Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.


The expression of 16 genes related to the cell of origin and immune response predicts survival in elderly patients with diffuse large B-cell lymphoma treated with CHOP and rituximab


Gene expression profiles have been associated with clinical outcome in patients with diffuse large B-cell lymphoma (DLBCL) treated with anthracycline-containing chemotherapy. Using Affymetrix HU133A microarrays, we analyzed the lymphoma transcriptional profile of 30 patients treated with CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) and 23 patients treated with rituximab (R)-CHOP in the Groupe d'Etude des Lymphomes de l'Adulte clinical centers. We used this data set to select transcripts showing an association with progression-free survival in all patients or showing a differential effect in the two treatment groups. We performed real-time quantitative reverse transcription-PCR in the 23 R-CHOP samples of the screening set and an additional 44 R-CHOP samples set to evaluate the prognostic significance of these transcripts. In these 67 patients, the level of expression of 16 genes and the cell-of-origin classification were significantly associated with overall survival, independently of the International Prognostic Index. A multivariate model comprising four genes of the cell-of-origin signature (LMO2, MME, LPP and FOXP1) and two genes related to immune response, identified for their differential effects in R-CHOP patients (APOBEC3G and RAB33A), demonstrated a high predictive efficiency in this set of patients, suggesting that both features affect outcome in DLBCL patients receiving immunochemotherapy.


Diffuse large B-cell lymphoma (DLBCL) represents roughly 30% of non-Hodgkin lymphomas in adult patients.1 Half of the patients can be cured with anthracycline-based chemotherapy. Although the highly variable outcome reflects underlying molecular heterogeneity, the therapeutic stratification of patients currently relies on the use of the International Prognostic Index (IPI) based on clinical prognostic factors.2 During the last years, the availability of DNA microarray technology made it possible to obtain important insights into the molecular heterogeneity of this disease and demonstrated the existence of distinct subgroups harboring specific gene expression signatures.3 The Lymphoma/Leukemia Molecular Profiling project showed that these lymphomas could be separated into three groups: those with a germinal center (GC) profile, those with an activated B-cell (ABC) profile and a third group called type III.4, 5 Later on, using a consensus-clustering approach, Monti et al.6 showed that DLBCL comprised three discrete subsets: OxPhos (oxidative phosphorylation), BCR/Prolif (B-cell receptor/proliferation) and HR (host response), which did not overlap with the cell-of-origin (COO) classification. DLBCLs with a GC profile were shown to be associated with a better overall survival (OS) than DLBCLs with an ABC profile in two independent large series,5, 6 whereas OxPhos, BCR/Prolif and HR DLBCL subsets did not show prognosis significance. Several groups attempted to identify gene signatures that would be specifically helpful in predicting outcome, and proposed predictors based on the expression of 6–17 genes, established in patients who had received conventional chemotherapy.5, 7, 8

The GELA LNH98.5 clinical trial demonstrated the benefit of rituximab association with CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy in elderly patients with DLBCL.9, 10 The addition of an anti-CD20 monoclonal antibody to chemotherapy has now become a standard in this disease.11 The mechanism of action of this drug was shown to depend on antibody-dependant cell cytotoxicity, complement-dependant cytotoxicity and direct induction of cell death in in vitro experiments, animal models and clinical studies.12 However, the mechanism prevailing in vivo, in the context of simultaneous administration of chemotherapy, remains elusive.13, 14 It is noteworthy that several individual prognostic markers established in patients treated with chemotherapy, such as BCL215 or BCL616 protein expression, were shown to lose their significance in patients receiving rituximab (R)-chemotherapy.

This prompted us to investigate the lymphoma transcriptional profile of patients included in the LNH98.5 trial, to determine which individual genes retain or gain prognostic significance and whether the COO molecular signature remains a valid prognostic marker, in the context of rituximab association with chemotherapy.

Materials and methods

Patients' characteristics and lymphoma samples

Frozen lymphoma samples were obtained retrospectively from 97 newly diagnosed, previously untreated patients with DLBCL. A total of 52 patients had been enrolled in the LNH98.5 clinical trial of the GELA between August 1998 and March 2000, and had signed an informed consent. Twenty-eight of these patients had received R-CHOP and 24 had received CHOP treatment. Details regarding the design, data management and clinical results of this trial have already been published.9, 10 Samples from 45 additional patients who presented the same criteria as those required to enter the LNH98.5 trial (that is, previously untreated DLBCL, 60–80 years of age, no previous history of indolent lymphoma), and who had received either CHOP (n=6) or R-CHOP (n=39) treatment in GELA centers (St. Louis (Paris), Créteil, Lyon, Nancy or Rouen) between March 1998 and July 2004 were added to this study. The Institutional Review Board ‘Comité de Protection de Personnes—Ile de France IX’ authorized the use of these samples and associated clinical data and approved this study. The study comprised a screening set, based mostly on the LNH98.5 trial, and a complementary set, corresponding mostly to additional patients (Supplementary Table 1).

All cases were reviewed by hematopathologists (TJM, JB and PG) and diagnosed as DLBCLs according to the criteria of the WHO classification. Tumor infiltration of the frozen samples was checked by hemalun eosin safran staining of tissue sections. Total RNA was extracted from frozen tissue samples with TRIZOL reagent (Invitrogen, Cergy-Pontoise, France), according to the manufacturer's instructions, and RNA integrity was analyzed by electrophoresis on a 2100 Agilent Bioanalyzer (Agilent, Massy, France) to select samples that could be used in Affymetrix experiments.

Microarray procedures

Microarray analyses were performed using 5 μg of total RNA of each sample as the starting material and 10 μg of cRNA per hybridization (GeneChip Fluidics Station 400). The total RNA was amplified and labeled following the manufacturer's one-cycle target-labeling protocol ( The labeled cRNAs were then hybridized to HU133A Affymetrix GeneChip arrays (Affymetrix, Santa Clara, CA, USA). The chips were scanned with an Affymetrix GeneChip Scanner 3000 and the subsequent images analyzed using GCOS 1.4. Raw feature data were normalized, and log 2 intensity expression summary values for each probe set were calculated using robust multiarray average (RMA package affy V1.4.32).17 Probe sets corresponding to control genes or having a ‘_x_’ annotation were masked yielding a total of 19 787 probe sets for further analyses. The microarray and associated clinical data used in this study have been deposited in the Array Express public database (accession number: E-TABM-346).

Taqman low-density arrays experiments

cDNAs were reverse-transcribed from 4 μg of total RNA in a volume of 100 μl, with the High capacity Archive Kit (Applied Biosystems, Courtaboeuf, France), according to the manufacturer's instructions. A total of 10 μl cDNA was diluted in 90 μl water and 100 μl 2 × Universal Taqman Master Mix (Applied Biosystems) to load the custom microfluidic card (96 assays × 4 per card). Thermocycling was performed with a 7900HT fast real-time PCR system (Applied Biosystems). Analysis was done using the SDS 2.2.1 software (Applied Biosystems), using a manual adjustment for the baseline and threshold levels to ensure identical settings for the 20 cards used. One assay (IGHM-Hs00385741_m1) proved defective and was therefore removed from the final analysis. One assay (CCND2-Hs00153380_m1) displayed poor quality amplification in 11 of the 67 samples analyzed and these data were discarded before the final analysis. Ten wells that showed abnormal amplification curves were also discarded. Cycle thresholds (Ct) superior or equal to 40 were set to 40. Eleven samples run in two separate experiments showed a high reproducibility of the raw Ct values and were used to optimize normalization to the six housekeeping genes (see Supplementary information). The remaining 56 samples were run in one experiment. For each cDNA, the level of expression of each gene (g) was calculated as Δ C t = - ( C t g - ( C t 18S + C t BRF2 + C t FBXO7 + C t MGC 15396 + C t PGK1 + C t TBP ) / 6 ) , and ΔCt values were averaged for duplicates.

Biostatistical analysis

All analyses were performed with the R-system software (v2.5) and Bioconductor (V2.0).18, 19 Comparisons of groups for quantitative data were done by Student's t-tests. Qualitative data were compared using the χ2-test or Fisher exact test, when appropriate. We defined the distance by the correlation coefficient to perform hierarchical clustering with the Ward agglomeration algorithm.

Progression-free survival (PFS) was defined as the time interval between the first chemotherapy cycle and death during induction treatment, disease progression or relapse and excluded late deaths unrelated to lymphoma. OS was calculated from the first chemotherapy cycle until death from any cause. Censored data were analyzed by Cox regression to evaluate the prognostic significance of gene expression levels.20 In the screening set, Cox analyses on individual probe sets, adjusted on IPI score and treatment arm, were performed to evaluate their prognostic values independently of these two factors. Interaction between gene expression levels and treatment arm were also tested to detect a differential effect between the CHOP and R-CHOP patients. In the Taqman low-density array (TLDA) experiments, Cox analyses on ΔCt values adjusted on IPI score and stratified by the analysis set (screening and complementary) were also computed. Interaction between ΔCt values and analysis set were also tested to detect a differential effect between the two sets. The hazard ratios were expressed with a 95% confidence interval and estimated the ratio of risk associated with a twofold increase in gene expression level (that is, a differential expression of 1 unit in log 2 or in ΔCt). A Cox model with L1 penalty function (lasso) and a path following algorithm was used to select the best gene combination among candidate genes.21, 22 C-index scores were computed to quantify the predictive ability of adjusted models to discriminate between subjects who failed and those who did not.23


Search for transcripts associated with PFS in patients treated with R-CHOP

To search for correlations between the tumor transcriptional profile and clinical outcome in patients receiving chemotherapy associated with rituximab, we collected available frozen lymphoma samples from patients included in the GELA LNH98.5 study and from patients treated with the same regimen in GELA centers. RNA from 53 lymphomas were analyzed with Affymetrix HU133A microarrays and constituted a screening set of 30 CHOP and 23 R-CHOP samples. The median age (69 years) and sex ratio (27 males/26 females) of this subset of patients were similar to those of the 399 patients included in the LNH98.5 trial (Supplementary Table 1).10 As reported in this trial, two parameters were strongly correlated with PFS in these patients: a high IPI (each +1 IPI increment, hazard ratio: 1.63 (1.14–1.24); P<0.01) and treatment regimen (CHOP versus R-CHOP, hazard ratio: 2.25 (1.07–4.78); P<0.05). We therefore used Cox models adjusted for IPI and treatment effects to screen this Affymetrix data set in two different ways. First, we selected probe sets that showed an association with PFS in all patients, with a low stringency P-value0.1, assuming that some genes might be of importance for the patient's outcome whether the patient receives rituximab or not. Secondly, we looked for probe sets that showed an association with PFS that was significantly different (P<0.05) in CHOP and R-CHOP patients, assuming that some genes might gain or their lose effects when patients receive rituximab (see Supplementary information for the complete selection strategy). To evaluate the prognostic significance of the candidate genes, we designed a real-time PCR TLDA that comprised Taqman inventoried assays for the 30 genes that were associated with outcome in all patients, 31 genes that showed a differential effect according to treatment arm, 2 genes that showed both effects, 26 additional genes discriminating GC and ABC signature or previously reported to be associated with outcome and six housekeeping genes (Supplementary Table 2). These arrays were used to analyze the transcriptional profile of the 23 R-CHOP samples of the screening set and of 44 complementary R-CHOP samples (Supplementary Table 1).

Genes of which expression is associated with outcome in R-CHOP patients

Cox models were used to analyze the correlation between TLDA gene expression levels and PFS or OS. A large fraction of the 63 candidate genes showed a significant association with PFS (29 genes) and/or OS (25 genes) in the screening group (n=23), as expected from the selection procedure and the high correlation of Affymetrix microarray and TLDA data. Few genes showed a significant association with PFS (two genes: VNN2 and APOBEC3G) and/or OS (three genes: VNN2, LPP and LMO2) in the complementary set (n=44). As some genes showed similar trends in both sets without reaching statistical significance, we analyzed the correlations in the whole R-CHOP group and used interaction tests to exclude the genes showing significantly different effects in the screening and complementary sets (see Supplementary Table 3 for detailed results).

In these 67 patients, 16 transcripts showed a significant association with OS and 6 of these were also significantly associated with PFS (Table 1 and Figure 1). These associations were independent of the IPI (data not shown). To further analyze the potential prognostic significance of these genes, we analyzed a completely independent set of data published by Monti et al.6 concerning a series of 129 patients treated with CHOP or CHOP-like regimen and observed a significant correlation with OS, with similar hazard ratios, for 9 of these 16 genes (Table 1). Quite interestingly, five out of the seven genes selected because of their effect in all patients (ANKRD15, BCL7A, MME, RAFTLIN, SYPL), but only one of the four genes selected because of a differential effect in CHOP and R-CHOP-treated patients (LPP) and three additional genes (LMO2, MYBL, FOXP1) showed a significant correlation with OS in this independent set. Altogether, these data suggest that some genes (ANKRD15, BCL7A, MME, RAFTLIN, SYPL, LMO2, MYBL, LPP, FOXP1) do have prognostic significance both in CHOP and R-CHOP-treated patients, whereas some genes (APOBEC3G, JAK2, RAB33) may have prognostic significance only in patients receiving chemotherapy associated with rituximab.

Table 1 Hazard ratio for survival in 67 R-CHOP patients
Figure 1

Overall survival (OS) and progression-free survival (PFS) according to LPP or APOBEC3G gene expression in 67 patients treated with R-CHOP. (a) OS according to LPP gene expression. (b) PFS according to LPP gene expression. (c) OS according to APOBEC3G gene expression. (d) PFS according to APOBEC3G gene expression. Dashed lines indicate patients with gene expression lower than median expression. Plain lines indicate patients with higher than or equal to median gene expression.

Prognostic significance of COO signature in R-CHOP patients

Wright et al.24 established a predictor for GC and ABC classification of DLBCL lymphomas, based on the expression of 27 genes. Affymetrix HU133A probe sets matched unambiguously to 19 of these genes and were used to classify the samples of the screening set (data not shown). Nineteen assays corresponding to these genes were included in the TLDA. The IgHM Taqman assay proved defective and the expression of one gene (DDB1) showed little variation and did not affect the classification of the samples based on hierarchical clustering (data not shown). The expression of the remaining 17 genes divided the 67 R-CHOP DLBCL samples into 2 clusters: 25 with a GC transcriptional profile and 42 with an ABC profile (Figure 2a). The GC group showed a significantly higher OS than the ABC group (hazard ratio=0.18 (0.04–0.76); P=0.02) (Figure 2b) and a non significant trend for higher PFS (P=0.15) (Figure 2c). The COO distinction remained statistically significant in a multivariate Cox model for OS that incorporated the IPI (GC versus ABC profile, hazard ratio=0.19 (0.04–0.83); P=0.03). Comparison of the OS and PFS curves suggested that after relapse or progression, patients with an ABC-type DLBCL had a high probability of dying (14 deaths out of 17 patients), whereas patients with a GC-type DLBCL could respond to salvage therapy (2 deaths out of 5 patients). Indeed, the COO classification and relapse or progression proved to be a significant and independent risk factor for death in a time-dependent Cox model (GC versus ABC profile, hazard ratio=0.24 (0.05–1); P=0.05; relapse or progression, hazard ratio=29 (9.2–91.7); P=10−8).

Figure 2

Progression-free survival (PFS) and overall survival (OS) according to the COO classification. (a) Hierarchical clustering of the 67 R-CHOP samples according to the expression of 17 genes included in Wright's predictor. The expression level of each gene corresponds to the ΔCt centered on the mean gene ΔCt value, and is depicted according to the color scale shown at the bottom of the heatmap. Yellow boxes indicate samples of the screening set, and orange boxes indicate samples from the additional set. (b) OS according to the COO classification of the samples by hierarchical clustering. (c) PFS according to the COO classification of the samples by hierarchical clustering. Dashed lines indicate patients with an ABC profile. Plain lines indicate patients with a GCB profile.

Selection of a multivariate model with high predictive efficiency

It was quite striking that 5 genes (LMO2, MME, MYBL1, BCL7A and FOXP1) showing significant association with outcome in the 67 R-CHOP patients were known to belong to the set of genes discriminating GC and ABC DLBCLs.4, 24, 25 Among the 11 other genes, 6 were significantly higher in the GC-type DLBCL subgroup and 3 were significantly higher in the ABC-type DLBCL subgroup (Table 2). In a multivariate Cox model adjusted for the COO effect, 2 of the 16 candidate genes (APOBEC3G and RAB33A) remained significantly associated with OS. We used a Cox with L1 penalty (lasso) model adjusted for the IPI effect to test the prognostic significance of the 16 candidate genes and selected a 7-variables model with the path algorithm (IPI+APOBEC3G, LMO2, MME, LPP, FOXP1 and RAB33A) (Supplementary information). We computed C-index values to evaluate the efficiency of these variables, as well as the COO classification and the IPI, to differentiate fatal versus non fatal disease (Table 3 and Supplementary Figure 1). The results showed that four genes of the COO signature had a strong prognostic impact (LMO2, MME, LPP and FOXP1), recapitulating the prognostic significance associated with the COO classification, and that the expression of two other genes (APOBEC3G and RAB33A), selected because of their differential effect in R-CHOP, further influenced the outcome in this series of patients.

Table 2 Differential expressions of the 16 candidate genes according to the COO classification
Table 3 Evaluation of the predictive power of candidate models by the C-index criterion


We analyzed the lymphoma transcriptional profile of patients with DLBCL treated with CHOP or R-CHOP in GELA clinical centers, to determine whether rituximab combined with chemotherapy affects prognostic biomarkers. We used a two-stage screening procedure, which identified 16 genes showing a significant association with OS in 67 R-CHOP-treated patients. The results revealed that the COO classification remained a strong prognostic biomarker in this therapeutic setting. Moreover, we showed that a few genes of the COO (LMO2, LPP, MME and FOXP1) bear most of the prognostic significance of this classification, and that two independent genes (APOBEC3G and RAB33A) could add significant prognostic information in these patients.

Overall, our data are in agreement with previous gene expression profiling studies. In a study that used reverse transcription-PCR to evaluate the expression levels of 36 genes in 66 patients, the only gene that showed a significant correlation with survival in univariate analysis was LMO2,8 indicating that few genes can reach the level of statistical significance in a limited series of patients. Indeed, using different statistical approaches, Segal26 showed that gene expression data delivers only limited predictions of post-therapy DLBCL survival. In this context, the use of corrections for tests multiplicity would exclude all candidate genes, even those already known to bear a prognostic value. Therefore, we chose to jointly analyze all the R-CHOP samples27, 28 and checked the consistency of results by testing interaction terms between the two subsets. Finally, a Cox model with L1 penalty was used to build the predictive IPI plus 6-gene model assessed in R-CHOP samples, and to control for the over-fitting bias (Supplementary information).

Our results show that the GC transcriptional signature retains prognostic significance in patients who received rituximab in addition to chemotherapy, whereas studies based on immunohistochemistry reported that BCL6 protein expression,16 or the COO classification based on BCL6, CD10 and IRF4 protein expression,29 did not remain as significant prognostic indicators in patients treated with immunochemotherapy.30 This discrepancy may be due to technical pitfalls associated with this technique (effect of fixative, optimal cutoff setup, low amount of tumor cells analyzed in tissue microarrays)31 or due to the fact that the immunohistochemical markers used in these studies do not efficiently represent the transcriptional COO classification. In this regard, it is quite interesting to note that LMO2 protein expression, which is a marker of GC origin, was very recently shown to correlate with outcome in both CHOP and R-CHOP-treated patients.32 On the other hand, it is clear that some but not all genes of the COO classification have a strong prognostic significance. In our study, the optimal prognostic model comprised four genes linked to the COO. Three of these genes (LMO2, MME, FOXP1) were previously reported to be prognostic biomarkers in DLBCL patients treated with chemotherapy, although with inconstant results for MME.8, 25, 32, 33, 34, 35, 36 The fourth gene (LPP) was not previously known to be associated with prognosis and its expression and function during B-cell differentiation are unknown. The LPP gene has been discovered as a fusion partner of the HMGA2 gene in lipomas.37 This gene was also involved in a t(3;11) translocation, fused to MLL gene, in a secondary myeloid acute leukemia,38 and in an interstitial deletion of chromosome 3q, fused to BCL6 gene, in a primary central nervous system lymphoma.39 It encodes a LIM-domain containing protein of the zyxin family, recently shown to be a coactivator for the PEA3 Ets transcription factor in breast cancer cells.40 Although these single genes represent markers of a more global transcriptional signature, it is also possible that they recapitulate essential features of GC B-cell differentiation stage, or that their expression makes the cells highly responsive to therapy.

Two other genes (APOBEC3G, RAB33A) showed differential effects in CHOP and R-CHOP patients and demonstrated a significant correlation with OS in R-CHOP patients independently of the COO classification. APOBEC3G encodes a cytoplasmic cytidine deaminase41 that exists in two forms: a low-molecular-mass, enzymatically active protein, which inhibits HIV infection, and a high molecular complex, devoid of enzymatic activity, which localizes to ribonucleoprotein complexes.42 APOBEC3G expression was shown to be regulated by mitogenic and cytokine activation and to depend on MAP kinases as well as JAK/STAT signaling in lymphocytes, monocytes, macrophages and dendritic cells, with cell-type-specific regulation patterns.43 The RAB33A gene is expressed in the brain, lymphocytes and normal melanocytes and encodes a protein belonging to the small-GTPase superfamily, which may be involved in vesicle transport.44 It is interesting to note that RAB33A was reported to be predominantly expressed in the CD8+ T cells and downregulated in the peripheral blood lymphocytes of patients with tuberculosis,45 suggesting a possible link between a low RAB33A expression and a poor host immune response.

In this study, we identified a list of candidate genes that could be useful in predicting outcome in patients with DLBCL treated with R-CHOP, and showed that GC-type DLBCLs indeed have a longer survival than ABC-type DLBCLs in the context of immunochemotherapy. The validation of the prognostic significance of these candidate genes and the elucidation of their mechanism of action will require further studies.

Accession codes




  1. 1

    Harris NL, Jaffe ES, Diebold J, Flandrin G, Muller-Hermelink HK, Vardiman J et al. World Health Organization classification of neoplastic diseases of the hematopoietic and lymphoid tissues: report of the Clinical Advisory Committee meeting-Airlie House, Virginia, November 1997. J Clin Oncol 1999; 17: 3835–3849.

    CAS  Article  PubMed  Google Scholar 

  2. 2

    Shipp MA, Harrington DP, Anderson JR, Armitage JO, Bonadonna G, Brittinger G et al. A predictive model for aggressive non-Hodgkin's lymphoma. The International Non-Hodgkin's Lymphoma Prognostic Factors Project. N Engl J Med 1993; 329: 987–994.

    Article  Google Scholar 

  3. 3

    Lossos IS . Molecular pathogenesis of diffuse large B-cell lymphoma. J Clin Oncol 2005; 23: 6351–6357.

    CAS  Article  PubMed  Google Scholar 

  4. 4

    Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000; 403: 503–511.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  5. 5

    Rosenwald A, Wright G, Chan WC, Connors JM, Campo E, Fisher RI et al. The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 2002; 346: 1937–1947.

    Article  PubMed  PubMed Central  Google Scholar 

  6. 6

    Monti S, Savage KJ, Kutok JL, Feuerhake F, Kurtin P, Mihm M et al. Molecular profiling of diffuse large B-cell lymphoma identifies robust subtypes including one characterized by host inflammatory response. Blood 2005; 105: 1851–1861.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7

    Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC et al. Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med 2002; 8: 68–74.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  8. 8

    Lossos IS, Czerwinski DK, Alizadeh AA, Wechser MA, Tibshirani R, Botstein D et al. Prediction of survival in diffuse large-B-cell lymphoma based on the expression of six genes. N Engl J Med 2004; 350: 1828–1837.

    CAS  Article  PubMed  Google Scholar 

  9. 9

    Coiffier B, Lepage E, Briere J, Herbrecht R, Tilly H, Bouabdallah R et al. CHOP chemotherapy plus rituximab compared with CHOP alone in elderly patients with diffuse large-B-cell lymphoma. N Engl J Med 2002; 346: 235–242.

    CAS  Article  Google Scholar 

  10. 10

    Feugier P, Van Hoof A, Sebban C, Solal-Celigny P, Bouabdallah R, Ferme C et al. Long-term results of the R-CHOP study in the treatment of elderly patients with diffuse large B-cell lymphoma: a study by the Groupe d'Etude des Lymphomes de l'Adulte. J Clin Oncol 2005; 23: 4117–4126.

    CAS  Article  Google Scholar 

  11. 11

    Coiffier B . State-of-the-art therapeutics: diffuse large B-cell lymphoma. J Clin Oncol 2005; 23: 6387–6393.

    CAS  Article  Google Scholar 

  12. 12

    Smith MR . Rituximab (monoclonal anti-CD20 antibody): mechanisms of action and resistance. Oncogene 2003; 22: 7359–7368.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  13. 13

    Cartron G, Watier H, Golay J, Solal-Celigny P . From the bench to the bedside: ways to improve rituximab efficacy. Blood 2004; 104: 2635–2642.

    CAS  Article  Google Scholar 

  14. 14

    Friedberg JW . Unique toxicities and resistance mechanisms associated with monoclonal antibody therapy. In: Gerwitz AM, Winter JN, Zuckerman K (eds). Hematology (Am Soc Hematol Educ Program). American Society of Hematology Inc.: Washington DC, 2005, pp 329–334.

    Google Scholar 

  15. 15

    Mounier N, Briere J, Gisselbrecht C, Emile JF, Lederlin P, Sebban C et al. Rituximab plus CHOP (R-CHOP) overcomes bcl-2—associated resistance to chemotherapy in elderly patients with diffuse large B-cell lymphoma (DLBCL). Blood 2003; 101: 4279–4284.

    CAS  Article  PubMed  Google Scholar 

  16. 16

    Winter JN, Weller EA, Horning SJ, Krajewska M, Variakojis D, Habermann TM et al. Prognostic significance of Bcl-6 protein expression in DLBCL treated with CHOP or R-CHOP: a prospective correlative study. Blood 2006; 107: 4207–4213.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  17. 17

    Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics (Oxford, England) 2003; 4: 249–264.

    Article  Google Scholar 

  18. 18

    Gentleman R, Ihaka R, the R Development Core Team. R: A Language and Environment for Statistical Computing—Reference Manual. R Foundation for Statistical Computing: Vienna, Austria, 2007, 1–1507.

    Google Scholar 

  19. 19

    Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004; 5: R80.

    Article  PubMed  PubMed Central  Google Scholar 

  20. 20

    Kalbfleisch JD, Prentice RL . Relative risk (Cox) regression models. In: Kalbfleisch JD, Prentice RL (eds). The Statistical Analysis of Failure Time Data, 2nd edn. John Wiley & Sons Inc.: Hoboken, NJ, 2002, pp 95–147.

    Google Scholar 

  21. 21

    Tibshirani R . Regression shrinkage and selection via the lasso. J R Statist Soc B 1996; 58: 267–288.

    Google Scholar 

  22. 22

    Park MY, Hastie T . L1-regularization path algorithm for generalized linear models. J R Statist Soc B 2007; 69: 659–677.

    Article  Google Scholar 

  23. 23

    Harrell Jr FE, Lee KL, Mark DB . Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 1996; 15: 361–387.

    Article  PubMed  Google Scholar 

  24. 24

    Wright G, Tan B, Rosenwald A, Hurt EH, Wiestner A, Staudt LM . A gene expression-based method to diagnose clinically distinct subgroups of diffuse large B cell lymphoma. Proc Natl Acad Sci USA 2003; 100: 9991–9996.

    CAS  Article  PubMed  Google Scholar 

  25. 25

    Barrans SL, Fenton JA, Banham A, Owen RG, Jack AS . Strong expression of FOXP1 identifies a distinct subset of diffuse large B-cell lymphoma (DLBCL) patients with poor outcome. Blood 2004; 104: 2933–2935.

    CAS  Article  PubMed  Google Scholar 

  26. 26

    Segal MR . Microarray gene expression data with linked survival phenotypes: diffuse large-B-cell lymphoma revisited. Biostatistics (Oxford, England) 2006; 7: 268–285.

    Article  Google Scholar 

  27. 27

    Skol AD, Scott LJ, Abecasis GR, Boehnke M . Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association studies. Nat Genet 2006; 38: 209–213.

    CAS  Article  PubMed  Google Scholar 

  28. 28

    Zehetmayer S, Bauer P, Posch M . Two-stage designs for experiments with a large number of hypotheses. Bioinformatics 2005; 21: 3771–3777.

    CAS  Article  PubMed  Google Scholar 

  29. 29

    Hans CP, Weisenburger DD, Greiner TC, Gascoyne RD, Delabie J, Ott G et al. Confirmation of the molecular classification of diffuse large B-cell lymphoma by immunohistochemistry using a tissue microarray. Blood 2004; 103: 275–282.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  30. 30

    Nyman H, Adde M, Karjalainen-Lindsberg ML, Taskinen M, Berglund M, Amini RM et al. Prognostic impact of immunohistochemically defined germinal center phenotype in diffuse large B-cell lymphoma patients treated with immunochemotherapy. Blood 2007; 109: 4930–4935.

    CAS  Article  PubMed  Google Scholar 

  31. 31

    de Jong D, Rosenwald A, Chhanabhai M, Gaulard P, Klapper W, Lee A et al. Immunohistochemical prognostic markers in diffuse large B-cell lymphoma: validation of tissue microarray as a prerequisite for broad clinical applications—a study from the Lunenburg Lymphoma Biomarker Consortium. J Clin Oncol 2007; 25: 805–812.

    Article  PubMed  Google Scholar 

  32. 32

    Natkunam Y, Farinha P, Hsi ED, Hans CP, Tibshirani R, Sehn LH et al. LMO2 protein expression predicts survival in patients with diffuse large B-cell lymphoma treated with anthracycline-based chemotherapy with and without rituximab. J Clin Oncol 2008; 26: 447–454.

    CAS  Article  PubMed  Google Scholar 

  33. 33

    Banham AH, Connors JM, Brown PJ, Cordell JL, Ott G, Sreenivasan G et al. Expression of the FOXP1 transcription factor is strongly associated with inferior survival in patients with diffuse large B-cell lymphoma. Clin Cancer Res 2005; 11: 1065–1072.

    CAS  PubMed  Google Scholar 

  34. 34

    Chang CC, McClintock S, Cleveland RP, Trzpuc T, Vesole DH, Logan B et al. Immunohistochemical expression patterns of germinal center and activation B-cell markers correlate with prognosis in diffuse large B-cell lymphoma. Am J Surg Pathol 2004; 28: 464–470.

    Article  PubMed  Google Scholar 

  35. 35

    Fabiani B, Delmer A, Lepage E, Guettier C, Petrella T, Briere J et al. CD10 expression in diffuse large B-cell lymphomas does not influence survival. Virchows Arch 2004; 445: 545–551.

    CAS  Article  PubMed  Google Scholar 

  36. 36

    van Imhoff GW, Boerma EJ, van der Holt B, Schuuring E, Verdonck LF, Kluin-Nelemans HC et al. Prognostic impact of germinal center-associated proteins and chromosomal breakpoints in poor-risk diffuse large B-cell lymphoma. J Clin Oncol 2006; 24: 4135–4142.

    CAS  Article  PubMed  Google Scholar 

  37. 37

    Petit MM, Mols R, Schoenmakers EF, Mandahl N, Van de Ven WJ . LPP, the preferred fusion partner gene of HMGIC in lipomas, is a novel member of the LIM protein gene family. Genomics 1996; 36: 118–129.

    CAS  Article  PubMed  Google Scholar 

  38. 38

    Daheron L, Veinstein A, Brizard F, Drabkin H, Lacotte L, Guilhot F et al. Human LPP gene is fused to MLL in a secondary acute leukemia with a t(3;11) (q28;q23). Genes Chromosomes Cancer 2001; 31: 382–389.

    CAS  Article  PubMed  Google Scholar 

  39. 39

    Schwindt H, Akasaka T, Zuhlke-Jenisch R, Hans V, Schaller C, Klapper W et al. Chromosomal translocations fusing the BCL6 gene to different partner loci are recurrent in primary central nervous system lymphoma and may be associated with aberrant somatic hypermutation or defective class switch recombination. J Neuropathol Exp Neurol 2006; 65: 776–782.

    CAS  Article  PubMed  Google Scholar 

  40. 40

    Guo B, Sallis RE, Greenall A, Petit MM, Jansen E, Young L et al. The LIM domain protein LPP is a coactivator for the ETS domain transcription factor PEA3. Mol Cell Biol 2006; 26: 4529–4538.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  41. 41

    Takaori-Kondo A . APOBEC family proteins: novel antiviral innate immunity. Int J Hematol 2006; 83: 213–216.

    CAS  Article  PubMed  Google Scholar 

  42. 42

    Gallois-Montbrun S, Kramer B, Swanson CM, Byers H, Lynham S, Ward M et al. Antiviral protein APOBEC3G localizes to ribonucleoprotein complexes found in P bodies and stress granules. J Virol 2007; 81: 2165–2178.

    CAS  Article  PubMed  Google Scholar 

  43. 43

    Stopak KS, Chiu YL, Kropp J, Grant RM, Greene WC . Distinct patterns of cytokine regulation of APOBEC3G expression and activity in primary lymphocytes, macrophages, and dendritic cells. J Biol Chem 2007; 282: 3539–3546.

    CAS  Article  PubMed  Google Scholar 

  44. 44

    Cheng E, Trombetta SE, Kovacs D, Beech RD, Ariyan S, Reyes-Mugica M et al. Rab33A: characterization, expression, and suppression by epigenetic modification. J Invest Dermatol 2006; 126: 2257–2271.

    CAS  Article  PubMed  Google Scholar 

  45. 45

    Jacobsen M, Repsilber D, Gutschmidt A, Neher A, Feldmann K, Mollenkopf HJ et al. Ras-associated small GTPase 33A, a novel T cell factor, is down-regulated in patients with tuberculosis. J Infect Dis 2005; 192: 1211–1218.

    CAS  Article  PubMed  Google Scholar 

Download references


We are grateful to Fabien Petel (Ligue Nationale Contre le Cancer) for the management of the Affymetrix and annotation databases and submission of the same to the Array Express. We thank E Jacquet, H Lévaique, E Garrido (CNRS, Gif sur Yvette, France) and E Côme for the realization of the TLDA experiments; A Allain and N Nio for their help in clinical data management. We are indebted to the pathologists and clinicians of the GELA who contributed pathological specimen and clinical data. This study was supported by grants of the Ligue Nationale Contre le Cancer (programme Carte d'Identité des Tumeurs), the Programme Hospitalier de Recherche Clinique (AOM 03060) and an unrestricted grant from Roche.

Author information




Corresponding author

Correspondence to K Leroy.

Additional information

Supplementary Information accompanies the paper on the Leukemia website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Jais, JP., Haioun, C., Molina, T. et al. The expression of 16 genes related to the cell of origin and immune response predicts survival in elderly patients with diffuse large B-cell lymphoma treated with CHOP and rituximab. Leukemia 22, 1917–1924 (2008).

Download citation


  • diffuse B-cell lymphoma
  • rituximab
  • gene expression
  • prognosis

Further reading


Quick links