Titanium-beta zeolites catalyze the stereospecific isomerization of d -glucose to l -sorbose via intramolecular C5-C1 hydride shift

Article abstract of DOI:10.1021/cs400273c

Pure-silica zeolite beta containing Lewis acidic framework Ti⁴⁺ centers (Ti-Beta) is shown to catalyze the isomerization of d-glucose to l-sorbose via an intramolecular C5-C1 hydride shift. Glucose-sorbose isomerization occurs in parallel to glucose-fructose isomerization on Ti-Beta in both water and methanol solvents, with fructose formed as the predominant product in water and sorbose as the predominant product in methanol (at 373 K) at initial times and over the course of >10 turnovers. Isotopic tracer studies demonstrate that ¹³C and D labels placed respectively at the C1 and C2 positions of glucose are retained respectively at the C6 and C5 positions of sorbose, consistent with its formation via an intramolecular C5-C1 hydride shift isomerization mechanism. This direct Lewis acid-mediated pathway for glucose-sorbose isomerization appears to be unprecedented among heterogeneous or biological catalysts and sharply contrasts indirect base-mediated glucose-sorbose isomerization via 3,4-enediol intermediates or via retro-aldol fragmentation and recombination of sugar fragments. Measured first-order glucose-sorbose isomerization rate constants (per total Ti; 373 K) for Ti-Beta in methanol are similar for glucose and glucose deuterated at the C2 position (within a factor of ～1.1), but are a factor of ～2.3 lower for glucose deuterated at each carbon position, leading to H/D kinetic isotope effects expected for kinetically relevant intramolecular C5-C1 hydride shift steps. Optical rotation measurements show that isomerization of d-(+)-glucose (92% enantiomeric purity) with Ti-Beta in water (373 K) led to the formation of l-(-)-sorbose (73% enantiomeric purity) and d-(-)-fructose (87% enantiomeric purity) as the predominant stereoisomers, indicating that stereochemistry is preserved at carbon centers not directly involved in intramolecular C5-C1 or C2-C1 hydride shift steps, respectively. This new Lewis acid-mediated rearrangement of glucose to sorbose does not appear to have a metalloenzyme analog.

Full text of DOI:10.1021/cs400273c

Research Article
pubs.acs.org/acscatalysis
Titanium-Beta Zeolites Catalyze the Stereospecific Isomerization of
D‑Glucose to L‑Sorbose via Intramolecular C5−C1 Hydride Shift
Rajamani Gounder and Mark E. Davis*
Chemical Engineering, California Institute of Technology, Pasadena, California 91125, United States
S
* Supporting Information
ABSTRACT: Pure-silica zeolite beta containing Lewis acidic
framework Ti4+ centers (Ti-Beta) is shown to catalyze the
isomerization of D-glucose to L-sorbose via an intramolecular
C5−C1 hydride shift. Glucose−sorbose isomerization occurs in
parallel to glucose−fructose isomerization on Ti-Beta in both
water and methanol solvents, with fructose formed as the
predominant product in water and sorbose as the predominant
product in methanol (at 373 K) at initial times and over the
course of >10 turnovers. Isotopic tracer studies demonstrate
that 13C and D labels placed respectively at the C1 and C2
positions of glucose are retained respectively at the C6 and C5 positions of sorbose, consistent with its formation via an
intramolecular C5−C1 hydride shift isomerization mechanism. This direct Lewis acid-mediated pathway for glucose−sorbose
isomerization appears to be unprecedented among heterogeneous or biological catalysts and sharply contrasts indirect base-
mediated glucose−sorbose isomerization via 3,4-enediol intermediates or via retro-aldol fragmentation and recombination of
sugar fragments. Measured first-order glucose−sorbose isomerization rate constants (per total Ti; 373 K) for Ti-Beta in methanol
are similar for glucose and glucose deuterated at the C2 position (within a factor of ∼1.1), but are a factor of ∼2.3 lower for
glucose deuterated at each carbon position, leading to H/D kinetic isotope effects expected for kinetically relevant intramolecular
C5−C1 hydride shift steps. Optical rotation measurements show that isomerization of D-(+)-glucose (92% enantiomeric purity)
with Ti-Beta in water (373 K) led to the formation of L-(−)-sorbose (73% enantiomeric purity) and D-(−)-fructose (87%
enantiomeric purity) as the predominant stereoisomers, indicating that stereochemistry is preserved at carbon centers not
directly involved in intramolecular C5−C1 or C2−C1 hydride shift steps, respectively. This new Lewis acid-mediated
rearrangement of glucose to sorbose does not appear to have a metalloenzyme analog.
KEYWORDS: fructose, glucose, hydride shift, isomerization, Lewis acid, sorbose, stereospecific, titanium-Beta
located along the glucose backbone; subsequent hydroxyl
deprotonation forms a bound alkoxide moiety with increased
electron density at its carbon center. Such bidentate
coordination of aldehyde and alkoxide groups on open-chain
sugars to a single Lewis acid site facilitates the nucleophilic
addition of electron-rich moieties at the alkoxide carbon center
to the electrophlic aldehyde C1 center preferentially over other
carbon centers in the sugar backbone.
1. INTRODUCTION
Glucose isomerization and epimerization reactions catalyzed by
bases proceed via abstraction of α-carbonyl protons to form
1,2-enediol intermediates, which undergo proton transfer-
mediated rearrangements to form fructose and mannose
products (Lobry de Bruyn−Alberda van Ekenstein rearrange-
ments; LdB−AvE).1−3 Double-bond isomerization of 1,2-
enediols leads to a mixture of 2,3- and 3,4-enediols that are
precursors to psicose, tagatose, and sorbose ketohexoses (the
C3, C4, and C5 epimers of fructose, respectively) and other
aldohexoses.4 Fructose is the preferred product of LdB−AvE
rearrangements of glucose in alkaline media, but it is formed
with a selectivity that decreases with increasing glucose
conversion as sequential 1,2-enediol isomerization4 and as
retro-aldol fragmentation and other degradation reactions of
monosaccharides4−7 become more prevalent. In contrast to
bases that initiate glucose isomerization via α-carbonyl proton
abstraction, Lewis acids coordinate with lone electron pairs in
oxygen atoms (O1) at glucose aldehyde carbons (C1), leading
to polarization of CO bonds and to electron-deficient C1
centers.8 A single Lewis acid center can also coordinate with a
second oxygen atom, such as those found in hydroxyl groups
Infrared (IR) and solid-state 13C nuclear magnetic resonance
(NMR) studies, together with quantum chemical calculations,
have shown that framework Sn centers in zeolite beta (Sn-Beta)
mediate glucose ring-opening and coordinate with glucose O1
and O2 atoms.9 In turn, glucose−fructose isomerization occurs
via an intramolecular hydride shift from C2 to C1 carbon atoms
on open glucose chains (Scheme S1, Supporting Informa-
tion).10 This isomerization mechanism is analogous to that
mediated by two divalent Lewis acid metal centers (e.g., Mg2+
or Mn2+) within hydrophobic pockets of metalloenzymes (e.g.,
Received: April 11, 2013
Revised: May 15, 2013
© XXXX American Chemical Society
1469
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
D-xylose isomerase) that are spatially positioned to bind glucose
via O1 and O2 atoms prior to intramolecular C2−C1 hydride
shift steps.11,12 Sn-Beta can also mediate glucose−mannose
epimerization in methanol,13 and in water in the presence of
borate salts,14 via a Lewis acid-mediated intramolecular carbon
shift known as the Bilik reaction.15−17 In the glucose−mannose
epimerization mechanism, C3 carbon centers bound to C2
atoms behave as nucleophiles and migrate, along with the rest
of the covalently bound sugar backbone, to electrophilic C1
centers (Scheme S1, Supporting Information).
The mechanisms for Sn-mediated glucose−fructose isomer-
ization and glucose−mannose epimerization are similar because
they first require bidentate glucose coordination to metal
centers via O1 and O2 atoms; they differ, in part, because
electron-rich H2 or C3 species located at glucose C2 centers
respectively act as the nucleophiles that add to electron-
deficient C1 centers (Scheme S1, Supporting Information).
These intramolecular hydride and carbon shifts are mediated
only by Lewis acidic framework Sn sites in Sn-Beta and not by
base sites located on extraframework SnO2 domains,13
reflecting the requirement of Lewis acid centers to facilitate
the redistribution of oxidation states between carbon atoms in
organic substrates at transition states for intramolecular10,13,18
or intermolecular19−21 Meerwein−Ponndorf−Verley aldehyde
and ketone reduction and Oppenauer alcohol oxidation
(MPVO) reactions.
Here, we report the first evidence for the direct isomerization
of D-glucose to L-sorbose, (the ketohexose C5 epimer of
fructose), which is mediated by Lewis acidic Ti4+ centers
isolated within the framework of pure-silica zeolite beta (Ti-
Beta). D-Glucose is used as feedstock in the Reichstein
synthesis of L-ascorbic acid (a form of vitamin C; ∼105 tons
produced annually worldwide22) via L-sorbose intermediates.
Current routes for D-glucose-to-L-sorbose conversion involve
the sequential hydrogenation of D-glucose to D-sorbitol over a
nickel-based catalyst and the selective oxidation of D-sorbitol
C2−OH groups using microbial enzymes to form L-
sorbose.22−25 Alkaline media can isomerize glucose into
sorbose, but only among a mixture of several aldohexose and
ketohexose isomers and only via indirect pathways, either via
the formation of 1,2-enediols and isomerization to 3,4-enediol
sorbose precursors4 or via isomerization to fructose, retro-aldol
fragmentation to triose intermediates, and recombination of
sugar fragments.26 Heterogeneous base resins (Amberlite XE-
48, Amberlite IRA-400) can also convert D-(+)-glucose to a
mixture of D-(+)-sorbose (∼68%) and L-(−)-sorbose
(∼32%),27 among several other hexose products, via 3,4-
enediol intermediates.28 The data and the mechanistic evidence
presented herein, to our knowledge, constitute the first report
of direct and stereospecific D-(+)-glucose to L-(−)-sorbose
isomerization mediated by a Lewis acid center or by any
catalytic entity, for that matter.
electron microprobe is denoted in the suffix of sample names
(e.g., Ti-Beta-79 contains a Si/Ti ratio of 79).
The crystal structures of all samples, determined from
powder X-ray diffraction (XRD) patterns collected using a
Rigaku Miniflex II diffractometer and Cu Kα radiation, was
consistent with zeolite beta (Figure S1, section S.2, Supporting
Information). N2 (77 K) adsorption isotherms were measured
using a Quantachrome Autosorb iQ automated gas sorption
analyzer, using protocols reported elsewhere,30 and gave
micropore volumes consistent with the beta topology (Figure
S2, section S.2, Supporting Information). Diffuse reflectance
UV−visible spectra of Ti-Beta samples (Figure S3, section S.2,
Supporting Information) showed bands centered at ∼200−220
nm, which have been assigned previously to Ti centers
incorporated within zeolite frameworks.31
2.2. Kinetic Studies of Glucose Reactions with Ti-Beta.
Reactions with D-glucose (Sigma-Aldrich, ≥ 99%) were
conducted in 10 mL thick-walled glass batch reactors (VWR),
with temperature control via an oil bath located on a digital
stirring hot plate (Fisher Scientific). Typical reactions with D-
glucose were carried out at a 1:50 metal/glucose molar ratio
and involved contacting 4 g of a 1% (w/w) glucose solution
(∼0.04 g of glucose) in water or in methanol with the catalytic
solids (∼0.01−0.04 g) in a stirred glass reactor sealed with a
crimp top (PTFE/silicone septum, Agilent). Kinetic studies
using isotopically labeled glucose were performed using 1% (w/
w) solutions of D-glucose-D2 (Cambridge Isotope Laboratories,
≥ 98%) or of D-glucose-D7-1,2,3,4,5,6,6 (Cambridge Isotope
Laboratories, ≥ 98%) in methanol.
Reactors were placed in the oil bath, and small aliquots
(∼50−100 μL) were extracted at various time intervals via
syringe (Hamilton, 700 series), filtered through a 0.2 μm PTFE
filter (National Scientific), and mixed with 1% (w/w) aqueous
D-mannitol (Sigma-Aldrich, ≥ 98%) solutions used as an
internal standard for quantification. The composition of
reaction aliquots was determined after separation of the
compounds in an Agilent 1200 high performance liquid
chromatograph (HPLC) equipped with an evaporative light
scattering (ELS) detector (Agilent 380 LC). Glucose, sorbose,
mannose, fructose, and mannitol fractions were separated using
a Hi-Plex Ca column (7.7 × 300 mm, 8 μm particle size,
Agilent) held at 353 K, with either ultrapure water (0.010 mL
s−1 flow rate) or a 70/30 (v/v) mixture of acetonitrile/water
(0.013 mL s−1 flow rate) as the mobile phase.
2.3. Isotopic and Stereochemical Characterization of
Sugars. Liquid NMR analysis of products formed from
isotopic tracer studies using D-glucose-D2 or D-glucose-13C-
C1 (Cambridge Isotope Laboratories, ≥ 98%) reactants
involved separation of the glucose, sorbose, and fructose
fractions by HPLC, evaporation of H2O, and dissolution in
1
D2O (Cambridge Isotope Laboratories, 99.9%). H and 13C
liquid NMR spectra were collected on a 400 MHz NMR
spectrometer (Varian) in the Caltech liquid NMR facility. After
collection of NMR spectra, the glucose, sorbose, and fructose
solids were subsequently isolated by evaporation of D2O and
dissolved in H2O prior to measurement of optical rotation at
589 nm and ambient temperature using a Jasco P-2000
polarimeter and a 100 mm path-length cell.
2. EXPERIMENTAL METHODS
2.1. Catalyst Synthesis and Characterization. Proce-
dures to synthesize Ti-Beta zeolites in fluoride media with
different Si/Ti ratios were adapted from reported protocols.29
Ti-Beta samples were treated in flowing air (1.67 cm3 s−1, Air
Liquide, breathing grade) at 853 K (0.0167 K s−1) for 12 h
prior to characterization and catalytic evaluation. Atomic Si and
Ti contents were measured using a JEOL 8200 electron
microprobe, operated in focused beam mode with a 40 μm spot
size, at 15 kV and 25 nA. The Si/Ti ratio determined by
3. RESULTS AND DISCUSSION
3.1. Kinetic Studies of Glucose Isomerization over Ti-
Beta. Monosaccharide yields resulting from the reaction of 1%
(w/w) glucose solutions with Ti-Beta samples (373 K) of
1470
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
a
Table 1. Monosaccharide Yields and Turnover Numbers from Glucose Reactions with Ti-Beta in Water and Methanol
monosaccharide yield (w/w %)
b
catalyst
solvent
glucose/metal ratio
glucose
sorbose
mannose
fructose
total
turnover no.
Ti-Beta-66
Ti-Beta-79
Ti-Beta-107
Ti-Beta-202
Ti-Beta-66
Ti-Beta-79
Ti-Beta-107
Ti-Beta-202
H2O
32
60
82
81
80
87
64
79
76
4
3
<0.1
<0.1
<0.1
<0.1
1.1
11
8
98
93
92
96
85
92
90
88
5.0
6.8
H2O
H2O
56
4
8
6.4
H2O
69
3
6
5.8
CH3OH
CH3OH
CH3OH
CH3OH
30
12
8
8
6.5
63
0.7
4
8.4
54
9
0.8
4
7.9
119
77
6
0.6
4
13.0
a
b
Reaction conditions: 1% (w/w) glucose solutions, 373 K, 2 h. Moles of product monosaccharides formed per moles of total Ti.
Figure 1. 13C NMR spectra of (a) unlabeled glucose and of the glucose fractions isolated after reaction of (b) glucose-13C-C1 and (c) glucose-D2
with Ti-Beta in water at 373 K for 6 h. 13C NMR spectra of (d) unlabeled sorbose together with assignments for each carbon position in α-L-
sorbopyranose32 and of the sorbose fractions isolated after reaction of (e) glucose-13C-C1 and (f) glucose-D2 with Ti-Beta in water at 373 K for 6 h.
varying Si/Ti content in water and in methanol are shown in
Table 1; characterization data for the samples used in this study
are provided in section S.2 of the Supporting Information.
Reactions of glucose with Ti-Beta in water formed predom-
inantly fructose, as reported previously,30,31 together with a
second previously unidentified hexose sugar that became the
predominant product of glucose reactions with Ti-Beta in
methanol solvent (Table 1). This unidentified hexose product
was retained at times similar to (within 0.2 min) that of
mannose during chromatographic separation with a Ca Hi-Plex
column (7.7 × 300 mm, 8 μm particle size, Agilent) using water
as the mobile phase (0.6 mL min−1, 353 K). Mannose and the
unidentified hexose product were resolved, however, upon
changing the mobile phase to a 70/30 (v/v) mixture of
rates that are first-order in glucose concentration (details in
section S.3, Supporting Information). Fructose and sorbose
were formed at nonzero initial turnover rates and in essentially
constant molar ratios as glucose conversion increased from 0 to
36% (Figures S4a and S4b, section S.3, Supporting
Information) and as the number of isomerization turnovers
(per total Ti) increased from 0 to ∼13 (Table 1), reflecting the
formation of both products from primary, parallel reactions of
glucose. This kinetic behavior is inconsistent with the formation
of sorbose via 3,4-enediol intermediates formed upon
sequential rearrangements of 1,2-enediols, or via sequential
retro-aldol fragmentation, isomerization, and recombination of
triose fragments, as has been reported with homogeneous4,26,33
or heterogeneous27,28 bases. Other ketohexoses (e.g., psicose,
tagatose) and trioses were absent in solution after glucose
reactions with Ti-Beta (Table 1), also inconsistent with their
expected formation during base-catalyzed glucose reac-
tions.4−7,26−28
1
acetonitrile/water (0.8 mL min−1, 353 K). H and 13C nuclear
magnetic resonance (NMR) spectra of this previously
unidentified hexose product were identical to that of authentic
sorbose,32 the C5 epimer of fructose.
Kinetic studies of glucose reactions with Ti-Beta (373 K)
performed in batch reactors using 1% (w/w) glucose solutions
in water and in methanol showed that the evolution of liquid-
phase fructose and sorbose concentrations with reaction time
were consistent with that expected from product formation
3.2. Isotopic Tracer Studies of Glucose Isomerization
over Ti-Beta. Isotopic tracer studies using glucose reactants
labeled with 13C at the C1 position (glucose-13C-C1) or with D
1
at the C2 position (glucose-D2), together with H and 13C
NMR spectroscopic analysis of sugar products isolated by
1471
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
Scheme 1. Parallel Reaction Scheme for Glucose−Fructose and Glucose−Sorbose Isomerization Mediated by Lewis Acidic Ti4+
a
Centers in Ti-Beta
a
Mechanistic evidence from isotopic tracer studies using D and 13C labels shown in reactants and products, which are depicted using Fischer
projections.
Scheme 2. Plausible Intermediate (1−3, 5−7) and Transition State (4) Structures Involved in the Proposed Intramolecular C5−
C1 Hydride Shift Reaction Mechanism for Glucose−Sorbose Isomerization on Open Sites in Ti-Beta
fractionation, were used to confirm the presence of Lewis acid
sites on Ti-Beta and to probe the mechanism of glucose−
sorbose isomerization. The 13C NMR spectrum of unlabeled
glucose is shown in Figure 1a for reference. The glucose
fraction isolated after reaction of glucose-13C-C1 with Ti-Beta
in water (373 K) showed resonances at δ = 95.8 and 92.0 ppm
(Figure 1b), which correspond to C1 positions in the β-
pyranose and α-pyranose anomers respectively, of glucose. The
13C NMR spectrum of the glucose fraction collected after
reaction of glucose-D2 with Ti-Beta in water (373 K) showed
low-intensity triplets present in place of sharp resonances at δ =
74.1 and 71.3 ppm (Figure 1c), which correspond to the C2
positions in β-glucopyranose and α-glucopyranose, respectively.
Deuterium atoms bonded to carbon centers suppress the
nuclear Overhauser enhancement (NOE) of carbon resonances
in 13C NMR spectra collected using proton broad-band
decoupling34 and lead to the appearance of low-intensity
triplets, confirmed in the case of glucose-D2 by the absence of
1
resonances for C2−H atoms in the corresponding H NMR
spectrum (Figure S5, section S.4, Supporting Information).
These data (Figures 1b, 1c) confirm that 13C and H/D
scrambling in glucose reactants does not occur during reaction
or chromatographic separation.
The fructose products formed from the reaction of
glucose-13C-C1 and glucose-D2 with Ti-Beta in water
contained a 13C label at its C1 position (fructose- 13C-C1)
and a D label at its C1 position (fructose-D1) (Figures S6 and
S7, section S.4, Supporting Information), respectively, as also
observed on Lewis acidic Sn-Beta.10,13 These isotopic tracer
studies confirm that glucose−fructose isomerization occurs via
1472
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
Figure 2. Liquid-phase concentrations of (a) fructose and (b) sorbose as a function of reaction time during reaction of a 1% (w/w) solution of
⧫
●
▲
glucose ( ), glucose-D2 ( ), or glucose-D7-1,2,3,4,5,6,6 ( ) with Ti-Beta-79 in methanol solvent (1:50 glucose/Ti molar ratio, 373 K).
Corresponding initial turnover rates are given in Table 2. Dashed curves represent best fits of the experimental data to kinetic models derived,
assuming parallel isomerization reactions and rates that are first-order in glucose concentration (details in section S.3, Supporting Information).
an intramolecular C2−C1 hydride shift, reflecting the sole
involvement of Lewis acidic framework Ti4+ sites in Ti-Beta for
glucose isomerization. In contrast, base sites would otherwise
have mediated reversible enolization to cause H/D scrambling
at the C2 position of glucose and, in turn, fructose products
formed without D atoms retained at their C1 position.10,13
The 13C NMR spectrum of unlabeled sorbose is provided in
Figure 1d, together with positional assignments for each carbon
atom in α-L-sorbopyranose.32 The 13C NMR spectrum of the
sorbose fraction isolated after reaction of glucose-13C-C1 with
Ti-Beta in water showed a single resonance at δ = 61.7 ppm
(Figure 1e), which corresponds to the C6 position. In contrast,
isotopic tracer studies using radioisotopically labeled glucose-14
C-C1 reactants showed that base-mediated isomerization via
symmetric 3,4-enediol intermediates form sorbose with 14C
labels at both the C1 and C6 positions,28 whereas base-
mediated isomerization via retro-aldol fragmentation and
recombination reactions redistribute the 14C label throughout
the sorbose backbone.4,28 The 13C NMR spectrum of sorbose
formed form reaction of glucose-D2 with Ti-Beta showed
resonances for all carbon atoms except that for the C5 position
at δ = 70.3 ppm (Figure 1f), whose NOE is suppressed by D
atoms bound to C5 centers. In contrast, base-mediated glucose
isomerization initiated via α-carbonyl proton abstraction would
have caused H/D scrambling at C2 positions of glucose and the
incorporation of H atoms at C5 positions of sorbose. Thus, we
conclude that C and H atoms at glucose C1 and C2 locations
are retained at C6 and C5 locations, respectively, in sorbose
upon isomerization with Lewis acidic Ti4+ centers in Ti-Beta
(Scheme 1). The atom rearrangements involved in glucose−
sorbose isomerization occur between opposite ends of ring-
opened glucose chains; at first glance, they appear to require
several skeletal rearrangement steps but, in fact, require only
glucose C1 aldehyde-to-alcohol reduction and C5 alcohol-to-
ketone oxidation (Scheme 1) in a concerted step, as we discuss
next.
step mediated by a hydride shift from the C5 to the C1 position
over open glucose chains (Scheme 2), analogous to the
intramolecular MPVO mechanism for glucose−fructose iso-
merization mediated by a C2−C1 hydride shift. Quantum
chemical studies of glucose−fructose isomerization on Sn-Beta
and Ti-Beta open sites (three framework −OSi bonds and one
−OH group)9 have shown that coordination of oxygen atoms
in C1−O−C5 hemiacetal linkages of cyclic glucose at Lewis
acidic framework metal (M = Sn, Ti) sites (1, Scheme 2) and
subsequent metal-mediated ring-opening (2, Scheme 2) results
in open-chain glucose bound to metal centers via O1 and O5
atoms (3, Scheme 2). These theoretical studies indicate that
intramolecular C2−C1 hydride shifts in glucose−fructose
isomerization, in fact, require proton transfer from M−(OH2)
groups to glucose O5 atoms in intermediate 3 (Scheme 2) to
form C5−OH groups, desorption of C5−OH groups from M
sites, adsorption of C2−OH moieties after rotation of glucose
coordinated solely via its O1 atom, and deprotonation of C2−
OH groups by M−OH moieties to enable bidentate
coordination of O1 and O2 atoms.9
Glucose−sorbose isomerization instead would require only
an alternate reaction sequence beginning with intermediate 3
(Scheme 2). In this alternate sequence, an intramolecular C5−
C1 hydride shift (4, Scheme 2) would form an open-chain
sorbose bound via O2 and O6 atoms (5, Scheme 2), and
subsequent protonation of sorbose O6 atoms to C6−OH
groups (6, Scheme 2) and ring-closing would form bound cyclic
sorbose (7, Scheme 2), analogous to the steps required to close
glucose−fructose isomerization cycles. In this proposal, the
reaction coordinates for glucose−fructose and glucose−sorbose
isomerization share common elementary steps for binding and
ring-opening of cyclic glucose at framework metal centers
(steps a and b, intermediates 1, 2, and 3, Scheme 2). These
mechanistic features are consistent with the sole formation of
fructose and sorbose with Lewis acidic Ti-Beta (Table 1) and
the absence of psicose and tagatose isomers, which are
otherwise formed concomitantly with fructose and sorbose
via interconvertible enediol intermediates on homogene-
ous4,26,33 or heterogeneous27,28 base catalysts. The purported
3.3. Proposed Mechanism for Glucose−Sorbose
Isomerization on Ti-Beta. We propose that glucose−sorbose
isomerization occurs via a concerted intramolecular MPVO
1473
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
open Ti site structures depicted in Scheme 2, which contain an
−OH group that mediates proton transfer steps with
oxygenated moieties on adsorbed sugars, are present in water
solvent; yet, Ti sites may differ in structure in methanol solvent,
which can coordinate with Lewis acidic Ti centers and
dissociate to form bound methoxy (−OCH3) groups, according
to EXAFS35 and Raman36 studies. We speculate that structural
differences prevalent among active Ti sites in water and
methanol solvent may influence the isomerization selectivity
differences observed with these two solvents (Table 1).
Initial turnover rates (per total Ti; 373 K) for glucose
isomerization to sorbose and isomerization to fructose with Ti-
Beta in methanol were determined from the evolution of
fructose (Figure 2a) and sorbose (Figure 2b) concentrations
with reaction time (additional details given in section S.3,
Supporting Information). Measured first-order isomerization
rate constants derived from these initial turnover rates (Figure
2a, b) with glucose reactants containing different locations and
amounts of isotopic deuterium labels are shown in Table 2.
3.4. Stereospecific Isomerization Mediated by Lewis
Acid Sites. The proposed glucose−sorbose isomerization
mechanism involves the reduction of C1 centers and the
oxidation of C5 centers in D-glucose, but does not change the
formal oxidation states or stereochemistry at C2, C3, and C4
centers (Scheme 1). Thus, D-glucose isomerization via a C5−
C1 hydride shift should selectively form L-sorbose, and
isomerization via C2−C1 hydride shift should, by a similar
argument, selectively form D-fructose. The glucose, sorbose,
and fructose fractions isolated after reaction of a 10% (w/w)
aqueous solution of glucose-D2 with Ti-Beta (373 K, 6 h) and
after collection of 1H and 13C NMR spectra (Figures 1,
Supporting Information S5−S7) and replacement of D2O with
H2O as the solvent were tested for optical activity at 589 nm
and at ambient temperature. The specific optical rotations of
the glucose, sorbose, and fructose fractions were 44.7 0.1°,
−19.8
0.6°, and −67.6
2.7°, respectively (Table 3),
Table 3. Specific Rotation and Enantiomeric Compositions
of Glucose, Sorbose, and Fructose Fractions Isolated After
Reaction of Glucose-D2 with Ti-Beta in Water
a
Table 2. Measured First-Order Rate Constants and H/D
Kinetic Isotope Effects (373 K) for Glucose−Fructose and
b
c
specific rotation
ee
D-enantiomer L-enantiomer
(%) (%)
a
Glucose−Sorbose Isomerization in Methanol
fraction
(deg)
(%)
glucose
sorbose
fructose
44.7 0.1
−19.8 0.6
−67.6 2.7
85 0.3
46 1.5
92
1
1
2
8
1
1
2
measured rate constant
(373 K)
27
87
73
13
(/10−6mol(molTi)−1 s−1
b
−1
73
3
((mol glucose) m−3
)
)
KIE
a
Reaction conditions: 10% (w/w) aqueous glucose-D2 solution, Ti-
Beta-79 (300:1 glucose/Ti ratio), 373 K, 6 h. Specific optical
rotations at 589 nm and ambient temperature. Calculated assuming
each sugar fraction contained only a mixture of the two enantiomers
and the following pure enantiomer optical rotation values: D-
(+)-glucose, 52.7°; L-(−)-sorbose, −42.7°; D-(−)-fructose, −92.4°.
reactant
fructose
sorbose
fructose
sorbose
b
c
glucose
11 0.6
5.0 0.2
5.4 0.3
27 1.3
25 1.2
12 0.6
glucose-D2
2.2 0.3
2.0 0.2
1.1 0.1
2.2 0.2
glucose-D7-
1,2,3,4,5,6,6
a
Calculated from turnover rates measured on Ti-Beta-79 (shown in
Figure 2). Given by the ratio of the rate constant for unlabeled
glucose relative to that for deuterated reactants.
b
reflecting the presence of predominantly D-(+)-glucose, L-
(−)-sorbose, and D-(−)-fructose stereoisomers within the
respective fractions. Table 3 also shows estimated values for
the enantiomeric excess and composition of each sugar fraction,
indicating that isomerization reactions of D-(+)-glucose-D2
(92% enantiomeric purity) with Ti-Beta formed L-(−)-sorbose-
D5 (73% enantiomeric purity) and D-(−)-fructose-D1 (87%
enantiomeric purity) with high stereospecificity.
The predominant formation of L-(−)-sorbose and D-
(−)-fructose from reactions of D-(+)-glucose with Ti-Beta is
consistent with intramolecular hydride shift isomerization
mediated by Lewis acidic Ti centers and specifically with the
intramolecular C5−C1 hydride shift proposed for glucose−
sorbose isomerization (Scheme 2). Such stereochemical
specificity is in sharp constrast to that expected from base-
catalyzed D-glucose isomerization via 3,4-enediol intermediates,
which leads to the predominant formation of D-(+)-sorbose
(68%) over L-(−)-sorbose (32%).27 Such stereospecificity also
contrasts sharply the racemization expected from base-
mediated retro-aldol fragmentation to L-glyceraldehyde and
recombination with dihydroxyacetone, which would also form
similar amounts of L-(−)-sorbose and L-(+)-fructose.27
Rate constants for glucose−fructose isomerization were higher
by a factor of ∼2.2 for glucose than for glucose-D2 (Table 2)
on Ti-Beta in methanol, consistent with isomerization via
kinetically relevant intramolecular C2−C1 hydride shift
(section S.5, Supporting Information), as also observed on
Sn-Beta and Ti-Beta in water.9,10,13 Initial turnover rates for
glucose−sorbose isomerization were essentially identical (with-
in a factor of ∼1.1, Table 2) for glucose and glucose-D2
reactants in methanol, indicating that C2−D bonds remain
intact during such isomerization cycles. The reaction of fully
deuterated glucose (glucose-D7-1,2,3,4,5,6,6) led to an
observed H/D KIE for glucose−fructose isomerization of
∼2.0 in methanol solvent (Table 2) because C2−D bonds are
broken in kinetically relevant steps and also led to a similar KIE
of ∼2.3 for glucose−sorbose isomerization (Table 2). The KIE
values of ∼1.1 and ∼2.3 for glucose−sorbose isomerization
when glucose reactants are deuterated at the C2 position and at
all positions, respectively, reflect the kinetic relevance of C−D
bond-breaking steps at a position other than C2. Although we
are unable to probe C5−D cleavage directly because glucose-
D5 reactants are unavailable, the KIE of ∼2.3 observed with
fully deuterated glucose reactants is expected from kinetically
relevant C5−D bond cleavage (section S.5, Supporting
Information), as required for the proposed intramolecular
C5−C1 hydride shift mechanism (Scheme 2), and consistent
with the isotopic tracer studies (Scheme 1) that led to this
proposal.
Direct glucose−fructose isomerization via intramolecular
C2−C1 hydride shifts mediated by Ti-Beta has known
metalloenzyme analogs (e.g., D-xylose isomerase), in which
two divalent cations in the enzyme active site pockets must
interact in a concerted manner to coordinate with glucose O1
and O2 atoms prior to isomerization.11,12 In contrast, direct
glucose−sorbose isomerization via intramolecular C5−C1
hydride shifts mediated by Ti-Beta does not appear to have a
1474
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
known metalloenzyme analog. The lack of enzymes that
mediate direct glucose−sorbose isomerization appears evident
in currently known routes for D-glucose to L-sorbose isomer-
ization, which require sequential reduction to a sugar alcohol
and oxidation to sorbose by a metal and an enzyme22−25 or by
two different enzymes37 operating in series. This observation
suggests that enzymatic active sites that selectively bind glucose
via O1 and O5 atoms may not be as prevalent as those that
bind glucose via O1 and O2 atoms and, in part, may be an
underlying factor contributing to the rarity of L-sorbose found
in nature.37 These findings indicate that Lewis acidic metal
centers in synthetic molecular sieve frameworks that can
coordinate selectively with two oxygenated moieties along sugar
backbones may be able to facilitate direct and stereospecific
sugar rearrangements that occur rarely in biological systems.
how these features may differ from those responsible for
glucose−fructose isomerization will provide specific guidance
for the synthesis of active site ensembles that selectively
mediate intramolecular sugar rearrangements.
ASSOCIATED CONTENT
■
S
* Supporting Information
Mechanistic scheme for glucose isomerization and epimeriza-
tion on Sn-Beta, catalyst characterization (X-ray diffractograms,
N2 adsorption isotherms, diffuse reflectance UV−visible
spectra), kinetic studies of glucose reactions with Ti-Beta-F,
product identification (1H and 13C liquid NMR spectra), and
H/D kinetic isotope effect estimates are provided. This material
is available free of charge via the Internet at http://pubs.acs.
org/.
4. CONCLUSIONS
AUTHOR INFORMATION
Lewis acidic Ti4+ centers in the framework of pure-silica zeolite
beta (Ti-Beta) mediate the isomerization of glucose to sorbose
in a direct step involving intramolecular C5−C1 hydride shift
(Scheme 2), consistent with isotopic tracer studies in which 13C
and D labels placed at the C1 and C2 positions of glucose are
retained at the C6 and C5 positions, respectively, of sorbose.
Glucose−fructose and glucose−sorbose isomerization reactions
are catalyzed in parallel by Ti-Beta in water and methanol
solvent (373 K), with the former reaction predominating in
water and the latter in methanol. Turnover rates of glucose−
fructose and glucose−sorbose isomerization sequences are
limited by a kinetically relevant intramolecular C2−C1 hydride
shift and intramolecular C5−C1 hydride shift steps, respec-
tively, consistent with observed H/D kinetic isotope effects
(373 K) of 2.0−2.3 for the former reaction with both glucose-
D2 and fully deuterated glucose reactants, and for the latter
reaction only with fully deuterated glucose reactants. Intra-
molecular C2−C1 or C5−C1 hydride shift steps preserve the
stereochemistry at unreacted glucose carbon centers (C3, C4,
and C5 or C2, C3, and C4, respectively, Scheme 1), leading to
the isomerization of D-(+)-glucose (92% enantiomeric purity)
to D-(+)-fructose (87% enantiomeric purity) and to L-
(−)-sorbose (73% enantiomeric purity) with high stereo-
specificity. The stereospecificity for D-glucose to L-sorbose
isomerization is inaccessible to base-catalyzed isomerization
initiated via α-carbonyl proton abstraction.
■
Corresponding Author
*E-mail: mdavis@cheme.caltech.edu.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was financially supported as part of the Catalysis
Center for Energy Innovation, an Energy Frontier Research
Center funded by the U.S. Department of Energy, Office of
Science, Office of Basic Energy Sciences under Award No. DE-
SC0001004. We thank Prof. Brian M. Stoltz (Caltech) for use
of the optical polarimeter.
REFERENCES
■
(1) Lobry de Bruyn, C. A.; Alberda van Ekenstein, W. Recl. Trav.
Chim. Pays-Bas. 1895, 14, 203.
(2) Angyal, S. J. Top. Curr. Chem. 2001, 215, 1−14.
(3) Speck, J. C. Adv. Carbohydr. Chem. 1958, 13, 63−103.
(4) Elkhadem, H. S.; Ennifar, S.; Isbell, H. S. Carbohydr. Res. 1987,
169, 13−21.
(5) Yang, B. Y.; Montgomery, R. Carbohydr. Res. 1996, 280, 27−45.
(6) Dewit, G.; Kieboom, A. P. G.; Vanbekkum, H. Carbohydr. Res.
1979, 74, 157−175.
(7) Kooyman, C.; Vellenga, K.; Dewilt, H. G. J. Carbohydr. Res. 1977,
54, 33−44.
(8) Roman
1580.
(9) Bermejo-Deval, R.; Assary, R. S.; Nikolla, E.; Moliner, M.;
Roman-Leshkov, Y.; Hwang, S.-J.; Pallsdottir, A.; Silverman, D.; Lobo,
́
-Leshkov, Y.; Davis, M. E. ACS Catal. 2011, 1, 1566−
This is the first report, to our knowledge, for direct or
stereospecific D-(+)-glucose to L-(−)-sorbose isomerization in a
single step mediated by a Lewis acid catalyst or by any catalytic
entity. Catalytic routes for direct D-glucose to L-sorbose
isomerization have useful applications in the Reichstein
synthesis of L-ascorbic acid, which currently requires sequential
conversion of D-glucose to D-sorbitol via metal-catalyzed
hydrogenation and conversion of D-sorbitol to L-sorbose via
enzyme-catalyzed selective oxidation C2-OH. In contrast with
Lewis acid-mediated glucose−fructose isomerization, for which
known metalloenzyme analogs (e.g., D-xylose isomerase) exist,
there does not appear to be a metalloenzyme analog for
glucose−sorbose isomerization. The demonstration that Ti-
Beta can mediate stereospecific D-glucose to L-sorbose
isomerization via intramolecular MPVO catalytic cycles offers
promise for the development of synthetic catalyst structures
with Lewis acid function that can mediate new sugar
rearrangements via intramolecular MPVO cycles. We expect
that further insights into the active site structural features
responsible for glucose−sorbose isomerization on Ti-Beta and
́
R. F.; Curtiss, L. A.; Davis, M. E. Proc. Natl. Acad. Sci. U.S.A. 2012, 109,
9727−9732.
́
(10) Roman-Leshkov, Y.; Moliner, M.; Labinger, J. A.; Davis, M. E.
Angew. Chem., Int. Ed. 2010, 49, 8954−8957.
(11) Collyer, C. A.; Blow, D. M. Proc. Natl. Acad. Sci. U.S.A. 1990, 87,
1362−1366.
(12) Kovalevsky, A. Y.; Hanson, L.; Fisher, S. Z.; Mustyakimov, M.;
Mason, S. A.; Forsyth, V. T.; Blakeley, M. P.; Keen, D. A.; Wagner, T.;
Carrell, H. L.; Katz, A. K.; Glusker, J. P.; Langan, P. Structure 2010, 18,
688−699.
(13) Bermejo-Deval, R.; Gounder, R.; Davis, M. E. ACS Catal. 2012,
2, 2705−2713.
(14) Gunther, W. R.; Wang, Y. R.; Ji, Y. W.; Michaelis, V. K.; Hunt, S.
T.; Griffin, R. G.; Roman-Leshkov, Y. Nat. Commun. 2012, 3, 1109.
(15) Bilik, V.; Petrus, L.; Farkas, V. Chem. Zvesti 1975, 29, 690−696.
(16) Bilik, V.; Petrus, L.; Zemek, J. Chem. Zvesti 1978, 32, 242−251.
(17) Osanai, S. Top. Curr. Chem. 2001, 215, 43−76.
(18) Lobo, R. F. ChemSusChem 2010, 3, 1237−1240.
1475
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476

ACS Catalysis
Research Article
(19) Boronat, M.; Corma, A.; Renz, M. J. Phys. Chem. B 2006, 110,
21168−21174.
(20) Corma, A.; Domine, M. E.; Nemeth, L.; Valencia, S. J. Am.
Chem. Soc. 2002, 124, 3194−3195.
(21) Corma, A.; Domine, M. E.; Valencia, S. J. Catal. 2003, 215,
294−304.
(22) Bremus, C.; Herrmann, U.; Bringer-Meyer, S.; Sahm, H. J.
Biotechnol. 2006, 124, 196−205.
(23) Hancock, R. D.; Viola, R. Trends Biotechnol. 2002, 20, 299−305.
(24) Boudrant, J. Enzyme Microb. Technol. 1990, 12, 322−329.
(25) Eggersdorfer, M.; Laudert, D.; Letinois, U.; McClymont, T.;
Medlock, J.; Netscher, T.; Bonrath, W. Angew. Chem., Int. Ed. 2012, 51,
12960−12990.
(26) Fischer, E. Ber. Dtsch. Chem. Ges. 1890, 23, 2114.
(27) Blair, M. G.; Sowden, J. C. J. Am. Chem. Soc. 1955, 77, 3323−
3325.
(28) Sowden, J. C.; Thompson, R. R. J. Am. Chem Soc. 1958, 80,
1435−1438.
(29) Blasco, T.; Camblor, M. A.; Corma, A.; Esteve, P.; Guil, J. M.;
Martinez, A.; Perdigon-Melon, J. A.; Valencia, S. J. Phys. Chem. B 1998,
102, 75−88.
(30) Gounder, R.; Davis, M. E. AIChE J. 2013, DOI: 10.1002/
aic.14016.
́
(31) Moliner, M.; Roman-Leshkov, Y.; Davis, M. E. Proc. Natl. Acad.
Sci. U.S.A. 2010, 107, 6164−6168.
(32) Que, L.; Gray, G. R. Biochemistry 1974, 13, 146−153.
(33) Crueger, A.; Crueger, W. Glucose Transforming Enzymes. In
Microbial Enzymes and Biotechnology, ;Fogarty, W. M., Kelly, C. T.,
Eds.; Springer Netherlands: The Netherlands, 1990; pp 177−226.
(34) Neuhaus, D.; Williamson, M. P. The Nuclear Overhauser Effect in
Structural and Conformational Analysis. VCH: Weinheim, 1989.
(35) Davis, R. J.; Liu, Z.; Tabora, J. E.; Wieland, W. S. Catal. Lett.
1995, 34, 101−113.
(36) Wang, L. L.; Xiong, G.; Su, J.; Li, P.; Guo, H. C. J. Phys. Chem. C
2012, 116, 9122−9131.
(37) Granstrom, T. B.; Takata, G.; Tokuda, M.; Izumori, K. J. Biosci.
Bioeng. 2004, 97, 89−94.
1476
dx.doi.org/10.1021/cs400273c | ACS Catal. 2013, 3, 1469−1476