Phase I Trial of a Novel Diphtheria Toxin/Granulocyte Macrophage Colony-stimulating Factor Fusion Protein (DT₃₈₈GMCSF) for Refractory or Relapsed Acute Myeloid Leukemia¹

AML³ is the most common acute leukemia in adults with 10,000 new cases/year in the United States (1). With combination induction and consolidation chemotherapy using cytarabine and topoisomerase II inhibitors such as daunorubicin and etoposide, complete remission rates of about 70% have been achieved (2). However, most patients ultimately relapse and die from the disease or complications of treatment. The prognosis is dismal for patients with relapsed or refractory AML. Except for the minority of patients who undergo allogeneic bone marrow transplants or have initial complete remissions of >12 months, patients receive salvage therapy similar to that used in their induction/consolidation and have a median survival of weeks to months (3). Two-year survival in this subgroup of relapsed and refractory patients is rare.

Chemotherapy-resistant blasts contribute to treatment failures in AML patients (4). These blasts are resistant to multiple drugs and often show altered expression of one or more resistance proteins that influence drug efflux, drug metabolism, substrate levels, or cell death regulation (5, 6). Because most of the multidrug resistance phenotypes target low-molecular-weight inhibitors of DNA synthesis or cell proliferation, we searched for an agent that could induce leukemic cell death by mechanisms other than damage to DNA or cell division.

One novel class of AML therapeutics are DT fusion proteins consisting of the catalytic and translocation domains of DT genetically fused to AML blast selective ligands (7). The ligand delivers the protein to blast cell surface receptors. After receptor-mediated endocytosis, the fusion protein reaches the early endosomes, to which it is cleaved by furin and, in the acidic environment, inserts into the vesicle membrane and facilitates the escape of the DT-A fragment to the cytosol. In the cytosol, the A fragment ADP-ribosylates elongation factor 2, which leads to the inactivation of protein synthesis and cell death.

Because GMCSF receptors are expressed on the majority of myeloid leukemias but are poorly expressed on early normal hematopoietic stem cells, we chose to fuse GMCSF to the catalytic and translocation domains of DT (DT₃₈₈; Ref.8). The DT₃₈₈GMCSF molecule was cytotoxic in vitro to chemotherapy-resistant cell lines and therapy-refractory AML patient progenitors but was nontoxic to normal human myeloid progenitors (9, 10, 11, 12). Dramatic antileukemic efficacy was observed in vivo administering DT₃₈₈GMCSF to severe combined immunodeficient mice bearing human leukemia (13). On the basis of these results, we manufactured DT₃₈₈GMCSF under good manufacturing practice and obtained an IND (BB no. 8153) to perform a Phase I dose-escalation trial in patients with relapsed or refractory AML. This report describes the results of that study.

Eligibility criteria for patient entry into the study included histological bone marrow evidence for AML and age ≥18 years old. Patients were required to have relapsed or refractory leukemia, a Zubrod performance status of ≤2, to have fully recovered from toxicities of prior chemotherapy or radiation therapy, and to have a life expectancy of ≥3 months. Eligible patients had a bilirubin ≤1.5 mg/dl; transaminases ≤5 × upper limit of normal, creatinine ≤1.5 mg/dl; forced expiratory volume (FEV1) ≥70% normal; cardiac ejection fraction ≥50% normal; had no serious concurrent medical problems, uncontrolled infections, central nervous system leukemia, myocardial infarctions in the last 6 months, or history of severe penicillin allergy; and were not pregnant. Written informed consent was obtained from each patient before entry into the study.

DT₃₈₈GMCSF was produced under good manufacturing practice and aliquoted in vials containing 1 ml of PBS at 1.5 mg/ml and stored at −80°C (14). Before treatment, vials were thawed at room temperature, and appropriate amounts of drug aliquoted in a laminar flow hood for individual daily doses in sterile Eppendorf tubes and were refrozen at −80°C. Each day of treatment, a single dose was rethawed in the laminar flow hood and diluted aseptically in 1 ml of normal saline or 3% saline. A 1-h incubation at room temperature with 3% saline was used after patient 28 to reduce protein aggregates. After premedications with acetaminophen, diphenhydramine, and corticosteroids, the drug was administered as a 15-min infusion through a rapidly flowing i.v. line. Treatments were repeated daily for up to 5 days. Patients could be retreated with Federal Drug Administration approval if they had no unresolved toxicities.

Patients were monitored in-hospital for 2 weeks for toxicities. The National Cancer Institute Common Toxicity Criteria scale was used. Vital signs were measured posttherapy every 15 min for 1 h, every hour for 8 h, and then every 4 h for 7–10 days. Physical exams were done daily throughout the treatment period and for at least 1 additional week. Blood counts, blood chemistries, and urinalysis were done daily during treatment. Drug pharmacokinetics were measured with blood sampling on each treatment day. A previously described bioassay was used to quantitate circulating DT₃₈₈GMCSF with an assay limit of 0.3 ng/ml (15). Circulating cytokines (IL-1β, IL-6, IL-8, IL-12, IL-18, macrophage inflammatory protein 1α, IFNγ, and TNFα) were measured on the blood samples obtained for pharmacokinetics using enzyme immunoassays (R&D Systems, Minneapolis, MN; or MBL, Nogoya, Japan) following the directions of the manufacturers. Humoral immune response to DT₃₈₈GMCSF was measured on serum that was obtained pretreatment, day 14, day 30, and day 60 by both a sandwich enzyme immunoassay and cytotoxicity neutralization assay following the methods we have reported previously (16). Clinical response was assessed by blood counts and differentials, exams, and bone marrow aspirate and biopsies on days 12 and 30 and as indicated. A patient was considered in complete remission if there were no circulating blast, no extramedullary leukemia, the marrow blast percentage was <5%, and there was reconstitution of normal hematopoiesis with normal peripheral hemoglobin, platelet count, and neutrophil count without the need for transfusions. A partial remission occurred with the same conditions except for a lack of recovery of normal hematopoeisis.

Patients were treated at one of six dose levels (1, 2, 3, 4, 4.5, or 5 μg/kg/day for 5 days). Three to six patients were treated at each dose level to establish the MTD. Additional patients were added in some cases to better define the toxicity at that dose level. The MTD was defined as the highest dose level at which zero or one patient, among at least six patients, had DLT. DLT was defined as a drug-related ≥ grade 4 nonhematological toxicity or drug-related grade 4 hematological toxicity of >28 days duration. Transient asymptomatic (<2 weeks) grade 4 elevations of transaminases, CPK, or decreases in serum calcium were not considered DLTs on this study.

The cell cytotoxicity assay used the HepG2 hepatocellular carcinoma cell line and was obtained from the American Type Culture Collection (Manassas, VA) in MEM containing 0.1 mm nonessential amino acids, 1 mm sodium pyruvate, 2 mm l-glutamine, Earle’s balanced salt solution with 1.5 g/liter sodium bicarbonate and 10% fetal bovine serum (Life Technologies, Inc., Grand Island, NY). Cells were trypsinized and aliquoted at 10⁴ cells/well in Costar 96-well flat-bottomed tissue culture plates in 150 μl of medium containing 12 different concentrations of DT₃₈₈GMCSF or DAB₃₈₉EGF (a gift from Ligand Pharmaceuticals, Inc, San Diego, CA; Ref. 17), and were incubated at 37°C 5% CO₂ for 48 h. Fifty μl of medium with 1 μCi [³H]thymidine (NEN DuPont, Wilmington, DE) was then added, and, after 18 h, media was removed and cells washed three times with serum-free media, solubilized with 50 μl of 2 m NaOH, neutralized with 50 μl of 2 m HCL, harvested onto glass fiber mats with a Skatron Cell Harvestor (Skatron Instruments, Lier, Norway), and counted on a Betaplate reader gated for ³H. The calculated IC₅₀s were the concentrations of toxin that inhibited thymidine incorporation by 50% compared with control wells.

Thirty-one patients were treated for a total of 39 courses (8 patients received a second course). The mean age was 54 years (range, 12–84), and there were 13 males and 18 females (Tables 1 and 2). Although the eligibility criteria stipulated age ≥18 years, one child was allowed under a Federal Drug Administration-approved exemption. Four patients were in first relapse, 2 patients were in second relapse, and 25 patients had refractory AML. Six patients had undergone autologous bone marrow transplants, 4 had previously received allogeneic bone marrow transplants, and 21 had not received transplants. The cytogenetics were poor risk in 15 patients, intermediate risk in 8 patients, and good risk in 1 patient. In seven patients, cytogenetics had not been performed. Prior myelodysplasia had been present in 5 of the 31 patients.

The DLT for this dose/schedule of DT₃₈₈GMCSF was liver injury. Eight patients were treated at the 4.5–5-μg/kg/day dose level, and two of these patients had severe hepatic toxicity. The dose was reduced after patient 24, although there were no DLTs, to more safely assess the effects of modified prophylaxis (rofecoxib and anti-TNF-α antibody) on the incidence and severity of toxicities. Patient 28 developed liver failure and died 1 week after treatment. This was coincident with florid leukemic progression. Necropsy showed centrilobular hepatic necrosis. Patient 31 developed transient hepatic encephalopathy with elevated ammonia levels 4 days after therapy. The liver dysfunction resolved, but the patient developed renal failure contributed to by aminoglycosides and amphotericin. Without dialysis, the patient died 1 week later. Transient elevations in AST were seen in most patients (Tables 3 and 4). Furthermore, the peak AST level correlated with the dose (P = 0.002; Fig. 1,A) and peak DT₃₈₈GMCSF level (P = 0.0002; Fig. 1,B). The AST rose disproportionately relative to ALT and GGT. Steroids, calcium channel blocker, and antibody to TNFα did not influence this toxicity. The mechanism for the liver injury is unknown. As shown in Fig. 2, DT₃₈₈GMCSF did not damage a hepatic cell line. DT₃₈₈GMCSF may indirectly injure hepatocytes by causing inflammatory cytokine release from liver Kupffer cells, which are known to have GM-CSF receptors. No liver biopsies have been performed in DT₃₈₈GMCSF-treated patients, but elevated circulating levels of two inflammatory cytokines, IL-8 and IL-18, have been observed (Table 5). There was a significant positive association of IL-18 levels with AST levels (P = 0.05; Table 6). The elevations of IL-18 occurred earlier (by day 2) than the elevations of IL-8 (day 4–8; data not shown).

Other drug-related side effects include fever, chills, hypoxemia, and transient hypotension postinfusion. These were rare with corticosteroid prophylaxis beginning with patient 10. A transient asymptomatic sinus bradycardia occurred in some patients that was responsive to oral theophyllines. Echocardiograms during and after treatment did not show reduced cardiac function. There were no cardiac enzyme elevations (CPK-MB fractions) and no evidence of heart block by electrocardiography. Asymptomatic, transient hypocalcemia occurred between days 4 and 10. The incidence of hypocalcemia was dose dependent. Neither neurological nor muscular abnormalities were observed. When measured, the vitamin D levels were normal and parathyroid hormone was slightly increased. Serum calcium values of <7 mg/dl were treated with i.v. calcium gluconate. Serum calciums were not corrected for the serum albumins. Transient CPK and lactate dehydrogenase elevations were observed in most patients and paralleled the changes in AST. However, there were no signs of muscle injury or RBC injury measured by serum aldolase and haptoglobin or peripheral smear. VLS with the combination of edema, weight gain, hypoxia, hypotension, and hypoalbuminemia was noted only in patients 9 and 11. Patient 9 had not been receiving thyroid medications for 1 week, and the symptoms resolved with diuresis and reinstitution of thyroxine. Patient 11 had fluid overload combined with corticosteroids that also resolved with diuresis. The other patients had signs of the components of VLS with weight gain and/or hypoalbuminemia, but there was no associated edema, hypoxemia, or hypotension. The incidence and frequency of toxicities to DT₃₈₈GMCSF are shown in Tables 3 and 4.

On the basis of the occurrence of grade 4 and 5 liver toxicities in patients 31 and 28, respectively, we determined the MTD to be 4 μg/kg/day, and nine patients have been treated at that dose to date without DLT.

Pharmacokinetic data were obtained for the first infusion of each course on all 31 patients and for the first and last infusion on 5 patients. The peak DT₃₈₈GMCSF serum level occurred at 2 min postinfusion, and the concentration decreased over time exponentially with a t_1/2 of ∼30 min (Fig. 3). The peak fusion protein concentration was not significantly correlated with the dose (P = 0.53; Fig. 4,B) but was correlated with the pretreatment antibody titer (P = 0.0001; Table 7; Fig. 4 A; and see “Immune Response”). Interestingly, in the five patients for whom data were available, the peak DT₃₈₈GMCSF concentrations were higher on day 5 than on day 1.

We do not have a vaccination history for the 31 patients in the trial. However, most patients likely received their full immunization against DT in childhood. The EIA showed that 28 (90%) of 31 were positive (i.e., had circulating anti-DT₃₈₈GMCSF antibodies) in concentrations ranging from 0.2 to 9.4 μg/ml with a median of 1.85 μg/ml (Table 7). Patients with low pretreatment antibody titers were more likely to have measurable peak circulating DT₃₈₈GMCSF. The median and range of antibody for those with undetectable peak fusion protein during the first course were 2.5 μg/ml (1.1–9.4 μg/ml; n = 15), and the same antibody levels for those with measurable peak fusion protein were 0.6 μg/ml (0–3.7 μg/ml; n = 16). The difference was highly significant (P < 0.001). Only 2 of 11 patients with EIA antibody concentration above 2.2 μg/ml had detectable peak DT₃₈₈GMCSF compared with 14 of 20 with EIA antibody concentrations of <2.2 μg/ml.

After 15–60 days, 21 of 25 evaluable patients had increased antibody titers ranging from 0.2 to 6613 μg/ml (Table 7). One patient had no change in antibody titer and three patients had decreased antibody titers. Of the three patients with no humoral immune response to DT₃₈₈GMCSF, one patient had received two prior allogeneic bone marrow grafts, one patient had undergone a prior autologous bone marrow transplant, and the last patient was heavily pretreated with fludarabine.

Three clinical remissions were observed (Table 8). Six of the 31 patients had relapsed disease. The three responses were seen in this group. No responses were seen in the refractory patients. Patient 22 was a 72-year-old female who developed AML in April 1996. She had normal cytogenetics and received idarubicin plus cytarabine (3 + 7), achieving a complete remission. She had no consolidation therapy, relapsed in January 2000 and received salvage therapy consisting of topotecan and high-dose cytarabine. However, she was refractory, and bone marrow biopsy on March 8, 2000, showed 28% blasts confirmed by flow cytometry (Fig. 5,A). She received a 5-day course of DT₃₈₈GMCSF at 5 μg/kg/day complicated only by asymptomatic and transient transaminasemia, elevated CPK, and hypocalcemia from April 10 to April 14, 2000. Before therapy, she had an ANC of 280/μl, a platelet count of 64,000/μl, and no circulating blasts. One and 2 months posttherapy, her bone marrow showed 1–3% blasts by morphology and flow cytometry (Fig. 5 B). She had recovery of platelets by day 60 to 158,000/μl but continued to be neutropenic (ANC of 279/μl). She was active spending most days out of the home and did not require antibiotics. By August 17, 2000, she had normalization of counts with an ANC of 1,320/μl, a platelet count of 238,000/μl, and a hemoglobin of 12.0 g/dl, and did not require transfusions or antibiotics. Repeat bone marrow exam on November 15, 2000, showed no morphological evidence of increased blasts, but there were 8% blasts by flow cytometry. Her blood counts remained normal with an ANC of 1,600/μl, a hemoglobin of 14.2 g/dl, and a platelet count of 180,000/μl. By March 29, 2001, she had a recurrence of pancytopenia (ANC of 420/μl, platelet count of 76,000/μl, and hemoglobin of 12.8 g/dl). There were no circulating blasts, but the bone marrow examination showed 8% blasts by morphology and 12% blasts by flow cytometry. She received a second course of DT₃₈₈GMCSF at 5 μg/kg/day for 5 days, and the day-12 bone marrow showed disappearance of the blasts by morphology and flow cytometry. By day 17 after the second course, her ANC was 1,000/μl However, she remained thrombopenic (platelet count of 7/μl)and anemic (hemoglobin of 7.1 g/dl). On day 21, she developed a pneumonia, confirmed by chest X-ray, but declined aggressive management and died on day 24 posttherapy.

Patient 24 was a 33-year-old female diagnosed with AML in November 1998 and induced with idarubicin plus cytarabine (3 + 7) followed by an allogeneic bone marrow transplant in June 1999. She developed bronchiolitis obliterans, requiring chronic corticosteroids, and had relapse of her AML in April 2000. Bone marrow exam showed 14% blasts. She had 300/μl circulating blasts and thrombopenia (platelet count of 23,000/μl). She received DT₃₈₈GMCSF at 5 μg/kg/day for 5 days, and, again, the only side effects were transient, asymptomatic transaminasemia and hypocalcemia. Her day-30 bone marrow showed 2% blasts, but she remained pancytopenic with an ANC of 75/μl, a platelet count of 25,000/μl and a hemoglobin of 10.8 g/dl. She was asymptomatic. She received a course of granulocyte colony-stimulating factor (G-CSF) to stimulate recovery of normal myelopoiesis, however, by day 60, a repeat bone marrow showed an increase in blast percentage to 15%. She was again treated with DT₃₈₈GMCSF at 5 μg/kg/day complicated only by asymptomatic and transient transaminasemia. Her marrow blasts at day 30 were 5–10%. She continued to have neutropenia and thrombopenia. She died 1 year after initiating fusion protein therapy in April 2001.

Patient 25 was a 70-year-old male diagnosed with AML in March 2000 and treated with idarubicin plus cytarabine for induction and consolidation. He relapsed in April 2000 with a bone marrow showing 15% blasts. He had circulating blasts (50/μl), thrombopenia (platelet count of 34,000/μl), and anemia (hemoglobin 8.3 g/dl). He was treated from 6/22/00 to 6/26/00 with DT₃₈₈GMCSF at 3 μg/kg/day. His course was complicated by transient renal insufficiency attributed to rofecoxib and transient, asymptomatic transaminasemia, hypocalcemia, and elevated CPK. The day-12 marrow showed 1% blasts, and the day-30 marrow showed 4% blasts. However, he remained anemic and thrombopenic. He declined further fusion protein or other therapy and died from progressive disease 40 days posttherapy.

The diphtheria fusion protein DT₃₈₈GMCSF unexpectedly produced liver injury characterized by transient transaminasemia and, rarely, more severe liver dysfunction. The MTD of 4 μg/kg/day for 5 days or 20 μg/kg total per course is much lower than that observed for the other systemically administered clinically active fusion proteins ONTAK, LMB–2, and BL22 (120, 120, and 135 μg/kg/total, respectively; Refs. 19, 21, 22) and resembles the low MTDs of 18 and 3 μg/kg/total DAB₃₈₉EGF and erb-38, respectively (23). Both DAB₃₈₉EGF and erb-38 bound hepatocytes and produced direct liver injury (see Fig. 3). In contrast, DT₃₈₈GMCSF does not directly bind or damage hepatocytes. We hypothesize that the fusion protein bound the GMCSF receptor-positive macrophages in the liver (Kupffer cells) and triggered cytokine release. We have observed extremely elevated circulating levels of IL-18 in treated patients. IL-18 is a known pro-inflammatory cytokine that can be released by stimulated Kupffer cells and can damage hepatocytes by inducing their expression of Fas, leading to Fas ligand-mediated cell death (24, 25). Other laboratories have shown that cyclosporin, FK506, and monoclonal anti-IL-18 antibodies can provide effective prophylaxis in vivo for IL-18-inducing liver toxins (26, 27). Whereas our data are consistent with the hypothesis of Kupffer cell cytokine-induced liver injury, the cause of the DT₃₈₈GMCSF hepatotoxicity is not identified definitively.

The humoral immune response to DT fusion proteins clearly affected circulating levels of DT₃₈₈GMCSF and likely influenced clinical benefit at the low doses used in this study. Approaches to circumvent the immune response are to carefully select patients with low pretreatment antibody titers or to alter the schedule and use fewer but higher doses. Another approach is to engineer and develop other targeted toxins that can intoxicate malignant myeloid cells. An antibody to CD33 has been humanized and conjugated to recombinant gelonin, a plant ribosome-inactivating protein (28). Human proteins such as eosinophil-derived neurotoxin, a member of the RNase family, may be engineered to target myeloblasts (29). This toxin, combined with a human-derived ligand peptide, should be less immunogenic in patients.

We observed clinical activity with DT₃₈₈GMCSF in highly chemoresistant AML patients. This observation confirms the antileukemic activity observed with this drug in tissue culture and animal models (10, 11, 12, 13). Similar fusion proteins (DAB₃₈₉IL-2, Tf-CRM107, LMB-2, and LMB-7) directed to other receptors have yielded 30–80% response rates in cutaneous T-cell lymphoma, high-grade gliomas, and hairy cell leukemia (19, 20, 21, 22). The observed response rate of 10% is significantly lower than with these other fusion proteins. There are several possible explanations for this difference. First, DT₃₈₈GMCSF may be less active in patients than are the other proteins. The preclinical data does not support this hypothesis. In tissue culture, leukemic progenitors were killed with picomolar concentrations of drug (9, 10, 11, 12). In vivo, long-term remissions were produced with DT₃₈₈GMCSF (13). These results compare favorably with the other clinically efficacious fusion proteins. A second hypothesis is that most patients were treated at suboptimal doses, and very few patients received therapeutic amounts of DT₃₈₈GMCSF. Only 18 of 31 patients received doses at or above the MTD (4 to 5 μg/kg/day). A slow interpatient dose escalation was done because of the novel nature of this drug and the comorbidities and non-drug-related toxicities that are common in these highly pretreated leukemic patients. A third hypothesis is that the early occurrence of DLT limited the dose escalation and prevented administration of adequate dose levels to achieve a higher response rate. The liver toxicity precluded dose escalation, and reducing the liver toxicity should permit dose escalation. A fourth factor that likely contributed to the low response rate was the presence of pretreatment anti-DT antibodies in many of the patients. Whereas the influence of anti-DT antibodies is less pronounced when higher doses of DT fusion proteins are given (19), when only 4–5 μg/kg fusion protein can be administered, the antibodies likely reduce blast exposure to drug. We failed to observe circulating drug in most patients with high anti-DT antibody titers. Fifth, there may be significant patient-to-patient variations in sensitivity to DT₃₈₈GMCSF that were not detected with the previously used colony-forming assay (AML-CFC; Refs. 9, 10, 11, 12). In support of this hypothesis, we have found dramatic differences in patient leukemic blast sensitivity to DT₃₈₈GMCSF using a 3-day proliferation assay.⁴ However, patient blast GM-CSF receptor content and DT₃₈₈GMCSF sensitivity was not measured prospectively in our study patients. Finally, the patients treated on this study, in most instances, had very advanced disease. In only 12 of 39 courses were patients treated who had <30% marrow blasts. Patients with extensive tumor burden may be less sensitive to biologicals such as DT₃₈₈GMCSF, particularly at the low dose levels used in this study. All of the responses observed occurred among the 12 patients with low marrow-blast percentage.

We are committed to discovering approaches to improve the preliminary response rate observed in the Phase I study. First, we are investigating the molecular mechanism for the liver toxicity and looking for methods of prophylaxis that will protect the liver without modifying the antileukemic efficacy. We are testing oral glycine prophylaxis to block Kupffer cell activation and liver injury from DT₃₈₈GMCSF (30). The selective glycine inhibition of liver macrophage should not impair the antileukemic activity of DT₃₈₈GMCSF. Second, we are evaluating prognostic factors that may identify a patient category more likely to respond, including low pretreatment anti-DT antibody titer, high sensitivity of pretreatment blasts to DT₃₈₈GMCSF, and low pretreatment leukemic burden. Third, we will give the fusion protein on a twice-weekly-for-2-weeks schedule as has been used successfully for the Pseudomonas exotoxin fusion proteins (i.e., LMB-2 and BL22). Such a schedule may permit higher individual doses to be given and yield better blast saturation both in the blood and marrow and other extravascular sites.

Although multiple issues need to be addressed to improve the therapeutic index of the DT₃₈₈GMCSF fusion protein for therapy of myeloid malignancies, the preliminary results are encouraging and suggest that additional preclinical and clinical development is warranted.

We thank Drs. Samer Renno, Itzvan Molnar, and Andrew Mone for their assistance in developing some of the in vitro cytokine assays, patient serum DT₃₈₈GMCSF assays, and patient serum cytokine assays used in this study. We also thank Jason Ramage and Jennifer Black for excellent technical assistance, Cynthia Patrick for clinical protocol nursing support, Dr. Mark Willingham for assistance with photomicroscopy, and Dr. Charles McCall for advice and guidance during this study. We also thank Dr. Dianna Howard at the University of Kentucky Medical Center for the care of two of the patients on protocol.