Genetic Heterogeneity of the Epidermal Growth Factor Receptor in Non–Small Cell Lung Cancer Cell Lines Revealed by a Rapid and Sensitive Detection System, the Peptide Nucleic Acid-Locked Nucleic Acid PCR Clamp

Lung cancer is one of the leading causes of cancer deaths worldwide (1). Although advanced lung cancers are treated by the combination of chemotherapy and radiotherapy (2), the outcome is still not satisfactory (3). Gefitinib is an inhibitor of tyrosine kinase activity of the epidermal growth factor receptor (EGFR) and has recently been introduced in the treatment of advanced lung cancers (4). Gefitinib is dramatically effective in some patients, whereas it is completely ineffective in others. In addition, Japanese show a high response rate (5). Factors that govern the responsiveness to gefitinib are under intense investigation. EGFR mutations in cancer cells are the first such factors reported (6–8). EGFR mutations are found in a portion of lung cancers, mostly those with adenocarcinoma histology. These mutations are activating mutations that enhance the tyrosine kinase activity of EGFR, which is effectively suppressed by gefitinib. The mutations are much more frequent in Japanese than in U.S. patients, consistent with the difference in the response rate (6, 9). The association of the presence of EGFR mutations and the responsiveness to gefitinib is so tight that the former may be used as a predictor of the latter.

Clinical specimens used to diagnose lung cancers (i.e., sputum, pleural effusion, bronchial washing, and surgically resected tissue) contain many normal cells. A method capable of detecting EGFR mutations in a large background of wild-type EGFR genes could provide a superior clinical test. To serve this purpose, we have developed a method able to detect all 11 different EGFR mutations in the initially reported (6, 7) in the presence of 100- to 1,000-fold wild-type EGFR background. These 11 mutations account for >95% of EGFR mutations found in Japan (9). As an application of the method and to investigate its performance, we screened non–small cell lung cancer (NSCLC) cell lines established from Japanese patients and found several cell lines that have EGFR mutations. Unexpectedly, we found that many cell lines have subpopulations that harbor specific EGFR mutations.

Plasmid. Wild-type fragments that contain exons 18, 19, or 21 of the EGFR gene (UniGene Cluster Hs.77432) were amplified by PCR from normal human genomic DNA. Primers used were F18, 5′-GGTAGCTGTTCAGTTAAAGAACACC-3′ and B18, 5′-CCTTTGGTCTGTGAATTGGTC-3′ for exon 18; F19, 5′-CTGGATGAAATGATCCACACG-3′ and B19, 5′-TGGGTAGATGCCAGTAATTGC-3′ for exon 19; and F21, 5′-CTGGATGGAGAAAAGTTAATGGTC-3′ and B21, 5′-CAGCAAGTACCGTTCCCAAAG-3′ for exon 21. The amplified fragments were cloned into pCR-Script vector (Stratagene, La Jolla CA). DNA fragments harboring individual EGFR mutations were made from the cloned fragments by site-directed mutagenesis using the Megaprimer method (10), and the fragments were then inserted into pCR-Script. Linear wild-type and mutant DNA fragments about 500 bp long were amplified from individual plasmids by PCR and used.

Cell lines and genomic DNA isolation. Human NSCLC cell lines which were established from Japanese patients with well-documented histologies were collected. 11-18, 1-87, LCSC#1, LCSC#2, and LK87 (adenocarcinomas); 86-2 (large cell carcinoma); Lu99 (giant cell carcinoma); EBc-1, LK2, LK79, Sq-1, Sq5, and Sq19 (squamous cell carcinomas); and A431 (squamous cell carcinoma of the skin) were obtained from the Cell Resource Center for Biomedical Research, Tohoku University, Japan. ABC-1, PC-3, RERF-LC-Ad1, RERF-LC-Ad2, and RERF-LC-MS (adenocarcinomas); RERF-LC Sq-1 (squamous cell carcinoma); and Lu65 (giant cell carcinoma) were from the Japanese Collection of Research Bioresources (Tokyo, Japan). LC2/ad, PC-14, RERF-LC-KJ, and RERF-LC-OK (adenocarcinomas) and RERF-LC-A1 (squamous cell carcinoma) were from the Riken Bioresource Center (Tsukuba, Japan). PC-7, PC-9, and PC-13 (adenocarcinomas) were from IBL (Takasaki, Japan). KTA-7 (adenocarcinoma) and KTSq-1 were kindly provided by Drs. Toru Kameya and Shi-Xu Jiang (Kitasato University, Japan). Genomic DNA was prepared by QIAamp DNA Blood Mini Kit (Qiagen, Venlo, The Netherlands) from the cells in the first passage of expansion after being obtained from the establisher (KTA-7 and KTSq-1) or from the cell banks (all other cells). Normal human genomic DNA was purified in the same way from the peripheral blood of normal volunteers.

PCR primers and solutions. All PCR reaction solutions (25 μL) were based on the Basic Mixture containing 25 mmol/L TAPS (pH 9.3), 50 mmol/L KCl, 2 mmol/L MgCl₂, 1 mmol/L 2-mercaptoethanol, 200 μmol/L each of deoxynucleotide triphosphates, and 1.25 units of Takara Ex Taq HS (Takara Bio, Shiga, Japan). For conventional PCR, PCR primers (200 nmol/L each) were added to the Basic Mixture. For peptide nucleic acid-locked nucleic acid (PNA-LNA) PCR clamp, PCR primers (200 nmol/L each), fluorogenic probes (100 nmol/L each), and a PNA clamp primer (5 μmol/L) were added to the Basic Mixture. PCR primers were designed manually or by using Primer 3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3.cgi) so that the T_m values were between 55°C and 60°C. Fluorogenic probes containing LNA were manually designed and confirmed by the LNA T_m prediction tool (http://lna-tm.com/) to have T_m values between 54°C and 56°C. PNA clamp primers, 14- to 18-mer in length, were designed according to the guidelines (11). LNA-containing oligos were synthesized by IDT (Coralville, IA) and PNA oligos were by Greiner Japan (Tokyo, Japan).

PCR reactions. PCR and the real-time amplification monitoring for both conventional and the PNA-LNA PCR clamp were done using Smart Cycler II (Cepheid, Sunnyvale, CA). PCR cycling was a 30-second hold at 95°C followed by 45 cycles of 95°C for 3 seconds and 62°C (exons 18 and 19) or 56°C (exon 21) for 30 seconds. Nested PCR for the PNA-LNA PCR clamp was done using the same reaction conditions except that the inner primers and 1 μL of a 1:10⁶ dilution of the first PCR reaction were used.

Nucleotide sequencing. PCR products from both conventional PCR and the PNA-LNA PCR clamp were purified by Wizard PCR Prep DNA purification kit (Promega, Madison, WI) and directly sequenced by an automated DNA sequencer.

Establishment of peptide nucleic acid-locked nucleic acid PCR clamp. The PNA-LNA PCR clamp is a method that can quickly detect specific mutation or deletions occur at known positions. In the PNA-LNA PCR clamp, PNA (12) and LNA (11) are used to construct PCR clamp reactions (13). Positive signals are detected by the 5′ nuclease assay (14). The PNA-LNA PCR clamp system is schematically presented in Fig. 1. Here, PNA clamp primers suppress amplification of the wild-type sequences, thereby enhance preferential amplification of the mutant sequences. LNA probes were employed to specifically detect mutant sequences in the presence of wild-type sequences. Because PNA clamp primers that have wild-type sequences and LNA probes that have mutant sequences are located in the position, PNA clamp primers competitively inhibit mutant LNA probes to bind to the wild type, further increasing the specificity of detection. Thus, individual EGFR mutations can be detected in the presence of a 100- to 1,000-fold wild-type EGFR background molecules (Fig. 2A and B). We multiplexed the reactions by using multiple probes labeled with different dyes to detect 11 mutations by five reactions (Table 1). Even after being multiplexed, each mutation was detected in the presence of 100- to 1,000-fold background (data not shown). The mutant/wild type ratio of a given sample can be semiquantified by plotting the second derivatives of the amplification curves (Fig. 2C). Samples with signals close to the baseline are further resolved by optional nested PCR reactions (Fig. 2D).

Screening of the non–small cell lung cancer cell lines established from Japanese patients. As an application of the method, we screened NSCLC cell lines established from Japanese patients. It has been reported that the rate of EGFR mutations shows apparent ethnic differences and mutations are frequently found in lung cancer specimens from Japanese patients (6, 9). Therefore, we considered that mutations would also be frequent in the cell lines. We found six different mutations in 11 of the 30 cell lines examined (Table 2, “mutation present in most of the cells”). In addition, we unexpectedly found many cell lines presented amplification curves that indicated a subpopulation(s) of the cells harbored EGFR mutation(s).

We first confirmed the presence of the subpopulations by nucleotide sequencing. Direct sequencing of the conventional PCR product revealed a wild-type sequence, whereas that of the PNA-LNA PCR clamp product revealed a mutant sequence (Fig. 3A). The mutant sequence is considered to be derived from the sample DNA, because the PNA-LNA PCR does not have any reaction steps that artificially produce specific mutation fragments. Consistently, nucleotide sequencing of both the conventional and the PNA-LNA PCR products from 17 normal DNA samples presented only the wild-type sequences (data not shown).

We next semiquantified the size of the subpopulations by two methods. First, the PCR cycle number that gives the peak of the second derivative of the amplification curve was compared with that of standards (Fig. 3B,, left). Second, a shift in the amplification curve was observed after incrementally adding mutant DNA fragments as an internal positive control. Here, adding mutant DNA fragments at a copy number less than or close to that of the mutant EGFR gene do not affect the curve while adding more shifts the curve to the left and upwards (Fig. 3B,, right). For all cell lines, these two methods gave consistent results that are summarized in Table 2 [mutation present in a subpopulation(s) of the cells]. Results suggesting a subpopulation size of ≤0.1% were removed from the table because 0.1% is close to the detection limit of the first round of the PNA-LNA PCR clamp; therefore, the result may be erroneous. Taken together, 19 of the 30 cell lines had an EGFR mutation(s) either in their entire population (11 cell lines) and/or in their subpopulation(s) (12 cell lines; Table 2).

The EGFR gene is amplified in some types of cancers such as glioblastomas (15). Our hypothesis that subpopulations of EGFR mutants exist may be disputed by the possibility that the EGFR gene is amplified in the cell lines and only a single copy within the amplified EGFR gene is mutated, resulting in what seems subpopulations. To disprove this argument, we semiquantified the copy number of the EGFR gene in a cell using two different methods. First, individual exons are amplified by the conventional PCR (i.e., a quantitative PCR) and the PCR cycle number that gave the peak of the second derivative of the amplification curve was compared with that of standards (Fig. 3C,, left). Second, a shift in the amplification curve was observed after incrementally adding wild-type DNA fragments as an internal control (Fig. 3C , right). Both methods gave consistent results for all cell lines and showed that the copy number of the EGFR gene in all the cell lines with EGFR mutations in the subpopulation(s) is close to 2, indicating that the EGFR gene is not amplified.

In this study, we have established a rapid and sensitive method, the PNA-LNA PCR clamp, for detecting EGFR mutations. The PNA-DNA heteroduplex with a completely matched sequence has a higher melting temperature (T_m), whereas that with a single base mismatch has a lower T_m than the corresponding DNA-DNA duplex (16). In addition, PNA is resistant to the 5′ nuclease activity of Taq DNA polymerase. These characteristics make a PNA oligo a superior clamp primer for inhibiting PCR amplification of wild-type sequences. The LNA-DNA heteroduplex with a completely matched sequence has a higher T_m, whereas that with a single base mismatch has a lower T_m than the corresponding DNA-DNA duplex (17). LNA can be mixed with other nucleotides to synthesize fluorogenic probes for detecting mutant sequences. The amount of DNA used for one reaction is 25 ng (i.e., the amount of DNA from 8,000 haploid genome). Therefore, the resolution of one copy of mutant in 100 to 1,000 copies of wild type is considered sufficient for many of clinical and experimental applications. This method is applicable to other mutations where known point mutations or small deletions at fixed positions need to be detected rapidly and at a high sensitivity. An example is the detection of K-ras mutations, in which we successfully detected in the presence of 100- to 1,000-fold background of the wild-type gene. We found many PCR primers and fluorogenic probes that have T_m described in Materials and Methods worked fine. For designing PNA primers A + G contents and positions are important so that the primer is soluble in water (16).

The finding that many cell lines are mixtures of the cells without EGFR mutations and of cells with specific EGFR mutations (i.e., genetic heterogeneity of EGFR) surprised us. Cell lines are usually isolated from advanced cancers and have expanded many passages in culture medium. Therefore, they are considered to better represent the later stages of cancers and to possess genetic changes that better provide a growth advantage (18, 19). Thus, they are often considered a clone, although they may actually contain a variety of subclones (20, 21). Coexistence of the wild-type cells and the mutant cells in the same culture suggests that mutation of the EGFR gene does not confer cells with a major growth advantage. It also suggests that the EGFR gene is unstable and that mutation(s) may occur in established cancers. Wild-type EGFR transmits growth signals to extracellular signal-regulated kinase (Erk) to promote cell division and to Akt and to signal transducer and activator of transcription to promote cell survival (22). L858R (T2573G) and L747-E749del A750P (2239-2247del, G2248C), two major mutations that together account for two thirds of the mutations in lung cancers (6, 7), produce EGFR protein that only stimulates the survival signaling to Akt (23) in cooperation of Erb-B3 (24) but not the growth signaling to Erk (23). Therefore, these EGFR mutations may not dramatically change the rate of cell growth, which in turn may allow the coexistence of both wild-type and mutant cells. In several cell lines, almost 100% of the cells have mutations. In such cells, the mutation may provide a strong survival advantage.

Genetic heterogeneity of EGFR may contribute to the variable clinical response to gefitinib. For example, we have seen patients in whom primary tumors responded poorly to gefitinib but metastatic legions responded well, and patients in whom tumors initially responded to gefitinib but became refractory during therapy and the tumor relapsed. Genetic heterogeneity of EGFR in the tumor may explain the disease in these patients. Genetic heterogeneity may also explain tumors with two different EGFR mutations (25), a situation we also observed in PC-3, PC-9, EBc-1, KTSq-1, and RERF-LC-A1 cells (Table 2). Serial examination of tumor samples for changes in EGFR mutations during the course of gefitinib therapy, and examination of different parts in a single tumor using microdissection are warranted.

Grant support: Japan Society of Promotion of Science grant-in-aid for scientific research 16390236 and Japanese Foundation for Multidisciplinary Treatment of Cancer grant-in-aid.

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 Drs. Shunsuke Kobayashi (Miyagi Cardiovascular and Respirology Center, Miyagi, Japan), Tohru Kameya, and Shi-Xu Jiang for permission to use their cell lines.