Identification of Crocin I as a positive modulator of GSTP1 with potential relevance to prostate cancer

Abstract

Glutathione S-transferase Pi 1 (GSTP1) is frequently silenced in prostate adenocarcinoma (PRAD), yet whether its residual catalytic activity can be chemically enhanced is largely unexplored. Here, we identify Crocin I as a positive modulator of GSTP1 through structure-based screening, biophysical characterisation, and biochemical validation. Bio-layer interferometry (BLI) supported a concentration-dependent association between Crocin I and GSTP1, and enzymatic assays demonstrated enhanced catalytic activity. Michaelis–Menten kinetics showed that Crocin I increased V_max without markedly altering the apparent K_m, consistent with apparent non-competitive activation of 1-chloro-2,4-dinitrobenzene conjugation under the tested conditions. Differential scanning fluorimetry showed that Crocin I altered GSTP1 thermal behaviour, with a comparable shift in the presence of the H-site inhibitor NBDHEX. Exploratory systems-level analyses further connected Crocin I–GSTP1 modulation with redox- and PRAD-associated networks. Together, these findings establish Crocin I as a previously unrecognised positive modulator of GSTP1 and provide a biochemical foundation for GSTP1-directed modulation in PRAD.

Introduction

Glutathione S-transferases (GSTs) constitute a superfamily of enzymes widely distributed in eukaryotes and prokaryotes, serving as crucial components of phase II metabolic detoxification and playing an indispensable role in maintaining cellular homeostasis¹. GSTs catalyse the conjugation of the sulfhydryl group of endogenous glutathione (GSH) with a broad spectrum of xenobiotic and endogenous toxic compounds, significantly enhancing the water solubility of these substrates and facilitating their cellular efflux, thereby effectively neutralising their genotoxic and cytotoxic effects². Based on subcellular localisation, structure, substrate specificity, and immunoreactivity, mammalian GSTs are primarily classified into three major categories: cytosolic, mitochondrial, and microsomal, with cytosolic GSTs further subdivided into α, μ, π, θ, and other isoforms^3,4.

Among the various GST isoforms, glutathione S-transferase Pi 1 (GSTP1) has attracted considerable attention due to its unique expression pattern and function. GSTP1 is notably highly expressed in the proximal digestive tract, cornea, and subcutaneous skin tissues, positioning it as a key molecular component of the first line of defence of the body against environmental toxins and carcinogens. Structurally, the GSTP1 protein exhibits the characteristic GSTs family fold: an N-terminal domain (G-site) responsible for GSH binding and a C-terminal domain (H-site) primarily involved in binding hydrophobic substrates, together forming a versatile electrophilic substrate-binding pocket⁵. This structural feature endows GSTP1 with a broad substrate spectrum, enabling it to efficiently eliminate various genotoxic agents, including polycyclic aromatic hydrocarbons and aflatoxins^6,7. Consequently, the expression level and activity of GSTP1 directly determine the cell defensive capacity against chemical stress and oxidative damage, establishing it as a critical intracellular barrier for protecting genomic integrity and preventing malignant transformation.

Interestingly, the role of GSTP1 is highly context-dependent in oncology. While it is overexpressed in many malignancies contributing to chemoresistance, its expression is frequently downregulated or epigenetically silenced during the initiation and progression of prostate adenocarcinoma (PRAD)^8,9. This loss of GSTP1 function deprives prostate epithelial cells of their primary antioxidant defence, leading to the accumulation of DNA damage and accelerating malignant transformation^10–12. Because GSTP1 serves as a critical node in maintaining cellular homeostasis, its dysregulation is deeply intertwined with PRAD progression. Historically, pharmacological research has predominantly focused on designing GSTP1 inhibitors to overcome chemoresistance in tumours where the enzyme is overexpressed. Conversely, the development of positive modulators—molecules capable of enhancing the residual GSTP1 activity in compromised tissues like the prostate—represents a significant, yet largely unexplored. Identifying small molecules that can positively modulate GSTP1 would not only provide essential chemical probes to investigate these survival networks but also offer a novel conceptual framework for chemopreventive or targeted interventions in PRAD^13–17.

Crocin is a pharmacologically active water-soluble carotenoids derived from Crocus sativus L. and Gardenia jasminoides Ellis., possessing a well-documented safety profile¹⁸. Preclinical studies have highlighted its pleiotropic cytoprotective and chemopreventive effects, including the modulation of oxidative stress and apoptosis in various models, including prostate cancer (PCa)^19,20. While recent reviews postulate that these protective effects may stem from the systemic modulation of phase II detoxifying enzymes, the precise direct molecular targets remain elusive^21–23. Whether Crocin I could directly interact with and modulate GSTP1 has not been investigated.

To address this gap, the present study integrates computational modelling with in vitro biophysical and biochemical approaches to identify and characterise modulators of GSTP1 from a natural product library. This work establishes an experimental framework for investigating GSTP1 modulation and provides a candidate chemical probe for further evaluation in PCa research.

Methods

Protein structure modelling

The target protein sequence was retrieved from the UniProt database (https://www.uniprot.org/) under accession number UniProt-P09211. Reference three-dimensional structures derived from X-ray diffraction or electron microscopy were obtained from the Protein Data Bank (https://www.rcsb.org/). To construct the native protein structure, the Protein Preparation Wizard module (Schrödinger, Inc.) was employed for structural optimisation. Firstly, hydrogen atoms were added to the protein structure according to the ionisation states of amino acid residues at pH 7.0. Subsequently, the structure was refined by optimising hydrogen bonding networks, eliminating atomic clashes, and assigning formal charges to heterogroups. Finally, extraneous water molecules were removed, and the resulting structure was energy-minimised using the OPLS4 force field to converge to the lowest energy conformation. The modelled and optimised protein structures were subjected to Ramachandran plot analysis for validation. Structures meeting quality criteria were subsequently used for binding site prediction.

Analysis of potential binding sites on protein surface

Binding site analysis was performed on the optimised GSTP1 structure using SiteMap (Schrödinger, Inc.). Five putative binding pockets were identified, with particular attention to groove regions capable of accommodating hydrogen bond donors and acceptors. Pocket topologies were validated against published crystal structures of GSTP1 in ligand-bound conformations (PDB: 3GUS, 5J41, 5X79). Grid maps for subsequent virtual screening of the natural product library were generated using the Glide module, with grid centres placed at each of the predicted binding regions.

Simulated docking

The natural product library (Selleck L1400) was prepared using the LigPrep module (Schrödinger, Inc.) to generate spatial and chiral derivatives. It is important to note that this process generates all theoretical stereoisomers, including non-natural configurations such as Entry 3 (Table 1). In contrast, Entry 8 corresponds to Crocin I, the dominant naturally occurring isomer. During this process, each compound was expanded into multiple conformations based on its chiral, ionisation, and tautomeric properties. Conformations were analysed across a pH range of 5.0–9.0 to simulate the compound’s potential states in different physiological environments, with up to 32 conformers generated per compound. A total of 47 624 derived compound structures were then docked into the binding site of GSTP1. The virtual screening was performed in three sequential steps: high-throughput virtual screening (HTVS) docking, standard precision (SP) docking, and extra precision (XP) docking.

Table 1.

Calculated binding free energies (ΔG_bind) of candidate compounds with GSTP1 derived from MM–GBSA analysis.

Ranking	Compound name	Docking free energy (kCal/mol)	Total bond energy (kCal/mol)	Electrostatic bond energy (kCal/mol)	Covalent bond energy (kCal/mol)	Hydrogen bond energy (kCal/mol)	Hydrophobic bond energy (kCal/mol)
1	Mogroside V	−16.3683	−86.7126	−69.4473	7.7578	−6.8755	−31.8171
2	Rebaudioside M	−16.3051	−42.8277	−55.1699	−4.8850	−7.6627	−12.4393
3	Crocin	−16.2239	−91.9572	−74.8123	8.3788	−5.0143	−26.8818
4	Macranthoidin A	−15.9692	−75.2470	−80.4467	11.4346	−6.9024	−19.8148
5	Timosaponin BII	−15.8823	−81.7982	−66.1901	11.1117	−5.3698	−27.4124
6	Hosenkoside K	−15.7566	−76.3089	−91.7461	10.0178	−5.2723	−23.3717
7	Rebaudioside D	−15.4986	−75.2599	−89.9572	8.8213	−6.8369	−18.6260
8	Crocin I	−15.2275	102.8660	−93.3734	4.1380	−7.5496	−24.5322
9	Methylprotodioscin	−14.9603	−79.5660	−120.0205	25.1539	−7.0028	−20.3255
10	Platycodin D	−14.7888	−50.5695	−50.9634	9.4363	−5.8832	−13.7156
11	9″-Methyl salvianolate B	−14.1817	−85.1315	−20.2834	12.3245	−4.8845	−28.8715
12	Crocin II	−14.1229	−78.7014	−72.0756	14.5029	−6.8435	−29.4968
13	Poliumoside	−13.8675	−75.2148	−60.8590	10.3626	−5.1189	−27.2027
14	Rebaudioside C	−13.6318	−75.4530	−58.5923	2.6878	−6.0005	−21.8758
15	Angoroside C	−13.5704	−77.1847	−60.6869	17.3597	−5.2214	−24.3321
16	Polyphyllin VII	−13.5604	−84.9222	−80.3975	10.8361	−7.7151	−17.9936
17	Forsythoside A	−13.5131	−68.2582	−54.3370	10.9184	−4.7188	−27.7117
18	Flavin adenine Dinucleotide	−13.4895	−65.6549	41.1200	10.0013	−5.4930	−14.2020
19	Complanatoside A	−13.4836	−60.3434	−56.5047	9.1193	−5.8301	−16.7773
20	Ligustroflavone	−13.4629	−75.9685	−79.8098	7.9187	−5.8347	−14.1424
21	Buddlejasaponin IVb	−13.4576	−89.8430	−74.0366	6.6783	−5.5782	−27.7625
22	Neohesperidin Dihydrochalcone	−13.3060	−70.7959	−115.8344	13.0155	−6.8937	−16.0812
23	Salvianolic acid B	−13.2978	−68.5221	17.7779	13.0174	−6.1492	−26.0266
24	Epmedin B	−13.1176	−82.6315	−66.2770	19.7369	−6.0088	−25.8659
25	Escin	−13.0857	−47.0128	−29.1472	15.7884	−5.8361	−18.6835
26	Positive control	−8.4503	−60.2485	25.1506	3.3563	−6.5812	−14.1596

HTVS screening operating parameters

During the molecular docking screening phase, we employed the standard Glide protocol for HTVS. Ligand conformational sampling was performed using flexible docking, with nitrogen inversion and ring conformation sampling enabled to comprehensively explore potential binding poses. To balance computational efficiency and conformational rationality, intramolecular hydrogen bond rewards and enhanced planarity of conjugated π groups were disabled during the calculation. Following docking, post-docking minimisation was conducted to optimise intermolecular interactions, and predefined spatial constraints were applied to ensure proper ligand positioning within the active site. To balance screening breadth and computational resources, a maximum of one representative pose per compound was generated for deduplication analysis, with the top 30% of scored compounds retained for subsequent screening stages.

SP screening operating parameters

For the SP screening phase, the molecular docking parameters were configured as follows: ligand conformations were sampled via flexible docking, with nitrogen inversion and ring conformation sampling enabled, the latter utilising an energy window of 2.5 kcal/mol. To enhance computational efficiency and conformational rationality, rewards for intramolecular hydrogen bonds and enhanced planarity of conjugated π groups were disabled. Post-docking minimisation was performed, and selected spatial constraints were applied to ensure ligand placement within the active site. Each compound generated one representative pose for deduplication, with the top 30% of ranked compounds retained for further screening.

XP screening operating parameters

In the XP screening phase, docking parameters were optimised based on the SP settings: each compound generated up to five binding poses to improve conformational coverage, while the retention rate was adjusted to the top 5% of scored compounds to enhance the precision and reliability of the screening results. The energy window for ring conformation sampling remained 2.5 kcal/mol.

MM/GBSA operation parameters

The molecular mechanics/generalised Born surface area (MM/GBSA) method was employed for binding free energy calculations, with the following parameters: the OPLS4 force field and the VSGB solvation model were used, the latter incorporating a variable-dielectric generalised Born model with residue-specific effects. Given that the ligands are administered in neutral salt form, partial charge corrections for ligands were not applied. As the target protein is non-membrane, the implicit membrane model was disabled. For protein flexibility, default settings were applied, and all residue atoms were minimised to achieve a stable conformational state.

Molecular dynamics simulation

Molecular dynamics (MD) simulations were conducted in an aqueous system designed to mimic physiological conditions. The system construction involved the following steps: the SPC water model was used within an orthorhombic box, with solute reorientation performed to minimise box volume. To simulate the in vivo buffer environment, sodium or chloride ions were added to neutralise the system charge, and 0.15 M NaCl was incorporated to replicate physiological saline concentration in body fluids.

Expression and purification of recombinant human GSTP1

The plasmid pET-28a(+) (GenScript, Nanjing, China) harbouring the human GSTP1 gene was transformed into Escherichia coli BL21 (DE3) (Bio-rabit, Shanghai, China) competent cells. Following overnight incubation on LB agar plates supplemented with kanamycin (50 µg/mL) at 37 °C, a single colony was selected and inoculated directly into liquid Luria-Bertani (LB) medium containing the same antibiotic. The bacterial culture was incubated at 37 °C with shaking until the optical density at 600 nm (OD₆₀₀) reached 0.6–0.8. Protein expression was induced by isopropyl β-D-1-thiogalactopyranoside (0.4 mM, final concentration), followed by incubation at 16 °C for 16 h with shaking. The cells were harvested by centrifugation and resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM PMSF, pH 7.4).

Cell disruption was performed by ultrasonication on ice at 60% amplitude for 30 min (with cycles of 5 s on and 5 s off). The lysate was centrifuged at 14 000 × g for 1 h at 4 °C to remove insoluble debris. The supernatant containing soluble GSTP1 was loaded onto a Ni-NTA affinity chromatography column. After washing with lysis buffer supplemented with 20 mM imidazole to remove non-specifically bound proteins, the target protein was eluted using elution buffer containing 200 mM imidazole. The eluted protein was dialysed overnight at 4 °C against a storage buffer (20 mM HEPES, 150 mM NaCl, 5% glycerol, pH 7.4) to remove imidazole and salts. Following dialysis, the recombinant GSTP1 protein was subjected to concentration measurement using a BCA protein assay kit (Vazyme, Nanjing, China) and purity analysis by SDS-PAGE. After confirmation, the protein was aliquoted and stored at −80 °C for subsequent use.

GSTP1 enzymatic activity assay

The 1-chloro-2,4-dinitrobenzene (CDNB; Aladdin, Shanghai, China) method was used for enzyme activity characterisation. The enzymatic activity of GSTP1 in the presence of Crocin was evaluated using the ultraviolet spectroscopy method with CDNB as the substrate. The assay was performed in a 100 μL reaction system containing (20 mM HEPES, 150 mM NaCl, 5% glycerol, pH 7.4) consistent with the storage buffer of the recombinant GSTP1 protein, 1 mM reduced GSH, and 1 mM CDNB. Recombinant human GSTP1 was pre-incubated with varying concentrations of Crocin (20, 40, 80 μM) (MUST, Chengdu, China) or a known GSTP1 inhibitor (4 μM) (NBDHEX, MedChemExpress, New Jersey, USA) as a negative control for 10 min at 37 °C. The absorbance at 340 nm was immediately measured using a microplate reader (SpectraMax iD3, Molecular Devices, USA) to determine the basal enzymatic activity under each condition. The specific activity was calculated using the extinction coefficient of 9.6 mM⁻¹ cm⁻¹ for the conjugated product and expressed as μM of product formed per minute per mg of protein. All assays were performed in at least three independent replicates.

Michaelis–Menten kinetics

Steady-state kinetic parameters were determined by varying the CDNB concentration across an eight-point gradient (0.025, 0.05, 0.1, 0.2, 0.4, 0.8, 1.2, and 1.6 mM) at a fixed GSH concentration of 2 mM. Reactions (200 μL) contained GSTP1 (80 nM) in potassium phosphate buffer (pH 6.5) and were performed at room temperature in the presence or absence of Crocin I (100 μM). Initial velocities were determined from the linear phase of each progress curve. Michaelis–Menten parameters (V_max and K_m) were estimated by nonlinear regression. The activation mode was evaluated by comparing independent fits with a shared-K_m global fit; Lineweaver–Burk transformation was used for graphical confirmation. All experiments were performed in three independent replicates, each in triplicate.

DSF assays

The binding interaction between Crocin I and GSTP1 was investigated by differential scanning fluorimetry (DSF). GSTP1 (final concentration 3.2 mg/mL) was mixed with 10× SYPRO Orange dye (Invitrogen, Thermo Fisher Scientific, USA) and Crocin I at a final concentration of 40 μM in HEPES buffer. For co-incubation experiments, NBDHEX was added at a final concentration of 20 μM. The apo protein (buffer only) served as the reference control. Reaction mixtures (60 μL) were loaded into a 96-well PCR plate, sealed optically, and centrifuged briefly. Thermal denaturation was performed on a Bio-Rad CFX real-time PCR system with a temperature gradient from 25 to 95 °C at a ramp rate of 0.2 °C/min. Fluorescence intensity (excitation/emission: 490/530 nm) was recorded continuously. The melting temperature (T_m) was determined from the minimum of the first derivative of the fluorescence curve (−d(RFU)/dT). A shift in T_m (ΔT_m) between Crocin I-treated and apo samples was interpreted as evidence of ligand-induced conformational perturbation. All experiments were performed in three independent replicates.

Bio-layer interferometry

Binding between Crocin I and GSTP1 was assessed by bio-layer interferometry (BLI) experiment on an Octet R4 system (Sartorius, Germany) using streptavidin (SA) biosensors (Sartorius, Germany). Recombinant GSTP1 was biotinylated with biotinylation reagent (1828 M, Genemore, Jiangsu, China) and immobilised onto SA sensors in HEPES buffer to a loading signal exceeding 4 nm. Crocin I was serially diluted in a two-fold series (15.6–500 μM). The association phase was monitored for 90 s, followed by a dissociation phase of 120 s. Data were reference-subtracted and analysed using the Octet data analysis software (Sartorius) to determine dissociation constant (KD).

Network pharmacology analysis obtaining targets associated with Crocin I and PCa

Instead of relying solely on computational prediction, we compiled a robust list of Crocin I-binding targets through extensive literature mining. The PubMed and Google Scholar databases were searched using keywords such as Crocin I targets, Crocin I mechanism, and Crocin I binding proteins. Only targets with experimental evidence of direct interaction or significant modulation by Crocin I were included in the final dataset (n = 15). Simultaneously, disease-associated targets for PCa were retrieved from the GeneCards database (https://www.genecards.org/)²⁴. To ensure data quality and relevance, a relevance score threshold (Score > 6.5) was applied to filter out low-confidence genes.

Intersection and network construction

The identified Crocin I-related targets were mapped against the PCa disease targets using Venny 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/index.html)²⁵ to identify overlapping genes, which represent potential therapeutic targets. To elucidate the interactions among these common targets, a protein–protein interaction (PPI) network was constructed using the STRING database (https://string-db.org/) and visualised via Cytoscape software.

Functional enrichment analysis

To explore the biological functions and signalling pathways associated with the intersection targets, Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using the Metascape platform (http://metascape.org/)²⁶. Terms with a p values <0.01 were considered statistically significant.

Bioinformatic analysis

Pan-cancer analysis of GSTP1 gene expression was performed using data from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/) and the Genotype-Tissue Expression (GTEx) (https://www.gtexportal.org/home/) databases. The mRNA and protein expression levels of GSTP1 in PRAD were analysed using the Human Protein Atlas (HPA) database (https://www.proteinatlas.org/)²⁷. Via the Xiantao Academic Website, the mRNA expression levels of GSTP1 in PRAD and diagnostic receiver operating characteristic (ROC) analysis were conducted by analysing data from TCGA-PRAD.

Data analysis

Statistical analyses were performed using GraphPad Prism (Version 9.5.1, GraphPad Software, LLC, San Diego, CA, USA) and Microsoft Excel (Professional 2021, Microsoft Corporation, Redmond, WA, USA). Graphical representations were generated using GraphPad Prism and Matplotlib. The outcomes of all statistical tests including sample sizes and p values are provided in the corresponding figure legends. Results were presented as means ± SD. The significance of p values is denoted as follows: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Results

Structural preparation and identification of druggable pockets

To identify novel ligands targeting GSTP1, we first established a high-quality structural model for virtual screening. The GSTP1 homodimer structure was prepared and energy-minimised to resolve steric clashes and optimise hydrogen bonding networks, resulting in a conformation that closely approximates the physiological state (Figure 1(A,B)). The structural overview of the monomer and the assembled homodimer is presented in Figure 1(C–E). Based on this optimised structure, surface topology analysis identified five potential druggable pockets characterised by distinct hydrophobic grooves and accessible hydrogen bond donor/acceptor sites (Figure 2(A–F)). These regions provided the structural basis for the subsequent docking campaign.

Figure 1.

Structural preparation and validation of the GSTP1 homodimer system. Ramachandran plots evaluating the stereochemical quality of the GSTP1 structure before energy minimisation (A) and after energy minimization (B). Structural overview of the GSTP1 complex: surface representation of the monomer (C); the complementary monomer within the homodimer interface (D); and the assembled GSTP1 homodimer (E).

Figure 2.

Identification and topological characterisation of potential druggable pockets on GSTP1. Global view of predicted binding pockets on the GSTP1 dimer surface (A). Detailed surface topology and spatial location of the five identified binding sites, namely site #1 (B), site #2 (C), site #3 (D), site #4 (E), and site #5 (F), highlighting the diversity of potential ligand-interaction regions.

Virtual screening and MD simulation

Virtual screening of 2540 natural products identified 26 candidates with favourable docking scores (Table 1). Among these, Mogroside V and Crocin (theoretical isomer) yielded slightly more favourable docking free energies (−16.37 and −16.22 kcal/mol, respectively) than Crocin I (−15.23 kcal/mol). Docking scores, however, primarily reflect the enthalpy of the static protein–ligand complex and may not fully capture binding stability. We therefore additionally evaluated total bond energy as a supplementary energetic metric, which incorporates electrostatic, hydrogen bonding, and hydrophobic contributions into a more comprehensive energetic metric. By this measure, Crocin I ranked highest (−102.87 kcal/mol), surpassing its structural analogue Crocin (theoretical isomer; −91.96 kcal/mol) and Crocin II (−78.70 kcal/mol). Energy decomposition indicated that the favourable total bond energy of Crocin I was driven predominantly by electrostatic interactions (−93.37 kcal/mol) and hydrogen bonding (−7.55 kcal/mol). On the basis of these computational results, Crocin I was selected for further experimental characterisation. Comparison of the predicted binding modes revealed distinct interaction profiles among the three Crocin isomers. In the docking pose of Crocin (theoretical isomer), the interaction landscape was dominated by negative electrostatic patches supplemented by polar contacts (Figure 3(C)). By contrast, Crocin I engaged a denser network of polar interactions involving residues Arg14, Ser66, and Asn67 within the active site, consistent with its higher computed hydrogen bond energy (Figure 3(G)). Crocin II displayed a binding mode dominated by hydrophobic contacts (Figure 3(K)), correlating with its less favourable total bond energy. These docking-level observations motivated further investigation of Crocin I. To assess whether the predicted binding modes persist under dynamic conditions, we performed MD simulations on the Apo, Crocin (theoretical isomer), Crocin I, and Crocin II systems. Backbone root-mean-square deviation (RMSD) values converged and remained stable over the simulation trajectory for all four systems (Figure 4(A–D)), indicating that ligand binding did not induce global structural perturbations. Root-mean-square fluctuation (RMSF) analysis indicated that residue flexibility in the ligand-bound states was comparable to that of the Apo state, with local reductions in flexibility at the predicted interaction sites (Figure 4(E–H)). Secondary structure conservation and protein–ligand contact persistence (hydrogen bond and hydrophobic interaction fractions) remained stable for all three Crocin isomers throughout the simulations. Taken together, the computational screening and MD results identified Crocin I as a candidate for experimental validation, while recognising that computational predictions require direct biophysical verification.

Figure 3.

Molecular interaction profiles and binding modes of Crocin stereoisomers with GSTP1. Chemical structures of Crocin (theoretical isomer) (A), Crocin I (natural isomer) (E), and Crocin II (I). 2D interaction diagrams highlighting key contacts in the binding pocket for Crocin (F), Crocin I (J), and Crocin II (L). Detailed non-covalent interaction networks for Crocin (C), Crocin I (G), and Crocin II (K). Electrostatic potential surface representations of the ligand–protein complex of GSTP1 with Crocin (D), Crocin I (H), and Crocin II (L). Surface colouring: red (negative potential), blue (positive potential), white (neutral/hydrophobic). Green mesh represents the ligand surface. Note: Purple: charged (positive); green: hydrophobic; red: charged (negative); nattier blue: polar; white: water. It is possible to have interactions with >100% as some residues may have multiple interactions of a single type with the same ligand atom.

Figure 4.

Comparative MD analysis of structural stability and flexibility across GSTP1 systems. Backbone RMSD trajectories over the simulation period, assessing the global stability of the protein in different states: Apo–GSTP1 (A), Crocin–GSTP1 (B), Crocin I–GSTP1 (C), and Crocin II–GSTP1 (D). The convergence of the curves indicates system equilibration. Per-residue RMSF profiles highlighting local protein flexibility and rigidity for Apo–GSTP1 (E), Crocin–GSTP1 (F), Crocin I–GSTP1 (G), and Crocin II–GSTP1 (H). Green vertical bars (in E–H) denote residues interacting with the ligand. Lower peaks generally indicate rigid regions, while higher peaks correspond to flexible loops or termini.

Crocin I is a positive modulator of GSTP1 enzymatic activity

We next investigated whether the computationally predicted interaction between Crocin I and GSTP1 translated into a functional effect on catalysis. Initial activity screening showed that Crocin I, Crocin II, and total Crocins increased GSTP1 activity at the tested concentrations, whereas NBDHEX produced the expected inhibitory effect (Figure 5(A)). Concentration–response analysis further showed an overall increase in GSTP1 activity with increasing Crocin I concentrations, with the greatest activation observed at 100 μM (Table 2). These findings identify Crocin I as a positive modulator of GSTP1 enzymatic activity.

Figure 5.

Experimental validation of Crocin I as a positive modulator of GSTP1. Enzymatic activity of GSTP1 in the presence of Crocin I, Crocin II, and Total Crocins relative to the untreated control (A). Michaelis–Menten kinetics of GSTP1 with and without Crocin I (100 μM) (B). Lineweaver–Burk transformation of the kinetic data (C). BLI sensorgrams of Crocin I binding to immobilised GSTP1(D). Thermal denaturation (melting) curve and corresponding first-derivative profile (melt peak) of GSTP1 determined by DSF (E).

Table 2.

Concentration–response of Crocin I on GSTP1 activity and EC50 assessment.

Treatment	Concentration (μM)	Relative GSTP1 activity (% vehicle, mean ± SD)
Vehicle	0	Normalisation reference
NBDHEX	4	59.6 ± 18.5
Crocin I	0.03	127.9 ± 28.1
Crocin I	0.10	106.7 ± 28.6
Crocin I	0.30	111.2 ± 18.4
Crocin I	1	103.9 ± 23.4
Crocin I	3	133.6 ± 44.9
Crocin I	10	166.1 ± 69.8
Crocin I	30	247.7 ± 106.5
Crocin I	100	279.9 ± 94.2
Crocin I	300	201.6 ± 38.9

To define the kinetic basis of this effect, Michaelis–Menten parameters were determined in the absence and presence of Crocin I. Shared-K_m analysis yielded an apparent K_m for CDNB of 1.974 mM (95% CI, 0.586–3.362 mM). Crocin I (100 μM) increased V_max from 0.007042 mM min⁻¹ (95% CI, 0.005761–0.008323 mM min⁻¹) to 0.012700 mM min⁻¹ (95% CI, 0.008410–0.016989 mM min⁻¹), corresponding to an approximately 1.8-fold increase (Figure 5(B) and Table 3). The treatment-associated increase in V_max was consistently observed, without producing a consistent change in the apparent K_m for CDNB. Lineweaver–Burk analysis produced a similar kinetic pattern (Figure 5(C)). These results are consistent with an apparent non-competitive activation pattern towards CDNB under the tested conditions.

Table 3.

Kinetic parameters of GSTP1.

Parameter	Unit	Mean	SD	95% CI lower	95% CI upper
Apparent shared Km, CDNB	mM	1.974	0.559	0.586	3.362
Vmax vehicle	mM min−1	0.007042	0.000516	0.005761	0.008323
Vmax + Crocin I	mM min−1	0.012700	0.001727	0.008410	0.016989
Delta Vmax (Crocin − vehicle)	mM min−1	0.005658	0.002214	0.000158	0.011158

BLI was subsequently used to examine the physical association between Crocin I and GSTP1. Crocin I produced concentration-dependent association responses on immobilised GSTP1, followed by rapid signal decay during the dissociation phase (Figure 5(D)). The concentration ordering of the sensorgrams provides qualitative evidence that Crocin I associates directly with GSTP1 and complements the functional enzyme data.

DSF further showed that Crocin I altered the thermal behaviour of GSTP1. Apparent T_m was defined as the temperature corresponding to the primary downward minimum of − d(RFU)/dT. Apo GSTP1 exhibited an apparent T_m of 47.9 ± 0.4 °C, whereas Crocin I shifted the apparent T_m to 43.8 ± 0.7 °C (ΔT_m = −4.1 °C; Figure 5(E)). NBDHEX alone produced only a limited shift, with an apparent T_m of 47.5 ± 0.6 °C. In the presence of both NBDHEX and Crocin I, the apparent T_m remained shifted to 43.6 ± 0.8 °C (ΔT_m = −4.3 °C), closely resembling the response to Crocin I alone. The apparent T_m in the presence of both compounds was thus comparable to that of Crocin I alone.

To test whether predicted contacts contribute to the functional response, Glu98 and Asn205—residues that formed persistent hydrogen bonds with Crocin I in the MD simulations while remaining outside the core catalytic residues (Tyr8 and Tyr109)—were replaced (E98Q, N205L) and the resulting variants evaluated (Figure S3(E)). Crocin I retained its ability to enhance the activity of both variants across the tested concentrations (Table S1). Therefore, neither E98 nor N205 was individually essential for Crocin I-associated activation under these assay conditions.

Collectively, the concentration–response profile, kinetic analysis, BLI association signals, DSF thermal shifts, and mutational results identify Crocin I as a positive modulator of GSTP1.

GSTP1 is downregulated in PRAD

GSTP1 expression is markedly reduced in PRAD relative to normal prostate tissue. Pan-cancer analysis of TCGA data showed significantly lower GSTP1 mRNA levels in PRAD compared to normal tissue (Figure 6(A)), a pattern consistently observed in both paired tumour-normal samples (Figure 6(B)) and unpaired cohorts (Figure 6(C)). Immunohistochemistry (IHC) recapitulated this finding at the protein level, with robust GSTP1 staining in normal prostate epithelium (Figure 6(E)) and near-complete loss of staining in PRAD tissues (Figure 6(D)). Correlation analysis of GSTP1 expression across tumour-infiltrating immune cell subsets and cancer cell lines is shown in Figure 6(G). Survival analysis showed a trend towards favourable prognosis with higher GSTP1 expression (HR = 0.67, 95% CI: 0.45–1.02, p = 0.060; Figure 6(F)), though this did not reach statistical significance. ROC curve analysis yielded an AUC of 0.929 (Figure 6(H)), indicating robust discriminatory power between tumour and normal samples. Together, these exploratory expression data reflect the epigenetic silencing of GSTP1 in PCa and may offer additional context for pharmacological strategies aimed at modulating GSTP1 activity.

Figure 6.

Clinical significance of GSTP1 downregulation in PRAD. Pan-cancer expression profile of GSTP1 across TCGA + GTEx datasets (A). Comparative expression analysis of GSTP1 in paired tumour-normal tissues (B) and unpaired samples (C). IHC staining validation in PRAD tissue (D) and normal prostate tissue (E). Kaplan–Meier survival analysis based on GSTP1 expression levels (F). Correlation analysis of GSTP1 expression across different cell lines/immune sets (G). ROC curve evaluating the diagnostic specificity and sensitivity of GSTP1 (H).

Network pharmacology identifies HRPC-associated pathways as candidate targets of Crocin I

To explore potential targets of Crocin I beyond GSTP1, we performed network pharmacology analysis. Intersection of predicted Crocin I targets with PCa-associated genes identified a set of shared candidates (Figure 7(A)), which were organised into a PPI network (Figure 7(B)). MCODE clustering resolved the PPI network into discrete, densely connected modules, which were then subjected to functional enrichment analysis using Metascape (Figure 7(C,D)). GO enrichment identified hormone-refractory prostate cancer (HRPC) as a prominently enriched disease term among the clustered targets (Figure 7(E)). KEGG pathway analysis further highlighted the NF-κB signalling pathway as a central node within the target network (Figure 7(F)). These associations, while computationally derived and requiring independent experimental validation, suggest that Crocin I may engage signalling pathways relevant to drug-resistant PCa in addition to its direct modulation of GSTP1.

Figure 7.

Network pharmacology analysis revealing the therapeutic mechanism of Crocin I in PCa. Venn diagram illustrating the intersection of Crocin-predicted targets and PCa-associated targets (A). PPI network of the overlapping targets (B). MCODE cluster analysis of the PPI network (coloured by cluster ID) (C). Network visualisation coloured by significance (p values) (D). GO enrichment analysis (E). KEGG pathway enrichment analysis highlighting the NF-κB signalling pathway (F).

Discussion and conclusion

The central question raised by this study is not simply whether Crocin I interacts with GSTP1, but how this interaction enhances catalytic capacity without producing conventional substrate-competitive behaviour. Our findings identify Crocin I as a positive modulator that alters GSTP1 catalytic performance rather than measurably improving its apparent recognition of CDNB. The integration of kinetic, mutational, thermal-shift, and binding evidence therefore supports a direct biochemical relationship between Crocin I and GSTP1, while allowing the structural basis of this modulation to remain open.

The kinetic behaviour of Crocin I provides the clearest functional distinction from conventional active-site inhibition. Crocin I increased the apparent V_max of GSTP1 by approximately 1.8-fold without producing a consistent change in the apparent K_m for CDNB, yielding an apparent non-competitive activation pattern under the tested conditions. Whereas conventional active-site inhibitors typically suppress catalysis by restricting substrate access or stabilising catalytically inactive configurations^28–31, the observed increase in V_max suggests that Crocin I influences a catalytic step that becomes rate-limiting after substrate association. This interpretation does not, by itself, establish an allosteric mechanism or define a particular conformational state; rather, it provides an experimentally testable framework in which Crocin I enhances GSTP1 turnover without acting as a conventional substrate competitor.

The computational and mutational results further refine this framework. Molecular modelling predicted that Crocin I could interact with a peripheral region extending towards the boundary of the G- and H-site environment rather than occupying the conventional substrate-binding cavity. Such a location is structurally plausible because GSTP1 catalytic competence depends on communication between the active-site architecture and the dimeric organisation of the enzyme³², and ligand-interaction networks can extend across the active-site and subunit-interface regions³³. However, replacement of either E98 or N205 did not abolish Crocin I-associated activation. These findings indicate that neither residue is individually essential for the functional response and argue against a binding mechanism dominated by a single predicted contact. They do not exclude either residue from contributing to the interaction, but instead suggest that Crocin I recognition may be supported by multiple contacts or by a region broader than the initially modelled residues.

The biophysical observations show that Crocin I association is accompanied by altered GSTP1 thermal behaviour without assigning a specific structural mechanism. Crocin I shifted the primary thermal transition of GSTP1 to a lower temperature, indicating a change in apparent thermal stability. Ligand-induced thermal shifts can occur in either direction because binding may alter local structural organisation, conformational heterogeneity, solvent exposure, or aggregation behaviour³⁴; Consequently, the negative shift should be interpreted as evidence of ligand-dependent perturbation rather than as direct proof of stabilisation or allostery. Complementing this observation, BLI produced reproducible concentration-ordered association responses between Crocin I and immobilised GSTP1. Although the rapid dissociation phase limits quantitative interpretation of binding affinity³⁵, the sensorgrams provide qualitative evidence of physical association and connect the thermal response with the enzymatic effect.

The NBDHEX experiment imposes an additional mechanistic constraint on this interaction. NBDHEX is an H-site-directed GST inhibitor whose inhibitory mechanism has been characterised in solution studies³⁶. Importantly, crystallographic analysis captured NBDHEX within the GSTP1 H-site without covalent attachment to the protein, whereas formation of the GSH–NBDHEX σ-complex was structurally observed for GSTM2 rather than GSTP1³⁷. In the present study, the Crocin I-associated thermal transition persisted when Crocin I and NBDHEX were added together and closely resembled that produced by Crocin I alone. This result argues against a simple model in which the biophysical effect of Crocin I requires an unoccupied conventional H-site configuration. Nevertheless, the DSF experiment does not independently establish simultaneous ligand binding or prove the existence of completely separate binding sites; instead, it supports the more restrained conclusion that the effect of Crocin I is not governed solely by conventional competition for the NBDHEX-recognised substrate-binding environment.

The biological implications of increasing GSTP1 activity are necessarily context-dependent. GSTP1 functions both as a detoxification enzyme and as a regulator of redox-sensitive signalling through PPIs³⁸. In PRAD, where GSTP1 expression is frequently reduced, the effect of a positive modulator will therefore depend on the amount of residual functional enzyme, cellular redox status, and the balance between GSTP1-mediated detoxification and signalling functions³⁹. This context dependence does not alter the biochemical conclusion that Crocin I enhances GSTP1 activity; rather, it identifies GSTP1 abundance and cellular state as testable determinants of the downstream response.

Exploratory transcriptomic and network analyses placed Crocin I–GSTP1 modulation within pathways associated with oxidative stress and the MAP3K7/NF-κB signalling axis^13,40. These analyses are hypothesis-generating rather than direct evidence of a cellular mechanism, but they predict that Crocin I-responsive cellular phenotypes should depend on GSTP1 expression and be accompanied by measurable changes in redox balance and pathway activity. This prediction provides a defined route for connecting the biochemical findings to PCa biology.

Crocin-related compounds have previously been associated with antioxidant and antitumour activities through multiple cellular pathways²³. The present study adds a specific biochemical component to this broader pharmacology by identifying GSTP1 as a directly responsive enzyme. Structural characterisation of the Crocin I–GSTP1 complex and additional functional substitutions across the predicted interaction region could distinguish whether the response arises from a peripheral binding surface, an extended G/H-site interaction network, or another mode of association. These alternatives constitute experimentally separable models rather than prerequisites for recognising the positive modulatory activity established here.

In conclusion, Crocin I is identified as a positive modulator of GSTP1 that enhances catalytic capacity through an apparent non-competitive activation pattern and produces reproducible binding and thermal responses. These findings establish a biochemical foundation for testing how Crocin I–GSTP1 modulation influences redox regulation and GSTP1-dependent signalling in PCa contexts.

Funding Statement

This work was funded by Ph.D. Start-up Foundation of Jiangxi Science and Technology Normal University (Grant No. 2023BSQD39), the National Natural Science Foundation of China (Grant No. 82304606), and Natural Science Foundation of Jiangxi Province (Grant No. 20242BAB25576).

Disclosure statement

The authors declare no conflict of interest.

Data availability statement

Data available on request from the authors.