SEARCH

SEARCH BY CITATION

Keywords:

  • microarray;
  • peritoneal mesothelioma;
  • cancer pathways;
  • patient survival

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

BACKGROUND:

Malignant peritoneal mesothelioma (MPM) is a rare malignancy of the serosal membranes of the abdominal cavity. This cancer is ultimately fatal in almost all afflicted individuals; however, there is marked variability in its clinical behavior: Some patients die rapidly, and others survive for many years. In the current study, the authors investigated the molecular nature of MPM to obtain insights into the heterogeneity of its clinical behavior and to identify new therapeutic targets for intervention.

METHODS:

Fresh pretreatment tumor samples were collected from 41 patients with MPM who underwent surgical cytoreduction and received regional intraoperative chemotherapy perfusion. From those samples, gene expression analyses were performed. The major cellular pathways that were identified in this cancer were inhibited using a pathway-specific inhibitor.

RESULTS:

Unsupervised clustering of genes identified 2 distinct groups of patients with significantly different survivals (Group A: median survival, 24 months; Group B: median survival, 69.5 months; P = .035). Phosphoinositide-3-kinase (PI3K) and the closely interacting mammalian target of rapamycin (mTOR) signaling pathways were overexpressed predominantly in the poor survival group; and the genes of these pathways, phosphoinositide-3-kinase, catalytic, α polypeptide (PIK3CA) and rapamycin-insensitive companion of mammalian target of rapamycin (RICTOR), were highly significantly predictive of shortened patient survival in Group A. The role of these pathways in MPM tumor progression was also investigated by treating 2 MPM cell lines with BEZ235, a dual-class PI3K and mTOR inhibitor, and the authors observed significant inhibition of downstream cell signaling and cell proliferation.

CONCLUSIONS:

Taken together, the results from this study revealed that, based on gene expression profiles, there were 2 distinct patient groups with significantly different survival and that targeting the PI3K and mTOR signaling pathways may have significant therapeutic value in patients with MPM. Cancer 2011. © 2010 American Cancer Society.

Malignant peritoneal mesothelioma (MPM) is a progressive and lethal malignancy of the serosal membranes of the abdominal cavity. MPM is a rare cancer; of the approximately 3500 patients who are diagnosed with malignant mesothelioma in the United States every year, those with MPM account for approximately 15% to 20% of all mesothelioma diagnoses. There is growing evidence that MPM differs from pleural mesothelioma; it has a distinct natural history and afflicts both sexes approximately equally, whereas pleural mesothelioma has a higher incidence among men, and the contribution of asbestos exposure as an etiologic factor is not as well established in MPM.1, 2 Macroscopically, MPM is characterized by innumerable tumor nodules of variable size located diffusely throughout the peritoneal cavity and frequently resulting in massive malignant ascites; morbidity and mortality almost always are because of disease progression within the peritoneum.

To date, there is no curative therapy for patients with MPM. Systemic treatment for this condition has modest clinical benefit; pemetrexed with cisplatin or gemcitabine has produced overall response rates of 26% and 15%, respectively, with stabilization of disease observed in up to an additional 50% of treated patients.3-5 Currently, operative cytoreduction and intraoperative or early postoperative intraperitoneal chemotherapy are treatment options offered to many patients who have a good performance status and a disease burden that is amenable to resection.6, 7

MPM has marked heterogeneity in its clinical behavior; some patients die rapidly after initial treatment, whereas others survive for many years,8 a phenomenon that was recognized many years ago.9, 10 However, because of the rare incidence of this cancer, there is limited information about the molecular basis of MPM tumor progression. Some studies,6, 11 but not all,7, 12 have noted that certain histologic subtypes, such as tubulopapillary or epithelioid tumors, have a better prognosis than biphasic or sarcomatiod tumors. One study analyzed the molecular characteristics of MPM tumors from 16 patients and identified a distinct gene expression pattern in the epithelioid subtype compared with the pattern in the biphasic and sarcomatous subtypes. That study also demonstrated that the ubiquitin-proteosome pathway was up-regulated in biphasic tumors.13 In the current study, we used global gene expression profiling of 41 fresh patient tumors to determine whether we could characterize important pathways that influence the molecular basis of MPM tumor progression and to identify possible targets for therapy. We were able to demonstrate that 2 distinct tumor gene expression patterns are associated significantly with different patient survival in a uniformly treated patient population and that the predominantly overexpressed phosphoinositide-3 kinase (PI3K) and mammalian target of rapamycin (mTOR) pathways identified in these patients are potential therapeutic targets.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Patient Characteristics and Treatment Parameters

The overall survival analyses encompass data from 41 consecutively treated patients with histologically proven MPM who 1) underwent laparotomy with the objective of achieving complete gross cytoreduction and 2) received a 90-minute hyperthermic intraoperative intraperitoneal perfusion with chemotherapy (HIPEC). All patients had tissue collected on institutional review board-approved clinical protocols at the National Cancer Institute (NCI), and informed consent was obtained from all patients. All patients were treated with a uniform approach: Briefly, patients underwent laparotomy, cytoreduction, and HIPEC as described previously.7 During HIPEC, the peritoneal cavity was maintained at a median temperature of 41°C, and cisplatin (Bristol-Meyers Squibb, NJ), mixed in 1 L of 0.9% sodium chloride solution (United States Pharmacopeia), was added to the perfusate at a dose of 250 mg/m2. Perfusion was continued for 90 minutes. Sodium thiosulfate was given as a renal protectant with a loading dose of 7.5 g/m2 intravenously over 20 minutes before the addition of cisplatin followed by a continuous postoperative infusion at 2.13 g/m2 per hour for 12 hours.14 Toxicity was assessed using the NCI Common Toxicity Criteria (version 2.0). Patients were evaluated 6 weeks postoperatively and then every 3 months for 1 year, every 4 months for 1 year, and every 6 months thereafter with laboratory tests, physical examination, and computed tomography scan of the chest, abdomen, and pelvis to assess for ascites or soft tissue masses indicative of tumor recurrence. Patients were considered to have stable disease until they had radiographic evidence of progression. Survival of patients was determined from the prospectively maintained database.

Clinical Specimens for Microarray

Gene expression profiles were generated from 41 consecutive, surgically resected, fresh-frozen MPM tumors and 6 nonmalignant peritoneal tissues that were obtained from patients who underwent other surgical procedures; these nonmalignant tissues were used to identify intrinsic, nonmalignant peritoneal gene expression for comparison with the tumor gene expression profile and to identify the genes that were altered predominantly in MPM tumors. Although the surgeon determined that all resected samples were tumor mass, for each patient, a portion of the resected tumor was processed for hematoxylin and eosin staining to determine that tumor cells were the predominant cell type in the resected tumor tissues. Immunohistochemical analyses for known markers of mesothelioma—calretinin and cytokeratin—were routinely performed to confirm the diagnosis.

DNA Microarrays

Total RNA was extracted from both tumor samples and nonmalignant peritoneum samples using TRIZOL reagent (Invitrogen Corp., Carlsbad, Calif) and passed through an RNeasy spin column (Qiagen, Valencia, Calif) for cleaning. We used 20 μg total RNA from samples and human universal reference RNA (BD Biosciences, San Jose) as the reference sample for hybridization; both were reverse transcribed using reverse transcriptase III (Invitrogen Corp.) according to the manufacturer's protocol. All nonmalignant peritoneum and tumor samples were hybridized against the common reference. The generated complementary DNA (cDNA) was labeled fluorescently with Cy5 mono-reactive dye (normal or tumor samples) or Cy3 mono-reactive dye (reference) from Amersham Biosciences (Piscataway, NJ). The human oligonucleotide array set (HS-OperonV2-vW2) for ∼32 K genes produced at the Advanced Technology Center at the NCI (Gaithersburg, Md) was used. Microarray hybridization and washings were performed as described previously.15 We used a dye-swap strategy to avoid any dye-labeling bias. Hybridized arrays were scanned at 10-μm resolution on a GenePix 4000A scanner. The resulting tagged image file (TIF) images were extracted using GenePix Pro 3.0 software (Axon Instruments, Foster City, Calif).

Statistical Analyses

The primary objective of the patient clinical data analysis was to determine the association between the tumor gene expression pattern derived from pretreatment tumor samples and various clinical parameters, including patient survival. Actuarial survival analyses were performed using the standard Kaplan-Meier method and 2-tailed log-rank tests.16, 17

For the microarray data analysis, we performed data transformation and normalization of genes using BRB ArrayTools (the BRB ArryTools 3.6 software package was developed by Dr. Richard Simon of the NCI). Spot filtering was performed and image spots that measured <10 μm in greatest dimension or with signal intensities below background intensity for any of the 2 fluorescent channels were excluded. The fluorescent intensity ratios were log2-transformed and normalized using Lowess smoother. Then, we averaged expression ratios of each gene when it was represented more than once in a microarray platform.18 To identify genes that had meaningful differences in expression data, we excluded any gene that had missing expression data in >40% of the datasets and genes based on variation less than the 75th percentile of their log-expression values compared with reference values across all arrays. We applied hierarchical cluster analysis (average linkage) in which gene clustering and tumor clustering were performed independently with or without normal peritoneal samples to study the segregation of samples based on gene expression data. A 3-dimensional plot was generated to evaluate the distance in experimental samples. To select gene signatures that were expressed differentially in 2 given groups, we used Student t tests with a significance level of P < .01.

Cox proportional hazards regression model analysis was also used to identify individual genes and gene sets for which expression was associated with overall patient survival (Wald test).19 To select genes that were expressed differentially between men and women, we used a significance analysis of microarray (SAM) as a method for 2-sample Student t tests.20

Quantitative Real-Time Polymerase Chain Reaction in Tumors

Two highly expressed genes of the PI3K and mTOR signaling pathways, phosphoinositide-3-kinase, catalytic, α polypeptide (PIK3CA) and rapamycin-insensitive companion of mTOR (RICTOR), that were identified in MPM tumors were validated by quantitative real-time polymerase chain reaction (PCR) to validate their differential expression in microarray Classes A and B. One microgram of total RNA was reverse transcribed in 20 μL of reaction volume using reverse transcriptase III (Invitrogen Corp.) according to manufacturer's protocol. Quantitative real-time PCR was performed with the 7500 Real-Time PCR System (Applied Biosystems, Foster City, Calif) in a 96-well reaction plate. The reaction was performed in a 25 μL PCR reaction mix containing 1 μL of the cDNA template and appropriate concentrations of forward and reverse primers and probe (Applied Biosystems). We used previously prepared probe and primer sets for PIK3CA (Hs.00907966_m1) and RICTOR (Hs.00380903_m1). The forward and reverse primers and the TaqMan probe for β-actin that was used for normalization of data were as follows: forward primer, 5′-GCGAGAAGATGACCCAGA-3′; reverse primer, 5′-CCAGTGGTACGGCCAGAG-3′; and TaqMan probe, 5′-FAM-CCAGCCATGTACGTTGCTA-3′. Relative gene expression was calculated and expressed as the mean ± standard error of the mean.

Cell Culture and Cell Proliferation Assay

We used the MPM tumor cell lines Meso-1 and Meso-2, which were derived from patients who underwent cytoreduction procedures at our institution. MPM cells isolated from patient tumors were cultured in RPMI-1640 media with 10% fetal bovine serum (FBS) for several generations, and stable immortal cell lines were generated. Cell lines were confirmed as MPM based on characteristic morphology by transmission electron microscopy and cellular expression of the known markers for MPM, calretinin and cytokeratin, by immunohistochemistry. We reconstituted the dual PI3K-mTOR inhibitor NVP-BEZ235-AN (Novartis Institutes for Biomedical Research, East Hanover, NJ) in dimethyl sulfoxide (DMSO) for experimental studies to test the therapeutic significance of the PI3K and mTOR signaling pathways. Cell proliferation assays were performed as described previously.21 Briefly, Meso-1 and Meso-2 cells were plated on flat-bottomed, 96-well plates (Corning, Inc., Corning, NY) at a density of 1500 per well and were allowed to grow overnight in complete RPMI-1640 medium with 10% FBS. Fresh medium with DMSO or medium that contained BEZ235 (250 nm/L) was added the day after the tumor cells were plated and incubated. We identified the optimal concentration of BEZ235 to inhibit cell proliferation in our dose-escalation studies. Tumor cell proliferation was assayed after 3 hours (Day 0) and once daily for 4 days by WST-1 assay (Roche Diagnostics Corp., Indianapolis, Ind). Cell proliferation was quantified by measuring the absorbance at 450 nm with a reference wavelength of 650 nm on a Bio-Rad microplate reader (Bio-Rad Laboratories, Hercules, Calif). The results were analyzed with an analysis of variance and are reported as the mean ± standard error of the mean.

Western Blot Analysis

Meso-1 and Meso-2 cells were plated (10,000 cell per well) in 6-well plates, the cells were treated with either DMSO (control) or BEZ235 250 nm/L for 2 days, and total protein was extracted from both control and treated cells using CelLytic (Sigma Chemical Company, St. Louis, Mo) with protease inhibitors. Total protein concentration was determined in the supernatant by using bicinchoninic acid protein assay (Pierce Biotechnology, Inc., Rockford, Ill). Cell lysate (30-50 μg of total protein) was loaded in a 10% Bis-Tris NuPAGE gel (Invitrogen Corp.) then transferred onto a 0.25-μm nitrocellulose membrane. The membranes were hybridized with the following primary antibodies from Cell Signaling Technology (Beverly, Mass); PI3K-p110α, phosphorylated protein 85 (phospho-p85), v-akt murine thymoma viral oncogene homolog (AKT), phospho-AKT, mTOR, RICTOR, and β-actin (Sigma-Aldrich Corp., St. Louis, Mo) in 5% nonfat milk. Rabbit horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology) was used in 3% nonfat milk. Protein-antibody complexes were detected by chemiluminescence with the SuperSignal West Dura Extended Duration Substrate (Pierce Biotechnology, Rockford, Ill), and images were captured using Kodak BioMax film (Eastman-Kodak, New York, NY).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Clinical Patient Profiles and Survival

The demographic, disease-related, and treatment parameters for 41 patients with MPM are listed in Table 1. The median age of the patient cohort was 51.4 years, and there was a preponderance of men. In this cohort, only 2 patients had received previous systemic chemotherapy. Seven patients had undergone a previous cytoreduction procedure, including 5 patients who also had received HIPEC. All patients in this study underwent laparotomy with the intent to resect all gross disease, including visceral organ resection and peritonectomy, if necessary, followed by 90-minute HIPEC, as described above (see Materials and Methods). The median actuarial overall survival for the entire cohort was 47.4 months (Fig. 1).

Table 1. Clinical and Treatment Characteristics of Patients With Malignant Peritoneal Mesothelioma
CharacteristicValue
  • a

    Five of these 7 patients also received hyperthermic intraoperative intraperitoneal perfusion with chemotherapy.

Total41
Sex: No. of Patients (%) 
 Women14 (34)
 Men27 (67)
Age, y 
 Range17.2-72.8
 Mean49
 Median51.4
Prior therapy 
 Systemic chemotherapy2
 Diagnostic procedure only23
 Cytoreductiona7
Operative time, h 
 Range3-11.3
 Mean7.24
 Median7.33
Residual disease (RD) status: No. of patients (%) 
 RD 0: Complete debulking8 (22)
 RD 1: <100 Lesions all measuring <5 mm6 (17)
 RD 2: >100 Lesions or any measuring >5 mm13 (36)
 RD 3: Any lesion measuring >1 cm7 (19)
 RD 4: Biopsy only2 (6)
 Missing5 (12)
Follow-up, mo 
 Range14-165
 Mean57.6
 Median50.5
thumbnail image

Figure 1. The actuarial overall survival of patients with malignant peritoneal mesothelioma is illustrated. The median overall survival was 47.4 months, and the probability of dying was 20% at 1 year and 37% at 2 years.

Download figure to PowerPoint

Clustering of Tumor Samples and Analysis of Cell Pathways

We generated a gene expression dataset of 41 individual patient tumors in which 63% were epithelioid, 32% were tubulopapillary, and 5% were the adenomatoid histologic type. First, we performed an unsupervised hierarchical cluster analysis of all patient tumors to determine whether an unsupervised gene expression analysis could be used to classify groups of patients based on molecular characteristics of tumors before treatment. Average linkage-agglomerative hierarchical clustering and 3-dimensional scaling of tumor samples revealed 2 groups of approximately equal size that had distinct gene expression patterns among the 41 MPM tumors that we analyzed (Fig. 2A). Members of the 2 clusters were separated easily when viewed in a multidimensional viewer, indicating that the 2 clusters were characterized by their overall similarity in gene expression patterns. In a separate analysis, we compared nonmalignant peritoneum tissues with MPM tumor tissues; the major proportion of genes in the nonmalignant tissues revealed the down-regulation of genes, and these genes formed a distinct cluster compared with gene expression in the tumor samples. For quality control of the microarray results, we performed dye-swap experiments and observed high correlation and reproducibility with the experiments described.

thumbnail image

Figure 2. An unsupervised hierarchical cluster analysis of gene expression data from 41 malignant peritoneal mesothelioma (MPM) tumors is shown. (A) This dendrogram of a 2-way cluster analysis of 4748 genes extracted from MPM tumors illustrates the 2 distinct classes of patients with significantly different survival. Data are presented in a matrix format in which the rows represent the individual gene, and the columns represent the tumor from each patient. Red represents high gene expression, and green represents low gene expression, as indicated in the scale bar (log2-transformed scale). (B) This Kaplan-Meier plot illustrates the actuarial overall survival of patients with MPM in Groups A and B as microarray classes clustered on the basis of gene expression similarities (P = .035). (C) Real-time polymerase chain reaction of phosphoinositide-3-kinase, catalytic, α polypeptide (PIK3CA) and rapamycin-insensitive companion of mammalian target of rapamycin (RICTOR) expression in tumors from patients who had poor survival (Group A) versus good survival (Group B) was based on microarray classes; relative messenger RNA (mRNA) levels are reported as the mean ± standard error of the mean. Asterisks indicate P < .001.

Download figure to PowerPoint

Having identified 2 distinct groups of patients with MPM, next, we analyzed the clinical relevance of these findings by comparing survival and other clinical parameters between the patients in each cluster. Kaplan-Meier actuarial survival curves revealed that patients in Group B had a significantly better prognosis compared with patients in Group A (P = .035) (Fig. 2B). Group B patients had an estimated 80% 3-year survival compared with a 38% 3-year survival for patients in Group A. The average age distribution and sex distribution between Groups A and B were approximately equal.

To obtain meaningful data from the 4748 differentially expressed genes that were clustered in Group A (poor survival) and Group B (good survival), we performed a pathway comparison analysis using the gene set analysis tools of BRB ArrayTools. The 2 significantly overexpressed pathways that were identified in patients who had a poor prognosis were associated with PI3K and mTOR signaling (P < .01) (Table 2). We observed that PIK3CA was the most highly significant (P < .001) gene overexpressed in Group A associated with a poor prognosis, and the closely interacting gene RICTOR also was associated highly significantly with a poor outcome in patients with MPM (P < .001). To confirm the overexpression of PIK3CA and RICTOR, we performed a real-time quantitative reverse transcriptase-PCR, which confirmed that these genes were overexpressed significantly (P < .001) in patients who had poor survival (Group A) compared with patients who had good outcomes (Group B).

Table 2. Differential Tumor Expression of the Phosphoinositide-3-Kinase and Mammalian Target of Rapamycin Signaling Pathways in Patients With Malignant Peritoneal Mesothelioma Who Had Poor Survival (Group A) Versus Good Survival (Group B)a
  Gene Expression Ratio  
Gene SymbolUG ClusterGroup AGroup BPDescription
  • mTOR indicates mammalian target of rapamycin; UG, UniGene (database sponsored by the National Center for Biotechnology Information; http://www.ncbi.nlm.nih.gov/unigene); Hs., homo sapiens; kD, kilodalton.

  • a

    P < .01.

Phosphatidylinositol signaling
 PIK3CAHs.5534984.4512521.205973<1e-07Phosphoinositide-3-kinase, catalytic, α polypeptide
 DGKZHs.5024610.3006710.72942<1e-07Diacylglycerol kinase, zeta, transcript variant 3
 PIP5K1CHs.2821770.315050.6197732.40E-06Phosphatidylinositol-4-phosphate 5-kinase, type I, gamma
 EIF2AK3Hs.5915893.9310652.0337719.00E-06Eukaryotic translation initiation factor 2-alpha kinase 3
 PIK3CGHs.329422.0370870.9373797.69E-05Phosphoinositide-3-kinase, catalytic, gamma polypeptide
 PLCG2Hs.4131111.8819481.125888.00031Phospholipase C, gamma 2
 DGKHHs.6594372.4279181.493399.000418Diacylglycerol kinase, eta
 PLCB1Hs.4311732.3700351.353466.000633Phospholipase C, beta 1, transcript variant 1
 PIK3C2BHs.4974870.6393721.033808.001167Phosphoinositide-3-kinase, class 2, beta polypeptide
 PLCE1Hs.6550332.1701791.258424.003016Phospholipase C, epsilon 1
 PRKCAHs.5317040.802721.486625.003439Protein kinase C, alpha
 PTENHs.5004663.0190812.345187.038219Phosphatase and tensin homolog
 CALM2Hs.6434831.1320631.426249.091226Calmodulin 2 (phosphorylase kinase, delta)
 CALM1Hs.2824100.7047420.820932.311983Calmodulin 1 (phosphorylase kinase, delta)
 SKIPHs.4363060.8215760.94904.456471Sphingosine kinase 1 interacting protein
 INPP5DHs.6019111.294631.440615.486488Inositol polyphosphate-5-phosphatase
 ITPR3Hs.657581.6550381.681271.923294Inositol 1,4,5-triphosphate receptor, type 3
mTOR signaling
 PIK3CAHs.5534984.4512521.205973<1e-07Phosphoinositide-3-kinase, catalytic, alpha polypeptide
 RPS6KB1Hs.4636423.5302651.356929<1e-07Ribosomal protein S6 kinase, 70-kD, polypeptide 1
 RPS6KA3Hs.4453874.267851.8597964.00E-07Ribosomal protein S6 kinase, 90-kD, polypeptide 3
 PGFHs.2528200.64351.2187071.80E-06Placental growth factor, vascular endothelial growth factor-related protein
 PIK3CGHs.329422.0370870.9373797.69E-05Phosphoinositide-3-kinase, catalytic, gamma polypeptide
 RICTORHs.4079262.9826611.859029.000282Rapamycin-insensitive companion of mTOR
 IGF1Hs.1605623.0644381.622047.000853Insulin-like growth factor 1
 ULK2Hs.1687621.1611032.081621.007347Unc-51-like kinase 2 (Caenorhabditis elegans)
 VEGFAHs.737931.352952.374731.007487Vascular endothelial growth factor A
 RHEBHs.6470681.5947511.187123.018201Ras homolog enriched in brain
 MAPK3Hs.8610.7094781.100081.018803Mitogen-activated protein kinase 3, transcript variant 1
 AKT3Hs.4982920.9981881.475271.025163v-akt Murine thymoma viral oncogene homolog 3, transcript variant 1
 CAB39Hs.6325361.6284452.043519.039796Calcium-binding protein 39
 INSHs.6545790.5853220.722639.073382Insulin
 DDIT4Hs.5230120.7443931.046804.075985DNA-damage-inducible transcript 4
 RPS6KB2Hs.5343450.130770.111368.207598Ribosomal protein S6 kinase, 70-kD, polypeptide 2
 RPS6Hs.4080732.2459121.959056.298606Ribosomal protein S6
 HIF1AHs.6546003.9929554.248296.760635Hypoxia-inducible factor 1, α subunit, transcript variant 2

Next, we analyzed various patient or tumor parameters that may have explained differences between the microarray groups. In Group A (poor survival), epithelioid and tubulopapillary histologic types were distributed nearly equally; and, in Group B (good survival), 80% of the samples were the epithelioid type, and 20% were the tubulopapillary histologic type. To identify any differences in gene expression between these histologic subtypes, we performed a class-comparison t test between the epithelioid and tubulopapillary types. Surprisingly, we observed that only 7 genes—ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 1 (ST8SIA1); proline-rich 13 (PRR13); cystatin B (CSTB); serpin peptidase inhibitor, clade B (ovalbumin), member 1 (SERPINB1); chromosome 7 open reading frame 30 (C7orf30); C2orf59; and the predicted S100 calcium-binding protein A14 isoform 1 (LOC730278)—had significantly different gene expression in epithelioid tumors compared with tubulopapillary tumors (P < .001). Next, we compared the gene expression data from men versus women to test whether there were any differences in gene expression based on sex. Only 5 genes—aspartic acid (D)-glutamic acid (E)-alanine (A)-aspartic acid (D) (DEAD) box polypeptide 3, Y-linked (DDX3Y); eukaryotic translation initiation factor 1A, Y-linked (EIF1AY); ubiquitin-specific peptidase 9, Y-linked (USP9Y); mucin 16, cell surface associated (MUC16); and X (inactive)-specific transcript antisense RNA (nonprotein coding) (TSIX) (mostly genes associated with the sex chromosomes)—were expressed differentially between men and women (P < .001). These findings indicate that the biology of disease progression is independent of tumor histology and patient sex and that pretreatment gene expression profiles are highly associated with survival.

The ability to achieve a complete or near complete cytoreduction (residual disease [RD] status ≤2) is a known favorable prognostic factor associated with outcome. We assessed the correlation of that parameter with microarray classes and observed that the number of patients who achieved complete or near complete cytoreduction (RD status ≤2) were not different between microarray Classes A and B (P2 > .3; Fisher exact test).

To identify pathways that were associated most strongly with survival irrespective of microarray class in patients with MPM, we applied a supervised gene set analysis. The Cox proportional hazards model, along with a permutation test (which identifies gene sets that have more genes expressed differentially with survival times than expected by chance) were used to identify the gene sets associated most significantly with patient survival (P < .01). We identified 4 gene sets that were associated significantly with survival: cell cycle regulation, regulation of actin cytoskeleton, the apoptosis pathway, and the proteasome pathway. Expression of known cell cycle markers, such as the cell division cycle (CDC) markers CDC2 and CDC7 and the cyclin family genes, such as cyclin-A2 (CCNA2), cyclin-B1 (CCNB1), and cyclin-B2 (CCNB2) were associated highly significantly with poor patient outcome in a Kyoto Encyclopedia of Genes and Genomes pathway gene set analysis (P < .01).

Inhibition of the PI3K and mTOR Signaling Pathways in MPM

Our gene expression studies indicated that PI3K and mTOR signaling are associated significantly with aggressive MPM growth and shortened patient survival. We reasoned that inhibition of these pathways might have antitumor activity by suppressing tumor cell proliferation. To investigate this possibility, we generated actively proliferating MPM cell lines, Meso-1 and Meso-2, from 2 patient tumor samples. These cells were treated with NVP-BEZ235, a dual-class PI3K and mTOR inhibitor. Cells that were treated with 250 nm BEZ235 revealed the marked suppression of PI3K and mTOR signaling and resulted in significant inhibition of cell proliferation (Fig. 3A,B). In both cell lines, BEZ235 inhibited PI3K-p110α, mTOR, and RICTOR protein expression and phosphorylation of PI3K-p85 and AKT (Fig. 3C), indicating that this agent is an effective inhibitor of PI3K and mTOR signaling in MPM.

thumbnail image

Figure 3. The inhibition of phosphoinositide-3-kinase (PI3K) and mammalian target of rapamycin (mTOR) signaling in malignant peritoneal mesothelioma (MPM) cells was demonstrated by using the dual-class inhibitor BEZ235. (A) Photomicrographs and (B) a WST-1 assay indicate that BEZ235 significantly (P < .001) inhibited MPM cell proliferation. (C) Western blot analysis of proteins of the PI3K and mTOR signaling pathways demonstrated decreases in PI3K isoform p110α, phosphorylated protein 85 (phospho-p85), phosphorylated v-akt murine thymoma viral oncogene homolog (phospho-AKT), mTOR, and rapamycin-insensitive companion of mTOR (RICTOR) in total lysates from the mesothelioma cell lines Meso-1 and Meso-2 when the cells were treated with BEZ235 for 2 days at a concentration of 250 nm/L.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

In the current cohort of patients with MPM who were analyzed in this study, the use of cytoreduction and HIPEC resulted in an actuarial overall median survival of 47.4 months. This outcome is similar to that reported previously by us and others6, 7, 11; therefore, we believe that the findings in this study are relevant to the general population of patients with MPM. The actuarial overall median survival of 47.4 months, coupled with a 20% probability of death within 1 year and a 37% probability of death within 2 years, demonstrated that many patients died of early disease progression, whereas others survived for many years. The heterogeneity in the clinical behavior of MPM has been recognized for many years, but the molecular basis for this phenomenon remains poorly understood.

Analysis of the gene expression patterns of patient tumors provides important insights into tumor biology and may identify novel targets for therapy. Most gene expression studies of mesothelioma have been performed in the pleural variant to understand the molecular nature of disease progression and to identify prognostic factors for this condition.22-24 The major emphasis of the current study was to determine whether molecular profiling of tumor before any therapeutic intervention can predict biologic behavior and identify pathways in MPM that may be targets for therapeutic intervention. We performed a global gene expression analysis in tumors from 41 patients with MPM to identify any association between gene expression patterns and pretreatment clinical or pathologic parameters, such as age, sex, and tumor histology. We observed that clustering of patient tumors based on similarities in gene expression resulted in 2 classes or groups of patients with distinct molecular profiles that had significantly different survivals. Although the patients were separated into 2 distinct groups based on their tumor gene expression patterns, the MPM tumor samples shared marked overall similarity in gene expression compared with nonmalignant peritoneal tissue samples. The nonmalignant peritoneum formed a distinct cluster, and most genes that were expressed in nonmalignant peritoneum primarily were down-regulated, reflecting its largely quiescent biologic activity. It has been reported that patients who have MPM with epithelioid histology have significantly prolonged disease-free survival compared with patients who have MPM with nonepithelioid histologies,6, 11 although this difference has not been observed consistently by others.7, 12 We observed no statistically significant relation between tumor histology and microarray classes. Other statistical analyses revealed no significant relation between sex and age compared with the microarray classes, indicating that the biology of disease progression was independent of these parameters. We also observed no relation between microarray classes and the ability to perform a complete or near complete cytoreduction, indicating that, with results from the available data, microarray profiling cannot be used to identify which patients should undergo cytoreduction.

Cell pathway analyses derived from gene expression studies have been used to understand important pathways associated with distinct tumor biology and patient survival. In the current study, we identified 2 important and closely interacting pathways, PI3K and mTOR signaling, that are overexpressed in MPM tumors and are associated with significantly poor patient survival; these pathways may represent targets for novel therapeutic intervention. In our study, there was markedly higher expression of the genes of these pathways among the patients in Group A (poor survival) than among the patients in Group B (good survival). A significant role in tumor progression has been identified for the PI3K and mTOR signaling pathways in several solid organ tumors.25-28 It is known that PIK3CA, the gene that encodes PI3K-P110α, regulates cell proliferation and angiogenesis29 and has a highly significant association with a poor prognosis in patients with MPM. This and other studies indicate that the p110α regulatory subunit of PIK3CA is essential for vascular development and growth factor signaling.29, 30 In addition, this gene is mutated or amplified in several cancers, suggesting the importance of PIK3CA in tumorigenesis and cancer progression.31, 32 Another isoform, PIK3CG, also was overexpressed in patients with MPM who had poor survival; and, together, the activation of these pathways may lead to an aggressive tumor phenotype by activating several downstream effectors, such as AKT. PI3K and AKT are important for mTOR signaling with the downstream activation of a variety cell survival, antiapoptotic, and angiogenic factors, such as B-cell lymphoma 2 (BCL2), cyclin-D1, hypoxia-inducible factor 1α (HIF1α), and vascular endothelial factor A (VEGFA), all of which are overexpressed in MPM tumors. RICTOR is 1 of the central genes of the mTOR complex that is highly activated in MPM and is associated with a poor outcome. Recent studies have demonstrated that RICTOR results in the phosphorylation of AKT and is relatively insensitive to rapamycin but sensitive to PI3K inhibitors.33, 34 Therefore, RICTOR may mediate several functions that are assigned to PI3K. Moreover, RICTOR is a direct target of the ribosomal protein S6 kinase-1 (S6K1), which regulates cell size and growth.35

Tumors from patients with shortened survival express several genes related to cell cycle regulation, actin cytoskeleton regulation, and antiapoptosis. In addition, these patients display higher expression of genes of the proteasome pathway, which is involved in cell cycle regulation, growth, and apoptosis and is known to promote cancer progression. Borczuk et al also demonstrated that the ubiquitin-proteasome pathway is associated significantly with the MPM tumors that have biphasic histology, a rare phenotype of MPM with a particularly virulent biology.13 Consistent with that report, we demonstrated that the proteasome pathway is up-regulated in MPM tumors and is associated with poor patient survival.

Differential expression of the genes PIK3CA and RICTOR was confirmed further by real-time PCR and produced results that were consistent with the microarray data. The PI3K-AKT-mTOR signaling pathway is central to the malignant phenotype of most cancer cells. Targeted therapies against the PI3K and mTOR signaling pathways identified in MPM tumors should be efficacious, because these pathways interact closely for cellular functions. It has been demonstrated that BEZ235, a dual-class PI3K and mTOR inhibitor, significantly reduces growth in human xenograft models of lung cancer and glioma.36, 37 In our study, treatment of the MPM cell lines Meso-1 and Meso-2 with BEZ235 significantly reduced cell proliferation, indicating that BEZ235 may be an effective inhibitor of MPM tumor progression. Protein analyses have indicated that BEZ235 results in the global inhibition of downstream PI3K and mTOR signaling. However, further studies will be essential to identify the antitumor activities of BEZ235, particularly in MPM xenograft models.

In summary, the biologic behavior of MPM is associated with distinct gene expression patterns, and overexpression of the closely interacting PI3K and mTOR pathways appears to be essential in regulating the malignant phenotype of the tumor. These results indicate that the survival of patients with MPM is associated strongly with distinct tumor gene expression profiles before therapy. Ideally, the findings described in this report should be confirmed in an independent population of patients. The poor prognostic genes and pathways identified in this cancer may represent targets for novel therapeutic intervention.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

We thank Dr. Yue Zhu, Hui Xu, and Ethan Hagan for technical assistance.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES

Funding support was received from the Marlene and Stewart Greenebaum Cancer Center, University of Maryland School of Medicine, Baltimore, Maryland, and the Center for Cancer Research, National Cancer Institute, Bethesda, Maryland.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. CONFLICT OF INTEREST DISCLOSURES
  8. REFERENCES