Silicate complexes of sugars in aqueous solution

Article abstract of DOI:10.1021/ja031748v

Certain sugars react readily with basic silicic acid to form soluble 2/1 (sugar/silicic acid) silicate complexes. Failure of monohydroxy compounds to give soluble products under these conditions indicates that the sugar silicates are chelates: five-membered diolato rings. Only furanose forms react. Pyranose sugars are stable under these conditions. Because all glycosides fail to react with silicic acid under these conditions, reaction appears to involve the anomeric position (C1 in aldoses, C2 in ketoses), which has a more acidic hydroxy group. Reaction is completed only when the anomeric hydroxy group is cis to an adjacent hydroxy group. The appropriate furanose form must have sufficient natural abundance and solubility to provide an observable product, as measured by ²⁹Si and ¹³C NMR spectroscopy. These structural and practical constraints rationalize the successful reaction of the monosaccharides ribose, xylose, lyxose, talose, psicose, fructose, sorbose, and tagatose and the disaccharides lactulose, maltulose, and palatinose. Glucose, mannose, galactose, and sucrose, among others, failed to form complexes. This high selectivity for formation of sugar silicates may have ramifications in prebiotic chemistry.

Full text of DOI:10.1021/ja031748v

Published on Web 07/17/2004
Silicate Complexes of Sugars in Aqueous Solution
Joseph B. Lambert,*,† Gang Lu,† Stephanie R. Singer,† and Vera M. Kolb‡
Contribution from the Department of Chemistry, Northwestern UniVersity, 2145 Sheridan Road,
EVanston, Illinois 60208, and Department of Chemistry, UniVersity of Wisconsin-Parkside,
Kenosha, Wisconsin 53141
Received December 16, 2003; E-mail: jlambert@northwestern.edu
Abstract: Certain sugars react readily with basic silicic acid to form soluble 2/1 (sugar/silicic acid) silicate
complexes. Failure of monohydroxy compounds to give soluble products under these conditions indicates
that the sugar silicates are chelates: five-membered diolato rings. Only furanose forms react. Pyranose
sugars are stable under these conditions. Because all glycosides fail to react with silicic acid under these
conditions, reaction appears to involve the anomeric position (C1 in aldoses, C2 in ketoses), which has a
more acidic hydroxy group. Reaction is completed only when the anomeric hydroxy group is cis to an
adjacent hydroxy group. The appropriate furanose form must have sufficient natural abundance and solubility
to provide an observable product, as measured by 29Si and 13C NMR spectroscopy. These structural and
practical constraints rationalize the successful reaction of the monosaccharides ribose, xylose, lyxose, talose,
psicose, fructose, sorbose, and tagatose and the disaccharides lactulose, maltulose, and palatinose.
Glucose, mannose, galactose, and sucrose, among others, failed to form complexes. This high selectivity
for formation of sugar silicates may have ramifications in prebiotic chemistry.
Introduction
NMR experiments. In these strongly basic solutions, they found
29Si signals that corresponded to pentacoordination and hexa-
Silicon is the second most abundant element in the earth’s
lithosphere, after oxygen, and is found largely in the form of
solid silicates and aluminosilicates. Nonetheless, and despite
being located just below carbon in the periodic table, silicon
has been found to have very little involvement in life pro-
cesses.1,2 Carbohydrates are the most abundant organic material
in the biosphere, largely as cellulose, so that there ought to be
considerable opportunity for silicates and carbohydrates to
interact. What biochemistry there is of silicon, however, is
related primarily to silica uptake by marine organisms and some
plants. Considerable progress has been made in understanding
proteins that control the biological production of silica nano-
structures.3 The critical connection between silicates and
carbohydrates, however, has not been fully established.
Important experiments have been performed on certain sugar
derivatives, such as the glycitols (polyhydroxy compounds, in
which the sugar aldehyde function is reduced to the alcohol)
and glyconic acids (in which the aldehyde function is oxidized
to the carboxylic acid group). Kinrade and co-workers found
that aqueous silicate solutions react with straight-chain poly-
hydroxy compounds (glycitols such as xylitol, threitol, and
sorbitol).4 Their evidence was based entirely on 13C and 29Si
coordination, depending on the pH. They observed several peaks
within the coordination ranges, indicating that several high-
coordination forms existed with sugars attached to silicon
presumably via oxygen chelates. The same group found NMR
evidence for similar complexes between sugars and glyconic
acids under less basic conditions.5 Klu¨fers and co-workers
suggested that the connection between the glycitols and silicic
acid takes the form of a five-membered diolato ring.6 They later
supported the suggestion with X-ray structures of the manni-
tolato, xylitolato, and threitolato complexes.7 These are 3/1
complexes with hexacoordinate silicon. The glycitol hydroxy
groups that are not covalently bonded to silicon provide
important hydrogen-bonding interactions. The authors observed
that the diolato ring forms with five-membered (furanoid) rather
than six-membered (pyranoid) rings. The necessary O-C-C-O
dihedral angle within the sugar derivative must be close to 0°
to form the silicon-diolato chelate ring, an angle more easily
achieved by furanoid than pyranoid rings.7 There has been a
report of the reaction of silicic acid with ribose.8
To date, all reported observations except for that with ribose
have been made on either the polyhydroxy compounds (glyci-
tols) or the monocarboxylic acids (glyconic acids).4-8 Far more
important would be the possible reactions of carbohydrates
† Northwestern University.
‡ University of Wisconsin-Parkside.
(1) Bond, R.; McAuliffe J. C. Curr. Chem. 2003, 56, 7-11.
(2) Brook, M. A. Silicon in Organic, Organometallic, and Polymer Chemistry;
Wiley: New York, 2000.
(3) Cha, J.; Shimizu, K.; Zhou, Y.; Christiansen, S. C.; Chmelka, B. F.; Stucky,
G. D.; Morse, D. E. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 361-365.
Kro¨ger, N.; Lorenz, S.; Brunner, E.; Sumper, M. Science 2002, 298, 584-
586.
(5) Kinrade, S. D.; Hamilton, R. J.; Schach, A. S.; Knight, C. T. B. J. Chem.
Soc., Dalton Trans. 2001, 961-963.
(6) Benner, K.; Klu¨fers, P.; Schuhmacher, J. Z. Inorg. Allg. Chem. 1999, 625,
541-543.
(7) Benner, K.; Klu¨fers, P.; Vogt, M. Angew. Chem., Int. Ed. 2003, 42, 1058-
1062.
(4) Kinrade, S. D.; Del Nin, J. W.; Schach, A. S.; Sloan, T. A.; Wilson, K. L.;
Knight, C. T. G. Science 1999, 285, 1542-1545.
(8) Kinrade, S. D.; Deguns, E. W.; Gillson, A.-M. E.; Knight, C. T. G. J. Chem.
Soc., Dalton Trans. 2003, 3713-3716.
9
10.1021/ja031748v CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 9611-9625
9611

A R T I C L E S
Lambert et al.
Scheme 1
Table 1. Percentage of Ring-Size and Anomeric Sugar Isomers
RP
âP
RF
âF
ribosea
21.5
60
36.5
70
14
27
38.0
64.9
16
30
42
22
2.5
93
71
b
58.5
35.5
63
28
77.5
43
62.0
34.2
81
64
29
6.5
2.5
<1
1.5
3.5
17
b
0.6
b
2.5
13.5
2
<1
0.5
5
arabinosea
xylosea
lyxosea
allosea
altrosea
13
glucosea
mannosea
gulosea
0.1
0.3
3
3.5
13
15
25
1
7.5
62.6
29.3
22.4
71.4
100
41
galactosea
talosea
16
39
6.5
psicosea
fructosea
sorbosea
tagatosea
erythrosec,d
lactulosee,f
maltulosee,f
palatinoseg
sucrose
24
65
2
18
4
2.5
25.5
7.6
12.1
28.6
b
b
b
b
b
b
61.5
64.0
b
b
39
themselves. Whereas glycitols and glyconic acids exist primarily
as open-chain compounds, carbohydrates exist predominantly
as rings with either five members (aldofuranoses, ketofuranoses)
or six members (aldopyranoses, ketopyranoses). We report
herein evidence for widespread reactions of aldoses and ketoses
with silicic acid.
turanoseh
<4
20
a Reference 10. b This form is impossible, inaccessible, or undetected.
c Snyder, J. R.; Johnston, E. R.; Serianni, A. S. J. Am. Chem. Soc. 1989,
111, 2681-2687. d The open-chain form also is present. e The keto form
also is present at equilibrium. f Pfeffer, P. E.; Hicks, K. B. Carbohydr. Res.
1982, 102, 11-22. g Jarrell, H. C.; Conway, T. F.; Moyna, P.; Smith, I. C.
P. Carbohydr. Res. 1979, 76, 45-57. h Angyal, S. J. AdV. Carbohydr. Chem.
Biochem. 1984, 42, 15-68.
Results
The experiment involved dissolution of the sugar in D2O and
addition of an aqueous sodium silicate solution of precisely
defined stoichiometry. The pH of these solutions was 11.7.9 We
examined all four aldopentoses (ribose, arabinose, xylose, and
lyxose), seven of the eight aldohexoses (allose, altrose, glucose,
mannose, gulose, galactose, and talose), all four ketohexoses
(psicose, fructose, sorbose, and tagatose), several disaccharides
(sucrose, lactulose, maltulose, palatinose, and turanose), 2-deoxy-
ribose, fucose (6-deoxygalactose), the 1-O-methylglycosides of
fructose, ribose, and xylose, and some nonsugar model com-
pounds. The purpose of this survey was to determine the
structural and stereochemical basis for reaction of sugars with
silicic acid.
Table 2. 29Si Resonance Positions (δ) for Solutions of Sugars
Dissolved in Highly Basic Silicic Acid
sodium silicate
D-ribose
-71.9, -80.0, -88.1, -97.0, -106.0 (broad)
-99.0, -99.4, -99.5
D-xylosea
-98.3, -98.5, -98.8
D-lyxosea
D-talose
D-psicose
D-fructose
L-sorbose
D-tagatose
lactulose
maltulose
-99.8, -100.1
-100.9 (br)
-100.0, -100.4, -139.8, -140.4, -141.0
-100.3, -100.6, -100.9, -101.2, -101.6
-98.7, -99.3, -99.6
-137.5, -138.3, -138.8, -139.1
-98.1, -99.0, -99.4
-100.3, -100.9, -101.2, -101.5
-100.8 (broad), 101.4 (broad)
-97.5, -98.3, -98.5
palatinose
1,4-anhydroerythritola
a These solutions also contained resonances from unreacted sodium
silicate.
Sugars usually are complex mixtures of five-membered and
six-membered rings, in which the C1 stereochemistry can be
either R or â. Scheme 1 illustrates the four forms for D-glucose,
which in general will be abbreviated RP, âP, RF, and âF. These
forms are in slow equilibrium with each other and with the open-
chain form in aqueous solution. The percentages have been
measured for neutral pH and are collected in Table 1.10 In all
cases but one, the open-chain form is present at less than 1%
and is ignored.
contains the 29Si resonances observed in reactions with various
sugars, as well as for control samples. Omissions from the table
imply that no signals were observed. By way of example, Figure
1a provides the 29Si spectrum of D-ribose silicate, Figure 2a
provides that of D-fructose silicate, Figure 3 provides that of
several disaccharide silicates, and Figure 4 provides that of 1,4-
anhydroerythritol silicate.
The sugar silicate solutions were characterized initially by
29Si NMR spectroscopy. Although previous workers used silicic
acid containing enriched 29Si, we used natural abundance
materials to examine a large number of substrates. As a result,
overnight runs were required for acquisition of data. When
reaction occurred to form a soluble product, one or more new
peaks were observed in the 29Si NMR spectrum. Table 2
The 1H spectra generally were broad and uninformative. The
13C spectra, however, were extremely useful in understanding
the stability of the solutions and the sites of coordination. For
most systems, they were recorded under three conditions: (1)
at ambient pH (simple dissolution in D2O) in the absence of
silicic acid, (2) at pH ca. 12 in the absence of silicic acid (in
D2O with the pH adjusted), and (3) dissolved in silicic acid at
pH 11.7. Comparison of the first two solutions provides the
effect of pH alone. Comparison of the last two solutions provides
(9) This solution provides the lowest pH (11.7) at which reactions occur
successfully with sugars. Although gelling occurs sometimes (but not
always, as with sucrose), high-resolution spectra of the materials in solution
still may be obtained. Solution normally used by other workers4-8 has
considerably higher basicity (pH 12.7-13.4).
(10) Collins, P. M.; Ferrier, R. J. Monosaccharides; Wiley: Chichester, 1995;
p 41. Lichtenhaler, F. W.; Ronninger, S. J. Chem. Soc., Perkin Trans. 2
1990, 1489-1497.
the effect of coordination with silicic acid alone. Individual 13
C
resonance positions are given in the Experimental Section for
solutions containing silicic acid. Figure 5 contains all three
spectra for D-ribose, Figure 6 contains those for D-talose, Figure
9
9612 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
Figure 1. (a) The 79 MHz 29Si NMR spectrum of D-ribose silicate. (b) The ESI-MS of D-ribose silicate.
7 contains those for D-fructose, and Figure 8 contains those for
D-palatinose.
O substitution), those of pentacoordinated (2) start around δ
-100, and those of hexacoordinated (3) start around δ -140.
The solutions that produced new, observable 29Si resonances
were considered to contain stable sugar silicates, so they also
were examined by electrospray mass spectrometry. Figure 1b
illustrates the spectrum for D-ribose, and Figure 2b illustrates
that for D-fructose.
Discussion
Table 2 shows that almost all of the sugar silicates fall into the
range for pentacoordinated silicate. Geometries of either the
trigonal bipyramid (2a) or the tetragonal pyramid (2b) are
implied. No resonances of tetracoordinated silicate were ob-
served, except for unreacted silicic acid in the cases of xylose
Coordination Number of Silicon. The coordination number
may be discerned from the resonance position in the 29Si
spectrum. For materials containing a single silicon atom bonded
only to oxygen, resonances of tetracoordinated Si (1) start
around δ -70 and continue to lower frequency (depending on
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9613

A R T I C L E S
Lambert et al.
Figure 2. (a) The 79 MHz 29Si NMR spectrum of D-fructose silicate. (b) The ESI-MS of D-fructose silicate.
and lyxose. The low-frequency resonances of the ketohexoses
psicose and tagatose suggest hexacoordination in these cases.
Oligomerization. Resonances of tetracoordinated silicates
also provide information about the number of additional silicons
bonded through oxygen to the resonating silicon. If there are
no other silicon atoms, the structure is monomeric silicic acid,
Si(OH)4 (or derivatives with carbon attached, SiOC), and the
resonance occurs in the previously quoted position, δ ca. -70.
This structure is termed Q0 and is illustrated by the peak at δ
-71.9 (Table 2) for sodium silicate. If one oxygen is bonded
to another Si (SiOSi), the structure is termed Q1 and the
resonance is shifted to lower frequency (more negative) by about
8 ppm, as illustrated by the peak at δ -80.0 for sodium silicate.
Progressive substitution of Si leads to Q2 [Si(OSi)2], Q3
[Si(OSi)3], and Q4 [Si(OSi)4] structures, illustrated respectively
by the peaks clustered at δ -88.1, -97.0, and -106.0 for
sodium silicate. Although there are several 29Si resonances for
each sugar silicate, they are found invariably in the narrow range
of δ -98.3 to -101.6 for pentacoordination. This range indicates
that all of the molecules lack additional silicon atoms. One or
more such silicon atoms would have produced multiple reso-
nances, presumably every 8 ppm or so. The substitution patterns
therefore can be designated as P0, according to the notation of
Kinrade et al.,4 by which is meant structures 2 and 3 in which
each O is attached to either H or C but not to Si. Thus, sodium
silicate is oligomeric, but the sugar silicates are not.
Stability in Base. Birchall11 has argued that silicic acid cannot
have an important role with sugars because the medium must
be highly basic and sugars are unstable in base. The stability
of SiOC linkages, however, was established for polyols4-7
although not for the parent sugars themselves, other than ribose.8
Consequently, we sought evidence to demonstrate that decom-
(11) Birchall, J. D. Chem. Soc. ReV. 1995, 24, 351-357.
9
9614 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
resulting 13C NMR spectra are presented in the middle of Figures
5-8. Various peaks broadened at pH 12, but upon neutralization
to return to the starting conditions, the original spectra were
obtained (bottom spectra) with no new resonances and all old
resonances restored. Thus, the sugars are stable at pH 12 for
the times involved in these experiments. No anomerization,
epimerization, ketonization, fragmentation, or other decomposi-
tion occurred.
It is worthwhile to examine the 13C spectrum of D-ribose in
detail, in anticipation of assessing the changes that occur in the
presence of silicic acid. The spectrum at neutral pH has already
been fully analyzed.12 At neutral pH (Figure 5, bottom),13
resonances for the anomeric carbon, for example, are found in
the region δ 95-105. The four resonances correspond from left
to right to the âF, RF, âP, and RP forms (see Scheme 1 and
Table 1). Other isolated peaks include those in the regions
around δ 85 (C4 of RF and âF) and δ 78 (C2 of âF). When the
pH was changed to ca. 12 (Figure 5, middle), all of the peaks
from the pyranose forms remained sharp and at essentially
unchanged positions. The peaks from the furanose forms,
however, were altered profoundly. The peaks from C1 and C2
nearly disappeared. Sometimes broad peaks could be seen. The
peaks for C3 (δ 72.79), C4 (δ 84.92), and C5 (δ 65.59) are
present as single rather than double peaks, presumably overlap-
ping for the R and â forms.
Figure 3. The 79 MHz 29Si NMR spectrum of the silicates of palatinose,
maltulose, lactulose, and D-fructose (from top to bottom).
position is not a factor in the study of sugars in silicic acid
solution. The primary reactions of aldoses in base are anomer-
ization and isomerization from the ring form to the open-chain
aldehyde (or ketone), followed by enolization and C2 epimer-
ization. A hydride shift can lead to ketoses from aldoses, and
reverse aldol condensations can result in chain cleavage. Even
C3 and C4 epimerization can occur by a sequence of reversible
reactions.
For each sugar, we performed a control experiment, whereby
the stability of the sugar was examined under conditions of
basicity similar to that of the experiment in silicic acid (pH
11.7). In these experiments, no silicic acid was present. The
These observations have been explained in terms of the
known rates for ring opening and closing.14 Furanose ring-
opening and -closing rates are much faster than pyranose rates.
Whereas the pyranose rates do not fall in the range to which
NMR is sensitive, furanose rates are sufficiently rapid to broaden
Figure 4. The 79 MHz 29Si NMR spectrum of 1,4-anhydroerythritol silicate.
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9615

A R T I C L E S
Lambert et al.
Figure 5. The 100 MHz 13C NMR spectra of D-ribose (top) as the silicate, (middle) as the free sugar at pH ca. 12, and (bottom) as the free sugar at pH ca.
7.
the peaks at high pH. Such broadening does not occur when
the anomeric carbon is substituted, preventing ring-chain
isomerism. Neutralization of the solution regenerates all of the
furanose peaks, unchanged. With some minor variations, this
phenomenon is repeated in the spectra of all of the sugars
studied, whereby some furanose peaks are broadened by the
ring-opening/closing process. These observations will be rel-
evant to our analysis of the spectra of sugar silicates, which are
measured at the same basic pH.
silicate at pH 12-12.5.15,16 Under our conditions, such materials
do react but form insoluble gels that give no 29Si resonances.
Our objectives were to characterize major, soluble products. The
reaction with aliphatic alcohols is further discussed in a later
section.
In contrast, catechol (1,2-dihydroxybenzene) readily formed
a soluble complex with silicic acid, with a 29Si resonance at δ
-144.3, as previously observed.15 Substituted catechols such
as 4-methylcatechol, pyrogallol (4), L-DOPA (5), and 3,3′,4,4′-
tetrahydroxybiphenyl also gave such resonances, whose posi-
tions (respectively δ -144.1, -142.3, -143.1, and -143.0) are
all indicative of hexacoordination (3).17 We substantiated this
Structural Factors. Among the controls we studied for
sugars were simple alcohols and phenols. In no case did a
monohydroxylic alcohol or phenol (e.g., methanol and phenol)
form a stable, soluble complex under our conditions. The result
usually was formation of insoluble gels or precipitates and never
a new, observable 29Si resonance (silicic acid peaks remained).
Aliphatic alcohols and polyols have been reported to form small
amounts of tetracoordinated species with tetrabutylammonium
(12) King-Morris, M. J.; Serianni, A. S. J. Am. Chem. Soc. 1985, 107, 3501-
3508. Bock, K.; Pedensen, C. AdV. Carbohydr. Chem. 1983, 41, 27-66.
(13) The solvent was D2O, but H2O also was present whenever silicic acid was
included. Protons also derived from the HO groups from the sugar and in
the control pH 12 solution from NaOH. Thus, the measured pH was a
combination of pH and pD (pD ) pH + 0.4) and should not be considered
exact. Nonetheless, the small differences are not significant.
conclusion by the observation of a parent peak for the catechol
complex at m/z 353 in the electrospray mass spectrum, corre-
(14) Goux, W. J. J. Am. Chem. Soc. 1985, 107, 4320-4327.
9
9616 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
Figure 6. The 100 MHz 13C NMR spectra of D-talose (top) as the silicate, (middle) as the free sugar at pH ca. 12, and (bottom) as the free sugar at pH ca.
7.
sponding to the doubly negatively charged 3/1 species (6)
described previously.18 Similar 3/1 structures were confirmed
proton more than the nominal structures. Consequently, these
numbers are one unit greater than the calculated figures for the
3/1 structures (respectively, m/z 352.0, 400.0, 613.1, and 676.1).
This is the first time mass spectrometry has been applied to
this subject.
Simple aliphatic diols such as ethylene glycol failed to give
stable, soluble complexes with silicic acid under our conditions.
Several groups succeeded in preparing such complexes only at
very high temperatures.19 Mixed aliphatic/aromatic systems such
as 2,6-bis(hydroxymethyl)-1,4-cresol (7) also failed to give
stable, soluble complexes under our conditions. The flexibility
of open-chain diols may be a deterrent.
Thus, both chelating and acidity factors seem to be operating.
The ability to form two C-O bonds in the diolato chelate
structure is clearly favorable in comparison with only a single
C-O bond. The failure of aliphatic 1,2-diols to form stable
silicates suggests that a pKa factor also may be important. The
pKa of catechol at 9.3 is typically at least 6 orders of magnitude
lower than those of aliphatic alcohols, which are in the range
15-19. At pH 12, phenols exist preponderately in the ionized
form, whereas alcohols are largely un-ionized. The ionized form
by ES-MS for pyrogallol (m/z 401.2), L-DOPA (614.2), and
3,3′,4,4′-tetrahydroxybiphenyl (676.9). This experiment gener-
ates singly charged species in which the parent ion carries one
(15) Kinrade, S.; Maa, K. J.; Schach, A. S.; Sloan, T. A.; Knight, C. T. G. J.
Chem. Soc., Dalton Trans. 1999, 3149-3150.
(16) Samadi-Maybodi, A.; Harris, R. K.; Azizi, S. N.; Kenwright, A. M. Magn.
Reson. Chem. 2001, 39, 443-446.
(17) Cella, J. A.; Cargioli, J. D.; Williams, E. A. J. Organomet. Chem. 1980,
186, 13-17.
(18) Boudin, A.; Cerveau, G.; Chuit, C.; Corriu, R. J. P.; Reye, C. Angew. Chem.,
Int. Ed. Engl. 1986, 25, 474-476. O¨ hman, L.-O.; Nordin, A.; Sedeh, I. F.;
Sjo¨berg, S. Acta Chem. Scand. 1991, 45, 335-341.
(19) Frye, C. L. J. Am. Chem. Soc. 1970, 92, 1205-1210. Kansal, P.; Laine, R.
M. J. Am. Ceram. Soc. 1994, 77, 875-882. Herreros, B.; Carr, S. W.,
Klinowski, J. Science 1994, 263, 1585-1587.
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9617

A R T I C L E S
Lambert et al.
Figure 7. The 100 MHz 13C NMR spectra of D-fructose (top) as the silicate, (middle) as the free sugar at pH ca. 12, and (bottom) as the free sugar at pH
ca. 7.
is more nucleophilic and hence is more likely to react with silicic
acid.
spectrum can provide detailed information concerning which
hydroxy groups are involved. The needed comparison is of the
spectra of the sugars in silicic acid (top spectra in Figures 5-8)
with those at pH ca. 11.7 in the absence of silicic acid (middle
spectra).
The angle of chelation bite also is important, as the anion of
2,4-pentanedione failed to form an acac complex with silicic
acid. It should be noted that resorcinol (1,3-dihydroxybenezene)
also failed. Because its chelation bite is essentially impossible,
resorcinol resembles a monophenol in reactivity. The strength
of the Si-O bond also is critical, as the nitrogen and sulfur
analogues of catechol (respectively, 1,2-phenylenediamine and
benzenedithiol) failed to form stable, soluble complexes with
silicic acid.
These experiments revealed a number of structural factors.
(1) The Si-O bond is a necessary component of complexation,
as analogous Si-N and Si-S bonds failed to form. (2) Chelation
is necessary, as mono alcohols and phenols failed. (3) The angle
of chelation bite is important, as acetylacetone chelation (flat
six-membered rings) failed whereas 1,2-dihydroxyphenyl che-
lation (five-membered rings) succeeded. (4) The acidity of the
hydroxy group is critical, as aromatic diols succeeded whereas
aliphatic diols failed. Exceptions are discussed in a later section.
Positions of Complexation. The heuristic approach of the
previous section led to the conclusion that 1,2-diols with
relatively acidic hydroxy groups react with silicic acid to form
five-membered ring chelates. The observation of new 29Si
resonances when sugars are dissolved in silicic acid indicated
that chelates apparently are formed. Examination of the 13C
The 13C resonance of C-OH should move in a reproducible
fashion on complexation with silicon to form the entity C-OSi.
The carbon attached to oxygen thus experiences a â effect of
the silicon atom. Its adjacent carbon (CC-OSi) experiences a
γ effect. Unfortunately, the electronic effects of silicon have
not been well characterized in the literature,20 and the effects
of a silicon bearing three oxygen substituents have not been
characterized at all. Replacement in simple hydrocarbons of a
CH bond with CSiMe3 causes an average â effect (CCSi) of
0.9 ppm (the positive sign indicates a shift to higher frequency)
and an average γ effect (CCCSi) of 1.4 ppm.20 The trimethylsilyl
group also introduces a trio of methyl groups that are γ to the
carbon atom being examined (CCSiMe3). The normal carbon
γ effect of -2.5 (multiplied by three) offsets the normal high
(20) Pouchert, C. J., Behnke, J., Eds. The Aldrich Library of 13C and 1H FT-
NMR Spectra; Aldrich Chemical Co.: Milwaukee, WI, 1992. Schraml, J.
Collect. Czech. Chem. Commun. 1979, 44, 854-865. Maciel, G. E.; Dorn,
H. C.; Green, R. L.; Kleschic, W.; Peterson, M. R.; Wahl, G. H. Org. Magn.
Reson. 1974, 6, 178-180. Schraml, J.; Chvalovsky, V.; Magi, M.; Lippmaa,
E. Collect. Czech. Chem. Commun. 1978, 43, 3179-3191. Chung, M. K.;
Orlova, G.; Goddard, J. D.; Schlaf, M.; Harris, R.; Beveridge, T. J.; White,
G.; Hallett, F. R. J. Am. Chem. Soc. 2002, 124, 10508-10518. Bock, K.;
Thogersen, H. Annu. Rep. NMR Spectrosc. 1982, 13, 1-57.
9
9618 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
Figure 8. The 100 MHz 13C NMR spectra of D-palatinose (top) as the silicate, (middle) as the free sugar at pH ca. 12, and (bottom) as the free sugar at pH
ca. 7.
Scheme 2
frequency shift, resulting in the low value of 0.9. Correction
for the carbon γ effects results in a value of 8.4 for the â effect
of silicon, in line with other â effects. The methyls also introduce
an offsetting δ effect on the γ effect of silicon, but the usually
negligible value of δ effects does not necessitate a correction.
There are three drawbacks to this calculation. (1) It involves
a large correction for the γ effect of carbon. (2) The pathway
does not contain the oxygen atom present in the sugars (COSi,
rather than CCSi for the â effect). (3) The sugars contain three
γ oxygens rather than γ methyls, which may impart an important
contribution. The second problem may be obviated by examining
systems with oxygen along the pathway, of which a number
exist.20 The â effect in five such systems was found to average
1.3 ppm, or 8.8 ppm after correction for the γ effects of methyl,
marginally larger than in the all-carbon system. The γ effect in
these systems was measured to be 0.5. Thus, oxygen along the
pathway plays no substantive role.
The assumptions in these calculations still are too uncertain
to allow confidence in the results. Therefore, we sought a system
in which we could measure the effects directly. There are few
systems in which the desired structural change (COH f
COSiO3) is present, and none in which the three γ oxygens
consist of two diolato oxygens and one adjacent hydroxyl
oxygen. 1,4-Anhydroethritol (cis-3,4-dihydroxytetrahydrofuran)
possesses the same ring structure as ribose but lacks 1-hydroxy
and 4-hydroxymethyl groups. It has been found to complex with
silicic acid.8 Scheme 2 illustrates the structural changes that
occur on complexation. Carbons 2 and 3 experience the â effect
of silicon and three γ effects of oxygen. Carbons 1 and 4
experience the γ effect of silicon and three δ effects of oxygen.
The 13C spectrum in aqueous solution contains two peaks at δ
70.0 and 71.4. The resonance at higher frequency presumably
is from C2 and C3 because of the electron-withdrawing effect
of oxygen. On complexation, several new peaks appear (Figure
9, upper spectrum). The two peaks at low frequency are from
unreacted material. There is a set of overlapping peaks at δ
75.4 and four peaks at δ 71.7-72.3. Thus, upon complexation,
all shifts caused by â/γ silicon and γ/δ oxygens are to high
frequency. We attribute the presence of multiple peaks to
stereochemical variants that are discussed in a later section. To
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9619

A R T I C L E S
Lambert et al.
Figure 9. (Upper) The 125 MHz 13C spectrum of 1,4-anhydroerythritol silicates. (Lower) The DEPT spectrum of the same material, in which CH peaks are
positive and CH2 peaks are negative.
assign the peaks in the complexed spectrum, we carried out a
DEPT experiment, given in the lower portion of Figure 9 (CH2
groups give negative peaks, and CH groups give positive peaks).
The DEPT experiment confirms that the higher frequency peak
of the starting material is from C2 and C3 and proves that the
higher frequency peaks in the silicate complex also are from
these carbons. This assignment is unambiguous and is the
reverse of that in a previous report based on 1D decoupling
experiments.8 The â effect of silicon in this context is found to
be 3.9 ppm and the γ effect 0.8, in line with the earlier models
but far more reliable. The fact that the â effect is larger than
the γ effect has important ramifications on the conclusions of
previous workers,8 altering their position of complexation.
We now can proceed with examination of the 13C spectra of
the sugars. According to the results just discussed, the position
at which complexation occurs will experience a â effect of about
4 ppm. Shifts of the more distant carbons will be less than a
ppm.
In the presence of silicic acid, the C1 resonances for the
pyranose forms of ribose (Figure 5, top) are found essentially
unchanged at δ 96.30 and 96.62 (originally, δ 96.14 and 96.45).
Consequently, no reaction has occurred at the site. A new cluster
of peaks is found at δ 100-101, which we attribute to the RF
isomer, because its position originally was at δ 98.73. Two broad
peaks occur in the region δ 106-108, which we attribute to
the âF isomer, whose original position was δ 103.38. These
shifts (â effects of 2-5 ppm) indicate coordination of C1 with
silicon in the furanose isomer only. Multiple C1 13C peaks
correspond to the multiple peaks observed in the 29Si spectrum.
We discuss the structural significance of the multiplicity in a
later section. Suffice it to say that there are multiple forms of
the silicate complexes present for each isomer. We can conclude
that one site of attachment of the sugar with Si is at C1 of the
furanose forms.
It is likely that reaction at C1 is related to its higher acidity.
Whereas the pKa of typical alcohol hydrogens is in the region
15-19 (which includes methanol, ethanol, and tert-butyl
alcohol), and the presence of an adjacent hydroxy group lowers
the pKa by about one unit, the pKa of the anomeric hydroxyl
group is usually about 12. In general, the anomeric hydroxyl is
the most acidic site for both pentoses and hexoses.21 Moreover,
the anomeric hydroxyls of furanoses are kinetically more acidic
than those of pyranoses.22 This difference may contribute to
the absence of complexation by all pyranoses (six members)
and the failure of some furanoses (five members) to complex.
The acidity of the anomeric hydroxy groups is reflected in the
13C chemical shifts of the anomeric carbons. For example, in
fructose the R and â furanose forms produce C2 chemical shifts
(21) Christensen, J. J.; Rytting, J. H.; Izatt, R. M. J. Chem. Soc. B 1970, 1646-
1648.
(22) Gillet, B.; Nicole, D. J.; Delpuech, J. J. J. Chem. Soc., Perkin Trans. 2
1981, 1329-1335.
9
9620 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
Scheme 3
at δ 103.9 and 106.9, whereas the â pyranose form produces a
chemical shift at 100.5 (C2 is anomeric in ketoses). The
significant difference is reproduced in many other sugars and
is related to the differences in acidity. The pKa values for ribose,
lyxose, and xylose, which successfully formed complexes with
silicic acid, are close to 12.1.23 Arabinose and 2-deoxyribose,
which did not form complexes, are less acidic, with respective
pKa’s of 12.34 and 12.61.23 Thus, it is possible that in some
cases lower acidity of the anomeric hydroxyl is a deterrent to
reaction with silicic acid, both for furanoses and pyranoses.
Other factors, such as stereochemistry, are assessed in a later
section.
The pyranose peaks for all of the other carbon atoms of ribose
also remained essentially unchanged. The resonance positions
at pH ca. 7 for furanose C2 are δ 77.60 for âF and δ 73.36 for
RF. The former peak disappears on introduction of silicic acid
and is replaced by a cluster of peaks at δ 80-81 (a â effect of
2.4-3.4 ppm), which we attribute to the complexed âF isomer.
New peaks in the δ 74-75 region are likely to come from the
complexed RF isomer. Changes in the C2 resonance position
thus suggest that complexation also is occurring at C2. The
resonances for the C3 and C5 carbons are relatively unchanged,
although peak overlap, particularly for C3, makes conclusions
uncertain. The peaks at δ 63.85 (C5, RF) and 65.02 (C5, âF)
are replaced by peaks at δ 62.9 and 64.6, respectively, again as
clusters. There is no hydroxy group on C4 in ribose. The
approximate constancy of the C3 and C5 resonance positions
suggests that these hydroxy groups are not involved in com-
plexation.
The conclusion based on the movement of 13C peaks is that
chelation occurs at C1 and C2 in ribose. Similar conclusions
are obtained by analysis of the remaining aldose spectra. For
xylose and talose (Figure 6), the pyranose resonances remain
unchanged, the C1 and C2 peaks shift, and the other peaks
remain unchanged. Talose has relatively high proportions of
furanose rings. Although the C1 and C2 resonances clearly
move, it is possible in this case that the C6 resonances also
move, indicating a possible alternative site of coordination.
In the case of fructose, a ketose, it is the C2 position that is
anomeric, and the C2 resonances clearly move on treatment with
silicic acid (Figure 7), as do the C1 and C3 resonances. The
C4 and C6 resonances appear to remain unchanged (C5 does
not bear a hydroxy group). We conclude from the 13C spectra
that silicon attachment occurs in ketoses at the anomeric C2
and the two adjacent carbons C1 and C3.
Previous workers4-7 observed complexation of glycitols and
glyconic acids, which have no analogue of the anomeric
hydroxyl. These materials exist in conformationally flexible
open-chain forms. Kinrade et al.4 suggested and Benner et al.6,7
confirmed through X-ray crystallography the important hydrogen-
bonding role of nonchelating hydroxy groups in complexes of
the glycitols (polyhydroxy derivatives). The rings in the sugars
generally make the nonchelating hydroxy groups sterically
unavailable for this supporting role. Sugars thus have more
stringent structural requirements than glycitols for complexation.
Blocking the Anomeric Carbon. Analysis of 13C chemical
shifts indicates that C1 and C2 appear to be the points of
attachment for chelation of aldoses. What then happens if one
of these positions is blocked? One simple test is to convert the
reducing sugars to their methyl glycoside derivatives. These
nonreducing sugars prohibit interconversion of the anomeric
center but also prevent any new reaction from occurring there.
We prepared the 1-O-methyl derivatives of ribose and xylose
and the 2-O-methyl derivative of fructose (ROH f ROCH3),
all sugars that gave silicates according to the appearance of new
29Si peaks and changes in the 13C chemical shifts. The three
glycosides showed no reaction on dissolution in basic silicic
acid according to these NMR criteria. Thus, it appears that it is
necessary that the anomeric carbon have a free hydroxy group
for reaction to occur with silicic acid.
Nucleosides provide another useful test of the effect of C1
substitution. We examined guanosine and adenosine under the
usual conditions and found no reaction. No new 29Si peaks
appeared, and the 13C peaks underwent no significant shifts
relative to each other. In these molecules, the C1 carbon is
blocked by the presence of the base, similar to the O-methyl
glycosides above. Previous authors reported that these molecules
underwent reaction.8 The 29Si spectra in their Figure 4, however,
contain almost entirely peaks only from silicic acid, which
appear (Table 2) as multiplets at δ ca. -100 (dotted line in
their figure), the expected location for the product peaks. These
reactions were carried out at higher pH than we used, and it is
possible that decomposition occurred.
One subtle fact needs to be emphasized. Ribose and other
reacting aldoses produce clearly observable C1 resonances in
silicic acid (top spectra in Figures 5-7). As already remarked,
at pH ca. 12 in the absence of silicic acid, these sugars exhibit
severely broadened C1 furanose resonances, or none at all,
because of rapid ring-chain isomerization (middle spectra in
Figures 5-7). The presence of silicon at C1 prevents this
process, as with glycosides and nucleosides, so that C1
resonances are sharper (top spectra in Figures 5-7). Thus, the
observation of well-defined C1 resonances in silicic acid at pH
11.7 is ipso facto evidence for reaction at C1, although reaction
elsewhere is not excluded.
Another approach to probing the location of bonding between
silicon and specific sugar positions is to examine disaccharides,
which have specific connections between the two rings (Scheme
3; only the â forms are shown). Sucrose, for example, is
composed of glucose and fructose rings that are connected
through their respective anomeric carbons. We found that
(23) Bethell, G. S.; Ferrier, R. J. J. Carbohydr. Res. 1973, 31, 69-80.
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9621

A R T I C L E S
Lambert et al.
Scheme 4
sucrose failed to react with silicic acid, consistent with a sugar
lacking an open hydroxy group on an anomeric carbon. The
disaccharide lactulose (see Scheme 3) is composed of a galactose
ring connected via its anomeric carbon with C4 of a fructose
ring. The galactose anomeric hydroxyl is blocked by the fructose
ring, but the fructose anomeric hydroxyl is free. As Figure 3
shows, lactulose gives four peaks in the 29Si spectrum with a
pattern very similar to that of fructose, although shifted to higher
frequency. The 4 position of the fructose ring is blocked by
attachment to galactose and could not be used for reaction with
silicic acid. Therefore, this position is not essential for com-
plexation. Palatinose is composed of a glucose ring attached
via its anomeric carbon with C6 of a fructose ring. It too reacts
with silicic acid (Figures 3 and 8). These observations with
lactulose and palatinose confirm that reaction at C4 and C6 of
fructose is not important (C5 lacks a hydroxy group), focusing
attention on C1 and C3 (C2 is the anomeric carbon and is
necessary for reaction). Maltulose (not shown in Scheme 3) is
composed of a glucose ring attached via its anomeric carbon to
C4 of fructose, and it too reacts with silicic acid (Figure 3).
Finally, turanose (Scheme 3) fails to react or decomposes on
exposure to silicic acid, as no new 29Si resonances appear. It
contains a glucose ring connected via its anomeric carbon to
C3 of a fructose ring. This latter position must then be unblocked
for reaction to occur with silicic acid.
Experiments with glycosides and nucleosides thus confirm
that the anomeric carbon (C1 of aldoses, C2 of ketoses) must
have an available hydroxy group for reaction to occur with silicic
acid. Blocking C4 or C6 of fructose does not prevent reaction,
indicating that these positions are not involved. Blocking C3
of fructose prevents reaction, so this position adjacent to the
anomeric carbon also is necessary for reaction. The role of C1
in ketoses is not clear.
Complexation Ratio from Mass Spectrometry. The 13C
spectra gave strong evidence for the location on the sugar for
complexation with silicic acid, but they provided no information
concerning the stoichiometry. The 29Si spectra indicated (in most
cases) pentacoordination without oligomerization. We therefore
need to specify how many of the five coordination sites are
occupied by sugar oxygens. The most likely alternatives are
1/1 complexation, as in 8, and 2/1 complexation, as in 9. In
complexes, despite the weakness of the 29Si spectra. Fructose
produced the simple spectrum illustrated in Figure 2b, with a
strong parent peak at m/z 401.3 indicative again of a 2/1
complex. The ketohexose sorbose also gave a parent peak at
m/z 401.0 consistent with a 2/1 complex. We were unable to
obtain mass spectra of the other two ketohexoses, psicose and
tagatose, because of gelling of the solution. These were the two
cases that produced hexacoordinate species either in part or
entirely according to their 29Si spectra. The disaccharide
palatinose also gave a parent peak (m/z 724.2) indicative of a
2/1 complex. We conclude that, in the preponderance of cases,
2/1 complexes of the type 9 formed. Kinrade et al. have come
to this same conclusion for ribose by the difficult process of
comparing intensities of 13C and 29Si peaks. They preferred
structures, however, with higher levels of oligomerization at
this ratio (4/2 or 8/4, their Species 2 and 3). We have found no
evidence for either 4/2 or 8/4 structures in the mass spectra,
but rather only monomeric 2/1 complexes.
Role of the Diol Dihedral Angle. It has already been
established that all pyranose rings fail to react with silicic acid,
whereas many furanose rings react readily. There may be a small
acidity factor that favors the five-membered ring, but it probably
is more important that the HO-C-C-OH dihedral angle in
six-membered rings is ill-suited to complex with silicic acid.
The resulting diolato ring contains five members, fused to a
five-membered ring for furanoses or to a six-membered ring
for pyranoses. Cis, vicinal substituents in the five-membered
diolato ring have very small dihedral angles, approaching 0°.
The HO-C-C-OH group in the sugar becomes O-C-C-O
in the diolato ring, with necessarily a very small dihedral angle.
The HO-C-C-OH dihedral angle present in a six-membered
ring (ca. 60° for axial-equatorial or equatorial-equatorial
relationships) is much too large to work. The small HO-C-
C-OH dihedral angles already present in the furanose rings
(ca. 0° for cis relationships) are ideally suited for formation of
the nearly flat diolato ring. Even in furanose rings, the trans
relationship (ca. 120°) present in half the forms is unsuitable.
Thus, for every sugar, only one of the four isomers is suitable
for forming a complex with silicic acid (Scheme 4). If the natural
percentage of this one form is very small, it is possible that we
would not observe complex formation by 29Si, because we are
examining 29Si at natural abundance.
pentacoordination, 3/1 complexation is not possible unless the
third sugar molecule is monocoordinated. This arrangement is
highly improbable, given the failure of monoalcohols and even
monophenols to form soluble silicate complexes. A 3/1 complex,
of course, is possible for the hexacoordinated species, analogous
to 6.
We carried out electrospray mass spectrometric experiments
directly on the newly prepared silicic acid solutions. Figure 1b
shows the results for ribose. A strong peak at m/z 341.1 in a
very simple spectrum clearly indicates a 2/1 (sugar to silicic
acid) complex. The aldopentoses xylose (m/z 341.1) and lyxose
(m/z 341.1) also gave parent peaks clearly indicative of 2/1
These structural constraints explain many of our negative
results. For example, the favorable furanose form of glucose is
the âF (in which the 1 and 2 hydroxyls are cis), but it is absent
(Table 1). The same form also is absent for gulose and is present
at the level of only 0.3% for mannose. The favorable form is
present at 6.5% for ribose and at decent levels for all four
9
9622 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
ketohexoses, which must have contributed to the positive results
for these systems.
availability, the trans diol stereochemistry does not permit
complexation.
Substrate Availability. The extent of formation of complexes
between silicic acid and sugars, aliphatic alcohols and diols,
phenols, and catechols has been rationalized by four factors:
(1) chelate formation, (2) HO acidity, (3) diol stereochemistry,
and (4) substrate availability. The first three already have been
discussed thoroughly. They do not, however, explain several
observations, so that exceptions must be possible. Harris,16
Kinrade,15 and their respective co-workers found that methanol
reacts with an ammonium silicate to form CH3OSi(OH)3.
Methanol provides neither chelation nor heightened acidity.
Harris’s conditions, however, involved 45% v/v methanol. In
all cases, reaction was only partial, and the equilibrium constant
[Si(OH)4 + CH3OH h CH3OSi(OH)3 + H2O] was always on
the side of silicic acid (K ) 0.46-0.80). Thus, even with an
overwhelming excess of alcohol, conversion is low. The poor
properties of methanol are overcome by higher substrate
availability (the concentration is raised to 45%). Higher avail-
ability is possible because of the high solubility of methanol in
water. Thus, complexation can occur in the absence of one or
more of the three principal factors, compensated by the fourth
factor of higher product availability. Sugars and catechols,
however, still react more completely.
Substrate availability also is an important factor in the
formation of sugar silicates, as outlined in the previous section.
The one anomer with a furanose ring and a cis stereochemistry
(Scheme 1) must be present in sufficient amount to provide an
observable product. Thus, glucose and mannose have very small
amounts of the favorable anomer, and we observe no product
with these substrates. Only sugars with an abundance of that
anomer gave positive results. Xylose and lyxose provide an
interesting borderline behavior. Both of these substrates did
produce an observable 29Si result (Table 2). The peaks, however,
were weak, and in only these two cases (of all of the successful
cases listed in Table 2) we also observed resonances from
unreacted silicic acid. In this aspect, these two substrates
resemble methanol and 1,4-dihydroerythritol (10), whose solu-
tions also contained resonances of unreacted silicic acid. Xylose
and lyxose suffer from lack of sufficient available substrate.
Keep in mind that the sugar isomers do not interconvert over
the time scale of the experiments.
Insufficient substrate availability can inhibit complex forma-
tion, because of either low solubility (11 and 12) or low anomer
abundance (glucose, mannose, and, to a lesser extent, xylose
and lyxose). The high substrate availability of methanol and
1,4-anhydroerythritol (10) can compensate for the absence of
diol structure or heightened HO acidity. The existence of sugar
silicates from methanol and 10 implies that it is possible that
complexation can occur with compounds that do not include
the anomeric carbon, as was already observed also with
polyhydroxy compounds.4-7 For example, glycosides in which
the anomeric position is blocked might form a chelate at the 2
and 3 positions, given cis stereochemistry. Such complexation,
however, is less favorable and would have to be augmented by
higher substrate availability. When both 1,2 and 2,3 positions
offer synperiplanar arrangements and are available, complex-
ation will always favor 1,2 reaction in aldoses because of the
heightened acidity of the anomeric OH. Nonetheless, 2,3 reaction
is a viable option, although we have not observed it under our
conditions.
We have already discussed 1,4-anhydroerythitol (10) as a
model for sugars. Lacking an anomeric carbon, it still has a
five-membered ring, analogous to furanoses, and a pair of cis
hydroxy groups. This molecule forms complexes with silicic
acid, producing the 29Si spectrum in Figure 4, previously
observed by Kinrade et al.8 Unlike almost all of the sugars for
which we observed complexation, 10 is not converted entirely
to the complex. Like methanol, 10 reacts only partially, as seen
in the 13C spectra in Figure 9. The major difference between
10 and reducing sugars or catechols is the absence of a hydroxy
group with heightened acidity. As with methanol, a critical factor
is the high solubility of 10 in water. Thus, greater substrate
availability compensates for the poorer acidity. In addition, the
absence of substituents on the 2 and 5 carbons may provide
better steric accessibility of the 3 and 4 hydroxy groups. Even
with higher solubility and possibly better steric accessibility,
10 still reacts less fully with silicic acid than do catechol and
reducing sugars such as ribose.
Gel Formation. Spectral characterization as described thus
far was carried out immediately after mixing of the reagents,
so that even the long-term NMR experiments were completed
in a few hours. Most of the sugar silicates formed a white gel
after standing 10-12 h, and almost all did so after a month.
Only sucrose and the glycitols (inositols) were stable indefi-
nitely in silicic acid solution. Although sugars appear to be
incorporated into the gels, we have no structural information
on them. This study is concerned only with the soluble silicate
products.
As a test of the hypothesis of substrate availability, we
examined cis-1,2-dihydroxycyclopentane (11) and cis-2,3-
dihydroxynorbornane (12). These molecules resemble 10 by
having cis diols attached to a five-membered ring. Neither
substrate, however, formed an observable stable, soluble
complex with silicic acid. Although 11 and 12 are stereochemi-
cally able to form the ring, we believe that their low solubility
in water (poor substrate availability) prevents sufficient sugar
silicate to be formed and detected under our conditions. We
also examined the trans form of 10, 1,4-anhydrothreitol, and
failed to observe any sugar silicate. Despite high product
Analogues to Sugar Silicates. Whereas complexation of
sugars with silicates was unknown prior to this study, at least
two analogous structures are quite common, the isopropylidene
acetals and the borates (compare structures 13-15). The acetals
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9623

A R T I C L E S
Lambert et al.
have been found to form with almost all pentoses and hexoses.24
Even double acetals form in many cases. Acetal formation
occurs not only when the two hydroxy groups are cis to each
other, but even in some trans cases. The anomeric hydroxyl
plays no special role, and the acetals form with a much wider
variety of sugars, including glucose.
Scheme 5
The borate anion has long been known to form complexes
with sugars.25 The reaction is more general than the silicate
reaction, as it occurs under acidic as well as basic conditions,
and it occurs with most 1,2- and 1,3-diols, including glucose.
Furanoses, however, are more reactive, and the anomeric
hydroxyl plays an important role (the 13C spectra of furanose
borates and silicates are remarkably similar for a given sugar).26
Despite similarities, the differences are important, in particular
the high selectivity of the silicate reaction. Whereas acetal and
borate formation is quite general with sugars, silicate formation
is rather specific. We attribute the selectivity to the ability of
silicon to hypercoordinate. The maximum coordination for both
carbon and boron is four, as they reside in the second row of
the periodic table. Silicon, from the third row, can hypercoor-
dinate. The resulting complexes have 90° angles (see 8 and 9).
Angles in the borates and acetals are closer to tetrahedral. The
smaller angle required in the hypercoordinate sugar silicates
apparently imposes a stronger restriction on the acceptable HO-
C-C-OH dihedral angles (probably demanding angles closer
to 0°), so that many sugars fail to have the structural require-
ments.
Stereoisomers. The sugar silicates typically exhibit 3-5
resonances in the 29Si NMR spectra (Figures 1-4). The presence
of several peaks is mirrored in the 13C spectra, particularly in
the resonances of the anomeric carbons on dissolution in silicic
acid (top spectra in Figures 5-8), and in 1,4-anhydroerythritol
(Figure 9). The narrow range of the 29Si resonances, as already
noted, indicates that all of the species are monomeric in silicon.
The multiple peaks of the sugar silicates are quite different from
the multiplicity of silicic acid itself, which extends over a range
from δ -70 to -106 (Table 2) as the result of oligomerization
with the presence of multiple silicon atoms in the molecules.
We believe that the multiple peaks for the sugar silicates result
from stereoisomerism deriving from the 2/1 stoichiometry, as
depicted in Scheme 5. The mass spectral experiments indicated
that the sugar silicates contain two sugar molecules for each
silicic acid. We have depicted the result diagrammatically as
structure 9. This structure is redrawn in Scheme 5 as structure
A, in which the five-membered diolato rings are shown. The
two carbons of the diolato rings are part of the sugar ring, not
shown. Structures 9 and A incorporate both rings through one
axial (or apical) and one equatorial bond. Structure B depicts
an alternative 2/1 complex in which one diolato ring is axial/
equatorial but the other is diequatorial (diaxial, of course, is
impossible). We do not, however, believe that such a species
contributes to the spectral mutliplicities. It requires an angle of
approximately 120° within the five-membered ring, which would
be prohibitively strained. The arrangement of A is generally
more stable.
Kinrade et al.8 Structures C-E illustrate different stereoisomeric
modes to place the sugar bonds. In structure C, both connections
to C1 of the sugar occur at the axial positions to Si, whereas
both connections to C2 occur at the equatorial positions. In
structure D, the roles are reversed, and both C1 connections
are equatorial and both C2 connections are axial. Finally, in
structure E, there is one axial connection to C1 and one axial
connection to C2. Thus, there are three distinct stereo modes
for constructing a 2/1 complex. In the case of the ketoses, there
can be both C1-C2 and C2-C3 ring fusions, which would
result in additional isomers, such as F. If the complexes are
square pyramids (2b) rather than trigonal bipyramids (2a), an
analogous set of cis/trans isomers exist.
The three peaks for ribose (an aldose) and the five peaks for
fructose (a ketose) thus can be explained by such stereo-
isomerism. The symmetrical molecule 10, however, cannot
exhibit this kind of stereoisomerism, because the two points of
attachment are identical. There are further modes of stereo-
isomerism, however, illustrated by structures G-I. In these
structures, the beginnings of the sugar furanose rings have been
indicated by a pair of bonds on the atoms held in common to
the diolato and furanose rings. In structure G, both furanose
rings are directed syn to the remaining hydroxy group. In
structure I, both furanose rings are anti to the hydroxy group.
In structure H, one furanose ring is syn and one is anti. Each
of the structures C-E can exist as each of the structures G-H,
so that there are a total of nine possible aldose stereoisomers
and more for the ketoses. The three observed 29Si peaks for
ribose may represent more than three species, if there is peak
overlap. The three peaks observed for 1,4-anhydroerythritol (10,
Figure 4) can be explained entirely by the syn/anti isomerism
of structures G-I. The X-ray structure of 10 complexed with
silicic acid proved to have the syn,syn structure G.27
Even within the axial/equatorial structure are there opportuni-
ties for stereochemical multiplicity, as was also recognized by
Conclusions
Certain sugars form complexes with silicic acid under highly
basic conditions. The resulting silicate has a 2/1 sugar/silicic
(24) Clode, D. A. Chem. ReV. 1979, 79, 491-513.
(25) Boeseken, J. AdV. Carbohydr. Chem. 1949, 4, 189-210. Dale, J. J. J. Chem.
Soc. 1961, 922-930.
(26) Chapelle, S.; Verchere, J.-F. Tetrahedron 1988, 44, 4469-4482.
(27) Benner, K.; Klhfers, P. Carbohydr. Res. 2000, 327, 287-292.
9
9624 J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004

Silicate Complexes of Sugars in Aqueous Solution
A R T I C L E S
Experimental Section
acid stoichiometry, in which each sugar is attached via a pair
of hydroxy groups to form a chelating diolato five-membered
ring. Only furanose forms offer a sufficiently small HO-C-
C-OH dihedral angle to allow formation of the nearly planar
five-membered ring. Pyranose rings remain unchanged under
the conditions of reaction. Reaction occurs with the anomeric
carbon (C1 for aldoses, C2 for ketoses), whose hydroxy group
hence must be free (no glycosidic linkage). The hydroxy groups
on the anomeric carbon and its adjacent carbon must be cis for
reaction to proceed to the chelate. The furanose ring with the
appropriate stereochemistry must be present in at least ca. 1%
to be observable with natural abundance 29Si, and the compound
must be sufficiently soluble. Most sugars that meet these criteria
form sugar silicates that may be observed by 29Si and 13C NMR
spectroscopy and mass spectrometry, including the aldoses
ribose, xylose, lyxose, and talose, the ketoses psicose, fructose,
sorbose, and tagatose, and the disaccharides lactulose, maltulose,
and palatinose. Noteworthy failures to form complexes were
glucose, mannose, galactose, and sucrose, which do not meet
these requirements.
General Methods. Sodium silicate solution (14% NaOH, 27% SiO2,
Aldrich) was used as received. NMR spectra were recorded on either
a Varian Unity Plus 400 (400 MHz, 1H; 100 MHz, 13C; 79 MHz, 29Si)
or a Varian Unity Plus 500 (500 MHz, 1H; 125 MHz, 13C) spectrometer.
Chemical shifts for 1H and 13C spectra were referenced against external
trimethylsilylpropionic acid sodium salt (TSP) and are reported relative
to TSP; 29Si was referenced with respect to external tetramethylsilane
(1 M in C6D6) and is reported relative to tetramethylsilane. Electrospray
mass spectrometry was performed with a Thermo Finnegan model LCQ
Advantage mass spectrometer. We mixed 6 M sodium silicate solution
with 4 M sugar solution in D2O to form a final solution 3.3 M in sugar
and 1.1 M in silicate.
1-O-Methyl glycosides were prepared by literature procedures.28
1,4-Anhydrothreitol. 3,4-Epoxytetrahydrofuran (0.5 g, 5.8 mmol)
was dissolved in 15 mL of 3 M sulfuric acid. The mixture was heated
to reflux for 24 h and then was allowed to cool to room temperature.
Solid Na2CO3 was added to the solution until the pH reached 8-9.
The mixture was then evaporated to dryness under vacuum at room
temperature. The residue was extracted with tetrahydrofuran (3 × 10
mL). After filtration, all organic portions were combined and dried
(MgSO4) overnight. The solvent was evaporated, and 1,4-anhydro-
threitol was obtained as a colorless oil (0.43 g, 72%): 1H NMR (D2O)
δ 4.26 (d, 2H), 3.91 (m, 2H), 3.65 (m, 2H); 13C NMR (D2O) δ 78.72,
75.38.
Ribose was the only aldose that formed a truly significant
amount of the silicate (the 29Si peaks for lyxose and xylose were
much weaker), possibly relating to the 6.5% abundance of the
R-furanose form able to form the complex. It is possible that
selection and stabilization of ribose silicates played a role in
the emergence of ribose as the key sugar in nucleotide
biochemical activity. Borate minerals have been suggested as
stabilizing agents for ribose in a prebiotic soup.29 The more
widespread silicate minerals may have performed a similar role
more effectively. Formation of stable, solid silicates would
reduce volatility and provide sequestering that might allow
sugars to survive interplanetary transport on meteors. The role
of sugar silicates in prebiotic evolution needs to be examined.
Synthesis of Hypervalent Silicates. In a typical experiment, the
sugar (2.8 mmol) in 0.7 mL of D2O was added to commercial sodium
silicate solution (0.15 mL, 0.9 mmol). The mixture was shaken
vigorously and sonicated for a few seconds to ensure thorough mixing.
Mass and NMR spectra were taken immediately. Specific results are
compiled in the Supporting Information.
Acknowledgment. This research was supported by the
National Science Foundation (Grant No. CHE-0091162).
Supporting Information Available: Additional experimental
details. This material is available free of charge via the Internet
at http://pubs.acs.org.
(28) Kiersznicki, T.; Szeja, W. Rocz. Chem. 1977, 51, 29-37.
(29) Ricardo, A.; Carrigan, M. A.; Olcott, A. N.; Benner, S. A. Science 2004,
303, 196.
JA031748V
9
J. AM. CHEM. SOC. VOL. 126, NO. 31, 2004 9625