The phytochemical piceatannol induces the loss of CBL and CBL-associated proteins

Piceatannol is a small molecule that was initially isolated as the antileukemic agent from the domesticated oilseed Euphorbia lagascae (1). In mammals and mammalian cell culture, piceatannol has also been shown to have beneficial effects as an anti-inflammatory agent (2, 3), an anti-histamine (4, 5), and a general anticancer agent (1, 6–8). Piceatannol is thought to accomplish these varied effects through its inhibition of specific tyrosine and serine/threonine protein kinases. Piceatannol was initially shown to be, and is still commonly used as, a Syk-selective protein tyrosine kinase (PTK) inhibitor (9). It can also inhibit other tyrosine kinases including Src, Lck, and FAK, albeit with lower efficiency (9–11). More recently, piceatannol has been shown to inhibit several serine/threonine kinases (12, 13). The cellular effects of piceatannol are not limited to kinase inhibition; it has been shown to induce apoptosis, which may be related to its inhibition of mitochondrial F₀F₁-ATPase activity (14, 15), and to induce DNA damage (16–18). Piceatannol has also been shown to have antioxidant properties (19, 20).

Despite the extensive characterization of its properties and cellular effects, piceatannol has never been associated with the loss of specific proteins. Herein, we show that piceatannol induces the loss of the Cbl family of proteins and specific Cbl-associated PTKs in a reactive oxygen species (ROS)-dependent manner.

The c-Cbl proto-oncogene is an adaptor protein and a RING finger E3 ubiquitin ligase; as such, it acts as both positive and negative regulators of many PTKs. On activation of receptor tyrosine kinases (RTK), c-Cbl is recruited from the cytosol to the activated receptor complex where it recruits signaling effector proteins and positively regulates signaling by acting as an adaptor protein. Cbl then ubiquitinates these receptors as well as many of their associated signaling proteins (21), causing them to be degraded by the proteasome and/or lysosome (21). Cbl itself is also regulated by ubiquitin-mediated proteolysis (21–24). Thus, through its E3 ubiquitin ligase activity, Cbl also functions as a potent negative regulator of PTK signaling.

We have found that piceatannol dramatically reduces cellular Cbl protein levels independently of the proteasome, the lysosome, and caspase activation. By screening a small library of piceatannol-like compounds, we determined that compound-induced Cbl protein loss was mediated directly by aromatic compounds that contain catechol rings through their oxidative conversion into a highly reactive O-benzoquinone. We further characterized the protein selectivity of this process and established the applicability of these observations to cancer treatment. This work characterizes a previously unrecognized piceatannol-dependent effect and shows a novel way in which this effect can be exploited as a therapeutic to inhibit disease-associated signaling pathways.

Murine 3T3 fibroblasts overexpressing wild-type c-Cbl (wtCbl), 70ZCbl, and p95Cbl were generated and maintained as described previously (25). The NH₂-terminal HA-tagged v-Cbl allele was generously provided by Dr. Wallace Langdon and cloned into pBabepuro3 as described previously for p95Cbl (25). The Cbl deletion mutant Δ1-355Cbl, a c-Cbl NH₂-terminal truncation mutant that begins at codon 356, was generated by PCR using pBabe/wtCbl(N) and a NH₂-terminal primer introducing a XhoI site in front of c-Cbl codon 356 [primer XhoCbl356(s) 5′-CCTCGAGACTCCCCAAGACCATATC-3′ and primer Cbl6(a) 5′-TCTCTGGAGGGACAGTCGC-3′]. This PCR product was digested with XhoI and BglII and ligated into XhoI/BglII-digested pBabe/wtCbl(N). The resultant plasmid, pBabe/Δ1-355Cbl, was stably expressed in murine 3T3 fibroblasts as above (25). The mouse embryonic fibroblasts were kindly provided by Dr. Hamid Band (26). The mouse embryonic fibroblasts and the A431 cells were maintained similar to the 3T3 fibroblasts in the absence of any selection. The 70Z/3 murine pre-B-cell line and the K562 human erythroleukemia cell line were maintained as described previously (25).

Piceatannol and resveratrol were purchased from two independent suppliers (Sigma and Biomol) to ensure specificity and reproducibility. Compounds 1 to 37 were supplied by Dr. Henrik Hansen and Karm Hans. All new compounds were identified by nuclear magnetic resonance and mass spectrometry. Based on the nuclear magnetic resonance spectra, the compounds had a purity of >95%. All stilbene analogues were prepared and tested as single isomers. All of the piceatannol-like compounds were solubilized in DMSO at a stock concentration of 200 mmol/L and stored as individual aliquots at −20°C until use. Geldanamycin (Alamone Labs), herbimycin A (Life Technologies), curcumin (Sigma), (CBZ-Phe-Arg)₂-R110 (Molecular Probes), and epigallocatechin gallate (Biomol) were made up at stock concentrations in DMSO. Ammonium chloride (BDH/VWR) was dissolved in 10 mmol/L Tris-HCl (pH 7.4) at a stock concentration of 2 mol/L and N-acetylcysteine (Sigma) was dissolved in PBS at a stock concentration of 100 mmol/L. STI571 was a kind gift from Dr. Aru Narendran (Tom Baker Cancer Centre). All other chemicals were obtained from Calbiochem and made up at stock concentrations in DMSO.

Murine 3T3 fibroblasts ectopically expressing Cbl, mouse embryonic fibroblasts, and A431 cells were plated in six-well plates with 5 × 10⁵ per well in 2 mL supplemented DMEM and grown overnight before their use in experiments. K562 and 70Z/3 cells were resuspended at 8 × 10⁵/mL in complete RPMI in a total volume of 4 mL in 5 mL round-bottomed polystyrene tubes the day they were used in experiments. Six-well plates with 5 × 10⁵ Cbl-expressing 3T3 cells per well were grown in 2 mL supplemented DMEM overnight before transfection. Transfection with pMT/HA-ubiquitin, a kind gift of Dr. Jane McGlade (Hospital for Sick Children, Toronto, Ontario, Canada), was accomplished using Fugene 6 (Roche) according to the manufacturer's protocol. Twenty-four hours after transfection, cells were treated with MG132 and/or piceatannol. The specific details regarding each experiment are included in the figure legends. At the indicated times, cells were lysed in NP-40 lysis buffer as described previously (25). Alternatively, cells were lysed in hot 2× SDS Laemmli's sample buffer after being washed in PBS.

Immunoprecipitations and Western blots were done as described previously (25) using the following antibodies: 7G10 monoclonal antibody (mAb) against c-Cbl (Upstate), rabbit anti-Cbl polyclonal antibody (Santa Cruz Biotechnology), 12CA5 mAb against HA, anti-actin mAb1501 (Chemicon), C2-10 mAb against poly(ADP-ribose) polymerase (Trevigen), rabbit anti-platelet-derived growth factor receptor (PDGFR) polyclonal antibody (Santa Cruz Biotechnology), rabbit anti-Syk (Santa Cruz Biotechnology), rabbit anti-p56Lyn polyclonal antibody (kind gift from Dr. Anthony L. DeFranco, University of California at San Francisco), rabbit anti-Abl (Cell Signaling Technology), Src327 mAb against Src (kind gift from Dr. J. Michael Bishop, University of California at San Francisco), anti-Grb2 mAb (Transduction Labs), and anti-ubiquitin (Zymed).

RNA was isolated with TRIzol reagent (Invitrogen) using the manufacturer's protocol. RNA was then DNase treated with 0.5 units RQ1 RNase-free DNase (Promega) in the presence of 20 units RNase Block (Invitrogen) for 30 min at 37°C. RNA was phenol/chloroform extracted and 2 μg RNA was used to make cDNA using Superscript II (Invitrogen) following the manufacturer's protocol. Reverse transcription-PCR was done using the primer drop method (27) using the Cbl4 (5′-GCCCTGACCTTCTGATTCCTGCCA-3′) and MiddleCblF (5′-GATTGATGGCTTCAGGGAAGG-3′) primers to detect c-Cbl/70ZCbl and the mGAPDH-F (5′-ACCACAGTCCATGCCATCAC-3′) and mGAPDH-R (5′-TCCACCACCCTGTTGCTGTA-3′) primers to detect GAPDH.

Mouse fibroblast cells (7.5 × 10⁵ per well) stably overexpressing 70ZCbl were plated in six-well dishes. The following day, cells were treated for 4 h with lysosomal inhibitors and then trypsinized from the plates. Trypsin was neutralized in 10 mL culture medium and cells were stained with (CBZ-Phe-Arg)₂-R110 and propidium iodide (Sigma) and analyzed by flow cytometry using the manufacturer's protocol. Propidium iodide-positive cells were subtracted from the (CBZ-Phe-Arg)₂-R110-positive cells for the determination of relative cathepsin activity. The average means of (CBZ-Phe-Arg)₂-R110 intensity of five experiments were normalized to their respective untreated controls. Values were then represented as relative percentages of normal cathepsin activity.

Cbl was originally identified based on its homology with the retrovirally encoded v-Cbl oncogene (28). In addition, we and others have identified two naturally occurring transforming mutant isoforms of c-Cbl: p95Cbl and 70ZCbl (25, 29). These mutant c-Cbl proteins contain deletions of the RING finger, or of areas immediately proximal to the RING finger, which cripple their E3 ligase activity (21, 25, 29). These transforming deletion mutations are also associated with the constitutive phosphorylation of several COOH-terminal tyrosine residues (25, 29). While examining the role of tyrosine phosphorylation in the ability of 70ZCbl to transform cells, we observed that, unlike other tyrosine kinase inhibitors, piceatannol treatment of 70ZCbl-overexpressing murine 3T3 fibroblasts caused the loss of Cbl protein (Fig. 1A). We further examined the effect of piceatannol on endogenous c-Cbl and Cbl-b protein levels in murine 3T3 fibroblasts as well as the effect on Cbl protein levels in murine 3T3 fibroblasts overexpressing wtCbl, 70ZCbl, p95Cbl, v-Cbl, and the NH₂-terminal truncation mutant Δ1-355Cbl. All of the Cbl proteins, endogenous, wild-type, and mutant, were lost on treatment with piceatannol as shown by Western blotting (Fig. 1A).

To ensure that the loss of Cbl was not due to an artifact of the lysis or Western blotting procedures, wtCbl-overexpressing murine 3T3 fibroblasts were lysed in NP-40 lysis buffer or hot SDS sample buffer and both soluble and pellet fractions of the NP-40 lysates were loaded on a SDS-PAGE gel along with the SDS lysates. The stacking gel was left intact and transferred to nitrocellulose along with the resolving gel to ensure that Cbl was not lost due to an inability to enter into the resolving gel (Fig. 1B).

To verify that the observed piceatannol-induced Cbl loss was not due to the modification or loss of the epitope recognized by the c-Cbl-specific mAb, Western blots of HA-tagged 70ZCbl-overexpressing murine 3T3 fibroblasts lysates were reprobed with a polyclonal antibody directed toward the COOH terminus of 70ZCbl as well as a monoclonal anti-HA antibody that recognized the HA tag fused to the NH₂ terminus of 70ZCbl. All three antibodies showed that the epitope-tagged 70ZCbl was lost in a dose-responsive manner in cells treated with increasing concentrations of piceatannol (Fig. 1C).

The piceatannol-induced loss of Cbl protein levels, while visible at time points earlier than 30 min, was most dramatic after ≥3 h (data not shown). This timeframe for piceatannol-induced loss suggested the possibility that this effect could be at least partially mediated at a transcriptional level. RNA was isolated from 70ZCbl expressing fibroblasts at 1, 3, 8, and 24 h following piceatannol treatment. Reverse transcription-PCR for Cbl, normalized to the loading control GAPDH, showed no decrease in the level of Cbl transcript in piceatannol-treated cells (Fig. 1D). These data indicated that the observed reduction of Cbl was not due to piceatannol-induced changes in Cbl transcription or Cbl transcript stability. Together, these data suggested that piceatannol is mediating its effect on Cbl at the protein level.

A distinct smearing of Cbl proteins was often observed on overexposed Western blots following exposure to piceatannol (Fig. 2A, right). There is evidence that c-Cbl is degraded in a ubiquitination-dependent manner (22–24) and it is conceivable that piceatannol promotes this event. To examine whether piceatannol was inducing the ubiquitination of Cbl, wtCbl-overexpressing murine fibroblasts, transfected with HA-tagged ubiquitin, were pretreated with the proteasome inhibitor MG132 before piceatannol treatment. By Western blotting with an anti-HA antibody, HA-ubiquitin was only detected on Cbl immunoprecipitated from cells treated with MG132 alone (Fig. 2A, left). Reprobing these Western blots for Cbl (Fig. 2A, right) also showed that MG132 did not induce the same pattern of Cbl smearing as seen after piceatannol treatment. We also tested whether inhibition of the proteasome and the lysosome with pharmacologic inhibitors could prevent the piceatannol-induced loss of 70ZCbl. Preincubation of 70ZCbl-overexpressing murine fibroblasts with the proteasome inhibitors ALLN, MG132, and epoxomycin or the lysosomal inhibitors bafilomycin A1, ammonium chloride, and EST was not sufficient to prevent the piceatannol-induced loss of 70ZCbl (Fig. 2B and C). The pharmacologic agents were used at effective doses, as shown by the presence of ubiquitin laddering (Fig. 2B, bottom) and by the 60% reduction in fluorescence of the lysosomal protease substrate (CBZ-Phe-Arg)₂-R110 (Fig. 2C, bottom).

Apoptosis-induced caspase activation has been shown to induce the degradation of Cbl (30). To examine whether the loss of Cbl was due to the induction of apoptosis, 70Z/3 cells were treated with several agents that promote apoptosis including hydrogen peroxide, UV radiation, and γ-radiation. All of the agents induced the caspase-mediated cleavage of poly(ADP-ribose) polymerase in the 70Z/3 cells; however, only piceatannol was able to induce the loss of Cbl (Fig. 2D).

To directly examine the potential role of caspases in piceatannol-mediated Cbl loss, 70Z/3 cells were treated with the broad-spectrum caspase inhibitor Boc-d-FMK before the addition of piceatannol. Although the use of this caspase inhibitor produced a dose-responsive inhibition of piceatannol-induced poly(ADP-ribose) polymerase cleavage, it was unable to prevent the reduction of Cbl protein levels (Fig. 2E). Taken together, these data suggested that a previously uncharacterized Cbl-regulatory mechanism was induced on treatment with piceatannol.

To further confirm that piceatannol was not affecting normal mechanisms of Cbl protein regulation to induce Cbl loss, we next examined the properties attributed to piceatannol to see which, if any, were involved in reducing Cbl protein levels. These properties include the ability of piceatannol to inhibit kinases, to induce apoptosis, and to act as an antioxidant. As we have already examined and excluded the ability of piceatannol to induce apoptosis as a cause for Cbl loss (Fig. 2D and E), we next examined whether kinase inhibition was involved. 70ZCbl-overexpressing murine fibroblasts were treated with a wide variety of tyrosine and serine/threonine kinase inhibitors. After 3 h treatment, only piceatannol was observed to induce a dramatic decrease in amount of Cbl detected without affecting actin levels (Fig. 3A, lanes 1P and 2P).

To further confirm this result, we next determined if the tyrosine phosphorylation status of c-Cbl played a role in piceatannol-induced Cbl loss. The K562 human erythroleukemia cell line was used to test this possibility, as these cells exhibit Cbl loss on treatment with piceatannol and they contain high levels of constitutively tyrosine-phosphorylated c-Cbl due to the presence of the Bcr-Abl oncogene (29, 31). After 8 h pretreatment with the Abl tyrosine kinase inhibitor STI571, c-Cbl tyrosine phosphorylation levels in the K562 cells were reduced to undetectable levels (Fig. 3B, top). A further 12 h treatment of these cells with piceatannol resulted in a reduction in the amount of c-Cbl protein that was comparable with that observed in cells treated with piceatannol alone (Fig. 3B, bottom).

The antioxidant capacity of piceatannol was also examined to determine if it was involved in piceatannol-induced Cbl loss. This was accomplished by comparing the piceatannol treatment of 70ZCbl-overexpressing murine fibroblasts with several different antioxidants including resveratrol, N-acetylcysteine, curcumin, and epigallocatechin gallate. Western blot analysis indicated that only piceatannol was able to induce Cbl loss (Fig. 3C). Collectively, these data indicate that individual properties commonly associated with piceatannol are not responsible for the induction of Cbl loss.

Our results indicated that piceatannol was inducing the loss of Cbl in treated cells independently of previously characterized Cbl protein regulatory mechanisms and that Cbl loss was due to a heretofore unrecognized property of piceatannol. They also indicated that the structurally similar compound, resveratrol, was unable to mediate this effect (Fig. 3C). To gain a better understanding of the mechanism by which piceatannol was inducing Cbl loss, we screened several piceatannol-like compounds in an attempt to determine the chemical structure(s) required to induce Cbl loss. The piceatannol-like molecules used had many chemical and structural similarities to piceatannol including aromaticity, size, presence of hydroxyl groups, and hydrophobicity (Fig. 4). 70Z/3 cells were used to screen the compounds, as these cells have shown the highest degree of sensitivity to piceatannol-induced Cbl loss. The cells were incubated with 100 μmol/L of each compound for 1 h to avoid artifacts due to the induction of apoptosis by higher concentrations and later time points, and cell lysates were examined for Cbl protein levels by Western blotting (Fig. 5). Compounds 7, 29, 31, and 33 were all able to induce Cbl loss. However, only compounds 7, 29, and 31 were able to induce the loss of Cbl without affecting actin protein levels, as was the case with piceatannol (Fig. 5). The common feature of these chemical compounds was the presence of an ortho-catechol ring structure. In piceatannol and in compound 29, the catechol rings are unprotected, whereas, in compounds 7 and 31, the catechol rings are protected by acetate esters. As cells contain numerous esterase enzymes, once these compounds enter into cells, they are de-esterified, exposing the catechol ring. The other common feature of these compounds is that they are all small, highly aromatic molecules with limited degrees of freedom. Compounds 7 and 29 are planar trans-stilbenes like piceatannol, whereas compound 31 has rotational freedom around its benzylic sulfur group. Whereas other compounds in the screen possessed these features individually, only piceatannol and these three compounds possessed them in combination, which suggested that these particular chemical and structural properties were indispensable for the induction of Cbl loss.

Due to the requirement for a catechol ring, and our observation that treatment of cells with piceatannol and compound 7 led to the medium and cell pellets acquiring a brown color after treatment (data not shown), we hypothesized that oxidation of the catechol ring into an O-benzoquinone was involved in piceatannol-induced Cbl loss (32, 33). This hypothesis could explain the observed smearing and loss of Cbl as the result of oxidized piceatannol causing protein cross-linking (34, 35) and chemical scission of the peptide backbone (36). To test this hypothesis, we designed an in vitro assay system to test the minimal requirements for piceatannol-induced Cbl loss. Immunoprecipitated Cbl was resuspended in PBS and aliquoted into several different reaction mixtures containing combinations of trans-stilbene, H₂O₂, and horseradish peroxidase (HRP). The H₂O₂ and HRP were included to serve as a source of ROS to promote the oxidation of piceatannol and the other trans-stilbenes. After 15 min at room temperature, the ROS in the reaction mixtures was neutralized by the addition of Laemmli's buffer containing β-mercaptoethanol, and the samples were analyzed by Western blotting. Only the combination of piceatannol, H₂O₂, and HRP was able to induce specific Cbl loss without affecting the levels of IgG heavy chain from the 7G10 antibody used to immunoprecipitate Cbl (Fig. 6A).

To further examine the requirements for piceatannol oxidation in piceatannol-induced Cbl loss, N-acetylcysteine was added to the in vitro assay conditions to neutralize ROS produced by the H₂O₂/HRP system. In the presence of this antioxidant, piceatannol-induced Cbl loss was dramatically decreased (Fig. 6B), further showing the requirement for oxidation of piceatannol in this process.

To verify that piceatannol and piceatannol-related compounds were being converted into O-benzoquinones, we examined the spectral properties of these compounds in the presence and absence of ROS. O-benzoquinones absorb light at a wavelength of 400 nm much more strongly than do catechols (37); therefore, by subtracting the absorbance at 400 nm (A₄₀₀) of an aqueous solution of compound from the A₄₀₀ of the compound solution in the presence of ROS, over time, we would be able to determine if these compounds are becoming O-benzoquinones, as this value, the ΔA₄₀₀, would increase. Aqueous solutions of piceatannol and compound 29 were prepared and their A₄₀₀ was monitored every 5 s for 150 s. These values were subtracted from the comparable A₄₀₀ readings from these compounds in the presence of 10 μmol/L NaIO₄, a source of ROS. Both piceatannol and compound 29 showed increasing ΔA₄₀₀ values over time compared with the negative control resveratrol, indicating the formation of O-benzoquinones (Fig. 6C). Resveratrol was included as a negative control, as this compound is unable to form an O-benzoquinone due to the positions of its hydroxyl groups (Fig. 4). To confirm the requirement for ROS in the conversion of the piceatannol-like compounds into O-benzoquinones, the same experiment was done in the presence of N-acetylcysteine. The presence of this antioxidant completely abrogated the increasing ΔA₄₀₀ over time, indicating that it blocked the formation of O-benzoquinones (Fig. 6C). These data showed that O-benzoquinones were formed when the catechol containing compounds piceatannol and compound 29 were exposed to an oxidant. These experiments show that the oxidation of piceatannol and piceatannol-like compounds induces the formation of O-benzoquinones and that the presence of O-benzoquinones is necessary and sufficient to induce Cbl loss.

As we were unable to identify a region of Cbl that gave it sensitivity to piceatannol (Fig. 1A), it seemed likely that piceatannol could react with many different regions of this protein. Western blots of piceatannol-treated cells examining c-Cbl and the Cbl-associated proteins PDGFRβ, c-Abl, c-Src, Lyn, actin, and Grb2 showed that piceatannol was also able to induce the loss of other proteins. c-Cbl, PDGFRβ, and c-Abl were all lost as a result of piceatannol treatment (Fig. 7A-C). That Lyn, c-Src, Grb2, and action were unaffected by piceatannol treatment also reinforced the observation that there was selectivity to the induction of protein loss of piceatannol (Fig. 7A-C). In addition to the selective loss of proteins in vivo, Lyn shifted to a lower apparent molecular weight after piceatannol treatment (Fig. 7C, 70Z/3). This molecular weight shift is reminiscent of a change in the tyrosine phosphorylation status of the protein and was likely the result of piceatannol-induced kinase inhibition.

To better understand the protein specificity for piceatannol-induced protein loss observed in piceatannol-treated cells, we tested several different immunoprecipitated proteins in our in vitro reaction with piceatannol and/or ROS. We hypothesized that in vitro most proteins would be sensitive to piceatannol-induced protein loss, considering that we have been unable to identify a specific region of Cbl that led to its sensitivity to piceatannol (Fig. 1A) and that there are many different amino acids with nucleophilic atoms in their side chains, none of which are unique to Cbl, which could react with O-benzoquinones including cysteine, methionine, histidine, tryptophan, serine, tyrosine, lysine, arginine, and proline (38). Western blotting of in vitro reactions containing c-Cbl, c-Src, Lyn, or PDGFRβ confirmed that, similar to c-Cbl, all of these proteins were lost in the presence of piceatannol and ROS (Fig. 7D). These data indicated that piceatannol-induced protein loss had a broader specificity in vitro than in cell culture.

Based on our observation that piceatannol treatment led to a loss of PDGFRβ, we were interested in determining if this loss was sufficient to affect how cells responded to the PDGFRβ ligand, PDGF-BB. To test this, we examined the mitogen-activated protein kinase (MAPK) activation status in murine embryonic fibroblasts that were pretreated with piceatannol for 8 h or 10 min before a 10 min stimulation with PDGF-BB (Fig. 7E). MAPK activation was not robustly affected after the 10 min pretreatment, showing that the activity of piceatannol as a kinase inhibitor does not have a major effect on PDGFRβ-induced MAPK activation. The 8 h pretreatment with piceatannol, which induced the loss of both c-Cbl and PDGFRβ, prevented PDGF-BB-induced ERK1/2 phosphorylation (Fig. 7D). These data showed that the piceatannol-induced loss of PDGFRβ protein also induces the loss of PDGFRβ function.

To examine if piceatannol treatment could be used to induce the loss of Cbl-associated oncogenic receptors in a cancer cell model, the A431 human squamous cell carcinoma cell line, which overexpresses epidermal growth factor receptor (EGFR), was treated with increasing concentrations of piceatannol before stimulation with EGF. Western blots of the A431 cell lysates indicated that, at higher concentrations of piceatannol, both Cbl and EGFR protein levels are reduced and that this corresponds to reduced activation of MAPK in response to EGF stimulation (Fig. 7F). This result suggested that the piceatannol-induced loss of Cbl and Cbl-associated signaling proteins may be beneficial as a treatment for a variety of cancers promoted by misregulated Cbl-associated PTKs.

Piceatannol is a naturally occurring small molecule known for its kinase inhibition and anticancer properties (1, 6–11). We have shown that piceatannol is also able to induce the loss of Cbl proteins (Fig. 1). By screening a small library of piceatannol-like compounds for their ability to induce Cbl loss in cells, we determined that the Cbl loss is associated with the catechol ring structure of piceatannol (Figs. 4 and 5). The results of this screen showed that only 3 of 19 ortho-catechol-containing compounds, compounds 7, 29, and 31, are able to induce Cbl loss to a similar extent as piceatannol (Fig. 5), indicating that, in addition to the catechol ring, there are further chemical requirements for Cbl loss. Catechol-containing compounds lacking an aromatic substituent, including compounds 12 and 15, were unable to induce Cbl loss (Fig. 4), indicating that the presence of an aromatic substituent was also important to the process.

These data suggest a mechanism of action in which piceatannol enters into cells and is subsequently oxidized into an electrophilic O-benzoquinone that readily reacts with nucleophilic atoms from the amino acid side chains of Cbl (38). This would then lead to the cross-linking of Cbl, explaining the smearing observed by Western blotting after piceatannol treatment (refs. 34, 35; Fig. 2A, right). The O-benzoquinones can also react with peptide bonds, inducing scission of the polypeptide backbone (36) and leading to the loss of Cbl observed after piceatannol treatment. Further supporting this plausible mechanism of action, using an in vitro reaction that recapitulates the minimum requirements for piceatannol-induced Cbl loss, we show that this phenomenon is directly mediated through the oxidative conversion of piceatannol into an O-benzoquinone (Fig. 6). In addition, similar observations and similar mechanisms of action have been shown for electrophilic α,β-unsaturated ketones, such as curcumin and acrolein (39, 40). These and other α,β-unsaturated ketones have been shown to covalently modify and cross-link proteins such as nuclear factor-κB (39), p53 (41), peroxisome proliferator-activated receptor γ (42), SNAP-25, and VAMP (40). As O-benzoquinones contain an α,β-unsaturated ketone, these observations support the requirement for the oxidation of piceatannol to induce Cbl loss.

Similar to these other α,β-unsaturated ketones, piceatannol also showed a limited protein specificity in cell culture. In addition to inducing the loss of Cbl, piceatannol treatment of cells induced the selective loss of three Cbl-associated proteins: PDGFRβ, c-Abl, and EGFR (Fig. 7). Other proteins were unaffected in piceatannol-treated cells including c-Src, Lyn, Syk, actin, and Grb2. This evidence suggests that piceatannol is associating specifically with structure(s), within cells that have restrictive binding requirements, because not all aromatic catechol-containing compounds were able to mediate protein loss. However, other experimental data argue against such a direct and specific interaction. (a) The absence of a specific domain of Cbl that is responsible for its sensitivity to piceatannol (Fig. 1A). (b) Syk and Src family kinases, proteins known to be functionally inhibited by piceatannol and thus assumed to physically associate with piceatannol (9, 10), are not affected by piceatannol-induced protein loss in cell culture experiments (Fig. 7A-C).

Other factors affecting specificity of protein loss may reside within the microenvironment of the cell including the local concentrations of ROS, which are required to activate piceatannol, the local concentrations of glutathione and other cellular antioxidants, and the compartmentalization of the specific proteins of interest. This is supported by the actions of compound 33, which, unlike piceatannol, already contains an O-benzoquinone structure (Fig. 4). This compound induced nonspecific protein loss (Fig. 5), suggesting that the location of catechol oxidation is involved in the selectivity for the protein loss reaction. Another explanation for the lack of specificity of piceatannol in vitro may be due to the nature of the in vitro experimental conditions. The immunoprecipitation of proteins in NP-40 lysis buffer may alter their conformation such that reactive residues previously hidden become exposed and available to react with oxidized piceatannol.

Considering that several signaling molecules are affected by piceatannol-induced protein loss (Fig. 7) and that several others have their catalytic activity inhibited by piceatannol (Fig. 7C), it is unlikely that Cbl loss is specifically responsible for all of the biological activities of piceatannol. Piceatannol-induced protein loss is more aptly viewed as a marker for piceatannol activity. This is illustrated by the effects of piceatannol on RTK protein levels and signaling (Fig. 7). Whereas the loss of Cbl has been associated with an increase in RTK-induced MAPK activation, piceatannol treatment induced the loss of Cbl, the loss of RTKs, and a decrease in RTK-induced MAPK signaling (Fig. 7E and F). It is also illustrated by piceatannol-induced apoptosis. Cellular sensitivity to piceatannol-induced Cbl-loss correlates with the sensitivity to piceatannol-induced apoptosis in several different cell lines (data not shown). However, these effects were independent of the isoform of Cbl present. Logically, the loss of a transforming Cbl mutant should make cells more susceptible to apoptosis and the loss wtCbl, a negative regulator of cell growth and survival, would promote cell survival. This suggests that a common piceatannol-sensitive element between the cell lines was responsible for the induction of apoptosis and also suggests that piceatannol-induced Cbl loss is an effective marker of the biological activity of piceatannol.

The ability of piceatannol to induce the loss of specific signaling proteins has implications for its use in the treatment of disease. The piceatannol-induced loss of RTK-mediated MAPK activation (Fig. 7E and F) and the piceatannol-induced loss of Cbl (Fig. 1) may be effective for the treatment of cancers in which these piceatannol-sensitive proteins are involved. In addition, the inhibition of multiple PTKs, which may be possible through the use of piceatannol as a therapeutic, has proven to be more effective as a cancer treatment than the inhibition of single PTKs (43–45). There also exists the possibility for synergy between piceatannol and several current therapeutics that are known to produce ROS (46–50). Although piceatannol has been studied as an anticancer agent, a lack of mechanistic insight has prevented any further development. Our methods and our data identify the active chemical structures involved in piceatannol-induced protein loss providing a platform from which more effective piceatannol-like compounds can be designed and tested (Figs. 4 and 5).

We have made the observation that piceatannol treatment of cells causes a reduction in protein levels of Cbl and specific Cbl-associated proteins. This work suggests a mechanism of piceatannol action in which oxidized piceatannol reacts with nucleophilic atoms from the amino acid side chains of Cbl, leading to protein cross-linking, and reacts with the peptide backbone of Cbl, inducing protein loss. The ability of piceatannol to induce protein loss also has important functional consequences with respect to its use as a kinase inhibitor: if piceatannol is to be used as a specific kinase inhibitor, it must be used at low concentrations for short treatment periods to avoid the potentially overlapping effects of Cbl and Cbl-associated protein loss. Cbl and its associated proteins are involved in many of the same signaling pathways as Syk and Src family kinases, and as such, this work also offers a strong cautionary note on the interpretation of the involvement of these PTKs when relying exclusively on the use of piceatannol as a tyrosine kinase inhibitor.

No potential conflicts of interest were disclosed.

We thank Laurie Robertson (University of Calgary Flow Cytometry Core Facility) for expertise and technical help and Erin Côté for help in optimizing the apoptosis assays.