Expression of Interleukin-13 Receptor α2 in Glioblastoma Multiforme: Implications for Targeted Therapies

Glioblastoma multiforme is the most common primary malignant brain tumor and despite treatment with surgery, radiation, and chemotherapy, the median survival of patients with glioblastoma multiforme is ∼1 year. Glioblastoma multiforme explants and cell lines have been reported to overexpress the interleukin-13 receptor α2 subunit (IL13Rα2) relative to nonneoplastic brain. Based on this finding, a recombinant cytotoxin composed of IL13 ligand and a truncated form of Pseudomonas aeruginosa exotoxin A (IL13-PE38QQR) was developed for the targeted treatment of glioblastoma multiforme tumors. In a recently completed phase III clinical trial, however, IL13-PE38QQR was found to be no more effective than an existing therapy in prolonging survival. To determine possible explanations for this result, we analyzed the relative expression levels of IL13Rα2 in glioblastoma multiforme and nonneoplastic brain specimens using publicly available oligonucleotide microarray databases, quantitative real-time reverse transcription PCR, and immunohistochemical staining. Increased expression of the IL13Rα2 gene relative to nonneoplastic brain was observed in 36 of 81 (44%) and 8 of 17 (47%) tumor specimens by microarray and quantitative real-time reverse transcription PCR analyses, respectively. Immunohistochemical staining of tumor specimens showed highly variable expression of IL13Rα2 protein both within and across specimens. These data indicate that prescreening of subjects may be of benefit in future trials of IL13Rα2 targeting therapies. [Cancer Res 2007;67(17):7983–6]

A highly sought after treatment for glioblastoma multiforme is one that eradicates neoplastic cells while sparing normal brain tissues. A cell surface antigen differentially overexpressed by glioblastoma multiforme cells could serve as a therapeutic target and the basis for such a treatment. Interleukin 4 (IL4) and interleukin 13 (IL13) are two immune regulatory cytokines that can compete for binding to a heterodimeric cell surface receptor consisting of a 140-kDa IL4 receptor β subunit and a 45-kDa IL13 receptor α1 subunit (1). In contrast to most normal body tissues, high-grade gliomas can also bind IL13 in the presence of excess quantities of IL4, an effect referred to as IL4-independent IL13 binding (2–5). The cell surface expression of a monomeric 42-kDa receptor capable of binding IL13 but not IL4, termed IL13 receptor α2 (IL13Rα2), was used to explain this effect (6). It was also initially reported that significant IL13Rα2 expression, as assessed by Northern blotting, occurs exclusively in the testes and in malignant glioma tissues (7). The putative relative restriction of IL13Rα2 expression to glioma cells was subsequently used as a rationale for the development of several IL13-based treatment strategies including a recombinant cytotoxin composed of IL13 and a truncated form of Pseudomonas aeruginosa exotoxin A (IL13-PE38QQR; ref. 8). IL13-PE38QQR internalized by cells expressing IL13 receptor complexes enzymatically inhibits protein synthesis and causes apoptotic cell death (9). The IC₅₀ of IL13-PE38QQR in IL13 receptor–expressing cells has been reported to be as low as 0.1 ng/mL (8). A phase III clinical trial assessing the efficacy of IL13-PE38QQR administered intratumorally via convection-enhanced delivery has been completed recently, but detailed results have not as yet been published. NeoPharm, the study sponsor, has however released preliminary information indicating that the median survival of patients treated with IL13-PE38QQR was 36.4 weeks, whereas that of patients treated with a slow-release carmustine wafer was 35.3 weeks (10). To identify factors that may have influenced the outcomes of IL13-PE38QQR–treated patients, we analyzed the relative expression levels of IL13Rα2 in glioblastoma multiforme and nonneoplastic brain specimens. Using two independent methods and two independent sets of patient specimens, we determined that the percentage of glioblastoma multiforme tumors with relative overexpression of IL13Rα2 is <50%. The relevance of our findings on the analysis of completed, as well as the design of future, IL13Rα2 targeting clinical trials is discussed.

Microarray analysis. The Oncomine Web site⁴

(11) was used to examine the differential expression of IL13Rα2 in brain tissues. The microarray data of Sun et al. (12) was identified as an appropriate data set for the detailed comparison of IL13Rα2 gene expression in glioblastoma multiforme and nonneoplastic brain specimens and could be downloaded from the National Center for Biotechnology Information Gene Expression Omnibus⁵ as record GDS1962. All statistical analyses were done using the program R-2.4.1.⁶ Examination of the data by covariance-based principal component analysis and Pearson's correlation analysis was used to confirm the suitability of this data set for investigation of IL13Rα2 gene expression. The differential expression of IL13Rα2 between glioblastoma multiforme and nonneoplastic brain specimens was examined by Tukey box plot, XY scatterplot, and density plot.

Quantitative real-time reverse transcription PCR. Seventeen glioblastoma multiforme and six nonneoplastic brain tissue samples, obtained via neurosurgical resection in accordance with the NIH (NIH, Bethesda, MD) institutional review board–approved human tissue collection protocols, were randomly selected for study. Informed consent for tissue collection was obtained from each subject. Total RNA was extracted from snap-frozen tissues using Trizol reagent (Invitrogen). The RNA was treated with DNase I (Invitrogen) to eliminate traces of genomic DNA and cDNAs were synthesized using SuperScript II Reverse Transcriptase (Invitrogen) and random primers (Invitrogen). All procedures were done according to the manufacturer's protocols. The integrity of the cDNA template was verified by standard PCR amplification of the human 18S rRNA product at an annealing temperature of 50°C for 15 cycles. The primer sequences for the 18S rRNA product were 5′-GGAATAATGGAATAGGACC-3′ (sense) and 5′-GCTCCACCAACTAAGAAC-3′ (antisense). Quantitative PCR was carried out using FastStart SYBR Green Master (Roche) in the Prism 7900HT sequence detection system (Applied Biosystems) for 40 cycles. The primers for amplification of IL13Rα2 were 5′-AATGGCTTTCGTTTGCTTGG-3′ (sense) and 5′-ACGCAATCCATATCCTGAAC-3′ (antisense; 13). A cycle threshold value in the linear range of amplification was selected for each sample and normalized for level of 18S rRNA expression. The relative IL13Rα2 expression level of each sample was calculated using the formula 2^ΔΔCT, where ΔΔC_T is the difference between the selected cycle threshold value of a particular sample and the mean of the cycle thresholds of the nonneoplastic brain samples (14). The mean IL13Rα2 expression level of the six nonneoplastic brain samples was assigned an expression value of 1.0 and the fold increase or decrease in IL13Rα2 expression was determined for each control and glioblastoma multiforme sample. Individual glioblastoma multiforme samples were considered to have significantly different expression of IL13Rα2 compared with nonneoplastic brain when the P value for a t test comparing the two was <0.05. Experiments were done in triplicate.

Immunofluorescence histochemistry. Five-micrometer-thick paraffin-embedded sections of one nonneoplastic brain specimen and six glioblastoma multiforme specimens were dewaxed, rehydrated, and subjected to heat antigen retrieval with 10 mmol/L sodium citrate at 95°C for 20 min. The sections were then incubated with goat anti-IL13Rα2 polyclonal antibodies (1:500 dilution; R&D Systems) followed by Alexa Flour 647–conjugated donkey anti-goat antibodies (1:500 dilution; Molecular Probes). Sections were counterstained with 300 nmol/L 4′,6-diamidino-2-phenylindole (DAPI; Sigma) and images were acquired using an LSM 510 confocal laser scanning microscope (Carl Zeiss, Inc.). To show the specificity of the goat anti-IL13Rα2 polyclonal antibodies, 293FT human embryonic kidney cells (Invitrogen) stably transfected with an IL13Rα2 cDNA (Origene) and nontransfected control 293FT cells were stained and imaged as described.

Microarray analysis. To determine the relative gene expression levels of IL13Rα2 in glioblastoma multiforme and nonneoplastic brain samples, we first analyzed a publicly available oligonucleotide microarray data set. Examination of the data set from the study by Sun et al. (12) using covariance-based principal component analysis and Pearson's correlation analysis indicated that this data set was suitable for evaluation of IL13Rα2 expression among glioblastoma multiforme and nonneoplastic brain samples. A Tukey box plot comparing glioblastoma multiforme samples to nonneoplastic brain samples indicates that the ranges of expression of IL13Rα2 among these groups are overlapping and the distribution of expression among glioblastoma multiforme samples is skewed (Fig. 1A). When this data is presented as an XY scatterplot, 36 of 81 (44.4%) glioblastoma multiforme samples exhibit expression of IL13Rα2 greater than two SDs from the expression mean for nonneoplastic brain samples (Fig. 1B). A sample density distribution plot was generated and confirmed that the distribution of expression of IL13Rα2 among glioblastoma multiforme samples and nonneoplastic brain samples is overlapping and mixed (Fig. 1C).

Quantitative real-time reverse transcription PCR. To confirm the oligonucleotide microarray finding that less than half of glioblastoma multiforme samples have overexpression of IL13Rα2, we used both a different set of patient samples and a different method, quantitative real-time reverse transcription PCR (QRT-PCR). QRT-PCR analyses of 17 glioblastoma multiforme and 6 nonneoplastic brain samples were done (Fig. 2). The mean IL13Rα2 expression level of the six nonneoplastic brain specimens was assigned an arbitrary value of 1.0 and the fold increase in expression for each of the glioblastoma multiforme specimens relative to this mean was determined. Individual glioblastoma multiforme samples were considered to have significantly different expression of IL13Rα2 compared with nonneoplastic brain samples when the P value for a t test comparing the two was <0.05. Using this criterion, 8 of 17 (47%) glioblastoma multiforme samples exhibited increased IL13Rα2 expression relative to nonneoplastic brain samples (Fig. 2).

Immunohistochemistry. As a third method for analyzing IL13Rα2 expression in glioblastoma multiforme tumors, we did immunohistochemical staining of glioblastoma multiforme and nonneoplastic brain paraffin-embedded sections. The relative specificity of the anti-IL13Rα2 antibodies used was first confirmed by staining control and IL13Rα2 cDNA–transfected human embryonic kidney–derived cell lines (Supplementary Fig. S1A and B). Immunohistochemical staining revealed no detectable IL13Rα2 protein expression in nonneoplastic brain (Fig. 3A). Specific membrane staining was observed in four of six glioblastoma multiforme specimens and the distribution of expression, when present, was highly variable from region to region within a single specimen. Two of the specimens that stained positively showed expression of IL13Rα2 throughout the section. The other two specimens showed mixed expression, with areas of high expression and clusters of neoplastic cells that were negative for IL13Rα2 expression. Photomicrographs of representative sections showing high, mixed, and negative expression of IL13Rα2 protein are shown (Fig. 3B–D). These qualitative results confirm the quantitative results obtained using oligonucleotide microarray and QRT-PCR and also show the intratumoral heterogeneity of IL13Rα2 expression.

In principle, cytotoxins that recognize and bind specific cell surface proteins carry great promise for the treatment of tumors. The success of such treatments, however, is dependent on the exclusive or enriched expression of the targeted protein on tumor or tumor-dependent cells. Previous reports have indicated that IL13Rα2 is expressed by a vast majority of glioblastoma multiforme explants. Autoradiography of glioblastoma multiforme specimens indicated that nearly all tumors possess IL4-independent IL13 binding sites (2–4), whereas nonquantitative RT-PCR analysis indicated that 14 of 17 (82%) tumors show expression of IL13Rα2 (15). Similarly, it has been reported that 11 of 11 glioblastoma multiforme specimens overexpress IL13Rα2 as determined by in situ hybridization and immunohistochemical staining (16). In contrast to these results, our analysis of publicly available oligonucleotide microarray data sets suggests that the percentage of glioblastoma multiforme tumors that overexpress IL13Rα2 is 44% (36 of 81 specimens). In concordance with this latter result, the percentage of glioblastoma multiforme tumors that overexpress IL13Rα2 as determined by QRT-PCR analysis is 47% (8 of 17 specimens).

The results presented here about the relative expression levels of IL13Rα2 in glioblastoma multiforme specimens compared with nonneoplastic brain tissue differ from those of previously published studies (15, 16). The discrepancies in the results may in part be accounted for by differences in the materials and methods used. Joshi et al. (15) used primary cell cultures derived from glioblastoma multiforme and nonneoplastic brain explants, whereas we and Sun et al. (12) used actual tissue specimens. Previous studies have shown that the gene expression signatures of tumor tissues can differ significantly from that of primary cultures derived from those tissues (17). With regard to differences in methods, we used oligonucleotide microarrays and quantitative RT-PCR rather than nonquantitative RT-PCR (15) and in situ hybridization (16), which is also nonquantitative.

Although the results of a recent phase III trial assessing the efficacy of IL13-PE38QQR have yet to be published, a preliminary information release indicates that it is no more effective than existing therapies. This is not unexpected given our finding that less than half of tumors overexpress IL13Rα2. Additional factors that may have contributed to the lack of efficacy of IL13-PE38QQR are suboptimal drug infusion technique in some patients (18) and the high variability of IL13Rα2 expression from region to region within some tumors. Although the intratumoral heterogeneity of expression could be an experimental artifact due to inherent variations in the fixation and immunohistochemical processing of clinical tissues or to expression levels of IL13Rα2 below the detection limit of the antibodies used, this heterogeneity has been reported by others as well (16). This is of potential clinical significance as the effect of IL13-PE38QQR, the induction of apoptosis, is reported to be highly specific to IL13Rα2-expressing cells (9). The IL13Rα2-expressing fraction of cells within an IL13Rα2-expressing tumor may therefore respond to IL13-PE38QQR, but the IL13Rα2-negative fraction will be unresponsive and continue to proliferate. This could serve to explain any partial or temporary responses to IL13-PE38QQR that may have occurred. Also of interest is the incidence and nature of any non–central nervous system toxicities that may have occurred during the trial. Although IL13Rα2 has been described as a cancer/testis antigen (7), its reported expression in the kidney, spleen, liver, lung, thymus, respiratory epithelium, and monocytes (1, 19) indicates that it is more likely a tumor-associated or tumor-overexpressed antigen.

Rather than dismiss the further testing and potential use of IL13-PE38QQR, we recommend a retrospective analysis of all available patient samples for the expression of IL13Rα2 and a stratification of patients by degree and or pattern of tumor expression. The data obtained from such retrospective analyses, although problematic for determining the success of already completed trials, could be used in the improved planning of future trials. At a minimum, demonstration of tumor IL13Rα2 expression should be an eligibility criterion for any such trials. In this way, patients with a higher likelihood of benefiting from the therapy can be considered for enrollment, whereas those with a lower likelihood can be encouraged to pursue other options.

Grant support: Intramural Research Program of the NIH, National Institute of Neurological Disorders and Stroke.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank A. Sedlock for providing technical assistance and G. Park for critical review of the manuscript.