Clinical significance of pretreatment serum amphiregulin and transforming growth factor-α, and an epidermal growth factor receptor somatic mutation in patients with advanced non-squamous, non-small cell lung cancer
Department of Respiratory Medicine, Graduate School of Medicine, Kyoto University, 54 Syogoin-Kawaharacho, Sakyo-ku, Kyoto 606-8507;
Circulating amphiregulin and transforming growth factor-α (TGF-α) have been found to be correlated with an unfavorable response to gefitinib based on the identification of patients with a higher probability of resistance to the drug. However, the association between an epidermal growth factor receptor (EGFR) somatic mutation and the overexpression of its ligands has not been determined. To verify the clinical significance of the two serum markers and EGFR mutation status, we determined serum amphiregulin and TGF-α levels by enzyme-linked immunosorbent assay in 93 patients with advanced non-squamous, non-small cell lung cancer and EGFR somatic mutation status using the peptic nucleic acid-locked nucleic acid clamp method in 46 cases. The relationship between each independent clinicopathological variable and the response to gefitinib therapy was examined. We also evaluated the risk factors associated with prognosis. Fourteen (41.0%) of 34 progressive disease cases were positive for amphiregulin (P = 0.007). Eleven (32.4%) of 34 progressive disease cases were positive for TGF-α (P = 0.005). The median survival time of patients with the EGFR somatic mutation was significantly longer (P = 0.01). The same was true of amphiregulin- (P = 0.046) and TGF-α-negative patients (P < 0.01). In multivariate analysis, serum TGF-α positivity (hazard ratio, 2.558; P = 0.005) and the wild type EGFR gene (hazard ratio, 1.894; P = 0.003) were significant independent prognostic factors. Our study demonstrates that the status of the serum EGFR ligand, in addition to EGFR activating mutation, is a predictive factor for response to gefitinib therapy. (Cancer Sci 2008; 99: 2295–2301)
Non-small cell lung cancer (NSCLC) is a major cause of cancer-related mortality worldwide and is expected to remain a major health problem for the foreseeable future. Chemotherapy is the cornerstone for the management of the disease; however, therapeutic impact on patient survival has been modest. Recent discoveries in biomedical research have provided greater understanding of the molecular basis of the disease and the success of the two small molecule kinase inhibitors, gefitinib and erlotinib, against epidermal growth factor receptor (EGFR) in the treatment of NSCLC has provided evidence for the effectiveness of the strategy.
Many reports have indicated that EGFR tyrosine kinase inhibitors (EGFR-TKI) have significant efficacy in specific subgroups of patients, such as the Asian population, patients with adenocarcinomas, non-smokers and females(1–3) and have indicated that the presence of somatic mutations in EGFR is a strong predictor for both clinical and in vitro sensitivity to EGFR-TKI.(2–18) In terms of the side-effects, interstitial lung disease with gefitinib therapy is a serious problem, especially in the Asian population.(2,3) In randomized phase II trials, there is an increased tendency for interstitial lung disease to develop secondary to gefitinib treatment in Japanese people.(2,3) Therefore, it is important to establish the optimal manner in which to select patients for therapy with EGFR-TKI.
In terms of EGFR ligands, few studies analyzing the expression of EGFR ligands in lung cancer have been reported. Transforming growth factor-α (TGF-α) expression has been detected in 60–80% of primary NSCLC tumors.(19–22) Amphiregulin is also expressed in primary lung carcinoma cells and one study has suggested that its strong expression is correlated with a worse outcome.(23) Ishikawa et al.(24) reported that circulating amphiregulin and TGF-α could be clinically applicable as indicators for an unfavorable response to gefitinib based on the identification of patients with a higher probability of resistance to the drug. They also raised the possibility that positivity of two serum markers might be a prognostic factor for response to treatment.
It is necessary to confirm the role of EGFR somatic mutations and the ligands of the EGFR as a predictor of EGFR-TKI therapy and as a prognostic factor. At the same time, the association between EGFR somatic mutations and the overexpression of its ligands is important. In this report, we confirmed that circulating amphiregulin and TGF-α are suitable indicators for an unfavorable response to gefitinib and that EGFR somatic mutation is a good predictor for gefitinib therapy. The positivity of at least one of the two serum proteins (amphiregulin and TGF-α) and a performance status (PS) of more than 2 is an independent predictor for an unfavorable response to gefitinib. We also demonstrated that circulating TGF-α and an EGFR somatic mutation could be an independent prognostic factor in patients with non-squamous NSCLC.
Materials and Methods
Study population and tissue procurement. Ninety-three patients with advanced non-squamous NSCLC who were treated with gefitinib between December 2002 and September 2007 at Kyoto University Hospital, were consecutively enrolled in this study. The Standard Response Evaluation Criteria in Solid Tumors was used for response evaluation. Trans-bronchial specimens from 44 patients were available for analysis of EGFR somatic mutations. Serum samples were collected from all the subjects before treatment with gefitinib and stored at –80°C. Written informed consent pertaining to the utilization of clinical materials was obtained from all patients. The study was approved by the Ethics Committee of the Kyoto University Graduate School and Faculty of Medicine.
Enzyme-linked immunosorbent assay of amphiregulin and TGF-α. Serum amphiregulin and TGF-α concentrations were measured using commercially available enzyme-linked immunoassays (human Amphiregulin DuoSet ELISA Development System and human TGF-A DuoSet ELISA Development System, respectively; R&D Systems, Minneapolis, MN, USA) according to the manufacturer's instructions in 96-well flexible microtiter plates (Nalge Nunc International, Rochester, NY, USA). The color intensity was determined by a photometer at a wavelength of 450 nm with a reference wavelength of 540 nm. A standard curve was drawn for each plate using recombinant amphiregulin or TGF-α proteins for reference. Minimum detection limits of the assays for serum amphiregulin and TGF-α were 7.81 and 3.91 pg/mL, respectively. Determination of positive or negative reactions was based on measured concentrations of each protein that fell above or below cut-off values, which had been set by drawing receiver-operated characteristic curves according to optimal diagnostic accuracy and likelihood ratios discriminating progressive disease (PD) cases from disease-controlled cases (partial response [PR] + stable disease [SD]) (Fig. 1).
Analysis of EGFR somatic mutation. DNA was extracted from cancer cells obtained from paraffin-embedded biopsy specimens by manual microdissection using a modification of the phenol-chloroform extraction method.(12,14) Briefly, formalin-fixed paraffin-embedded tissues were cut into 5-µm sections and mounted on pretreated glass slides. One slice was used to confirm the presence of cancer cells under microscopy, whereas the others were for DNA analyses. Non-cancer cells and necrotic portions were manually removed from the slide under the microscope. The slides were deparaffinized and DNA was extracted with phenol-chloroform and ethanol precipitation. The peptic nucleic acid-locked nucleic acid polymerase chain reaction (PNA-LNA PCR) clamp, which was designed to detect 11 different EGFR mutations, was used for determination of the EGFR gene mutation status.(25–27) In this study, we adopted the PNA-LNA PCR clamp for analysis of the EGFR somatic mutations. In the PNA-LNA PCR clamp, other types of EGFR mutations are evidenced as an escape of the inhibition of amplification by the clamp primer and, in this case, a direct sequencing method was employed to seek other types of EGFR mutations. The overall sensitivity and specificity of this system using clinical samples is 97% and 100%, respectively.(26) Exons 18, 19 and 21 were amplified by PCR. Exons 18, 19 and 21 were selected for examination because most of the reported mutations known to have an effect on EGFR-TKI therapy in patients with NSCLC have been described in these loci. Exon 20 mutations (insertion and T790M) are reported to be rare in tumor specimens obtained from patients before gefitinib treatment(28) and we did not examine that status. Primers and cycling conditions for PCR amplification were performed by a modification of previously published methods.(12,14) Sequencing reactions were electrophoresed on the ABI PRISM 3100 (Applied Biosystems, Foster City, CA, USA). Direct sequencing of the PCR products was performed in both sense and antisense directions.
Statistical analysis. The univariate relationship between each independent clinicopathological variable, including the EGFR mutation status and serum amphiregulin and/or TGF-α status, and the response to gefitinib therapy was examined using Fisher's exact tests. Categorical variables were assessed with the χ2-test. Variables were selected for further analysis if the probability values were less than 0.05 by univariate analysis. In order to avoid problems of multicolinearity, correlation among predictor variables was analyzed first and then multiple logistic regression analysis was performed. All tests were two-tailed and probability values of 0.05 or less were considered statistically significant.
In order to evaluate risk factors associated with prognosis, a Cox proportional hazards regression model with a step-down procedure was used. Proportional hazards assumptions were checked and satisfied; only those variables with statistically significant results in univariate analysis were included in a multivariate analysis. The criterion for removing a variable was the likelihood ratio statistic, which was based on the maximum partial likelihood estimate (default P-value of 0.05 for removal from the model). Survival time was calculated from the date of diagnosis. Survival curves ware determined using the Kaplan–Meier method. The log–rank test allowed us to evaluate the differences between survival curves. All preceding statistical analyses were performed using JMP ver. 6 software (SAS Institute, Cary, NC, USA).
The characteristics of the 93 enrolled patients are shown in Table 1. All patients were Japanese, and comprised 39 (41.9%) men and 54 (58.1%) women, with a median age of 62 years (range, 27–81) and were treated with 250 mg gefitinib daily. Enrolled patients had either adenocarcinoma (82/93, 88.1%) or unspecified NSCLC (11/93, 11.9%). Forty-five (48.4%) patients were non-smokers and 48 (51.6%) patients were former or current smokers. The Eastern Cooperative Oncology Group (ECOG) PS was 0–1 for 84 patients and 2–3 for nine patients. Seventy (75.3%) patients had been treated with at least two regimens of chemotherapy and 23 (24.7%) patients had received gefitinib therapy as the first-line regimen.
Table 1. Association between non-small cell lung cancer patients’ characteristics and response to gefitinib therapy (n = 93)
Positive or negative serum markers were determined by the cut-off values of 27.0 pg/mL for amphiregulin and 25.0 pg/mL for TGF-α. Positive EGFR somatic mutations include exon 21 L858R, del 746–750, and del 752–759 in exon 19.
The associations between clinicopathological factors and responses to gefitinib therapy are shown in Table 1. Fourteen (41.0%) of 34 PD cases were positive for amphiregulin, whereas 50 (84.7%) of 59 disease-controlled cases were negative (P = 0.007 by Fisher's exact test). Eleven (32.4%) of 34 PD cases were positive for TGF-α, whereas 54 (91.5%) of 59 disease-controlled cases were negative (P = 0.005 by Fisher's exact test). At least one of the two proteins was positive in 18 (52.9%) of 34 PD cases, whereas 47 (79.6%) of 59 disease-controlled cases were negative (P = 0.002 by Fisher's exact test). Twenty-three (79.3%) of 29 patients with EGFR somatic mutations had a disease-controlled response to gefitinib therapy and five (83.3%) of six PD cases with EGFR somatic mutations were positive for at least one of two serum markers. Distant metastases before gefitinib therapy (yes vs no; P = 0.0025 by Fisher's exact test) and PS (≥2 vs 0–1; P = 0.002 by Fisher's exact test) were associated with a response to gefitinib therapy. No correlation existed between response to gefitinib therapy and sex, histological type, disease stage, number of prior chemotherapy regimens, prior thoracic radiation therapy, smoking history or EGFR somatic mutation. Among 16 patients who showed TGF-α positivity, we analyzed the clinicopathological features including sex, age, histlogical type, disease stage, PS and smoking history, but we found no correlations with statistical significance.
Distant metastases before gefitinib therapy were correlated with PS (P = 0.025). A multiple logistic regression analysis revealed that positivity of at least two serum markers and a PS more than 2 were significant indicators of an unfavorable response to gefitinib therapy (Table 2).
Table 2. Multiple logistic regression model analysis of predictive factor with gefitinib therapy
The median survival time of patients with a EGFR somatic mutation treated with gefitinib was significantly longer than those with wild-type EGFR (P = 0.01 by log–rank test; Fig. 2). The same was true of amphiregulin-negative patients compared with amphiregulin-positive patients (P = 0.046 by log–rank test; Fig. 3a) and of TGF-α-negative patients compared with TGF-α-positive patients (P < 0.01 by log–rank test; Fig. 3b). However, with respect to median survival time, there was no statistical significance between patients showing at least one of the two serum markers and patients negative to both of these markers (Fig. 3c). A Cox regression analysis was performed on the 93 patients to determine the correlation between patient prognosis and clinicopathological factors in whom age (≥70 years vs <70 years), sex (female vs male), smoking history (smoker vs never a smoker), PS (0–1 vs 2–3), disease stage (other vs IIIB), EGFR mutation status (wild-type vs positive), serum amphiregulin status (positive vs negative), and serum TGF-α status (positive vs negative) were available (Table 3). Among these factors, the wild-type EGFR gene (hazards ratio [HR], 1.752; 95% confidence interval [CI], 1.176–2.624; P = 0.006), serum amphiregulin (HR, 1.340; 95% CI, 1.176–2.624; P = 0.06), serum TGF-α positivity (HR, 1.580; 95% CI, 1.082–2.193; P = 0.06), poor PS (HR, 1.553; 95% CI, 1.282–2.212; P = 0.04), and advanced disease stage (HR, 1.420; 95% CI, 1.018–2.101; P = 0.038) had a significant prognostic effect on survival based on univariate analysis. Based on multivariate analysis, serum TGF-α positivity (HR, 2.558; 95% CI, 1.348–4.717; P = 0.005) and the wild-type EGFR gene (HR, 1.894; 95% CI, 1.255–2.911; P = 0.003) were significant independent negative prognostic factors for survival.
Table 3. Cox proportional hazards model analysis of prognostic factors in patients with advanced non-small cell lung cancer who were treated with gefitinib
Among 70 patients who received gefitinib therapy after second line setting, we performed the subset analysis. Twelve (50.0%) of 24 PD cases were positive for amphiregulin, whereas 41 (89.1%) of 46 disease-controlled cases were negative (P < 0.001 by Fisher's exact test). Eight (33.3%) of 24 PD cases were positive for TGF-α, whereas 45 (97.8%) of 46 disease-controlled cases were negative (P = 0.017 by Fisher's exact test). At least one of the two proteins was positive in 14 (58.3%) of 24 PD cases, whereas 40 (87.0%) of 46 disease-controlled cases were negative (P < 0.001 by Fisher's exact test). Disease stage (IIIB vs IV; P = 0.046 by Fisher's exact test) was associated with a response to gefitinib therapy. A multiple logistic regression analysis revealed that positivity of at least two serum markers and disease stage were significant indicators of an unfavorable response to gefitinib therapy. The median survival time of patients with an EGFR somatic mutation treated with gefitinib was significantly longer than those with wild-type EGFR (P = 0.04 by log–rank test). The same was true of amphiregulin-negative patients compared to amphiregulin-positive patients (P = 0.035 by log–rank test) and of TGF-α-negative patients compared to TGF-α-positive patients (P < 0.001 by log–rank test). On a Cox regression analysis, the wild-type EGFR gene (HR, 1.908; 95% CI, 1.159–3.207; P = 0.012), serum TGF-α positivity (HR, 2.373; 95% CI, 1.494–3.517; P < 0.001), poor PS (HR, 1.594; 95% CI, 1.013–2.338; P = 0.045), and smoking history (HR, 1.360; 95% CI, 1.002–1.860; P = 0.049) had a significant prognostic effect on survival based on univariate analysis. Based on multivariate analysis, the wild-type EGFR gene (HR, 1.750; 95% CI, 1.011–3.135; P = 0.045) was a significant independent negative prognostic factor for survival. Serum TGF-α positivity could not be analyzed because the number was very small (n = 9).
Many reports have indicated that EGFR-TKI have significant efficacy in specific subgroups, such as patients with adenocarcinomas, non-smokers and females.(1–3) EGFR mutation status is thought to be the strongest predictive factor for response to EGFR-TKI in patients with NSCLC.(29) Lynch et al.(12) reported that clinical responsiveness to EGFR-TKI was correlated with specific mutations in the EGFR gene in a subgroup of patients with NSCLC. Although the potential usefulness of EGFR somatic mutational analysis has been reported, analysis of expression profiles or mutations requires acquisition of tissue specimens by surgery or biopsy, which is not routine in cases of advanced NSCLC, and sometimes these procedures themselves cause various complications. Practical clinical tests using serological markers that can predict the sensitivity or resistance of lung cancers to EGFR-TKI therapy are urgently required. Moreover, as a predictive factor for EGFR-TKI therapy, EGFR somatic mutations may only be useful for discriminating super-responders and/or patients who show PR. EGFR somatic mutations might not be good predictive markers for identifying patients with SD, which is also considered to be beneficial in the management of many solid tumors.(30) It is important to establish the optimal method by which to select patients for therapy with EGFR-TKI.
There are fewer studies analyzing the expression of EGFR ligands in lung cancer. TGF-α expression has been detected in 60–80% of primary NSCLC tumors.(19,20,22,31) Amphiregulin is also expressed in primary lung carcinoma and one study suggested that it is correlated with a worse outcome.(24) It has been shown that genes encoding two EGFR ligands, amphiregulin and TGF-α, are among those overexpressed in tumor tissues obtained from non-responders to gefitinib.(23) Thus, we selected these two serum markers as candidate surrogate markers for gefitinib therapy. In Ishikawa's report(24) there is reference to an association between EGFR somatic mutations and the overexpression of its ligands; however, there are no reports that revealed definitive results.
In this study, serum amphiregulin and TGF-α positivity in non-responders to gefitinib therapy was 14 of 34 (41.1%) and 11 of 34 (32.4%), respectively; a combination of the two serum markers improved the sensitivity for discriminating patients with PD from patients with disease-control to 52.3%. We also revealed that both positivity of serum TGF-α and wild-type of EGFR independently affect the survival of patients with non-squamous NSCLC, findings which are consistent with a previous report.(24) TGF-α specifically binds and activates EGFR and was reported to have oncogenic effects in many cancer cell types. Several EGFR ligands including TGF-α and amphiregulin are synthesized as membrane-anchored precursor forms that are later cleaved to generate soluble ligands in a metalloproteinase-dependent manner.(32) Moreover, in certain circumstances, the membrane-anchored isoforms of tumor cells may also act as biologically active ligands. Thus, activation of the EGFR by these ligands can occur through autocrine, paracrine and juxtacrine mechanisms. Besides tumor cells themselves, amphiregulin and TGF-α released from pulmonary fibroblasts were reported to stimulate the growth of lung adenocarcinoma.(33) Serum concentration of the ligands may represent the total intensity of activation signal for EGFR and tumorigenic capacity of existing neoplasms.
In this study, five (83.3%) of six PD cases with EGFR somatic mutations were positive for at least one of two serum markers. There is a possibility that the ligands may compete with EGFR-TKI at the cell-surface EGFR molecules and this observation may offer an explanation for primary PD cases (not acquired resistance) with EGFR somatic mutations who are treated with EGFR-TKI. For these cases, increasing the dose of EGFR-TKI or administration of ‘decoy’ molecules which selectively bind to the excessive ligands and decrease stimuli for membrane-anchored EGFR may be an alternative treatment option.
Limitations of our study include a small sample size, heterogeneity of treatment regimens and the retrospective nature of the study. We were able to examine the association between concentration of the ligands and EGFR mutation status in only 46 patients. The difference in the prognostic power of amphiregulin shown in Fig. 3a and Table 3 (P = 0.046 vs P = 0.06) seems to demonstrate the limitation of sample size in this pooled survival analysis of various patients with different clinical backgrounds. Only the prospective study with more strict criteria for the selection of cases could settle these limitations.
In conclusion, we demonstrated that the status of serum amphiregulin and TGF-α, and an EGFR somatic mutation, are predictive factors for gefitinib therapy. In addition, we revealed that both positivity of serum TGF-α and wild-type EGFR were independent poor prognostic factors in patients with non-squamous NSCLC. To determine whether these two clinical markers are predictive markers for gefitinib therapy or independent prognostic factors and to definitively demonstrate the association between EGFR somatic mutations and its ligands, we should evaluate EGFR somatic mutations, and serum amphiregulin and TGF-α before treatment with EGFR-TKI in a randomized clinical study.