Deep learning assessment of senescence-associated nuclear morphologies in mammary tissue from healthy female donors to predict future risk of breast cancer: a retrospective cohort study.

BACKGROUND: Cellular senescence has been associated with cancer as either a barrier mechanism restricting autonomous cell proliferation or a tumour-promoting microenvironmental mechanism that secretes proinflammatory paracrine factors. With most work done in non-human models and the heterogeneous nature of senescence, the precise role of senescent cells in the development of cancer in humans is not well understood. Furthermore, more than 1 million non-malignant breast biopsies are taken every year that could be a major resource for risk stratification. We aimed to explore the clinical relevance for breast cancer development of markers of senescence in mammary tissue from healthy female donors. METHODS: In this retrospective cohort study, we applied single-cell deep learning senescence predictors, based on nuclear morphology, to histological images of haematoxylin and eosin-stained breast biopsy samples from healthy female donors at the Komen Tissue Bank (KTB) at the Indiana University Simon Cancer Center (Indianapolis, IN, USA). All KTB participants (aged ≥18 years) who underwent core biopsies for research purposes between 2009 and 2019 were eligible for the study. Senescence was predicted in the epithelial (terminal duct lobular units [TDLUs] and non-TDLU epithelium), stromal, and adipose tissue compartments using validated models, previously trained on cells induced to senescence by ionising radiation (IR), replicative exhaustion (or replicative senescence; RS), or antimycin A, atazanavir-ritonavir, and doxorubicin (AAD) exposures. To benchmark our senescence-based cancer prediction results, we generated 5-year Gail scores-the current clinical gold standard for breast cancer risk prediction-for participants aged 35 years and older on the basis of characteristics at the time of tissue donation. The primary outcome was estimated odds of breast cancer via logistic modelling for each tissue compartment based on predicted senescence scores in cases (participants who had been diagnosed with breast cancer as of data cutoff, July 31, 2022) and controls (those who had not been diagnosed with breast cancer). FINDINGS: 4382 female donors (median age at donation 45 years [IQR 34-57]) were eligible for the study. As of data cutoff (median follow-up of 10 years [7-11]), 86 (2·0%) had developed breast cancer a mean of 4·8 years (SD 2·84) after date of donation and 4296 (98·0%) had not received a breast cancer diagnosis. Among the 86 cases, we found significant differences in adipose-specific IR and AAD senescence prediction scores compared with controls. Risk analysis showed that individuals in the upper half (above the median) of scores for the adipose tissue IR model had higher odds of developing breast cancer (odds ratio [OR] 1·71 [95% CI 1·10-2·68]; p=0·019), whereas the adipose AAD model revealed a reduced odds of developing breast cancer (OR 0·57 [0·36-0·88]; p=0·013). For the other tissue compartments and the RS model, no significant associations were found (except for stromal tissue via the IR model, had higher odds of developing breast cancer [OR 1·59, 1·03-2·49]). Individuals with both of the adipose risk factors had an OR of 3·32 (1·68-7·03; p=0·0009). Participants with 5-year Gail scores above the median had an OR for development of cancer of 2·33 (1·46-3·82; p=0·0012) compared with those with scores below the median. When combining Gail scores with our adipose AAD risk model, we found that individuals with both of these predictors had an OR of 4·70 (2·29-10·90; p<0·0001). When combining the Gail score with our adipose IR model, we found that individuals with both predictors had an OR of 3·45 (1·77-7·24; p=0·0002). INTERPRETATION: Assessment of senescence-associated nuclear morphologies with deep learning allows prediction of future cancer risk from normal breast biopsy samples. The combination of multiple models improved prediction of future breast cancer compared with the current clinical benchmark, the Gail model. Our results suggest an important role for microscope image-based deep learning models in predicting future cancer development. Such models could be incorporated into current breast cancer risk assessment and screening protocols. FUNDING: Novo Nordisk Foundation, Danish Cancer Society, and the US National Institutes of Health.

Breast cancer risk based on a deep learning predictor of senescent cells in normal tissue.

Background: The ability to predict future risk of cancer development in non-malignant biopsies is poor. Cellular senescence has been associated with cancer as either a barrier mechanism restricting autonomous cell proliferation or a tumor-promoting microenvironmental mechanism that secretes pro-inflammatory paracrine factors. With most work done in non-human models and the heterogenous nature of senescence the precise role of senescent cells in the development of cancer in humans is not well understood. Further, more than one million non-malignant breast biopsies are taken every year that could be a major source of risk-stratification for women. Methods: We applied single cell deep learning senescence predictors based on nuclear morphology to histological images of 4,411 H&E-stained breast biopsies from healthy female donors. Senescence was predicted in the epithelial, stromal, and adipocyte compartments using predictor models trained on cells induced to senescence by ionizing radiation (IR), replicative exhaustion (RS), or antimycin A, Atv/R and doxorubicin (AAD) exposures. To benchmark our senescence-based prediction results we generated 5-year Gail scores, the current clinical gold standard for breast cancer risk prediction. Findings: We found significant differences in adipocyte-specific IR and AAD senescence prediction for the 86 out of 4,411 healthy women who developed breast cancer an average 4.8 years after study entry. Risk models demonstrated that individuals in the upper median of scores for the adipocyte IR model had a higher risk (OR=1.71 [1.10-2.68], p=0.019), while the adipocyte AAD model revealed a reduced risk (OR=0.57 [0.36-0.88], p=0.013). Individuals with both adipocyte risk factors had an OR of 3.32 ([1.68-7.03], p<0.001). Alone, 5-year Gail scores yielded an OR of 2.70 ([1.22-6.54], p=0.019). When combining Gail scores with our adipocyte AAD risk model, we found that individuals with both of these risk predictors had an OR of 4.70 ([2.29-10.90], p<0.001). Interpretation: Assessment of senescence with deep learning allows considerable prediction of future cancer risk from non-malignant breast biopsies, something that was previously impossible to do. Furthermore, our results suggest an important role for microscope image-based deep learning models in predicting future cancer development. Such models could be incorporated into current breast cancer risk assessment and screening protocols. Funding: This study was funded by the Novo Nordisk Foundation (#NNF17OC0027812), and by the National Institutes of Health (NIH) Common Fund SenNet program (U54AG075932).

Anti-fibrotic activity of a rho-kinase inhibitor restores outflow function and intraocular pressure homeostasis.

Glucocorticoids are widely used as an ophthalmic medication. A common, sight-threatening adverse event of glucocorticoid usage is ocular hypertension, caused by dysfunction of the conventional outflow pathway. We report that netarsudil, a rho-kinase inhibitor, decreased glucocorticoid-induced ocular hypertension in patients whose intraocular pressures were poorly controlled by standard medications. Mechanistic studies in our established mouse model of glucocorticoid-induced ocular hypertension show that netarsudil both prevented and reduced intraocular pressure elevation. Further, netarsudil attenuated characteristic steroid-induced pathologies as assessed by quantification of outflow function and tissue stiffness, and morphological and immunohistochemical indicators of tissue fibrosis. Thus, rho-kinase inhibitors act directly on conventional outflow cells to prevent or attenuate fibrotic disease processes in glucocorticoid-induced ocular hypertension in an immune-privileged environment. Moreover, these data motivate the need for a randomized prospective clinical study to determine whether netarsudil is indeed superior to first-line anti-glaucoma drugs in lowering steroid-induced ocular hypertension.

Integral role for lysyl oxidase-like-1 in conventional outflow tissue function and behavior.

Lysyl oxidase-like-1 (LOXL1), a vital crosslinking enzyme in elastin fiber maintenance, is essential for the stability and strength of elastic vessels and tissues. Variants in the LOXL1 locus associate with a dramatic increase in risk of exfoliation syndrome (XFS), a systemic fibrillopathy, which often presents with ocular hypertension and exfoliation glaucoma (XFG). We examined the role of LOXL1 in conventional outflow function, the prime regulator of intraocular pressure (IOP). Using Loxl1-/- , Loxl1+/- , and Loxl1+/+ mice, we observed an inverse relationship between LOXL1 expression and IOP, which worsened with age. Elevated IOP in Loxl1-/- mice was associated with a larger globe, decreased ocular compliance, increased outflow facility, extracellular matrix (ECM) abnormalities, and dilated intrascleral veins, yet, no dilation of arteries or capillaries. Interestingly, in living Loxl1-/- mouse eyes, Schlemm's canal (SC) was less susceptible to collapse when challenged with acute elevations in IOP, suggesting elevated episcleral venous pressure (EVP). Thus, LOXL1 expression is required for normal IOP control, while ablation results in altered ECM repair/homeostasis and conventional outflow physiology. Dilation of SC and distal veins, but not arteries, is consistent with key structural and functional roles for elastin in low-pressure vessels subjected to cyclical mechanical stress.

SOCR Analyses - an Instructional Java Web-based Statistical Analysis Toolkit.

The Statistical Online Computational Resource (SOCR) designs web-based tools for educational use in a variety of undergraduate courses (Dinov 2006). Several studies have demonstrated that these resources significantly improve students' motivation and learning experiences (Dinov et al. 2008). SOCR Analyses is a new component that concentrates on data modeling and analysis using parametric and non-parametric techniques supported with graphical model diagnostics. Currently implemented analyses include commonly used models in undergraduate statistics courses like linear models (Simple Linear Regression, Multiple Linear Regression, One-Way and Two-Way ANOVA). In addition, we implemented tests for sample comparisons, such as t-test in the parametric category; and Wilcoxon rank sum test, Kruskal-Wallis test, Friedman's test, in the non-parametric category. SOCR Analyses also include several hypothesis test models, such as Contingency tables, Friedman's test and Fisher's exact test.The code itself is open source (http://socr.googlecode.com/), hoping to contribute to the efforts of the statistical computing community. The code includes functionality for each specific analysis model and it has general utilities that can be applied in various statistical computing tasks. For example, concrete methods with API (Application Programming Interface) have been implemented in statistical summary, least square solutions of general linear models, rank calculations, etc. HTML interfaces, tutorials, source code, activities, and data are freely available via the web (www.SOCR.ucla.edu). Code examples for developers and demos for educators are provided on the SOCR Wiki website.In this article, the pedagogical utilization of the SOCR Analyses is discussed, as well as the underlying design framework. As the SOCR project is on-going and more functions and tools are being added to it, these resources are constantly improved. The reader is strongly encouraged to check the SOCR site for most updated information and newly added models.

SOCR Analyses: Implementation and Demonstration of a New Graphical Statistics Educational Toolkit.

The web-based, Java-written SOCR (Statistical Online Computational Resource) tools have been utilized in many undergraduate and graduate level statistics courses for seven years now (Dinov 2006; Dinov et al. 2008b). It has been proven that these resources can successfully improve students' learning (Dinov et al. 2008b). Being first published online in 2005, SOCR Analyses is a somewhat new component and it concentrate on data modeling for both parametric and non-parametric data analyses with graphical model diagnostics. One of the main purposes of SOCR Analyses is to facilitate statistical learning for high school and undergraduate students. As we have already implemented SOCR Distributions and Experiments, SOCR Analyses and Charts fulfill the rest of a standard statistics curricula. Currently, there are four core components of SOCR Analyses. Linear models included in SOCR Analyses are simple linear regression, multiple linear regression, one-way and two-way ANOVA. Tests for sample comparisons include t-test in the parametric category. Some examples of SOCR Analyses' in the non-parametric category are Wilcoxon rank sum test, Kruskal-Wallis test, Friedman's test, Kolmogorov-Smirnoff test and Fligner-Killeen test. Hypothesis testing models include contingency table, Friedman's test and Fisher's exact test. The last component of Analyses is a utility for computing sample sizes for normal distribution. In this article, we present the design framework, computational implementation and the utilization of SOCR Analyses.

Global genechip profiling to identify genes responsive to p53-induced growth arrest and apoptosis in human lung carcinoma cells.

To identify critical genes that mediate p53-induced growth arrest and apoptosis at a global level, we profiled a human lung carcinoma cell model in which cells undergo growth arrest and apoptosis in a p53 and DNA damage-dependent manner. Profiling of the Affymetrix human HG-U1333 GeneChip, covering the entire human transcriptome, revealed about 3, 000 unique genes either induced or repressed during p53-induced growth arrest or apoptosis, respectively. A total of 1, 057 genes, including many well-known p53 targets, responded to both conditions. A mini apoptotic protein database was generated from 3, 033 unique apoptosis responsive genes. Analysis of this database yielded 23 proteins with a pro-apoptotic BH3 domain and three with anti-apoptotic BIR2/BIR3 domains, including well-known p53 targets: Bax, Puma, Noxa and survivin. In addition, 14 mitochondrial proteins were identified that contain a pro-apoptotic AVPI-like motif, and 15 proteins were identified that contain a DAVPI-like domain with the potential of being cleaved by caspases during apoptosis to release the AVPI motif. Many of the genes we identified with these domains do contain p53-binding sites either in the promoter or in the first three introns, suggesting a high probability of being direct p53 targets. Pathway analysis revealed that p53 might control the Wnt pathway through transcriptional regulation of some of its components. Thus, global chip profiling coupled with bioinformatics analysis is a powerful tool in identification of genes critical for p53-induced apoptosis. Further characterization of these genes will lead to a better understanding of the mechanism of p53 action and p53 regulation of other signaling pathways. It will also provide novel cancer drug targets for further validation.