Alternative Cytotoxic Effects of the Postulated IGF-IR Inhibitor Picropodophyllin In Vitro

Growth factors and their receptors play crucial roles in the establishment and development of cancer. The insulin-like growth factor receptor 1 (IGF-IR) is important in the transformation and growth of malignant cells and in preventing programmed cell death (apoptosis; ref. 1). It is often overexpressed in malignant tumors, such as breast, prostate, and lung cancers, as well as myelomas and malignant melanomas (2). The binding of IGF-I to the α-subunit of IGF-IR results in the autophosphorylation of tyrosine residues in the intracellular β-subunit of the receptor, followed by the activation of the receptor kinase. This enables the downstream activation of signaling cascades that are important for proliferation and survival, including the phosphoinositide 3-kinase (PI3K)/Akt and Ras/Erk1/2 mitogen-activated protein kinase (MAPK) pathways. Notably, these pathways are also activated downstream of most tyrosine kinase receptors, including the EGF receptor (EGFR). Like IGF-IR, EGFR is important in the control of cellular growth, proliferation, and differentiation (3). Moreover, it is often strongly expressed in esophageal cancer cells and has been associated with poor prognosis (4).

Although several inhibitors of EGFR signaling have been developed and are in current clinical use, the development of IGF-IR inhibitors has not yet produced any drugs with clinical benefits. Several recently reported phase III clinical trials of anti-IGF-IR antibodies have yielded disappointing results, clearly showing that the tested antibodies do not have substantial favorable effects in unselected patients with cancer (5). Unanticipated and dose-limiting toxicity accompanied by adverse metabolic adverse events caused significant problems in several of these trials, which prompted the abandonment of several antibodies and small-molecule tyrosine kinase inhibitors. It has also been suggested that the efficacy of targeted agents may be impacted by cross-talk between pathways and downregulation of negative feedback loops (6).

The cyclolignan picropodophyllin (PPP) has been described as a potent and selective inhibitor of tyrosine phosphorylation of the IGF-IR, with low systematic toxicity. According to literature reports, PPP efficiently blocks IGF-IR activity, reduces the phosphorylation of Akt (pAkt) and extracellular signal–regulated kinase (Erk) 1 and 2 (pErk1/2), induces apoptosis in cultured IGF-IR–positive tumor cells, and causes complete tumor regression in xenografted and allografted mice (7). PPP does not affect the insulin receptor or compete with ATP in an in vitro kinase assay, suggesting that it may inhibit IGF-IR autophosphorylation at the substrate level (8). Furthermore, PPP specifically blocks the phosphorylation of a tyrosine residue, Tyr1136, within the activation loop of the IGF-IR (9). PPP is currently being investigated in clinical trials (under the name AXL1717) and the results that have been presented to date suggest that it has at least some useful clinical activity and exhibits only modest toxicity (10–12). As such, it is quite different to previously tested IGF-IR antagonists in both respects.

PPP is the C2 epimer of podophyllotoxin (PPT; structures shown in Supplementary Fig. S1), a naturally occuring compound isolated from several species of Podophyllum. PPT inhibits microtubule assembly by binding to the colchicine-binding site of tubulin (13) and has been evaluated as an anticancer agent as well as an agent for treating venereal warts. On the basis of their ID₅₀ values, PPP is at least 50 times less effective than the diasteromer PPT at inhibiting microtubule assembly (14). The 2 epimers seem to spontaneously interconvert until equilibrium is reached. In the presence of mildly basic catalysts, PPT epimerizes smoothly to PPP. However, it has been reported that the PPP-to-PPT conversion does not proceed so readily (15). The underlying mechanism of PPP-induced apoptotic cell death, including the absence of tubulin-affecting activity at relevant concentrations has been the subject of a scientific controversy (16). In this respect, it is remarkable that while there have been more than 30 publications discussing the effects of PPP on the IGF-IR signaling, very few have mentioned the PPP to PPT equilibrium. Because PPT is a known and very potent inhibitor of microtubule assembly, even small amounts of PPT in samples (which could theoretically be formed during the experiments) could potentially have dramatic effects on experimental results, especially in assays conducted using living cells.

This report summarizes our experiences with PPP in various cell lines and in vitro assays. The work presented was conducted using cells of several different histologies, using different methods and with PPP that was chemically pure at the beginning of experiments. In silico methods were used to investigate the potential direct interaction of PPP and tubulin. The results obtained are compared with data for established tubulin inhibitors, PPT, and a low-molecular weight IGF-IR tyrosine kinase inhibitor, AEW541.

The human esophageal squamous cell carcinoma (ESCC) cell lines Kyse30, Kyse70, Kyse140, Kyse150, Kyse180, Kyse410, Kyse450, Kyse510, and Kyse520 (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH) were cultivated in RPMI-1640 medium, supplemented with 10% FBS, 2 mmol/L l-glutamine, 50 U penicillin, and 50 μg streptomycin/mL (Sigma-Aldrich Sweden AB). Cells were split twice per week and they were used for further experiments after reaching 70% to 90% confluency. Cell counting was done with a Coulter Z2 particle counter and size analyzer. Data from additional cell line panels are presented in the Supplementary Data. Because of the multitude of different cell lines used (n = 27), their diverse origins, and the lack of conclusions linked to a specific diagnosis, no authentication experiments were carried out.

AEW541 was kindly donated by Novartis. Cisplatin, 5-fluorouracil (5-FU), gemcitabine, etoposide, docetaxel, vincristine, and vinorelbine (Apoteket AB) were used as standard agents These standards were dissolved as specified by the manufacturer and further diluted with sterile PBS (Sigma). PPP was kindly donated by Prof. Magnus Axelson (Karolinska Institute, Solna, Sweden). PPT was purchased from Sigma-Aldrich. PPP and PPT were dissolved in dimethyl sulfoxide (DMSO) and then diluted further with sterile PBS immediately before use. PPT and PPP in cell growth medium were analyzed by high-performance liquid chromatography following purification (analysis conducted by Prof. Magnus Axelson, details of method will be published elsewhere).

Cell lysates were prepared according to Lennartsson and colleagues (17). Briefly, total protein concentration was determined using the BCA Protein Assay Kit (Pierce). Total cell lysates (TCL) were submitted to SDS-PAGE. Immunoprecipitation was achieved by adding antibodies against IGF-IRβ to each lysate at a concentration of 1 μg/mL. Protein A beads were then added to collect the resulting immunocomplexes. The beads were then washed and boiled in reducing sample buffer and the liberated proteins were separated by SDS-PAGE. The separated proteins were electrotransferred to polyvinylidene difluoride (PVDF) membranes (Millipore), which were then blocked with 5% bovine serum albumin and incubated with the appropriate primary antibody overnight at 4°C. The primary antibodies were IGF-IRβ (C-20; Santa Cruz Biotechnology), anti-phosphotyrosine PY99 (Santa Cruz Biotechnology), and anti-actin mouse monoclonal antibodies (Sigma), and were used at the concentrations recommended by the suppliers. After washing, the membranes were incubated with horseradish peroxidase–conjugated anti-mouse or anti-rabbit immunoglobulin G (IgG) antibodies (Amersham Biosciences), and the proteins were visualized using enhanced chemiluminescence (ECL) Western blotting detection systems from Roche Applied Science with a cooled charge-coupled device (CCD) camera (Fuji). The band intensities in the blots were quantified using the Aida Image Analyzer.

The fluorometric microculture cytotoxicity assay (FMCA; ref. 18) was used to investigate the cytotoxic effects of PPP alone and in combination with other cytostatic drugs in vitro. Briefly, 96- or 384-well microtiter plates (Nunclon surface; NUNC Brand Products) were prepared containing drug solutions in duplicate at 10 times the desired final drug concentration. Cells were seeded into the drug-containing microtiter plates at a cell density of 0.1 × 10⁶ cells/mL and incubated for 72 hours at 37°C in a humidified 5% CO₂ atmosphere. FMCA was conducted using an automated Optimized Robot for Chemical Analysis (Orca; Beckman Coulter) programmed using the SAMI software package (Beckman Coulter). The plates were washed, fluorescein diacetate (10 μg/mL; Sigma-Aldrich) was added, and the fluorescence generated was measured (485/520 nm) after 40 to 50 minutes incubation. The resulting fluorescence is proportional to the number of intact cells in each well. A successful assay required a ratio of more than 10 between the signal in the control wells and the blank wells and a coefficient of variation of less than 30% in the control wells. Cell survival is reported in terms of the survival index (SI), which was defined as the fluorescence intensity from drug-treated cells compared with the intensity from untreated cells. All concentrations were tested in duplicate and all experiments were repeated at least twice. The cytotoxic IC₅₀ values for the tested drugs in the cell lines in vitro were determined from log concentration-effect (SI%) curves in GraphPad Prism (GraphPad software Inc.) using nonlinear regression analysis. For the in vitro results, comparisons of activity in different groups were conducted using two-sided t tests. Outliers were rejected using Dixon Q test with a 95% confidence interval. Correlations between the effects of different drugs on the cell line panels were analyzed on the basis of the survival index values for concentrations that yielded substantial variation in survival between the different cell lines (PPP, 10 μmol/L; PPT, 16 nmol/L; AEW541 10 μmol/L). In addition, combinations of fixed doses of PPP and standard agents were investigated; the corresponding results are presented in the Supplementary Table S1.

Molecular modeling was conducted using the Schrödinger software package with version 9.0 (r211) of the Maestro interface (19, 20). The structure of the complex found between tubulin and PPT (PDB 1SA1; ref. 21) was used as a basis for this study. The ligand and the A and B chains of tubulin were removed from the structure, and the missing residues in the protein were added using v. 2.1 (r211) of the Prime application. The structures of tubulin, PPT, and PPP were then energy minimized in the standard way. Docking was conducted using Glide v. 5.5 (r211) with the extra precision (XP) settings and standard parameters for ligand docking. Conformation analysis was conducted using MacroModel with the OPLS 2005 force field (method: PRCG; energy limit: 21.0 kJ/mol) and was followed by Multiple Minimization, also using the OPLS 2005 force field (method: TNCG; energy limit: 21.0 kJ/mol).

The drugs' effects on tubulin polymerization were investigated using the Tubulin Polymerization Assay Kit (porcine tubulin- and fluorescence-based), which use fluorescent reporter enhancement (Cytoskeleton Inc.). The standard assay protocol was used (22) to determine how treatment with 10 μmol/L PPP affected tubulin polymerization, and the system's fluorescence was measured using a FLUOstar OPTIMA instrument. The experiment was repeated twice and mean values are presented. The pharmaceutical compositions of docetaxel (Taxotere, a microtubule-stabilizing agent that was used as a positive control) and vincristine (a microtubule-destabilizing agent that was used as a second positive control) were diluted with PBS and used at final concentrations of 3 μmol/L.

Cells were grown on coverslips overnight and then treated with the appropriate drug for 24 hours as indicated. Cells were fixed in 2% formaldehyde and permeabilized with 0.2% Triton X-100. Coverslips were blocked with 10 mmol/L glycine at room temperature for 1 hour, after which they were incubated with primary mouse anti-α-tubulin antibody (Sigma-Aldrich) and then with a secondary polyclonal goat anti-mouse antibody labeled with fluorescein isothiocyanate (FITC; DAKO). The nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI). Coverslips were mounted on object slides using a Fluoromount-G instrument (Southern Biotechnology Associates). Microtubule staining was visualized using a Zeiss immunofluorescence microscope at ×40 magnification.

IGF-IR is overexpressed in several malignant tumors. Its phosphorylation, induced by ligand binding, enables the downstream activation of signaling pathways that are important for proliferation and survival (1). The expression and phosphorylation of IGF-IR in 9 ESCC cell lines was investigated by immunoprecipitating lysates from each cell line using an antibody against the β-subunit of IGF-IR, followed by Western blot analysis. IGF-IR was expressed in all 9 analyzed ESCC cell lines and exhibited a low level of spontaneous phosphorylation in most of them after cultivation in serum-supplemented growth medium (Fig. 1). However, this level of IGF-IR phosphorylation was much lower than that induced by the addition of recombinant IGF-I, as shown in Fig. 3A. Various detectable levels of pIGF-IR/IGF-IR expression were also observed in 8 different lung cancer cell lines (not shown).

Various approaches have been used to inhibit the function of IGF-IR in malignant cells. PPP has been reported to be a potent and selective inhibitor of the tyrosine phosphorylation of the IGF-IR. Despite their expression of IGF-IR, the 9 ESCC cell lines exhibited only modest sensitivity toward PPP in the FMCA. At the tested concentration range of 10 μmol/L, treatment with PPP only produce survival index values less than 50% in 4 of 9 cell lines (Fig. 2A). The cell lines varied as expected in their sensitivity to standard agents (data not shown; combination effects between PPP and standard agents were mostly antagonistic or additive; data shown in Supplementary Table S1). Furthermore, there was no significant (P > 0.05) correlation between PPP activity and IGF-IR expression (linear correlation coefficient; R² = 0.02) or phosphorylation (R² = 0.57; see Supplementary Fig. S2). However, the exclusion of 4 nonresponding cell lines resulted in a weak correlation for the remaining five (data not shown). Moreover, there was no correlation of the effects of PPP and those of the unrelated IGF-IR inhibitor AEW541 (R² = 0.18; Fig. 2C).

Similar results were found in a panel of 8 lung cancer cell lines and a phenotypic cell line panel designed to predict drug's mechanism of action (Supplementary Table S2 and Supplementary Figs. S3–S5). These results strongly suggest that the cytotoxic effects of PPP identified using the FMCA method are not due to IGF-IR inhibition.

Our finding that the activity of PPP did not correlate with either the level of IGF-IR expression or the activity of the IGF-IR kinase inhibitor AEW541, prompted us to analyze the effects of PPP treatment on the IGF-I–induced activation of this receptor. To this end, we treated cells with increasing concentrations of PPP (0.5, 5 and 10 μmol/L) for 24 or 48 hours, after which they were stimulated with IGF-I for 5 minutes. PPP had no effect on the phosphorylation or stability of the IGF-IR in any case (Fig. 3A). In contrast, AEW541 effectively inhibited ligand-induced IGF-IR phosphorylation in both of the tested cell lines (Fig. 3B).

These results suggest that PPP does not act as a direct IGF-IR kinase inhibitor in these 2 ESCC cell lines.

Interestingly, the activity of PPP at 10 μmol/L in the 9 ESCC cell lines correlated strongly with that of PPT at 16 nmol/L (R² = 0.90; Fig. 2B), and also to that of two other standard agents that target microtubule assembly: docetaxel (20 μmol/L; R² = 0.75) and vinorelbine (10 μmol/L; R² = 0.91; for details see Supplementary Fig. S6A and S6B). The activity of PPP either did not correlate or only correlated weakly with those of three other standard cytotoxic drugs: cisplatin, gemcitabine, and 5-FU (Supplementary Fig. S6C and S6D). All the tested drug concentrations were selected on the basis of empirical data to give the greatest possible variation among the cell lines (23).

Results consistent with those presented earlier for ESCC lines were also obtained in a phenotypic cell line panel, which was designed to reveal drug's mechanisms of action, and in a lung cancer cell line panel (Supplementary Data, sections 4 and 5 and Supplementary Figs. S3 and S5).

Although PPP had no detectable effect on IGF-IR levels or phosphorylation, its activity in the ESCC panel clearly correlated with that of tubulin inhibitors. Therefore, the relevance and contribution of tubulin inhibition to its mechanism of action was explored further. ESCC cells exposed to cytotoxic levels of PPP, PPT, docetaxel, and vincristine for 24 or 48 hours were stained to visualize their microtubule networks. The resulting micrographs (Fig. 4A) revealed blurry tubulin staining after PPP exposure, which is consistent with the hypothesis that it inhibits tubulin assembly. The results obtained using PPP were similar to those seen with PPT and vincristine; docetaxel yielded a more distinct pattern of microtubules. Treatment with a PPP concentration that is too low to have cytotoxic effects in the ESCC panel (i.e., 0.5 μmol/L) did not cause dissolution of the microtubule network, which suggests that the cytotoxic effects of PPP seen in these cells are associated with tubulin inhibition. Indeed, PPP could be docked in silico into the colchicines-binding site of tubulin in a model constructed using the PPT–tubulin complex (PDB 1SA1) as a template (Fig. 4B–E). As described previously, PPT binds to a hydrophobic pocket of tubulin that features two hydrophobic regions. The first is surrounded by the side chains of Met259, Ala316, and Lys352 and is occupied by the benzodioxole fragment of PPT, whereas the second is surrounded by the side chains of Leu242, Ala250, and Leu255 and is occupied by the trimethoxyphenyl moiety of PPT (24). In addition, the hydroxyl hydrogen of PPT can form a hydrogen bond with the carbonyl oxygen of Thr179 in the protein backbone (2.24 Å, –OH···O=C 146°; Fig. 4D). Docking of PPP into the colchicines-binding site (Fig. 4C) revealed that it can bind in a similar way. However, because of the difference between PPP and PPT with regards to the chirality of the C2 carbon, the fused lactone ring of PPP is oriented in a different direction than in PPT. It has previously been suggested that the orientation of this lactone ring is important for the microtubule-destabilizing activity of PPT (25), which is consistent with the relatively high concentrations of PPP that must be used to induce cytotoxicity and microtubule disruption. In addition, a conformational analysis of PPT conducted using molecular mechanisms (OPLS, MacroModel) suggested that PPT is fairly rigid and has only one low-energy conformation. In contrast, PPP is more flexible and can adopt several favorable conformations, mainly due to different arrangements of the fused lactone ring and the oxazole moiety (Fig. 4E). The docking study also revealed that unlike PPT, PPP cannot form a stabilizing hydrogen bond with the tubulin backbone because the distance between its hydroxyl group and Thr179 in the protein backbone is too great.

The possibility that PPP could directly affect tubulin polymerization was investigated with the commercially available Tubulin Polymerization Assay Kit (Cytoskeleton Inc.), which uses purified porcine tubulin. In this assay, PPP clearly inhibited the polymerization process at a concentration of 10 μmol/L (Fig. 4F). This suggests that while PPP can act as a tubulin inhibitor, it is not a potent one, as its effects are observed only at relatively high concentrations. This is consistent with our in silico docking studies, which showed that PPP can bind to tubulin but not as tightly as PPT.

The PPP used in all of the investigations discussed earlier (including the microtubule assay) was pure when the experiments were started. However, PPP may undergo nonenzymatic conversion to PPT, which could potentially have influenced the results obtained. The conversion of PPP into PPT was investigated by monitoring initially pure PPP solutions of various concentrations (0.5, 5, and 10 μmol/L) in cell growth medium at 37°C. After 1 hour, the level of PPT in the solutions corresponded to 0, 0.4, and 0.3% of the initially added PPP; after 24 hours, the corresponding levels were 1.8%, 1.7%, and 1.5%; and after 48 hours they were 2.1%, 2.5%, and 2.6%, respectively. The 48-hour values are very close to the expected equilibrium concentration (15). The PPP-to-PPT conversion was measured by Dr. Magnus Axelson, and used with permission.

We have previously found that the dissolution of microtubule networks using established tubulin-binding agents can result in dephosphorylation and degradation of the EGFR but not IGF-IR (26). Therefore, we investigated whether PPP had similar effects on EGFR, which was expressed in all of the ESCC cell lines in the panel (not shown). PPP did indeed inhibit the phosphorylation of EGFR and reduced its stability in a dose- and time-dependent manner (Fig. 5A). It therefore seemed reasonable to investigate whether downstream signaling pathways such as Ras/Erk1/2 MAPK and PI3K/Akt were also affected by PPP. In keeping with the observed inhibition of EGFR, PPP also reduced the phosphorylation of the signaling proteins Erk1/2 and Akt in the Kyse70 cell line, whereas these effects were minimal in Kyse140 cells (Fig. 5B). This may relate to the observation that in Kyse140 cells exhibit relatively high levels of spontaneous IGF-IR phosphorylation in serum-supplemented growth medium (Fig. 1); this may be sufficient to compensate for the PPP-induced reduction in EGFR phosphorylation and to promote Akt and Erk1/2 phosphorylation even in the presence of PPP. In fact, a degree of functional cross-talk (including a physical interaction), between IGF-IR and EGFR has been described (reviewed in ref. 27). It is therefore not appropriate to use reductions in the phosphorylation of Erk1/2 and Akt (both of which are activated by most receptor tyrosine kinases) as measures of IGF-IR inhibition phosphorylation by PPP as our data show that PPP has no effect on IGF-IR despite inhibiting Erk1/2 and Akt phosphorylation in some cases. The reduced phosphorylation of Erk1/2 and Akt seen in Kyse70 cells may be partly due to the downregulation of EGFR. As the molecular target(s) of PPP have not been unambiguously identified, it remains possible that other receptors or proteins may also be affected. Moreover, we have observed that microtubule disruption in ESCC cell lines by PPT or vincristine also inhibits EGF-mediated phosphorylation of Akt and Erk1/2 (26).

Growth factors and their receptors play pivotal roles in the establishment and development of cancer. The IGF-IR is an attractive target for cancer therapy because it is often overexpressed and activated in various malignant tumors (1), including esophageal cancer (28). Patients with ESCC have significantly higher serum levels of IGF-I than healthy donors, which may be an important predictor of risk for esophageal cancer (29). Strong IGF-I signaling correlates significantly with depth of invasion, pathologic stage, and poor prognosis in esophageal cancers (28, 30). Blockage of IGF signaling by dominant-negative IGF-IR suppresses proliferation and motility of ESCC cell lines (29). Therefore, the level of IGF can be used as a tumor marker in esophageal carcinoma, and drugs that inhibit the functioning of IGF-I could potentially be used to treat esophageal cancer (29). Presumably, candidates for such treatment would be identified on the basis of clinical immunohistochemical detection of the IGF-IR/IGF-II complex (30). In addition, laboratory studies have suggested that IGF-I stimulation plays an important role in repair processes in the esophagus (31). The IGF-I peptide antagonist JB1 is active against the esophageal carcinoma cell line CE81T/VGH in vitro and in vivo, and also seems to have favorable effects when used in combination with chemotherapy (32).

Various approaches for interfering with the IGF-IR pathway have been developed, including methods based on neutralizing antibodies, a dominant-negative receptor, IGF-IR antisense/siRNA and small-molecule tyrosine kinase inhibitors with varying degrees of selectivity. However, as IGF-IR displays 84% similarity with the insulin receptor (whose functioning is essential for most mammalian cells), the development of anticancer drug candidates targeting this receptor has been difficult. Crystallographic studies conducted to identify structural differences between IGF-IR and the insulin receptor have been used as the basis for attempts to develop selective inhibitors. Several inhibitors of the IGF-IR pathway have been evaluated in clinical trials, most of which have yielded discouraging results and resulted in the abandonment of the corresponding drug candidates. Some cyclolignans have been identified as potent and selective inhibitors of the tyrosine phosphorylation of IGF-IR. One such compound is PPP, which is almost nontoxic (oral LD₅₀ >500 mg/kg in rodents; ref. 8). In addition to inhibiting the tyrosine kinase activity of IGF-IR, PPP has also been claimed to inhibit IGF-IR by downregulating the protein (33). Although most IGF-IR inhibitors have failed because of a lack of activity or adverse effects, PPP seems to be different, in that it exhibits clinical activity, with no metabolic toxicity, and only moderate general toxicity. In our view, this suggests that PPP may have a primary target (or targets) other than IGF-IR.

Our molecular modeling studies indicated that PPP can adopt several stable conformations (Fig. 4E), and it is possible that these have distinct biologic effects.

Although a dual mechanism of IGF-IR antagonism has been proposed for PPP, we could not confirm either of these effects in the ESCC cell lines or in other cell lines (see the Supplementary Data for details). PPP exhibited cytotoxic/cytostatic activity at micromolar concentrations in vitro in the tested cell lines, and there was no overall correlation between PPP sensitivity and the expression or phosphorylation of IGF-IR (Supplementary Fig. S2). The exclusion of 4 nonresponding cell lines resulted in a weak correlation for the remaining five. However, the significance of these results is questionable because some of the excluded cell lines exhibited clear spontaneous IGF-IR phosphorylation. Furthermore, in experiments with recombinant IGF-I, we were unable to observe any effect of PPP on the phosphorylation or stability of IGF-IR in 2 ESCC cell lines (Fig. 3A). In contrast, AEW541 efficiently inhibited IGF-IR phosphorylation under similar conditions (Fig. 3B). A recent study using four different colon cancer cell lines, showed that all four cell lines were sensitive to PPP treatment, but PPP only inhibited the phosphorylation and stability of IGF-IR in HT-29 (34). Because all cells were sensitive to PPP, this indicates that the cytotoxic effects of PPP may be independent of its postulated IGF-IR antagonism. Another recent study showed that PPP does not inhibit the phosphorylation of IGF-IR in adrenocortical adenocarcinoma cells, but that PPP treatment nevertheless causes accumulation in G₂–M phase and apoptosis (35). Interestingly, the authors concluded that these effects are not mediated by antitubulin effects because PPP used was “ultra-pure,” and previous investigations had shown that PPP only binds very weakly to tubulin. The PPP used in the investigations described herein was obtained from the same source as that used in the previous investigations.

We have previously shown that the disruption of the microtubule system by PPT or vincristine causes the dephosphorylation and degradation of the EGFR (26). We obtained similar results in this work using PPP (Fig. 5A). This inhibition correlated with reductions in the activity of the downstream signaling proteins Erk1/2 and Akt in Kyse70 cells but not in Kyse140 cells (Fig. 5B), which exhibit relatively high levels of spontaneous IGF-IR phosphorylation (Fig. 1); this spontaneous phosphorylation was unaffected by exposure to PPP. The existence of this effect on EGFR means that care should be taken when using secondary reporters such as reductions in the activity of Erk1/2 or Akt to monitor IGF-IR inhibition in vitro. The cross-talk between IGF-I and EGFR signaling is another factor that could confound such evaluations, and it has been shown that the activation of Erk1/2 by IGF-I requires the EGFR (36, 37). These 2 factors mean that loss of EGFR function would have results that could be misinterpreted as evidence for the inhibition of IGF-I–induced signaling to Erk1/2. This significantly complicates the interpretation of studies in which the inhibition of Erk1/2 was used as a measure of PPP-mediated IGF-IR inhibition.

The cytotoxic activity of PPP in various cell line panels measured by the FMCA method correlated strongly with that of various tubulin inhibitors but not with that of the IGF-IR inhibitor NVP-AEW541 (Fig. 2B and C). Using computer modeling, we showed that PPP is theoretically able to fit into the colchicine-binding pocket of tubulin (Fig. 4C). This is partly due to structural flexibility of PPP allowing it to adopt conformations that are compatible with tubulin binding (Fig. 4E). In keeping with observation, PPP inhibited microtubule assembly in a cell-free assay (Fig. 4F) and disrupted the tubulin staining patterns at relevant (i.e., cytotoxic and clinically achievable in patients) concentrations in cultured cells (Fig. 4A). However, PPP did not fit as tightly as PPT into the binding pocket of tubulin; together with the relatively high PPP concentration needed to interfere with tubulin polymerization, this suggests that PPP is a weak tubulin inhibitor.

An equilibrium between PPP and PPT was established rather rapidly when PPP solutions were left to stand in cell growth media, and concentrations of PPT that would be expected to have biologic effects were detected after only 1 hour of incubation. It is therefore possible that the results observed in our experiments were partly due to the presence of PPT in small quantities. Importantly, this would also be the case for all previous experimental investigations using PPP (as an IGF-IR inhibitor), and our results strongly suggest that the equilibrium between PPP and PPT should be considered in future studies on these compounds. PPT is a known and potent inhibitor of microtubule assembly, and even small amounts of PPT in experimental samples could potentially have dramatic effects on the results obtained, especially in assays based on whole cells. The impact of the PPP-to-PPT conversion on the results will depend on the concentration of PPP that is needed to observe cytotoxic effects. It has previously been suggested that PPP may have no effect whatsoever on tubulin (38), and data were presented suggesting that highly purified PPP does not bind to tubulin or binds very weakly to at concentrations up to 50 μmol/L. This is in sharp contrast with our results, which confirmed that PPP has direct effects on the microtubule assembly in a 45-minute cell-free assay (Fig. 4F). The PPP used in this assay was pure at the beginning of the experiment (coming from freshly thawed DMSO stock), but it is certainly possible that minute amounts of PPT could be formed under the experimental conditions, as this seems to happen spontaneously under standard laboratory conditions. Furthermore, it cannot be excluded that PPT is actively being removed from solution due to its strong binding to tubulin [the binding constant for the formation of tubulin–PPT complex is 1.8 × 10⁶ (μmol/L)⁻¹ (13)], and that this removal drive the further conversion of PPP into PPT to maintain the equilibrium between the 2 isomers in solution. It is possible that this conversion also happens in vivo, but may not be detected as the levels of free PPT would be negligible: tubulin is very abundant in all cells and would rapidly sequester the PPT as it is formed, thus removing it from serum. Moreover, the interaction between tubulin and PPP may enhance the rate of PPP to PPT conversion due to the tighter binding of PPT to tubulin. In such a scenario, one would expect to find negligible or very low levels of PPT in serum as it is formed in the binding pocket of tubulin. Indeed, no circulating PPT was detected in a recent investigation (with a limit of quantification of 0.01 μmol/L) that examined plasma from patients treated with oral PPP in a phase I study (39). Plasma levels PPP in excess of 1 μmol/L were detected following an oral dosing with 520 mg PPP, which justifies the selection of the concentrations used in our experiments. However, the cellular concentrations and PPP or PPT (i.e., their concentrations in the tubulin-rich compartments of the body) were not analyzed.

Interestingly, genomic arrays of cells selected for PPP resistance show commonly altered expression of tubulin or microtubule-associated genes including overexpression of MARK1 and TUBA4A and underexpression of EML5 and MICAL2 (40), although other changes in gene expression was also found.

The viability of the ESCC cells used in this investigation was affected by PPP at micromolar concentrations, which are high but may be achieved in vivo after oral dosing of PPP (39). On the basis of the results presented herein, we propose that under such conditions, PPP's primary mechanism of action does not involve IGF-IR inhibition and that it instead causes microtubule destabilization. This is associated with EGFR downregulation and decreased activity of downstream signaling pathways. Similar results were obtained in a lung cancer cell line panel (n = 8), and a phenotypic cell line panel (n = 10; Supplementary Table S2), which have different resistance mechanisms to those of the ESCC lines. PPP may act directly on microtubules, as it inhibited tubulin polymerization in a cell-free assay (Fig. 4F) and molecular modeling investigations did not exclude the possibility that PPP might bind directly to tubulin (Fig. 4C and E). However, it is likely that at least some of PPP's apparent effects on microtubules are indirect and stem from its conversion into PPT, which was observed to occur within an hour under standard laboratory conditions. Interestingly, although several publications have reported that PPP affects the phosphorylation and/or stability of IGF-IR, we could not reproduce this in our ESCC cells despite using relatively high PPP concentrations (Fig. 3A). However, we did observe an analogous effect on EGFR (Fig. 5A). The mechanism responsible for this effect is not readily apparent, and it is possible that the microtubule-destabilazing agents may affect different receptors in different cell types. In addition, the effect on EGFR was only observed after prolonged incubation with PPP (24 hours or more), suggesting that the tubulin-mediated effect targets receptor turnover. Consequently, treatment with PPP also downregulates tyrosine kinase receptor signaling via proteins such as pAkt (Fig. 5B). This could potentially mislead investigators into believing that PPP has anti-IGF-IR effects in various cell lines.

In summary, we suggest that the postulated IGF-IR inhibitor PPP primarily acts by disrupting the microtubule cytoskeleton, either by directly binding to tubulin or via conversion into its isomer, PPT, and that this indirectly results in EGFR downregulation. We strongly suggest that the alternative effect on micotubules should be accounted for in all further studies with that use PPP. Furthermore, our results suggest that it is not appropriate to use levels of phosphorylated Akt or Erk1/2 as a readout for anti-IGF-IR activity in vitro as we clearly show that EGFR is inhibited by PPP. Our results are in sharp contrast to previous reports, and we have put forward a model that explains the inconsistent results presented in some previous publications. Although the results presented in this article strongly suggest that PPP does not have a IGF-IR-mediated mechanism of action, we would like to emphasize that this in no way means that it is necessarily without therapeutic effects, potentially beneficial to patients. On the contrary, the continous moderate inhibition of tubulin assembly (either by PPP itself or via its gradual conversion to PPT, which would simulate a slow-release preparation) is of high interest and could be a promising strategy for treating solid malignancies. For example, in a study using metronomic oral vinorelbine dosing in 62 patients with advanced refractory cancer, 13% of the patients exhibited objective antitumor responses, and 32% of patients experienced disease stability for at least 6 months. Three responders received nonstop treatment for more than 3 years without overt toxicity (including absence of neurotoxicity). The authors suggest an antiangiogenetic mechanism (41), which would be consistent with the disease stabilization and the central necrotic areas observed in patients with non–small cell lung carcinoma (NSCLC) treated with PPP (10). Because PPP has shown activity in clinical trials, investigations aimed at identifying its cellular target(s) are highly warranted.

J. Gullbo is co-founder with ownership interest (including patents) in Oncopeptides AB, which is a small research and development company that for a long period of time (until July 2012) shared the same principal investor/owner (Karolinska Development AB) as Axelar AB (now developing PPP for clinical use). J. Gullbo has no ownership interest in Karolinska Development AB or Axelar AB. No potential conflicts of interest were disclosed by the other authors.

Conception and design: X. Wu, M. Wickström, J. Lennartsson, J. Gullbo

Development of methodology: J. Gullbo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Sooman, M. Wickström, J. Lennartsson, J. Gullbo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Wu, L. Sooman, M. Wickström, M. Fryknäs, C. Dyrager, J. Gullbo

Writing, review, and/or revision of the manuscript: X. Wu, L. Sooman, M. Wickström, M. Fryknäs, C. Dyrager, J. Lennartsson, J. Gullbo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Fryknäs

Study supervision: J. Lennartsson, J. Gullbo

The authors thank Prof. Magnus Axelson for quantifying the conversion of PPP to PPT.

This work was financially supported by the Ludwig Institute for Cancer Research (to J. Lennartsson), the Swedish Cancer Society (J. Lennartsson, CAN2012/365; J. Gullbo, co-investigator CAN 2010/517), and the Swedish Research Council (J. Lennartsson, K2011-67X-21859-01-6), LIONS cancerforskningsfond i Uppsala (J. Gullbo), Stiftelsen Onkologiska Klinikens i Uppsala forskningsfond (J. Gullbo).

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.