Patterns of EGFR, HER2, TP53, and KRAS Mutations of p14^arf Expression in Non–Small Cell Lung Cancers in Relation to Smoking History

Introduction

Lung cancer accounts for over a million annual deaths worldwide, ∼90% attributable to smoking. Non–small cell lung cancer (NSCLC) represents 75% of primary lung cancers and includes squamous cell carcinomas (SCC), adenocarcinomas (ADC), and large-cell carcinomas (LCC). Irrespective of histologic type, common genetic alterations in NSCLC are mutations in TP53, defects in the CDKN2a/RB pathway, loss of alleles on chromosome 3p encompassing FHIT, SEMA3B, and RASSF1A, and aberrant promoter methylation in O⁶MGMT, CDKN2a, DAPK, TIMP-3, or RASSF1A. Mutations at codon 12 in KRAS are found in 30% to 40% of ADC but are rare in other histologies ( 1). In NSCLC of smokers, there is a characteristic pattern of TP53 mutations ( 2, 3), compatible with the effects of DNA damage by tobacco carcinogens, such as polycyclic aromatic hydrocarbons. Mutations in EGFR, encoding the epidermal growth factor receptor, have emerged as a frequent molecular alteration in NSCLC of never smokers. These mutations have attracted considerable clinical interest due to their association with tumor sensitivity to the antiproliferative effects of small-molecule tyrosine kinase (TK) inhibitors, erlotinib and gefinitib ( 4– 6). Most of described EGFR mutations (85%) fall into two categories: point mutations in exon 21 (L858R; 40%) and in-frame deletions of two to nine residues in exon 19, encompassing residues of a conserved LREA motif (residues 748–751; 45%; ref. 7). Other mutations include rare insertions in exon 20, a minor hotspot at exon 18 (5%) and scattered missense mutations in exons 18 to 21. There is structural and biochemical evidence that the L858R mutation and the short deletions in exon 19 modify the geometry of the ATP binding cleft in the TK of EGFR ( 7), resulting in a hyperactive form of the receptor. EGFR mutations are inversely correlated with tobacco consumption and are reportedly frequent in ADC of women of Asian descent ( 8). R1-C1: one sentence mutations in EGFR may coexist with mutations in TP53 ( 9, 10) but are mutually exclusive of mutations in KRAS, in agreement with the notion that KRAS is a downstream effector of EGFR signaling. Moreover, HER2 (a member of the ERBB family to which EGFR also belongs) is dysregulated in many cancers. The most common alteration of HER2 is overexpression with amplification, which is frequent in breast and ovarian cancers and is associated with poor prognosis ( 11). HER2 can dimerize with other members of the ERBB family. The strongest and the most powerful heterodimer formed is EGFR/HER2 ( 12). Recent studies have reported that mutations in the TK domain of HER2 are detected in lung cancers ( 13).

In this retrospective study, we have analyzed EGFR (exons 18–21) and HER2 (exons 19 and 20) mutations in 116 NSCLC of well-characterized current, former, and never smokers, analyzed previously for KRAS and TP53 mutations ( 2). In parallel, we have assessed by immunohistochemistry the expression of p14^arf and p16^ink4a, the products of the CDKN2a locus ( 10, 14). p16^ink4a is a cyclin kinase inhibitor that controls cyclin-dependent kinase 4/cyclin D1 complexes during progression into the G₁ phase of the cell cycle. p14^arf exerts complex functions in cell cycle, apoptosis, and DNA repair ( 15). Its expression is regulated by E2F, a transcription factor activated as end product of the growth signaling cascade initiated by EGFR ( 16, 17). p14^arf complexes with and neutralizes Mdm2, the main controller of wild-type (WT) p53 protein stability, thus allowing p53 to escape degradation and to exert antiproliferative effects. We reasoned that p14^arf represents an important connection between the EGFR and/or HER2 signaling pathways and the p53-dependent suppressive machinery. Here, we show that mutations in EGFR as well as HER2 are associated with TP53 mutation and/or with deficiency of p14^arf expression in NSCLC of never and former smokers. This observation indicates that inactivation of the p14^arf/p53 connection is required for the development of tumors containing EGFR or HER2 mutations.

Materials and Methods

Study subjects and tumors. The settings of the study have been described previously ( 2). The study is based on the INCO Central Europe Health Study, a hospital-based case-control study on lung cancer involving 600 cases and 600 controls. Each participant answered a questionnaire on lifestyle (residence, past health, direct and indirect exposure to tobacco, diet, alcohol consumption, sexual history, and occupational exposure to selected cancer risk factors). NSCLC from a total of 116 subjects from the Moscow area were selected into three groups: 33 never smokers (less than 100 cigarettes smoked in a lifetime; with or without exposure to involuntary smoking from the spouse or at workplace), 38 former smokers (smoking cessation for 2 years or more before diagnosis; average cumulative tobacco smoking, 27.6 pack-years), and 45 current smokers. The latter category was composed of heavy smokers (over 35 pack-years; average consumption, 48.8 pack-years). The characteristics of the patients and their cancers are summarized in Table 1 .

Mutation analysis. DNA was extracted from areas of fresh formalin-fixed, paraffin-embedded tumor sections as selected by the pathologist and analyzed for TP53 (exons 4–10) and KRAS mutations (codon 12) as described elsewhere ( 2). EGFR mutations were detected using PCR-based direct sequencing of the four exons of the TK domain (exons 18–21) using primers and annealing conditions as described by Pao et al. ( 6). HER2 was amplified using the following sense and antisense primers for exons 19 and 20: 5′-GGATCCAGCCCACGCTCTT-3′ (19 forward) and 5′-CTGCAGCCATGGGGTCCTT-3′ (19 reverse) and 5′-CCATACCCTCTCAGCGTA-3′ (20 forward) and 5′-GCTCCGGAGAGACCTGCAA-3′ (20 reverse). Amplifications were done by using a touchdown protocol from 65°C to 62°C for exon 19 and from 61°C to 58°C for exon 20.

Aliquots of PCR products were examined by electrophoresis on 2% agarose gel containing ethidium bromide. PCR products were treated with 2 μL ExoSAP-IT (Amersham Biosciences) at 37°C for 15 min followed by inactivation at 80°C for 15 min and direct sequencing using Applied Biosystems Prism dye terminator cycle sequencing method (Perkin-Elmer) on ABI Prism 3100 Genetic Analyzer (Applied Biosystems).

Immunochemistry. Deparaffinized tissue sections were labeled with specific antibodies against p16^ink4a and p14^arf. The monoclonal antibody p16^INK4a (16P07) (Neomarkers) was used at 1:200 dilution after 60-min Tris-citrate (pH 6) retrieving treatment on a Ventana automated system. The secondary antibody was biotinylated antibody (goat anti-rabbit/goat anti-mouse) followed by streptavidin-peroxidase complex (View kit, Ventana Systems). Detection was done using 3,3′-diaminobenzidine with H₂O₂ (View kit). Results were expressed as expression scores combining the percentage of positive cells for the marker and the staining intensity (1–3 on an arbitrary, semiquantitative scale). Cases were considered as positive when the expression score was >60. This value corresponds to the minimal score of normal pulmonary cells ( 18). p14^arf primary antibody was a polyclonal rabbit antibody (Neomarkers) used at 1:50 after 25-min treatment in citrate (pH 6). As described previously, the secondary antibody was a biotin retrieving labeled donkey anti-rabbit (1:1,250; Jackson ImmunoResearch) followed by detecting the streptavidin-biotin/horseradish peroxidase complex (DAKO). Normal lymphocytes, stromal cells, and normal epithelial cells (bronchial cells and alveolar cells) were considered as positive controls. p14^arf was considered as down-regulated in tumors when the expression score was <60.

Statistical analysis. Relative risks and 95% confidence intervals (95% CI) were calculated after adjustment for sex, age, and education, based on unconditional multivariate logistic regression model using the SAS System for Windows (release 9.1), as were pooled t tests for independent samples with equal variances, Satterthwaite t test for samples with unequal variances (when folded F tests P < 0.05), Pearson χ² analyses, Fisher's exact tests, and ΰ statistics.

Results

EGFR and HER2 mutations in relation to histology, gender, and smoking history. A total of 23 EGFR mutations (exons 18–21) were found in 20 of 116 (17%) NSCLC cases (see detailed results in Supplementary Table). Of these mutated cases, 15 (75%) were ADC, 3 (15%) SCC, and 2 (10%) adeno-squamous (ADC-SCC) cases. Mutations were of the following types: deletions of variable size in exon 19, all encompassing codons 746 to 750 (14 mutations), missense mutations in exon 18 (2 mutations, I706T and S720P), exon 20 (2 mutations, S768I), and exon 21 (4 mutations, L858R, and 1 mutation L858Q). Three cases contained two mutations: in two cases, short deletions in exon 19 coexisted with a missense mutation (L858R or I706T); in the other case, two missense mutations were detected (L858R and S768I).

Mutations were inversely correlated with tobacco consumption, with no mutation in 45 current smokers, 4 in 38 (10%) former smokers, and 16 in 33 (48%) never smokers [odds ratio (OR), 0.05; 95% CI, 0.02–0.18]. In former smokers, 2 of the tumors with mutations were SCC and 2 were ADC. In never smokers, 13 of the tumors with mutations were ADC, 2 were ADC-SCC, and 1 was a SCC ( Table 2 ).

Of patients with EGFR mutations, 15 were women and 5 were men. Thus, EGFR mutations were detected in 48% (15 of 31) of the women included in the study and in 6% (5 of 85) of men. EGFR mutations were found in 50% of ADC in never smokers, irrespective of gender (2 of 4 males and 11 of 21 females).

On the other hand, only 5 (4.3%) mutations were found in HER2 [4 (10.5%) in former smokers and 1 (3%) in never smokers]. Of the 5 mutated cases, 3 were ADC, 1 was SCC, and 1 was ADC-SCC. Four of the five mutations were detected in former smokers, including 2 silent (G787G and L823L) and the 2 missense (E744G and R745D) mutations. The mutation T791I was found in a tumor of never smoker that harbored a KRAS mutation. All HER2 mutations were detected in men and the three missense mutations were associated with mutant TP53 and/or low expression of p14^arf (see details in Supplementary Table).

Patterns of EGFR, HER2, KRAS, and TP53 mutations according to smoking history. Figure 1 shows the prevalence of EGFR mutations according to smoking history, combined with the prevalence of TP53 and KRAS mutations as reported previously ( 2). None of the cases was found to contain mutations in all three genes ( Table 3 ). In never smokers, TP53 mutations were detected in 11 of 16 (69%) cases with EGFR mutation but in only 2 of 17 cases without EGFR mutation. Only one never smoker (N11) had a mutation in KRAS, without TP53 or EGFR mutation but with HER2 mutation. In former smokers, EGFR mutations were found in 10% (4 of 38) of the cases, KRAS mutations in 24%, and TP53 in 59%. Of the four patients with EGFR mutations, none had a TP53 or a KRAS mutation. In contrast, two of the eight cases with KRAS mutation also had a TP53 mutation. In current smokers, EGFR mutations were absent, and KRAS and TP53 mutations were found in, respectively, 16% and 84% of the cases. All but one case with KRAS mutation contained a TP53 mutation. Overall, these results show that although TP53 mutation load was particularly high in current smokers, TP53 mutations were also detected in never smokers in relation to the presence of EGFR mutation. In contrast, mutations in KRAS were relatively rare in never smokers (6%) but more common in current (24%) and former smokers (21%). These differences were mostly due to differences in mutation patterns for ADC: KRAS mutations were detected twice as frequently in ADC of former smokers (38%) than in ADC of never or current smokers (19% in each group).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Prevalence of TP53, KRAS, and EGFR mutations in relation to smoking history. Percentage of tumors with mutation in EGFR (exons 18–21), TP53 (exons 4–9), and KRAS (codon 12), in relation to smoking status (never smokers, n = 33; former smokers, n = 38; and current smokers, n = 45). TP53 and KRAS mutation data are from Le Calvez et al. ( 2).

Expression of p14^arf and p16^ink4 in relation to TP53, KRAS, and EGFR/HER2 mutation patterns. The observation that TP53 mutation is more common in never smokers with than without EGFR mutation led us to formulate the hypothesis that inactivation of p53 function may be required for carcinogenesis driven by mutant EGFR. In vitro studies in human diploid fibroblasts have shown that activation of growth signaling cascades resulted in increased expression of p14^arf, leading to accumulation of p53 and growth suppression ( 16). This model is consistent with the fact that TP53 mutation is uncommon in knockout mice lacking p14^arf expression ( 19). We thus analyzed by immunohistochemistry the expression of p14^arf and of p16^ink4a, the two product of CDKN2a, a locus often altered by deletion, mutation, or hypermethylation in lung cancer (see Introduction). Results were expressed as an expression score combining staining intensity and distribution in tumor cells (see methods). Figure 2 shows an example of staining patterns in tumors with low (expression score < 60) and normal (expression score ≥ 60) p14^arf or p16^ink4a expression. Results are presented in Table 4 . Loss of p14^arf expression seemed to be more common in never smokers (63%) than in former smokers (47%) or current smokers (25%). This trend was not observed with p16^ink4a, the loss of expression of which was marginally more common in current smokers (68%) and former smokers (61%) than in never smokers (53%). Among the 20 tumors containing EGFR mutations, 17 (85%) showed low p14^arf expression. Figure 3 shows the interrelation between TP53 status and p14^arf expression in tumors containing EGFR mutations. All the tumors with WT TP53 showed deficiency of p14^arf expression. Thus, overall, the p14^arf/p53 pathway was altered in all tumors containing mutant EGFR, either by mutation of TP53 (3 cases), by loss of expression of p14^arf (9 cases), or by both (8 cases). Despite the low number of HER2 alterations, a similar concordance between mutation and p14^arf expression was observed with HER2. All cases with missense HER2 mutation were deficient for p14^arf expression, two of them also harboring a TP53 mutation (see Supplementary Table).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Immunohistochemical detection of p14^arf an p16^INK4a in lung tumors. Results were expressed as expression scores combining the percentage of cells positive for each marker and the staining intensity (1–3 on an arbitrary, semiquantitative scale). Representative examples of staining with different scores in SCC (D) and ADC (A–C). p14^arf staining in ADC (A and B). Expression scores, 120 (A) and 20 (B). p16^ink4a staining in ADC (C) and SCC (D). Expression scores, 80 (C) and 0 (D). Note the presence of positive nontumor cells in the stroma of the specimen in (D). Bars, 10 μm.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Concordance TP53, EGFR mutations and p14^arf expression in never smokers. A, percentage of tumors with down-regulated (gray) or normal (white) p14^arf expression among tumors with WT or mutant EGFR. B, among tumors with mutant EGFR, proportion of cases with down-regulated p14^arf expression (gray), TP53 mutation (white), or both (shaded).

Discussion

Somatic mutations in the TK domain of EGFR have emerged recently as common alterations in ADC of never smokers ( 8), whereas mutations in HER2 are not frequent ( 13, 20, 21). EGFR mutation prevalence is inversely correlated with tobacco consumption ( 22, 23). Furthermore, EGFR and KRAS mutations exhibit a mutually exclusive distribution in ADC, suggesting that they belong to redundant oncogenic pathways. EGFR/HER2 and KRAS mutations have been reported to be mutually exclusive ( 24); however, thus far, little is known on whether these mutations may coexist with TP53 mutations. TP53 has been reported as mutated in ADC of never smokers with prevalences between 15% and 50% according to different studies ( 2). In a study by Kosaka et al. ( 10), TP53 mutations were found at similar prevalences in tumors with or without EGFR or KRAS mutations, suggesting that mutation in TP53 occur independently of mutation in EGFR or KRAS.

In the present study, conducted on a series of 116 patients with well-defined smoking history, we found that HER2 mutations are detectable but rare in lung cancer (4.3%), whereas EGFR mutations are more common in (a) tumors of never smokers (16 of 33, 48%) as opposed to former smokers (4 of 38, 10%) or current smokers (0%) and (b) ADC (15 of 20 tumors, 75%) as opposed to ADC-SCC (2 of 20, 10%) and to SCC (3 of 20, 15%). EGFR mutations are mutually exclusive with KRAS but not TP53 mutations. On the other hand, HER2 mutations are mutually exclusive with EGFR mutations but can coexist with KRAS or TP53 mutations. The overall type and distribution of EGFR mutations is compatible with previous reports: mutation in exon 19 represented 61% of the mutations and mutations in exons 20 and 21 represented, respectively, 9% and 22%.

It has been suggested that the prevalence of EGFR mutations was particularly high in ADC of never smoking women ( 25). In the present study, cases harboring EGFR mutations were equally distributed among male and female never smokers (50% in each group). This observation suggests that the high prevalence of EGFR mutations in females is mostly an effect of the overrepresentation of women in the “never smoker” category, compatible with patterns of tobacco consumption among genders.

The CDKN2a locus encodes two distinct suppressor proteins, p16^Ink4a and p14^arf, that are often down-regulated in lung cancers by several mechanisms, including promoter methylation, loss of heterozygosity, point mutations, and loss of expression ( 26). How alterations in these genes correlate with EGFR mutation in lung cancer is not clearly understood. In a recent study, Toyooka et al. ( 27) have shown that the probability of having EGFR mutation was lower among tumors with methylated CDKN2a gene. The patterns of alterations of p14^arf in lung cancers are much less well described as those of p16^ink4a. Eymin et al. ( 28) have shown that loss of p14^arf expression tended to correlate with high expression of Mdm2 in primary lung cancers, suggesting a role for Mdm2/p14^arf in modulating the activity of p53. Nicholson et al. ( 29) have detected homozygous deletions, but not intragenic mutations of p14^arf in 19% of NSCLC, equally distributed among TP53 mutated and WT tumors. CDKN2a promoter methylation has been described in lung cancer cell lines ( 30), in DNA from bronchial lavages ( 31), and in primary SCC ( 32), but how methylation correlates with expression is poorly documented. To determine whether down-regulation of p16^ink4a and p14^arf correlates with the mutation status of EGFR, HER2, KRAS, and TP53 in lung cancers, we have analyzed their expression by immunohistochemistry, a method that captures inactivation of these products by several mechanisms. In agreement with previous reports, p16^ink4a was frequently down-regulated in lung cancers, with a nonsignificant tendency for more frequent loss of expression in current smokers. In contrast, p14^arf was significantly more frequently down-regulated in never smokers than in current smokers. Among patients with EGFR mutations, 85% showed down-regulation of p14^arf expression, whereas the 15% of the cases that retained p14^arf expression harbored a mutation in TP53. Furthermore, the three tumors carrying missense mutation of HER2 were down-regulated for p14^arf, and two of them were mutant for TP53. These results suggest that mutation in EGFR/HER2 requires the functional inactivation of the p14^arf/p53 connection, either by down-regulation of p14^arf, by mutation of TP53, or by both mechanisms. This interpretation is consistent with a model in which hyperproliferative signals generated by mutant EGFR may induce p14^arf, resulting in the activation of p53 as part of a fail-safe mechanism to counter mitotic signaling. The fact that some cancers show both TP53 mutation and low p14^arf expression could reflect a selection for the loss of one or both the INK4 genes that flank the ARF locus. Alternatively, mutation of TP53 could reflect the fact that loss of p14^arf might not completely abrogate p53 function, perhaps due to its participation to other signaling pathways ( 33– 35). Further dissection of these mechanisms will require the analysis of the status of other partners in the p53/p14^arf connection, in particular mdm2.

According to this model, only cells with a functional deficiency in the p14^arf/p53 connection may bypass growth suppression and clonally expand to cause cancer. This hypothesis requires that EGFR mutation is an early event in lung carcinogenesis, which may occur in normal lung, and that mutation in TP53 and/or inactivation of p14^arf is the rate-limiting event that permits clonal expansion. It is intriguing to note that a concordance with p14^arf/p53 inactivation was not observed for KRAS mutations. Indeed, constitutive activation of KRAS is expected to induce an oncogenic stress that activates the p14^arf/p53 pathway, leading to the same type of selection pressure for inactivating this pathway as with EGFR mutation. One possible explanation is that these mutations trend to arise in distinct etiologic context, perhaps as the consequence of distinct mechanisms of carcinogenesis. Of the 16 cases with KRAS mutation in the present series, only 1 was a never smoker, whereas 8 and 7 were former or current smokers, respectively. In line with this hypothesis, Toyooka et al. ( 27) have shown that there were differences in the evolvement of epigenetic alterations between EGFR- and KRAS-mediated tumorigenesis. Further studies are needed to determine whether such differences in genetic and epigenetic alterations may predict different tumor behavior and response to therapy.