Scalable production of muscle and adipose cell-laden microtissues using edible macroporous microcarriers for 3D printing of cultured fish fillets

Abstract

Cellular agriculture is a novel platform for addressing the issues of protein scarcity, environmental pressures, and food safety. However, expanding seed cells at a large scale remains a prerequisite for achieving industrial cultured meat production. We here propose the production of large-pore-sized edible porous microcarriers (EPMs) by using NaCl to precisely control ice crystal growth during cryogenic crosslinking. Muscle satellite cells (SCs) and adipose-derived stem cells (ASCs) from large yellow croakers exhibit remarkable adhesion, proliferation, and differentiation on gelatin-based EPMs. Following consecutive expansion, SCs and ASCs densities reach 6.25 × 10⁵ and 5.77 × 10⁵ cells/mL, leading to a 499-fold and 461-fold increase in cell numbers, respectively. Subsequently, the mature microtissues are incorporated into a bioink, thereby enabling successful three-dimensional printing of cultured fish fillets with sensory properties similar to native fish fillets. EPM-based cell expansion and food structuring techniques are promising in facilitating large-scale cultured fish meat production.

Introduction

The demand for animal protein is rapidly increasing because of the growing global population, increased pursuance of high-quality protein, and urbanization-mediated changes in consumption patterns. Fish and seafood are indispensable protein sources that are rich in omega-3 fatty acids and trace nutrients¹. Aquaculture, akin to traditional livestock farming, faces substantial hurdles due to environmental pollution, global greenhouse gas emissions, and animal welfare concerns^2,3,4. Meat contributes to more than 18% of the global protein intake and has a huge demand⁵. However, a question arises about the ideal strategy for providing meat substitution with eating experiences similar to those offered by animal-derived meat.

As a sustainable and eco-friendly meat substitution, cultured meat is drawing the marked attention of researchers attempting to explore available high-quality meat for the growing global population, reduce environmental impact, and increase animal welfare^6,7. Nevertheless, research on cell-cultured meat is in its infancy because of a series of challenges, primarily related to the establishment of robust seed cell lines, implementation of scalable cell expansion techniques, development of low-cost serum-free culture media, and advancement of food structuring technology for cell-to-meat conversion^8,9,10,11. Among these challenges, developing efficient methodologies for scalable cell expansion is essential for generating abundant cells for cell-to-meat processing. However, most seed cells obtained from animal biopsies are adherent-dependent cells¹², and their growth is limited by adherent-dependent proliferation, which hampers their self-expansion capability¹³.

Microcarrier suspension culture platforms have enabled large-scale adherent cell cultivation¹⁴. However, commercially available microcarriers frequently used in the medical field lack food-grade properties as they contain toxic substances introduced by the material itself or the manufacturing process¹⁵. These microcarriers serve as temporary scaffolds during expansion, and the expenses and cell loss incurred during subsequent cell harvesting are unwanted¹⁶. Unlike the standards used in tissue engineering and regenerative medicine, microcarrier scaffolds used for cellular agriculture must ideally be edible, nutritionally sound, safe, cost-effective, and amenable to large-scale production. Compared with conventional spherical microcarriers, edible porous microcarriers (EPMs) have an intricate support network that allows increased cellular adhesion, growth, and diffusion¹⁷. After the proliferated cells converge on EPMs, these high-density cell-carrying microcarriers serve as building blocks for assembling meat analogs approaching the true meat size by using suitable food structuring techniques. Three-dimensional (3D) printing technology has been widely used for producing plant-based meat analogs¹⁸. This innovative technology allows for designing and assembling customized, personalized, and nutritionally controlled food products by precisely controlling each component’s proportion and distribution in the product matrix^18,19. Therefore, cell-laden microcarrier-based bioinks can potentially be integrated into the cultured meat supply chain for large-scale production.

We here developed EPMs and established a method for introducing macroporous structures within these EPMs. We also demonstrated the scale-up expansion and maturation of seeded cells on the EPMs and ultimately incorporated mature cell-laden microtissues into a bioink for 3D printing cultured fish fillets.

Results

Production and characterization of EPMs with macroporous structures

Fish gelatin and microbial transglutaminase (mTG) were used for crosslinking in the microcarrier scaffold. The precursor solution was emulsified, and the temperature of the solution was decreased to −20 °C. During cryogenic crosslinking, spherical porous microcarriers were formed (Fig. 1a). To obtain microcarriers with larger pore structures, different reaction temperatures and matrix concentrations (gelatin) were initially tested (Supplementary Fig. 1), wherein the reduction in the gelatin concentration caused no alterations in the average pore size of the microcarriers (Supplementary Fig. 1f). At a −30 °C reaction temperature, the prepared EPMs exhibited the largest average pore size of 6.28 μm (Supplementary Fig. 1b, e). However, macroporous structure-lacking microcarriers cannot rely on the internal pore surface area to increase the cell density (and thereby increase productivity)²⁰. Although studies have reported that reaction temperature²¹ is positively correlated with polymer content²² with pore size in bulk cryogel systems, this relationship does not appear to hold for a gelatin microcarrier scaffold system.

Fig. 1: Control of EPMs pore size by NaCl modulation of ice crystal growth.

a Schematic diagram of scaffolds fabricated from fish gelatin solution by cryogenic emulsification in the presence of a food-grade cross-linker (mTG). White scale bar: 100 μm, black scale bar: 500 μm. b Mechanism of NaCl-enabled pore size control by regulating ice crystal growth. The presence of NaCl (e.g., 1%−5%) in the gelatin mTG precursor solution induced a decrease in the cross-linking density of the gelatin network, resulting in a larger ice crystal and corresponding pore size by further ice crystal growth. c Characterization of Young’s modulus of gelatin-mTG hydrogels containing different concentrations of NaCl (n = 3 independent experiments). d The size of the ice crystals depends on the NaCl concentration in the gelatin mTG precursor solution. Scale bar: 100 μm. e, f Control of pore size of drying (e) and swelling (f) EPMs scaffolds by different concentrations of NaCl (n = 5 independent experiments, with scatter points representing individual EPMs’ pore size). g, h Swelling ratio (g) and porosity (h) of EPMs scaffolds fabricated by incorporating different concentrations of NaCl (n = 5 independent experiments). i SEM images of EPMs scaffolds demonstrating the correlation between pore size and NaCl concentration. Scale bar: 50 μm. For (a, b), diagrams were created using biorender.com. For (c, e–h), error bars indicate means ± SD, ** p < 0.01, **** p < 0.0001, significance was determined by one-way ANOVA with Tukey’s post hoc analysis. For (d, i), similar results were obtained in three independent experiments.

On the other hand, introducing NaCl can simply control the EPM pore size (Fig. 1b). Regarding the mechanism underlying NaCl-mediated regulation of the EPM pore size, we hypothesized that NaCl enrichment during freezing alters the ice crystal size in the aqueous phase, thereby affecting the formation of corresponding pores. Specifically, the ion concentration in polymers is known to determine the aggregation state of the polymers because of the Hofmeister effect^23,24. When NaCl is present, electrostatic interactions occur between NaCl and gelatin chains in the precursor solution. These interactions reduce the availability of cross-linkable groups in gelatin, thereby weakening the mTG and cryogenic-induced gelatin network. In this scenario, more free water is available for crystallization in the gelatin polymer, and ice crystals grow further, leading to the formation of larger pores in the EPMs. Thus, the initial NaCl concentration in the precursor solution can be altered to control the size of the growing ice crystals and the resulting pores in the frozen scaffold. To validate the hypothesis, changes in the Young’s modulus of gelatin hydrogels with varying NaCl concentrations were quantified. A gradual decrease in the modulus was observed with an increase in the NaCl concentration from 0% to 5% (w/w) (Fig. 1c). Next, NaCl-mediated regulation of the ice crystal size in the gelatin hydrogel system was directly visualized. Higher NaCl concentrations led to the formation of larger ice crystals (Fig. 1d, Supplementary Movies 1–3). Finally, NaCl-mediated pore-size regulation of the porous microcarrier scaffold was analyzed through scanning electron microscopy (SEM) (Fig. 1i). As the NaCl concentration increased from 0% to 5%, the pore size increased from 4.34 to 27.64 μm (Fig. 1e), which caused a corresponding decrease in the specific surface area by 63.63% (Supplementary Fig. 2a). The porosity of the EPMs synthesized with different NaCl concentrations remained consistently above 80% (Fig. 1h), with no significant differences observed.

Because EPMs are used in culture media, the structural characteristics and mechanical properties of the EPMs were assessed upon hydration in the medium. Dry EPMs rapidly hydrated once they came in contact with water, which led to an evident increase in the EPM size (Supplementary Fig. 2b–d) and an 18-fold increase in weight (Fig. 1g). Additionally, as the NaCl concentration increased, the EPM swelling ratio significantly increased (Fig. 1g). Interestingly, even after the EPMs were swollen, NaCl regulated the pore size of the porous EPMs. Specifically, because of the swelling effect, the adjustable pore size range of the porous EPMs in water increased to 15.56–69.90 μm (Fig. 1f, Supplementary Fig. 2i). After long-term incubation in culture media, no significant mass loss was observed in different EPM scaffolds (Supplementary Fig. 2e). EPMs fabricated with varying NaCl concentrations have Young’s modulus within the 30.26–56.20 kPa range in culture media (Supplementary Fig. 2f), thus allowing for the refined customization of cell scaffolds in subsequent studies. Collectively, these results unveiled that the NaCl-mediated pore control method is effective for producing edible porous biomaterial scaffolds. A simple washing procedure can keep residual NaCl in the scaffold within 0.001% (w/w), thereby preventing adverse effects on cells (Supplementary Fig. 2h). Importantly, cyro-emulsion based EPM production can be easily scaled up, as evidenced through our successful porous microcarrier production in a 100-L reaction system (Supplementary Fig. 2j). In detail, 600 g fish gelatin can yield 485.21 g EPMs, which are presented as a white powder after sterilization (Supplementary Fig. 2k), with approximately 2000 microcarriers/mg (Supplementary Fig. 2g).

Expansion and maturation of SCs and ASCs on EPMs

Seed cells are the source of cultured meat and provide muscle and adipose tissue components². Muscle satellite cells (SCs) and adipose-derived stem cells (ASCs) were isolated from the epaxial muscle and ventral coelomic fat tissues of large yellow croakers (Larimichthys crocea, Supplementary Fig. 3a, b). To characterize the cells isolated from the muscle tissue, immunofluorescence staining was performed for PAX7 and MYOD1. PAX7 is a specific marker for SCs that plays a pivotal role in cellular development and functional maintenance²⁵. Upon activation from quiescence, MYOD expression is initiated, which indicates the onset of satellite cell proliferation. PAX7 and MYOD coexpression in SCs signifies the transition of these cells into the proliferative phase²⁶. According to the results, 44% of the epaxial muscle-derived cells exhibited PAX7 positivity, 72% displayed MYOD1 positivity, and 42% coexpressed PAX7 and MYOD1 (Supplementary Fig. 3c, e). ASCs bind to the extracellular matrix (ECM) through surface antigens involved in cell-cell and cell-matrix interactions, and the specific surface markers CD29 and CD44 serve as identifiers for ASCs²⁷. In this study, the ventral coelomic fat tissue-derived cells exhibited 92% and 98% positive rates for CD29 and CD44, respectively, whereas the cells exhibited negativity for the hematopoietic cell surface marker CD45 (Supplementary Fig. 3d, f). These findings suggest that the epaxial muscle- and ventral coelomic fat tissue-derived cells correspond to SCs and ASCs, respectively. Induction studies were conducted to evaluate the muscle and adipose differentiation potentials of SCs and ASCs. After 6 days of myogenic induction culture, aligned and fused cells were observed in SCs (Supplementary Fig. 3i). Immunofluorescence staining revealed the presence of MYH-positive multinucleated myotubes on days 3 and 6 (Supplementary Fig. 3g, i), with a fusion index of 78% on day 6. In adipogenic induction differentiation, 92% of ASCs exhibited lipid droplets on day 3, and the number and size of these droplets significantly increased by day 6 (Supplementary Fig. 3h, j). These results demonstrate that SCs and ASCs serve as ideal progenitor cells for the development of cultured meat. To obtain adequate amounts of muscle and fat tissue, spinner flasks were used for SC and ASC expansion and differentiation (maturation) (Fig. 2a). Regarding cell scaffolding, the cultivation performances of EPMs having three pore sizes were compared. To ensure that cells effectively attached to the EPM surface, we used a stir-stop seeding strategy. Both SCs and ASCs displayed an adhesion rate of >80%, with no significant difference noted in the seeding efficiency among EPMs having different pore sizes (Fig. 2b). With further cultivation, cells on EPMs were imaged and quantified by conducting live/dead cell staining, live cell quantification, and viability test. The images clearly demonstrated that the confluence of SCs and ASCs on different pore-sized EPMs gradually increased over 7 days of proliferation in the spinner flasks, eventually completely covering the microcarriers (referred to as microtissues, Fig. 2c). Interestingly, SCs and ASCs fused to form aggregates during differentiation (Fig. 2c). SEM images revealed that the microtissues were densely covered with the cells (Fig. 2d).

Fig. 2: SCs and ASCs expansion and maturation on EPMs.

a Schematic representation of the culture conditions of SCs and ASCs. Cells were cultured continuously in proliferation medium (Pif) for 7 days and subsequently replaced with differentiation medium (Dif) containing horse serum till Dif D6. b Cell seeding efficiency of different EPMs scaffolds (n = 4, independent experiments). c Representative Live-dead cell staining images. Green: live cells (Calcein AM); Red: dead cells (Propidium iodide, PI). Scale bar: 500 μm. d Representative SEM images of SCs and ASCs after seven days of proliferation and 6 days of differentiation in the bioreactor. Scale bar: 50 μm. e, f Growth and viability curves of SCs (e) and ASCs (f) on different EPMs scaffolds after 13 days of culture (7 days for proliferation and 6 days for differentiation) (n = 3 independent experiments). g Representative immunofluorescence staining images of SCs and ASCs after 7 days of proliferation and 6 days of differentiation in the spinner flask. Scale bar: 100 μm. h The mRNA expression of genes related to myogenic differentiation (MYOD1, MYOG, MYH2) and lipogenic differentiation (LPL, PPARγ) was quantified by qRT-PCR (n = 3 independent experiments). For (b, e, f, h), error bars indicate means ± SD. For (b, h), * p < 0.05, *** p < 0.001, **** p < 0.0001 and ns indicates p > 0.05, significance was determined by one-way ANOVA with Tukey’s post hoc analysis. For (c, d, g), similar results were obtained in three independent experiments. For (d, g, h), data were derived from cells cultured on EPM fabricated by incorporating 3% NaCl.

The cell density was quantitatively analyzed. According to this analysis, SCs and ASCs reached their highest cell density on EPMs after 5–7 days of cultivation (Fig. 2e, f). However, the number of live cells gradually decreased during SCs and ASCs differentiation. Notably, EPMs-3NaCl and EPMs-5NaCl with larger pore sizes exhibited higher cell density during SC and ASC cultivation than EPMs-0NaCl with a smaller pore size (Fig. 2e, f). This phenomenon may be attributable to the larger pore size of EPM scaffolds, which provide greater space for cell adhesion and growth. Confocal laser scanning microscopy observations demonstrated that cells on EPMs-3NaCl with a larger pore size were deposited on the EPMs and achieved multilayered growth within their macroporous structure. By contrast, cells on EPMs-0NaCl with a smaller pore size only deposited on the monolayer surface of the EPMs (Supplementary Fig. 5, Supplementary Movies 4–7). Next, RNA-seq analysis of cells on EPMs with different pore sizes was conducted to comprehensively evaluate the effect of various EPMs on cell behavior. The gene expression profiles of SCs and ASCs differed significantly on EPMs synthesized using varying NaCl concentrations (Supplementary Fig. 6a, b). Cell cycle-related genes such as CCNB1, CCNB3, CCND1, CCND3, and CKC2 were highly expressed in cells on large-pore-sized EPMs (manufactured with 3% and 5% NaCl). According to the Gene Ontology (GO) biological process enrichment analysis, cell cycle- and cell division-associated terms were enriched (Supplementary Fig. 6c–f). This suggests that compared with EPMs with smaller pores, EPMs with larger pores are more conducive to promoting SCs and ASCs proliferation (Fig. 2c, e, f), affirming the notion that EPMs with larger pore sizes offer greater space and flexibility for cell growth and reshaping. Studies have also reported the promotive effects of larger-pore-sized scaffolds on cell proliferation^17,21. Additionally, EPMs supported the maturation of adhered SCs and ASCs. Immunofluorescence staining presented a clear expression of the mature skeletal muscle fiber marker MYH in the porous structure and interstices of muscle microtissues (Fig. 2g, Supplementary Fig. 5, Supplementary Movie 5). On the other hand, the cytoplasm of ASCs present on adipose microtissues was filled with lipid droplets, a typical characteristic of mature adipocytes (Fig. 2g, Supplementary Fig. 5, Supplementary Movie 7). High expression of MYOD1, MYOG, MYH2, and PPARγ genes in the qPCR results also suggested the maturity of muscle and adipose microtissues (Fig. 2h), thereby demonstrating a differentiation efficiency comparable to that of C2C12²⁸ and 3T3L1²⁹ cells.

Scale-up production of SCs and ASCs

Cell quantities required for the commercial-scale cultivation of anchorage-dependent seed cells and their progeny for cultured meat production are higher than those provided by traditional adherent culture techniques³⁰. Therefore, we have developed a method for the consecutive expansion of anchorage-dependent cells. This expansion method was initiated with 5.00 × 10⁶ SCs isolated from one T75 culture flask and seeded into a 125-mL spinner flask containing 100 mL culture medium. After 5 days of suspension culture of SCs on EPMs, the viable cell density was 5.03 × 10⁵ cells/mL (Supplementary Fig. 7b), with a total of 5.03 × 10⁷ viable cells (Supplementary Fig. 7c). At this point, the SCs had completely covered the EPMs. However, further cell growth was constrained because of the limited availability of adhesive surfaces (Supplementary Fig. 7a). To facilitate continued cell expansion, the cells must be transferred onto new EPMs, and an appropriate procedure was required for this transfer. The cell amplification protocols using bead-to-bead transfer and food-grade collagenase digestion were compared (Supplementary Fig. 7). In bead-to-bead transfer, cell-laden microtissues, and new EPMs are transferred to the next-level spinner flask (500 mL). With time, cell viability gradually decreased (Supplementary Fig. 7c). Images obtained after live/dead staining unveiled ineffective cell transfer onto the new EPMs, with apoptotic cells exhibiting a notable increase (Supplementary Fig. 7a). By contrast, the collagenase digestion-based cell amplification procedure facilitated uniform cell seeding in the next-level spinner flask (500 mL), which resulted in stable cell proliferation and cell viability of >80% throughout the subsequent cultivation period (Supplementary Fig. 7a, d, e). Therefore, collagenase digestion-based reseeding is an effective approach for supporting continuous cell expansion. Based on this amplification approach, continuous cell expansion was carried out for 16 days. The final 4-L culture system achieved a viable cell density of 6.25 × 10⁵ cells/mL and a total of 2.50 × 10⁹ live cells, which corresponded to a 499-fold expansion (Fig. 3a, b, g, h, Supplementary Fig. 8). After excess moisture was removed from the harvested microtissues, the yield amounted to 30.97 g (Supplementary Fig. 8d). Each gram of microtissue contained over 80 million cells. During consecutive expansion, individual EPMs accommodated up to 128 SCs and 118 ASCs (Supplementary Fig. 8f), which resulted in cellular contents of 13.14% and 10.51% (w/w, dry basis) within the harvested muscle and fat microtissues, respectively. Moreover, the high cell viability rate noted during the consecutive expansion steps ensured robust cell expansion (Fig. 3b, Supplementary Fig. 8b). Remarkably, the doubling time of SCs during consecutive expansion in the bioreactor was significantly shorter than that observed during the two-dimensional (2D) monolayer culture (Fig. 3c). The efficient expansion process was also extended to ASCs. After 17 days of consecutive expansion, the final density of viable cells reached 5.77 × 10⁵ cells/mL, totaling 2.31 × 10⁹ cells in the 4-L culture system (Fig. 3d, e, g), with a doubling time significantly lower than that of the 2D monolayer culture (Fig. 3f). Quality control measures were implemented throughout the seeding cell expansion and maturation processes. No exogenous bacterial contamination was detected (Supplementary Fig. 8e). Finally, muscle and adipose microtissues were harvested through static settling, washed to remove residual culture medium, and produced as a milky paste-like material for the subsequent food structuring protocol (Fig. 3i).

Fig. 3: Scalable expansion of microtissues for cultured fish meat.

a, b, c Representative run showing cell densities (a), total cell number (b), viability (b) and doubling time (c) of SCs were monitored throughout consecutive expansion in 125 mL spinner flask, 500 mL spinner flask, and 5 L bioreactor. d, e, f Representative run showing cell densities (d), total cell number (e), viability (e) and doubling time (f) of ASCs were monitored throughout consecutive expansion in 125 mL spinner flask, 500 mL spinner flask, and 5 L bioreactor. g Summary of three independent consecutive expansion runs (n = 3 independent experiments). h Photograph of biotech-5JGC-7000ASC System (Baoxing bio-engineering, China) with a 5 L glass vessel. i Photographs of scaled-up and matured microtissues as raw material for cultured fish fillets product manufacturing. For (a, b, d, e), n = 3 independent experiments, error bars indicate means ± SD.

Transcriptome characterization of SCs and ASCs in consecutive expansion

When porous microcarriers are used for cell expansion in a bioreactor, as with any 3D culture system, various aspects of cell behavior and growth are affected³¹. After confirming substantial growth rates (Fig. 3a–f) and cell morphology using this culture system (Supplementary Fig. 8b), we performed the RNA-seq analysis to investigate the effect of consecutive expansion on the behavior of SCs and ASCs. In total, 10,670 differentially expressed genes (DEGs) were identified in each cell population, and the expanded strains of both cell types clustered together (Fig. 4a). Notably, various myogenesis-, cell cycle regulation-, and ECM remodeling-related genes were significantly upregulated, whereas lipogenesis-related genes exhibited minimal changes (Fig. 4b). Specifically, several genes related to mid-late-stage myogenic transcription factors, such as MYOG, MYOD1, and multiple MYH subtypes, and ECM remodeling were upregulated in SCs-EPMs (Fig. 4b). Surprisingly, the expression of myogenic genes was also significantly enhanced in ASCs-EPMs (Fig. 4b). During the consecutive expansion phase, we did not switch to myogenic differentiation media. These aberrant myogenic markers were upregulated, possibly because EPMs have a specific stiffness and structure, such as the higher myogenic differentiation degree of bovine myoblasts in scaffolds with hardness equivalent to those of natural muscle tissues³². Additionally, porcine skeletal muscle satellite cells and C2C12 cells exhibited spontaneous myogenesis after 7-day cultivation on commercial microparticles²⁹. This could potentially confer an advantage in producing skeletal muscle tissues with native tissue characteristics. Furthermore, cell cycle proteins were upregulated, reflecting the reduced cell doubling time during consecutive expansion (Fig. 3c, f). The enrichment analysis unveiled those genes involved in system development, cell communication, multicellular organism development, animal organ development, and transcription regulation were differentially expressed after consecutive expansion (Fig. 4c, d). These GO terms shared between the two seed cell types indicate that they had common mechanisms. The principal component analysis demonstrated the occurrence of transcriptional changes before and after consecutive expansion (Fig. 4e). In summary, the results indicate that the expanded cell strains displayed transcriptional alterations beneficial for cultured meat production.

Fig. 4: Clustering analysis of differential gene expression patterns in SCs and ASCs during consecutive expansion on EPMs.

a Heatmap of total genes in SCs and ASCs before (SCs-2D and ASCs-2D) and after (SCs-EPMs and ASCs-EPMs) consecutive expansion. b Heatmap showing the expression patterns of genes corresponding to myogenesis, lipogenesis, cell cycle, and extracellular matrix in SCs and ASCs before (SCs-2D and ASCs-2D) and after (SCs-EPMs and ASCs-EPMs) consecutive expansion. c, d Top 10 enrichment GO analysis for significantly upregulated after consecutive expansion on EPMs (SCs-EPMs and ASCs-EPMs) compared to before (SCs-2D and ASCs-2D) consecutive expansion on SCs (c) and ASCs (d). e Principal component analysis (PCA) plot of RNA-Seq data from SCs and ASCs before and after consecutive expansion. For (c, d), p-values were obtained using a two-tailed hypergeometric test without adjustment for multiple comparisons.

Construction of 3D printed fish fillets using bioink incorporating muscle and adipose microtissues

Acceptance of cultivated meat by mainstream consumers improves when it contains both muscle and adipose tissues². Muscle tissue imparts ample amino acids and chewiness to mimic native meat³³, animal fat offers unique flavor, aroma, and juiciness³². Therefore, a method to integrate expanded and mature muscle and adipose microtissues into bioink was developed. In this method, a small amount of pea protein was used to improve printability, and mTG was used to generate cohesive structures (Fig. 5a). We successfully printed 3D cultured fish fillets (length: 100 mm, height: 15 mm) by using a commercial 3D bioprinter (Fig. 5b, Supplementary Fig. 9a–d, Supplementary Movie 8). The cross-section of the 3D cultured fish fillet was similar to that of the native fish meat, and the Maillard reaction was evident (Fig. 5b, Supplementary Fig. 9c). The 3D-printed cultured fish fillet was comprehensively characterized in comparison with the native fish fillet. The DNA concentration was significantly lower in the cultured fish fillet than in the native fish fillet (Fig. 5d), which indicated that scaffold optimization is necessary for increasing cell loading for equivalence to native meat. Compared with the typical translucent white color of fresh large yellow croakers, the cultured fish fillet was more yellow because of the addition of pea protein and mTG (Fig. 5e). The color of the cooked cultured fish fillet deepened because of the Maillard reaction (Fig. 5e). Additionally, the water content of the cultured fillet and the native fish fillet was similar (~70%). The weight loss of ~35% that occurred during steaming was also comparable (Fig. 5f). Cooking increased the hardness and chewiness of both cultured and native fish fillets (Fig. 5g). By contrast, the cell-free cultured fish fillet had a cooking loss of up to 51.21% (Supplementary Fig. 11a). The cooked cultured and native fish fillets had equivalent hardness. However, the chewiness and cohesiveness of the cooked cultured fish fillet were lower than those of the cooked native fish fillet (Fig. 5g).

Fig. 5: Manufacturing and characterization of 3D printed cultured fish fillets.

a Properties of microcarrier-based cellular microtissues bioinks. The diagram was created using BioRender. b The appearance of raw and cooked 3D printed cultured fish fillet prototypes. c Nutritional analysis of cultured fish fillets compared to Native fish fillets (n = 3 independent experiments, data were presented as means ± SD). d DNA per gram of native fish fillet, muscle microtissues and fat microtissues. e Color of raw and cooked cultured fish fillet and native fish fillet. L*: lightness, a*: redness, and b*: yellowness. f Moisture and cooking loss of cultured fish fillet and native fish fillet. g Textural properties of raw and cooked cultured fish fillet and native fish fillet. For (d–g), n = 3 independent experiments, error bars indicate means ± SD, * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, and ns indicates p > 0.05, For (d, e, g), significance was determined by one-way ANOVA with Tukey’s post hoc analysis. For (f), significance was determined using two-tailed student’s t test.

According to the nutritional analysis, the cultured and native fish fillets had comparable energies (Fig. 5c). The cultured fish fillet exhibited 68.92% and 87.93% reductions in fat content and cholesterol levels, respectively, compared with the native fish fillet, with 8.5 g more protein per 100 g of product (Fig. 5c). Moreover, the cultured fish fillet had a higher sodium content, exceeding that of the native fish fillet by 192.7 mg/100 g (Fig. 5c). The cultured fish fillet contained the most characteristic omega-3 fatty acids found in fish oil, particularly C22:6 (5.99 g/100 g oil), and its unsaturated fatty acid content was similar to that of the native fish fillet (Supplementary Fig. 9e, f). On analyzing the amino acid profile, we noted that the cultured fish fillet had 51% higher essential amino acid content than the native fish fillet (Supplementary Fig. 9g, h). Flavor analysis revealed that the flavor compounds differed significantly between the cultured and native fish fillets (Supplementary Fig. 12a–c).

The nutritional attributes of cells directly impact the quality of cultured meat³⁴. Following consecutive expansion and maturation of cells on EPMs, SCs loaded on each EPM contributed 41.53 ± 2.25 ng of protein and 16.15 ± 0.88 ng of lipid (on a dry basis), while ASCs loaded on each EPM contributed 17.59 ± 1.03 ng of protein and 27.85 ± 1.63 ng of lipid (on a dry basis) (Supplementary Fig. 10b). To gain a more comprehensive insight of nutrient deposition in cellularized microtissues across different cultivation stages, the nutritional profiles were analyzed. Compared to uninoculated EPMs, cell-laden EPMs exhibited reduced protein content but accumulated higher levels of lipids, cholesterol, and carbohydrates (Supplementary Fig. 10a). Compared to undifferentiated cellularized microtissues, the protein and fat content of muscle microtissues after myogenic differentiation increased by 4.43% and 45.45%, respectively (Supplementary Fig. 10a). The accumulation of lipids during myogenic differentiation may indicate that SCs are thrust toward an adipogenic phenotype in differentiation media containing horse serum^35,36. This lipid accumulation could offer potential advantages for cultured meat production, warranting further investigation in future research. After adipogenic differentiation, the protein content of adipose microtissues decreased by 4.91%, while the fat content increased by 50.98% (Supplementary Fig. 10a). Furthermore, the differentiation process restructured the amino acid and fatty acid profiles of the muscle and fat cellularized microtissues (Supplementary Fig. 10c, d). These findings confirmed that the cell differentiation steps affected the nutritional value of the final cultured meat. Therefore, to produce customer-oriented cultured meat, the texture, flavor, and other nutritional components of cultured meat must be precisely controlled.

Discussion

Cultured meat production necessitates substantial amounts of mature muscle fibers and fat cells³⁷. Various types of seed cells, such as satellite cells, mesenchymal stem cells, fibroblasts, and embryonic stem cells, can be used for meat product cultivation¹². However, the anchorage-dependent growth characteristic of these cell types limits their self-expansion capability³⁸. A scalable approach involves the use of high-surface-area microcarrier scaffolds³⁹. Nevertheless, microcarriers commonly used in the medical field lack edible attributes, because the materials themselves or the preparation processes introduce toxic substances¹⁶. These microcarriers thus can only be used as temporary scaffolds during expansion, and the subsequent cell harvesting steps are thus associated with undesirable costs and cell loss³¹. Using edible microcarriers circumvents the aforementioned issues. Seed cells expanded on these edible microcarriers can be directly transformed into cultured meat by using food structuring techniques, and this obviates the need for additional harvesting steps¹⁶. In the present study, fish gelatin derived from waste materials obtained from fish food processing (fish skin, scales, and bones) was used as the microcarrier matrix, thereby representing a sustainable biomaterial⁴⁰. Through enzymatic crosslinking during cryogenic emulsification, we generated an edible porous microcarrier (Fig. 1a). The abundant RGD sequences within the gelatin microcarriers allowed effective cell adhesion⁴¹ (Fig. 2b). However, the submicron or nanoscale pore network strictly restricted cell growth and diffusion⁴². NaCl introduced during the formation of porous microcarriers induced the development of larger pores. Specifically, NaCl weakened the gelatin network²⁴, enabling ice crystals to grow further and thus introducing macropores into the EPM scaffold (Fig. 1b, d, i). SCs and ASCs with robust differentiation potential were isolated from the white epaxial muscle tissue and adipose tissue of large yellow croakers (Supplementary Fig. 3g–j). In contrast to EPMs with smaller pore sizes, SCs and ASCs presented increased cellular loading within EPMs having larger pore sizes (Fig. 2e, f, Supplementary Fig. 5). Furthermore, seed cells expanded on EPMs also exhibited significant differentiation characteristics (Fig. 2g, h).

Scaling up of anchorage-dependent cell production is associated with a substantial challenge¹⁰. Previous studies have utilized various bulk scaffolds or microgel scaffolds for cultured meat development^8,28,32; however, these traditional static culture or small-scale systems are challenging to scale up⁴³. To achieve scalable cell expansion, we demonstrated that SCs and ASCs successfully transitioned from a 100 mL system to a 4 L system through consecutive expansion (Fig. 3). For industrial-scale production, the consistency of the entire culture process will be crucial to ensure standardized and reproducible continuous production³¹. As illustrated by our results, microcarriers fully covered with cells were dissociated using collagenase, and the seeding protocol enabled even distribution of cells onto additional microcarriers for further production (Supplementary Fig. 8b). The process of scaling up production involved increasing the volume of the bioreactor while supplementing with more microcarriers to achieve scalable surface area. Theoretically, a linear scale-up to a 100-L bioreactor produces 750 g of microtissues per batch. Perfusion culture with an increased microcarrier concentration is an effective strategy for augmenting the cell yield⁴⁴. Compared to traditional monolayer culture, growing cells on scalable EPMs significantly reduces both cell culture media costs and labor expenses, highlighting its potential for commercial viability²⁹. Additionally, the growth kinetics of SCs and ASCs achieved during consecutive expansion are comparable to those observed in the large-scale expansion of human mesenchymal stem cells⁴⁵ and human adipose-derived stem cells⁴⁶, thus providing a more relevant proof of concept for industrial-scale production in cell therapy. Maintaining optimal growth conditions for cells in large systems is complex, and dynamic stirring in the bioreactor ensures uniformity of oxygen, nutrients, and temperature. We also observed that the average doubling time during consecutive expansion in the bioreactor was approximately 33% lower than that in 2D culture plates (Fig. 3c, f). Therefore, this consecutive expansion circumvents labor-intensive 2D culture systems and increases cell culture efficiency by approximately a third. Previous reports have indicated that cells cultured on microcarriers in 3D systems exhibit higher density and proliferation rates compared to 2D monolayer cultures^47,48. Microcarrier-based suspension cultures enable 3D cell growth on the porous surfaces of microcarriers, providing a larger surface area for cell attachment. Dynamic stirring ensures the uniform distribution of oxygen, nutrients, and temperature, thereby enhancing cell growth within the bioreactor. The consecutive expansion process of seed cells triggered major transcriptional changes beneficial for cultured meat production, including spontaneous myogenesis, ECM remodeling, and cell cycle upregulation (Fig. 4a, b). A minute population of initial cells yielded a substantial amount of cultivated meat through marked proliferation. Throughout this process, the heterogeneity within the initial cell population may decelerate cell growth rates, diminish differentiation capacity, and alter cellular population dynamics⁴⁹. Although the two types of seeding cells used exhibited multiplicative and differentiative capabilities, the respective contributions of varied cell abundances to cell behaviors and fates during prolonged culturing must be determined. Methodologies available for seeding cell purification include selective pre-plating³², magnetically activated cell sorting (MACS)⁵⁰, and fluorescence-activated cell sorting (FACS)⁵¹. Sorting techniques (MACS and FACS) allow high-purity cells to be efficiently isolated from extensive cell populations based on cell surface-specific markers⁵². Nevertheless, not all species have accessible antibodies or microbeads. Conversely, expensive commercial antibodies and microbeads may escalate the cost of industrial scaling up of cultured meat production. Selective pre-plating is the simplest, most economical cell purification method⁹. However, adherence times may substantially differ among various species and cell types, thereby necessitating the development of improved selective pre-plating techniques for acquiring high-purity seeding cells.

Appropriate food structuring techniques are required to mimic the tissue-like composition of native fish muscles by using mature muscle and fat microtissues. As a food structuring technique for cultured meat, 3D printing allows the customization of the shape, texture, and nutritional composition of cultured meat according to consumer preferences. By incorporating muscle and fat microtissues into bioinks based on the texture of sliced fish fillets, cultured fish fillets were successfully printed (Fig. 5a, b). The visual and texture analysis revealed striking similarities between the cultured and native fish fillets (Fig. 5b, g). Nutritional profiles of the cultured and native fish fillets displayed notable differences (Fig. 5c). The omega-3 fatty acid content in cultured fish fillets was 59.01% of that in native fish fillets (Supplementary Fig. 9f). In summary, we achieved the initial production of cultured fish fillets derived from cellularized microtissues. To ensure overall consumer acceptability, more diverse cultured meat products must be developed. The texture, nutrition, and flavor attributes of cultured meat products can be further adjusted by modulating the degree of cell differentiation³². Another potential approach involves optimization of the bioink formulation for controlling the mechanical and structural properties of the product, and for incorporating various macro and micronutrients into the bioink to improve the nutritional profile of cultured meat.

This work indicates the potential of EPM-based cell expansion and EPM-based bioink 3D printing platforms in cultured meat production. Further studies are warranted for exploring cost-effective alternatives to serum, which currently proves expensive³⁵. Accordingly, the focus must be shifted to scaling up the production of growth factors, recombinant proteins, and serum substitutes that adhere to industry standards, and thus reduce the cost of culture media. Advancing high-density microcarrier culture systems to enable dense cell cultivation is pivotal for optimizing cost-efficiency and improving production scalability in the cultured meat industry. Additionally, the cultured fish fillets in the present study failed to replicate the dense fiber alignment observed in native fish fillets, thus indicating a disparity in fine structural details between microtissue-derived cultured meat and fish meat. In conclusion, our study overcomes challenges associated with scaling up cell expansion and food structuring technologies and provides a technological pathway for sustainably producing cultured fish meat and enhancing food security.

Methods

Animal treatment and ethics statements

All animal slaughter procedures were performed in accordance with the Animal Care and Use Guidelines of Ocean University of China and were approved by the Institutional Animal Care and Use Committee of Ocean University of China (OUC-AE-2022-010/OUC-AE-2024-071).

Production of EPMs

EPMs were produced using a low-temperature emulsification technique. Specifically, fish gelatin powder (Sigma, Cat# G7041) was dissolved in deionized water at 40 °C with continuous stirring to prepare a 2–4% w/v gelatin aqueous solution. Different concentrations of NaCl (0–5% w/v) and 0.2% (w/v) mTG (Longda Biotechnology, 4000 u Customized model) were mixed and continuously stirred at room temperature for 20 min to prepare the crosslinking agent solution. The gelatin solution and crosslinking agent solution were then mixed in a 4:1 (v/v) ratio to produce the precursor solution (dispersed phase). The continuous phase consisted of corn oil (Macklin, Cat# C805618) containing 5% (v/v) Span 80 (Macklin, Cat# H811141). At room temperature with magnetic stirring at 400 rpm, the 100 mL precursor solution was added dropwise to 400 mL of the continuous phase, thereby forming a W/O emulsion. After the emulsion was stirred for 10 min, it was transferred to a low-temperature reaction vessel at −30 °C to −10 °C with another 2 h of continuous stirring. Once the reaction was completed, the cold w/o emulsion was allowed to stand for 10 min. Following phase separation, the upper layer of corn oil was removed. Subsequently, a 10% ethanol solution was added to the emulsion and stirred at room temperature for 30 min. After allowing the mixture to stand, the supernatant was removed. The process was repeated 2–5 times to eliminate residual NaCl, corn oil, and Span 80. After completing the washing steps, sterile powder was obtained through freeze-drying and high-pressure sterilization.

EPMs were produced at a large scale in a 100-L low-temperature reactor vessel (RAT-100L, Kexing instrument, China). The vessel was equipped with a low-temperature circulation temperature control system (DHJF-4050, Great Wall Technology Industry, and Trade, China) that created a chilled environment by circulating through the jacket of the reactor vessel. Initially, a 16 L solution of 3% gelatin and a 4 L solution of the crosslinking agent containing 3% NaCl (w/v) and 0.2% mTG (w/v) were prepared. At room temperature, 80 L corn oil containing 5% (v/v) Span 80 was introduced into the low-temperature reactor vessel. Then, mechanical stirring at 400 rpm was initiated, while the pre-formed precursor solution containing a prepared 3% (w/v) gelatin solution and crosslinking agent solution was rapidly mixed and added to the low-temperature reactor vessel. After stirring the solutions for 10 min, ethylene glycol pre-cooled to −20 °C in the low-temperature circulation temperature control system was introduced into the reactor vessel, and the mixture was stirred continuously for an additional 2 h. Once the reaction was completed, residual impurities were removed through the washing steps, and a sterile EPM powder was obtained through freeze-drying and high-pressure sterilization.

Young’s modulus of gelatin-mTG hydrogels

To examine the variation in Young’s modulus of gelatin-mTG hydrogels with different NaCl concentrations (0–5%), a 3% gelatin aqueous solution was swiftly mixed with an identical crosslinking agent solution at a 4:1 (v/v) ratio, promptly injected into cylindrical molds (5 mm × 25 mm, Φ×L), and incubated at 25 °C, thereby yielding hydrogels with various crosslinking durations (30–180 min). Once the reaction was completed, the hydrogels were subjected to immediate uniaxial compression testing by using a cylindrical probe (Φ 38 mm) (CT3, Brookfield, USA) to generate stress-strain curves. The Young’s modulus of the hydrogels was determined as the slope within the linear range of the stress-strain curve spanning from 0% to 30% strain.

Visualization of ice crystal growth

A total of 100 μL of the precursor solution comprising fish gelatin and mTG was applied to the surface of a 25 × 75 mm glass coverslip (Sigma, Cat# S8902). A coverslip was then placed under an optical microscope to image the precursor solution. Subsequently, the precursor solution was rapidly cryotransferred from liquid nitrogen onto the front of the coverslip to induce ice crystal formation and growth in the crosslinking system. The entire process was documented under the microscope.

Structural analysis of EPMs

The surface morphology of the EPMs in swelling and dry states (in DMEM/F12 medium (Gibco, Cat# 11320033) supplemented with 10% fetal bovine serum (FBS; Viva cell, Cat# C04001-500)) was observed under an optical microscope (Axio Vert.A1, Carl Zeiss AG, Germany) and a scanning electron microscope (Bruker 5on1, Bruker, USA). The mean particle size and pore size of the microcarriers were determined by measuring the long and short axes of each particle and pore using Image J software (NIH, USA). For each individual EPM sample, two images were recorded, and 20 microcarriers and pores were randomly selected from each image. The final results were obtained using 10 images per group, comprising five independent samples.

Characterization of EPMs

The swelling ratio and porosity of EPMs were calculated based on previous studies¹⁷, and they were defined using the following Eqs. (1) and (2).

$${{\mbox{Swelling}}}\, {{\mbox{ratio}}}\, (\%)=({W_{wet}}-{W_{dry}})/{W_{dry}}\times 100$$

(1)

Where, W_wet and W_dry represent the wet weight and dry weight of the microcarriers, respectively. (W_wet − W_dry) is the weight of the water absorbed by the microcarriers. The dry weight of the carrier was determined after the complete removal of residual moisture through freeze-drying. Each experimental arrangement included 5 independent samples (n = 5).

$${{\mbox{Porosity}}}\, (\%)=({V_0}-{V_1})/({V_T}-{V_1})\times 100$$

(2)

Here, V₀ denotes the initial volume of dehydrated alcohol used to immerse the dried microcarriers, V_T defines the total volume of the system when the microcarriers are immersed in dehydrated alcohol, and V₁ is the volume of residual liquid after the removal of the immersion frame. Each study involved 5 independent samples (n = 5).

The specific surface area was characterized through a nitrogen adsorption-desorption isotherm analysis of 0.1 g samples at 77.3 K. This analysis was performed using a specific surface area and pore size analyzer (MicrotracBEL, BELSORP-miniII, Japan). EPMs were quantified by counting the number of particles in 3 mg EPMs by using the Agilent 8700 LDIR laser infrared imaging system (Agilent, USA).

EPM stability was assessed by calculating the mass remaining during prolonged incubation in cell culture media. Specifically, 0.2 g sterilized EPMs were added to a 125-mL spinner flask (CytoNiche, China) containing 100 mL complete culture medium (DMEM/F12 medium supplemented with 10% FBS) and continuously stirred at 28 °C and 40 rpm. The samples were collected every 7 days, washed three times with distilled water to remove the culture medium, lyophilized, and measured to determine the scaffold mass. The ratio of postincubation mass to initial mass represents the remaining mass.

The Young’s modulus of EPMs in the complete culture medium was calculated using a biological nanoindenter (Optics11 life, Netherlands). Before being tested, EPMs were fragmented using a tissue homogenizer. The resulting fragments were affixed onto poly-L-lysine-coated culture dishes (Nest, Cat# 706001). During the experimental procedure, a 10-μm probe was controlled to approach the fragment at a 20 μm/s velocity until contact was made with the fragment. When a force of >200 kPa was generated between the probe and scaffold fragment, separation occurred. The Young’s modulus of EPMs was calculated from the force-displacement curve obtained throughout the entire process.

The total sodium residue in EPMs was analyzed through inductively coupled plasma optical emission spectroscopy. First, a 1 g sample was digested in a 10 mL solution of nitric-perchloric acid (10:1, v/v). After the digested sample was diluted to an appropriate volume, the spectral intensity of the sodium element was measured using an inductively coupled plasma optical emission spectrometer (iCAP PRO ICP-OES, Thermo Fisher Scientific, USA). The total sodium content of the sample was calculated based on sodium’s standard curve within the 0–20 mg/L range.

Isolation of SCs and ASCs

SCs and ASCs were isolated and purified from large yellow croakers, as previously described⁵³. Juvenile large yellow croakers were obtained from a commercial breeding facility. The fish were sprayed with 75% ethanol and wiped clean. Fresh white epaxial muscle tissue was collected from above the lateral line, minced, and digested with 0.3% type I collagenase (Sigma, Cat# SCR103) at room temperature for 1 h. After filtering the digested tissue sample through a 40-μm cell strainer (Sigma, Cat# CLS431750), the cells were dispersed in DMEM/F12 medium containing 10% FBS, 10 ng/mL bFGF (Sigma, Cat# GF003AF), and 1% penicillin-streptomycin (PS; Biochannel, Cat# BC-CE-007). The cells were then seeded onto a 6-well plate (Nest, Cat# 703002) coated with mouse tail collagen (Solarbio, Cat# C8062) and cultured for 3 h at 28 °C before they were transferred to new wells to remove fibroblasts.

After the visceral organs were removed from the abdominal cavity, adipose tissue was scraped from the abdomen and digested with 0.3% type I collagenase solution at room temperature with agitation for 1 h. ASCs were isolated; passed through a 40-μm cell strainer; and resuspended in DMEM/F12 medium containing 10% FBS, 10 ng/mL bFGF, and 1% PS. The cells were finally seeded onto a 6-well plate coated with mouse tail collagen and cultured for 3 h at 28 °C before they were transferred to new wells to remove fibroblasts.

2D cell culture and differentiation

Standard passaging procedures were followed when the primary cultures reached 80% confluence. A 0.25% trypsin-EDTA solution (Sbjbio, Cat# BC-CE-005) was used for cell detachment. The cells were passaged at a 1:2 ratio twice a week. DMEM/F12 medium supplemented with 10% FBS and 1% PS was used to maintain the passaged cultures.

When the myoblasts of the yellow croakers reached 100% confluency, the culture medium was switched to DMEM/F12 medium containing 2% horse serum, 100 ng/mL VD₃ (Sbjbio, Cat# V8070), and 1% PS to induce myogenic differentiation. Every 3 days, half of the medium was refreshed, and differentiation was maintained for 6 days.

When the adipose-derived stem cells from the yellow croakers attained 100% fusion, the culture medium was changed to DMEM/F12 medium containing 10% horse serum (Gibco, Cat# 26050088), 10 μg/mL insulin (Sigma, Cat# Y0001717), 0.5 μmol/L IBMX (Sigma, Cat# 410957), 0.4 μmol/L dexamethasone (Sigma, Cat# 265005), and 1% PS to induce adipogenic differentiation. Half of the medium was refreshed every 3 days, and differentiation was maintained for 6 days.

SCs and ASCs expansion on EPMs

First, 2 mg/mL EPMs were dispersed in 100 mL cell culture medium and subsequently co-transferred to a sterile 125-mL spinner flask. Both SCs and ASCs were harvested from the T75 culture flasks (Nest, Cat# 708001) and seeded onto the microcarriers in the spinner flask at a density of 5 × 10⁴ cells/mL. The spinner flask was placed on a magnetic stir plate (CytoNiche, China) and programmed to undergo 10 cycles (40 rpm for 5 min, followed by 0 rpm for 30 min) to promote cell adhesion. After the 10 cycles were completed, both SCs and ASCs were stirred to a constant speed of 40 rpm for 13 days. Differentiation medium was added after 7 days of culture. Half of the medium was refreshed every 3 days.

Scalable expansion of SCs and ASCs

Bead-to-bead transfer

The microcarrier scalable cultivation protocol based on bead-to-bead transfer was implemented using the optimized approach, as previously described⁵⁴. In brief, after the expansion phase, the proliferated cellularized microtissues were transferred to the next stage spinner flask/bioreactor. Fresh EPMs and culture medium were introduced into the scaled-up spinner flask/bioreactor, maintaining the EPM concentration at 2 mg/mL. Subsequently, intermittent agitation was performed for 24 h (40 rpm for 5 min, followed by 0 rpm for 2 h, totaling 12 cycles) to facilitate cell bead-to-bead transfer. Then, continuous agitation was performed at 40 rpm for 4 days, with half-volume medium changes being performed every 3 days.

Collagenase digestion

Following expansion, a 3-min static period was maintained to achieve cellularized microtissues sedimentation. The supernatant was removed and replaced with collagenase solution (Longda Biotechnology, Cat# DC1-7) at a microcarrier-to-collagenase ratio of 100 mg:5 mL (1 mg/mL) for EPM digestion. Digestion was performed for 5 min at 28 °C and 40 rpm. Subsequently, the supernatant was discarded through centrifugation at 300 g for 5 min. The cells were resuspended in the cell culture medium and transferred to the next stage spinner flask/bioreactor. Then, fresh EPMs and culture medium were introduced into the scaled-up spinner flask/bioreactor, maintaining the EPM concentration at 2 mg/mL, and 10 cycles of agitation (40 rpm for 5 min, followed by 0 rpm for 30 min) were applied to promote cell adhesion. After the 10 cycles, agitation was switched to a constant speed of 40 rpm for cultivation, with half-volume medium changes performed every 3 days.

The maximum expansion factor (Max. expansion factors) denoted the multiplicative increase in the final cell number (N_max) relative to the initial seeded cell number (N₀) following consecutive expansion. The Max. expansion factor was calculated by the following Eq. (3):

$${\mbox{Max.expansion factor}}=\left(\frac{{N}_{\max }-{N}_{0}}{{N}_{0}}\right)$$

(3)

Quantitative RT-PCR

A 5-mL microtissue suspension was sampled from the spinner flask. RNA was isolated and purified using the Animal Total RNA Isolation Kit (Sangon Biotech, Cat# B518621). RNA concentration and purity were determined using the NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, USA). DNA was synthesized from 1 μg of total RNA by using reverse transcriptase (Beyotime, China). Gene expression was analyzed using the specified primers and BeyoFast SYBR Green qPCR Mix (Beyotime, Cat# D7260), as indicated in the supplementary information (Supplementary Table 1). Gene transcription, which was normalized to β-actin, was evaluated using the 2^−ΔΔCt method.

Live cell Counting, viability assay, and seeding efficiency on EPMs

A 1-mL suspension of cell-containing microtissues was sampled from the spinner flask/bioreactor with thorough mixing to ensure an even sampling for cell growth monitoring. The supernatant was discarded. The microtissues were incubated with 0.1 mg/mL collagenase at 37 °C for 5 min to dissolve the microcarrier scaffolds. Cell density and viability were measured using the Cellaca MX high-throughput automated cell counter (Nexcelom, USA).

The maximum cellular load per EPM was determined by the ratio of the maximum number of cells attained on the EPMs during consecutive expansion to the number of microcarrier particles introduced. Thus, the maximum cellular load per EPM was computed according to Eq. (4).

$${{{\rm{Maximum}}}}\,{{{\rm{cellular}}}}\,{{{\rm{load}}}}\,{{{\rm{per}}}}\,{{{\rm{EPM}}}}=\frac{{{{\rm{Maximum}}}}\,{{{\rm{cell}}}}\,{{{\rm{number}}}}}{{{{\rm{Microcarrier}}}}\,{{{\rm{particles}}}}\,{{{\rm{number}}}}}$$

(4)

Cell seeding efficiency was investigated as the ratio of the number of cells adhering to EPMs to the initial seeded cell number, following a 6 h incubation in a spinner flask, as previously described⁵⁵.The cell seeding efficiency was calculated using the following Eq. (5):

$${{{\rm{Seeding}}}}\,{{{\rm{efficiency}}}}(\%)=\frac{{{{\rm{Adhered}}}}\,{{{\rm{cell}}}}\,{{{\rm{number}}}}}{{{{\rm{Initial}}}}\,{{{\rm{seeded}}}}\,{{{\rm{cell}}}}\,{{{\rm{number}}}}}\times 100$$

(5)

Live-dead staining and imaging

To perform live and dead cell staining, a 0.5-mL microtissue suspension was sampled from the spinner flask. The supernatant was discarded. The microtissues were incubated at 37 °C in PBS (Sbjbio, Cat# P1020), protected from light, and treated with a 1:1000 dilution of calcein AM (Beyotime, Cat# C2012) and propidium iodide dye (Beyotime, Cat# ST511) for 30 min.

Immunofluorescence staining and imaging

The cells were fixed with 4% paraformaldehyde (Sigma, Cat# P1110) at room temperature for 15 min and washed three times with PBS. Cell permeabilization was achieved following treatment with 0.1% Triton X-100 (Sigma, Cat# T8787) at room temperature for 15 min. Blocking was performed using 5% (w/v) bovine serum albumin (Sigma, Cat# SRE0096). The cells were incubated with the primary antibody overnight (Supplementary Table 2). Subsequently, the cells were washed three times with PBS and incubated at room temperature for 1 h with the secondary antibody diluted in the blocking solution (Supplementary Table 2). Then, the cells were washed twice with PBS and counterstained with 1 μg/mL DAPI (Sigma, Cat# D9542) for 20 min. Before being imaged, the cells were washed once with PBS. Imaging was performed under a Nikon confocal microscope by using a C-Apochromat 20 × water immersion objective. The images were analyzed using NIS-Elements Viewer software (Nikon, Japan). The myotube fusion index was determined by calculating the percentage of nuclei within MYH-stained myofibers relative to the total number of nuclei, based on measurements from three randomly selected images. For the validation of the specificity of immunofluorescence staining antibodies, samples established without primary or secondary antibodies were utilized as negative controls, while samples derived from cell lines (Mouse myoblast cells (C2C12), Jurkat T lymphocyte cells (Jurkat) and mouse adipose-derived mesenchymal stem cells (mASCs)) with known positive expression served as positive controls (Supplementary Fig. 4). The positive control cell lines were sourced from Pricella Biotechnology Co., Ltd. and authenticated using STR analysis, with confirmation provided by the analysis certificate supplied at purchase. C2C12 cells were cultured in high-glucose DMEM supplemented with 10% fetal bovine serum and 1% P/S. Jurkat cells were maintained in RPMI-1640 medium with the addition of 10% fetal bovine serum and 1% P/S. mASCs were cultured in a complete medium specifically designed for mASCs (Pricella Biotechnology, Cat# CM-M138). Cultivation of C2C12, Jurkat, and mASCs was maintained at 37 °C in a 5% CO2 environment, and cells used between passages 2–3.

Microscopy of lipid accumulation

The lipid was stained using Nile Red (Sigma, Cat# 19123) neutral lipid stain, followed by counterstaining with 1 μg/mL DAPI. The cells were imaged as described above (immunofluorescence staining and imaging). The adipogenesis index was determined by calculating the percentage of nuclei within lipid droplets relative to the total number of nuclei, based on measurements from three randomly selected images.

RNA sequencing

Library construction and sequencing

Total RNA was extracted using the Trizol reagent (Thermo Fisher Scientific, Cat# 15596018), and its quantity and purity were assessed using a Bioanalyzer 2100 with the RNA 6000 Nano LabChip Kit (Agilent, Cat# 5067-1511). High-quality RNA samples with RIN values of >7.0 were used for library construction. Then, mRNA was purified from the total RNA (Sug) by using Dynabeads Oligo (dT) (Thermo Fisher Scientific, USA). mRNA was fragmented to form short fragments by using divalent cations at elevated temperatures (Magnesium RNA Fragmentation Module (NEB, Cat# e6150), 5–7 min at 94 °C). Subsequently, Super Script^TM II reverse transcriptase (Invitrogen, Cat# 1989649) was used to reverse transcribe and cleave the RNA fragments to generate cDNA. cDNA was then used to synthesize U-labeled second-stranded DNAs by using Escherichia coli DNA polymerase I (NEB, Cat# m0209), RNase H (NEX, Cat# m0297, USA), and dUTP solution (Thermo Fisher Scientific, Cat# R03133). U-labeled second-stranded DNA was processed with the thermolabile UDG enzyme (NEB, Cat# m0280) and PCR amplified for library preparation. Finally, paired-end sequencing (PE150) was performed on the Illumina NovaseqTM 6000 (LC-Bio Technology CO., Ltd., China) to generate 2 × 150-bp reads.

Raw data quality control, alignment to the genome, and analysis of differential expression genes

To obtain high-quality clean reads, Cutadapt (version: cutadapt-1.9) was used to filter the clean reads by implementing the following criteria: (1) removal of adapter-containing reads; (2) elimination of reads containing polyA and polyG sequences; (3) exclusion of reads with >5% unidentified nucleotides (N); and (4) discarding reads with >20% low-quality bases (Q score <20). Subsequently, FastQC (version 0.11.9) was used to assess sequence quality, encompassing Q20, Q30, and GC content of the clean data. All samples’ reads were aligned to the Larimichthys crocea reference genome (L_crocea_2.0) by using HISAT2. Reads were partially filtered based on accompanying quality information before being mapped to the reference genome. HISAT2 allows for multiple alignments per read (up to 20 by default) and permits a maximum of two mismatches during read-to-reference mapping. Additionally, it establishes a potential splice junction database and validates these junctions by comparing previously unmapped reads with the presumed junction database. DESeq2 software was used to analyze intergroup gene differential expression (edgeR for intergroup comparison), where genes with a false discovery rate of <0.05 and an absolute fold change of 2 were considered DEGs. Subsequently, the GO enrichment analysis was performed on the DEGs.

Principal component analysis

The PCA was performed using the princomp function in R.

GO enrichment analysis

The GO enrichment analysis was performed. This analysis identified all significantly enriched GO terms in DEGs compared to the genomic background. Initially, all DEGs were mapped to GO terms in the Gene Ontology database (http://www.geneontology.org/), with the gene count for each term being calculated. Subsequently, through hypergeometric testing, significant enrichment of GO terms in DEGs was determined in comparison to the genomic background.

3D printing process

A bioprinter (Luckybot One, Wiiboox, China) was used for 3D printing. The system consists of a three-axis positioning platform, with automatic movement along the X, Y, and Z axes for controlling the motion of the receiving platform and nozzle. The working space dimension is 20 × 20 × 16 cm. After the sample was thoroughly mixed at 4 °C, the edible ink was loaded into a 60-mL syringe equipped with a tapered needle (inner diameter: 1.2 mm). The complete 3D printing process was conducted at 4 °C. Supplementary Table 3 lists the printing parameters.

Moisture content

The samples were weighed and freeze-dried. The water content was determined according to the following Eq. (6):

$${{{\rm{Moisture}}}}\,{{{\rm{content}}}}(\%)=\frac{{{{\rm{Raw}}}}\,{{{\rm{weight}}}}({{{\rm{g}}}})-{{{\rm{Lyophilized}}}}\,{{{\rm{weight}}}}\, ({{{\rm{g}}}})}{{{{\rm{Raw}}}}\,{{{\rm{weight}}}}\, ({{{\rm{g}}}})}$$

(6)

Cooking loss

Cultured fish fillet, native fish fillet, and blank (prepared by using cell-free EPMs) were pan-fried in a skillet at 150 °C for 5 min, followed by an additional 1 min of frying at 250 °C, and subsequently cooling to 25 °C. The cooking loss (CL) was determined as the weight of the sample expressed as a percentage and calculated according to the following Eq. (7):

$${{{\rm{CL}}}}\, (\%)=\frac{{{{\rm{Raw}}}}\,{{{\rm{weight}}}}\, ({{{\rm{g}}}})-{{{\rm{Cooked}}}}\,{{{\rm{weight}}}}\, ({{{\rm{g}}}})}{{{{\rm{Raw}}}}\,{{{\rm{weight}}}}\, ({{{\rm{g}}}})}$$

(7)

Textural properties

The samples were subjected to the following analyses by using a texture analyzer (CT3, Brookfield, USA). They were compressed twice at a constant speed of 50 mm/min to 50% of their initial height against a stainless steel plate (diameter: 11.5 cm). Load sensors of 10 or 50 N were used. The cultured fish fillet, native fish fillet, and blank had a diameter of 1.8–2 cm and a thickness of 2 cm. Texture parameters, including hardness, adhesiveness, cohesiveness, and chewiness at 50% deformation, were calculated using TexturePro CT software version v1.9 (Brookfield, USA).

Amino acid profile

The method used for determining the amino acid profile was slightly modified⁵⁶. Approximately 25 mg of accurately weighed sample was placed in an ampoule. Then, 10 mL of 6 mol/L hydrochloric acid (Aladdin, Cat# H399545) containing 5% mercaptoethanol (Sigma, Cat# M6250) was added to the ampoule. The ampoule was evacuated, filled with nitrogen gas, and digested in a 110 °C oven for 22 h. After the sample was digested, it was diluted to a final volume of 25 mL with ultrapure water and allowed to stand. Then, 1 mL of the diluted sample was dried with nitrogen gas and resuspended with 1 ml of 0.02 M hydrochloric acid for sample recovery. The resuspended solution was vortexed, filtered through a 0.22-μm filter, and prepared for further analysis. The total amino acid content of the sample was directly measured using an automated amino acid analyzer (Hitachi L-8900, Japan).

Instrumental color measurement

The color was measured using a Minolta Chroma Meter CR-300 colorimeter (Minolta, USA). This instrument was calibrated using a standard white porcelain tile under D65 illumination. Results (n = 3 independent experiments) were expressed in the CIELab system as L* (lightness), a* (redness), and b* (yellowness), and quantitatively calculated based on the Minolta spectral curve.

Nutritional evaluation

Nutrient composition of the cellularized microtissues, cultured fish fillets and native fish fillets were analyzed in a certified laboratory (Qingdao Zhongyi Testing Co., Ltd).

The protein and lipid production per EPM was assessed by correlating the protein and lipid content in mature cells to the number of EPM particles introduced. To obtain scaffold-free cellular biomass, cells were dissociated from a 4-L bioreactor system using 0.1 mg/mL collagenase and subsequently washed twice with PBS. The resulting scaffold-free cellular biomass was analyzed for protein and lipid content at a certified laboratory (Qingdao Zhongyi Testing Co., Ltd). Consequently, the protein and lipid production per EPM was calculated according to Eq. (8).

$${{{\rm{Protein}}}}\,{{{\rm{and}}}}\,{{{\rm{lipid}}}}\,{{{\rm{production}}}}\,{{{\rm{per}}}}\,{{{\rm{EPM}}}}\, ({{{\rm{ng}}}})=\frac{{{{\rm{Protein}}}}\,{{{\rm{and}}}}\,{{{\rm{lipid}}}}\,{{{\rm{content}}}}\, ({{{\rm{ng}}}})}{{{{\rm{Microcarrier}}}}\,{{{\rm{particles}}}}\,{{{\rm{number}}}}}$$

(8)

Fatty acid profile

One gram of the freeze-dried sample was mixed with 10 mL of n-hexane for 1 min, ultrasonically extracted for 30 min in an ice bath, and centrifuged at 6790 g for 10 min at 4 °C. The n-hexane layer was collected. This complete extraction process was repeated three times. The combined extract was dried using nitrogen gas to obtain the lipid fraction. Methylation and fatty acid profiling were performed using the method described in the Chinese national standard GB 5009.168-2016. Fatty acid methyl esters were analyzed using a gas chromatography system (8890 GC System, Agilent Technologies, USA) equipped with a fused silica capillary column (100 m × 0.25 mm, 0.20 μm film thickness) and a flame ionization detector. The fatty acid profile was determined using the area normalization method.

Flavor analysis

Volatile organic compounds (VOCs) in the samples were analyzed using a gas chromatography-ion mobility spectrometry (GC-IMS) instrument (FlavourSpec G.A.S., Germany) with slight modifications. Briefly, a 5 g sample was placed in separate 20-mL headspace vials and incubated at 500 rpm (60 °C, 10 min). Then, after incubation, 1 mL of headspace gas was injected into the injector heated at 85 °C by using a syringe in a non-split mode. Subsequently, the sample was separated using the FS-SE-54-CB-1 capillary column (15 m × 0.53 mm, CS Chromatographic Service GmbH, Germany). Nitrogen (purity: 99.99%) was used as the carrier gas, and the following programmed flow rates were used: 2 mL/min for 2 min, followed by 10 mL/min for 10 min. After the analytes were separated on the column (60 °C), they were ionized in the IMS ionization chamber at a constant voltage of 5 kV and a drift gas flow rate of 150 mL/min (nitrogen) at 45 °C. The analysis was performed with three independent replicates. The retention indices (RIs) of VOCs were determined using normal ketones (C4–C9) as external standards. A calibration curve was constructed by correlating the RIs (known) of normal ketones (C4–C9) with their retention times. For unknown substances, the RI was calculated based on the retention time. The RIs and drift times of the VOCs were compared to standards in the GC-IMS database to identify the VOCs. The GC-IMS instrument’s laboratory analysis viewer, reporter, and gallery plots (VOCal version 0.4.03, FlavourSpec G.A.S., Germany) were used to construct 2D terrain plots and gallery plots of the VOCs.

Statistical analysis

Statistical analyses were conducted using GraphPad Prism 9 software. Data depicted in the figures are expressed as the mean ± SD, unless otherwise indicated. Comparative assessments between two experimental groups were conducted using two-tailed Student’s t-tests, while one-way analysis of variance (ANOVA) was employed to discern differences among multiple groups, complemented by Tukey’s post hoc test for pairwise comparisons. The experiments were independently replicated a minimum of three times, with exceptions duly noted. Sample sizes (n) are delineated within the figure captions, with the corresponding raw data presented in the accompanying source data file. Statistical differences are quantified by p value, accessible for each dataset upon analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.