According to the method described by Gowda et al. (2020), dopamine hydrochloride was conjugated to gelatin A (from porcine skin) in the presence of EDC and NHS. Specifically, 1.0 g of gelatin-A and 250 mg of dopamine hydrochloride were dissolved in 50 mL of MES buffer. Subsequently, appropriate amounts of EDC and NHS were added to the mixture. The mixture was then shaken at 100 rpm in the dark at 37 °C for 24 h. The resulting solution was purified using a dialysis membrane against deionized water for four cycles. The final Gel-DOPA product was lyophilized and stored at −20 °C (Figure 2B) (Brennan et al., 2017; Lee et al., 2006). Approximately 1 mg of lyophilized powder sample was thoroughly mixed with 100 mg of spectroscopically pure potassium bromide powder in an agate mortar and ground homogeneously. The mixture was then compressed under a pressure of 10 MPa to form a transparent pellet for scanning. Fourier transform infrared spectroscopy (FT-IR) was performed to detect the surface functional groups in the synthesized Gel-DOPA, raw gelatin, and dopamine hydrochloride, confirming the formation of the target product. FT-IR was performed over a range of 350–7,800 cm^-1 at a scanning resolution of 2 cm^-1 over 32 scans using an FT-IR spectrometer (Nicolet IZ10, Thermo Fisher Scientific, United States).

The backing-layer solution was prepared by mixing 75% ethanol, zein powder, and glycerol at a mass ratio of 250:25:1 (Yan et al., 2024). Subsequently, 7 mL of this solution was coated onto a 90-mm diameter Petri dish and vacuum-dried at 65 °C for 2 h to form a backing layer film (Figure 2A).

The adhesive and drug-loaded layers of the film were fabricated using 3D printing. The ink formulations for 3D printing are as follows: doxepin hydrochloride, 250 mg; Gel-DOPA, 200 mg; HPMC, 200 mg; CARBOPOL, 200 mg; ethanol, 4.7 mL; and water, 4.7 mL. Ethanol, water, and a stirring bar were placed in a screw-capped glass vial. Under continuous stirring, gelatin-polydopamine, CARBOPOL, HPMC, and doxepin hydrochloride were sequentially added (Takashima et al., 2022). The vials were left overnight at room temperature. Subsequently, the formed gel was loaded into a syringe with a 27G nozzle and placed in a 3D printer (INKREDIBLE, CELLINK BIO X, Sweden). The printing parameters were set to a circular area with a diameter of 90 mm, height of 1 mm, and an air pump pressure of 42 kPa. The printed gel was air-dried overnight at room temperature and a doxepin hydrochloride-loaded oral double (DL) film was fabricated (Figure 2D).

Powder XRD patterns of the DL film and related powder samples were analyzed. They were obtained using a MiniFlex 600 (Rigaku Co., Tokyo, Japan) by irradiating with Cu-Kα X-rays. The tube voltage and amperage were 35 kV and 25 mA, respectively. Samples were scanned from a 2θ of 5°–90°.

The DSC peaks of the DL film, doxepin hydrochloride, and the excipient were measured using a differential scanning calorimeter (TGA/SDTA 851e; Mettler-Toledo, Switzerland). Approximately 2 mg of the sample was placed at the bottom of the sample pan and the temperature was increased from 25 °C to 600 °C at a rate of 10 °C/min.

The WCA of the backing and drug-carrying layers were determined using a goniometer (Theta, Biolin, Sweden). Water (5 μL) was automatically dropped onto the film. The WCA was immediately and automatically measured on both sides of the droplet. All measurements were repeated at least three times for each sample and the results were averaged.

Three batches of DL films were prepared according to the methods outlined in Sections 2.2.2, 2.2.3. Three patches were randomly selected from each batch, and a 10 × 10 mm section was excised from each patch. These samples were dissolved in 30 mL of artificial saliva, then filtered through a 0.22-μm aqueous membrane filter. The concentration of doxepin hydrochloride in the filtered solution was analyzed using liquid chromatography (LC20A, Shimadzu 145, Japan) with reference to the method described by Sanz et al. (2015) Single-point quantification was performed using doxepin hydrochloride solutions of known concentrations and the doxepin hydrochloride content in each patch was calculated.

The in vitro release test was performed using a Franz diffusion cell. A cellulose acetate membrane (0.22 μm) was immersed in phosphate buffer (pH 6.8) for 12 h and placed on top of the receptor compartment. A DL film (1.5 × 1.5 cm) was attached to the membrane with its drug-containing layer facing the receptor compartment. The receptor compartment was filled with 9 mL artificial saliva. The system was maintained at 37 °C with a stirring rate of 100 rpm. Aliquots were collected at predetermined time intervals (10–250 min) and the concentration of doxepin hydrochloride in the dissolution medium was quantified using HPLC (as described in Section 2.3.7) using a validated standard curve (r² > 0.99).

The ex vivo drug permeation experiment was conducted using porcine buccal mucosa mounted on Franz diffusion cells because its histological structure closely resembles that of the human buccal mucosa (Figueiras et al., 2009). Artificial saliva (8 mL) and a magnetic stirrer bar were added to the receptor compartment. A porcine buccal mucosal tissue sample (approximately 2 × 2 cm) was clamped between the donor and receptor compartments. A 1.5 × 1.5 cm DL film was moistened and applied to the mucosal surface. The Franz diffusion cell temperature was maintained at 37 °C with a stirring speed of 100 rpm. At designated time intervals (0.5, 1, 1.5, and 2 h), the film was removed and the receptor fluid was collected. The mucosal tissue was minced and ultrasonicated in acetone for 10 min to extract the residual drug. The amount of doxepin hydrochloride in both the receptor fluid and acetone solution was quantified to determine the total amount of permeated doxepin hydrochloride.

We selected two additional commercially available patch-type drugs in China for the treatment of oral ulcers as references: propolis oral film (ZIZHU PHARMACEUTICAL, China) and dexamethasone acetate oral mucoadhesive patches (Shenzhen Taitai Pharmaceutical Co., Ltd., China). The weights of the film samples were measured using a precision electronic balance with an accuracy of 0.1 mg and the relative weight of three commercially available oral ulcer patches in China were calculated based on their dimensions. The thicknesses of the film samples were measured at three randomly selected points using Vernier calipers with an accuracy of 0.001 mm (n = 5).

Each type of oral patch adhered to the buccal mucosa of the anesthetized rats (n = 5) and the adhesion of the patches was visually monitored within 0–150 min. The average time required for patch detachment from the mucosa was recorded as the in vivo adhesion time.

Following the method described by Perioli et al. (2004), porcine buccal mucosa was first adhered to the inner wall of a beaker using an adhesive. A patch was moistened with 50 μL of phosphate-buffered saline (PBS) and pressed onto the porcine mucosa surface for 20 s. The beaker was filled with 800 mL of PBS maintained at 37 °C. After 2 min, stirring was initiated at 150 rpm and the adhesion time of the patch was monitored for over 300 min. Three samples were tested for each patch type and the average adhesion time was calculated as the in vitro adhesion time of the patches.

As described in previous reports, we combined 5-FU injections with the topical application of 70% glacial acetic acid to induce chemotherapeutic oral mucosal ulcers in the buccal mucosa of rats (Çelebi et al., 2022; Lima Gde et al., 2015). Specifically, on days 1, 3, and 5, the rats received intraperitoneal injections of 0.5% 5-FU solution at a concentration of 60 mg/kg. Following the final injection on day 5, the rats were anesthetized and a small cotton ball (approximately 8-mm in diameter) soaked in 70% glacial acetic acid was applied to the buccal mucosa for 40 s to induce cauterization until the local mucosal surface appeared white. Overall experimental protocol was shown in Figure 1. In our preliminary pilot experiments, the observed effect size was f = 0.78, with a significance level (α) set at 0.05 and statistical power (1-β) set at 0.9. With a total of four groups, and using one-way ANOVA for comparisons between different groups, the calculated sample size per group was 5, resulting in a total sample size of 20 in order to evaluate the efficacy and biosafety of DL film in CIOM. To further explore the mechanism of DL film in treating CIOM, totally 35 SD rats were utilized.

Rats were randomly divided into the following four groups: Group 1 (control group): rats that were modeled but not treated (n = 5); Group 2: modeled rats treated with doxepin patch (n = 5); Group 3: modeled rats treated with the propolis patch (n = 5); and Group 4: modeled rats treated with the dexamethasone patch (n = 5). All groups received intraperitoneal anesthesia with pentobarbital sodium administered once daily by designated experimenters at fixed time points before treatment. The modeling process was conducted in two batches. In the first batch, photographic documentation of buccal mucosal ulcer lesions was performed during treatment and ulcer area changes were analyzed using ImageJ. Macroscopic scoring was conducted on days 3, 6, and 9 of treatment using the following criteria: 1. intact oral mucosa without abnormalities; 2. moderate mucosal erythema and congestion with punctate erosions, and absence of ulcers or abscesses; 3. severe mucosal erythema and congestion with hemorrhage and small ulcers, and no abscess formation; and 4. severe mucosal erythema and congestion with hemorrhage, extensive ulceration, and abscess formation. In the second batch, the rats were euthanized after 3, 6, or 9 days of treatment. Buccal ulcer tissues were collected, fixed in neutral-buffered formalin, embedded in paraffin, and subjected to H&E staining for pathological evaluation. The major organs (heart, liver, spleen, lungs, and kidneys) were harvested after treatment for additional H&E staining to assess tissue damage and pathological alterations.

The modeling method was conducted as described in Section 2.4.1, with rats divided into three groups: Group 1, healthy rats (n = 5); Group 2, untreated modeling group (n = 15); and Group 3, doxepin-treated modeling group (n = 15). After 3, 6, and 9 days of treatment, five rats were randomly selected from Groups 2 and 3 for buccal mucosal lesion tissue collection. Part of the tissue was analyzed via RT-qPCR to detect expression levels of inflammatory factors (IL-1b, IL-6, and TNF-a) and antioxidant stress factors (superoxide dismutase [SOD] and heme oxygenase-1 [HO-1]). The reagent kits and primers used in RT-qPCR are listed in Supplementary Material. The remaining tissue was fixed in neutral-buffered formalin, embedded in paraffin, and subjected to immunohistochemical analysis for Keap1 and Nrf2 expression. ImageJ software was used to select the yellow-stained immunohistochemical reaction areas on the images, followed by measurement of the mean optical density and semi-quantitative analysis of these regions.

As shown in Figure 2C, the absorption band of Gel-DOPA near 3,434 cm^-1 corresponds to N-H/O-H stretching vibrations. Gelatin inherently exhibits a broad band owing to protein hydrogen bonding interactions. The vibrations of the phenolic hydroxyl groups (-OH) and amines (-NH-) in polydopamine also fell within this range. The superposition of these two factors resulted in a strong and broad signal in this region. The band at 1,635 cm^-1 (amide I, C=O and C-N stretching vibrations) serves as a sensitive indicator of protein secondary structures, such as β-sheets. The band at 1,550 cm^-1 (amide II) arises from N-H bending vibrations and C-N stretching vibrations, with additional contributions from aromatic ring vibrations (e.g., C=C stretching in benzene rings) in polydopamine. The band at 1,240 cm^-1 (amide III, C-N and N-H vibrations) provides critical information about protein conformations. These characteristic bands are attributed to Gel-DOPA. The emergence of a new C-O vibration peak indicates the successful grafting of catechol groups from dopamine hydrochloride onto the gelatin molecular chains. The band at 2,945 cm^-1 represents asymmetric stretching vibrations of CH₂, while the band at 2,875 cm^-1 corresponds to symmetric stretching vibrations of CH₂. Bending and rocking vibrations of CH₂ are observed at 1,455 and 1,333 cm^-1, respectively. Finally, the band at 1,032 cm^-1 corresponds to CH₃ amide groups and the C-O-C band appears at 620 cm^-1. This analysis conclusively demonstrates the successful synthesis of Gel-DOPA.

The sharp diffraction peaks of HPMC and doxepin hydrochloride indicated their highly crystalline nature (Figure 2E). In contrast, a broad hump was observed between 20° and 30° for carbomer 974P, gelatin-polydopamine, and the drug-loaded layer mixture reflected amorphous characteristics. The random molecular chain network of the carbomer, which is a cross-linked polyacrylic acid, inhibits crystalline ordering. After polydopamine modification, the original collagen triple-helix structure of gelatin was disrupted, forming a disordered, physically cross-linked network. The amorphous state of the drug-loaded mixture stems from complex intermolecular interactions; hydroxyl groups from HPMC, carboxyl groups from carbomers, and drug molecules engage in hydrogen bonding and ionic interactions that prevent molecular alignment. This amorphous structure enhanced the dissolution rate and facilitated drug release.

DSC curves (Figure 2F) revealed distinct thermal behaviors of HPMC, doxepin hydrochloride, carbomer 974P, gelatin-polydopamine, and synthetic gels. The endothermic peak of carbomer 974P (e) near 160 °C was likely associated with the rupture of intramolecular hydrogen bonds. The dual endothermic peaks of doxepin hydrochloride (b) at 190 °C and 270 °C might correspond to its melting process and hydrochloride salt decomposition, respectively, while the exothermic bulge at 560 °C could represent the exothermic response from oxidative decomposition of organic carbon chains. The exothermic peak of HPMC (a) at approximately 450 °C was related to thermal oxidative decomposition of its cellulose backbone, with heat release from cleavage of methyl and hydroxypropyl substituents at elevated temperatures. The sharp exothermic peak of gelatin-polydopamine (d) at 505 °C might originate from carbonization exotherm of the polydopamine cross-linked network, where energy release occurs through aromatic ring structural reorganization above 500 °C. These results are also consistent with previous reports. (Khan et al., 2022; Kanwar et al., 2016; Ghaderi et al., 2019). The mixed component system (e) showed no distinct thermal events, indicating the formation of a stable amorphous composite system through intermolecular interactions (e.g., hydrogen bonding between HPMC hydroxyl groups and carbomer carboxyl groups, π-π stacking between gelatin-polydopamine and drugs), which suppressed characteristic thermal behaviors of individual components. This observation is consistent with the XRD results, suggesting that such an amorphous drug dispersion facilitates drug dissolution and enables uniform distribution within the film matrix.

WCA analysis confirmed that the backing layer is hydrophobic, whereas the drug-loaded and adhesive layers are hydrophilic. As shown in Figure 3A, the backing layer exhibited a significantly higher WCA (104.6° ± 3.2°, 1 s) compared with that of the drug-loaded layer (55.1° ± 0.7°, 1 s). The hydrophobicity of the backing layer helps prevent the dilution of the drug concentration caused by saliva washing in the oral cavity to a certain extent, thereby achieving a localized unidirectional release effect of the drug (Tehrani-Bagha, 2019). The hydrophilic surfaces of the drug-loaded and adhesive layers facilitate patch adhesion and drug release.

Furthermore, the drug content in DL membranes fabricated via 3D printing technology demonstrated strong homogeneity. As shown in Figure 3B, no significant differences were observed in the average drug content (0.4 mg/cm²) among three randomly selected batches of DL membranes.

The DL film developed in this study demonstrated an ultrathin profile of 128.3 ± 1.5 μm, comparable to propolis patches and significantly thinner than dexamethasone acetate patches (1,546.0 ± 26.9 μm) (Figure 3C). Moreover, its areal density was merely 19.8 ± 1.6 mg/cm², substantially lower than that of dexamethasone patches (Figure 3D). These metrics highlight the notable thinness and lightweight characteristics of the film, which enhance patient comfort by minimizing foreign body sensations and improving mucosal conformity during localized administration. The results of the in vitro dissolution test demonstrated that doxepin hydrochloride was released gradually as the components of the drug-loaded layer dissolved, with nearly 80% of the drug released within 90 min (Figure 3E). Meanwhile, data from the in vitro permeability study indicated that the dissolved doxepin could permeate the local mucosal tissue. Although approximately 40% and 50% of the total doxepin hydrochloride was released at 0.5 h and 1 h, respectively, the amount permeating into the mucosa remained below the detection limit of the chromatographic assay (Figures 3E,F). However, at later time points, the drug levels within and beneath the mucosa showed a detectable increase (Figure 3F). This confirms that the drug is released gradually over time and can continuously permeate into the mucosal tissue.

An ideal buccal adhesive system requires adequate mucoadhesion to ensure localized retention and sustained therapeutic effects (Shipp et al., 2022). Therefore, we evaluated the in vivo and in vitro adhesion duration of the three patch types to assess the mucoadhesive performance of the DL film. As depicted in Figure 3G, the DL film incorporating the wet adhesive material Gel-DOPA demonstrated a significantly prolonged in vivo adhesion duration (average of 57 min) compared to that with propolis patches. In vitro adhesion testing showed comparable adhesion times between the DL film and the dexamethasone patch (194.2 min vs 198.8 min), both markedly exceeding that of propolis patches (21.6 min). This discrepancy likely stems from the absence of a backing layer in propolis patches, leading to rapid dissolution in high-moisture environments. The adhesion performance of three types of patches on rat buccal mucosa is shown in Figure 3H. It can be observed that although the dexamethasone patch demonstrates extended adhesion time, its lack of flexibility and fixed size prevent complete mucosal conformal adhesion and personalized dosing adaptation based on the specific dimensions of ulcers.