Growth Factor Receptor Expression Varies among High-Grade Gliomas and Normal Brain: Epidermal Growth Factor Receptor Has Excellent Properties for Interstitial Fusion Protein Therapy

High-grade gliomas including anaplastic astrocytoma and glioblastoma multiforme are diagnosed in close to 20,000 people per year in the United States (1). Despite aggressive treatment with surgery, irradiation, and chemotherapy, the 2-year survival rate is less than 20% (2). Two recent developments may yield improved clinical outcomes. First, direct administration of antiglioma agents into the tumor bed has been achieved using CED² or interstitial therapy (3). Catheters are placed in the tumor, and fluid bathes the tumor under pressure. This technique bypasses the blood-brain barrier, achieves high local concentrations of agent with low systemic concentrations, and improves delivery to tumor cells of high molecular weight compounds. Second, fusion proteins have been developed that selectively target and kill brain tumor cells. The peptide ligand of the fusion protein binds glioma cell surface receptors, the fusion protein-receptor complex undergoes receptor-mediated endocytosis, the toxin catalytic domain of the fusion protein translocates across vesicle membranes and reaches the cytosol, protein synthesis is enzymatically irreversibly inactivated, and cell death ensues (4). Brain tumor cells with resistance to DNA-damaging agents, including radiation and chemotherapy, are sensitive to the protein synthesis-inhibiting fusion proteins. Four different brain tumor fusion proteins have been administered by CED to patients in Phase I/II clinical studies including Tf-CRM107, IL13PE38QQR, IL4(38-37)PE38KDEL, and TP-38 (3–6). Additional fusion proteins have shown antiglioma activity in tissue culture and animal models (7–10). We have tested the diphtheria fusion protein DAB₃₈₉EGF on brain tumor cells in tissue culture and observed dramatic antitumor activity (11, 12). The cytotoxic potency was proportional to EGFR density. Based on these observations, we examine in this paper the expression of growth factor receptors and interleukin receptors on brain tumors and normal brain tissues for the current clinically available fusion proteins on individual patients. The expression levels of these receptors may be surrogate markers for tumor sensitivity. Because physicians and patients have the opportunity to choose among these novel agents for therapy, it is vital to know which agent may be best for a particular patient. Furthermore, physicians and patients need to know the risks to the normal brain tissue with these different agents. The results provide evidence that EGFR should be an excellent target for interstitial fusion protein therapy and suggest that different fusion proteins may be optimal for different patients. Finally, each fusion protein may be associated with distinct toxicity profiles.

Freshly resected tumor samples were obtained on 31 glioblastoma multiforme and 7 anaplastic astrocytomas and snap-frozen and stored in liquid nitrogen. All samples were obtained under Wake Forest protocol (BG02-526). Under the conditions of Institutional Review Board approval, patient identifiers including age, sex, treatment history, or clinical course were not provided with the samples.

With a cryostat, 5-μm sections were cut and thaw-mounted to Superfrost/Plus microscope slides (Fisher, Pittsburgh, PA). Slides were immediately fixed with 3.7% formaldehyde in PBS; washed with PBS; blocked with 10% normal goat serum; reacted with (a) 10 μg/ml MOPC21 IgG1 isotype control antibody (Zymed Laboratories Inc., South San Francisco, CA), (b) 2 μg/ml 31G7 IgG1 mouse monoclonal antibody to EGFR (Zymed Laboratories Inc.), (c) 10 μg/ml IgG1 mouse monoclonal antibody to IL13Rα2 (Diaclone SAS, Besancon, France), (d) 5 μg/ml 454A12 IgG1 mouse monoclonal antibody to the human TfR (Chiron, Emeryville, CA), or (e) 10 μg/ml X2/45-12 IgG1 mouse monoclonal antibody to the human IL4Rα (eBioscience, San Diego, CA) in 1% normal goat serum-PBS overnight at 4°C; rewashed with PBS; reacted for 1 h at room temperature with 15 μg/ml rhodamine-conjugated goat antimouse immunoglobulin (Jackson ImmunoResearch) in 1% normal goat serum-PBS; and rewashed. Slides were then mounted with 90% glycerol/10% PBS and immediately examined under a Zeiss epifluorescence microscope. Adjacent sections were used for all immunostaining, and representative sections were stained with H&E to verify adequate cryopreservation and the presence of readily identified collections of malignant cells in each case. Staining was graded as + with >90% glioma cells stained strongly, +/− with 1–10% of glioma cells stained or weak staining intensity, and − with <1% glioma cells stained or absent staining. A subset of cases (patients 1, 4, 7, 9, 10, 13, 17, 24, 28, 33, and 36) expressing homogenously positive cells (>90%) typical for variations in intensity of labeling with the different antibodies was analyzed by morphometric quantitation using Adobe Photoshop (13).

Strong cell surface reactivity was observed on the high-grade gliomas for monoclonal antibodies to EGFR, IL13Rα2, and TfR. In contrast, cytoplasmic and surface reactivity was seen on a subset of tumor cells with the antibody to IL4Rα (Table 1 and Fig. 1). The frequency of high receptor expression was 32 of 38 (84%) for EGFR, 30 of 38 (79%) for IL13R, 25 of 38 (66%) for TfR, and 17 of 38 (45%) for IL4R. Reactivity of normal brain endothelium was observed for TfR, and reactivity of normal brain astrocytes was observed for IL4R. To validate the interpretation of differences in intensity of labeling, a subset of cases representing positive versus negative examples for each receptor was quantified (Fig. 2). The results confirm the accuracy of the interpretations described in Tables 1 and 2.

EGFR and IL13Rα2 were not always expressed on the same tumors (Table 2). Six tumors (tumors 12, 17, 18, 24, 26 and 34) had strong EGFR but absent or weak IL13Rα2 expression. Conversely, four tumors (tumors 2, 9, 30, and 36) had strong IL13Rα2 but absent EGFR expression. Strong TfR expression was found primarily on EGFR-positive tumors. The only exceptions were tumors 2 and 30. Ten tumors had strong EGFR or IL13Rα2 expression but lacked or had weak TfR expression, including tumors 5, 8, 9, 17, 18, 29, 30, 33, and 37. IL4Rα expression was infrequent, and even when it was expressed, it was generally observed only on a subset of the tumor cells. This receptor was positive only in cases with strong positivity of all of the other three receptors.

Therapy of refractory high-grade gliomas with fusion protein CED has shown promising results. Tf-CRM107, human transferrin chemically coupled to CRM107 receptor binding mutant of diphtheria toxin (S525F), produced two complete remissions and seven partial remissions among 15 patients in a Phase I study (3). Median survival of responding patients was 74 weeks compared with 36 weeks for the nonresponders. Factors that predicted response were the Tf-CRM107 concentration administered, the total dose of Tf-CRM107, and the patient pretreatment tumor volume. In a Phase II study, there were four complete remissions and seven partial remissions among 44 recurrent malignant glioma patients (14). IL4(38-37)PE38KDEL consists of IL4 residues 38–129 followed by a GGNGG linker, amino acid residues 1–37 of IL4, and amino acid residues 253–613 of Pseudomonas exotoxin with deletion of residues 365–380 and modification of residues 609–613 from REDLK to KDEL. This IL4R targeted fusion protein produced one durable (>18 months) complete remission among nine treated patients (4). IL13PE38QQR (composed of IL13 fused to Pseudomonas exotoxin residues 253–613 with deletion of residues 365–380 and replacement of lysines at residues 590, 606, and 613 with glutamine, glutamine, and arginine) was administered by CED to six patients with recurrent supratentorial malignant gliomas, and two remissions were observed (5). TP-38 consists of transforming growth factor α fused to a M_r 38,000 fragment of Pseudomonas exotoxin (PE38; Ref. 6). Eleven patients were treated by CED with 1 or 2 μg of fusion protein. One patient achieved a partial remission, and two patients had stable disease. Whereas the antitumor activity is striking, clearly not every patient responds to each fusion protein.

What factors may predict fusion protein efficacy? The first step in fusion protein cell is binding to cell surface receptors. For ligands with similar receptor binding affinity and similar concentrations in the tumor interstitial space, the tumors with greater receptor homogeneity and density will have more cell surface receptor occupancy and, potentially, sensitivity to fusion protein. The affinity of transferrin, IL4, IL13, and EGF for their receptors is similar, with dissociation constants of approximately 1 nm. Furthermore, in the above-mentioned CED clinical trials, the infusion concentration of fusion proteins was similar, around 0.5 μg/ml. Thus, the major variable for fusion protein binding is the presence and number of receptors/cell. The current study was undertaken to evaluate the available fusion protein receptor homogeneity and heterogeneity of different receptors. The results yield insight into the clinical behavior of each agent in patients with recurrent malignant gliomas and suggest methods to enhance the response rate.

EGFR status of high-grade gliomas has been studied previously by immunohistochemistry, ¹²⁵I-EGF binding, immunoblots, and enzyme-linked immunoassay. The percentage of tumors with high EGFR were 71%, 51%, 69%, 50%, and 50% by frozen section immunohistology (15), paraffin section immunohistology (16), radiolabeled ligand binding (17), immunoblots (18), and enzyme-linked immunoassay (19), respectively. In our study, using only frozen sections, we observed EGFR reactivity in 84% of high-grade gliomas, which compares well with the mean value from the other frozen section immunohistology studies. Thus, a very high percentage of high-grade gliomas display EGFR. Some patients’ tumors also have expression of truncated, constitutively activated EGFR proteins (e.g., EGFRvIII), but, in each case, the tumors also show overexpression of wild-type EGFR, which can bind EGF (20). It is possible that some tumors may only express mutated EGFR, which cannot be detected by the 317G5 monoclonal antibody. However, most of the cases within our study were positive with 317G5, making that possibility less of a concern. Because EGFR binds the EGF ligand with high affinity, and the EGF-EGFR complex internalizes efficiently by receptor-mediated endocytosis, the EGFR is an excellent target for fusion proteins with toxins such as ricin, Pseudomonas exotoxin, and diphtheria toxin (7, 10, 11, 21, 22). IL13Rα2 expression on malignant gliomas has been measured by radiolabeled IL13 binding and immunohistochemistry. Autoradiography studies of ¹²⁵I-labeled IL13 binding to tumor tissue sections showed reactivity with 96% of high-grade gliomas (23). Immunohistochemistry with a monoclonal anti-IL13Rα2 antibody was performed on cultured cells (after mincing and digesting tumor specimens with collagenase and hyaluronidase and incubating them in media). Sixty-five percent of tumor specimens showed reactivity (24). Thus, most tumors have significant amounts of surface IL13R. Lesser but still measurable amounts of IL13R protein were observed on normal human astrocytes, based on tissue culture immunofluorescence and autoradiography (25). The mean percentage of gliomas with IL13Rα2 reactivity (87%) was similar to our finding of 79%. We did not observe antibody reactivity in normal brain. Because the antibody only reacts with IL13Rα2 and does not react with the low level of IL4Rα or IL13Rα1 (400–500/cell) on normal brain, these results were expected. TfR was overexpressed on 83% of high-grade gliomas by immunohistochemistry (26). However, TfR was also abundantly expressed on normal brain capillaries (27). The normal brain TfR was also observed on immunoblots (27). The mean percentage of tumors with TfR overexpression from the previous reports (82%) was similar to our observation of 66%. IL4R consists of a dimer of IL4Rα and IL13Rα1. Anti-IL4Rα antibody immunostaining has been used to show expression of the receptor on glioblastoma multiforme (83%; Ref. 28). However, reactivity was heterogeneous, with only 45% of samples showing >90% positive cells. Reactivity with normal astrocytes has also been observed (29). Our result of 45% glioma samples bearing IL4Rα on >90% of tumor cells matches well with the previously reported 45%. We also found marked heterogeneity with IL4Rα staining.

Coordinate expression of multiple receptors on gliomas has only been previously reported twice (30, 31). Discordant expression was observed in both studies. Our study is the first to use immunohistochemistry to compare receptor expression. We found EGFR overexpression without IL13Rα2 in 6 of 38 (16%) of high-grade gliomas; IL13Rα2 overexpression without EGFR was seen in 4 of 38 (11%) of samples. Thus, the current work confirms discordant receptor overexpression in gliomas.

Several conclusions can be drawn from the current study. First, EGFR is an excellent target for CED fusion protein therapy. Its high frequency of overexpression and lack of expression in normal brain tissue are ideal for this type of treatment. Second, measurement of receptors before therapy may permit optimal selection of fusion proteins. The ease and rapid determination of receptor status by immunohistochemistry make it an attractive methodology for this testing. Third, the heterogeneity with regard to receptor expression suggests that combination treatment with two or more fusion proteins (e.g., DAB₃₈₉EGF and IL13PE38QQR) may yield additive or supra-additive antitumor efficacy.

We thank Raj K. Puri for helpful suggestions regarding the manuscript.