Search

Background: Tumor immune markers such as tumor infiltrating lymphocytes (TILs) or expression-based profiles have been correlated with both response to neoadjuvant chemotherapy and prognosis in early breast cancer (BC) patients. Some chemotherapies, such as paclitaxel, lead to the development of TILs and in some cases, suppression of regulatory T-cells. Therefore, assessment of the tumor microenvironment (TME) could provide important information for clinical decision-making. The aim of this study was to test if RNAseq-based TME classification of BC tumors is predictive of pathological complete Response (pCR) and prognosis in the neoadjuvant GeparSepto (G7) trial (NCT01583426). Methods: We performed a retrospective-prospective analysis of a subset of 810 subjects of the total of 1207 patients of the G7 trial. In G7 HER2-negative early high-risk BC patients were studied to determine if nab-paclitaxel is superior to solvent-based paclitaxel. In addition to the taxane paclitaxel, both treatment arms received epirubicin plus cyclophosphamide before surgery. For this analysis, a subset of 279 HER2 negative patients with sufficient quality of pretherapeutic core biopsies to perform whole-transcriptome RNAseq (~200x106 reads per tumor) was used. Based on RNAseq data, immune activity classification was provided by ImmunityBio (Culver City, CA) by comparison of expression of 23 immune-cell-specific gene signatures as described by Bindea et al. (Immunity, 2013) to those from a reference population of 1467 similarly-profiled unselected tumor samples from a large tumor database (NantOmics, Culver City, CA). Unsupervised hierarchical clustering of inferred immune activities revealed 3 distinct groups termed “hot”, “warm”, and “cold” clusters. Logistic regression analysis based on age, trial arm, tumor size, nodal status, Ki-67, hormone-receptor (HR) status and immune activity cluster (hot/warm vs. cold) as independent variables was performed to predict pCR (ypT0/ypN0). Cox regression analysis with the same covariates was also performed to predict disease-free survival (DFS) and overall survival (OS). Results: Of the 279 patients, 67 had a pCR (24%). The analyzed subset was similar to the main HER2 negative population (pCR-rate: 22%). Patients with a “hot/warm” or “cold” immune activity assessment had a pCR in 30% and 13% of the cases, respectively. The odds-ratio of the multivariate logistic regression analysis was 2.17 (95% CI: 1.00-4.71, p=0.0512). With regard to DFS and OS, T follicular helper cell B-cell, and T-cell signatures seemed to play a prominent role, and the hazard ratios (also “hot/warm” vs. “cold”) for the multivariate analyses were 0.38 (95% CI: 0.21-0.66; p=0.0007) and 0.34 (95%CI: 0.16-0.72, p= 0.0045), respectively. Within the 23 individual immune-cell-specific gene signatures, CD56dimNatural Killer (NK), type 1 helper T-cells, and CD8+ T-cell signatures seemed to be closely associated with achievement of a pCR. RNAseq-based deconvolution of immune-cell activity was corroborated by IHC-based TIL scoring. Immune-hot/warm patients had more intratumoral lymphocytes compared to cold tumors (mean: 11.6% vs. 4.9%, p Citation Format: Peter A Fasching, Carsten Denkert, Stephen Benz, Karsten E Weber, Christopher Szeto, Jan Budczies, Andreas Schneeweiss, Elmar Stickeler, Sabine Schmatloch, Christian Jackisch, Thomas Karn, Hans Peter Sinn, Mathias Warm, Marion van Mackelenbergh, Sharooz Rabizadeh, Christian Schem, Ernst Heinmöller, Volkmar Müller, Frederik Marmé, Patrick Soon-Shiong, Valentina Nekljudova, Michael Untch, Sibylle Loibl. Tumor immune-cell activity assessed by RNAseq is an independent predictor of therapy response and prognosis after neoadjuvant chemotherapy in HER2 negative breast cancer patients - An analysis of the GeparSepto trial [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr PD5-08.

Background: Previous studies of breast samples from cosmetic surgeries, benign biopsies associated with abnormal mammograms or cancer-adjacent tissue have identified at least two different transcriptional phenotypes of “normal” human breast tissue (Troester, 2009; Haakensen, 2011), including an “Active” phenotype linked to increased risk of later breast cancer mortality (Roman-Perez, 2012; Troester, 2016). This study compares breast histology and transcriptional phenotypes from healthy parous women with no prior history of breast disease who donated breast core biopsies for research and supplied reproductive histories enabling breast cancer risk calculation. Since the Active transcriptome phenotype was recently associated with increased mammary adipocyte content, we focused on the possibility that adipocyte activation contributes to the Active transcriptome and drives breast cancer risk. Methods: RNA from paraffin-embedded tissue sections sufficient for RNAseq analysis (~100ng) was extracted from 151/200 core biopsies donated to the Komen Tissue Bank by healthy, parous white women (age range: 27-66, median = 45) with no history of breast cancer. Questionnaire data enabled breast cancer risk (Gail) score calculation; and digitized H&E images were used for histologic analyses. A previously validated classifying signature was used in unsupervised hierarchical clustering to identify samples with Active (78/151) vs. Inactive (73/151) transcriptome phenotypes for comparison with donor risk factors, breast tissue composition, and expression of candidate genes and gene signatures. Results: Mean (+/-SD) BMI and Gail score values were 29.60 (+/-7.92) and 1.27 (+/- 1.34), respectively; BMI scores were not significantly different by phenotype, but Gail scores were significantly higher for donors with an Active phenotype (1.46 vs. 1.18; p=0.007, Wilcoxon rank-sum). Active normal breast tissue samples possessed significantly more (%) adipocyte nuclei (p=3.9e-11) and greater adipocyte size (p80% of this donor cohort would not qualify for breast cancer chemoprevention, those with Active transcriptomes had significantly higher Gail scores supporting their increased future risk for breast cancer development. The Active breast transcriptome is strongly associated with increased adipocyte content, size, and overexpression of signatures and genes (including those previously linked to breast cancer progression) indicating a differentially activated adipocyte population. This dysregulated mammary adipocyte microenvironment not only appears to underlie the Active transcriptome phenotype but also precedes and potentially predicts the future histologic development of breast cancer. Citation Format: Christopher C Benz, Taekyu Kang, Christina Yau, Chris Wong, Yulia Newton, Charlie Vaske, Stephen C Benz, Gregor Krings, Roman Camarda, Jill E Henry, Josh Stuart, Mark Powell. The normal breast Active transcriptome associated with future breast cancer risk is driven by a dysregulated adipocyte microenvironment [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P3-08-20.

Background Breast cancer intrinsic molecular subtype (IMS) as classified by the expression-based PAM50 assay is considered a strong prognostic feature, even when controlled for by standard clinicopathological features such as age, grade, and nodal status, yet the molecular testing required to elucidate these subtypes is not routinely performed. Furthermore, when such bulk assays as RNA sequencing are performed, intratumoral heterogeneity that may affect prognosis and therapeutic decision-making can be missed. Methods As a more facile and readily available method for determining IMS in breast cancer, we developed a deep learning approach for approximating PAM50 intrinsic subtyping using only whole-slide images of H&E-stained breast biopsy tissue sections. This algorithm was trained on images from 443 tumors that had previously undergone PAM50 subtyping to classify small patches of the images into four major molecular subtypes—Basal-like, HER2-enriched, Luminal A, and Luminal B—as well as Basal vs. non-Basal. The algorithm was subsequently used for subtype classification of a held-out set of 222 tumors. Results This deep learning image-based classifier correctly subtyped the majority of samples in the held-out set of tumors. However, in many cases, significant heterogeneity was observed in assigned subtypes across patches from within a single whole-slide image. We performed further analysis of heterogeneity, focusing on contrasting Luminal A and Basal-like subtypes because classifications from our deep learning algorithm—similar to PAM50—are associated with significant differences in survival between these two subtypes. Patients with tumors classified as heterogeneous were found to have survival intermediate between Luminal A and Basal patients, as well as more varied levels of hormone receptor expression patterns. Conclusions Here, we present a method for minimizing manual work required to identify cancer-rich patches among all multiscale patches in H&E-stained WSIs that can be generalized to any indication. These results suggest that advanced deep machine learning methods that use only routinely collected whole-slide images can approximate RNA-seq-based molecular tests such as PAM50 and, importantly, may increase detection of heterogeneous tumors that may require more detailed subtype analysis.

Background Metastatic triple negative breast cancer (mTNBC) is an inherently diverse disease and while molecular classification of mTNBC has assisted in treatment decisions, if based on only an initial biopsy, it does not take into account the evolution of metastatic cancer. Characterization of emerging metastases is needed to reveal both new resistance or sensitivity to available therapeutics. The goal of “Intensive Trial of OMics in Cancer (ITOMIC) - Intensive Longitudinal Monitoring in Subjects With Triple-Negative Breast Cancer” (NCT01957514) - was to determine the feasibility of longitudinal collection of patient biopsies that would be subjected to molecular analysis to provide actionable, relevant and timely information to guide treatment decisions.Methods Multiple biopsies were collected longitudinally, including pre- and post-treatment, from 29 mTNBC patients enrolled in the ITOMIC study and subjected to multi-dimensional molecular profiling including WES, WGS, cancer gene panel sequencing, RNA-seq, and proteomics and/or IHC for tumor biomarkers. This information was used to guide iterative, patient- and tumor- individualized treatment recommendations made by a multi-institutional ITOMIC Tumor Board (ITB) and conveyed to each subject’s oncologist.Results Longitudinal biopsy collection was found to be safe. Molecular profiling revealed that 2 of an original 31 enrolled subjects likely had lung cancer rather than mTNBC, supporting the merit of repeated tissue analysis. While the other 29 subjects had all been given a diagnosis of mTNBC before entering the trial, estrogen receptor, progesterone receptor, and/or HER2 were found to be over-expressed in at least one sample for 12 subjects; appearance of receptor positivity suggests targeted therapy may be effective. Tumor evolution in response to the first on-study treatment for most subjects (cisplatin) was revealed by copy number alterations, changes in single nucleotide variants, and insertions/deletions in pre-/post-treatment biopsies. Over the course of the study, the ITB convened 54 times and 39 of 182 recommended treatments were evaluated and accessed through either an existing clinical trial, a single patient IND, approved off label or label indication. While not all ITB treatment recommendations were followed, 24 subjects did receive at least one ITB-recommended drug, frequently as part of a clinical trial. Currently, for 27 subjects (2 withdrew) median survival is ~31 months. There are 4 surviving patients in treatment with a remarkable median survival of >51 months.Conclusion Collection and molecular analysis of multiple biopsies during the course of patient’s disease, shown here to be safe and feasible, provides information vital to appropriate treatment choice and reveals new targets for and resistance to therapy in metastatic TNBC. Citation Format: Kimberly A Burton, Eric Q Konnick, Sibel Blau, Michael O Dorschner, Julie Gralow, Rahul Parulkar, Elisabeth Mahen, Patricia Spilman, Stephanie Parker, Francis M Senecal, Colin Pritchard, Christopher Szeto, Jing Zhu, Vijayakrishna K Gadi, Stephen C Benz, Shahrooz Rabizadeh, Patrick Soon-Shiong, Carl Anthony Blau. Multidimensional molecular profiling of repeated metastatic TNBC biopsies in the intensive trial of omics safely guides treatment decisions [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS11-13.

BackgroundTherapeutic regimens designed to augment the immunological response of a patient with breast cancer (BC) to tumor tissue are critically informed by tumor mutational burden and the antigenicity of expressed neoepitopes. Herein we describe a neoepitope and cognate neoepitope-reactive T-cell identification and validation program that supports the development of next-generation immunotherapies.MethodsUsing GPS Cancer, NantOmics research, and The Cancer Genome Atlas databases, we developed a novel bioinformatic-based approach which assesses mutational load, neoepitope expression, human leukocyte antigen (HLA)-binding prediction, and in vitro confirmation of T-cell recognition to preferentially identify targetable neoepitopes. This program was validated by application to a BC cell line and confirmed using tumor biopsies from two patients with BC enrolled in the Tumor-Infiltrating Lymphocytes and Genomics (TILGen) study.ResultsThe antigenicity and HLA-A2 restriction of the BC cell line predicted neoepitopes were determined by reactivity of T cells from HLA-A2-expressing healthy donors. For the TILGen subjects, tumor-infiltrating lymphocytes (TILs) recognized the predicted neoepitopes both as peptides and on retroviral expression in HLA-matched Epstein-Barr virus–lymphoblastoid cell line and BC cell line MCF-7 cells; PCR clonotyping revealed the presence of T cells in the periphery with T-cell receptors for the predicted neoepitopes. These high-avidity immune responses were polyclonal, mutation-specific and restricted to either HLA class I or II. Interestingly, we observed the persistence and expansion of polyclonal T-cell responses following neoadjuvant chemotherapy.ConclusionsWe demonstrate our neoepitope prediction program allows for the successful identification of neoepitopes targeted by TILs in patients with BC, providing a means to identify tumor-specific immunogenic targets for individualized treatment, including vaccines or adoptively transferred cellular therapies.

Different endogenous and exogenous mutational processes act over the evolutionary history of a malignant tumor, driven by abnormal DNA editing, mutagens or age-related DNA alterations, among others, to generate the specific mutational landscape of each individual tumor. The signatures of these mutational processes can be identified in large genomic datasets. We investigated the hypothesis that genomic patterns of mutational signatures are associated with the clinical behavior of breast cancer, in particular chemotherapy response and survival, with a particular focus on therapy-resistant disease.Whole exome sequencing was carried out in 405 pretherapeutic samples from the prospective neoadjuvant multicenter GeparSepto study. We analyzed 11 mutational signatures including biological processes such as APOBEC-mutagenesis, homologous recombination deficiency (HRD), mismatch repair deficiency and also age-related or tobacco-induced alterations.Different subgroups of breast carcinomas were defined mainly by differences in HRD-related and APOBEC-related mutational signatures and significant differences between hormone-receptor (HR)-negative and HR-positive tumors as well as correlations with age, Ki-67 and immunological parameters were observed. We could identify mutational processes that were linked to increased pathological complete response rates to neoadjuvant chemotherapy with high significance. In univariate analyses for HR-positive tumors signatures, S3 (HRD, P0.001) and S13 (APOBEC, P = 0.001) as well as exonic mutation rate (P = 0.002) were significantly correlated with increased pathological complete response rates. The signatures S3 (HRD, P = 0.006) and S4 (tobacco, P = 0.011) were prognostic for reduced disease-free survival of patients with chemotherapy-resistant tumors.The results of this investigation suggest that the clinical behavior of a tumor, in particular, response to neoadjuvant chemotherapy and disease-free survival of therapy-resistant tumors, could be predicted by the composition of mutational signatures as an indicator of the individual genomic history of a tumor. After additional validations, mutational signatures might be used to identify tumors with an increased response rate to neoadjuvant chemotherapy and to define therapy-resistant subgroups for future therapeutic interventions.

Background Previous studies have identified and validated a risk-associated Active transcriptome phenotype commonly expressed in the cancer-adjacent and histologically normal epithelium, stroma, and adipose containing peritumor microenvironment of clinically established invasive breast cancers, conferring a 2.5- to 3-fold later risk of dying from recurrent breast cancer. Expression of this Active transcriptome phenotype has not yet been evaluated in normal breast tissue samples unassociated with any benign or malignant lesions; however, it has been associated with increased peritumor adipocyte composition. Methods Detailed histologic and transcriptomic (RNAseq) analyses were performed on normal breast biopsy samples from 151 healthy, parous, non-obese (mean BMI = 29.60 ± 7.92) women, ages 27–66 who donated core breast biopsy samples to the Komen Tissue Bank, and whose average breast cancer risk estimate (Gail score) at the time of biopsy (1.27 ± 1.34) would not qualify them for endocrine prevention therapy. Results Full genome RNA sequencing (RNAseq) identified 52% (78/151) of these normal breast samples as expressing the Active breast phenotype. While Active signature genes were found to be most variably expressed in mammary adipocytes, donors with the Active phenotype had no difference in BMI but significantly higher Gail scores (1.46 vs. 1.18; p = 0.007). Active breast samples possessed 1.6-fold more (~ 80%) adipocyte nuclei, larger cross-sectional adipocyte areas (p p Active breast samples. Active samples were enriched in gene sets associated with adipogenesis and fat metabolism (FDR q ≤ 10%), higher signature scores for cAMP-dependent lipolysis known to drive breast cancer progression, white adipose tissue browning (Wilcoxon p Conclusions The risk-associated Active transcriptome phenotype first identified in cancer-adjacent breast tissues also occurs commonly in healthy women without breast disease who do not qualify for breast cancer chemoprevention, and independently of breast expressed cancer-associated mutations. The risk-associated Active phenotype appears driven by a pro-tumorigenic adipocyte microenvironment that can predate breast cancer development.

Background: Recent studies have identified at least two different transcriptional subtypes of benign human breast tissue (Haakensen et al., 2011), including an “active” subtype associated with increased risk of later life mortality from breast cancer (Troester et al., 2016). However, previous studies used "normal" breast samples from cosmetic surgery, biopsies from abnormal mammograms, or cancer-adjacent tissue. The present study evaluates normal breast transcriptome phenotypes from healthy women who donated for this study purpose only. Methods: 200 formalin-fixed paraffin-embedded (FFPE) breast tissue biopsy samples were analyzed from healthy, parous non-Hispanic white women ranging in age from 27–66 (median = 45) with no prior history of breast cancer ( Results: As expected, the normal histologic composition of these KTB samples varied considerably with mean % adipose area twice that of fibrous area, and epithelial content averaging Conclusion: Over 30% of healthy adult women with histologically normal breast tissue carry an “active” breast transcriptome phenotype previously linked to co-existent breast cancer and now shown to be associated with increased future risk of developing breast cancer assessed by Gail score. This “active” transcriptome phenotype is characterized by increased endogenous IGF-1 activity, a known breast cancer promoting growth factor, along with an inverse reduction in IGF1R expression and activity as previously seen in some breast cancers. Further morphologic and molecular characterization of this risk-associated subset of normal breast tissues is underway. Citation Format: Benz C, Yau C, Benz S, Krings G, Powell M. Normal breast biopsies reveal an "active" transcriptome associating with higher breast cancer risk (Gail) scores and increased IGF-1 growth factor expression [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P3-10-05.

Background: Tumor microenvironment (TME) characteristics are gaining acceptance as important biomarkers across all subtypes of breast cancer (BC). For example, stromal Tumor Infiltrating Lymphocytes (sTILs) have been demonstrated to be predictive in triple-negative and HER2+ BC on immune therapies (Hudeček, 2020), and TILs have been associated with recurrence-risk in HER2- BC (Kolberg-Liedtke, 2020). However, standardizing TME biomarkers remains a challenge to these prognostic & predictive studies (Kos, 2020). Methods: Diagnostic H&E-stained pathology images from 506 HER2- breast cancer patients were acquired from TCGA sources. Pre-trained convolutional neural networks were used to classify each 100μm2 region as containing tumoral, stromal, and lymphocyte-infiltrated tissue as well as map their spatial co-distribution. Nine TME summary features were derived from these spatial maps, including total lymphocyte area, tumor-infiltrating lymphocytes (iTILs), stromal-infiltrating lymphocytes (sTILs), and tumor-adjacent lymphocytes (aTILs). Prognostic models relating these TME features to risk were fitted using Cox multiple-regression trained on 60% of patients and tested in the remaining 40%. Additional Cox models that incorporate seven standard clinicopathological features such as TNM staging, age, ethnicity, treatment type, and hormone-receptor status were also analyzed to establish independence of the TME features. Results: A prognostic model developed using 9 TME-summarizing features accurately stratified unseen patients (HR=0.67 p=0.002) into high-risk (N=95) and low-risk (N=107) categories. Interestingly, tumor-adjacent stroma was significantly associated with higher risk (proportional HR=1.02, p=0.005) whereas tumor-infiltrating stroma was associated with lower risk (proportional HR=0.97, p=0.02). Incorporating standard clinicopathological features increased prognostic performance in test patients (HR=0.63, p=0.0005). The prognostic effect of tumor-adjacent and tumor-infiltrating stroma remained significant in multiple regression with clinicopathological features (p=0.003 & p=0.02 respectively). Conclusions: Here we present using machine-vision to automate and standardize describing the TME, as well as demonstrate the prognostic potential of these descriptions by successfully stratifying risk in HER2- breast cancer. These TME descriptions add independent prognostic power to standard clinicopathological features. Machine-vision tools that produce interpretable features such as this can inform current pathology practices, as well as provide facile and scalable biomarkers for clinical studies going forward. Citation Format: Mustafa I. Jaber, Liudmila Beziaeva, Stephen C. Benz, Shahrooz Rabizadeh, Patrick Soon-Shiong, Christopher W. Szeto. Deep-learning image-based features stratify risk in HER2- breast cancer patients [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 3171.

Background Sebaceous carcinoma (SeC) is an uncommon malignancy arising from sebaceous glands of the conjunctiva and skin. Recurrent mutations in the ZNF750 were recently identified in ocular SeC. We assessed whether ZNF750 loss is a specific feature of ocular SeC or a general feature of sebaceous tumors. Methods Immunostaining for ZNF750 expression was performed in 54 benign and malignant sebocytic proliferations. Staining for ZNF750 was scored on a three-tier scale: positive (>75%), partially positive (5%-74%), and negative ( Results ZNF750 expression was negative in 4/11 ocular SeC, and partially positive in 4/11 ocular SeC and 6/13 cutaneous SeC. No extraocular tumors were negative. No loss was found in sebaceous adenoma or sebaceous hyperplasia. In nine previously sequenced ocular SeCs, two lacked detectable somatic mutations in ZNF750, but showed complete loss of staining, indicating non-mutational inactivation of ZNF750. Conclusion We show complete loss of the ZNF750 epidermal differentiation regulator in about half of ocular SeC, highlighting the most common genetic defect in this cancer type. Loss of ZNF750 expression is seen even in tumors without truncating mutations and reduced in many of the remaining ocular and cutaneous SeC. In contrast, no ZNF750 loss was detected in benign sebaceous proliferations.

Progressive tumor growth is associated with deficits in the immunity generated against tumor antigens. Vaccines targeting tumor neoepitopes have the potential to address qualitative defects; however, additional mechanisms of immune failure may underlie tumor progression. In such cases, patients would benefit from additional immune-oncology agents targeting potential mechanisms of immune failure. This study explores the identification of neoepitopes in the MC38 colon carcinoma model by comparison of tumor to normal DNA and tumor RNA sequencing technology, as well as neoepitope delivery by both peptide- and adenovirus-based vaccination strategies. To improve antitumor efficacies, we combined the vaccine with a group of rationally selected immune-oncology agents. We utilized an IL15 superagonist to enhance the development of antigen-specific immunity initiated by the neoepitope vaccine, PD-L1 blockade to reduce tumor immunosuppression, and a tumor-targeted IL12 molecule to facilitate T-cell function within the tumor microenvironment. Analysis of tumor-infiltrating leukocytes demonstrated this multifaceted treatment regimen was required to promote the influx of CD8+ T cells and enhance the expression of transcripts relating to T-cell activation/effector function. Tumor-targeted IL12 resulted in a marked increase in clonality of T-cell repertoire infiltrating the tumor, which when sculpted with the addition of either a peptide or adenoviral neoepitope vaccine promoted efficient tumor clearance. In addition, the neoepitope vaccine induced the spread of immunity to neoepitopes expressed by the tumor but not contained within the vaccine. These results demonstrate the importance of combining neoepitope-targeting vaccines with a multifaceted treatment regimen to generate effective antitumor immunity.

Background: Stromal and lymphocyte density have each been implicated in differential survival in pancreatic cancer (Torphy, 2018 & Orhan, 2020). In this study, we developed an automated deep-learning system to provide risk-assessment upon spatial relationships between tumor, stroma, and lymphocyte regions in pancreatic pathology images. Methods: Diagnostic H&E-stained pathology images from 82 pancreatic adenocarcinoma patients who underwent chemotherapy were acquired from TCGA sources. Thirty-two patients were held out for testing purposes. Tumor, stroma, and lymphocytes image masks were generated using pre-trained convolutional neural networks, and their co-distribution was summarized in nine numerical image-based features. Optimal thresholds in these image-based features were identified using 2-way Gaussian mixture models. This process found four spatial image features that significantly contributed to low-risk of early death: Low lymphocyte count, lymphocytes adjacent to tumor regions, and stromal adjacency to tumor regions. Ability to separate patients based on these features was evaluated using silhouette score, concordance index, and Cox proportional hazards ratios (HR). Results: Without using image-based features, exhaustive search of this cohort’s clinicopathological features found an optimal Cox proportional hazards model can yield a HR = 0.22 (p = 0.02) in 50 training examples and HR = 0.41 (p = 0.29) on 32 unseen test patients, ultimately utilizing just pathologically-determined T, and N information. The developed image-based risk predictor improved performance with HR = 0.51 (p = 0.06) on training data and HR = 0.52 (p = 0.09) on unseen test data. Combining the image-based risk models to selected clinicopathological features enhanced performance further to HR = 0.25 (p = 0.01) on the training set and HR = 0.37 (p = 0.07) on unseen test patients. Conclusions: Our interpretable image-based risk predictor shows high-risk pancreatic cancer patients have higher lymphocyte count overall but proportionally fewer tumor-infiltrating lymphocytes (TILs). In addition, this system shows high-risk patients have less stromal tissue within 100um from tumor compared to low risk patients. By aggregating both standard clinicopathological features with the proposed image-based risk assessment, superior separation in survival curves was achieved for both training and testing sets compared to either risk-model alone. Thus, our study demonstrates that image-based risk-associated features are independently prognostic of clinicopathological features. Despite the very limited sample-size of similarly-treated patients within the training dataset, these results trend towards significance and warrant further study within a larger cohort. Citation Format: Mustafa I. Jaber, Liudmila Beziaeva, Robert J. Torphy, Stephen C. Benz, Shahrooz Rabizadeh, Patrick Soon-Shiong, Christopher W Szeto. Deep-learning image-based tumor, stroma, and lymphocytes spatial relationships and clinicopathological features that affect survival in pancreatic cancer patients [abstract]. In: Proceedings of the AACR Virtual Special Conference on Pancreatic Cancer; 2020 Sep 29-30. Philadelphia (PA): AACR; Cancer Res 2020;80(22 Suppl):Abstract nr PO-030.

3578 Background: Pathologist inspection of biopsy slides is the gold-standard for diagnosis and is crucial for effective therapy decisions. However, expert shortage is resulting in turnaround times exceeding College of American Pathologists’ (CAP) standards (Alshieban, 2015). Further, discrepancy between diagnoses can exceed 4% (Mukhopadhyay, 2018), and 2% of cases are designated as ‘carcinoma of unknown primary’ (CUP) negatively affect outcomes due to difficulty selecting therapies (Rassy, 2020). Here we sought to aid in diagnosing patients from whole-slide images (WSIs) using deep neural networks. Methods: > 6.3K high-resolution H&E-stained diagnostic WSI of formalin-fixed paraffin-embedded (FFPE) tumor block slices were selected from TCGA sources. Slide images were obtained from 30 different cancer subtypes including 368 Breast (5.6%), 324 Colon (5.12%), 287 Lung Adenocarcinoma (LUAD) (4.5%), Lung Squamous-Cell carcinoma (LUSC) (4.5%), and Stomach Adenocarcinoma (4.3%). Local regions containing tumor tissue were automatically identified by training an Inception V3 deep-learning network as previously presented. A separate Inception V3 network was trained to classify the primary tissue of 200mm2 tumor regions in 60% of the images, which was validated in the remaining 40% testing cohort. Results: The proposed deep-learning model was 92.7% accurate in identifying the primary tissue within the test set of WSIs. As expected, most misclassification occurred in highly-related tissue-types: Rectal cancers misclassified as colon (25%) and vice versa (4.8%), uveal melanomas misclassified as cutaneous melanomas (18.6%), cholangiocarcinomas as hepatocellular carcinomas (8.6%), and LUSC misclassified as LUAD (6.0%) and vice versa (3.4%). Combining related tissues, the classifier achieves 94.6% accuracy across 24 primary types. Unexpectedly, cutaneous melanomas samples were misclassified as breast (9.1%) and LUSC (5.6%), suggestive of related molecular phenotypes. Conclusions: By focusing machine-vision attention on tumor regions, the automated system approaches pathologist accuracy. Used in conjunction with molecular profiling, rates of CUP or misdiagnosis can feasibly be minimized to improve patient care. This system is currently being validated in an external set of > 4K unselected clinical cases from the NantHealth database.

3131 Background: Tumor mutation burden (TMB) and PD1/L1 expression are independent biomarkers for immune checkpoint blockade therapy, as seen in the Checkmate227 trial. Here we explore whether chemokine activity, an intermediate step between neoantigen presentation and immune-infiltration, can resolve this lack of association between existing biomarkers. We further use this novel biomarker to corroborate recent findings from Crowther et al. for a role of MHC class 1-related gene ( MR1) downregulation in immune evasion. Methods: 1,395 clinical samples from the NantHealth database with matched tumor:normal whole exomes and deep whole-transcriptomic sequencing (> 200M reads) were available for analysis. The most common indications in the cohort were Breast (18%), Colon (9.8%), Lung (7.8%), Soft-tissue/Sarcomas (7.7%), and Pancreatic (6.1%). TMB was calculated by counting non-synonymous exonic mutations as per Rizvi, 2015. Immune-infiltration and chemokine signaling were inferred from RNAseq expression of published immune-cell-specific genesets (Bindea, 2013) and chemokine ligands (Nagarsheth, 2017) respectively. Significant associations between TMB, chemokine activity, immune-infiltration, and checkpoint expression were analyzed by ranksums test and corrected for multiple-hypothesis testing using Benjamini-Hochberg adjustment. Results: As expected, TMB and PD1/L1 mRNA expression were not correlated in this cohort (r = 0.08 and r = 0.07 respectively). 36.3% of patients classified as highly immune-infiltrated by unsupervised clustering of immune-cell scores, and this subgroup significantly overexpressed all 11 targetable checkpoint genes analyzed including PD1, PDL1, CTLA4, IDO1, and VISTA (adj. p 9.7e-68 to 4e-168). There was no association between immune-infiltrated samples and TMB (t = 0.9, p = 0.35). Twice as many patients classified as chemokine-active (70.0%) and there was significant agreement between immune-infiltrated and chemokine-active patients (OR = 34.8, p = 6.5e-81). Interestingly, there was a weak but significant association between high chemokine-activity and increased TMB (t = 3.3, p = 0.001). Within patients that were chemokine-active but lacked immune-infiltration, MR1 expression was significantly depleted (t = -10.7, p = 1e-26). Conclusions: Chemokine signatures can help resolve discordance between TMB and checkpoint expression. Analysis of discordance between chemokine-active but immune-depleted tumors may aid in identifying targets for converting from cold to hot tumor microenvironments.

Adipocytes are the predominant cell population in the normal breast and while recent attention has pointed to adipocyte-tumor cell crosstalk as a driver of breast cancer biology there have been few reports on the potential role of adipocytes in driving breast cancer initiation. Because normal breast tissue studies have invariably used reduction mammoplasty, benign biopsy or cancer-adjacent tissues, we studied random breast core biopsy samples donated by 145 healthy, parous, non-obese, white women (median age = 45, range 27-66 y) without any history of breast cancer. Using questionnaire data to calculate future breast cancer risk (Gail scores), we compared digitized microscopic breast tissue (H&E) images with whole genome transcriptome profiling (RNAseq) from FFPE-extracted RNA. We used unsupervised hierarchical clustering of 1487 genes (normalized, median centered, log2-scaled RSEM values) to identify 32% of normal samples with an “Active” (vs. 68% “Inactive”) transcriptome phenotype previously associated with later-life risk of death from breast cancer. Despite slightly lower BMI values, donors with the Active transcriptome phenotype showed significantly higher Gail scores as well as higher mammary adipocyte nuclei counts (median 80% vs. 60%, p=2.3e-6). Tissue resident leukocytes were uncommon but Active transcriptome tissues expressed significantly altered immune modules enriched in TGFβ, interferon and macrophage gene signatures (including single gene increase in CD68) and depleted (relative to Inactive samples) of CD8+ T-cell and serum response/inflammation/wound healing signatures. Active samples were not enriched in cell senescence, SASP or DNA damage response gene signatures but were enriched in an autophagy-to-senescence-transition (AST) signature with increased CAV1 (caveolin-1, p=2.7e-12) and BNIP3 (Bcl2 interacting protein-3, p=4.7e-05) expression, genes that also regulate lipoprotein digestion/mobilization and adipocyte remodeling. Strongest among significant associations linking Active with adipocyte-enriched normal breast samples were increases in two adipokine growth factors, IGF-1 (p=2.2e-16) and FGF2 (p=3.0e-11), the adipokine (resistin) receptor CAP1 (p=0.04) recently linked to poor breast cancer outcomes, and a cAMP-dependent pro-lipolytic signature (p=0.01) known to drive breast cancer progression which, in these samples, correlated positively with average adipocyte area values. Altogether, the collective histologic and molecular features characterizing the normal breast tissue of >30% of healthy parous and non-obese women with increased predicted breast cancer risk seem to implicate a dysregulated mammary adipocyte microenvironment similar to but distinct from that associated with established breast tumors, that precedes microscopic and clinical evidence of breast tumorigenesis. Citation Format: Taekyu Kang, Christina Yau, Stephen Benz, Gregor Krings, Roman Camarda, Jill E. Henry, Mark Powell, Christopher C. Benz. Normal breast tissue at risk for cancer development: A breast cancer initiating role for mammary adipocytes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3314.

528 Background: Dysregulation of DNA mismatch repair pathway can lead to microsatellite instability in many GI tumors, and microsatellite instability is an important diagnostic and prognostic marker. Microsatellite instable (MSI) tumors comprise about 15% of colorectal malignancies and can be found in other gastrointestinal (GI) tumor types. We present results of analysis of genomic and immune infiltration differences between MSI and microsatellite stable (MSS) GI tumors spanning multiple cancer types. Methods: A total of 521 GI patients with deep whole exome sequencing (WES) of tumor and blood samples, and whole transcriptomic sequencing (RNA-Seq) (∼200M reads per tumor) were available for this analysis from a commercial database. Variant calling was performed through joint probabilistic analysis of tumor and normal DNA reads, with germline status of variants being determined by heterozygous or homozygous alternate allele fraction in the germline sample. Results: Gene expression and pathway analysis found significantly higher immune signaling in MSI cohort and higher metabolic signaling in MSS cohort. We also found upregulation of structural cellular integrity pathways in MSI tumors. Per-sample deconvolution of immune infiltration using cell type gene markers shows some MSI samples with high CD8 T-cells. Co-expression analysis of checkpoint and TME genes shows higher correlation of FOXP3 and CTLA4 in the MSS cohort compared to the MSI samples, whereas correlation between FOXP3 and PDL1 is decreased. TIM3, LAG3, and OX40 are significantly more expressed in MSI samples than MSS samples. Within the subset of colorectal tumors, additional checkpoints are significantly differentially overexpressed in MSI malignancies. 50 somatic variants are significantly differential in MSI tumors. Conclusions: MSI tumors demonstrably exhibit higher immune signaling, with many immune and checkpoint markers expressed at higher levels in MSI tumors. Some cellular integrity pathways also appear to be up in MSI cohort. A number of potentially important somatic variants are associated with MSI samples.

610 Background: Tumor mutational burden (TMB) is emerging as an important biomarker for immune checkpoint therapy (ICT) response. Yet even in the context of high TMB, ICT are likely ineffective in an immuno-suppressed microenvironment. Here we demonstrate that a well-characterized subtype of CRC, CMS2, associated with Wnt pathway activation, is immunosuppressive despite high TMB. Methods: Tumor/normal-paired DNAseq (WGS or WES) and deep RNAseq (∼200x106reads per tumor) was performed on 464 GI tumors from a commercial database. Samples were classified as high TMB if they had > 200 non-synonymous exonic mutations as previously established (Rizvi et al, 2015). Each sample was assigned to one of the colorectal Consensus Molecular subtypes (CMS) based on RNA classification. A curated panel of 109 genes that discriminate between 22 immune subsets was identified. For each of these immune signatures, a database containing 1880 unselected tumors was used to define a distribution of expression. The study samples were then scored for their deviances within such distributions. Significant enrichment was analyzed between immune subsets, CMS types, TMB status, and somatic mutational status. Results: Compared to other subtypes, CMS1 & CMS2 were significantly high-TMB (adj. p < 3.8E-4 & p < 4.7E-3 respectively). Perplexingly, CMS2 had significantly lower expression of 11 well-established checkpoint and TME markers including LAG3 and PDL1 (adj. p 1.5E-2 & 2.9E-9 respectively), while CMS1 (MSI-enriched) expresses selected TME markers more than other subtypes (PDL1 adj. p < 4.0E-6 & LAG3 adj. p < 1.0E-6). As expected, CMS2 tumors were significantly enriched for likely pathogenic variants in the Wnt-associated gene APC (adj. p < 1.3E-8). Immune-deconvolution indicated substantial exclusion of Tem cells from CMS2 tumors, in line with Wnt/b-catenin blockade of TcmàTem maturation for immunoreactivity. Conclusions: The most common subtype of CRC, CMS2 (~37%), is highly immunosuppressive despite high TMB. ICT is only effective in an immunologically active microenvironment. TMB alone as a biomarker likely is insufficient to indicate the effectiveness of immunotherapy.

579 Background: Immune contexture shapes oncogenesis, tumor progression and response to therapy. Traditional techniques for studying the microenvironment, such as flow cytometry, are limited by low throughput. Recently, computational methods using transcriptomic data have enabled deconvolution of tumor immune landscape, particularly in colorectal cancer (CRC). However, the immune profiles of other GI and HN cancers remain unclear. We proposed that elucidating the immune compositions in those cancers would reveal biological insights and potentially therapeutic targets. Methods: 464 GI tumors with RNA-Seq data (∼200x106reads per tumor) from a commercial database were available for analysis. Tumors were categorized into CRC, gastroesophageal (GE), HN, and biliary cancer. A curated panel of 122 genes that discriminate between 28 immune subsets was identified. For each of these immune signatures, a database containing 1880 unselected tumors was used to define a distribution of expression. The study samples were then scored for their deviances within such distributions.In addition to immune deconvolution, each of the 464 tumors were assigned to one of the colorectal Consensus Molecular subtypes (CMS). Significant enrichment for immune subsets between locations and CMS was analyzed. Results: Unsupervised clustering revealed 2 clusters with distinct immune profiles, which largely separated CRC from HN/GE tumors (silhouette coefficient = 0.14). Eosinophils are more abundant in GI cancers than others. HN/GE tumors were characterized by abundant NK and Tregs (adj. p < 0.001), and were predominantly classified as the immune-active CMS1 (adj. p < 0.001). CRC was significantly associated with high eosinophils and fibroblasts (adj. p < 0.0001). Biliary tumors spanned both immune-type clusters, and were frequently classified as the stromal-rich CMS4 (adj. p < 0.0001). Immature dendritic cells were sparse in GI tumors, especially GE. Conclusions: Upper and lower GI tumors have distinct immune contextures, which may differentially impact the efficacy of immunotherapy. Comprehensive immune profiling may facilitate the identification of targetable immune subsets.

High-throughput data sets such as genome-wide protein–protein interactions, protein–DNA interactions and gene expression data have been published for several model systems, especially for human cancer samples. The University of California, Santa Cruz (UCSC) Interaction Browser (http://sysbio.soe.ucsc.edu/nets) is an online tool for biologists to view high-throughput data sets simultaneously for the analysis of functional relationships between biological entities. Users can access several public interaction networks and functional genomics data sets through the portal as well as upload their own networks and data sets for analysis. Users can navigate through correlative relationships for focused sets of genes belonging to biological pathways using a standard web browser. Using a new visual modality called the CircleMap, multiple ‘omics’ data sets can be viewed simultaneously within the context of curated, predicted, directed and undirected regulatory interactions. The Interaction Browser provides an integrative viewing of biological networks based on the consensus of many observations about genes and their products, which may provide new insights about normal and disease processes not obvious from any isolated data set.

Background: However hormone receptor–positive, early-stage breast cancer is a disease with a long natural history and improved survival with 5-year adjuvant endocrine treatments. Yet late recurrence remains an important issue in adjuvant therapy. Predictors of late recurrence are not yet well characterized. Method: A total of 252 breast primary tumors were selected at the Netherlands Cancer Institute from retrospective series of ER+, HER2− breast cancer patients with a follow-up of at least 10 years. Gene expression analysis was performed using Agilent 4×44K microarrays. In order to identify genes associated to late survival differences, we used the survdiff function implemented in the R package survival and we set the parameter rho to −1 to give weight to the later part of the survival curve. The survdiff function was applied to each probe individually for DMFS time considering the probe as a covariate dichotomized into 2 groups (above and below the median expression across all samples). The parameter “strata” was used to stratify the 140 patients based on additional clinico-pathological parameters (Grade, Diameter, Lymph node status and MammaPrint), in order to find genes that add prognostic value to those parameters already known. This approach uses the distant-metastasis free survival (DMFS) time as a continuous variable. Results: After univariate analysis, MammaPrint, diameter, lymph node status and grade were significantly associated to late DMFS differences (Chi-square test p-values equal to 0.016, 0.004, In order to independently validate the prognostic power of these two genes, we tested their performance in the validation set of treated patients (n = 112) and in three publicly available datasets. In all datasets, the CH25H gene confirmed to be significantly associated to metastasis-free survival time in all tested series. Conclusions: These results might indicate that CH25H is an independent marker of late metastatic relapses. CH25H catalyzes the formation of 25-hydroxycholesterol from cholesterol, leading to repress cholesterol biosynthetic enzymes. In the last years, it is emerging that lipid metabolism plays an important role in breast cancer development and progression. Taken together, these findings make the CH25H gene a potential target for late metastasis control in breast cancer. These results warrant further prospective investigation and functional characterization of CH25H in this setting. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-09-05.

Background Previous reports have described the use of microarrays to assess the molecular classification of human breast cancers and defined new subgroups based on gene expression that are relevant to patient management through their ability to predict metastatic relapse and survival relapse. However, different mechanisms may be associated with the development of early and late distant metastases. With the hypothesis that tumors may lead to early or distant metastases based on their intrinsic biological initial features, we aimed at defining molecular profiles for several subgroups of patients based on their outcome over time. Material and methods Breast primary tumors were selected from retrospective series of patients with frozen material available. These series include patients of all ages, LN- and LN+; Estrogen or Progesteron-receptor positive, Her2-negative, no adjuvant treatment, with a follow-up of more than 10 years (y) for the control group or distant metastatic relapse as first event (DM) for the study group (n=144). Patients tumors were classified in 4 groups: no relapse at 10 y (M0), DM before 3 y (M0-3, n=30), DM between 3 and 7 y (M3-7), DM after 7 y (M7+). Samples were collected in 2 different institutions (NKI series for identifying the signature and IGR series for validation). Gene expression analysis of breast tumor samples was performed using custom-made Agilent 44K high-density microarrays and hybridized against the Mammaprint® reference pool (MRP). Tumors were also assessed for their Mammaprint® status, wound-healing signature status and their intrinsic subtypes based on the Blueprint® signature. Moreover, we identified the pathway-level activities of the patient groups using PARADIGM. Results and Discussion For the NKI series, A subset of 144 samples was included based on the selection criteria: 57 M0, 31 M0-3, 25 M3-7, 31 M7+. None of the 3 previously mentioned signatures correctly identified M0 vs. M7+ patients. In order to identify a predictive signature of late relapse (after 7y) we considered M0 and M7+ MammaPrint-Low Risk patients and we split them in a training (n=41) and in a test (n=23) sets. A 73-gene signature was able to classify M7+ patients with 75% of sensitivity and 66% of specificity on the test set. DM after 7yr showed significant activation of pathway related to inflammatory response and angiogenesis. Detailed results and validation results on the independent IGR series will be presented at the meeting. Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr S1-6.

e15597Background: Responses to anti-PD1 is ~35% in microsatellite unstable (MSI-h) and near 0% in microsatellite stable (MSS) Colorectal cancer (CRC). Understanding the composition of immune cell p...

12113Background: Targeting immune checkpoints has led to clinical benefit across a variety of tumor types, and employing combinations has enhanced response rates even further. We hypothesize that p...

12117Background: Next generation sequencing (NGS) Gene panel testing is used to imputed tumor mutational burden (iTMB) and has shown rough correlation with TMB derived from whole exome sequencing (...

6053Background: Adenoid cystic carcinoma (ACC) is a rare cancer of secretory glands accounting for 10% of salivary gland cancers and 1% of head and neck cancers. ACC rarely responds to chemotherapy...

e15663Background: Detection of microsatellite instability (MSI) status has become a vitally important biomarker for the use of the checkpoint inhibitor immuno-oncology. Traditional MSI detection re...

e12643Background: Around 40% of HER2-positive (HER2+) breast cancer patients who receive trastuzumab-based neoadjuvant therapy (TNT) achieve pCR (the FDA-recommended outcome measure in this setting...

574Background: Different mutational processes act over the evolutionary history of a malignant tumor, driven by e.g. abnormal DNA editing, mutagens or age-related DNA alterations. Many of these pro...

572 Background: Comprehensive molecular profiling of CRC can inform treatment decisions by identifying patient subgroups at varying risks of death. Microsatellite instability (MSI) is prognostic in CRC and is used to select patients for immunotherapy. High tumor mutational burden (TMB) is associated with genomic instability and is prognostic in melanoma. Expression of p16 protein is prognostic in many tumor types. We used proteomic and genomic profiling to measure MSI, TMB and p16 in CRC tumors and to assess associations with patient survival. Methods: In archived clinical samples of CRC, 76 proteins were quantitated with mass spectrometry-based proteomics. MSI was measured by WGS and RNA-seq; unstable loci were quantified in tumor and normal samples. Cutoffs were derived via ROC analysis: high TMB was defined as > 4.5 somatic mutations per megabase; p16 as > 108 amol/ug. Patients were grouped by microsatellite status (MSI vs. microsatellite stable [MSS]), TMB (high vs. low), and p16 protein expression level. Survival curves were compared with the Mantel-Cox log-rank test. Results: Of 145 samples, 39 (27%) had high TMB and 29 (20%) had MSI. Patients with MSI tumors had longer OS than patients with MSS tumors (HR: 0.096; p = 0.003). Similarly, patients with high TMB had longer OS than those with low TMB (HR: 0.076; p < 0.001). High p16 expression was prognostic of poor survival (HR: 2.874; p = 0.019). Among patients with MSS tumors or low TMB, those with low p16 levels had longer OS than patients with high p16 (HR: 0.257; p = 0.002 and HR: 0.249; p = 0.002, for MSS and low TMB, respectively). A combination of MSS, low TMB, and low p16 also differentiated between long and short survivors (HR: 0.249; p = 0.002). These associations remained after adjustment for tumor sidedness. Further analyses of clinical correlates will be presented. Conclusions: A combination of MSS, low TMB and low p16 expression characterized a subset of patients with longer survival. This is important because patients with MSS tumors have limited treatment options but may respond to CDK4/6 inhibitors due to low p16 expression. Molecular profiling of CRC may identify patient subgroups with a relatively poor prognosis who could benefit from personalized therapy.

Introduction Breast cancer risk of recurrence is known to span 20 years, yet existing prognostic signatures are best at predicting early recurrences (≤5 years). There is a critical need to identify those patients at risk of late-relapse (>5 years), in order to select potential candidates for further treatment and to identify molecular targets for such treatment. Methods A total of 252 breast primary tumors were selected at the Netherlands Cancer Institute from a retrospective series of ER+, HER2− breast cancer patients with a follow-up of at least 10 years. Gene expression analysis was performed using Agilent 4x44K microarrays. Patients were classified in 3 groups: no relapse (M0); relapse before 5 years (M0-5) or after 5 years (M5-15). We assessed the correlation of clinico-pathological variables with late Distant Metastases (DM). We divided the patient series into a training set of untreated patients ( n = 140) and a test set of treated patients ( n = 112), to investigate whether a gene-signature or single genes could be identified for predicting late DM. Pathway level late DM correlates were identified using PARADIGM and DAVID. Results Of the clinico-pathologic variables tested, only lymph node status associated with late DM. A 241-gene signature developed on the NKI training set was able to classify M5-15 patients in the test set with a sensitivity of 77% and a specificity of 33% (AUC 0.654). This signature showed enrichment in genes involved in immune response and extracellular matrix. An alternative analysis of individual genes identified CH25H as an independent predictor of distant metastasis in our patient series. Conclusions We identified a gene signature for late metastasis in breast cancer. Our data are consistent with a model in which suppressed anti-tumoral immunity enables dormant tumor cells to re-enter the cell cycle to form metastases in response to extrinsic events in the microenvironment.

11606Background: Anti-HER2 therapies have demonstrated success in improving the outcomes of patients (pts) with HER2-positive breast cancer; however, a high proportion of pts either do not respond ...

Introduction: Selection of drugs to treat patients with cancer is typically based on the anatomical site in which the tumor is located. Here we report a treatment decision for a patient with relapsed, advanced cervical cancer that was based on a comprehensive omics analysis using whole genome sequencing (WGS) combined with quantitative proteomics. Methods: The patient was a 44-year-old female whose disease had progressed following surgery and more than 4 lines of chemotherapy. WGS was performed on the patient's formalin-fixed, paraffin-embedded (FFPE) metastatic tumor sample and a matched-normal reference sample. Quantitative proteomics was performed on the FFPE tumor sample by Selected Reaction Monitoring Mass Spectrometry and was quantitated at the atomolar level. Results: WGS found somatic mutations and rearrangements and reads mapping to human papillomavirus type 18 (HPV 18). Mutations more commonly found in breast cancer (ERBB2, CDH1, and CLTCL1) were noted. The HPV 18 genome was integrated into chromosome 17 in close proximity to a 7-fold amplification of the ERBB2 gene. Proteomic analysis of the FFPE tumor validated and quantitated overexpression of HER2 protein resulting from ERBB2 gene amplification, with 11,322 amol/μg of tissue protein. Clinically observed ranges for breast or gastric cancer are 150-500 amol/μg, with levels above 750 amol/μg correlating with FISH-positive amplification and clinical efficacy of trastuzumab (unpublished observation). Based on these comprehensive omic findings, trastuzumab, a therapy approved for breast and gastric cancer, was administered. The patient experienced a reduction in the size of her tumor (by CT/PET) and stabilization of her disease for 5 months. Conclusion: WGS and proteomic profiling of this patient's disease identified, confirmed, and quantitated an appropriate target for pharmaceutical intervention. The patient presented with cervical cancer; however, the WGS analysis pointed towards a potentially causative integration of the HPV 18 genome resulting in ERBB2 amplification along with genomic mutations more commonly found in breast cancer. Proteomic analysis further validated and quantitated the HER2 expression resulting from ERBB2 gene amplification, leading to the patient's treatment with trastuzumab. Our findings argue for the use of comprehensive omics analysis to guide decision support for personalized management of cancer care with therapies determined based on a quantitative proteomic signature, independent of anatomical tumor type. Citation Format: Stephen Benz, J Zackary Sanborn, Nicole S. Hensley, Todd Hembrough, Charles J. Vaske, Jon Burrows, Shahrooz Rabizadeh, Ivor Royston, Patrick Soon-Shiong. Whole genome sequencing and quantitative proteomics reveal HPV integration and HER2 overexpression in a patient with cervical cancer: Comprehensive omics analysis driving clinical treatment decisions. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Integrating Clinical Genomics and Cancer Therapy; Jun 13-16, 2015; Salt Lake City, UT. Philadelphia (PA): AACR; Clin Cancer Res 2016;22(1_Suppl):Abstract nr 25.

11005 Background: Treatment decision support by next generation sequencing of gene panels is currently limited to the analysis of somatic (tumor) data from DNA sequencing without taking into consid...

Statins are cholesterol-lowering drugs with pleiotropic activities including inhibition of isoprenylation reactions and reduction of signals driving cell proliferation and survival responses. The objectives of this study were to examine the effects of statins on breast cancer cells, both in vitro and in vivo, and to begin to determine their mechanism of action. We evaluated the effects of statins on breast cancer cell growth, phosphoprotein signaling intermediates, survival/apoptosis regulators, cell cycle regulators, and activated transcription factors. We also examined the in vivo effect of statin administration in a mouse ErbB2+ breast cancer model. Only lipophilic statins had direct anticancer activity in vitro. Breast cancer cells with activated Ras or ErbB2 pathways seemed to be more sensitive than those overexpressing estrogen receptor, and this correlated with endogenous levels of activated nuclear factor κB (NF-κB). Key intermediates regulating cell survival by NF-κB activation, as well as cell proliferation by the mitogen activated protein kinase cascade, were among the earliest phosphoproteins influenced by statin treatment. These early effects were followed by declines in activator protein-1 and NF-κB activation and concordant changes in other mediators of proliferation and apoptosis. In vivo results showed that oral dosing of statins significantly inhibited the growth of a mouse mammary carcinoma. Lipophilic statins can exert direct anticancer activity in vitro by reducing proliferation and survival signals in susceptible breast cancer phenotypes. Tumor growth inhibition in vivo using a clinically relevant statin dose also seems to be associated with reduced tumor cell proliferation and survival. These findings provide supporting rationale for future statin trials in breast cancer patients. (Cancer Res 2006; 66(17): 8707-13)

Background: TN and BRCA1-deficient breast tumors share clinicopathologic characteristics and are highly sensitive to DNA-damaging agents. This single-arm phase II study was designed to assess efficacy and safety of iniparib, whose mechanism of action is under investigation, in combination with GC in early-stage triple-negative and BRCA1/2 mutation-associated breast cancer. Our objective was to use multi-omic profiling of breast tumors to identify markers associated with therapy response. Methods: Fresh-frozen breast tumor core biopsies were collected under ultrasound guidance (n=75 tumors from 74 patients (one patient with bilateral tumors)) prior to treatment. Tumor RNA and DNA was extracted, as well as matched peripheral whole blood DNA for germline comparison. Affymetrix U133 plus 2.0 array was used for whole genome expression analysis and Affymetrix Molecular Inversion Probe (MIP) was used for copy number variation analysis on 75 tumors. After normalization, we used the PARADIGM algorithm, a pathway-based approach for combining multi-omics data types on individual tumors. Results: We identified pathway features that correspond to pathologic response, assessed using the residual cancer burden (RCB) index. All patients were tested for BRCA germline mutation and 16 (22%) patients were positive. Overall, a low RCB (0,1) was observed in 43 (58%) patients and a high RCB (2,3) in 31 (42%) patients. Pathway-based analysis of these breast tumors revealed significant alterations in 459 of 1466 (31%) of the cancer-related pathways. Gene set enrichment analysis on the PARADIGM results identified differentially expressed pathways in tumors with RCB 0 (65 pathways at p Conclusions: Using a pathway-based approach, predictors of pathologic response to GC plus iniparib were identified from pre-treatment tumor biopsies. Tumors with complete pathologic response were enriched for immune-related pathways, including interleukin signaling pathways. Least responsive tumors to this therapy were enriched for NOTCH and WNT signaling pathways. We plan to validate these findings in external datasets. Platinum-based therapies are under clinical investigation for TN and BRCA1/2-associated breast cancer and identification of response predictors can guide patient selection. Citation Format: Shaveta Vinayak, Stephen C. Benz, Charles J. Vaske, Melinda L. Telli, James M. Ford. Multi-omic profiling to predict response to gemcitabine/ carboplatin (GC) plus iniparib (BSI-201) as neoadjuvant therapy for triple-negative (TN) and BRCA1/2 mutation-associated breast cancer using a pathway-based approach. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2826. doi:10.1158/1538-7445.AM2014-2826

One goal of characterizing the genome-wide landscape of cancer cells is to identify predictive signatures of onset, progression, and treatment outcomes. Many computational approaches have been developed to discover gene signatures with a range of success. The challenge still remains to identify the best approach that, when trained on one cohort, remains accurate in predicting outcomes on an unseen cohort. Thus far, no clear themes have emerged that might provide clues about which method works for a particular task. We have built a system called TopModel that facilitates the identification of top-performing machine-learning algorithms for a series of cancer-genomics challenges. The four components of the system include: 1) a benchmark that includes several cancer genomics datasets with outcome variables as targets to predict; 2) a database of results derived from the application of thousands of machine-learning and feature selection combinations; 3) a web interface that allows bioinformaticians to evaluate their own prediction results; and 4) a web interface that allows a biomedical researcher to upload data on a sample or set of samples in order to receive a report on the signatures predicted to exist in the sample(s). The cancer benchmark component provides a common ground for the development and evaluation of prediction methods for variables such as cancer subtype, drug response, survival, and others. Several datasets have been loaded including predicting survival in the TCGA cohorts, and the hundreds of drug sensitivities in several cancer cell line cohorts. We demonstrate the utility of the resource by comparing state-of-the art feature selection methods to a new approach that uses locality on a genetic interaction network. We evaluate the performance in terms of how well the features generalize across datasets as a trade-off to the accuracy of prediction. In addition to identifying high-value genome features, we explore the robustness of the cancer state in the absence of these features. We simulate gene knock outs by disconnecting these features in our pathway models, inferring the pathway interaction network in the absence of these features, and then reassessing using the top-performing predictive models of cancer. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 5085. doi:1538-7445.AM2012-5085