Model Concepts
In order to avoid extensive enzyme engineering and trial-and-error modifications of
substrate structures, several useful ‘models’ for the more commonly used enzymes
have been developed to predict the stereochemical outcome of enzymatic reactions
on nonnatural substrates. These models provide a rationale to ‘redesign’ a substrate
or an enzyme, when initial results are not satisfying with respect to reaction rate
and/or selectivity. Since the application of such ‘models’ holds a couple of potential
pitfalls, the most important principles underlying their construction and application
are discussed here.
Molecular Modeling The structure of an enzyme in crystallized form can be
accurately determined by X-ray crystallography [351–354]. Since the tertiary
structure of most enzymes is closely related to the preferred form in a dissolved
state [355], this method provides the most accurate 3D-description of an active site.
However, X-ray data only represent a static protein structure, while the chiral
recognition process during formation of the enzyme–substrate complex is a complex dynamic process. Thus, any attempt of predicting the selectivity of an enzymatic reaction based on X-ray data is comparable to explaining the complex
movements in a somersault from a single photographic snapshot.
Although the rapidly increasing number of crystal structures of proteins,
11 which
are available through the Protein Data Bank (PDB), encompass widely used
enzymes, such as α-chymotrypsin [121], subtilisin [196], and lipases from Mucor
spp. [9], Geotrichum candidum [356], Candida rugosa (formerly cylindracea)
[357], Candida antarctica B [358], and Pseudomonas glumae [359], for a large
number of synthetically useful enzymes, such as pig liver esterase, relevant structural data are not available.
If the amino acid sequence of an enzyme is known either entirely or even in part,
computer-assisted calculations can provide a model for its three-dimensional structure [360]. This is done by comparing the amino acid sequence of the enzyme in
question with that of other enzymes with known sequence and three-dimensional
structure, which serve as blueprint. Depending on the percentage of the homology,
i.e., ‘overlap’, of the amino acid sequences, the results are more or less accurate. In
general, an overlap of about ~50–60% is sufficient for good results; less is considered too inaccurate. Tools for the construction of enzyme models, such as Modeller
12 or Phyre
2 13 are available via Internet. The utility of various protein structure
prediction methods was recently reviewed by Zhang [361].
11 To date (2017), approx. 111.000 protein crystal structures are available.
12 https://salilab.org/modeller/
13 http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id¼index
80
2 Biocatalytic Applications
In order to avoid extensive enzyme engineering and trial-and-error modifications of
substrate structures, several useful ‘models’ for the more commonly used enzymes
have been developed to predict the stereochemical outcome of enzymatic reactions
on nonnatural substrates. These models provide a rationale to ‘redesign’ a substrate
or an enzyme, when initial results are not satisfying with respect to reaction rate
and/or selectivity. Since the application of such ‘models’ holds a couple of potential
pitfalls, the most important principles underlying their construction and application
are discussed here.
Molecular Modeling The structure of an enzyme in crystallized form can be
accurately determined by X-ray crystallography [351–354]. Since the tertiary
structure of most enzymes is closely related to the preferred form in a dissolved
state [355], this method provides the most accurate 3D-description of an active site.
However, X-ray data only represent a static protein structure, while the chiral
recognition process during formation of the enzyme–substrate complex is a complex dynamic process. Thus, any attempt of predicting the selectivity of an enzymatic reaction based on X-ray data is comparable to explaining the complex
movements in a somersault from a single photographic snapshot.
Although the rapidly increasing number of crystal structures of proteins,
11 which
are available through the Protein Data Bank (PDB), encompass widely used
enzymes, such as α-chymotrypsin [121], subtilisin [196], and lipases from Mucor
spp. [9], Geotrichum candidum [356], Candida rugosa (formerly cylindracea)
[357], Candida antarctica B [358], and Pseudomonas glumae [359], for a large
number of synthetically useful enzymes, such as pig liver esterase, relevant structural data are not available.
If the amino acid sequence of an enzyme is known either entirely or even in part,
computer-assisted calculations can provide a model for its three-dimensional structure [360]. This is done by comparing the amino acid sequence of the enzyme in
question with that of other enzymes with known sequence and three-dimensional
structure, which serve as blueprint. Depending on the percentage of the homology,
i.e., ‘overlap’, of the amino acid sequences, the results are more or less accurate. In
general, an overlap of about ~50–60% is sufficient for good results; less is considered too inaccurate. Tools for the construction of enzyme models, such as Modeller
12 or Phyre
2 13 are available via Internet. The utility of various protein structure
prediction methods was recently reviewed by Zhang [361].
11 To date (2017), approx. 111.000 protein crystal structures are available.
12 https://salilab.org/modeller/
13 http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id¼index
80
2 Biocatalytic Applications