Laleh Montaser-Kouhsari

Computational Pathology to Discriminate Benign from Malignant Intraductal Proliferations of the Breast

The categorization of intraductal proliferative lesions of the breast based on routine light microscopic examination of histopathologic sections is in many cases challenging, even for experienced pathologists. The development of computational tools to aid pathologists in the characterization of these lesions would have great diagnostic and clinical value. As a first step to address this issue, we evaluated the ability of computational image analysis to accurately classify DCIS and UDH and to stratify nuclear grade within DCIS. Using 116 breast biopsies diagnosed as DCIS or UDH from the Massachusetts General Hospital (MGH), we developed a computational method to extract 392 features corresponding to the mean and standard deviation in nuclear size and shape, intensity, and texture across 8 color channels. We used L1-regularized logistic regression to build classification models to discriminate DCIS from UDH. The top-performing model contained 22 active features and achieved an AUC of 0.95 in cross-validation on the MGH data-set. We applied this model to an external validation set of 51 breast biopsies diagnosed as DCIS or UDH from the Beth Israel Deaconess Medical Center, and the model achieved an AUC of 0.86. The top-performing model contained active features from all color-spaces and from the three classes of features (morphology, intensity, and texture), suggesting the value of each for prediction. We built models to stratify grade within DCIS and obtained strong performance for stratifying low nuclear grade vs. high nuclear grade DCIS (AUC = 0.98 in cross-validation) with only moderate performance for discriminating low nuclear grade vs. intermediate nuclear grade and intermediate nuclear grade vs. high nuclear grade DCIS (AUC = 0.83 and 0.69, respectively). These data show that computational pathology models can robustly discriminate benign from malignant intraductal proliferative lesions of the breast and may aid pathologists in the diagnosis and classification of these lesions.

Extensive rewiring of epithelial-stromal co-expression networks in breast cancer

BackgroundEpithelial-stromal crosstalk plays a critical role in invasive breast cancer pathogenesis; however, little is known on a systems level about how epithelial-stromal interactions evolve during carcinogenesis.ResultsWe develop a framework for building genome-wide epithelial-stromal co-expression networks composed of pairwise co-expression relationships between mRNA levels of genes expressed in the epithelium and stroma across a population of patients. We apply this method to laser capture micro-dissection expression profiling datasets in the setting of breast carcinogenesis. Our analysis shows that epithelial-stromal co-expression networks undergo extensive rewiring during carcinogenesis, with the emergence of distinct network hubs in normal breast, and estrogen receptor-positive and estrogen receptor-negative invasive breast cancer, and the emergence of distinct patterns of functional network enrichment. In contrast to normal breast, the strongest epithelial-stromal co-expression relationships in invasive breast cancer mostly represent self-loops, in which the same gene is co-expressed in epithelial and stromal regions. We validate this observation using an independent laser capture micro-dissection dataset and confirm that self-loop interactions are significantly increased in cancer by performing computational image analysis of epithelial and stromal protein expression using images from the Human Protein Atlas.ConclusionsEpithelial-stromal co-expression network analysis represents a new approach for systems-level analyses of spatially localized transcriptomic data. The analysis provides new biological insights into the rewiring of epithelial-stromal co-expression networks and the emergence of epithelial-stromal co-expression self-loops in breast cancer. The approach may facilitate the development of new diagnostics and therapeutics targeting epithelial-stromal interactions in cancer.

MERIT40 Is an Akt Substrate that Promotes Resolution of DNA Damage Induced by Chemotherapy

Resistance to cytotoxic chemotherapy drugs, including doxorubicin, is a significant obstacle to the effective treatment of breast cancer. Here, we have identified a mechanism by which the PI3K/Akt pathway mediates resistance to doxorubicin. In addition to inducing DNA damage, doxorubicin triggers sustained activation of Akt signaling in breast cancer cells. We show that Akt contributes to chemotherapy resistance such that PI3K or Akt inhibitors sensitize cells to doxorubicin. We identify MERIT40, a component of the BRCA1-A DNA damage repair complex, as an Akt substrate that is phosphorylated following doxorubicin treatment. MERIT40 phosphorylation facilitates assembly of the BRCA1-A complex in response to DNA damage and contributes to DNA repair and cell survival following doxorubicin treatment. Finally, MERIT40 phosphorylation in human breast cancers is associated with estrogen receptor positivity. Our findings suggest that combination therapy with PI3K or Akt inhibitors and doxorubicin may constitute a successful strategy for overcoming chemotherapy resistance.

Clinical Utility of a Blood-Based BRAFV600E Mutation Assay in Melanoma

BRAF inhibitors (BRAFi) have led to clinical benefit in patients with melanoma. The development of a blood-based assay to detect and quantify BRAF levels in these patients has diagnostic, prognostic, and predictive capabilities that could guide treatment decisions. Blood BRAFV600E detection and quantification were performed on samples from 128 patients with stage II (19), III (67), and IV (42) melanoma. Tissue BRAF analysis was performed in all patients with stage IV disease and in selected patients with stage II and III disease. Clinical outcomes were correlated to initial BRAF levels as well as BRAF level dynamics. Serial analysis was performed on 17 stage IV melanoma patients treated with BRAFi and compared with tumor measurements by RECIST. The assay was highly sensitive (96%) and specific (95%) in the stage IV setting, using a blood level of 4.8 pg as “positive.” BRAF levels typically decreased following BRAFi. A subset of these patients (5) had an increase in BRAFV600E values 42 to 112 days before clinical or radiographic disease progression (PD). From 86 patients with resected, stage II or III melanoma, 39 had evidence of disease relapse (45.3%). Furthermore, BRAF mutation in the blood after surgical resection in these patients was not associated with a difference in relapse risk, although tissue BRAF status was only available for a subset of patients. In summary, we have developed a highly sensitive and specific, blood-based assay to detect BRAFV600 mutation in patients with melanoma. Mol Cancer Ther; 13(12); 3210–8. ©2014 AACR.

Crowdsourcing image annotation for nucleus detection and segmentation in computational pathology: evaluating experts, automated methods, and the crowd.

The development of tools in computational pathology to assist physicians and biomedical scientists in the diagnosis of disease requires access to high-quality annotated images for algorithm learning and evaluation. Generating high-quality expert-derived annotations is time-consuming and expensive. We explore the use of crowdsourcing for rapidly obtaining annotations for two core tasks in com- putational pathology: nucleus detection and nucleus segmentation. We designed and implemented crowdsourcing experiments using the CrowdFlower platform, which provides access to a large set of labor channel partners that accesses and manages millions of contributors worldwide. We obtained annotations from four types of annotators and compared concordance across these groups. We obtained: crowdsourced annotations for nucleus detection and segmentation on a total of 810 images; annotations using automated methods on 810 images; annotations from research fellows for detection and segmentation on 477 and 455 images, respectively; and expert pathologist-derived annotations for detection and segmentation on 80 and 63 images, respectively. For the crowdsourced annotations, we evaluated performance across a range of contributor skill levels (1, 2, or 3). The crowdsourced annotations (4,860 images in total) were completed in only a fraction of the time and cost required for obtaining annotations using traditional methods. For the nucleus detection task, the research fellow-derived annotations showed the strongest concordance with the expert pathologist- derived annotations (F-M =93.68%), followed by the crowd-sourced contributor levels 1,2, and 3 and the automated method, which showed relatively similar performance (F-M = 87.84%, 88.49%, 87.26%, and 86.99%, respectively). For the nucleus segmentation task, the crowdsourced contributor level 3-derived annotations, research fellow-derived annotations, and automated method showed the strongest concordance with the expert pathologist-derived annotations (F-M = 66.41%, 65.93%, and 65.36%, respectively), followed by the contributor levels 2 and 1 (60.89% and 60.87%, respectively). When the research fellows were used as a gold-standard for the segmentation task, all three con- tributor levels of the crowdsourced annotations significantly outperformed the automated method (F-M = 62.21%, 62.47%, and 65.15% vs. 51.92%). Aggregating multiple annotations from the crowd to obtain a consensus annotation resulted in the strongest performance for the crowd-sourced segmentation. For both detection and segmentation, crowd-sourced performance is strongest with small images (400 × 400 pixels) and degrades significantly with the use of larger images (600 × 600 and 800 × 800 pixels). We conclude that crowdsourcing to non-experts can be used for large-scale labeling microtasks in computational pathology and offers a new approach for the rapid generation of labeled images for algorithm development and evaluation.

Expression of estrogen receptor, progesterone receptor, and Ki67 in normal breast tissue in relation to subsequent risk of breast cancer

Although expression of estrogen receptor (ER), progesterone receptor (PR), and cell proliferation marker Ki67 serve as predictive and prognostic factors in breast cancers, little is known about their roles in normal breast tissue. Here in a nested case–control study within the Nurses’ Health Studies (90 cases, 297 controls), we evaluated their expression levels in normal breast epithelium in relation to subsequent breast cancer risk among women with benign breast disease. Tissue microarrays were constructed using cores obtained from benign biopsies containing normal terminal duct lobular units and immunohistochemical stained for these markers. We found PR and Ki67 expression was non-significantly but positively associated with subsequent breast cancer risk, whereas ER expression was non-significantly inversely associated. After stratifying by lesion subtype, Ki67 was significantly associated with higher risk among women with proliferative lesions with atypical hyperplasia. However, given the small sample size, further studies are required to confirm these results.

NFAT1 promotes intratumoral neutrophil infiltration by regulating IL8 expression in breast cancer

NFAT transcription factors are key regulators of gene expression in immune cells. In addition, NFAT1‐induced genes play diverse roles in mediating the progression of various solid tumors. Here we show that NFAT1 induces the expression of the IL8 gene by binding to its promoter and leading to IL8 secretion. Thapsigargin stimulation of breast cancer cells induces IL8 expression in an NFAT‐dependent manner. Moreover, we show that NFAT1‐mediated IL8 production promotes the migration of primary human neutrophils in vitro and also promotes neutrophil infiltration in tumor xenografts. Furthermore, expression of active NFAT1 effectively suppresses the growth of nascent and established tumors by a non cell‐autonomous mechanism. Evaluation of breast tumor tissue reveals that while the levels of NFAT1 are similar in tumor cells and normal breast epithelium, cells in the tumor stroma express higher levels of NFAT1 compared to normal stroma. Elevated levels of NFAT1 also correlate with increased neutrophil infiltrate in breast tumors. These data point to a mechanism by which NFAT1 orchestrates the communication between breast cancer cells and host neutrophils during breast cancer progression.

Breast cancer risk factors in relation to estrogen receptor, progesterone receptor, insulin-like growth factor-1 receptor, and Ki67 expression in normal breast tissue

Studies have suggested that hormone receptor and Ki67 expression in normal breast tissue are associated with subsequent breast cancer risk. We examined the associations of breast cancer risk factors with estrogen receptor (ER), progesterone receptor (PR), insulin-like growth factor-1 receptor (IGF-1R), and Ki67 expression in normal breast tissue. This analysis included 388 women with benign breast disease (ages 17–67 years) in the Nurses’ Health Studies. Immunohistochemical staining was performed on tissue microarrays constructed from benign biopsies containing normal breast epithelium and scored as the percentage of epithelial cells that were positively stained. Ordinal logistic regression (outcomes in tertiles), adjusting for age and potential confounders, was performed to estimate odds ratios (OR) and 95% confidence intervals (CI) for the associations with risk factors. Alcohol consumption was positively associated (≥2.5 vs.<0.4 drink/wk: OR = 2.69, 95% CI = 1.26–5.75, p-trend = 0.008) and breastfeeding was inversely associated (≥6 months vs. never: OR = 0.11, 95% CI = 0.04–0.35, p-trend = 0.0003) with ER expression. Height (≥66 vs.<64 inches: OR = 2.50, 95% CI = 1.34–4.67, p-trend = 0.005) and BMI at age 18 (≥22 vs.<20 kg/m2: OR = 2.33, 95% CI = 1.18–4.62, p-trend = 0.01) were positively associated with PR expression. Body size at age 5–10 years was inversely associated with Ki67 (Level ≥ 2.5 vs. 1: OR = 0.55, 95% CI = 0.30–1.01, p-trend = 0.03). Premenopausal BMI (≥25 vs.<20 kg/m2) was positively associated with cytoplasmic IGF-1R (OR = 5.06, 95% CI = 1.17–21.8, p-trend = 0.04). Our data suggest that anthropometrics, breastfeeding, and alcohol intake may influence the molecular characteristics of normal breast tissue, elucidating the mechanisms by which these risk factors operate. However, larger studies are required to confirm these results.Risk factors: Alcohol, breastfeeding and body traits tied to molecular biomarkersBody size, alcohol intake and breastfeeding may affect the molecular features of normal breast tissue to influence cancer risk. A team led by Hannah Oh from the Harvard T. H. Chan School of Public Health in Boston, USA, examined the link between behavioral and physiological risk factors for breast cancer and the expression levels of certain proteins (such as hormone receptors) that also promote cancer formation. By examining healthy breast tissue biopsied from 388 women diagnosed with benign breast diseases, the researchers showed that alcohol consumption, height and body mass index were all positively associated with the expression of various breast tissue markers, while breastfeeding and early-life body size were inversely associated. The findings point to a connection between lifestyle and breast tissue-specific molecular characteristics that underpin cancer risk.

Image-guided Coring for Large-scale Studies in Molecular Pathology.

Sampling of formalin-fixed paraffin-embedded (FFPE) tissue blocks is a critical initial step in molecular pathology. Image-guided coring (IGC) is a new method for using digital pathology images to guide tissue block coring for molecular analyses. The goal of our study is to evaluate the use of IGC for both tissue-based and nucleic acid–based projects in molecular pathology. First, we used IGC to construct a tissue microarray (TMA); second, we used IGC for FFPE block sampling followed by RNA extraction; and third, we assessed the correlation between nuclear counts quantitated from the IGC images and RNA yields. We used IGC to construct a TMA containing 198 normal and breast cancer cores. Histopathologic analysis showed high accuracy for obtaining tumor and normal breast tissue. Next, we used IGC to obtain normal and tumor breast samples before RNA extraction. We selected a random subset of tumor and normal samples to perform computational image analysis to quantify nuclear density, and we built regression models to estimate RNA yields from nuclear count, age of the block, and core diameter. Number of nuclei and core diameter were the strongest predictors of RNA yields in both normal and tumor tissue. IGC is an effective method for sampling FFPE tissue blocks for TMA construction and nucleic acid extraction. We identify significant associations between quantitative nuclear counts obtained from IGC images and RNA yields, suggesting that the integration of computational image analysis with IGC may be an effective approach for tumor sampling in large-scale molecular studies.

Abstract 3477: 3D morphological hallmarks of breast carcinogenesis: Diagnosis of non-invasive and invasive breast cancer with Lightsheet microscopy

BACKGROUND: Since the early 20th century, the pathological classification of breast cancer has been based primarily on the visual analysis of HE 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 3477. doi:10.1158/1538-7445.AM2015-3477