RNA-binding proteins (RBPs) regulate many posttranscriptional cellular activities. Accumulating evidence suggests associations between RBPs with colonic tumorigenesis and chemosensitivity. We investigated the prognostic and predictive values of SNPs of genes encoding RBPs in metastatic colorectal cancer (mCRC), using clinical and genomic data from three randomized clinical trials of standard first-line chemotherapy for mCRC (TRIBE, FIRE-3, and MAVERICC). Genomic DNA extracted from blood samples was genotyped using an OncoArray. We tested 30 candidate SNPs of 10 major RBP-related genes with additive models. Prognostic values were estimated by meta-analysis approach. Treatment-by-SNP interactions were tested to estimate predictive values for targeted drugs and cytotoxic backbone chemotherapies. This study included 884 patients. The meta-analysis revealed prognostic values of LIN28B rs314277 [HR, 1.26; 95% confidence interval (CI), 1.06–1.49, P = 0.005, FDR-adjusted P = 0.072 for overall survival (OS)] and LIN28B rs314276 (HR, 1.25; 95% CI, 1.08–1.44, P = 0.002, FDR-adjusted P = 0.062 for OS). Although some SNPs showed potentially predictive values, these associations were not confirmed after FDR adjustment. In conclusion, the results of this study are warranting additional studies to provide the evidence that RBP-related SNPs may be associated with the prognosis of patients with mCRC treated with standard first-line chemotherapies. In addition, further studies are warranted to study the predictive value.

RNA-binding proteins (RBPs) are highly conserved molecules involved in the posttranscriptional regulation of many cellular activities that bind to and form ribonucleoprotein complexes with various classes of RNA, including mRNA, miRNA, transfer RNA, siRNA, telomerase RNA, small nucleolar RNA, and spliceosomal small nuclear RNA (1). The functions of RBPs include RNA splicing, modifications, transport, localization, stability, degradation, and translation (2). As reported in 2014, high-throughput screening identified 1,542 human RBPs, which is believed to represent up to 7.5% of all protein-coding genes, and nearly 50% of RBPs were related to mRNA metabolic pathways (3). This large number of RBPs underscores the importance of RNA metabolism in various cellular systems. Thus, the study of RBPs is a novel research field to elucidate the pathogenesis and identify therapeutic targets of various diseases, including cancer (4, 5).

A growing body of experimental evidence has linked the dysregulation of RBPs to the carcinogenesis of colorectal cancer (6). For instance, LIN28A and LIN28B enhance the invasive capabilities of colorectal cancer cells in association with the Wnt pathway, which is crucial for carcinogenesis of colonic epithelial cells (7). RNA-binding motif protein 3 (RBM3) accelerates the proliferation of colorectal cancer cells by stabilizing the expression of proangiogenic factors (e.g., COX-2, IL8, and VEGF) and markers of cancer stem cells (8, 9). The mammalian IGF2 mRNA-binding protein (IGF2BP) belongs to a distinct family of RBPs that stabilizes expression of several proliferative factors. For instance, IGF2BP1 regulates MYC, IGFBP2 regulates NRAS and RAF1, and IGFBP3 regulates IGF2, HMGA2, and CCND1/2 (10–13). Finally, Musashi-1 (MSI1) and Musashi-2 (MSI2) are well-studied oncogenic RBPs that promote intestinal transformation via the regulation of several key pathways, such as NUMB/Notch, PTEN/mTOR, TGFβ/SMAD3, MYC, and Cmet (14).

Further evidence suggests that RBPs are related to the sensitivity and resistance to antitumor agents clinically used for the treatment of patients with metastatic colorectal cancer (mCRC). For example, overexpression of human antigen R (HuR, also known as ELAVL1) induces resistance to oxaliplatin in colorectal cancer cells by upregulation of CDC6, which is essential for the initiation of DNA replication (15). Other studies of ovarian cancer cells showed that RBM3 is associated with sensitivity to platinum agents via the regulation of DNA damage response factors, such as Chk1/2 and MCM3. This is consistent with another study showing that downregulation of RBM3 leads to the activation of Chk1/2 in colorectal cancer cells (8).

Despite promising preclinical evidence indicating associations between RBPs and the pathogenesis and chemosensitivity of colorectal cancer, little is known about possible clinical implications of these molecules. Thus, the aim of the present study was to investigate the prognostic and predictive values of SNPs of major RBP-encoding genes using clinical data from three randomized trials of first-line chemotherapy agents for mCRC, including TRIBE (16), FIRE-3 (17), and MAVERICC (18).

Patient population and study design

The study cohort included patients with mCRC enrolled in three first-line randomized trials: TRIBE (NCT00719797; ref. 16), FIRE-3 (NCT00433927; ref. 17), and MAVERICC (NCT01374425; ref. 18). In the TRIBE trial, patients received either FOLFIRI plus bevacizumab or FOLFOXIRI plus bevacizumab; in the FIRE-3 trial, patients received either FOLFIRI plus bevacizumab or FOLFIRI plus cetuximab; in the MAVERICC trial, patients received either FOLFIRI plus bevacizumab or mFOLFOX6 plus bevacizumab. Patients without sufficient peripheral whole blood samples for analysis were excluded from the study. All patients provided informed written consent for molecular research prior to study enrollment. The study protocol was approved by the Institutional Review Boards of each participating institution and was conducted in accordance with the tenets of the Declaration of Helsinki, as well as the Good Clinical Practice and REMARK guidelines.

Genotyping and selecting polymorphisms

Genomic DNA was extracted from peripheral whole blood collected before treatment initiation using the QIAmp Kit (QIAGEN, Inc.) in accordance with the manufacturer's protocol (www.qiagen.com). Genomic DNA was quality controlled using Quant-iT PicoGreen dsDNA (Thermo Fisher Scientific; P7589) following the manufacturer's protocol and then processed using Infinium OncoArray-530K BeadChip for genotyping (Illumina, Inc.; ref. 19). Data generated were packaged, and data quality was assessed using Illumina Genome Studio version 2. The following steps were run to impute untyped variants from the 1000 Genomes Project Phase 3 database using the quality-controlled genotypes as the inference set: exclude SNPs with call rate below 98% and minor allele frequency < 0.01 in either Europeans or Asian samples; remove monomorphic SNPs; and remove SNPs with 1000 Genomes Project Phase 3 concordance <95% (further details about the imputation are shown in Supplementary Methods).

We focused on 10 major genes encoding RBPs linked to colorectal cancer pathogenesis: MSI1, MSI2, ELAVL1, RBM3, LIN28A, LIN28B, IGF2BP1, IGF2BP2, IGF2BP3, and ZEP36. The candidate SNPs for this study were carefully selected according to the following criteria filtering and prioritizing the SNPs from dbSNP variants (https://www.ncbi.nlm.nih.gov): (i) minor allele frequency in whites (defined as “European” in the 1000 Genomes Project Phase 3) of at least 10% in the Ensemble Genome Browser (https://www.ensembl.org), (ii) having potential biological functions, including nonsynonymous coding, stop codon, transcription factor–binding site, splicing regulation, and miRNA-binding site, based on public databases (https://snpinfo.niehs.nih.gov), (iii) tag SNPs chosen from HapMap genotype data with an r2 threshold = 0.8 (https://snpinfo.niehs.nih.gov), and (iv) being associated with human diseases and/or drug sensitivity based on published literatures. In total, 30 SNPs were selected. The characteristics of the selected SNPs are shown in Supplementary Table S1.

Statistical analysis

Selected SNPs were evaluated for relevance to tumor response (TR), progression-free survival (PFS), and overall survival (OS). TR was defined as the proportion of patients achieving a complete response or partial response according to the RECIST 1.1 criteria. PFS was defined as the time from randomization to disease progression or death from any cause. OS was defined as time from randomization to death from any cause. Patients who experienced no events were censored at the last follow-up date. Associations between the SNPs and clinical outcomes were tested using a logistic regression model for TR or a Cox proportional hazards model for PFS and OS. These models were multivariable using the following study-specific adjusting covariates: the TRIBE trial included sex, age, Eastern Cooperative Oncology Group (ECOG) performance status (PS), primary tumor site, liver-limited disease, adjuvant chemotherapy, BRAF status, and RAS status; the FIRE-3 trial included sex, age, ECOG PS, primary tumor site, liver-limited disease, BRAF status, and RAS status; and the MAVERICC trial included ethnicity, sex, age, ECOG PS, primary tumor site, primary tumor resected, number of metastases, and KRAS status. SNPs were coded using additive and dominant genetic models for the number of variant alleles. In the additive model, the common homozygote is represented by 0, the heterozygote by 1, and the variant homozygote by 2.

To deeply assess the prognostic value of each SNP across all treatment arms based on the additive genetic model, a statistically powerful meta-analysis was conducted using METASOFT software (http://genetics.cs.ucla.edu/meta/). A fixed-effects model based on inverse-variance–weighted effect size was used to evaluate the overall effect of each SNP in all treatment arms. Heterogeneity of effects across all arms was tested using Cochran's Q statistic. After the meta-analysis, we performed an additional analysis to further test each SNP's contribution to the clinical outcomes using the dominant genetic model. In the additional analysis, correlation between each SNP and TR was examined using the χ2 test. To test the association between each SNP and PFS or OS, univariable and multivariable analyses using the Cox proportional hazards model and the log-rank test were performed.

To assess the predictive value of each SNP, treatment-by-SNP interaction was tested for all clinical outcomes within the FIRE-3 and MAVERICC cohorts based on the OR and HR calculated in the multivariable analyses using the additive genetic model. In the FIRE-3 cohort, the predictive value of the targeting agents (i.e., cetuximab or bevacizumab) was assessed. In the MAVERICC cohort, the cytotoxicity of backbone chemotherapy (i.e., FOLFOX vs. FOLFIRI) was assessed.

A two-sided probability (P) value of <0.05 was considered statistically significant. In multiple testing in the meta-analysis, this statistical significance was determined after adjusting for the Benjamini and Hochberg FDR. All analyses were performed using SAS version 9.4 software (SAS Institute).

We assessed the associations between selected SNPs and the gene expression status in colon tissue using Genotype-Tissue Expression (GTEx) Portal V8 (https://www.gtexportal.org/home/).

Patient characteristics

A total of 884 patients were included in this study (Fig. 1). Some characteristics were unbalanced among the six treatment arms. Specifically, the proportions of patients aged >65 years and with the RAS wild-type were higher in the FIRE-3 arms, whereas that of ECOG PS 0 was higher in the TRIBE arms, and that of more than two metastatic organs, a right-sided primary tumor, and the absence of primary tumor resection were higher in the MAVERICC arms (Table 1).

Figure 1.

CONSORT diagram. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Figure 1.

CONSORT diagram. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Close modal
Table 1.

Comparison of patient characteristics between cohorts.

TRIBEMAVERICCFIRE-3
TotalFOLFIRIFOLFOXIRIFOLFIRIFOLFOX6FOLFIRIFOLFIRI
CharacteristicsN = 884BEVBEVBEVBEVBEVCETP value
Sex        0.06 
 Male 571 132 (61%) 66 (61%) 103 (63%) 101 (63%) 70 (65%) 99 (77%)  
 Female 313 83 (39%) 43 (39%) 60 (37%) 60 (37%) 37 (35%) 30 (23%)  
Age        0.005 
 ≤65 589 156 (73%) 78 (72%) 101 (62%) 117 (73%) 62 (58%) 75 (58%)  
 >65 295 59 (27%) 31 (28%) 62 (38%) 44 (27%) 45 (42%) 54 (42%)  
ECOG PS        <0.001 
 ECOG 0 586 177 (82%) 95 (87%) 97 (60%) 81 (50%) 56 (52%) 80 (62%)  
 ECOG 1 296 37 (17%) 14 (13%) 66 (40%) 79 (49%) 51 (48%) 49 (38%)  
 Unknowna 1 (1%) 0 (0%) 0 (0%) 1 (1%) 0 (0%) 0 (0%)  
Primary tumor site        <0.001 
 Right-sided 261 53 (25%) 30 (28%) 67 (41%) 64 (40%) 25 (23%) 22 (17%)  
 Left-sided 599 147 (68%) 73 (67%) 96 (59%) 97 (60%) 81 (76%) 105 (81%)  
 Unknowna 24 15 (7%) 6 (6%) 0 (0%) 0 (0%) 1 (1%) 2 (2%)  
Number of metastases        <0.001 
 ≤2 657 178 (83%) 89 (82%) 106 (65%) 101 (63%) 83 (78%) 100 (78%)  
 >2 219 37 (17%) 20 (18%) 57 (35%) 60 (37%) 20 (19%) 25 (19%)  
 Unknowna 0 (0%) 0 (0%) 0 (0%) 0 (0%) 4 (4%) 4 (3%)  
Liver limited disease        0.64 
 No 379 150 (70%) 70 (64%) NA NA 75 (70%) 84 (65%)  
 Yes 181 65 (30%) 39 (36%) NA NA 32 (30%) 45 (35%)  
Primary tumor resected        <0.001 
 No 447 80 (37%) 31 (28%) 153 (94%) 148 (92%) 12 (11%) 23 (18%)  
 Yes 437 135 (63%) 78 (72%) 10 (6%) 13 (8%) 95 (89%) 106 (82%)  
Adjuvant chemotherapy        0.08 
 No 760 188 (87%) 94 (86%) 143 (88%) 146 (91%) 86 (80%) 103 (80%)  
 Yes 123 27 (13%) 15 (14%) 20 (12%) 15 (9%) 21 (20%) 25 (19%)  
 Unknowna 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (1%)  
RAS status        <0.001 
 Wild 233 50 (23%) 34 (31%) NA NA 66 (62%) 83 (64%)  
 Mutant 203 110 (51%) 57 (52%) NA NA 17 (16%) 19 (15%)  
 Unknowna 124 55 (26%) 18 (17%) NA NA 24 (22%) 27 (21%)  
BRAF status        0.59 
 Wild 432 168 (78%) 88 (81%) NA NA 81 (76%) 95 (74%)  
 Mutant 30 10 (5%) 9 (8%) NA NA 4 (4%) 7 (5%)  
 Unknowna 98 37 (17%) 12 (11%) NA NA 22 (21%) 27 (21%)  
TRIBEMAVERICCFIRE-3
TotalFOLFIRIFOLFOXIRIFOLFIRIFOLFOX6FOLFIRIFOLFIRI
CharacteristicsN = 884BEVBEVBEVBEVBEVCETP value
Sex        0.06 
 Male 571 132 (61%) 66 (61%) 103 (63%) 101 (63%) 70 (65%) 99 (77%)  
 Female 313 83 (39%) 43 (39%) 60 (37%) 60 (37%) 37 (35%) 30 (23%)  
Age        0.005 
 ≤65 589 156 (73%) 78 (72%) 101 (62%) 117 (73%) 62 (58%) 75 (58%)  
 >65 295 59 (27%) 31 (28%) 62 (38%) 44 (27%) 45 (42%) 54 (42%)  
ECOG PS        <0.001 
 ECOG 0 586 177 (82%) 95 (87%) 97 (60%) 81 (50%) 56 (52%) 80 (62%)  
 ECOG 1 296 37 (17%) 14 (13%) 66 (40%) 79 (49%) 51 (48%) 49 (38%)  
 Unknowna 1 (1%) 0 (0%) 0 (0%) 1 (1%) 0 (0%) 0 (0%)  
Primary tumor site        <0.001 
 Right-sided 261 53 (25%) 30 (28%) 67 (41%) 64 (40%) 25 (23%) 22 (17%)  
 Left-sided 599 147 (68%) 73 (67%) 96 (59%) 97 (60%) 81 (76%) 105 (81%)  
 Unknowna 24 15 (7%) 6 (6%) 0 (0%) 0 (0%) 1 (1%) 2 (2%)  
Number of metastases        <0.001 
 ≤2 657 178 (83%) 89 (82%) 106 (65%) 101 (63%) 83 (78%) 100 (78%)  
 >2 219 37 (17%) 20 (18%) 57 (35%) 60 (37%) 20 (19%) 25 (19%)  
 Unknowna 0 (0%) 0 (0%) 0 (0%) 0 (0%) 4 (4%) 4 (3%)  
Liver limited disease        0.64 
 No 379 150 (70%) 70 (64%) NA NA 75 (70%) 84 (65%)  
 Yes 181 65 (30%) 39 (36%) NA NA 32 (30%) 45 (35%)  
Primary tumor resected        <0.001 
 No 447 80 (37%) 31 (28%) 153 (94%) 148 (92%) 12 (11%) 23 (18%)  
 Yes 437 135 (63%) 78 (72%) 10 (6%) 13 (8%) 95 (89%) 106 (82%)  
Adjuvant chemotherapy        0.08 
 No 760 188 (87%) 94 (86%) 143 (88%) 146 (91%) 86 (80%) 103 (80%)  
 Yes 123 27 (13%) 15 (14%) 20 (12%) 15 (9%) 21 (20%) 25 (19%)  
 Unknowna 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 1 (1%)  
RAS status        <0.001 
 Wild 233 50 (23%) 34 (31%) NA NA 66 (62%) 83 (64%)  
 Mutant 203 110 (51%) 57 (52%) NA NA 17 (16%) 19 (15%)  
 Unknowna 124 55 (26%) 18 (17%) NA NA 24 (22%) 27 (21%)  
BRAF status        0.59 
 Wild 432 168 (78%) 88 (81%) NA NA 81 (76%) 95 (74%)  
 Mutant 30 10 (5%) 9 (8%) NA NA 4 (4%) 7 (5%)  
 Unknowna 98 37 (17%) 12 (11%) NA NA 22 (21%) 27 (21%)  

Note: P values were estimated by the χ2 test.

Abbreviations: BEV, bevacizumab; CET, cetuximab; NA, not assessed.

aUnknown groups were not included in the analysis.

Prognostic values of each SNP

Many potential associations (raw P < 0.05) between SNPs and clinical outcomes were identified by analyses performed separately for each treatment arm (Supplementary Fig. S1). However, these associations did not appear to be clustered by SNPs. The meta-analysis results are shown in Table 2 and Supplementary Table S2. Cochran's Q statistic showed no evidence of a heterogeneous effect across the six treatment arms, except for IGF2BP2 rs11705701 for PFS (Q = 21.56, FDR-adjusted P = 0.019). In the meta-analysis, outstanding associations with clinical outcomes were observed in two SNPs: LIN28B rs314277 with OS [HR, 1.26; 95% confidence interval (CI), 1.06–1.49, raw P = 0.005, FDR-adjusted P = 0.072]; LIN28B rs314276 with OS (HR, 1.25; 95% CI, 1.08–1.44, raw P = 0.002, FDR-adjusted P = 0.062). Given that the HRs were >1.00 for all of these results, these SNPs seem to serve as prognostic factors of poorer OS. Forest plots were constructed to visualize the effects of these SNPs in each treatment arm and the summary (Figs. 23). Specifically, the prognostic value of LIN28B rs314277 was consistent among all clinical outcomes (TR, PFS, and OS). The adverse prognostic effect of the SNP for OS was consistent across all treatment arms, except for the FOLFIRI plus cetuximab arm of the FIRE-3 trial (Fig. 2). LIN28B rs314276 was prognostic of only OS, which was consistent across all treatment arms, except for the FOLFIRI plus cetuximab arm of the FIRE-3 trial (Fig. 3). Additional analyses using a dominant genetic model showed patients carrying any A allele of LIN28B rs314277 had a worse OS than those with C/C genotype in the FOLFIRI plus bevacizumab arm of the TRIBE trial (HR, 1.86; 95% CI, 1.31–2.65, P < 0.001, in multivariable analysis), and patients carrying any A allele of LIN28B rs314276 had a worse OS than those with C/C genotype in the FOLFIRI plus bevacizumab arm of the TRIBE trial (HR, 1.64; 95% CI, 1.19–2.27, P = 0.002, in multivariable analysis; Supplementary Table S3). Consistently, a trend for worse OS in patients with any A allele of both SNPs was observed in the FOLFOXIRI plus bevacizumab arm in the TRIBE trial, in the FOLFIRI plus bevacizumab arm in the MAVERICC trial, and in the FOLFIRI plus bevacizumab arm in the FIRE-3 trial (Supplementary Table S3). Separate analyses by RAS status showed a consistent adverse effect of A alleles of LIN28B rs314277 and rs314276 on OS both in patients with the RAS wild-type and with the RAS mutation in the TRIBE trial arms (Supplementary Tables S4 and S5). However, inconsistent results were observed in the FOLFOX plus bevacizumab arm in the MAVERICC trial: patients having any A allele of LIN28B rs314276 had better OS than those with C/C genotype in patients with the RAS wild-type, which was opposite trend to the patients with the RAS mutation (Supplementary Tables S4 and S5).

Table 2.

Meta-analysis results for TR, PFS, and OS.

TRPFSOS
GeneSNPP value for FEQP value for QP value for FEQP value for QP value for FEQP value for Q
MSI1 rs1076205 0.81 8.09 0.15 0.83 4.75 0.45 0.69 4.16 0.53 
 rs2522137 0.33 4.97 0.42 0.58 7.53 0.18 0.37 5.82 0.32 
 rs1179434 0.32 0.59 0.99 0.46 8.34 0.14 0.16 7.27 0.20 
 rs1179442 0.77 5.94 0.31 0.78 5.05 0.41 0.39 6.28 0.28 
MSI2 rs9892791 0.83 3.47 0.63 0.84 3.61 0.61 0.89 0.94 0.97 
 rs11657292 0.58 7.51 0.19 0.24 4.33 0.50 0.03 4.17 0.53 
 rs3826301 0.74 10.09 0.07 0.46 3.16 0.68 0.46 0.84 0.98 
 rs1822381 0.02 4.32 0.51 0.12 3.20 0.67 0.14 3.23 0.67 
ELAVL1 rs2042920 0.40 4.29 0.51 0.36 8.24 0.14 0.06 6.68 0.25 
 rs12983784 0.92 9.72 0.08 0.04 5.32 0.38 0.048 3.44 0.63 
 rs4804244 0.56 2.93 0.71 0.58 5.26 0.39 0.07 2.48 0.78 
RBM3 rs926152 0.63 1.47 0.92 0.58 3.67 0.60 0.06 7.58 0.18 
 rs2249585 0.87 1.21 0.94 0.74 2.34 0.80 0.05 6.93 0.23 
LIN28A rs12728900 0.31 2.44 0.79 0.03 5.91 0.32 0.03 2.50 0.78 
 rs6697410 0.02 10.02 0.08 0.93 5.72 0.33 0.64 6.98 0.22 
 rs6598964 0.67 6.41 0.27 0.12 2.83 0.73 0.25 2.28 0.81 
 rs3811463 0.33 8.02 0.16 0.34 3.54 0.62 0.76 2.85 0.72 
 rs3811464 0.89 2.46 0.78 0.17 8.15 0.15 0.76 12.52 0.03 
LIN28B rs314277 0.14 1.62 0.90 0.09 4.97 0.42 0.005 (0.072) 10.83 0.06 
 rs221634 0.52 6.51 0.26 0.20 8.21 0.15 0.62 5.10 0.40 
 rs221635 0.59 11.14 0.049 0.88 1.98 0.85 0.10 7.45 0.19 
 rs314276 0.86 5.50 0.36 0.54 2.36 0.80 0.002 (0.062) 3.10 0.68 
IGF2BP1 rs2969 0.86 6.37 0.27 0.78 2.96 0.71 0.24 3.45 0.63 
 rs6504593 0.60 6.62 0.25 0.40 4.22 0.52 0.14 3.85 0.57 
 rs11655950 0.24 5.70 0.34 0.17 2.37 0.80 0.04 4.177 0.52 
IGF2BP2 rs11705701 0.70 6.57 0.25 0.54 21.56 0.001 (0.019) 0.86 8.47 0.13 
 rs7651090 0.50 2.04 0.84 0.45 10.01 0.08 0.66 10.03 0.07 
IGF2BP3 rs274035 0.28 5.36 0.37 0.89 12.32 0.03 0.23 12.92 0.02 
 rs433395 0.31 3.43 0.64 0.86 5.25 0.39 0.77 3.92 0.56 
ZFP36 rs251864 0.44 2.24 0.82 0.11 8.07 0.15 0.56 7.73 0.17 
TRPFSOS
GeneSNPP value for FEQP value for QP value for FEQP value for QP value for FEQP value for Q
MSI1 rs1076205 0.81 8.09 0.15 0.83 4.75 0.45 0.69 4.16 0.53 
 rs2522137 0.33 4.97 0.42 0.58 7.53 0.18 0.37 5.82 0.32 
 rs1179434 0.32 0.59 0.99 0.46 8.34 0.14 0.16 7.27 0.20 
 rs1179442 0.77 5.94 0.31 0.78 5.05 0.41 0.39 6.28 0.28 
MSI2 rs9892791 0.83 3.47 0.63 0.84 3.61 0.61 0.89 0.94 0.97 
 rs11657292 0.58 7.51 0.19 0.24 4.33 0.50 0.03 4.17 0.53 
 rs3826301 0.74 10.09 0.07 0.46 3.16 0.68 0.46 0.84 0.98 
 rs1822381 0.02 4.32 0.51 0.12 3.20 0.67 0.14 3.23 0.67 
ELAVL1 rs2042920 0.40 4.29 0.51 0.36 8.24 0.14 0.06 6.68 0.25 
 rs12983784 0.92 9.72 0.08 0.04 5.32 0.38 0.048 3.44 0.63 
 rs4804244 0.56 2.93 0.71 0.58 5.26 0.39 0.07 2.48 0.78 
RBM3 rs926152 0.63 1.47 0.92 0.58 3.67 0.60 0.06 7.58 0.18 
 rs2249585 0.87 1.21 0.94 0.74 2.34 0.80 0.05 6.93 0.23 
LIN28A rs12728900 0.31 2.44 0.79 0.03 5.91 0.32 0.03 2.50 0.78 
 rs6697410 0.02 10.02 0.08 0.93 5.72 0.33 0.64 6.98 0.22 
 rs6598964 0.67 6.41 0.27 0.12 2.83 0.73 0.25 2.28 0.81 
 rs3811463 0.33 8.02 0.16 0.34 3.54 0.62 0.76 2.85 0.72 
 rs3811464 0.89 2.46 0.78 0.17 8.15 0.15 0.76 12.52 0.03 
LIN28B rs314277 0.14 1.62 0.90 0.09 4.97 0.42 0.005 (0.072) 10.83 0.06 
 rs221634 0.52 6.51 0.26 0.20 8.21 0.15 0.62 5.10 0.40 
 rs221635 0.59 11.14 0.049 0.88 1.98 0.85 0.10 7.45 0.19 
 rs314276 0.86 5.50 0.36 0.54 2.36 0.80 0.002 (0.062) 3.10 0.68 
IGF2BP1 rs2969 0.86 6.37 0.27 0.78 2.96 0.71 0.24 3.45 0.63 
 rs6504593 0.60 6.62 0.25 0.40 4.22 0.52 0.14 3.85 0.57 
 rs11655950 0.24 5.70 0.34 0.17 2.37 0.80 0.04 4.177 0.52 
IGF2BP2 rs11705701 0.70 6.57 0.25 0.54 21.56 0.001 (0.019) 0.86 8.47 0.13 
 rs7651090 0.50 2.04 0.84 0.45 10.01 0.08 0.66 10.03 0.07 
IGF2BP3 rs274035 0.28 5.36 0.37 0.89 12.32 0.03 0.23 12.92 0.02 
 rs433395 0.31 3.43 0.64 0.86 5.25 0.39 0.77 3.92 0.56 
ZFP36 rs251864 0.44 2.24 0.82 0.11 8.07 0.15 0.56 7.73 0.17 

Note: FE indicates fixed effects based on inverse-variance–weighted effect size. OR and HR are shown in Supplementary Table S2. Q denotes Cochran's Q statistic for assessing heterogeneity across all treatment cohorts. P values that achieved a nominal 0.05 significance level are shown in bold. P values that achieved 0.1 level after adjustment by FDR are shown in the parentheses.

Abbreviation: FE, fixed effects.

Figure 2.

Forest plots of meta-analysis for LIN28B rs314277: log OR or log HR is shown with SE. The summary row shows the inverse-variance–weighted effect size with SE, combining the six estimates for the individual arms into a single summary measure. A negative log OR or a positive HR implies a negative influence on TR, PFS, and OS, respectively. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Figure 2.

Forest plots of meta-analysis for LIN28B rs314277: log OR or log HR is shown with SE. The summary row shows the inverse-variance–weighted effect size with SE, combining the six estimates for the individual arms into a single summary measure. A negative log OR or a positive HR implies a negative influence on TR, PFS, and OS, respectively. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Close modal
Figure 3.

Forest plots of meta-analysis for LIN28B rs314276: log OR or log HR are shown with SE. The summary row shows the inverse-variance–weighted effect size with SE, combining the six estimates for the individual arms into a single summary measure. A negative log OR or a positive HR implies a negative influence on TR, PFS, and OS, respectively. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Figure 3.

Forest plots of meta-analysis for LIN28B rs314276: log OR or log HR are shown with SE. The summary row shows the inverse-variance–weighted effect size with SE, combining the six estimates for the individual arms into a single summary measure. A negative log OR or a positive HR implies a negative influence on TR, PFS, and OS, respectively. Abbreviations: BEV, bevacizumab; CET, cetuximab.

Close modal

Predictive values of each SNP

In the FIRE-3 cohort, three SNPs were potentially predictive of the treatment outcome (interaction raw P < 0.05): RBM3 rs926152 for OS (raw P = 0.03), IGF2BP2 rs11705701 for PFS (raw P = 0.03), and IGF2BP2 rs7651090 for PFS (raw P = 0.04). While in the MAVERICC cohort, only one SNP had potential predictive value: IGF2BP2 rs11705701 for PFS (raw P = 0.008). However, the significance of these interactions was not confirmed after FDR adjustment (Table 3).

Table 3.

Results of treatment-by-SNP interaction test.

FIRE-3MAVERICC
TRPFSOSTRPFSOS
RawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adj
GeneSNPP valueP valueP valueP valueP valueP valueP valueP valueP valueP valueP valueP value
MSI1 rs1076205 0.35 0.90 0.95 0.99 0.76 0.95 0.24 0.87 0.68 0.92 0.90 0.94 
 rs2522137 0.70 0.90 0.99 0.99 0.45 0.84 0.24 0.87 0.71 0.92 1.00 1.00 
 rs1179434 0.70 0.90 0.57 0.96 0.98 0.98 0.56 0.87 0.66 0.92 0.47 0.90 
 rs1179442 0.49 0.90 0.85 0.99 0.56 0.84 0.49 0.87 0.90 0.92 0.63 0.90 
MSI2 rs9892791 0.28 0.90 0.10 0.96 0.67 0.89 0.95 0.95 0.92 0.92 0.59 0.90 
 rs11657292 0.81 0.90 0.54 0.96 0.59 0.84 0.36 0.87 0.37 0.92 0.91 0.94 
 rs3826301 0.22 0.90 0.16 0.96 0.28 0.84 0.60 0.87 0.59 0.92 0.55 0.90 
 rs1822381 0.80 0.90 0.49 0.96 0.87 0.98 0.61 0.87 0.47 0.92 0.27 0.90 
ELAVL1 rs2042920 0.60 0.90 0.26 0.96 0.95 0.98 0.36 0.87 0.65 0.92 0.45 0.90 
 rs12983784 0.79 0.90 0.38 0.96 0.40 0.84 0.34 0.87 0.89 0.92 0.52 0.90 
 rs4804244 0.71 0.90 0.97 0.99 0.89 0.98 0.30 0.87 0.71 0.92 0.90 0.94 
RBM3 rs926152 0.97 0.97 0.35 0.96 0.03 0.60 0.51 0.87 0.81 0.92 0.28 0.90 
 rs2249585 0.49 0.90 0.58 0.96 0.06 0.60 0.81 0.91 0.58 0.92 0.63 0.90 
LIN28A rs12728900 0.78 0.90 0.21 0.96 0.49 0.84 0.58 0.87 0.78 0.92 0.30 0.90 
 rs6697410 0.84 0.90 0.46 0.96 0.36 0.84 0.40 0.87 0.86 0.92 0.72 0.93 
 rs6598964 0.54 0.90 0.42 0.96 0.52 0.84 0.85 0.91 0.58 0.92 0.90 0.94 
 rs3811463 0.16 0.80 0.96 0.99 0.29 0.84 0.58 0.87 0.54 0.92 0.74 0.93 
 rs3811464 0.69 0.90 0.83 0.99 0.69 0.89 0.77 0.91 0.52 0.92 0.20 0.90 
LIN28B rs314277 0.83 0.90 0.92 0.99 0.06 0.60 0.54 0.87 0.42 0.92 0.46 0.90 
 rs221634 0.62 0.90 0.67 0.99 0.96 0.98 0.27 0.87 0.053 0.80 0.46 0.90 
 rs221635 0.13 0.79 0.97 0.99 0.94 0.98 0.22 0.87 0.78 0.92 0.36 0.90 
 rs314276 0.56 0.90 0.53 0.96 0.37 0.84 0.82 0.91 0.21 0.92 0.71 0.93 
IGF2BP1 rs2969 0.74 0.90 0.56 0.96 0.33 0.84 0.24 0.87 0.47 0.92 0.34 0.90 
 rs6504593 0.06 0.79 0.69 0.99 0.24 0.84 0.53 0.87 0.87 0.92 0.51 0.90 
 rs11655950 0.11 0.79 0.85 0.99 0.55 0.84 0.85 0.91 0.70 0.92 0.85 0.94 
IGF2BP2 rs11705701 0.13 0.79 0.03 0.56 0.14 0.83 0.56 0.87 0.008 0.23 0.13 0.90 
 rs7651090 0.93 0.96 0.04 0.56 0.23 0.84 0.89 0.92 0.08 0.83 0.09 0.90 
IGF2BP3 rs274035 0.13 0.79 0.86 0.99 0.19 0.84 0.76 0.91 0.52 0.92 0.54 0.90 
 rs433395 0.83 0.90 0.29 0.96 0.49 0.84 0.40 0.87 0.53 0.92 0.17 0.90 
ZFP36 rs251864 0.75 0.90 0.25 0.96 0.08 0.62 0.76 0.91 0.83 0.92 0.18 0.90 
FIRE-3MAVERICC
TRPFSOSTRPFSOS
RawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adjRawFDR-adj
GeneSNPP valueP valueP valueP valueP valueP valueP valueP valueP valueP valueP valueP value
MSI1 rs1076205 0.35 0.90 0.95 0.99 0.76 0.95 0.24 0.87 0.68 0.92 0.90 0.94 
 rs2522137 0.70 0.90 0.99 0.99 0.45 0.84 0.24 0.87 0.71 0.92 1.00 1.00 
 rs1179434 0.70 0.90 0.57 0.96 0.98 0.98 0.56 0.87 0.66 0.92 0.47 0.90 
 rs1179442 0.49 0.90 0.85 0.99 0.56 0.84 0.49 0.87 0.90 0.92 0.63 0.90 
MSI2 rs9892791 0.28 0.90 0.10 0.96 0.67 0.89 0.95 0.95 0.92 0.92 0.59 0.90 
 rs11657292 0.81 0.90 0.54 0.96 0.59 0.84 0.36 0.87 0.37 0.92 0.91 0.94 
 rs3826301 0.22 0.90 0.16 0.96 0.28 0.84 0.60 0.87 0.59 0.92 0.55 0.90 
 rs1822381 0.80 0.90 0.49 0.96 0.87 0.98 0.61 0.87 0.47 0.92 0.27 0.90 
ELAVL1 rs2042920 0.60 0.90 0.26 0.96 0.95 0.98 0.36 0.87 0.65 0.92 0.45 0.90 
 rs12983784 0.79 0.90 0.38 0.96 0.40 0.84 0.34 0.87 0.89 0.92 0.52 0.90 
 rs4804244 0.71 0.90 0.97 0.99 0.89 0.98 0.30 0.87 0.71 0.92 0.90 0.94 
RBM3 rs926152 0.97 0.97 0.35 0.96 0.03 0.60 0.51 0.87 0.81 0.92 0.28 0.90 
 rs2249585 0.49 0.90 0.58 0.96 0.06 0.60 0.81 0.91 0.58 0.92 0.63 0.90 
LIN28A rs12728900 0.78 0.90 0.21 0.96 0.49 0.84 0.58 0.87 0.78 0.92 0.30 0.90 
 rs6697410 0.84 0.90 0.46 0.96 0.36 0.84 0.40 0.87 0.86 0.92 0.72 0.93 
 rs6598964 0.54 0.90 0.42 0.96 0.52 0.84 0.85 0.91 0.58 0.92 0.90 0.94 
 rs3811463 0.16 0.80 0.96 0.99 0.29 0.84 0.58 0.87 0.54 0.92 0.74 0.93 
 rs3811464 0.69 0.90 0.83 0.99 0.69 0.89 0.77 0.91 0.52 0.92 0.20 0.90 
LIN28B rs314277 0.83 0.90 0.92 0.99 0.06 0.60 0.54 0.87 0.42 0.92 0.46 0.90 
 rs221634 0.62 0.90 0.67 0.99 0.96 0.98 0.27 0.87 0.053 0.80 0.46 0.90 
 rs221635 0.13 0.79 0.97 0.99 0.94 0.98 0.22 0.87 0.78 0.92 0.36 0.90 
 rs314276 0.56 0.90 0.53 0.96 0.37 0.84 0.82 0.91 0.21 0.92 0.71 0.93 
IGF2BP1 rs2969 0.74 0.90 0.56 0.96 0.33 0.84 0.24 0.87 0.47 0.92 0.34 0.90 
 rs6504593 0.06 0.79 0.69 0.99 0.24 0.84 0.53 0.87 0.87 0.92 0.51 0.90 
 rs11655950 0.11 0.79 0.85 0.99 0.55 0.84 0.85 0.91 0.70 0.92 0.85 0.94 
IGF2BP2 rs11705701 0.13 0.79 0.03 0.56 0.14 0.83 0.56 0.87 0.008 0.23 0.13 0.90 
 rs7651090 0.93 0.96 0.04 0.56 0.23 0.84 0.89 0.92 0.08 0.83 0.09 0.90 
IGF2BP3 rs274035 0.13 0.79 0.86 0.99 0.19 0.84 0.76 0.91 0.52 0.92 0.54 0.90 
 rs433395 0.83 0.90 0.29 0.96 0.49 0.84 0.40 0.87 0.53 0.92 0.17 0.90 
ZFP36 rs251864 0.75 0.90 0.25 0.96 0.08 0.62 0.76 0.91 0.83 0.92 0.18 0.90 

Note: Interaction P values less than 0.05 are shown in bold.

Abbreviation: FDR-adj, false discovery rate-adjusted.

Associations between selected SNPs and gene expression status in colon tissue from GTEx analysis

The results of GTEx analysis were obtained on 18 of 30 tested SNPs. We found eight SNPs significantly affecting gene expression status in normal colon tissue (sigmoid and/or transverse colon): MSI1 rs1076205, MSI1 rs2522137, MSI1 rs1179434, MSI1 rs1179442, ELAVL1 rs2042920, IGF2BP2 rs11705701, IGF2BP2 rs7651090, and IGF2BP3 rs433395 (Supplementary Fig. S2).

This study is the first to investigate the associations between the SNPs of major RBP-encoding genes and the clinical outcomes of patients with mCRC treated with standard first-line chemotherapeutic agents. Using a powerful meta-analysis approach, we demonstrated reliable prognostic values of the SNPs across all patients. However, there was no distinct predictive value for targeting drugs (i.e., cetuximab or bevacizumab) or backbone chemotherapy regimens (i.e., FOLFOX or FOLFIRI).

In the meta-analysis of the current study, two RBP-related SNPs associated with outstanding prognostic values were identified, despite the prespecified statistical significance was not demonstrated: LIN28B rs314277 and LIN28B rs314276 which are not linked in linkage disequilibrium. Our results indicate that these SNPs were associated with an increasing risk for death, given the HR of >1.00 for OS. Consistently, an adverse prognostic effect of LIN28B rs314277 was observed on TR and PFS. On the other hand, that of LIN28B rs314276 was not observed on TR and PFS, which implies a presence of some unidentified factors such as different contribution of second-line treatment affecting the associations between biomarker and clinical outcomes. LIN28B is reportedly overexpressed in colorectal cancer (20, 21) and promotes the migration, invasion, and transformation of colorectal cancer cells in a let-7 miRNA-dependent manner (i.e., repressing let-7) and let-7–independent manner (e.g., regulation of LGR5 and PROM1, which are markers of intestinal stem cells; ref. 21). In addition, LIN28B cooperates with Wnt signaling to promote tumor formation (7). Consistent with these preclinical findings, LIN28B overexpression is correlated with an invasive tumor phenotype, poorer survival, and increased tumor recurrence in colorectal cancer (20, 22). As a novel finding of the current study, A allele of LIN28B rs314277 and A allele of LIN28B rs314276 are associated with poorer OS of patients with mCRC treated with standard first-line chemotherapies. This effect was independent of RAS status in the TRIBE trial. The detailed molecular mechanisms involved in how these SNPs affect colorectal cancer biology remain unclear. However, according to the public database to predict the functional effects of SNPs, we found that both polymorphisms have potential biological function as a transcription factor–binding site which may affect the level, location, or timing of gene expression (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html; Supplementary Table S1). Thus, our findings showing associations between these SNPs and clinical outcomes can be biologically explained. Given that our results showed A alleles of these SNPs had a worse impact on OS, we hypothesize that these alleles may increase transcription of LIN28B contributing to a more aggressive phenotype. These results are strongly supported by those of a previous study showing that A allele of LIN28B rs314277 is associated with worse survival and recurrence of patients with stage II colorectal cancer (23). Furthermore, several studies have identified the disease-specific risk of these SNPs, strongly suggesting phenotypical functions (23–26). However, other tested SNPs were not shown to have significant prognostic values, which implies these SNPs may not have much impact on biology and pathogenesis of colorectal cancer.

The choices of a targeting drug (i.e., anti-VEGF or anti-EGFR drugs) and backbone chemotherapy regimen (i.e., oxaliplatin-based or irinotecan-based) are important clinical considerations for first-line treatment of mCRC. Predictive biomarkers are crucial to guide these decisions. Thus, statistical analysis of treatment-by-biomarker interaction was conducted to determine whether a biomarker is predictive (27). Although no SNPs met the prespecified significance level of interaction (FDR-adjusted P value < 0.05), a trend was observed in some SNPs: RBM3 rs926152, IGF2BP2 rs11705701, and IGF2BP2 rs7651090 in the FIRE-3 cohort; IGF2BP2 rs11705701 in the MAVERICC cohort. RBM3 is linked to angiogenesis, supported by a preclinical finding that RBM3 overexpression increases the proliferation of colorectal cancer cells by stabilizing COX-2, IL8, and VEGF (8). Thus, the potential predictive effect of RBM3 rs926152 in the FIRE-3 cohort may attribute to the relationship between RBM3 and the efficacy of bevacizumab. In contrast, IGF2BP2 enhances the mRNA stability of NRAS and RAF1 that are essential components of the MAPK pathway (10, 13). IGF2BP2 also promotes the IGF2/IGF-1R pathway that may be associated with resistance to EGFR-targeted therapies (28, 29). These findings suggest IGF2BP2 may be linked to the efficacy of cetuximab, supporting our results, showing the predictive effect of SNPs of IGF2BP2 in the FIRE-3 cohort. Moreover, another result showing that of IGF2BP2 rs11705701 in the MAVERICC cohort is supported by previous studies that have shown IGF-1R is associated with sensitivity to cytotoxic agents (30–32). However, it was unclear whether the SNP is specifically related to the efficacy of oxaliplatin or irinotecan. Hence, further investigations are needed to validate our results.

There were some key limitations to this study. First, fixed-effects models were used for the meta-analysis of several trials, which were performed independently. Although the fixed-effects model is based on an assumption that all of the trials included in the analysis were functionally identical, it is unlikely that all of the trials included in this study had completely equivalent functions. If the inherent heterogeneity is larger than expected, it is difficult to assume a common effect size. However, we believe that this statistical method was reasonable for the screening and discovery of prognostic SNPs in this study, as all patients were treated with standard first-line chemotherapy. In fact, the Cochran's Q statistic could statistically deny heterogeneity. Second, this was a prospective–retrospective study to identify biomarkers; thus, the results are not necessarily applicable to other populations. According to published guidelines, one or more validation studies are required to reach a level of evidence for medical utility (33, 34). Therefore, further studies are needed to validate our results. Third, the GTEx analysis was performed on normal colon tissue. Finally, because we do not have the exome sequencing data from tumor samples, we tested only germline polymorphisms from blood samples. However, based on the previous information notifying a high concordant rate of genotype for polymorphisms in DNA isolated from peripheral blood and colorectal cancer tumor samples (35, 36), we believe the impact of discordance between them is very small and would not change outcome.

In conclusion, RBP-related SNPs may be associated with the prognosis of patients with mCRC treated with standard first-line chemotherapy. The predictive values for targeted drugs and cytotoxic backbone chemotherapy regimens warrant further prospective evaluations. However, these novel findings may have translational therapeutic implications.

No disclosures were reported.

H. Arai: Conceptualization, writing–original draft. S. Cao: Formal analysis, methodology, writing–review and editing. F. Battaglin: Writing–review and editing. J. Wang: Writing–review and editing. N. Kawanishi: Writing–review and editing. R. Tokunaga: Writing–review and editing. F. Loupakis: Resources, writing–review and editing. S. Stintzing: Resources, writing–original draft. S. Soni: Writing–review and editing. W. Zhang: Writing–review and editing. C. Mancao: Resources, writing–review and editing. B. Salhia: Resources, writing–review and editing. S.M. Mumenthaler: Resources, writing–review and editing. C. Cremolini: Resources, writing–review and editing. V. Heinemann: Resources, writing–review and editing. A. Falcone: Resources, writing–review and editing. J. Millstein: Formal analysis, methodology, writing–review and editing. H.-J. Lenz: Resources, supervision, funding acquisition, writing–review and editing.

The authors thank all patients who contributed to this study.

This work was supported by the NCI (P30CA 014089 to H.-J. Lenz), Gloria Borges WunderGlo Foundation, Dhont Family Foundation, San Pedro Peninsula Cancer Guild, and Daniel Butler Research Fund.

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

1.
Pereira
B
,
Billaud
M
,
Almeida
R
. 
RNA-binding proteins in cancer: old players and new actors
.
Trends Cancer
2017
;
3
:
506
28
.
2.
Dreyfuss
G
,
Kim
VN
,
Kataoka
N
. 
Messenger-RNA-binding proteins and the messages they carry
.
Nat Rev Mol Cell Biol
2002
;
3
:
195
205
.
3.
Gerstberger
S
,
Hafner
M
,
Tuschl
T
. 
A census of human RNA-binding proteins
.
Nat Rev Genet
2014
;
15
:
829
45
.
4.
Brinegar
AE
,
Cooper
TA
. 
Roles for RNA-binding proteins in development and disease
.
Brain Res
2016
;
1647
:
1
8
.
5.
Mohibi
S
,
Chen
X
,
Zhang
J
. 
Cancer the ‘RBP’eutics-RNA-binding proteins as therapeutic targets for cancer
.
Pharmacol Ther
2019
:
107390
.
6.
Chatterji
P
,
Rustgi
AK
. 
RNA binding proteins in intestinal epithelial biology and colorectal cancer
.
Trends Mol Med
2018
;
24
:
490
506
.
7.
Tu
HC
,
Schwitalla
S
,
Qian
Z
,
LaPier
GS
,
Yermalovich
A
,
Ku
YC
, et al
LIN28 cooperates with WNT signaling to drive invasive intestinal and colorectal adenocarcinoma in mice and humans
.
Genes Dev
2015
;
29
:
1074
86
.
8.
Sureban
SM
,
Ramalingam
S
,
Natarajan
G
,
May
R
,
Subramaniam
D
,
Bishnupuri
KS
, et al
Translation regulatory factor RBM3 is a proto-oncogene that prevents mitotic catastrophe
.
Oncogene
2008
;
27
:
4544
56
.
9.
Venugopal
A
,
Subramaniam
D
,
Balmaceda
J
,
Roy
B
,
Dixon
DA
,
Umar
S
, et al
RNA binding protein RBM3 increases beta-catenin signaling to increase stem cell characteristics in colorectal cancer cells
.
Mol Carcinog
2016
;
55
:
1503
16
.
10.
Ye
S
,
Song
W
,
Xu
X
,
Zhao
X
,
Yang
L
. 
IGF2BP2 promotes colorectal cancer cell proliferation and survival through interfering with RAF-1 degradation by miR-195
.
FEBS Lett
2016
;
590
:
1641
50
.
11.
Lederer
M
,
Bley
N
,
Schleifer
C
,
Huttelmaier
S
. 
The role of the oncofetal IGF2 mRNA-binding protein 3 (IGF2BP3) in cancer
.
Semin Cancer Biol
2014
;
29
:
3
12
.
12.
Noubissi
FK
,
Elcheva
I
,
Bhatia
N
,
Shakoori
A
,
Ougolkov
A
,
Liu
J
, et al
CRD-BP mediates stabilization of betaTrCP1 and c-myc mRNA in response to beta-catenin signalling
.
Nature
2006
;
441
:
898
901
.
13.
Li
Z
,
Zhang
Y
,
Ramanujan
K
,
Ma
Y
,
Kirsch
DG
,
Glass
DJ
. 
Oncogenic NRAS, required for pathogenesis of embryonic rhabdomyosarcoma, relies upon the HMGA2-IGF2BP2 pathway
.
Cancer Res
2013
;
73
:
3041
50
.
14.
Kudinov
AE
,
Karanicolas
J
,
Golemis
EA
,
Boumber
Y.
Musashi RNA-binding proteins as cancer drivers and novel therapeutic targets. Clin Cancer Res
2017
;
23
:
2143
53
.
15.
Cai
J
,
Wang
H
,
Jiao
X
,
Huang
R
,
Qin
Q
,
Zhang
J
, et al
The RNA-binding protein HuR confers oxaliplatin resistance of colorectal cancer by upregulating CDC6
.
Mol Cancer Ther
2019
;
18
:
1243
54
.
16.
Loupakis
F
,
Cremolini
C
,
Masi
G
,
Lonardi
S
,
Zagonel
V
,
Salvatore
L
, et al
Initial therapy with FOLFOXIRI and bevacizumab for metastatic colorectal cancer
.
N Engl J Med
2014
;
371
:
1609
18
.
17.
Heinemann
V
,
von Weikersthal
LF
,
Decker
T
,
Kiani
A
,
Vehling-Kaiser
U
,
Al-Batran
S-E
, et al
FOLFIRI plus cetuximab versus FOLFIRI plus bevacizumab as first-line treatment for patients with metastatic colorectal cancer (FIRE-3): a randomised, open-label, phase 3 trial
.
Lancet Oncol
2014
;
15
:
1065
75
.
18.
Parikh
AR
,
Lee
FC
,
Yau
L
,
Koh
H
,
Knost
J
,
Mitchell
EP
, et al
MAVERICC, a randomized, biomarker-stratified, phase II study of mFOLFOX6-bevacizumab versus FOLFIRI-bevacizumab as first-line chemotherapy in metastatic colorectal cancer
.
Clin Cancer Res
2019
;
25
:
2988
95
.
19.
Amos
CI
,
Dennis
J
,
Wang
Z
,
Byun
J
,
Schumacher
FR
,
Gayther
SA
, et al
The OncoArray Consortium: a network for understanding the genetic architecture of common cancers
.
Cancer Epidemiol Biomarkers Prev
2017
;
26
:
126
35
.
20.
King
CE
,
Cuatrecasas
M
,
Castells
A
,
Sepulveda
AR
,
Lee
JS
,
Rustgi
AK
. 
LIN28B promotes colon cancer progression and metastasis
.
Cancer Res
2011
;
71
:
4260
8
.
21.
King
CE
,
Wang
L
,
Winograd
R
,
Madison
BB
,
Mongroo
PS
,
Johnstone
CN
, et al
LIN28B fosters colon cancer migration, invasion and transformation through let-7-dependent and -independent mechanisms
.
Oncogene
2011
;
30
:
4185
93
.
22.
Madison
BB
,
Liu
Q
,
Zhong
X
,
Hahn
CM
,
Lin
N
,
Emmett
MJ
, et al
LIN28B promotes growth and tumorigenesis of the intestinal epithelium via Let-7
.
Genes Dev
2013
;
27
:
2233
45
.
23.
Ye
Y
,
Madison
B
,
Wu
X
,
Rustgi
AK
. 
A LIN28B polymorphism predicts for colon cancer survival
.
Cancer Biol Ther
2012
;
13
:
1390
5
.
24.
Lu
L
,
Katsaros
D
,
Mayne
ST
,
Risch
HA
,
Benedetto
C
,
Canuto
EM
, et al
Functional study of risk loci of stem cell-associated gene lin-28B and associations with disease survival outcomes in epithelial ovarian cancer
.
Carcinogenesis
2012
;
33
:
2119
25
.
25.
Fu
W
,
Liu
GC
,
Zhao
Z
,
Zhu
J
,
Jia
W
,
Zhu
SB
, et al
The correlation between LIN28B gene potentially functional variants and Wilms tumor susceptibility in Chinese children
.
J Clin Lab Anal
2018
;
32
:
e22200
.
26.
Yang
Z
,
Deng
Y
,
Zhang
K
,
Bai
Y
,
Zhu
J
,
Zhang
J
, et al
LIN28B gene polymorphisms modify hepatoblastoma susceptibility in Chinese children
.
J Cancer
2020
;
11
:
3512
8
.
27.
Ballman
KV
. 
Biomarker: predictive or prognostic?
J Clin Oncol
2015
;
33
:
3968
71
.
28.
Dai
N
,
Ji
F
,
Wright
J
,
Minichiello
L
,
Sadreyev
R
,
Avruch
J
. 
IGF2 mRNA binding protein-2 is a tumor promoter that drives cancer proliferation through its client mRNAs IGF2 and HMGA1
.
eLife
2017
;
6
:
e27155
.
29.
Oliveira-Mateos
C
,
Sanchez-Castillo
A
,
Soler
M
,
Obiols-Guardia
A
,
Pineyro
D
,
Boque-Sastre
R
, et al
The transcribed pseudogene RPSAP52 enhances the oncofetal HMGA2-IGF2BP2-RAS axis through LIN28B-dependent and independent let-7 inhibition
.
Nat Commun
2019
;
10
:
3979
.
30.
Shen
K
,
Cui
D
,
Sun
L
,
Lu
Y
,
Han
M
,
Liu
J
. 
Inhibition of IGF-IR increases chemosensitivity in human colorectal cancer cells through MRP-2 promoter suppression
.
J Cell Biochem
2012
;
113
:
2086
97
.
31.
Dallas
NA
,
Xia
L
,
Fan
F
,
Gray
MJ
,
Gaur
P
,
van Buren
G
 2nd
, et al
Chemoresistant colorectal cancer cells, the cancer stem cell phenotype, and increased sensitivity to insulin-like growth factor-I receptor inhibition
.
Cancer Res
2009
;
69
:
1951
7
.
32.
Flanigan
SA
,
Pitts
TM
,
Eckhardt
SG
,
Tentler
JJ
,
Tan
AC
,
Thorburn
A
, et al
The insulin-like growth factor I receptor/insulin receptor tyrosine kinase inhibitor PQIP exhibits enhanced antitumor effects in combination with chemotherapy against colorectal cancer models
.
Clin Cancer Res
2010
;
16
:
5436
46
.
33.
Simon
RM
,
Paik
S
,
Hayes
DF
. 
Use of archived specimens in evaluation of prognostic and predictive biomarkers
.
J Natl Cancer Inst
2009
;
101
:
1446
52
.
34.
Polley
MY
,
Freidlin
B
,
Korn
EL
,
Conley
BA
,
Abrams
JS
,
McShane
LM
. 
Statistical and practical considerations for clinical evaluation of predictive biomarkers
.
J Natl Cancer Inst
2013
;
105
:
1677
83
.
35.
Shao
W
,
Ge
Y
,
Ma
G
,
Du
M
,
Chu
H
,
Qiang
F
, et al
Evaluation of genome-wide genotyping concordance between tumor tissues and peripheral blood
.
Genomics
2017
;
109
:
108
12
.
36.
van Huis-Tanja
L
,
Kweekel
D
,
Gelderblom
H
,
Koopman
M
,
Punt
K
,
Guchelaar
HJ
, et al
Concordance of genotype for polymorphisms in DNA isolated from peripheral blood and colorectal cancer tumor samples
.
Pharmacogenomics
2013
;
14
:
2005
12
.