Abstract Background: Better understanding of prognostic factors in tubo-ovarian high-grade serous carcinoma (HGSC) is critical, as diagnosis confers an aggressive disease course. Variation in tumor DNA methylation shows promise predicting outcome, yet prior studies were largely platform-specific and unable to evaluate multiple molecular features. Methods: We analyzed genome-wide DNA methylation in 1,040 frozen HGSC, including 325 previously reported upon, seeking a multi-platform quantitative methylation signature that we evaluated in relation to clinical features, tumor characteristics, time to recurrence/death, extent of CD8+ tumor-infiltrating lymphocytes (TIL), gene expression molecular subtypes, and gene expression of the ATP-binding cassette transporter TAP1. Results: Methylation signature was associated with shorter time to recurrence, independent of clinical factors (N = 715 new set, hazard ratio (HR), 1.65; 95% confidence interval (CI), 1.10–2.46; P = 0.015; N = 325 published set HR, 2.87; 95% CI, 2.17–3.81; P = 2.2 × 10−13) and remained prognostic after adjustment for gene expression molecular subtype and TAP1 expression (N = 599; HR, 2.22; 95% CI, 1.66–2.95; P = 4.1 × 10−8). Methylation signature was inversely related to CD8+ TIL levels (P = 2.4 × 10−7) and TAP1 expression (P = 0.0011) and was associated with gene expression molecular subtype (P = 5.9 × 10−4) in covariate-adjusted analysis. Conclusions: Multi-center analysis identified a novel quantitative tumor methylation signature of HGSC applicable to numerous commercially available platforms indicative of shorter time to recurrence/death, adjusting for other factors. Along with immune cell composition analysis, these results suggest a role for DNA methylation in the immunosuppressive microenvironment. Impact: This work aids in identification of targetable epigenome processes and stratification of patients for whom tailored treatment may be most beneficial.
Prognostic value of deep learning based mitotic count in hormonal receptor- and HER2-positive breast cancer
Abstract The discovery of molecular subtypes has corroborated the heterogeneity of breast cancers. But in clinical routine, treatment selection relies on measuring the aggressiveness of the tumor via histopathology, which is routinely based on grading, a prognostic yet laborious, poorly reproducible, and subjective procedure defined more than 30 years ago, before molecular subtypes existed. In this paper, we present an interpretable deep learning-based computer-aided diagnosis method to automate mitotic count, one of the components of grading; the method complies with current guidelines and automates hot-spot finding on full digital whole-slide images of routinely prepared pathologic slides stained with hematoxylin and eosin. We use the computer model to query the current “one-size-fits-all” mitotic count procedure by investigating subtype-specific prognostic value of mitotic count using an external independent large-scale multicentric dataset of 1,709 breast cancers. We show that for human epidermal growth factor-2 (HER2) positive tumors, mitotic count was not prognostic. In hormone-receptor positive/HER2 negative cancers, an average mitotic count density different from clinically used cut offs predicted worse recurrence free survival in multivariable analysis, adjusted for known clinical prognosticators. We provided some first experimental evidence that the prognostic role of mitotic count should be reviewed.
Benign Breast Disease, NSAIDs, and Postmenopausal Breast Cancer Risk in the CPS-II Cohort
Abstract Nonsteroidal anti-inflammatory agents (NSAID) are associated with modest inconsistent reductions in breast cancer risk in population-based cohorts, whereas two focused studies of patients with benign breast disease (BBD) have found lower risk with NSAID use. Given that BBD includes fibroinflammatory lesions linked to elevated breast cancer risk, we assessed whether NSAID use was associated with lower breast cancer risk among patients with BBD. Participants were postmenopausal women in the Cancer Prevention Study-II (CPS-II), a prospective study of cancer incidence and mortality, who completed follow-up surveys in 1997 with follow-up through June 30, 2015. History of BBD, NSAID use, and covariate data were updated biennially. This analysis included 23,615 patients with BBD and 36,751 patients with non-BBD, including 3,896 incident breast cancers over an average of 12.72 years of follow-up among participants. NSAID use, overall and by formulation, recency, duration, and pills per month was analyzed versus breast cancer risk overall and by BBD status using multivariable-adjusted Cox models; BBD status and NSAID use were modeled as time-dependent exposures. Patients with BBD who reported using NSAIDs experienced lower breast cancer risk (HR, 0.87; 95% CI, 0.78–0.97), with similar effects for estrogen receptor (ER)-positive breast cancers [HR, 0.85; 95% confidence interval (CI), 0.74–0.97] and ER-negative breast cancers (HR, 0.87; 95% CI, 0.59–1.29); among women without BBD, NSAID use was unrelated to risk (HR, 1.02; 95% CI, 0.92–1.13; Pinteraction = 0.04). Associations stratified by age, obesity, menopausal hormone use, and cardiovascular disease were similar. Among patients with BBD, NSAID use appears linked to lower breast cancer risk. Further studies to assess the value of NSAID use among patients with BBD are warranted. Prevention Relevance: We examined whether NSAID use, a modifiable exposure, is associated with breast cancer risk in postmenopausal women from the Cancer Prevention Study-II with self-reported benign breast disease, an often inflammatory condition associated with higher rates of breast cancer.
DNA barcoded competitive clone-initiating cell analysis reveals novel features of metastatic growth in a cancer xenograft model
Abstract A problematic feature of many human cancers is a lack of understanding of mechanisms controlling organ-specific patterns of metastasis, despite recent progress in identifying many mutations and transcriptional programs shown to confer this potential. To address this gap, we developed a methodology that enables different aspects of the metastatic process to be comprehensively characterized at a clonal resolution. Our approach exploits the application of a computational pipeline to analyze and visualize clonal data obtained from transplant experiments in which a cellular DNA barcoding strategy is used to distinguish the separate clonal contributions of two or more competing cell populations. To illustrate the power of this methodology, we demonstrate its ability to discriminate the metastatic behavior in immunodeficient mice of a well-established human metastatic cancer cell line and its co-transplanted LRRC15 knockdown derivative. We also show how the use of machine learning to quantify clone-initiating cell (CIC) numbers and their subsequent metastatic progeny generated in different sites can reveal previously unknown relationships between different cellular genotypes and their initial sites of implantation with their subsequent respective dissemination patterns. These findings underscore the potential of such combined genomic and computational methodologies to identify new clonally-relevant drivers of site-specific patterns of metastasis.
Immune cell density differences in normal breast lobules of BRCA1/2 mutation carriers, women with benign breast disease and healthy research volunteers
Abstract Purpose Breast cancer risk is elevated in pathogenic germline BRCA 1/2 mutation carriers due to compromised DNA quality control and elevated mutagenic load. We hypothesized that if immunosurveillance promotes tumor suppression, then normal breast lobules from BRCA carriers may demonstrate higher immune cell densities. Methods We assessed immune cell composition in normal breast lobules from age-matched women with progressively increased breast cancer risk, including 1) 19 women who donated normal breast tissue to the Komen Tissue Bank (KTB) at Indiana University Simon Cancer Center, 2) 15 women with biopsy-identified benign breast disease (BBD), and 3) 19 prophylactic mastectomies from women with germline mutations in BRCA1/2 genes. We performed immunohistochemical stains and analysis to quantitate immune cell densities from digital images in up to 10 representative lobules per sample. Median cell counts per mm2 were compared between groups using Wilcoxon rank-sum tests. Results Compared to BBD tissues, normal breast lobules from BRCA carriers had significantly higher densities of CD8 + cells (median 354.4 vs 200.0 cells/mm2, p = 0.01). Compared to KTB normal donor breast tissues, normal breast lobules from BRCA carriers also had significantly higher densities of CD4+ (median 116.3 vs 17.7 cells/mm2), CD8+ (354.4 vs 150.9 cells/mm2), CD11c+ (3.5% vs 0.4% pixels positive), and CD68+ (237.5 vs 57.8 cells/mm2) immune cells (each p < 0.001). Conclusions These preliminary analyses show that normal/benign breast lobules of BRCA mutation carriers contain increased immune cells compared with BBD and normal donated tissues. Future studies will be helpful to define relationships of mutations and immune responses.
Abstract Polycystic ovarian syndrome (PCOS) is a complex health condition associated with metabolic disturbances and infertility. Recent data suggest that the prevalence of PCOS is increasing among women globally, although the etiology of these trends is undefined. Consequently, preclinical models that better reflect the biology of PCOS are urgently needed to facilitate research that can lead to the discovery of prevention strategies or improved management. The existing animal models have several limitations as they do not reflect all the PCOS features metabolically and/or phenotypically. Therefore, there is no clear consensus on the use of appropriate animal model and selection of the most appropriate PCOS-inducing agent. To that end, we have established a Swiss albino mouse model of PCOS based on 3 weeks of daily treatment with letrozole (50 μg/day; intraperitoneal) and dehydroepiandrosterone (DHEA, 6 mg/100 g body weight; subcutaneous) in 5-week-old female mice fed on normal or high-fat diet (HFD). Mice were regularly assessed for body weight, blood glucose, and estrous cycle. Three weeks after drug administration, mice were sacrificed and assessed for blood-based metabolic parameters as well as ovarian function. Our results indicate that DHEA combined with HFD produces changes mimicking those of clinical PCOS, including elevated serum testosterone and luteinizing hormone, dyslipidemia, poor ovarian microenvironment, and development of multiple ovarian cysts, recapitulating cardinal features of PCOS. In comparison, normal diet and/or letrozole produced fewer features of PCOS. The data from the experimental models presented here can improve our understanding of PCOS, a growing concern in women’s health.
Abstract 2203: Towards prediction of breast cancer risk in benign biopsies with high-plex GeoMx spatial protein profiling
Abstract Background: Biopsy diagnoses of benign breast disease (BBD) confer a 1.5- to 4-fold increased risk of developing breast cancer (BC) compared with women without BBD. Previously, we reported that decreased numbers of specific immune cell types in lobules of BBD biopsies predicted increased BC risk, suggesting the promise of future tissue biomarker studies to define BC risk markers among BBD patients. Thus, we applied protein-based GeoMx® Digital Spatial Profiling (DSP) to BBD biopsies preceding BC (cases) and to BBD biopsies from cancer-free patients (controls) to identify possible BC risk markers, which we then evaluated in subsequent BC tissues and in surrounding normal lobules of cases. Methods: Archived pathology slides of BBD biopsies were reviewed and used to guide preparation of TMAs containing 1.0-mm diameter FFPE cores of lobules from an age- and cohort-period-matched set of 91 cases and 88 controls from the Mayo Clinic BBD Cohort. For patients who later developed BC, we prepared TMAs of BC tissue and surrounding mapped normal lobules. We applied GeoMx® DSP (immune and canonical signaling proteins) to both sets of TMAs. Following QC and data normalization, associations of case status with log-transformed biomarker expression in lobules of BBD biopsies were carried out using linear mixed modeling approaches, accounting for multiple ROIs per individual. Biomarkers significantly associated with case status (p<0.05) were further examined in BCs and adjacent normal lobules, using similar approaches. Results: The mean age at BBD biopsy was 52 years, and at BC diagnosis of cases, 61.4 years (mean time from BBD to BC was 10.2 years). A family history of BC was more frequent among cases (70% versus 43%; chi-square p=0.002). Of 72 biomarkers tested, 46 (64.4%) were evaluable after QC and normalization and 5 were associated with BC risk after adjustment for family history of BC: BCL2 (p=0.005), STING (p=0.006), CD44 (p=0.02), S100 protein (p=0.03) and pan-AKT (p=0.05); each showed higher levels in lobules of BBD biopsies of controls than cases. Three unique patterns appeared when examining these biomarkers across tissue type within cases: for BCL2 (p=5 x 10-9) and STING (p=2 x 10-19), levels were high in both BC and lobules surrounding BC but low in preceding BBD; for GAPDH (p=4 x 10-53) and pan-AKT (p=2 x 10-33), levels were high in BC, low in preceding BBD, and moderate in lobules surrounding BC; and for CD44 (p=2 x 10-6) and S100B (p=2 x 10-49), levels were low in BC, high in lobules surrounding BC and moderate in preceding BBD. Conclusions: Using a novel TMA of lobules in combination with DSP, we preliminarily identified immune-based and PI3 kinase-related protein biomarkers in BBD biopsies associated with BC risk. In case-only analyses, these markers demonstrated complex differences between lobules in BBD biopsies, subsequent BCs and adjacent normal lobules. Citation Format: Robert Alan Vierkant, Jodi M. Carter, Stacey J. Winham, Chen Wang, Jennifer M. Kachergus, Ji Shi, Raymond M. Moore, Bryan M. McCauley, Laura M. Pacheco-Spann, E A. Thompson, Derek C. Radisky, Amy C. Degnim, Mark E. Sherman. Towards prediction of breast cancer risk in benign biopsies with high-plex GeoMx spatial protein profiling [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2203.
Abstract 6333: Quantification of oxylipins as biomarkers of inflammation and cancer risk in breast tissue
Abstract Arachidonic acid (AA) metabolizing enzymes, including cyclooxygenases (COX), lipoxygenases, and cytochrome P450s (CYP450), are expressed in breast cancer (BC). Elevated levels of metabolites of COX-derived, pro-inflammatory prostaglandin (PGs) metabolites in urine have been linked to increased risk of developing breast cancer (BC) in two large prospective cohorts. Further, CYP450-derived EETs in BC tumors are associated with metastasis. We hypothesize that measuring levels of AA-derived oxylipins in breast tissues could enable improved BC risk prediction, provide insights into pro-carcinogenic inflammatory mechanisms, and serve as potential surrogate endpoints in chemoprevention trials. Thus, we performed a pilot study to assess the feasibility of measuring these analytes in small breast tissue samples. We analyzed two sample sets, collected under Institutional Review Board (IRB) approved protocols: 50 normal breast tissue cores donated to the Komen Tissue Bank (KTB) via vacuum-assisted biopsy and 10 tissues removed surgically at Mayo Clinic beyond resection margins of lumpectomy samples performed for atypical ductal hyperplasia (ADH). Tissues were frozen and maintained at -80°C. Normal KTB tissue cores were split in half. Tissues for analysis were weighed, extracted in cold methanol containing a cyclooxygenase inhibitor, and assayed for a panel of 42 oxylipins using validated mass spectrometric methods with a limit of detection per analyte of 0.010 ng/mL in supernatant and CVs of approximately <11%. Analyte measurements in paired cores (ng/g tissue) were compared with Spearman correlations and intra-class correlation coefficients. Measurements in surgical samples were assessed descriptively and levels were compared with normal KTB breast tissues.Median weight of the 50 split KTB cores was 0.055 grams. 17 oxylipins were detected in the tissue samples. PGD2 demonstrated the weakest correlation between split cores (r=0.12; p=0.40), whereas 15-HETE demonstrated the strongest correlation (r=0.68; p<0.0001). ICCs ranged from non-significant to nearly perfect, with highest values for 12-HETE (ICC=0.98, 95%CI: 0.97-0.99). Exploratory analyses of the KTB cores demonstrated significant associations between two BC risk factors, alcohol use and family history of breast/ovarian cancer, and several analytes. Levels of most eicosanoids tested were substantially and highly significantly elevated in tissues surrounding ADH versus KTB normal breast tissues. We conclude that eicosanoids can be readily measured in small frozen breast tissue samples obtained via percutaneous biopsy or at surgery. Future studies to assess the associations of eicosanoid levels with BC risk factors and development of BC after a benign biopsy may inform risk prediction and potentially inhibitable inflammatory mechanisms of breast carcinogenesis. Citation Format: Ginger L. Milne, Robert A. Vierkant, Amy C. Degnim, Anna Maria Storniolo, Jill E. Henry, Laura M. Pacheco-Spann, Derek C. Radisky, Mark E. Sherman. Quantification of oxylipins as biomarkers of inflammation and cancer risk in breast tissue [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6333.
Methylated DNA markers discriminate ovarian cancer from benign tissue in BRCA carriers.
e17610 Background: Methylated DNA markers (MDMs) associated with sporadic ovarian cancer (OC) are detected in plasma from patients with treatment naïve tumors, but not cancer-free controls. To assess potential benefits to patients at risk for hereditary OC, the aim of this study was to measure the performance of validated sporadic OC MDMs in OC and benign ovarian tissue (BOT), surgically obtained from germline BRCA1 or BRCA2 ( BRCA1/2) mutation carriers. Methods: BRCA1/2 carriers with OC and BRCA1/2 carriers who had undergone risk-reducing salpingo-oophorectomy (SO) were identified at a comprehensive cancer center. BRCA1/2 carriers in both OC and BOT groups were balanced on age and year of surgery with non- BRCA carriers with sporadic OC or benign indications for SO. OC and BOT formalin fixed paraffin embedded (FFPE) tissues were macrodissected following pathologist selection of representative sites; DNA was extracted, and bisulfite converted. 15 OC MDMs ( GDF6, IFFO1, MAX.chr1.147790358, MAX.chr6.10382190, MAX.chr11.14926602, C2CD4D, PDRM14, NCOR2, SKI, DSCR6, SIM2, PALLD, CDO1, GPRIN1, BCAT1) previously identified in sporadic OC were assayed by quantitative methylation specific PCR, normalized by β-actin, by blinded personnel. Areas under the receiver operating characteristic curve (AUC) were generated for each MDM to assess discrimination of OC from benign tissue and compared between BRCA1/2 and sporadic (non- BRCA) patients, using a z-test. Results: Among BRCA1/2 carriers, there were 48 OC ( BRCA-OC) and 48 BOT ( BRCA-BOT); among non- BRCA, we matched 48 OC (Sporadic-OC) and 48 BOT (non- BRCA-BOT). The median AUC for OC MDM discriminated between OC and BOT with AUCs in BRCA1/2 carriers was 0.87, (IQR, 0.83-0.93). Sporadic-OC vs non- BRCA-BOT median AUC was 0.88 (IQR, 0.83-0.92). AUCs were not significantly different for each MDM when stratified by BRCA or sporadic status. MDMs with an AUC of ≥0.9 in either group are shown (Table). Conclusions: OC MDMs identified and validated in sporadic OC are highly discriminant for OC from BOT in DNA extracted from tissues of people who carry a BRCA1/2 mutation. Testing the performance of these OC MDMs in plasma from BRCA1/2 carriers with and without OC may be of clinical value for high-risk patients and inform multi-cancer early detection testing strategies.[Table: see text]