Topical Garcinia Mangostana Pericarp Extract Attenuates TPA-Induced Cutaneous Inflammation and Oxidative Stress in Mice: Preclinical Implications for Tumor Promotion-Related Inflammatory Mechanisms

Abstract

Purpose

Persistent cutaneous inflammation and oxidative stress drive the promotion stage of skin carcinogenesis. Most evidence for Garcinia mangostana pericarp comes from purified α-mangostin tested in vitro or given systemically. We evaluated whether a standardized crude G. mangostana pericarp extract (GMPE) applied topically produces concentration-related anti-inflammatory and antioxidant effects in acute 12-O-tetradecanoylphorbol-13-acetate (TPA)-induced skin inflammation in mice.

Methods

Thirty-five male Swiss albino mice were randomized to seven groups (n = 5): negative control, TPA control, TPA + vehicle cream, TPA + 1% hydrocortisone, and TPA + GMPE cream at 2.5%, 5%, and 10%. GMPE was standardized for α-mangostin by ultra-performance liquid chromatography. Cutaneous tumor necrosis factor-alpha (TNF-α; ELISA) and malondialdehyde (MDA; TBARS assay) were measured, and epidermal thickness, neutrophil infiltration, and immunohistochemical TNF-α and MDA expression (immunoreactive score, IRS) were assessed by a blinded observer.

Results

GMPE contained 17.54 ± 0.03% (w/w) α-mangostin. TPA raised TNF-α from 22.03 ± 11.43 to 77.03 ± 17.50 pg/mL and MDA from 0.48 ± 0.13 to 0.92 ± 0.10 nmol/mL, increased epidermal thickness fivefold, and induced neutrophil infiltration (all p < 0.001). GMPE 10% lowered TNF-α to 24.81 ± 11.85 pg/mL and MDA to 0.48 ± 0.10 nmol/mL, values not different from healthy skin or from hydrocortisone, and restored epidermal thickness (50 ± 20 µm). Immunohistochemistry confirmed these changes. A linear concentration–response was significant for MDA (p = 0.001).

Conclusion

In this experimental animal study, topical GMPE exerted concentration-related anti-inflammatory and antioxidant effects in acute TPA-induced skin inflammation, and the 10% cream performed comparably to 1% hydrocortisone. Because signaling pathways and tumor endpoints were not measured, GMPE may serve as a promising preclinical candidate for mechanistic and long-term chemoprevention studies rather than a proven modulator of tumor promotion.

Introduction

Skin cancer is the most common malignancy worldwide, with about 1.5 million new cases each year, of which roughly 330,000 are melanoma and 1.2 million are non-melanoma skin cancers.¹ Cutaneous carcinogenesis proceeds through initiation, promotion, and progression. Initiation depends on genetic damage, whereas promotion is a prolonged and potentially reversible phase that is sustained by chronic inflammation and oxidative stress and is therefore a rational target for chemoprevention.^2–4

Two measurable processes link inflammation to tumor promotion. Tumor necrosis factor-alpha (TNF-α) is an early mediator that activates nuclear factor-kappa B (NF-κB) and thereby sustains cytokine production, leukocyte recruitment, keratinocyte proliferation, and survival signaling.^5,6 In parallel, reactive oxygen species (ROS) attack membrane polyunsaturated fatty acids, and the resulting lipid peroxidation yields malondialdehyde (MDA), a stable and widely used marker of oxidative damage.⁷ Cytokine signaling and oxidative damage reinforce each other, creating a self-perpetuating loop that favors neoplastic progression.^8,9

Topical 12-O-tetradecanoylphorbol-13-acetate (TPA) is the classic tumor promoter used in the two-stage mouse skin carcinogenesis model. TPA mimics diacylglycerol, activates protein kinase C, and triggers activator protein-1, mitogen-activated protein kinase, and NF-κB signaling, which leads to cytokine release, ROS generation, neutrophil infiltration, and epidermal hyperplasia.^2,3 Repeated short-term TPA application reproduces these promotion-associated inflammatory changes within days and is widely used to screen candidate anti-inflammatory and chemopreventive agents.^10,11 The model captures the inflammatory component of promotion but does not itself measure tumor formation, a distinction that frames the interpretation of the present study.

Garcinia mangostana L. (Clusiaceae), the mangosteen, has a long history of topical use in Southeast Asia for skin infections, wounds, and localized inflammation, usually as a poultice or decoction of the pericarp.^12,13 The pericarp is rich in prenylated xanthones, of which α-mangostin is the most abundant; γ-mangostin, gartanin, garcinone E, and related compounds are also present.^12,14 α-Mangostin scavenges ROS, blocks NF-κB nuclear translocation, and reduces cyclooxygenase-2 (COX-2) and cytokine expression in activated macrophages,^15,16 and systemic α-mangostin lowered tumor burden and inflammatory markers in the two-stage 7,12-dimethylbenz[a]anthracene (DMBA)/TPA model through the phosphoinositide 3-kinase (PI3K)/Akt/mammalian target of rapamycin (mTOR) axis.¹⁷ A phenolic-rich pericarp extract also showed anti-skin cancer properties in vitro.¹⁸

Topical formulations of mangosteen have been examined less often. Our group previously reported that a topical pericarp extract cream reduced neutrophil infiltration and epidermal thickening in TPA-treated mouse skin, but that study used histology alone, an extract that was not chemically standardized, and no active comparator.¹⁹ Other topical preparations include a pericarp extract that attenuated inflammatory-cell infiltration in a murine superficial skin infection model,²⁰ pericarp extract-loaded liposomes,²¹ α-mangostin nanoparticles that improved acne in a split-face human trial,²² an α-mangostin nanoparticle gel intended for skin cancer,²³ and a cellulose nanofiber film containing the ethanolic pericarp extract.²⁴ Topical botanicals with combined anti-inflammatory and antioxidant activity are of clinical interest because they may reduce reliance on corticosteroids in inflammation-associated skin disorders, and because chemoprevention of skin cancer is applied to the skin surface, where local drug concentrations can be high.^14,25

We chose a crude standardized extract rather than purified α-mangostin for three reasons. First, the crude pericarp extract is the preparation that traditional practice and low-cost local production actually use. Second, the co-extracted xanthones have complementary anti-inflammatory and antioxidant activity, so a whole extract may act on more than one target,^26–28 which cannot be tested with a single purified compound. Third, purified α-mangostin is expensive and poorly water-soluble, whereas standardizing the extract to its α-mangostin content preserves reproducibility at lower cost. Topical rather than systemic administration was chosen because the skin is the target organ, because α-mangostin has very low oral bioavailability and extensive first-pass metabolism,²⁹ and because lipophilic xanthones partition readily into the stratum corneum, so topical delivery achieves local concentrations that systemic dosing cannot.^14,23

The knowledge gap is therefore specific: it is not known whether a chemically standardized crude G. mangostana pericarp extract (GMPE), applied topically, modulates the biochemical, histological, and immunohistochemical markers of promotion-associated inflammation, or how it compares with a standard topical corticosteroid. We hypothesized that topical GMPE would reduce cutaneous TNF-α, MDA, epidermal hyperplasia, and neutrophil infiltration in a concentration-dependent manner and that the highest concentration would be comparable to 1% hydrocortisone. The aim of this study was to test this hypothesis in an acute TPA-induced model of skin inflammation in mice. The study was designed to characterize the concentration–response of the whole extract; it did not measure signaling proteins or tumor endpoints, and its findings are interpreted within that limit.

Materials and Methods

Study Design and Ethical Approval

This was a randomized, blinded, controlled, post-test-only experimental study reported in accordance with the ARRIVE 2.0 guidelines.³⁰ The protocol was approved before the study began by the Ethics Review Committee of the Faculty of Medicine, Hasanuddin University (approval no. 289/UN4.6.4.5.3.1/PP66/2025, issued 3 April 2025). This committee is the institutional body responsible for reviewing all health research at the faculty, including research using laboratory animals; Universitas Hasanuddin does not operate a separate animal-only committee, and the approval covered the use of mice in this protocol. All procedures followed the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the principles of replacement, reduction, and refinement (3Rs).³¹ The animal experiments were carried out between April and June 2025. All chemicals were of analytical or chromatographic grade; suppliers, catalogue numbers, and purities of all reagents, kits, and antibodies are listed in Supplementary Table S5.

Plant Material, Botanical Identification, and Extraction

Ripe fruits of G. mangostana were collected in September 2024 from cultivated trees in Bua District, Luwu Regency, South Sulawesi, Indonesia (3°28′12″ S, 120°10′36″ E; 92 m above sea level) with the landowner’s permission. The species is a widely cultivated fruit crop that is neither protected nor listed under the Convention on International Trade in Endangered Species, and no additional institutional or governmental permit was required for its collection or use under Indonesian regulations. Botanical identity was confirmed by Dr. Muhammad Ridwan, plant taxonomist, Laboratory of Plant Taxonomy and Systematics, Department of Biology, Faculty of Mathematics and Natural Sciences, Universitas Hasanuddin, and a voucher specimen (accession no. GM/UH-BIO/2024-0117) was deposited in the Herbarium of the Department of Biology, Universitas Hasanuddin, Makassar, where it is publicly accessible on request.

Pericarps were separated within 24 h of harvest, washed, cut into pieces of about 1 cm, and dried in a cabinet air dryer protected from light at 40–50°C to constant weight (72 h; residual moisture 8.4% w/w by loss on drying). The dried pericarp was milled in short pulses and passed through a 60-mesh (250 µm) sieve. The powder was macerated in 96% ethanol (Merck, Darmstadt, Germany; 1:10 w/v) for 72 h at 28 ± 2°C with intermittent agitation, filtered through Whatman No. 1 paper, and the marc was re-macerated twice under the same conditions. The combined filtrates were evaporated under reduced pressure in a rotary evaporator at 40°C to give a dark reddish-brown viscous crude extract. The extraction yield was 11.8% (w/w) relative to the dried pericarp powder. A single batch of extract was used for all experiments; it was stored in nitrogen-flushed amber vials at −20°C, protected from light, until formulation. Post-harvest handling, drying, milling, and storage are detailed in Supplementary Methods S1.

Preparation of the Topical Formulations

The vehicle was an oil-in-water vanishing cream base composed of stearic acid 12.0%, cetyl alcohol 1.0%, liquid paraffin 4.0%, propylene glycol 8.0%, triethanolamine 1.5%, methylparaben 0.18%, propylparaben 0.02%, and purified water to 100% (w/w); all excipients were of pharmacopoeial grade (Bratachem, Jakarta, Indonesia). The crude extract was first dispersed in propylene glycol (2.0% w/w of the finished cream) and incorporated into the cooled base by geometric dilution, the amount of base being reduced correspondingly, to give creams containing 2.5%, 5%, or 10% (w/w) GMPE. The vehicle control cream was the identical base containing the same added propylene glycol but no extract, so that the treatment arms differed in extract content alone. One batch of each cream was prepared for the study. The pH of every batch was between 5.5 and 6.5, and all batches were homogeneous, free of grittiness, and showed no phase separation after centrifugation at 1500 × g for 30 min. Creams were stored in amber jars at 25 ± 2°C and used within 14 days. Batch-to-batch consistency was not assessed because a single batch of each formulation was used. The preparation procedure is described in Supplementary Methods S2.

Standardization of the Extract by UPLC

α-Mangostin was quantified in the crude extract by ultra-performance liquid chromatography (UPLC) on a Waters ACQUITY UPLC H-Class system (Waters Corporation, Milford, MA, USA) fitted with an ACQUITY UPLC BEH C18 column (1.7 µm, 2.1×100 mm) maintained at 25 ± 1°C, with the autosampler at 10 ± 1°C. Mobile phase A was 0.1% (v/v) orthophosphoric acid in water and mobile phase B was acetonitrile, delivered at 1.0 mL/min with a 2 µL injection volume and photodiode array detection with the chromatogram extracted at 320 nm; the gradient program is given in Supplementary Table S1. α-Mangostin eluted at 7.42 min in a 12-min run and was identified by retention-time matching with an authentic reference standard (≥98%; Sigma-Aldrich, St. Louis, MO, USA; cat. no. M3824); no peak-purity assessment was performed. Quantification used an external calibration curve prepared from six standard concentrations spanning 5–80 µg/mL, each injected in triplicate; the regression of peak area (y) on concentration (x, µg/mL) was y = 21,800x + 8,438 with a coefficient of determination of 0.9998. The limits of detection and quantification, calculated from the standard deviation of the response and the slope, were 0.42 and 1.27 µg/mL. System suitability was verified before each sequence (theoretical plates > 5,000, tailing factor < 1.5, relative standard deviation of replicate standard injections < 2%). Two portions of the crude extract were weighed independently (23.4 mg each), dissolved in methanol with sonication for 15 min, made up to 100 mL (nominal 234 µg/mL), filtered through a 0.22 µm polytetrafluoroethylene membrane, and each solution was injected in duplicate. Data were processed in Empower 3 software (Waters). Representative chromatograms of the reference standard and of the extract are provided in Supplementary Figure S1. Accuracy by spike recovery was not assessed, and precision was estimated from two independently prepared solutions of one batch, which constrains the reported precision (see Limitations). Calibration and sample preparation are detailed in Supplementary Methods S3.

Animals, Sample Size, Randomization, and Blinding

Thirty-five healthy male Swiss albino mice (8–10 weeks old, 25–35 g) were obtained from an accredited breeding facility and acclimatized for 7 days at 28 ± 2°C and 55 ± 10% relative humidity under a 12-h light/dark cycle with free access to standard chow and water. A single sex was used to limit variance in a small cohort. The group size was determined with the Federer formula, (t − 1)(n − 1) ≥ 15, which for seven groups requires at least four animals per group; one animal was added per group to allow for attrition, giving n = 5 (35 animals). No formal a priori power analysis was performed. A sensitivity analysis performed during revision showed that with seven groups, five animals per group, and α = 0.05, a one-way analysis of variance has 80% power to detect an effect of Cohen’s f ≥ 0.70 (η² ≥ 0.33); the design is therefore powered for large effects only (see Statistical Analysis).

Animals were allocated to the seven groups by block randomization using a computer-generated random sequence prepared by an investigator who was not involved in treatment administration or outcome assessment. Cages were identified by code only, and the allocation list was kept by that investigator and not disclosed to the personnel performing the biochemical assays or the histopathological and immunohistochemical scoring until all data had been entered. All histological and immunohistochemical evaluations were performed by a pathologist blinded to group allocation. All 35 animals completed the protocol; there were no deaths or exclusions. The groups were: (1) negative control (no TPA, no treatment); (2) TPA control (TPA, no treatment); (3) TPA + vehicle cream; (4) TPA + 1% hydrocortisone cream; (5) TPA + GMPE 2.5%; (6) TPA + GMPE 5%; and (7) TPA + GMPE 10%.

Induction of Skin Inflammation and Treatment Protocol

Dorsal hair was clipped over an area of about 2×2 cm 24 h before induction. TPA (Sigma-Aldrich; cat. no. P1585; ≥99%; 10 nmol dissolved in acetone) was applied to the clipped area at 0, 12, and 24 h (three applications). This multiple-application regimen was chosen because a single dose of TPA produces a transient edematous response, whereas repeated application over 24 h produces sustained epidermal hyperplasia, neutrophil infiltration, and cytokine expression that persist for several days and more closely resemble the inflammatory milieu of tumor promotion.^2,10,11 The regimen follows the protocol used in our earlier study of this extract.¹⁹ Topical treatment began 30 min after the last TPA application, so that the interventions were tested therapeutically on established inflammation rather than prophylactically, and was continued twice daily for three consecutive days. Each application delivered approximately 2 mg of cream per cm² of skin (about 8 mg per application over the 4 cm² treatment area), spread evenly over the treated area. Animals in the negative control and TPA control groups received no cream. The vehicle group received the unmedicated cream base, the positive control group received a commercially available 1% hydrocortisone cream (PT Kalbe Farma Tbk, Jakarta, Indonesia), which is the conventional low-potency reference comparator in TPA models, and the treatment groups received GMPE cream at 2.5%, 5%, or 10%.

The three GMPE concentrations were spaced twofold to permit a concentration–response analysis and were taken from the range (2.5–10%) over which our earlier study of the same extract observed graded histological responses in TPA-treated mouse skin.¹⁹ Twenty-four hours after the final application, mice were deeply anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg) (ketamine–xylazine solution; Sigma-Aldrich; cat. no. K113), and once the pedal withdrawal reflex was lost, cervical dislocation was performed as a secondary method to ensure death, in accordance with the American Veterinary Medical Association Guidelines for the Euthanasia of Animals (2020).³²

Tissue Collection and Biochemical Assays

Dorsal skin from the treated site was excised and divided into two halves. One half was snap-frozen and stored at −80°C for biochemical assays; the other was fixed in 10% neutral-buffered formalin for histology. Frozen tissue was homogenized in phosphate-buffered saline (pH 7.4) containing a protease inhibitor cocktail and centrifuged at 10,000 × g for 15 min at 4°C, and the supernatant was used for the assays. TNF-α was measured with a mouse-specific sandwich ELISA kit (Mouse TNF alpha Uncoated ELISA Kit; Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA; cat. no. 88–7324-22) according to the manufacturer’s protocol, reading absorbance at 450 nm on a microplate reader; concentrations were interpolated from a standard curve run on each plate. MDA was measured by the thiobarbituric acid reactive substances (TBARS) method with a TBARS Assay Kit (Cayman Chemical, Ann Arbor, MI, USA; cat. no. 10009055): homogenate supernatants were reacted with thiobarbituric acid under acidic conditions at 95°C for 60 min, cooled, and read at 532 nm, and MDA was interpolated from a standard curve. Both assays were run in duplicate wells for each of the five animals per group, and the mean of the duplicates was used. Results are expressed per mL of homogenate supernatant (pg/mL for TNF-α, nmol/mL for MDA); values were not normalized to tissue protein (see Limitations).

Histopathological Evaluation

Formalin-fixed tissue was embedded in paraffin, sectioned at 4–5 µm, and stained with hematoxylin and eosin (H&E). Sections were examined with an Olympus CX43 bright-field light microscope (Olympus Corporation, Tokyo, Japan) fitted with a digital camera, and images were analyzed with Olympus cellSens Standard software; the imaging system was calibrated using a stage micrometer at each magnification before measurement. Neutrophil infiltration was graded on a semi-quantitative 0–3 scale (0, absent; 1, scattered neutrophils; 2, moderate focal aggregates; 3, dense diffuse infiltrate) in five random non-overlapping high-power fields (400×) per section, and the mean grade was used as the per-animal value. Epidermal thickness was measured digitally from the basal layer to the stratum corneum with the calibrated software at ten positions in each of five non-overlapping fields per section, and the mean of the 50 measurements was used as the per-animal value; values are reported to the nearest 10 µm, the resolution of the measurement. Edema, epidermal hyperplasia, and increased vascular permeability (dermal vasodilatation with perivascular extravasation) were recorded as present or absent for each animal. All assessments were made by one pathologist blinded to group allocation. Representative images were captured at 100× (H&E) and 400× (immunohistochemistry).

Immunohistochemistry

Serial 4 µm sections were deparaffinized and subjected to heat-induced epitope retrieval in 10 mM citrate buffer (pH 6.0) for 20 min at 95°C. Endogenous peroxidase was quenched with 3% hydrogen peroxide for 10 min and non-specific binding was blocked with 5% normal goat serum for 30 min. Sections were incubated overnight at 4°C with a rabbit polyclonal anti-TNF-α antibody (Abcam, Cambridge, UK; cat. no. ab6671, RRID AB_305641; 1:200) or a mouse monoclonal anti-MDA antibody (clone 1F83; JaICA, Shizuoka, Japan; cat. no. MMD-030n; 1:100), followed by a horseradish peroxidase-conjugated polymer detection system for 30 min at room temperature. Immunoreactivity was visualized with 3,3′-diaminobenzidine and sections were counterstained with hematoxylin. Negative controls, in which the primary antibody was replaced by an isotype-matched immunoglobulin at the same concentration, and positive control sections were included in every run. Expression was quantified with the immunoreactive score (IRS) of Remmele and Stegner,³³ the product of the percentage of positive cells (0, none; 1, < 10%; 2, 10–50%; 3, 51–80%; 4, > 80%) and staining intensity (0, none; 1, weak; 2, moderate; 3, strong), giving a score of 0–12 classified as negative (0–1), weak (2–3), moderate (4–8), or strong (9–12). Scores were assigned in five fields at 400× per section by the blinded pathologist and averaged per animal. Detailed immunohistochemical procedures are provided in Supplementary Methods S4.

Statistical Analysis

Analyses were performed in IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA). Data are presented as mean ± standard deviation (SD) with 95% confidence intervals (CIs) for group means, or as counts and percentages. Normality was assessed with the Shapiro–Wilk test and homogeneity of variance with Levene’s test and the ratio of the largest to the smallest group variance. TNF-α, MDA, and epidermal thickness met the normality assumption and were compared by one-way analysis of variance (ANOVA) with Bonferroni-adjusted pairwise comparisons across all 21 contrasts; the letters in the tables and figures denote these comparisons. Because the group variances of MDA were unequal (variance ratio 16.9) and the negative-control variance of epidermal thickness was below the measurement resolution, these two endpoints were re-analyzed with Welch’s ANOVA and the Games–Howell post hoc test as a sensitivity analysis (Supplementary Table S2). Neutrophil grades were compared with the Kruskal–Wallis test followed by Dunn’s pairwise test with Bonferroni adjustment. IRS values were compared by one-way ANOVA with Tamhane’s T2 post hoc test, which does not assume equal variances, because group SDs differed more than fourfold; the Kruskal–Wallis test with Dunn’s test gave the same pattern (Supplementary Table S3). Categorical outcomes were analyzed with the Fisher–Freeman–Halton exact test for the overall association and Fisher’s exact test for each group versus the TPA control with Bonferroni adjustment for six comparisons; the chi-square test was not used because expected cell counts were below five. Effect sizes are reported as η² and ω² for ANOVA and ε² for the Kruskal–Wallis test, and mean differences from the TPA control are reported with Bonferroni-adjusted 95% CIs. The concentration–response relationship across the 2.5%, 5%, and 10% groups was tested with a linear contrast. Correlations used Pearson’s coefficient for pairs of continuous variables and Spearman’s coefficient where the ordinal neutrophil grade was involved. Post hoc achieved power for the observed effect sizes was computed from the non-central F distribution and is reported for completeness. All tests were two-sided with p < 0.05 considered significant (Supplementary Table S4).

Results

Standardization of the Extract

UPLC analysis of the two independently prepared extract solutions gave mean peak areas of 904,200 and 901,802, corresponding to α-mangostin concentrations of 41.09 and 40.98 µg/mL. Against the nominal total extract concentration of 234 µg/mL, these values correspond to 17.56% and 17.51% (w/w), a mean α-mangostin content of 17.54 ± 0.03% with a relative standard deviation of 0.19% (Table 1). Representative chromatograms are shown in Supplementary Figure S1. Minor peaks eluting before and after α-mangostin were not identified because authentic standards for the minor xanthones were unavailable; the extract is therefore standardized with respect to its principal xanthone only.

Table 1.

α-Mangostin Content of the Crude Garcinia Mangostana Pericarp Extract Determined by UPLC

Extract Solution	Total Extract (µg/mL)	Mean Peak Area	α-Mangostin (µg/mL)	α-Mangostin (% w/w)
I	234	904,200	41.09	17.56
II	234	901,802	40.98	17.51
Mean ± SD	234	903,001 ± 1,696	41.04 ± 0.08	17.54 ± 0.03

Notes: Two independently prepared solutions of the same extract batch were each injected in duplicate; the mean peak area of the duplicate injections is shown. Concentrations were read from the calibration line y = 21,800x + 8,438 (peak area against µg/mL) and expressed as a percentage of the nominal total extract concentration. The relative standard deviation of the two determinations was 0.19%.

Abbreviations: SD, standard deviation; UPLC, ultra-performance liquid chromatography.

Induction of Inflammation by TPA

All 35 animals completed the protocol. Three applications of TPA produced marked acute inflammation and oxidative stress (Table 2 and Figure 1). Compared with the negative control, cutaneous TNF-α rose 3.5-fold (77.03 ± 17.50 vs 22.03 ± 11.43 pg/mL; p < 0.001; Figure 1A) and MDA almost doubled (0.92 ± 0.10 vs 0.48 ± 0.13 nmol/mL; p < 0.001; Figure 1B). Epidermal thickness increased fivefold (300 ± 40 vs 60 ± 0 µm; p < 0.001; Figure 1C), and neutrophil infiltration rose from a mean rank of 3.00 to 30.00 (p < 0.001; Figure 1D). Epidermal hyperplasia and increased vascular permeability were present in all five TPA-treated animals and edema in four of five (Table 3).

Table 2.

Effects of the Topical Interventions on Continuous Inflammatory, Oxidative, and Histological Variables

Group	TNF-α (pg/mL)	MDA (nmol/mL)	Epidermal Thickness (µm)	Neutrophil Infiltration (Mean Rank)
Negative control	22.03 ± 11.43 (7.84–36.22)a	0.48 ± 0.13 (0.32–0.64)a	60 ± 0 (NE)a	3.00a
TPA control	77.03 ± 17.50 (55.30–98.76)b	0.92 ± 0.10 (0.80–1.04)b	300 ± 40 (250–350)b	30.00b
TPA + vehicle cream	68.72 ± 17.95 (46.43–91.01)bc	0.83 ± 0.10 (0.71–0.95)ab	340 ± 100 (216–464)b	24.40b
TPA + hydrocortisone 1%	28.01 ± 15.54 (8.71–47.31)a	0.59 ± 0.09 (0.48–0.70)ab	160 ± 100 (36–284)a	14.10ab
TPA + GMPE 2.5%	37.60 ± 17.95 (15.31–59.89)ac	0.88 ± 0.37 (0.42–1.34)b	60 ± 10 (48–72)a	16.00ab
TPA + GMPE 5%	32.89 ± 24.29 (2.73–63.05)ac	0.67 ± 0.10 (0.55–0.79)ab	90 ± 40 (40–140)a	24.40b
TPA + GMPE 10%	24.81 ± 11.85 (10.10–39.52)a	0.48 ± 0.10 (0.36–0.60)a	50 ± 20 (25–75)a	14.10ab
Test statistic	F(6,28) = 8.330	F(6,28) = 5.898	F(6,28) = 22.350	H(6) = 27.127
Overall p value	< 0.001	< 0.001	< 0.001	< 0.001
Effect size	η2 = 0.641; ω2 = 0.557	η2 = 0.562; ω2 = 0.461	η2 = 0.824; ω2 = 0.781	ε2 = 0.75

Notes: Values are mean ± SD (95% CI of the mean) for n = 5 animals per group, except neutrophil infiltration, an ordinal grade for which the mean rank is given. Epidermal thickness is reported to the nearest 10 µm, the resolution of the measurement; the SD of the negative control is below that resolution and is shown as 0, and its CI is not estimable (NE). Superscript letters denote the pairwise comparisons; within a column, groups that do not share a superscript letter differ significantly (Bonferroni-adjusted pairwise comparisons across all 21 contrasts, adjusted p < 0.05; Dunn’s test with Bonferroni adjustment for neutrophil infiltration). Group means with 95% CIs were computed with the t distribution (4 degrees of freedom).

Abbreviations: CI, confidence interval; GMPE, Garcinia mangostana pericarp extract; MDA, malondialdehyde; NE, not estimable; SD, standard deviation; TNF-α, tumor necrosis factor-alpha; TPA, 12-O-tetradecanoylphorbol-13-acetate.

Figure 1.

Effects of topical Garcinia mangostana pericarp extract (GMPE) on inflammatory, oxidative, and histological endpoints in TPA-induced skin inflammation in mice. (A) Cutaneous TNF-α concentration measured by ELISA. (B) Cutaneous malondialdehyde (MDA) concentration measured by the TBARS method. (C) Epidermal thickness. (D) Neutrophil infiltration expressed as the mean rank of the semi-quantitative grade. Bars are group means and error bars are one standard deviation (n = 5 per group); no error bar is shown in (D) because a standard deviation is not defined for mean ranks. The italic p value in each panel is that of the overall group comparison (one-way ANOVA for (A–C), Kruskal–Wallis test for (D). Letters above the bars denote the post hoc pairwise comparisons: bars that share a letter do not differ significantly (Bonferroni-adjusted pairwise comparisons across all 21 contrasts for (A–C), Dunn’s test with Bonferroni adjustment for D; adjusted p ≥ 0.05). The letter assignment is identical to that in Table 2.

Abbreviations: HC, hydrocortisone; NC, negative control; TPA, 12-O-tetradecanoylphorbol-13-acetate; VC, vehicle cream.

Table 3.

Categorical Histological Outcomes Across Treatment Groups

Group	Edema, n (%)	Adjusted p vs TPA	Epidermal Hyperplasia, n (%)	Adjusted p vs TPA	Increased Vascular Permeability, n (%)	Adjusted p vs TPA
Negative control	0 (0)	0.286	0 (0)	0.048	0 (0)	0.048
TPA control	4 (80)	Reference	5 (100)	Reference	5 (100)	Reference
TPA + vehicle cream	2 (40)	1.000	5 (100)	1.000	5 (100)	1.000
TPA + hydrocortisone 1%	4 (80)	1.000	3 (60)	1.000	5 (100)	1.000
TPA + GMPE 2.5%	4 (80)	1.000	0 (0)	0.048	5 (100)	1.000
TPA + GMPE 5%	1 (20)	1.000	2 (40)	1.000	2 (40)	1.000
TPA + GMPE 10%	3 (60)	1.000	0 (0)	0.048	1 (20)	0.286
Overall exact p value	0.058		< 0.001		< 0.001

Notes: Values are the number (percentage) of animals in which the feature was present (n = 5 per group). The overall association between treatment group and each outcome was tested with the Fisher–Freeman–Halton exact test; the Pearson chi-square test was not used because expected cell counts were below five. Each group was compared with the TPA control by Fisher’s exact test, and the p values were Bonferroni-adjusted for the six comparisons within each outcome; adjusted values of 1.000 are Bonferroni products truncated at unity.

Abbreviations: GMPE, Garcinia mangostana pericarp extract; TPA, 12-O-tetradecanoylphorbol-13-acetate.

Effects of GMPE on TNF-α and MDA

TNF-α differed significantly among groups (F(6,28) = 8.330, p < 0.001; η² = 0.641, ω² = 0.557; Table 2 and Figure 1A). Every GMPE concentration lowered TNF-α relative to the TPA control (Bonferroni-adjusted p = 0.023, 0.007, and 0.001 for 2.5%, 5%, and 10%), with mean differences of −39.43 (Bonferroni 95% CI −75.61 to −3.25), −44.14 (−80.32 to −7.96), and −52.22 (−88.40 to −16.04) pg/mL (Supplementary Table S2). GMPE 10% reduced TNF-α by 67.8% to 24.81 ± 11.85 pg/mL (95% CI 10.10–39.52), a value not significantly different from the negative control or from 1% hydrocortisone (28.01 ± 15.54 pg/mL; 95% CI 8.71–47.31). The vehicle cream had no detectable effect (68.72 ± 17.95 pg/mL; adjusted p = 1.000 vs TPA). The 2.5% and 5% creams, although lower than the TPA control, did not differ significantly from the vehicle group after adjustment. The TNF-α means fell monotonically with concentration (37.60, 32.89, and 24.81 pg/mL), but the linear trend across the three concentrations was not significant (t(28) = −1.18, p = 0.248).

MDA also differed among groups (F(6,28) = 5.898, p < 0.001; η² = 0.562, ω² = 0.461; Figure 1B). The response was graded: GMPE 2.5% did not differ from the TPA control (0.88 ± 0.37 nmol/mL; adjusted p = 1.000), GMPE 5% produced an intermediate value (0.67 ± 0.10 nmol/mL; adjusted p = 0.575), and GMPE 10% returned MDA to the negative-control level (0.48 ± 0.10 nmol/mL; 95% CI 0.36–0.60; mean difference from TPA −0.44, Bonferroni 95% CI −0.80 to −0.08; adjusted p = 0.007), a 47.8% reduction. Hydrocortisone gave 0.59 ± 0.09 nmol/mL (adjusted p = 0.099 vs TPA). The linear trend across the three concentrations was significant for MDA (t(28) = −3.72, p = 0.001). The sensitivity analysis with Welch’s ANOVA and Games–Howell comparisons confirmed the direction and significance of the principal contrasts (TPA vs negative control, TPA vs GMPE 10%); the individual comparisons whose significance differed between the two procedures are indicated in Supplementary Table S2.

Histopathological Findings

Epidermal thickness differed significantly among groups (F(6,28) = 22.350, p < 0.001; η² = 0.824, ω² = 0.781; Figure 1C). All three GMPE concentrations reduced epidermal thickness relative to the TPA control (adjusted p < 0.001 for each), to 60 ± 10, 90 ± 40, and 50 ± 20 µm for 2.5%, 5%, and 10%, values statistically indistinguishable from the negative control (60 ± 0 µm). Hydrocortisone also reduced thickness (160 ± 100 µm; adjusted p = 0.011 vs TPA) and did not differ from the GMPE groups or the negative control. The vehicle cream did not attenuate TPA-induced thickening (340 ± 100 µm). The ordering across GMPE concentrations was not monotonic (the 5% group was slightly thicker than the 2.5% group), and the linear trend was not significant (t(28) = −0.27, p = 0.788). The observed effect sizes and corresponding achieved power for the principal continuous endpoints are summarized in Supplementary Table S4.

Neutrophil infiltration differed among groups (Kruskal–Wallis H = 27.127, p < 0.001; ε² = 0.75; Figure 1D). The TPA control showed the highest infiltration (mean rank 30.00), followed by the vehicle and GMPE 5% groups (24.40 each), GMPE 2.5% (16.00), and hydrocortisone and GMPE 10% (14.10 each). After Bonferroni adjustment, Dunn’s test separated the negative control from the TPA, vehicle, and GMPE 5% groups; the reductions under hydrocortisone, GMPE 2.5%, and GMPE 10% relative to the TPA control were not significant with five animals per group (Table 2).

Representative H&E sections are shown in Figure 2. Negative-control skin (Figure 2A) had a thin, regular epidermis with an intact stratum corneum and no dermal infiltrate. After TPA (Figure 2B), the epidermis was markedly thickened with acanthosis and a dense, neutrophil-rich dermal infiltrate; the vehicle-treated section (Figure 2C) was indistinguishable from the TPA control. Hydrocortisone (Figure 2D) reduced thickening and infiltration without fully restoring normal architecture. In the GMPE groups the improvement was graded: at 2.5% (Figure 2E) the epidermis approached normal thickness but a residual infiltrate persisted, at 5% (Figure 2F) both features were further reduced, and at 10% (Figure 2G) the epidermis was thin and well organized with a sparse infiltrate, closely resembling the negative control.

Figure 2.

Representative hematoxylin and eosin-stained sections of dorsal skin at the end of treatment (original magnification 100×; 10× objective). (A) Negative control; (B) TPA control; (C) TPA + vehicle cream; (D) TPA + hydrocortisone 1%; (E) TPA + GMPE 2.5%; (F) TPA + GMPE 5%; (G) TPA + GMPE 10%. TPA induced marked epidermal thickening with acanthosis and a dense dermal inflammatory infiltrate, which persisted with the vehicle cream; hydrocortisone and GMPE reduced epidermal thickening and cellular infiltration, and the GMPE 10% section closely resembles the negative control. Quantitative measurements are given in Table 2.

Abbreviations: GMPE, Garcinia mangostana pericarp extract; TPA, 12-O-tetradecanoylphorbol-13-acetate.

Categorical Outcomes

The categorical findings followed the same direction but were not uniformly significant (Table 3). Epidermal hyperplasia and increased vascular permeability each showed a significant overall association with treatment group (Fisher–Freeman–Halton exact test, p < 0.001). Hyperplasia was present in all TPA-control animals and in none of the animals receiving GMPE 2.5% or 10% (adjusted p = 0.048 for each versus TPA), whereas the reduction with hydrocortisone (3/5) was not significant. Increased vascular permeability was present in every animal of the TPA, vehicle, hydrocortisone, and GMPE 2.5% groups, in 2/5 animals at GMPE 5%, and in 1/5 at GMPE 10%; no reduction reached significance after adjustment (adjusted p = 0.286 for GMPE 10%). Edema did not show a significant overall association with treatment (exact p = 0.058). Its distribution was not monotonic: edema was present in 4/5 TPA-control, 2/5 vehicle, 4/5 hydrocortisone, 4/5 GMPE 2.5%, 1/5 GMPE 5%, and 3/5 GMPE 10% animals. With a binary endpoint and five animals per group, the difference between the 5% and 10% groups corresponds to two animals and lies well within sampling variation; the vehicle group showed a similar unexplained departure. Edema is therefore not a positive finding of this study, and the apparent inferiority of the 10% cream to the 5% cream for this endpoint should not be interpreted as a true reversal of effect.

Immunohistochemical Expression of TNF-α and MDA

Immunohistochemistry corroborated the biochemical findings (Table 4; Figures 3 and 4). In negative-control skin, TNF-α (Figure 3A) and MDA (Figure 4A) immunoreactivity was weak and confined to scattered cells, whereas after TPA both markers stained intensely throughout the thickened epidermis and the papillary dermal infiltrate. IRS values differed significantly among groups for TNF-α (F(6,28) = 27.42, p < 0.001; η² = 0.855) and MDA (F(6,28) = 41.38, p < 0.001; η² = 0.899). TPA raised the TNF-α IRS from 1.60 ± 0.55 to 10.00 ± 1.87 (Tamhane’s T2, p = 0.006) and the MDA IRS from 1.40 ± 0.55 to 10.60 ± 1.95 (p = 0.005), and the vehicle cream did not alter either score (p = 1.000). GMPE 2.5% produced only a non-significant reduction (TNF-α IRS 6.20 ± 1.79, p = 0.210; MDA IRS 9.20 ± 1.64, p = 0.998). GMPE 5% significantly reduced both scores (TNF-α 4.60 ± 1.34, p = 0.022; MDA 4.80 ± 1.10, p = 0.020), to levels comparable with hydrocortisone (4.40 ± 1.67 and 3.40 ± 0.89). GMPE 10% produced the largest reduction (TNF-α 1.80 ± 0.45, p = 0.008; MDA 1.60 ± 0.55, p = 0.006), with scores that did not differ from the negative control (p = 1.000 for both) (Figures 3B and 4B). Because the IRS is an ordinal product of two ordinal scales, these data are presented as semi-quantitative support for the ELISA and TBARS results rather than as an independent quantitative outcome.

Table 4.

Immunohistochemical Expression of TNF-α and MDA in Dorsal Skin Expressed as the Immunoreactive Score

Group	TNF-α IRS, Mean ± SD (95% CI)	Median (Range)	Adjusted p vs TPA	MDA IRS, Mean ± SD (95% CI)	Median (Range)	Adjusted p vs TPA
Negative control	1.60 ± 0.55 (0.92–2.28)a	2 (1–2)	0.006	1.40 ± 0.55 (0.72–2.08)a	1 (1–2)	0.005
TPA control	10.00 ± 1.87 (7.68–12.32)b	9 (8–12)	Reference	10.60 ± 1.95 (8.18–13.02)b	12 (8–12)	Reference
TPA + vehicle cream	10.00 ± 1.87 (7.68–12.32)b	9 (8–12)	1.000	9.80 ± 2.05 (7.26–12.34)bc	9 (8–12)	1.000
TPA + hydrocortisone 1%	4.40 ± 1.67 (2.32–6.48)a	4 (2–6)	0.023	3.40 ± 0.89 (2.29–4.51)ad	4 (2–4)	0.008
TPA + GMPE 2.5%	6.20 ± 1.79 (3.98–8.42)ab	6 (4–9)	0.210	9.20 ± 1.64 (7.16–11.24)b	9 (8–12)	0.998
TPA + GMPE 5%	4.60 ± 1.34 (2.93–6.27)a	4 (3–6)	0.022	4.80 ± 1.10 (3.44–6.16)cd	4 (4–6)	0.020
TPA + GMPE 10%	1.80 ± 0.45 (1.24–2.36)a	2 (1–2)	0.008	1.60 ± 0.55 (0.92–2.28)a	2 (1–2)	0.006
Overall test	F(6,28) = 27.42; p < 0.001; η2 = 0.855			F(6,28) = 41.38; p < 0.001; η2 = 0.899

Notes: n = 5 animals per group. The IRS is the product of the percentage of positive cells (0–4) and staining intensity (0–3), range 0–12. Superscript letters denote the pairwise comparisons; adjusted p values are from Tamhane’s T2 post hoc test (all 21 pairwise comparisons) for each group versus the TPA control; within a column, groups that do not share a superscript letter differ significantly (Tamhane’s T2, adjusted p < 0.05). The Kruskal–Wallis test with Dunn’s test gave the same pattern (Supplementary Table S3).

Abbreviations: CI, confidence interval; GMPE, Garcinia mangostana pericarp extract; IRS, immunoreactive score; MDA, malondialdehyde; SD, standard deviation; TNF-α, tumor necrosis factor-alpha; TPA, 12-O-tetradecanoylphorbol-13-acetate.

Figure 3.

Immunohistochemical expression of TNF-α in mouse dorsal skin. (A) Representative sections from each group (original magnification 400×; scale bars 50 µm); positive immunoreactivity is shown by brown 3,3′-diaminobenzidine staining with hematoxylin counterstain. (B) TNF-α immunoreactive score (IRS; percentage of positive cells × staining intensity, range 0–12). Bars are group means and error bars one standard deviation (n = 5 per group). The italic p value is that of the overall one-way ANOVA. Letters above the bars denote the results of Tamhane’s T2 post hoc comparisons; bars that share a letter do not differ significantly (adjusted p ≥ 0.05). The letter assignment is identical to that in Table 4.

Abbreviations: GMPE, Garcinia mangostana pericarp extract; HC, hydrocortisone; NC, negative control; TNF-α, tumor necrosis factor-alpha; TPA, 12-O-tetradecanoylphorbol-13-acetate; VC, vehicle cream.

Figure 4.

Immunohistochemical expression of malondialdehyde (MDA) in mouse dorsal skin. (A) Representative sections from each group (original magnification 400×; scale bars 50 µm); positive immunoreactivity is shown by brown 3,3′-diaminobenzidine staining with hematoxylin counterstain. (B) MDA immunoreactive score (IRS; percentage of positive cells × staining intensity, range 0–12). Bars are group means and error bars one standard deviation (n = 5 per group). The italic p value is that of the overall one-way ANOVA. Letters above the bars denote the results of Tamhane’s T2 post hoc comparisons; bars that share a letter do not differ significantly (adjusted p ≥ 0.05). The letter assignment is identical to that in Table 4.

Abbreviations: GMPE, Garcinia mangostana pericarp extract; HC, hydrocortisone; NC, negative control; TPA, 12-O-tetradecanoylphorbol-13-acetate; VC, vehicle cream.

Correlations Among Inflammatory, Oxidative, and Histological Parameters

The endpoints were not independent of one another (Table 5). TNF-α correlated strongly with epidermal thickness (r = 0.775, p < 0.001) and moderately with MDA (r = 0.454, p = 0.006). Neutrophil infiltration correlated moderately with TNF-α (ρ = 0.496, p = 0.002) and with epidermal thickness (ρ = 0.429, p = 0.010), and most strongly with MDA (ρ = 0.598, p < 0.001). These cross-sectional associations, measured at a single time point across all seven groups, indicate that cytokine concentration, lipid peroxidation, and structural change varied together in this model; they do not establish causal direction.

Table 5.

Correlations Among Inflammatory, Oxidative, and Histological Parameters (n = 35)

Variable Pair	Coefficient	r or ρ	p Value
TNF-α vs epidermal thickness	Pearson r	0.775	< 0.001
TNF-α vs MDA	Pearson r	0.454	0.006
MDA vs epidermal thickness	Pearson r	0.429	0.010
TNF-α vs neutrophil infiltration	Spearman ρ	0.496	0.002
MDA vs neutrophil infiltration	Spearman ρ	0.598	< 0.001
Epidermal thickness vs neutrophil infiltration	Spearman ρ	0.429	0.010

Notes: Pearson’s coefficient was used for pairs of continuous variables and Spearman’s coefficient wherever the ordinal neutrophil grade was involved. Correlations were computed across all 35 animals at a single time point and do not establish causal direction.

Abbreviations: MDA, malondialdehyde; TNF-α, tumor necrosis factor-alpha.

Discussion

In this experimental animal study, topical application of a standardized G. mangostana pericarp extract attenuated TPA-induced cutaneous inflammation and oxidative stress in mice. GMPE reduced cutaneous TNF-α, MDA, epidermal thickening, and neutrophil infiltration, and the 10% cream restored TNF-α, MDA, and epidermal thickness to values that did not differ from healthy skin or from 1% hydrocortisone. The immunohistochemical findings paralleled the biochemical data. The concentration dependence was, however, not uniform: a significant linear trend was found for MDA, the TNF-α means fell monotonically without reaching significance for trend, and the histological endpoints did not order by concentration. The effect is therefore established for lipid peroxidation and supported, but not formally demonstrated, for the other endpoints.

The study builds directly on our earlier report, which three of the present authors co-authored, in which a topical pericarp extract cream reduced neutrophil infiltration and epidermal thickness in TPA-treated mouse skin.¹⁹ That study used histology alone, an extract that was not chemically characterized, and no active comparator, so it established the direction of the effect but not its magnitude or its biochemical correlates. The present study adds a chemically defined preparation standardized to 17.54% α-mangostin, two independent biochemical axes (a cytokine and a lipid peroxidation product) each confirmed by immunohistochemistry, a full cream formula, and a benchmark against a clinically used corticosteroid. The findings are consistent with other topical mangosteen preparations: pericarp extract reduced inflammatory-cell infiltration in a murine skin infection model,²⁰ pericarp extract-loaded liposomes and α-mangostin were effective against staphylococcal skin infection,²¹ and α-mangostin nanoparticles improved acne in a split-face human trial without irritation.²² To our knowledge, this is the first study to combine chemical standardization, biochemical and immunohistochemical measurement of TNF-α and MDA, and a corticosteroid comparator for a crude topical GMPE.

The mechanism by which GMPE produced these effects was not investigated in this study, and the interpretation that follows is a hypothesis drawn from the published pharmacology of mangosteen xanthones rather than a result of this work. α-Mangostin, the dominant xanthone in the extract, blocks NF-κB nuclear translocation and reduces COX-2, TNF-α, and interleukin-6 expression in activated macrophages,¹⁵ and systemic α-mangostin suppressed inflammation and tumor development in the DMBA/TPA model with down-regulation of PI3K/Akt/mTOR signaling, COX-2, and NF-κB.¹⁷ If the same actions occur in TPA-treated skin, they would account for the two-thirds reduction in cutaneous TNF-α under the 10% cream, and the parallel fall in MDA is compatible with the radical-scavenging and nuclear factor erythroid 2-related factor 2 (Nrf2)-activating activity reported for pericarp xanthones.^28,34,35 Minor constituents may also contribute: γ-mangostin suppresses interleukin-6/NF-κB signaling,²⁷ several pericarp xanthones modulate NF-κB/mitogen-activated protein kinase signaling and macrophage polarization,²⁶ and α-mangostin inhibits the keratinocyte channel TRPV3 while reducing cytokine release.³⁶ None of these pathways was measured here, and direct measurement of NF-κB p65, COX-2, and Nrf2 targets in treated skin is required before any of them can be attributed to GMPE.

The strong correlation between TNF-α and epidermal thickness and the moderate correlations linking MDA to neutrophil infiltration are consistent with the interdependence of cytokine signaling, oxidative damage, and structural remodeling described in TPA-treated skin,^3,8 but they are cross-sectional associations across treatment groups and cannot indicate which change drives which. The non-monotonic behavior of some endpoints deserves comment. For edema, the difference between the 5% and 10% groups amounts to two animals in a binary endpoint, the vehicle group showed a similar departure, and the overall association was not significant, so this pattern is most plausibly sampling variation. In addition, TPA-induced edema, which is largely TNF-α dependent, peaks within hours of application and subsides thereafter,¹¹ whereas our assessment was made about 96 h after the first TPA dose, when the endpoint is insensitive; the persistent endpoints at this time point, epidermal hyperplasia and cellular infiltration, were the ones that responded consistently to treatment. For epidermal thickness and neutrophil grade, all GMPE concentrations produced near-maximal reductions, so a concentration–response relationship could not be resolved within the range tested.

Topical delivery of mangosteen xanthones has both advantages and uncertainties. α-Mangostin has very low oral bioavailability and extensive first-pass metabolism,²⁹ so topical application places the compound where it is needed at concentrations that systemic dosing cannot achieve. Its lipophilicity favors partitioning into the stratum corneum, but the same property limits diffusion through the hydrated viable epidermis, and nanoparticle and gel carriers have been developed to improve cutaneous deposition of α-mangostin.^22,23 We used a simple oil-in-water vanishing cream and did not measure skin penetration, tissue concentration, or systemic exposure, so the amount of xanthone that reached the dermis is unknown. A second uncertainty is intrinsic to crude extracts: xanthone content varies with cultivar, ripeness, and extraction method,^12,13 the extract used here was standardized for α-mangostin only, the minor constituents were not identified, and a single batch was used, so batch-to-batch consistency remains to be demonstrated. These points do not undermine the observed effects but define what must be controlled before the preparation could be developed further.

The 10% cream did not differ significantly from 1% hydrocortisone on any continuous endpoint. This is an absence of a detected difference, not a demonstration of equivalence, since no equivalence margin was specified and the design cannot reliably detect a moderate difference between two active treatments. Nevertheless, a botanical preparation with short-term anti-inflammatory activity of the same order as a low-potency corticosteroid is of translational interest, because long-term corticosteroid use is limited by atrophy and tachyphylaxis and because mangosteen pericarp is an inexpensive and abundant agricultural by-product in Southeast Asia.^14,25 Any dermatological application would, however, require formal irritation and sensitization testing, chronic-use data, and evaluation in human skin.

Limitations

Several limitations constrain the interpretation. First, the model was an acute, three-day TPA challenge. It reproduces the inflammatory and oxidative features of tumor promotion, but it does not measure tumor formation, and the results say nothing about chemoprevention, chronic or relapsing inflammation, efficacy on repeated exposure, or cumulative safety. Second, no signaling molecule was measured: NF-κB, COX-2, PI3K/Akt/mTOR, and Nrf2 were not assayed at the protein, transcript, or activity level, and the mechanism discussed above remains an inference from the literature. Third, the sample size was small (five animals per group) and was set by the Federer formula rather than by an a priori power analysis; the design is powered for large effects only, the observed effect sizes for the primary endpoints were well above that threshold (f = 1.13–2.16, achieved power ≥ 0.998), but moderate effects and the categorical endpoints, including edema, could not be adjudicated. Fourth, only male mice were used, so the findings cannot be generalized to females. Fifth, no group received GMPE on uninflamed skin and no formal irritation or toxicity assessment was made; the normal epidermal thickness under the 10% cream argues against overt irritancy but does not establish tolerability. Sixth, skin penetration, tissue concentration, and systemic exposure to xanthones were not measured. Seventh, all endpoints were assessed at a single time point by a single blinded observer, the histological and immunohistochemical scores are semi-quantitative, and the IRS is ordinal, so those panels are supportive rather than independently quantitative. Eighth, the biochemical values were expressed per mL of homogenate rather than per mg of protein. Ninth, the analytical characterization has limits: α-mangostin identity rested on retention-time matching with a single standard without peak-purity assessment, precision was estimated from two determinations, accuracy was not assessed by spike recovery, the minor xanthones were not identified, and α-mangostin was measured in the extract but not in the finished creams. Finally, the extract and creams were single batches, so variability between batches of a crude preparation was not evaluated. Future work should measure NF-κB p65, COX-2, and Nrf2-dependent targets in treated skin, extend the model to two-stage DMBA/TPA carcinogenesis with tumor endpoints, quantify cutaneous xanthone delivery, include sex-balanced cohorts sized on the effect sizes reported here, and add dermal irritation and sensitization testing.

Conclusion

In this experimental animal study, topical application of a G. mangostana pericarp extract standardized to 17.54% α-mangostin attenuated TPA-induced cutaneous inflammation and oxidative stress in mice. GMPE reduced cutaneous TNF-α, MDA, epidermal thickening, and neutrophil infiltration, with a concentration-dependent effect demonstrated for lipid peroxidation, and the 10% cream performed comparably to 1% hydrocortisone on every continuous endpoint. Because the study used an acute inflammation model and did not measure signaling proteins or tumor formation, it does not establish that GMPE modulates tumor-promotion pathways. The findings identify topical GMPE as a promising preclinical candidate whose anti-inflammatory and antioxidant activity warrants direct mechanistic confirmation and evaluation in long-term carcinogenesis models before any claim of chemopreventive activity or clinical application can be made.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Abbreviations

ANOVA, analysis of variance; AP-1, activator protein-1; ARRIVE, Animal Research: Reporting of In Vivo Experiments; AVMA, American Veterinary Medical Association; CI, confidence interval; COX-2, cyclooxygenase-2; DAB, 3,3′-diaminobenzidine; DMBA, 7,12-dimethylbenz[a]anthracene; ELISA, enzyme-linked immunosorbent assay; GMPE, Garcinia mangostana pericarp extract; H&E, hematoxylin and eosin; IRS, immunoreactive score; LOD, limit of detection; LOQ, limit of quantification; MDA, malondialdehyde; mTOR, mammalian target of rapamycin; NF-κB, nuclear factor-kappa B; Nrf2, nuclear factor erythroid 2-related factor 2; PI3K, phosphoinositide 3-kinase; PKC, protein kinase C; ROS, reactive oxygen species; SD, standard deviation; TBARS, thiobarbituric acid reactive substances; TNF-α, tumor necrosis factor-alpha; TPA, 12-O-tetradecanoylphorbol-13-acetate; UPLC, ultra-performance liquid chromatography.

Data Sharing Statement

The per-animal datasets generated and analyzed during this study, together with the chromatographic raw files, are available from the corresponding author on reasonable request. The data are not publicly available because they form part of an ongoing research project.

Ethics Approval

The animal study protocol was approved by the Ethics Review Committee of the Faculty of Medicine, Hasanuddin University, Makassar, Indonesia (approval no. 289/UN4.6.4.5.3.1/PP66/2025; approved on 3 April 2025), which is the institutional committee responsible for the ethical review of research involving laboratory animals at the faculty. All procedures were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the ARRIVE 2.0 guidelines.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

Three of the authors (Khairuddin Djawad, Siswanto Wahab, and Upik Anderiani Miskad) co-authored the earlier study of this extract by Mokoagow et al (2020),¹⁹ which is cited and compared with the present work in the Discussion; no animals, tissues, or data are shared between the two studies. The authors report no other conflicts of interest in this work.