Mitochondrial HSP90 Paralog TRAP1 Deletion Drives Glutamine Addiction in Tumor Cells via Destablization of the Cys/Glu Antiporter SLC7A11/xCT

Abstract

TRAP1, the mitochondrial isoform of HSP90, has emerged as a key regulator of cancer cell metabolism, yet the mechanisms by which it rewires nutrient utilization remain poorly understood. We previously reported that TRAP1 loss increases glutamine (Gln) dependency of mitochondrial respiration following glucose (Glc) withdrawal. In this study, we investigate how TRAP1 deletion impacts Glc metabolism and the mechanisms enabling Gln retention to support mitochondrial respiration via reductive carboxylation and the oxidative TCA cycle. TRAP1 knockout (KO) in bladder and prostate cancer cells recapitulates the carbon source–specific metabolic rewiring previously observed. Stable isotope tracing reveals that although Glc oxidation remains functional, TRAP1 KO reduces overall Glc uptake and its contribution to glycolysis and the pentose phosphate pathway. This effect is consistent across multiple cell lines. Concurrently, TRAP1-deficient cells exhibit increased Gln retention and reliance, potentially due to downregulation of the cystine/glutamate antiporter SLC7A11/xCT. Supporting this, xCT overexpression reduces Gln-dependent respiration in TRAP1 KO cells. qPCR and proteasome inhibition assays suggest that xCT is regulated posttranslationally via protein stability. Notably, xCT suppression does not trigger ferroptosis, indicating a selective adaptation rather than induction of cell death. Together, our findings suggest that TRAP1 loss decreases Glc uptake while preserving its metabolic fate, promoting Gln conservation through xCT downregulation to maintain mitochondrial respiration without inducing ferroptosis.

Implications:

These results reveal a TRAP1-dependent mechanism of metabolic rewiring in cancer cells and identify xCT-mediated Gln conservation as a key adaptive response, underscoring TRAP1 as a potential metabolic vulnerability and therapeutic target in tumors with altered nutrient utilization.

Introduction

TRAP1, a paralog of the molecular chaperone HSP90 primarily localized in the mitochondrial matrix (1), was described almost a decade ago to promote aerobic glycolysis by altering central carbon metabolism (CCM; refs. 2, 3). Whereas significant efforts have focused on TRAP1’s role in the inhibition or activation of major electron transport chain components (4–7), no study has explored the rewiring of major carbon sources that drive the CCM in the presence or absence of this protein. This information is particularly important because cancer cells are known to rewire metabolism (8, 9) by specifically switching carbon source dependence between glucose (Glc) and glutamine (Gln), both central pillars of the CCM (10, 11), with a particular emphasis on addiction to Gln (12–14). Given that presence or absence of the TRAP1 protein alters metabolic preference (3, 7, 15), it is likely to impact specific utilization of these two carbon sources, thereby making it an attractive protein from the perspective of cancer therapy. In fact, TRAP1 has been implicated as a drug target in oncology (16–19) because of its ability to rewire the CCM and force aerobic glycolysis (7, 20). Multiple studies over the years have focused on its targeted inhibition (21–23), even though its role as either an oncogene or tumor suppressor is still not clearly understood (19, 20, 24). To this end, we recently reported that deletion of TRAP1 led to an inability of Glc to support the tricarboxylic acid cycle (TCA) cycle while inducing a concomitant increase in Gln anaplerosis to support oxidative phosphorylation (OxPhos; ref. 15).

In this study, we further extend our previous findings by tracing the path of uniformly labeled ¹³C-Glc and ¹³C-Gln in TRAP1-null cells to investigate a possible mechanism of retention of Gln in the absence of TRAP1. We first determined whether Glc is directed into the pentose phosphate pathway (PPP) instead of the TCA cycle to balance the oxidative load generated in TRAP1-null cells (25–28). Next, we queried whether Gln is directed into both the oxidative arm of the TCA cycle and reductive carboxylation to generate pools of TCA cycle intermediates and sustain an impaired mitochondrial metabolism (29). Finally, we explored the role of SLC7A11 (alternatively called xCT or system Xc⁻), a sodium-independent cystine/glutamate (Cys/Glu) antiporter of particular interest in cell survival under Glc withdrawal (30–33), as one of the possible mechanisms by which TRAP1-deficient cells preserve Gln for its use as an anaplerotic substrate to sustain the TCA cycle.

Materials and Methods

Cell culture

All cell lines (UMUC3, PC3, HEK293T, HCT116, and A549) were obtained from the ATCC between 2017 and 2022 and authenticated by the ATCC using short tandem repeat profiling. The biological sex of the cell lines is as follows: UMUC3 (male), PC3 (male), HEK293T (female), HCT116 (male), and A549 (male). Mycoplasma contamination was routinely monitored using MycoStrip Mycoplasma Detection Kit (InvivoGen), with the most recent test performed in December 2024. All cells were used within 15 passages from thawing. Unless specified, all cells were cultured at 37°C with 5% CO₂ in a standard incubator with DMEM containing GlutaMAX, 4.5 g/L Glc, and 1 mmol/L pyruvate (Pyr; Thermo Fisher Scientific) supplemented with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin.

TRAP1 knockout generation

TRAP1 knockout (KO) PC3 and UMUC3 cells were generated using the GeneCopoeia CRISPR-Cas9 Gene Engineering platform as illustrated in Supplementary Fig. S1A. Cells were transiently transfected with an all-in-one vector, pCRISPR-CG12, harboring a mCherry reporter, against three target site sequences (Supplementary Fig. S1A) using Lipofectamine 3000 (Thermo Fisher Scientific) according to the manufacturer’s protocol. Twenty-four hours after transfections, cells were resuspended in flow cytometry sorting medium [5% FBS with 1× HyClone antibiotic/antimycotic solution (Thermo Fisher Scientific)] and transferred to a 35-μm mesh tube (Thermo Fisher Scientific) to be sorted using a BD FACSAria II flow cytometer under aseptic conditions. mCherry-positive cells were collected in 96-well plates, with one cell per well, for clonal selection. The single cells were allowed to grow for 2 weeks and then picked, subcultured, and finally analyzed by Western blotting with antibodies against TRAP1 to identify clones not expressing the protein. No obvious growth defects or reduction in viability were detected in PC3 and UMUC3 TRAP1 KOs. Unless specified, each parental wild-type (WT) was compared with a single KO clone in various experiments. As mentioned earlier, all cell lines used for experiments were routinely tested for Mycoplasma using Mycostrip Mycoplasma Detection Kit (InvivoGen).

Cell culture for oxygen consumption rate experiments

Before performing single carbon source oxygen consumption rate (OCR) experiments, the cells were grown overnight in a medium with the carbon source to be tested in order to acclimatize and stabilize them metabolically. Carbon sources were then added to DMEM lacking Glc, Pyr, and Gln (A14430-01) with 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin as follows: (i) Glc only, 4.5 g/L Glc and (ii) Gln only, 2 mmol/L Gln.

Real-time metabolic assays

Mitochondrial OCR was monitored in real-time using a Seahorse XF analyzer (XF^e96, Agilent). Depending on the experiment, 4 × 10⁴ HEK293T, 1.5 × 10⁴ UMUC3, or 1.5 × 10⁴ PC3 cells were cultured overnight in custom XF96 microplates (coated with 0.01% poly-L-lysine) with either DMEM GlutaMAX or DMEM-A14430-01 supplemented with Glc or Gln. The standard assay medium used for all mitochondrial stress tests was buffered XF DMEM (pH7.4; Agilent, 103575-100) without Glc, L-Gln, sodium Pyr, sodium bicarbonate, phenol red, and FBS. Depending on the experiment, the XF DMEM medium was supplemented with the desired carbon source as indicated above. Prior to experiments, the cells were washed with and then incubated in XF DMEM medium containing the respective carbon source in the absence of CO₂ for 1 hour to acclimatize them to the assay medium. After preincubation, basal OCR was determined before recording mitochondrial stress test profiles by sequential injections of oligomycin, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone, and rotenone/antimycin A in combination. In some experiments, described in Results, the ferroptosis inhibitor erastin was injected once a stable baseline OCR was achieved. All OCR profiles presented were normalized to 50,000 cells for each cell type.

For all assays involving transfections, 2 × 10⁵ cells were first seeded in six-well plates and allowed to grow overnight in DMEM GlutaMAX. They were transfected on day 2 with 2 μg DNA using the JetPrime transfection reagent (Polyplus) for 2.5 hours and further incubated overnight in DMEM GlutaMAX. For Gln-only analysis, on day 3, 4 × 10⁴ transfected cells were seeded in poly-L-lysine–coated XF96 microplates and incubated in in DMEM (A14430-01) supplemented with Gln overnight.

For overnight treatments with erastin, 4 × 10⁴ HEK293T cells were cultured with DMEM (A14430-01) supplemented with Gln and 5 or 10 μmol/L erastin. The XF DMEM medium used for washes, and preincubation and analysis were premixed with erastin for continuity.

Total cell reactive oxygen species

Total intracellular reactive oxygen species (ROS) levels in A549, UMUC3, and PC3 cells were measured by using the fluorescent dye H₂DCFDA (Thermo Fisher Scientific) according to the manufacturer’s protocol. ROS levels were measured based on the fluorescence intensity of H₂DCFDA using a BD FACSCanto II and its software BD FACSDiva. Final data analysis and mean fluorescence intensity (MFI) were calculated using FlowJo software (Tree Star).

Total metabolite and flux analysis using uniformly labeled ¹³C-Glc and ¹³C-Gln

Metabolic flux analysis using uniformly ¹³C-labeled Glc (U-¹³C-Glc) was performed by Human Metabolome Technologies (HMT), Inc. (https://humanmetabolome.com/en/targeted.html). Three biological replicates each of HCT116 and HEK293T cells were used for this experiment. The cells were first grown in standard growth medium containing unlabeled Glc, Pyr, and Gln until confluency reached 90%. Then the growth medium was replaced with Glc-deficient media (containing Pyr and Gln only) for 2 hours. After Glc starvation, HEK293T and HCT116 cells were incubated in media containing unlabeled Gln and Pyr and U-¹³C-Glc for 6 and 12 hours respectively. The labeling time was chosen based on the ability of the cells to incorporate Glc using DMEM supplemented with 150 μg/mL 2-NBDG, a nonhydrolyzable fluorescent analog of Glc. Samples were prepared according to guidelines issued by the service provider (5 × 10⁶ cells/ replicate) and resuspended in 50 μL ultrapure water before measurements were taken. The samples were analyzed by capillary electrophoresis time-of-flight mass spectrometry (Agilent Technologies) in two modes to detect both anionic and cationic metabolites (34–36). Detected peaks were then extracted using MasterHands ver. 2.17.1.11 to obtain m/z, migration time, and peak area. Putative metabolites were assigned based on HMT’s target library and their isotopic ions on the basis of m/z and migration time.

¹³C-Gln tracing was also performed by HMT. Two biological replicates of UMUC3 and HCT116 cells were used for this experiment. The cells were first grown in standard growth medium containing unlabeled Glc, Pyr, and Gln until confluency reached 90%. Then the growth medium was replaced with Gln-deficient media (containing Pyr and Glc only) for 6 hours. After Gln starvation, UMUC3 and HCT116 cells were incubated in media containing unlabeled Glc and Pyr and uniformly ¹³C-labeled Gln (U-¹³C-Gln) for 12 hours. Samples were then prepared according to the service provider’s guidelines as stated above.

Glc consumption

Glc depletion from culture medium was measured by collecting medium from 12-well plates incubated for 24 hours either with or without cells. Glc concentration in medium was measured using a YSI Biochemistry Analyzer 2900 (Xylem, Inc; ref. 37). Glc consumption by cells was calculated by subtracting the value in cell culture medium obtained from wells containing cells from the value in non–cell-containing medium. Values were normalized to cell number.

Immunoblotting

WT and KO cells were grown in six-well plates until 95% (not 100%) confluency and then trypsinized, washed, and pelleted in ice-cold PBS (1,000 rpm, 5 minutes). The pellet was resuspended in ice-cold lysis buffer [10 mmol/L Tris-HCl, pH 7.5, 50 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L DTT, 10% glycerol, 10 mmol/L sodium molybdate, 0.1% Triton X-100, and protease (4693124001, Roche) and phosphatase (4906837001, Roche) inhibitor cocktails] and lysed by sonication (AMP 55, 15 minutes, 30 seconds pulse on, and 40 seconds pulse off) using a chilled bath sonicator (Qsonica). Lysate protein concentration was estimated using Bradford assay (Bio-Rad), and proteins were separated by SDS-PAGE using Mini-PROTEAN TGX precast gels (Bio-Rad). For all whole-cell lysate Western blots, 50 μg protein was loaded into each well. Transfer to nitrocellulose membranes was performed using the Trans-Blot turbo system (Bio-Rad), and transfer efficiency was checked using Ponceau S solution (P7170, Sigma). Antibodies against TRAP1, xCT, glutathione peroxidase 4 (GPX4), HSP60, and GAPDH and their respective dilutions used for immunoblotting are outlined in Supplementary File S5: Table 5: List of reagents and resources. Proteins were visualized with a Li-COR Odyssey imager using the SuperSignal West Pico PLUS chemiluminescence detection system (34580, Thermo Fisher Scientific).

RT-qPCR analysis

RNA extraction and RT-qPCR were performed as previously described (38). Briefly, RNA was isolated using RNeasy RNA Isolation Kit according to the manufacturer’s protocol (Qiagen). For reverse transcription (RT) reactions, 200 ng of total RNA was used in a reaction mixture containing 1× TaqMan RT buffer (Applied Biosystems) and MultiScribe Reverse Transcriptase. RT was performed at 25°C for 10 minutes, then at 48°C for 30 minutes, and finally at 95°C for 5 minutes using the PE9700 thermal cycler (Applied Biosystems). All PCR primers were designed using Primer Express software (Applied Biosystems). Relative RNA levels were calculated by the comparative Ct method as described by the manufacturer. Experiments were performed in triplicate. The data normalization was based on one of the three biological replicates, not a combination of all three.

Proteasome inhibitor treatments

For treatment with the proteasome inhibitor MG132, HEK293T WT and TRAP1 KO cells were seeded at a density 5 × 10⁵ cells/ well in six-well plates. At approximately 95% confluency (not 100%), cells were treated with 20 μmol/L MG132 for 2, 4, and 6 hours. Lysate preparation and immunoblotting were done as described above. Poly-ubiquitin (poly-Ub) antibody and its dilution used for immunoblotting are outlined in Supplementary File S5: Table 5: List of reagents and resources.

Statistical analysis

All data analyses were performed using a combination of GraphPad Prism 8, Microsoft Excel, Wave 2.0, BD FACSDiva and Li-COR Image studio software. Differences between groups were analyzed using a two-tailed Student t test. Unless specified, the error bars represent the SEM with *, P < 0.05; **, P < 0.01; and ***, P < 0.001 denoting the difference between the means of two compared groups considered to be statistically significant. Each real-time OCR tracing profile presents a cumulative plot of three technical replicates per cell type. Unless specified, each bar graph represents a cumulative plot for three independent experiments.

Data availability

All data generated in this study are available upon request to the corresponding author. Supplementary figures and tracing data are presented in the attached Supplementary Figs. S1 through S5 and Supplementary Files S1, S2, S3, and S4 (Tables 1 through 4). Supplementary File S5: Table 5 provides a comprehensive list of reagents and resources used in our experiments.

Results

TRAP1 deletion forces mitochondrial respiration to rely on Gln

The TRAP1 gene was disrupted in UMUC3 (bladder cancer) and PC3 (prostate cancer) cells using the CRISPR/Cas9 workflow presented in Supplementary Fig. S1A. HEK293T, HCT116, and A549 TRAP1 KOs were generated previously (15). First, we confirmed that TRAP1 KO resulted in increased mitochondrial respiration as measured by the cellular OCR, a direct measure of mitochondrial OxPhos. The OCR was recorded in real-time in TRAP1 WT and KO UMUC3 and PC3 cells (Fig. 1A and B). When grown in medium containing the standard carbon sources Glc, Pyr, and Gln, we observed a significant increase in basal OCRs in both UMUC3 and PC3 TRAP1 KO cells when compared with their isogenic WT controls (Fig. 1C and D). To further corroborate these findings, we measured electron transport chain complex I activity in both UMUC3 and PC3 TRAP1 KOs and found it to be significantly elevated (Fig. 1E and F). It should be noted that although we observe a consistent increase in the OCR upon TRAP1 KO, the impact is not comparable across multiple cell lines, likely due to their distinct metabolic preferences (39, 40).

Figure 1.

Metabolic profiling of human TRAP1 KO cells. A and B, Representative real-time OCR traces of WT and TRAP1 KO UMUC3 (A) and PC3 (B) cells with Glc + Pyr + Gln as carbon sources. Measurement of basal OCR was followed by sequential injections of the ATP synthase inhibitor (oligomycin; 5 μmol/L), an uncoupler of ATP-dependent OCR (carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; 2 μmol/L), and a combination of complex I (rotenone; 1 μmol/L) and complex III (antimycin; 1 μmol/L) electron transport chain inhibitors (n = 3). C and D, Quantitation and comparison of basal respiratory rates in WT vs. TRAP1 KO UMUC3 (C) and PC3 (D) cells (n = 3). E and F, Relative complex I activities comparing WT with TRAP1 KO UMUC3 (E) and PC3 (F) cells (n = 3). G and H, Comparative OCR traces of WT and TRAP1 KO UMUC3 cells with Gln (G) and Glc (H) as the only carbon sources (n = 3). I and J, Quantitation of basal respiratory rates in WT and TRAP1 KO UMUC3 cells with Gln (I) and Glc (J) as the only carbon sources (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; DHAP, dihydroxyacetone phosphate; PEP, phosphoenol pyruvic acid; PGA, phosphoglyceric acid.

After establishing new TRAP1 KOs whose metabolic phenotype was consistent with previous reports (3, 4, 7, 15), we investigated the impact of TRAP1 KO on Glc and Gln metabolism. Our previous study was the first to report that loss of TRAP1 induced a metabolic defect in cells which led to a decline in Glc contribution to mitochondrial respiration while increasing Gln anaplerosis to support the TCA cycle (15). We performed similar experiments with UMUC3 TRAP1 KO cells and recorded their OCR in real-time in growth media supplemented with either Gln or Glc as the only carbon source (Fig. 1G and H). Previous observations could be replicated in these bladder cancer cells, as we observed a significant increase in mitochondrial respiration (basal OCR) when TRAP1 KO UMUC3 cells were grown in Gln-only media (Fig. 1I), whereas a significant decline in mitochondrial respiration was observed when the KO cells were provided with Glc as the only carbon source (Fig. 1J). Similar data were obtained with prostate cancer–derived PC3 TRAP1 KO cells (Supplementary Fig. S1B and S1C). These data indicate that the dynamics of Glc and Gln metabolism are replicable and consistent in a panel of TRAP1 KO cells derived from different tumor types.

Loss of TRAP1 does not redirect Glc to the PPP but attenuates Glc uptake and Glc-derived OxPhos

Reduced Glc influx into the TCA cycle may indicate its r-direction into other metabolic pathways to balance TRAP1 KO-associated anomalies. Loss of TRAP1 has been reported to induce ROS in a variety of cell models (5, 26, 41, 42). Indeed, quantitative measurement of H₂DCFDA fluorescence intensity by flow cytometry revealed increased total cell ROS across all TRAP1 KO cell lines, including A549, UMUC3, and PC3 KO cells (Fig. 2A). Based on these data, one possible explanation for the reduction in Glc contribution to the TCA cycle may be its redirection into the PPP to counter oxidative stress (43, 44) generated in TRAP1 KO cells (5, 26, 41, 42). To test this hypothesis, a metabolic flux analysis using stable isotope tracing was performed with U-¹³C-Glc to trace its path in glycolysis and the PPP (Fig. 2B). Isotopically labeled ¹³C-Glc was added to the growth medium in addition to unlabeled Pyr and Gln as carbon sources. Detected metabolites were compared across three biological replicates of TRAP1 KO and WT cells for both PC3 and HCT116. Tracing data are presented in Supplementary Files S1 and S2 (Tables 1 and 2). For the analysis, we focused on the fractional distribution of ¹³C (in %, normalized to natural existence) into specific isotopologue populations of detected glycolytic and PPP metabolites (Supplementary Fig. S2A) and their specific concentrations (Fig. 2C). Surprisingly, we did not observe any significant difference in the fractional distribution of ¹³C (%) to isotopologues of interest from most glycolytic and PPP metabolites when comparing TRAP1 KO with WT cells in both PC3 and HCT116 (Supplementary Fig. S2A). There were a few outliers, but there did not seem to be a trend across the two cell lines (Supplementary Fig. S2A). However, when comparing the concentrations of these isotopologues in WT versus TRAP1 KO cells from both PC3 and HCT116, a trend emerged, indicating an overall decline in the amount of ¹³C-tagged metabolites in TRAP1 KO cells (Fig. 2C). Many metabolite-specific isotopologues from both glycolysis and the PPP showed a significant reduction in concentration compared with WT cells in both cell lines (Fig. 2C). Lactic acid (m+3 isotopologue), the classic byproduct of Glc metabolism, showed a strong decline in concentration in PC3 TRAP1 KO cells and a significant decline in HCT116 TRAP1 KOs when compared with their respective WT controls (Fig. 2C). Taken together, these data suggested that whereas Glc may be processed similarly in TRAP1 KO and WT cells, there is likely an overall decline in the total Glc uptake and input to glycolysis and the PPP in TRAP1 KO cells.

Figure 2.

Total cellular ROS and Glc tracing in human tumor-derived TRAP1 KO cells. A, Comparative assessment of total cell ROS in WT and TRAP1 KO A549, UMUC3, and PC3 cells using the cell-permeant fluorescent ROS probe H₂DCFDA. ROS levels in cells are represented as the mean fluorescent intensity (MFI) which indicates an average of H₂DCFDA fluorescence measured from 10,000 individual cells by flow cytometry (n = 3). B, Flowchart showing the path of isotopically labeled ¹³C-Glc through the CCM. C, Comparison of concentrations of specific ¹³C isotopologues from various glycolytic and PPP metabolites between WT and TRAP1 KO PC3 and HCT116 cells. #, indicates a metabolite common to both glycolysis and the PPP (n = 3). D, Measurement of Glc consumption by WT and TRAP1 KO A549, UMUC3, HCT116, and PC3 cells using a YSI 2500 biochemistry analyzer. Glc consumption was calculated as (Glc concentration in wells with no cells) − (Glc concentration in wells with cells). This value was normalized to cell number and represented as mmol/L/day (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonsignificant.

To test this further, equal numbers of WT A549, UMUC3, HCT116, and PC3 cells, and their TRAP1 KO pairs, were seeded in 12-well plates with complete medium (Glc + Pyr + Gln). Empty wells with an equal amount of the same medium were used as controls. After 24 hours of incubation, Glc concentration in the medium from cell-containing and control wells was measured using a YSI biochemistry analyzer 2900 (Xylem, Inc). Glc consumption was calculated by subtracting values obtained from control wells. A significant decline in Glc consumption was recorded for A549, HCT116, and PC3 TRAP1 KO cells compared with their WT counterparts (Fig. 2D). This corroborates tracing data (in Fig. 2C) and shows that TRAP1 KOs display reduced Glc uptake. Of interest, growth rate analysis of PC3, UMUC3, HEK293T, and HCT116 cells using an Incucyte instrument (Sartorius) showed that the KO cells might not always proliferate at the same rate as the WT controls when grown in media with the standard carbon sources (Supplementary Fig. S2B). Although not significant, variable growth rates might lead to slightly different cell numbers during biochemical analyses and metabolite extraction for tracing experiments. This variable was mitigated by normalizing all tracing (Fig. 2C; Supplementary Fig. S2A; Supplementary Files S1 and S2) and Glc uptake (Fig. 2D) data to cell number.

Gln anaplerosis drives both reductive carboxylation and the oxidative arm of the TCA cycle in TRAP1 KO cells

Upon establishing that Glc uptake and input into glycolysis (Fig. 2C and D) and the TCA cycle (Fig. 1H and J) are reduced in a panel of TRAP1 KO cells, further experiments were designed to solely focus on Gln metabolism which was consistently found to be affected across various WT and KO pairs [Fig. 1G and data from Joshi and colleagues (15)].

Metabolic analyses with different cell types such as HEK293T (15), HCT116 (15), UMUC3 (Fig. 1G), and PC3 (Supplementary Fig. S1B) have consistently shown that Gln may act as an anaplerotic substrate to compensate for Glc withdrawal from mitochondrial respiration in TRAP1 KO cells. To confirm an anaplerotic metabolic shift in cells lacking the TRAP1 protein, we performed flux tracing experiments with uniformly labeled U-¹³C-Gln in UMUC3 and HCT116 WT cells and their TRAP1 KO counterparts [Fig. 3; Supplementary Files S3 and S4 (Tables 3 and 4); refer to sheet 1 for fractional contribution and sheet two for concentrations]. Unlike unidirectional flow of Glc, Gln can be directed both into reductive carboxylation [counterclockwise from α-ketoglutaric acid (α-KG)] or into the oxidative arm (clockwise from α-KG) of the TCA cycle (Fig. 3A), both of which activities are critical to drive respiration in defective mitochondria (29, 45). To verify both arms of ¹³C-Gln flow, we chose to focus on specific isotopologue populations (normalized to natural existence) that demonstrate ¹³C contribution to reductive carboxylation of Gln (Fig. 3A, m+5 citrate counterclockwise, depicted by blue squares) and to the oxidative arm of the TCA cycle (Fig. 3A, m+4 citrate clockwise). Starting with reductive carboxylation and focusing on m+5 isotopologues for glutamic acid, α-KG, and citric acid (Fig. 3B–D), we noted a consistent increase in ¹³C incorporation into these metabolites generated in TRAP1 KO cells from both UMUC3 and HCT116 cell lines. The m+3 isotopologues of malic acid and aspartic acid generated from the m+5 citrate also showed a consistent increase in Gln-derived ¹³C incorporation in both UMUC3 and HCT116 TRAP1 KO cells (Fig. 3E and F).

Figure 3.

Gln flux in mitochondrial respiration of TRAP1 KO cells. A, Schematic representation showing ¹³C flow and distribution from uniformly labeled Gln as it is processed through reductive carboxylation (blue dotted squares) or through the oxidative arm of the TCA cycle. B–F, Comparison of fractional contribution of ¹³C (%) to isotopologue populations in (B) glutamic acid, (C) α-KG, (D) citric acid, (E) malic acid, and (F) aspartic acid between WT and TRAP1 KO UMUC3 and HCT116 cells (n = 2). Colored underlines highlight isotopologues derived from reductive carboxylation (dark blue) or the oxidative arm (light blue) of the TCA cycle. *, P < 0.05; **, P < 0.01.

Focusing next on the oxidative arm of the TCA cycle, we noted an increase in ¹³C incorporation into m+4 isotopologues of malic and aspartic acid in TRAP1 KO cells from both cell lines when compared with WT controls (Fig. 3E and F) indicating that in TRAP1 KO cells, Gln fuels the oxidative arm as well as reductive carboxylation.

The Cys/Glu antiporter SLC7A11/xCT is downregulated in TRAP1 KO cells to support Gln retention and anaplerosis without inducing ferroptosis

For Gln to be utilized as a primary carbon source in mitochondrial respiration under Glc withdrawal, it must be actively retained in cells. We have previously reported that metabolic tracing with U-¹³C-Gln revealed an increase in total Gln levels in HEK293T and A549 TRAP1 KO cells when compared with WT cells (15), supporting the possibility of Gln anaplerosis in TRAP1 KO cells. The Cys/Glu antiporter xCT (SLC7A11), which exports Glu in exchange for Cys, is an important protein implicated in metabolic flexibility, nutrient dependency, redox homeostasis, ferroptosis, an tumor growth and progression (30, 33, 46–49). Disruption of the SLC7A11/xCT antiporter has been shown to greatly increase cell viability under Glc deficiency as it preserves Gln and enables cells to sustain mitochondrial respiration via Gln anaplerosis (32, 49). Based on these observations, we tested whether xCT dysregulation in TRAP1 KO cells is a potential means to retain Gln. Studies often predict xCT’s molecular weight to be ∼55 kDa, but it lies between 30 and 40 kDa. We verified our xCT antibody [xCT (D2M7A) Rb mAb #12691] and consistently observed a band at ∼35 kDa across all tested cell lines, with a faint band below it (Supplementary Fig. S3A–S3C). The manufacturer indicated that the lower faint band may result from posttranslational cleavage, with the 35 kDa band being the target. No higher molecular weight bands were observed in any of the tested cell lines. Immunoblotting analysis of whole-cell extracts showed a significant decline in xCT protein levels in HEK293T, UMUC3, and PC3 TRAP1 KO cells when compared with their isogenic WT controls (Fig. 4A; Supplementary Fig. S3B and S3C). HSP60 and GAPDH were used as mitochondrial and cytosolic loading controls, respectively. Although these data were promising, specific pharmacologic inhibition of xCT (50–52) or genetic ablation of the SLC7A11 gene (53) have been reported to induce ferroptosis (54–56), a type of programmed cell death that is biochemically distinct from apoptosis. Inhibition or deletion of xCT leads to decreased Cys-dependent glutathione (GSH) levels, inhibiting GSH-dependent GPX4 antioxidant activity (57, 58) which is a primary factor for regulating ferroptosis. Because TRAP1 deletion led to a decline in endogenous levels of xCT, we determined whether TRAP1 KO might itself induce ferroptosis. We first performed immunoblots to monitor the expression of GPX4 protein (Fig. 4B; Supplementary Fig. S3D), whose expression is inversely correlated with ferroptosis (59, 60). We quantified and compared GPX4 protein levels side-by-side in HEK293T and PC3 TRAP1 WT and KO cells, with or without treatment with erastin, a pharmacologic xCT inhibitor (Fig. 4B and C; Supplementary Fig. S3D and S3E; refs. 51, 52, 61, 62). We also evaluated HEK293T TRAP1 KO cells overexpressing the xCT protein (Fig. 4B and C). Interestingly, a decline in endogenous xCT protein level in HEK293T or PC3 TRAP1 KO cells did not affect GPX4 expression when compared with isogenic WT cells (Fig. 4B and C; Supplementary Fig. S3D and S3E), thereby showing that TRAP1 KO-dependent reduction in xCT protein level by itself may not be sufficient to induce ferroptosis. After comparing TRAP1 KO cells with TRAP1 KOs overexpressing the xCT protein in HEK293T, we concluded that wide variations in xCT protein levels may not impact GPX4 expression (Fig. 4B and C; for PC3 cells refer to Supplementary Fig. S3F). Most notably, and in agreement with other published data (63, 64), GPX4 levels only declined when WT or TRAP1 KO cells (both HEK293T and PC3) were treated overnight with 10 μmol/L erastin. Of note, erastin displayed a significantly stronger impact in WT cells when compared with their TRAP1 KO counterparts (refer to Fig. 4C and Supplementary Fig. S3E). To confirm whether overexpression of the xCT protein in the TRAP1 KO background restored “WT-like” erastin sensitivity, we performed immunoblots to compare GPX4 expression in erastin-treated HEK293T and PC3 TRAP1 KO cells compared with TRAP1 KOs overexpressing xCT (Fig. 4D; Supplementary Fig. S3F). Indeed, we found that overexpression of the xCT protein in TRAP1 KO cells, both HEK293T and PC3, restored a WT-like phenotype with respect to erastin sensitivity and associated GPX4 levels (Fig. 4D; Supplementary Fig. S3F).

Figure 4.

SLC7A11/xCT expression and ferroptosis in TRAP1 KO cells. A and B, Immunoblots of denaturing protein gels showing xCT and TRAP1 expression in whole-cell extracts of WT vs. TRAP1 KO HEK293T cells (A), and GXP4 expression in WT vs. TRAP1 KO HEK293T cells treated or not with the xCT inhibitor erastin (10 μmol/L), and in TRAP1 KO cells overexpressing an untagged xCT protein (B). HSP60 and/or GAPDH were used as loading controls (mitochondrial and cytosolic, respectively) to check the quality of the extracts. C, Densitometric analysis of immunoblots comparing GPX4 expression in WT and TRAP1 KO HEK293T cells treated or not with erastin and in KO cells overexpressing an untagged xCT protein (n = 3). D, Immunoblot of a denaturing gel comparing GPX4 expression, with or without erastin treatment, in HEK293T TRAP1 KO cells vs. TRAP1 KO cells overexpressing an untagged xCT protein. E, Comparison of total Cys, GSH, and GSSG levels, WT vs. TRAP1 KO, in A549 and HEK293T cells (n = 2). *, P < 0.05; Mw, molecular weight; ns, nonsignificant.

Importantly, and supporting the findings above, xCT-dependent reduction in intracellular Cys is considered be a primary factor in inducing ferroptosis (30, 31, 49, 65–67). Because TRAP1 deletion impacted xCT protein level without apparently inducing ferroptosis, we examined endogenous Cys levels in TRAP1 WT and KO cells. Analyzing two biological replicates for each cell type, we found that A549 and HEK293T TRAP1 KO cells displayed a similar trend of increased Cys and glutathione [both reduced glutathione (GSH) and glutathione disulphide (GSSG)] levels when compared with their isogenic WT controls (Fig. 4E). These data suggest that intracellular Cys levels may be maintained only in part by the xCT protein and that there are other factors at play in regulating intracellular Cys in TRAP1 KO cells. Finally, growth analyses using the Incucyte instrument (Supplementary Fig. S2B) also did not show any stark differences between growth rates of TRAP1 KO HEK293T, HCT116, UMUC3, and PC3 cells compared with their WT controls. Taken together, our GPX4 expression, metabolite concentration, and growth analysis data are not consistent with the hypothesis that TRAP1 KO spontaneously induces ferroptosis.

Overexpression of the Cys/Glu antiporter SLC7A11/xCT in TRAP1 KO cells restores “WT-like” Gln metabolism

Next, we asked whether mitochondrial respiration itself was differentially sensitive to erastin treatment in WT versus TRAP1 KO cells. According to our GPX4 immunoblots, higher xCT expression in WT cells renders them more susceptible to erastin when compared with TRAP1 KOs. Based on these data, we hypothesized that xCT inhibition with erastin should also have a stronger impact on mitochondrial OCR in WT cells when compared with TRAP1 KOs. To test this, we performed real-time OCR analyses using HEK293T TRAP1 WT and KO cells treated with erastin but with Gln as the only available carbon source for mitochondrial respiration. The cells had either direct injection of 10 μmol/L erastin (Supplementary Fig. S4A and S4B) or were incubated with 5 μmol/L (orange line, Supplementary Fig. S4C and S4D) and 10 μmol/L erastin (red line, Supplementary Fig. S4C and S4D) overnight. All OCR analyses were performed simultaneously on the same plate using both TRAP1 WT and KO cells, and the data are presented using the same scale in Supplementary Fig. S4A–S4D. Direct injection of 10 μmol/L erastin had a negligible impact on basal mitochondrial OCR in both TRAP1 WT and KO cells (Supplementary Fig. S4A and S4B, after the blue arrow) indicating no immediate metabolic response to the drug. In contrast, overnight treatments with erastin resulted in a decline in basal OCRs of both TRAP1 KO and WT cells (Fig. 5A). As predicted, TRAP1 WT cells were significantly more sensitive to erastin than were TRAP1 KO cells. The basal OCR in HEK293T WT cells showed a greater decline when compared with TRAP1 KO cells after treatment with both 5 μmol/L (∼21% in WT vs. ∼9.5% in KO) and 10 μmol/L (∼61% in WT vs. ∼37% in KO) erastin when compared with their respective controls (Fig. 5A).

Figure 5.

SLC7A11/xCT overexpression and Gln metabolism in TRAP1 KO cells. A, Comparison of basal OCR in WT and TRAP1 KO HEK293T cells with or without overnight treatment with 5 and 10 μmol/L erastin (n = 3). B, Immunoblot of a denaturing protein gel (7.5% SDS-PAGE) showing xCT overexpression in xCT transfected TRAP1 KO HEK293T cells. Mw, molecular weight. C, OCR traces comparing WT with TRAP1 KO HEK293T cells with or without overexpression of xCT protein. The mitochondrial stress test profile was obtained by a sequential injection of oligomycin (5 μmol/L), the uncoupler carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; 2 μmol/L), and the electron transport chain (ETC) complex I and III inhibitors rotenone (1 μmol/L) and antimycin (1 μmol/L), respectively (n = 3). D, Basal OCR of WT and TRAP1 KO HEK293T cells with or without overexpression of xCT (n = 3). *, P < 0.05; **, P < 0.01.

After ruling out the possibility of spontaneous ferroptosis in TRAP1 KO cells and confirming that xCT is involved in Gln-dependent mitochondrial respiration, we examined the role of xCT in Gln anaplerosis. To do this, we transiently overexpressed xCT in HEK293T TRAP1 KO cells (Fig. 5B) and recorded their OCR in real-time with Gln as the only available carbon source (Fig. 5C). Following xCT overexpression, a significant (∼40%) decline in basal mitochondrial OCR was observed in KO cells when compared with untransfected controls (Fig. 5D). The large increase in xCT levels in KO cells following xCT overexpression (refer to Figs. 5B and 4B or 4 days for comparisons) also decreased basal OxPhos below WT levels with Gln as the only carbon source (Fig. 5D). Similar effects were observed in PC3 cells, in which transient overexpression of xCT in TRAP1 KOs led to a significant decline in basal mitochondrial OCR with Gln as the sole carbon source (Supplementary Fig. S4E and S4F).

SLC7A11/xCT expression is regulated at the level of protein stability in TRAP1 KO cells

After confirming reduced xCT expression in a panel of TRAP1 KO cells (Fig. 4A; Supplementary Fig. S3B and S3C) and its direct involvement in promoting Gln-driven mitochondrial OCR in HEK293T and PC3 cells (Fig. 5C and D; Supplementary Fig. S4E and S4F), we investigated how its expression/stability may be affected by the presence or absence of the TRAP1 protein. Extramitochondrial protein expression is known to be affected by dynamic alterations in the mitochondria (68–71), and change in xCT expression may be an indirect consequence of TRAP1-associated alterations in mitochondrial metabolism (3, 4, 7, 15, 20). In case of TRAP1 deficiency, a further level of complexity is added by transcriptional and posttranslational regulation of xCT expression (31).

The transcription factors NRF2, ATF4, and ETS-1 are reported to be involved in cooperative upregulation of xCT expression (31, 48, 72, 73) in cells under mitochondrial stress (31, 74–76). We first questioned whether a decline in the levels of these transcription factors may play a role in reduced expression of xCT in TRAP1 KO cells. This seemed highly unlikely in the case of NRF2 and ATF4 in particular, as both transcription factors are upregulated in the context of elevated ROS (74, 75), a phenotype common to TRAP1 KO cells (Fig. 2A). We performed RT-qPCR experiments with HEK293T and PC3 WT and TRAP1 KO cell extracts and calculated fold change in the mRNA expression of these molecules (Fig. 6A; Supplementary Fig. S5A). Unsurprisingly, we observed a significant increase in mRNA expression of NRF2 and ATF4 in HEK293T TRAP1 KO cells when compared with WT cells (Fig. 6A), which is not consistent with the decline in xCT protein (Fig. 4A). In line with the observations in HEK293T cells, we noted a significant increase in ATF4 mRNA expression in PC3 cells; however, there was no difference in NRF2 expression (Supplementary Fig. S5A). ETS-1 expression is reported to be inversely correlated with mitochondrial respiration, and an increase in mitochondrial OCR should lead to a decline in ETS-1 levels (77). Indeed, we observed a decline in ETS-1 mRNA expression in both HEK293T and PC3 TRAP1 KO cells (Fig. 6A; Supplementary Fig. S5A). Finally, we did not observe a decline in SLC7A11/xCT mRNA expression in HEK293T or PC3 TRAP1 KO cells when compared with their isogenic WT controls (Fig. 6A; Supplementary Fig. S5A).

Figure 6.

Regulation of SLC7A11/xCT expression in TRAP1 KO cells. A, Quantitative RT-qPCR analysis of the mRNA levels for ETS-1 (A and B), ATF4, NRF2, and SLC7A11 in HEK293T WT and TRAP1 KO cells. All data are reported as means $\mathit{\pm }$ SEM (n = 3). B, A representative immunoblot of a denaturing protein gel (7.5% SDS-PAGE) and quantitative densitometric analysis of three independent biological replicates (mean $\mathit{\pm }$ SEM) of xCT stabilization in WT (left) vs. TRAP1 KO HEK293T cells (right) after 2, 4, and 6 hours of exposure to the proteasome inhibitor MG132. A pan poly-Ub antibody was used to check MG132 efficacy (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; Mw, molecular weight; ns, nonsignificant.

Because RT-qPCR analysis showed that xCT transcription (Fig. 6A; Supplementary Fig. S5A) did not correlate with its protein expression (Fig. 4A; Supplementary Fig. S3C), we performed experiments with the proteasome inhibitor MG132 to assess impact on xCT protein stability. After exposing HEK293T TRAP1 KO cells to 2, 4, and 6 hours of 20 μmol/L MG132 [Fig. 6B (right)], we observed a steady and significant increase in xCT protein level by immunoblotting (and its quantitation thereof), suggesting that xCT protein expression in TRAP1 KO cells is regulated at the level of protein turnover/stability. GAPDH was used as a loading control, and a pan poly-Ub antibody was used to detect MG132 efficacy for proteasome inhibition. In contrast to HEK293T WT cells which showed a negligible elevation in xCT protein levels after MG132 compared with DMSO control [Fig. 6B (left)], TRAP1 KO cells responded more robustly to MG132 [Fig. 6B (right)]. Proteasomal degradation often involves ubiquitination, which typically increases a protein’s molecular weight. Accordingly, MG132-induced stabilization of xCT in TRAP1 KO cells could, in principle, result in the accumulation of a higher molecular weight, Ub-conjugated form of xCT. However, at the tested time points, xCT remained near its expected molecular weight of ∼35 kDa. This was evident in high-exposure anti-xCT (D2M7A) immunoblots of MG132-treated HEK293T TRAP1 KO cell extracts, in which only faint signals above 250 kDa were detected under extreme exposure conditions and are likely nonspecific (Supplementary Fig. S5B). The observation that xCT migrates near its predicted molecular weight suggests that although the proteasome likely contributes to its degradation, this process may not involve stable accumulation of ubiquitinated xCT species or that alternative degradation mechanisms may also be involved. In light of our findings, one possible explanation is a proteasome-dependent but Ub-independent mode of degradation or the involvement of monoubiquitination or transient Ub modifications that are not readily detectable under our experimental conditions. Alternatively, rapid Ub turnover could prevent stabilization of ubiquitinated xCT intermediates. Whereas our data support at least a partial role for the proteasome in xCT turnover, the absence of detectable Ub-conjugated species means that contributions from other degradation pathways cannot be ruled out. Clarifying the precise mechanisms will require further investigation beyond the scope of the present study.

MG132 experiments with PC3 WT and TRAP1 KO cells yielded results comparable with HEK293Ts (Supplementary Fig. S5C). Consistent with the immunoblotting data from HEK293T TRAP1 KO cells [Fig. 6B (right)], we observed a significant and steady increase in xCT protein expression in PC3 TRAP1 KO cells [Supplementary Fig. S5C (right)]. Densitometric analyses from three biological replicates showed a significant increase in xCT protein levels in PC3 WT cells at the 6-hours MG132 treatment timepoint when compared with the DMSO control [Supplementary Fig. S5C (left)]. However, whereas WT cells exhibited a modest ∼1.2-fold increase, PC3 TRAP1 KOs showed a substantial ∼10-fold increase in xCT levels when compared with their DMSO controls. All MG132 data were confirmed by three biological replicates in both HEK293T and PC3 cell lines (mean $\mathit{\pm }$ SEM; Fig. 6B; Supplementary Fig. S5C) and are consistent with a model in which TRAP1 KO cells actively degrade xCT protein to conserve Gln. Despite undergoing degradation in TRAP1 KO cells, xCT did not accumulate as high–molecular weight species following MG132 treatment, suggesting that whereas Ub-independent proteasome-mediated degradation is a potential mechanism, alternative degradation pathways should not be ruled out.

Discussion

TRAP1 is uniquely positioned in the HSP90 chaperone family because of its ability to alter the metabolic preference of cells (3, 7, 15) without having a significant impact on the mitochondrial proteome (15). Specifically, the presence of TRAP1 drives cells toward aerobic glycolysis (3, 7), whereas its deletion leads to an increase in mitochondrial OxPhos (3, 15). Collectively, these findings indicate that whereas TRAP1 promotes a metabolic shift toward aerobic glycolysis, its absence results in a metabolic defect that is compensated by an increased reliance on Gln (15). This Gln-dependent anaplerosis signals mitochondrial dysfunction and is crucial for replenishing and maintaining the mitochondrial carbon pool when Glc is withdrawn from the CCM (12, 31, 32, 78).

In this study, we sought to further understand the metabolic “defect” induced by the loss of the TRAP1 protein wherein Glc is directed away from the TCA cycle, and we sought to uncover a possible mechanism by which TRAP1 KO cells may retain Gln to compensate for Glc withdrawal. A possible direction in which Glc may be channeled away from the TCA cycle is in the oxidative arm of the PPP, which is used to generate NADPH (43). This molecule is critically important as it provides the reducing power required to fuel protein-based antioxidant systems and to recycle oxidized glutathione (43). Given that oxidative stress is a frequently observed phenotype associated with TRAP1 dysfunction or loss (5, 25, 26, 41, 42), we hypothesized that redirection of Glc to the PPP to generate reducing power may be a general consequence in cells lacking the TRAP1 protein (5, 25–28). Surprisingly, we found that Glc was not redirected into the PPP to counter the increased oxidative load generated in TRAP1-null cells. In fact, Glc uptake and input into glycolysis and the PPP is consistently reduced in cells lacking TRAP1.

Glc withdrawal from the TCA cycle in the absence of TRAP1 induces Gln-mediated anaplerosis via both reductive carboxylation and the oxidative arm of the TCA cycle. Our data suggest that downregulation of the Cys/Glu antiporter xCT or SLC7A11 may be one way in which TRAP1-deficient cells conserve Gln to power the TCA cycle. xCT expression is typically regulated by transcription factors ETS-1, NRF2, and ATF4, whose levels are under direct influence of mitochondrial respiration and oxidative stress (31, 74–76). Variations in these transcription factors do not seem to influence xCT mRNA levels in TRAP1 knockout cells. Instead, our findings indicate that xCT expression in TRAP1-deficient cells is regulated, at least in part, posttranslationally through proteasome-dependent degradation, although the involvement of additional degradation pathways cannot be excluded. Notably, robust stabilization of xCT following MG132 treatment in TRAP1 KO cells did not coincide with the accumulation of higher molecular weight species typically associated with polyubiquitinated intermediates. This observation suggests that xCT turnover may occur via noncanonical proteasomal degradation mechanisms that do not rely on the stabilization of detectable polyubiquitin chains. One possibility is that xCT undergoes monoubiquitination or transient Ub modifications that are rapidly reversed or processed, rendering them undetectable under standard immunoblotting conditions. Alternatively, degradation may occur through Ub-independent mechanisms, which have been described for certain membrane proteins and may involve recognition by alternative proteasome adapters or chaperones. It is also plausible that the apparent lack of ubiquitinated intermediates reflects a highly efficient degradation process with minimal accumulation of substrate. Whereas our data support a role for the proteasome in xCT turnover, the precise mechanism by which xCT is targeted for degradation in the absence of TRAP1 remains unresolved. Further studies will be required to define the molecular signals governing xCT proteostasis and to determine whether TRAP1 plays a direct role in modulating its degradation pathway. Genetic ablation of the Cys/Glu antiporter xCT has been reported to induce ferroptosis (53), a form of regulated cell death characterized by iron-dependent lipid peroxidation (54–56). Despite the reduction in xCT levels in TRAP1 KO cells, our data indicate that Cys levels remain sufficient to prevent ferroptosis. This suggests that TRAP1-deficient cells use alternative mechanisms to maintain Cys homeostasis and avoid ferroptotic cell death.

Taken together, the findings of this study demonstrate that a peripheral analysis of TRAP1-associated metabolic rewiring under nonlimiting conditions may not always represent an accurate picture of what this protein does with respect to metabolism and that loss of TRAP1 does not necessarily boost mitochondrial respiration as was previously thought. TRAP1 seems to be critical for Glc uptake and routing Glc into the TCA cycle. TRAP1 KO cells compensate for this metabolic defect by downregulating xCT protein expression at the level of protein stability to conserve Gln and support the TCA cycle by anaplerosis without inducing ferroptosis.

Although the data are robust, it is important to acknowledge two major limitations of this study. The first unresolved question is how TRAP1, an intramitochondrial protein, influences Glc uptake and metabolism outside the mitochondria. One plausible hypothesis may be an indirect effect via succinate, as TRAP1 inhibits succinate dehydrogenase activity, leading to succinate accumulation, stabilization of hypoxia-inducible factor 1-α, and increased glycolysis (4). Loss of TRAP1 has a limiting effect on glycolysis (7, 15), and this observed reduction in the glycolytic rate (and Glc uptake in general) might be due to feedback generated by a decline in succinate levels and destabilization of hypoxia-inducible factor 1-α in TRAP1-null cells. Testing this hypothesis remains challenging with the current set of tools available, as alterations in succinate levels can also have other complex, widespread, and interconnected effects on cellular metabolism. The regulation of xCT expression presents the second significant issue. Our qPCR data show that the expression levels of NRF2, ATF4, and ETS-1 do not correspond with xCT levels. As mentioned earlier, xCT seems to be downregulated at the protein level in TRAP1 KO cells. This unique mechanism utilized by TRAP1 KO cells may reflect either unique posttranslational modification(s; PTM) on xCT or a secondary effect on its stability that is not directly associated with the loss of TRAP1 but may be indirectly affected by the metabolic defect introduced in the cell due to the absence of this protein. Because identification of validated xCT PTM(s) is in its early stages, we can only speculate. Exploring secondary effects of PTMs or other factors are beyond the scope of this study. Future analyses will likely shed more light on these possibilities.

Supplementary Material

Supplementary Legends

This file has legends for supplementary figures and tables

Supplementary Figure S1

Generation of TRAP1 KO cells using CRISPR/Cas9

Supplementary Figure S2

Glucose tracing and growth rates of various TRAP1 KO cell lines

Supplementary Figure S3

SLC7A11/xCT expression in HEK293T, UMUC3 and PC3 TRAP1 KO cells

Supplementary Figure S4

OCR traces comparing erastin treatments and xCT overexpression

Supplementary Figure S5

Regulation of SLC7A11/xCT expression in PC3 and HEK293T TRAP1 KO cells

Supplementary File 1: Table 1

This table has 13C Glc tracing data for WT and TRAP1 KO HCT116 cells

Supplementary File 2: Table 2

This table has 13C Glc tracing data for WT and TRAP1 KO PC3 cells

Supplementary File 3: Table 3

This table has 13C Gln tracing data for WT and TRAP1 KO UMUC3 cells

Supplementary File 4: Table 4

This table has 13C Gln tracing data for WT and TRAP1 KO HCT116 cells

Supplementary file 5: Table 5

Reagents and resources

Authors’ Disclosures

No disclosures were reported.

Authors’ Contributions

A. Joshi: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Dai: Methodology. M. Maisiak: Methodology. S. Lee: Data curation, formal analysis, validation, methodology. E. Lopez: Methodology. T. Ito: Writing–review and editing. L. Neckers: Conceptualization, resources, data curation, supervision, funding acquisition, project administration, writing–review and editing.