Annals of Surgical Oncology 14(7):2150–2158

DOI: 10.1245/s10434-007-9409-5

Laboratory Research

Pancreatic Cancer Epidermal Growth Factor Receptor (EGFR) Intron 1 Polymorphism Influences Postoperative Patient Survival and in vitro Erlotinib Response Ching-Wei D. Tzeng, MD,1 Andrey Frolov, MD, PhD,1 Natalya Frolova, MD,2 Nirag C. Jhala, MD,2 J. Harrison Howard, MD,1 Selwyn M. Vickers, MD,3 Donald J. Buchsbaum, PhD,4 Martin J. Heslin, MD,1 and J. Pablo Arnoletti, MD1

1

Department of Surgery, University of Alabama at Birmingham, Birmingham, AL, USA Department of Pathology, University of Alabama of Birmingham, Birmingham, AL, USA 3 Department of Surgery, University of Minnesota, Minneapolis, MN, USA 4 Department of Radiation Oncology, University of Alabama at Birmingham, Birmingham, AL, USA 2

Background: Epidermal growth factor receptor (EGFR) has a highly polymorphic CA repeat region that affects transcription efficiency and anti-EGFR drug sensitivity in carcinomas. Erlotinib is an EGFR tyrosine kinase inhibitor approved for pancreatic cancer treatment. We analyzed the impact of EGFR intron 1 CA repeat lengths in pancreatic cancer clinical outcome and in vitro response to erlotinib. Methods: Allele-specific EGFR intron 1 lengths were analyzed in 30 microdissected pancreatic cancer surgical specimens, matched peripheral blood samples, and 9 pancreatic cancer cell lines treated with erlotinib. CA repeat lengths were correlated with survival, tumor parameters, molecular markers of EGFR pathway activation, and in vitro antiproliferative effects of erlotinib. Results: Both patient samples and cell lines displayed the full spectrum of EGFR CA repeat lengths (14–22 per allele). Patients with shorter sum of total CA repeats (<36) had worse median survival than patients with ‡36 repeats (13.7 vs 30.6 months, P = .002). Shorter patient EGFR intron 1 length correlated with EGFR expression (P = .026). Tumor intron 1 length was identical to that of matched peripheral blood specimens. There was no correlation between EGFR intron 1 length and pancreatic cancer stage, nodal status, grade, or expression of p-EGFR, p-ERK and p-Akt. Shorter EGFR intron 1 length was associated with in vitro response to erlotinib treatment (P = .02). Conclusions: Shorter EGFR intron 1 CA repeat length is associated with worse pancreatic cancer clinical prognosis and in vitro response to erlotinib. EGFR intron 1 length can be reliably measured in peripheral blood and may translate into a quantitative predictive marker of both pancreatic cancer aggressiveness and erlotinib sensitivity. Key Words: EGFR—Epidermal growth factor receptor—Pancreatic cancer—Intron 1—Polymorphism—Erlotinib.

Received January 18, 2007; accepted March 2, 2007; published online: April 24, 2007. Presented at The Society of Surgical Oncology 60th Annual Cancer Symposium, Washington, DC, March 17, 2007. Recipient of the SSO Harvey Baker Traveling Fellow Award and the SSO Resident/Fellow Essay Award for Best Basic Science Research Paper. Address correspondence and reprint requests to: J. Pablo Arnoletti, MD; E-mail: [email protected]

Pancreatic cancer remains the fourth deadliest cancer in the United States, with an estimated 33,730 new cases and 32,300 deaths expected in 2006.1 Long-term survival approaches 5%, regardless of therapeutic interventions. This has led to the study of molecular targeting therapy with the hope of improving response to traditional cytotoxic agents. The epidermal growth

Published by Springer Science+Business Media, LLC
2007 The Society of Surgical Oncology, Inc.

2150

2151

PANCREATIC CANCER EGFR INTRON 1 POLYMORPHISM

factor receptor (EGFR) is overexpressed in up to 60% of pancreatic cancers,2 making it a logical target for anti-EGFR molecular targeting. Erlotinib is an EGFR tyrosine kinase inhibitor (TKI) recently approved for treatment of advanced pancreatic cancer. A recent phase III clinical trial showed a modest but statistically significant improvement in overall survival when erlotinib plus gemcitabine were compared with gemcitabine alone.3 In this study, 53% of analyzed patients expressed EGFR protein by immunohistochemistry (IHC). However, as described in other cancers, there was no statistical correlation between pancreatic tumor EGFR expression and tumor response to EGFRtargeted therapy. It has therefore become evident that EGFR protein expression by itself does not constitute an accurate marker of tumor-related EGFR pathway activation. In addition, unlike non-small-cell lung cancer, EGFR activating tyrosine kinase mutations, which have been correlated with clinical prognosis and TKI sensitivity, are either nonexistent or exceedingly rare in pancreatic cancer.4–7 There remains a need to identify reliable surrogate markers of active EGFR pathway signaling that may reflect pancreatic cancer EGFR pathway dependence and sensitivity to antiEGFR therapy. EGFR intron 1 has a highly polymorphic CA repeat region (typically consisting of 14–22 repeats, but as diverse as 9–26 repeats), which affects transcription efficiency and anti-EGFR drug sensitivity in colorectal,8 head and neck,9 and breast cancers.10 In these carcinomas, increased CA repeat length decreased transcription efficiency by theoretically affecting the promoter/enhancer region around exon 1.11 Thus, with decreased CA repeat length, there was greater EGFR transcript level and, more importantly, improved tumor response to anti-EGFR drug therapy. Another very important finding was the consistently identical length of EGFR intron 1 CA repeats analyzed in matched tumor and normal tissues, indicating this was a germline polymorphism, rather than a tumor-specific somatic mutation.8 This raises the possibility of employing EGFR intron 1 length as a clinical biomarker that can be easily assessed in normal tissues, including peripheral blood samples. In the present study, we analyzed the association between EGFR intron 1 CA repeat length and the clinical outcome of 30 surgically treated pancreatic cancer patients. We then tested the influence of intron 1 CA repeat length on the in vitro pancreatic cancer cell response to erlotinib therapy to see if this polymorphism could translate into a marker of both clinical prognosis and therapeutic response. Finally, we sought to corroborate past studies in other car-

TABLE 1. Patient characteristics

Age Less than 70 70 or greater Race White Black Sex Male Female Tumor location Head Body/tail Stage I II III IV Node positive (n = 26) Resection (n = 26) R0 R1 Positive perineural invasion Pathologic grade (n = 24) Well differentiated Moderately differentiated Poorly differentiated

n

%

24 6

80 20

28 2

93 7

14 16

47 53

26 4

87 13

4 25 0 1 8

13 83 0 3 31

21 4 8

84 16 27

5 14 5

21 58 21

cinomas, in which tumor EGFR intron 1 length could be accurately estimated with matched peripheral blood samples. MATERIALS AND METHODS Patients and Tumor Specimens After IRB-approved informed consent was obtained, tumor specimens were collected from 30 pancreatic cancer patients who underwent laparotomy with curative intent at the University of Alabama at Birmingham. Twenty-nine patients underwent surgical resection. One patient was found to have unresectable disease at the time of laparotomy. Patient demographics and tumor characteristics are provided in Table 1. Specimens were snap-frozen at the time of operation and stored at )80C. Followup was obtained from hospital records, except in 2 cases, for which the information was unavailable. One patient who died within 30 days of operation was excluded from survival analysis. Human Pancreatic Cancer Cell Lines S2-013 and S2-VP10 cell lines, cloned sublines of SUIT-212 (a gift from Dr. Michael Hollingsworth, University of Nebraska Medical Center), were culAnn. Surg. Oncol. Vol. 14, No. 7, 2007

2152

C. W. D. TZENG ET AL.

tured in DulbeccoÕs modified EagleÕs medium (DMEM) supplemented with L-glutamine and 10% fetal bovine serum (FBS), in a 37C incubator with 5% CO2. ASPC-1, BxPC-3, CAPAN-1, HPAC, HPAF-II, MIA PaCa-2, and PANC-1 were obtained from the American Type Culture Collection (Rockville, MD) and propagated according to providerÕs recommendations. Laser Capture Microdissection of Patient Samples Fresh-frozen pancreatic cancer specimens were embedded in optimal cutting temperature compound before 8-lm sections were stained to visualize cells for microdissection. The sections were reviewed by an experienced pancreaticobiliary pathologist (N.C.J.) to localize and verify the presence of adenocarcinoma cells. Target cells were then selectively harvested using a PixCell II Laser Capture Microdissection (LCM) System (Arcturus, Mountain View, CA). Approximately 10,000, 7500, or 3000 adenocarcinoma cells were captured on each LCM cap for protein, DNA, or RNA extraction, respectively.

Patient Tissue and Cell Line DNA Extraction Microdissected tumor specimen DNA was isolated from the tumor cells on the LCM cap by incubating the surface of the cap with 50 lL of a DNA extraction solution at 56C overnight. The DNA extraction solution consisted of 100 mM Tris-HCl, 2 mM EDTA, 1% Tween-20, and 0.42 mg/mL Proteinase K. Genomic DNA from nine pancreatic cancer cell lines was isolated using the AquaPure Genomic DNA Isolation Kit (Bio-Rad, Hercules, CA).

Protein and RNA Extraction from LCM Specimens Total protein was extracted from each LCM cap by applying 30 lL of a protein buffer containing 50% ‘‘Buffer C,’’13 containing buffer of 10 mM Tris-HCl (pH 7.4), 0.1% Triton X-100, 1.5 mM EDTA, and 10% glycerol, and 50% 2x Laemmli Sample Buffer (Sigma, St. Louis, MO). The buffer was applied to the LCM cap for several minutes at room temperature before storage at )80C until Western blot analysis. Total RNA from the microdissected sections was isolated and purified with the RNAqueous Micro Kit (Ambion, Austin, TX) using the manufacturerÕs protocol. The yield of 18 lL of RNA solution per sample was subsequently stored at )80C. Ann. Surg. Oncol. Vol. 14, No. 7, 2007

EGFR Intron 1 Polymorphism (CA Repeat) Analysis After extraction, 25 ng of DNA was put into a PCR reaction using an unlabeled forward EGFR primer (5¢-GGGCTCACAGCAAACTTCTC-3¢) and a fluorescent HEX-labeled reverse EGFR primer (5¢AAGCCAGACTCGCTCATGTT-3¢).8 Conditions were as follows: initial denaturation at 95C for 5 min; 30 cycles of denaturation at 95C for 45 seconds, annealing at 60C for 45 seconds, and extension at 72C for 45 seconds, and final extension of 72C for 10 min. Final concentrations in a 50 lL PCR reaction were 5 lL of 10x PCR Buffer II (Applied Biosystems, ABI), 2.5 lM MgCl2, 200 lM each dNTP, 0.25 lM forward primer, 0.25 lM labeled reverse primer, 1.25 units AmpliTaq Gold Polymerase (ABI), and 1 lL DMSO. Genotypes were resolved on an ABI Prism 3130XL Genetic Analyzer (Applied Biosystems) to determine allele lengths and number of CA repeats. Cell Proliferation Assays after Treatment with Erlotinib Nine pancreatic cell lines were treated with 12 lM erlotinib for 48 hours. Five thousand cells were seeded in 100 lL of total media per well in a 96-well plate, and allowed to grow overnight before the first of two oncedaily treatments for a total 48-hour treatment period. At the end of the treatment period, CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI) was used per the manufacturerÕs guidelines to assess relative proliferation in control (untreated) versus treated cells for each cell line. All cell assays were performed in triplicate wells and repeated in triplicate independent runs. Immunoblot for Downstream Adaptor Proteins in Patient Samples Protein lysates from microdissected tumor specimens were prepared, and standard SDS-PAGE was performed as previously described.14 Primary antibodies were obtained for antiphosphorylated ERK1/ 2 (p42/44), and antiphosphorylated Akt (Serine 473) from Cell Signaling Technology (Beverly, MA), and for b-actin from Sigma. All primary antibodies were diluted 1:1000, and Western blot was performed according to the manufacturerÕs protocol.

Reverse Transcriptase PCR (RT-PCR) of EGFR CDNA was synthesized using a High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City,

2153

PANCREATIC CANCER EGFR INTRON 1 POLYMORPHISM

CA). Forward and reverse primers for EGFR were previously described.15 Primers for b-actin were forward 5¢-CTCACCATGGATGATGATATCGC-3¢ and reverse 5¢-CATGATGGAGTTGAAGGTAGTT TCGT-3¢. EGFR RT-PCR was performed in a 20 lL reaction containing iProof 5x GC buffer (Bio-Rad, Hercules, CA), 0.2 mM each dNTP, 0.2 lL iProof High Fidelity DNA Polymerase, 0.4 lL DMSO, 0.5 lM forward primer, 0.5 lM reverse primer, and 4 lL of cDNA. The reaction was initially denatured at 98C for 30 seconds, before 35 cycles of denaturing at 98C for 10 seconds, annealing at 57C for 30 seconds, and extension at 72C for 1 minute, with a final extension at 72C for 7 min. Similar reactions were performed for b-actin with annealing temperature of 55C. Immunohistochemistry for Receptor Activation in Patient Specimens Frozen tumor specimens were sectioned to 5 lm and affixed on slides in 10% formalin for 30 min before placing in a Tris buffer, containing 0.05 M Tris (pH 7.6), 0.15 M NaCl, and 0.01% Triton X-100, for 15 min. After endogenous peroxidase quenching with 3% H2O2 for 8 min and 10% goat serum blocking for 1 hour, the rabbit antiphospho-EGFR (Cell Signaling Technology, Danvers, MA) antibody was incubated at 1:25 dilution in PBE (PBS solution containing 1% BSA, 1 mM EDTA, and 0.01% NaN3, at pH 7.6) for 1 hour at room temperature. Secondary USA HRP 500 Test secondary biotinylated antibody (Signet Kit No. 2254) was applied for 10 minutes, followed by the matching ultrastreptavidin HRP complex for 5 minutes and diaminobenzidine (DAB) chromogen (BioGenex, San Ramon, CA) for 7 minutes. Slides were counterstained before dehydration and permanent mounting. Cells were graded by IHC intensity scores that ranged from 0 (none) to 4+ (strongest) and by percentage of tumor cells positive. The threshold percentage to be considered positive was 10%. EGFR Intron 1 Length, EGFR Pathway Activation, Prognostic Variables, and in vitro Sensitivity EGFR intron 1 length was retrospectively compared to survival, node-positivity, margin status, tumor grade, and stage. EGFR intron 1 length was also compared with presence or absence of the following molecular markers: EGFR transcript, p-EGFR, pERK1/2, and p-Akt. Allele-specific EGFR intron 1 lengths for nine pancreatic cell lines were compared to erlotinib sensitivity.

Matched Patient Blood Sample EGFR Intron 1 Length Matched patient blood samples were available for 10 patients. DNA from blood was extracted using AquaPure RBC Lysis Solution and AquaPure Genomic DNA Isolation Kit (Bio-Rad). PCR conditions and measurement of CA repeats were identical to methods used for tumor sample DNA. Statistical Analysis Survival was measured from day of surgery. All statistics were performed with SPSS software (SPSS, Chicago, IL). Cell line intron 1 status was compared with erlotinib sensitivity and EGFR relative transcript level with SpearmanÕs rho. Nonparametric tests were also used to correlate the following: erlotinib sensitivity, EGFR intron 1 length, EGFR, p-EGFR, pERK1/2 and p-Akt expression, node-positivity, margin status, tumor grade, and stage. Kaplan-Meier method and Life Table analysis were used for patient survival. Log-rank analysis was used to test differences in survival. Statistical significance was defined at P < .05. RESULTS Patient Characteristics Among 30 patients with tumor samples microdissected and analyzed for EGFR intron 1 length, 27 were included in the follow-up analysis. Median patient follow-up was 17.5 months. Median postoperative survival was 18.8 months (range 1.5–30.6), consistent with expected pancreatic cancer surgical outcomes.16 Median patient age was 63 (range 51– 83). Patient and tumor characteristics are detailed in Table 1. Distribution of EGFR Intron 1 length Among Pancreatic Cancer Patients The median allele-specific CA repeat length was 18 (range 14–21). The most frequent allele-specific length was 16, in 35% patients, followed by the length 20, in 28% patients (Fig. 1A). This corroborated the interethnic distribution seen by Liu et al.,17 who found that CA repeat length 16 was the most common allele-specific CA repeat length among Americans (42–43% of all alleles). When allele-specific CA repeats were added, the median sum of repeats was 36 (range 29–40), with 47% patients having less than 36 repeats. No sum of repeats represented more than Ann. Surg. Oncol. Vol. 14, No. 7, 2007

2154

C. W. D. TZENG ET AL.

Patient EGFR Intron 1 Length, EGFR Pathway Activation, and Clinical Prognosis

FIG. 1. Distribution of allele-specific EGFR intron 1 CA repeats for pancreatic cancer patients. (A) Our American patient cohort showed an expected predominance of the allele-specific CA repeat length 16, with a full range of 14–21. (B) Our patients were almost evenly distributed in number by the sum of both allelesÕ CA repeats, based on a cutoff of 36 CA repeats. About 47% have less than 36 total repeats (the median), and 53% are 36 or greater, with a wide range of 29–40.

20% of patients (Fig. 1B), producing a wide spectrum of total CA repeats, from 29 to 40. Patient EGFR Pathway Activation The majority of tumors (23 of 30, 77%) demonstrated EGFR transcript expression, consistent with past studies showing EGFR protein overexpression in most pancreatic cancer patients.2,3 Approximately half of the available EGFR-expressing tumors (11 of 21, 52%), representing one-third of all patients (11 of 30, 33%), displayed constitutively phosphorylated EGFR (p-EGFR). Downstream EGFR pathway activation was also present in a majority of pancreatic cancer patients, with 26 of 30 (87%) and 19 of 30 (63%) patients displaying p-ERK1/2 and p-Akt, respectively. Ann. Surg. Oncol. Vol. 14, No. 7, 2007

Pancreatic cancer patients with shorter sums of EGFR intron 1 CA repeats (<36) had worse median survival (13.7 months) than patients with ‡36 CA repeats (30.6 months, log-rank, P = .002). In Fig. 2, the Kaplan-Meier curve shows improved survival in patients with longer EGFR intron 1 CA repeat length. Actuarial postoperative survival at 12, 18, and 24 months, was 33%, 22%, and 0%, respectively, in the shorter intron 1 group, versus 86%, 66%, and 66%, respectively, in the longer intron 1 group. Patients with shorter EGFR intron 1 length were more likely to express EGFR transcript (12 of 14 patients, P = .026). While presence or absence of EGFR transcript did not correlate with worse clinical prognosis, shorter EGFR intron 1 CA repeat length, which is linked to improved EGFR transcription, was associated with worse overall survival. Univariate log-rank analyses of EGFR intron 1 length, stage, node-positivity, margin status, tumor grade, EGFR, p-EGFR, and p-Akt, revealed that only shorter EGFR intron 1 CA repeats (P = .002) negatively affected survival (Table 2). Because past studies in other cancers also examined allele-specific CA repeats in addition to the sum of both alleles, we analyzed specific lengths in the short and long alleles as well as the homozygous 16/16 and 20/20 genotypes (Table 3). Our analysis also suggested that EGFR intron 1 CA repeats greater than 16 in the long allele were associated with improved survival (P = .06).

Pancreatic Cancer Cell Line EGFR Intron 1 Length and Erlotinib Sensitivity The nine pancreatic cancer cell lines displayed the expected spectrum of CA repeat lengths (allele-specific range 15–22). Two cell lines, MIA PaCa-2 (43 total repeats) and PANC-1 (38), had total EGFR intron 1 CA repeat lengths greater than 36. This group with longer CA repeat lengths had median relative cell proliferation of 89% (range 82–98%) and exhibited resistance to erlotinib therapy. The pancreatic cancer cells with shorter sums of EGFR Intron 1 CA repeats (<36) displayed median relative cell proliferation of 55% (range 34–60%) following treatment with erlotinib. The degree of inhibition of proliferation showed a significant correlation between shorter CA repeat length and erlotinib sensitivity (P = .02), as seen in Fig. 3.

2155

PANCREATIC CANCER EGFR INTRON 1 POLYMORPHISM

TABLE 2. Analysis of factors for patient survival

n EGFR intron 1 CA repeats Less than 36 12 36 or greater 15 EGFR Negative 6 Positive 21 p-EGFR Negative 11 Positive 10 p-Akt Negative 11 Positive 16 Stage I 3 II 23 III 0 IV 1 Node positive Negative 18 Positive 8 Resection R0 20 R1 2 Pathologic grade Not poorly 18 differentiated Poorly differentiated 4

Hazard ratio % (HR)

Confidence interval (CI)

44 4.36 56

(1.46–12.95) .002

log-rank P-value

22 78 0.66

(0.18–2.38)

.41

52 48 0.79

(0.25–2.45)

.35

41 59 0.79

(0.27–2.28)

.58

11 85 1.97 (II/I) (0.42–9.27) 0 4

.15

69 31 2.13

(0.58–7.76)

.24

91 9 0.8

(0.12–5.26)

.83

(0.28–3.77)

.96

82 18 1.03

TABLE 3. Negative influence of allelic lengths on survival

n

FIG. 2. (A) Kaplan-Meier curve showing that our cohort of pancreatic cancer patients had a median survival of 18.8 months (n = 27), which is typical of surgically treated patients. (B) KaplanMeier curve for pancreatic cancer patients stratified by EGFR intron 1 total CA repeat length. Shorter total CA repeat length (lower line, n = 12) yielded 13.7-month median survival. Patients with longer total CA repeat length (upper line, n = 15) had not reached median survival at 30.6 months.

EGFR Intron 1 Length in Peripheral Blood and Matched Pancreatic Tumor Samples Allele-specific EGFR intron 1 CA repeat lengths were identical between 10 available microdissected tumor tissue specimens and their respective matched peripheral blood samples. This corroborated past studies in other tumor types, where EGFR intron 1 status could be reliably assessed from normal tissue, such as skin and blood.8 Blood analysis required <0.3 mL of peripherally drawn blood.

Sum of both alleles Less than 36 12 36 or greater 15 Short allele Less than 17 16 17 or greater 11 Short allele Less than 18 19 18 or greater 8 Long allele Less than 17 6 17 or greater 21 Homozygous 16/16 Not 16/16 23 16/16 4 Homozygous 20/20 Not 20/20 23 20/20 4

Hazard ratio (HR)

Confidence interval

Log-rank P-value

44 56

4.36

(1.46–12.95)

.002

59 41

2.05

(0.75–5.65)

.13

70 30

1.89

(0.64–5.59)

.24

22 78

2.62

(0.66–5.59)

.06

3.23

(0.46–22.9)

.06

3.32

(0.92–12.01)

.21

%

85 15 85 15

DISCUSSION This is the first study of pancreatic cancer patients to show that a shorter sum of CA repeats in a polymorphic region of EGFR intron 1 is associated with Ann. Surg. Oncol. Vol. 14, No. 7, 2007

2156

C. W. D. TZENG ET AL.

FIG. 3. A panel of nine pancreatic cancer cell lines shows the relationship between CA repeat length and relative cell proliferation under erlotinib treatment at 12 lM. Longer total CA repeat length (PANC-1 and MIA PaCa-2) was associated with erlotinib resistance. Even without separating the cell lines into two groups, an increased sum of CA repeats was associated with in vitro erlotinib resistance (P = .02). Allele-specific and total CA repeats are displayed for each cell line, as well as relative cell proliferation.

worse postoperative survival as well as in vitro inhibition of proliferation in erlotinib-treated pancreatic cancer cells. We suggest that EGFR intron 1 CA repeat length can be used as a surrogate marker for EGFR-dependent tumor aggressiveness and prognosis. We also propose that, as shown in other cancers with overexpressed molecular targets, these more aggressive tumors may be more susceptible to EGFRtargeted therapy.18–21 EGFR protein expression is associated with decreased survival in several epithelial cancers, such as lung, colorectal, and head and neck cancers.22–24 However, the consistent difficulty with the identification of EGFR expression has been its lack of power in predicting response to anti-EGFR treatment. In other words, the presence or absence of the target molecule (EGFR protein in this case), as determined by traditional methods such as IHC, cannot be consistently correlated with tumor response to specific inhibitors. This is certainly the case in pancreatic cancer.3 Even in lung cancer, where EGFR mutations play a significant role in aberrant pathway activation, there is a persistent need for reliable biomarkers beyond these mutations.25 The combination of patient survival analysis and in vitro erlotinib sensitivity we provide in our work Ann. Surg. Oncol. Vol. 14, No. 7, 2007

suggests that, in pancreatic cancer, EGFR intron 1 CA repeat length may provide an association between EGFR expression (as estimated by the polymorphismÕs influence on EGFR transcription efficiency) and anti-EGFR therapy sensitivity. EGFR intron 1 polymorphism affects EGFR expression through transcription efficiency via its location near a promoter and 2 enhancers around EGFR exon 1. While our study was not designed mechanistically to analyze EGFR expression levels (only to measure presence or absence of transcript), our patients did have an association between shorter EGFR intron 1 length and EGFR expression (P = .026), corroborating studies in other cancers. Early studies of the sequences around EGFR exon 1 showed that these polymorphic regions could regulate transcription.26 Haley et al. showed that transcription activity was highly concentrated in this CG-rich region, where bidirectional transcription originated.27 Gebhardt et al. described 80% transcription inhibition, by quantitative nuclear runoff, in alleles with 21 CA repeats (on the high end of the spectrum of repeats). Using PCR to test a region of EGFR, which included the promoter, 2 enhancers, and the CA repeat polymorphism region, they documented declining transcription activity with increasing CA repeat number. The authors acknowledge, in this and in another related study, that there are other transcription regulation mechanisms that can augment or negate this particular polymorphismÕs impact on transcription efficiency.28 Zhang et al. also pointed out that other polymorphisms that affected portions of the EGFR pathway, including cyclin D1 and EGF, could also affect EGFRÕs impact on prognosis and tumor sensitivity to anti-EGFR therapy.29 Across a spectrum of epithelial cancers, shorter EGFR intron 1 CA repeats are associated with more aggressive disease and, ultimately, decreased survival. For example, Zhang et al. studied 105 colorectal cancer patients and found that those with CA repeats <20 per allele were more likely to show disease progression despite chemotherapy when compared to patients with 20 or greater CA repeats (P = .019).30 Zhang et al. saw a similar trend in rectal cancer patients, who were more likely to have recurrences, despite chemoradiation, if they had CA repeats <20.31 Our pancreatic cancer patients fit this pattern of association between shorter EGFR intron 1 CA repeats and worse overall survival (P = .002). Buerger et al. also looked into the influence of each individual allele on both EGFR expression and prognosis. Looking at the shorter allele, shorter sequences correlated with higher EGFR levels in breast

PANCREATIC CANCER EGFR INTRON 1 POLYMORPHISM

tumors (P < .05). In tumors with loss of heterozygosity, the loss of the shorter allele yielded less EGFR protein than if the opposite happened.10 Their homozygous patients were also predominately 16/16 genotype (16 CA repeats on both alleles) with significantly more EGFR. Although some EGFR intron 1 studies focus on the sum of total CA repeats and others describe allele-specific CA repeats, the trend remains the same across different cancer types. Shorter CA repeat length is associated with worse prognosis, theoretically because of improved EGFR transcription efficiency. Table 3 details a summary of our allele- and genotype-specific survival analyses of our patients. The 16/16 genotype patients seem to have worse survival (P = .06), which is expected since their sum of repeats is 32, well below the proposed <36/ ‡36 stratification for worse/better prognosis. This association, however, did not reach statistical significance, perhaps due to the small sample size. While we could not demonstrate a significant influence of the shorter allele on patient survival, the length of the longer/second allele did show an association between CA repeat length shorter than 17 and worse clinical prognosis (P = .06). This is an intriguing epidemiological cutoff point as the most common allele among Americans. This means that American patients could be more likely to have genotype combinations involving shorter allele-specific repeat lengths, especially 16 (42–43%), when compared with Asian patients (17% alleles with 16 repeats).17 This could correspond with the suggestion that Americans have more aggressive pancreatic tumors compared to Asian patients.32 Further study of larger populations might better define what subpopulation of American patients would fit into this poorer prognosis (yet better erlotinib response) category. It is evident that alternative transcription regulation mechanisms, which are difficult to quantify, are likely to impact EGFR transcription efficiency, which results in variable EGFR protein expression levels. This may explain why EGFR intron 1 length correlates with EGFR transcript levels, but is inconsistently associated with EGFR protein expression.8 In addition, there are several possible translational and post-translational modifications that may influence EGFR protein levels, rendering EGFR protein an unreliable marker of tumor pathway dependence. User-dependent techniques such as IHC can also introduce bias in the process of quantifying EGFR protein expression. One interesting finding that supports tumor-specific transcriptional or translational

2157

changes affecting the role of EGFR intron 1 polymorphism was Etienne-Grimaldi et al.Õs finding that, while increasing CA repeats inversely correlated with EGFR expression in tumor samples, that was not the case for normal tissue.9 This could mean that, while normal tissue may be sampled to measure the tumorÕs EGFR intron 1 length, intron 1 polymorphismÕs influence on EGFR expression and EGFR pathway activation/dependence is muted, or not upregulated, in normal tissue. We agree that EGFR intron 1 polymorphism length may not correlate exactly with EGFR protein expression levels, but perhaps that is not its most relevant significance. We propose that it may act as a closely related, quantitatively measured surrogate marker for both adverse clinical prognosis and improved response to anti-EGFR therapy in tumors that exhibit EGFR pathway dependence. Successful patient selection for anti-EGFR therapy ultimately depends on identifying markers of pathway dependence, or ‘‘oncogene addiction,’’33 which are biologically relevant and clinically practical. In our patient population, we were able to confirm that EGFR intron 1 length is identical between microdissected surgical tumor specimens and matched peripheral blood DNA. Because most pancreatic cancer patients present with locally advanced and/or metastatic disease, evaluating tumor markers from peripheral blood instead of surgical specimens certainly has many advantages. One limitation of our study is the lack of erlotinibtreated pancreatic cancer patients, which limits our drug response studies to in vitro data. Unlike lung, colorectal, and head and neck cancers, where there have been multiple anti-EGFR trials, there is only one large anti-EGFR clinical trial reported for pancreatic cancer thus far.3 Ideally, the molecular characterization of a large population of erlotinibtreated pancreatic cancer patients would allow better identification of reliable markers of prognosis and response. In conclusion, shorter EGFR intron 1 CA repeat length is associated with worse clinical prognosis in surgically treated pancreatic cancer patients. However, as in other cancers with overexpressed molecular targets, shorter EGFR intron 1 CA repeat length is also associated with in vitro erlotinib sensitivity in pancreatic cancer. Adding to its practicality, EGFR intron 1 polymorphism status can be reliably identified in small peripheral blood samples. EGFR intron 1 CA repeat length may translate clinically into a quantitative predictive marker of pancreatic cancer aggressiveness, patient prognosis, and erlotinib sensitivity. Ann. Surg. Oncol. Vol. 14, No. 7, 2007

2158

C. W. D. TZENG ET AL.

ACKNOWLEDGMENT We would like to thank Maria Salazar in the UAB Core Sequencing Facility for her assistance with EGFR intron 1 polymorphism length analysis and Dr. Andra Frost for her guidance in the UAB Laser Microdissection Facility. This study was supported by the John W. Kirklin Research and Education Fellowship Award (J.P.A.), the James Ewing Oncology Fellowship Award for Basic Research (J.P.A.), and UABÕs P20 Pancreatic S.P.O.R.E. (CA10195-01).

15.

16. 17.

18. 19.

REFERENCES 1. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. CA Cancer J Clin 2006; 56(2):106–30. 2. Friess H, Wang L, Zhu Z, et al. Growth factor receptors are differentially expressed in cancers of the papilla of vater and pancreas. Ann Surg 1999; 230(6):767–74; discussion 74–5. 3. Moore M, Goldstein D, Hamm J, et al. Erlotinib plus gemcitabine compared to gemcitabine alone in patients with advanced pancreatic cancer. A Phase III trial of the National Cancer Institute of Canada Clinical Trials Group (NCICCTG). Proc Am Soc Clin Oncol 2005 2005; p. 24. 4. Lynch TJ, Bell DW, Sordella R, et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med 2004; 350(21):2129–39. 5. Kwak EL, Jankowski J, Thayer SP, et al. Epidermal growth factor receptor kinase domain mutations in esophageal and pancreatic adenocarcinomas. Clin Cancer Res 2006; 12(14 Pt 1):4283–7. 6. Immervoll H, Hoem D, Kugarajh K, Steine SJ, Molven A. Molecular analysis of the EGFR-RAS-RAF pathway in pancreatic ductal adenocarcinomas: lack of mutations in the BRAF and EGFR genes. Virchows Arch 2006; 448(6):788– 96. 7. Tzeng C, Frolov A, Frolova N, et al. Epidermal growth factor receptor (EGFR) is highly conserved in pancreatic cancer. Surgery 2007; 141(4):464–9. 8. Amador ML, Oppenheimer D, Perea S, et al. An epidermal growth factor receptor intron 1 polymorphism mediates response to epidermal growth factor receptor inhibitors. Cancer Res 2004; 64(24):9139–43. 9. Etienne-Grimaldi MC, Pereira S, Magne N, et al. Analysis of the dinucleotide repeat polymorphism in the epidermal growth factor receptor (EGFR) gene in head and neck cancer patients. Ann Oncol 2005; 16(6):934–41. 10. Buerger H, Gebhardt F, Schmidt H, et al. Length and loss of heterozygosity of an intron 1 polymorphic sequence of egfr is related to cytogenetic alterations and epithelial growth factor receptor expression. Cancer Res 2000; 60(4):854–7. 11. Gebhardt F, Zanker KS, Brandt B. Modulation of epidermal growth factor receptor gene transcription by a polymorphic dinucleotide repeat in intron 1. J Biol Chem 1999; 274(19):13176–80. 12. Iwamura T, Katsuki T, Ide K. Establishment and characterization of a human pancreatic cancer cell line (SUIT-2) producing carcinoembryonic antigen and carbohydrate antigen 19-9. Jpn J Cancer Res 1987; 78(1):54–62. 13. Simone NL, Remaley AT, Charboneau L, et al. Sensitive immunoassay of tissue cell proteins procured by laser capture microdissection. Am J Pathol 2000; 156(2):445–52. 14. Frolov A, Chahwan S, Ochs M, et al. Response markers and the molecular mechanisms of action of Gleevec in

Ann. Surg. Oncol. Vol. 14, No. 7, 2007

20. 21. 22.

23.

24.

25. 26.

27.

28.

29.

30.

31.

32.

33.

gastrointestinal stromal tumors. Mol Cancer Ther 2003; 2(8):699–709. Bor MV, Sorensen BS, Rammer P, Nexo E. Calibrated userfriendly reverse transcriptase-PCR assay: quantitation of epidermal growth factor receptor mRNA. Clin Chem 1998; 44(6 Pt 1):1154–60. Cameron JL, Riall TS, Coleman J, Belcher KA. One thousand consecutive pancreaticoduodenectomies. Ann Surg 2006; 244(1):10–5. Liu W, Innocenti F, Chen P, Das S, Cook EH Jr, Ratain MJ. Interethnic difference in the allelic distribution of human epidermal growth factor receptor intron 1 polymorphism. Clin Cancer Res 2003; 9(3):1009–12. Mendelsohn J, Baselga J. Epidermal growth factor receptor targeting in cancer. Semin Oncol 2006; 33(4):369–85. Ono M, Kuwano M. Molecular mechanisms of epidermal growth factor receptor (EGFR) activation and response to gefitinib and other EGFR-targeting drugs. Clin Cancer Res 2006; 12(24):7242–51. Meric-Bernstam F, Hung MC. Advances in targeting human epidermal growth factor receptor-2 signaling for cancer therapy. Clin Cancer Res 2006; 12(21):6326–30. Richter M, Zhang H. Receptor-targeted cancer therapy. DNA Cell Biol 2005; 24(5):271–82. Hirsch FR, Varella-Garcia M, Bunn PA Jr, et al. Molecular predictors of outcome with gefitinib in a phase III placebocontrolled study in advanced non-small-cell lung cancer. J Clin Oncol 2006; 24(31):5034–42. Galizia G, Lieto E, Ferraraccio F, et al. Prognostic significance of epidermal growth factor receptor expression in colon cancer patients undergoing curative surgery. Ann Surg Oncol 2006; 13(6):823–35. Hitt R, Ciruelos E, Amador ML, et al. Prognostic value of the epidermal growth factor receptor (EGRF) and p53 in advanced head and neck squamous cell carcinoma patients treated with induction chemotherapy. Eur J Cancer 2005; 41(3):453–60. Baselga J, Arteaga CL. Critical update and emerging trends in epidermal growth factor receptor targeting in cancer. J Clin Oncol 2005; 23(11):2445–59. Haley J, Whittle N, Bennet P, Kinchington D, Ullrich A, Waterfield M. The human EGF receptor gene: structure of the 110 kb locus and identification of sequences regulating its transcription. Oncogene Res 1987; 1(4):375–96. Haley JD, Waterfield MD. Contributory effects of de novo transcription and premature transcript termination in the regulation of human epidermal growth factor receptor protooncogene RNA synthesis. J Biol Chem 1991; 266(3):1746–53. Gebhardt F, Burger H, Brandt B. Modulation of EGFR gene transcription by a polymorphic repetitive sequence—a link between genetics and epigenetics. Int J Biol Markers 2000; 15(1):105–10. Zhang W, Gordon M, Press OA, et al. Cyclin D1 and epidermal growth factor polymorphisms associated with survival in patients with advanced colorectal cancer treated with Cetuximab. Pharmacogenet Genomics 2006; 16(7):475–83. Zhang W, Stoehlmacher J, Park DJ, et al. Gene polymorphisms of epidermal growth factor receptor and its downstream effector, interleukin-8, predict oxaliplatin efficacy in patients with advanced colorectal cancer. Clin Colorectal Cancer 2005; 5(2):124–31. Zhang W, Park DJ, Lu B, et al. Epidermal growth factor receptor gene polymorphisms predict pelvic recurrence in patients with rectal cancer treated with chemoradiation. Clin Cancer Res 2005; 11(2 Pt 1):600–5. Longnecker DS, Karagas MR, Tosteson TD, Mott LA. Racial differences in pancreatic cancer: comparison of survival and histologic types of pancreatic carcinoma in Asians, blacks, and whites in the United States. Pancreas 2000; 21(4):338–43. Weinstein I. Cancer Cell Addiction to oncogenes—the Achilles heal of cancer. Science 2002; 297(5578):63–4.