Germline Polymorphisms in Genes Involved in the IGF1 Pathway Predict Efficacy of Cetuximab in Wild-type KRAS mCRC Patients

Colorectal cancer (CRC) is the second most lethal malignancy in the United States. In 2009, 146,970 new cases of CRC and 49,920 deaths were recorded (1). Monoclonal antibodies (mAb) targeting the epidermal growth factor receptor (EGFR), including the chimeric immunoglobulin G1 (IgG1) anti-EGFR mAb cetuximab, have been proven effective in combination with chemotherapy or as single agent for the treatment of metastatic CRC (mCRC; refs. 2–5). Activating KRAS mutation has recently emerged as a major predictor of resistance to the EGFR-targeted mAbs, and patient selection based on KRAS mutational status allows more accurate treatment selection with improved efficacy, reduction of unnecessary toxicities, and improved overall cost effectiveness (5, 6). Although the KRAS mutation is a highly specific negative biomarker of response (93% specificity), it does lack sensitivity (47% sensitivity; ref. 7). This indicates the existence of additional but, as of yet, unknown determinants of efficacy to the anti-EGFR mAbs. Previous studies have investigated additional determinants of EGFR mAb sensitivity within the EGFR signaling network, including BRAF mutational status (8), epiregulin and amphiregulin mRNA expression (9), high EGFR gene copy number (10), loss of PTEN protein expression (11), and PIK3CA mutation status (12), in wild-type (wt) KRAS mCRC patients treated with cetuximab. Although several of these molecular markers seem promising, their utility as predictive determinants will require evaluation in prospective clinical trials.

Insulin-like growth factor 1 (IGF1) signaling mediated by IGF1 receptor (IGF1R) is an important growth-regulatory pathway that plays a crucial role in CRC cell proliferation, migration, and apoptosis (13–17). IGF1 is a potent mitogenic activator via the Ras/Raf/mitogen-activated protein kinase signaling pathway and a powerful antiapoptotic molecule through the phosphatidylinositol 3-kinase (PI3K)/Akt pathway (18). An analysis of functional cross-talk between IGF1R and EGFR has shown that activation of the IGF1 downstream signaling cascade is crucial for the mitogenic and transforming activity of EGFR (19). More specifically, the IGF1R pathway induces both transforming growth factor α–mediated activation of EGFR and stimulation of EGFR-independent PI3K/Akt activity (20). Both cetuximab and the IgG2 EGFR-targeting mAb panitumumab function principally by inhibiting ligand binding to EGFR, thereby suppressing downstream signaling. Consequently, IGF1-driven PI3K/Akt overstimulation due to hyperactivation and/or pathway aberrations provides a rational explanation, at least in part, for the lack of efficacy observed in a notable fraction of patients with wt KRAS CRC treated with EGFR-targeting mAbs. Recently, Scartozzi et al. (17) reported that high IGF1 protein expression correlates with poor clinical outcome in wt KRAS mCRC patients treated with cetuximab and irinotecan. In addition, IGF1 and IGF1R polymorphisms have been associated with cancer risk (21, 22) and increased IGF1 plasma levels (23), suggesting functional and clinical significance.

The current study sought to evaluate whether functional polymorphisms in the IGF1 and IGF1R genes, alone or in combination, can augment the prediction of sensitivity to cetuximab treatment in drug-refractory wt KRAS mCRC patients treated with single-agent cetuximab in a phase II clinical trial (IMC-0144).

Formalin-fixed, paraffin-embedded (FFPE) tumor tissue of 130 (38%) of 346 mCRC patients enrolled in a multicenter, multinational, open-label, phase II trial of cetuximab in patients with drug-refractory mCRC (IMCL-0144) was available for analysis of IGF1 and IGF1R polymorphisms. Patients were enrolled from November 2002 to December 2005. Cetuximab was administered as a 120-minute intravenous infusion at 400 mg/m² followed by weekly 60-minute infusions of 250 mg/m². Eligibility for the IMCL-0144 study required that mCRC patients failed chemotherapy consisting of oxaliplatin, irinotecan, and fluoropyrimidines (4). The tissue analysis presented in present study was conducted at the University of Southern California/Norris Comprehensive Cancer Center (USC/NCCC) following approval by USC Institutional Review Board for Medical Sciences. All patients provided their written informed consent for tissue and blood collection to allow study of molecular correlates.

Objective tumor response was assessed every 6 weeks during the course of the study, and criteria were based on modified WHO guidelines (4). Response to cetuximab was determined by an independent response assessment committee that was blinded to the investigator-reported measurements, and assessments were reported in the study as previously reported (4). A partial response (PR) required at least a 50% reduction in the sum of the bidimensional products of all measurable lesions documented at least 4 weeks apart. Treatment was continued in the absence of intolerable toxicity or progressive disease, defined as at least a 25% increase in measurable disease, unequivocal growth of existing nonmeasurable disease, the appearance of one or more new lesions, or reappearance of old lesions.

DNA was extracted from FFPE tumor samples using the QIAamp kit (Qiagen). The genes, reference single-nucleotide polymorphisms (SNP) identification numbers, location, function, forward and reverse primer, and restriction enzymes are summarized in Table 1. The polymorphisms we tested were selected by an IGF1 pathway approach with the goal of selecting common and functional SNPs previously associated with other tumors. The following criteria were used to select the candidate polymorphisms: (a) a minor allele frequency ≥10% in Caucasians according to the HapMap Project database (http://www.hapmap.org), (b) potential functional polymorphisms located in the 3′UTR (untranslated region) and/or were shown to be of biological significance according to the literature review, and (c) were associated with cancer risk and/or IGF1 plasma level in previous studies (Table 1). KRAS mutation status was determined from microdissected tumor DNA by direct sequencing as previously described (24). IGF1 pathway polymorphisms were tested by using either PCR-RFLP technique or direct sequencing as previously described. For quality control purposes, a total of 5% positive and negative duplicate controls were matched for each polymorphism and were analyzed by direct DNA sequencing where applicable. Genotype concordance was ≥99%.

IGF1 and IGF1R polymorphisms were related to parameters reflecting clinical outcome, including progression-free survival (PFS), OS, and objective response as defined in the clinical trial (7). The PFS was calculated from the date of the first date of cetuximab treatment until disease progression or death from any cause or censored at the last follow-up if the patients were still alive and not progressed. The OS time was the period from the first day of cetuximab infusion until death from any cause or the last follow-up, at which point the data were censored. The objective response rate (RR) was defined as the total number of complete response or PR divided by the number of evaluable patients.

The distributions of polymorphisms across baseline demographic and clinical characteristics were examined using Fisher's exact test. The association between each polymorphism with OS and PFS was analyzed using Kaplan-Meier curves and the log-rank test. RR was analyzed using contingency tables and the exact conditional test. Subgroup analysis was conducted in patients carrying wt KRAS only. The false discovery rate (FDR) of multiple testing was controlled using the Benjamini and Hochberg method (25). The FDR-adjusted P values <15% were considered as significant. In the multivariable Cox proportional hazards regression model, known prognostic, predictive variables, such as KRAS mutation status, and polymorphisms that had significant associations with clinical outcome in our previous study were included to investigate the independent effects of IGF1 and IGF1R polymorphisms. The genetic model of inheritance for IGF1 and IGF1R polymorphisms was unknown, so we considered the dominant, recessive, co-dominant, or additive model whenever appropriate.

A regression tree method based on recursive partitioning (RP) was used to identify homogenous subgroups for tumor response from genetic markers we previously tested (24).

Allelic distribution of all polymorphisms in each race/ethnic group was tested for deviation from Hardy-Weinberg equilibrium (HWE). Linkage disequilibrium among polymorphisms in IGF1 and IGF1R was assessed using D′ and r² values, and the haplotype frequencies of the two genes were inferred using HaploView version 4.1 (http://www.broad.mit.edu/mpg/haploview). All statistical tests were two-sided and performed using the SAS statistical package version 9.2 (SAS Institute, Inc.).

DNA was extracted for analysis from 130 (38%) of 346 eligible tumor specimens. The median PFS [1.3 months; 95% confidence interval (CI), 1.3-1.5], OS (6.3 months; 95% CI, 4.3-7.7), and RR (9.2%; 95% CI, 4.9-15.6%) of patients with eligible tumor samples were similar to those patients on trial without tissue and blood samples available (n = 216): median PFS (1.5 months; 95% CI, 1.4-2.6), OS (6.8 months; 95% CI, 5.8-8.1), and RR (13.0%; 95% CI, 8.8-18.2%; ref. 4). The median follow-up for this translational study was 12.3 months (range, 2.2-17.3 months). Sixteen (12%) of 130 patients were not evaluable for tumor response. In 114 patients assessable for tumor response, 77 patients had wt KRAS and 37 patients had mutant KRAS. The allelic frequencies observed for IGF1 and IGF1R polymorphisms analyzed were within the probability limits of HWE (P > 0.05). The patient characteristics, KRAS mutation status, and clinical outcome were described previously and summarized in Table 2 (4, 24).

In univariate analysis, three IGF1 polymorphisms—IGF1 rs6214, IGF1 rs2946834, and IGF1 rs7136446—were significantly related to PFS in all patients (P = 0.048, P < 0.001, and P = 0.034, respectively). Two of these polymorphisms, IGF1 rs2946834 and IGF1 rs713664, were also related to PFS in the subgroup of wt KRAS patients (P < 0.001 and P = 0.022, respectively). Univariate analysis for OS was subsequently performed. One IGF1 (rs7136446) and the recessive model of two IGF1R polymorphisms (IGF1R rs2272037 and IGF1R rs2016347) were significantly associated with shorter OS in all patients (P = 0.026, P = 0.039, and P = 0.038, respectively), and IGF1R rs2016347 was significantly associated with shorter OS in the subgroup of wt KRAS patients only (P = 0.004; Table 3).

Response to treatment with cetuximab monotherapy was related to SNPs in IGF1 and IGF1R genes. Three polymorphism—IGF1 rs6214, IGF1 rs2946834, and IGF1R2016347—were significantly related to reduced responsiveness to cetuximab treatment (Table 4). RP was used to construct a decision tree as a predictive model to classify patients based on the presence of these molecular markers and the likelihood of response to cetuximab. This comprehensive RP analysis incorporated a total of 16 potential markers including the IGF1 and IGF1R polymorphisms in the current study, a panel of polymorphisms previously evaluated in this patient cohort (FCGR2Ars1801274 and FCGR3Ars396991, EGFRrs11543848, CyclinD1rs17852153, IL8rs4073, VEGFrs3025039, COX-2rs20417, COX-2rs5275, EGFrs4444903, and NRP-1rs3750733; ref. 24), and the development of skin rash during cetuximab treatment. In the resultant decision tree, the most important factor that determined the RR in these patients was the IGF1 rs2946834 polymorphism (node 1). Patients carrying IGF1 rs2946834 A/A genotype had a RR of 50% to cetuximab treatment. Patients in node 2 or 3 segregated based on their COX2-765 genotype had RR of 18% versus 5%, respectively (Fig. 1).

The multivariable model was adjusted by skin rash severity, KRAS mutation, Eastern Cooperative Oncology Group (ECOG) performance status, and three significant polymorphisms previously published by our group (EGFRrs11543848, COX-2rs5275, and EGFrs4444903; ref. 24) and stratified by race. IGF1 rs6214 (adjusted P = 0.008), IGF1 rs6220 (adjusted P = 0.028), and IGF1 rs2946834 (adjusted P = 0.002) remained significantly associated with PFS, and IGF1R rs2016347 (adjusted P = 0.033) remained significantly associated with OS (Table 5). Because these four polymorphisms were consistently associated with either PFS or OS, an unfavorable genotype analysis was performed to determine the effect of the number of risk alleles (Table 5; Fig. 2). Both the additive models for PFS and OS were significantly associated with the number of risk alleles [hazard ratio (HR), 2.028; 95% CI, 1.478-2.783; P < 0.001 versus HR, 1.828; 95% CI, 1.272-2.627; P = 0.001, respectively].

The multivariable model for the wt KRAS subgroup was adjusted for the same factors mentioned above. IGF1 rs6214 (adjusted P = 0.044), IGF1 rs2946834 (adjusted P = 0.009), and the number of risk alleles (P = 0.004) remained significantly associated with PFS (Table 4). No polymorphism was found to be significant for OS. Using unfavorable genotype analysis, the additive model showed a significant association with the number of risk alleles for PFS (HR, 1.895; 95% CI, 1.299-2.765; P = 0.001) and OS (HR, 1.745; 95% CI, 1.092-2.787; P = 0.02; Table 5).

After adjusting for the FDR at <15% level, IGF1 rs6214 (FDR-adjusted P = 0.022), IGF1 rs2946834 (FDR-adjusted P = 0.007), and the number of risk alleles (FDR-adjusted P = 0.001) remained significant for PFS. IGF1 rs6214 (FDR- adjusted P = 0.043) and the number of risk alleles (FDR- adjusted P = 0.009) also remained significant for OS in all patients. In wt KRAS patients, IGF1 rs2946834 (FDR-adjusted P = 0.038) and the number of risk alleles (FDR-adjusted P = 0.007) remained significant for PFS.

IGF1 rs2946834 and IGF1 rs7136446 variants showed linkage disequilibrium, with D′ value of 0.8 and r² value of 0.44. Haplotypes were constructed from these two polymorphisms. However, there were no significant relationships between these variants and the clinical outcome parameters PFS, OS, and RR.

The presence of activating mutations in KRAS has been associated with resistance to the anti-EGFR mAbs in the treatment of mCRC (5). However, only 10% to 40% of chemorefractory mCRC patients with wt KRAS respond to anti-EGFR mAbs, indicating the presence of additional determinants of sensitivity and resistance. The results of these translational studies using tumor tissues from chemorefractory mCRC patients receiving cetuximab monotherapy in a multicenter phase II clinical trial provide evidence that IGF1 and IGF1R polymorphisms are significantly associated with PFS, OS, and RR, particularly in wt KRAS patients. These preliminary findings also provide support for the IGF1R-EGFR cross-talk hypothesis described by Hu and colleagues (20). These results remained significant after adjusting for other potential predictors of patients' outcome and adjusting the FDR for multiple comparisons. To the best of our knowledge, this is the first report showing that IGF1 pathway polymorphisms may serve as potential predictive and/or prognostic determinants for mCRC patients undergoing cetuximab monotherapy, particularly those with wt KRAS tumors.

Several studies have suggested that the IGF1 pathway is a key mediator of resistance to cytotoxic chemotherapy and anti-EGFR treatment (26–28). In the present study, chemorefractory wt KRAS mCRC patients possessing the IGF1 rs2946834 variant A/A genotype had a significantly higher RR to cetuximab of 50% compared with 0% RR for those with the A/G genotype. Moreover, the effect of IGF1 rs2946834 variant A/A genotype in determining response was further supported by subsequent RP analysis that included a total of 16 current and previously analyzed markers. The IGF1 rs2946834 polymorphism was identified as the most important determinant of response to cetuximab in the decision tree analysis. Collectively, these data provide strong evidence suggesting that the IGF1 rs2946834 variant A/A genotype is a predictive factor for cetuximab efficacy in wt KRAS mCRC patients (Fig. 1). A potential explanation for these findings relates to the fact that the IGF1 rs2946834 polymorphism is located in the 3′UTR of the IGF1 gene and that 3′UTR plays an important role in regulating mRNA stability via the presence of several regulatory elements, including selenocysteine insertion sequences, mRNA binding protein sites, and microRNA binding sites. Alteration of mRNA sequence by the presence of a polymorphic variant in the 3′UTR can have measurable effects on mRNA stability and translational expression (29).

Previous investigations have shown that the IGF1 rs2946834 A/G genotype is associated with the highest circulating IGF1 plasma level in breast cancer patients (30). In addition, high circulating level of IGF1 was correlated with increased risk (31), enhanced tumor growth, and metastasis in CRC (32). Hyperactivation of the IGF1R pathway by IGF1 has been associated with resistance to several chemotherapeutics, particularly cisplatin and etoposide (27, 28), and anti-EGFR (26) mAbs through continued activation of PI3K signaling. Scartozzi et al. (17) showed that wt KRAS CRC with IGF1-positive protein expression had significantly lower RRs to cetuximab and irinotecan than tumors with IGF1-negative protein expression (22% versus 65%; P = 0.002). These findings are consistent with our results, which show that the IGF1 rs2946834 A/G genotype is associated with resistance to cetuximab treatment in wt KRAS patients. We hypothesize that the IGF1 rs2946834 G allele leads to increased expression of IGF1 and the subsequent hyperactivation of IGF1R, resulting in EGFR-independent stimulation of the PI3K pathway.

The association between IGF1 and IGF1R polymorphisms and clinical outcome has recently been investigated and established in various other cancer types. Zhang et al. (33) showed that a 3′UTR polymorphism in IGF1 predicts survival of non–small cell lung cancer patients. Dong et al. (34) suggested that genetic variations in the IGF pathway predict worse survival in patients with pancreatic cancer. Additionally, the IGF1 and IGF1R polymorphisms evaluated in the present study were recently shown to be associated with an increased cancer risk and/or elevated IGF1 plasma levels (Table 1), further suggesting functional and clinical significance. We identified five of six variants of IGF1 and IGF1R to be significantly associated with PFS and/or OS in univariate and multivariate analysis (Tables 3 and 5). In multivariate analysis, wt KRAS patients treated with cetuximab had a significant increase in their risk for tumor progression with two IGF1 polymorphisms [rs6214 and rs2946834; HR, 2.155 (95% CI, 1.023-4.540) and 3.066 (95% CI, 1.316-7.143), respectively] and in the combined risk allele analysis (additive model P = 0.001; Fig. 2B). Although these results require further validation, our findings also suggest a rationale for combination treatment with both EGFR and IGF1R inhibitors (35, 36).

Several mAbs and small molecules targeting IGF1R and related signal transduction pathways are currently undergoing preclinical and clinical evaluations (37). These therapeutics inhibit either the binding of IGF1 to its receptor or IGF1R tyrosine kinase activity. Clinical studies have shown that the IGF1R is overexpressed in ∼90% of CRC compared with normal tissues, suggesting a potentially broad therapeutic application for agents targeting this pathway (38). In addition, IGF1 signaling has been shown to protect tumor cells from EGFR-targeted treatment. Indeed, recent evidence suggests that resistance to the anti-Her2 mAb trastuzumab in breast cancer (39) and the anti-EGFR antibody AG1478 in human glioblastoma cells (26) is due to activation of IGF1R signaling and that blockage of IGF1R can restore sensitivity. These studies further support the notion of IGF1R-EGFR pathway cross talk and the possibility that inhibition of both pathways may be required to achieve complete and sustained PI3K/Akt inhibition.

The retrospective design and relatively small numbers of patients involved in the present translational analysis indicate that the results should be considered hypothesis generating and confirmed in prospective randomized controlled clinical trials. Nevertheless, it should be stressed that the results remained significant even after adjusting for a total of 16 markers in RP and after adjusting the FDR for multiple comparisons in multivariate analysis. These data provide the first evidence that polymorphisms within the IGF1 pathway are significantly associated with RR, PFS, and OS in wt KRAS mCRC patients treated with cetuximab monotherapy. Moreover, these observations may aid in the selection of patients with an increased likelihood of cetuximab response or who are candidates for combined EGFR and IGF1R treatment.

Dr. Lenz has received honoraria from Merck KG and Bristol-Myers Squibb. Eric Rowinsky is employed by Imclone Systems, Inc.; David Mauro is employed by Merck Co., Inc.; Christiane Langer is employed by and has an ownership interest in Bristol-Myers Squibb.

Grant Support: NIH grant 5 P30CA14089-27I and Dhont Family Foundation. T. Winder is supported in part by a Research Grant of the Austrian Society of Hematology and Oncology and the "Kurt und Senta-Herrman Foundation." This work was performed in the Sharon A. Carpenter Laboratory at USC/NCCC and in memory of David Donaldson.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.