ALBERT

Description: Abstract 2933 Poster Board II-909 The challenge of clinical proteomic is to link protein expression profile variations to specific disease phenotypes and identify relevant biomarkers in order to develop straightforward diagnostic or prognosis tools. Blood, a tissue that interfaces with virtually every part of the body, is considered to be a deep source of native and secreted diagnostic analytes. Despite this great potential, the first decade of the proteomics era met with little success, not only because of the vast and complex nature of the proteome but also due to proteins dynamic range and complex degradation pathways, and to the heterogeneity of plasma protein profiles in the human population. Altogether, progress in clinical proteomics will reside in the elaboration of standardized preanalytic procedures, cross-comparisons between samples and independent validation. In aggressive diffuse large B-cell lymphomas (DLBCL), diagnostic and prognostic biomarkers are mandatory to optimize treatment, and include patients in future trials. The aim of the present study was first to identify diagnostic blood biomarkers of DLBCL based on the 075 French GOELAMS ongoing trial -which involved adults younger than 60 suffering from an aggressive form of DLBCL- that randomized patients between CHOP-14 Rituximab or intensive chemotherapy plus Rituximab including autologous stem-cell support. This protocol was built after our group demonstrated the high efficiency of high-dose chemotherapy and autologous stem-cell transplantation as frontline therapy in this disease compared to conventional CHOP (New Engl J Med, 2004). In this study, 200 patients were compared to 100 controls matched for sex and age. Well-defined pre-analytic steps were applied and plasma was collected, at the time of diagnosis, on the specific anti-proteasic P100v1.1 tube from Becton Dickinson. All samples were centralized and aliquoted in a unique platform prior to analytic steps. The whole series of 300 samples was randomly assigned on chips to be analyzed with SELDI-TOF/MS using three different chemistry protocols (CM10, Q10 & IMAC30) and two beam intensities (2000 and 4000, respectively). There was a longer delay in the process of patient's samples before plasma isolation compared to controls; this time had to be considered since it could participate to the protein degradation process and lead to proteomic modifications. Statistical analyses were implemented with the R package software [R development core team. R: A Language for Environmental and Statistical Computing. Vienna: R Foundation, 2008.]. Univariate analyses comparison resulted in 185 peaks differential of the case-control status (t-test, FDR=5%). Multivariate analyses were then performed according to chemistry and beam intensity using stepforward logistic regression. This resulted in 78 peaks related to DLBCL diagnosis. In order to reduce dimension, partial least square regression [A. L. Boulesteix and K. Strimmer (2005). Predicting Transcription Factor Activities from Combined Analysis of Microarray and ChIP Data: A Partial Least Squares Approach] was applied, resulting in two components corresponding to a weighted sum of the 78 peaks. These two components were introduced as covariates in logistic regression so that the 78 peaks could be ranked according to their global coefficient, allowing then top peaks to be studied. Sparse partial least square was also considered as another approach to reduce dimension and select peaks among the 78 identified. These two approaches were compared and proteins studied in greater detail. Altogether, this study allowed to identify promising candidate cancer biomarkers that are currently being validated through the analysis of additional plasma issued from other types of lymphoma (follicular, mantle cell and low burden DLBCL) and non-cancerous septic patients. Highly specific peak combinations will be considered before peptide characterization in order to end up with a diagnostic set of proteins useful for the diagnosis and management of aggressive DLBCLs. Disclosures: No relevant conflicts of interest to declare.

Description: Abstract 276 Treatment strategies of acute myeloid leukaemia (AML) largely depend on their cytogenetic status. However this stratification remains insufficient for almost half of the patients, requiring subsequent molecular investigations. In our study, we used mass spectrometry-based proteomic approaches to characterize de novo-AML. Samples (blood mononuclear cells collected and frozen before any treatment) were retrospectively selected from two independent sets of newly diagnosed AML patients, included in prospective clinical trials of the GOELAMS (Groupe d'Etude Ouest-Est des Leucémies aigues). We showed that protein signature of leukemic cells defined 2 groups of patients that displayed significant variation of overall and disease free survival (Fig-1A). This proteomic classification refined cytogenetic classes. Combination of proteomic and cytogenetic allowed a new stratification highly correlated with outcome. In particular, AML with intermediate and high risk cytogenetic could be respectively subdivided according to their protein profiles in two subgroups with significantly different survival (Fig-1B). Interestingly in both type 2 and type 3 cytogenetic groups, a good proteomic profile identified subsets of patients with significantly increased probability of survival, suggesting that the weigh of a good proteomic risk might predominate over a bad predictive cytogenetic. Among proteins expressed by leukemic cells, we isolated a 10800 Da marker that retained the highest discriminative value between alive and deceased patients. The median intensity of the 10800 peak was 94.3 ± 97.7 (7.5-368.7) in living patients compared to 214.2 ± 163.7 (9.0-693.4) in dead patients (p= 0.009). Among normal cytogenetic patients, intensity levels of the marker was also significantly different between dead (217.4 ± 155.7 range 10.4-612.8) and alive patients (26.0 ± 73.1, range 6-168.1; p= 0.008) (Fig-1C). Using a thresholds of 100 (calculated by ROC curves) we were able to correctly identify 82% of patients who died (patients greater than 100) and 70% of patients who stayed alive (patients lower than 100) with a specificity of 65% and 82% respectively. These data were confirmed in a second independent set of patients. 10800 Da marker was identified by MS peptide sequencing as S100A8 (also designated MRP8 or calgranulin A), a cytosolic protein of mature granulocytes. Western blot analysis confirmed its expression mainly in AML patients with the worst prognostic but not in all AML patients, neither in some leukemic and lymphoma cell lines, arguing for a selective deregulation associated with poor prognosis. These results show that expression of S100A8 in leukemic cells at diagnosis might be considered as a predictor of low survival. Disclosures: No relevant conflicts of interest to declare.

Description: Non-germinal centre small B-cell lymphomas represent a heterogeneous group of non-hodgkin lymphomas which most frequent histologic subtypes are small lymphocytic lymphoma (SLL), marginal zone lymphoma (MZL) and mantle cell lymphoma (MCL). These three lymphoma entities have very different clinical outcomes but may be difficult to distinguish either histologically or clinically. We previously identified transcriptomic signatures specific of these 3 lymphoma subtypes. We further analyzed these lymphomas using Surface-Enhanced Laser Desorption/Ionisation Time of Flight (SELDI-TOF). A total of 58 tumors, including 20 SLL (all lymph nodes), 20 MZL (1 lymph node and 19 spleens) and 18 MCL (19 lymph nodes and 1 spleen) were analyzed. In addition, we included 7 controls obtained from traumatic normal spleens. The spectra were generated on weak cation exchange (CM10), strong anion exchange (Q10) and reversed-phase (H50) ProteinChip arrays. Protein patterns of all samples were comparatively analysed using two distinct strategies. We first used a binary recursive partitioning method with the Biomarker Pattern software (Ciphergen®), and second a hierarchical clustering method to visualized patterns of protein peaks completed with a supervised method (discriminating score) to point out individual peaks distinguishing the three histological subtypes (SLL, MZL and MCL). Spectra analyses revealed a very homogeneous protein patterns among all lymphoma samples. However specific SLL, MZL and MCL signatures based on 34 protein peaks with differential expression could be identified and allowed to classify 95% of the samples in their respective entity. SLL signature included 9 peaks, MZL signature 16 peaks and MCL signature 9 peaks. The binary recursive partitioning analysis was concordant but identified only the five most discriminant peaks. Further identification of the discriminating peaks is currently realized using SELDI-assisted purification. We are focusing on peaks at 9942 Da for SLL and at 11324 Da for MCL. Functional genomic studies can distinguish non-germinal small B-cell lymphomas at the transcriptomic level (our previous study) and at the proteomic level. This will provide new markers for diagnosis and potentially new therapeutic targets.