Enhanced production of extracellular proteolytic enzyme excreted by a newly isolated Bacillus subtilis FBL-1 through combined utilization of statistical designs and response surface methodology

Abstract

An extracellular protease producing strain FBL-1 was newly isolated from soil, and it was identified as Bacillus subtilis through 16S rDNA sequence analysis. B. subtilis FBL-1 was used for extracellular protease production, and culture conditions were optimized by statistical methods. Three statistical approaches such as Plackett–Burman design, steepest ascent path analysis, and Box–Behnken design were successfully combined to optimize protease production, which resulted in significant enhancement of production. Through Plackett–Burman experimental design, fructose and yeast extract were screened as the significant components for the production of extracellular protease. The center points of each parameter were predetermined by the steepest ascent method. Based on the results obtained by Box–Behnken experimental design, the optimized level of significant parameters for protease production was determined by multiple regression analysis. The optimized levels of parameters were fructose 32.42 g L⁻¹, yeast extract 15.02 g L⁻¹, and incubation time 35.78 h. Protease activity predicted by the quadratic model developed in this work was 578.55 U mL⁻¹, which was only 2.67% different from the experimental protease activity obtained by verification studies. The protease activity was significantly enhanced by 7.0-fold compared with the activity obtained from basal medium.

---

1. Introduction

Several extracellular enzymes are commercially available and widely used in industry. Among them, proteases are the class of hydrolytic enzymes catalyzing cleavage of peptide linkages in protein molecules, and they form a large and complex group of enzymes in nature.¹ Proteases are an industrially significant class of enzymes, since they account for approximately 60–65% of total worldwide sales of the enzymes.^2,3 They have been one of the most versatile enzymes with a long history of catalytic applications in a large variety of industries. For example, proteases have been used in the pharmaceutical and food industries for peptide synthesis, in the leather industry for de-hairing, and in the detergent industry as an additive in detergent formulation.⁴ Though the use of proteases in laundry detergents accounts for approximately 25% of total worldwide sales of enzymes, the commercial importance of proteases is due to their applicability for use in the field of food industry, where they are required in large quantity to prepare the functional foods such as bean paste, bean curd, and soymilk.⁵ Proteases have been also used for natural cheese making by accelerating many biochemical changes which occur during natural cheese ripening. Functional compounds can be drawn out or can be synthesized from different industrial waste with the help of proteases. In addition, proteases have been used to a lesser extent for other applications such as contact lens cleaning, pest control, the isolation of nucleic acid, degumming of silk, and selective delignification of hemp stemwood.^6,7 Recently, the roles of proteases in synthesis of bioactive peptides and as additive in commercial detergents gain attention.²

Proteases are commonly found in all living systems such as plant, animals, and microorganisms. Production of protease from plant resources depends on many time consuming factors such as cultivation of plants and suitable climatic conditions for their growth. The production of protease from animal resources is more convenient than protease production from plant resources but still not suitable for industrial production because it is dependent on the availability of livestock for slaughter. Microorganisms become attractive source of proteases because of their biochemical diversity and the ease of genetic manipulation.⁸ Several microorganisms including bacteria, yeasts, and fungi are capable of producing microbial proteases, but only those microorganisms that produce enough amounts of extracellular proteases are of industrial importance. Microbial proteases are widely different not only in their functions but also in their properties. The proteases are classified into acid, neutral, and alkaline proteases depending on the optimum pH for their activity, and they are also divided into four different groups such as serine proteases, thiol proteases, aspartic proteases, and metallo proteases according to the presence of catalytic residue on the active sites.⁹ The production of extracellular proteases in microorganisms is significantly affected by medium components and fermentation conditions.^10,11 Therefore, a process optimization as well as isolation of protease overproducing microorganisms should be significantly important to develop the industrial protease production process.

Bioprocess optimization has become an important topic in industrial production processes, and statistical experimental methods could offer ideal ways for process optimization studies in biotechnology.^1,8 In addition, experimental designs based on statistics can eliminate the limitations of single factor optimization and can provide information about the optimum levels of various medium components and culture conditions to improve the maximum production of valued products.¹² Plackett–Burman experimental design has been frequently used for selection of critical parameters from a large number of process parameters, and it is thus useful tool in preliminary studies where the major objective is to evaluate each parameter that can be fixed or eliminated in further optimization processes.^13,14 The steepest ascent path analysis is an effective experimental technique for moving sequentially along the direction of the maximum increase in the response, and thus, can approach the optimum neighborhood rapidly and efficiently.¹⁵ Response surface methodology is a collection of mathematical and statistical techniques that can be used for modelling and analysis in applications where a response of interest is affected by several parameters and the objective is to find the optimum conditions for the parameters to accomplish the desired responses.¹⁶ Recently, the response surface methodological approaches have been used for optimization studies in several biotechnological and industrial processes.^17–21 These kinds of statistical approaches are versatile techniques for evaluation of multiple process variables since they make the process easily optimized with smaller amounts of experimental trials and enable interactions between variables to be readily identified.^1,22 Recently, most optimization studies for microbial protease production were carried out by employment of two statistical methods, Plackett–Burman design and response surface methodology such as Box–Behnken design or central composite rotatable design. However, it is necessary to combine the steepest ascent path analysis with two conventional designs, since it should be expected to give more accurate information of optimum neighborhood for the next statistical design such as response surface methodology.

In the present study, a newly isolated strain FBL-1 as a potential candidate of extracellular protease producer was reported, and systematic and sequential optimization strategies were applied to enhance the production of microbial extracellular protease from strain FBL-1. For this objective, combined utilization of three statistical approaches such as Plackett–Burman design, steepest ascent path analysis, and Box–Behnken design was proposed. Screening of significant parameters affecting extracellular protease production was carried out by 2-level factorial designs using the Plackett–Burman experimental design. Then, further optimization of the screened parameters was performed by the steepest ascent path analysis and response surface methodology for improved production of extracellular protease from strain FBL-1. To the best of our knowledge, this study might be the first to report that the steepest ascent path analysis was combined with Plackett–Burman design and Box–Behnken design for optimization of microbial protease production.

2. Experimental details

Microorganism

One gram of soil sample, collected from Yeungnam University campus, South Korea, was suspended in 9 mL of sterile saline solution. Then 100 μL of this was serially diluted and amplified on sterilized agar plates with the following composition (g L⁻¹): tryptic soy broth (TSB; BD, Sparks, MD, USA), 3.0; skim milk (BD), 2.0; agar (Yakuri Pure Chemicals Co., Kyoto, Japan), 2.0. Colonies showing the largest clear zone in their surroundings were selected and streaked onto fresh TSB agar plates. The isolated strain named FBL-1 was grown on TSB agar medium, which was preserved at −70 °C in medium containing 50% (v/v) glycerol until further use.

The identity of strain FBL-1 was confirmed by using 16S rDNA sequence analysis. Chromosomal DNA of strain FBL-1 was separated by using Wizard genomic DNA purification kit (Promega Co., Medison, WI, USA). The DNA extracts were used for polymerase chain reaction (PCR) with the universal primers such as 27f (5′-AGAGTTTGATCCTGGCTCAG-3′) and 1492r (5′-GGCTACCTTGTTACGACTT-3′). Amplified 16S rDNA sequences were purified with a Wizard SV Gel and PCR clean-up system (Promega Co.), which were sequenced with a ABI PRISM 377 DNA analyzer (Applied Biosystems, Foster City, CA, USA). The 16S rDNA sequence of strain FBL-1 was compared with other sequences on the EzTaxon-e server (http://eztaxon-e.ezbiocloud.net/).²³ Multiple sequence alignment program, CLUSTAL-X, was used to align the 16S rDNA sequences of strain FBL-1 and its close relatives.²⁴ The neighbor-joining method was used, based on consensus phylogenetic tree constructed using MEGA software (Ver. 5).²⁵ The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test was obtained from 1000 bootstrap replicates.

Fermentation

The cells in the stock cultures were transferred to 100 mL of a growth medium and dispensed into 250 mL Erlenmeyer flasks, followed by incubation for 15 h at 37 °C. The growth medium used was TSB (BD) medium. This was then inoculated aseptically at 2% (v/v) into 250 mL Erlenmeyer flask containing 100 mL of the production medium, which were incubated on a shaking incubator (Vision Scientific Co., Daejeon, Korea). The production medium was composed of the following composition (g L⁻¹): fructose, 0–40; yeast extract, 0–20; K₂HPO₄, 0.1–1.0; KH₂PO₄, 0.1–1.0; MgSO₄, 0.01–0.1; MnCl₂, 0.01–0.1. The inoculated flasks were incubated at 37 °C under shaking condition at 200 rpm for 6–48 h.

Plackett–Burman experimental design

Plackett–Burman experimental design was used for selection of significant parameters for extracellular protease production.¹³ Six culture parameters (fructose, yeast extract, K₂HPO₄, KH₂PO₄, MgSO₄, and MnCl₂) were evaluated at three different levels (−1, 0, and +1), based on Plackett–Burman experimental design matrix. The independent parameters were screened from 16 combinations of experiments including 4 replications of center point (Table 1). The major objective of Plackett–Burman designed experiment is to identify significant parameters and their corresponding coefficients. Factors with the highest t-value, p-value, and confidence level over 95% were considered showing significant effect on protease production. The Plackett–Burman experimental design is based on the first-order model:

y = β₀ + ∑β_ix_i

where y is the predicted response, β₀ and β_i are the constant coefficients, and x_i is the coded independent parameters.

Table 1 Plackett–Burman experimental design and observed response for selection of significant parameters affecting extracellular protease production by B. subtilis FBL-1^a

Run	Actual levels (g L^−1)						Protease activity (U mL^−1)	Dry cell weight (g L^−1)
	x1	x2	x3	x4	x5	x6
a x1, fructose; x2, yeast extract; x3, K2HPO4; x4, KH2PO4; x5, MgSO4; x6, MnCl2.
1	10.000	0.000	1.000	0.100	0.010	0.010	0.0 ± 0.0	0.000 ± 0.000
2	10.000	5.000	0.100	1.000	0.010	0.010	258.7 ± 3.5	1.462 ± 0.020
3	0.000	5.000	1.000	0.100	0.100	0.010	5.3 ± 0.1	2.131 ± 0.020
4	10.000	0.000	1.000	1.000	0.010	0.100	0.0 ± 0.0	0.762 ± 0.002
5	10.000	5.000	0.100	1.000	0.100	0.010	283.1 ± 2.1	1.268 ± 0.023
6	10.000	5.000	1.000	0.100	0.100	0.100	331.8 ± 5.3	1.058 ± 0.012
7	0.000	5.000	1.000	1.000	0.010	0.100	2.1 ± 0.2	1.688 ± 0.010
8	0.000	0.000	1.000	1.000	0.100	0.010	0.0 ± 0.0	0.132 ± 0.002
9	0.000	0.000	0.100	1.000	0.100	0.100	4.2 ± 0.1	0.054 ± 0.001
10	10.000	0.000	0.100	0.100	0.100	0.100	0.0 ± 0.0	1.283 ± 0.022
11	0.000	5.000	0.100	0.100	0.010	0.100	1.1 ± 0.0	1.805 ± 0.018
12	0.000	0.000	0.100	0.100	0.010	0.010	0.0 ± 0.0	0.218 ± 0.001
13	5.000	2.500	0.550	0.550	0.055	0.055	254.4 ± 3.3	0.700 ± 0.005
14	5.000	2.500	0.550	0.550	0.055	0.055	265.0 ± 2.1	0.747 ± 0.003
15	5.000	2.500	0.550	0.550	0.055	0.055	272.5 ± 1.6	0.785 ± 0.001
16	5.000	2.500	0.550	0.550	0.055	0.055	264.0 ± 4.6	0.832 ± 0.001

---

The steepest ascent path analysis

Plackett–Burman design gives a local approximation in a narrow region close to the initial operating conditions, but far from where the process evidenced curvature.²⁶ Thus, the significant parameters selected by Plackett–Burman designed experiments were roughly optimized in terms of extracellular protease activity through the steepest ascent path analysis to determine center points of each parameter for further optimization step, Box–Behnken experimental design for response surface methodology. The path of the steepest ascent started from the zero level of selected parameter in Plackett–Burman design, and a series of exploratory runs were carried out. The point where the path of the steepest ascent reached its plateau would be near the optimum point and could be used as center points for further optimization study.²⁷

Box–Behnken experimental design

Box–Behnken design for three independent parameters, each at three levels, was used to develop a second-order polynomial model which determined the optimum levels of each parameter for extracellular protease production. As selected by the Plackett–Burman experimental design, fructose and yeast extract were utilized as independent parameters for the Box–Behnken experimental design. In addition, incubation time was added to the design matrix to determine the appropriate culture period of strain FBL-1 for maximum production of extracellular protease. The parameters of experiments were coded on the basis of the following equation to calculate statistical results:
where x_i is the dimensionless coded value of the independent parameter, X_i is the actual value of the independent parameter, X₀ is the actual value of the independent parameter at the center point, and ΔX_i is the step change value.

The behavior of each parameter, their interactions, and statistical analysis to obtain predicted response (protease activity) was expressed by the following second-order polynomial equation:

y = b₀ + ∑b_ix_i + ∑b_ijx_ix_j + ∑b_iix_i²

where y is the predicted response, b₀ is the offset term, b_i is the linear effect, b_ii is the squared effect, b_ij is the interaction effect, and x_i is the ith independent parameter. The terms x_ix_j and x_i² represented the interaction and quadratic terms, respectively. A total of 18 experimental runs with various combinations of fructose, yeast extract, and incubation time were conducted (Table 2). The response surface and contour plots were constructed in order to evaluate the interaction effects of the parameters and to determine the optimal values of each parameter.

Table 2 Box–Behnken experimental design and responses for optimization of extracellular protease production by B. subtilis FBL-1^a

Run	Actual levels			Protease activity (U mL^−1)
	x1 (g L^−1)	x2 (g L^−1)	x7 (h)	Experimental	Predicted
a x1, fructose; x2, yeast extract; x7, incubation time.
1	1	0.5	24	18.0 ± 0.5	−30.625
2	39	0.5	24	1.1 ± 0.1	62.175
3	1	19.5	24	212.0 ± 5.1	150.925
4	39	19.5	24	323.4 ± 7.2	372.025
5	1	10	6	50.9 ± 0.6	125.5
6	39	10	6	32.9 ± 0.4	−2.2
7	1	10	42	10.6 ± 0.3	45.7
8	39	10	42	561.9 ± 3.3	487.3
9	20	0.5	6	1.1 ± 0.0	−24.875
10	20	19.5	6	67.9 ± 0.3	54.375
11	20	0.5	42	0.0 ± 0.0	13.525
12	20	19.5	42	399.7 ± 2.2	425.675
13	20	10	24	481.3 ± 5.6	489.083
14	20	10	24	500.4 ± 4.5	489.083
15	20	10	24	505.7 ± 4.0	489.083
16	20	10	24	483.4 ± 6.8	489.083
17	20	10	24	482.4 ± 7.1	489.083
18	20	10	24	481.3 ± 6.2	489.083

---

Verification of the developed model

The optimum values developed by response surface methodological model were evaluated for its validity, and an optimized set of experimental combinations was utilized to confirm protease production by strain FBL-1. The shake-flask fermentation for model validation was conducted at 37 °C and 200 rpm for 48 h in 250 mL Erlenmeyer flasks containing 100 mL aliquots of the fermentation medium. Samples were withdrawn aseptically at appropriate intervals, which were subjected to further analysis in terms of protease activity and cell growth.

Analytical methods

Cell growth was monitored as dry weight by optical density measurements at 660 nm with a spectrophotometer (Shimadzu Co., Kyoto, Japan). The cells were then centrifuged for 10 min at 13 000 × g using a bench-top centrifuge (Vision Scientific Co.), washed twice with deionized water, and dried in a drying oven (Vision Scientific Co.) until a constant weight was achieved (24 h). Then, the optical density was converted to dry cell weight (g L⁻¹) using an appropriate calibration curve.

Protease activity was measured by Folin & Ciocalteu's method²⁸ with some modifications. A 20 μL of crude enzyme solution was mixed with 500 μL of 1% (w/v) casein solution dissolved in 50 mM glycine–NaOH buffer (pH.9.0), which was then incubated at 40 °C for 10 min. The enzyme reaction was stopped by addition of 500 μL of 10% (w/v) trichloroacetic acid to the reaction mixture, which was then centrifuged at 16 000 × g for 10 min to remove the insoluble precipitates. After the supernatant (500 μL) was mixed with 2.5 mL of 500 mM Na₂CO₃, 500 μL of 20% (v/v) 2 N Folin-Ciocalteu Phenol Reagent (Sigma-Aldrich Co., St. Louis, Mo, USA) was added. The mixture was allowed to incubate at 40 °C for 10 min. Absorbance of the mixture was measured with a spectrophotometer (Shimadzu Co.) at 660 nm. One unit of protease activity was defined as the amount of enzyme required to liberate 1 μg tyrosine per mL per min under the standard assay conditions used in this study.

Statistical analysis

All experiments were carried out in triplicate, and the average value including standard error was taken as the response. The Plackett–Burman experimental design was done with Minitab software package (Ver. 16.1.0; Minitab Inc., State College, PA, USA), and the main effects of each parameter were evaluated. The Box–Behnken experimental design was done with Design-Expert statistical software package (Ver. 7.1; Stat-Ease, Inc., Minneapolis, MN, USA). Statistical analysis of the models was used to evaluate the analysis of variance (ANOVA). The quality of the polynomial model equations was determined statistically by the coefficient of determination (R²), and its significance was determined by the F-test. The significance of the regression coefficients was tested by Student's t-test.

3. Results and discussion

Isolation and identification of strain FBL-1 by 16S rDNA sequence analysis

Many studies have been extensively carried out on alkaline and acidic protease from Bacillus species, but there are few reports on microbial proteases that are stable under neutral condition. The protease activity stable under neutral condition is quite important to the food industry since it possesses a specific functions in hydrolyzing hydrophobic amino acid bonds at a neutral pH, thereby reducing the bitterness of food protein hydrolyzates. Therefore, in this study, we tried to isolate a new microorganism for production of protease which is stable under broad pH ranges near neutral conditions. Strain FBL-1 could produce relatively high amount of extracellular protease stable at broad pH range between pH 6–12. Protease from strain FBL-1 was most stable under neutral condition at pH 7, and the enzyme activity was sustained to >60% at pH 6–12 (data not shown).

The 16S rDNA sequence (1482 bp) of a newly isolated strain FBL-1 was analyzed and compared with other sequences of related taxa. The results showed clear similarities with the 16S rDNA sequences of B. subtilis subsp. inaquosorum BGSC 3A28^T (99.8%) and B. subtilis subsp. spizizenii NRRL B-23049^T (99.8%). Fig. 1 shows the phylogenetic tree constructed using 16S rDNA sequences of strain FBL-1 and close relatives. Strain FBL-1 was identified as B. subtilis based upon the results of 16S rDNA sequence analysis, and the 16S rDNA sequence was deposited to GenBank database as an accession number KR336550. For further experiments, strain FBL-1 was named as B. subtilis FBL-1.

Fig. 1 Phylogenetic tree showing the position of strain FBL-1 and other related taxa. Numbers at nodes are percentage bootstrap values based on 1000 replications. GeneBank accession numbers of the sequences are indicated in parentheses. Bar represents 5 nt substitution per 1000 nt.

Screening of significant parameters affecting extracellular protease activity through Plackett–Burman experimental design

Plackett–Burman designed experiments of 16 total runs with high and low levels of each parameter was done to evaluate significance of the parameters affecting extracellular protease production by B. subtilis FBL-1. As shown in Table 1, six parameters were contained in the design matrix, and both extracellular protease activity and dry cell weight were resulted as the measured responses. A wide variation of protease production (0–331.8 U mL⁻¹) and dry cell weight (0–2.131 g L⁻¹) was obtained. This results suggest that optimization of medium composition should be important to get a high yield of product and cells. Fig. 2 shows the estimated effects of each parameter on protease production and cell growth. Among the parameters investigated, fructose, yeast extract, K₂HPO₄, and KH₂PO₄ had positive effect on protease production, and MgSO₄ and MnCl₂ had negative effect (Fig. 2A). However, yeast extract and MnCl₂ had positive effect on the growth of B. subtilis FBL-1, and fructose, K₂HPO₄, KH₂PO₄, and MgSO₄ had negative effect (Fig. 2B). Based on the analysis of regression coefficients and p-value of six parameters, yeast extract was found to be a significant parameter for protease production and cell growth. The t-test for individual effect helps to evaluate the probability of finding observed effect purely by chance. Several researchers reported that confidence levels >70% should be acceptable.²⁹ In this study, the parameters with confidence levels >95% (p < 0.05) were considered to influence the responses significantly. The parameters showing significant positive effects on protease production were fructose (p = 0.010) and yeast extract (p = 0.009). The most significant parameter for cell growth was yeast extract (p = 0.003). Metal ions and phosphate sources had confidence levels <95% (p > 0.05) and were considered not significant for protease production and cell growth. The coefficient of determination (R²) of the polynomial model for protease production and cell growth were 0.8531 and 0.7176, respectively. Polynomial equation of the model for protease production and cell growth are as follows:

y_protease (U mL⁻¹) = −87.0140 + 14.3479x₁ + 29.2612x₂ − 38.4809x₃ + 38.8735x₄ + 671.454x₅ − 384.809x₆

y_cell (g L⁻¹) = 0.9884 − 0.0325x₁ + 1.1605x₂ − 0.0532x₃ − 0.1882x₄ − 0.0015x₅ + 0.2398x₆

where y_protease is the predicted extracellular protease production (U mL⁻¹), y_cell is the predicted dry cell weight (g L⁻¹), and x₁, x₂, x₃, x₄, x₅, and x₆ are fructose, yeast extract, K₂HPO₄, KH₂PO₄, MgSO₄, and MnCl₂, respectively.

Fig. 2 Estimated effects of experimental parameters on extracellular protease production (A) and cell growth (B) of B. subtilis FBL-1 based on the results obtained by Plackett–Burman experimental design. x₁, fructose; x₂, yeast extract; x₃, K₂HPO₄; x₄, KH₂PO₄; x₅, MgSO₄; x₆, MnCl₂.

The results suggested that fructose and yeast extract should be significant parameters showing positive effect on extracellular protease production. However, yeast extract alone showed significant positive effect on cell growth of B. subtilis FBL-1. There have been no generalized media for protease production by different microbial strains.³⁰ Each microorganism evidenced its own characteristic physicochemical and nutritional requirements for cell growth and enzyme secretion. Several research groups have attempted to produce protease from glucose or starch, coupled with yeast extract, peptone, or casamino acid.^31,32 Though a further study has to be performed, nutrient requirement of B. subtilis FBL-1 seemed to be very simple since it did not require salts and metals for protease production. This simple nutrient requirement of microorganism is important to industrial production of extracellular protease. The optimum levels of two significant parameters, fructose and yeast extract, for extracellular protease production by B. subtilis FBL-1 were further evaluated by response surface methodology.

Steepest ascent path analysis to determine center points for Box–Behnken experimental design

The Plackett–Burman design was valuable tool for selection of significant parameters affecting extracellular protease production, but it was unable to predict exact optimum levels of the parameters. On the basis of the 1st-order model, the steepest ascent path was analyzed to find the proper direction of changing parameters, increasing the levels of each parameter to improve protease production. In addition, incubation time was added to the design matrix since protease yield might be normally dependent on fermentation period during batch cultivation of microorganisms. The results obtained by Plackett–Burman design indicated that fructose and yeast extract had significant positive effect, which suggested that increasing levels of these parameters should result in higher level of protease activity. The experimental design and results of the steepest ascent path analysis are given in Table 3. It was shown that the highest protease activity was 509.6 ± 6.9 U mL⁻¹ at run 2, 0 + 1Δ. Therefore, this point was considered close to the region of maximum protease activity. Consequently, these culture parameters (fructose 20 g L⁻¹, yeast extract 10 g L⁻¹, and incubation time 24 h) were chosen as the proper center point for further optimization study.

Table 3 Steepest ascent path analysis and results for determination of center point values of each parameter for Box–Behnken experimental design^a

Run	Step change value	Actual levels			Protease activity (U mL^−1)
		x1 (g L^−1)	x2 (g L^−1)	x7 (h)
a x1, fructose; x2, yeast extract; x7, incubation time.
1	0	10	5	12	259.4 ± 9.6
2	0 + 1Δ	20	10	24	509.6 ± 6.9
3	0 + 2Δ	30	15	36	455.2 ± 11.8
4	0 + 3Δ	40	20	48	446.7 ± 13.2

---

Box–Behnken experimental design for response surface methodological optimization of microbial extracellular protease production

The Box–Behnken designed experiments were carried out in the vicinity of the optimum to locate true optimum levels of fructose (x₁), yeast extract (x₂), and incubation time (x₇). The levels of the parameters for the Box–Behnken design were selected according to the results of the previous experiments. The design matrix and the experimental data are given in Table 2. A total of 18 runs were conducted to optimize three independent parameters. The experimental results of the Box–Behnken design were fit with the following 2nd-order polynomial equation:

y_protease (U mL⁻¹) = −203.63 + 9.18x₁ + 41.73x₂ − 18.17x₇ + 0.18x₁x₂ + 0.42x₁x₇ + 0.49x₂x₇ − 0.42x₁² − 2.20x₂² − 0.53x₇²

where y_protease is the predicted extracellular protease production (U mL⁻¹) and x₁, x₂, and x₃ are codded parameters of fructose, yeast extract, and incubation time, respectively.

The predicted response of the Box–Behnken design calculated by the above polynomial equation is given in Table 2. The fitness of the model was evaluated by the coefficient of determination (R²), which was 0.9683, indicating that 96.83% of the variability in the response could be explained by the quadratic model (Table 4). The statistical significance of the model equation was evaluated by an ANOVA, which revealed that the regression is statistically significant (p < 0.0001) at 99% confidence level. The adjusted R² value was 0.9326, and the coefficient of variation of the model was 23.13%. These results imply that the model equation should afford a better precision and reliability for the experiments.

Table 4 Analysis of variance (ANOVA) for the response quadratic model developed by Box–Behnken experimental design^a

	DF	SS	MS	F-Value	p-Value
a DF, degree of freedom; SS, sum of squares; MS, mean square; R^2 = 0.9683; adjusted R^2 = 0.9326; adequately precision = 11.762; CV = 23.13%; statistically significant at p-value <0.05.
Model	9	858 829.37	95 425.49	27.16	<0.0001
Residual	8	28 112.42	3514.05
Lack of fit	3	27 510.71	9170.24	76.20	0.0001
Pure error	5	601.71	120.34
Total	17	886 941.80

---

The relation of experimental results and predicted response is given in Fig. 3, where all the predicted responses were in close agreement with experimental results (R² = 0.9683). This implies that the quadratic model developed in this study has to be enough to represent the response of the experimental results associated with extracellular protease production. According to the statistical analysis, two independent parameters (x₂ and x₇), two interaction terms (x₁x₇ and x₂x₇), and three quadratic terms (x₁², x₂², and x₇²) were significant for extracellular protease production by batch fermentation of B. subtilis FBL-1. Though the effect of fructose (x₁) was not significant, its quadratic term (x₁²) was significant for the production of extracellular protease from B. subtilis FBL-1.

Fig. 3 Relation between experimental and predicted values calculated by response surface methodological model based on Box–Behnken experimental design.

Response surface and contour plots were visualized to evaluate the interaction effects of the parameters on extracellular protease production and to analyze their optimum levels for maximization of protease production. As shown in Fig. 4, all response surface plots shape in convexity, which indicates that there were well-defined optimum levels. The response surface plots showing the combined effects of fructose–yeast extract, fructose–incubation time, and yeast extract–incubation time had almost circular contour plots. The optimized levels for the parameters obtained from the maximum point of the model were 32.42 g L⁻¹ for fructose, 15.02 g L⁻¹ for yeast extract, and 35.78 h for incubation time. The model predicted a maximum response of 578.55 U mL⁻¹ protease activity for this point.

Fig. 4 Response surface and contour plots of extracellular protease production by B. subtilis FBL-1 showing mutual interactions between fructose and yeast extract (A), fructose and incubation time (B), and yeast extract and incubation time (C). Another parameter in each figure was adjusted to zero level in coded units.

Verification of the model

Experiments were carried out in triplicate using the optimized conditions representing the maximum point of protease activity to verify the modelling results. Fig. 5 shows time courses of extracellular protease production and cell growth in the shake flask fermentation using the optimized medium that is composed of fructose 32.42 g L⁻¹ and yeast extract 15.02 g L⁻¹.

Fig. 5 Time courses of extracellular protease activity and dry cell weight during fermentation under the statistically optimized conditions. The culture medium used in this experiment was composed of fructose 32.42 g L⁻¹ and yeast extract 15.02 g L⁻¹. The error bars represent standard deviation within triplicate samples.

The average protease activity found by experiments was 594.4 ± 6.0 U mL⁻¹ at 35.78 h of incubation time. The maximum extracellular protease activity obtained by experiments gives only 2.7% difference from the predicted response. This result suggests that the model should be in good agreement between predicted and experimental values.

Table 5 shows the enhancement degree of extracellular protease production by B. subtilis FBL-1 during each optimization step. Protease activity was only 84.82 U mL⁻¹ when the basal medium (TSB) was used as the major medium, but it could be improved to 331.8 U mL⁻¹ by Plackett–Burman experimental design. The steepest ascent method resulted in the increase of protease activity to 509.6 U mL⁻¹, and further the protease activity was enhanced to 594.4 U mL⁻¹ by response surface methodology using Box–Behnken experimental design. The final protease activity obtained from this study corresponds to 7.0-fold increased amount compared to the fermentation using TSB as a medium. Beg et al.⁸ reported that 4.2-fold increased protease activity was obtained by batch culture of B. mojavensis through response surface methodology. According to Chauhan and Gupta,³² the protease activity from Bacillus sp. RGR-14 was enhanced to 12.85-fold increased value by statistical optimization method. Oskouie et al.³³ also reported 6-fold enhancement of protease production from B. clausii cultivation by statistical optimization method. Therefore, it can be deduced that the application of statistical designs for selection and optimization of culture conditions should allow rapid identification and optimization of the significant parameters and interactions between them. Maximum protease activity obtained in this study (594.4 U mL⁻¹) was relatively high compared with previous reports by Joo et al. (207.8 U mL⁻¹ from B. horikoshii),³⁴ Tari et al. (306.5 U mL⁻¹ from Bacillus sp. L21),³⁵ Patel et al. (410 U mL⁻¹ from Bacillus sp. Ve-1),³⁶ and Suganthi et al. (141.5 U mL⁻¹).³⁷ In addition, as shown in Table 6, volumetric productivity of protease activity from B. subtilis FBL-1 was relatively high (16.5 U mL⁻¹ h⁻¹) compared with other Bacillus species cited. Thus, it could be concluded that a newly isolated protease-producer employed in this study, B. subtilis FBL-1, proved a potential microbial resource for extracellular protease production.

Table 5 Enhancement of extracellular protease production from B. subtilis FBL-1 in this study

Optimization step	Protease activity (U mL^−1)	Production (fold)
Unoptimized (TSB)	84.8	1.0
Plackett–Burman design	331.8	3.9
Steepest ascent method	509.6	6.0
Box–Behnken design	594.4	7.0

---

Table 6 Comparison of protease production by B. subtilis FBL-1 with other Bacillus species recently reported

Microorganism	Substrate	Protease activity (U mL^−1)	Time (h)	Productivity (U mL^−1 h^−1)	Ref.
B. horikoshii	Soybean meal 1.5%	207.8	18	11.5	Joo et al.^34
	Casein 1%
Bacillus sp. L21	Soybean meal 2%	306.5	88	3.5	Tari et al.^35
	Maltose 2%
	Tween 80 0.035%
Bacillus sp. Ve-1	Gelatin 1%	410.0	66	6.2	Patel et al.^36
	Casein 1%
	NaCl 10%
B. licheniformis TD4	Xylose 1% urea 1%	141.5	24	5.9	Suganthi et al.^37
B. subtilis FBL-1	Fructose 3.242%	594.4	36	16.5	This study
	Yeast extract 1.502%

---

4. Conclusions

Several Gram-positive bacteria including Bacillus have been conventionally used as sources of proteases that are widely used in the food and detergent industries. The proteases are the most economically important enzyme produced by Bacillus. Therefore, we report the significant enhancement of protease production from a newly isolated soil microorganism which has been identified as B. subtilis. The combined utilization of statistical methods such as Plackett–Burman design, steepest ascent path analysis, and Box–Behnken design was successfully applied for optimization of extracellular protease production by B. subtilis FBL-1. In addition, response surface methodology was used to study the effects of several parameters affecting the response by varying them simultaneously in a limited number of experiments. The protease activity obtained from the optimized conditions was enhanced to 7.0-fold increased value compared to the fermentation using basal medium. Determination of enzymatic and biochemical properties of the protease as well as further enhancement of protease production by B. subtilis FBL-1 through fed-batch or continuous bioreactor are proposed for further study.

Acknowledgements

This study was financially supported by 2014 Yeungnam University Research Grant (214A367018).

Notes and references

1. L. V. Reddy, Y. J. Wee and H. W. Ryu, J. Chem. Technol. Biotechnol., 2008, 83, 1526.
2. D. Jain, I. Pancha, S. K. Mishra, A. Shrivastav and S. Mishra, Bioresour. Technol., 2012, 115, 228.
3. I. Younes, R. Nasri, I. Bkhairia, K. Jellouli and M. Nasri, Food Bioprod. Process, 2015, 94, 453.
4. H. S. Joo and C. S. Chang, Process Biochem., 2005, 40, 1263.
5. R. Oberoi, K. Beg, S. Puri, R. K. Saxena and R. Gupta, J. Microbiol. Biotechnol., 2001, 17, 493.
6. J. Dorado, J. A. Field, G. Almendros and R. S. Alvarez, Appl. Microbiol. Biotechnol., 2001, 57, 205.
7. S. Puri, O. Khalil and R. Gupta, Curr. Microbiol., 2002, 44, 286.
8. Q. K. Beg, V. Sahai and R. Gupta, Process Biochem., 2003, 39, 203.
9. K. Morihara, Adv. Enzymol., 1974, 41, 179.
10. Q. K. Beg, R. K. Saxena and R. Gupta, Process Biochem., 2002, 37, 1103.
11. A. Hameed, T. Keshavarz and C. S. Evans, J. Chem. Technol. Biotechnol., 1999, 74, 5.
12. X. Mao, Y. Shen, L. Yang, S. Chen and Y. P. Yang, Process Biochem., 2007, 42, 78.
13. R. L. Plackett and J. P. Burman, Biometrika, 1946, 33, 305.
14. Y. Han, Z. Y. Li, X. L. Miao and F. L. Zhang, Process Biochem., 2008, 43, 1088.
15. M. Y. Chang, G. J. Tsai and J. Y. Houng, Enzyme Microb. Technol., 2006, 38, 407.
16. C. J. De, S. Bouquelet, V. Dumortier and V. Duyme, J. Ind. Microbiol. Biotechnol., 2000, 24, 285.
17. C. Tari, N. Gogus and F. Tokatli, Enzyme Microb. Technol., 2007, 40, 1108.
18. J. He, Q. Zhen, N. Qiu, Z. Liu, B. Wang, Z. Shao and Z. Yu, Bioresour. Technol., 2009, 100, 5922.
19. S. Kaur, B. C. Sarkar, H. K. Sharma and C. Singh, Food Bioprocess Technol., 2009, 2, 96.
20. P. Chen, C. J. Chiang and Y. P. Chao, Biochem. Eng. J., 2010, 49, 395.
21. E. Tiwary and R. Gupta, Bioresour. Technol., 2010, 101, 6103.
22. Y. J. Wee, L. V. Reddy, K. C. Chung and H. W. Ryu, J. Chem. Technol. Biotechnol., 2009, 84, 1356.
23. O. S. Kim, Y. J. Cho, K. Lee, S. H. Yoon, M. Kim, H. Na, S. C. Park, Y. S. Jeon, J. H. Lee, H. Yi, S. Won and J. Chun, Int. J. Syst. Evol. Microbiol., 2012, 62, 716.
24. J. D. Thomson, T. J. Gibson, F. Plewniak, F. Jeanmougin and D. G. Haggins, Nucleic Acids Res., 1997, 25, 4876.
25. K. Tamura, D. Peterson, N. Peterson, G. Stecher, M. Nei and S. Kumar, Mol. Biol. Evol., 2011, 28, 2731.
26. L. L. Yuan, Y. Q. Li, Y. Wang, X. H. Zhang and Y. Q. Xu, J. Biosci. Bioeng., 2008, 105, 232.
27. X. J. Tang, G. Q. He, Q. H. Chen, X. Y. Zhang and M. A. Ali, Bioresour. Technol., 2004, 93, 175.
28. O. Folin and V. Ciocalteu, J. Biol. Chem., 1927, 73, 627.
29. R. A. Stowe and R. P. Mayer, J. Ind. Eng. Chem., 1996, 58, 36.
30. A. Pandey, P. Nigam, C. R. Soccol, V. T. Soccol, D. Singh and R. Mohan, Biotechnol. Appl. Biochem., 2000, 1, 135.
31. G. Dey, A. Mitra, R. Banerjee and B. R. Maiti, Biochem. Eng. J., 2001, 7, 227.
32. B. Chauhan and R. Gupta, Process Biochem., 2004, 39, 2115.
33. S. F. G. Oskouie, F. Tabandeh, B. Yakhchali and F. Eftekhar, Biochem. Eng. J., 2008, 39, 37.
34. H. S. Joo, C. G. Kumar, G. C. Park, K. T. Kim, S. R. Paik and C. S. Chang, Process Biochem., 2002, 38, 155.
35. C. Tari, H. Genckal and F. Tokatli, Process Biochem., 2006, 41, 659.
36. R. Patel, M. Dodia and S. P. Singh, Process Biochem., 2005, 40, 3569.
37. C. Suganthi, A. Mageswari, S. Karthikeyan, M. Anbalagan, A. Sivakumar and K. M. Gothandam, J. Gen. Eng. Biotechnol., 2013, 11, 47.

---

Footnote

† These authors equally contributed to this study.

---