Influence of fermentation time on P. pastoris GS115 growth and protein output
The optical density at 600 nm (OD₆₀₀) and the protein yield served as indicators to evaluate the impact of fermentation duration on the growth of P. pastoris GS115 and its protein production. The findings indicated a positive correlation between fermentation time and protein yield, with a plateau observed after 96 hours of fermentation. A significant increase was noted when comparing the 96-hour mark to the 72-hour mark (Figure 1; P ≤ 0.0001). At 96 and 120 hours, the protein yield was 726% and 780% higher than at 24 hours, respectively, with no significant difference between these two time points (Figure 1A). This suggested that the engineered strain achieved a stable protein yield by the 96-hour fermentation mark. Secreted protein is linked to the cell density, as mirrored by the OD₆₀₀ values, which is indicative of P. pastoris GS115 growth. Compared to the 24 hours, the OD₆₀₀ values were 176.33% and 195.61% higher after 72 and 96 hours of fermentation, respectively, with a significant difference between these two time points (Figure 1B; P ≤ 0.05). However, at 120 hours, the OD₆₀₀ value reached 202.15%, with no significant difference observed between the 96 hour and 120 hour marks. Mao et al.²² found that the concentration of rFIP-glu expressed by P. pastoris GS115 increased from 2-5 days and the OD₆₀₀ value increased from 3-5 days. Externally expressed protein might vary due to different strains and incubation time.²³ Appropriate incubation time was one of the important factors for obtaining high proteins expression.

Optimization of fermentation conditions via RSM
Employing the Box-Behnken design principle, a three-factor, three-level RSM experiment was devised, encompassing 17 test points. These tests were categorized to assess experimental error, with fermentation temperature, initial pH, and methanol concentration as independent variables, and protein yield as the dependent response. The experimental and model-predicted values from the response surface analysis are detailed in Table 2. Design-Expert 13.0 software was utilized to fit the experimental data through a quadratic polynomial regression model. The relationship between the response value Y (protein yield) and the independent variables could be described by the following equation:

Y = 260.85 – 8.40 X₁ + 6.07 X₂ + 4.80 X₃ – 5.07 X₁X₂ – 0.49 X₁X₃ – 1.24 X₂X₃ + 15.36 X₁² + 50.37 X₂² – 22.86 X₃²

In this equation, Y, X₁, X₂, and X₃ represented the predicted T36N yield, temperature, initial pH, and methanol content, respectively. The effects of independent  variables and their interaction on the production yield of T36N were analyzed and the results are shown in Figure 2. Further analysis found that the higher the initial pH tending to 6.5 and the lower the temperature tending to 26 °C, accordingly resulted in the 3D surface higher (Figure 2A) and the contoured lines denser (Figure 2B). At this point, the yield of T36N revealed a greater increase. The initial pH played a relatively dominant role in the interaction between temperature and initial pH, but it was not obvious. On the other hand, at the level of interaction between temperature and methanol content, when the methanol content was inclined to 1.00%, the 3D surface was relatively high (Figure 2C) and the contours were relatively dense (Figure 2D), in the direction of the lower temperature. At this point, the yield of T36N showed a greater increase. The fermentation temperature obviously played a dominant role in the interaction between temperature and methanol content. In a similar analysis, initial pH played a more dominant role than methanol content, and when methanol content was 1.00%, the higher initial pH was, the greater yield of T36N reached (Figures 2E-F).

In summary, the degree of influence of three factors on protein yield was ranked as B > A > C, that is, initial pH> temperature > methanol content. As a result, the optimal combination conditions of the model for protein yield were fermentation temperature of 26 °C, initial pH 6.5, methanol content of 1.00%.

Model variance analysis
The model’s variance analysis is presented in Table 3. The regression model was subjected to variance analysis, revealing an overall P-value of ≤0.0001, signifying the model’s high statistical significance. The determination coefficient R² equaled 0.9767, indicating that 97.67% of the response variability was attributed to the model’s nine factors, suggesting a robust fit. Additionally, the adjusted determination coefficient R²_Adj equaled 0.9467, further validating the model’s predictive accuracy.

Table (3):
ANOVA of regression model

To test the reliability of the regression fitting model, 9 sets of validation experiments were conducted (Table 4). The results showed that the correlation coefficient between measured and predicted protein yield was 0.9356 > 0.9, which demonstrated that the regression fitting model was valid. According to the regression fitting model of response surface optimization, the predicted yield of T36N was 346.13 mg^.L^-1 under the optimal combination conditions (temperature 26 °C, initial pH 6.5, and methanol content 1.00%) for inducing expression of T36N by P. pastoris GS115. At the same time, the P. pastoris GS115 fermentation experiment was conducted under the above optimal combination conditions for 5 replicates. The results showed that the actual expression yield of T36N was 346.64 ± 3.47 mg^.L^-1 under the optimal condition, while the original T36N yield was 260.85 ± 2.45 mg^.L^-1, obtaining an increase of 32.89% (Supplementary Table S1). These data indicated that the optimal fermentation combination conditions obtained by RSM were consistent with the actual conditions. Lin et al.²⁴ obtained the expression of rFIP-fve produced by P. pastoris GS115, with the optimal yield of 258.2 mg^.L^-1. The same host, under different fermentation conditions or expressing different heterologous proteins, will lead to different yields.

Table (4):
Validation results of the model

P. pastoris expression system is one of the most popular and standard tools to produce recombinant protein in the biomedical industry. However, the fermentation conditions of engineered P. pastoris are different from those of wild strains. In addition to temperature and pH, it is also necessary to consider the induction time of foreign protein expression and the concentration of inducer.^23,25 For example, methanol is the inducer of gene expression in most P. pastoris expression systems. However, high concentration of methanol is toxic to the cells, while low concentration of methanol is not conducive to improving the productivity of heterologous proteins.²³ To obtain a lot of FIP-glu, the FIP-glu had successfully expressed in various host cells, and the yield of rFIP-glu in P. pastoris ranged from 191.2 mg^.L^-1 to 270 mg^.L^-1, which was not enough for a pharmaceutical use. After optimization of the fermentation parameters, Mao et al. ²² improved the yield of rFIP-glu up to 368.71 mg^.L^-1 by P. pastoris in the shake-flask, under the condition of 1.0% methanol, initial pH 6.5 and 26 °C at 280 rpm for 4 days.²² In the present study, the gene FIP-glu_T36N was a gene FIP-glu variants with a N-linked glycan modification. It was still unknown whether the engineered yeast had the potential to produce protein T36N on a large scale after the gene FIP-glu_T36N was constructed into P. pastoris. This experiment demonstrated that T36N yield under these optimal fermentation conditions reached up to 346.64 ± 3.47 mg^.L^-1, which was increased by 32.9% comparing with not optimized.

In this study, three factors including temperature, initial pH and methanol content were taken into consideration. It is well-known that the suitable fermentation temperature is conducive to the growth of P. pastoris and the synthesis or hydrolysis degradation of protein.^26,27 Some studies have suggested that glycosylation could affect the thermostability of proteins to some extent.^28,29 In our study, the optimal fermentation temperature of the engineered P. pastoris was 26 °C that generally consistent with the previous results.²² The glycosylation modification does not seem to have much effect on the yield and thermal stability of the protein, which need further experiments confirmed. Secondly, initial pH of media also played an important role in cell growth, protein formation, and protein stability. In our study, initial pH was the most important of the three variables during P. pastoris GS115 fermentation for T36N glycoprotein. Under the condition of initial pH 6.5, the T36N yield reached up to the highest, which was basically consistent with the results of similar studies in previously.²² In Mao et al. study, the initial pH had a better effect on the rFIP-glu production of P. pastoris GS115 culture than fermentation temperature and methanol concentration, and under the optimum fermentation parameters (fermentation temperature 26 °C, initial pH 6.5, methanol concentration 1%), rFIP-glu concentration reached 368.71 ± 10.34 μg·mL^-1.²² Generally, the optimal initial pH was obtained by running a series of fermentations at different pH. The yield of recombinant porcine lactoferrin expressed in P. pastoris increased to 12 mg·L^-1 through increasing initial pH from 6.0 to 7.0.³⁰ Therefore, it is necessary to control initial pH in the media and the best pH is related to protein properties, especially stability. Besides, methanol contributed to cell growth and was an inducer for the expression of foreign proteins in methanotrophic P. pastoris. It was reported that in the P. pastoris expression system at least 0.5% concentration of methanol was necessary for recombinant protein production, and methanol concentration should not exceed 2-2.5% for cell toxic effect of high level of methanol.²³