Abstract
Chemical conjugates of anti-CD22 monoclonal antibodies and toxins have been used to treat CD22+ hematological malignancies. A new anti-CD22 recombinant immunotoxin RFB4(dsFv)-PE38, composed of the Fv portion of the monoclonal antibody RFB4 fused to a truncated form of Pseudomonas exotoxin A, is being developed to target CD22+ tumor cells. To explore the potential clinical utility of this recombinant toxin in treating patients with B-cell malignancies, the fresh cells of patients were incubated ex vivo with RFB4(dsFv)-PE38. Specific cytotoxicity was demonstrated in the malignant cells of 25 of 28 patients with a variety of B-cell malignancies, including acute and chronic lymphocytic leukemias and large cell, mantle cell, and follicular lymphomas. The IC50s, the concentrations necessary for 50% inhibition of protein synthesis, were 3–10 ng/ml in five patients and 10–50 ng/ml in seven patients. Cytotoxicity correlated with cell death upon direct examination of the malignant cells. Significant cytotoxicity was observed with cells containing as few as 350 CD22 sites/cell. A more active derivative of RFB4(dsFv)-PE38, RFB4(dsFv)-PE38KDEL, was produced and was slightly to more than 10-fold more cytotoxic toward patient cells and about twice as toxic to mice. Thus, RFB4(dsFv)-PE38 was specifically cytotoxic toward malignant cells from patients with B-cell leukemias. These data support the testing of RFB4(dsFv)-PE38 in patients with CD22+ leukemias and lymphomas, which is presently under way.
INTRODUCTION
B-cell lymphomas and leukemias together constitute a major public health problem in the United States and elsewhere; it is estimated that in 1999, 7800 new cases and 5100 deaths will be reported for CLL,2 and 56,800 new cases and 25,700 deaths will be reported for non-Hodgkin’s lymphomas (1) . The vast majority of patients with these disorders have tumor cells that display CD22, a 135-kDa phosphoglycoprotein adhesion molecule (2 , 3) . To treat B-cell malignancies, anti-CD22 antibodies or their Fab′ fragments have been chemically conjugated to toxins, and responses in animals bearing human lymphoma have been observed (4 , 5) . Clinical trials have been performed with immunotoxins containing a dgA chain conjugated to either the anti-CD22 antibody RFB4 (6 , 7) or its Fab′ fragment (2) . RFB4-Fab′-dgA gave partial responses in 5 of 13 patients, and RFB4-IgG-dgA gave a complete response and nine partial responses in 30 evaluable patients. The dose was limited by vascular leak syndrome, which in several cases was fatal (2 , 6 , 7) . Studies indicate that dgA is cytotoxic to endothelial cells, which become leaky even before cytotoxicity can be measured (8) . In contrast, such cells are resistant to a truncated form of PE, termed PE38, which contains amino acids 253–364 and 381–613 of the toxin (9) . Moreover, ligands can be fused to truncated PE to create recombinant chimeric toxins, which due to their relatively small molecular weight, might exit the vasculature more rapidly and have less chance of damaging endothelium in vivo.
To target PE to CD22+ malignant cells, chemical conjugates were initially made between the anti-CD22 MAbs LL2 or RFB4 and truncated PE (5 , 10 , 11) . These chemical conjugates showed potent cytotoxic and antitumor activity against human Burkitt’s lymphoma cell lines in vivo. The cloned variable domains of RFB4 were used to create the disulfide-stabilized (12 , 13) recombinant immunotoxin RFB4(dsFv)-PE38, which contains a disulfide bond linking VL with VH via cysteine residues engineered into the framework regions of VH and VL (14) . The VH domain is fused to PE38, which contains the translocating and ADP ribosylating domains of the toxin, but not the binding domain that binds to normal cells (15, 16, 17, 18, 19, 20, 21) .
RFB4(dsFv)-PE38 is very cytotoxic toward CD22+ cell lines and induces complete regressions of the human CD22+ cell line CA46 as s.c. xenografts in mice (14 , 22) . However, it is known that malignant cells in patients express lower numbers of CD22 sites per cell compared to cell lines. Moreover, it is possible that CD22+ cells from some patients may not metabolize the toxin properly to produce the active fragment that must be translocated to the cytosol. Indeed, B-cell leukemia cells have been reported to be resistant to toxins containing truncated PE without the KDEL carboxyl terminus (23) . Thus, studies of immunotoxin efficacy must include direct sensitivity tests of fresh patient cells. In the present study, we partially purified malignant cells from 28 patients with CLL and other B-cell malignancies and performed cytotoxicity assays of such cells with RFB4(dsFv)-PE38. Cytotoxicity was measured by a variety of methods to confirm that malignant cells were being killed. In most cases, we measured the number of CD22 sites per cell to determine if sensitivity could be predicted by the level of CD22 expression. A more active derivative, RFB4(dsFv)-PE38KDEL, which contains the amino acids KDEL (19 , 24) replacing REDLK at the carboxyl terminus of PE, was produced and compared with RFB4(dsFv)-PE38 with respect to cytotoxicity toward the fresh patient cells and toxicity toward mice. The relationship of time of incubation to cytotoxic activity was studied to explore the potential utility of continuous infusion to provide constant blood levels. Finally, we show in preliminary clinical data that low doses of RFB4(dsFv)-PE38 administered to patients with CLL cause decreases in circulating malignant cells, particularly in a patient with an IC50 of <10 ng/ml after ex vivo incubation of the cells.
MATERIALS AND METHODS
Plasmid Construction and Recombinant Toxins.
As shown in Fig. 1⇓ , RFB4(dsFv)-PE38, as previously described (14) , is composed of RFB4(VL) disulfide bonded to a fusion of RFB4(VH) and PE38 via engineered cysteine residues replacing Arg44 of VH and Gly100 of VL. The RFB4(VH)(R44C)-PE38 fusion is encoded by pEM15 and RFB4(VL)(G100C) by pEM16. RFB4(dsFv)-PE38KDEL, which is identical to RFB4(dsFv)-PE38 except that KDEL replaces amino acids 609–613 at the carboxyl terminus of PE38, is encoded by pEM16 and pRKRFH9K. The latter plasmid encodes RFB4(VH)(R44C)-PE38KDEL. pRKRFH9K was produced by ligating the 0.42-kb XbaI-HindIII fragment of pEM15, which encodes the RFB4(VH)(R44C) segment, to the 4.0-kb HindIII-XbaI fragment of pRK7H9K. The intermediate plasmid pRK7H9K encodes anti-Tac(VH)(G44C) fused to PE38KDEL (25) . The control molecules B1(dsFv)-PE38 (26) and LysPE38KDEL (5) were previously described, and the RFB4 MAb (27) was kindly supplied by Drs. Amlot (Royal Free Hospital, London, United Kingdom) and Vitetta (University of Texas-Southwestern, Dallas, TX).
Recombinant toxins and control molecules. RFB4(dsFv)-PE38 is composed of RFB4(VL) disulfide bonded to a fusion of RFB4(VH) and PE38. The disulfide is formed between engineered cysteine residues replacing Gly100 of VL and Arg44 of VH. PE38 is a truncated fragment of PE, which contains domain II (amino acids 253–364), part of domain Ib (amino acids 381–399), and domain III (amino acids 400–613), which includes the carboxyl terminal amino acids REDLK. In RFB4(dsFv)-PE38KDEL, the five carboxyl terminal amino acids REDLK are mutated to KDEL. The control molecule B1(dsFv)-PE38 (26) contains the same toxin domains as RFB4(dsFv)-PE38 but binds to a different antigen. LysPE38KDEL (5) contains the same toxin as RFB4(dsFv)-PE38KDEL, but instead of an Fv-binding domain, contains only the amino acids MANLAEEAFK preceding PE38. Thus, LysPE38KDEL is 39 kDa, whereas the other proteins are ∼63 kDa.
Immunotoxin Purification.
The RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL proteins used for the patient cells were similar to clinical grade material in terms of purity and endotoxin content and were produced as previously described (22) . Plasmids were expressed in Escherichia coli BL21/λDE3, and the inclusion bodies were purified from cell paste by washing with triton-X-100. Inclusion body protein was denatured, solubilized, and reduced in guanidine-dithioerythritol solution. To renature the recombinant immunotoxins, RFB4(VL)(G100C) was mixed either with RFB4(VH)(R44C)-PE38 to make RFB4(dsFv)-PE38 or with RFB4(VH)(R44C)-PE38KDEL to make RFB4(dsFv)-PE38KDEL. Solutions containing VL- and VH-toxin were refolded by 100-fold dilution into a redox buffer, and the dialyzed protein was purified by anion exchange and sizing chromatography.
Cell Lines and Patient Cells.
The CA46 Burkitt’s lymphoma cell line, available from the American Type Culture Collection, was kindly provided by Dr. I. Magrath. The WSU-CLL line was kindly provided by Dr. A. Al-Katib (Wayne State University, Detroit, MI). Blood was collected from patients as part of approved protocols at the NIH. Patients 1–24 represented consecutive patients with B-cell malignancies having malignant cells in the peripheral blood and were unselected for CD22 expression. Patients 25–28 were tested later. Blood in tubes containing sodium heparin was mixed 1:1 with PBS, layered over 15 ml of Ficoll (Pharmacia, Piscataway, NJ) in 50-ml tubes, and centrifuged to obtain mononuclear cells. The PBMCs were collected, washed with PBS, and resuspended in leucine-poor medium (Leucine-free RPMI, RPMI, and fetal bovine serum in a 88:2:10 ratio). The patient cells were incubated with recombinant toxins and control molecules in 100-μl aliquots in 96-well plates for either 3, 4, 6, or 7 days and then pulsed with[ 3H]leucine (2 μCi/well) for an additional 6–8 h. After freezing and thawing to liberate cells that may have attached, the cells were harvested onto protein-binding glass-fiber filters (Pharmacia-Wallac) and counted in a Betaplate (Pharmacia-Wallac) scintillation counter. To correlate inhibition of protein synthesis to cell death, cells in parallel were pulsed with WST-1 (Boehringer-Mannheim, Gaithersburg, MD) instead of[ 3H]leucine, which reacts with microsomal dehydrogenases and increases the absorbance at 450 nm in wells containing viable cells. Each well was pulsed with 10 μl of WST-1 reagent diluted in 40 μl of leucine-free RPMI, and like[ 3H]leucine incubated for 6–8 h prior to freezing the cells. To partially correct for background, the difference in absorbance at 450 and 680 nm was calculated for each well. The IC50 was defined as the concentration of recombinant immunotoxin necessary for 50% inhibition of either protein synthesis by [3H]leucine incorporation or cell viability by the WST-1 assay. Fifty percent inhibition was defined as the midpoint between maximal inhibition, as assessed by incubation with cycloheximide 10,000 ng/ml, and no inhibition, as assessed by incubation with media alone.
Binding Assay.
RFB4-IgG (150 μg) in a volume of 85 μl containing 175 mm sodium phosphate (pH 7.5) was treated with chloramine T (10 μg in 3 μl of H2O). Na[125I] (10 μl/1 mCi) was added, and after 2 min, the reaction was stopped by the addition of an aqueous solution of sodium metabisulfite (1.6 μl/83 μg). The radiolabeled MAb was purified on a Sephadex G-25 column (PD-10, Pharmacia) equilibrated and eluted with 0.2% HSA in PBS (HSA-PBS). The specific activity was between 3 and 6 μCi/μg. To keep radiolysis to a minimum, 125I-labeled RFB4-IgG was stored in the presence of 1% HSA by the addition of 25% HSA to the sample to increase the HSA concentration from 0.2 to 1%. Ficoll-purified PBMCs (106-107/well) from patients with leukemias were incubated at 4°C for 1.5–2 h in 100-μl aliquots of binding buffer (DMEM containing 0.1% BSA and 0.2% sodium azide) with 125I-labeled RFB4-IgG (0.125–4 nm, 2-fold dilutions) in the presence or absence of a 100-fold excess of unlabeled RFB4(dsFv)-PE38. The cells were then washed twice by centrifugation and counted.
FACS Analysis.
FACS was performed in the Clinical Pathology branch of the National Cancer Institute at the NIH clinical center in a laboratory approved for determining surface antigen expression for patient care purposes. Intensity of antigen expression was determined by College of American Pathologists guidelines. Like the cytotoxicity and binding assays described above, blood for FACS analyses was collected in sodium heparin-containing tubes.
Toxicity in Mice of Recombinant Immunotoxins Containing RFB4(dsFv).
Athymic nude mice, obtained from the National Cancer Institute (Frederick, MD), were injected i.v. via the tail vein with RFB4(dsFv)-PE38 or RFB4(dsFv)-PE38KDEL every other day for three doses (QOD ×3). Animals were observed for survival to determine the maximum tolerated doses of the two agents.
Treatment of Two CLL Patients with Low Doses of RFB4(dsFv)-PE38.
CLL patients were treated as part of an ongoing Phase I trial recently begun in patients with CD22+ B-cell malignancies. The Investigational New Drug was held by the Cancer Therapy and Evaluation Program of the National Cancer Institute, and the protocol with informed consent was approved by the National Cancer Institute Institutional Review Board and the Food and Drug Administration. Eligible patients had lymphoma CD22+ by immunohistochemistry or leukemia CD22+ by either FACS analysis or radiolabeled binding assay (>400 sites/cell). Patients had negative assays for neutralizing antibodies to RFB4(dsFv)-PE38. The immunotoxin diluted in 50 ml of saline containing 0.2% HSA was infused over 30 min three times QOD, and patients were eligible for retreatment at 25–28-day intervals in the absence of neutralizing antibodies or progressive disease.
RESULTS
RFB4(dsFv)-PE38 is a recombinant anti-CD22 disulfide-stabilized immunotoxin, which was previously shown to kill CD22+ cell lines in tissue culture and in vivo (14 , 22) , but its cytotoxicity toward fresh malignant cells from patients with CD22+ leukemias was unknown. To assess its clinical utility for treating such patients and to explore which patients might benefit most from such therapy, malignant CD22+ cells freshly obtained from patients were incubated with RFB4(dsFv)-PE38 and its more active derivative RFB4(dsFv)-PE38KDEL, and the cytotoxicity of these agents was studied.
Structure of Recombinant Toxins.
Shown in Fig. 1⇓ are schematic structures for RFB4(dsFv)-PE38 (14) and the new more active derivative RFB4(dsFv)-PE38KDEL. Both recombinant toxins containing RFB4(dsFv) have RFB4(VL) disulfide bonded to a genetic fusion of RFB4(VH) and truncated PE. The disulfide bond between VL and VH is created by a conversion of Gly100 of VL and Arg44 of VH to cysteine residues. The control molecules B1(dsFv)-PE38 (26) and Lys-PE38KDEL do not bind CD22 but contain the same toxin domains as RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL, respectively.
Cytotoxicity of RFB4(dsFv)-PE38 toward Fresh Malignant Cells.
To determine whether RFB4(dsFv)-PE38 would kill malignant cells in patients with CD22+ malignancies, the recombinant immunotoxin was incubated with the leukemic cells from such patients for 3 days at 37°C and pulsed with [3H]leucine for 6–8 h, and inhibition of protein synthesis was measured. Twenty-eight patients with B-cell malignancies whose cells survived in short-term culture were tested. Cytotoxicity curves on 24 of these patients, showing[ 3H]leucine incorporation into protein (cpm) at each concentration of RFB4(dsFv)-PE38 used, are displayed in Fig. 2⇓ . Dashed lines show the level of protein synthesis that is halfway between maximum inhibition of protein synthesis, as assessed by incubation with cycloheximide, and no inhibition, as assessed by incubation with media alone. The IC50s, which are the concentrations of recombinant protein necessary for 50% inhibition, are listed in Table 1⇓ and are shown in Fig. 2⇓ as the concentration where the cytotoxicity curve crosses the dashed line. All samples showed some sensitivity to RFB4(dsFv)-PE38, with the level of protein synthesis decreasing with increasing toxin concentration. Malignant cells from nearly half (12 of 28) of the patients were quite sensitive with IC50s of <50 ng/ml, and five of these were very sensitive with IC50s of <10 ng/ml. The most sensitive sample was from patient 24, with B-cell acute lymphoblastic leukemia, whose cells displayed an IC50 of 2.9 ng/ml. The protein synthesis of these cells was inhibited 99.8% with 1000 ng/ml of RFB4(dsFv)-PE38 (Fig. 2X).⇓ Cells from some patients, such as patients 5, 6, 10, and 12, were inhibited <50% at 1000 ng/ml of RFB4(dsFv)-PE38 (Fig. 2, E, F, J, and L).⇓ As shown in Table 1⇓ , malignant cells from six patients with non-B-cell leukemias were resistant to RFB4(dsFv)-PE38 at 1000 ng/ml, which was evidence of the B-cell specificity of the anti-CD22 immunotoxin.
Cytotoxicity of RFB4(dsFv)-PE38 toward fresh malignant cells. Ficoll-purified mononuclear cells from patients 1–24 (A-X, respectively) were incubated with RFB4(dsFv)-PE38 for 60 h and then [3H]leucine for 6–8 h. The graphs show protein synthesis as measured by cpm of[ 3H]leucine incorporated into protein. Points on the Y axes indicate [3H]leucine incorporation in the absence of immunotoxin. Dashed lines indicate 50% inhibition of protein synthesis, which is halfway between the level of incorporation in the absence of toxin and that in the presence of 10μ g/ml of cycloheximide. Error bars, SD from the means of triplicate experiments.
%Cytotoxicity of RFB4(dsFv)-PE38 toward fresh patient cells
Diagnoses included CLL, large cell lymphoma (LCL), follicular lymphoma (FL), acute lymphocytic leukemia (ALL), mantle cell lymphoma (MCL), and prolymphocytic leukemia (PLL). Control cells were obtained from patients with non-B-cell leukemias, including T-cell CLL (T-CLL), acute myelogenous leukemia (AML), and adult T-cell leukemia (ATL).
Correlation of Cytotoxicity to CD22 Sites Per Cell.
To determine the relationship between cytotoxicity by RFB4(dsFv)-PE38 and CD22 expression on the surface of the fresh malignant patient cells, the cells were incubated with increasing concentrations of 125I-labeled RFB4-IgG in the presence or absence of a 100-fold excess of unlabeled RFB4(dsFv)-PE38. An excess of RFB4(dsFv)-PE38 was used to determine nonspecific binding because it would not displace 125I-labeled RFB4-IgG from binding to cells via the Fc region or via any other part of the antibody other than the variable domains. Fig. 3⇓ shows representative Scatchard plots for patients 13, 21, and 16. The data are shown as RFB4 specifically bound in terms of sites per cell at different ratios of specific bound and free antibody. At maximum saturation, where the ratio of bound:free is minimal, the amount of RFB4(dsFv)-PE38 bound should equal the total number of CD22 sites per cell. The numbers of sites per cell, when possible to determine, are listed in Table 1⇓ . The binding constants (Kd), determined from the inverse-slopes of the Scatchard plots, were similar at 0.8–1.9 nm. As shown in Table 1⇓ , the most sensitive cells had the highest numbers of CD22 sites per cell. There were exceptions, as for patient 15 with mantle cell lymphoma, where the cells displayed only 350 sites/cell but were still quite sensitive with an IC50 of 39 ng/ml. All of the samples with> 2000 sites/cell were very sensitive to RFB4(dsFv)-PE38. Fig. 3D⇓ shows IC50 versus sites per cell for all of the assays on patients with IC50s of <1000 ng/ml, including some repeat assays on some patients. In addition, a point is shown corresponding to CA46 cells, which contain 22,000 sites/cell and have an IC50 of 0.44 ng/ml. Results are also plotted for the CLL cell line WSU-CLL, which is considered particularly representative of proliferating CLL cells because unlike other CLL cell lines, it is not EBV-transformed (28) . WSU-CLL cells have an IC50 of 4.7 ng/ml with 5500 CD22 sites/cell. The distribution shows that with 400-1900 sites/cell, there is no correlation between CD22 expression and sensitivity. However, the double-log plot in Fig. 3D⇓ does show that with CD22 levels> 2000 sites/cell, a correlation exists with an r2 value of 0.72. Thus, malignant CD22+ B-cells with 400-1900 sites/cell were sensitive to RFB4(dsFv)-PE38, but the IC50s varied widely between 10 and 1000 ng/ml, whereas cells with 4000–15,000 sites/cell were very sensitive. No correlation could be found linking sensitivity of malignant cells to RFB4(dsFv)-PE38 and their expression of other antigens, including CD5, CD10, CD19, CD20, CD23, CD25, and FMC7 (data not shown).
CD22 expression in patients. The number of sites per cell of CD22 was determined on the patients cells by incubating Ficoll-purified mononuclear cells at 5–10 × 106/well in 200-μl aliquots of 0.125–4 nm 125I-labeled RFB4 for 1–2 h at 4°C and then counting the washed cells. Nonspecific binding was determined by coincubating with a 100-fold excess of unlabeled RFB4(dsFv)-PE38. Scatchard plots are shown for patients 13 (A), 16 (B), and 21 (C), where the X axis is specific-bound in units of sites per cell and where the Y axis is bound/free in units of sites per cell per nm. Error bars, the SDs from the means of duplicate experiments. D, all experiments where sites per cell was determined simultaneously with cytotoxicity, often from more than one assay per patient. The regression line was drawn only using those assays with> 2000 sites/cell, where the correlation coefficient (r2) was 0.72.
Effect of the KDEL Carboxyl Terminal Mutation on the Cytotoxicity of Anti-CD22 Recombinant Immunotoxin.
To determine whether converting the carboxyl terminus of RFB4(dsFv)-PE38 from REDLK to KDEL, which has been reported to improve the cytotoxic activity of immunotoxins targeted toward a variety of antigens (11 , 19 , 24 , 29) , would result in increased cytotoxicity toward CD22+ cells, RFB4(dsFv)-PE38KDEL was constructed, purified, and tested on the patient cells in parallel to RFB4(dsFv)-PE38. As shown in Table 2⇓ , RFB4(dsFv)-PE38KDEL was usually severalfold more cytotoxic than RFB4(dsFv)-PE38. The largest difference was observed in patient 9. The malignant cells contained 1800 sites/cell and showed an IC50 of 8.1 ng/ml with RFB4(dsFv)-PE38KDEL compared to 125 ng/ml with RFB4(dsFv)-PE38. Malignant cells from patient 4, which displayed 6200 sites/cell, had an IC50 of 3.1 ng/ml with RFB4(dsFv)-PE38 and <1 ng/ml with RFB4(dsFv)-PE38KDEL. The IC65 of these cells was 8.9 ng/ml for RFB4(dsFv)-PE38 and 1.3 ng/ml for RFB4(dsFv)-PE38KDEL. Malignant cells from patient 5, which had only 360 CD22 sites/cell, showed IC50s of >1000 ng/ml to both recombinant toxins, but the IC35 of these cells was 720 ng/ml for RFB4(dsFv)-PE38 and 38 ng/ml for RFB4(dsFv)-PE38KDEL. In contrast, the least difference (<2-fold) in the two recombinant immunotoxins was seen toward malignant cells from patients 14 and 24, which had the highest CD22 expression (8600–14,800 sites/cell). Thus RFB4(dsFv)-PE38KDEL was more cytotoxic than RFB4(dsFv)-PE38, particularly when cytotoxicity was limited by the number of CD22 sites per cell. The difference in toxicity of the two immunotoxins was explored by injection of mice with different doses i.v. for three alternate days (QOD × 3). The calculated LD50 for RFB4(dsFv)-PE38 was 1030 μg/kg QOD × 3, compared to 600 μg/Kg QOD × 3 for RFB4(dsFv)-PE38KDEL, indicating a significant (P < 0.05 by χ2 analysis) but <2-fold difference in toxicity.
Effect of the KDEL carboxyl terminus of PE on cytotoxicity of RFB4(dsFv)-PE38 toward malignant CD22+ cells
Proportion of Cells Killed with RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL.
Shown in Fig. 4, A and C⇓ are cytotoxicity graphs of RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL on the malignant cells of patients 13 and 7, respectively. Not only are the IC50s lower for RFB4(dsFv)-PE38KDEL compared to RFB4(dsFv)-PE38, but also the maximum inhibition of protein synthesis is greater at high concentrations of RFB4(dsFv)-PE38KDEL compared to high concentrations of RFB4(dsFv)-PE38. In fact, the cytotoxic effect of 1000 ng/ml of RFB4(dsFv)-PE38KDEL is equal to that of cycloheximide in both patients. Thus it appears that a greater proportion of malignant CD22+ cells are potentially sensitive to RFB4(dsFv)-PE38KDEL than to RFB4(dsFv)-PE38, and this may contribute to the lower IC50s of RFB4(dsFv)-PE38KDEL.
Correlation of inhibition of protein synthesis with cell death. Cells of patients 13 (A and B) and 7 (C and D) were incubated with the indicated concentrations of RFB4(dsFv)-PE38 (○) or RFB4(dsFv)-PE38KDEL (•) for 3 days. To determine the inhibition of protein synthesis, the cells of patients 13 and 7 (A and C, respectively) were pulsed with[ 3H]leucine, harvested, and counted. To determine cell death, cells of patients 13 and 7 (B and D, respectively) were pulsed with 10 μl of WST-1 reagent (Boehringer-Mannheim), and the difference in A450 and A680 is shown.
Specificity of the Cytotoxic Activity of Recombinant Toxins Containing RFB4(dsFv).
To determine whether the cytotoxic activity of RFB4(dsFv)-PE38 or RFB4(dsFv)-PE38KDEL was specific and required both binding to CD22 and the action of the toxin within the fresh patient leukemic cells, several control experiments were performed. As shown in Fig. 5A⇓ using CLL cells from patient 22, the cytotoxicity of RFB4(dsFv)-PE38 was reversed by coincubation with an excess (5 μg/ml) of RFB4 MAb, indicating that the cytotoxic activity of RFB4(dsFv)-PE38 requires its binding to CD22. In Fig. 5B,⇓ the follicular lymphoma cells from patient 4 displayed potent sensitivity to RFB4(dsFv)-PE38KDEL and RFB4(dsFv)-PE38, but no sensitivity to Lys-PE38KDEL (5) , which has no binding domain, or to B1(dsFv)-PE38 (26) , which binds to Ley but not to CD22. This supports the specificity of the cytotoxicity of RFB4(dsFv)-PE38 or RFB4(dsFv)-PE38KDEL in that the malignant patient cells do not nonspecifically internalize toxins containing truncated PE. Finally, as shown in Fig. 5B,⇓ the malignant cells from patient 4 are resistant to RFB4-IgG even at 2500 ng/ml, indicating that although binding to CD22 is required for the cytotoxicity of RFB4(dsFv)-PE38KDEL and RFB4(dsFv)-PE38, binding alone is insufficient, and the activity of the toxin is also required to kill the target cells. The lack of cytotoxic activity of LysPE38KDEL was verified on malignant cells from all 28 patients. RFB4 was tested on all patient samples except from patient 12, and it was also not cytotoxic. B1(dsFv)-PE38 was tested against leukemic samples from all patients except 10–12, and all samples tested were resistant. Thus, RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL internalize specifically and not nonspecifically into those cells.
Determination of cytotoxic specificity. A, cells from patient 22 were incubated with the indicated concentrations of RFB4(dsFv)-PE38 in the presence (○) and absence (•) of an excess (5 μg/ml) of RFB4-MAb. B, cells from patient 4 were incubated with the indicated concentrations of RFB4(dsFv)-PE38 (○), RFB4(dsFv)-PE38KDEL (•), PE38KDEL (▾), B1(dsFv)-PE38 (▿), and RFB4-IgG (▪). After incubations (3 days), the cells were pulsed with [3H]leucine, harvested, and counted.
Analysis of the Maximum Protein Synthesis Inhibition Achieved in Patient Samples.
As shown in Fig. 2⇓ , the 24 patients differed with respect to the maximum inhibition of protein synthesis achieved with the highest concentrations of RFB4(dsFv)-PE38. For example, patient 4 had malignant cells with an IC50 of 3.1 ng/ml, but the protein synthesis inhibition with 100 or 1000 ng/ml reached a plateau of∼ 80%. Patient 13, with an IC50 of 9.7 ng/ml, reached a plateau of ∼60% inhibition of protein synthesis with 100 or 1000 ng/ml of RFB4(dsFv)-PE38. In contrast, patients 14 and 24, with IC50s of 3.8 and 2.9 ng/ml, respectively, had cells that were nearly completely inhibited by RFB4(dsFv)-PE38. Possible reasons for <100% inhibition include (a) contamination of the cell samples with nonmalignant cells, (b) contamination of the malignant cells with CD22-negative malignant cells, and (c) lack of sensitivity of at least some CD22+ malignant cells to RFB4(dsFv)-PE38. Table 1⇓ for most of the patients shows the percent of malignant cells in the PBMC sample tested. Based on this data, 33% of the PBMCs from patient 4 and 31% of the PBMCs from patient 21 are nonmalignant, which at least partly accounts for the plateaus seen in Fig. 2, D and U,⇓ respectively. In contrast, only 2% of the PBMCs of patient 13 and <3% of the PBMCs of patient 22 are nonmalignant, and yet the protein synthesis is inhibited only 60% for patient 13 and 80% for patient 22. This suggests that not all of the malignant cells in these PBMC samples are sensitive to RFB4(dsFv)-PE38.
Correlation of Inhibition of Protein Synthesis with Cell Death.
It has been shown previously that inhibition of protein synthesis inhibition in leukemias by immunotoxins leads to cell death (23 , 30) . Nevertheless, it remains formally possible that CD22+ leukemia cells, particularly those resistant to undergoing apoptosis, might respond to RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL with inhibition of protein synthesis and not cell death. To determine if this is the case, malignant PBMCs were exposed in parallel to WST-1, which reacts with microsomal dehydrogenases present in living cells and is a sensitive indicator of cell viability. As shown in Fig. 4, A and B,⇓ cells from patient 13 were incubated with either RFB4(dsFv)-PE38KDEL or RFB4(dsFv)-PE38 and either pulsed with [3H]leucine or exposed to WST-1. The solid line in the graphs indicates the level of protein synthesis or absorbance corresponding to maximum inhibition achieved after incubation with cycloheximide 10,000 ng/ml. The dashed line constitutes half-maximal inhibition. The IC50s of RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL toward cells from patient 13 are 14.1 and 2.9, respectively, by protein synthesis inhibition (Fig. 4A)⇓ and are 13.5 and 2.3 ng/ml, respectively, by the WST-1 assay (Fig. 4B),⇓ indicating that protein synthesis inhibition does correlate with cell death in these cells. Similarly, with cells from patient 7, The IC50s of RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL were 31 and 4.5 ng/ml, respectively, by inhibition of protein synthesis (Fig. 4C)⇓ and were 19 and 3.7 ng/ml, respectively, by WST-1. Thus, protein synthesis inhibition by the anti-CD22 recombinant immunotoxins correlates with cell death.
Relationship of Cytotoxicity to Time of Incubation with Recombinant Toxin.
The 3-day time point was chosen as the standard amount of time for incubating recombinant immunotoxins with B-leukemia cells based on previous data with anti-CD25 immunotoxins showing that the 3-day incubation with CLL cells results in improved cytotoxicity compared to 1- or 2-day incubations. To determine the effect of longer incubations on the cytotoxicity of RFB4(dsFv)-PE38 toward CD22+ leukemia cells, incubations were extended to 4 or to 6–7 days. As shown by the IC50s in Table 3⇓ , in 6 of 19 patents, the malignant PBMCs were at least severalfold more sensitive after just 1 extra day of incubation (4 days). PBMCs from patients 6 and 23 were over 10-fold more sensitive after 4 days compared to 3 days of incubation. The IC50 after 3 days of incubation with RFB4(dsFv)-PE38 was 14 ng/ml for both patients 8 and 13, and although it decreased only slightly after an additional day of incubation (11–12 ng/ml), the IC50 decreased substantially to 4–4.5 after 6–7 days of incubation. PBMCs from a minority of patients showed an increase in IC50 with time of incubation. In general, however, the time course studies showed a slight to severalfold improvement in sensitivity of the fresh CD22+ malignant cells with time of incubation, suggesting that if sustained levels of immunotoxin were achieved by continuous infusion, the antitumor response might improve.
Relationship of cytotoxicity with time of incubationa
Fig. 6A⇓ shows the cytotoxicity curves of RFB4(dsFv)-PE38 toward malignant PBMCs from patient 8, where the IC50s were 14, 11, and 4.5 ng/ml after 3, 4, and 7 days of incubation, respectively. After 30-min i.v. infusion in patients, preliminary data indicate that the half-life of other recombinant PE38-containing toxins having the same size and stability is 3–6 h (data not shown). To determine if a 4-h exposure to RFB4(dsFv)-PE38 would be sufficient for cytotoxicity, PBMCs from patient 8 were incubated with RFB4(dsFv)-PE38 for only 4 h, and the washed cells were incubated 7 days prior to pulsing with[ 3H]leucine. ⇓ shows that the IC50 was 60 ng/ml, about 4-fold higher than the 14 ng/ml IC50 observed after 3 days of incubation. Thus, RFB4(dsFv)-PE38 is cytotoxic after short-term exposure and should also be effective by intermittent bolus injection.
Influence of time of incubation on cytotoxicity of RFB4(dsFv)-PE38. Leukemic cells from patient 8 (A) or normal PBMCs (B) were incubated with RFB4(dsFv)-PE38 for 3 (♦), 4 (□), or 7 (▪) days before pulsing with[ 3H]leucine. The malignant cells of patient 8 in A were also incubated in parallel for 4 h, washed free of RFB4(dsFv)-PE38, and after 7 days of incubation, pulsed with[ 3H]leucine (⋄).
Confirmation of Malignant Cell Death as a Result of RFB4(dsFv)-PE38.
Many of the fresh patient PBMC samples contained >95% malignant cells (Table 1)⇓ , as determined by FACS analysis. Nevertheless, it remained theoretically possible that the inhibition of protein synthesis or WST-1 reaction in the samples was due solely to cytotoxicity toward small numbers of nonmalignant, possibly activated CD22+ cells mixed with metabolically inactive malignant cells. To determine whether the cytotoxicity of anti-CD22 immunotoxins was really directed toward malignant cells, PBMC samples containing high percentages of CLL cells were incubated with recombinant toxins and examined directly for viability either by trypan blue staining or by FACS analysis using propidium iodide staining. The cells were incubated for approximately 1 week before direct examination to allow sufficient time for protein synthesis inhibition to cause cell death. Fig. 7A⇓ shows that the PBMC sample from patient 27 displayed inhibition of [3H]leucine incorporation with an IC50 of 70 ng/ml for RFB4(dsFv)-PE38 and 6 ng/ml for RFB4(dsFv)-PE38KDEL. By inhibition of cell viability as assessed by trypan blue staining in Fig. 7C,⇓ the IC50s were 20 and 0.7 ng/ml for these respective immunotoxins. By performing FACS analysis on each 100-μl aliquot of cultured cells (106 cells) using propidium iodide for viability staining, the IC50 of RFB4(dsFv)-PE38 was 20 ng/ml. Similar results were obtained toward cells from patient 28 (Fig. 7, B, D, and F)⇓ and PBMCs from other patients (data not shown). Interestingly, FACS analysis using annexin staining showed that the CLL cells were resistant to apoptosis (data not shown), indicating that the protein synthesis inhibition due to RFB4(dsFv)-PE38 eventually resulted in cell death without apoptosis. Thus the malignant cells, regardless of their terminal differentiation, undergo cell death as a result of exposure to the anti-CD22 immunotoxins. Moreover, the control molecules RFB4-IgG, B1(dsFv)-PE38 or PE38KDEL, tested above in Fig. 5⇓ , showed no significant cytotoxicity (Fig. 7, C-D),⇓ indicating that the killing of CLL cells by RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL was specifically mediated by both binding and toxin activity.
Direct examination of CLL cells for cytotoxicity. PBMCs containing 95–99% CLL from patients 27 (A) and 28 (B) were incubated with RFB4(dsFv)-PE38 (○), RFB4(dsFv)-PE38KDEL (•), PE38 (▾), B1(dsFv)-PE38 (▿), and RFB4-IgG (▪), and [3H]leucine incorporation determined as in Fig. 2⇓ . The same assays were repeated in C and D, respectively, except that percent cell viability was determined by trypan blue exclusion. The percent viability after RFB4(dsFv)-PE38 (○) was also determined by FACS analysis on cells from patients 27 (E) and 28 (F) using propidium iodide.
Sensitivity of Normal PBMCs to RFB4(dsFv)-PE38.
Despite the known small percentage of B cells making up normal PBMCs, such cells were tested under the same conditions as the patient cells for sensitivity to RFB4(dsFv)-PE38. As shown in Fig. 6B,⇓ there is a slight decrease in leucine incorporation at 1000 or 10,000 ng/ml of RFB4(dsFv)-PE38 of borderline significance at 3 or 4 days, and an IC50 of 2600 ng/ml was observed after 7 days of incubation. Thus the malignant CD22+ cells judged to be sensitive to RFB4(dsFv)-PE38 were much more sensitive than were normal cells, although it is expected that some normal B-cells, which make up a small percentage of the lymphocytes in the peripheral blood, may be susceptible to the cytotoxic effects of RFB4(dsFv)-PE38.
Decreases in Circulating CLL Cells in Patients Treated with Low Doses of RFB4(dsFv)-PE38.
The ex vivo cytotoxicity results shown above for recombinant toxins toward CD22+ leukemic cells do not necessarily indicate that the proliferating malignant cells, which may be much less differentiated than those in the peripheral blood, are actually targeted. These proliferating CLL cells cannot be isolated even from bone marrow because the vast majority of malignant cells in CLL marrows are similar to circulating CLL cells. The cytotoxicity of RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL toward proliferating cells was tested using the cell line WSU-CLL (28) , which was sensitive with IC50s of 4.7 and 2.6 ng/ml, respectively. This CLL cell line is probably more representative of proliferating CLL cells than other CLL cell lines that have undergone viral transformation. Nevertheless, direct determination of cytotoxicity toward proliferating malignant patient cells requires clinical testing in leukemic patients. RFB4(dsFv)-PE38 has begun clinical testing in patients with CD22+ malignancies, and preliminary data are available at early nontoxic dose levels. As shown in Fig. 8A,⇓ cells from patient 25 were very sensitive ex vivo to RFB4(dsFv)-PE38 with an IC50 of 5.5 ng/ml This patient showed repeated reductions of leukemic cells when treated with 30-min infusions i.v. at 6–10 μg/kg three times QOD (Fig. 8B).⇓ The maximal reduction was 98% during cycle 3. As shown in Fig. 8C,⇓ cells from patient 26 were not as sensitive as those from patient 25, and the decrease in circulating malignant cells was less pronounced and more transient. Plasma levels of RFB4(dsFv)-PE38 were assessed by cytotoxicity assay using the cell line Raji as described previously (31) . For the first two cycles, peak plasma levels were 42–51 ng/ml for patient 25 and 57–79 ng/ml for patient 26, and for both patients, peak levels during the third cycle were 90–93 ng/ml. The half-lives were ∼1–2 h. These experiments suggest that cytotoxicity ex vivo may lead to antitumor activity in patients, but additional clinical data are needed for determining both the mechanism of malignant cell reductions in the patients and also the relationship between ex vivo and in vivo results in humans.
Reduction of CLL counts in patients receiving low doses of RFB4(dsFv)-PE38. PBMCs containing 98% of CLL cells from patients 25 (A) and 26 (C) were incubated in A and C ex vivo with RFB4(dsFv)-PE38 (○) and [3H]leucine incorporation determined. Patients 25 (B) and 26 (D) were then treated with RFB4(dsFv)-PE38 at doses of 6μ g/kg i.v. QOD × 3 for cycles 1 and 2 and 10 μg/kg i.v. QOD × 3 for cycle 3. CLL counts were determined by morphological examination of peripheral blood or FACS analysis at the indicated time points.
DISCUSSION
Our goal in the present study was to determine whether the anti-CD22 recombinant immunotoxin RFB4(dsFv)-PE38, which was recently reported to possess cytotoxic activity toward CD22+ cell lines (14) , would also be cytotoxic toward fresh malignant cells from patients. We found that leukemic PBMCs from about half of the patients with B-cell leukemias are quite sensitive to RFB4(dsFv)-PE38, although the cells from several patients contained only 350–680 CD22 sites/cell. We also showed that the cytotoxicity is specific for CD22+ cells, that cytotoxicity is usually related to both the levels of CD22 expression and time of incubation, and that the recombinant immunotoxins cause the death of fresh malignant cells.
The analysis of the sensitivity of malignant cells directly obtained from patients is a key part of an agent’s preclinical development because fresh malignant cells are often much less sensitive than cell lines, and such studies might be able to identify patients who would be more likely to respond. Other than the 1989 report of the cytotoxicity of HD6-Saporin and HD39-Saporin toward fresh malignant cells from CLL patients (32) , we were unable to find previous data showing that anti-CD22 immunotoxins could effectively kill freshly obtained malignant cells from patients. Our results are contrary to the ex vivo study of fresh CLL cells showing that the KDEL carboxyl terminus was essential in targeting an anti-CD25 recombinant immunotoxin to CLL cells (23) . The present study shows that RFB4(dsFv)-PE38, which is less toxic in mice than RFB4(dsFv)-PE38KDEL, is cytotoxic toward CLL cells from most patients. It is likely that CLL cells from some patients internalize CD22 more efficiently than CD25 or that after internalization by CD22 the toxin is better able to traffic intracellularly toward the cytosol.
Predicting Which Patients Have Sensitive Cells.
When possible, malignant cells from patients were analyzed for CD22 expression by both radiolabeled binding assays and by FACS analysis. Sensitivity to RFB4(dsFv)-PE38 was associated with more than dim positivity by FACS analysis and >4000 sites/cell by radiolabeled binding assay, although strictly adhering to either of these criteria will cause one to miss a significant number of patients who have immunotoxin-sensitive malignant cells. Patients whose cells had >4000 sites/cell were all sensitive, and such patients would make the most promising candidates for clinical testing. Only data from clinical trials will appropriately address this issue and allow more effective patient selection for later, more advanced trials.
RFB4(dsFv)-PE38KDEL, a Recombinant Anti-CD22 Immunotoxin with Higher Activity.
Our results showed that in most of the patient samples, RFB4(dsFv)-PE38KDEL was >4-fold more cytotoxic to patient samples and nearly 2-fold more toxic to mice than RFB4(dsFv)-PE38 (Table 2)⇓ . Although this suggests that RFB4(dsFv)-PE38KDEL may be more optimal for testing in patients than RFB4(dsFv)-PE38, this conclusion is premature for the following reasons: First, it has been found with other recombinant toxins, such as those containing anti-Tac(Fv), that although KDEL imparts only slightly more toxicity to mice, the toxicity in monkeys is more severe. Secondly, although a variety of PE38-containing immunotoxins have been or are being given to patients, PE38KDEL-containing toxins have never been administered systemically to patients, and the potential human toxicity is thus unknown. Fig. 4, A and C,⇓ suggests that the KDEL carboxyl terminus, which improves intracellular binding of the carboxyl terminus to the KDEL receptor by 100-fold compared to REDL (19) , allows the toxin to kill essentially all of the CD22+ cells rather than just a majority. However, toxicity to normal organs might also be more pronounced or widespread. Finally, it is possible that the proliferating CD22+ malignant cells, which usually make up a low percentage of those malignant cells in the peripheral blood, are just as sensitive to RFB4(dsFv)-PE38 as to RFB4(dsFv)-PE38KDEL. The fact that RFB4(dsFv)-PE38 and RFB4(dsFv)-PE38KDEL have nearly equal activity toward cells from the B-ALL cells from patient 24 supports this hypothesis because it is a less differentiated tumor than CLL where a higher percentage of the PBMCs are proliferating.
Relevance of Time of Incubation to Method of Administration.
Because maximum toxicity in animal models to recombinant immunotoxins is almost always observed by 48 h after the dose, we have adopted a dosing scheme of three times QOD to begin Phase I testing with recombinant immunotoxins targeting other antigens, including B3-Lys-PE38 (LMB-1), B3(Fv)-PE38 (LMB-7), and B3(dsFv)-PE38 (LMB-9), which target Ley, e23(dsFv)-PE38 (Erb-38), which targets Erb-B2, and anti-Tac(Fv)-PE38 (LMB-2), which targets CD25 (12 , 33, 34, 35, 36) . The data from Fig. 6A⇓ suggest even a short exposure of RFB4(dsFv)-PE38 to malignant patient cells, as would be observed in the QOD × 3 bolus regimen, is sufficient to result in significant cytotoxicity. Perhaps continuous infusion for 5–7 days would result in much better efficacy by allowing improved penetration of the recombinant toxin into extravascular masses. It is interesting that continuous infusion applied to other immunotoxins has not yielded a dramatic improvement in therapeutic index compared to bolus injection (7 , 37) . Nevertheless, because previous agents have had longer half-lives than RFB4(dsFv)-PE38 and because cells from many patients are much more sensitive to RFB4(dsFv)-PE38 with prolonged incubation times, it would be reasonable to consider continuous infusion as an option in clinical trials of this new agent.
In this report, we have shown that RFB4(dsFv)-PE38, a stable 63-kDa recombinant anti-CD22 immunotoxin, is cytotoxic to fresh malignant cells from patients with a wide range of B-cell leukemias. Recent in vivo experiments with RFB4(dsFv)-PE38 show very high tolerance in monkeys (2 mg/kg i.v. QOD × 3) without clinically significant toxicity to normal tissues.3 In these studies, peak serum levels were obtained, which are >1000-fold higher than the IC50s of malignant cells from half of the patients shown in Table 1⇓ . Based on preclinical toxicity and the efficacy data shown here, RFB4(dsFv)-PE38 has begun clinical testing (under the name of BL22) in patients with B-cell lymphomas and leukemias.
Acknowledgments
We thank Dr. Ellen Vitetta and Dr. Peter Amlot for providing RFB4 monoclonal antibody for these studies. We also thank Dr. Al-Katib for providing the WSU-CLL cell line.
Footnotes
-
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.
-
↵1 To whom requests for reprints should be addressed, at Laboratory of Molecular Biology, National Cancer Institute, NIH, 37/4E16, 37 Convent Drive, MSC 4255, Bethesda, MD 20892. Phone: (301) 496-4797; Fax: (301) 402-1344.
-
↵2 The abbreviations used are: CLL, chronic lymphocytic leukemia; dgA, deglycosylated ricin A; PE, Pseudomonas exotoxin; MAb, monoclonal antibody; PBMC, peripheral blood mononuclear cell; HSA, human serum albumin; FACS, fluorescent-activated cell sorting.
-
↵3 Unpublished observations.
- Received October 18, 1999.
- Revision received December 15, 1999.
- Accepted December 16, 1999.
References
Volume 6, Issue 4
