*3.5. Molecular Docking*

The initial receptor structure was constructed based on the crystal structure of aldose reductase in complex with cofactor NADP+ and the inhibitor idd594 (PDB code: 1US0) from the Protein Data Bank [20]. All nonstandard groups (HETATM) were deleted except for the inhibitor and cofactor when preparing for the receptor structure. The Protonate 3D module in the Molecular Operating Environment (MOE) program was used to estimate the protonated state of titratable residues and add hydrogen atoms. Meanwhile, MOE was also used to construct the ligand structures (compound **14**). The subsequent molecular docking was performed with an induced fit protocol. During the process, the protein–cofactor complex was defined as the receptor and the position of the inhibitor idd594 was defined as the docking site. The triangle matcher placement with the London ΔG initial scoring methodology was set for conformational sampling and 100 poses were recorded, then the forcefield post-placement refinement with GBVI/WSA ΔG rescoring methodology was utilized to further screen the best docking pose. In addition, the MMFF94x force field was adopted for the whole process.

#### *3.6. ALR2 Expression and Purification*

For the AKR1B1 enzyme assay and SPR measurements, the protein was expressed and purified using the protocol related to the one reported previously [15]. The plasmid (pET28b, Novagen) containing the open reading frame of the human ALR2 gene was kindly provided by Atagenix, Wuhan, Hubei, China. The E. coli strain BL21 gold (DE3) (Novagen) was used to express the hexa-histidine tagged protein after induction with IPTG (Roth) for 16 h at 293 K. The pellet from a 4 L culture was resuspended in a buffer containing 20 mM Tris and 500 mM NaCl (pH 8.0) before being sonicated and centrifuged. A HiTrap chelating HP column (GE Healthcare) was loaded with the supernatant. After a short washing step with a low imidazole concentration, the fusion protein was eluted by applying a gradient of imidazole. The buffer was exchanged with 20mM Tris-HCl, 10 mM NaCl, 1mM EDTA, and the tag cleaved by thrombin (yuanyeBio-Technology Co. Ltd. Shanghai, China). A Hiprep DEAE FastFlow 16/10 column (GE Healthcare) was loaded with the remaining solution. A NaCl gradient was used to elute the ALR2 from the column.

#### *3.7. ALR2 Enzyme Assay (In Vitro)*

In the first step of the polyol pathway, glucose reduction is accompanied by the conversation of NADPH to NADP, where only NADPH has an obvious spectral absorption around 340 nm and NADP has none. Thus, the decrease in OD340nm can represent the consumption of NADPH. Therefore, we detect the change of OD340nm before and after reaction to screen the effective ARIs or evaluate the AR activity; assays with epalrestat were used as positive controls. Briefly, the incubation system contains 10 μL ALR2 enzyme (20 μg/mL), 80 μL DL-glyceraldehyde (0.16 mmol/L) as the substrate, 1.0 mmol/L NADPH·4Na as a coenzyme, and 0.1 mol/L PBS (pH = 6.2). The incubation mixture was minimized to a total volume of 100 μL and ongoing in a 96-well ultraviolet plate. Then, test wells were treated with the tested compounds for 10 min at 25 ◦C. Then, the absorbance was measured using FlexStation 3 Reader (Molecular Devices, San Francisco, CA, USA) at 340 nm. The absorbance of wells treated with PBS and NADPH was considered as 100% (OD1) and the absorbance of wells treated with PBS and DL-glyceraldehyde was considered as 0% (OD2); the inhibitory rate was calculated by the formula ODcompounds − OD2/OD1 − OD2. IC50 represents the concentration that inhibits the ALR2 enzyme by 50%.

#### *3.8. Surface Plasmon Resonance (SPR) Binding Assay*

SPR binding analysis methodology can be used to study molecular interactions. Herein, SPR was applied to measure the interactions between compounds **13**/**14** and ALR2. Initially, ALR2 was prepared in 10 mM sodium acetate (pH 5.0) and then immobilized covalently by an amine-coupling reaction on a CM5 sensor chip. The remaining binding sites of the sensor chip were then blocked by ethanolamine. The addition of compound **13**/**14**, the flow-through analyte, to the chamber resulted in binding to the immobilized protein ligand, producing a small change in the refractive index at the gold surface. In this step, the compound was diluted in PBS-P+ buffer to the desired concentration and was injected over the chip with a flow rate of 10 μL/min. All of the above buffers, solutions, and sensor chips were placed at room temperature before the run. The association time and dissociation time were both set at 60 s. Binding affinities were obtained from the ratio of rate constants to directly characterize the protein-molecular interactions. Data analysis was completed via the state model in T200 evaluation software (Cytiva Danaher, Marlborough, MA, USA).
