Targeted Toxins for Glioblastoma Multiforme: pre-clinical studies and clinical implementation

Abstract

Glioblastoma multiforme (GBM) is most common primary brain tumor in adults. GBM is very aggressive due to its poor cellular differentiation and invasiveness, which makes complete surgical resection virtually impossible. Therefore, GBM’s invasive nature as well as its intrinsic resistance to current treatment modalities makes it a unique therapeutic challenge. Extensive examination of human GBM specimens has uncovered that these tumors overexpress a variety of receptors that are virtually absent in the surrounding non-neoplastic brain. Human GBMs overexpress receptors for cytokines, growth factors, ephrins, urokinase-type plasminogen activator (uPA), and transferrin, which can be targeted with high specificity by linking their ligands with highly cytotoxic molecules, such as Diptheria toxin and Pseudomonas exotoxin A. We review the preclinical development and clinical translation of targeted toxins for GBM. In view of the clinical experience, we conclude that although these are very promising therapeutic modalities for GBM patients, efforts should be focused on improving the delivery systems utilized in order to achieve better distribution of the immuno-toxins in the tumor/resection cavity. Delivery of targeted toxins using viral vectors would also benefit enormously from improved strategies for local delivery.

INTRODUCTION

Glioblastoma multiforme (GBM) is highly malignant due to its poor cellular differentiation, invasiveness, and surgical inaccessibility to complete resection. GBM is characterized by the presence of necrosis, microvascular proliferation, nuclear atypia, and invasion into the surrounding non-neoplastic brain ¹. GBM’s infiltrative nature and intrinsic resistance to current treatment modalities make it a unique therapeutic challenge ². However, extensive scrutiny of GBM samples from human patients has uncovered that these tumors overexpress several receptors that distinguish them from the surrounding normal brain parenchyma. Human GBMs overexpress receptors, which are virtually absent in the normal brain such as those for cytokines ^3–9, growth factors ^10–14, ephrins ^{9, 15–17}, urokinase-type plasminogen activator (uPA) ¹⁸, transferrin ^{19, 20}. Thus, specific targeting of cytotoxic molecules towards tumor cells has been attempted by linking them with ligands for these GBM specific receptors (Figure 1).

Figure 1. Illustration depicting the use of various targeted approaches under development for the treatment of GBM.

A) The receptors (left) and their ligands (right) under development for targeted toxin therapies are shown. Various toxins (T) are linked to specific ligands in order to elicit the binding and internalization of the targeted toxins to receptors that are overexpressed on GBM cells. The rationale underlying the use of targeted toxins is shown in (B).

Receptors for IL-13 have been detected in 50–75% of human GBM specimens ^{3, 9, 11, 21}. In normal tissues, receptors for IL-13 consist of the IL13α1R chain which requires heterodimerization with the IL4Rα chain in order to bind IL-13 with high affinity ^{4, 22, 23}. IL-13 and IL-4 compete for the shared physiological signaling receptor IL13/IL4R present in normal cells ^{4, 22, 23}. GBM cells overexpress a distinct monomeric receptor for IL-13, which consists of the IL13Rα2 chain ^{3, 24–26}. IL-13 binds to GBM-associated IL-13Rα2 with high affinity, while IL-4 does not interact with this receptor ^{3, 24–26}. Although the heterodimeric receptor complex composed of IL-13α1 receptor and IL-4 receptor (IL13/IL4R) is widely expressed among normal tissues ²⁷, GBM-associated IL13Rα2 seems to be restricted mainly to the testes among normal peripheral tissues, and virtually absent from the CNS ³.

Overexpression of EphA2, a receptor tyrosine kinase for ephrinA1, has been observed in most of human GBM specimens ^{16, 17} but is undetectable in normal brain ¹⁷. EphA2 is an attractive target due to the fact that it is not only expressed on GBM cells, but also in the tumor microvasculature ⁹. In addition, expression of EphA2 in GBM has been shown to be associated with a more aggressive phenotype and a poorer clinical outcome ^{16, 17}.

Amplification of the epidermal growth factor receptor (EGFR) gene is the most common genetic alteration found in human GBM ²⁸. This results in overexpression of EGFR protein, which has been detected in at least half of human GBM specimens ^{11, 29, 30}. Between 50–60% of EGFR⁺ GBMs also express a mutated form of the receptor, EGFRvIII, which is a truncated and constitutively active form of EGFR that is not present in normal tissues ³¹. The association of EGFR overexpression and clinical outcome for GBM patients remains controversial ^{29, 30}.

Receptors for transferring, TfR1 and TfR2, have been detected in the majority of GBM patients ^{11, 32, 33}. TfRs are mainly restricted to the luminal surface of the brain capillaries with minimal dispersion throughout the normal brain tissue ^34–36, thus making TfRs attractive candidates for targeted toxins. The binding of the ligand transferrin (Tf) to its receptor TfR mediates the transport of iron into actively dividing cells. Each Tf molecule has two lobes, each of which serves as an iron-binding site; thus, each Tf molecule is capable of binding and transporting two ferric ions (Fe³⁺) into a cell. TfR is only able to bind Tf when both lobes of the transferrin molecule are bound to ferric ions (holo-Tf). After binding of holo-Tf to TfR, the holo-Tf/TfR complex undergoes receptor-mediated endocytosis. Upon entry of the holo-Tf/TfR complex into the cell, two Fe³⁺ ions are released from holo-Tf, thus re-forming the Tf ligand in its iron-free state (apo-Tf) while remaining bound to the TfR. The apo-Tf/TfR complex is then recycled back to the cell surface where the iron-free apo-Tf is released from the TfR back into the extracellular space. The unbound apo-Tf can then bind two more ferric irons and repeat the endocytic cycle ³⁷. The endocytic capacity of TfR has been exploited to introduce cytotoxic toxins within GBM cells.

Overexpression of receptors for urokinase-type plasminogen activator (uPAR) has also been observed in human GBM and is associated with poorer prognosis due to its role in invasion ^{18, 38–40}. These receptors are virtually confined to GBM cells and that their presence seems to be associated with a more aggressive tumor phenotype, targeting of highly cytotoxic compounds towards tumor cells has been attempted using their natural ligands. Considering that uPAR is found in tumor microvasculature ⁴¹, this receptor is also an attractive target to impact tumor angiogenesis ⁴².

Targeting of IL-13Rα2

A number of recombinant proteins consisting of catalytic toxins fused to IL-13 have been developed to deliver highly toxic agents to IL13α2R⁺ GBM cells ^{4, 43, 44}. Pseudomonas exotoxin A [PE] is a cytotoxic bacterial protein which encompasses three functional domains. Domain I binds the α2-macroglobulin receptor, which is ubiquitously expressed in normal tissues; the exotoxin-α2-macroglobulin receptor complex undergoes internalization by endocytosis ⁴⁵. Domain II is a site of proteolytic cleavage that activates PE and is required for catalyzing the translocation of the catalytic domain III into the cytosol. Once in the cytosol, Domain III directs the processed fragment of the toxin to the endoplasmic reticulum where it inactivates the elongation factor 2 through ADP ribosylation, inhibiting protein synthesis and leading to cell death ⁴⁵. The mutant exotoxin, PE38QQR ⁴⁶, does not bind to the ubiquitous α2-macroglobulin receptor due to the deletion of domain I ⁴⁶, hence it can be linked to various ligands in order to promote its internalization into target tumor cells. To target the PE toxin to human glioma cells, a fusion protein was constructed that links the N-terminal domain of PE38QQR to native hIL-13 (hIL-13-PE, Cintredekin Besudotox) ⁴. The sensitivity of GBM cells to hIL-13-PE correlates positively with the density of IL-13Rα2 ⁴⁷ and it has been demonstrated that this toxin exerts an anti-tumor effect in preclinical models of human GBM ^47–50. Systemic administration of the hIL-13-PE is limited by dose ⁴⁸ due to toxic side effects. Preclinical ^{4, 7, 48, 51, 52} and clinical trials ⁵³ have shown that local administration can be used to successfully deliver chimeric toxins to brain tumors. However, since the toxin remains active for only up to six hours at the site of injection, intratumoral administration requires multiple injections to be effective ^{51, 54}. Convection enhanced delivery (CED) of the hIL-13-PE has been tested in a Phase III clinical trial for GBM ⁵⁵ and is associated with dose-related neurological side effects in most of the enrolled patients ⁵⁶ (Table 1)

Table 1.

Summary of clinical trials using convection enhanced delivery (CED) of targeted toxin approaches for the treatment of brain tumors

Receptor target	Ligand	Toxin	Clinical Stage	# of patients	Main findings	Publication
IL-13Rα2	IL-13	PE38	Phase I	46	Interim analysis of three Phase I trials.	Kunwar et al., 2003
IL-13Rα2	IL-13	PE38	Phase I	22	MTD was established as 0.5 µg/ml administered via CED.	Vogelbaum et al., 2007
IL-13Rα2	IL-13	PE38	Phase I	51	Final analysis of three Phase I trials: MTD was established as 0.5 µg/ml administered via CED. Infusion duration of up to 6 days was well tolerated. Postoperative catheter placement is critical for optimal drug distribution.	Kunwar et al., 2007
IL-13Rα2	IL-13	PE38	Phase III	296	Efficacy of drugs delivered by CED may be hampered by ineffective delivery.	Sampson et al., 2010
IL-13Rα2	IL-13	PE38	Phase III	296	No survival difference between IL13-PE38QQR when compared to Gliadel wafers	Kunwar et al., 2010
IL-4R	IL-4	PE38	Phase I	9	No systemic toxicity observed. Necrosis was observed in 6 of 9 patients.	Rand et al., 2000
IL-4R	IL-4	PE38	Phase I/II	31	No drug-related systemic toxicity; treatment-related adverse effects were limited to the central nervous system. MTD was defined as 6 µg/ml × 40ml.	Weber et al., 2003
IL-4R	IL-4	PE38	Phase I/II	31	Case study of long-term survivor.	Rainov et al., 2005
EGFR	TGFα	PE38	Phase I	20	Progress Report: 3 of 15 patients, with residual disease at the time of therapy, demonstrated radiographic responses and one patient with a complete response that survived longer than 83 weeks	Sampson et al., 2003
EGFR	TGFα	PE38	Phase I	20	Case Study of long-term survivor.	Sampson et al., 2005
EGFR	TGFα	PE-38	Phase I	20	Final Results: MTD was not reached. Dose escalation was stopped at 100 ng/ml due to inconsistent drug delivery.	Sampson et al., 2008
Transferrin-R	Transferrin	DT	Phase I	18	No symptomatic systemic toxicity occurred. MTD was defined as 0.67 µg/ml × 40ml.	Laske et al., 1997
Transferrin-R	Transferrin	DT	Phase II	31	35% of the patients displayed complete or partial responses, and only 38% of the patients had progressive disease after treatment. Toxicities associated with Tf-CRM107 were progressive cerebral edema and seizures	Weaver et al., 2003

Initially it was thought that binding sites for human IL-13 were exclusively localized in GBM cells and not present in the non-neoplastic brain ⁵⁷. However, although IL-13Rα2 seems to indeed be confined to GBM cells ^{9, 21, 26, 58}, several reports show that normal brain cells express IL4αR. Expression of physiological IL4αR chain was observed in human non-neoplastic astrocytes in vitro ⁵⁸ and in vivo in reactive astrocytes surrounding human astrocytoma specimens ⁵⁹. These findings suggest that the specificity of the hIL-13-PEQQR may be lower than hitherto anticipated. In fact, clinical trials in GBM patients showed that intracranial administration of hIL-13-PE38QQR led to dose limiting toxicities in some patients, including neurological symptoms secondary to necrotic and inflammatory processes as well as irreversible hemiparesis and the death of one patient due to neurologic decline possibly related to Cintredekin Besudotox ⁵⁶. Grade III and IV imaging changes were observed in tumor infiltrated and normal brain parenchyma that were indicative of tissue damage ⁶⁰. Brain tissue damage was regarded as a result of nonspecific internalization of hIL-13-PE38QQR by normal brain cells ⁵⁶. In fact, we have recently demonstrated that a single injection of hIL-13-PE38QQR into the naïve mouse brain leads to acute neurological deterioration and severe neuropathological changes, even at low doses ⁵⁰.

To optimize the targeting of the toxin to the GBM-associated IL-13Rα2, the IL-13 gene has been engineered to generate a mutant form of IL-13 (mhIL-13, IL-13.E13K) ⁶¹. It has been shown that IL13.E13K fused to PE (mhIL-13-PE) binds to GBM-associated IL13α2R with 50-fold higher affinity compared to native hIL-13 ^{61, 62}. Importantly, unlike its native counterpart, mhIL13-PE does not interact with the physiological IL13/IL4R ^{61, 62}, hence, decreasing the capacity of the chimeric toxin to bind to normal cells. Although this second-generation cytotoxin exerts a stronger anti-tumor effect than first-generation hIL-13-PE in intracranial human GBM models, its administration into naïve mouse brain leads to dose-dependent neurotoxicity ⁵⁰.

In a recent publication from our lab, we showed the development of a novel third-generation IL-13-based cytotoxin ⁵⁰. We developed an adenoviral vector (Ad, Fig. 2) encoding mhIL13-PE to provide long-term high local expression of the targeted toxin, leading to an effective cytotoxic response in IL-13Rα2-expressing GBM cells without adverse side effects to surrounding normal brain tissue. The expression of mhIL-13 fused to PE toxin from an Ad vector allows direct and continued targeting of mhIL-13-PE to GBM cells in situ. We demonstrated that a single intratumoral injection of the therapeutic vector in intracranial human GBM xenografts and syngeneic GL26 tumors implanted in immune-competent mice leads to tumor regression and long-term survival in 50–70% of the animals (Fig. 3) ⁵⁰. To further increase the safety of this therapeutic vector, we also encoded a mutated IL-4 (mIL4, IL-4.Y124D). Since mIL-4 is known to bind and block the IL13R/IL4R present in normal cells without interacting with IL-13Rα2 ^{61, 62}, we hypothesized that mIL4 would block any potential binding of the mhIL-13-PE to normal cells, without effecting the binding of the chimeric toxin to malignant cells. The expression of these transgenes is under the tight control of the regulatable bidirectional TRE promoter, allowing the inhibition of transgene expression by withdrawal of the inducer if adverse side effects were to occur. The TetON system consists of rtTA2S-M2 (TetON), which triggers expression of mhIL-13-PE and mIL-4 in the presence of Dox, and the Tet-repressor tTSkid, which inhibits transgene expression in the absence of Dox; both systems function to inhibit potential leakiness of this system ^{63, 64}. This approach has several advantages over traditional protein formulations of IL-13 cytotoxins (Fig. 4): i) once synthesized by infected cells, mhIL13-PE is released ⁵⁰ and exerts a powerful bystander effect, inducing apoptosis of GBM cells expressing the IL-13α2R located within the diffusion range of the toxin and amplifying the therapeutic efficiency of our approach; ii) this approach is highly specific and exhibits negligible toxicity towards normal brain tissue ⁵⁰ since mhIL-13-PE specifically binds to GBM cells expressing IL-13Rα2, hence sparing normal brain cells. Additionally, the co-expression of mIL4 blocks any putative negligible binding of the toxin to normal cells.

Figure 2. Regulated transgene expression from the therapeutic adenoviral vectors.

Ad-mediated regulated therapeutic cytotoxin expression. Ad.mhIL-4.TRE.mhIL-13-PE expresses mhIL-4 and mhIL-13-PE under the control of the bidirectional TRE promoter, which is activated by rtTA2^sM2 in the presence of Dox (ON state, A). In the absence of the inducer (OFF state, B) the transactivator is unable to induce transgene expression.

Figure 3. Intratumoral administration of Ad.mhIL-4.TRE.mhIL-13 mediates tumor regression and long term survival in orthotopic models of GBM.

Nude and Rag1^−/− mice were implanted in the brain with human U251 glioma cells and human primary GBM12 cells, respectively. Five days later, they were treated with either a single intratumoral injection of control vector Ad.mhIL-4.TRE.mhIL-13 or saline. Animals were fed Dox-chow to activate transgene expression.

Figure 4. Gene therapy-mediated delivery of mhIL-13-PE leads to anti-tumor efficacy and long term survival in the absence of neurological toxicity.

The diagram depicts the mechanism of action of the targeted gene therapeutic strategy. The targeting of IL-13α2 receptor overexpressed in glioma cells has been approached by constructing an Ad vector encoding the truncated form of PE toxin fused to a mutated form of human IL-13 (mhIL-13), which has higher affinity for the glioma-associated IL-13α2 receptor and negligible binding to the physiological receptor, IL13/IL4R. Binding of mhIL-13-PE to IL13Rα2 promotes its internalization into glioma cells. Domain II (PEII) mediates the translocation of the toxin into the endosomes by endocytosis. Once in the endosomes, furin mediates proteolytic cleavage that activates the catalytic Domain III (PEIII). Due to the low pH of the endosome, the processed fragment of the toxin is translocated to the cytosol and inhibits protein synthesis, leading to glioma cell death. The absence of IL-13Rα2 protects normal cells from mhIL-13-PE mediated cell death. The safety of our approach has been further enhanced by co-expressing a mutated form of IL-4 (mhIL-4) encoded in the therapeutic bi-directional Ad which acts as an antagonist of the physiological receptor comprising the IL-13α1R and IL-4αR chains. Secreted IL-4 would inhibit the already negligible binding of mhIL-13-PE to normal brain cells, without affecting the binding of the mhIL-13-PE to GBM cells.

Previous reports aiming at controlling the expression of ADP-ribosylating toxins consisted of plasmids expressing native or attenuated Diphtheria toxin (DT) ^{65, 66}. Plasmids expressing DT under the control of the first generation Tet-OFF system showed superior control of transgene expression in vitro when compared to lacR-IPTG system ⁶⁵. While the first generation Tet system fails to completely inhibit transgene expression in the “OFF” state ⁶⁷, the third generation Tet system used by us constitutes a non leaky inducible system ideal for delivering such toxic genes. Another approach to increase the specificity and reduce the toxicity of targeted toxins is to express them under the control of a cell-type specific` promoter. Native DT expressed under the control of a prostate specific antigen (PSA) promoter and delivered using an Ad vector in a mouse model of prostate cancer led to ~80% long term survival ⁶⁸. However, transgene expression is much higher when driven by ubiquitous promoters than when using cell-type specific promoters ^{69, 70}. The use of strong promoters elicits high transgene expression levels using low doses of Ads, which minimizes the risk of neurotoxicity due to high viral vector loads ^{71, 72}. Considering that the specificity of our approach relies on the targeting of mIL-13-PE to IL-13Rα2, which is only expressed in GBM cells and absent in the normal brain, we expressed the components of the Tet-ON system under the control of the potent murine CMV promoter ⁵⁰. We have previously demonstrated that this promoter elicits high levels of transduction in vitro in human GBM cells and in vivo in mouse and dog brain ^{63, 73, 74}

Since mIL-13-PE kills GBM cells that express IL-13Rα2, treatment with Ad.mIL-4.TRE.mIL-13-PE could promote the selection of IL-13Rα2-negative GBM cells, which could lead to the development of a mIL-13-PE-resistant tumor. Thus, combination with other chemotherapeutic, proapoptotic agents or immunotherapeutic approaches could be attempted ⁷⁵. Additionally, patients that do not display overexpression of IL-13Rα2 may benefit from PE toxin fused to ligands that bind other receptors that are overexpressed in GBM. For example, the receptor tyrosine kinase Eph2R ^{9, 15} is also overexpressed in human GBM specimens, but virtually absent in the non-neoplastic brain ^{9, 17}. Targeting of this receptor has been attempted by fusing the ligand of Eph2R, ephrinA1, to PE38QQR, which exhibits a potent and specific cytotoxic effect against GBM cells ⁷⁶.

Targeting the epidermal growth factor receptor (EGFR)

Two EGFR ligands, TGF-α and EGF, have been fused to toxic proteins and assessed both in pre-clinical animal models of GBM, and in human patients in Phase I clinical trials. In these studies, TGF-α was fused to the mutated variant of pseudomonas exotoxin PE-38 ^{77, 78}. Efficacy of TGF-α-PE-38 fusion protein (TP-38) was assessed in an intracranial brain tumor model consisting of epidermoid carcinoma A431 cells mixed with TP-38 and implanted into the caudate nuclei of athymic mice. Mice receiving tumor cells mixed with either 0.03 µg or 0.1 µg of the fusion protein displayed 100% and 90% survival compared to mice injected with cells alone, which exhibited a median survival of 19 days. Mice treated with higher doses of TP-38 died before the control mice, most likely due to toxicity of the drug ⁷⁷. Toxicity of TP-38 was then assessed in both athymic rats and Rhesus macaques to determine the maximum tolerated dose (MTD). Naïve athymic rats were injected with 20 µL in the caudate nucleus with increasing doses of TP-38. Animals treated with high dose of TP-38 exhibited evidence of demyelination and necrosis and the MTD was determined to be 0.66 ⁷⁷. To assess toxicity studies in Rhesus macaques, 200 µL of TP-38 was infused into the brain. A neurological assessment was performed 7 days later followed by a histopathological analysis of the brain. Neurotoxicity was only observed at the highest dose tested (6 µg), and thus the MTD was determined to be 6 µg ⁷⁷. Based on these data and extrapolation from brain mass, body mass, and body surface area, 1 µg was selected as the starting “low” dose for a human Phase I clinical trial ⁷⁷.

A Phase I clinical trial of TP-38 was undertaken at Duke University and University of California San Francisco with the primary objective to determine the MTD of TP-38 in twenty adult human patients with recurrent brain tumors following infusion of TP-38 into the brain ^{77, 79–81}. The dose escalation study commenced at a total dose of 1 µg, at a concentration of 25 ng/mL, and a rate of 0.4 mL/hour, for a total of 40 mL. Doses were escalated to 2 µg and 4 µg. The authors concluded that most toxicities encountered were solely neurological and most likely unrelated to TP-38, rather instead a consequence of infusion volume, recurrent tumor, or stereotactic catheter placement ⁷⁹. The MTD of TP-38 was never reached in this study, however, two dose-limiting toxicities (DLT) were observed and were of neurologic nature ⁷⁹. The dose escalation of TP-38 was stopped at 4 µg due to inconsistent drug delivery. The secondary objective of this study was to assess efficacy of TP-38. The median survival after TP-38 was 28 weeks. Two patients demonstrated radiographic responses, including one patient who demonstrated a nearly complete response and remained alive >83 weeks post-therapy. The sponsors of the study describe that the efficacy of this approach was limited due to ineffective transfusion of TP-38 in many patients; only 3 catheters produced intraparenchymal infusate distribution ⁷⁹.

EGF has also been fused to the diphtheria toxin to target EGFR in brain tumors. Investigators used a His-Ala linker to fuse the catalytic and translocation domains of diphtheria toxin (DAB₃₈₉) to human EGF with promising results obtained in vitro in a battery of brain tumor cell lines ^{7, 82}. In vitro, efficacy of DAB₃₈₉EGF was strongly correlated with EGFR density on the GBM target cell lines. One report of the efficacy and safety of DAB₃₈₉EGF has been published to date. In this dose escalation study, athymic mice were injected (s.c.) with U87MG human glioma cells in the flank ¹⁰. When a large peripheral tumor mass developed, animals were treated with intratumoral injections of either 1, 3, 5, or 10 µg of DAB₃₈₉EGF, given every other day for three to six doses. The MTD of DAB₃₈₉EGF was demonstrated to be 3 µg given every other day. Animals receiving high doses exhibited clinical symptoms of weight loss, diminished activity, and dehydration. These clinical symptoms correlated with altered blood chemistry including abnormal blood urea nitrogen, creatinine, aspartate transaminase, and alanine transaminase. Histopathological analysis of the liver and kidney at the high doses revealed renal tubular necrosis ¹⁰. At the MTD, tumor regression was observed in all animals, while 25% of the animals (4 out of 16 mice) exhibited a tumor relapse within one month. Relapsed tumors were found to retain their EGF receptor, and responded to a second round of intratumoral DAB₃₈₉EGF.

Another approach uses an antibody-toxin approach to target EGFR overexpressing tumor cells. One example of this approach consists of variable domains of the antagonistic monoclonal antibody 14E1, which is specific for human full-length EGFR, fused to a truncated form of the pseudomonas exotoxin A (ETA). Importantly, scFv(14E1)-ETA specifically recognizes wildtype human EGF receptor and its constituently active mutant EGFRvIII ^83–85. Systemic delivery of scFv(14E1)-ETA in mice was found to completely suppress or reduce pulmonary metastases of intravenously injected renal carcinoma cells overexpressing EGFR or EGFRvIII ⁸³. The cytotoxic effects of scFv(14E1)-ETA and TP-38 were tested on neuroblastoma cells, which displayed elevated levels of EGFR, both in chemosensitive and chemoresistant cell lines. Both scFv(14E1)-ETA and TP-38 resulted in decreased cell viability in cell lines that were insensitive to cetuximab or EGFR tyrosine kinase inhibitors ⁸⁶ currently in use in clinical trials. Furthermore, it was demonstrated that cisplatin treatment resulted in an increase in EGFR expression in neuroblastoma cells and a concomitant increase in cytotoxicity of both scFv(14E1)-ETA and TP-38. Cisplatin-resistant cell lines also displayed an increased sensitivity to both targeted toxin approaches ⁸⁶. These highly encouraging data demonstrate that the therapeutic efficacy of scFv(14E1)-ETA ought to be further explored in orthotopic models of neuroblastoma, and also glioblastoma multiforme.

Targeting the transferrin receptor (TfR)

The transferrin receptor (TfR) is another receptor that is overexpressed on GBM cells ³² and has been exploited for its use to specifically target toxins to GBM. To take advantage of TfR endocytic pathway, researchers have fused human transferrin to an mutated diphtheria toxin (CRM107), joined together by a stable, non-reducible thioether bond ⁸⁷. The CRM107 diphtheria variant has a single point mutation in its subunit B that prevents the toxin from binding to the cell surface without effecting its cytotoxic effects, i.e., inhibiting protein synthesis ⁸⁸. In vitro, Tf-CRM107 was 1,000 to 100,000 more toxic than the unconjugated CRM107 ⁸⁹. To assess efficacy and safety in vivo, subcutaneous tumors comprised of U251 cells in the flanks of nude mice were treated with escalating doses of Tf-CRM107. Each mouse was treated with one of three escalating doses every 2 days. The identical dose was delivered for a total of four times in each tumor bearing animal. Tumor regression was observed at all three doses ⁵³. Toxicity studies revealed that when injected into the brain parenchyma, the dose of Tf-CRM107 required to elicit significant toxicity was 4–5 orders of magnitude greater than the therapeutic dose used to treat subcutaneous U251 tumors ⁸⁷.

A dose escalation Phase I clinical trial was undertaken to assess the toxicity and determine the MTD of Tf-CRM107 in patients with progressive malignant brain tumors (primary or metastatic) who had previously failed conventional therapies ^{53, 90}. In total 18 patients, 10 of which were diagnosed with glioblastoma multiforme, underwent treatment involving the sterotactic placement of multiple catheters followed by infusion of the drug starting at a rate of 0.5 µL/min, eventually increasing to a maximum rate of 4–10 µL/min. The concentrations of drug delivered started at 0.1 µg/mL and escalated until the MTD was achieved. Total infusion volumes of 20, 40, 60, 80, and 120 mL were tested, with the optimal infusion volume that could be safely administered defined at 40mL. Patients were re-treated with Tf-CRM107 at every 4–6 week intervals until there was evidence of tumor resolution by MRI, or until toxicities were noted. Treatment with Tf-CRM107 resulted in a ≥50% decrease in tumor volume in 9 out of 15 patients, and tumor response appeared to be dose dependent. Two patients showed complete response, one of which was tumor free for 23 months after a single infusion of Tf-CRM107. Intratumoral infusions were well tolerated with no treatment related deaths of life threatening toxicity. Localized toxicity, including peritumoral focal brain injury, was observed in patients receiving the higher doses (≥1.0 µg/mL), but there was no evidence of Tf-CRM107-induced tissue damage outside of the CNS. This Phase I clinical trial established the MTD at 0.67 µg/mL delivered in 40 mL.

Using this MTD, a Phase II clinical trial was then begun using CED infusion of Tf-CRM107 from two catheters. The primary objectives of this study were to evaluate the efficacy of Tf-CRM107 in patients with refractory or progressive GBM or anaplastic astrocytoma and to further evaluate the safety of Tf-CRM107. All patients received 0.67 µg/mL of Tf-CRM107 over a period of 4–5 days until the entire 40mL was delivered and then received additional treatments 4–10 weeks later. 44 patients at 9 study sites were treated, of which 31 of the patients received the second administration of Tf-CRM107. In total, 35% of the patients displayed complete or partial responses, and only 38% of the patients had progressive disease after treatment. Toxicities associated with Tf-CRM107 were progressive cerebral edema and seizures ⁹⁰.

A recent publication described the development of the next generation of transferrin-diphtheria toxin ligand conjugate and the assessment of its therapeutic efficacy ⁹¹. Yoon et al. tested the hypothesis that sub-optimal efficacy of Tf-CRM107 was due to the rapid cycling of transferrin through the cell that resulting in a very limited time to deliver the diphtheria toxin. Given the physiological cycling of native human holo-Tf (iron loaded) to apo -Tf (devoid of Fe³⁺ ions), they speculated that up to 30 cycles of transferrin trafficking might be required for efficacious delivery of transferrin bound toxins into the cell. To overcome this limitation, Yoon et al. used mathematical modeling of Tf/TfR trafficking to identify that decreasing the Fe³⁺ release rate following receptor mediated endocytosis could induce the cell to traffic holo-Tf towards the cell surface, effectively “recycling” the transferrin-diphtheria ligand conjugate and increasing the probability of efficacious delivery of the toxin to the cytosol. Yoon et al. had previously developed recombinant mutant variants of Tf that exhibited reduced iron release kinetics ⁹². These mutant Tf-diphtheria toxin conjugates displayed higher efficiency of drug delivery compared to wildtype Tf in both HeLa cells ⁹² and also the GBM cell lines U87 and U251 ⁹¹. When these mutant Tf-diphtheria toxin conjugates were used to treat nude mice bearing flank tumors U87 tumors overexpressing EGFRvIII, they exhibited increased tumor regression compared to animals treated with Tf-CRM107 ⁹¹. These promising data demonstrate the renewed clinical promise of using transferrin to deliver toxins for the treatment of GBM.

Targeting of Eph2

The endogenous ligand for EphA2 receptor tyrosine kinase, ephrin A1, has been fused to PE38QQR for the treatment of GBM ^{9, 76}. EphrinA1-PE38QQR exerts a potent cytotoxic effect on human GBM cells depending on their level of EphA2 expression ⁹. The cytotoxic effect of ephrinA1-PE38QQR appears specific for EphA2⁺ tumor cells since it is not toxic to normal human endothelial cells and EphA2⁻ tumor cells ⁷⁶.

Although EphA2 and IL-13Rα2 are overexpressed in a fraction of GBM patients ⁹, when combined they are expressed in virtually all patients with GBM ⁹. Thus, it has been proposed that a combinatorial approach using IL-13- and ephrin1-based cytotoxins would improve the outcome of patients with GBM ^{9, 76}.

Targeting of uPAR

Tallying with the idea of targeting more than one receptor overexpressed in GBM to achieve therapeutic efficacy in as many GBM patients as possible, bispecific toxins have been developed that target uPAR and EGFR or IL-13Rα2 ^{42, 93}. These bispecific toxins improved efficacy and reduced toxicity than the monovalent toxins.

A bispecific toxin consisting of EGF and a urokinase fused to truncated pseudomonas exotoxin PE38 was generated to target EGFR in GBM cells, and uPAR in the tumor microvasculature ⁴². This toxin, EGFATFKDEL, has the advantage of being less immunogenic since immunodominant B-cell epitopes of the PE38 molecule were mutated ⁴². In preclinical studies, EGFATFKDEL showed superior efficacy when compared to toxins that target EGFR or uPAR alone ⁴².

A recombinant diptheria toxin-uPA fusion protein (DTAT) was developed that contains the translocation and catalytic portions of diphtheria toxin (responsible for cell entry and killing, respectively), fused to the non-internalizing amino terminal fragment portion of human plasminogen activator ⁹³. The DTAT fusion protein targets malignant glioma cells and the endothelial cells of the neovasculature that express the urokinase-type plasminogen activator receptor (uPAR). In the same study, a second protein was engineered (DTAT13) that targets uPAR and IL13Rα2 expressing GBM cells. In preclinical models of GBM, DTAT13 exhibited similar efficacy as DTAT but with reduced toxicity ⁹³. Authors proposed that the bi-specific DTAT13 might target a broader range of GBM patients and that the bi-specificity reduces the amount of toxin required for therapeutic efficacy than if the two agents were administered separately ⁹³.

CONCLUSIONS

The use of targeted toxins for glioma therapeutics is the focus of intense research, related to both the development of novel improved cytotoxins, with enhanced efficacy and lower “off target” toxicities, the implementation of improved in vivo delivery systems including devices and biologicals (i.e., viral vectors, plasmids), and the discovery of novel glioma specific receptors. Several of these approaches have moved from preclinical GBM models to phase I, II and III clinical trials. The results from the preclinical models have been very encouraging ^47–50; in spite of this, the clinical experience has not been as successful ⁹⁴. For example, results from the PRECISE clinical trial have not reported a significant improvement in the median survival of GBM patients and several adverse events were observed ⁵⁶. One of the strongest reasons which could explain the lack of more encouraging efficacy results from the PRECISE phase III trial could be the lack of consistency in the delivery of the immunotoxin Cintredekin Besudotox ^{55, 56, 94–96}. In order to overcome the short half life of the protein Cintredekin Besudotox, clinical trials used convection enhanced delivery (CED) to administer Cintredekin Besudotox intracranially in GBM patients, who received 96h infusions through 1–3 catheters placed before and/or after tumor resection into the brain tumor mass or the brain parenchyma surrounding the tumor cavity, respectively ⁹⁵. Although CED produces widespread drug dissemination in some patients, this technique is complex and leads to varied, discrepant results when comparing different centers ^{95, 96}. The anatomy of the target area as well as the accuracy of catheter positioning had strong influence on the distribution of the protein, resulting in heterogeneous and limited distribution of the drug ^{95, 96}. In fact, the efficacy of Cintredekin Besudotox in GBM patients was strongly dependent on the accuracy of catheter placement ^{55, 56}. In addition, preliminary information from a recent Phase III clinical trial indicated that CED of Cintredekin Besudotox did not significantly improve the median survival of GBM patients when compared to patients treated with Gliadel wafers ⁹⁷. In summary, although this is a very exciting field which could lead to novel improved treatments for GBM patients, substantial efforts should be devoted to improving the delivery approaches utilized in order to enable distribution of the immuno-toxins safely and accurately into the tumor/resection cavity. Also, delivery of targeted immuno-toxins using viral vectors would benefit enormously from improved devices for local delivery.