Directed Evolution To Produce
Although protocols for improving enzyme
enantioselectivity are not standardized, changes to increase flexibility
work on a bacterial lipase
Karl-Erich Jaeger, Manfred T. Reetz, and Bauke W.
Enzymes direct biochemical reactions in all living
cells, enabling organisms to live and reproduce. Enzymes ordinarily work
under mild conditionstypically, in an aqueous medium at room
temperature, atmospheric pressure, and near neutral pH. They also
typically exhibit high substrate specificity, including regio- and
enantioselectivity. These properties make enzymes candidates for
catalyzing not only specific biochemical reactions in living cells but
also for replacing classical chemical catalysts as part of the movement
toward an industry based on "green chemistry," one in which
large-scale production processes might produce smaller amounts of toxic
wastes and thus would be easier on our environment.
However, the production demands of the chemical industry
do not readily coincide with the properties of naturally occurring
biocatalysts that evolved to fulfill tasks inside living cells.
Therefore, apart from screening diverse cells for biocatalysts with
desired properties, researchers are designing strategies, based in part
on determining enzyme crystal structures and also identifying and then
modifying essential amino acid residues, to obtain enzymes with
Recognizing that these methods can be time-consuming and
cumbersome, researchers are also using an alternative process, called
"directed evolution," as a means to develop biocatalyts with
specific, sought-after properties. This process resembles natural
evolution, and involves improving a particular biocatalyst by following
several discrete steps: (i) a gene encoding a biocatalyst of interest is
randomly mutagenized, thereby creating a library of mutant genes; (ii)
these mutant genes are (preferably over-) expressed to generate a
library of variant catalytic proteins; (iii) a variant exhibiting the
desired property is identified by screening or selection; and (iv) the
gene encoding the best variant is chosen for another round of
optimization. Frances Arnold and her collaborators at California
Institute of Technology in Pasadena, Calif., were among the first to
apply directed evolution to the creation of improved biocatalysts. More
recently, Pim Stemmer of Maxygen in Redwood City, Calif., developed a
method called DNA shuffling which is designed for the in vitro
recombination of homologous genes. This method eliminates deleterious
mutations while combining beneficial ones.
Lipases Chosen as Model Biocatalysts
About a century ago, the chemists Jokichi Takamine in
Japan and Otto Rohm in Germany began using specific enzymes, namely
a-amylases and proteases, as catalysts for reactions outside cells.
Since then, enzyme biotechnology has expanded considerably, with enzymes
now used in industrial-scale processes for synthesizing fine chemicals,
pharmaceuticals, and agrochemicals. In addition, enzymes are added to
many detergents, and they are used in many applications in the food,
leather, and paper industries. Examples of high-yield reactions
catalyzed by bacterial enzymes include production of 6-aminopenicillanic
acid (40,000 tons/year) by an amidase from Escherichia coli,
isomerisation of D-glucose to D-fructose (100,000 ton/year) by glucose
isomerase from Streptomyces, and synthesis of acrylamide from
acrylonitrile (30,000 tons/year) by a nitrile hydratase from Pseudomonas
On a laboratory scale, lipases (triacylglycerol ester
hydrolases, EC 126.96.36.199) are widely used by organic chemists. They are
stable in aqueous media as well as in organic solvents, exhibit broad
substrate specificity and selectivity, and a large number is
commercially available. Many of these lipases are obtained from bacteria
Because of their usefulness, versatility, and ready
availability, we decided to focus on a bacterial lipase as a model
biocatalyst to test the potential of directed evolution. Before starting
our experiments, we needed to identify which property of this catalyst
was biotechnologically relevant and thus worth modifying.
The Relevant Property
The stereoselective synthesis of chiral organic
compounds is of immense academic and industrial interest. The "chiral
market" for enantiomerically pure or enriched organic compounds
continues to expand rapidly, and total sales of chiral pharmaceuticals
alone exceeded $100 billion in 2000. Organic chemists currently prepare
many of these compounds by asymmetric catalysis, traditionally a process
that depends on use of one or another transition metal catalyst.
Considerable experience, intuition, knowledge of reaction mechanisms,
molecular modeling, and time-consuming trial-and-error experiments are
required to develop efficient, chiral, transition-metal catalysts.
However, success in this field is highly appreciated as demonstrated by
the fact that the 2001 Nobel prize in chemistry was awarded to William
S. Knowles, Ryoji Noyori, and K. Barry Sharpless for the development of
chiral catalysts and their industrial applications.
Although biocatalysts represent an attractive
alternative, how can chiral biocatalysts be obtained? In some cases,
naturally occurring biocatalysts are being used to produce certain
chiral intermediates. For example, a lipase from Serratia marcescens is
used for making an epoxy-ester which is an intermediate in the synthesis
of the antihypertensive therapeutic Dilthiazem®.
However, the enantioselectivity of biocatalysts often
may be poor for a synthetic transformation of interest, A ? B.
Does this mean that biocatalysis is not a method to be seriously
considered? With that question in mind, our three research teams joined
forces to evaluate the potential of directed evolution to create
enantioselective biocatalysts. The three teams consisted of molecular
microbiologists, at that time located at the Ruhr-Universität Bochum,
who provided appropriate lipase genes and the corresponding expression
and secretion systems; organic chemists at the nearby Max-Planck-Institut
für Kohlenforschung in Mülheim an der Ruhr, who developed
high-throughput screening methods for enantioselective biocatalysts and
applied them to identify lipase variants with improved
enantioselectivity; and crystallographers at the University of Groningen,
the Netherlands, who solved the three-dimensional structures of the
lipases, helping team members to interpret the results of the directed
Providing Bacterial Lipases
For more than a decade, our group at Ruhr-Universität
Bochum has been studying bacterial lipases. Earlier, Uli Winkler
directed the characterization of extracellular lipases from S.
marcescens and P. aeruginosa, and also the purification of a Pseudomonas
lipase to electrophoretic homogeneity. Since then, we have studied
how lipase gene expression is regulated, mechanisms governing folding
and secretion of lipases, and also how lipases from gram-negative and
gram-positive bacteria might be used in biotechnological applications.
Extracellular lipases (with a molecular mass of 19 kDa)
produced by Bacillus subtilis are smaller than other bacterial
lipases, are tolerant to basic pH (optimum at pH 10), hydrolyze a
variety of different glycerol ester substrates, and prefer fatty acids
of medium chain lengths containing eight carbon atoms. We have developed
a heterologous overexpression system for producing B. subtilis
lipases LipA and LipB, have purified both enzymes to electrophoretic
homogeneity, and have compared their biochemical properties. Recently,
we obtained crystals of the B. subtilis LipA, enabling us to
determine its X-ray structure at 1.5 Å resolution (Fig. 1A). This
lipase contains a compact minimal a/ß hydrolase fold that is
characteristic for lipases. However, unlike other lipases, it contains
no lid domain, meaning its active site serine residue is exposed to the
Recently, we also solved the 3D structure of P.
aeruginosa LipA at 2.54 Å resolution (Fig. 1B). This lipase shows
an a/ß hydrolase fold with a six-stranded parallel ß-sheet
surrounded by a-helices on both sides. The catalytic triad is
composed of residues Ser-82, Asp-229, and His-251. The loop containing
His-251 is stabilized by an octahedrally coordinated calcium ion. The
enzyme can adopt a closed or an open conformation, depending on the
absence or presence of a substrate by movement of a-helices that
belong to the cap domain (indicated in red in Fig. 1B).
The extracellular lipase LipA from P. aeruginosa
and its encoding gene are unusual in several ways: (i) the lipase gene lipA
is located in an operon with a second gene lif, which encodes a
lipase-specific "foldase" required for folding the lipase
within the periplasm and properly secreting it; (ii) the lipase operon
is regulated in a complex way that includes a two-component regulatory
system LipQ/R; (iii) the type II secretory pathway, which consists of at
least 12 different proteins, secretes LipA of P. aeruginosa; (iv)
lipase acts as a virulence factor; and (v) LipA is stable indefinitely
at room temperature and shows a broad substrate specificity and
enantioselectivity towards various alcohols, amines, esters, and
triglyceridesthereby making it a good candidate for a variety of
High-Throughput Screening for Enantioselectivity
One of the major challenges in devising a specific
directed-evolution protocol concerns the need for high-throughput
screening for enantioselectivity. The degree of selectivity, expressed
as E or ee values, traditionally is determined by
separating the stereoisomer products by gas chromatography (GC) or
high-pressure liquid chromatography (HPLC) using chiral phases. However,
only a few dozen samples can be analyzed per day. Indeed, when we began
to study enantioselective enzymes in 1995, high-throughput assays for
such research were unavailable. Since then, several methods were
developed, two of which we used to identify enantioselective lipases
derived from P. aeruginosa and B. subtilis (Fig. 2).
One of those analytic methods, electrospray ionization
mass spectrometry (ESI-MS), enables one to separate the (R)- and
(S)-enantiomers of a particular chiral product, even though they
have identical mass spectra and, ordinarily in the absence of
chromatographic separation procedures, cannot be distinguished. However,
if one of the enantiomers is deuterium-labeled, the parent peaks appear
separately in the mass spectrum of the mixture, and integration then
provides the ee value. Adding the unlabeled starting compound to
the product mixture as an external standard allows one to determine how
much is converted to each of the two enantiomer products.
Culture supernantants containing B. subtilis
lipase catalyze an asymmetric hydrolysis of meso-1,4-diacetoxy-2-cyclopentene,
going to completion and forming chiral alcohols (Fig. 2A). The
deuterium-labeled pseudo-meso-substrate is hydrolyzed to give two
chiral products which are pseudo-enantiomers. In a typical example, the
wild-type B. subtilis lipase yields an ESI-MS spectrum in which
the ee-value is 38% (Fig. 2B).
A second analytic method depends on spectrophotometry to
determine the degree of enantioselectivity of the lipases being studied.
To prepare samples, we grew P. aeruginosa clones in wells of a
microtiter plate overnight, removed aliquots of the lipase-containing
culture supernatants, and then added them to other microtiter plates,
which contained the (R)- or (S)-enantiomer, respectively,
of the substrate 2-methyldecanoic acid p-nitrophenyl ester. The
reactions are monitored at 410 nm using a spectrophotometer and apparent
E-values are determined. Those clones showing improved
enantioselectivity as compared to the "parent" of the
corresponding generation (see Fig. 2D and E) are further analyzed by
reacting a racemic substrate mixture with culture supernatant and
subsequently analyzing the reaction products by chiral gas
Directed Evolution Put to the Test for Several
In an ongoing study of Bacillus subtilis lipase,
we use epPCR, saturation mutagenesis, and DNA shuffling to seek
improvements in the enantioselectivity of this enzyme. Use of the
wild-type enzyme leads to an ee value of only 38% in the
asymmetric hydrolysis of the substrate meso-1,4-diacetoxy-2-cyclopentene
(see Fig. 2A). Following an initial round of epPCR-based random
mutagenesis, we found a mutant with an ee value of 60%. This
value increased to about 70% ee in a second round, and we
continue to seek further increases in enantioselectivity by combining
saturation mutagenesis and DNA shuffling.
We followed a somewhat different experimental strategy
to improve the enantioselective properties of P. aeruginosa
lipase (Fig. 4). Before improvement, the enantioselectivity for the
reaction (Fig. 2C) is poor, yielding an ee of 5% in favor of the
(S)-acid at about 50% conversion. The selectivity factor E,
which reflects the relative rate of the reactions of the (S)- and
(R)-substrates, is only 1.1.
When doing random mutagenesis as one way of improving
these properties, one needs first to consider the problem of exploring
an individual protein's potential sequence space. In this case of an
enzyme consisting of 285 amino acids, allowing for all possible
permutations would result theoretically in 20,285 different mutant
enzymes, whose collective mass would greatly exceed the mass of the
universe, even if only one molecule of each of those enzymes was
produced. The other end of the scale entails the minimum amount of
structural change, namely the substitution of just one amino acid per
molecule of enzyme. On the basis of the algorithm N = 19M
· 285!/[(285 - M)! · M!] , in which M = number of
amino acid substitutions per enzyme molecule (here M = 1), the
library of mutants N would theoretically contain 5,415 membersonly
about one third of which is accessible by epPCR.
In our initial strategy, we aimed for a mutagenesis
frequency of M = 1, relying on such limited changes to provide
stepwise improvements in enantioselectivity. Upon generating a library
of only 1,000 mutants in the first generation, about 12 variants of this
enzyme with improved properties were identified, the best one resulting
in an ee value of 31% (E = 2.1) in the test reaction. We
repeated the process in a second, third, and fourth rounds of
mutagenesis, each time forming slightly larger libraries of mutants
(2,000 - 3,000). By the fourth round, we identified an altered lipase
with an ee of 81%, corresponding to a selectivity factor of E
= 11.3. We then created a larger, fifth-generation library of mutant
enzymes with further improvements.
We also developed alternative, more efficient ways to
explore protein sequence space with respect to enantioselectivity.
Before beginning, we sequenced the genes encoding the best members of
the various mutant generations to identify the position and nature of
the amino acid substitutions that were responsible for increasing
enantioselectivitytentatively classifying them as "hot
spots" while also assuming that, even if those were the correct
positions, they might not yet contain the optimal amino acids.
Therefore, we applied saturation mutagenesis at one or more of these
An example is amino acid leucine155 that, during the
initial epPCR experiments, was replaced by serine, yielding a mutant
enzyme with slightly increased enantioselectivity. However, saturation
mutagenesis led to phenylalanine replacing leucine at this site, and
resulted in a significant increase in enantioselectivity. We continued
to use improved mutants as the starting point for further rounds of
epPCR, leading us eventually to a variant with an E value of 25 (ee
= 90%), which contains five mutations (V47G, V55G, S149G, S155F, and
Thus, by combining epPCR with saturation mutagenesis, we
developed a highly efficient way to explore protein sequence space with
respect to enantioselectivity. To further increase the
enantioselectivity of P. aeruginosa lipase, we explored DNA
shuffling, but our initial experiments proved disappointing and failed
to produce better variants. As an alternative, we tried combining
high-error-rate epPCR with combinatorial multiple cassette mutagensis.
By introducing an average of three amino acid exchanges into each lipase
molecule, we obtained libraries that contain several variants with
significantly increased enantioselectivities (E = 3- 32). In
parallel, a cassette was isolated and saturated with mutations at
positions corresponding to amino acids 155 and 162. These positions were
identified as being important for enantioselectivity of P. aeruginosa
lipase from previous directed evolution experiments as well as from
analyzing the crystal structure. Genes obtained from both libraries were
reassembled by DNA shuffling, which yielded a highly S-selective
variant (E = 51) that contains six amino acid changes (D20N,
S53P, S155M, L162G, T180I, T234S) and appears to be particularly active
and stable (Fig. 4).
Another intriguing challenge concerns the possibility of
reversing the sense of an enzyme's enantioselectivity. We carefully
examined the libraries we had obtained by high-error-rate epPCR and
identified several (R)-selective enzymes in addition to the (S)-selective
analogs. Thus, the same library contains (S)- as well as (R)-selective
variants. In a second round of high-error-rate mutagenesis, we found
additional (R)-selective variants that provided genes for further
improvements through DNA shuffling (Fig. 4). An appropriate combination
of epPCR at high error rate and DNA shuffling led us to identify a
highly (R)-selective variant (E = 30) that carries 11
amino acid substitutions (M16L, A34T, P86L, T87S, V94A, D113G, S147N,
T150A, L208H, V232I, and S237T).
What Makes an Enzyme Enantioselective?
Knowing the 3D structure of the wild-type P.
aeruginosa lipase enables us to identify which amino acids are
altered among the best variants and provides us with valuable
information with which to explain the effects of these mutations. For
instance, the lipase variant showing an enantioselectivity of E =
25 towards the (S)-enantiomer of the substrate contains five key
amino acid substitutions, four of which introduce glycine residues that
most probably increase the overall conformational flexibility of this
lipase. Surprisingly, none of the five amino acid substitutions is in
direct contact with the bound substrate at the active center, thus
excluding a direct spatial effect on the enantioselectivity of the
reaction. Instead, the substitutions are located directly or in close
vicinity to loops that are involved in the enzyme's transition from a
closed to an open conformation.
Similarly, for the (S)-selective variant obtained
by combinatorial multiple cassette mutagenesis (E = 51), the
mutations identified by directed evolution are also located remote from
the active site, with the exception of a change at position 162, which
makes direct contact with the substrate.
It is too early to establish a standardized protocol for
improving an enzyme's enantioselective properties. However, our results
suggest that certain domains of a given enzyme may require more
flexibility to improve its enantioselective properties. Moreover,
binding of the desired substrate should be optimized.
These results for a P. aeruginosa lipase
constitute a comprehensive effort to explore the protein sequence space
and optimize specific hot spots along it with respect to improving the
enantioselectivity of a particular enzyme-catalyzed reaction. The
experiments do not depend on theoretical predictions regarding either
the reaction mechanism or the enzyme's 3D structure. Next we plan to
purify and crystallize the best enantioselective variants as a way
toward better understanding the structural basis of enantioselectivity.
We predict that directed evolution will be used more widely to produce
novel biocatalysts that are active, stable, and highly enantioselective,
making them good candidates for a variety of practical applications in
Research in the lab of K.E.J. was
supported by the European Commission in the framework of the programs
BRIDGE (BIOT-CT91-0272) and BIOTECHNOLOGY (BIO4-CT96-0119,
BIO4-CT98-0249, QLK3-CT2001-00519, and QLRT-2001-02086). Research in the
lab of M.T.R was supported by European Community grants BIO4-CT98-0249,
QLK3-2000-00426, and QLK3-CT2001-00519.
Farinas, E.T., T. Bulter, and Frances H.
Arnold. 2001. Directed enzyme evolution. Curr.
Opin. Biotechnol. 12:545-551.
Jaeger, K.-E, B. W. Dijkstra, and M. T.
Reetz. 1999. Bacterial biocatalysts: molecular
biology, three-dimensional structures, and biotechnological applications
of lipases. Annu. Rev. Microbiol. 53:315-351.
Jaeger, K.-E., and T. Eggert. 2002.
Lipases for biotechnology. Curr. Opinion Biotechnol. 13:390-397.
Jaeger, K.-E., T. Eggert, A. Eipper, and
M. T. Reetz. 2001. Directed evolution and the
creation of enantioselective biocatalysts. Appl. Microbiol. Biotechnol. 55:519-530.
Jaeger, K.-E., and M. T. Reetz.
1998. Microbial lipases form versatile tools for biotechnology. Trends
Jaeger, K.-E., and M. T. Reetz.
2000. Directed evolution of enantioselective enzymes for organic
chemistry. Curr. Opin. Chem. Biol. 4:68-73.
Powell, K. A., S. W. Ramer, S. B. del
Cardayre, W. P. C. Stemmer, M. B. Tobin, P. F. Longchamp, and G.W.
Huisman. 2001. Directed evolution and
biocatalysis. Angew. Chem. Int. Ed. 40:3948-3959.
Reetz, M. T.
2001. Combinatorial and evolution-based methods in the creation of
enantioselective catalysts. Angew. Chem. Int. Ed. 40:1000-1026.