Optimized Co-Fermentation of Seed Melon and Z. bungeanum Seed Meal with Saccharomyces cerevisiae L23: Valorization into Functional Feed with Enhanced Antioxidant Activity

Abstract

This study aimed to enhance the value of agricultural by-products by developing seed melon compound fermented feed (SMFF) using Saccharomyces cerevisiae L23. A two-stage optimization strategy was implemented. First, seed melon juice seed culture medium (SMCM) composition and fermentation conditions were optimized to maximize S. cerevisiae L23 biomass through single-factor and response surface methodology (RSM) approaches. The SMCM medium was optimized to contain 0.06% MgSO₄·7H₂O, 0.2% KH₂PO₄, 0.65% (NH₄)₂SO₄, 0.1% pectinase, and 1.0% urea, and fermentation conditions with inoculation amount, fermentation time, fermentation temperature, and glucose addition were 6%, 28 h, 30 °C, and 0.5%, respectively. Furthermore, SMFF fermentation parameters were optimized via RSM, achieving S. cerevisiae L23 (10.35 lg CFU/g) and sensory evaluation score (83.1) at substrate ratio of 7:3 (seed melon juice: Zanthoxylum bungeanum seed meal), inoculation amount of 8%, and fermentation time of 36 h. Fermentation process significantly improved the nutritional profile of SMFF, increasing crude protein (13%) and vitamin C (VC) content (21%) while reducing neutral detergent fiber/acid detergent fiber (NDF/ADF) levels. SMFF also improved in vitro antioxidant capacity, with higher DPPH, ABTS, hydroxyl radical, and superoxide anion scavenging activities compared to SMFF control. This process efficiently valorized agricultural by-products into nutritionally enriched functional feed.

1. Introduction

Lanzhou seed melon (Citrullus lanatus), a creeping herb indigenous to Northwest China, is rich in diverse nutrients, including citrulline, phenolic antioxidants, and minerals (calcium, phosphorus, and potassium); annual production exceeds 1 million tons. Its high sugar content establishes it as an ideal carbon source for microbial fermentation [1,2]. Currently, the utilization of seed melon primarily focuses on the extraction of bioactive compounds such as cucurbitacin E, total saponins, citrulline, and polysaccharides, with studies demonstrating the hypoglycemic and hypolipidemic activities of seed melon juice [3,4,5,6,7]. In the context of feed applications, previous research has explored the fermentation of seed melon by-products. Mai et al. utilized a mixed consortium of Lactobacillus acidophilus, Saccharomyces cerevisiae, and Bacillus subtilis for solid-state fermentation of seed melon pulp and skin, optimizing for lactic acid production and general nutritional improvement [8]. However, the potential of seed melon in functional feed systems remains underexplored, particularly in multi-substrate strategies. Zanthoxylum bungeanum seed meal (ZSM), a by-product of Z. bungeanum oil extraction, exhibits high protein and unsaturated fatty acids content. Microbial fermentation further has been shown to enhance its nutritional profile by reducing anti-nutritional factors and increasing the crude protein levels [9]. Both crops represent major characteristic industries in Gansu Province. Nevertheless, seed melon utilization remains largely restricted to seed extraction, with a utilization rate below 10% for the fruit itself. Approximately 90–95% of the by-products (peel and pulp) are discarded post-seed extraction, leading to significant resource waste and environmental pollution [10]. With the expansion of seed melon and Z. bungeanum cultivation, the volume of these by-products continues to increase. Developing integrated strategies for their utilization could not only enhance economic value but also address challenges in agricultural waste management.

Saccharomyces cerevisiae has considerable efficacy as a microbial cell protein producer, capable of secreting enzymes that degrade macromolecules, improve feed conversion rates, and enhance utilization efficiency [11]. S. cerevisiae combined with Lactobacillus plantarum has been shown to improve the nutritional quality of fermented feed, regulate the rumen environment, and enhance digestibility [12]. Similarly, combinations of S. cerevisiae, Lactiplantibacillus plantarum, and Bacillus velezensis have been reported to reduce spoilage risks and improve storage stability in silage and fermented feed [13]. However, these studies predominantly focused on conventional substrates and mixed cultures.

Compared to previous approaches, in the present study, we propose a novel multi-substrate fermentation system integrating seed melon juice and Z. bungeanum seed meal (ZSM), leveraging the high sugar content of seed melon juice to support microbial growth and the protein-rich matrix of ZSM to achieve nutritional balance. Furthermore, we employ a specialized strain, S. cerevisiae L23, under a two-stage optimization strategy to precisely control the fermentation process and enhance functional properties. Beyond conventional nutritional parameters, we place a strong emphasis on evaluating in vitro antioxidant activities—an aspect not thoroughly explored in earlier seed melon fermentation studies. No previous research has reported on the co-fermentation of these two agricultural by-products, nor comprehensively analyzed the resultant nutritional and functional characteristics.

Response surface methodology (RSM) is a collection of statistical and mathematical techniques used for developing, improving, and optimizing processes, which is particularly useful for modeling and analyzing the interactive effects of multiple variables to determine optimum conditions [8]. Therefore, this study pioneers a two-stage response surface methodology (RSM) optimization strategy for producing seed melon compound fermented feed (SMFF) using S. cerevisiae L23. Initially, single-factor and RSM approaches were employed to maximize S. cerevisiae L23 biomass in a seed melon juice-based seed culture medium (SMCM), identifying critical additives and fermentation conditions. Subsequently, RSM was applied to synchronize substrate ratio, inoculation amount, and fermentation time to achieve high S. cerevisiae L23 counts while maintaining superior sensory quality. Finally, a comprehensive evaluation of nutritional parameters, in vitro antioxidant activities (DPPH, ABTS, hydroxyl radical, and superoxide anion scavenging), and sensory attributes was conducted to validate the functional superiority of the product. This research not only enhances the value of underutilized agricultural by-products but also establishes a replicable framework for developing functional feeds using tailored microbial fermentation, thereby contributing to sustainable livestock production.

2. Materials and Methods

2.1. Experimental Materials and Chemicals

The Lanzhou seed melon used as the substrate for fermentation was purchased from local vegetable trading market located in Lanzhou, Gansu Province, China. The raw materials of Z. bungeanum Maxim seed meal (ZSM) were obtained from a local Z. bungeanum retail market (Tianshui, Gansu, China). Saccharomyces cerevisiae L23 was isolated and preserved at −80 °C in the Central Laboratory of School of Life Sciences and Engineering, Northwest Minzu University. This specific strain had significant enhancements in protein content and antioxidant activity in preliminary laboratory trials. S. cerevisiae L23 was activated and cultivated in YPD liquid medium, which contained (w/v) 1% yeast extract, 2% peptone, 2% glucose, and 2% sodium chloride. Fermentation bags with a one-way venting device (size 20 cm × 25 cm) were acquired from Aoke experimental equipment department (Lanzhou, Gansu, China). Pectinase (from Aspergillus niger, light-yellow powder, activity 40 U/mg) was purchased from Solarbio Technology Co., Ltd. (Beijing, China). All other chemicals and reagents used were analytical grade.

2.2. Pretreatment of Raw Materials

Fresh and non-rotten Lanzhou local seed melon was processed into seed melon juice and obtained filtrate, which served as the substrate for the seed melon juice seed culture medium (SMCM) and as one of the main components of seed melon compound fermented feed (SMFF). ZSM was ground into powder using a home-scale grinder machine (HBS-4500A, Guangzhou, China)and stored in an airtight container (Rongma Co., Ltd., Suqian, China) at 4 °C for further utilization.

2.3. Optimization of SMCM

The initial composition of the SMCM and the ranges for fermentation conditions were selected based on prior study from literature on Saccharomyces cerevisiae cultivation [8] and supported by preliminary single-factor experiments conducted in our laboratory to identify approximate feasible ranges. Single-factor experiments of SMCM were conducted with the S. cerevisiae L23 count as the optimization index. The additives MgSO₄·7H₂O (0.03%, 0.04%, 0.05%, 0.06%, and 0.07%, w/v); KH₂PO₄ (0.05%, 0.10%, 0.15%, 0.20%, and 0.25%, w/v); (NH₄)₂SO₄ (0.35%, 0.45%, 0.55%, 0.65%, and 0.75%, w/v); pectinase (0.05%, 0.10%, 0.15%, 0.20%, and 0.25%, w/v); and urea addition (0.6%, 0.8%, 1.0%, 1.2%, and 1.4%, w/v) were selected as single factors to optimize the composition of SMCM. Furthermore, inoculation amount (4%, 5%, 6%, 7%, and 8%, v/v), fermentation time (16, 20, 24, 28, and 32 h), fermentation temperature (26, 28, 30, 32, and 34 °C), and glucose addition (0, 0.5%, 1.0%, 1.5%, and 2%, w/v) were selected as influencing single-factors to optimize the fermentation conditions of SMCM.

2.4. Preparation of SMCM

A loop of S. cerevisiae L23 was aseptically inoculated into a 50 mL triangular flask containing 20 mL of YPD liquid medium. The culture was incubated at 28 °C and 180 rpm for 24 h in a temperature-controlled shaker (Shanghai scientific experimental instrument Co., Ltd., Beijing, China). The resulting seed culture (cell density of 2.00 × 10⁸ CFU/mL) was transferred into the SMCM and incubated under the same conditions (28 °C, 180 rpm, and 24 h). The count of S. cerevisiae L23 was determined using the plate count method according to the Chinese National Standard GB 20287-2006 for agricultural microbial inoculants. The seed melon juice seed culture medium (SMCM) was formulated with S. cerevisiae L23 as the fermentation strain, seed melon juice as the carbon source, and urea and (NH₄)₂SO₄ as the nitrogen sources and supplemented with appropriate amount of MgSO₄·7H₂O, KH₂PO₄, and pectinase. The culture was incubated to enhance the biomass yield of S. cerevisiae L23 in the SMCM for use in subsequent experiments.

2.5. Preparation of SMFF

The sample SMFF was prepared by mixing the seed melon juice and ZSM in a certain ratio and then loaded into the fermentation bag with an amount of 500 g/bag, fully stirred and mixed to evenly ferment the SMFF under certain fermentation conditions.

2.6. Standard Curve of S. cerevisiae L23

S. cerevisiae L23 was inoculated into the SMCM, which was cultivated for 120 h under the optimal conditions. In order to plot the biomass growth curve of S. cerevisiae L23 in the SMCM, regular sampling plate counts were required, and three replicates for all time points within three dilution gradients and the results were averaged.

2.7. Optimization of SMFF Fermentation Processing

The fermentation process of SMFF was optimized to obtain the best set of process parameters. Response surface methodology (RSM) was used to generate the experimental designs, statistical analysis, and regression model with the help of Design Expert Software Version 13 (Stat-ease Inc.) [14]. S. cerevisiae L23 count of SMFF was determined by plate counting as the optimization index. Three independent variables, namely substrate ratio (8:2, 7:3, 6:4, 5:5, 4:6, and 7:3), inoculation amount (4%, 5%, 6%, 7%, 8%, and 9%), and fermentation time (12, 24, 36, 48, and 60 h) of SMFF, were selected for single-factor experiments to determine the main influencing factors. Based on the results of the single-factor experimental data, the optimization of the SMFF fermentation process was conducted using a Box–Behnken design with three factors at three levels (high, medium, and low, code +1, 0, and −1, respectively) designed for each independent variable condition [15]. The experimental design levels for the response surface are exhibited in

Table 1. Experimental factors and levels design.

Level	Substrate Ratio (g/g): A	Inoculation Amount (%): B	Fermentation Time (h): C
−1	8:2	7%	24
0	7:3	8%	36
1	6:4	9%	48

.

2.8. Sensory Evaluation of SMFF

This research was approved by the Ethics Committee of Food Science, Northwest Minzu University (Lanzhou, China; approval no. SM20240515) on 15 May 2024. The privacy of the subjects was respected during the experiment, and informed consent was obtained from them. The SMFF was investigated for sensory evaluation by 10 semi-trained panelists (aged between 28 and 45, consisting of 5 women and 5 men) with sensory evaluation experience. Sensory evaluation was conducted according to the description by ISO 13299: 2016 [16]. The panelists were mostly animal husbandry scholars and teaching faculty from the Department of Animal Science, having rich experience in sensory evaluation parameters and indicators [17]. The sensory evaluation are shown in

Table 2. Sensory evaluation standard of SMFF.

Index	Evaluation Standard
	Good	General	Worse
Color	Chocolate-brown (21–25)	Yellowish-brown (11–20)	Brown-black (1–10)
Flavor	The unique fragrance of seed melon; pleasant and savory sour fragment (21–25)	Faint sour taste; delicate fragrance of seed melon; light pepper fragrance (11–20)	Heavier sour taste and numb-taste (1–10)
Texture	Well structured and loose (21–25)	Well structured with a few clumps (11–20)	Decay with large clumps (1–10)
Fermentation degree	Uniform, completely(21–25)	Uncompletely (11–20)	Decay and deterioration (1–10)

. The average scores for each parameter of SMFF samples were calculated.

2.9. Determination of Physiochemical Properties

The lactic acid (LA), acetic acid (AA), propionic acid (PA), and butyric acid (BA) content of SMFF were determined by1260 Infinity II HPLC (Lichrospher C18 reversed phase column 250 × 4.6 mm, 5 μm, DAD detector, Agilent Technologies Co., Ltd., Hangzhou, China) and 3 mmol/L perchloric acid as mobile phase [18]. The pH value of SMFF was measured using pH meter (Mettler–Toledo International Inc., Shanghai, China). The moisture content was evaluated with the direct drying according to GB 5009.3-2016, and the crude ash content was assessed with the method of drying method as GB 5009.4-2016, which is expressed as the percentage of the weight of the sample. The crude protein content was conducted by the Kjeldahl nitrogen determination as GB/T 6572-2018 [19], and the crude fat content was analyzed according to the Soxhlet extraction described in GB 5009.6-2016. The reducing sugar content was determined based on Fehling’s reagent method [20]. The VC content was determined using the 2,6-dichlorophenol titration method as GB 5009.86-2016. The content of dry matter (DM), neutral detergent fiber (NDF), acid detergent fiber (ADF), Ca, and phosphorus was evaluated using the methods proposed by [18].

2.10. Verification of Optimal Condition of SMFF

In order to determine the S. cerevisiae L23 count, fermented SMFF was prepared as the optimal fermentation conditions obtained from RSM by using the gradient dilution and plate paint isolation methods, while another portion of the samples was dried at 60 °C, ground, and sieved through a 60-mesh sieve to determine the physiochemical properties of SMFF.

2.11. Determination of In Vitro Antioxidant Activity

In vitro antioxidant capacity of SMFF was evaluated by measuring the 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,2′-azinobis (3-ethylbenzthiazoline-6-sulfonic acid) (ABTS), hydroxyl radical scavenging ability, and superoxide anion radical scavenging ability. VC was used as a positive control.

2.11.1. DPPH Radical Scavenging Activity

The samples were processed as proposed by [17] with modifications, which were dried at 50 °C to reach a constant weight in an oven (GRX-9203A, Zhengji Instrument Co., Ltd. Suzhou China). Then they were ground with a mortar and sieved through a 60-mesh sieve. In total, 50 mg of the SMFF ground samples were mixed with 1 mL of the 75% alcohol. The mixture was extracted at 45 °C for 40 min and centrifuged at 12,000 rpm for 12 min at 4 °C. The supernatant of SMFF was harvested and kept in a 4 °C refrigerator (Haier, Shanghai, China)as samples further. The DPPH radical scavenging activity of SMFF was determined as described by [21] with slight modifications. Equal volume sample was mixed with 0.2 mmol/L DPPH anhydrous ethanol and reacted in the dark at room temperature for 1 h. The absorbance values were measured at 517 nm. The DPPH radical scavenging activity was calculated using the formula:

$$\mathrm{DPPH} (\%)=[1-(A_S-A_C)/A_0]\times 100$$

(1)

where A_S is the absorbance value of the sample, A_C is the absorbance value of the control with anhydrous ethanol instead of DPPH, and A₀ is the blank control absorbance of deionized water instead of the sample.

2.11.2. ABTS Radical Scavenging Capacity

The ABTS solution was prepared as described by [22]. Amounts of 150 μL of ABTS working solution and 50 μL of the sample supernatant were added to a 96-well plate and reacted for 10 min in the dark at room temperature. The absorbance value was measured at 734 nm. The scavenging activity was calculated as follows:

$$\mathrm{ABST} (\%)=[1-(A_S-A_C)/A_0]\times 100$$

(2)

where A_S is the absorbance value of the sample, A_C is the absorbance value of the control with deionized water instead of ABTS working solution, and A₀ is the absorbance of blank control as deionized water instead of the sample supernatant.

2.11.3. Hydroxyl Radical Scavenging Activity

The hydroxyl radical scavenging activity was measured as described by [21]. An amount of 1 mL of sample solution was mixed with 1 mL of 9 mmol/L FeSO₄, 3 mL of 9 mmol/L salicylic acid-ethanol, and 1 mL of 8.8 mmol/L H₂O₂. After incubation at 37 °C for 15 min, absorbance was measured at 510 nm. The scavenging activity was calculated as follows:

$$\mathrm{Scavenging\; ability} (\%)=[1-(A_S-A_C)/A_0]\times 100$$

(3)

where A_S is the absorbance value of the sample, A_C is the absorbance value of the control with deionized water instead of H₂O₂, and A₀ is the absorbance of blank control as deionized water instead of the sample supernatant.

2.11.4. Superoxide Anion Radical Scavenging Activity

The samples were processed as described proposed by [17] with modifications. According to the method by [23], 3 mL Tris-HCl (pH 8.2) was added into the centrifuge tube for 30 min in water bath at 25 °C, sequentially mixed with 1 mL of 25 mmol/L pyrogallol and 1 mL of sample solution, and reacted for 15 min in a water bath at 25 °C. The absorbance was measured at 320 nm. The scavenging activity was calculated as follows:

$$\mathrm{Scavenging\; ability} (\%)=[1-(A_S-A_C)/A_0]\times 100$$

(4)

where A_S is the absorbance value of the sample, A_C is the absorbance value of the control with physiological saline instead of Tris-HCl and pyrogallol, and A₀ is the absorbance of blank control with physiological saline instead of the sample supernatant.

2.12. Statistical Analysis

All trials were conducted in triplicate and averaged, and the results were presented as mean ± standard deviation. Statistical analyses and visualization of experimental data were performed using GraphPad Prism 9.0 software. Differences at p < 0.05 were considered statistically significant by a one-way analysis of variance (ANOVA), and response surface design and analysis were performed using Design-Expert software version 8.06. A significance level was established at p < 0.05.

3. Results

3.1. Optimization of the Culture Conditions of SMCM

The S. cerevisiae L23 count of SMCM was determined by plate counting as the optimization index to explore the effect of different culture media. The single-factor experimental results of SMCM culture conditions are shown in Figure 1. The S. cerevisiae L23 count of SMCM increased with the increase of the MgSO₄·7H₂O addition until 0.06% and then decreased. The S. cerevisiae L23 count was the highest at the MgSO₄·7H₂O addition of 0.06% (Figure 1A). The highest S. cerevisiae L23 count of SMCM was observed when fermentation was conducted at KH₂PO₄ addition of 0.2% (Figure 1B) and (NH₄)₂SO₄ addition of 0.65% (Figure 1C). Throughout the fermentation period of 24 h, S. cerevisiae L23 count was the highest at the pectinase addition of 0.1% (Figure 1D) and urea addition of 1.0% (Figure 1E). The above findings demonstrated that different culture media conditions of SMCM affected the S. cerevisiae L23 count, and the optimal MgSO₄·7H₂O addition, KH₂PO₄ addition, (NH₄)₂SO₄ addition, pectinase addition, and urea addition were 0.06%, 0.2%, 0.65%, 0.1% and 1.0%, respectively.

3.2. Optimization of SMCM Fermentation Process

The experimental results of SMCM fermentation with S. cerevisiae L23 as the fermentation strain are shown in Figure 2. The S. cerevisiae L23 count of SMCM increased with the increase of the inoculation amount until 6% and then decreased, which was the highest at the inoculation amount of 6% (Figure 2A). Throughout the fermentation period of 32 h, the S. cerevisiae L23 count of SMCM reached the maximum until 28 h and then decreased (Figure 2B). The highest S. cerevisiae L23 count of SMCM was observed when fermentation was conducted at 30 °C (Figure 2C). Furthermore, SMCM showed the highest S. cerevisiae L23 count when glucose was added at 0.5% (Figure 2D). All the experimental results indicated that the best fermentation process of SMCM was achieved when the optimal inoculation amount, fermentation time, fermentation temperature, and glucose addition of the SMCM were 6%, 28 h, 30 °C, and 0.5%, respectively.

3.3. Growth Curve of S. cerevisiae L23

Throughout the fermentation period of 96 h under the optimal culture conditions and fermentation process parameters of SMCM compared with the seed melon juice only as the control, the S. cerevisiae L23 count was determined at a certain time to plot the growth curves, which was obtained as shown in Figure 3. The S. cerevisiae L23 count of control increased rapidly with the fermentation time from 4 to 24 h, which was the highest at the fermentation time of 28 h, and declined after 28 h. Furthermore, the S. cerevisiae L23 count of SMCM increased rapidly with the fermentation time until 28 h and then slightly decreased. The S. cerevisiae L23 count was the highest at the fermentation time of 28 h.

3.4. Optimization of SMFF Fermentation Processing

3.4.1. The Single-Factor Experimental of SMFF Fermentation Processing

The single-factor test results of SMFF fermentation processing are shown in Figure 4. The SMFF was inoculated with substrate ratio (7:3), inoculation amount (5%), and fermentation time (24 h). The viable counts of S. cerevisiae L23 of SMFF significantly increased with the increase of the seed melon juice proportion, and 8.53 ± 0.05 lg CFU/g was reached after fermentation (Figure 4A). The sensory score of SMFF increased with the increase of the seed melon juice proportion as substrate ratio of 7:3 (Figure 4B) and then gradually decreased. Furthermore, the number of S. cerevisiae L23 of SMFF significantly rose to 8.72 ± 0.03 lg CFU/g as the optimal inoculation amount of 8% (Figure 4C). The highest sensory score of SMFF was observed when fermentation was conducted at inoculation amount of 8% (Figure 4D). The S. cerevisiae L23 count of SMFF was observed with a significant rise up to 10.35 ± 0.05 lg CFU/g with the increase of the fermentation time of 36 h (Figure 4E). Throughout the fermentation period of 60 h, the sensory score of SMFF reached the maximum until 36 h and then decreased (Figure 4F). The results demonstrated that the optimal substrate ratio, inoculation amount, and fermentation time were 7:3, 8%, and 36 h, respectively, which was beneficial to the growth of S. cerevisiae L23 of SMFF.

3.4.2. RSM Optimization of SMFF Fermentation Processing

On the basis of the single-factor experimental results, S. cerevisiae L23 count, and sensory score of SMFF as the optimization index, three independent variables, namely substrate ratio (A), inoculation amount (B), and fermentation time (C), significantly affected the S. cerevisiae L23 count and sensory score of SMFF, and their corresponding condition ranges were confirmed. The results of the Box–Behnken design consisted of 17 experimental points are shown in

Table 3. Experimental results for Box–Behnken design.

Runs	A Substrate Ratio (%)	B Inoculation Amount (%)	C Fermentation Time (h)	R1 Sensory Score	R2 S. cerevisiae L23 Count (lg(CFU/g))
1	−1.00	1.00	0.00	76	9.15
2	1.00	0.00	1.00	72	8.98
3	1.00	0.00	−1.00	68	8.56
4	0.00	1.00	1.00	74	8.79
5	0.00	1.00	−1.00	70	8.68
6	0.00	0.00	0.00	83	10.38
7	0.00	0.00	0.00	80	10.02
8	0.00	0.00	0.00	84	10.65
9	1.00	−1.00	0.00	69	8.46
10	−1.00	0.00	−1.00	73	9.02
11	1.00	1.00	0.00	70	8.67
12	0.00	−1.00	−1.00	68	8.49
13	0.00	−1.00	1.00	74	8.63
14	−1.00	−1.00	0.00	73	8.78
15	−1.00	0.00	1.00	77	9.18
16	0.00	0.00	0.00	83	10.38
17	0.00	0.00	0.00	84	10.53

. Then, two quadratic polynomial equations describing the relationship between these three variables and the S. cerevisiae L23 count were obtained through multiple regression fitting analysis of the data in Table 2 as follows:

$$R_1=+ 82.80-2.50\mathrm{A}+0.75\mathrm{B}+2.25\mathrm{C}-0.50\mathrm{AB}+0.000\mathrm{AC}-0.50\mathrm{BC}-4.90{\mathrm{A}}^2-5.90{\mathrm{B}}^2-5.40{\mathrm{C}}^2$$

$$R_2=+ 10.39-0.18\mathrm{A}+0.12\mathrm{B}+0.10\mathrm{C}-0.040\mathrm{AB}+0.065\mathrm{AC}-7.500\times {\mathrm{10}}^{\mathrm{-3}}\mathrm{BC}-0.67{\mathrm{A}}^2-0.96{\mathrm{B}}^2-0.79{\mathrm{C}}^2$$

where R₁ and R₂ refer to the sensory score and S. cerevisiae L23 count of SMFF, respectively, and A, B, and C refer to substrate ratio, inoculation amount, and fermentation time, respectively.

The ANOVA analysis of the developed model equations in

Table 4. ANOVA analysis of the sensory score quadratic models.

Variance Source	Sum of Squares	df	Mean Square	F-Value	p-Value	Significant
Sensory score
Modle	510.67	9	56.74	33.66	<0.0001	****
A	50.00	1	50.00	29.66	0.0010	**
B	4.50	1	4.50	2.67	0.1463
C	40.50	1	40.50	24.03	0.0017	**
AB	1.00	1	1.00	0.59	0.4664
AC	−5.684 × 10−14	1	−5.684 × 10−14	−3.372 × 10−14	1.0000
BC	1.00	1	1.00	0.59	0.4664
A2	101.09	1	101.09	59.97	0.0001	***
B2	146.57	1	146.57	86.95	< 0.0001	****
C2	122.78	1	122.78	72.83	< 0.0001	****
Residual	11.80	7	1.69
Lack of Fit	1.00	3	0.33	0.12	0.9415	not significant
Pure Error	10.80	4	2.70
Cor Total	522.47	16
S. cerevisiae L23 count
Model	9.79	9	1.09	30.85	< 0.0001	****
A	0.27	1	0.27	7.55	0.0286	*
B	0.11	1	0.11	3.06	0.1235
C	0.086	1	0.086	2.44	0.1622
AB	6.4 × 10−3	1	6.4 × 10−3	0.18	0.6830
AC	0.017	1	0.017	0.48	0.5112
BC	2.25 × 10−4	1	2.25 × 10−4	6.378 × 10−3	0.9386
A2	1.89	1	1.89	53.54	0.0002	***
B2	3.86	1	3.86	109.36	< 0.0001	****
C2	2.61	1	2.61	73.97	< 0.0001	****
Residual	0.25	7	0.035
Lack of Fit	0.023	3	7.558 × 10−3	0.13	0.9343	not significant
Pure Error	0.22	4	0.056
Cor Total	10.04	16

* Significant at 0.05 level, ** Significant at 0.01 level, *** Significant at < 0.001 level, **** Significant at < 0.0001 level.

showed that the two models were significant (p < 0.0001), which suggested that the regression equation fitted the experiment well. In addition, the coefficient of determination (R²) of sensory score and S. cerevisiae L23 count models were 0.9774 and 0.9754, respectively, indicating the goodness of fit of the SMFF regression models. The adjusted R² (R²_adj) of sensory score and S. cerevisiae L23 count models were 0.9484 and 0.9438, respectively, and the coefficients of variation (C.V.%) were 1.73 and 2.03, respectively, indicating that the predicted values were highly precise with a good degree of reliability, and the fermentation conditions of SMFF could be predicted by them. According to the F-value, the order of the effects of independent variables on the sensory score was A > C > B (substrate ratio > fermentation time > inoculation amount). The order of the effects of independent variables on the S. cerevisiae L23 count of SMFF was A > B > C (substrate ratio > inoculation amount > fermentation time). To sum up, this model can be used to analyze and predict the fermentation processing of SMFF.

The RSM can intuitively reflect the effects of substrate ratio, inoculation amount, and fermentation time on sensory score and S. cerevisiae L23 count of SMFF; the interactions can be observed from the contour shape of the projection surface. Circles suggested that there was no significant interaction between the two components, while ellipses indicated significant [24]. The three-dimensional response surface plots of independent variables on the sensory score and S. cerevisiae L23 count are shown in Figure 5. The substrate ratio (p = 0.001) and fermentation time (p = 0.0017) had a significant impact on the sensory score (model p < 0.0001). In addition, the substrate ratio (p = 0.0286) had a significant effect on S. cerevisiae L23 count (model p < 0.0001). Considering the feasibility of practical applications, the highest S. cerevisiae L23 count and best sensory were obtained under the following conditions: substrate ratio of 7:3, inoculation amount of 8%, and fermentation time of 36 h.

3.5. Verification Experiment

The validation experiment was conducted with the optimal fermentation conditions parameters (substrate ratio of 7:3, inoculation amount of 8%, and fermentation time of 36 h). The sensory score and S. cerevisiae L23 count of SMFF were measured to be 83.1 and 10.35, respectively. The theoretical predicted sensory score and S. cerevisiae L23 count of SMFF by the model were 83.4 and 10.4, respectively, and the good fit between the theoretical and experimental values confirmed the effectiveness of the model, indicating that the model can well reflect the actual fermentation situation of SMFF.

3.6. Determination of Fermentation Quality and Physiochemical Properties

The fermentation quality of SMFF control and experiments (test1, test2, and test3) is shown in

Table 5. The fermentation quality of SMFF.

Run	Variables			LA/%	AA/%	PA/%	BA/%
	Substrate Ratio (g/g)	Inoculation Amount (%)	Fermentation Time (h)
Control	7:3	-	24	0.43 ± 0.05 c	0.59 ± 0.04 d	0.27 ± 0.08 c	0.17 ± 0.04 a
Test 1	7:3	5	24	0.76 ± 0.18 b	0.72 ± 0.15 b	0.22 ± 0.07 d	0.07 ± 0.06 b
Test 2	7:3	8	24	0.78 ± 0.14 b	0.70 ± 0.16 c	0.31 ± 0.05 b	0.06 ± 0.07 b
Test 3	7:3	8	36	0.85 ± 0.12 a	0.75 ± 0.13 a	0.35 ± 0.09 a	0.06 ± 0.08 b

Same letters indicate no significant difference (p > 0.05), while different letters indicate significant differences (p < 0.05), the same applies below.

. The lactic acid (LA), acetic acid (AA), and propionic acid (PA) content of SMFF were increased by the addition of S. cerevisiae L23. As shown in Figure 6, the physiochemical properties of SMFF in different fermentation condition were different. After fermentation, the pH value in the SMFF at the different tests with S. cerevisiae L23 were 5.53 ± 0.21, 4.83 ± 0.23, 4.54 ± 0.25, and 4.18 ± 0.16, respectively, and SMFF with S. cerevisiae L23 inoculation showed the significant reduction in pH compared with SMFF control. This may be attributed to the elevated levels of lactic acid (LA), acetic acid (AA), and propionic acid (PA) generated during the fermentation process. These findings indicated that S. cerevisiae has the excellent acids-producing capacity and led to the higher acid level in SMFF. The same change trend in SMFF with or without S. cerevisiae L23 of reducing sugar content, NDF, ADF, Ca, and phosphorus content was observed. The NDF content in SMFF control presented a significant decrease, 23%. There were no significant differences in the moisture, crude ash, and DM content between the control of SMFF and the three tests in the absence or presence of S. cerevisiae L23. The SMFF with S. cerevisiae L23 significantly enhanced the crude protein (from 13.65 ± 0.52 to 15.76 ± 0.75) and VC content (from 0.99 ± 0.08 to 1.25 ± 0.25).

3.7. Determination of In Vitro Antioxidant Activity

In vitro antioxidant activities were evaluated in SMFF control or SMFF with S. cerevisiae L23. As shown in Figure 7, compared with SMFF control, SMFF with S. cerevisiae L23 inoculation significantly improved DPPH, ABTS, hydroxyl radical, and superoxide anion radical scavenging activities. The total in vitro antioxidant capacity of SMFF with S. cerevisiae L23 was notably higher than that observed in SMFF control.

4. Discussion

The optimization of seed melon compound fermented feed (SMFF) with S. cerevisiae L23 significantly enhanced nutritional quality, sensory attributes, and antioxidant capacity. These improvements aligned with established principles of yeast-mediated bioconversion, wherein metabolic activity (e.g., proteolysis and phenolic compound liberation) elevated feed value through bioactive compound generation and anti-nutritional factor degradation [23]. Critically, the integration of seed melon juice and ZSM as primary substrates represented an innovative approach for feed fermentation. Seed melon juice provided bioaccessible citrulline and phenolic antioxidants (e.g., flavonoids), while ZSM contributed phytochemically active alkylamides and unsaturated fatty acids [25]. Their co-fermentation synergistically utilized the agro-industrial waste seed melon peel/pulp and ZSM. This strategy directly aligned with circular bioeconomy principles by converting low-value biomass into high-value functional feeds, simultaneously addressing resource waste and environmental pollution. The process optimization of S. cerevisiae L23 enabled high-density fermentation in a low-cost medium SMCM [26].

The single-factor and RSM approaches revealed that a substrate ratio of 7:3, an inoculation amount of 8%, and a fermentation time of 36 h maximized S. cerevisiae L23 count (10.35 lg CFU/g) and sensory scores (83.1). The decline in microbial biomass beyond the optimal substrate ratio suggests a potential shift from carbon-limited to carbon-excess conditions. While the exact critical carbon threshold was not quantified in this study, the observed growth inhibition at higher seed melon juice proportions aligns with the classic microbial growth phenomenon where excessive substrate concentration can induce osmotic stress and catabolite repression, ultimately inhibiting yeast proliferation [27]. The decline in biomass at excessive concentrations of nitrogen sources like urea or salts (Figure 1) can be attributed to substrate inhibition or increased osmotic stress, which can disrupt cellular homeostasis and impede metabolic functions [13]. S. cerevisiae L23 biomass increased to an optimum point before declining, aligning well-established microbial growth principles (Figure 2C). The optimum at 30 °C represented the strain’s ideal balance between enzymatic reaction rates and protein stability. Temperatures above this optimum likely induce thermal stress, leading to the denaturation of critical enzymes and proteins, thereby reducing metabolic activity and growth yield [15,17]. The reduction in viability after the optimal fermentation time (Figure 2B) is typically caused by nutrient depletion and the subsequent accumulation of inhibitory metabolites, such as ethanol and organic acids, which can suppress further microbial growth in batch cultures [18]. The excess carbon sources promote microbial biomass but may inhibit growth beyond optimal ratios due to osmotic stress, and subsequent declines in S. cerevisiae L23 counts after 36 h likely reflected nutrient depletion or accumulation of inhibitory metabolites. These results were slightly different from the previously reported research. It mainly is related to the fermentation strain, fermentation substrates, and the determining criteria of fermentation feed [8].

The high viability of S. cerevisiae L23 (10.35 lg CFU/g) achieved in the SMFF under optimized conditions compares favorably with previous studies on fermenting agro-industrial wastes, which reported a viable bacterial count of approximately 7 lg CFU/g after 30 d of fermentation [8]. The significantly higher cell density obtained in our study in a shorter time (36 h) underscores the efficiency of our two-stage optimization strategy and the high fermentability of the seed melon–ZSM substrate mixture for supporting yeast proliferation. The high-quality feed fermentation critically depends on three core parameters: suitable moisture content, soluble carbohydrate concentration, and strict fermentation conditions. The moisture content of SMFF with S. cerevisiae L23 was 65–75%, which was a suitable moisture content. The combination of ZSM and seed melon juice to produce mixed fermented feed would be a new way to improve the fermentation quality of feed and the rumen environment.

The butyric acid concentration served as a critical indicator of feed spoilage severity, with elevated levels correlating directly with compromised nutritional quality and aerobic stability. In this study, all the tests exhibited significantly lower butyric acid content (65% reduction) compared to control, which aligned with the study of Wang et al. [28]. This reduction was a key indicator of successful fermentation dominated by S. cerevisiae L23. Butyric acid was primarily produced by clostridial bacteria during spoilage, often under conditions of inadequate acidification or the presence of obligate anaerobes. The vigorous growth and metabolic activity of S. cerevisiae L23, leading to rapid production of lactic and acetic acids, caused a swift and significant drop in pH to 4.2 ± 0.1. This acidic environment was profoundly inhibitory to clostridial bacteria, thereby effectively suppressing butyric acid formation [29,30]. Therefore, the achievement of high S. cerevisiae biomass was not contradictory to low butyric acid levels. The high yeast biomass created conditions that prevented the proliferation of spoilage microorganisms, ensuring fermentation quality. Mechanistically, when fermentation pH declined below 4.5, a threshold achieved rapidly via S. cerevisiae-driven organic acid production (lactic, acetic, and propionic) and clostridial metabolism was inhibited, reducing proteolysis [29]. S. cerevisiae L23 mediated lactate/acetate accumulation and lowered pH to 4.2 ± 0.1 within 24 h, which enhanced the SMFF preservation and inhibited pathogens. The results of this study showed that the content of crude protein and VC in SMFF with S. cerevisiae L23 was significantly increased, indicating that S. cerevisiae L23 fermentation can improve the nutritional composition of the feed [27,30]. The increase in crude protein content (from 13.65% to 15.76%) observed in the SMFF is a critical indicator of successful microbial bioconversion. This improvement aligned with the findings of Sharawy et al., which reported a 44% increase in protein content in soybean meal fermented with Saccharomyces cerevisiae, attributing it to microbial biomass production [30]. The significant increase in crude protein content in SMFF was maybe attributed to the substantial accumulation of S. cerevisiae L23 microbial biomass, which is rich in protein. The Kjeldahl method used for determination measures total nitrogen, encompassing both intracellular microbial protein and extracellular protein, though the former constitutes the primary contributor to the observed increase [19,30]. Furthermore, the reduction in NDF content (by 23%) in our study suggests enhanced fiber degradation, which is comparable to the results reported by Wang et al. using biological pretreatment of corn stover [28]. These parallel outcomes confirmed the role of S. cerevisiae in improving the nutritional accessibility of fibrous by-products. The increase in the content of these substances may be related to the secretion of amino acids and various bioactive substances by S. cerevisiae L23 fermentation [31], but the specific mechanism still needs further research.

The remarkable enhancement in the in vitro antioxidant activity of SMFF post-fermentation is a key finding of this work. The recorded DPPH and ABTS radical scavenging activities were substantially higher than those reported in unfermented seed melon juice [10]. This boost in antioxidant capacity can be attributed to the release of bioactive phenolic compounds from the seed melon and ZSM matrices through the enzymatic activity of S. cerevisiae L23; this phenomenon was also observed in the fermentation of other plant substrates [31]. The results observed underscore the fermentation potential to amplify the value of agricultural waste streams. High sensory scores for SMFF reflected desirable color, flavor, and texture, attributable to reaction products formed during fermentation. The “pleasant sourness” noted by panelists aligned with optimal organic acid levels and the degradation of pungent compounds in ZSM with a certain aroma and taste [32]. This process valorized agricultural by-products into stable, nutrient-dense, feed-addressing waste and livestock nutrition challenges, while RSM model analysis of SMFF fermentation and real-time monitoring (e.g., transcriptomics) were needed to elucidate interactions between S. cerevisiae L23 and native microbiota. Quantification of active compounds in ZSM and their degradation by S. cerevisiae L23 enzymes remained unaddressed. Furthermore, the validation of animal performance was essential, as microbial dynamics may shift in industrial production. Future research would evaluate in vivo impacts on livestock growth and gut health, exploring synergistic effects of co-fermentation with lactic acid bacteria and assessing economic feasibility and analysis for large-scale fermentation process.

5. Conclusions

Co-fermentation of seed melon juice and Z. bungeanum seed meal with S. cerevisiae L23 significantly improved SMFF fermentation quality, achieving high yeast viability (10.35 lg CFU/g), sensory attributes (score: 83.1), and enhanced nutritional profiles. The optimal process reduced pH to 4.18, elevated organic acid production, and decreased NDF/ADF content. SMFF demonstrated superior in vitro antioxidant activities across all tested radicals. These results indicated that S. cerevisiae L23 fermentation effectively converts agricultural by-products (seed melon juice and Z. bungeanum seed meal) into value-added functional feed, mitigating resource waste while enhancing preservation and bioactive compound availability. Future study should validate in vivo efficacy and scale-up feasibility.