S-nitrosylation of the Mitochondrial Chaperone TRAP1 Sensitizes Hepatocellular Carcinoma Cells to Inhibitors of Succinate Dehydrogenase

S-nitrosylation is a cysteine posttranslational modification emerging as the main modification underlying nitric oxide (NO) bioactivity (1). The extent of S-nitrosocysteines (SNO) depends on: (i) NO production by NO synthase (NOS); (ii) how NO is conveyed by S-nitrosylase complexes on target proteins (2); (iii) SNO removal by denitrosylases, among which S-nitrosoglutathione reductase (GSNOR) is the best-characterized example (3, 4). GSNOR catalyzes S-nitrosoglutathione (GSNO) reduction, thus indirectly controlling cellular levels of protein SNOs (PSNO; refs. 3, 5).

GSNOR-KO mice develop hepatocellular carcinoma (HCC) in association with S-nitrosylation and proteasomal degradation of the DNA repair enzyme O⁶-alkylguanine-DNA alkyltransferase (AGT; ref. 6). Studies performed on human patients with HCC showed a significant decrease of GSNOR protein levels and activity (7), arguing for a functional link between GSNOR-dependent S-nitrosylation and HCC existing also in humans.

We recently observed that skeletal muscle of GSNOR-KO mice show altered mitochondrial network (8), suggesting that excessive S-nitrosylation affects mitochondrial homeostasis. This is in line with the evidence arguing for mitochondria representing a preferential target of NO (9), and respiratory chain being susceptible to S-nitrosylation (10).

Here, we demonstrate that GSNOR deficiency induces mitochondrial complex II (succinate dehydrogenase, SDH) upregulation due to S-nitrosylation and subsequent degradation of the TNF receptor–associated protein 1 (TRAP1). TRAP1 is a mitochondrial chaperone: (i) suppressing mitochondrial ROS production (11, 12); (ii) inhibiting permeability transition pore opening; (iii) preserving mitochondrial bioenergetics (11, 13); and (iv) acting as SDH activity inhibitor (14, 15). We provide evidence that TRAP1 degradation by S-nitrosylation upregulates SDH and sensitizes GSNOR-deficient HCC cells to SDH inhibitors, paving the way for the development of novel chemotherapeutic approaches.

HepG2 cell line was obtained from the Banca Biologica e Cell Factory and grown in RPMI1640. Hepa1-6 and Huh-7-12 were purchased from Sigma and grown in DMEM. Media were supplemented with 10% FBS at 37°C in 5% CO₂. All cell lines were passaged for fewer than 2 months after resuscitation and were used from the third to the fifteenth passage in culture. Cell line validation was carried out by the manufacturer by means of DNA Profile STR (short tandem repeat). Cell lines were routinely screened for mycoplasma contamination by a PCR-based assay. Primary hepatocytes were obtained from GSNOR-KO and WT mice as described previously (16) and cultured in mammary epithelial cell growth medium (MEGM). Compounds and relative concentrations used in this study are as follows: 40 or 60 μmol/L α-TOS, 20 μmol/L 3BrPA, 5 mmol/L NAC, 200 μmol/L TROLOX, 100 μmol/L DHQ, 1 μmol/L 3AP, 1 mmol/L DTT, 50 μmol/L CHX, 500 μmol/L L-NAME (Sigma). 5 μmol/L DHE, 5 μmol/L H₂DCFDA and 20 μmol/L DAF-FM-DA (Life Technologies); 10 μmol/L MG132 (Merck Millipore); 0.5 μmol/L epoxomicin and 10 μmol/L necrostatin-1 (NEC-1; Santa Cruz Biotechnology); 20 μmol/L z-VAD (MP Biomedicals). Cells transfections, Western blot analysis, and qRT-PCR analysis are described in Supplementary Materials.

Dead cells were evaluated by direct cell count upon Trypan blue staining or by fluorescence microscopy upon staining with LIVE/DEAD Cell Imaging Kit (488/570; Life Technologies). Cell viability was determined following fluorescence emission at 590 nm after 4-hour incubation with AlamarBlue Reagent (Life Technologies). Apoptotic cells were evaluated upon staining with 50 μmol/L propidium iodide (Sigma) by a FACSCalibur flow cytometer (BD Biosciences; ref. 17). Apoptotic cells refer to the hypodiploid nuclei (sub-G₁) population. Clonogenic survival assay was performed as described previously (18). Electron and immunofluorescence microscopy analyses are described in Supplementary Materials.

Thirty minutes before the end of treatment with SDH inhibitors (6 hours), cells were incubated with DHE or H₂DCFDA at 37°C for the detection of superoxide or H₂O₂, respectively. Endogenous nitric oxide production was assessed in cells 24 hours after plating by 30-minute incubation with DAF-FM-DA. Fluorescence intensity was analyzed cytofluorometrically.

Succinate:coenzyme Q oxidoreductase activity was assayed spectrophotometrically in purified mitochondria after the reduction of 2,6-dichlorophenolindophenol, DCPIP (Sigma) as reported previously (19). Lactate dehydrogenase (LDH) activity was measured in cell media after the oxidation of NADH (Roche). GSNOR activity was assayed as described by Jensen and colleagues (5).

PSNOs were evaluated by biotin switch assay as described previously (8, 20). To identify TRAP1 S-nitrosylated sites, human recombinant TRAP1 (ADI-SPP-848; Enzo Life Sciences) was incubated with 2 mmol/L DTT for 1 hour, washed with PBS, and then maintained for 2 hours with 5 mmol/L DPTA-NONOate (Sigma). As control, equal amount of protein was left in DTT. Proteins were then subjected to biotin switch assay and digested with endopeptidase LysC (Wako) followed by trypsin (sequencing grade, Promega; ref. 21). Resulted peptides were analyzed by nanoLC-MS/MS as in ref. 22 and measurements were performed on an EASY-nLC system (Proxeon Biosystems) connected to the Q-Exactive mass spectrometer (Thermo Scientific) equipped with a nanoelectrospray ion source (Proxeon Biosystems). The Q-Exactive was operated in data-dependent mode with the MS survey scan ranging from 300 to 1,750 m/z and a resolution of 70,000. The 12 most abundant ions with charge ≥ 2+ from each survey scan were sequentially isolated (isolation window of 2 m/z) and subjected to MS/MS fragmentation by high-energy collisional dissociation (HCD; ref. 23). Raw data were processed with MaxQuant software suite version 1.3.0.5 searching peak lists agains UniProt human database. Methylthiolation and HPDP-Biotinylation were manually configured as variable peptide modifications. Maximum false discovery rates were set to 0.01 for proteins, peptides, and modification sites.

Mouse experiments were conducted at the mouse facility of the Danish Cancer Society Research Center with the approval of the Regional Committee for Medical Research Ethics in Denmark. GSNOR-KO mice were generated in J.S. Stamler's laboratory on C57BL/6J background (4). Livers were collected from 2-month-old mice and used for Western blot analyses or processed to evaluate mitochondrial oxygen consumption. Xenograft experiments were conducted on NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (SCID) mice strains. Tumors were generated by subcutaneous injection, into the right lower flank, of 5 × 10⁶ HepG2 cells stably downregulating GSNOR (n = 16) or scrambled counterparts (n = 16). When palpable tumors were recognized, each group was split into α-TOS–treated (n = 10) and vehicle-treated (n = 6) groups. α-TOS was dissolved in Corn oil/4% ethanol as reported previously (24) and used at 5 μmol (2 injections/week). Treatments started when tumor diameter reached 5 mm, and tumor size was measured twice per week for 2 weeks.

ATP levels were measured by the ATP Bioluminescence Assay Kit CLS II (Roche) using a microplate luminometer (PerkinElmer).

Total mitochondrial mass and Δψ_m were analyzed cytofluorometrically by incubating the cells with 50 nmol/L MitoTracker Green-FM and 200 nmol/L MitoTracker Red CMXRos (Life Technologies), respectively. Alternatively, Δψ_m was evaluated by fluorescence microscopy upon JC-1 staining (Life Technologies; ref. 8).

Mitochondria were isolated and oxygen consumption determined as described previously (25) in intact mitochondria, or in mitochondrial membranes obtained upon osmotic lysis in water as previously reported (26). P:O ratio calculated in intact mitochondria of WT and GSNOR-KO livers was 2.34 ± 0.13 versus 2.09 ± 0.13, respectively.

Protein concentration was determined by the method of Lowry and colleagues (27).

Values are expressed as means ± SD (or SEM when indicated). The statistical significance of the differences between means was assessed by independent Student t test or one-way ANOVA test using GraphPad Prism.

GSNOR-KO mice spontaneously develop HCC (7) and show mitochondrial fragmentation in skeletal muscle (8). On the basis of this, we investigated whether mitochondria of liver obtained from 2-month-old GSNOR-KO mice (which did show no sign of HCC hitherto) exhibited defects in mitochondrial respiration and function. Respiration, measured upon malate/glutamate addition (involving complex I), was significantly dampened in GSNOR-KO mitochondria, whereas respiration measured upon succinate addition (starting from complex II) was about 40% higher than the WT counterparts, notwithstanding complex IV efficiency was decreased (Fig. 1A, left). To untie complex I and II activities from complex IV oxygen consumption, we repeated the analyses in mitochondrial membranes obtained upon osmotic lysis. Figure 1A (right) shows no difference of complex I activity between the two genotypes. In contrast, complex II activity was even more increased in GSNOR-KO samples, suggesting that this effect was masked in intact mitochondria due to the downstream inhibition of complex IV. These results were strengthened by Western blot analysis of complex IV (subunit I), which were reduced in GSNOR-KO liver extracts, and complex II (as revealed by its subunit A) that, conversely, significantly increased (Fig. 1B).

To investigate the effects of GSNOR deficiency in the context of HCC, we silenced GSNOR in the human HCC cell line HepG2 (siGSNOR HepG2) with two different pools of siRNAs (Fig. 1C and D), inducing 2-fold increase in PSNOs content (Fig. 1E). However, they did not affect NOS isoform expression pattern (with nNOS being undetectable, see Fig. 1C), or alter NO production (Fig. 1F), indicating that changes in S-nitrosylation were independent on NO synthesis.

In line with results obtained in liver extracts, SDHA levels and Complex II activity were upregulated in GSNOR-silenced cells (Fig. 1G and H). However, mitochondrial proteins and mass (Supplementary Fig. S1A and S1B) and gene expression of all four SDH subunits (Supplementary Fig. S1C) did not change, indicating that SDH upregulation was not due to any transcriptional regulation or change in the number of mitochondria. We also observed that ATP levels tend to decrease (Supplementary Fig. S1D and S1E), whereas Δψ_m significantly dropped with respect to siScr cells (Supplementary Fig. S1F and S1G).

To unveil whether SDH upregulation upon GSNOR depletion was related to S-nitrosylation, we focused on the mitochondrial chaperone TRAP1, which has recently been found to affect SDH activity (14). Western blot analysis revealed that TRAP1 protein levels were reduced in GSNOR-depleted cells (Fig. 2A; Supplementary Table S3), despite no change in mRNA was detected (Supplementary Fig. S2A). We then hypothesized that TRAP1 could be destabilized and undergo degradation owing to S-nitrosylation, with this assumption strengthening by the evidence that: (i) incubations with the proteasome inhibitor MG132 completely restored TRAP1 levels in siGSNOR cells (Fig. 2B) and (ii) pretreatment with the NOS inhibitor Nω-nitro-l-arginine methyl ester (L-NAME) prevented TRAP1 decrease and SDHA upregulation (Supplementary Fig. S2B). We then evaluated TRAP1 protein levels by Western blot analysis on a pool of PSNOs obtained by biotin switch assays. Results shown in Fig. 2C indicate that S-nitrosylated TRAP1 accumulated upon proteasome inhibition only in siGSNOR cells, arguing for S-nitrosylation being associated with TRAP1-selective degradation. Consistently, siGSNOR or siTRAP1 HepG2 cells showed the same expression pattern as SDHA and TRAP1, with the former upregulated and the latter decreased (Fig. 2D). Moreover, TRAP1 overexpression was able to downregulate SDH (Fig. 2E) and to accelerate its turnover in the presence of the protein translation inhibitor cycloheximide (Fig. 2F).

The possibility that TRAP1 could be degraded via the proteasome has been suggested (28). However, mitochondrial proteins are usually targets of specialized mitochondrial proteases (29). Therefore, to clarify whether S-nitrosylation–induced degradation of TRAP1 was mediated by the proteasome, we purified cytosolic and mitochondrial fractions of HepG2 cells and mouse livers, confirming that TRAP1 significantly decreased only in mitochondria of GSNOR-deficient systems (Fig. 2G and H). Moreover, incubation with MG132 showed that TRAP1 accumulated exclusively in the cytosol of siGSNOR cells (Fig. 2G), where the levels of ubiquitylated TRAP1 increased even in the absence of MG132 (Supplementary Fig. S2C). The evidence that the highly selective proteasome inhibitor epoxomicin produced the same effects of MG132 (Fig. 2I) reinforced the hypothesis that TRAP1 degradation occurred via the proteasome. In agreement, incubations with cycloheximide resulted in a more rapid turnover of TRAP1 in siGSNOR cells (Fig. 2J).

To identify the cysteine residue(s) being modified by NO, we treated recombinant human TRAP1 with DETA-NONOate and performed a biotin switch assay, followed by analysis with high accuracy mass spectrometry (Fig. 2K). The mass spectrometric measurements unambiguously identified TRAP1 peptide bearing biotinylation at Cys⁵⁰¹ residue, implementing Cys⁵⁰¹ as S-nitrosylation site (Fig. 2L), and confirming previous high-throughput proteomic studies (30). On the basis of these results, we generated a human TRAP1-C501S mutant and induced its expression in siGSNOR cells. Western blots, obtained upon incubation with cycloheximide, indicate that TRAP1-C501S was more stable (Fig. 2M), confirming that S-nitrosylation of Cys⁵⁰¹ prepares the way for TRAP1 degradation.

To understand whether Cys-to-Ser substitution was still able to maintain TRAP1 functional relationship with SDH, we overexpressed TRAP1-C501S mutant in parental HepG2. Under these conditions, SDH protein levels were reduced (Fig. 2N) and its turnover accordingly decreased (Fig. 2O). In agreement, TRAP1-C501S was able to revert the increase of Complex II activity observed in siGSNOR cells (Fig. 2P).

We attempted to exploit the unusual TRAP1 decrease and take advantage by SDH upregulation to selectively target HCC cells. We analyzed the occurrence of cell death upon treatment with two different SDH inhibitors listed as mitocans, namely: 3-bromopyruvate (3-BrPA), which alkylates SDH sulfhydryl groups (31), and α-tocopheryl succinate (α-TOS), which specifically binds to ubiquinone binding site of SDH (32). As negative control, we used malonate, the physiologic SDH inhibitor, at a concentration of 20 mmol/L to abolish Complex II activity. Direct cell counts showed that siGSNOR cells were insensitive to malonate, but more susceptible to treatments with α-TOS and 3-BrPA used at concentrations in a range that provides a similar extent of cell death (Supplementary Fig. S3A). Further cell viability and clonogenic survival assays performed with α-TOS produced similar results, confirming that GSNOR downregulation sensitized HepG2 cells to α-TOS treatment (Fig. 3A and B).

Oxidative stress is one of the main effects induced by mitocans, thus contributing to their antitumor activity (24, 33). Therefore, we measured superoxide and H₂O₂ levels and found that both increased significantly when GSNOR was downregulated (Fig. 4C and D). Interestingly, when we preincubated the cells with the ROS scavenger TROLOX, as well as when we ectopically overexpressed either mitochondrial superoxide dismutase (SOD2) or catalase fused with a mitochondrial import sequence (mitoCAT; Supplementary Fig. S3C and S3D), ROS production was inhibited and protein oxidative damage was prevented (Fig. 3C–E). This was associated with a complete recovery of cell viability (Fig. 3F and G and Supplementary Fig. S3B). We then transfected HepG2 cells with siRNA against GSNOR in combination with siRNAs against each SDH subunit (SDH A–D). In these conditions, superoxide production was prevented and α-TOS cytotoxicity was completely abolished (Supplementary Fig. S3E and S3F), confirming the selectivity of α-TOS towards mitochondrial complex II. Moreover, TRAP1-C501S rendered siGSNOR cells insensitive to α-TOS toxicity (Fig. 3H and I) and abolished superoxide production (Supplementary Fig. S3G), implying that TRAP1 S-nitrosylation is required for α-TOS toxicity. In agreement, TRAP1 silencing enhanced α-TOS vulnerability of parental HepG2 cells (Supplementary Fig. S3H).

By looking at the molecular mechanisms underlying the cell death process observed, we analyzed the possible occurrence of apoptosis. Western blot analyses indicated that cleavage of caspase-9 and caspase-3, as well as proteolysis of PARP1, were not induced in α-TOS–treated siGSNOR cells (Fig. 4A). No increase of apoptotic cells was observed (Supplementary Fig. S4A), and no rescue of cell viability was achieved upon the inhibition of caspases (Supplementary Fig. S4B). Subcellular localization analyses of the apoptosis inducing factor 1 (AIF1) also excluded a caspase-independent apoptosis (Supplementary Fig. S4C). Conversely, we observed an increase of LDH release in the extracellular medium (Fig. 4B), suggesting that treatment with α-TOS reasonably induced necrosis in siGSNOR cells.

Several findings have reported that caspases undergo S-nitrosylation and inactivation (34–36). Consistently, we observed that caspase-3 underwent massive S-nitrosylation in siGSNOR cells (Supplementary Fig. S4D), suggesting that apoptosis, even though induced, could be inhibited by excessive S-nitrosylation. To confirm this hypothesis, we preincubated the cells with the thiol-reducing agent dithiotreitol (DTT) and treated them with higher concentration of α-TOS to also induce a significant extent of cell death in siScr cells. DTT significantly switched cell death response in siGSNOR cells from full necrosis to apoptosis at the same extent measured in siScr cells (Supplementary Fig. S4E), thus arguing for the denitrosylating effect of DTT being entirely compensatory of GSNOR deficiency. In agreement, PARP1 was cleaved, signaling the effective accomplishment of the apoptotic program (Supplementary Fig. S4F).

In the search for the molecular determinants of mitocan-induced necrosis, we pondered the evidence that full-length PARP1 seemed actually to be induced upon α-TOS treatment (see Fig. 4A). Consistently, polyADP-ribosylated (pADPr) protein levels also increased in siGSNOR cells, even in the absence of α-TOS (Supplementary Fig. S4G). Long-lasting PARP1 hyperactivation results in necrosis (37). Therefore, we investigated whether PARP1 could have a role in α-TOS–induced cell death. Figure 4C and D indicate that the extent of dead cells significantly decreased by inhibiting PARP1 activity either genetically (siRNA against PARP1), or pharmacologically (pretreatments with PARP1 inhibitors).

We then verified whether PARP1 triggered programmed necrosis by the activation of the receptor-activating protein 1 (RIP1; ref. 38). Particularly, we observed that RIP1 was induced in siGSNOR cells treated with α-TOS (Fig. 4E) and that preincubation with the RIP1 inhibitor NEC-1 reverted cell death (Fig. 4F). Moreover, PARP1 inhibition reduced RIP1 levels (Fig. 4G), confirming the crucial role of PARP1/RIP1 signaling axis in transducing death signals under nitrosative stress. Indeed, both NEC-1 and DTT restored clonogenic survival of siGSNOR cells treated with α-TOS (Fig. 4H). To widen the biologic significance of our findings, we also performed experiments with 3-BrPa obtaining similar results (Supplementary Fig. S4H).

To finally confirm the occurrence of necroptosis, we analyzed the cell ultrastructure through electron microscopy (39). siGSNOR cells exposed to α-TOS generally display mitochondrial swelling and damage, whose severity increased in a dose-dependent manner (Fig. 4I). Particularly, upon treatment with 40 μmol/L α-TOS, the normal structure of inner mitochondrial cristae appeared almost completely lost. Nuclei were still intact but swollen and cytoplasmic volume was substantially reduced. Conversely, in siScr cells, α-TOS induced changes in subcellular structures only at the highest concentrations used (60 μmol/L), with many of them resembling the canonical features of apoptosis (Supplementary Fig. S4I).

We next tested α-TOS effects on cell viability of other two different HCC cell lines, HuH7 and Hepa1-6, along with primary murine hepatocytes (Fig. 5A). α-TOS did not affect cell viability of hepatocytes as previously reported (40), but induced only a mild cytostatic effect above 80 μmol/L, with this being independent on GSNOR expression (Fig. 5B and C). In contrast, α-TOS induced cell death in HCC cells, although to different extents. In particular, Hepa1-6 were less susceptible, displaying an EC₅₀ = 175.04 μmol/L versus 74.07 and 60.06 μmol/L of HuH7 and HepG2, respectively. Furthermore, GSNOR downregulation did not enhance α-TOS toxicity in Hepa1-6 cells (Fig. 5D). Going into depth on this different response, we found that Hepa1-6 cells displayed a halved mitochondrial content and 8-fold decrease of SDH activity (Supplementary Fig. S5A and Fig. 5E), confirming the role of mitochondria as elective target for HCC selective killing. On the other hand, GSNOR-downregulating HuH7 cells recapitulated what previously observed in HepG2. In particular, GSNOR deficiency resulted in TRAP1 downregulation and SDH increase (Fig. 5F), condition enhancing α-TOS toxicity (Fig. 5G and H). Cell death was reverted by preincubation with DTT and NEC-1, as well as by overexpressing TRAP1-C501S mutant (Fig. 5I). Similarly, α-TOS induced cell death by means of massive superoxide and H₂O₂ production that were dampened upon preincubation with TROLOX or by overexpression of SOD2 and mitoCAT (Fig. 5K and L and Supplementary Fig. S5B and S5C).

We next moved to HepG2 xenografts implanted in SCID mice, and evaluated α-TOS antitumor efficacy. We generated HepG2 cell lines stably expressing shRNAs against GSNOR (shGSNOR) (Fig. 6A). Then, we selected two clones showing TRAP1 downregulation, SDHA increase (Fig. 6B), and sensitivity to α-TOS cytotoxicity (Fig. 6C) to exclude any cellular compensatory mechanism(s) emerging during clonal selection. We injected subcutaneously shGSNOR or control shSrc HepG2 cells and once tumor masses reached 5 mm diameter, we treated the mice for 2 weeks by intraperitoneal administrations of α-TOS, provided at low doses (5 μmol each three days). Results shown in Fig. 6D and E indicate that volumes, as well as final weights and size, were significantly reduced in shGSNOR HCC xenografts with respect to the vehicle-treated counterparts, whereas no significant difference was observed in shScr xenografts. shGSNOR xenografts treated with α-TOS showed increased protein carbonylation (Fig. 6F), and induction of RIP1 expression (Fig. 6G), confirming that α-TOS susceptibility was due to necroptosis-like cell death induced by oxidative stress.

Conflicting lines of evidence argue that NO can carry out anti- or protumor activities, as well as having a still not clarified role in tumorigenesis. HCC is a case in point, as iNOS is often increased in pathologies predisposing to HCC (41) but iNOS^−/− mice revealed only very little effects of NO in hepatocarcinogenesis (42). In reconciling these apparently contradictory results, several pieces of evidence about the establishment of excessive S-nitrosylation in HCC due to GSNOR downregulation are accumulating (6). Moreover, human GSNOR gene is located at approximately 4q23, a region frequently subjected to chromosomal deletion in HCC (43–45) and GSNOR downregulation has been found associated with de novo hepatocarcinogenesis after tumor resection and a poor prognosis in patients with HCC (46). Finally, meta-analyses present in the Oncomine platform (Supplementary Fig. S6A) and quantitative evaluation of human HCC samples obtained by The Human Protein Atlas collection by IHC profiler (Supplementary Fig. S6B; ref. 47) show a direct correlation between GSNOR underexpression and HCC.

Here, we have characterized that GSNOR downregulation induces a rearrangement of the mitochondrial electron transfer chain. Although the downregulation of complex IV was already reported as being negatively affected by S-nitrosylation, and, for this reason, expected to occur in conditions of GSNOR deficiency, the evidence that SDH is upregulated is new and noteworthy, and may constitute a mechanism of cell adaptation. In support of this hypothesis, it should be noticed that, notwithstanding the fact that GSNOR-downregulating cells show a decreased Δψ, ATP levels are relatively preserved, indicating that a metabolic reprogramming is nevertheless activated to sustain cell energetic needs.

From a mechanistic point of view, we have demonstrated that TRAP1 S-nitrosylation at Cys⁵⁰¹ is the causative event underlying the increase of SDH levels and activity. TRAP1 is essential for mitochondrial protein quality control (48, 49). As it preferentially accumulates in the matrix (50), it is plausible that TRAP1 could interact with SDH, mainly with subunits A and B that are exposed to this mitochondrial compartment, and, in turn, regulate SDH turnover. This hypothesis would support our data, and would also partly agree with recent results highlighting a negative correlation between TRAP1 and SDH activity (14). However, the role of TRAP1 downregulation that we have here outlined as responsible for HCC sensitivity to mitochondrial drugs, to the best of our knowledge does not represent a condition concurring to HCC onset. This would be inconsistent with the current literature that, in contrast, indicates that TRAP1 is upregulated in several cancers (51, 52). We do not currently know the kinetics of TRAP1 modulation during the complex process of hepatocarcinogenesis. We only indicate that at least in the GSNOR-downregulating subtype of HCC, TRAP1 is target of S-nitrosylation and subjected to accelerated degradation. This condition, resulting in SDH upregulation, makes it possible to envision that SDH inhibitors might selectively target this HCC subtype.

It is worthwhile noting that excessive S-nitrosylation, because of GSNOR deficiency, also has an effect on the modality of cell death induced by SDH inhibitors, with the RIP1-mediated necroptosis activated in place of apoptosis. This depends both on the high levels of ROS produced and on the incapability of GSNOR-downregulating HCC cells to correctly activate caspase-3, as caspases undergo S-nitrosylation on catalytic cysteine and inactivation of their proteolytic activity (34, 35). The incapability to correctly activate apoptosis could be, in principle, another factor that contributes to hepatocarcinogenesis in conditions of GSNOR deficiency.

Altogether, if on the one hand, our results provide the molecular basis for designing highly selective SDH-targeting drugs aimed at improving chemotherapy against HCC, on the one other they also set the stage for comprehending a redox-based regulation of TRAP1 stability.

J.S. Stamler reports receiving a commercial research grant from Novartis and has ownership interest (including patents) in Adamas and Lifehealth. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Rizza, G. Filomeni

Development of methodology: B. Blagoev, J.S. Stamler

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Rizza, C. Montagna, E. Maiani, V. Sanchez-Quiles, B. Blagoev, D.D. Zio

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Rizza, V. Sanchez-Quiles, B. Blagoev, A. Rasola, G. Filomeni

Writing, review, and/or revision of the manuscript: S. Rizza, S. Cardaci, B. Blagoev, A. Rasola, J.S. Stamler, F. Cecconi, G. Filomeni

Study supervision: B. Blagoev, F. Cecconi, G. Filomeni

Other (performed oxygen consumption measurements): G.D. Giacomo

Other (mass spectrometry analysis): V. Sanchez-Quiles

The authors thank Vanda Turcanova for laboratory assistance; Petra Hamerlik and Elena Papaleo for assistance and discussion; and Magdalena Acuña Villa, Laila Fisher, and Martin W. Bennett for secretarial and editorial work.

This work has been supported by grants from the Danish Cancer Society (KBVU R72-A4408 to F. Cecconi; KBVU R72-A4647 to G. Filomeni); Danish Research Foundation (CARD); the Italian Association for Cancer Research, (AIRC-MFAG 2011 n.11452 to G. Filomeni; IG2013 to F. Cecconi; IG2014 n.15863 to A. Rasola); the Italian Ministry of Health (GR-2008-1138121 to G. Filomeni). We are also grateful to the Bjarne Saxhof Foundation, Lundbeck Foundation (R167-2013-16100 and R52-A4895 to F. Cecconi), NovoNordisk (7559 to F. Cecconi), FNU (DFF – 1323-00191 to B. Blagoev).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.