**1. Introduction**

Collagen is the main component of the extracellular matrix. Collagen is the predominant constituent of skin, tendons and cartilage and is the main organic component of bones, teeth and corneas [1,2]. Collagen is not only a structural protein with high tensile strength, but also a protein that affects cell differentiation, migration and attachment. Collagen is an inexpensive and resourceful meat by-product that is used extensively as a food additive to increase the texture, water-holding capacity and stability of several food products. The main sources of collagen, such as bovine and porcine skin and bone, are derived from land-based animals. Recently, components of marine organisms, including fish skin, bone and scales, have received increasing attention as sources of collagen [3]. Additionally, with collagen being the most abundant protein in all higher organisms, collagenases have diverse biotechnological applications [4–6].

Collagenases, a class of proteases with high specificity for collagen, can cleave peptide bonds in collagen [7]. Collagenases are usually considered virulence factors and normally target the connective tissue in muscle cells and other organs. On the other hand, collagenases can be used to produce bioactive collagen peptides, which have been widely used in the pharmaceutical, food and cosmetic industries. To date, a number of collagenases of bacterial origin have been identified and characterized, such as MCP01 from *Pseudoalteromonas* sp. SM9913 [8], Col H from *Clostridium histolyticum* [9] and collagenases from *Vibrio alginolyticus* [10]. However, the productivity of most collagenases is not high enough for industrial application. Recently, marine bacteria have become known as important sources for the identification of novel enzymes. Compared with collagenases from land animals, marine collagenases usually have higher catalytic efficiency towards marine collagen from fish skin and bone [8]. Therefore, collagenases obtained from marine bacteria in particular have received attention owing to their diverse properties. Simultaneously, there have been attempts to increase the production of marine-derived collagenases and to optimize the fermentation conditions for the production of these enzymes [11].

Statistical methods such as factorial design, central composite design and response surface methodology (RSM) have been frequently used to optimize process parameters for the production of different kinds of enzymes [12]. Statistically-designed experiments are more effective than those designed using classical optimization strategies because they can be used to study many variables simultaneously with a low number of observations, saving time and expense. The Plackett–Burman design provides a way to rapidly screen for main variables with significant effects on a specified parameter from a large number of variables, and the information obtained can be retained for further optimization [13,14]. Response surface methodology (RSM) can identify interactions among various factors while requiring less experimental resources. Therefore, RSM, as a collection of statistical techniques that are useful for designing experiments, building models, evaluating the effects of different factors and searching for optimal conditions of studied factors for predictable responses, has been successfully applied in many areas of biotechnology, including the protein enzyme industry [15]. *Pseudoalteromonas* sp. SJN2, isolated from the inshore environment of the South China Sea, can produce extracellular collagenases that have high catalytic efficiency. The aim of this study was to use RSM to optimize the fermentation medium to increase collagenase production by *Pseudoalteromonas* sp. SJN2. In addition, marine collagen extracted from seafood by-products was digested by crude extracellular collagenases secreted by *Pseudoalteromonas* sp. SJN2, and the antioxidant activity of hydrolysates was measured.

#### **2. Results**

### *2.1. Purification and Enzymatic Properties of Ps sp. SJN2 Collagenases*

Collagenases SJN2 was sequentially purified using ammonium sulfate precipitation, anion exchange and size exclusion chromatography; Col SJN2 purification is shown in Figure S1 and Table S1 (Supplementary Material). Currently, as a highly effective feed additive, enzymes used in the food industry are crude enzymes because the cost of pure enzymes is high and because the operation is tedious [16]. In addition, combinations of enzymes can result in a range of possible biological properties for the corresponding hydrolysates.

Compared with other strains isolated from the inshore environment of the South China Sea at the same time, *Pseudoalteromonas* sp. SJN2 collagenases, which appear as bright strips in Figure 1a Line 1, have some advantages in terms of the number and brightness of the strips, which suggest a strong ability to hydrolyze collagen.

The effects of ions on the enzymatic activities of *Ps* sp. SJN2 collagenases was measured, as shown in Figure 1b. Some ions (Zn2+, Mg2+ and Ca2+; at lower concentrations; red column) were seen to promote *Ps* sp. SJN2 collagenase activity or were less toxic to strain *Ps* sp. SJN2 than EDTA and Cd2+.

Swelling of insoluble collagen after hydrolysis by *Ps* sp. SJN2 collagenases was observed (Figure 1c–g). With increasing enzyme hydrolysis time, the porosity of the insoluble collagen increased. As observed by SEM, the collagen structure in the control group remained compact and tough, while in the experimental group, the bulky collagen fiber bundles changed into small dispersed collagen fibers (Figure 1e), and the sub-fibers (Figure 1f) were exposed, which indicated the collagen-hydrolysis ability of *Ps* sp. SJN2 collagenases.

**Figure 1.** Enzymatic properties of *Pseudoalteromonas* sp. SJN2 collagenases. (**a**) Comparison of catalytic ability on gelatin of collagenases from *Ps* sp. SJN2, *Ps* sp. SBN2-2, *Ps* sp. SGS2-2, *Ps* sp. SWN-1, *Ps* sp. SWN-2, *Vibrio* sp. HK3-2, *Vibrio* sp. SJN4, and *Planococcus* sp. SYT1; (**b**) effect of ions against *Ps* sp. SJN2 collagenases' enzyme activity, red column for 2 mM, blue column for 10 mM; (**c**) swelling effect of *Ps* sp. SJN2 collagenase on bovine collagen type I: Tube 1: 200 μL 20 mM PBS treated for 24 h at 37 ◦C; Tubes 2–5: 200 μL *Ps* sp. SJN2 crude enzyme treated for 1 h, 5 h, 12 h and 24 h; (**d**–**g**) swelling of insoluble collagen was observed under SEM, the (d) control group with 8000× magnification and (**e**–**g**) the experimental group with 8000×, 20,000× and 250,000× magnification.

### *2.2. Catalytic Efficiency of Collagenases from Ps sp. SJN2*

Compared with commercially available terrestrial Col H, collagenases from *Ps* sp. SJN2, as enzymes from marine sources, have a competitive edge in the degradation of collagen from marine-biological sources. Figure 2a shows the hydrolysis of fish skin collagen by collagenases from *Ps* sp. SJN2 into smaller molecular collagen-peptides than those produced using Col H, after a 1-min reaction at <sup>45</sup> ◦C with both collagenases having low concentrations (2 mg·mL−<sup>1</sup> Col H and 0.2 mg·mL−<sup>1</sup> crude enzymes from *Ps* sp. SJN2), which indicated that crude collagenases from *Ps* sp. SJN2 had higher catalytic efficiency than Col H. In the degradation of marine collagen, marine collagenases exhibit properties such as low enzyme concentrations and rapid catalysis, and the same results can be obtained for the degradation of collagen from fish scales (Figure 2b) and fish bone (Figure 2c). This result showed that crude collagenases from *Ps* sp. SJN2 could be more suitable for the digestion of marine collagen, which was seen to be spliced into small polypeptide fragments. Improving the production of collagenases from *Ps* sp. SJN2 would be useful for future application of this enzyme.

**Figure 2.** Differences between Col H and collagenases from *Ps* sp. SJN2 in the degradation of fish collagen and the antioxidative test. (**a**–**c**) Fish-collagen substrates from skin, scale and bone after hydrolysis are shown in Line 1: fish collagen; Line 2: fish collagen hydrolysed by Col H with a reaction time of 1 min; Line 3: Col H; Line 4: fish collagen hydrolysed by collagenases from *Ps* sp. SJN2 at a reaction time of 1 min; Line 5: collagenases from *Ps* sp. SJN2. (**d**,**e**) DPPH scavenging of five collagen peptides (d) and the extracted collagen (e). The reaction concentration of extracted collagen for the black column is 0.4 mg·mL−1, and that for the blue column is 0.2 mg·mL−1. The green is vitamin C at 0.1 mg·mL−1, as a positive control. (**f**–**i**) ORAC assay for four kinds of marine collagen peptides: (f) octopus flesh, (g) seabream scale, (h) Spanish mackerel bone and (i) salmon skin. The red curve is vitamin C at 100 <sup>μ</sup>g·mL<sup>−</sup>1, as a positive control, while the black is PBS, as a negative control. For each substrate, the blue curve is 400 <sup>μ</sup>g·mL−<sup>1</sup> of hydrolysed collagen; the purple curve is 200 <sup>μ</sup>g·mL−1; the green curve is 100 <sup>μ</sup>g·mL<sup>−</sup>1; and the navy curve is 50 <sup>μ</sup>g·mL<sup>−</sup>1.

#### *2.3. Initial Screening of Significant Fermentation Conditions*

The effects of temperature, initial pH, seed inoculation, culture time, corn meal concentration, bran liquid concentration and soybean meal concentration, which are the seven variables associated with collagenase production in the Plackett–Burman design, are shown in Tables 1 and 2. A random experimental program was devised by using Design-Expert-8.0.6 software. Analysis of variance (ANOVA) results are listed in Table 3.

**Table 1.** Variables tested for collagenase production using Plackett–Burman designs and the levels for each variable.


**Table 2.** Plackett–Burman design matrix with corresponding results.


**Table 3.** Variance analysis of the Plackett–Burman linear model.


<sup>a</sup> *R*<sup>2</sup> = 0.9447. <sup>b</sup> Model terms are significant.

Mathematical analysis of this model shows that the model was significant (*F*-value = 9.77, *p*-value = 0.0219), with a complex correlation *R*<sup>2</sup> = 0.9447, indicating that the model can explain 94.47% of the experimental results. The multiple regression Equation (1) describes the mathematical relationship between collagenase production and fermentation conditions:

$$\mathbf{Y} = 198.38 - 26.16\mathbf{X}\_1 + 0.86\mathbf{X}\_2 - 11.01\mathbf{X}\_3 - 18.89\mathbf{X}\_4 + 1.45\mathbf{X}\_5 + 8.16\mathbf{X}\_6 + 17.52\mathbf{X}\gamma \tag{1}$$

For every decrease in *X1*, a decrease of 26.16 units in *Y* is predicted; a similar effect is predicted for the other variables. This relationship between *Y* and *X* was further used in the steepest ascent experiment. Table 3 indicates that the terms with *p*-values less than 0.05, i.e., *X1* (temperature, *p* = 0.0054), *X4* (culture time, *p* = 0.0169) and *X7* (soybean meal concentration, *p* = 0.0216), had relatively significant effects on collagenase yield. Therefore, the steepest ascent experiment was designed with the three significant variables mentioned above.

The path of steepest ascent was determined based on a decrease of 2.00 ◦C for *X1* (temperature), a decrease of 0.64 d for *X4* (culture time) and an increase of 3.00 g·L−<sup>1</sup> for *X7* (soybean concentration). The movement was generated along the path until no improvement of *Y* (yield of collagenases from *Ps* sp. SJN2) occurred. After the first step of the coordinates, decreasing collagenase yield was observed, as shown in Table 4. Consequently, this combination (temperature, 16 ◦C; culture time, 3.36 d; and soybean concentration, 33 g·L−1) was selected as the middle point (zero level) for a final optimal design by response surface methodology.


**Table 4.** Experimental details and results of the steepest ascent design with four steps approaching the response region.

<sup>a</sup> Coefficient estimate in Table 3. <sup>b</sup> −1/+1 level in the Plackett–Burman design in Table 1. <sup>c</sup> Average value of the −1 level and +1 level in Table 1. <sup>d</sup> An appropriate ratio determined by the experimenter, based on experiential knowledge and laboratory conditions, to reasonably adjust the step coefficient in this model. <sup>e</sup> Step length, calculated as Equation (4).

#### *2.4. Further Optimization by Response Surface Methodology*

The fermentation conditions were further explored by response surface methodology based on central composite design. The design matrix and the corresponding results of RSM are shown in Tables 5 and 6, and the ANOVA results of the RSM are displayed in Table 7.


**Table 5.** Levels of significant variables for the response surface design.


**Table 6.** The matrix of the response surface experiment and the corresponding results.

**Table 7.** Variance analysis of the response surface methodology.


<sup>a</sup> *R*<sup>2</sup> = 0.9694, C.V. = 7.01%. <sup>b</sup> Model terms are significant.

The adequacy of the model was assessed using ANOVA (Table 7). The coefficient of determination (*R*2) was 0.9694 for collagenase production, indicating good agreement between the experimental and predicted values [17]. The results demonstrated that the 96.94% variability in the response experiment could be explained by this model. The R2 is always between zero and 1.0, and a value closer to 1.0 indicates a stronger model and better response. A very low value of the coefficient of variation (C.V., 7.01%) indicated high reliability and precision of the experimental simulation. The *F*-value of the model was 35.19, which implied that the model was statistically significant. The smaller the *p*-value, the more significant the corresponding coefficient could be considered. The *p*-value of 0.0001, which is less than 0.05, suggested that the model terms were significant. These results indicated that temperature, culture time and soybean concentration have a direct relationship with collagenase production.

A multivariate regression model with interaction terms was characterized by the following regression Equation (2):

$$\begin{aligned} Y &= 287.87 + 26.22A + 18.21B + 33.95\text{C} + 10.15AB + 1.58AC + 8.93BC \\ &- 24.74A^2 - 25.18B^2 - 48.46C^2 \end{aligned} \tag{2}$$

where *<sup>Y</sup>* is the predicted collagenase production (U·mL<sup>−</sup>1) and *<sup>A</sup>*, *<sup>B</sup>* and *<sup>C</sup>* are the values of temperature ( ◦C), culture time (d) and soybean powder concentration (g·L<sup>−</sup>1), respectively.

The three-dimensional response surface plots and two-dimensional contour plots were used to elucidate the relationship and interaction effects of the chosen fermentation conditions for maximal production of collagenases from *Ps* sp. SJN2, as shown in Figure 3. In each sketch, two variables linearly changed within the experimental range, while the other variable remained constant at the central point. For the two-dimensional contour plots, the shape of the contour described the interaction significance of the paired variables, and the elliptical appearance suggested an extremely significant interaction.

**Figure 3.** Response surface plots (left) and two-dimensional contour plots (right) of the effects of (**a**) culture time vs. temperature, (**b**) soybean concentration vs. temperature and (**c**) soybean concentration vs. culture time on the yield of collagenases from *Ps* sp. SJN2.

Figure 3a shows the response surface and the corresponding contour plot of culture time (*A*) vs. temperature (*B*), keeping soybean concentration (*C*) at the zero level. It can be noticed from the surface that the optimal yield of collagenases from *Ps* sp. SJN2 was observed when the culture time was approximately at the −1 level, while the temperature was nearly at the +1 level. The two-dimensional contour plot of culture time (*A*) vs. temperature (*B*) showed an elongated pattern, suggesting that the interaction between culture time and temperature has a significant effect on the yield of collagenases from *Ps* sp. SJN2. Similar profiles were also observed in Figure 3b,c.

The experimental data were fitted to Equation (2), and the optimal levels of the significant variables were determined to be as follows: culture time, 3.72 d; temperature, 17.32 ◦C; and soybean concentration, 34.23 g·L<sup>−</sup>1; with a predicted maximum production of collagenases from *Ps* sp. SJN2 at 306.683 U·mL<sup>−</sup>1.

#### *2.5. Experimental Validation of the Model*

The validation of the statistical model and the regression equation was conducted using the optimized conditions. The predicted response for collagenase production was 306.68 U·mL<sup>−</sup>1, and the observed experimental value was 322.58 U·mL−1. The experimental production was close to the predicted response, and the yield of collagenases from *Ps* sp. SJN2 increased 2.2-fold compared to the yield from the original fermentation scheme. The optimal fermentation scheme was further examined by the continuous determination of collagenase activity by bacterial biomass evaluation and gelatine-immersing zymography, as shown in Figure 4.

As shown in Figure 4a, the activity of the collagenases from *Ps* sp. SJN2 increased steadily as *Ps* sp. SJN2 grew in the logarithmic-growth phase, and the activity was highest when the biomass entered the stationary phase at approximately 72 h of cultivation time. Then, within approximately 24 h, the activity of the collagenases from *Ps* sp. SJN2 started to decrease, which may be explained by the decrease in biomass over the same period or by transformation of part of collagenases into other non-catalytic proteins. This finding affirmed that collagenase activity is related to biomass, which is characterized by the phenomenon that collagenase activity and biomass both exhibit maximum values at approximately 3.5 d, which is in agreement with the optimal model prediction.

**Figure 4.** Time course of collagenase yield vs. biomass of *Ps* sp. SJN2 (**a**) and gelatine-immersing zymography (**b**).

In Figure 4b, the gelatine-substrate immersing zymography validation showed that when the proteases reach a maximum yield at 3.5 d, which was also proven by the time course experiment, there were eight bright strips, which represented activated zymogen and proteases (primarily metalloproteases and serine proteases; Figure S2, Supplementary Material) that might possess the ability to hydrolyze collagen, which proved that the increased activity of the crude enzyme under optimized conditions was caused by the increased yield. Furthermore, the visible strips of proteases with high molecular weight grew progressively darker over culture time, which might partly be due to enzyme autolysis into other mature forms.
