73% - April 2990 [Voi. 121

how valid is feiffer’s rule? ~~~ti~~ers:

Richard Barlow in tire 1950s Pfeiffer observed thf trte relative potencies of fire enanfio,ners (mirror-image isot*rers) of sezyeral drugs appeared to be related to fhe dose of the racemate used clinically. The difference was greatest with fhe most active compc~ds and he suggested reasons for what has become known as ‘Pfeiffer’s rule’. ln this shrf arfick Dick Barlow poinfs out that there is a corollary to the rule - fkaf the activity of fhe weaker ena~~fiom~r is defermif~ed by tke acfioify of fke more potent one - which is intzfifiveiy i~nprobable. With more examples and in a simpler sifuafion, fhe picture is more co.mplex and there are marked exceptiofrs. The ideas behirui Pfeiffer’s rule overlook differencesiti molecular flexibility and there is a need for a revrsed model that takes entropy differences info account, particalariy as recent developments in fke elucidation of receptor sfrucfure are likely to revive interest in Pfeiffer’s rule. In a note entitled ‘Optical isomerism and pharmacologica action, a generalization’, C. C. Pfeiffer’ observed a striking correlation between the relative potencies of the mirror-image forms (enantiomers) of some clinicafly used drugs and the average human dose of the drug, given as the racemate. With the more active drugs there was a bigger difference between the activities of the enantiomers. For 14 compounds the optical isomerit ratios (Y) plotted on a logarithmic scale against the average human dose (X), also on a logarithmic scale (Fig. l), appeared to lie on a straight line described by the equation: log Y = 1.19 - 0.354 (log X) The correlation coefficient r appeared to be > 0.9. Pfeiffer’s observation was the more remarkable because no allowance was made for differences in the metabolism of the enantiome~, differences between the receptors involved and differences in the type of action of the drugs (some were agonists, others were antagonists). An additional problem is ensuring that the weaker enantiomer is stereochemically pure and does not conR. 6. Barlow is Render irr Chenricnl Pbnnnncology at the Departrt~entof Phwnmology, University University

of Bristol, Tk Medical School, Walk, BrisfolBSS ITD, UK.

tain traces of the stronger enantiomer. On the basis of the corrc!ations generated by these data Pfeiffer’s rule appears to make sense. In the simplest situation, where activity depends only on binding, as with competitive antagonists, it is plausible that the higher the affinity of a compound, the more it matters how groups are arranged about a chiral centre. If mirrorimage forms have different affinity, binding must involve an interaction at three points (at Ieast); the less active form would perhaps only form an attachmznr at two of them. If the binding is measured as an affinity constant K then log K is directly proportional to the Gibbs free energy of binding [from the van? Hoff relation -AC = R T(ln K), where R is the gas constant and T is the absolute temperature]. The difference between log K for the enantiomers is the logarithm of their relative potency. Various names have been given to the relative potency of mirror-image forms: if it is referred to as the stereospecific index (SSI; Ref. 2) and this is 100, then log SSI is 2 and the difference between the free energy of binding of the two forms is 11.8 kJ mol-r at 37°C.

Further implications There is, however, a corollary to

Pfeiffer’s rule which it is much more difficult to believe. If 1ogSSI is linearly related to 1ogK for the racemate, it must also be at least approximately linearly related to log K for the stronger enantiomer: this should not be more than twice as active as the racemate so the difference in 1ogK should be less than 0.3 (log 2). Lehmann et aL3 have re-calculated Pfeiffer’s observations in this way and obtained a slope of 0.34 with a correlation coefficient, r, of 0.96. If 1ogK for the stronger enantiomer determines logs%, it must also determine 1ogK for the weaker enantiomer. Is this always true? It does not seem probable, if only because it takes no account of the flexibility of the molecule. In a flexible molecule it would be expected that there would be less difference between enantiomers because the weaker form might adapt its conformation to achieve a degree of fit that would be denied to a more rigid molecule. If flexibility is associated with a smaller difference between enantiomers, Pfeiffer’s rule would suggest that flexible molecules should have lower affinity, which does not seem likely to be true and is not what is found. For example, results obtained with several series of antagonists at muscarinic cholinoceptors4,5 are shown in Fig. 2. There is a trend; a least-squares fit to a straight line gives: log SSI = 0.380 (log K) - 1.906 with the r value of 0.52 (n = 54). The slope is similar to that observed with Pfeiffer’s data, indicating that SSI roughly doubles for a tenfold increase in affinity: there is, however, a lot of scatter and the correlation could not be used to predict results for new pairs of compounds with any reliability. The highest value of 1ogSSI (3.05, indicating a > lOOO-foid difference between enantiomers) is for an aza-analogue of hyostyamine (IogK 8.36). The calculated value of iog SSZ is 1.19 so the experimental value of SSZ is nearly 100 times the calculated value. Although the experimental value falls within the 95% confidence limits, these range from -0.82 to

3.20 (i.e. > lOOOO-fold). Better correlations were

obtained by Lehmann et ~1.3 who analysed the same data as separate

TiPS - April 1990 [Vol. 1 ZJ

149 for some compounds it is greater at higher temperature.

_

2.29

I

c

0.5 1.0

5 10 50 100 1000 5000 Average Human Dose

Fig. 7. Pfeiffer’s original observations. The weaker compounds, for which the dose is bigger, lie to the right (Redrawn using data from Ref. 1.)

_I sets of homologues. The results for a particular type of compound often form a pattern (as in Fig. 2) and Lehmann et al. interpreted these and other results in terms of equilibria involving binding at four parts of the molecule to complementary sites in a rigid receptor. The slopes of the lines differ from one set to another, because the binding groups differ in relative importance from one set to another. The main problem with this approach is in defining a set. The values for hyoscyamine, for instance, are 1ogSSi = 2.5 and 1ogK = 9.38 for the (-)-isomer: should its aza-analogue belong to the same set? If it does, the compounds clearly do not obey Pfeiffer’s rule. There are thermodynamic reasons why there should be differences between sets and even between compounds. The Gibbs free energy, AG, depends on the enthalpy of binding, AH, and on the entropy of binding, AS (AG = AH - T AS; T is the temperature); there is no reason why AH and AS should be related, though they might be similar in sets of compounds that bind to similar areas of the receptor. Differences in flexibility are likely to be associated with differences in AS. Differences in AH may be detected experimentally because the enthalpy determines

the effects of temperature on binding [dlogKId(lIT) = -AH/R] and it is known that these are variable-. Usually affinity is greater at lower temperature but

With the recent developments in the elucidation of the structures of many types of receptor and the possibility of doing molecular modelling in some instances, it is important to re-assess the validity of Pfeiffer’s rule. Because it looks reasonable at first sight there is a danger that it becomes part of established teaching without its implications being explored fully. The corollary to Pfeiffer’s rule that the affinity of the weaker enantiomer is determined by that of the stronger one - is improbable. There are many exceptions besides the examples given here. Moreover, as methods for measuring affinity become more reliable there is increasing information about the variable effects of temperature on binding; this information gives thermodynamic reasons for exceptions to the rule. The entropy of binding is likely to involve what happens to water in the binding process as well as what happens to the conformation

z2cn

5

9

7

11

Log K

Fig. 2. Affinities of some antagonists for muscarinic acetylcholine receptors in guineapig ileum. The weaker compounds, with lower affinity, lie to Ihe left. Values of log SSI are plotted against log K for the stronger enantiomer. Mandelic esters (0); phenylcyclohexyl acety/ esters (0); phenylcyclohexyl g/yco//ic esters (81; u-methylfropic esters (a); quaternary derivatives of hyoscyamine (m); qualernary derivatives of h.“osclne (0); quaternaty derivatives of homalropine (A). The line is the least-squares fit for aIf54 compounds and has the equation Y = 0.380 (X - 1.906) (X-intercept = 5.02). Results for the aza-analogue of hyoscyamine and its methiodide (0) were nol included in fhe fit (Data from Refs 4 and 5.)

TiPS - April 1990 [Vol. 221

150 of the drug. The effects on water important will be particular@ where hydrophobic binding is invalved. In assessi:lg hydrophobic effects it is more appropriate to think in terms of overlapping areas, rather than interactions between points. There is a need for a revised model of the interactions between drugs and receptors

which takes account of the reasons why there are exceptions to Pfeiffer’s rule. References I Pfeiffer, C.

C. (1956) Scir~r 124. 29-30 2 Barlow, R. 8. (1971) /. Plrnn~~.Plu~~~ol. 23,9C-97 3 Lehmann, P. A., de Miranda, J. F. R. and Ari&s, E. J. (1976) Prog. Dr~rg Rrs. 18, 101-142

firnary sequence of cyclic nucleotide phos hodiesterase isozymes and the design of selective inhibitors Joseph A. Beavo and David H. Reifsnyder Primary seqzle~tceiuforrnntion hns been reported for more than 25 drfferent nmmunlinn cyclic uucleotide phosplzodiesternses. Moreover, recent observntious suggest that mnn!y of these isozyznes me selectively expressed in LI limited number of cell types. The fuct that nearly all these different phosphodiestercses hnzle unique prinmy sequences in their cntnlytic or regulntory dotnnim and thnt they me often selectively expressed itnplies that it tnny be possible to modulate individunl isozynzes using specific drugs. Joe Beavo and David Reifsnyder summrize much of the evidence that hns led to oldr current understanding of multipIe isozyrnes of phosphodiesterczse, with emphasis on aspects thnt my be relezmt to drug design. They also discuss 7ulry ninny previous attempts to isolate isozyme-selective inhibitors may have failed. Although many texts imply that cyclic nucleotides are degraded by a single enzymatic activity, it is now apparent that inactivation of CAMP and cGMP is catalysed by not one, but rather a large number of different cyclic nucleotide phosphodiesterases. Recent data suggest that at least five different isozyme families exist and more than 20 distinct enzymes are now recognized (see Box). More importantly, biological reasons for this great diversity are beginning to be appreciated. For example, many of the isozymes are differentially expressed and regulated in difI. A. Benoo

is Professor nt

tlte Dcparhwn~ 51-3U. Utllvcrsity of WA 98195. US& atld D. H. Rrifsnydrr is n Scientist at Geueutech, Sm Fraucisco, CA 94080, USA.

of Pharmncolo~y. Wasltiqton, Stvtttc.

ferent cell types. From a pharmacological perspective, just as multiple receptors controlling the synthesis of CAMP and cGMP offer opportunity for selective therapeutic intervention., multiple ‘receptors’ for cyclic nucleotide degradation should offer equally good possibilities. Here we discuss the way in which recent data on the structure, regulation and localization of multiple phosphodiesterases provide a conceptual rationale for the design of selective inhibitors and activators of these enzymes (see Refs 1 and 2 for more general reviews on phosphodiesterases).

Early phosphodiesterase inhibitors Since Butcher

the original reports by and Sutherland3 in 1962

4 Barlow, R. B., Franks, F. M. and Pearson, J. D. M. (1973) /. Med. U&w. 16, 439-446 5 Barlow; R. B. (1973) /. Med. Clte,,r. 16, 1037-1038 6 Barlow, R. 6.. Berry, K. J., Clenl,>n, I’. A. M., Niko!aou, N. M. and Soh, K. S. (1978) Br. 1. Pltarr~mcol. 58, 613-620 7 Barlow. R. B. and Burston, K. N. (1979) Br. /. Plu~n~mcol.66, 581-585 8 Barlow, R. B., Birdsall, N. j. M. and Hulme, E. C. (1979) Br. 1, PJxnr~t~ncol.66, 587-590

that methylxanthines such as caffeine and theophylline inhibit CAMP hydrolysis, many studies have been carried out to identify drugs that inhibit cyclic nucleotide phosphodiesterase activity. In the first 20 years of this period no new therapeutically important agents were identified. Although drugs like papaverine and theophylline were found to be relatively potent and effective competitive inhibitors of CAMP and later cGMP hydrolysis, most new agents based on these structures were either ineffective or tou toxic tu be of widespread clinical use. We now know that the relatively poor therapeutic index of these agents was due at least in part to the fact that they were not isozyme specific and therefore increased cyclic nucleotide levels in many non-target cells. Moreover most had additional modes of action. For example, theophylline and its more potent congener 3-isobutyl-l-methyl xanthine (IBMX) are also potent antagonists at adenosine receptors. Dipyridamole, a classical antithrombotic agent that is now known to be a selective phosphodiesterase inhibitor, is also a potent inhibitor of adenosine transport. It is now evident that the design of most early screening experiments was inappropriate. In nearly all of them, whole tissue homogenates or extracts were used as the source of phosphodiesterase activity. Because most tissues are heterogeneous with respect to cell type and many cells contain multiple phosphodiesterases, the screening studies were all conducted with a mixture of isozymes and were therefore unlikely to identify inhibitors selective for an individual isozyme. For a few agents it was noted that l&20% of the total phosphodiesterase activity was inhibited by relatively low drug