http://www.paper.edu.

cn

Enhanced production of extracellular ribonuclease from Aspergillus niger

by optimization of culture conditions using response surface methodology
Ya-Hong Xiong, Jian-Zhong Liu∗ , Hai-Yan Song, Liang-Nian Ji1
Biotechnology Research Center and The Key Laboratory of Gene Engineering of Ministry of Education,
Zhongshan University, Guangzhou 510275, China
Received 31 August 2003; received in revised form 29 April 2004; accepted 29 April 2004

Abstract
Response surface methodology were used to study the cumulative effect of the culture conditions and to enhance the production of
extracellular ribonuclease in shake flask fermentation by Aspergillus niger. These conditions considered include initial pH, inoculum size,
total carbon, ratio of glucose to corn powder, NH4 NO3 and K2 SO4 . The relative importance of these conditions on ribonuclease production
was investigated by using the Plackett–Burman experimental design. Initial pH, ratio of glucose to corn powder and NH4 NO3 were found
to significantly influence the ribonuclease production. For obtaining the mutual interaction between the variables and optimizing these
variables, a 23 factorial central composite design using response surface methodology was employed. Initial pH has a significant positive
effect on the ribonuclease production. NH4 NO3 has a significant negative effect. The mutual interactions have no significant effect. The
ribonuclease production was increased by 23.2% compared to that under unoptimized conditions. The fermentation time was also shortened
after optimization.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Ribonuclease; Aspergillus niger; Optimization; Response surface methodology

1. Introduction Therefore, microbial RNases are considered as alternative

Ribonucleases (RNases) are important analytical enzymes
and widely distributed in nature. RNases from microorgan-
isms are widely used in molecular biological study, food,
and pharmaceutical industry. As an important analytical tool,
they have played a major role in study on the structure and
function of RNA [1,2]. In single-cell protein production, they
are used to remove RNA in cell [3]. RNases have also been
applied commercially to produce nucleotides for clinical
use or for seasoning nucleotides [4,5]. Many ribonucleases
are highly cytotoxic. They have several important biologi-
cal roles such as antitumor and antiviral activity [6]. Factors
that determine the cytotoxicity of microbial RNases include
catalytic activity, ability to escape natural inhibitors such
as RNase inhibitor protein (RI), stability, positive charge
on the molecule, and interaction with cell membrane [7].
Site-directed mutagenesis and chemical modification can en-
hance the resistance to RI and the positive charge of RNases.
∗ Corresponding author. Tel.: +86 20 8411 0115;
fax: +86 20 8403 5497.
E-mail address: lssljz@zsu.edu.cn (J.-Z. Liu).
1 Co-corresponding author.
chemotherapeutic drugs [7,8].
Many RNases of fungi have been studied widely and
some RNases of Aspergillus have been isolated and purified
[9–12]. But only two papers described optimization of cul-
ture conditions of ribonuclease production [10,13], where a
classical approach of the one-factor-at-a-time optimization
was employed. This method is generally time-consuming,
requires a large number of experiments to be carried out, and
does not include interactive effects among the variables [14].
Response surface methodology (RSM) is the most widely
used statistical technique for bioprocess optimization. It can
be used to evaluate the relationship between a set of con-
trollable experimental factors and observed results. The in-
teraction among the possible influencing parameters can be
evaluated with limited number of experiments [15]. It has
been successfully employed for optimization in many bio-
processes [14–16].
In our previous paper [17], we reported the effects of car-
bon, nitrogen source, and metal ions on RNase production
from Aspergillus niger mutant of A. niger ATTC 26550 us-
ing the one-factor-at-a-time methodology. In order to fur-
ther investigate the effect of conditions and improve RNase
yield, Plackett–Burman design, and a 23 central composite

1369-703X/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.bej.2004.04.010 转载
中国科技论文在线 http://www.paper.edu.cn
28 Y.-H. Xiong et al. / Biochemical Engineering Journal 21 (2004) 27–32

design using response surface methodology were employed 2.3.1. Plackett–Burman design
for optimization. Plackett–Burman experimental design [18] based on the
first order model:


2. Materials and methods Y = β0 + βi xi (1)

was used to screen the important variables that influence

2.1. Microorganism
The strain used for RNase production was A. niger
SA-13-20, which is a mutant resistant to sodium azide,
screened from A. niger ATCC 26550 (same as A. niger
NRC-A-1-23) using combined UV-nitrosoguanidine muta-
genesis method (data unpublished). The strain was main-
tained on malt agar slants containing (g l−1 ): malt, 130;
peptone, 1.0; glucose, 20; agar, 20, cultured for 6 days at
25 ◦ C, then stored at 4 ◦ C and subcultured every 2 months.
2.2. Media and cultivation
The inoculum medium was the same as above slant
medium without agar. The fermentation medium used for
optimization studies contained (g l−1 ): glucose, 30; corn
powder, 120; NH4 NO3 , 2.0; K2 SO4 , 0.087; pH 2.2.
All growth experiments were carried out in 250 ml Erlen-
meyer flasks each with 40 ml medium sterilized at 121 ◦ C
for 30 min. The flasks were placed on a rotary shaker at
180 rpm with 5 cm amplitude at 30 ◦ C for a period.
Spores suspension (1 ml, about 107 spores) was inoculated
into each flask with inoculum medium. After incubation for
24 h, the medium was used as the inoculum. The fermen-
tation medium inoculated with 10% (v/v) of 24 h inoculum
was incubated for 120 h to produce RNase.
After fermentation, the fungal biomass was separated
from the culture fluid by filtration and then the filtrate was
used to determine the enzyme activity.
2.3. Optimization procedure
RNase production. This design does not consider the in-
teraction effects among the variables. Seven variables were
screened in eight experiments with one dummy variable.
The experimental design is shown in Table 1 (Hadamard
matrix). Each variable is represented in two levels namely a
high level denoted by (+) and a low level designated by (−).
All experiments were performed in duplicate and the aver-
age of RNase activity after 5 days was taken as the response
(Y). The variables whose confidence levels were greater than
85% were considered to significantly influence RNase pro-
duction.
2.3.2. Central composite design
A 23 factorial central composite experimental design with
four start points (α = 1.68) and six replicates at the central
point, resulting in 20 experiments (generated by Design Ex-
pert, Version 2.0.5, Stat-Ease Inc., Minneapolis, MN) was
used to optimize the screened variables grouped as initial
pH (x1 ), the ratio of glucose to corn powder (x2 ) and am-
monium nitrate (x3 ). The experimental design is shown in
Table 2. The variables were coded according to the Eq. (2):
(Xi − X0 )
xi = i = 1, 2, 3, . . . , j (2)
X
where xi = coded (dimensionless) value of the variable Xi ,
X0 = the value of Xi at the center point, and X = the step
change.
The behavior of the system was explained by the folllow-
ing second order polynomial equation:




y = β0 + βi xi + βii xi 2 + βij xi xj (3)

The optimization of culture conditions for RNase produc- where y = predicted response, β0 = offset term, βi = linear
tion by A. niger SA-13-20 was carried out in two stages. effect, βii = squared effect, and βij =interaction effect.
Table 1
Plackett–Burman experimental design matrix for screening of culture conditions of RNase production by Aspergillus niger SA-13-20
Run no. Variables/levels RNase (U ml−1 )

X1 X2 X3 X4 X5 X6 X7

1 + + + + − − − 1256.9
2 − + + + + − − 1118.3
3 − − + + + + − 1200.1
4 − − − + + + + 1074.1
5 + − − − + + + 1206.5
6 + + − − − + + 1467.8
7 + + + − − − + 1143.5
8 − − − − − − − 1045.8
X1 , initial pH at a high level of 2.50 and a low level of 2.20; X2 , inoculum size at a high level of 12.5% (v/v) and a low level of 10.0 (v/v); X3 , dummy
variable; X4 , total carbon at a high concentration of 187.5 g l−1 and a low concentration of 150.0 g l−1 ; X5 , ratio of glucose to corn powder at a high
level of 1:3 and a low level of 1:4; X6 , NH4 NO3 at a high concentration of 2.50 g l−1 and a low concentration of 2.00 g l−1 ; and X7 , K2 SO4 at a high
concentration of 0.110 g l−1 and a low concentration of 0.087 g l−1 .
中国科技论文在线 http://www.paper.edu.cn

Y.-H. Xiong et al. / Biochemical Engineering Journal 21 (2004) 27–32 29

Table 2
Central composite design matrix of independent variables with corresponding experimental and predicted value
Run order no. Initial pH Ratio of glucose to corn powder NH4 NO3 (g l−1 ) RNase activity (U ml−1 )

X1 Code x1 X2 Code x2 X3 Code x3 Experimental value Predicted value

1 3.00 1 0.30 1 2.00 −1 1791.6 1665.6

2 3.00 1 0.20 −1 2.00 −1 1643.0 1494.0
3 2.50 0 0.25 0 2.50 0 1557.4 1495.2
4 2.50 0 0.25 0 2.50 0 1542.2 1495.2
5 2.00 −1 0.20 −1 2.00 −1 1353.3 1349.8
6 2.00 −1 0.20 −1 3.00 1 1081.2 1081.9
7 2.50 0 0.25 0 2.50 0 1403.7 1495.2
8 3.00 1 0.30 1 3.00 1 1459.0 1337.3
9 3.00 1 0.20 −1 3.00 1 1244.9 1125.4
10 3.34 1.68 0.25 0 2.50 0 1141.6 1388.2
11 1.66 −1.68 0.25 0 2.50 0 1164.2 1094.8
12 2.50 0 0.25 0 2.50 0 1534.8 1495.2
13 2.50 0 0.25 0 3.30 1.68 816.5 885.0
14 2.50 0 0.25 0 1.70 −1.68 1277.6 1386.3
15 2.50 0 0.33 1.68 2.50 0 1625.5 1701.7
16 2.50 0 0.17 −1.68 2.50 0 1413.8 1514.8
17 2.00 −1 0.30 1 2.00 −1 1365.8 1360.1
18 2.00 −1 0.30 1 3.00 1 1108.8 1132.6
19 2.50 0 0.25 0 2.50 0 1421.3 1495.2
20 2.50 0 0.25 0 2.50 0 1542.2 1495.2

The Design Expert was used for regression analysis of the above explanation. So, the analytic method may be reliable,
data obtained and to estimate the coefficients of the regres- especially after 2-day fermentation.
sion equation. Isoresponse contour plots were also obtained Analyses were carried out in duplicate. The data given
by using Design Expert. here are the average of the measurements.

2.4. Analytical method

RNase assay was carried out essentially according to
Horitsu et al. [10]. The total reaction mixture of 1.50 ml
contained 0.25 ml of 0.5% yeast RNA in 0.10 mol l−1 cit-
rate buffer pH 3.5, and the appropriate amount of enzyme
solution. The reaction was initiated by the addition of RNA
followed by incubation at 30 ◦ C for 20 min. The reaction
was then stopped by the addition of 0.25 ml of MacFadyen’s
reagent (0.75% (w/v) uranyl acetate in 25% (v/v) perchloric
acid). After standing for 30 min, the precipitate was removed
by centrifugation (4000 × g, 10 min). Supernatant solution
(0.10 ml) was diluted to 5.00 ml with distilled water and the
absorbance was measured at 260 nm. Controls were run in
each experiment by the addition of enzyme solution after
the addition of 0.75% MacFadyen’s reagent. One unit of
enzyme activity was defined as the increase in absorbance
of 1.00 per 20 min under the condition described above.
Ten milliliter cultures were filtered through thick filter
cloth and the mycelium was washed twice with distilled
water and then squeezed to remove small particle insol-
uble material. The biomass was dried at 105 ◦ C, cooled
in a desiccator, and weighted [19]. Corn powder in the
fermentation medium were first hydrolyzed partly during
sterilization because of low pH (about 3.0) and further cat-
alytically hydrolyzed completely by amylase excreted by
A. niger SA-13-20 after 2-day fermentation. The fact that
the fermentation broth became clear after 2 days proved the
3. Results and discussion
3.1. Screening of culture conditions for RNase production
by Aspergillus niger
According to our previous results [17], the relative impor-
tance of initial pH, inoculum size, total carbon, ratio of glu-
cose to corn powder, NH4 NO3 and K2 SO4 concentration for
RNase production was investigated using Plackett–Burman
design. The experimental design and corresponding RNase
yields are shown in Table 1. When the sign of the effect Exi
of the tested variable is positive, the influence of the vari-
able on RNase production is greater at a high level. And
when negative, the effect of the variable is greater at a low
level. The effect of variables X1 (initial pH), X2 (inoculum
size), X4 (total carbon), X5 (ratio of glucose to corn pow-
der), X6 (NH4 NO3 ), and X7 (K2 SO4 ) were 636.4, 460.0,
−214.2, −315.0, 380.0, and 270.3, respectively (Table 3).
Hence, the influence of X1 (initial pH), X2 (inoculum size),
X6 (NH4 NO3 ), and X7 (K2 SO4 ) is greater at a high level
and that of X4 (total carbon) and X5 (ratio of glucose to corn
powder) is greater at a low level. It was also found that the
variables X1 , X2 , X5 , and X6 had confidence levels above
85% and hence were considered to significantly influence
RNase production. The variables X4 and X7 , respectively,
having confidence levels below 85% were considered in-
significant.
中国科技论文在线 http://www.paper.edu.cn
30 Y.-H. Xiong et al. / Biochemical Engineering Journal 21 (2004) 27–32

Table 3
Results of the screening experiment for RNase production by Aspergillus
niger SA-13-20 Table 5
The least-squares fit and the parameter estimates (significance of regres-
Variable Exi β t-values Confidence level (%) Model Parameter Degree of Computed t P (P > |t|)
sion coefficient). The number in parentheses is the standard error
term estimate freedom
X1 636.4 318.2 8.440 **
X2 460.0 230.0 6.101 * Intercept 1495.2 (56.16) 1
X4 −214.2 −107.1 −2.841 x1 87.2 (37.26) 1 2.340 0.0412
X5 −315.0 −157.5 −4.178 * x2 55.6 (37.26) 1 1.492 0.1668
X6 380.0 190.0 5.093 * x3 −149.0 (37.26) 1 −3.999 0.0025
X7 270.3 135.1 3.585 x1 x2 40.3 (48.68) 1 0.828 0.4268
x1 x3 −25.2 (48.68) 1 −0.518 0.6160
x2 x3 10.1 (48.68) 1 0.207 0.8402
x12 −89.7 (36.27) 1 −2.473 0.0329
Significant levels of regression coefficient are given as (**) 90% and (*) x22 40.0 (36.27) 1.103 0.2964
85% by t-test. x32 −127.1 (36.27) 1 −3.504 0.0057
3.2. Optimization of the screening culture conditions for
RNase production

Because 12.5% (v/v) of inoculum size was already high
enough, the other variables (initial pH, ratio of glucose to
corn powder and NH4 NO3 ) showing confidence levels above
85% in the Plackett–Burman design were selected and op-
timized using central composite design. The experimental
and predicted response of RNase production at 5 days of
cultivation by A. niger SA-13-20 are given in Table 2. By
applying multiple regression analysis on the experimental
data, the following second order polynomial equation was
found to explain the RNase production:
y = 1495.2 + 87.2x1 + 55.6x2 − 149.0x3 − 89.7x12
+40.0x22 − 127.1x32 + 40.3x1 x2 − 25.2x1 x3 + 10.1x2 x3
(4)
where y = predicted response; x1 , x2 , and x3 are the coded
values of initial pH, ratio of glucose to corn powder and
NH4 NO3 , respectively.
Statistical testing of the model was done by the Fisher’s
statistical test for analysis of variance (ANOVA) and the re-
sults are shown in Table 4. The analysis of variance of the
quadratic regression model demonstrates that the model is
highly significant, as the computed F-value (Fmodel = 4.92)
is much greater than the tabular F10,9 value (3.02) at the
5% level. The closer the value of R (multiple correlation
coefficient) to 1, the better the correlation between the ob-
served and predicted values. Here the value of R (0.9032)
indicates a good agreement between the experimental and
predicted values. The coefficient of variation (CV) indicates
the degree of precision with which the treatments are com-
Table 4
Analysis of variance (ANOVA) for the selected quadratic modela
Source Sum of Degrees of Mean F-value P>F
squares freedom square
Model 839700 9 93305.3 4.92 0.0102
Error 189600 10 18960.5
Corrected 1029000 19
total
a Coefficient of variation (CV) = 10.02%; coefficient of determination
(R2 ) = 0.8158; correlation coefficient (R) = 0.9032, and adjusted R2
= 0.6500.
pared. Usually, the higher the value of CV, the lower is
the reliability of experiment. Here, a lower value of CV
(10.02%) indicates a greater reliability of the experiments
performed.
The Student t-distribution and the corresponding
P-values, along with the parameter estimate, are given in
Table 5. The P-values are used as a tool to check the sig-
nificance of each of the coefficients which, in turn, are nec-
essary to understand the pattern of the mutual interactions
between the best variables. The smaller the magnitude of
the P, the more significant is the corresponding coefficient.
The parameter estimate and the corresponding P-values
(Table 5) suggest that among the independent variables
initial pH (x1 ) and NH4 NO3 (x3 ) have a significant effect
on RNase production. The quadratic term of these two
variables also have a significant effect. But the independent
variable, ratio of glucose to corn powder (x2 ), has no effect.
And there is no interaction. The effect of NH4 NO3 (x3 ) is
greater than that of pH (x1 ). It was also found that RNase
production increased with the increase of initial pH and
with the decrease of NH4 NO3 concentration.
A representative contour plot which is more or less spher-
ical is shown in Fig. 1. Fig. 1 shows the relative effect of
initial pH and NH4 NO3 on RNase production. The regres-
sion equation (Eq. (4)) was solved by using Design Expert.
The optimal values of test variables in the coded units were
x1 = 0.95, x2 = 1.68, and x3 = −0.61. At these values,
the initial pH, the ratio of glucose to corn powder and the
concentration of NH4 NO3 were 2.98, 0.33, and 2.20 g l−1 ,
respectively. The maximum predicted value of RNase yield
obtained was 1815.4 U ml−1 .
A repeated fermentation of RNase under optimal condi-
tions was carried out. The result is given in Fig. 2. The
maximal RNase level obtained was 1750.0 U ml−1 . This
value was found to be 3.6% less than the predicted value.
This discrepancy might be due to the slight variation in ex-
perimental conditions. The optimization resulted in 23.2%
increase of RNase production (from 1420 U ml−1 of maxi-
mal RNase level at the sixth day under unoptimized condi-
tions to 1750 U ml−1 at the fifth day under optimized con-
ditions) and 19.3% increase of dry cell weight. The fer-
中国科技论文在线 http://www.paper.edu.cn
Y.-H. Xiong et al. / Biochemical Engineering Journal 21 (2004) 27–32 31

3.30 4. Conclusion

Statistical optimization of culture conditions using

2.90
Plackett–Burman and central composite design appears to
be a valuable tool for the production of RNase by A. niger
SA-13-20. RNase production increased with the increase of
NH4NO3 (g l-1)

1370. 0
1423. 2 initial pH and with the decrease of NH4 NO3 concentration.
2.50 The optimization enhanced the RNase production by 23.2%
and shortened the fermentation time.
156 7.2
2.10
1529. 6 Acknowledgements
1468. 1 1423 .2

1.70
We are grateful to the National Natural Science Founda-
tion of China, the Natural Science Foundation of Guangdong
1.66 2.08 2.50 2.92 3.34
Province for their financial support.
Initial pH
Fig. 1. Isoresponse contour plot of RNase production for initial pH and
NH4 NO3 . References

mentation time was shortened from 6 to 5 days. Optimiza-
tion also resulted in the increase of the growth rate and
RNase production rate. Cell growth was very fast in the
first day and then became slow to the end of fermentation.
The level of RNase was sharply increased after the first day.
From Fig. 2, it is clear that RNase production was linked to
growth.
Horitsu et al. reported that RNase production by A. niger
NRC-A-1-23 (the parent of this strain used in this study) in-
creased with initial pH range from 1.6 to 2.2 and decreased
when initial pH further increased to 3.0 [10]. They also con-
sidered NH4 NO3 was the best nitrogen source. The opti-
mal initial pH of RNase production by Rhizopus stolonifer
was 6.0 [13]. Inorganic nitrogen salts used as sole nitrogen
source did not support the growth of R. stolonifer. The opti-
mal nitrogen source of R. stolonifer RNase production was
the peptone obtained from Sarabhai M. Chemicals Ltd.
30
25
[1] K. Takahashi, S. Moore, Ribonuclease T1, in: P.D. Boyer (Ed.), The
enzymes: Nucleic Acids, Part B, vol. 15, Acadmic Press, New York,
1982, pp. 435–468.
[2] K. Zhao, L. Hua, Application of RNase TCS in RNA sequence
analysis, Prog. Biochem. Biophys. 23 (1996) 257–260 (in Chinese).
[3] G.L. Reddy, V. Shankar, Immobilized nucleases, Crit. Rev.
Biotechnol. 13 (1993) 255–273.
[4] M. Lindblom, H. Morgen, Enzymatic RNA reduction in disintegrated
cells of Saccharomyces cerevisiae, Biotechnol. Bioeng. 16 (1974)
1123–1133.
[5] H. Liu, H.C. Pan, H. Ding, X.L. Wang, Studies on the immobilization
of nuclease P (1), Food Ferment. Ind. 4 (1994) 44–48 (in Chinese).
[6] S.K. Saxena, M. Gravell, Y.-N. Wu, S.M. Mikulski, K. Shogen,
W. Ardelt, R.J. Youle, Inhibition of HIV-1 production and selective
degradation of viral RNA by an amphibian ribonuclease, J. Biol.
Chem. 271 (1996) 20783–20788.
[7] A.A. Makarov, O.N. Ilinkaya, Cytotoxic ribonucleases: molecular
weapons and their targets, FEBS Lett. 540 (2003) 15–20.
[8] Y. Zhang, W. Li, Z. Su, Trends in the pharmaceutical research of
ribonucleases and their therapeutic uses, J. Biomed. Eng. 18 (2001)
456–460.
[9] H. Horitsu, N. Hara, M. Tohgane, K. Kawai, Purification of a
base-specific ribonuclease Lu from Aspergillus niger, Agric. Biol.
2000 Chem. 46 (1982) 1145–1151.
[10] H. Horitsu, Y. Higashi, M. Tomoyeda, Production, purification and
properties of ribonuclease from Aspergillus niger, Agric. Biol. Chem.
RNase activity (U ml )

1600
38 (1974) 933–940.
Mycelium (g l )

20 [11] M. Irie, M. Harada, T. Negi, Some chemical and physical properties

-1

1200
of ribonuclese from Aspergillus niger, J. Biochem. 69 (1971) 881–
15 892.
800 [12] E. Gomes, R. de Silva, A. Serzedello, Ribonuclease production by
10
Aspergillus species, Rev. Microbiol. 29 (1998) 187–192.
5
400 [13] R. Chacko, M. Deshpande, V. Shankar, Extracellular ribonucleae
production by Rhizopus stolonifer: influence of metal ions, Curr.
-1

0 0 Microbiol. 32 (1996) 246–251.

0 2 4 6
Time (d)
Fig. 2. Growth and RNase production by Aspergillus niger SA-13-20
under optimized conditions compared to that under unoptimized condi-
tions. Under optimized conditions: (䊏) mycelium and (䊉) RNase; under
unoptimized conditions: (䊐) mycelium and (䊊) RNase.
8 [14] G. Dey, A. Mitra, R. Banerjee, B.R. Maiti, Enhanced production of
amylase by optimization of nutritional constituents using response
surface methodology, Biochem. Eng. J. 7 (2001) 227–231.
[15] F. Francis, A. Sabu, K.M. Nampoothiri, S. Ramachandran, S. Ghosh,
G. Szakacs, A. Pandey, Use of response surface methodology for
optimizing process parameters for the production of ␣-amylase by
Aspergillus oryzae, Biochem. Eng. J. 15 (2003) 107–115.
中国科技论文在线 http://www.paper.edu.cn
32 Y.-H. Xiong et al. / Biochemical Engineering Journal 21 (2004) 27–32

[16] J.-Z. Liu, L.-P. Weng, Q.-L. Zhang, H. Xu, L.-N. Ji, Optimization
of glucose oxidase production by Aspergillus niger in a benchtop
bioreactor using response surface methodology, World J. Microbiol.
Biotechnol. 19 (2003) 317–323.
[17] Y.-H. Xiong, J.-Z. Liu, T.-L. Wang, L.-N. Ji, Preliminary optimization
of culture condition for the production of extracellular ribonuclease
from Aspergillus niger using one-factor-at-a-time methodology,
Acta Sci. Nat. Univ. Sunyatseni 42 (Suppl. 2) (2003) 191–
194.
[18] R.L. Plackett, J.P. Burman, The design of optimum multifactorial
experiments, Biometrika 33 (1946) 305–325.
[19] J.-Z. Liu, Y.-Y. Huang, J. Liu, L.-P. Weng, L.-N. Ji, Effects of metal
ions on simultaneous production of glucose oxidase and catalase by
Aspergillus niger, Lett. Appl. Microbiol. 32 (2001) 16–19.