Prognostic and predictive biomarkers for epidermal growth factor receptor-targeted therapy in colorectal cancer: Beyond KRAS mutations

Transcription

1 Critical Reviews in Oncology/Hematology 85 (2013) Prognostic and predictive biomarkers for epidermal growth factor receptor-targeted therapy in colorectal cancer: Beyond KRAS mutations Ana Custodio, Jaime Feliu Medical Oncology Department, La Paz Universitary Hospital, IDiPAZ, RTICC (RD06/0020/1022), Spain Accepted 4 May 2012 Contents 1. Introduction EGFR as a target EGFR protein expression EGFR gene mutations EGFR gene copy number (GCN) Alternative EGFR ligands Germline polymorphisms within the EGFR signalling pathway EGFR polymorphisms EGF and cyclin-d1 polymorphisms Polymorphisms in the fragment c gamma (Fc- ) receptors Cyclooxygenase-2 (COX-2) polymorphisms MicroRNAs polymorphisms KIRSTEN-RAS (KRAS) status Introduction KRAS mutations in colorectal cancer KRAS mutation status as a predictor of resistance to anti-egfr mabs Effect of KRAS mutation on response to anti-egfr therapy in first-line treatment Effect of KRAS mutation on response to anti-egfr therapy in pretreated patients Variation of the treatment effect by KRAS specific mutations Prognostic value of KRAS mutation status Neuroblastoma-RAS (NRAS) status BRAF status Introduction BRAF mutation status as a predictor of resistance to anti-egfr mabs Prognostic value of BRAF mutation status PIK3CA status Introduction PIK3CA mutation status as a predictor of resistance to anti-egfr mabs Prognostic value of PIK3CA mutation status PTEN status Other potential biomarkers HER2 gene status Corresponding author at: Medical Oncology Department, Hospital Universitario La Paz, Paseo de la Castellana 261, Madrid, Spain. Tel.: ; fax: address: anabcustodio@gmail.com (A. Custodio) /$ see front matter 2012 Elsevier Ireland Ltd. All rights reserved.

2 46 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) c-met and insulin-like growth factor receptor 1 (IGF1R) pathways TP53 mutations Use of multiple biomarkers to predict clinical outcome to ANTI-EGFR mabs Early response evaluation Early radiological tumour size decrease Skin rash as a biomarker of efficacy of anti-egfr therapy Hypomagnesemia as a biomarker of efficacy of anti-egfr therapy Conclusions and future directions Conflict of interest Reviewers References Biographies Abstract The advent of the epidermal growth factor receptor (EGFR)-targeted monoclonal antibodies (mabs), cetuximab and panitumumab has expanded the range of treatment options for metastatic colorectal cancer (CRC). Despite these agents have paved the way to individualized therapy, our understanding why some patients respond to treatment whereas others do not remain poor. The realization that detection of positive EGFR expression by IHC does not reliably predict clinical outcome of EGFR-targeted treatment has led to an intense search for alternative predictive biomarkers. Data derived from multiple phase III trials have indicated that KRAS mutations can be considered a highly specific negative biomarker of benefit to anti-egfr mabs. Oncologists are now facing emerging issues in the treatment of metastatic CRC, including the identification of additional genetic determinants of primary resistance to EGFR-targeted therapy for further improving selection of patients, the explanation of rare cases of patients carrying KRAS-mutated tumours who have been reported to respond to cetuximab and panitumumab and the discovery of mechanisms of secondary resistance to EGFR-targeted therapy. Current data suggest that, together with KRAS mutations, the evaluation of EGFR gene copy number (GCN), BRAF, NRAS, PIK3CA mutations or loss of PTEN expression could also be useful for selecting patients with reduced chance to benefit from anti-egfr mabs. This review aims to provide an updated of the most recent data on predictive and prognostic biomarkers within the EGFR pathway, the challenges this emerging field presents and the future role of these molecular markers in CRC treatment Elsevier Ireland Ltd. All rights reserved. Keywords: Prognostic factor; Predictive factor; Molecular marker; Epidermal growth factor receptor (EGFR) inhibitors; KRAS; BRAF; PIK3CA; PTEN 1. Introduction Colorectal cancer (CRC) is the third most common cause of cancer related death in western societies, accounting for approximately 10% of all cancer incidence and mortality [1]. It is widely accepted that CRC is a heterogeneous disease defined by different activating mutations in receptor tyrosine kinases (RTK) or activating loss-of-functions mutations in downstream components of RTK-activated intracellular pathways [2,3]. Therefore, the realization of that treatment should be tailored on an individual case-specific basis is becoming prevalent. Treatment on an individualized basis now involves a simultaneous case-specific analysis of clinical and pathological characteristics and analysis of a patientǐs genetic and tumour biomarker profile [4]. The improvement of our understanding of the biology of CRC has pointed towards the epidermal growth factor receptor (EGFR) pathway as a critical mechanism in CRC tumourigenesis [2,5]. The EGFR (ERB-1 or HER-1) is a member of the human epidermal growth factor receptor (HER)-erbB family of RTKs, that includes three other RTK: HER2/C-neu (ErbB2), HER3 (ErbB3) and HER4 (ErbB4). Since the establishment of the EGFR gene as an oncogene, it has become an important target for cancer treatment because its activation stimulates key processes involved in tumour growth and progression, including proliferation, angiogenesis, invasion and metastases. The binding of epidermal growth factor (EGF) or other ligands to EGFR initiates a mitogenic signalling cascade through two main axes. On one side, the KRAS RAF-mitogen-activated protein kinase (MAPK) pathway, responsible for gene transcription, cell-cycle progression and cell proliferation. The other axis involves membrane localization of the lipid kinase phosphatidylinositol 3-kinase (PI3K), which promotes Akt-mammalian target of rapapycin (mtor) activation, responsible for antiapoptosis and prosurvival signals. Importantly, the two axes (KRAS/BRAF and PIK3CA) are closely related and strictly interconnected, as the p110 subunit of PI3K can also be activated via interaction with RAS proteins. Cetuximab and panitumumab are both monoclonal antibodies (mabs) that recognize and inactivate the extracellular domain of EGFR, thus leading to inhibition of the downstream signalling pathways [6]. Both agents have been approved for metastatic CRC treatment based on the improvement of progression free survival (PFS) and overall survival (OS) when used either as single agents or in combination with chemotherapy (CT) [7 13]. These costly and potentially toxic treatments are, however, efficient in only a small percentage of patients and it is therefore extremely important to identify specific factors who will lead to a clearer definition of those

3 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) patients who will benefit from anti-egfr treatments. KRAS mutation is the first established molecular marker which precludes responsiveness to EGFR-targeted treatment with cetuximab and panitumumab [14 16]. In addition, among CRC tumours carrying wild-type (WT) KRAS, EGFR gene copy number (GCN), mutations of BRAF or PIK3CA or loss of phosphatase and tensin homolog (PTEN) expression may be associated with resistance to anti-egfr therapy [17 19], although these additional biomarkers require further validation before incorporation into clinical practice. Up until now, each of these markers has been mainly assessed as a single event, but these molecular alterations display overlapping patterns of occurrence, thus adding considerable complexity for drawing an algorithm suitable for clinical decision-making. Therefore, it has been suggested that the future comprehensive analysis of the entire oncogenic pathway triggered by the EGFR should be performed, thus enhancing the prediction ability of the markers individually used. In this review we summarize current progress in the search for molecular markers within the EGFR signalling cascade and further discuss any inconsistent or conflicting findings for these molecular classifiers in early-stage and advanced CRC treatment. 2. EGFR as a target 2.1. EGFR protein expression Positive EGFR protein expression, as determined by immunohistochemistry (IHC), was initially selected as an entry criterion for early studies evaluating EGFR inhibitors on the assumption that sensitivity to such agents was associated with EGFR expression [20]. However, a large body of evidence from patients who were treated with mabs for metastatic CRC [6,11,21,22] or tyrosine kinase inhibitors (TKI) for other solid tumours [23] indicates that this biomarker is poorly associated with response. Retrospective analysis of multiple series and data from the PRIME trial confirmed that, even in KRAS WT tumours, EGFR expression was not predictive for anti-egfr therapies efficacy [24 27]. Moreover, several authors reported that cetuximab was also active in tumours which were EGFR-negative by IHC yielding response rates (RR) up to 25% [28,29]. Many technical explanations have been advocated for the lack of association between EGFR protein detection by IHC and response to EGFR-targeted agents [30]. These reasons include disparity between the form of epitope of EGFR detected by IHC and that targeted by anti-egfr mabs, as well as issues related to processing and handling of tumour tissue samples, such as prolonged storage, and the fact that IHC could be actually a suboptimal technique for assessing EGFR status. In fact, IHC has shown a low specificity for predicting EGFR gene amplification [30]. EGFR is overexpressed in 60 80% CRC [31], but only a fraction of IHC-positive tumours also show EGFR gene amplification (17% in primary and 23% in metastatic tumours). Copy number variations of EGFR can be caused by polysomy in addition to gene amplification, without a significant increase of the receptor protein [32,33]. IHC is also a semiquantitative method that lacks a standardized scoring system and is subject to interobserver variation. Moreover, EGFR expression differences between primary CRC and their metastases have been reported, indicating that reliance on such marker in the primary tumour to predict treatment response of metastatic growths may be inappropriate [34]. Finally, some tumour specimens contain both low- and high-affinity EGFR binding sites. Because IHC-based methods cannot distinguish between them, these findings may provide further explanation for the lack of correlation between EGFR immunostaining and clinical response to EGFR-targeted treatment [28] EGFR gene mutations Activating mutations, including in-frame deletions and amino acid substitution in exons 18, 19 and 21 in the EGFR catalytic domain, are seen frequently in lung cancer and play an important role in determining responsiveness to TKIs [35]. However, studies on large cohorts of untreated metastatic CRC and on small cohorts of tumour samples taken from treated patients, quickly and unequivocally clarified that EGFR somatic mutations are extremely rare in CRC, and when they do occur, are not associated with response to anti-egfr mabs [36,37] EGFR gene copy number (GCN) Contrary to other situations in which the genomic locus corresponding to an oncogene is frankly amplified (>20- fold) (MET, PIK3CA), thus resulting in largely increased expression of the corresponding protein, it seems clear that the increase in EGFR GCN is often modest (3 5-fold) and caused by polysomy rather than gene amplification, without a significant increase of the receptor protein, as mentioned before [32,33]. Nevertheless, the association between an increase in EGFR GCN and positive clinical outcome to anti-egfr agents certainly exists, as confirmed by different series (Table 1) [14,29,30,32,33,37 45]. This molecular alteration can be detected by fluorescence in situ hybridization (FISH), chromogenic in situ hybridization (CISH) or polymerase chain reaction (PCR)-based methods. The comparability of these methods and their differential impact on results still needs to be defined. Moreover, the cutoffs defining EGFR amplification varied among series between 2- and 6-fold. As a reflection of methodological uncertainties, the prevalence of EGFR gene amplification shows a wide variation of 6 51% in different reports [30,32]. Increased GCN was found in at least 30% of patients when a threshold value of approximately three EGFR copies per nucleus was used, as determined by FISH, compared with only 10% of patients when a threshold of six or more EGFR copies per nucleus

4 48 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) Table 1 Tumour epidermal growth factor receptor (EGFR) gene copy number (GCN) and outcome of anti-egfr treatment in advanced CRC. Author (year) Treatment (type of patients) GCN cutoff (methodology) RR (%) PFS (months) OS (months) Frattini (2007) [43] Capuzzo (2007) [33] Sartore-Bianchi (2007) [39] Lièvre (2006) [14] Moroni (2005) [32] Personeni (2008) [42] Laurent-Puig (2009) [44] (KRAS WT population) Scartozzi (2009) [45] (KRAS WT population) Cetuximab ± CT (CT naïve or CT refractory) Cetuximab ± CT (CT refractory) Panitumumab (CT refractory) Cetuximab ± CT (CT naïve or CT refractory) Cetuximab/panitumumab ± CT (CT naïve or CT refractory) Cetuximab ± CT (CT refractory) Cetuximab ± CT (CT refractory) Cetuximab + irinotecan (CT refractory, KRAS WT) 4 22% NA NA <4 (FISH) 14% (p = 0.05) % <2.92 (FISH) 2.4% (p = ) 3.5 (p = 0.02) 8.5 (p = 0.8) % 8 15 <2.47 (FISH) 0% (p = ) 3 (p = 0.039) 10 (p = 0.015) 6 27% NA NA <6 (CISH) 0% (p = 0.04) 3 89% NA NA <3 (FISH) 5% (p < 0.001) 2.83 NA <2.83 (FISH) 4 (p = 0.25) 8.3 (p = 0.037) % <2.0 (FISH) 37% (p = 0.015) 7 (NS) 11.8 (NS) % 7.7 NA <2.6 (FISH) 9% (p = 0.002) 2.9 (p = 0.04) % 6.4 NA <2.12 (CISH) 6% (p = 0.03) 3.1 (p = 0.02) Lenz (2006) [29] Cetuximab (CT refractory) [PCR] NA NA NA (NS) (NS) (p = 0.03) CT = chemotherapy; FISH = fluorescence in situ hybridization; CISH = chromogenic in situ hybridization; PCR = polymerase chain reaction; WT = wild type; RR = response rate; PFS = progression-free survival; OS = overall survival; NA = not available; NS = not statistically significant. was used, as determined by CISH [38,39]. In addition, poor concordance rates between primary and metastatic tumours have been reported for EGFR GCN [40]. Moroni et al. first demonstrated the association between EGFR GCN, as determined by FISH analysis, and response to anti-egfr mabs in unselected populations [32]. Authors described an 89% RR in a subgroup of nine patients with CRC whose tumours had an increased EGFR GCN. These findings were confirmed from a retrospective analysis of a subgroup of patients participating in the pivotal phase III trial of panitumumab monotherapy [39]. The mean EGFR GCN per nucleus and the percentage of tumour cells with chromosome 7 polysomy ( 3 EGFR signals per nucleus) were analyzed by FISH and the association between these parameters and clinical outcome was assessed. None of the patients with a mean of 2.47 or less EGFR gene copies per nucleus, or fewer than 43% of tumour cells with chromosome 7 polysomy, respectively, achieved objective response, compared with six (30%) of the 20 patients (p = 0.001) and six (32%) of the 19 patients (p = 0.001) who had values above these thresholds. Patients treated with panitumumab with <2.47 EGFR copies/nucleus or <43% of tumour cells displaying chromosome 7 polysomy predicted for shorter PFS (p = and p = 0.029, respectively) and OS (p = and p = 0.014, respectively). However, EGFR GCN and chromosome 7 polysomy did not draw a parallel with PFS in patients receiving only best supportive care (BSC), suggesting that this parameter is not prognostic in metastatic CRC. The association between EGFR GCN increase and response to cetuximab- or panitumumabbased treatment was confirmed with different cut-offs by subsequent analysis, which are summarized in Table 1 [14,29,32,33,39,42 44]. More recent studies have evaluated the added value of EGFR GCN status in patients with known KRAS WT tumours who received cetuximab-based CT in the second and subsequent treatment lines [44,45]. Laurent-Puig et al. conducted a comprehensive analysis including the evaluation

5 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) of EGFR amplification/polysomy status by FISH and CISH in 96 KRAS WT tumours [44]. An EGFR FISH-positive phenotype was found in 17 patients (17.7%) who showed a statistically significant higher RR (71%) compared with patients with normal EGFR GCN (37%) (p = 0.015). A trend towards longer PFS and OS was found in patients with FISH-positive phenotype but without reaching a threshold of significance. This methodology was also employed by Scartozzi et al. [45] who evaluated the role of EGFR GCN in 44 irinotecan-refractory KRAS WT CRC patients treated with irinotecan and cetuximab. They reported a statistical significant association between GCN and RR, with a tumour regression in 9 (60%) and 2 (9%) cases with an increased and low FISH EGFR GCN, respectively (p = 0.002) and in 10 (36%) and 1 (6%) cases with an increased and low CISH EGFR GCN, respectively (p = 0.03). Median time to progression(ttp) was 7.7 and 6.4 months in patients showing increased FISH and CISH EGFR GCN, whereas it was 2.9 and 3.1 months in those with low FISH and CISH EGFR GCN (p = 0.04 and 0.02, respectively). These data contrast with earlier findings based on quantitative PCR analysis, showing that when EGFR GCN was assessed by this method, no association was found between this parameter and clinical outcome of cetuximab- or panitumumab-based treatment, probably because of tumour DNA dilution by DNA from normal cells during DNA extraction [23,29]. Lenz et al. did not find a relationship between EGFR GCN and response or PFS in 34 intensively treated patients exposed to single agent cetuximab. However, a significant correlation with survival was observed (p = 0.03) [29]. In a further analysis of 39 patients, the same group reported no correlation between EGFR gene expression and response to cetuximab, but indicated a significantly longer survival in patients with low gene expression levels of EGFR, IL-8 and Cox-2 [23]. In this context, it still needs to be clarified to which extent EGFR gene expression is not only a predictive factor of outcome but also has a prognostic importance. Although the most recent data are promising for the use of increased EGFR GCN as a positive predictive factor of clinical outcome to EGFR-targeted mabs, the reproducibility of data remains the largest obstacle for clinical applicability of this molecular determinant. Methods of tissue processing and EGFR scoring systems need to be standardized before using it for selecting patients for EGFR-targeted therapy [46] Alternative EGFR ligands The overexpression of alternative EGFR ligands, such epiregulin (EREG) and amphiregulin (AREG), is likely to be a key factor in determining sensitivity to anti-egfr therapies via ligand-driven autocrine oncogenic EGFR signalling [47 49]. The proposed model suggests that high expression of activating ligands may contribute to the cellular addiction to EGFR signalling in KRAS WT CRC. Such tumours demonstrate increased dependence on ligandactivated EGFR signalling and will experience the greatest oncogenic shock following rapid loss of this signalling as a result of ligand withdrawal during EGFR blockade. Retrospective studies have shown the potential predictive value of AREG and EREG expression in KRAS WT tumours. The level of sensitivity to cetuximab was shown to be proportional to the intensity of AREG and EREG tumour mrna expression [47,48,50,51]. Khambata-Ford et al. compared clinical outcomes for patients with high and low levels of these ligands in fresh-frozen tissues from CRC metastases [47]. Patients with high levels of EREG and AREG expression are more likely to have disease control with cetuximab (EREG, p = ; AREG, p = ) and also have significantly longer PFS than patients with low expression (AREG: median, vs. 57 days, hazard ratio [HR], 0.44; p = ; EREG: median, vs. 57 days; HR, 0.47; p = ). The exclusive use of either KRAS status or AREG or EREG gene expression profiles does not result in the selection of identical patient populations who are likely to benefit from treatment with cetuximab: among patients with WT KRAS, those whose tumours expressed high levels of AREG or EREG were likely to experience disease control, whereas patients whose tumours expressed low levels of these genes were not, thus providing important complementary information to KRAS status [48]. Jacobs et al. showed that it is also technically feasible to measure EGFR ligands expression in archival formalin-fixed paraffinembedded (FFPE) tissue of primary CRC [49]. High ligand expression levels were significantly associated with response (AREG, p = ; EREG, p = ), PFS (AREG: HR, 0.43; 95% confidence interval [CI], ; p < 0.001; EREG: HR, 0.41; 95% CI, ; p < 0.001) and OS (AREG: HR, 0.40; 95% CI, ; p = ; EREG: HR, 0.42; 95% CI; ; p = ) in KRAS WT metastatic CRC treated with the combination of cetuximab and irinotecan. There was no predictive power of ligand expression in patients with KRAS mutation. In the NCI-CTG 017 phase III trial of cetuximab versus BSC in chemorefractory CRC, high EREG gene expression was also found to be associated with improved median PFS (5.4 vs. 1.9 months) and OS (9.8 vs. 5.1 months) in KRAS WT patients treated with cetuximab [52]. In addition to its predictive value, EREG mrna expression appeared to be a useful prognostic marker in KRAS WT patients regardless of receiving anti-egfr therapy [53].In a cohort of 120 metastatic CRC patients not treated with anti- EGFR mabs, those with KRAS WT status and low EREG mrna levels showed significantly better OS than those with high levels (p = 0.006). AREG expression showed the same tendency but did not reach significant difference. Similar to EGFR copy number, the lack of standardization of the assays has prevented AREG and EREG expression levels from being used as clinical biomarkers for directing treatment with EGFR-targeted agents.

6 50 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) Germline polymorphisms within the EGFR signalling pathway 3.1. EGFR polymorphisms The regulation of EGFR transcription is not yet completely understood and there is growing interest to explore whether EGFR polymorphisms may influence expression of the receptor and clinical outcome to anti-egfr mabs [54 64]. In particular, the EGFR gene contains a highly polymorphic sequence in intron-1, which consists of a variable number tandem repeats (VNTR) of CA dinucleotides ranging from 15 to 26. Due to variable impact on DNA binding, this sequence has been shown to affect the efficiency of gene transcription, mrna amount and EGFR expression. Preclinical and clinical data have suggested that the longer the CA repeat, the lower the EGFR expression [54]. In experimental models, the transcription of the EGFR gene was found to be inhibited by approximately 80% in (CA) 21 -repeat alleles of the EGFR intron-1 variant, whereas decreasing the number of CA pairs down to (CA) 12 enhances transcription as much as five-fold [54]. Han et al. [56] and Liu et al. [57] found that a low number of the EGFR intron-1 (CA) n variants was associated with gefitinib responsiveness in non-small-cell lung cancer patients. Relative to the clinical response to anti-egfr treatment in CRC patients, Amador and coworkers reported that the VNRT of CA dinucleotides in EGFR intron-1 was associated with the pharmacodynamics of gefitinib [54]. Graziano et al. observed a significant association with longer PFS (HR, 0.45; 95% CI, ; p = 0.01) and OS (HR, 0.41; 95% CI, ; p = 0.006) for the EGFR intron-1 short/short (S/S) genotype (17 or less CA repeats VNTR) in 110 patients with refractory advanced CRC treated with cetuximab plus irinotecan therapy [58]. EGFR intron-1 S/S carriers also showed more frequent treatment response (p = 0.008) and grade 2 3 skin toxicity (p = 0.001) than EGFR intron-1 long/long (L/L) carriers, whereas no significant association was found between EGFR expression and EGFR intron-1 status. Therefore, a plausible mechanism for explaining the EGFR intron-1 variant influence on cetuximab activity is EGFR upregulation, which in turn is a determinant for the activity of anti-egfr therapeutics [54,55,59]. The lack of association between EGFR intron-1 variant and EGFR expression is likely to be related to methodological issues regarding the use of poorly validated and non-qualitative IHC methods for EGFR analysis in vivo [60,61]. These findings were confirmed by Pander et al. [62] but are in conflict with results from two previous studies that include 39 patients from a phase II multicenter trial of cetuximab [63,64]. In these studies, Zhang et al. did not find any association between EGFR intron-1 (CA) n and outcome. However, this polymorphism was studied by simply subdividing patients into two groups: 16 carriers of both (CA) n < 20 alleles, and 18 carriers of any (CA) n 20 alleles, with five missing cases. The sample size of the study by Graziano allowed for a more precise distinction of genotypes [58], with three distinct groups including two homozygous (S/S and L/L) and the heterozygous S/L genotype. Moreover, Zhang et al. [63,64] may have missed a dose-dependent effect of the EGFR intron-1 (CA) n variant with both S alleles present (CA repeats < 17) in the EGFR intron-1 S/S genotype. EGFR +497G>A is a single nucleotide polymorphism (SNP) affecting exon 13 at residue 521 (previously identified as residue 497, rs ), which has been associated with an arginine (Arg) by lysine (Lys) substitution in the extracellular domain of EGFR within subdomain IV. Moriai et al. were able to show that the Lys/Lys (A/A) genotype confers an attenuated function in EGFR ligand binding, growth stimulation, tyrosine kinase activation, and induction of proto-oncogenes [65]. The Arg/Arg genotype has been linked with improved OS in women with metastatic CRC (vs. Lys/Lys and/or Lys/Arg variants), although the reverse pattern was observed in men [66]. This same polymorphism has been linked to cetuximab response in other studies. Lurje et al. showed that EGFR +497 A/A genotype was associated with poor clinical outcome and shorter PFS (median PFS of 1.2 months), compared with other genotypes (median PFS 1.3 months and 1.8 months in patients homozygous and heterozygous for the G-allele, respectively; p = 0.017) [67] EGF and cyclin-d1 polymorphisms Modulation of the EGFR ligand, EGF, and of the downstream EGFR signalling, including the cyclin-d1 gene, may also play a role on cetuximab activity [58,62,63]. Functional variants have been described in the EGF 5 -untranslated region (EGF 61 G>A) [68] and in exon 4 of the cyclin-d1 gene (870 A>G) [69,70]. In the analysis by Graziano, EGF 61 G/G homozygous genotype was significantly associated with favourable OS (HR = 0.44; 95% CI, ; p = 0.01) and a borderline association between EGF 61 A/G heterozygous genotype (HR = 0.58; 95% CI, ; p = 0.04) and favourable OS was also observed [58]. However, this association was not described by other authors [62,63]. Zhang et al. found that EGF 61 any A allele genotype was correlated with favourable outcome [63]. In addition, a significant association was showed in this study between the cyclin D1 (CCND1) A870G polymorphism and OS. Patients with the A/A homozygous genotype survived for a median of 2.3 months (95% CI ), whereas those with any G allele (A/G, G/G genotype) survived for a median of 8.7 months (95% CI = ) (p = 0.019). When the CCND1 and EGF polymorphisms together were analyzed, patients with favourable genotypes (EGF any A allele and CCND1 any G allele) showed a median survival time of 12 months (95% CI = ), whereas patients with any two unfavourable genotypes (EGF GG or CCND1 A/A) showed a median survival time of 4.4 months (95% CI = ) (p = 0.004). These results have not been confirmed by Pander et al.: the CCND1 A870G polymorphism was not

7 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) significantly associated with PFS regardless of receiving cetuximab treatment [62] Polymorphisms in the fragment c gamma (Fc-γ) receptors Modulation of the immune response could be another important mechanism of action to anti-egfr mabs. The immunological mechanism antibody-dependent cellular cytotoxicity (ADCC) mediated through the fragment c gamma receptors (Fc- R) carried by innate immune cells such as macrophages and natural killers (NK) plays an important role in the antitumour effect of immunoglobulin (Ig) G1 antibodies [71]. In vitro studies have shown that cetuximab was able to induce ADCC [72]. Although it was initially believed that panitumumab, as an IgG2 mab, would not elicit ADCC, this phenomenon has recently been demonstrated in preclinical studies at concentrations that are analogous to therapeutic doses [73]. Constitutional polymorphisms have been demonstrated on genes encoding for the activating receptors Fc RIIa (CD 32, mainly expressed on macrophages) and Fc RIIIa (CD16, expressed on NK cells and macrophages), affecting their affinity to human IgG: a histidine (H)/arginine (R) polymorphism at position 131 for Fc RIIa (131 H/R) and a valine (V)/phenylalanine (F) polymorphism at position 158 for Fc RIIIa (158 V/F) [74]. On the basis of the different binding affinities, patients harbouring Fc RIIa-131H/H and Fc RIIIa-158V/V genotypes would be expected to mediate a more potent ADCC antitumour response after mab treatment [75]. Clinical studies have shown that Fc RIIa-131H/H and Fc RIIIa-158V/V genotypes were associated with better clinical outcome to rituximab as first-line treatment of follicular lymphoma [76,77] and to trastuzumab-based therapy in metastatic breast cancer [78]. Conflicting results have been published in chemorefractory advanced CRC studies, most of them developed in unselected KRAS populations [58,64,79 82]. Zhang et al. reported that Fc RIIa-H131R and Fc RIIIa-V158F polymorphisms were independently associated with PFS (p = and 0.055, respectively) in 39 EGFR-expressing metastatic CRC patients treated with cetuximab monotherapy [64]. Combined analysis of these two polymorphisms showed that patients with the favourable genotypes (FC RIIa any H allele, and FC RIIIa any F allele) showed a median PFS of 3.7 months (95% CI, months), whereas patients with any unfavourable genotypes (FC RIIa R/R or V/V) had a PFS of 1.1 months (95% CI, months; p = 0.004). On the contrary, in the analysis performed by Bibeau et al. [79], Fc RIIa-131H and Fc RIIIa-158V were favourable alleles, as observed in the studies with rituximab and trastuzumab. Irinotecanrefractory metastatic CRC patients treated with cetuximab plus irinotecan and with combined Fc RIIa-131 H/H and/or Fc RIIIa-158V/V genotypes had longer PFS than 131R and 158F carriers (5.5 vs. 3.0 months; p = 0.005) for the whole population as well as for the subpopulation of patients with WT KRAS (9.6 vs. 4.6 months; p = 0.015) and mutated KRAS (3.2 vs. 2.8 months; p = 0.015). Fc RIIa-131H/H and/or 158V/V polymorphisms also demonstrated significant association with OS for whole group (p = 0.032) and WT KRAS patients (p = 0.027). A pooled analysis of these two polymorphisms in 3 phase III trials of metastatic CRC patients (n = 317) treated with either single agent cetuximab (IMCL0144 and CA225045) or with second-line irinotecan plus cetuximab (EPIC) showed that patients with Fc RIIa H/H genotype had shorter PFS (median PFS, 1.3 months; 95% CI, ) compared to those with H/R or R/R genotypes (median PFS, 2.5 months; 95% CI, ; p = 0.036). No statistically significant associations were observed between Fc RIIIa polymorphisms and efficacy of cetuximab-based treatment [80]. Moreover, other studies have not found a significant association between the Fc RIIa or Fc RIIIa polymorphisms and the efficacy of cetuximab [58,81]. Finally, in order to provide the statistical power to test the definitive role of these SNPs on cetuximab efficacy, a large international consortium study has been performed including 900 unselected chemorefractory metastatic CRC patients treated with cetuximab alone or in combination [82]. No clinically significant correlation was found between Fc R SNPs and cetuximab efficacy in unselected KRAS status patients. However, an increased in disease control rate (DCR) was seen in KRAS mutant patients harbouring the FC RIIIa FF genotype (71.3% vs. 47.9% in non-ff; p = 0.049), as well as a trend for increased median PFS (median PFS, 16 vs. 12 weeks for FF and non-ff genotypes, respectively; p = 0.072). In the first-line setting, Pander et al. found that the Fc RIIIa V allele (F/V and V/V genotypes combined) was associated with decreased PFS compared with the Fc RIIIa F/F genotype (median PFS, 8.2 vs months, respectively; HR, 1.57; 95% CI ; p = 0.025) in a cohort of WTKRAS metastatic CRC patients treated with capecitabine, oxaliplatin and bevacizumab with cetuximab in the CAIRO- II study [26], whereas it was not significantly associated with PFS in the arm without cetuximab (p = 0.832) [62]. The Fc RIIa polymorphism was not significantly associated with PFS neither in the cetuximab arm nor in the arm without cetuximab. A possible mechanism for the opposite association of thefc RIIIa polymorphism with outcome in the studies by Pander and Bibeau could be that the high affinity V- allele [83] results in increased activation of tumour associated macrophages (TAMs) by cetuximab through cross-linking ofthe Fc R [84] in the study by Bibeau, instead of increasing ADCC in the study by Pander et al. [62]. As a result of TAM activation, proangiogenic mediators are released in the tumour microenvironment, such as vascular endothelial growth factor (VEGF) and matrix metalloproteinases (MMPs) [85]. In the CAIRO-II trial [26], patients had not received palliative CT before, whereas patients in the other studies [58,67,79] had been exposed to irinotecan and/or other lines of CT prior to cetuximab, which could have altered the infiltration of cells of the myeloid lineage such

8 52 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) as TAMs. However, it must be noted that the FC RIIIa F-allele was associated with increased efficacy of the IgG1- type mabs rituximab inlymphoma [76,77] and trastuzumab [78] in advanced breast cancer, as well as cetuximab in the study by Zhang et al. [64]. Therefore, fundamental research should be performed to support this highly speculative hypothesis Cyclooxygenase-2 (COX-2) polymorphisms Another SNP investigated as a predictive factor to pharmacoresistance was found in the COX-2 gene [67]. COX-2 is involved in the regulation of a broad range of cellular processes including tumour onset and progression, metastases, angiogenesis and resistance to CT [86]. The relationship between COX-2 and the EGF/EGFR signalling pathway is still controversial [87]. COX-2 is thought to be a downstream effector of EGFR and was found to be induced by EGF mediated stimulation of EGFR tyrosine kinase in human cell lines [88]. Other studies showed that COX-2 may be an upstream effector of EGFR in human colon cancer cells lines, suggesting that it induces colon cancer carcinogenesis by the activation of EGFR [89]. Furthermore, COX-2 has been reported to be a predictive and an adverse prognostic factor in a variety of malignancies, including CRC [23,86]. COX G>C is a frequent SNP located 765 base pairs upstream of the COX-2 transcription start site. The 765 C allele was shown to be associated with significantly lower COX-2 promoter activity compared with the 765 G variant [90]. Other common variant within the COX-2 gene include the COX T>C SNP. It locates within the functional region of 3-untranslated region of the gene and, therefore, may have a potential functional relevance in carcinogenesis, perhaps through control of mrna-stability and degradation [91]. The C allele is significantly less common in patients with cancer compared with healthy control patients, suggesting a protective effect [91]. Lurje et al. found low-expression variants of COX-2 (COX C and COX C) to be significantly associated with higher PFS in metastatic CRC patients treated with single agent cetuximab (COX G>C C/C genotype: HR, 0.31; 95% CI, ; p = 0.032; COX T>C C/C genotype: HR, 0.67; 95% CI, ; p = 0.003) when compared with other genotypes [67]. In multivariable analysis, COX T>C (adjusted p = 0.013) remained significantly associated with PFS, independent of skin rash toxicity and KRAS mutation. In summary, data on the predictive and prognostic value of EGFR signalling pathway polymorphisms at the present time are conflicting and need further investigation and validation. Given the retrospective design of all these studies, further larger statistically powered clinical trials are needed to evaluate the significance of these polymorphisms on anti-egfr mabs efficacy MicroRNAs polymorphisms MicroRNAs (mirnas) are small, noncoding RNAs that have revealed a new level of gene regulation in the cell. After being processed by Drosha and Dicer RNase III endonucleases, mature mirnas can regulate gene expression by degrading and/or suppressing the translation of target messenger mrna (mrna) by base pairing in the 3 -untranslated region (UTR) of mrna [92]. Exerting an effect as either tumour suppressors or oncogenes, mirnas regulate several genes that have important roles in cancer [93]. Variable levels of mirna in vivo may affect apoptosis, angiogenesis and specific molecular pathways such as the RAS cascade [94,95]. Recently, mirna polymorphisms were discovered and are becoming increasingly important on the fast growing field of the personalized medicine. MiRNA polymorphisms could present at or near a mirna-binding site of functional genes and can affect gene expression by interfering with mirna function. They have been shown to affect drug response and have the potential to confer drug resistance [96]. The let-7 (Lethal-7) family of mirna showed RAS regulating activity [97]. Let-7 induced RAS downregulation after binding to specific sites in the 3 -UTR of the human KRAS mrna [98]. A functional SNP has been described and characterized in the Let-7 complementary site (LCS) in the KRAS 3 -UTR mrna (LCS6) [99]. The LCS6 SNP (rs ) consists in a T-to-G base change and it was found to alter the binding capability of the mature Let-7 to the KRAS mrna. In experimental models, this variant attenuated the Let-7 control on KRAS with oncogene overexpression [99]. In addition, as a consequence of a possible negative feedback loop, the presence of the LCS6 G variant allele was associated with Let-7 downregulation [99]. The frequency of the LCS6 G- allele in Caucasian population is estimated to be approximately 5 10% in healthy individuals, but it was found to be markedly increased up to 20% in patients with lung cancer [99]. Furthermore, this polymorphism was found to be associated with increased cancer risk in non-small-cell lung cancer patients [99] and reduced OS in squamous cell carcinomas of the head and neck [100], suggesting functional and clinical significance. Moreover, it have been demonstrated that the Let-7 mirna can also modulate tumour sensitivity to chemotherapeutic agents. Nakajima et al. found the level of expression of Let-7g strongly related to responsiveness to S-1 treatment in 46 patients with recurrent of refractory advanced CRC [101]. The prognostic value of the LCS6 T>G variant in metastatic CRC patients treated with anti-egfr therapy has been evaluated in two studies [10,103]. Graziano et al. analyzed the KRAS 3 -UTR LCS6 polymorphism, KRAS mutation in codons 12, 13, 61 and the BRAF V600E mutation in the tumour DNA ofpatients with irinotecan-refractory metastatic CRC who underwent salvage irinotecan-cetuximab therapy [102]. In 134 patients there were 100 carriers of the wild-type LCS6 T/T

9 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) genotype (75%) and 34 carriers of the LCS6 G variant allele (T/G and G/G) genotypes (25%). The distribution of carriers of the LCS6 genotypes was significantly different among carriers of KRAS and BRAF mutation. In particular, all the 13 BRAF V600E mutation carriers were LCS6 T/T WT and G- allele carriers were significantly more frequent in the KRAS mutation group than in patients with KRAS WT (p = 0.004). In the 121 patients without BRAF V600E mutation, OS and PFS times were compared between carriers of the LCS6 G-allele genotypes and carriers of the WT T/T genotype. LCS6 G-allele carriers showed worse OS (p = 0.001) and PFS (p = 0.004) than T/T genotype carriers and these data were confirmed in the multivariate model including the KRAS status. In the exploratory analysis of the 55 unresponsive patients with KRAS mutation, LCS6 G-allele carriers and T/T carriers showed adverse PFS and OS times. Median PFS was 2.5 and 3.4 months (HR = 1.78; 95% CI, ) and 5.9 and 9.7 months (HR = 1.77; 95% CI, ) in G-allele carriers and T/T carriers, respectively. The G-allele could also have a role in patients with KRAS/BRAF WT status and its presence could represent an unfavourable predictive marker to the anti-egfr therapy. However, there were 12 G-allele carriers only among the 63 patients with KRAS/BRAF WT tumours. With this limitation, the log-rank comparison of PFS and OS times showed a trend for unfavourable outcomes in G allele carriers, but the difference was not significant. G allele and T/T carriers showed median OS of 9 and 14.2 months (HR = 1.19; 95% CI, ) and median PFS of 3.7 and 5.3 months (HR = 1.45; 95% CI, ), respectively. In general, the worse survival times in G-allele carriers than in WT T/T genotype carriers would suggest that preserved Let- 7 function may exert some control on the RAS pathway with additive effect to the anti-egf blockade. If the anti-egfr blockade does not work, because of the presence of an activating KRAS mutation, the downstream control of Let-7 may still ensure some KRAS downregulation, provided that there is preserved binding between the mirna and the mrna of the target gene. In a second study, Zhang et al. have studied the relationships between the KRAS Let-7 LCS6 T>G SNP and outcome in 130 metastatic CRC patients who were refractory to fluoropyrimidines, irinotecan, and oxaliplatin, and treated with cetuximab as monotherapy in the phase II study IMCL [103]. Analysis of KRAS let-7 LCS6 polymorphism was available in 111 patients and 13 of them (12%) were not assessable for tumour response. In 98 patients assessable for tumour response, 67 patients had WT KRAS and 31 patients had mutant KRAS. Fifty-five (82%) of the 67 KRAS WT patients had the KRAS Let-7 LCS6 TT genotype, whereas there are 12 KRAS WT patients harbouring at least one Let-7 LCS6 variant G allele (TG or GG). These 12 patients had a 42% ORR compared with KRAS WT patients with WT TT genotype who had a 9% ORR (p = 0.02). None of the 31 KRAS mutant patients had an objective response to cetuximab regardless of KRAS Let-7 LCS6 polymorphism. KRAS WT patients with TG/GG genotypes had a trend of longer median PFS (3.9 vs. 1.3 months; p = 0.25) and OS (10.7 vs. 6.4 months; p = 0.33) compared with those with T/T genotypes (p = 0.02), but did not reach statistical significance due to the small sample size. Apparently, these two studies [102,103] report conflicting results, but to reach a firm conclusion, it would be of interest to know the percentage of mutated 61 KRAS and the BRAF status in the 55 patients with LCS T/T genotype analyzed by Zhang et al. [103]. In fact, if it is considered that about 18% of WT KRAS patients could carry a mutated codon 61 KRAS or BRAF (8% of 61 KRAS plus 10% of BRAF, therefore approximately 9 10/55 cases), the non-responders to cetuximab among the LCS6 T/T patients could be attributed at least in part to the presence of KRAS or BRAF mutations. The same considerations could apply to PFS and OS analyses. In addition, in the abstract presentation of the Zhang et al. work at the 2009 American Society of Clinical Oncology (ASCO) meeting, authors reported an analysis of the LSC6 variant in patients enrolled in the IMCL-0144 trial and in the EPIC trial independent of KRAS mutation status in the tumour [104]. In this report, T/T carriers with mutant KRAS who were treated with irinotecan/cetuximab in the EPIC trial had significantly better PFS of 12 weeks (95% CI, ) compared with those harbouring a G-allele with median PFS of 6.4 weeks (95% CI, 5.7 7) (p = 0.037). There was no association between this polymorphism and clinical outcome in patients with WT KRAS enrolled in the EPIC. In a multivariate analysis the polymorphism remained independently associated with PFS in EPIC study. Notably, this finding parallels results of Graziano and colleagues [105]. In summary, these studies support the role of Let-7 LCS6 polymorphisms as a predictive marker of cetuximab efficacy in metastatic CRC patients. However, data were derived retrospectively and involve a relatively small number of patients, and therefore should be considered hypothesis generating and subject to confirmation in prospective and randomized controlled studies. 4. KIRSTEN-RAS (KRAS) status 4.1. Introduction KRAS, a member of the rat sarcoma virus (ras) gene family of oncogenes (including KRAS, HRAS, and NRAS), is located on the short arm of chromosome 12. It is a protooncogene encoding a small 21 kd guanosine diphosphate (GDP)/guanosine triphosphate (GTP) binding protein RAS that acts as a self-inactivating intracellular signal transducer [106]. Following Grb2/SOS mediated activation, GTP-bound KRAS recruits the serine protein BRAF, thus starting a cytoplasmic phosphorylation cascade leading to the activation of transcription factors [CREB, SRF, Fos, nuclear factor B (NF- B)] playing important roles in cell growth, differentiation, and survival. The oncogene PIK3CA encodes the p110 subunit of PI3K, which can be also activated via interaction with RAS proteins, as mentioned before [2,107].

10 54 A. Custodio, J. Feliu / Critical Reviews in Oncology/Hematology 85 (2013) The KRAS WT protein is transiently activated during tightly regulated signal transduction events. The binding of mabs to EGFR normally induces receptor internalization, causing a direct inhibition of TK activity and the blockage of downstream RAS/RAF/MAPK signalling. However, activating KRAS mutations result in a constitutively active GTP-bound protein which consequently renders the downstream pathway permanently switched on irrespective of the activation status of upstream receptors including EGFR. In such an instance, the binding of an anti-egfr mab to EGFR and theinhibition of ligand-mediated receptor activation will fail to elicitany pathway suppressive effects. Therefore, this constitutive pathway activation leads tounregulated proliferation, impaired differentiation, and resistance to anti-egfr therapies [2,107] KRAS mutations in colorectal cancer KRAS is the most commonly mutated gene in the RAS/RAF/MAPK pathway, with approximately 35% to 45% of metastatic CRC patients harbouring KRAS mutations [7,10,26]. Mutations in KRAS or BRAF and PIK3CA may coexist within the same tumour [19,108], but KRAS and BRAF mutations appear to be mutually exclusive [14,17,109]. Up to 90% of activating mutations of the KRAS gene are detected in codons 12 (82 87%) and 13 (13 18%), but less frequently in codons 61, 63 and 146. These are point mutations and are generally observed as somatic mutations. The most common types of KRAS mutations in CRC are G-to-A transitions and G-to-T transversions. The codons 12 and 13 code for two adjacent glycine residues located in the proximity of the catalytic site of RAS [110]. Different KRAS mutations result in an exchange of different amino acids at these catalytic sites, and therefore, may be responsible for the different levels of intrinsic GTPase activity reduction. As a consequence, variable KRAS mutations may be responsible for different biological alterations [110]. Codon 12 mutations of the KRAS gene were associated with a mucinous phenothype of CRC. By contrast, CRCs associated with codon 13 mutations were rather non-mucinous, but were characterized as more aggressive tumours with a greater metastatic potential [111]. KRAS mutation is thought to be an early event in tumourigenesis [112]. Several series have demonstrated a KRAS mutational status concordance 95% between primary tumours and paired metastases [40, ]. These data have recently been confirmed in a large and homogenous Dutch study, which found a discordance between primary tumours and corresponding liver metastases in only 11 of 305 cases (3.6%) [119], thus suggesting that KRAS mutations analyses from primary tumours can be used with a great reliability for treatment decisions in metastatic setting. However, high rate of heterogeneity of KRAS mutation status (up to 30%) was observed between primary tumours and lymph node metastases in a limited number of studies [116,117,120]. In addition, intratumoural heterogeneity of KRAS mutation status has also been detected in up to 8% of samples when tumour centers and invasion fronts of primary CRC were compared [116]. Therefore, the site most suitable and reliable for diagnostic mutation analysis has been defined as the tumour center. Interestingly, in a retrospective Spanish report of 230 CRC patients, a higher percentage of KRAS mutations was detected in primary tumours of patients with lung metastases than in those with liver metastases (57% vs. 35%, p = 0.006), suggesting a role for KRAS mutations in the propensity of primary CRC to metastatize to the lung [121]. Despite a general consensus favouring the introduction of KRAS testing in clinical practice as a powerful means to select patients before drug administration, the choice of the most appropriate method for KRAS mutation analysis remains a complex challenge, and it is still not known what level of test sensitivity is required in order to provide useful and predictive information in clinical practice [ ]. The hotspots and thus the most frequent mutations given in the recommended genetic nomenclature [122] are g.34g>c (p.g12r), g.35g>c (p.g12c), g.34g>a (p.g12s), g.35g>a (p.g12d), g.35g>c (p.g12a), g.35g>t (p.g12v), g.38g>a (p.g13d) and rarely g.183g>t (p.q13h). Currently, the most widely applied methods for assessing KRAS gene status is direct dideoxy DNA sequencing and the PCR-based assays, which have a relatively low sensitivity because mutant alleles must be present in at least 20% and 10% of cells, respectively, to be reproducibly detected [123,126]. More sensitive methods are available for KRAS analysis. Some are laboratory-made techniques, such as mutant enriched-pcr (ME-PCR), and others are commercial test for diagnostic use, such as matrix-assisted laser desorption/ionization-time-of-flight (MALDI-TOF) technology or amplification refractory mutation system (ARMS). Several studies have compared different methodologies of KRAS analysis but without showing a correlation with the clinical response to anti-egfr agents in metastatic CRC patients [127,128]. Mollinari et al. retrospectively evaluated objective tumour responses in metastatic CRC patients treated with cetuximab- or panitumumab-based regimen [125]. KRAS codons 12 and 13 mutations were examined by direct sequencing, MALDI-TOF mass spectrometry (MALDI-TOF MS), ME-PCR, and engineered ME-PCR (emr-pcr), which have a sensitivity of 20%, 10%, 0.1% and 0.1%, respectively. In addition, KRAS codon 61 mutations, BRAF and PIK3CA mutations by direct sequencing and PTEN expression by IHC. Direct sequencing revealed mutations in codons 12 and 13 of KRAS in 43 of the 111 patients considered (39%) and BRAF mutations in 9/111 (8%). Using highly sensitive KRAS analysis methods, additional KRAS mutations were detected in up to 13/68 (19%) patients after directed sequencing had shown them to be WT in codons 12 and 13 of KRAS. With the exception of 2 patients with KRAS G13D mutations and 1 patient with the rare KRAS G60D mutation in exon 3, the patients with KRAS, BRAF, or PIK3CA mutations, as detected by direct sequencing, did