Several types of epidermal growth factor receptor (EGFR)gene mutations have been reported in glioblastomas, and in nearly all cases the alterations have been reported in tumors with EGFR amplification. The objectives of this study were to determine the frequency and diversity of EGFR mutations in glioblastomas and to determine whether gene mutation is inevitably associated with increased EGFR gene dosage. To accomplish these aims, we sequenced cDNA products representing the entire EGFR coding region in 44 glioblastomas, half of which had EGFR amplification. Coding sequence alterations were identified in 17 of the tumors, and each of these cases had amplified EGFR. No mutations were identified in the 22 tumors without EGFR amplification. An additional 26 glioblastomas with EGFR amplification were then examined to establish more reliable frequencies for each type of mutation identified in the tumors for which the entire gene was sequenced. Transcripts associated with the most common mutation lacked coding sequence for amino acids 6–273 (67%). This mutation has been described extensively in the literature. Transcripts encoding receptors that would truncate at amino acid 958 and transcripts encoding receptors that would lack amino acids 521–603 were the next most common types of alteration. Each of these were observed in 15% of the tumors with EGFR amplification. Other mutations were observed at lower frequencies, but among these were three cases with missense mutations. Sixteen of the 48 tumors with EGFRamplification showed multiple types of EGFR mutations(33%), and in one case it was determined that multiple alterations had occurred in the same transcript. In total, these data are consistent with EGFR mutation being exclusively and frequently associated with EGFR amplification. Furthermore, the determination of multiple EGFR mutations within individual tumors suggests that glioblastomas with EGFRamplification have the capacity to produce a variety of functionally distinct EGFRs.
The involvement of elevated and/or aberrant EGFR3activity in human cancer is well established, especially in the central nervous system tumor glioblastoma, some 40% of which show EGFR gene amplification (1, 2, 3). Initially, it was thought that EGFR amplification promoted tumor development solely by increasing mitogen-activated, growth-stimulatory signaling by wild-type EGFR. Such an interpretation has been shown to be incomplete, however, because it is now known that many glioblastomas with EGFR amplification have EGFR mutations(4, 5). Consequently, the types of EGFR that exist in individual tumors, as well as their corresponding roles in contributing to the highly proliferative nature of glioblastoma, are equivocal.
The majority and possibly all of the EGFR mutations and EGFR alterations that have been described in the literature were initially identified through the use of Southern blot or Western blot analysis,respectively (4, 5, 6, 7, 8, 9). Both of these methods, however, have a limited ability to recognize subtle mutations that do not result in significant changes to restriction fragment patterns or protein molecular weights. The potential for subtle alterations occurring in EGFR is supported by the discovery of missense mutations in several receptor tyrosine kinase genes, including those that encode Ret(10, 11), c-Kit (12, 13), Met (14, 15), and fibroblast growth factor receptors 2 and 3 (16, 17). Another matter of interest that involves the alteration of EGFR in glioblastoma is whether mutations of this cellular oncogene can occur independently of its amplification.
As a result of these concerns, we have conducted an analysis of EGFR mutations in glioblastoma by sequencing cDNAs that represent the entire EGFR coding region for each member of a series of tumors containing equal numbers of cases with and without EGFR amplification. In addition to revealing that EGFR mutations are limited to tumors with EGFRamplification and include single nucleotide substitutions, the data collected from this investigation show that multiple types of EGFR mutations can be detected in individual tumors. The concept involving the ability of a tumor to produce multiple mutant forms of a specific protein product is not unprecedented but appears to occur frequently for EGFR in glioblastoma.
MATERIALS AND METHODS
Tumor Specimens and Nucleic Acid Extractions.
All tumors in this study were obtained from patients undergoing surgical treatment at the Mayo Clinic and were classified as glioblastomas based upon WHO criteria. DNAs from frozen tumor tissue and corresponding peripheral blood leukocytes were isolated and purified as described previously (18). RNAs were extracted from ∼2-mm square sections of frozen tumor tissue with Trizol (Life Technologies, Inc., Grand Island, NY).
EGFR Amplification Detection.
DNAs were digested with HindIII restriction enzyme,electrophoresed through 0.8% agarose gels, and blot-transferred to reinforced nitrocellulose membranes. After baking for 2 h in a vacuum oven at 80°C, membranes were hybridized with one or more EGFR genomic probes described previously (19, 20). After washing, filters were exposed to X-ray film for 6–24 h. Filters were then stripped of EGFR probe and rehybridized with a probe from a syntenic, nonamplified locus (21).
cDNA Synthesis and Sequence Analysis.
One-μg samples of total RNA were reverse transcribed at 37°C for 1 h in 20-μl reaction volumes containing random hexamer primers,Moloney murine leukemia virus reverse transcriptase, and buffer supplied by the manufacturer (Life Technologies). After a 2-min, 95°C heat denaturation step, cDNA amplifications were performed in 50-μl reaction volumes containing 1 μl of product from the reverse transcription reaction, 20 pmol of forward and reverse primers, 200μm deoxynucleotide triphosphates (Perkin-Elmer, Foster City,CA), 1.25 units of Taq polymerase (AmpliTaq Gold; Perkin-Elmer), 10 mm Tris-HCl (pH 8.3), 50 mm KCl, 1.5 mm MgCl2, and 0.001% gelatin. After sample denaturation at 95°C for 9 min, PCR amplifications were performed for 43 cycles at 95°C for 30 s, 55°C for 30 s,and 72°C for 1 min (final extension at 72°C for 10 min). The sequence identities for the primers used for cDNA synthesis, according to GenBank accession number X00588, are: S1, 128–149; AS1, 3839–3815;S2, 95–115; AS2, 808–790; S3, 131–150; AS3, 1356–1334; S4,708–728; S4, 1356–1334; S5, 1187–1208; AS5, 1845–1826; S6,1670–1689; AS6, 2194–2175; S7, 2106–2125; AS7, 2637–2618; S8,2460–2479; AS8, 3542–3515; S9, 2867–2894; AS9, 3832–3813; S10,3111–3130; and AS10, 2637–2618 (see Fig. 1). Preliminary analysis and fractionation of PCR reaction products was by agarose gel electrophoresis. Each distinct cDNA product that could be visualized in the gel was excised and purified (gel extraction kit;Qiagen) or incubated with 40 units of exonuclease I and 8 units of shrimp alkaline phosphatase (PCR Product Pre-Sequencing kit; Amersham)at 37°C for 15 min and then incubated at 80°C for 15 min. Approximately 100-ng quantities of purified or treated PCR product were sequenced using 2 pmol primer and reagents from Amersham’s Thermosequenase kit. Cycling parameters were 20 s at 95°C,30 s at 58°C, and 1 min at 72°C for 30 cycles. Stop solution was added to each reaction, and these were subsequently denatured at 95°C for 2 min, quenched on ice, and electrophoresed through a 6%polyacrylamide gel with 15% formamide and 7 m urea. Gels were dried for 1 h and exposed to film overnight. Preparatory PCR and sequencing conditions for the analysis of amplified genomic DNAs were the same as described for cDNAs (unless as specified below for long distance PCR).
Genomic DNA from tumors with transcripts showing deletion of coding sequence for amino acids 521–603 were amplified in 50-μl reaction volumes using long distance-PCR. These reactions included ∼100 ng of tumor DNA, 1× XL buffer II (GeneAmp XL PCR kit; Perkin-Elmer, Foster City, CA), 200 μm deoxynucleotide triphosphates, 20 pmol each of upstream (EGFR exon 13, bases 1770–1800) and downstream (EGFR exon 16, bases 2104–2076) primer, 1 mm Mg(OAc)2, and 1 unit of rTth polymerase. Reaction profiles consisted of a 1-min sample denaturation at 93°C, followed by 35 cycles of 30 s denaturation at 93°C and 13 min annealing/extension at 68°C, followed by a 10 min extension at 72°C. Reaction products were electrophoresed through 0.8% gels and stained with ethidium bromide.
Forty-four tumors, characterized previously for the presence or absence of EGFR amplification (22), were selected for detailed analysis of EGFR coding sequence variations. For this analysis, overlapping sets of cDNAs spanning the entire EGFR coding region were generated from corresponding tissue RNA (locations of primer pairs used in the synthesis reactions are shown in Fig. 1). Products from amplification reactions were electrophoresed through agarose gels and isolated for sequence analysis. Sequencing of the cDNA fragments revealed several alterations, the corresponding protein structures for which are summarized in Fig. 1. A total of 25 alterations were identified in 17 of the tumors, and all but two of these involved the deletion of large portions of coding sequence. For these cases, there was one instance in which exons 18–26 (codons 664-1030) were duplicated, and in two instances, missense mutations were detected, both within the second cysteine-rich region of the extracellular domain at codons 572 and 604(Figs. 1 and 2). Several polymorphisms were also identified, as indicated by the presence of sequence variants in patient constitutional DNA (Table 1). In total, 17 of the 44 tumors were determined to have one or more EGFR mutations, and all of these were identified among the 22 cases with EGFR amplification; none of 22 cases without increased EGFR gene dosage showed any tumor-specific sequence abnormalities.
To obtain more accurate estimates of the frequency of occurrence for each type of alteration identified in the initial group of tumors, RNAs from an additional 26 glioblastomas with EGFR amplification were examined. The most frequent type of transcript alteration identified among the total of 48 tumors with EGFRamplification was the loss of coding sequence for amino acids 6–273(67%; Table 2). Alterations resulting in the loss of this sequence were most often observed in the absence of any other mutation; in 18 of 32 cases with the Δ6–273 alteration (56%), there was no other mutation evident. Transcripts encoding receptors that would truncate at amino acid 958 and transcripts encoding receptors lacking amino acids 521–603 were each detected in 15% of the tumors but were generally observed in tumors also having the Δ6–273 alteration. To examine the gene alterations responsible for the Δ521–603 alteration, long distance-PCR was performed on corresponding tumor DNA samples, and this analysis revealed truncated genomic fragments lacking exons 14 and 15[as defined by Callaghan et al. (23)] in each instance (examples shown in Fig. 3). Genomic rearrangements associated with the Δ6–273 and C-958 receptors have been described previously (19, 20).
One additional case of coding sequence duplication-insertion was discovered, and this involved exons 18–25 (codons 664-1014) as opposed to the 18–26 duplication-insertion that was identified in the initial group of tumors. For the two cases with transcripts having duplicated coding sequence, it was possible to amplify corresponding, aberrant genomic DNA fragments (Fig. 4). Finally, an additional missense mutation within the second cysteine-rich domain of the receptor was identified at codon 574. In total, EGFR mutations were determined in 36 of the 48 cases with EGFR amplification (75%; Table 2).
Of the 36 tumors having EGFR mutations, 16 showed multiple transcript alterations (one-third of all tumors with EGFRamplification), and the loss of coding sequence for amino acids 6–273 was involved in the majority of such cases. In fact, most of the other types of mutations occurred in tumors with the Δ6–273 alteration,but this relationship was not statistically significant(P = 0.206 for tumors with EGFRamplification, χ2 test).
Although these data indicate the presence of multiple EGFRmutations in individual tumors, they do not address whether multiple mutations occur in individual EGFR genes. Because the distance between each of the three EGFR rearrangement regions (Δ6–273, Δ521–603, and C-958 or Δ959–1030) precludes such a determination by Southern analysis, we used RT-PCR, with primers that flanked combinations of deleted regions, to address this issue in an indirect manner. Although the majority of the tumor RNAs were not of sufficient quality to permit synthesis of the larger cDNAs bordered by the primers, four cases yielded RT-PCR products having only single coding sequence deletions (data not shown). In a fifth case, however, a cDNA was produced that lacked coding sequence for amino acids 521–603 and also showed the deletion mutation resulting in the C-958 truncation(tumor 23; protein model shown in Fig. 1). Furthermore, the point mutation in tumor 26 was identified in cDNAs having the Δ6–273 alteration.
Although several studies have reported the identification of EGFR gene rearrangements and/or EGFR protein alterations in glioblastomas (4, 5, 6, 7, 8, 9), nothing has been reported that involves an analysis of the entire coding region of EGFR in any tumor or group of tumors. Here we have conducted such an investigation because, as noted previously, determination of reported alterations have been based on initial detection by Southern or Western blot analysis, and because it seems likely that sequence changes not having significant effects on DNA restriction enzyme fragment patterns or protein molecular weights could easily be overlooked. Furthermore, EGFR mutations have been shown to occur in tumors that harbor and express wild-type genes and proteins, the presence of which can obscure detection of corresponding mutants.
The majority of mutations detected in the current study are the result of intragene deletion rearrangements and have been reported previously (4, 5, 19). The protein associated with the most common of these, Δ6–273, has been extensively studied and found to display constitutive, ligand-independent kinase activity(24, 25, 26, 27). Recently, we have examined cells expressing C-958 EGFR and found this mutant receptor to display increased,ligand-dependent kinase activity,4similar to that previously described for recombinant COOH-terminal EGFR truncation mutants (28, 29). The Δ521–603 mutant has only been described in two glioblastoma xenografts (4, 7)and has not been examined previously in a series of primary tumors for the purpose of determining its incidence. Recently, an in-frame, tandem duplication of EGFR exons 18–26 was identified in a glioblastoma cell line (30), and this prompted our analysis of the 70 glioblastomas examined here to determine the incidence of this type of alteration in primary tumors. One such alteration was identified, as was an in-frame tandem duplication of exons 18–25 (Fig. 4). Functional consequences associated with the tandem duplication of exons 18–25 or 18–26, which encode the receptor’s tyrosine kinase domain and portions of its internalization domain, as well as the consequences associated with the loss of amino acids 521–603, have yet to be determined (30, 31). One would expect, however, for these mutations to confer a selective growth advantage, as appears to be the case for the Δ6–273 and C-958 mutations (24, 25, 26, 27, 32).4
In addition to the more common mutations, three cases were identified in which there was an in-frame deletion mutation that eliminates coding sequence for the internalization domain, and one tumor contained transcripts encoding a receptor with an extracellular domain truncation of amino acids 6–185 (Fig. 1). Perhaps the most unique finding in this study, however, was the identification of three missense mutations in tumors with EGFR amplification. In one of these tumors,there was no other identifiable EGFR alteration. Each of these mutations occurred within the second cysteine-rich region of the extracellular domain, and two were within the Δ521–603 deletion region. Missense mutations of the extracellular juxtamembrane region have also been reported for RET, FGFR2, and FGFR3and have been shown to result in receptor activation (16, 17, 33). Functional analysis of the EGFR missense mutations should prove interesting.
In addition to the identification of missense mutations in EGFR, it was also determined that a significant proportion of glioblastomas with EGFR amplification show evidence of multiple gene alterations. In two of these cases (nos. 23 and 26), it was apparent that multiple mutations had occurred, not only within the same tumor but within the same gene, because we were able to amplify cDNAs containing two alterations.
It is worth emphasizing that all mutations were detected in tumors with EGFR amplification, suggesting that amplification precedes EGFR mutation. The frequent combination of amplification and mutation (including intragene rearrangements) appears to be unique to EGFR in glioblastoma and suggests a currently undefined mechanism that promotes each type of gene alteration. It would be of significance to determine whether this is attributable to increased genomic instability in certain of these tumors or whether their occurrence together simply indicates a 2-step selective process: the first for increased EGFR signaling (amplification) and the second for signaling deregulation (mutation). A combination of these factors is certainly a possibility as well.
In total, these data suggest that glioblastomas with EGFRamplification have the capacity to overexpress multiple aberrant forms of EGFR, each of which would be expected to have distinct properties. Glioblastomas are known to display substantial heterogeneity at both phenotypic and molecular levels, and the finding of multiple forms of mutant EGFR in these tumors is consistent with this concept. It seems likely that the ability to generate multiple, functionally distinct forms of an oncoprotein would contribute to the well-documented ability of this cancer to evade multimodal therapeutic regimens.
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.
Study supported by Grant CA55728 from the National Cancer Institute (to C. D. J.).
The abbreviations used are: EGFR, epidermal growth factor receptor; RT-PCR, reverse transcription-PCR.
X-Y. Wang, A. Blahnik, and C. D. James,unpublished data.