Purine-Rich Box-1–Mediated Reduced Expression of CD20 Alters Rituximab-Induced Lysis of Chronic Lymphocytic Leukemia B Cells

Contrary to the long-held belief that chronic lymphocytic leukemia (CLL) results from accumulation of B cells (1), the current interpretation is that the increased growth rate of the tumor has overtaken its decreased apoptosis (2). As CLL refers to heterogeneous entities, delineation of patient groups with a given prognosis remains essential (3). For example, patients at risk of a poor outcome are characterized by the expression of CD38 and the absence of VH gene mutations (4). This latter peculiarity may be replaced (5) by the ratio of transcripts for lipoprotein lipase (LPL) to those for a disintegrin and metalloproteinase 29 (ADAM29). Severity is also associated with expression of a full-length form of CD79b (6), presence of ZAP70 (7), and sensitivity of malignant cells to CD5-induced (8) and IgM-induced apoptosis (9).

In CLL, the B-cell marker CD20 is the ideal target for B-cell–depleting antibodies. Consequently, the first anti-B lymphocyte monoclonal antibody (mAb) approved by the Food and Drug Administration (10) to treat non–Hodgkin's lymphoma (NHL) was an anti-CD20 mAb: rituximab. It was then tested in CLL, but why the response rate to rituximab is lower in CLL than in NHL remains to be established (11). Theoretically, to kill the cells (12–14), rituximab combines apoptosis, complement-mediated lysis (CML), and antibody-dependent cellular cytotoxicity (ADCC). One possibility is therefore that diverging response rates reflect different usages of effector mechanisms. If this interpretation is correct, variations in the treatment could be due to low membrane expression of CD20 in CLL compared with NHL (15). Briefly, the more rituximab molecules that bind to the cells increase the chances of the cells being killed and provide a rationale for increasing the density of CD20 on CLL B cells (16).

One step further, the reduced expression of CD20 may result from a defect in transcription factors (TF). Given interactions of purine-rich box-1 (PU.1) with related binding sites in the promoter of cd20 (17), investigation of this TF is appropriate. It belongs to a complex (18), containing also the PU.1-interacting partner (Pip), the octamer-binding protein 2 (Oct), and the B-cell Oct-binding protein 1 coactivator (BOB). Transcription of cd20 is initiated by the binding of PU.1/Pip and Oct2/BOB.1 to its promoter. Combinatorial effects are then induced on granulocyte macrophage colony-stimulating factor (GM-CSF; ref. 19) and all-trans-retinoic acids (ATRA) by PU.1 (20), whereas BOB.1 facilitates posttranslational modifications (21). In addition, the FMS proto-oncogene–like tyrosine kinase 3 (Flt3) receptor represses PU.1 (22) and is in turn repressed by PU.1 (23). Internal tandem duplication (ITD) in its juxtamembrane domain results in Flt3 ligand (FL)-independent activation of Flt3 (reviewed in ref. 24), but this is normally activated by a cognate molecule termed FL.

The cd20 gene can be silenced by abnormal transcription, crippling mutations, methylation of the promoter of one TF, or methylation of the binding sites for one of them. Given the DNA methyltransferase (DNMT)-induced widespread nonrandom CpG island methylation in CLL (25), the possibility exists that pu.1 transcription is repressed by such methylation. Transfection of pu.1 cDNA into CLL B cells up-regulates CD20 expression and improves rituximab-induced CML and ADCC. Rather than methylation-mediated blockade of the gene, excessive engagement of Flt3 may preclude transcription of pu.1 and thereby diminish transcription of cd20. We failed to detect ITP and observed that the level of FL is elevated in the majority of the CLL sera. This increase might engage Flt3 in leukemia.

Patients and controls. Thirty-six untreated patients fulfilling the criteria for the diagnosis of CLL (26) participated in the study. At disease stage A, there were 16 men plus 9 women ranging in age from 54 to 82 years, and at disease stage B, there were 4 men plus 7 women ranging in age from 55 to 83 years (3). Circulating B cells from two patients with mantle cell lymphoma and another two with marginal zone lymphoma served as disease controls (27). Venous blood was also collected from 14 healthy volunteers, and tonsils from 7 children at elective surgery. Informed consent was obtained from CLL patients, normal volunteers, and the guardians of the children. The protocol was approved by the Institutional Review Board at Brest University.

Cell preparation. Peripheral blood mononuclear cells (PBMC) were separated by density gradient centrifugation, and B lymphocytes were purified using a B-cell isolation kit (Miltenyi). Tonsil single-cell suspensions were filtered through a 70-mm nylon mesh strainer, and B lymphocytes were subsequently purified as above. B-cell–enriched populations were then incubated with unconjugated anti-CD5 mAb (BD PharMingen). For further studies, goat anti-mouse IgG antibody-coated magnetic beads selected distinguished negatively the CD5⁻ B cells and positively the CD5⁺ B cells. Natural killer (NK) cells for ADCC were isolated from normal PBMC by removing non-NK cells using the Miltenyi NK cell isolation kit. Their purity was verified by fluorescence-activated cell sorting (FACS) analysis using three mAbs: FITC-anti-CD3, phycoerythrin (PE)-anti-CD56, and PE-cyanin 7 (PC7)-anti-CD16 (Beckman Coulter).

Staining with FITC-anti-CD4, PC7-anti-CD8, and PE-anti-CD19 mAb established that isolated B cells were 98% pure, and staining with FITC-labeled Annexin V (Beckman Coulter) that they were 98% viable.

FACS analyses. Forward and side scatter gates were set to acquire 5,000 events per experiment. Unstained cells and isotype-matched FITC-, PE-, PC7-, and PC5-labeled irrelevant mAbs were used as internal controls. The mean fluorescence intensity (MFI) of CD20 on B cells was measured by FITC- or PC5-conjugated anti-CD20, and that of Flt3 by PE-anti-CD135 (BD PharMingen). The results were validated by calibration beads to compensate for inconsistency. In experiments on PU.1-induced expression of CD20, CD72 (stained with PE-anti-CD72 mAb from Abcam) and CD5 (stained with PC5-anti-CD5 mAb from Beckman Coulter) were taken as positive and negative controls, respectively.

Functional B lymphocyte subsets can be distinguished using combinations of markers. In particular, the relative expression of IgD and CD38 has led to a model for mature B-cell homeostasis from naive Bm1 cells (IgD⁺, CD38⁻) to activated Bm2 cells (IgD⁺, CD38⁺), germinal center founder Bm2' cells (IgD⁺, CD38⁺⁺), centroblast Bm3 and centrocyte Bm4 cells (IgD⁻, CD38⁺⁺), and early memory and memory (IgD⁻, CD38⁻) Bm5 cells (28). Quadruple combinations of markers were devised to phenotype tonsillar B-cell–enriched suspensions. They were stained with PC7-anti-CD5, FITC-anti-CD20, PE-anti-IgD (BD PharMingen), and PC5-anti-CD38 mAb. CD5⁺ B cells were gated and distributed into CD20^low and CD20^high populations. Each of them was subdivided according to the expression of IgD and CD38. To serve as an additional control, normal blood was stained with PE-anti-CD3, PC5-anti-CD19, and PC7-anti-CD5. The MFI of FITC-anti-CD20 was then measured in CD3⁻CD19⁺CD5⁺ cells.

Cell culture. B cells were seeded at 5 × 10⁶ per well and maintained in 200 μL of RPMI 1640 with 10% FCS (Life Technologies), 2 mmol/L l-glutamine, and antibiotics. The cells were incubated for 24 h alone, with 5,000 to 10,000 units of GM-CSF (R&D Systems) or with 1 to 100 μmol/L of ATRA. Methylation was inhibited by 10 to 50 μmol/L of 5-azacytidine or procainamide (both from Sigma).

Western blot analysis. For Western blot analysis, 2 × 10⁷ cells were lysed by incubation for 30 min at 4°C with Triton X-100 in a buffer made of 140 mmol/L NaCl, 1 mmol/L EDTA, and Tris-HCl and supplemented with 1 mmol/L phenylmethylsulfonyl fluoride, 10 μg/mL aprotinin, and 1 mmol/L sodium orthovanadate.

The lysates were centrifuged at 10,000 × g at 4°C for 10 min, run on 10% SDS-PAGE, and blotted to nitrocellulose membrane (Bio-Rad). Unbound sites were quenched with 5% nonfat dry milk in 10 mmol/L Tris, 0.1% Tween 20 for 1 h, washed thrice in the same buffer, and probed with mouse anti-CD20 (BD PharMingen), anti-PU.1 (Santa Cruz Biotechnology), anti-Pip (Ozyme), and anti-actin or anti-BOB.1 (Sigma) mAb or with rabbit anti-Oct2 (Abcam) antibody for 1 h. After three more washes, membranes were incubated with horseradish peroxidase (HRP)-labeled goat anti-mouse or goat anti-rabbit immunoglobulin (both from Jackson ImmunoResearch Laboratories) for 1 h. Bound anti-CD20 mAb was revealed with 5-chloronaphtol, and other bound antibody by enhanced chemiluminescence (ECL). Absorbances of specific bands were normalized to actin and quantified using the Biocapt Express software.

The lysates of leukemic B cells were subjected to immunoprecipitation with rabbit anti-Flt3 antibody (Ozyme), resolved on 10% SDS-PAGE, transferred to polyvinylidene difluoride membrane (Amersham), and probed with antibody to Flt3 and tyrosine phosphorylated Flt3 (BD Biosciences; ref. 29) and revealed as above. The negative and the positive controls were, respectively, tonsillar B cells and acute lymphoblastic cell line cells, referred to as SEM cells and kindly donated by Dr. O. Heidenreich (University of Tübingen, Tübingen, Germany).

Quantitative reverse transcription-PCR. The RNeasy Mini kit (Qiagen) was used to extract mRNA. Its integrity was verified by ethidium bromide staining of the 18S and 28S species on agarose gel electrophoresis. mRNA (1 μg) was reverse transcribed into cDNA using a High-Capacity cDNA Archive kit with random hexamers according to the manufacturer's instructions (Applied Biosystems).

Quantitative reverse transcription-PCR (RT-PCR) was performed with the ABI PRISM 7000 Sequence Detector where the 18S rRNA gene enabled to normalize mRNA. Some primers have been described (30), and the remainder were designed using Primer Express software version 2.0 after Genbank sequences (Table 1): NM_001379 for DNMT1, BC043617 for DNMT3A, AF331857 for DNMT3B, NR_003286 for 18S, NM_152866 for CD20, X52056 for PU.1, NM_002460 for Pip/IRF4, NM_002698 for Oct2, and NM_004119 for Flt3. Temperatures were 50°C for 2 min, 95°C for 10 min, 95°C for 15 s, and 60°C for 1 min for 40 cycles. Numbers of transcripts were determined from threshold cycle numbers, the standard curves were normalized to 18S, and the levels were expressed relative to those of transcripts in normal B cells, which were assigned a value of 1.

PCR analyses of pu.1 promoter methylation and ITD of flt3 and FL. The methylation status of the pu.1 promoter was determined essentially as described by Ushmorov and colleagues (31). Briefly, genomic DNA was purified using a DNA blood kit (Qiagen), and 50 ng were digested with 20 units of MspI or HpaII (Invitrogen) for 3 h at 37°C. Then, 2 μL of the digest were amplified with 5′-TTAGCCCCCAAAGTCATCCCTCTCA-3′, plus 5′-ACCCTTCCATTTTCGACTCCTGTAAC-3′, to flank the 5′-CCGG-3′ binding sites of the pu.1 promoter. PCR comprised denaturation at 94°C for 5 min, 40 cycles at 94°C for 30 s, 62°C for 1 min, 72°C for 1 min, and a 10-min final extension at 72°C. The products were resolved in 1% agarose gel and stained with ethidium bromide.

For analysis of ITD of the flt3 gene, 50 ng of genomic DNA were amplified with 5′-CAATTTAGGTATGAAAGCC-3′ plus 5′-CAAACTCTAAATTTTCTC-3′ in 50 μL reaction volume by 10 cycles at 95°C for 1 min, 50°C for 1 min, and 72°C for 1 min followed by 35 cycles at 95°C for 1 min, 47°C for 30 s, and 72°C for 1 min, ending with a final extension at 72°C for 10 min. The final products were checked as above. The wild-type allele was ∼129 bp, and the ITD allele ranged from 149 to 240 bp.

To validate our search for mutations in the flt gene, the promyelocytic leukemia cell line HL-60 (American Type Culture Collection) and the acute monocytic leukemia cell line MV4-11 (gift of G. Kromer, Institut National de la Sante et de la Recherche Medicale, Paris, France) were taken as positive and negative controls, respectively. Primer pairs specific for FL were also used to amplify a 352-bp sequence in B cells from two CLL samples and two normal tonsils.

Transfection of malignant B cells. The method for isolating the pu.1 probe has been described elsewhere (32). It was inserted into a pcDNA-3.1 expression plasmid, and two suspensions of CLL B lymphocytes were electroporated at 40% efficiency using the B-cell nucleofactor kit (Amaxa). The first suspension was transfected with pu.1-containing pcDNA-3.1 constructs, and the second with empty pcDNA-3.1 vectors. Both were cotransfected with green fluorescent protein (GFP)-containing pMax vector. After a 24-h culture, CD20 and PU.1 were evaluated at the mRNA level by quantitative RT-PCR, and at the protein level by FACS.

Killing assays. B lymphocytes from patients and controls were distributed into three groups of aliquots of 10⁶ B cells each. The first group served for apoptosis, the second for CML, and the third for ADCC. In each group, aliquots underwent a 4-h incubation with rituximab (Roche) at the doses of 0, 1, 10, 25, and 50 μg/mL, respectively, and lysis of aliquots without rituximab was subtracted from that of the rituximab-containing aliquots with rituximab.

In the CML assay, B cells were incubated with 15% human AB serum for 6 h. Propidium iodide (PI) was added 10 min before harvest for FACS analysis. In the ADCC assay, B cells were incubated with 10⁶ NK cells for 6 h. In the apoptosis assay, B cells were incubated in medium without serum and evaluated by FACS analysis through a combination of PI and PE-Annexin V.

ELISA for FL. The level of FL was measured in 23 sera from CLL patients and 10 sera from normal volunteers. The sandwich ELISA kit from R&D Systems was used according to the manufacturer's instructions. Values greater than 75 pg/mL (i.e., above 3 SDs of the mean of 30 normal sera) were considered to be increased.

Statistical analyses. Results are expressed as mean ± SD. Comparisons were made using the Mann-Whitney U test for unpaired data and the Wilcoxon test for paired data. Correlations were established using the Spearman's test.

Expression of CD20 at the mRNA and protein levels. There was less CD20 mRNA in CD5⁺ B cells of CLL blood than in CD5⁺ B cells of normal blood and control tonsils. To compensate for inconsistency, a value of 1 was assigned to the mean of 14 normal samples of blood B cells, and other samples were rated at 1.8 ± 0.09 in 7 control tonsils and 0.17 ± 0.14 in 36 CLL patients (P < 0.01, compared with normal blood and tonsils).

Western blot analysis detected fewer CD20 molecules (Fig. 1A) in CLL than in normal blood and tonsil B cells [0.02 ± 0.01 versus 0.22 ± 0.05 (P < 0.01) and versus 0.29 ± 0.04 (P < 0.01)]. Defective expression of CD20 was confirmed by FACS analysis (Fig. 1B , left), where the CD20 MFI of gated CD5⁺ B cells was 10.8 ± 1.1 in CLL, 41.2 ± 7.2 in normal blood (P < 0.01, compared with CLL), and 55.5 ± 7.2 in control tonsils (P < 0.01, compared with CLL). Contrary to CLL B cells, CD20 was overexpressed in B cells from lymphoma.

Clearly, the CD20 MFI distinguished two populations of tonsillar B cells. The presence of CD38 and the absence of IgD (Fig. 1B,, right) localized the CD20^high B cells within the Bm3 centroblast/Bm4 centrocyte population (28). Interestingly, there were more CD20 molecules in CD5⁻ (Fig. 1C,, black histograms) than in CD5⁺ B cells (Fig. 1C , white histograms) from normal blood and tonsils.

Reductions in the number of mRNA copies for CD20 were not different in the two disease stages studied in our patient group, neither were they in patients with a different VH mutational status, as defined by the LPL/ADAM29 ratios (Table 2).

Expression of the PU.1/Pip and the Oct2/BOB.1 complexes. The role of some TFs in the reduced expression of CD20 in CLL was then determined. There were fewer transcripts for PU.1 in CLL than in control B cells (Fig. 1A), and therefrom a correlation (P < 0.01) between PU.1 mRNA and CD20 mRNA (Fig. 1C) in CLL (Fig. 1D,, closed circles) but not in lymphoma (Fig. 1D,, open circles). These data suggest that PU.1 might affect CD20 expression in CLL, although its expression did not correlate with the disease stage or the VH mutational status (Table 2). We found less PU.1 protein (Fig. 1A) in CLL than in normal blood and tonsil B cells [0.02 ± 0.01 versus 0.30 ± 0.05 (P < 0.05) and versus 0.25 ± 0.04 (P < 0.05)]. Compared with CLL, in non-CLL B-cell lymphoproliferative disorders, there were more transcripts for CD20 (1.28 ± 0.29 versus 0.17 ± 0.14; P < 0.05) as well as for PU.1 (0.66 ± 0.11 versus 0.08 ± 0.01; P < 0.05).

In contrast, Oct2, BOB.1, and Pip had comparable levels of mRNA and protein in CLL and normal blood CD5⁺ B cells. They were also unrelated to disease stage and VH mutational status (Table 2). The absence of difference indicates that the reduction in the density of CD20 could not be ascribed to any of these three factors.

PU.1-induced expression of CD20. Next, we explored the possible role of PU.1 in the decrease in the expression of CD20 in CLL B lymphocytes. To this end, B cells from eight patients were transfected with GFP-pu.1–containing pcDNA-3.1 constructs or with empty vectors. Transfection of pu.1 enhanced the transcription of pu.1 (P < 0.05, between pu.1-containing constructs and empty vectors in Fig. 2A). This promoted an increase in the expression of CD20 at the mRNA (P < 0.05 between pu.1-containing constructs and empty vectors in Fig. 2A) and the protein levels (P < 0.05, between pu.1-containing constructs and empty vectors in Fig. 2B). Consequently, increases in PU.1 mRNA correlated to increases in CD20 mRNA (P < 0.05 in Fig. 2A). An example is shown in Fig. 2C (left), where the specificity of PU.1 was confirmed by raised expression of CD72, the gene of which is controlled by PU.1 (33), in contrast to CD5, the gene of which is not (34).

A high density of CD20 improves the rituximab-induced lysis. A standing question is whether raised expression of CD20 ameliorates the rituximab-mediated cell lysis. CML (Fig. 3A,, left) and ADCC (Fig. 3B,, left) were significantly increased in a dose-dependent manner [Fig. 3A  (right) and B (right)]. Changes in surface expression of CD20 (7.5 ± 5.4%) correlated with improvements of CML (6.1 ± 3.1%; P < 0.05) as well as with those of ADCC (21.4 ± 8.7%; P < 0.05). Not surprisingly, apoptosis was not influenced by the expression level of CD20, as efficient rituximab-induced apoptosis requires cross-linking of rituximab with a second-layer antibody (35).

Is pu.1 methylation regulated? Because both GM-CSF (19) and ATRA (20) increase the expression of PU.1 in myeloid cells, we looked for similar mechanisms in B cells. Although GM-CSF and ATRA were added at doses as high as 10,000 units and 100 μmol/L, respectively, the levels of mRNA for PU.1 and CD20 were unchanged. In other words, these agents failed to modify membrane expression of CD20.

Alternatively, undue activity of DNMTs could methylate and thereby silence the pu.1 gene in CLL B cells. Supporting this view, DNMT1 and DNMT3A were 2.0 ± 0.9–fold and 1.8 ± 0.7–fold increased, respectively, in 10 CLL patients relative to 4 healthy controls. We investigated this further by evaluating DNMT-induced methylation of the pu.1 promoter. The assay was based on the inability of some enzymes to digest a methylated 5′-CC^mGG-3′ site. Genomic DNA of B cells from 14 CLL patients and 4 healthy controls was digested with the methylation-sensitive enzyme HpaII or the methylation-insensitive enzyme MspI and subsequently amplified (Fig. 4A , left).

The group of 14 patients was distributed into a first subgroup of 8 individuals whose CpG included in the restriction sites were methylated, and a second subgroup of 6 individuals whose CpG were not methylated. To confirm such a dichotomy, five methylated CLL samples were incubated with DNMT inhibitors after pilot experiments had ensured that they did not alter cell viability (Fig. 4A , right). Three nonmethylated CLL samples served as negative controls. Although favoring the transcription of pu.1, DNMT blockade did not restore expression of CD20. Furthermore, there was hardly any difference in the MFI of CD20 comparable in methylated and unmethylated groups (6.66 ± 3.86 versus 5.83 ± 1.85). That is, aberrant methylation of the pu.1 gene does not explain its repression in CLL.

Flt3 kinase receptor is overexpressed and activated. We found 15.8-fold more mRNA for Flt3 in CLL than in normal B cells (Fig. 4B,, left), and protein in excess was revealed by FACS analysis of CLL B cells (Fig. 4B , right). This is consistent with the restriction in pu.1 expression by Flt3 signaling (36) and the low levels of Flt3 transcripts in PU.1^−/− fetal liver hematopoietic progenitors (37). Of intriguing note, the amount of mRNA for Flt3 was higher in stage A than stage B patients (P < 0.01) and in VH mutated than VH unmutated cases (P < 0.01).

In addition to its overexpression, Flt3 was activated. As expected, Western blot analysis of immunoprecipitate (Fig. 4C , left) detected a 130-kDa nonglycosylated and a 160-kDa glycosylated bands of Flt3. The anti-phosphorylated antibody showed that both forms were activated, although normal tonsillar B cells were negative and SEM cells were positive.

Two explanations deserve special attention. The first is the FL-independent activation where ITD activates Flt3 and, by doing so, prevents the expression of PU.1 (38). This option was invalidated by the absence of ITD on genomic DNA (Fig. 4C,, right). The second option is the FL-dependent phosphorylation. This seemed to be likely, as the serum levels of FL (reviewed in ref. 24) were elevated (Fig. 4D) in 20 of 23 CLL sera compared with none of 10 sera from healthy volunteers. Our pilot experiments reveal more transcripts for FL in CLL than in normal B cells (data not shown). Further experiments are currently in progress to gain insight into this mechanism.

Controversy persists over the relationship between efficacy of the anti-CD20 mAb treatment and expression level of its CD20 target. A correlation has indeed been reported by some (39) but not all groups (40). In view of the scarce expression of CD20 by CLL B cells (15), we postulated that enhancing its expression might improve their susceptibility to rituximab. To assess this possibility, we transfected malignant cells with pu.1 cDNA. The density of CD20 was enhanced, and indeed, rituximab-induced CML and ADCC were improved accordingly.

We thus wanted to decipher the mechanisms of the lowered expression of CD20 in CLL. Because one way to modulate a protein synthesis is to interfere with its TF, we first evaluated the importance of some TFs previously shown to act on cd20. Whereas the Pip cofactor of PU.1, Oct2, and its cofactor BOB.1 were normal, PU.1 was defective, and the following expression of CD20 diminished. Consistent with our interpretation, the transcription of PU.1 mRNA was elevated in other non-CLL malignant. Transfection confirmed the role of PU.1 in the regulation of CD20 by the presence of effect on the positive control CD72 (33) and its absence of effects on the negative control CD5 (34).

Because our goal was to normalize CD20 expression on CLL B lymphocytes, the results of the first set of experiments focused our attention on PU.1 up-regulating agents. Like several groups (see ref. 41 for example), we failed to modify the expression of PU.1 and CD20 using GM-CSF or ATRA. Thus, normalization of PU.1 by ATRA in myeloid leukemic blasts (20), and the absence of normalization in CLL B lymphocytes, reflects differences between the mechanisms activating the pu.1 promoter in the former and the latter cells (42).

In addition, our data raise the issue of the mechanisms restricting the transcription of pu.1. It is unclear as to what TF is involved because the relative contribution of the growth factors to this transcription is not fully understood. Gene regulation was therefore considered, and it made sense to compare transcripts for DNMT in leukemic and normal B cells because one of the hallmarks of malignant B cells is hypermethylation of DNA (24). Despite the high expression level of DNMT, the CpG islands of the pu.1 promoter were hypermethylated in only 8 of the 14 CLL cases tested. Such a variability renders methylation an unlikely mechanism for pu.1 repression. Not only may demethylation of pu.1 be assumed to be unnecessary to normalize CD20 expression, simply because some CLLs are not methylated, but it is also insufficient because blockade of DNMT did not restore CD20 expression.

Instead of methylation, CD20 expression might be inhibited by Flt3. Consistent with this option is our finding of more Flt3 in stage A than in stage B patients and in VH mutated than VH unmutated cases. The influence of Flt3 on CD20 is reminiscent of the association of a low CD20 expression (43) with stage A and with the presence of VH gene mutations (4). However, this comparing of facts is contradicted by the absence of correlation between PU.1 expression and the stage or the mutational status of CLL B cells, suggesting that additional agents interfere with the relationship between pu.1 and Flt3. Indeed, PU.1 functions as a transcription repressor for the flt3 gene (23), and, in turn, its deficiency up-regulates the expression of Flt3. In either case, defective PU.1 is an independent prognostic factor for CLL. Flt3 in excess might be due to overproduction of FL, in an autocrine loop for the Flt3/FL system, as put forward by Brasel and colleagues (37). Expectedly, Flt3 inhibition would release PU.1 and thereby increase the expression of CD20. Unlike acute myeloblastic leukemia, Flt3 was internalized and degraded by 5 min in leukemic B lymphocytes as well as normal B lymphocytes. Similar results were obtained in three separate experiments so that the Flt3 blockade experiment was rendered impossible. However, our pilot experiments showed that the expression of mRNA for FL was higher in leukemic cells compared with normal B cells.

More effective strategies are required in the treatment of CLL with rituximab. Recent findings have indicated that rituximab-opsonized B cells are sequestered by the reticuloendothelial system and released back into the circulation after the rituximab-CD20 complexes have been taken up by the phagocytes (44). To prevent the CD20 loss, and enhance targeting, the use of low doses of rituximab has been advised (45). Additional insight comes from the demonstration in a mouse transgenic for the human cd20 gene, that resistance to CD20 involves circulatory dynamics, as access of B cells to the circulation is required (46, 47). Finally, aberrant expression of B-cell activating factor of the tumor necrosis factor family (BAFF) by CLL B cells might be an additional mechanism for survival (48). This excess accounts for B-cell resistance to rituximab (49) and suggests to associate inhibitors of BAFF with rituximab.

In conclusion, our belief is that these results, although preliminary, are robust enough to provide a strong basis for future work. Another area that requires further understanding is the role of Flt3 in the down-regulation of PU.1. These findings may provide new opportunities in the treatment of B-cell malignancies.

No potential conflicts of interest were disclosed.

Grant support: “La Ligue Contre le Cancer” and “Association Céline et Stéphane.”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. G. Kromer and O. Heidenreich for the generous gift of cell lines, Dr. E. Hardy for technical help, C. Séné and S. Forest for expert secretarial assistance, and R. Budd (Birmingham, United Kingdom) for help with the writing of the article.