Development of deep learning models for microglia analyses in brain tissue using DeePathology™ STUDIO

Targeted therapy and deep learning insights into microglia modulation for spinal cord injury

Spinal cord injury (SCI) is a devastating condition that can cause significant motor and sensory impairment. Microglia, the central nervous system's immune sentinels, are known to be promising therapeutic targets in both SCI and neurodegenerative diseases. The most effective way to deliver medications and control microglial inflammation is through nanovectors; however, because of the variability in microglial morphology and the lack of standardized techniques, it is still difficult to precisely measure their activation in preclinical models. This problem is especially important in SCI, where the intricacy of the glia response following traumatic events necessitates the use of a sophisticated method to automatically discern between various microglial cell activation states that vary over time and space as the secondary injury progresses. We address this issue by proposing a deep learning-based technique for quantifying microglial activation following drug-loaded nanovector treatment in a preclinical SCI model. Our method uses a convolutional neural network to segment and classify microglia based on morphological characteristics. Our approach's accuracy and efficiency are demonstrated through evaluation on a collection of histology pictures from injured and intact spinal cords. This robust computational technique has potential for analyzing microglial activation across various neuropathologies and demonstrating the usefulness of nanovectors in modifying microglia in SCI and other neurological disorders. It has the ability to speed development in this crucial sector by providing a standardized and objective way to compare therapeutic options.

Physiological expression of mutated TAU impaired astrocyte activity and exacerbates β-amyloid pathology in 5xFAD mice

BACKGROUND

Alzheimer's disease (AD) is the leading cause of dementia in the world. The pathology of AD is affiliated with the elevation of both tau (τ) and β-amyloid (Aβ) pathologies. Yet, the direct link between natural τ expression on glia cell activity and Aβ remains unclear. While experiments in mouse models suggest that an increase in Aβ exacerbates τ pathology when expressed under a neuronal promoter, brain pathology from AD patients suggests an appearance of τ pathology in regions without Aβ.

METHODS

Here, we aimed to assess the link between τ and Aβ using a new mouse model that was generated by crossing a mouse model that expresses two human mutations of the human MAPT under a mouse Tau natural promoter with 5xFAD mice that express human mutated APP and PS1 in neurons.

RESULTS

The new mouse model, called 5xFAD TAU, shows accelerated cognitive impairment at 2 months of age, increased number of Aβ depositions at 4 months and neuritic plaques at 6 months of age. An expression of human mutated TAU in astrocytes leads to a dystrophic appearance and reduces their ability to engulf Aβ, which leads to an increased brain Aβ load. Astrocytes expressing mutated human TAU showed an impairment in the expression of vascular endothelial growth factor (VEGF) that has previously been suggested to play an important role in supporting neurons.

CONCLUSIONS

Our results suggest the role of τ in exacerbating Aβ pathology in addition to pointing out the potential role of astrocytes in disease progression. Further research of the crosstalk between τ and Aβ in astrocytes may increase our understanding of the role glia cells have in the pathology of AD with the aim of identifying novel therapeutic interventions to an otherwise currently incurable disease.

Artificial intelligence in liver cancers: Decoding the impact of machine learning models in clinical diagnosis of primary liver cancers and liver cancer metastases

Liver cancers are the fourth leading cause of cancer-related mortality worldwide. In the past decade, breakthroughs in the field of artificial intelligence (AI) have inspired development of algorithms in the cancer setting. A growing body of recent studies have evaluated machine learning (ML) and deep learning (DL) algorithms for pre-screening, diagnosis and management of liver cancer patients through diagnostic image analysis, biomarker discovery and predicting personalized clinical outcomes. Despite the promise of these early AI tools, there is a significant need to explain the 'black box' of AI and work towards deployment to enable ultimate clinical translatability. Certain emerging fields such as RNA nanomedicine for targeted liver cancer therapy may also benefit from application of AI, specifically in nano-formulation research and development given that they are still largely reliant on lengthy trial-and-error experiments. In this paper, we put forward the current landscape of AI in liver cancers along with the challenges of AI in liver cancer diagnosis and management. Finally, we have discussed the future perspectives of AI application in liver cancer and how a multidisciplinary approach using AI in nanomedicine could accelerate the transition of personalized liver cancer medicine from bench side to the clinic.

Code-free machine learning for classification of central nervous system histopathology images

Machine learning (ML), an application of artificial intelligence, is currently transforming the analysis of biomedical data and specifically of biomedical images including histopathology. The promises of this technology contrast, however, with its currently limited application in routine clinical practice. This discrepancy is in part due to the extent of informatics expertise typically required for implementation of ML. Therefore, we assessed the suitability of 2 publicly accessible code-free ML platforms (Microsoft Custom Vision and Google AutoML), for classification of histopathological images of diagnostic central nervous system tissue samples. When trained with typically 100 to more than 1000 images, both systems were able to perform nontrivial classifications (glioma vs brain metastasis; astrocytoma vs astrocytosis, prediction of 1p/19q co-deletion in IDH-mutant tumors) based on hematoxylin and eosin-stained images with high accuracy (from ∼80% to nearly 100%). External validation of the predicted accuracy and negative control experiments were found to be crucial for verification of the accuracy predicted by the algorithms. Furthermore, we propose a possible diagnostic workflow for pathologists to implement classification of histopathological images based on code-free machine platforms.

Time- and Sex-Dependent Effects of Fingolimod Treatment in a Mouse Model of Alzheimer's Disease

Alzheimer's disease (AD) is the most common cause of dementia. Fingolimod has previously shown beneficial effects in different animal models of AD. However, it has shown contradictory effects when it has been applied at early disease stages. Our objective was to evaluate fingolimod in two different treatment paradigms. To address this aim, we treated male and female APP-transgenic mice for 50 days, starting either before plaque deposition at 50 days of age (early) or at 125 days of age (late). To evaluate the effects, we investigated the neuroinflammatory and glial markers, the Aβ load, and the concentration of the brain-derived neurotrophic factor (BDNF). We found a reduced Aβ load only in male animals in the late treatment paradigm. These animals also showed reduced microglia activation and reduced IL-1β. No other treatment group showed any difference in comparison to the controls. On the other hand, we detected a linear correlation between BDNF and the brain Aβ concentrations. The fingolimod treatment has shown beneficial effects in AD models, but the outcome depends on the neuroinflammatory state at the start of the treatment. Thus, according to our data, a fingolimod treatment would be effective after the onset of the first AD symptoms, mainly affecting the neuroinflammatory reaction to the ongoing Aβ deposition.

Annotating for Artificial Intelligence Applications in Digital Pathology: A Practical Guide for Pathologists and Researchers

Training machine learning models for artificial intelligence (AI) applications in pathology often requires extensive annotation by human experts, but there is little guidance on the subject. In this work, we aimed to describe our experience and provide a simple, useful, and practical guide addressing annotation strategies for AI development in computational pathology. Annotation methodology will vary significantly depending on the specific study's objectives, but common difficulties will be present across different settings. We summarize key aspects and issue guiding principles regarding team interaction, ground-truth quality assessment, different annotation types, and available software and hardware options and address common difficulties while annotating. This guide was specifically designed for pathology annotation, intending to help pathologists, other researchers, and AI developers with this process.

A Novel Huntington's Disease Assessment Platform to Support Future Drug Discovery and Development

Huntington's disease (HD) is a lethal neurodegenerative disorder without efficient therapeutic options. The inefficient translation from preclinical and clinical research into clinical use is mainly attributed to the lack of (i) understanding of disease initiation, progression, and involved molecular mechanisms; (ii) knowledge of the possible HD target space and general data awareness; (iii) detailed characterizations of available disease models; (iv) better suitable models; and (v) reliable and sensitive biomarkers. To generate robust HD-like symptoms in a mouse model, the neomycin resistance cassette was excised from zQ175 mice, generating a new line: zQ175^Δneo. We entirely describe the dynamics of behavioral, neuropathological, and immunohistological changes from 15-57 weeks of age. Specifically, zQ175^Δneo mice showed early astrogliosis from 15 weeks; growth retardation, body weight loss, and anxiety-like behaviors from 29 weeks; motor deficits and reduced muscular strength from 36 weeks; and finally slight microgliosis at 57 weeks of age. Additionally, we collected the entire bioactivity network of small-molecule HD modulators in a multitarget dataset (HD_MDS). Hereby, we uncovered 358 unique compounds addressing over 80 different pharmacological targets and pathways. Our data will support future drug discovery approaches and may serve as useful assessment platform for drug discovery and development against HD.

A novel automated morphological analysis of Iba1+ microglia using a deep learning assisted model

There is growing evidence for the key role of microglial functional state in brain pathophysiology. Consequently, there is a need for efficient automated methods to measure the morphological changes distinctive of microglia functional states in research settings. Currently, many commonly used automated methods can be subject to sample representation bias, time consuming imaging, specific hardware requirements and difficulty in maintaining an accurate comparison across research environments. To overcome these issues, we use commercially available deep learning tools Aiforia^® Cloud (Aifoira Inc., Cambridge, MA, United States) to quantify microglial morphology and cell counts from histopathological slides of Iba1 stained tissue sections. We provide evidence for the effective application of this method across a range of independently collected datasets in mouse models of viral infection and Parkinson's disease. Additionally, we provide a comprehensive workflow with training details and annotation strategies by feature layer that can be used as a guide to generate new models. In addition, all models described in this work are available within the Aiforia^® platform for study-specific adaptation and validation.

Isotope-labeled amyloid-β does not transmit to the brain in a prion-like manner after peripheral administration

Findings of early cerebral amyloid-β deposition in mice after peripheral injection of amyloid-β-containing brain extracts, and in humans following cadaveric human growth hormone treatment raised concerns that amyloid-β aggregates and possibly Alzheimer's disease may be transmissible between individuals. Yet, proof that Aβ actually reaches the brain from the peripheral injection site is lacking. Here, we use a proteomic approach combining stable isotope labeling of mammals and targeted mass spectrometry. Specifically, we generate ¹³ C-isotope-labeled brain extracts from mice expressing human amyloid-β and track ¹³ C-lysine-labeled amyloid-β after intraperitoneal administration into young amyloid precursor protein-transgenic mice. We detect injected amyloid-β in the liver and lymphoid tissues for up to 100 days. In contrast, injected ¹³ C-lysine-labeled amyloid-β is not detectable in the brain whereas the mice incorporate ¹³ C-lysine from the donor brain extracts into endogenous amyloid-β. Using a highly sensitive and specific proteomic approach, we demonstrate that amyloid-β does not reach the brain from the periphery. Our study argues against potential transmissibility of Alzheimer's disease while opening new avenues to uncover mechanisms of pathophysiological protein deposition.

Cyclic multiplex fluorescent immunohistochemistry and machine learning reveal distinct states of astrocytes and microglia in normal aging and Alzheimer's disease

BACKGROUND

Astrocytes and microglia react to Aβ plaques, neurofibrillary tangles, and neurodegeneration in the Alzheimer's disease (AD) brain. Single-nuclei and single-cell RNA-seq have revealed multiple states or subpopulations of these glial cells but lack spatial information. We have developed a methodology of cyclic multiplex fluorescent immunohistochemistry on human postmortem brains and image analysis that enables a comprehensive morphological quantitative characterization of astrocytes and microglia in the context of their spatial relationships with plaques and tangles.

METHODS

Single FFPE sections from the temporal association cortex of control and AD subjects were subjected to 8 cycles of multiplex fluorescent immunohistochemistry, including 7 astroglial, 6 microglial, 1 neuronal, Aβ, and phospho-tau markers. Our analysis pipeline consisted of: (1) image alignment across cycles; (2) background subtraction; (3) manual annotation of 5172 ALDH1L1+ astrocytic and 6226 IBA1+ microglial profiles; (4) local thresholding and segmentation of profiles; (5) machine learning on marker intensity data; and (6) deep learning on image features.

RESULTS

Spectral clustering identified three phenotypes of astrocytes and microglia, which we termed "homeostatic," "intermediate," and "reactive." Reactive and, to a lesser extent, intermediate astrocytes and microglia were closely associated with AD pathology (≤ 50 µm). Compared to homeostatic, reactive astrocytes contained substantially higher GFAP and YKL-40, modestly elevated vimentin and TSPO as well as EAAT1, and reduced GS. Intermediate astrocytes had markedly increased EAAT2, moderately increased GS, and intermediate GFAP and YKL-40 levels. Relative to homeostatic, reactive microglia showed increased expression of all markers (CD68, ferritin, MHC2, TMEM119, TSPO), whereas intermediate microglia exhibited increased ferritin and TMEM119 as well as intermediate CD68 levels. Machine learning models applied on either high-plex signal intensity data (gradient boosting machines) or directly on image features (convolutional neural networks) accurately discriminated control vs. AD diagnoses at the single-cell level.

CONCLUSIONS

Cyclic multiplex fluorescent immunohistochemistry combined with machine learning models holds promise to advance our understanding of the complexity and heterogeneity of glial responses as well as inform transcriptomics studies. Three distinct phenotypes emerged with our combination of markers, thus expanding the classic binary "homeostatic vs. reactive" classification to a third state, which could represent "transitional" or "resilient" glia.

A New Tool for the Analysis of the Effect of Intracerebrally Injected Anti-Amyloid-β Compounds

BACKGROUND

A wide range of techniques has been developed over the past decades to characterize amyloid-β (Aβ) pathology in mice. Until now, no method has been established to quantify spatial changes in Aβ plaque deposition due to targeted delivery of substances using ALZET® pumps.

OBJECTIVE

Development of a methodology to quantify the local distribution of Aβ plaques after intracerebral infusion of compounds.

METHODS

We have developed a toolbox to quantify Aβ plaques in relation to intracerebral injection channels using Zeiss AxioVision® and Microsoft Excel® software. For the proof of concept, intracerebral stereotactic surgery was performed in 50-day-old APP-transgenic mice injected with PBS. At the age of 100 days, brains were collected for immunhistological analysis.

RESULTS

The toolbox can be used to analyze and evaluate Aβ plaques (number, size, and coverage) in specific brain areas based on their location relative to the point of the injection or the injection channel. The tool provides classification of Aβ plaques in pre-defined distance groups using two different approaches.

CONCLUSION

This new analytic toolbox facilitates the analysis of long-term continuous intracerebral experimental compound infusions using ALZET® pumps. This method generates reliable data for Aβ deposition characterization in relation to the distribution of experimental compounds.