Control of cell surface expression of GABAA receptors by a conserved region at the end of the N-terminal extracellular domain of receptor subunits.

Type A γ-aminobutyric acid receptors (GABAARs) represent a family of pentameric GABA-gated Cl-/HCO3- ion channels which mediate inhibitory transmission in the central nervous system. Cell surface expression of GABAARs, a prerequisite for their function, is dependent on the appropriate assembly of the receptor subunits and their transient interactions with molecular chaperones within the endoplasmic reticulum (ER) and Golgi apparatus. Here, we describe a highly conserved amino acid sequence within the extracellular N-terminal domain of the receptor subunits adjoining the first transmembrane domain as a region important for GABAAR processing within the ER. Modifications of this region in the α1, ß3, and γ2 subunits using insertion or site-directed mutagenesis impaired GABAAR trafficking to the cell surface in heterologous cell systems although they had no effect on the subunit assembly. We found that mutated receptors accumulated in the ER where they were shown to associate with chaperones calnexin, BiP, and Grp94. However, their surface expression was increased when ER-associated degradation or proteosome function was inhibited, while modulation of ER calcium stores had little effect. When compared to the wt, mutated receptors showed decreased interaction with calnexin, similar binding to BiP, and increased association with Grp94. Structural modeling of calnexin interaction with the wt or mutated GABAAR revealed that disruption in structure caused by mutations in the conserved region adjoining the first transmembrane domain may impair calnexin binding. Thus, this previously uncharacterized region plays an important role in intracellular processing of GABAARs at least in part by stabilizing their interaction with calnexin.

Benign overfitting in linear regression.

The phenomenon of benign overfitting is one of the key mysteries uncovered by deep learning methodology: deep neural networks seem to predict well, even with a perfect fit to noisy training data. Motivated by this phenomenon, we consider when a perfect fit to training data in linear regression is compatible with accurate prediction. We give a characterization of linear regression problems for which the minimum norm interpolating prediction rule has near-optimal prediction accuracy. The characterization is in terms of two notions of the effective rank of the data covariance. It shows that overparameterization is essential for benign overfitting in this setting: the number of directions in parameter space that are unimportant for prediction must significantly exceed the sample size. By studying examples of data covariance properties that this characterization shows are required for benign overfitting, we find an important role for finite-dimensional data: the accuracy of the minimum norm interpolating prediction rule approaches the best possible accuracy for a much narrower range of properties of the data distribution when the data lie in an infinite-dimensional space vs. when the data lie in a finite-dimensional space with dimension that grows faster than the sample size.

The Perils of Being Unhinged: On the Accuracy of Classifiers Minimizing a Noise-Robust Convex Loss.

van Rooyen, Menon, and Williamson (2015) introduced a notion of convex loss functions being robust to random classification noise and established that the "unhinged" loss function is robust in this sense. In this letter, we study the accuracy of binary classifiers obtained by minimizing the unhinged loss and observe that even for simple linearly separable data distributions, minimizing the unhinged loss may only yield a binary classifier with accuracy no better than random guessing.

On the Effect of the Activation Function on the Distribution of Hidden Nodes in a Deep Network.

We analyze the joint probability distribution on the lengths of the vectors of hidden variables in different layers of a fully connected deep network, when the weights and biases are chosen randomly according to gaussian distributions. We show that if the activation function φ satisfies a minimal set of assumptions, satisfied by all activation functions that we know that are used in practice, then, as the width of the network gets large, the "length process" converges in probability to a length map that is determined as a simple function of the variances of the random weights and biases and the activation function φ. We also show that this convergence may fail for φ that violate our assumptions. We show how to use this analysis to choose the variance of weight initialization, depending on the activation function, so that hidden variables maintain a consistent scale throughout the network.

Gradient Descent with Identity Initialization Efficiently Learns Positive-Definite Linear Transformations by Deep Residual Networks.

We analyze algorithms for approximating a function f(x)=Φx mapping ℜd to ℜd using deep linear neural networks, that is, that learn a function h parameterized by matrices Θ1,…,ΘL and defined by h(x)=ΘLΘL-1…Θ1x . We focus on algorithms that learn through gradient descent on the population quadratic loss in the case that the distribution over the inputs is isotropic. We provide polynomial bounds on the number of iterations for gradient descent to approximate the least-squares matrix Φ , in the case where the initial hypothesis Θ1=…=ΘL=I has excess loss bounded by a small enough constant. We also show that gradient descent fails to converge for Φ whose distance from the identity is a larger constant, and we show that some forms of regularization toward the identity in each layer do not help. If Φ is symmetric positive definite, we show that an algorithm that initializes Θi=I learns an ε -approximation of f using a number of updates polynomial in L , the condition number of Φ , and log(d/ε) . In contrast, we show that if the least-squares matrix Φ is symmetric and has a negative eigenvalue, then all members of a class of algorithms that perform gradient descent with identity initialization, and optionally regularize toward the identity in each layer, fail to converge. We analyze an algorithm for the case that Φ satisfies u⊤Φu>0 for all u but may not be symmetric. This algorithm uses two regularizers: one that maintains the invariant u⊤ΘLΘL-1…Θ1u>0 for all u and the other that "balances" Θ1,…,ΘL so that they have the same singular values.

Field Application of 238U/235U Measurements To Detect Reoxidation and Mobilization of U(IV).

Biostimulation to induce reduction of soluble U(VI) to relatively immobile U(IV) is an effective strategy for decreasing aqueous U(VI) concentrations in contaminated groundwater systems. If oxidation of U(IV) occurs following the biostimulation phase, U(VI) concentrations increase, challenging the long-term effectiveness of this technique. However, detecting U(IV) oxidation through dissolved U concentrations alone can prove difficult in locations with few groundwater wells to track the addition of U to a mass of groundwater. We propose the 238U/235U ratio of aqueous U as an independent, reliable tracer of U(IV) remobilization via oxidation or mobilization of colloids. Reduction of U(VI) produces 238U-enriched U(IV), whereas remobilization of solid U(IV) should not induce isotopic fractionation. The incorporation of remobilized U(IV) with a high 238U/235U ratio into the aqueous U(VI) pool produces an increase in 238U/235U of aqueous U(VI). During several injections of nitrate to induce U(IV) oxidation, 238U/235U consistently increased, suggesting 238U/235U is broadly applicable for detecting mobilization of U(IV).

Physico-Chemical Heterogeneity of Organic-Rich Sediments in the Rifle Aquifer, CO: Impact on Uranium Biogeochemistry.

The Rifle alluvial aquifer along the Colorado River in west central Colorado contains fine-grained, diffusion-limited sediment lenses that are substantially enriched in organic carbon and sulfides, as well as uranium, from previous milling operations. These naturally reduced zones (NRZs) coincide spatially with a persistent uranium groundwater plume. There is concern that uranium release from NRZs is contributing to plume persistence or will do so in the future. To better define the physical extent, heterogeneity and biogeochemistry of these NRZs, we investigated sediment cores from five neighboring wells. The main NRZ body exhibited uranium concentrations up to 100 mg/kg U as U(IV) and contains ca. 286 g of U in total. Uranium accumulated only in areas where organic carbon and reduced sulfur (as iron sulfides) were present, emphasizing the importance of sulfate-reducing conditions to uranium retention and the essential role of organic matter. NRZs further exhibited centimeter-scale variations in both redox status and particle size. Mackinawite, greigite, pyrite and sulfate coexist in the sediments, indicating that dynamic redox cycling occurs within NRZs and that their internal portions can be seasonally oxidized. We show that oxidative U(VI) release to the aquifer has the potential to sustain a groundwater contaminant plume for centuries. NRZs, known to exist in other uranium-contaminated aquifers, may be regionally important to uranium persistence.

Nonproliferative and Proliferative Lesions of the Rat and Mouse Skeletal Tissues (Bones, Joints, and Teeth).

The INHAND (International Harmonization of Nomenclature and Diagnostic Criteria for Lesions in Rats and Mice) Project (www.toxpath.org/inhand.asp) is an initiative of the Societies of Toxicological Pathology from Europe (ESTP), Great Britain (BSTP), Japan (JSTP) and North America (STP) to develop an internationally accepted nomenclature for proliferative and nonproliferative lesions in laboratory animals. The purpose of this publication is to provide a standardized nomenclature for classifying microscopic lesions observed in the skeletal tissues and teeth of laboratory rats and mice, with color photomicrographs illustrating examples of many common lesions. The standardized nomenclature presented in this document is also available on the internet (http://www.goreni.org/). Sources of material were databases from government, academic and industrial laboratories throughout the world.

Disturbed subsurface microbial communities follow equivalent trajectories despite different structural starting points.

Microbial community structure, and niche and neutral processes can all influence response to disturbance. Here, we provide experimental evidence for niche versus neutral and founding community effects during a bioremediation-related organic carbon disturbance. Subsurface sediment, partitioned into 22 flow-through columns, was stimulated in situ by the addition of acetate as a carbon and electron donor source. This drove the system into a new transient biogeochemical state characterized by iron reduction and enriched Desulfuromonadales, Comamonadaceae and Bacteroidetes lineages. After approximately 1 month conditions favoured sulfate reduction, and were accompanied by a substantial increase in the relative abundance of Desulfobulbus, Desulfosporosinus, Desulfitobacterium and Desulfotomaculum. Two subsets of four to five columns each were switched from acetate to lactate amendment during either iron (earlier) or sulfate (later) reduction. Hence, subsets had significantly different founding communities. All lactate treatments exhibited lower relative abundances of Desulfotomaculum and Bacteroidetes, enrichments of Clostridiales and Psychrosinus species, and a temporal succession from highly abundant Clostridium sensu stricto to Psychrosinus. Regardless of starting point, lactate-switch communities followed comparable structural trajectories, whereby convergence was evident 9 to 16 days after each switch, and significant after 29 to 34 days of lactate addition. Results imply that neither the founding community nor neutral processes influenced succession following perturbation.

Long-term in situ oxidation of biogenic uraninite in an alluvial aquifer: impact of dissolved oxygen and calcium.

Oxidative dissolution controls uranium release to (sub)oxic pore waters from biogenic uraninite produced by natural or engineered processes, such as bioremediation. Laboratory studies show that uraninite dissolution is profoundly influenced by dissolved oxygen (DO), carbonate, and solutes such as Ca(2+). In complex and heterogeneous subsurface environments, the concentrations of these solutes vary in time and space. Knowledge of dissolution processes and kinetics occurring over the long-term under such conditions is needed to predict subsurface uranium behavior and optimize the selection and performance of uraninite-based remediation technologies over multiyear periods. We have assessed dissolution of biogenic uraninite deployed in wells at the Rifle, CO, DOE research site over a 22 month period. Uraninite loss rates were highly sensitive to DO, with near-complete loss at >0.6 mg/L over this period but no measurable loss at lower DO. We conclude that uraninite can be stable over decadal time scales in aquifers under low DO conditions. U(VI) solid products were absent over a wide range of DO values, suggesting that dissolution proceeded through complexation and removal of oxidized surface uranium atoms by carbonate. Moreover, under the groundwater conditions present, Ca(2+) binds strongly to uraninite surfaces at structural uranium sites, impacting uranium fate.

Uranium bioreduction rates across scales: biogeochemical hot moments and hot spots during a biostimulation experiment at Rifle, Colorado.

We aim to understand the scale-dependent evolution of uranium bioreduction during a field experiment at a former uranium mill site near Rifle, Colorado. Acetate was injected to stimulate Fe-reducing bacteria (FeRB) and to immobilize aqueous U(VI) to insoluble U(IV). Bicarbonate was coinjected in half of the domain to mobilize sorbed U(VI). We used reactive transport modeling to integrate hydraulic and geochemical data and to quantify rates at the grid block (0.25 m) and experimental field scale (tens of meters). Although local rates varied by orders of magnitude in conjunction with biostimulation fronts propagating downstream, field-scale rates were dominated by those orders of magnitude higher rates at a few selected hot spots where Fe(III), U(VI), and FeRB were at their maxima in the vicinity of the injection wells. At particular locations, the hot moments with maximum rates negatively corresponded to their distance from the injection wells. Although bicarbonate injection enhanced local rates near the injection wells by a maximum of 39.4%, its effect at the field scale was limited to a maximum of 10.0%. We propose a rate-versus-measurement-length relationship (log R' = -0.63 log L - 2.20, with R' in µmol/mg cell protein/day and L in meters) for orders-of-magnitude estimation of uranium bioreduction rates across scales.

Speciation and reactivity of uranium products formed during in situ bioremediation in a shallow alluvial aquifer.

In this study, we report the results of in situ U(VI) bioreduction experiments at the Integrated Field Research Challenge site in Rifle, Colorado, USA. Columns filled with sediments were deployed into a groundwater well at the site and, after a period of conditioning with groundwater, were amended with a mixture of groundwater, soluble U(VI), and acetate to stimulate the growth of indigenous microorganisms. Individual reactors were collected as various redox regimes in the column sediments were achieved: (i) during iron reduction, (ii) just after the onset of sulfate reduction, and (iii) later into sulfate reduction. The speciation of U retained in the sediments was studied using X-ray absorption spectroscopy, electron microscopy, and chemical extractions. Circa 90% of the total uranium was reduced to U(IV) in each reactor. Noncrystalline U(IV) comprised about two-thirds of the U(IV) pool, across large changes in microbial community structure, redox regime, total uranium accumulation, and reaction time. A significant body of recent research has demonstrated that noncrystalline U(IV) species are more suceptible to remobilization and reoxidation than crystalline U(IV) phases such as uraninite. Our results highlight the importance of considering noncrystalline U(IV) formation across a wide range of aquifer parameters when designing in situ remediation plans.

Expression of integrin and CD44 receptors recognising osteopontin in the normal and LPS-lesioned rat substantia nigra.

The multifunctional protein osteopontin (OPN) is expressed in the substantia nigra (SN) and protects nigral dopaminergic neurones against toxic insult in animal models of Parkinson's disease, although the mechanisms involved are uncertain. In the periphery, OPN regulates inflammatory processes by interacting with integrin and CD44 receptors but the presence and distribution of these sites in SN is unknown. We investigated the expression of integrin receptor subunits and CD44 receptors in the normal SN and after induction of inflammation by lipopolysaccharide (LPS), and their interaction with OPN. In normal rat SN, integrin αv , ß3 and ß1 , and CD44, receptors were expressed on neurones including TH-positive cells but not on glia. LPS administration induced a loss of TH-positive neurones in SN and increased expression of glial cells as shown by GFAP, OX-6 and ED-1 immunoreactivity. In LPS-lesioned SN, there was up-regulation of the expression of integrin ß3 and CD44 receptors. Co-localisation studies showed that this related to their increased expression on OX-6-, ED-1- and GFAP-positive cells. Furthermore, OPN interacted with integrin and CD44 receptors in the normal rat SN as demonstrated by co-immunoprecipitation and pull-down techniques. These data show that integrin and CD44 receptors are present on neurones in normal rat SN and that they are up-regulated on glial cells following LPS-mediated inflammation in SN, suggesting that they are functionally important in the inflammatory process. The interaction of OPN with these receptors suggests a role in the neuroprotective effect of this protein in the LPS model of Parkinson's disease.

Profiling in situ microbial community structure with an amplification microarray.

The objectives of this study were to unify amplification, labeling, and microarray hybridization chemistries within a single, closed microfluidic chamber (an amplification microarray) and verify technology performance on a series of groundwater samples from an in situ field experiment designed to compare U(VI) mobility under conditions of various alkalinities (as HCO(3)(-)) during stimulated microbial activity accompanying acetate amendment. Analytical limits of detection were between 2 and 200 cell equivalents of purified DNA. Amplification microarray signatures were well correlated with 16S rRNA-targeted quantitative PCR results and hybridization microarray signatures. The succession of the microbial community was evident with and consistent between the two microarray platforms. Amplification microarray analysis of acetate-treated groundwater showed elevated levels of iron-reducing bacteria (Flexibacter, Geobacter, Rhodoferax, and Shewanella) relative to the average background profile, as expected. Identical molecular signatures were evident in the transect treated with acetate plus NaHCO(3), but at much lower signal intensities and with a much more rapid decline (to nondetection). Azoarcus, Thaurea, and Methylobacterium were responsive in the acetate-only transect but not in the presence of bicarbonate. Observed differences in microbial community composition or response to bicarbonate amendment likely had an effect on measured rates of U reduction, with higher rates probable in the part of the field experiment that was amended with bicarbonate. The simplification in microarray-based work flow is a significant technological advance toward entirely closed-amplicon microarray-based tests and is generally extensible to any number of environmental monitoring applications.

Molecular analysis of the in situ growth rates of subsurface Geobacter species.

Molecular tools that can provide an estimate of the in situ growth rate of Geobacter species could improve understanding of dissimilatory metal reduction in a diversity of environments. Whole-genome microarray analyses of a subsurface isolate of Geobacter uraniireducens, grown under a variety of conditions, identified a number of genes that are differentially expressed at different specific growth rates. Expression of two genes encoding ribosomal proteins, rpsC and rplL, was further evaluated with quantitative reverse transcription-PCR (qRT-PCR) in cells with doubling times ranging from 6.56 h to 89.28 h. Transcript abundance of rpsC correlated best (r(2) = 0.90) with specific growth rates. Therefore, expression patterns of rpsC were used to estimate specific growth rates of Geobacter species during an in situ uranium bioremediation field experiment in which acetate was added to the groundwater to promote dissimilatory metal reduction. Initially, increased availability of acetate in the groundwater resulted in higher expression of Geobacter rpsC, and the increase in the number of Geobacter cells estimated with fluorescent in situ hybridization compared well with specific growth rates estimated from levels of in situ rpsC expression. However, in later phases, cell number increases were substantially lower than predicted from rpsC transcript abundance. This change coincided with a bloom of protozoa and increased attachment of Geobacter species to solid phases. These results suggest that monitoring rpsC expression may better reflect the actual rate that Geobacter species are metabolizing and growing during in situ uranium bioremediation than changes in cell abundance.

No measurable changes in (238)U/(235)U due to desorption-adsorption of U(VI) from groundwater at the Rifle, Colorado, integrated field research challenge site.

Groundwater samples were collected from the Integrated Field Research Challenge field site in Rifle, Colorado, over the course of a bicarbonate-induced U desorption-adsorption experiment. Uranium concentrations and high precision U isotopic compositions ((238)U/(235)U) of these groundwater samples were determined and used to assess the impact of bicarbonate-induced U(VI) desorption from contaminated sediments on the (238)U/(235)U of groundwater. The (238)U/(235)U of groundwater was not significantly impacted by bicarbonate-induced desorption of U(VI) from mineral surfaces or by adsorption of advecting U(VI) from upgradient locations onto those surfaces after the treatment. Assuming this absence of a significant shift in U isotopic composition associated with desorption-adsorption applies to other systems, reduction of U(VI) to U(IV) is expected to be the dominant source of U isotopic fractionation associated with removal of U(VI) from pore water as a result of natural and stimulated reductive pathways. Thus, changes in the (238)U/(235)U composition of uranium-bearing fluids should be useful in quantifying the extent of reduction.

A yeast two-hybrid screen reveals that osteopontin associates with MAP1A and MAP1B in addition to other proteins linked to microtubule stability, apoptosis and protein degradation in the human brain.

Osteopontin (OPN) expression is reduced in surviving dopaminergic neurones in the substantia nigra (SN) in Parkinson's disease (PD), and protects against MPP(+) -induced cell death in primary mesencephalic cultures and 6-OHDA-induced cell loss in the rat, while inactivation of OPN aggravates cell death. OPN is thought to act through interactions with integrin receptors or CD44. However, the specific protein interactions involved in OPN-mediated neuroprotection are unknown and are the focus of this study. The yeast two-hybrid (YTH) technique was utilised to investigate OPN-protein interactions, using full-length human OPN to screen a human foetal brain cDNA library. Proteins involved in apoptosis, protein degradation and microtubule stability were identified as OPN binding partners. These included: MAP1A and MAP1B, which regulate microtubule stability; RNF138, an E3 ubiquitin-ligase; proteasome ß1 subunit, a subunit of the 20S proteasome involved in the ubiquitin-dependent cleavage of peptides; BAG6, SGTΑ and EF1A, proteins implicated in control of apoptosis; DnaJB1, a co-chaperone of Hsp70s; and pleiotrophin, a growth factor. The use of site-directed mutagenesis to modify known OPN protein binding sites outside the RGD integrin binding domain, specifically Y165A and D139E, inhibited some of these interactions. Further investigation using affinity pull-down assays, co-immunoprecipitation and immunohistochemistry confirmed that OPN associates with MAP1A and MAP1B in rat SN and striatum. These findings indicate a role for OPN in the regulation of microtubule dynamics, apoptosis and proteolysis in the brain, suggesting that OPN may act as an endogenous multifunctional protective protein in PD.

Evaluation of a genome-scale in silico metabolic model for Geobacter metallireducens by using proteomic data from a field biostimulation experiment.

Accurately predicting the interactions between microbial metabolism and the physical subsurface environment is necessary to enhance subsurface energy development, soil and groundwater cleanup, and carbon management. This study was an initial attempt to confirm the metabolic functional roles within an in silico model using environmental proteomic data collected during field experiments. Shotgun global proteomics data collected during a subsurface biostimulation experiment were used to validate a genome-scale metabolic model of Geobacter metallireducens-specifically, the ability of the metabolic model to predict metal reduction, biomass yield, and growth rate under dynamic field conditions. The constraint-based in silico model of G. metallireducens relates an annotated genome sequence to the physiological functions with 697 reactions controlled by 747 enzyme-coding genes. Proteomic analysis showed that 180 of the 637 G. metallireducens proteins detected during the 2008 experiment were associated with specific metabolic reactions in the in silico model. When the field-calibrated Fe(III) terminal electron acceptor process reaction in a reactive transport model for the field experiments was replaced with the genome-scale model, the model predicted that the largest metabolic fluxes through the in silico model reactions generally correspond to the highest abundances of proteins that catalyze those reactions. Central metabolism predicted by the model agrees well with protein abundance profiles inferred from proteomic analysis. Model discrepancies with the proteomic data, such as the relatively low abundances of proteins associated with amino acid transport and metabolism, revealed pathways or flux constraints in the in silico model that could be updated to more accurately predict metabolic processes that occur in the subsurface environment.

Microbial functional gene diversity with a shift of subsurface redox conditions during In Situ uranium reduction.

To better understand the microbial functional diversity changes with subsurface redox conditions during in situ uranium bioremediation, key functional genes were studied with GeoChip, a comprehensive functional gene microarray, in field experiments at a uranium mill tailings remedial action (UMTRA) site (Rifle, CO). The results indicated that functional microbial communities altered with a shift in the dominant metabolic process, as documented by hierarchical cluster and ordination analyses of all detected functional genes. The abundance of dsrAB genes (dissimilatory sulfite reductase genes) and methane generation-related mcr genes (methyl coenzyme M reductase coding genes) increased when redox conditions shifted from Fe-reducing to sulfate-reducing conditions. The cytochrome genes detected were primarily from Geobacter sp. and decreased with lower subsurface redox conditions. Statistical analysis of environmental parameters and functional genes indicated that acetate, U(VI), and redox potential (E(h)) were the most significant geochemical variables linked to microbial functional gene structures, and changes in microbial functional diversity were strongly related to the dominant terminal electron-accepting process following acetate addition. The study indicates that the microbial functional genes clearly reflect the in situ redox conditions and the dominant microbial processes, which in turn influence uranium bioreduction. Microbial functional genes thus could be very useful for tracking microbial community structure and dynamics during bioremediation.

Optimal gene expression analysis by microarrays.

DNA microarrays make possible the rapid and comprehensive assessment of the transcriptional activity of a cell, and as such have proven valuable in assessing the molecular contributors to biological processes and in the classification of human cancers. The major challenge in using this technology is the analysis of its massive data output, which requires computational means for interpretation and a heightened need for quality data. The optimal analysis requires an accounting and control of the many sources of variance within the system, an understanding of the limitations of the statistical approaches, and the ability to make sense of the results through intelligent database interrogation.