Intracranial therapy of glioblastoma with the fusion protein DTIL13 in immunodeficient mice
Version of Record online: 15 DEC 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 10, pages 2594–2601, 15 May 2006
How to Cite
Rustamzadeh, E., Hall, W. A., Todhunter, D. A., Low, W. C., Liu, H., Panoskaltsis-Mortari, A. and Vallera, D. A. (2006), Intracranial therapy of glioblastoma with the fusion protein DTIL13 in immunodeficient mice. Int. J. Cancer, 118: 2594–2601. doi: 10.1002/ijc.21647
- Issue online: 28 FEB 2006
- Version of Record online: 15 DEC 2005
- Manuscript Accepted: 15 SEP 2005
- Manuscript Received: 9 SEP 2005
- William Peyton Foundation
- Peyton Society of the University of Minnesota
- National Institutes of Health. Grant Number: NCI R01CA 108637
- diphtheria toxin;
- brain tumor;
A fusion protein consisting of human interleukin-13 and the first 389 amino acids of diphtheria toxin was assembled in order to target human glioblastoma cell lines in a murine intracranial model. In vitro studies to determine specificity indicated that the protein called DTIL13 was highly selective for human glioblastoma. In vivo, the maximum tolerated dose of DTIL13 was 1 μg/injection given every other day and repeated for 3 days. Doses that exceeded this amount resulted in weight loss and liver damage as determined by histology and enzyme assay. Experiments in IL-4 receptor knockout mice revealed that liver toxicity was receptor-related. This same dose given to nude mice with established U373 MG brain tumors resulted in significant reductions in tumor volume and significantly prolonged survival (p < 0.0001). Magnetic resonance imaging (MRI) proved to be extremely useful in (i) determining the ability of DTIL13 to reduce tumor size and (ii) for studying toxicity since diffusion-weighted and gradient echo-weighted MRI revealed that vascular leak syndrome was not a limiting toxicity at this dose. These results suggest that DTIL13 is as effective in an intracranial rodent model as it was in a flank model in previous studies and that DTIL13 might be an effective treatment for glioblastoma multiforme. © 2005 Wiley-Liss, Inc.
Glioblastoma multiforme (GBM) is an incurable, heterogeneous, high-grade astrocytic glioma, thought to originate from glial nonneuronal cells.1 More recent studies have identified a subpopulation of cancer stem-cells within GBM that retain the capacity for self-renewal and the ability to differentiate into a phenotypically diverse population of cells.2, 3, 4 With a 2-year-patient survival rate of <30%,5 effective treatment has been hindered by the lack of tumor-specific markers in the majority of patients. Recently, the IL-13 receptor (IL-13R) has been found to be overexpressed on cultured human GBM cell lines and surgical GBM specimens, but is not detectable in normal brain tissue.6, 7, 8 Targeting the IL-13 receptor with an immunotoxin (IT) has already proven to have potential for treating brain tumors.8, 9, 10, 11, 12 ITs are not mainstream pharmaceuticals, and treatment of systemic tumors with ITs has been limited by their failure to localize in tumor. The promise of these drugs is considerably greater for brain-cancer therapy, since IT can be directly administered intracranially. In an earlier study, we assembled an IL-13 IT made by fusing gene fragments encoding for human IL-13 and the first 389 amino acids of diphtheria toxin.13In vivo studies with this agent showed that it had a potent antitumor effect against IL-13R-positive GBM cells grown in the flanks of nude mice without significant toxicity.
Although encouraging, the flank model has limitations as a model for intracranial therapy. First, intratumoral pressure within the flank model does not reach the pressures that occur in intracranial tumors due to the fixed volume of the skull, thus limiting the depth of IT penetration. Second, flank tumor models are known to autoinfarct once a critical mass threshold has been reached. Third, flank xenograft models can elicit an immune response whereas the intracranial model is privileged due to the blood–brain-barrier.14 Finally, toxicity may be different in the intracranial and flank models. For example, the nonspecific toxicity that can occur with IT treatment, such as vascular leak syndrome, may be more deleterious in an intracranial model limiting the dose that can be safely administered.15, 16
DT390 was chosen to construct DTIL13 because of previous studies describing a series of internal in-frame deletion mutations that established 389 as the optimal site for genetic fusion of DT and targeting ligands.17 In our own experience, fusion proteins made with this mutation retain full enzymatic and translocation enhancing activity, but exclude the native DT binding domain.18, 19, 20 IL-13/IL-13 receptor were chosen because their biology renders them an excellent choice for targeting glioblastoma. IL-13 is structurally homologous to IL-4, containing 4 α helices (A, B, C, and D). IL-13 and IL-4 bind to an IL-13/4 heterodimeric receptor complex composed of the IL-13Rα1 chain (IL-13Rα1) and the IL-4Rα chain (IL-4Rα), transducing the IL-13 signals.21, 22 This receptor is expressed on many normal organs and on adenocarcinomas.23, 24, 25 The fact that IL-13 bound malignant glioblastoma cells independently of IL-4 indicated the existence of another binding moiety specific for IL-13.6, 26 Thus, the monomeric IL-13 receptor α2 (IL-13Rα2) was cloned.27 This protein binds IL-13, but does not bind IL-4.28, 29 Studies have shown that malignant glioblastomas overexpress this restricted receptor for IL-13, and that IL-13Rα2 is the molecular entity responsible for IL-13 binding to glioblastoma tumors.7, 8, 12, 30, 31
An IL-13-based Pseudomonas exotoxin A (PE) fusion protein (FP) showed promising clinical antitumor activity, but no toxicity to normal endothelial, lymphoid, or bone marrow precursor cells.32 DT and PE have identical mechanisms of action and are known to induce cell death with as little as 1 toxin molecule.33 We chose to incorporate a modified diphtheria toxin DT390 in constructing an IL-13 IT, because DT has a clinical history and it may be advantageous to switch therapy to DTIL13 if immunity develops to IL13PE. The fact that humans are mostly preimmunized to DT may be problematic, but advantages of using 1 toxin over the other are currently unknown. Since IL-13 cross-reacts among species,34 we evaluated both efficacy and toxicity of human IL-13 IT in mouse models.
In these studies, we show in mice that DTIL13 can be used as an anti-cancer drug in the treatment of GBM. No apparent vascular toxicity is associated with DTIL13 as measured by MRI and histology. Liver toxicity that is seen is receptor-related, and can be minimized by limiting the dose administered.
Material and methods
Production of recombinant DTIL13
The cytokine fusion toxin gene was assembled using DNA fragments encoding amino acids 21–132 of human IL-13 spliced to DT390 (Fig. 1), as previously described.13 The hybrid gene was ligated into expression vector pET21d (Invitrogen, Carlsbad, CA), and successful cloning of the designed gene was confirmed by restriction endonuclease digestion and sequencing analysis at the University of Minnesota Microchemical Facility (Minneapolis, MN). Plasmid pDThIL13.pET21d was then transformed into the Escherichia coli strain BL21(DER) (Novagen, Madison, WI). Expression was induced by addition of isopropyl-b-D-thiogalactopyranoside (IPTG) (GIBCO, Gaithersburg, MD) and protein from inclusion bodies was solubilized, refolded and purified, as described previously.13 The purified fusion protein DTIL13, Mr 55,539, was then analyzed using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) on a 10% nonreducing gel (Bio-Rad, Richmond, CA) and a Mini-protein II gel apparatus (Bio-Rad). Control DT-based FP used in these studies has been reported previously.17, 18, 19, 35, 36
The cell lines U87 MG and U373 MG, established from human GBM, were obtained from ATCC (Rockville, MD). The cells were maintained in RPMI 1640 medium (Biowhittaker, Walkersville, MD) supplemented with 10% heat-inactivated fetal bovine serum (Biowhittaker), 2 mM L-glutamine (GIBCO), 1.0 mM sodium pyruvate (GIBCO), 100 U/ml penicillin and 100 μg/ml streptomycin (GIBCO). All cell lines were maintained at 37°C in a humidified incubator of 5% CO2, 95% air and passaged twice per week.
Proliferation inhibition activity
The bioassay for measuring the effect of DTIL13 and other fusion proteins was previously described.36 U373 MG or Neuro 2a cells at 1 × 104/well or Daudi cells at 1 × 105/well were plated in a 96-well flat-bottomed tissue culture plate. Adherent cells were incubated at 37°C prior to the addition of ITs. Various concentrations of ITs were diluted in medium and added to cells. All concentrations were tested in triplicate. Cells were incubated for 72 hr and 1 μCi [methyl-3H]-thymidine (Amersham Pharmacia Biotech UK Limited, Buckinghamshire, England) was added to each well for the last 8 hr of incubation. The cell-associated radioactivity was counted, and the effect of the IT was expressed as percent control response, according to the formula: (treated well cpm/untreated well cpm) × 100.
Six–eight-week-old female homozygous athymic nu/nu mice, BALB/c mice, C57BL/6 mice or nu/nu rats were purchased from NCI Frederick Cancer Research and Development Center Animal Production Area (operated by Charles Rivers Laboratories, Hartford, CT) and housed in an AAALAC-accredited specific pathogen-free facility under the care of the Department of Research Animal Resources, University of Minnesota. IL-4R α-knockout mice (BALB/cJ-IL4Rαtm1) were obtained from Taconic (Germantown, NY). These IL4Rα-deficient mice were generated by gene-targeting (exons 4–12 of the IL-4Rα) in BALB/c embryonic stem-cells.37 All animals were maintained in microisolator cages, and sentinel mice were maintained in the facility and routinely monitored for MHV, Sendai and other infectious viruses.
Intracranial mouse tumor model
Athymic nu/nu mice were anesthetized with intraperitoneal injection of 1 mg of Ketamine and 0.1 mg of Xylocaine and placed in a Kopf stereotactic head frame. The scalp was swabbed with betadine, and a midline incision was made with a scalpel. A burr hole was placed 2-mm lateral to the sagittal sinus at the midpoint between the bregma and lambda. To assess intracranial toxicity, various concentrations of DTIL13 in 1-μl media were injected on alternate days for a total of 3 injections, using a Hamilton syringe/needle. Controls were injected with saline. In developing an intracranial tumor model, similar landmarks were used, and various concentrations of U373MG cells were implanted using a Hamilton syringe/needle. The mice were followed to assess the natural history of the tumor concentration.
Serum chemistry study
All organs were harvested 24 hr after the third and final intracranial injection of DTIL13. Individual serum samples were obtained from blood via a heart puncture. Only 1 time-point was possible, since the tests we ran required >0.5 ml of serum. This requires all of the blood we can get from a single mouse. Therefore, sequential measurements were not possible on the same mice. Blood was centrifuged immediately at 5,000 rpm, and serum was frozen at −80°C. Serum samples were analyzed on a Kodak ECTA-CHEM 950 by the Clinical Chemistry Laboratory, Fairview University Medical Center (Minneapolis, MN).
Tissue specimens of brain, heart, lung, spleen, liver and kidney were taken from mice. Histology studies were performed as described previously.38 All samples were embedded in OCT compound (Miles, Elkhark, IN), snap frozen in liquid nitrogen and stored at −80°C until sectioned. Serial 4-μm-sections were cut and mounted on glass slides and fixed for 5 min in acetone. Slides were stained with hematoxylin and eosin (H&E).
Intracranial toxicity imaging
To assess the accuracy of MRI to detect vascular toxicity and cellular toxicity, rodents were injected with 5 μg of DTIL13 or control of 0.9% NaCl intracranially, using a Hamilton needle. Seventy-two hours after injection, the rats were sedated using ketamine/xylocaine, and then imaged using a 1.5-Tesla MR system (NT-ACS, Philips Medical System). Coronal MRI images of the brain were obtained using diffusion-weighted or gradient echo-weighted scans. Diffusion-weighted imaging (DWI) is very sensitive in detecting stroke. Gradient echo imaging is useful in detecting acute hemorrhage. A nu/nu rat was scanned after injection with 3 μl of tail vein blood, to provide a positive control.
Construction and purification of DTIL13
DNA sequencing analysis (University of Minnesota Microchemical facility) verified that the hybrid gene was cloned in frame. SDS-PAGE analysis was performed, and the gel was stained using Coomassie Brilliant Blue (Fig. 1). Lane 1 shows inclusion bodies preinduction with IPTG. Lane 2 shows inclusion bodies after induction with IPTG and a prominent inclusion body band at 55 kDa. Lane 3 shows purified DTIL13 following ion-exchange chromatography. A comparison to the molecular weight standards in Lane 4 revealed a band of about 55 kDa, the expected size of DTIL13 with a purity of 95%.
DTIL13 proliferation assay
To determine the potency of DTIL13, U373 MG cells were cultured in the presence of increasing concentrations of DTIL13, DTmIL2 or DTmIL4 for 72 hr. Figure 2 depicts the selective cytotoxicity of DTIL13 against U373 MG cells with an IC50 of 0.0004 nM. The IC50 is the concentration inhibiting 50% control activity. Irrelevant controls DTIL2 (mouse) and DTIL4 (mouse) made with the same DT390 cassette had no activity against U373MG cells. DTIL13 did not have any cytotoxicity against control cell lines Daudi and Neuro 2a (data not shown).
U373 MG intracranial tumor model
To verify that injected glioblastoma cells could form tumors in the nude-mouse brain, histopathology studies were performed. Hematoxylin and eosin stained slides confirmed active mitosis in the xenograft tumor cells transplanted into the murine brain (data not shown). To determine the capability of the tumor to grow over time, we performed serial MRI scans so that we would not have to euthanize animals unnecessarily for serial histopathology sections (data not shown). Post-gadolineum-injected images of the murine brain showed exponential progression of the tumor size. In fact by post implantation day 42, the tumor volume had nearly reached the size of 1 hemisphere. We conclude that the tumor shows robust growth in a nude murine brain model. To determine the optimum cell concentration for injection and the natural course of tumor growth in the xenograft model, 103 to 5 × 106 cells/3 μl were injected into athymic mice (10 mice per cell concentration). Figure 3 depicts the survival curve of mice administered various cell concentrations of U373 MG intracranially. Since 2 million cells diluted in 3 μl of media were sufficient to kill all mice with an average survival of 53 days, this concentration was chosen to study the efficacy of DTIL13 in an intracranial tumor model. Together, these data indicate that U373MG tumor development is dose–dependent, and this tumor makes a useful intracranial model of human glioblastoma.
DTIL13 intracranial toxicity
To determine whether DTIL13 would be tolerated when injected intracranially, C57BL6 mice were given various concentrations of IT every-other-day for a total of 3 injections (n = 5–10/group). Figure 4 depicts that the 1 μg/injection of DTIL13 had no associated mortality. In addition to assessment of survival following injection of DTIL13, changes in the body weight were measured on a daily basis. The 2-μg injections resulted in a 14% loss of total body weight, whereas the 1-μg injections did not result in any loss of body weight (data not shown).
Brain histology of mice injected with 1-μg DTIL13 did not show any focal necrosis; however, injections with higher doses of DTIL13 resulted in some necrosis. For example, Figure 5 shows a brain section from a mouse treated with 7-μg DTIL13. The arrows indicate the area of infarct. Interestingly, histology findings correlated well with the MRI scans obtained post-DTIL13 injection that showed focal infarct on DWI.
Figure 6 shows coronal DWI scans of nu/nu rats postinjection with DTIL13 or negative control normal saline. For positive controls, the middle cerebral artery was surgically occluded to demonstrate how stroke would appear using DWI. The arrow points to an area of infarct that appears as a hyperdense signal on the scan. Although higher doses of DTIL13 resulted in local stroke, these images of rats injected with 1 μg of DTIL13 showed no overt evidence of infarct.
To address the reported side effect of intracranially administered ITs, specifically vascular leak syndrome,15, 16, 39, 40 MRI scans sensitive to endothelial cell damage were obtained 1 and 72 hr after injection of DTIL13. No direct vascular damage was seen in terms of hemorrhage on gradient echo scans (Fig. 7). Positive controls depict the signal changes following a vascular injury in both figures. In Figure 7, the positive control is a scan of a nu/nu rat injected with 3 μl of blood obtained from a tail vein. Acute hemorrhage appears dark on gradient echo imaging. Together, these findings indicate that the 1 μg dose regimen of DTIL13 did not induce vascular damage by MRI.
Systemic toxicity was evaluated with organ histology and blood serum enzymes. Tissue samples of the kidneys showed mild monocyte infiltration into glomeruli with no damage to the tubular or peritubular structures (data not shown). BUN levels were slightly elevated (not shown), but based on our previous published studies,38 these had minimal effect on kidney function. Creatinine values also were unaffected. Additionally, histology of the heart, spleen, gastrointestinal tract and lungs showed no evidence of toxicity.
In contrast, liver histology showed that high-dose DTIL13 damaged the liver as evidenced by the perivascular fibrosis and fatty degeneration, shown in Figure 8. Enzyme-testing results on the same mice correlated with histology findings and revealed liver enzymes (ALT and AST) also markedly increased in a dose-dependent fashion in BALB/c mice (Fig. 9a). Enzymes were measured at only 1 time point. However, it is unlikely that these measurements were transient given the level of histologic damage to the liver. To determine whether toxicity findings were strain-specific, the study was repeated using C57BL/6 mice. Similar dose-dependent toxicity was noted in both histology and serum enzyme levels (not shown).
To differentiate between receptor specific and nonspecific toxicity, we performed similar toxicity studies in IL-4 α-receptor knockout mice (Fig. 9b). Since a functioning IL-13 receptor requires the IL-4-α receptor as a subunit, transgenic mice lacking this subunit should not have a functioning IL-13R41, 42 The results of this experiment revealed no damage to the liver cytostructure and no significant increase in the serum levels of AST and ALT compared to the BALB/c and C57BL/6 mice. These studies indicated that the damage observed at these higher doses was related to the presence of a functional IL13 receptor or IL4 receptor; and therefore, the liver toxicity was related to the direct binding of DTIL13 to the IL-13 or IL-4 receptor in the liver. Furthermore, Akaiwa et al. have demonstrated the presence of both the IL-13R and IL-4R on hepatocytes and bile duct cells in the liver.43
DTIL13 intracranial efficacy
Based on the toxicity studies that established the maximal tolerated dose (MTD), intracranial efficacy studies were performed with the 1-μl dose regimen of DTIL13. As in the toxicity studies, mice given U373 MG tumor cells were injected with either DTIL13 or normal saline on an every-other-day basis for a total of 3 injections at posttumor implantation day 43. Based on the established tumor-survival curve, this was ∼10 days before death. MRI scans were obtained on postimplantation day 50. Figure 10 depicts postcontrast coronal T1-weighted MRI scans of mice on day 43 (before treatment) and on day 50 (posttreatment). A marked decrease in the size of the tumor burden can be seen. Treatment of the mice with DTIL13 increased survival by 35%. Kaplan–Meyer survival curve (Fig. 11) shows a statistically significant (p < 0.0001) increase in survival time in 1-μg/injection DTIL13-treated mice as compared to controls. Together, these data show that U373 MG provides an aggressive intracranial model to study the efficacy of DTIL13, and that MRI can be used to assess tumor growth, treatment toxicity and efficacy.
The major contribution of this work is the description of the efficacy of the fusion protein DTIL13 in an intracranial xenograft tumor model along with a description of its toxic side effects. DTIL13 showed a statistically significant (p < 0.0001) decrease in the tumor burden with a 35% increase in survival time. The increase in survival is impressive in this model, because the mice were treated at a time point when the tumor burden was far advanced, mimicking the true clinical state. Although the U373 MG model used in these studies brings us closer to a clinical model than do flank tumors, some would argue that it is not an exact clinical correlation since the U-373 MG tumor grows in a noninfiltrative manner. In the future, it will be important to test DTIL13 in an infiltrative model with a prolonged infusion of DTIL13.
The advantage of our study lies not only in demonstrating efficacy of the DTIL13 IT, but also in establishing a more clinically-relevant, murine brain-tumor model. Although MRI scans of the brain obtained after tumor treatment with DTIL13 showed a greater than 75% reduction in tumor volume, complete tumor eradication was not accomplished. One explanation for this result is the limited diffusion of locally injected ITs. Previous work has been published with regards to the intracranial distribution of ITs.44, 45, 46, 47, 48
Since intracranial injection of ITs relies on concentration-based diffusion, the distance of penetration decreases exponentially as the IT concentration declines. Thus, recent studies show that convection-enhanced delivery (CED) of IT is superior to bolus injection of IT.49 Investigators compared the concentration of 14C-sucrose in rat brain via intravenous, intrathecal and convection-enhanced delivery and found that convection-enhanced delivery resulted in 10,000–fold increase in tissue concentration of 14C-sucrose versus intravenous delivery.50 Others have demonstrated that CED of 125I-boronated epidermal growth factor into rat brains with glioma tumor resulted in 47.4% injected dose/gram (IDg) of the IT being detected within the tumor bed even after 24 hr, whereas intratumoral injection of the IT resulted in only 33% ID/g being detected within the tumor bed 24 hr after injection.47 To overcome this limitation of DTIL13 in a clinical trial, it is logical that convection-enhanced delivery should be used. We are currently studying the ability of infusion pumps to maintain a steady IT concentration in rodents.
Although one disadvantage of the xenograft brain-tumor model is the lack of invasion of the tumor into the surrounding brain parenchyma, our study shows that DTIL13 was able to kill the U373 MG cells without any toxicity to the surrounding brain tissue as evidenced by a lack of encephalomalacia on posttreatment scans. Furthermore, no animal demonstrated a neurological deficit posttreatment with DTIL13.
Because brain tumors release factors that result in vascular leak and brain edema, 1 major concern with the use of an IT is the potential for brain herniation due to the exacerbation of preexisting brain edema. We addressed this with various MRI scanning sequences intended to detect deleterious effects of DTIL13 on the brain. Specifically, gradient echo imaging, which is very sensitive to intracranial hemorrhage, showed no toxin-related hemorrhage. There was a small focus of infarction on DWI at the site of the DTIL13 injection that was only evident with higher toxic doses of the IT. Histological analysis of toxin-treated murine brain corroborated the results of MRI in that DTIL13 was not directly toxic to vascular endothelium. Together, these results confirm an advantage of DTIL13 with regards to cytotoxicity of brain vasculature.15, 16, 39, 40
The dose-limiting toxicity of DTIL13 was hepatotoxicity. Perhaps, this was not surprising since previous clinical reports documented hepatotoxicity with IT administration.51, 52, 53, 54 Our studies showed a direct relationship in the elevation of liver function enzymes with an increase in the concentration and total dose of DTIL13. We suspected hepatotoxicity of DTIL13, because investigators previously showed that IL13R and IL-4R were expressed on hepatocytes and bile duct cells in the liver.43, 55 Therefore, an IL4Rknockout model was used. IL-13 has been shown to bind IL-4R.28, 36 Our studies showed that toxicity was reduced in the knockouts that did not have IL-4 receptor. These studies indicated that the hepatotoxicity was attributed to specific binding of the IT, and not to the nonspecific presence of DT. Our studies on knockouts also indicate that the IL-4 receptor plays an important role in hepatic toxicity and argue that investigators are correct in attempting to mutate IL-13 so as to improve its binding and lessen its toxicity to normal organs. Studies have already detected hot spot areas in the α helix D region of IL-13 that are crucial for binding and mutations in this region enhance binding.56, 57. Further mutations to lessen toxicity are currently being explored.
In clinical trials with DTIL13, we anticipate a small therapeutic window based on our current animal toxicity studies. Thus, the clinical trial should proceed in small dose increments. Still, several scenarios might decrease the hepatotoxicity of DTIL13 in the clinical setting. First, convection-enhanced delivery of DTIL13 directly into the tumor bed may result in a “tumor sink” effect (i.e., binding to the tumor receptors may diminish the systemic circulation of the intact DTIL13). Second, the relatively small amount of the IT administered into the tumor bed in relation to the intravascular volume may decrease the effective systemic concentration of the IT with regards to liver damage. Third, pharmacokinetic experiments performed in our lab involving tissue distribution of intracranially injected radiolabeled DTIL13 have shown that DTIL13 is mainly cleared by the kidneys (data not shown); therefore, keeping the patient well hydrated and using gentle diuresis may help remove the circulating DTIL13 rapidly.
These studies did not measure the immunogenicity of the various IT. However, based on the known immunogenicity of bacterial toxins, one could predict that the IT would elicit an antibody response. Still, it is uncertain whether such responses will be problematic in cancer patients receiving immunosuppressive chemotherapy.
In summary, these studies show for the first time that when DTIL13 is administered directly into the cranium, it is highly efficacious with an acceptable therapeutic index. Also, these studies show that pre-and posttreatment MRI can be a useful tool to study antitumor effectiveness in mouse brain tumors. Finally, the agent appears highly effective, even in instances where the tumor burden has been permitted to reach a clinically relevant size before treatment.
We wish to acknowledge David Kuroki and Vincent Vallera (University of Minnesota) for their helpful suggestions and assistance.
- 15Immuntoxin and vascular leak syndrome. Cancer 2000; 6: 218–24..