The Correct Structures of "Dihydrothiamine". Resolution of a Long-Standing Controversy

Article abstract of DOI:10.1021/jo970782n

Reduction of thiamin (1) by a borohydride in water first gives bicyclic perhydrofuro[2,3-d]thiazole 3. Heating 3 in water generates fused tricyclic pyrimido[4,5-d]thiazolo[3,4-a]pyrimidine 4. X-ray crystal structures indicate the stereocenters are cis in 3 and cis-cis in 4. The kinetics of reduction of both 3 and 4 to tetrahydrothiamin 5 by additional aqueous borohydride show that 3 is the kinetic product of the reduction of 1 while 4 is the thermodynamic product produced from 3 in an acid-catalyzed ring-opening ring-closing isomerization process. Several structural assignments of "dihydrothiamine" presented some 40 years ago are incorrect. Subsequent reports based on the incorrect structures need to be reinterpreted.

Full text of DOI:10.1021/jo970782n

6760
J . Org. Chem. 1997, 62, 6760-6766
Th e Cor r ect Str u ctu r es of “Dih yd r oth ia m in e”. Resolu tion of a
Lon g-Sta n d in g Con tr over sy
J ohn A. Zoltewicz,* Carlton D. Dill, and Khalil A. Abboud
Department of Chemistry, University of Florida, Gainesville, Florida 32611-7200X
Reduction of thiamin (1) by a borohydride in water first gives bicyclic perhydrofuro[2,3-d]thiazole
3. Heating 3 in water generates fused tricyclic pyrimido[4,5-d]thiazolo[3,4-a]pyrimidine 4. X-ray
crystal structures indicate the stereocenters are cis in 3 and cis-cis in 4. The kinetics of reduction
of both 3 and 4 to tetrahydrothiamin 5 by additional aqueous borohydride show that 3 is the kinetic
product of the reduction of 1 while 4 is the thermodynamic product produced from 3 in an acid-
catalyzed ring-opening ring-closing isomerization process. Several structural assignments of
“dihydrothiamine” presented some 40 years ago are incorrect. Subsequent reports based on the
incorrect structures need to be reinterpreted.
In 1950 Karrer and Krishna reported that the reduc-
tion of thiamin1 (vitamin B1, 1) using lithium aluminum
hydride gave the sparingly water soluble thiazoline 2 (mp
150 °C) by the addition of hydride ion to position 2 of the
thiazolium ring.2-4 In 1957 Bonvicino and Hennessy
reported the same structure 2 (mp 151 °C) but prepared
the material by trimethoxyborohydride reduction of 1 at
a reduced temperature in methanol-water. They also
claimed that merely heating this dihydro material in
water gave an isomer, mp 175 °C, to which they assigned
structure 3, a substituted bicyclic perhydrofuro[2,3-
d]thiazole.5 About the same time, scientists at the
Takeda laboratories were preparing several dihydrothia-
mins by a number of routes, including hydride reduction
of the vitamin and cyclization reactions that also made
derivatives of the reduced materials. They reported three
“dihydrothiamines”: normal (mp 150 °C), iso (mp 160 °C),
and pseudo (mp 175 °C). None of these was said to be
the simple hydride adduct 2. Instead, normal and iso
were postulated to be ring-fused 3 having trans and cis
stereochemistry, respectively, and the pseudo was said
to be the fused tricyclic pyrimido[4,5-d]thiazolo[3,4-
a]pyrimidine 4. Oxidation of any of these three gave back
1.6-10 The IR spectra of the normal and iso materials in
solution (chloroform) were reported to be essentially
identical;7 they matched that of Bonvicino and Hennessy.5
The stereochemistry about the several chiral centers in
3 and 4 was never published.
mediately invalidated the originally proposed achiral
structure 2 but did not eliminate the possibility that
structural changes might have taken place in the solid
over the many elapsed years.
The following work provides a resolution to the old
controversy concerning the structures of dihydrothiamin,
nicely outlined in detail in Beilstein.11 X-ray crystal
structures were obtained for 3 with its cis stereochem-
istry and 4 with cis-cis geometry. Kinetic studies making
use of hydride reducing agents in the presence of dihydro
3 and 4 gave tetrahydrothiamin 55,12 in aqueous buffers
and provided information to show that 3 is the kinetic
product of the hydride reduction of 1 while 4 is the
thermodynamic product. While chiral 3 is homogenous
in solution, 4 with three stereocenters, one of which is a
nitrogen invertomer, exists as a mixture of diastereomers
under some conditions. “Normal dihydrothiamine” (mp
150 °C), said to be trans 3, is cis 3 and the mysterious
and unstable “iso-dihydrothiamine” (mp 160 °C), claimed
to be cis 3, is likely to be only a polymorphic form of cis
3 with a higher melting point.
Recently, we obtained from Dr. Bonvicino the same
sample of dihydrothiamin that he prepared by the
hydride reduction route during his graduate studies, a
sample that generated the 1957 publication.5 A proton
NMR spectrum of this partially soluble old material in
CDCl3 indicated it was a mixture of substances, not
including 2. The obvious presence of diastereotopic
protons, indicative of one or more chiral centers, im-
X Abstract published in Advance ACS Abstracts, September 15, 1997.
(1) Zoltewicz, J . A.; Uray, G. Bioorg. Chem. 1994, 22, 1.
(2) Compounds reported in the literature are conveniently identified
by their melting points.
(3) Karrer, P.; Krishna, H. Helv. Chim. Acta 1950, 33, 555.
(4) Karrer, P.; Krishna, H. Helv. Chim. Acta 1952, 35, 459.
(5) Bonvicino, G. E.; Hennessy, D. J . J . Am. Chem. Soc. 1957, 79,
6325.
(6) Iwatsu, T. Yakugaku Zasshi 1955, 75, 677.
(7) Matsukawa, T.; Hirano, H.; Iwatsu, T.; Yurugi, S. J . Vitaminol.
1957, 3, 213.
(8) Matsukawa, T.; Hirano, H.; Iwatsu, T.; Yurugi, S. J . Vitaminol.
1957, 3, 218.
Resu lts a n d Discu ssion
Two isomers of dihydrothiamin were prepared in
accordance with the method of Bonvicino and Hennessy.5
The lower melting isomer (149-151 °C), which we now
show to have structure 3, was made by the mild reduction
of a neutralized aqueous-methanolic solution of thiamin
chloride hydrochloride using NaBH(OCH3)3 at -12 °C,
(9) Matsukawa, T.; Iwatsu, T. J . Vitaminol. 1955, 1, 305.
(10) Hirano, H.; Iwatsu, T.; Yurugi, S. Yakugaku Zasshi 1957, 77,
241.
(11) Beilsteins Handbuch der Organische Chemie, Supplement III/
IV, Vol. 27/ 3; Springer-Verlag: New York, 1983.
(12) Hirano, H. Yakugaku Zasshi 1958, 78, 1387.
S0022-3263(97)00782-2 CCC: $14.00 © 1997 American Chemical Society

Correct Structures of “Dihydrothiamine”
J . Org. Chem., Vol. 62, No. 20, 1997 6761
F igu r e 1. X-ray structure of perhydrofuro[2,3-d]thiazole 3.
protons at 2.41 and 2.09 ppm and therefore they were
assigned to CH2-CH2-O group. The signal at 2.41 ppm
also was coupled to the methinyl site at 3.78 ppm.
Distinguishing between the two remaining methylene
groups was accomplished by adding 1 to D2O, first
allowing the thiazole proton NdCHsS to exchange to
NdCD-S,17,18 prior to reduction with NaBH(OCH3)3 in
D2O-CH3OD to give N-CDH-S. The deuterium labeled
compound showed a loss of geminal coupling at this
partially exchanged methylene; the associated doublets
(J ) 8 Hz) at 3.78 and 3.62 ppm collapsed to apparent
singlets. Therefore, the more highly coupled geminal
methylene protons (14 Hz) at 3.92 and 3.62 ppm that did
not undergo isotope exchange correspond to the bridging
Py-CH 2-N unit. Moreover, the methinyl multiplet at
3.78 ppm now was largely isotope exchanged (>85%),
thereby causing the signal at 2.41 ppm which is associ-
ated with the proton of the CH2-CH2-O unit anti to it
to lose some of its multiplicity. Isotope exchange at the
methinyl site is an observation of some significance when
considering subsequently the mechanism of ring forma-
tion.
The presence of isotopes in the N-CDH-S group
resulted in an interesting long range four-bond spectral
change. This newly formed chiral center, present equally
in both configurations, caused the bridging methylene
protons (Py-CH2-N) to assume slightly different chemi-
cal shifts; the doublet for each proton now was split into
two additional lines with separations on the order of 2-4
Hz, giving an apparent doublet of doublets. However,
the magnitude of the new separation was different for
each of the methylene protons, and the magnitude also
changed with the field strength (300 vs 500 MHz),
showing that the new multiplicity was not due to spin
coupling but rather to differential shielding by the chiral
center.
and the higher melting isomer (173-175 °C), 4, was
prepared from the lower melting substance 3 by dissolv-
ing it in hot water, adding Na2CO3, and extracting it into
CHCl3. Both preparations attempted to reproduce the
original work.5
Cr ysta l Str u ctu r e of 3. Slow evaporation of a
solution of 3 in ethyl acetate gave crystals that led to
the X-ray generated structure in Figure 1 having a unit
cell containing both of the enantiomers; it is a racemic
compound and not a racemic mixture or conglomerate.13
The pair of cis-fused five-membered rings has the bicyclic
perhydrofuro[2,3-d]thiazole structure and form concave
dihedral angles of 110.8° (N3-C-O4) and 113.6° (S1-C-
C6). The thiazolidine methyl group and the methinyl
hydrogen atom are slightly staggered and both are
directed outside and above the envelope of the fused rings
while one hydrogen atom on each of the N-CH2-S and
O-CH2-CH2 methylene units is directed inside the
envelope. The two bridging Py-CH2-N hydrogen atoms
have different environments; one of these is located closer
to the thiazolidine methyl group than the other by 0.69
Å while the second hydrogen is nearer by 0.93 Å to the
pyrimidine proton, distances of significance in the sub-
sequent NOE experiments. One hydrogen atom of the
NH2 group is directed toward the N-atom of the thiazo-
lidine ring, forming a bent H-bond with a distance of 2.29
Å between the connected hydrogen and nitrogen sites.
The resultant six-membered ring is a common feature of
crystal structures.14
The molecular structure of 3 has some resemblance to
the crystal structure of tetracyclic 6 (mp 175 °C), formed
by simple deprotonation of 1 under mild conditions.
Again, the cis perhydrofuro[2,3-d]thiazole unit is present
among the four fused rings formed as a consequence of
the addition of the amino group of the free base of 1 to
the 2 position of the thiazolium ring to produce tricyclic
7 (dihydrothiochrome) followed by additional cyclization
to give the two fused five-membered rings.15,16
NMR An a lysis of 3. The structure of 3 in the solid
state and in CDCl3 solution is the same. Figure 2
contains chemical shift and stereochemical data and the
results of NOE investigations.
NOE experiments on a CDCl3 solution established the
environment about the Py-CH2-N grouping. On ir-
radiation of the aromatic proton signal at 7.96 ppm, only
one signal (3.62 ppm, 14 Hz) associated with one of the
two protons of the Py-CH2-N bridge was enhanced
(5.2%). But on irradiation of the thiazolidine methyl
signal at 1.56 ppm, the other proton of this methylene
bridge at 3.92 ppm was enhanced (8.7%), along with the
methinyl multiplet at 3.78 ppm (13%). The different
NOE values for the methylene protons indicate there is
likely to be a preferred conformation, one that may make
use of H-bonding between the amino group and the
tertiary nitrogen atom, as in the crystal.
A COSY analysis indicated that the closely spaced,
unsymmetrical, two proton multiplet at 4.03 and 4.05
ppm was coupled with both of the high field CH2-CH2-O
(13) Eliel, E. L.; Wilen, S. H. Stereochemistry of Organic Compounds;
J . Wiley & Sons: New York, 1994.
(14) Etter, M. C. Acc. Chem. Res. 1990, 23, 120.
(15) Takamizawa, A.; Hirai, K.; Ishiba, T.; Makino, I. Chem. Pharm.
Bull. 1971, 19, 759.
(16) Siro, M.; Nakai, H.; Makino, I.; Takamizawa, A. Acta Crystal-
logr. 1978, B34, 3424.
(17) Kemp, D. S.; O'Brien, J . T. J . Am. Chem. Soc. 1970, 92, 2554.
(18) Clarke, G. M.; Sykes, P. J . Chem. Soc. 1967, 1411.

6762 J . Org. Chem., Vol. 62, No. 20, 1997
Zoltewicz et al.
F igu r e 2. Proton NMR spectrum at 300 MHz of perhydrofuro[2,3-d]thiazole 3 in CDCl3. The pyrimidine proton at 7.96 and the
amino protons at 5.8 ppm are not included. Numbering corresponds to that in the X-ray structure. The arrows and percentages
designate NOE values.
A COSY analysis (DMSO-d6) allowed all the protons
to be assigned, Figure 4. The structure is consistent with
that found in the solid state. The multiplet at 1.69 ppm
due to one of the protons of the CH2-CH2-O methylene
group was coupled to the methinyl proton at 3.43 ppm;
the other CH2-CH2-O multiplet at 2.18 ppm was
coupled to both of the mutually coupled one proton
multiplets at 3.58 and 3.70 ppm, the latter two being
assigned to the CH2-CH2-O group. These CH2-O
protons are upfield relative to those in 3. The pair of
doublets (9 Hz) at 4.16 and 3.84 ppm were assigned to
the N-CH2-S methylene group on the basis of their
small coupling constant;19 they are shifted downfield
F igu r e 3. X-ray structure of pyrimido[4,5-d]thiazolo[3,4-
relative to those for the equivalent group in 3. The Py-
CH2-N bridge protons at 3.98 and 3.88 ppm are also
shifted downfield with a coupling constant, 16 Hz, similar
to the value of 14 Hz found for 3.
a]pyrimidine 4.
The APT 13C spectrum of 3 contained signals for four
carbons of the CH and CH3 types and eight of the C and
CH2 types, in agreement with the proposed structure. The
O-C-N carbon had a shift of 104.3 ppm making it
distinguishable from the equivalent atom (N-C-N)
found in isomer 4 at 84 ppm.
Cr ysta l Str u ctu r e of 4. A crystal grown from chlo-
roform has molecular structure 4, containing three fused
rings that do not include the hydroxyethyl group. The
central six-membered ring was formed by the addition
of the pyrimidyl amino group to the C-methyl site of the
reduced thiazolium ring; the tertiary nitrogen atom in
this ring resides outside the plane formed by the other
atoms in the ring, Figure 3. The three stereocenters
defined by atoms NCC give rise to the cis-cis diastere-
omer. The five-membered ring has a puckered envelope-
like conformation.
NOE experiments, Figure 4, allowed the stereochem-
istry to be assigned. Irradiation of the thiazolidine
methyl group at 1.45 ppm of a sample (D2O) highly
enriched in a single diastereomer by recrystallization
caused enhancement of five signals at 4.16 (11%), 3.98
(6.5%), 3.43 (3.9%), 2.18 (4.3%), and 1.69 ppm (3.9%)
corresponding to one syn proton in the methylene groups
of N-CH2-S and Py-CH2-N, as well as the methinyl
site and two protons in the CH2-CH2-O methylene unit,
respectively. Irradiation of the methinyl proton caused
a 5.5% enhancement of the Py-CH2-N proton syn to it
at 3.88 ppm. As expected, irradiation of the pyrimidyl
proton generated NOEs (3.1%) in both of the adjacent
methylene protons, in contrast with the observations for
3 where only a single site was enhanced.
The 13C APT spectrum again demonstrated four CH
and CH3 types and eight C and CH2 types, in agreement
with structure 4. Also, the N-C-N carbon was present
at 84 ppm, 20 ppm higher than the analogous carbon (N-
C-O) in 3. These two signals at 104 and 84 ppm stand
apart and therefore serve as convenient indicators of 3
and 4, respectively.
Con d ition s for th e Con ver sion of 3 to 4. The
proton spectra of 3 and 4 separately dissolved in acidic
phosphate buffer (pD ) 6.25) at room temperature
indicated that 3 after 22 h had converted to 4 with
deuterium exchange of both the thiazolidine methyl and
the methinyl protons; 4 experienced hydrogen-deute-
NMR An a lysis of 4 a n d Its Dia ster eom er s. The
proton spectrum of the compound 4 (mp 175 °C) dissolved
in alkaline water in order to suppress any pH-dependent
isomerization, Figure 4, shows sharp signals and is
consistent with the presence of just a single structure or
a set of rapidly equilibrating structures at ambient
temperatures, except for the presence of a small methyl
peak at 1.46 ppm believed to be due to a diastereomer.
The same compound when dissolved in unbuffered water
gives signals that are less well resolved, possibly indica-
tive of signal averaging due to isomerization, including
minor peaks among the multiplets. Spectra of 4, espe-
cially when dissolved in CDCl3, show a number of small
peaks in addition to the main spectrum, again suggesting
the possibility of the presence of a diastereomer.
(19) Leeper, F. J .; Lowe, P. N. Heterocycles 1983, 20, 65.

Correct Structures of “Dihydrothiamine”
J . Org. Chem., Vol. 62, No. 20, 1997 6763
F igu r e 4. Proton NMR spectrum at 300 MHz of pyrimido[4,5-d]thiazolo[3,4-a]pyrimidine 4 in D2O containing carbonate, not
including the aromatic proton at 7.94 ppm. The minor methyl peak at 1.46 ppm is due to a diastereomer. The X-ray structure
contains the same numbering pattern. The arrows and percentages designate NOE values.
Ta ble 1. Con cen tr a tion s a n d Ra te Con sta n ts for th e
rium exchange at the same two sites but otherwise
Red u ction of Dih yd r o 3 a n d 4 to Tetr a h yd r oth ia m in 5 by
remained unchanged structurally but several new peaks
did appear, most noticeably at 3.88 (d) and 4.20 (d) ppm,
again suggesting the presence of a stereoisomer.
Bor oh yd r id es in Aqu eou s Bu ffer s a t 25 °C
substrate [3] or [4], M [hydride], M pD
kobs (s-1
)
k3 or k2
3
3
3
4
4
4
4
4
1.1 × 10-2 6.4 × 10-2 a
1.1 × 10-2 2.9 × 10-2 a
9.77c 2.45 × 10-4 2.6 × 107 h
In alkaline carbonate (pD ) 9.77) at the same tem-
perature and after the same time 3 was only partially
converted to 4 while a separate sample of 4 was com-
pletely unchanged. Therefore, both isomers must form
a common set of intermediates to allow isotope exchange
to take place at the same sites in an acid dependent
process, likely by way of opening of the fused rings in
order to generate sites of unsaturation.
9.77c 1.36 × 10-4 2.3 × 107 h
2.1 × 10-2 5.6 × 10-2 a 10.71d 1.71 × 10-5 1.6 × 107 h
2.3 × 10-2 6.4 × 10-2 b 6.25e 5.07 × 10-4 90i
1.2 × 10-2 6.4 × 10-2 b 5.75f 2.19 × 10-4 123i
1.0 × 10-2 3.0 × 10-2 b 5.75f 1.68 × 10-4 94i
1.2 × 10-2 9.1 × 10-2 b 5.75f 2.12 × 10-4 119i
3.7 × 10-2 9.1 × 10-2 b 6.25g 5.56 × 10-4 99i
NaBH4. NaBH3CN. c Buffer 1. Buffer 2. e Buffer 3. f Buffer
a
b
d
4. Buffer 5. k3 M-2 s-1
.
k2 M-1 s-1
g
h
i
Kin etics of Tr a p p in g w ith Hyd r id e. Knowing that
further reduction of both dihydro isomers resulted in the
formation of tetrahydro 5,5,18-20 and that hydride is a
carbocation trapping agent,21 the reactivity of each isomer
toward hydride was examined in order to obtain informa-
tion about the conditions required to generate the
expected ring-opened intermediate.
hydride is some 108 times less reactive than the parent
hydride toward water.23 No attempt was made to control
the ionic strength of the medium, which generally was
high from the buffer, present in much higher concentra-
tion than the hydride in order to prevent increases in
the pD value. Pseudo-first-order rate constants, kobs
,
were obtained by following changes in the proton peak
intensities of the closely spaced thiazolidine methyl
groups in the starting material and products.
Tetrahydro 5 exists in cis and trans diastereomeric
forms that are distinguishable by proton NMR. The
thiazolidine methyl substituent appears as a pair of
doublets (J = ) 6 Hz) at 1.30 ppm (minor, trans19) and
1.05 ppm (major, cis19). Their formation served as a
convenient way to follow the reduction process.
On reduction of 3 with NaBH4 in aqueous solution, one
tetrahydro diastereomer was formed consistently and
preferentially in a 3:2 (cis/ trans) ratio. However, 4 did
not react with NaBH4 in aqueous carbonate but it did
undergo reduction with NaBH3CN in the more acidic
phthalic acid buffer (pD ) 4.2). Tetrahydrothiamin was
produced from 4 in a 9:11 (cis/ trans) ratio, the other
diastereomer now predominating slightly. Curiously,
NaBH4 reduction of 1 consistently gave the cis product
as the major diastereomer (2.5:1) as reported18,19 and in
our hands as well (3.1:1), in contrast to our observations
with this same hydride and 3 that gave a smaller ratio.
Table 1 summarizes the kinetic results using carbonate
buffer with NaBH4 for 3, and phosphate and acetate
buffers with NaBH3CN for 4. Attempted reduction of 3
with NaBH3CN at high pD resulted in only a meager
amount of 5 along with ring degradation products. The
data show that (1) the rate of reduction of 3 was first-
order in hydride ion concentration, but the rate for 4 was
independent of the hydride concentration, and (2) the
rates for both substrates were clearly first-order in
hydronium ion concentration. For 3, the apparent third-
order rate constant, k3, was calculated making use of eq
2. The third entry in Table 1 for 3 shows a k3 value 35%
smaller than the average of the first two, perhaps because
the initial ionic strength was twice as large or because
some 2 was present at the higher pD as indicated by a
consideration of the following mechanism.
Under our conditions the rate of reaction of both
hydrides with water is negligible,22,23 although some
bubbles were always evolved on mixing. The cyano
With substrate 4 the apparent second-order rate
constant, k2, was calculated using eq 3.
(20) Bau, R.; Hyun, Y.-S.; Lim, J .; Choi, H.; Yannopoulos, A.;
Hadjiliadis, N. Inorg. Chim. Acta 1988, 150, 107.
(21) Brown, H. C.; Bell, H. M. J . Org. Chem. 1962, 27, 1928.
(22) Schlesinger, H. I.; Brown, H. C.; Hoekstra, H. R.; Rapp, L. R.
J . Am. Chem. Soc. 1953, 75, 199.
(23) Lane, C. F. Synthesis 1975, 135.

6764 J . Org. Chem., Vol. 62, No. 20, 1997
k3 ) kobs/[D3O+][NaBH4]
k2 ) kobs/[D3O+]
Zoltewicz et al.
Sch em e 1
(2)
(3)
Mech a n ism of Rin g-Op en in g a n d Rin g-Closin g.
Hydrogen-deuterium isotope exchange took place at the
same methyl and methinyl sites of the thiazolidine ring
in 3 and 4, and hydride ion reduction of each gave the
same product 5, suggesting a common set of ring-opened
intermediates for the reduction of both of these to 5 in a
pH dependent process, Scheme 1. Thus, the mechanism
appears to involve protonation of the oxygen atom in 3
and a nitrogen atom in 4 to give iminium ion 8 resulting
from ring cleavage. The iminium ion then can undergo
a number of reactions: (a) isotope exchange at the
methinyl and the thiazolidine methyl sites by deproto-
nation of 8 to an enamine such as 2 in the case of isotope
exchange at the methinyl site, and this then is followed
by deuteronation to give back 8 now isotopically labeled
or (b) be trapped by hydride to give tetrahydro product
518 or (c) ring close in an intramolecular reaction to form
either 3 or 4 following the loss of a proton.
The cis/ trans ratio for 5 is not the same starting from
aromatic 1 (2.5-3.1 to 1) and from dihydro 3 (3/2) or 4
(9/11). The former two reactants make use of NaBH4 and
the latter NaBH3CN. Although the ratios are similar
starting from 3 and 4, these values are clearly smaller
than that beginning with 1,18 showing that more trans
tetrahydro material is generated from the two dihydro
isomers. The reason for the larger cis/ trans ratio for 5
from 1 than from 3 with the same hydride donor is not
apparent. However, under the conditions starting from
1, reducible tricyclic enamine 7 is present,24-27 providing
another possible route to reduced material, one in which
the order of addition of the two hydrogen atoms is likely
to be reversed.
Cyclization to form a five-membered ring, here 5-exo-
trig, generally is much faster than closure to a six-
membered ring, 6-exo-trig.28 Moreover, the pyrimidyl
amino group should undergo some rotation and rehy-
bridization on addition to give the six-membered central
ring and this change should contribute to the energy
barrier; in the microscopic reverse ring-cleavage process,
the transition state is expected to adopt an E1cb-like,
antiparallel geometry to generate 8. Hence, 3 is a kinetic
product and 4 is a thermodynamic product of cyclization.
Cyclization back to 3 is faster than hydride reduction and
so there is a kinetic dependence on the [BH4-], reduction
becoming the rate-limiting step in the forward direction
to yield 5. By contrast, cyclization back to 4 is so slow
that trapping by the less reactive BH3CN- is more rapid.
Thus, in the forward process involving the reduction of
4 to 5, ring-opening of protonated 4 becomes the rate-
limiting step. At low pH any conversion of 4 to 3 by way
of the open-chain intermediate 8 should be rapidly
reversible, giving back the lower energy 4.
atom of 4 that leaves as an amino group on ring-cleavage
rather than protonation at a nitrogen atom of the
pyrimidine ring. This protonation avoids the formation
of a high energy imino tautomer of the amino pyrimidine
on departure. However, the favored protonation process
requires the addition to 8 of the weakly basic30,31 and
weakly nucleophilic amino group in the forward direction.
But a similar addition of the amino group to the 2
position of the thiazolium ring to give tricyclic 7 has been
advanced.25-27
Ring opening to iminium ion 8 followed by ring closure
may change the stereochemistry about the thiazolidine
methyl group, giving cis-trans isomers about the CC
bond. Inversion of configuration at the stereogenic
tertiary nitrogen atom is expected to be facile, leading
to equilibrating cis and trans ring-fused structures.32
Likely, all four diastereomers are formed on the initial
conversion of 3 to 4 as suggested by NMR spectra, clearly
showing a set of minor peaks that are consistent with
the presence of another diastereomer. Recrystallization
enriches that set of diastereomers having the two chiral
carbon centers cis as found in the crystal structure.
“Nor m a l a n d Iso-d ih yd r oth ia m in e”. We were not
able to repeat the reported (no yield) conversion of the
lower melting (150 °C) “normal” to the higher melting
“iso” (160 °C) dihydro material by adding 10% NaOH to
an ethanolic solution of the 150 °C substance followed
by recrystallization from ethanol.9 But we were able to
generate a material (60%) that melted at 160 °C by
adding 10% aqueous hydroxide to a methanolic solution
of the lower melting substance resulting in the rapid
formation of a precipitate. Recrystallization avoided the
use of a hydroxylic solvent that might be conducive to a
rearrangement. This solid on addition to the 150 °C
substance gave a mixed melting range between that of
the lower and the higher melting (160 °C) compounds as
originally reported.6,9 But the NMR spectra of the two
materials separately dissolved in CDCl3 were identical.
All attempts to grow a crystal for X-ray analysis caused
reversion of the higher back to the lower melting mate-
rial.
Con clu sion s a n d Con sid er a tion of th e Old Lit-
er a tu r e Bonvicino and Hennessy misidentified both
isomers of dihydrothiamin, attributing structure 2 to the
151 °C isomer (now known to be 3) and the structure 3
to the material melting at 175 °C that we now know as
4.
The J apanese scientists correctly identified the 175 °C
material as 4 but misidentified the structure of the
materials melting at 150 °C and 160 °C, considering them
to be the trans and cis diastereomers of 3, respectively.
We believe both of these latter two materials to be
polymorphic forms of cis 3, one as the racemic compound
There is some question about where to locate the site
of protonation in 4. In the case of 1 protonation is known
to take place at the pyrimidyl ring nitrogen atom shown
in 1.29 But we favor protonation at the amino nitrogen
(24) Maier, G. D.; Metzler, D. E. J . Am. Chem. Soc. 1957, 79, 4387.
(25) Chahine, J . M. E. H.; Dubois, J . E. J . Chem. Soc., Perkin Trans.
2 1989, 25.
(26) Chahine, J . M. E. H. J . Chem. Soc., Perkin Trans. 2 1990, 1045.
(27) Washabaugh, M. W.; Yang, C. C.; Hollenbach, A. D.; Chen, P.
Bioorg. Chem. 1993, 21, 170.
(29) Cain, A. H.; Sullivan, G. R.; Roberts, J . D. J . Am. Chem. Soc.
1977, 99, 6423.
(30) Schellenberger, A. Angew. Chem., Int. Ed. Engl. 1967, 6, 1024.
(31) Schellenberger, A. Chem. Ber. 1990, 123, 1489.
(32) Crabb, T. A.; Katritzky, A. R. Adv. Heterocycl. Chem. 1984, 36,
1.
(28) Kirby, A. J . Adv. Phys. Org. Chem. 1980, 17, 183.

Correct Structures of “Dihydrothiamine”
J . Org. Chem., Vol. 62, No. 20, 1997 6765
(D2O): 7.94 (1H, s), 4.16 (1H, d J ) 9 Hz), 3.98 (1H, d J ) 16
Hz), 3.88 (1H, d, J ) 16 Hz), 3.84 (1H, d, J ) 9 Hz), 3.72 (1H,
m), 3.62 (1H, m), 3.43 (1H, d, d, J ) 3 and 12 Hz), 2.38 (3H,
s), 2.18 (1H, m), 1.69 (1H, m), 1.45 (3H, s). 13C NMR (D2O):
170.7, 161.6, 156.2, 111.4, 84.1, 65.4, 57.5, 55.7, 48.6, 38.0, 28.4,
25.5. Anal. Calcd for C12H18N4OS: C, 54.11; H, 6.81; N, 21.03.
Found: C, 53.84; H, 6.89; N, 20.80.
(mp 150 °C) and other as the metastable racemic con-
glomerate (mp 160 °C).13
The correct structure of “dihydrothiamine” as ena-
mine 2 generally has been accepted. Therefore, those
reports33-37 claiming 2 need to be reevaluated.
Gen er a l P r oced u r e t o Ob ser ve H -D Isot op e E x-
ch a n ge. Two samples of 3 and 4 (4.0 mg) in four NMR tubes
were dissolved in 0.1 mL of CD3OD. Buffers (1 and 3) below
were added. After 22 h at room temperature their 1H NMR
spectra were recorded. Observations are reported in the text.
Exp er im en ta l Section
Gen er a l. Thiamin chloride hydrochloride, NaBH(OCH3)3,
NaBH4, NaBH3CN, CD3CO2D, and CD3CO2Na were com-
mercially available and were used without further purification.
NMR data were collected on either a 300 or a 500 MHz
instrument.
Tetr a h yd r oth ia m in 5.5,18 The reductions to tetrahydro 5
were accomplished by dissolving 50 mg (0.19 mmol) of 3 or 4
in 4 mL of MeOH and adding borax buffer and a molar excess
of NaBH4 along with 3 or phthalate buffer and excess
NaBH3CN with 4. The reactions were monitored by TLC on
silica gel. Tetrahydrothiamin was isolated by extracting the
aqueous mixtures with chloroform, then drying over K2CO3
and evaporating to afford the white product consisting of a
mixture of diastereomers, mp from 3 122-127 °C and from 4
120-125 °C (lit.18 129-131 °C for the mixture and 145 °C for
purified material5). The average yield for several experiments
was 75%. Anal. Calcd for C12H20N4OS.1/3H2O (unrecrystal-
lized raw product) C, 52.53; H, 7.59; N, 20.42. Found: C,
52.27; H, 7.93; N, 20.27.
The original preparation of 5 from 1 with NaBH4 under
alkaline conditions18 was repeated at 0° C and at room
temperature. The unrecrystallized product had a cis/ trans
ratio of 3.1:1 (lit.18 2.5:1). Under these conditions with the
intermittent addition of the hydride, the solution had a yellow
color indicative of the presence of 7 and its bicyclic sulfide ring-
opened derivative.24 Attempts to reduce 1 with NaBH3CN in
water resulted in the precipitation of a slightly soluble
borohydride salt.
Dih yd r oth ia m in e P r ep a r ed by Dr . G. Bon vicin o in th e
1950’s. The needles of our 40-year old sample had largely
changed to powder, the combustion analysis was essentially
correct for the original empirical formula for dihydrothiamin,
and the melting point, curiously, was virtually unchanged
(150-155 °C). Mixed melting of equal amounts of the old
sample and 3 gave a substantial depression (135-142 °C) but
mixed melting with 4 gave no depression (151-156 °C). The
melting range of an equimolar mixture of authentic 3 and 4
was 135-142 °C. Partial extraction into CDCl3, leaving some
unidentified yellow residue, showed the presence of dissolved
4 by proton NMR. The old material now consisted of some 3,
mostly 4, and possibly a third substance that was not identi-
fied. It is likely that 4 originated in the solid state over the
years because the sample when added to D2O gave the
spectrum of 4 without deuterium incorporation at the methinyl
site, indicating that protonation of enamine 2 and cyclization
had taken place prior to dissolution of the old solid in D2O.
Anal. Calcd for C12H18N4OS: C, 54.11; H, 6.81; N, 21.03.
Found: C, 53.62; H, 6.81; N, 20.90.
Dih yd r oth ia m in 3 or cis-[3-(2′-Meth yl-4′-a m in o-5′-p y-
r i m i d y l )m e t h y l ]-3 a -m e t h y l p e r h y d r o fu r o [2 ,3 -d ]-
[1,3]th ia zole. This procedure follows that published in 1957.5
Thiamin chloride hydrochloride (5.00 g, 14.8 mmol), dissolved
in water (15 mL) and cooled to 0 °C, was treated with 1 N
NaOH (15 mL, 15 mmol) and MeOH (20 mL). After cooling
to -12 °C, NaBH(OCH3)3 (2.40 g, 19.0 mmol) was added in
small portions over 20 min. The mixture was warmed to room
temperature, and a heavy white precipitate formed after
several hours. The solid was recrystallized from EtOAc and
hexanes to afford 2.37 g (8.90 mmol, 60.1% yield, mp 149-
151 °C). 1H NMR (CDCl3): 7.96 (1H, s), 5.79 (2H, s, NH2),
4.03 (2H, dd J ) 4 and 9 Hz), 3.92 (1H, d J ) 14 Hz), 3.78
(1H, dd, J ) 2 and 6 Hz), 3.77 (1H, d, J ) 8 Hz), 3.62 (1H, d,
J ) 14 Hz), 3.62 (1H, d, J ) 8 Hz), 2.48 (3H, s), 2.41 (1H, m),
2.09 (1H, dtd, J ) 1, 4, and 12 Hz), 1.56 (3H, s). 13C NMR
(CDCl3): 167.1, 162.6, 154.7, 110.2, 104.3, 68.5, 55.0, 54.5, 47.7,
35.4, 25.5, 23.1. Anal. Calcd for C12H18N4OS: C, 54.11; H,
6.81; N, 21.03. Found: C, 54.38; H, 6.71; N, 20.95.
Deu ter a ted Dih yd r oth ia m in 3. The method is substan-
tially the same as that above except that the solvent contained
OD in place of OH. Na2CO3 was used to make an alkaline
solution and CH3OD was the cosolvent. Product (0.86 mmol,
57% yield, mp 147-149 °C) had the same proton NMR (CDCl3)
as 3 except that the doublets at 3.77 and 3.62 ppm had
collapsed to singlets, and the methinyl site at 3.78 ppm was
>85% exchanged.
Bu ffer s Used in Kin etic Ru n s w ith Hyd r id e in D2O.
Buffer (1): pD 9.77; carbonate 0.25 M, bicarbonate 0.25 M;
(2): pD 10.71; carbonate 0.625 M, bicarbonate 0.125 M; (3):
pD 6.25 deuterophosphate 0.050 M, dideuterophosphate 0.10
M; (4): pD 5.75; acetate 0.75 M, acetic acid 0.15 M; (5): pD;
6.25 acetate 0.75 M, acetic acid 0.050 M.
Gen er a l P r oced u r e for th e Kin etics of Red u ction of 3
a n d 4 w ith a Bor oh yd r id e a t 500 MHz. To 3 in 0.1 mL of
CD3OD in an NMR tube was quickly added 0.6 mL of a freshly
prepared D2O solution of carbonate- buffered NaBH4, prepared
by adding the solid hydride to the alkaline solution in order
to minimize decomposition. Timing began upon mixing. The
heights of the thiazolidine methyl peak in the starting material
and the two methyl diastereomeric peaks in reduced product
5 were measured at 25 °C. (Peak heights proved to be more
accurate than peak areas due to the proximity of other signals.)
The logarithm of the mole fraction of 3, ([3]/([3] + [5]), versus
time was subjected to a least squares analysis to give the rate
constant. Plots were constructed for at least two half-lives.
With 4, phosphate and acetate buffers and NaBH3CN were
employed. The same kinetic treatment was applied.
“Iso-Dih yd r oth ia m in e”. When 50 mg (0.19 mmol) of
dihydrothiamin 3 (mp 149-151 °C) in MeOH (3 mL) was
treated with 10% NaOH (5 mL) and stirred for a few minutes,
a white precipitate formed. The solid was collected, carefully
washed with cold water to remove excess alkali, dried, and
recrystallized from EtOAc and hexanes to afford 22 mg (0.083
mmol, 43% yield, mp 156-158 °C) of product. 1H, NMR
(CDCl3): δ 7.96 (1H, s), 5.79 (2H, s), 4.03 (2H, dd J ) 4 and 9
Hz), 3.92 (1H, d J ) 14 Hz), 3.78 (1H, dd, J ) 2 and 6 Hz),
3.77 (1H, d, J ) 8 Hz), 3.62 (1H, d, J ) 14 Hz), 3.62 (1H, d, J
) 8 Hz), 2.48 (3H, s), 2.41 (1H, m), 2.09 (1H, dtd, J ) 1, 4,
and 12 Hz), 1.56 (3H, s). Anal. Calcd for C12H18N4OS: C,
54.11; H, 6.81; N, 21.03. Found: C, 54.27; H, 6.71; N, 21.09.
When the solid was dissolved in a mixture of EtOH, hexanes,
and EtOAc and allowed to stand, the resultant crystals had a
lower melting point corresponding to the isomer (mp 149-151
°C) from which it was prepared.
Dih yd r oth ia m in 4 or cis,cis-9-(2-Hyd r oxyeth yl)-2,9a -
d im eth yl-5,9,9a ,10-tetr a h yd r o-7H-p yr im id o[4,5-d ]th ia z-
olo[3,4-a ]p yr im id in e. This procedure follows that published
in 1957.5 The white solid was recrystallized from EtOAc,
hexanes, and MeOH (22% yield, mp 173-175 °C). 1H NMR
(33) Hadjiliadis, N.; Markopoulos, J . J . Chem. Soc., Dalton Trans.
1981, 1635.
(34) Risinger, G. E.; Breaux, E. J .; Hsieh, H. H. J . Chem. Soc., Chem.
Commun. 1968, 841.
(35) Sorensen, H. C.; Ingraham, L. L. Biochim. Biophys. Acta 1970,
208, 343.
(36) Sorensen, H. C.; Ingraham, L. L. J . Heterocycl. Chem. 1970, 7,
711.
(37) Moorthy, P. N.; Haynon, E. J . Org. Chem. 1977, 42, 879.
X-r a y In for m a tion . Slow solvent evaporation of a solution
of 3 in EtOAc resulted in the formation of needles, mp 149-

6766 J . Org. Chem., Vol. 62, No. 20, 1997
Zoltewicz et al.
151 °C, that were submitted for X-ray analysis. CH3Cl was
used to grow a crystal of 4, mp 173-175 °C.
of refinement and were riding on the atoms to which they are
bonded. In each compound one of the methyl H atoms was
disordered, and they were refined in riding models in two parts
with 50% occupancy for each part. For compound 3, 173
parameters were refined in the final cycle of refinement using
2658 reflections with I > 2σ(I) to yield R1 and wR2 of 3.82%
and 8.99%, respectively. And for compound 4, 173 parameters
were refined in the final cycle of refinement using 2280
reflections with I > 2σ(I) to yield R1 and wR2 of 4.82% and
9.26%, respectively. Refinement was done using F2.
Ack n ow led gm en t. The National Science Foundation and
the University of Florida kindly provided funds for the
purchase of X-ray equipment.
Compound 3: C12H18N4OS, Mr ) 266.36, monoclinic, P2(1)/
n, a ) 6.3749(1) Å, b ) 23.1887(3) Å, c ) 9.724 Å, R ) 90°, â
) 106.94°, γ ) 90°, V ) 1375.04(3) Å3, Z ) 4, Dcalcd ) 1.287 g
cm-3, Mo KR (λ ) 0.71073 C), T ) 173 K. Compound 4:
C12H18N4OS, Mr ) 266.36, orthorhombic, Pbca, a ) 6.7405(1)
Å, b ) 16.9618(1) Å, c ) 23.0344(2) Å, R ) 90°, â ) 90°, γ )
90°, V ) 2633.55(5) Å3, Z ) 8, Dcalcd ) 1.344 g cm-3, Mo KR (λ
) 0.71073 Å), T ) 173 K.
Data for both compounds were collected at 173 K on a
Siemens CCD SMART PLATFORM equipped with a CCD area
detector and a graphite monochromator utilizing Mo KR
radiation (λ ) 0.71073 Å). Cell parameters were refined using
5969 (compound 3) and 6943 (compound 4) reflections from
each data set. A hemisphere of data (1381 frames) was
collected using the ω-scan method (0.3° frame width). The first
50 frames were remeasured at the end of data collection to
monitor instrument and crystal stability (maximum correction
on I was < 1%). Absorption corrections were applied based
on the psi scan using the entire data sets.
Su p p or tin g In for m a tion Ava ila ble: Crystallographic
parameters for 3 and 4 (15 pages). This material is contained
in libraries on microfiche, immediately follows this article in
the microfilm version of the journal, and can be ordered from
the ACS; see any current masthead page for ordering
information.
Both structures were solved by the Direct Methods38 in
SHELXTL5, and refined using full-matrix least squares on F2.
The non-H atoms were refined with anisotropic thermal
parameters. All of the H atoms were included in the final cycle
J O970782N
(38) Sheldrick, G. M. 1995, SHELXTL5. Siemens Analytical Instru-
mentation. Madison, WI.