Modification of nanocellulose filler via amination and subsequent grafting of gallic acid for enhancing the performance of biodegradable soy protein films in banana packaging

Abstract

The excessive use of petroleum-based packaging necessitates the need for sustainable, biodegradable alternatives. This study developed a high-performance active packaging film by incorporating chemically modified cellulose nanocrystals (CNCs) into a soy protein isolate (SPI) matrix. CNCs were functionalized via silanization with 3-aminopropyltriethoxysilane (APTES) followed by covalent gallic acid grafting. Compared to glycerol-plasticized SPI, the composite films showed significantly enhanced mechanical strength (237.3% increase in tensile strength, from 2.17 to 7.32 MPa), improved water vapor barrier (30.16% reduction in permeability), nearly 100% UV blocking at 200–350 nm, antioxidant activity (12.4% DPPH scavenging), and antibacterial properties. In a 7-day banana preservation test, the optimized film effectively delayed weight loss, softening, and browning while remaining biodegradable, showing visible degradation within 20 days in soil. This work demonstrates a feasible interfacial design strategy for developing multifunctional SPI-based nanocomposite films with high potential for fresh-keeping packaging of perishable fruits such as bananas.

Graphical abstract

Highlights

• •A biodegradable film based on soy protein with enhanced performance was prepared.
• •Silanized bamboo CNCs-GA boosted SPI film strength by 237% (from 2.17 to 7.32 MPa).
• •SPI/CNCs composite film shows visible degradation after 20 days of soil burial.
• •It provided nearly 100% UV absorption at 200–350 nm.
• •SPI/CNCs composite film significantly inhibits the browning process of bananas during the 7-day storage period.

1. Introduction

The extensive use of petroleum-based polymers, including polyethylene, polypropylene, and polystyrene, has led to severe “white pollution” (Tian et al., 2018), posing increasingly serious threats to both human health and ecosystems. To address this issue, there is an urgent need to advance the development of renewable and biodegradable biopolymer materials (Song et al., 2017; Xu, Yang, et al., 2020). Natural polymer films are typically fabricated from biopolymer-based materials (e.g., polysaccharides (Gao et al., 2019), lipids (Sartori & Menegalli, 2016), and proteins (Han et al., 2023; Zareie et al., 2020)) by incorporating plasticizers, crosslinking agents, and other additives, which enhance their performance and functionality through intermolecular interactions (Hassan et al., 2018).

Among various biopolymers, soy protein isolate (SPI) has garnered significant attention due to its wide availability, low cost, environmental friendliness, and favorable transparency and ductility (Mostafa et al., 2023). The SPI molecules contain abundant amino acid side chains bearing polar groups, such as hydroxyl, amino, and carboxyl groups. Its stable structure is maintained primarily through hydrogen bonds, ionic interactions, disulfide linkages, hydrophobic interactions, and van der Waals forces, which contribute to the excellent film-forming properties (Gu et al., 2023; Tian et al., 2018). However, compared to conventional petroleum-based polymers, such as polyvinyl chloride (PVC), polyethylene terephthalate (PET) and polypropylene (PP), pure SPI films still face considerable challenges in practical applications due to their poor mechanical strength, weak barrier properties, and high sensitivity to moisture (Gu et al., 2023; Pawde et al., 2025), which severely restricts their applicability and protective lifespan. These limitations, largely attributed to the strong hydrophilicity imparted by the high content of polar amino acids in SPI molecular chains (Hassan et al., 2018), severely restrict their applicability and protective lifespan, particularly in meeting the requirements for packaging materials during transportation and storage. Therefore, a modification of SPI films is necessary to enhance their properties.

Various approaches, including physical, chemical, and enzymatic treatments, have been explored to optimize the structure and improve the overall performance of SPI films (Benbettaïeb et al., 2016). The chemical modification involves the use of cross-linking agents such as glutaraldehyde and ethylene glycol to introduce stable covalent bonds between protein molecules, significantly enhancing mechanical strength and water resistance. However, the potential cytotoxicity of such cross-linkers limits their application in food-contact materials (Cheng et al., 2025) Enzymatic modification, typically achieved using transglutaminase, is often time-consuming, costly, and operationally complex (Benbettaïeb et al., 2016). In contrast, physical blending with polysaccharides (e.g., chitosan, carboxymethyl cellulose) (Pan et al., 2014) or polyol plasticizers (e.g., glycerol) (Kumar & Zhang, 2009) is considered a safer, simpler, and more cost-effective modification strategy. Such methods enhance the intermolecular physical entanglement and hydrogen bonding, and improve effectively film flexibility and stability. Cellulose nanocrystals (CNCs), recognized as a promising reinforcing material, have been widely investigated (Douglass et al., 2017). Building upon physical blending, the introduction of nano-sized reinforcing phases to establish stronger interfacial interactions and form a dense composite network has emerged as a promising frontier strategy.

Currently, CNCs are used as nanofillers due to their excellent mechanical strength, environmental friendliness, and high aspect ratio. Compared with conventional inorganic fillers (such as carbon nanotubes and metal nanoparticles), CNCs are the most abundant renewable materials on Earth, sourced from widely available natural materials including wood, bamboo, and agricultural waste (Phuong et al., 2022). They offer several advantages for bio-composites: they maintain the biodegradability of the polymer matrix, exhibit superior dispersibility and stronger chemical affinity within polymer matrices for enhanced reinforcement efficiency, and possess surfaces rich in hydroxyl groups that facilitate further functionalization (Abdellaoui & Bouhfid, 2020). In protein-based composite films, CNCs can enhance hydrophobic interactions, induce conformational changes in the protein matrix, and establish hydrogen-bonding networks, working synergistically to form a more robust three-dimensional film structure (Dissanayake et al., 2023). However, because natural cellulose is inherently hydrophilic and its surface is rich in hydroxyl groups, direct blending with SPI may lead to uneven dispersion and poor interfacial compatibility, resulting in inferior film properties (Raza & Abu-Jdayil, 2022). To address this issue, surface modification of CNCs can be performed by substituting some of the surface hydroxyl groups to reduce or alter their polarity. Shi et al. successfully prepared carboxylated nanocellulose using a deep eutectic solvent, achieving an aspect ratio as high as 2500 and a carboxyl group content of 1.5 mmol/g, demonstrating excellent stability (Shi et al., 2024). Current surface modification methods include acetylation (Xu, Wu et al., 2020), carboxymethylation (Wei et al., 2021), cationization (Littunen et al., 2016), oxidation (Isogai & Zhou, 2019), and silylation (Rana et al., 2021). Among these, silanization has been extensively studied owing to its high reaction efficiency, excellent modification outcomes, and mild reaction conditions, particularly in aqueous-phase systems (Ludovici et al., 2022). 3-aminopropyltriethoxysilane (APTES), known for its low cost, low toxicity, high reactivity, and ability to introduce amino functional groups, is a suitable silanization modifier for CNCs (Amit et al., 2024). In addition to surface modification, cross-linking agents are crucial for constructing high-performance bio-based composites. To avoid the environmental issues associated with the toxicity of traditional aldehyde-based cross-linkers, green alternatives are preferable. Gallic acid (GA), an environmentally friendly, low-cost natural polyphenol with demonstrated antioxidant, antimicrobial, and anticancer properties (Badhani et al., 2015), serves this purpose well. As a cross-linking agent, GA can form a tighter cross-linked network with the matrix, improving the mechanical strength, antioxidant activity, and antimicrobial properties of SPI (Huang et al., 2021). These modification methods offer a promising approach for preparation of protein-based composite films with enhanced performance.

Indeed, silane-modified CNCs can significantly enhance the mechanical and barrier properties of composite films (Gomri et al., 2022), while the incorporation of polyphenols such as tea or date leaf extracts can impart antioxidant, antibacterial, and UV-blocking functions to SPI films (Mostafa et al., 2023; Yong et al., 2024). This study proposes a synergistic strategy combining interfacial coupling and active grafting: APTES is used to improve CNCs dispersion and interfacial adhesion within the SPI matrix, and GA is covalently grafted to reinforce the network while introducing antioxidant and antibacterial functionalities. Based on this design, we prepared an SPI-based composite film by blending SPI with CNCs modified using GA as a crosslinker, activated with 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC) and N-hydroxysuccinimide (NHS), and functionalized with APTES as a silane modifier. This approach offers a method to fabricate biodegradable, safe, protein-based nanocomposite films with improved performance. The films were evaluated for mechanical properties, UV–visible light barrier ability, water resistance, antioxidant and antibacterial activity to explore their potential in preserving fresh products such as bananas.

2. Materials and methods

2.1. Materials

Bamboo pulp from moso bamboo (Phyllostachys edulis) was utilized as the primary cellulose source. Soy protein isolate (SPI, 98% protein content) was supplied by Ruikang Biotechnology Co., Ltd. Sodium hydroxide (NaOH, AR), glycerol (Gl, AR), sulfuric acid (H₂SO₄, AR), and glacial acetic acid (C₂H₄O₂, AR) were purchased from Sinopharm Chemical Reagent Co., Ltd. 3-aminopropyltriethoxysilane (APTES, 99%), 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (EDC, 98.5%), and N-hydroxysuccinimide (NHS, 98%) were obtained from Shanghai Macklin Biochemical Co., Ltd. Gallic acid (GA, 99%) was obtained from Shanghai Aladdin Biochemical Technology Co., Ltd. The distilled water (DW) was obtained from the laboratory.

2.2. Preparation of nanocellulose and subsequent surface modification

Bamboo pulp (10 g) was mixed with 200 mL of H₂SO₄ (45 wt%) in a three-necked flask equipped with a reflux condenser to prevent evaporation. The reaction was maintained at 45 °C for 30 min, followed by immediate dilution with 10-fold DW to terminate the hydrolysis (Xiao et al., 2021). The resulting mixture was vacuum filtered through a Buchner funnel and washed with DW until the pH increased from an initial value of 1.2 to neutrality. The obtained CNC sample was collected, concentration was determined, and finally stored at 4 °C.

For surface modification, 12 mL APTES was mixed with 300 mL DW, and 3 g CNC was added. The pH was adjusted from 11.2 to 5.0 using glacial acetic acid, and the mixture was stirred at room temperature for 2 h. The solution was then vacuum-filtered to remove unreacted silane coupling agent and dried at 80 °C for 1 h. The silanized CNC, designated as CNC-APTES, was re-dispersed in DW to prepare a 2 wt% suspension. A few drops of glacial acetic acid were then added to adjust the pH from an initial value of 6.8 to 5.0. For subsequent modification, 5 g GA was dissolved in DW at 70 °C, and pH was adjusted from 2.3 to 5.0 using NaOH. Then, 5.8 g EDC was added to the GA solution with constant stirring for 10 min, followed by addition of 3.4 g NHS and continued stirring for 2 h. The CNC-APTES suspension was combined with the activated GA solution and stirred at room temperature for 12 h (Huang et al., 2024; Salih et al., 2024). The final product (designated as CNCN) was collected by vacuum filtration and stored at 4 °C.

2.3. Preparation of different SPI derived nanocomposite films

CNC and CNCN suspensions were homogenized at 10,000 rpm for 10 min to achieve uniform dispersions. For film preparation, 6 g of soy protein isolate (SPI) were dissolved in 100 mL of deionized water, with the addition of a 10 wt% sodium hydroxide solution to facilitate dissolution (Xu et al., 2022). After stirring for 4 h at room temperature, the pH was adjusted from an initial value of 6.7 to 9.2. Then, 2.5 g of glycerol (corresponding to 41.7 wt% relative to SPI solid) was added as a plasticizer, and the solution was subsequently stirred at 85 °C for 30 min. At this stage, the solid content of the SPI-glycerol solution was approximately 7.8 wt%. Upon cooling to room temperature, the CNC or CNCN suspension was incorporated into the SPI mixture under stirring for homogenization. The resulting film-forming solution was then cast onto polytetrafluoroethylene molds and air-dried at room temperature for 48 h. The dried film was carefully peeled off and transferred to a constant temperature and humidity chamber maintained at 25 °C and 45% relative humidity (RH) for over 24 h prior to further use.

The prepared films were designated as SPIC5, SPIC10, SPIC15, SPIC20 (SPI/CNC series), based on dry CNC loading of 5, 10, 15, and 20 wt% relative to the dry weight of SPI solid, respectively). Accordingly, SPICN5, SPICN10, SPICN15, and SPICN20 (SPI/CNCN series) refer to films containing CNCN at the corresponding dry-weight percentages. A control film (SPI) without nanocellulose addition was prepared under identical conditions. The detailed formulation is provided in Table 1.

Table 1.

Formulations of the prepared composite films.

Samples	SPI solution/g	Gl/g	CNC/g	CNCN/g
SPI	50	1.25	0	0
SPIC5	50	1.25	0.15	0
SPIC10	50	1.25	0.30	0
SPIC15	50	1.25	0.45	0
SPIC20	50	1.25	0.60	0
SPICN5	50	1.25	0	0.15
SPICN10	50	1.25	0	0.30
SPICN15	50	1.25	0	0.45
SPICN20	50	1.25	0	0.60

2.4. Fourier transform infrared spectroscopy (FTIR)

The interactions among the components in APTES, GA, CNC, CNC-APTES and CNCN, as well as SPI, SPI/CNC and SPI/CNCN films were investigated using a Fourier transform infrared spectrometer (FTIR) (Thermo Nicolet IS5) in attenuated total reflection (ATR) mode. Each sample was scanned 32 times in the wavenumber range of 400–4000 cm⁻¹ with a resolution of 4 cm⁻¹.

2.5. X-ray diffraction (XRD)

The crystallinity of CNC, CNC-APTES, CNCN, as well as SPI, SPI/CNC and SPI/CNCN were determined using an X-ray diffractometer (Bruker D8 Advance, USA). The instrument used CuKα as the radiation source (30 kV, 30 mA).

XRD patterns of CNC and modified CNC samples were recorded in the 2θ range of 5° to 90°. The crystallinity index (CrI) was determined based on an empirical X-ray diffraction method (Segal et al., 1959), where I₀₀₂ and I_am represent the diffraction peak intensity of the crystalline (002) plane and the amorphous background, respectively. The CrI values were calculated using Eq. (1):

$$\mathit{CrI}\left(\%\right)=\frac{I_{002}-I_{\mathit{am}}}{I_{002}}\times 100$$ (1)

2.6. X-ray photoelectron spectroscopy (XPS)

The surface elemental composition was analyzed via X-ray photoelectron spectroscopy (XPS) (Thermo Scientific K-Alpha, USA). A monochromated Al Kα X-ray source (photon energy = 1486.6 eV) was used for excitation. Survey spectra were collected with an energy step size of 1.000 eV and a pass energy of 117.4 eV. High-resolution spectra were acquired using an energy step size of 0.05 eV and a pass energy of 46.95 eV.

2.7. Mechanical properties

The mechanical properties of the films were evaluated using a universal testing machine (Model SUST 5569). The samples were cut into rectangular strips with dimensions of 150 mm in length and 15 mm in width. Tensile tests were conducted at room temperature (25 °C) and 45% RH, with a crosshead speed of 50 mm/min. Five replicates were performed for each sample. A one-way analysis of variance (ANOVA) was employed to evaluate the effects of CNC and CNCN on the mechanical properties of soy protein films, with film type as the factor and tensile strength, elongation at break, and elastic modulus as dependent variables. The experiment was performed in triplicate.

2.8. Film thickness, moisture content and water solubility

The film thickness was measured at 10 randomly selected points on each film sample using an Everte digital thickness gauge, following a random sampling method. For moisture content (MC) and water solubility (WS), the films were cut into 2 cm × 2 cm squares, with the initial mass recorded as (m₀, g). After drying in an oven at 103 ± 2 °C until constant weight, the mass was accurately measured (m₁, g). The dried film samples were then placed in Petri dishes containing 30 mL of DW for 24 h. The solution was filtered to remove insoluble residues, and the remaining material was dried again at 103 ± 2 °C to constant weight (m₂, g). With three samples per group, the experiment was independently repeated three times. MC and WS were calculated using Eqs. (2), (3), respectively:

$$\mathit{MC}\left(\%\right)=\frac{m_0-m_1}{m_0}\times 100$$ (2)

$$\mathit{WS}\left(\%\right)=\frac{m_1-m_2}{m_1}\times 100$$ (3)

2.9. Water vapor permeability (WVP)

WVP was studied using the permeation cup method (Peng et al., 2013) with the same modifications. Films were cut into circles with a diameter of 40 mm and used to seal cups (100 mm depth and 30 mm diameter) containing 3 g of completely dried anhydrous calcium chloride (0% RH). The sealed cups were then placed in a desiccator containing 1 L of DW to establish a 100% RH gradient. All tests were conducted at room temperature (25 ± 2 °C). The weight of the cups was recorded every 12 h over a period of 3 consecutive days, with the experiment repeated in triplicate. The WVP was calculated using Eq. (4):

$$\mathit{WVP}=\frac{∆m\times x}{t\times S\times ∆P}$$ (4)

where Δm refers to the increase in weight of the sample tube (g), x reveals the film thickness (mm), t points out the duration for the increase in weight of the test tube (s), S explains the permeable area of the film sample (m²), and ΔP is the water vapor partial pressure difference across the film (Pa, taken as 3169 Pa, the saturated vapor pressure at 25 °C under 0–100% RH gradient).

2.10. Surface morphology analysis

Transmission electron microscopy (TEM) was employed to examine the general morphology of the CNC, CNC-APTES and CNCN. The suspension was deposited onto carbon-coated copper grids, negatively stained with 2% uranyl acetate, and analyzed using a transmission electron microscope (Hitachi HT7700, Japan) operated at an accelerating voltage of 80 kV. The cross-sectional morphology of the films was observed using a scanning electron microscope (SEM) (Zeiss Sigma 300, Germany). Each sample was cryo-fractured in liquid nitrogen, and the fractured cross-sections were sputter-coated with gold prior to observation.

2.11. Thermogravimetric analysis (TGA)

The thermal degradation behavior of CNC, CNC-APTES and CNCN, as well as SPI, SPI/CNC and SPI/CNCN films were evaluated using a thermogravimetric analyzer (NETZSCH TG 209F3). About 5–10 mg of the samples were heated at a rate of 10 °C/min from 30 °C to 600 °C in a nitrogen atmosphere.

2.12. UV–vis light barrier properties

The UV–visible light transmittance of the films was measured at wavelengths ranging from 200 to 800 nm using a UV–visible spectrophotometer (A260, Mettler Toledo, Switzerland). Samples were cut into 1 cm × 4 cm strips and measured in triplicate. The opacity was calculated according to Eq. (5):

$$\text{Opacity}=\frac{{\mathit{Abs}}_{600}}{x}$$ (5)

where Abs₆₀₀ is the absorbance at 600 nm and x is the film thickness (mm). Three replicates were performed for each sample, and the average value was reported.

2.13. Antioxidant and antibacterial activity

The antioxidant activity was determined by the DPPH radical scavenging assay (Bersuder et al., 1998) with slight modifications. Briefly, 0.1 g of film sample was immersed in 20 mL of DW for 30 min to form a solution. Then, 4 mL of a 0.15 mmol/L 2,2-diphenyl-1-picrylhydrazyl (DPPH) ethanol solution was mixed with 1 mL of the prepared aqueous solution. The mixture was allowed to react in the dark at room temperature for 30 min. The reduction of DPPH radicals was quantified by measuring the absorbance at 517 nm using a UV–visible spectrophotometer. The radical scavenging activity (RSA) was determined using Eq. (6):

$$\mathit{RSA}\left(\%\right)=\left(1-\frac{A_1}{A_0}\right)\times 100$$ (6)

where A₀ is the absorbance of the control sample at 517 nm and A₁ is the absorbance of the film sample solution at 517 nm.

The antibacterial activity of the films against Escherichia coli (E. coli) (ATCC 25922, Gram-negative) and Staphylococcus aureus (S. aureus) (ATCC 29213, Gram-positive) was evaluated following the method described by Venkatesan (Venkatesan et al., 2024), with minor modifications. A single colony was inoculated into nutrient broth and incubated at 37 °C with shaking at 200 rpm for 12 h. The bacterial suspension was then diluted with sterile PBS to 10⁶ CFU/mL. After UV sterilization for 30 min, 100 mg of film sample was mixed with 1 mL of the bacterial suspension and incubated at 37 °C with shaking at 120 rpm for 24 h. The co-culture was serially diluted 10-fold with sterile PBS, and 100 μL of each dilution was spread onto LB agar plates. Following incubation at 37 °C for 24 h, the bacterial growth was visually assessed. Colony density changes and morphological abnormalities were compared among the blank control, SPI, SPI/CNC and SPI/CNCN groups to determine the antibacterial performance. All experiments described above were repeated three times.

2.14. Banana preservation performance test

The preservation performance of the films on bananas was evaluated following the method described by Lee (Lee et al., 2023), with minor modifications. Detailed information is provided in the Supplementary Materials.

2.15. Soil burial degradation test

A soil burial induced degradation test was employed to evaluate the biodegradability of SPI, SPI/CNC, and SPI/CNCN films (Gan et al., 2021). With three replicates for each type, the films were first cut into rectangular specimens measuring 50 mm × 100 mm and buried in outdoor soil at a longitude of 102.73, latitude of 25.03, and a depth of 10–15 cm. The soil pH was 8.13 with a typical moisture content of 20.6%. Each film sample was enclosed in a plastic mesh. Samples were taken and photographed every six days to evaluate the degree of biodegradation of the films. The complete experiment was independently repeated three times.

2.16. Statistical analysis

All experiments were performed in multiple replicates. Data are presented as mean ± standard deviation. Statistical analysis was done by one-way ANOVA (SPSS Statistics; version 27). The Duncan test was used to compare means. Mean values were considered statistically significant when P$\le$0.05.

3. Results and discussion

3.1. Preparation and surface modification of CNCs

The FTIR spectra in Fig. 1a show that the absorption peak at 2928 cm⁻¹ can be attributed to the stretching vibration of CH₂ in the propyl chain of APTES. The peak at 1072 cm⁻¹ corresponds to the stretching vibration of Si—O—C or Si—O—Si bonds. The enhanced absorption at around 2903 cm⁻¹ in CNC-APTES indicates the successful introduction of organic silane chains (—CH₂—) onto the CNC surface. In the spectrum of GA, the peaks at 3494 cm⁻¹ and 3281 cm⁻¹ are assigned to the O—H stretching vibration of phenolic hydroxyl groups and the broad O—H stretching of carboxylic acid dimers, respectively. A gradual decrease in the intensity of the broad O—H absorption band around 3350 cm⁻¹ is observed from CNC to CNC-APTES and further to CNCN, which can be attributed to the consumption of surface hydroxyl groups during the grafting process (Qin et al., 2019). The appearance of a new peak at 1563 cm⁻¹ in CNCN serves as the most direct evidence for the formation of amide bonds and is clearly identified as the amide II band (Huang et al., 2024). This band originates from the coupling of N—H bending and C—N stretching vibrations in the amide linkage, confirming the successful occurrence of the amidation reaction. The enhanced peak at 821 cm⁻¹, assigned to the out-of-plane bending vibration of C—H in a 1,2,4-trisubstituted benzene ring, verifies the presence of the gallic acid aromatic framework in the final product.

Fig. 1.

(a) FTIR spectra of APTES, GA, CNC, CNC-APTES and CNCN; (b) XPS full survey spectra of CNC, CNC-APTES and CNCN, (c) C1s of CNC, (d) O1s of CNC, (e) C1s of CNCN, (f) O1s of CNCN, (g) N1s of CNCN, (h) Si2p of CNCN; (i) XRD diffractograms of CNC, CNC-APTES and CNCN; (j) TEM images of CNC, CNC-APTES and CNCN.

Fig. 1b to Fig. 1h depict the XPS spectra of C 1s, O 1s, N 1s, and Si 2p for CNC and CNCN, while the corresponding spectra for CNC-APTES are shown in Fig. S1. As illustrated in Fig. 1c, compared with CNC, the carbon and oxygen signals of CNC-APTES and CNCN show continuous enhancement. Additionally, characteristic nitrogen and silicon signals, which are absent in pristine CNC, are detected, confirming the successful immobilization of APTES. In the C 1s spectrum of CNC-APTES, the C—C component increases, likely due to the introduction of the propyl chains from APTES. For CNCN, the C—C signal is further enhanced owing to the aromatic rings of GA, and a new peak appears at 288.05 eV, assigned to the newly formed amide bond (N—C O). The N 1s spectra directly track the amidation reaction: CNC-APTES shows a primary amine (—NH₂) at 399.74 eV and a protonated ammonium group at 401.96 eV, resulting from reaction with atmospheric CO₂ or H₂O (Magalhaes et al., 2002). In contrast, CNCN exhibits a new strong peak at 400.16 eV, identified as the N—C O peak, which provides a direct evidence of covalent bond formation (Huang et al., 2024). The O 1s spectra reveal a significant increase in the O C component intensity from CNC-APTES to CNCN, originating from the carboxyl and phenolic hydroxyl groups of the GA. Meanwhile, in the Si 2p spectra, each chemical state is represented by a pair of spin-orbit split doublets (Dietrich et al., 2016). The signal at a Si 2p_3/2 binding energy of 101.94 eV is attributed to fully condensed siloxane networks (Si—O—Si), indicating that APTES has formed stable covalent bonds on the CNC surface, while the signal at 103.15 eV corresponds to incompletely condensed silanol groups (Si-OH). The stability of the Si 2p signals in both CNC-APTES and CNCN is because the chemical state of the siloxane network remained unaltered throughout the two-step reaction. These conclusions, consistent with the FTIR results, are further supported by the TGA data in Fig. S2, collectively demonstrating the successful synthesis of CNCN.

To investigate whether there was any effect of the surface chemical modification on the crystal structure of CNC, XRD analysis was conducted on CNC, CNC-APTES, and CNCN. As shown in Fig. 1i, all samples exhibit characteristic diffraction patterns of cellulose I, which is the dominant crystalline form in a native cellulose. The persistence of the cellulose I pattern in all samples indicates that the modifications on the CNC were predominantly surface-level, without altering its bulk crystalline structure. However, compared with CNC, the diffraction patterns of the modified samples show significant changes: the intensity of their characteristic diffraction peaks, particularly the main (2 0 0) peak at 22.6° and the (0 0 4) peak at 34.6° representing the crystalline regions (Ojogbo et al., 2025), decreases markedly. Additionally, the diffraction peak at approximately 16.22°, which can be assigned to the overlapping of the (1 −1 0) and (1 1 0) planes of cellulose, broadens (Yang et al., 2017). These observations indicate a gradual decrease in crystallinity from CNC to CNC-APTES and further to CNCN, which is primarily attributed to the silanization reaction with APTES and the subsequent EDC/NHS-mediated grafting of GA. This caused progressive disruption and replacement of the original hydroxyl groups and hydrogen-bonding network on the CNC surface, introducing amorphous components into the crystal structure and re-orientation of the hydrogen bonding network. This effectively confirms the successful implementation of surface chemical modification.

The morphological features of CNC and CNCN are displayed in Fig. 1j. TEM images reveal that the inherent needle-like morphology of CNC remained intact after both modification steps. However, while maintaining the original structure, CNC-APTES and CNCN exhibited low-contrast regions with blurred edges and bright halos surrounding and encapsulating the fibers. The sporadic low-contrast spots observed in CNC-APTES can be attributed to small amounts of siloxane oligomers generated during the silanization process, diluting the hydrogen bonding framework via expansion. In the case of CNCN, more densely distributed low-contrast regions formed a continuous encapsulating layer, reflecting the amorphous organic molecular layer resulting from the successful grafting of GA via amide bonds. This conclusion is consistent with the XRD results, indicating that the two-step surface chemical modification successfully introduced organic molecules, while perfectly preserving the inherent morphology and crystal structure of CNC.

3.2. Chemical structure and potential interactions of the composite films

The SPI/CNCN nanocomposite films were fabricated via a solvent-casting method (Fig. 2a). Fig. 2b presents the FTIR spectra of the SPI, SPI/CNC, and SPI/CNCN films. The spectrum of pure SPI displays characteristic bands at 1630, 1539, and 1238 cm⁻¹, corresponding to amide I (C O stretching), amide II (N—H bending), and amide III (C—N and N—H stretching), respectively, consistent with previous reports (Zhang et al., 2016). The broad peak at 3273 cm⁻¹ is assigned to O—H and N—H stretching vibrations, primarily originating from the intermolecular hydrogen bonding network among protein molecules. A reduction in intensity is observed at 1038 cm⁻¹, likely due to the formation of new hydrogen bonds between the abundant —OH groups on the surface of CNC and CNCN and functional groups such as C—O and N—H on the SPI molecular chains. No absorption peak is detected at 1157 cm⁻¹ in the SPI film, whereas a distinct peak emerges in both SPI/CNC and SPI/CNCN films. This peak corresponds to the asymmetric stretching vibration of the C—O—C pyranose ring in the cellulose backbone, indicating the incorporation of CNC or CNCN into the modified composite films.

Fig. 2.

(a) Schematic illustration of modified CNC and SPI/CNCN nanocomposite film; (b) FTIR spectra of SPI/CNC and SPI/CNCN nanocomposite films; (c) XRD spectra of SPI/CNC and SPI/CNCN nanocomposite films.

The XRD patterns of SPI, SPI/CNC, and SPI/CNCN films are presented in Fig. 2c. The SPI film displays a broad and diffuse diffraction peak at 20.76°, which is ascribed to its distinctive β-sheet structure (Sun et al., 2022). In contrast, the SPI/CNC and SPI/CNCN composite films exhibit a new sharp diffraction peak at 22.54°, characteristic of the (2 0 0) crystallographic plane of cellulose Iβ, confirming the introduction of CNC and CNCN into the SPI matrix while retaining their highly crystalline and rigid structure (Huang et al., 2020). The significantly enhanced diffraction peak intensity at 22.54° for the SPI/CNCN film, compared to the SPI/CNC film, indicates a higher degree of crystallinity or structural ordering. This improvement is attributed to the superior dispersion of CNCN within the SPI matrix and the enhanced interfacial interactions between the phenolic hydroxyl groups of grafted GA on CNCN and the functional groups of SPI.

3.3. Surface morphology

To gain deeper insight into the interactions between SPI and CNC or CNCN, the internal microstructure of the films was investigated using SEM. Representative micrographs at various magnifications are presented in Fig. 3 and S3. Consistent with previous reports (Li et al., 2017), the pure SPI film exhibits an overall uniform but slightly rough surface morphology. This is typical of SPI molecules undergoing cross-linking, aggregation, and drying solidification via intermolecular interactions such as hydrogen bonding and hydrophobic interactions during film formation.

Fig. 3.

SEM micrographs of the cryo-fractured cross-sections of SPI, SPI/CNC and SPI/CNCN nanocomposite films.

Upon incorporation of CNC, the composite film develops micro-scale pores on its surface, a phenomenon attributable to the insufficient interfacial compatibility of CNC nanoparticles within the SPI matrix, leading to their localized aggregation during processing. In contrast, the composite film incorporating modified CNCN displays a rougher fracture surface, yet the microstructure appears denser and more uniform. The improved homogeneity is likely due to the enhanced dispersibility, and stronger interfacial bonding facilitated by CNCN. The incorporation of CNCN further promotes chemical and physical interactions within the matrix, resulting in composite films with higher tensile strength and improved elastic modulus compared to those reinforced with unmodified CNC.

3.4. Tensile strength of modified soy protein film

The mechanical properties of SPI/CNC and SPI/CNCN nanocomposite films are shown in Fig. 4. The incorporation of CNCs into the SPI matrix significantly enhanced the tensile strength (TS) and elastic modulus (EM) of the composite films. This improvement can be attributed to the overall reinforcement of SPI through the introduction of nanofillers and formation of physical or chemical forces, causing reinforcement (Siqueira et al., 2010). Typically, strong hydrogen bonds along with other physical cross-linking were formed between the CNC and the SPI matrix upon the addition of CNC, leading to a cellulose-protein network structure (Mostafa et al., 2023). The TS value increased progressively from 2.17 MPa to 4.83 MPa, representing a 122.58% increase compared to the pure SPI film (P < 0.05). However, the tightly packed framework generally reduced the molecular mobility of the SPI polymer chains, resulting in a decrease in elongation at break (EB) which is in conformity with literature findings (Yu et al., 2017).

Fig. 4.

Strain-stress curves (a), tensile strength (b), elongation at break (c), and elastic modulus (d) of SPI, SPI/CNC and SPI/CNCN nanocomposite films; (e) the water vapor permeability of SPI, SPI/CNC and SPI/CNCN nanocomposite films; (f) A comparative analysis of mechanical properties in a transverse direction.

When CNCN, modified with silane and GA, was incorporated into the SPI film, the tensile strength increased from 2.17 MPa to 7.32 MPa (P < 0.05), corresponding to a 237.33% improvement over the pure SPI film. Although the EB values of both CNC- and CNCN reinforced films decreased compared to the pure SPI film, the EB in the case of the CNCN derived film was higher than that of the CNC reinforced film at equivalent loading levels, likely expressing stronger interaction forces. FTIR and XRD analyses (as shown in Fig. 1a and i) confirmed that the enhancement in mechanical properties of the modified nanocomposite films is due to intermolecular hydrogen bonding and chemical cross-linking between CNCN and the SPI matrix. The reduction in EB is attributed to the increased cross-linking density and diminished surface activity, which results in the depletion of available hydrogen bonds and reduced interfacial stress transfer. Furthermore, the additional chemical cross-linking from the silane and GA modifications contributed to the formation of a more rigid network, thereby reducing EB and toughness. The elastic modulus of the composite films increased by 317.25% and 417.73% with the incorporation of CNC and CNCN, respectively. The increase in both elastic modulus and tensile strength indicates stronger interactions among the internal components of the film network, resulting in a more robust film structure. Compared to previous studies (as summarized in Table S1), the modified SPI film demonstrates superior tensile strength, but a lower elongation at break (Fig. 4f) (Amado et al., 2019; Amado et al., 2023; González et al., 2015; González et al., 2019; Insaward et al., 2014; Li et al., 2015; Liu et al., 2008; Mikus et al., 2021; Yu et al., 2019), indicating that subsequent improvements should focus on enhancing the strength without sacrificing the flexibility. Furthermore, comparative data with conventional petroleum-based polymers are provided in Table S2.

3.5. The thickness, moisture content, water solubility and water vapor permeability

The thickness, moisture content, and water solubility of the obtained composite films are presented in Table 2. The pure SPI film exhibited a thickness of 0.144 mm. While unmodified CNC showed no significant effect on film thickness, CNCN, which was modified with APTES and GA, influenced this parameter. At a CNCN loading of 15%, the SPICN15 film reached its maximum thickness of 0.161 mm, attributed to enhanced interfacial interactions between the amine and phenolic hydroxyl groups on CNCN and SPI molecular chains, which likely restricted the densification of the protein network during film formation. However, when the CNCN content increased to 20%, the thickness of SPICN20 decreased, possibly due to localized agglomeration of CNCN, impairing its uniform dispersion and structural regulation within the matrix.

Table 2.

Thickness, moisture content and water solubility of SPI, SPI/CNC and SPI/CNCN films.

Sample	Thickness (mm)	MC (%)	WS (%)
SPI	0.144 ± 0.011c	28.68 ± 1.55a	32.68 ± 1.88a
SPIC5	0.142 ± 0.007c	28.30 ± 3.19ab	32.46 ± 1.75a
SPIC10	0.141 ± 0.006c	26.67 ± 1.94abc	29.69 ± 2.08bc
SPIC15	0.142 ± 0.007c	25.59 ± 1.30bcd	29.35 ± 2.06bcd
SPIC20	0.148 ± 0.006bc	23.97 ± 1.56cde	28.61 ± 2.79cd
SPICN5	0.141 ± 0.015c	25.96 ± 0.55abcd	31.50 ± 0.70ab
SPICN10	0.157 ± 0.005ab	23.98 ± 0.91cde	29.50 ± 0.56bcd
SPICN15	0.161 ± 0.008a	23.46 ± 1.93de	28.80 ± 0.85cd
SPICN20	0.147 ± 0.007c	21.94 ± 1.14e	26.80 ± 1.39d

Notes: Mean values are presented as ± standard deviation. Data labeled with different lower case letters indicate statistically significant differences (P ≤ 0.05).

The MC of the SPI film decreased from 28.68% to 26.67% after incorporating 10% CNC and further declined to 23.97% with 20% CNC. This reduction is primarily due to interactions introduced by CNC, which reduce interstitial spaces between protein chains and hinder water penetration. With the addition of 20% CNCN, MC decreased to 21.94%. The increased cross-linking density, as supported by the FTIR results (Fig. 1a), restricted film swelling. Silane-induced cross-linking and amide bond formation reduced free hydrophilic groups. Furthermore, the hydrophobic aromatic rings in GA participate in hydrophobic interactions and π-π stacking (Almeida et al., 2023). This enhances the compactness of the matrix and forms a more effective moisture barrier, which consequently leads to a lower water content.

WS tests, which can indirectly indicate changes in physicochemical interactions between SPI and nanofillers, showed that WS of SPIC20 decreased from 32.68% to 28.61%, and SPICN20 dropped to 26.80%, compared to pure SPI. The aforementioned FTIR and XRD results (Fig. 1a and i) indicate that hydrogen bonding between CNC surface hydroxyl groups and SPI consumed hydrophilic sites, and the high crystallinity and strong interactions of CNC reduced protein solubility. The reduction in WS suggests that nanofillers enhance matrix interactions, thereby progressively decreasing film solubility.

An important barrier parameter of packaging films for preventing perishable fruits (e.g., bananas) from environmental exposure is WVP. Fig. 4e presents the evaluation of WVP in SPI composite films. Both CNC and CNCN reduced WVP, with CNCN-modified films showing superior performance. WVP is influenced by component hydrophilicity, intermolecular interactions, and matrix microstructure (Deng et al., 2023). SPICN15 and SPICN20 exhibited WVP values 34.23% and 30.16% lower than pure SPI, respectively. This improvement stems from hydrophobic groups introduced via APTES and GA modification, along with enhanced interfacial bonding between CNCN and the SPI network. These findings align with previous studies on phenolic-compound-modified soy protein-cellulose films (Yu et al., 2018), confirming that higher CNCN loading enhances water vapor barrier performance in SPI films.

3.6. Thermal stability, optical and UV blocking performance

Generally, the thermal stability of composite materials increases after cross-linking (Masutani & Kimura, 2014). The thermal stability of nanocomposite films is shown in Fig. 5a and b. The initial weight loss (30–130 °C) corresponds to moisture evaporation. The second stage (130–270 °C) involves bond cleavage and glycerol loss, and the third (270–390 °C) results from protein and cellulose decomposition (Martelli-Tosi et al., 2018; Zhang et al., 2016). Compared to the pure SPI film, composite films containing CNC and CNCN exhibited lower degradation rates and higher residual weights starting from the second stage. The thermogravimetric (TG) curves of the SPI nanocomposites shifted toward higher temperatures relative to the pure SPI film. This indicates that the incorporation of nanofillers, along with silane and GA modifications, and the resulting hydrogen bonding and crosslinked structures, collectively enhanced the interaction within the soy protein-cellulose network, thereby improving the thermal stability of the material.

Fig. 5.

(a) and (b) Thermogravimetric (TG) curves and thermogravimetric derivative (DTG) curves of SPI, SPI/CNC and SPI/CNCN films, respectively; The opacity (c) and UV–vis transmission spectra (d) of SPI, SPI/CNC and SPI/CNCN composite films; (e) Effects of SPI and SPI/CNCN films on the growth of E. coli and S. aureus.

During transportation and storage, UV radiation can readily induce photooxidation and quality deterioration of fresh produce, particularly light-sensitive fruits like bananas. Therefore, the development of packaging materials with UV-blocking capabilities is of significant importance for their preservation (Lee et al., 2023). Pure soy protein films exhibit a pale yellow color and are nearly transparent. As the content of CNC and CNCN increases, the opacity of the films also increases (P ≤ 0.05), as illustrated in the corresponding Fig. 5c. The light transmittance of pure SPI, SPI/CNC, and SPI/CNCN films was measured in the wavelength range of 200–800 nm, and presented in Fig. 5d. The nanocomposite films demonstrated excellent UV-shielding properties, blocking 100% of both UV-C (200–280 nm) and UV-B (280–320 nm) radiation. However, approximately 7.28% to 16.75% of UV-A (320–400 nm) light was still able to transmit through the protein-based films. As the content of CNC and CNCN increased, the composite films generally exhibited enhanced UV barrier performance. The SPICN20 film showed the most effective UV blocking, preventing 92.72% of UV-A transmission. This remarkable UV barrier performance is attributed to the benzene rings and phenolic hydroxyl groups of GA, whose conjugated system enhances the UV absorption capacity of the film through π → π* electron transitions (Xing et al., 2019). Modification with APTES and GA also improves the interfacial adhesion between CNC and the SPI matrix, further increasing the compactness of the film and reducing UV penetration. These results indicate that SPI/CNCs composite films possess excellent UV barrier properties and can effectively protect bananas from photooxidation and degradation.

3.7. Antibacterial and antioxidant activity of composite films

Effective antibacterial properties are crucial for preserving perishable fruits such as bananas. This study systematically evaluated the antibacterial performance of composite films against two common foodborne spoilage bacteria (Fadiji et al., 2023), E. coli and S. aureus, with results shown in Fig. 5e (supplemented by Fig. S4). Compared to the blank control group, neither the pure SPI film nor the SPI/CNC film exhibited significant antibacterial activity. In contrast, the SPI/CNCN films demonstrated notable antibacterial efficacy, which can be attributed to the phenolic and other active functional groups introduced by CNCN. These compounds disrupt bacterial cell membrane structures, thereby inhibiting enzymatic activity and ultimately leading to bacterial death (Gangadharan et al., 2024). Furthermore, the antibacterial effect increased markedly with higher CNCN loading. The SPICN20 film exhibited the strongest inhibitory activity, nearly completely suppressing visible colony formation of S. aureus and inhibiting a large portion of E. coli. It is noteworthy that all SPI/CNCN films showed greater efficacy against S. aureus than against E. coli, which aligns with findings reported by Venkatesan (Venkatesan et al., 2024).

The antioxidant activity of SPI-based films was evaluated using the DPPH assay. Results indicated that the SPI film exhibited low activity (4.21 ± 1.1%), while the incorporation of CNCN grafted with GA significantly enhanced the antioxidant activity of the SPI/CNCN film to 12.42 ± 0.4% (Fig. 6a). Although CNC itself showed limited radical scavenging capacity (Han et al., 2015), the covalent grafting of GA onto CNC via EDC/NHS-mediated carboxyl activation enabled efficient free radical scavenging. Each GA molecule can scavenge up to six DPPH radicals (Huang et al., 2021). The covalent bonding introduced by APTES improved the stability of grafted GA, and the antioxidant activity increased with higher CNCN content. Therefore, CNCN modification effectively enhances the antioxidant performance of SPI films, demonstrating promising potential for extending perishable fruits such as bananas.

Fig. 6.

(a) DPPH radical scavenging performance of SPI, SPI/CNC and SPI/CNCN films; (b) weight loss of untreated bananas with storage time; (c) color difference of untreated bananas with storage time; (d) firmness of untreated bananas with storage time; (e) changes in the appearance of untreated bananas over different periods of storage.

3.8. Impact of packaging on banana preservation

To investigate the preservation performance of SPI/CNCs composite films, bananas were selected and divided into untreated and pre-treated groups. Key quality indices, namely weight loss, firmness, color difference (ΔE), and visual browning, were evaluated under packaging conditions (SPI, SPI/CNC, and SPI/CNCN) and in air. Detailed data of the pre-treated group are shown in Fig. S5. The pre-treated bananas exhibited more severe browning than the untreated ones, likely due to removal of the natural wax layer during washing, which accelerated moisture evaporation.

Fig. 6b–d presents the weight loss, ΔE, and firmness of untreated bananas during storage, while Fig. 6e displays photographs of bananas stored for 7 days in plastic containers covered with different film samples (Fig. S6 and S7). Over the 7-day storage, all untreated banana samples showed increasing trends in weight loss, ΔE, and browning, while firmness decreased. The UC group exhibited the most severe browning, with weight loss reaching 29.92 ± 2.45%, ΔE reaching 23.32 ± 0.41, and firmness reduced to 5.80 ± 0.44 N. In contrast, all film-packaged groups effectively suppressed browning. The pure SPI film reduced weight loss to 20.81 ± 0.87%, ΔE to 18.65 ± 0.32, and maintained firmness at 7.12 ± 0.28 N, while inhibiting browning. For SPI/CNC and SPI/CNCN composite films, weight loss systematically decreased and browning was alleviated with increasing nanofiller content, consistent with the gradual enhancement of barrier properties (Fig. 4e). Among them, SPICN20 showed the lowest weight loss (11.75 ± 0.62%) and ΔE (8.96 ± 0.42), the highest firmness (11.97 ± 0.23 N), and the most pronounced suppression of browning. At the same loading level, it outperformed SPIC20 (weight loss 15.88 ± 0.61%, ΔE 15.14 ± 0.35, firmness 9.83 ± 0.21 N), indicating superior preservation performance of the SPI/CNCN composite film. This improvement is attributed to the stronger antioxidant activity of the SPICN20 film (Fig. 6a), in which the covalently grafted GA effectively scavenges free radicals and quinone intermediates generated during banana storage, thereby blocking the oxidative browning reaction (Yu et al., 2024). Additionally, the higher cross-linking density and lower WVP collectively reduced moisture loss, thus delaying both physical desiccation and enzymatic browning processes.

3.9. Film soil burial degradation test

Through soil degradation experiments on SPI films and their composite films, along with a comparison of sample photographs at 0, 6, 12 and 20 days (Fig. 7, with supplementary images in Fig. S8), their degradation capacity can be visually evaluated. It is well known that protein molecular chains are susceptible to hydrolysis by enzymes secreted by soil microorganisms, thus the soy protein matrix itself possesses inherent biodegradability potential (Xu et al., 2024). All film samples absorbed water, softened, and deformed markedly within 20 days. Initial signs of decomposition appeared in all composite films by day 6, though their overall morphology remained largely intact. Further degradation occurred by day 12, accompanied by loss of original shape, adhesion, and shrinkage. After 20 days of burial, the films had reached a high degree of biodegradation, with most material integrated into the soil and no longer identifiable as discrete film structures. These results indicate that the incorporation of CNC and CNCN did not impair the degradation capability of the films. However, the introduction of fillers had a noticeable impact on the degradation rate. The degradation rates of SPIC20 and SPICN20 films were evidently slower, with SPICN20 in particular demonstrating relatively better integrity. This could be attributed to the higher degree of chemical cross-linking within the film network, which temporarily acts as a barrier against microbial erosion, a conclusion consistent with observations made during the degradation of SPI films modified with hyperbranched polyester-cardanol derivatives (Gu et al., 2019).

Fig. 7.

Photographs of SPI, SPI/CNC, and SPI/CNCN films after soil burial for different time intervals. (Burial conditions: depth 10–15 cm, soil pH 8.13, moisture content 20.6%).

4. Conclusions

This study successfully transformed CNC into multifunctional reactive nanomodifiers (CNCN) via a two-step approach involving APTES silanization followed by EDC/NHS-mediated grafting of GA. The core of this strategy lies in the integration of interface coupling and active grafting, which not only improves the dispersion and interfacial compatibility of CNCs but also endows them with antioxidant, antibacterial, and UV-shielding functionalities, thereby synergistically enhancing the overall performance of SPI/CNCs nanocomposite films. Specifically, the silanization constructs a stable siloxane network on the CNC surface, providing covalent anchoring sites for subsequent reactions and improving dispersion. The grafted GA then forms a multifunctional active interface that interacts with the SPI matrix through both covalent and non-covalent bonds, enabling CNCN to act as an efficient multi-point cross-linking center. Consequently, the composite structure transitions from a physically blended, hydrogen-bond-dominated system (SPI/CNC) to a densely cross-linked, multi-level hybrid network (SPI/CNCN). This structural transformation fundamentally explains the synergistic improvements in mechanical strength, water vapor barrier properties, and thermal stability. Furthermore, the grafted GA enhances the film's UV-blocking and antioxidant capabilities while imparting potential antibacterial activity. Combined with the dense cross-linked network of the SPI/CNCN films, the composite material simultaneously blocks water vapor transport, inhibits enzymatic browning, and retards microbial activity, achieving multi-path suppression of banana spoilage and significantly delaying deterioration. These results demonstrate its strong potential for banana fresh-keeping packaging applications. In summary, the strategy presented in this work offers a promising and effective approach for developing sustainable and biodegradable SPI-based active packaging materials.

CRediT authorship contribution statement

Yan Zhu: Writing – original draft, Software, Formal analysis. Haizhu Wu: Writing – review & editing, Methodology, Formal analysis. Weijia Yang: Software, Methodology, Formal analysis. Bertrand Charrier: Writing – review & editing, Methodology. Hisham Essawy: Writing – review & editing, Methodology. Antonio Pizzi: Writing – review & editing, Resources, Methodology. Xiaojian Zhou: Writing – review & editing, Resources, Methodology, Funding acquisition. Xinyi Chen: Writing – review & editing, Software, Methodology, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material 1

Supplementary experimental methods, including the banana preservation performance test in Section S1, as well as supporting data in Figs. S1–S8, Tables S1–S2, and additional explanatory notes.

Data availability

Data will be made available on request.