In Table VII, the effect of some interfering ions (900 pg/ml) on the atomic fluorescence signal of magnesium (9 pg/ml) in the premixed Nz0/C2H2and unpremixed A/Hz flames is shown. In all instances, there is a marked improvement using the premixed NzO/CZHzflame. The effect due to silicate is further reduced by the addition of small amounts of EDTA at pH 12. The corresponding atomic emission signals in the premixed NzO/CzHzflames only showed interference in the presence of silicate. These experiments indicate that the use of a high temperature, reducing flame is essential in any practical analysis. Furthermore, it is possible to use such flames in conjunction with dc measurement systems even in high background regions. D. Extension to Analysis of Elements Forming Stable Monoxides. Atomic fluorescence was observed using the premixed N20/CzH2flame for a number of elements which fall into this class-e.g., aluminum (3961 A), beryllium (2349 A), and germanium (2652 A). However, for the purposes of evaluating the premixed flame, the atomic fluorescence characteristics of beryllium were examined. Other workers (8, 9) have also studied the atomic fluorescence of beryllium using premixed laminar NzO/C2Hz flames. As with other elements prone to forming stable monoxides, the greatest sensitivity toward beryllium was obtained with a fuel-rich flame (NzO, 17 psi, and C2H2,10 psi) and with a
height of measurement about 4 cm above the top of the burner head. In Figures 12 and 13, the change in atomic fluorescence signal for a solution containing 240 pg/ml beryllium with varying C2H2 and NzO gas pressures and height of measurement is shown. A linear analytical working curve (log-log plot) was obtained up to 24 pg/ml beryllium; the limit of detection (signa1:noise = 1) was 0.04 pg/ml using a slit width of 0.25 mm. This compares favorably with the value of 0.01 p g / d obtained by Hingle et at. (9) using a quartz separated premixed laminar flame and a specially designed burner in conjunction with an ac measurement system. It is also considerably better than the value of 0.5 p g / d obtained by Robinson and Hsu (8) using again a specially designed burner for a premixed NzO/CzHz flame, an ac measurement system, and a high intensity hollow cathode lamp as source of excitation. Scatter is not observed under our experimental conditions. In Figure 14, the atomic emission profiles for beryllium in the very fuel-rich, slightly fuel-rich, and fuel-lean premixed NzO/C2Hzflames NzO (17 psi and CzH2 11, 8, and 6 psi, respectively) are shown. The emission profiles for other refractory oxide forming elements-e.g., aluminum and barium were similar. Interferences were not investigated because Robinson and Hsu (8) have already made an extensive study of this aspect.
(8) J. W. Robinson and C. J. Hsu, Anal. Chim. Acta, 43, 109 (1968). (9) D. H. Hingle, G . F. Kirkbright, and T. S. West, Analyst, 93, 522 (1968).
RECEIVED for review March 10, 1969. Accepted August 4, 1969. This work was supported by AFOSR(SRC)-OAR, U.S.A.F. Grant No. AF-AFOSR-69-1685.
Determination of a Protonation Scheme for Isochlortetracycl ine Using Nuclear Magnetic Resonance Ulrich W. Kesselring' and Leslie Z. Benet2 College of Pharmacy, Washington State University, Pullman, Wash. 99163
The macro- and microdissociation constants for isochlortetracycline in a 50-50 wt/wt methanol-water solvent mixture were determined by the use of potentiometric and NMR techniques. A modification of the equations used to calculate microconstants for compounds exhibiting widely separated macroconstants is discussed and shown to yield more accurate results. Equations, after Edsall and Wyman, are developed to calculate the microconstants for isochlortetracycline where there is strong overlap between pKz and pK,. Calculations involving chemical shift data from two different sites are shown to yield almost identical microdissociation constants. On the basis of the comparable shift values for the phenolic diketone system in isochlortetracyline and in other tetracyclines, it appears that the site of dissociation in this system may be assigned to the hydroxyl group at carbon 10.
BEFORE THE EXACT STRUCTURES of the tetracycline compounds were known, Stephens et al. ( I ) reported that the hydrochloride salts of oxy- and chlortetracycline exhibited three dissociation constants whose acid pK values were approximately 3, 7, and 'Present address, Ecole de Pharmacie, Place du Chitleau 3, Lausanne, Switzerland 2 Present address. School of Pharmacy, University of California, San Francisco Medical Center, San Francisco, Calif. 94122
9. In the decade following this report, a minor controversy arose as to the assignment of the particular functional groups giving rise to the observed macroscopic constants. This controversy was reviewed by Rigler et al. (2). It is interesting to note that this disagreement was largely the result of a misprint in a footnote of a 1956 paper by Stephens et al. (3), for which the correction (4) has been generally ignored. Rigler et al. (2)attempted to resolve the lack of agreement as to the pK assignment of the particular functional groups on tetracycline by determining the microscopic dissociation schemes of three members of the tetracycline series using nuclear magnetic resonance. By investigating tetracycline, its 4-epimer, and its quaternary methyl iodide, these authors concluded that the dissociation of the first proton occurs primarily from site A (see Structure' I). For 4-epi-tetracycline, the second dissociating hydrogen was found to come primarily from site C while the third dissociating hydrogen came pri(1) C. R. Stephens et al., J. Amer. Chem. SOC.,74,4979 (1952). (2) N. E. Rigler, S.P. Bag, D. E. Leyden, J. L. Sudmeier, and C. N. Reilley, ANAL.CHEM., 37, 872 (1965). (3) C. R. Stephens, K. Murai, K. J. Brunings, andR. B. Woodward, J. Amer. Chem. SOC.,78, 4155 (1956). (4) Zbid.,p. 6425. VOL. 41, NO. 12, OCTOBER 1969
1535
marily from site B. For tetracycline they found a strong overlap of dissociation for the second and third protons from sites C and B ; however, dissociation from site B slightly preceded dissociation from site C. Rigler et al. (2) also detected a fourth ionizable proton in tetracycline methiodide; the pK of which was found by potentiometric titration and confirmed by spectrometry to be 10.67. The C' moiety was assigned as the probable site of dissociation for this proton. H,
CH. OH
NHH(cH,),
9-
a
II-
-
D-RING
'
9-
-
T 7, I ,
-
n 5-
-
3-
-
I
I In this paper we have applied the Rigler et al. (2) technique to one of the degradation products of the tetracyclines, specifically isochlortetracycline (11). This compound is well suited for an NMR determination of microconstants, because phthalide formation of ring C breaks the conjugation of the phenolic diketone system and minimizes proton shifts at sites remote from the site of deprotonation. In addition, we have discussed the possible errors and hidden assumptions in the calculations used by Rigler et al. (2) and have proposed an alternative method of calculation.
I
l
l
1
d
l IOl 20 l l 30 l SHIFT DOWNFIELD
40 ~
l
Figure 1. Chemical shift data for isochlortetracycline at various p[H+]values
A-60 high resolution spectrometer, whose probe was maintained at 25 + 0.5 "C. The NMR spectrometer was tuned up on a standard acetal sample which resulted in resolution always better than 0.4 cps. In order to get more accurate readings of the chemical shifts, the sweep width used was 250 cps and the scanning speed 0.5 cps/sec. Therefore, the spectra of the Dring protons had to be recorded separately from the 4-dimethyl and 6-methyl groups. RESULTS
II EXPERIMENTAL
Our data for isochlortetracycline, donated by Lederle Laboratories, in a 50-50 wt/wt methanol-water solution was obtained using conditions and procedures similar to those employed by Rigler et al. (2) with three exceptions: a) Both NMR titrations and pH determinations were carried out at 25 0.5 "C.; b) Concentrations of drug ranged from 0.007M to 0.017M (it is much easier to see the ring D protons in isochlortetracycline than in chlortetracycline or tetracycline) ; c) Because of the rapid rate of epimerization of isochlortetrathe drug was always placed into solution in a basic cycline (3, medium and then back titrated to the desired pH. Therefore, in the pH range essential to determine the microconstants for pK2and pKa the epimerization problem was avoided. To individual samples of either 8 or 12 mg isochlotetracycline neutral, from 0.02 to 0.11 ml of COz-free KOH, 1.098N, and 1.00 ml of a water-alcohol mixture (62.56 parts methanol, 50.15 parts of water, and 1.00 part tertiary butyl alcohol) were added. From a microburet, enough HC1 1.011N was added to reach the desired pH. In order to maintain the correct methanol percentage, methanol was added from a semi-micro pipet corresponding to the total volume of aqueous titrant added, The exact pH was determined at 25 + 0.5 OC using a Beckrnan Research pH Meter equipped with #39166 probe assembly. The NMR spectra were recorded using a Varian (5) J. R. D. McCormick et al., J. Amer. Chem. SOC.,79, 2849
(1957). 1536
0
ANALYTICAL CHEMISTRY
The stoichiometric dissociation constants (Kat) were determined potentiometrically from the same solutions as were used in measuring the NMR shifts. The values were determined by the methods previously described by Benet and Goyan (6, 7), assuming that the drug sample was pure as received except for an unknown number of waters of hydration. Three dissociation constants were observed, the pK's of which were found to be 4.19,7.23, 7.83. Structural and general considerations (shielding effects on neighboring groups, absolute possible peak position, area under peak, qualitative pH dependence) and comparison with published values for tetracycline and some of its derivatives (2,8) lead to the following peak assignments for isochlortetracycline in acid medium at pH 1.7, relative to tertiary butyl alcohol as an internal standard: at 6.2 ppm downfield- doublet due to H at C-8 at 5.7 ppm downfield- doublet due to H at C-9 at 5.4 ppm downfield- amide protons at 1.9 ppm downfield- 4-dimethylamino protons at 0.4 ppm downfield- 6-methyl protons Because of the low solubility of the compound, the unavailability of a CAT (computer of average transients), and because of the broad sidebands due to the solvent mixture, no attempt has been made to obtain a fully defined NMR spectrum of isochlortetracycline. The peak positions of the D-ring and 4-dimethyl protons are directly related to the degree of protonation of the molecule and therefore to hydrogen ion concentration of the solution. Figure 1 shows a plot of chemical shift values of 4-dimethyl(6) L.Z.Benet and J. E. Goyan, J . Pharm. Sci., 54,983 (1965). (7) L.Z.Benet and J. E. Goyan, ibid., 56, 665 (1967). (8) M.Schach von Wittenau and R. K. Blackwood, J. Org. Chem., 31, 613 (1966).
~
,
~~
Figure 2. Microscopic acid-base equilibria for tetracyclines, after Rigler et al. (2)
amino singlet and of the low field doublet for C-8 in the D-ring pattern for isochlortetracycline. The plots of the shift values in cps us. pH yield titration curves with inflection points at about pH 7.45 for the D-ring protons, and pH 7.57 for dimethylamino singlet. A total shift of 0.70 ppm was observed for the dimethylamino protons and 0.35 ppm for the first signal of the D-ring protons. Rigler et al. (2) found total chemical shift changes of approximately 0.65 and 0.28 ppm for these two signals in tetracycline and 4-epi-tetracycline. Because the accuracy of estimated activity coefficients for the various ionic species of isochlortetracycline would be questionable, and because all equilibrium equations necessary to calculate both macro- and microdissociation constants are a function of concentration, all pH values discussed in this work and plotted on Figure 1 are, in fact, the negative log of the hydrogen ion concentrations. The conversion of hydrogen ion activity read on the pH meter to hydrogen ion concentration was carried out as previously described (6, 3. DISCUSSION
As would be expected for a compound with strongly overlapping second and third macrodissociation constants, the proton dissociation from the dimethylamino group shows strong overlapping with the dissociating proton measured by the D-ring proton shift (Figure 1). However, in contrast to the work of Rigler et al. with tetracycline, we find that the shift of the D-ring protons occurs previous to the shift of the protons at the dimethylamino group. Therefore, qualitatively in the present study it appears that the dissociation occurs from site C before dissociation at site B, in contrast to the tetracycline data where dissociation at site B occurs before dissociation at site C. Although the qualitative assignment of relative order of dissociation seems clear, a problem does arise in attempting to assign microconstants to the dissociation. The appropriate microconstants may be calculated if one can determine the percent protonation or deprotonation at each site as a function of pH. The percentage deprotonation for a site may be determined at each pH by dividing the shift of the nonlabile protons adjacent to the dissociation site by the total shift for that site observed over the entire pH range of the titration. However, this calculation is only valid if the shift of nonlabile protons is exclusively a function of the dissociation at the adjacent site and is independent of proton dissociation at remote sites. Rigler et al. (2) were not able to tell from tetracycline data alone whether the shifts at the measurable sites were independent of protonation at remote sites. Therefore, they also observed the proton shifts for tetracycline methiodide and used these shifts as corrections for the effect of dissociation at site A on the shift at sites B and C, and the effect of dissociation at site C on the shift observed at site B. This was possible because any observed shifts with tetracycline methiodide could not be caused by dissociation at the nonlabile site B in tetracycline methiodide. Because of the unavailability of 4-epi-tetracycline methiodide, Rigler et al. also had to use the shifts for tetracycline methiodide as corrections for 4-epi-tetracycline. Knowing that there appears to be such a
Table I. Macrodissociation Constants for Four Tetracyclines in 50-50 wt/wt Methanol-Water PKI' pKzC pK3' Tetracycline (2) 4.40 7.80 9.40 CEpi-tetracycline(2) 4.80 8.00 9.50 Chlortetracycline (14) 4.26 7.61 9.36 Isochlortetracycline 4.19 7.23 7.83
marked difference in the protonation scheme of epi-tetracyclines (very little overlap of dissociation for sites B and C ) with respect to tetracyclines (strong overlap), it seems questionable that the influence of the dissociation at site A would appear the same for an epi-methiodide as for a non-epi-methiodide. However, because these corrections are quite small (about 4 to 5 cps shift for the effect of the dissociation of A on site B, and zero shift for the effect of dissociation of C on site B and A on site C), the possible errors in utilizing tetracycline methiodide as a correction for 4-epi-tetracycline are probably within limits of the experimental error involved in the measurements. Because of the unavailability of isochlortetracycline methiodide we were also required, in one case, to use the effect of dissociation of A on the shift of site B for tetracycline methiodide as a correction for isochlortetracycline. Calculation of Microscopic Dissociation Constants. Edsall and Wyman (9) have shown that the microscopic dissociation constants may be calculated from values for the per cent protonation of the various sites and from the macroscopic dissociation constants. Figure 2, after Rigler et al. (2), depicts all possible dissociation equilibria for tetracycline, with superscripts indicating the electrostatic charge on the corresponding sites in Structure I. Site C' is omitted because it is assumed to be negligibly dissociated under the experimental conditions. Rigler and coworkers (2) have developed the following equations and methodology for calculating the microconstants. At a pH value corresponding to pK1, Equation 1 is assumed to be true, and at a pH value corresponding to pK3, Equation l a is assumed to be valid. [A'B'CO]
=
[A-B'CO]
[A-B'C-] = [A-B'C']
+ [AoBoCo]+ [AOBtC-] = 0.5T + [A-WC-] + [AoBoC-]= 0.5T
(1)
(la) where T equals the stoichiometric concentration of tetracycline (the total of all possible species). Implicit in Equations 1 and l a are the assumption that the macrodissociation constants are sufficiently separated so that at pH = pK1 and at pH = pK3the only species present are those listed in the above equations. Using the microconstants reported by Rigler et al. the macroconstants for tetracycline and 4-epi-tetracycline may be calculated as presented in Table I. Thus, it may be seen that Equation 1 is valid for tetracycline and 4-epi-tetracycline since there is such a wide spread between pK1 and pK2. Although pK2 and pK3 are closer, species other than those listed in Equation la, would be present at a level no greater than 3 and therefore can be safely ignored. Two of the microconstants (as examples) are described by the following equations, defining pH = pKI in Equation 2 and pH = pK3 in Equation 2a. [H+][A--B+CO] - K*[A-B+C"] k1 = [A'BtC"] 0.5T (9) J. T. Edsall and J. Wyman, "Biophysical Chemistry," Vol. I , Academic Press, New York, 1958, p. 495. VOL. 41, NO. 12, OCTOBER 1969
1537
Table 11. Per Cent Protonation at Various Values of n for Isochlortetracycline n PH A B C 3.0 1.69 100 100 100 2.5 4.19 56.5 93.5 100 2.0 5.71 17 83 100 1.0 7.53 6 51 43 0.5 8.01 2 29 19 0 10.31 and11.96 0 0 0
Now in order to calculate the microconstants in the above equations, Rigler et at. related the unknown concentrations to the degree of protonation of the particular sites as in Equations 3 and 3a. [A-B+Co]= 0.5 (1 - &)T (3) (3a)
[A-BOC"] = 0.5 (fic)T
where hAis defined as the degree of protonation for site A at a pH where 2 total equivalents of dissociable hydrogen (ignoring the possibility of dissociation at C') are present per molecule of tetracycline-Le., at pH = (pK1 pK2)/2. And 6, equals the fractional degree of protonation of site C when pH = (pK2 pK3)/2--i.e., when there is one total equivdent of dissociable hydrogen per tetracycline molecule. Note that the degrees of protonation are measured at a different pH than those specified in defining Equations 2 and 2a. Therefore, Equation 3a contains the implicit assumption that the relative percentages of protonation for the three sites will be the same at pH = pK3 as when pH = (pK2 pK3)/2. Equation 3 contains a similar assumption related to pK1 and pK2. In addition Equation 3a contains either a second implicit assumption that the only species existing at pH = (pK2 pK3)/2 are those on the right hand side of Equation l a or a third assumption that if other species do exist at this pH, they will only dissociate through one specific pathway-e.g., if A-B+Co and A o B f C oexist at pH = (pK2 pK3)/2 they may only be converted into A-B+C- as pH is raised, and if AoBoCo and A-B+Co exist they may only be converted into A-BOC" as the pH is raised. The second assumption could only be valid if pK2 and pK3 were separated by at least 3 to 4 pK units. The third assumption is necessary to maintain a constant relative percentage for each dissociation site in going from n = 1 to n = l/?, where n is defined as the total equivalents of dissociable hydrogen per molecule of tetracycline. But this third assumption is impossible because in the example above A-B+Co was required to go exclusively by one pathway for the measurement of the degree of protonation on site B and exclusively by another pathway for the degree of protonation on site C. However, the necessity for these three assumptions is only the result of using degrees of protonation in Equations 3 and 3a which are not measured at the pH's specified in Equations 2 and 2a. If instead Equations 4 and 4a are substituted into Equations 2 and 2a the only assumptions necessary are those used in Equations 1 and l a which were shown above to be valid.
+
+
+
+
+
proton dissociations. The error in using Equation 3a would be more substantial than that for Equation 3 because pK2 and pK3 are much closer than pK1 and pK2. Because the microconstants for the second dissociating hydrogens are determined using the calculated values of the microconstants for the first and thud dissociating hydrogens, an error in calculating k123 would also be reflected in the calculated values for k12and k21. The errors introduced in using Equation 3a will probably not substantially affect the relative relationship of the microconstants as determined by Rigler et a1.-Le., for tetracycline k?1> kl2 and k132 > k1~3-but they could have quite an effect on the calculated A pk values which were determined as measures of the interactions between dissociating sites. Microscopic Dissociation Constants for Isochlortetracycline. Table I1 shows the per cent protonation of isochlortetracycline at sites A, B, and C (see structure 11) for various values of n. From Figure 1, it can be seen that there is no measurable shift in the D-ring protons until a pH above 5.3 was reached. Thus, it may be concluded that negligible dissociation takes place at site C during the dissociation of the first proton from site A , and microconstants k3, k31,and k32 may be eliminated from the dissociation scheme (Figure 2) for isochlortetracycline. The first microscopic dissociation constants may be calculated using Equation 4 for the per cent protonation of the A and B sites at n = 2l/2 with the following results: pkl = 4.25 and pkn = 5.07. If Equation 3 is used: pkl = 4.27 and pkz = 4.96. As explained above, the error involved in using Equation 3 for this calculation is not very great, because pKl and pK2 are separated by 3 pK units. Although the first microscopic dissociation constants may be determined unambiguously, the third microscopic constants (k123 etc.) may not be determined for isochlortetracycline using either Equations 4a or 3a, because pK2 and pK3 are so close for this compound that the assumptions necessary for Equation l a are not valid. However, we may get an estimate of the remaining microconstants if we assume that the concentration of AoBoC- is negligible at pH values close to the point in the isochlortetracycline titration where n = 1.0. This is probably a good assumption, as Table I1 and Figure 2 show that at n = 1 the total concentration of species with unionized protons at site A (AoBoCoand AoBoC-) is only 6%. With this approximation, the remaining microconstants may be determined by an adaption of the method Edsall and Wyman (IO) utilized in treating the cysteine data of Benesch and Benesch (11). At 50% ionization of site C (pH = 7.45 from Figure 1): [A-BfCO]
+ [AoBoCo]+ [A-BOCO] = [ A - B f C ] + [A-BOC] (5)
The concentration of AoB+Cowould be a negligible quantity at this pH, due to the wide separation of pK1 and pK2,and it may be safely omitted from Equation 5. Replacing the concentrations of the microspecies in terms of hydrogen ion concentration and microconstants, Equation 5 becomes:
Rearranging Equation 6 yields : (ki kz) [H+I2 ki[H+] (kiz - k13) = klk1zk123
+
+
____
Because Rigler et al. (2) used Equations 3 and 3a instead of Equations 4 and 4a, some of their microconstants are probably slightly inaccurate, especially those for the second and third 1538
ANALYTICAL CHEMISTRY
~
~
~
~
_
_
_
_
(7)
_
(10) J. T. Edsall and J. Wyman, "Biophysical Chemistry," Vol. I, Academic Press, New York, 1958, p 500. (11) R. E. Benesch and R. Benesch, J. Amer. Chem. SOC.,77, 5877 (1955).
Utilizing the relationships between macro- and microdissociation constants, as presented by Edsall and Wyman (9), Equation 7 becomes : K1[H+l2
+ ki[H+] (klz - ku) = K1K2K3
(8)
Under the assumption made above, Kz has been defined as:
Kz =
+
([A-B"C"] [A-B+C-l) [H+l ([A-BfCo] [AoB"CoI)
+
(9)
Table 111. Microscopic Dissociation Constants for Isochlortetracycline Method of calculation Shift at Shift at site C site B Equation 4 Equation 3 Pkl 4.25 4.27 Pk2 5.07 4.96 7.63 7.62 pkii $13
Substituting for the concentrations of microspecies as was done in going from Equation 5 to Equation 6, and cancelling out similar terms in the numerator and denominator, results in the following definition for Kz:
Therefore, at a pH corresponding to 5 0 z dissociation of site C, we have two equations (Equations 8 and 10) and two unknowns (klz and k13), and we may therefore solve for the five remaining microconstants by utilizing the additional rela= K1KzK3and klklz = kzkzl. tionships: klkl~klz3= klk~3k~32 This calculation results in the values presented in Table I11 where "Shift at Site C" is listed as the method of calculation. The microconstants may also be calculated at 50 ionization of site B (pH = 7.57 from Figure 1) by using a variant of Equation 5 where the concentrations of species containing a positively charged site B are set equal to the concentrations of species containing a neutral site B. Values calculated at 50 ionization of site B are presented in Table 111. The close correspondence between the values calculated at each site seems to confirm the validity of the assumption that the concentration of AoBoC- is negligible at the pH values relevant to the calculations. In addition, it should be noted that the shift values used in the calculations involving site C required no corrections for the influence of proton dissociations at other sites (see above), while the shift values used in the calculations involving site B were corrected for the effect of dissociation of A as measured by Rigler et al. (2) for tetracycline methiodide. The close correspondence between the two sets of microconstants in Table 111 shows that the tetracycline methiodide corrections for site B are probably valid for isochlortetracycline. This is not too surprising because the major structural and steric differences between isochlortetracycline and tetracycline occur at positions in the molecules remote from sites A and B. A comparison of the macroconstants for isochlortetracycline, chlortetracycline, and tetracycline in Table I shows that the breakdown in the B-C-D ring conjugation causes a large shift for pK3. Although all of the microconstants for the second and third dissociating hydrogens increase in strength as compared to those of tetracycline (2), those microconstants for site C dissociation increase 3 to 4 times more than those for site B. Due to the similarity of the total NMR shifts for isochlortetracycline and tetracycline, it would appear that we would confirm the results of Rigler et al. (2) as to site C being the dissociating site in the tetracyclines as opposed to C'. Finally it is possible that the 5-member lactone ring opens in the basic medium, and that this opening is consuming base and being reflected in the shift of the C-8 hydrogen. For this to be true, it must take place in about 10 minutes, the time required to bring the solution in the NMR tube to 25 "C, and must be a relatively rapid reversible reaction. Otherwise it could be detected, especially from the NMR spectra. Evidence against an irreversible breakdown of the molecule in
z
z
pkn
pk123 pkm
7.45 6.81 7.37 7.55
7.46 6.79 7.38 7.53
basic medium has been acquired by comparison of the NMR spectra obtained from fresh solutions in the pH range 7 to 8.5 with those from samples at pH 10.3 and 12.0 which are backtitrated to the 7 to 8.5 pH range. At any particular pH, no significant difference in shift was observed between the fresh samples and those taken to a higher pH and then backtitrated to the same pH as the fresh sample. If isochlortetracycline were stable under the experimental conditions, or if the lactone ring opening were completely reversible and pH dependent, a back-titration (with acid) should yield exactly the same values for the dissociation constants as the normal forward titration (with base). The back-titration of isochlortetracycline yielded slightly lower values for pK2 and pK3, whereas pK1 was found to be exactly the same. The observed differences, however, are relatively small (0.04 and 0.08) and it is difficult to conclude whether there is a real variation in the observed pK's or whether the difference is due only to the accumulated errors in concentration and volume. On the basis of the repeatability of the NMR shifts and dissociation constant determinations in both deprotonation and protonation titrations, it would appear (but it cannot be stated unambiguously) that dissociation does take place from the phenolic group at C-10, a completely reversible reaction, rather than by opening of the lactone ring, which should require different mechanisms in opening and closing the C ring. We would not expect these two reactions to show an identical pH profile. The above calculations and the work of Rigler et at. point out the uncertainty in the methodology employed in calculating stability constants for tetracycline complexes with metal ions (6, 12, 13). In these calculations it has always been assumed that the tetracycline species with a single negative charge--i.e., when n < 2 . 0 4 s the complexing ligand. However, when there are overlapping microconstants, dissociation is taking place at two or three different sites on the molecule, yet in the stability constant calculation, chelation at these sites is considered equivalent. A more detailed treatment of the tetracycline stability constants will be discussed in a future publication (14). ACKNOWLEDGMENT
The authors thank Lewis J. Leeson for his most helpful comments on the manuscript and for suggesting the use of Equations 9 and 10, Donald R. Galpin for his aid in interpreting the NMR spectra, and Lederle Laboratories for supplying tetracycline compounds. Received for review March 14, 1969. Accepted July 14, 1969. (12) A. Albert and C. W. Rees,Nature, 177,433 (1956). (13) J. T. Doluisio and A. N. Martin, J . Med. Chem., 6, 16 (1963). (14) U. W. Kesselring and L. Z . Benet, Washington State University, unpublished work, 1968.
VOL. 41, NO, 12, OCTOBER 1969
1539