Disease-driven domain generalization for neuroimaging-based assessment of Alzheimer's disease.

Development of deep learning models to evaluate structural brain changes caused by cognitive impairment in MRI scans holds significant translational value. The efficacy of these models often encounters challenges due to variabilities arising from different data generation protocols, imaging equipment, radiological artifacts, and shifts in demographic distributions. Domain generalization (DG) techniques show promise in addressing these challenges by enabling the model to learn from one or more source domains and apply this knowledge to new, unseen target domains. Here we present a framework that utilizes model interpretability to enhance the generalizability of classification models across various cohorts. We used MRI scans and clinical diagnoses from four independent cohorts: Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 1821), the Framingham Heart Study (FHS, n = 304), the Australian Imaging Biomarkers & Lifestyle Study of Ageing (AIBL, n = 661), and the National Alzheimer's Coordinating Center (NACC, n = 4647). With this data, we trained a deep neural network to focus on areas of the brain identified as relevant to the disease for model training. Our approach involved training a classifier to differentiate between structural neurodegeneration in individuals with normal cognition (NC), mild cognitive impairment (MCI), and dementia due to Alzheimer's disease (AD). This was achieved by aligning class-wise attention with a unified visual saliency prior, which was computed offline for each class using all the training data. Our method not only competes with state-of-the-art approaches but also shows improved correlation with postmortem histology. This alignment with the gold standard evidence is a significant step towards validating the effectiveness of DG frameworks, paving the way for their broader application in the field.

Disease-driven domain generalization for neuroimaging-based assessment of Alzheimer's disease.

Development of deep learning models to assess the degree of cognitive impairment on magnetic resonance imaging (MRI) scans has high translational significance. Performance of such models is often affected by potential variabilities stemming from independent protocols for data generation, imaging equipment, radiology artifacts, and demographic distributional shifts. Domain generalization (DG) frameworks have the potential to overcome these issues by learning signal from one or more source domains that can be transferable to unseen target domains. We developed an approach that leverages model interpretability as a means to improve generalizability of classification models across multiple cohorts. Using MRI scans and clinical diagnosis obtained from four independent cohorts (Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 1,821), the Framingham Heart Study (FHS, n = 304), the Australian Imaging Biomarkers and Lifestyle Study of Ageing (AIBL, n = 661), and the National Alzheimer's Coordinating Center (NACC, n = 4,647)), we trained a deep neural network that used model-identified regions of disease relevance to inform model training. We trained a classifier to distinguish persons with normal cognition (NC) from those with mild cognitive impairment (MCI) and Alzheimer's disease (AD) by aligning class-wise attention with a unified visual saliency prior computed offline per class over all training data. Our proposed method competes with state-of-the-art methods with improved correlation with postmortem histology, thus grounding our findings with gold standard evidence and paving a way towards validating DG frameworks.

Neural Differentiation of Induced Pluripotent Stem Cells for a Xenogeneic Material-Free 3D Neurological Disease Model Neurulation from Pluripotent Cells Using a Human Hydrogel.

Alzheimer's Disease (AD) is characterized by synapse and neuronal loss and the accumulation of neurofibrillary tangles and Amyloid ß plaques. Despite significant research efforts to understand the late stages of the disease, its etiology remains largely unknown. This is in part because of the imprecise AD models in current use. In addition, little attention has been paid to neural stem cells (NSC), which are the cells responsible for the development and maintenance of brain tissue during an individual's lifespan. Thus, an in vitro 3D human brain tissue model using induced pluripotent stem (iPS) cell-derived neural cells in human physiological conditions may be an excellent alternative to standard models to investigate AD pathology. Following the differentiation process mimicking development, iPS cells can be turned into NSCs and, ultimately, neural cells. During differentiation, the traditionally used xenogeneic products may alter the cells' physiology and prevent accurate disease pathology modeling. Hence, establishing a xenogeneic material-free cell culture and differentiation protocol is essential. This study investigated the differentiation of iPS cells to neural cells using a novel extracellular matrix derived from human platelet lysates (PL Matrix). We compared the stemness properties and differentiation efficacies of iPS cells in a PL matrix against those in a conventional 3D scaffold made of an oncogenic murine-matrix. Using well-defined conditions without xenogeneic material, we successfully expanded and differentiated iPS cells into NSCs via dual-SMAD inhibition, which regulates the BMP and TGF signaling cascades in a manner closer to human conditions. This in vitro, 3D, xenogeneic-free scaffold will enhance the quality of disease modeling for neurodegenerative disease research, and the knowledge produced could be used in developing more effective translational medicine.

Methods and Protocols for Using Extracellular Vesicles as Delivery Vehicles in Neuronal Research.

Extracellular vesicles (EVs) transport nucleic acids, proteins, and lipid molecules for intercellular communication. The biomolecular cargo from EVs can modify the recipient cell genetically, physiologically, and pathologically. This innate ability of EVs can be harnessed to deliver the cargo of interest to a specific organ or a cell type. Importantly, due to their ability to cross the blood-brain barrier (BBB), the EVs can be used as delivery vehicles to transport therapeutic drugs and other macromolecules to inaccessible organs such as the brain. Therefore, the current chapter includes laboratory techniques and protocols focusing on the customization of EVs for neuronal research.

Coculture with Neural Stem Cells May Shift the Transcription Profile of Glioblastoma Multiforme towards Cancer-Specific Stemness.

Glioblastoma multiforme (GBM) possesses a small but significant population of cancer stem cells (CSCs) thought to play a role in its invasiveness, recurrence, and metastasis. The CSCs display transcriptional profiles for multipotency, self-renewal, tumorigenesis, and therapy resistance. There are two possible theories regarding the origin of CSCs in the context of neural stem cells (NSCs); i.e., NSCs modify cancer cells by conferring them with cancer-specific stemness, or NSCs themselves are transformed into CSCs due to the tumor environment created by cancer cells. To test the theories and to investigate the transcriptional regulation of the genes involved in CSC formation, we cocultured NSC and GBM cell lines together. Where genes related to cancer stemness, drug efflux, and DNA modification were upregulated in GBM, they were downregulated in NSCs upon coculture. These results indicate that cancer cells shift the transcriptional profile towards stemness and drug resistance in the presence of NSCs. Concurrently, GBM triggers NSCs differentiation. Because the cell lines were separated by a membrane (0.4 µm pore size) to prevent direct contact between GBM and NSCs, cell-secreted signaling molecules and extracellular vesicles (EVs) are likely involved in reciprocal communication between NSCs and GBM, causing transcription modification. Understanding the mechanism of CSC creation will aid in the identification of precise molecular targets within the CSCs to exterminate them, which, in turn, will increase the efficacy of chemo-radiation treatment.

Exposure to a Pathological Condition May Be Required for the Cells to Secrete Exosomes Containing mtDNA Aberration.

Exosomes, nanovesicles secreted by all cells, carry out intercellular communication by transmitting biologically active cargo comprising DNA, RNA, and proteins. These biomolecules reflect the status of their parent cells and can be altered by pathological conditions. Therefore, the researchers have been investigating differential sequences and quantities of DNA associated with exosomes as valuable biomarkers of diseases. Exosomes carry different types of DNA molecules, including genomic, cytoplasmic, and mitochondrial (mtDNA). The mtDNA aberrations are reported to be a hallmark of diseases involving oxidative stress, such as cancer and neurodegenerative diseases. Establishing robust in vitro models comprising appropriate cell lineages is the first step towards investigating disease-specific anomalies and testing therapeutics. Induced pluripotent stem (iPS) cells from patients with diseases have been used for this purpose since they can differentiate into various cells. The current study investigated mtDNA aberrations in exosomes secreted by primary cancer cells and neural stem cells (NSCs) differentiated from iPS cells. The primary cancer cells were isolated from surgically removed glioblastoma multiforme (GBM) tissue, and the iPS cells were produced from control and Alzheimer's disease (AD) subjects' B lymphocytes. We detected aberrations in mtDNA associated with exosomes secreted from GBM cells but not from the NSCs. This result indicates that the cells may not secrete exosomes carrying mtDNA aberration without exposure to a pathological condition. Thus, we may need to consider this fact when we use iPS cell-derived cells as an in vitro disease model.