Analyst, January 1996, Vol. 121 (43-48) 43 Existence of Two Basic Sites in Triazolo-I ,4=diazepines: Determination of Two p& Values for a Model Compound in Water Beatrice Legouin and Jean-Louis Burgot UFR des Sciences Pharmaceutiques et Biologiques, Dkpartement d' Etudes Physicochimiques et BiocinCtiques des Pharmacosyst2mes, Laboratoire de Chimie Analytique, 2 av. du Pr Lkon-Bernard, 35043 Rennes-Cedex, France By a UV/VIS spectrophotometric study in the pH range -1.6 to 10.1 and by a polarographic study of a water soluble model compound, the occurrence of two basic sites in water has been ascertained for triazolo-1,4-diazepines. The pK, values found for this model were -0.24 and +1.81. Owing to the overlapping of the two pK, values, microforms exist simultaneously. Corresponding ionization microconstant values have been tentatively assigned. Keywords: Triazolo-I ,4-thienodiazepine; ultraviolet-visible spectrophotometry ; polarography; pK,; ionization microconstant Introduction 1,4-Benzodiazepines are compounds of important phar- maceutical interest.' It is now well known that their 4,Sazome- thine bond suffers hydrolysis in acidic media to give the corresponding aminobenzophenones, which reversibly cyclize into the original closed forms by elevating the pH of the solution.As a result, 1,4-benzodiazepines exist in such media as pH-dependent, equilibrated mixtures of ring opened and closed forms.2-12 Given the fact that these compounds generally possess multibasic sites, this hydrolytic process is accompanied by acid-base reactions. Hence, the knowledge of the different ionization constants is of utmost importance both for carrying out mechanistic studies concerning their hydrolytic process and for firmly grounding their pharmacological properties.Some of these diazepine derivatives have their 1,2-bond fused to a 1,2,4-triazolo ring. This is the case with alprazolam, estazolam and triazolam, which are triazolobenzodiazepines, and also with brotizolam, etizolam and of the We'973 compound of Gallo et al.,l which are triazolothienodiaze- pines. R I = H R,=H ESTAZOLAM R, = CH3 R2 = Br BROTIZOLAM RIeCHj Rt=Cl TRIAZOLAM R, = C6H1, R2 = Bt We'973 R, = CH3 R2= H ALPRAZOLAM R, = CH, Rt = CzHs ETEOLAM The pK, values in water of these triazolodiazepines (2.76, 2.76,2.84 and 5.9,2.26 and 1.52 for brotizolam,8,11 etizolam,*O estazolam10.12 and triazolam,6.9 respectively) are systematically attributed to the conjugated acid form of the imine group.It is important to recall that these determinations were performed with low percentages of methanol in water. Surprisingly, no PKa data concerning the basic character of the triazolo ring have been given in the aforementioned works, although the conju- gated acids of simple derivatives of this heterocycle exhibit pK, values falling in the 2.2-3.4 range.13 As a result, owing to the overlapping of these two ranges of pK,, the attribution of the above pKa values to the azomethine group of triazolodiazepines that has been carried out so far is questionable. In this paper we demonstrate that triazolodiazepines actually possess two pK, values.We used for our demonstration a UV/ VIS spectrophotometric study of the following compound: 4,7,8,10-tetrahydro- 1 -methyl-6-(2-chlorophenyl)-[4',3'- 4,5]pyrid0-[3,2--thieno- 1,2,4[4,3-a]-triazolo- 1,4-diazepine (NHPTT)14: NHm 3.5-DLM~HYL4-PHENYL-l.2.4-TRIAZOLE The pK, value of the protonated imine group of this model compound was tentatively assigned by a polarographic study and that of the triazolo ring by an independent determination of the pKa value of 3,5-dimethyl-4-phenyl- 1,2,4-triazole. The aqueous solubility of NHPTT, conferred by the protonation of the piperidino moiety in the pH range of our study, allowed us to work without using any additional organic solvent. Experimental Apparatus Measurements of pH were performed by use of a Tacussel LPH430T pH meter that was calibrated daily with six NBS buffers (commercial buffers manufactured according to the National Institute for Standards and Technology recom- mendations) and using an Ingold 9811 (pH 0-14) glass electrode.All UV/VIS spectra were recorded by using a Uvikon 930 spectrophotometer with 1 cm silica cells. Dc and differential-pulse polarograms were recorded on a Tacussel EPL3. Potentials were measured versus a saturated44 Analyst, January 1996, Vol. 121 calomel electrode (Tacussel XR 100). The following conditions were used: scan rate, 5 mV s-l; drop time, 2 s; pulse duration, 60 ms; pulse amplitude, 60 mV. Reagents The water used throughout this work was de-ionized by a set of ion exchanging columns (Bioblock Scientific, Illkirch, France) to p > 2 MQ cm-1.NHPTT was kindly purchased for us by Beaufour-Ipsen Industry.14 3,5-Dimethyl-4-phenyl- 1,2,4-triazole was synthe- sised according to Reiter's method.15 Physical properties were in full agreement with those given in the literature. The buffers for solutions of pH >2 were the Britton- Robinson buffers.16J7 Table 1 Determination of the pKa of the piperidino moiety : results obtained by treating spectrophotometric data for pH values from 6.51 to 10.1 (spectra recorded immediately after the preparation of solutions) C/moll-' A = 2 4 0 m A = 275 nm 10-4 ~Ka3 EB" EBH+ 1.5 x 10-4 p~~~ EB EB (experimental) EBH+ (experimental) * E in 1 mol-1 cm-1. 7.96 r7.86. 8.111 14 000 [13 900; 14 1001 16 200 [16 000; 16 4001 8.14 [8.10; 8.171 14 550 [13 500 13 6001 15 650 [ 15 600; 15 7001 13 850 16 500 8.25 [8.18; 8.331 6750 [6700; 68001 5400 [5350; 55001 8.09 [8.07; 8.121 6550 [6500; 66001 5380 [5360; 54001 6200 5100 BH 3'.For solutions of pH < 2 Bascombe and Bell's acidity functions18J9 (aqueous sulfuric acid solutions) were used. Methods All measurements were performed at 25 "C. Care was taken to avoid the opening of the benzodiazepine ring (by hydrolysis) in acidic media. Preliminary polarographic studies had indicated that, in acidic media (--1 < pH < 3), no 'opened' product could be detected for time durations below 1 min from the beginning of the preparation of solutions ('closed' and 'opened' products exhibited two very well differentiated waves). For UV/VIS studies we selected two working wavelengths, which differed from the pH ranges of the solutions studied: for -1.6 < pH < 3.1, h = 250 nm and h = 300 nm, and for 6.5 < pH < 10.1, h = 240 nm and h = 275 nm.Two wavelengths were selected for the sake of comparison of the K, values which, of course, had to be the same. For the same reason, we selected two analytical concentrations: 1 X mol I-' and 1.5 X 10-4 moll-'. In the two ranges -1.6 < pH < 3.1 and 6.5 < pH < 10.1, 19 and 16 working pH were used, respectively. Moreover, for each pHj value, 6 measurements on theoretically identical (but prepared in a totally independent way) solutions were replicated. Solutions were prepared by mixing aliquots of 2 ml of a stock solution of NHPTT (1 X 10-3 or 1.5 X 10-3 moll-1) with 18 ml of the appropriate buffer.Spectra were recorded in the 190-300 nm range at fast scan (scan speed, 2000 mm min-1) against the corresponding blank. The whole process (mixing of solution and recording) did not exceed 1 min. The fact that with our data treatment we obtained by calculation (see below) the same molar absorptivities of species BH+ for the most acidic media as those found experimentally at higher pH values was a strong argument in favour of the existence of only the closed form in the experimental conditions. In our calculations, unknowns were K, values and, when the species could not exist alone in solution, E values. Moreover, in some cases, we deliberately considered E values as supplementary unknowns. This provided a way of checking the accuracy of our data and calculations by comparison of the experimental and H' "p Ka, &" BH' OC' B &"' B*H,*+ Fig.1 Investigated ionization pathways of NHPTT. * The site of protonation on this nitrogen atom of the triazole was arbitrarily chosen.Analyst, January 1996, Vol. 121 45 I I -0.4 calculated E values. Treatment of experimental data was performed by two non-linear squares procedures. In the first one, data which were simultaneously treated by the regression procedure were chosen in such a way that for each pHi, one absorbance AFp was kept at random among the six replicates. All the selected data allowed the determination of the search for parameters by minimization of the cost function: I I 3.00 I 2.00 1 .oo 0.00 200 250 300 350 Wavelengthhm Fig.2 Absorption spectra of NHPTT in solutions of different pH: - 1.6; -1; -0.5; 0; +0.5; +l; +1.5; +2; +2.5; +3. (Scan speed 2000 nm min-I; path length 1 cm; C,,, = mol I-'). Table 2 Determination of the pKa of the imino and triazolo groups: results obtained for the first ionization scheme at different concentrations and wavelengths for pH between - 1.6 and +3.1 (same experimental conditions as in Table 1) EBH+* 1.5 x 10-4 PK,, EBH+ EBH+ (experimental) (experimental) EBH33+ * E in 1 mol-1 cm-1. h = 250nm -0.36 [-0.45; -0.261 1.70 [1.63; 1.791 16 050 [15 900 16 2001 12 400 [ 12 250; 12 5001 9 600 [9450; 97001 -0.28 [-0.39; -0.151 1.69 [1.59; 1.801 15 550 [14 480; 15 6501 12 450 [12 250; 12 6001 9 750 [9650; 99001 15 850 9 550 A = 300nm -0.10 [-0.18; 01 1.92 [1.89; 1.941 2 650 [2600; 27001 9 350 [9200; 95001 13 OOO [12 850; 13 2001 -0.24 [-0.32; -0.131 1.82 [1.79; 1.851 2 700 [2650; 27501 9 600 [9400; 98001 13 250 [13 100; 13 4001 2 500 13 300 where A?'" stood for the absorbance calculated according to the model and wi was a weighting factor defined classically by wi = 1/02, a? being the variance of the whole of the replicates of absorbance measurements for a given pHi value.Strictly, the same process was repeated seven times. Variances for the parameters were calculated from the preceeding results in the usual manner. In the second procedure, pooling of the 19 X 6 and 16 X 6 data was performed. The objective function was calculated according to Weighting factors were the same as above. The variances of the search for parameters were calculated through the variance- covariance matrix which takes place naturally during the process of minimization of the U function.It is relevant to note that the variances found with the two procedures (see results) were the same. The algorithm of minimization of the U function that we used was a direct one, according to the Hooke and Jeeves20 principle. For polarographic studies, systematic deoxygenation of solutions was accomplished by the bubbling of purified nitrogen through the buffer and through the stock solutions separately and by passing nitrogen over the solution in the cell throughout the experiments. The concentration of solutions in the cell was maintained at 1 X 10-3 moll-'. The lack of electroactivity of 3,5-dimethyl-4-phenyl- 1,2,4-triazole in our experimental con- ditions had been previously ascertained by an independent polarographic study.The triazolo ring of estazolam has also been found to be inactive on the mercury drop12 by other authors. Determination of the pK, of 3,5-dimethyl-4-phenyl- 1,2,4-triazole was performed potentiometrically by titration of 5 X mol 1-1 solutions with 2.5 X 10-2 moll-' hydrochloric acid. Results and Discussion Since the goal of this work was either to confirm or to invalidate the existence of a basic site on the triazolo ring, two ionization v) z? E I -0.6 1 PH Fig. 3 EIl2 versus pH. Experimental conditions: EO = -0.3 V; 'c - 2 s; t = 25 "C; scan rate, 5 mV s-1; pulse duration, 60 ms; pulse amplitude, 60 mV. Table 3 Half-wave potential values at pH between - 1 and +1 PH -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 E l f l versus SCE -0.46 -0.46 -0.47 -0.48 -0.49 -0.50 -0.52 -0.53 -0.55 -0.56 -0.5746 Analyst, January 1996, Vol.121 sequences were investigated: on the one hand, the three Acid-Base Pairs BHz2+IBH+ and BH$+lBH22+ (or equilibria sequences B‘Hz2+IBH+) K” 1 K”2 K“3 BH33+ BH22+ + H+ BH22+ BH++H+ BH+ B + H+ which took into account the protonation of the triazolo ring; on the other, the two equilibria sequences K’a , B’H22+ BH+ + H+ Ultraviolet-visible spectra (of NHPTT) recorded immediately after preparation of solutions (see Experimental) exhibited a continuous evolution with the pH in the range -2-4 (Fig. 2). The absence of an isosbestic point in the pH range was, at first sight, already an argument in favour of the occurrence for more than one equilibrium.Nevertheless, we performed a non-linear squares treatment of the experimental UV-visible data accord- ing to the two ionization schemes. Hence, the two mathematical models allowing the calculation of absorbances A:a1c were: Acalc - C&a 1 Ka2 - &BH+ Ka1 Ka2 + Ka, (H+) + (H+I2 K“3 COKa, (H+) BH+ B + H+ + &BH22+ Kal Ka2 + Ka1 (H+) + which did not (Fig. 1). In this work, no activity correction was performed because of the necessary use of acidity functions. c o (H+>2 K.1 Ka2 + Ka1 (H+) + (H+I2 + &BH33+ CO K’ai co (H+) Acid-Base Pair BH+IB In order to be sure that overlapping between the acidities of BH+ and of BH33+, BH22+ (or of B’H22+) does not exist, we (2) pale - - &B’H22+ + EBH+ K’ai + (H+> K’ai + (H+) determined the pKa of the acid-base pair BH+/B which intervenes in the two ionization schemes.The values found for the two concentrations at 240 and 275 nm, together with those of molar absorptivities of BH+ and B obtained simultaneously (see experimental part), are given in Table 1. The values are self-consistent. We note that calculated EBH+ and EB are in good agreement with the experimentally obtained values. The pKa found is somewhat lower than the pKa values of benzylamines.2l A pKa value (BH+/B) = 8.10 allowed us to be sure that in the pH range of study of the pairs BH22+/BH+ and BH33+/BH22+ (or B’H22+/BH+) (see below) all the piperidino moiety was protonated and, hence, no overlapping conferred by this acidity and no concentration problems could exist.which result from the laws of matter conservation and of equilibria in water.22 In expression (l), the only unknowns were Ka1, Ka2 and &BH22+, this last one being experimentally unattainable because of the low ratio of the two Ka values. The consequence is that the species BH22+ did not exist alone whatever the pH of the solution was. As above, &BH+ and &BH33+ were also considered as supplementary unknowns for the sake of comparison with their experimental values. Alternatively, in expression (2), the only unknown was Pa!. For the same reason as previously discussed, we also considered &B’H22+ and &BH+ as supplemen- tary unknowns. The results for the first ionization scheme are given in Table 2. B 1 ~ , 2 + N’ H *. BH+ B~H:+ Fig.4 Ionization microconstants scheme.Analyst, January 1996, Vol. 121 47 It appears that pKal and pKa2.values did not vary significantly with the analytical concentrations and with the wavelengths. The same was true for the &BH33+, &BH22+ and EBH+ values for each wavelength. Moreover, the molar absorptivities obtained were in agreement with the experimental values. Therefore, this ionization process was satisfactory. A supplementary study of these results was also carried out. The experimental absorbances were treated by pooling all their values (again according to the same ionization scheme). This was possible because, for each pH value, six replicates were available. Generally, whatever the nature of the experimental data studied, such a treatment allows the calculation of the particular cost function Uo, which takes into account only random errors.Then, comparison of Uo and U provides an estimate of the lack of fit of the mode1.23 For the investigated ionization pathway, the values were Uo = 95 and U(minimum minimorum)3oonm - 166 and U(minimum minimor~m)~5~ nm = 1 10. These values indicated the quality of the model and confirmed the preceeding results. The likelihood of the pKal value being -0.24 and the pKa2 value 1.81 is reinforced by the pKa value of 3.7 for 3,5-dimethyl-4-phenyl- 1,2,4-triazole that we determined by pH measurement and by polarographic study of NHPTT. Its polarograms exhibited only one wave, the half-wave potentials of which shifted in the range -0.46 to -0.57 V versus SCE with pH (Fig.3 and Table 3). A plot of EIR versus pH exhibited a break at pH -0.5. Therefore, the pKa value of NHPTT was close to this pH. The results obtained with the second ionization process (B’H22+/E3H+) are given in Table 4. Discrepancies found for pK’,,, EB’H22-b and EBH+ values, according to the analytical concentrations and the wavelengths used, as well as the very Table 4 Determination of the pKa of the imino and triazolo groups: results obtained according to the second ionization scheme at different concentra- tions and wavelengths Clmol l-1 h = 250nm h = 300nm 10-4 +0.66 EBH+* 15 000 &B‘HZ2+ 10 000 lJ 965 1.5 x 10-4 PK’,, +0.84 EBH+ 15 200 U 1018 &B’H22+ 10 600 * E in 1 mol-1 cm-1. +1.59 3 000 11 000 4 000 3 000 11 500 4 224 +1.55 1.8 T 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2 4 1 k,, - k,, -0- 4 2 - k22 -I- Fig.S Ionization microconstants kij versus kl I for the range kl < K,,. high objective function, are obvious, which strongly indicates a lack of fit of the model. These results confirm, a contrario, the preceeding ones. As a result, it can be concluded that the basic character of the triazolo ring does exist. Assignment of K,, and K,, Values to the Ionization Sites By comparison with the polarographic behaviour of a reference compound (chl~ramphenicol~~) it was found that the observed polarographic wave corresponded to a two-electron transfer. Owing to the well known fact that the protonated imines are reduced by a two-electron transfer,2325 we can deduce that the reduced function was the imine.Since a pKa = -0.5 value was found by the study of the displacement of the half-wave potential ,5112 versus pH, it could be deduced that pKal = -0.24 was the ionization constant of the protonated imine and hence a pKa2 value of 1.8 1 was that of the protonated triazole. Actually, this IS not quite exact because the ratio Ka!/Ka2 = 112 indicates that ionization of the two conjugated acids occurs somewhat simultaneously. The two species B1H22+ and BZH22+ coexist and the four ionization microconstants, kl 1, kzl, kI2 and k22 are effective (Fig. 4). (By accident, the microform B2Hz2+ was identical to the conjugated acid B’H22+ considered above in the second model.) As a result the K,, and Ka2 values found previously were only the macroscopic ones. They correspond to the ionization steps Ka , Ka2 BH33+ B1H22+ + B2H22+ + H+ BH+ + H+ I BH2*+ and the species we symbolized above by BH22+ were actually a mixture of the microforms B1H22+ and B2H22+ in a constant ratio.26 Between macroscopic and microscopic constants only three independent relations can be drawn: (4) k l l k12 = k21 k22 (5 1 Owing to this fact, it was not possible to obtain microscopic kj,j values without supplementary assumptions, nor was it possible to assign the found Kal and Ka2 values to a definite ionization scheme, i.e., to assimilate K,, to kll (or k21) and K.2 to k12 (or k22).We tentatively assigned the ranges of kj, values in the following manner. In a first step we plotted different values of kll, k21, k12 and k22 versus kl1 for the range of values kll < K.1 (Fig.5). In a second step we accepted the assumption that kl1 > kz1 owing to the preceding discussion. This allowed us to exclude from the range of interest values k l l < 0.8689, k21 > 0.8689, k12 > 0.031 1 and k22 < 0.031 1. Finally, considering that k22 > k21 and that k l l > k12 (for the same reason as in the second step) limits the range of interest to values 1.5678 < k l l < 1.7000 3.78 X 10-2 < k2l < 0.1692 1.58 X < k12 < 1.75 X 10-2 0.1692 < k22 < 0.7120 These values are satisfactory from a chemical standpoint. The triprotonated imine, BH33+, is a stronger acid than the diprotonated one B2H22+ and likewise for the triprotonated triazole BH33+ and the diprotonated one B1HZ2+. This can, indeed, be well explained by inductive effects induced by supplementary positive charges.The range -0.23 < pKa1, < -0.20 confirms the polarographic results. It is worth noting that48 Analyst, January 1996, Vol. 121 the pKI 1 and pK22 values are considerably lower than the known pK, values of N-alkylamines.27 To conclude, triazolodiazepines must be absolutely con- sidered as dibases, the protonation occurring on the azomethine group and on one of the nitrogen atoms of the triazolo ring. For our model compound, we found values of -0.24 and + 1.8 1 for the macroscopic ionization constants which corre- spond mainly to the ionization of the protonated imine group and of the protonated triazolo ring. Plausible ranges of microscopic constants values have been defined.The authors are grateful to Guy Bouer for his technical assistance. References Yang, S. K., J. Pharm. Sci., 1994, 83, 898. Triballet, C., Boucly, P., and Guemet, M., Bull. Chem. Soc. Fr., 1981, 2-3, 113. Pfendt, L. B., and Popovic, G. V., J. Chem. SOC., Perkin Trans. 2, 1994, 1845. Gallo, B., Alonso, R. M., Vicente, F., Ortiz, I., Irabien, A., Patriarche, G. J., and VirC, J.-C., Pharmazie, 1988, 43, 212. VirC, J.-C., Gallo Hermosa, B., and Patriarche, G. J., Analusis, 1987, 15, 499. JimCnez, R. M., Dorninguez, E., Badia, D., Alonso, R. M., Vicente, F., and Hernandez, L., J. Heterocyclic Chem., 1987, 24, 421. VirC, J.-C., and Patriarche, G. J., J. Electroanal. Chem., 1986, 214, 275. Gallo, B., Alonso, R. M., Madariaga, J. M., Patriarche, G. J., and Vir6, J.-C., Anal. Lett., 1986, 19, 1853. Konishi, M., Hirai, K., and Mori, Y., J. Pharm. Sci., 1982, 71, 1328. 10 I1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Inotsume, N., and Nakano, M., Chem. Pharm. Bull., 1980, 28, 2536. Gallo, B., Alonso, R. M., Lete, E., Badia, M. D., Patriarche, G. J., and GebeIke, M., J. Heterocyclic Chem., 1988, 25, 867. Jimenez, R. M., Alonso, R. M., Oleaga, E., Vicente, F., and Hernandez, L., Fresenius Z. Anal. Chem., 1987,329,468. Barton, D., Ollis, W. D., Comprehensive Organic Chemistry, Pergamon, Oxford, 1979, vol. 4, p. 365, Ipsen-Beaufour Industry, 35, rue Spontini, 75 116 Paris, France. Reiter, A. L., and Berg, E. G., Heterocycles, 1992, 34, 771. Perrin, D. D., and Dempsey, B., in Buffers for pH and Metal Zon Control, Chapman and Hall, London, 1974, P. 155. Britton and Robinson, J. Chem. SOC., 1931, 458. Rochester, C. H., Acidity Functions, Academic Press, London, 1970, vol. 17, p. 155. Bascombe, K. N., and Bell, R. P., J. Chem. Soc., 1959, 1096. Hooke, R., and Jeeves, T. A., J. Assoc. Comput. Mach., 1961, 8, 212. Blakwell, L. F., Fischer, A., Miller, I. J., Topsom, R. D., and Vaughan, J., J. Chem. SOC., 1964, 3588. Butler, J. N., Ionic Equilibrium. A Mathematical Approach, Addison- Wesley, Reading, MA, USA, 1964, p. 61. Draper, N., and Smith, H., Applied Regression Analysis, John Wiley, New York, USA, 2nd edn., 1981, pp. 33-40. Fossdal, K., and Jacobsen, E., Anal. Chim. Acta, 1971, 56, 105. Lund, H., Acta Chem. Scand., 1959, 13, 249. Edsall, J. T., Martin, R. B., and Hollingworth, B. R., Proc. Natl. Acad. Sci., 1958, 44, 505. Bouzard, D., Weber, A., and Le Henaff, P., Bull. SOC. Chim. Fr., 1972, 9, 3385. Paper 5104881 H Received July 24, 1995 Accepted September 20, I995