The impact of MET, IGF-1, IGF1R expression and EGFR mutations on survival of patients with non-small-cell lung cancer

Introduction To compare the efficacy of silver in situ hybridization (SISH) and immunohistochemistry (IHC) in detecting MET and IGF1R alterations and to investigate their prevalence and prognostic significance. A possible correlation between MET receptor expression, MET gene alterations and the two most frequent occurring EGFR gene mutations was also investigated. Materials and methods Stage I to IIIA tumors from 326 patients with NSCLC were immunohistochemically tested for protein expression of MET and IGF-1. Their cytoplasmic expression was compared with the gene copy number of the MET and IGF1Rgenes by SISH in paraffin-embedded, formalin-fixed material. Correlations were made with the immunohistochemical expression of two frequent EGFR mutations and clinicopathological variables. Univariate and multivariate survival analyses was used to evaluate the prognostic efficacy of the tested markers. Results In univariate analyses, high cytoplasmic MET expression showed a significant negative prognostic effect in adenocarcinoma patients (p = 0.026). MET gene to chromosome 7 ratio was a significant positive prognostic marker (p = 0.005), probably only due to the highly negative prognostic significance of chromosome 7 polysomy (p = 0.002). High IGF1R gene copy number was a negative prognostic marker for all NSCLC patients (p = 0.037). In the multivariate analysis, polysomy of chromosome 7 in tumor cells correlated significantly and independently with a poor prognosis (p = 0.011). In patients with adenocarcinoma, a high cytoplasmic MET expression was an independent negative prognostic marker (p = 0.013). In males a high IGF1R gene copy number to chromosome 15 count ratio was significantly and independently correlated to a poor prognosis (p = 0.018). Conclusion MET protein expression provides superior prognostic information compared with SISH. Polysomy of chromosome 7 is an independent negative prognostic factor in NSCLC patients. This finding has an important implication while examining genes located on chromosome 7 by means of SISH. High IGF1R gene copy number to chromosome 15 count ratio is an independent predictor of inferior survival in male patients with primary NSCLC.


Introduction
To compare the efficacy of silver in situ hybridization (SISH) and immunohistochemistry (IHC) in detecting MET and IGF1R alterations and to investigate their prevalence and prognostic significance. A possible correlation between MET receptor expression, MET gene alterations and the two most frequent occurring EGFR gene mutations was also investigated.

Materials and methods
Stage I to IIIA tumors from 326 patients with NSCLC were immunohistochemically tested for protein expression of MET and IGF-1. Their cytoplasmic expression was compared with the gene copy number of the MET and IGF1Rgenes by SISH in paraffin-embedded, formalinfixed material. Correlations were made with the immunohistochemical expression of two frequent EGFR mutations and clinicopathological variables. Univariate and multivariate survival analyses was used to evaluate the prognostic efficacy of the tested markers.

Results
In univariate analyses, high cytoplasmic MET expression showed a significant negative prognostic effect in adenocarcinoma patients (p = 0.026). MET gene to chromosome 7 ratio was a significant positive prognostic marker (p = 0.005), probably only due to the highly negative prognostic significance of chromosome 7 polysomy (p = 0.002). High IGF1R gene copy number was a negative prognostic marker for all NSCLC patients (p = 0.037). In the multivariate analysis, polysomy of chromosome 7 in tumor cells correlated significantly and independently with a poor prognosis (p = 0.011). In patients with adenocarcinoma, a high cytoplasmic MET expression was an independent negative prognostic marker (p = 0.013). In males a high IGF1R gene copy number to chromosome 15 count ratio was significantly and independently correlated to a poor prognosis (p = 0.018). PLOS (3) inadequate paraffin-embedded tissue blocks. Thirty-six patients fell into these three categories (criteria 1: n = 10; criteria 2: n = 13; criteria 3: n = 13) and were excluded from the study. Adjuvant chemotherapy had not yet been introduced as a therapeutic option in Norway during this time span (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004). In total, 335 patients with complete medical records and adequate paraffin embedded tissue blocks were included in this study. The tumors were subtyped and histologically graded according to the recent World Health Organization (WHO) guidelines. [14] The patients were staged corresponding to the 7th edition of the UICC TNM classification, where 9 patients were regarded as having an in-situ disease regarding the new lung cancer classification resulting in a total of 326 patients eligible for this study. [15] The Regional Committee for Medical and Health Research Ethics, as well as the Norwegian Data Inspectorate, approved this study.

Microarray constructions
Two experienced pathologists (S.A.S. and K.A.S.) investigated all the lung cancer specimens thoroughly. Tissue microarray (TMA) blocks were constructed using a tissue-array instrument (Beecher Instruments, Silver Springs, MD, USA) as previously described [16].

Immunohistochemistry and silver in situ hybridization
The applied antibodies have been previously subjected to in-house validation by the manufacturer for IHC analysis of paraffin-embedded material. The antibodies used in this study were as follows:  [16,17].
Scoring of immunohistochemistry (IHC) and silver in situ hybridization (SISH) The tissue cores were scored by light microscopy to determine the degree of cytoplasmic and nuclear expression. Examples of various markers' expressions are shown (Fig 2). Staining for bound protein 2 (GRB2). The MET pathway is modulated by cell surface molecules, including the HER family (HER1, HER2 and HER3) and IGF1R.MET can activate the effector molecule phosphatidylinositol 3-kinase (PI3K), and signals through the AKT/protein kinase B axis, which activates the mammalian target of rapamycin (mTOR) axis stimulating cell growth and protein synthesis. Moreover, the activation of either PI3K-AKT can activate NF-κB, which then can be released and stimulate gene transcription after being translocated to the nucleus.MET activation can result in the down-stream activation of the RAS-MAPK pathway. The nucleotide exchanger protein Son of Sevenless (SOS) activates the rat sarcoma viral oncogene homolog RAS by binding with the GRB2 (GRB2-SOS complex). This complex can activate the v-raf murine sarcoma viral oncogene homolog B1 (RAF) kinases, which successively activate MAPK effector kinase (MEK), and finally results in the activation of the mitogen-activated protein kinase (MAPK).Activation of MET cytoplasmic signalling cascades can additionally alter cell invasiveness, motility, and cytoskeleton, mediated through the RAS-related protein 1 (RAP1), the focal adhesion complex (FAC) as well as integrin connections. MET can additionally activate the downstream axis of the Signal transducer and activator of transcription 3 (STAT3) pathway. STAT3 can be activated through phosphorylation after binding directly to MET resulting in dimerization and translocation to the nucleus and consequently mediating cell proliferation, transformation, as well as tumorigenesis and invasion. [6][7][8][9]. MET and IGF1R genes resulted in signals as black dots on the corresponding chromosomes for both genes, while centromeres of chromosome 7 and 15 were stained as red dots. Regarding SISH scoring, uniform guidelines exist and were strictly followed for the interpretation of gene and chromosome signals. [18]Even though these guidelines were developed for breast cancer testing, we found similar staining results in NSCLC biopsies. Evaluation ofHER2sish is reported in breast cancer as the ratio of the average number of HER2 gene copies to the average number of chromosome 17 copies (HER2:chr17) per cell. Because no clear guidelines have been established for measuring MET or IGF1Rgene amplification in NSCLC, we also sought to determine whether the absolute number of MET and IGF1Rgene copies detected by SISH (i.e., the number of black dots observed in the nuclei of tumor cells) would add prognostic significance beyond that established by the gene copy number to chromosome count ratio. From each tumor, four cores were eligible for scoring. In each core, we counted genes and centromere signals in 20 cells at least in two cores, where one core was taken from the central part of the tumor and the other core was taken at the advancing edge of tumor. An overall average was taken for both gene and centromere count. The other two cores included for the most stromal tissue surrounding epithelial cells of NSCLC. Heterogeneity was not observed while scoring immunostains. Regarding SISH, in cases of heterogeneity hot spots with the highest gene or centromere count were scored. The number of gene copies was assessed according to the manufacturer's protocols for INFORMHER2 DNA. Briefly, a discrete dot was counted as a single copy of MET, IGF1R, chromosome 7 or chromosome 15. Some nuclei showed multiple discrete copies. Clusters of dots representing many copies of the targetgenes were also apparent; a small cluster of multiple signals was counted as six copies and a large cluster was counted as 12 copies. Staining for MET, IGF-1, EGFRmut and EGFRdel resulted in a homogenous cytoplasmic staining (Fig 2). Due to homogenous staining, there was no need to score the density of positive cells. Scoring the intensity of staining was considered as representative to measure the protein expression by immunohistochemical staining. Tissue sections for the two first mentioned proteins were scored semi quantitatively for staining intensity in tumor epithelial cells as follows: 0 = negative, 1 = weak, 2 = intermediate, and 3 = strong. Regarding the immunohistochemical expression for IGF-1 and MET, tumors showing intermediate or strong positivity were regarded as the high expression group, while tumors with weak or negative staining were regarded as the low expression group. When evaluating EGFRmut and EGFRdel, only clear strong positive tumors where regarded as harboring gene mutations.
All of the anonymized samples were semi quantitatively and independently scored by two experienced pathologists (S.A.S. and E.R). In the event of disagreement, the slides were re-examined and a consensus was reached by the observers. When assessing one variable for a given core, the observers were blinded to the scores of the other variables and the outcome. The reproducibility of the IHC and SISH evaluation in randomly selected specimens is high. The IHC and SISH scores from each observer were compared for interobserver reliability using a 2-way random effects model with absolute agreement definition, yielding an intra-class correlation coefficient (reliability coefficient) and Cohen's kappa. There was an excellent scoring agreement for two tested markers (MET SISH and EGFRdel) between the pathologists, with an intra-class correlation coefficient of 0.91 (P < 0.001) for MET SISH and 0.93 (P < 0.001) for EGFRdel.

Statistical analyses
The statistical analyses presented in this study were performed using the statistical package IBM SPSS, version 24 (SPSS Inc., Chicago, IL, USA).Chi-square and Fisher's exact test were used to examine the correlation among different molecular markers and clinicopathological factors. The r-values represent Spearman's rank correlation coefficients. The Kaplan-Meier method was used for drawing curves for univariate analysis of the association between marker expression and disease-specific survival (DSS). DSS was determined from the date of surgery until the time of lung cancer death. Statistical significance between the survival curves was assessed utilizing the log-rank test. Cut-offs were chosen by a minimal p-value approach, with regard to the association between markers and survival endpoints. The survival curves were terminated at 120 months as fewer than 10% of patients were at risk after this point. Statistically significant variables from the univariate analysis were included in the multivariate analysis, applying the Cox proportional hazards model. The data were run in a backward stepwise Cox regression with a probability for stepwise entry and a removal set at 0.05 and 0.10. The significance level was set at a P-value less than 0.05.

Clinicopathological variables
We retrospectively examined a non-selected group of NSCLC patients, which was examined in previous works by our research group. [17]Clinical, demographic and histopathological variables are presented in Table 1.Of the 326 NSCLC patients, the majority were male (76%) and nearly all (96%) were previous or present smokers. The median age was 67 years (range 28-85 years) and the median follow-up of survivors was 105 months (range 73-234 months). Histologically subtypes presented as: (58.6%) 191 cases of squamous cell carcinoma (SCCs), (31.9%) 104 cases of adenocarcinomas (ACs) and (9.5%) 31 cases of large cell (anaplastic) carcinomas (LCCs).Fifty-nine patients (18%) were administered adjuvant radiotherapy due to nodal metastasis or non-radical surgical margins verified during surgery. The median follow-up of the survivors was 105 months (range 73-234 months).

Biomarker expression and correlation in NSCLC tissue
Of 326 cases, all were immunohistochemically evaluable for the expression of MET and IGF-1, while 295 were evaluable for the silver in situ hybridization analysis of the MET gene copy number on chromosome 7, and 237 cases were eligible for the IGF1R gene copy number analysis on chromosome 15.
For the immunohistochemical analysis of EGFRgene mutations, 313 cases were eligible for EGFRmut, while 318 cases were eligible for the detection of the EGFRdel mutation.
MET, IGF-1, EGFRmut and EGFRdel showed a homogenous cytoplasmic staining pattern. We found high (moderate to strong) cytoplasmic expression of MET and IGF-1 in 82.8% and 6.1% of valid tumor samples, respectively. The reported prevalence of phosphorylated MET in our material is consistent with results demonstrated by other investigators. [19][20][21]. However, a lower grade of prevalence has been observed by other reports [22][23][24]. Using various methods to detect the prevalence of phosphorylated MET in FFPE tissue, Dua et al [25] demonstrated that their c-MET FFPE assay could detect and quantify c-MET receptor levels in FFPE tumor specimens, and that these measurements would correlate well with measurements obtained by conventional methods.
A high chromosome 7 count, higher than 2 copies was found in 21.5% of valid tumor samples, while 6% of tumor samples showed polysomy, i.e. a chromosome 7 count higher than 3. A MET gene to chromosome 7 ratio higher than 1 was observed in 7.1% of valid tumors, while an IGF1R gene to chromosome 15 ratio higher than 1 was observed in 21.6% of valid tumors. In the male cohort population, 6.1% of valid tumors showed an IGF1R gene to chromosome 15 ratio higher than 1, p = 0.015, while this was observed in 15.5% of all females with NSCLC, p = 0.021. An EGFRdel mutated protein indicating a gene mutation was found in 6% of all valid patients (5.8% of males and 6.6% of women) while 5.4% showed an EGFRmut mutated protein, indicating a mutated gene (4.6% of males and 8% of women). Prevalence of investigated factors

Univariate analysis
Results from the univariate analysis regarding the clinical variables and their impact on DSS are presented in Table 2. T-stage (P<0.001), N-stage (P<0.001), pathological stage (P<0.001), WHO performance status (p = 0.016), histology (P = 0.028), vascular infiltration (P = 0.001), differentiation (p<0.001) and surgical procedure (p = 0.007) were significant prognosticators for the total patient population. High cytoplasmic MET expression showed a significant negative prognostic effect only in patients with adenocarcinoma (p = 0.026), but not for the whole cohort (p = 0.411; Table 3 and Fig 3).MET gen copy count to chromosome 7 ratio>1 was a significant positive prognostic marker (p = 0.005). We sought to determine if different cut-off points for the MET gene/chromosome 7 ratio would provide an additional prognostic significance. A MET gene/chromosome 7 ratio >1.5 (p = 0.21) did not show a prognostic significance. The same applies for a MET gene/chromosome 7 ratio >2 (p = 0.43). However, the polysomy of chromosome 7 emerged as a highly specific (p = 0.002) negative prognosticator for all patients.IGF1Rgene copy number was a highly negative prognostic marker for all NSCLC patients (p = 0.037), but was even more significant in males (p = 0.015) than in females (p = 0.021).Finally, a high IGF-1 expression showed a trend as a negative prognostic marker in males (p = 0.053) but not for the whole cohort population. There was no significant correlation between DSS and tumor epithelial cell expression of the EGFR gene mutations EGFRmut (p = 0.628) and EGFRdel (p = 0.498).

Multivariate Cox proportional hazards analysis
Significant clinic pathological and molecular variables from the univariate analyses were entered into the multivariate analysis. The statistically significant results are presented in Table 3.
Polysomy of chromosome 7 in tumor cells correlated significantly and independently with a poor prognosis (HR: 2.29; 95% CI: 1.21-4.35; p = 0.011). In patients with adenocarcinoma, a high cytoplasmic MET expression was an independent negative prognostic factor (HR: 2.85; 95% CI: 1.25-6.50; p = 0.013). In males with NSCLC a high IGF1R gene copy number to chromosome 15 count ratio higher than one, was significantly and independently correlated to a poor prognosis (HR: 2.67; 95% CI: 1.18-6.00; p = 0.018). For the whole cohort, the histologic subtypes did not reach a statistically prognostic significance, neither for the group of adenocarcinoma (p = 0.165), nor for the group of squamous cell carcinoma (p = 0.7) or for the large cell carcinoma group (p = 0.466).

Discussion
The chief aim of our study was to investigate the prognostic role of MET expression and the MET gene copy number gain in NSCLC. In addition to IHC we used SISH to investigate the METgene. While, in the adenocarcinoma patient population, independent of other clinicopathological variables, a high cytoplasmic MET expression was a significant negative prognosticator, as determined by IHC, a similar correlation was not found in patients with MET gene copy number gain, i.e. the absolute MET gene copy number in tumor cells did not affect the prognosis. Surprisingly, investigating the ratio of MET gene to chromosome 7, we found a trend of better survival in patients with a higher ratio in the whole cohort in the univariate analysis, but not in the multivariate analysis. However, a high MET gene to chromosome 7 ratio appeared to be a rather non-frequent event. We observed a MET gene to chromosome 7 ratio higher than 1 in 7.1% of our non-selected patients. This is in agreement with recent published data by Noonan et al [26], where-using fluorescence in situ hybridization in lung adenocarcinoma-they observed a MET gene to chromosome 7 ratio of 1.8 or higher only in 4.5% of adenocarcinoma patients. The rather confusing positive prognostic effect of a high MET gene to chromosome 7 ratio led us to investigate a possible role of chromosome 7 polysomy in NSCLC patients. Interestingly, we found a high chromosome 7 count, higher than three chromosome copies as a highly independent negative prognostic factor for the whole cohort population. We assumed that a dysfunction in the transcriptional or posttranscriptional controlling mechanisms could partly explain the discrepancies between the IHC and SISH results; still we wanted to explore other genes located on chromosome 7. Chromosome 7 is known to harbor genes whose alteration my play an important role in multiple diseases as cystic fibrosis [27], but also in tumorigenesis [28]with over than 1150 protein-coding genes, 605 of which have been validated by transcript sequences.
[29] Nevertheless, three genes; EGFR, MET, and BRAF emerge as of special interest in NSCLC. Both MET and EGFR gene amplification are described to have a critical predictive role in NSCLC.
[30]BRAF appears to have a role as a predictive  MET, IGF-1, IGF1R and EGFR mutations prognosis in lung cancer proposing a mechanism for MET and EGFR axis regulation mediated by miRNAs. Additionally, MET protein activation has been associated with primary resistance to EGFR tyrosine kinase inhibitor (TKI) therapy in NSCLC patients (Fig 1). [13,37] These observations have initiated a scientific debate about novel bispecific EGFR/MET inhibitors to obtain better therapeutic results. [10,38,39]In the light of this, we sought to determine any correlation between MET expression, MET gene copy number count (GCNC) and two of the most observed EGFR gene mutations in NSCLC; EGFRmut and EGFRdel [40,41]in untreated patients. A critical issue in investigating EGFR mutations using immunohistochemistry was finding antibodies with an acceptable sensitivity and specificity. The specificity of the used antibodies has been certified by the manufacturer both by means of Western blot and flow cytometric analysis [42,43] Additionally, Several groups [44][45][46] have concluded with an acceptable specificity and sensitivity of the aforementioned antibodies. Brevet et al. [47] reported that IHC using the EGFR L858R specific antibody showed a sensitivity of 95.2% and a specificity of 98.8%. They further found that the EGFR exon 19 mutant specific antibody would detect 100% of 15-bp (base pair) deletions with a high specificity, however, a significant lower sensitivity of about 48.6% in non-15-bp exon 19 deletions was observed. According to the COSMIC database, Non 15-bp exon 19 deletions account for about 35% of exon 19 deletions [48]. We did not find a statistic correlation between MET expression, MET GCNC and any of the examined EGFR gene mutations. Further, none of the investigated EGFR mutations had a prognostic significance, despite conflicting results by other groups[40, 49,50].
The essential role of the HGF-MET cellular pathway has been further established following observations on MET-null mutant mice embryos, with malformation of liver, placenta, [7] melanocytes, [51] and testis. [52] Meanwhile, it has been observed that overexpression of MET can have an oncogenic potential by itself and can induce hepatocellular carcinoma in liver cells [53]. There is established evidence [54] that MET is sufficient for transformation of normal human osteoblasts causing an osteosarcoma-like disease in vivo. Furthermore, dysregulation of the HGF-MET pathway has been demonstrated in malignancies of epithelial cell origin, represented by carcinomas of the lung, mamma, hepatic cells, pancreas ovaries, papillary renal carcinoma, papillary thyroid carcinoma, and carcinomas of the colorectal system. [6]Dysregulation of the HGF-MET cellular axis may due to MET gene mutations, MET amplification, chromosomal rearrangement, MET transcriptional upregulation or changes in the autocrine or paracrine signaling. Several studies have investigated the prognostic role of the MET receptor and the MET gene alteration in NSCLC. While some groups investigated a rather small group of patients [55][56][57], other groups found a negative prognostic effect of high MET protein expression and MET gene copy number gain either independent of the histologic type [58][59][60][61][62][63] or only in patients with adenocarcinoma [64] or squamous cell carcinoma [65].
Exploring both the MET protein expression and the MET GCNC in 140 NSCLC patients, Dziadziuszko et al [66] concluded that neither was associated with prognosis. Meanwhile, Tran et al [67] observed that MET overexpression and MET high (GCNC) occur in a low proportion of primary NSCLCs and are associated with a good prognosis. Awad et al [68]reported MET exon 14 mutations to occur in rather older patients and that they may represent a clinically unique molecular subtype and a possible important therapeutical target in NSCLC. Preliminary findings from the PROFILE 1001 trial [69] show that crizotinib demonstrates a meaningful antitumor activity in patients with NSCLC harboring MET exon 14 alterations. Similar results were also recently published by Lu et al [70].
Recent studies have proposed a ligand-independent MET activation. [71,72] Using prostate cancer cell lines, Varkaris et al, [73] proposed a full but rather delayed activation of MET through IGF1R (Fig 1). Consequently, we aimed to determine a possible correlation between the MET receptor expression, MET GCNC, and the IGF-1 axis.
However, investigating the prognostic significance of IGF-1 expression in epithelial NSCLC showed a trend towards worse survival. Furthermore, investigating the ratio of IGF1R GCNC to chromosome 15, this ratio immerged as a highly significant and independent negative prognostic indicator for disease specific survival in males, whereas it did not show prognostic influence in females. Gender differences in cancer tumorigenesis and survival are most likely associated with different sex hormone effects on various genes. While there are promising preliminary results among breast cancer and NSCLC patients treated with the IGF1R inhibitor Dalotuzumab, [74] according to our results, the subgroup of male patients with NSCLC appears to benefit the most of such a treatment. Promising preclinical trials investigating the role of IGF1R as a therapeutical target has resulted in the initiating of clinical trials on patients with multiple myeloma. [75]There are studies [76] suggesting IGF1R as a potential target in NSCLC treatment. While Tsuata et al [77] and Capuzzo et al [78]reported that the IGF1R expression did not represent a prognostic factor in resected NSCLC patients, other research groups [79][80][81] have either found a negative prognostic significance of high IGF1R expression or IGF1R GCNC. [82] However, future results of stratified treatment-trials among NSCLC patients will be needed to fortify these results.
Our current prognostic findings as detected by SISH for IGF1R and by immunohistochemistry for MET have a relevant practical implementation. While IGF1R SISH analyses seems to give additional information about the subgroup of NSCLC patients, who most likely would benefit of an anti-IGF1R therapy regimen, the MET SISH analyses seem to be biased when investigating the ratio of gene copy number to chromosome count due to the highly negative prognostic significance of chromosome 7 polysomy, higher than 3 chromosome copies. Our results would most probably apply to any SISH analysis investigating a gene to chromosome ratio located on chromosome 7. Even thought our findings would apply to small subgroups of patients, still finding therapeutic aid to these subgroups would be regarded as a significant step towards personalized NSCLC treatment. Until this, every effort should be made to find more specific and even more personalized potential molecular targets whose status in tumor samples might impact therapeutic responses.
Supporting information S1