Escherichia coli Rosetta (DE3) cells were obtained from Tiangen Biotech. The HEK 293T and HLE-B3 cell lines were acquired from the American Type Culture Collection.

Isopropyl 1-thio-β-D-galactopyranoside (IPTG), Triton X-100, and NP40 solution were obtained from Sangon Biotech. The phenylmethylsulfonyl fluoride (PMSF) and 1-anilinonaphthalene-8-sulfonate (ANS) solutions were purchased from Sigma-Aldrich. Dulbecco’s modified Eagle’s medium (DMEM), DMEM/F-12 (1:1) medium, Lipofectamine 2000, Lipofectamine 3000, and Alexa Fluor Plus 555 antibodies were provided by Invitrogen. The GFP Tag Mouse antibody was procured from Proteintech. Propidium Iodide (PI) and Annexin V-Alexa Fluor 647 were obtained from Yeasen Biotech. All other reagents were local products with analytical grade.

The human βB1–crystallin gene was cloned from a human lens cDNA library, and the S93R mutant of βB1 was constructed via site-directed mutagenesis. The WT and S93R genes were then inserted into the prokaryotic vector pET28a and the eukaryotic vector pEGFP-N1 for exogenous expression in E. coli Rosetta (DE3) cells and human cells, respectively, following previously described steps (Qi et al., 2016; Zhu et al., 2018). Briefly, the E. coli cells were first cultivated in Luria-Bertani medium at 37°C for 4 h and then treated with 0.01 mM isopropyl 1-thio-β-D-galactopyranoside (IPTG) at 16°C for 16 h to induce recombinant protein overexpression. Next, the target recombinant proteins were isolated from the supernatant of cell lysates using a Ni-NTA affinity column and further purified on the ÄKTA explorer through the HiLoad 16/600 Superdex 200 prep-grade column with 1 mM phenylmethylsulfonyl fluoride (PMSF). The purified proteins were dissolved in buffer A (20 mM Na₂HPO₄, 150 mM NaCl, and 1 mM EDTA, pH = 7.4) and concentrated before being stored at −80°C. The purity of the target proteins was verified using SDS-PAGE, which revealed that WT was more than 95% pure.

According to the SDS-PAGE analysis, the S93R proteins barely dissolved in the supernatant of cell lysates regardless of the cultivation temperatures and the concentrations of IPTG, indicating that the majority of the mutant proteins existed in inclusion bodies. Therefore, unlike the WT proteins, the S93R proteins were extracted from inclusion bodies in fully denatured status and then renatured using previously described methods (Qi et al., 2016). Briefly, the inclusion bodies were purified from the precipitation of cell lysates and then dissolved in buffer B (20 mmol/L Na₂HPO₄, 500 mmol/L NaCl and 6 mol/L GdnHCl, pH 7.0). The denatured S93R proteins were soluble and retrievable using Ni-NTA affinity and HiPrep 26/10 desalting columns in the same buffers as the WT proteins but with additional 6 mol/L GdnHCl. The purified proteins were concentrated to 1 ml and carefully diluted in buffer A at 0.1 ml/min until protein aggregation was observed. Next, the protein solution was incubated at 4°C for 16 h. The refolded S93R proteins were then found in the supernatant fraction after centrifugation. The refolded proteins were concentrated, and SDS-PAGE was used to ensure that the purity was greater than 95%.

The concentrated WT and the S93R proteins were diluted in buffer C (20 mM Na₂HPO₄, 150 mM NaCl, 0.8 mol/L GdnHCl and 1 mM EDTA, pH = 7.4) to the concentration of 10 μM for all samples measured using spectroscopy. Far-UV circular dichroism (CD) was measured using a Chirascan spectrophotometer. The path-length was 0.5 mm, and the range of scanning wavelength was from 200 to 250 nm. The intrinsic fluorescence and the extrinsic ANS fluorescence were measured using a F-4600 fluorescence spectrophotometer (Hitachi, Tokyo, Japan) using previously described methods and parameters (Yang et al., 2020; Fu et al., 2021; Liu et al., 2022). The excitation wavelength for intrinsic fluorescence was 280 nm for Trp and Tyr fluorescence and 295 nm for Trp fluorescence. The scanning wavelength of the intrinsic fluorescence ranged from 260 to 400 nm. For the temperature-gradient heating experiments, the protein samples were heated stepwise from 20 to 90°C at an interval of 2°C using the F-4600 fluorescence spectrophotometer. At each temperature, the samples were incubated for 2 min before Trp fluorescence was measured. The melting midpoint temperature (Tm) was calculated using the equation described previously (Zhu et al., 2018). Extrinsic ANS fluorescence was excited at 380 nm light, and the range of scanning wavelength was 400–700 nm. The 1-anilinonaphthalene-8-sulfonate (ANS) solution was added to protein samples at a 50:1 M ratio (ANS:protein) before measurement. A₄₀₀ values was obtained by measuring the absorbance of protein samples at 400 nm light using an Evolution 300 Security UV/Vis Spectrophotometer (Thermo Fisher).

The HEK 293T cells were cultivated in DMEM with 10% FBS, whereas the HLE-B3 cells were cultured in DMEM/F-12 (1:1) with 20% FBS. Both cell lines were incubated in 5% CO₂ at 37°C. The WT and S93R genes were exogenously overexpressed in both cell lines to establish the cataract cell models, and the protein distribution within the cells were visualized using immunofluorescence, as previously described (Xu et al., 2021). In brief, the WT and S93R-fused plasmids were transfected into the HEK 293T cells using Lipofectamine 2000 and the HLE-B3 cells were transfected using Lipofectamine 3000. After 5 h of transfection, the cells were cultured in fresh DMEM with 10% FBS or DMEM/F-12 (1:1) medium with 20% FBS at 37°C for 24 h, and fixed by 4% paraformaldehyde for 20 min. The cells were then washed three times with PBS buffer before being treated with 0.4% Triton X-100 for 10 min and blocked by 10% FBS for 50 min. The nuclei were identified using DAPI, the p62-detected aggregates were visualized using Alexa Fluor Plus 555 antibody, whereas the recombinant WT and S93R proteins were observed using the GFP tag. Immunofluorescence images were captured using a Leica DMi8 confocal microscope system, whereas a Fiji software was used to analyze the percentages of cells with aggregates and the area of protein aggregates in 10 random viewing fields that were from three independent experiments.

The HEK 293T cells were lysed in 1% NP40 solution after transfection and 24 h of cultivation. To perform western blot analysis, the total (T) protein solution was divided into a supernatant (S) fraction and a precipitated (P) fraction using high-speed centrifugation. The GFP-tagged WT and S93R proteins were detected using the GFP Tag Mouse antibody, and incubated at 4°C for 12 h before being treated with an anti-mouse secondary antibody. The Image Lab software (BioRad) was used to visualize and analyze the intensities of the protein bands. For SDS-PAGE, the WT and S93R proteins expressed in E. coli Rosetta cells were also divided into total (T), supernatant (S), and precipitated (P) samples. The gels were submerged in Coomassie blue dye for 30 min and rinsed with water to visualize the protein bands. The intensity of the protein bands was visualized using Image Lab software.

Cell viability of the transfected cells was determined using the Cell Counting Kit-8 (CCK-8) according to the manufacturer’s instructions. Briefly, the WT and S93R-fused plasmids were transfected into the HEK 293T cells seeded in 96-well culture plates. After 6, 24, 48, and 72 h of cultivation, 10 μl of the CCK-8 solution was added to each well. Then the cells were incubated at 37°C for 1 h, and the OD values were measured at wavelength 450 nm using a microplate reader (BioRad). For cell apoptosis assay, HEK 293T cells transfected with the WT and S93R genes were collected after 24 h of cultivation and stained with PI and Annexin V-Alexa Fluor 647. The percentages of apoptotic and necrotic cells were determined using flow cytometry on a CytoFLEX LX flow cytometer (Beckman Coulter).

The structure file of βB1-crystallin (PDB code: 1OKI) was downloaded from the Protein Data Bank, and the S93R variant was constructed by Pymol, which is a molecular graphics system (http://www.pymol.org/). All simulations were performed using GROMACS (Version. 2018.5.1.2) under CHARMM36 force field. The molecular dynamics (MD) simulations were carried out in a cubic water box with 150 mM NaCl. The minimal distance from the edges of the cubic box to the protein surface was set to 0.6 nm. After energy minimization, the system was equilibrated for 5 ns under NVT and NPT conditions at 310 K. The simulations were run at 2 fs time steps for 100 ns. Then, the values of backbone root mean square deviation (RMSD), backbone root mean square fluctuation (RMSF), and solvent accessible surface area (SASA) were analyzed. Finally, the Visual Molecular Dynamics V.1.9.3 (Phillips et al., 2005) was used to process and analyze the simulated trajectories. All structural figures were generated with Pymol or Chimera software.

The βB1-crystallin amino acid sequence alignment revealed that the 93rd serine residue remained highly conserved during the evolution of vertebrates (Figure 1A). Furthermore, previous studies of sequence alignment of the β/γ-crystallins reported that the S93 residue is conserved in the β/γ-crystallin superfamily in many species (Bloemendal et al., 2004; Jin et al., 2020). From the dimeric structure of the βB1-WT, S93 is located in the first Greek key motif and structurally close to the second Greek key motif (Figure 1B), implying that this site probably plays an important role in maintaining structural stability of βB1–crystallin.

To observe the behaviors and distribution of the WT and S93R proteins in both human and prokaryotic cells, the WT and S93R genes were exogenously expressed in two kinds of human cell lines and E. coli Rosetta (DE3) cells (Figure 2). In the human cells, the GFP-fused WT protein dispersed uniformly in the cytosol and nucleus, which was consistent with the distribution pattern of the control group transfected with an empty pEGFP-N1 vector. However, the S93R mutant formed many intracellular protein aggregates that colocalized with p62, a classic aggresome marker protein (Figures 2A,B). A quantitative analysis of protein aggregation showed that the percentages of cells with protein aggregates differed significantly between the WT and S93R groups (Figure 2C). The proteins overexpressed in HEK 293T cells were then prepared for a western blot analysis. The western blot and the densitometric analyses showed that the WT protein mainly existed in the supernatant, while a considerable amount of S93R protein appeared in the precipitate (Figure 2D). For E. coli Rosetta (DE3) cells, the results of SDS-PAGE and densitometric analysis suggested that the amount of WT proteins in the supernatant fraction increased when the cultivation temperature or IPTG concentration decreased. The optimal culture condition for obtaining soluble WT proteins was at 16°C with 0.01 mM IPTG. However, the S93R mutated proteins were predominant in the precipitate fraction regardless of the culture conditions, implying that the mutated proteins mainly accumulated in inclusion bodies and could hardly be obtained by purification procedures of soluble proteins (Figure 2E).

The WT proteins were successfully purified from the supernatant fraction of E. coli cell lysates, however the S93R proteins were difficult to extract using the same procedures as the WT proteins. Therefore, the S93R proteins were completely denatured using 6 mol/L GdnHCl, purified from the inclusion bodies of E. coli cells and then renatured by dilution. To determine if the production protocols bias results, the obtained WT proteins were denatured and renatured following the same steps as the S93R proteins, and the intrinsic Trp fluorescence and far-UV CD showed that there were no major differences between the basic biophysical properties of WT proteins purified in native condition and in denaturant condition (Supplementary Figure S1). The SDS-PAGE analysis on the obtained protein samples revealed that the purities of both proteins were greater than 95% (Figure 3A). Furthermore, the obtained WT and S93R protein samples were compared to 6 mol/L GdnHCl treated groups using far-UV CD and intrinsic Trp & Tyr fluorescence. The results implied that the refolded S93R proteins were partially renatured as compared to the fully denatured status (Figures 3B,C). For intrinsic fluorescence, the E_max (maximum emission wavelength) of the refolded S93R proteins was comparable to that of the WT in both Trp & Tyr fluorescence and Trp fluorescence with a slight red-shift, as previously reported (Qi et al., 2016). However, S93R had significantly higher RLS (Rayleigh light scattering) intensity than the WT, suggesting that the refolded S93R formed small protein aggresomes and had a strong aggregation potential (Figures 3D,E). For extrinsic ANS fluorescence, S93R had a slight increase in ANS fluorescence in comparison to the WT, implying that the S93R proteins had higher hydrophobic residue exposure (Figure 3F).

As indicated by previous work, protein stability and aggregation propensity are highly related (Booth et al., 1997). In this study, the results of temperature-gradient heating experiments showed that the WT remained in a native status as the temperature rose and started to denature at around 52°C, whereas the denaturation of S93R occurred at around 28°C. We also analyzed the T_m value of βB1-WT and S93R mutant, which was 60 and 32°C, respectively, (Supplementary Figure S2). Because the refolded S93R had an aggregation tendency, the protein samples were immediately subjected to different treatments and measurements after purification in order to assess the potential effects of S93R mutation on the βB1-crystallin (Figure 4). According to the intrinsic and extrinsic ANS fluorescence spectra, the WT had a slight increase on RLS intensity after incubation at 37°C, whereas the S93R mutant showed much higher RLS and ANS increase, indicating severe protein aggregation and hydrophobic residue exposure, as well as a slight red-shift of E_max. (Figures 4A,B). In terms of turbidity, the results suggested that the WT was stable at physiological temperature. Nevertheless, the S93R aggregated rapidly after only 0.5 h of incubation at 37°C, and the turbidity increase steadily as the incubation duration prolonged, implying that the S93R was extremely unstable at physiological temperatures (Figure 4C). The fluorescence spectra results revealed that UV treatment had little effect on the WT, with just a slight increase of RLS intensity after 90 min of UV irradiation. Compared with the WT, the S93R exhibited higher RLS and ANS intensity, as well as a red-shift of E_max, indicating that the S93R was more sensitive to UV irradiation and prone to aggregation (Figures 4D,E). The turbidities of the WT and S93R were consistent with the results of fluorescence spectra (Figure 4F). After being incubated with H₂O₂ for 24 h, both WT and S93R showed increase in RLS intensity, ANS intensity, and turbidity when compared to the control. However, the S93R had a relatively greater increase than the WT, indicating that the S93R mutation reduced the resistance of βB1–crystallin to oxidative stress (Figures 4G–I).

To further mimic the occurrence of cataract, the disease cell models were subjected to various environmental stresses and visualized using immunofluorescence (Figure 5). The HEK 293T cells transfected with WT and S93R-fused plasmids were cultivated for 24 h before and 2 h after treatments to let protein aggregates form. There was no significant change in the distribution pattern of the GFP-fused WT protein after heat shock, UV irradiation, or oxidative stress, except for the appearance of some tiny protein aggregates within a few cells after H₂O₂ treatment. However, the S93R mutant group exhibited more intracellular p62-positive aggregates after all treatments, especially under oxidative stress (Figures 5A–C). The quantitative analysis revealed an increase in the percentages of cells with typical S93R aggregates after treatments, whereas there was no significant difference in the WT, which was consistent with the confocal results (Figure 5E). Furthermore, the distribution of protein aggregation area of the S93R suggested that the protein mutants could form larger aggregates under stressful conditions (Figure 5F). According to the western blot and the densitometric analyses, the proportion of WT proteins in the precipitate increased slightly after the treatments, but the supernatant fraction remained dominant. However, when compared to the result with no treatment, the S93R proteins showed a decrease in the supernatant fraction and an increase in the precipitate fraction. The western blot results further proved that the S93R mutant was more sensitive to heat shock, UV irradiation, and oxidative stress, all of which promoted protein aggregation in HEK 293T cells (Figure 5D).

The effects of exogenously expressed S93R proteins on the survival and growth of HEK 293T cells were studied by performing flow cytometry and CCK-8 assay. Because the cells were relatively more sensitive to oxidative stress based on the confocal results and quantitative analysis, the H₂O₂-treated cells were also subjected to flow cytometry (Figure 6). When compared to cells expressing WT proteins, cells expressing S93R proteins had higher percentages of apoptotic and necrotic cells, as well as significantly greater increase in dead cells after H₂O₂ treatment, indicating that intracellular S93R aggregates induced cell death (Figures 6A,B). According to the cell viability results, the proliferation of cells expressing the S93R proteins was significantly inhibited after 72 h of cultivation (Figure 6C).

MD simulations were applied to monitor the effects of S93R mutation on the structural stability (chain A and chain B) of βB1-crystallin. From the alignment of the dimeric structure of the WT and S93R mutant, we found that S93R mutation impaired some secondary structures of βB1-crystallin. The α-helixes (A156-F158, W174-Y176, and W215-W218) of βB1-chain A were transformed into loop structures after the S93R mutation. The α-helixes (E67-F69, S106-F108, and A156-F158) and β-strands (R73-S77, and T162-Q166) of βB1-chain B were also transformed into loop structures after the S93R mutation (Figure 7F). In addition, the S93 in the WT formed four hydrogen bonds with F64, L66 and F69 in chain B, and five hydrogen bonds with F64, N68, F69 and P122 in chain A, resulting in strong hydrogen bond networks that stabilized local β-sheet and loop structures (Figure 7A). Notably, P122 is located in the second Greek key motif of βB1-crystallin (Figures 7A,F), suggesting that the interaction between S93 and P122 might play an important role in maintaining structural stability of the N-terminal domain. However, the S93R mutant only interacted with F64 through two hydrogen bonds in both chain A and chain B (Figure 7A). Loss of the stable α-helixes, β-strands and part of the hydrogen bonds in the Greek-key motifs caused severe disruptions to local secondary structures. Furthermore, the RMSD of the S93R mutant was much higher than that of the WT from 80 ns (Figure 7B), implying that the mutation affected the protein structural flexibility. The S93R had higher RMSD, RMSF and SASA between residue 65 to 77 and 190 to 210 (Figures 7C–E). More interestingly, the S93R mutant had more hydrophobic residues at the surfaces around the mutation site and the dimer linkage site (Figure 7G). Unstable structural conformation and more exposed hydrophobic surfaces might contribute to the aggregation propensity of the S93R mutant.