Clinical Implications and Quality Assurance of Molecular Testing for EGFR-Targeting Agents in Colorectal Cancer
- First Online:
- Cite this article as:
- Vecchione, L., Saridaki, Z. & Tejpar, S. Curr Colorectal Cancer Rep (2012) 8: 42. doi:10.1007/s11888-011-0112-3
- 50 Views
The introduction in clinical practice of anti-epidermal growth factor receptor (EGFR) antibodies has improved the clinical outcome of metastatic colorectal cancer (mCRC) patients. Nevertheless, only 10% of mCRC tumors respond to these treatments, thus rendering the efforts made to maximize their therapeutic index justified. Although several biomarkers have been identified, we do not know yet how to administer these drugs in colorectal cancer patients in a “personalized–targeted manner.” With this review we will try to demonstrate that we need to go beyond the assumption of a binary relationship between one genetic event and response or resistance to anti-EGFR drugs and that several factors can influence the response to these agents. Therefore, the introduction in future approaches of a holistic genomic discovery plan instead of an individual and specific identification of alterations is needed.
KeywordsMetastatic colorectal cancerEGFR dependencyAnti-EGFR treatmentsCetuximabPanitumumabAnti-EGFR moAbsPersonalized cancer medicineColorectal cancer subgroupsMolecular subgroupsKRASBRAFPI3KCANRASPTENAmphiregulinEpiregulinHER2 amplificationGene moduleGene expression profileFcγRIIaFcγRIIΙaLet7Anti-EGFR sensitivityAnti-EGFR resistanceKRAS testing
The development of large-scale sequencing projects has helped us to understand that neoplasms arising in a given tissue are highly heterogeneous at their molecular level and carry hundreds of genomic alterations. This has led the scientific community to look at neoplasms originating from a given tissue as consisting of several molecular subtypes that can define a specific population of patients with distinct prognosis and clinical manifestations, thus justifying different treatment approaches. At the same time, we have also learned that only few of these alterations present in each neoplasm are the “drivers” of the tumorigenic process. These “driving” alterations are selected during tumorigenesis in order to render cancer cells “addicted” to their constitutive activation. Pharmacological blockade of the alterations on which cancer cells are depending on often leads to an “addiction shock” which affects cell growth and induces cell death. The “addiction” paradigm has been pharmacologically exploited and drugs designed to specifically target “driving” alterations have led to the so-called “personalized medicine” in cancer treatment .
Although the classical tumor progression model for colorectal cancer (CRC) development introduced by Fearon and Vogelstein  does not capture this complexity , a common agreement still exists on the different stages that characterize its evolution: from normal mucosa to invasive adenocarcinoma through dysplastic lesions, adenoma, and adenocarcinoma in situ . This multistage CRC formation process is driven, on one hand, by the inactivation of recessive growth inhibitory genes, mostly due to genetic (mutations) and epigenetic (DNA methylation and histones acetylation) events, and on the other hand, by the activation of dominant growth-enhancing genes. The genomic landscape of CRC is complex, with each tumor comprising an estimate median of 76 non-silent mutations of which approximately 15 can be considered as “drivers” in the tumorigenic pathogenesis and progression of the primary tumor.
Much less is known about the alterations leading to and maintaining the metastatic growth of carcinoma cells, whereas this is the main population we are treating today. Our current animal models, based on transgenic mice, are good at mimicking the tumor initiation process, providing us functional information on the key genetic drivers there, but we are sorely lacking models of metastatic disease. This means that the key “drivers” of metastatic disease and their interactions with the metastatic environment still remain enigmatic.
The EGFR is a member of the human epidermal growth factor receptor (HER)-erbB family of receptor tyrosine kinases. It is selectively activated by ligands belonging to the EGF family of peptide growth factors [5, 6]. Upon the binding to the extracellular domain, EGFR forms homodimers or heterodimers with its family members ErbB2/Neu, Erbb3/HER3, and Erbb4/HER4, thus undergoing to the autophosphorylation of its intracellular domain. EGFR can also be activated through E-cadherin  interaction and heterodimerization with insulin growth factor receptor . The receptor autophosphorylation leads to the activation of multitude pathways, including the Ras/Raf mitogen-activated protein kinase (MAPK), the PI3K/AKT and the Jak2/Stat3 pathways, which can be responsible for cancer cell proliferation, survival, invasion, metastases, and neo-angiogenesis [9, 10]
The importance of EGFR and its ligands in normal and malignant epithelial cell biology  and in the growth and aggressiveness of various epithelial tumors  has been the rationale for the development of drugs against it [13, 14]. Nevertheless, it is relevant to underline that this rationale has only been based on the knowledge of the role of EGFR in the early stage of tumor formation. On the contrary, little is known about the precise function of EGFR in metastatic disease and its interactions with the niche. The lack of knowledge of the EGFR role in mCRC physiology is one of the reasons why we are still looking for mCRC patients to selectively target.
Two monoclonal antibodies targeting EGFR (anti-EGFR moAbs), the chimeric IgG1 moAb cetuximab and the fully humanized IgG2 moAb panitumumab, have been developed and introduced into clinical practice. By binding the EGFR extracellular domain these moAbs inhibit its dimerization, its subsequent phosphorylation, and its downstream signaling. Several clinical trials have confirmed the efficacy of these two compounds in the treatment of mCRC patients [15–17]. Although the EGFR is expressed in the majority of colon adenocarcinoma and seems to have a role in microadenoma initiation with still unknown mechanisms , responses to treatment with anti-EGFR moAbs used as single agents have been observed only in 10% of mCRC patients .
Moreover, even if the first clinical trials were designed for an unselected population of chemorefractory mCRC patients [15–17], the observation of low response rates was already the proof of concept of primary resistance to be common and that only a specific subset of mCRC might benefit from these anti-EGFR moAbs. This can be explained by the hypothesis that only a limited number of tumors are EGFR dependent or, better, EGFR addicted, therefore, resulting in clinical benefit with anti-EGFR moAbs even after failure of standard cytotoxic chemotherapies.
Biomarkers for anti-EGFR moAbs sensitivity
Negative selection, but robust negative predictive value and large incidence (40%), high impact
Negative selection, small incidence (5%–10%); low impact
Negative selection, small incidence (15%); contrasting results depending on the mutation; small impact
Negative selection, small incidence (2,6%); small impact
Contrasting results, small impact
Positive selection, mainly overexpressed in KRAS wild-type mCRC; impact of relevance in particular when they are considered together with other genes
The aim of this review is to critically analyze what we have been taught until now in the field of anti-EGFR personalized medicine and the future perspectives related to it. We will go beyond the assumption of a binary relationship between one genetic event and response or resistance to a given therapy and we will describe how several factors can influence that. Furthermore, we will try to justify why our future approaches should be characterized by a holistic genomic discovery plan and not by individual and specific identification of alterations.
Negative Predictors of anti-EGFR moAbs
Although it seemed to have some rationale at that time, the evaluation of the EGFR expression in CRC primary tumors by immunohistochemistry (IHC) failed to be considered as a marker of response, since it does not correlate neither with clinical responses nor with clinical benefit from anti-EGFR moAbs [15–17, 19, 20]. EGFR amplification and EGFR gene copy number (GCN) variation have also been investigated but few limited evidences can support their use as predictive markers of response [21–24], while EGFR kinase domain mutations in CRC tumors are highly unlikely to determine response to anti-EGFR targeted agents because of their low incidence in these tumors [25–27].
It became soon evident that activating KRAS alterations, which are present in nearly 30%–40% of CRC tumors [28, 29], were associated with lack of activity of anti-EGFR moAbs. Several independent non-randomized retrospective studies [30–34], several retrospective analyses of prospective randomized trials [35–39], and a European consortium study [40•] have been the basis for this proof of concept: the presence of KRAS-activating mutations correlates with primary resistance to anti-EGFR moAbs. This has led the American and European health authorities to introduce the KRAS test as mandatory for cetuximab and panitumumab treatment in mCRC patients (European Medicine Agency EMEA-H-C-741 and H-C-558, and US Food and Drug Administration FDA Application No. [BLA] 125084 and No. [BLA] 125147).
In spite of these first encouraging results, the KRAS status is not able alone to optimize the selection of mCRC patients who will benefit from anti-EGFR moAbs treatments at least for two reasons.
The first reason is that even if KRAS mutations can be considered a specific negative biomarker of response with a specificity of 93%, their sensitivity, defined as the capability of the test to recognize the wild-type patients who will benefit from anti-EGFR agents, is low, around 47%. This means that the absence of KRAS mutations is not a guarantee for anti-EGFR agents response. Indeed, the percentage of wild-type (WT) KRAS patients who can benefit from anti-EGFR moAbs is around 20%–40%, thus indicating that other mechanisms of primary resistance might be involved [33, 39, 40•].
The second reason is that recent evidence indicates that not all KRAS mutations are equal regarding their impact on mediating EGFR resistance [41•]. Various reports have described that patients with KRAS-mutated tumors may occasionally respond to either cetuximab or panitumumab. Initially these findings were considered as erroneous, and were thought to be related to either the inaccuracy of the mutational test, or, to response to chemotherapy that was administered in combination with anti-EGFR targeted compounds . KRAS mutations are mainly detected in codons 12 (70% of cases) and 13 (20% of cases) and less frequently (1%–4%) in codons 61 and 146 [42–44]. Recently our group has unexpectedly found a positive association between KRAS G13D mutations and better progression free and overall survival for patients who received either cetuximab alone or in combination with chemotherapy compared to KRAS G12V [41•]. The evidence in our dataset of some KRAS G13D not achieving a clinical benefit from anti-EGFR targeted treatment indicates that these results have to be considered with caution and their impact should be carefully assessed in well-designed prospective trials. Nevertheless, we can hypothesize that KRAS mutations might have different roles depending on the context where they develop and thus to which subgroup of colorectal cancer they belong.
Several groups have then been focusing on the molecular analysis of additional genes involved in the downstream of the EGFR signaling such as BRAF, NRAS, and PIK3CA. BRAF is one of the primary downstream effectors of KRAS signaling and the V600E point mutation is the most common alteration that involves this gene, with a frequency in mCRC of about 10%. Some retrospective studies suggest that the BRAF V600E mutation is associated with an unfavorable prognosis regardless of the treatment administrated and with primary resistance to anti-EGFR moAbs [40•, 45–49]. Nevertheless, it seems important to underline that these studies investigated the role of the BRAF mutation in response to anti-EGFR agents used as monotherapy or in combination with irinotecan-based regimens in a chemorefractory population. On the contrary, the combined analysis of two randomized trials of first-line chemotherapy, the CRYSTAL and the OPUS, reported a trend in favor to cetuximab treatment for BRAF-mutant patients. Based on these contradictory results, we cannot yet conclude if the BRAF V600E mutation is a negative predictive marker of response to anti-EGFR moAbs but we can certainly hypothesize that it might play a different role in heavily pre-treated and chemo-naive patients. Prospective trials evaluating its role should be then taken into account.
Furthermore, NRAS mutations might impair responsiveness to cetuximab treatment [40•]. Even if KRAS, NRAS, and BRAF mutations are mutually exclusive in CRC, presumably because there is no advantage for a tumor cell to alter both genes since they act in the same linear signaling pathway, all of them can contribute to the selection of non-eligible patients for anti-EGFR agents treatment. By considering the 40% of KRAS and the 10% of BRAF and NRAS mutated tumors the selection of non-responders can be improved up to 45%–55% [40•].
Besides the Ras/Raf MAPK pathway, EGFR also activates the PI3K/AKT signaling pathway. The latter can be deregulated, either by the inactivation (often through epigenetic mechanisms) of the PTEN phosphatase, or by activating mutations of the PIK3CA p110 subunit. Initial analyses on the impact of PIK3CA mutations on response to cetuximab were conflicting with some groups reporting association with resistance and some others reporting association with response [40•, 47, 48, 50–52]. Moreover, the role of PIK3CA mutations became more complicated due to the fact that these mutations can co-occur with KRAS and BRAF mutation, probably indentifying another CRC molecular subgroup different from the single mutation ones. The role of PTEN loss and its relationship with response to anti-EGFR targeted agents is still under investigation. Four retrospective studies [50, 51, 53, 54] have provided evidence that PTEN status is associated with objective responses in cetuximab-treated mCRC patients, suggesting that PTEN-positive tumors tend to have a better outcome than negative ones; however, two other ones failed to confirm this observation [48, 52]. The reasons of this discordance could be probably attributed to methodological differences, such as the anti-PTEN antibodies, the IHC scoring algorithms, and the cutoff criteria that have been used. The above-mentioned information underlines the fact that PTEN results have to be considered with caution and that the lack of validation for PTEN analysis does not allow it to be used yet in clinical practice. Nevertheless, it should be kept under consideration in the planning process of prospective biomarker studies. Another biomarker that should be better validated in retrospective studies and also introduced in new prospective trials is c-ERB2/neu receptor. Bertotti et al. have recently reported c-Erb2 amplification to be responsible for cetuximab resistance both in in vitro and in in vivo studies [55, 56]. These results are in line with previous data in mCRC chemo-refractory patients, where HER2 amplification was not associated with cetuximab response . These findings provide the rationale of combining anti-EGFR moAbs and anti-HER2 compounds when the validation of HER2 role in mCRC will be obtained. In summary, negative predictors still have a lot of uncertainty about them. First, because they have not been studied in randomized settings and thus it is still unclear whether they are prognostic or predictive, and second because the biology behind them is not completely clear yet. For example, KRAS and BRAF might not be equal in terms of EGFR signaling regulation. We are not to date certain on the activity of KRAS codon 146 mutations and of the role of NRAS in the activation of the MAPK pathway. It is also unclear which distinctions have to be made regarding PIK3CA mutations and which is the role of PTEN. The above-mentioned uncertainties show us that without solid proof and in-depth knowledge of CRC biology, we risk over-adopting negative markers and under-treating patients. It would probably be more correct to pursue a holistic approach based on repeated observations and biological rationale.
Positive Predictors of anti-EGFR moAbs
KRAS, BRAF, NRAS, and PIK3CA are all downstream effectors of EGFR. Based on the above-mentioned studies [30–39, 40•, 45–52], patients affected by mCRC that carry the mutated active forms of these genes are associated with lack of response to anti-EGFR moABs. In this context we might postulate that these mutated tumors are non-addicted to EGFR and belong to different CRC subgroups. This means that the resistance to cetuximab and panitumumab might not be only related to the constitutive downstream activation of the EGFR pathway by these mutations, but it could indicate that these mutations occur in a different subset of tumors, that does not bear the EGFR upstream oncogenic activation hallmarks. An example of this is given by the epiregulin (EREG) expression, an activator of EGFR, which is clearly less or not expressed in the KRAS-mutant subpopulation, whereas it is present in a subset of the KRAS wild-type mCRC population . Moreover, a single gene mutation will never be able to capture the whole complexity of a molecular subtype; as it has been previously described and it will be further better explained, some of these single gene mutation groups are still heterogeneous both at their clinical manifestations and at their molecular level.
To improve our ability to predict the response to anti-EGFR agents we must identify what is driving EGFR dependency. The finding of positive predictors of response would allow us to administer targeted drugs to a predefined selected population of patients which would benefit from them . Examples for this statement are some solid tumors like GIST, lung, and breast cancer where c-KIT mutations, EGFR mutations, and HER2 overexpression, respectively, predict the response to the corresponding targeted agents [59, 60]. In CRC we are not yet in the position to claim that there are predictive markers that can capture the responding population to anti-EGFR treatments with high specificity and sensitivity. Nevertheless, some evidence toward this direction exists. Khambata-Ford et al.  were the first to indentify amphiregulin (AREG) and epiregulin (EREG), two EGFR ligands, to be among the top genes predicting the response to cetuximab. This gene signature, obtained from snap-frozen liver metastasis from mCRC patients uniformly treated with cetuximab monotherapy, was thought to be predictive of response to cetuximab. These data were later confirmed by our group where the analysis of primary CRC formalin-fixed paraffin-embedded (FFPE) tumors from refractory metastatic patients treated with cetuximab-based therapy revealed AREG and EREG to predict cetuximab sensitivity . In line with these results, the analysis performed by Tabernero et al.  in a cohort of 106 patients which received cetuximab as first line in combination with an irinotecan-based chemotherapy regimen revealed that AREG and EREG were overexpressed in the KRAS wild-type subgroup which achieved clinical benefit. Similar results were obtained by Saridaki et al.  in a total of 112 mCRC patients. Based on these data we might conclude that AREG and EREG could be considered as markers of sensitivity to anti-EGFR moAbs and therefore markers of EGFR dependency in CRC tumors. In reality, this is not completely true since there are also some KRAS WT tumors with low ligands expression that respond to cetuximab . The model of EGFR addiction has to be improved and additional markers need to be found. In line with this hypothesis, Baker et al. [62••] analyzed, by high-throughput RT-PCR, the mRNA expression of 110 candidate genes in a total of 144 KRAS wild-type mCRC patients treated with cetuximab monotherapy. Most of these genes were found to be strongly associated with all the clinical outcome variables considered (disease control, response rate, progression-free survival) and a number of them were compatible to the EGFR biology in CRC development: AREG, EREG, and VAV3, which activates the EGFR signaling pathway, were associated with an increased clinical benefit, while DUSP6, a phosphatase known as a feedback inhibitor of the MAPK pathway, was associated with a decreased clinical benefit. Moreover, the authors also performed a multivariate analysis on the entire set of identified genes, and a four-gene classifier including EREG, AREG, DUSP6, and SCL26A3 was generated. This four-gene classifier, alongside KRAS status, significantly improved the specificity and the predictive positive value of cetuximab benefit (in terms of disease control and response rate) compared to the selection of patients based solely on KRAS status. When this model was applied to progression-free survival it performed better than AREG, EREG, or KRAS status alone. This means that the use of one single marker might lack the capability to define the EGFR-dependent tumors while a group of markers, like genes and/or proteins, might better identify the EGFR-addicted/cetuximab-sensitive colorectal cancer. A first step in this direction has been done by Rhodes et al. [63••], who, by an unsupervised hierarchical clustering analysis of gene expression profile of mCRC treated with cetuximab, identified gene modules associated with cetuximab response. The importance of their work is based on the finding of a RAS activation module associated with lack of response to cetuximab that they defined as surrogate of RAS mutations, a carcinoma-like proliferating module, which included EREG and AREG expression, associated with response to cetuximab and other modules associated with resistance, mainly linked to metastasis/invasion, leukocytes, stromal response, cellular defense, and interferon response. The identification of a gene expression module based on RAS activation that could serve as a surrogate for KRAS mutational status, as well as the identification of modules of resistance independent of KRAS mutations, clearly demonstrate that we should go beyond a single-gene mutational status in stratifying the target population. As previously reported, a single gene alteration in terms of overexpression or downregulation at its mRNA level is not representative for different subgroups. On the contrary, the evaluation of a set of genes belonging to modules or to a specific signature might help in identifying the different colorectal tumors [64••].
Besides the tumoral genetic characteristics, the germline genome of a CRC patient might play a role in granting resistance or sensitivity to anti-EGFR moAbs. This hypothesis has been validated by the finding that polymorphisms in the genes encoding for the FcγRIIa, FcγRIIΙa [65, 66], and the EGFR  have been associated with outcome in mCRC patients treated with cetuximab. Moreover, some studies indicate KRAS to be a direct target of the Lethal-7 (let-7) micro RNA (miRNA) family. Upon binding specific sites in the 3′ untranslated region (3′-UTR) of the KRAS mRNA, let-7 seems to induce KRAS downregulation . A functional single nucleotide polymorphism (SNP) has been described and characterized in the Let-7 complementary site (LCS) in the KRAS 3′-UTR mRNA, thus resulting in increased expression of the KRAS oncogenic protein. Nevertheless, the role of SNP still remains unclear since conflicting data are emerging from small clinical studies [69, 70], thus underlining the need of further research in finding the host/germline characteristics that might track with EGFR sensitivity.
Quality Assessment for KRAS Testing
Every biomarker, either positive or negative, needs extensive technical validation. As good as a marker may be in predicting outcome, if it cannot be reliably assessed on patient material and provide a clear interpretable result, the marker is of no use to the community.
Even for KRAS determination, and although its assessment is obligatory before the administration of anti-EGFR moAbs–based therapy, technical issues remain in the whole process of its evaluation in CRC tumor specimens. As is the case for many markers, the process of technical optimization can take years, but must remain a priority for the marker information to translate efficiently into the clinic. For KRAS and CRC issues such as the kind of sample to be used, concordance between primary and metastasis and genetic drift have been previously reviewed [71, 72]. Here we would like to highlight an aspect that touches on decentralized assessment of markers in laboratories and the use of validated assays. As it was clearly revealed recently after the set up of a KRAS external quality assessment (EQA) scheme in an effort to evaluate KRAS testing in Europe, an optimal, uniform, and reliable community-based evaluation does not exist . The KRAS EQA scheme included the evaluation of the identification of mutations, of tumor cells percentage calculation, and of a test results report. Only 70% of the participating laboratories correctly identified all KRAS mutations in all samples, with both false-negative and false-positive results affecting negatively patient care. Among the 30% of laboratories that made at least one error, 22% made genotyping errors and 8% reported technical failures. Furthermore, serious mistakes were also detected in the report of the results. Concluding, the authors state that if the EQA scheme reflects KRAS testing on a routine basis, at least one in 10 samples is wrongly genotyped in >30% of laboratories and that these observations should provide the basis for remedial measures and harmonization . As previously mentioned, a primary tumor’s KRAS status became mandatory for the treatment of metastatic CRC with an anti-EGFR moAb in both the European Medicines Agency (European Medicine Agency EMEA-H-C-741 and H-C-558) and the US Food and Drug Administration (FDA Application No. [BLA] 125084 and No. [BLA] 125147). Following that, in 2009 the FDA issued a recommendation against the use of these drugs in patients with tumors mutated in codon 12 or 13 of KRAS. Nevertheless, in order for this recommendation to be translated into a label change of the drugs two additional elements are required: first, a single assay for the 7 KRAS mutations in codons 12 and 13 must be validated fulfilling class III premarket approval requirements, and second, all the trials should be reassessed uniformly with this assay, with consistent results on the negative predictive impact. In this front results are eagerly awaited. In Europe, EMEA took from the beginning a slightly different, more open-minded approach when changing the approval of these drugs, declaring that they should be used only in metastatic patients with KRAS wild-type tumors. This has important implications because neither exact mutations to be tested must be specified, nor with which exact methodology .
The introduction in clinical practice of targeted agents against EGFR has improved the outcome of mCRC patients but has also changed the way of thinking of this disease. The finding that not all the mCRC patients respond to anti-EGFR moAbs has helped the scientific community to go further from the simplistic histological definition and to look for markers of response and/or resistance to these compounds. KRAS mutation can be considered as the first accepted negative marker of response to cetuximab and panitumumab in mCRC setting that has been introduced into clinical practice, although uncertainty still exists on whether this application can be considered adequate and homogenous. Several evidences point to the direction of CRC being a heterogeneous disease, providing a possible explanation on why not all mCRC patients respond to anti-EGFR moABs. Individual mutations within a single cancer gene may play distinct roles and define different tumors subtypes within a given histology with completely different gene expression profiles. Currently the information brought by the KRAS mutation status for patient selections is the only with large, robust evidence and warrants its clinical application for cost-effective use of anti-EGFR agents in mCRC. But at the same time, the “KRAS story” has taught us that more comprehensive approaches are needed in the future in order to lead to a better marker identification, and should be integrated upfront in the drug development process. This is much easier said than done, and necessitates a prohibitively large set of unselected patients to be treated and retrospectively analyzed to generate robust biomarker signals, which then need to be prospectively validated. This assumption goes against many of the current drug development strategies, as well as the constraints applied by regulatory agencies, finances, and the community. A new dialogue will be needed to weigh the continuation and consequence of each approach on the rocky path of personalized medicine.
Loredana Vecchione and Zenia Saridaki equally contributed in manuscript writing. Zenia Saridaki is a recipient of a research fellowship from the Hellenic Society of Medical Oncology.
L. Vecchione: none; Z. Saridaki: none; S. Tejpar: honoraria from and educational presentations/speakers’ bureau for Merck Serono.