Malignant Progenitors From Patients With Acute Myelogenous Leukemia Are Sensitive to a Diphtheria Toxin–Granulocyte-Macrophage Colony-Stimulating Factor Fusion Protein

ACUTE MYELOGENOUS leukemia (AML), the most common acute leukemia in adults, is associated with a 70% complete remission (CR) rate after standard induction chemotherapy regimens.1 Intensive postremission therapy in combination with allogeneic bone marrow transplantation offers the possibility of cure to some of these individuals; however, fewer than 20% of all patients with the diagnosis of AML will have prolonged disease-free survival.2 Resistance to standard chemotherapeutic drugs is an important cause of the relapsed, refractory leukemia to which most patients succumb.3 

A number of identified drug resistance phenotypes are due to overexpression of specific proteins, and the concentration of these molecules in leukemic blasts has been correlated with response to cytotoxic chemotherapy.4-8 In several cases, the resistance protein transports or inactivates xenobiotics such as anthracyclines. Pharmacologic reversal of MDR-1–related anthracycline resistance has, to date, been associated with toxicities to marrow, gastrointestinal tract, and the central nervous system; altered anthracycline pharmacodynamics; and minimal improvements in response rate or disease-free survival.9 An antibody-targeted cytotoxic drug, anti-CD33-calicheamicin, was recently tested in a clinical trial and had reduced side effects.10However, most patients rapidly developed resistance due to active drug efflux. Thus, novel agents that are cytotoxic to leukemic blasts and bypass multidrug resistance phenotypes are urgently needed.

One such class of therapeutics are protein toxins covalently linked to peptide ligands. The ligand directs the molecule to the surface of specific cell types. The toxin moiety then enters the cell and catalytically inactivates protein synthesis. Toxins constructed that target AML blasts include the following: anti-CD33-blocked ricin, anti-CD33-gelonin, diphtheria toxin-interleukin-3 (IL-3), anti-transferrin receptor-ricin A chain, granulocyte-macrophage colony-stimulating factor (GM-CSF)-ricin, GM-CSF-Pseudomonas exotoxin, and diphtheria toxin-GM-CSF.11-21 Each of these reagents inhibited protein synthesis by 50% (IC₅₀) in cell lines in a dose- and time-dependent manner. The least active drugs were anti-transferrin receptor-ricin A chain, GM-CSF-ricin, and GM-CSF-Pseudomonas exotoxin, with IC₅₀s of approximately 3 × 10⁻¹⁰ mol/L. Anti-CD33-blocked ricin and anti-CD33-gelonin had intermediate IC₅₀s of 10⁻¹⁰ mol/L, and diphtheria toxin-IL3 and diphtheria toxin-GM-CSF produced IC₅₀s of 3 × 10^-11mol/L. The variability in sensitivity that AML cell lines show to targeted toxins has been attributed to premature intracellular routing of ricin, ricin A chain, gelonin, and Pseudomonas exotoxin conjugates to lysomes. In contrast, the diphtheria toxin fusion proteins are able to translocate to the cytosol from a prelysosomal intracellular compartment. Specificity of cell kill has been demonstrated for several of these conjugates, including anti-CD33-blocked ricin, anti-CD33-gelonin, diphtheria toxin-IL3, and diphtheria toxin-GM-CSF. Reductions in normal human or murine marrow progenitors were seen after toxin conjugate exposure, but, in each case, the effect was much smaller than seen with AML cell lines.17,19,20 

A critical question is whether these chimeric proteins, which require functional cell surface receptors, intact intracellular routing pathways, and sensitive protein synthesis machinery for intoxication, will kill malignant cells isolated directly from patients with leukemia. With several previously studied lymphoid malignancy-targeted toxins, including anti-CD5-ricin A chain and anti-CD25 (sFv)-Pseudomonas exotoxin, fresh leukemia cells were much less sensitive than cell lines.22,23 Cytotoxicity was enhanced for anti-CD5-ricin A chain by adding the monocarboxylic acid ionophore, monensin, and for anti-CD25-Pseudomonas exotoxin by prolonged incubation. Both of these interventions improve intracellular toxin transport to compartments optimal for translocation to the cytosol. Few similar experiments have been conducted with toxins targeted to myeloid leukemia.

Roy et al11 tested the cytotoxicity of anti-CD33-blocked ricin to leukemic colony-forming cells (CFU-L) from 12 AML patients and found that reductions in CFU-L colony formation were dose- and time-dependent. Perentesis et al21 observed a 1 to 3 log kill of CFU-L from 7 of 9 therapy-refractory AML patients after exposure to diphtheria toxin-GM-CSF. However, it is possible that some of the observed effects in the latter studies were due to the presence of GM-CSF in the control but not in the toxin-treated test cultures.

To perform a comprehensive study of the sensitivity of leukemic clonogenic cells to diphtheria toxin-GM-CSF, we collected a series of 50 AML samples representing every French-American-British (FAB) subtype except M3 and M6. These included samples from patients who were known at the time to have disease both sensitive to and refractory to conventional cytotoxic agents. Also included were several samples from patients with lymphoid leukemias and normal bone marrow controls. We determined the GM-CSF receptor density and kd for the malignant blasts from each of the malignant samples and then assessed the ability of DT388-GMCSF to inhibit the growth of normal and leukemic clonogenic cells in standard semisolid growth factor-supplemented medium. As expected, some of the (untreated) leukemic cell samples failed to yield any discrete colonies in this assay. Nevertheless, for 33 of the 53 samples, we were able to determine for the toxin both the log kill and IC₅₀ of CFU-L. The results demonstrate that leukemic progenitors from the majority of primary human AML patients are sensitive to the cytotoxic effects of this conjugate, regardless of their responsiveness to conventional chemotherapy drugs, whereas normal clonogenic cells largely are not.

Heparinized peripheral blood samples from 50 patients with a diagnosis of AML and 3 patients with lymphoid leukemias and normal bone marrow samples from 4 individuals donating marrow for allogeneic transplantation were obtained after informed consent was obtained (MUSC protocol #7123, Terry Fox Laboratory protocol #C96-0429, and SWOG protocol #8600). Low-density (<1.077 g/mL) cells were isolated using a Ficoll-Hypaque gradient.

Freshly isolated or thawed, cryopreserved leukemic samples were suspended in RPMI 1640 medium (Irvine Scientific, Santa Ana, CA) with 15% fetal calf serum (FCS). After incubation for 1 hour at 37°C/5% CO₂ in 75-cm² tissue culture flasks (Costar Scientific Corp, Cambridge, MA), the nonadherent cells were collected and mixed with 0.1 mL anti-CD2 immunobeads (Dynal, Oslo, Norway). The bead-cell mixture was gently rocked at 4°C for 30 minutes, and then CD2⁺ cells were depleted by magnetic separation. The procedure removed greater than 95% of CD2⁺T cells from initial preparations based on flow cytometry using anti-CD3-phycoerythrin binding (data not shown). Cells were then counted and aliquoted for the GM-CSF receptor and DT388-GMCSF sensitivity studies described below.

DT388-GMCSF was prepared and purified as previously described.17 Material was stored as aliquots at 840 μg/mL in phosphate-buffered saline (PBS) plus 1% human serum albumin at −20°C until used. The material used in this study was found to kill the HL60 human AML cell line with an IC₅₀ of 2 × 10⁻¹² mol/L in a 48-hour³H-thymidine incorporation assay and to produce a maximum 3.5 log depletion of HL60 cells forming colonies in semisolid medium.15 

Aliquots of 1 to 6 × 10⁶ cells in RPMI 1640 plus 2.5% bovine serum albumin and 20 mmol/L HEPES and 0.2% sodium azide were mixed with varying amounts of ¹²⁵I Bolton-Hunter-labeled human GM-CSF (80 to 120 μCi/μg; NEX249; DuPont, Boston, MA) with or without excess (1,500 ng) cold GM-CSF (Immunex, Seattle, WA) in a total volume of 150 μL in 1.5-mL Eppendorf tubes. Cells were incubated at 37°C for 30 minutes and then layered over a 200 μL oil phthalate mixture (1 part dioctylphthalate and 1.5 parts dibutylphthalate; Aldrich, Milwaukee, WI). After centrifugation at 12,000 rpm for 1 minute in a microfuge at room temperature, both pellets and supernatants were saved and counted in an LKB-Wallac 1260 Multi-gamma counter (Gaithersburg, MD) gated for ¹²⁵I with 50% counting efficiency. Background cpm was calculated by linear extrapolation from incubations with excess cold GM-CSF. Scatchard plots of specific bound/free versus specific bound cpm were made. Receptor number/cell was calculated using the following equation: the value for the x intercept / (specific activity in μCi/μg × the cell number × [4.2 × 10⁻⁸]). kd was calculated as the x intercept times 2.7 × 10⁻¹³ divided by the y intercept times the specific activity. The Statistical software package (Statsoft, Tulsa, OK) was used for linear regression with separate analysis of the 6 lowest concentration ¹²⁵I-GM-CSF points and the 4 highest concentration ¹²⁵I-GM-CSF points. Nonlinear regression using the Radlig program (Biosoft) confirmed estimates from Scatchard plots for 4 samples. Receptor densities of 0 were recorded based on the absence of specific ¹²⁵I binding or negative values for kd.

Sensitivities of fresh leukemic blast progenitors and normal bone marrow CFC to DT388-GMCSF were tested in suspension culture. Aliquots of 10⁶ AML blasts or light density marrow cells were placed in suspension culture with different concentrations of DT388-GMCSF (0 to 4 × 10⁻⁹ mol/L) in 24-well flat-bottomed Costar plates. AML cells were cultured in 1 mL of RPMI 1640 with 20% FCS and 50 ng/mL granulocyte colony-stimulating factor (G-CSF; Amgen, Thousand Oaks, CA) plus 1 of 10 different toxin concentrations over the indicated range. Normal marrow cells were cultured with 1 of 4 different DT388-GMCSF concentrations over the same range in 0.5 mL serum-free medium (StemCell Technologies, Vancouver, British Columbia, Canada) without added cytokines. Suspension cultures without the fusion toxin but the same in all other respects served as the untreated controls for determining the percentage of survival of clonogenic cells from both leukemic and normal samples.

After 48 hours of incubation at 37°C in 5% CO₂, 100 μL from each AML suspension culture containing different concentrations of DT388-GMCSF was mixed with 3 mL RPMI + 15% FCS plus 50 ng/mL G-CSF and GM-CSF and 0.3% agarose (SeaPlaque; FMC Bioproducts, Rockland, ME) and poured into 35-mm gridded petri dishes (Nunc, Naperville, IL). In some experiments, 10% medium conditioned by the 5637 human bladder carcinoma cell line was added in the agarose colony assay instead of recombinant growth factors. After 10 minutes at 4°C to solidify the medium, dishes were placed in humidified chambers at 37°C/5% CO₂ for 14 to 21 days, after which colonies containing greater than 20 cells were counted. Both the concentrations of toxin reducing colony formation by 50% (IC₅₀) and the maximal log cell kill compared with controls were calculated as previously described.15 

In one experiment, cells from 4 AML samples treated with DT388-GMCSF were assayed for colony formation in parallel in both the agarose-based assay described above and in methylcellulose medium (StemCell Technologies) with 30% FCS and 3 U/mL human erythropoietin (Epo; StemCell), 10 ng/mL GM-CSF (Sandoz International, Basel, Switzerland), 10 ng/mL IL-3 (Sandoz), 50 ng/mL Steel factor (SF; Terry Fox Laboratories, Vancouver, British Columbia, Canada), and 50 ng/mL flt-3 ligand (FL; Immunex) with equivalent results. All subsequent assays were performed using the agarose-based assay and form the basis of the experiments reported here.

After 4 to 48 hours of incubation in suspension culture, cells to be assayed for burst-forming units-erythroid (BFU-E), colony-forming units-erythroid (CFU-E), CFU–granulocyte-macrophage (CFU-GM), and CFU-granulocyte/macrophage/erythroid/megakaryocyte (CFU-GEMM) were plated in methylcellulose-containing medium supplemented with 30% FCS and 3 U/mL Epo (StemCell) to which was added 50 ng/mL SF and 20 ng/mL each of human IL-6 (Terry Fox Laboratories), IL-3, GM-CSF, and G-CSF (Amgen). After 18 to 21 days at 37°C, erythroid, GM, and multilineage colonies were scored in situ. CFU-megakaryocyte (CFU-Mk) were detected as previously described in a serum-free assay modified to use a collagen base rather than agarose, to which was added 10 ng/mL IL-3, 10 ng/mL IL-6, and 50 ng/mL thrombopoietin (Zymogenetics Inc, Seattle, WA).24 After 18 to 21 days of incubation, Mk-containing colonies were identified in situ using immunocytochemical staining with an anti-CD41 monoclonal antibody (provided by P. Lansdorp, Terry Fox Laboratories) followed by alkaline phosphatase/anti-alkaline phosphatase detection.24 

Fifty previously untreated AML patients were studied. The age, type of leukemia, and responsiveness of the patient to chemotherapy after collection of the sample are shown in Table1. There were 2, 7, 13, 11, 11, and 1 patient with the subtypes M0, M1, M2, M4, M5, and M7, respectively, whereas for 5 patients the FAB type was not specified. No M3 or M6 subtypes were represented. A large proportion (78%) had presenting white blood cell counts in excess of 50 × 10⁹/L. The median age was 48.5 years (range, 9 to 82 years). Three non-AML samples (patients no. 51 and 52 with acute lymphoblastic leukemia [ALL] and patient no. 53 with prolymphocytic leukemia [PLL]) were also tested. Of the 47 patients who received remission induction chemotherapy, 22 (47%) achieved CR.

Because the GM-CSF receptor population includes both high-affinity α, β chain complexes and low-affinity receptors consisting of the α subunit only, we analyzed the presence of both types of receptor complex on leukemic blasts. An example of one of the Scatchard plots is shown in Fig 1. The results for all 53 leukemic patients are shown in Table 2. The PLL sample and 1 ALL sample (no. 52) had no detectable high-affinity receptors. The other ALL sample (no. 51) showed 93 high-affinity receptors/cell, with a kd of 4.5 × 10⁻¹¹ mol/L and 216 low-affinity receptors, with a kd of 5 × 10⁻⁹ mol/L. Eighty-eight percent of AML patients (44/50) had ≥35 high-affinity receptors/cell, and 74% (37/50) had greater than 100 high-affinity receptors/cell. Among all 50 AML patients, for the high-affinity receptor, the mean ± SEM dissociation constant (kd) was 5 ± 1.3 × 10⁻¹¹ mol/L, and the median kd was 2 × 10⁻¹¹ mol/L. The mean and median kd of the low-affinity receptor were both 1 ± 0.2 × 10⁻⁹ mol/L. Nonlinear regression analysis yielded similar values (within 30%) for receptor numbers and kd on samples from 4 normal marrow donors.

As shown in Table 2, 30 of 50 (60%) patient samples formed discrete colonies of greater than 20 cells after 14 to 21 days in semisolid medium. The plating efficiency of the various leukemic samples varied by almost 3 orders of magnitude (from 8 to 5,950 CFU-L per 10⁵ cells plated).

DT388-GMCSF reduced blast colony formation by at least 50% from 27 of 30 (90%, with 95% confidence interval 78%-98%) AML patient samples (Table 2 and Fig 2). It had no effect on colony formation from 2 AML samples (no. 46 and 47) that had less than 35 GM-CSF high-affinity receptors/cell. In contrast, the fusion toxin inhibited colony growth by at least 50% from all but 1 AML sample with ≥35 high-affinity receptors/cell (no. 25). The 3 non-AML patients' blasts were either insensitive (no. 53 and 52) or showed minimal sensitivity to the toxin (no. 51). The median IC₅₀ for CFU-L among the 30 AML samples that formed colonies was 4 × 10⁻¹² mol/L (range, 5 to >4,000 × 10⁻¹² mol/L), whereas the corresponding values for log kill of CFU-L in the same samples was 1.6 (range, 0 to 3.7; Table2). In logistic regression analyses based on the 25 AML patients who received remission induction chemotherapy, who had ≥ 35 high-affinity receptors per cell, and whose samples grew colonies, there were no significant associations between inhibition of colony growth (log cell kill) and probability of complete response (two-tailed P = .36) or of resistant disease (P = .34). Similarly, there was no significant association of the IC50 with CR (P = .33) or resistant disease (P = .42).

Light-density clonogenic cells from 4 normal bone marrow samples were also tested for sensitivity to the fusion toxin. As shown in Table 3, an average of 56% of total normal CFC survived (range, 33% to 102%), regardless of their lineage, after being incubated for 48 hours with the maximum concentration of DT388-GMCSF tested against both normal and malignant cells (4 × 10⁻⁹ mol/L). Failure of the fusion toxin to kill normal CFC was not attributable to competition from endogenous GM-CSF release during the 48-hour incubation period, because coincubated HL60 cells showed the same percentage of kill as in the absence of marrow cells (data not shown).

Recent clinical trials continue to show that the majority of AML patients develop disease resistant to standard cytotoxic chemotherapy drugs.25,26 Because DT388-GMCSF works by a unique mechanism (inhibition of protein synthesis), its efficacy should not be affected by most multidrug resistance phenotypes,15,17 and its in vivo toxicity profile may be different than current cytotoxic chemoradiotherapy regimens.27 Thus, such a reagent could be useful in the treatment of both newly diagnosed and relapsed/refractory patients.

Before further clinical development, we sought experimental evidence that AML blast progenitors from newly diagnosed patients could be killed by the fusion toxin. Although we intended to test toxin sensitivity in a series of different AML samples representative of the entire spectrum of phenotypes in this disease, the need for relatively large numbers of leukemic blasts to perform these experiments led to a selection bias for patients presenting with high circulating blasts counts. A large proportion of these patients (15/47 or 32%) failed to respond to standard remission induction chemotherapy, consistent with results from clinical trials showing that such patients have a poor prognosis.25,26 

Almost all (47/50) of the AML patients in this study had GM-CSF receptors on their blasts. Our rate of receptor frequency paralleled those reported by others.28,29 Kelleher et al30reported a slightly lower receptor frequency on a small group of patients using a cold saturation assay and two different types of recombinant GM-CSF. Our higher receptor frequency may reflect our use of the more sensitive hot saturation assay and/or the myelomonocytic differentiation of AML blasts from the M4 or M5 FAB subgroups, which made up half of the samples we tested.

Although it is well known that clonogenic progenitors exist among the malignant cells in patients with AML,31 their frequency and the size and morphology of the colonies they produce are highly variable (Table 2).32 To demonstrate the sensitivity of CFU-L to DT388-GMCSF quantitatively, it was necessary that the malignant blasts from a given sample form a sufficient number of discrete colonies in the semisolid assay used. Thirty of 50 (60%) AML samples analyzed here met this criterion. Conclusions regarding the sensitivity of CFU-L to the fusion toxin are thus restricted to this group. Similarly, the maximum log kill detected in these assays was, in many cases, determined by the clonogenicity of the untreated AML sample (ie, even if 100% of CFU-L were killed by the toxin at low concentrations, if only 10 colonies formed in the control assay the maximum log kill detectable is 1). Patients whose leukemic blasts contain a higher proportion of CFU-L have been reported to have a poorer prognosis than AML patients in general.33Nevertheless, even in this subgroup of patients, sensitivity of clonogenic progenitors to DT388-GMCSF was observed here, with at least 50% inhibition of colony growth in 90% of patients.

The presence of GM-CSF receptors on CFU-L against which the toxin showed cytotoxicity would be expected, because the drug must first bind to target cells and internalize before intoxication. Both subunits of the receptor must be present for ligand internalization.34In fact, AML blasts from all the samples against which DT388-GMCSF had activity were shown to have at least 35 high-affinity receptors per cell. We did not find a significant correlation between either the number of high-affinity receptors or their kd and their IC50 for the drug. With this study's limited sample size, there may have been insufficient statistical power to detect such associations. Another possible explanation for this fact is that the receptor studies were performed on the entire blast population rather than on the small subset of clonogenic AML cells against which the toxicity of the drug was tested. These latter cells may differ from the population as a whole in their expression of GM-CSF receptors. Cells from the patient with PLL and 1 ALL patient were both receptor negative and insensitive to DT388-GMCSF, whereas the ALL patient sample that showed intermediate numbers of high-affinity receptors had modest sensitivity (IC₅₀ 6 × 10⁻¹⁰ mol/L) to the drug. The blasts of the 2 AML patients (no. 46 and 47), with 6 and 33 high-affinity receptors/cell, were insensitive to DT388-GMCSF. Although these observations on lymphoid leukemias and AMLs with low numbers of GM-CSF receptors on circulating blasts should be confirmed with larger numbers, at present it appears appropriate to restrict future clinical development of this reagent to AML patients with GM-CSF receptor-positive blasts (≥35 receptors/cell).

CFU-L from the eight patients with refractory AML and blasts bearing ≥ 35 GM-CSF receptors per cell were assayed for DT388 GM-CSF sensitivity in this study. All eight showed significant (≥ 1 order of magnitude) inhibition of colony growth with low values of IC50 (eg, < 2 × 10^-11 mol/L) for seven of the eight. These results extend an earlier report on nine refractory patients by Parentesis et al21 in which seven had DT-GMCSF–sensitive blasts21 and are consistent with our results on drug-resistant AML cell lines.15,35 We hypothesize that targeted toxins are, in general, less affected by multidrug resistant phenotypes induced by conventional cytotoxic drugs or radiotherapy. Confirmation of this hypothesis will require testing of a larger group of relapsed/refractory patients' blasts.

In contrast to the marked toxicity of DT388-GMCSF on AML progenitors, its effect on most normal bone marrow clonogenic cells was insignificant, even after exposure to the highest concentrations of toxin tested for 48 hours. These results (Table 3) are consistent with those reported by others.17,19,20 Further evidence for the lack of effect of this reagent on more primitive normal hematopoietic precursors that form cobblestone areas in long-term culture has also recently been published.36 Taken together, these findings support the further preclinical development of DT388-GMCSF as a novel treatment agent for AML patients with both chemotherapy-sensitive and drug-resistant disease.