Search

Background Second-line therapy has consistently demonstrated survival benefit if compared with best supportive care; however, there is limited evidence whether further lines of treatment may improve the prognosis of advanced gastric cancer (AGC) patients. Materials and methods Starting from a real-world cohort of 868 AGC patients, we retrospectively analyzed baseline parameters, tumor characteristics, and treatment data of those treated with at least three lines. Categorical features were described through cross-tables and chi-square test. We explored the impact of treatment intensity and progression-free survival (PFS) experienced in previous lines on PFS and overall survival in third-line by uni- and multivariate Cox regression models and described by Kaplan-Meier estimator plot with log-rank test. Results Overall, 300 patients were included in the analysis. The most common site of primary tumor was gastric body; 45.3% of cancers had an intestinal histotype, 14% were human epidermal growth receptor 2 positive. In third-line, 45.7% of patients received a single-agent chemotherapy, 49.7% a combination regimen. Patients who had experienced a first-line PFS ≥6.9 months had a better prognosis compared with those who had achieved a shorter one. Consistently, a second-line PFS ≥3.5 months positively influenced the prognosis. Patients receiving a third-line combination regimen had better outcomes compared with those treated with a single-agent chemotherapy. Conclusion Our real-world study confirms that selected AGC patients may receive third-line treatment. Longer PFS in previous lines or a more intense third-line treatment positively influenced prognosis. Further efforts are warranted to define the best therapeutic sequences, and to identify the optimal candidate for treatment beyond second-line. Implications for practice The benefit of third-line treatment to advanced gastric cancer patients is controversial. This study depicts a real scenario of the clinical practice in Italy, confirming that a non-negligible proportion of patients receive a third-line therapy. Longer progression-free survival in previous treatment lines or higher third-line treatment intensity positively influenced prognosis. Including a large number of real-world patients, this study provides information on third-line treatment from the daily clinical practice; moreover, its results help in defining the best therapeutic sequence and offer some hints to select the optimal candidate for treatment beyond second-line.

Regorafenib is a multi-targeting kinase inhibitor approved for the treatment of metastatic colorectal cancer patients in refractory to standard chemotherapy. Similarly to sorafenib, this agent was originally developed as a RAF1 inhibitor. However, the kinase inhibitory profile is distinct from sorafenib. A broad-spectrum of kinase inhibition induces wide-range drug sensitivity, irrespective of mutation status of major oncogenes. This agent's main therapeutic effects are anti-angiogenesis and the remodeling of tumor microenvironment through several mechanisms of action. The dual blockade of VEGF receptors and TIE2 can lead to both additive anti-angiogenesis effects and the suggestive unique regulation of vessel stability. Additionally, it inhibits molecular escape pathways to VEGF inhibition (e.g., FGF, PIGF, and PDGF signaling), enabling its continuous antiangiogenic effect even in tumors resistant to VEGF inhibitors. Furthermore, regorafenib has the important effect of enhancing anti-tumor immunity via macrophage modulation. Based on this concept, clinical trials have been recently launched for the development of a combination strategy with immune checkpoint inhibitors. Contrary to regorafenib induced clinical benefits and advances in the novel strategy, currently no predictive biomarkers have been identified. In the present review, we revisit and summarize regorafenib's unique mechanisms of action. The review could highlight molecular insights and provide some perspective for the search of predictive biomarkers used in metastatic colorectal cancer patients treated with regorafenib.

Background Though molecular classifications have been well studied, targeted therapies for GC are still limited. CDH1 mutations are characteristics of genomically stable GC and associated with poor prognosis. Herein, to understand the specific mutation profile and enriched pathways in CDH1-mutated (MT) GC could help to discover novel drug targets. Methods GC (N = 1596) were analyzed by next-generation sequencing (NGS) and immunohistochemistry (Caris Life Sciences, Phoenix, AZ). Tumor mutational burden (TMB) was calculated based on somatic nonsynonymous mutation, and microsatellite instability (MSI) was evaluated by NGS of known MSI loci. Gene fusions were evaluated using Archer (N = 253) or Whole Transcriptome Sequencing (WTS, N = 96). Results The overall mutation frequency of CDH1 was 9.7%. A significant association of CDH1 mutations with cancer types (GC, 12% vs gastroesophageal junction adenocarcinoma, 2%), sex (female, 13% vs male, 8%), Lauren subtypes (diffuse, 27% vs intestinal, 1%) and younger age (58 vs 62) were observed (p 0) (57% vs. 73%, p Conclusions This is the largest study to explore the distinct genomic landscape in CDH1-MT GC. It indicates CDH1-MT GC patients could potentially benefit from agents targeting WRN, POT1, CDK12 and IGF1R, while immunotherapy may be of lower efficacy in this cohort based on lower CPS score. Efficiency of these therapeutic targets and enriched pathways in CDH1-MT GC warrant further investigation. Legal entity responsible for the study Norris Comprehensive Cancer Center, Keck School of Medicine, University of Southern California. Funding Has not received any funding. Disclosure All authors have declared no conflicts of interest.

Background Although rare, the incidence of BM is increasing due to the improvement in metastatic CRC treatment and longer survival. Genomic analyses of small series revealed that BM can harbor potentially unique driver mutations. We aimed to comprehensively characterize the molecular profile of BM and explore the differences between BM vs other distant metastases (OM) and primary tumors (PT) in CRC. Methods Tumor samples from BM (n = 81), PT (n = 6898) and OM (n = 5918) were analyzed using NGS (TruSEQ on 45 genes or NextSEQ on 592 genes), in situ hybridization and immunohistochemistry (Caris Life Sciences, Phoenix, AZ). Tumor mutational burden (TMB) was calculated based on somatic nonsynonymous missense mutations, and microsatellite instability (MSI) was evaluated by NGS of known MSI loci. Results The most frequently mutated genes in BM were TP53 (80%), APC (73%), KRAS (68%), ARID1A (18%), PIK3CA (13%) and FBXW7 (9%). The most prevalent copy number increase was seen in CDX2 (20%), CCND2 (7%), FLT1 (7%), FLT3 (7%) and FOXO1 (7%). When compared to OM and PT, mutations in KRAS (BM: 68%, OM: 49%, PT: 48%), CDKN2A (5%, 1%, 1%), ERCC2 (2%, 0, 0) and HRAS (1%, 0, 0) were significantly higher in BM (P 17mut/MB) and MSI-high were seen in BM vs PT (2 vs 9% and 1 vs 8%, P=.049 and .031, respectively) but not compared to OM. No RSPO3 nor NTRK1 fusions were seen in BM (n = 5). Female gender was associated with younger age in BM (53.5 vs 62 yr, P=.0014) and OM (58.8 vs 60.2 yr, P Conclusions This is the largest and most extensive profiling study to investigate the molecular makeup of BM and the differences with PT and OM in CRC. Our data show distinct mutations and CNA characterizing BM and lower expression of immune related markers, supporting the rationale to develop tailored approaches to the treatment of this metastatic site. Legal entity responsible for the study The authors. Funding National Cancer Institute grant number P30CA014089, Gloria Borges WunderGlo Foundation-The Wunder Project, Dhont Family Foundation, San Pedro Peninsula Cancer Guild, Daniel Butler Research Fund, Call to Cure Research Fund and Fong Research Project. Disclosure J. Xiu: Full / Part-time employment: Caris Life Sciences. Y. Baca: Full / Part-time employment: Caris Life Sciences. R.M. Goldberg: Research grant / Funding (self), Travel / Accommodation / Expenses: Caris Life Sciences. A. Grothey: Travel / Accommodation / Expenses: Caris Life Sciences. A.F. Shields: Research grant / Funding (self), Travel / Accommodation / Expenses: Caris Life Sciences. A. Seeber: Advisory / Consultancy: Caris Life Sciences. M.E. Salem: Travel / Accommodation / Expenses: Caris Life Sciences. P.A. Philip: Travel / Accommodation / Expenses: Caris Life Sciences. J.L. Marshall: Advisory / Consultancy: Caris Life Sciences. W..M. Korn: Full / Part-time employment: Caris Life Sciences. H.J. Lenz: Travel / Accommodation / Expenses: Caris Life Sciences. All other authors have declared no conflicts of interest.

Purpose: MAEA, also known as Emp (Erythroblast Macrophage protein), plays an important role in actin cytoskeleton rearrangement in macrophages and erythroid cells. A recent study indicated that down-regulation of MAEA in mouse monocyte macrophage results in abnormal cell motility and higher expression of mitogen-activated protein kinase 1 (MAPK 1) and thymoma viral proto-oncogene 1 (Akt 1). However, the precise mechanism of MAEA in cancer remains unclear. In this study, we examined the role of MAEA in cell motility, invasion and chemosensitivity in colorectal cancer (CRC) cell lines. Experimental Procedure: CRC cell lines were selected based on MSI/MSS status. MAEA was overexpressed in HCT116, SW480, LoVo, SW626 cell lines using vector pRP[Exp]-EGFP/Puro-CMV>hMAEA[ORF013172]. MAEA expression was quantified by qRT-PCR and immunoblotting. To determine the role of MAEA in migration and invasion, wound closure and transwell invasion assays were performed. Immunofluorescent staining was carried out to study actin cytoskeletal remodeling. The effect of MAEA on chemosensitivity was evaluated by using MTS assay after treatment with 5-FU, oxaliplatin and irinotecan. Both cell motility and DNA repair related genes expression profile were analyzed by human RT2 Profiler PCR array. Result: Our results showed that up-regulation of MAEA significantly suppressed migration and invasion in HCT116 cell line (MSI, RAS mutant, BRAF wild type, APC wild type). Furthermore, immunofluorescence staining indicated actin cytoskeleton projections were disrupted with up-regulation of MAEA. In MTS assay, MAEA overexpressed HCT116 demonstrated enhanced chemo sensitivity to 5-FU, oxaliplatin, and irrinotecan (35%, 40%, and 40% respectively). PCR array analysis showed significant downregulation of cell motility related genes AKT1, BCAR1, HGF, ITGA4, ITGB3, MAPK1, RHOB, SRC (2.4, 2.4, 50, 9.1, 2.0, 1.56, 2.38, and 2.38 folds respectively). Similarly, DNA repair related genes DMC1, ERCC1, ERCC2, ERCC3, ERCC4, RAD51D, RAD52 were downregulated (1.63, 1.64, 1.56, 1.66, 1.66, 2.0, 1.75 folds respectively). Conclusion: Our findings show MAEA suppresses migration, invasion and enhances chemo sensitivity to first-line of treatment in colorectal cancer cell lines. Downregulation of both MAPK1 and AKT1 suggests that MAEA may be involved in PI3K-ATK and/or RAS-ERK pathways regulating cell motility and invasion. Downregulation of DNA repair related genes can be a plausible reason for enhanced chemosensitivity in MSI-high CRC cell lines. Citation Format: Jae Ho Lo, Shivani Soni, Ryuma Tokunaga, Francesca Battaglin, Madiha Naseem, Alberto Puccini, Hiroyuki Arai, Tricia Ning, Martin D. Berger, Shu Cao, Wu Zhang, Josh Millstein, Heinz-Josef Lenz. MAEA (macrophage erythroblast attacher) suppresses migration, invasion and enhances chemosensitivity in colorectal cancer cell lines [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2864.

4085 Background: Biliary tract cancers constitute ~3% of cancers worldwide with incidence increasing, especially for intrahepatic cholangiocarcinoma (IHC). The prognosis of these tumors remains dismal and novel treatment strategies are needed to improve overall survival. BRCA mutations occur in biliary tract cancers but their frequency in distinct sites of biliary tract cancer is unknown. Moreover, no data are available correlating BRCA mutation with immunogenic markers such as TMB, MSI, or PD-L1 expression. Methods: Tumor samples from 1288 primary biliary tract cancers, comprising IHC (n = 746), extrahepatic cholangiocarcinoma (EHC) (n = 189), gallbladder (GBC) (n=353) were profiled at Caris Life Sciences, Phoenix, AZ. Testing included NextGen SEQ (MiSeq on 47 genes, NextSeq on 592 genes) and PD-L1 IHC (SP142). TMB was calculated based on somatic nonsynonymous missense mutations, and MSI was evaluated by NGS of known MSI loci. Results: BRCA mutations were detected in 3.6% (N = 46) of samples ( BRCA1 0.6%, BRCA2 3%), no differences were seen based on the site of the tumor. In GBC and IHC BRCA2 mutations (4.0% and 2.7%) were more frequent than BRCA1 (0.3% and 0.4, p < 0.05) while in EHC, similar frequency was observed ( BRCA1: 2.1%; BRCA2: 2.6%). There was no significant association with gender or age. In BRCA-mutant biliary tract cancer the most frequently mutated genes were TP53 (55.6%), ARID1A (52.2%) and KRAS (26.1%), KMT2D/C (20%, 13%) and CDKN2A(13%). Overall, BRCA mutations were associated with a higher rate of MSI-H (19.5% vs 1.7%, p = 0.001) and higher TMB in both MSI-H and MSS tumors (p

3545 Background: RNA binding proteins (RBPs) post-transcriptionally regulate gene expression by stabilizing or destabilizing target messenger RNA. Although alteration of RBPs affects many steps of cancer development, its clinical implication in mCRC remains unclear. Methods: We analyzed data from mCRC patients (pts) enrolled in three first-line randomized trials (FIRE-3, TRIBE, and MAVERICC). Genomic DNA from blood samples of pts was genotyped through the OncoArray, a custom array manufactured by Illumina. Candidate 30 SNPs in 10 RBP genes (MSI1, MSI2, ELAVL1, RBM3, LIN28A, LIN28B, IGF2BP1, IGF2BP2, IGF2BP3, ZFP36) were tested on association with progression-free survival (PFS) and overall survival (OS). To evaluate prognostic effects and heterogeneities across treatment arms, meta-analysis approach using the METASOFT software was conducted. We also tested interaction between each SNP and treatment within each trial, i.e. FIRE-3 cohort (FOLFIRI+cetuximab (Cet) vs FOLFIRI+bevacizumab (Bev)) and MAVERICC cohort (FOLFIRI+Bev vs FOLFOX6+Bev). For multiple testing, p values were adjusted by the false discovery rate (FDR) method. Results: A total of 884 pts’ SNPs data were available (FIRE-3: n = 236, TRIBE: n = 324, and MAVERICC: n = 324). Meta-analysis combining three trials showed RBM3 rs926152 (adjusted p = 0.045) and RBM3 rs2249585 (adjusted p = 0.016) were significantly prognostic for PFS. Whereas, in terms of OS, only LIN28B rs314277 (adjusted p = 0.045) was significant, and RBM3 rs926152 (adjusted p = 0.057) and RBM3 rs2249585 (adjusted p = 0.059) had a trend. Interaction test showed several SNPs were potentially predictive (raw p < 0.05), although without any significance after FDR adjustment: in FIRE-3 cohort, MSI2 rs1822381, RBM3 rs926152, LIN28B rs221635, IGF2BP1 rs2969 for OS; in MAVERICC cohort, MSI1 rs1179442 and MSI2 rs3826301 for OS, ELAVL1 rs4804244 for PFS. Conclusions: Our results indicate prognostic potential of SNPs in RBP genes, such as RBM3 and LIN28B, in mCRC. However, we find no distinct evidence that these SNPs can predict differential effect between Cet and Bev, or between oxaliplatin- and irinotecan-based chemotherapy.

3145 Background: BRCA muts in CRC are associated with a higher tumor mutation burden irrespective of microsatellite instability, which highlights the possibility of using PARP-inhibitors(i) in CRC in the future. Early phase studies have shown that combination of PARP-i with oxaliplatin or irinotecan enhances tumor lysis in CRC. In this study, we investigated the influence of mutations in the Homologous Repair Pathway genes on survival outcomes among mCRC pts treated with oxaliplatin or irinotecan-based regimens. Methods: The impact of selected SNPs within 4 genes (BRCA1, BRCA2, RAD51, BARD1) on OS/PFS was analyzed through the OncoArray, a custom array manufactured by Illumina, on genomic DNA from blood samples of 431 pts enrolled in 2 randomized trials. TRIBE FOLFIRI/bev arm (n = 215, mPFS/OS: 9.7/26.2 mo) served as discovery cohort, FIRE-3 FOLFIRI/bev arm (n = 107, mPFS/OS: 11.5/31.4 mo) as validation and TRIBE FOLFOX/bev arm (n = 109, mPFS/OS: 10.8/26 mo) as control. Results: Significant associations were found among carriers of BRCA1 rs8176318 SNP, where C > A base change is known to reduce BRCA1 expression among CRC cells. In the discovery cohort, pts with A/A had shorter OS (22.4 vs 27.3 mo, P = .009) and PFS (7.5 vs 10.5 mo, P = .0006) compared to carriers of any C allele in both univariate and multivariate analysis. Same results were observed in pts with left-sided CRCs (PFS-7.5 vs 11 mo, P = .005; OS- 25.6 vs 32.3, P = .034) and among males (PFS- 7.5 vs 10.3 mo, P = .008; OS- 25.7 vs 31.3 mo, P = .008) in both uni and multivariate analysis. These results were also seen in the validation cohort: A/A carriers in left-sided CRCs had poor OS (26.1 vs 36.0 mo, P = .027) and PFS (9.5 vs 11.7 mo, P = .002. Males with A/A genotype also had poor OS (24.7 vs 32.5 mo, P = .028) and PFS 96.9 vs 12.2 mo; P = .0002). In the control cohort, A/A genotype carriers had poor tumor response in overall (P = .011) and left-sided disease (P = .034). These outcomes were independent of KRAS mutation status. No significant relationship was observed among females with mCRC. Conclusions: This is the first study to report that BRCA1 mut influence survival outcomes among mCRC pts, particularly among males and those with left-sided disease. Prospective trials are warranted to assess the utility of routine BRCA mut testing and the role of PARP-i in improving survival outcomes in this pt population.

3595 Background: PD is one of the most common age-related neurodegenerative disorders. Large epidemiological studies have consistently reported a reduced risk of CRC in PD patients (pts), but the biology behind this evidence is unclear. The methylation status of SNCA, one of the causal PD genes, has been identified as a tool for CRC screening and early diagnosis when detected in stool samples, and alterations in core PD genes are prevalent across human malignancies including CRC. Methods: The impact on outcome of 13 SNPs within 6 core PD genes ( SNCA, PRKN, UCHL1, PINK1, DJ-1, LRRK2) was analyzed in pts enrolled in the randomized FIRE-3 trial. Genomic DNA from blood samples of pts treated with first-line FOLFIRI-cetuximab (cet, n = 129) and FOLFIRI-bevacizumab (bev, n = 107) was genotyped through the OncoArray, a custom array manufactured by Illumina. Gene expression levels were measured from 102 tumor samples of pts in the cet arm by HTG EdgeSeq Oncology Biomarker Panel. Results: In the cet cohort, pts carrying the G/G variants of SNCA rs356165 and rs2736990 had significantly shorter mOS (30 vs 41.1 mo) compared to any A genotype in both uni- and multivariable analysis (adjusted P[ Padj] = .047 and .042, respectively). LRKK2 rs3761863 T/T allele carriers showed shorter mPFS (9.5 vs 13.3 mo, Padj = .01), while rs11564148 any A carriers had longer mPFS (14.2 vs 10.2 mo, Padj = .01) compared to reference genotypes. LRKK2 rs11564148 any A carriers also showed longer mOS in multivariable analysis (43.7 vs 33.2 mo, Padj = .044). Any C allele carriers of PINK1 rs1043424 showed longer mPFS in uni- and multivariable analysis ( Padj < .001). No significant interaction was found with gender, tumor location and RAS status. These associations were not observed in bev arm. High SNCA expression was associated with worse mPFS (log2 > 7.89, 5.9 vs 11.2 mo) and mOS (log2 > 7.68, 17.9 vs 31.1 mo) in FIRE-3 cet arm ( P < .05). Conclusions: We provide the first evidence that gene expression and genetic variants in PD genes may have a predictive value in mCRC pts receiving first-line cetuximab-based treatment. Our findings open new perspectives on the role of PD genes in CRC biology warranting further investigation.

2633 Background: T-cell activation plays a key role in maintaining an effective host immunity and antitumor control. Targeting costimulatory immune checkpoint proteins can lead to increased antitumor immunity. We hypothesize, that variations in genes encoding for T-cell activation molecules may predict outcome in stage II and III colon cancer patients. Methods: The impact of 4 functional single nucleotide polymorphisms (SNPs) within the LIGHT, ICOS, CD80 and GITR genes on time to recurrence and overall survival was evaluated in 209 patients with stage II and III colon cancer. Genomic DNA was extracted from formalin-fixed paraffin embedded tissue and the SNPs were analyzed by PCR-based direct sequencing. Results: Baseline characteristics were as follows: median age = 70y (19-91); female/male ratio = 41.6% / 58.4%; 111 patients had stage II, and 98 stage III colon cancer. The LIGHT rs3760746 SNP showed significant association with recurrence rate in the overall population. Patients harboring any G allele had a lower 3-years recurrence rate compared to those with a A/A genotype (16% vs 30%) in both univariate (HR 0.53, 95% Confidence intervall (CI) 0.29-0.96, p = 0.033) and multivariate analyses (HR 0.52, 95% CI 0.29-0.95, p = 0.034). This trend was most evident among patients with stage III colon cancer. Here again, G allele carriers had both a lower 3-years recurrence and a longer 5-years overall survival rate compared to those having a A/A genotype (21% vs 40% and 77% vs 43%) in both univariate (HR 0.42, 95% CI 0.19-0.90, p = 0.021 and HR 0.51, 95% CI 0.25-1.03, p = 0.046) and multivariate analyses (HR 0.43, 95% CI 0.20-0.93, p = 0.033 and HR 0.30, 95% CI 0.14-0.65, p = 0.002). Conclusions: Our results provide the first evidence that polymorphisms within the T-cell costimulatory LIGHT gene might serve as prognostic markers in patients with stage II and III colon cancer. Targeting LIGHT might be a promising approach to further optimize treatment options and to improve outcome of colon cancer patients in the adjuvant setting.

3048 Background: Strong evidence supports the critical role of the gut-brain axis in modulating gastrointestinal function and homeostasis. Available data suggest an involvement of the dopaminergic pathway in CRC dynamics. DA could inhibit proliferation and migration of tumor endothelial cells and enhanced 5-fluorouracil efficacy in CRC preclinical models. Hence, we hypothesized that genetic variants in DA signaling may predict treatment outcomes in mCRC pts. Methods: The impact on outcome of 22 selected single nucleotide polymorphisms (SNPs) in 9 genes of the DA signaling pathway ( DRD1, DRD2, DRD3, DRD4, DRD5, TAAR1, SLC6A3, SLC18A2, PPP1R1B) was analyzed on a total of 884 pts enrolled in three independent randomized first-line trials: TRIBE (n = 324), MAVERICC (n = 324), and FIRE-3 (n = 236). Genomic DNA from blood samples of pts was genotyped through the OncoArray, a custom array manufactured by Illumina. A meta-analysis approach using the METASOFT software was used to quantify SNPs prognostic effects and heterogeneities across treatment arms. P values were adjusted for multiple testing using the false discovery rate (FDR) method. Results: Overall, DRD3 rs3732790, rs9817063 and rs2134655 showed a significant nominal p value ( P) in association with tumor response (TR) across trials ( P= 0.032, P= 0.021, P= 0.027, respectively). TAAR1 rs8192620 showed an association with both progression free survival (PFS) ( P= 0.01) and overall survival (OS) ( P= 0.033), similar to DA transporter SLC6A3 rs6347 ( P= 0.016 and P= 0.002, respectively). SLC6A3 rs6347 association with OS remained significant after FDR ( PFDR= 0.045). Subgroup analyses showed a significant association with PFS for DRD1 rs267410 and SLC6A3 rs2652510 in females ( PFDR= 0.056), and between SLC6A3 rs6347 and OS ( PFDR= 0.041) and SLC6A3 rs6876890 and TR ( PFDR= 0.05) in KRAS wild type. Conclusions: Our results suggest that SNPs in DA signaling may have a prognostic value in mCRC pts receiving first-line treatment. Upon validation, these findings may provide novel insight on the role of DA signaling in CRC and possibly contribute to open novel therapeutic perspectives.

4133 Background: In the U.S. 56,000 Americans are expected to be diagnosed with pancreatic cancer in 2019. Prognosis in pancreatic cancer is poor. Therefore, new treatment strategies are urgently needed to improve survival. BRCA1 and BRCA2 mutations have been described to be the most common genetic mutations involved in familial pancreatic cancer. The optimal treatment regimen to use in BRCA-mutant pancreatic cancer has still to be established. Moreover, no data are available on association of BRCA mutation with immune-associated markerssuch as tumor mutational burden (TMB), microsatellite instability (MSI) or PD-L1 expression. Methods: Tumor samples of 2824 patients with pancreatic ductal adenocarcinoma were analyzed for BRCA mutation by NGS and for other genes (MiSeq on 47 genes, NextSeq on 592 genes) at Caris Life Sciences, Phoenix, AZ. TMB was calculated based on somatic nonsynonymous missense mutations, and MSI was evaluated by NGS of known MSI loci. PD-L1 expression was evaluated using immunohistochemistry. Results: In 4.4% (N = 124) of all pancreatic adenocarcinoma samples BRCA mutations were detected. BRCA2 mutations were more common: 3.1% (N = 89) vs 1.1% BRCA1 mutations (N = 35). BRCA mutations were associated with younger age ( BRCA1: 61 yrs for mutated vs. 64 for wild-type, p = 0.07; BRCA2: 61 yrs vs. 64, p = 0.002; both: p < 0.001). BRCA mutations were associated with higher MSI-H frequency (4.8% vs. 1.2%, p = 0.002), elevated PD-L1 expression (22% vs. 11%, p < 0.001) and higher TMB (mean 8.7 mut/MB vs. 6.5, p < 0.001); the differences remain significant in MSS tumors (p < 0.05). BRCA-mutant pancreatic carcinomas showed a significantly lower mutation frequencies in TP53 (59% vs 73%, p = 0.001), CKDN2A (13% vs 25%, p = 0.006), but higher frequencies in APC (6.5% vs 2.2%), KMT2A (1.9% vs 0.2%), AMER1 (1.9 vs 0.5%) and SETD2 (3.7% vs 0.4%) mutations (p < 0.05 for all comparisons). Conclusions: BRCA mutations are found in a significant subgroup of pancreatic ductal adenocarcinoma and these carcinomas are associated with an immunogenic tumor profile. These data suggest evaluating PARP inhibitors in combination with immunotherapy in patients with BRCA-mutant pancreatic adenocarcinoma especially in tumors that are MSS.

3600 Background: The SWI/SNF complex includes proteins produced by 29 genes that regulate chromatin structure remodeling through effects upon transcription, replication, and repair. Understanding how SWI/SNF gene mutations interact to affect cancer progression could lead to new treatment strategies. Methods: We analyzed 7,370 colorectal cancer (CRC) samples with immunohistochemical stains (IHC) and Next-Generation Sequencing (NGS) using a 592-gene panel to examine the association between gene mutations of the SWI/SNF complex ( ARID1A, ARID2, PBRM1, SMARCA4, SMARCB1, SMARCE1, BCL11A, BCL11B, BCL7A, SS18, and SS18L1) and molecular features. Results: The overall mutation rate of the SWI/SNF complex genes was 11.3% ( ARID1A: 7.7% , ARID2: 1.7% , SMARCA4: 1.3% , PBRM1: 1.2% , BCL11A: 1.0% , SMARCB1: 0.5% , BCL11B: 0.5% , SMARCE1: 0.3% , SS18: 0.3% , BCL7A: 0.1% , SS18L1: 0.1%). When compared to tumors with SWI/SNF wild-type genes, those tumors with SWI/SNF gene mutations showed significantly higher rates of microsatellite instability (MSI)-high (40.9% vs 2.4%, P < 0.001), tumor mutational burden (TMB)-high (>= 10mut/MB) (56.8% vs 21.6%, P < 0.001) and PD-L1 positivity (17.9% vs 5.5%, P < 0.001). Tumors with each gene mutant also had strong association with the immune profile (MSI-high, TMB-high, and PD-L1 positivity) (Table). Furthermore, even SWI/SNF gene mutant samples with microsatellite stable status were significantly associated with TMB-high (28.2%, P < 0.001) and PD-L1 positivity (10.0%, P < 0.001). Conclusions: Gene mutations of the SWI/SNF complex exhibit findings that suggest that this subgroup of CRCs may have a higher likelihood of response to PD-1 and PD-L1 targeting monoclonal antibodies. If validated in other data sets, these findings can be used to justify clinical trials with eligibility based upon the presence of mutations within the SWI/SNF complex. [Table: see text]

3130 Background: LNs mets are thought to occur before distant mets. However, lymphatic and distant mets arise from independent subclones of the primary tumor, suggesting that LNs are not essential intermediaries for distant mets. We aimed to comprehensively characterize the molecular profile of LN mets and to explore the differences between LN vs distant mets and primary tumors. Methods: Tumor samples from primary CRCs, LNs, and distant mets were analyzed using NGS (MiSeq on 47 genes, NextSeq on 592 genes), immunohistochemistry. Tumor mutational burden (TMB) was calculated based on somatic nonsynonymous missense mutations, and microsatellite instability (MSI) was evaluated by NGS of known MSI loci. Results: In total, 11871 tumors samples were examined, comprising primaries (N = 5862), distant (N = 5605) and LNs mets (N = 404). The most frequently mutated genes in LNs were TP53 (72%), APC (61%), KRAS (39%), ARID1A (20%), and PIK3CA (12%). LNs showed a higher mean TMB (13 mut/MB) vs distant mets (9, P < .0001). TMB-high (17mut/MB) was more frequent in primaries and LNs vs distant mets (9.5% and 8.8% vs 4.2%, P < .001 and P = .001, respectively), as well as MSI-H (8.8% and 6.9% vs 3.7%, P < .001 and P = .017, respectively). TMB-high is significantly higher in LNs vs distant mets and primaries (P < .0001), independent of MSI-H status. Analyzing distant mets by location, LNs showed higher TMB compared to lung, liver and peritoneum mets (P < .0001). Overall, LNs showed significantly different rates of mutations in APC, KRAS, PI3KCA, KDM6A, and BRIP1 (P < .01 for all comparisons) vs primaries; while presenting a distinct molecular profile compared to distant mets ( TP53 72 vs 67%; KRAS 39 vs 50%; RNF43 7 vs 4%; ATM 5 vs 3%; KDM6A 4 vs 1%; BRCA2 4 vs 2%; MSH6 3 vs 2%; PTCH1 4 vs 1%; BRCA1 2 vs 1%; GNAS 2 vs 5%; P < .05 for all comparisons). Our cohort of 30 paired samples confirmed the molecular heterogeneity between primaries, LNs, and distant mets. Conclusions: This is the largest study to investigate the molecular differences between LNs mets, distant mets and primary tumors in CRC patients. Our data support the hypothesis that lymphatic and distants mets harbor different mutation profiles which suggests that they may arise from distinct subclones.

566 Background: Telomere/telomerase interplay is involved in the regulation of genomic stability and cellular replicative potential. In CRC, shortening of telomeres promotes early steps of carcinogenesis but the prognostic role of telomere length is debated. High circulating levels of telomerase reverse transcriptase (TERT) appear to be a negative prognostic biomarker and polymorphisms in the telomerase RNA component (TERC) are associated with increased cancer risk. We hypothesized that genetic variants in the telomerase/telomere capping complex may predict first-line treatment outcome in mCRC pts. Methods: The impact on outcome of 21 selected SNPs within 13 genes ( TERT, TERC, CLPTM1L, TERF1/2, POT1, ACD, NOP10, NHP2, GAR1, STN1, CTC1) was analyzed through the OncoArray, a custom array manufactured by Illumina, on genomic DNA from blood samples of 451 pts enrolled in two independent randomized trials. TRIBE FOLFIRI/bevacizumab (bev) arm (n=215, mPFS/OS: 9.7/26.2 mo) served as discovery cohort, FIRE-3 FOLFIRI/bev arm (n=107, mPFS/OS: 11.5/31.4 mo) as validation and FOLFIRI/cetuximab (cet) arm (n=129, mPFS/OS: 12.8/49.8 mo) as control. Results: In the discovery cohort, pts carrying any A allele of TERT rs2075786 showed longer mPFS (11.1 vs 9.3 mo, P= .021) and OS (33.5 vs 25 mo, P= .005) compared to the G/G genotype in multivariable analysis. Same results were observed for pts carrying any T allele of CLPTM1L rs401681 compared to the C/C genotype ( P= .022). Any C allele carriers of TERC rs2293607 also showed longer mOS in multivariable analysis ( P= .041) and the A/A genotype of ACD rs6979 was associated with longer mPFS in both uni- and multivariable analysis. Interestingly, the T/T genotype of CLPTM1L rs401681 was associated with shorter mPFS (7.8 vs 13.5 mo) in the cet cohort in uni- ( P= .003) and multivariable analysis ( P= .041). In subgroup analyses based on RAS mutational status several SNPs in additional genes were associated with PFS and/or OS in different cohorts. Conclusions: Our results suggest that SNPs in core telomerase/telomere capping complex genes may have a predictive and prognostic value in mCRC pts receiving targeted first-line treatments.

564 Background: Large metagenomic studies associate disrupted gut microbiome signatures, comprising more prevalently gram-negative bacteria, with CRC carcinogenesis. TLR4, the LPS receptor, has been involved in microbiota-mediated tumorigenesis. High TLR4 expression is associated with poor prognosis in CRC and TLR4 can activate EGFR through ligands epiregulin and amphiregulin. Hence, we hypothesized that genetic variants in the LPS receptor complex and TLR4 expression levels may predict cetuximab efficacy in mCRC pts. Methods: Genomic DNA from blood samples of pts enrolled in the randomized FIRE-3 trial was genotyped through the OncoArray, a custom array manufactured by Illumina. The impact on outcome of 5 functional SNPs within TLR4, MD2 and CD14 was analyzed in 129 pts treated with first-line FOLFIRI/cet (discovery, mPFS/OS: 12.8/49.8 mo) and 107 pts treated with FOLFIRI/bevacizumab (bev) (mPFS/OS: 11.5/31.4 mo). Gene expression levels were measured from 102 tumor samples of pts in the cet arm by HTG EdgeSeq Oncology Biomarker Panel. Results: In the discovery cohort, pts carrying the C/T variant of TLR4 rs4986791 had significantly shorter mPFS and OS compared to the C/C genotype in both uni- and multivariable analysis (PFS: 5.4 vs 13.3 mo, adjusted P[ Padj] = .01; OS: 21.7 vs 40.7 mo, Padj= .03). Conversely, C/C carriers of rs12377632 had a longer mPFS compared to any T in uni- and multivariable analysis (15.8 vs 12.2 mo, Padj< .001). Any A allele carriers of MD2 rs12546552 and any G allele carriers of CD14 rs2569190 showed shorter mPFS in multivariable analyses. These associations were not observed in bev arm. Significant interaction was found between TLR4 rs12377632 and RAS status ( P= .015). High TLR4 expression (log2 > 10.93) was associated with worse mPFS (7.2 vs 11 mo, P= .003) and OS (19.1 vs 29.8 mo, P< .001) in FIRE-3 cet arm. Conclusions: Our results provide the first evidence that polymorphisms in the LPS receptor complex and TLR4 expression levels may have a predictive value in mCRC pts receiving first-line cetuximab-based treatment and, overall, contribute to resistance to anti-EGFRs.

558 Background: The introduction of immunotherapy has significantly improved outcome in various tumors. Immune stimulating proteins exert an anti-tumor effect mainly through enhancing T-cell mediated immune response. Additionally, preliminary data suggest a major role of immune stimulating proteins in modulating angiogenesis. We therefore hypothesize that variations in genes involved in the immune activation pathway may predict outcome in pts with mCRC treated with first-line FOLFIRI/ bevacizumab (bev). Methods: The impact of 4 functional SNPs within the CD40L, Light, OX40L and ICOS genes on outcome was evaluated in 322 pts with mCRC treated with first-line FOLFIRI/bev in two randomized phase III trials. We used TRIBE as a discovery (n = 215) and FIRE-3 as a validation set (n = 107). One hundred twenty-nine pts treated with FOLFIRI/cetuximab (cet) served as a control cohort (FIRE-3). OncoArray, a custom array manufactured by Illumina was used for data extraction. Genomic DNA was extracted from blood. Results: Baseline characteristics: FOLFIRI/bev, discovery set (TRIBE), median PFS/OS/FU 9.7/26.2/48.9 mo; FOLFIRI/bev, validation set (FIRE-3), PFS/OS/FU 11.5/32.4/71.1 mo; FOLFIRI/cet, control set (FIRE-3) PFS/OS/FU 12.8//23.9/70.7 mo. The CD40L rs1126535 SNP showed significant association with OS. Pts in the discovery cohort harboring any T allele and treated with FOLFIRI/bev had a longer median OS compared to C/C carriers (27.9 vs. 20.0 mo) in both univariate (HR 1.83, 95% CI 1.19-2.81, p = 0.005) and multivariate analyses (HR 1.62, 95% CI 1.03-2.56, p = 0.038). Similarly, any T allele carriers in the validation cohort had a significantly longer median OS than those harboring a C/C genotype (40 vs. 19.0 mo) in the multivariate analysis (HR 2.80, 95% CI 1.05-7.50, p = 0.040). However, this association could not be shown in pts receiving FOLFIRI/cet (HR 0.60, 95% CI, 0.18-1.94, p = 0.38). Conclusions: We show for the first time that the CD40L polymorphism rs1126535 might serve as a predictive marker in pts with mCRC treated with FOLFIRI/bev. Targeting CD40L might be promising to further improve treatment against mCRC and to overcome resistance to anti-angiogenic therapy.

63 Background: Signet ring cell carcinoma (SRCC) is a rare variant of adenocarcinoma, accounting for about 10% of gastric cancer (GC) and 1% of colorectal cancer (CRC). SRCC is associated with poor prognosis, however little is known about the underlying molecular characteristics. Herein, we aimed to characterize the molecular features of SRCCs, and to compare the molecular profile of SRCC to adenocarcinoma; further, we assessed the impact of tumor location on the molecular profile of SRCC. Methods: SRCCs were analyzed using NGS (MiSeq on 47 genes, NextSeq on 592 genes), immunohistochemistry, and in-situ hybridization. Tumor mutational burden (TMB) was calculated based on somatic nonsynonymous missense mutations, and microsatellite instability (MSI) was evaluated by NGS of known MSI loci. Chi-square and t-tests were used for comparative analyses. Results: A total of 8,500 CRC and 1,100 GC were screened for SRCC histology. Seventy-six SRCC were identified from the CRC cohort (

594 Background: Th17 cells constitute a subset of T-helper cells, and play a role in immune response to extracellular pathogens in the human intestinal tract. Further, Th17 cells are associated with tumor angiogenesis and enhanced efficacy of 5-FU treatment. We thus investigated associations between the Th17 cell pathway-related SNPs and clinical outcomes in patients with metastatic colorectal cancer (mCRC) treated with conventional chemotherapy. Methods: We analyzed a total of 884 patients with mCRC enrolled in three randomized clinical trials (TRIBE, MAVERICC, and FIRE-3: where patients were treated with FOLFIRI, mFOLFOX6, or FOLFOXIRI combined with bevacizumab or cetuximab as the first-line chemotherapy). Multivariable logistic regression and Cox regression were performed to evaluate the association between candidate SNPs in the Th17 cell pathway and clinical outcomes [tumor response (TR), progression-free survival (PFS), and overall survival (OS)] in each treatment cohort. The meta-analysis approach using the METASOFT software were implemented to quantify the prognostic effect of each SNP using the inverse-variance-weighted effect size, and also to evaluate the heterogeneity across cohorts using the Q statistic. SNPs were coded as additive, dominant, or recessive in the analysis. The Pegasus analysis was also used to identify effects across multiple SNPs and treatment arms. Results: Pathway analysis showed that the Th17 cell pathway was significantly associated with TR ( P = 0.011). There were suggestive associations of IL17F rs763780 with TR (log OR = 0.64, SE = 0.31; P = 0.038), of IL23R rs10889677 with TR (log OR = 0.37, SE = 0.18; P = 0.039), of IRF4 rs872071 with TR (log OR = -0.26, SE = 0.13; P = 0.037), and of IL21 rs2221903 with PFS (log HR = 0.33, SE = 0.15; P = 0.026), although these results were not significant after FDR adjustment. In addition, IL23R rs10889677 had suggestive heterogeneity of effects for PFS across the six cohorts after Cochran’s Q statistic ( P = 0.013). Conclusions: Th17 cell pathway-related SNPs may be predictors for the first-line chemotherapy in mCRC. Upon validation, our findings would provide novel insight for selecting treatment strategies for mCRC.

606 Background: Between 3-5% of CRC patients have BRCA1/2 pathogenic mutations. This study aims to identify associations between BRCA1 and BRCA2 mutations and clinical characteristics in CRC. Methods: A total of 6396 CRC tumor samples were tested with Next-Generation Sequencing (NGS) on a 592-gene panel, pathogenic or presumed pathogenic variants were counted as mutations (mt). Microsatellite instability (MSI) and tumor mutational burden (TMB) were tested by NGS. Statistical correlations were investigated using ANOVA, Chi-square and t-test. Results: Among tumors sampled, 53% derived from male patients and median age was 60 years. BRCA1 mt were detected in 1.1% (n = 72) of tumors, while BRCA2 in 2.8% (n = 179). BRCA1 mt were more frequent in women (W;65%) than men (M;35%) (p = 0.0019) while no relationship with sex was seen for BRCA2 mt (42% F vs. 58% M). No significant associations with age were noticed. Majority of pathogenic mt in BRCA1 (52%; n = 34) and BRCA2 (62%; n = 103) occurred in MSI-High (MSI-H) cases. MSI-H pts had more frameshift mt in both BRCA1/2 than MSS pts. MSS cases had lower rates of BRCA1 and 2 pathogenic mt (44% and 37%, respectively). Right-sided tumors were significantly associated with BRCA1 (p = 0.0056) and BRCA2 (p < 0.0001) mt in MSI-H cases only. BRCA1/2 mt were associated with higher TMB in all CRCs, including MSI-H and MSS cases (p < 0.001). POLE mt (n = 31) were associated with higher BRCA1/2 mt rates (9.6%, 55% respectively). Among MSS cases with POLE wild-type status, BRCA1 (p = 0.0269) and BRCA2 (p = 0.0151) mt were associated with high TMB and combining both BRCA1/2 mutations led to an even higher TMB (3.6%; p = 0.001). Conclusions: This is the first study to show that BRCA1/2 mutations are more frequent in MSI-H, and independently associated with higher TMB, pathogenic POLE mutations, and right-sided tumors in MSI-H CRCs. Given their relationship with TMB, the presence of BRCA1/2 mutations may be potential predictive biomarkers for checkpoint or PARP inhibitors in CRC, a finding that should be prospectively evaluated.

Background Regorafenib and TAS-102 have recently demonstrated statistically significant survival gains in patients with refractory metastatic colorectal cancer (mCRC). Life expectancy ≥12 weeks was an inclusion criterion in registrative trials, and the identification of proper clinical selection tools for the daily use of these drugs in heavily pre-treated patients is needed to improve the cost-benefit ratio. We aimed at building a nomogram able to predict death probability within 12 weeks from the date of assessment of refractory mCRC. Patients and methods Four hundred eleven refractory mCRC patients with ECOG performance status (PS) ≤2 receiving regorafenib, TAS-102 or other treatments were used as developing set. Putative prognostic variables were selected using a random forest model and included in a binary logistic model from which the nomogram was developed. The nomogram was externally validated and its performance was evaluated by examining calibration (how close predictions were to the actual outcome) and discriminative ability (Harrell C index) both on developing (internal validation) and validating (external validation) sets. Results Four variables were selected and included in the nomogram: PS (P

Purpose Tumor screening for Lynch syndrome is recommended in all or most patients with colorectal cancer (CRC). In metastatic CRC, sequencing of RAS/BRAF is necessary to guide clinical management. We hypothesized that a next-generation sequencing (NGS) panel that identifies RAS/BRAF and other actionable mutations could also reliably identify tumors with DNA mismatch repair protein deficiency (MMR-D) on the basis of increased mutational load. Methods We identified all CRCs that underwent genomic mutation profiling with a custom NGS assay (MSK-IMPACT) between March 2014 and July 2015. Tumor mutational load, with exclusion of copy number changes, was determined for each case and compared with MMR status as determined by routine immunohistochemistry. Results Tumors from 224 patients with unique CRC analyzed for MMR status also underwent MSK-IMPACT. Thirteen percent (n = 28) exhibited MMR-D by immunohistochemistry. Using the 341-gene assay, 100% of the 193 tumors with < 20 mutations were MMR-proficient. Of 31 tumors with ≥ 20 mutations, 28 (90%) were MMR-D. The three remaining tumors were easily identified as being distinct from the MMR-D tumors with > 150 mutations each. Each of these tumors harbored the P286R hotspot POLE mutation consistent with the ultramutator phenotype. Among MMR-D tumors, the median number of mutations was 50 (range, 20 to 90) compared with six (range, 0 to 17) in MMR-proficient/POLE wild-type tumors (P < .001). With a mutational load cutoff of ≥ 20 and < 150 for MMR-D detection, sensitivity and specificity were both 1.0 (95% CI, 0.93 to 1.0). Conclusion A cutoff for mutational load can be identified via multigene NGS tumor profiling, which provides a highly accurate means of screening for MMR-D in the same assay that is used for tumor genotyping.

INTRODUCTION Right- and left-sided colorectal cancers (CRCs) differ in clinical and molecular characteristics. Some retrospective analyses suggested that patients with right-sided tumors derive less benefit from anti-epidermal growth factor receptor (EGFR) antibodies; however, molecular selection in those studies was not extensive. PATIENTS AND METHODS Patients with RAS and BRAF wild-type metastatic CRC (mCRC) who were treated with single-agent anti-EGFRs or with cetuximab-irinotecan (if refractory to previous irinotecan) were included in the study. Differences in outcome between patients with right- and left-sided tumors were investigated. RESULTS Of 75 patients, 14 and 61 had right- and left-sided tumors, respectively. None of the right-sided tumors responded according to RECIST, compared with 24 left-sided tumors (overall response rate: 0% vs. 41%; p = .0032), and only 2 patients with right-sided tumors (15%) versus 47 patients with left-sided tumors (80%) achieved disease control (p < .0001). The median duration of progression-free survival was 2.3 and 6.6 months in patients with right-sided and left-sided tumors, respectively (hazard ratio: 3.97; 95% confidence interval: 2.09-7.53; p < .0001). CONCLUSION Patients with right-sided RAS and BRAF wild-type mCRC seemed to derive no benefit from single-agent anti-EGFRs. IMPLICATIONS FOR PRACTICE Right- and left-sided colorectal tumors have peculiar epidemiological and clinicopathological characteristics, distinct gene expression profiles and genetic alterations, and different prognoses. This study assessed the potential predictive impact of primary tumor site with regard to anti-epidermal growth factor receptor (EGFR) monoclonal antibody treatment in patients with RAS and BRAF wild-type metastatic colorectal cancer. The results demonstrated the lack of activity of anti-EGFRs in RAS and BRAF wild-type, right-sided tumors, thus suggesting a potential role for primary tumor location in driving treatment choices.

3543Background: The combination of triple chemotherapy regimens with an anti-EGFR reported promising activity with some safety concerns in phase II trials. MACBETH trial aimed at evaluating the activity and safety of 1st-line mFOLFOXIRI plus cet and at exploring the role of maintenance with cet or bev in unresectable mCRC patients (pts). The study initially enrolled KRAS wt pts and, after an amendment in Oct 2013, only RAS/BRAF wt pts. Methods: Eligibility criteria included measurable, unresectable, RAS/BRAF wt disease (centrally screened), age 18-75 yrs, no prior therapy for advanced disease. Pts were randomized to receive up to 8 cycles of mFOLFOXIRI + cet (cet 500 mg/sqm, iri 130 mg/sqm, oxa 85 mg/sqm, I-LV 200 mg/sqm, 5FU 2400 mg/sqm over 48h q2w), followed by cet (arm A) or bev (arm B) until progression. The trial was a phase II randomized non-comparative study, with primary endpoint 10 months-Progression Free Rate (10m-PFR). Results: Between Nov 2011 and Feb 2015, 323 pts from 21 Italian centers wer...

Although second-line therapy is often considered for advanced gastric cancer patients, the optimal candidates are not well defined.We retrospectively collected baseline parameters, tumour features, and treatment data for 868 advanced gastric cancer patients exposed to multiple treatment lines at 19 Italian centres. Cross-tables and chi-square tests were used to describe categorical features. To predict the impact of clinical variables on progression-free survival and overall survival, Kaplan-Meier and Cox regression analyses were performed.At the start of second-line therapy, median age was 64.8 years (25th-75th percentiles: 55.2-71.9 years). Overall, 43% of patients received single-agent chemotherapy, 47.4% a doublet, and 7.3% a triplet. Median second-line progression-free survival was 2.8 months (25th-75th percentiles: 1.8-5.2 months) and median second-line overall survival was 5.6 months (25th-75th percentiles: 2.9-10.0 months). Multivariate analysis showed that performance status, LDH level, neutrophils/lymphocytes ratio, and progression-free survival in the first-line therapy all impacted on prognosis. Based on these four prognostic factors, a prognostic index was constructed that divided patients into good, intermediate, and poor risk groups; median second-line overall survival for each group was 7.7, 4.5, and 2.0 months, respectively (log-rank p 0.0001).Advanced gastric cancer patients with a favourable ECOG performance status, lower LDH levels, and a lower neutrophils/lymphocytes ratio at the start of second-line therapy seem to have better outcomes, regardless of age and intensity of treatment. A longer progression-free survival in the first-line therapy also had positive prognostic value. Our real-life study might help clinicians to identify the patients who may benefit most from a second-line therapy.

12098Background: Glucocorticoids (GC) have important anti-inflammatory and pro-apoptotic activities. CRC cells produce immunoregulatory GC, a process regulated by nuclear receptor liver receptor ho...

12106Background: PER1 encodes for one of the main negative regulator of the clock genes pathway, which modulates the circadian expression of key target genes at the cellular level. Downregulation o...

4098Background: GEP-NETs are rare malignancies and their molecular characteristics are largely undefined. Here, we explored the underlying biology of GEP-NETs and the differences between gastrointe...

e15578Background: Th17 cells are recognized for their role in immune activation and angiogenesis. Further, Th17 cells are associated with enhanced efficacy of 5-FU treatment, which was the rational...

3611Background: The natural history and prognosis of appendiceal adenocarcinomas (AA) differ from that of adenocarcinomas arising in other large bowel sites. Compared to CRC, AA has more peritoneal...

12107Background: Cetuximab(cet) is an anti-EGFR mAb which enhances the antibody-dependent cellular cytotoxicity (ADCC) by Natural Killer (NK) cells in EGFR+ CRCs. Overexpression of MHC class I anti...

3576Background: The clock machinery comprises a complex network of transcription-translation feedback loops which regulates the circadian expression of target genes involved in key cellular functio...

The treatment scenario of colorectal cancer (CRC) has been evolving in recent years with the introduction of novel targeted agents and new therapeutic strategies for the metastatic disease. An extensive effort has been directed to the identification of predictive biomarkers to aid patients selection and guide therapeutic choices. Pharmacogenomics represents an irreplaceable tool to individualize patients treatment based on germline and tumor acquired somatic genetic variations able to predict drugs response and risk of toxicities. The growing knowledge of CRC molecular characteristics and complex genomic makeup has played a crucial role in identifying predictive pharmacogenomic biomarkers, while supporting the rationale for the development of new drugs and treatment combinations. Clinical validation of promising biomarkers, however, is often an issue. More recently, a deeper understanding of resistance mechanisms and tumor escape dynamics under treatment pressure and the availability of novel technologies are opening new perspectives in this field. This review aims to present an overview of current pharmacogenomic biomarkers and future perspectives of pharmacogenomics in CRC, in an evolving scenario moving from a single drug-gene interactions approach to a more comprehensive genome-wide approach, comprising genomics and epigenetics.

646 Background: DNA hypermethylation in CpG islands (CIMP) has been shown to promote CRC by silencing the expression of tumor suppressor genes. However, global DNA hypomethylation is now considered a common characteristic of CRC. A new family of genes (TET) has been shown to play a crucial role in actively demethylating the DNA. We explored whether genetic polymorphisms in the genes which promote DNA methylation and demethylation are associated with outcome in pts enrolled in TRIBE and FIRE-3 trials. Methods: Genomic DNA from blood samples was genotyped through the OncoArray, a custom array by Illumina including 530K SNP markers. 17 functional SNPs within 11 genes related to methylation (DNMT1/3A/3B, MDM2) and demethylation pathways (TET1/2/3, IDH1/2, MBD4, TDG) were analyzed. Results: A total of 451 pts were included. FOLFIRI/bevacizumab (bev) arms from TRIBE (N = 215, median PFS/OS: 9.7/26.3 months (mts)) and FIRE3 (N = 107, median PFS/OS: 11.6/31.5 mts) served as discovery and validation cohorts respectively. FIRE-3 FOLFIRI/cetuximab arm (N = 129, median OFS/OS: 12.8/49.8 mts) served as control. In the discovery cohort, overall pts carrying TET1 rs3814177 C/C variant showed lower response rate (RR) (42.1% vs 61.8%, p = 0.026) and shorter median PFS (9.5 vs 10.3 mts; HR = 1.54, 95%CI = 1.04-2.27; p = 0.024, univariate analysis) compare to those with any T allele. The effect was more significant among KRAS mutant pts in RR (21.4% vs 60.8%, p = 0.007) and median PFS (8.6 vs 9.5 mts) both in univariate (HR = 2.00, 95%CI = 1.07-3.73; p = 0.020) and multivariable analysis (HR = 2.01, 95%CI = 1.03-3.92 p = 0.041). These findings were validated in overall pts in FIRE-3 cohort in median OS both in univariate (23.7 vs 41.9 mts; HR = 2.08, 95%CI = 1.07-4.05; p = 0.023), and multivariable analysis (HR = 2.36, 95%CI = 1.15-4.81; p = 0.019). No significant association was observed in the control arm. Conclusions: Our results provide the first evidence that TET1 polymorphism may be used as a predictive biomarker in mCRC pts treated with FOLFIRI/bev. Therefore, TET enzymes could serve as promising new treatment targets for mCRC pts.

651 Background: B cells are recognized for their immune activation and lymph-angiogenesis, which attract us to clarify the correlation between B cells and the effect of bevacizumab in cancer patients. We thus investigated whether genetic polymorphisms in Tfh cell/ B cell pathway could predict clinical outcomes in metastatic colorectal cancer (mCRC) patients with bevacizumab-based chemotherapy in first-line treatment. Methods: Genomic DNA was extracted from 451 samples of three independent cohorts: a discovery cohort of 215 patients with bevacizumab plus FOLFIRI in TRIBE (NCT00719797); a validation cohort of 107 patients with bevacizumab plus FOLFIRI in FIRE-3 (NCT00433927); a control cohort of 129 patients with cetuximab plus FOLFIRI in FIRE-3. The OncoArray including 530K single nucleotide polymorphisms (SNPs) from Illumina was used in this study. The correlation between selected SNPs in Tfh cell/ B cell pathway ( CCL19, CCL21, CCR7, CXCL13, CXCR5, and S1PR2) and clinical outcomes were analyzed. Results: In the discovery cohort, patients with any CXCL13 rs355687 C allele had shorter OS than those with T/T variant (22.9 vs. 30.8 months, HR 1.36, 95%CI: 1.00–2.86, P = 0.050), which remained significant in multivariable analysis (HR 1.51, P = 0.013); they also showed a trend of lower response rate (51 vs. 63%, P = 0.093); patients with any CCL21 rs11574915 C allele had shorter OS than those with A/A variant (25.6 vs. 27.9 months) in both univariate (HR 1.43, 95%CI: 1.01–2.02, P = 0.043) and multivariable analyses (HR 1.50, P = 0.030). In the validation cohort, patients with any CXCL13 rs355687 C allele had a trend which showed lower response rate than those with T/T variant (56 vs. 71%, P = 0.099). The findings were not detected in the control cohort. Conclusions: Genetic polymorphisms of CXCL13 and CCL21 may impact the angiogenic environment and patients survivals. They are impacting efficacy of bevacizumab-based chemotherapy in first-line treatment.

645 Background: Tumor microenvironment plays a critical role for CRC development and progression. CAFs are the most dominant cellular components which orchestrate tumor cell proliferation, survival and invasion and have been shown to be associated with prognosis in pts with CRC. We aimed to evaluate whether single-nucleotide polymorphisms (SNPs) in CAFs related genes predict outcome in mCRC pts treated with first-line FOLFIRI/bevacizumab (bev) in TRIBE and FIRE-3 trials. Methods: The OncoArray, a custom array by Illumina including 530K SNP markers, was used to genotype genomic DNA from blood samples and served as our database, from which 15 functional SNPs within 9 genes (OPN, CD44, TCN, CD63, TIMP1, SLIT2, ROBO1, LOXL2, INHBA) related to CAF pathway were analyzed. Results: A total of 451 pts were included. Those treated with FOLFIRI/bev in TRIBE (N = 215, median PFS/OS: 9.7/26.3 months (mts)) and FIRE-3 (N = 107, median PFS/OS: 11.6/31.5 mts) served as discovery and validation cohorts respectively, while FIRE-3 FOLFIRI/cetuximab arm (N = 129, median PFS/OS: 12.8/49.8 mts) was used as the control. In overall pts in discovery set, OPN (osteopontin) rs11728697 any C variants showed longer PFS (median 10.8 vs 9.0 mts; HR = 0.68, 95%CI = 0.49-0.95; p = 0.025) and OS (median 30.6 vs 23.8 mts; HR = 0.69, 95%CI = 0.50-0.95; p = 0.025) than the homozygous wild type genotype (T/T) in multivariable analysis. These data were validated in the FIRE-3 cohort in OS both in univariate (median 49.3 vs 23.7 mts; HR = 0.40, 95%CI = 0.21-0.77; p < 0.001) and in multivariable analyses (HR = 0.31, 95%CI = 0.16-0.60; p < 0.001). However, the favorable impact on outcome was not shown among C allele carriers treated with FOLFIRI/cetuximab in the control arm. Conclusions: Our findings confirm that CAFs play a crucial role in mCRC and suggest that OPN polymorphisms could serve as predictive biomarkers for mCRC pts treated with first-line FOLFIRI/bev.

e15042 Background: Chemotherapy (CT) associated with angiogenesis inhibitor bevacizumab (bev) is a standard first line treatment in metastatic colorectal cancer (mCRC). Among adverse events related to bev, anal canal ulcers have never been extensively described. The aim of our study was to evaluate symptomatic anal ulcers development and its impact on clinical practice and proctologic management. Methods: We retrospectively reviewed clinical records of 601 consecutive mCRC patients (pts) undergoing first line CT from January 2010 to December 2015 at our Institute. Pts’ characteristics, treatments and proctology reports were analyzed. Anal pain was graded in a semi-quantitative scale. Vascularization of the anal canal was compared with contrast MRI in 5 pts treated with bev and 5 healthy controls. Non parametric statistics was performed. Results: Fifty out of 601 pts reported perianal complaints during treatment and underwent proctologic evaluation. Among those, 16 were found to have an anal ulcer. Symptomatic anal ulcers occurred only in pts receiving CT plus bev (16 out of 384, 4.2%) while none was observed in 217 pts treated with other regimens (p = 0.009). The peak incidence was 4-8 weeks after treatment start. Vascularization of anal canal was significantly lower in pts treated with bev (p = 0.03). Hypertension and hemorrhoids were associated with a lower risk of anal ulcer occurrence (p = 0.009 and p = 0.036) while a previous neoadjuvant pelvic CT-RT was not associated with a higher risk. Pain intensity was severe. All attempts at symptomatic treatment with topical and sistemic drugs only led to transient benefit. Pts with symptomatic anal ulcers suspended their first line therapy significantly earlier than other pts (HR = 0.30, 95%CI 0.07-0.80). Conclusions: Symptomatic anal ulcer in mCRC pts treated with CT plus bev is a common adverse event in our series. It can compromise oncologic treatment and pts’ quality of life. We recommend an early proctologic evaluation in case of anal symptoms with the aim to prevent and timely manage such complication in pts treated with bev. A more extensive analysis is ongoing considering subsequent treatment lines and other angiogenesis inhibitors.

TPS3627Background: Based on available data, a reasonable upfront treatment for elderly mCRC pts is a fluoropyrimidine-based monotherapy plus bevacizumab, irrespectively of RAS status. Up today, dat...

3540Background: The impact of gender differences in terms of OS and of association with clinical and molecular characteristics remains unclear. Previous analyses were limited by small sample size, ...

11589Background: Peptidyl-Prolyl Cis/Trans Isomerase NIMA-Interacting 1 (Pin1) is associated with cell cycle progression, cellular proliferation and neoplastic transformation. Pin1 is also involved...

733 Background: Few data are available on outcome of clinical practice unselected patients (pts) with mCRC. Methods: We retrospectively collected data of pts with mCRC followed at our Institution from January 2010 to December 2013 evaluating the clinical characteristics, treatments and survival outcomes. Results: A total of 584 pts were evaluated, 461 were followed at our Center while 123 were seen for a second opinion. Median age was 66 ys (25-94), 59% were male, 63% had an ECOG PS = 0 while 11% ≥2. 33% had right colon primary, 68% synchronous metastatic (mts) disease and 70% a single mts site. 81% underwent surgery on primary and 41% on metastases. 51% were RAS mutated (mut) and 5% BRAF mut. 57 pts didn't receive any systemic treatment, 33 due to frail clinical conditions and 24 due to radical surgical approach (R0). Among 404 treated pts, 239 received all 3 available cytotoxic agents (oxaliplatin, irinotecan, 5FU), 324 bevacizumab and 98 anti-EGFR; 153 (38%) were enrolled in clinical trials. Median overall survival (OS) was 27.6 months (mo) for the entire mCRC population, 3.7 mo for untreated frail pts, 28.7 mo for treated pts while it is still not reached for untreated R0 pts. OS was significantly longer for pts receiving first line combination therapy (29 vs 17 mo, p < 0.01) while a poor prognosis was confirmed for BRAF mut pts (p < 0.001). In a multivariate analysis age < 70, PS 0 and R0 surgery on mts disease showed a positive prognostic impact on OS while a right site of primary was a negative predictor of outcome. At logistic regression older age, low PS and peritoneal disease negatively affected the possibility to receive all 3 active drugs. Conclusions: Despite being an unselected population our outcomes are comparable with results of clinical trials in the corresponding period. We feel that such positive evidence derives from a personalization of treatment and a multidisciplinary approach to mts disease.

673 Background: Induction chemotherapy (CT) is a promising option in LARC. FOLFOXIRI + BV is an effective treatment in metastatic colorectal cancer. Methods: Patients (pts) with LARC at < 12 cm from the anal verge, N+ or cT4 or high risk cT3 (MRI criteria) underwent 6 cycles of FOLFOXIRI + BV followed by CRT (50.4 Gy + 5FU 225 mg/m2/day or capecitabine 800 mg/m2/bid 5 days/week + BV 5 mg/kg on days 1, 15, 28). Surgery was planned 8 weeks after CRT. Primary endpoint is 2-year disease-free survival (DFS). Results: From April 2012 to April 2015 48 pts were enrolled. At now, 46 pts completed induction CT, 43 completed CRT and 39 underwent surgery (5 pts ongoing). Pts characteristics were: median age, 53 years (range 30-74); cT2/cT3/cT4, 4%/60%/36%; cN0/N+, 4%/96%. Main grade (G) 3/4 toxicities during induction were neutropenia (42%), febrile neutropenia (4.2%), diarrhea (12.5. Two pts did not complete induction: one died due to bowel perforation and sepsis and one discontinued CT after acute kidney injury. Response rate (RR) after induction was 77%. Forty-five pts started CRT (1 pts underwent surgery after induction because of SAE). After the first 13 patients, protocol was amended and schedule of capecitabine modified due to an excessive rate of G3 toxicities during CRT: hand-foot syndrome (23%), proctalgia (23%), proctitis (23%) and diarrhea (15%). After amendment all pts completed CRT with acceptable toxicity: G3 proctitis 6.7% and proctalgia 3.3%. One pts died due to early progressive disease (PD) after CRT. RR after CRT was 88%. Surgery was low anterior resection 87%, abdomino-perineal resection 8%, other 5%. Radical resection was achieved in 95% of pts. Early post-surgical complication rate was 31%. Pathologic complete response was reached in 33% and pathological downstaging in 44% of pts. At a median follow up of 17 months, 8 pts had PD and estimated 2y-DFS is 72%. Conclusions: Induction CT with FOLFOXIRI + BV is feasible and highly active. A protocol amendment was needed due to toxicities during CRT. The rate of early post-surgical complications is not negligible. A longer follow up is needed for DFS. Clinical trial information: 2011-003340-45.