Regionally defined proteomic profiles of human cerebral tissue and organoids reveal conserved molecular modules of neurodevelopment.

Cerebral organoids have emerged as robust models for neurodevelopmental and pathological processes, as well as a powerful discovery platform for less-characterized neurobiological programs. Toward this prospect, we leverage mass-spectrometry-based proteomics to molecularly profile precursor and neuronal compartments of both human-derived organoids and mid-gestation fetal brain tissue to define overlapping programs. Our analysis includes recovery of precursor-enriched transcriptional regulatory proteins not found to be differentially expressed in previous transcriptomic datasets. To highlight the discovery potential of this resource, we show that RUVBL2 is preferentially expressed in the SOX2-positive compartment of organoids and that chemical inactivation leads to precursor cell displacement and apoptosis. To explore clinicopathological correlates of this cytoarchitectural disruption, we interrogate clinical datasets and identify rare de novo genetic variants involving RUVBL2 in patients with neurodevelopmental impairments. Together, our findings demonstrate how cell-type-specific profiling of organoids can help nominate previously unappreciated genes in neurodevelopment and disease.

Integrating morphologic and molecular histopathological features through whole slide image registration and deep learning.

BACKGROUND: Modern molecular pathology workflows in neuro-oncology heavily rely on the integration of morphologic and immunohistochemical patterns for analysis, classification, and prognostication. However, despite the recent emergence of digital pathology platforms and artificial intelligence-driven computational image analysis tools, automating the integration of histomorphologic information found across these multiple studies is challenged by large files sizes of whole slide images (WSIs) and shifts/rotations in tissue sections introduced during slide preparation. METHODS: To address this, we develop a workflow that couples different computer vision tools including scale-invariant feature transform (SIFT) and deep learning to efficiently align and integrate histopathological information found across multiple independent studies. We highlight the utility and automation potential of this workflow in the molecular subclassification and discovery of previously unappreciated spatial patterns in diffuse gliomas. RESULTS: First, we show how a SIFT-driven computer vision workflow was effective at automated WSI alignment in a cohort of 107 randomly selected surgical neuropathology cases (97/107 (91%) showing appropriate matches, AUC = 0.96). This alignment allows our AI-driven diagnostic workflow to not only differentiate different brain tumor types, but also integrate and carry out molecular subclassification of diffuse gliomas using relevant immunohistochemical biomarkers (IDH1-R132H, ATRX). To highlight the discovery potential of this workflow, we also examined spatial distributions of tumors showing heterogenous expression of the proliferation marker MIB1 and Olig2. This analysis helped uncover an interesting and unappreciated association of Olig2 positive and proliferative areas in some gliomas (r = 0.62). CONCLUSION: This efficient neuropathologist-inspired workflow provides a generalizable approach to help automate a variety of advanced immunohistochemically compatible diagnostic and discovery exercises in surgical neuropathology and neuro-oncology.

Topographic mapping of the glioblastoma proteome reveals a triple-axis model of intra-tumoral heterogeneity.

Glioblastoma is an aggressive form of brain cancer with well-established patterns of intra-tumoral heterogeneity implicated in treatment resistance and progression. While regional and single cell transcriptomic variations of glioblastoma have been recently resolved, downstream phenotype-level proteomic programs have yet to be assigned across glioblastoma's hallmark histomorphologic niches. Here, we leverage mass spectrometry to spatially align abundance levels of 4,794 proteins to distinct histologic patterns across 20 patients and propose diverse molecular programs operational within these regional tumor compartments. Using machine learning, we overlay concordant transcriptional information, and define two distinct proteogenomic programs, MYC- and KRAS-axis hereon, that cooperate with hypoxia to produce a tri-dimensional model of intra-tumoral heterogeneity. Moreover, we highlight differential drug sensitivities and relative chemoresistance in glioblastoma cell lines with enhanced KRAS programs. Importantly, these pharmacological differences are less pronounced in transcriptional glioblastoma subgroups suggesting that this model may provide insights for targeting heterogeneity and overcoming therapy resistance.

The promise of organoids for unraveling the proteomic landscape of the developing human brain.

Cerebral organoids offer an opportunity to bioengineer experimental avatars of the developing human brain and have already begun garnering relevant insights into complex neurobiological processes and disease. Thus far, investigations into their heterogeneous cellular composition and developmental trajectories have been largely limited to transcriptional readouts. Recent advances in global proteomic technologies have enabled a new range of techniques to explore dynamic and non-overlapping spatiotemporal protein-level programs operational in these humanoid neural structures. Here we discuss these early protein-based studies and their potentially essential role for unraveling critical secreted paracrine signals, processes with poor proteogenomic correlations, or neurodevelopmental proteins requiring post-translational modification for biological activity. Integrating emerging proteomic tools with these faithful human-derived neurodevelopmental models could transform our understanding of complex neural cell phenotypes and neurobiological processes, not exclusively driven by transcriptional regulation. These insights, less accessible by exclusive RNA-based approaches, could reveal new knowledge into human brain development and guide improvements in neural regenerative medicine efforts.

Cerebral organoids: emerging ex vivo humanoid models of glioblastoma.

Glioblastoma is an aggressive form of brain cancer that has seen only marginal improvements in its bleak survival outlook of 12-15 months over the last forty years. There is therefore an urgent need for the development of advanced drug screening platforms and systems that can better recapitulate glioblastoma's infiltrative biology, a process largely responsible for its relentless propensity for recurrence and progression. Recent advances in stem cell biology have allowed the generation of artificial tridimensional brain-like tissue termed cerebral organoids. In addition to their potential to model brain development, these reagents are providing much needed synthetic humanoid scaffolds to model glioblastoma's infiltrative capacity in a faithful and scalable manner. Here, we highlight and review the early breakthroughs in this growing field and discuss its potential future role for glioblastoma research.

Unsupervised Resolution of Histomorphologic Heterogeneity in Renal Cell Carcinoma Using a Brain Tumor-Educated Neural Network.

PURPOSE: Applications of deep learning to histopathology have proven capable of expert-level performance, but approaches have largely focused on supervised classification tasks requiring context-specific training and deployment. More generalizable workflows that can be easily shared across subspecialties could help accelerate and broaden adoption. Here, we hypothesized that histology-optimized feature representations, generated by a convolutional neural network (CNN) during supervised learning, are transferable and can resolve meaningful differences in large-scale, discovery-type unsupervised analyses. METHODS: We used a CNN, previously trained to recognize brain tumor histomorphologies, to extract 512 feature representations from > 550 digital whole-slide images (WSIs) of renal cell carcinomas (RCCs) from The Cancer Genome Atlas and other previously unencountered tumors. We use these extracted feature vectors to conduct unsupervised image-set clustering and analyze the clinical and biologic relevance of the intra- and interpatient subgroups generated. RESULTS: Within individual WSIs, feature-based clustering could reliably segment tumor regions and other relevant histopathologic subpatterns (eg, adenosquamous and poorly differentiated regions). Across the larger RCC cohorts, clustering extracted features generated subgroups enriched for clinically relevant subtypes (eg, papillary RCC) and outcomes (eg, survival). Importantly, individual feature activation mapping highlighted salient subtype-specific patterns and features of malignancies (eg, nuclear grade, sarcomatous change) contributing to subgroupings. Moreover, some proposed clusters were enriched for recurring, human-based RCC-subtype misclassifications. CONCLUSION: Our data support that CNNs, pretrained on large histologic datasets, can extend learned representations to novel scenarios and resolve clinically relevant intra- and interpatient tissue-pattern differences without explicit instruction or additional optimization. Repositioning of existing histology-educated networks could provide scalable approaches for image classification, quality assurance, and discovery of unappreciated patterns and subgroups of disease.

Unsupervised Machine Learning in Pathology: The Next Frontier.

Applications of artificial intelligence and particularly deep learning to aid pathologists in carrying out laborious and qualitative tasks in histopathologic image analysis have now become ubiquitous. We introduce and illustrate how unsupervised machine learning workflows can be deployed in existing pathology workflows to begin learning autonomously through exploration and without the need for extensive direction. Although still in its infancy, this type of machine learning, which more closely mirrors human intelligence, stands to add another exciting layer of innovation to computational pathology and accelerate the transition to autonomous pathologic tissue analysis.

Shifts in Ribosome Engagement Impact Key Gene Sets in Neurodevelopment and Ubiquitination in Rett Syndrome.

Regulation of translation during human development is poorly understood, and its dysregulation is associated with Rett syndrome (RTT). To discover shifts in mRNA ribosomal engagement (RE) during human neurodevelopment, we use parallel translating ribosome affinity purification sequencing (TRAP-seq) and RNA sequencing (RNA-seq) on control and RTT human induced pluripotent stem cells, neural progenitor cells, and cortical neurons. We find that 30% of transcribed genes are translationally regulated, including key gene sets (neurodevelopment, transcription and translation factors, and glycolysis). Approximately 35% of abundant intergenic long noncoding RNAs (lncRNAs) are ribosome engaged. Neurons translate mRNAs more efficiently and have longer 3' UTRs, and RE correlates with elements for RNA-binding proteins. RTT neurons have reduced global translation and compromised mTOR signaling, and >2,100 genes are translationally dysregulated. NEDD4L E3-ubiquitin ligase is translationally impaired, ubiquitinated protein levels are reduced, and protein targets accumulate in RTT neurons. Overall, the dynamic translatome in neurodevelopment is disturbed in RTT and provides insight into altered ubiquitination that may have therapeutic implications.

Physician perspectives on integration of artificial intelligence into diagnostic pathology.

Advancements in computer vision and artificial intelligence (AI) carry the potential to make significant contributions to health care, particularly in diagnostic specialties such as radiology and pathology. The impact of these technologies on physician stakeholders is the subject of significant speculation. There is however a dearth of information regarding the opinions, enthusiasm, and concerns of the pathology community at large. Here, we report results from a survey of 487 pathologist-respondents practicing in 54 countries, conducted to examine perspectives on AI implementation in clinical practice. Despite limitations, including difficulty with quantifying response bias and verifying identity of respondents to this anonymous and voluntary survey, several interesting findings were uncovered. Overall, respondents carried generally positive attitudes towards AI, with nearly 75% reporting interest or excitement in AI as a diagnostic tool to facilitate improvements in workflow efficiency and quality assurance in pathology. Importantly, even within the more optimistic cohort, a significant number of respondents endorsed concerns about AI, including the potential for job displacement and replacement. Overall, around 80% of respondents predicted the introduction of AI technology in the pathology laboratory within the coming decade. Attempts to identify statistically significant demographic characteristics (e.g., age, sex, type/place of practice) predictive of attitudes towards AI using Kolmogorov-Smirnov (KS) testing revealed several associations. Important themes which were commented on by respondents included the need for increasing efforts towards physician training and resolving medical-legal implications prior to the generalized implementation of AI in pathology.

Defining Protein Pattern Differences Among Molecular Subtypes of Diffuse Gliomas Using Mass Spectrometry.

Molecular characterization of diffuse gliomas has thus far largely focused on genomic and transcriptomic interrogations. Here, we utilized mass spectrometry and overlay protein-level information onto genomically defined cohorts of diffuse gliomas to improve our downstream molecular understanding of these lethal malignancies. Bulk and macrodissected tissues were utilized to quantitate 5,496 unique proteins over three glioma cohorts subclassified largely based on their IDH and 1p19q codeletion status (IDH wild type (IDHwt), n = 7; IDH mutated (IDHmt), 1p19q non-codeleted, n = 7; IDH mutated, 1p19q-codeleted, n = 10). Clustering analysis highlighted proteome and systems-level pathway differences in gliomas according to IDH and 1p19q-codeletion status, including 287 differentially abundant proteins in macrodissection-enriched tumor specimens. IDHwt tumors were enriched for proteins involved in invasiveness and epithelial to mesenchymal transition (EMT), while IDHmt gliomas had increased abundances of proteins involved in mRNA splicing. Finally, these abundance changes were compared with IDH-matched GBM stem-like cells (GSCs) to better pinpoint protein patterns enriched in putative cellular drivers of gliomas. Using this integrative approach, we outline specific proteins involved in chloride transport (e.g. chloride intracellular channel 1, CLIC1) and EMT (e.g. procollagen-lysine, 2-oxoglutarate 5-dioxygenase 3, PLOD3, and serpin peptidase inhibitor clade H member 1, SERPINH1) that showed concordant IDH-status-dependent abundance differences in both primary tissue and purified GSC cultures. Given the downstream position proteins occupy in driving biology and phenotype, understanding the proteomic patterns operational in distinct glioma subtypes could help propose more specific, personalized, and effective targets for the management of patients with these aggressive malignancies.

Proteomic analysis of meningiomas reveals clinically distinct molecular patterns.

BACKGROUND: Meningiomas represent one of the most common brain tumors and exhibit a clinically heterogeneous behavior, sometimes difficult to predict with classic histopathologic features. While emerging molecular profiling efforts have linked specific genomic drivers to distinct clinical patterns, the proteomic landscape of meningiomas remains largely unexplored. METHODS: We utilize liquid chromatography tandem mass spectrometry with an Orbitrap mass analyzer to quantify global protein abundances of a clinically well-annotated formalin-fixed paraffin embedded (FFPE) cohort (n = 61) of meningiomas spanning all World Health Organization (WHO) grades and various degrees of clinical aggressiveness. RESULTS: In total, we quantify 3042 unique proteins comparing patterns across different clinical parameters. Unsupervised clustering analysis highlighted distinct proteomic (n = 106 proteins, Welch's t-test, P < 0.01) and pathway-level (eg, Notch and PI3K/AKT/mTOR) differences between convexity and skull base meningiomas. Supervised comparative analyses of different pathological grades revealed distinct patterns between benign (grade I) and atypical/malignant (grades II‒III) meningiomas with specific oncogenes enriched in higher grade lesions. Independent of WHO grade, clinically aggressive meningiomas that rapidly recurred (<3 y) had distinctive protein patterns converging on mRNA processing and impaired activation of the matrisome complex. Larger sized meningiomas (>3 cm maximum tumor diameter) and those with previous radiation exposure revealed perturbed pro-proliferative (eg, epidermal growth factor receptor) and metabolic as well as inflammatory response pathways (mitochondrial activity, interferon), respectively. CONCLUSIONS: Our proteomic study demonstrates that meningiomas of different grades and clinical parameters present distinct proteomic profiles. These proteomic variations offer potential future utility in helping better predict patient outcome and in nominating novel therapeutic targets for personalized care.

Deep learning for image analysis: Personalizing medicine closer to the point of care.

The precision-based revolution in medicine continues to demand stratification of patients into smaller and more personalized subgroups. While genomic technologies have largely led this movement, diagnostic results can take days to weeks to generate. Management at, or closer to, the point of care still heavily relies on the subjective qualitative interpretation of clinical and diagnostic imaging findings. New and emerging technological advances in artificial intelligence (AI) now appear poised to help bring objectivity and precision to these traditionally qualitative analytic tools. In particular, one specific form of AI, known as deep learning, is achieving expert-level disease classifications in many areas of diagnostic medicine dependent on visual and image-based findings. Here, we briefly review concepts of deep learning, and more specifically recent developments in convolutional neural networks (CNNs), to highlight their transformative potential in personalized medicine and, in particular, diagnostic histopathology. Understanding the opportunities and challenges of these quantitative machine-based decision support tools is critical to their widespread introduction into routine diagnostics.

Visualizing histopathologic deep learning classification and anomaly detection using nonlinear feature space dimensionality reduction.

BACKGROUND: There is growing interest in utilizing artificial intelligence, and particularly deep learning, for computer vision in histopathology. While accumulating studies highlight expert-level performance of convolutional neural networks (CNNs) on focused classification tasks, most studies rely on probability distribution scores with empirically defined cutoff values based on post-hoc analysis. More generalizable tools that allow humans to visualize histology-based deep learning inferences and decision making are scarce. RESULTS: Here, we leverage t-distributed Stochastic Neighbor Embedding (t-SNE) to reduce dimensionality and depict how CNNs organize histomorphologic information. Unique to our workflow, we develop a quantitative and transparent approach to visualizing classification decisions prior to softmax compression. By discretizing the relationships between classes on the t-SNE plot, we show we can super-impose randomly sampled regions of test images and use their distribution to render statistically-driven classifications. Therefore, in addition to providing intuitive outputs for human review, this visual approach can carry out automated and objective multi-class classifications similar to more traditional and less-transparent categorical probability distribution scores. Importantly, this novel classification approach is driven by a priori statistically defined cutoffs. It therefore serves as a generalizable classification and anomaly detection tool less reliant on post-hoc tuning. CONCLUSION: Routine incorporation of this convenient approach for quantitative visualization and error reduction in histopathology aims to accelerate early adoption of CNNs into generalized real-world applications where unanticipated and previously untrained classes are often encountered.

Spatiotemporal Proteomic Profiling of Human Cerebral Development.

Mass spectrometry (MS) analysis of human post-mortem central nervous system (CNS) tissue and induced pluripotent stem cell (iPSC)-based directed differentiations offer complementary avenues to define protein signatures of neurodevelopment. Methodological improvements of formalin-fixed, paraffin-embedded (FFPE) protein isolation now enable widespread proteomic analysis of well-annotated archival tissue samples in the context of development and disease. Here, we utilize a shotgun label-free quantification (LFQ) MS method to profile magnetically enriched human cortical neurons and neural progenitor cells (NPCs) derived from iPSCs. We use these signatures to help define spatiotemporal protein dynamics of developing human FFPE cerebral regions. We show that the use of high resolution Q Exactive mass spectrometers now allow simultaneous quantification of >2700 proteins in a single LFQ experiment and provide sufficient coverage to define novel biomarkers and signatures of NPC maintenance and differentiation. Importantly, we show that this abbreviated strategy allows efficient recovery of novel cytoplasmic, membrane-specific and synaptic proteins that are shared between both in vivo and in vitro neuronal differentiation. This study highlights the discovery potential of non-comprehensive high-throughput proteomic profiling of unfractionated clinically well-annotated FFPE human tissue from a diverse array of development and diseased states.

The pluripotency factor Nanog regulates pericentromeric heterochromatin organization in mouse embryonic stem cells.

An open and decondensed chromatin organization is a defining property of pluripotency. Several epigenetic regulators have been implicated in maintaining an open chromatin organization, but how these processes are connected to the pluripotency network is unknown. Here, we identified a new role for the transcription factor NANOG as a key regulator connecting the pluripotency network with constitutive heterochromatin organization in mouse embryonic stem cells. Deletion of Nanog leads to chromatin compaction and the remodeling of heterochromatin domains. Forced expression of NANOG in epiblast stem cells is sufficient to decompact chromatin. NANOG associates with satellite repeats within heterochromatin domains, contributing to an architecture characterized by highly dispersed chromatin fibers, low levels of H3K9me3, and high major satellite transcription, and the strong transactivation domain of NANOG is required for this organization. The heterochromatin-associated protein SALL1 is a direct cofactor for NANOG, and loss of Sall1 recapitulates the Nanog-null phenotype, but the loss of Sall1 can be circumvented through direct recruitment of the NANOG transactivation domain to major satellites. These results establish a direct connection between the pluripotency network and chromatin organization and emphasize that maintaining an open heterochromatin architecture is a highly regulated process in embryonic stem cells.