OUR CHOICE: PAPER OF THE WEEK
July 24, 2018
July 10, 2018
June 26, 2018
In this recent work, Ben Readhead, and his colleagues enlighten the possible role of populational common viruses over the sporadic Alzheimer´s disease (AD) causal factors and developmental characteristics. Postmortem brains from three independent and geographically dispersed cohorts were classified as control, pre-clinical AD, and AD. Then, a multi-omic analysis was conducted using brain regions classically associated with AD pathology, namely Superior Temporal Gyrus, Anterior Prefrontal Cortex, Inferior Frontal Gyrus and Parahippocampal Gyrus. These analyses identified increased levels of HHV-6 and HHV-7 viruses, up to now known mainly for causing herpes infections, in pre-clinical AD and AD brains. Also, they found significant associations between these viruses and genetic, clinical and neuropathological aspects associated to AD pathology. With its conclusion, this paper paves the way for additional studies discussing causality or opportunism from viruses in AD.
June 12, 2018
The term Biomarker refers to a measurable parameter with accurate and reproducible metrics that indicates or associates to a clinical condition. Biomarkers can be used for detecting Alzheimer’s Disease (AD) alterations indexed by functional Magnetic Resonance Imaging (fMRI) in resting state (rs-fMRI) since the Default Mode Network (from temporal parietal, posterior cingulate, and precuneus regions) is sensitive to AD pathology. Cerebral Glucose metabolism can be also evaluated by using 18F-2-deoxy-2-fluoro-D-glucose positrons emission tomography (FDG-PET). Combining these techniques in a hybrid PET/MRI scanner allows for simultaneous information acquisition of brain glucose metabolism and functional activity. Marchitelli & Rocco, et al. (2018) proposed, in a deep technical article, a voxelwise analysis of such relation, targeting healthy elderly people and elderly presenting with Mild Cognitive Impairment/Alzheimer’s Disease (aMCI/AD). The authors concluded that both imaging modalities show abnormalities in the aMCI/AD patients’ posterior part of the brain (posterior cingulate cortex). They also concluded that the rs-fMRI measures were capable of differentiating patients from healthy elderly in fewer DMN cortical regions than the FDG-PET measures. Moreover, about the correlation between the FDG-PET and rs-fMRI imaging, the authors conclude that there is a high correlation (p = 0.8) for healthy elderly, which is reduced by about 17% in aMCI/AD. Therefore this observations may imply in the use of hybrid FDG-PET/fMRI resting state scanners for AD detection.
May 22, 2018
Published by Sophia Schedin-Weiss and her colleagues, this study is very interesting because it shows the potential connection between MAO-B (monoamine oxidase β), γ -secretase and Aβ (amyloid β-peptide) in Alzheimer's disease (AD). Through immunohistochemistry they revealed MAO-B staining in neurons, mainly in the frontal cortex and hippocampus, in postmortem human brain. Surprisingly, the neuronal staining in AD brain is more intense than in control brain. It was concluded that MAO-B levels have increased in the regions cited. These results suggest that MAO-B is related to γ -secretase in the neurons, contributing to increased Aβ formation.
May 08, 2018
This very interesting work published in Nature recently by Wendeln A.C. and collaborators shows the effect of epigenetic changes on microglial response to β-amyloid. The authors repeated intraperitoneal injections of lipopolysaccharides (LPS – lipopolysaccharides present on most of gram-negative bacteria and that have the property activate the innate immune system) in a mouse model of Alzheimer’s pathology. The study suggests that when consecutively exposed to pherypheral LPS, the microglial cells become “tolerant” and promote a HIF1α signaling cascade through epigenetic changes at H3K4 and H3K27. The tolerant microglia exacerbates the β-amyloid deposition while the microglia of animals only exposed to LPS once, called “trained” microglia, aided in plaque clearance. This findings show that microglia is sensitive and creates a memory when in contact with peripheral inflammation, potentially changing its response to pathology development.
April 24, 2018
This interesting review by Trovato Salinaro and collaborators discusses the role of neuroinflammation in the progression of Alzheimer's disease (AD) and other neurodegenerative pathologies. The authors argue that the excessive production of reactive oxygen species (ROS) leads to an inflammatory cascade, which contributes to the pathogenesis of AD. Thus, the reported theory is that low levels of ROS can produce protective responses to these pathogenic processes. In this context, modulation of this system by extracts of nutritional mushrooms such as Coriolus versicolor and Hericium erinaceus are suggested as potential neuroprotective agents, which could potentially reduce or halt this inflammatory cascade.
April 17, 2018
April 10, 2018
In this very interesting paper published by Wijesekara and colleagues, they investigated the role of amyloidogenic proteins in the crosstalk between Alzheimer’s disease (AD) and type 2 diabetes (T2D). Epidemiologic studies show these diseases are a significant risk to each other, and so far, animal models presenting both amyloid-β protein and islet amyloid polypeptide (IAPP, a common factor in T2D) were lacking. This work presents a novel double transgenic model (DTG) that expresses human Aβ and hIAPP. When exposed to a high fat diet (HFD), these animals were considerably more hyperglycemic and glucose intolerant when compared to individual hIAPP and Aβ models, as well as presenting increased cerebral and pancreatic pathology. The study also brings forth evidence that Aβ may be involved in peripheral insulin resistance, and that the combination of IAPP, Aβ, and tau exacerbates the pathological development of both AD and T2D, and contributes greatly to the discussion of the molecular mechanisms that connect these diseases.
April 03, 2018
Zhang et al. tested the effects of chronic administration of cromolyn sodium, an asthma therapeutic drug, and ibuprofen, an antiinflammatory agent, on Aβ-aggregation and microglial activation. They treated 5 months Tg2576 mice, which harbors a human amyloid precursor protein (APP) mutation, with cromolyn or ibuprofen, or in combination for three months. The treatments decreased insoluble Aβ levels and could potentially attenuate amyloid-β deposition, increasing, however, the soluble pools of Aβ. This work also shows cromolyn positive effects on microglial activation favoring amyloid-β phagocytosis. The effects on cognition were not analyzed, emphasizing the need of further investigations. These early findings support a promising use of cromolyn and cromolyn-Ibuprofen as a potential therapeutic strategy targeting amyloid-β in AD.
March 27, 2018
In this recent work, Garranzo-Asensio et al. performed a high-throughput analysis of 706 molecules implicated in cell communication and signalling in the prefrontal cortex of Alzheimer’s disease patients. By comparing AD patients and other types of dementia, the authors found 40 specific proteins altered in AD samples only. This work paves the way for future research, using those proteins as potential biomarkers or therapeutic targets.
March 20, 2018
Recent studies are pointing out that soluble Tau, rather than the aggregated form associated with neurofibrillary tangles, is the main toxic component, being able to induce early synaptic defects before synapse and neuronal loss in several neurodegenerative disorders. Zhou L. and co-workers here show in flies and rat neurons a potential mechanism by which abnormal soluble forms of Tau could be causing damage to the brain. On this incredible work, it was found that mutant forms of Tau interact with synaptic vesicles impairing its release and mobility. Disruptions on this interaction were able to rescue the defects on synaptic function. This work shows promising results, indicating a new potential target for early treatment in Tauopathies like Frontotemporal Dementia and Alzheimer’s Disease.
March 13, 2018
Dubois and collaborators present an alternative scenario for the actual dominant amyloid-dependent view of progressive deterioration of cognitive abilities in Preclinical Alzheimer’s Disease (AD). This new scenario suggests that cortical amyloid-β deposition per se, indexed by PET amyloid imaging, is insufficient to identify patients at high risk of rapid progression to the Prodromal AD phase, the first symptomatic phase in AD continuum. That happens because, despite the underlying brain lesions, compensatory mechanisms on the brain maintain cognitive functions preserved. Based on that, it seems intuitive that a panel of biomarkers would be needed for identifying patients at-risk. The field moves toward the use of biomarkers and computational algorithms, based on artificial intelligence, which are expected to play a crucial role in identifying patients at risk to develop AD.
March 06, 2018
Generally speaking, Machine Learning techniques use some sort of mathematical decision criterium in order to classify patterns, usually described by feature vectors. A set of feature vectors assemble a vector space, generally called feature space by researchers in the field. The dimensionality of such spaces may vary wildly, from dozens to thousands of dimensions, and it is a matter of choice/necessity to work with a "huge" or a "tiny" feature space. Tipically, it is assumed that low-dimensional feature spaces are linear. However, this assumption is less adequate to describe high dimensional spaces, since they might be nonlinear, and in this case, it is said that the data lies on a manifold (i.e. a topoligcal space that is locally Euclidean). When working with medical images (e.g. MRI and PET), scientist often apply machine learning predictive models to patterns that lie on high-dimensional feature space and this task tends to be a hard and computationally costly one. In this report, researchers from The Alzheimer's Disease Neuroimaging Initiative (ADNI) describe a mathematical way of handling this problem. R. Guerrero et al, introduce a group constraint in order to modify the, so called, similarity matrix used in dimensionality reduction algorithms, such as Laplacian Eigenmaps  and Isomap . By reducing the dimensionality of Local Binary Patterns , extracted from learned ROIs of MR images, and applying the kNN classifier, the authors proposal methodology yielded substantive improvements, in comparison to other dimensionality reduction methods, in the estimation of the mini mental state examination scores, among other measures, used for the prediction of MCI to AD subjects.
(1) doi: 10.1162/089976603321780317
 doi: 10.1126/science.290.5500.2319
 doi: 10.1016/0031-3203(95)00067-4
February 27, 2018
In the last decades, the computational neuroscientists' community experienced a deluge of neuronal electrophysiological measures and reconstruction data. Such richness allowed the theoretical principles of neural function and dynamics to be expanded way beyond the nobelized Hudgey-Huxley's 60's experiments, on a squid giant axon, through open science initiatives. However, despite simulation neuroscience field being currently in a ramp-up phase by means of great projects such as the Blue Brain Project and White House BRAIN initiative, we are still missing important parts of the brain puzzle. The Astrocyte is a glial cell that has its role in neuronal firing overly underestimated. Recently, Herculano-Houzel (1), demonstrates that in the human cortex glial cells outnumber neurons by far (three glial for each neuron). Given that, the strict neurocentric point of view allowed only a partial understanding of brain function and behavior. Such preference can be seen in a popular database called modelDB (https://senselab.med.yale), which curates a total of 1308 computational models (at 26/02/2018), with only 12 acknowledging astrocytes. But why are you telling me that? Astrocytes play an important role in many neurodegenerative and brain disorders (2), which we are trying to understand (luckily by means of computer simulation) and potentially cure, also they contribute to plasticity and protection against excessive excitation. Tiina Manninen and colleagues bring attention to the main issues of reproducibility in computational neuroscience, highlighting good practices to increase reuse and comparability for astrocytes modeling.
(1) doi: 10.1002/cne.24040
(2) doi: 10.1101/cshperspect.a020628
This incredible work presents the largest known cohort to date of individuals with autosomal dominant Alzheimer’s disease studied longitudinally with multiple neuroimaging biomarkers. In this paper, Gordon, Blazey and co-authors followed for 6 years individuals from families with autosomal dominant Alzheimer’s disease. The study found that mutation carriers first display amyloid β plaques accumulation accessed by ¹¹C-PiB, followed by hypometabolism accessed by [¹⁸F-FDG] PET, and finally structural atrophy accessed by MRI. Temporally the study suggests a pattern on which β-amyloidosis, detectable hypometabolism, and structural atrophy emerge more than 20, 15, and 10 years, respectively, before the expected onset of clinical dementia. On the other side, the spatial progression and distribution of imaging biomarkers abnormalities was particularly heterogeneous among the individuals. A special feature found was the presence of no metabolic decline on regions exhibiting advanced β-amyloidosis and structural atrophy. This work reveals complex patterns of biomarker alteration across the brain, contributing with the understanding of the pathophysiological progression of Alzheimer’s disease.
February 13, 2018
Bate & Williams demonstrated that amyloid-beta (Aβ) may be neurotoxic or neuroprotective, depending on its conformational state. In vitro data support that Aβ monomers are neuroprotective against Aβ oligomer-induced synapse damage. The authors suggest that synaptotoxicity is a result from a pathological Aβ monomer:oligomers unbalance rather than total Aβ concentration in brain.
February 06, 2018
In this very interesting report, Kobayashi et al. demonstrate that people presenting with AD pathology but cognitively normal seem to have high levels of astrocytic GLT-1. This article highlights the potential neuroproctetive role of an asctroyctic transporter.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
In this work published in Scientific Reports, Kim and colleagues created the Social Event Memory Test (SEMT), a task in which participants were asked to recall events after watching a 360º video-clip that simulates real-life social interactions. Head-mounted displays were used to increase the patients' sense of reality during the examination.
Participants were segregated into three groups: Subjective Cognitive Impairment (SCI), Mild Cognitive Impairment (aMCI) and Early Stage Alzheimer's Disease. All underwent detailed neuropsychological testing, blood sampling, MRI and PET imaging. SEMT showed its capability to predict amyloid positivity in individuals of the same group and correlated significantly positively with Seoul Verbal Learning Test (SVLT) scores and with hippocampal volume. These findings suggest the possibility of using Virtual Reality-based neuropsychological tests as screening tests, since there are many advantages in comparison with conventional cognitive tasks, such as reduced time, high sensitivity and financial cost.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Artificial dense neural networks are computational simplifications inspired in biological neural networks. Typically, these networks consist of many layers of nodes and each node from each layer is connected to all adjacent layer nodes (these are known as "fully connected neural networks"), which means that for every layer in a network, all connecting weights have to be computed, which tends to increase the model's computational complexity. Decebal C. M. and colleagues, introduce a method that, as in biological neurons, assumes sparse connected neural networks (contrarily to fully connected) and, therefore is lighter and faster than the current approaches. The algorithm uses graph theory (Erdős–Rényi random graph topology) and it is capable of generating sparse models for many neural networks architectures, such as restricted Boltzmann machines (RBM), multi-layer perceptrons (MLP), and convolutional neural networks (CNN). Their model achieved better accuracy than the classic architecture in three different datasets using MLPs and also had better results on CNNs and RBMs using down to 1% of the amount of weights. This is an innovative method that has the potential to enable the processing of, until then, impossible or inviable algorithms with high potential for analyzing big medical databases.