Proc. Nat. Acad. Sci. USA Vol. 72, No. 5, pp. 1899-1903, May 1975
Phenothiazine Drugs: Structure-Activity Relationships Explained by a Conformation That Mimics Dopamine (catecholamine/neurotransmitter/receptor/chlorpromazine/schizophrenia)
ANDREW P. FEINBERG AND SOLOMON H. SNYDER Departments of Pharmacology and Experimental Therapeutics and Psychiatry and the Behavioral Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland 21205
Communicated by Seymour S. Kety, February 21, 1975 ABSTRACT The antischizophrenic activity of phenothiazine drugs and their tendency to elicit extrapyramidal symptoms are thought to involve blockade of synaptic dopamine receptors in the brain. Space filling molecular models show how favorable Van der Waal's interactions between the side chain amino of phenothiazines and the 2substituent on ring A can promote a conformation mimicking dopamine. These Van der Waal's attractive forces can explain (i) the greater potency of drugs with trifluoromethyl rather than chlorine as a 2-substituent; (ii) the enhanced activity of phenothiazines with piperazine instead of alkylamino side chains; (iii) the increased potency associated with hydroxyethylpiperazines as contrasted to piperazine side chains ; (iv) the greater potency of cis rather than trans thioxanthenes; and (v) the crucial location of the ring A substituent at carbon no. 2. Potential energy calculations support the observations with molecular models and suggest an active conformation for the phenothiazines.
An abundance of recent research suggests that a major mechanism whereby antischizophrenic phenothiazine drugs exert their therapeutic actions and extrapyramidal side effects involves a blockade of synaptic receptor sites for dopamine in the brain (1, 2). The activity of a dopamine-sensitive adenylate cyclase correlates with dopamine receptor activity (3). Relative potencies of several phenothiazine and related drugs as inhibitors of the dopamine-sensitive adenylate cyclase parallel their antischizophrenic potency (4-6). Earlier we proposed a molecular model wherein dopamine could be superimposed upon a portion of the chlorpromazine molecule (7). While explaining some structure-activity characteristics of phenothiazines, this model did not deal with important features such as the greater potencies of phenothiazines with piperazine rather than alkylamino side chains, nor did it explain the greater potency of trifluoromethyl than of chlorine ring substituents. Moreover, it provided no explanation for the mechanism whereby the A ring substituent caused the side chain to tilt toward the A ring, nor did it explain why the A ring substituent must be located in the number 2 position. In the present study, Corey-Pauling-Koltun molecular models and computer calculations support a model in which phenothiazines assume a conformation that mimics that of dopamine, explaining the role of trifluoromethyl, piperazine, and hydrox-yethvlpiperazine groups as well as the mechanism whereby A ring substituents influence the side chain. MATERIALS AND METHODS
Molecular models of promazine, chlorpromazine, triflupromazine, prochlorperazine, trifluoperazine, thiethylperazine,
1899
perphenazine, and fluphenazine were constructed with CoreyPauling-Koltun kits (Ealing Corp.; Cambridge, Mass.). To confirm the apparent influence of Van der Waal's forces on the side chain conformation, we performed potential energy calculations on the following three compounds with differing ring substituents but the same side chain: promazine, chlorpromazine, and triflupromazine. Calculations were performed on alkylamino side chain phenothiazines, since the great increase in conformational variables introduced by a flexible piperidine or piperazine ring was beyond the practical capability of the computer. The calculations were performed on a PDP-12 computer with floating point processor and 8K of memory, using a form of the Buckingham 6-exp potential function proposed by A. I. Kitaygorodskii (8): V = 3.5 (8600 e-13Z - 0.04 Z-6) where z is the ratio of the distance between atom centers and G. N. Ramachandran's K2 values for Van der Waal's radii (9). This function has shown good agreement with observed x-ray crystal structures (10). Two groups of computer programs, developed at Washington University Computer Systems Laboratory, were used. The first group, CHEMAST (11, 12), was used to generate lists of atomic coordinates, atom types, connectivity and rotation parameters, and stereoscopic images of the molecules on a point-plotting cathode ray tube. The second group, BURLESK (13, 14), was used to perform iterative rotations over the bonds indicated in Fig. 1 and the trifluoromethyl group in triflupromazine, computing the potential energy between all atom pairs whose relative orientations were changed by the rotations; and to display the data on the cathode ray tube as isoenergy curves. To obviate calculations on absurd conformers, the programs made a "bump check", for which Van der Waal's radii were set at 90% of those used by Leach et al. (15). The data were compared to control data obtained without a "bump check" to ascertain that no legitimate conformers were excluded. A special connectivity table was prepared for the phenothiazine ring system (16), and its bond lengths and bond angles were obtained from the published crystal structure of thiethylperazine (17). It might have been more natural to use published data on chlorpromazine (18), but these yielded bizarre ring atom placements. For each molecule, between 16,000 and 150,000 conformations were examined using 5-15 hr of computer time. In criti-
1900
Medical Sciences: Feinberg and Snyder
19l 20
(ECH2\
I*He
(1975)
5
4
CH3
Proc. Nat. Acad. Sci. USA 72
9
-C1
'qH2
T3 162
4
CH 3 19
T,
a
T(C,,-NO-C15"C16 )
T2=aT (NI0_CI5 -C6 -CI7 ) T3 T(Ci5-CI6-CI7 -NIS) a
T(C16-CI7
-N18 -C19)
FIG. 1. Rotatable bonds of chlorpromazine, drawn with = T4 = 00 and T3 = 1800. The convexity of the phenothiazine ring system is upward. The torsion angle r(A-X-Y-B) is the dihedral angle between planes AXY and XYB, measured clockwise when viewed down the bond X-Y (25). T1. =2
cal regions, the data were compared to control data obtained density of conformations to confirm their completeness. A Hewlett-Packard 7005B X-Y Recorder was used to draw the isoenergy curves on graph paper, and they were then retraced. The figures of ball and stick models were made with a DECSYSTEM 10 time-sharing computer, AGT-30 ADAGE graphics system, and model 200 ZETA plotter, using the program package CHEM (19), and they were then redrawn by an artist. over a greater
C
RESULTS AND DISCUSSION
Space-filling models of chlorpromazine (Fig. 2) reveal that tilting of the side chain toward ring A permits favorable Van der Waal's interactions of the side chain with the chlorine substituent. These Van der Waal's attractive forces would be greatly lessened in the case of promazine, which lacks the chlorine substituent. This conformation permits the superimposition of dopamine (Fig. 3), thus explaining how chlorpromazine can interact with dopamine receptors. The spacefilling models provide a causal explanation for the folding of the side chain toward ring A. The notion that Van der Waal's attractions between the side chain and the ring A substituent could account for the ability of phenothiazines to assume the dopamine-like conformation allows several predictions about the structure of phenothiazines that might be expected to be more or less potent in interactions with the dopamine receptors. For instance, this model requires that the ring A substituent be located at the number 2 position. A 1-substituent would sterically hinder the ability of the side chain to approach ring A, while a 3substituent would be too distant from the side chain to provide Van der Waal's attractions of a significant degree. Extensive experience with many phenothiazine structures has shown that optimal neuroleptic activity occurs only when the ring A substituent is in the 2-position (20, 21). The nature of the 2-substituents in our model would be expected to influence the conformation of the side chain. A trifluoromethyl substituent provides a greater number of favorable Van der Waal's contacts with the side chain than
FIG. 2. Phenothiazines with the side chain "tilted" toward the A ring, drawn with relatively small atom size for ease of viewing. ri and T2 are set at (-68°, 1350), the relative minimum of conformational zone no. 2 in Fig. 5. (a) Chlorpromazine (the side chain N to Cl distance is 3.3 A); (b) triflupromazine (the distance of the side chain N to the closest fluorines is 3.0-3.5 A); (c) trifluoperazine; (d) fluphenazine.
the chlorine substituent. Our model predicts that phenothiazines with trifluoromethyl substituents should be more potent than those with chlorine substituents. Of the clinically used phenothiazines, those with trifluoromethyl substituents at position 2 of ring A are uniformly more potent than those with chlorine substituents, both clinically and in their inhibition of the dopamine-sensitive adenylate cyclase (4-6, 20, 21). The nature of the atom in proximity to the side chain amine might also be anticipated to influence the conformation of the phenothiazine side chain. A piperazine side chain affords more Van der Waal's contacts with the 2-substituent than does an alkylamino side chain (Fig. 2). The resultant predictions that
Proc. Nat. Acad. Sci. USA 72
Phenothiazines and Dopamine
(1975)
1901
TABLE 1. Phenothiazine and thioxanthene drug structures and effects on a dopamine-sensitive adenylate cyclase
Relative potency in inhibiting
Drug
Ca-Flupenthixol (aO6)-Flupenthixol Fluphenazine Trifluoperazine Triflupromazine a-Clopenthixol Perphenazine a-Chlorprothixene Prochlorperazine Chlorpromazine Promazine
jS-Chlorprothixene ,6-Clopenthixol
fl-Flupenthixol
Ring thio. thio. pheno. pheno. pheno. thio. pheno. thio. pheno. pheno. pheno. thio. thio. thio.
R, HEP HEP HEP P A HEP HEP A P A A A HEP HEP
dopamine-sensitive adenylate cyclase of rat corpus striatum (chlorpromazine = 100) t Results of Results of Miller and Clement-Courmier Iversen (5) et al. (4) 4545 1333 1087 857 250 600
Relative clinical potency* 1 2 3 4 5 6 7 8 9 10 11 12 13 14
R2 CF3 CF3 CF3 CF3 CF3 Cl Cl Cl C1 Cl H Cl Cl CF3
150
303 128 45 100 17 5 1.7 1
120 100
171t
Inhibition of dopamine-sensitive adenylate cyclase, relative to chlorpromazine = 100. R1 and R2 refer to Fig. 4. Abbreviations: thio. = thioxanthene, pheno. = phenothiazine, A = alkylamino, P = piperazine, HEP = hydroxyethylpiperazine. * Drugs are listed in approximate descending order of milligram potency in treating schizophrenia and eliciting extrapyramidal side effects (20, 21). t Data are derived from the study of Miller et al. (5) and Clement-Courmier et al. (4). Karobath and Leitich (6) obtained similar results. t Values for promazine represent the major discrepancy between the data of Miller et al. (5) and Clement-Courmier et al. (4). An abundance of pharmacologic data (1, 2, 24) indicates that promazine is only a weak dopamine receptor blocker, which is most consistent with the results of Miller et al. (5).
piperazine phenothiazines should be more potent clinically than alkylamino phenothiazines is borne out by extensive clinical experience showing that the piperazine phenothiazines are consistently more potent in their antischizophrenic effects, their ability to elicit extrapyramidal side effects, and in their affinity for the dopamine-sensitive adenylate cyclase than alkylamino phenothiazines (Fig. 4 and TAble 1) (5, 20, 21). Hydroxyethylpiperazine side chain phenothiazines are in turn more potent in inhibiting the dopamine-sensitive adenylate cyclase and in their clinical actions than are the simple piperazine side chain phenothiazines (Table 1) (5, 20, 21). Space-filling models indicate that hydroxyethylpiperazine side chain phenothiazines would be anticipated to display more favorable Van der Waal's interactions with ring A than simple piperazines (Fig. 2). Relatively rigid analogs of the phenothiazines possess interesting structure-activity relationships. The thioxanthenes
and xanthenes contain an exocyclic double bond which replaces nitrogen in the ring system (Fig. 4). If the ring is substituted, they can exist in a cis form with the side chain turned toward ring A or in a trans form with the side chain turned toward ring C. The cis forms of these compounds are considerably more potent neuroleptics than the trans isomers
R2
N
R2
B
CH
CIH2
C&H
R.I THIOXANTHENES
PHENOTHIAZINES
.-CIS
R.
e -TRANS
-RHI-CH2N CHo -CHg-CH2-NH O\CH3
ALKYLAMINO (A)
PIPERAZINE (P) -CH2-CH2-
.,,CH2-CH2 H
C 3
HYDROXYETHYLPIPERAZINE (HEP)
CH2-CH2
C -N
@DCH2-CH2j
HYDROGEN -H CHLORINE -CI
FIG. 3. Drawings of chlorpromazine as in Fig. 2a, with dopamine, in an extended conformation, superimposed.
2 2-CHCHLOH
/F
TRIFLUOROMETHYL -C-F IF
FIG. 4. Structure of phenothiazine and thioxanthene drugs.
Proc. Nat. Am& Sci. USA 72
Medical Sciences: Feinberg and Snyder
1902
a -601-
0.2
0.4 0.6
0.8
1.0
-1201.
2 0.4 0.6
0.6
l20d 0.4
60'
0.3- i04
60
.0
0-
60-
I8O-
120-
-120-
0fI0
-60-
T. PROMAZINE
~~~~-
1.04
0~~~~~~~~~~
-6(
0.02
0.6 0.8
-12( 0
v
1.0 3 2 a . ~~~~~~06
~~~
.
0
0.4-
18(
0.3 0.6 0.8
0.6
~~~~~~~~~~~~~~1.0
121 0
61 0.2
0.4
0.6
0.8
0~~~~~~~~~~~~~. 0o
60
120"
80"
-120"
-60"
C-o
CHLORPROMAZINE
CM
TRIFLUPROMAZINE
FIG. 5. Van der Waal's potential energy contour maps. Isoenergy curves are in kcal/mol above the absolute minimum energy. The coordinate axes indicate bond torsion angles ri and 72. (a) Promazine; (b) chlorpromazine; (c) triflupromazine.
(1975)
(20, 21). Our original model (7) had posited that the ring A substituent "directs" the side chain away from the midline to assume a dopamine-like conformation by tilting toward either ring C or ring A. Accordingly, compounds with the exocyclic double bond whose side chain tilts away from the midline regardless of the presence or absence of any ring substituent, might not be expected to display differential potency in the cis or trans forms. We propose that by directing the side chain away from the midline, the exocyclic double bond of cis forms of drugs such as the thioxanthenes provides a closer approximation of the side chain to ring A, enabling the 2-substituent to enter into Van der Waal's attractive forces with the side chain. This proposal accounts for the greater neuroleptic activity of the cis than of the trans forms of thioxanthenes. Moreover, this reasoning suggests that the exocyclic double bond synergizes with other structural features such as a trifluoromethyl group at the 2-position and a piperazine side chain, thereby enhancing phenothiazine potency. In accordance with this prediction, both clinically and in effects on the dopamine sensitive adenylate cyclase, phenothiazine analogs with an exocyclic double bond are more potent than corresponding compounds that lack them (Table 1) (5, 20, 21). Of numerous phenothiazines examined for inhibition of the dopamine-sensitive adenylate cyclase, the most potent agent is flupenthixol, which combines all the chemical features that would enhance the ability of phenothiazines to assume the dopamine conformation. Thus, flupenthixol possesses a trifluoromethyl substituent on the A ring, an hydroxyethylpiperazine side chain, and an exocyclic double bond. Potential energy calculations were performed on the following three phenothiazines with differing 2-substituents but the same alkylamino side chain: promazine, chlorpromazine, and triflupromazine (Fig. 4). Potential energy contour maps reveal four conformational zones for the first two bonds of the side chain (Fig. 5). In all cases, the side chain is directed toward the convexity of the ring system. Direction of the side chain toward the concavity of the ring system incurred Van der Waal's repulsions [1.9-2.1 kcal/mol (7950-8790 joules/mol)] using the generally accepted angle of t = 1400 between rings A and C (17, 18). The global energy minimum in our calculations is zone no. 1 for all three drugs. However, the energy contours for conformations no. 1, no. 3, and no. 4 are closely similar. By contrast, the energy contour for conformation no. 2, which represents the dopamine-like conformation depicted in Fig. 3, changes systematically and progressively among the three drugs. As one proceeds from promazine to chlorpromazine to triflupromazine, the size and depth of the energy wells increase, with the relative minimum energy for zone no. 2 decreasing from 0.4 to 0.3 to 0.1 kcal/mol (1674 to 1255 to 418.4 joules/mol). Indeed, the energy minimum for conformation no. 2 in triflupromazine is essentially the same as the absolute energy minimum for this compound. Similarly, the areas of the 0.6 kcal (2510 joules) contour for chlorpromazine and triflupromazine, respectively, are 2.3 and 3.5 times that for promazine. Moreover, for the phenothiazines examined here, the distance at the minimum energy conformation of zone no. 2 from the side chain nitrogen to the 2-substituent is optimal for Van der Waal's attractions, e.g., 3.30 A for chlorpromazine (9). The N-methyl substituents would be expected to enhance Van der Waal's attractions. To determine the extent to which the energetic differences between promazine, chlorpromazine, and triflupromazine
Proc. Nat. Acad. Sci. USA 72
(1975)
influence the relative probabilities for assuming conformation no. 2, we calculated the integral of e-VlkT over conformational zone no. 2, assuming the minima and contour areas of Fig. 5 delimited to an approximately ellipsoidal 0.6 kcal (2510 joule) contour with T = 370C. The calculated relative probability for assuming conformation no. 2 increases 3.9-fold from promazine to chlorpromazine and 3.0-fold from chlorpromazine to triflupromazine, correlating with the relative potencies of these drugs in inhibiting the dopamine-sensitive adenylate cyclase, especially using the data of Miller et al. (5). Thus, whereas the other conformational zones change very little in proceeding from promazine to chlorpromazine to triflupromazine, conformational zone no. 2, the dopaminelike conformation, becomes increasingly desirable. This supports the hypothesis derived from molecular models that a chlorine 2-substituent promotes the dopamine-like conformation for phenothiazine drugs. Moreover, a trifluoromethyl 2-substituent favors the dopamine-like conformation to a greater extent than a chlorine substituent. Zone no. 2, the conformation proposed by us, represents one of the relative energy minima (Ti = -68°, T-2 = 1350), but it is not the conformation with the absolute least energy. In the case of triflupromazine, the potential energy of the proposed conformation roughly equals the absolute minimum, and for piperazines and hydroxyethylpiperazines it might be the absolute minimum. There is some precedent for biologically active compounds exerting their physiological effects in conformations other than the one with the absolute energy minimum (22). For instance, the conformation of triiodothyronine that binds to thyroglobulin is not the absolute minimal energy conformation (23). Interestingly, the observed crystal structure of chlorpromazine (18) falls within zone no. 2, with (TI, TO) = (-690, 1640) but not at the relative minimum. A thiethyl 2-substituent displaces the side chain from a dopamine-like conformation, so that the crystal conformation of thiethylperazine (17), (TI, T2) = (-1430, 1750), lies in a 1.2 kcal/mol (5021 joules/mol) zone in our contour maps for the compounds with smaller 2-substituents. Indeed, thiethylperazine is weak in both antipsychotic actions and Parkin-
sonian side effects (20, 21). Our potential energy calculations were performed while varying Ti, T2, T3, T4, and the trifluoromethyl group, to consider all conceivable conformations, but the graphs depict only Tj and T2. T3 is also important in determining the position of the amine nitrogen, but we found many possible values for T3 within the zones depicted for Ti and T2. This variability in 73 is consistent with a gauche, trans, or intermediate conformation for dopamine that would correspond to the ring A and the amine nitrogen of the phenothiazines. Fig. 3 depicts one such correspondence. The present calculations consider only Van der Waal's forces, not electrostatic forces or interactions with the receptor medium. Their close correspondence to biological and clinical data suggests that, for phenothiazines, Van der Waal's interactions may be major determinants of the drugs'
Phenothiazines and Dopamine
1903
conformation at receptor sites, despite the fact that in aqueous solution these forces may not play major roles. This work was supported by USPHS Grant MH-18501, grants of the John A. Hartford and Scottish Rite Foundations, and USPHS RSDA Award MH-33128 to S.H.S. We thank Drs. L. M. Amzel, H. E. Bosshard, J. Heinz, R. Meltzer, and R. J. Feldmann for computational facilities and helpful suggestions. A.P.F. is a Year IV medical student. 1. Carlsson, A. & Lindquist, M. (1963) Acta Pharmacol. Toxicol. 20, 140-144. 2. Snyder, S. H., Banerjee, S. P., Yamamura, H. I. & Greenberg, D. (1974) Science 184, 1243-1253. 3. Kebabian, J. W., Petzold, G. L. & Greengard, P. (1972) Proc. Nat. Acad. Sci. USA 69, 2145-2149. 4. Clement-Courmier, Y. S., Kebabian, J. W., Petzold, G. L. & Greengard, P. (1974) Proc. Nat. Acad. Sci. USA 71, 11131117. 5. Miller, R. J., Horn, A. S. & Iversen, L. L. (1974) Mol. Pharmacol. 10, 759-766. 6. Karobath, M. & Leitich, H. (1974) Proc. Nat. Acad. Sci. USA 71, 2915-2918. 7. Horn, A. S. & Snyder, S. H. (1971) Proc. Nat. Acad. Sci. USA 68, 2325-2328. 8. Kitaygorodskii, A. I. (1961) Tetrahedron 14, 230-236. 9. Venkatachalam, C. M. & Ramachandran, G. N. (1967) in Conformation of Biopolymers, ed. Ramachandran, G. N. (Academic Press, New York), pp. 83-105. 10. Ramachandran, G. N. & Sasisekharan, V. (1968) Advan. Protein Chem. 23, 284-437. 11. Dickson, C. (1972) Technical Memorandum No. 156 (Washington Univ. Computer Systems Laboratory). 12. Marshall, G. R., Bosshard, H. E. & Ellis, R. A. (1973) in Proc. NATO Advan. Study Inst. on Computer Representation and Manipulation of Chem. Info. (Wiley, New York), pp. 203-237. 13. Bosshard, H. E., Barry, C. D., Fritsch, J. M., Ellis, R. A. & Marshall, G. R. (1972) Proc. Computer Summer Simulation Conference, San Diego, pp. 581-585. 14. Marshall, G. R. & Bosshard, H. E. (1972) Circulation Res. 31 (Suppl. 2), 143-150. 15. Leach, S. J., Nemethy, G. & Scheraga, H. A. (1966) Biopoly. Symp. 4, 369-407. 16. Feinberg, A. P. (1972) Technical Memorandum No. 166 (Washington Univ. Computer Systems Laboratory). 17. McDowell, J. J. H. (1970) Acta Crystallogr. Sect. B 26, 954964. 18. McDowell, J. J. H. (1969), Acta Crystallogr. Sect. B 25, 2175-2181. 19. Feldmann, R. J. (1973) in Proc. NATO Advan. Study Inst. on Computer Representation and Manipulation of Chem. Info. (Wiley, New York), pp. 55-81. 20. Gordon, M. (1967) in Psychopharmacological Agents (Academic Press, New York), Vol. II, pp. 2-198. 21. Klein, D. F. & Davis, J. M. (1969) Diagnosis and Drug Treatment of Psychiatric Disorders (Williams and Wilkins, Baltimore, Md.). 22. Green, J. P., Johnson, C. L. & Kang, S. (1974) Annu. Rev. Pharmacol. 14, 319-342. 23. Schussler, G. C. (1972) Science 178, 172-174. 24. Jannsen, P. A. J. (1965) in International Review of Neurobiology, eds. Pfeiffer, C. C. & Smythies, J. R. (Academic Press, New York-London), Vol. 8, pp. 221-263. 25. Klyne, W. & Prelog, V. (1960) Experienitia 16, 521-568.
Phenothiazine Drugs: Structure-Activity Relationships Explained by a Conformation That Mimics Dopamine (catecholamine/neurotransmitter/receptor/chlorpromazine/schizophrenia)
ANDREW P. FEINBERG AND SOLOMON H. SNYDER Departments of Pharmacology and Experimental Therapeutics and Psychiatry and the Behavioral Sciences, Johns Hopkins University School of Medicine, Baltimore, Maryland 21205
Communicated by Seymour S. Kety, February 21, 1975 ABSTRACT The antischizophrenic activity of phenothiazine drugs and their tendency to elicit extrapyramidal symptoms are thought to involve blockade of synaptic dopamine receptors in the brain. Space filling molecular models show how favorable Van der Waal's interactions between the side chain amino of phenothiazines and the 2substituent on ring A can promote a conformation mimicking dopamine. These Van der Waal's attractive forces can explain (i) the greater potency of drugs with trifluoromethyl rather than chlorine as a 2-substituent; (ii) the enhanced activity of phenothiazines with piperazine instead of alkylamino side chains; (iii) the increased potency associated with hydroxyethylpiperazines as contrasted to piperazine side chains ; (iv) the greater potency of cis rather than trans thioxanthenes; and (v) the crucial location of the ring A substituent at carbon no. 2. Potential energy calculations support the observations with molecular models and suggest an active conformation for the phenothiazines.
An abundance of recent research suggests that a major mechanism whereby antischizophrenic phenothiazine drugs exert their therapeutic actions and extrapyramidal side effects involves a blockade of synaptic receptor sites for dopamine in the brain (1, 2). The activity of a dopamine-sensitive adenylate cyclase correlates with dopamine receptor activity (3). Relative potencies of several phenothiazine and related drugs as inhibitors of the dopamine-sensitive adenylate cyclase parallel their antischizophrenic potency (4-6). Earlier we proposed a molecular model wherein dopamine could be superimposed upon a portion of the chlorpromazine molecule (7). While explaining some structure-activity characteristics of phenothiazines, this model did not deal with important features such as the greater potencies of phenothiazines with piperazine rather than alkylamino side chains, nor did it explain the greater potency of trifluoromethyl than of chlorine ring substituents. Moreover, it provided no explanation for the mechanism whereby the A ring substituent caused the side chain to tilt toward the A ring, nor did it explain why the A ring substituent must be located in the number 2 position. In the present study, Corey-Pauling-Koltun molecular models and computer calculations support a model in which phenothiazines assume a conformation that mimics that of dopamine, explaining the role of trifluoromethyl, piperazine, and hydrox-yethvlpiperazine groups as well as the mechanism whereby A ring substituents influence the side chain. MATERIALS AND METHODS
Molecular models of promazine, chlorpromazine, triflupromazine, prochlorperazine, trifluoperazine, thiethylperazine,
1899
perphenazine, and fluphenazine were constructed with CoreyPauling-Koltun kits (Ealing Corp.; Cambridge, Mass.). To confirm the apparent influence of Van der Waal's forces on the side chain conformation, we performed potential energy calculations on the following three compounds with differing ring substituents but the same side chain: promazine, chlorpromazine, and triflupromazine. Calculations were performed on alkylamino side chain phenothiazines, since the great increase in conformational variables introduced by a flexible piperidine or piperazine ring was beyond the practical capability of the computer. The calculations were performed on a PDP-12 computer with floating point processor and 8K of memory, using a form of the Buckingham 6-exp potential function proposed by A. I. Kitaygorodskii (8): V = 3.5 (8600 e-13Z - 0.04 Z-6) where z is the ratio of the distance between atom centers and G. N. Ramachandran's K2 values for Van der Waal's radii (9). This function has shown good agreement with observed x-ray crystal structures (10). Two groups of computer programs, developed at Washington University Computer Systems Laboratory, were used. The first group, CHEMAST (11, 12), was used to generate lists of atomic coordinates, atom types, connectivity and rotation parameters, and stereoscopic images of the molecules on a point-plotting cathode ray tube. The second group, BURLESK (13, 14), was used to perform iterative rotations over the bonds indicated in Fig. 1 and the trifluoromethyl group in triflupromazine, computing the potential energy between all atom pairs whose relative orientations were changed by the rotations; and to display the data on the cathode ray tube as isoenergy curves. To obviate calculations on absurd conformers, the programs made a "bump check", for which Van der Waal's radii were set at 90% of those used by Leach et al. (15). The data were compared to control data obtained without a "bump check" to ascertain that no legitimate conformers were excluded. A special connectivity table was prepared for the phenothiazine ring system (16), and its bond lengths and bond angles were obtained from the published crystal structure of thiethylperazine (17). It might have been more natural to use published data on chlorpromazine (18), but these yielded bizarre ring atom placements. For each molecule, between 16,000 and 150,000 conformations were examined using 5-15 hr of computer time. In criti-
1900
Medical Sciences: Feinberg and Snyder
19l 20
(ECH2\
I*He
(1975)
5
4
CH3
Proc. Nat. Acad. Sci. USA 72
9
-C1
'qH2
T3 162
4
CH 3 19
T,
a
T(C,,-NO-C15"C16 )
T2=aT (NI0_CI5 -C6 -CI7 ) T3 T(Ci5-CI6-CI7 -NIS) a
T(C16-CI7
-N18 -C19)
FIG. 1. Rotatable bonds of chlorpromazine, drawn with = T4 = 00 and T3 = 1800. The convexity of the phenothiazine ring system is upward. The torsion angle r(A-X-Y-B) is the dihedral angle between planes AXY and XYB, measured clockwise when viewed down the bond X-Y (25). T1. =2
cal regions, the data were compared to control data obtained density of conformations to confirm their completeness. A Hewlett-Packard 7005B X-Y Recorder was used to draw the isoenergy curves on graph paper, and they were then retraced. The figures of ball and stick models were made with a DECSYSTEM 10 time-sharing computer, AGT-30 ADAGE graphics system, and model 200 ZETA plotter, using the program package CHEM (19), and they were then redrawn by an artist. over a greater
C
RESULTS AND DISCUSSION
Space-filling models of chlorpromazine (Fig. 2) reveal that tilting of the side chain toward ring A permits favorable Van der Waal's interactions of the side chain with the chlorine substituent. These Van der Waal's attractive forces would be greatly lessened in the case of promazine, which lacks the chlorine substituent. This conformation permits the superimposition of dopamine (Fig. 3), thus explaining how chlorpromazine can interact with dopamine receptors. The spacefilling models provide a causal explanation for the folding of the side chain toward ring A. The notion that Van der Waal's attractions between the side chain and the ring A substituent could account for the ability of phenothiazines to assume the dopamine-like conformation allows several predictions about the structure of phenothiazines that might be expected to be more or less potent in interactions with the dopamine receptors. For instance, this model requires that the ring A substituent be located at the number 2 position. A 1-substituent would sterically hinder the ability of the side chain to approach ring A, while a 3substituent would be too distant from the side chain to provide Van der Waal's attractions of a significant degree. Extensive experience with many phenothiazine structures has shown that optimal neuroleptic activity occurs only when the ring A substituent is in the 2-position (20, 21). The nature of the 2-substituents in our model would be expected to influence the conformation of the side chain. A trifluoromethyl substituent provides a greater number of favorable Van der Waal's contacts with the side chain than
FIG. 2. Phenothiazines with the side chain "tilted" toward the A ring, drawn with relatively small atom size for ease of viewing. ri and T2 are set at (-68°, 1350), the relative minimum of conformational zone no. 2 in Fig. 5. (a) Chlorpromazine (the side chain N to Cl distance is 3.3 A); (b) triflupromazine (the distance of the side chain N to the closest fluorines is 3.0-3.5 A); (c) trifluoperazine; (d) fluphenazine.
the chlorine substituent. Our model predicts that phenothiazines with trifluoromethyl substituents should be more potent than those with chlorine substituents. Of the clinically used phenothiazines, those with trifluoromethyl substituents at position 2 of ring A are uniformly more potent than those with chlorine substituents, both clinically and in their inhibition of the dopamine-sensitive adenylate cyclase (4-6, 20, 21). The nature of the atom in proximity to the side chain amine might also be anticipated to influence the conformation of the phenothiazine side chain. A piperazine side chain affords more Van der Waal's contacts with the 2-substituent than does an alkylamino side chain (Fig. 2). The resultant predictions that
Proc. Nat. Acad. Sci. USA 72
Phenothiazines and Dopamine
(1975)
1901
TABLE 1. Phenothiazine and thioxanthene drug structures and effects on a dopamine-sensitive adenylate cyclase
Relative potency in inhibiting
Drug
Ca-Flupenthixol (aO6)-Flupenthixol Fluphenazine Trifluoperazine Triflupromazine a-Clopenthixol Perphenazine a-Chlorprothixene Prochlorperazine Chlorpromazine Promazine
jS-Chlorprothixene ,6-Clopenthixol
fl-Flupenthixol
Ring thio. thio. pheno. pheno. pheno. thio. pheno. thio. pheno. pheno. pheno. thio. thio. thio.
R, HEP HEP HEP P A HEP HEP A P A A A HEP HEP
dopamine-sensitive adenylate cyclase of rat corpus striatum (chlorpromazine = 100) t Results of Results of Miller and Clement-Courmier Iversen (5) et al. (4) 4545 1333 1087 857 250 600
Relative clinical potency* 1 2 3 4 5 6 7 8 9 10 11 12 13 14
R2 CF3 CF3 CF3 CF3 CF3 Cl Cl Cl C1 Cl H Cl Cl CF3
150
303 128 45 100 17 5 1.7 1
120 100
171t
Inhibition of dopamine-sensitive adenylate cyclase, relative to chlorpromazine = 100. R1 and R2 refer to Fig. 4. Abbreviations: thio. = thioxanthene, pheno. = phenothiazine, A = alkylamino, P = piperazine, HEP = hydroxyethylpiperazine. * Drugs are listed in approximate descending order of milligram potency in treating schizophrenia and eliciting extrapyramidal side effects (20, 21). t Data are derived from the study of Miller et al. (5) and Clement-Courmier et al. (4). Karobath and Leitich (6) obtained similar results. t Values for promazine represent the major discrepancy between the data of Miller et al. (5) and Clement-Courmier et al. (4). An abundance of pharmacologic data (1, 2, 24) indicates that promazine is only a weak dopamine receptor blocker, which is most consistent with the results of Miller et al. (5).
piperazine phenothiazines should be more potent clinically than alkylamino phenothiazines is borne out by extensive clinical experience showing that the piperazine phenothiazines are consistently more potent in their antischizophrenic effects, their ability to elicit extrapyramidal side effects, and in their affinity for the dopamine-sensitive adenylate cyclase than alkylamino phenothiazines (Fig. 4 and TAble 1) (5, 20, 21). Hydroxyethylpiperazine side chain phenothiazines are in turn more potent in inhibiting the dopamine-sensitive adenylate cyclase and in their clinical actions than are the simple piperazine side chain phenothiazines (Table 1) (5, 20, 21). Space-filling models indicate that hydroxyethylpiperazine side chain phenothiazines would be anticipated to display more favorable Van der Waal's interactions with ring A than simple piperazines (Fig. 2). Relatively rigid analogs of the phenothiazines possess interesting structure-activity relationships. The thioxanthenes
and xanthenes contain an exocyclic double bond which replaces nitrogen in the ring system (Fig. 4). If the ring is substituted, they can exist in a cis form with the side chain turned toward ring A or in a trans form with the side chain turned toward ring C. The cis forms of these compounds are considerably more potent neuroleptics than the trans isomers
R2
N
R2
B
CH
CIH2
C&H
R.I THIOXANTHENES
PHENOTHIAZINES
.-CIS
R.
e -TRANS
-RHI-CH2N CHo -CHg-CH2-NH O\CH3
ALKYLAMINO (A)
PIPERAZINE (P) -CH2-CH2-
.,,CH2-CH2 H
C 3
HYDROXYETHYLPIPERAZINE (HEP)
CH2-CH2
C -N
@DCH2-CH2j
HYDROGEN -H CHLORINE -CI
FIG. 3. Drawings of chlorpromazine as in Fig. 2a, with dopamine, in an extended conformation, superimposed.
2 2-CHCHLOH
/F
TRIFLUOROMETHYL -C-F IF
FIG. 4. Structure of phenothiazine and thioxanthene drugs.
Proc. Nat. Am& Sci. USA 72
Medical Sciences: Feinberg and Snyder
1902
a -601-
0.2
0.4 0.6
0.8
1.0
-1201.
2 0.4 0.6
0.6
l20d 0.4
60'
0.3- i04
60
.0
0-
60-
I8O-
120-
-120-
0fI0
-60-
T. PROMAZINE
~~~~-
1.04
0~~~~~~~~~~
-6(
0.02
0.6 0.8
-12( 0
v
1.0 3 2 a . ~~~~~~06
~~~
.
0
0.4-
18(
0.3 0.6 0.8
0.6
~~~~~~~~~~~~~~1.0
121 0
61 0.2
0.4
0.6
0.8
0~~~~~~~~~~~~~. 0o
60
120"
80"
-120"
-60"
C-o
CHLORPROMAZINE
CM
TRIFLUPROMAZINE
FIG. 5. Van der Waal's potential energy contour maps. Isoenergy curves are in kcal/mol above the absolute minimum energy. The coordinate axes indicate bond torsion angles ri and 72. (a) Promazine; (b) chlorpromazine; (c) triflupromazine.
(1975)
(20, 21). Our original model (7) had posited that the ring A substituent "directs" the side chain away from the midline to assume a dopamine-like conformation by tilting toward either ring C or ring A. Accordingly, compounds with the exocyclic double bond whose side chain tilts away from the midline regardless of the presence or absence of any ring substituent, might not be expected to display differential potency in the cis or trans forms. We propose that by directing the side chain away from the midline, the exocyclic double bond of cis forms of drugs such as the thioxanthenes provides a closer approximation of the side chain to ring A, enabling the 2-substituent to enter into Van der Waal's attractive forces with the side chain. This proposal accounts for the greater neuroleptic activity of the cis than of the trans forms of thioxanthenes. Moreover, this reasoning suggests that the exocyclic double bond synergizes with other structural features such as a trifluoromethyl group at the 2-position and a piperazine side chain, thereby enhancing phenothiazine potency. In accordance with this prediction, both clinically and in effects on the dopamine sensitive adenylate cyclase, phenothiazine analogs with an exocyclic double bond are more potent than corresponding compounds that lack them (Table 1) (5, 20, 21). Of numerous phenothiazines examined for inhibition of the dopamine-sensitive adenylate cyclase, the most potent agent is flupenthixol, which combines all the chemical features that would enhance the ability of phenothiazines to assume the dopamine conformation. Thus, flupenthixol possesses a trifluoromethyl substituent on the A ring, an hydroxyethylpiperazine side chain, and an exocyclic double bond. Potential energy calculations were performed on the following three phenothiazines with differing 2-substituents but the same alkylamino side chain: promazine, chlorpromazine, and triflupromazine (Fig. 4). Potential energy contour maps reveal four conformational zones for the first two bonds of the side chain (Fig. 5). In all cases, the side chain is directed toward the convexity of the ring system. Direction of the side chain toward the concavity of the ring system incurred Van der Waal's repulsions [1.9-2.1 kcal/mol (7950-8790 joules/mol)] using the generally accepted angle of t = 1400 between rings A and C (17, 18). The global energy minimum in our calculations is zone no. 1 for all three drugs. However, the energy contours for conformations no. 1, no. 3, and no. 4 are closely similar. By contrast, the energy contour for conformation no. 2, which represents the dopamine-like conformation depicted in Fig. 3, changes systematically and progressively among the three drugs. As one proceeds from promazine to chlorpromazine to triflupromazine, the size and depth of the energy wells increase, with the relative minimum energy for zone no. 2 decreasing from 0.4 to 0.3 to 0.1 kcal/mol (1674 to 1255 to 418.4 joules/mol). Indeed, the energy minimum for conformation no. 2 in triflupromazine is essentially the same as the absolute energy minimum for this compound. Similarly, the areas of the 0.6 kcal (2510 joules) contour for chlorpromazine and triflupromazine, respectively, are 2.3 and 3.5 times that for promazine. Moreover, for the phenothiazines examined here, the distance at the minimum energy conformation of zone no. 2 from the side chain nitrogen to the 2-substituent is optimal for Van der Waal's attractions, e.g., 3.30 A for chlorpromazine (9). The N-methyl substituents would be expected to enhance Van der Waal's attractions. To determine the extent to which the energetic differences between promazine, chlorpromazine, and triflupromazine
Proc. Nat. Acad. Sci. USA 72
(1975)
influence the relative probabilities for assuming conformation no. 2, we calculated the integral of e-VlkT over conformational zone no. 2, assuming the minima and contour areas of Fig. 5 delimited to an approximately ellipsoidal 0.6 kcal (2510 joule) contour with T = 370C. The calculated relative probability for assuming conformation no. 2 increases 3.9-fold from promazine to chlorpromazine and 3.0-fold from chlorpromazine to triflupromazine, correlating with the relative potencies of these drugs in inhibiting the dopamine-sensitive adenylate cyclase, especially using the data of Miller et al. (5). Thus, whereas the other conformational zones change very little in proceeding from promazine to chlorpromazine to triflupromazine, conformational zone no. 2, the dopaminelike conformation, becomes increasingly desirable. This supports the hypothesis derived from molecular models that a chlorine 2-substituent promotes the dopamine-like conformation for phenothiazine drugs. Moreover, a trifluoromethyl 2-substituent favors the dopamine-like conformation to a greater extent than a chlorine substituent. Zone no. 2, the conformation proposed by us, represents one of the relative energy minima (Ti = -68°, T-2 = 1350), but it is not the conformation with the absolute least energy. In the case of triflupromazine, the potential energy of the proposed conformation roughly equals the absolute minimum, and for piperazines and hydroxyethylpiperazines it might be the absolute minimum. There is some precedent for biologically active compounds exerting their physiological effects in conformations other than the one with the absolute energy minimum (22). For instance, the conformation of triiodothyronine that binds to thyroglobulin is not the absolute minimal energy conformation (23). Interestingly, the observed crystal structure of chlorpromazine (18) falls within zone no. 2, with (TI, TO) = (-690, 1640) but not at the relative minimum. A thiethyl 2-substituent displaces the side chain from a dopamine-like conformation, so that the crystal conformation of thiethylperazine (17), (TI, T2) = (-1430, 1750), lies in a 1.2 kcal/mol (5021 joules/mol) zone in our contour maps for the compounds with smaller 2-substituents. Indeed, thiethylperazine is weak in both antipsychotic actions and Parkin-
sonian side effects (20, 21). Our potential energy calculations were performed while varying Ti, T2, T3, T4, and the trifluoromethyl group, to consider all conceivable conformations, but the graphs depict only Tj and T2. T3 is also important in determining the position of the amine nitrogen, but we found many possible values for T3 within the zones depicted for Ti and T2. This variability in 73 is consistent with a gauche, trans, or intermediate conformation for dopamine that would correspond to the ring A and the amine nitrogen of the phenothiazines. Fig. 3 depicts one such correspondence. The present calculations consider only Van der Waal's forces, not electrostatic forces or interactions with the receptor medium. Their close correspondence to biological and clinical data suggests that, for phenothiazines, Van der Waal's interactions may be major determinants of the drugs'
Phenothiazines and Dopamine
1903
conformation at receptor sites, despite the fact that in aqueous solution these forces may not play major roles. This work was supported by USPHS Grant MH-18501, grants of the John A. Hartford and Scottish Rite Foundations, and USPHS RSDA Award MH-33128 to S.H.S. We thank Drs. L. M. Amzel, H. E. Bosshard, J. Heinz, R. Meltzer, and R. J. Feldmann for computational facilities and helpful suggestions. A.P.F. is a Year IV medical student. 1. Carlsson, A. & Lindquist, M. (1963) Acta Pharmacol. Toxicol. 20, 140-144. 2. Snyder, S. H., Banerjee, S. P., Yamamura, H. I. & Greenberg, D. (1974) Science 184, 1243-1253. 3. Kebabian, J. W., Petzold, G. L. & Greengard, P. (1972) Proc. Nat. Acad. Sci. USA 69, 2145-2149. 4. Clement-Courmier, Y. S., Kebabian, J. W., Petzold, G. L. & Greengard, P. (1974) Proc. Nat. Acad. Sci. USA 71, 11131117. 5. Miller, R. J., Horn, A. S. & Iversen, L. L. (1974) Mol. Pharmacol. 10, 759-766. 6. Karobath, M. & Leitich, H. (1974) Proc. Nat. Acad. Sci. USA 71, 2915-2918. 7. Horn, A. S. & Snyder, S. H. (1971) Proc. Nat. Acad. Sci. USA 68, 2325-2328. 8. Kitaygorodskii, A. I. (1961) Tetrahedron 14, 230-236. 9. Venkatachalam, C. M. & Ramachandran, G. N. (1967) in Conformation of Biopolymers, ed. Ramachandran, G. N. (Academic Press, New York), pp. 83-105. 10. Ramachandran, G. N. & Sasisekharan, V. (1968) Advan. Protein Chem. 23, 284-437. 11. Dickson, C. (1972) Technical Memorandum No. 156 (Washington Univ. Computer Systems Laboratory). 12. Marshall, G. R., Bosshard, H. E. & Ellis, R. A. (1973) in Proc. NATO Advan. Study Inst. on Computer Representation and Manipulation of Chem. Info. (Wiley, New York), pp. 203-237. 13. Bosshard, H. E., Barry, C. D., Fritsch, J. M., Ellis, R. A. & Marshall, G. R. (1972) Proc. Computer Summer Simulation Conference, San Diego, pp. 581-585. 14. Marshall, G. R. & Bosshard, H. E. (1972) Circulation Res. 31 (Suppl. 2), 143-150. 15. Leach, S. J., Nemethy, G. & Scheraga, H. A. (1966) Biopoly. Symp. 4, 369-407. 16. Feinberg, A. P. (1972) Technical Memorandum No. 166 (Washington Univ. Computer Systems Laboratory). 17. McDowell, J. J. H. (1970) Acta Crystallogr. Sect. B 26, 954964. 18. McDowell, J. J. H. (1969), Acta Crystallogr. Sect. B 25, 2175-2181. 19. Feldmann, R. J. (1973) in Proc. NATO Advan. Study Inst. on Computer Representation and Manipulation of Chem. Info. (Wiley, New York), pp. 55-81. 20. Gordon, M. (1967) in Psychopharmacological Agents (Academic Press, New York), Vol. II, pp. 2-198. 21. Klein, D. F. & Davis, J. M. (1969) Diagnosis and Drug Treatment of Psychiatric Disorders (Williams and Wilkins, Baltimore, Md.). 22. Green, J. P., Johnson, C. L. & Kang, S. (1974) Annu. Rev. Pharmacol. 14, 319-342. 23. Schussler, G. C. (1972) Science 178, 172-174. 24. Jannsen, P. A. J. (1965) in International Review of Neurobiology, eds. Pfeiffer, C. C. & Smythies, J. R. (Academic Press, New York-London), Vol. 8, pp. 221-263. 25. Klyne, W. & Prelog, V. (1960) Experienitia 16, 521-568.
