Homodimers but not monomers of Rituxan (chimeric anti-CD20) induce apoptosis in human B-lymphoma cells and synergize with a chemotherapeutic agent and an immunotoxin

In a previous study, we demonstrated that monoclonal antibodies (mAbs) (ie, anti-CD19, anti-CD20, anti-CD21, and anti-CD22) that have little or no antigrowth activity on neoplastic B-cells can become potent antitumor agents when they are converted into homodimers.1 These homodimers exert their antigrowth activity either by signaling G₀/G₁arrest or by inducing apoptosis, depending upon the cell surface molecule to which they bind. Our results confirmed and extended previous reports in which a chimeric mAb (Chi BR96) homodimer killed tumor cells 10-fold more effectively than its monomeric counterpart.2 

The chimeric anti-CD20 mAb (Rituxan/Rituximab; Genentech, San Francisco, CA) has recently been approved by the Federal Drug Administration in the United States for the treatment of patients with relapsed or refractory, low-grade or follicular, B-cell non-Hodgkin lymphoma, and in Europe for patients with relapsed stage III/IV follicular lymphoma (reviewed in Grillo-Lopez et al3 and Gopal and Press4). The overall response was 50% when it was used as a single agent,4,5 and response rates were significantly higher when it was used in combination with chemotherapy.6 It is not yet known whether Rituxan will improve the long-term survival of patients with low-grade lymphoma, although early results look encouraging.3,4 

In vitro studies, as well as studies in both animals and humans, suggest that the antitumor activity of Rituxan can be mediated by antibody-dependent cellular cytotoxicity (ADCC) or complement-dependent cytotoxicity (CDC).3,4,7 It has also been shown that Rituxan can induce apoptosis7-10 in DHL-4 or DHL-4–derived cell lines with bcl2 gene rearrangements.9,10 However, in accord with studies using other anti-CD20 mAbs such as 1F5 and B1,11 we could induce only modest levels of apoptosis with Rituxan (even in DHL-4 cells) unless we hypercrosslinked the bound Rituxan with rabbit antihuman immunoglobulin (Ig)–G (RAHIg) or goat antimouse immunoglobulin (GAMIg). In accord with these observations, it has recently been reported that Rituxan can induce apoptosis in lymphoma cells from patients only when GAMIg is added.10 

Because hypercrosslinking with a secondary antibody would be impractical in vivo, in this study we evaluated tetravalent F(ab′)₂ or IgG homodimers of Rituxan for their ability to exert antigrowth activity and to induce apoptosis in B-lymphoma cell lines. We found that homodimers had superior antigrowth activity and induced apoptosis in several neoplastic B-cell lines in vitro. Apoptotic activity did not require the participation of Fc receptors (FcRs) on target cells and was caspase dependent. Treatment with homodimers, compared with monomers, also rendered tumor cell lines more sensitive to chemotherapeutic agents and synergized with an anti–CD22-immunotoxin (IT) in vitro.

Five human Burkitt lymphoma cell lines, including Daudi, Raji, Ramos, Namalwa (ATCC, Rockville, MD); the P-glycoprotein (P-gp)–transfected drug-resistant Namalwa/multidrug-resistant 1 (MDR1)12 (a gift from Dr Rosemary O'Connor, Immuno Gen, Cambridge, MA); 1 diffuse histiocytic cell line (DHL-4); and 2 T-leukemia cell lines, MOLT-4 and Jurkat (ATCC), were maintained in culture by serial passage in RPMI-1640 containing 25 mM Hepes, 10% heat-inactivated fetal bovine serum, 100 U/mL penicillin, 100 μg/mL streptomycin (complete medium [CM]), and 100 mM glutamine. The cells were grown in a humidified atmosphere of 5% CO₂ and air. Cell viability was determined by trypan blue exclusion. Cell lines were maintained in culture for 6 weeks and were then replaced with frozen stock.

The chimeric anti-CD20 mAb (Rituxan) developed by IDEC Pharmaceuticals13 (San Diego, CA) was purchased from Genentech (San Francisco, CA). We used 2 controls: (1) an isotype-matched mouse IgG₁ of irrelevant specificity (3F12), prepared in our laboratory by purification of hybridoma cell supernatants on a protein A–sepharose column; and (2) a human IgG (Sigma, St Louis, MO), purified by chromatography on protein G–sepharose. GAMIg and GAHIg were affinity-purified in our laboratory. Fluorescein isothiocyanate (FITC)–GAMIg (H + L) was purchased from Kirkegaard & Perry Laboratories (Gaithersburg, MD). Rabbit antiserum against human IgG Fc was purchased from Cappel, Biochemical Division (Aurora, OH). Goat antihuman IgM (GAHIgM) was purchased from Chemicon (Temecula, CA). FITC-labeled annexin V apoptosis detection kits were purchased from BioVision Research Products (Palo Alto, CA). Caspase inhibitors (Z-VAD-FMK, Ac-YVAD-CHO, and Z-DEVD-FMK) were purchased from Calbiochem (San Diego, CA).

Indirect immunofluorescence assays were carried out by means of the panel of mAbs described above and FITC-GAMIg. The mAbs (monomers and homodimers) were labeled with FITC as described elsewhere14 and were used to determine the cell-binding and cell-dissociation rates of mAbs. Cells were suspended at 10⁶ cells per milliliter in CM and were treated with 0.1 to 1.0 μg/mL mAbs for 30 minutes on ice, washed twice with CM containing 0.1% sodium azide, resuspended in 100 μL CM, and treated with 2 to 3 μL FITC-GAMIg. The direct immunofluorescence assay was carried out by means of FITC-mAbs. To determine the dissociation rate of FITC-mAbs (monomers vs homodimers), cells were treated as described above with saturating concentrations of FITC-mAbs. The excess mAbs were washed out, and the cells were incubated at 37°C for 2, 4, 6, 24, 48, and 72 hours in CM. At the indicated intervals, samples were taken, washed, and analyzed on a FACScan (Becton Dickinson, Franklin Lakes, NJ) to determine the percentage of the remaining cells that were still stained.

Homodimers of Rituxan and the mouse IgG₁ control mAb 3F12 were prepared and purified as previously described.1Briefly, we used 2 hetero-bifunctional cross-linkers, succinimidyl 4-(maleimidemethyl) cyclohexane-1-carboxylate (SMCC) and N-succinimidyl S-acethylthio-acetate (SATA) (Pierce, Rockford, IL) to bind 2 IgG molecules through a stable thioether bond. The final product was a mixture of polymers, trimers, dimers, and monomers, which were further separated on a preparative 21.5 mm inner diameter (ID) × 60 cm TSKG3000SW (Tosohaas, Montgomeryville, PA) high-performance liquid chromatography (HPLC) column. The protein fractions obtained by HPLC were concentrated to 5 to 10 mg/mL, filter-sterilized, and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and HPLC. The purity of the homodimer fraction was 80% to 90%.

HPLC-purified homodimers were digested with insoluble pepsin (Sigma) as previously described.15 The F(ab′)₂fraction was affinity-purified on a protein G–Sepharose column.

The mAb, RFB4, which recognizes human CD22, was coupled to deglycosylated (dg) ricin toxin A chain (RTA). RFB4-dgRTA was prepared in our good laboratory practice (GLP) scale-up laboratory as described.16 RFB4-dgRTA, has been evaluated in severe combined immunodeficient (SCID)/Daudi mice (reviewed in Ghetie et al17) and in patients with B-cell lymphoma (reviewed in Sausville and Vitetta18 and Engert et al19).

Proteins were analyzed under both reducing and nonreducing conditions by SDS-PAGE on 4% to 15% gels by means of the Pharmacia PhastSystem Separation Unit.20 Protein bands were visualized by staining the gels with Coomassie blue.

The molecular weights of the dimers were determined by analytical HPLC with the use of a 7.5 mm ID × 60 cm TSKG4000SW column (Tosohaas). A standard protein mixture of apoferritin (450 kd), IgA (300 kd), and IgG (150 kd) was fractionated on the same HPLC column to establish the retention time of known molecular weight standards.

We incubated CD20⁺ human cell lines [Ramos, DHL-4, Daudi, Raji] and the CD20⁻MOLT-4 cells at 2 × 10⁵ cells per milliliter in CM at 37°C for different time periods in the presence and absence of different concentrations of anti-CD20 (1, 10, and 100 μg/mL). The irrelevant IgG₁ isotype–matched mAb 3F12 and a human IgG (100 μg/mL) were used as a controls. Cells were stained with trypan blue and counted. Cell growth rates were then determined.

The antiproliferative activity of different mAb monomers and homodimers was determined by means of a ³H-thymidine incorporation assay on Ramos, DHL-4, Daudi, Raji, Namalwa, Namalwa/MDR1, and MOLT-4 cells. Briefly, 5 × 10⁴ cells in 100 μL CM were plated in 96-well plates (Costar, Corning, NY) and incubated for 24 to 48 hours with 100 μL of different concentrations (0.01 to 1.0 mg/mL) of mAbs (monomers and homodimers) diluted in CM. To determine the cytotoxic effect of the immunotoxin (IT), the cells were incubated with 6 serial dilutions (10⁻⁸ to 10⁻¹³ M) of either IT alone or IT in combination with a constant amount (20 μg/mL) of Rituxan monomers or homodimers. Doxorubicin was used in a range of 1 to 70 nM, either alone or in the presence of Rituxan as described above. After incubating the cells at 37°C, the plates were pulsed for 4 hours with 1 μCi³H-thymidine (Amersham Pharmacia Biotech, Piscataway, NJ), harvested, and counted in a liquid scintillation spectrometer. The percentage reduction in ³H-thymidine incorporation, compared with untreated controls, was used to quantitate the cytotoxic effect.

We treated 2 × 10⁵ cells with different concentrations (1 to 100 μg/mL) of Rituxan or irrelevant mAbs (monomers and homodimers); with Rituxan monomers (10 μg/mL) and GAMIg (50 μg/mL) or rabbit antihuman immunoglobulin (RAHIg) (10 μg/mL); or with GAMIg or RAHIg alone for 2, 4, 8, and 24 hours at 37°C. Cells were washed, treated with FITC–Annexin V and propidium iodide (PI), washed again, and analyzed within 1 hour on a FACScan. FL1-positive cells (stained with FITC-Annexin) were considered apoptotic and FL2-positive cells (permeable to PI) were considered necrotic. NaN₃ (5 to 10 mg/mL) was used as a positive control for the induction of apoptosis. The caspase inhibitors (Z-VAD-FMK, Ac-YVAD-CHO, and Z-DEVD-FMK) were used at 50 μM.

Rituxan multimers were separated on an HPLC column and were characterized by SDS-PAGE and by binding to, and dissociation from, cells.

The initial Rituxan contained approximately 100% monomers (150 kd). Homodimers of Rituxan were purified by preparative HPLC (Figure1). We obtained 4 fractions and analyzed them by SDS-PAGE. Fraction 1 consisted of high molecular weight polymers (greater than 450 kd); fraction 2 consisted of approximately 80% trimers (450 kd); fraction 3 consisted of 80% to 90% dimers (300 kd); and fraction 4 consisted of greater than 90% monomer (150 kd).

The 6 CD20⁺ B-cell lines and the CD20⁻ T-cell line were stained with FITC-Rituxan to determine their levels of expression of CD20. As Table 1 shows, 4 of the 6 B-cell lines were more than 90% positive and expressed varying densities of CD20. Raji cells expressed the highest density and DHL-4 cells the lowest density of CD20. Namalwa and Namalwa/MDR1 were less positive (Figure 2A), and Jurkat cells were negative. When the relative binding affinities to all the cell lines were determined in the same experiment (by plotting the percentage of positive cells vs concentration and calculating the concentration required to reach 50% saturation), Rituxan showed the highest relative binding affinity on DHL-4 cells and the lowest on Namalwa and Namalwa/MDR1 cells. The relative binding affinities on Daudi, Ramos, and Raji cells were very similar (Figure 2A and Table 1). The binding of Rituxan monomers, dimers, and trimers to cells was also compared by means of direct immunofluorescence on Daudi, Raji, Namalwa, Namalwa/MDR1, Ramos, DHL-4, and Jurkat cells. When the relative binding affinity to Ramos cells was determined, it was found to be similar for monomers, homodimers, and trimers, indicating that increasing the valency of the mAb did not change its relative binding affinity for CD20 (Figure 2B and Table 1).

Dissociation experiments were performed by means of a direct immunofluorescence assay in which the DHL-4 or Ramos cells were saturated with FITC-mAbs and the positively stained cells were quantified after different time intervals in culture. As shown in Figure 2C, the monomers dissociated much more rapidly than the homodimers (T_1/2 of 2 vs more than 72 hours), and the trimers showed no dissociation even after 72 hours. The increased rate of dissociation as the valency of the mAb increased indicates a stronger and more prolonged interaction with the target cells. These results support the previously reported finding that CD20 does not internalize.21,22 

Taken together, these experiments demonstrate that homodimers and trimers have similar binding affinity but a slower rate of dissociation than monomers.

Several different lymphoma cell lines were cultured with different amounts of Rituxan monomers or homodimers for 6 days. At different time intervals, the cells were counted, and their growth indexes and viabilities were compared with those of untreated cells. As shown in Figure 3 the growth indexes of cells treated with either monomers or homodimers (either intact IgG or as F(ab′)₂) were lower than those of untreated cells. DHL-4 and Ramos cells treated with homodimers grew more slowly than those treated with monomers, demonstrating that the antigrowth activity of Rituxan was increased by using it in homodimeric form. The F(ab′)₂ homodimers also had significant antigrowth activity, suggesting that interaction with cellular FcγRII (CD32) is not involved in this growth inhibition (Figure 3). In addition, since Ramos cells lack CD3223 and their growth was inhibited, the presence of CD32 on target cells is not required for growth inhibition in vitro. We also demonstrated that the binding of FITC-Rituxan (monomer or dimer) to CD20⁺ cells (eg, Ramos, DHL-4) was not impaired by blocking CD32 with specific antibodies prior to adding Rituxan (data not shown). Rituxan had no effect on the growth of CD20⁻MOLT-4 or Jurkat cells (data not shown).

The FITC–Annexin V assay demonstrated that monomers of Rituxan did not induce apoptosis unless cell-bound mAbs were cross-linked with a secondary antibody (GAMIg) (Table 2). Annexin V positivity was achieved after incubating cells with Rituxan monomers cross-linked with GAMIg at 37°C for 24 hours. A comparable effect was observed when Ramos cells were treated for 24 hours with homodimers or with monomers and GAMIgM (Table 2). The positive control, sodium azide, induced a very profound apoptotic effect after 24 hours. The apoptotic activity of Rituxan homodimers on different B-cell lines is shown in Table 3. Apoptosis was completely inhibited by the irreversible caspase inhibitors Z-VAD-FMK and Z-DEVD-FMK and partially inhibited by the reversible inhibitor (Ac-YVAD-CHO) of PARP cleavage (Figure4).

The Rituxan monomers, homodimers, and trimers were incubated for different intervals of time with Ramos, DHL-4, Daudi, Raji, Namalwa, and Namalwa/MDR1 cells, and the levels of ³H-thymidine incorporation were measured and compared with those of untreated cells (Table 3). CD20⁻MOLT-4 cells were used as negative control cells. The Rituxan monomers did not inhibit ³H-thymidine incorporation in any of the cell lines tested, even at high concentrations (greater than 10⁻⁶ M). Both the homodimers and trimers reduced ³H-thymidine incorporation in a dose-dependent manner. Neither the irrelevant 3F12 mAb nor the human IgG had cytotoxic activity in any form (data not shown). The same assay was used to determine the cytotoxic activity of Rituxan in combination with doxorubicin as described below.

In comparing Table 3 with Table 1, we found no obvious correlations between the mean fluorescence intensity (MFI) of CD20 on the cell lines, and either the IC₅₀ of the Rituxan dimers (³H-Thymidine assay) or the percentage of positive cells in the Annexin V assay.

We used 3 cell lines with different sensitivities to doxorubicin: DHL-4, Namalwa, and Namalwa/MDR1. When treated with doxorubicin alone, the IC₅₀ for drug-sensitive Namalwa cells was 3- to 4-fold lower than that of either drug-resistant Namalwa/MDR1 cells (Figure5C) or the less drug-sensitive DHL-4 cells (Figure 5B). When Rituxan monomers (20 μg/mL) were added to different concentrations of doxorubicin (1 to 70 nM), their IC₅₀ on DHL-4 (Figure 5A) or Namalwa/MDR1 (Figure 5C) cells decreased in a dose-dependent manner, suggesting that both the drug-resistant Namalwa/MDR1 cells and the DHL-4 cells became more sensitive to the cytotoxic effect of doxorubicin. When added to the same amount of homodimer (at a noncytotoxic concentration), doxorubicin became even more toxic for DHL-4 (Figure 5A) and Namalwa/MDR1 (Figure5C) cells. In contrast, both homodimers and monomers increased the sensitivity of Namalwa cells to doxorubicin to the same extent (Figure5B), suggesting that the Rituxan homodimers are most advantageous when cells are insensitive to doxorubicin.

The IT RFB4-dgRTA kills human B-lymphoma cells in vitro and in vivo.24,25 In the presence of a noncytotoxic concentration of Rituxan monomers or homodimers, the IC₅₀ of the IT decreased by 3-fold and greater than 1000-fold, respectively (Figure6A). RFB4-mAbs plus either Rituxan monomers or homodimers had no effect on the cells (Figure 6B).

Monoclonal antibodies that exert antitumor activity may do so by virtue of their effector function (CDC and ADCC)26-28 and/or of their ability to signal growth arrest or cell death.29 In this study, we have evaluated the negative signaling activity of the mAb Rituxan, a chimeric antihuman CD20 mAb that has excellent activity in humans with lymphoma. Previous in vitro studies have shown that Rituxan arrests cell growth most effectively in vitro following cross-linking with a secondary anti-immunoglobulin antibody or with CD32-expressing cells.11 To take advantage of these observations, we have generated mAb homodimers (tetravalent antibody molecules) of Rituxan in order to optimize its ability to hypercrosslink. In general, homodimers have higher avidity, cross-link their target antigen more efficiently, and dissociate from cells more slowly.1,2 Previously, Wolff et al2 and Caron et al30 reported that homodimers had more potent antitumor activity than monomers, and we reported that mAbs that have little or no negative signaling activity can become potent antitumor agents when they are converted to homodimers.1 In vitro, some homodimers can exert antigrowth activity by signaling G_o/G₁ arrest and/or apoptosis, depending upon which cell surface molecule they bind.1 In some cases F(ab′)₂ homodimers work as well as, or better than, IgG homodimers in vitro.1 

Rituxan has exhibited in vitro antigrowth activity on Ramos, DHL-4, and Raji cells, but apoptosis has been demonstrated only in Ramos and DHL-4 cells.7-11 It has also been shown that exposure of some, but not all, cell lines to anti-CD20 mAbs induces tyrosine phosphorylation.7 

Our studies were designed to determine whether homodimers of Rituxan have better antigrowth activity than monomers. The major findings to emerge are the following: (1) homodimers are far more effective than monomers (and as effective as monomers plus GAMIg) in inducing growth arrest; (2) homodimers but not monomers induce both apoptosis and necrosis, and these effects are not dependent upon the presence of CD32 on the tumor cells; (3) the apoptotic effect of homodimers is inhibited by the caspase inhibitors (Z-VAD-FMK and Z-DEVD-FMK); (4) as compared with monomers, homodimers of Rituxan show decreased dissociation from, CD20⁺ cells; (5) the inhibitory activity of Rituxan on different cell lines does not correlate with their density of CD20; (6) Rituxan homodimers show marked synergy with an anti-CD22 IT in vitro; (7) Rituxan homodimers chemosensitize doxorubicin-resistant or insensitive tumor cells more effectively than monomers in vitro.

Taken together with the results of other studies, the results presented here suggest that Rituxan homodimers might have better antitumor activity in vivo by virtue of enhanced negative signaling. This may, of course, be the case only if homodimers cross-link better in vivo where FcR⁺ cells are present.

With regard to the mechanism of action of Rituxan in vivo, it has been clearly demonstrated that its FC portion and effector functions play major roles in its therapeutic efficacy in genetically engineered mice.30 Whether the effector function of homodimers will be enhanced in vivo remains to be determined. Hence, some mAb homodimers show improved ADCC and CMC30 whereas others do not.2 In an in vivo setting, an improvement may depend upon half-life and the binding of homodimers vs monomers of a particular mAb to effector cells.

Regardless of whether homodimers will outperform monomers in vivo by virtue of enhanced negative signaling, effector function, or both, another way to improve the activity of Rituxan is to use it in combination with other antitumor agents. For this reason, we also tested Rituxan in combination with ITs and chemotherapy in vitro. With regard to ITs, we have previously reported that an anti-CD19 mAb (HD37) acts synergistically with the anti-CD22 IT RFB4-dgRTA in SCID mice with human lymphoma.24 In this study, we demonstrate that both Rituxan monomers and homodimers can increase the cytotoxic effect of RFB4-dgRTA but not an RFB4 mAb in vitro and that homodimers were more effective than monomers. This finding suggests that Rituxan homodimers plus ITs should be considered for further in vivo development. Our preliminary data in SCID mice with human lymphomas support this hypothesis.

Combination therapy with Rituxan and chemotherapeutic agents (eg, CHOP) has been evaluated in patients with indolent B-cell lymphoma; a 95% overall response (55%, complete remissions; 40%, partial remissions) was achieved6; it is too early to predict what impact this will have on long-term survival. In a previous in vitro study, it was shown that Rituxan sensitized a B-lymphoma cell line to several cytotoxic drugs.8 In the present study, we demonstrate that both monomers and homodimers of Rituxan can potentiate the in vitro effect of doxorubicin on partially resistant or MDR⁺tumor cells, but that homodimers are much more effective. The fact that Rituxan can render MDR⁺ cells sensitive to chemotherapy is in accord with our previous finding that an anti-CD19 mAb rendered MDR⁺ cells (P-gp⁺) sensitive to chemotherapeutic drugs in vitro.31 Taken together, these findings suggest exploring a strategy using homodimers followed by chemotherapy.

In conclusion, Rituxan homodimers are in all respects superior to monomers in the in vitro killing of the CD20⁺ cell lines used in these studies, and this superiority can be attributed to enhanced negative signaling by Rituxan. Homodimers might also have improved effector function, but this remains to be tested. Finally, homodimers were also highly effective in synergizing with an IT and in chemosensitizing cells to doxorubicin, suggesting that they might be useful for the treatment of patients with MDR⁺ lymphomas. What remains to be determined is whether homodimers will also show improved antitumor activity in vivo.

The authors thank Drs J. Uhr and V. Ghetie for their critical reviews of this work, and Ms S. Flowers, Ms T. Bowdler, and Ms R. Lewis for secretarial assistance.