Intragenic Mutation Analysis of the Human Epidermal Growth Factor Receptor (EGFR) Gene in Malignant Human Oral Keratinocytes

INTRODUCTION

EGFR, 4 one of the most extensively studied tyrosine kinase receptors (1, 2, 3, 4, 5) , is believed to play important roles in the control of cell growth and differentiation. Gene amplification and overexpression of EGFR have been reported in various human tumors, including head and neck/oral cancer (6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20) . Overexpression of EGFR has been associated with increased proliferation of oral epithelium (21) . Many of the ligands that bind to EGFR are expressed in malignant tissues. One of its ligands, TGF-α, is expressed in a wide variety of epithelial tumors (22) . The expression of both EGFR and TGF-α within tumors is consistent with an autocrine or paracrine mechanism of growth stimulation (23 , 24) .

Mature EGFR is a 170-kDa transmembrane glycoprotein composed of a single polypeptide chain of 1186 amino acid residues (4) . EGFR has four functional domains: an extracellular ligand-binding domain, a transmembrane domain, the catalytic protein kinase domain, and the COOH-terminal regulatory domain. The extracellular domain can be further divided into subdomains (I-IV), which include two cysteine-rich regions (II and IV) and two regions (I and III) involved in ligand binding (25 , 26) . The 170-kDa human EGFR is encoded by two major transcripts of 5.8- and 10.5-kb (27) . A 2.8-kb transcript produced by alternative splicing encodes only the cytoplasmic tyrosine kinase domain (27) . Similar rearrangement and amplification of the EGFR gene occur frequently in glioblastomas (28 , 29) .

In head and neck cancer, EGFR overexpression and/or amplification is associated with the majority of the cases examined (6 , 8 , 30, 31, 32, 33) . Ishitoya et al. (6) have shown that 53% of 15 SCCs exhibited EGFR overexpression. Using Northern blot analysis, Grandis and Tweardy (34) showed that EGFR is overexpressed in 92% of the tumors examined, based on a study of 24 patients. We have found EGFR overexpression in 8 of 10 cases of surgically resected human oral cancer examined by in situ hybridization (21) . Although overexpression of EGFR is a frequent finding in oral cancer, gene amplification could only account for a fraction of these cases. Thus, additional mechanisms likely contribute. We hypothesize that promoter alterations and intragenic mutations are candidate mechanisms. In this report, we present the findings of an intragenic mutation analysis approach.

We hypothesize that there are specific and unique mutations in the EGFR gene in human oral cancers. Intragenic mutations of EGFR were examined by RT-PCR and a NIRCA using well-characterized normal and malignant human oral keratinocyte cell lines. We report two consistent intragenic EGFR alterations in malignant human oral keratinocytes: a truncated form of EGFR and a tumor-associated polymorphism at position 2073.

MATERIALS AND METHODS

Cell Lines and Specimens.

Two normal, three advanced SCC lines, and A431 cells were used in this study. OKB2 and OKF4 are primary cultures derived from normal oral keratinocytes; SCC15, SCC25, and SCC66 are derived from oral SCCs (35 , 36) . A431 is an epidermoid carcinoma cell line known to exhibit amplification and overexpression of the EGFR gene (27 , 37 , 38) . Four cases of oral cancer and two normal oral mucosa tissues were also analyzed. All human oral cancer biopsies were obtained prior to surgery, radiotherapy, or chemotherapy.

Northern Blot Analysis.

Total RNA isolation and Northern blot analysis were performed on confluent cultures of normal (OKB2 and OKF4) and malignant (SCC15, SCC25, and SCC66) keratinocyte cultures as well as the normal and tumor tissues according to methods described previously (39) . A 1.6-kb human EGFR cDNA (pAW10) was labeled with [³²P]dCTP by random priming.

RT-PCR and Southern Blot Analysis.

Total RNA was isolated from cell lines and tissue specimens. SuperScript II reverse transcriptase (Life Technologies, Gaithersburg, MD) and EGFR gene-specific-primers (5′-TTTTTAGGGCTCATACTATCCTCCGTGG-3′) were used for reverse transcription to synthesize first-strand cDNA. The size of the first-strand cDNA was 3.7-kb, and it contained the entire coding region of the EGFR mRNA. This first-strand cDNA was then used as the template to generate a pool of PCR products containing the entire coding region by the use of primer set 1A/1B. The sequence of 1A is 5′-CCCCTGACTCCGTCCAGTATTG-3′, corresponding to EGFR positions 128–149, and the sequence of 1B is 5′-GCTCATACTATCCTCCGTGGTCATG-3′, corresponding to positions 3815–3839. PCR was performed using an XL PCR kit (Perkin-Elmer Cetus Instruments, Norwalk, CT) to permit the generation of long-distance PCR products. The size of the expected PCR product is 3.7-kb. Southern blotting with a ³²P-labeled human EGFR cDNA (40) was used to determine whether the PCR products were related to EGFR.

Cloning and Sequence Analysis.

PCR products of EGFR were cloned using the CloneAmp system (Life Technologies) followed by DNA sequencing using the dideoxy-chain termination method. The PCR products were purified from agarose gel using the QIAEX II agarose gel extraction system (Qiagen Inc., Chatsworth CA). The products were treated with Uracil/DNA glycosylase to remove uracils creating basic sites in the DNA, destabilizing base pairs, and exposing the single-stranded 3′ termini. The Uracil/DNA glycosylase-modified PCR products were then cloned into the pAMP1 vector.

Western Blotting.

Keratinocytes were lysed directly in radioimmunoprecipitation buffer containing a cocktail of protease inhibitors. Frozen tissues (normal and tumor) were pulverized in dry ice and homogenized immediately in a buffer [1% SDS, 50 mm Tris-HCl (pH 7.6)] containing a cocktail of protease inhibitors. The cell lysate was centrifuged at 1500 × g to remove unbroken cells and large particulates. The supernatant fraction was diluted in loading buffer, and 25μl were used for Western blot analysis. Proteins were fractionated by 10% SDS-PAGE electrophoresis and then electroblotted onto polyvinylidene difluoride membranes. Two antihuman EGFR antibodies were used. A rabbit antihuman EGFR polyclonal antibody raised against a peptide corresponding to amino acids 1005–1016 of EGFR (sc-03; Santa Cruz Biotechnology, Santa Cruz, CA) and a mouse monoclonal IgG_2a against a peptide corresponding to amino acids 351–364 of EGFR (05-104; Upstate Biotechnology, Lake Placid, NY). Signals were detected by chemiluminescence using the ECL kit (Amersham Corp., Arlington Heights, IL).

Detection of Point Mutations Using the NIRCA and Direct DNA Sequencing.

Intragenic point mutations were detected by a NIRCA system (MisMatch Detect II; Ambion Inc., Austin, TX). This system is based on the property of the enzyme RNase that cleaves single unpaired bases (mismatches) of hybridized RNA duplexes produced in vitro by generating sense and antisense transcripts of each of the eight EGFR-coding domains using the specific primer pairs (Fig. 1) ⇓ . Bacteriophage RNA promoter sequences T7 and Sp6 were integrated into the sense and antisense primers to permit the generation of the needed transcripts for the RNA mismatch analyses. Sense and antisense RNA transcripts from normal and SCC lines were hybridized in a reciprocal and complementary experiments. Hybridized RNA duplexes were then treated with RNase A to cleave the duplexes at the mismatch positions. The RNase A-digested products were then analyzed by 2% agarose gel electrophoresis. The primer sets used are shown in Table 1 ⇓ .

RESULTS

Overexpression of EGFR in Human Oral Cancer Cell Lines.

To assess the status of EGFR in the normal and malignant human oral keratinocyte cell lines, the expression of EGFR mRNA was examined by Northern blot analysis. Fig. 2A ⇓ shows that all cell lines expressed detectable levels of EGFR mRNA. OKF4 shows a faint hybridized band. All three malignant oral SCC lines demonstrated overexpression of EGFR mRNA. Western blot was used to detect cellular levels of EGFR protein in the same keratinocyte cultures (Fig. 2B) ⇓ . When a monoclonal antibody against human EGFR (05-104; Upstate Biotechnology, Lake Placid, NY) was used, the 170-kDa EGFR was readily detectable in all keratinocyte lines examined. OKF4 clearly expressed EGFR. Malignant oral keratinocyte cell lines demonstrated higher levels of EGFR protein.

A Novel Truncated EGFR in Human Oral Keratinocytes.

We began the examination of intragenic mutations in human EGFR by generating first-strand cDNA templates from each cell line that encompassed the entire coding region. When the primer set 1A/1B (Fig. 1) ⇓ was used, the RT-PCR products from the normal and tumor keratinocyte cell lines showed only one product, at the expected size of 3.5-kb (Fig. 3A) ⇓ . The amounts of RT-PCR products from the tumor cell lines seemed higher, confirming the data in Fig. 2 ⇓ . To confirm that the observed RT-PCR products were EGFR related, Southern blotting was performed using a human EGFR cDNA probe. Interestingly, in addition to the 3.5-kb, there was an addition band at ∼1.5-kb that was hybridizable in the tumor oral keratinocytes, including A431, suggestive of a truncated form of EGFR (Fig. 3B ⇓ , asterisk).

To examine the nature of the truncated EGFR, the 1.5-kb RT-PCR products from the three SCC cell lines were cloned and sequenced. Sequence analyses revealed that in all three SCC lines, this shortened form of EGFR represented a deletion of ∼2.2 kb of coding sequence from either positions 3263–1072 or 3266–1075. This truncation brought the proline at position 273 to the proline at 1003, resulting in only one of the two prolines being retained. This truncation spans the transmembrane, the juxtamembrane, and the protein kinase domain (Fig. 3C) ⇓ .

To determine whether the truncated EGFR transcript was expressed in the other cells, a PCR primer pair was designed to span the deleted region. Forward primer 533A (5′-CCTATGCCTTAGCAGTCTTATC-3′, positions 533–554) and reverse primer 3414B (5′-CTTGACTGAGGACAGCATAGA-3′, positions 3414–3434) were used (Fig. 3C) ⇓ . Fig. 3D ⇓ shows that this primer set detected the wild-type 2880-bp and the truncated 714-bp transcripts in all of the cell lines examined, including the normal keratinocytes OKB2 and OKF4. The 533A and 3414B primer set was used similarly to amplify RNA isolated from two normal gingival tissues and four cases of freshly resected human oral cancers. Both the wild-type 2880-bp and truncated 714-bp products were detected in all samples examined (Fig. 3E) ⇓ .

To examine whether the truncated EGFR is translated into protein, Western blots were performed on lysates from the normal and tumor oral keratinocyte cell lines using different EGFR antibodies to detect the truncated form. When an antibody preparation that recognizes the sequence 1005–1016 (sc-03; Santa Cruz Biotechnology, Santa Cruz CA), which is slightly adjacent to the COOH-terminal of truncated region was used, a 53-kDa band was detected in all of the cell lines examined, in addition to the 170-kDa wild-type EGFR band (Fig. 3F) ⇓ . It is interesting to note that, unlike the normal or tumor oral keratinocytes, the level of the 53-kDa EGFR was only marginally detectable in the A431 cells. The size of the 53-kDa band is consistent with the size of the truncated form of the EGFR detected in the human oral keratinocytes.

Point Mutation Detection Using NIRCA.

The results of screening for point mutations using NIRCA are summarized in Table 2 ⇓ . The results indicated that all three oral SCC cell lines have a common alteration in the extracellular region 6A-6B (Fig. 1) ⇓ close to the transmembrane domain (Fig. 4) ⇓ . In addition, both SCC25 and SCC66 showed a second alteration in the extracellular region 2A-2B. To determine the precise nature of these alterations, RT-PCR products were cloned, sequenced, and analyzed. The alterations detected in 2A-2B (SCC25 and SCC66) indicate a polymorphisms at position 660 (C/T). Sequence analysis of the 6A-6B region revealed that all clones from OKB2 (three of three clones) and OKF4 (nine of nine clones) showed A at position 2073. All four clones from SCC15 showed a T at this position. Two of four clones from the SCC25 cell line showed a T and the other two clones showed an A, suggesting heterozygosity. Interestingly, one of four SCC66 clones showed T and the other three clones showed A, again suggesting heterozygosity. The published sequence from A431 cells shows a T at this position. This is, therefore, a silent mutation because it does not encode an altered amino acid (both ACT and ACA sequences encode for threonine).

The change from A to T at 2073 created a new restriction enzyme site for BsrI (Fig. 5A) ⇓ . The NIRCA and sequence data suggest that BsrI can cut SCC cDNA at this position (ACTGGN), either partially (SCC25 and SCC66) or completely (SCC15 and A431). BsrI cannot cut this site in normal the cell lines OKB2 and OKF4 (ACAGGN). To confirm this in cells, 6A-6B PCR products were generated from normal and tumor cells and digested with BsrI (Fig. 5B) ⇓ . The 6A-6B region has three BsrI sites at positions (1926, 2133, and 2267). BsrI digestion of wild-type 6A-6B cDNA should, therefore, give rise to four fragments of sizes 65-, 134-, 134-, and 207-bp. The two 134-bp fragments should comigrate on agarose gel. The additional BsrI site created by the A→T base change at 2073 would result in the further cleavage of the 207-bp wild-type fragment into a 150- and a 57-bp fragment. Fig. 5B ⇓ shows the BsrI digestion results of the RT-PCR products from the normal and tumor oral keratinocyte cell lines (including A431), confirming the presence of the predicted products. In normal keratinocytes (OKB2 and OKF4), BsrI digestion resulted in three bands, at 207-, 134-, and 65-bp (Fig. 5B ⇓ , Lanes 1 and 2). BsrI digestion of SCC15 and A431 region 6A-6B RT-PCR products generated four bands, at 150-, 134-, 65-, and 57-bp. Complete disappearance of the 207-bp fragment indicates homozygosity at this site, and appearance of the 150- and 57-bp bands suggests tumor-association. As predicted, BsrI digestion of SCC25 and SCC66 region 6A-6B RT-PCR products resulted in a pattern consistent with heterozygosity at this site.

A preliminary attempt was made to determine whether the 2073 polymorphism could be detected in transformed keratinocytes in vivo. Four cases of freshly resected human oral cancers were used. These were the same four cases used previously for detecting the truncated EGFR mRNA (Fig. 3E) ⇓ . Fig. 5C ⇓ showed that position 2073 is wild type in the two cases of normal gingival tissues (Fig. 5C ⇓ , Lanes 1 and 2). For the four oral cancer specimens, BsrI digestion yielded the 150-bp band in all four tumors. Interestingly, the 207-wild-type fragment seemingly was not present in three of the four tumors (Fig. 5C ⇓ , Lanes 4–6), suggesting the EGFR mRNAs expressed in these tumors are predominantly polymorphic at this site. One of the four tumors (Fig. 5C ⇓ , Lane 3) clearly showed a heterozygous pattern: both 207- and 150-bp bands are present. The 65- and 57-bp bands were only faintly detectable in all of these specimens.

DISCUSSION

In this study, we report two novel findings in the human EGFR coding region. First, there is a novel truncated form of the human EGFR, and second, there is a polymorphism at position 2073 creating a tumor-associated BsrI site. It should be noted that our approach is one that began with the examination of intragenic mutations in oral keratinocyte cell lines and then validation in tumor tissues. In vivo mutations in the EGFR gene may not be retained in cell lines. For example, the most well-studied EGFR mutation (EGFRvIII) is restricted to tumors and is not found in cell lines (41) .

We detected a 1.5-kb novel EGFR transcript from normal and transformed human oral keratinocytes. Sequence analysis of the 1.5-kb EGFR transcript revealed that it lacks the transmembrane-spanning region, the juxtamembrane region, and the protein kinase domain, suggesting that it cannot span the cell membrane and cannot act as a signal transducer. This deletion did not result in frame shifting, truncating the region from proline 702 to proline 1003. A human-specific EGFR antibody raised against amino acid region 1005–1016 detected the predicted 53-kDa truncated EGFR peptide in all normal and tumor oral keratinocyte cell lines examined. The presence of this truncated form of EGFR in normal oral keratinocytes suggests that it may play a role in the regulation of normal growth and differentiation. In addition, its presence in normal and malignant oral keratinocytes suggests it may not contribute to the malignant transformation process. The detection of this 53-kDa truncated EGFR was due to the initial need to create cDNA templates for intragenic mutations analysis. Upon Southern blot hybridization of the amplified 1A/1B PCR products, we observed the 1.5-kb truncated EGFR message (Fig. 3B) ⇓

The region spanning proline 1003 has been associated with EGFR function. Mutation of proline 1003 to glycine has been shown to enhance receptor function (27 , 42) and also binds actin (43) . Several possibilities exist that may eventually account for the mechanism of the detected truncated EGFR protein. For example, the truncated EGFR may modulate the activity of transmembrane receptor. The 170-kDa EGFR binds to members of the EGF family, including EGF, TGF-α, amphiregulin, betacelluin, and heparin-binding EGF-like growth factor. In vitro studies have suggested that soluble c-erbB1 product, which lacks the extracellular EGF-binding domain, can block TGF-α-dependent soft agar colony formation in chicken embryo fibroblast cells (44) . The avian c-erbB1 soluble product is fully capable of binding to human TGF-α in vitro (44) . Thus, protein encoded by the 1.5-kb truncated EGFR may bind to a subset of EGF family of proteins and modulate the activity of the receptor. It should be noted that a soluble EGFR-related inhibitor (also called astrocyte mitogen inhibitor) has been identified in normal rat brain extracts that cross-reacts with anti-EGFR antibodies (43) .

Another potential mechanism is that the truncated EGFR may be able to form inactive heterodimers with the cell surface EGFR and interfere with signal transduction. In A431 cells, soluble truncated EGFR has been shown to inhibit tyrosine kinase activity of the transmembrane receptor (45) . Whereas this truncated EGFR is 100 kDa, our detected truncated receptor was 53 kDa. Structurally, the 100-kDa truncated EGFR is missing the intracellular tyrosine kinase domain, whereas the 53-kDa truncated EGFR predicts a deletion of the tyrosine kinase, transmembrane, and juxtamembrane domains. For the 100-kDa truncated EGFR, inhibition did not appear to result from competition for ligand binding but rather through a direct interaction of the soluble receptor with the full-length 170-kDa EGFR. If this hypothesis is true, then it is interesting to speculate whether the full-length EGFR/truncated EGFR ratio would be useful to determine aggressiveness and malignant potential of the primary tumor.

The use of the NIRCA assay to screen for intragenic mutations in EGFR led to the finding that there is one polymorphism that is consistently associated with tumor oral keratinocytes. The tumor-associated base change at 2073 generated a novel BsrI site. When the RT-PCR products were digested with BsrI, this tumor-associated BsrI site generated two diagnostic fragments of 150 and 57 bp. These findings suggest that the base change at the position 2073 could be a potential diagnostic marker. Preliminary testing using four cases of human oral cancers supported this hypothesis. All four tumors tested yielded the BsrI-digested 150-bp band. It should be noted that the 2073 site is not retained in the truncated form of EGFR. Additional experiments are in progress to further validate this finding in vivo.