A Potential of Agro-Industrial Biowaste as Low-Cost Substrates for Carotenoid Production by Rhodotorula mucilaginosa

Abstract

The sustainable production of natural pigments is gaining attention as industries seek alternatives to synthetic additives. This study explored agro-industrial biowastes as feedstocks for carotenoid biosynthesis by Rhodotorula mucilaginosa (natural isolate from Jerusalem artichoke), aiming to identify an optimal substrate that combines high productivity with economic and environmental feasibility. Thirteen biowastes, including grape pomace, crude glycerol, chicken feathers, sugar beet juice, and pea protein isolate, were systematically evaluated for their impact on yeast growth and pigment accumulation. Carotenoid yields ranged from 21.4 to 187.2 mg/100 g dry weight, with the highest volumetric productivity achieved in pea protein isolate (14.98 mg/L), untreated white grape pomace (14.09 mg/L), and crude glycerol (13.87 mg/L). To assess scalability, a simplified techno-economic and sustainability analysis was applied, revealing that although pea protein isolate offered the best yields, its high market cost limited industrial feasibility. In contrast, untreated grape pomace and crude glycerol emerged as low-cost, abundant alternatives with strong circular bioeconomy potential. Fed-batch bioreactor validation using untreated grape pomace confirmed its suitability, achieving a 43% improvement in carotenoid productivity (20.1 mg/L) compared to shake-flask trials. These results position untreated grape pomace as the optimal substrate–strategy combination for sustainable carotenoid production linking agro-waste valorization with high-value bioproduct generation. This study provides both experimental evidence and economic rationale for integrating winery residues into industrial pigment production chains, advancing yeast biotechnology toward more circular and resource-efficient models.

1. Introduction

Every year vast amounts of waste are generated by various food and agricultural industries, posing a serious environmental challenge. Improper waste disposal poses a serious threat to both human health and the environment, as it contaminates the air, water and soil [1,2]. On the other hand, certain processing industry effluents could represent valuable sources of bioactive compounds and have rich nutritional profiles making them suitable for animal feed or as growth medium for microorganisms and thus excellent substrates for various bioprocesses [3]. In addition to the use of waste as raw material for production, biobased products (products obtained through biotechnological processes) have great potential for application across various industries and target markets. These products are already widely used commercially and successfully replace those derived through chemical processes, particularly given their significantly lower content of environmentally harmful substances and greater ecological acceptability [4,5].

Carotenoids are natural pigments in shades of yellow, orange, and red, responsible for the coloration of many fruits, vegetables, flowers and leaves. They can also be found in certain animals, such as insects, birds and fish, as well as in various microorganisms. While plants, bacteria, fungi and algae are capable of synthesizing carotenoids, humans and animals must obtain them through their diet [6]. Carotenoids belong to a class of terpenoid pigmented compounds that are non-polar, hydrophobic and lipid-soluble and, as such, are important biomolecules for human health. In addition to being used as coloring agents, they have anticancer and antioxidant properties, as well as serving as provitamin A compounds. Their roles have already been established in phytomedicine, chemistry, pharmaceuticals, cosmetics, the food industry and the animal feed industry, and the demand for their production is growing annually [7]. Carotenoids are present in a wide variety of plants and microorganisms and most of them are currently extracted either from plants such as annatto, tomato, carrot, pepper and grapes, which involve seasonal and geographical variability, or are chemically synthesized, which results in the accumulation of hazardous by-products as waste [8]. Demand for carotenoid production continues to rise, with their primary use as food colorants, and they are also widely utilized as dietary supplements [9].

In recent years, interest in synthetically produced pigments has declined due to their toxic, carcinogenic and teratogenic effects, leading to a growing shift toward biotechnological production as a safer and more sustainable alternative. Various types of algae, fungi, yeasts and bacteria are already being used for commercial pigment production, while some microorganisms remain under investigation [6]. Red yeasts are named for the distinctive orange, pink or red coloration of their colonies, which results from their high production of carotenoid pigments and intracellular lipids. Thanks to their remarkable ability to convert carbon sources into a wide range of primary and secondary metabolites, red yeasts hold significant industrial potential, particularly in the food, pharmaceutical, cosmetic and animal feed sectors [10,11]. Most yeasts belonging to the genus Rhodotorula are mesophilic and aerobic organisms, although some strains can thrive at lower temperatures as well. Their cells exhibit spherical, ellipsoidal or elongated shapes. These yeasts possess the ability to utilize a wide variety of compounds as carbon sources, including glucose, galactose, sucrose, maltose, trehalose, ethanol, glycerol, etc. [11]. Among them, Rhodotorula mucilaginosa is an ubiquitous species, widely distributed in nature and it can even be found in polluted environments and within the rhizosphere of plants. Most strains of this species are considered safe and non-pathogenic [8].

To reduce the number of waste streams, carotenoid biosynthesis can be carried out using raw materials and agro-industrial waste, given that red yeasts exhibit tolerance to inhibitors that may arise as by-products of agro-industrial processes [12]. In recent years, R. mucilaginosa has been actively studied as a promising production microorganism within the red yeast group. It is a natural producer of carotenoids, lipids and enzymes, and offers several advantages over bacteria, microalgae and plants in carotenoid production. These advantages include its unicellular form, rapid growth rate, ability to grow on various substrates, tolerance to inhibitory compounds present in those substrates and suitability for cultivation in large-scale bioreactors [8,13]. In terms of nutritional requirements, Rhodotorula species do not require growth on media with a specialized composition. Due to the previously mentioned need to reduce the release of waste streams into the environment, there is growing interest in the use of natural raw materials and agro-industrial waste as growth substrates. The use of natural substrates such as grape pomace, sugarcane and sugar beet molasses, whey, and similar materials could contribute to reducing environmental pollution and the energy consumption associated with their disposal [14].

The aim of this study is to investigate the use of agro-industrial biowaste as a substrate for the production of biomass and carotenoids using the best-performing yeast isolate R. mucilaginosa. This research can contribute to ongoing studies focused on the identification, quantification, and optimization of microbial carotenoid production. The use of low-cost starting substrates for the biotechnological production of carotenoids would contribute to defining an economically and environmentally sustainable solution for carotenoid production.

2. Materials and Methods

2.1. Collection of Secondary Raw Materials

Between May 2022 and February 2023, 13 types of agro-industrial secondary raw materials (biowaste) were collected from food industry producers in the Autonomous Province of Vojvodina (APV), Serbia. Selection criteria included regional availability, generation as by-products in existing industrial processes, and potential for yeast cultivation. The substrates included in the research are presented in

Table 1. Biowastes with the used pretreatments and additions for the preparation of an appropriate cultivation medium.

Biowaste	Physical Pretreatment 1	Chemical Pretreatment	C:N Ratio (Before Supplementation)	Additional C-Source (g/L) ***	Additional N-Source (g/L) ***
Crude glycerol	Filtration, dilution	Sterilization *	~45:1	—	3
Treated 2 white grape pomace	Milling, decantation, filtration	pH adjustment	~30:1	15	2
Untreated 3 white grape pomace	Milling, decantation, filtration	pH adjustment	~28:1	15	2
Treated 2 red grape pomace	Milling, decantation, filtration	pH adjustment	~27:1	15	2
Untreated 3 red grape pomace	Milling, decantation, filtration	pH adjustment	~26:1	15	2
Pea protein isolate	Milling	Sterilization **	~7:1	10	-
Sugar beet juice	Filtration	Sterilization *	~40:1	—	3
Whey	Filtration	Sterilization *	~18:1	15	2
Molasses	Dilution, filtration	Sterilization *	~38:1	—	3
Hydrolyzed corn waste flour	Milling	Sterilization **	~25:1	15	2
Chicken feathers	Milling	Alkaline hydrolysis 4, neutralization	~5:1	20	-
Potato peels	Milling	Sterilization *	~22:1	15	2
Sweet potato peels	Milling	Sterilization *	~23:1	15	2

Note: ¹ in the case of multiple methods, they are shown in the order in which they were applied; ² subjected to heat treatment (pasteurization at 80–85 °C for 15–20 min) and washing with distilled water prior to storage; ³ fresh solid residue obtained immediately after grape pressing during winemaking, consisting of skins, seeds, and small amounts of pulp, without any further processing; ⁴ chicken feathers required alkaline hydrolysis to break down keratin proteins into peptides and amino acids, enabling better bioavailability of nitrogen; DAP-(NH₄)₂HPO₄; *—autoclave (at 121 °C for 15 min); **—dry heat (at 160 °C for 2 h); ***—per liter of prepared cultivation medium.

.

All samples were collected in hygienically acceptable conditions at the production facilities and transported in cold chain containers to the Laboratory of Microbiology, Faculty of Technology, University of Novi Sad. Each raw material underwent pre-treatment tailored to its physical and chemical properties (Table 1). The choice of pretreatment methods for each type of biowaste was based on its physical properties (solid, semi-solid, or liquid) and microbiological safety requirements for cultivation media preparation.

Physicochemical characterization of each substrate was performed to determine carbon (C) and nitrogen (N) content. Total carbon content was determined using the Walkley–Black dichromate oxidation method [15], which is widely applied for organic material analysis in food and agricultural samples. Nitrogen content was determined using the Kjeldahl method (AOAC Official Method 984.13 [16]). When C-sources were insufficient, glucose (Sigma-Aldrich, Burlington, MA, USA, ≥99% purity) was added. For N-source supplementation, diammonium phosphate ((NH₄)₂HPO₄) was incorporated. pH was adjusted to 6.5 ± 0.1 prior to inoculation using sterile NaOH (1 M) or HCl (1 M) solutions. The decision to add supplementary nutrients was made following elemental composition analysis and literature-based C:N ratio optimization for yeast growth [17,18]. Briefly, glucose was added in the range 10–20 g/L to reach an optimal C:N ratio of ~15:1 in the medium, while DAP additions were calculated to supply ~0.5 g/L nitrogen, which supports yeast pigment production without triggering excessive biomass at the expense of carotenoids [19]. These pretreatment and supplementation strategies aimed to maximize biomass yield and pigment production, while also ensuring the microbiological safety of the cultivation process.

2.2. Yeast and Inoculum Preparation

The red yeast Rhodotorula mucilaginosa (natural isolate from Jerusalem artichoke, marked as “top 30” in the Collection of microorganisms at Faculty of Technology Novi Sad) was used as the test organism because it was the best-performing strain for carotenoid production during the prior comprehensive selection [20]. Initial inoculation (first passage) was performed on YPD agar medium (1% yeast extract, 2% peptone, 2% glucose, 2% agar, pH 6.5). After 48 h incubation at 30 °C, cells were harvested, washed with sterile distilled water, and resuspended in sterile saline to an optical density corresponding to 4 log CFU/mL. This suspension served as the inoculum for cultivation experiments.

2.3. Cultivation Conditions

All cultivation experiments were conducted in 250 mL Erlenmeyer flasks containing 100 mL of prepared medium (in triplicate). Incubation conditions included 28 °C, 150 rpm shaking (orbital incubator, New Brunswick Scientific, Edison, NJ, USA), in the dark for 168 h for biomass evaluation and carotenoid yield analysis. Control culture was maintained on commercial YPD medium under identical conditions. At the end of fermentation, biomass was centrifuged at 6000 rpm for 10 min at 4 °C (ROTINA 380R, Hettich, Tuttlingen, Germany). The supernatant was discarded, and the wet biomass was weighed. Visual assessment of pigmentation was recorded, and non-pigmented biomass samples were excluded from carotenoid analysis.

2.4. Carotenoid Extraction

After harvesting, yeast biomass was frozen at −20 °C and subsequently lyophilized in a Martin Christ Alpha 1–4 LSC (Osterode, Germany) freeze-dryer under the following conditions: −40 °C condenser temperature, 0.01 mbar vacuum, and a total drying time of 28 h. The endpoint was defined as a constant dry weight achieved in three consecutive measurements. As noted by Saini and Keum [21], this step is crucial for removing the high water content naturally present in yeast cells. Since water can hinder the efficient extraction of carotenoids due to their hydrophobic nature, dehydration enhances extraction efficiency. Moreover, lyophilization minimizes the risk of thermal degradation and isomerization that may occur in other drying methods, making it the most suitable pretreatment for preserving carotenoids. Based on an optimized protocol for maximizing carotenoid extraction [20], the dried biomass was subjected to mechanical milling (1 min, 5000 rpm) using a batch mill tube mill control system (LLG, Meckenheim, Germany) to cause cell lysis prior to carotenoid extraction. Carotenoids were then extracted using a conventional extraction (CE) approach with acetone as the solvent. The solvent was added to the biomass at a 1:2.5 (m/V) ratio, and the mixture was vortexed for 2 min. The extraction cycle was repeated 2–3 times until the biomass pellet became colorless, with samples being centrifuged at 8000 rpm for 5 min between repetitions and the supernatant being collected each time. The combined acetone extracts were pooled for subsequent carotenoid quantification.

2.5. Quantification of Carotenoids and Yeast Biomass Concentration

Carotenoid content was quantified spectrophotometrically using a microtiter plate format. Aliquots (200 µL) of carotenoid extract were transferred into individual wells, and absorbance was measured at 663, 645, 505, and 453 nm. Carotenoid concentration (mg/100 g dry weight) was calculated using Equation (1), and all measurements were conducted in triplicate.

Carotenoid yield = 0.216·A₆₆₃ − 1.22·A₆₄₅ − 0.304·A₅₀₅ + 0.452·A₄₅₃

(1)

For yeast biomass, in which carotenoids are expected to be dominated by torulene, torularhodin, and β-carotene, matrix effects may differ significantly from those of plant tissues. Therefore, absorbance at 453 nm was used as the primary basis for carotenoid quantification, in line with established protocols for yeast acetone extracts [17]. Comparison between the two methods showed only minor differences (<2%), and therefore Equation (1) are used in this study.

To ensure consistency in productivity metrics, yeast biomass was quantified using both viable cell counts (CFU/mL) and gravimetric dry weight (g/L). For dry weight determination, an aliquot (10 mL) of fermentation broth was centrifuged at 6000 rpm for 10 min at 4 °C, and the biomass pellet was washed, freeze-dried, and weighed to obtain biomass concentration in g/L. Parallel viable counts (log CFU/mL) were conducted as described using serial dilution and YPD agar plating. A calibration curve correlating log CFU/mL and dry biomass (g/L) was generated, with a high correlation coefficient (R² > 0.98), enabling conversion of CFU data to gravimetric biomass values. Carotenoid productivity was calculated as the product of biomass concentration and carotenoid content, using Equation (2).

Productivity (mg/L) = Biomass (g DW/L) × Carotenoid yield (mg/g DW)

(2)

This method ensures accurate cross-comparison of microbial growth and pigment production across all tested substrates.

2.6. Bioreactor Validation of Carotenoid Production on Best-Performing Substrate

To validate carotenoid production under controlled conditions, fed-batch fermentation was performed in a 2 L stirred-tank bioreactor (Biostat B-Plus, Sartorius, Göttingen, Germany) using untreated white grape pomace, the optimal substrate identified in shake-flask trials. The initial medium composition matched that used in flask experiments, with supplementation adjusted to achieve a working C:N ratio of ~15:1. Bioreactor operating conditions were: temperature 28 °C, pH 6.5 ± 0.1 (controlled automatically with 2 M NaOH or HCl), aeration 1 vvm, and agitation 300–600 rpm in cascade mode to maintain dissolved oxygen > 20% saturation. Antifoam (polypropylene glycol, 0.01%, w/v) was added as required. The reactor was inoculated with 5% (v/v) seed culture of R. mucilaginosa (OD600 ~0.5). After 48 h, a fed-batch regime was initiated by pulsed addition of sterile glucose (10 g/L) and diammonium phosphate (0.5 g/L N) every 24 h until 120 h, to prolong exponential growth and enhance carotenoid biosynthesis. Biomass (g/L dry weight), viable cell counts (CFU/mL), residual sugars, and carotenoid concentration (mg/L) were monitored every 24 h up to 168 h. Carotenoids were extracted and quantified as described in Section 2.4 and Section 2.5. Online sensors recorded dissolved oxygen, pH, agitation, and gas exchange (O₂ uptake and CO₂ evolution). Fed-batch productivity (mg/L) was compared with shake-flask data to evaluate the effect of controlled aeration and nutrient feeding on carotenoid yield.

2.7. Statistical Analysis

The sustainability potential of each biowaste substrate for carotenoid production was assessed using a standardized scoring framework adapted from microbial substrate evaluation approaches [22]. The scoring system considered both the intrinsic nutrient profile of the substrate and the degree of supplementation required to achieve optimal fermentation conditions. Three parameters were used for evaluation: (i) the C:N ratio prior to supplementation, (ii) the amount of additional carbon source (g/L) required to reach the optimal C:N ratio, and (iii) the amount of additional nitrogen source (g/L) required. For yeast-based carotenoid biosynthesis, an optimal C:N ratio of ~15:1 was assumed, consistent with previous studies on Rhodotorula spp. growth and pigment production [17,18]. A C:N suitability score was first calculated as the negative absolute deviation from the optimal ratio:

C:N suitability = −∣C:N ratio − 15∣

(3)

This scoring favors substrates whose native C:N ratio is closer to the optimal value.

This favors substrates whose native C:N ratio is closer to the optimal range. Since high supplementation demand reduces both environmental and economic sustainability, the amounts of added carbon and nitrogen were treated as negative indicators. To integrate different variables into a single metric, all parameters were standardized using z-score normalization:

$$\mathrm{z} = {\displaystyle \frac{x-\overline{x}}{\sigma }}$$

(4)

where $x$ is the measured value, $\overline{x}$ is the mean, and σ is the standard deviation for the given parameter. The final sustainability score for each substrate was calculated as

Sustainability score = z_{C:N suitability} − z_{Additional C} − z_{Additional N}

(5)

where z_{C:N suitability} is the standardized C:N suitability score, z_{Additional C} and z_{Additional N} are the standardized nutrient addition amounts (negative sign applied so that lower supplementation yields higher scores). Substrates were ranked from highest to lowest total sustainability score, with higher scores indicating substrates that are closer to the optimal C:N ratio and require minimal nutrient supplementation.

The biomass concentration obtained during fermentation was described using a four-parameter sigmoidal Equation (6), which captures the kinetics of yeast growth on different biowaste substrates.

$$y(t)=d+{\displaystyle \frac{a-d}{1+{\left(\frac{t}{c}\right)}^b}}$$

(6)

The model was expressed through y(t) is the yeast biomass concentration (log CFU/mL) at incubation time t (h), a denotes the minimum experimentally observed value (initial biomass concentration at t = 0), b is the Hill’s slope, describing the steepness of the curve at the inflection point, c corresponds to the inflection point, the time at which half of the maximum biomass is reached, and d represents the maximum asymptotic biomass concentration achieved at the stationary phase [23]. Nonlinear regression was performed to fit the experimental biomass data (0–168 h) to the sigmoidal function using the least-squares method. Model fitting was evaluated using multiple statistical error indices, including the coefficient of determination (R²), root mean square error (RMSE), chi-square (χ²), mean bias error (MBE), mean percentage error (MPE), and skewness. These metrics were calculated following standard approaches for assessing the accuracy and bias of predictive microbial growth models [24,25,26]. To further validate model performance, predicted values were compared with experimental biomass measurements, and the fitted growth curves were plotted against experimental data for visual inspection of goodness-of-fit [26].

In order to evaluate differences in yeast biomass yield (g/L) and carotenoid yield (mg/100 g dry biomass) across the tested biowaste substrates, a one-way analysis of variance (ANOVA) was performed. The independent factor was the type of biowaste substrate, while the dependent variables were the biomass yield and carotenoid yield, respectively. The ANOVA tested the null hypothesis (H₀) that there were no significant differences in mean yields among biowaste groups, against the alternative hypothesis (H₁) that at least one group differed. Prior to analysis, assumptions of normality (Shapiro–Wilk test) and homogeneity of variance (Levene’s test) were verified. Where ANOVA indicated statistically significant differences (p < 0.05), post hoc pairwise comparisons were conducted using Tukey’s Honestly Significant Difference (HSD) test. This test controls the family-wise error rate and allows grouping of substrates into statistically homogeneous subsets based on their mean yields. All statistical analyses were conducted in Python (Statsmodels version 0.14, SciPy v1.11) and verified with Statistica v13. The results are presented with groupings indicated by different superscript letters.

3. Results

Figure 1 illustrates the sustainability scores obtained for various biowaste substrates tested in the cultivation of R. mucilaginosa for carotenoid production. Among the evaluated materials, pea protein isolate demonstrated the most favorable outcome, achieving a score of 2.93. Chicken feathers (1.23) and whey (0.75) followed, both characterized by C:N ratios close to the optimal value of approximately 15:1, which minimized the need for additional supplementation with external carbon or nitrogen. Potato peels (0.24) and orange sweet potato peels (0.11) also performed relatively well, indicating above-average sustainability. In contrast, crude glycerol (−1.56), sugar beet juice (−0.92), and treated white grape pomace (−0.78) yielded the lowest scores. Their poor performance was largely attributed to highly imbalanced initial C:N ratios, requiring substantial nitrogen adjustment. Substrates such as grape pomace, molasses, and hydrolyzed corn waste flour occupied an intermediate position in terms of sustainability.

Following the first step in experimental procedure, i.e., cultivations in flasks, three substrates, whey, potato peels, and molasses, were excluded from subsequent experiments because no yeast biomass was detected at the end of fermentation. As a result, these substrates also failed to support the development of the desired pigmentation. This outcome was reproducible across replicates, confirming the lack of biomass formation. In contrast, the remaining substrates supported yeast growth, and the biomass accumulation kinetics of Rhodotorula were well described by the four-parameter sigmoidal model [20]. The model captured the characteristic S-shaped growth profile, which included an initial lag phase, a rapid exponential increase, and a final plateau corresponding to the maximum biomass concentration (Figure 2).

The parameters of the kinetic model (minimum biomass (a), maximum biomass (d), inflection point (c), and Hill’s slope (b)) are presented in

Table 2. Kinetic models parameters and verification of fitting of models.

Biowaste	Kinetic Parameters				Verification of Kinetic Models
	a	d	c	b	R2	RMSE	χ2	MBE	MPE	Skew.
Crude glycerol	4.06	8.35	55.52	5.30	0.9996	0.028	0.362	0.458	5.703	0.572
Treated white grape pomace	3.26	3.94	99.00	10.00	0.941	0.077	0.259	−0.274	−8.373	−0.400
Untreated white grape pomace	4.68	10.00	113.22	1.48	0.991	0.125	0.090	0.105	1.946	0.235
Treated red grape pomace	3.86	7.15	95.52	4.09	0.996	0.079	0.078	0.058	0.520	−0.669
Untreated red grape pomace	4.15	5.37	39.73	5.32	0.996	0.036	0.050	0.122	2.176	0.452
Pea protein isolate	4.99	7.84	37.87	3.21	0.987	0.138	0.216	0.353	5.037	1.038
Sugar beet juice	4.30	10.00	200 *	1.09	0.947	0.226	0.183	0.283	4.964	−0.351
Hydrolyzed corn waste flour	3.57	4.18	92.21	4.84	0.984	0.031	0.171	−0.29	−7.79	−0.279
Chicken feathers	4.12	6.07	114.72	8.60	0.991	0.073	0.160	−0.108	−3.366	−0.043
Sweet potato peels	4.31	6.62	58.83	1.34	0.935	0.185	0.140	0.284	5.135	−0.087

Note: R²—coefficient of determination; RMSE—root mean square error; χ²—reduced chi-square; MBE—mean bias error; MPE—mean percentage error; Skew.—skewedness; * for sugar beet juice, the model estimated the inflection point (c) beyond the experimental timeframe (200 h); this extrapolation reflects slow growth and required additional analysis.

together with the experimental values, predicted time-point values, and model performance statistics (R², RMSE, χ², MBE, MPE, Skewness). Among the examined substrates, crude glycerol, untreated white grape pomace, and pea protein isolate yielded the highest maximum biomass concentrations, surpassing 8 log CFU/mL after 168 h. These substrates also demonstrated excellent agreement with the model, as reflected by R² values greater than 0.98. In contrast, treated white grape pomace and sugar beet juice supported only limited biomass growth, plateauing at approximately 3–4 log CFU/mL. This outcome was consistent with their lower d values and weaker model performance (R² < 0.95). Considerable variation was observed in the estimated inflection point, ranging from an early acceleration phase in pea protein isolate (c ≈ 48–60 h) to a delayed exponential onset in chicken feathers (c ≈ 120 h). Such variation highlights the substrate-specific adaptation patterns and metabolic utilization dynamics of R. mucilaginosa.

Error analysis of the four-parameter sigmoidal model applied to yeast biomass growth demonstrated low χ² values for most biowaste substrates, confirming strong agreement between experimental and predicted data (Table 2). The best fits were obtained for untreated red grape pomace (χ² = 0.050), treated red grape pomace (0.078), and untreated white grape pomace (0.090), all of which exhibited excellent predictive accuracy. By contrast, crude glycerol produced the highest χ² (0.362), reflecting greater deviation from experimental results. Mean bias error (MBE) values suggested minimal systematic bias, with most ranging from −0.27 to 0.46 log CFU/mL. For treated white grape pomace, a negative MBE (−0.274) indicated a slight underestimation of biomass, whereas crude glycerol displayed the strongest positive bias (0.458). Mean percentage errors (MPE) varied between −8.37% and 5.70%, confirming that overall deviations were small and within the limits of biological variability. Skewness values remained close to zero for most cases, indicating symmetric error distribution and the absence of systematic under- or over-prediction. Only crude glycerol (0.572) and untreated red grape pomace (0.452) showed moderately higher skewness, pointing to minor underestimation tendencies at specific time points.

The central outcome of this study was the assessment of carotenoid productivity, calculated as the product of biomass yield (g/L) and carotenoid yield (mg/g DW) using the optimized extraction protocol defined by Šovljanski et al. [20]. As summarized in

Table 3. Carotenoid productivity of yeast cultivated on biowaste substrates.

Rank	Biowaste	Biomass (log CFU/mL)	Dry Biomass (g/L)	Carotenoid Yield (mg/g DW)	Productivity (mg/L)
1	Pea protein isolate	8.00 ± 0.12 a	4.5 ± 0.20 a	3.33 ± 0.05 a	14.98 ± 0.42 a
2	Untreated white grape pomace	8.26 ± 0.10 a	4.6 ± 0.15 a	3.06 ± 0.04 ab	14.09 ± 0.38 a
3	Crude glycerol	8.30 ± 0.11 a	4.7 ± 0.18 a	2.95 ± 0.03 ab	13.87 ± 0.36 a
4	Chicken feathers	6.00 ± 0.09 b	3.0 ± 0.12 b	2.84 ± 0.02 b	8.53 ± 0.28 b
5	Sugar beet juice	7.15 ± 0.13 ab	3.3 ± 0.14 b	2.31 ± 0.02 d	7.65 ± 0.26 bc
6	Sweet potato peels	6.00 ± 0.08 b	2.2 ± 0.10 c	2.54 ± 0.02 bc	5.59 ± 0.19 cd
7	Hydrolyzed corn waste flour	3.60 ± 0.07 c	1.9 ± 0.08 c	1.67 ± 0.02 e	3.17 ± 0.12 de
8	Treated red grape pomace	6.85 ± 0.10 b	1.6 ± 0.07 cd	2.44 ± 0.0 c	3.91 ± 0.14 de
9	Untreated red grape pomace	5.32 ± 0.09 c	1.2 ± 0.05 d	2.125 ± 0.01 f	2.55 ± 0.10 e
10	Treated white grape pomace	3.30 ± 0.06 d	0.4 ± 0.02 e	1.775 ± 0.01 f	0.71 ± 0.05 f

Note: Statistical groupings (a–f letters) reflect Tukey’s HSD test (p < 0.05).

, carotenoid productivity varied substantially among substrates. Pea protein isolate (~4.5 g/L), untreated white grape pomace (~4.6 g/L), and crude glycerol (~4.7 g/L) supported the highest dry biomass concentrations, corresponding to microbial densities above 8 log CFU/mL. These same substrates also produced the largest carotenoid yields, exceeding 167 mg/100 g DW. When integrated into the productivity formula, they generated between 13.87 and 14.98 mg/L of carotenoids after 168 h of fermentation. Statistical analysis by one-way ANOVA confirmed that productivity differed significantly across substrates (F = 1.18 × 10⁵, p < 0.0001). Post hoc testing with Tukey’s HSD further revealed that pea protein isolate, untreated white grape pomace, and crude glycerol formed a statistically homogeneous group (p > 0.05), while all other substrates exhibited significantly lower carotenoid outputs (p < 0.05). Collectively, these findings emphasize the critical role of substrate composition and C:N balance in regulating both biomass accumulation and carotenoid biosynthesis, particularly when evaluated in terms of volumetric productivity (mg/L).

Hierarchical clustering (Figure 3) further highlighted the variation in carotenoid productivity (mg/L) among the tested biowastes, with distinct grouping patterns that delineated clear performance tiers. Results from the one-way ANOVA combined with Tukey’s HSD test confirmed that pea protein isolate, untreated white grape pomace, and crude glycerol formed a statistically homogeneous cluster (group “a”), each achieving productivity values above 13.8 mg/L.

At the opposite end, treated white grape pomace represented the lowest-performing substrate (group “f”), with carotenoid productivity more than twenty-fold lower than that of the leading group. The vertical dendrogram identified three principal clusters of substrates. Cluster 1 comprised the high-productivity group, including pea protein isolate, untreated white grape pomace, and crude glycerol. Cluster 2 represented substrates with intermediate productivity, namely chicken feathers, sugar beet juice, and orange sweet potato peels. Cluster 3 contained the lowest-performing substrates, which included treated and untreated red grape pomace, hydrolyzed corn waste flour, and treated white grape pomace.

Fed-batch fermentation of R. mucilaginosa on untreated white grape pomace was successfully validated in a 2 L stirred-tank bioreactor under controlled aeration and pH regulation (

Table 4. Growth and carotenoid accumulation of R. mucilaginosa during fed-batch bioreactor cultivation on untreated grape pomace after 168 h cultivation.

Parameter	Shake Flask	Fed-Batch Bioreactor	Improvement (%)
Maximum biomass (g/L)	4.6	6.2	+35%
Maximum biomass (log CFU/mL)	8.26	8.5	+3%
Carotenoid yield (mg/100 g DW)	170.7	195.4	+14%
Volumetric productivity (mg/L)	14.1	20.1	+43%

). Biomass growth showed a characteristic sigmoidal pattern, with exponential increase up to ~120 h, followed by a plateau. The maximum dry biomass concentration reached 6.2 g/L (8.5 log CFU/mL), which represents a ~35% increase compared to shake-flask trials (4.6 g/L).

Carotenoid accumulation was markedly enhanced under fed-batch conditions. The maximum yield reached 195.4 mg/100 g DW, corresponding to a volumetric productivity of 20.1 mg/L, which represents a 43% increase compared with flask cultivation (14.09 mg/L). Carotenoid production followed a growth-associated pattern, reaching its peak between 144 and 168 h, in parallel with the late exponential phase. Continuous online monitoring confirmed stable process parameters, with dissolved oxygen maintained above 20% saturation and pH effectively controlled throughout the run. Each feeding pulse of glucose and diammonium phosphate (administered at 48, 72, 96, and 120 h) was followed by a rise in both biomass accumulation and pigment synthesis, reflecting the successful extension of exponential growth. Comparative analysis further demonstrated that the fed-batch bioreactor process delivered a 35% higher biomass yield and a 43% increase in carotenoid productivity relative to shake-flask trials. Collectively, these findings underline the scalability and robustness of untreated grape pomace as a sustainable feedstock for industrial carotenoid production.

4. Discussion

The sustainability scoring (Figure 1) highlighted that protein-rich biowastes, such as pea protein isolate, chicken feathers, and whey, are inherently more favorable for carotenoid production by Rhodotorula species. Their high nitrogen content reduced the need for external supplementation, making them efficient substrates. These observations are consistent with previous reports indicating that protein-rich agricultural residues can enhance fermentation performance, improving both process economics and environmental outcomes [27,28]. Similarly, potato and sweet potato peels performed well, reflecting their balanced nutrient composition and high starch levels, which can be readily hydrolyzed into fermentable sugars [29]. In contrast, sugar-dense but nitrogen-deficient substrate, such as crude glycerol, molasses, and sugar beet juice, required extensive nitrogen supplementation, which reduced their sustainability ranking. Comparable trends were documented in many studies [6,30,31] where highlight that carbon-rich substrates lacking nitrogen supplementation produced suboptimal carotenoid yields. Interestingly, grape pomace showed only intermediate to low sustainability scores, despite its richness in phenolic compounds with potential antioxidant benefits. This was mainly attributed to imbalanced C:N ratios and the inhibitory effects of polyphenols on yeast growth [32,33]. Taken together, the analysis supports high-nitrogen agro-residues as strong candidates for sustainable carotenoid bioproduction. Meanwhile, sugar-rich waste streams remain viable if paired with integrated nitrogen recovery or co-fermentation strategies that balance their nutrient profile. The absence of biomass formation on whey, potato peels, and molasses indicates that these substrates were unsuitable for supporting carotenoid biosynthesis by R. mucilaginosa under the tested conditions. Several factors likely contributed to this outcome, including imbalanced C:N ratios, the presence of inhibitory compounds, or the lack of metabolic precursors required for carotenoid synthesis. Similar substrate-dependent variability in pigment production by Rhodotorula spp. has been reported in previous studies, where nutrient composition played a decisive role [28,30].

Consequently, the exclusion of whey, potato peels, and molasses from further experimentation was justified based on their nutrient limitations and potential growth-inhibitory characteristics. As additionally explanation can be fact that whey exhibits a relatively balanced C:N ratio (~18:1), but its primary carbon source is lactose, which R. mucilaginosa cannot efficiently metabolize without enzymatic pretreatment [10,34,35]. Specifically, the strain employed in this study lacks the ability to utilize lactose [20], which likely restricted both biomass accumulation and secondary carotenoid synthesis. Potato peels, while more balanced nutritionally, contain high levels of phenolic compounds and glycoalkaloids such as solanine and chaconine which are molecules known to inhibit microbial growth and metabolic activity, including pigment production [36]. In the case of molasses, its inherently high C:N ratio results in nitrogen limitation, while additional inhibitory factors such as heavy metals, Maillard reaction products, and osmotic stress-inducing sugars further constrain yeast growth and secondary metabolite formation [31,37].

These findings are consistent with previous reports demonstrating that carotenoid biosynthesis in Rhodotorula species is strongly influenced by nutrient balance and the absence of growth-inhibitory compounds [28,30]. Substrates with unbalanced nutrient profiles or elevated levels of inhibitors often require extensive pretreatment or supplementation to be viable as fermentation feedstocks which are steps that ultimately reduce their sustainability potential. Consequently, focusing subsequent experiments on substrates that already support biomass accumulation and visible pigmentation is likely to enhance both process efficiency and the ecological relevance of biowaste valorization. While nutrient imbalance emerged as a major limiting factor, the presence of inhibitory compounds also contributed to suppressed carotenoid production in certain substrates. In particular, polyphenols in grape pomace and residual heavy metals in molasses may have restricted metabolic activity. Polyphenols are known to exert antimicrobial effects by compromising membrane integrity, altering protein function, and disrupting oxidative balance, thereby interfering with secondary metabolite pathways, including carotenoid biosynthesis [38,39]. Likewise, molasses may contain trace metals such as Fe, Zn, and Cu, along with Maillard reaction products and osmotic stress-inducing sugars, all of which have been reported to inhibit microbial metabolism [37]. To clarify the role of these compounds, future work should quantify the total phenolic content of candidate substrates and correlate these values with carotenoid yields. Detoxification strategies, such as phenolic adsorption using activated carbon or removal with polyvinylpolypyrrolidone, could also be employed to test whether pigment production improves once inhibitors are removed [40,41].

To better understand how substrate composition influences pigment yields, it is useful to consider the carotenoid biosynthetic pathway in R. mucilaginosa. In this yeast, carotenoid formation proceeds via the eukaryotic mevalonate pathway, beginning with acetyl-CoA and progressing through HMG-CoA to generate the isoprenoid precursors isopentenyl pyrophosphate and dimethylallyl pyrophosphate [42]. These intermediates are condensed into geranylgeranyl pyrophosphate, the branching point for carotenoid synthesis, which involves key enzymes such as CrtI, CrtYB, CrtS, and CrtR [13]. The choice of substrate can modulate carotenoid production in two primary ways: (1) by altering precursor availability, particularly acetyl-CoA and NADPH, thereby impacting flux through the HMG-CoA reductase step, and (2) by inducing oxidative stress, which stimulates carotenoid biosynthesis as part of an antioxidant defense response [13,43]. For instance, substrates funneled through the pentose phosphate pathway can increase NADPH generation and enhance pigment production [12]. Insights from metabolic engineering confirm the centrality of these regulatory nodes. Overexpression of HMG1, which encodes HMG-CoA reductase, has been shown to increase carotenoid yields, particularly when combined with sterol biosynthesis inhibitors such as ketoconazole that redirect flux toward isoprenoids [17]. Similarly, high-performing R. mucilaginosa mutants identified via whole-genome resequencing illustrate how genetic changes in MVA-linked genes can elevate carotenoid synthesis [6]. Collectively, these findings suggest that agro-industrial substrates that provide favorable carbon precursors and redox balance, or moderately induce stress responses, can enhance carotenoid production through both metabolic flux regulation and adaptive upregulation.

The application of the four-parameter sigmoidal model (Figure 2, Table 1) allowed precise characterization of Rhodotorula growth kinetics on diverse agro-industrial wastes. High R² values for crude glycerol, untreated white grape pomace, and pea protein isolate confirmed that these substrates provide balanced nutrient profiles with minimal inhibitors, supporting rapid adaptation and sustained exponential growth [41]. Crude glycerol, as a biodiesel byproduct, supported early biomass accumulation and reached near-maximum levels within 72–120 h, consistent with its efficient assimilation via glycerol kinase and glycerol-3-phosphate dehydrogenase, which channel carbon into lipid and carotenoid biosynthesis without significant growth delays [44]. Pea protein isolate also supported high biomass levels, attributable to its rich nitrogen content, which has been linked to enhanced yeast proliferation in protein-rich media [45]. By contrast, grape pomace-derived substrates displayed more variable performance. Untreated white grape pomace supported high yields, reaching d = 10.0 log CFU/mL in the model, likely due to residual sugars and micronutrients in the raw material [46,47,48]. However, treated pomace generally performed worse, reflecting both the loss of soluble carbohydrates and the accumulation of stress-inducing polyphenols during processing [47]. Substrates such as hydrolyzed corn waste flour and treated white grape pomace showed low maximum biomass values, probably due to suboptimal C:N ratios and inhibitory breakdown products (e.g., furfurals, phenolics), consistent with reports on lignocellulosic wastes [49,50]. The inflection point (c) provided further insight into biomass dynamics. A rapid inflection, as observed in untreated red grape pomace (39.7 h), suggests substrates that promote fast initial growth but lack long-term nutrient availability, leading to early plateauing.

Overall, the modeling confirmed that crude glycerol, pea protein isolate, and untreated white grape pomace are the most promising substrates for high-density cultivation, combining fast growth with high maximum biomass yields. These results are consistent with recent findings highlighting the potential of protein-rich and sugar-rich agro-industrial residues for carotenoid production [28,51]. Statistical validation further underscored the robustness of the four-parameter sigmoidal model. Low χ² values across most substrates align with established thresholds for reliable microbial growth modeling, where values below 0.5 are considered indicative of strong fits [25,52]. Near-zero MBE values confirmed that the model neither systematically over- nor underestimated biomass trends, reflecting accurate calibration [53]. The slight underestimation for treated white grape pomace may be linked to irregular biomass dynamics caused by nutrient limitations or residual inhibitors [54]. Importantly, MPE values under 10% fell within the range reported in microbial growth prediction studies [55], while low skewness values indicated that residuals were randomly rather than systematically distributed. Together, these results reinforce the reliability of the sigmoidal model for predicting yeast biomass kinetics across diverse biowaste substrates. Notably, the model performed best for untreated red grape pomace, treated red grape pomace, and untreated white grape pomace, which also ranked among the highest in carotenoid productivity—highlighting the link between accurate growth prediction and overall bioprocess efficiency.

The results summarized in Table 2 clearly indicate that pea protein isolate, untreated white grape pomace, and crude glycerol represent the most promising biowastes for carotenoid production by Rhodotorula spp., each achieving productivities above 13 mg/L. The strong performance of pea protein isolate can be explained by its low C:N ratio (~7:1), which ensures readily available nitrogen to support yeast proliferation. This observation is consistent with the previous findings [17,18] where demonstrated that nitrogen-rich substrates enhance pigment biosynthesis. Untreated white grape pomace also proved highly effective, likely due to its residual sugars and phenolic compounds, which may act as metabolic modulators, as similarly reported in valorization studies by Dulf et al. [39]. Crude glycerol, a biodiesel industry by-product, supported high biomass and carotenoid yields without the need for additional carbon supplementation, aligning with Saenge et al. [30], who identified crude glycerol as an efficient carbon source for oleaginous yeast cultivation. In contrast, substrates such as treated white grape pomace, untreated red grape pomace, and treated red grape pomace exhibited markedly reduced productivity. This reduction can be attributed to nutrient loss during processing steps or to the presence of inhibitory phenolic compounds and tannins [38]. These results confirm that both nutrient composition and processing history are decisive factors in determining carotenoid productivity, emphasizing that minimally processed, nutrient-rich, and non-inhibitory substrates offer the best prospects for sustainable pigment production.

The statistical grouping analysis (Figure 3) further supports these conclusions. Pea protein isolate, untreated white grape pomace, and crude glycerol all produced 13.87–14.98 mg/L of carotenoids and clustered within the same top-performing group (“a”). Their comparable productivity, despite contrasting biochemical profiles, nitrogen-rich (pea protein isolate) versus carbon-rich (crude glycerol), underscores the importance of balanced C:N ratios in achieving optimal pigment levels, a finding consistent with Kot et al. [18] and Saenge et al. [30]. Chicken feathers and sugar beet juice occupied intermediate positions (groups “b” and “bc”), likely reflecting their more complex nutrient composition and the need for additional metabolic adaptation by the yeast. Substrates in the lowest-ranking groups (“cd,” “de,” “e,” and “f”), including treated pomaces and hydrolyzed corn waste flour, likely performed poorly due to nutrient depletion during processing or the presence of inhibitory compounds such as phenolic tannins [38]. Taken together, this grouping-based interpretation highlights the importance of prioritizing optimization efforts for substrates in group “a,” while those in lower-performing groups may require targeted pretreatments or supplementation strategies to achieve competitive yields. The agreement between Tukey’s HSD groupings and hierarchical clustering (Table 2, Figure 3) confirms that the three leading substrates (Cluster 1, Group “a”) are robust candidates for further process optimization. Their comparable productivity, despite having contrasting nutrient profiles, demonstrates that Rhodotorula spp. can adapt to both nitrogen-rich substrates such as pea protein isolate and carbon-rich substrates such as crude glycerol, provided that the C:N ratio is properly balanced. Substrates in Cluster 2, which displayed intermediate productivity, may benefit from targeted supplementation strategies, for example, amino acid enrichment for chicken feathers or phenolic reduction in sugar beet juice, to improve carotenoid synthesis [38]. By contrast, Cluster 3 substrates consistently showed low productivity, likely due to nutrient depletion during processing (treated pomaces) or inhibitory compounds such as tannins in red grape pomace, in agreement with findings by Dulf et al. [39]. These results indicate that while high-yield substrates can be optimized directly, low-performing substrates will require pretreatment or microbial adaptation strategies before they can achieve competitive productivity. Overall, the combined use of parametric statistical testing and unsupervised clustering not only validated the robustness of the productivity ranking but also provided a rational framework for prioritizing biowaste substrates in carotenoid production.

Although the sustainability score offered a useful proxy for substrate suitability, a more comprehensive evaluation requires integrating economic and environmental considerations. Techno-economic analysis (TEA), as basic version given in

Table 5. Basic techno-economic comparison of selected substrates for carotenoid production by R. mucilaginosa.

Substrate	Estimated Raw Material Cost (€/ton) *	Productivity (mg/L)	Sustainability Score	Relative Feasibility **
Pea protein isolate	~1800–2000	14.98	2.93	Low (too costly despite high yield)
Crude glycerol	~100–200	13.87	−1.56	High (cheap, abundant, scalable)
Untreated grape pomace	~0–50 (disposal cost avoided)	14.09	−0.78	High (zero-cost waste, valorization benefits)
Chicken feathers	~50–100	8.53	1.23	Moderate (low cost, but processing required)
Sugar beet juice	~200–300	7.65	−0.92	Moderate (regional availability, requires supplementation)

Notes: * Costs are estimated from published reports on agro-industrial residues [2,57,58]; actual costs vary regionally; Sustainability Score values are taken from Figure 1; ** Relative Feasibility integrates productivity, cost, and availability.

for this research, provide powerful tools to determine whether top-yielding substrates also align with cost-effectiveness and reduced environmental burden. For instance, pea protein isolate, despite its superior biological performance, is a relatively high-value commodity in food and feed markets, which constrains its feasibility as a large-scale fermentation feedstock [56]. In contrast, crude glycerol, a biodiesel by-product, is inexpensive and abundantly available, making it an attractive option despite its requirement for nitrogen supplementation [57]. Likewise, untreated grape pomace represents a zero-cost substrate commonly discarded by wineries, offering additional valorization benefits such as reduced disposal costs and mitigation of environmental pollution [58]. A simplified cost comparison suggests that while protein-rich substrates may deliver maximum carotenoid yields per unit volume, low-cost carbon-rich residues such as glycerol and grape pomace may be more scalable due to their reduced raw material expenses. From an environmental perspective, valorizing agro-industrial residues diverts waste from landfilling or incineration and contributes to greenhouse gas emission reductions, thereby enhancing the overall sustainability of the process [2]. Future studies should therefore combine experimental productivity results with economic and environmental metrics, including substrate cost profiles and CO₂-equivalent emissions avoid, to provide a holistic picture of scalability.

Although pea protein isolate delivered the highest carotenoid productivity (14.98 mg/L), its relatively high market price (~€1800–2000/ton) limits its industrial feasibility. By contrast, crude glycerol and grape pomace (both low-cost by-products) combined competitive productivity (13.87–14.09 mg/L) with scalability potential due to their abundance and valorization benefits (Table 5). These comparisons emphasize that economic viability and substrate availability are as critical as biological performance when selecting feedstocks for large-scale carotenoid bioproduction. Taken together, the comparative evaluation identifies pea protein isolate, crude glycerol, and untreated white grape pomace as the most promising substrates. However, from a techno-economic and sustainability perspective, untreated grape pomace stands out as the optimal choice, offering high carotenoid yields while simultaneously valorizing an abundant agro-industrial residue. This aligns with circular bioeconomy principles by transforming winemaking waste into high-value pigments, while also reducing disposal costs and environmental impact. Crude glycerol, as a biodiesel by-product, similarly represents a viable alternative, particularly in regions with established biodiesel production chains.

Fed-batch bioreactor cultivation of R. mucilaginosa on untreated grape pomace demonstrated a substantial enhancement in both biomass accumulation and carotenoid productivity compared to flask-scale trials, with volumetric yields increasing by approximately 43%. This improvement is consistent with previous studies showing that controlled aeration, pH regulation, and nutrient feeding significantly promote carotenoid biosynthesis in Rhodotorula spp. Saenge et al. [30] reported a 1.5-fold improvement in carotenoid production under fed-batch operation with glycerol as the carbon source, while Kot et al. [18,19] demonstrated that targeted nitrogen supplementation enhanced both growth and pigment accumulation in bioreactor systems. The carotenoid yield obtained in this study (195.4 mg/100 g DW, 20.1 mg/L) is comparable to, or higher than, values previously reported for Rhodotorula species cultivated under optimized fed-batch conditions. For example, Bonadio et al. [12] achieved maximal carotenoid production from R. rubra grown on sugarcane juice with nutrient supplementation (magnesium addition), and Sharma et al. [59] reported 717.35 μg/g of β-carotene from R. mucilaginosa cultivated on agro-industrial residues. The superior performance of untreated grape pomace highlights its dual advantage: high carotenoid productivity combined with environmental and economic benefits derived from valorizing winery residues. Unlike refined feedstocks such as glucose or protein isolates, grape pomace is abundant, inexpensive, and requires minimal pretreatment, making it a highly attractive candidate for sustainable pigment bioprocessing. These findings reinforce the feasibility of integrating winemaking by-products into circular bioeconomy models, where agricultural residues are transformed into value-added bioproducts.

In this study, the focus was on a wild-type strain of R. mucilaginosa, since limitation to work with genetically engineered microorganisms. While this may partly explain the relatively low carotenoid yields compared with reports in the literature, it enabled us to establish a baseline comparison of diverse agro-industrial biowastes under conditions that are broadly accessible to many laboratories. Authors acknowledge that metabolic engineering of Rhodotorula spp. has been successfully employed to enhance carotenoid yields, for example, through overexpression of phytoene synthase (crtB) or pathway rewiring to favor torulene and torularhodin biosynthesis [14,18,19,28]. Future work could combine the substrate suitability identified in this study with such engineering strategies to further improve titers and develop scalable bioprocesses.

5. Conclusions

This study systematically investigated 13 agro-industrial biowastes as substrates for carotenoid production by Rhodotorula mucilaginosa. Through an integrated approach combining sustainability scoring, kinetic modeling, and productivity assessment, the results demonstrated that substrate composition plays a decisive role in both biomass growth and pigment biosynthesis. Among the tested materials, pea protein isolate, crude glycerol, and untreated white grape pomace yielded the highest carotenoid productivities (>13.8 mg/L). Although pea protein isolate exhibited excellent biological performance, its high market value and strong demand in the food and feed sectors reduce its practicality as a fermentation feedstock. Crude glycerol, a low-cost by-product of biodiesel production, also supported strong productivity but required nitrogen supplementation to balance its nutrient profile. By contrast, untreated grape pomace emerged as the most promising option, combining competitive carotenoid productivity (14.09 mg/L) with cost-free availability and significant environmental benefits through the valorization of winery residues. This positions grape pomace as an optimal substrate for sustainable carotenoid production, directly supporting circular bioeconomy principles by converting agricultural waste into high-value bioproducts. The fed-batch bioreactor validation further confirmed the feasibility of this approach. Under controlled fermentation, R. mucilaginosa efficiently converted untreated grape pomace into carotenoids, achieving a 35% increase in biomass yield and a 43% improvement in carotenoid productivity compared with shake-flask trials, ultimately reaching 20.1 mg/L. These findings highlight untreated grape pomace, when coupled with fed-batch processing, as the optimal substrate–strategy combination for scalable and sustainable carotenoid production. Overall, this work provides a rational framework for integrating agro-industrial residues into biotechnological platforms, demonstrating how waste streams such as grape pomace can be valorized into natural pigments of commercial relevance. Future research should prioritize techno-economic analyses and process optimization to support industrial implementation at scale.