Purpose: Although most children with B-lineage acute lymphoblastic leukemia (ALL) and non–Hodgkin lymphoma are cured, new agents are needed to overcome drug resistance and reduce toxicities of chemotherapy. We hypothesized that the novel anti-CD22 immunotoxin, RFB4(dsFv)-PE38 (BL22, CAT-3888), would be active and have limited nonspecific side effects in children with CD22-expressing hematologic malignancies. We conducted the first preclinical and phase I clinical studies of BL22 in that setting.
Experimental Design: Lymphoblasts from children with B-lineage ALL were assessed for CD22 expression by flow cytometry and for BL22 sensitivity by in vitro cytotoxicity assay. BL22 was evaluated in a human ALL murine xenograft model. A phase I clinical trial was conducted for pediatric subjects with CD22+ ALL and non–Hodgkin lymphoma.
Results: All samples screened were CD22+. BL22 was cytotoxic to blasts in vitro (median IC50, 9.8 ng/mL) and prolonged the leukemia-free survival of murine xenografts. Phase I trial cohorts were treated at escalating doses and schedules ranging from 10 to 40 μg/kg every other day for three or six doses repeated every 21 or 28 days. Treatment was associated with an acceptable safety profile, adverse events were rapidly reversible, and no maximum tolerated dose was defined. Pharmacokinetics were influenced by disease burden consistent with rapid drug binding by CD22+ blasts. Although no responses were observed, transient clinical activity was seen in most subjects.
Conclusions: CD22 represents an excellent target and anti-CD22 immunotoxins offer therapeutic promise in B-lineage hematologic malignancies of childhood. Clin Cancer Res; 16(6); 1894–903
- Acute lymphoblastic leukemia
- Non–Hodgkin lymphoma
- Childhood cancer
B-lineage hematologic malignancies remain a leading cause of cancer-related mortality in pediatrics and current therapies are associated with a wide array of toxicities. New agents are needed to overcome drug resistance and reduce nonspecific adverse effects. We report the results of the first preclinical studies and phase I clinical trial of a novel anti-CD22 immunotoxin, RFB4(dsFv)-PE38, in the setting of childhood hematologic malignancies. An acceptable toxicity profile and transient clinical activity was observed. This phase I clinical trial serves as proof-of-principle that this immunotoxin construct is cytotoxic to chemotherapy-resistant CD22+ blasts and that it can be administered to children in doses that achieve serum levels which exceed the expected in vitro IC50. The trial established a dose and schedule for subsequent testing of RFB4(dsFv)-PE38 with a modified Fv sequence that confers higher binding affinity for CD22. Future trials are planned in combination with standard chemotherapy agents.
There has been great progress in the curative treatment of hematologic malignancies in childhood (1). Acute lymphoblastic leukemia (ALL), the most common pediatric cancer, is highly curable and 80% of children with B-precursor ALL achieve long-term relapse-free survival (2). However, the outlook remains guarded for individuals with certain high-risk features at diagnosis and for those who relapse. Hematologic malignancies remain a leading cause of cancer-related mortality in pediatrics (3, 4). Additionally, current therapies carry the risk of treatment-associated morbidity and mortality (5, 6). Thus, novel approaches that can overcome chemotherapy resistance and decrease nonspecific toxicities are needed to improve the outcome for children with hematologic malignancies.
CD22 is a B-lineage–restricted surface molecule that modulates B-cell receptor signaling and mediates cellular adhesion (7). Immunotoxins are proteins that consist of two primary components: a targeting moiety responsible for cell binding and a bacterial or plant toxin that induces cell death upon internalization (8). The recombinant immunotoxin, RFB4(dsFv)-PE38 (BL22, CAT-3888), contains the variable domains of the anti-CD22 monoclonal antibody (MoAb) RFB4 fused to a 38-kDa fragment of Pseudomonas aeruginosa exotoxin A (PE; refs. 9, 10). BL22 is cytotoxic toward CD22+ cell lines and malignant cells from patients, and it is active in murine xenograft models (11–13). In phase I and II human clinical trials, BL22 induced complete remissions in adults with hairy cell leukemia resistant to purine analogue therapy and exhibited a safety profile conducive to continued development (14–16). We hypothesized that this novel anti-CD22 immunotoxin would be active and have limited nonspecific side effects in children with CD22-expressing hematologic malignancies. We conducted the first preclinical studies and phase I clinical trial of BL22 for pediatric ALL and non–Hodgkin lymphoma.
Materials and Methods
Fresh bone marrow or peripheral blood blasts were collected from children with B-lineage ALL.
In vitro cytotoxicity
Seventy-two–hour in vitro cytotoxicity assays were conducted using protein synthesis inhibition ([3H]leucine incorporation) and colorimetric viability (WST-1). Results were expressed as the IC50 value (concentration of BL22 required to reduce viability/protein synthesis by 50% compared with nontreated controls) as previously described (12).
Flow cytometry and antigen binding site determination
CD22 antigen expression and absolute peripheral blast counts were determined by flow cytometry. Antigen site density was quantified by determining the anti-CD22 antibody binding capacity per cell (17) using the BD Biosciences QuantiBRITE system for fluorescence quantitation.
Cells from the human ALL line EU-1 were used for xenograft studies. This cell line was established and authenticated as previously described (18), and phenotype was reconfirmed by serial flow cytometric analyses including at the time of the xenograft studies. EU-1 cells were injected by tail vein into 5-wk-old female C.B-17 severe combined immunodeficient mice−/− (107 cells per mouse). Seventy-two hours after injection, cohorts of 10 (treatment) or 5 (control) xenografts were treated with BL22 at dose levels of 1.5, 3, or 4.5 μg/dose, or control agents through i.p. injection every other day for nine doses. Xenograft recipients were euthanized at hind-limb paralysis and evaluated for the presence of human leukemia by histopathology.
BL22 and control agents
Recombinant immunotoxins were produced as previously described (11, 12, 19). Clinical grade BL22 for human use was produced by the Monoclonal Antibody and Recombinant Protein Facility of the National Cancer Institute (NCI) and provided by MedImmune Cambridge (formerly Cambridge Antibody Technology, Inc., a subsidiary of MedImmune, LLC). Control reagents for preclinical studies included PBS, anti-CD22 MoAb (RFB4-IgG provided by the Developmental Therapeutics Program of the NCI), and the anti–CD25-PE immunotoxin anti–Tac(Fv)-PE38 (LMB2).
Phase I clinical trial
A phase I trial was conducted at the NIH Clinical Center (ClinicalTrials.gov no. NCT00077493).
Individuals between 6 mo and 24 y of age with relapsed or refractory ALL or non–Hodgkin lymphoma who had exhausted available curative therapies were eligible. Measurable or evaluable disease that was CD22+ by flow cytometry (>30%) or immunohistochemistry (>15%) was required. Subjects must have been off other investigational agents for at least 30 d and systemic chemotherapy for at least 14 d. Individuals with isolated testicular relapse and active central nervous system malignancy were excluded, but concurrent prophylactic intrathecal chemotherapy was permitted. Concurrent corticosteroids were allowed for patients who had been previously treated with such to reduce the likelihood of rapid disease progression during the time required to travel to the NIH and undergo eligibility screening. The doses of corticosteroids could not have been increased for at least 7 d before trial enrollment and patients were required to have persistent or progressive (i.e., not decreasing) disease burden. Eligibility required aspartate aminotransferase and alanine aminotransferase (ALT) of five or more times the upper limit of the normal, total bilirubin of ≤2 mg/dL, and age-adjusted normal creatinine.
BL22 was administered i.v. over 30 min every other day for three or six doses. Intravenous hydration was initiated 6 h before BL22 using 5% dextrose and 0.45% sodium chloride at a rate of 90 mL/m2/h. Premedication consisted of acetaminophen, diphenhydramine, and ranitidine.
Cohorts of three to six subjects were treated at doses of 10, 20, 25, 30, and 40 μg/kg/dose. During the course of the trial, the escalation scheme was modified to shorten the treatment interval from 28 to 21 d and to increase the number of doses per cycle from three to six (Table 1). If dose-limiting toxicity (DLT) was encountered in one subject in a cohort of three, an additional three patients were entered at that dose level. If DLT occurred in two or more patients at any given dose level, the maximum tolerated dose was considered to have been exceeded. Re-treatment required the recovery of BL22-related toxicity to lower than grade 2 and the absence of DLT, high-titer neutralizing antibodies, and disease progression.
Toxicity grading and definition
Version 3.0 of the NCI Common Terminology Criteria for Adverse Events6 was used for toxicity and adverse event reporting. DLT was defined as a nonhematologic toxicity of grade 3 or more with the following exceptions: tumor lysis syndrome, abnormal electrolytes responding to supplementation, grade 3 hepatic dysfunction with resolution before the next cycle, and grade 3 fever, hypertriglyceridemia, hypercholesterolemia, and hypoalbuminemia in the absence of vascular leak syndrome. Hematologic DLT was defined as grade 4 hematologic toxicity (with the exception of lymphocytes) lasting >5 d or any platelet transfusion. Subjects with abnormal blood counts due to bone marrow infiltration were not evaluable for hematologic toxicity.
Plasma levels of BL22 were determined by incubating dilutions of plasma with Raji cells and comparing cytotoxicity as assessed by [3H]leucine incorporation to that obtained by a BL22 standard as previously described (15). Samples were obtained before, immediately after, and at 1/2, 1, 1 1/2, 2, 4, 8, 12, and 18 h after each day 1 dose, and before and immediately after subsequent doses. Area under the curve was calculated on the first dose of the cycle from either a monoexponential or biexponential model based on Aikake's rule (20). For biexponential kinetics, the β half-life was used in analyses.
Neutralizing antibody assay
To assay for the presence of neutralizing antibodies, mixtures containing 90% serum and 10% BL22 (final BL22 concentration, 1,000 ng/mL) were incubated at 37°C for 15 min, diluted, and cultured with Raji cells, and the percentage of inhibition in the presence of subject serum was calculated as previously described (15). High-titer neutralizing antibodies were defined as levels that resulted in >75% neutralization of 1,000 ng/mL of BL22 in this assay.
For mice evaluated in preclinical studies, the probability of survival as a function of time, according to treatment administered, was determined by the Kaplan-Meier method with a log-rank test used to determine the difference of the probability of survival between the three pooled control groups and the combined BL22-treated groups. For the dose effect, pooled data from the two higher dose level groups were analyzed compared with the lowest dose. Results were not adjusted for multiple comparisons.
The relationship between clearance and peripheral blast count per microliter in subjects treated on the clinical trial was determined using Spearman rank correlation, with ∣r∣ > 0.70 indicating a strong correlation and the P value indicating a result from a test of whether r = 0. Changes in peak BL22 levels between the first and last dose were evaluated using a paired t test after verifying that the differences followed a normal probability distribution. Although patients received a varying number of doses, the difference is meant to illustrate the effect associated with maximal dosing on the trial. All P values were two-tailed and are reported without adjustment for multiple comparisons.
Human subjects protections
All studies were approved by the Investigational Review Boards of the NCI (preclinical studies, clinical trial NCT00077493) or Emory University School of Medicine (preclinical studies).
To assess the frequency of CD22 expression in B-lineage ALL, blasts from 93 children with pre-B or Burkitt-type ALL were evaluated by flow cytometry. All cases were CD22+. CD22 expression distribution and density were further quantified in 54 of these cases. One hundred percent of the blasts within individual cases expressed CD22 in 52 of 54 cases. Minor populations of blasts without demonstrable CD22 expression were detected in two individuals (∼80% and 90% CD22+). Determination of antibody binding capacity per cell revealed that the average CD22 site density within cases ranged from 451 to 16,523 sites per blast (median, 4,062), and only four cases showed <1,000 CD22 sites per cell (Fig. 1).
In vitro cytotoxicity assays were conducted on blast samples obtained from 42 children with B-precursor ALL (Fig. 1). BL22-induced killing was observed in all samples, with IC50s that ranged from 0.5 to 100 ng/mL (median, 9.8).
Murine xenografts treated with BL22 had a significant prolongation of leukemia-free survival (Fig. 1). Treatment cohorts were analyzed against the controls (using grouped data for both), and differences (treatment versus control) were significant (P < 0.001). To evaluate dose response, 3 and 4.5 μg cohorts were combined after determining that they had similar survival, and were analyzed compared with the 1.5 μg cohort. The difference was significant (P < 0.05).
Phase I trial
Twenty-three subjects ranging in age from 3 to 22 years (median, 13) were treated on the phase I trial. Twenty had ALL with marrow relapse, and one each had ALL with extramedullary relapse, stage 4 lymphoblastic lymphoma, and Burkitt lymphoma. All had been heavily pretreated, having received a median of four prior regimens (range, 2-9) and 20 were refractory to chemotherapy at the time of enrollment (Table 1). Twenty-two subjects completed at least one cycle of BL22 and one subject at dose level 7 received three of six doses. All subjects were evaluable through the DLT evaluation period for the primary study end points (i.e., toxicity, pharmacokinetics, and immunogenicity). During the course of the trial, the treatment schedule was amended to escalate the dose intensity (Table 1) based on safety and activity data from preceding cohorts. Cohort 5 was terminated early to increase the number of doses administered per cycle from three to six.
BL22 treatment was associated with an acceptable safety profile. No subject experienced infusion reactions, allergic events, vascular leak syndrome, or hemolytic uremic syndrome. All adverse events were self-limited and most were grades 1 and 2 (Fig. 2). Grades 3 and 4 events were rapidly reversible. Two subjects treated at dose level 7 experienced grade 4 ALT elevation of 1 to 2 days in duration, which met the original protocol definition of DLT. However, both of these resolved to levels required for ongoing treatment as scheduled supporting a revision in the DLT definition.
There was a dose-related increase in the peak plasma levels of BL22 with wide interpatient variation in pharmacokinetic parameters (Table 2). BL22 was rapidly cleared from the circulation with a plasma half-life on cycle 1 that ranged from 35 to 248 minutes (median, 106). Cycle 1 clearance correlated with increased peripheral blast count (Spearman correlation, r = 0.73; 95% confidence interval, 0.45-0.88; P < 0.0001; Fig. 3). Peak plasma levels increased with progressive dosing in individuals who experienced blast reduction treated at the highest doses (30 and 40 μg/kg; P = 0.01; data not shown). This was in contrast to subjects without peripheral blast count reduction. Thus, pharmacokinetics seemed to be influenced by disease burden, consistent with rapid BL22 binding by CD22+ blasts.
Only 3 of 23 subjects (13%) developed neutralizing antibodies. One had preexisting low-titer antibodies that increased to 93% after cycle 1. Two developed de novo antibodies with 78% neutralization after cycle 1 and 65% after cycle 3.
Most subjects had high disease burden and rapidly progressive disease at the time of protocol enrollment. No responses were observed, however, transient clinical activity was seen in 16 of 23 subjects (70%), which in some cases was dramatic and clinically significant (Table 1; Fig. 4). For example, four subjects had >2 log10 reduction in circulating blasts and four had recovery of normal blood counts. Decreased blast infiltration of bone marrow (6) and extramedullary sites (3) were also observed. Notably, the most significant clinical activity was seen at the highest dose levels, including the four individuals with the largest reduction in peripheral blast counts (two each treated at dose levels 6 and 7) and all those with normalization of blood counts (treated at doses of 30 μg/kg or higher). Of the seven subjects without disease progression, two were ineligible to remain on the study (neutralizing antibodies, grade 4 ALT elevation), two chose to discontinue treatment after two to three cycles, and three developed progressive disease after two to three cycles, one of which was associated with the development of neutralizing antibodies.
Despite significant progress in the curative treatment of childhood hematologic malignancies, relapse remains one of the greatest challenges in pediatric oncology (3). Furthermore, survivors have life-long risks of treatment-associated morbidity and mortality (6). New therapeutic approaches are needed to overcome chemotherapy resistance and to reduce side effects (23).
CD22 is rapidly internalized upon antibody or immunotoxin binding (24). As shown, this antigen is expressed in high frequency in childhood ALL. Unconjugated MoAbs may induce cytotoxicity by direct and indirect (e.g., immune mediated) mechanisms, the latter of which are expected to be defective in individuals with ALL (25). Unconjugated MoAb against CD22 (epratuzumab) has recently been studied in childhood ALL and its activity as a single agent in the setting of relapsed ALL seems to be limited (26). Notably, anti-RFB4 control showed no activity against ALL xenografts or in cytotoxicity assays with primary samples from children with ALL (Fig. 1).
The activity of MoAbs can be dramatically increased by linkage to toxic moieties. Plant and bacterial toxins cause cellular cytotoxicity through the inhibition of protein synthesis after internalization. These are highly potent and active in minute quantities, such that even a single molecule in the cytoplasm is sufficient to kill a cell (27). There have been limited studies of immunotoxins in childhood hematologic malignancies, and previously evaluated agents have been associated with severe adverse events and a high incidence of immunogenicity (28, 29). The clinical development of immunotoxins in general has been hampered by nonspecific toxicities, immunogenicity, and production complexities. Serial modifications in the Pseudomonas-based immunotoxin constructs used at the NCI have reduced nonspecific toxicities, increased stability, enhanced tissue penetration, and improved targeted cellular killing (30).
BL22 is a potent immunotoxin that targets CD22, which as shown, is expressed in relatively high density on the surface of 100% of the blasts in the vast majority (96%) of cases of childhood B-lineage ALL. BL22 was shown to have clinical activity with acceptable toxicity in adults with relapsed and refractory hairy cell leukemia, in which a maximum tolerated dose of 40 μg/kg every other day for 3 doses every 28 days was defined (15). This pediatric phase I trial extends those observations and establishes that activity can be achieved in highly resistant childhood ALL with acceptable toxicity. Notably, BL22 was tolerated at a greater dose intensity (i.e., six doses every other day every 21 days) compared with adults, and hemolytic uremic syndrome, which was the DLT in adults, was not observed. Importantly, antileukemia activity was seen at all dose levels; however, clinical benefits in this highly refractory population were modest and transient at the doses tested. There are several possible explanations for the limited observed activity. Higher doses are likely required to achieve maximal benefit. Furthermore, although peak levels at the upper doses exceeded the concentrations required for in vitro cytotoxicity, drug exposure was limited in most subjects due to the rapid clearance associated with large disease burden. CD22 expression has been shown to be a determinant of response to BL22 in vitro (12), although there was no obvious influence of antigen density on clinical activity in this trial (Table 1). However, small numbers preclude definitive conclusions in this regard, and it is notable that all subjects without progressive disease had site densities that exceeded 3,000 sites per cell, whereas 6 of 16 with progressive disease had lower levels of expression.
This trial shows that BL22 can be administered at doses of up to 40 μg/kg every other day for 6 doses in children with ALL. No maximum tolerated dose was defined. Although two subjects treated at the highest dose level developed brief grade 4 ALT elevations, this was not dose-limiting given the short duration (1-2 days). We subsequently chose to close the trial and apply the schedule developed in this study (every other day for 6 doses every 21 days ) to phase I testing of a modified BL22 construct with higher affinity for the CD22 antigen. This second-generation agent, HA22 or CAT-8015, was engineered to replace three amino acid residues in the heavy chain complementary determining region 3 of the BL22 binding domain. This modification increased the binding affinity for CD22 by 14-fold, which resulted in an approximately 1 log10 improvement in cytotoxicity against a variety of CD22+ malignancies (31, 32).
In summary, CD22 represents an excellent target for pediatric B-lineage hematologic malignancies. These studies offer proof-of-principle that anti-CD22 Pseudomonas-based immunotoxins can be administered to children and have the potential to overcome chemotherapy resistance and induce cytotoxicity of CD22+ blasts refractory to standard therapy. Anti-CD22 immunotoxins hold therapeutic promise in this common subtype of pediatric cancer.
Disclosure of Potential Conflicts of Interest
R.J. Kreitman, D.J. FitzGerald, and I. Pastan are coinventors on patents assigned to the NIH for the investigational product used in this research.
We thank the clinical trial participants and their families, referring physicians, Dr. David Waters, Dr. Lubing Gu, CURE Childhood Cancer, Inc., and the staff of the NCI, NIH Clinical Center, Aflac Cancer Center and Blood Disorders Service, Emory University/Children's Healthcare of Atlanta, Genencor, Inc., and MedImmune, LLC.
Grant Support: Intramural Research Program of the NIH, National Cancer Institute, the Center for Cancer Research, and a Cooperative Research and Development Agreement between the NIH, National Cancer Institute, and MedImmune, LLC.
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.