Piperlongumine treatment inactivates peroxiredoxin 4, exacerbates endoplasmic reticulum stress, and preferentially kills high-grade glioma cells

Abstract

Backgrounds

Piperlongumine, a natural plant product, kills multiple cancer types with little effect on normal cells. Piperlongumine raises intracellular levels of reactive oxygen species (ROS), a phenomenon that may underlie the cancer-cell killing. Although these findings suggest that piperlongumine could be useful for treating cancers, the mechanism by which the drug selectively kills cancer cells remains unknown.

Methods

We treated multiple high-grade glioma (HGG) sphere cultures with piperlongumine and assessed its effects on ROS and cell-growth levels as well as changes in downstream signaling. We also examined the levels of putative piperlongumine targets and their roles in HGG cell growth.

Results

Piperlongumine treatment increased ROS levels and preferentially killed HGG cells with little effect in normal brain cells. Piperlongumine reportedly increases ROS levels after interactions with several redox regulators. We found that HGG cells expressed higher levels of the putative piperlongumine targets than did normal neural stem cells (NSCs). Furthermore, piperlongumine treatment in HGG cells, but not in normal NSCs, increased oxidative inactivation of peroxiredoxin 4 (PRDX4), an ROS-reducing enzyme that is overexpressed in HGGs and facilitates proper protein folding in the endoplasmic reticulum (ER). Moreover, piperlongumine exacerbated intracellular ER stress, an effect that was mimicked by suppressing PRDX4 expression.

Conclusions

Our results reveal that the mechanism by which piperlongumine preferentially kills HGG cells involves PRDX4 inactivation, thereby inducing ER stress. Therefore, piperlongumine treatment could be considered as a novel therapeutic option for HGG treatment.

High-grade glioma (HGG) is the most common primary brain malignancy and has a dismal prognosis.¹ For instance, glioblastoma is the most common and aggressive HGG and has a median patient survival of 14.6 months, which has increased only 2.5 months in the last 2.5 decades despite vigorous efforts in clinics and laboratories. The best combination of current cancer therapies marginally prolongs patient survival but is not curative. Therefore, innovative cancer therapies are urgently needed.

We discovered that most glioblastomas overexpress the reactive oxygen species (ROS)-degrading enzyme peroxiredoxin 4 (PRDX4).² We also found that PRDX4 knockdown increases ROS levels, kills glioblastoma cells in vitro, and prolongs mouse survival in an orthotopic model with decreased tumor size.² These results suggest that pharmaceutical PRDX4 inactivation effectively kills the cancer cells by increasing ROS levels.

Increasing ROS levels via piperlongumine treatment selectively kills multiple cancer types including melanomas, lymphomas, and glioblastomas.^3–5 Piperlongumine is a naturally occurring alkaloid present in the plant Long pepper.⁶ Importantly, piperlongumine treatment selectively kills cancer cells without substantially affecting growth of normal counterpart cells.^3,5 The drug treatment increases ROS levels several folds in cancer cells but not in normal cells,³ which may underlie piperlongumine's selective cancer cell-killing action. While piperlongumine has been known to interact with several antioxidant proteins, including PRDX1,³ it remains unknown how the drug treatment selectively increases ROS levels and kills cancer cells. In addition, a recent report has shown that piperlongumine treatment selectively kills glioblastoma cells cultured in a serum-containing condition.⁵ However, it is unknown whether the drug treatment also kills glioblastoma or HGG cells grown as spheres, which more closely resemble in situ tumors in terms of genetic mutations and biological properties than serum-containing adherent cultures.⁷ In addition, such sphere cultures enrich glioma stem/initiating cells,⁸ which have been considered to critically contribute to malignant characteristics and therapeutic responses of HGGs.^8,9

In the present study, we found that piperlongumine treatment increased ROS levels and preferentially killed HGG cells with little effect on normal brain cells. Mechanistically, we showed that piperlongumine treatment inactivated PRDX4 and exacerbated endoplasmic reticulum (ER) stress in HGG cells.

Materials and Methods

HGG and Primary Neural Cell Cultures

We used human patient-derived HGG sphere cultures that had been previously established and maintained under the protocols approved by the Institutional Review Board.⁸ We established and maintained sphere cultures of neural stem cells (NSCs) from the subventricular zone of wild-type (wt) mice and of HGGs from a mouse genetic model.^2,10 Detailed characterizations of the sphere cultures are described in our reports.^2,8,10 We established primary cultures of granule neurons from the cerebellum of postnatal days 6–7 (P6–7) mice and of astrocytes from the cortex of P1–2 mice, as described.^11,12

Drug Treatment and Cell Growth Assay

One day after plating 5.0 × 10³ of cells per well in 96-well plates or 3.0 × 10⁴ cells per well in 24-well plates, we treated them with 0–10 μM of piperlongumine (BioVision) dissolved in dimethyl sulfoxide (0.1% v/v) and incubated for 1–3 days. We assessed cell growth by 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) cell proliferation assay or direct cell counting, using the Vi-cell Viability Analyzer (Beckman Coulter). We previously reported that these assays resulted in similar data.² For MTS assay, we used the CellTiter 96 Aqueous One Solution Cell Proliferation Assay kit (Promega) and measured optical density at 490 nm using the Multiskan Spectrum 96-well plate reader (Thermo Fisher Scientific).

Proxiredoxin 4 Activity Assay

We assessed PRDX enzyme activity using the 2-Cys-Peroxiredoxin Activity Assay kit (Redoxica). Briefly, we incubated 30 μg of cell lysate or 3 μg of purified recombinant human PRDX4 (ProSpec) with thioredoxin, thioredoxin reductase, and NADPH. We added 10 mM of hydrogen peroxide to initiate the enzymatic reaction and measured NADPH consumption by spectrophotometry at 340 nm 60, 90, and 120 seconds after the addition of hydrogen peroxide.

Statistical Analysis

We used the 2-sample Student t test to analyze the data including 2 groups. For the experiments with more than 2 groups, analysis of variance (ANOVA) was used, and Holm's procedure was applied to adjust multiplicity to control the overall family-wise error rate at α = 0.05.¹³ Data were analyzed by SAS 9.3 (SAS Institute Inc.) and GraphPad Prism (GraphPad Software, Inc.). We displayed all data as mean ± standard error of the mean.

All of the other and further materials and methods are described in Supplementary Methods.

Results

Piperlongumine Treatment Increases ROS Levels and Preferentially Kills HGG Cells

To determine the effect of piperlongumine on a more physiologically relevant, in vitro HGG model, we used sphere cultures in a majority of experiments. We used 2 patient-derived human HGG sphere cultures⁸ as well as 2 mouse HGG sphere cultures derived from an Nf1: Trp53: Pten mouse genetic model.^2,10 The human HGG sphere cultures belonged to the more aggressive and therapy-resistant mesenchymal HGG subtype⁸ and showed high levels of mesenchymal stemness markers, including CD44 and WT1 (Supplementary Fig. 1). By showing that NSC-specific deletion of Nf1, Trp53, and Pten tumor suppressor genes induced HGGs, we demonstrated that NSC lineage is a cellular origin of the brain tumor.¹⁴ Incorporating this finding into the current work, we used NSC sphere cultures derived from wt mice^10,14 as a control for HGGs. We also used primary cultures of 2 major brain cell types, neurons and astrocytes, as additional controls.

We found that piperlongumine treatment preferentially suppressed growth of all 4 HGG sphere cultures tested in a dose-dependent manner (Fig. 1). For instance, 10 μM of piperlongumine suppressed the growth of HGG cultures by 60%–90% upon 3 days of treatment. By contrast, piperlongumine treatment did not suppress neuronal growth and had little effect on NSC growth (Fig. 1B and C). However, at the highest dose of 10 μM, piperlongumine treatment suppressed the growth of primary astrocyte culture by 58%.

Fig. 1.

Effect of piperlongumine (PL) treatment on the growth of HGG and normal brain cells. (A) Representative mouse HGG spheres 3 days after PL treatment (10 μM). Scale bar = 100 μm. (B) 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium cell-growth assay data upon PL treatment in HGG and normal brain cell cultures. (C) Relative growth of mouse HGG (ID: 7080), wt NSC (ID: 19740), neuron, and astrocyte cultures 3 days after PL treatment. ** P < .005 compared with vehicle. (A–C) Representative data from 2 independent experiments (with n = 5 replicates).

We next examined whether the killing effect of piperlongumine on HGG cells and astrocytes involved increasing ROS levels, as it does in other cancer types.³ ROS levels, assessed by the ROS-sensitive dye CellROX Deep Red reagent,¹⁵ were increased by 35%–65% in all 4 HGG cultures and by 23% in the astrocyte culture after piperlongumine treatment (Fig. 2A). In concurrence with cell-growth data (Fig. 1B and C), the same dose of drug treatment did not increase ROS levels in wt NSC cultures. This result suggested that the cell-killing effect of piperlongumine is attributable to increases in ROS levels. Astrocytes are sensitive to ROS-mediated oxidative stress.^16,17 Therefore, we mainly used the piperlongumine-insensitive wt NSCs as a control for HGG cells in the remaining studies.

Fig. 2.

Piperlongumine (PL) treatment increases ROS and apoptosis levels in HGG spheres. (A) Left: flow cytometric analysis after incubation with ROS-sensitive dye CellROX Deep Red reagent represents fluorescent ROS levels in mouse HGG (ID: 6989), a wt NSC (ID: 19739), and an astrocyte culture12 hours after PL treatment (10 μM). Middle: summary of flow cytometric analyses compares ROS levels in mouse HGG (ID: 7080), a wt NSC, and an astrocyte culture after PL treatment. * P < .05, ** P < .005. Right: summary of relative ROS levels in wt NSC, HGG, and astrocyte cultures after PL treatment. (B) Left: representative flow cytometric analysis shows relative numbers of Annexin V-positive mouse HGG cells (ID: 6989) 24 hours after PL treatment. Middle: summary of 3 flow cytometric analyses compares relative numbers of Annexin V-positive mouse HGG cells (sum of propidium iodide-positive late and propidium iodide-negative early apoptotic cells) after PL treatment. Right: summary of Annexin V labeling analysis on HGG cultures after PL treatment. (C) Left: summary of 3 ROS assays as in A compares relative ROS levels in mouse HGG culture (ID: 7080) 12 hours after PL treatment (10 μM), antioxidant NAC (5 mM), or both. Right: representative cell counting data 2 days after the drug treatments in the HGG culture. * P < .05. (A–C) Representative or summarized data from experiments with n = 3–4 replicates per group.

Since excess ROS levels can be cytotoxic,² we examined apoptosis level in the HGG cultures by Annexin V staining and subsequent fluorescence-activated cell sorting analysis. Indeed, piperlongumine treatment increased the number of Annexin-V-positive cells by 82%–376% in all 4 HGG cultures (Fig. 2B). Treatment with an antioxidant N-acetyl cysteine (NAC) at 5 mM significantly reduced ROS levels in the presence of piperlongumine and partially rescued the piperlongumine-induced HGG growth suppression (Fig. 2C). Therefore, these data indicated that piperlongumine treatment induces apoptotic death of HGG cells in part by increasing ROS levels, as in other cancer cells.³

Piperlongumine Inactivates PRDX4 in HGG Cells

Piperlongumine interacts directly with at least a dozen different proteins, as shown by a quantitative proteomics approach combined with affinity isotope labeling in EJ and U2OS cell lines.³ Interestingly, 7 of the piperlongumine-interacting proteins are known to regulate intracellular ROS levels. To see if that is the case in HGG cells, we compared gene expression levels of the piperlongumine-interacting ROS regulators between HGG and normal brain cells. Our quantitative PCR (qPCR) analysis revealed that mRNA levels of 4 out of the 7 piperlongumine-interacting proteins were 1.5–2.0-fold higher in mouse HGG spheres than in wt NSCs (Fig. 3A). Similar to our findings from the mouse cells, analysis of the Repository of Molecular Brain Neoplasia DATA (REMBRANDT) database revealed that mRNA levels of 5 of the 7 piperlongumine-interacting proteins were 1.3–2.0-fold higher in human glioblastomas than in normal brain tissues (Fig. 3B). These data suggested that increased expression of the piperlongumine-interacting ROS regulators in HGG cells underlies the selective killing of the cancer cells by piperlongumine treatment.

Fig. 3.

Piperlongumine (PL) treatment inactivates PRDX4 in HGG spheres. (A) Representative qPCR analysis on PL-interacting ROS regulators between mouse HGG and NSC sphere cultures (n = 4 cultures). * P <.05. (B) A heat map of expressions of the PL-interacting ROS regulators in human glioblastomas in comparison with normal brain tissues. These data are from analysis of the REMBRANDT database. (C) qPCR analysis on Prdx family members in mouse HGG versus wt NSC sphere cultures. (D) Western blotting (WB) analysis for oxidized forms of Prdx (Oxi-Prdx) 12 hours after PL treatment (10 μM) in 2 mouse HGG sphere cultures. A portion of the cell lysates was immunoprecipitated (IP) with PRDX4 antibody before WB. Band densities (target/loading control) relative to that of vehicle (Veh) are shown below Western blotting data. (*) and (#): Lower and higher exposures of Prdx4 and Prdx1/2/4 of the second blot, respectively, for more accurate analysis. (E) PRDX enzymatic activity measured by incubation of cell lysate or purified recombinant PRDX4 protein with hydrogen peroxide and assessing NADPH consumption. The cell lysate (30 μg) was from mouse HGG cultures (6989 and 7080) 12 hours after PL treatment (10 μM). PRDX4 protein (3 μg) was treated with 50 μM of PL for 30 minutes. * P < .05. (F) Western blotting analysis 12 hours after PL treatment (10 μM) in mouse wt NSC and mouse HGG cultures. NAC (5 mM) was treated 1 hour before PL treatment. Band densities relative to that of vehicle (Veh) treatment in wt NSCs, except for oxi-Prdx1/2/4 (relative to Veh treatment in HGG cells), are shown below Western blotting data. (A, C–F) Representative data from 2 independent experiments.

Overexpression of 2 putative piperlongumine target genes, carbonyl reductase 1 (CBR1) and glutathione S-transferase pi 1 (GSTP1), in EJ cells appeared to reduce piperlongumine-mediated ROS elevation and apoptosis, while knocking down the genes had no apparent effect on ROS or apoptosis levels.³ Among the putative piperlongumine target proteins, CBR1 and GSTZ1 were highly expressed in both mouse HGGs and human glioblastomas (Fig. 3A and B). To determine whether these 2 genes were necessary for piperlongumine-mediated apoptosis of HGG cells, we combined piperlongumine treatment with gene knockdown in U251 and U373 glioblastoma cell lines. In concurrence with the published report,³ knocking down CBR1 and GSTZ1 expression did not change the growth suppression by piperlongumine treatment (Supplementary Fig. 2), even though piperlongumine treatment effectively suppressed growth of the glioblastoma cells. These results suggested that piperlongumine-mediated growth suppression in HGG cells occurs via a different target.

In looking for the HGG target(s) responsible for piperlongumine-induced cytotoxicity, we noted that piperlongumine treatment mimicked the effects of PRDX4 knockdown in HGG; suppression of HGG cell growth with increases in ROS levels.² Although piperlongumine has been known to interact with PRDX1 in both EJ and U2OS cell lines,³ only the Prdx4 level was highly increased among 6 Prdx family members in mouse HGG cells compared with wt NSCs (Fig. 3C). In addition, knocking down PRDX1 in U251 glioblastoma cells with small interfering RNA had no apparent effect on the growth of cancer cells (Supplementary Fig. 3), as previously reported,¹⁸ or influence on piperlongumine-mediated growth suppression (Supplementary Fig. 3). Therefore, we tested the effects of piperlongumine treatment on PRDX4 activity in HGG cells. For this, we used an antibody specific to oxidized forms of PRDX.^19,20 When PRDX reduces peroxide, the active cite Cys51 is oxidized to cysteine sulfenic acid, and PRDX becomes inactive.²¹ We found that piperlongumine treatment in mouse HGG cultures increased the oxidized forms of Prdx4, a ∼27 KDa form²⁰ (oxi-Prdx4) and a ∼18 KDa form,¹⁹ presumably a hyperoxidized form with a similar size to oxi-Prdx1/2 (oxi-Prdx1/2/4, Fig. 3D). The levels of oxi-Prdx4 and oxi-Prdx1/2/4 were highly increased following the drug treatment both in total cell lysates and after immunoprecipitation with PRDX4 antibody. In support of the increase in oxidized Prdx4, enzymatic activity of Prdx in the HGG cell lysate measured by NADPH consumption occurring upon hydrogen peroxide degradation by Prdx was decreased after piperlongumine treatment (Fig. 3E, left and middle columns). However, this enzyme assay did not distinguish different Prdx isomers in HGG cell lysate. In fact, the levels of oxi-Prdx3 were also increased in HGG cells upon piperlongumine treatment. However, piperlongumine treatment decreased the ROS-degrading activity of purified human PRDX4 (Fig. 3E, right columns). Given the overexpression of Prdx4 and not Prdx1–3 in mouse HGG cells compared with wt NSCs (Fig. 3C), these results suggested that the decrease in Prdx activity in HGG cells by piperlongumine treatment was largely attributable to Prdx4. In wt NSCs, piperlongumine alone or in cotreatment with the antioxidant NAC did not substantially change the levels of oxi-Prdx (Fig. 3F). However, NAC pretreatment prevented the piperlongumine-mediated oxidation of Prdx1–4 in HGG cells (Fig. 3F, columns 6 and 8). Therefore, collectively these results indicated that piperlongumine treatment inactivates PRDX4 in HGG cells but not in wt NSCs, which may underlie the piperlongumine-mediated ROS increase and growth suppression in HGG cells but not in wt NSCs.

Piperlongumine Increases ER Stress Level in HGG Cells

Consistent with the increased levels of ROS and apoptosis upon piperlongumine treatment (Fig. 2), we found that the drug treatment increased the levels of a marker for complex DNA strand breaks P-H2AX and an apoptosis marker, cleaved caspase 3, in all 4 HGG sphere cultures tested (Fig. 4A and Supplementary Fig. 4A). However, the molecular mechanism by which piperlongumine causes cell death appeared to differ in HGG spheres from previous reports: piperlongumine reportedly kills cancer cells with marked elevations of a p53 proapoptotic target p53-upregulated modulator of apoptosis (PUMA) in both p53-expressing and p53-null cell lines.^3,4 However, PUMA levels were not highly increased in the HGG cells (Fig. 4A and Supplementary Fig. 4A), which suggested a different or additional mechanism by which piperlongumine treatment kills HGG cells.

Fig. 4.

Piperlongumine (PL) treatment increases ER stress level in HGG spheres. (A) Western blotting analysis for DNA damage marker P-H2AX and apoptosis regulators 1 day after PL treatment in HGG cultures. (B) Western blotting analysis 3–24 hours after PL treatment (10 μM) in the HGG cultures. Tunicamycin (Tu), an ER stress inducer was used as a positive control (5 μg/mL for 6 h). (C) qPCR analysis for ER stress-response genes 6 hours after PL treatment (10 μM) in mouse HGG cultures. Relative mRNA levels to vehicle treatment are shown. ** P < .005. (D) Western blotting analysis (left) 12 hours after and MTS cell-growth assay data and (right) 24 hours after PL treatment (5 μM) in mouse HGG culture. Tauroursodeoxycholic acid (1 mM) was treated 1 hour before the PL treatment. (E) Relative ROS levels in mouse HGG cultures upon PL treatment (10 μM). * P < .05, ** P < .005. (F) Western blotting analysis 12 hours after PL treatment (10 μM) in mouse wt NSC and mouse HGG cultures. NAC (5 mM) was treated 1 hour before PL treatment. (G) Western blotting analysis 3–24 hours after PL treatment (10 μM) in primary mouse astrocytes. (A–G) Representative data from 2 independent experiments.

Excess ROS levels can exacerbate ER stress, which in turn can induce the death of cancer cells.^22,23 Therefore, we hypothesized that exacerbation of ER stress contributes to HGG cell killing by piperlongumine treatment. Accumulation of misfolded or unfolded proteins in the ER initiates ER stress response.²² While ER molecular chaperones sensors target misfolded or unfolded proteins for degradation and neutralize ER stress via unfolded protein response (UPR), severe or persistent ER stress switches the prosurvival effort to proapoptotic signaling. A transcription factor CCAAT/enhancer binding protein homologous protein (CHOP) promotes cell death by inducing proapoptotic genes and suppressing antiapoptotic genes.²⁴ We found that piperlongumine treatment rapidly and substantially increased CHOP protein levels in all 4 HGG sphere cultures (Fig. 4B and Supplementary Fig. 4B). As with CHOP, other UPR protein levels were also increased upon piperlongumine treatment in the HGG cells, such as phospho-eukaryotic translation initiation factor 2A (P-eIF2a), activating transcription factor 4 (ATF4), growth arrest and DNA damage inducible protein 34 (GADD34), and glucose-regulated protein of molecular weight 78 (GRP78) (Fig. 4B, 4C, Supplementary Fig. 4B, and 4C). In addition, the piperlongumine-mediated increases in P-eIF2a and CHOP levels as well as in cell death were reduced by pretreatment with the ER stress inhibitor tauroursodeoxycholic acid²⁵ (Fig. 4D). These results indicated that the piperlongumine-mediated growth suppression of HGG cells includes induction of ER stress and UPR.

Since excess ROS levels can exacerbate ER stress and vice versa,^22,23 we compared the piperlongumine-mediated changes in ROS and ER stress levels over time in HGG spheres. While the levels of ER stress were increased 3–6 hours after the drug treatment in mouse HGG cultures (Fig. 4B and Supplementary Fig. 4B), the increases in ROS levels were not observed until 12 hours after piperlongumine treatment (Fig. 4E), suggesting that the early increase in ER stress level was independent of the ROS elevation in the HGG cultures. However, pretreatment with NAC prevented the increases in CHOP, P-eIF2a, and cleaved caspase 3 levels in HGG cells 12 hours after piperlongumine treatment (Fig. 4F, lanes 6 and 8), suggesting that the late increases in ER stress, UPR, and apoptosis levels depended on the piperlongumine-mediated ROS elevation. In wt NSCs, piperlongumine treatment did not increase P-eIF2a level (Fig. 4F, lanes 1 and 2), although it had mild effects on CHOP and cleaved caspase 3 levels. These results together suggested that the piperlongumine-mediated PRDX4 inactivation increases both ER stress and ROS levels, which may act together to efficiently suppress HGG cell growth. In support of this, piperlongumine treatment in primary mouse astrocytes increased the levels of oxidized Prdx3 and P-eIF2a but not oxidized Prdx4 or CHOP (Fig. 4G), which correlates with the less growth suppression effect in the astrocytes than in mouse HGG cells (Fig. 1C).

PRDX4 Level Inversely Correlates with ER Stress Level in HGG Cells and Survival of Glioma Patients

Since PRDX4 contributes to protein folding in the ER by detoxifying hydrogen peroxide,²⁶ suppression of PRDX4 may also increase unfolded proteins, an excess of which can lead to apoptotic cell death by exacerbating ER stress.²² To examine whether PRDX4 contributed to ER-stress regulation, we employed our Tet on/off small hairpin RNA-expressing lentivirus system.² Western blotting and qPCR analysis revealed that PRDX4 knockdown increased ER stress in all 4 HGG cultures: the levels of all ER stress (CHOP) and UPR (P-eIF2a) markers and ER stress-response genes (ATF4, GADD34, and GRP78) were elevated upon PRDX4 knockdown (Fig. 5A and B, Supplementary Fig. 5, and not shown). Since the level of CHOP was also increased upon PRDX4 knockdown, these results indicated that proapoptotic signaling was initiated by aggravated ER stress in the HGG cells. These results also suggested that piperlongumine treatment exacerbates ER stress in HGG cells by inactivating PRDX4, which may contribute to the preferential killing of the cancer cells.

Fig. 5.

PRDX4 knockdown in sphere-cultured HGG cells increases ER stress. Western blotting (A) and qPCR analysis (B) for ER stress and UPR markers and ER stress-response genes 3 days after PRDX4 knockdown in human HGG cultures. Relative mRNA levels to control (Cnr) knockdown are shown for qPCR data. * P < .05, ** P < .005. (C) Kaplan-Meier survival curves from analysis of the REMBRANDT database reveal ∼470 and 1270 days of median survivals in glioma patients with a > 2-fold upregulated and an intermediate level of PRDX4 in their gliomas, respectively (P < .0001). PRDX4 expression fold changes are in comparison with normal brain tissues. (A and B) Representative data from 2 independent experiments.

Our results indicate that PRDX4 regulates both ROS and ER stress levels in HGG cells and that piperlongumine inactivates PRDX4, suggesting a prognostic value of PRDX4 expression in HGGs. Indeed, analysis of the REMBRANDT database reveals that PRDX4 expression inversely correlates with glioma patient survival: Patients with an intermediate level of PRDX4 in their gliomas survive ∼800 days longer than those with a > 2-fold increase in intratumoral PRDX4 level (Fig. 5C). These data not only demonstrate the prognostic value of PRDX4 in HGGs but also suggest that PRDX4 inactivation can extend survival of HGG patients.

Discussion

We have previously reported that PRDX4 is highly expressed in a majority of glioblastomas and that PRDX4 knockdown increases ROS levels and kills glioblastoma cells.² Extending our previous study, we sought to pharmacologically increase ROS levels in HGG cells. By employing piperlongumine, we found that the drug treatment kills sphere-cultured HGG cells but has little effect on the growth of normal brain cells.

ROS contribute to numerous normal cellular processes. However, excess ROS levels can cause oxidative stress, DNA damage, and even genomic instability. Therefore, excess ROS levels have long been considered carcinogenic, and antioxidant mechanisms have been sought as putative cancer therapies. However, recent reports indicate that a majority of cancers contain higher levels of antioxidant proteins than normal tissues.²⁷ These high levels of antioxidant proteins function to detoxify elevated levels of ROS in cancers. Typical cancer cells are highly proliferative, with accelerated protein synthesis.²⁸ One outcome of elevated protein synthesis is increased ROS levels because one hydrogen peroxide molecule forms each time a disulfide bond forms during protein synthesis in the ER. Hence, many cancers harbor higher ROS levels than normal tissues²⁹ and need increased levels of antioxidant proteins to reduce the lethal toxicity associated with excess ROS levels.²⁷ Accordingly, this detoxification mechanism is a cancer-specific inherent vulnerability that can be exploited in new therapeutic paradigms.

Several recent reports support the idea that increasing ROS levels could be an effective cancer therapy. For example, treatment with the natural compound obtusaquinone kills glioblastoma and breast cancer cells both in serum-containing cultures and in orthotopic mouse models, with several-fold increases of ROS.³⁰ Piperlongumine treatment also kills a range of cancer types, including glioblastoma,^3–5 with several-fold ROS increases.³ Since excess ROS levels can be cytotoxic to normal cells, it is critical for cancer chemotherapies to leave ROS and growth levels unaffected in normal cells. Piperlongumine treatment reportedly does not alter ROS levels in nontransformed cells.³ We found that piperlongumine treatment preferentially increases ROS levels and kills HGG cells with few effects in normal brain cells. However, the increases in ROS levels and growth suppression in astrocytes by the highest dose of piperlogumine indicate that careful dosage determination is needed for in vivo use of the drug. Although these results point to a chemotherapeutic potential for piperlongumine, the means by which the drug treatment preferentially increases ROS levels and kills cancer cells remains unknown.

Piperlongumine is known to interact with several ROS regulators.³ In the current study, we found that HGG cells expressed the piperlongumine-interacting ROS regulators at higher levels than normal brain cells. However, knockdown of CBR1, GSTZ1, and PRDX1 did not apparently affect piperlongumine-induced growth suppression in HGG cells. These findings suggest that HGG cells have a different piperlongumine target responsible for growth arrest. Through Western blotting for oxidized (inactivated) forms of PRDX and functional enzymatic assay, we found that piperlongumine treatment inactivated PRDX4. The oxidized forms of PRDX4 were apparently increased in HGG cells after the drug treatment, as evidenced by results obtained via immunoprecipitation with PRDX4 antibody. In addition, piperlongumine treatment inhibited enzyme reaction of purified PRDX4 out of the cells, suggesting a direct inhibition of PRDX4 by piperlongumine. The lack of Prdx oxidation in wt NSCs after piperlongumine treatment could result from poor penetration of piperlongumine in the cells, which may underlie their insensitivity to the drug treatment.

Through a series of microarray-based gene expression analyses in a mouse genetic model of HGG¹⁰ versus wt NSCs, a cellular origin of HGG,¹⁴ in conjunction with publically available TCGA human database,³¹ we found that most glioblastomas (>98%; n = 214/217) overexpress PRDX4.² The PRDX4 mRNA level is, on average, 5.0-fold higher in HGGs than in normal brain tissues, which is a considerably higher level than those of putative piperlongumine-interacting ROS regulators (1.3–2.0-fold, Fig. 3). Even more convincing is that both PRDX4 knockdown and piperlongumine treatment increased ROS, ER stress, and apoptosis levels in HGG cells based on the findings from our current and previous studies,² both PRDX4 knockdown and piperlongumine treatment increased ROS, ER stress, and apoptosis levels in HGG cells.

In addition to the elevation of ROS levels, exacerbation of ER stress in cancer cells has been under active investigation for cancer therapy.²² Numerous chemical compounds that enhance ER stress and increase cancer cell death have been developed for use in cancer clinics. However, since both ROS and ER stress also contribute to homeostasis of normal cells, the need to identify cancer-specific targets for these endpoints is essential. Given this necessity, we propose that PRDX4 inactivation by lower doses of piperlongumine treatment in HGG cells may have an important clinical implication. Among the PRDX gene family, PRDX4 is the only member present in the ER, where it detoxifies hydrogen peroxide generated during protein folding.²⁶ Therefore, suppression of PRDX4 can increase both ROS and ER stress levels, as demonstrated by our results from the current and previous studies on HGG cells.² Since an excess of either ROS or ER stress levels can trigger apoptotic cell death,²² PRDX4 inactivation in HGG cells by piperlongumine may amplify the cell-death signaling as supported by the increased growth suppression in mouse HGG cells (PRDX4 oxidation plus increases in both ROS and ER stress levels) compared with mouse astrocytes (no substantial changes in PRDX4 oxidation or ER stress level). Importantly, Prdx4 knockout mice are viable and fertile,³² and Prdx4 knockdown does not apparently affect growth of mouse embryonic fibroblasts,³³ suggesting that systemic PRDX4 targeting would be minimally toxic to normal cells.

Optimal chemotherapy for brain cancers must use drugs that cross the blood-brain barrier. Piperlongumine can cross the blood-brain barrier, as documented by a recent report.³⁴ Therefore, inactivation of PRDX4 could be considered as a putative therapy for patients with HGG as well as other PRDX4-overexpressing cancer types including prostate cancers, triple-negative breast cancers, and osteosarcomas.^35–37

Supplementary Material

Supplementary material is available online at Neuro-Oncology (http://neuro-oncology.oxfordjournals.org/).

Funding

NIH R01NS064607, R01CA150153, P01 CA163205, and P30NS045758 to B.K. and American Cancer Society Institutional Seed Grant, Ohio State University (OSU) Comprehensive Cancer Center (CCC) Intramural Research Program Idea Grant, and OSU CCC Start-up Fund to C.-H.K.