HK-2 (ATCC® CRL-2190TM) cells are maintained in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum (FBS), 100 units/ml penicillin, and 100 μg/ml streptomycin at 37°C in a humidified incubator with 5% CO₂. The cell culture medium was refreshed every other day. All culture reagents were purchased from Invitrogen Technologies (Carlsbad, CA, United States).

Agilent 7500 ICP-MS was used to study the cellular metal content. All cells were washed with PBS, digested with trypsin, and collected by centrifugation at 1,000 rpm for 5 min. The cells were dissolved in 0.2 ml of 50% HNO₃ and 50% H₂O₂ (volume ratio 1:1), incubated at 60°C for 1 h, and diluted to an appropriate concentration. Metal quantification was repeated three times.

HK-2 cells were incubated with 500 µM bismuth citrate in a cystine-saturated medium for 48 h. The cells were harvested, lysed in 1% SDS, and centrifuged at 15,000 rpm for 15 min. The pellet was extracted with a two-phase system of 1:2 octanol and water. After 24 h, the particles in the bottom phase of the tube were collected and re-extracted with chloroform. The upper phase was carefully transferred to a new tube and centrifuged (×10,000 g, 30 min). The particles were sequentially washed with ethanol and distilled water and sonicated for 1 h. The morphology and composition of the particles were, respectively, determined by transmission electron microscopy (TEM) and energy-dispersive X-ray spectroscopy (EDS).

The medium was aspirated, and the cells were washed with ice-cold PBS three times. The cells were scraped and collected in 1.5-ml vials. The cells were pelleted by centrifugation for 5 min at ×800 g at 4°C. The cell pellets were snap-frozen in liquid nitrogen and stored overnight at −80°C until further analysis. The pellets were thawed, and the cells were lysed with buffer (Beyotime) 50 mM Tris (pH 7.4), 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, and 0.1% SDS, and supplemented with a complete protease inhibitor mixture (Beyotime). The cells were plated on the ice for 30 min, and sonicated for 15 cycles of 30 s. The total protein concentration was determined using a BCA assay (Thermo). Then, the cells were heated for 5 min at 95°C. Proteins were desalted with acetone and reduced and alkylated with 7M DTT and 1M IAA, respectively. The proteins were digested overnight at 37°C with trypsin (Promega) and with an enzyme-to-substrate ratio of 1:50. The samples were desalted using Sep-Pak C18 cartridges (Waters), eluted with 60% acetonitrile (ACN)/0.1% formic acid (FA), and directly subjected to peptide enrichment. The total peptide concentration was determined using the Pierce™ Quantitative Colorimetric Peptide Assay (Thermo). The samples for the proteome analysis were instead dried down and stored at −80°C until subjected to nLC-MS analysis.

Peptides were subjected to reverse phase nLC-MS/MS analysis using an Easy nLC1200 system coupled to a Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) for proteome analysis. The chromatograph system was equipped with a trap column (Acclaim PepMapTM 100, 75 μm × 2 cm, C18, 3 μm, 100 Å) and an analytical column (Acclaim PepMapTM RSLC, 75 μm × 25 cm, C18, 2 μm, 100 Å). Trapping was performed for 3 min in solvent A (0.1% FA in water) at 2 μl min−1, while for the elution, the flow rate was passively split to 300 nl min−1. The gradient was as follows: 0–1 min, 8% buffer B (0.1% FA in 80% ACN); 1–98 min, 8–28% buffer B; 98–112 min, 28–36% buffer B; 112–116 min, 36%–100% buffer B; and 116–120 min, 100% buffer B. The mass spectrometer was operated via Xcalibur (version 4.1, Thermo Scientific) in the data-dependent mode. The Orbitrap full-scan MS spectra from m/z 350–1800 were acquired at a resolution of 60,000 after accumulation to a target value of 3e6. The maximum ion injection time was 20 ms. Up to 20 most intense precursor ions were selected for fragmentation, with the isolation window set to 1.6 m/z. HCD fragmentation was performed at a normalized collision energy of 28% after accumulation to a target value of 2e5, with the maximum injection time of 25 ms. MS/MS was acquired at a resolution of 15,000. Dynamic exclusion was set to 60 s.

Proteome Discoverer 2.1 software (Thermo Fisher Scientific) was applied for identification of peptide and protein. The latest Human UniProtKB/SwissProt Release (2018_10 has 551,681 entries) was used for searching MS/MS spectra. Peak lists were generated with a precursor signal-to-noise ratio of 1.5, and the Sequest algorithm was set as default. Trypsin was set as the specific enzyme, and one missed cleavage was tolerated. The precursor ion mass tolerance was set to 7 ppm, and the product ion mass tolerance was set to 0.06 Da. The peptide false discovery rate (FDR) was determined by performing a decoy database search with the target decoy PSM (Peptide Map Matching) validator module. The peptide FDR threshold is 1%.

The survival rate of the cell was measured using Cell Counting Kit-8 (Dojindo). HK-2 cells were cultured in a 96-well plate. CCK-8 solution was added to the cell culture medium at a 1: 10 ratio and incubated for 2 h at 37°C. The absorbance at wavelength 450 nm was measured using an automatic detector (Biotek Synergy 2).

5,5-dithiobis (2-nitrobenzoic acid) (DTNB)–based Glutathione Assay Kit (Beyotime, S0052) and Glutathione Reductase Assay Kit (Beyotime, S0055) were used. The Glutathione S-Transferase (GST) Assay Kit was purchased from Sigma (CS0410), and details of all assays can be found in the manufacturer’s protocol.

The data are expressed as mean ± standard error of mean. Each experiment was repeated at least three times. The statistical significance was determined using Student’s t-test. Statistical significance is p < 0.05.

To investigate bismuth’s effect on proximal tubular cells, we studied the proteomic alterations it induces in the proximal tubular cell line HK-2. HK-2 cells were treated with or without 500 μM CBS for 48 h, and the cells were collected, processed, and analyzed via LC-MS/MS. A total of 6,230 proteins containing at least one unique peptide were identified with confidence values 99% of FDR estimation. Among all proteins identified, 312 were differentially expressed proteins (DEPs, fold of change >2 or <0.5, p value <0.05), including 138 up and 174 downregulated proteins (Figure 1A).

The DEPs were categorized according to the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) databases, while the enrichment analysis was performed. The KEGG result shows that in HK-2 cells, CBS affects the metabolism of a series of widely different metabolites: glutathione, cysteine and methionine, amino acids, fatty acids, sugars, folate, and purine/pyrimidine (Figure 1B). The GO result shows that among all cellular components, actin cytoskeleton and vesicle are the most affected (Figure 1C). As the protein–protein interaction network and volcano plot have showed, the expression of the actin cytoskeleton was greatly increased upon CBS treatment (Figures 1D,E). Rho GTPases are central regulators of actin reorganization and its subsequent functions (Heasman and Ridley, 2008). The biological process of Rho protein signal transduction was significantly affected by CBS, suggesting an upstream mechanism behind the changes of the actin filament (Figure 1C, Supplementary Figure S1). In contrast to actin, the expression level of the proteins in vesicles was greatly reduced, especially those from the secretory granule (Figures 1D,E). The key proteins of glycolysis were all downregulated by CBS, which includes GAPDH, ENO1, ENO2, PGK1, PGAM1, and TPI1 (Figure 1D). The upregulation of actin and the downregulation of the vesicles and glycolysis were also validated by the enrichment analysis separately using upregulated and downregulated genes (Supplementary Figures S2A,B). The total lists of DEP enrichment analysis results of CBS-treaded HK-2 can be found in Supplementary Data S1.

We then performed GSEA analysis on proteomic data among control and CBS-treated HK-2 cells, in perspective of KEGG (Supplementary Figure S3A). It also shows that CBS downregulated the metabolism of multiple metabolites, including amino acids, fatty acids, folic acid, pyruvate, and glycolysis (Supplementary Figure S3B). In HK-2 cells pretreated with L-buthionine-sulfoximine (L-BSO), GSH was pre-depleted, and such alterations to the metabolism did not occur, suggesting that it is a consequence of the activation of massive GSH biosynthesis (Supplementary Figure S3A). CBS enhanced Notch signaling in HK-2 cells with or without GSH depletion (Supplementary Figures S3A,B). The Notch heterodimer is exposed on the cell surface, which has cysteine-rich Lin Notch repeats (LNR) that could be a potential target of a CBS attack (Weng et al., 2004). Suppression of Notch signaling was found to induce actin accumulation, suggesting another possible upstream mechanism of the increase of actin upon CBS treatment (Wesley et al., 2011). Among the metabolic pathways identified, glutathione metabolism, cysteine, and methionine metabolism and glycolysis have the maximum negative enrichment score (ES), and porphyrin and chlorophyll metabolism has the maximum positive ES (Supplementary Figure S3C). We also performed GSEA analysis in perspective of GOCC, downregulations were found in the ribosome, vesicular lumen, and ER, and upregulation was mainly found in filamentous actin and vesicle tethering complex (Supplementary Figure S3D). The ES of the aforementioned GOCC terms in GSH-depleted HK-2 cells was not as outstanding (Supplementary Figure S3E). The GSEA results of the ribosome, vesicle tethering complex, and actin filament are shown (Supplementary Figure S3F). GSEA results further substantiate the findings in the DEP enrichment analysis, and all GSEA leading-edge proteins in the pathways referred can be found in Supplementary Data S2.

To further investigate the bismuth effect on the proteome of HK-2 in the absence of GSH, HK-2 cells were incubated with 500 μM L-BSO for 24 h before being treated with 500 μM CBS for 24 h. The cells were collected, processed, and analyzed via LC-MS/MS. Among all proteins identified, 393 were differentially expressed (fold of change >2 or <0.5, p value <0.05), including 189 up and 104 downregulated proteins (Figure 2A).

To gain insight into the biological functions affected by bismuth in GSH-depleted HK-2 cells, the DEPs were categorized according to Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) databases and enrichment analysis was performed. Unlike in HK-2 cells without GSH depletion, bismuth enhanced ubiquitin-mediated proteolysis, suggesting that without GSH, proteins are vulnerable upon bismuth treatment (Figure 2B, Supplementary Figure S2C). Meanwhile, without GSH sequestration, bismuth also greatly affected the chromosome and mitochondrion. The KEGG enrichment analysis shows that bismuth selectively enhanced mineral absorption in GSH-depleted HK-2 cells (Supplementary Figure S2C). Examples of proteins of mineral absorption include MT1E, MT1F, MT2A, ATOX1, and HMOX1, which proves that GSH plays an important protective role in bismuth-treated HK-2 cells before MT.

Furthermore, the DEPs were involved in biological processes including cytoskeleton organization (Figure 2C). The protein–protein interaction network showed a strong interaction network between the set of upregulated proteins which are mainly actin cytoskeleton and microtubule cytoskeleton, and the downregulated DEPs are less well-connected to each other and did not form clique in the graph (Figure 2D). The proteins belonging to GO terms actin cytoskeleton and the ubiquitin protein ligase activity are showed as volcano plots (Figure 2E). The total lists of the DEP enrichment analysis results of CBS-treated HK-2 can be found in Supplementary Data S3.

We also perform GSEA analysis on proteomic data among GSH-depleted HK-2 cells and CBS-treated GSH-depleted HK-2 cells (Supplementary Figure S4). Without GSH sequestration, in perspective of KEGG pathways, bismuth causes severe protein and DNA damage, increased ubiquitin-mediated proteolysis, DNA mismatch repair, and a cytosolic DNA-sensing pathway (Supplementary Figures S4A–C). In perspective of GOCC, bismuth increased the presence of the inclusion body but severely suppressed the levels of proteins from ER and Golgi lumen (Supplementary Figures S4D–F). GSEA results further substantiate the findings in the DEP enrichment analysis (Supplementary Figure S2), and all GSEA leading-edge proteins in the pathways referred can be found in Supplementary Data S4.

By comparing the proteomes of control and CBS-treated HK-2 cells, we found alterations mainly toward the major metabolic pathways, including the metabolism of GSH. In GSH-depleted HK-2 cells, the cells underwent severe cytotoxicity but not alteration to major metabolisms. For further investigation, we performed enrichment analysis for HK-2 cells without and with GSH depletion, in the perspective of protein domain (Interpro knowledgebase was used) (Figure 3A). Indeed, Interpro term Glutathione S-transferase, N-terminal, and Metallothionein, vertebrate, and metal binding sites were found outstanding in CBS-treated HK-2 cells without and with GSH-depletion. Proteins GSTZ1, GSTP1, GSS, IDH1, and GLO1 are differentially expressed in normal HK-2 cells but not in HK-2 cells with the L-BSO inhibition of GSH biosynthesis (Figure 3B). GSEA analysis also showed that CBS induced expressional differences in proteins of glutathione metabolism only in HK-2 cells with GSH (Figure 3C). In GSH-depleted HK-2 cells, not HK-2 cells with GSH, the expression levels of MT isoforms MT2A, MT1E, and MT1F were greatly increased (Figure 3D). Cystine is the rare amino acid in the biosynthesis of both GSH and MT; we can see that CBS induced the expression of cystine importer SLC7A11 only when GSH biosynthesis is not suppressed by L-BSO. The metabolism of sulfur-containing amino acids cysteine and methionine was also altered only in normal HK-2 cells (Figure 3E).

In order to study the protective effect of bismuth on cisplatin, we first studied the dose-dependent effect of cisplatin on the human proximal tubular cell line HK-2. A dose–response curve of cisplatin in HK-2 cells was established, and the LD50 is 11.63 μM (Figure 4A). We used 20 μM cisplatin (LD80) in subsequent experiments to evaluate bismuth-induced cytotoxicity mitigation (Figure 4B). HK-2 cells were pretreated with 500 μM colloidal bismuth subcitrate (CBS) for different periods of time. HK-2 cells pretreated with CBS for 36 h or 48 h showed the best protection against cisplatin cytotoxicity, and the cell survival rate increased from 20% to about 50%. The morphology of HK-2 cells without or with CBS pretreatment before cisplatin insult was also imaged, and compared to CBS-pretreated HK-2, the cisplatin-treated cells showed more severe apoptosis (Figure 4C). Glutathione is critical for bismuth uptake and detoxification (Hong et al., 2015). We found that GSH was significantly increased by 2.5 times in HK-2 cells after 48 h of CBS treatment (Figure 4D). The glutathione-S-transferase (GST) and glutathione reductase (GR) activity of HK-2 cells treated with CBS with different periods of time was also increased (Figure 4E), which is consistent with the change in the GSH level (Hong et al., 2015). These data suggest a strong correlation between GSH and bismuth-induced protection against cisplatin toxicity.

In a previous study, we found that bismuth was conjugated to glutathione and then transported into vesicles by an MRP transporter. Since CBS induces the expression of cystine importer SLC7A11 in HK-2 cells (Figure 3D), we further supplied the cells with 0.1 mg/ml of cystine in the culture medium, which dramatically increased the absorption of bismuth (Figure 5A). Furthermore, by inhibition of the MRP transporters (which transport metal-GSH conjugates) using MK571, the absorption of bismuth was greatly reduced. Black particles are synthesized within the vesicles of HK-2 cells after being incubated with CBS. These particles are visualized by the staining of the cell nucleus with Acridine Orange (AO) and imaging with fluorescent microscopy (Figure 5B). We isolate these particles by dissolving away the soluble part of the cell with 1% SDS. The black particles were collected by 15,000 rpm centrifugation for 15 min. The morphology of the particle was determined by transmission electron microscopy (TEM) (Figure 5C). The mean diameter of the particles is approximately 20 nm. We then studied element composition of particles using energy-dispersive X-ray spectrometry (EDS) (Figure 5D). According to EDS result, the atom ratio between bismuth and sulfur is close to 1:1.7 (Table 1). To further investigate whether bismuth pretreatment influences cisplatin absorption and sequestration, HK-2 cells were treated with 500 μM bismuth for 48 h prior to 20 μM cisplatin cells for 24 h. We then quantified the amount of bismuth and platinum in the whole cell, the SDS-dissolved part of the cell, and the particles. The total amount of bismuth absorbed is 150 ng/10⁶ cells (Figure 5E). The platinum content in the cells pretreated with bismuth is around two folds higher than that of the cells treated only with cisplatin. We also compared the amount of bismuth and platinum in the particles and the SDS-dissolved part of cells (supernatant) (Figure 5F). The amount of bismuth and platinum in the particles are both around two times higher than that in the supernatant. These data indicated that both bismuth and platinum are transported to the vesicles and incorporated into the black particles.

The timing of the upregulation of intracellular GSH after CBS treatment concurs that of the reduction of cisplatin toxicity in HK-2 cells and CBS-enhanced cisplatin sequestration to the vesicles, suggesting that GSH plays the major role (compared to MT) against cisplatin toxicity. However, whether the intracellular scavenging of reactive oxygen species or cisplatin by GSH is the main mechanism behind the protective effect of CBS remains unknown. For further investigation, we systematically examined the protective effect induced by CBS on HK-2 cells against a series of soft metal compounds (including anticancer drugs such as carboplatin and arsenic trioxide and common heavy metal pollutants such as cadmium and mercury), hard metal compounds, and oxidant. HK-2 cells were treated with 500 μM CBS for 48 h before being exposed to different compounds. We found that CBS pretreatment only reduced the toxicity of soft metal compounds: carboplatin, arsenic trioxide (As₂O₃), arsenic pentoxide (As₂O₅), antimony citrate (SbCit), cadmium chloride (CdCl₂), and mercury chloride (HgCl₂) (Figures 6A–F). CBS shows no protection against either oxidant hydrogen peroxide (H₂O₂) and sodium chromate (Na₂CrO₄) or hard metal compounds: Na₂CrO₄ and titanocene dichloride (Figures 6G–I). It shows that the CBS-induced protection is general against different heavy metal compounds but not the hard metal compounds and oxidants. It suggests that CBS-induced GSH upregulation mitigates cisplatin toxicity through direct conjugation, instead of reducing cisplatin-induced oxidative stress.