Medulloblastoma sensitivity to 17-allylamino-17-demethoxygeldanamycin requires MEK/ERKM.

ERBB2 increases the sensitivity of breast cancer cells to the HSP90 inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG). This has been attributed to the disruption of ERBB3/ERBB2 heterodimers that maintain a crucial cell survival signal via phosphatidylinositol 3-kinase/AKT. ERBB2 confers a poor clinical outcome in medulloblastoma, the most common malignant pediatric brain tumor. Here, we show that medulloblastoma cell sensitivity to 17-AAG is directly related to ERBB2 expression level. Furthermore, overexpression of exogenous ERBB2 in these cells induces spontaneous homodimerization, further enhancing cell sensitivity to 17-AAG. In contrast to breast cancer cells, this increased sensitivity to 17-AAG does not result from cell dependence on AKT1 activity. Rather, we show that 17-AAG generates a dose- and time-dependent increase in MEK/ERK signaling that is required for the drug to inhibit the proliferation of medulloblastoma cells and that ERBB2 sensitizes medulloblastoma cells to 17-AAG by up-regulating basal MEK/ERK signaling. We further show that down-regulation of MEK1 activity markedly reduces the sensitivity of medulloblastoma, breast, and ovarian cancer cells to 17-AAG, whereas expression of a constitutively active MEK1 potentiates the activity of 17-AAG against these cells. Therefore, intact MEK/ERK signaling may be required for optimal 17AAG activity against a variety of tumor cell types. These data identify a new mechanism by which 17-AAG inhibits the proliferation of cancer cells. Defining the precise mode of action of these agents within specific tumor cell types will be crucial if this class of drugs is to be efficiently developed in the clinic.

ERBB2 increases the sensitivity of breast cancer cells to the HSP90 inhibitor 17-allylamino-17-demethoxygeldanamycin (17-AAG). This has been attributed to the disruption of ERBB3/ERBB2 heterodimers that maintain a crucial cell survival signal via phosphatidylinositol 3-kinase/AKT. ERBB2 confers a poor clinical outcome in medulloblastoma, the most common malignant pediatric brain tumor. Here, we show that medulloblastoma cell sensitivity to 17-AAG is directly related to ERBB2 expression level. Furthermore, overexpression of exogenous ERBB2 in these cells induces spontaneous homodimerization, further enhancing cell sensitivity to 17-AAG. In contrast to breast cancer cells, this increased sensitivity to 17-AAG does not result from cell dependence on AKT1 activity. Rather, we show that 17-AAG generates a dose-and time-dependent increase in MEK/ ERK signaling that is required for the drug to inhibit the proliferation of medulloblastoma cells and that ERBB2 sensitizes medulloblastoma cells to 17-AAG by up-regulating basal MEK/ERK signaling. We further show that down-regulation of MEK1 activity markedly reduces the sensitivity of medulloblastoma, breast, and ovarian cancer cells to 17-AAG, whereas expression of a constitutively active MEK1 potentiates the activity of 17-AAG against these cells. Therefore, intact MEK/ERK signaling may be required for optimal 17AAG activity against a variety of tumor cell types. These data identify a new mechanism by which 17-AAG inhibits the proliferation of cancer cells. Defining the precise mode of action of these agents within specific tumor cell types will be crucial if this class of drugs is to be efficiently developed in the clinic.
HSP90 plays a key role in the conformational maturation of important cell signaling proteins, including ERBB2, AKT1, and RAF1 (9,10). Inhibition of HSP90 targets these proteins for proteasomal degradation (11)(12)(13)(14)(15), resulting in growth inhibition and apoptosis of tumor cells (16 -21). Although 17-AAG down-regulates a large number of cellular proteins (9), the anti-tumor properties of this drug may be attributed to effects on specific signal pathways, most notably those driven by ERBB2.
ERBB2 is one of the most sensitive HSP90-dependent client proteins (22). Overexpression of this receptor by breast and ovarian cancer cells renders them acutely sensitive to growth inhibition by 17-AAG (18,23). This enhanced sensitivity has been attributed to the disruption of ERBB2/ERBB3 heterodimers, which block activation of phosphatidylinositol 3-kinase/AKT, thereby inhibiting a crucial cell survival signal (18,19). Although ERBB2-dependent phosphatidylinositol 3-kinase/AKT signaling appears to be an important target of 17-AAG, this drug also inhibits components of the RAS/mitogenactivated protein kinase pathway (20,22). Indeed, depletion of N-RAS, Ki-RAS, and RAF1 from colon carcinoma cells inhibits ERK1/2 activation, resulting in growth inhibition and apoptosis (16).
Cells, DNA Constructs, and Transfections-The MHH-MED-1 and MEB-MED-8A cells lines (33) were provided by Dr. Torsten Pietsch (University of Bonn, Bonn, Germany). The Daoy medulloblastoma, MCF-7 and SKBR3 breast cancer, and SKOV3 ovarian cancer cell lines were obtained from the American Type Culture Collection. Cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (BioWhittaker, Walkersville, MA). ERBB2-overexpressing Daoy cell clones (designated Daoy.1 and Daoy.2) were generated by stable transfection with human ERBB2 cDNA inserted into pcDNA3.1 as described previously (26). Control cells transfected with empty vector alone were designated Daoy.V. We generated the AKT1 and green fluorescent protein retroviral vectors as described previously (26). The pUSEamp/MEK1* expression vector that contains an HA-tagged constitutively activated (S218D/S222D) MEK1 was obtained from Upstate Biotechnology, Lake Placid, NY. Daoy cells were infected with retroviral vectors using standard procedures as described previously (26). Tumor cells were transiently transfected with pUSEamp/MEK1* or empty vector using the LipofectAMINE reagent as detailed by the manufacturer (Invitrogen). Expression of exogenous AKT1 and activated MEK1 was confirmed by HA-specific Western blotting. Green fluorescent protein expression was detected by fluorescence microscopy (34). MEK1 oligonucleotides were added to serum-free Dulbecco's modified Eagle's medium containing 16 l of LipofectAMINE (Invitrogen) per 1 g of DNA to give a final oligonucleotide concentration of 10 M. Cells were incubated with oligonucleotides for 8 h before the addition of 17-AAG and processed as described.
Dimer Assay-Cultured cells were treated on ice with either phosphate-buffered saline or epidermal growth factor 50 ng/ml for 2 h. 1 mM bis-(sulfosuccinimidyl)-suberate (BS 3 ) was then added, and cells were incubated for a further 45 min. Total protein lysates were then generated and analyzed by Western blotting as described.
Drug Treatments-For growth inhibition studies, 3 ϫ 10 3 cells were seeded into each well of a 96-well plate in 10% fetal bovine serum Dulbecco's modified Eagle's medium and incubated for 24 h. 17-AAG or vehicle only (0.1% Me 2 SO) was then added, and plates were incubated for the indicated time period, up to a maximum of 96 h. For drug exposures less than 96 h, drug-containing medium was replaced with drug-free medium, and incubation was continued for a total of 96 h. After incubation, cell growth was determined using an XTT-based assay (Roche Applied Science). IC 50 was defined as the drug concentration that inhibited cell proliferation by 50% for a 96-h exposure. Morphological detection of apoptosis was performed by epifluorescence microscopy of Hoechst 33258-stained cells. One observer who was blinded to all preceding cell treatments performed the apoptotic counts using nuclei from triplicate experiments.
For drug combination studies, cells were pretreated for 8 h with MEK1 oligonucleotide (or scrambled control) as described above or with the MEK1/2 inhibitors PD98059 or U0126 (or 0.1% Me 2 SO control) for 2 h. 17-AAG (or 0.1% Me 2 SO) was then added, and growth assays were conducted as described. To determine the impact of drug treatment on protein expression, cells were treated as described for growth assays, and the total protein lysates (29) were analyzed by Western blotting.

ERBB2 Expression Level Correlates with Medulloblastoma
Cell Sensitivity to 17-AAG-Breast cancer cells that overexpress ERBB2 are dependent on ERBB3/ERBB2 heterodimer signaling. This renders them acutely sensitive to treatment with 17-AAG (18,19,23). We investigated whether ERBB2 similarly dictates 17-AAG activity against medulloblastoma cells. To do this, we determined the 17-AAG inhibitory concen-tration (IC) 50 values of 3 medulloblastoma cell lines after 96 h of continuous drug exposure and compared these to cell ERBB2 expression levels. Daoy cells that expressed low levels of ERBB2 ( Fig. 1A) were relatively resistant to 17-AAG (Fig. 1B); indeed these cells displayed an IC 50 value that was considerably higher than that previously reported for 40 human cancer cell lines treated under similar conditions (18,20,21). In contrast, MHH-MED-1 and MEB-MED-8A cells, which express moderate to high levels of ERBB2, were relatively sensitive to 17-AAG (Fig. 1, A and B). Therefore, ERBB2 expression level correlates with medulloblastoma cell sensitivity to 17-AAG. However, in contrast to breast cancer cells, this enhanced sensitivity cannot be mediated via ERBB3/ERBB2 heterodimers, since ERBB3 was not detected in any of the medulloblastoma cell lines (Fig. 1A). Furthermore, whereas Daoy and MHH-MED-1 cells express low level ERBB1 and high level ERBB4, respectively, ERBB2 is expressed in isolation in MEB-MED-8A cells (Fig. 1A).

ERBB2 Homodimerization Activates AKT1 and ERK1/2 Signaling in Medulloblastoma Cells and Is Associated with
Increased Sensitivity to 17-AAG-ERBB2 homodimers spontaneously arise when the receptor is expressed at high levels in cells (3,35). Therefore, we reasoned that ERBB2 homodimer signaling might increase the sensitivity of medulloblastoma cells to 17-AAG. To investigate this we induced spontaneous ERBB2 homodimerization in Daoy cells by overexpressing exogenous ERBB2 ( Fig. 2A, second lane). This resulted in activation of AKT1 and ERK1/2, but not p38, JNK, STAT3, or STAT5 ( Fig.  2B and data not shown). Activation of phosphatidylinositol 3-kinase/AKT signaling by spontaneous ERBB2 homodimers in the absence of growth factors has been described in a number of cell types including NIH3T3 fibroblasts (36). However, the precise mechanism by which ERBB2 homodimers activate phosphatidylinositol 3-kinase/AKT remains to be determined. Our data indicate that overexpression of ERBB2 did not increase cell signaling via ERBB1/ERBB2 heterodimerization, since ERBB1 was not phosphorylated (Fig. 2B) and ERBB1/ ERBB2 heterodimers could only be detected after treatment with exogenous epidermal growth factor (EGF) ( Fig. 2A, third  lane).
We next investigated the impact of ERBB2 homodimer signaling on cell sensitivity to 17-AAG. After 96 h of continuous exposure to 17-AAG, two independent ERBB2-transfected Daoy clones were significantly more sensitive to growth inhibition by 17-AAG than control cells (Fig. 2C). These data confirm that ERBB2 sensitizes medulloblastoma cells to 17-AAG and support the hypothesis that this is mediated by ERBB2 homodimer signaling via AKT1 and/or ERK1/2. Importantly, the expression of key chaperone proteins was not affected by ERBB2 transfection and, thus, did not account for differences in 17-AAG sensitivity (Fig. 2B).
ERBB2 Homodimers Sensitize Medulloblastoma Cells to 17-AAG by Up-regulating ERK1/2 Signaling-We reasoned that ERBB2 homodimer signaling could increase medulloblastoma cell sensitivity to 17-AAG through two alternative mechanisms. Up-regulation of AKT1 or ERK1/2 activity might provide a crucial survival signal for medulloblastoma cells. Thus, analogous to the disruption of ERBB3/ERBB2 heterodimers by 2) were treated on ice with either phosphate-buffered saline (Ϫ) or 50 ng/ml epidermal growth factor (EGF) (ϩ) for 2 h. 1 mM bis-(sulfosuccinimidyl)-suberate (BS 3 ) was then added, and cells were incubated for a further 45 min. Total protein lysates were generated and analyzed by Western blotting using an antibody specific for phospho-Tyr-1248 of ERBB2. Note that spontaneous ERBB2 homodimers (D 1 ) were only detected in ERBB2-overexpressing Daoy.2 cells (compare first and second lanes). Epidermal growth factor induced the formation of a distinct ERBB1/ERBB2 heterodimer complex (D 2 ). M, monomers. B, total protein lysates were prepared from exponentially growing cultures of Daoy.V and two independent ERBB2-transfected Daoy clones (Daoy.1 and Daoy.2). Fifty micrograms of each lysate were analyzed by phospho (p)specific Western blotting for activation of ERBB1, ERBB2, ERK1/2, and AKT1 signaling. Levels of phosphorylated proteins were compared with total amounts of the respective protein. Expression levels of RAF1, TP53, and the key chaperone components HSP90, HSP70, and GRP94 were also determined. C, Daoy.V, Daoy.1, and Daoy.2 cells were seeded in 96-well plates and exposed to the indicated concentrations of 17-AAG (or vehicle alone) for 96 h. The percentage growth inhibition of 17-AAG relative to vehicle-treated cells was then determined using an XTT-based assay.
17-AAG in breast cancer cells (18,19), down-regulation of ERBB2 homodimers could inhibit a pro-survival signal. Conversely, AKT1 or ERK1/2 could play an active role in 17-AAGmediated growth inhibition. In this situation, up-regulation of one or both of these pathways by ERBB2 homodimers could sensitize medulloblastoma cells to the drug.
To investigate these alternative hypotheses we exposed Daoy cells to a range of 17-AAG concentrations and studied the impact of this treatment on cell protein expression levels. Continuous exposure of control cells to 17-AAG (48 h) abolished ERBB2 expression and generated a dose-dependent depletion of phosphorylated AKT1 (Fig. 3A). 17-AAG also caused a decrease in total AKT1 protein expression in these cells, albeit to a lesser extent than the phosphorylated form. In contrast, 17-AAG was relatively ineffective at inhibiting AKT1 in ERBB2-transfected cells (Fig. 3A). Indeed, phosphorylation of AKT1 was unaffected by 250 nM 17-AAG, which inhibited the growth of these cells by Ն75% (Fig. 2C). This signaling was maintained by active ERBB2 homodimers that could still be detected in the presence of 250 nM 17-AAG ( Fig. 3A and data not shown). Therefore, we concluded that inhibition of ERBB2/ AKT1 signaling is unlikely to account for the increased sensitivity of ERBB2-overexpressing Daoy cells to 17-AAG.
Unexpectedly, 17-AAG generated a dose-dependent increase in ERK1/2 phosphorylation in Daoy cells (Fig. 3A). This was not affected by ERBB2 expression status and occurred despite concurrent degradation of RAF1. As previously reported (37-39), 17-AAG did not affect expression of either total MEK1 or ERK1. To investigate whether up-regulation of ERK1/2 activity by 17-AAG is a general feature of medulloblastoma cells, we analyzed the impact of 17-AAG treatment on ERBB2, RAF-1, and pERK1/2 expression levels in MHH-MED-1 and MEB-MED-8A cells. 17-AAG generated a dose-dependent increase in ERK1/2 phosphorylation independent of ERBB2 and RAF1 in these cells (Fig. 3B). In other cell line systems, ansamycins have been reported to decrease MEK/ERK signaling by depleting upstream components and activators of the mitogen-activated protein kinase pathway (37,38,40). Conversely, our data indicate that 17-AAG can positively regulate MEK/ERK signaling independent of RAF-1 in medulloblastoma cells, presumably by inducing alternative activators of this signal cassette. Furthermore, these data support the hypothesis that induction of ERK1/2 plays a positive role in the anti-proliferative activity of 17-AAG. Therefore, although the induction of ERK1/2 by 17-AAG appears independent of ERBB2 and RAF1 (Fig. 3, A  and B), up-regulation of basal ERK1/2 signaling by ERBB2 homodimers (Fig. 2B) may nonetheless sensitize medulloblastoma cells to 17-AAG. Indeed, the high ERBB2-expressing, 17-AAG-sensitive Daoy.1, Daoy.2, MHH-MED-1, and MEB-MED-8A cells all demonstrate detectable basal pERK1/2, which is absent in low ERBB2-expressing, 17-AAG-resistant Daoy.V cells (Fig. 3, A and B).
To investigate further the role of ERK1/2 signaling in 17-AAG activity, we used the isogenic empty vector and ERBB2transfected Daoy cells to study the temporal relationship between drug-mediated activation of ERK1/2 and growth inhibition. As expected, phosphorylation of ERK1/2 was undetectable in untreated Daoy.V control cells (Fig. 4A). However, exposure of these cells to 250 nM 17-AAG for 16 h activated ERK1/2 (Fig. 4A). This coincided exactly with the duration of drug exposure required to inhibit the growth of these cells (Fig.  4B). In contrast, ERK1/2 activation was readily detected in vehicle and 17-AAG-treated ERBB2-transfected cells (Fig. 4A), and only 8 h of drug exposure were required to significantly inhibit cell growth (Fig. 4B). Similar results were observed in the other ERBB2-transfected clone (Daoy.1). In contrast, AKT1 activity in both control and ERBB2-transfected cells was largely unaffected by drug exposures of less than 24 h (Fig. 4A).
These data support the hypothesis that inhibition of medulloblastoma cell growth by 17-AAG involves the induction of ERK1/2 signaling and that up-regulation of basal ERK1/2 activity by ERBB2 homodimers sensitizes these cells to 17-AAG. Furthermore, they provide additional evidence that inhibition of AKT1 is not required for 17-AAG-mediated growth inhibition of medulloblastoma cells.
Inhibition of MEK1 Antagonizes 17-AAG-To determine the dependence of 17-AAG activity on MEK/ERK signaling, we studied the effect of blocking this pathway on drug-mediated growth inhibition. First, we confirmed that the MEK1-specific inhibitors U0126 and PD98059 alone do not significantly impact the growth of ERBB2 transfected or control Daoy cells (data not shown). However, pretreatment of Daoy cells with either inhibitor markedly reduced cell sensitivity to 17-AAG independent of ERBB2 expression status (Fig. 5A). To confirm that this effect was mediated by MEK1 blockade, we selectively depleted MEK1 expression from Daoy cells using an antisense oligonucleotide (M1AS) (Fig. 5B). In keeping with the results obtained using the pharmacological MEK1 inhibitors, M1AS significantly reduced the sensitivity of both ERBB2-transfected and control Daoy cells to 17-AAG (Fig. 5C). Although M1AS curtailed the activity of 17-AAG, it was less effective than either U0126 or PD98059. This likely resulted from the inability of M1AS to totally abolish ERK1/2 signaling (Fig. 5B).
To investigate whether MEK/ERK signaling mediates 17-AAG activity in other tumor types we expanded these studies to include the high ERBB2-expressing SKBR3 breast cancer and SKOV3 ovarian cancer cell lines. Both U0126 and PD98059 significantly impaired the activity of 17-AAG against SKBR3 breast cancer cells, although this effect was less pronounced than that observed in Daoy cells (Fig. 5A). PD98059, which was the most effective compound at blocking 17-AAG activity against Daoy and SKBR3 cells, only marginally reduced 17-AAG activity against SKOV3 ovarian cancer cells (Fig. 5A); however, MEK/ERK signaling in these cells was relatively resistant to treatment with MEK1 inhibitors (data not shown). These data confirm that MEK/ERK signaling plays a direct role in 17-AAG-mediated growth inhibition of medulloblastoma cells. Furthermore, they provide preliminary data that this pathway may also determine the sensitivity of other forms of cancer to 17-AAG.
Up-regulation of MEK/ERK Signaling Sensitizes Cancer Cells to 17-AAG-Having shown that inhibition of MEK1 antagonizes 17-AAG-mediated growth inhibition of tumor cells, we reasoned that up-regulation of MEK1 might potentiate the anti-proliferative activity of 17-AAG. To test this we compared the growth inhibitory properties of 17-AAG against tumor cells that we had transiently transfected with either constitutively active MEK1 or empty vector (Fig. 6). To assess the importance of cell background for MEK1 modulation of 17-AAG activity, we performed these experiments in Daoy and MEB-MED-8A medulloblastoma cells, MCF-7 and SKBR3 breast cancer cells, were then prepared and analyzed by Western blotting for expression of the indicated phosphorylated (p) proteins. Actin was employed as a loading and transfer control. B, cells were seeded into 96-well plates and exposed to 250 nM 17-AAG or vehicle-only for the indicated times. Medium was then replaced with drug-free medium, and incubation was continued for a total of 96 h. The percentage growth inhibition of 17-AAG relative to vehicle-treated cells was then determined using an XTT-based assay. and the SKOV3 ovarian cancer cell line. Despite numerous attempts we were unable to transfect MHH-MED-1 cells with the pUSEam/MEK1* construct. No significant difference in proliferation or basal apoptosis was seen among cells transfected with either pUSEam/MEK1* or vector alone (data not shown).
As expected transient transfection with constitutively active MEK1 (but not empty vector) up-regulated ERK1/2 signaling in Daoy, MEB-MED-8A, MCF-7, SKBR3, and SKOV3 cells (Fig.  6). Up-regulation of MEK/ERK signaling in Daoy, MEB-MED-8A, SKBR3, and SKOV3 cells also significantly increased the sensitivity of these cells to 17-AAG, as measured by growth inhibition assay and apoptotic index (Fig. 6). In contrast, although activated MEK1 increased ERK1/2 activity in MCF-7 cells (Fig. 6), this did not appear to impact the sensitivity of these cells to 17-AAG (Fig. 6). MEK1 modulation of 17-AAG activity was especially striking in SKOV3 cells, confirming the suggestion from our MEK1 inhibitor studies (Fig. 5A) that this signal system plays a role in dictating the sensitivity of these cells to 17-AAG. These data confirm that up-regulation of MEK/ERK signaling sensitizes human cancer cells to 17-AAG. However, our analysis of MCF-7 cells indicates that this effect is not universal and may vary with cell background.
Inhibition of AKT1 Is Not Required for 17-AAG Activity against Medulloblastoma Cells-Finally, we sought to provide more direct evidence that inhibition of AKT1 is not required for 17-AAG activity against medulloblastoma cells. To do this we studied the impact of up-regulating wild-type AKT1 expression on Daoy cell sensitivity to 17-AAG. First, we confirmed that retroviral infections proceeded with equal efficiency in Daoy cells independent of ERBB2 expression level (Fig. 7A). In contrast to endogenous AKT1, overexpressed exogenous AKT1 was 17-AAG was then added to Daoy cells (250 nM final concentration). SKBR3 and SKOV3 cells were treated with 20 nM 17-AAG that we had previously determined to inhibit the growth of these cells by Ն50%. Cells were cultured for a further 94 h, and the percentage growth inhibition of cells treated with 17-AAG relative to vehicle-only was then determined using an XTT-based assay (differences in 17-AAG-mediated growth inhibition in cells treated with U0126 or PD98059 relative to vehicle-only are marked as a single asterisk (p Ͻ 0.05) or triple asterisk (p Ͻ 0.0005)). B, Daoy.V and Daoy.2 cells were treated for 8 h with MEK1 antisense (M1AS) or scrambled control (M1SCR) oligonucleotide. Total cell lysates derived from treated cells were then analyzed by Western blotting for expression levels of MEK1 and phosphorylated (p) ERK1/2. C, Daoy.V and Daoy.2 cells were treated for 8 h with the M1SCR or M1AS oligonucleotides followed by exposure to 250 nM 17-AAG for 96 h. The percentage growth inhibition of 17AAG relative to vehicle-onlytreated cells was then determined using an XTT-based assay (differences in 17-AAG-mediated growth inhibition in cells treated with M1AS relative to M1SCR are marked as single asterisk (p Ͻ 0.05) or double asterisk (p Ͻ 0.005)). resistant to inhibition by 250 nM 17-AAG (compare Fig. 3 with Fig. 7B). This presumably occurred as a consequence of increasing the total amount of cellular AKT1. Despite this, 17-AAG readily inhibited the growth of Daoy.V control cells and activated ERK1/2 (Fig. 7, B and C). Therefore, inhibition of AKT1 signaling is not required for 17-AAG activity against Daoy cells. Interestingly, although phosphorylated AKT1 is relatively stable in 17-AAG-treated ERBB2-transfected cells (Figs. 3 and 7B), up-regulation of AKT1 reduced the sensitivity of ERBB2-overexpressing cells to 17-AAG, returning it to that of control cells (Fig. 7C). We have previously demonstrated that cross-talk between AKT1 and RAF1 negatively regulates the MEK/ERK pathway in Daoy cells, thereby blocking ERBB2 homodimer signaling via ERK1/2 (26). Indeed, in the current study up-regulation of AKT1 in ERBB2-transfected cells blocked basal ERK1/2 signaling (Fig. 7B, third lane) but did not prevent 17-AAG from inducing ERK1/2 activation (Fig. 7B,  fourth lane). Therefore, up-regulation of AKT1 in ERBB2transfected Daoy cells down-regulates ERBB2 homodimer signaling via MEK/ERK, preventing this pathway from sensitizing Daoy cells to 17-AAG. DISCUSSION The majority of published studies report ansamycins to inhibit rather than enhance MEK/ERK signaling (13,16,37,38,40). In human colonic carcinoma cells (16), mouse splenic B lymphocytes (37), and NIH3T3 cells (39,40), ansamycins degrade upstream components of the mitogen-activated protein kinase pathway, most notably RAF1. This uncouples activators of the pathway from downstream MEK/ERK. Conversely, we found that 17-AAG induces MEK/ERK signaling in a dose-and time-dependent fashion independent of RAF1 in three separate medulloblastoma cell lines. Furthermore, we show that inhibition of MEK1 antagonizes the anti-proliferative activity of 17-AAG against medulloblastoma cells, whereas direct up-regulation of MEK/ERK signaling sensitizes these cells to 17-AAG. HSP90 blockade has been shown to similarly affect MEK/ERK signaling in N2A cells that are also derived from a neuroectodermal tumor (41). Exposure of these cells to 2 M geldanamycin for 10 h resulted in a marked degradation of RAF1, a 15-fold induction of ERK1/2 kinase activity, and significant cell differentiation by 24 h (41). Together, these data suggest that tumor cell type may be an important determinant of how ansamycins affect signal transduction pathways. Indeed, our finding that Daoy cell sensitivity to 17-AAG is AKT1-independent provides further evidence of this. AKT1 activity in ERBB2-overexpressing Daoy cells was unaffected by time-concentration schedules of 17-AAG that induced MEK/ERK signaling and inhibited proliferation by Ն75%. Furthermore, the sensitivity of Daoy control cells to 17-AAG was unaffected by increased expression of exogenous AKT1. These data are in stark contrast to human breast cancer cells in which AKT1 inhibition occurs rapidly after exposure to 17-AAG and is critical for the drug to inhibit cell proliferation (18,19). Cell context-dependent activity of drugs against signal pathways is not without precedent. For example, a diverse range of anti-cancer drugs including paclitaxel, vinca alkaloids (42), cisplatin (43), and topoisomerase II inhibitors (44) modulate MEK/ERK signaling in a manner that varies with cell background.
Although modulation of second messenger pathways by 17-AAG varies between cell lines, our data and those in the literature indicate that ERBB2 uniformly enhances the sensitivity of cancer cells to HSP90 blockade (18,19,23). This may result from a complex interaction between the pattern of ERBB2 dimerization, consequent second messenger activation, and the role of these second messengers in cell survival. In this regard, whereas phosphatidylinositol 3-kinase/AKT-dependent survival of SKBR3 cells is maintained by ERBB2/ERBB3 heterodimers that are acutely sensitive to disruption by ansamy-cins (18,19), we observed no consistent pattern of ERBB2 heterodimerization in medulloblastoma cell lines. Rather, we show that ERBB2 overexpression predisposes medulloblastoma cells to growth inhibition by 17-AAG by up-regulating basal ERBB2 homodimer/MEK/ERK signaling. Nonetheless, our finding that MEK1 activity modulates 17-AAG activity in a variety of human cancer cells, including breast cancer cells, suggests that some degree of overlap exists in the mode of action of these drugs between cell types. Unraveling these complex interactions between cell-specific signaling and drug action will be crucial for the successful clinical development of 17-AAG and related compounds. Total protein lysates were prepared from cells and analyzed by Western blotting for the indicated proteins. C, Daoy.V and Daoy.2 cells were infected with control or AKT1 retrovirus and exposed to 250 nM 17-AAG for 96 h. The percentage growth inhibition of 17AAG relative to vehicle-only-treated cells was then determined using an XTT-based assay. Differences in 17-AAG-mediated growth inhibition in cells infected with AKT1 relative to control are marked with a double asterisk (p Ͻ 0.005).
Our study raises two additional questions that we are currently investigating. The first relates to the mechanism by which 17-AAG activates MEK/ERK signaling independent of RAF1. It has been suggested that the disruption of HSP90/ RAF1 interaction by ansamycins could allow transient oligomerization and activation of RAF1 before degradation (41). An alternative hypothesis is that novel positive regulators of MEK/ERK signaling are responsible for activation by ansamycins. Indeed, a number of positive and negative regulators of the mitogen-activated protein kinase pathway have been identified (45,46). It is possible that primitive neuroectodermal cells express a variety of alternative mitogen-activated protein kinase activators. These might include the oncogenic protein kinase MOS (47,48) that is expressed in a variety of normal and malignant neuronal tissues (49,50). Finally, it will also be important to establish how 17-AAG-induced MEK/ERK signaling generates an anti-proliferative effect in medulloblastoma cells. A number of chemotherapeutic agents have been shown to induce tumor cell apoptosis via MEK/ERK (51,52). Phenethyl isothiocyanate treatment of PC-3 human prostate cancer cells results in sustained activation of ERK1/2 and its substrate ELK-1 and apoptosis that is readily prevented by the MEK1 inhibitor PD98059 (52). In other cell line systems drug-induced interaction between the mitogen-activated protein kinase pathway and caspases has been suggested to account for MEK/ ERK-mediated apoptosis (53). This has clear clinical relevance given the increasing enthusiasm to combine the use of signal transduction inhibitors in the clinic, including the ansamycins and MEK1 inhibitors (51,54,55). Defining the precise mode of action of these agents within specific tumor contexts will be crucial for the efficient clinical development of these compounds.