Chemomimetic Biocatalysis: Exploiting the Synthetic Potential of
Cofactor-Dependent Enzymes To Create New Catalysts
Christopher K. Prier and Frances H. Arnold
*
Division of Chemistry and Chemical Engineering, California Institute of Technology, 1200 East California Boulevard, MC 210-41,
Pasadena, California 91125, United States
ABSTRACT:
Despite the astonishing breadth of enzymes
in nature, no enzymes are known for many of the valuable
catalytic transformations discovered by chemists. Recent
work in enzyme design and evolution, however, gives us
good reason to think that this will change. We describe a
chemomimetic biocatalysis approach that draws from
small-molecule catalysis and synthetic chemistry, enzymol-
ogy, and molecular evolution to discover or create
enzymes with non-natural reactivities. We illustrate how
cofactor-dependent enzymes can be exploited to promote
reactions
fi
rst established with related chemical catalysts.
The cofactors can be biological, or they can be non-
biological to further expand catalytic possibilities. The
ability of enzymes to amplify and precisely control the
reactivity of their cofactors together with the ability to
optimize non-natural reactivity by directed evolution
promises to yield exceptional catalysts for challenging
transformations that have no biological counterparts.
■
INTRODUCTION
Challenges in catalysis demand the creation of enzymes with
activities not yet found in the biological world. Nature has
evolved a certain set of synthetic strategies and uses an
impressive array of enzyme catalysts to construct everything
from simple metabolites to complex natural products. For the
production of medicinal compounds, fuels, materials, or
chemicals,
1
however, nature
’
s synthetic strategies may not be
ideal or even appropriate. Toward this end, one might desire
enzymes that act on nonbiological functional groups or
promote non-natural bond constructions while still capitalizing
on enzymes
’
extraordinary powers of rate acceleration and
selectivity. Developing genetically encoded catalysts for non-
natural chemical transformations will expand the reach of
biocatalysis and facilitate construction of biocatalytic routes for
the synthesis of valuable chemical products
in vitro
and
in vivo
.
2
Enzymes constructed only of the 20 canonical amino acids
catalyze a remarkable range of chemistries. To achieve certain
types of activity, however, proteins are often augmented with
organic metabolites or metal ions known as cofactors; these
species have functional groups and properties that enable the
protein
−
cofactor complex to catalyze reactions that the protein
alone cannot.
3
In turn, the protein sequence plays a critical role
in controlling and amplifying the reactivity of the cofactor,
enabling the ensemble to e
ff
ect transformations that the
cofactor often cannot perform alone or dictating the regio-,
diastereo-, or enantioselectivity of those transformations.
Furthermore, a given cofactor can often catalyze a multitude
of chemically diverse transformations, and the protein structure
acts to guide reactivity down one out of many possible
pathways.
Many cofactor-dependent enzymes have been studied in
depth with regard to their reaction mechanisms and the
complex interactio
ns between protein and cofactor that
promote catalysis. The enzymologists carrying out these studies
almost always focus on the natural function and substrate(s). At
the same time, synthetic chemists have developed catalysts for a
broad range of reactions that are completely absent in biology,
either because nature has not found it advantageous to use
them or because they require reagents not normally found in
biology. In many cases, the small-molecule catalysts resemble
natural cofactors; sometimes their creation was inspired by
enzymes, in a biomimetic chemistry approach to catalyst
design.
4
Similarities between many chemical catalysts and
natural cofactors, both structural and functional, raise the
possibility that the non-natural activities of small-molecule
catalysts can be translated back into the corresponding
cofactor-dependent enzymes. As proteins can provide exquisite
control over reaction pathways, this chemomimetic biology
strategy can improve on the e
ffi
ciencies and selectivities of
small-molecule catalysts just as natural enzymes improve on the
activities and selectivities of their cofactors (
Figure 1
).
This Perspective will demonstrate how a chemomimetic
approach can generate new biocatalysts from existing cofactor-
dependent enzymes. Other approaches, including catalytic
antibodies
5
and
de novo
designed enzymes,
6
have also delivered
biocatalysts that catalyze reactions not known in nature.
However, despite extensive e
ff
orts, their reactions have been
limited to a relatively narrow set of transformations, and most
of the new enzymes do not catalyze reactions at useful rates. In
contrast, by repurposing existing cofactor-dependent enzymes
for new chemistry, protein engineers and chemists have created
enzymes that execute a diverse range of synthetically
challenging nonbiological reactions. Cofactors enable the
generation of unique reactive intermediates in enzyme active
sites, whereas synthetic chemistry serves as a guide for the types
of activity that can be achieved with a given reactive motif, even
if they have not been observed in nature. This Perspective is
not intended as an extensive review of the literature but, rather,
a discussion of case studies that illustrate how new activities
may be introduced into existing enzymes. We will discuss novel
activities for enzymes that use thiamine and heme cofactors as
well as proteins that use natural amino acids for non-native
aminocatalysis.
7
We will also brie
fl
y consider the introduction
Received:
September 3, 2015
Published:
October 26, 2015
Perspective
pubs.acs.org/JACS
© 2015 American Chemical Society
13992
DOI: 10.1021/jacs.5b09348
J.Am.Chem.Soc.
2015, 137, 13992
−
14006
This is an open access article published under an ACS AuthorChoice License, which permits
copying and redistribution of the article or any adaptations for non-commercial purposes.
of arti
fi
cial cofactors into proteins to create new enzymes.
Throughout, we will focus on reactions that are not biological
rather than on creating new catalysts for transformations that
are already known in nature, for which there is ample
precedent.
8
Interest in engineering non-natural enzyme activity is
growing rapidly. Apart from better understanding of enzyme
structures and mechanisms, one key driving force has been
ready access to powerful methods of protein engineering. As
natural enzymes are rarely pro
fi
cient at performing non-natural
chemistries, optimization of the protein structure is required to
access synthetically useful catalysts; this is now possible and is
in fact relatively straightforward. A hallmark of enzymes is that
they can evolve and adapt under selective pressure, and this
evolvability can be exploited in the laboratory to optimize
enzymes via an iterative process of mutagenesis and screening
for a desired outcome. This engineering approach, known as
directed evolution, enables rapid tuning of key catalyst features
such as selectivity, activity, and stability and circumvents our
still poor understanding of how sequence a
ff
ects enzyme
function.
9
Thus, once a small amount of activity for a given
transformation is discovered, the activity can often be greatly
improved by introducing one or a few mutations at a time.
Although mutagenesis guided by mechanistic understanding
can sometimes be a successful approach to improving enzyme
activities, the creation of exceptional catalysts almost always
relies (at least in part) on a wider exploration of protein
sequence space. In the realm of cofactor-dependent enzymes,
the manner in which protein structure impacts the inherent
reactivity of the cofactor provides fertile ground for protein
engineers to alter the course of chemistry just by mutation of
the protein sequence. In the world of chemical catalysis, there is
no general strategy equivalent to evolution for optimizing
catalyst structure. Catalyst modi
fi
cation often requires laborious
resynthesis (as opposed to relatively straightforward gene
modi
fi
cation), and subtle bene
fi
cial structural mutations rarely
accumulate over generations of small-molecule catalyst
optimization.
The creation of enzymes for nonbiological processes is an
emerging
fi
eld full of promise at the interface between
chemistry and biology. Opportunities abound for protein
engineers to exploit the wealth of knowledge gained from
mechanistic enzymology and synthetic chemistry. Our goal here
is to introduce concepts in this
fi
eld and point to some of the
opportunities; we encourage chemists to look at enzymes in a
new way and contribute their intuition and insights to creating
enzymes with new, synthetically useful activities.
■
THIAMINE-DEPENDENT ENZYMES
Chemists and protein engineers have developed new reactions
using enzymes dependent on thiamine diphosphate (ThDP) by
taking advantage of the unusual catalytic mechanisms enabled
by this cofactor. Thiamine diphosphate comprises an
N
-alkyl
thiazolium core, a tethered pyrimidine ring, and a diphosphate-
terminated side chain (
Scheme 1
). The thiamine cofactor is
noncovalently bound to the enzyme, with the diphosphate
group binding a second metal cofactor (typically, magnesium).
Enzymes containing ThDP possess the unique ability to forge
or break C
−
C bonds between two oxidized carbon centers;
these enzymes catalyze decarboxylations (as in pyruvate
decarboxylase), carboligation
s (as in transketolases), and
oxidative transformations (as in pyruvate dehydrogenase).
10
Pyruvate decarboxylase (PDC), a particularly well-studied
thiamine-dependent enzyme,
catalyzes the conversion of
pyruvate to acetaldehyde and carbon dioxide (
Scheme 1
).
Early studies demonstrated that ThDP can perform certain
functions of thiamine-dependent enzymes, indicating that the
cofactor itself contains all of the functionality required for
catalysis.
11
It was not until 1957, however, that the now-
accepted mechanism of thiamine catalysis was put forward by
Breslow, who provided evidence for a mechanism involving
deprotonation of the thiazolium ring at C2 (p
K
a
≈
18)
12
to give
the thiazolium ylide
1
(
Scheme 2
).
13
A resonance form of this
ylide is the nucleophilic carbene
2
, in which a carbon atom
bearing a sextet of electrons is stabilized by
σ
-electron
withdrawal and
π
-electron donation from the adjacent
heteroatoms. In the mechanism of pyruvate decarboxylation
by PDC, addition of the nucleophilic carbene carbon (C2) to
the pyruvate keto-group gives the covalent adduct
3
. The
electron-withdrawing nature of the thiazolium ring then
facilitates decarboxylation to yield the enaminol moiety
4
,
known as the Breslow intermediate. This species is strongly
nucleophilic; it may undergo protonation to provide thiazolium
Figure 1.
Chemomimetic strategies guide the creation of biocatalysts
for reactions not known in nature.
Scheme 1. Thiamine Diphosphate and the Native Activity of
Thiamine-Dependent Pyruvate Decarboxylase (PDC)
10
Journal of the American Chemical Society
Perspective
DOI: 10.1021/jacs.5b09348
J.Am.Chem.Soc.
2015, 137, 13992
−
14006
13993
5
, which then collapses to release the product acetaldehyde and
regenerate the active carbene form of the cofactor.
Whereas the cofactor alone decarboxylates pyruvate, the
protein dramatically accelerates catalysis: by up to a factor of
10
12
for yeast PDC.
14
Various mechanisms have been
implicated in this rate acceleration. Conserved polar
interactions with the pyrimidine ring (provided by E51 and
G413 in PDC) increase the basicity of the 4
′
-nitrogen, whereas
a conserved hydrophobic residue (I415 in yeast PDC) acts as a
fulcrum between the two heteroaromatic rings, enforcing a V-
shaped conformation of the cofactor and positioning the 4
′
-
nitrogen in a favorable orientation to perform an intramolecular
deprotonation of C2 (
Scheme 2
).
15
These active site features
e
ff
ectively lower the p
K
a
at C2 and accelerate deprotonation of
the thiazolium.
16
Next, upon formation of the covalent adduct
3
, studies on other ThDP-dependent enzymes suggest that the
protein promotes decarboxylation by enforcing maximal orbital
overlap between the scissile C
−
C bond and the thiazolium
π
-
system in the decarboxylation transition state.
17
Finally, speci
fi
c
residues have been implicated in both protonation of the
Breslow intermediate and deprotonation of the alcoholic
proton in the acetaldehyde-forming step.
10
The protein
sequence of PDC thus functions to accelerate many steps
throughout the catalytic cycle.
Concurrent with the enzymology studies delineating the
mode of action of ThDP-dependent enzymes, chemists
examining small molecules related to thiamine found that a
broad range of heterocyclic structures (such as those shown in
Scheme 3
) undergo deprotonation to yield nucleophilic
carbenes analogous to intermediate
2
. These species, termed
N-heterocyclic carbenes (NHCs), have found broad application
in catalysis; similar to the mechanism of thiamine catalysis,
these carbenes condense with carbonyl compounds to give
Breslow intermediates.
19
This activation mode is particularly
useful, as it reverses the typical reactivity of a carbonyl group:
carbonyl carbon atoms are typically electrophilic, but formation
of the Breslow intermediate (an acyl anion equivalent) renders
the carbonyl carbon nucleophilic, enabling unique bond
constructions. Using this catalytic manifold, chemists have
accomplished many reactions that are not known to be
catalyzed by thiamine-dependent enzymes in nature; these
include benzoin and aza-benzoin condensations as well as
Stetter, hydroacylation, and various annulation reactions.
19
Of
the various catalyst systems that have been developed,
triazolium-derived N-heterocyclic carbenes have been found
to be especially useful for achieving enantioselective trans-
formations.
20
Recognizing the similarities between enzyme and NHC
catalysts, biochemists and protein engineers have sought to use
thiamine-dependent enzymes to perform some of these
nonbiological synthetic transformations.
21
Many ThDP-de-
pendent enzymes have been found to promote benzoin-type
condensations between two aldehydes; in the catalytic
mechanism, generation of the Breslow intermediate
6
is
achieved via
α
-deprotonation of the initial covalent adduct
rather than
α
-decarboxylation, as in the mechanism of PDC
(
Scheme 4
A).
22
Nucleophilic addition to a second equivalent of
aldehyde followed by expulsion of thiamine then gives the
α
-
hydroxyketone
7
. Various other enzymatic heterocouplings
between aromatic and aliphatic aldehydes have also been
established.
23
Particularly notable applications of ThDP-dependent en-
zymes have been found for reactions that have historically
proven to be challenging for their small-molecule equivalents.
In one such transformation, the asymmetric cross-benzoin
reaction, one aldehyde must exclusively react with the carbene
while a second (chemically very similar) aldehyde must
function only as an acceptor. Due to this chemoselectivity
problem, small-molecule methods are typically limited to
aldehyde homocoupling. Mu
̈
ller and co-workers identi
fi
ed
two thiamine-dependent enzymes, benzaldehyde lyase (BAL)
from
Pseudomonas
fl
uorescens
and a variant of benzoylformate
decarboxylase (BFD) from
Pseudomonas putida
, that success-
fully execute the reaction.
24
Several aldehydes bearing ortho-
substituents (such as 2-chlorobenzaldehyde,
8
) were selected as
Scheme 2. Mechanism of Thiamine Catalysis in Pyruvate
Decarboxylase and Active Site Architecture of Pyruvate
Decarboxylase from
Saccharomyces cerevisiae
10
,
a
a
ThDP is shown in gray, active site residues are in yellow, and
magnesium is in green; PDB: 1PVD.
18
Scheme 3. Small-Molecule Thiamine Equivalents (N-
Heterocyclic Carbene Precursors) Used To Catalyze Diverse
Organic Transformations
19
Journal of the American Chemical Society
Perspective
DOI: 10.1021/jacs.5b09348
J.Am.Chem.Soc.
2015, 137, 13992
−
14006
13994
acceptor substrates because they do not undergo enzyme-
catalyzed homocoupling, suggesting that their condensation
with the thiamine cofactor is not possible. Thus, for the
example shown in
Scheme 4
, only benzaldehyde
9
reacts with
the carbene to give the Breslow intermediate
10
; subsequent
addition to the favored acceptor
8
yields the cross-benzoin
adduct with high chemo- and enantioselectivity. Using 2-
chlorobenzaldehyde as the acceptor, selective cross-benzoin
couplings may be achieved with a range of electronically diverse
donor aldehydes.
Directed evolution of thiamine-dependent enzymes has been
performed, enabled by a colorimetric high-throughput screen
for the detection of
α
-hydroxyketone products. These e
ff
orts
have enhanced the activities and enantioselectivities, as well as
expanded the substrate scope, of enzymes performing benzoin-
type condensations.
25
Instead of adding to carbonyls in a 1,2-fashion, the Breslow
intermediate may also add in a 1,4-conjugate fashion to
α
,
β
-
unsaturated carbonyl compounds; the resulting transformation
is known as the Stetter reaction. Some of the earliest reported
Stetter reactions were actually achieved using thiazolium
11
,
which features the core of the thiamine cofactor (
Scheme 5
).
26
However, achieving intermolecular, enantioselective Stetter
reactions has been very challenging for small-molecule NHC
catalysts; although exampl
es of this activity have been
developed, signi
fi
cant limitations remain with regard to scope
and enantioselectivity.
27
Dresen et al. demonstrated that the
thiamine-dependent enzyme PigD from
Serratia marcescens
performs the intermolecular Stetter coupling of an acetaldehyde
unit (derived from pyruvate) with enones (
Scheme 5
).
28
This
enzyme was postulated to perform a Stetter reaction as its
native function in the biosynthesis of prodigiosin,
29
but
experiments conducted with puri
fi
ed PigD and the proposed
enal substrates for such reactions provided only the products of
1,2-addition. Upon evaluation of enone substrates, however,
Stetter activity was observed and no 1,2-addition could be
detected. A variety of enones having aliphatic, aromatic, and
heteroaromatic functionalities at the 4-position undergo the
PigD-catalyzed Stetter reaction, in many cases with excellent
enantioselectivity.
30
Subsequent studies identi
fi
ed two PigD
homologues that also display
“
Stetterase
”
activity.
31
Benzaldehyde lyase (BAL) has also been engineered to
perform the formose reaction, in which dihydroxyacetone is
produced from the condensation of three equivalents of
formaldehyde. Whereas chemical catalysts, including thiazolium
salts, are known to perform this transformation, the reaction is
not known in biology. Siegel et al. used computational design
and directed evolution to identify a variant having seven
mutations and 100-fold improved
“
formolase
”
activity relative
to that of BAL.
32
This variant was used in a biosynthetic
pathway for the conversion of formate into three-carbon
metabolites.
In all of these examples, a mechanistic understanding of
thiamine catalysis combined with experience from the synthetic
chemistry of small-molecule thiamine analogues guided the
discovery of new enzyme activities. By mimicking reactions
fi
rst
achieved with chemical catalysts related to the thiamine
cofactor, native enzymes could be used to perform desired
non-natural functions and, notably, even provide solutions to
long-standing synthetic challenges. These activities proceed via
the same key intermediate (the Breslow intermediate) as that in
the natural enzyme transformations, but they access or utilize
the intermediate in a nonbiological manner via selection of
appropriate chemical reagents.
■
HEME-DEPENDENT ENZYMES
Heme-containing enzymes have been useful starting points for
new enzyme activities. Particularly versatile are the cytochrome
P450s (CYPs), a remarkable class of iron porphyrin-dependent
enzymes that participate in xenobiotic metabolism and natural
product biosynthesis.
33
These enzymes activate dioxygen, in a
process requiring two electrons from NAD(P)H, to perform a
multitude of oxygenation reactions including C
−
H hydrox-
ylation, epoxidation, sulfoxi
dation, and heteroatom deal-
kylation. The mechanism of P450-catalyzed hydroxylation
proceeds via a series of distinct iron intermediates to achieve
Scheme 4. Benzoin (A) and Cross-Benzoin (B) Reactions
Promoted by Thiamine-Dependent Benzoylformate
Decarboxylase (BFD)
22
−
24
Scheme 5. Thiamine-Dependent Enzyme PigD Performs the
Intermolecular Stetter Reaction
28
Journal of the American Chemical Society
Perspective
DOI: 10.1021/jacs.5b09348
J.Am.Chem.Soc.
2015, 137, 13992
−
14006
13995
the insertion of an atom from dioxygen into a C
−
H bond.
34
First, substrate binding displaces a water ligand of the resting
ferric state of the cofactor, giving the penta-coordinate
intermediate
12
(
Scheme 6
). This event induces a transition
of the heme iron from low to high spin, increasing its reduction
potential (by 140 mV for P450
BM3
)
35
and triggering electron
transfer from a reductase partner. Upon reduction of the
cofactor to its ferrous state (
13
), molecular oxygen binds to
give a ferric
−
superoxide complex
14
; a subsequent second
electron transfer followed by protonation then delivers the
iron
−
hydroperoxy species
15
. This species, termed Compound
0, is protonated to release water and form a key iron(IV)
−
oxo
porphyrin radical cation intermediate termed Compound I.
This intermediate is the species that performs most of the
oxygenation chemistry characteristic of P450s; in hydroxylation,
Compound I abstracts a hydrogen atom from the substrate to
generate an organic radical as well as the iron(IV)
−
hydroxyl
species, Compound II. Radical rebound delivers the oxygenated
product and returns the cofactor to its ferric resting state.
Protein engineers have targeted a range of P450s for
applications in biocatalysis, but one of the most widely used
is P450
BM3
from
Bacillus megaterium
(CYP102A1). This soluble
protein contains heme and di
fl
avin reductase domains fused in
a single polypeptide chain and naturally performs the
subterminal hydroxylation of long chain fatty acids.
36
Although all of the catalytic intermediates in the cytochrome
P450 monooxygenation cycle are heme-bound, the protein
’
s
primary sequence makes critical contributions to catalysis. In all
P450s, the heme iron is ligated by an axial cysteine thiolate
residue (C400 in P450
BM3
). Coordination by an electron-rich
ligand decreases the reduction potential of the cofactor,
preventing initiation of the catalytic cycle in the absence of
substrate. The thiolate ligand is also postulated to promote
heterolytic cleavage of the O
−
O bond in the iron
−
hydroperoxy
intermediate
15
.
34
Furthermore, as demonstrated by recent
studies,
37
thiolate ligation increases the basicity (p
K
a
)of
Compound II, causing Compound I to favor abstraction of a
hydrogen atom from the substrate over single-electron
oxidation events that would be destructive to the protein.
Another highly conserved residue is an active-site threonine
(T268 in P450
BM3
) that has been implicated in protonation and
stabilization of heme-bound intermediates through active-site
water molecules.
38
Protonation of the iron
−
hydroperoxy
intermediate
15
mediated by this threonine likely promotes
heterolytic O
−
O bond scission in the generation of Compound
I.
Synthetic chemists have long sought to replicate the
remarkable reactivity of cytochrome P450s, and many small
molecules have been developed that mimic their oxene transfer
activity; some of these catal
ysts are metalloporphyrin
complexes structurally analogous to the native heme cofactor.
40
At the same time, many metalloporphyrins perform reactions
unknown in biology; prominent among these is the transfer of
carbenes and nitrenes to organic substrates.
41
In such reactions,
an activated chemical precursor such as a diazo or azido species
reacts with a transition metal (typically, Ru, Rh, Cu, Fe, Co, or
Mn) to give a metal carbenoid or metal nitrenoid, respectively
(
Scheme 7
). These electrophilic species, electronically analo-
gous to the iron(IV)
−
oxo intermediate Compound I, may
subsequently transfer the carbene or nitrene to an organic
substrate. This mechanism has been employed to achieve a
number of challenging non-natural transformations including
cyclopropanation, C
−
H alkylation, and C
−
H amination.
42
Whereas porphyrin complexes are often highly active toward
many of these reactions, they are typically not highly
enantioselective. Instead, a number of nonporphyrin chiral
sca
ff
olds, such as copper bis(oxazolines) and dirhodium
carboxylates or carboxamidates, have been more broadly useful
for asymmetric catalysis (
Scheme 7
).
Several years ago, our group demonstrated that cytochrome
P450
BM3
can, in fact, perform the cyclopropanation of styrenes
via nonbiological metal carbenoid intermediates.
43
In this mode
of catalysis, sodium dithionite (Na
2
S
2
O
4
)
fi
rst converts the
resting ferric state of P450
BM3
to an active ferrous state
(
Scheme 8
). Reaction with the reagent ethyl diazoacetate
(EDA,
16
) then yields the iron carbenoid
17
with concomitant
Scheme 6. Mechanism of Monooxygenation Catalyzed by
Cytochrome P450s and Active Site Structure of Cytochrome
P450
BM3
Bound to
N
-Palmitoylglycine
34
,
a
a
The heme is shown in gray,
N
-palmitoylglycine is in green, active site
residues are in yellow, and iron is in orange; PDB: 1JPZ.
39
Journal of the American Chemical Society
Perspective
DOI: 10.1021/jacs.5b09348
J.Am.Chem.Soc.
2015, 137, 13992
−
14006
13996