PathoGraph: An Attention-Based Graph Neural Network Capable of Prognostication Based on CD276 Labelling of Malignant Glioma Cells.

Computerized methods have been developed that allow quantitative morphological analyses of whole slide images (WSIs), e.g., of immunohistochemical stains. The latter are attractive because they can provide high-resolution data on the distribution of proteins in tissue. However, many immunohistochemical results are complex because the protein of interest occurs in multiple locations (in different cells and also extracellularly). We have recently established an artificial intelligence framework, PathoFusion which utilises a bifocal convolutional neural network (BCNN) model for detecting and counting arbitrarily definable morphological structures. We have now complemented this model by adding an attention-based graph neural network (abGCN) for the advanced analysis and automated interpretation of such data. Classical convolutional neural network (CNN) models suffer from limitations when handling global information. In contrast, our abGCN is capable of creating a graph representation of cellular detail from entire WSIs. This abGCN method combines attention learning with visualisation techniques that pinpoint the location of informative cells and highlight cell-cell interactions. We have analysed cellular labelling for CD276, a protein of great interest in cancer immunology and a potential marker of malignant glioma cells/putative glioma stem cells (GSCs). We are especially interested in the relationship between CD276 expression and prognosis. The graphs permit predicting individual patient survival on the basis of GSC community features. Our experiments lay a foundation for the use of the BCNN-abGCN tool chain in automated diagnostic prognostication using immunohistochemically labelled histological slides, but the method is essentially generic and potentially a widely usable tool in medical research and AI based healthcare applications.

Neuroinflammation: The Abused Concept.

Scientific progress requires the relentless correction of errors and refinement of hypotheses. Clarity of terminology is essential for clarity of thought and proper experimental interrogation of nature. Therefore, the application of the same scientific term to different and even conflicting phenomena and concepts is not useful and must be corrected. Such abuse of terminology has happened and is still increasing in the case of "neuroinflammation," a term that until the 1990s meant classical inflammation affecting the central nervous system (CNS) and thereon was progressively used to mostly denote microglia activation. The resulting confusion is very wasteful and detrimental not only for scientists but also for patients, given the numerous failed clinical trials in acute and chronic CNS diseases over the last decade with "anti-inflammatory" drugs. Despite this failure, reassessments of the "neuroinflammation" concept are rare, especially considering the number of articles still using the term. This undesirable situation motivates this article. We review the origins and evolution of the term "neuroinflammation," discuss the unique tissue defense and repair strategies in the CNS, define CNS immunity, and emphasize the notion of gliopathies to help readdress, if not bury, the term "neuroinflammation" as it stands in the way of scientific progress.

Artificial intelligence techniques for neuropathological diagnostics and research.

Artificial intelligence (AI) research began in theoretical neurophysiology, and the resulting classical paper on the McCulloch-Pitts mathematical neuron was written in a psychiatry department almost 80 years ago. However, the application of AI in digital neuropathology is still in its infancy. Rapid progress is now being made, which prompted this article. Human brain diseases represent distinct system states that fall outside the normal spectrum. Many differ not only in functional but also in structural terms, and the morphology of abnormal nervous tissue forms the traditional basis of neuropathological disease classifications. However, only a few countries have the medical specialty of neuropathology, and, given the sheer number of newly developed histological tools that can be applied to the study of brain diseases, a tremendous shortage of qualified hands and eyes at the microscope is obvious. Similarly, in neuroanatomy, human observers no longer have the capacity to process the vast amounts of connectomics data. Therefore, it is reasonable to assume that advances in AI technology and, especially, whole-slide image (WSI) analysis will greatly aid neuropathological practice. In this paper, we discuss machine learning (ML) techniques that are important for understanding WSI analysis, such as traditional ML and deep learning, introduce a recently developed neuropathological AI termed PathoFusion, and present thoughts on some of the challenges that must be overcome before the full potential of AI in digital neuropathology can be realized.

Microglia and Brain Macrophages as Drivers of Glioma Progression.

Evidence is accumulating that the tumour microenvironment (TME) has a key role in the progression of gliomas. Non-neoplastic cells in addition to the tumour cells are therefore finding increasing attention. Microglia and other glioma-associated macrophages are at the centre of this interest especially in the context of therapeutic considerations. New ideas have emerged regarding the role of microglia and, more recently, blood-derived brain macrophages in glioblastoma (GBM) progression. We are now beginning to understand the mechanisms that allow malignant glioma cells to weaken microglia and brain macrophage defence mechanisms. Surface molecules and cytokines have a prominent role in microglia/macrophage-glioma cell interactions, and we discuss them in detail. The involvement of exosomes and microRNAs forms another focus of this review. In addition, certain microglia and glioma cell pathways deserve special attention. These "synergistic" (we suggest calling them "Janus") pathways are active in both glioma cells and microglia/macrophages where they act in concert supporting malignant glioma progression. Examples include CCN4 (WISP1)/Integrin α6ß1/Akt and CHI3L1/PI3K/Akt/mTOR. They represent attractive therapeutic targets.

Microglia morphophysiological diversity and its implications for the CNS.

Microglia are mononuclear phagocytes of mesodermal origin that migrate to the central nervous system (CNS) during the early stages of embryonic development. After colonizing the CNS, they proliferate and remain able to self-renew throughout life, maintaining the number of microglia around 5-12% of the cells in the CNS parenchyma. They are considered to play key roles in development, homeostasis and innate immunity of the CNS. Microglia are exceptionally diverse in their morphological characteristics, actively modifying the shape of their processes and soma in response to different stimuli. This broad morphological spectrum of microglia responses is considered to be closely correlated to their diverse range of functions in health and disease. However, the morphophysiological attributes of microglia, and the structural and functional features of microglia-neuron interactions, remain largely unknown. Here, we assess the current knowledge of the diverse microglial morphologies, with a focus on the correlation between microglial shape and function. We also outline some of the current challenges, opportunities, and future directions that will help us to tackle unanswered questions about microglia, and to continue unravelling the mysteries of microglia, in all its shapes.

Microglia states and nomenclature: A field at its crossroads.

Microglial research has advanced considerably in recent decades yet has been constrained by a rolling series of dichotomies such as "resting versus activated" and "M1 versus M2." This dualistic classification of good or bad microglia is inconsistent with the wide repertoire of microglial states and functions in development, plasticity, aging, and diseases that were elucidated in recent years. New designations continuously arising in an attempt to describe the different microglial states, notably defined using transcriptomics and proteomics, may easily lead to a misleading, although unintentional, coupling of categories and functions. To address these issues, we assembled a group of multidisciplinary experts to discuss our current understanding of microglial states as a dynamic concept and the importance of addressing microglial function. Here, we provide a conceptual framework and recommendations on the use of microglial nomenclature for researchers, reviewers, and editors, which will serve as the foundations for a future white paper.

An Open-Source AI Framework for the Analysis of Single Cells in Whole-Slide Images with a Note on CD276 in Glioblastoma.

Routine examination of entire histological slides at cellular resolution poses a significant if not insurmountable challenge to human observers. However, high-resolution data such as the cellular distribution of proteins in tissues, e.g., those obtained following immunochemical staining, are highly desirable. Our present study extends the applicability of the PathoFusion framework to the cellular level. We illustrate our approach using the detection of CD276 immunoreactive cells in glioblastoma as an example. Following automatic identification by means of PathoFusion's bifocal convolutional neural network (BCNN) model, individual cells are automatically profiled and counted. Only discriminable cells selected through data filtering and thresholding were segmented for cell-level analysis. Subsequently, we converted the detection signals into the corresponding heatmaps visualizing the distribution of the detected cells in entire whole-slide images of adjacent H&E-stained sections using the Discrete Wavelet Transform (DWT). Our results demonstrate that PathoFusion is capable of autonomously detecting and counting individual immunochemically labelled cells with a high prediction performance of 0.992 AUC and 97.7% accuracy. The data can be used for whole-slide cross-modality analyses, e.g., relationships between immunochemical signals and anaplastic histological features. PathoFusion has the potential to be applied to additional problems that seek to correlate heterogeneous data streams and to serve as a clinically applicable, weakly supervised system for histological image analyses in (neuro)pathology.

Cytokine Signalling at the Microglial Penta-Partite Synapse.

Microglial cell processes form part of a subset of synaptic contacts that have been dubbed microglial tetra-partite or quad-partite synapses. Since tetrapartite may also refer to the presence of extracellular matrix components, we propose the more precise term microglial penta-partite synapse for synapses that show a microglial cell process in close physical proximity to neuronal and astrocytic synaptic constituents. Microglial cells are now recognised as key players in central nervous system (CNS) synaptic changes. When synaptic plasticity involving microglial penta-partite synapses occurs, microglia may utilise their cytokine arsenal to facilitate the generation of new synapses, eliminate those that are not needed anymore, or modify the molecular and structural properties of the remaining synaptic contacts. In addition, microglia-synapse contacts may develop de novo under pathological conditions. Microglial penta-partite synapses have received comparatively little attention as unique sites in the CNS where microglial cells, cytokines and other factors they release have a direct influence on the connections between neurons and their function. It concerns our understanding of the penta-partite synapse where the confusion created by the term "neuroinflammation" is most counterproductive. The mere presence of activated microglia or the release of their cytokines may occur independent of inflammation, and penta-partite synapses are not usually active in a neuroimmunological sense. Clarification of these details is the main purpose of this review, specifically highlighting the relationship between microglia, synapses, and the cytokines that can be released by microglial cells in health and disease.

Ground state depletion microscopy as a tool for studying microglia-synapse interactions.

Ground state depletion followed by individual molecule return microscopy (GSDIM) has been used in the past to study the nanoscale distribution of protein co-localization in living cells. We now demonstrate the successful application of GSDIM to archival human brain tissue sections including from Alzheimer's disease cases as well as experimental tissue samples from mouse and zebrafish larvae. Presynaptic terminals and microglia and their cell processes were visualized at a resolution beyond diffraction-limited light microscopy, allowing clearer insights into their interactions in situ. The procedure described here offers time and cost savings compared to electron microscopy and opens the spectrum of molecular imaging using antibodies and super-resolution microscopy to the analysis of routine formalin-fixed paraffin sections of archival human brain. The investigation of microglia-synapse interactions in dementia will be of special interest in this context.

PathoFusion: An Open-Source AI Framework for Recognition of Pathomorphological Features and Mapping of Immunohistochemical Data.

We have developed a platform, termed PathoFusion, which is an integrated system for marking, training, and recognition of pathological features in whole-slide tissue sections. The platform uses a bifocal convolutional neural network (BCNN) which is designed to simultaneously capture both index and contextual feature information from shorter and longer image tiles, respectively. This is analogous to how a microscopist in pathology works, identifying a cancerous morphological feature in the tissue context using first a narrow and then a wider focus, hence bifocal. Adjacent tissue sections obtained from glioblastoma cases were processed for hematoxylin and eosin (H&E) and immunohistochemical (CD276) staining. Image tiles cropped from the digitized images based on markings made by a consultant neuropathologist were used to train the BCNN. PathoFusion demonstrated its ability to recognize malignant neuropathological features autonomously and map immunohistochemical data simultaneously. Our experiments show that PathoFusion achieved areas under the curve (AUCs) of 0.985 ± 0.011 and 0.988 ± 0.001 in patch-level recognition of six typical pathomorphological features and detection of associated immunoreactivity, respectively. On this basis, the system further correlated CD276 immunoreactivity to abnormal tumor vasculature. Corresponding feature distributions and overlaps were visualized by heatmaps, permitting high-resolution qualitative as well as quantitative morphological analyses for entire histological slides. Recognition of more user-defined pathomorphological features can be added to the system and included in future tissue analyses. Integration of PathoFusion with the day-to-day service workflow of a (neuro)pathology department is a goal. The software code for PathoFusion is made publicly available.

Synapses, Microglia, and Lipids in Alzheimer's Disease.

Alzheimer's disease (AD) is characterised by synaptic dysfunction accompanied by the microscopically visible accumulation of pathological protein deposits and cellular dystrophy involving both neurons and glia. Late-stage AD shows pronounced loss of synapses and neurons across several differentially affected brain regions. Recent studies of advanced AD using post-mortem brain samples have demonstrated the direct involvement of microglia in synaptic changes. Variants of the Apolipoprotein E and Triggering Receptors Expressed on Myeloid Cells gene represent important determinants of microglial activity but also of lipid metabolism in cells of the central nervous system. Here we review evidence that may help to explain how abnormal lipid metabolism, microglial activation, and synaptic pathophysiology are inter-related in AD.

Automated Machine-Learning Framework Integrating Histopathological and Radiological Information for Predicting IDH1 Mutation Status in Glioma.

Diffuse gliomas are the most common malignant primary brain tumors. Identification of isocitrate dehydrogenase 1 (IDH1) mutations aids the diagnostic classification of these tumors and the prediction of their clinical outcomes. While histology continues to play a key role in frozen section diagnosis, as a diagnostic reference and as a method for monitoring disease progression, recent research has demonstrated the ability of multi-parametric magnetic resonance imaging (MRI) sequences for predicting IDH genotypes. In this paper, we aim to improve the prediction accuracy of IDH1 genotypes by integrating multi-modal imaging information from digitized histopathological data derived from routine histological slide scans and the MRI sequences including T1-contrast (T1) and Fluid-attenuated inversion recovery imaging (T2-FLAIR). In this research, we have established an automated framework to process, analyze and integrate the histopathological and radiological information from high-resolution pathology slides and multi-sequence MRI scans. Our machine-learning framework comprehensively computed multi-level information including molecular level, cellular level, and texture level information to reflect predictive IDH genotypes. Firstly, an automated pre-processing was developed to select the regions of interest (ROIs) from pathology slides. Secondly, to interactively fuse the multimodal complementary information, comprehensive feature information was extracted from the pathology ROIs and segmented tumor regions (enhanced tumor, edema and non-enhanced tumor) from MRI sequences. Thirdly, a Random Forest (RF)-based algorithm was employed to identify and quantitatively characterize histopathological and radiological imaging origins, respectively. Finally, we integrated multi-modal imaging features with a machine-learning algorithm and tested the performance of the framework for IDH1 genotyping, we also provided visual and statistical explanation to support the understanding on prediction outcomes. The training and testing experiments on 217 pathologically verified IDH1 genotyped glioma cases from multi-resource validated that our fully automated machine-learning model predicted IDH1 genotypes with greater accuracy and reliability than models that were based on radiological imaging data only. The accuracy of IDH1 genotype prediction was 0.90 compared to 0.82 for radiomic result. Thus, the integration of multi-parametric imaging features for automated analysis of cross-modal biomedical data improved the prediction accuracy of glioma IDH1 genotypes.

Author Correction: Selective, high-contrast detection of syngeneic glioblastoma in vivo.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Selective, high-contrast detection of syngeneic glioblastoma in vivo.

Glioblastoma is a highly malignant, largely therapy-resistant brain tumour. Deep infiltration of brain tissue by neoplastic cells represents the key problem of diffuse glioma. Much current research focuses on the molecular makeup of the visible tumour mass rather than the cellular interactions in the surrounding brain tissue infiltrated by the invasive glioma cells that cause the tumour's ultimately lethal outcome. Diagnostic neuroimaging that enables the direct in vivo observation of the tumour infiltration zone and the local host tissue responses at a preclinical stage are important for the development of more effective glioma treatments. Here, we report an animal model that allows high-contrast imaging of wild-type glioma cells by positron emission tomography (PET) using [18 F]PBR111, a selective radioligand for the mitochondrial 18 kDa Translocator Protein (TSPO), in the Tspo-/- mouse strain (C57BL/6-Tspotm1GuMu(GuwiyangWurra)). The high selectivity of [18 F]PBR111 for the TSPO combined with the exclusive expression of TSPO in glioma cells infiltrating into null-background host tissue free of any TSPO expression, makes it possible, for the first time, to unequivocally and with uniquely high biological contrast identify peri-tumoral glioma cell invasion at preclinical stages in vivo. Comparison of the in vivo imaging signal from wild-type glioma cells in a null background with the signal in a wild-type host tissue, where the tumour induces the expected TSPO expression in the host's glial cells, illustrates the substantial extent of the peritumoral host response to the growing tumour. The syngeneic tumour (TSPO+/+) in null background (TSPO-/-) model is thus well suited to study the interaction of the tumour front with the peri-tumoral tissue, and the experimental evaluation of new therapeutic approaches targeting the invasive behaviour of glioblastoma.

Driving innovation through collaboration: development of clinical annotation datasets for brain cancer biobanking.

BACKGROUND: A key component of cancer research is the availability of clinical samples with appropriately annotated clinical data. Biobanks facilitate research by collecting/storing various types of clinical samples for research. Brain Cancer Biobanking Australia (BCBA) was established to facilitate the networking of brain cancer biobanking operations Australia-wide. Maximizing biospecimen utility in a networked biobanking environment requires the standardization of procedures and data across different sites. The aim of this research was to scope and develop a recommended clinical annotation dataset both for pediatric and adult brain cancer biobanks. METHODS: A multidisciplinary working group consisting of members from the BCBA Consortium was established to develop clinical dataset recommendations for brain cancer biobanks. A literature search was undertaken to identify any published clinical dataset recommendations for brain cancer biobanks. An audit of data items collected and stored by BCBA member biobanks was also conducted to survey current clinical data collection practices across the BCBA network. RESULTS: BCBA has developed a clinical annotation dataset recommendation for pediatric and adult brain cancer biobanks. The clinical dataset recommendation has 5 clinical data categories: demographic, clinical and radiological diagnosis and surgery, neuropathological diagnosis, patient treatment, and patient follow-up. The data fields have been categorized into 1 of 3 tiers; essential, preferred, and comprehensive. This enables biobanks and researchers to prioritize appropriately where resources are limited. CONCLUSION: This dataset can be used to guide the integration of data from multiple existing biobanks for research studies and for planning prospective brain cancer biobanking activities.

Hippocampal CA2 Lewy pathology is associated with cholinergic degeneration in Parkinson's disease with cognitive decline.

Although the precise neuropathological substrates of cognitive decline in Parkinson's disease (PD) remain elusive, it has long been regarded that pathology in the CA2 hippocampal subfield is characteristic of Lewy body dementias, including dementia in PD (PDD). Early non-human primate tracer studies demonstrated connections from the nucleus of the vertical limb of the diagonal band of Broca (nvlDBB, Ch2) to the hippocampus. However, the relationship between Lewy pathology of the CA2 subfield and cholinergic fibres has not been explored. Therefore, in this study, we investigated the burden of pathology in the CA2 subsector of PD cases with varying degrees of cognitive impairment and correlated this with the extent of septohippocampal cholinergic deficit. Hippocampal sections from 67 PD, 34 PD with mild cognitive impairment and 96 PDD cases were immunostained for tau and alpha-synuclein, and the respective pathology burden was assessed semi-quantitatively. In a subset of cases, the degree of CA2 cholinergic depletion was quantified using confocal microscopy and correlated with cholinergic neuronal loss in Ch2. We found that only cases with dementia have a significantly greater Lewy pathology, whereas cholinergic fibre depletion was evident in cases with mild cognitive impairment and this was significantly correlated with loss of cholinergic neurons in Ch2. In addition, multiple antigen immunofluorescence demonstrated colocalisation between cholinergic fibres and alpha-synuclein but not tau pathology. Such specific Lewy pathology targeting the cholinergic system within the CA2 subfield may contribute to the unique memory retrieval deficit seen in patients with Lewy body disorders, as distinct from the memory storage deficit seen in Alzheimer's disease.

The emerging clinical potential of circulating extracellular vesicles for non-invasive glioma diagnosis and disease monitoring.

Diffuse gliomas (grades II-IV) are amongst the most frequent and devastating primary brain tumours of adults. Currently, patients are monitored by clinical examination and radiographic imaging, which can be challenging to interpret and insensitive to early signs of treatment failure and tumour relapse. While brain biopsy and histologic analysis can evaluate disease progression, serial biopsies are invasive and impractical given the cumulative surgical risk, and may not capture the complete molecular landscape of an evolving tumour. The availability of a minimally invasive 'liquid biopsy' that could assess tumour activity and molecular phenotype in situ has the potential to greatly enhance patient care. Circulating extracellular vesicles (EVs) hold significant promise as robust disease-specific biomarkers accessible in the blood of patients with glioblastoma and other diffuse gliomas. EVs are membrane-bound nanoparticles shed from most if not all cells of the body, and carry DNA, RNA, protein, and lipids that reflect the identity and molecular state of their cell-of-origin. EVs can cross the blood-brain barrier and their release is upregulated in neoplasia. In this review, we describe the current knowledge of EV biology, the role of EVs in glioma biology and the current experience and challenges in profiling glioma-EVs from the circulation.

Nucleo-cytoplasmic transport of TDP-43 studied in real time: impaired microglia function leads to axonal spreading of TDP-43 in degenerating motor neurons.

Transactivating DNA-binding protein-43 (TDP-43) deposits represent a typical finding in almost all ALS patients, more than half of FTLD patients and patients with several other neurodegenerative disorders. It appears that perturbation of nucleo-cytoplasmic transport is an important event in these conditions but the mechanistic role and the fate of TDP-43 during neuronal degeneration remain elusive. We have developed an experimental system for visualising the perturbed nucleocytoplasmic transport of neuronal TDP-43 at the single-cell level in vivo using zebrafish spinal cord. This approach enabled us to image TDP-43-expressing motor neurons before and after experimental initiation of cell death. We report the formation of mobile TDP-43 deposits within degenerating motor neurons, which are normally phagocytosed by microglia. However, when microglial cells were depleted, injury-induced motor neuron degeneration follows a characteristic process that includes TDP-43 redistribution into the cytoplasm, axon and extracellular space. This is the first demonstration of perturbed TDP-43 nucleocytoplasmic transport in vivo, and suggests that impairment in microglial phagocytosis of dying neurons may contribute towards the formation of pathological TDP-43 presentations in ALS and FTLD.

Calcium-axonemal microtubuli interactions underlie mechanism(s) of primary cilia morphological changes.

We have used cell culture of astrocytes aligned within microchannels to investigate calcium effects on primary cilia morphology. In the absence of calcium and in the presence of flow of media (10 µL.s-1) the majority (90%) of primary cilia showed reversible bending with an average curvature of 2.1 ± 0.9 × 10-4 nm-1. When 1.0 mM calcium was present, 90% of cilia underwent bending. Forty percent of these cilia demonstrated strong irreversible bending, resulting in a final average curvature of 3.9 ± 1 × 10-4 nm-1, while 50% of cilia underwent bending similar to that observed during calcium-free flow. The average length of cilia was shifted toward shorter values (3.67 ± 0.34 µm) when exposed to excess calcium (1.0 mM), compared to media devoid of calcium (3.96 ± 0.26 µm). The number of primary cilia that became curved after calcium application was reduced when the cell culture was pre-incubated with 15 µM of the microtubule stabilizer, taxol, for 60 min prior to calcium application. Calcium caused single microtubules to curve at a concentration ≈1.0 mM in vitro, but at higher concentration (≈1.5 mM) multiple microtubule curving occurred. Additionally, calcium causes microtubule-associated protein-2 conformational changes and its dislocation from the microtubule wall at the location of microtubule curvature. A very small amount of calcium, that is 1.45 × 1011 times lower than the maximal capacity of TRPPs calcium channels, may cause gross morphological changes (curving) of primary cilia, while global cytosol calcium levels are expected to remain unchanged. These findings reflect the non-linear manner in which primary cilia may respond to calcium signaling, which in turn may influence the course of development of ciliopathies and cancer.