Search

177 Background: LRP1B is a member of the low-density lipoprotein receptor family and a tumor suppressor found to be downregulated in colon cancer (CRC). GRM3 is a receptor of glutamate, an amino acid and neurotransmitter. Inhibition of GRM3 reduces CRC cell growth. Recent data from CALGB/SWOG 80405 suggests that mutations (MT) of either LRP1B or GRM3 are associated with better and worse overall survival (OS) in patients treated with bevacizumab (Bev), respectively. We investigate the association of LRP1B or GRM3 mRNA levels with outcomes. Methods: A total of 13,780 CRC tumors (male 7,497, female 6,283) underwent comprehensive molecular profiling (Caris Life Sciences). Analyses included next-generation sequencing of DNA (592 genes, NextSeq, WES, NovaSEQ) and RNA (NovaSeq). Significance with multiple correction was indicated with q, otherwise p value. Gene Set Enrichment Analyses (GSEA) were performed (significance p

598 Background: PDAC of the H and B/T differ in embryonic origin, cell composition, blood supply, lymphatic and venous drainage, and innervation. H tumors tend cause symptoms earlier and to present at earlier stages compared to B/T cancers. The impact of PDAC tumor location on patient presentation and survival has been shown in large national data-based analyses, although with conflicting results. We aimed to compare the molecular and tumor immune microenvironment (TIME) profiles of PDAC of the H vs. B/T. Methods: A total of 3499 PDAC samples were analyzed via next-generation sequencing (NGS) of RNA (whole transcriptome, NovaSeq), DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and immunohistochemistry (IHC, Caris Life Sciences, Phoenix, AZ). RNA deconvolution was performed using QuantiSeq (Finotello 2019, Genome Medicine) to quantify the immune cell infiltration. Pathway gene enrichment analyses were done using Gene Set Enrichment Analysis (GSEA, Subramaniam 2015, PNAS). Significance was determined as p values adjusted for multiple correction (q) of < 0.05. Results: Anatomic subsites of PDAC tumors were grouped by primary tumor sites into H (N = 2058) or B/T (N = 1384). There were significantly more metastatic tumors profiled from H vs. B/T (57% vs. 44%, p < 0.001). KRAS mutations (93.8% vs. 90.2%), genomic loss of heterozygosity (12.7% vs. 9.1%), and several copy number alterations ( FGF3, FGF4, FGF19, CCND1, ZNF703, FLT4, MUTYH, TNFRS14) trended higher in B/T when compared to H (p < 0.05 but q > 0.05). GNAS mutations (2.2% vs. 0.7%) trended higher in H vs. B/T (p < 0.05). No significant difference in immuno-oncology (IO) markers (TMB, PD-L1, MSI-H) were observed, but expression analysis of IO-related genes showed significantly higher expression of CTLA4 and PDCD1 in H (q < 0.05, fold change 1.2 and 1.3) and IDO1 and PDCD1LG2 expression trended higher in B/T (p < 0.05, fold change 0.95). When comparing median cell abundance values as part of TIME analysis, H had increased immune infiltration of B cells (0.045 vs. 0.043), M2 macrophages (0.035 vs. 0.032), neutrophils (0.056 vs. 0.052), NK cells (0.027 vs. 0.026), CD8+ T cells (% > 0: 48.2% vs. 43.2%), while B/T had increased infiltration of M1 macrophages (0.035 vs. 0.032) (all q < 0.05). GSEA showed enrichment of CTLA4 (normalized enrichment score (NES) 1.6, false discovery rate (FDR) 0.19) and primary immunodeficiency pathway enrichment (NES 1.7, FDR 0.11) in H. Conclusions: To our knowledge, this is one of the largest cohort of PDAC tumors subjected to broad molecular profiling. Differences in IO-related gene expression and TIME cell distribution suggest that response to IO therapies may differ in PDAC arising from H vs B/T. Subtle differences in the genomic profliles of H vs. B/T tumors were also observed.

Microsatellite instability (MSI) is a key secondary effect of a defective DNA mismatch repair mechanism resulting in incorrectly replicated microsatellites in many malignant tumors. Historically, MSI detection has been performed by fragment analysis (FA) on a panel of representative genomic markers. More recently, using next‐generation sequencing (NGS) to analyze thousands of microsatellites has been shown to improve the robustness and sensitivity of MSI detection. However, NGS‐based MSI tests can be prone to population biases if NGS results are aligned to a reference genome instead of patient‐matched normal tissue. We observed an increased rate of false positives in patients of African ancestry with an NGS‐based diagnostic for MSI status utilizing 7317 microsatellite loci. We then minimized this bias by training a modified calling model that utilized 2011 microsatellite loci. With these adjustments 100% (95% CI: 89.1% to 100%) of African ancestry patients in an independent validation test were called correctly using the updated model. This poses not only a significant technical improvement but also has an important clinical impact on directing immune checkpoint inhibitor therapy., This work presents observations of a significant population bias in an NGS‐based MSI test and minimizes this bias using a computational‐based approach. The research poses not only a significant technical improvement to the NGS‐based MSI diagnostic, but has an important clinical impact on directing immune‐check point inhibitor therapy for late‐stage cancer patients of non‐European descent.

Introduction: Oncogenic gene fusions are frequently identified in different cancers; however, the exact incidence of oncogenic fusions in melanoma is not well defined. Several targeted therapies are approved and considered the standard of care for patients whose solid or hematologic tumors harbor particular gene fusions, but their role as targets in melanoma also remains undelineated. We sought to determine the prevalence of oncogenic fusions in metastatic or locally advanced melanoma. Methods: We retrospectively analyzed data from formalin-fixed paraffin-embedded (FFPE) tumor samples sent to a commercial CLIA-certified laboratory (Caris Life Sciences) from February 2019 to July 2020. Samples were profiled by next-generation sequencing of a 592-gene DNA panel, whole transcriptome sequencing, and immunohistochemistry (IHC). Results: Melanoma specimens were analyzed from 1,255 subjects, of whom 478 (38.1%) were female and 777 (61.9%) were male, with a median age of 67 years. Of these specimens, 780 (63.1%) were from metastatic sites. We identified 33 (2.6%) cases with in-frame oncogenic fusions (14 novel) including 21 BRAF fusions and 4 RAF1 fusions, as well as fusions involving PRKCA (n=4), TERT (n=2), AXL (n=1), and FGFR3 (n=1). PD-L1 expression by IHC (SP142 or 28-8 antibody) was detected in 512 (42.5%) specimens, including 10 (32.3%) of the fusion-positive tumors, while 572 (47.4%) specimens were TMB-High (≥10 mutations/Mb), including 11 (33.3%) of the fusion-positive tumors, suggesting these patients may respond to immunotherapy. We identified 796 (63.4%) cases with RAS/RAF pathogenic or likely pathogenic mutations, including 373 (30.0%) BRAF p.V600X mutations. With the exception of a single BRAF p.S467L (Class 3, “kinase-dead”) mutation, the absence of BRAF mutations in BRAF fusion-positive tumors suggests these fusions are oncogenic drivers. However, tumors harboring PRKCA and TERT fusions were each detected with at least one MAPK pathway co-alteration (NRAS, NF1, or BRAF p.V600E mutation). Fusion transcripts with unknown pathogenicity were also detected in 668 (53.2%) cases. RNA expression analysis of key molecular pathways (including Wnt/β-catenin, PI3K/AKT/MTOR, DNA repair, INFG, and JAK/STAT) showed a high degree of variability among fusion-positive tumors and other mutational subgroups, which may reflect the heterogeneity common to melanoma, prior treatments or development of various resistance mechanisms. Conclusions: Oncogenic gene fusions are rare in melanoma when compared with other genetic variants. We identified potentially actionable fusions as well as co-alterations in fusion-positive cases, with notably mutual exclusivity of BRAF fusions and p.V600X mutations. The data suggest that targetable variants, including oncogenic fusions, may be identified in melanoma with comprehensive tumor profiling and provide patients with personalized treatment or clinical trial options. Citation Format: Sourat Darabi, Andrew Elliott, David R. Braxton, Jia Zeng, Kelsey Poorman, Jeffrey Swensen, Geoffrey T. Gibney, Justin C. Moser, Thuy Phung, Michael B. Atkins, Gino K. In, Wolfgang Michael Korn, Burton L. Eisenberg, Michael J. Demeure. Whole transcriptome sequencing reveals oncogenic fusions in melanoma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2221.

e20533 Background: Nuclear protein transport is essential in guiding the organized traffic of important proteins and RNAs between the nucleus and cytoplasm of the cell. Export of proteins from the nucleus is exclusively regulated by Exportin 1(XPO1). In cancer, XPO1 is universally hyperactive and can promote the export of important tumor suppressors to the cytoplasm, leading to their functional inactivation. XPO1 is aberrantly over expressed in NSCLC and this over expression has been linked to poor overall survival. The underlying mechanisms of XPO1 over expression are not known. Currently there are no studies evaluating the impact of XPO1 mutations on NSCLC incidence and therapy resistance. Additionally, there are no studies that examined the XPO1 related pathways in NSCLC harboring co-alterations with other driver mutations such as EGFR or ALK. Methods: Tumor samples were analyzed using next-generation sequencing (NextSeq, 592 Genes), IHC, and whole transcriptome sequencing (WTS ,NovaSeq) (Caris Life Sciences, Phoenix, AZ). PD-L1 expression was tested by IHC using 22c3 (Dako) and TPS scores were reported (cutoff > 1%). TMB was measured by totaling somatic mutations (TMB-high cut-off ³ 10 mutations per Megabase). Gene fusions were detected by RNA sequencing using either the Archer FusionPlex panel or WTS. Survival was extracted from insurance claims data and calculated from time of tissue collection to last contact using Kaplan-Meier estimate. Statistical significance was determined using chi-square and Wilcoxon rank sum test and adjusted for multiple comparisons. Results: Among 18,218 NSCLC tumors sequenced, 26 harbored XPO1 mutations. XPO1 mutant tumors were more likely to be TMB High(79% vs. 52%, p = 0.007) and less likely to have high PDL1(32% vs. 68%, p = 0.03). KRAS mutations were seen in 19%(n = 5), EGFR mutation were rare (n = 2), and no targetable fusions were seen. Among the 17,449 NSCLC tumors with clinical data, there were 24 XPO1 mutant tumors with no histology imbalance observed in mutant vs. wild-type(WT). Comparison of survival in the NSCLC group between XPO1 mutant and WT showed a negative association with a hazard ratio(HR) of 1.932 (95% CI: 1.144- 3.264 p = 0.012). Comparing the survival within the subgroup with confirmed adenocarcinoma histology (9973 XPO1 WT and 14 XPO1 mutant) showed a similar negative correlation in survival with a HR of 2.156 (95% CI: 1.027- 4.525 P = 0.037). Conclusions: Presence of XPO1 pathogenic mutations was associated with a poor survival in both the entire NSCLC cohort and the adenocarcinoma subgroup. Further studies of this negative association at the molecular level along with effect of other co-existing mutations can result in development of novel treatment strategies.

1034 Background: The Ras-MAPK pathway is a known driver of tumorigenesis and therapeutic target in a variety of cancers. Alterations in this pathway have been linked to decreased tumor immunogenicity. However, molecular alterations in the Ras-MAPK are rare in breast cancer (BC) and their clinical implications remain unclear. As mutational status does not accurately correlate with transcriptional activity, a MAPK pathway activity score (MPAS, Wagle et al., 2018, npj Precision Medicine) is indicative of MAPK activation and correlates with response to MEK (MEKi) or BRAF inhibition (BRAFi). Our goal was to determine the frequency of molecular alterations in the Ras-MAPK and correlate to MAPK pathway activation in MBC. Methods: A total of 6464 BC samples underwent comprehensive molecular profiling at Caris Life Sciences. Analyses included next generation sequencing of DNA (592 Gene Panel, NextSeq; whole exome sequencing, NovaSEQ), RNA (NovaSeq, whole transcriptome sequencing, WTS) and IHC. MPAS and immune cell fraction (ICF, Quantiseq) were assessed by mRNA analysis. Wilcoxon, Fisher’s exact, or Dunnett’s test was used. All results shown were statistically significant (p < 0.05). Results: The predominant alteration of RAS genes was mutation followed by amplification, no fusions were detected (Table). Only 0.17% of all tumors harbor KRAS G12c mutations. The highest MPAS scores were found in KRAS mutants (mut), HRAS mut (Q61, G1213), BRAF V600 (class 1) mut and NRAS Q61 mut (Table) and therefore used to define Genomic MAPK Activated Tumors (GMAT). GMAT compared to wild type (WT) had significantly higher PD-L1 expression, TMB and MSI/dMMR. GMAT had less B cells (3.4% vs 4.4%), more M1 Macrophages (4.4% vs 3.4%) and neutrophils (5.5% vs 2.7%) regardless of HR status but less NK cells (2.3% s 3.0%), MSDCs (0.9% vs 3.0%) only in HR- tumors with respect to WT. GMAT tumors showed more frequent mutation rate (mr) of PIK3CA (HR+: 57.3% vs 40%; HR-: 41.9% vs 17.9%). HR+ tumors had a higher mr of MSH3 (11.8% vs 0.6%) while HR- tumors had higher mr of PIK3R1 (9.6% vs 3.8%), RhoA (5.3% vs 0.5%), DNA repair genes (TERT, 18.2% vs 1.0%; ARID1A, 18.2% vs 5.9%; PRKDC, 3.9% vs 0) and lower TP53 mr (54.5% vs 85.8%) compared to WT. Conclusions: Our study demonstrates that RAS, BRAF and MEK1 mutations are associated with MAPK pathway activation indicative of benefit from MEKi or BRAFi. GMAT warrant further investigation for combinations targeting the RAS-MAPK pathway and immune checkpoint inhibitors.[Table: see text]

8508 Background: The dominant expression of four lineage-defining transcription factors ( ASCL1, NEUROD1, YAP1, or POU2F3) has enabled the classification of small cell lung cancer (SCLC) into four subtypes (SCLC-A/N/Y/P, respectively). Emerging evidence suggests that YAP1 expression is associated with a T-cell inflamed phenotype, and SCLC has significant intra-tumor heterogeneity mediated by MYC-driven activation of NOTCH signaling. We performed a large-scale analysis of real-world SCLC patient samples to examine the expression of clinically relevant biomarkers across SCLC subtypes. Methods: Comprehensive molecular profiling of 437 small cell lung neuroendocrine tumors (including 7.3% high-grade neuroendocrine lung carcinomas) was performed using next-generation DNA sequencing (592-gene panel), RNA sequencing (whole transcriptome), and immunohistochemistry at Caris Life Sciences (Phoenix, AZ). Tumors were stratified into 5 subgroups (SCLC-A/N/Y/P and -mixed) based on the relative expression of the four transcription factors. RNA expression of key genes and previously validated immune signatures (T-cell inflamed, NK cell, and STING pathway signatures) were evaluated across subgroups. Significance was tested by Chi-square, Fisher’s exact test, or Mann-Whitney U test. Results: Median age of the study cohort was 66 years (IQR: 59-72) and 50.6% of patients were female. The majority (67.3%) of samples were derived from metastatic sites. Stratification of tumors by expression resulted in 35.7% SCLC-A, 17.6% SCLC-N, 21.1% SCLC-Y, 6.4% SCLC-P, and 19.2% SCLC-mixed samples. Compared to tumors from metastatic sites, YAP1 expression was significantly increased (p < 0.001) in primary tumors. Amongst the 14 tumors obtained from the CNS, SCLC-N (36%, n = 5) was the most common subtype identified. dMMR/MSI-high (negative MMR protein expression/ ≥46 altered loci per tumor) was rare overall (0.5%, n = 2); TMB (median of 9-10 mut/Mb) was similar between the SCLC subtypes. SCLC-Y was associated with the highest expression of T-cell inflamed, NK cell and STING pathway signatures (p < 0.0001 each). MYC and NOTCH gene expression ( NOTCH1/2/3/4) strongly correlated with YAP1 expression. Analysis of co-mutations revealed that EGFR-sensitizing mutations (L858R and Exon 19 deletions) were recurrent (5.2%, n = 4) in SCLC-N tumors. The expression of SNF11, SSTR2, and MYC varied significantly among SCLC subtypes (p < 0.001 each), with the highest median expression of SNF11 and SSTR2 observed in SCLC-N, while MYC expression was highest in SCLC-P. Conclusions: Our analysis represents the largest real-world dataset of human SCLC tumors profiled by whole transcriptomic sequencing. The differential expression of immune genes and predictive biomarkers across SCLC subtypes may inform therapeutic vulnerabilities for rational and personalized treatment approaches in SCLC.

4021 Background: Immunotherapy is considered ineffective in the majority of patients with pancreatic ductal adenocarcinoma (PDAC), a consequence of a highly immunosuppressive tumor microenvironment (TME). However, treatment induced inhibition of CXC chemokine receptor 4 (CXCR4) and programmed cell death protein-1 (PD-1) in the COMBAT trial caused T cell infiltration and tumor regression in a subset of PDAC patients. Elucidating a phenotype that predicts response is clinically relevant. We performed a comprehensive molecular landscape study in PDAC evaluating CXCR4 RNA expression. Methods: 3,647 PDAC specimens were centrally analysed. NextGen DNA sequencing (NextSeq, 592 gene panel or NovaSeq, whole-exome sequencing), whole-transcriptome RNA sequencing (NovaSeq) and immunohistochemistry (Caris Life Sciences, Phoenix, AZ) were performed. Gene expression is reported as TPM (Transcripts per million). Pathway gene enrichment analyses were done using GSEA (Subramaniam 2015, PNAS). Immune cell fraction was calculated by QuantiSeq (Finotello 2019, Genome Medicine). The cohort was stratified in quartiles according to CXCR4 RNA expression status. Results: Overall, CXCR4 expression was higher in primary tumors compared to distant metastasis (38 vs. 28 TPM, p < 0.0001). CXCR4-high (top quartile: > 59 TPMs), when compared to CXCR4-low (bottom quartile: < 17 TPM) PDACs, were characterized by a high prevalence of mutations in signal transduction pathway genes (e.g. GNAS: 3.6 vs. 0.5%), an increased infiltration of immune cells (e.g. CD8+ T cells, M1 macrophages), and a higher expression of HLA-DRA and HLA-E (all p < 0.0001). We detected an upregulation of CXCL9, CXCL10, CXCL12, CCL5, IDO1 and LAG3 in CXCR4-high compared to CXCR4-low tumors. In contrast, lower PD-L1 expression (17.4 vs. 13.1%, p = 0.02), genomic loss of heterozygosity (17.4 vs. 10.8%), and a lower frequency of gene amplifications in ERBB2 (2.1 vs. 0.1%), TNFRSF14 (2.0 vs. 0.1%), and TP53 (82 vs. 73%, all p < 0.0001) were observed. Moreover, CXCR4-high expression was associated with a better survival (HR: 1.417, 95% CI [1.168-1.72], p < 0.001). Conclusions: This is the first study comprehensively investigating the molecular landscape of PDACs according to CXCR4 RNA expression. High CXCR4 expression is associated with an improved survival and a pro-inflammatory phenotype that may identify a subset of tumors with greater responsiveness to immunotherapeutic approaches.

1091 Background: Tumor mutational burden (TMB) has emerged as an imperfect biomarker of immune checkpoint inhibition (ICI) outcomes in solid tumors. Despite the approval for pembrolizumab in all TMB-high (TMB-H) solid tumors, the optimal clinical approach to TMB-H or hypermutated advanced/metastatic breast cancer (MBC) is unknown with sparse prospective data. We hypothesize that TMB-H MBC will have unique genomic alterations compared to TMB-low (TMB-L) breast cancer that could inform novel therapeutic approaches. Methods: Tumor samples (N = 5621) obtained from patients with MBC were analyzed by next-generation sequencing (NGS) of DNA (592-gene panel or whole exome sequencing) and RNA (whole transcriptome sequencing) at Caris Life Sciences (Phoenix, AZ). TMB was calculated based on recommendations from the Friends of Cancer Research TMB Harmonization Project (Merino et al., 2020), with the TMB-H threshold set to ≥ 10 muts/Mb. IHC was performed for PD-L1 (Ventana SP142 ≥1% immune cells). Deficient mismatch repair (dMMR)/high microsatellite instability (MSI-H) was tested by IHC and NGS, respectively. Results: TMB-H was identified in 8.2% (n = 461) of MBC samples, with similar frequencies observed across molecular subtypes (7.8-8.6%, p = 0.85): HR+/HER2- (n = 3087) 7.8%, HR+/HER2+ (n = 266) 8.3%, HR-/HER2+ (n = 179) 7.8%, TNBC (n = 1476) 8.6%. The frequency of TMB-H was significantly increased in lobular (16%) versus ductal (5%) MBC (p < 0.01). TMB-H samples were enriched in genitourinary (42%), soft tissue (20%), and gastrointestinal non-liver (16%) biopsy specimens. Compared to TMB-L tumors, TMB-H tumors exhibited significantly higher mutation rates for TP53 (60 v 52%), PIK3CA (55 vs 31%), ARID1A (34 vs 11%), CDH1 (27 vs 11%), NF1 (22 vs 9%), RB1 (14 vs 5%), KMT2C (12 vs 7%), PTEN (12 vs 7%), ERBB2 (7 vs 2.9%), and PALB2 (3.3 vs 1%) genes (p < 0.05 each). Copy number alteration and fusion rates did not differ between TMB-H and TMB-L breast cancers. PI3K/AKT/MTOR, TP53, Histone/Chromatin remodeling, DNA damage repair (DDR), RAS, and cell cycle pathway alterations were detected in > 25% TMB-H MBCs (p < 0.05 each). dMMR/MSI-High (7.2 vs 0.3%, p < 0.01) and PD-L1 positivity (36 vs 28%, p < 0.05) frequencies were significantly increased in TMB-H tumors. DNA signature analyses including APOBEC and homologous recombination repair deficiency, as well as gene expression profiling to assess immune-related signatures and tumor microenvironment are underway. Conclusions: TMB-H breast cancers contain a unique genomic profile enriched with targetable mutations such as PIK3CA, ARID1A, NF1, PTEN, ERBB2, and PALB2. Concurrent predictive biomarkers of response to immune checkpoint inhibition such as MSI-H and PDL-1 positivity are also more prevalent in TMB-H MBC. These findings suggest novel combination strategies within TMB-H MBC could be explored.

e15530 Background: The TP53 gene product is a multifunctional protein and prototypical tumor suppressor. Certain missense mutations occurring in hotspots in the TP53 gene may confer gain of function (GOF) oncogenic properties. A recent mouse model (Kadosh et. al., Nature, 2020) demonstrated that secreted factors in the gut microbiome permit GOF p53 to be oncogenic in lower GI tract cancers (LT). By contrast, the GOF p53 mutants in the upper tract cancers (UT) retained tumor suppressor functions. We aimed to provide evidence of the plasticity of TP53 GOF mutants in human GI cancers. We hypothesized that UT should have a lower rate of TP53 GOF than the LT, and UT cancers with putative GOF TP53 mutations should display a higher degree of co-occurring oncogenic alterations such as dysregulation of CCND1, MYCS, or WNT. We also explored outcome differences for TP53 GOF variants. Methods: Tumors of UT (stomach, esophagus, small intestine) and LT (colorectum, anus) were tested at Caris Life Sciences (Phoenix, AZ) by NGS (NextSeq, 592 or NovoSeq, whole exome). MSI/MMRP status was determined by IHC and NGS. Real-world overall survival (OS) was obtained from payor claims data and Kaplan-Meier estimates were calculated. P values adjusted for multiple correction (q) of < 0.05 was considered significant. Six p53 variants were classified as GOF (p.R175, p.G245, p.R248, p.R249, p.R273, and p.R282). Results: Of the 5311 UT and 14810 LT tumors sequenced, 4799 harbored a TP53 GOF variant. UT had TP53 GOF in 22.9% of cases compared to 27.0% of LT (q < 0.05). 67 genes were enriched for amplification (q < 0.05) in the UT TP53 GOF cancers compared to the LT TP53 GOF, including cell cycle regulators( CCND1, CCNE1, CDK6, CDK12), and oncogenes ( KRAS, ERBB2, EGFR). LT TP53 GOF cancers were enriched for eight gene amplifications, including CDX2, FLT3, and SRC. Differences in SNV/InDel patterns were not remarkable. OS analysis revealed TP53 GOF in UT had shorter OS when compared to TP53 wild-type (WT) (Median OS: 418 vs 515 days; HR: 0.864; 95%CI [0.772-0.966]; p = 0.011), but not when MSI/MMRP status was considered separately. LT TP53 GOF did not show OS difference in aggregate, however, MSI-High TP53 GOF cancers showed a shorter OS (Median OS: 762 vs 1479 days; HR: 1.522; [1.068 - 2.171]; p = 0.019). Sub-analysis of LT MSI-High cancers treated by immune checkpoint blockade showed a non-significant trend toward shorter OS of TP53 GOF versus TP53 wild type cancers (HR: 2.55; [0.90-7.19], p: 0.067), but no such trend was seen in UT (HR: 1.434; [0.304-6.758], p:0.646). Conclusions: Our study is the first to compare putative TP53 GOF variants across UT and LT GI adenocarcinoma. TP53 GOF variants are associated with decreased OS in both UT cancers and MSI-High LT cancers. Our findings support the functional plasticity of TP53 GOF variants in human GI cancers. Therefore, the purported effects of secreted microbiome factors on TP53 GOF variants cannot be ruled out.

5580 Background: HER2 is an emerging prognostic and therapeutic target in uterine serous carcinoma (USC). Testing algorithms and platforms in breast and gastric cancers are well studied and validated, but optimal HER2 testing in uterine cancer is not yet established. We aimed to assess the concordance of chromogenic in situ hybridization (CISH), immunohistochemistry (IHC), and next generation sequencing (NGS) platforms to aid in the development of USC specific testing guidelines. We also evaluated the rate of downstream mutations that may affect response to HER2 directed therapy. Methods: A total of 2,192 USC tumors were analyzed using NGS (NextSeq, 592 Genes and WES, NovaSEQ), a subset of 1,423 tumors were also tested by IHC and CISH (Caris Life Sciences, Phoenix, AZ). HER2 positivity through IHC (4B5, Ventana) and CISH (INFORM DUAL HER2 ISH Assay, Ventana) was determined based on 2007 and 2018 ASCO/CAP HER2 breast cancer guidelines. PD-L1 expression was tested by IHC using SP142 (Spring Biosciences) (positive cut-off >1%). Microsatellite instability (MSI) was tested by fragment analysis (FA), IHC and NGS. Tumor mutational burden (TMB) was measured by totaling somatic mutations per tumor (TMB-high cut-off > 10 mutations per Mb). Statistical significance was determined using chi-square. Results: Rates of HER2 positivity were comparable using the 2018 and 2007 breast cancer guidelines (19.5% vs 17.5%; p=0.25). Based on 2018 guidelines, the concordance between IHC and CISH was 98.9%. Specifically, 229/1423 patients (16%) were IHC+/CISH+, 5 patients (0.4%) were IHC+/CISH- and 11 patients (0.8%) were IHC-/CISH+ (Table). Common pathway alterations in HER2+ USC include TP53, RTK RAS, PI3K, NOTCH, chromatin remodeling and cell cycle genes. Single gene alterations in HER2+ tumors that may implicate HER2 therapy resistance (based on pathway analyses in other tumor types) included PI3K (36%), KRAS (2.6%), and PTEN (2.1%). HER2+ tumors had low immunotherapy biomarker profiles (0.3% MSI-H, 0.8% TMB, 17.1% PD-L1). Conclusions: High concordance rates were observed between CISH and IHC. Ultimately these testing platforms need to be validated by response to HER2 targeted therapies in order to develop USC specific HER2 testing guidelines.[Table: see text]

11539 Background: PEComa is a rare mesenchymal neoplasm composed of perivascular epithelioid cells. Due to its rarity, diagnosis is challenging and no standardized treatment guidelines have been established. A subgroup of PEComas are characterized by a loss of function mutation in TSC1/2 that activates the PIK3-Akt-mTOR pathway. In the majority of patients, however, the molecular landscape and the composition of the tumor microenvironment (TME) remain largely unclear. Thus, we conducted this study to elucidate the genetic landscape of PEComas. A comparative analysis was performed with melanoma as a representative immunogenic tumor type. Methods: Thirty-five PEComa specimens were centrally analysed at the Caris Life Sciences laboratory. NextGen DNA sequencing (NextSeq, 592 gene panel or NovaSeq, whole-exome-sequencing), whole-transcriptome RNA sequencing (NovaSeq) and immunohistochemistry (Caris Life Sciences, Phoenix, AZ) were performed. Gene expression profiling (GEP) was performed by unsupervised hierarchical clustering. RNA deconvolution analysis was performed using the Microenvironment Cell Populations (MCP)-counter method to quantify immune cell populations (Becht 2016, Genome Biology). Results: The most common mutations detected in this cohort were TP53 (47%), ATRX (32%), TSC1/2 (11%/29%) and MSH3 (17%). Interestingly, TP53 mutations occurred less frequently (25 vs 60%, p = 0.055) in TSC1/2-mutated ( TSC1/2-mt) compared to TSC1/2-wildtype ( TSC1/2-wt) tumors, whereas MSH3 (25%, n = 1/4) and ERCC2 (14%, n = 2/14) mutations were exclusively observed in TSC1/2-mt cases. TSC1/2 mutations and other mTOR signalling pathway alterations, including two TFE gene fusion transcripts, were mutually exclusive. Of note, we found that 33.3% (n = 2) of TSC2-mt tumors were associated with high PIK3-Akt-mTOR pathway expression, while 100% (n = 3) of TSC1-mt tumors demonstrated lower expression. Deficient mismatch repair/microsatellite instability-high and high tumor mutational burden were rare (2.9%, n = 1 each) and observed concurrently in absence of PD-L1 expression. Overall, PD-L1 expression was observed in 21.9% (n = 7) of patients. An exploratory comparison with melanoma revealed that PEComa TMEs were characterized by a significant increase of NK cells and fibroblasts, as well as a relevant decrease of CD8+ T cells and B cells. Conclusions: Within this study we discovered a heterogeneous molecular landscape with a high prevalence of TSC1/2 mutations that were in part associated with transcriptional up-regulation of the PIK3-Akt-mTOR pathway. Furthermore, the relatively immune-cold TME compared to melanoma suggests increased lymphocyte infiltration may be required to increase the efficacy of immune checkpoint inhibitors for PEComa.

3091 Background: Gene fusions often represent critical therapeutic targets across cancer subtypes. Fusions within the ErbB family of receptor tyrosine kinases, including EGFR, ERBB2 ( HER2) and ERBB4 ( HER4), have been previously described and represent potentially actionable alterations. Here, we report the relative incidence and functional characterization of these rare genomic events. Methods: Tumor samples (n = 64,354; representing > 40 tumors types) submitted to Caris Life Sciences (Phoenix, AZ) were molecularly profiled by next-generation sequencing of DNA (NextSeq, 592-gene panel; or NovaSeq, whole exome) and RNA (whole transcriptome). Gene fusion partners, in/out-of-frame status, retention of ERBB kinase domain, and topology of fusion breakpoints were characterized for each ERBB fusion transcript detected. Fusion prevalence was further examined in public data sets (TCGA, MSK-IMPACT and AACR GENIE). Results: From the Caris database, a total of 64 EGFR fusion isoforms were detected in 59 tumors (incidence 0.09%); 83% were in-frame and 91% retained the EGFR kinase domain. 206 ERBB2 fusion isoforms were detected in 114 tumors (0.18%); 37% were in-frame and 34% retained the ERBB2 kinase domain. 131 ERBB4 fusion isoforms were detected in 108 tumors (0.17%); 62% were in-frame and 51% retained the kinase domain. All fusions were detected at low incidence across all tumor types. EGFR fusions were most common in high grade glioma (1.7%, n = 35), largely driven by recurrent EGFR-SEPT14 fusions (n = 20). ERBB2 fusions were most common in esophageal/gastroesophageal junction carcinoma (1.1%, n = 20), with recurrent fusion to PGAP3 observed in multiple tumor types (n = 37). ERBB4 fusions were most common in ovarian (0.7%, n = 40) and bladder (0.7%, n = 15) cancers, which often resulted from recurrent fusion with IKZF2 (n = 36). EGFR and ERBB2 fusions were generated predominantly (44-48%) from inversion events, while ERBB4 fusions arose more frequently and at similar rates (27-32%) from deletions, duplications, or translocations. Mining of public data sets corroborated the prevalence of ERBB gene fusions: the frequency of EGFR fusions was 0.63%, ERBB2 was 0.14% and ERBB4 was 0.04%. TP53 mutations frequently co-occurred with ERBB2 and ERBB4 fusions ( > 60% average across public data sets), with higher co-mutation rates ( > 70%) observed for samples in the Caris database. Conclusions: ERBB gene fusions are detectable at low frequency in various tumor types and may represent a unique genomic subset of cancer. Identification of novel ERBB gene fusions warrants further investigation to determine the potential pathogenicity and actionability of these fusions.

3563 Background: The tumor microenvironment (TME) of colorectal cancers (CRC) is modulated by oncogenic drivers such as KRAS. The TME comprises a broad landscape of immune infiltration. How tumor genomics associates with the immune cell landscape is less known. We aim to characterize immune cell types in RAS wild-type (WT) and mutant (MT) CRC, and to examine the prevalence of immuno-oncologic (IO) biomarkers (e.g. tumor mutation burden (TMB), PD-L1, MSI-H/dMMR) in these tumors. We performed genomic and transcriptomic analysis to confirm associations of mutant RAS with immune infiltration of the TME conducive to metastasis vs. potential response to immunotherapies. Methods: A total of 7,801 CRC were analyzed using next-generation sequencing on DNA (NextSeq, 592 Genes and WES, NovaSEQ), RNA (NovaSeq, whole transcriptome equencing) and IHC (Caris Life Sciences, Phoenix, AZ). MSI/MMR was tested by FA, IHC and NGS. TMB-H was based on a cut-off of > 10 mutations per MB). Immune cell fraction was calculated by QuantiSeq (Finotello 2019, Genome Medicine). Significance was determined by X2 and Fisher-Exact and p adjusted for multiple comparisons (q) was = than 50 yrs. In MSS tumors, there was a significantly higher neutrophil infiltration in KRAS MT (median cell fraction 6.6% vs. 5.9%) and NRAS MT (6.9%) overall and also when individual codons were studied. B cells, M2 macrophages, CD8+ T cells, dendritic cells and fibroblasts were lower in KRAS mutant tumors; B cells and M1 macrophages are lower in NRAS (q

e18050 Background: Adenoid cystic carcinoma (ACC) is a rare malignancy of glandular tissue with a high rate of local recurrence and metastatic disease for which clinical behavior varies widely. Currently, no standard therapy exists, and available therapies have low response rates. Therefore, prognostic biomarkers to determine optimal treatment strategies are urgently needed. Here we investigate the molecular landscape and therapeutic targets of ACC. Methods: ACC samples were analyzed using next generation sequencing (NGS) (NextSeq, 592 gene panel) or whole exome sequencing (NovaSeq) (Caris Life Sciences, Phoenix, AZ). Pathogenic fusion events were detected using whole transcriptome sequencing (NovaSeq). Statistical significance was determined using the Chi-square test and adjusted for multiple comparisons. Tumor infiltrating Immune cell fractions were determined using the QuanTIseq deconvolution algorithm and WTS data. Results: 327 ACC patients were identified, median age of 58 years, 62% female and 62% metastatic disease. The most frequent mutations (excluding indeterminate results) involved genes in the NOTCH pathway ( NOTCH1 16% [41 of 298 cases]), chromatin remodeling ( ARID1A 24.3% [27 of 138 cases], KDM6A 12.5% [28 of 224 cases], KMT2C 10.7% [21 of 197 cases]) and TP53 10.2% (25 of 246 cases). Chromatin remodeling genes were co-mutated with NOTCH1. NOTCH1 mutations were found in 41 cases, 18 primary and 23 metastatic were associated with reduced median overall survival (mOS 1.8 years vs 3.8 years, p = 0.01). Mutations were most frequent in the PEST (n = 35), NRR (n = 18), PEST+NRR (n = 16) and EGF-like domains (n = 11). PEST domain mutations were found exclusively in ACC originating from the head and neck (H&N). GSEA showed that MYC, E2F and G2M target pathways were enriched in NOTCH1 mutated ACC regardless of the presence of MYB fusions. All three pathways were enriched in PEST domain mutations whereas only MYC targets were enriched in NRR and NRR+PEST domain mutations indicating that PEST domain mutations drive transcriptomic changes in NOTCH1 mutated ACC. MYC gene expression was significantly upregulated in NOTCH1 mutated ACC. Fusion events were detected in 45% (104/231). MYB:NFIB was the most common fusion (40%) and was more frequent in H&N origin than other sites. There was no difference in the mOS between MYB fusion positive and negative ACC (4.4 years vs 5.5 years, p = 0.14). M2-polarized macrophages, myeloid dendritic cells and neutrophils were significantly higher in metastatic compared to primary ACC while NK cells were more abundant in primary tumor. Conclusions: ACC is a genetically heterogenous disease with an immune-excluded microenvironment. NOTCH1 mutations were associated with worse, while MYB:NFIB fusion was not associated with prognosis. Our study shows that M2 macrophages and MYC are potential novel therapeutic targets in ACC.

5588 Background: Recent data has shed light on molecular profiles of uterine carcinosarcoma (UCS), but few have correlated molecular profiles with prognosis. In a preliminary data analysis, we found that hormone receptors (HR)—estrogen receptor (ER) and progesterone receptor (PR)—expression was associated with improved OS. Here, we investigate the molecular profile differences between ER+/- and PR +/- tumors. Methods: Tumor samples were analyzed using Next-Generation sequencing of DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and RNA (NovaSeq, whole transcriptome sequencing) and immunohistochemistry (IHC) at the Caris Life Sciences Laboratory (Phoenix, AZ).ER and PR tested by IHC on whole tumor (cut-off: +1, 10%). PD-L1 IHC used SP-142 (cut-off >1%). MSI was tested by FA, IHC an NGS. TMB was measured by totaling somatic mutations per tumor (high > 10 mutations per MB). Immune cell fraction was calculated by QuantiSeq. Overall survival (OS) information was obtained from insurance claims data and Kaplan-Meier estimates were calculated for molecularly defined cohorts. Statistical significance was determined by chi-square and Wilcoxon rank sum test and p-values adjusted for multiple comparisons (q) to be

e18024 Background: Multiple studies investigating immune checkpoint inhibitors (ICI) in head and neck squamous cell cancer (HNSCC) patients have demonstrated prolonged survival of p16+ versus p16- tumors. However, the data has been conflicting. Methods: We queried the Caris Life Sciences CODEai database to assess survival outcomes of HNSCC patients who received ICIs comparing p16+ and p16- subgroups. A standard cut-off of 2+, >70% p16 staining was used. Presence of HPV16/18 genomes was tested using whole exome sequencing. Patients were considered smokers if they had >15 pack-years of tobacco use. PD-L1 expression was assessed by the 22c3 antibody, ≥1 being positive. Gene mutations were evaluated using Next-Generation Sequencing (NGS) (Caris Life Sciences, Phoenix, AZ). Tumor mutational burden (TMB) was assessed by somatic nonsynonymous missense mutations. Real world overall survival (rwOS) was obtained from insurance claims data and Kaplan-Meier estimates were calculated from the date of treatment start to the date of last contact. Time on treatment (TOT) was calculated from date of start to completion of ICIs. Results: 2905 patients with HNSCC were identified in the Caris database. 41% (215/525) were smokers. Among patients who were tested for p16 and/or HPV, 32% (251/791) expressed p16 and 28% (91/236) were HPV+. The majority of p16+ tumors were oropharynx (OP) in origin (68%, 171/251). 87% (970/1111) of tumors expressed PD-L1 and 16% (216/1362) had TMB ≥10/Mb. TP53 (54%, n=1115/2076) and CDKN2A (17%, n=281/1649) were the most common mutations. When compared to p16-, the p16+ group had a higher prevalence of TMB ≥10/Mb (26% vs 17%) and RB1 mutations (21% vs 4%) but lower number of smokers (15% vs 34%) and TP53 mutations (32% vs 58%). Similar to previous reports, p16+ oropharynx squamous cell carcinoma (OPSCC) patients survived longer than p16- patients, rwOS, 47 vs. 20 months (HR=0.55, p = 0.014) respectively. Among patients who were treated with ICIs, p16+ and p16- non-OP HNSCC, rwOS was not reached (NR) for both groups at follow up of 22 months while TOT for p16+ and p16- respectively, was 3.5 vs. 2.7 months, HR 0.507, p=0.039. No statistically significant difference was found in rwOS (19.3 vs. 24 months, HR 1.8, p=0.27) or TOT (3.7 vs. 4.1 months, HR 0.78, p=0.38) between p16+ and p16- OPSCC groups treated with ICIs, respectively. Conclusions: P16+ non-OP HNSCC patients receiving ICIs were found to remain on treatment longer compared to p16- patients which was not reproduced in the OPSCC subgroup. Randomized controlled trials are needed to verify p16 as a prognostic marker for ICI therapy.[Table: see text]

5063 Background: ATM mutations, one of a family of DNA repair defects prevalent in prostate cancer, have been included in a list of actionable mutations for PARP inhibitor (PARPi) therapeutic trials. Despite preclinical evidence, PARPi have shown minimal clinical activity in ATM mutant prostate cancer (ATMmPCa). The present analysis explores co-occurring genomic alterations that may drive outcomes of metastatic PCa (mPCa) patients with tumors harboring ATM mutations and provide clues for understanding therapy resistance and potential targets. Methods: This study included molecular profiling analysis of 1375 cases of mPCa. Tumors were analyzed using next-generation sequencing (NGS), whole transcriptome sequencing (WTS), and immunohistochemistry (IHC) (Caris Life Sciences, Phoenix, AZ). dMMR/MSI-H status was determined by IHC, NGS, and fragment analysis and tumor mutational burden (TMB) was calculated based on somatic nonsynonymous missense mutations. We performed differential gene expression analysis of HR-associated transcripts such as ATR, PARP1-3, RAD50, RAD51A/B/C/D and RAD54. Significance was determined using the ꭓ2 test and Benjamini-Hochberg method. Results: Fifty-nine (4.2%) cases harbored pathogenic ATM mutations, 84 (6.2%) harbored BRCA2 mutations. 1018 tumors (74%) were deemed homologous recombination proficient (HRP) and 155 tumors (11.3%) were HR Deficient (HRD); harboring one or more mutation in HR-related genes excluding ATM and BRCA2. The mutation rate of TP53 was significantly lower in ATMmPCa (12.0%) compared to BRCA2mPCa (35%), HRD (35%) and HRP (46.6%) tumors. ATMmPCa showed higher rates of SMAD2 (3.7%/1%) and FLCN (5.2%/0.3%) alterations compared to HRP cases. PARP1 and RAD51D gene expression was reduced in ATMmPCa compared to HRP (p < 0.05) and BRCA2mPCa ( p < 0.05) tumors, respectively. No differences in gene expression levels were detected for ATR, PARP2, PARP3, RAD50, and RAD54. Chromosomal segments demonstrating differential CNA in ATMmPCa vs HRP, HRD, or BRCA2mPCa included FGF19, FGF4, PTPN11, ALDH2, DAXX, BCL7A, CCND1, BMPR1A and MEF2B (Q-value < 0.05 determined by ꭓ2). The most common CNA in ATMmPCa was CCND1, present in approximately 13% (7/55) of cases. Compared to BRCA2mPCa and HRD cases, ATMmPCa cases are less likely to display markers of immunotherapy response such as dMMR/MSI-H or TMB ≥10 mutations/MB. Conclusions: ATMmPCa demonstrated several differences in co-occurring alterations compared to BRCA2mPCa, HRD and HRP mPCa. ATMmPCa tumors were less likely to harbor alterations in TP53 compared to BRCA2, HRD or HRP tumors. CNA in ATMmPCa occurred in 9 genes across distinct mPCa molecular subtypes and were enriched for those associated with the 11q13 amplicon harboring Cyclin D1. The FGF and PTPN11 related pathways are potentially targetable pathways in ATMmPC and may merit further study.

5582 Background: Uterine carcinosarcoma (UCS) is subtype of endometrial cancer (EC) with aggressive behavior and poor prognosis. UCS has not traditionally been included in EC clinical trials and treatment options are limited. Immune-oncology (IO) therapy has shown promise UCS, but it is unknown which patients benefit most. We sought to identify immunogenic markers in UCS and explore treatment response to IO therapy. Methods: Tumor samples were analyzed using Nex-Gen sequencing of the DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and RNA (NovaSeq, whole transcriptome sequencing) and immunohistochemistry (IHC) (Caris Life Sciences, Phoenix, AZ).PD-L1 IHC used SP-142 (cut-off >1%). MSI was tested by FA, IHC and NGS. TMB was measured by totaling somatic mutations per tumor (high > 10 mutations per MB). Immune cell fraction was calculated by QuantiSeq. Overall survival (OS) information was obtained from insurance claims data and Kaplan-Meier estimates were calculated for molecularly defined patient cohorts. Statistical significance was determined using chi-square and Wilcoxon rank sum test and p values adjusted for multiple comparisons (q) to be

e14033 Background: Advanced age is associated with poorer outcomes in GBM and current NCCN guidelines distinguish older GBM patients (≥70 years, oGBM) from their younger counterparts ( < 70 years, yGBM). We aim to characterize age-related comprehensive mutational profiles with the long-term goal of improving treatment strategies, outcomes, and rational clinical trial design. Specifically, we focused on IDH-wildtype (WT) oGBM, investigated if oGBM are more likely to acquire temozolomide-induced hypermutations, and finally, compared the frequency of high-tumor mutational burden (TMB) between oGBM vs. yGBM. Methods: Comprehensive molecular profiles of 1,657 adult IDH-WT GBM tumors tested at Caris Life Sciences (Phoenix, AZ) were queried. Tests included NGS of DNA (NextSeq, 592 Genes and NovaSEQ, WES) and RNA (NovaSeq) sequencing. SBS11 gene signature (i.e temozolomide-induced hypermutational profile) was queried using SigProfiler (Alexandrov 2020, Nature). Significance was determined by X2 and Fisher-Exact and p adjusted for multiple comparisons ( q) was < 0.05. Results: We identified 1,657 patients (range 21-89 years old, median 61 years) with IDH-wildtype GBM, 22% (360) of whom were ≥ 70 years. There was a slight male predominance (60%) for all ages. The most prevalent alterations in oGBM were TERT promoter mutation (105/131,80%), MGMT promoter methylation (pMe) (175/346, 51%), and PTEN mutation (129/349, 37%). EGFR amplification was seen in 35% (125/356) and EGFRvIII in 23% (81/360); Overall, fusions were seen in 12% (44 of 360) oGBM; events > 1% included MET (3.6%), FGFR3 (3.1%), EGFR (2.6%) and ROS1 (1.4%). 17% (56/349) of oGBM had positive PD-L1 by IHC. High TMB ( > 10mt/Mb) tumors were rare (3.1%) and MSI-high tumors even rarer (0.8%) in oGBM. When compared to yGBM, MGMT pMe was more prevalent (51% vs 38%, risk ratio (RR) 1.35 [1.19-1.52], q < 0.05) and NF1 mutations were less frequent in oGBM (21% v 34%, RR 0.62 [0.50-0.77], q < 0.05). No significant differences were seen in other key markers examined. The prevalence of SBS11 gene signature across all ages (data available for 1,141 patients) was 1.2% and was comparable across the age spectrum; no significant difference seen in the MGMT pMe group when oGBM was compared to yGBM (3.9% vs. 1.8 %, p= 0.3). Conclusions: This study represents the largest comprehensive molecular characterization of older IDH-WT GBM patients. We show that molecular profiles of IDH-WT GBM are remarkably similar across the age spectrum, including immunotherapy-associated markers, gene fusion landscape, EGFR amplification, and TMB. The significantly higher prevalence of MGMT pMe and lower NF-1 mutation rate in the older population bear significant prognostic and therapeutic implications.

9087 Background: Recent data suggest inferior responses to immune checkpoint inhibitors (ICIs) in STK11-mt NSCLC. TP53 is a critical tumor suppressor gene regulating DNA repair by arresting cells in the G1 phase in response to critical double strand breaks. We hypothesized that accumulated DNA damage from mutations in the TP53 gene might increase immunogenicity and potentially enhance benefit of ICIs in STK11-mt NSCLC. Methods: A total of 16,896 NSCLC tumors submitted to Caris Life Sciences (Phoenix, AZ) for targeted NGS (DNA-Seq, 592 genes) were analyzed. A subset (N = 5034 tumors) had gene expression profiling (RNA-Seq, whole transcriptome). PD-L1 (TPS) was tested with 22c3 antibody (Dako). Exome-level neoantigen load for STK11-mt NSCLC was obtained from published TCGA Pan-immune analysis (Thorsson et al. 2018). Non-parametric tests were used for comparing differences in tumor mutational burden (TMB) and neoantigen load. Transcriptomic analysis included differential gene expression and hierarchical clustering. Tumor immune cell content was obtained from transcriptome using Microenvironment Cell Population-counter (MCP). Publicly available data from the POPLAR/OAK trials of atezolizumab in advanced NSCLC were used to model PFS and OS for STK11-mt with TP53-mt (n = 14) and without TP53-mt (n = 20). Results: Of 16,896 NSCLC samples, 12.6% had an STK11-mt with the proportions of TMB-high (≥10 Mut/Mb), PD-L1 ≥ 50% and MSI-high being 55.9%, 11.8%, and 0.72%, respectively. STK11-mt vs. STK11-wt NSCLC did not differ in median TMB (Caris:10 vs. 10 Mut/Mb; p > 0.1) or neoantigen load (TCGA: 154.5 vs. 165; p > 0.1). Median TMB (13 vs. 9 Mut/Mb; p < 0.001) and neoantigen load (263 vs. 134; p < 0.001) were higher in STK11-mt/ TP53-mt vs. STK11-mt/ TP53-wt. MCP analysis showed higher CD8, NK-cell and lower myeloid dendritic cell infiltration in STK11-mt/ TP53-mt vs. STK11-mt/ TP53-wt (p < 0.01). Expression of MYC and HIF-α were increased in the STK11-mt/ TP53-mt vs. STK11-mt/ TP53-wt (p < 0.01) along with higher expression (p < 0.01) of genes associated with both glycolysis ( HK2, LDHA, ALDOA) and glutamine metabolism ( GOT2, PPAT2). Hierarchical clustering of STK11-mt adenocarcinomas (n = 463) for STING pathway genes (CCL5, CXCL10, cGAS) identified a STING-high and a STING low cluster. The STING high cluster was significantly enriched in TP53-mt (48 vs. 32%; p < 0.01).In the OAK/POPLAR cohort, median OS (HR is 1.14, 95% CI 0.53 - 2.48); p > 0.1) and PFS (HR 1.88, 95% CI 0.89-3.97, p = 0.098) were not statistically different between STK11-mt/ TP53-mt vs. STK-mt/ TP53-wt. However, the 15-months PFS was 21% in the STK11-mt/ TP53-mt vs 0% in the STK11-mt/ TP53-wt. Conclusions: STK11-mt NSCLC with TP53-mt are associated with an immunologically active TME with metabolic reprogramming. These intrinsic properties could be exploited to improve outcomes to ICIs in combination with metabolically directed agents.

5587 Background: In a preliminary data analysis to identify prognostic molecular biomarkers in uterine carcinosarcoma (UCS), we found that alterations in KMT2C, a gene involved in the chromatin remodeling pathway, correlated with improved survival. We sought to explore relevant biomarkers of KMT2C-mutated (KMT2C-mut) tumors compared to wildtype (KMT2C-wt) tumors. Methods: Tumor samples were analyzed using next generation sequencing (NGS) of the DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and RNA (NovaSeq, whole transcriptome sequencing) and immunohistochemistry (IHC) (Caris Life Sciences, Phoenix, AZ).PD-L1 IHC used SP-142 (cut-off >1%). Microsatelite instability (MSI) was tested by FA, IHC and NGS. TMB was measured by totaling somatic mutations per tumor (high > 10 mutations per MB). Immune cell fraction was calculated by QuantiSeq. Overall survival (OS) information was obtained from insurance claims data and Kaplan-Meier estimates were calculated for molecularly defined patient cohorts. Statistical significance was determined using chi-square and Wilcoxon rank sum test and p values adjusted for multiple comparisons (q) to be

3537 Background: Aberrant NT signaling has been shown to activate uncontrolled proliferation and dissemination in several gastrointestinal cancer types. Neurotransmitters have been shown to affect endothelial cells and immune cells in the tumor microenvironment to promote tumor progression. We previously showed that single nucleotide polymorphisms in the dopamine and GABA pathways are associated with outcome in patients with metastatic CRC receiving first-line treatment. Here we further evaluated the distribution and molecular context of NT pathway alterations in CRC. Methods: A total of 7,595 CRC tumors tested at Caris Life Sciences (Phoenix, AZ) with NextGen Sequencing on DNA (Next Seq, 592 genes or NovaSeq, WES) and RNA (NovaSeq, WTS) were analyzed. ssGSEA (single-sample gene set enrichment analysis) was used to calculate pathway enrichment scores (ES) of 7 NT gene sets (GABA, nicotinic, muscarinic, dopamine (DA), reelin, glial cell line-derived neurotrophic factor and neurotrophins). X2/Fisher-Exact was used for comparison and significance was determined as p-value adjusted for multiple comparison of ( q) < 0.05. Results: ES based on sample sites showed a substantial heterogeneity in NT enrichment. Notably, when compared to primary tumors, all 7 gene sets were significantly enriched in brain metastases (mets; ES ratio 1.14-1.55), while abdomen, liver, and peritoneal mets displayed significant decreases in most NT gene sets. DA was enriched in ovarian and lung mets (ES ratio: 1.18 and 1.09, respectively), the latter also showing increased neurotrophins ES (1.06) (all q < 0.05). When investigating primary tumors grouped according to overall ES by unsupervised clustering, right-sided and CMS4 CRCs were more prevalent in the high ES cluster compared to the low ES cluster (32 vs 29%, P = 0.02 and 46 vs 30%, P < 0.001, respectively). In addition, tumors in the high ES cluster showed lower prevalence of TMB-H (≥ 10mt/MB) (7 vs 10%), MSI-H (6 vs 10%) and PD-L1 (2 vs 6%), while higher CNA rates were noted in 9 genes (all q < 0.05). High ES tumors showed significant positive associations with microenvironment infiltration of B cells, T cells (NK, CD4+ and CD8+ T cells, but not Treg), M2 Macrophages, Myeloid Dendritic Cell, Neutrophils, and an inverse association with M1 Macrophages, regardless of MSI status ( q < 0.05). Conclusions: This is the first and most extensive molecular profiling study to investigate NT signaling pathway alterations in CRC. Our data show a distinct distribution of pathway enrichment according to metastatic site, distinct molecular features in high vs low ES clusters in primary tumors (including CMS subtypes, TMB, MSI and PD-L1 rates), and differential immune cell infiltration. These findings support the role of NT signaling in the metastatic spread of CRC and modulation of tumor immune microenvironment.

e18026 Background: Previous studies have consistently shown that p16+ oropharynx squamous cell cancer (OPSCC) patients have better outcomes. The significance of molecular and transcriptional signatures and its correlation with p16 expression remains unclear. Methods: We queried the Caris Life Sciences database to assess the molecular and transcriptional signatures related to p16+ and p16- head and neck squamous cell cancer (HNSCC) patients. Comprehensive molecular profiling including whole exome sequencing (WES), targeted Next-Generation Sequencing (NGS), and immunohistochemistry (IHC) 22c3 for PD-L1 was performed (CPS ≥1 considered positive) (Caris Life Sciences, Phoenix, AZ). Tumor mutational burden (TMB) was calculated based on somatic nonsynonymous missense mutations. A standard cut-off of 2+, > 70% p16 staining was used. HPV16/18 was tested using WES. Patients were considered smokers if they had > 15 pack-years. Real world overall survival (rwOS) was obtained from insurance claims data and Kaplan-Meier estimates were calculated from the date of start of treatment to the date of last contact. Results: 948 cases of OPSCC were identified in the Caris database. 41% (82/199) were smokers. Where p16 and HPV data was available, 41% were p16+ (171/420) while HPV positivity rate was 52% (71/148). We noted a small number of patients with discordant p16 and HPV status (7 p16+/HPV-, 8 p16-/HPV+) in 327 patients who had HPV status available. Most common mutations were TP53 (33%), PIK3CA (17%) and KMT2D (10.6%). 87% were PD-L1 positive (342/394), with high expression in both p16+ and p16- subgroups, 90% and 85% respectively. 10% had TMB≥10/Mb (48/463). TP53 mutations were more common in p16- (49%) tumors in contrast to p16+ (10%) (p < 0.0005), while no statistical difference was detected in TMB≥10/Mb between the groups. CDKN2A, TERT and NOTCH1 mutations were more prevalent in tumors that were p16- or HPV- in contrast to tumors that were p16+ or HPV+ (p < 0.05). FGF3, CCND1, FGF4, FGF19 copy number alterations (CNA) were less common in p16+ OPSCC compared to p16- or HPV16- OPSCC (p < 0.0005). P16+ patients had longer rwOS when compared to p16-, 47 months versus 20 months (HR = 0.55, p = 0.014) respectively. Conclusions: Higher frequency of CDKN2A, TERT and NOTCH1 mutations among p16- (versus p16+) raise the possibility of potential targets for treatment in this group with poor prognosis. However, it remains unclear if these can serve as independent predictors of survival.[Table: see text]

3092 Background: ADP-ribosylation factor 6 ( ARF6) is a small GTPase in the RAS superfamily, which regulates membrane trafficking, remodeling and tumor progression. Preclinical study shows that TP53 and KRAS cooperatively activate the ARF6-AMAP1 pathway which serves as a link by which pancreatic driver mutations promote tumor invasion, PD-L1 dynamics and immune evasion properties in pancreatic ductal adenocarcinoma (PDAC). The clinical impact of ARF6 on cancer progression and prognosis remains unclear. Methods: A total of 2,948 PDAC samples were analyzed using next-generation sequencing of RNA (whole transcriptome, NovaSeq) and DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and immunohistochemistry (IHC) (Caris Life Sciences, Phoenix, AZ). QuantiSeq (Finotello 2019, Genome Medicine) was used to quantify immune cell infiltration. Overall survival (OS) was obtained from insurance claims, and Kaplan-Meier estimates were calculated for molecularly defined cohorts. Significance was determined as p values adjusted for multiple correction ( q) of

5590 Background: Uterine serous carcinoma (USC) is an aggressive subtype of endometrial cancer with poor prognosis and limited treatment options. Immune-oncology (IO) agents have shown promise USC, however data is limited regarding which patients benefit most from IO therapy. In other malignancies, PD-L1, MSI-H status and high TMB have been predictive of IO response. We sought to characterize the immune profiles of USC and investigate treatment response to IO therapy. Methods: Tumor samples were analyzed using Nex-Gen sequencing of the DNA (NextSeq, 592 genes or NovaSeq, whole exome sequencing) and RNA (NovaSeq, whole transcriptome sequencing) and immunohistochemistry (IHC) (Caris Life Sciences, Phoenix, AZ).PD-L1 IHC used SP-142 (cut-off >1%). MSI was tested by FA, IHC and NGS. TMB was measured by totaling somatic mutations per tumor (high > 10 mutations per MB). Immune cell fraction was calculated by QuantiSeq. Real-world overall survival (OS) information was obtained from insurance claims data and Kaplan-Meier estimates were calculated for molecularly defined patient cohorts. Statistical significance was determined using chi-square and Wilcoxon rank sum test and p values adjusted for multiple comparisons (q) to be < 0.05. Results: Molecular analysis was performed on 2,806 USC tumors. The median age was 67 years. 65.3% were from primary tumors, 34.7% from metastatic sites. In total, 92 patients were treated with IO therapy and had a significantly longer median survival than those not treated with IO (months: 59.6 vs 31.2; HR(95% CI): 0.38(0.24-0.61) p < 0.001), resulting in a survival advantage of 867 days. PD-L1 expression was present in 19.1% of cases, but only 2.3% of tumors were MSI-H and 4.2% were TMB-H. Patients with these markers trended toward a better median survival, but this was not significant; PD-L1 (months: 34.4 vs 31.2; HR(95% CI): 0.90 (0.74-1.1), MSI-H (OS not yet reached vs 31.6 months; HR(95% CI): 0.69(0.38-1.25) and TMB-H (months: 36.4 vs 31.6; HR(95% CI): 0.84(0.50-1.39). Regarding the immune microenvironment, the most common infiltrating immune cells were M2 Macrophages (5.35%), B cells (4.71%), myeloid dendritic cells (3.45%), NK cells (2.94%) and regulatory T cells (1.59%). There were few CD8 T cells and non-regulatory CD4 T cells. Conclusions: IO therapy was associated with a median survival benefit of more than 2 years in USC. We did not identify any prognostic markers of IO-therapy response. MSI-H and TMB-H are rare in USC, but PD-L1 is present in nearly 20% of cases. Notably these markers did predict a significant survival benefit, which has important clinical implications. Further study is warranted.

4022 Background: The prognosis of biliary tract cancers (BTC) remains dismal and novel treatment strategies are needed to improve survival. Polybromo-1 ( PBRM1) is a subunit of the PBF chromatin-remodeling complex and preclinical studies suggest induction of synthetic lethality by PARP inhibitors in PBRM1-mutated cancers. Therefore, we aimed to describe the molecular landscape in BTC harboring PBRM1 mutations. Methods: 1,848 BTC samples were included in this study. Specimens were analyzed using NextGen DNA sequencing (NextSeq, 592 gene panel or NovaSeq, whole-exome sequencing), whole-transcriptome RNA sequencing (NovaSeq) and immunohistochemistry (Caris Life Sciences, Phoenix, AZ). Pathway gene enrichment analyses were done using GSEA (Subramaniam 2015, PNAS). Immune cell fraction was calculated by QuantiSeq (Finotello 2019, Genome Medicine). Survival was calculated from time of tissue collection to last contact using Kaplan-Meier estimates. Results: PBRM1 mutations were identified in 8.1% (n = 150) of BTC tumors and were more prevalent in intrahepatic BTC (9.9%) than in gallbladder cancer (6%, p = 0.0141) and in extrahepatic BTC (4.5%, p = 0.008). In PBRM1-mutated tumors, we found a higher rate of MSI-H/dMMR (8.7% vs. 2.1%, p < 0.0001) and a higher median TMB (4 vs. 3 mt/MB, p < 0.0001). When compared to PBRM1-wildtype cancers higher rates of co-mutations in chromatin-remodeling genes (e.g. ARID1A, 31% vs. 16% , p < 0.0001) and DNA damage repair pathway (e.g. ATRX, 4.4% vs. 0.3%, p < 0.0001) were detected. Within PBRM1-mutated tumors, a significant higher frequency of infiltrating M1 macrophages was observed (p < 0.0001). Gene set enrichment analysis revealed that genes associated with tumor inflammation (e.g. HLA-DRA, HLA-DRB1, IFNGR1) were enriched in PBRM1-mutated tumors (NES = 2.02, FDR = 1.3%, p < 0.0001). Overall survival analysis showed that PBRM1 mutations were associated with a favorable outcome (HR 1.502, 95% CI [1.013-2.227], p = 0.041). This relationship was also present in MSS subgroup (HR: 1.667, [1.026-2.71], p = 0.037). Conclusions: This is the largest and most extensive molecular profiling study focusing on PBRM1-mutated BTC. Co-mutations in chromatin-remodelling and DNA damage repair genes might set the stage for clinical testing of PARP inhibitors in PBRM1-mutated BTC. Moreover, a distinct tumor microenvironment characterized by high M1 macrophages infiltration and an enrichment of inflammatory genes suggest a potential benefit of immunotherapy.

3527 Background: Globo H is a carbohydrate antigen that is highly expressed on the cell surface of epithelial cancers but not in normal tissue, and has been reported to correlate with poor prognosis. An attractive therapeutic target, Globo H-targeted agents are being tested in early clinical trials (e.g., OBI-833, a Globo H antigen conjugated to a mutated diphtheria toxin with potential antineoplastic activities, and OBI-999, an antibody-drug conjugate (ADC) consisting of a Globo H monoclonal antibody with a synthetic antineoplastic agent). We aim to describe the molecular features associated with Globo H expression in CRC. Methods: A total of 7,604 CRC tumors were tested by Caris Life Sciences (Phoenix, AZ) by NextGen DNA and RNA sequencing. The expression of β3GalT5, FUT-1 and FUT-2 were evaluated as surrogates for Globo H expression as they are the key enzymes in its biosynthesis. An average z-score of the 3 genes (GloboH) and of β3GalT5 (B3) alone were calculated; tumors with top quartile z-scores were considered expression-high (Q4) and bottom quartile, expression-low (Q1). QuantiSEQ was used to assess immune cell infiltration in the tumor microenvironment (TME). Statistical significance was determined using chi-square/Fisher-Exact and adjusted for multiple comparisons (q=10) (11.4% vs. 8.3%), PD-L1 positive (5.7% vs. 3.4%) and MSI-H/dMMR (8.3% vs. 5.5%). Strong positive associations were seen with mutations in BRAF, KRAS, RSPO3 fusion, and cMYC amplification with B3 alone and GloboH (all q

e18056 Background: Adenoid cystic carcinoma (ACC) is a rare malignancy of glandular tissue with a high rate of local recurrence and metastatic disease. Despite being regarded as an indolent disease, the clinical course of recurrent and metastatic ACC (R/M ACC) is highly variable. Responses to chemotherapy (chemo) are uniformly poor. Several multi-targeted tyrosine-kinase inhibitors (mTKIs), EGFR inhibitors (EGFRi) and other targeted agents have been studied in single-arm early phase trials with response rates ranging from 0-16% and progression free survival ranging from 2.5-17 months. However, there have been no comparative clinical trials and it is not known if one treatment strategy is superior. We undertook this retrospective study to assess the real-world clinical outcomes in patients with adenoid cystic carcinoma using the Caris Life Sciences database. Methods: Real world overall survival (rwOS) for cases of ACC was obtained from insurance claims data and Kaplan-Meier estimates were calculated from the date of collection to the date of last contact. Cases were divided into subgroups based on treatment received – chemo (including platinum agents, taxanes, 5FU, topoisomerase inhibitors, anthracyclines), EGFRi (cetuximab, erlotinib, lapatinib), mTKIs (pazopanib, axitinib, sunitinib, cabozantinib, lenvatinib, sorafenib) and immune checkpoint inhibitors (ICIs) (atezolizumab, ipilimumab, pembrolizumab, nivolumab). Results: 368 patients (pts) were identified with ACC, 16 were locally recurrent and 216 tumors were taken from metastatic sites. 50 pts received chemo, 6 were treated with EGFRi and 15 with mTKIs. Pts who received combination EGFRi and chemo or mTKI and chemo were excluded. The median overall survival (mOS) all patients with metastatic ACC was 2.8 years (yrs). The mOS of pts with R/M ACC was 3 yrs for chemo, 2.9 yrs for EGFRi and 1.5 yrs for mTKIs. There was no significance in mOS between chemo vs mTKIs (HR 0.85, 95% CI 0.3 - 2, p = 0.72) and chemo vs EGFRi (HR = 0.88, 95% CI 0.3 - 2.5, p = 0.78). We further compared the outcomes of those treated with EGFRi (n = 8) with mTKIs (n = 19) in the entire cohort. For most pts, these agents were given as front line therapy. 25% (2/8) of patients had received treatment prior to EGFRi and 20% (4/9) prior to mTKIs (p = 1). There was no significant difference in mOS with HR 0.6 (95% CI 0.16 - 2.6), p = 0.6. We also compared the mOS of patients who received ICIs (n = 22) with those who did not (n = 346) but there was no significant difference (mOS 3.19 vs 3.17 yrs respectively, HR 0.87, 95% CI 0.47- 1.61, p = 0.65). Conclusions: There was no significant difference in the mOS between pts with R/M ACC who were treated with chemo, EGFRi or TKIs and in those who received ICIs compared to those who did not in our limited patient population. This highlights the need for predictive biomarkers for better patient selection with the goal of personalizing treatment strategies for this disease.

This study aimed to determine the safety and effectiveness of bevacizumab (B) plus FOLFOX or CAPOX in advanced neuroendocrine tumors (NETs) by performing a combined analysis of 2 separate prospective phase II studies.In the FOLFOX/B study, patients received chemotherapy without scheduled breaks in 3 cohorts: carcinoid, pancreatic NET, and poorly differentiated neuroendocrine carcinomas. In the CAPOX/B study, NET subtypes were pooled, and patients were treated with 4 cycles of CAPOX/B followed by optional maintenance therapy. Primary end points were radiographic response rate (RR) after 12 cycles (FOLFOX/B), progression-free survival (PFS) (CAPOX/B), and toxicity (both).Seventy-six patients (FOLFOX/B, n = 36; CAPOX/B, n = 40) were included. In FOLFOX/B, RR for carcinoid at 12 cycles 3/22 (13.6%), median PFS 19.3 months; RR for pancreatic NET at 12 cycles 4/12 (41.7%), median PFS 21 months; RR 1/2 (50%) in poorly differentiated neuroendocrine carcinoma; pooled RR 25% and median PFS 21 months (1-year PFS 68%). In CAPOX/B (pooled NET), RR 18% and median PFS 16.7 months (1-year PFS 65%). Predictable toxicity was observed.Neither study met its primary end point, but radiographic responses and prolonged disease stability in previously progressing patients suggest that selected patients with NET may benefit from oxaliplatin-fluoropyrimidine chemotherapy plus bevacizumab and that the combination may warrant further study.

Highlights • CUP occurs in as many as 3–5% of patients when standard diagnostic tests are not able to determine the origin of cancer. • MI GPSai (Genomic Prevalence Score) is an AI that uses genomic and transcriptomic data to elucidate tumor origin. • The algorithm was trained on molecular data from 57,489 cases and validated on 19,555 cases. • MI GPSai predicted the tumor type out of 21 options in the labeled data set with an accuracy of over 94% on 93% of cases. • When also considering the second highest prediction, the accuracy increases to 97%., Cancer of Unknown Primary (CUP) occurs in 3–5% of patients when standard histological diagnostic tests are unable to determine the origin of metastatic cancer. Typically, a CUP diagnosis is treated empirically and has very poor outcomes, with median overall survival less than one year. Gene expression profiling alone has been used to identify the tissue of origin but struggles with low neoplastic percentage in metastatic sites which is where identification is often most needed. MI GPSai, a Genomic Prevalence Score, uses DNA sequencing and whole transcriptome data coupled with machine learning to aid in the diagnosis of cancer. The algorithm trained on genomic data from 34,352 cases and genomic and transcriptomic data from 23,137 cases and was validated on 19,555 cases. MI GPSai predicted the tumor type in the labeled data set with an accuracy of over 94% on 93% of cases while deliberating amongst 21 possible categories of cancer. When also considering the second highest prediction, the accuracy increases to 97%. Additionally, MI GPSai rendered a prediction for 71.7% of CUP cases. Pathologist evaluation of discrepancies between submitted diagnosis and MI GPSai predictions resulted in change of diagnosis in 41.3% of the time. MI GPSai provides clinically meaningful information in a large proportion of CUP cases and inclusion of MI GPSai in clinical routine could improve diagnostic fidelity. Moreover, all genomic markers essential for therapy selection are assessed in this assay, maximizing the clinical utility for patients within a single test.

116 Background: Cachexia affects many cancer patients. Growth differentiation factor-15 (GDF15) is a protein that regulates weight and the stress response of cells. The GDF15 gene encodes a ligand of TGF-beta that triggers cachexia and modulates the progression from tumorigenesis to metastasis. Inhibition of GDF15 with an antibody restored muscle mass and fat in animal models. Serum levels rise in proportion to the progression of colon cancer, predict outcome, and have been correlated with CEA. Methods: We retrospectively reviewed 7607 CRC tumors profiled by Caris Life Sciences (Phoenix, AZ) from 2019 to 2020. Profiling included whole transcriptome sequencing (RNA-Seq by NovoSeq). Tumor mutational burden, mismatch repair status, and pathway genomic alterations were evaluated. QuantiSEQ was used to assess immune cell infiltration in the tumor microenvironment. Results: GDF15 expression ranged from 0 to 593 transcripts per million (TPM) with median of 30 (IQR = 15.02). There was no association with age, sex, or primary tumor sidedness. MSI-H/dMMR tumors had higher GDF15 expression (median 37 vs 30, p = 0.0004); TMB > = 17 tumors was seen in 5.9% of bottom quartile (Q1) GDF15 expressors and 8.3% of top quartile (Q4). PDL1 IHC positivity was inversely correlated with GDF15 expression (7.1% in Q1 vs. 2.6% in Q4, p < 0.0001). Genomic alterations associated with higher GDF15 expression (Q4 vs Q1) included genes on TGF-B (SMAD2/4), PI3K (PIK3CA, MTOR), chromatin remodeling (ARID1A, KMT2C), DDR (ATM) and Wnt pathway (APC); those inversely associated included MYC CNA and TP53. Q1 tumors had higher CNA of ERBB2 and FGFR1. Relative neutrophils and NK cells in the TME increased from Q1 to Q4 (p < 0.001). There was a decrease in CD8+ T-cells and Treg cells from Q1 to Q4. Conclusions: GDF15 expression correlates with increased dMMR/MSI-H and TMB, but not with PDL1 expression. Mutations and activated pathways associated with GDF15 expression may explain increased cachexia with more aggressive disease. The association with chromatin remodeling may warrant therapies targeting histone modification and epigenetics. The increase in NK cells but decrease in CD8+ T cells in the TME with increasing GDF15 suggests approaches to treatment. Higher CD8+ lymphocyte counts correlate with PFS with immunotherapy. Anti-PD-L1 therapy reinvigorates the killing function of CD8+ T cells. The decrease in CD8+ T cells and PDL1 positivity with rising GDF15 suggests worse outcome and a lack of response to anti-PDL1 therapy. NK cell checkpoint inhibitors, CARs, and an anti-GFRAL antibody are now in clinical trials and might be utilized in high GDF15 cancers. GDF15 is emerging as a target in the treatment of obesity and cachexia and as a prognostic marker in oncology. Understanding its expression in metastatic colon cancer may reveal which patients could benefit from developing anti-GDF15 targeted therapies against cancer progression.

BACKGROUND There is a clear survival benefit to neoadjuvant chemoradiation prior to esophagectomy for patients with stages II-III esophageal cancer. A minimally invasive esophagectomy approach may decrease morbidity but is more challenging in a previously radiated field and therefore patients who undergo neoadjuvant chemoradiation may experience more postoperative complications. METHODS A prospective database of all esophageal cancer patients who underwent attempted hybrid minimally invasive Ivor Lewis esophagectomy was maintained between 2006 and 2015. The clinical characteristics, neoadjuvant treatments, perioperative complications, and survival outcomes were reviewed. RESULTS Overall 30- and 90-day mortality rates were 0.8% (1/131) and 2.3% (3/131), respectively. The majority of patients 58% (76/131) underwent induction treatment without significant adverse impact on mortality, major complications, or hospital stay. Overall survival at 1, 3, and 5 years was 85.9%, 65.3%, and 53.9%. Five-year survival by pathologic stage was stage I 68.9%, stage II 54.0%, and stage III 29.6%. CONCLUSIONS The hybrid minimally invasive Ivor Lewis esophagectomy approach results in low perioperative morbidity and mortality and is well tolerated after neoadjuvant chemoradiation. Good long-term overall survival rates likely resulted from combined concurrent neoadjuvant chemoradiation in the majority of patients, which did not impact the ability to safely perform the operation or postoperative complications rates. J. Surg. Oncol. 2016;114:838-847. © 2016 2016 Wiley Periodicals, Inc.

Background: Neoadjuvant chemotherapy followed by radical cystectomy is the optimal treatment for patients with muscle invasive bladder cancer. Between 30 and 40% of patients receiving neoadjuvant chemotherapy will have pathological complete response at the time of radical cystectomy. Presence of somatic mutations in DNA damage repair genes correlate with chemoresponse, likely stemming from a loss-of-function in the ability to repair DNA damage brought on by cytotoxic chemotherapy. However, many chemoresponders have WT DNA repair status. As such, there is a keen interest to identify additional mechanisms associated with chemoresponse. Results: A higher tumor mutation/neoantigen burden is associated with response to immune checkpoint blockade, so we sought to investigate a correlation between mutation burden and chemoresponse. We found that responders had a higher neoantigen burden and higher signature of tumor infiltrating CD8+ T cells (TILs). We interrogated a database of tumors that underwent clinical tumor sequencing (NextGen SEQ, Caris Life Sciences; 592 genes; n=897) and PD-L1 staining (Combined Positive Score [CPS] using 22C3 antibody) and found that mutation burden was directly proportional to CPS. In addition, we found that CPS was positively associated with KRAS and TP53 mutation, whereas it was negatively associated with FGFR3 activating alterations, MDM2 amplification and GATA3 amplification after correction for multiple comparisons. There was an overall increase in the proportional representation of TIL-derived T cell receptors (TCR) repertoire in the periphery after chemotherapy independent of responder status. This implies that the presence of immune response even in patients without pathologic response. Additionally, TIL-derived TCRs were also found in perivesical lymph nodes in both responders and non-responders, however, there was a distinct profile of activation/exhaustion associated with either patient group. Responders showed a unique CD69+/PD1−(activated/non-exhausted) population amongst the nodal CD8+ T cells suggesting that this population is required to achieve pathologic complete response. Conclusions: These results further our understanding of immune responses to chemotherapy in bladder cancer and may enable the development of chemo-immunotherapy regimens to improve response rates to neoadjuvant chemotherapy. Citation Format: Rashida Ginwala, Alexander Macfarlane, IV, David Liu, R K. Alpaugh, Joanne Xiu, Zoran Gatalica, W M. Korn, Eliezer M. Van Allen, Kerry S. Campbell, Alexander Kutikov, Elizabeth Plimack, Philip H. Abbosh. Immune correlates of response to neoadjuvant chemotherapy in muscle-invasive bladder cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6624.

2549 Background: TMB-H in gliomas is caused by molecular alterations or alkylator treatment- induced genomic changes characterized by a large number of G:C>A:T transitions. Our study describes the molecular features of TMB-H gliomas. Methods: Gliomas were tested with NextGen sequencing (592 genes), MGMT promoter methylation (MGMT-m) fragment analysis and IHC at Caris Life Sciences. Microsatellite instability (MSI) was test by NGS, FA/IHC. The GC:AT transition rate was calculated as the prevalence of G:A and C:T changes seen in each tumor and > 80% was regarded as high transition(TR-H). TMB values were compared using Wilcoxon Rank Sum. TMB-H was defined as the top quartile of all TMB values (TMB>9). Results: TMB in the 3129 gliomas ranged from 0 to 372 mutations/MB (mean: 8.5, median: 6). TMB-H was observed in 31% of glioblastomas, 16% of astrocytomas (astro) (22% of grade III, 7% of grade I/II) and 22% of oligodendrogliomas (oligo) (32% of grade III and 15% of grade I/II). MGMT-m (58% vs. 47%; p=0.0001), pathogenic (p) or likely p (lp) EGFR (14% vs 10%, p=0.004) and PIK3CA mutations (13% vs. 9%, p=0.002), as well as p/lp in 30 other genes were more prevalent in TMB-H cases (p100 (median 264) and TR-L; 4 of the 5 were also MSI-H. PDL1 IHC had no correlation with TMB, MSI or transition rates. In 89 paired samples taken >150 days apart (regardless of intervening treatment), acquisition of TMB-H was seen in 11 pairs: 8 glioblastomas, 2 grade II/III astro and 2 oligo. In the paired tumors that acquired TMB-H status compared to those that did not, a significantly higher prevalence of MGMT-m (82% vs. 37%, p=0.008) and IDH mutation (64% vs. 19%, p=0.004) were seen. 10 of the 11 recurrent tumors with acquisition of TMB-H had TR-H while none in the other 78 pairs. Conclusions: TMB varies significantly in gliomas and associates with POLE, TR-H and MSI-H, but not with an increase of PD-L1. POLE-mutated tumors had the highest TMB levels. TR-H, an indicator of alkylator-induced phenotype, is associated with a higher TMB than MSI-H, however, TR-H may synergize with MSI-H to further increase TMB. Tumors with an IDH mutation and MGMT-m are more prevalent in tumors with high TMB gain. Further understanding of molecular and immune profile of the TMB-H may facilitate more individualized treatment planning.

4077 Background: In the TRIBE2 study molecularly unselected and untreated mCRC patients were randomized to receive FOLFOXIRI/bevacizumab (bev) followed by the same agents after disease progression (PD) or FOLFOX/bev followed by FOLFIRI/bev after PD. We performed a comprehensive NGS analysis of samples from randomized patients in order to investigate the prognostic impact of tumor mutational load (TML), its additional value with respect to the assessment of microsatellite instability (MSI), and the overall prevalence of potentially actionable alterations. Methods: Tumor DNA was obtained from formalin-fixed, paraffin-embedded blocks from primary tumors of 296 (44%) out of 679 randomized patients and underwent NGS analysis using the Caris MI TumorSeek panel, assessing 592 genes. TML was defined low, intermediate or high if < 7, 7-16 or > 16 mutations/Mb were found. MSI status was determined both by NGS and by IHC. Results: TML and MSI were successfully determined by NGS in 224 (76%) cases. NGS and IHC results were concordant in 221 (99%) cases. TML was low, intermediate or high in 56 (25%), 157 (70%) and 11 (5%) cases, respectively. When compared with TML low and intermediate tumors, TML high were more frequently right-sided (p = 0.013), mucinous (p < 0.001) and MSI-high (p < 0.001). TML high tumors were MSI-high or MSS in 8 (73%) and 3 (27%) cases, respectively. Two out of 3 TML high and MSS tumours showed a pathogenic POLE mutation (p.S459F and p.P286R). The other TML high, MSS and POLE wt tumor was dMMR at IHC (loss of MSH6 expression) and showed a pathogenic MSH6 mutation (p.F1040fs). As compared with low and intermediate TML, high TML was associated with longer PFS (median PFS: 17.3 vs 10.6; HR: 0.54 [95%CI: 0.35-1.09], p = 0.098) and OS (median OS: not reached vs 23.7: HR: 0.45 [95%CI:0-28-1.13], p = 0.106). No interaction effect between TML and treatment arm was observed, and no difference between TML low and intermediate tumors was reported in terms of baseline characteristics and prognosis. Actionable alterations ( HER2 mutations [N = 2] and amplifications [N = 4], KRAS G12C [N = 10] and BRAF V600E mutation [N = 39]) were found in 55 (19%) out of 296 cases. No NTRK/ROS/ALK or MET amplification was found. Conclusions: TML high tumors are not limited to MSI-high ones but showed POLE or MSH6 somatic mutation and shower longer PFS and OS. No differences are reported between TML low and intermediate tumors. Molecular alterations predictive of benefit from targeted strategies currently available are detectable only in a small percentage of mCRCs.

3113 Background: NRG1 fusions are actionable genomic alterations detected across tumor types. The NRG1 gene encode for neuregulin, which serves as a ligand for ERBB3 and ERBB4 receptors and activates downstream signaling through the MAPK and PI3K pathways. Here, we update the detection of NRG1 gene fusions across tumor types and further describe fusion characteristics. Methods: Samples submitted for clinical molecular profiling that included RNA-sequencing (Archer Dx or Caris MI transcriptome) were retrospectively analyzed for NRG1 fusion events. All NRG1 fusions with ≥ 3 junction reads were identified for manual review and for characterization of fusion class, intact functional domains, domain prediction, breakpoints, frame retention and co-occurring alterations by NGS. Results: A total of 82 NRG1 fusion events (0.2% of 44,570) were identified. Among the fusions identified, the distribution across tumor types was as follows: non-small cell lung cancer (NSCLC, 54%), breast cancer (11%), ovarian cancer (7%), pancreatic cancer (7%), cholangiocarcinoma (6%), colorectal cancer (5%), and other (10%). Forty-two unique fusion partners were identified, the most common being CD74 (23%), ATP1B1 (9%), SLC3A2 (7%), RBPMS (6%) and SDC4 (4%). Almost half (47%) of all fusion events are expected to include the transmembrane domain contributed by the NRG1 fusion partner. Lung and pancreatobilliary cancers had the highest rates of transmembrane domain retention from their fusion partners (63.6% and 54.5%, respectively). In all other tumor groups, most fusion partners lacked transmembrane domains. In 15% of cases, the chimeric transcripts are predicted to lead to increased expression of NRG1. The most commonly reported breakpoints in NRG1 occur in exon 6 and exon 2. While fusions with the NRG1 breakpoint at exon 2 retain the immunoglobulin (Ig) domain and all downstream portions (including EGF-like domain), those at exon 6 do not contain the Ig portion and result in shorter chimeric proteins. The breakpoints in all CD74:NRG1 fusions, the most common fusions in NSCLC, occur at exon 5 or 6 and cause truncation of domains upstream of the EGF-like domain. In ATP1B1:NRG1 fusions, the most common fusions in pancreatobilliary cancers, the breakpoints are at exon 1 or 2 and retain the Ig domain. Conclusions: NRG1 fusion products are diverse across tumor types, but the significance of these variations is not clear. The biological and clinical implications of retaining certain domains of NRG1 (such as the Ig domain) and of fusion partners warrants further investigation.

4558 Background: The increased PD-L1 expression evaluated by combined positive score (CPS) is associated with improved efficacy of immunotherapy in GE cancers. The impact of tumor molecular alterations on PD-L1 expression is still not well-studied. We aimed to characterize specific molecular features of tumors with different CPS levels in GE cancers. Methods: 2,707 GE tumors [1,662 gastric/GE junction adenocarcinoma (GA), 856 esophageal adenocarcinoma (EA), 75 esophageal squamous (ES) and 114 GE unspecified] collected between 2000.8 and 2019.7 were analyzed using NextGen DNA sequencing (NGS), immunohistochemistry (IHC) and fragment analysis (FA) (Caris Life Sciences, Phoenix, AZ). Tumor mutation burden (TMB) was calculated based on somatic nonsynonymous missense mutations. dMMR/MSI status was evaluated by a combination of IHC, FA and NGS. PD-L1 expression measured by IHC (22c3) was evaluated by CPS scores. Molecular alterations were compared in three groups (CPS ≥ 10, H; CPS = 1~9, M; CPS = 0, L) using Fisher-Exact or Chi-square and adjusted for multiple comparison by Benjamini-Hochberg. Significance was determined by adjusted (adj) p < .05. Results: Overall, CPS-H, M, and L were seen in 18% (n = 494), 28% (n = 765) and 53% (n = 1,448) of GE tumors respectively. CPS-H was the most prevalent in ES (43%) followed by GA (19%) and lowest in EA (14%). Overall, TMB was similar between CPS-L and M, but was significantly increased in H (average TMB = 8.4 vs. 8.6 vs. 11 mt/MB, adj p < .0001); the effect was seen in EA and GA, but not in ES. An overall significant association between MSI/dMMR status and PD-L1 expression levels was seen (2%, 3.2% and 12% in CPS-L, M and H, adj p < .05) in GE tumors; the significance was seen in GA, but not in EA or ES. Amplifications of PD-L1 (H: 1.5%, M: 0.1% and L: 0) and PD-L2 (H: 1.1%, M: 0.1%, L: 0) were the highest in CPS-H, while ASPSCR1 (H: 0, M: 0, L: 1%) and TNFRSF14 (H: 0, M: 0.4, L: 2%) were the lowest (adj p < .01). Genes involved in epigenetic modification (top 5: ARID1A, ASXL1, BCL9, BCOR, CREBBP), MAPK ( KRAS, MAP2K1) and mismatch repair ( MLH1, MSH6) had the highest mutation rates in CPS-H, compared to M and L ( p < .0001). In contrast, CDH1 had higher mutation rates in CPS-L (12%), compared to M and H (5% and 5%) ( p < .0001). Conclusions: This is the largest study to investigate the distinct molecular landscape of pts with different PD-L1 expression levels in GE cancers. Our data may provide novel insights for pt selection (e.g. pts with gene mutations involved in epigenetic modification) and the development of rational combination immunotherapy (e.g. drugs targeting MAPK pathway).

4629 Background: Genomic profiling has identified KRAS mutations in 88-90% of PC. KRAS WT tumors represent a molecularly heterogeneous group that may harbor targetable alterations (TA). We studied KRAS WT PC using NextGen sequencing (NGS) and whole transcriptome sequencing (WTS) to characterize the molecular landscape of this unique group and to assess the prevalence of TA. Methods: A total of 1164 PC tumors were tested at Caris Life Sciences by NGS (592 genes), WTS (NovaSeq), IHC and fragment analysis. Comparison of KRAS WT vs. mutant (MT) was done by Fisher-Exact or Chi2 and was corrected for multiple tests. Results: The KRAS WT cohort included 144 tumors (12.4%). No differences were seen in gender (female: 46% in both WT & MT) and age (median: 66 & 67) compared to KRAS MT. In KRAS WT tumors, targetable fusions tested by WTS and pathogenic mutations by NGS were seen in 22% (32 of 144) and 52% (75 of 144) respectively; potentially targetable amplifications (amp) were seen in 5 additional tumors. No TA were seen in 22% of WT tumors. Key alterations are in Table. Alterations inducing MAPK activation, including BRAF, RAF1, NF1 and GNAS changes were seen in 38 (26%) tumors. Frequent alterations were seen in FGFR genes (11 tumors), MET (4, including 1 exon 14 skip), and ERBB receptor and ligands (10). Fusions in ALK, ROS1, RET and NOTCH1 were seen (8), largely exclusive of other drivers. Wnt, PI3K, chromatin remodeling (CR) and DDR changes were common and sometimes seen concurrent with other alterations. Compared to KRAS MT, no difference of PD-L1 or TMB-H was seen. BRAF, APC, PBRM1, CTNNB1 mutations, MDM2 amp, gene fusions and MSI-H/dMMR were all more frequent in KRAS WT tumors (corrected p < 0.05). Conclusions: KRAS WT PC is enriched with TA (e.g., BRAF, ALK, ROS1, NRG1, MSI-H). The use of WTS in combination with NGS identifies activated molecular pathways in the majority of KRAS WT tumors. Based on our findings, comprehensive profiling of PC at the DNA and RNA level is recommended to provide patients with therapeutic opportunities beyond standard treatments. [Table: see text]

4113 Background: Inactivating alterations in MAP3K1/MAP2K4 occur in various solid tumors, sensitize cancer models to MEK inhibitors, and have co-mutation partners which may enable therapeutic targeting. Methods: We retrospectively reviewed 20290 GI malignancy patients (pts), comprised of 9986 colorectal carcinoma (CRC) and 10304 non-CRC, whose tumors were profiled with Caris Life Sciences from 2015-2019. Profiling included immunohistochemistry (IHC) with programmed death ligand-1 (PD-L1), next-generation sequencing (NGS), tumor mutational burden (TMB) and deficient mismatch repair or microsatellite instability-high status (dMMR/MSI-H). Results: MAP3K1/MAP2K4-alteration ( MAP3K1/MAP2K4-MT) was more frequent in CRC than non-CRC pts (2.0% v. 1.2%, p

9056 Background: Malignant pleural mesothelioma (MPM) is a relatively uncommon malignancy with poor prognosis and no major therapeutic breakthroughs over the past decade. Better understanding of the genomic landscape and distribution of immune biomarkers in this disease has the potential to enable development of novel therapies. Methods: We analyzed molecular profiles of 222 MPM tumors using next-generation sequencing of 592 genes utilizing Caris Life Sciences platform. Genes were grouped into pathways: DNA damage repair (DDR) ( ATM, BRCA2, BRIP1, BAP1, CHEK2, ERCC2, FANCA/D2/E/L, MLH1, MSH6, MUTYH, NBN, PMS2, RAD50/51B, WRN), cell cycle regulation including TP53 ( RB1,CCNE1, CDKN2A, CCND1, CCND3, CDKN1B), chromatin remodeling (CR) ( ARID2, ASXL1, DNMT3A, EP300, EZH2, KDM6A, KMT2C, KMT2D, NSD3, PBRM1, SMARCB1/A4, SETD2), RAS/MAPK ( KRAS, MAP2K1, NF1, NF2), and PI3K/AKT ( AKT, PIK3CA, PIK3R1/R2, PTEN, RICTOR, TSC1, TSC2, ZNF703). Tumor mutational burden (TMB), PD-L1 expression (SP142 IHC, tumor staining), and MSI/MMR were analyzed. Seventy-two cases also had whole transcriptome sequencing data. Differences in alterations were compared for age, gender, and pathways. Results: Median age of patients (pts) was 72 yr (range, 37-90), 73% were male. Gene pathway alterations were seen in 81% of cases. DDR, specifically homologous recombination (HR), was the most commonly mutated pathway (36.9%), followed by RAS (25.2%) and CR (18.9%). Genes mutated in ≥5% of cases included BAP1 (26.3%), NF2 (23.5%), TP53 (15.5%), SETD2 (10.2%). PD-L1 was high (≥50% tumor cells positive) in 11.4% (n = 24), intermediate (1-49%) in 31.4% (n = 66), and negative ( < 1%) in 57.1% (n = 120) pts. TMB was high (≥10 mutations/Mb) in 9.6% of tumors (n = 20). None of the tumors were dMMR/MSI-H. HR gene BAP1 and CR gene SETD2 mutations trended to be more prevalent in pts ≥70 yo (p = 0.02). CR trended to be more commonly mutated in females (p = 0.02). No other significant differences were found in specific gene/pathway alterations, PD-L1 expression, or TMB in the context of age and gender. Distribution of PD-L1 expression was not different among various pathways. No highly recurrent, targetable fusion isoforms were seen among the 85 identified (mean 1.1 fusions/tumor), which have not yet been characterized for pathogenicity. Conclusions: The majority of MPM tumors harbor alteration in one of the key cellular pathways. HR pathway mutations are the most common. The majority of tumors were PD-L1 negative and carry low TMB indicating low immunogenicity. No age and gender specific differences exist except for BAP1 and SETD2 mutations.

9544 Background: KRAS is the most commonly mutated oncogene in NSCLC and the development of direct KRAS inhibitors has renewed interest in this molecular subtype. However, there are several different KRAS mts, representing unique biology and different prognostic and therapeutic impact. A more comprehensive understanding of the genomic landscape relative to each KRAS mt subset will help guide therapeutic development. Methods: Molecular profiles of 17,113 NSCLC specimens were obtained using next-generation sequencing of 592 genes (Caris Life Sciences) and classified based on presence and types of KRAS mt. Incidence of KRAS mts was noted across the cohort and by histology. Co-occurring genomic alterations, tumor mutational burden (TMB) and PD-L1 IHC (22C3, TPS score) were analyzed by KRAS mt type. Results: Across the entire cohort, 4706 (27%) of samples harbored a KRAS mt (Table). The most common was G12C (40%), followed by G12V (19%) and G12D (15%). The prevalence of KRAS mt was 37.2% among adenocarcinoma and only 4.4% in squamous. High TMB, defined by > 10 mts/Mb, varied across the different KRAS mt types, most common in G13X (68.3%) and least common in G12D (43.2%). PD-L1 expression also varied. G12C was the most likely to be PD-L1 positive, with 65.5% TPS > 1%, and the most likely to be PD-L1 high, with 41.3% TPS > 50%. STK11 was mutated in 8.6% of KRAS wild type NSCLC but more frequently noted in every KRAS subtype, with the highest rate in G13X (36.2%) and the lowest in G12D(14.2%). TP53 mts were more frequent in KRAS wild type NSCLC (73.6%), with the highest rate among KRAS mutants at 55.4% (G12other) and the lowest at 36.8% (Q61X). NF1 was noted to be mutated in 21.4% of KRAS G13X cases, while all other KRAS mts had a lower frequency of NF1 mts than KRAS wild type (11.5%). Conclusions: KRAS mts are relatively common in lung adenocarcinoma and KRAS G12C is the most common variant. The different KRAS mts have different co-occurring mutations and a different genomic landscape. KRAS G12C was associated with the highest rate of PD-L1 expression. The clinical relevance of these differences in the context of therapeutic intervention warrants investigation. [Table: see text]

Background Werner syndrome gene (WRN) encodes a DNA helicase with an exonuclease activity that contributes to DNA repair. In cancer, WRN mutations lead to genomic instability. It is known that WRN is necessary to sustain in-vivo growth of cancers cells with microsatellite instability (MSI), including CRC. WRN is a very promising new target especially in cancers with MSI. There is still a lack of knowledge about the frequency of WRN alterations and their association with immunological and molecular phenotypes. Methods Tumour samples from 6854 CRC patients were analyzed using NGS (NextSEQ on 592 genes), in-situ hybridization and immunohistochemistry (Caris Life Sciences, Phoenix, AZ, USA). Tumour mutational burden (TMB) was calculated based on somatic non-synonymous missense mutations, and MSI was evaluated by NGS of known MSI loci. Results WRN mutations (WRN-mut) were observed in 80 of 6854 samples (1.2%). A higher prevalence of WRN-mut was detected in right- compared to left-sided CRC (2.4% vs 0.7%, p Conclusions This is the largest profiling study to investigate the molecular and immunological landscape of WRN-mut CRCs. We show the high prevalence of MSI in WRN-mut tumours and their association with higher TMB and PD-L1 expression. Furthermore, it revealed that WRN-mut CRC is characterized by a distinct genetic profile. Our data might serve to tailor treatment in WRN-mut CRC. Legal entity responsible for the study The authors. Funding Has not received any funding. Disclosure All authors have declared no conflicts of interest.