Polysaccharide from Asimina triloba fruits alleviates DSS-induced ulcerative colitis by modulating inflammatory signaling and gut microbiota

Abstract

The treatment of ulcerative colitis (UC) remains a clinical challenge due to its frequent recurrence, underscoring the urgent need for novel functional food-based therapeutic strategies. In this study, a polysaccharide (ATP-W-1) was isolated from Asimina triloba fruit, structurally characterized, and evaluated for its therapeutic efficiency in a DSS-induced UC mouse model. ATP-W-1, with a molecular weight of 164,504 g/mol, was composed of mannose, rhamnose, glucose, galactose, and arabinose. In vivo study revealed that ATP-W-1 significantly ameliorated UC symptoms by preventing body weight loss and modulating inflammatory cytokines (IL-6, IL-10, and TNF-α) expression. Specifically, it downregulated IL-6 and TNF-α associated with the NF-κB pathway and restoration of epithelial barrier integrity, as evidenced by increased expression of tight junction proteins (Claudin 1, ZO-1, and Occludin). Furthermore, ATP-W-1 positively modulated the gut microbiota composition and enhanced the prodcution of short-chain fatty acids (SCFAs), thereby contributing to the reestablishment of intestinal homeostasis. Collectively, these findings highlight that ATP-W-1 is a promising functional food ingredient with therapeutic potential for UC prevention and management.

Introduction

Inflammatory bowel disease (IBD) refers to a nonspecific inflammatory disorder of the gastrointestinal tract; this condition is classified into two forms, ulcerative colitis (UC) and Crohn’s disease¹. IBD is associated with several pathogenic mechanisms, including immune reactions, genetic factors, environmental factors, and intestinal barrier dysfunction². The immune defense against inflammation occurs in response to oxidative stress, infection, and pathogenic invasion³. Furthermore, UC damages the integrity of the entire intestinal epithelium, thus causing abdominal pain, diarrhea, mucopurulent discharge, and bloody stools^4,5. The development of UC is mainly associated with gut microbiome (GM) diversity⁶. Recent studies suggest that UC is characterized by reduced microbial richness and an abnormal microbial composition (dysbiosis), marked by a decrease in short-chain fatty acid (SCFAs)-producing bacteria and an overabundance of pro-inflammatory pathogens such as Enterobacteriaceae spp. and Enterococcus faecium^7–9. Dysbiosis promotes inflammatory responses^10–12. Moreover, the GM is crucial in regulating intestinal mucosal immune functions, epithelial cell renewal, differentiation, mucus secretion, intestinal motility, and pathogen invasion^13–16. The UC-associated intestinal mucosal inflammation leads to tissue damage and an increased risk of neoplasia development^17,18. Acidic mucins and glycogen analysis are essential to clarify the pathological alterations of the mucosa¹⁹.

The dextran sodium sulfate (DSS)-induced UC mouse model has often been used in research studies because of its reproducibility and rapid induction of colitis^20,21. Although the exact mechanism of DSS-induced UC is unknown, earlier studies reported that DSS induces colitis by triggering proinflammatory cytokines, reducing tight junction proteins (ZO-1, Occludin, and Claudin-1), and altering the GM composition^22,23. Moreover, DSS resembles the colitis conditions observed in humans. Therefore, DSS is commonly used to develop the UC mouse model for scientific research^22,24. However, 5-aminosalicylic acid (5-ASA), glucocorticoids, and immunosuppressants are commonly used to treat UC in clinical regimens. However, the long-term use of these drugs can result in drug tolerance and adverse effects, including nephropathy, pancreatitis, hepatotoxicity, vomiting, and nausea²⁵. Hence, UC treatment strongly demands more effective and non-toxic therapeutics. Plant polysaccharides have attracted considerable interest due to their safety and non-toxic nature^26,27. In this context, the polysaccharides derived from plants, such as Rheum tanguticum, Morinda citrifolia L, Astragalus, Plantago asiatica L, Chrysanthemum morifolium Ramat, and Dendrobium officinaleon have been reported to relieve the symptoms of UC by regulating GM, enhancing SCFA production, promoting the mucosa, regulating the IL-6/JAK2/STAT3 signaling pathway and tight junction proteins, and inhibiting NF-κB activation and pro-inflammatory cytokines (TNF-α and IL-6)^28–32.

A. triloba is a member of the Annonaceae family, containing bioactive compounds such as phenolics, annonaceous acetogenins, and alkaloids with promising bioactivities, including antioxidant, anticancer, antitumor, pesticidal, antimalarial, piscicidal, anthelmintic, antiviral, and antimicrobial effects^33–35. However, the therapeutic effect of polysaccharides extracted from A. triloba fruits against UC has not been explored. Therefore, the present study hypothesized that the polysaccharide from A. triloba fruits exerts protective effects against UC by modulation of GM composition, increasing SCFA generation, rebuilding intestinal barrier function, and reducing inflammatory responses. Thus, this study aimed to isolate and characterize the polysaccharides (ATP-W-1) from the fruit of A. triloba, and investigate its mechanism of action in the DSS-induced UC by analyzing GM profiles, SCFA production, inflammatory genes, and tight junction proteins.

Results and discussion

Isolation, purification, and molecular weight of ATP-W-1

The isolation and purification of ATP-W-1 from the A. triloba fruit are demonstrated in the flow chart in Fig. 1A. The crude ATP extract was obtained by ethanol precipitation, yielding 18.43%. This was then purified by DEAE-Sepharose Fast Flow through elution with water and different molar concentrations of NaCl, which resulted in a water fraction of ATP-W (59.2%), and two different fractions at 0.1 and 0.2 M NaCl: ATP-1 (8.7%) and ATP-2 (6.4%) respectively (Fig. 1B). After lyophilization, ATP-W was further purified using Sephacryl S-300/HR column. A symmetric single peak was obtained by water elution, with a yield of 72.1%, and named ATP-W-1 (Fig. S1). The purity and molecular weight were assessed using HPGPC. The GPC profile of ATP-W-1 revealed a single, symmetrical peak, suggesting homogeneity with an average molecular weight (Mw) of 164,504 g/mol and a number-average molecular weight (Mn) of 143,046 g/mol for ATP-W-1 (Fig.1C). The polydispersity index (PDI) was calculated as the M_w/M_n ratio. The PDI was 1.15, calculated for ATP-W-1, indicating the molecular size range distribution in ATP-W-1. The morphological characteristics of ATP-W-1 were examined using SEM at two different magnifications, ×500 and ×20,000 and the results revealed that the polysaccharide exhibited an irregular, slice-like structure (Fig. 1D). The TGA curve evidenced the thermal stability and decomposition behavior of ATP-W-1 under nitrogen atmosphere at 100 to 800 °C (Fig. 1E). A significant weight loss of 60.98% was observed in the second stage at 270 to 470 °C, which signifies the primary thermal decomposition of the ATP-W-1. This phase is associated with the degradation of ATP-W-1 through the cleavage of glycosidic bonds³⁶. The elevated decomposition temperature indicated that the ATP-W-1 possesses moderate thermal stability, potentially advantageous for specific industrial or biomedical applications.

Fig. 1. Isolation and characterization of ATP-W-1.

Schematic experimental workflow for isolating ATP-W-1 from the A. trioba fruit (A). DEAE-52 Sepharose fast flow column fractionation of ATP-W-1 (B), determination of molecular weight of ATP-W-1 by HP-GPC chromatogram (C), observation of the surface morphology of ATP-W-1 by scanning electron microscopy (SEM) (D), analysis of thermal stability of ATP-W-1 by TGA analysis (E). Elucidation of active functional groups of ATP-W-1 FTIR spectrum (F).

FT-IR analysis of ATP-W-1

The FTIR spectrum of ATP-W-1 is shown in Fig. 1F. Broad vibration peaks at 3286 and 2933 cm⁻¹ were assigned to the O–H stretching and C–H stretching vibrations, respectively, as characteristic peaks of carbohydrates. These functional group vibrations are commonly observed in plant polysaccharides, including those from Malus halliana³⁷, Inonotus hispidus³⁸ and purple sweet potato³⁹, which indicate the structural similarity of polysaccharides. An absorption peak observed at 1641 cm⁻¹ corresponds to O–H bending vibrations⁴⁰. The peaks at 1404 cm⁻¹, was attributed to C–H vibrations⁴¹. The absorption peak at 1240 cm⁻¹ corresponds to the asymmetric stretching vibration of the C–O–C bond, which is characteristic of glycosidic linkages in polysaccharide structures³⁹. The peak at 1048 cm⁻¹ was ascribed to C–O bending vibrations associated with C–O–H or C–O–C linkages, suggesting the presence of pyranose ring structures⁴². Moreover, an absorption peak at 890 cm⁻¹ indicated the existence of β-glycosidic linkages⁴³.

Monosaccharide analysis

The monosaccharide composition analysis revealed that ATP-W-1 was composed of mannose (Man), rhamnose (Rha), glucose (Glc), galactose (Gal), and arabinose (Ara) (Fig. 2). The calculated molar ratios were 2.23, 6.43, 50.12, 8.36, and 31.91, respectively (Table 1), indicating that Glc and Ara were the predominant constituents. No acidic sugars, such as glucuronic acid (GlcA) or galacturonic acid (GalA), were detected in ATP-W-1, suggesting a distinct neutral polysaccharide profile. These results contrast with previously reported fruit-derived polysaccharides, which typically exhibit higher proportions of acidic residues such as GalA^44–46. The measurement of each monosaccharide was verified by standard calibration curves (R² > 0.995 for all sugars) ranging from 2.5 to 500 µg/mL. The lower limit of detection (LOD) ranged from 0.72 to 1.13 µg/mL. In contrast, the limit of quantification (LOQ) ranged from 1.75 to 3.45 µg/mL (Supplementary Table S1). These LOD and LOQ results confirmed the validity of the HPLC analysis.

Fig. 2. Analysis of monosaccharide composition in ATP-W-1 by high-performance liquid chromatography (HPLC).

The monosaccharide composition was identified and quantified by comparing the HPLC chromatograph of ATP-W-1 and monosaccharide standard (1, mannose; 2, rhamnose; 3, glucuronic acid; 4, galacturonic acid; 5, glucose; 6, galactose; 7, xylose; 8, arabinose; 9, fucose).

Table 1.

Monosaccharide composition (molar ratio, %)

	Man	Rha	Glc	Gal	Ara
ATP-W-1	2.23	6.43	50.12	8.36	31.91

Methylation analysis

Methylation analysis was used to determine the type of glycosidic bonds in polysaccharides⁴⁷. The GC–MS results indicated that ATP-W-1 contained 10 partially methylated alditol acetates (PMAAs), suggesting a structure composed of eight distinct sugar residue fragments (Fig. S2). These fragments were identified as: T-Araf, 1,5-Araf, T-Glcp, 1,3-Rhap, 1,2,5-Araf, 1,4-Glcp, 1,4,6-Gal, 1,3,6-Manp (Fig. S3)^48–55. These sugar residues had a molar ratio of 3.29: 4.16: 2.42: 1.55: 2.80: 36.95: 3.81: 1.00, respectively. Among all sugar residues analyzed, 4-linked α-D-glucose was the most abundant, comprising 36.95 of the total sugar residues, which might be a primary structural component of ATP-W-1 (Table 2).

Table 2.

Methylation analyses of ATP-W-1

PMAAs	Major mass fragments (m/z)	Linkages	Molar ratios
1,4-Di-O-acetyl-1- deuterio-2,3,5-tri-O-methyl-l-arabinitol	59, 71, 87, 102, 118, 129, 161	l-Araf-(1→	3.29
1,4,5-Tri-O-acetyl-1-deuterio-2,3-di-O-methyl-l-arabinitol	59, 71, 87, 102, 118, 129, 189	→5)-l-Araf-(1→	4.16
1,5-Di-O-acetyl-1-deuterio-2,3,4,6-tetra-O-methyl-d-glucitol	59, 71, 87, 102, 118, 129, 145, 162, 205	d-Glcp-(1→	2.42
1,3,5-Tri-O-acetyl-1-deuterio-6-deoxy-2,4-di-O-methyl-l-mannital	59, 71, 87, 101, 118, 131, 190	→3)-l-Rhap-(1→	1.55
1,2,4,5-Tetra-O- acetyl-1-deuterio-3-O-methyl-l-arabinitol	59, 87, 100, 129, 159, 189	→2,5)-l-Araf-(1→	2.80
1,4,5-Tri-O-acetyl-1-deuterio-2,3,6-tri-O-methyl-D-glucitol	59, 87, 99, 118, 129, 159, 233	→4)-d-Glcp-(1→	36.95
1,4,5,6-Tetra-O-acetyl-1-deuterio-2,3-di-O-methyl-D-galactitol	59, 85, 102, 118, 127, 159, 261	→4,6)-d-Galp-(→	3.81
1,3,5,6-Tetra-O-acetyl-1-deuterio-2,4-di-O-methyl-d-mannitol	87, 101, 118, 129, 139, 189, 234	→3,6)-d-Manp-(1→	1.00

NMR spectroscopy analysis

The NMR measurements demonstrated the chemical shifts of hydrogen and carbon atoms for each significant sugar residue, allowing for the inference of the connectivity between sugar units (Fig. 3). The NMR spectra and GC–MS results confirmed that ATP-W-1 consists of eight sugar residues labeled A, B, D–G, I, and J. The complete assignment of the proton and carbon chemical shifts for ATP-W-1 was analyzed through 1D and 2D NMR spectra (Table S2). These results were validated by comparing the chemical shifts reported in previous studies^56–64.

Fig. 3. Characterization of ATP-W-1 by 1D and 2D-NMR spectra analysis.

¹H NMR spectra of ATP-W-1 (A), ¹³C NMR spectra of ATP-W-1 (B), DEPT 135 spectrum of ATP-W-1 (C), HSQC spectra (D), HMBC spectra (E), COSY spectra (F) of ATP-W-1.

Anomeric proton signals of α-configurational sugar residues were attributed to the six signals at 5.12, 5.03, 5.19, 4.88, 5.09, and 5.35 ppm, and those of β-configurational sugar residues at 4.48 and 4.60 ppm (Fig. 3A)^50,65–67. Based on the ¹³C NMR data, the anomeric carbons were located at 99.4, 99.6, 99.7, 103.4, 104.0, 106.8, 107.3, and 109.2 ppm (Fig. 3B). The analysis of the DEPT-135 NMR spectrum revealed inverted peaks at 60.3, 63.4, and 65.9 ppm, corresponding to the C-5 carbons of Residues A, B, and D, respectively (Fig. 3C). Meanwhile, the peaks observed at 59.8, 60.3, 65.8, and 66.0 ppm were identified as the C-6 carbon residues F, G, I, and J, respectively. Anomeric protons and carbon atoms were observed in eight cross-peaks in the HSQC spectrum of ATP-W-1 (Fig. 3D). These relatively stronger signals at δ 5.35/99.6 ppm were assigned to H-1/C-1 of →4)-α-d-Glcp-(1→ (residue G), which was found in high abundance in the GC-MS results. The 2D HSQC signals at 5.12/106.8, 5.03/107.3, 5.19/109.2 ppm, were assigned to H-1/C-1 of α-l-Araf-(1→ (residue A), →5)-α-l-Araf-(1→ (residue B), →2,5)-α-l-Araf-(1→ (residue D), respectively. These chemical shifts are attributed to arabinofuranose residues reported in plant polysaccharides isolated from Malus halliana³⁷, and Sida cordifolia⁶⁸. The signal at 4.88/99.4 ppm was assigned to →3)-α-l-Rhap-(1→ (residue E), while 5.09/99.7 ppm was attributed to α-d-Glcp-(1→ (residue F), both matching with previous reported plant polysaccharides from Euphorbia humifusa⁶⁹, Inonotus hispidus³⁸, and Calvatia gigantea⁷⁰. The signals at 4.48/103.3 and 4.60/104.0 ppm were attributed to H-1/C-1 of →4,6)-β-d-Galp-( → (residue I) and→3,6)-β-d-Manp-(1→ (residue J). These units are commonly found in branched plant polysaccharides isolated from Inonotus hispidus³⁸, Malus halliana³⁷, and Radix Puerariae thomsonii⁷¹. The signal at 16.6 ppm could be attributed to C-6 of →3)-α-l-Rhap-(1 → . These results further confirm the presence of 6-deoxy sugar residues. These data collectively indicate that ATP-W-1 contains a mixture of α- and β-linked pyranose and furanose residues.

In addition, the linkage sites and sequences of ATP-W-1 were further confirmed using HMBC and COSY spectra (Fig. 3E, F). The cross signal at 5.03/65.9 ppm (B H-1/D C-5) indicated that O-1 of residue B was linked to C-5 of residue D. The cross signals at 5.19/66.0 ppm (D H-1/J C-6), 5.09/76.5 ppm (F H-1/I C-4), 5.35/73.4 (G H-1/E C-3) and 5.12/84.9 (A H1/D C-2) showed that O-1 of residue D was connected to C-6 of residue J, O-1 of residue F was connected to C-4 of residue I, O-1 of residue G was connected to C-3 of residue E, and O-1 of residue A was connected to C-2 of residue D. The cross signal at 3.67/104.0 ppm (I H-6/J C-1), and 3.68/103.4 ppm (G H-4/I C-1) indicating that O-6 of residue I was connected to C-1 of residue J, and O-4 of residue G was connected to C-1 of residue I. Furthermore, ATP-W-1 appeared to contain recurring moieties 5.03/66.4 ppm cross signal (B H-1/B C-5), 5.19/65.9 ppm cross signal (D H-1/D C-5), and 3.68/99.6 ppm cross signal (G H-4/G C-1). The structure of ATP-W-1 was proposed based on these NMR and molecular weight results (Fig. 4).

Fig. 4. The predicted structure of ATP-W-1 is based on structural characterization analysis.

The proposed structure formula of ATP-W-1 (A) and the proposed structure symbols of ATP-W-1 (B).

ATP-W-1 ameliorated DSS-induced UC symptoms

To investigate the therapeutic potential of ATP-W-1 in UC, a mouse model was established using a 2.5% DSS solution; the experimental design is shown in Fig. 5A. A significant difference between the DSS-treated mice and the control mice was observed throughout the 3-day DSS administration period. As shown in Fig. 5B, the control group mice remained healthy and displayed no noticeable signs of perianal swelling or inflammation. In contrast, the model group exhibited significant UC symptoms, including severe perianal inflammation, redness, and swelling. The L-ATP-W-1 and M-ATP-W-1 treatments significantly improved UC symptoms by inhibiting these inflammatory symptoms. However, the mice treated with H-ATP-W-1 showed marked recovery, evidenced by minimal redness and swelling in the perianal region. The control group mice exhibited smooth, structurally intact, healthy colons free from inflammation or fibrosis, with a colon length of 8 cm. In contrast, the model group showed a marked reduction in colon length, increased swelling, structural abnormalities, severe inflammation, and tissue damage (Fig. 5C, D). This characteristic feature, such as inflammation and tissue injury in the colon of mice, is associated with DSS-induced UC⁷². Moreover, the L-ATP-W-1 and M-ATP-W-1-treated groups showed a moderate improvement in colon length, ranging from 5 to 6 cm. Although all three concentrations, including low, medium, and high of ATP-W-1 improved the colon morphology and reduced inflammation, the H-ATP-W-1 treatment significantly increased colon length, implying effective mitigation of pro-inflammatory cytokines (IL-6, IL-1β) and upregulation of tight junction proteins (ZO-1, Occludin)^73,74. These effects collectively reduce mucosal damage and tissue contraction, thereby maintaining colonic structure. Similarly, 5-ASA treatment improved colon length, paralleling the effects observed in the H-ATP-W-1 group. The Disease Activity Index (DAI) is a composite metric that evaluates the severity of colitis by considering weight loss, stool consistency, and bleeding⁷⁵. The results demonstrated that administration of DSS significantly elevated DAI scores evidencing the severe colitis (Fig. 5E).

Fig. 5. Experimental design of model DSS induced UC model development and ATP-W-1 treatment to ameliorate the symptoms of UC.

The experimental schedule and treatments strategies (A), pictorial representation of changes in mouse perianal region due to UC on day 10 (B), photographic visualization and measurement of colon length (C, D), analysis of DAI score (E), Body weight analysis at different time intervals (F), pictorial observation of the spleen from different treatment groups (G), and analysis of the spleen index (H). ^##p < 0.01, ^###p < 0.001, ^####p < 0.0001 vs. control group; *p < 0.05, **p < 0.01, ***p < 0.001 vs. DSS group; ^nsp > 0.05.

The body weight was significantly varied between the treatment groups (Fig. 5F). The results revealed that the control group maintained a consistent body weight during the study. In contrast, the model group resulted in significant body weight loss (Fig. 5F). The low and medium-dose treatments alleviated weight loss to some extent and gradually improved body weight over time. However, the H-ATP-W-1 treatment exhibited the most pronounced effect, with body weight nearing control group levels. The spleen morphology and spleen index revealed that the spleen collected from the control group exhibited normal size, indicating the absence of pathological enlargement or immune system activation, whereas the model group showed a marked increase in spleen size, indicating significant inflammation (Fig. 5G and H). However, the ATP-W-1 group showed that spleen size was significantly reduced compared to the model group, indicating that treatment effectively mitigated the inflammatory response (Fig. 5G). The spleen index was calculated as spleen weight relative to body weight (mg/g), which is a quantitative measure of spleen enlargement and systemic inflammation⁷⁶. The spleen index was significantly higher in the model group compared to the control group. The increased spleen index in the model group was associated with DSS-induced systemic inflammation. However, the ATP-W-1 treatment significantly reduced spleen index, indicating its efficacy in reducing inflammation and alleviating systemic immune responses (Fig. 5H). These findings suggest that the ATP-W-1 exerted a robust, dose-dependent therapeutic effect in mitigating DSS-induced UC.

Fig. 6. Histopathological evaluation of colon tissues in DSS-induced UC mice following ATP-W-1 treatment.

H&E staining (×100) (A), Alcian blue (AB) staining (B), and PAS staining (C) of colon tissues collected from different treatment groups. Quantification of staining intensity (D). ^###p < 0.001, ^####p < 0.0001  vs. control group; **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. DSS group.

Furthermore, the therapeutic activities of ATP-W-1 were compared with those of other plants or marine-originated polysaccharides reported in DSS-induced UC models. Fucoidan (M_w: 428) derived from Undaria pinnatifida alleviates UC through modulating the GM, enhancing intestinal barrier function, and suppressing inflammatory pathways such as MAPK/NF-κB⁷⁷. Another polysaccharide, pectins (M_w: 26–270 kDa) derived from Aconitum carmichaelii leaves, demonstrated a significant protective effect against DSS-induced UC through modulation of GM and TLR4 activation⁷⁸. β-D-glucan (M_w = 2.42 × 10⁶ g/mol) from Ganoderma lucidum significantly recovered mice from UC through inhibiting the pro-inflammatory cytokines (TNF-α, IL-1β, and IL-6) (Liu et al. 2025)⁷⁹. In contrast, ATP-W-1 (M_w: 164.5 kDa) showed similar or better effects on the recovery of GM diversity, SCFA production, and DAI scores by a relative dose (400 mg/kg), indicating that ATP-W-1 is a potential bioactive polysaccharide for UC treatment.

ATP-W-1 alleviated colonic injury in UC mice

H&E staining revealed that the control group mice exhibited intact and continuous colorectal walls, with clearly visible intestinal epithelial cells, regular and orderly arranged glandular structures, abundant goblet cells, and minimal acute inflammatory cell infiltration (Fig. 6A–D). In contrast, the model group displayed significant epithelial destruction, loss of crypt structure, depletion of goblet cells, and substantial infiltration of inflammatory cells. Following treatment with ATP-W-1, these pathological changes were markedly ameliorated, with the H-ATP-W-1 group demonstrating the most pronounced improvement, characterized by glandular arrangement and discernible basic tissue structure (Fig. 6A). AB-PAS staining results indicated abundant goblet cells in the control group that were distinctly visible. In contrast, there was a significant reduction in goblet cells in the model group. A substantial increase in goblet cells was observed after ATP-W-1 administration (Fig. 6B, C and D). Furthermore, the H & E staining of the organs indicated the non-toxic effect of ATP-W-1 administration while the model group showed a significant change in diseased organs (Fig. S4). These findings suggest that ATP-W-1 administration can effectively mitigate damage caused by DSS-induced UC in mice.

ATP-W-1 restores tight junction protein expression in DSS-induced colitis

To evaluate the impact of ATP-W-1 on intestinal barrier integrity, the expression levels of key tight junction (TJ) proteins, including Claudin-1, Occludin, and Occludens-1 (ZO-1), were assessed through IHC staining (Fig. 7A–E). In the control group, the abundant and continuous expression of Claudin-1, Occludin, and ZO-1 was observed along the epithelial lining, reflecting a well-maintained epithelial barrier. Conversely, the model group exhibited significantly disrupted architecture, as evidenced by diminished staining intensity and fragmented localization of TJ proteins, indicating a compromised barrier and increased paracellular permeability. Administration of ATP-W-1 notably mitigated the DSS-induced downregulation of these TJ proteins in a dose-dependent manner. Specifically, mice treated with medium and high doses of ATP-W-1 (M-ATP-W-1 and H-ATP-W-1) exhibited restoration of TJ protein expression, with the H-ATP-W-1 group achieving levels comparable to those seen in the 5-ASA-treated group. Quantitative analysis confirmed these observations (Fig. 7D), showing significant upregulation of Claudin-1, Occludin, and ZO-1 in ATP-W-1-treated groups. This suggests that ATP-W-1 restored the intestinal epithelial barrier by stabilizing tight junction complexes. To further investigate intestinal barrier function, FITC-dextran permeability assays were conducted to assess systemic leakage. The DSS group exhibited significantly elevated serum fluorescence, indicating increased intestinal permeability. In contrast, the ATP-W-1-treated groups showed markedly reduced fluorescence (Fig. 7E), supporting the protective effect of ATP-W-1 against barrier dysfunction. Together, these results highlight that ATP-W-1 not only restores tight junction proteins (ZO-1, Occludin, Claudin-1) but also recovers the mucosal layer through mucin preservation, thereby contributing to intestinal barrier repair via multiple mechanisms. The therapeutic potential of ATP-W-1 in UC is due to its ability to modulate both structural and biochemical components of epithelial cells.

Fig. 7. IHC analysis of intestinal tight junction proteins.

Claudin-1 (A), Occludin (B), and ZO-1 (C) in colon tissues of mice. Quantification of staining intensity for Claudin-1, Occludin, and ZO-1 (D). The level of FITC-dextran in serum (E). ####p < 0.0001 vs. control group; **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. DSS group.

Inflammatory signals in UC mice

Inflammation is a primary contributor to colonic damage in UC, and the severity of the condition is associated with levels of inflammatory factors and mediators⁸⁰. These proinflammatory signals compromise epithelial barrier function and enhance microbial translocation, exacerbating mucosal inflammation. Among them, TNF-α and IL-6 are key activators of the NF-κB and MAPK signaling pathways, which are well-characterized in UC for promoting transcription of proinflammatory genes and perpetuating immune cell activation^81,82. In contrast, IL-10 acts through TLR4-NF-κB negative loops to suppress inflammatory signaling and restore immune balance^83,84. The model group exhibited a significant upregulation of TNF-α and IL-6, along with suppression of IL-10, consistent with hyperactivation of the NF-κB and MAPK pathways (Fig. 8A–D). However, treatment with ATP-W-1 effectively reversed this pattern by reducing TNF-α and IL-6 levels, while restoring IL-10 expression, suggesting ATP-W-1 inhibits NF-κB and/or MAPK activation while enhancing anti-inflammatory signaling. This is further supported by IHC staining of colon tissues, which showed restored IL-10 levels in the ATP-W-1 group (Fig. 8A–D). Although direct pathway markers (p65 or p-ERK) were not measured in this study, the cytokine profile modulation strongly implies that ATP-W-1 attenuates colonic inflammation through inhibition of NF-κB and MAPK signaling cascades and potentially acts via the TLR4/IL-10 axis to promote immune tolerance.

Fig. 8. The anti-inflammatory effect of ATP-W-1 treatments in UC mice.

The expression of cytokines IL-6 (A), TNF-α (B), and IL-10 (C). Immunofluorescence analysis of inflammatory proteins in colon tissues collected from different treatment groups (D). ###p < 0.001, ####p < 0.0001 indicated the statistical difference between control and DSS group; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.001 indicated the statistical difference between DSS and ATP-W-1 treatment group.

ATP-W-1 modulated GM dysbiosis and SCFAs production in UC mice

The development of UC is associated with GM dysbiosis, which disrupts the symbiotic relationship between the host and its GMs, leading to inflammation and colon tissue damage⁸⁵. To examine the effect of ATP-W-1 treatment on α- and β- diversity of intestinal GM in DSS-induced UC mice, we employed 16S rDNA sequencing of the cecal contents of mice. The α- diversity indices such as species richness, diversity, complexity and evenness were assessed by Ace, Chao1, Simpson, and Shannon indices respectively (Fig. 9A). The results revealed that the model group exhibited a significantly lower level of Ace, Chao, Simpson and Shahnon indices than the control group, indicating that DSS exhibited a profound impact on the richness, diversity, complexity and evenness of GM and which is in agreement with earlier findings⁸⁶. However, the ATP-W-1 treatment group significantly increased these indices compared to the control group, implying that ATP-W-1 mitigated GM dysbiosis.

Fig. 9. Microbial diversity and composition analysis in different treatment groups based on 16S rRNA sequencing.

The comparison of Ace index, Chao index, Sobs index, Shannon index of different treatment group (A), PCoA analysis (B), and NMDS analysis (C). Phylum barplot analysis (D, E), Genus barplot analysis (F, G). ^##p < 0.01, ^###p < 0.001, ^####p < 0.001 vs. control group; **p < 0.01, ***p < 0.001, ****p < 0.0001 vs. DSS group; ^nsp > 0.05.

Principal coordinate analysis (PCoA) demonstrated that the control group, model group, and ATP-W-1 group were independently clustered (Fig. 9B). These results indicated that the OTUs of GM were significantly varied between the control group and model group. This distant shift in model group reflects substantial alterations in GM dysbiosis resulting from inflammation. Moreover, the ATP-W-1-treated group exhibited many beneficial bacterial OTUs in mitigating DSS-induced dysbiosis. Moreover, non-metric multidimensional scaling (NMDS) analysis demonstrated the β-diversity of GM in different experimental groups, such as control, model, and ATP-W-1 group (Fig. 9C). The results revealed that the control group formed a separate cluster compared to the model group, which displayed the disruptive impact of DSS on GM diversity. However, the ATP-W-1 treatment group clustering pattern was more closely aligned with the control group, demonstrating that ATP-W-1 contributes to restoring GM diversity.

The relative abundance of phylum and genus in different treatment groups was demonstrated in Fig. 9D–G. The results illustrated that the relative proportions of phylum and genus were significantly varied between the treatment groups. The GM in the control group comprised Firmicutes (52.80%) and Bacteroidetes (39.85%). In contrast, the model group exhibited a significant alteration in increasing the Firmicutes (67.95%) and decreasing the Bacteroidetes (14.52%) compared to the control group. The Firmicutes/Bacteroidetes (F/B) ratio is an indicator for evaluating the changes in microorganisms⁸⁷. The present study demonstrated that the F/B ratio was significantly higher in the DSS group (F/B 4.680) compared to the control group (F/B 1.32). The changes in F/B ratio indicated GM dysbiosis, and these findings are consistent with earlier results^88,89. However, the ATP-W-1 treatment modulated dysbiosis specifically; the F/B ratio in the ATP-W-1 group was restored to 2.037, indicating the partial recovery towards normal GM balance (Fig. 9E). Proteobacteria were found to be higher in the model group (10.52%) compared to the control group (1.50%). These results agreed with earlier studies, which demonstrated its increased abundance in UC^90,91. Moreover, H-ATP-W-1 treatment significantly reduced the relative abundance of Proteobacteria compared to the model group.

Figure 9F displays the GM composition at the genus level across different groups. The control group showed the predominance of beneficial genera, such as Lactobacillus, Muribaculum, and Turicibacter, which are associated with gut homeostasis and immune regulation^92–94. In the contract, the model group exhibited a significantly decreased Lactobacillus and Muribaculum level, reflecting the inflammation and dysbiosis^95,96. These changes are consistent with the phylum-level shifts observed, specifically increased Firmicutes and decreased Bacteroidetes, indicating microbial imbalance (Fig. 9E). Following treatment with H-ATP-W-1, the abundance of key probiotic genera significantly increased, restoring the microbiota profile closer to that of the control group (Fig. 9G). The reduction of inflammation-associated genera reflects an improvement in microbial balance and mitigation of dysbiosis⁹⁷. Lactobacillus and Muribacuum known for producing SCFAs (propionate and butyrate), recovered substantially after ATP-W-1 administration. Their abundance is critical for maintaining epithelial integrity and regulating inflammation, both compromised under DSS-induced UC^98,99. Besides, Turicibacter is recognized for its key roles in host immune modulation and inflammatory response⁹⁹. The marked increase in Turicibacter abundance due to ATP-W-1 treatment indicates anti-inflammatory effects. Meanwhile, Bifidobacteria had been previously reported to play a role in maintaining mucosal and epithelial barriers and regulating intestinal microbiota¹⁰⁰. In the model group, Bifidobacteria decreased significantly but were restored after ATP-W-1 treatment, suggesting their importance in maintaining gut homeostasis and overall health.

SCFA (acetic acid, propionic acid, and butyric acid) are key microbial metabolites produced by the fermentation of dietary polysaccharides. These compounds play crucial roles in maintaining intestinal homeostasis by enhancing epithelial barrier integrity, modulating immune responses¹⁰¹. Specifically, butyrate is known to upregulate tight junction proteins (ZO-1, occludin), reduce intestinal permeability, and exert anti-inflammatory effects by inhibiting histone deacetylases (HDACs) and suppressing NF-κB signaling. Acetate supports mucosal immunity and helps maintain epithelial function, while propionate has been shown to regulate T cell differentiation and reduce pro-inflammatory cytokine production. In the present study, ATP-W-1 treatment significantly restored the SCFA concentrations, including acetic acid, propionic acid, and butyric acid, which were notably decreased in the DSS-induced model group (Fig. 10A–C). The elevation of these SCFAs in the ATP-W-1-treated groups suggests a recovery of gut microbial fermentation capacity. This recovery correlates with improved intestinal barrier function, as seen by increased expression of tight junction proteins (ZO-1, occludin), and attenuation of inflammatory responses observed in colon tissue (Fig. 10D).

Fig. 10. The analysis of SCFA.

Concentration of acetic acid (A), propionic acid (B), and butyric acid (C) in different groups, correlation analysis of SCFA vs. primary UC markers (D). ^##p < 0.01 indicated the statistical difference between control and model group; *p < 0.05, **p < 0.01 indicated the statistical difference between model and H-ATP-W-1 group.

Given its demonstrated efficacy in modulating GM, restoring SCFA levels, and reducing colonic inflammation, ATP-W-1 shows potential as a candidate for incorporation into functional foods or nutraceutical supplements. Polysaccharides like ATP-W-1 can be formulated as prebiotic dietary fibers in capsules, powders, or beverages to improve gut health. Microencapsulation and spray-drying techniques may enhance ATP-W-1 stability during processing and gastrointestinal delivery. From a pharmaceutical perspective, its integration into colon-targeted delivery systems such as pH-sensitive hydrogels or enteric-coated tablets could ensure controlled release in the inflamed intestine. Future studies should explore bioavailability of ATP-W-1 and its safety, and optimal dosing to support its development as a dietary therapeutic or adjunct to conventional UC treatments.

In conclusion, this research investigated the structural characteristics and therapeutic potential of ATP-W-1, a water-soluble polysaccharide derived from the fruit of A. triloba. Structural analysis revealed that ATP-W-1 was primarily composed of →4)-α-D-Glcp-(1→ linkages, with glucose and arabinose as the major monosaccharide components. In a DSS-induced colitis model, ATP-W-1 treatment significantly alleviated UC symptoms by preventing weight loss, preserving colonic tissue architecture, and enhancing intestinal barrier function. Mechanistically, these protective effects were associated with restoring SCFA (acetic, propionic, and butyric acids), regulating inflammatory cytokines (TNF-α, IL-6, IL-10), and rebalancing the GM composition. Despite these promising findings, this study has some limitations. The efficacy of ATP-W-1 was assessed in a single animal model, and dose-response relationships were not fully explored. The long-term safety, pharmacokinetics, and optimal delivery strategies remain to be investigated. Future studies should focus on validating these effects in chronic colitis models, conducting dose-optimization experiments, and evaluating the translational potential of ATP-W-1 in human clinical trials. Overall, ATP-W-1 is a promising candidate for developing therapeutic agents or functional food supplements targeting UC by modulating the gut microenvironment and immune responses.

Methods

Materials and reagents

Fresh A. triloba fruits were sourced from the Korean Pawpaw Tree Cooperative (Jeollanam-do, South Korea). DEAE-Sepharose Fast Flow and Sephadex G-75 were procured from GE Healthcare Ltd. (Stockholm, Sweden). Standard monosaccharides were provided by DeSiTe Biological Technology Co., Ltd. (Chengdu, China). Dextran sulfate sodium salt (DSS, M_w: 36–50 kDa) was purchased from MP Biomedicals (OH, USA). All other chemicals of analytical grades were sourced from Daejung Chemicals and Metals (Siheung-si, Gyeonggi-do), Republic of Korea.

Isolation and purification of ATP-W-1

The unpeeled A. triloba fruit (10 kg) was extracted thrice with distilled H₂O at 80 °C, each extraction for 3 h. The resulting extract was concentrated and precipitated with 95% ethanol, then stored at 4 °C for 24 h. The precipitates were collected by centrifugation at 5000 rpm for 10 min and dissolved in distilled H₂O. The solution was deproteinized using the Sevage solution (n-butanol: dichloromethane, 1:4 v/v), then dialyzed (M_w cutoff: 3500 Da), concentrated using a rotary evaporator at 50 °C, and freeze-dried to obtain the crude polysaccharide, designated as ATP. The ATP (50 mg/mL) was fractionated using a DEAE-sepharose fast flow (DEAE-FF) column (⌀ 3.5 × 40 cm), sequentially eluting with water, and NaCl solutions (0.1, 0.2, and 0.3 M). The fractions (10 mL/tube) were collected using an automatic collector (CombiFlash NextGen 100 apparatus), and total carbohydrate content was measured by phenol-sulfuric acid assay using a UV spectrophotometer (Optizen 2120), measuring an OD at 490. The water fraction was labeled as ATP-W, while the fractions eluted with 0.1 and 0.2 M NaCl were labeled ATP-1 and ATP-2, respectively. The primary fraction of ATP-W was further purified using a Sephacryl S-300/HR column (⌀ 16 × 600 mm) eluted with distilled H₂O at a 0.2 mL/min flow rate. The resulting purified fraction, ATP-W-1, was freeze-dried and stored at 4 °C for subsequent experiments.

SEM analysis of ATP-W-1

The microstructural features of ATP-W-1 were observed by a field emission scanning electron microscope (FE-SEM; Gemini 500 + EDS (Oxford). The samples were coated with gold and detected with a 2 kV accelerating voltage.

Monosaccharide composition analysis of ATP-W-1

5 mg of ATP-W-1 was hydrolyzed in 2 mL of 2 M trifluoroacetic acid (TFA) at 110 °C for 4 h. After cooling at room temperature, anhydrous methanol was added, and the solution was repeatedly evaporated under reduced pressure to remove residual TFA. The resulting hydrolysate was derivatized in 0.5 M of 1-phenyl-3-methyl-5-pyrazolone (PMP) in methanol solution under alkaline conditions at 70 °C for 30 min. The reaction was terminated by adding 0.3 M hydrochloric acid, and the mixture was extracted twice with chloroform to remove the PMP. The aqueous phase was collected and filtered through a 0.22 µm membrane filter and used for high-performance liquid chromatography (HPLC) analysis. The HPLC analysis was carried out using a Phenomenex C18 column (250 × 10 mm), with a mobile phase composed of 0.05 M phosphate buffer solution (pH 6.70) and acetonitrile (80:20 v/v), at a flow rate of 1.0 mL/min, column temperature of 30 °C, and UV detection at 250 nm. A standard mixture of monosaccharides (mannose, rhamnose, galacturonic acid, glucuronic acid, glucose, galactose, xylose, arabinose, and fucose) was derivatized with PMP and analyzed under the same conditions to determine the retention times.

HP-GPC and FTIR analysis

The molecular weight of ATP-W-1 was measured using a high-performance gel-permeation chromatography (HP-GPC) system according to the method reported in previous work¹⁰². In brief, ATP-W-1 (1 mg/mL) was dissolved in NaNO₃ (0.1 M) and filtered through a 0.45 μm filter paper for HP-GPC analysis. The functional groups of ATP-W-1 were analyzed by Fourier transform infrared (FT-IR) spectroscopy (PerkinElmer Paragon 500). For the analysis, the dried powder of ATP-W-1 (2 mg) was mixed with potassium bromide (KBr) powder (200 mg) to form pellets. These pellets were analyzed using FT-IR at wavelength ranges from 4000 to 400 cm⁻¹.

Methylation analysis

The dried sample (5 mg) was dissolved in anhydrous dimethyl sulfoxide (DMSO) by sonication for 30 min. Subsequently, NaOH (400 mg) was added to anhydrous DMSO (5 mL). Methyl iodide (CH₃I, 1.5 mL) was added to the reaction mixture three times in the dark conditions. To terminate the methylation reaction, 1 mL of distilled water was added. Subsequently, 2 mL of chloroform was added, and the mixture was shaken vigorously. The chloroform layer was then extracted three times with an equal volume of distilled water to obtain the methylated product of ATP-W-1. The methylated product was then hydrolyzed with 2 M trifluoroacetic acid (TFA), and the excess TFA was removed by co-distillation with methanol. The hydrolyzate was dissolved in distilled water, sodium borodeuteride (NaBD₄) was added, and the hydrolysate was reduced at 40 °C for 30 min. Acetic acid was then added to terminate the reduction reaction, and the resulting solution was concentrated to dryness. The final product reacted with acetic anhydride and pyridine at 95 °C for one hour to obtain partially methylated alditol acetate (PMAA). The resulting sample was concentrated, dissolved in chloroform, and extracted three times with distilled water to remove impurities. Finally, these acetylated derivatives were analyzed using GC–MS (Agilent 7890A-5977 B).

Nuclear magnetic resonance (NMR) spectroscopy analysis

The freeze-dried ATP-W-1 (60 mg) was dissolved in 0.5 mL of D₂O, and both 1D NMR spectra, such as ¹H, ¹³C, and DEPT-135, and 2D NMR spectra, such as ¹H–¹H COSY, HSQC, and HMBC, were recorded at 25 °C using a Bruker 400 MHz NMR (Bruker, Germany). All the spectra data were processed for the structure prediction of ATP-W-1 using Bruker software and MestNova.

Thermogravimetric analysis

The thermal stability of the ATP-W-1 was assessed using thermogravimetric analysis (TGA) using an STA 449 F3 from NETZSCH-Gerätebau GmbH (Germany). 5 mg each sample was heated from 50 to 800 °C at a constant ramp rate of 5 °C/min, purged with nitrogen gas at 50 mL/min, and placed in 70 μL Al₂O₃ crucibles for analysis.

Effect of ATP-W-1 treatment on DSS-induced UC mice

Healthy male mice (C57BL/J) 6–8 weeks old weighing 20–25 g were procured from Samtako Co., Ltd., Korea. Before beginning the experiments, the mice were acclimatized for a week (temperature: 23 °C ± 2 °C, relative humidity: 40–75%, light–dark cycle: 14/10 h, and free access to water and feed). All animal experiments were conducted in accordance with the guidelines of the National Research Council’s Guide for the Care and Use of Laboratory Animals, and the study protocol was approved by the Ethical Committee of Animal Use of the College of Pharmacy, Chonnam National University, Korea (approval number: CNU IACUC-YB-2024-115). The mice were randomly assigned to treatment groups (n = 6/group) as follows: Control, model (2.5% DSS), L-ATP-W-1 (ATP, 100 mg/kg), M-ATP-W-1 (ATP, 200 mg/kg), H-ATP-W-1 (ATP, 400 mg/kg), and 5-ASA (Mesalazine, 200 mg/kg). To induce UC, 2.5% DSS was administered in the drinking water for 3 days, a concentration widely used to induce reproducible, moderate colitis that resembles human UC pathology without severe toxicity or mortality^75,103. After DSS exposure, each treatment group received the corresponding treatments for 10 days. The parameters, including body weight, fecal status, and occurrence of bloody stools, were recorded at regular intervals during the animal experiment. Body weight loss was calculated based on initial and subsequent measurements. The Disease Activity Index (DAI) was assessed based on weight loss, diarrhea severity, and bleeding. On day 14, mice were fasted for five hours, then anesthetized with inhaled isoflurane (3–5% for induction, maintained at 1–2%) to ensure complete unconsciousness. The depth of anesthesia was confirmed by a loss of reflex response (toe pinch), after which euthanasia was performed by cervical dislocation in accordance with IACUC guidelines. Colon lengths and spleen index were measured. Blood, colon, spleen, kidney, liver, cecal contents, and entire colorectum were collected for histology analysis. Serum levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), and interleukin-10 (IL-10) were quantified using enzyme-linked immunosorbent assay (ELISA) kits according to the manufacturer’s instructions.

Assessment of UC and histopathology

Colon tissue sections were fixed, dehydrated, and embedded in paraffin. These sections were then stained with H&E and AB-PAS¹⁰⁴. Followed by immunohistochemistry (IHC) staining was conducted according to the methods reported earlier^105,106, with minor modifications. In brief, tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval by boiling in a citrate buffer at 100 °C. The sections were then incubated overnight at 4 °C with primary antibodies, including TNF-α (1:300, Abcam, 52B83, ab1793), IL-6 (1:500, Abcam, EPR23819-103, ab290735), IL-10 (1:300, Bioss, BS-20373R), ZO-1 (1:100, Invitrogen, 61-7300), Occludin (1:200, Invitrogen, 33-1500), and Claudin-1 (1:500, Novusbio, NBP1-77036). After incubation with the primary antibodies, sections were washed with PBS and treated with secondary antibodies (rabbit anti-mouse IgG H&L (HRP), Abcam, ab6728) at 37 °C for 1 h. Following further PBS washes, staining was developed using 3,3-diaminobenzidine (DAB) and counterstained with hematoxylin. Images were captured using a full-slide scanner (ZEISS Axio Scan.Z1).

16 rDNA sequencing

The fecal samples were collected from different treatment groups in sterile tubes and stored at −80 °C. The genomic DNA was isolated and quantified using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s guidelines for subsequent next-generation sequencing (NGS) analysis. PCR amplification was conducted with the StepOne Plus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) under the following conditions: an initial denaturation at 95 °C for 3 min, followed by 25 cycles of 30 s at 95 °C, 30 s at 55 °C, and 30 s at 72 °C. The NGS library was prepared using the Illumina iSeq library preparation protocol (Illumina, San Diego, CA, USA). Amplicons were purified with AMPure beads to eliminate residual primers. An additional eight cycles of PCR were performed with Illumina barcoded adapters to create sequencing libraries. Following the manufacturer’s instructions, the pooled amplicons were sequenced using the Illumina iSeq100 system (Illumina, San Diego, CA, USA). For phylogenetic analysis and taxonomic classification, 16S rRNA gene sequences were processed using the Microbiome Taxonomic Profiling (MTP) pipeline available on EzBioCloud (https://www.ezbiocloud.net/contents/16smtp; CJ Bioscience, Seoul, Korea).

Short-chain fatty acids (SCFA) analysis

To measure SCFA concentrations, ~100 mg of cecal content was collected, mixed with 1 mL of 0.5% phosphoric acid, vortexed thoroughly, and centrifuged at 12,000 rpm for 10 min at 4 °C. The supernatant was filtered through a 0.22 µm membrane filter before injection into the GC–MS system. The production of acetic acid, butyric acid, and propionic acid was quantified using an Agilent 7890A GC–MS system equipped with flame ionization detection and a DB-FATWAX Ultra Inert column (30 m × 0.32 mm × 0.25 μm, Agilent).

Statistical analysis

Statistical analyses were performed using GraphPad Prism 8.0 (GraphPad Software, USA). Data are presented as means ± standard deviation (SD). The means were evaluated using t-tests or two-way analysis of variance (ANOVA). A multiple comparison test was performed on all the treated groups compared with DSS, and the DSS group was compared with the control group using Tukey’s post hoc test.

Author contributions

Zijun Li: Writing—review and editing, writing—original draft, methodology, investigation, data curation, and conceptualization. Kandasamy Saravanakumar: Writing—review and editing, writing—original draft, methodology, investigation, data curation, and conceptualization. Lulu Yao: Investigation and formal analysis. Yunyeong Kim: Investigation and formal analysis. Sang Yoon Choi: Formal analysis. Guijae Yoo: Formal analysis. Phil Jun Leec: Investigation. Soeun Kim: Formal analysis. Namki Cho: Writing—review & editing, writing—original draft, visualization, supervision, resources, project administration, formal analysis, data curation, and conceptualization.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41538-025-00587-5.