Patent Application
20130288250
The present invention refers to a method for in vitro predicting survival and/or metastatic outcome of gastrointestinal stromal tumors (GISTs), a kit for (i) the in vitro prediction of the survival outcome of a patient suffering from GIST, (ii) and/or the development of metastases in a patient treated for or suffering from GIST, (ii) and/or the prediction of the efficacy of a treatment for GIST. The present invention also refers to a method for screening for compounds for the use in the treatment of GISTs, and to a compound for its use in the treatment of GISTs.
Therefore, the present invention has utility in the medical and pharmaceutical fields, especially in the field of diagnosis.
In the description below, the numeral reference in brackets (“number”) refers to the respective listing of references situated at the end of the text.
Gastrointestinal stromal tumors (GISTs) are the most frequent mesenchymal tumors of the gastrointestinal tract and account for approximately 25% of soft tissue sarcomas. GISTs are thought to arise from the intestinal cells of Cajal (1), or from a common progenitor cell (2).
The KIT tyrosine kinase or the platelet-derived factor receptor a (PDGFRA) activating mutations are early oncogenic events in GISTs. Most GISTs (80%) are characterized by activating mutations of the KIT tyrosine kinase receptor, while a subset (8%) harbours platelet-derived factor receptor α (PDGFRA mutations (3,4). In addition to these mutations, other genetic changes do occur, the most frequent alterations reported being 14q, 22q and 1p deletions (5). Overall, GIST cytogenetics is quite simple and imbalances mainly involve full chromosomes or chromosome arms. Notably, GIST molecular and cytogenetic profiles correlate with disease progression. Nevertheless, it has been observed that changes are more frequent and more complex in advanced tumors (6). Furthermore, the genetic basis of the metastatic outcome of GISTs is still poorly understood.
Clinical management of GISTs consists mainly of surgical resection with adjuvant or neo-adjuvant targeted therapy with Imatinib Mesylate (Gleevec, formerly STI571, Novartis Pharma AG) which has been demonstrated to target the KIT- or PDGFRA-aberrant signaling induced by activating mutations (7). The majority of cases can be cured by surgical resection alone, but 20-40% of patients relapse with distant liver metastasis being the most common manifestation of the recurrent disease.
Many pathological criteria based on tumor site, size, cell type, degree of necrosis and mitotic rate have been proposed for predicting the outcome of patients with GISTs. A consensus was found by the National Institute of Health (NIH) in 2001 to estimate the relative risk of GISTs based on tumor size and mitotic count (8) and in 2006, the Armed Forces Institute of Pathology (AFIP) proposed an updated system taking into account also a tumor location (9). Even if these two systems are particularly efficient in determining the metastatic risk of GISTs, they are based on an indirect histopathological reflection of tumor aggressiveness. Moreover, the cutoff values for these criteria have been determined empirically leading to subjectivity that is inevitable in skilled pathologists' assessments. Hence, there is a need to more deeply understand the biology underlining the aggressiveness of GISTs in order to identify objective biomarkers that enhance the specificity and the reproducibility of outcome prediction.
The development of a valid and reliable, investigator-independent method of GIST prognostication is essential for the proper clinical management of GIST patients, especially in the context of adjuvant treatment, where many patients are exposed to imatinib while only a small proportion will likely benefit from such treatment (19).
To achieve this purpose, genomic and expression profiling has already been used but only partial and heterogeneous results have been reported. At the genomic level, it has been shown that the genome complexity level increases with tumor stage (6, 10), but no threshold has ever been defined and no specific alteration has been proposed except for p16INK4A alterations whose role in metastasis development is still controversial (11-16). At the expression level, Yamaguchi and colleagues have proposed a gene-expression signature: they identified CD26 as a prognostic marker but only in GISTs of gastric origin. Nevertheless, the authors concluded that CD26 might not be the cause of malignant progression of gastric GISTs. Moreover, this signature is limited as it has been established on only a few cases (32 GISTs), it predicts outcome only in gastric GISTs (but not in GIST of the small intestine) and it has not been compared to histopathological grading methods considered as “gold standard” (17).
In the few genome or expression profiling analyses using smaller numbers of GISTs that have been conducted before this study, only one (35) presents an integrative analysis gathering genome and transcription profiling as we present here. This study was based on 25 patients and aimed to identify target genes located in altered regions described within the last 15 years. Essentially, many studies have described GISTs genome and demonstrated that GISTs cytogenetics is quite simple, reflected by only few aberrations, deletions being more frequent than gains (6, 10, 11, 12, 36, 37). All these studies concluded that chromosome 14, 22 and 1p deletions are the most frequent aberrations. It has also been shown that changes are more frequent in high-risk and overly malignant GISTs than in low/intermediate-risk GISTs (6, 10), but a strong association between one alteration and prognosis has not yet been identified, except CDKN2A alterations as discussed above. At the expression level, most of the studies have been set up to enhance delineation of diagnosis (38, 39) or to identify expression differences according to KIT or PDGFRA mutation status (40-42).
The AURKA, encoded for a gene that maps to chromosome 20q13, is a mitotic centrosomal protein kinase (20). It is a well known oncogene, which main role in tumor development is the control of chromosome segregation during mitosis (21). Gene amplification and AURKA overexpression have been widely described in many cancer types (22). In particular, as it has been clearly demonstrated that AURKA overexpression induces centrosome duplication-distribution abnormalities and aneuploidy leading to transformation in breast cancer cells (23). Actually, centrosomes maintain genomic stability through the establishment of the bipolar spindle body during cell division, ensuring equal segregation of replicated chromosomes to two daughter cells. The AURKA expression has also been associated with poor prognosis mainly in breast carcinoma (24), colon carcinoma (25, 26), neuroblastoma (27) and head and neck squamous cell carcinoma (28). Taken together, these data indicate that up-regulation of AURKA expression could be a major driving event in establishing genome complexity leading to wild gene expression reprogramming, creating optimal conditions for development of metastasis. AURKA inhibitors are currently under clinical studies (29-33).
In view of all these elements, it clearly still exists a need of new tools allowing to predict outcome of GISTs, notably palliating the failures, drawbacks and obstacles of the state of the art.
In some aspects, the present invention is directed to a method for in vitro predicting survival and/or metastatic outcome of gastrointestinal stromal tumors (GISTs), characterized in that it comprises the measure of the level, in a patient-derived biological sample of GIST, of a pool of polypeptides or polynucleotides consisting in Aurora kinase A (AURKA).
In some aspects, said measure of the level of the pool of polypeptides is a measure of the expression level of a pool of polynucleotides consisting in AURKA.
In some aspects, GIST is classified in a group with high risk to develop metastases within 5 years, i.e. with a risk to develop metastases within 5 years of more than 80%, when AURKA is up-regulated compared to a group with no risk to develop metastases within 5 years when AURKA is down-regulated.
In some aspects, the calculation of the Genomic Index (GI), i.e. the number and type of alterations of the GIST genome, according to the formula as follows:
GI=A2×C,
wherein A is the number of alterations in GIST genome and C is the number of involved chromosomes in GIST.
In some aspects, GIST is classified in a group of metastasis- and disease-free survival group when AURKA is down-regulated and the GI is equal or less than 10. In some aspects, AURKA expression is less than 9.13.
In some aspects, GIST is classified in a group with low risk to develop metastases within 5 years, i.e. with a risk to develop metastases within 5 years equal to 0%, when AURKA expression is equal or less than the mean of AURKA expression and GI is equal or less than 10, said mean being the mean of AURKA expression in several GISTs.
In some aspects, GIST is classified in a group with high risk to develop metastases within 5 years, i.e. with a risk to develop metastases within 5 years more than 75%, when AURKA expression is more than the mean of AURKA expression and GI is more than 10, said mean being the mean of AURKA expression in several GISTs.
In some aspects, the present invention is directed at a kit for the in vitro prediction of the survival outcome of a patient suffering from GIST, and/or the development of metastases in a patient treated for or suffering from GIST, and/or the prediction of the efficacy of a treatment for GIST, characterized in that it comprises means for detecting and/or quantify, in a sample, AURKA expression or level, and means for the calculation of the GI.
In some aspects, the present invention is directed at a method for screening for compounds for the use in the treatment of GISTs, characterized in that it comprises the steps of contacting a test compound with a patient-derived biological sample containing GISTs cells, measuring the expression or the level of AURKA, comparing said expression or level of AURKA with the expression of AURKA before the contact between said test compound and said sample, and selecting said test compound that allows a down-regulation of the expression of AURKA.
In some aspects, the method comprises calculating GI, comparing said GI with the GI before the contact between said test compound and said sample, and selecting said test compound allowing a down-regulation of the GI to 10 or less.
In some aspects, the present invention is directed at an AURKA inhibitor for its use in the treatment of GISTs. In some aspects, the AURKA inhibitor is selected among PHA-739358, MLN8237 and MK-5108.
After important researches, the Applicant found surprisingly a one gene-expression signature prognostic for clinical outcome of primary GISTs.
Surprisingly, it has been demonstrated by the Applicant that the CINSARC signature (CINSARC for Complexity INdex in SARComa, a 67 genes-expression prognostic signature related to genome complexity in sarcomas, PCT/FR2010/000323, [55]) and/or a new one-gene-expression signature predict metastatic outcome in GIST and that the combination of each of these signatures with genome imbalances outperforms current histopathological grading method in determining patient prognosis. More specifically, these molecular signatures identify “at-risk patients” within cases stratified as intermediate-risk according to the Armed Forces Institut of Pathology classification.
The Applicant manages to show that a positive correlation exists between GI (Genomic Index) and AURKA expression.
The Applicant surprisingly manages to construct a decisional algorithm based on GI and AURKA expression.
Advantageously, application of the signature permits more selective imatinib therapy leading to decreased iatrogenic morbidity and improved outcomes for individual patients.
Accordingly, in a first aspect, the invention provides for a method for in vitro predicting survival and/or metastatic outcome of gastrointestinal stromal tumors (GISTs), the method comprising the measure of the level, in a patient-derived biological sample of GIST, of a pool of polypeptides or polynucleotides consisting in Aurora kinase A (AURKA).
“Predicting survival and/or metastatic outcome” refers herein to predicting whether a patient has a chance to survive, or a risk to develop metastases following the outcome of a GIST. The survival or development of metastases may be calculated from the date of initial diagnosis to the date of first metastases, relapse, last follow-up or death for patients with diagnosis of metastasis. According to a particular embodiment of the invention, GIST may be classified in a group with high risk to develop metastases within 5 years of an outcome of GIST, or in a group with no risk to develop metastases within 5 years, or in an intermediate group. More particularly, the group of patient with high risk to develop metastases within 5 years is characterized by a risk to develop metastases within 5 years of more than 80%, when AURKA is up-regulated, compared to a group with no risk to develop metastases within 5 years when AURKA is down-regulated.
“Patient-derived biological sample of GIST” refers herein to any biological sample containing GIST cells and obtained from a patient treated for or suffering from GIST. For example, GIST may be primary untreated tumors.
“Polypeptide” refers herein to the AURKA protein (Genbank accession number NM—198433; SEQ ID NO: 1), a AURKA protein fragment, a AURKA protein region or a derivative of AURKA protein. For example, the polypeptide may be a polypeptide having at least 70% of sequence identity with the peptidic sequence of AURKA protein, or a polypeptide having at least 80% of sequence identity with the peptidic sequence of AURKA protein, or a polypeptide having at least 90% of sequence identity with the peptidic sequence of AURKA protein.
“Polynucleotide” refers herein to any polynucleotide coding for the polypeptide as defined above, or to any polynucleotide hybridizing under stringent conditions to a polypeptide coding for the polypeptide as defined above. The polynucleotide of the invention may be any of DNA and RNA, for example the sequence SEQ ID NO: 2. The DNA may be in any form of genomic DNA, a genomic DNA library, cDNA or a synthetic DNA. Moreover, the polynucleotide of the present invention may be any of those amplified directly by an RT-PCR method using total RNA or an mRNA fraction prepared from a GIST. The polynucleotide of the present invention includes a polynucleotide that hybridizes under stringent conditions to a polynucleotide.
The measure of the level of polypeptides may be realized by any appropriate technique known by the man skilled in the art. It may be, for example, an immunohistochemistry technique, in which the expression of the protein is measured after hybridization of an antibody recognizing specifically the AURKA protein.
The measure of the level of polynucleotides may be realized be any appropriate technique known by the man skilled in the art. It may be, for example, a method of genomic qPCR (quantitative polymerization chain reaction), CGH-array (Comparative Genomic Hibridization) or RT-qPCR (real time qPCR) in order to check copy number of genomic DNA or quantify expression of genomic DNA.
Advantageously, the AURKA expression allows to predicting survival and/or metastatic outcome of GISTs, and no other gene or protein expression has to be measured.
The method of the invention may comprise the calculation of the Genomic Index (GI), i.e. the number and type of alterations of the GIST genome, according to the formula as follows:
GI=A2×C,
wherein A is the number of alterations in GIST genome and C is the number of involved chromosomes in GIST.
“Number of alterations in GIST genome” refers herein to different numerical and segmental gains and losses. The alterations may for example involve whole chromosome arms or chromosome without rearrangement, or intra-chromosome gains or losses. It may be measured by techniques known in the art, such as CGH-array.
“Number of involved chromosomes in GIST” refers herein to the number of chromosomes of GIST cells having an alteration. The number of chromosome may be measured by CGH-array.
Advantageously, GIST is classified in a group of metastasis- and disease-free survival group when AURKA is down-regulated and the GI is equal or less than 10. In this case, AURKA expression may be less than 9.13, or less than 9, or less than 8, or less than 7, or less than 6, or less than 5. In this case, there is a survival of 5 years, i.e. there are no metastasis or disease during 5 years after GIST outcome or after the end of a treatment. In a particular embodiment, GIST may be classified in a group with low risk to develop metastases within 5 years, i.e. with a risk to develop metastases within 5 years comprises between 0% and 10%, when AURKA expression is equal or less than the mean of AURKA expression, and GI is equal or less than 10, said mean being the mean of AURKA expression/level in several GISTs, for example a series of 50 to 70 GISTs, for example 60 GISTs. In this case, there is no metastasis or disease during 5 years.
Alternatively, GIST may be classified in a group with high risk to develop metastases within 5 years, i.e. with a risk to develop metastases within 5 years more than 75%, when AURKA expression is more than the mean of AURKA expression and GI is more than 10, said mean being the mean of AURKA expression in several GISTs, for example in a series of 50 to 70 GISTs, for example 60 GISTs. In this case, there are 75% of cases of metastasis or disease during 5 years after GIST outcome or after the end of a treatment.
Another object of the invention is a kit for the in vitro prediction of the survival outcome of a patient suffering from GIST, and/or the development of metastases in a patient treated for or suffering from GIST, and/or the prediction of the efficacy of a treatment for GIST, characterized in that it comprises means for detecting and/or quantify, in a sample, AURKA expression/level, and means for the calculation of the GI.
“Means for detecting and/or AURKA expression/level” may be any means for detecting levels of proteins or of polynucleotides known by the man skilled in the art. The means may be for example the means to realize an immunohistochemistry analysis, a western blot or a q-PCR.
“Means for the calculation of the GI” refers herein to any means allowing the calculation of the number of alterations in GIST genome and of the number of involved chromosomes in GIST.
Another object of the invention is a method for screening for compounds for the use in the treatment of GISTs, comprising the steps of:
“Down-regulation” refers herein to any diminution of the expression or level of AURKA protein or polynucleotide.
The method may also comprise the steps of:
Another object of the invention is an AURKA inhibitor for its use in the treatment of GISTs.
“Inhibitor” refers herein to any compound allowing a decrease of the expression/level of AURKA protein, or in a decrease of a biological effect of AURKA, when the inhibitor is contacted with GIST. The AURKA inhibitor may be PHA-739358 (29, 54), MLN8237 (30), MK-5108 (33).
Another object of the invention is a method of treatment of GIST, in a subject in need thereof, comprising the step of administering to the patient a pharmaceutically effective dose of a inhibitor as defined above.
Sixty seven fresh frozen GIST tumors were selected from the European GIST database CONTICAGIST (www.conticagist.org) which contains data from GIST tissues, including information regarding patients, primary tumor, treatment, follow-up and availability of tumor samples. All GISTs selected were primary untreated tumors. Their characteristics are presented in supplementary table 1. Most GISTs (59/67) were studied by both CGH-array and gene expression profiling (a combination of 66 by CGH-array and 60 by gene expression profiling).
DNA Isolation and Array-CGH
Genomic DNA was extracted using the standard phenol-chloroform method. Reference DNA (female), was extracted from a blood sample. The concentration and the quality of DNA were measured using NanoDrop ND-1000 Spectrophotometer and gel electrophoresis. Tumor and control DNAs were hybridized to 8×60K whole-genome Agilent arrays (G4450A). Briefly, for each sample, 350 ng of DNA were fragmented by a double enzymatic digestion (AluI+RsaI) and checked with LabOnChip (2100 Bioanalyzer System, Agilent Technologies) before labeling and hybridization. Tumor and control DNAs were labeled by random priming with CY5-dUTPs and CY3-dUTP, respectively, hybridized at 65° C. for 24 h and rotating at 20 rpm. Arrays were scanned using an Agilent G2585CA DNA Microarray Scanner and image analysis was done using the Feature-Extraction V10.1.1.1 software (Agilent Technologies). Normalization was done using the ranking-mode method available in the Feature-Extraction V10.1.1 software, with default value for any parameter. Raw copy number ratio data were transferred to CGH Analytics v4.0.76 software. The ADM-2 algorithm of CGH Analytics v4.0.76 software (Agilent) was used to identify DNA copy number anomalies at the probe level. A low-level copy number gain was defined as a log 2 ratio>0.25 and a copy number loss was defined as a log 2 ratio<−0.25. A high-level gain or amplification was defined as a log 2 ratio>1.5 and a homozygous deletion is suspected when ration is below −1. To establish decision criteria for prognosis, alterations involving more than 100 probes have been automatically computed using an aberration filter.
Real-Time Genomic Quantitative PCR and Sequencing
To determine the copy number status of p16, p15 and p14, real-time PCR was performed on genomic DNA using TaqMan® Universal Master Mix (Applied Biosystems). For normalizing the results, we used three reference genes: GAPDH, ALB and RPLP0, in order to have, for each tumor, at least two of these reference genes in normal copy number (array-CGH). Primers and probe used for RPLP0 are as follow: (F) 5′-TGGATCTGCTGGTTGTCCAA-3′ (SEQ ID NO: 3); (R) 5′-CCAGTCTTGATCAGCTGCACAT-3′ (SEQ ID NO: 4); (probe) 5′-AGGTGTTTACTGCCCCACTATTATCTGGTTCAGA-3′ (SEQ ID NO: 5). Other primers and probes used were previously described (52). Tumor data were normalized against data obtained for normal DNA. The results were then calculated as previously described (52). A normal status corresponds to 0.8≦ratio≦1.2, 0.1<ratio<0.8 is considered as a hemizygous deletion. When ratio is inferior to 0.1, the deletion is considered as homozygous. CDKN2A locus has been submitted to sequencing as previously described (|52|) and RB1 gene was sequenced using genomic DNA according to Houdayer et al (53) or cDNA with following primers: (F1) 5′-TCATGTCAGAGAGAGAGCTTGG-3′ (SEQ ID NO: 6), (R1) 5′-CGTGCACTCCTGTTCTGACC-3′ (SEQ ID NO: 7); (F2) 5′-AATGGTTCACCTCGAACACC-3′ (SEQ ID NO: 8), (R2) 5′-CTCGGTAATACAAGCGAACTCC-3′ (SEQ ID NO: 9); (F3) 5′-CCTCCACACACTCCAGTTAGG-3′ (SEQ ID NO: 10), (R3) 5′-TGATCAGTTGGTCCTTCTCG-3′ (SEQ ID NO: 11); (F4) 5′-GCATGGCTCTCAGATTCACC-3′ (SEQ ID NO: 12), (R4) 5′-TCGAGGAATGTGAGGTATTGG-3′ (SEQ ID NO: 13); (F5) 5′-TCTTCCTCATGCTGTTCAGG-3′ (SEQ ID NO: 14), (R5) 5′-TGTACACAGTGTCCACCAAGG-3′ (SEQ ID NO: 15).
RNA Isolation and Gene Expression Profiling by One Color Assay
Total RNAs were extracted from frozen tumor samples with TRIzol reagent (Life Technologies, Inc.) and purified using the RNeasy® Min Elute™ Cleanup Kit (Qiagen) according to the manufacturer's procedures. RNA quality was checked on an Agilent 2100 bioanalyzer (Agilent Technologies). RNAs with a RNA Integrity Number (RIN)>6.5 were used for microarray.
Gene expression analysis was carried out using Agilent Whole human 44K Genome Oligo Array (Agilent Technologies). This specific array represents over 41 000 human genes and transcripts, all with public domain annotations. Total RNA (500 ng) was reverse transcribed into cRNA by incorporating a T7 oligo-dT promoter primer prior to the generation of fluorescent cRNA using an Agilent Quick Amp Labeling Kit (Agilent Technologies). The labeled cRNA was purified using a Qiagen RNeasy Mini Kit (Qiagen) and quantified using a NanoDrop ND-1000 instrument. In these experiments Cy3-labeled (sample) cRNAs were hybridized to the array using a Gene Expression Hybridization Kit (Agilent Technologies). The hybridization was incubated in Agilent SureHyb chambers for 17 hours in a Hyb Oven set to 65° C. and rotating at 10 rpm. The microarray slides were washed according to the manufacturer's instructions and then scanned on an Agilent G2565BA DNA Microarray Scanner and image analysis was done using the Feature-Extraction V 10.1.1.1 software (Agilent Technologies).
All microarray data were simultaneously normalized using the Quantile algorithm. The t-test was performed using Genespring (Agilent Technologies) and P-values were adjusted using the Benjamini-Hochberg procedure. The P-value and fold change cut-off for gene selection were 0.001 and 3, respectively. Gene ontology analysis was performed to establish statistical enrichment in GO terms using Genespring (Agilent Technologies).
Real-Time PCR
Reverse transcription and real-time PCR were performed as previously described (52). We used TaqMan® Gene Expression assays (Applied Biosystems): Hs01582072_m1 for AURKA; Hs01078066_m1 for RB1; Hs99999905_m1 for GAPDH; Hs99999903_m1 for ACTB and Hs99999902_m1 for RPLP0. p14 and p16 expression level was assessed as previously described (52). In order to normalize the results, we used GAPDH, ACTB and RPLP0 genes as reference genes. Triplicates were performed for each sample for each gene. A reference CT (threshold cycle) for each sample was defined as the average measured CT of the three reference genes. Relative mRNA level of AURKA in a sample was defined as: ΔCT=CT (gene of interest)−CT (mean of the three reference genes).
Statistical Analysis.
To assign prognosis, we applied the nearest centroid method. Centroids represent a centered mean of expression for the signature genes for each patient outcome (metastatic and non-metastatic). Thus, centroids were calculated from the cohort 1 samples (17) and then each sample of our series (thus considered as a validation set) was allocated to the prognostic class (centroid) with the highest Spearman correlation.
Metastasis- and disease-free survival was calculated by the Kaplan-Meier method from the date of initial diagnosis to the date of first metastasis, relapse, last follow-up or death for patients within diagnosis of metastasis. Survival curves were compared with the log rank test. Hazard ratios were performed with the Cox proportional hazard model. All statistical analyses were performed using R software version 2.11.11 and the package “survival”.
CINSARC is a Significant Prognosis Factor in GISTs
To assess the issue whether our previously published signature could have prognostic value in GISTs, we performed expression profiling in a series of 67 GISTs (Table 1).
Among them, we obtained sufficient mRNA quality for 60 cases (89.5%). We applied the CINSARC nearest-centroid signature (18) to GISTs, using a published series (17) as a training set to retrain centroids and the present series as the validation set. Kaplan-Meier analysis (
Gene Expression Changes Associated with Metastatic Outcome
The results presented above indicate that expression of genes involved in mitosis control and chromosome integrity (CINSARC) is associated with survival outcomes in GISTs. We thus asked whether the reciprocal phenomenon was true, that is whether the differential expression between metastatic and non-metastatic cases can identify such genes. To assess this issue, we performed supervised t-test comparing tumor expression profiles stratified according to outcomes (
Homo sapiens RecQ protein-like 4
Homo sapiens cell division cycle
Homo sapiens aurora kinase A (AURKA),
Homo sapiens non-SMC condensin II
Homo sapiens clone HQ0327 PRO0327
Homo sapiens extra spindle pole bodies
Homo sapiens MCM2 minichromosome
Homo sapiens structural maintenance of
Homo sapiens thyroid hormone receptor
Homo sapiens cell division cycle
Homo sapiens stromal antigen 3 (STAG3),
Homo sapiens pituitary tumor-
Homo sapiens transforming, acidic coiled-
Homo sapiens pituitary tumor-
Homo sapiens TPX2, microtubule-
Homo sapiens NUF2, NDC80 kinetochore
Homo sapiens polymerase (DNA directed),
Homo sapiens centrosomal protein 72 kDa
Homo sapiens chromosome 15 open
Homo sapiens hypothetical protein
Homo sapiens cyclin B1 (CCNB1), mRNA
Homo sapiens cDNA clone
Homo sapiens hypothetical protein
Homo sapiens chromosome 13 open
Homo sapiens cell division cycle
Homo sapiens protein kinase, membrane
Homo sapiens DEP domain containing 1B
Homo sapiens kinesin family member C1
Homo sapiens BUB1 budding uninhibited
Homo sapiens ATPase family, AAA
Homo sapiens ubiquitin-conjugating
Homo sapiens centromere protein A
Homo sapiens centromere protein F,
Homo sapiens cDNA FLJ32129 fis, clone
Homo sapiens family with sequence
Homo sapiens trophinin associated protein
Homo sapiens WD repeat domain 51A
Homo sapiens cDNA, mRNA sequence
Homo sapiens NIMA (never in mitosis
Homo sapiens polo-like kinase 1
Homo sapiens v-myb myeloblastosis viral
Homo sapiens pituitary tumor-
Homo sapiens SHC SH2-domain binding
Homo sapiens meiotic nuclear divisions 1
Homo sapiens baculoviral IAP repeat-
Homo sapiens establishment of cohesion 1
Homo sapiens kinesin family member 11
Homo sapiens chromosome 18 open
Homo sapiens FLJ20105 protein
Homo sapiens family with sequence
Homo sapiens kinesin family member 15
Homo sapiens polo-like kinase 1
Homo sapiens non-SMC condensin I
Homo sapiens cell division cycle
Homo sapiens SPC25, NDC80 kinetochore
Homo sapiens hypothetical protein
Homo sapiens cell division cycle 2, G1 to
Homo sapiens family with sequence
Homo sapiens chromosome 15 open
Homo sapiens asp (abnormal spindle)
Homo sapiens Bloom syndrome (BLM),
Homo sapiens hypothetical protein
Homo sapiens chromosome 15 open
Homo sapiens sorting nexin 5, mRNA
Homo sapiens cytoskeleton associated
Homo sapiens cyclin-dependent kinase
Homo sapiens forkhead box M1
Homo sapiens kinesin family member 11
Homo sapiens kinesin family member 2C
Homo sapiens hyaluronan-mediated
Homo sapiens MCM10 minichromosome
Homo sapiens chromosome 17 open
Homo sapiens centromere protein I
Homo sapiens NDC80 homolog,
Homo sapiens anillin, actin binding protein
Homo sapiens G-2 and S-phase expressed
Homo sapiens hypothetical protein
Homo sapiens chromatin licensing and
Homo sapiens chromosome 19 open
Homo sapiens discs, large homolog 7
Homo sapiens CDC45 cell division cycle
Homo sapiens TTK protein kinase (TTK),
Homo sapiens GINS complex subunit 1
Homo sapiens G-2 and S-phase expressed
Homo sapiens maternal embryonic leucine
Homo sapiens Fanconi anemia,
Homo sapiens PDZ binding kinase (PBK),
Homo sapiens cDNA clone
Homo sapiens cell division cycle
Homo sapiens cancer susceptibility
Homo sapiens centromere protein F,
Homo sapiens kinesin family member 20A
Homo sapiens cyclin B2 (CCNB2), mRNA
Homo sapiens NIMA (never in mitosis
Homo sapiens keratin 18 (KRT18),
Homo sapiens chromatin licensing and
Homo sapiens ASF1 anti-silencing
Homo sapiens topoisomerase (DNA) II
Homo sapiens non-SMC condensin I
Homo sapiens asp (abnormal spindle)
Homo sapiens chromosome 1 open reading
Homo sapiens germ cell associated 2
Homo sapiens cell division cycle 25
Homo sapiens lamin B1 (LMNB1), mRNA
Homo sapiens centrosomal protein 55 kDa
Homo sapiens RAD51 associated protein 1
Homo sapiens ovostatin 2, mRNA (cDNA
Homo sapiens enhancer of zeste homolog
Homo sapiens protein regulator of
Homo sapiens E2F transcription factor 8
Homo sapiens nucleolar and spindle
Homo sapiens ZW10 interactor (ZWINT),
Homo sapiens MCM10 minichromosome
Homo sapiens cDNA FLJ13607 fis, clone
Homo sapiens DEP domain containing 1
Homo sapiens centromere protein I
Homo sapiens MLF1 interacting protein
Homo sapiens gamma-glutamyl hydrolase
Homo sapiens kinesin family member 4A
Homo sapiens kinesin family member 18A
Homo sapiens CHK1 checkpoint homolog
Homo sapiens family with sequence
Homo sapiens kinesin family member 14
Homo sapiens homeobox C9 (HOXC9),
Homo sapiens cyclin A2 (CCNA2),
Homo sapiens keratin 18 (KRT18),
Homo sapiens hypothetical protein
Homo sapiens chromosome 9 open reading
Homo sapiens X-ray repair complementing
Homo sapiens keratin 18 (KRT18),
Homo sapiens SPC24, NDC80 kinetochore
Homo sapiens shugoshin-like 1 (S. pombe)
Homo sapiens galactose-3-O-
Homo sapiens shugoshin-like 1 (S. pombe)
Homo sapiens aurora kinase B (AURKB),
Homo sapiens denticleless homolog
Homo sapiens cadherin, EGF LAG seven-
Drosophila) (CELSR3), mRNA
Homo sapiens cancer susceptibility
Homo sapiens collagen, type II, alpha 1
Homo sapiens high-mobility group
Homo sapiens DEP domain containing 1
Homo sapiens immunoglobulin
Homo sapiens ets variant gene 4 (E1A
Homo sapiens ets variant gene 4 (E1A
Homo sapiens E2F transcription factor 2
Homo sapiens kinesin family member 23
Homo sapiens Opa interacting protein 5
sapiens cDNA clone IMAGE: 2766163 3′,
Homo sapiens MARCKS-like 1
Homo sapiens 5′-nucleotidase domain
Homo sapiens creatine kinase, muscle
Homo sapiens high-mobility group box 3
Homo sapiens antigen identified by
Homo sapiens testicular cell adhesion
Homo sapiens centromere protein M
Homo sapiens RAD51 homolog (RecA
Homo sapiens olfactory-like receptor
Homo sapiens KIAA0101 (KIAA0101),
Homo sapiens cancer susceptibility
Homo sapiens centromere protein K
Homo sapiens NDC80 homolog,
Homo sapiens phosphatidic acid
Homo sapiens telomerase reverse
Homo sapiens mRNA for KIAA0599
Homo sapiens cystathionine-beta-synthase
Homo sapiens exonuclease 1 (EXO1),
Homo sapiens zinc finger protein 695
Homo sapiens ribonucleotide reductase M2
Homo sapiens nei endonuclease VIII-like 3
Homo sapiens mRNA; cDNA
Homo sapiens tumor necrosis factor
Homo sapiens heat shock transcription
Homo sapiens melanoma antigen family A,
Homo sapiens Rho GTPase activating
Homo sapiens matrilin 3 (MATN3),
Homo sapiens polymerase (DNA directed),
Homo sapiens cDNA FLJ25193 fis, clone
Homo sapiens RAD54-like (S. cerevisiae)
Homo sapiens olfactory receptor, family 7,
Homo sapiens E2F transcription factor 7
Homo sapiens laeverin (FLJ90650),
Homo sapiens coronin, actin binding
Homo sapiens olfactory receptor, family 7,
Homo sapiens butyrophilin-like 8
Homo sapiens chromosome 10 open
Homo sapiens melanoma antigen family A,
Homo sapiens olfactory receptor, family 7,
Homo sapiens diaphanous homolog 3
Homo sapiens chorionic
Homo sapiens homogentisate 1,2-
Homo sapiens cDNA clone
Homo sapiens scrapie responsive protein 1
Homo sapiens zinc finger protein 483
Homo sapiens MCF.2 cell line derived
Homo sapiens solute carrier family 24
Homo sapiens SH3-domain GRB2-like
Homo sapiens catenin (cadherin-associated
Homo sapiens cDNA clone
Homo sapiens KIAA1467 (KIAA1467),
Homo sapiens carboxypeptidase E (CPE),
Homo sapiens calcium binding protein 39-
Homo sapiens SH3-domain GRB2-like
Homo sapiens catenin (cadherin-associated
Homo sapiens similar to expressed
Homo sapiens C1q and tumor necrosis
Homo sapiens cDNA FLJ42672 fis, clone
Homo sapiens coiled-coil domain
Homo sapiens C1q and tumor necrosis
Homo sapiens myozenin 2 (MYOZ2),
Homo sapiens v-akt murine thymoma viral
Homo sapiens Fas apoptotic inhibitory
Homo sapiens ankyrin repeat domain 35
Homo sapiens latent transforming growth
Homo sapiens clone 23688 mRNA
Homo sapiens GLI pathogenesis-related 1
Homo sapiens receptor accessory protein 1
Homo sapiens resistance to inhibitors of
Homo sapiens GLI pathogenesis-related 1
Homo sapiens GLI pathogenesis-related 1
Homo sapiens cDNA FLJ20767 fis, clone
Homo sapiens cyclin D2 (CCND2),
Homo sapiens stearoyl-CoA desaturase 5
Homo sapiens FXYD domain containing
Homo sapiens microtubule-associated
Homo sapiens retinoic acid induced 2
Homo sapiens solute carrier family 24
Homo sapiens hypothetical protein
Homo sapiens GLI pathogenesis-related 1
Homo sapiens leiomodin 3 (fetal)
Homo sapiens leiomodin 3 (fetal)
Homo sapiens RIC3 isoform d (RIC3)
Homo sapiens cDNA clone UI-E-DX0-
Homo sapiens FLJ41603 protein
Homo sapiens resistance to inhibitors of
Homo sapiens cDNA FLJ12235 fis, clone
Homo sapiens low density lipoprotein-
Homo sapiens microtubule-associated
Homo sapiens dynein, light chain,
Homo sapiens receptor accessory protein 2
Homo sapiens adaptor-related protein
Homo sapiens DMRT-like family C1
Homo sapiens MOB1, Mps One Binder
Homo sapiens tubby homolog (mouse)
Homo sapiens DKFZP564O0823 protein
Homo sapiens coiled-coil domain
CcDNA clone
Homo sapiens clone 24626 mRNA
Homo sapiens cDNA FLJ13735 fis, clone
sapiens mRNA for KIAA0672 protein.
Homo sapiens DAB2 interacting protein
Homo sapiens cyclin D2 (CCND2),
Homo sapiens DAB2 interacting protein
Homo sapiens sortilin-related VPS10
Homo sapiens ankyrin 2, neuronal
Homo sapiens ADAM metallopeptidase
Homo sapiens fibroblast growth factor 12
Homo sapiens discs, large homolog 2,
Homo sapiens cDNA 3′ end similar to EST
Homo sapiens podocan (PODN), mRNA
Homo sapiens filamin A interacting protein
Homo sapiens HtrA serine peptidase 1
Homo sapiens myeloid/lymphoid or
Homo sapiens Rho GTPase activating
Homo sapiens hypothetical protein
Homo sapiens coagulation factor II
Homo sapiens serine/threonine kinase 33
Homo sapiens ISL2 transcription factor,
Homo sapiens chromogranin B
Homo sapiens Rho GTPase activating
Homo sapiens chromogranin B
Homo sapiens cDNA FLJ31517 fis, clone
Concerning the 70 down-regulated genes, no significantly enriched pathways were identified. In contrast, we observed that 45 of the 227 up-regulated genes belonged to the CINSARC signature (
Moreover, gene enrichment analysis of the 182 genes not included in CINSARC showed that this gene set was also enriched by genes involved in the same pathways as CINSARC genes, i.e. mitosis control and chromosome integrity (Table 4).
AURKA is a Significant Marker of Metastasis Outcome
We took advantage of the supervised analysis results to test the possibility of reducing the CINSARC signature. Among the top-ranked significant genes sorted in the supervised t-test, AURKA (Aurora kinase A, previously designated STK6 or STK15) was the best ranked gene that also belonged to the CINSARC signature (Table 1). We thus tested whether AURKA alone could predict outcomes as well as CINSARC and we stratified samples according to their AURKA expression (with the mean expression of 9.13 as a cut-off, table 5).
For this purpose, we considered the present 67-GISTs series as the training set and the Yamaguchi's one as the validation set. Expression data were then validated by qRT-PCR and we found a high correlation between both techniques (Pearson correlation coefficient=0.94; P<1×10−15). Survival analyses revealed that the two groups obtained had very different outcomes, both in the training set (Present series, MFS: P=5.31×10−11 and DFS: P=3.61×10−12,
Chromosomal Complexity is a Significant Prognosis Factor of GISTs
We have previously shown that the CINSARC signature is associated with the genome complexity (18), therefore the question arises whether the alteration level of the GISTs genome is correlated with the CINSARC signature and with the metastatic outcome. Genome profiling with arrays containing 60 000 oligonucleotides (see material and methods) has been performed on 66 GISTs with sufficient DNA quality. Different profiles were obtained, ranging from simple, i.e. without any detectable changes, to complex, with numerical and segmental gains and losses (
Integrative Analysis Allows Identification of a No-Risk Group of Patients
Considering these results as a whole, we construct a decisional algorithm based on GI and AURKA expression. More specifically, a positive correlation exists between GI and AURKA expression (Pearson correlation r=0.65,
P16/RB1 Pathway is Associated to Metastatic Outcome.
As a result of these findings, we reconsidered CGH array data to examine whether any specific alterations were associated with patients' outcome. We compared alteration frequency of each probe set between GISTs with or without metastatic outcome (
We thus hypothesized that another genomic alteration could lead to p16 pathway inactivation in cases without p16 homozygous deletion. We observed one homozygous deletion and 13 hemizygous deletion of the RB1 locus (Table 6).
Eleven tumors harboring RB1 deletions are classified AG2 and eight developed metastases. Interestingly tumors with a p16 homozygous deletion are without any RB1 deletion although they are highly rearranged. Sequencing RB1 in all patients with available RNA and DNA (66 cases) we indentified one mutation (c.1959—1960del/p.Lys653AsnfsX14) in the retained copy in GIST #61. qPCR analysis confirmed that deleted tumors had a significantly down regulated expression of RB1 (Table 6).
We demonstrated that CINSARC is a very powerful signature to predict metastatic outcome. CINSARC is composed of 67 genes which are all involved in chromosome integrity and mitosis control pathways, indicating that such mechanisms appear to be driving the development of metastasis in this tumor type, as we recently demonstrated in sarcomas (18). This is in line with results from the reciprocal approach, which is the identification of genes differentially expressed between GISTs with or without metastatic outcome. Actually, among the 227 up-regulated genes in the 18 metastasizing tumors, 45 were common with the CINSARC signature and the activated pathways were almost all the same. These results indicate that genome integrity and mitosis control are the effective restrain mechanisms underlying development of metastasis, and moreover, that these mechanisms appear to be sufficient, or at least the strongest. In line with this, we show that expression of the top ranked gene in both approaches, AURKA, is as efficient as CINSARC to predict metastatic outcome in both series of GISTs.
Following our results demonstrating the central role of the genome integrity control, we hypothesized that a defect of such mechanisms should lead to chromosome imbalances and that the resulting genome complexity should also predict outcome in GISTs. This is exactly what we show here since tumor stratification according to a CGH Genome Index (GI) which integrates the number of alterations and of altered chromosomes forms two groups of clearly distinct outcomes. This is clearly in agreement with the AURKA expression results, since whole chromosome losses are the most frequent alterations observed in GISTs and these are assumed to originate from the chromosome segregation deficiency induced by mitosis check-point defects, such as the AURKA overexpression (34).
In contrast to results of Yamagushi and colleagues, our study demonstrates that CINSARC, AURKA expression or CGH prognostic values are irrespective to tumor location. Furthermore, as mentioned above, the biological meaning of CINSARC and its association to genomic changes strongly indicate that CINSARC genes are involved in malignant progression and are not just a consequence of the process. This hypothesis is supported by the association we observe here between CDKN2A deletions (homozygous deletions in 6 cases and hemizygous deletions in 9 cases among 18 cases with metastatic outcome versus 5 hemizygous deletions in 48 non-metastatic patients), CINSARC prognosis groups and metastatic occurrence. Previous studies have already pointed out the potential association of CDKN2A alterations or expression of p16INK4a to tumor progression (11-16, 43). Nevertheless, data remain controversial mainly due to lack of clear delineation of the targeted gene at the 9p21 locus. It is still unclear which gene of CDKN2A, CDKN2B or MTAP is driving the association to poor prognosis. At the genomic level, even if CDKN2A and 2B appear to be systematically codeleted (37, 44, 45), two studies indicate that 9p21 deletions are likely to target the MTAP gene and not exclusively CDKN2A and CDKN2B (35, 46). Here, CGH and genomic qPCR analyses demonstrated that homozygous deletions specifically target CDKN2A and that the common region of deletion excludes CDKN2B and MTAP. Surprisingly, we did not find any harmful CDKN2A mutations in any of GISTs case tested. Schnieder-Stock and colleagues (14) reported 9 so-called mutations in a series of 43 GISTs. But two of them are identical and have been detected in a tumor and its recurrence, one is now referenced as a SNP, two are silent mutations and in one case no interpretable sequence was obtained. Considering all this, the authors evidenced only four CDKN2A mutations (4/43=9%). According to these data we expected around five mutations in our study and we have identified three changes, two SNP and one silent change not referred so far. One explication for this discrepancy could be sampling bias, but it is of interest to note that, we detected twice more homozygous CDKN2A deletions than reported in the study of Schneider-Stock et al (7/63 vs 2/43). Following the idea that another exclusive alteration could explain aggressive tumors (CINSARC C2, AG2) without p16 inactivation we identified two tumors without RB1 functional copy and 12 significantly down-regulated due to loss of one RB1 copy (p value). We did not detect any truncating mutation in these tumors but we hypothesize that micro-deletions, that we did not identify because of the lowest resolution of the arrays, could account for this second inactivation, as in sarcomas (47). An exclusive occurrence of p16 and RB1 alterations is highly supported by the observation that none of the tumors with CDKN2A homozygous deletion harbors any RB1 deletion and among the 29 GISTs with one of these deletions, only three cases harbor both deletions (table 1). Altogether, p16/RB1 pathway is inactivated or down regulated in 14/18 (78%) and in 3/48 (6%) patients with and without metastatic outcome, respectively, which clearly means that inactivation of p16/RB1 pathway is associated to metastatic development.
CDKN2A codes for two key tumor suppressor proteins, the p16INK4a and the p14ARF, which are involved in the regulation of the cell cycle G1 and G2/M transition. Together, these proteins regulate two important cell cycle checkpoints, the p53 and the RB1 pathways for p14 and p16INK4a, respectively. Loss of these genes can lead to replicate senescence, cell immortalization and tumor growth (48-51). Most of the CINSARC genes are under the transcriptional control of E2F, which is tightly regulated by RB 1 interaction. Actually, RB 1 sequestrates E2F which is delivered upon RB1 phosphorylation by CDK4 (Cyclin Dependent Kinase 4) and p16INK4a inhibits CDK4. Therefore, our results allow us to hypothesize that inactivation of the p16/RB1 pathway in GISTs, mainly by deletion, is likely to be the causative alteration that leads to the over-expression of genes involved in mitosis control. This deregulation triggers cell genome rearrangements until a combination is naturally selected and fixed. Thus, the resulting genome complexity and its related expression confer the tumor cell aggressiveness and metastatic potential. Although this hypothesis has to be experimentally validated in cellular and mouse models, it is supported by the expression analysis of the GISTs with or without functional p16/RB1 pathway which shows that 42/225 (19%) genes up regulated in GISTs without functional p16/RB1 pathway are common with CINSARC signature. Moreover these 225 genes are involved in the same pathways than those enriched in CINSARC and metastases signatures (Supplementary table 3).
Imatinib mesylate has been proven to target KIT-aberrant signaling inhibiting the proliferation and survival in GIST cells. Until 2009, imatinib therapy was restricted to disseminated or advanced disease at the time of diagnosis. Since then, adjuvant treatment has been approved and the necessity to apply selection criteria to identify patients susceptible to benefit from such management has emerged. Patient selection foreseen by FDA (Food and Drug Administration) and to a lesser extent by EMA (European Medicines Agency) is essentially based on the histological risk evaluation. Both AFIP (9) and NIH (8) histological-based staging systems are widely accepted as “gold standards” in determining tumor metastatic risk and to determine whether a GIST patient is eligible or not for adjuvant therapy with imatinib. Here we show that the CINSARC signature and AURKA expression outperform the AFIP classification (survival analysis according to AFIP classification is presented in
We thus propose two possible decisional methods either to enhance the AFIP or NIH grading systems or to replace these histopathological methods. Firstly, when using the AFIP or NIH classifications, intermediate-risk cases are problematic for therapeutic management and our results demonstrate that the use of CGH profiling can easily and rapidly solve such a problem. Secondly, our results suggest that the combined use of GI and AURKA expression offer a better selection of patients for imatinib therapy than the AFIP classification does. Both methods offer equally efficient treatments for patients with metastatic risk, but CINSARC/AURKA-based selection, which is totally investigator-independent, would diminish consistently the number of patients, without metastatic risk, who are falsely declared eligible for imatinib therapy.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10013806.4 | Oct 2010 | EP | regional |
The present application is a National Phase entry of PCT Application No. PCT/IB2011/054688, filed Oct. 20, 2011, which claims priority from EP Application No. 10013806.4, filed Oct. 20, 2010, which applications are hereby incorporated by reference herein in their entirety.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/IB2011/054688 | 10/20/2011 | WO | 00 | 7/9/2013 |
