Chem. Res. Toxicol. 1994, 7, 307-312
307
Structure-Activity Relationships for Contact Allergenic Potential of y,y-Dimethyl-y-butyrolactoneDerivatives. 2. Quantitative Structure-Skin Sensitization Relationships for a!-Substituted-a!-methyl-y,y-dimethyl-y-butyrolactones Christelle Franot,? David W. Roberts,$ David A. Basketter,Q Claude Benezra,tJ and Jean-Pierre Lepoittevin'vt Laboratoire de Dermatochimie associk au CNRS, Universit6 Louis Pasteur, Clinique Dermatologique, CHU, F-67091 Strasbourg, France, Port Sunlight Laboratory, Unilever Research, Quarry Road East Bebington, Wirral, Merseyside L63 3JW, United Kingdom, and Environmental Safety Laboratory, Unilever Research, Colworth House, Sharnbrook, Bedford MK44 lLQ, United Kingdom Received June 14, 199P
A skin sensitization cross-challenge dataset for a series of a-(X-substituted-methyl)-y,ydimethyl-y-butyrolactones is analyzed in terms of the relative alkylation index (RAI) model. The data analyzed consist of guinea pig sensitization response data for tests in which one lactone derivative is used for the induction stage and then the animals are challenged with another lactone derivative to elicit the response. RAI values are based on calculated log P (octanol/ water) values together with measured relative rate constants for reactions of the lactones with n-butylamine to form a-methylene-y ,y-dimethyl-y-butyrolactone. Plots of biological response against RAI for induction (RAIi) for sets of data in which the same compound is used for challenge have the double-sigmoid shape typical of sensitization responseRAIi plots, with some points in the overload region. Relative elicitation potential (REP) values, defined in terms of the biological response when sensitized animals are challenged with the compound in question relative to the response when the same animals are challenged with a chosen reference compound, are obtained. Consistent with the RAI model, plots of REP against RAI,, the RAI value corresponding to challenge, are linear and the slopes of plots corresponding to different reference compounds are, within the limits of experimental error, in the same ratio as the REP values of the reference compounds. Finally, multiple linear regression analysis gives a quantitative structure-activity relationship (QSAR) covering the complete set of cross-challengedata, relating the biological responses to the RAIi and RAI, values. The coefficients and intercept in this QSAR are similar to those of a QSAR reported previously for sensitization by sulfonate esters. The findings validate the applicability of the RAI model to datasets in which both reactivity and lipophilicity are varied and support the argument that for the lactones studied the elimination reaction with n-butylamine is a good model for the rate-determining step of in vivo carrier protein haptenation. The similarity between the QSARs for the lactones and for the sulfonate esters suggeststhat although the two series are very different chemically, the intrinsic antigenicities of the groups transferred to carrier protein are similar. Introduction In the preceding paper (1)we described the synthesis and chemical reactivity studies of three series of (Y-(osubstituted-alkyl)-y,ydimethyl-y-butyrolactones,in which the w-substituted alkyl group is CHzX (series l), CH2CHzX (series 21, or CH2CH2CH2X (series 31, and discussed the variation in sensitization potential and cross-reactivity between the different series in terms of the observed reaction chemistry. *Address correspondence to this author at the Laboratoire de Dermatochimie, Clinique Dermatologique, CHU, F-67091 Straebourg, France. + Universite Louie Pasteur. t Port Sunlight Laboratory, Unilever Research. Environmental Safety Laboratory, Unilever Research. 11 Accidentally deceased, January 20,1992. 0 Abstract published in Advance ACS Abstracts, March 15,1994. 1 Abbreviations: ACD, allergic contact dermatitis; FCA, Freund's complete adjuvant; MSIAT, modified single-injection adjuvant test; QSAR, quantitativestructure-activity relationships;RAI, relativealkylation index; RAL, relative alkylation index at challenge: RAI,, relative alkylation index at induction; REP, relative elicitation potential.
Here we report quantitative structure-activity relationship (QSAR)1 studies, based on the use of RAI (relative alkylation index) values calculated from the kinetic data reported in the preceding paper, on the cross-reactivity data obtained for the a-methyl-substituted a-methyl-y,ydimethyl-ybutyrolactones, series 1. Structure-sensitizationrelationships have been studied for other series of compounds (21,although in none of these cases has the range of relative reactivity and hydrophobicity been as varied as in the present study. In these earlier studies it has always been found that sensitization potential varies with the relative alkylation index (RAI), which in its most general form (3)can be expressed as RAI = log D
+ a log I t , + b log P
The RAI is a quantifier of the relative degree of carrier protein haptenation (Le., in the present case the number of lactone groups which become covalentlybound to carrier
0893-228~/94/2707-0307$04.50/00 1994 American Chemical Society
308 Chem. Res. Toxicol., Vol. 7, No. 3, 1994
protein in the skin) by a dose D of the compound under consideration. Besides being dependent on the dose given, the degree of carrier haptenation also depends on the compound's chemical reactivity (modeled by the log krel term in the RAI) toward carrier protein nucleophiles and on its partitioning properties, modeled by the log P term ( P being the octanol/water partition coefficient). Reference 2 gives a detailed account of how the RAI was originally derived and subsequently refined by a mathematical modeling approach. The values of the coefficients a and b should be constant for a given series of compounds all sensitizing by the same molecular mechanism and all tested by the same protocol [except that where the compounds range from hydrophilic (log P negative) to hydrophobic (log P positive), the b coefficient is not constant]. The aims of the present study are to test the RAI model on a dataset with a more varied multiplicity of lipophilicity and reactivity values than hitherto, and to test the hypothesis that the elimination reaction of compounds of series 1 with n-butylamine is a suitable model for the ratedetermining step in the carrier protein haptenation reaction leading to sensitization. Materials and Methods Caution: Skin contact with all lactone derivatives must be avoided. As sensitizing substances, these compounds must be handled with care. Chemicals. The synthesis and physical characteristics of chemicals used in this study are reported in the preceding paper (1).
Biological Testing. The guinea pig sensitization testa were carried out using the modified single-injection adjuvant test (MSIAT) (4). Preliminary irritation testa were used to determine the concentration range suitable for induction and to ensure that the challenge w a ~conducted a t an optimal nonirritant concentration. In general, all the chemicals were tested a t equimolar concentrations of 0.31 M; exceptions are indicated in Table 3. In brief, the MSIAT protocol was as follows. Sensitization was induced in groups of 10 albino Dunkin-Hartley guinea pigs by a series of six 0.1-mL intradermal injections of test substance in 0.9% w/v sodium chloride vehicle, in combination with Freund's complete adjuvant (FCA), in the shoulder region. This part of the procedure was identical to that used for the guinea pig maximization test (5). After a 2-week interval, the initial challengewas normally conducted by giving a 6-h occluded topical patchapplication of the test chemicalin Finn chambers to clipped and shaved flank skin. The Challenge vehicle was 70/30 v/v acetone/poly(ethylene glycol) 400, and the challenge concentration was in all cases 0.31 M. Skin reactions were scored for erythema (scale 0-3) and the presence of edema approximately 24 and 48 h after removal of the patch. Subsequent crosschallenges were made at weekly intervals on alternate flanks using the same patch procedure and scoring system. For each animal, erythema was scored as 0,0.5,1,2, or 3 on each occasion. For each challenge the erythema scores for all animals a t 24 and 48 h were totaled and the totals expressed as a percentage of the maximum possible score (this being 6 times the number of animals involved in the challenge), this figure being used as the quantifier of the biological response. Results a n d Discussion Sensitization Dataset. As described in the preceding paper (I), a series of compounds of structure la-i was synthesized for study of the relationship between chemical structure and skin sensitization. It was found that all of these compounds react with butylamine, used as a simple
Franot et al. Scheme 1. Reaction Pathway of Compounds la-i with n-Butylaminea
1a-i
2
a a: X = C1. b: X = Br. c: X = OS02Me. d: X = OSOzArMe. e: X = OSOzArOMe. fi X = OS02Ar. g: X = OSO&rC. h: X = OS02ArNOz. i: X = SCN.
Table 1. Cross-Challengeand Self-challenge Matrix substituent compd XO
no. la lb IC Id le lf
76 biolo ical cross-cdenge reaction* 10~kc~OK la P lb
IC Id le If 18 lh li
c1 0 1.717 66 15 Br 70 23 40 24 0.95 1.857 OS02CHs 0.176 0.781 38 5 21 24 OSOtArCHs 0.097 2.519 48 22 OSOp4rOCHs 0.024 2.264 40 12 30 15 OSOzAr 0.117 2.020 59 20 10 38 23 OSOzArCl lg 0.328 2.733 94 78 84 OSOZArNOZ l h 1.699 1.763 62 24 38 37 31 42 15 8 49 SCN li 0.179 1.280 a All inductions carried out at 0.31 M, except for le and lh for which induction was at 0.16 M. *See Materials and Methods. c Relative rate constant with n-butylamine. Compound la was used as reference. d Partition coefficient (octanol/water) calculated according to the method of Leo and Hansch (4).
chemical model for biological nuleophiles in skin, by elimination of HX to produce the a-methylene-y-lactone 2 (Scheme 11,which can then react with protein by Michael addition. Table 1summarizes the results from sensitization testing and cross-challenge and also gives log P values calculated by the Leo and Hansch method (6)and log krel values. The log krelvalues were obtained from competition experiments in which pairs of compounds were allowed to react with a deficiency of n-butylamine, and the composition of the mixture before and after reaction was determined by NMR. Sensitization Potential. The compounds of the present study are all hydrophobic. For testing of other series of compounds by the MSIAT method a value of 0.48 has been used for b in the RAI expression and found to lead to good correlations between RAI and sensitization (3). Therefore, 0.48 is taken as the value for b here. The appropriate value of the log krel coefficient a has to be found-no previous studies have been carried out on compounds which sensitize by an elimination-Michael addition reaction sequence. In studies on series of compounds reacting by the s N 2 mechanism a value of 1 has been found appropriate (3) although in view of experimental scatter and the small number of different reactivity values in these studies, it is possible that other values for a would have served equally well. For the self-challenge results, Le., the biological responses obtained when the same compound is tested both at induction and at challenge, no obvious pattern of variation of response with structure was found-we attribute this partly to the fact that not only the RAI for induction (RAIi) but also the RAI for challenge (RAI,) varies as the structure of the test compound changes. We next examined more closely the cross-challenge results. For a dataset in which induction is carried out with a series of compounds and the challenges are done with a single compound (provided of course that all compounds react in the same way and transfer an identical
Chem. Res. Toxicol., Vol. 7, No. 3, 1994 309
QSAR for Lactone Derivatives
Table 3. Cross-Challenge with 2
Table 2. Cross-Challenge with l b substituent Xa OS02CeH4C1 Br OS02CsH4N02 OS02C&
SCN OSO2C&OMe OSO2CH3
compd no. la lb lh If li le IC
% biological responseb 94 70 62 59 42 40 38
RAIf 1.97 2.79 3.94 1.21 0.97 0.79 0.73
a All inductions carried out at 0.31 M, except for l e and l h for which induction was at 0.155 M. Therefore, log D is taken as 0 for all compounds except l e and l h for which log D is taken as -0.30. b See Materials and Methods. RAIi calculated as log D + 2 log kre1 + 0.48 log P.
‘0°
1
substituent Xa
compd no.
% biological responseb
RAP
1g Id lb la lh
89 59 48 32 30 13 13 13 0
1.97
OSO&H&l OSO&&Me Br
c1
OSO&H4NO2 OSOzCH3
IC
SCN
li If le
OSO&& OSO&&OMe
1.40 2.79 0.82 3.94 0.73 0.97 1.21 0.79
a All inductions carried out at 0.31 M, except for le and l h for which induction was at 0.156 M. Therefore, log D is taken as 0 for all compounds except le and l h for which log D is taken as -0.30. b See Materials and Methods. c RAIi calculated as log D + 2 log kd + 0.48 log P.
A
Table 4. Cross-Challenge with IC, -d, and -i
.
A
substituent
Xa
% biologicalb response to challenge 4 t h li le IC
compd no.
RAIf
.-
404
&A A
1
30 0
1
2
3
RAli
Figure 1. Biological response against RAIi for compound lb.
chemical group to the biological nucleophile) the analysis should be simpler since there is variation only in the biological insult at induction, not at challenge. Such an analysis should, therefore, provide a quantitative interpretation of how potential to sensitize, as opposed to potential to elicit a response, varies with structure. For most of the compounds tested at the induction stage, cross-challengeshad been carried out with the compound l b (X = Br). This subset of the full data matrix is shown in Table 2. It was found that, with RAIi calculated as log D + 2 log krel 0.48 log P, the data was well fitted by the RAI model. The fit was less good with a value of 1 for the log krel coefficient. Figure 1 shows a plot of biological response against RAIi. The plot has the typical shape of sensitization response-RAI plots (2,3,7).Two of the points are in the overload region where the degree of sensitization decreases with increasing carrier haptenation. The overload effect has been recognized for some considerable time (8)although there does not appear to be general agreement as to the underlying biological mechanism. In the development of the original RAI model the overload effect was rationalized in terms of induction of suppressor T-cellmediated tolerance (7). For all of the sensitization tests, cross-challenges had been carried out with 2, the HX elimination product. This subset of the data matrix is shown in Table 3. Again, a plot of biologicalresponse against RAIi (Figure 2), with RAIi calculated as previously, has the shape expected from the RAI model. Further RAI plots can be made for challenges with li (X = SCN), Id (X = OSO&eH*Me), and IC (X = OS02Me) (these subsets of the data matrix are shown in Table 4). These plots each consist of a cluster of points in the low RAIi region and a few in the overload region, with no points in the intermediate region where higher biological responses are to be expected. Hence, these plots on their
+
100
A
t
80-
0
n
p
60-
A
K
-3 .-ul .-- m
A
40-
0 0
A
A
20
0
. A .
- ,
1
310 Chem. Res. Toxicol., Vol. 7, No. 3, 1994
IC
1 0.63
0.55 0.34
Franot et al.
Table 5. Relative Elicitation Potentials (REP) Referred to lb as Standard animals sensitized toa challenge li If le lh compd (x) RALb mean REPC 1 1 1 1 Br lb 2.79 1 (by definition) 0.64 0.38 0.50 SCN li 0.97 0.47 0.36 0.39 0.30 OSOzCH3 IC 0.73 0.33 0.19 0.34 OS02C&CH3 Id 1.40 0.36 0.39 OSOzC& If 1.21 0.39 0.64 0.61 OSOzC&Cl 1g 1.97 0.63 0.31 0.22 0.48 2 0.34
av of ~
~~
4 2 3 1 2 4
All induction carried out at 0.31 M except for le and l h for which induction was at 0.16 M. RAI, calculated as log D + 2 log k,l+ 0.48 log P.c Relative elicitation potential, compound lb being used as reference. 0
Table 6. Relative Elicitation Potentials (REP) Referred to 2 as Standard la
lb 0.48 0.50
animals sensitized toa IC li Id 1.15 1.85 2.92 3.23 0.81 1.62 0.62
If
lh
challenge compd (X) OSOzCH3 SCN Br c1 OS02C&CH3 OSOzC& OS0~C6H~C1
RAI, 0.73 0.97 2.79 0.82 1.40 1.21 1.97
mean REPc 0.82 1.29 3.19 0.81 1.02 0.80 2.10 1 (by definition)
IC li lb la 0.47 0.83 1.54 Id If 0.80 2.92 1.27 lg 1 1 1 1 1 1 1 2 a All induction carried out at 0.31 M except for le and l h for which induction was at 0.16 M. RAI, calculated aa log D + 2 log log P. c Relative elicitation potential, compound 2 being used as reference. 1.76 4.53
1.03 2.07
response to 1,X = XI, divided by the response to 1, X = Xz. If the data are such that this can be done for more than one set of animals, then an average can be taken. In performing this exercise, self-challenge data (i.e., when the same compound is used both for induction and challenge) were excluded from the calculations. It has previously been observed that even with a set of sensitizers which all transfer the same chemical grouping to protein, self-challengescan have an "unfair advantage" over crosschallenges since the epitopes produced at induction and challenge should be identical. A published example is provided by a study in which a full sensitization and crosschallenge matrix was produced for three methyl-transfer agents of the general formula RSOsCH3, where the R groups were docecyl, hexadecyl, and hexadec-2-enyl. In each case, self-challengewas found to give a higher response than did cross-challenges (9). Although the "unfair advantage" effect does not appear to have been discussed elsewhere in the literature, it is not uncommon in our experience. We presume that it arises as a result of the polyclonal nature of the sensitization response and a closer match of epitope distribution and density patterns when the same compound with the same reactivity and partitioning properties is used both at induction and at challenge. Table 5, based on a subset of the data in Table 1,shows how averaged REP values relative to lb (X = Br) are obtained. Table 6 shows how the same approach can be applied to obtain REP values relative to 2 as the reference standard. For these calculations it is not appropriate to use data from challenges on sets of animals which are so highly sensitized that they give very high responses on all challenges-since responses cannot be greater than 100% , there is not the scope for differences in averaged REP to be expressed. For this reason, data for challenges onto animals sensitized to lg (X = OS02CsH4Cl) have been excluded from the REP analysis. It is clear from Tables 5 and 6 that compound 2 has a lower REP than several of the compounds 1, even though it is the "ultimate hapten" into which the compounds 1
:
41 1 Le W n
1; 1:
av of 2 4 4 1 5 1 2
+ 0.48 +/
2
1
0
+
/ Slope 0.31
2
1
1 3
RAlc
Figure 3. Relative elicitation potential (REP)against RAI, for l b ( 0 )and 2 (+I.
are presumed to be converted before haptenation of the biological nucleophiles. A possible explanation is that whereas compound 2 can only sensitize by the Michael addition pathway, the compounds 1 may in part sensitize via an s N 2 reaction pathway, as discussed in the preceding paper (1). If this is so, then compound 2 will not produce antigen which can be recognized by those clones of T-cells, expanded in the process of sensitization by a series 1 compound, which specifically recognize antigen produced via the sN2 pathway from series 1. It is also possible that compound 2, being more reactive than compounds 1, is lost to reaction with stratum corneum protein in the course of skin penetration to agreater extent than are compounds 1. If all the compounds used in challenge penetrate skin equally well, or if the extent of penetration is correlated with RAI,, then there should be a positive correlation between REP and RAI,. Figure 3 shows plots of relative elicitation potential agianst RAI, for both lb (X = Br) and 2 as reference standards. Both plots are reasonably linear, and the ratio (0.27) of the slope for the l b (X= Br) standard to the slope for the 2 standard is not eignificantly different from the REP value (0.34)for lb based on the 2 standard. This indicates good consistency of the data and suppo& the validity of the approach.
QSAR for Lactone Derivatives
Chem. Res. Toxicol., Vol. 7, No. 3,1994 311
Regression Analysis. The theoretically derived relationship between biological response (expressed, for example, as percent of maximum theoretically possible response or as the corresponding probit value) and RAI values for a given series of compounds is of the form:
100
1
/
n
r) 0 Y
L
biological response = a(RAIi) - b(RAIi)2+ c(RAI,)
0
+d
P
(1)
where a, b, and c are positive coefficientsand d is a constant for the series studied. For the present dataset, regression analysis for all of the cross-challenge data shown in Table 1(self-challenge data excluded to avoid the “unfair advantage” effect)gives the equations: % response =
v) Q)
a
0
20
40
% Response
60
80
100
(calc)
Figure 4. Percent (%) response observed vs percent (%) response calculated.
73.O(RAIi) - 7.2(RAIJ2 + 17.5(RAIC)- 208 (2)
n = 24, R = 0.874, s = 11.4, F = 21.5 probit (% response) = 2.38(RAIi) - O.24(RAIS2+ O.52(RAIc) - 3.12 (3) I
n = 24, R = 0.856, s = 0.38, F = 18.3 In these equations, RAI values were calculated as log D 0.48 log P. However, the units of D are in ppm mol, so that the RAI values are all 2.49 [=log(0.31 X lO3)] greater than those used in the REP analysis. The reason for changing units at this point is to facilitate comparison with quantitative structure-activity relationships (QSAR) analyses for other series of compounds (see below). Since over the biological response range covered, the percent responses are highly colinear with the correspondingprobit values, these two equations are equivalent. The percent response equation gives a slightly better fit to the data than does the probit equation,but the difference is insignificant. Figure 4 shows a plot of percent response observed against percent response calculated from the regression equation. In the course of carrying out the regression analysis,the effects of changing a and b, the log lz,l and log Pcoefficients, respectively, in the RAI formula, were explored. The best correlations were found with values of 2.2 and 0.6 for a and b, respectively:
+ 2 log & +
% response = 79.1(RAIi) - 6.2(RAIi)’
+ 15.6(RAIc)- 290
(4)
Figure 5. F ratio against a and b parameters.
fit becomes progressively and significantly worse. It can also be seen that increasing b to 1.0 gives a significant decrease in the quality of the fit. It is interesting to compare the probit-based eq 3 with a QSAR/dose response relationship reported previously by two of us (3)for a series of six alkyl alkanesulfonates. The data in that study were all based on self-challenges, and most of the variation in RAI values came from the use of severaldifferent induction and challengeconcentrations for some of the compounds and to a lesser extent from variation in log P values. Only two different K,1 values were represented-a value of 1 for compounds of type MeS03R and a value of 30 for compounds of the type RS03Me. In contrast, the variation in RAI values in the present study comes mainly from each compound having a unique krel value and a unique log P value. Nevertheless, the coefficients of the two equations are very similar: sulfonates:
n = 24, R = 0.876, s = 11.3, F = 22
probit = 2.24(RAIi) - 0.26(RAIi)’ + O.54(RAIc) - 2.23
(5)
probit = 2.85(RAIi) - 0.24(RAIi)’ + O.52(RAIc) - 3.12
However, eqs 4 and 5 are not significantly different from eqs 2 and 3, and over the ranges 2-2.5 for a and 0.48-0.7 for b, the differences in quality of fit were insignificant. This is illustrated in Figure 5, which shows a block scatterplot of the F ratio for the percent response regressions against a and b. Similar plots were obtained using R and s as quantifiers of the goodness of fit. A key point to note from Figure 5 is that as a is reduced from the 2-2.5 range to 1.5, then to 1.0, and then to zero, the
The intercepts are also quite similar, the difference corresponding to the sulfonates giving on the percentage scale a ca, 30 units greater response than the lactones if compounds of both series were tested at the same RAIi value and the same RAI, value. This difference is of the same order of magnitude as the “unfair advantage” effect (see below), suggesting that the two QSARs would be almost identical if both related to self-challenge or both related to cross-challenge. Self-challenge Data. It would be expected that the biological responses from self-challenges,as shown in Table
probit (% response) = 2.17(RAI,>- 0.2O(RAIi)’
lactones:
+ 0.46(RAIC)- 2.88
n = 24, R = 0.858, s = 0.37, F = 18.6
312 Chem. Res. Toxicol., Vol. 7, No. 3, 1994
Franot et al.
Table 7. Comparison of Self-challenge Results with Predictions of Equation 4 compd % response % response difference substituent Xu no. obsdb calcd‘ (calcd - obsd) 12 66 54 la 70 4 lb 66 ~~
IC
Id le If 1g lh li
10 35 6 24 50 76 16
5 22 30 10 84
37 49
- 5 -13 24 -14 34 -39 33
a All induction carried out at 0.31 M, except for le and l h for which induction was at 0.16 M. b See Materials and Methods. Calculated as % response = 79.1(RAIi) - 6.2(RAId2+ 15.6(RAIC)
- 290.
1, should either be well predicted by the regression eqs 2-4 or be underestimated as a result of the “unfair advantage” effect. Table 7 shows a comparison of the observed biological responses with those predicted from eq 4. It can be seen that some of the self-challenge responses are quite well predicted and others, as expected, are quite markedly underestimated. A striking exception is the selfchallenge result for the compound lh (X= p-nitrobenzenesulfonate), which is markedly overestimated by the regression equation. Possibly this may be explained in terms of reduced penetration of this compound at challenge-because of its amphiphilic structure it might be expected to bind strongly in the lamellar-phase lipid barrier in the stratum corneum. An alternative possibility is that because of its high chemical reactivity a substantial proportion of the challenge dose is lost by reaction with stratum corneum proteins. Overall,the mean value for the magnitude of the “unfair advantage” effect, calculated by taking the average of the values in the fifth column of Table 7 is 10 units (15 if compound lh is excluded from the calculation) on the percentage scale.
Conclusions The dataset analyzed here was designed with the aim of providing the most stringent test so far of the RAI model. The findings provide good validation of the model by demonstrating its applicability to interpretation of sensitization datasets based on compounds with diverse reactivities and lipophilicities. The present dataset is particulary interesting in that the compounds sensitize by a prohapten mechanism. The chemical reactivity parameter in this instance relates not to the reaction in which the hapten couples to carrier protein, but to a preceding activation step in which the compounds are converted via an elimination reaction to a single ”ultimate hapten” y ,y-dimethyl-a-methylene-ybutyrolactone, which then undergoes Michael addition. The preceding paper described how in vitro kinetic studies revealed the elimination reaction in all cases to be faster than the subsequent Michael addition reaction (1). If this were also true of the in vivo coupling of these compounds to skin nucleophiles, then there should be no need to use a krel term in the RAI expression, since the rate-determining step would have the same rate constant in all cases. However, as shown above, interpretation of the sensitization dataset in terms of the RAI model requires the use of the relative rate constants for the elimination
reaction. This strongly supports the argument in the preceding paper that in the in vivo coupling of the compounds to skin protein the elimination reaction is rate determining because of the lower basicity of the in vivo medium and also because of the higher Michael reactivity of some skin nucleophiles as compared with butylamine. A further outcome of the present study, which was not anticipated at the commencement, is the finding that the QSAR equations for the cross-challenge dataset on the lactones and a self-challenge dataset on sulfonate esters are very similar. That the coefficients of the RAI terms in the equations are well matched is not particularly surprising and servesto further substantiate the robustness of the RAI model. What is more unexpected is that the difference between the intercepts is not major and may indeed be negligible if the “unfair advantage” effect of self-challengeas compared to cross-challengeis taken into account. If the relative rate constants used in calculation of the RAI values were expressed on the same scale for both datasets, then the difference in intercepts would represent the difference in intrinsic antigenicity, Le., the relative abilities of the different chemical groupings, when attached to carrier protein, to be percieved by the immune system as nonself. The relative rate constants used in the sulfonate ester study were not based on measured values and cannot be compared directly with those used in the lactones study. However, it seems unlikely that the relative reactivities of the reference compounds for the two series (laand primary higher alkyl methane sulfonate, MeSOaR) would differ very greatly. We are therefore led to the so far tentative but potentially far-reaching conclusion that the intrinsic antigenicities of groups as different as the lactone entity and simple alkyl groups may be very similar. Acknowledgment. We thank the Centre National de la Recherche Scientifique (CNRS, France) for financial support to C.F. References Franot, C., Roberts, D. W., Smith, R. G., Basketter, D. A., Benezra, C., and Lepoittevin, J.-P. Structure-activity relationships for contact allergenic potential of y,y-dimethyl-y-butyrolactone derivatives. 1. Synthesis and electrophilic reactivity studies of a-(a-substitutedalkyl)-7,~-dimethyl-y-butyrolactones and correlation of skin sensitization potential and cross-sensitization patterns with structure. Chem. Res. Toxicob. (preceding paper in this issue). Fraginals, R., Roberts, D. W., Lepoittevin, J.-P., and Benezra, C. (1991)Refinement of the relative alkylation index (RAI) model for skin sensitization and application to mouse and guinea-pigtest data for alkylsulfonates. Arch. Dermatol. Res. 283, 387-394. Roberta, D. W., and Basketter, D. A. (1990)A quantitative structure activity/dose response relationship for contact allergic potential of alkyl group transfer agents. Contact Dermatitis 23, 331-335. Goodwin,B.F. J., and Johnson, A. W. (1985)Singleinjection adjuvant test. Current Problems in Dermatology (Andersen, K. E., and Maibach, H. I., Eds.) Vol. 14,pp 201-207,Karger, Basel. Magnusson, B., and Kligman, A. M. (1970) Allergic contact dermatitis in the guinea pig. Identification of contact allergens, Charles C . Thomas, Springfield, IL. Leo, A., Hansch, C., and Elkins, D. (1971)Partition coefficients and their uses. Chem. Reu. 71,525-616. Roberta, D. W., and Williams, D. L. (1982) The derivation of quantitative correlation between skin sensitisation and physicochemical parameters for alkylating agents and their application to experimental data for sultone. J. Theor. Biol. 99,807-825. Sy, M. S.,Miller,S. D., andClaman,H. N. (1977)Immunesupression
with supraoptimal doses of antigen in contact sensitivity. I. Demonstration of suppressor cells and their sensitivity to cyclophosphamide. J . Immunol. 119,240. Roberts, D. W., Goodwin,B.F. J., andBasketter,D. A. (1988)Methyl groups as antigenic determinants in akin sensitisation. Contact Dermatitis 18,219-225.
