Crystal Structures and Catalytic Mechanism of l-erythro-3,5-Diaminohexanoate Dehydrogenase and Rational Engineering for Asymmetric Synthesis of β-Amino Acids

Article abstract of DOI:10.1002/anie.202017225

Amino acid dehydrogenases (AADHs) have shown considerable potential as biocatalysts in the asymmetric synthesis of chiral amino acids. However, compared to the widely studied α-AADHs, limited knowledge is available about β-AADHs that enable the synthesis of β-amino acids. Herein, we report the crystal structures of a l-erythro-3,5-diaminohexanoate dehydrogenase and its variants, the only known member of β-AADH family. Crystal structure analysis, site-directed mutagenesis studies and quantum chemical calculations revealed the differences in the substrate binding and catalytic mechanism from α-AADHs. A number of rationally engineered variants were then obtained with improved activity (by 110–800 times) toward various aliphatic β-amino acids without an enantioselectivity trade-off. Two β-amino acids were prepared by using the outstanding variants with excellent enantioselectivity (>99 % ee) and high isolated yields (86–87 %). These results provide important insights into the molecular mechanism of 3,5-DAHDH, and establish a solid foundation for further design of β-AADHs for the asymmetric synthesis of β-amino acids.

Full text of DOI:10.1002/anie.202017225

Angewandte
Research Articles
Chemie
How to cite: Angew. Chem. Int. Ed. 2021, 60, 10203–10210
International Edition: doi.org/10.1002/anie.202017225
German Edition: doi.org/10.1002/ange.202017225
Asymmetric Synthesis
Crystal Structures and Catalytic Mechanism of l-erythro-3,5-
Diaminohexanoate Dehydrogenase and Rational Engineering for
Asymmetric Synthesis of b-Amino Acids
Na Liu, Lian Wu, Jinhui Feng, Xiang Sheng,* Jian Li, Xi Chen, Jianjiong Li, Weidong Liu,
Jiahai Zhou,* Qiaqing Wu,* and Dunming Zhu*
Abstract: Amino acid dehydrogenases (AADHs) have shown
considerable potential as biocatalysts in the asymmetric syn-
thesis of chiral amino acids. However, compared to the widely
studied a-AADHs, limited knowledge is available about b-
AADHs that enable the synthesis of b-amino acids. Herein, we
report the crystal structures of a l-erythro-3,5-diaminohexa-
noate dehydrogenase and its variants, the only known member
of b-AADH family. Crystal structure analysis, site-directed
mutagenesis studies and quantum chemical calculations re-
vealed the differences in the substrate binding and catalytic
mechanism from a-AADHs. A number of rationally engi-
neered variants were then obtained with improved activity (by
110–800 times) toward various aliphatic b-amino acids without
an enantioselectivity trade-off. Two b-amino acids were
prepared by using the outstanding variants with excellent
an attractive straightforward approach to access b-amino
acids and derivatives.[3] Recently, direct amination of b-
hydroxyl acid esters was achieved using a borrowing hydrogen
methodology involving cooperative catalysis of a ruthenium
catalyst and a Brønsted acid additive.[4] Biocatalytic reductive
amination of b-keto acids has been reported using trans-
aminases, which catalyze the amino transfer from an amino
acid or amine to b-keto acids to give optically active b-amino
acids.[5] While this represents a promising method for the
preparation of chiral b-amino acids, the efficiency and scope
still desire further improvements for practical applications. A
paradigm of biocatalytic asymmetric reductive amination of
carbonyl group is the a-amino acid or amine dehydrogenase-
catalyzed reductive amination of a-keto acids or ketones in
the presence of ammonium as amino donor,[6] that can be
enantioselectivity (> 99% ee) and high isolated yields (86– coupled with alcohol dehydrogenase-catalyzed oxidation of
87%). These results provide important insights into the
molecular mechanism of 3,5-DAHDH, and establish a solid
foundation for further design of b-AADHs for the asymmetric
synthesis of b-amino acids.
alcohols to directly convert hydroxyl group to amino group in
the enantioselective manner.[7] For example, the asymmetric
synthesis of l-tert-leucine via leucine dehydrogenase
(LeuDH)-catalyzed reductive amination of 3,3-dimethy-2-
oxobutanoic acid is a representative case at industrial scale,
showing advantages in terms of atomic economy and environ-
mental impact due to the by-products being H2O and CO2
when formate dehydrogenase is used for co-factor regener-
ation and ammonium formate serves as reactant for both
enzymatic reactions.[8] We thus envisioned that b-amino acid
dehydrogenase (b-AADH) could serve as a useful biocatalyst
for the asymmetric synthesis of chiral b-amino acids.
Introduction
b-Amino acids are not only found as essential component
in natural compounds with interesting biological properties,
but also have emerged as a class of important building blocks
for synthetic pharmaceuticals and promising scaffolds in
peptide research.[1] Therefore, both chemical and biocatalytic
methods have been developed for the construction of b-amino
acid moieties.[2] Among them, the reductive amination of b-
keto or b-hydroxyl carboxylic acids and derivatives represents
However, compared to a-amino acid dehydrogenase (a-
AADH), the b-counterpart has been much less studied. To
the best of our knowledge, the only reported b-AADH is l-
[*] N. Liu, Prof. Dr. J. Feng, Prof. Dr. X. Sheng, Prof. Dr. X. Chen, J. Li,
Prof. Dr. W. Liu, Prof. Dr. Q. Wu, Prof. Dr. D. Zhu
National Engineering Laboratory for Industrial Enzymes and Tianjin
Engineering Research Center of Biocatalytic Technology, Tianjin
Institute of Industrial Biotechnology, Chinese Academy of Sciences
and National Technology Innovation Center for Synthetic Biology
Tianjin 300308 (China)
Technology, Chinese Academy of Sciences
Shenzhen 518055 (China)
E-mail: jiahai@siat.ac.cn
L. Wu, J. Li, Prof. Dr. J. Zhou
State Key Laboratory of Bio-organic and Natural Products Chemistry
Center for Excellence in Molecular Synthesis, Shanghai Institute of
Organic Chemistry, Chinese Academy of Sciences
Shanghai 200032 (China)
E-mail: shengx@tib.cas.cn
wu_qq@tib.cas.cn
zhu_dm@tib.cas.cn
L. Wu
The Key Laboratory of Synthetic Biology, and CAS Center for
Excellence in Molecular Plant Sciences, Institute of Plant Physiology
and Ecology, Chinese Academy of Sciences
Shanghai 200032 (China)
N. Liu, L. Wu, Prof. Dr. J. Feng, J. Li, Prof. Dr. Q. Wu, Prof. Dr. D. Zhu
University of Chinese Academy of Sciences
19A Yuquan Road, Beijing 100049 (China)
Prof. Dr. J. Zhou
Supporting information and the ORCID identification number(s) for
the author(s) of this article can be found under:
https://doi.org/10.1002/anie.202017225.
CAS Key Laboratory of Quantitative Engineering Biology, Shenzhen
Institute of Synthetic Biology, Shenzhen Institutes of Advanced
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210
ꢀ 2021 Wiley-VCH GmbH
10203

Angewandte
Research Articles
Chemie
erythro-3,5-diaminohexanoate dehydrogenase (3,5-DAHDH, Results and Discussion
EC.1.4.1.11), which has been identified to participate in the
degradation pathway of lysine from some bacterial species.[9]
This NAD(P)H-dependent enzyme catalyzes the reversible
deamination/ amination reaction between (3S,5S)-diamino-
hexanoate (1a) and 3-keto-5-aminohexanoate (1b,
Scheme 1).[9a] However, 3,5-DAHDHs exhibited relatively
strict substrate specificity toward native substrate 1a, thus
limiting their synthetic applications. In this context, we
engineered the 3,5-DAHDH from Candidatus Cloacamonas
acidaminovorans (3,5-DAHDHcca, Uniprot ID: B0VJ11)[10]
by screening a domain scanning mutagenesis library to
broaden the substrate scope.[11] Although some variants were
obtained to show enhanced activities toward several aliphatic
and aromatic b-amino acids, little is known about this enzyme
and the 3,5-DAHDHs in general.
Crystal structures of 3,5-DAHDHcca
We solved the structures of 3,5-DAHDHcca in apo form
(WT), complexed with cofactor (WT·NADPH), and variant
structure E310G/A314Y·NADPH that exhibited a broad
substrate-scope in our previous study,[11] with eight, four and
two molecules in the asymmetric unit, respectively (Table S1).
Unfortunately, the attempts to obtain the structures bound
with amino acid substrate or keto product failed (Section 1.3
in Supporting Information). The overall monomeric structure
of 3,5-DAHDHcca was composed of an N-terminal catalytic
domain (Domain I) and a C-terminal cofactor-binding
domain (Domain II) separating by a deep cleft (Figure 1a),
that showed similarities to the structures of a-AADHs and
the recently reported native amine dehydrogenases (AmDH)
at the single subunit level,[12] although there was no significant
sequence homology between them. The cofactor-binding
domain of 3,5-DAHDHcca contained the typical Rossmann
fold observed in all NAD(P)H-dependent dehydrogenases,[13]
but the catalytic domain was significantly different from
them. The catalytic domain of 3,5-DAHDHcca was featured
by six anti-paralleled b-strands almost bent into U-shape with
two a-helices, while the b-sheets were straight in LeuDH,
phenylalanine dehydrogenase (PheDH), meso-diaminopime-
late dehydrogenase (DAPDH) and AmDH.[12b–e] In the
structures of 3,5-DAHDHcca, every two monomers were
packed into a homodimer by hydrogen bond interactions
between the cofactor-binding domain of each subunit. The
dimerization resulted in 12-stranded b-sheets spanning two
molecules in an antiparallel manner as presented in Fig-
ure 1a. This is different from the crystal structures of LeuDH,
PheDH, DAPDH, and AmDH in which the dimeric archi-
tecture is constructed by the b-sheets in catalytic domain
rather than the cofactor-binding domain.[12b–e]
Scheme 1. l-erythro-3,5-Diaminohexanoate dehydrogenase (3,5-
DAHDH) catalyzed reaction.
On the other hand, it is still quite a challenge to engineer
an enzyme to expand its substrate scope and/or improve its
activity while keeping its excellent stereoselectivity. In order
to achieve this rational engineering goal, a prerequisite is to
understand the catalytic mechanism and substrate-binding of
the target enzyme. Herein we report the X-ray crystal
structures of 3,5-DAHDHcca and its variants, together with
the site-directed mutagenesis and quantum chemical calcu-
lations, to reveal the catalytic mechanism of 3,5-DAHDHcca
and the molecular basis of substrate specificity, which offered
directions for engineering b-AADHs to achieve their syn-
thetic potentials. Subsequently, structure-guided directed
evolution of 3,5-DAHDHcca was performed to enhance the
activity toward non-native substrates without sacrifice of
enantioselectivity, and the obtained variants were applied for
the asymmetric synthesis of b-amino acids at preparative
scale.
The electron density of WT·NADPH unequivocally
revealed that two cofactors were bound on the interface of
two monomers (monomer A and C) forming a homodimer
with the nicotinamide ring deep in the active cleft (Fig-
ure 1b). The ensemble framework of this homodimer binding
with NADPH had no significant difference to the ligand-free
structure (Figure S1), with a root mean square deviation
Figure 1. Crystal structures of 3,5-DAHDH from Candidatus Cloacamonas acidaminovorans (3,5-DAHDHcca). a) Secondary structure elements of
monomer and homodimer in wild-type structure (WT) are shown as cartoon representations. b) Dimer structure with cofactors binding in
monomers A and C of WT·NADPH. c) Zoom-in view of the interaction between the amide of NADPH binding in monomer A (blue) and residue
E310 in monomer C (red). The cofactors in (b) and (c) are indicated by the simulated annealing omit maps (blue mesh, contoured at 1.0 s).
10204 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210

Angewandte
Research Articles
Chemie
(RMSD) value of 0.63 Š for 700 Ca atoms. While a NADPH
allosteric, it is not uncommon that conformation changes
was mainly bound to the residues of monomer A (Figure S2),
the amide group of nicotinamide also interacted with the
residue E310 of monomer C by hydrogen bond (Figure 1c).
The binding mode of one NADPH shared by two monomers
was not formed by crystal packing, indicating that the
dimerization event was required for enzymatic catalysis and
each homodimer was a minimal integral functional unit.
However, in the reported structures of LeuDH, PheDH,
DAPDH and native AmDH, the substrate-binding pockets
are constructed by one subunit.[12b–e]
are involved in substrate binding and product release.[15] In
the functionally similar NAD(P)H-dependent dehydrogen-
ases, it is also generally believed that the apo state has an open
cleft which closes upon the binding of reactants.[14b]
Active sites and catalytic mechanism of 3,5-DAHDHcca
To understand how 3,5-DAHDHcca recognizes the b-
amino acid substrate and cofactor NADP+, and the reaction
mechanism, quantum chemical calculations[16] were per-
formed by using a large active site model of the ternary
complex (Figure S4) constructed on the basis of the E310G/
A314Y·NADPH structure by modifying the two mutated
residues back to their original forms. The native substrate 1a
was docked into the active sites. According to the pH
dependence of 3,5-DAHDHcca (Figure S5), the oxidative
deamination activity toward 1a increased rapidly under
alkaline conditions from pH 9.0, and the optimum pH was
11.0, which is similar to that for some PheDHs and
LeuDHs.[17] Therefore, the titratable residues (D49, D177,
K203 and E310) and 1a in the employed model were set as
deprotonated states.
As shown in the optimized structure of the ternary
complex (called E:DAH, Figure S6), the carboxylate and C5-
amino groups of 1a are captured by the side chains of S125
and E310, respectively. The other amino group at C3 forms
two hydrogen bonds with D49 and D177. For NADP+, the 2’-
hydroxyl of nicotinamide ribose and the amide group interact
with D49 and E310, respectively. Some main chain groups are
also involved in the binding of 1a and NADP+, including G94,
G95, G180, G321 and G323.
We also compared the structures of WT·NADPH and
E310G/A314Y·NADPH. Most obviously, E310G destroyed
the hydrogen bond interaction with the amide group of
NADPH observed in the wild-type structure. While mutation
showed no significant effect on the conformation of cofactor-
binding domain, the catalytic domain of the NADPH-bound
subunit (monomer B) in E310G/A314Y·NADPH exhibited
a tremendous conformational change by the superimposition
with monomer A of WT·NADPH (Figure 2a), with an RMSD
of 3.89 Š for 352 Ca atoms. Among the a-AADHs, inter-
domain flexibility is characterized by the angle of domain
cleft.[12c,14] In 3,5-DAHDHcca, the angle was quantified by
three Ca atoms of T91 in Domain I, D177 in the helix a3 at the
base of the cleft, and M298 in Domain II. For the structures of
3,5-DAHDHcca in apo state and complexed with NADPH,
the angle between two domains ranged from 46.18 to 56.68,
which were observed to be an open form. By contrast, the
monomer B in E310G/A314Y·NADPH exhibited a relative
closed conformation with a smaller angle of 22.28 (Figure 2a).
It is also worth to note that, the ꢀ 308 inter-domain movement
enabled the side chains of D49 and S50 to shift about 8.2 Š
and 8.8 Š toward NADPH, respectively, resulting in the
formation of hydrogen bonds with the nicotinamide ribose
and pyrophosphate moiety (Figure 2b). Additionally, the
flexibility of catalytic domain was also confirmed by the much
higher B-factors of E310G/A314Y·NADPH compared to
WT·NADPH (Figure S3). We thus propose that the confor-
mation of E310G/A314Y·NADPH structure is probably
closer to the reaction state and the proposed substrate
binding pocket indicated by C4 of NADPH is shown in
Figure 2c. Considering that the enzyme is dynamic and
In the proposed mechanism (Figure 3), the catalytic cycle
starts with the hydride transfer from C3 of 1a to C4 of
NADP+ with a barrier of 18.8 kcalmolÀ1 (TS1, Figure S7 for
structure). The generated iminium intermediate (Int1) is
1.8 kcalmolÀ1 lower than E:DAH. In the next hydration step,
both the deprotonated C5-amino group of 1a and D177 can
be the catalytic base to activate the water molecule that
attacks the electrophilic carbon of Int1. The barrier is
8.3 kcalmolÀ1 (TS2) or 12.4 kcalmolÀ1 (TS2’) relative to Int1
Figure 2. Structural comparison of WT·NADPH and E310G/A314Y·NADPH, and the proposed substrate pocket in E310G/A314Y·NADPH.
a) Superimposed overall structures of WT·NADPH monomer A (light blue) and E310G/A314Y·NADPH monomer B (wheat) with an angle of 48.98
and 22.28, respectively. b) Zoom-in view of the interactions between D49, S50 and NADPH (green in E310G/A314Y·NADPH and orange in
WT·NADPH). c) Substrate pocket of E310G/A314Y·NADPH indicated by the coordinate of C4 of NADPH. The substrate pocket was relatively
hydrophobic and mainly comprised of residues both from monomer A (red, L307, G310, G311) and B (blue, M96, L123, V124, S125, V159, L176,
D177, V178, A179, G180, G323, Y324 and H328).
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10205

Angewandte
Research Articles
Chemie
Figure 3. Proposed catalytic mechanism of 3,5-DAHDHcca. The relative energies of all the intermediates and transition states are given in
kcalmolÀ1 in parentheses. Optimized structures for all the intermediates and transition states are given in Figure S7.
for the pathways with C5-amino group or D177, respectively.
For the former pathway, after the formation of hydrated
intermediate (Int2), a proton transfer takes place from the
protonated C5-amino group to D177 via the newly formed
hydroxyl group with a barrier of 10.7 kcalmolÀ1 (TS3) relative
to Int1, arriving at the same intermediate (Int3) as the latter
pathway. Then, a proton transfer from D177 to C3-amino
group results in intermediate Int4 with À0.1 kcalmolÀ1
energy, in which the protonated amino group can be
significantly stabilized by the carboxylate groups of D49
ca. 2 kcalmolÀ1. This explained why the wild-type enzyme
exhibited a hundreds of times lower activity (10 mUmgÀ1)
toward (S)-3-aminohexanoic acid (2a) lacking C5-amino
group compared to 1a (3.0 UmgÀ1).[11] These results thus
provide the molecular insights into the substrate specificity of
3,5-DAHDHcca toward 1a.
To further verify the contributions of the above-identified
amino acid residues during the catalysis, we performed site-
directed mutagenesis and measured the enzymatic activities
(Figure 4) and kinetic parameters (Table S2 for 1a, Table S3
for NADP+). Replacement of S125 with alanine resulted in
almost loss of the enzyme activity, but S125T showed ꢀ 6%
activity of the wild-type enzyme, and was ꢀ 300 times higher
than S125A. The 20-fold increased Km value of S125A for
substrate 1a and the partial rescue of this almost inactive
variant by S125T suggested that the hydroxyl group of this
residue is important for the binding of carboxyl group of the
substrate, and these results are consistent with the above
analysis of the substrate binding mode. All the variants of
E310 exhibited much lower activities than the wild-type
enzyme for the native substrate 1a.[11] And the variant E310G
exhibited a ꢀ 90-fold decreased kcat (6.25 minÀ1) toward 1a,
and two times increased Km values for both NADP+ and 1a.
These results proved that residue E310 was vital for catalysis
and substrate recognition since it was bound with the amide of
NADPH and the C5-amino group of 1a. The mutation of
S50A did not significantly affect the enzyme activity, and this
was supported by the evolutionary analysis that across 831
diverse 3,5-DAHDH sequences this residue was conserved as
either serine or alanine at a similar frequency (Figure S8). In
contrast, substituting S50 with threonine that possesses the
À
and D177. Finally, the products are formed by the C N bond
cleavage, concurrently with intramolecular proton transfer
from the C3-hydroxyl group to the C5-amine (TS4). This step
has a barrier of 18.2 kcalmolÀ1 relative to Int1. The energy of
enzyme-product complex E:KAH is 1.9 kcalmolÀ1 lower than
E:DAH. In E:KAH, the protonated C5-amino is stabilized by
the interaction with the side chain of E310.
Both the hydride transfer and the deamination can be
rate-limiting steps (RDSs) for 3,5-DAHDHcca, since they
have very similar barriers (18.8 kcalmolÀ1 vs. 18.2 kcalmolÀ1).
The calculated barrier of the overall reaction is in good
agreement with the experimentally measured rate constant
(kcat) of 553 minÀ1 (Table S2), corresponding to an activation
energy of ca. 16 kcalmolÀ1 according to classical transition
state theory. For a-AADHs that share a similar catalytic
mechanism as 3,5-DAHDHcca, their RDSs are still not
established.[6a,18] However, it was proposed that the hydride
transfer is unlikely to be RDS for the oxidation reaction of
PheDH.[18c] An interesting finding for 3,5-DAHDHcca is that
D177 and the C5-amino group of 1a could serve as the
catalytic group to accept the proton, in favor of the latter by
10206 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210

Angewandte
Research Articles
Chemie
enzyme. In addition, replacement of the only basic residue
H328 in the active sites of 3,5-DAHDHcca with alanine did
not significantly affect the activity, suggesting that it was not
essential for the enzyme function. Overall, the site-directed
mutagenesis studies confirmed the likely roles of these
specific residues of 3,5-DAHDHcca, that showed some
differences from a-AADHs.
Structure-guided directed evolution of 3,5-DAHDHcca for non-
native substrates
Although 3,5-DAHDHs offer a new venue for the
asymmetric synthesis of b-amino acids, it is limited due to
their quite strict substrate specificity. In previous work by
screening 85 variant libraries of 3,5-DAHDHcca, we identi-
fied variants E310G/A314Y and E310S/A314N with im-
proved activities toward (R)-b-homomethionine (3a) by
ꢀ 200 times. Further mutagenesis studies showed that the
replacement of E310 rather than A314 significantly affected
the substrate scope.[11] This is consistent with the substrate
binding mode that the C5-amino group of the native substrate
interacts with the carboxyl group of E310. Therefore, further
engineering of 3,5-DAHDHcca to enhance the activity
toward b-amino acids was performed on the basis of variant
E310G.
The crystal structures of 3,5-DAHDHcca and the opti-
mized ternary complex E:DAH indicated that the substitu-
tion of E310 with glycine would destroy its binding with the
amide of NADPH. The optimized structure of enzyme-
substrate (3a) complex for the E310G variant further
revealed a flip of the amide group due to the absence of
a hydrogen bond with E310 (Figure S9). As such, we
hypothesized that the catalytic activity of this enzyme could
be improved by reconstructing this interaction between the
amide group of NADPH and nearby residues. The E310G/
A314Y·NADPH structure showed that the residues within
a 4 Š radius sphere centered on the C-atom of amide included
F296, M298 and G323 (Figure S10). Due to the direct p-p
interaction between F296 with the nicotinamide ring of
NADPH, F296 was excluded from mutation. The side chain
of M298 was located in the side of nicotinamide ring of
NADPH that is away from the amide, this residue was not
considered for mutation either. Considering that the Ca atom
of G323 was only 3.8 Š away from the amide group of
NADPH, it was firstly replaced with serine that has a shortest
side chain with a hydroxyl group. The result showed that
variant E310G/G323S improved the specific activity toward
3a by 17-fold compared to variant E310G (Table S4).
Furthermore, the enzyme activity of E310G/G323T was
increased by 5 times toward 3a (Table S4), which proved
the importance of the hydroxyl on this residue. The kinetic
data showed that E310G/G323S had slight lower Km toward
NADP+ (13.0 mM vs. 16.9 mM, Table S3) and 3a (30.4 mM vs.
40.5 mM, Table S5) compared to E310G. The structures of
E310G/G323S·NADPH and E310G/A314Y·NADPH (Ta-
ble S1) indicated that the substitution of G323 with serine
may introduce an additional hydrogen bond between the
hydroxyl side chain of S323 and the amide group of NADPH
Figure 4. Relative activity of 3,5-DAHDHcca variants for oxidative
deamination toward 1a. The activity of wild-type enzyme was set as
100%. Asterisks indicate that the activity was not measurable. Error
bars represent the standard deviations of three replicates.
similar property as serine caused a 100-fold loss of activity,
suggesting that the extra methyl group in the side chain might
prevent the hydrogen bond interaction between the hydroxyl
group with pyrophosphate moiety of NADPH (as shown in
Figure 2b) and thus led to a decreased activity. The substi-
tution of residue D49 with alanine led to the complete
deactivation, and the activity was less than 0.1% of the wild-
type 3,5-DAHDHcca by replacing D49 with N or E. Mutation
of D49 to N might not significantly change its hydrogen bond
network with the nicotinamide ribose of NADP+ and the C3-
amino group of 1a, resulting in only about 1.4-fold and 3.8-
fold increase in Km values toward NADP+ and 1a, respec-
tively, compared to wild-type enzyme. However, variant
D49N had a significantly lower ability to stabilize the Int4
+
in Figure 3 due to lack of the ionic interaction with C3-NH3 ,
leading to the catalytic constant kcat dropping 675-fold for 1a.
The removal of carboxyl group at residue D177 was
detrimental to enzyme activity, both variants D177A and
D177N decreased the activity nearly a thousand times. The
Km value of D177A for 1a was approximate to that of wild-
type, but the kcat was decreased by ꢀ 3000-fold. In addition,
the enzyme activity could be restored to ꢀ 7% of the wild-
type enzyme by D177E which also has a carboxyl group but
one methylene longer than aspartic acid, indicating that the
carboxyl group at this position may play a key role in the
catalytic process and the length of amino acid side chain also
has an important effect on catalysis. These results are
consistent with the proposed catalytic mechanism (Figure 3)
that D177 serves as a catalytic residue being involved in the
proton transfer with C3-NH2 group of the substrate, rather
than a role of binding (recognition) substrate. For the
DAPDHs and native AmDHs, D or E is also proposed as
the key catalytic residue.[12e,14c] In the catalytic mechanism of
LeuDH and PheDH, lysine is the catalytic residue partici-
pated in the hydration step of the deamination direction.[18b,c]
However, for 3,5-DAHDHcca, the activity of the variant
D177K was barely detected, and the enzyme activities of
D177R and D177H were less than 1% that of wild-type
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10207

Angewandte
Research Articles
Chemie
(Figure S11). This interaction was corroborated by quantum
chemical calculations on the enzyme-substrate (3a) complex
of E310G/G232S variant (Figure S12). Moreover, the com-
parison between the optimized structures of E310G and
E310G/G323S variants showed that there was no significant
difference in the binding mode of 3a.
(6a) was increased by 2.2 times relative to E310G. And it
extended the substrate scope to (S)-b-homophenylalanine
(7a), although the activity was only 5.3 mUmgÀ1. The activity
toward (R)-b-homoserine (8a) was not enhanced, and this
variant did not show measurable activity toward (S)-b-
homotyrosine (9a), (S)-b-homoglutamic acid (10a) and (S)-
b-homoglutamine (11a). These results demonstrated that the
structure-guided mutagenesis of 3,5-DAHDHcca give ready
access to enzyme variants for the target b-amino acids
although further expansion of the substrate scope is required.
Subsequently, the substrate binding pocket was modified
to better accommodate R-groups of aliphatic substrates by
using 3a as a model substrate. Subsequent mutagenesis was
carried out using E310G/G323S as the starting chassis. Three
site-directed saturation mutagenesis libraries of residues
L123, A179 and H328 locating around the R-group of 3a
(as shown in Figure S12) were constructed and screened
against substrate 3a. The substitution of L123 did not enhance
the enzyme activity toward 3a. When A179 was substituted
with S, V, Tor C, or H328 was substituted with S, V, Tor F, the
activity was enhanced by 2–3 times, and E310G/G323S/
H328N exhibited the highest activity (5.1 UmgÀ1) that was 4.6
times relative to E310G/G323S (Table S4). Next, the combi-
natorial mutagenesis was performed by the replacement of
A179 with S, V, Tand C on the basis of E310G/G323S/H328N,
which further enhanced the activities toward 3a by 14–145%
compared to E310G/G323S/H328N. The best combinatorial
variant A179S/E310G/G323S/H328N exhibited a ꢀ 200 times
enhanced activity (Figure 5), and the lower Km value com-
pared to E310G/G323S (14.3 mM vs. 30.4 mM, Table S5)
suggested a higher affinity toward 3a. The > 104-fold in-
creased activity compared to wild-type of 3,5-DAHDHcca
toward non-native substrate 3a indicated a great potential of
protein engineering for this enzyme.
b-Homologous amino acids are a class of valuable
building blocks for novel analogues of bioactive peptides.[19]
Therefore, the substrate scope of A179S/E310G/G323S/
H328N was tested toward various b-homologous amino acids
(Table S6). The results showed that this variant had good
activities toward aliphatic substrates 2a and (S)-b-homolysine
(4a) with a > 102-fold increase compared to E310G, except
the activity toward (S)-b-aminobutyric acid (5a) was en-
hanced to about 6 times. For aromatic substrates, the activity
of A179S/E310G/G323S/H328N toward (R)-b-phenylalanine
Asymmetric preparation of chiral b-amino acids
To evaluate the feasibility of the biocatalytic reductive
amination using these enzyme variants for the synthesis of b-
amino acids, the reductive aminations of the b-keto acids 5-
(methylthio)-3-oxopentanoic acid (3b) and 3-ketohexanoic
acid (2b) were scaled up to 30 mL to prepare 3a and 2a,
respectively. These b-keto acids were obtained by hydrolyzing
the corresponding b-keto esters with Novozym 435. Since b-
keto acids decompose easily in aqueous solution, the fed-
batch method was applied for the substrate loading and the
final substrate concentration was up to ꢀ 154 mM. The
recombinant E. coli cells overexpressing variant A179S/
E310G/G323S/H328N or E310G/G323S/H328N were used
as the biocatalysts and freeze-dried d-glucose dehydrogenase
(GDH) from Bacillus subtilis was used for the cofactor
recycling.
The preparation of 3a was carried out using the most
efficient variant A179S/E310G/G323S/H328N, achieving
93% analytical yield in 4 h (Figure S13) with ee > 99%
(Figure S14). For the synthesis of 2a, we speculated that the
substitution of A179 with serine in A179S/E310G/G323S/
H328N may be unfavorable for the hydrophobic substrate 2a.
Indeed, the variant E310G/G323S/H328N had higher oxida-
tive deamination activity (5.7 UmgÀ1) toward 2a than A179S/
E310G/G323S/H328N (2.0 UmgÀ1). The reductive amination
of 2b was thus performed by using variant E310G/G323S/
H328N. As shown in Figure S15 and Figure S16, 2a was
obtained at 95% analytical yield in 4 h with ee > 99%. The
optically pure products were isolated in 86% (3a) and 87%
(2a) yields, respectively (Figure S17 and Figure S18). There-
fore, these engineered b-AADHs showed great potential as
biocatalysts for the asymmetric synthesis of the target b-
amino acids.
Conclusion
The crystal structures of 3,5-DAHDHcca show that the
catalytic domain is significantly different from that of a-
AADHs.[12] The binding mode of one NADPH shared by two
monomers indicates that the substrate pocket of 3,5-
DAHDHcca is constructed by homodimer, while each
monomer of a-AADHs is a catalytic unit.[12b–d] The subse-
quent quantum chemical calculations and mutagenesis studies
provide valuable insight into the substrate binding mode and
catalytic mechanism of 3,5-DAHDHcca. It was shown that
Figure 5. The iterative stepwise improvement process of directed
evolution of 3,5-DAHDHcca toward substrate 3a. Enzyme activity of
E310G was taken as 1.0.
10208 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210

Angewandte
Research Articles
Chemie
D49, E310 and S125 contribute to the binding of the C3-
Conflict of interest
amino, C5-amino and carboxyl groups of the native substrate
1a, respectively. Moreover, residues D49 and E310 also play
important roles in capturing the 2’-hydroxyl of nicotinamide
ribose and the amide group of NADPH, respectively. The
oxidative deamination reaction of 3,5-DAHDHcca consists of
hydride transfer, the successive formation of iminium and
The authors declare no conflict of interest.
Keywords: asymmetric synthesis ·biocatalysis ·
catalytic mechanism ·protein engineering ·
b-amino acid dehydrogenase
À
hydrated intermediates, and the final cleavage of C N bond.
Although this is similar to that of a-AADHs,[6a,18] an
interesting finding of 3,5-DAHDHcca is that both the
deprotonated C5-amino group of 1a and the active site
residue D177 can be the proton acceptor to activate the water
in the hydration step. The calculations also showed that both
the hydride transfer and C-N cleavage steps are rate-limiting.
The X-ray and optimized structures further guided the
rational engineering of 3,5-DAHDHcca starting from
E310G that has been shown with an expanded substrate
scope.[11] The highly efficient variant A179S/E310G/G323S/
H328N was obtained by reconstructing the hydrogen bond
interaction of the amide of NADPH with nearby residue S323
and subsequently reshaping the substrate pocket. This variant
exhibits a ꢀ 200-fold increased activity toward 3a compared
to E310G, that is ꢀ 104-fold compared to the wild-type
enzyme. It also displays enhanced activities toward other
aliphatic substrates such as 2a and 4a by > 100-fold compared
to E310G. For 2a, another variant E310G/G323S/H328N
exhibited even ꢀ 3 times higher activity than A179S/E310G/
G323S/H328N. Finally, the synthetic abilities of these two
outstanding b-AADH variants were established by the
successful preparation of 2a and 3a with high analytical
yields (93%-95%) and excellent enantioselectivity (> 99%
ee). It is notable that the rational engineering enabled the
substrate scope expansion without trade-off of the stereose-
lectivity, an on-going challenge in enzyme engineering.
This study not only extends our understanding of
the substrate binding and catalytic mechanism of
3,5-DAHDHcca, the first characterized member of b-
AADHs, but also shows the potential of this engineered b-
AADH in the asymmetric synthesis of b-amino acids and lays
the foundation for further rational design of this enzyme to
achieve the asymmetric synthesis of target b-amino acids.
[1] a) F. Kudo, A. Miyanaga, T. Eguchi, Nat. Prod. Rep. 2014, 31,
1056 –1073; b) C. Cabrele, T. A. Martinek, O. Reiser, Ł.
Berlicki, J. Med. Chem. 2014, 57, 9718 –9739; c) R. W. Cheloha,
T. Watanabe, T. Dean, S. H. Gellman, T. J. Gardella, ACS Chem.
Biol. 2016, 11, 2752 –2762; d) L. Kiss, I. M. Mµndity, F. Fülçp,
Amino Acids 2017, 49, 1441 –1455; e) A. Henninot, J. C. Collins,
J. M. Nuss, J. Med. Chem. 2018, 61, 1382 –1414; f) V. K. Outlaw,
D. F. Kreitler, D. Stelitano, M. Porotto, A. Moscona, S. H.
Gellman, ACS Infect. Dis. 2020, 6, 2017 –2022; g) T. Katoh, T.
Sengoku, K. Hirata, K. Ogata, H. Suga, Nat. Chem. 2020, 12,
1081 –1088.
[2] a) J.-A. Ma, Angew. Chem. Int. Ed. 2003, 42, 4290 –4299; Angew.
Chem. 2003, 115, 4426 –4435; b) A. Liljeblad, L. T. Kanerva,
´
Tetrahedron 2006, 62, 5831 –5854; c) B. Weiner, W. Szymanski,
D. B. Janssen, A. J. Minnaard, B. L. Feringa, Chem. Soc. Rev.
2010, 39, 1656 –1691; d) M. Ashfaq, R. Tabassum, M. M.
Ahmad, N. A. Hassan, H. Oku, G. Rivera, Med. Chem. 2015,
5, 295 –305; e) H. Noda, M. Shibasaki, Eur. J. Org. Chem. 2020,
2350 –2361; f) R. Li, H. J. Wijma, L. Song, Y. Cui, M. Otzen, Y.
Tian, J. Du, T. Li, D. Niu, Y. Chen, J. Feng, J. Han, H. Chen, Y.
Tao, D. B. Janssen, B. Wu, Nat. Chem. Biol. 2018, 14, 664 –670;
g) F. Parmeggiani, N. J. Weise, S. T. Ahmed, N. J. Turner, Chem.
Rev. 2018, 118, 73 –118; h) N. J. Turner, Curr. Opin. Chem. Biol.
2011, 15, 234 –240.
[3] a) Q. Yin, Y. Shi, J. Wang, X. Zhang, Chem. Soc. Rev. 2020, 49,
6141 –6153; b) K. Murugesan, T. Senthamarai, V. G. Chandra-
shekhar, K. Natte, P. C. J. Kamer, M. Beller, R. V. Jagadeesh,
Chem. Soc. Rev. 2020, 49, 6273 –6328; c) T. Irrgang, R. Kempe,
Chem. Rev. 2020, 120, 9583 –9674.
[4] A. Afanasenko, T. Yan, K. Barta, Commun. Chem. 2019, 2, 127.
[5] a) S. Mathew, H. Bea, N. S. Prabhu, T. Chung, H. Yun, J.
Biotechnol. 2015, 196, 1 –8; b) S. Mathew, S.-S. Jeong, T. Chung,
S.-H. Lee, H. Yun, Biotechnol. J. 2016, 11, 185 –190; c) S.
Mathew, S. P. Nadarajan, U. Sundaramoorthy, H. Jeon, T. Chung,
H. Yun, Biotechnol. Lett. 2017, 39, 535 –543; d) G.-H. Kim, H.
Jeon, T. P. Khobragade, M. D. Patil, S. Sung, S. Yoon, Y. Won, S.
Sarak, H. Yun, ChemCatChem 2019, 11, 1437 –1440; e) X. Feng,
J. Guo, R. Zhang, W. Liu, Y. Cao, M. Xian, H. Liu, ACS Omega
2020, 5, 7745 –7750; f) I. Slabu, J. L. Galman, R. C. Lloyd, N. J.
Turner, ACS Catal. 2017, 7, 8263 –8284.
Acknowledgements
[6] a) M. Sharma, J. Mangas-Sanchez, N. J. Turner, G. Grogan, Adv.
Synth. Catal. 2017, 359, 2011 –2025; b) Y.-P. Xue, C.-H. Cao, Y.-
G. Zheng, Chem. Soc. Rev. 2018, 47, 1516 –1561.
[7] F. G. Mutti, T. Knaus, N. S. Scrutton, M. Breuer, N. J. Turner,
Science 2015, 349, 1525 –1529.
[8] a) A. S. Bommarius, K. Drauz, W. Hummel, M. R. Kula, C.
Wandrey, Biocatalysis 1994, 10, 37 –47; b) U. Kragl, W. Kruse, W.
Hummel, C. Wandrey, Biotechnol. Bioeng. 1996, 52, 309 –319;
c) U. Kragl, D. Vasic-Racki, C. Wandrey, Bioprocess Eng. 1996,
14, 291 –297.
The work was financially supported by the National Natural
Science Foundation of China (Grant Nos. 21778072 and
92056101) to D.Z., Tianjin Synthetic Biotechnology Innova-
tion Capacity Improvement Project (TSBICIP-KJGG-001) to
Q.W. , the National Key Research and Development Program
of China (2019YFA09005000 and 2018YFA0901900) to J.Z.
The authors also thank the staff of beamlines BL18U1 and
BL19U1 in Shanghai Synchrotron Radiation Facility for
access and help with the X-ray data collection.
[9] a) J. J. Baker, I. Jeng, H. A. Barker, J. Biol. Chem. 1972, 247,
7724 –7734; b) S. L. Hong, H. A. Barker, J. Biol. Chem. 1973,
248, 41 –49; c) J. J. Baker, C. van der Drift, Biochemistry 1974,
13, 292 –299; d) H. A. Barker, J. M. Kahn, L. Hedrick, J.
Bacteriol. 1982, 152, 201 –207; e) H. A. Barker, J. M. Kahn, S.
Chew, J. Bacteriol. 1980, 143, 1165 –1170; f) A. Kreimeyer, A.
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210
ꢀ 2021 Wiley-VCH GmbH
www.angewandte.org 10209

Angewandte
Research Articles
Chemie
Perret, C. Lechaplais, D. Vallenet, C. MØdigue, M. Salanoubat, J.
Weissenbach, J. Biol. Chem. 2007, 282, 7191 –7197.
[15] T. C. Pochapsky, S. S. Pochapsky, Acc. Chem. Res. 2019, 52,
1409 –1418.
[10] E. Pelletier, A. Kreimeyer, S. Bocs, Z. Rouy, G. Gyapay, R.
Chouari, D. Rivi›re, A. Ganesan, P. Daegelen, A. Sghir, G. N.
Cohen, C. MØdigue, J. Weissenbach, D. L. Paslier, J. Bacteriol.
2008, 190, 2572 –2579.
[11] D. Zhang, X. Chen, R. Zhang, P. Yao, Q. Wu, D. Zhu, ACS Catal.
2015, 5, 2220 –2224.
[12] a) P. J. Baker, K. L. Britton, P. C. Engel, G. W. Farrants, K. S.
Lilley, D. W. Rice, T. J. Stillman, Proteins Struct. Funct. Genet.
1992, 12, 75 –86; b) P. J. Baker, A. P. Turnbull, S. E. Sedelnikova,
T. J. Stillman, D. W. Rice, Structure 1995, 3, 693 –705; c) G.
Scapin, S. G. Reddy, J. S. Blanchard, Biochemistry 1996, 35,
13540 –13551; d) J. L. Vanhooke, J. B. Thoden, N. M. W. Brun-
huber, J. S. Blanchard, H. M. Holden, Biochemistry 1999, 38,
2326 –2339; e) O. Mayol, K. Bastard, L. Beloti, A. Frese, J. P.
Turkenburg, J. L. Petit, A. Mariage, A. Debard, V. Pellouin, A.
Perret, V. de Berardinis, A. Zaparucha, G. Grogan, C. Vergne-
Vaxelaire, Nat. Catal. 2019, 2, 324 –333.
[16] a) X. Sheng, M. Kazemi, F. Planas, F. Himo, ACS Catal. 2020, 10,
6430 –6449; b) F. Himo, J. Am. Chem. Soc. 2017, 139, 6780 –
6786; c) M. R. A. Blomberg, T. Borowski, F. Himo, R.-Z. Liao,
P. E. M. Siegbahn, Chem. Rev. 2014, 114, 3601 –3658.
[17] a) E. Omidinia, A. Samadi, H. Taherkhani, S. Khatami, N.
Moazami, R. R. Pouraie, Y. Asano, World J. Microbiol. Biotech-
nol. 2002, 18, 593 –597; b) Y. Zhao, T. Wakamatsu, K. Doi, H.
Sakuraba, T. Ohshima, J. Mol. Catal. B Enzym. 2012, 83, 65 –72;
c) W. Zhu, Y. Li, H. Jia, P. Wei, H. Zhou, M. Jiang, Biotechnol.
Lett. 2016, 38, 855 –861.
[18] a) J. E. Rife, W. W. Cleland, Biochemistry 1980, 19, 2328 –2333;
b) T. Sekimoto, T. Matsuyama, T. Fukui, K. Tanizawa, J. Biol.
Chem. 1993, 268, 27039 –27045; c) N. M. W. Brunhuber, J. B.
Thoden, J. S. Blanchard, J. L. Vanhooke, Biochemistry 2000, 39,
9174 –9187.
[19] S. Liu, S. H. Gellman, J. Org. Chem. 2020, 85, 1718 –1724.
[13] A. M. Lesk, Curr. Opin. Struct. Biol. 1995, 5, 775 –783.
[14] a) T. Oliveira, S. Panjikar, J. B. Carrigan, M. Hamza, M. A.
Sharkey, P. C. Engel, A. R. Khan, J. Struct. Biol. 2012, 177, 543 –
552; b) T. Oliveira, M. A. Sharkey, P. C. Engel, A. R. Khan, Acta
Crystallogr. Sect. F 2016, 72, 462 –466; c) W. Liu, Z. Li, C.-H.
Huang, R.-T. Guo, L. Zhao, D. Zhang, X. Chen, Q. Wu, D. Zhu,
ChemBioChem 2014, 15, 217 –222.
Manuscript received: December 28, 2020
Revised manuscript received: February 2, 2021
Accepted manuscript online: February 24, 2021
Version of record online: March 24, 2021
10210 www.angewandte.org
ꢀ 2021 Wiley-VCH GmbH
Angew. Chem. Int. Ed. 2021, 60, 10203 –10210