Molecular Alterations in Tumors and Response to Combination Chemotherapy with Gefitinib for Advanced Colorectal Cancer

The epidermal growth factor receptor (encoded by the EGFR gene, OMIM #131550) tyrosine kinase mediates cellular responses to growth signals through RAS, signal transduction and activator of transcription (STAT), and protein kinase C pathways. Inhibition of EGFR by either oral tyrosine kinase inhibitors or monoclonal antibodies has shown tumor regression in patients with lung and colorectal adenocarcinomas (1). Activating mutations of EGFR have been discovered in non–small cell lung cancers (NSCLC) that were sensitive to EGFR tyrosine kinase inhibitor gefitinib (ZD1839, “Iressa”) but not in gefitinib-resistant cancers (2, 3). These discoveries have opened possibilities for targeted therapies against other tumors with activating EGFR mutations, as well as against tumors with alterations in downstream events that might predict responsiveness to EGFR-directed therapy. NSCLC cells expressing mutant EGFR undergo extensive apoptosis after small interfering RNA against mutant EGFR or treatment with pharmacologic inhibitors of AKT and STAT signaling but are relatively resistant to conventional chemotherapeutic agents (4), emphasizing the importance of the EGF/AKT/STAT pathways in NSCLC. In contrast, mutations in KRAS have been shown to predict resistance of lung adenocarcinoma to gefitinib and erlotinib (5).

Depending on methods of detection, ∼50% to 80% of human colorectal cancers have been shown to express EGFR, suggesting that EGFR represents an attractive target in colorectal cancer patients (6, 7). Phase II trials of monoclonal antibodies directed against EGFR's extracellular domain have shown objective tumor responses in patients with colorectal cancer (8–10). Among patients with advanced colorectal cancer, a combination of gefitinib with chemotherapy showed a 78% response rate among chemotherapy-naive patients and 36% among those who had progressed through first-line therapy (11). A number of combinations of radiation and/or chemotherapy with agents directed against EGFR have been under investigation either in vitro or in vivo (12–20).

Unlike NSCLC, it remains unclear whether activating mutations in EGFR have a role in colorectal carcinogenesis. EGFR mutations were identified in 4 of 33 (12%) patients with colorectal cancer in Japan (21), although in other studies, an EGFR mutation was identified in only 1 of 293 (0.34%) colorectal cancer cases (22) and in none of 20 colorectal cancer cases (2). Moreover, no study to date has examined the influence of EGFR mutations on the responsiveness to gefitinib-based therapy in advanced colorectal cancer. We therefore investigated the influence of EGFR mutations and protein expression as well as other events related to downstream pathways [including alterations of KRAS, BRAF, p21 (CDKN1A/CIP1), p27 (CDKN1B/KIP1), p53 (TP53), and phosphorylated AKT1] on the treatment response among patients who were enrolled in phase I/II trials of gefitinib and chemotherapy for chemotherapy-naive metastatic colorectal cancer.

Patient selection and treatment. Patients for this analysis were enrolled into one of two phase I/II trials that examined the maximum tolerated dose of daily oral gefitinib in combination with i.v. irinotecan, leucovorin, and 5-fluorouracil (5-FU) in previously untreated metastatic colorectal cancer. The study has been approved by the Dana-Farber Harvard Cancer Center Institutional Review Board. All patients provided written informed consent, permitting us to do genomic/genetic analysis on their samples, before participation in the clinical trials. In both trials, patients had histologically confirmed metastatic colorectal cancer. Prior therapy for metastatic disease was not permitted, and all patients had measurable disease based on the Response Evaluation Criteria in Solid Tumors (23). Eligibility criteria also included the following: age of ≥18; Eastern Cooperative Oncology Group performance status of 0 to 2; life expectancy of ≥12 weeks; adequate hematopoietic function (absolute neutrophil count of ≥1,500/mm³, platelets ≥ 100,000/mm³); and adequate renal (serum creatinine ≤ 1.5 mg/dL) and hepatic function (serum bilirubin ≤ 1.5 mg/dL).

Patients in the first trial (gefitinib plus bolus IFL) received daily oral gefitinib in combination with irinotecan 100 (or 125) mg/m², leucovorin 20 mg/m², and 5-FU 400 (or 500) mg/m² delivered i.v. weekly for two consecutive weeks followed by 1-week rest. In the second trial (gefitinib plus FOLFIRI), daily oral gefitinib was given with irinotecan 180 mg/m² i.v. bolus, leucovorin 200 mg/m² i.v. bolus, 5-FU 400 mg/m² i.v. bolus followed by 5-FU 2400 mg/m² via continuous infusion over 46 hours; chemotherapy was repeated every 2 weeks. Therapy was continued until disease progression, unacceptable toxicity, or withdrawal of patient consent. In both trials, doses of oral gefitinib were started at 250 mg/d and the dose was escalated to 500 mg/d in cohorts of three patients. In both studies, the maximum tolerated dose of gefitinib was 250 mg/d.

Assessment of objective tumor response. In each trial, computed axial tomographic scans were done at baseline and every 6 weeks thereafter. Response evaluation was assessed according to the Response Evaluation Criteria in Solid Tumors (23).

Tumor tissue samples and genomic DNA extraction. Of 35 patients enrolled in the two trials, primary tumor specimens were available from 31 patients. Paraffin-embedded tissue blocks were obtained at the time of resection of the primary tumor. Tumor tissue (and normal tissue if applicable) was reviewed and dissected from tissue sections obtained from the paraffin tissue blocks, and genomic DNA was extracted using QIAmp DNA Mini Kit (Qiagen, Valencia, CA).

PCR and sequencing of EGFR, KRAS, and BRAF. For whole genome amplification, genomic DNA was PCR amplified using random 15-mer primers (24, 25). Each PCR mix contained 20 pmol of the random primers, 1.69 nmol each of deoxynucleotide triphosphate, 2.5 mmol/L of MgCl₂, 1× PCR buffer (Applied Biosystems, Foster City, CA), 0.5 unit of AmpliTaq Gold (Applied Biosystems), and 2 μL of template DNA solution in a total volume of 27 μL. PCR condition consisted of initial denaturing at 94°C (1 minute); 50 cycles of 94°C (30 seconds), 37°C (2 minutes), 41.5°C (30 seconds), 46°C (30 seconds), 50.5°C (30 seconds), 55°C (2 minutes), and 68°C (30 seconds); and final extension at 72°C (2 minutes) (25). PCR of EGFR was targeted for exons 18, 19, and 21 in the critical kinase domain. PCR Primers for EGFR, KRAS (exon 1), and BRAF (exon 15) were EGFR-E18F, 5′-TTGTGGAGCCTCTTACACCC-3′; EGFR-E18R, 5′-ATGAGAGGCCCTGCGGCCCA-3′; EGFR-E19F, 5′-TGCCAGTTAACGTCTTCCTTC-3′; EGFR-E19R, 5′-GGGCCTGAGGTTCAGAGCCA-3′; EGFR-E21F, 5′-ACCGTCGCTTGGTGCACCGC-3′; EGFR-E21R, 5′-GTCAGGAAAATGCTGGCTGA-3′; KRAS-F14, 5′-TGTAAAACGACGGCCAGTTGTGTGACATGTTCTAATATAGTCAC-3′; KRAS-R7, 5′-AGAATGGTCCTGCACCAGTAA-3′; BRAF-F7, 5′-TCCTTTACTTACACCTC-3′; and BRAF-R7, 5′-AAATAGCCTCAATTCTTACC-3′. Each PCR mix contained the forward and reverse primers (each 10 pmol), 1.69 nmol each of deoxynucleotide triphosphate, 3 mmol/L of MgCl₂, 1× PCR buffer, 0.75 unit of AmpliTaq Gold, and 2 μL of template whole genome amplification product in a total volume of 27 μL. PCR condition consisted of initial denaturing at 94°C (1 minute); 50 cycles of 95°C (20 seconds), annealing (20 seconds; 48°C for BRAF, 50°C for KRAS, 52°C for EGFR exon 19, 55°C for EGFR exons 18 and 21), and 72°C (40 seconds); and final extension at 72°C (2 minutes). The PCR products were purified using QIAquik PCR Purification Kit (Qiagen), cycle-sequenced with BigDye Terminator kit (Applied Biosystems), and analyzed by ABI 3730 (Applied Biosystems). All forward sequencing results were confirmed by reverse sequencing. KRAS sequencing was further validated by Pyrosequencing (25). A mutation and polymorphism in EGFR were further confirmed by a duplicated run from the original DNA samples. By our sequencing methods, we could detect EGFR mutations in NSCLC cell lines 3255 (with the p.Leu858Arg mutation in exon 21) and H1650 (exon 19 deletion mutation; both provided by Matthew Meyerson, Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA).

Immunohistochemical analyses. For EGFR immunohistochemistry, paraffin sections of colorectal cancers were deparaffinized, incubated with 3% H₂O₂ (20 minutes) to block endogenous peroxidase, and incubated with pepsin at 37°C (10 minutes). Protein block (Vector Laboratories, Burlingame, CA; 20 minutes) was followed by application of primary anti-EGFR antibody (Zymed Laboratories, South San Francisco, CA; dilution 1:100; overnight at 4°C). Then, secondary anti-mouse antibody (Vector Laboratories) was applied (20 minutes), avidin biotin complex was added, and sections were visualized by diaminobenzidine (5 minutes) and methyl-green counterstain. EGFR expression was recorded as negative (0), weakly positive (1+), positive (2+), or strongly positive (3+).

For phosphorylated AKT1 (p-AKT1) immunohistochemistry, antigen retrieval was done by incubating deparaffinized tissue sections in citrate buffer by a microwave at 92.8°C for 15 and 30 minutes. Tissue sections were then incubated with 3% H₂O₂ (30 minutes) to block endogenous peroxidase and incubated with protein block (Vector Laboratories; 30 minutes). Primary anti-p-AKT1 antibody (Cell Signaling, Beverly, MA; dilution 1:50) was applied for overnight at 4°C. Then, secondary anti-rabbit antibody (Vector Laboratories) was applied (30 minutes), avidin biotin complex was added, and sections were visualized by 3,3′-diaminobenzidine (5 minutes) and methyl-green counterstain. p-AKT1 expression was interpreted as negative or positive, using normal epithelial cells and lymphocytes as reference.

For p21 (CDKN1A/CIP1) and p27 (CDKN1B/KIP1) immunohistochemistry, antigen retrieval was done by incubating deparaffinized tissue sections in citrate buffer at high power in a microwave for 30 minutes (in a pressure cooker) and 15 minutes, respectively. Tissue sections were then incubated with 3% H₂O₂ (10 minutes) to block endogenous peroxidase and incubated with protein block (Vector Laboratories; 10 minutes). Primary anti-p21 antibody (PharMingen, San Diego, CA; dilution 1:50) or anti-p27 antibody (Transduction Laboratories, San Diego, CA; dilution 1:200) was applied for 30 minutes at room temperature. Then, biotinylated secondary Multi-Link antibody (Biogenex, San Ramon, CA) was applied (20 minutes), horseradish peroxidase avidin complex (Biogenex) was added, and sections were visualized by 3,3′-diaminobenzidine (30 seconds) and methyl-green counterstain. p21 expression was recorded as negative (<10% of cells staining) or positive (≥10% of cells staining). p27 expression was recorded as negative (≤20% of cells staining) or positive (>20% of cells staining).

We assessed for mutations in p53 (TP53) by immunohistochemistry. Methods for p53 immunohistochemistry were previously described (26). Only strong and unequivocal staining with ≥10% of cells staining was interpreted as positive. Weak or no staining for p53 was regarded as negative.

All immunohistochemical and molecular results were interpreted as blinded from patients' identity and clinical outcomes data.

Statistical analysis. Statistical analysis was done using the SAS program (version 9.1, SAS Institute, Cary, NC). Given the limited sample size of patient population, Fisher's exact test was used for the analysis on categorical data. Statistical significance was set at P = 0.05. All P values are two sided.

Clinical and pathologic characteristics. Primary tumor specimens were obtained from 31 evaluable patients enrolled in two trials assessing the efficacy and safety of a combination of gefitinib with irinotecan, leucovorin, and 5-FU in previously untreated metastatic colorectal cancer. All 31 patients (22 men and 9 women) were Caucasians with age ranging from 31 to 83 years (median, 57 years). Clinical characteristics of these patients are listed in Table 1. Forty-three percent and 57% of patients had Eastern Cooperative Oncology Group performance status of 0 and 1, respectively, and Eastern Cooperative Oncology Group performance status was not a predictor of chemoresponsiveness. All 31 patients had measurable disease by the Response Evaluation Criteria in Solid Tumors; 12 patients (39%) experienced a partial response as their best response; 16 patients (52%) experienced stable disease; and one patient (3%) experienced only progressive disease. Responsiveness to therapy was not associated with pathologic features such as tumor differentiation, or presence of mucinous or signet ring cell component (data not shown).

Epidermal growth factor receptor mutation and protein expression. Of 31 available tumor specimens from patients treated with gefitinib and chemotherapy, we successfully sequenced exons 18, 19, and 21 in 30 specimens (Table 1). Of the 30 colorectal cancer tumors analyzed for gene mutations, we found one novel mutation of EGFR, c.2170G>A (p.Gly724Ser; Fig. 1A-B), in a tumor with wild-type KRAS and BRAF. This mutation resides in exon 18, the critical kinase domain of EGFR. We confirmed the presence of this mutation by duplicating all steps of the PCR-sequencing assays and excluded the possibility of a germ line polymorphism by sequencing normal DNA from the same patient. This c.2170G>A (p.Gly724Ser) mutation in the kinase domain confers a glycine-to-serine substitution at codon 724. Notably, a previously described mutation at codon 719 (c.2155G>A, p.Gly719Ser) also confers a glycine-to-serine substitution and has been associated with gefitinib sensitivity in NSCLC (3). However, in the current study, the patient whose colorectal cancer possessed the c.2170G>A (p.Gly724Ser) mutation did not experience any disease regression following treatment with gefitinib and chemotherapy; the patient achieved stable disease for only 6 months before developing disease progression. In a second patient, we found a germline polymorphism, c.2203G>A (p.Ala735Thr), in exon 18, and confirmed its presence in both the tumor and normal tissue (Fig. 1C). Interestingly, in this tumor, the normal allele seemed deleted to create a loss of heterozygosity pattern. However, this second patient also did not experience any disease regression following treatment with gefitinib and chemotherapy and achieved stable disease for only 6 months before developing disease progression. No other patient was found to harbor a somatic EGFR exon 18, 19, or 21 mutation.

We also examined tumor expression of EGFR as measured by immunohistochemistry (Table 2). Of the 27 tumors for which EGFR expression could be evaluated, 12 (44%) exhibited 2+ or 3+ (strong) EGFR expression and 11 (41%) exhibited 1+ (weak) expression. However, levels of EGFR expression were not associated with tumor responsiveness to gefitinib and chemotherapy (P = 0.45; Table 2).

KRAS and BRAF status and response to therapy. We considered the possibility that responsiveness to gefitinib and chemotherapy may reflect other somatic events in the tumors. Among the 30 tumors that were sequenced for mutations in KRAS, we observed mutations in 10 patients [33%; four were c.38G>A (codon 13 GAC; p.Gly13Asp), three c.34G>A (codon 12 AGT; p.Gly12Ser), two c.35G>A (codon 12 GAT; p.Gly12Asp), and one c.35G>T (codon 12 GTT; p.Gly12Val)]. Moreover, using sequence analysis, we found two cases with mutations in BRAF [6.7%; both were c.1799T>A (p.Val600Glu; previously called as “V599E”) mutations]. Of note, we found no significant association between KRAS status (Table 2) or BRAF status and response to gefitinib and chemotherapy.

Expression of p21 (CDKN1A/CIP1), p27 (CDKN1B/KIP1), p53, and phosphorylated AKT1 and response to therapy. We examined alterations in events that are potentially downstream from EGFR, including p21, p27, p53, and p-AKT1. Loss of p21 expression was found in 43% of cases and was associated with a higher response rate to therapy; 67% of cases with p21 loss experienced an objective tumor response, whereas 25% of cases with p21 positivity responded to therapy (P = 0.05; Table 2). Furthermore, we observed a striking trend towards resistance to therapy among patients whose tumors maintained p21 expression and had a mutation in p53. Only 1 of 11 (9%) patients with p21 expression and mutated p53 experienced a partial response, whereas 11 of 16 (69%) patients with either p21 loss or wild-type p53 responded to therapy (P for comparison = 0.005; Table 2). With regard to p-AKT1, 14% of tumors with p-AKT1 overexpression responded to therapy, whereas 50% of tumors without p-AKT1 responded, although this difference did not reach a level of statistical significance (P = 0.19). In contrast, expression of p27 was not related to response to therapy.

Recent research efforts have found that activating mutations in the EGFR kinase domain represent an important predictor for responsiveness to gefitinib in NSCLC (2, 3). These specific activating mutations have been limited to exons 18, 19, and 21, which are in the tyrosine kinase domain of EGFR (2, 3). Activating EGFR mutations selectively activate AKT and STAT signaling pathways (4). NSCLC cells expressing mutant EGFR undergo extensive apoptosis after small interfering RNA against mutant EGFR or treatment with pharmacologic inhibitors of AKT and STAT signaling but are relatively resistant to conventional chemotherapeutic agents (4).

Lynch et al. (2) sequenced EGFR exons 19 and 21 but not exon 18, in colorectal cancer cell lines (COLO-205, HCT-116, HCT-15, HT-29, and SW-620) and found no mutations. Barber et al. (22) similarly identify only one EGFR mutation (p.Gly719Ser) in 293 cases of colorectal cancer. In contrast, Nagahara et al. (21) recently identified somatic EGFR mutations in 4 of 33 (12%) cases of colorectal cancer. Our results confirm the rarity of activating EGFR mutation in colorectal cancer; we found one novel mutation (c.2170G>A, p.Gly724Ser) in exon 18 of EGFR (in the critical kinase domain). This mutation is particularly interesting because it represents an identical amino acid substitution (glycine to serine) as another well-described activating mutation (c.2155G>A, p.Gly719Ser; refs. 3, 22), which is only five amino acids away from the currently described mutation. Although functional analysis is lacking in this study, it remains a possibility that the p.Gly724Ser mutation represents an activating EGFR mutation given the importance of the glycine-to-serine substitution at codon 719. Nonetheless, in contrast to NSCLC, this EGFR mutation was identified in a patient who did not experience an objective tumor response to gefitinib and chemotherapy. In addition, EGFR expression, as measured by immunohistochemistry, was not associated with responsiveness to the chemotherapy with gefitinib. Studies of cetuximab (a monoclonal antibody to EGFR) in colorectal cancer has similarly failed to show a relation between EGFR expression and response to therapy (8).

Although several studies suggest that a monoclonal antibody to EGFR can result in tumor regression in patients with colorectal cancer (8–10), as a single agent, oral tyrosine kinase inhibitors such as gefitinib seem less active (27). The potential diminished activity of oral tyrosine kinase inhibitors in colorectal cancer may reflect the rarity of activating EGFR mutations in colorectal cancer, although other mechanisms of resistance may exist. Of note, if EGFR mutations are uncommon in colorectal cancer, other downstream events such as alterations in p-AKT1 or p21 may influence responsiveness to EGFR inhibition.

We observed a significantly lower response to gefitinib and chemotherapy among patients whose tumors had intact p21 expression. Moreover, the prediction for treatment resistance was most striking in patients whose tumors maintained p21 expression in the setting of mutated p53. Still, we cannot exclude the possibility that the inverse relation between p21 expression and response to gefitinib and chemotherapy was principally related to the effect of 5-FU and irinotecan rather than any interaction with gefitinib. However, in contrast to our findings, a recent review of 1,809 colorectal cancer cases in the published literature suggest that, overall, greater p21 expression predicted a greater responsiveness to fluoropyrimidine-based chemotherapy that did not include an EGFR inhibitor (28). We are unaware of any other study that has examined predictors of responsiveness to gefitinib-based therapy in patients with colorectal cancer. Previous studies have shown that overexpression of p21-activated kinase-1 (PAK1) is associated with greater metastatic potential in colorectal cancer (29) and that gefitinib can suppress PAK1 pathway and the subsequent invasiveness of cancer cells (30). Taken together, these data could suggest that p21 overexpression may overcome the inhibitory effect of gefitinib through activation of PAK1 pathway. Gefitinib has been shown to up-regulate p27 and p21 in head and neck squamous cell cancer cells, and suppression of p21 and p27 by antisense constructs can decrease the growth inhibitory effect of gefitinib (31), indicating the important roles of p21 and p27 in modulating the effect of gefitinib.

Previous studies have examined the influence of p53 and p-AKT1 status on interventions that inhibit EGFR. MDM2 antisense oligonucleotides, which potentiate p53 activity, enhance the effect of EGFR inhibitors by affecting in vitro and in vivo proliferation, apoptosis, and protein expression in hormone-refractory and hormone-dependent human prostate cancer cells (32). Moreover, combination of gefitinib and antisense MDM2 synergistically inhibits the growth of hormone-independent prostate cancer cells, and this effect is accompanied by the inhibition of p-AKT1 (32), suggesting a role for intact p53 in augmenting the effect of gefitinib. Our results are consistent with these previous data; we observed that loss of functional p53 limited the response to gefitinib and chemotherapy, particularly in tumors with intact p21.

With regard to p-AKT1, previous studies have shown that response to gefitinib is associated with p-AKT1 overexpression in NSCLC (33–36). In contrast, our results showed a trend towards gefitinib sensitivity in p-AKT1-negative colorectal cancer. However, different mechanisms likely underlie sensitivity to EGFR inhibition in colorectal cancer than those observed in NSCLC. Gefitinib can down-regulate p-AKT1 or up-regulate p27 expression in some patients with metastatic colorectal cancer when pretreatment and posttreatment tumor specimens are compared (37). Further investigation is necessary to confirm the effect of p53 and p-AKT1 on gefitinib sensitivity in colorectal cancer.

In conclusion, activating EGFR mutations are rare in colorectal cancer and do not seem to confer responsiveness to gefitinib and chemotherapy. Nonetheless, p21 expression in colorectal cancer, especially in combination with p53 mutation, seems a predictor of resistance to the combination of chemotherapy with gefitinib. Further studies of EGFR inhibitors in colorectal cancer should examine the role of such downstream events on treatment efficacy.

Microsatellite instability was assessed in 28 patients, and all 28 tumors showed stable microsatellites of BAT25, BAT26, and BAT40 markers. Methods of microsatellite instability analysis are described in Ogino et al. (38).

We thank Matthew Meyerson for providing the control cell lines with the EGFR mutations and patients who participated in the clinical trials.