Carbon 13 Magnetic Resonance Studies of DL-2-(a-Hydroxyethyl)thiamin and Related Compounds RELATION OF KINETIC ACIDITY TO ELECTRONIC FACTORS IN THIAMIN CATALYSIS*

13 NMR spectra have been obtained for aqueous solutions of oL-2-(tu-hydroxyethyl)thiamin, DL-2-(CV-hydroxybenzyl)thiamin, DL-2-(a-hydroxybenzyl)oxythiamin, and related N-3 methyl and N-3 benzyl analogs. The unusually large downfield shift of the 13C resonance of C-2 of hydroxyethylthiamin suggests that this carbon bears a partial positive charge. This result stands in contrast to results of x-ray crystallographic studies of hydroxyethylthiamin, which place a partial negative charge on C-2 (Pletcher, J., and Sax, M. (1974) J. Am. Chem. Sot. 96, 155-165). A partial positive charge on C-2 helps to explain the facility of carbanion formation at the cy carbon both enzymatically and in model systems. The rates of proton-deuteron exchange of (C-(U)-H with solvent deuterium, and of release of aldehyde to regenerate thiamin have been measured for hydroxyethylthiamin and analogs. The differences in kinetic acidity of (C-n)-H and of rates of aldehyde release are rationalized

A partial positive charge on C-2 helps to explain the facility of carbanion formation at the cy carbon both enzymatically and in model systems.
The rates of proton-deuteron exchange of (C-(U)-H with solvent deuterium, and of release of aldehyde to regenerate thiamin have been measured for hydroxyethylthiamin and analogs. The differences in kinetic acidity of (C-n)-H and of rates of aldehyde release are rationalized in terms of differing electron-withdrawing abilities of the substituents attached to N-3, and appear not to be related to intramolecular basic catalysis of these processes by the C-4' amino group.
Since Breslow (1,2) proposed the mechanism of thiamin action which is broadly accepted today, numerous attempts have been made to correlate the catalytic efficiency of thiamin in each step of the mechanism with a description of the electronic structure of this catalyst, but these attempts have been only modestly successful. Molecular orbital calculations of the thiamin rings by Pullman and Pullman (3,4) placed a partial negative charge on C-2 of the thiazolium ring of thiamin and suggested that the amount of electron excess on C-2 plays an important role in determining the ease of deprotonation of thiazolium salts. When the relative order of kinetic acidity of the C-2 protons of thiazolium and oxazolium salts were found to be opposite to that predicted is discussed in the miniprint following this paper.3 The compounds studied are shown in Fig. 1P.

RESULTS
The complete 13C NMR spectrum of 0.6 M nL-P-(cu-hydroxyethyl)thiamin (Ib, Fig. 1P) in D,O is shown in Fig. 2P. The assignment is based on the splitting of the resonance lines under conditions in which proton noise decoupling is absent (Table  IS) and on the reported assignment of thiamin (7). The detailed rationale for the assignment of HET and its analogs and a tabular presentation of chemical shifts are presented in the miniprint.
The resonance of C-2 of 46 undergoes a downfield shift of 21.4 ppm (Table  IIS) relative to the same carbon of 4~. This shift is similar to that observed for the quaternary carbon of cu-phenylethanol (C-l", 6, Fig. 1P) of 17.4 ppm downfield relative to benzene (Table  IIS). Carbons 4 and 5 of 46 were assigned by the similarity in chemical shift of these resonances of 46 and 4a (Table  IIS). The C-cu and a-CH, carbon resonances were assigned based on the similarity in chemical shift of these resonances to those of the corresponding resonances of n-phenylethanol, 6, (Table IIS). When the 13C NMR spectrum of HET, lb, is compared to thiamin, la (Table  IS), an even larger downfield shift, of 24.0 ppm, is observed for the C-2 resonance.
The assignment of C-4 and C-5 of HET is assisted by the small downfield shift of C-4 and the small upfield shift of C-5 expected for these resonances upon substitution at C-2 (compare 4b with 4~). There is no ambiguity in assigning C-4, C-5, and C-6' since C-6' is a doublet and C-4 and C-5 are singlets under conditions of proton coupling ( assigned to the C-2", (C-3", C-4"), and C-1" carbons.
The most upfield of this group is assigned to C-2" and the most downfield resonance to C-l", based on the reported assignment of n-phenylethanol, benzyl alcohol, and toluene (13, 14). The C-3" and C-4" resonances were not resolved.
There is no confusion between the resonances of the thiazolium carbon atoms, C-4 and C-5, and the phenyl resonances, C-2" and (C-3", C-4"), since these latter resonances appeared as doublets and the former as singlets under conditions of proton coupling (Table  IS). Another doublet resonance, C-6', is not confused with the phenyl carbon resonances because this signal was shifted downfield upon deprotonation of N-l', as in thiamin and HET (Fig. lS, panel B). A completely unambiguous assignment of C-5 and C-l" cannot be made because these resonances differ by only 1.7 ppm and both appear as singlets with proton coupling. However, the near identity of the chemical shifts of C-5 of Ic with that of lb, (Table  IS) which has no C-l" resonance, strongly supports the original assignment. Further arguments bearing on this point are presented in the supplement.
The assignment of C-2' and C-4' resonances was achieved by comparison with the spectra of model compounds, 4c and 5c, which have no pyrimidine ring (Table  IIS) "The (C-3", C-4") and (C-3"', C-4"') resonances and the C-2" and C-2"' resonances are unresolved constant of HBT. which is similar to that coupling in thiamin (7). This additional information reinforces the assignment of resonances in the spectrum of lc. The 0, value for a phenyl group is + 0.10, relative to hydrogen = 0 (33) and the U* value for a benzyl group is + 0.22, relative to methyl = 0 (33). Although cr values for an aminopyrimidyl substituent have not been published, such a ring should be more electronwithdrawing than a benzene ring, and this is confirmed by pK measurements (30). We have also reported the rates of exchange of the C-2 proton of thiamin and 4a (7). For thiamin this exchange is 25 times faster than for the N-3 methyl analog. Similarly, the rate for 3-benzyl-4-methylthiazolium bromide is 3 times as fast as that for 3,4-dimethylthiazolium bromide7 (2). The relative rates of exchange in thiazolium compounds with an aminopyrimidyl, a benzyl, or a methyl substituent are thus 25:3:1.
For the same substituents, the relative rates of exchange of the C-a position of hydroxybenzyl derivatives (Table  IIP) are 2.9:1.3:1.0, respectively. It is remarkable that the compound with a benzyl substituent at N-3 occupies the same relative position between the aminopyrimidyl and methyl substituent in these two series, suggesting that a common factor controls both relative orders, i.e. the inductive effect of the substituent at N-3. The attenuation of substituent effects on exchange at the C-o position relative to exchange at C-2 is understandable, since the reaction center is moved 1 carbon further away from the substituent. The inductive effect of the substituent at N-3 as measured by the C-2 chemical shift and by the reported pK measurement of methylamine, benzylamine and 4-amino-5-aminomethyl-2-methylpyrimidine correlate well with n-CH acidity (Fig. 3P) On the other hand, if step B is slower than A, * the rate of aldehyde release may also be accelerated somewhat by an electron-withdrawing substituent at N-3, since in this step the negative charge is maintained but moved closer to the substituent. The substituent at N-3 may also affect the pK of the cu-OH group.
In this case, even if step B is slower than A and unresponsive to differing inductive effects at N-3, the amount of 9 present would be increased by an electron-withdrawing substituent at N-3, and the overall rate would be accelerated. The results in Table IIP suggest that an inductive effect is indeed operative, since the ratio of rates of aldehyde release for lc, 3c and 4c, with an aminopyrimidyl substituent, a benzyl, and a methyl substituent at N-3 are >9.6: >5.4:1.
This suggestion is further supported by the observation that the cu.hydroxybenzyl derivative of P'-trifluoromethylthiamin (35) releases benzaldehyde 2.7 times as fast as HBT.9 In this case. the ?'-trifluoromethylpvrimidine ring presumably exerts a greater electron-withdrawing effect than the normal 2'.methylpyrimidine ring.
The results of the present study contribute to a developing understanding of the chemistry of thiamin and further suggest that the electronic structure of thiamin is in accord with the natural suitability of thiamin to enzymatic carbanion-forming processes.
'The absence of general base catalysis of aldehyde release of HB? (Fig. 2s) suggests that carbon-carbon bond cleavage (step B, Scheme 11 may be the rate-determining step, since in this case, the rate would be specific base-catalyzed (34). 9 G. Bantle, unpublished observation.