Enhanced Ubiquitinylation of Heat Shock Protein 90 as a Potential Mechanism for Mitotic Cell Death in Cancer Cells Induced with Hypericin

Treatment of cancerous cells with chemotherapeutic agents such as cisplatin (1), daunorubicin (2), and paclitaxel (3), as well as radiotherapy with high dose γ radiation, can kill tumor cells via a newly described mitotic catastrophe or MCD⁴ mechanism (4, 5). This phenomenon, which prevails in many tumors bearing mutant p53 and which exhibit resistance to genotoxic agents, is characterized by cell accumulation at G₂-M, increased cell volume, and multinucleation (6). A variety of seemingly unrelated cellular functions appear to be altered as cells approach mitotic death. The G₁-S checkpoint diminishes in activity, cells are delayed at G₂, apoptosis is uncoupled, and polygenomic giant cells are formed (7). The processes, which lead to development of these abnormal polykaryons remain controversial, and theories fluctuate between aberrant endomitotic replication cycles and cell fusion (5).

The anticancer activities of the signal transduction inhibitor hypericin, a photodynamic dianthraquinone, elicit light-dependent inhibition of protein kinase C (8) and Erk1/2 MAP kinases (9). However, hypericin is unique in its ability to also maintain activity within biological systems in the absence of light (Refs. 10, 11; dark effects). It results in a portfolio of antitumoral (10), antimetastatic,⁵ immunomodulatory (12), and antiviral (13, 14) activities in vitro, and in animal models in the dark.

In exploring mechanisms for its anticancer activities we considered that hypericin can cause reduction in intracellular pH by proton transfer to surrounding molecules (15) leading to pH-dependent structural changes in proteins (16). Furthermore, in cells this compound localizes to the endoplasmic reticulum and Golgi apparatus (17), which are folding sites of newly synthesized proteins. The broad range of activities may, thus, result from hypericin targeting chaperone networks. One abundant chaperone that assists in protein folding and protects de novo synthesized proteins from catabolism is Hsp90 (18, 19). This chaperone is pivotal for stability and function of a large group of client proteins including ser/thr kinases, tyrosine kinases, mutated p53, cyclin D-associated Cdk4 (20) and Cdk6, and wild-type Plk (21). Replication mechanisms in cells that lose Hsp90-protective function are severely compromised, disabling cell progression through the cell cycle (22). Indeed, Hsp90 modulators such as geldanamycin or radicicol, which interact with the ATP binding pocket of Hsp90, exhibit potent anticancer effects (18).

In this study we examined the correlation between stability of members of chaperone families and cell functionality in two murine tumor cell models treated with hypericin in the dark. A link through Hsp90 is unraveled between several key cell cycle regulators, which are destabilized as a result of loss of protective chaperone activity leading to MCD. The potential role of such mechanisms in generating diverse biological activities is discussed.

Monoclonal antibodies to Cdk1, Cdk2, Cdk7, Raf-1, p27^kip1, Plk, ubiquitin, and polyclonal anticyclin B₁ and anti-Plk (H-152) antibodies (for immunoprecipitations) were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); monoclonal antibodies to p53, cdk4, cyclin A, and proliferating cell nuclear antigen were from Oncogene (Boston, MA). Anti-HSP-90 and anti-Hsc70 antibodies were purchased from StressGen Biotechnologies (Victoria, British Columbia, Canada); anti-Erk1/2 MAP kinases (antibody Erk1/2-CT) and anti-Active MAP kinase pAb (pTEpY) polyclonal antibodies from Upstate Biotechnology (Lake Placid, NY) and from Promega (Madison, WI), respectively. Anti-p50^cdc37 antibody was a gift from Dr. Nicholas Grammatikakis (Queen’s University, Kingston, Ontario, Canada). Secondary horseradish peroxidase-conjugated or FITC-labeled antibodies were from Jackson ImmunoResearch Laboratories, Inc. (West Grove, PA).

Hypericin (10, 11-dimethyl-1, 3, 4, 6, 8, 13-hexahydroxy-naphtho-dianthrone) was synthesized by Yehuda Mazur (Department of Organic Chemistry, Weizmann Institute of Science, Rehovot, Israel) as described previously (12). The compound was dissolved in 70% aqueous ethanol to a 3 mg/ml stock solution. Subsequent dilutions were made in sterile double distilled H₂O to obtain a final ethanol concentration <0.5%.

[³⁵S]methionine/cysteine NEG-772 EASYTAG express protein labeling mix was purchased from Perkin-Elmer Life Sciences (Boston, MA) and the proteasome inhibitor MG132 from Calbiochem (Darmstadt, Germany).

Murine DA3 mammary adenocarcinoma cell line was obtained from Dr. Diana Lopez (University of Miami School of Medicine, Miami, FL), SQ2 squamous cell carcinoma cell line was generated in our laboratory (10), and B16.F10, a highly metastatic melanoma cell line, from Dr. Lea Eisenbach (Weizmann Institute). All of the cell lines were grown in DMEM supplemented with 10% fetal bovine serum, 2 mm l-glutamine, and 100 units/ml penicillin-streptomycin (Life Technologies, Inc., Paisley, Scotland). After administration of hypericin the cultures were maintained in strict darkness (ambient light kept ≤0.03 mW/cm²) throughout the entire duration of the experiments. Light incidence was measured with an IL 1350 Radiometer/Photometer (International Lighting, Newburyport, MA).

DNA synthesis measured as BrdUrd incorporation was assayed on cells grown on glass coverslips and pulsed with BrdUrd. Incorporation of BrdUrd into DNA was visualized by immunostaining with monoclonal antibromo-deoxyuridine antibody followed by FITC-conjugated rabbit antimouse F(ab′)₂ using a commercial kit (Amersham, Piscataway, NJ).

Cell viability was assessed using the Hemacolor assay (reagents obtained from Merck, Darmstadt, Germany) as described previously (16). The percentage of cells remaining in hypericin treated cultures (or in untreated samples) is presented as percentage of cell viability using the formula:

Cellular DNA content was determined by cell staining with propidium iodide using a Coulter DNA-Prep kit (Beckman Coulter Int. S.A., Nyon, Switzerland). Cell cycle distribution was determined using Coulter EPICS XL-MCL flow cytometer (Coulter) and Multicycle software (Phoenix, San Diego, CA) for data analysis. Polykaryons occurring during MCD were visualized by concomitant staining of cell cultures with Hemacolor reagents (10).

Nuclear and cytosolic extracts were prepared from DA3 and SQ2 cells grown to confluence; cytosolic extracts using buffer A [10 mm HEPES (pH 7.9), 10 mm KCl, 1 mm EDTA, 1 mm EGTA, 1 mm DTT] and complete Protease Inhibitor mixture, 40 μg/ml (Boehringer, Mannheim, Germany), containing 0.6% NP40. Nuclear extracts were prepared by dissolution of nuclei in buffer C [20 mm HEPES (pH 7.9), 400 mm KCl, 1 mm EDTA, 1 mm EGTA, 1 mm DTT, and complete Protease Inhibitor mixture 40 μg/ml]. The protein content was calibrated using the BCA protein assay reagent kit (Pierce, Rockford, IL). Samples were separated on 10–15% SDS-PAGE and transblotted onto nitrocellulose filters (Schleicher & Schuell, Dassel, Germany). Equal protein loading was verified by staining with Ponceau S (Sigma, St. Louis, MO). The membranes were probed with several primary antibodies, with peroxidase-tagged second antibodies and developed with enhanced chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate; Pierce).

Immunoprecipitates were prepared from cytosolic extracts. After overnight incubation with primary antibodies, samples were immobilized with Protein A/G PLUS-Agarose (Santa Cruz Biotechnology) at 4°C. Immunoprecipitated proteins were separated on 10% SDS-PAGE and subjected to Western blot analyses.

DA3 cells were cultured in DMEM (met⁻, cys⁻; Life Technologies, Inc.) containing 10% dialyzed FCS and 100 μCi ml⁻¹ [³⁵S]methionine/cysteine for 1 h. The cells were washed with PBS, cultured in chase medium, lysed, and cytosolic extracts prepared. Plk was immunoprecipitated and the precipitates separated on SDS-PAGE. Radiolabeled proteins were visualized by fluorography.

mRNA expression was analyzed using the Titan one-step RT-PCR system (Roche Molecular Biochemicals, Mannheim, Germany). β-Actin served as a reference housekeeping gene. Specific primers were designed from sequences derived from GenBank database using the Primer3 Software (Whitehead Institute) to yield an annealing temperature of 55°C. The specificity of these primers was confirmed using Standard nucleotide-nucleotide BLAST. Primers used were: Plk: 5′-aagtctttgctgctcaagcc-3′ (forward) 5′- aggtccctgtgaatgacctg-3′ (reverse); Hsp90β: 5′- agaaggctgaggcagacaaa-3′ (forward) ′-ctccaagtcatcatgagcca-3′ (reverse); and β-Actin: 5′-atggatgacgatatcgct-3′ (forward) 5′-atgaggtagtctgtcaggt-3′ (reverse).

Amplified cDNAs were subjected to electrophoresis and photographed under UV light.

The data were analyzed using the two-tail Student’s t test. Ps < 0.05 were considered statistically significant.

Treatment of DA3 breast and SQ2 squamous carcinoma cells with hypericin for 72 h in strict darkness (to exclude light-induced phototoxicity) had two outcomes. Doses ≤10 μm induced cytostasis; BrdUrd incorporation into DNA declined to below detection (Fig. 1,A); and [³H]thymidine incorporation into DNA diminished (data not shown). Concentrations >10 μm reduced cell viability with LD₅₀ of ∼10 μm in SQ2 cells, and 12.5 μm in DA3 cells (Fig. 1 B). Complete loss of cell viability in both cell lines occurred after cell exposure to ≥40 μm hypericin.

The proportion of cells sequestered in G₂-M increased from 28.1% in untreated controls to 41.8% (P = 0.025; n = 7) after 72 h treatment of DA3 cells with 20 μm hypericin (Fig. 1 C). No evidence for occurrence of apoptosis could be detected: sub-G₁ DNA peaks in propidium iodide-stained cells were absent and DNA ladder not noted in agarose gel electrophoresis in both cell lines.

Microscopic analyses revealed the formation of enlarged polynucleated cells in DA3 cells, SQ2 cells, and in B16.F10 melanoma cells (Fig. 1 D). Polykaryon abundance began to increase 48 h after administration of ≥10 μm hypericin and peaked at 72 h. The decline in G₁ checkpoint activity, retardation at G₂-M phase, and multinucleation are hallmarks of MCD (5) and point to induction of this phenomenon by hypericin in the dark.

Expression of several cell cycling regulators that govern cell progression through the replication cycle were examined in DA3 and SQ2 cells after treatment with hypericin for 72 h. Western blot analyses revealed diminished expression of mutant p53, cdk4, and Raf-1 in DA3 cells (Fig. 2,A). Most of these proteins declined to below detection primarily in the cytosolic fraction of the cells in hypericin dose-dependent manners. Catabolism of mutant p53 showed evidence of ubiquitinylation (Fig. 2 A, top panel). The mutated nature of p53 in DA3 cells was confirmed by immunofluorescence using the monoclonal PAb 240 anti-p53 antibody (data not shown). Under nondenaturing conditions this antibody has been shown to react selectively with mutant p53 (23, 24).

The cytoplasmic content of Plk, a prominent regulator of mitotic events such as spindle formation, chromosome segregation, and cytokinesis (21), also diminished after the treatment in a hypericin dose-dependent manner (Fig. 2 A). Mutant p53, cdk4, Raf-1, and Plk are all client proteins of Hsp90, assisted in folding and subsequently stabilized by this chaperone (20).

Similar effects were noted in the SQ2 cell model in which cellular contents of the Hsp90 client proteins Raf-1 and Plk also declined dramatically after treatment with hypericin in the dark (Fig. 2 B). SQ2 cells bear wild-type p53 (not immunostained with PAb 240 anti-p53 antibody). Wild-type p53 has been documented not to form stable complexes with Hsp90 (25) and, indeed, p53 expression in SQ2 cells treated with hypericin was unaltered (data not shown).

The diminution in cytosolic Raf-1 prevents the activating phosphorylation of the downstream p44/42 Erk1/2 MAP kinases. In DA3 cells Erk1/2 phosphorylation dissipated after 72 h of treatment with hypericin, whereas the overall content of Erk1/2 remained unaffected (Fig. 2,C). In SQ2 cells the effect was mainly on Erk1 phosphorylation (Fig. 2 C). Erk1/2 deactivation deprives cells of several active transcription factors downstream to the mitogen-activated Ras/Raf pathway (26), which are essential for cell proliferation.

The temporal relationship between administration of hypericin to DA3 cells and the enhancement of Hsp90 client protein turnover has been examined for Plk using [³⁵S]methionine/cysteine pulse-chase analysis. Cells treated with 10 μm of hypericin were subjected to 1 h of met/cys pulses 12, 24, 48, and 72 h before cell lysis. Plk was immunoprecipitated from the lysates and separated on SDS-PAGE. An autoradiogram of Plk cell content at these different time points is shown in Fig. 2,D. The results indicate that whereas Plk synthesized between 12 and 48 h, posthypericin administration increased in content and peaked at 48 h compared with untreated controls; Plk pulsed 72 h before cell lysis was almost completely degraded at an accelerated pace (Fig. 2,D, right panel). These data suggest that the treatment with hypericin shortened the half-life of Plk, however, only within the last 24 h. The transcription of Plk remained unchanged throughout the 72-h cell treatment with hypericin, (Fig. 2 E).

To determine whether the enhanced turnover of Hsp90 client proteins in hypericin-treated cells involves the proteasome pathway, we examined the effects of combinations of hypericin and the MG132 proteasome inhibitor on the cellular content of the Hsp90 client proteins Plk and Raf-1 in detergent soluble and insoluble fractions. Fig. 3,A shows that addition of 100 nm of MG132 administered for the last 48 of 72 h cell treatment with hypericin (the highest nonlethal dose of MG132 in a 48 h treatment schedule) enhanced hypericin-induced turnover of each of the two Hsp90 client proteins in the NP40 soluble fraction. Whereas MG132 alone induced accumulation of these client proteins in the NP40 insoluble fraction, no such accumulation was noted in combination with hypericin (Fig. 3 B). The presence of MG132 was not found to affect the MCD phenomenon induced by hypericin (data not shown).

The destabilization and enhanced turnover of Hsp90 client proteins was associated with marked declines in the cellular content of additional cell cycling regulators. Cyclin A, cyclin B1, and cyclin H (the regulatory subunit of cyclin activating kinase; Refs. 27, 28) diminished in a hypericin dose-related manner (Fig. 4), whereas Cdk1, Cdk2, and Cdk7 remained unaltered. The cell content of p27^kip1 also decreased and was associated with elevated cyclin E, predominantly in the nuclei (Fig. 4). This latter combination may have contributed to diminishing residual G₁ checkpoint activity and facilitated cell progression to S phase after treatment with hypericin. mRNA levels of cyclin B₁, cyclin H, Cdk4, and p27^kip1 were unchanged by the action of hypericin, indicating that reductions in the cellular content of these proteins were not an outcome of diminished transcription (data not shown).

To identify the causes for hypericin-induced diminution in cellular content of the Hsp90 client proteins, we examined possible effects of this compound on intracellular Hsp90. The protein was immunoprecipitated from cytosolic fractions of hypericin-treated DA3 and SQ2 cells with anti-Hsp90 antibody and Western blots developed using an antiubiquitin monoclonal antibody.

In DA3 cells, Hsp90 was seen to undergo extensive ubiquitinylation (predominantly monoubiquitinylation), which became evident after 48 h of treatment with 10 μm (Fig. 5,A, left top panel) and after 72 h of treatment with ≥5 μm hypericin (Fig. 5,A, right top panel). Decreases in Hsp90 became noticeable only after 72 h of cultivation with ≥5 μm of this compound (Fig. 5,A, 2nd right panel). Stripping the nitrocellulose membranes and reprobing with anti-p53 antibody revealed diminished mutant p53 protein in complex with Hsp90 (Fig. 5,A, bottom panel) concomitantly with the ubiquitinylation of Hsp90 (Fig. 5,A, top panel) and before its own turnover (Fig. 5,A, 2nd right panel). Transcription levels of Hsp90 in the cells remained unchanged throughout 72-h treatment periods with hypericin as shown by RT-PCR (Fig. 5 C).

Enhancement of Hsp90 ubiquitinylation induced by hypericin was also noted in SQ2 cells. Exposure to hypericin for 72 h resulted in hypericin dose-dependent increases in the amounts of monoubiquitinylated Hsp90 (Fig. 5 B, top panel).

We next examined whether ubiquitinylation of Hsp90 affects its chaperonal functioning, analyzed as capability of Hsp90 to associate with the p50^cdc37 cochaperone. Cytosolic fractions of DA3 and SQ2 cells treated with 10 μm hypericin for 48–72 h, confirmed to contain ubiquitinylated Hsp90, were immunoprecipitated with anti-p50^cdc37 antibody. The ability of this cochaperone to pull down Hsp90 was examined by immunoblotting with anti Hsp90. Fig. 5,D (top panels) shows that the treatment with hypericin strongly diminished the interaction of p50^cdc37 with Hsp90 in both cell lines. The overall contents of p50^cdc37 and also Hsp90 were unaltered in these two cell lines (Fig. 5 D, middle and bottom panels).

The ubiquitinylation of Hsp90 by hypericin prompted evaluation of possible effects of this compound on Hsp70 chaperones. There was no evidence for ubiquitinylation of the heat shock cognate protein (Hsc70), a member of the Hsp70 family of chaperones, after cell treatment with hypericin neither for 48 h nor for 72 h in DA3 or SQ2 cells (data not shown). Conversely, Hsc70 increased in content in hypericin-treated DA3 cells at doses that induced Hsp90-ubiquitinylation at the 48- and 72-h time points (Fig. 5,A, 3rd left and right panels, respectively). In SQ2 cells the levels of Hsc70 remained unchanged (Fig. 5 B, bottom panel). Thus, hypericin-mediated chaperone ubiqutinylation appears to be unique to Hsp90.

MCD, which can be generated by the action of exogenous agents, occurs after diminution in functions of several cellular signaling pathways. Their effects neutralize the G₁-S checkpoint, delay cells at G₂-M, lead to formation of enlarged cells, and generate multinucleation (5). The different pathways that elicit MCD appear to be unrelated and to involve mutually exclusive molecular players, requiring the existence of a mechanism that can link together these different aberrant functions.

We found that hypericin induces MCD in various tumor cell lines after treatment for 48–72 h in the dark. This model was used to identify mechanisms that can generate the multitargeted deficiencies leading to MCD.

We show that cell exposure to hypericin accelerates turnover and leads to diminution in the content of Cdk4, mutant p53, Raf-1, and Plk (Fig. 2, A and B); others noted reduced cellular levels of ErbB2 in ovary carcinoma cells (11). All are client proteins of the Hsp90 chaperone. Catabolism of Hsp90 client proteins also affected their downstream signaling targets as shown for Raf-1, the turnover of which resulted in diminished activation of Erk type MAP kinases (Fig. 2 C). Because Hsp90 client proteins participate in G₁ cyclin activities, G₁-S cell cycle phase transition checkpoints, and various cellular signaling pathways, reducing their levels can promote cell entry into aberrant cycling, lead to premature mitosis (19, 20), and culminate in MCD.

A prominent role in cell failure to complete M phase-related events in MCD is played by Plk. Plk regulates centrosome maturation (29, 30), bipolar spindle formation (30, 31), microtubule-contraction, cdc2 activation (32), anaphase promoting complex activity (33), and cytokinesis (34, 35). Depletion of Plk or mutations in its substrates have been shown to increase the frequency of multinucleated cells (30, 36). Destabilization of Plk after treatment of tumor cells with hypericin (Fig. 2) appears to generate similar deficiencies in Plk-dependent functions that culminate in MCD.

Another Hsp90 client protein, the destabilization of which can potentially contribute to mitotic catastrophe, is Wee1. This cell cycle checkpoint regulator acts to protract G₂ and prevent immature cell entry into mitosis (37). Loss of Wee1 kinase activity can increase the propensity of cells to enter premature mitosis and consequently to generate the MCD phenomenon.

Analyses of the causes for inactivation of Hsp90 chaperone activity, also exemplified by diminished interaction between Hsp90 and the p50^cdc37 cochaperone (Fig. 5,D), revealed that hypericin induces ubiquitinylation of Hsp90 (Fig. 5, A and B). This effect appears to be exclusive to Hsp90 and not to affect Hsp70 chaperones. Conversely, cell content of the cognate Hsc70 increased in hypericin-treated cells in a manner similar to the effects of geldanamycin, another Hsp90 inhibitor (38).

The hypericin-induced enhanced turnover of Hsp90 client proteins was accompanied by changes in the cellular content of several additional positive cell cycle regulators. Cyclin A, cyclin B1, and cyclin H were found to decline, whereas the content of cyclin E was elevated (Fig. 4). Similar modulations in cell cycle regulators were reported after cell treatment with depsipeptide FR901228 (FK228), a histone deacetylase inhibitor (39). Depsipeptide inactivates Hsp90 by acetylation, leading to depletion of Hsp90 client proteins. These changes are also accompanied by a secondary decrease in cyclin A and elevation of cyclin E (39).

The diminishing cellular content of p27^kip1 observed in this study (Fig. 4) may result from the G₂-M phase arrest caused by hypericin. p27^kip1 was reported to interact with Hsc70 in a cell cycle-dependent manner with maximal binding at the G₁-S transition (40). Prolonged cell sequestration in S phase and G₂-M may have prevented this interaction and accelerated p27^kip1 degradation (41).

Thus, hypericin emerges as a new class of Hsp90 inhibitors that differs from others in the mode by which it affects this chaperone and in the form of cell death that it induces. Unlike geldanamycin and radicicol, which inactivate Hsp90 through interaction with the ADP/ATP binding pocket (18, 38, 42, 43), or depsipeptide (FK228), which induces Hsp90 acetylation (39), hypericin inactivates the chaperone via ubiqutinylation.

An additional dissimilarity between the activities of hypericin and other Hsp90 inhibitors is in the mechanism of protein elimination. In contrast to geldanamycin and radicicol, in which Hsp90 client proteins can be rescued in the presence of MG132 (44), administration of this proteasome inhibitor to hypericin-treated cells is shown to have no effect on accumulation of Hsp90 client proteins in detergent-insoluble fractions (Fig. 3 B) as seen with other Hsp90 inhibitors. These differences may result from reduction in intracellular pH known to be elicited by hypericin (15). The reduced pH appears to activate lysosome-mediated protein degradation. Alternatively, proteasome susceptibility to MG132 may diminish under the reduced pH conditions.

The profound differences between the effects of hypericin and those of other Hsp90 inhibitors also extend to the mode of cell death that is induced. Whereas geldanamycin and depsipeptide mainly induce apoptosis (18, 39), hypericin is shown here to cause MCD. The reasons for these disparities are at present unclear. The reduced intracellular pH caused by hypericin might affect caspase activation. It may also be the cause for the reduced catabolism of Plk seen in the pulse/chase experiments during the first 48 h (Fig. 2,D), whereas the dramatic shift to accelerated catabolism noted at 72 h may be the outcome of Hsp90 ubiquitinylation, noted to occur after 48–72 h of exposure to hypericin (Fig. 5, A and B).

The anticancer activities of hypericin in the dark may, thus, be viewed in the context of inducing Hsp90 inactivation via ubiquitinylation, leading to destabilization of Hsp90 client proteins culminating in MCD.