Phosphoprotein Pathway Mapping: Akt/Mammalian Target of Rapamycin Activation Is Negatively Associated with Childhood Rhabdomyosarcoma Survival

Rhabdomyosarcoma arises from undifferentiated mesenchymal cells bearing skeletal muscle features (1, 2). Rhabdomyosarcoma is the most common soft tissue sarcoma in children, consisting of three histologic subtypes—alveolar, embryonal, and botyroid. Despite the recent advances in combination chemotherapy, and the molecular knowledge of the translocations t(2;13)(q35;q14) and t(1;13)(p36;q14) in alveolar rhabdomyosarcoma, the overall survival of all patients with childhood rhabdomyosarcoma has remained in the range of 60% to 70% (3, 4).

The Children's Oncology Group (COG) completed a series of treatment protocols evaluating the activity of a combination regimen using vincristine, actinomycin D and cyclophosphamide (VAC), VA plus ifosfamide (VAI), or VI and etoposide (VIE) in newly diagnosed children who presented with nonmetastatic rhabdomyosarcoma (3, 4). IRS-IV accrued patients between 1991 and 1997 with patients randomized to chemotherapy regimens (except for those with group I paratesticular and group I/II orbital primary tumors and patients with preexisting renal dysfunction). Additionally, patients were randomized to conventional or hyperfractionated radiotherapy for patients with group III disease. No difference in outcome was seen among the chemotherapy regimens or between the radiotherapy regimens for group III patients (4).

D9502 was a pilot study to assess the feasibility of cyclophosphamide dose intensification during the first 12 weeks of VAC therapy for patients with intermediate risk rhabdomyosarcoma. Dose intensification therapy did not result in outcome differences compared with VAC with lower cyclophosphamide cycle doses (3). D9803 is an on-going study comparing VAC to VAC alternating with vincristine, topotecan, and cyclophosphamide for patients with intermediate risk rhabdomyosarcoma.

The overall disease-free survival rate in these studies was 67%, but the newer VA plus ifosfamide or VI and etoposide combinations did not improve overall outcome and survival compared with the standard VAC regimen (3, 4). Unfortunately, there is no way to identify the 33% of children destined to fail initial therapy, regardless of disease stage or histologic subtype. On the other hand, the 60% to 70% of children that respond to standard therapy do so exceedingly well, with a vast majority of these patients currently living disease-free. Consequently, an urgent clinical goal is to identify functionally important molecular networks associated with the 30% to 40% nonresponder rhabdomyosarcoma subjects to develop new treatment strategies for this group. Because kinases are important drug targets, kinase network information could become the basis of therapeutic strategies for improving treatment outcome (5).

Although gene microarrays can provide important information about somatic genetic taxonomy, they are unable to provide an effective recapitulation of the posttranslational, fluctuating signaling events that occur at the proteomic level. The phosphorylation, or activation state, of kinase-driven signaling networks contains important information concerning both the disease pathogenesis as well as potential for therapeutic target selection (5, 6). It is for this reason that modulation of ongoing cellular kinase activity represents one of the most rapidly growing arenas in new drug development (7). The phosphorylation status of proteins can be detected and measured using specific antiphosphoprotein antibodies. Antibodies have been developed to specifically recognize the phosphorylated isoform of kinase substrates. Through the use of these phosphospecific antibodies, it is now possible to evaluate the state of entire portions of a signaling pathway or cascade, although the cell is lysed, by looking at dozens of kinase substrates at once through multiplexed phosphospecific antibody analysis (5, 8, 9). Applying these antibodies to reverse-phase protein arrays provides the opportunity to profile the state of the ongoing cellular signaling events within small numbers of human tumor cells obtained by biopsy (5, 9, 10). Sandwich two-site antibody pairs are not available for most phosphorylated epitopes. The advantage of the reverse phase arrays is that this format requires only one antibody to measure each phosphorylated epitope.

We applied the reverse phase array phosphoprotein pathway mapping to evaluate the hypothesis that prosurvival pathways, including Akt/mammalian target of rapamycin (mTOR), are differentially activated for rhabdomyosarcoma tumors associated with a poor versus favorable outcome. A secondary goal was to functionally test the effect of treatment on the pathway we found to be activated in a xenograft model of human rhabdomyosarcoma. This study shows the importance of phosphoprotein cell signaling events and how they may constitute a new and functionally relevant analyte for deriving therapeutic insights for rhabdomyosarcoma.

Specimens and patient data. All specimens (n = 59) and relevant clinical data were obtained from the Intergroup Rhabdomyosarcoma Study (IRS) IV, D9502 and D9803, studies from the COG with appropriate institutional review board approval (3, 4).

All specimens were snap frozen in liquid nitrogen and procured before therapy. The sample set was analyzed in two groups, 1A and 1B (Fig. 1A). Figure 1B, to C shows the survival characteristics for the two study sets. Samples were anonymized and blinded as to clinical survival outcome before final data analysis. The samples representing study set 1A (Fig. 1A) consisted of nine snap-frozen surgical specimens and 290 frozen section slides for 33 different patients with a pathologic diagnosis of rhabdomyosarcoma. All patients used here had stage III (tumors <5 cm or regional lymph node involvement) disease and group III tumors (gross residual disease remaining after treatment) before study entry. An additional set of 26 frozen section samples and clinical data were provided by the COG for patients from the same protocols (Fig. 1A). Pathologic diagnosis was rendered before therapy. An independent board-certified pathologist verified the diagnosis before laser capture microdissection. The histologic subtypes represented alveolar, embryonal, botryoid, and mixed morphologic types. Pure tumor cell populations were microdissected from the tissue sections with a PixCell II laser capture microdissection instrument (Molecular Devices, Sunnyvale, CA).

Reverse phase protein microarrays. The microdissected cells (5, 9, 10) were subjected to lysis and reverse phase protein microarrays were printed in duplicate with the whole-cell protein lysates as described by Sheehan et al. (10). Briefly, the lysates were printed on glass backed nitrocellulose array slides (FAST Slides Whatman, Florham Park, NJ) using a GMS 417 arrayer (Affymetrix, Santa Clara, CA) equipped with 500 μm pins. Each lysate was printed in a dilution curve representing neat, 1:2, 1:4, 1:8, 1:16, and negative control dilutions. The slides were stored with desiccant (Drierite, W.A. Hammond, Xenia, OH) at −20°C before immunostaining.

Protein microarray immunostaining. Immunostaining was done on an automated slide stainer per manufacturer's instructions (Autostainer CSA kit, DAKO, Carpinteria, CA). Each slide was incubated with a single primary antibody at room temperature for 30 min. Polyclonal primary antibodies were as follows: glycogen synthase kinase-3 (GSK3) α/β Tyr^279/216 (Invitrogen-Biosource, Carlsbad, CA), BCL-2, HIF-1α (BD, Franklin Lakes, NJ), 4EBP1, FKHR ser²⁵⁶, eIF4E, eIF4E ser²⁰⁹, eIF4G, eIF4G Ser¹¹⁰⁸, IGFR-β, IRS-1, IRS-2, IRS-1 Ser⁶¹², SGK, Bak, Bax, BAD, BAD Ser¹¹², BAD Ser¹³⁶, BAD Ser¹⁵⁵, B-Raf, mTOR, mTOR Ser²⁴⁴⁸, p70S6 Thr³⁸⁹, p70S6 kinase, p70S6 Ser³⁷¹, S6 kinase Ser^240/244, Akt, Akt Ser⁴⁷³, Akt Thr³⁰⁸, 4EBP1 Ser⁶⁵, 4EBP1 Ser⁷⁰, and 4EBP1 Thr^37/46 (Cell Signaling Technology, Danvers, MA). The negative control slide was incubated with antibody diluent. Secondary antibody was goat anti-rabbit IgG H+L (1:5,000; Vector Labs, Burlingame, CA).

Bioinformatics method for microarray analysis. Each array was scanned; spot intensity was analyzed; data were normalized; and a standardized, single data value was generated for each sample on the array (Image Quant v5.2, GE Healthcare, Piscataway, NJ). Spot intensity was integrated over a fixed area. Local area background intensity was calculated for each spot with the unprinted adjacent slide background. This resulted in a single data point for each sample, for comparison with every other spot on the array. The Ward method for two-way hierarchical clustering was done using JMP v5.0 (SAS Institute, Cary, NC). Wilcoxon two-sample rank sum test was used to compare values between two groups. P values <0.05 were considered significant. When we could not assume a normal distribution of the variables, we used nonparametric methods. We used Kaplan-Meier (log-rank) survival estimates for univariate survival analysis.

In vivo xenograft tumor model. Animal studies were done in accordance with guidelines of the NIH Animal Care and Use Committee. Female 4- to 6-week-old beige-severe combined immunodeficient (SCID) mice were purchased from Charles River Laboratories (Wilmington, MA). Two million viable cells harvested from mid confluent cultures of either Rh30 alveolar or RD embryonal cells in 0.2 mL diluent (5% Tween 80, 5% polyethylene glycol 400; Sigma, St. Louis, MO) were injected orthotopically into the gastrocnemius muscle in the left hind leg, and after 1 week mice were randomly assigned to control (n = 8) or CCI-779 treatment groups (n = 8). Mice were treated i.p. every 3 days for 30 consecutive days with 20 mg/kg/i.p. of CCI-779 (Developmental Therapeutics Program, National Cancer Institute and Wyeth, Madison, NJ) or vehicle alone. Tumor growth was measured every 3 days with calipers, and tumor volume was calculated by the formula V (mm³) = a × b², where a is the longest tumor axis and b is the shortest tumor axis. All mice were sacrificed by asphyxiation with CO₂ and underwent necropsy for confirmation of tumor growth. Tumors were excised and snap frozen at −80°C until analysis.

Exploratory data analysis of rhabdomyosarcoma tumor set 1A. Enrichment of tumor cells by laser capture microdissection was done before analysis to ensure that the cells for analysis came from within the cancer cell population, without contamination by noncancer cells (5, 11). For study set 1A (n = 33), 15 specific signaling proteins (Fig. 2A) were initially chosen for reverse phase protein microarray analysis because they constituted a broad survey of multiple prosurvival related events related to the Akt/mTOR pathways that have been shown to play a role in rhabdomyosarcoma (12, 13). Unsupervised hierarchical clustering analysis of the 15 protein end points revealed two major classes of tumors: one cluster with Akt/mTOR activation/phosphorylation and the other with a comparatively low level of signaling (Fig. 2A). After clinical outcome data was obtained from the COG, these two clusters were compared by Fisher's exact test based on patient characteristics of age, sex, primary site, histology, invasion, and lymph node involvement (Fig. 2B; ref. 14). Although none of the characteristics reached P < 0.05 statistical significance, patients with parameningeal head and neck primary site tumors comprised 62% of cluster 1, whereas cluster 2 had 27% of patients with parameningeal primary site tumors (Fisher's exact test, P = 0.06). Additionally, cluster 2 contained 73% alveolar tumors, whereas cluster 1 had 62% embryonal tumors (Fisher's exact test, P = 0.06). Typically, patients with embryonal rhabdomyosarcoma tumors from orbital or nonparameningeal sites have the best prognosis (15). These two clusters were not statistically different for commonly accepted prognostic/clinical factors associated with rhabdomyosarcoma.

We proceeded to correlate the protein analyte values with disease-free and overall survival clinical outcome data provided by the COG for study set 1A. A clear partitioning of the tumors emerged after clinical outcome data was obtained from the COG. A decision tree analysis of three proteins—4EBP1, phosphorylated 4EBP1 Thr^37/46, and eIF4E—all components of the Akt/mTOR pathway, partitioned patients who were in continuous complete remission from those who recurred and died after being treated with standard therapy (data not shown). Among these end points, 4EBP1 and 4EBP1 Thr^37/46 individually were found to be significantly correlated with survival by Wilcoxon one-way analysis, 4EBP1 (P < 0.0064) and 4EBP1 Thr^37/46 (P < 0.0135; Fig. 3A). A log-rank univariate survival analysis (Kaplan-Meier) supported the association of 4EBP1 with outcome in overall and recurrence-free survival in study set 1A (Fig. 3B; overall survival P = 0.0177, recurrence-free survival P = 0.0370; ref. 16).

For recurrence-free survival in study set 1A, 4EBP1 level (P₂ = 0.0074; hazard ratio, 7.44; 95% confidence interval, 1.71–32.36) emerged as significant prognostic factor (17, 18). Thus, for study set 1A (Figs. 1–3), individual components within the Akt/mTOR pathway seemed to correlate with survival.

Disease-free and overall survival in rhabdomyosarcoma patients is associated with phosphorylated components of the Akt and mTOR pathways. Based on the findings of study set 1A, an independent set of samples (set 1B, Fig. 1A) were obtained from COG (n = 26) for analysis of an expanded set of proteins associated with the Akt/mTOR pathway. Univariate log-rank analysis of the two heterogeneous sample sets (set 1A and 1B) revealed no significant difference in overall or recurrence-free survival by sample set (overall survival P = 0.2111, recurrence-free survival P = 0.5824; Fig. 1B) or histology (overall survival P = 0.4103, recurrence-free survival P = 0.4312; Fig. 1C). We analyzed set 1B by laser capture microdissection and reverse phase protein microarray as in set 1A. We expanded the number of end points to 27 to include additional signaling proteins upstream and downstream of Akt and mTOR for an independent evaluation of pathway activation.

Following unblinding of the data, the results for study set 1B (Fig. 4) showed a significant association of disease-free and overall survival with phosphorylated components of the Akt-mTOR pathway. High levels of Akt Ser⁴⁷³, 4EBP1 Thr^37/46, eIF4G Ser¹¹⁰⁸, and p70S6 Thr³⁸⁹ were all significantly associated with poor overall and poor disease-free survival: Akt Ser⁴⁷³ (overall survival P < 0.001, recurrence-free survival P < 0.0009), 4EBP1 Thr^37/46 (overall survival P < 0.0110, recurrence-free survival P < 0.0106), eIF4G Ser¹¹⁰⁸ (overall survival P < 0.0017, recurrence-free survival P < 0.0072), and p70S6 Thr³⁸⁹ (overall survival P < 0.0085, recurrence-free survival P < 0.0296; Fig. 4A). Each of the 27 components was also evaluated individually for statistical correlation with survivor versus nonsurvivor status. Six end points—again, all components of the Akt/mTOR network (4EBP1 Thr³⁷, Akt Ser⁴⁷³, eIF4G Ser¹¹⁰⁸, p70S6 Thr³⁸⁹, Bak, and GSK3α/β Tyr^279/216)—correlated independently with survival [Wilcoxon one-way analysis 4EBP1 Thr^37/46 (P < 0.0348), GSK3α/β Tyr^279/216 (P < 0.0348), eIF4G Ser¹¹⁰⁸ (P < 0.0196), Akt Ser⁴⁷³ (P < 0.0227), Bak (P < 0.0321), and p70S6 Thr³⁸⁹ (P < 0.0373); Fig. 4B].

IRS-1/Akt/mTOR feedback loop is dysregulated in nonsurvivor cohort. Although tyrosine phosphorylated insulin receptor substrate-1 (IRS-1) activates Akt/mTOR signaling through phosphatidylinositol 3-kinase (PI3K), serine phosphorylation of IRS-1 at serine⁶¹² by mTOR and p70S6 down-regulates IRS-1 tyrosine activation (19–21). Thus, IRS-1 is subject to negative feedback regulation in response to Akt/mTOR activation (Fig. 5A). We examined levels of phosphorylated members of the IRS-1/Akt/mTOR feedback loop by reverse phase protein microarray for the tumors in study set 1B (n = 26). Although levels of IRS-1 Ser⁶¹² were no different between the survivors and nonsurvivors, phosphorylation of IRS-1 Ser⁶¹² correlated strongly with phosphorylation of mTOR at Ser²⁴⁴⁸ in the survivor cohort (Spearman's Rho nonparametric P < 0.0027), suggesting a linkage between these two signaling events (Fig. 5B). By contrast, the phosphorylation of these same two signaling proteins was not correlated in the nonsurvivor cohort (Spearman's Rho nonparametric P = 0.7358; Fig. 5B–C). This lack of correlation with IRS-1 Ser⁶¹² phosphorylation also prevailed for the mTOR downstream components eIF4E Ser²⁰⁹ (survivor P = 0.0006, nonsurvivor P = 0.102) and p70S6 Thr³⁸⁹ (survivor P = 0.00004, nonsurvivor P = 0.1827; Fig. 5B and D). Thus, the interrelationship between IRS-1 activity and the mTOR pathway proteins may be altered in the tumors of patients, which subsequently are found to have poor survival after chemotherapy compared with the tumors of patients who have long-term survival (Fig. 5A–D).

Interrogation of the phosphorylated versus nonphosphorylated state of proteins. Phosphorylation is an important posttranslational modification that has potential significance as a readout for the activation state of pathways and kinase inhibitor targets. To further investigate potential significant cell signaling proteins within the IRS-1/Akt/mTOR pathway, we extended our analysis to include the following additional end points: BAD, eIF4G, IRS-1, IRS-2, IGFR-β, and S6 Ser^240/244. We conducted Wilcoxon one-way analysis and Kaplan-Meier survival analysis for the phosphorylated protein, the total protein form, and the ratio of the phosphorylated to total forms of key protein end points (Fig. 5B; Supplementary Data). The results clearly show that the specific phosphorylated forms of the protein end points within the Akt-mTOR and associated pathways are independently associated with survival (P < 0.05) compared with the nonphosphorylated total form of the analyte protein (4EBP1 Thr^37/46 P < 0.035, p70S6 Thr³⁸⁹ P < 0.037, Akt Ser⁴⁷³ P < 0.023, eIF4G Ser¹¹⁰⁸ P < 0.02). This is an important distinction because it is likely that the population of the total protein in a signal pathway node is in excess compared with the phosphorylated form. The phosphorylated form, actively engaged in signaling, constitutes a subpopulation of the total protein. Thus, the phosphorylated form constitutes a variable that is independently correlated with survival compared with the total protein (Fig. 5; Supplementary Data).

Suppression of the mTOR pathway in a mouse xenograft model reduces tumor growth. To validate the functional significance of our IRS-1/Akt/mTOR network analysis, we used rapamycin analogues, which are well-characterized inhibitors of the mTOR protein kinase pathway, using a mouse xenograft treatment model. Either RD embryonal cells or Rh30 alveolar cells were injected orthotopically into the hind leg of beige SCID mice. These two different cell lines were used to determine the effects of mTOR inhibition in different histologic tumor categories. The rapamycin analogue CCI-779 (Wyeth, Madison, NJ) dosage was 20 mg/kg, which corresponds to dosages currently administered to humans in phase I and II clinical trials (22, 23). Administration of CCI-779 at doses that were verified to suppress the phosphorylation of mTOR downstream targets profoundly reduced the growth of rhabdomyosarcoma xenografts as measured in the beige SCID murine model (Rh30 xenograft group P = 0.0002; RD xenograft group P = 0.00008, n = 8 for both groups; Fig. 6A–D). Suppression of the mTOR pathway was monitored by measuring the phosphorylation of 4EBP1 and S6 ribosomal protein, which are well-established downstream targets of mTOR (12). CCI-779 inhibited the phosphorylation of these downstream targets commensurate with a blockade in mTOR signaling in both the Rh30 alveolar– and RD embryonal xenograft–derived tumors. In addition, there was a slight but definite increase in phosphorylated Akt at Ser⁴⁷³ in both the RD and Rh30 xenograft tumors over the course of treatment with CCI-779 (Fig. 6D).

Recent real-time reverse transcription-PCR and fluorescence in situ hybridization assays for PAX3-FKHR and PAX7-FKHR fusion transcripts have been developed and applied to rhabdomyosarcoma. Nevertheless, genomic and transcriptomic assays do not provide an effective recapitulation of the posttranslational, fluctuating signaling events that occur at the proteomic level (24–26). Therefore, we used reverse phase protein microarrays as a means of monitoring the in vivo state of selected kinase pathways.

In the present investigation, analysis of protein signaling pathways was conducted blinded to treatment or survival using two independent rhabdomyosarcoma tumor study sets for which 12-year follow-up data was available. Two independent study sets (Fig. 1A) were procured randomly from the pool of frozen specimens. Each study set represented a variety of treatment modalities, histologic subtypes, and tumor sites. The two sets differed in the proportion of samples with alveolar versus embryonal histology (Fig. 1A and C; refs. 3, 4). Although the sample sets were heterogeneous, there was no statistically significant difference in either overall survival or recurrence-free survival between the two sample sets (overall survival P = 0.2111, recurrence-free survival P = 0.5824; Fig. 1B).

Current prognostic indicators for patients diagnosed with rhabdomyosarcoma are age, stage, group, histology, and primary site, with patients in the 1- to 8-year age group with embryonal rhabdomyosarcoma from orbital or nonparameningeal head and neck sites having the best prognosis (15). Using unsupervised clustering analysis, we sought to determine if any protein signaling signature correlated with histologic subtype. For the first study set, 15 specific signaling proteins (Figs. 2A and 3A) were initially chosen because they constituted a broad survey of multiple prosurvival-related events. A multiplexed measurement of the chosen phosphorylation states provided an averaged portrait of the ongoing kinase activity events within selected networks that drive cellular growth or survival.

The initial unsupervised clustering analysis was not significantly associated with histology but there was clear partitioning of the samples into two clusters, with one cluster exhibiting activation of Akt/mTOR proteins (Fig. 2A). Therefore, clinical outcome data was obtained from the COG for further exploratory associations between the protein end points and clinical data. The results of set 1A revealed a statistically significant association between survival and the activation/suppression of proteins linked to the Akt/mTOR signaling pathway (Fig. 3A–B).

Based on the results of set 1A, we expanded this exploratory analysis to 27 end points applied to a second independent set of samples (Fig. 1A; and Supplementary Data). Proteins that seemed to correlate with survival or failure in the second study set were linked together in the Akt/mTOR kinase pathway (12, 27, 28). Phosphorylated components of the Akt/mTOR pathway, specifically Akt, eIF4G, 4EBP1 (elongation binding factor), GSK3α/β, and p70S6 were found to be associated with outcome (Fig. 4B). IRS-1, Akt, and GSK3β are associated with cell growth, survival, insulin response, and glucose metabolism. mTOR, 4EBP1, and p70S6 are essential components of protein translation, in which phosphorylation of 4EBP1 releases 4EBP1 from eIF4E, activating cap-dependent translation (29). These pathways are involved in the regulation of prosurvival and translation for a group of proteins that are important for cell cycle and apoptosis, including several known oncogenes such as cyclin D, c-myc, and Hif-1α (30). Although Akt Ser⁴⁷³ correlated with survival (P < 0.02) for study set 1B, it did not correlate with survival in set 1A (P = 0.2460). This may have been due to differences in the relative composition of tumor histologies and sites of origin between the two groups (Fig. 1A).

A variety of autocrine and paracrine stimuli, including hormones, growth factors, mitogens, cytokines, and G-protein–coupled receptor agonists, elicit 4EBP1 hyperphosphorylation and concomitant loss of eIF4E-binding activity in the mTOR pathway (12, 27, 28, 30). Activation of PI3K or the downstream effector kinase Akt leads to 4EBP1 hyperphosphorylation, affecting its release from eIF4E. Phosphorylation of 4EBP1 on multiple loci is associated with linkage to the insulin receptor pathway and the PI3K pathway. Six phosphorylation sites have been identified on 4EBP1. Thr³⁷, Thr⁴⁶, Ser⁶⁵, and Thr⁷⁰ become phosphorylated after insulin stimulation, and such phosphorylation can be blocked by rapamycin (inhibitor of mTOR) and wortmannin (inhibitor of PI3K; refs. 31, 32). It has been shown that mTOR itself, as well as an mTOR-associated kinase, directly phosphorylates sites on 4EBP1. Gingras et al. (31) established that phosphate groups are first added to Thr³⁷ and Thr⁴⁶. This priming phosphorylation is required for the phosphorylation of other sites necessary for binding. Thus, multiple phosphorylation events triggered from multiple kinases, primed by Thr^37/46, are involved in the release of 4E-BP1 from eIF4E.

Tyrosine-phosphorylated IRS-1 activates Akt/mTOR signaling through PI3K, serine phosphorylates IRS-1 (at Ser⁶¹²) by mTOR, and p70S6 down-regulates IRS-1 tyrosine activation (19–21). Thus, it is hypothesized that IRS-1 is subject to negative feedback regulation in response to Akt/mTOR activation through p70S6 (Fig. 5A; ref. 33). Based on this negative feedback between mTOR and IRS-1, O'Reilly et al. (33) recently hypothesized that inhibition of mTOR may cause augmented phosphorylation of Akt through disruption of the negative feedback loop between p70S6 and IRS-1. We examined the IRS-1 feedback loop interrelationship with components of the Akt and mTOR pathway by nonparametric correlations (Fig. 5B–D). Interrogation of IRS-1 Ser⁶¹² and various potential interacting proteins provided a means to assess the protein interactions with the actual phosphorylation site involved in the negative feedback regulation of IRS-1 (20, 21). The average level of IRS-1 Ser¹² was not statistically different (P < 0.098) between tumors from patients with favorable outcome compared with those with poor outcome (Fig. 4B), suggesting that the level of IRS-1 upstream activity was similar. Although the average level of IRS-1 Ser⁶¹² phosphorylation was similar in the favorable versus poor outcome cases, the correlation of individual IRS-1 phosphorylation levels in each tumor with phosphorylation levels of Akt and mTOR pathway proteins was highly dissimilar in these two phenotypes. In addition Bak, FKHR Ser²⁵⁶, and IGFR-β were significantly correlated for both groups (Supplementary Data). As shown in Fig. 5, there was a strong positive correlation (P = 0.00269) of IRS-1 Ser⁶¹² with mTOR Ser²⁴⁴⁸ and with p70S6 Thr²⁸⁹ (P = 0.00004) in tumors with favorable outcome (Fig. 5B and C). This suggests a linkage or correlation consistent with a feedback loop between mTOR and IRS-1 in these tumors with favorable outcomes. These data support a selective alteration in the active interrelationship between IRS-Ser⁶¹² and downstream components of the mTOR pathway in tumors with poor outcome.

The implications of these differences in the IRS-Akt-mTOR interconnectivity of survivors and nonsurvivors are 2-fold. First, the apparent lack of interconnection between IRS-1 and mTOR could disrupt the normal negative feedback regulation as described (20, 33). This could result in increased phosphorylation of Akt as we noted in the tumors from patients with poor outcomes and illustrated in Fig. 5A. Baseline levels of phosphorylated Akt and mTOR may be elevated in aggressive tumors in which the negative feedback regulation of mTOR through IRS-1 is disrupted, or phosphorylation of Akt is supplemented by IRS-1–independent mechanisms.

The identified 4E-BP1 phosphorylation sites are specifically inhibited by rapamycin treatment (29, 31). To validate the functional significance of our network analysis, revealing mTOR pathway suppression observed in patients who had a favorable treatment outcome, we exploited the existence of rapamycin analogues, which are well-characterized inhibitors of the mTOR protein kinase pathway. Some of these analogues are currently in phase I and II clinical trials of adults with cancer (22, 23). Suppression of the mTOR pathway was monitored by measuring the state of phosphorylation of 4EBP1 and S6 kinase, which are well-established downstream substrates of mTOR (12, 27, 28, 30, 31). CCI-779 inhibited the expected phosphorylation of the downstream targets commensurate with a blockade in mTOR signaling in xenograft tumors derived from Rh30 alveolar or RD embryonal cells (Fig. 6). Phosphorylation of Akt was present in both of these cell lines grown in vivo (Fig. 6D) and there was a slight increase in the RH30 (alveolar) xenograft and a perceptible augmentation in the RD (embryonal) xenograft of phosphorylated Akt on Ser⁴⁷³ over time during CCI-779 treatment (Fig. 6D). This is consistent with the negative feedback loop as described by O'Reilly et al. (33). The growth of both histologic tumor types was inhibited by CCI-779 treatment. In Fig. 4, both embryonal and alveolar rhabdomyosarcoma subtypes were represented in the cohort with poor outcomes, which had higher levels of mTOR pathway activation.

The findings in the present study for both the patient's tumor tissue and the xenograft model are supported and complemented by small interfering RNA S6K1 knockdown experiments in Rh30 and RD cell lines (34). Both the patient tumor and mouse xenograft findings (Figs. 5B–D and 6) support a key regulation of Akt signaling through mTOR and IGF-IRS-1 as shown by the correlation between IRS-1 Ser⁶¹² and other members of these pathways. Moreover, a recent report from the COG using oligonucleotide gene expression analysis revealed an association of elevated platelet-derived growth factor receptor and insulin-like growth factor gene transcripts with decreased survival for patients from the IRS-IV study set (35). Our finding of an increased amount of total IGFR-β associated with poor survival (Fig. 5B) is in keeping with this finding. Thus, the present data support the conclusion of Wan et al. (34) that an optimal therapy strategy would be to combine an mTOR inhibitor with an IGF-IR antibody (or small molecule) inhibitor.

The IGF-I/Akt/mTOR pathway found herein to be associated with rhabdomyosarcoma has previously been associated with regulation of hypertrophy and atrophy of muscle, the cellular progenitor of rhabdomyosarcoma Song et al. (36) have reported that IGF-I, via the mTOR pathway, regulates angiotensin II–induced skeletal muscle wasting. Rat skeletal muscle atrophy is associated with inactivation of the mTOR pathway. Activation of the mTOR pathway is also associated with active remodeling of the heart in vivo and protection against atrophy, whereas rapamycin treatment of rat hearts augmented atrophy (37). Cardiac myocyte reperfusion after transient ischemia induced by reversible occlusion was associated with activation of the PI3K/Akt/mTOR pathway (38). Reperfusion was associated with activation of Akt and phosphorylation of GSK3β, and associated stimulation of signaling through mTOR as evidenced by phosphorylation of eIF4E and S6 kinase. Activation of the mTOR pathway and phosphorylation of S6K1 is also associated with hypertrophy and pressure overload hypertrophy. Post ischemia, surviving myocytes undergo 10% to 15% hypertrophy (39). Phosphorylation of S6K1 was unaltered during ischemia but increased during reperfusion (38).

These results in nonneoplastic muscle indicate that a physiologic cycle of ischemia followed by reperfusion activates the Akt/mTOR pathway, is associated with myocyte hypertrophy, and protects against atrophy. This may be relevant during the precursor state before neoplastic growth, but they may also apply to cycles of ischemia, necrosis, and angiogenesis associated with aggressive neoplastic growth. Because rhabdomyosarcoma is thought to arise from a myocyte lineage, signaling pathways involved in normal muscle cell regrowth/remodeling could be similar to those involved in tumorigenesis.

Mirk, a protein recently identified to possess antiapoptotic functions in rhabdomyosarcoma cell lines and tumors, exhibits both prosurvival and growth arrest functions in differentiating skeletal myoblasts (40, 41). Mirk has been shown to be activated by MKK3, a stress-activated mitogen-activated kinase kinase (42). RNA interference (RNAi) depletion of endogenous Mirk reduced the clonogenicity of RD embryonal and Rh30 alveolar rhabdomyosarcoma cell lines in colony formation experiments, indicating a prosurvival role in rhabdomyosarcoma (41). In addition, RNAi depletion of Mirk has been shown to block myoblast differentiation in skeletal muscle primary cell culture (40). These findings further strengthen the role of cell survival mechanisms in rhabdomyosarcoma and muscle development.

In summary, protein pathway analysis of microdissected human rhabdomyosarcoma clinical specimens, procured before treatment, revealed a strong association between activation (phosphorylation) of multiple interconnected Akt/mTOR pathway components and a poor disease-free or overall survival in this initial, exploratory analysis. This observation was found to be consistent between two independently analyzed clinical study sets. Moreover, the functional significance of IRS-1/Akt/mTOR pathway activation in rhabdomyosarcoma was verified using the specific targeted inhibitor CCI-779 to suppress tumor growth in a beige SCID rhabdomyosarcoma xenograft model. These data provide impetus for testing rapamycin analogues in this tumor type as a potential way to modulate poor prognosis patients into more durable outcomes. In support of this combination therapy concept, MacKiegan et al. (43) used RNAi libraries to search for kinases whose inhibition resulted in increased rates of apoptosis. mTOR knockdown was found to sensitize cells to Taxol-induced apoptosis (43). Based on the interrelationship of the Akt/mTOR pathway and the IGF-IRS pathway (Fig. 5), future combination therapy strategies could be aimed at blocking both upstream IRS-1–mediated signaling factor activation, as well as downstream mTOR signaling, as a means of augmenting standard cytotoxic rhabdomyosarcoma therapy.

Grant support: Intramural Research Program of the NIH, National Cancer Institute, and George Mason University.

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.

The authors thank James Anderson for critical reading of the manuscript, and William Mayer for assistance in providing samples.