Speech Enhancement from Additive Noise and Channel Distortion - a Corpus-Based Approach

This paper presents a new approach to single-channel speech enhancement involving both noise and channel distortion (i.e., convolutional noise). The approach is based on finding longest matching segments (LMS) from a corpus of clean, wideband speech. The approach adds three novel developments to our previous LMS research. First, we address the problem of channel distortion as well as additive noise. Second, we present an improved method for modeling noise. Third, we present an iterative algorithm for improved speech estimates. In experiments using speech recognition as a test with the Aurora 4 database, the use of our enhancement approach as a preprocessor for feature extraction significantly improved the performance of a baseline recognition system. In another comparison against conventional enhancement algorithms, both the PESQ and the segmental SNR ratings of the LMS algorithm were superior to the other methods for noisy speech enhancement. Index Terms: corpus-based speech model, longest matching segment, speech enhancement, speech recognition

Calmodulin/KCa3.1 channel interactions as determinant to the KCa3.1 Ca2+ dependent gating : theoretical and experimental analyses

Differentes études ont montré que la sensibilité au Ca2+ du canal KCa3.1, un canal potassique indépendant du voltage, était conférée par la protéine calmoduline (CaM) liée de façon constitutive au canal. Cette liaison impliquerait la région C-lobe de la CaM et un domaine de $\ikca$ directement relié au segment transmembranaire S6 du canal. La CaM pourrait égalment se lier au canal de façon Ca2+ dépendante via une interaction entre un domaine de KCa3.1 du C-terminal (CaMBD2) et la région N-lobe de la CaM. Une étude fut entreprise afin de déterminer la nature des résidus responsables de la liaison entre le domaine CaMBD2 de KCa3.1 et la région N-lobe de la CaM et leur rôle dans le processus d'ouverture du canal par le Ca2+. Une structure 3D du complexe KCa3.1/CaM a d'abord été générée par modélisation par homologie avec le logiciel MODELLER en utilisant comme référence la structure cristalline du complexe SK2.2/CaM (PDB: 1G4Y). Le modèle ainsi obtenu de KCa3.1 plus CaM prévoit que le segment L361-S372 dans KCa3.1 devrait être responsable de la liaison dépendante du Ca2+ du canal avec la région N-lobe de la CaM via les résidus L361 et Q364 de KCa3.1 et E45, E47 et D50 de la CaM. Pour tester ce modèle, les résidus dans le segment L361-S372 ont été mutés en Cys et l'action du MTSET+ (chargé positivement) et MTSACE (neutre) a été mesurée sur l'activité du canal. Des enregistrements en patch clamp en configuration ``inside-out`` ont montré que la liaison du réactif chargé MTSET+ au le mutant Q364C entraîne une forte augmentation du courant, un effet non observé avec le MTSACE. De plus les mutations E45A et E47A dans la CaM, ont empêché l'augmentation du courant initié par MTSET+ sur le mutant Q364C. Une analyse en canal unitaire a confirmé que la liaison MTSET+ à Q364C cause une augmentation de la probabilité d'ouverture de KCa3.1 par une déstabilisation de l'état fermé du canal. Nous concluons que nos résultats sont compatibles avec la formation de liaisons ioniques entre les complexes chargés positivement Cys-MTSET+ à la position 364 de KCa3.1 et les résidus chargés négativement E45 et E47 dans la CaM. Ces données confirment qu'une stabilisation électrostatique des interactions CaM/KCa3.1 peut conduire à une augmentation de la probabilité d'ouverture du canal en conditions de concentrations saturantes de Ca2+.

Mechanism of Ca(v)1.2 channel modulation by the amino terminus of cardiac beta2-subunits

L-type calcium channels are composed of a pore, alpha1c (Ca(V)1.2), and accessory beta- and alpha2delta-subunits. The beta-subunit core structure was recently resolved at high resolution, providing important information on many functional aspects of channel modulation. In this study we reveal differential novel effects of five beta2-subunits isoforms expressed in human heart (beta(2a-e)) on the single L-type calcium channel current. These splice variants differ only by amino-terminal length and amino acid composition. Single-channel modulation by beta2-subunit isoforms was investigated in HEK293 cells expressing the recombinant L-type ion conducting pore. All beta2-subunits increased open probability, availability, and peak current with a highly consistent rank order (beta2a approximately = beta2b > beta2e approximately = beta2c > beta2d). We show graded modulation of some transition rates within and between deep-closed and inactivated states. The extent of modulation correlates strongly with the length of amino-terminal domains. Two mutant beta2-subunits that imitate the natural span related to length confirm this conclusion. The data show that the length of amino termini is a relevant physiological mechanism for channel closure and inactivation, and that natural alternative splicing exploits this principle for modulation of the gating properties of calcium channels.

Comparison of the ion channel characteristics of proapoptotic BAX and antiapoptotic BCL-2

The BCL-2 family of proteins is composed of both pro- and antiapoptotic regulators, although its most critical biochemical functions remain uncertain. The structural similarity between the BCL-XL monomer and several ion-pore-forming bacterial toxins has prompted electrophysiologic studies. Both BAX and BCL-2 insert into KCl-loaded vesicles in a pH-dependent fashion and demonstrate macroscopic ion efflux. Release is maximum at ≈pH 4.0 for both proteins; however, BAX demonstrates a broader pH range of activity. Both purified proteins also insert into planar lipid bilayers at pH 4.0. Single-channel recordings revealed a minimal channel conductance for BAX of 22 pS that evolved to channel currents with at least three subconductance levels. The final, apparently stable BAX channel had a conductance of 0.731 nS at pH 4.0 that changed to 0.329 nS when shifted to pH 7.0 but remained mildly Cl− selective and predominantly open. When BAX-incorporated lipid vesicles were fused to planar lipid bilayers at pH 7.0, a Cl−-selective (PK/PCl = 0.3) 1.5-nS channel displaying mild inward rectification was noted. In contrast, BCL-2 formed mildly K+-selective (PK/PCl = 3.9) channels with a most prominent initial conductance of 80 pS that increased to 1.90 nS. Fusion of BCL-2-incorporated lipid vesicles into planar bilayers at pH 7.0 also revealed mild K+ selectivity (PK/PCl = 2.4) with a maximum conductance of 1.08 nS. BAX and BCL-2 each form channels in artificial membranes that have distinct characteristics including ion selectivity, conductance, voltage dependence, and rectification. Thus, one role of these molecules may include pore activity at selected membrane sites.

A human intermediate conductance calcium-activated potassium channel

An intermediate conductance calcium-activated potassium channel, hIK1, was cloned from human pancreas. The predicted amino acid sequence is related to, but distinct from, the small conductance calcium-activated potassium channel subfamily, which is ≈50% conserved. hIK1 mRNA was detected in peripheral tissues but not in brain. Expression of hIK1 in Xenopus oocytes gave rise to inwardly rectifying potassium currents, which were activated by submicromolar concentrations of intracellular calcium (K0.5 = 0.3 μM). Although the K0.5 for calcium was similar to that of small conductance calcium-activated potassium channels, the slope factor derived from the Hill equation was significantly reduced (1.7 vs. 3.5). Single-channel current amplitudes reflected the macroscopic inward rectification and revealed a conductance level of 39 pS in the inward direction. hIK1 currents were reversibly blocked by charybdotoxin (Ki = 2.5 nM) and clotrimazole (Ki = 24.8 nM) but were minimally affected by apamin (100 nM), iberiotoxin (50 nM), or ketoconazole (10 μM). These biophysical and pharmacological properties are consistent with native intermediate conductance calcium-activated potassium channels, including the erythrocyte Gardos channel.

Functional consequences of NR2 subunit composition in single recombinant N-methyl-d-aspartate receptors

Single-channel recordings were obtained from Chinese hamster ovary cells transfected with the N-methyl-d-aspartate (NMDA) receptor subunit NR1 in combination with NR2A, NR2B, NR2C, or NR2A/NR2B. NMDA-activated currents were recorded under control conditions and in the presence of a thiol reductant (DTT), an oxidant (5,5′-dithio-bis[2-nitrobenzoic acid], DTNB), or the noncompetitive antagonist CP101,606 (CP). For all subunit combinations, DTT increased the frequency of channel opening when compared with DTNB. In addition, channels obtained from NR1/NR2A-transfected cells also exhibited a pronounced difference in mean open dwell-time between redox conditions. CP dramatically reduced both the open dwell-time and frequency of channel opening of NR1/NR2B-containing receptors, but only modestly inhibited NR1/NR2A and NR1/NR2C channel activity. A small number of patches obtained from cells transfected with NR1/NR2A/NR2B had channels with properties intermediate to NR1/NR2A and NR1/NR2B receptors, including insensitivity to CP block but redox properties similar to NR1/NR2B, consistent with the coassembly of NR2A with NR2B. Hence, NMDA receptors containing multiple types of NR2 subunits can have functionally distinguishable attributes.

Basolateral membrane targeting of a renal-epithelial inwardly rectifying potassium channel from the cortical collecting duct, CCD-IRK3, in MDCK cells

We recently cloned an inward-rectifying K channel (Kir) cDNA, CCD-IRK3 (mKir 2.3), from a cortical collecting duct (CCD) cell line. Although this recombinant channel shares many functional properties with the “small-conductance” basolateral membrane Kir channel in the CCD, its precise subcellular localization has been difficult to elucidate by conventional immunocytochemistry. To circumvent this problem, we studied the targeting of several different epitope-tagged CCD-IRK3 in a polarized renal epithelial cell line. Either the 11-amino acid span of the vesicular stomatitis virus (VSV) G glycoprotein (P5D4 epitope) or a 6-amino acid epitope of the bovine papilloma virus capsid protein (AU1) was genetically engineered on the extreme N terminus of CCD-IRK3. As determined by patch-clamp and two-microelectrode voltage-clamp analyses in Xenopus oocytes, neither tag affected channel function; no differences in cation selectivity, barium block, single channel conductance, or open probability could be distinguished between the wild-type and the tagged constructs. MDCK cells were transfected with tagged CCD-IRK3, and several stable clonal cell lines were generated by neomycin-resistance selection. Immunoprecipitation studies with anti-P5D4 or anti-AU1 antibodies readily detected the predicted-size 50-kDa protein in the transfected cells lines but not in wild-type or vector-only (PcB6) transfected MDCK cells. As visualized by indirect immunofluorescence and confocal microscopy, both the tagged CCD-IRK3 forms were exclusively detected on the basolateral membrane. To assure that the VSV G tag was not responsible for the targeting, the P5D4 epitope modified by a site-directed mutagenesis (Y2F) to remove a potential basolateral targeting signal contained in this tag. VSV(Y2F) was also detected exclusively on the basolateral membrane, confirming bona fide IRK3 basolateral expression. These observations, with our functional studies, suggest that CCD-IRK3 may encode the small-conductance CCD basolateral K channel.

Cell surface expression of the epithelial Na channel and a mutant causing Liddle syndrome: A quantitative approach

The epithelial amiloride-sensitive sodium channel (ENaC) controls transepithelial Na+ movement in Na+-transporting epithelia and is associated with Liddle syndrome, an autosomal dominant form of salt-sensitive hypertension. Detailed analysis of ENaC channel properties and the functional consequences of mutations causing Liddle syndrome has been, so far, limited by lack of a method allowing specific and quantitative detection of cell-surface-expressed ENaC. We have developed a quantitative assay based on the binding of 125I-labeled M2 anti-FLAG monoclonal antibody (M2Ab*) directed against a FLAG reporter epitope introduced in the extracellular loop of each of the α, β, and γ ENaC subunits. Insertion of the FLAG epitope into ENaC sequences did not change its functional and pharmacological properties. The binding specificity and affinity (Kd = 3 nM) allowed us to correlate in individual Xenopus oocytes the macroscopic amiloride-sensitive sodium current (INa) with the number of ENaC wild-type and mutant subunits expressed at the cell surface. These experiments demonstrate that: (i) only heteromultimeric channels made of α, β, and γ ENaC subunits are maximally and efficiently expressed at the cell surface; (ii) the overall ENaC open probability is one order of magnitude lower than previously observed in single-channel recordings; (iii) the mutation causing Liddle syndrome (β R564stop) enhances channel activity by two mechanisms, i.e., by increasing ENaC cell surface expression and by changing channel open probability. This quantitative approach provides new insights on the molecular mechanisms underlying one form of salt-sensitive hypertension.

Stretch-activated single K+ channels account for whole-cell currents elicited by swelling

Functionally significant stretch-activated ion channels have been clearly identified in excitable cells. Although single-channel studies suggest their expression in other cell types, their activity in the whole-cell configuration has not been shown. This discrepancy makes their physiological significance doubtful and suggests that their mechanical activation is artifactual. Possible roles for these molecules in nonexcitable cells are acute cell-volume regulation and, in epithelial cells, the complex adjustment of ion fluxes across individual cell membranes when the rate of transepithelial transport changes. We report the results of experiments on isolated epithelial cells expressing in the basolateral membrane stretch-activated K+ channels demonstrable by the cell-attached patch-clamp technique. In these cells, reversible whole-cell currents were elicited by both isosmotic and hyposmotic cell swelling. Cation selectivity and block by inorganic agents were the same for single-channel and whole-cell currents, indicating that the same entity underlies single-channel and whole-cell currents and that the single-channel events are not artifactual. In these cells, when the rate of apical-membrane NaCl entry increases, the cell Na+ content and volume also increase, stimulating the Na+,K+-ATPase at the basolateral membrane, i.e., both Na+ extrusion and K+ uptake increase. We speculate that, under these conditions, the parallel activation of basolateral K+ channels (by the swelling) elevates conductive K+ loss, tending to maintain the cell K+ content constant (“pump-leak parallelism”). This study describes a physiologically relevant stretch-activated channel, at both the single-channel and whole-cell levels, in a nonneural cell type.

Inositol 1,4,5-tris-phosphate activation of inositol tris-phosphate receptor Ca2+ channel by ligand tuning of Ca2+ inhibition

Inositol 1,4,5-tris-phosphate (IP3) binding to its receptors (IP3R) in the endoplasmic reticulum (ER) activates Ca2+ release from the ER lumen to the cytoplasm, generating complex cytoplasmic Ca2+ concentration signals including temporal oscillations and propagating waves. IP3-mediated Ca2+ release is also controlled by cytoplasmic Ca2+ concentration with both positive and negative feedback. Single-channel properties of the IP3R in its native ER membrane were investigated by patch clamp electrophysiology of isolated Xenopus oocyte nuclei to determine the dependencies of IP3R on cytoplasmic Ca2+ and IP3 concentrations under rigorously defined conditions. Instead of the expected narrow bell-shaped cytoplasmic free Ca2+ concentration ([Ca2+]i) response centered at ≈300 nM–1 μM, the open probability remained elevated (≈0.8) in the presence of saturating levels (10 μM) of IP3, even as [Ca2+]i was raised to high concentrations, displaying two distinct types of functional Ca2+ binding sites: activating sites with half-maximal activating [Ca2+]i (Kact) of 210 nM and Hill coefficient (Hact) ≈2; and inhibitory sites with half-maximal inhibitory [Ca2+]i (Kinh) of 54 μM and Hill coefficient (Hinh) ≈4. Lowering IP3 concentration was without effect on Ca2+ activation parameters or Hinh, but decreased Kinh with a functional half-maximal activating IP3 concentration (KIP3) of 50 nM and Hill coefficient (HIP3) of 4 for IP3. These results demonstrate that Ca2+ is a true receptor agonist, whereas the sole function of IP3 is to relieve Ca2+ inhibition of IP3R. Allosteric tuning of Ca2+ inhibition by IP3 enables the individual IP3R Ca2+ channel to respond in a graded fashion, which has implications for localized and global cytoplasmic Ca2+ concentration signaling and quantal Ca2+ release.

Flexibility of the Kir6.2 inward rectifier K+ channel pore

Interactions of sulfhydryl reagents with introduced cysteines in the pore-forming (Kir6.2) subunits of the KATP channel were examined. 2-Aminoethyl methanethiosulfonate (MTSEA+) failed to modify Cd2+-insensitive control-Kir6.2 channels, but rapidly and irreversibly modified Kir6.2[L164C] (L164C) channels. Although a single Cd2+ ion is coordinated by L164C, four MTSEA+ “hits” can occur, each sequentially reducing the single-channel current. A dimeric fusion of control-Kir6.2 and L164C subunits generates Cd2+-insensitive channels, confirming that at least three cysteines are required for coordination, but MTSEA+ modification of the dimer occurs in two hits. L164C channels were not modified by bromotrimethyl ammoniumbimane (qBBr+), even though qBBr+ caused voltage-dependent block (as opposed to modification) that was comparable to that of MTSEA+ or 3-(triethylammonium)propyl methanethiosulfonate (MTSPTrEA+), implying that qBBr+ can also enter the inner cavity but does not modify L164C residues. The Kir channel pore structure was modeled by homology with the KcsA crystal structure. A stable conformation optimally places the four L164C side chains for coordination of a single Cd2+ ion. Modification of these cysteines by up to four MTSEA+ (or three MTSPTrEA+, or two qBBr+) does not require widening of the cavity to accommodate the derivatives within it. However, like the KcsA crystal structure, the energy-minimized model shows a narrowing at the inner entrance, and in the Kir6.2 model this narrowing excludes all ions. To allow entry of ions as large as MTSPTrEA+ or qBBr+, the entrance must widen to >8 Å, but this widening is readily accomplished by minimal M2 helix motion and side-chain rearrangement.

A Calcium-Selective Channel from Root-Tip Endomembranes of Garden Cress1

A Ca2+ channel from root-tip endomembranes of garden cress (Lepidium sativum L.) (LCC1) was characterized using the planar lipid-bilayer technique. Investigation of single-channel recordings revealed that LCC1 is voltage gated and strongly rectifying. In symmetrical 50 mm CaCl2 solutions, the single-channel conductance was 24 picosiemens. LCC1 showed a moderate selectivity for Ca2+ over K+ (9.4:1) and was permeable for a range of divalent cations (Ca2+, Ba2+, and Sr2+). In contrast to Bryonia dioica Ca2+ channel 1, a Ca2+-selective channel from the endoplasmic reticulum of touch-sensitive tendrils, LCC1 showed no bursting channel activity and had a low open probability and mean open time (2.83 ms at 50 mV). Inhibitor studies demonstrated that LCC1 is blocked by micromolar concentrations of erythrosin B (inhibitor concentration for 50% inhibition [IC50] = 1.8 μm) and the trivalent cations La3+ (IC50 = 5 μm) and Gd3+ (IC50 = 10 μm), whereas verapamil showed no blocking effect. LCC1 may play an important role in the regulation of the cytoplasmic free Ca2+ concentration in root-tip and/or root-cap cells. The question of whether this ion channel is part of the gravitropic signal transduction pathway deserves further investigation.

Ion channel genes and human neurological disease: Recent progress, prospects, and challenges

What do epilepsy, migraine headache, deafness, episodic ataxia, periodic paralysis, malignant hyperthermia, and generalized myotonia have in common? These human neurological disorders can be caused by mutations in genes for ion channels. Many of the channel diseases are “paroxysmal disorders” whose principal symptoms occur intermittently in individuals who otherwise may be healthy and active. Some of the ion channels that cause human neurological disease are old acquaintances previously cloned and extensively studied by channel specialists. In other cases, however, disease-gene hunts have led the way to the identification of new channel genes. Progress in the study of ion channels has made it possible to analyze the effects of human neurological disease-causing channel mutations at the level of the single channel, the subcellular domain, the neuronal network, and the behaving organism.

Stretching dependence of the vibration modes of a single-molecule Pt-H2-Pt bridge

A conducting bridge of a single hydrogen molecule between Pt electrodes is formed in a break junction experiment. It has a conductance near the quantum unit, G0=2e2∕h, carried by a single channel. Using point-contact spectroscopy three vibration modes are observed and their variation upon isotope substitution is obtained. The stretching dependence for each of the modes allows uniquely classifying them as longitudinal or transversal modes. The interpretation of the experiment in terms of a Pt-H2-Pt bridge is verified by density-functional theory calculations for the stability, vibrational modes, and conductance of the structure.

Selective modulation of neuronal nicotinic acetylcholine receptor channel subunits by G(o)-protein subunits

protein modulation of neuronal nicotinic acetylcholine receptor ( nAChR) channels in rat intrinsic cardiac ganglia was examined using dialyzed whole-cell and excised membrane patch-recording configurations. Cell dialysis with GTP gamma S increased the agonist affinity of nAChRs, resulting in a potentiation of nicotine-evoked whole-cell currents at low concentrations. ACh- and nicotine-evoked current amplitudes were increased approximately twofold in the presence of GTP gamma S. In inside-out membrane patches, the open probability (NPo) of nAChR-mediated unitary currents was reversibly increased fourfold after bath application of 0.2mM GTP gamma S relative to control but was unchanged in the presence of GDP gamma S. The modulation of nAChR-mediated whole- cell currents was agonist specific; currents evoked by the cholinergic agonists ACh, nicotine, and 1,1-dimethyl-4-phenylpiperazinium iodide, but not cytisine or choline, were potentiated in the presence of GTP gamma S. The direct interaction between G-protein subunits and nAChRs was examined by bath application of either G(o)alpha or G beta gamma subunits to inside-out membrane patches and in glutathione S-transferase pull-down and coimmunoprecipitation experiments. Bath application of 50 nM G beta gamma increased the open probability of ACh- activated single-channel currents fivefold, whereas G(o)alpha( 50 nM) produced no significant increase in NPo. Neuronal nAChR subunits alpha 3-alpha 5 and alpha 2 exhibited a positive interaction with G(o)alpha and G beta gamma, whereas beta 4 and alpha 7 failed to interact with either of the G-protein subunits. These results provide evidence for a direct interaction between nAChR and G-protein subunits, underlying the increased open probability of ACh-activated single-channel currents and potentiation of nAChR-mediated whole-cell currents in parasympathetic neurons of rat intrinsic cardiac ganglia.