Proteasome Inhibition Potentiates Antitumor Effects of Photodynamic Therapy in Mice through Induction of Endoplasmic Reticulum Stress and Unfolded Protein Response

Reactive oxygen species (ROS) and, in particular, singlet oxygen (¹O₂) are responsible for the cytotoxic effects induced by photodynamic therapy (PDT; refs. 1, 2). ¹O₂ reacts with proteins, lipids, and DNA (3–7). Radical-induced protein modifications include fragmentation, multimerization, misfolding, and/or structural alterations, resulting in functional inactivation, changes in mechanical properties, aggregation, formation of further reactive species, or accelerated degradation (8).

Multiple mechanisms have evolved to interact with oxidants to form less reactive products. These include enzymes that directly react with ROS such as superoxide dismutases, catalase, glutathione peroxidase, and a number of secondary scavengers participating in the reduction of oxidized biomolecules. However, during robust oxidative stress conditions, such as those occurring during PDT, these cytoprotective mechanisms are insufficient and a significant damage to cellular constituents may ensue. Therefore, other mechanisms have developed to enable restoration of normal protein function within cells. Oxidized proteins, which lose their physiologic function due to conformational changes, can be refolded by molecular chaperones such as heat shock proteins (HSP). Notably, PDT was found to induce expression of a variety of HSPs including HS1, HSP27, HSP60, HSP70, HSP90, GRP78, and GRP94, and for some of these proteins a protective role in PDT-treated cells has been shown (9–13). Inefficient restoration of protein structure that leads to accumulation of misfolded proteins within endoplasmic reticulum (ER) is called ER stress (14).

ER stress triggers several independent adaptive responses collectively referred to as unfolded protein response (UPR). Its function is to reestablish protein homeostasis and normal ER function. UPR triggers at least three signaling pathways: activation of protein kinase RNA–like ER kinase, activating transcription factor 6, and inositol-requiring enzyme 1 (15). Activation of inositol-requiring enzyme 1 leads to unusual cytosolic splicing of mRNA for X-box binding protein 1 (XBP-1), which on translation produces sXBP-1, an active transcription factor. UPR-induced genes encode chaperones (including numerous HSPs), enzymes that participate in red-ox reactions and management of oxidative stress, and genes for ER-associated degradation (ERAD). ERAD involves the retrograde translocation of unfolded proteins from the lumen of the ER to the cytosol, where they undergo degradation in proteasomes (16). The latter are large multisubunit protease complexes endowed with three main proteolytic activities: trypsin-like, chymotrypsin-like, and caspase-like (17, 18). Excessive ER stress or a failure of UPR to counteract it can trigger cell death.

One of the typical biomarkers for oxidative protein damage is carbonylation. Carbonylation leads to exposure of hydrophobic patches within proteins, resulting in their partial unfolding, which favors their ubiquitination followed by recognition and degradation by proteasomes (19). The ubiquitin-proteasome system (UPS) has been shown to play an important cytoprotective role through degradation of oxidatively modified proteins (19, 20). Robust carbonylation of proteins results in the formation of large protein aggregates or “aggresomes,” which entrap elements of the UPS, decreasing the cellular capacity of proteasomal proteolysis (21). Treatment with proteasome inhibitors is also associated with formation of intracellular protein aggregates, increased ER stress, and UPR induction in different tumor models (22, 23).

It was previously shown that PDT with purpurin-18 induces protein carbonylation in tumor cells (24). However, the fate of carbonylated proteins in tumor cells has not been investigated. Therefore, we decided to study the proteasome-mediated degradation of PDT-induced protein carbonyls to better understand the mechanisms of PDT-mediated cytotoxicity and to find out whether inhibition of this degradation would affect the antitumor effects of PDT.

Tumor cells. Human cervical cancer (HeLa) and murine breast carcinoma (EMT6) cell lines were purchased from American Type Culture Collection. Murine colon adenocarcinoma (C-26) was obtained from Dr. Danuta Dus (Institute of Immunology and Experimental Medicine, Wroclaw, Poland). HeLa cells stably expressing HA-tagged δCD3, HA-tagged αTCR, α1-antitrypsin Hong Kong mutant (α1AT), and Ub_G76VGFP were used. Expression plasmids encoding reporter substrates were obtained from Dr. Allan Weissman (Laboratory of Protein Dynamics and Signaling, National Cancer Institute, Frederick, MD; δCD3), Dr. Kazuhiro Nagata (Department of Molecular and Cellular Biology, Kyoto University, Kyoto, Japan; α1AT), Dr. Ron Kopito (Department of Biology, Stanford University, Stanford, CA; αTCR), and Dr. Maria Masucci (Department of Cell and Molecular Biology, Karolinska Institutet, Stockholm, Sweden; Ub_G76VGFP), and their sequences were verified before transfection. Stably transfected HeLa clones expressing reporter substrates were chosen for experiments. Cells were cultured in RPMI 1640 (C-26) or DMEM (HeLa and EMT6) supplemented with 10% fetal bovine serum (Invitrogen), antibiotic/antimycotic solution (Sigma), and 400 μg/mL geneticin (Sigma; transfected HeLa cells) to ensure stable expression of plasmid vectors.

Mice. BALB/c mice, 8 to 12 wk of age, were used in the experiments. Breeding pairs were obtained from the Animal House of the Polish Academy of Sciences, Medical Research Center (Warsaw, Poland). All in vivo experiments were done in accordance with the guidelines approved by the Ethical Committee of the Medical University of Warsaw.

Reagents. Photofrin (Axcan Pharma, Inc.), verteporfin (a generous gift of QLT PhotoTherapeutics, Inc.), aminolevulinic acid (Sigma), and hypericin (prepared, purified, and stored as described; ref. 25) were used as photosensitizers. MG132 and PSI were purchased from Calbiochem/EMD and were dissolved in cell culture quality DMSO (Sigma). Bortezomib (Millenium Pharmaceuticals) was dissolved in 0.9% NaCl.

Tumor treatment and monitoring. Exponentially growing EMT6 and C-26 cells were harvested, resuspended in PBS, and injected into the footpad of the right hind limb of experimental mice. Tumor cell viability, measured by trypan blue exclusion, was >98%. Tumor treatment was started when all mice developed tumors with a minimum size of 3 × 4 mm. Photofrin (in 5% dextrose) was administered i.p. at a dose of 10 mg/kg 24 h before illumination with 632.8-nm light (day 6 after inoculation of tumor cells). Control mice received 5% dextrose. The light source was a He-Ne ion laser. The light at a fluence rate of 40 mW/cm² was delivered on day 7 of the experiments using a fiber optic light delivery system as previously described (26, 27). The total light dose delivered to the tumors was 90 J/cm². During illumination, mice were anesthetized with ketamine (87 mg/kg) and xylazine (13 mg/kg) and were restrained in a specially designed holder at 37°C. PSI at a dose of 20 nmol (dissolved in DMSO) was administered intratumorally (i.t.) for 7 consecutive days, with the first dose given on day 7 of the experiment. Control animals received DMSO. Bortezomib at a dose of 1 mg/kg was administered i.p. in two different schedules: on days 5 and 7 after inoculation of tumor cells (before PDT) or on days 7, 9, 11, and 13 after inoculation of tumor cells (after PDT). Control mice received 0.9% NaCl i.p. Local tumor growth was determined with calipers as previously described (28, 29) with the following formula: tumor volume (mm³) = (longer diameter) × (shorter diameter)².

Statistical analyses. Data were calculated using Microsoft Excel 2007. Differences in in vitro cytotoxicity assays and tumor volume were analyzed for significance by Student's t test. Kaplan-Meier plots were generated using days of animal death (after inoculation of tumor cells) as a criterion, and survival time of animals was analyzed for significance by log-rank survival analysis. Significance was defined as two-sided P < 0.05. The nature of the interaction observed between proteasome inhibitors and PDT was analyzed using the Calcusyn software (Biosoft), which uses the combination index method of Chou and Talalay (30) based on multiple drug effect equation. The advantage of this method is the automatic construction of a fraction-affected combination index table, graph, and calculation of dose reduction indices by the software. Combination indices of <1 indicate greater than additive effects (synergism; the smaller the value, the greater the degree of synergy); combination indices equal to 1 indicate additivity; and combination indices of >1 indicate antagonism.

Additional procedures are described in Supplementary Materials and Methods.

PDT induces carbonylation and accumulation of ubiquitinated proteins. Time course experiments with approximately equitoxic light fluencies revealed that PDT led to a robust protein carbonylation detectable 8 hours after illumination of HeLa and C-26 cells (Fig. 1A and C). Increased protein carbonylation coincided with the accumulation of polyubiquitinated proteins at 8 hours after illumination (Fig. 1B and D). However, despite a decrease in the amount of carbonylated proteins in cellular lysates, there was a further accumulation of polyubiquitinated proteins 24 hours after illumination.

To study the influence of photosensitization on proteasome activity, purified 20S proteasome complexes were incubated with Photofrin and illuminated with laser light. PDT did not affect the trypsin-, chymotrypsin-, or caspase-like activities of purified proteasomes (Fig. 2A). However, despite the lack of direct effect of photosensitization on proteasome activity, there was a time-dependent (Fig. 2B) and dose-dependent (Fig. 2C) induction of chymotrypsin-like activity of proteasomes in whole-cell lysates of PDT-treated C-26 cells. These observations indicate that PDT does not inhibit the activity of proteasomes (found in extramembranous compartments of cells). The increased activity of proteasomes in PDT-treated cells might be an adaptive mechanism to remove oxidatively damaged proteins.

To investigate subcellular localization of photodamaged and polyubiquitinated proteins, we used two HeLa cells stably transfected with reporter constructs: an ER membrane–associated model protein (HA-tagged δCD3 chain) and a cytosolic Ub_G76VGFP protein (GFP-GV). Most δCD3 resides on the cytosolic side of the ER membrane and is a substrate for the ERAD pathway. GFP-GV is a cytosolic protein that has a noncleavable ubiquitin fused in frame with green fluorescent protein (GFP) that targets the protein for degradation through the ubiquitin-fusion degradation pathway. Photofrin-mediated PDT induced accumulation of ER-associated δCD3, but not cytosolic GFP-GV proteins (Fig. 2D). Because Photofrin is a hydrophobic photosensitizer that accumulates in membrane compartments (31), it is likely that PDT is inducing a selective damage to membrane-associated proteins that leads to ER overloading and thereby impairment of ERAD.

Oxidative damage and impaired degradation of ER-associated proteins might lead to ER stress. Indeed, PDT increased the expression of BiP and calnexin in HeLa cells (Supplementary Fig. S1A and B); reverse transcription-PCR revealed that PDT induced a time-dependent unconventional cytosolic splicing of XBP-1 (Supplementary Fig. S1C); and electron microscopy studies revealed that PDT led to widening of the ER lumen and to occasional vacuolization of the cytoplasm of EMT6 cells. These changes were accompanied by mitochondrial swelling and formation of infrequent lysosomal vacuoles that could represent autophagosomal vesicles (Supplementary Fig. S2). Similar changes were noticed in PDT-treated HeLa cells (see below).

Preincubation of tumor cells with proteasome inhibitors augments accumulation of carbonylated and polyubiquitinated proteins. Cells remove carbonylated proteins to avoid the toxic effects of protein aggregation by directing them for proteasome-mediated degradation (32, 33). Therefore, we examined whether inhibition of proteasomal protein degradation would affect the amount of carbonylated proteins in PDT-treated tumor cells. EMT6 cells were preincubated with bortezomib and Photofrin for 24 hours and illuminated at a fluence of 2.4 J/cm². A marked accumulation of carbonylated proteins was observed within 8 hours after PDT, which significantly decreased during the next 16 hours (Fig. 3A). In the presence of bortezomib, the amount of carbonylated proteins was higher at 8 hours after illumination, and their amount further increased during the next 16 hours. Similar effects were detected in HeLa cells (Fig. 3A). Immunofluorescence studies also revealed increased accumulation of polyubiquitinated proteins in cells treated with PDT + proteasome inhibitors (Fig. 3B). Similarly, with Western blotting we observed that proteasome inhibitors increased the amount of polyubiquitinated proteins that accumulate in PDT-treated tumor cells (Fig. 3C).

Next, we performed immunofluorescence studies with HeLa cells stably transfected with ER-localized model proteins: δCD3, α1AT, and αTCR. All these proteins are ERAD substrates. In normal conditions, they undergo proteasomal degradation and are undetectable by immunofluorescence. Impairment of ERAD leads to accumulation of these proteins. α1AT is an entirely lumenal ERAD substrate (with no association to the ER membrane), whereas αTCR and δCD3 are single-span transmembrane proteins. Both bortezomib and PDT, when used alone, induced slight accumulation of reporter proteins in tumor cells, but only ER membrane–associated proteins (δCD3 and αTCR) accumulated to a significantly higher extent in HeLa cells preincubated with proteasome inhibitor and treated with PDT (Supplementary Fig. S3A). Immunoprecipitation of polyubiquitinated proteins followed by immunochemical detection of carbonyl groups revealed that proteins that accumulate in the treated cells were also carbonylated (Supplementary Fig. S3B). Similarly, immunoprecipitated δCD3 was heavily carbonylated (Supplementary Fig. S3C).

Preincubation of tumor cells with proteasome inhibitors potentiates the cytotoxic effects of PDT in vitro. Considering that PDT induces robust oxidative protein damage followed by carbonylation, and that proteasomes participate in the degradation of carbonylated proteins, we decided to investigate the outcome of the combination treatment including PDT and proteasome inhibitors. To this end, tumor cells (EMT6, C-26, and HeLa) were preincubated with Photofrin and/or three different proteasome inhibitors (bortezomib, MG132, and PSI) for 24 hours and illuminated at fluencies that produced suboptimal rates of tumor cell death. Potentiated cytotoxic effects of the combination of PDT + proteasome inhibitors in all studied cell lines were observed (Fig. 4). The proteasome inhibitor concentrations shown are the lowest that produced potentiated cytotoxicity. At higher concentrations of bortezomib, potentiated cytotoxicity was not better than that at low doses (data not shown), indicating that it might not be necessary to use high-dose treatment to elicit maximal cytotoxic response. Chow and Talalay calculations revealed that proteasome inhibitors and PDT exert synergistic and strongly synergistic cytotoxic effects against tumor cells. Incubation of tumor cells with proteasome inhibitors did not increase Photofrin uptake (Supplementary Table S1), indicating that it is an unlikely mechanism of potentiated photosensitization.

To further study the mechanism of potentiated cytotoxicity, we incubated HeLa cells with 250 nmol/L MG132 for either 24 hours or 1 hour before illumination. A 24-hour exposure to MG132 potentiated the cytotoxic effects of PDT to a higher degree than did incubation of tumor cells for 1 hour before illumination (Supplementary Fig. S4). A significant cytotoxicity was observed already at 4 hours after illumination with 2.4 J/cm² in HeLa cells preincubated with MG132, whereas the cytotoxic effects of PDT alone and PDT combined with a 1-hour preincubation with MG132 were recorded 24 hours after illumination (Supplementary Fig. S5A). Potentiated cytotoxicity was associated with accumulation of higher amounts of carbonylated proteins in cells preincubated with 250 nmol/L MG132 for 24 hours and treated with PDT (Supplementary Fig. S5B). Because proteasome inhibitors induce accumulation of proteins that, under normal conditions, should be degraded (34), including misfolded proteins in the ER, it can be hypothesized that the potentiated cytotoxicity results from the accrual of misfolded and/or oxidatively damaged proteins beyond the proteasomal degradative capacity. Furthermore, the influence of proteasome inhibitors on the cytotoxic effects of PDT with different photosensitizers was evaluated. For these studies, we used photosensitizers that localize predominantly either in mitochondria (verteporfin and aminolevulinic acid) or in ER (hypericin). The ER-localizing photosensitizer hypericin was found most effective in eliciting potentiated cytotoxic effects in the combined regimen (Supplementary Fig. S6).

To gain further insight into the mechanisms of potentiated cytotoxicity of the combination treatment, electron microscopy studies were done. The most remarkable lesions observed within 8 hours of illumination in cells treated with the combination of bortezomib and PDT included robust vacuolization of the cytoplasm, with frequent lysosomal/autophagosomal vesicles and extended ER (Fig. 5).

Preincubation of tumor cells with proteasome inhibitors potentiates the antitumor effects of PDT in vivo. In the initial in vivo experiment, we compared the antitumor effects of the combination treatment with bortezomib administered either before or after suboptimal PDT in an immunogenic murine breast carcinoma (EMT6) model (Fig. 6A and B). Interestingly, proteasome inhibition significantly potentiated the antitumor effects of PDT in both experimental paradigms, with complete responses observed when bortezomib was administered before or after PDT. Furthermore, PSI, another proteasome inhibitor, significantly potentiated the antitumor effects of PDT in murine colon (C-26) and breast (EMT6) adenocarcinoma models (Fig. 6C and D). Altogether, in all experiments, a marked potentiation of the antitumor effects of PDT in vivo was observed, resulting in retardation of tumor growth, prolongation of the survival time, and complete disappearance of tumors in at least 60% of animals (no tumor regrowth for 120 days of observation).

It was previously shown that PDT with ER- and mitochondria-localizing purpurin-18 induces protein carbonylation, which is a typical oxidative protein modification (24). Interestingly, many of the carbonylated proteins identified in this report are normally functioning within ER and include BiP, calreticulin, phosphate disulfide isomerase, and HSP cognate 71. The major mechanism for the elimination of carbonylated proteins is their degradation by the UPS (32, 33). Therefore, we studied the influence of Photofrin-mediated PDT on carbonylation and ubiquitination of proteins. Interestingly, although the level of protein carbonylation was maximal at 8 hours after PDT and then decreased (Fig. 1A and C), the amount of polyubiquitinated proteins was still increasing until 24 hours after light exposure (Fig. 1B and D). There are several potential mechanisms for this discrepancy. It is possible that carbonylated proteins are not only ubiquitinated but are also continuously accumulating in PDT-treated cells. It is also possible that heavily carbonylated proteins are preferential substrates for proteasomes, thereby resulting in accumulation of other proteins destined for degradation. Accumulation of polyubiquitinated proteins might also be caused by impaired proteasomal degradation, as has been observed with radiotherapy (35). However, the fact that neither the function of purified 20S proteasomes after direct photosensitization nor the proteasome activity in PDT-treated cells was impaired indicates that PDT does not damage the UPS (Fig. 2A–C). Immunofluorescence studies of HeLa cells expressing model substrates of the UPS indicated that PDT leads to selective accumulation of ER-associated reporter protein (δCD3), but not a cytosolic one (GFP-GV), although both undergo carbonylation after PDT. This observation indicates that PDT might lead to a selective impairment of the ERAD pathway, which is responsible for the retrotranslocation of damaged or misfolded proteins from the ER for their cytoplasmic degradation within proteasomes (36). Similar changes might also occur in other cellular membranes, including mitochondrial or plasma membrane, but were not further studied here.

We and others have shown that PDT leads to a rapid ROS-mediated loss of function of proteins, enzymes, and transporters colocalized in various cellular compartments with photosensitizers (37–41). For example, the ER-localizing hypericin leads to a rapid disappearance of SERCA2 Ca²⁺ transporters, which is associated with increased cytoplasmic Ca²⁺ levels (37). Rapid disappearance of SERCA2 might be caused by a number of mechanisms such as fragmentation or cross-linking, but it might also be caused by its carbonylation resulting in change of the antibody binding sites (epitopes). The possibility that PDT leads to ROS-mediated damage to the components of ERAD machinery will require further studies. It is also possible that elimination of carbonylated proteins observable within 24 hours results from engagement of other mechanisms such as autophagy. Indeed, PDT was shown to induce autophagy in tumor cells (42). Electron microscopy studies revealed that there are lysosomal/autophagosomal vesicles forming in PDT-treated cells. The potential role of autophagy in the elimination of carbonylated proteins remains to be elucidated.

Accumulation of carbonylated proteins within ER is accompanied by induction of ER stress and UPR after PDT (Supplementary Fig. S1). Although it was previously shown that PDT induces damage to ER (42–44) and increases the expression of ER chaperones (45–47), the signaling pathways associated with UPR have not been studied. We observed that PDT leads to cytoplasmic splicing of XBP-1 mRNA (Supplementary Fig. S1C), a transcription factor associated with UPR. Together, these observations indicate that PDT-induced damage leads to aberrant protein folding within ER. The balance between the extent of ER stress and the adaptive UPR response can dictate cell fate (14). Excessive loading of ER with unfolded proteins that results in extensive ER stress can overwhelm the capacity of UPR, leading to cell death (14, 48).

Therefore, we decided to study the cytotoxicity of a combined regimen consisting of preincubation of tumor cells with proteasome inhibitors followed by PDT. In all cell lines and combinations studied, we have observed potentiated cytotoxicity as compared with either PDT alone or proteasome inhibitor alone groups (Fig. 4). It is possible that the potentiated cytotoxicity results from accumulation of proteins within the ER membrane at the time of illumination and not from any direct interaction of PDT and proteasome inhibitors because only 24 hours, and not 1 hour, of preincubation of HeLa cells with MG132 significantly sensitized HeLa cells to PDT (Supplementary Fig. S4 and S5A). Combined treatment led to accumulation of more carbonylated and polyubiquitinated proteins in tumor cells as compared with single modality–treated cells (Supplementary Fig. S5B), and the proteins that preferentially accumulate were associated with the ER membrane (δCD3 and αTCR) but not with the ER lumen (α1AT) or cytoplasm (GFP-GV; Supplementary Fig. S3A). These observations indicate that most severely damaged are ER membrane–associated proteins. Indeed, transmission electron microscopy studies revealed that there is a marked vacuolization of the cytoplasm in bortezomib + PDT–treated cells (Fig. 5), which might be caused by robust extension of the ER lumen. However, it cannot be excluded that large vacuoles observed in tumor cells treated with bortezomib and PDT are of autophagic origin. In PDT-treated cells, we observed numerous swollen mitochondria (which also accumulate Photofrin), but these structures were not present in tumor cells after combination treatment. It is likely, although not further addressed, that severely damaged mitochondria were removed in a process of mitophagy.

The most important observation is that proteasome inhibitors significantly potentiate the antitumor effects of Photofrin-mediated PDT in vivo (Fig. 6). There were complete tumor regressions in mice treated with the combined regimen, and after 120 days of observation, tumors relapsed in only up to 40% of animals. Intriguingly, potentiated antitumor effects were observed independently of the timing of proteasome inhibitor administration and occurred when bortezomib was administered either before or after PDT. Because PDT exerts potent antivascular effects, it is possible that PDT-treated tumors undergo ischemia-reperfusion episodes that lead to prolonged oxidative stress. Therefore, administration of proteasome inhibitors before or after PDT might elicit potentiated antitumor effects that would result from either sensitization to photooxidation or postillumination oxidative stress caused by ischemia-reperfusion.

In conclusion, we observed that Photofrin-mediated PDT leads to carbonylation and ubiquitination of ER membrane–associated proteins, accompanied by induction of ER stress and UPR. Compounds that induce ER stress, such as proteasome inhibitors, can effectively sensitize tumor cells to PDT-mediated cytotoxicity. These observations are of clinical significance because proteasome inhibitors are successful drugs approved for use in oncology. However, caution should be exercised in the treatment of tumors growing at certain anatomic locations (such as esophagus, bronchi, or bladder) because potential increase in the cytotoxicity to normal cells might be associated with side effects such as perforations or bleeding.

No potential conflicts of interest were disclosed.

Grant support: Ministry of Science and Higher Education-Poland grants N40112331/2736, R0504303 (J. Golab), and N401324033 (D. Nowis); Medical University of Warsaw grants 1M19/N (M. Jakóbisiak), 1M19/WB1/07 (M. Jakóbisiak), 1M19/WB2/07 (D. Nowis), and 1M19/NM2/07 (P. Salwa); K.U.Leuven grant OT/06/49 (P. Agostinis); “Fonds voor Wetenschappelijk Onderzoek” (FWO)-Vlanderen grant G.0492.05 (P. Agostinis); U.S. NIH grant R01-CA/AI838801 (M.R. Hamblin and P. Mróz); and an institutional appropriation of the American Cancer Society grant IRG-84-002-22 (C. Wójcik). T. Issat is a recipient of the START stipend from the Foundation for Polish Science.

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.