Efficient racemization of N-phenylacetyl-D-glufosinate for L-glufosinate production

Article abstract of DOI:10.1002/chir.23076

Most amino acids contain chiral centres and exist as both D-enantiomer and L-enantiomer. The optically pure enantiomer is often more valuable than the racemate. Enzymatic resolution provides an effective strategy to obtain optically pure amino acids but often results in large amounts of unwanted isomer. In this study, optically pure L-glufosinate (L-PPT) was obtained by coupling amidase-mediated hydrolysis of N-phenylacetyl-D,L-glufosinate with racemization of N-phenylacetyl-D-glufosinate (NPDG), which exclusively exhibits effective herbicidal properties compared with its D-enantiomer. To improve the yield of L-PPT, the racemization reaction conditions were optimized, and through single-factor experiments, the optimal reaction temperature, reaction time, and mole ratio of phenylacetic acid to NPDG were determined to be 150°C, 30?minutes, and 1.5, respectively. The response surface methodology was applied to further optimize the racemization conditions, and the final yield of L-PPT reached 96.13% with optimum reaction temperature of 154°C, reaction time of 23?minutes, and phenylacetic acid/NPDG mole ratio of 1.7, respectively. Moreover, adding a small amount of acetic anhydride further raised the yield of L-PPT to 97.02%.

Full text of DOI:10.1002/chir.23076

Received: 18 January 2019
DOI: 10.1002/chir.23076
Revised: 7 March 2019
Accepted: 9 April 2019
R E G U L A R A R T I C L E
Efficient racemization of N‐phenylacetyl‐D‐glufosinate for
L‐glufosinate production
1
,2
1,2
1,2
1,2
Jian‐Miao Xu
| Fang‐Long Li | Ya‐Ping Xue
| Yu‐Guo Zheng
1
Engineering Research Center of
Bioconversion and Biopurification of
Ministry of Education, Zhejiang
University of Technology, Hangzhou, PR
China
Abstract
Most amino acids contain chiral centres and exist as both D‐enantiomer and L‐
enantiomer. The optically pure enantiomer is often more valuable than the
racemate. Enzymatic resolution provides an effective strategy to obtain opti-
cally pure amino acids but often results in large amounts of unwanted isomer.
In this study, optically pure L‐glufosinate (L‐PPT) was obtained by coupling
amidase‐mediated hydrolysis of N‐phenylacetyl‐D,L‐glufosinate with racemiza-
tion of N‐phenylacetyl‐D‐glufosinate (NPDG), which exclusively exhibits effec-
tive herbicidal properties compared with its D‐enantiomer. To improve the
yield of L‐PPT, the racemization reaction conditions were optimized, and
through single‐factor experiments, the optimal reaction temperature, reaction
time, and mole ratio of phenylacetic acid to NPDG were determined to be
2
Key Laboratory of Bioorganic Synthesis
of Zhejiang Province, College of
Biotechnology and Bioengineering,
Zhejiang University of Technology,
Hangzhou, PR China
Correspondence
Ya‐Ping Xue, Engineering Research
Center of Bioconversion and
Biopurification of Ministry of Education,
Zhejiang University of Technology,
Hangzhou 310014, PR China.
Email: xyp@zjut.edu.cn
150°C, 30 minutes, and 1.5, respectively. The response surface methodology
was applied to further optimize the racemization conditions, and the final yield
of L‐PPT reached 96.13% with optimum reaction temperature of 154°C, reac-
tion time of 23 minutes, and phenylacetic acid/NPDG mole ratio of 1.7, respec-
tively. Moreover, adding a small amount of acetic anhydride further raised the
yield of L‐PPT to 97.02%.
Funding information
National Natural Science Foundation of
China, Grant/Award Number: 21676254
KEYWORDS
enantiomer, L‐glufosinate, optically pure, optimization, racemization, RSM‐BBD
1
| INTRODUCTION
optically pure enantiomer is of great significance. Kinetic
resolution is a widely used method which can afford
As the basic units of protein, amino acids are widely used
in pharmaceutical, food, and cosmetics industries. Most
amino acids contain chiral centres and exist as two enan-
almost enantiopure compounds with a theoretical yield
of 50%, while the dynamic kinetic resolution can over-
come the yield limit by racemization of the unreacted
1
5
tiomers, D‐ and L‐, which have different or even opposite
enantiomer. Therefore, to obtain a maximum theoretical
2
physiological activities. L‐amino acids can be directly
yield of optical pure isomer from the racemate, the race-
mization approaches are required. In some cases, racemi-
zations have to be carried out in a separate pot to
overcome its incompatibility with kinetic resolution
methods.
Generally, complete racemization of the unwanted iso-
mer is of critical importance for economically and envi-
ronmentally acceptable resolution. In the development
utilized by organisms, so they play an important role in
3
the feed industry. As important raw materials in the
pharmaceutical and food fields, D‐amino acids and their
derivatives are widely used in the synthesis of semisyn-
thetic antibiotics, peptide hormones, pyrethroids, insecti-
4
cides, and sweeteners. Because of the functional
differences between enantiomers, separation of an
Chirality. 2019;1–9.
wileyonlinelibrary.com/journal/chir
© 2019 Wiley Periodicals, Inc.
1

2
XU ET AL.
of an industrial resolution procedure, the racemization of
the unwanted enantiomer often presents as one of the
most difficult obstacles, which require harsh conditions,
of continuous circulation, and the losses it causes are
gradually accumulated; therefore, the current reported
yield is not enough economical. Optimization of the
racemization conditions is thus an important goal.
Response surface methodology (RSM) is a powerful
statistical‐based technique for modeling complex systems,
evaluating the simultaneous effects of several factors, and
thus searching the optimum conditions for desirable
6,7
leading to product decomposition. The racemization
of amino acids and their derivatives mainly involves ther-
8
9
mal racemization, base‐catalyzed racemization, acid‐
10
catalyzed racemization, and enzyme‐catalyzed racemi-
11
zation. Thermal racemization is a very common method
because of its operational simplicity and low cost, espe-
cially in industrial applications. However, decomposition
of the substrate is a recurring problem when high temper-
atures are required. Glufosinate (PPT) is an amino acid
derivative with effective herbicidal properties exclusively
in the L‐form, while the enantiomeric D‐form is ineffec-
16
responses. The Box‐Behnken design (BBD), one of the
important RSM experimental designs, is based on a
three‐level factorial design with its experimental points
located on a hypersphere equidistant from the central
17
point. In this study, the above one‐factor‐at‐a‐time
experiment clearly demonstrated the effects of single fac-
tors (reaction temperature, time, and mole ratio of
phenylacetic acid to NPDG) on the yield of L‐PPT. How-
ever, the racemization of NPDG is controlled by the inter-
action of all three factors. For example, as the reaction
temperature increased, the mole ratio and reaction time
could decrease accordingly. Therefore, BBD was used to
investigate and optimize the influence of the key process
variables of reaction temperature, time, and mole ratio of
phenylacetic acid to NPDG on the yield of L‐PPT.
12
tive. Compared with glyphosate, PPT has better herbi-
cidal effects on some perennial malignant weeds and is
more competitive for crops of high economic value. The
key enzyme involved in PPT synthesis is glutamine syn-
thetase, which can eliminate the toxicity of ammonium
released from nitrate reduction, amino acid degradation,
13,14
and photorespiration.
Artificial PPT is always produced as a racemic com-
pound and needs to be optically resolved. There are sev-
eral ways for the bio‐enzymatic production of L‐PPT. As
an example, α‐chymotrypsin was used to resolve
bialaphos‐ethyl ester to prepare L‐PPT. However, the
atomic economy is poor because D‐enantiomer cannot
2
| MATERIALS AND METHODS
.1 | Materials
2
15
be reused after derivatization. Catalyzation of N‐ace-
tyl‐glufosinate by deacetylase is another way to obtain
L‐PPT. However, there are some problems such as
insufficient enzyme activity and difficulties in separation
of D‐isomer. In contrast, the use of acylase to catalyze
the resolution of N‐phenylacetyl‐glufosinate (NPG)
avoids above problems. In this way, the D‐isomer is
easy for separation, racemization, and recycling. Enzy-
matic resolution provides as a promising approach to
obtain optically pure L‐PPT. However, while the above
method can be used for the racemization of N‐
phenylacetyl‐D‐glufosinate (NPDG), in most cases, no
robust enzymes are available. There are several advan-
tages to racemic with phenylacetic acid. Firstly, it is
not necessary to introduce additional substance. Sec-
ondly, the reaction is very efficient. Finally, after the
conditions are optimized, the yield is higher than the
other reported racemization methods. For example,
Fed et al reported the use of anhydrous inorganic acids
such as phosphoric acid or sulphuric acid to racemize
NPDG at temperatures ranging from 100°C to 150°C,
and the racemization degree reached 100% with 90%
yield. Fed et al mixed NPDG with phenylacetic acid
and heated the reaction mixture to 180°C for 30 minutes,
resulting in 92% yield and 100% degree of racemization.
For the production of L‐PPT, racemization is a process
NPG was supplied by Shandong Luba Chemical Co., Ltd.
Phenylacetic acid and dichloromethane were both
obtained from Yonghua Chemical Technology Co., Ltd.
and were of analytical grade. HPLC grade methanol was
used as the mobile phase in HPLC analysis and was pur-
chased from J&K Scientific, Ltd. Immobilized amidase
was provided by our laboratory.
2.2 | Optical rotation detection
NPDG (0.5 g) was dissolved in 1 mol/L aqueous ammonia
and diluted to give a concentration of 10 g/L. Its optical
rotation was measured by a polarimeter, and the specific
rotation was calculated, which was then used as the
starting data. After the racemization reaction was com-
pleted, the sample was again formulated into a solution
of 10 g/L, and the optical rotation was again detected
and the specific rotation calculated. Finally, Equation (1)
was used to calculate the degree of racemization.
Inital specific rotation‐Final specific rotation
Racemization degree ¼
Inital specific rotation
(1)

XU ET AL.
3
2
.3 | Detection of PPT
2
× Molar quantities of L‐PPT
Yield of L‐PPT ¼
Molar quantities of NPDG ðbefore racemizationÞ
To separate optically pure L‐PPT, a sample was first
reacted with o‐phthalaldehyde/N‐acetyl‐L‐cysteine
OPA/NAC) to yield a pair of diastereomers. Then, the
(3)
(
2.7 | Optimization of mole ratio of
chiral separation was carried out on a Waters C18 col-
umn with a mixture of 0.05 mol/L ammonium acetate
buffer (pH 5.7)‐methanol (90:10, v/v) as a mobile phase
at a flow rate of 1.0 mL/min. The column temperature
was 35°C with UV‐Vis detection at 232 nm. The peak
times of L‐PPT and D‐PPT were 9.0 and 12.0 minutes,
respectively.
phenylacetic acid to NPDG
The racemization was carried out at 150°C for 30 minutes
with mole ratio of phenylacetic acid to NPDG as 0.50,
0.75, 1.00, 1.25, 1.50, 1.75, and 2.00, respectively. After
the racemization, the yield of NPG and L‐PPT were calcu-
lated consistently with the above method “optimization
of racemization temperature.”
2.4 | Detection of NPG
2.8 | Optimization of racemization time
The mobile phase used for the detection was 40% metha-
nol with pH 2.8 adjusted by phosphoric acid. NPG was
detected by UV‐Vis at 230 nm with the column tempera-
ture of 30°C and the flow rate of 1.0 mL/min. The peak
time of NPG was 7.5 minutes.
The reaction time for the racemization were set as 10, 20,
30, 40, 50, and 60 minutes, respectively, with the temper-
ature set at 150°C and mole ratio held at 1.5. After the
racemization, the yield of NPG and L‐PPT were calcu-
lated consistently with the above method “optimization
of racemization temperature.”
2.5 | Racemization of NPDG
First, 0.1 mol NPDG was mixed with 0.15 mol
phenylacetic acid, and the mixture was heated rapidly
to 170°C with stirring. The mixture was kept at this tem-
perature for 20 minutes. After the reaction was com-
pleted, phenylacetic acid was removed after extraction
with dichloromethane, and then the pH was adjusted to
2.9 | Response surface methodology to
optimize reaction conditions
To study the mutual effects of the variables on the race-
mization reaction and optimize the reaction process, an
RSM experimental design based on the BBD was used.
2
to obtain the precipitate of racemization products.
To describe the effects of temperature (X ), mole ratio
1
Finally, the optical rotation was detected by polarimetry.
(X ), and time (X ) on the yield of L‐PPT (Y), batch exper-
2
3
iments were conducted based on the BBD. The low, cen-
ter, and high levels of each variable were designated as
2
.6 | Optimization of racemization
−1, 0, and +1, respectively, as shown in Table 1. The
temperature
dependent variable was the yield of L‐PPT. The response
variable was fitted to a nonlinear quadratic model as fol-
lows:
The racemization of NPDG was carried out at 120°C,
1
30°C, 140°C, 150°C, 160°C, and 170°C, respectively,
with mole ratio of phenylacetic acid to NPDG fixed at
.5 and reaction time fixed for 30 minutes. After the
Y ¼ β þ ∑ β X þ ∑ βiiX2i
k
k
i¼1
0
i¼1
∑
i
i
1
k−1
i¼1
k
þ ∑
β X X þ ε
(4)
i
j
racemization reaction was completed, the mixture was
extracted twice with dichloromethane, and the pH of
the solution was adjusted to 8.5 with substrate concen-
tration of 150 g/L. Then, 2.5 g immobilized amidase
was added and reacted at 30°C for 8 hours. Finally,
the degree of racemization, the concentration of NPG,
and the concentration of L‐PPT were detected. The yield
of NPG and L‐PPT were calculated by the Equations (2)
and (3).
j¼iþ1 ij
where Y is the predicted response, X , X are the coded
i
j
independent variables, β is a constant, β is the linear
0
i
TABLE 1 Box‐Behnken design of variables and their levels
Coded Levels of Variables
Factors
Variables
−1
0
1
X
1
X
2
X
3
Temperature (°C)
Mole ratio
145
155
165
Molar quantities of NPG ðafter racemizationÞ
Yield of NPG ¼
1.0
10
1.5
20
2.0
30
Molar quantities of NPDG ðbefore racemizationÞ
Time (min)
(
2)

4
XU ET AL.
effect, β is the quadratic effect, β is the coefficient of the
interaction factor, and ε is the random error, which
which was consistent with the HPLC analysis results.
Therefore, comelting with phenylacetic acid was sug-
gested as an effective procedure for the racemization of
NPDG (Scheme 1).
ii
ij
allows for description of uncertainties between predicted
18
and measured data.
The optimization procedure
involves studying the response to a set of statistically
designed parameter combinations, estimating the coeffi-
cients by fitting the experimental data to the response
functions, predicting the response of the fitted model,
and checking the adequacy of the model.
3.2 | Optimum temperature of
racemization
The racemization of NPDG mainly occurs after its
comelting with phenylacetic acid. The reaction tempera-
ture and phenylacetic acid/NPDG mole ratio are the
main factors affecting the melting of NPDG. The temper-
ature not only affects the melting of the NPDG but also
affects the rate of racemization. In addition, temperature
also affects the side reaction rate. As shown in Figure 1,
with the change of temperature, the yield of NPG and
the degree of racemization showed completely opposite
trends. Higher temperature led to a higher degree of race-
mization but resulted in greater loss of NPG because it
promoted the side reaction. Accordingly, the yield of L‐
PPT first increased and then declined. The yield was
increased rapidly with temperature increased from
120°C to 150°C, and a maximum yield of 92.89% was
obtained at 150°C. When the temperature exceeded
150°C, the yield of L‐PPT began to decrease. This can be
attributed to the decomposition of NPG at high tempera-
tures. This phenomenon indicated that, when the mole
ratio is 1.5, NPG is eutectic with phenylacetic acid while
the temperature exceeds 150°C and the racemization
2
.10 | Effect of adding acetic anhydride on
racemization
NPDG (0.1 mol) was mixed with 0.17 mol phenylacetic
acid, and the mixture was heated rapidly to 149°C.
Afterthen, 0.02 mol acetic anhydride was added, and the
mixture was kept at this temperature for 13 minutes.
Then, the optical rotation and the concentration of NPDG
were detected.
3
3
| RESULTS AND DISCUSSION
.1 | Racemization of NPDG
The specific rotation of the NPDG and the postreaction
2
2
sample were detected to be 23.2 °·cm /g and 0 °·cm /g,
respectively. According to eq. 1, the degree of racemiza-
tion was 100%, which indicated that half of the D‐form
NPDG was converted into the L‐form by the reaction,
SCHEME 1 The racemization of NPDG
FIGURE 1 Effect of temperature on the
degree of racemization, the yield of NPG,
and the yield of L‐PPT

XU ET AL.
5
reaction mainly occurs after the mixture melts. Therefore,
the temperature of 150°C was deemed optimum.
yield of 92.70%. However, the L‐PPT yield decreased
above the mole ratio of 1.5. This indicated that excess
phenylacetic acid led to the loss of NPG, resulted in
increased economic cost. Hence, the mole ratio of 1.5
was selected.
3.3 | Optimum mole ratio of phenylacetic
acid to NPDG
Phenylacetic acid functioned as a solvent, as well as an
acid in the reaction. The melting point of phenylacetic
acid is 78°C, and it can be eutectic with NPDG, which
lowers the melting temperature of NPDG. The more the
amount of phenylacetic acid in the mixture, the easier
the NPDG will be melted, while the higher the acidity
of the reaction system made, the more efficient the race-
mization reaction. As shown in Figure 2, when the mole
ratio of phenylacetic acid to NPDG was increased from
3.4 | Optimum reaction time of
racemization
Complete racemization requires long‐time reaction; how-
ever, overreaction will increase the side reaction and
cause the loss of NPG. Generally, longer reaction times
increased the degree of racemization but lowered the
yield of NPG. As shown in Figure 3, the yield of L‐PPT
increased markedly as the reaction time was increased
from 10 to 30 minutes and then declined from 30 to
60 minutes. The maximum L‐PPT yield of 92.58% was
obtained within 30 minutes reaction, while with 10 and
0
.5 to 1.5, the racemization degree was increased, and
the NPG yield was decreased. The L‐PPT yield was
highest when the mole ratio was 1.5, with a quantitative
FIGURE 2 Effect of phenylacetic acid/
NPDG mole ratio on the degree of
racemization, the yield of NPG, and the
yield of L‐PPT
FIGURE 3 Effect of time on the degree
of racemization, the yield of NPG, and the
yield of L‐PPT

6
XU ET AL.
6
0 minute reaction time, the yields only reached 49.51%
TABLE 3 Analysis of variance (ANOVA) for quadratic model
Sum of Mean F‐ P‐Value
and 82.75%, respectively. Therefore, the optimum reac-
tion time was determined to be 30 minutes.
Variables
Squares Square Value Prob>F Significant
Model
2856.69 317.41 424.36 <0.0001 Significant
X
X
X
X
X
X
X
1
2
3
1
1
2
1
332.3
332.3
444.27 <0.0001
3
.5 | Response surface methodology to
354.45 354.45 473.88 <0.0001
396.63 396.63 530.28 <0.0001
182.52 182.52 244.02 <0.0001
565.49 565.49 756.03 <0.0001
278.72 278.72 372.64 <0.0001
509.27 509.27 680.86 <0.0001
optimize reaction conditions
Table 2 shows the design matrix in the Box‐Behnken
model, along with the experimental and predicted values
of the response (yield of L‐PPT). The center point (0, 0, 0)
was repeated five times, and the same predicted value
was obtained each time, demonstrating a perfect fit
between the factors and response. The precise values of
the five replicates differed slightly, but within the margin
of error. Using the Design Expert software, a multiple
regression analysis of the experimental data was per-
formed to obtain the coefficients and the corresponding
equation to estimate the response. The obtained second‐
order polynomial equation shows the relationship
between the yield of L‐PPT and the three variables:
X
2
3
3
X
X
2
2
2
X2
90.27
85.84
5.24
90.27 120.69 <0.0001
85.84 114.77 <0.0001
0.75
X
3
19
Residual
Lack of fit
3.87
1.29
3.79 0.1153 Not
significant
Pure error
1.36
0.34
Corrected
total
2861.93
Y ¼ −3474:99 þ 39:14X1 þ 311:67X2 þ 23:44X3
2
1
−
1:35X X − 0:12X X3 − 1:67X X − 0:11X
1
2
1
2
3
Analysis of variance (ANOVA) was performed, as
summarized in Table 3, to examine the goodness of the
2
2
−
18:52X − 0:045X :
2
3
20,21
model and the significance of the variables.
The value
2
of R (0.9982) shows that the response function predic-
tions agree well with the experimental data, implying that
99.82% of the total variation could be explained by the
TABLE 2 Experimental design with response of variables
Temperature, Mole Time, Actual Predicted
2
regression model. The value of adjusted R (0.9958) is in
2
reasonable agreement with the value of predicted R
°
C
Ratio min Value, % Value, %
(
0.9776), which further illustrates the good predictive
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
145
165
145
165
145
165
145
165
155
155
155
155
155
155
155
155
155
1.0
1.0
2.0
2.0
1.5
1.5
1.5
1.5
1.0
2.0
1.0
2.0
1.5
1.5
1.5
1.5
1.5
20
20
20
20
10
10
30
30
10
10
30
30
20
20
20
20
20
58.40
85.42
84.19
84.19
52.75
88.80
91.31
79.80
62.30
93.36
92.40
90.05
94.50
93.33
93.02
93.53
94.01
58.19
84.59
85.02
84.40
52.79
89.46
90.65
79.76
62.49
92.50
93.26
89.88
93.68
93.68
93.68
93.68
93.68
ability of the model. The statistical significance of the
quadratic model was calculated based on the F‐value
22
and the Prob>F. The F‐value measures the variance of
the data about the mean, expressed as the ratio of the
mean square of group variance to the mean square of
the error. A model term is said to be significant if it pos-
sesses a high F‐value and Prob>F less than 0.05. As
shown in Table 3, the P‐value of the model is
P < 0.0001, implying that the model is significant. In
addition, the linear effects of the coefficients X , X , and
X , the interactive effects of the coefficients X X , X X ,
and X X , and the quadratic terms X , X , and X are
significant model terms with respect to the response
because they all have P‐values of P < 0.0001.
The model indicated that there is only a 0.01% chance
that a model F‐value this large could occur because of
noise. The F‐value (3.79) and P‐value (0.1153) of lack‐of‐
fit indicated that the lack‐of‐fit had an insignificant effect
on the pure error and that the model was well fitted over
the entire regression area.
2
3
1
2
3
1
2
1 3
0
1
2
3
4
5
6
7
2
1
2
2
2
3
2
3

XU ET AL.
7
Model adequacy diagnostics were used to further ana-
was set at a constant level while the other two factors
were changed within the experimental ranges.
24
lyze the applicability of the model. Figure 4A shows the
diagnostic diagrams for model adequacy. The data points
of the actual experiment and the predicted values show a
normal distribution and are close to the 45° lines, demon-
The undersurface of the 3D response surface plots was
the corresponding contour plot. Significant interaction
effects were observed between the temperature, mole
ratio, and time, particularly between the temperature
and time. As shown in Figure 5B, the yield of L‐PPT ini-
tially increased with increasing temperature when the
reaction time was longer than 20 minutes; however, with
further increase of temperature, the yield decreased. As
shown in Figure 5A, the same trend was observed
between the temperature and mole ratio. These results
indicate that the use of an intermediate temperature in
combination with suitable mole ratio and reaction time
could lead to an increase in the yield of L‐PPT. Figure 5
C shows that the interaction between mole ratio and reac-
tion time is very significant because the same yield of L‐
PPT could be obtained by either a long reaction time with
low mole ratio or a short reaction time with high mole
ratio. Furthermore, the yield of L‐PPT could be increased
by increasing the mole ratio and reaction time within cer-
tain ranges. However, when the mole ratio exceeded 1.8
and the reaction time was more than 30 minutes, the
yield of L‐PPT decreased.
25
strating that the model has good adaptability. Figure 4B
shows the internally studentized residuals versus the pre-
dicted values. The plots of the internally studentized
residuals are dispersed randomly, showing that the origi-
26
nal variance was constant for all values. The normal
probability plot of the residuals under the normality
assumption is shown in Figure 4C. The residual curve is
almost a straight line, confirming that the normality
assumption is appropriate. As shown in Figure 4D, the
plots of internally studentized residuals versus experi-
mental run numbers are all in the effective range. Over-
all, the ANOVA and the model adequacy diagnostics
indicated that the model has high statistical significance
and the experimental values are highly reliable and
accurate.
To analyze in more detail the effect of the variables on
the yield of L‐PPT and the interactions between the three
variables, 3D response surface plots were generated by
the Design Expert software. In these graphs, one factor
FIGURE 4 Diagnostics diagrams for the model adequacy. A, Predicted versus actual; B, residuals versus predicted; C, normal plot of
residuals; and D, residuals versus run numbers

8
XU ET AL.
model provided by the response surface analysis method
was well fitted to the actual chemical process and that
the optimization of the racemization conditions by RSM
is not only theoretically rigorous but also convenient
and effective in practice.
3.6 | Effect of adding acetic anhydride on
racemization
Previous study showed that N‐acyl amino acids were
racemized in the presence of anhydrides or ketenes. The
result showed that on the basis of optimizing racemic
reaction conditions, the addition of acetic anhydride
could reduce the reaction temperature from 154°C to
1
49°C and reaction time from 23 to 13 minutes. In addi-
tion, the racemization was complete with a yield of
7.02%. The main reason for the improvement might be
9
that the in‐situ formation of azlactones with anhydride,
which racemize readily under acidic conditions, greatly
facilitates racemization.
4
| CONCLUSION
In this study, racemization of NPDG was developed to
obtain optically pure L‐PPT, and racemization conditions
were optimized. Using a single‐factor experiment,
together with BBD‐RSM design, the optimal conditions
were obtained as follows: temperature, 154°C;
phenylacetic acid/NPDG mole ratio, 1.7; and reaction
time, 23 minutes. Accordingly, the experimental yield of
L‐PPT reached 96.13%, which was consistent with the
predicted yield of 96.32%. Moreover, addition of a small
amount of acetic anhydride further increased the yield
of L‐PPT. The results have greatly improved the yield of
racemization reaction, reduced the cost of L‐PPT produc-
tion by enzymatic resolution, and provided a theoretical
basis for the racemization of other amino acids.
FIGURE 5 3D response‐surface plots. A, Temperature and mole
ratio; B, temperature and time; and C, mole ratio and time
Using the same software, numerical optimization was
utilized to find the specific point that maximizes the
27
desired function, ie, the yield of L‐PPT, within reason-
able conditions. In this software, the “goal” field for the
response has five options: “none,” “target,” “maximum,”
“
minimum,” and “in range.” The goal for temperature,
mole ratio, and reaction time was assigned as in range,
and the corresponding ranges were chosen as 140°C to
ACKNOWLEDGMENTS
1
70°C, 1 to 2, and 10 to 30 minutes, respectively, within
This work was supported by the National Natural Science
Foundation of China (No. 21676254).
which the model was found to be well fitted. The goal
for the yield of L‐PPT was assigned as maximum. When
carrying out the optimization procedure under these
boundaries and settings, it was found that the optimal
temperature is 154°C, the optimal mole ratio is 1.7, and
the optimal reaction time is 23 minutes, at which the pre-
dicted yield of L‐PPT is 96.32%. To verify the reliability of
the response surface method, the solid experiments were
carried out, and the results found that the actual yield
of L‐PPT was 96.13%, which was in good agreement with
the model prediction of 96.32%. This showed that the
ORCID
Jian‐Miao Xu https://orcid.org/0000-0002-6230-4101
Ya‐Ping Xue https://orcid.org/0000-0003-0370-2502
REFERENCES AND NOTES
1. Hernández H, Asanza ML, Alcaraz I, Hidalgo AM, Máximo MF,
Bastida J. Production of optically pure L‐alanine in batch and

XU ET AL.
9
continuous stirred tank reactors using immobilized Pseudomo-
nas sp. Biotechnol Lett. 2001;23(11):887‐891.
17. Zhang J, Wei Y, Li H, Zeng EY, You J. Application of Box‐
Behnken design to optimize multi‐sorbent solid phase extraction
for trace neonicotinoids in water containing high level of matrix
substances. Talanta. 2017;170:392‐398.
2
3
4
5
6
7
. Xu JM, Chen B, Wang YS. Production of l‐tryptophan by
enantioselective hydrolysis of d,l‐tryptophanamide using a
newly isolated bacterium. Chem Papers. 2013;67(10):1262‐1270.
18. Hashemi SH, Kaykhaii M, Keikha AJ, Sajjadi Z. Application of
Box‐Behnken design in response surface methodology for the
molecularly imprinted polymer pipette‐tip solid phase extraction
of methyl red from seawater samples and its determination by
spectrophotometery. Mar Pollut Bull. 2018;137:306‐314.
. Leuchtenberger W, Huthmacher K, Drauz K. Biotechnological
production of amino acids and derivatives: current status and
prospects. Appl Microbiol Biotechnol. 2005;69(1):1‐8.
. Takahashi E, Furui M, Shibatani T. D‐Amino acid production
from racemic amino acids by a microbial asymmetric degrada-
tion. Biotechnol Tech. 1997;11(12):913‐916.
1
9. Li H, van den Driesche S, Bunge F, Yang B, Vellekoop MJ. Opti-
mization of on‐chip bacterial culture conditions using the Box‐
Behnken design response surface methodology for faster drug
susceptibility screening. Talanta. 2019;194:627‐633.
. Xue YP, Zheng YG, Liu ZQ, Liu X. Efficient synthesis of non‐
natural L‐2‐aryl‐amino acids by a chemoenzymatic route. ACS
Catal. 2014;4(9):3051‐3058.
2
0. Hao J, Gao Y, Zhao J, et al. Preparation and optimization of res-
veratrol nanosuspensions by antisolvent precipitation using
Box‐Behnken design. AAPS PharmSciTech. 2015;16(1):118‐128.
. Yamada S, Hongo C, Yoshioka R, Chibata I. Method for the
racemization of optically active amino acids. Jorgchem.
2
1. Maran JP, Manikandan S, Priya B, Gurumoorthi P. Box‐
Behnken design based multi‐response analysis and optimization
of supercritical carbon dioxide extraction of bioactive flavonoid
compounds from tea (Camellia sinensis L.) leaves. Journal of
Food Science and Technology. 2015;52(1):92‐104.
1
983;48(6):483‐486.
. Ariaans GJA, Houbiers JPM, Bruggink A, Zwanenburg B. Con-
trolled racemization of optically active organic compounds:
prospects for asymmetric transformation. Tetrahderon.
1
997;53(28):9417‐9476.
2
2. Vozza G, Danish M, Byrne HJ, Frias JM, Ryan SM. Application
of Box‐Behnken experimental design for the formulation and
optimisation of selenomethionine‐loaded chitosan nanoparticles
coated with zein for oral delivery. Int J Pharm. 2018;551(1–
8
. Takahashi E, Furui M, Shibatani T. Scale‐up of D‐lysine produc-
tion from L‐lysine by successive chemical racemization and
microbial
asymmetric
997;19(3):245‐250.
degradation.
Biotechnol
Lett.
1
2):257‐269.
9
. D'Arrigo P, Arosio D, Cerioli L, et al. Base catalyzed racemiza-
tion of amino acid derivatives. Tetrahedron Asymmetry.
2
2
3. Sulaiman NS, Hashim R, Mohamad Amini MH, Danish M,
Sulaiman O. Optimization of activated carbon preparation from
cassava stem using response surface methodology on surface
area and yield. J Clean Prod. 2018;198:1422‐1430.
011;22(8):851‐856.
1
0. Yoshioka R. Racemization,
optical
resolution
and
crystallization‐induced asymmetric transformation of amino
acids and pharmaceutical intermediates. In: Sakai K,
Hirayama N, Tamura R, eds. Novel Optical Resolution Technol-
ogies. Berlin, Heidelberg: Springer Berlin Heidelberg;
2
4. Tian S, Hao C, Xu G, Yang J, Sun R. Optimization conditions for
extracting polysaccharide from Angelica sinensis and its antiox-
idant activities. J Food Drug Anal. 2017;25(4):766‐775.
2
5. Kumar M, Sharma MP. Optimization of transesterification of
Chlorella protothecoides oil to biodiesel using Box‐Behnken
2
007:83‐132.
1
1
1. Tokuyama S. Discovery and application of a new enzyme N‐
acylamino acid racemase. J Mol Catal B: Enzym. 2001;12(1):3‐14.
design
method.
Waste
and
Biomass
Valorization.
2016;7(5):1105‐1114.
2. Khodadady M, Ramezani MK, Mahdavi V, Ghassempour A,
Aboul‐Enein HY. Enantioseparation and enantioselective phyto-
toxicity of glufosinate ammonium on catechin biosynthesis in
wheat. Food Anal Methods. 2014;7(4):747‐753.
2
6. Thangam R, Suresh V, Kannan S. Optimized extraction of poly-
saccharides from Cymbopogon citratus and its biological
activities. Int J Biol Macromol. 2014;65:415‐423.
2
7. Rahman N, Sameen S, Kashif M. Application of Box‐Behnken
design and desirability function in the optimization of spectro-
photometric method for the quantification of WADA banned
drug: acetazolamide. J Mol Liq. 2019;274:270‐277.
1
1
1
1
3. Herold A, Wendler C, Wild A. The effect of phosphinothricin
(
1
glufosinate) on glutathione synthesis in plants. Plant Biol.
990;103(1):68‐71.
4. Jalaludin A, Yu Q, Zoellner P, Beffa R, Powles SB. Characterisa-
tion of glufosinate resistance mechanisms in Eleusine indica.
Pest Manag Sci. 2017;73(6):1091‐1100.
5. Natchev IA. Total synthesis and enzyme‐substrate interaction of
D‐, DL‐ and L‐phosphinotrine, ‘Bialaphos’ (SF‐1293) and its
cyclic analogues. J Chem Soc Perkin Trans 1. 1989;1(1):125‐131.
How to cite this article: Xu J‐M, Li F‐L, Xue Y‐P,
Zheng Y‐G. Efficient racemization of N‐
phenylacetyl‐D‐glufosinate for L‐glufosinate
production. Chirality. 2019;1–9. https://doi.org/
6. Tak B‐Y, Tak B‐S, Kim Y‐J, Park Y‐J, Yoon Y‐h, Min G‐h. Opti-
mization of color and COD removal from livestock wastewater
by electrocoagulation process: application of Box–Behnken
design (BBD). J Ind Eng Chem. 2015;28:307‐315.
10.1002/chir.23076