Introduction

B-chronic lymphocytic leukemia (CLL) is a heterogeneous disease with regard to prognosis and clinical course. Rai and Binet staging systems have significantly contributed to identification of major prognostic subgroups, although they fail to predict which patients among the good prognostic groups (Rai 0 and Binet A) will shift into progressive disease. Phenotypically, CLL cells are monoclonal CD5⁺ B cells corresponding to mantle zone naive B cells and would be expected to express unmutated IgV_H genes.¹ In fact, more than half of CLL cases harbor somatic mutations of V_H genes, as if they had matured in a lymphoid follicle.² Unmutated V_H genes are associated with a markedly poorer prognosis even within the stage A group^3,4 and mutational status appears as the most relevant prognostic parameter.

Since the ability to sequence IgV_H genes is not available in most laboratories, a valid and easily performed surrogate assay is desirable. Damle et al⁵ suggested that CD38 determination might be a useful alternative. However, correlation of Ig mutational status with CD38 expression remains controversial.^6,7,8

The absence of V_H gene mutations is strongly linked to advanced stages of the disease, suggesting that these cells may show a higher proliferative potential than cells with somatic mutations. In this report, based on the clinical and biological data of 58 previously untreated CLL patients, we tried to define whether proliferation markers – serum thymidine kinase (s-TK),^9,10 2 microglobulin, LDH – or cell cycle regulatory protein expression – p27^Kip1, cyclin D^11,16 – could help predict mutational status of IgV_H genes.

Thymidine kinase (TK), adenosine triphosphate (ATP) thymidine 5'-phosphotransferase is a cellular enzyme known to be involved in a 'salvage pathway' for DNA synthesis. The s-TK activity in CLL patients correlates with the proliferative activity of CLL cells. In addition, previous work in CLL has shown that s-TK adds independent prognostic information to the definitions of smoldering and non-smoldering CLL in Binet stage A.¹⁰

The cyclin-dependent kinase inhibitor p27^Kip1 plays a major role in G0/G1 cell cycle arrest. We have shown previously that p27^Kip1 is an adverse prognostic factor in B-CLL, and that its expression was linked to the defect in apoptosis.¹¹ In a previous study, we observed that both p27^Kip1 and s-TK levels were strongly associated with lymphocyte doubling time.¹²

The aim of this study was to identify predictive parameters of the mutational status. Our results demonstrate a straight relationship between high levels of s-TK and an unmutated profile, whereas low levels of s-TK were associated with a mutated profile of these V_H genes. Consequently, we aimed, through cluster analysis, at using these biological parameters reflecting the mechanisms of cell accumulation for the identification of subgroups of patients. Our results show that these parameters allow segregation into two groups of patients on their proliferative potential that will experience a significantly different outcome.

Materials and methods

Peripheral blood samples from 58 previously untreated CLL patients were obtained after informed consent. The diagnosis of B-CLL relied on characteristic cytological features of mature lymphocytes, and a typical phenotype (CD5⁺, CD23⁺, low expression of surface immunoglobulins and of CD79b). Blood mononuclear cells were isolated by Ficoll–Hypaque density gradient and monocytes removed by plastic adherence.

p27^Kip1 and cyclin D2 expression

These were assessed by Western blot. Briefly, protein extracts were electrophoresed in SDS polyacrylamide gels, transferred on to Immobilon membranes and incubated with specific anti-p27^Kip1 (N-20; Santa-Cruz Laboratories, CA, USA) and cyclin D2 (G132-43; Pharmingen, USA) antibodies and signals were measured by an enhanced chemoluminescent system (ECL, Amersham, UK). For each sample, the p27^Kip1 value was expressed as the ratio of the signal obtained to that of a positive control.¹¹

Serum thymidine kinase activity

The s-TK assay was performed with a commercially available radioenzyme assay (Prolifigen; Sangtec Medical, Bromma, Sweden). The TK activity is expressed in units per liter. For TK values, the intra-assay variability was between 5.4% and 7.5%, and the inter-assay variability between 5.7% and 8.3%.

Cloning and sequencing of V_H immunoglobulin genes

RNA was isolated from 10⁶ PBMC. cDNA synthesis and PCR amplification using specific V_H5' primers and consensus J_H3' primers were perfomed as previously reported.¹³ PCR products were introduced into a pGEM-T Easy cloning vector (Promega Life Sciences, Charbonnieres, France) and sequenced by the dideoxy chain termination method from multiple independent clones. Nucleotide sequence data were analyzed and compared by Gene Runner software and Vbase plot of Genebank-EMBL databases. The complete sequences have been contributed to GenBank.

Statistical methods

Associations of IgV_H genes mutational status with clinical and biological variables such as surrogate markers of proliferation and cell cycle protein expression markers were assessed using Fisher's exact test for categorical variables and Wilcoxon rank-sum test for quantitative continuous variables. A set of factors independently associated with mutational status was identified by means of generalized additive modeling, followed by a multiple logistic regression model, with a backward stepwise procedure for variable selection. All tests were two-sided, with a 5% significance level. Additionally, cluster analysis was used to identify subgroups of patients with similar characteristics. For clustering, a nonhierarchical method¹⁴ based on a modified K-means algorithm was used. The optimal number of clusters was assessed through the pseudo-F statistic. All clinical and biological variables except Binet staging were considered for cluster construction. Reproducibility of the obtained classification was tested using a nonparametric bootstrap procedure¹⁵ based on 999 independent samples randomly constituted from the original observations. The clustering method was then applied to each of these samples, and misclassification rates as compared to the original classification were then derived. All analyses were performed on SAS v8.1 software (SAS Institute, Cary, NC, USA).

Results

Analysis of clinical and biological parameters of mutated and unmutated cases

The following data were collected from 58 B-CLL cases: age (median 60 years), sex (F: 31%), presence of lymph nodes (41%), spleen enlargement (18%), Binet stage (A: 53%, B: 30%, C: 17%), lymphocyte absolute count and doubling time (DT), sTK level, serum 2 microglobulin, LDH, expression of cyclin-dependent kinase inhibitor p27^Kip1 and cyclin D2, and the mutational status of IgV_H genes and summarized in Table 1. CD38 expression was not available for all patients of this retrospective study. In 33 cases (57%), more than 2% somatic mutations of IgV_H gene were observed, which allowed these cases to be considered as 'mutated' and the 25 cases with less than 2% mutations as 'unmutated'. We compared the clinical and biological parameters of the 33 cases with somatic mutations with the 25 cases with unmutated IgV_H genes. Twenty-one of 31 stage A patients harbored somatic mutations but stage B and C patients appeared distributed in both groups (12 in the mutated group and 15 in the unmutated group). Importance of tumoral mass, as reflected by presence of adenopathy or splenomegaly, was not significantly different when considering only the stage B and C patients of each group. Conversely, most of the cases with unmutated IgV_H genes had a short lymphocyte doubling time as compared with the mutated cases (82% vs 27%) (Table 1).

Table 1 - Clinical and biological characteristics of the 58 patients and following their mutational status. The only significant differences between groups were observed for s-TK levels (P = 0.0019, Wilcoxon rank-sum test) and doubling time <1 yr (P = 0.0025, Fisher's exact test).

Full table (8K)

Follow-up was available for 38 patients. Overall survival was significantly different according to mutational status, with a median survival of 84 months for unmutated cases (n = 16), and not reached for mutated cases (n = 22) (log-rank test, P = 0.025) (Figure 1).

Figure 1.

Overall survival of patients regarding mutational status. Median survival was 84 months for unmutated cases, and not reached for mutated cases (log-rank test, P = 0.025).

Full figure and legend (12K)

s-TK is the only independent predictor of mutational status

We tried to define whether cell cycle protein expresssion (p27^kip1, cyclin D2) or surrogate markers of proliferation (s-TK, 2 microglobulin, LDH) could help predict the mutational status of IgVH genes. The expression of p27^kip1 was distributed over a wide range in both groups and did not differ significantly between the two groups. Cyclin D2 was highly expressed in B-CLL, and 80% of mutated cases overexpressed cyclin D2, whereas less than 50% of unmutated cases expressed detectable levels of cyclin D2 protein.

Neither 2 microglobulin nor LDH were found to discriminate between mutated and unmutated patients (data not shown). The median value of s-TK in the mutated group was 7.4 U/l (min 2.9–max 74.0) vs 29 U/l (4.8–65) in the unmutated group. The variable selection procedure in the multiple logistic regression model yielded s-TK level as the only biological parameter independently associated with the presence of unmutated IgV_H genes. A cut-off value of 15 U/l was found well suited, using generalized additive modelling. A higher sTK level (>15 U/l) proved then to be a strong independent predictor of the presence of unmutated Ig-V genes (odds ratio: 31.7, 95% confidence interval: 7.1–142.0, P < 0.0001) (Figure 2). With such a cut-off value for sTK, and according to the prevalence of mutated cases in our cohort, predictive value of a higher sTK level for the absence of mutation was 86.4% and predictive value of lower sTK level for the presence of mutation was 83.3%.

Figure 2.

s-TK level is related to the percentage of VH homology. Dashed lines represent the cut-offs considered (15 U/l defining higher s-TK levels and 98% homology defining unmutated IgV_H genes).

Full figure and legend (19K)

Cluster analysis allows identification of subgroups

We aimed at identifying subgroups of patients with similar disease characteristics, taking account of the mutational status. Clusters with similar characteristics were constructed using a statistical analysis algorithm (K-means). When considering all the clinical and biological data, except Binet staging, the most representative number of clusters (as determined by pseudo-F statistics) proved to be two, including 25 and 13 cases, respectively. The bootstrap procedure used to assess the reliability of the classification confirmed the selection of two clusters in 75.3% of simulated samples, while three, four and five clusters were selected in 12.5%, 5.3% and 2.% of cases, respectively. The procedure failed to select an optimal number of clusters in 4% of cases. For 752 out of 999 samples (75.3%) where two clusters were considered as optimal, the global misclassification rate was 15.6% (13.3% for cluster 1 and 16.9% for cluster 2), illustrating a good reproducibility of the classification, owing to the moderate sample size. Moreover, we applied the clustering algorithm with all previously used parameters except the mutational status of IgVH genes. This procedure also yielded two clusters as optimal and no significant difference between the two classifications was found using a McNemar test.

The clinical and biological parameters of these two clusters are summarized in Table 2, and depict the existence of two populations of patients with distinct features regarding proliferation. Cluster 1 included patients with parameters mostly reflecting slow progressing disease: low s-TK levels (96%), mutated IgVH genes (80%) and slow lymphocyte doubling time (80%). Conversely, in cluster 2, patients displayed high s-TK levels (92%), unmutated IgV_H genes (84%) and short lymphocyte doubling time (92%). In cluster 2, cyclin D2, which was shown to be highly expressed in non-proliferative cells, was undetectable in 54% of cases.^16,17 Moreover, five patients in cluster 2 had a very low p27^kip1 expression within the range observed in cells with proliferative potential.¹⁸

Table 2 - Clinical and biological characteristics of the two clusters.

Full table (6K)

Expectedly, patients within these two clusters had a very different clinical course and outcome. Overall survival was significantly different between the two clusters, with a median survival of 84 months for cluster 2, and not reached for cluster 1 (log rank test, P = 0.004) (Figure 3).

Figure 3.

Overall survival of patients according to clusters 1 and 2. Cluster 2 has a median survival of 84 months while it is not reached for cluster 1 (log-rank test, P = 0.004).

Full figure and legend (11K)

Repartition of the Binet stages among the two clusters

According to Binet classification, 64% of patients in cluster 1 were in stage A, while 77% of patients in cluster 2 were in advanced stage (46% B and 31% C). Most of the stage A patients (16/19) fell into cluster 1. Among these 16 patients, only three became progressive after 1, 3 and 4 years, respectively, and were treated. All stage A patients within cluster 1 are still alive. Three stage A patients were included in cluster 2, displaying a high s-TK and unmutated IgV_H genes. The three patients progressed and were treated, and two of them died of refractory disease.

If considering only the B and C stage patients (n = 19), they appeared to be distributed in both groups (cluster 1 = 9, cluster 2 = 10). All patients were treated. Out of nine patients in cluster 1, two died, one of Richter's syndrome, and the other one of cardiac failure. Out of 10 patients in cluster 2, four died of progression and one was lost to follow-up.

Stage B and C patients still significantly differed by their mutational status (89% vs 9% mutated cases, respectively) and s-TK level >15 U/l (11% vs 100%) between the two clusters (Table 3).

Table 3 - Differences in mutational status and s-TK levels in stages B and C patients in clusters 1 and 2.

Full table (3K)

Discussion

Mutational status of the immunoglobulin genes recently proved to be a strong pronostic parameter identified in B-CLL, even among cases with a good prognosis (Binet stage A patients). The determination of the presence of somatic mutations is a difficult, expensive and time-consuming technique which will not be readily available for most patients within the next years. Therefore, the identification of predictive parameters for mutational status in B-CLL is warranted. CD38 has been previously noted as a potential candidate, but with controversial results. CD38 appears as a prognostic factor by itself, but is probably not related to the mutational status.⁶ Moreover, its determination has encountered technical problems, related to reproducibility among laboratories, and the threshold is still under discussion. CD38 expression varies during disease evolution, which contrasts with the invariability of the mutational status.

We focused our study on previously identified prognostic parameters related to proliferation and apoptosis. Defective apoptosis is the hallmark of CLL and it is supposed to play a major role in the pathogeny of the disease. On the other hand, proliferation is low as compared to other lymphoproliferative disorders, and the contribution of proliferation in disease induction remains unclear as yet. We considered commonly used biological parameters such as LDH, 2 microglobulin and lymphocyte doubling time. We also used two other parameters that proved of interest in our previous studies. We have shown previously that the cyclin-dependent kinase inhibitor p27^kip1 was highly expressed in CLL as compared with other lymphoproliferative disorders, and that its high expression was an adverse prognostic factor in B-CLL.¹¹ Serum TK levels have also been shown to correlate with disease progression in various lymphoid malignancies and to be able to identify patients at risk of progression in early CLL.¹⁰ s-TK level seems to reflect the proliferative activity of the tumor, whereas p27^kip1 is involved in cell cycle G0/G1 arrest and its expression is linked to the defect in apoptosis. p27^kip1 and s-TK levels were both strongly associated with either doubling time or disease progression (both P < 0.0001) in our preliminary study on 72 patients.¹²

In the present study, we found that s-TK, which reflects the proliferative potential of the malignant clone, was an independent predictor of mutational status and the only one among all the studied parameters. The other studied parameters reflecting the tumor burden such as LDH and 2 microglobulin lacked sensitivity. Cyclin D2 expresion also shed some interesting light. In previous work, we reported that cyclin D2 was highly expressed in most CLL, as compared with other lymphoproliferative disorders.¹⁶ Cyclin D2 in other models, has also been shown to reflect a cellular quiescent state.¹⁹ We found that cyclin D2 was detectable in 80% of mutated cases, and expressed in less than 50% of unmutated cases. Conversely, p27^kip1 expression was not found to be associated with the existence of somatic mutations suggesting that mutational status might not be related to defective apoptosis. We even observed that in five unmutated cases, p27^kip1 expression was very low, within the range observed in more proliferative lymphomas. Therefore, it appears that the absence of somatic mutations is predominantly reflected by the presence of a proliferative potential.

Biological prognostic parameters are increasingly assessed in routine analysis, but decisions to treat still rely exclusively on clinical staging dividing patients into three categories based on the extent of nodal involvement and the existence of cytopenias. These groups proved to have a significantly distinct outcome, and treatment recommendations have been edited after large long-term trials.^20,21 Treatment can be appropriately deferred for stage A, whereas early treatment is commonly indicated for stage B and C patients. We searched for a model that could help to design a classification taking account of the new biological prognostic factors.

Cluster analysis is increasingly applied in biomedical research, and notably in analyzing data arising from microarray experiments. We tried to define subgroups of patients with similar disease patterns. The interest of clusterization resides in the segregation of groups while ignoring the outcome of patients, conversely to usual approaches to prognostic classification. Such a classification is meant to be nosographic, but interestingly, the two clusters turned out to exhibit very discriminant properties in survival analysis (Figure 3).

As Binet staging was purposely not included in the parameters used for clusterization, we were able to analyze the repartition of the different stages within the two clusters. Expectedly, most of the A stage (16/19) patients fell into cluster 1, with good prognostic features. Stage B and C cases were distributed in both clusters and had a very different course. This raises the question of whether all B and C stages should be considered for the same treatment. Hematologists have long suspected that a slowly progressing B stage with small lymph nodes will experience a different course from that of a stage B with bulky disease. Although the presence of anemia and/or thrombocytopenia is associated with poor prognosis, it remains to be established whether the mechanism related to anemia and/or thrombopenia (autoimmune or bone marrow insufficiency by massive replacement) could play a role in stage C prognosis. This is particularly important, since most published trials dealing with stage C patients, have systematically excluded for therapeutic reasons, patients initially displaying autoimmune anemia or thrombocytopenia. This heterogeneity, presently unaccounted for, is likely to undermine the power of on-going clinical trials.²² It is becoming apparent that these subgroups of patients may have specific molecular signatures. Therefore, a stratification on biological parameters reflecting the molecular features is warranted. Further studies on a larger cohort of patients are on-going to answer this question.

In conclusion, mutational status is strongly correlated with the presence of a proliferative potential and may be predicted by s-TK level. These results are consistent with the recent results of gene expression profiling in CLL showing that upregulation of genes modulated during mitogenic stimulation was likely to discriminate two subgroups of CLL, and correlated with absence of the IgV_H mutation.^23,24 s-TK level is determined through an easily performed and standardized technique and should be routinely assessed in CLL.

References

1. Schroeder, HWJr & Dighiero, GThe pathogenesis of chronic lymphocytic leukemia: analysis of the antibody repertoire. Immunol Today, (1994). 15, 288–294. | Article | PubMed | ChemPort |
2. Fais, F, Ghiotto, F, Hashimoto, S, Sellars, B, Valetto, A, Allen, SL, Schulman, P, Vinciguerra, VP, Rai, K, Rassenti, LZ, Kipps, TJ, Dighiero, G, Schroeder, HWJr, Ferrarini, M & Chiorazzi, NChronic lymphocytic leukemia B cells express restricted sets of mutated and unmutated antigen receptors. J Clin Invest, (1998). 102, 1515–1525. | PubMed | ISI | ChemPort |
3. Hamblin, TJ, Davis, Z, Gardiner, A, Oscier, DG & Stevenson, FKUnmutated Ig V(H) genes are associated with a more aggressive form of chronic lymphocytic leukemia. Blood, (1999). 94, 1848–1854. | PubMed | ISI | ChemPort |
4. Maloum, K, Davi, F, Merle-Beral, H, Pritsch, O, Magnac, C, Vuillier, F, Dighiero, G, Troussard, X, Mauro, FF & Benichou, JExpression of unmutated VH genes is a detrimental prognostic factor in chronic lymphocytic leukemia. Blood, (2000). 96, 377–379. | PubMed | ISI | ChemPort |
5. Damle, RN, Wasil, T, Fais, F, Ghiotto, F, Valetto, A, Allen, SL, Buchbinder, A, Budman, D, Dittmar, K, Kolitz, J, Lichtman, SM, Schulman, P, Vinciguerra, VP, Rai, KR, Ferrarini, M & Chiorazzi, NIg V gene mutation status and CD38 expression as novel prognostic indicators in chronic lymphocytic leukemia. Blood, (1999). 94, 1840–1847. | PubMed | ISI | ChemPort |
6. Hamblin, TJ, Orchard, JA, Gardiner, A, Oscier, DG, Davis, Z & Stevenson, FKImmunoglobulin V genes and CD38 expression in CLL. Blood, (2000). 95, 2455–2457. | PubMed | ISI | ChemPort |
7. Matrai, Z, Lin, K, Dennis, M, Sherrington, P, Zuzel, M, Pettitt, AR & Cawley, JCCD38 expression and Ig VH gene mutation in B-cell chronic lymphocytic leukemia. Blood, (2001). 97, 1902–1903. | Article | PubMed | ISI | ChemPort |
8. Thunberg, U, Johnson, A, Roos, G, Thorn, I, Tobin, G, Sallstrom, J, Sundstrom, C & Rosenquist, RCD38 expression is a poor predictor for VH gene mutational status and prognosis in chronic lymphocytic leukemia. Blood, (2001). 97, 1892–1894. | Article | PubMed | ISI | ChemPort |
9. Di Raimondo, F, Giustolisi, R, Lerner, S, Cacciola, E, O'Brien, S, Kantarjian, H & Keating, MJRetrospective study of the prognostic role of serum thymidine kinase level in CLL patients with active disease treated with fludarabine. Ann Oncol, (2001). 12, 621–625.
10. Hallek, M, Langenmayer, I, Nerl, C, Knauf, W, Dietzfelbinger, H, Adorf, D, Ostwald, M, Busch, R, Kuhn-Hallek, I, Thiel, E & Emmerich, BElevated serum thymidine kinase levels identify a subgroup at high risk of disease progression in early, nonsmoldering chronic lymphocytic leukemia. Blood, (1999). 93, 1732–1737. | PubMed | ISI | ChemPort |
11. Vrhovac, R, Delmer, A, Tang, R, Marie, JP, Zittoun, R & Ajchenbaum-Cymbalista, FPrognostic significance of the cell cycle inhibitor p27Kip1 in chronic B-cell lymphocytic leukemia. Blood, (1998). 91, 4694–4700. | PubMed | ISI | ChemPort |
12. Levy, V, Nataf, J, Zompi, S, Thomas, F, Brouzes, C, Casadevall, N, Hallek, M, Delmer, A & Ajchenbaum-Cymbalista, FPrognostic significance of serum thymidine kinase levels and p27KIP1 expression in chronic lymphocytic leukemia. Blood, (1999). 94, 321a
13. Pritsch, O, Troussard, X, Magnac, C, Mauro, FR, Davi, F, Payelle-Brogard, B, Dumas, G, Pulik, M, Clerget, F, Mandelli, F, Chiorazzi, N, Schroeder, HWJr, Leporrier, M & Dighiero, GVH gene usage by family members affected with chronic lymphocytic leukaemia. Br J Haematol, (1999). 107, 616–624. | PubMed |
14. Everitt, BSCluster analysis of subjects. Non hierarchical methods. In: Armitage P, Colton T (eds). Encyclopedia of Biostatistics, John Wiley and Sons: Chichester (1998). 713–731.
15. Efron, B & Tibshirani, RAn Introduction to Bootstrap, Chapman & Hall: London (1993).
16. Delmer, A, Ajchenbaum-Cymbalista, F, Tang, R, Ramond, S, Faussat, AM, Marie, JP & Zittoun, ROverexpression of cyclin D2 in chronic B-cell malignancies. Blood, (1995). 85, 2870–2876. | PubMed | ISI | ChemPort |
17. Solvason, N, Wu, WW, Parry, D, Mahony, D, Lam, EW, Glassford, J, Klaus, GG, Sicinski, P, Weinberg, R, Liu, YJ, Howard, M & Lees, ECyclin D2 is essential for BCR-mediated proliferation and CD5 B cell development. Int Immunol, (2000). 12, 631–638. | Article | PubMed | ISI | ChemPort |
18. Sanchez-Beato, M, Saez, AI, Martinez-Montero, JC, Sol Mateo, M, Sanchez-Verde, L, Villuendas, R, Troncone, G & Piris, MACyclin-dependent kinase inhibitor p27KIP1 in lymphoid tissue: p27KIP1 expression is inversely proportional to the proliferative index. Am J Pathol, (1997). 151, 151–160. | PubMed | ISI | ChemPort |
19. Meyyappan, M, Wong, H, Hull, C & Riabowol, KTIncreased expression of cyclin D2 during multiple states of growth arrest in primary and established cells. Mol Cell Biol, (1998). 18, 3163–3172. | PubMed | ISI | ChemPort |
20. Dighiero, G, Maloum, K, Desablens, B, Cazin, B, Navarro, M, Leblay, R, Leporrier, M, Jaubert, J, Lepeu, G, Dreyfus, B, Binet, JL & Travade, PChlorambucil in indolent chronic lymphocytic leukemia. French Cooperative Group on Chronic Lymphocytic Leukemia. N Engl J Med, (1998). 338, 1506–1514. | Article | PubMed | ISI | ChemPort |
21. Leporrier, M, Chevret, S, Cazin, B, Boudjerra, N, Feugier, P, Desablens, B, Rapp, MJ, Jaubert, J, Autrand, C, Divine, M, Dreyfus, B, Maloum, K, Travade, P, Dighiero, G, Binet, JL & Chastang, CRandomized comparison of fludarabine, CAP, and ChOP in 938 previously untreated stage B and C chronic lymphocytic leukemia patients. Blood, (2001). 98, 2319–2325. | Article | PubMed | ISI | ChemPort |
22. Betensky, R, Louis, N & Cairncross, JInfluence of unrecognized molecular heterogeneity on randomized clinical trials. J Clin Oncol, (2002). 20, 2495–2499. | Article | PubMed | ISI |
23. Rosenwald, A, Alizadeh, AA, Widhopf, G, Simon, R, Davis, RE, Yu, X, Yang, L, Pickeral, OK, Rassenti, LZ, Powell, J, Botstein, D, Byrd, JC, Grever, MR, Cheson, BD, Chiorazzi, N, Wilson, WH, Kipps, TJ, Brown, PO & Staudt, LMRelation of gene expression phenotype to immunoglobulin mutation genotype in B cell chronic lymphocytic leukemia. J Exp Med, (2001). 194, 1639–1647. | Article | PubMed | ISI | ChemPort |
24. Klein, U, Tu, Y, Stolovitzky, GA, Mattioli, M, Cattoretti, G, Husson, H, Freedman, A, Inghirami, G, Cro, L, Baldini, L, Neri, A, Califano, A & Dalla-Favera, RGene expression profiling of B cell chronic lymphocytic leukemia reveals a homogeneous phenotype related to memory B cells. J Exp Med, (2001). 194, 1625–1638. | Article | PubMed | ISI | ChemPort |