Integrin Genetic Variants and Stage-Specific Tumor Recurrence In

ORIGINAL ARTICLE Integrin genetic variants and stage-speciﬁc tumor recurrence in patients with stage II and III colon cancer

P Bohanes1, D Yang2, F Loupakis1, MJ LaBonte1, A Gerger1, Y Ning1, C Lenz1, F Lenz1, T Wakatsuki1, W Zhang1, L Benhaim1, A El-Khoueiry1, R El-Khoueiry1 and H-J Lenz1

Integrins (ITGs) are key elements in cancer biology, regulating tumor growth, angiogenesis and lymphangiogenesis through interactions of the tumor cells with the microenvironment. Moving from the hypothesis that ITGs could have different effects in stage II and III colon cancer, we tested whether a comprehensive panel of germline single-nucleotide polymorphisms (SNPs) in ITG genes could predict stage-specific time to tumor recurrence (TTR). A total of 234 patients treated with 5-fluorouracil-based chemotherapy at the University of Southern California were included in this study. Whole-blood samples were analyzed for germline SNPs in ITG genes using PCR–restriction fragment length polymorphism or direct DNA sequencing. In the multivariable analysis, stage II colon cancer patients with at least one G allele for ITGB3 rs4642 had higher risk of recurrence (hazard ratio (HR) = 4.027, 95% confidence interval (95% CI) 1.556–10.421, P = 0.004). This association was also significant in the combined stage II–III cohort (HR = 1.975, 95% CI 1.194–3.269, P = 0.008). The predominant role of ITGB3 rs4642 in stage II diseases was confirmed using recursive partitioning, showing that ITGB3 rs4642 was the most important factor in stage II diseases. In contrast, in stage III diseases the combined analysis of ITGB1 rs2298141 and ITGA4 rs7562325 allowed to identify three distinct prognostic subgroups (P = 0.009). The interaction between stage and the combined ITGB1 rs2298141 and ITGA4 rs7562325 on TTR was significant (P = 0.025). This study identifies germline polymorphisms in ITG genes as independent stage-specific prognostic markers for stage II and III colon cancer. These data may help to select subgroups of patients who may benefit from ITG-targeted treatments.

The Pharmacogenomics Journal (2015) 15, 226–234; doi:10.1038/tpj.2014.66; published online 9 December 2014

INTRODUCTION attachment to the surrounding extracellular matrix and, for some Adjuvant chemotherapy has improved survival in patients with ITGs, also to other cells. ITGs are expressed on both tumor cells colon cancer. However, being active in patients with stage III colon and normal cells (for example, endothelial cells, platelets and cancer (and in a subgroup of stage II patients), adjuvant chemo- leucocytes). Therefore, they promote tumor growth not only by therapy fails to eradicate micro-metastatic disease in 30% of direct interaction of tumor cells with the microenvironment but patients.1,2 In contrast, as many as 30–40% of patients with stage also by regulating endothelial cell’s survival and migration during 6 III colon cancer are given unnecessary post-operative treatment angiogenesis and lymphangiogenesis. Recent evidence showing because their tumor has good intrinsic biologic behavior.3 For that ITGs are critical in cancer dormancy suggests that their stage II patients, there are currently no validated clinical param- differential expression or activity may be responsible for tumor eters or biomarkers that identify patients that should be given recurrence.7 post-operative chemotherapy. Therefore, both prognostic factors ITGs are heterodimeric cell surface receptors. Eighteen α- and and predictive factors for oxaliplatin and 5-fluorourcil efficacies are eight β-subunits associate to form 24 different non-covalently of critical importance. associated heterodimers with specific tissue distribution and Recent evidence indicates that molecular biomarkers may have unique ligand specificity.8 ITGs transmit signals in both direction a stage-specific distribution and influence in colon cancer,4,5 of the cell membrane: ITG binding to extracellular matrix changes which has led some authors to consider stage II and III colon the shape and composition of the cytoskeleton; intracellular signal cancer as distinct entities. This data has lead to active research to mechanisms activate the ITGs (conformational change) and identify key pathways that would be responsible for a different increase their affinity.9 These signals are integrated with those clinical behavior across colon cancer stages. A candidate group of originating from the growth factor receptors enabling a specific molecules is the adhesion cell receptors that interact with the cellular response in various biologic situations.10 Multiple ITG extracellular matrix, having a critical role in cell survival, prolifera- heterodimers have been shown to play a role in cancers, including tion, differentiation and migration. The major group of proteins colon cancer. However, most of the attention has been on αV promoting these interactions is the integrins (ITGs), a superfamily heterodimers, notably αVβ3, as it has been shown to be important of transmembrane-related adhesion receptors that regulate in tumor angiogenesis. Several drugs have thus been developed

1Sharon A. Carpenter Laboratory, Division of Medical Oncology, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA and 2Department of Preventive Medicine, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, Los Angeles, CA, USA. Correspondence: Dr H-J Lenz, Sharon A. Carpenter Laboratory, Division of Medical Oncology, Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California, 1441 Eastlake Avenue, Los Angeles, CA 90033, USA. E-mail: [email protected] Received 1 April 2014; revised 18 August 2014; accepted 28 August 2014; published online 9 December 2014 Integrin variants and stage II and III colon cancer P Bohanes et al 227 to target αV heterodimers.11 Another important heterodimer that of about 0.5. Deviation of allelic distribution of each ITG polymorphism has caught attention for drug development is α5β1 heterodimer, from Hardy–Weinberg equilibrium was tested using χ2-test in each ethnic as it has been involved in tumor angiogenesis and more recently group. The distribution of polymorphisms by baseline demographic, in cancer dormancy.12 However, recently, interest is growing on clinical and pathological characteristics was examined using Fisher’s exact other ITG heterodimers as many of them have been linked to test. The true mode of inheritance of all ITG polymorphisms tested is not known and this study assumed a codominant, additive, dominant or cancer prognosis. recessive genetic model wherever appropriate. The association between There is substantial genetic variability within the genes coding polymorphisms and TTR was analyzed using Kaplan–Meier curves and log- for the ITGs that may alter the gene expression and/or activity by rank test. In the multivariable Cox regression analysis, the model was affecting various steps in protein synthesis, including transcription, adjusted for stage and type of adjuvant chemotherapy, and stratified by translation or splicing. This, in turn, may alter the tumor behavior race. Other clinico-pathologic parameters were not entered in the model, and/or sensitivity to chemotherapy drugs, thereby causing inter- as they did not seem to be a significant prognostic impact in this cohort of 16 tumor stage or interindividual differences. Based on the hypoth- patients. Interactions between polymorphisms and stage on TTR were esis that specific germline single-nucleotide polymorphism (SNP) tested by comparing likelihood ratio statistics between the baseline and in ITG genes could have different effect across stage II and III colon nested Cox regression models that include the multiplicative product term. P-values for all polymorphisms were adjusted for multiple testing using a cancer patients, we investigated a comprehensive panel of ITG modified test of Conneely and Boehnke17 that was applied for the SNPs in patients with stage II and III colon cancer and linked them correlated tests due to linkage disequilibrium and the different modes of with outcome. Only ITG genes that have been previously asso- inheritance considered. We considered a pACT of o0.15 as being ciated with colon cancer were selected. potentially important given the candidate pathway approach. Recursive partitioning (RP), incorporating cross-validation, was used to explore and identify polymorphism profiles and interactions associated with TTR using MATERIALS AND METHODS the rPart-function in S-plus. Case-wise deletion for missing polymorphisms Eligible patients was used in univariate and multivariable analyses. In the RP analysis, all patients with at least one polymorphism result available were included. All A total of 234 patients with stage III and high-risk stage II colon cancer analyses were performed using the SAS statistical package version 9.2 (SAS were included in this study cohort. Whole blood was available from 206 Institute, Cary, NC, USA) and S-PLUS 7.0 (TIBCO, Palo Alto, CA, USA). patients. All patients were treated with 5-fluorouracil-based adjuvant chemotherapy at the Norris Comprehensive Cancer Center/University of Southern California or the Los Angeles County/University of Southern RESULTS California. At the completion of post-operative therapy, patients were followed clinically every 3 months for the first 2 years and then every The baseline characteristics of the 234 patients included in this 6 months. Annual follow-up computed tomography scans were performed. analysis are summarized in Table 1. The median age at time of Whole blood was collected at the time of diagnosis and stored at − 80 °C. diagnosis was 59 years (range 22–87 years), with a median follow- Patient data were collected retrospectively through chart review. The up time of 4.4 years (range 0.4–16.8 years). Ninety patients Institutional Review Boards at University of Southern California approved showed tumor recurrence, stage III and high-risk stage II proba- the study and all study participants signed informed consent for the bility of 3-year recurrence was 0.45 ± 0.047 and 0.21 ± 0.043, analysis of molecular correlates. This study was conducted adhering to the 13 respectively. Median overall survival has not been reached yet. REporting recommendations for tumor MARKer prognostic studies. The genotyping quality control by direct DNA sequencing provided a genotype concordance of ⩾ 99%. Genotyping was Candidate polymorphisms successful in at least 90% of cases for each polymorphism Common and putatively functional polymorphisms within the ITG genes analyzed. In failed cases, genotyping was not successful because were selected using public literature resources and databases (NCBI of limited quantity and/or quality of extracted genomic DNA. The PubMed, Ensemble). Pre-defined criteria were as follows: (a) minor allele allelic frequencies for all polymorphisms were within the frequency ⩾ 10% in Caucasians; (b) polymorphism could alter the function probability limits of Hardy–Weinberg equilibrium, except ITGB1 of the gene in a biologically relevant manner (either published data or rs2153875 in Asians, ITGB4 rs743554 in Whites and ITGB5 predicted function using Functional-Single-Nucleotide-Polymorphism 14,15 rs1803825 in Asians. There were no significant associations database, http://compbio.cs.queensu.ca/F-SNP); and (c) published clinical associations (Supplementary Table 1). between the polymorphisms and baseline demographic, clinical or pathological characteristics, except the allele distributions of ITGB1 rs2153875, ITGB4 rs743554, ITGB4 rs871443, ITGB5 Genotyping rs1803825, IGTA1 rs2270756, ITGA3 rs1062484 and IGTA4 Genomic DNA was extracted from peripheral blood using the QIAmp kit rs1143676 that varied significantly by race (data shown in (Qiagen, Valencia, CA, USA). The majority of the samples were tested using Supplementary Table S1). PCR-based restriction fragment length polymorphism analysis. Briefly, forward and reverse primers were used for PCR amplification, PCR products were digested by restriction enzymes (New England Biolab, Ipswich, MA, Univariate analysis for ITG SNPs and TTR USA) and alleles were separated on 4% NuSieve ethidium bromide-stained Stage II colon cancer. Only one polymorphism was significantly agarose gel. If no matching restriction enzyme could be found, samples associated with TTR in stage II colon cancer patients. Patients were analyzed by direct DNA sequencing. For quality control purposes, a carrying at least one G allele for ITGB3 rs4642 had a significantly total of 5% PCR–restriction fragment length polymorphism analyzed higher risk of recurrence (Table 2). samples were re-analyzed by direct DNA sequencing for each SNP. The investigator analyzing the polymorphisms was blinded to the clinical data set. Stage III colon cancer. Two polymorphisms demonstrated a significant association with TTR in stage III colon cancer patients. Patients harboring at least one G allele of ITGB1 rs2298141 Statistical analysis showed a higher risk of recurrence, while patients carrying the The endpoint of the study was time to tumor recurrence (TTR). TTR was homozygous T/T ITGA4 rs7562325 genotype had lower risk of calculated from the date of diagnosis of stage II or III colon cancer to the recurrence compared with both patients with the C/T and T/T date of first documented tumor recurrence. TTR was censored at the time of death or at the last follow-up if the patient did not develop disease genotype (Table 2). recurrence at that time. With 206 patients with specimen available for genotyping ITG SNPs, this study had an 80% power to detect a minimum Combined stage II and III colon cancer. In the whole study group, hazard ratio (HR) of 1.84–1.95 for TTR in a dominant model with the minor three polymorphisms were significantly associated with TTR. ITGB3 allele frequency of 0.2–0.5 or a recessive model with minor allele frequency rs4642, being already associated with stage II colon cancer’s

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 226 – 234 Integrin variants and stage II and III colon cancer P Bohanes et al 228 at least one G allele for ITGB3 rs4642 having a higher risk of Table 1. Baseline patient characteristics recurrence (HR = 4.027, 95% confidence interval (95% CI): 1.556– n % 10.421, P = 0.004 in stage II, Table 3 and Figures 1a and b). This association was not retrieved in stage III colon cancer. No Sex significant interaction was found between stages and ITGB3 Female 107 45.72 rs4642 on TTR (P = 0.12 Table 3). Male 127 54.28 In stage III colon cancer patients, both ITGB1 rs2298141 (HR = 1.909, 95% CI: 1.054–3.459, P = 0.033) and ITGA4 rs7562325 Ethnicity (HR = 0.260, 95% CI: 0.077–0.879, P = 0.030) (Figures 1c and d) Asian 34 14.53 remained significant. Only ITGA4 rs7562325 showed a significant African American 15 6.41 α Caucasian 123 52.56 interaction with stage on TTR (Table 3). As the 4 subunit binds Hispanic 62 26.5 mainly with the β1 subunit to form the α4β1 heterodimer, we combined both SNPs and were able to define three prognostic T stage groups of patients with stage III colon cancer (Figure 2). The inter- T1 2 0.85 action between stage and the combination of ITGB1 rs2298141 T2 14 5.98 and ITGA4 rs7562325 on TTR was significant (P = 0.025). The SNPs T3 187 79.91 combination was not associated with TTR in stage II patients (log- T4 27 11.54 rank P = 0.55). Tx 4 1.72 ITGB5 rs849019 remained significantly associated with TTR in – Grade the combined stage II III group, with heterozygous patients Well 11 2.18 having higher risk of recurrence compared with patients with Moderate 151 64.53 homozygous genotypes (HR = 1.743, 95% CI: 1.027–2.958, P = 0.039 Poor/undifferentiated 54 23.08 Table 3). This association did not remain significant when we Missing 18 10.21 applied the recessive model and therefore may represent a discovery by chance. N stage Negative 105 44.87 N1 72 30.77 Multiple testing for ITG SNPs and TTR N2 57 24.36 After adjusting the significant SNPs for multiple testing, including all ITG polymorphisms analyzed, only ITGB3 rs4642 in stage II AJCC Stage colon cancer had a pACT value below 0.15. High-risk II 105 44.87 III 129 55.13 Recursive partitioning for ITG SNPs and baseline characteristics Resected lymph nodes RP was used to construct a decision tree as a predictive model to ⩽ 12 70 29.91 classify patients in high- and low-risk subgroups based on the 4 12 145 61.97 gene variants and baseline characteristics (Figure 3). The most Missing 19 8.12 important factor that determined TTR in the resultant tree was Tumor location tumor stage. Left 110 47.01 In patients with stage II disease, the most important factor was Right 115 49.15 ITGB3 rs4642 polymorphism. Patients carrying the AA genotype Left and right 4 1.71 for ITGB3 rs4642 demonstrated a lower risk of tumor recurrence. Missing 5 2.13 In patients with stage III disease, the most important factor that determined clinical outcome was the number of resected lymph Adjuvant treatment nodes. The patients with stage III disease and ⩾ 12 lymph nodes 5-FU 151 64.53 resected could be further segregated based on their ITGA3 5-FU/LV/Oxaliplatin 60 25.64 5-FU/LV/Irinotecan 23 9.83 rs1062484 status, with patients harboring at least one T allele having a higher risk of tumor recurrence. Abbreviations: 5-FU, 5-fluorouracil; LV, leucovorin. DISCUSSION outcome, showed a significant association with TTR in the This study identifies several SNPs in ITG genes that seem to combined stage II–III cohort. Carriers of the G allele also have a distinct role in stage II and III colon cancer patients, demonstrated a higher risk of recurrence, albeit to a lesser degree independently of clinico-pathologic characteristics. Although in than in the stage II cohort. In the recessive model, carrying an A our study the rs4642 SNP located in the ITGB3 gene demon- allele for ITGB5 rs849019 was associated with higher risk of strates a predominant role in stage II colon cancer, both ITGB1 recurrence. In the co-dominant model, heterozygote patients for rs2298141 and ITGA4 rs7562325 SNPs have a predominant role in ITGA1 rs1531545 had lower risk of recurrence than homozygous stage III colon cancer. Moreover, our data suggest a distinct role patients for the wild and mutant allele (Table 2). This association between stages for the combined ITGB1 rs2298141 and ITGA4 fi did not remain significant when we applied the dominant or rs7562325, as we were able to demonstrate a signi cant recessive models and therefore may represent a discovery by interaction between stage II and III diseases. In contrast, it is chance (false positive). No significant association for both ITGB5 likely that ITGB3 rs4642 SNP has also an effect in stage III diseases, rs849019 and IGTA1 rs1531545 polymorphisms was present when albeit probably small, as the test for interaction by stage was not significant. looked by tumor stage. Our results suggest a predominant role of α4β1 in stage III colon cancer. To our knowledge, little information has been reported Multivariable analysis for ITG SNPs and TTR about α4β1 ITG heterodimer and colon cancer. However, these ITGB3 rs4642 remained significant in both stage II and the findings are consistent with recent data showing that α4β1 combined stage II–III colon cancer cohorts, with patients carrying expression in the tumor microenvironment is critical, promoting

Table 2. Polymorphisms of integrin genes and univariate analysis for TTR in patients with stage II or III colon cancer

Polymorphism N Stage II and III Stage II only Stage III only

3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) probability ± SE probability ± SE probability ± SE

ITGB1 rs2153875 A/A 86 0.34 ± 0.06 1 (Reference) 0.24 ± 0.08 1 (Reference) 0.40 ± 0.07 1 (Reference) A/C 93 0.31 ± 0.05 0.91 (0.56, 1.48) 0.18 ± 0.06 0.94 (0.40, 2.18) 0.44 ± 0.08 1.01 (0.55, 1.87) C/C 20 0.32 ± 0.11 1.05 (0.50, 2.21) 0.00 ± 0.00 0.48 (0.10, 2.23) 0.58 ± 0.16 1.75 (0.74, 4.14) P valuea 0.89 0.61 0.39

ITGB1 rs2298141 A/A 128 0.28 ± 0.04 1 (Reference) 0.19 ± 0.06 1 (Reference) 0.35 ± 0.06 1 (Reference) A/Gb 70 0.37 ± 0.06 1.20 (0.75, 1.92) 0.15 ± 0.06 0.69 (0.30, 1.62) 0.58 ± 0.09 1.82 (1.03, 3.23) G/Gb 4 P-valuea 0.45 0.39 0.036

ITGB3 rs3809865 A/A 92 0.29 ± 0.05 1 (Reference) 0.19 ± 0.06 1 (Reference) 0.38 ± 0.08 1 (Reference) A/T b 92 0.34 ± 0.05 1.35 (0.84, 2.16) 0.15 ± 0.06 0.96 (0.43, 2.15) 0.46 ± 0.07 1.59 (0.88, 2.88) T/T b 19 P-valuea 0.21 0.92 0.12

ITGB3 rs4642 A/A 93 0.25 ± 0.05 1 (Reference) 0.05 ± 0.04 1 (Reference) 0.41 ± 0.08 1 (Reference) A/Gb 86 0.37 ± 0.05 1.88 (1.16, 3.06) 0.27 ± 0.07 3.11 (1.28, 7.60) 0.44 ± 0.07 1.39 (0.78, 2.49) G/Gb 19 P-valuea 0.009 (0.17) 0.008 (0.14) 0.26

ITGB3 rs5918 T/T 152 0.31 ± 0.04 1 (Reference) 0.16 ± 0.05 1 (Reference) 0.42 ± 0.06 1 (Reference) T/Cb 47 0.35 ± 0.07 1.46 (0.89, 2.40) 0.19 ± 0.09 1.32 (0.54, 3.20) 0.46 ± 0.10 1.51 (0.83, 2.76) C/Cb 4 P-valuea 0.13 0.53 0.17

ITGB4 rs9367 T/T 107 0.26 ± 0.05 1 (Reference) 0.15 ± 0.05 1 (Reference) 0.36 ± 0.07 1 (Reference) T/Cb 80 0.41 ± 0.06 1.30 (0.82, 2.07) 0.21 ± 0.07 1.10 (0.48, 2.52) 0.52 ± 0.07 1.24 (0.70, 2.20) C/Cb 13 P-valuea 0.26 0.82 0.45

ITGB4 rs743554 G/G 140 0.27 ± 0.04 1 (Reference) 0.17 ± 0.05 1 (Reference) 0.37 ± 0.06 1 (Reference) G/A 46 0.42 ± 0.08 1.25 (0.73, 2.14) 0.17 ± 0.09 0.93 (0.34, 2.53) 0.56 ± 0.10 1.38 (0.73, 2.63) P-valuea 0.41 0.89 0.32

ITGB4 rs871443 C/C 102 0.31 ± 0.05 1 (Reference) 0.22 ± 0.06 1 (Reference) 0.38 ± 0.07 1 (Reference) C/T 76 0.35 ± 0.06 1.05 (0.64, 1.71) 0.16 ± 0.07 0.68 (0.28, 1.66) 0.47 ± 0.08 1.17 (0.64, 2.14) T/T 21 0.26 ± 0.10 1.06 (0.47, 2.39) 0.08 ± 0.08 0.55 (0.13, 2.40) 0.48 ± 0.18 1.90 (0.71, 5.09) P-valuea 0.98 0.56 0.42

ITGB5 rs849019 T/T 94 0.22 ± 0.05 1 (Reference) 0.12 ± 0.05 1 (Reference) 0.32 ± 0.08 1 (Reference) T/A 72 0.46 ± 0.06 1.98 (1.18, 3.31) 0.29 ± 0.09 1.70 (0.69, 4.20) 0.57 ± 0.08 1.92 (1.01, 3.62) A/A 28 0.28 ± 0.09 1.09 (0.52, 2.25) 0.09 ± 0.09 0.95 (0.26, 3.48) 0.40 ± 0.13 1.01 (0.42, 2.44) P-valuea 0.020 0.44 0.075 T/A or A/A 100 0.41 ± 0.05 1.66 (1.02, 2.70) 0.23 ± 0.07 1.42 (0.61, 3.30) 0.52 ± 0.07 1.59 (0.87, 2.92) Pvaluea 0.039 0.41 0.13 T/T or T/A 166 0.33 ± 0.04 1 (reference) 0.18 ± 0.05 1 (reference) 0.44 ± 0.06 1 (reference) A/A 28 0.28 ± 0.09 0.79 (0.40, 1.54) 0.09 ± 0.09 0.76 (0.22, 2.61) 0.40 ± 0.13 0.72 (0.32, 1.61) P-valuea 0.48 0.66 0.42

ITGB5 rs1803825 A/A 115 0.29 ± 0.05 1 (Reference) 0.18 ± 0.06 1 (Reference) 0.38 ± 0.07 1 (Reference) A/G 65 0.38 ± 0.06 1.47 (0.89, 2.43) 0.23 ± 0.08 1.71 (0.72, 4.09) 0.48 ± 0.09 1.31 (0.71, 2.42) G/G 21 0.25 ± 0.10 1.19 (0.53, 2.66) 0.08 ± 0.07 0.78 (0.18, 3.46) 0.50 ± 0.18 1.80 (0.68, 4.73) P-valuea 0.30 0.35 0.41

ITGA1 rs1531545 T/T 73 0.38 ± 0.06 1 (Reference) 0.26 ± 0.09 1 (Reference) 0.46 ± 0.08 1 (Reference) T/C 86 0.22 ± 0.05 0.57 (0.34, 0.97) 0.09 ± 0.04 0.66 (0.26, 1.64) 0.36 ± 0.08 0.61 (0.31, 1.18)

Table. 2. (Continued )

Polymorphism N Stage II and III Stage II only Stage III only

3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) probability ± SE probability ± SE probability ± SE

C/C 41 0.45 ± 0.09 1.15 (0.62, 2.11) 0.27 ± 0.12 1.34 (0.43, 4.20) 0.58 ± 0.12 1.18 (0.57, 2.44) P-valuea 0.035 0.35 0.17 T/C or C/C 127 0.29 ± 0.04 0.71 (0.44, 1.13) 0.14 ± 0.05 0.78 (0.33, 1.84) 0.43 ± 0.07 0.77 (0.43, 1.35) P-valuea 0.14 0.57 0.35 TT or T/C 159 0.29 ± 0.04 1 (Reference) 0.15 ± 0.05 1 (Reference) 0.41 ± 0.06 1 (Reference) C/C 41 0.45 ± 0.09 1.55 (0.89, 2.71) 0.27 ± 0.12 1.75 (0.64, 4.81) 0.58 ± 0.12 1.49 (0.76, 2.93) P-valuea 0.12 0.25 0.24 IGTA1 rs2270756 C/C 110 0.31 ± 0.05 1 (Reference) 0.13 ± 0.05 1 (Reference) 0.45 ± 0.07 1 (Reference) C/Ab 72 0.33 ± 0.06 1.08 (0.67, 1.73) 0.22 ± 0.08 1.34 (0.59, 3.07) 0.42 ± 0.08 0.94 (0.52, 1.69) A/Ab 16 P-valuea 0.75 0.47 0.83

IGTA2 rs1801106 G/G 162 0.31 ± 0.04 1 (Reference) 0.18 ± 0.05 1 (Reference) 0.41 ± 0.06 1 (Reference) G/A b 35 0.39 ± 0.08 1.49 (0.87, 2.54) 0.14 ± 0.09 1.22 (0.45, 3.29) 0.56 ± 0.11 1.68 (0.89, 3.19) A/A b 3 P-valuea 0.14 0.69 0.10

IGTA2 rs1126643 C/C 75 0.38 ± 0.06 1 (Reference) 0.26 ± 0.09 1 (Reference) 0.45 ± 0.08 1 (Reference) C/T 104 0.29 ± 0.05 0.68 (0.41, 1.11) 0.09 ± 0.04 0.47 (0.19, 1.20) 0.46 ± 0.07 0.90 (0.50, 1.60) T/T 22 0.26 ± 0.10 0.74 (0.34, 1.62) 0.25 ± 0.13 1.58 (0.56, 4.44) 0.25 ± 0.15 0.32 (0.08, 1.37) P-valuea 0.28 0.053 0.27

ITGA3 rs1062484 C/C 135 0.28 ± 0.04 1 (Reference) 0.16 ± 0.05 1 (Reference) 0.38 ± 0.06 1 (Reference) C/T b 57 0.40 ± 0.07 1.47 (0.92, 2.35) 0.20 ± 0.08 1.58 (0.70, 3.56) 0.55 ± 0.09 1.41 (0.79, 2.50) T/T b 7 P-valuea 0.10 0.27 0.24

IGTA4 rs1449263 T/T 52 0.41 ± 0.08 1 (Reference) 0.18 ± 0.09 1 (Reference) 0.57 ± 0.11 1 (Reference) T/C 99 0.26 ± 0.05 0.86 (0.48, 1.53) 0.12 ± 0.05 1.09 (0.38, 3.17) 0.36 ± 0.07 0.74 (0.37, 1.48) C/C 47 0.39 ± 0.08 1.20 (0.64, 2.26) 0.30 ± 0.11 1.80 (0.59, 5.52) 0.46 ± 0.10 0.93 (0.43, 2.03) P-valuea 0.47 0.46 0.64

IGTA4 rs1143676 A/A 109 0.32 ± 0.05 1 (Reference) 0.15 ± 0.05 1 (Reference) 0.48 ± 0.07 1 (Reference) A/G b 79 0.33 ± 0.05 1.06 (0.67, 1.69) 0.21 ± 0.07 1.54 (0.68, 3.50) 0.40 ± 0.07 0.81 (0.46, 1.42) G/G b 10 P-valuea 0.80 0.30 0.45

ITGA4 rs7562325 C/C 71 0.40 ± 0.06 1 (Reference) 0.14 ± 0.06 1 (Reference) 0.60 ± 0.09 1 (Reference) C/T 99 0.33 ± 0.05 0.91 (0.56, 1.50) 0.19 ± 0.06 1.21 (0.48, 3.04) 0.45 ± 0.08 0.83 (0.46, 1.49) T/T 28 0.12 ± 0.06 0.49 (0.21, 1.12) 0.19 ± 0.12 1.38 (0.40, 4.78) 0.06 ± 0.06 0.23 (0.07, 0.78) P-valuea 0.22 0.86 0.039 (0.31) C/C or C/T 170 0.36 ± 0.04 1 (reference) 0.17 ± 0.04 1 (reference) 0.52 ± 0.06 1 (reference) T/T 28 0.12 ± 0.06 0.52 (0.24, 1.13) 0.19 ± 0.12 1.23 (0.41, 3.65) 0.06 ± 0.06 0.26 (0.08, 0.84) P-valuea 0.089 0.70 0.014

ITGAV rs9333288 A/A 106 0.31 ± 0.05 1 (Reference) 0.12 ± 0.05 1 (Reference) 0.45 ± 0.07 1 (Reference) A/G b 75 0.35 ± 0.05 1.05 (0.66, 1.68) 0.24 ± 0.07 1.08 (0.48, 2.42) 0.44 ± 0.08 1.06 (0.60, 1.89) G/G b 16 P-valuea 0.82 0.85 0.83

ITGAV rs9333289 T/T 117 0.32 ± 0.05 1 (Reference) 0.17 ± 0.05 1 (Reference) 0.44 ± 0.07 1 (Reference) T/C b 69 0.34 ± 0.06 0.97 (0.60, 1.57) 0.18 ± 0.07 0.84 (0.36, 1.99) 0.45 ± 0.08 1.00 (0.56, 1.77) C/C b 12 P-valuea 0.91 0.69 0.99

IGTA6 rs1920979 A/A 76 0.26 ± 0.05 1 (Reference) 0.16 ± 0.07 1 (Reference) 0.33 ± 0.08 1 (Reference) A/G 92 0.38 ± 0.06 1.51 (0.90, 2.53) 0.26 ± 0.08 1.46 (0.61, 3.51) 0.46 ± 0.08 1.52 (0.80, 2.89)

Table. 2. (Continued )

Polymorphism N Stage II and III Stage II only Stage III only

3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) 3-Year recurrence HR (95% CI) probability ± SE probability ± SE probability ± SE

G/G 30 0.32 ± 0.09 1.03 (0.49, 2.17) 0.00 ± 0.00 0.38 (0.08, 1.82) 0.72 ± 0.14 2.29 (0.96, 5.50) P-valuea 0.23 0.15 0.13 ITGA6 rs11895564 G/G 96 0.28 ± 0.05 1 (Reference) 0.17 ± 0.06 1 (Reference) 0.36 ± 0.07 1 (Reference) G/A b 82 0.36 ± 0.05 1.23 (0.77, 1.97) 0.15 ± 0.06 0.95 (0.41, 2.20) 0.52 ± 0.07 1.46 (0.82, 2.60) A/A b 19 P-valuea 0.38 0.91 0.19 Abbreviations: CI. Conﬁdence interval; HR, hazard ratio; TTR, time to recurrence. aP-value was based on log-rank test. Adjusted P-value for multiple testing 17 b using (PACT) from methods of Conneely and Boehnke was shown in the parentheses. PACT > 0.5 was omitted. Dominant model. Signiﬁcant results are in bold.

Table 3. Polymorphisms of integrin genes and multivariable analysis for TTR in patients with stage II or III colon cancer

Polymorphism Stage II and II Stage II only Stage III only

HR (95% CI) P-valuea HR (95% CI) P-valuea HR (95% CI) P- valuea

ITGB1 rs2298141 A/A 1 (Reference) 1 (Reference) 1 (Reference) A/G or G/Gb 1.286 (0.793, 2.085) 0.31 0.687 (0.282, 1.673) 0.41 1.909 (1.054, 3.459) 0.033 P for interaction 0.054

ITGB3 rs4642 A/A 1 (Reference) 1 (Reference) 1 (Reference) A/G 2.300 (1.358, 3.894) 0.002 5.939 (2.110, 16.718) o0.001 1.594 (0.842, 3.016) 0.15 G/G 1.141 (0.482 (2.700) 0.76 1.388 (0.266, 7.247) 0.70 1.093 (0.393, 3.040 0.86 A/G or G/Gb 1.975 (1.194, 3.269) 0.008 4.027 (1.556, 10.421) 0.004 1.476 (0.801, 2.721) 0.21 P for interaction 0.12

ITGB5 rs849019 T/T 1 (Reference) 1 (Reference) 1 (Reference) T/A 1.743 (1.027, 2.958) 0.039 1.767 (0.663, 4.711) 0.25 1.699 (0.881, 3.278) 0.11 A/A 0.991 (0.470, 2.088) 0.98 1.097 (0.282, 4.262) 0.89 0.927 (0.369, 2.331) 0.87 T/A, A/Ab 1.482 (0.898, 2.446) 0.12 1.524 (0.610, 3.805) 0.37 1.434 (0.768, 2.679) 0.26

ITGA1 rs1531545 T/T 1 (Reference) 1 (Reference) 1 (Reference) T/C 0.588 (0.341, 1.012) 0.055 0.614 (0.238, 1.586) 0.31 0.524 (0.266, 1.031) 0.061 C/C 1.112 (0.594, 2.082) 0.74 1.056 (0.298, 3.748) 0.93 1.018 (0.478, 2.165) 0.96 T/C or C/Cb 0.719 (0.442, 1.170) 0.18 0.682 (0.274, 1.695) 0.41 0.658 (0.365, 1.188) 0.17 C/Cc 1.478 (0.835, 2.617) 0.18 1.481 (0.498, 4.404) 0.48 1.406 (0.701, 2.820) 0.34

ITGA4 rs7562325 C/C 1 (Reference) 1 (Reference) 1 (Reference) C/T 0.885 (0.535, 1.463) 0.63 1.726 (0.551, 5.412) 0.35 0.796 (0.437, 1.448) 0.45 T/T 0.453 (0.191, 1.074) 0.072 2.364 (0.578, 9.667) 0.23 0.227 (0.064, 0.804) 0.022 C/T or T/Tb 0.778 (0.478, 1.266) 0.31 1.856 (0.616, 5.591) 0.27 0.649 (0.360, 1.173) 0.15 T/Tc 0.487 (0.216, 1.098) 0.083 1.637 (0.527, 5.089) 0.39 0.260 (0.077, 0.879) 0.030 P for interaction 0.048 aP-value was based on Wald test within Cox regression model adjusted stage, type of adjuvant therapy when appropriate and stratified by race. bDominant model. cRecessive model. Significant results are in bold. tumor-induced lymphangiogenesis in various tumor models In contrast to the α4 subunit, β1 is able to bind most of the α- thereby facilitating metastasis to lymph nodes.18 It is interesting subunits. Therefore, we have to consider that its effect may be to notice that in contrast to the effect of the α4β1 expression in independent of α4. A key mechanism that has been recently the microenvironment, its expression on the tumor cells is recognized to lead to chemotherapy resistance is cancer associated with a loss of tumorigenicity,19,20 which expression dormancy. Cancer cells are quiescent, not dividing and thus may differ by stage.21 Further supporting our results, the resistant to agents that target actively dividing cells. Preclinical prognostic value of ITGA4 hypermethylation in patients with data support that the ITGB1 activation result to metastatic growth squamous cell esophageal cancer has been shown to be stage from cellular dormancy,12 suggesting that ITGB1 differential specific.22 expression or activity may be responsible for tumor recurrence

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 226 – 234 Integrin variants and stage II and III colon cancer P Bohanes et al 232 a c ITGB3 rs4642 ITGB1 rs2298141 1.0 in Stage II and III colon cancer 1.0 in Stage III Colon Cancer

0.9 0.9 Adjusted P value = 0.008 Adjusted P value = 0.033 0.8 0.8

0.7 0.7 A/A/A (n=93) 0.6 0.6 A/A (n=70) 0.5 0.5

0.4 0.4 A/ A/G or G/G (n=105) A/G or G/G (n=37) 0.3 0.3

0.2 0.2 Estimated Recurrence-Free Probability Estimated Recurrence-Free Probability 0.1 0.1

0.0 0.0 024681012141618 02468101214 Years Since Diagnosis of Stage II or III Colon Cancer Years SinceDiagnosis ofStageIIIColonCancer

b ITGB3 rs4642 d ITGA4 rs7562325 1.0 1.0 in Stage II Colon Cancer in Stage III Colon Cancer 0.9 0.9 Adjusted P value = 0.004 Adjusted P value = 0.030 0.8 0.8 T/T (n=17) 0.7 0.7 A/A (n=45) 0.6 0.6

0.5 0.5 A/G or G/G (n=47) 0.4 0.4 C/C or C/T (n=88) 0.3 0.3

0.2 0.2 Estimated Recurrence-Free Probability Estimated Recurrence-Free Probability 0.1 0.1

0.0 0.0 0 2 4 6 8 10 12 14 16 18 02468101214 Years Since Diagnosis of Stage II Colon Cancer Years Since Diagnosis of Stage III Colon Cancer Figure 1. Time to recurrence by (a) ITGB3 rs4642 in the whole cohort, (b) ITGB3 only in stage II colon cancer, (c) ITGB1 rs2298141 in stage III colon cancer and (d) ITGA4 rs7562325 in stage III colon cancer.

during tumor angiogenesis and promote endothelial cell survival.23 Interestingly, ITGB5 has also been suggested to be a tumor suppressor gene in various cancer models including colon cancer,24,25 suggesting that α5β1 role depends on the tumor and/or microenvironment, which may explain the impact of ITGB1 rs2298141 limited to stage III disease. Our comprehensive RP analysis, strengthen the predominant role of ITGB3 in stage II disease. Although there are no data currently supporting a differential β3 protein or mRNA expression between stage II and III colon cancer, the functional role of β3 may be different between stages. β3 protein combines with either the αVorαIIb subunit to form two distinct heterodimeric ITGs. Although αIIbβ3 ITG (also known has GPIIb/IIIa) expression is limited to platelets, αVβ3 ITG is widely distributed, being mainly expressed on the activated endothelium. Both αIIbβ3 and αVβ3 ITG are also expressed on various tumors with αVβ3 expression being correlated with an aggressive phenotype and metastatic dissemination.8,26–29 In colon cancer models, αVβ3 ITG has an impact on tumor growth by increasing tumor cell invasion, Figure 2. Time to recurrence by ITGB1 rs2298141 and ITGA4 adhesion and proliferation.30,31 However, data generated in rs7562325 genotypes in stage III colon cancer. melanoma models suggest that the function of αVβ3 is modulated by the co-expression of αIIbβ3.32 Therefore, a stage-speciﬁc in patients treated with chemotherapy. Currently, the α5β1 relative expression of both heterodimers may explain our heterodimer, which is present on mesenchymal cells and generally prognostic differences seen between stage II and III. Another absent on cells from epithelial origin, has been involved in cancer way to modulate β3 heterodimers’ activity is by its interaction with dormancy.12 α5β1 is also expressed on activated endothelium various proteins such as metalloprotease domain 23,33 that have

The Pharmacogenomics Journal (2015), 226 – 234 © 2015 Macmillan Publishers Limited Integrin variants and stage II and III colon cancer P Bohanes et al 233 simulation using the Functional-Single-Nucleotide-Polymorphism Stage database14,15 predicted changes in the splicing regulation for all (n=234) three synonymous polymorphisms. Of note, the predicted functional effect was particularly stringent for the ITGB3 rs4642 SNP. Despite our hypothesis-driven approach evaluating variants in II III genes known to have a biologic role in colon cancer, we sought to adjust for multiple comparisons to decrease the number of false o ITGB3 Lymph node positives. Considering a pACT value of 0.15 as potentially rs4642 resection meaningful, only ITGB3 rs4642 in stage II colon cancer patients (n=105) (n=129) remained significant. As the combined analysis of ITGB1 rs2298141 and ITGA4 rs7562325 does not follow the genetic A/A A/G or G/G > 12 ≤ 12 mode of inheritance, we could not use the pACT method to adjust P-value. Therefore, in the context of a non-agnostic approach and 9/48 21/57 because multiple testing adjustment techniques increase false HR: 0.25 HR: 1 ITGA3 20/31 negatives, both ITGB3 rs4642 and the combined analysis of ITGB1 (0.10-0.64) (reference) rs1062484 rs2298141 and ITGA4 rs7562325 deserve validation in a large (n=163) cohort of stage II and III colon cancer patients, respectively. Our data should therefore be seen as hypothesis generating and C/C C/T, T/T should be further validated. The authors acknowledge that a limitation of this study is a lack of a validation cohort to support 27/75 13/23 our findings. HR: 1 HR: 2.12 In summary, this study provides the first evidence that germline (reference) (1.01-4.47) polymorphisms in ITG genes predict stage-specific tumor recurrence in patients with colon cancer. Our data strengthen the role Figure 3. Recursive partitioning analysis that identifies five distinct of tumor dormancy in early colon cancer and may help to select patient groups based on time to recurrence. Blue ovals represent fi intermediate subgroups; blue squares represent terminal nodes. subgroups of patients who may bene t from more aggressive Brown rectangles indicate predictive polymorphism or patient treatment strategies or from ITG-targeted treatments. These baseline characteristics. Fractions within terminal nodes indicate polymorphisms may have important drug development implica- the number of patients who recurred/total patients. tions as agents targeting the αVβ3 ITGs are currently in clinical trials for patients with advanced colorectal cancer. Furthermore, agents targeting the α4 subunit have been developed in other fl been reported numerically more frequently in stage II than in diseases, including in ammatory bowel disease, and may be stage III (albeit not significant).34 useful in selected patients with colon cancer. The comprehensive RP analysis also demonstrated that ITGA3 rs1062484 was the most important prognostic factor influencing CONFLICT OF INTEREST TTR in patients with stage III disease treated with an adequate fl surgery (⩾12 lymph nodes removed). These data suggest that in The authors declare no con ict of interest. stage III colon cancers, α3β1 is a prognostic marker for micrometastatic disease. In case of inadequate surgery, its ACKNOWLEDGMENTS prognostic value is probably offset by overt macrometastatic α β This work was funded by the Lanni Family Charitable Foundation and by the P30 CA disease. The functional role of 3 1 in cancer is not yet fully P30 CA014089-27S1 grant. AG is supported in part by a research grant from the understood, with some studies suggesting that its downregulation Austrian Society of Hematology and Oncology, the Bank Austria Visiting Scientists 35,36 is associated with an increased metastatic potential, while Program and the ‘Verein fuer Krebskranke’ of the Medical University Graz. Statement 37,38 others suggesting the reverse association. It has therefore of translational relevance: In the era of personalized treatments, the identification of been suggested that an interaction between different ITG prognostic biomarkers in early stages of colon cancer is of critical importance for risk- heterodimers, which may be tissue specific, could explain these adapted treatments/follow-up and for drug development. Recent data have shown results.39 In colon cancer patients, a previous study showed that that meaningful biomarkers may not be similar in stage II or in stage III colon cancer. low α3β1 protein and mRNA expression was associated with In this study, we identified ITG variants that give stage-specific prognostic worse outcome. In addition, contrasting with our data the α3β1 information, identifying patients at higher risk of tumor recurrence. Our data shows prognostic influence was mainly seen in patients with node- that ITGB3 is relevant is stage II colon cancer and suggest that drugs under negative disease.40 Major differences between our studies such as development targeting this pathway should be tested in this setting. In contrast, our data support a leading role for the α4β1 heterodimer in stage III colon cancer, which the inclusion of stage I and low-risk stage II tumors, Asian is of major importance for future drug development. Furthermore, the variants ethnicity, the lack of adjuvant chemotherapy and the lack of indi- identified in this study could predict benefit from agents targeting these pathways. cation on the quality of surgery may explain these conflicting data. The detailed molecular mechanisms in how these polymorphisms exert effects on colon cancer are unclear. ITGB1 rs2298141, DISCLAIMER ITGB3 rs4642 and ITGA4 rs7562325 SNPs are synonymous poly- ′ The manuscript is solely the work of the authors stated. Neither the submitted morphisms, whereas ITGA3 rs1062484 is located within the 3 - paper nor any similar paper, in whole or in part has been submitted, published untranslated region. Synonymous polymorphisms are called silent or is in press in any other scientific journal. All authors have read and approved polymorphisms, because they do not lead to amino acid changes. all versions of the manuscript, its content and its submission. However, synonymous substitutions can lead to functional changes by affecting mRNA structure and stability, translational kinetics and small RNA binding by alternative splicing.41,42 REFERENCES ′ Similarly, polymorphisms within the 3 -untranslated region may 1 Kuebler JP, Wieand HS, O'Connell MJ, Smith RE, Colangelo LH, Yothers G et al. affect gene expression at the transcriptional level mainly through Oxaliplatin combined with weekly bolus fluorouracil and leucovorin as surgical 43 the control of mRNA stability and translational efficiency. adjuvant chemotherapy for stage II and III colon cancer: results from NSABP C-07. Supporting the translational effects seen in our study, in silico J Clin Oncol. 2007; 25: 2198–2204.

© 2015 Macmillan Publishers Limited The Pharmacogenomics Journal (2015), 226 – 234 Integrin variants and stage II and III colon cancer P Bohanes et al 234 2 André T, Boni C, Navarro M, Tabernero J, Hickish T, Topham C et al. Improved 24 Stallmach A, von Lampe B, Orzechowski HD, Matthes H, Riecken EO. Increased overall survival with oxaliplatin, fluorouracil, and leucovorin as adjuvant treatment fibronectin-receptor expression in colon carcinoma-derived HT 29 cells decreases in stage II or III colon cancer in the MOSAIC trial. J Clin Oncol. 2009; 27: 3109–3116. tumorigenicity in nude mice. Gastroenterology 1994; 106:19–27. 3 International Multicentre Pooled Analysis of Colon Cancer Trials (IMPACT) inves- 25 Varner JA, Emerson DA, Juliano RL. Integrin alpha 5 beta 1 expression negatively tigators. Efficacy of adjuvant fluorouracil and folinic acid in colon cancer. Inter- regulates cell growth: reversal by attachment to fibronectin. Mol Biol Cell 1995; 6: national Multicentre Pooled Analysis of Colon Cancer Trials (IMPACT) 725–740. investigators. Lancet 1995; 345: 939–944. 26 Chen YQ, Trikha M, Gao X, Bazaz R, Porter AT, Timar J et al. Ectopic expression of 4 Roth AD, Tejpar S, Delorenzi M, Yan P, Fiocca R, Klingbiel D et al. Stage-specific platelet integrin alphaIIb beta3 in tumor cells from various species and prognostic value of molecular markers in colon cancer: Results of the translational histological origin. Int J Cancer 1997; 72:642–648. study on the PETACC 3-EORTC 40993-SAKK 60-00 trial (abstract). J Clin Oncol 27 Hosotani R, Kawaguchi M, Masui T, Koshiba T, Ida J, Fujimoto K et al. Expression of (Meeting Abstract) 2009; 27: 4002. integrin alphaVbeta3 in pancreatic carcinoma: relation to MMP-2 activation and 5 Sargent DJ, Marsoni S, Monges G, Thibodeau SN, Labianca R, Hamilton SR et al. lymph node metastasis. Pancreas 2002; 25:e30–e35. Defective mismatch repair as a predictive marker for lack of efficacy of 28 Trikha M, Zhou Z, Timar J, Raso E, Kennel M, Emmell E et al. Multiple roles for fluorouracil-based adjuvant therapy in colon cancer. J Clin Oncol 2010; 28: platelet GPIIb/IIIa and alphavbeta3 integrins in tumor growth, angiogenesis, and 3219–3226. metastasis. Cancer Res 2002; 62: 2824–2833. 6 Avraamides CJ, Garmy-Susini B, Varner JA. Integrins in angiogenesis and lym- 29 Vellon L, Menendez JA, Lupu R. AlphaVbeta3 integrin regulates heregulin (HRG)- 8 – phangiogenesis. Nat Rev Cancer 2008; :604 617. induced cell proliferation and survival in breast cancer. Oncogene 2005; 24: 7 Páez D, Labonte MJ, Bohanes P, Zhang W, Benhanim L, Ning Y et al. Cancer 3759–3773. dormancy: a model of early dissemination and late cancer recurrence. Clin Cancer 30 Karadag A, Fedarko NS, Fisher LW. Dentin matrix protein 1 enhances invasion 18 – Res 2012; : 645 653. potential of colon cancer cells by bridging matrix metalloproteinase-9 to integrins 8 8 Takada Y, Ye X, Simon S. The integrins. Genome Biol 2007; : 215. and CD44. Cancer Res 2005; 65: 11545–11552. 9 Schwartz MA, Ginsberg MH. Networks and crosstalk: integrin signalling spreads. 31 Conti JA, Kendall TJ, Bateman A, Armstrong TA, Papa-Adams A, Xu Q et al. The 4 – Nat Cell Biol 2002; :E65 E68. desmoplastic reaction surrounding hepatic colorectal adenocarcinoma metas- fi 10 Alam N, Goel HL, Zarif MJ, Butter eld JE, Perkins HM, Sansoucy BG et al. The tases aids tumor growth and survival via alphav integrin ligation. Clin Cancer Res 213 – integrin-growth factor receptor duet. J Cell Physiol 2007; :649 653. 2008; 14: 6405–6413. 11 Liu Z, Wang F, Chen X. Integrin alpha(v)beta(3)-Targeted Cancer Therapy. Drug 32 Trikha M, Timar J, Zacharek A, Nemeth JA, Cai Y, Dome B et al. Role for beta3 Dev Res 2008; 69: 329–339. integrins in human melanoma growth and survival. Int J Cancer 2002; 101: 12 Barkan D, Chambers AF. β1-integrin: a potential therapeutic target in the battle 156–167. against cancer recurrence. Clin Cancer Res 2011; 17: 7219–7223. 33 Verbisck NV, Costa ET, Costa FF, Cavalher FP, Costa MD, Muras A et al. ADAM23 13 McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM et al. negatively modulates alpha(v)beta(3) integrin activation during metastasis. Can- Reporting recommendations for tumor marker prognostic studies. J Clin Oncol cer Res 2009; 69: 5546–5552. 2005; 23: 9067–9072. 34 Choi JS, Kim KH, Jeon YK, Kim SH, Jang SG, Ku JL et al. Promoter hypermethylation 14 Lee PH, Shatkay H. F-SNP: computationally predicted functional SNPs for disease of the ADAM23 gene in colorectal cancer cell lines and cancer tissues. Int J Cancer association studies. Nucleic Acids Res 2008; 36: D820–D824. 2009; 124:1258–1262. 15 Lee PH, Shatkay H. An integrative scoring system for ranking SNPs by their 35 Dedhar S, Saulnier R, Nagle R, Overall CM. Specific alterations in the expression of potential deleterious effects. Bioinformatics 2009; 25: 1048–1055. alpha 3 beta 1 and alpha 6 beta 4 integrins in highly invasive and metastatic 16 Winder T, Bohanes P, Zhang W, Yang D, Power DG, Ning Y et al. GRP78 promoter variants of human prostate carcinoma cells selected by in vitro invasion through polymorphism rs391957 as potential predictor for clinical outcome in gastric and reconstituted basement membrane. Clin Exp Metastasis 1993; 11:391–400. colorectal cancer patients. Ann Oncol 2011; 22:2431–2439. 36 Sampson-Johannes A, Wang W, Shtivelman E. Colonization of human lung grafts 17 Conneely KN, Boehnke M. So many correlated tests, so little time! Rapid adjust- in SCID-hu mice by human colon carcinoma cells. Int J Cancer 1996; 65: 864–869. ment of p values for multiple correlated tests. Am J Hum Genet 2007; 81: 37 Takatsuki H, Komatsu S, Sano R, Takada Y, Tsuji T. Adhesion of gastric carcinoma 1158–1168. 18 Garmy-Susini B, Avraamides CJ, Schmid MC, Foubert P, Ellies LG, Barnes L et al. cells to peritoneum mediated by alpha3beta1 integrin (VLA-3). Cancer Res 2004; 64 – Integrin alpha4beta1 signaling is required for lymphangiogenesis and tumor : 6065 6070. ’ metastasis. Cancer Res 2010; 70:3042–3051. 38 Pontes-Júnior J, Reis ST, de Oliveira LC, Sant anna AC, Dall'oglio MF, Antunes AA 19 Qian F, Vaux DL, Weissman IL. Expression of the integrin alpha 4 beta 1 on et al. Association between integrin expression and prognosis in localized 70 – melanoma cells can inhibit the invasive stage of metastasis formation. Cell 1994; prostate cancer. Prostate 2010; : 1189 1195. 12 – 77: 335–347. 39 Kreidberg JA. Functions of alpha3beta1 integrin. Curr Opin Cell Biol 2000; : 548 553. 20 Park J, Song SH, Kim TY, Choi MC, Jong HS, Kim TY et al. Aberrant methylation of 40 Hashida H, Takabayashi A, Tokuhara T, Taki T, Kondo K, Kohno N et al. Integrin integrin alpha4 gene in human gastric cancer cells. Oncogene 2004; 23: alpha3 expression as a prognostic factor in colon cancer: association with MRP-1/ 97 – 3474–3480. CD9 and KAI1/CD82. Int J Cancer 2002; :518 525. 21 Karmakar S, Mukherjee R. Integrin receptors and ECM proteins involved in pre- 41 Yu JJ, Lee KB, Mu C, Li Q, Abernathy TV, Bostick-Bruton F et al. Comparison of two ferential adhesion of colon carcinoma cells to lung cells. Cancer Lett 2003; 196: human ovarian carcinoma cell lines (A2780/CP70 and MCAS) that are equally 217–227. resistant to platinum, but differ at codon 118 of the ERCC1 gene. Int J Oncol 2000; 22 Lee EJ, Lee BB, Han J, Cho EY, Shim YM, Park J et al. CpG island hypermethylation 16:555–560. of E-cadherin (CDH1) and integrin alpha4 is associated with recurrence of 42 Sauna ZE, Kimchi-Sarfaty C, Ambudkar SV, Gottesman MM. Silent polymorphisms early stage esophageal squamous cell carcinoma. Int J Cancer 2008; 123: speak: how they affect pharmacogenomics and the treatment of cancer. Cancer 2073–2079. Res 2007; 67: 9609–9612. 23 Stoeltzing O, Liu W, Reinmuth N, Fan F, Parry GC, Parikh AA et al. Inhibition of 43 Mandola MV, Stoehlmacher J, Zhang W, Groshen S, Yu MC, Iqbal S et al. A6bp integrin alpha5beta1 function with a small peptide (ATN-161) plus continuous polymorphism in the thymidylate synthase gene causes message instability and is 5-FU infusion reduces colorectal liver metastases and improves survival in mice. associated with decreased intratumoral TS mRNA levels. Pharmacogenetics 2004; Int J Cancer 2003; 104: 496–503. 14:319–327.

Supplementary Information accompanies the paper on the The Pharmacogenomics Journal website (http://www.nature.com/tpj)