., 2005). Studies with coexpressed δ- and κ-opioid receptors in cultured cells have established they are organized as heterodimers and have led to the suggestion that the putative κ-2 subtype may be a heterodimerized κ receptor (Jordan and Devi, 1999). The putative subtypes of δ- and κ-opioid receptors were originally proposed based on differential pharmacological selectivity in vivo (Horan et al., 1991; Jiang et al., 1991; Sofuoglu et al., 1991).

More recently, studies using selective opioid ligands as tools to target δ-1 and κ-2 putative subtypes in the mouse spinal cord have suggested they arise from their organization as δ-κ heterodimers (Portoghese and Lunzer, 2003; Waldhoer et al., 2005). The physical association of these spinal receptors has been demonstrated in vivo using a specifically designed bivalent ligand, KDN-21, that is capable of bridging δ-κ heterodimers (Bhushan et al., 2004). Significantly, KDN-21 possessed different spinal and supraspinal selectivity profiles, presumably because there seems to be no δ-κ heterodimers in the brain that mediate antinociception. Another study with the spinally selective opioid agonist 6′GNT1, which selectively targets δ-κ heterodimers, produced potent analgesia only by the spinal route of administration, consistent with the localization of δ-κ heterodimers in the spinal cord but not in the brain (Waldhoer et al., 2005).

Allosteric interactions have been proposed for δ and κ receptors organized as heterodimers in the spinal cord and in cultured cells. This was attributed to the antagonism of [d-Pen²,d-Pen⁵]-enkephalin (DPDPE) by the κ-opioid antagonist norbinaltorphimine (norBNI), at its neighboring receptor in a δ-κ heterodimer in the mouse spinal cord, and to the ability of norBNI to enhance the binding of [³H]naltrindole in human embryonic kidney 293 cells transfected with δ- and κ-opioid receptors (Portoghese and Lunzer, 2003; Xie et al., 2005).

These studies have suggested that ligands known to be highly selective for a homogeneous population of opioid receptors may display different pharmacological or binding properties for heterodimeric opioid receptors. This distinction has also been observed for coexpressed δ/κ- and δ/μ-heterodimeric receptors (Jordan and Devi, 1999; George et al., 2000). Consequently, it is possible that the pharmacological selectivity of an agonist or antagonist may differ from tissue to tissue depending on the organization of receptors.

In view of the aforementioned considerations concerning the reliability of conclusions drawn from the use of selective opioid ligands, we have evaluated both i.t. and i.c.v. selectivity of several widely used, pharmacologically selective opioid antagonists in mice to compare selectivity as a function of route of administration. Here, we report on the i.t. and i.c.v. antagonist ED₅₀ values of frequently used selective antagonist ligands used as research tools in vivo. The results of this study have revealed that the pharmacological selectivity of an antagonist in mice depends on the route of administration and is therefore a function of the tissue (e.g., spinal cord versus brain) targeted by the ligand.

Materials and Methods

Animals. Male ICR mice (15–20 g; Harlan, Madison, WI), were housed in groups of 5 to 10 in a temperature- and humidity-controlled environment with unlimited access to food and water. They were maintained on a 12-h light/dark cycle. All experiments were approved by the Institutional Animal Care and Use Committee of the University of Minnesota (Minneapolis, MN).

Analgesic Studies. Antinociception was measured using the modified radiant heat tail-flick test (Tulunay and Takemori, 1974). In brief, a radiant heat source was applied to the dorsal side of the tail, and the latency to flick away from the heat source was recorded. The data were made quantal by designating a positive antinociceptive response of an animal as those that increased their latency to tail-flick (after drug treatment) by at least 3 S.D.s above the mean of the baseline latency of the whole group (Tallarida, 2000). The light source was manually turned off if the mouse did not flick its tail after the 3 S.D. criteria for a positive response. At least three groups of eight to 10 mice were used for each drug paradigm, and each mouse was used only once. ED₅₀ values and 95% confidence intervals were calculated by using the parallel line assay (Finney, 1964).

Drugs. Benzylidenenaltrexone (BNTX), naltriben (NTB), naltrindole (NTI), and norBNI were synthesized as described previously (Portoghese et al., 1988, 1991, 1992; Miyamoto et al., 1993). DPDPE (cyclic), U50488, and H-Tyr-d-Ala-Phe-Glu-Val-Val-Gly-NH (deltorphin II) were gifts from the National Institute on Drug Abuse (Szmuszkovicz and Von Voigtlander, 1982; Mosberg et al., 1983; Erspamer et al., 1989). Bremazocine (Simonin et al., 2001) was provided by Ping-Yee Law (Department of Pharmacology, University of Minnesota).

Experimental Protocols. The antagonist ED₅₀ value was determined by challenging an ED₈₀ or ED₉₀ dose of the selected agonist with graded doses of the antagonist, so that the data points in the regression either flanked or were at the antagonist ED₅₀ value. All solutions were dissolved in distilled water. Controls with only distilled water showed no antinociception. All drugs were administered ina5-μl volume in conscious mice according to the method of Hylden and Wilcox (1980) for i.t. injections and Haley and McCormick (1957) for i.c.v. injections. The drugs were administered so that the antagonist and agonist effects would peak simultaneously. The i.t. doses and peak times used for agonists were as follows: DPDPE, 12 nmol/mouse (10 min); deltorphin II, 6 nmol/mouse (10 min); U50488, 40 nmol/mouse (10 min); and bremazocine, 0.1 nmol/mouse (20 min). Agonist doses and peak times for i.c.v. administration were as follows: DPDPE, 12 nmol/mouse (20 min); deltorphin II, 12 nmol/mouse (10 min); U50488, 90 nmol/mouse (10 min); and bremazocine, 0.5 nmol/mouse (10 min). The peak times for the antagonists were as follows: BNTX and NTB, both i.t. and i.c.v, 10 min; and NTI and norBNI, both 20 min.

Results

Routes of Administration. The i.t. and i.c.v. ED₅₀ values of pharmacologically selective δ (DPDPE and deltorphin II) and κ (U50488 and bremazocine) agonists are displayed in Table 1. It is noteworthy that, with the exception of U50488, there were no significant potency differences between i.t. and i.c.v. administration. In this regard, U50488 was more potent by the i.t. route by a factor of ∼2.

The antagonist ED₅₀ values expressed as δ-1/δ-2 and κ-1/κ-2 selectivity ratios of the antagonists were generally dependent on the route of administration, and in all but one case (i.c.v. δ-1/δ-2 ratio for BNTX), they were significantly different from unity (Table 2). It should be noted that the δ-1/δ-2 and κ-1/κ-2 selectivity ratios in Table 2 were obtained from values across rows, rather than from columns of data.

In contrast to the agonist data, the i.c.v./i.t. antagonist potency ratios (Table 3) frequently differed from unity compared with the agonist potency ratios. The norBNI i.c.v./i.t. antagonist potency ratios for DPDPE, U50488, and bremazocine were in the 6 to 40 range. BNTX exhibited an i.c.v./i.t. antagonist ratio of 4 for DPDPE, whereas deltorphin II, U50488, and bremazocine were antagonized to the same degree via the two routes. The antagonist potency ratios for NTB were significantly different from unity for DPDPE, U50488, and bremazocine, with values of 2.1, 0.1, and 0.3, respectively; no significant difference was seen for deltorphin II. With NTI the i.c.v./i.t. antagonist potency ratios were significantly different for deltorphin II (7.6) and bremazocine (0.4); no differences in i.c.v./i.t. potency ratio was observed for the antagonism DPDPE and U50488. The antagonism curves for the i.t. and i.c.v. routes of administration were essentially linear within the dose ranges of the antagonists used (Fig. 1)

Norbinaltorphimine. By the i.t. route, the κ antagonist norBNI, selectively antagonized bremazocine (κ-2). The antagonism was ∼12 times greater relative to that of U50488 (κ-1), suggesting that norBNI functions as a selective κ-2 antagonist in the cord. The 40-fold greater antagonism of i.t. bremazocine over the i.c.v. route by norBNI may reflect divergent phenotypic κ receptors in the cord versus the brain. Significantly, i.t. norBNI more potently antagonized DPDPE (δ-1) over deltorphin II (δ-2) by a factor of 8 (δ-1/δ-2 = 0.13), which is consistent with earlier proposals for allosteric antagonism via coupled δ-κ-opioid receptor heterodimers in the spinal cord (Portoghese and Lunzer, 2003; Bhushan et al., 2004). Importantly, the observation that i.t. DPDPE was more potently antagonized than U50488 by norBNI illustrates the need for caution in pharmacological characterization based on selectivity. This is further illustrated by the data that i.t. norBNI is ∼7-fold more potent than the δ antagonist NTI, in antagonizing DPDPE. The discrimination of phenotypic receptors by i.c.v. norBNI was relatively low (selectivity ratio ∼2) compared with i.t. administration.

Benzylidenenaltrexone. In the cord DPDPE (δ-1) was antagonized 8-fold more potently than deltorphin-II, and it exhibited a κ-1/κ-2 selectivity ratio of 5. These data are consistent with the well known δ-1 antagonism of BNTX. It is not known whether BNTX antagonizes bremazocine by an allosteric mechanism through interaction with a δ-1 receptor component of a δ-κ heterodimer or by direct interaction with the κ-2 phenotype. In view of the absence of significant selectivity for i.c.v.-administered BNTX for phenotypic δ receptors, heterodimers containing δ-1 and κ-2 phenotypic receptors that mediate analgesia are probably not abundant in the brain.

Naltriben. Greater antagonism of deltorphin II relative to DPDPE (δ-1/δ-2 = 4- to 6-fold) was observed by both routes of administration. NTB antagonized the κ-opioid agonists, U50488 and bremazocine, with selectivity ratios that were in the same range as the δ agonists. Interestingly, of the four i.c.v.-administered agonists, bremazocine was most potently antagonized by NTB.

Naltrindole. Because NTI is a selective δ antagonist, but generally is considered not to have “subtype” selectivity, its ability to antagonize selective agonists was examined. Upon i.t. administration, NTI blocked deltorphin II 4-fold more potently than it blocked DPDPE-induced antinociception. This was in contrast to a selectivity factor of 0.4 by the i.c.v. route. Although there was no significant difference between the i.c.v. and i.t. antagonism of DPDPE by NTI, i.t. deltorphin II was antagonized more potently than DPDPE by a factor of 8. The κ-1/κ-2 selectivity ratios for NTI by the i.t. and i.c.v. routes were in the range of 2 to 5.

Discussion

Heterodimerization of GPCRs has raised the possibility of greater pharmacological diversity in vivo compared with model systems that contain homogeneous populations of receptors (Jordan and Devi, 1999; George et al., 2000). Consequently, the in vivo properties of agonists and antagonists may not correlate well with data derived from cell-based assays or in other types of in vitro assays if the target receptors are organized differently. Because the use of selective opioid ligands used as pharmacological tools in vivo have in many cases been based upon in vitro binding selectivity or function in homogeneous populations of receptors, it is possible there may be a mismatch between in vivo and in vitro selectivity. Thus, it is conceivable that in vivo tissue-specific localization of heterodimers could give rise to erroneous assignment of an opioid receptor type involved in a pharmacological response if the recognition and functional properties of a heterodimer differ from those of a homodimer. Such differences could even account for the commonly observed greater in vitro binding selectivity of opioid ligands compared with in vivo pharmacological selectivity.

Consistent with this concept, the results of the present study have revealed that the antinociceptive receptor selectivity ratios of selective antagonists administered i.t. often differ significantly from those obtained by the i.c.v. route. For example, the κ-opioid antagonist norBNI was 12 times more potent in the antagonism of i.t. bremazocine (κ-2) relative to U50488 (κ-1), but only minimal difference was observed on i.c.v. administration. In view of reports suggesting the presence of δ-κ-heterodimeric opioid receptors in the cord but not the brain (Portoghese and Lunzer, 2003; Bhushan et al., 2004; Waldhoer et al., 2005), the present results may reflect differential distribution of phenotypic κ-1 and κ-2 receptors. Moreover, the more potent antagonism by norBNI of DPDPE relative to U50488 is in keeping with the proposed interaction of DPDPE with spinally localized heterodimers containing δ-1 and κ-2 phenotypic receptors as proposed by Bhushan et al. (2004). Consistent with the antagonism of DPDPE by norBNI, binding studies in cultured cells using selective antagonists have suggested cooperativity between δ and κ receptors organized as heterodimers (Xie et al., 2005). Thus, norBNI antagonism is mediated via competitive interaction at the κ-1 receptor, which allosterically leads to antagonism at the δ-1 receptor. Viewed from this perspective, the in vivo selectivity of norBNI would depend upon the phenotypic κ receptor targeted, as suggested by the relatively low κ-1/κ-2 selectivity ratio obtained i.c.v. where such heterodimers either are not present or do not mediate antinociception.

The 8-fold greater i.t. antagonism by BNTX of DPDPE over deltorphin II is consistent with the well known pharmacologic selectivity of this δ-1 antagonist (Portoghese et al., 1992). However, in contrast to the i.t. data, both DPDPE and deltorphin II were equally antagonized by BNTX given i.c.v. Likewise, the selective δ-kappa bivalent ligand antagonist, KDN-21, is reported to possess divergent selectivity by these different routes (Bhushan et al., 2004). These results may reflect differences between the organization of δ-opioid receptors in the spinal cord versus the brain. The report that the δ-κ-heterodimer-selective agonist 6′GNT1 produces potent spinal analgesia, but only weak, partial agonist activity i.c.v., also is consistent with the present results (Waldhoer et al., 2005).

The δ-2 antagonist NTB exhibited 4- to 6-fold greater selectivity in favoring deltorphin II over DPDPE, which is in harmony with its generally accepted selectivity. In view of the similar i.t. and i.c.v. δ-1/δ-2 selectivity ratios for NTB antagonism of DPDPE and deltorphin II antinociception, it seems that NTB targets identical phenotypic δ-opioid receptors in the cord and brain by the i.t. and i.c.v. routes of administration. The organization of this phenotypic δ-2 receptor recently has been proposed to differ from δ-1 based on studies with the bivalent ligand, KDAN-18, which contains both κ-1 agonist and δ-antagonist pharmacophores (Daniels et al., 2005). It was proposed that KDAN-18 bridges between oligomerized κ- and δ-receptor homodimers, and that κ-1 and δ-2 ligands selectively target such receptors. These receptors are distinct from the κ-2 and δ-1 phenotypes that have been proposed to be organized as heterodimers. In view of the report that KDAN-18 exhibits similar i.t. and i.c.v. agonist selectivity profiles, the NTB antagonism data provide additional support for the presence of both δ-2 and κ-1 phenotypic opioid receptors in the cord and the brain. Interestingly, of the four antagonists challenged by i.c.v. NTB, bremazocine was most potently antagonized. These results emphasize the importance of using multiple selective antagonists to evaluate the selectivity of agonists in vivo.

NTI is a δ antagonist, and it is considered to be nonselective in regard to distinguishing between phenotypic δ-1 and δ-2 receptors. Although this seemed to be the case when administered i.c.v., we have found that i.t. NTI exhibited 4-fold greater antagonism of deltorphin II relative to DPDPE. The lack of significant i.c.v. selectivity at phenotypic δ receptors and the observation that i.t. NTI antagonized deltorphin-II ∼8-fold more potently than by the i.c.v. route, suggests that NTI targeted different phenotypic δ receptors in the cord and brain. Given these data and the finding that i.c.v. NTI most potently antagonized κ-2 receptors, again suggests that several selective opioid antagonists should be used to establish the in vivo selectivity of an agonist.

In addition to affecting opioid receptor selectivity, the route of administration of opioid antagonists often affects potency, in part due to differences in the distribution of heterodimers in the brain and spinal cord. The finding that the i.c.v./i.t. ratios of agonists were either close to or not significantly different from unity (Table 1), whereas the antagonists more often exhibited i.c.v./i.t. ratios significantly greater or less than 1 (Table 3), reflects the complexity of the in vivo system. The complexity introduced by heterodimerization of opioid receptors is staggering when one considers the number of possible combinations that can exist for opioid receptor-containing heterodimers, when based upon studies of coexpressed receptors in cultured cells. In this regard, the standard selective agonists and antagonists for δ and κ receptors used in the present study may interact with heterodimers containing other opioid receptors and a variety of nonopioid GPCRs to produce analgesia. Unfortunately, it is not known how δ and κ receptors that are heterodimerized with nonopioid receptors would function in vivo. Thus, without knowledge of the tissue localization and identity of heterodimers containing δ-or κ-opioid receptors, the use of the standard armamentarium of opioid antagonists remains problematic for identifying specific opioid receptors in vivo. Clearly, the development of ligands that selectively target opioid receptor heterodimers would be key to their pharmacological characterization and localization in vivo.

The most dramatic example of divergent opioid antagonist potency as a function of route of administration was seen in the antagonism of DPDPE and bremazocine by norBNI, where the i.c.v./i.t. antagonist potency ratios were 12 and 40, respectively (Table 3). The much greater antagonist potency of norBNI by the i.t. route may reflect allosteric interactions within δ-1/κ-2 heterodimers present in the cord but not the brain (Bhushan et al., 2004: Waldhoer et al., 2005; Xie et al., 2005). However, correlation of the i.c.v./i.t. potency ratios with the receptor selectivity of other antagonists are not easily explained, given the paucity of information on their interaction with heterodimers. For example, since BNTX efficiently antagonizes DPDPE at δ-1 phenotypic receptors, it might be expected that BNTX would antagonize bremazocine through an allosteric mechanism. However, such an allosterically mediated mechanism may not occur, since this does not correlate with the finding that the i.c.v./i.t. antagonist potency ratio for BNTX antagonism of bremazocine, which was not significantly different from unity. Alternately, if DPDPE also interacts with a phenotypic δ receptor that is not present in the putative δ-1/κ-2 heterodimer, this may also be an explanation for the lack of a significant difference between the i.t. and i.c.v. antagonism of bremazocine by BNTX. Likewise, there are multiple possible explanations for the significantly high or low i.c.v./i.t. antagonist potency ratios for NTB and NTI.

In conclusion, antagonist ED₅₀ selectivity ratios derived from the administration graded doses of pharmacologically selective opioid antagonists in the presence an ED_80–90 dose of selective agonists has provided a reliable estimate of the selectivity profiles at spinal and supraspinal opioid receptors. These results are in qualitative agreement with earlier studies that used a single i.t. dose of antagonist, expressed as an ED₅₀ ratio in evaluating the spinal selectivity of agonists at phenotypic δ- and κ-opioid receptors. Importantly, in the present study the selectivity of an antagonist was dependent on the route of administration. The significant selectivity differences between the i.t. and i.c.v. routes of administration may be a consequence of different populations of phenotypic opioid receptors that reside in different tissues. Thus, it is suggested that the mouse spinal cord possesses δ-κ-opioid receptor heterodimers containing the δ-1 and κ-2 phenotypes that have somewhat different recognition properties from putative homomeric receptors or other δ and κ receptor-containing heterodimers in the brain. Given this tissue-dependent selectivity, the use of selective antagonists to characterize receptor selectivity may in some instances be problematic. Since the use of only one selective antagonist could lead to erroneous assignment of receptor selectivity, it is suggested that the antagonist ED₅₀ or ED₅₀ ratio paradigms be used with several selective antagonists to obtain a more reliable assignment of selectivity. Finally, the possible existence of cooperativity among receptors organized as heterodimers introduces new challenges in the pharmacological characterization of receptors, where transactivation or allosteric antagonism complicates the assignment of selectivity. It also offers the opportunity to develop new approaches to developing analgesics that act selectively, given the potential differential tissue distribution of heterodimers and the divergent recognition properties among phenotypic opioid receptors.