**2. Results**

#### *2.1. Isolation and Screening of Keratinase Producing Bacterium*

In the present study, five prevalent colonies are that competent for sustainable growth on feather meal agar (FMA) were successfully isolated based on hydrolysis zone on FMA indicate the use of feather keratin as both carbon and nitrogen sources. The four isolates were able to utilize keratin in FMA for its growth. The morphology of each isolate is shown in Table 1. For further analysis, all pure strains were subjected to endospore screening for the best keratinase producing *Bacillus*. For this purpose, endospore-forming species was confirmed by the formation of green-colored spore after staining with malachite green and safranin. Among the tested isolate, three isolates were spore positive within 2 days of incubation in sporulation media signifying a potential member of *Bacillus* sp. The isolate was isolated UPM-AAG1, UPM-AAG6, and UPM-AAG14. A further screening process to select the highest keratinase producer was conducted based on bacterial growth and keratinase activity in 1% feather as sole carbon and nitrogen sources. The result suggests that the highest keratinolytic activity was isolate UPM-AAG1 (35.23 U/mL), followed by isolate UPM-AAG14 (33.97 U/mL), while isolate UPM-AAG6 resulted in the lowest keratinase production at only 25.56 U/mL for the same incubation time. The bacterial growth showed the same pattern where isolate UPM-AAG1 gave the highest bacterial count at 7.771 Log Colony Forming Unit or CFU/mL followed by isolate UPM-AAG6 at 7.628 Log CFU/mL and isolate UPM-AAG14 at 7.573 Log CFU/mL (Figure 1). Based on the results, isolate UPM-AAG1 was subjected further for identification study.

**Table 1.** Morphology of isolated microorganism.


**Figure 1.** Screening result of three isolates with 1% feather. Error bars represent mean ± standard deviation (n = 3).

#### *2.2. Identification of Keratinolytic Microorganism*

Micromorphology of isolate AAG1 was examined microscopically and demonstrated rod-shaped blue color bacterial cells, signifying their Gram-positive characteristic. Biochemical analysis showed positive results towards oxidase, catalase, Voges-Proskauer, and citrate test but negative result towards nitrate production. Further identification was supported by the 16S rRNA sequencing. BLASTn result showed that isolate AAG1 belonged to the *Bacillus* genus with high similarity percentage of (>99%). The phylogenetic tree constructed using partial 16S rRNA sequence and *Escherichia coli* strain U5/41 as the outgroup demonstrated that isolate AAG1 was not attached to any know species in the clade. However, bootstrap result AAG1 shows sequence similarity to *Bacillus safensis* strain FO-36b, *Bacillus pumilis* strain ATCC 7061, *Bacillus pumilis* strain SBMP2, and *Bacillus stratosphericus* strain 41KF2a with a bootstrap value of 78% (Figure 2). Therefore, UPM-AAG1 isolate was identified as *Bacillus* sp. strain UPM-AAG1 and deposited in the GenBank with the Accession No. MK285608.1.

#### *2.3. Optimization of Keratinase Activity Using Plackett Burman and Response Surface Methodology*

#### 2.3.1. Pre-Screening of Significant Parameters Using Plackett-Burman

Four independent factors (i.e., inoculum size (v/v), feather concentration (w/v), pH, and temperature) were screened to evaluate their effects on keratinase production using Plackett-Burman design. A total of 12 experimental variables generated using software screening for keratinase production, and their corresponding responds, as shown in Table 2a. The adequacy of the model was calculated using ANOVA analysis and presented in Table 2b. The model F value 70.33 indicates the model is significant with only 0.25% chance that a "Model F-value" this large could occur due to noise. The factors with *p* < 0.05 were considered to have a significant effect on the response. As presented in the table, all factors—temperature, inoculum size (v/v), pH, and feather concentration (w/v)—exert a positive effect on the model. Therefore, all four significant factors screened were further brought into the central composite design.

**Figure 2.** Phylogram (neighbor-joining method) showing the genetic relationship between strain UPM-AAG1 and other related reference micro-organisms based on the 16S rRNA gene sequence analysis. Species names are followed by the strain of their16S rRNA sequences. The numbers at branching points or nodes refer to bootstrap values, based on 1000 resamplings (GenBank MK285608.1).


**Table 2.** Prescreening of significant parameters using Plackett-Burman design matrix with keratinase activity as the response (±standard deviation, n = 3).


**Table 2.** *Cont*.

2.3.2. Optimization of Significant Variables Using Central Composite Design (CCD)

CCD was used to determine the optimum condition of the four selected significant variables (temperature, inoculum, pH, and feather concentration) for keratinase production using keratinase activity as the output response. A total of 30 experiments with di fferent combinations of the four selected variables were performed. The experimental designs used are shown in Table 3.

**Table 3.** Optimization of keratinase activity by strain AAG-1 using central composite design (CCD) with six center points showing observed and predicted values (±standard deviation, n = 3).


The responses were studied using four independent variables with six center point showing both observed and predicted values for keratinase activity. The multiple regression analysis of the observed responses resulted in the below quadratic equation:

Keratinase Activity = + 47.10 + 3.65\*A + 0.65\* B − 9.70\*C − 8.72\*D − 6.11\*A2 − 0.78\*B2 + 0.75\*C2 − 3.55\*D2 + 0.46\*A\*B + 2.70\*A\*C − 1.02\*A\*D + 5.50\*B\*C − 0.18\*B\*D − 16.97\*C\*D,

where A, B, C, and D, each represent concentrations (coded values) of temperature, pH, inoculum, and feather concentrations, respectively. From Table 4, it can be observed that all four linear terms (A, B, C, D), three squared terms (A2, C2, D2), and two quadratic terms (BC and CD) of the model were significant to the response, suggesting that keratinase production highly depends on the interactions between these factors.


**Table 4.** ANOVA analysis of CCD for optimization of keratinase activity by *Bacillus* sp. strain UPM-AAG1.

Based on the coded value below, the effects of inoculum and feather concentrations outweigh the effect of other factors.

Final equation in terms of actual factors:

Keratinase Activity = −114.67125 + 13.59533\*Temperature − 7.93667\*pH − 13.85500\*Inoculum + 29.97458 \* Feather − 0.24450\*Temperature2 − 0.49800\*pH2 + 0.12000\*Inoculum2 − 0.88750\*Feather 2 + 0.073000\*Temperature\*pH + 0.21600\*Temperature\*Inoculum − 0.10250\*Temperature\*Feather + 1.7600\*pH\*Inoculum − 0.070000\*pH\*Feather − 3.39500\*Inoculum\*Feather.

The predicted model was assessed further by RSM analysis. The 3D response plot for keratinase activity represents the interaction between two parameters at a time, while fixing the other parameter at zero levels (constant) for maximum keratinase production (Figure 3a–f). The predicted optimum points for temperature, pH, and inoculum and feather concentrations were 31.66 ◦C, 6.87, 5.01 (w/v),

and 4.53 (w/v), respectively, with an optimum keratinase activity of 60.5539 U/mL. Verification of the value obtained showed a close value of 60.02 U/mL indicating good agreement.

**Figure 3.** Response surface 3D plot showing the interaction of factors affecting keratinase production (**a**) pH and temperature, (**b**) inoculum size and temperature, (**c**) feather concentration and temperature, (**d**) pH and inoculum size, (**e**) pH and feather concentration, and (**f**) feather concentration and inoculum size.

Data fitness into the selected model was examined using diagnostic model plots (Supplementary Figure S2a–d). The plots are especially important in the evaluation of data error which varies from model predictions, which helps to assess and improve model adequacy. The actual versus predicted response plot obtained from the experiment (Figure S2a) showed a similar relationship between the predicted and actual values as the data points were clustered near the line dividing the plot into identical halves (45◦). Plotting the predicted values and studentized residuals (Figure S2b) further verified the suitability of the model. Studentized residues are utilized to indicate differences between the predicted value and the actual model responses. The experimental data exhibit slight or no abnormality based on visual observation of the normal probability plot (Figure S2c). To visualize the

distantly standout standard deviation, an outlier plot (Figure S2d) can show the presence of outlier(s). The result shows that the data falls between 3.5 and −3.5, suggesting the absence of outlier.
