Vacuum Impregnation Process Optimization for Tilapia with Biopreservatives at Ice Temperature

The vacuum impregnation (VI) process was used to pretreat tilapia fillets with biopreservatives at −2 °C. Response surface methodology (RSM) was utilised to optimize processing conditions, including vacuum pressure (pv), vacuum maintenance time (t1), and atmospheric pressure recovery time (t2), which were determined to be 67.73 kPa, 23.66 min, and 8.87 min, respectively. The anticipated values for the aerobic plate count (APC), total volatile basic nitrogen (TVB-N), and comprehensive score (CS) were 5.17 lg CFU/g, 14.04 mg/100 g, and 0.98, respectively. Verification experiments were conducted, and the experimental results for APC and TVB-N deviated from the predicted values by 0.19% and 0.64%, respectively. After 30 days of storage following VI and atmosphere impregnation (AI) pretreatment, the water-holding capacity (WHC), APC, TVB-N, hardness, and whiteness were determined. On the 30th day, the results for VI pretreatment were 63.38%, 6.27 lg CFU/g, 17.41 mg/100 g, 3.11 N, and 47.73, respectively. Compared with AI pretreatment, WHC, hardness, and whiteness increased by 14.8%, 18.6%, and 6.3%, respectively, whereas APC and TVB-N decreased by 11.3% and 29.6%, respectively. This study demonstrates that when biopreservatives are applied during the pretreatment process, VI technology can be utilised to facilitate their penetration into the interior of tilapia, hence significantly enhancing the effect of ice-temperature preservation.


Introduction
China is the largest producer of tilapia in the world. Production is anticipated to reach 13.25 million tonnes by 2025 [1]. Tilapia is a high-protein perishable fish. Improper storage increases the total amount of bacteria in the product, and protein degradation generates volatile basic nitrogen, which alters the water-holding capacity, texture, and colour of the fish. Consequently, tilapia product preservation technology is essential for processing, transport, storage, and marketing. Ice-temperature storage is common practice for preserving the quality of aquatic products during storage and distribution. Because the hardness, colour, water-holding capacity, and other characteristics of seafood are better in ice-temperature samples than in frozen storage, consumers prefer ice temperature storage products more [2,3]. Usually, a sufficient quantity of biopreservatives is added to extend the storage time at ice temperature. Compared to chemical preservatives, biopreservatives are benign, non-toxic, and edible, which is why customers choose biopreserved products.
For a long time, impregnation was the most prevalent method of food preservation. However, because the impregnation process takes longer, preservatives remain on the surface and are unable to penetrate the interior of the food tissue; hence, the expected preservation conditions are not satisfied.
As a result, various researchers have investigated increased impregnation processes, with vacuum impregnation (VI) being the most prevalent. The VI process includes submerging the food in a solution containing a specified concentration of preservatives or In a previous study, the research team optimised a formulation of biopreservatives (sodium alginate 8 g/L, Nisin 0.8 g/L, and sodium erythorbate 7.5 g/L, in mass concentrations) for the preservation of tilapia at ice temperature (−2 • C) [36]. Compared to the control, the shelf life of tilapia fillets treated with the biopreservatives were extended to 22 days. In this study, VI technology was used to improve the fresh-keeping effect and to investigate the effects of the VI process parameters p v , t 1 , and t 2 on the quality of tilapia fillets. Using response surface methodology (RSM), the study intended to optimise the VI process parameters. Additionally, experiments were conducted with the optimized parameters to determine the influence of biopreservatives' supplemental VI technology on the quality of tilapia stored at −2 • C.

Materials, Reagents, and Equipment
Tilapia (Oreochromis niloticus) was purchased at the Huguang Market in Mazhang District of Zhanjiang City, Guangdong Province, China. This experiment employed only food-grade reagents. Sodium alginate was obtained from Yuanye Biotechnology Co., Ltd., Shanghai, China. Nisin was acquired from Wanlida Biotechnology Co., Ltd., Hebei, China. Sodium erythorbate was purchased from Liyuan Food Additives Co., Ltd., Guangzhou, China. 2-thiobarbituric acid was obtained from Sinopharm Chemical Reagent Co., Ltd., Shanghai, China; plate counting agar was obtained from Land Bridge Technology Co., Ltd., Beijing, China.

Preparation of Tilapia Fillets
After purchasing live tilapia, approximately 0.9 kg each, they were pre-cultured in a laboratory tank at a room temperature of 24 • C. In preparation for conducting the studies, the head, bone, and skin of the tilapia were removed, and the meat on both sides of the back was sliced into approximately 1 cm-thick fillets with a stainless steel knife. The fillets were cut into 12 cm × 5 cm × 1 cm rectangles and weighed around 80 ± 2 g.
The fish fillets were placed in beakers containing preservative liquid (25%, w/w) produced with sterile water [36]. The beakers were then placed within the vacuum chamber of the vacuum apparatus. Experiments were conducted in triplicate.

Impregnation Weight Gain Rate, WGR
At the end of the impregnation period, the fish fillets were taken out, placed in a sterile environment, drained until there were no remaining dops, and weighed. After absorbing the surface moisture of the fish fillets with sterile filter sheets, the fillets were weighed again. The weight growth rate of the impregnated fish fillets is calculated using Equation (1): where WGR is the impregnation weight gain rate, %; m 1 is the mass of the fish fillet before treatment, g; m 2 is the mass of the fish fillet at the end of treatment, g.

TVB-N
TVB-N concentration was determined using a Kjeldahl nitrogen analyzer in accordance with the "Chinese National Food Safety Standard, Determination of Volatile Base Nitrogen in Food, GB/5009.228-2016".

Aerobic Plate Count, APC
The total number of bacterial colonies was determined using the aerobic plate count method during storage. The APC was determined following the procedure outlined in "Chinese National Food Safety Standard, Food microbiology test-Determination of aerobic plate count, GB/4789.2-2016". The expressed amount of APC was in log colony-forming units (CFU).

Water-Holding Capacity, WHC
According to Lakshmanan's method [37], weigh 2 g of fish meat accurately, wrap it in two layers of known-weight filter paper, and place it in a centrifuge tube. After centrifuging at 10 • C and 3000 rpm for 10 min, the filter paper was weighed. WHC was calculated according to Equation (2): where m 0 is the mass of the tilapia meat, g; m 1 is the weight of the filter paper before centrifugation, g; m 2 is the weight of the filter paper after centrifugation, g.

Hardness
The samples were evaluated using the texture analyzer's texture profile analysis (TPA) mode. The diameter flat-bottomed column probe P/5 was used, the induction force was set to 1000 N, the test speed was set to 60 mm/min, the starting force was set to 0.5 N, the deformation was set to 50%, and the interval between two depressions was set to 1 s. Five measurement were taken of each piece of fish, and the average was calculated.

Whiteness
The value of whiteness can be used to determine the colour of the fish fillets. The colorimeter was used to determine the L* value (lightness on a scale of 0 to 100 from black to white), the a* value (+for red or −for green), and the b* value (+for yellow or −for blue) of the fish fillets [38]. The value of whiteness is calculated using Equation (3): where W represents the whiteness index; L* is lightness, a* is the red-green value, and b* is the yellow-blue value.
2.9. Experimental Design and Statistical Analysis 2.9.1. Single-Factor Experiments To investigate the effect of a single parameter on the experimental results, single-factor experiments using the control variables method were conducted, which means that only one variable was used to represent the influencing factors on the experimental results, and all other variables remained constant [39]. This section investigates the effect of p v , t 1 , and t 2 on the quality index of tilapia fillets for WGR, APC, and TVB-N.
Each time, five groups of prepared fish fillet samples (total 150 fillets) were processed for single-factor tests. Based on previous experiments, fist, t 1 , and t 2 were fixed at 15 and 6 min, respectively, and p v was changed to 0 (atmospheric pressure), 20, 40, 60, and 80 kPa. Then, p v and t 2 were held constant at 60 kPa and 6 min, respectively, while t 1 was adjusted to 5, 10, 15, 20, and 25 min. Finally, p v and t 1 were set to 60 kPa and 15 min, respectively, and t 2 was varied to 3, 6, 9, 12, and 15 min. Following the preceding procedures, the fillets were placed in a sterile environment to drain until no water remained. The drained fish fillets were subsequently vacuum-packed and stored at −2 • C. The indicators of WGR, APC, and TVB-N were measured every 5 days in all experimental groups.

RSM Experiments
RSM was used to optimise the VI process for tilapia fillets. The Box-Behnken design (Table 2) was employed, because this investigation required three levels with equal spacing and fewer runs [32,40]. The RSM experiment was created, analyzed, optimized, and plotted by Design Expert software (Version 8.0, Stat-Ease, Inc., Minneapolis, Minnesota, USA). On day 25, two response parameters were evaluated experimentally: APC and TVB-N. Through calculation, a single response, the comprehensive score (CS), was simultaneously determined. The statistics are summarized in Table 3. Using regression analysis [41], second-order polynomial equations were fitted to the responses in terms of the independent variables. Equation (4) is a quadratic model that was applied to characterise the response variables. This model describes the response variables Y (APC, TVB-N, and CS) as a function of the factor variables X i (i assuming values from 1 to 3) for all applicable VI solutions:  In this study, Equation (5) was derived for the solutions based on Equation (4), Y = a 0 + a 1 p v + a 2 t 1 + a 3 t 3 + a 12 p v t 1 + a 13 p v t 2 + a 23 t 1 t 2 + a 11 p v 2 + a 22 t 1 2 + a 33 t 2 2 (5) where a 0 represents the constant, a i represents the linear effect, a ii represents the quadratic effect, and a ij represents the factor interaction effect. Consequently, the main effect coefficients are a 1 , a 2 , and a 3 , the quadratic main effect coefficients are a 11 , a 22 , and a 33 , and the two factor interaction coefficients are a 12 , a 13 , and a 23 . Positive or negative coefficient values in Equation (4) imply a reaction-promoting or reaction-inhibiting influence, respectively. Following the development of polynomial models, analysis of variance (ANOVA) was used to examine how well the model described the data at a 95% confidence level, and p-value criteria were applied to establish statistical significance. R 2 is the model's coefficient of determination, and the lack of fit is used to evaluate the model's precision. The statistical significance of the coefficients was determined using the software's built-in ANOVA tool.
To initiate the optimization process, a composite score was calculated for APC and TVB-N using the Membership comprehensive scoring method, with a lower value indicating superior performance. This is because the smaller the aforementioned statistics were, the higher the storage quality of the fillets. According to Equation (6), the APC and TVB-N memberships were weighted independently: where c i is the index value, c imin is the minimum value of the index, c imax is the maximum value of the index, and I = 1 or 2, where 1 stands for APC, and 2 stands for TVB-N. Comprehensive scores (CS) were calculated with Equation (7): where l 1 represents the Membership of APC, l 2 represents the Membership of TVB-N, coefficient a represents the weight of APC, and coefficient b represents the weight of TVB-N. Given that both metrics have an equal impact on the fish fillets' quality, a = b = 0.5. The CS should be as large as possible, according to the analysis of Equation (6). The mean ± standard deviation of the results was calculated. The statistical significance was determined using SPSS (Version 18.0, SPSS Inc., Chicago, IL, USA) statistical analysis software for the one-way analysis of variance (ANOVA) with Duncan's test. The significance of differences was established at the 5% probability level (p < 0.05). The analysis figures of the experimental data were designed with Origin software (Version 8.0, OriginLab, Northampton, MA, USA). Table 4 depicts the impact of various p v on the WGR, APC, and TVB-N. The WGR increased significantly (p < 0.05) when p v was increased from 0 to 60 kPa; however, when p v was increased from 60 kPa to 80 kPa, the WGR changed insignificantly (p > 0.05). The APC and TVB-N values for each fillet group increased as storage duration increased. The APC tended to initially decrease and subsequently increase as p v increased over the same storage time. This tendency was at its lowest point when sustained for 10, 15, and 20 d at 60 kPa, but it increased when p v approached 80 kPa. The TVB-N value declined as p v increased on 5 d and 10 d, but on 15 d and 20 d, it decreased to a minimum at 60 kPa and then increased at 80 kPa. This demonstrated that p v does not imply the higher the better. In this investigation, the optimal p v was 60 kPa; when p v was greater than 60 kPa, the storage quality of the tilapia fillets was no longer significantly improved, and it might even deteriorate. According to the research of J.W. Kang et al. [42], this is mostly likely because the biopreservative liquid has completely filled the fish cell spacing and reached a saturated state at 60 kPa, the internal and external pressures have reached equilibrium, and the biopreservatives can no longer penetrate the fish tissue. Therefore, increasing p v above 60 kPa has no obvious impact on the preservation effect. Moreover, as p v increased, the fish tissues might have been progressively damaged, resulting in a decrease in storage quality.  Table 5 demonstrates the effect of different t 1 on the WGR, APC, and TVB-N. When p v and t 2 are held constant, the WGR value of the fish fillets steadily increased as t 1 increased from 5 to 20 min, but became negligible after 15 min. When t 1 was extended to 25 min, the WGR decreased. At the same storage time, as t 1 increased, the APC and TVB-N values of each group decreased significantly from 5 to 20 min, but increased at 25 min, indicating that a longer impregnation period is not always preferable for tilapia fillets.  Table 6 depicts the influence of various t 2 on the WGR, APC, and TVB-N. As can be seen, increasing t 2 from 3 to 15 min had no discernible effect on WGR, although it tended to increase and then decrease. The WGR increased as t 2 grew at and before 9 min, but there was a slight decrease at 12 and 15 min, indicating that t 2 does not imply that a longer impregnation period is better. The APC and TVB-N decreased initially (3~9 min) and then increased (9~15 min) in response to an increase in t 2 for the same storage duration. This may be because during the t 1 stage, the cell structure of the fish fillets acquired a vacuum pressure, and then during the t 2 stage, the system pressure was restored to atmospheric pressure. Obviously, the pressure in the t 2 stage was greater than the pressure in the t 1 stage. Under the pressure difference, the external biopreservation solution gradually permeated the interior of the fish fillets, and the permeation amount reached a saturation point at 9 min. Increasing t 2 beyond 9 min had no significant effect on the biopreservatives' fresh-keeping effect. This phenomenon is similar to Fito's experiments with fruits [43], but fish meat is quite different from fruits, and the effect of t 2 in this investigation was insignificant (p > 0.05). According to the results of the single-factor experiments, the RSM experiment's parameters, p v , t 1 , and t 2 , were selected as indicated in Table 2. Table 7 presents the linear, quadratic, and interaction effects of the independent variables on APC, TVB-N, and CS. In addition, Table 8 shows the regression coefficients of responses in terms of actual factors specified in the original units for each factor. The correlation coefficient R 2 is significantly greater than 0.9, suggesting that the model fits the data well. In this investigation, the R 2 values for APC, TVB-N, and CS were reported to be 0.955, 0.934, and 0.955, respectively, indicating that the model significantly matched the experimental data. Nearly all quadratic factors exhibited highly significant effects (p < 0.01) on response variables. Among the linear factors, t 1 had a significant impact on all of the response parameters, whereas p v had a significant effect on APC and CS, but an insignificant influence on TVB-N. The linear factor t 2 and all interaction factors were determined to be insignificant on all response variables. Note: SS Sum of squares, DF Degree of freedom. * Significant at p < 0.05; ** Significant at p < 0.01; *** Significant at p < 0.001. APC was significantly influenced by the linear factors p v (p < 0.01) and t 1 (p < 0.001), as well as the quadratic factors p v , t 1 , and t 2 (p < 0.01). Because the quadratic regression coefficients of APC were all positive, when p v , t 1 , and t 2 increased, the APC was observed to initially decrease and then increase; the three-dimensional response surface must be a surface with an upward opening (Figure 1). TVB-N was significantly impacted by linear factor t 1 (p < 0.01), and the quadratic factors p v (p < 0.01), t 1 (p < 0.05), and t 2 (p < 0.01). The remaining factors were insignificant. It was revealed that as p v , t 1 , and t 2 grew, TVB-N first decreased and then increased due to all the positive quadratic regression coefficients; the three-dimensional response surface must be a surface with an upward opening, too ( Figure 2).     Computed from APC and TVB-N, which have identical weights for the effect on fish storage quality, CS was significantly influenced by linear terms of p v (p < 0.05), t 1 (p < 0.001), and all the quadratic factors (p < 0.01). In accordance with the mathematical relationship (Equations (6) and (7)), the smaller APC and TVB-N are, the higher the storage quality and the greater the CS. Furthermore, CS should be opposite to APC and TVB-N, as suggested by the results of the variance analysis (Table 8), which revealed that the quadratic terms were characterised by negative regression coefficients. Consequently, when the independent variables p v , t 1 , and t 2 increased, CS initially increased and then decreased; the threedimensional response surface must be a surface with a downward opening (Figure 3).

Optimisation of the Process Variables
Using a numerical optimization strategy based on the desirability approach and Design Expert 8.0 software, the optimal settings of the independent variables for VI technology were identified. The optimisation criterion for restrictions was to minimise APC and TVB-N while maximising CS. The independent variables for the optimization technique were kept within the experimental range. The optimised values for pv, t1, and t2 were 67.73 kPa, 23.66 min, and 8.87 min, respectively. The estimated APC, TVB-N, and CS values under ideal conditions were 5.17 lg CFU/g, 14.04 mg/100 g, and 0.98, respectively. To accommodate laboratory operations, experiments were conducted in triplicate with the In conclusion, there must be a combination of p v , t 1 , and t 2 that simultaneously minimises APC and TVB-N while maximising CS.
According to the F test (F value in Table 7), the influence sequence of the three independent variables on the three response parameters is consistent: t > p v > t 2 , indicating that t 1 had a larger influence than p v , and p v has a greater effect than t 2 . Figures 1-3 also support this conclusion. Therefore, if VI technology is used to enhance the effect of biopreservatives on the storage of tilapia at ice temperature, the appropriate vacuum duration should be chosen first, followed by the determination of the vacuum pressure, and the atmospheric pressure restore time has the least influence on storage quality.

Optimisation of the Process Variables
Using a numerical optimization strategy based on the desirability approach and Design Expert 8.0 software, the optimal settings of the independent variables for VI technology were identified. The optimisation criterion for restrictions was to minimise APC and TVB-N while maximising CS. The independent variables for the optimization technique were kept within the experimental range. The optimised values for p v , t 1 , and t 2 were 67.73 kPa, 23.66 min, and 8.87 min, respectively. The estimated APC, TVB-N, and CS values under ideal conditions were 5.17 lg CFU/g, 14.04 mg/100 g, and 0.98, respectively. To accommodate laboratory operations, experiments were conducted in triplicate with the adjusted parameters pv = 68 kPa, t1 = 23 40 , and t2 = 8 50 . The experimental values for APC and TVB-N were 5.17 ± 0.06 lg CFU/g and 14.23 ± 0.18 mg/100 g, respectively. CS was determined to be 0.96 compared to the predicted value of 0.98; the relative error is around 2.6%, indicating a satisfactory fit between the predicted and experimental values.
Pretreated with the biopreservatives, the storage quality of AI and optimal VI was compared for 30 days of storage at −2 • C. Every five days, indicators of APC, TVB-N, WHC, hardness, and whiteness were tested, and data were collected (Figure 4). The subsequent sections investigated the variation in these indicators.
APC and TVB-N were 5.17 ± 0.06 lg CFU/g and 14.23 ± 0.18 mg/100 g, respectively. CS was determined to be 0.96 compared to the predicted value of 0.98; the relative error is around 2.6%, indicating a satisfactory fit between the predicted and experimental values.
Pretreated with the biopreservatives, the storage quality of AI and optimal VI was compared for 30 days of storage at −2 °C. Every five days, indicators of APC, TVB-N, WHC, hardness, and whiteness were tested, and data were collected (Figure 4). The subsequent sections investigated the variation in these indicators. •

TVB-N
The TVB-N value was applied as an indicator of fish spoilage [44,45], so it is an important indicator of fish storage quality. The majority of the increase in TVB-N resulted from the degradation of amino acids, protein, and some nitrogen-containing substances by spoilage bacteria and endogenous enzymes, resulting in the formation of volatile bases [28]. These enzymes resulted in the synthesis of fish-off-flavoring nitrogen compounds, including ammonia, monoethylamine, dimethylamine, and trimethylamine [46]. On day 25, the TVB-N value of AI exceeded the maximum limit standard for TVB-N in "China's national food safety standards, fresh and frozen animal and aquatic products" (20 mg/100 g), whereas the TVB-N value of the fish fillets pretreated with VI remained within the standard range on day 30, but exceeded the limit standard on day 35. By day 35, the fish fillets had already begun to show deterioration; therefore, other quality-related indicators, APC, WHC, hardness, and whiteness were only detected for 30 days.
The TVB-N values of the fish fillets increased with storage time in both groups (Figure 4a), and these changes were almost identical to the changes in APC (Figure 4b). At the completion of the storage period, the TVB-N concentration increased from 9.79 mg/100 g to 24.71 mg/100 g (AI) and 17.41 mg/100 g (VI). With VI pretreatment, the TVB-N value did not increase significantly until day 15, especially during the first five days (p > 0.05). This demonstrates that VI treatment enhances the preservation effect of the •

TVB-N
The TVB-N value was applied as an indicator of fish spoilage [44,45], so it is an important indicator of fish storage quality. The majority of the increase in TVB-N resulted from the degradation of amino acids, protein, and some nitrogen-containing substances by spoilage bacteria and endogenous enzymes, resulting in the formation of volatile bases [28]. These enzymes resulted in the synthesis of fish-off-flavoring nitrogen compounds, including ammonia, monoethylamine, dimethylamine, and trimethylamine [46]. On day 25, the TVB-N value of AI exceeded the maximum limit standard for TVB-N in "China's national food safety standards, fresh and frozen animal and aquatic products" (20 mg/100 g), whereas the TVB-N value of the fish fillets pretreated with VI remained within the standard range on day 30, but exceeded the limit standard on day 35. By day 35, the fish fillets had already begun to show deterioration; therefore, other quality-related indicators, APC, WHC, hardness, and whiteness were only detected for 30 days.
The TVB-N values of the fish fillets increased with storage time in both groups (Figure 4a), and these changes were almost identical to the changes in APC (Figure 4b). At the completion of the storage period, the TVB-N concentration increased from 9.79 mg/100 g to 24.71 mg/100 g (AI) and 17.41 mg/100 g (VI). With VI pretreatment, the TVB-N value did not increase significantly until day 15, especially during the first five days (p > 0.05). This demonstrates that VI treatment enhances the preservation effect of the biopreservatives, effectively extending the shelf life of tilapia fillets stored at ice temperature, because VI treatment effectively increases the penetration of preservatives into the fish tissue and inhibits protein oxidation. This result is comparable to the findings of X. Zhao et al. [28], who observed the lowest TVB-N value in tilapia fillets throughout storage in the combined pretreatment with fish gelatin and grape seed extract assisted by VI, compared to the control and the groups without VI. A. Andres-Bello et al. [30] studied the gilthead sea bream fillets pretreated with VI in solution containing lactic acid bacteria and nisin, respectively, and stored at 4 • C for 15 days; the TVB-N value assisting with VI was also lower than the control throughout the storage time. However, the TVB-N and other physico-chemical properties exhibited insignificant differences between the experiment groups; this could be because the VI process parameters used in the research were not large enough compared to the other researchers (Table 1).
• APC APC growth curves in fish fillets were fitted with the Baranyi equation [47]. With and without VI, the APC values increased gradually (Figure 4b). The average initial APC concentration of samples was 4.49 lg CFU/g. In the VI group, the APC rose more slowly and did not reach a significant level (p < 0.05) until day 15, particularly during the first 5 days (p > 0.05). On day 25, the APC of AI and VI pretreatment was 6.34 ± 0.25 lg CFU/g and 5.16 ± 0.10 lg CFU/g, respectively. The final APC of AI and VI pretreatment was 6.51 ± 0.16 lg CFU/g and 5.78 ± 0.21 lg CFU/g, respectively.
The hydrodynamic mechanism (HDM) [43] of VI may have led to a decrease in APC concentrations with VI pretreatment. This demonstrates that the auxiliary VI technology improves component penetration into the fish fillet and inhibits APC proliferation during the preservative impregnation pretreatment.
• WHC WHC refers to the capacity of fish to retain water during processing and storage. WHC is mostly related to the quantity of water that flows with difficulty, the protein gel, and the electrostatic charge of the fish. The higher the WHC, the stronger the binding ability of the fish network structure to water and other substances, indicating a denser spatial network structure in the gel [48,49]. Therefore, the WHC of fish meat is an important indicator of its preservation quality.
During storage, the multiplying bacteria in the fillets produce proteases that degrade the proteins and disrupt their gel structure, resulting in a reduction in WHC (Figure 4c). The initial WHC of both groups of fillets was 82.2%. During the storage period, the WHC of the tilapia fillets declined gradually in both pretreatment groups, but it decreased more dramatically in the AI pretreatment group, particularly in the first 10 days, decreasing to 63.80%. In the VI pretreatment group, the decline was slower, with no significant decrease in WHC over the first 5 days and a substantial decrease on day 10 to 72.21%. At day 30, the WHC of AI and VI pretreatment was 55.20% and 63.38%, respectively. This indicates that the adoption of VI technology in pretreatment reduced the degradation of the protein gel structure and maintained the amount of electrostatic charge, hence increasing the WHC of the fillets. Wang Z Y et al. [50] revealed that pulsed vacuum brining enhanced the WHC of lamb compared to atmosphere brining. Leal-Ramos et al. [20] determined that VI is an effective approach for increasing the moisture content of meat. This may be due to the fact that VI can expand muscle fibre to enhance the WHC [51].

• Hardness
Hardness is considered the most essential textural property of fish [38]. The highactivity autolytic enzymes hydrolyze proteins and other connective tissues, thereby accelerating muscle deterioration [52] and diminishing tissue hardness. Dunajski et al. [53] reported that fish with higher moisture content had a softer texture, suggesting that water content can also affect the hardness of fish. In other words, the fish with lower water content had a harder texture. In this investigation, the hardness of fish samples from both groups decreased during the preservation period (Figure 4d). In the AI group, the hardness declined dramatically over the first 15 days, from 5.44 N to 2.48 N, whereas in the VI group, the hardness decreased significantly in the first 10 days, from 5.44 N to 3.68 N, and then slowed. At the same storage time, the hardness difference between the two groups was not statistically significant. But the hardness of the VI group was still greater than that of the AI group. The AI group's lowest point of hardness occurred at 15 d (2.48 N), after which it increased slightly, whereas the VI group had a low point-though not its lowest-at 15 d (3.21 N). In the first 15 days, the effect of moisture loss on increasing hardness was much smaller than the effect of protein gel structure degradation on decreasing hardness, causing a rapid reduction in hardness. At 15 d, the degradation of the gel structure reached an extreme value, and the water loss continued to increase the hardness, resulting in a slight increase in hardness at 20 d. At 20 d, the water loss may have reached an extreme value, whereas the breakdown of the protein continued to decrease the hardness. The hardness did not change much after 20 d, but decreased gradually overall. The lowest hardness of VI pretreatment was observed at 30 d (3.11 N). This suggests that the VI technology can assist the biopreservatives in protecting the protein structure, delaying the deterioration of hardness and preserving the fillets' aesthetic appearance. In contrast, increasing water loss during the VI dehydration of apples [54] and bamboo shoots [55] softened the tissue and reduced the hardness. This may be because fish have a completely different tissue structure than that of plants.
• Whiteness It has been demonstrated that colour change in fillets is highly correlated with storage quality [38], and whiteness is the most important colour indicator for white-meat fish such us tilapia, whose fresh meat is a beautiful white colour. Whiteness is the comprehensive value of L*, a*, and b* according to Equation (3), where L* is positively correlated with whiteness, and a* and b* are negatively related with whiteness. Therefore, whiteness decreases as the value of red, green, blue, or yellow increases. During ice temperature storage, the formation of ice crystals causes the rupture of muscle and fat cells, causing the infiltration of water and lipids onto the surface and the aggregation of myofibrillar proteins, which can enhance the reflection of light and increase brightness, thereby enhancing whiteness. However, the disruption of protein structures by enzymatic reaction and bacterial degradation leads to the oxidation of myoglobin to metmyoglobin, which has a reddish hue, and consequently reduces the whiteness. Moreover, the oxidation of lipids causes the aggregation of aldehydes and migration to the surface, resulting in a yellowish tinge that decreases the whiteness of the fillets [56][57][58][59].
In this study, the whiteness of VI was consistently greater than that of AI during the same period of storage (Figure 4e). Both groups reached their lowest value at 30 d, with 44.92 (AI) and 47.73 (VI). The initial whiteness of the fresh fillets was 52.73, and it decreased significantly over the course of the first 10 days in both groups, to 46.33 (AI) and 48.58 (VI). This indicated that during the initial period of storage, protein structural degradation and lipid oxidation predominated. After that, for AI pretreatment, the whiteness increased slightly at 15 d, and then decreased slowly, whereas for VI pretreatment, whiteness increased slightly until 20 d and subsequently declined slowly. Increasing whiteness during the middle of the storage period indicated that moisture migration was the dominant process at this stage. This transformation was similar to the change in hardness. The VI treatment facilitated effective penetration of the biopreservatives into the interior of the fillets, establishing a protective layer both inside and outside the fillets, slowing the process of protein denaturation, water loss, and fat oxidation, thereby preserving the colour of the fish fillets.

Conclusions
In this investigation, RSM was employed to optimise the process parameters of VI pretreatment of tilapia fillets in order to improve the preservation impact of biopreservatives stored at −2 • C. The following are our conclusions: 1. Using APC, TVB-N, and CS as response variables, the optimised VI process parameters obtained by RSM were p v = 67.73 kPa, t 1 = 23.66 min, and t 2 = 8.87 min; according to ANOVA analysis, among the three parameters, the importance on quality effect was t 1 > p v > t 2 ; that is, t 1 had the greatest impact on the quality of tilapia, whereas t 2 had the least impact.
2. The optimal parameters for verification experiments were adjusted to p v = 68 kPa, t 1 = 23 40 , and t 2 = 8 50 . After 25 days of storage, the difference between the obtained and predicted CS value was 2.6%.
3. Using the optimised process parameters, tilapia fillets were pretreated with biopreservatives, and the quality indicators of tilapia fillets stored at −2 • C for 30 days were monitored. WHC, hardness, and whiteness increased by 14.8%, 18.6%, and 6.3%, respectively, in comparison to the AI effect, and APC and TVB-N decreased by 11.3% and 29.6%, respectively. The storage time for the tilapia was extended to 30 days. With the optimal parameters, it is obvious that VI technology can enhance the impregnation effect of biopreservatives, improve the storage quality of tilapia, and prolong the shelf life of tilapia at ice temperature.