Stéphane Liva

VAMP: Visualization and analysis of array-CGH, transcriptome and other molecular profiles

MOTIVATION Microarray-based CGH (Comparative Genomic Hybridization), transcriptome arrays and other large-scale genomic technologies are now routinely used to generate a vast amount of genomic profiles. Exploratory analysis of this data is crucial in helping to understand the data and to help form biological hypotheses. This step requires visualization of the data in a meaningful way to visualize the results and to perform first level analyses. RESULTS We have developed a graphical user interface for visualization and first level analysis of molecular profiles. It is currently in use at the Institut Curie for cancer research projects involving CGH arrays, transcriptome arrays, SNP (single nucleotide polymorphism) arrays, loss of heterozygosity results (LOH), and Chromatin ImmunoPrecipitation arrays (ChIP chips). The interface offers the possibility of studying these different types of information in a consistent way. Several views are proposed, such as the classical CGH karyotype view or genome-wide multi-tumor comparison. Many functionalities for analyzing CGH data are provided by the interface, including looking for recurrent regions of alterations, confrontation to transcriptome data or clinical information, and clustering. Our tool consists of PHP scripts and of an applet written in Java. It can be run on public datasets at http://bioinfo.curie.fr/vamp AVAILABILITY The VAMP software (Visualization and Analysis of array-CGH,transcriptome and other Molecular Profiles) is available upon request. It can be tested on public datasets at http://bioinfo.curie.fr/vamp. The documentation is available at http://bioinfo.curie.fr/vamp/doc.

Spatial normalization of array-CGH data

BackgroundArray-based comparative genomic hybridization (array-CGH) is a recently developed technique for analyzing changes in DNA copy number. As in all microarray analyses, normalization is required to correct for experimental artifacts while preserving the true biological signal. We investigated various sources of systematic variation in array-CGH data and identified two distinct types of spatial effect of no biological relevance as the predominant experimental artifacts: continuous spatial gradients and local spatial bias. Local spatial bias affects a large proportion of arrays, and has not previously been considered in array-CGH experiments.ResultsWe show that existing normalization techniques do not correct these spatial effects properly. We therefore developed an automatic method for the spatial normalization of array-CGH data. This method makes it possible to delineate and to eliminate and/or correct areas affected by spatial bias. It is based on the combination of a spatial segmentation algorithm called NEM (Neighborhood Expectation Maximization) and spatial trend estimation. We defined quality criteria for array-CGH data, demonstrating significant improvements in data quality with our method for three data sets coming from two different platforms (198, 175 and 26 BAC-arrays).ConclusionWe have designed an automatic algorithm for the spatial normalization of BAC CGH-array data, preventing the misinterpretation of experimental artifacts as biologically relevant outliers in the genomic profile. This algorithm is implemented in the R package MANOR (Micro-Array NORmalization), which is described at http://bioinfo.curie.fr/projects/manor and available from the Bioconductor site http://www.bioconductor.org. It can also be tested on the CAPweb bioinformatics platform at http://bioinfo.curie.fr/CAPweb.

Bioinformatics for precision medicine in oncology: principles and application to the SHIVA clinical trial

Precision medicine (PM) requires the delivery of individually adapted medical care based on the genetic characteristics of each patient and his/her tumor. The last decade witnessed the development of high-throughput technologies such as microarrays and next-generation sequencing which paved the way to PM in the field of oncology. While the cost of these technologies decreases, we are facing an exponential increase in the amount of data produced. Our ability to use this information in daily practice relies strongly on the availability of an efficient bioinformatics system that assists in the translation of knowledge from the bench towards molecular targeting and diagnosis. Clinical trials and routine diagnoses constitute different approaches, both requiring a strong bioinformatics environment capable of (i) warranting the integration and the traceability of data, (ii) ensuring the correct processing and analyses of genomic data, and (iii) applying well-defined and reproducible procedures for workflow management and decision-making. To address the issues, a seamless information system was developed at Institut Curie which facilitates the data integration and tracks in real-time the processing of individual samples. Moreover, computational pipelines were developed to identify reliably genomic alterations and mutations from the molecular profiles of each patient. After a rigorous quality control, a meaningful report is delivered to the clinicians and biologists for the therapeutic decision. The complete bioinformatics environment and the key points of its implementation are presented in the context of the SHIVA clinical trial, a multicentric randomized phase II trial comparing targeted therapy based on tumor molecular profiling versus conventional therapy in patients with refractory cancer. The numerous challenges faced in practice during the setting up and the conduct of this trial are discussed as an illustration of PM application.

CAPweb: a bioinformatics CGH array Analysis Platform

Assessing variations in DNA copy number is crucial for understanding constitutional or somatic diseases, particularly cancers. The recently developed array-CGH (comparative genomic hybridization) technology allows this to be investigated at the genomic level. We report the availability of a web tool for analysing array-CGH data. CAPweb (CGH array Analysis Platform on the Web) is intended as a user-friendly tool enabling biologists to completely analyse CGH arrays from the raw data to the visualization and biological interpretation. The user typically performs the following bioinformatics steps of a CGH array project within CAPweb: the secure upload of the results of CGH array image analysis and of the array annotation (genomic position of the probes); first level analysis of each array, including automatic normalization of the data (for correcting experimental biases), breakpoint detection and status assignment (gain, loss or normal); validation or deletion of the analysis based on a summary report and quality criteria; visualization and biological analysis of the genomic profiles and results through a user-friendly interface. CAPweb is accessible at .

Quantitative PCR analysis of c-erb B-2 (HER2/neu) gene amplification and comparison with p185HER2/neu protein expression in breast cancer drill biopsies†

A PCR assay using capillary electrophoresis was designed for the detection of c‐erbB‐2 gene amplification in alcohol‐formalin‐acetic acid (AFA)–fixed, paraffin‐embedded biopsies from 81 consecutive breast tumors. c‐erbB‐2 expression was analyzed in the same samples using immuno‐histochemistry (IHC). In the competitive PCR assay, a single pTag plasmid containing a 4‐nucleotide (nt)‐deleted copy of a 124‐nt sequence of c‐erbB‐2 and a 4‐nt‐deleted copy of a 120‐nt sequence of GAPDH was co‐amplified with genomic DNA extracted from 3 10‐μm‐thick tissue sections of the tumor biopsy. The percentage of tumor cells in the biopsy specimen and the percentage of tumor cells stained with the membrane anti‐c‐erbB‐2 monoclonal antibody CB11 were recorded by a single pathologist on 2 consecutive sections. Among 81 consecutive tumor biopsies assayed by PCR, 21 (26%) displayed unequivocal c‐erbB‐2 amplification (actual gene copy number, AGCN > 4), 47 (58%) displayed no c‐erbB‐2 amplification (AGCN ≤ 2) and 7 (9%) could not be analyzed due to an insufficient amount of DNA. Six samples (7%) were considered inconclusive since the percentage of tumor cells was <20%. Analysis of c‐erbB‐2 expression by IHC showed that among the 21 amplified specimens 15 displayed strong staining, while all non‐amplified samples (47) displayed no or only weak staining. The concordance of the 2 techniques was 91%. We conclude that c‐erbB‐2 gene amplification can be accurately quantitated by competitive PCR performed on small, fixed and embedded tumor samples. Int. J. Cancer 83:157–161, 1999.

Preferential Occurrence of Chromosome Breakpoints within Early Replicating Regions in Neuroblastoma

Neuroblastoma (NB) is a frequent paediatric extra cranial solid tumour characterized by the occurrence of unbalanced chromosome translocations, frequently, but not exclusively, involving chromosomes 1 and 17. We have used a 1 Mb resolution BAC array to further refine the mapping of breakpoints in NB cell lines. Replication timing profiles were evaluated in 7 NB cell lines, using DNAs from G1 and S phases flow sorted nuclei hybridised on the same array. Strikingly, these replication timing profiles were highly similar between the different NB cell lines. Furthermore, a significant level of similarity was also observed between NB cell lines and lymphoblastoid cells. A segmentation analysis using the Adaptative Weights Smoothing procedure was performed to determine regions of coordinate replication. More than 50% of the breakpoints mapped to early replicating regions, which account for 23.7% of the total genome. The breakpoints frequency per 108 bases was therefore 10.84 for early replicating regions, whereas it was only 2.94 for late replicating regions, these difference being highly significant (p

p53 mutations in BRCA1-associated familial breast cancer.

Crook and colleagues report that p53 mutations in germline BRCA1-mutation-associated breast cancers can be found with high frequency (100%) by preferential localisation in exon 5 around codon 133–165 of the p53 gene. We did a study of 13 breast and ovarian tumours from 16 patients with a known BRCA1 or BRCA2 germline mutation (table). Each sample was controlled by a pathologist to ensure that tumoral DNA was present. In each tumour, the presence of the BRCA1 mutation was controlled and loss of heterozygosity at the BRCA1 locus was studied by the three polymorphic intragenic markers D17S855, D17S1322, and D17S1323. We did molecular analysis of the p53 gene by direct sequencing of DNA extracted from frozen tumours. Exons 4 to 9 were sequenced in both directions. The immunohistochemical analysis was done on the same tumour sample with the DO7 antibody. p53 mutations were found in three (20%) of 14 tumours with a BRCA1 germline mutation. One mutation was also detected in the tumour of patient 127 with a BRCA2 germline mutation. There was no clustering of these four p53 mutations, which is typical of sporadic breast cancers. Furthermore, no mutation was found in exon 4–9 of the remaining tumours. Although, we did not screen the entire p53 sequence, the probability of mutation in the remaining exons is very low, since it is widely accepted that more than 95% of p53 mutations are localised in this region. p53 accumulation was found in six (40%) of 14 BRCA1-associated tumours and in one BRCA2associated tumours. The three tumours with missense mutations in the p53 gene were positive for protein accumulation, whereas the stop mutation did not lead to stable synthesis of p53. Three tumours presented p53 accumulation without any p53 mutation. Such findings have been reported for sporadic breast carcinomas, with nuclear accumulation of p53 protein in tumours expressing wild-type p53. Our results are similar to those reported for the status of the p53 gene in sporadic breast cancers. A recent study on 51 tumours from Swedish breast-cancer families had similar results but BRCA1 status was not checked. We have no explanation for the discrepancy between our results and those published by Crook and colleagues, but important features should be taken into account. In our study, the panel of BRCA1 mutations was diverse, with either missense or truncating mutations scattered along the coding sequence. Although Crook and colleagues do not fully describe the BRCA1 mutations, a specific BRCA1/p53 mutation association cannot be excluded. p53 mutations have, however, been detected in many cancers and have led to the discovery of specific mutational events that could be taken to show a pattern of carcinogen exposure. Hartman and colleagues stress the finding that there is an important geographical variation in the frequency and pattern of somatic p53 mutations in breast cancer, which suggests that their origin could be influenced by differences in exposure, sensitivity to carcinogens, or both, or to other genetic factors. In BRCA1-associated breast cancers such variation in p53 mutation patterns could be expected since the BRCA1 protein has a putative function in controlling genome integrity. Therefore, the link between germline BRCA1 mutation and the type of p53 alteration could be heterogeneous, dependent on the origin of the latter. Similar studies are needed in other geographical areas to explore this important aspect of breast cancer carcinogenesis.

ACTuDB, a new database for the integrated analysis of array-CGH and clinical data for tumors

In recent years, an increasing number of projects have investigated tumor genome structure, using microarray-based techniques like array comparative genomic hybridization (array-CGH) or single nucleotide polymorphism (SNP) arrays. The forthcoming studies have to integrate these former results and compare their findings to the existing sets of copy number data for validation. These sets also form the basis from which many comparative retrospective analyses can be carried out. Nevertheless, exploitation of this mass of data relies on a homogeneous preparation of copy number data, which will make it possible to compare them together, and their integration into a unified bioinformatics environment with ad hoc analysis tools and interfaces. To our knowledge, no such data integration has been proposed yet. Therefore the biologists and clinicians involved in cancer research urgently need such an integrative tool, which motivated us to undertake the construction of a database for array-CGH and other DNA copy number data for tumors (ACTuDB). When available, the associated clinical, transcriptome and loss of heterozygosity data were also integrated into ACTuDB. ACTuDB contains currently about 1500 genomic profiles for tumors and cell lines for the bladder, brain, breast, colon, liver, lymphoma, neuroblastoma, mouth and pancreas, together with data for replication timing experiments. The CGH array data were processed, using ad hoc algorithms (probe mapping, breakpoint detection, gain or loss status assignment and visualization) developed at Institut Curie. The database is available from http://bioinfo.curie.fr/actudb/ and can be browsed with a user-friendly interface. This database will be a useful resource for the genomic profiling of tumors, a field of highly active research. We invite research groups involved in tumor genome profiling to submit their data to ACTuDB.

Aberrant ERBB4-SRC Signaling as a Hallmark of Group 4 Medulloblastoma Revealed by Integrative Phosphoproteomic Profiling

The current consensus recognizes four main medulloblastoma subgroups (wingless, Sonic hedgehog, group 3 and group 4). While medulloblastoma subgroups have been characterized extensively at the (epi-)genomic and transcriptomic levels, the proteome and phosphoproteome landscape remain to be comprehensively elucidated. Using quantitative (phospho)-proteomics in primary human medulloblastomas, we unravel distinct posttranscriptional regulation leading to highly divergent oncogenic signaling and kinase activity profiles in groups 3 and 4 medulloblastomas. Specifically, proteomic and phosphoproteomic analyses identify aberrant ERBB4-SRC signaling in group 4. Hence, enforced expression of an activated SRC combined with p53 inactivation induces murine tumors that resemble group 4 medulloblastoma. Therefore, our integrative proteogenomics approach unveils an oncogenic pathway and potential therapeutic vulnerability in the most common medulloblastoma subgroup.

Caractérisation des altérations génomiques des carcinomes infiltrants du col utérin

Le cancer du col uterin est le deuxieme cancer le plus frequent chez la femme dans le monde. Les carcinomes infiltrants s’accompagnent d’une integration genomique cellulaire d’ADN de papillomavirus humain (PVH), en particulier les PVH a haut risque de type 16 et 18. L’oncogenese du col uterin necessite la presence des oncoproteines E6 et E7 du PVH qui induit la proliferation cellulaire par degradation des proteines TP53 et pRB. Il s’ensuit une instabilite genetique, necessaire a la transformation maligne. Les consequences moleculaires de cette instabilite genetique ont ete peu caracterisees jusqu’a present. La technique de puce genome (CGH array, Comparative Genomic Hybridization array) permet de caracteriser les alterations du nombre de copies d’ADN d’un prelevement tumoral. Le principe consiste en une hybridation genomique comparative d’un melange d’ADN tumoral et d’ADN normal sur de multiples sondes genomiques representatives de l’ensemble des locus chromosomiques. Nous avons utilise cette methodologie pour 37 prelevements de carcinome du col uterin : 8 lignees cellulaires et 29 carcinomes primitifs infiltrants. Apres coupure enzymatique, l’ADN tumoral a ete marque par un fluorochrome (cyanine 5) et l’ADN normal par un autre fluorochrome (cyanine 3). Les deux ADN ont hybride de facon competitive sur la puce genome, sur laquelle ont ete places 3 500 clones de BAC (Bacterial Artificial Chromosome). Les niveaux d’intensite de chaque fluorochrome ont ete mesures pour chaque clone. Les pertes et gains d’ADN tumoral ont ete deduits du rapport d’intensite des 2 fluorochromes. Les pertes chromosomiques les plus frequentes etaient observees pour les bras chromosomiques 2q, 3p, 9p, 11q et 16q. Les gains chromosomiques les plus frequents portaient sur les bras chromosomiques 1q, 3q, 5p, 8q, 16p et 20q. L’analyse detaillee de ces desequilibres alleliques a permis de delimiter precisement les bornes chromosomiques de ces alterations genomiques. Nous avons egalement identifie 20 amplifications genomiques presentes dans 14 tumeurs. Quatre locus chromosomiques comportaient des amplifications recurrentes : 11q22.2 contenant les genes MMP7 et MMP20 (4 cas), 1p21.1 (3 cas), 8q24.21 contenant le gene MYC (2 cas) et 11q13.3 contenant le gene CCND1 (2 cas). Ces resultats, couples a ceux de l’analyse du transcriptome, permettent de caracteriser precisement les anomalies moleculaires des carcinomes du col uterin et d’identifier de nouveaux genes potentiellement impliques dans l’oncogenese du col uterin.