The dried twig of AT was purchased from Omniherb (Gyeongbok, Korea). TNF-α (T6674), anti-β-actin antibody (A1978), thiazolyl blue tetrazolium bromide (MTT) (M2128), streptozotocin (S0130), poly-L-lysine (P4707), and bovine serum albumin (BSA) (A8806) were purchased from Sigma-Aldrich (St. Louis, MO, United States). Lipofectamine 2000, goat anti-rabbit IgG (#65-6120), and goat anti-mouse IgG (#62-6520) were purchased from Invitrogen (Carlsbad, CA, United States). Anti-intercellular adhesion molecule-1 (ICAM-1) (sc-7891 and sc-8439), anti-NF-κB (sc-8008), anti-SIRT1 (sc-74504), anti-vascular cell adhesion molecule-1 (VCAM-1) (sc-8304) antibodies, anti-IgG-FITC (sc-2010), protein G agarose (sc-2002), DAPI (sc-3598), and Immuno In situ Mount (sc-45088) were obtained from Santa Cruz Biotechnology (CA, United States). Calcein-AM (C3100MP), goat anti-mouse Alexa Fluor 488 (A32723), and goat anti-rabbit Alexa Fluor 568 (A11011) were purchased from Thermo Fisher Scientific (Waltham, MA, United States). Anti-phospho-ERK1/2 (#9101), anti-ERK1/2 (#9102), anti-phospho-p38 MAPK (#9211), anti-p38 MAPK (#9212), anti-phospho-SAPK/JNK (#9255), and anti-JNK2 (#9258) antibodies were obtained from Cell Signaling Technology (Beverly, MA, United States). All materials used for SDS-PAGE were purchased from Bio-Rad (Hercules, CA, United States). The Sirt1 small interfering RNA (siRNA, sc-40986) and non-targeting scrambled RNA (control siRNA, sc-37007) were obtained from Santa Cruz Biotechnology (Sparks, MD, United States).

Extraction of AT was performed as previously reported, with minor modifications (Yu et al., 2010). AT (400 g) was powdered and extracted with 2.5 L of 70% ethanol at 80°C for 2 h. The ethanol extract was filtered through a Whatman #2 filter paper (Whatman International, Maidstone, United Kingdom) and concentrated using a rotary evaporator (model VV2000; Heidolph, Walpersdorfer, Germany) to obtain the final extraction. The samples were frozen in −70°C before being freeze-dried using a FD8505S freeze-dryer (Ilshin Biobase Co., Ltd., Busan, South Korea), and a voucher (AT-0100) specimen was deposited in the pharmacology laboratory at the College of Oriental Medicine, Kyungju, South Korea. The final yield of the ethanol extract was 19.6%. The content of salidroside, the major component of AT, was calculated from the relevant peak area using an external standard method and quantified as 6.41% (w/w) in AT (Bae et al., 2022) (Supplementary Figure 2).

A total of 300 mg AT extract was fermented with 100 mL Luria–Bertani (LB) media containing 0.2–20% B. subtilis NCDO 1769T (X60,646) culture for 7 days (ATF) and then frozen at −70°C before use (ATF-0100) (Gum et al., 2017). Negative AT samples were subjected to 100 mL LB media without B. subtilis for 7 days (AT). After fermentation, both AT or ATF samples were stored in −70°C for further experiments.

HPLC analyses were performed as previously published with minor modifications (Meng et al., 2006). The HPLC analyses of samples were carried out using an Agilent 1260 HPLC system (1260 Infinity II; Agilent Technologies, Santa Clara, CA, United States) coupled to the Agilent G7115A-1260 DAD WR detector (diode array detector). In brief, 10 μL of AT or ATF samples (3 mg/mL) were injected into the Agilent ZORBAX Eclipse plus C18 column (4.6 mm × 250 mm, 5 μm) at 25°C. The mobile phase was 0.1% formic acid in H₂0 (A) and 100% MeOH (B), run at 1 mL/min; peaks were detected with a UV detector at 225/278 nm. The ratios of the A/B solvent were 80/20, 80/20 to 0/100 (linear gradient), and 0/100 at running times of 0–10, 11–30, and 31–35 min, respectively. SAL or TYR at 500 ppm was used as the standard. The SAL and TYR peaks were observed at 7.5 and 9.5 min, respectively.

EA. Hy926, a human umbilical vein endothelial cell line fused with the human carcinoma cell line (CRL-2922), was purchased from the ATCC (Rockville, MD, United States). Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (LM 001-05, Welgene, Gyeongsangbuk-do, Korea) containing 10% FBS, 50 μg/mL streptomycin, and 50 units/mL penicillin at 37°C in a humidified 5% CO₂ atmosphere. THP-1 cells (ATCC^® TIB-202™), a monocyte cell line, were purchased from the ATCC (Rockville, MD, United States). The cells were cultured in RPMI 1640 medium (LM 011-01, Welgene, Gyeongsangbuk-do, Korea) containing 10% FBS, 0.05 mM 2-mercaptoethanol, 50 μg/mL streptomycin, and 50 units/mL penicillin. Cells were treated with TNF-α for the indicated time with or without candidates (AT, ATF, SAL, or TYR) pretreated for 1 h.

Cells were seeded in a 96-well culture plate. After reaching 80% of confluence, cells were incubated with candidates at different concentrations for 24 h. The MTT solution (5 mg/mL) was incubated for 2 h. The assays were stopped by adding DMSO, and the cell viability was observed at 520 nm using a Sunrise™ Absorbance microplate reader (Tecan, Bionics, Seoul, Korea).

The monocyte–endothelial interaction was evaluated by the measurement of fluorescent labeling monocytes, as previously described (Nguyen et al., 2019). Endothelial cells were seeded in a 96-well culture black plate and then incubated with candidates for 1 h, following exposure of 10 ng/mL TNF-α for 6 h. Human monocyte cells were labeled with calcein-AM at 37°C for 1 h. Calcein-AM-labeled cells were co-cultured with the monolayer of endothelial cells for 1 h. After the incubation, the recruited monocytes were measured using microplate reader (Bio-Tek Instruments, Winooski, VT, United States) with excitation wavelength 490 nm and emission wavelength 510–570 nm. The average fluorescence intensity represented the number of monocytes adhered to endothelial cells.

MCP-1 cytokine levels in the culture supernatants were detected by ELISA using a commercial Human MCP-1 ELISA kit (Pierce Biotechnology, IL, United States), according to manufacturer’s protocol. The OD values were measured using a Sunrise™ Absorbance microplate reader (Tecan, Bionics Co., Seoul, Korea). Concentrations of MCP-1 cytokines were extrapolated from the standard curves provided.

RT-PCR was performed as previously reported (Kim et al., 2008). The primers encoding MCP-1 (forward primer 5′ -GAT​GCA​ATC​AAT​GCC​CCA​GTC -3′ and reverse primer 5′- TTT​GCT​TGT​CCA​GGT​GGT​CCA​T -3′) and β-actin (forward primer 5′- GACTACCTCATGAAGATC -3′ and reverse primer 5′- GATCCACATCTGCTGGAA -3′) were synthesized from Bioneer (Cheongwon, Korea).

Immunoprecipitation and immunoblotting were performed, as previously reported (Nguyen et al., 2019). For immunoprecipitation, cell lysates were incubated with antibodies against SIRT1 or NF-κB at 4°C for 2.5 h and protein G plus-agarose beads at 4°C for 1 h. The immunocomplexes were washed with IP buffer and dissolved by boiling. Samples were then run on SDS-PAGE gels. In immunoblotting, primary antibodies were incubated at 4°C overnight, and secondary antibodies were incubated at room temperature for 1 h. The target proteins were developed using an ECL chemiluminescence detection kit (Amersham Biosciences, Bucks, United Kingdom). The protein expressions were normalized with β-actin immunoblot. The densitometry was measured using Gel-Pro Analyzer software (Media Cybernetics, MD, United States).

The full-length VCAM-1 and ICAM-1 firefly luciferase reporter gene construct (VCAM-1: -1,716 to +119 bp, ICAM-1: -1,350 to +45 bp) were transfected as previously reported (Kim et al., 2008; Nguyen et al., 2019). The luciferase activities in cell lysates were measured using a GloMax 20/20 Luminometer (Tuner BioSystems, Sunnyvale, United States). The relative luciferase signal was normalized by dual transfection with the pSV-β-galactosidase control vector.

Cells were seeded in an 8-well chamber slide system (Nunc Lab-Tek II, Seoul, Korea). Immunofluorescence-labeled cells were evaluated as previously reported (Wang et al., 2016b). After drug treatment for the indicated time, cells were fixed with ice-cold MeOH for 15 min. After permeabilizing and blocking, the anti-NF-κB primary antibody was incubated overnight at 4°C. The secondary antibody against anti-mouse-FITC was then incubated. The images were taken in random regions using a BioTek Lionheart FX Automated Microscope (Agilent, VT, United States).

Cells were transfected with siRNAs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, United States), following the manufacturer’s protocol. Human Sirt1 or scrambled siRNA (10 nM) was transiently transfected for 3 h to knock down Sirt1. After an 18 h-recovery period, promoter-luciferase assay, immunoblotting, and adhesion assays were performed, as described above.

The full sequence alignment of human SIRT1 (NP_036370.2) was obtained from NCBI by FASTA. The crystal structures of SIRT1 PDB IDs 4ZZH and 4ZZJ were retrieved from the Protein Data Bank online database (PDB). The sequences were imported, and the structure-based sequence alignment was performed using BioEdit software version 7.2 (Pannek et al., 2017).

Three-dimensional sequences of SIRT1 with the NTD region were retrieved from PDB: 4ZZH and 4ZZJ (www.rcsb.org). All structures were optimized by the removal of ligands, substrates, ions, and water molecules. The active compounds’ (SAL and TYR) structures were downloaded from the NCBI PubChem online database (http://pubchem.ncbi.nlm.nih.gov/) and energy minimized. Molecular docking of PDB and compounds was analyzed by AutoDockZn and PyRx softwares, to predict the binding affinity and ligand conformational change. The proper docking conformations were chosen based on binding affinity scores and the root-mean-square deviation (RMSD) values. The binding pockets and interactions between the ligand and protein were visualized by PyMOL and Biovia Discovery Studio 2021 software applications.

The docked complexes of the active compounds and SIRT1 were used as the initial coordinates for molecular dynamics (MD) simulations using the GROningen MAchine for Chemical Simulations (GROMACS) (Shamsi et al., 2024). MD simulations were performed with GROMACS version 5.1.2 with the CHARMM36-jul2022 force field. The complexes were immersed in a dodecahedron box and solvated using the CHARMM-modified TIP3P water model. The topologies of SAL and TYR were generated using the CHARMM General Force Field (CGenFF). The systems were neutralized with counter ions. The energy minimization process was initiated with restraints on both the ligand and protein, followed by an unrestrained minimization step using the steepest descent method. The equilibration phases were conducted by NVT ensemble for 100 ps, followed by NPT ensemble for an additional 100 ps, with the temperature maintained at 300 K. The final MD trajectories were run for 100,000 ps (100 ns) with a 2.0-fs time step. Post-simulation analysis was performed using the following GROMACS utilities: root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), solvent-accessible surface area (SASA), and hydrogen bond analysis. The conformational changes in the ligand–protein complexes were visualized at the 20-ns interval using PyMOL software.

The animal study was conducted in accordance with the institutional guidelines based on the principles of laboratory animal care (NIH publications No. 85-23, revised 1996) and Korean laws of laboratory animal care (Health & Welfare Committee Registration No. 9025, revised 2010) and was approved by the Institutional Animal Care and Use Committee of the Dongguk University (protocol number IACUC-2023-15). Male C57Bl/6JBomTac mice were purchased from Daehan Bio Link (Chungbuk, South Korea). The 5-week-old mice received three consecutive intraperitoneal injections of streptozotocin (STZ) at a dose of 55 mg/kg in citrate buffer (0.1 M, pH 4.5). One week after STZ-treatment, mice with glucose levels exceeding 250 mg/dL were divided into four groups: control, STZ, STZ + AT, and STZ + ATF. AT or ATF (50 mg/kg) was orally administered daily for 4 weeks. All mice were anesthetized with pentobarbital sodium, and their eyes were collected and stored in 4% paraformaldehyde until further processing.

Eye samples were soaked in 30% sucrose (0.2 M phosphate buffer) at 4°C overnight before fixing in the optimal cutting temperature (OCT) compound for cryosectioning. Retina radial sections (8 μm) were placed on poly-L-lysine-coated slides and heated at 60°C for 2 h. Samples were permeabilized with washing buffer (0.1% Triton X-100) at room temperature for 20 min and then rinsed with PBS twice for 5 min. Retinal samples were blocked with buffer containing 5% fetal bovine serum, 2% BSA, and 0.1% Triton X-100 at room temperature for 30 min. Tissue samples were stained with primary antibodies for VCAM-1 and ICAM-1 (1:400 dilution) overnight at 4°C. The sections were then incubated with secondary antibodies conjugated to goat anti-mouse Alexa Fluor 488 and goat anti-rabbit Alexa Fluor 568 (1:600 dilution) for 30 min at room temperature. For counterstaining, DAPI was incubated with the sections for 5 min. Sections were analyzed using the BioTek Lionheart FX Automated Microscope (Agilent, VT, United States).

Statistical analyses were performed using SPSS 13.0. (IBM SPSS, Chicago, United States), and all data are presented as the mean ± standard error (S.E.). The statistical comparisons between the groups were determined using one-way ANOVA, followed by Dunnett’s post hoc test.

In vascular inflammatory conditions, circulating monocytes bind and migrate through the blood vessels’ endothelial monolayer, which plays a pivotal role in the initiation of vascular dysfunction (Nguyen et al., 2019). Therefore, we investigated the effects of AT, SAL (the main bioactive compound of AT), and TYR (the aglycone of SAL) on the TNF-α-stimulated adhesion of monocytes to endothelial cells. The concentrations of candidates used in this study did not affect cell viability (Supplementary Figure 1). Endothelial cells were pretreated with AT, SAL, or TYR for 1 h, and following TNF-α exposure for the next 6 h. We then measured the attached calcein-AM-labeled monocytes to the endothelial cells. TNF-α induced monocyte adhesion (∼5.5-fold) compared to the control (Figure 1). AT, SAL, and TYR suppressed the elevated monocyte attachment by TNF-α with IC₅₀ values of 1.62 μg/mL, 0.94 μM, and 0.14 μM, respectively (Figures 1A, B). TYR exhibited high potency in reducing TNF-α-induced monocyte recruitment on endothelial cells. These results suggest that TYR may exert beneficial effects as an active metabolite in inhibiting monocyte adhesion to endothelial cells.

SAL is the major compound of AT, whereas TYR is very minor (Hwang et al., 2013; Lee et al., 2014). We previously demonstrated that fermentation with B. subtilis subsp. subtilis NCDO 1769T (X60646) showed enhanced beneficial effects through secondary metabolite production (Gum et al., 2017). Therefore, we performed AT fermentation with B. subtilis subsp. subtilis NCDO 1769T (X60646). The fermented product of AT showed a bioconversion from SAL to TYR (Supplementary Figure 2), analyzed by HPLC. In ATF, SAL and TYR contents were 4.44% (w/w) and 0.75% (w/w), respectively. The elevation of monocyte recruitment under TNF-α exposure was inhibited by ATF with an IC₅₀ value of 0.49 μg/mL (Figure 1C). The lower IC₅₀ value of ATF indicates that ATF is more effective than AT in inhibiting monocyte attachment to endothelial cells, which may be attributed to the biotransformation of SAL in AT.

TNF-α induces endothelial cell activation, leading to transactivation of adhesion molecules such as VCAM-1 and ICAM-1, which represents the initial step in vascular inflammation (Nguyen et al., 2019). The previous report showed that TNF-α increased VCAM-1 expression, which peaked at 12 h, while ICAM-1 expression showed a sustained induction pattern until 24 h (Nguyen et al., 2019). Therefore, we performed immunoblotting with VCAM-1 and ICAM-1 after the indicated times of TNF-α incubation with AT or ATF pretreatment. The protein expressions of VCAM-1 and ICAM-1 were elevated ∼4-folds under TNF-α exposure, which were significantly inhibited by AT or ATF (Figure 2A).

We performed reporter gene assays in cells transfected with the promoter of VCAM-1 or ICAM-1 to determine transcriptional activity under TNF-α exposure. TNF-α increased the luciferase activity of VCAM-1 by 5-fold and ICAM-1 by 4-fold (Figure 2B). Incubation with AT or ATF strongly attenuated TNF-α-mediated transactivation. AT and ATF inhibited VCAM-1 luciferase activity with a similar effective range, with IC₅₀ values of 2.86 μg/mL and 2.09 μg/mL, respectively. The stronger protective effect of ATF was shown in ICAM-1 luciferase activity with an IC₅₀ value of 1.27 μg/mL compared to AT with an IC₅₀ value of 3.09 μg/mL. The significant suppression of ATF in the attenuation of TNF-α-mediated transcriptional activation of adhesion molecules was consistent with the inhibition of those expressions.

MCP-1 is one of the key cytokines that regulates the migration and infiltration of monocytes/macrophages, which initiates further inflammatory cellular responses (Tedgui and Mallat, 2006). MCP-1 is induced and secreted in endothelial cells stimulated by TNF-α (Meng et al., 2022). Therefore, we tested whether AT or ATF prevented TNF-α-induced MCP-1. We first measured MCP-1 secretion in the culture media after TNF-α incubation for 36 h. AT or ATF pretreatment abrogated TNF-α-induced MCP-1 production (Figure 3A). Next, we evaluated the level of MCP-1 mRNA. After 6 h of TNF-α treatment, MCP-1 mRNA levels increased 4-fold (Figure 3B). AT and ATF significantly suppressed MCP-1 mRNA, and inhibition by ATF was stronger than that by AT. These results support the hypothesis that ATF is potent due to its conversion to the active compound TYR.

MAPKs play an important role in the regulation of inflammatory signaling cascades (Dong et al., 2020). We performed immunoblotting of phosphorylated forms of ERK1/2, p38, and JNK to analyze the upstream signaling of AT and ATF in response to TNF-α exposure. Cells were treated with AT or ATF at a dose range of 1–10 μg/mL. Then, TNF-α was incubated for the next 30 min to detect phosphorylated forms of p38 and ERK1/2. Phosphorylated JNK was observed after incubation with TNF-α for 5 h. TNF-α significantly increased the phosphorylation of ERK1/2, p38, and JNK, which were ameliorated by AT or ATF pretreatment (Figure 4).

NF-κB is a master transcriptional factor that regulates inflammatory gene expression (Baker et al., 2011). Therefore, the effects of AT and ATF on NF-κB transcriptional activation under TNF-α exposure were determined. Pretreatment of AT or ATF abrogated TNF-α-induced transcriptional activity of NF-κB with IC₅₀ values of 3.77 μg/mL and 0.90 μg/mL, respectively (Figure 5A). TNF-α-induced translocation of NF-κB from the cytoplasm into the nucleus was observed, which was inhibited by AT or ATF (Figure 5B). These results suggest that AT and ATF inhibit the transcription of the inflammatory genes by blocking TNF-α-induced nuclear translocation and transcriptional activity of NF-κB.

We then confirmed the role of MAPKs and NF-κB in the regulation of adhesion molecule expressions and adhesion between endothelial cells and monocytes using chemical inhibitors. Cells were preincubated with the indicated doses of inhibitors, including PD98059 (ERK1/2 inhibitor), SB203580 (p38 inhibitor), SP600125 (JNK inhibitor), and Bay11-7082 (NF-κB inhibitor), and following TNF-α incubation. The TNF-α-induced luciferase activities of adhesion molecules VCAM-1 and ICAM-1 were mitigated by inhibitors of MAPK and NF-κB (Figure 6A). The attachment of calcein-AM labeled monocytes to endothelial cells induced by TNF-α was also inhibited by the blocking of MAPKs and NF-κB (Figure 6B). These results confirmed that MAPK/NF-κB signaling pathways play an important role in TNF-α-regulated endothelial cell inflammation.

NF-κB posttranslational modification is a critical step in regulating inflammatory gene expression (Chaturvedi et al., 2011). SIRT1, an NAD⁺-dependent deacetylase, binds to NF-κB and inhibits transcriptional activity of NF-κB by deacetylating the RelA/p65 subunit at Lys³¹⁰ (Yeung et al., 2004). Therefore, we tested whether SIRT1 was involved in AT or ATF-mediated NF-κB inhibition. NF-κB was bound to SIRT1 in the resting state, which failed to induce NF-κB transcriptional activity. The NF-κB/SIRT1 complex was dissociated in response to TNF-α, which was inhibited by AT or ATF with high potency in ATF (Figures 7A, B). TNF-α exposure resulted in the release of NF-κB from SIRT1 and increased acetylation of NF-κB at Lys³¹⁰. AT or ATF inhibited the acetylation of Lys³¹⁰ by promoting the interaction between NF-κB and SIRT1, thereby maintaining NF-κB in a deacetylated state. ATF significantly maintained the binding of NF-κB and SIRT1, implying that ATF exhibited a stronger effect in enhancing the NF-κB/SIRT1 interaction than AT did. Thus, the inhibitory effect of AT and ATF on NF-κB transactivation is due to the suppression of acetylated p65, which is attributed to SIRT1-mediated posttranslational modifications of NF-κB.

The role of SIRT1 in the AT and ATF against TNF-α-mediated MAPK/NF-κB signaling pathway and the regulation of adhesive molecules were evaluated by Sirt1 gene knockdown. Silencing Sirt1 significantly reduced its expression (Figure 8A). SIRT1-deficient cells increased NF-κB transactivation along with K310 acetylation of RelA/p65 in the absence of TNF-α, which is consistent with previous findings (Nguyen et al., 2019). The inhibitory effects of AT and ATF on NF-κB transactivation in response to TNF-α were diminished in SIRT1-deficient cells compared to the control group, indicating that SIRT1 is required for AT and ATF to suppress NF-κB acetylation (Figure 8B). Additionally, the differential regulation of MAPK phosphorylation by AT and ATF was attenuated in Sirt1 siRNA-transfected cells (Figures 8C, D). In these cells, AT and ATF also failed to inhibit TNF-α-induced VCAM-1 and ICAM-1 transactivations (Figure 8E). Furthermore, Sirt1 silencing mitigated the inhibitory effects of AT and ATF on TNF-α-induced adhesion, as measured by the fluorescent calcein-AM signal on endothelial cells, which is consistent with NF-κB regulation (Figure 8F). These findings suggest that SIRT1 plays a critical role in the protective effects of AT and ATF against TNF-α-mediated vascular inflammation.

We found that SIRT1 is an important target of AT or ATF for regulating the anti-inflammatory responses. We next determined SIRT1 expression by AT, ATF, SAL, or TYR after 24 h. Compared to AT, ATF significantly increased the expression of SIRT1 by 3.5-fold (Figure 9A). SAL and TYR also exhibited 5-fold and 8-fold upregulation of SIRT1 protein expression, respectively, supporting the idea that these candidates increase both SIRT1 activity and SIRT1 expression (Figure 9B).

We next performed molecular docking to analyze how SAL or TYR (the active compounds in AT and ATF) interacts with SIRT1. The multiple alignment results of hSIRT1 and two SIRT1 homologous residues (PDB IDs 4ZZH and 4ZZJ) indicate that all structures show strong homology (>90%) with the NTD and the catalytic domains of hSIRT1 (Supplementary Figure 3). The interaction of bioactive compounds with the allosteric region of the NTD can initiate conformational changes, leading to stabilization and induction of deacetylase activity of SIRT1 (Bonkowski and Sinclair, 2016). Hence, we generated molecular docking of the NTD region of SIRT1 with SAL or TYR, and detailed interactions were illustrated; all four docking results showed the same binding pocket regions (Figure 10). The glycol moiety of SAL interacted with SIRT1 via hydrogen bonds to Leu²⁰⁵, Pro²⁰⁷, Thr²⁰⁹ (4ZZH), Glu²⁰⁸, and Thr²⁰⁹ (4ZZJ); tyrosol residue of SAL showed interactions with Ile²²⁷ and Glu²³⁰ (Figures 10A, B). Interestingly, TYR also interacted with 4ZZH and 4ZZJ at Ile²²⁷ and Glu²³⁰, which were identical to the tyrosol residue of SAL (Figures 10C, D). These docking data indicate that the tyrosol moiety in both SAL and TYR bind to the same binding pocket of SIRT1 and that the binding distance toward SIRT1 is closer in TYR than in SAL, which could potentially regulate SIRT1 activity.

We evaluated the structural stability and dynamic behavior of ligand–protein complexes using MD simulation. The RMSD plot showed that the initial fluctuations of native SIRT1 and SIRT1–TYR persisted up to 50 ns, and then, the structures stabilized up to 100 ns (Figure 11A). Interestingly, the SIRT1–TYR complex showed movement of the allosteric region toward the catalytic region of SIRT1 after 80 ns, suggesting that TYR’s interaction with SIRT1 NTD may contribute to SIRT1 activation (Figure 11A, inset). The average RMSD and RMSF of SIRT1–SAL were 0.373 nm and 0.343 nm, respectively, which were lower than those of SIRT1–TYR (Table 1). The initial fluctuations in the RMSF plot of combined trajectories may be due to the random interactions in the NTD region. TYR interacted with the SIRT1 NTD, vibrating more strongly than SAL. The slight vibration of the TYR–SIRT1 complex in the catalytic region indicates the role of TYR in the regulation of SIRT1 interaction (Figure 11B). Next, the trajectories of the MD simulation were analyzed by Rg, SASA, and the number of hydrogen bonds. The Rg plot revealed distinct deviation patterns between the SAL–SIRT1 and TYR–SIRT1 complexes (Figure 11C). The Rg values of SIRT1–SAL and SIRT1–TYR were unstable during the initial 40 ns. After 60 ns, the Rg of SIRT1–TYR stabilized and remained below 2.2 nm, which was lower than the Rg of SIRT1–SAL (2.2–2.6 nm). The mean SASA value of SIRT1–TYR was the lowest compared to native SIRT1 and SIRT1–SAL (Table 1). The SASA plot showed that SIRT1–SAL fluctuated after 60 ns, suggesting slight flexibility of SIRT1 upon ligand binding. In contrast, the SASA value of SIRT1–TYR significantly decreased compared to that of SIRT1–SAL at the end of the simulation (after 60 ns) (Figure 11D). SAL exhibited a higher number of hydrogen bonds with SIRT1, with a maximum of five bonds, while TYR showed a more variable number of hydrogen bonds throughout the simulation (Figures 11E, F). These MD results revealed important insights into the interaction of SAL and TYR with SIRT1, suggesting that SAL and TYR have different interaction mechanisms with SIRT1.

Diabetic retinopathy (DR) is characterized by retinal nerve deformation and microcirculatory disturbances caused by an inflammatory process (Gerhardinger et al., 2005; Yang et al., 2022). Hyperglycemia induces adhesive molecules like VCAM-1 and ICAM-1, which contribute to damage in the retina, the neuro-vascular coupling tissue. Therefore, we evaluated the levels of VCAM-1 and ICAM-1 in STZ-treated retinal sections. STZ induced VCAM-1 in the inner plexiform layer of the retina (Figures 12A, C). AT and ATF attenuated STZ-induced VCAM-1 expression. ICAM-1 expression increased in the retinal nerve fiber layer of the STZ-treated retina sections, and AT and ATF moderately inhibited ICAM-1 distribution in this layer, with ATF showing a more potent effect (Figures 12B, C). These results suggest that AT and ATF reduce STZ-induced diabetic retina damage by inhibiting adhesion molecules.