Receptor reserve and affinity of mu opioid agonists in mouse antinociception: correlation with receptor binding

Life Sciences, Vol. 57, No. 23, pp. 2113-2125, 1995 Copyright Q 1995 Elscvicr Science Inc. Printed in the USA. All rights reserved OW-3205/95 $950 t .M)

Pergamon

0024-3205(95)02204-X

RECEPTOR RESERVE AND AFFINITY OF MU OPIOID AGONISTS IN MOUSE ANTINOCICEPTION: CORRELATION WITH RECEPTOR BINDING

Gerald Zernig*, Tom Issaevitchs, Jillian H. Broadbear*?**, Timothy F. Burke*, John W. Lewis$, George A. Brine# and James H. Woods*?** Departments of *Pharmacology and **Psychology, University of Michigan, Ann Arbor, USA; 3Wolfram Research, Inc., Champaign, USA; #Research Triangle Institute, Research Triangle Park, USA; and $Department of Chemistry, University of Bristol, England (Received in final form September 20, 1995)

Summarv

In order to quantitate the extent to which opioid agonist potencies obtained in behavioral assays are determined by the apparent in vivo affinity and efficacy of the agonist, the antinociceptive effects of the mu opioid agonists morphine, fentanyl, etonitazene, and NIH 10741 were assessed before and after administration of the insurmountable mu opioid antagonist clocinnamox (CCAM) in a 55oC warm-water tail withdrawal test in Swiss albino mice. Under control conditions, all four mu opioid agonists produced a full antinociceptive response with the following ED50 values (in mg/kg): morphine, 12; fentanyl, 0.47; etonitazene, 0.039; NIH 10741, 0.005 1, Analysis of CCAM’s effects according to Black and Leff gave the following agonist efficacy or tau values: Morphine, 4; fentanyl 15, etonitazene, 7; and NIH 10741, 59. The respective KA values were (in mg/kg): morphine, 29; fentanyl, 7.3; 0.22; and NIH 10741, 0.30. The major determinant of the etonitazene, experimentally observed ED50 values seemed to be the apparent in vivo affinity of the respective agonist and not its efficacy. KA values (expressed as mol/kg) correlated with the Ki values (in mol/l) obtained with [3H]DAMG0 radioligand binding (r=O.96 for pKA vs. pKi), although being on average 11,OOO-fold higher. Values for q, the available receptor fraction as determined in the behavioral experiments, correlated strongly (r=0.96) with the q values determined by ex vivo [3H]DAMGOand [3H]naltrexone equilibrium binding (i.e., Bmax,clocinnamox/Bmax,control), the relationship approaching unity. Key words; mu opioid receptors, efficacy, apparent in vivo affinity, receptor reserve, thermal antmociception, analgesia, clocinnamox, morphine, fentanyl, etonitazene, NIH 10741 The potency of an agonist acting at a single receptor (expressed as ED50,control) in a functional assay is determined in principle both by its affinity (measured as its apparent dissociation constant, KA) as well as its efficacy, e, in the following way (1): ED50,control = KA / (e + 1). If one takes into account that a response might not be a rectangular hyperbolic function Corresponding author: Gerald Zernig, M.D., Department of Pharmacology, University of Michigan, 1301 Medical Science Research Building III, Ann Arbor, MI 48 109-0632, USA. Phone: (313) 7649133, FAX: (313) 764 -7118.

2114

Mu Opioid Receptor Reserve in Mice

Vol. 57, No. 23, 1995

of the agonist concentration, this relationship expands to: ED50,control = KA / ((2+taun)l/n (equation [9] of (2)) where the transducer ratio, tau, is an operational definition of efficacy and n Ih the slope factor of a transducer function relating receptor occupancy to observed response as proposed by Black and Leff (3); see Methods for more detailed definition of terms). It should be emphasized that both in Furchgott’s original model (1) and in its refinement and extension by Black and Leff (2, 3) the efficacy value, e, or the transducer function, tau, is the reciprocal value of the fraction of receptors necessary to give half the maximum possible effect. It would be desirable to obtain estimates for efficacy and apparent in vivo affinity in order to determine if potency differences among agonists actin g at a single site are based primarily on functional affinity differences, if the observed potency differences are predominantly mediated by differences in efficacy (reflecting receptor reserve), or if differences in both affinity and efficacy contribute to the observed potency difference among agonists. These questions are of even more concern if the same agonist shows different potencies across behavioral assays or across species as is the case for opioid agonists (see, e.g., (4-7)). Both efficacy and affinity can be determined by the method of partial insurmountable antagonism as originally developed by Furchgott (I), and modified by Black and Leff (2, 3). By administering a single dose of the insurmountable mu opioid antagonist clocinnamox a 14-cinnamoylaminomorphinone (S), and evaluating the changes in alfentanil and morphine dose-response curves in a rhesus monkey warm-water tail withdrawal assay, it was possible to non-invasively follow the time course of mu opioid receptor recovery in rhesus monkeys (7). It was shown that after a single dose of CCAM which inactivated 90% of the mu opioid receptors involved in thermal antinociception, receptors recovered with a half-life of 6 - 7 days. (CCAM),

In rhesus monkeysthe prototypical mu opioid agonists morphine and alfentanil displayed efficacies that were inversely proportional to the intensity of the warm-water stimulus: the higher the temperature of the warm water, the smaller the receptor reserve. Morphine, for example, had an efficacy value of 15 at 450C which means that under control conditions, occupation of only 7% of the available mu opioid receptors was necessary to give a half-maximal antinociceptive response (7). In other words, the receptor reserve at this temperature was 13-fold. At 55OC, however, morphine’s efficacy was only 3. To obtain the half-maximal antinociceptive response at this temperature, 33% of the available receptors had to be occupied by morphine. Thus, its receptor reserve at 55OC was only 2-fold. Morphine’s apparent in vivo dissociation constant, KA, however, did not differ across temperatures, indicating that the same receptors mediated the antinociceptive response. The present study extends this type of investigation to another species to compare the size of the receptor populations involved in thermal antinociception and their affinity profiles. To that end, mu opioid agonists of lower (morphine, fentanyl) and higher antinociceptive potency (etonitazene, NIH 1074 1) were evaluated before and after insurmountable mu opioid receptor blockade by CCAM in the warm-water tail withdrawal assay. The mathematical model used for the calculation of efficacy and the slope factor of the transducer function is that of Black and Leff (2, 3), as it can account also for cases in which the effect-agonist concentration- relationship is not a simple rectangular hyperbolic function. NIH 10741 was chosen because it is a high-potency analog of fentanyl (9). Finally, the receptor characteristics obtained by the analysis according to Black and Leff (3) were compared to those determined in both in vitro and ex vivo radioligand binding studies. Methods Subjects. Outbred male NIH Swiss albino mice (20 - 35 g) were obtained from Harlan Sprague-Dawley, Inc. (Indianapolis, IN) and housed in standard laboratory cages in a temperaturecontrolled colony room maintained on a 12-h light/dark cycle (lights off at 19:OOhr). Purina Rodent Chow (Purina Mills, St. Louis, MO) and water were available ad libitum. Animals were housed at least 24 h under these conditions before being tested. In each subject, only one dose-response curve was obtained. The experimental protocol was approved by the University of Michigan’s University Committee on the Use and Care of Animals.

Vol. 57,No. 23, l!BS

Mu Opioid Receptor Reserve in Mice

2115

Warm-water tail withdrawal exaeriments. The antinociceptive effect of the opioid agonists was determined using a warm-water tail withdrawal procedure as originally devised by Janssen and coworkers (10) for rats as modified by Comer et al. (I 1). Tail withdrawal latencies were obtained at 55oC. If the mouse did not withdraw its tail within 15 s, the experimenter pulled out the animal’s tail (cutoff latency); under these conditions, no tissue damage was observed (see Results). The s tail withdrawal latency was converted to % normalized response using the following formula: c/c antinociceptive response = ((test latency [s] - baseline latency [~])I(15 s - baseline latency [s])) * 100. Drugs were administered intraperitoneally. CCAM was given 1 h and naltrexone 15 min before the assay; either antagonist was administered in a single dose. Cumulative dose-response curves were obtained by repeatedly injecting the animals with agonist in half-log unless indicated otherwise. The total number of injections given to determine a complete dose-response curve for each individual mouse varied across animals. Tail withdrawal latenciec were tested 15 min after the drug injection. Baseline withdrawal latencies were obtained after vehicle (sterile water) injections at the start of the test. The interinjection interval was 30 min. In an experiment designed to distinguish the effects of a noxious 55oC water stimulus from the effects of a innocuous but otherwise identical stimulus on tail withdrawal latencies, animals were injected with vehicle alone and tail withdrawal measured at 360C (cutoff latency. 120 s). Radiolieand binding experiments. Opioid radioligand binding to whole mouse brain (i.e.. including cerebellum) membranes was determined according to Medzihradsky et al. (12, 13). Mosberg ef ul. (l4), and Pasternak et ~1. (for a review see (15)) with the following modifications: Male Swiss albino mice (20-35 g) were sacrificed by cervical dislocation and their brains immediately placed in ice-cold Tris buffer (50 mM Tris-HCI, pH 7.4 at 25oC; 0.1 mM phenylmethylsulfonyl fluoride). After mincing and washing in 40 ml Tris buffer, membranes were homogenized by 10 strokes in a Dounce homogenizer (pestle clearance, 60 - 90 microm) and centrifuged at 40,000 x g for 10 min. The pellets were resuspended at a concentration of 5 volumes [ml] per g wet weight and stored at -700C until used. Membranes were incubated in a total volume of 0.5 ml for 2-4 h at 25OC. MU opioid receptor binding was determined with [3H]DAMGO using I microM unlabeled DAMGO as the nonspecific binding definition. Specific [3H]DAMGO binding was linear up to 300 microg protein (determined according to Bradford, 1976; data not shown), Delta opioid receptor binding was measured with [3H]-p-CI-DPDPE (nonspecific binding definition, IO microM DPDPE). Kappa opioid receptor binding was determined with [3H]-(-)-bremazocine in the presence of I microM DAMGO and 1 microM DPDPE to prevent any [3H]-(-)-bremazocine binding to mu or delta receptors (nonspecific binding definition, 10 microM bremazocine). Drugs were diluted in 100% dimethyl sulfoxide (DMSO) and added to the assay volume to give a final DMSO concentration of 1%. DMSO up to a concentration of 5% did not inhibit [3H]-DAMGG binding (not shown). Radioligand displacement experiments were performed with compounds of interest and Ki values were calculated according to Linden (I 6). Data analysis. Functional data (% antinociceptive response) were used for subsequent analysis according to Black and Leff (2, 3). Black and Leffs mode1 gives the apparent in vivo dissociation constant, KA, of the agonist, and the transducer ratio, tau, of the agonist for the receptor system tested, which is “a logical operationai definition of efficacy” (3). Black and Leffs model accounts for non-rectangular hyperbolic effect-agonist concentration (E/[A]) -relationships by introducing a slope factor, n, into a transducer function. The model also allows for backcalculation of the ED50 of an agonist (see equation [9] of (2); Introduction). Black and Leffs mode1 also allows to calculate the maximum effect of the agonist (see equation [lo] of (3)) and compares it with the theoretically attainable maximum effect, Em, of the system under investigation. By dividing the tau value obtained after partial inactivation of the receptors by an irreversible antagonist with the tau value obtained under control conditions, one gets an estimate of the fraction of receptors still available foi interaction with the reversible agonist after blockade with a certain dose of irreversible antagonist, assuming that KE, i.e., the value of the agonist-receptor complex concentration for half the maximum possible effect ((2), p. 562) is not changed by the irreversible antagonist. As tau equals [RoVKE (equation [61 of (311, where [ROI represents the total receptor concentration, one can rearrange:

2116

Mu Opioid Receptor Reserve in Mice

ta”clocinnamox

Vol. 57,No. 23, 1995

[RO]clocinnamox / KE

[II taucontrol

=

lRO]control / KE

= ’

Thus, ]RO]cl ocinnamox = q * ]RO]control and: tauclocinnamox = q * taucontrol. We chose the letter “q” to designate this fraction because it corresponds to the q value of Furchgott (I) which signifies exactly the same conceptual entity, viz., the fraction of recptors available for interaction with an agonist after partial irreversible blockade with an antagonist. Equation [8] of Black et al. (2) states that:

Em * taun * [A]”

121

E= (KA + [A])n + taun * [A]”

where E is the effect (% maximum antinociceptive response in the present study); Em, the maximum attainable antinociceptive response; [A], the agonist concentration; n, the slope factor of the transducer function. Putting the above equation into semilogarithmic form, extending it by c, the baseline response (i.e., the response in absence of any mu agonist), and expressing any apparent tau value as (q * tatlcontrol), one gets:

Em*(q*taucontrol)” E= (lOlog(KA)+lO’og]A])n

* 10(‘“g]A]*n) +c +(q*taucontrol)n

131

*lO(log[A]*n)

Rearranging it into a more economical expression gives:

E = Em / (((lO(‘“g(KA)-‘og]A])+

1) / (q*taucontrol))”

+ 1)

+ c

[41

All agonist dose-response curves obtained with one mu opioid agonist (either before or after administration of various doses of CCAM) were simultaneously fitted to equation [4] ([A] given in mg/kg) using a nonlinear fitting program that was developed by two of the authors (17) using the genera1 mathematical software package MathematicaR (Wolfram Research, Champaign, USA; (18)). The program is available from the authors. It yielded best-fits for all curve parameters (including the q values) and two different estimates of variance (see below). Statistical analvsis. Efficacy and apparent in vivo affinity values were calculated as described above. Lo istic dose-response curve fits and statistical analysis were performed using the InPlotR and InStat $ computer packages (GraphPad, San Diego, USA). Unless indicated otherwise, values are means + S.E. of n determinations. ED50 refers to the dose (in mg/kg) causing 50% of the gruded maximum response. Whenever the mean and the S.E. of an ED50 value is given for statistical comparison, both values are given in their logarithmic form, as only the logarithms of the ED50 values are normally distributed ((19); cited in (20)). Whenever linear regressions are presented, the p value is given for the two-tailed comparison. Estimates of the variance of Black and Leff parameters obtained by the simultanous nonlinear fit of all dose-response curves of an agonist were obtained in two different ways: (i) Constrained 95% confidence intervals (CIs) for each parameter were determined by holding all other fitted curve parameters constant and (ii) unconstrained CIs were obtained by allowing all curve parameters to vary at the same time. As can

Vol. 57, No. 23, l!W

Mu Opioid Receptor Reserve in Mice

2117

be seen in Table I, the unconstrained CIs for efficacy were considerably larger than the unconstrained ones. In most other cases, however, the unconstrained CIs were equal to or smaller than the constrained ones. This occured because the constrained CIs were calculated on the basis of the observed non-Gaussian nature of the posterior distributions for the parameters, something that could not be accomplished for the unconstrained CIs (and thus were calculated assuming Gaussian distribution). Because most posterior parameter distributions were indeed non-Gaussian, no formal statistical comparison (e.g., analysis of variance) is given for these values, the more as the 95% Cl should give a more intuitive measure of “true” differences between fitted parameters. The reported confidence intervals are approximate and consistently underestimate the uncertainty in the inferred parameters. Drugs. Clocinnamox (CCAM, 14R-(p-chlorocinnamoylamino)-7,8-dihydro-N-cyclopropylmethylnor-morphinone mesylate) was synthesized by Dr. John Lewis and coworkers (University of Bristol, England) and NIH 1074 1 ((ZS,3R,4S)-cis-N-[ I -(2-hydroxy-2-phenylethyf)-3-methyl+ piperidyll-N-phenylpropanamide) was synthesized by Dr. George Brine and coworkers (Research Triangle Institute, Research Triangle Park, NC). [3H]DAMG0. [D-Ala2,N-Me-Phe4,Glys~ ol][t rosyl-3,5-3H]enkephalin; DAMGO, [D-A1a2,N-Me-Phe4,Gly5-ol]-enkephalin, DPDPE, [DPen 1 .D-Pen2]enkepha1in, [3H]-p-Cl-DPDPE, [Tyrosyl-3,5-3H]-(2,5-D-penici11amine,4-Clphenylalanine), and all other drugs were obtained from commercial sources or the National Institute on Drug Abuse. Drugs were dissolved in DMSO as a IO mM stock solution (final concentration in the assay, 1 %) for the binding experiments and in sterile water at varying concentrations for the behavioral assays, as DMSO was shown to affect behavior (Jillian Broadbear, unpublished observation). DMSO at concentrations up to of 5%~did not affect opioid radioligand binding. uResults Exnerimental cutoff latencv and the effect of an innocuous warm-water stimulus on the tail withdrawal latencv. The 15 s cutoff originally chosen by Janssen et al. (IO) was based on the fact that quanta1 dose-response curves of fentanyl did not change appreciably in potency once a 12 s tail withdrawal latency at 55OC for defining antinociception was exceeded. In order to investigate how closely this 15 s cutoff, which has been used routinely in our laboratory, corresponds to full antinociception, mice were subjected to an innocuous water temperature (i.e., 360C) under otherwise identical experimental conditions. For practical reasons, a cutoff of 120 s was observed in this experiment. For a total of 117 mice tested by three different experimenters, the median tail withdrawal latency was 21 s, the 10th percentile at 5 s, the 25th percentile at 9 s, the 75th percentile at 46 s, and the 90th percentile at 94 s. Alternatively, mice were tested with fentanyl at 55oC in a cumulative dosing procedure either observing a 15 s cutoff (n=48) or a 120 s cutoff (n=29); the respecive median latency was 4.6 s and 4. I s. In the 120 s cutoff group, 11 of 29 mice showed tail withdrawal latencies > 15 S. One of these withdrew its tail after 58 s, one after 100 s, the remaining 9 mice had their tail removed by the experimenter at 120 s. These mice showed observable tissue damage (petechial bleeding, blistering, partial skin avulsion) while the animals in the 15 s cutoff groups showed only hyperemia. Clocinnamox effects on mu ouioid agonist-mediated antinocicention in mouse warm-water tail withdrawal. Under control conditions, all four agonists tested produced a full antinociceptive response. Their respective ED50 values are listed in Table 1. Fig. 1 shows the effect of fentanyl on tail withdrawal latencies before and after administration of 1 and 3.2 mg/kg CCAM. CCAM shifted the fentanyl dose-response curve to the right in a dose-dependent manner. At 1 mg/kg, it slightly depressed the maximum effect of fentanyl (left panel); at 3.2 mg/kg, it induced a biphasic fentanyl dose-response curve (right panel). The addition of 32 mg/kg naltrexone in another group of CCAMpretreated mice completely blocked the lower-dose part of this biphasic dose-response function, whereas it was unable to antagonize the higher-dose effect of fentanyl (right panel). Thus, fentanyl exerted opioid-mediated inhibition of the warm-water tail withdrawal response at doses up to 10 - 32 mg/kg, whereas at doses > 32 mg/kg its inhibitory effect on tail withdrawal was not modified by naltrexone. Furthermore, 5 of the 26 mice tested with 32 mg/kg fentanyl in presence of 3.2 mg/kg CCAM (Fig.1, right panel) died when the test dose was increased to 100 mg/kg fentanyl; at 320 mg/kg fentanyl, 4 of 9 mice died. Similarly, 3 of 12 mice that had been tested with 32 mg/kg fentanyl in presence of both 3.2 mg/kg CCAM and 32 mg/kg naltrexone died when the test dose was

2118

Mu Opioid

Receptor

Reserve

in

Vol. 57, No. 23, lW5

Mice

increased to 100 mg/kg fentanyl; at 320 mg/kg fentanyl, 8 of 9 mice died. Thus, the sudden steep increase in tail withdrawal latency when increasing the dose of fentanyl from 32 mg/kg to 100 mg/kg in presence of CCAM and/or naltrexone (Fig. 1) was most likely due to severe disruption of vital functions, Therefore, in subsequent experiments, warm-water tail withdrawal latencies obtained with only up to 32 mg/kg fentanyl were used for further analysis. Fig. 2A shows that CCAM produced a dose-dependent rightward shift and depression of the fentanyl dose-response curve, i.e.. insurmountable antagonism of fentanyl’s effect in the warm-water tail withdrawal test. CCAM also considerably decreased the maximum effect of morphine (Fig. 2B) after shifting the agonist’s doseresponse curve only about 3-fold to the right, indicating that morphine was of lower efficacy than fentanyl (roughly IO-fold shift; Fig. 2A) in this procedure. Surprisingly, the dose-response curves of the high-potency agonist etonitazene (control ED50, 0.037 mg/kg; Table 1) were depressed at a low CCAM dose (1 mg/kg; Fig. 2C) and showed a maximal rightward shift that was comparable to morphine and fentanyl.

MOUSE

WARM-WATER

TAIL

WITHDRAWAL

55’C

1

“01

1

1

10

100

FENTANYL

1000

“01

1

1

10

100

1000

bw/W

Fig. 1. Dose response curves for the antinociceptive effect of fentanyl before and after treatment with the insurmountable opioid antagonist CCAM and naltrexone in the mouse warm-water tail withdrawal at 55oC. Shown are means + S.E. of the normalized antinociceptive response of 6 - 54 mice. Open circles, dashed line: control experiments. Left panel: One hour after pretreatment with 1 mg/kg CCAM, fentanyl dose-response curves were obtained in the absence (open triangles) or presence (filled triangles) of 32 mg/kg naltrexone given 15 min before the agonist. Right panel: After pretreatment with 3.2 mg/kg CCAM, fentanyl dose-response curves were obtained in absence (open squares) or presence (filled squares) of 32 mg/kg naltrexone. The form of the etonitazene dose-response curve shifts by CCAM suggests that the efficacy of etonitazene was also comparable to that of morphine and fentanyl, despite the fact that under control conditions, etonitazene was 324-fold more potent than morphine and 13-fold more potent than fentanyl (Table 1). Fig. 2D shows the effects of CCAM on the antinociceptive effect of highpotency fentanyl analog NIH 1074 1 (9). NIH 1074 l’s dose-response curve was shifted roughly 30fold to the right before being depressed, indicative of an efficacy around 30 (i.e., a 29-fold receptor reserve under control conditions).

2119

Mu Opioid Receptor Reserve in Mice

Vol. 57, No. 23, 1995

Calculation of efficacies and annarent in vivo dissociation constants from the antinociception assays. Table 1 summarizes the values for efficacy and KA and compares the experimentally observed ED50 values for the control experiments (i.e., in absence of CCAM) with the ED50 values that were calculated on the basis of the efficacy, n, and KA values according to Black and coworkers (2): ED50,control = KA / ((2+tau”) 1’n - 1). This feedback calculation showed that, over an approximately 2,400-fold range of experimentally observed ED50 values, the average difference between observed and calculated values was only 1. l-fold (range, 1.O - 1.3). Thus, the algorithm used to derive apparent in vivo affinity and efficacy estimates (see Methods) proved to be internally consistent. The values for Em, i.e., the theoretically attainable maximum effect of the system, were (in % of the experimentally observed maximum value; values in parenthesis are unconstrained CIs): Morphine, 100 (98 - 103), fentanyl, 102 (75 -127); etonitazene, 101 (95 - 106); and NIH 10741, 102 (75 - 126). This shows that although the parameter was allowed to vary freely, the fit converged on a value that was very close to that observed experimentally.

MOUSE

5”

WARM-WATER

WITHDRAWAL

55-C

.JT t

0

TAIL

*

-I

1

10 MORPHINE

100

1000

[mg/kgl

Fig. 2. Effects of clocinnamox on mu opioid agonist-mediated antinociception in the mouse warmwater tail-withdrawal test. Shown are means f S.E. of the normalized antinociceptive response of 6 - 54 mice. Dose-response curves were determined in the absence (open circles) or after a 1 h pretreatment with 0.32 (filled circles), 1 (filled triangles), 3.2 (filled squares), or IO (diamonds) mg/kg CCAM. A, fentanyl; B, morphine; C, etonitazene; D, NIH 10741.

Opioid receptor binding. In vitro opioid radioligand saturation equilibrium binding experiments using whole mouse brain membranes yielded a Kd of 2.4 + 0.36 nM and a Bmax of 157 + 1I fmol/mg for [3H]DAMG0 (n = 5). [3H]-p-Cl-DPDPE saturably bound to the homogenate with a Kd of 1.O + 0.17 nM and a Bmax of 102 _+8.4 fmol/mg (n = 4). [3H]-(-)-Bremazocine binding in the presence of 1 microM DAMGO and 1 microM DPDPE yielded two binding site populations with Kd values of 0.28 2 0.07 and 5,400 f 4,100 nM, respectively (n = 6). Their Bmax values were 88 + 0.12 fmol/mg for the high-affinity and 28,000 + 6,000 fmol/mg for the low-affinity population.

Vol. 57, No. 23, 1995

Mu Opioid Receptor Reserve in Mice

2120

TABLE 1 Comparison Of Efficacy, Apparent In Vivo Affinity, And Observed Vs. Calculated ED50 Values For Opioid Agonists In The 55OC Warm-Water Tail Withdrawal Test. DRUG

Morphine

Fentanyl

Etonitazene

NIH 10741 only two curves analyzed

all three curves analyzed

Values for the efficacy, tau, and the apparent in vivo dissociation constant, KA, were obtained for each agonist by simultaneous analysis of dose-response curve obtained before and after administration of CCAM using the model by Black and Leff (3) as described in Methods. Shown are the fitted parameters and - in parentheses - both the constrained (upper pair) and the unconstrained (lower pair) 95% confidence intervals (see Methods for definition). The doseresponse curves used for the calculation of tau, KA, and n values are shown in Fig.2. In the case of NIH 10741, inclusion of the dose-response curve obtained in presence of 10 mg/kg CCAM, which was essentially flat, inflated the variance of the fitted parameters; furthermore, a satisfactory fit was obtained only after allowing the slope factor, n, to vary across individual dose-response curves. Therefore, the left NIH 10741-column lists a fit using only two of the three dose-response curve. The results of a fit of all three NIH 10741 curves, however, are listed for the benefit of the reader in the right NIH 10741-column. ED50 values for control dose response curves were calculated using the following relationship given by equation [9] of Black and et al. (2): ED50,control = KA / ((2+taun)1’n _ 1).

2121

Mu Opioid Receptor Reserve in Mice

Vol. 57, No. 23,199s

Thus, in all subsequent binding assays, care was taken never to exceed the IQ of the high-affinity receptor population because at 0.28 nM radioligand the low-affinity binding sites can contribute to only 3% of the specific binding. Table 2 summarizes the binding characteristics of these drugs. The affinity rank order for mu opioid binding was NIH 10741 (Ki = 0.084 nM) 2 etonitazene (0.11 nM) > morphine (3.4 nM) 2 fentanyl (3.5 nM). Thus, the agonists differed maximally 42-fold in their mu opioid receptor affinity. In contrast, maximum affinity differences for delta receptors were only S-fold, those for kappa receptors 12-fold. Etonitazene and NIH 10741 showed remarkable muselectivities (etonitazene, 1,100 for mu:delta and 17,000 for mu:kappa; NIH 10741,550 for mu:delta and 2,500 for mu:kappa). From these data, it seems highly unlikely that any of the antinociceptive effects (in absence of any opioid antagonist) of at least these two ligands might be mediated by other than mu opioid receptors.

TABLE 2 Comparison Of Binding Affinities And Behaviorally Affinities In Male Swiss Albino Mice

Drug

mu opioid receptors

delta opioid receptors

K bM1

kappa opioid receptors

Ki [nMl @Ki)

Wi)

Determined

Ki tnM1 @W

Apparent

Binding selectivity

au: ielta

In Vivo

PKA warmwater tail withdrawa

mu: kappa

Morphine 3.4 (8.47 _+ 0.07)

119 (6.93 f 0.12)

1450 (5.84 + 0.08)

35

3.5 (8.46 f 0.18)

380 (6.42 F 0.07)

2600 (5.59 + 0.08)

109

43 1.05

Fentanyl

Etonitazene

I

1.71

1

I055 (9.Z

lO.09) / (6.94Z.09)

74

16811

/ (5.7::00.06)

6.29

NIH 10741 551 10.00;oi840.06) j (7.33yO.06) I

/ (6.682100.08)

2491 6.13

I

Ki values for the different agonists were determined as described in Methods and converted to -log(Ki) = PKj values to allow statistical comparison. Shown are means + S.E. of n determinations. The numbers of individual radioligand displacement experiments used to determine mu-, delta -, and kappa -opioid receptor affinities were the following: for morphine, 5, 2, 2; fentanyl, 3, 3, 2; etonitazene, 3, 2, 2; NIH 10741, 8, 2, 3. Hill slopes did not significantly differ from unity. Values for the apparent in vivo dissociation constant, KA. were taken from Table 1 (dimension: mg/kg) and converted to nmol/kg and finally into -log(mol/kg) to allow direct comparison with the PKi values obtained in binding studies. In the case of NIH 10741, the dose-response curve obtained in presence of 10 mg/kg CCAM, which was essentially flat (Fig. 2), was not used for the calculation of the pKA listed befow, because its inclusion inflated the variance of the fitted parameters; when it was used, the pKA for NIH 10741 changed to 5.56. The molecular weights were: morphine HCI, 322; fentanyl HCI, 374; etonitazine HCI, 433; NIH 10741 HCI, 403. n.d., not determined.

2122

Mu Opioid Receptor Reserve in Mice

Vol. 51, No. 23, 1995

Correlation between recentor copulation parameters derived from functional exneriments and radioligand binding studies. For all compounds, the comparison of behaviorally determined apparent in vivo affinity and mu opioid receptor binding affinity yielded a correlation coefficient of 0.96 (n=4, p=O.O38), the relationship being: pKA,function = 1.2 (+ 0.2) * pKi,binding - 5.5 (? 2.2). In contrast, there was statistically a much weaker correlation between pKA values and pKi values for delta (r = 0.53, p = 0.47) or kappa (r = 0.46, p = 0.54) opioid receptors. Comparison of the pED50 values for antinociception and the pKi values for mu opioid receptor binding gave the following relationship (n = 4, r = 0.9, p = 0.1): pED50 = 1.5 (+ 0.5) * pKi,binding - 7.6 (? 1.5). Table 3 compares the q values determined in the antinociception experiments described above with q values determined by ex vivo radioligand binding (21). In both cases, CCAM was present in the mice for the same period, i.e., 1 h, before either the antinociceptive test was performed or the animals were sacrificed in order to obtain and homogenize their brains. Linear regression analysis for all CCAM doses tested both in functional and ex vivo binding experiments (averaged q values; n = 5, r=0.96, p=O.O093) gave the following relationship: qfunction = 0.99 (* 0.17) * (Ibinding 0.035 (+ 0.096). The slope was not significantly different from unity. The ED50 value of the ex viva binding inhibition by CCAM was 0.79 mg/kg and the Hill slope was 1.1.

Comparison Of Clocinnamox By Ex Viva [3H]-DAMGOExperiments.

Clocinnamox dose

hW

Available mu opioid receptor fraction PmaxlBmax,col

TABLE 3 Effects On Mu Opioid Receptor Population As Determined And [3H]-Naltrexone Binding Assays And Functional

Individual q values

Average q value

Fentanyl

Morphine I

Etonitazene I

NIH 1074 1 I

0

1

1

1

1

0.1 0.32 1 3.2

n.d. 0.69 0.26 0.17

n.d. 0.62 0.2 0.056

n.d. n.d. 0.5 0.13

n.d.

n.d.

0.064

n.d. n.d. n.d. 0.026 (0.013) n.d. (0.00004)

10* E,r vivo

mouse brain saturation binding experiments performed 1 h after CCAM pretreatment using [3H]DAMG0 (control B max, 67 fmol/mg; protein determined according to Lowry et al. (195 1) and [3H]naltrexone (control B max, 149 fmol/mg) were taken from Table 1 of Burke et al. (21), averaged and the fraction of available m receptors expressed as Bmax,CCAM/Bmax,control. *, the fraction of available mu opioid recpeptors after pretreatment with 10 mg/kg CCAM was not directly determined in the binding assays but extrapolated from a dose-response curve obtained with all other CCAM doses (ED50, O.Smg/kg; Hill slope, 1.1). For the determination of cl in the mouse warm-water tail withdrawal test, see Methods. In the case of NIH 1074 1, the dose-response curve obtained in presence of 10 mg/kg CCAM, which was essentially flat (Fig. 2) was excluded from the analysis, because its inclusion inflated the variance of the fitted parameters. However, q values from a fit that included this curve are given in parentheses. Linear regression of the averaged values (n=5, r=0.96, p=O.O093) gave the following relationship: qfunction = 0.99 (& 0.17) * qbinding - 0.035 (k 0.096). Comparison of the averaged ‘Ibinding values with the individual qfunction values (n=14, r=0.95, p
Vol. 57, No. 23, 1995

Mu Opioid Receptor Reserve in Mice

2123

Discussion In the warm-water tail withdrawal assay, the insurmountable mu opioid antagonist clocinnamox (CCAM) dose-dependently shifted the dose-response curves for all tested mu opioid agonists to the right and caused a decrease in the maximum antinociceptive response, the latter effect strongly suggesting an irreversible mode of action of CCAM (e.g., (1); for a detailed discussion see (7)). These findings corroborate similar results obtained in mice both in behavioral assays (1 1) and binding studies (2 1).

In rhesus monkeys, data obtained with CCAM has been used to calculate the apparent in vivo dissociation constant, the efficacy (and thus, the receptor reserve), and the fraction of available receptors for morphine and alfentanil after insurmountable inactivation with CCAM (7). In the rhesus monkey, the apparent in vivo dissociation constant for morphine in the warm-water tail withdrawal was calculated to be 24 mg/kg (95% CI 17 - 35 mg/kg, n=33) which is essentially equivalent to that obtained in the present study in mice (i.e. 29 mg/kg (CI 29.1 - 29.); Table I). Similar values have been reported in rats (25 mg/kg) in the warn-water tail withdrawal procedure by Tallarida and Cowan (22) and by Porreca et al. (22 mg/kg; (23)). The values are also comparable to the Km of 25 mg/kg (s.c. dose) obtained by Blaesig et ul. (24) who used a vocalization threshold antinociception assay in naive and morphine-tolerant rats. Finally, the efficacies of morphine in the 55oC warm-water tail withdrawal assay were essentially the same for the rhesus monkey (3 + I. n=9; (7)) and the mouse (4 (CI, 3.2 - 4.9): Table 1). Etonitazene, a high-potency agonist, had an efficacy of only 6.9 (Table I), i.e., 14 % of all relevant mu opioid receptors are necessary to produce a half-maximal antinociceptive response under control conditions. In comparison, fentanyl’s efficacy was 15 (i.e., only 6 % of all receptors are necessary). However, feedback calculation gave calculated ED50 values for the control doseresponse curves that on average differed only 1.1 -fold from the experimentally observed ED50 values which varied 2,400-fold. Thus, the mathematical analysis proved to be internally consistent and the obtained efficacy rank order of high confidence, despite all the shortcomings and caveats of the analytical method used (see (7) for a detailed discussion). Overall, the major determinant of the observed control ED50 values was the apparent in vivo affinity of the drug: In four out of six possible comparisons between the four tested compounds, the differences in experimentally observed ED50 values were matched by much larger differences in KA values than in efficacy (i.e., tau) values (Table 1). In one case (i.e., fentanyl vs. morphine), differences in KA and tau were equal; in only one case (i.e., etonitazene vs. NIH 10741, was the difference in tau larger than that in KA). For all six comparisons, the products of the differences in KA and tau corresponded very well to the differences in ED50,control. Thus, until mu opioid agonists of much higher efficacy (i.e., much better signal transducing ability) are found, a reasonable prediction of an agonist’s antinociceptive potency could be based on its affinity in opioid receptor radioligand binding displacement experiments. For all tested compounds and assays, the best correlation was obtained when comparing apparent in vivo affinity and mu opioid receptor binding affinity. In contrast, there was a much weaker correlation between pKA values and pKi values foi delta or kappa opioid receptors. As the ED50 values are determined primarily by the apparent in vivo affinities and these in turn correlate well with the affinities as determined by in vitro radioligand binding, one would expect that control ED50 values correlate well with Ki values, too. This was indeed the case: The correlation between the pED50 for the functional assays (-log[ED50 in mol/kg]) and the pKi for the binding experiments(-log[Ki in mol/l]) for all 4 tested agonists was 0.9 (p=O. 1). A compound-by-cmpound comparison of the affinities also showed that the functionally determined KA was on average 11,OOO-fold higher than the dissociation constant of the same mu opioid agonist observed in in vitro equilibrium binding experiments (Table 2). Absorption of the drug from the peritoneal cavity, distribution, e.g., poor penetration of the blood-brain barrier, and compartmentalization presumably account to some degree for this difference. However, the slope of the relationship between apparent

2124

Mu Opioid Receptor

Reserve

in Mice

Vol. 57, No. 23, 1995

in vivo dissociation constant and equilibrium binding dissociation constant approached unity, suggesting that the binding affinity rank order adequately predict the functional affinity rank order. The good correlation between behavioral and binding data found for apparent in vivo dissociation constants and Ki values obtained in equilibrium binding displacement experiments was also observed when behavioral and binding experiments were compared with respect to their quantitation of receptor population changes after administration of the insurmountable opioid antagonist CCAM. Comparison of functionally obtained q values (i.e., the fraction of available receptors after administration of the insurmountable antagonist) and q values determined by ex vivo equilibrium binding of [3H]DAMG0 and [3H]naltrexone (i.e., B max,clocinnamox/Bmax,controI) yielded a correlation coefficient of 0.96 (p = 0.0093). As already observed in the case of the pKA-pKicomparison (see above), the slope of the regression did not significantly differ from unity, strongly suggesting that q values determined by functional experiments are identical to the q values obtained in binding experiments. Thus, the non-invasive determination of receptor population changes in the intact, behaving animal is well suited to yield quantitative data on receptor population changes of reasonably high confidence. It should be emphasized that for the definition of the full antinociceptive response an arbitrary cutoff of 15 s tail withdrawal latency was used. This value had originally been chosen by Janssen and coworkers (1963) because above cutoff values of 12 s, the antinociceptive potency of the test compound fentanyl in quanta1 dose-response analyses did not change. In retrosprect, we found that an innocuous but otherwise identical stimulus (i.e., tail immersion into 36OC instead of 55OC water) led to a median tail withdrawal latency of 21 s. Converseley, extending the experimental cutoff from 15 s to 120 s did not appreciably change the median tail withdrawal latency (4.6 s to 4.1 s). Raising the cutoff to 120 s caused considerable tissue damage and still was not sufficient to have all mice withdraw their tails before reaching the experimental cutoff. Most importantly, pretreatment with high enough doses of CCAM lowered the maximum tail withdrawal latencies for al] agonists tested well below the IS s cutoff. For example, 3.2 mg/kg CCAM lowered maximum tail withdrawal latencies in the 5 s-range (i.e., approximately 26% maximum antinociceptive response; Fig. 2). According to the theory , the ratio of the maximum agonist effect after partial irreversible blockade, E’Am, can be given as: E’Am / Em = (q *tau control) / (I + (q*taUcontro])“) in extension of equation [lo] of Black et al. (2). If the true maximum tail withdrawal latency (Em) had been 60 s and not I5 s as defined by the cutoff, a 5 s maximum tail withdrawal latency after CCAM pretreatment (E’Am) would correspond to a q value of 0.0091 assuming a the tau control of 10 and a q value of 0.00091 assuming an efficacy of 100 (all assuming that n=l), values which are well below any of the q values that were either directly measured by ex vivo radioligand binding assays or calculated by Black and Leff analysis of the functional data (Table 3). Furthermore, although the fitting routine used in the present study allowed the E m to vary freely, it converged at Em values essentially identical to the experimentally observed ones. Estimates of Em in the operational model are known to be very sensitive to deviations from ideal conditions, e.g., considerable ternary complex formation (25). For all these reasons, we think that the choice of a 15 s cutoff for the definition of full antinociceptive response is adequate both for the description of the agonist potencies and the efficacy and apparent in vivo affinity estimations based on them. Acknowleements Drs. Fedor Medzihradsky, Henry Mosberg, and Gavril Pasternak are thanked for their advice on opioid receptor radioligand binding techniques. Chris Rowan, Jason Vieder, Sandra Lee, and Rob Hutchman provided expert technical assistance. This work was supported by the Austrian Science Foundation Grants J0697-MED and J0882-MED, by a NIDA INVEST Fellowship (to G.Z.) and USPHS Grant DA 00254. References 1. 2. 3.

R.F. FURCHGOTT, Adv. Drug Res. 2 21-56 (1966). J.W. BLACK, P. LEFF, N.P. SHANKLEY and J. WOOD, Br. J. Pharmacol. (1985). J.W. BLACK and P. LEFF, Proc. R. Sot. Lond. B 220 141-162 (1983).

@ 561-57 1

Vol. 57, No. 23, 199.5

4. 5. 6. 7. 8. 9. 10. 1 1. 12. 13, 14. 15. 16. 17.

18. 19. 20. 2 I. 22. 23. 24. 25.

Mu

Opioid Receptor Reserve in Mice

2125

A.M. YOUNG, J.H. WOODS, S. HERLING and D.W. HEIN, The Neurobioloev Of Ouiate Reward Processes, J.E. Smith and J.D. Lane (Eds), 147-174, Elsevier, Amsterdam (1983). C.P. FRANCE, A.E. JACOBSON and J.H. WOODS, J Pharmacol Exp Ther 230 652-7 (1984). G. WINGER, P. SKJOLDAGEr and J.H. WOODS, J. Pharmacol. Exp. Ther. 261 31 l-317 (1992). G. ZERNIG, E.R. BUTELMAN, J.W. LEWIS, E.A. WALKER and J.H. WOODS, J. Pharmacol. Exp. Ther. 269 57-65 (1994). J.W. LEWIS, C. SMITH, P. MCCARTHY, D. WALTER, R. KOBYLECKI, M. MYERS, A. HAYNES, C. LEWIS and K. WALTHAM, NIDA Res. Monogr. 90 136-143 (1988). G.A. BRINE, P.A. STARK, F.I. CARROLL. H. XU and R.B. ROTHMAN, Med. Chem. Res. 2.34-40 (1992). P.A.J. JANSSEN, J.E. NIEMEGEERS and J.G.H. DONY, Arzneim. Forsch./Drug Res. fi 502-507 (1963). S.D. COMER, T.F. BURKE, J.W. LEWIS and J.H. WOODS, J. Pharmacol. Exp. Ther. 262 1051-1056 (1992). S.V. FISCHEL and F. MEDZIHRADSKY, Mol. Pharmacol. 2Q 269-279 (198 1). A.E. REMMERS and F. MEDZIHRADSKY, J. Neurochem. 57 1265- 1269 (199 1). D.L. HEYL and H.I. MOSBERG, J. Med. Chem. 55 1535-1541 (1992). G.W. PASTERNAK, Clin. Neuropharmacol. 16 1-l 8 (1993). J. LINDEN, J. Cycl. Nucl. Res. 8 163-172 (1982). G. ZERNIG and T. ISSAEVITCH, Software For The Calculation Of Agonist Efficacv And Annarent In Vivo Affinitv Bv Simultaneous Analvsis Of Several Dose-Resnonse Curves. Wolfram Research, Champain (1995). S. WOLFRAM, MathematicaR: A Svstem For Doing Mathematics Bv Comnuter. (2nd Ed,) Addison-Wesley, Reading (1991) W.W. FLEMING, D.P. WESTFALL, IS. DELALANDE and L.B. JELLETT, J. Pharmacol. Exp. Ther. 181339-345 (1972). T.P. KENAKIN, Pharmacologic Analvsis Of Drug-Receptor Interaction (2nd Ed.). Raven Press, New York (1993) T.F. BURKE, J.H. WOODS, J.W. LEWIS and F. MEDZIHRADSKY, J. Pharmacol. Exp. Ther. 271715-721 (1994). R.J. TALLARIDA and A. COWAN, J. Pharmacol. Exp. Ther. 222 198-201 (1982). F. PORRECA, A. COWAN, R.B. RAFFA and R.J. TALLARIDA, Life Sci. 31 2355-2358 ( 1982). J. BLAESIG, G. MEYER, V. HOELLT, J. HENGSTENBERG, J. DUM and A. ‘Herz. Neuropharmacology u 473-48 1 (1979). P. LEFF, D. HARPER, IA. DAINTY and I.G. DOUGALL, Br. J. Pharmacol. 101 55-60 ( 1990).