Association of Polymorphisms in AKT1 and EGFR with Clinical Outcome and Toxicity in Non–Small Cell Lung Cancer Patients Treated with Gefitinib Free

This article is featured in Highlights of This Issue, p. 521

Non–small cell lung cancer (NSCLC) is the leading cause of cancer death worldwide. Chemotherapy increases survival in advanced disease, but overall prognosis remains poor.

Several new agents against critical molecular targets in cancer cell survival have been tested. Among these, inhibitors of the epidermal growth factor tyrosine kinase (EGFR-TKI), such as gefitinib and erlotinib, have shown antitumor activity in 4% to 27% of unselected NSCLC patients and have been registered in several countries for advanced NSCLC patients who failed chemotherapy (1–4).

Although no selection of patients was done in the pivotal studies, several clinical and biological signatures were shown to affect the sensitivity or resistance to EGFR inhibitors (5, 6). Many studies documented a relationship between female gender, adenocarcinoma histology, Asian ethnicity, and never having smoked and higher response rates to EGFR-TKIs (6). Specific activating mutations within the EGFR-TK domain are associated with these characteristics and are predictive of EGFR-TKI activity (7, 8). Prospective studies with gefitinib or erlotinib in patients harboring EGFR-activating mutations reported response rates ranging from 60% to 82%, suggesting that these drugs might be a valuable alternative to chemotherapy in selected patients (9, 10). However, in the BR.21 trial, the treatment with erlotinib resulted in statistically significant improvements in overall survival (OS) and quality of life in all of the subsets of patients (11, 12), and several studies reported that patients without mutations responded to EGFR-TKIs (13, 14). Similarly, some NSCLC cell lines with normal EGFR status are sensitive to these agents (15), suggesting that the question of who should receive EGFR-TKI therapy is still not completely answered and other molecular factors need to be identified.

Better outcome in gefitinib-treated patients was also associated with increased EGFR gene copy number (16, 17). In contrast, recent studies identified mutations in EGFR that are associated with resistance (18), and emerging data suggest that resistance to EGFR inhibition may be also due to the activation of proteins downstream of the receptor (K-Ras, mitogen-activated protein kinase, and signal transducers and activators of transcription 3), tumor dedifferentiation (so-called epithelial-mesenchymal transition), and other cell surface proteins, such as cMET (5, 6, 19–21). Nonetheless, all these changes do not completely explain the variable clinical outcomes, and identification of other biomarkers of EGFR-TKI sensitivity/resistance may help in optimal patient selection, especially in populations with a low frequency of EGFR mutations, such as Caucasians.

Germ line DNA variations of EGFR might also predict response to EGFR-TKIs. The length of a CA-repeat in intron-1 was associated with gefitinib response and/or time-to-progression(TTP)/OS in several retrospective studies (22–26). However, this polymorphism was not correlated with EGFR expression in cell lines and NSCLC tissues (24, 27), and no association with clinical outcome was observed in the largest pharmacogenetic analysis in NSCLC Caucasian patients treated with gefitinib (28). Interethnic differences in CA-intron-1-repeat frequency and linkage disequilibrium with other polymorphisms may explain these controversial results (29, 30). The variants −216G/T and −191C/A have been associated with increased EGFR promoter activity and gene expression, whereas a single nucleotide polymorphism (SNP) at codon 497 is associated with decreased EGFR activity (30, 31). A combination of −216G/T and R497K polymorphisms was weakly correlated with drug response in the NCI60 panel (27), whereas an integrated analysis of genotypic/pharmacokinetic variability showed a strong association between diarrhea in erlotinib-treated patients and the two polymorphisms in the EGFR promoter (32). Further studies are warranted to evaluate the interaction between mutations, gene amplification/expression, and polymorphisms in EGFR and how these factors might affect responsiveness to EGFR-TKIs.

Clinical response to EGFR-TKIs may also be influenced by changes affecting downstream EGFR signal transducers. The serine-threonine kinase Akt is a central player in the phosphoinositide 3-kinase oncogenic pathway and is involved in antiapoptosis or in pro-cell proliferation effects. Retrospective studies showed that patients with phospho-Akt–positive tumors had a better response, disease control rate, and TTP, suggesting that gefitinib may be most effective with basal Akt activation (20, 33). However, in the ONCOBELL prospective trial, only the EGFR fluorescence in situ hybridization, and not Akt immunohistochemistry, predicted response to gefitinib (34). Recent studies discovered an oncogenic mutation in the AKT1 subunit, which stimulates Akt signaling and induces cellular transformation, but its rare incidence suggests that it may not play a role in NSCLC development or response to EGFR-TKIs (35). A candidate gene approach focusing on apoptotic pathways identified two functional polymorphisms (AKT1-SNP3 and SNP4) affecting the expression and activity of Akt (36). The haplotype including these polymorphisms was associated with lower protein levels in tissues from Caucasians and contributed to the lowest apoptotic response of EBV-transformed lymphoblastoids to radiation (36, 37).

A recent trial reported that other genetic variations in AKT1 were associated with increased recurrence and significantly shorter survival in esophageal cancer patients treated with regimens including fluoropyrimidines, platinum compounds, and taxanes, but not with gefitinib, suggesting that genetic variations in the phosphoinositide 3-kinase/AKT pathway may be prognostic and/or predictive factors of drug response (38). However, no data are available on the effect of the AKT1-SNP3 and SNP4 polymorphisms on drugs targeting the EGFR/phosphoinositide 3-kinase/AKT signaling.

The purpose of this study was to retrospectively evaluate the associations between selected functional EGFR and AKT1 polymorphisms and clinical outcomes in gefitinib-treated NSCLC patients.

Fifty-one and 45 patients were treated within the Expanded-Access-Program of gefitinib at Humanitas Clinical Institute (Rozzano-Milano, Italy) and Livorno Civil Hospital (Livorno, Italy), respectively. The selection was based on the diagnosis of histologically confirmed NSCLC with measurable, locally advanced, or metastatic disease, progressing/relapsing after chemotherapy, or with contraindications for chemotherapy. The clinical trial and the study on patients' specimens was approved by the Ethics Committees (ClinicalTrials.gov-ID-NCT00831454) and conducted in accordance with principles stated in the Declaration of Helsinki. Patients received 250 mg/d gefitinib orally and were evaluated for response after 2 mo according to the Response Evaluation Criteria in Solid Tumors criteria. Tumor response was assessed by computed tomography scan, with a confirmatory evaluation repeated in patients with complete/partial response (PR) and stable disease (SD) at least 4 wk after the initial determination of response. Toxicities were assessed using the National Cancer Institute Common toxicity criteria manual version 3.0.⁹

Based on the analysis of the results obtained in the gefitinib-treated patients, an additional population of 127 chemotherapy-treated/gefitinib-nontreated NSCLC patients (39) was used to evaluate the prognostic versus predictive role of the polymorphic genotypes that were associated with OS.

DNA was isolated from blood or paraffin-embedded tumor samples using the microDNA-kit (Qiagen). DNA yields and integrity were checked at 260 to 280 nm with the NanoDrop-1000-Detector (NanoDrop-Technologies).

Nested PCR to amplify EGFR (exons 18–21) and K-Ras (exon 1–2), and sequencing of PCR products on a ABI-3100 genetic analyzer (Applied Biosystems) was done as previously described (20). The CA repeat length in the intron-1 (CA)n and the −216G/T (dbSNP-ID:rs712829) EGFR polymorphisms were studied as previously described (28, 32). The EGFR −191C/A (dbSNP-ID:rs712830) and R497K (G/A, dbSNP-ID:rs11543848), and the AKT1-SNP3 C/T (dbSNP-ID:rs3730350) and SNP4 G/A (dbSNP-ID:rs1130233) were studied with Taqman probe–based assays using the ABI PRISM-7500HT. Specific primers and probes for EGFR R497K and AKT1 polymorphisms were obtained from Applied Biosystems (C__16170352_20, C__193157_10 and C__7489835_10, respectively), whereas primers and probes for EGFR −191C/A were designed using the FileBuilder-2.0 software (Applied Biosystems), on the basis of the gene sequence from Genbank, (F): 5′-CCCCGCACGGTGTGA-3′; (R): 5′-GGCTAGCTCGGGACTCC-3′; (P): 5′-CCTCGGCCGCG[G/T]CG-3′, for the FAM/VIC-fluorescent reporter, respectively. The PCRs were done using 20 ng of DNA diluted in 5.94 μL DNase-RNase–free water, with 6.25 μL of Taqman Universal PCR Master Mix and 0.31 μL of the mix, including the primers and probes, in a total volume of 12.5 μL. After thermal cycling, the 7500HT instrument determined the allelic content of each sample in the plate by reading the generated fluorescence using the SDS-2.0 software (40).

Previous studies evaluated the relationships of EGFR polymorphisms and expression with EGFR-TKI sensitivity (27), but no data are available on the correlation between AKT1 polymorphisms and AKT1 expression or gefitinib sensitivity. Therefore, we evaluated AKT1 genotypes and mRNA expression in 15 NSCLC cell lines, including cells already studied for determinants of gefitinib sensitivity (20). All the cells were obtained from the American Type Culture Collection within the last 5 y and tested within the last 6 mo by morphology check by microscope and growth curve analysis according to the Cell Line Verification Test Recommendations (American Type Culture Collection Technical Bulletin no. 8, 2008). Periodic assays were done to detect Mycoplasma contamination (the last test was done on March 2009), and the identity of the cells was further confirmed by sequence analysis of their specific mutational status in EGFR and K-Ras, as already described by the Cancer Cell Line Project of the Wellcome Trust Sanger Institute.¹⁰

DNA extraction and genotyping were done as described above. RNA was extracted using Trizol (Invitrogen), and 500 ng were reverse transcribed using random hexamers from the DyNAmo-cDNA-Synthesis kit (Finnzymes-Oy). The AKT1 mRNA expression was studied by Taqman-PCR with the Hs00920503_m1 kit (Applied Biosystems). The relative levels of AKT1 mRNA were measured using a standard curves method and normalized to β-actin. The standard curves were derived from dilutions of cDNA from the Quantitative-PCR Human-Reference Total-RNA (Stratagene). All the samples were amplified in triplicate, with appropriate nontemplate controls, and the CV was <2%.

Demographic and clinical information were compared across genotype using Pearson χ² two-sided test. Patients achieving complete response or PR were defined as “responders,” whereas patients with SD or progressive disease (PD) were defined as “nonresponders.” Additional analyses were done by grouping patients with complete response/PR and SD (“patients with clinical benefit”) versus PD (40).

TTP was calculated from the date of the first dose of gefitinib to the date of clinical and/or radiological evidence of progression/death, whichever occurred first, whereas OS was calculated from the day of treatment start to the end point (death or censoring). The Kaplan-Meier method was used to plot TTP and OS, and the log-rank test was used to compare curves.

The univariate analysis included clinical characteristics, including baseline demographic characteristics, such as sex, performance status (PS; 0–1 versus 2–3), age (<median versus ≥median), clinical stage (IIIB versus IV), histology [adenocarcinoma versus other histology, and bronchiolo-alveolar carcinoma (BAC) versus other histology], smoking history [never, i.e., <100 cigarettes/lifetime versus former (quit smoking >12 mo before starting gefitinib) + current smokers] and prior chemotherapy (yes versus no), as well as clinical treatment effects, such as toxicity (grade 0–1 versus grade 2+), and response to gefitinib, categorized as described above. Regarding polymorphisms, statistical analyses were done grouping patients according to the three different genotypes for each polymorphism and to different allelic activity, as reported previously, i.e., EGFR −191C/C versus C/A-A/A, EGFR −216G/G versus G/T-T/T, EGFR R497K G/G-G/A versus A/A, AKT1-SNP3 C/C-C/T versus T/T, and SNP4 G/G-G/A versus A/A (30, 32, 36). For the EGFR (CA)n polymorphism, patients were classified as S/S, L/L, and S/L if the number of repeats was ≤16 on both alleles, >16 on both alleles and ≤16 in one allele, and >16 in the other (25, 32).

The significant prognostic variables in the univariate analysis were included in the multivariate analyses using Cox's proportional hazards model to identify factors of independent significance. A step-down procedure was used based on the likelihood ratio test, and hazard ratios were calculated to estimate the magnitude and the direction of the effect. According to the usual practice in genetic epidemiology, the multivariate statistical model treated the most common polymorphic variants as the baseline.

The Pearson χ² test was also used to determine the correlation between EGFR mutations and polymorphisms, and to evaluate whether the baseline demographic and biological characteristics were significantly different in the gefitinib-treated and gefitinib-nontreated patients.

The Student's t test, one-way ANOVA, and the nonparametric Kruskal-Wallis test were used when analyzing the correlation of AKT1-SNP4 polymorphisms with AKT1 mRNA expression and in vitro gefitinib sensitivity, as well as EGFR and K-Ras mutations and gefitinib sensitivity, whereas the Spearman correlation test was used to investigate the correlation between gene expression and gefitinib IC₅₀s.

All the analyses of the samples were done in a blinded fashion relative to clinical outcome. Data were analyzed using the SPSS-16 software (SPSS, Inc.). P values between 0.05 and 0.1 were regarded as marginally suggestive of an association (“trend toward a significant correlation”), but only P < 0.05 were considered statistically significant. However, in the univariate analyses for clinical outcome according to polymorphisms, a Bonferroni correction required a P < 0.05/6 = 0.008 for statistical significance.

Table 1 summarizes the clinical characteristics and their accordance with the outcome. Response and TTP data were available from all, and OS data were available from 95 patients. There were 83 deaths (event rate, 86.5%), whereas 6 individuals were alive without PD at last contact.

Of the 96 patients evaluated, 17.7% had PR, whereas 38.6% and 43.7% had SD and PD, respectively. Female gender and never-smoker status were significantly associated with a better response. Female gender was also associated with clinical benefit (i.e., PR+ SD) as well as BAC histology.

Median TTP and OS were 3.2 and 7.7 months, respectively. BAC histology was associated with significantly longer TTP and OS, whereas negative smoker history was associated with longer TTP. Surprisingly, older age (≥64 years) was also associated with a longer TTP but not with OS. Finally, a significant association was detected between both TTP and OS, and response. Similarly, TTP and OS were associated with clinical benefit. Median OS of responders, patients with clinical benefit, and nonresponders were 18.8, 11.3, and 4.1 months, respectively.

Seven of the 62 patients with available tissues for mutation analysis harbored EGFR exon-19 deletions, whereas 2 had the L858R EGFR exon-21 and 15 carried K-Ras mutations. No tumor with K-Ras mutations harbored EGFR mutations. EGFR-activating mutations were significantly associated with response [67% of the patients harboring EGFR mutations achieved response versus 6% of the patients without mutations (P < 0.01)] and showed a trend toward an association with clinical benefit: 8 of the 9 patients with mutations achieved clinical benefit, i.e., 89%, versus 55% of the patients without mutations (P = 0.07). EGFR mutations were also significantly associated with longer TTP [median TTP, 9.0 months in patients with EGFR mutations versus 3.0 months in patients without mutations (P < 0.01)]. Similar results were reported for OS, with median OS of 18.0 versus 6.0 months in patients with and without mutations (P < 0.01), respectively. In contrast, no significant associations were reported between K-Ras mutations and TTP and OS, but none of the patients bearing K-Ras mutations achieved PR after gefitinib treatment (P = 0.08).

Previous studies showed no differences in polymorphisms analyzed in tumors and normal tissues (41, 42). However, because high-level gene amplification of EGFR in NSCLC cells might result in homozygous genotypes in individuals who are heterozygous in the germline, we performed preliminary studies of EGFR −191C/A and R497K polymorphisms in 45 paired samples of germ line and lung cancer DNA, showing identical interindividual genotypes between normal and malignant tissues. Therefore, for the 96 patients enrolled in this study, genotyping was done in DNA extracted from the available 51 tumors and 45 blood samples. However, in 17 of the 96 patients enrolled, we obtained data from both the blood and the tumors.

All polymorphisms followed the Hardy-Weinberg's equilibrium (Table 2), as calculated with the SNP analyzer software,¹¹

Polymorphisms were not associated with response, although none of the patients harboring the AKT1-SNP3 T/T or the SNP4 A/A variants achieved response after gefitinib (Table 2). No significant differences in TTP and OS were detected with respect to EGFR and AKT1-SNP3 genotypes. However, the AKT1-SNP4-A/A genotype was associated with shorter TTP and OS. In particular, the median OS between patients harboring the AKT1-SNP4 A/A variant was 2.2 months, whereas patients carrying the G/A and the G/G genotype had a median OS of 7.0 and 8.0 months, respectively (Fig. 1A). To keep at minimum the probability to find a statistically significant difference purely by chance, the usual nominal level (P = 0.05) has been lowered to 0.008 by Bonferroni adjustment for multiple comparisons. After the Bonferroni adjustment, no statistically significant differences were detected for TTP, whereas the AKT1-SNP4 genotype was still significantly correlated to OS (P = 0.007).

None of the patients harboring the AKT1-SNP4-A/A genotype had EGFR or K-Ras mutations. The median age of these patients (four males and two females) was 67.5 y (range, 49–80). Regarding the histology, three had adenocarcinoma, two squamocellular carcinoma, and one large-cell carcinoma. Five of six of these patients were smokers, whereas four had a PS of 0, and two had a PS of 2.

Because the polymorphism results may reflect a generally worse prognosis with the presence of the A/A variant as opposed to a pharmacogenetic response, 127 gefitinib-nontreated NSCLC patients were also assessed for the AKT1-SNP4. These patients were enrolled in the NVALT-7 study and treated only with pemetrexed (n = 62) or carboplatin-pemetrexed (n = 65) regimen without receiving EGFR-TKIs as salvage therapy. In this study, the patients who received carboplatin-pemetrexed had a significantly longer TTP than patients receiving pemetrexed alone, but no significant differences were detected for OS (39).

Although the purpose of this analysis was to make a preliminary evaluation whether the polymorphic genotypes associated with OS were predictive of drug activity or prognostic factors, and not to perform a case-control study, we analyzed whether the baseline demographic characteristics were significantly different in the gefitinib-treated and gefitinib-nontreated patients. No data on BAC histology and smoking status in patients of the NVALT-7 study were available, but no significant differences were measured for age (≤65 versus >65, P = 0.47), sex (P = 0.70), PS (0-1 versus 2-3, P = 0.10), and histology (adenocarcinoma versus others, P = 0.06) in this additional series of NSCLC patients who did not receive EGFR-TKI treatment with respect to the NSCLC patients treated with gefitinib.

AKT1-SNP4 genotypes were detectable from 124 of 127 patients. The allelic frequencies of G/G, G/A, and A/A SNP4 genotypes were similar to those in gefitinib-treated patients (62.1, 31.4, and 6.5%, respectively, P = 0.80 in the Pearson χ² test of the G/G-G/A and A/A groups with respect to the gefitinib-treated patients). No significant differences were reported in OS in patients with the AKT1-SNP4-A/A and patients with the AKT1-SNP4-G/G-G/A genotypes (Fig. 1B).

To evaluate the risk of disease progression and death, we carried out two Cox regression analyses entering variables significantly associated with TTP and OS in the univariate model. Among biological characteristics (Table 3), the AKT1-SNP4 A/A variant showed a trend toward a significant association with risk of progression, in the analysis including the EGFR mutation covariate. However, in the regression model done on all the patients evaluable for the polymorphisms analyses (n = 94), excluding the EGFR mutational status, the AKT1-SNP4-A/A genotype showed a significantly higher risk of progression (P = 0.04). Regarding the risk of death, AKT1-SNP4-A/A was an independent prognostic factor. In particular, the A/A genotype was significantly associated with increased risk of death, in the model of 62 and of 94 patients, including and excluding the mutational status, respectively.

Considering all the significant variables in the univariate analysis (i.e., response, age, and smoking status for TTP and response for OS), we found that response was always associated with both TTP and OS, whereas age and smoking status were not independent predictors of TTP and OS in the model including and excluding the EGFR mutational status. In this analysis, the AKT1-SNP4-A/A genotype showed a trend toward a significant association with TTP in the model of 94 patients (P = 0.09), and it was still a significant predictor of OS in both models.

Of the 89 patients evaluable for skin toxicity, 24 patients (27%) developed grade 1 skin rash, whereas 21% and 6% had grade 2 and 3 skin rash, respectively. Regarding the 87 patients evaluable for gastrointestinal toxicity, 29 patients (33%) had grade 1 diarrhea and 7% developed grade >1 diarrhea. According to previous analyses (32), patients were grouped in patients with toxicity versus patients without and patients with grade 0 to 1 versus patients with “high” toxicity (grade 2–3). No significant correlations were observed between toxicities and most patients' characteristics, TTP, or OS. However, patients affected by adenocarcinomas experienced higher skin toxicity (P = 0.04) than patients with other tumor types.

No correlations were observed grouping patients according to EGFR or K-Ras mutational status.

The correlations between polymorphisms and toxicity are described in Table 4. EGFR −191C/A polymorphism were associated with grade >1 diarrhea (P < 0.001). In particular, five of the six patients who experienced severe diarrhea carried the EGFR −191C/A variants C/A and A/A (i.e., one third and 50% of the C/A and A/A patients had grade 2–3, whereas only 1 of 70 C/C patients had grade >1 diarrhea). Indeed, the Fisher's exact test for 3 × 2 table showed a strong association (P < 0.0001). The EGFR −216G/T polymorphism was also associated with severe diarrhea: all the patients who experienced grade >1 diarrhea carried the G/G variant (P < 0.01). Furthermore, the G/G variant showed a trend toward a significant association with any grade of diarrhea (P = 0.06). Similar results were found for the EGFR R497K polymorphism, showing high toxicity in 4.6%, 3.2%, and 30% of the patients carrying the genotype G/A, G/G, and A/A, respectively (P = 0.01 in the 3 × 2 table). Furthermore, the frequency of the A/A genotype was significantly higher in patients with grade >1 (50%) than in patients with grade 0 to 1 diarrhea (10%).

No significant association was reported between EGFR polymorphisms and skin toxicity, as well as between EGFR (CA)n and AKT1-SNPs, and both skin and gastrointestinal toxicity.

The significant association between AKT1-SNP4 and clinical outcome prompted us to perform an in vitro study on a panel NSCLC cells already characterized for EGFR/K-Ras mutations and gefitinib sensitivity (Table 5). AKT1-SNP4 genotypes were associated with AKT1 mRNA expression, and when combining the AKT1-SNP4 G/G and G/A versus A/A, there was a significant association with both AKT1 mRNA and gefitinib IC₅₀s (Fig. 1C and D). Finally, there was a significant inverse correlation between AKT1 mRNA levels and gefitinib IC₅₀s (Spearman r = −0.54; P = 0.04). Gefitinib sensitivity was also significantly associated with the occurrence of EGFR-activating mutations (P = 0.04), while there was a trend toward significant association with reduced sensitivity in the cells harboring K-Ras mutations (P = 0.07).

This study evaluated the effect of selected functional polymorphisms on outcome in advanced NSCLC patients and, to our knowledge, is the first to show the effect of a polymorphism in AKT1 on the TTP and OS of gefitinib-treated patients. Furthermore, we reported a significant association between EGFR polymorphisms with gastrointestinal toxicity after gefitinib treatment, as reported previously in erlotinib-treated patients (32).

A growing body of studies is trying to identify the molecular predictors of response and toxicity to EGFR-TKIs in NSCLC. This includes tumor-related molecular markers, such as EGFR mutations, copy number, and protein expression, but also germ line markers such as polymorphisms in EGFR or EGFR-related genes (44).

Most previous studies focused on polymorphisms affecting EGFR expression, and reported controversial results on their effect on EGFR-TKIs activity (23, 24, 28, 29). More recently, two linked promoter polymorphisms (−216G/T and −191C/A) and the R497K polymorphism have been associated with the modulation of EGFR expression and activity, and showed some correlation with response to erlotinib and 11 other EGFR-TKIs in the NCI60 panel (27, 30, 31). Furthermore, Rudin et al. (32) showed a significant association between the polymorphisms in the EGFR promoter and diarrhea, independent from erlotinib plasma concentration, in 80 NSCLC, head and neck, and ovarian cancer patients. We observed similar results in our uniform population of 96 NSCLC patients treated with gefitinib, with severe diarrhea occurring significantly more frequently in patients harboring the EGFR −191C/A and A/A, the EGFR -216G/G, and the EGFR R497K A/A variants. The pathophysiology of anti-EGFR–induced diarrhea is thought to result from excessive chloride secretion, inducing a secretory diarrhea (45). Therefore, the diarrhea might result from the higher EGFR expression in the intestinal lumen associated with the EGFR promoter polymorphisms variants as suggested previously (32). In contrast, the A allele in the R497K polymorphism is associated with alterations in EGFR ligand binding, and studies in colorectal cancer tissues showed a decreased phosphorylation of EGFR, whereas no differences were detected for EGFR mRNA expression (42). However, the R497K polymorphism was also associated with the decreased activation of c-Myc (42), whose activity is also downregulated by the Escherichia coli heat-stable enterotoxin STa, a major causative agent of secretory diarrhea (46).

In the present study, no associations were detected with skin toxicity, which was correlated with erlotinib exposure levels in the study by Rudin et al. (32). Data from multiple studies with erlotinib showed a consistent relationship between rash and response, as well as between rash and survival (2, 47). The relationship between rash and outcome is currently less consistent for gefitinib, and no significant association was detected in our study. These results may be due to the lower plasma concentration of gefitinib compared with erlotinib when administered at the recommended doses of 250 and 150 mg/d, respectively. Furthermore, the 250 mg/d is well below the maximum tolerated dose and the incidence of skin toxicity is significantly lower than with 500 mg/d gefitinib (47).

Female gender, never-smoker status, and BAC histology were associated with response and improved TTP or OS. These results corroborate the findings of several published studies, showing that our population is similar to other Caucasian populations treated with gefitinib. EGFR-activating mutations were also predictive for response, progression, and OS, as reported previously (6, 44), but not for toxicity. However, the detection of mutations requires tumor material, whose handling can be problematic in the advanced cancer setting in which diagnosis is usually done from small needle biopsy samples. From an analysis standpoint, polymorphisms are convenient also because there are no differences in tumor and normal tissues as reported previously (41, 42).

Several polymorphisms of determinants of drug activity/metabolism have already been shown to play a critical role in predicting toxicity and response of anticancer treatment (48). Previous studies identified functional polymorphisms affecting the expression and activity of Akt (36, 37), and this is the first study aimed at evaluating their possible effects in gefitinib-treated patients. Our data showed that the AKT1-SNP4 A/A genotype was associated with shorter OS. Given the small number (n = 6) of patients harboring the AKT1-SNP4-A/A genotype, to evaluate whether other poor prognostic factors could potentially explain their short survival, we carefully checked their baseline demographic and biological characteristics, which were similar to the average of the studied population. Furthermore, at multivariate analysis, the AKT1-SNP4 polymorphism remained an independent predictive parameter of progression and death risk.

Possible factors involved in resistance to anti-EGFR therapies include the occurrence of mutations in the target or in the downstream pathways, such as the K-Ras and the EGFR exon-20 T790M mutations, which are mainly associated with primary and secondary resistance to EGFR-TKIs, respectively (5, 18). Gene amplification, such as cMET amplification, involved in 20% of acquired resistance cases in NSCLC patients treated with EGFR-TKIs, and modulation of other pathways also contribute to resistance to EGFR-TKIs in NSCLC (19–21). Previous studies showed that SRC activation induced gefitinib resistance by modulation of both the extracellular signal-regulated kinase and Akt signaling pathways in human gallbladder adenocarcinoma cells (49). Oncogenic SRC protein also induced the epithelial-mesenchymal transition, which is associated with the loss of E-cadherin complexes at cell-cell adhesions and, consequently, with an increased invasive/metastatic potential and worse NSCLC prognosis (50). Previous immunohistochemical analyses for SRC, E-cadherin, pAkt, pERK, and p-cMET performed in tumor tissues from 51 patients enrolled in this study showed that cMET[pY1003] immunoreactivity was associated with shorter TTP (20). Interestingly, the majority of NSCLC patients with cMET[pY1003]-positive tumors had positive SRC expression and negative E-cadherin, consistent with the role of cMET in epithelial-mesenchymal transition (20). However, the identification of other biomarkers of gefitinib resistance may help in optimal patient selection, and the analysis of polymorphisms can be easily done in blood and is easier to adopt in the routine clinical setting than tumor expression arrays.

To evaluate whether the AKT1-SNP4 polymorphism is a candidate biomarker predictive of drug activity or a prognostic factor, we used a population of gefitinib-nontreated advanced NSCLC enrolled in the NVALT-7 study. These patients were treated only with the pemetrexed (n = 62) or carboplatin-pemetrexed (n = 65) regimen, without receiving EGFR-TKIs as salvage therapy. In this study, the patients who received carboplatin-pemetrexed had a significantly longer TTP, but no significant differences were detected for OS (39). The lack of correlation between the AKT1-SNP4-A/A genotype and survival in these gefitinib-nontreated patients suggests that it is not a prognostic factor, whereas it might be a predictive factor of gefitinib activity. These results have to be validated in a larger cohort of patients, in prospective multicenter trials, as well as in case-control studies (48).

To gain further insight into the mechanisms behind our findings, we performed in vitro studies showing a significant association with both AKT1 mRNA expression and gefitinib IC₅₀s, in agreement with the clinical results. Because sensitivity to gefitinib was correlated with the occurrence of EGFR-activating mutations as previously described (15), whereas the only 2 of the 15 available NSCLC cell lines harboring the A/A genotype have also K-Ras mutations, we cannot draw major conclusions from these results. However, there was a significant association considering the three different genotypes and AKT1 mRNA levels, which were clearly correlated with sensitivity to gefitinib. Previous studies reported controversial relationship between Akt and EGFR expression and EGFR-TKI activity both in the preclinical and clinical setting (15, 34), and more genotype-phenotype correlation studies and functional analyses of the critical genes involved in the EGFR pathway are warranted.

In conclusion, the present study describes for the first time a correlation of the AKT1-SNP4 polymorphism with TTP and OS in advanced NSCLC patients treated with gefitinib. Furthermore, EGFR −191C/A, -216G/T, and R497K polymorphisms were associated with gastrointestinal toxicity after gefitinib, as reported previously in erlotinib-treated patients (32). Validation of the role of these polymorphisms in further clinical trials will ultimately offer new tools for treatment optimization of currently available EGFR-TKIs in selected NSCLC patients.

No potential conflicts of interest were disclosed.

We thank Dr. Francesco Crea, M.D., and Dr. Davide Torti, M.D., Scuola Superiore di Studi Universitari e Perfezionamento Sant'Anna, Pisa, Italy for their initial contributions and Dr. Marielle I Gallegos-Ruiz, Ph.D., VUmc, Amsterdam, the Netherlands, for sequencing analysis and Fondazione A.R.C.O. for grant support.

Grant Support: Fondazione A.R.C.O. (C. Tibaldi).

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.