Stereoselective hydrogenation of olefins using rhodium-substituted carbonic anhydrase - A new reductase

Article abstract of DOI:10.1002/chem.200801673

One useful synthetic reaction missing from nature's toolbox is the direct hydrogenation of substrates using hydrogen. Instead nature uses co-factors like NADH to reduce organic substrates, which adds complexity and cost to these reductions. To create an enzyme that can directly reduce organic substrates with hydrogen, researchers have combined metal hydrogenation catalysts with proteins. One approach is an indirect link where a ligand is linked to a protein and the metal binds to the ligand. Another approach is direct linking of the metal to protein, but nonspecific binding of the metal limits this approach. Herein, we report a direct hydrogenation of olefins catalyzed by rhodium(I) bound to carbonic anhydrase (CA-[Rh]). We minimized nonspecific binding of rhodium by replacing histidine residues on the protein surface using site-directed mutagenesis or by chemically modifying the histidine residues. Hydrogenation catalyzed by CA-[Rh] is slightly slower than for uncomplexed rhodium(I), but the protein environment induces stereoselectivity favoring cis- over transstilbene by about 20:1. This enzyme is the first cofactor-independent reductase that reduces organic molecules using hydrogen. This catalyst is a good starting point to create variants with tailored reactivity and selectivity. This strategy to insert transition metals in the active site of metalloenzymes opens opportunities to a wider range of enzyme-catalyzed reactions.

Full text of DOI:10.1002/chem.200801673

DOI: 10.1002/chem.200801673
Stereoselective Hydrogenation of Olefins Using Rhodium-Substituted
Carbonic Anhydrase—A New Reductase
[
a]
Qing Jing, Krzysztof Okrasa, and Romas J. Kazlauskas*
Abstract: One useful synthetic reaction
missing from natureꢀs toolbox is the
direct hydrogenation of substrates
using hydrogen. Instead nature uses co-
factors like NADH to reduce organic
substrates, which adds complexity and
cost to these reductions. To create an
enzyme that can directly reduce organ-
ic substrates with hydrogen, research-
ers have combined metal hydrogena-
tion catalysts with proteins. One ap-
proach is an indirect link where a
ligand is linked to a protein and the
metal binds to the ligand. Another ap-
proach is direct linking of the metal to
protein, but nonspecific binding of the
metal limits this approach. Herein, we
report a direct hydrogenation of olefins
catalyzed by rhodium(I) bound to car-
bonic anhydrase (CA-[Rh]). We mini-
mized nonspecific binding of rhodium
by replacing histidine residues on the
protein surface using site-directed mu-
tagenesis or by chemically modifying
the histidine residues. Hydrogenation
catalyzed by CA-[Rh] is slightly slower
than for uncomplexed rhodium(I), but
the protein environment induces ste-
reoselectivity favoring cis- over trans-
stilbene by about 20:1. This enzyme is
the first cofactor-independent reduc-
tase that reduces organic molecules
using hydrogen. This catalyst is a good
starting point to create variants with
tailored reactivity and selectivity. This
strategy to insert transition metals in
the active site of metalloenzymes
opens opportunities to a wider range of
enzyme-catalyzed reactions.
Keywords: carbonic anhydrase
·
enzyme catalysis · hydrogenation ·
metalloenzymes · protein modifica-
tions
Introduction
positioning in an active site. This likely makes the enzymes
more efficient than they would be with hydrogen. Cofactors
differ in their redox balance in the cell. Cells contain ap-
Although natureꢀs enzymes efficiently catalyze a wide range
of chemical reactions, their catalytic range is limited to reac-
tions that offer a selective advantage to living organisms.
One notable reaction missing from natureꢀs toolbox is the
direct hydrogenation of substrates using hydrogen gas. This
is a common and useful reaction in organic synthesis, but
nature does not use it. Hydrogenases are enzymes that
cleave hydrogen to protons and electron, but they do not
+
proximately equal amounts of NADH and NAD , but they
+
contain 200-fold more NADPH than NADP . This differ-
ence allows NADPH-utilizing enzymes to be more efficient
at reducing substrates than they would be if the entire cell
had a single redox balance. Finally, cofactors are charged
and do not diffuse out of cells. These evolutionary advantag-
es do not accrue to enzyme-catalyzed synthesis using isolat-
ed enzymes in vitro. In these cases, cofactors are a disad-
vantage since they are unstable and add complexity and cost
to the syntheses.
[1]
reduce organic molecules.
To reduce organic molecules during metabolism nature
uses cofactors such as NAD(P)H or FADH , which provides
2
evolutionary advantages over the direct use of hydrogen.
These cofactors are large to allow tight binding and precise
Recognizing this need for enzymes that directly reduce
substrates with hydrogen, researchers created hybrids of
[2]
chemical hydrogenation catalysts and proteins. One ap-
proach was an indirect link. Researchers bound the metal
catalyst to a water-soluble ligand (often a phosphine) and
then linked the ligand to the protein. Most researchers used
a phosphorus ligand containing a biotin moiety, which binds
[
a] Dr. Q. Jing, Dr. K. Okrasa, Prof. R. J. Kazlauskas
Department of Biochemistry, Molecular Biology & Biophysics, and
the Biotechnology Institute
University of Minnesota
1
479 Gortner Avenue, Saint Paul, MN 55108 (USA)
[3]
tightly to the protein avidin. In the best cases—hydrogena-
tion of a-acetamidoacrylic acid and a-acetamidocinnamic
acid—the catalysts were highly enantioselective (ꢀ96% ee)
Fax : (+1)612-625-5780
E-mail: rjk@umn.edu
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.200801673.
[3c–d]
and gave high yields.
Another group bound a rhodium–
1370
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 1370 – 1376

FULL PAPER
[4]
phosphine complex to a monoclonal antibody. The catalyst
showed high enantioselectivity for one substrate, but the
yield was disappointingly low (23%). The disadvantage of
this indirect linking is that optimization still involves multi-
step synthesis of the ligands and linkers or generation of
monoclonal antibodies.
A second approach is directly linking hydrogenation cata-
lysts such as rhodium(I) or rhodium(II) complexes to pro-
tein surfaces. Unfortunately, rhodium and similar metals
bind at multiple sites so the re-
Results
Replacing zinc with rhodium in carbonic anhydrase: Previ-
ous researchers developed methods to replace the active site
[12]
zinc in carbonic anhydrase with another metal ion. Dialy-
sis of carbonic anhydrase against zinc chelator—2,6-pyridi-
nedicarboxylate—removes the zinc to form apo(carbonic an-
hydrase). A new metal ion added to the solution will bind
forming metal-substituted carbonic anhydrase (Figure 1) We
sulting catalysts are nonselec-
tive. For example, a rhodiu-
m(II)
Rh (carboxylate) ]
complex—
binds
[
2
4
weakly to serum albumin (K
3
ꢁ
10 ) in a molar ratio of about
[5]
8
Rh per serum albumin. The
rhodium probably binds to the
side chains of cysteine (one res-
idue in human serum albumin)
and histidine (sixteen residues).
Figure 1. Dialysis of carbonic anhydrase against a zinc chelator—2,6-pyridinedicarboxylate—removed 90–95%
of the active-site zinc. Subsequent dialysis against a solution of [Rh ACHTUNGTERNN(NUG cod) ]BF yielded rhodium(I)-substituted
2 4
Circular dichroism shows a loss carbonic anhydrase. 9*His hCAII is a variant of human carbonic anhydrase isoenzyme II, in which site-direct-
in helicity indicating that the ed mutatgenesis replaced nine of the histidine residues on the surface by arginine, alanine, or phenylalanine
residues. cod=1,5-cyclooctadiene.
binding distorts the proteinꢀs
secondary structure. In some
cases, the binding increased the
stability of the albumin to unfolding, possibly by forming
used bovine carbonic anhydrase isoenzyme II (bCAII, com-
mercially available), human carbonic anhydrase isoenzy-
me II (hCAII, produced by a recombinant strain of E. coli
according to Fierkeꢀs procedure ) as well as variants of
hCAII where selected histidine residues were replaced by
other amino acid residues, see below. We tested several
[6]
cross-links between several imidazoyl groups. Marchetti
and co-workers used a similar complex (serum albumin and
a 30-fold molar excess of rhodium(I)) to catalyze hydrofor-
mylation. A preliminary report suggested some selectivity,
[13]
[7]
but the full paper did not report any stereoselectivity.
A
direct-linking approach that did yield a size-selective hydro-
genation catalyst was the entrapment of 460 palladium
atoms within the 12 nm-diameter cavity of apoferritin. Pre-
sumably, the larger substrates could not enter the pores of
the apoferritin and thus did not react.
rhodium
ACHTUNGTRENNGNU( III) and rhodium(I) salts as metal sources. [Rh-
AHCTUNGTREG(NNNU cod) ]BF (cod=1,5-cyclooctadiene) gave a rhodium-car-
2
4
[8]
bonic anhydrase complex with the highest activity in hydro-
genation and the most stable to precipitation. We used [Rh-
AHCTUNGTNERUNG( cod) ]BF for all subsequent experiments.
2 4
A third approach—not yet realized—is to replace the
active-site metal of a metalloenzyme with rhodium or a sim-
ilar metal. Carbonic anhydrase is a good starting point for
several reasons. First, both zinc (atomic number 30) and
Like previous researchers that prepared rhodium–albumin
[5,6]
complexes,
we found that carbonic anhydrase forms
stable, pale yellow complexes with the rhodium salts and
that each carbonic anhydrase binds multiple rhodium ions
(Table 1). Inductively coupled plasma mass spectrometry
(ICP-MS) revealed the amount of rhodium, while a Brad-
[9]
rhodium (atomic number 45) have similar ionic radii, so
the rhodium may not distort the active site. Second, the
[10]
[14]
metal ligands are the imidazoyl groups of three histidines,
ford dye-binding assay revealed the amount of carbonic
which are also good ligands for rhodium. Third, bovine car-
bonic anhydrase lacks cysteine residues that could interfere
with binding of rhodium in the active site; human carbonic
anhydrase has one deeply buried cysteine residue. Fourth,
we, and others, previously replaced the active-site zinc in
carbonic anhydrase (CA) with manganese to create a cata-
anhydrase. Both bovine carbonic anhydrase (Table 1,
entry 1) and human CA (Table 1, entry 3) complexed an
average of 7.5 and 6.5 rhodium ions, respectively.
Minimizing the surface-bound rhodium: We hypothesized
that the extra rhodium binds on the protein surface most
likely to the side chains of histidine or lysine. Bovine CA II
contains eight histidine and eighteen lysine residues outside
[11]
lyst for the enantioselective epoxidation of styrene.
Herein, we describe our replacement of the active site
zinc in carbonic anhydrase with rhodium, our strategies to
minimize binding of extra rhodium ions at the proteinꢀs sur-
face and the cis-selective hydrogenation of stilbene with the
resulting catalyst.
[15]
the active site, while the human enzyme contains nine and
[16]
twenty-four, respectively. We tested this hypothesis by se-
lective chemical modification of the lysine and either site-di-
rected mutagenesis or a combination of chemical modifica-
tion and site-directed mutagenesis for the histidine residues.
Chem. Eur. J. 2009, 15, 1370 – 1376
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1371

R. J. Kazlauskas et al.
Table 1. Rhodium to protein ratios of various rhodium–carbonic anhy-
drase complexes.
rected mutagenesis with chemical modification—minimized
the binding of rhodium to the surface of carbonic anhydrase
while allowing a rhodium to bind to the active site. As a
result, we could test whether this hybrid catalyzes hydroge-
nation and whether the protein controls the selectivity of
the hydrogenation.
[
a]
Entry
Protein
Total Rh
Extra Rh
1
bCAII-[Rh]
Acetyl-bCAII-[Rh]
hCAII-[Rh]
H4/10R+H17F-[Rh]
H3/4/10R+H17F-[Rh]
9*His hCAII-[Rh]
DEPC-H4/10R+H17F-[Rh]
7.5
10.0
6.5
4.3
4.2
1.8
2.3
6.5
9.0
5.5
3.3
3.2
0.8
1.3
[
b]
2
3
4
5
6
7
[
[
c]
Hydrogenation: As expected, native carbonic anhydrase,
which contains zinc in the active site, does not catalyze hy-
drogenation of cis-stilbene (Table 2, entry 1). Substituting
d]
[
a] Rhodium content was measured by inductively coupled plasma mass
spectrometry, while the protein content was measured by Bradford dye
assay at 595 nm. bCAII=bovine carbonic anhydrase II (entries 1 and 2);
hCAII=human carbonic anhydrase II (entries 3–7). All variants refer to
hCAII. [b] Treated with acetic anhydride to block the e-amino group of
lysine residues. [c] 9*His refers to the mutant H3/4/10/36R+H15/17/107/
Table 2. Hydrogenation and isomerization of cis-stilbene catalyzed by
rhodium-carbonic anhydrase complexes.
[
a]
1
22F+H64A, in which the 9 histidine residues were replaced by arginine,
phenylalanine, or alanine as indicated. [d] hCAII was treated with dieth-
ylpyrocarbonate to block the imidazole amino group of the histidine resi-
dues before replacement of zinc with rhodium.
[
b]
[c]
[d]
Entry (Enzymatic) catalysts
H
2
H
I
H/I
Acetylation of the e-amino group of lysine decreases their
ACHTUNGTREN[NUGN atm] [%] [%] ratio
basicity and ability to bind metals. Acetylation of bovine
1
2
3
4
5
6
7
8
9
0
1
bCAII-[Zn]
bCAII-[Rh]
hCAII-[Rh]
hCAII-[Rh]
H4/10R+H17F-hCAII-[Rh]
H4/10R+H17F- hCAII-[Rh]
H3/4/10R+H17F- hCAII-[Rh]
H3/4/10R+H17F- hCAII-[Rh]
DEPC-H4/10R+H17F-hCAII-[Rh]
9*His- hCAII-[Rh]
1
1
1
5
1
5
1
5
5
5
5
–
–
–
[17]
carbonic anhydrase with acetic anhydride added an aver-
age of fourteen acetyl groups (see mass spectra in the Sup-
porting Information). Surprisingly, this acetylation increased
the number of bound rhodium ions from 7.5 to 10.0
15.8 16.4
4.2 5.8
55.5 15.8
21.3 21.2
80.5 19.0
11.6
42.5
46.8
55.6
0.9
0.7
3.5
1.0
4.2
1.2
7.6
[
e]
(
Table 1, entries 1 and 2). The increase may be due to non-
9.8
5.6
2.8 16.7
2.7 20.6
specific binding caused by the increased negative charge on
the acetylated carbonic anhydrase. Bovine carbonic anhy-
drase is slightly negatively charged at pH 7 (its isoelectric
point is 6.4), and acetylation of fourteen lysine residues in-
creases the overall negative charge by fourteen. This in-
crease in the number of bound rhodium ions indicates that
lysine residues probably do not bind the extra rhodium ions.
Diethylpyrocarbonate (DEPC), a histidine-selective re-
agent, adds an ethoxycarbonyl group to one of the imidazole
nitrogens. This addition of an electron-withdrawing group
makes the remaining imidazole nitrogen a poorer ligand.
Unfortunately, treatment of hCAII with DEPC to modify
all nine non-active-site histidines (see Figure S1 in the Sup-
[f]
1
1
[Rh
A
H
U
G
R
N
N
(cod)
2
]BF
4
80.0 12.6
6.3
[
[
a] Reaction time 12 h. MES=2-(N-morpholino)ethanesulfonic acid.
b] Hydrogenation. [c] Isomerization. [d] Hydrogenation/isomerization
ratio. [e] Entries 3–9 refer to human CA or variants. [f] 9*His-hCAII
refers to variant H3/4/10/36R+H15/17/107/122F+H64A, in which nine
histidine residues were replaced by the amino acids indicated.
the zinc by rhodium yields a hydrogenation catalyst for cis-
stilbene (Table 2, entries 2–4; note the change in H pressure
2
for entries 4, 6, 8–11). Complexation to the protein slightly
slows the hydrogenation as indicated by the lower hydroge-
nation conversion for hCAII-[Rh] as compared to uncom-
plexed rhodium, 55 versus 80% (Table 2, entries 4 vs. 11). In
both cases, isomerization of cis-stilbene to the more stable
trans-stilbene accompanied the hydrogenation—the ratio of
hydrogenation to isomerization is 3.5 (Table 2, entry 4) and
6.3 (Table 2, entry 11).
We hypothesized that the rhodium outside the active site
catalyzed most of the isomerization, while the rhodium in
the active site catalyzed mostly hydrogenation. Consistent
with this hypothesis, variants of carbonic anhydrase binding
fewer extra rhodium atoms showed less isomerization and
therefore elevated ratios of hydrogenation to isomerization.
The DEPC-H4/10R+H17F-hCAII-[Rh] variant showed a
hydrogenation to isomerization ratio of 16.7 (Table 2,
entry 9) and 9*His-hCAII-[Rh] showed a ratio of 20.6
(Table 2, entry 10). These ratios are 2.7 and 3.3-fold higher
[18]
porting Information), yielded an unstable protein, so we
used either site-directed mutagenesis or a combination of
site-directed mutagenesis and modification with DEPC to
block the binding of rhodium to surface histidines. In the
site-directed mutagenesis approach, we replaced these nine
histidine residues in hCAII with arginine (maintains positive
charge), phenylalanine (maintains shape and hydrophobic
character), or alanine (creates more space nearby the active
site). This variant yielded a rhodium–CA complex that had
only 0.8 extra Rh. In the chemical modification approach,
we used variant H4/10R+H17F because it was more stable
than the wild-type hCAII. Treatment of H4/10R+H17F-
hCAII with DEPC to modify the six remaining surface histi-
dines followed by replacement of the active site zinc with
rhodium yielded a protein that bound only 1.3 extra rhodi-
um ions. These results indicate that the extra rhodium likely
binds to the non-active-site histidine residues. Thus, two ap-
proaches—site-directed mutagenesis and combined site-di-
than the ratio for [Rh ACHTUNGTRNEGNU( cod) ]BF (Table 1, entry 11). These
2 4
1372
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 1370 – 1376

Stereoselective Hydrogenation of Olefins
FULL PAPER
results demonstrate that the protein environment controls
reactivity—it reduces the amount of isomerization to the
trans isomer during hydrogenation.
To confirm that the rhodium at the active site catalyzes
hydrogenation, we compared the hydrogenation of a rhodi-
um–carbonic anhydrase hybrid to a variant where zinc re-
mained in the active site, but rhodium bound to the surface.
This variant catalyzed mainly isomerization (Table 3,
hCAII-[Rh]), 55.6% versus 2.1% (Table 4, entries 5 vs. 6,
for 9*His-hCAII-[Rh]).
Computer modeling was consistent with the notion that
cis-stilbene fits in the active site of human carbonic anhy-
drase II better than trans-stilbene. Reliable molecular me-
chanics parameters were not available for rhodium, so the
model contained zinc in the active site. This modeling iden-
tifies the fit of the two stereoisomers in the active site, but
[19]
does not model the hydrogenation reaction itself.
One
phenyl group of the stilbene stereoisomers fits into a hydro-
Table 3. Hydrogenation of cis-stilbene catalyzed by rhodium-substituted
phobic pocket in the active site largely defined by Val121,
hCAII and a variant that retains zinc in the active site, but has rhodium
[
a]
[20]
on the surface.
Val143, Leu198, and Trp209 (Figure 2).
The geometry-
[
b]
[c]
[b]
Entry Enzymatic catalysts
H
[
I
H
at active site
[
d]
%] [%] [%]
21.3 21.2 80.3
H4/10R+H17F-hCAII-[Zn]-[Rh] 4.2 16.0
, 12 h reaction time, at room tempera-
1
2
H4/10R+H17F-hCAII-[Rh]
–
[
a] Reaction conditions: 1 atm H
2
ture. The H4/10R+H17F variant contains 3.3 extra rhodium ions. We
chose this variant for the experiment because it is more stable than other
variants. [b] Hydrogenation. [c] Isomerization. [d] (21.3–4.2)/21.3=
8
0.3%.
entry 2), whereas the variant containing rhodium both in the
active site and the surface catalyzed both hydrogenation and
isomerization (Table 2, entry 1). If we assume that the sur-
face-bound rhodium behaves similarly in the two cases, then
80% of the hydrogenation occurs in the active site for the
variant containing rhodium in both the active site and on
the surface. Similarly, about 25% of the isomerization
occurs in the active site.
Stereoselective hydrogenation of cis-stilbene over trans-stil-
bene: The lower amount of isomerization of cis-stilbene to
trans-stilbene during hydrogenation by the rhodium–carbon-
ic anhydrase hybrids with the least extra rhodium ions sug-
gests that these catalysts show stereoselective hydrogenation
favoring cis-stilbene over trans-stilbene. [Rh ACHTUNGTENRUNG( cod) ]BF cata-
2 4
lyzed hydrogenation of both cis- and trans-stilbene with sim-
ilar degrees of conversion under our conditions: 80% versus
57% conversion (Table 4, entries 1 vs. 2; stereoselectivity
approximately 1.4). In contrast, rhodium–carbonic anhy-
drase hybrids favored the hydrogenation of cis-stilbene over
trans-stilbene by more than 20-fold: 80.5% versus 3.5%
conversion (Table 4, entries 3 vs. 4, for H4/10R+H17F-
[
a]
Table 4. Stereoselective hydrogenation of cis- over trans-stilbene.
[
b]
[c]
[d]
Figure 2. Modeling fits the double bond of cis-stilbene facing the metal
ion in human carbonic anhydrase II (a), but the double bond of trans-stil-
bene does not face the metal ion (b). The zinc ion is the orange ball and
amino acids nearest the active site are shown as sticks in CPK colors.
Two histidines close the active site (His64, His107) are among those re-
placed by other amino acids. The residues are shown as the replaced
amino acids (Ala64, Phe107) in pink. The surrounding protein atoms are
in grey space-filling representation and the cis- or trans-stilbene are
shown as sticks in cyan. Modeling included the water molecules defined
in the X-ray crystal structure, but these are not shown in the figure for
clarity. The hydrophobic pocket of the active site is largely defined by
Entry Catalysts
Substrate
isomer
H
I
cis/trans
[%] [%]
1
2
3
4
5
6
[Rh
[Rh
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
(cod)
2
2
]BF
]BF
4
4
cis
trans
80.0 12.6
57.5
80.5 19.0 23.0
3.5
55.6
2.1
1.4
A( cod)
C
H
T
U
N
G
T
R
E
N
N
U
N
G
5.7
H4/10R+H17F-hCAII-[Rh] cis
H4/10R+H17F-hCAII-[Rh] trans
9*His- hCAII-[Rh]
9*His- hCAII-[Rh]
[
e]
–
cis
trans
2.7 26.4
[
e]
–
[
a] Reactions at 5 atm H
2
, room temperature, 12 h. [b] Hydrogenation.
c] Isomerization. [d] Hydrogenation conversion ratio of cis/trans by the
same catalyst. [e] Conversion was too low to determine.
[
[
20]
Val121, Val143, Leu198, and Trp209.
Chem. Eur. J. 2009, 15, 1370 – 1376
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1373

R. J. Kazlauskas et al.
optimized structures placed cis-stilbene close to the metal
enzyme classification 1.12.99: oxidoreductase acting on hy-
drogen as a donor with other acceptors. Current examples in
this classification reduce cofactors or other enzymes; none
reduce an organic substrate, so this hybrid catalyst is the
first example.
Several issues still limit the application of this catalyst to
synthesis. First, even with all the histidines blocked or re-
placed, one extra rhodium remains, which may contribute to
catalysis and lower the selectivity. Reducing the overall neg-
ative charge of the carbonic anhydrases may remove this
last extra rhodium. The modifications we made to remove
or modify histidine (partially positively charged amino acid
side chain) likely increased the overall negative charge of
the protein. This increased negative charge may help bind
the extra rhodium ion to the surface. Second, the protein
does not accelerate the hydrogenation reaction. Hydrogena-
ion (3.8 and 4.3 ꢂ between the olefinic carbons and the
[21]
zinc ), but placed trans-stilbene farther away (4.7 and
.7 ꢂ). Hydrogenation would require that the p-electron
5
cloud of the double bond face the metal center. This is the
orientation seen for cis-stilbene, but trans-stilbene orients in
a manner that is unlikely to be catalytically productive.
Manually reorienting trans-stilbene closer to metal ion with
its double bond facing the metal creates steric clashes with
Leu198, Thr200, His94/96, Asn62, and Ala64.
No stereoselectivity for cis- versus trans-3-hexen-1-ol: A
smaller substrate, 3-hexen-1-ol, showed similar conversions
in the hydrogenation of either cis or trans isomers. Further,
cis–trans isomerization accompanied the hydrogenation (see
the Supporting Information), suggesting that the rhodium–
carbonic anhydrase is not stereoselective for this olefin.
Computer modeling supported this suggestion since the
model placed both cis- and trans-3-hexen-1-ol at similar dis-
tances above, rather than in, the active site, possibly due to
only weak interactions with the hydrophobic region of the
active site.
Preliminary hydrogenation experiments with prochiral
substrates (e.g. methyl a-acetamidoacrylate or methyl a-
acetamidocinnamate) yielded <10% ee in the products. This
poor enantioselectivity may be due to the spaciousness
around the active site.
tion by [Rh ACHTGNUTRNEUNG( cod) ]BF gives a slightly higher conversion
2 4
than 9*His-hCAII-[Rh] (Table 2, entries 10 and 11) suggest-
ing that all the catalytic power comes from the rhodium.
Modification of the active site should be able to increase the
rate of hydrogenation. Third, the substrate range of the cat-
alyst has not been explored. If it proves to be limited, muta-
genesis guided by either rational design or directed evolu-
tion may expand it.
Nature also used the metal replacement approach to
create new catalytic activities. For example, the vicinal
oxygen chelate superfamily consists of proteins with similar
structure, but different metals bound at the active site to
[22]
catalyze different reactions. Glyoxylase I (zinc or nickel
in active site) and methylmalonyl-CoA epimerase (cobalt in
active site) catalyze rearrangements via acid–base reactions
involving similar enediol intermediates. In contrast, extra-
diol dioxygenases (iron or manganese in active site) catalyze
oxidative meta ring cleavage of catechol or similar sub-
strates. Fosfomycin resistance protein (manganese at active
site) catalyzes nucleophilic attack of glutathione on the ep-
oxide ring of fosfomycin.
Discussion
Removal of the active site zinc from bovine or human car-
bonic anhydrase isoenzyme II followed by addition of the
rhodium(I) salt [Rh ACHTUNGTRENNUNG( cod) ]BF gave rhodium–carbonic anhy-
2 4
drase complexes each containing 6.5–7.5 rhodium ions. We
hypothesized that one of these rhodium ions bound to the
active site and the others on the surface of the carbonic an-
hydrase. Modification of the histidine residues on the sur-
face either by site-directed mutagenesis or a combination of
site-directed mutagenesis and chemical modification re-
duced the amount of rhodium bound to approximately two.
Presumably one binds in the active site and the other on the
protein surface. This ability to minimize nonspecific binding
allowed the behavior of a metal hydrogenation catalyst to
be tested for the first time at the active site of a metalloen-
zyme.
The rhodium–carbonic anhydrase complexes with minimal
extra rhodium showed about 20:1 selectivity for cis-stilbene
over trans-stilbene. Evidence for this stereoselectivity is the
faster hydrogenation of cis-stilbene as compared to trans-
stilbene and the lower amount of isomerization of cis-stil-
bene to trans-stilbene during hydrogenation. Modeling sug-
gests that the shape of the active site (steric effects) favors
binding cis-stilbene, but not trans-stilbene, in a catalytically
productive orientation. This is the first biocatalyst that di-
rectly reduces a substrate with hydrogen gas without an
added phosphine ligand. This biocatalyst would fit into
Previous work on changing the catalytic activity of metal-
loenzymes by replacing the active site metal ion involved
only oxidation reactions. The first example was replacement
of the active-site zinc(II) in carboxypeptidase with copper-
[23]
(II) to create a slow oxidase. Recently, we and others re-
placed the active-site zinc in carbonic anhydrase with man-
[11]
ganese to create an oxidase. There are two additional bor-
derline examples. Sheldon and co-workers removed the
active site zinc from the protease thermolysin and added not
a metal ion, but an oxoanion—tungstate, molybdate, or sele-
[24]
nate. The resulting complex catalyzed oxidations, but this
experiment was unlikely a simple replacement. Zinc and the
oxoanions have opposite charge and differ dramatically in
size, so they likely bind in different places. Another border-
line example is not a replacement because the starting
enzyme did not contain a metal ion. Two groups started
with a phosphatase that contains no metal ion and added va-
[25]
nadate, a transition state analog for phosphate hydrolysis.
The resulting complexes catalyzed the enantioselective oxi-
dation of sulfides.
1374
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 1370 – 1376

Stereoselective Hydrogenation of Olefins
FULL PAPER
In the future this active-site-metal-replacement approach
may greatly expand the catalytic range of enzymes. Natural
enzymes do not contain precious metals because they are
rare in nature. Examples of interesting transition-metal-cata-
lyzed reactions that occur in water include hydroformyla-
tion, cycloaddition, olefin metathesis and Heck-type reac-
nation of stilbene was analyzed on a chiral column Chirasil-Dex CB
2
5mꢃ0.25 (oven temperature: 1508C, injector temperature 2508C, detec-
À1
tor temperature 3008C); flow rate: 1.2 mLmin ; retention time of cis-
stilbene, 5.2 min; retention time of bibenzyl, 5.5 min; retention time of
trans-stilbene 13.6 min. The details for the hydrogenation of 3-hexene-1-
ol are in the Supporting Information.
Molecular modeling: Molecular modeling was performed with Maestro
[26]
(
version 8.0, Schrçdinger, New York, NY) using the Merck molecular
tions. An important advantage of protein-based catalysts
is ease of optimization using directed evolution. Unlike tedi-
ous chemical synthesis of chemical ligands, molecular biol-
ogy methods can create thousands or millions of protein var-
iants and screening can identify those with improved proper-
ties. Carbonic anhydrases occur with three different protein
[
30]
force field (MMFF) starting with the X-ray crystal structure of zinc-
containing human carbonic anhydrase II at 2.0 ꢂ resolution, PDB file
ID: 4CAC. Even though the zinc-containing enzyme does not catalyze
[15]
hydrogenation, we did not replace it with rhodium because molecular
mechanics parameters for rhodium are limited and because we were in-
terested in the fit of the isomeric alkenes in the active site, not the metal-
olefin interaction. The dielectric constant was set to 1 to mimic the solva-
tion effects of water. Hydrogen atoms were added and His 119 was
changed to its less common tautomer because the X-ray structure shows
that this tautomer coordinates to zinc. The structure was geometry opti-
mized stepwise. First, the geometry of the hydrogen atoms on water were
optimized, next the hydrogen atoms on the protein, and then the whole
protein with water molecules, finally the water, whole protein and the
bound substrate (cis- or trans-stilbene). Each geometry optimization step
used the PolaRibiere conjugated gradient (PRCG) method with a maxi-
[27]
folds, so there are several starting points for creation of
new catalysts; in addition, other metalloenzymes may also
be good starting points.
Experimental Section
mum of 500 interations and
a
convergence threshold of
General: The Supporting Information contains details for preparation of
human carbonic anhydrase isoenzyme II and its variants. Dialysis experi-
À1
À1
0
.05 kcalꢂ mol .
ments used Spectra/Por 2 tubing with a molecular weight cut off of 12–
À1
1
4000 gmol
.
Replacement of zinc in carbonic anhydrase with rhodium: A dialysis
tube containing carbonic anhydrase (75 mg, 2.4 mmol) in buffer (0.1m 2-
Acknowledgements
(
N-morpholino)ethanesulfonic acid (MES), pH 6.0, 5.0 mL) was dialyzed
overnight against 2,6-pyridinedicarboxylic acid (2,6-PDCA, 557 mg,
.34 mmol) in sodium acetate buffer (500 mL, 0.1m, pH 5.5). The tube
was removed and dialyzed for 20 h against buffer (50 mm MES, pH 6.0,
00 mL) to remove the excess of 2,6-PDCA and yield apo(carbonic anhy-
drase). The rhodium ion was introduced by dialysis for 10 h against [Rh-
(cod) ]BF (203 mg, 0.5 mmol in 500 mL of 50 mm MES buffer, pH 6.0).
We thank Rick Knurr (Department of Geology and Geophysics) for
ICP-MS analysis, Thomas Krick and LeeAnn Higgins (Center for Mass
Spectrometry and Proteomics) for ESI-MS spectra, Carol Fierke and
Andrea Stoddard (University of Michigan) for the plasmid encoding
human carbonic anhydrase II, the University of Minnesota and General
Mills Corporation for financial support.
3
5
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
2
4
To remove the traces of unbound rhodium the dialysis tubing was dia-
lyzed twice for 2.5 h against buffer (50 mm MES, pH 6.0). The total
volume of Rh-substituted CA solution increased by about 10% (to
[
1] Reviews: a) P. M. Vignais, B. Billoud, J. Meyer, FEMS Microbiol.
Rev. 2001, 25, 455–501; b) M. Frey, ChemBioChem 2002, 3, 153–
5
.5 mL) during the entire process. During the preliminary work, we
tested RhCl , [(Cp*RhCl ] (Cp*=pentamethylcyclopentadiene), and
Rh(CO) (acac)] (acac=acetylacetonato), but found that these yielded
either lower hydrogenation or less stable protein. A ruthenium metal
precursor, [{Ru(p-cymene)Cl ], was also tested, but after dialysis against
1
60; c) The Hmd class of hydrogenases use a cofactor, but hydrogen
3
2 2
)
atoms still end up as protons in water: S. Shima, R. K. Thauer,
Chem. Rec. 2007, 7, 37–46.
[
2
ACHTUNGTRENNUNG
[
[
2] Reviews: a) D. Qi, C.-M. Tann, D. Haring, M. D. Distefano, Chem.
Rev. 2001, 101, 3081–3112; b) C. M. Thomas, T. R. Ward, Chem.
Soc. Rev. 2005, 34, 337–346; c) C. Letondor, T. R. Ward, ChemBio-
chem 2006, 7, 1845–1852.
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
2 2
}
II
buffer, most of the Ru was lost suggesting that it did not bind the pro-
tein.
Modification of human carbonic anhydrase by diethylpyrocarbonate
3] a) M. E. Wilson, G. M. Whitesides, J. Am. Chem. Soc. 1978, 100,
(
DEPC): Human carbonic anhydrase was modified by DEPC according
3
06–307; b) C. C. Lin, C. W. Lin, A. S. C. Chan, Tetrahedron: Asym-
[28]
to a literature procedure for protein modification.
DEPC solution
metry 1999, 10, 1887–1893; c) J. Collot, J. Gradinaru, N. Humbert,
M. Skander, A. Zocchi, T. R. Ward, J. Am. Chem. Soc. 2003, 125,
(
69 mL, 27.6 mmol of 400 mm DEPC in 30% ethanol) was added to a solu-
À1
tion of human carbonic anhydrase II (20 mgmL , 0.685 mmol, in 2.0 mL
of a mixture of 50 mm MES buffer and 0.10m borate buffer, pH 7) and
stirred for 20 min at room temperature. The mixture was dialyzed over-
night against MES buffer (400 mL of 50 mm, pH 6.0). ESI-MS spectra
9
030–9031; d) M. Skander, N. Humbert, J. Collot, J. Gradinaru, G.
Klein, A. Loosli, J. Sauser, A. Zocchi, F. Gilardoni, T. R. Ward, J.
Am. Chem. Soc. 2004, 126, 14411–14418; e) T. R. Ward, Chem. Eur.
J. 2005, 11, 3798–3804; f) M. Skander, C. Malan, A. Ivanova, T. R.
Ward, Chem. Commun. 2005, 4815–4817; g) C. Letondor, N. Hum-
bert, T. R. Ward, Proc. Natl. Acad. Sci. USA 2005, 102, 4683–4687;
h) L. Panella, J. Broos, J. Jin, M. W. Fraaije, D. B. Janssen, M. Jeroni-
mus-Stratingh, B. L. Feringa, A. J. Minnaard, J. G. de Vries, Chem.
Commun. 2005, 5656–5658; i) C. Letondor, A. Pordea, N. Humbert,
A. Ivanova, S. Mazurek, M. Novic, T. R. Ward, J. Am. Chem. Soc.
2006, 128, 8320–8328; j) G. Klein, N. Humbert, J. Gradinaru, A. Iva-
nova, F. Gilardoni, U. E. Rusbandi, T. R. Ward, Angew. Chem. 2005,
117, 7942–7945; Angew. Chem. Int. Ed. 2005, 44, 7764–7767.
(
see Supporting Information, Figure S2) showed a mixture of proteins
with increased molecular weight consistent with addition of 8–12 carbox-
yethoxyl moieties.
[
29]
Typical hydrogenation procedure:
Metal-substituted carbonic anhy-
drase (200 mL of 0.225 mmolmL , 0.045 mmol) followed by substrate
18 mL of 0.25m in CH CN, 4.5 mmol) was added to a 5 mL vial contain-
À1
(
3
ing a stirring bar (7 mm) and MES buffer (200 mL, 50 mm, pH 6.0). The
vial was placed in an autoclave; the autoclave was sealed, filled with hy-
drogen to 7 atm and then released. The fill–release was repeated twice
more and the hydrogen pressure was adjusted to 1 or 5 atm. The reaction
was stirred for 12 h, then extracted with ethyl acetate and the extract an-
alyzed by gas chromatography. This procedure was also used for hydroge-
nation of trans-stilbene, cis- or trans-3-hexene-1-ol and hydrogenation
[4] H. Yamaguchi, T. Hirano, H. Kiminami, D. Taura, A. Harada, Org.
Biomol. Chem. 2006, 4, 3571–3573.
[5] a) B. P. Espꢄsito, E. de Oliveira, S. B. Zyngier, R. Najjar, J. Braz.
Chem. Soc. 2000, 11, 447–452; b) B. P. Espꢄsito, A. Faljoni-Alꢅrio,
J. F. S. de Menezes, H. F. de Brito, R. Najjar, J. Inorg. Biochem.
2 4
catalyzed by [Rh ACHTUNGTRENNUNG( cod) ]BF (in 50 mm MES buffer, pH 6.0). The hydroge-
Chem. Eur. J. 2009, 15, 1370 – 1376
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
1375

R. J. Kazlauskas et al.
1
999, 75, 55–61; c) the interest in the binding of these complexes to
[17] K. L. Gudiksen, I. Gitlin, J. Yang, A. R. Urbach, D. T. Moustakas,
G. M. Whitesides, J. Am. Chem. Soc. 2004, 126, 4707–4714.
[18] Wild-type hCAII and several variants precipitated after treatment
with DEPC or after subsequent exchange of the zinc with rhodium.
The triple mutant H4/10R+H17F was an exception since it retained
90% activity in the hydrolysis of 4-nitrophenyl acetate after treat-
ment with DEPC and remained soluble after exchange of the zinc
with rhodium.
proteins stems from reports of their anti-cancer activity: J. L. Bear,
H. B. Gray, L. Rainen, I. M. Chang, R. A. Howard, G. Serio, A. P.
Kimball, Cancer Chemother. Rep. 1975, 159, 611–620.
[
6] a) L. Trynda, F. Pruchnik, J. Inorg. Biochem. 1995, 58, 69–77; b) L.
Trynda-Lemiesz, F. Pruchnik, J. Inorg. Biochem. 1997, 66, 187–192.
7] C. Bertucci, C. Botteghi, D. Giunta, M. Marchetti, S. Paganelli, Adv.
Synth. Catal. 2002, 344, 556–562; a preliminary communication
from this group (M. Marchetti, G. Mangano, S. Paganelli, C. Botte-
ghi, Tetrahedron Lett. 2000, 41, 3717–3720) mentioned enantioselec-
tivity for one substrate and selective hydroformylation of smaller al-
dehydes over larger ones, but this full paper did not confirm the pre-
liminary report.
[
[19] Reaction rates needed to determine K
M
and kcat are difficult to
measure accurately with the high pressure reaction chambers. For
this reason, we focused on selectivity where we measure the relative,
not absolute, amounts of products.
[20] R. S. Alexander, S. K. Nair, D. W. Christianson, Biochemistry 1991,
30, 11064–11072.
[
8] T. Ueno, M. Suzuki, T. Goto, T. Matsumoto, K. Nagayama, Y. Wata-
nabe, Angew. Chem. 2004, 116, 2581–2584; Angew. Chem. Int. Ed.
[21] X-ray structures of rhodium(I)–olefin complexes show RhÀC distan-
2
004, 43, 2527–2530.
ces of approximately 2 ꢂ: J. A. Evans, D. R. Russell, J. Chem. Soc.
D 1971, 197–198.
[
9] Zinc(II) has a radius of 0.74 ꢂ, while rhodiumACHTUNGTRENNUNG
0
1
[22] R. N. Armstrong, Biochemistry 2000, 39, 13625–13632.
[23] K. Yamamura, E. T. Kaiser, J. Chem. Soc. Chem. Commun. 1976,
830–831.
[24] M. Bakker, F. van Rantwijk, R. A. Sheldon, Can. J. Chem. 2002, 80,
622–625.
likely similar to that for zinc(II).
10] Review: D. W. Christianson, J. D. Cox, Annu. Rev. Biochem. 1999,
8, 33–57.
[
[
6
11] a) K. Okrasa, R. J. Kazlauskas, Chem. Eur. J. 2006, 12, 1587–1596;
b) A. Fernꢅndez-Gacio, A. Codina, J. Fastrez, O. Raint, P. Soumil-
lon, ChemBioChem 2006, 7, 1013–1016.
[25] a) N. Tanaka, V. Dumay, Q. Liao, A. J. Lange, R. Wever, Eur. J. Bio-
chem. 2002, 269, 2162–2167; b) F. van de Velde, L. Kçnemann, F.
van Rantwijk, R. A. Sheldon, Chem. Commun. 1998, 1891–1892;
c) F. van de Velde, L. Kçnemann, F. van Rantwijk, R. A. Sheldon,
Biotechnol. Bioeng. 2000, 67, 87–96.
[
12] a) R. G. Wilkins, K. R. Williams, J. Am. Chem. Soc. 1974, 96, 2241–
2
245; b) Y. Kidani, J. Hirose, H. Koike, J. Biochem. 1976, 79, 43–51;
c) Y. Kidani, J. Hirose, J. Biochem. 1977, 81, 1383–1391.
13] S. K. Nair, T. L. Calderone, D. W. Christianson, C. A. Fierke, J. Biol.
Chem. 1991, 266, 17320–17325.
14] a) M. Bradford, Anal. Biochem. 1976, 72, 248–254. b) We did not
observe a color characteristic of the hybrid enzyme. Both [Rh-
[26] Aqueous-Phase Organometallic Catalysis, 2nd ed. (Eds: B. Cornils,
W. A. Herrmann), Wiley-VCH, Weinheim, 2004.
[27] A. Liljas, M. Laurberg, EMBO Rep. 2000, 1, 16–17.
[28] W. Asmara, U. Murdiyatmo, A. J. Baines, A. T. Bull, D. J. Hardman,
Biochem. J. 1993, 292, 69–74.
[
[
A
C
H
T
U
N
G
T
R
E
N
N
U
N
G
(cod)
2
]BF
4
and bipyridine-[Rh] solutions are pale yellow and have
[29] Pressure is not the cause of the unstability since the protein precipi-
tates over several hours even under 1 atm of hydrogen.
[30] a) T. A. Halgren, J. Comput. Chem. 1996, 17, 490–519; T. A. Halg-
ren, J. Comput. Chem. 1996, 17, 520–552; T. A. Halgren, J. Comput.
Chem. 1996, 17, 553–586; b) T. A. Halgren, R. B. Nachbar, J.
Comput. Chem. 1996, 17, 587–615.
indistinguishable UV/Vis spectra. Similarly, the hCAII 9*His-[Rh]
complex shows a similar spectrum with an additional strong absorp-
tion at 280 nm due to protein.
[
[
15] X-ray structure: A. E. Eriksson, P. M. Kylsten, T. A. Jones, A. Liljas,
Proteins 1988, 4, 283–293. The Supporting Information includes a
graphic showing the location of these histidines.
16] Bovine CA II has no cysteine residues, but the human enzyme has
one at position 206, which is buried. Site-directed mutagenesis sug-
gested that it does not bind rhodium.
Received: August 12, 2008
Published online: December 29, 2008
1376
www.chemeurj.org
ꢁ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2009, 15, 1370 – 1376