Mini Review: The Last Mile-Opportunities and Challenges for Machine Learning in Digital Toxicologic Pathology.

The 2019 manuscript by the Special Interest Group on Digital Pathology and Image Analysis of the Society of Toxicologic pathology suggested that a synergism between artificial intelligence (AI) and machine learning (ML) technologies and digital toxicologic pathology would improve the daily workflow and future impact of toxicologic pathologists globally. Now 2 years later, the authors of this review consider whether, in their opinion, there is any evidence that supports that thesis. Specifically, we consider the opportunities and challenges for applying ML (the study of computer algorithms that are able to learn from example data and extrapolate the learned information to unseen data) algorithms in toxicologic pathology and how regulatory bodies are navigating this rapidly evolving field. Although we see similarities with the "Last Mile" metaphor, the weight of evidence suggests that toxicologic pathologists should approach ML with an equal dose of skepticism and enthusiasm. There are increasing opportunities for impact in our field that leave the authors cautiously excited and optimistic. Toxicologic pathologists have the opportunity to critically evaluate ML applications with a "call-to-arms" mentality. Why should we be late adopters? There is ample evidence to encourage engagement, growth, and leadership in this field.

Society of Toxicologic Pathology Digital Pathology and Image Analysis Special Interest Group Article*: Opinion on the Application of Artificial Intelligence and Machine Learning to Digital Toxicologic Pathology.

Toxicologic pathology is transitioning from analog to digital methods. This transition seems inevitable due to a host of ongoing social and medical technological forces. Of these, artificial intelligence (AI) and in particular machine learning (ML) are globally disruptive, rapidly growing sectors of technology whose impact on the long-established field of histopathology is quickly being realized. The development of increasing numbers of algorithms, peering ever deeper into the histopathological space, has demonstrated to the scientific community that AI pathology platforms are now poised to truly impact the future of precision and personalized medicine. However, as with all great technological advances, there are implementation and adoption challenges. This review aims to define common and relevant AI and ML terminology, describe data generation and interpretation, outline current and potential future business cases, discuss validation and regulatory hurdles, and most importantly, propose how overcoming the challenges of this burgeoning technology may shape toxicologic pathology for years to come, enabling pathologists to contribute even more effectively to answering scientific questions and solving global health issues. [Box: see text].

Validation of a Muscle-Specific Tissue Image Analysis Tool for Quantitative Assessment of Dystrophin Staining in Frozen Muscle Biopsies.

CONTEXT.­: Duchenne muscular dystrophy is a rare, progressive, and fatal neuromuscular disease caused by dystrophin protein loss. Common investigational treatment approaches aim at increasing dystrophin expression in diseased muscle. Some clinical trials include assessments of novel dystrophin production as a surrogate biomarker of efficacy, which may predict a clinical benefit from treatment. OBJECTIVES.­: To establish an immunofluorescent scanning and digital image analysis workflow that provides an objective approach for staining intensity assessment of the immunofluorescence dystrophin labeling and determination of the percentage of biomarker-positive fibers in muscle cryosections. DESIGN.­: Optimal and repeatable digital image capture was achieved by a rigorously qualified fluorescent scanning process. After scanning qualification, the MuscleMap (Flagship Biosciences, Westminster, Colorado) algorithm was validated by comparing high-power microscopic field total and dystrophin-positive fiber counts obtained by trained pathologists to data derived by MuscleMap. Next, the algorithm was tested on whole-slide images of immunofluorescent-labeled muscle sections from Duchenne muscular dystrophy, Becker muscular dystrophy, and control patients. RESULTS.­: When used under the guidance of a trained pathologist, the digital image analysis tool met predefined validation criteria and demonstrated functional and statistical equivalence with manual assessment. This work is the first, to our knowledge, to qualify and validate immunofluorescent scanning and digital tissue image-analysis workflow, respectively, with the rigor required to support the clinical trial environments. CONCLUSIONS.­: MuscleMap enables analysis of all fibers within an entire muscle biopsy section and provides data on a fiber-by-fiber basis. This will allow future clinical trials to objectively investigate myofibers' dystrophin expression at a greater level of consistency and detail.

The Emergence of Microphysiological Systems (Organs-on-chips) as Paradigm-changing Tools for Toxicologic Pathology.

Microphysiological systems (MPS), commonly known as organs-on-chips, are a rapidly advancing technology that promises to impact many areas of medical and toxicological pathology. In this minireview, the history of MPS and its potential utility in safety assessment are described with the toxicologic pathologist in mind. Several MPS development focus areas are defined, and recent progress in the area is highlighted. MPS will likely become an important tool for the toxicologic pathologist as part of our role in the safety assessment process within the pharmaceutical, biotechnology, medical device, and cosmetic and agrichemical industries.

Duchenne and Becker Muscular Dystrophies: A Review of Animal Models, Clinical End Points, and Biomarker Quantification.

Duchenne muscular dystrophy (DMD) and Becker muscular dystrophy (BMD) are neuromuscular disorders that primarily affect boys due to an X-linked mutation in the DMD gene, resulting in reduced to near absence of dystrophin or expression of truncated forms of dystrophin. Some newer therapeutic interventions aim to increase sarcolemmal dystrophin expression, and accurate dystrophin quantification is critical for demonstrating pharmacodynamic relationships in preclinical studies and clinical trials. Current challenges with measuring dystrophin include the variation in protein expression within individual muscle fibers and across whole muscle samples, the presence of preexisting dystrophin-positive revertant fibers, and trace amounts of residual dystrophin. Immunofluorescence quantification of dystrophin can overcome many of these challenges, but manual quantification of protein expression may be complicated by variations in the collection of images, reproducible scoring of fluorescent intensity, and bias introduced by manual scoring of typically only a few high-power fields. This review highlights the pathology of DMD and BMD, discusses animal models of DMD and BMD, and describes dystrophin biomarker quantitation in DMD and BMD, with several image analysis approaches, including a new automated method that evaluates protein expression of individual muscle fibers.

The Gold Standard Paradox in Digital Image Analysis: Manual Versus Automated Scoring as Ground Truth.

CONTEXT: - Novel therapeutics often target complex cellular mechanisms. Increasingly, quantitative methods like digital tissue image analysis (tIA) are required to evaluate correspondingly complex biomarkers to elucidate subtle phenotypes that can inform treatment decisions with these targeted therapies. These tIA systems need a gold standard, or reference method, to establish analytical validity. Conventional, subjective histopathologic scores assigned by an experienced pathologist are the gold standard in anatomic pathology and are an attractive reference method. The pathologist's score can establish the ground truth to assess a tIA solution's analytical performance. The paradox of this validation strategy, however, is that tIA is often used to assist pathologists to score complex biomarkers because it is more objective and reproducible than manual evaluation alone by overcoming known biases in a human's visual evaluation of tissue, and because it can generate endpoints that cannot be generated by a human observer. OBJECTIVE: - To discuss common visual and cognitive traps known in traditional pathology-based scoring paradigms that may impact characterization of tIA-assisted scoring accuracy, sensitivity, and specificity. DATA SOURCES: - This manuscript reviews the current literature from the past decades available for traditional subjective pathology scoring paradigms and known cognitive and visual traps relevant to these scoring paradigms. CONCLUSIONS: - Awareness of the gold standard paradox is necessary when using traditional pathologist scores to analytically validate a tIA tool because image analysis is used specifically to overcome known sources of bias in visual assessment of tissue sections.

Using Automated Image Analysis Algorithms to Distinguish Normal, Aberrant, and Degenerate Mitotic Figures Induced by Eg5 Inhibition.

Modulation of the cell cycle may underlie the toxicologic or pharmacologic responses of a potential therapeutic agent and contributes to decisions on its preclinical and clinical safety and efficacy. The descriptive and quantitative assessment of normal, aberrant, and degenerate mitotic figures in tissue sections is an important end point characterizing the effect of xenobiotics on the cell cycle. Historically, pathologists used manual counting and special staining visualization techniques such as immunohistochemistry for quantification of normal, aberrant, and degenerate mitotic figures. We designed an automated image analysis algorithm for measuring these mitotic figures in hematoxylin and eosin (H&E)-stained sections. Algorithm validation methods used data generated from a subcutaneous human transitional cell carcinoma xenograft model in nude rats treated with the cell cycle inhibitor Eg5. In these studies, we scanned and digitized H&E-stained xenografts and applied a complex ruleset of sequential mathematical filters and shape discriminators for classification of cell populations demonstrating normal, aberrant, or degenerate mitotic figures. The resultant classification system enabled the representations of three identifiable degrees of morphological change associated with tumor differentiation and compound effects. The numbers of mitotic figure variants and mitotic indices data generated corresponded to a manual assessment by a pathologist and supported automated algorithm verification and application for both efficacy and toxicity studies.

Tailored Pig Models for Preclinical Efficacy and Safety Testing of Targeted Therapies.

Despite enormous advances in translational biomedical research, there remains a growing demand for improved animal models of human disease. This is particularly true for diseases where rodent models do not reflect the human disease phenotype. Compared to rodents, pig anatomy and physiology are more similar to humans in cardiovascular, immune, respiratory, skeletal muscle, and metabolic systems. Importantly, efficient and precise techniques for genetic engineering of pigs are now available, facilitating the creation of tailored large animal models that mimic human disease mechanisms at the molecular level. In this article, the benefits of genetically engineered pigs for basic and translational research are exemplified by a novel pig model of Duchenne muscular dystrophy and by porcine models of cystic fibrosis. Particular emphasis is given to potential advantages of using these models for efficacy and safety testing of targeted therapies, such as exon skipping and gene editing, for example, using the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated system. In general, genetically tailored pig models have the potential to bridge the gap between proof-of-concept studies in rodents and clinical trials in patients, thus supporting translational medicine.

High molecular weight polyethylene glycol cellular distribution and PEG-associated cytoplasmic vacuolation is molecular weight dependent and does not require conjugation to proteins.

Conjugation of therapeutic proteins with high molecular weight polyethylene glycols (HMW PEGs) is used to extend the half-life of biologics. To evaluate the effects of HMW PEGs in animals, we used an immunohistochemical procedure to study the tissue distribution and toxicity of unconjugated HMW PEGs in rats given 100 mg/kg (10K)PEG, (20K)PEG, or (40K)PEG intravenously. Both the PEG cellular distribution and the histology were different between groups. In (10K)PEG and (20K)PEG groups, PEG immunoreactivity was most prominent in the renal tubule epithelium and in alveolar macrophages and hepatic Kupffer cells and cellular vacuolation was absent. In contrast, rats given (40K)PEG had strong PEG immunoreactivity in splenic subcapsular red pulp macrophages, renal interstitial macrophages, and choroid plexus epithelial cells that was frequently associated with cytoplasmic vacuolation. While the vacuolation appeared to be an adaptive response, there was focal renal tubular epithelial degeneration associated with strong PEG immunoreactivity in one rat given (40K)PEG. These data indicate that both the tissue distribution and the vacuolation observed with unconjugated HMW PEGs are markedly influenced by the molecular weight of the PEG and that when vacuolation is observed it is likely an adaptive change that is associated with PEG cytoplasmic immunoreactivity.

On-target and off-target-based toxicologic effects.

Adverse toxicologic effects are categorized as chemical-based, on-target, or off-target effects. Chemical-based toxicity is defined as toxicity that is related to the physicochemical characteristics of a compound and its effects on cellular organelles, membranes, and/or metabolic pathways. On-target refers to exaggerated and adverse pharmacologic effects at the target of interest in the test system. Off-target refers to adverse effects as a result of modulation of other targets; these may be related biologically or totally unrelated to the target of interest. Both the risk assessment and development strategies used for xenobiotics are influenced by the understanding of the mechanism of toxicity. It is imperative that the toxicologic pathologist use the toxicologic and biologic data at hand and literature information on the target to form testable hypotheses related to whether a toxicity is chemical-based, on-target, or off-target. The objective of this session at the 2012 Society of Toxicologic Pathologists Symposium in Boston, Massachusetts, was to discuss chemical-based, on-target, and off-target-based effects and the scientific approaches used to aid in their human risk assessment.

Rat-specific decreases in platelet count caused by a humanized monoclonal antibody against sclerostin.

LY2541546 is a humanized monoclonal antibody (IgG(4)) that has been optimized for neutralizing activity against sclerostin. In 5-week and 6-month nonclinical safety studies in rats, LY2541546 caused dose-dependent reversible decreases in platelet counts accompanied by accelerated platelet production, increased megakaryocytes, and altered megakaryocyte morphology. These treatment-related effects resulted in altered primary hemostasis as manifested by prolonged bleeding after phlebotomy or incidental toenail break. In some cases, the defects in hemostasis were sufficient to result in death of the affected rats. There was no evidence in rats of general bone marrow suppression or processes (e.g., disseminated intravascular coagulopathy) that may result in thrombocytopenia. Cynomolgus monkeys given LY2541546 for 5 weeks or 9 months had no changes in platelet count or megakaryocytes. In vitro cross-reactivity studies in rats, cynomolgus monkeys, and humans revealed LY2541546-bound rat but not cynomolgus monkey or human platelets and megakaryocytes. These data taken together demonstrated that the platelet and megakaryocyte effects in rats had a species-specific pathogenesis which likely involved LY2541546 binding of a rat-specific antigen on the surface of platelets and megakaryocytes resulting in the increased clearance of platelets and megakaryocyte hyperplasia. The species-specific nature of these reversible toxicological findings combined with the ease of clinical monitoring using standard hematology enabled the safe initiation of clinical studies in human volunteers.

Qualification of cardiac troponin I concentration in mouse serum using isoproterenol and implementation in pharmacology studies to accelerate drug development.

Cardiac troponin I is a useful biomarker of myocardial injury, but its use in mice and application to early drug discovery are not well described. The authors investigated the relationship between cTnI concentration in serum and histologic lesions in heart tissue from mice treated with isoproterenol (ISO). Cardiac TnI concentrations in serum increased in a dose-dependant manner and remained increased twenty-four to forty-eight hours after a single administration of isoproterenol. Increased cTnI concentration was of greater magnitude and longer duration than increased fatty acid binding protein 3 concentration, aspartate aminotransferase activity, and creatine kinase activity in serum. Isoproterenol-induced increases in cTnI concentrations were both greater and more sustained in BALB/c than in CD1 mice and correlated with incidence and severity of lesions observed in heart sections from both strains. In drug development studies in BALB/c mice with novel kinase inhibitors, cTnI concentration was a reliable stand-alone biomarker of cardiac injury and was used in combination with measurements of in vivo target inhibition to demonstrate an off-target contribution to cardiotoxicity. Additional attributes, including low cost and rapid turnaround time, made cTnI concentration in serum invaluable for detecting cardiotoxicity, exploring structure-activity relationships, and prioritizing development of compounds with improved safety profiles early in drug discovery.

Structure-activity relationships of SERMs optimized for uterine antagonism and ovarian safety.

Structure-activity relationship studies are described, which led to the discovery of novel selective estrogen receptor modulators (SERMs) for the potential treatment of uterine fibroids. The SAR studies focused on limiting brain exposure and were guided by computational properties. Compounds with limited impact on the HPO axis were selected using serum estrogen levels as a biomarker for ovarian stimulation.

The rat mammary gland: morphologic changes as an indicator of systemic hormonal perturbations induced by xenobiotics.

The development and morphology of the rat mammary gland are dependent upon several hormones including estrogens, androgens, progesterone, growth hormone and prolactin. In toxicology studies, treatment with xenobiotics may alter these hormones resulting in changes in the morphology of reproductive tissues such as the mammary gland. In the rat, male and female mammary glands exhibit striking morphologic differences that can be altered secondary to hormonal perturbations. Recognizing these morphologic changes can help the pathologist predict potential xenobiotic-induced perturbations in the systemic hormonal milieu. This review examines the development of the rat mammary gland and the influence of sex hormones on the morphology of the adult male and female rat mammary gland. Specific case examples from the literature and data from our laboratory highlight the dynamic nature of the rat mammary gland in response to hormonal changes.

Androgen dependent mammary gland virilism in rats given the selective estrogen receptor modulator LY2066948 hydrochloride.

A selective estrogen receptor modulator (SERM) is a nonsteroidal compound with tissue specific estrogen receptor (ER) agonist or antagonist activities. In animals, SERMs may produce morphologic changes in hormonally-sensitive tissues like the mammary gland. Mammary glands from female rats given the SERM LY2066948 hydrochloride (LY2066948) for 1 month at >or= 175 mg/kg had intralobular ducts and alveoli lined by multiple layers of vacuolated, hypertrophied epithelial cells, resembling in part the morphology of the normal male rat mammary gland. We hypothesized that these SERM-mediated changes represented an androgen-dependent virilism of the female rat mammary gland. To test this hypothesis, the androgen receptor antagonist flutamide was co-administered with LY2066948 (175 mg/kg) to female rats for 1 month. Female rats given SERM alone had hyperandrogenemia and the duct and alveolar changes described here. Flutamide cotreatment did not affect serum androgen levels but completely blocked the SERM-mediated mammary gland change. In the mouse, a species that does not have the sex-specific differences in the mammary gland observed in the rat, SERM treatment resulted in hyperandrogenemia but did not alter mammary gland morphology. These studies demonstrate that LY2066948 produces species-specific, androgen-dependent mammary gland virilism in the female rat.

A selective estrogen receptor modulator designed for the treatment of uterine leiomyoma with unique tissue specificity for uterus and ovaries in rats.

The design of a novel selective estrogen receptor modulator (SERM) for the potential treatment of uterine leiomyoma is described. 16 (LY2066948-HCl) binds with high affinity to estrogen receptors alpha and beta (ERalpha and ERbeta, respectively) and is a potent uterine antagonist with minimal effects on the ovaries as determined by serum biomarkers and histologic evaluation.

A new selective estrogen receptor modulator with potent uterine antagonist activity, agonist activity in bone, and minimal ovarian stimulation.

The use of selective estrogen receptor modulators for the treatment of estrogen-dependent diseases in premenopausal women has been hindered by undesirable ovarian stimulation and associated risks of ovarian cysts. We have identified a selective estrogen receptor modulator compound (LY2066948) that is a strong estrogen antagonist in the uterus yet has minimal effects on the ovaries of rats. LY2066948 binds with high affinity to both estrogen receptors and has potent estrogen antagonist activity in human uterine and breast cancer cells. Oral administration of LY2066948 to immature rats blocked uterine weight gain induced by ethynyl estradiol with an ED50 of 0.07 mg/kg. Studies in mature rats demonstrated that LY2066948 decreases uterine weight by 51% after 35 d treatment, confirming potent uterine antagonist activity over several estrous cycles. This strong uterine response contrasted with the minimal effects on the ovaries: serum estradiol levels remained within the normal range, whereas histologic evaluation showed granulosa cell hyperplasia in few of the rats. Bone studies demonstrated that LY2066948 prevented ovariectomy-induced bone loss and treatment of ovary-intact rats caused no bone loss, confirming estrogen receptor agonist skeletal effects. Collectively, these data show that LY2066948 exhibits a tissue-specific profile consistent with strong antagonist activity in the uterus, agonist activity in bone, and minimal effects in the ovaries.

Epididymal and systemic phospholipidosis in rats and dogs treated with the dopamine D3 selective antagonist PNU-177864.

Repeat dose oral toxicity studies were conducted in rat and dog to assess the safety for human clinical testing of the potent dopamine D3 receptor antagonist, PNU-177864. Systemic phospholipidosis was the principal treatment-related change with epididymal epithelial cell phospholipidosis being the most prominent finding in rats and dogs. Epididymal epithelial cells had no histologic evidence of degeneration; sperm density and morphology were normal histologically in both species; and sperm concentration, morphology, and motility in the dog were comparable to dogs given vehicle. Other sites with phospholipidosis included lymphoid tissues (lymph nodes, Peyer's patches, and/or spleen), pulmonary alveolar macrophages, and peripheral blood lymphocytes in rats and dogs and adrenal cortex, liver, pituitary, hair follicles, bone marrow lymphocytes and plasma cells, and skeletal muscle in rats only. The phospholipidosis was resolved after a 6-week recovery period in all tissues but epididymis. There was no evidence of cell injury in tissues that had phospholipid accumulations except in rat skeletal muscle that had multifocal myofiber degeneration and necrosis. For clinical trials, serum AST and CK and peripheral blood lymphocyte vacuolation were considered potential safety biomarkers for skeletal muscle degeneration and phospholipidosis, respectively.