In the present study we address two essential questions: first, the incidence of activating EGFR mutations in NSCLC metastases to the CNS, and second, the level of concordance between highly sensitive molecular methods routinely used for EGFR molecular diagnostics in primary lung tumours.
Incidence of EGFR mutations in NSCLC brain metastases
Brain metastases are one of the most frequent complications of lung cancer and are associated with significant morbidity and mortality [
11‐
13]. Despite this, published data concerning the
EGFR mutation status of metastatic tumours and corresponding primary lung cancers are limited, particularly in the Caucasian population.
There are several studies evaluating the presence of
EGFR mutations in CNS lung cancer metastases in the Asian population [
14,
15]. Matsumoto et al. [
14], examined 21 metastatic brain tumours from 19 NSCLC patients (68 % smokers) and eight samples from corresponding primary lung tumours.
EGFR mutations were detected in CNS samples from 63 % (12/19) of NSCLC patients, including ten short-in frame deletions in exon 19 and two L858R substitutions in exon 21. In six cases, mutations were identical to those detected in the corresponding primary tumour, while two mutations identified in primary tumours were not consistent with mutations detected in metastatic tumours [
14]. In another study, Han et al. [
15], observed a 60 % incidence of
EGFR mutations in NSCLC brain metastases; however, this study was based on a small patient cohort including five NSCLC patients with primary and corresponding brain tumours. In one patient, the L858R substitution present in the primary tumour was not detected in a corresponding metastatic brain sample [
15]. In both studies, the frequency of
EGFR mutations in CNS metastases was typical for the Asian population; however, these observations need to be verified in larger patient cohorts.
In the Caucasian population, the percentage of patients with
EGFR mutations in NSCLC primary tumours is lower compared with Asian patients (approximately 10–16 %) [
2,
16]. A study conducted by Munfus-McCray et al. [
17], demonstrated a 40 % incidence of activating
EGFR mutations in NSCLC brain metastases; however, this study also involved a small cohort (ten examined patients). Grommes et al. also reported a study involving treatment of a very small group of patients (
n = 9) treated with pulsatile, high-dose erlotinib with CNS metastases and with
EGFR mutation diagnosed outside of CNS metastases. A partial response of CNS metastases was observed in six patients. Corresponding tissue from the CNS metastases was available for four patients with response after tyrosine kinase inhibitor (TKI)-EGFR therapy diagnosed with
EGFR mutation matching to those diagnosed outside the CNS metastases (three L858R substitution and one deletion in exon 19) [
18].
In the present study, CNS metastases from 143 NSCLC Caucasian patients were examined. We observed activating
EGFR mutations in 6.29 % of patients. Importantly, complete compliance between
EGFR mutational status of 32 corresponding primary tumours and brain metastases was observed. Although the calculated incidence of
EGFR mutations in CNS metastases is lower than previously reported, we are unable to attribute this to technical difficulties since this analysis was performed by two laboratories routinely performing NSCLC molecular diagnostics and involved a large patient cohort. In support of our findings, studies by Lublin and Poznan and Warsaw [
19] identified exon 19 and 21
EGFR mutations in 10.5 % (
n = 460) and 9.11 % (
n = 384) of NSCLC samples, respectively. These results are also compatible with a recent meta-analysis including six randomized studies with a total of 2,797 Caucasian patients with NSCLC (not exclusively lung adenocarcinoma), where the estimated frequency of
EGFR mutations (exon 19 or 21) was 12.98 %. [
20]. The discrepancy between these studies and those reporting higher percentages of
EGFR-positive patients (>10 %) may be due to the pre-selection of patients based on clinical factors (e.g. histopathological diagnosis, smoking status or qualification for TKI-EGFR therapy). Since our study was not based on a pre-selected patient cohort (with the exception of tumour tissue accessibility), this may account for the lower percentage of
EGFR-mutated patients. It should be noted, however, that the patient characteristics in our study are not entirely representative of a non-selected NSCLC population in other European countries. The low frequency of patients with adenocarcinoma in our study likely accounts for the low percentage of
EGFR mutations detected in our group. Alternatively, this may be due to the high percentage of patients within the NSCLC-NOS pathological category (first patients were treated surgically in 2003), as a result of the retrospective nature of the study and since we did not utilize immunohistochemistry antibodies. It should be noted that no
EGFR mutations were detected in patients with NSCLC-NOS histology, perhaps owing to a lack of adenocarcinoma patients in this group. In addition, our population included a very high percentage of patients with present or past smoking history, characteristic for Polish NSCLC patient populations, which may also account for the low percentage of detectable
EGFR mutations.
To the best of our knowledge, these are the only publications concerning the frequency of
EGFR gene mutations in NSCLC CNS metastases. However, a study by Togashii et al. revealed that 50 % (11/22) of patients with
EGFR gene mutations were also diagnosed with different distant metastases. Moreover, metastasis was diagnosed much less frequently (12 %) in cases of lung adenocarcinoma with wt
EGFR [
21]. Studies by Sun et al. assessed the status of
EGFR and
KRAS genes in a cohort of 80 NSCLC patients for whom material from both the primary tumours and the lymph node metastases was available.
EGFR gene mutations were identified in 21 primary tumours and 26 lymph node metastases, with mutations in primary tumours confirmed in metastases in all cases [
22]. Taken together, the role of
EGFR gene mutations in the occurrence of distant metastases remains controversial.
To date, molecular diagnostics and lung cancer staging are predominantly performed using histological or cytological material [
23‐
25]. Consequently, the quantity of samples is often limited, with a cancer cell percentage below 50 %, and the DNA yield is correspondingly low. Low cancer cell content is an important issue, since the minimal requirement for accurate detection may be as high as 50 % for Sanger sequencing. Previously, we demonstrated that the median concentration of DNA isolated from intrabronchial forceps biopsy is 38.3 ng/μl [
19]. However, commercially available in vitro diagnostic real-time PCR-based tests (CE-IVD) specifically designed for of the detection of
EGFR activating mutations are not validated to analyse samples with less than 150–800 ng of DNA or 10 % of neoplastic cells [
2,
5,
25,
26]. Thus, the development of highly sensitive molecular methods appropriate for more technically demanding samples has become a major focus in lung cancer diagnostics. Techniques based on allele-specific amplification or on the inhibition of wt gene amplification and the simultaneous enhancement of mutated gene amplification have proven particularly useful owing to the high specificity, relative simplicity and cost effectiveness [
26‐
28].
Both allele-specific methods utilized in the present study (PCR followed by DNA fragment analysis and ASP–PCR) demonstrated high detection sensitivity. Previous analyses by Pan et al. utilized an assay to detect exon 19 mutations based on length analysis of fluorescently labelled PCR products. Deletion of exon 19 was readily detected in 6.25 % of DNA from H1650 cells [
8]. However, Dahse et al. [
28], were able to detect the mutant exon 21 T allele in a mixed sample containing a four fold excess of normal DNA, using an allele-specific PCR for L858R in exon 21.
In our study, the PNA–LNA PCR clamp technique, which inhibits wt gene amplification and simultaneously enhances amplification of the mutated allele, achieved very high sensitivity (1 % of tumour cells for both exons), in accordance with other reports [
6,
9,
24]. In an experimental setting, PNA–LNA PCR clamp not only clearly identified mutated alleles intermixed as 1 % of the normal human diploid genome, but also detected one mutant allele in 1,000 diploid human genomes (i.e. 0.1 %) [
9]. The reliability of PNA–LNA PCR clamp has been also confirmed in clinical settings, with high sensitivity (97 %) and specificity (100 %) demonstrated in variety of cytological specimens (bronchoscopy samples, sputum, pleural and pericardial effusion) in addition to paraffin-embedded tissues [
1,
24,
26]. Accordingly, Yamada et al. [
24], demonstrated that the PNA–LNA PCR clamp method allowed positive diagnosis in 33.6 % of 122 cytological samples from Asian NSCLC patients . Studies by Ikeda et al. [
26], compared the effectiveness of several highly sensitive PCR methods (ME-PCR, PNA–LNA PCR clamp and PCR invader) to detect
EGFR mutations in paraffin-embedded tumour sections, frozen cytology specimens obtained by bronchoscopy (washing and brushing) or from malignant pleural effusions. These studies revealed that all methods displayed similar sensitivity, and activating
EGFR mutations were detected in 28 % (14/50 samples) in a cohort of Asian patients with advanced NSCLC [
26].
To our knowledge, this study is the first to compare the consistency of highly sensitive methods in the molecular analysis of intracranial NSCLC metastases. Conflicting results were observed in three of 143 patients evaluated. Since the quantity of specimens available for diagnostic evaluation was generally low, these reported discrepancies were likely due to low material quality. As previously mentioned, pre-amplification of DNA using nested primers was performed owing to DNA fragmentation or low DNA concentration in 37 brain samples. Based on experience with both methods, which are routinely used in our laboratories for NSCLC molecular diagnostics, as well as assumptions based on methodological differences, we hypothesize that PNA–LNA PCR clamp may be more effective in samples with very low tumour cell number, while ASP–PCR may be more sensitive in samples with fragmented DNA.
Our findings and those of other groups, particularly Ikeda et al. [
26], provide a rationale for applying at least two molecular techniques in the routine diagnostics of difficult, low-volume or low-quality NSCLC samples, both from primary tumour or metastases. We believe that the use of substantially different methods may allow more consistent results and verification of negative results. Accordingly, discrepant results provided by highly sensitive and specific molecular methods should be rather accepted as true positive rather than false negative results, as exemplified in the Ikeda study. Consequently, we are inclined to recognize the three discrepant results reported in our study as true positives.
The sensitivity of molecular techniques used for the detection of
EGFR gene mutations is a critical factor in NSCLC diagnosis and subsequent treatment, since the results of these tests may affect qualification for TKI-EGFR-based therapy and the effectiveness of such therapies. Techniques with low sensitivity may lead to disqualification from TKI-EGFR therapy in patients harbouring
EGFR mutations. Conversely, techniques that are too sensitive may lead to the detection of mutations in rare cell clones within heterogeneous tumours. A study by Kim et al. [
29], showed that progression after TKI-EGFR therapy occurs significantly less frequently in patients when
EGFR mutations are detected by two different techniques (direct sequencing and PNA–LNA PCR clamp), compared with only one method (PNA–LNA PCR clamp, 11.5 vs 22.7 %) .
In conclusion, our analysis of EGFR mutations in a homogenous group of 143 Caucasian patients with NSCLC demonstrates that activating EGFR mutations are present in 6.29 % of patients, and include exon 19 mutations (2.1 %) and exon 21 mutations (4.2 %). We demonstrate that detection of EGFR mutations in NSCLC brain metastases is feasible using highly specific molecular techniques. However, the use of at least two independent molecular methods will ensure a more accurate identification of EGFR mutations.