Enzymatic preparation of an R-amino acid intermediate for a γ-secretase inhibitor

Article abstract of DOI:10.1021/op400013e

(R)-5,5,5-Trifluoronorvaline, an intermediate for a γ-secretase inhibitor (BMS-708163) under development, was initially prepared from the corresponding keto acid using a commercially available d-amino acid dehydrogenase for reductive amination and glucose dehydrogenase for cofactor recycling. This amino acid could also be prepared using a d-amino acid transaminase with alanine as the amino donor, but the transamination also requires lactate dehydrogenase, NAD, formate, and formate dehydrogenase to remove pyruvate in order to bring the reaction to completion. An effective proprietary d-amino acid dehydrogenase was constructed by modification of the d-diaminopimelic acid dehydrogenase gene from Bacillus sphaericus, and a glucose dehydrogenase gene was cloned from Gluconobacter oxidans. Both genes were expressed in the same strain of Escherichia coli, and the glutamate dehydrogenase gene was inactivated in the expression strain to eliminate background production of the S-amino acid and improve the ee of the product to 100%. The amino acid could be isolated or converted without isolation to a p-chlorophenylsulfonamide carboxamide intermediate needed for the synthetic route to the γ-secretase inhibitor development candidate.

Full text of DOI:10.1021/op400013e

Article
pubs.acs.org/OPRD
Enzymatic Preparation of an R‑Amino Acid Intermediate for a
γ‑Secretase Inhibitor
Ronald L. Hanson,* Robert M. Johnston, Steven L. Goldberg, William L. Parker, and Animesh Goswami
Chemical Development, Bristol-Myers Squibb, One Squibb Drive, New Brunswick, New Jersey 08903, United States
ABSTRACT: (R)-5,5,5-Trifluoronorvaline, an intermediate for a γ-secretase inhibitor (BMS-708163) under development, was
initially prepared from the corresponding keto acid using a commercially available D-amino acid dehydrogenase for reductive
amination and glucose dehydrogenase for cofactor recycling. This amino acid could also be prepared using a D-amino acid
transaminase with alanine as the amino donor, but the transamination also requires lactate dehydrogenase, NAD, formate, and
formate dehydrogenase to remove pyruvate in order to bring the reaction to completion. An effective proprietary D-amino acid
dehydrogenase was constructed by modification of the D-diaminopimelic acid dehydrogenase gene from Bacillus sphaericus, and a
glucose dehydrogenase gene was cloned from Gluconobacter oxidans. Both genes were expressed in the same strain of Escherichia
coli, and the glutamate dehydrogenase gene was inactivated in the expression strain to eliminate background production of the S-
amino acid and improve the ee of the product to 100%. The amino acid could be isolated or converted without isolation to a p-
chlorophenylsulfonamide carboxamide intermediate needed for the synthetic route to the γ-secretase inhibitor development
candidate.
explored the use of D-amino acid dehydrogenases and D-
transaminases to convert the corresponding keto acid 2 to the
R-amino acid. This report describes alternative preparations of
1 using D-amino acid transaminase (Scheme 2) or D-amino acid
dehydrogenase (D-AADH, Scheme 1) and development of the
latter process, initially using a commercially available D-amino
acid dehydrogenase and glucose dehydrogenase (GDH).
Further development of the process included construction of
a D-amino acid dehydrogenase and cloning and expressing a
glucose dehydrogenase for NADPH regeneration together with
the D-amino acid dehydrogenase, knockout of glutamate
dehydrogenase from the Escherichia coli expression strain to
improve ee, and conversion of the R-amino acid to the required
p-chlorophenylsulfonamide carboxamide intermediate.
INTRODUCTION
■
Amyloid-β peptides (Aβ) are a major component of the plaques
that are found in the brains of Alzheimer’s disease patients and
have been proposed to play a causative role in the disease.1
These peptides are produced from amyloid precursor protein
by initial cleavage at the β site by β-secretase followed by
further cleavage at sites near the C-terminus by γ-secretase to
generate Aβ forms, with Aβ42 most closely associated with
Alzheimer’s disease and Aβ40 most abundant. Inhibition of γ-
secretase has been pursued as a pharmaceutical target to
decrease the formation of Aβ peptides.2 γ-Secretase has other
substrates and functions in addition to the processing of
amyloid precursor protein, with Notch processing an important
activity. BMS-708163, a γ-secretase inhibitor (6, Figure 1),
causes a significant decrease of Aβ40 levels at concentrations
having minimal effects on Notch signaling.2
The initial synthetic route to 6 used an asymmetric Strecker
reaction to prepare (R)-5,5,5-trifluoronorvaline amide (7,
Figure 1), a key intermediate.2 (R)-5,5,5-Trifluoronorvaline
(1, Figure 1) has been prepared by synthesis of the racemic N-
acetyl amino acid and resolution with an L-amino acid acylase,
followed by hydrolysis of the separated (R)-N-acetyl-5,5,5-
trifluoronorvaline.3 Other enzymatic resolution approaches
have also been used for preparation of R-amino acids, including
transformations with D-acylases,4 D-hydantoinases,5,6 D-ami-
dases,7 and S-amino acid oxidases combined with an excess of
reducing agent to recycle the imine product.8 When the
corresponding keto acid can be synthesized, enzymatic
reductive amination or transamination are attractive approaches
for preparation of R- or S-amino acids since a resolution is not
required. D-Amino acid dehydrogenases catalyzing reductive
aminations with broad substrate specificity have become
available as a result of directed evolution of Corynebacterium
glutamicum meso-2,6-diaminopimelic acid D-dehydrogenase,9
and D-transaminases10,11 have also been used for preparation
of R-amino acids. As an alternative approach to 1, we have
RESULTS AND DISCUSSION
■
Screen for Conversion of 5,5,5-Trifluoro-2-oxopenta-
noic Acid, 2, to (R)-5,5,5-Trifluoronorvaline, 1. Six D-
amino acid dehydrogenases (Biocatalytics) were screened for
conversion of keto acid 2 to R-amino acid 1 (Scheme 1). All of
the enzymes carried out the conversion and gave a single
enantiomer. Seven L-amino acid dehydrogenases from various
sources all gave the opposite enantiomer as a product. When
the D-amino acid dehydrogenases were tested using 1% enzyme
weight, the most effective D-amino acid dehydrogenase was D-
AADH-102 which gave complete conversion in 16 h and was
selected for further development.
Biocatalytics D-transaminase and an in-house D-transaminase
(from a soil isolate of Bacillus thuringiensis cloned and expressed
in E. coli)10 also converted keto acid 2 exclusively to R-amino
acid 1, using alanine as an amino donor. Whereas complete
conversion occurred with the amino acid dehydrogenases,
Received: January 18, 2013
Published: March 7, 2013
© 2013 American Chemical Society
693
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
Figure 1. Structure of γ-secretase inhibitor 6 and (R)-5,5,5-trifluoronorvaline 1.
Scheme 1. Reaction scheme for the preparation of R-amino acid 1 from keto acid 2 and further conversion to a synthetic
intermediate 5 for a γ-secretase inhibitor
∼50% of the keto acid remained with the transaminases. These
results are expected on the basis of the equilibrium constants
for the two types of reaction. The D-transaminases can
effectively convert keto acid 2 to R-amino acid 1, provided
NAD, formate, lactate dehydrogenase (LDH), and formate
dehydrogenase (FDH) are added to remove inhibitory pyruvate
(Scheme 2) from the reaction (Table 1). However, because of
Table 1. Preparation of R-amino acid 1 with transaminase
time
(h)
keto acid 2 R-amino acid 1 yield
D-transaminase
Biocatalytics
(mg/mL)
(mg/mL)
(%)
21
44
21
44
21
10.66
9.27
9.78
9.47
0.00
10.82
9.95
54
50
48
48
94
in-house
9.76
9.73
Biocatalytics + LDH, FDH,
NAD, formate
18.98
Scheme 2. Preparation of R-amino acid 1 with D-
transaminase
in-house + LDH, FDH,
NAD, formate
44
0.00
18.34
94
substrates. Not surprisingly, in all cases enzyme activity with the
keto amide was <1% of the activity with the keto acid. The
slight activities observed in the spectrophotometric assays may
be due to traces of acid in the amide sample. An attempted
reaction using the in-house D-transaminase did not give any
product. In addition to the lack of enzyme activity, the low
water solubility of the ketoamide makes it a problematic
substrate for an enzyme reaction.
Process Optimization. Initial development was done with
D-amino acid dehydrogenase-D-AADH-102 and a commercial
glucose dehydrogenase (Amano) for NADPH regeneration.
Reductive aminations gave solution yields by HPLC of 84, 74,
and 63% at 5, 7.5, and 10% keto acid concentrations,
respectively, with only 1−2% keto acid remaining at the end
of the reactions. Keto acid 2 was found to be stable in the
reaction buffer for 1 day at 2 mg/mL. When stability of the keto
acid was tested at higher concentrations, however, the stability
was markedly lower, suggesting occurrence of a bimolecular
reaction such as an aldol condensation. Because of the
the anticipated difficulty in separating the product 1 from the
remaining alanine and the need for adding additional enzymes,
the transamination approach was not pursued further.
The intermediate used in the initial synthesis of 6 was the
primary amide 7. The activities of the six D-amino acid
dehydrogenases and two D-transaminases were compared with
those of 5,5,5-trifluoro-2-oxopentanamide and keto acid 2 as
694
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
instability at higher concentrations, subsequent reactions were
carried out with 50 mg/mL keto acid 2 and an enzyme
concentration sufficient to complete most of the reaction in 2 h.
With keto acid 2 as substrate, the pH optimum for D-AADH-
102 is ∼9.3 (Figure 2), and the apparent Km for NADPH is
coli BL21(DE3)-Gold, the resulting D-amino acid dehydrogen-
ase was highly efficient for the reductive amination of keto acid
2 despite the fact that the sequence identity of B. sphaericus
DAPD compared to that of C. glutamicum DAPD is only 51%
(Figure 3). The amino acid sequence of the evolved B.
sphaericus D-amino acid dehydrogenase is shown in Figure 4.
Preparation of R-Amino Acid 1 with In-House D-
Amino Acid Dehydrogenase and Procedure for p-
Chlorobenzenesulfonylation of 1 and Conversion of
the Resulting Acid, 3, to the Carboxamide, 5. For the in-
house D-amino acid dehydrogenase, the Km for NADPH was
determined to be 0.63 mM, and the pH optimum was ∼9.5
(Figure 2), similar to those for D-amino acid dehydrogenase-
102, and the same reaction conditions were used for the in-
house enzyme as had been used for the commercial enzyme.
The enzymatic transformation mixture containing R-amino
acid 1 was the starting material for preparing acid 3 (Scheme 1)
by arylsulfonylation with p-chlorobenzenesulfonylchloride. Acid
3 can be isolated as a crystalline solid, but a satisfactory way to
do this in good yield was not developed. The crude acid was
converted to the acid chloride 4 with oxalyl chloride and then
to carboxamide 5 with aqueous ammonia.
Enzymatic synthesis of 1 from 2 using the in-house D-AADH
followed by N-sulfonylation to 3 and conversion to
carboxamide 5 was done at a 3.5 mol scale (600 g input of
the keto acid). The sequence was done without isolation of
intermediates, giving a 65% overall yield of 5 with a purity of
98.4 wt % and an ee of >99.9%. The ee of the intermediate
amino acid 1 was 98.9%, and thus, the final isolation by
recrystallization from n-butanol gave good purification with
respect to chiral homogeneity.
Cloning of Glucose Dehydrogenase and Develop-
ment of a Dual-Expression Strain. A putative NADP-
dependent glucose 1-dehydrogenase gene was identified from
the sequenced genome of Gluconobacter oxidans (National
Center for Biotechnology Information (NCBI) accession
number NC 006677). Primers based on this sequence were
used for PCR amplification of the gene from G. oxidans
chromosomal DNA; the gene was ligated into plasmid
pBMS2004 and expressed in E. coli strain BL21. Maximum
activity was obtained at pH 9.0 and 40 °C. Near maximal
activity was seen between pH 8.0 to 10.0 and 35 to 50 °C. The
resulting glucose dehydrogenase activity was demonstrated to
be effective for NADPH regeneration during the reductive
▲
Figure 2. pH optima for D-amino acid dehydrogenases. ( ) In-house
⧫
D-AADH; ( ) D-AADH-102.
0.49 mM. The pH optimum for glucose dehydrogenase is 8−9,
and the Km for NADP is 0.043 mM (Amano Technical
Bulletin). A pH of 9 and 0.5 mM NADP were used for the
reaction.
Development of an In-House D-Amino Acid Dehydro-
genase. meso-Diaminopimelic acid D-dehydrogenase (DAPD)
from Corynebacterium glutamicum was the starting point for
evolution of D-amino acid dehydrogenases with broad substrate
specificity reported by Vedha-Peters et al.9 DAPD has also been
reported from Bacillus sphaericus. Using the GenBank amino
acid sequence submission for a B. sphaericus DAPD to design
primers, the gene from B. sphaericus ATCC 4525 was cloned
and expressed in E. coli BL21(DE3)-Gold. From the nucleotide
sequence, the encoded protein displayed 95% identity (311
amino acids out of 327) with the GenBank submission for B.
sphaericus DAPD and was shown to catalyze the oxidative
deamination of meso-diaminopimelic acid to (S)-α-amino-ε-
ketopimelic acid but had no detectable activity for the reductive
amination of keto acid 2. Oligonucleotide mutagenesis was
used to introduce nucleotide substitutions into the B. sphaericus
DAPD gene, changing five amino acids to those previously
found to be important for broad substrate specificity in the
evolved C. glutamicum DAPD.9 After the resulting modified
gene was ligated into plasmid pBMS2004 and expressed in E.
Figure 3. Amino acid alignment of DAPD from B. sphaericus and C. glutamicum. Those residues that are changed during the subsequent mutation of
the original sequence are underlined and bolded.
695
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
Figure 4. Amino acid sequence of evolved D-AADH. Those residues that were changed by mutation of the original sequence are underlined and
bolded.
amination of 2 with either D-amino acid dehydrogenase-102 or
our in-house D-amino acid dehydrogenase under the same
conditions used for the commercial glucose dehydrogenase.
A bicistronic expression plasmid, pBMS2004-GDH+D-
AADH was subsequently created to produce both D-amino
acid dehydrogenase and glucose dehydrogenase from the same
mRNA transcript. In this construct the D-AADH gene is
downstream from the termination codon of the GDH gene and
possesses its own ribosome binding site before the ATG
initiation codon. The plasmid was transformed into E. coli,
forming dual-expression strain BL21-Gold/pBMS2004-GDH+
(D)-AADH.
Preparation of R-Amino Acid with the Dual-Expres-
sion Strain. The dual-expression strain was grown in 250-L
fermentations. An extract containing both enzyme activities was
prepared by microfluidization of a 15% w/v cell (11.75 kg cell
paste) suspension in potassium phosphate buffer, then clarified
by treatment with 0.25% polyethylene imine and centrifugation.
Two preparations of clarified extract contained 38 and 33 U/
mL of D-amino acid dehydrogenase and 2174 and 2304 U/mL
of glucose dehydrogenase and gave 87% and 88% yields of 1,
with 98.8% and 98.6% ee in 1-g use tests. The extracts were
stored at −80 °C until used for preparation of R-amino acid 1
and conversion to 5 on a 50-kg scale.12
with keto acid 2 as substrate, and activity of amino acid
dehydrogenase with α-ketoglutarate as substrate (glutamate
dehydrogenase activity). The results in Table 2 show that
Table 2. Fractionation of cell extract enzyme activities by
chromatography on Q-Sepharose
amino acid 1
dehydrogenase
(U/mL)
glutamate
dehydrogenase
(U/mL)
ee of amino acid 1
fraction
sonicate
produced (%)
89.8
98.8
21.2
0.31
1.72
0.12
flow
through
0 M
98.8
98.9
99.8
99.6
83.6
0.05
0.11
0.48
13.9
4.39
0.08
0.12
0.12
0.13
1.77
NaCl
0.1 M
NaCl
0.2 M
NaCl
0.3 M
NaCl
0.4 M
NaCl
glutamate dehydrogenase activity eluted mainly at 0.4 M NaCl
which correlated with a low ee of R-amino acid 1 produced.
Amino acid dehydrogenase activity with keto acid 2 as substrate
peaked at 0.3 M NaCl, which correlated with 99.6% ee of R-
amino acid produced. The fraction giving a product with the
lowest ee had the highest amount of glutamate dehydrogenase
activity, and the elevated activity in this fraction is believed to
result from endogenous L-glutamate dehydrogenase found in E.
coli. Low levels of glutamate dehydrogenase activity measured
in the other fractions can result from D-AADH activity with α-
ketoglutarate as a substrate. These results provided an impetus
for eliminating the L-glutamate dehydrogenase activity from the
expression strain to obtain consistently higher ee for the R-
amino acid preparations.
The TargeTron Gene Knockout System (Sigma-Aldrich) was
used to disrupt the chromosomal gdhA gene in BL21-Gold by
directed insertion of a genetic element. Colonies containing the
knockout were identified by PCR of the gdhA gene showing an
∼1000 bp size increase due to the presence of the intron and
confirmed by glutamate dehydrogenase enzyme assays showing
the depletion of the enzyme activity. The gdhA knockout strain,
BL21-Gold(gdhAminus), was used as an expression strain for the
bicistronic GDH+D-AADH plasmid construct. Growth of the
BL2Gold(gdhAminus)/pBMS2004GDH+D-AADH strain in a 15-
L fermentation produced cells used to prepare a cell extract that
was utilized to prepare R-amino acid 1 on a 1-g scale with 89%
yield and 100% ee (undetectable S-enantiomer).
Glutamate Dehydrogenase gdhA Knockout Strain.
The ee for 1 prepared using Biocatalytics D-amino acid
dehydrogenase-102 was close to 100% (S-enantiomer was not
detectable), whereas the extract containing the in-house D-
amino acid dehydrogenase from the dual-expression strain gave
98−99% ee. Although crystallization of 5 increased the ee to
99.9%, analysis13 of the ternary phase diagram of the
enantiomers and solvent indicated that 98.2% ee of 5 was
required to purge all S-enantiomer by crystallization under
equilibrium conditions.12 During development of the fermen-
tation process, some batches yielded enzyme preparations that
produced the R-amino acid with considerably lower ee.
Although in practice excellent upgrade of 5 was obtained by
crystallization (ee 75% improved to 99.6%) apparently because
of failure of the racemic compound to nucleate, achieving a
higher ee for 1 in the enzymatic process is desirable to ensure
the ee of 5 is >99.5. E. coli contains L-glutamate dehydrogenase,
and we reasoned that this endogenous enzyme was responsible
for converting some of the keto acid 2 to (S)-5,5,5-
trifluoronorvaline, resulting in a lower ee. In support of this
possibility, a commercial preparation of L-glutamate dehydro-
genase from beef liver was found to completely convert keto
acid 2 to the S-amino acid.
To further test the idea, an extract from cells that produced
R-amino acid 1 with only 89.8% ee was fractionated on a 1-mL
column of Q-Sepharose. Fractions were assayed for the ee of R-
amino acid produced, activity of amino acid dehydrogenase
696
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
a final volume of 990 μL. The reactions were initiated by the
addition of 10 μL of 100 mM NADP. The absorbance increase/
min at 340 nm was used to calculate enzyme activity. A blank
was run with no enzyme.
CONCLUSION
■
(R)-5,5,5-Trifluoronorvaline was prepared from the corre-
sponding keto acid using a D-amino acid dehydrogenase and
converted without isolation to a p-chlorophenylsulfonamide
carboxamide intermediate needed for the synthesis of a γ-
secretase inhibitor development candidate. A proprietary strain
was constructed to produce enzymes for the reductive
amination reaction. The strain expressed a D-amino acid
dehydrogenase developed by modification of B. sphaericus D-
diaminopimelic acid dehydrogenase and also expressed a
glucose dehydrogenase from G. oxidans for NADPH regener-
ation. Extracts from this strain were used for production of
intermediate 5 on a 50-kg scale. The endogenous glutamate
dehydrogenase gene gdhA was subsequently inactivated in the
expression strain to improve the ee of R-amino acid 1, resulting
in amino acid yields of 89% with no detectable S-enantiomer.
Keto Acid 2. Keto acid 2 is known14 and was prepared at
Bristol-Myers Squibb.
Transaminase Reactions. A solution, containing keto acid
2 (100 mg, 0.588 mmol), D,L-alanine (200 mg, 2.24 mmol), and
0.02 mM pyridoxal phosphate, in 0.1 M potassium phosphate
buffer, pH 7.5, was incubated with D-transaminase AT-103 from
Biocatalytics (5 mg, 44 U) or with cloned D-transaminase from
B. thuringiensis10 (0.5 mL, 10 U) at 30 °C in a total volume of 5
mL. For reactions with auxiliary enzymes to reduce pyruvate to
lactate, NAD (3.31 mg, 5 μmol), L-lactate dehydrogenase
cloned from rabbit muscle (Biocatalytics LDH-103, 0.107 mg,
15 U), and formate dehydrogenase (0.5 mL, 15 U cloned from
Pichia pastoris and expressed in Escherichia coli) were also
added. Samples were taken at 21 and 44 h for HPLC analysis.
Conversion of Keto Acid 2 to R-Amino Acid 1 with
Commercial D-Amino Acid Dehydrogenase and Glucose
Dehydrogenase. Keto acid 2 (60 g, 0.353 mol), NH4Cl (64.2
g, 1.2 mol), glucose (86.4 g, 0.48 mol), Na2CO3 (12.72 g, 0.12
mol), NADP (458 mg, 0.60 mmol), glucose dehydrogenase
(Amano, 67 mg, 10490 U), D-AADH-102 (Biocatalytics, 600
mg, 2129 U) in 975 mL of water were incubated for 19 h in a 2-
L reactor at 30 °C, pH 9 (maintained by addition of 5 N NaOH
from a pH stat). The solution yield was 50.5 g (84%), and the
ee was 100%.
EXPERIMENTAL SECTION
■
HPLC Analysis. Samples of 0.02 mL containing R-amino
acid 1 were diluted with 0. 98 mL water and placed in a boiling
water bath for 2 min. After cooling, samples were filtered into
HPLC vials. Samples were analyzed with a Regis Davankov
Ligand Exchange 15 cm × 0.46 cm column. The mobile phase
was 20% methanol/80% water/1 mM CuSO4, flow rate was 1
mL/min, detection was at 235 nm, temperature was 40 °C, and
injection volume was 10 μL. Retention times were: S-
enantiomer of 1, 4.7 min; R-amino acid 1, 12.7 min; keto
acid 2, 5.6 min.
Production of E. coli Cells and Cell Extract Containing
D-Amino Acid Dehydrogenase and Glucose Dehydro-
genase. E. coli BL21-Gold/pBMS2004-GDH+ D-AADH was
used to inoculate 1 L of MT5mod2/kan medium (40 g/L
glycerol, 20 g/L pea hydrolysate, 18.5 g/L Tastone, 6 g/L
Na2HPO4, 1.25 g/L (NH4)2SO4, 50 mg/L kanamycin sulfate)
in a 4-L shake flask. After overnight incubation at 37 °C and
225 rpm, the 1-L culture was used to inoculate 250 L of
MT5mod2/kan in a 275-L Braun fermentor, yielding an initial
OD600 of ∼0.15. The cells were grown at 37 °C, 150 L per min
air input, 10 psig pressure, 320 rpm, and pH 7.2. The pH level
was not controlled and drifted downward during the course of
the fermentation. When the OD600 of the culture reached ∼5,
sterile isopropyl-β-D-thiogalactoside (IPTG) was added to a
final concentration of 1.0 mM for induction. The fermentation
was continued until the CO2 off-gas dropped precipitously,
indicating depletion of the growth medium. The cells were
harvested by centrifugation, and the cell paste, 11.75 kg from
two pooled tanks, was analyzed for both GDH and D-AADH
activity.
For small-scale use, 180 g of frozen E. coli cell paste was
added to 60 mL of 1 M potassium phosphate buffer pH 7 in
960 mL of water to give 15% w/v cells in 50 mM potassium
phosphate buffer, pH 7. The cells were suspended in the buffer
with an Ultraturrax homogenizer, and the cell suspension was
passed through a microfluidizer one time at 12000 psi to give
1200 mL of extract. To 1100 mL of magnetically stirred extract
was added 16.5 mL of 20% w/v poly(ethyleneimine) to give
0.3% final concentration. Stirring was continued for 10 min,
and then the extract was kept on ice for another 30 min. The
extract was centrifuged for 2 min at 18000g, and the clarified
supernatant (960 mL containing 21.9 U/mL D-AADH and
1886 U/mL GDH) was stored at −20 °C until use. The
procedure was scaled up to process 11.75 kg of cells to give
In later experiments, samples were analyzed with a
Phenomenex Chirex 3126 (D-Penicillamine Ligand Exchange)
50 mm × 4.6 mm column. The mobile phase was 2 mM CuSO4
in 5% isopropanol/95% water, flow rate 1 mL/min, detection at
235 nm, temperature 40 °C, and injection volume 10 μL.
Retention times were: S-enantiomer of 1, 3.75 min; R-amino
acid 1, 5.86 min; keto acid 2, 26.3 min.
Solutions of 3 and 5 in 40% MeCN were analyzed by HPLC
on a Phenomenex Luna C18(2), 5 μm, 4.6 mm × 50 mm
column, eluting at 2 mL/min with MeCN/water/TFA,
300:700:1 and detected at 235 nm. Retention times were
7.2−7.7 min for compound 3 and 3.3 min for compound 5.
Enzyme Assays. Enzyme assays were done using 1 cm
path-length cuvettes in a spectrophotometer at 30 °C. One unit
(U) of enzyme activity gives 1 μmol of product/min. The D-
AADH assay solution contained 5 mg/mL (29.4 mM) keto acid
2, 1 M NH4Cl, 0.1 M Na2CO3, 0.2 mM NADPH at pH 9.0 in a
volume of 1 mL. After addition of a suitably diluted enzyme
solution, the absorbance decrease/min at 340 nm was used to
calculate enzyme activity. A blank was run with no keto acid.
The glucose dehydrogenase assay solution contained 0.1 M
potassium phosphate buffer pH 8, 0.5 mM NADP, and 0.1 M
glucose in a volume of 1 mL. After addition of diluted enzyme
solution, the absorbance increase/min at 340 nm was used to
calculate enzyme activity. A blank was run with no enzyme.
The glutamate dehydrogenase assay contained 5 mg/mL
(29.7 mM) α-ketoglutarate monosodium salt, 1 M NH4Cl, 0.1
M Na2CO3, and 0.2 mM NADPH at pH 9.0 in a volume of 1
mL. After addition of a diluted enzyme solution, the absorbance
decrease/min at 340 nm was used to calculate enzyme activity.
A blank was run with no keto acid.
The meso-2,6-diaminopimelic acid dehydrogenase assay
contained 100 mM glycine-KCl buffer (pH 10.5), 25 mM
meso-2,6-diaminopimelic acid, 10 μL of cell lysate, and water to
697
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
76.2 kg of enzyme solution which was stored at −80 °C until
use.
acid 1 derived from 600 g (3.53 mol) of keto acid 2. The
solution from the conversion of keto acid to amino acid was
acidified to pH 2.10 (HCl), and after the solution was held at 5
°C for several days, the precipitated protein was filtered out.
The filtrate was concentrated in vacuo to 8.5 L, adjusted to pH
12.30 (NaOH), and further concentrated in vacuo with
addition of water to remove ammonia. The resulting solution,
6.5 L, contained 556 g (3.25 mol, 92% in process yield) of R-
amino acid 1, ee 98.9%.
Conversion of Keto Acid 2 to Amino Acid 1 Using
Extract Containing D-AADH and GDH. Water (800 mL)
was added to a 2-L reactor and stirred magnetically. NH4Cl
(26.75 g, 500 mmol), glucose (62.5 g, 347 mmol), and keto
acid 2 (50 g, 294 mmol) were added, followed by 29.5 mL of
10 N NaOH. Additional NaOH was added dropwise with
continued stirring to bring the pH to 9. NADP (382 mg, 0.5
mmol) and a solution containing D-AADH (1250 U) and GDH
(75800 U) were added to start the reaction. The pH was
maintained at 9.00 with 5 N NaOH from a pH stat, and the
reaction temperature was kept at 30 °C. After 22 h the pH was
adjusted to 2.0 with 64 mL of concentrated HCl. The final
reaction mixture contained 44.5 g (260 mmol, 88.5% solution
yield, 98.9% ee) of R-amino acid 1 in 1100 mL.
Conversion of Keto Acid 2 to R-Amino Acid 1 Using
Extract Containing D-AADH and GDH from E. coli-
gdhAminus Cells. Keto acid 2 (1 g, 5.88 mmol), NH4Cl
(0.535 g, 10 mmol), glucose (1.25 g, 6.94 mmol), and water
(16 mL) were charged to a 20-mL reactor, and the mixture was
stirred with a magnet at 30 °C to dissolve the solids. NaOH (10
N, 0.59 mL) was added and then more dropwise to bring the
pH to 9. NADP (7.65 mg, 0.01 mmol) and extract from
gdhAminus cells (0.919 mL containing 25 U of D-AADH and
1319 U of GDH) were then added in that order. The reaction
mixture was stirred at 30 °C and maintained at pH 9.00 by
addition of 1 N NaOH from a pH stat. After 20 h the solution
yield of R-amino acid 1 was 0.889 g, 89% yield, 100% ee.
Isolation of R-Amino Acid 1. A reaction mixture obtained
from the keto acid to amino acid conversion using the in-house
D-AADH, 1200 g, pH 2.0, containing 44.5 g of R-amino acid 1,
was filtered to remove precipitated protein. The filtrate was
adjusted to pH 7.0 with NaOH, diluted with 1-butanol (to
prevent foaming and bumping), and concentrated in vacuo to
286 g of wet solid. This was mixed with 1430 mL of MeOH and
the mixture refluxed briefly. The hot mixture was filtered, and
the solids were washed with a little MeOH. The solids, 123 g
after drying, contained 22 g of the amino acid 1. The solids
were mixed with 500 mL of MeOH and refluxed, and the hot
mixture was filtered and washed with 100 mL of MeOH. The
remaining solids were extracted with another 500-mL portion
of methanol in the same way. The resulting methanol-insoluble
solids, 106 g, contained 7 g of residual amino acid 1.
p-Chlorobenzenesulfonyl chloride, 1099 g (5.20 mol), and
2.2 L of water were cooled to 5 °C, combined and blended with
a homogenizer to give a fine suspension of the solid (warmed
to 14 °C during the homogenization). The pH of the amino
acid solution was adjusted to 10.5 (HCl) and the p-
chlorobenzenesulfonyl chloride−water mixture added in one
portion. The reaction was stirred, keeping the pH between 10.3
and 10.7 for 8 h by addition of 10 M NaOH. Water (a total of 8
L) was also added as needed to maintain effective stirring.
Stirring was continued overnight, and the pH, 4.7, was adjusted
to 10.5 (NaOH). After an additional 5 h the mixture, 20 L, was
stirred with 8 L of MTBE and the pH adjusted to 2.5 (HCl).
The lower phase was separated and discarded. The organic
phase contained 910 g (2.63 mol, 80.9% in process step yield)
of (R)-2-(4-chlorophenylsulfonamido)-5,5,5-trifluoropentanoic
acid, 3, by HPLC analysis. A sample of 3 isolated by
crystallization from heptane−dichloromethane melted at
1
115−116.5 °C. H NMR (400 MHz, CDCl3, δ relative to
internal TMS): δ 1.89 (m, 1H), 2.15 (m, 1 H), 2.25 (m, 2H),
4.00 (dt, 1H, J = 8.6, 8.6, 4.6 Hz), 5.28 (d, 1H, J = 8.6 Hz), 7.50
(dt, 2H, J = 8.8, 2.2, 2.2 Hz), 7.79 (dt, 2H, J = 8.7, 2.3, 2.3 Hz).
Anal. Calcd for C11H11ClF3NO4S: C, 38.12; H, 3.21; Cl, 10.25;
F 16.47; N, 4.05; S, 9.27. Found: C, 38.16; H, 3.04; Cl, 10.42;
F, 16.20; N, 3.96; S, 9.27.
The solution was concentrated in vacuo, adding portions of
2-methyltetrahydrofuran (MeTHF) until the water content in
the distillate was <0.1 weight%. The solution was diluted to 4 L
with MeTHF, and oxalyl chloride (335 mL, 3.96 mol) was
added in 30-mL portions at 1 min intervals. After 30 min, four
26-mL portions of a 5 vol % solution of DMF in MeTHF were
added at 10-min intervals. By 75 min gas evolution (CO and
CO2) had stopped. The solution was concentrated in vacuo
with addition of MeTHF until the distillate gave negligible gas
evolution when mixed with water. Compound 4 was not
isolated.
The combined methanol filtrates were concentrated to
dryness in vacuo and the residue, 92 g, was dissolved in 370 mL
of water at the boiling point. To remove a small quantity of
precipitated protein, the hot solution was filtered, rinsing with
50 mL of hot water. Crystallization proceeded as the filtrate
cooled. The mixture was cooled to 4 °C and filtered, washing
with 40 mL of ice-cold water. Drying in vacuo at room
temperature gave 21.4 g of R-amino acid 1 as nacreous platelets,
ee >99.8%, yield 42.5%. Recrystallization from water gave 1, mp
289 °C (dec). 1H NMR (500 MHz, D2O, δ relative to internal
3-(trimethylsilyl)propionic acid-d4 sodium salt): δ 2.14 (m,
2H), 2.37 (m, 2H), 3.81 (t, 1H, J = 6.3 Hz). Anal. Calcd for
C5H8F3NO2: C, 35.09; H, 4.71; N, 8.19; F, 33.31. Found: C,
34.88; H, 4.76; N, 8.04; F, 33.00.
A mixture of 7.8 L of MeTHF and 7.8 L of 5 M ammonium
hydroxide was cooled to 10 °C and the acid chloride solution
(2.8 kg) added rapidly with stirring, which was continued for 10
min. The lower phase was discarded. The organic phase was
stirred and adjusted to pH 2.8 with 0.5 M sulfuric acid. The
resulting aqueous phase was discarded and the organic phase
washed with 500-mL portions of water until the extract was
neutral.
The organic phase was concentrated in vacuo with addition
of 1-butanol to replace the MeTHF, and the resulting mixture
was diluted with 1-butanol to bring the volume to 5.4 L.
Heating to 100 °C dissolved all but a small quantity of the solid.
The mixture was stirred while cooling to 0 °C over 6 h and
then at 0 °C for an additional 10 h. The mixture was filtered
and the solid washed with cold 1-butanol and dried on the
funnel with suction, giving 800 g of (R)-2-(4-chlorophenylsul-
fonamido)-5,5,5-trifluoropentanamide, 5, as an off-white free-
flowing granular solid, mp 211−213.5 °C, ee >99.9, purity 98.4
wt %, corrected weight 787 g, 2.28 mol, 86.9% step yield, 65%
p-Chlorobenzenesulfonylation of R-Amino Acid 1 and
Conversion of the Resulting Acid 3 to the Acid Chloride
4 and Then to Carboxamide 5. (R)-2-(4-Chlorophenylsul-
fonamido)-5,5,5-trifluoropentanamide, 5, was prepared from 15
kg of an enzymatic reaction solution containing R-amino acid
698
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
overall yield from 2. Characterization of 5 is presented in
reference 2.
This mutagenized plasmid preparation was used to transform
competent E. coli strain XL10-Gold, selecting transformants by
growth on LB/kan agar plates. Plasmid was prepared from 20
isolates, and DNA sequencing identified six colonies that
contained a plasmid containing an evolved version of the
DAPD with all five of the desired mutations (Q155L/D159G/
T174I/R200M/H250N). One such plasmid was selected and
used for subsequent experiments.
Presence of Glutamate Dehydrogenase and Its Effect
on Reaction Yield and Ee. Extract from cells expressing D-
AADH and GDH was prepared by sonication of 835 mg of cells
in 5 mL of 50 mM potassium phosphate buffer, pH 7. All
further steps were carried out at 4 °C. The extract was
centrifuged for 10 min at 43000g, and 2 mL of the supernatant
was added to a 1-mL column of Q-Sepharose equilibrated with
20 mM tris-chloride, pH 7.4. The column was eluted with 2-mL
portions of 20 mM tris-chloride (pH 7.4) containing 0, 0.1, 0.2,
0.3, and 0.4 M NaCl, and 2-mL fractions were collected. Each
fraction was assayed for glutamate dehydrogenase activity, D-
AADH activity, and ee of 1 produced by the fraction.
The mother liquor/wash contained 68 g of 2-(4-
chlorophenylsulfonamido)-5,5,5-trifluoropentanamide, 5. A
small portion, purified by preparative TLC (not recrystallized),
had ee 74.5%. The mother liquor therefore contained 59 g of
the(R-enantiomer. A second crop was obtained by crystal-
lization from 1-butanol, giving 30 g of (R)-2-(4-chlorophe-
nylsulfonamido)-5,5,5-trifluoropentanamide, 5, ee 99.6%, purity
98.9 wt %.
Cloning and Expression of meso-Diaminopimelic
Dehydrogenase from B. sphaericus. B. sphaericus (ATCC
4525) was grown at 30 °C in LB medium, and chromosomal
DNA was prepared from the harvested cell paste using a
standard proteinase K/SDS/NaCl/CTAB bacterial DNA
purification protocol.15 This chromosomal DNA was used for
a target for PCR reactions to amplify the B. sphaericus DAPD
gene. The GenBank amino acid sequence submission for a
comparable B. sphaericus DAPD protein (GenBank accession
number BAB07799) was used to design synthetic oligonucleo-
tides to prime amplification of the DAPD gene from
chromosomal DNA. The completed PCR reaction was analyzed
by ethidium bromide-stained agarose gel electrophoresis. A
single DNA band of about 1000 base pairs was strongly
amplified, consistent with the expected size of a B. sphaericus
DAPD gene.
The ee of 1 produced by the fraction was determined with a
reaction mixture containing in a total volume of 1 mL at pH
9.0: 5 mg/mL (29.4 mM) of keto acid 2, 0.5 M NH4Cl, 0.347
M glucose, 0.5 mM NADP, and 0.1 mL of the fraction being
assayed. After 15 h incubation at 30 °C, the ee of 1 was
analyzed by HPLC.
Knockout of gdhA Gene and Production of E. coli
gdhAminus Strain. The nucleotide sequence of the gdhA gene
from E. coli strain B, the parental strain of the BL21-Gold
expression strain, was obtained from the GenBank sequence
database. On the basis of this sequence, oligonucleotide primers
were prepared for amplification of the gdhA gene by PCR. The
amplified fragment was cloned into pCR4-TOPO and
confirmed to be the gdhA gene by DNA sequence analysis.
The nucleotide sequence from the gdhA gene was submitted
to Sigma and analyzed using their proprietary algorithm to
determine the optimal region for insertion of an intron
intended to disrupt the gdhA coding sequence. The intron-
containing plasmid pACD4 was modified using oligonucleotide
primers in order to direct disruption of the gdhA locus. All
subsequent gene knockout experiments (i.e., amplification of a
gdhA-modified intron fragment, cloning of the modified intron
into the pACD4 vector, and transformation of BL21-Gold to
disrupt the native gdhA gene) were conducted according to the
TargeTron manufacturer’s protocol. Since the pACD4 has no
selective marker to identify potential intron-disrupted colonies,
colony PCR using the original gdhA terminal primers was used
to detect which colonies contained disrupted gdhA genes.
Approximately 20% of the colonies amplified a PCR product
∼1000 bp larger than the control (the purified pCR4-TOPO
+gdhA vector) indicating they had incorporated the intron
within the gdhA coding region. One of these colonies, named
BL21-Gold(gdhAminus), was selected for further analysis.
Both BL21-Gold and BL21-Gold(gdhAminus) were grown in
shake flasks containing MT5(mod2) medium at 37 °C to late-
log phase. Cell pellets of two wild-type (parental strain of the
BL21-Gold expression strain) and two knockout samples were
suspended in 5 mL of 50 mM potassium phosphate buffer pH
7, sonicated for 3 min, and then centrifuged for 15 min at
48000g. The supernatants were assayed for glutamate
dehydrogenase activity. Glutamate dehydrogenase activity in
knockouts measured in this assay was <10% of wild-type
control activity. The residual activity (ΔA 340/min) found in
The DNA from the band was purified and ligated into
cloning vector pCR4-TOPO (Invitrogen), using the manufac-
turer’s protocol, and then transformed into E. coli strain TOP10
by electroporation. Plasmid DNA purified from transformed
cells was used for DNA sequence analysis using plasmid based
primers that flanked the fragment insert site in PCR4-TOPO.
The DAPD gene was cloned into expression vector pET30a
and transformed into electrocompetent E. coli strain BL21-
(DE3)-Gold (Agilent, Santa Clara, CA). A single positive
colony of the transformed strain was grown overnight in
MT5mod2/kan broth at 37 °C, 250 rpm and then subcultured
in the same medium at 30 °C at a starting optical density at 600
nm (OD600) of 0.15. At OD600 of 1.0, isopropyl β-D-
thiogalactoside (IPTG) was added to a final concentration of
200 μM to induce transcription of the DAPD gene from the T7
promoter. Growth was continued overnight at 30 °C. A cell
extract derived from the DAPD expression culture displayed a
novel, highly overexpressed protein with an apparent molecular
weight of ∼36 kD on an SDS gel. The extract catalyzed the
oxidative deamination of meso-2,6-diaminopimelic acid to (S)-
α-amino-ε-ketopimelate. The oxidative deamination activity
level (10.3 U/mg protein) in the recombinant DAPD E. coli
expression strain was about 1000-fold higher than the level
observed in the native strain used to isolate the DAPD gene,
clearly demonstrating that the cloned gene encoded a DAPD.
Construction of the Evolved D-AADH Enzyme. A
QuikChange Multisite Mutagenesis kit (Stratagene) was used
to introduce nucleotide substitutions into the B. sphaericus
DAPD gene, altering the amino acid sequence of the encoded
protein to include five amino acid mutations. Four specific
mutagenic oligonucletide primers were used to initiate PCR
reactions that amplified a modified version of the native DAPD
gene. Each primer replicated the sense strand of the DAPD
gene in the region flanking the site of the desired mutation but
substituted an alternate codon at the point of the amino acid
change.
699
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700

Organic Process Research & Development
Article
Decicco, C. P.; Macor, J. E.; Olson, R. E. ACS Med. Chem. Lett. 2010,
1, 120.
(3) Ojima, I.; Koji, K.; Nakahashi, K. J. Org. Chem. 1989, 54, 4511.
(4) Moriguchi, M.; Ideta, K. Appl. Environ. Microbiol. 1988, 54, 2767.
(5) Drauz, K. Chimia 1997, 51, 310.
(6) Pietzsch, M.; Syldatk, C. In Enzyme Catalysis in Organic Synthesis,
2nd ed.; Drauz, K., Ed.; Wiley-VCH: Weinheim, 2002; Vol. 2, p 761.
(7) Komeda, H.; Asano, Y. Eur. J. Biochem. 2000, 267, 2028.
(8) Alexandre, F.-R.; Pantaleone, D. P.; Taylor, P. P.; Fotheringham,
I. G.; Ager, D. J.; Turner, N. J. Tetrahedron Lett. 2002, 43, 707.
(9) Vedha-Peters, K.; Gunawardana, M.; Rozzell, J. D.; Novick, S. J. J.
Am. Chem. Soc. 2006, 128, 10923.
(10) Hanson, R. L.; Davis, B. L.; Goldberg, S. L.; Johnston, R. M.;
Parker, W. L.; Tully, T. P.; Montana, M. A.; Patel, R. N. Org. Process
Res. Dev. 2008, 12, 1119.
the knockouts may be due to the presence of a ketoreductase
rather than glutamate dehydrogenase.
Expression of D-AADH and GDH in the E. coli-
gdhAminus Strain. E. coli BL21-Gold(gdhAminus) was trans-
formed with plasmid pBMS2004-GDH+D-AADH. Transform-
ants were identified by growth on LB/kan agar plates and
verified by colony PCR specific for amplification of the GDH
+D-AADH gene cassette. A transformant was grown in
MT5(mod2)/kan medium with induction at OD600 ∼1.5 with
0.2 mM IPTG and continued overnight growth. SDS-PAGE
and enzyme activity analyses of an extract of the induced cells
revealed that both the GDH and D-AADH proteins were highly
overexpressed in the gdhAminus background.
Production of E. coli-gdhAminus Producing Both D-
AADH and GDH. E. coli BL21-Gold(gdhAminus)/
pBMS2004GDH+D-AADH was used to inoculate 500 mL of
MT5mod2/kan broth in a 2-L shake flask. The culture was
grown overnight at 37 °C and 225 rpm, and then used to
inoculate 15 L of MT5mod2/kan in a 21-L Braun fermentor,
yielding an initial OD600 of ∼0.25. The cells were grown at 37
°C, 150 LPM air input, 10 psig pressure, 320 rpm, at pH 7.2.
The pH level was not controlled and drifted downward during
the course of the fermentation. When the OD600 of the culture
reached ∼5, sterile IPTG was added to a final concentration of
1.0 mM. The fermentation was continued until the CO2 off-gas
dropped precipitously, indicating depletion of the growth
medium. The cells were harvested by centrifugation, and the
cell paste was stored at −80 °C.
Preparation of D-AADH and GDH Enzyme Mixture
from E. coli gdhAminus Cells. A 15% w/v suspension of the
harvested culture was prepared in 50 mM potassium phosphate
buffer, pH 7.0 (6.0 g frozen cells +34 mL buffer). The cells
were suspended in the buffer with an Ultraturrax T25
homogenizer and then sonicated for 3 min. The extract was
centrifuged for 15 min at 48000g, 4 °C. The supernatant
contained 27.2 U/mL D-AADH and 1436 U/mL GDH.
Other Methods. Expression of evolved D-AADH in E. coli,
cloning and expression of the G. oxidans glucose dehydrogenase
gene, and construction of the E. coli strain expressing both
evolved D-AADH and glucose dehydrogenase have been
described.16
(11) Taylor, P. P.; Pantaleone, D. P.; Senkpeil, R. F.; Fotheringham,
I. G. Trends Biotechnol. 1998, 16, 412.
(12) Cassidy, M.; Chen, C.-K.; Goswami, A.; Guzikowski, S.;
Hanson, R.; Johnston, R.; Lai, C.; Okuniewicz, F.; Parker, W.; Razler,
T.; Randazzo, M.; Reiff, E.; Xu, X. Manuscript in preparation.
(13) Chen, A. M.; Wang, Y.; Wenslow, R. M. Org. Process Res. Dev.
2008, 12, 271.
(14) Urata, H.; Ishii, Y.; Fuchikami, T. Tetrahedron Lett. 1989, 30,
4407.
(15) Ausubel, F. M.; Brent, R.; Kingston, R. E.; Moore, D. D.;
Seidman, J. D.; Smith, J. A.; Struhl, K. Current Protocols in Molecular
Biology; John Wiley and Sons: New York, NY, 1990; Vol. 1,
Supplement 27, section 2.4.1.
(16) Goswami, A.; Goldberg, S. L.; Johnston, R. M.; Hanson, R. L.;
Parker, W. L. Mutant amino acid dehydrogenase and its use in
preparing amino acids from keto acids. Patent Application No. WO/
2012/106579, 2012. Chem. Abstr. 2012, 157, 354578.
AUTHOR INFORMATION
Corresponding Author
*Telephone: (732)-227-6222. E-mail: ronald.hanson@bms.
com.
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Hyei-Jha Chung, Thomas Razler, and Michael
Cassidy for development of procedures to synthesize keto acid
2. We thank Emily Reiff and Stephen Guzikowski for
optimization of the conversion of amino acid 1 to intermediate
5, and we thank Michael Randazzo and Ramesh Patel for
helpful advice on all of the work.
REFERENCES
(1) Roberts, S. B. Adv. Drug Delivery Rev. 2002, 54, 1579.
(2) Gillman, K. W.; Starrett, J. E., Jr.; Parker, M. F.; Xie, K.; Bronson,
J. J.; Marcin, L. R.; McElhone, K. E.; Bergstrom, C. P.; Mate, R. A.;
Williams, R.; Meredith, J. E.; Burton, C. R.; Barten, D. M.; Toyn, J. H.;
Roberts, S. B.; Lentz, K. A.; Houston, J. G.; Zaczek, R.; Albright, C. F.;
■
700
dx.doi.org/10.1021/op400013e | Org. Process Res. Dev. 2013, 17, 693−700