Proton flux across artificial vesicle bilayers

The study of proton conductance across artificial membranes has revealed a surprisingly high permeability for H+, (Pnet H+). A high Pnet H+ is difficult to reconcile with the biological requirement for the maintenance of pH gradients across the plasma membranes of cells, organellar study was undertaken to examine the role played by cholesterol and phospholipid fatty acid side chain composition in determining how well a membrane will function as a barrier to acid. The effects of counter-ion movement on acidification rates were examined in order to interpret the data obtained from variations in membrane composition. In phosphate buffered saline solutions, vesicle membranes composed of unsaturated fatty acid phosphatidylcholines proved to be poorer barriers to acid than membranes composed of saturated fatty acids. The barrier properties of these membranes could be ranked in the following order: DPL, (palmitic) $>$ Egg PC, (mixed chains) $>$ DLL, (linoleic), with DPL being the most effective in maintaining a one pH unit gradient near neutrality. Cholesterol decreased acidification rates of membranes made from the unsaturated phosphatidylcholines Egg PC and DLL, but enhanced acidification rates in vesicle membranes composed of the saturated phospholipid DPL. The cholesterol and fatty acid side chain effects were mediated by changes in membrane fluidity, with more rigid bilayers forming better barriers to acid. Experimental evidence was obtained which confirmed the Pnet H+ is very high relative to the permeabilities of other ions. Counter-ion controlled acidification rates depended on the size and charge of the ion which was moving in order to maintain electroneutrality. The biological relevance of a high intrinsic Pnet H+ and the possible role of counter-ion controlled acidification were discussed. ^

Effect of indomethacin on bile acid-phospholipid interactions: implication for small intestinal injury induced by nonsteroidal anti-inflammatory drugs.

The injurious effect of nonsteroidal anti-inflammatory drugs (NSAIDs) in the small intestine was not appreciated until the widespread use of capsule endoscopy. Animal studies found that NSAID-induced small intestinal injury depends on the ability of these drugs to be secreted into the bile. Because the individual toxicity of amphiphilic bile acids and NSAIDs directly correlates with their interactions with phospholipid membranes, we propose that the presence of both NSAIDs and bile acids alters their individual physicochemical properties and enhances the disruptive effect on cell membranes and overall cytotoxicity. We utilized in vitro gastric AGS and intestinal IEC-6 cells and found that combinations of bile acid, deoxycholic acid (DC), taurodeoxycholic acid, glycodeoxycholic acid, and the NSAID indomethacin (Indo) significantly increased cell plasma membrane permeability and became more cytotoxic than these agents alone. We confirmed this finding by measuring liposome permeability and intramembrane packing in synthetic model membranes exposed to DC, Indo, or combinations of both agents. By measuring physicochemical parameters, such as fluorescence resonance energy transfer and membrane surface charge, we found that Indo associated with phosphatidylcholine and promoted the molecular aggregation of DC and potential formation of larger and isolated bile acid complexes within either biomembranes or bile acid-lipid mixed micelles, which leads to membrane disruption. In this study, we demonstrated increased cytotoxicity of combinations of bile acid and NSAID and provided a molecular mechanism for the observed toxicity. This mechanism potentially contributes to the NSAID-induced injury in the small bowel.

Regulation of phospholipid biosynthesis in {\it Saccharomyces cerevisiae\/} by CDP-diacylglycerol synthase

Amine-containing phospholipid synthesis in Saccharomyces cerevisiae starts with the conversion of CDP-diacylglycerol (CDP-DAG) and serine to phosphatidylserine (PS) while phosphatidylinositol (PI) is formed from CDP-DAG and inositol (derived from inositol-1-phosphate). In this study a gene (CDS1) encoding CDP-DAG synthase in S. cerevisiae was isolated and identified. The CDS1 gene encodes the majority, if not all, of the synthase activity, and is essential for cell growth. Overexpression of the CDS1 gene resulted in an elevation in the apparent initial rate of synthesis and also steady-state level of PI relative to PS in both wild type yeast and the cds1 mutant. Down-regulation of CDS1 expression resulted in an inositol excretion phenotype and an opposite effect on the above phospholipid synthesis in the cds1 mutant. This regulation of phospholipid biosynthesis is mediated by changes of the phospholipid biosynthetic enzymes via a mechanism independent of the expression of the INO2-OPI1 regulatory genes. Reduction in the level of CDP-DAG synthase activity resulted in an increase in PS synthase activity which followed a similar change in the CHO1/PSS (encodes PS synthase) mRNA level. INO1 (encodes inositol-1-phosphate synthase) mRNA also increased but only after CDP-DAG synthase activity fell below the wild type level. PI synthase activity followed the decrease of the CDP-DAG synthase activity, but there was no parallel change in the level of PIS1 mRNA. A G$\sp{305}$/A$\sp{305}$ point mutation within the CDS1 gene which causes the cdg1 phenotype was identified. A human cDNA clone encoding CDP-DAG synthase activity was characterized by complementation of the yeast cds1 null mutant. ^

Identification and characterization of genes encoding phospholipid biosynthetic enzymes of {\it Saccharomyces cerevisiae\/}

Phospholipids are the major component of cellular membranes. In addition to its structural role, phospholipids play an active and diverse role in cellular processes. The goal of this study is to identify the genes involved in phospholipid biosynthesis in a model eukaryotic system, Saccharomyces cerevisiae. We have focused on the biosynthetic steps localized in the inner mitochondrial membrane; hence, the identification of the genes encoding phosphatidylserine decarboxylase (PSD1), cardiolipin synthase (CLS1), and phosphatidylglycerophosphate synthase (PGS1).^ The PSD1 gene encoding a phosphatidylserine decarboxylase was cloned by complementation of a conditional lethal mutation in the homologous gene in Escherichia coli strain EH150. Overexpression of the PSD1 gene in wild type yeast resulted in 20-fold amplification of phosphatidylserine decarboxylase activity. Disruption of the PSD1 gene resulted in 20-fold reduction of decarboxylase activity, but the PSD1 null mutant exhibited essentially normal phenotype. These results suggest that yeast has a second phosphatidylserine decarboxylation activity.^ Cardiolipin is the major anionic phospholipid of the inner mitochondrial membrane. It is thought to be an essential component of many biochemical functions. In eukaryotic cells, cardiolipin synthase catalyzes the final step in the synthesis of cardiolipin from phosphatidylglycerol and CDP-diacylglycerol. We have cloned the gene CLS1. Overexpression of the CLS1 gene product resulted in significantly elevated cardiolipin synthase activity, and disruption of the CLS1 gene, confirmed by PCR and Southern blot analysis, resulted in a null mutant that was viable and showed no petite phenotype. However, phospholipid analysis showed undetectable cardiolipin level and an accumulation of phosphatidylglycerol. These results support the conclusion that CLS1 encodes the cardiolipin synthase of yeast and that normal levels of cardiolipin are not absolutely essential for survival of the cell.^ Phosphatidylglycerophosphate (PGP) synthase catalyzes the synthesis of PGP from CDP-diacylglycerol and glycerol-3-phosphate and functions as the committal and rate limiting step in the biosynthesis of cardiolipin. We have identified the PGS1 gene as encoding the PGP synthase. Overexpression of the PGS1 gene product resulted in over 15-fold increase in in vitro PGP synthase activity. Disruption of the PGS1 gene in a haploid strain of yeast, confirmed by Southern blot analysis, resulted in a null mutant strain that was viable but had significantly altered phenotypes, i.e. inability to grow on glycerol and at $37\sp\circ$C. These cells showed over a 10-fold decrease in PGP synthase activity and a decrease in both phosphatidylglycerol and cardiolipin levels. These results support the conclusion that PGS1 encodes the PGP synthase of yeast and that neither phosphatidylglycerol nor cardiolipin are absolutely essential for survival of the cell. ^

Platelet-activating factor is crucial in psoralen and ultraviolet A-induced immune suppression, inflammation, and apoptosis.

Psoralen plus UVA (PUVA) is used as a very effective treatment modality for various diseases, including psoriasis and cutaneous T-cell lymphoma. PUVA-induced immune suppression and/or apoptosis are thought to be responsible for the therapeutic action. However, the molecular mechanisms by which PUVA acts are not well understood. We have previously identified platelet-activating factor (PAF), a potent phospholipid mediator, as a crucial substance triggering ultraviolet B radiation-induced immune suppression. In this study, we used PAF receptor knockout mice, a selective PAF receptor antagonist, a COX-2 inhibitor (presumably blocking downstream effects of PAF), and PAF-like molecules to test the role of PAF receptor binding in PUVA treatment. We found that activation of the PAF pathway is crucial for PUVA-induced immune suppression (as measured by suppression of delayed type hypersensitivity to Candida albicans) and that it plays a role in skin inflammation and apoptosis. Downstream of PAF, interleukin-10 was involved in PUVA-induced immune suppression but not inflammation. Better understanding of PUVA's mechanisms may offer the opportunity to dissect the therapeutic from the detrimental (ie, carcinogenic) effects and/or to develop new drugs (eg, using the PAF pathway) that act like PUVA but have fewer side effects.

Reversal of myocyte hypertrophy by ventricular unloading: cardiac improvement without adrenergic receptor up-regulation and relocalization.

In previous studies, we found that the improved contractile ability of cardiac myocytes from patients who have had left ventricular assist device (LVAD) support was due to a number of beneficial changes, most notably in calcium handling (increased sarcoplasmic reticulum calcium binding and uptake), improved integrity of cell membranes due to phospholipid reconstruction (reduced lysophospholipid content), and an upregulation of adrenoreceptors (increased adrenoreceptor numbers). However, in the case presented here, there was no increase in adrenoreceptor number, which is something that we usually find in core tissue at the time of LVAD removal or organ transplantation; also, there was no homogeneous postassist device receptor distribution. However, the patient was well maintained for 10 months following LVAD implantation, until a donor organ was available, regardless of the lack of adrenoreceptor improvement. We conclude from these studies that cardiac recovery is the result of the initiation of multiple repair mechanisms, and that the lack of expected changes, in this case increased adrenoreceptors, is not always an accurate indicator of anticipated outcome. We suggest that interventions and strategies have to consider multiple, beneficial changes due to unloading and target a number of biochemical and structural areas to produce improvement, even if not all of these improvements occur.

Lipid-protein interactions drive membrane protein topogenesis in accordance with the positive inside rule.

Transmembrane domain orientation within some membrane proteins is dependent on membrane lipid composition. Initial orientation occurs within the translocon, but final orientation is determined after membrane insertion by interactions within the protein and between lipid headgroups and protein extramembrane domains. Positively and negatively charged amino acids in extramembrane domains represent cytoplasmic retention and membrane translocation forces, respectively, which are determinants of protein orientation. Lipids with no net charge dampen the translocation potential of negative residues working in opposition to cytoplasmic retention of positive residues, thus allowing the functional presence of negative residues in cytoplasmic domains without affecting protein topology.

Cryo-EM model of the bullet-shaped vesicular stomatitis virus.

Vesicular stomatitis virus (VSV) is a bullet-shaped rhabdovirus and a model system of negative-strand RNA viruses. Through direct visualization by means of cryo-electron microscopy, we show that each virion contains two nested, left-handed helices: an outer helix of matrix protein M and an inner helix of nucleoprotein N and RNA. M has a hub domain with four contact sites that link to neighboring M and N subunits, providing rigidity by clamping adjacent turns of the nucleocapsid. Side-by-side interactions between neighboring N subunits are critical for the nucleocapsid to form a bullet shape, and structure-based mutagenesis results support this description. Together, our data suggest a mechanism of VSV assembly in which the nucleocapsid spirals from the tip to become the helical trunk, both subsequently framed and rigidified by the M layer.

AUTOIMMUNE RESPONSES TO ATHEROSCLEROTIC

Atherosclerosis is a chronic, complex arterial disease characterized by intimal lipid accumulation and inflammation. A unique lipid-binding molecule, namely cluster of differentiation 1d (CD1d), may impact atherosclerosis. Structurally, CD1d acts as a nonpolymorphic cell-surface receptor, resembling the major histocompatibility complex-I (MHC-I). While MHC-I restricts peptide antigen presentation to T cells, CD1d presents lipid antigens to T cells named CD1d-restrictedd T cells. Although increased expression of CD1d has been found in human plaques, the exact nature of CD1d-recognized lipids in atherosclerosis remains to be determined. Three groups of lipids may undergo oxidation in atherosclerosis producing atherogenic lipids: phospholipids, fatty acids, and cholesterol. The central hypothesis is that CD1d recognizes and present oxidative lipids to activate CD1d-restricted T cells, and trigger proinflammatory signal transduction In the first part of this study, oxidative phospholipids were identified and characterized as potential autoantigen for CD1d-restricted T cells. Derived from phospholipid 1-palmitoyl-2-arachidonoyl-sn-glycero-3-phosphorylcholine by oxidization, 1-palmitoyl-2-glutaryl-sn-glycero-3-phosphocholine (PGPC) is commonly found in atherosclerotic plaques. Upon stimulation with PGPC, spleen-derived CD1d-restricted T cells produced higher levels of cytokines and proliferated at higher rates than those without PGPC stimulation. CD1d deficiency compromised the PGPC-triggered T cell activation, suggesting that PGPC may function as a potentially novel autoantigen for T cells in atherosclerosis. In the second part of this study, CD1d-mediated proinflammatory signaling was evaluated in murine models. Enhanced CD1 expression occurred in spleens of db/db mice with hyperlipidemia. Tumor necrosis factor-alpha (TNF-α) was increased in db/db spleen, while TNF-α receptor expression augmented in the db/db murine heart, in comparison with those in normal mice. The nuclear factor-κ B (NF-κB) expression was enhanced in the db/db heart, whereas CD1d-null mice showed lower NF-κB, implying the involvement of CD1d in inflammation of the spleen and heart tissues in the mice with hyperlipidemia. The current study has identified PGPC as a novel lipid antigen recognized by CD1d-restricted T cells in atherosclerosis. The animal study has also provided evidence that CD1d regulates NF-κB-mediated proinflammatory signaling. Hence, CD1d-restricted T cell responses to autolipid antigen and mediated inflammatory signal may represent a new molecular pathway that triggers cardiovascular tissue injury in atherosclerosis and hyperlipidemia.

Calcification of a collagen-liposome complex

The purpose of the study was to evaluate in vitro calcification potential among liposomes composed of phospholipids with variations in fatty acid chains and polar head groups. The liposome was also modified by utilizing mixed phospholipids, incorporation of different types of protein to the liposome, or complexing with various collagen preparations. The samples were then incubated in a metastable calcium phosphate solution for the proposed time period. Calcium and phosphate uptake were measured. Resulting precipitates were processed for x-ray diffraction and electron microscopy. Acidic phospholipid, Dioleoylphosphatidic acid and mixed phospholipids, Dioleoylphosphatidic acid/Dipalmitoylphosphatidylethanolamine liposomes calcified at a faster rate and to a greater degree than other phospholipids tested. The incorporation of polylysine, fibronectin, bone protein, or the complexing with collagen decreased the rate and amount of calcification. Electron microscopy demonstrated the similarity of the calcified collagen-liposome complex to the natural calcification matrix. These preparations may be used as a model to study the role of membrane lipids and collagen-phospholipid during the process of calcification.^ The in vivo study was designed to determine whether the potential existed for the promotion of bone healing by the synthetic liposome-collagen complex. The implant materials were modified to provide decreased antigenicity, biocompatability while maintaining their bone conduction properties. The samples were placed subcutaneously and/or subperiosteally and/or in 8 mm calvarium defects of adult rats. Histological and immunological studies demonstrated that the implant itself retained minimal antigenicity and did not inhibit bone formation. However, modification of the implant may contain the bone induction property and be utilized to stimulate bony healing. ^

Studies on the {\it in vivo} function of a phosphatidylinositol transfer protein from {\it Saccharomyces cerevisiae}

Phosphatidylinositol transfer proteins (PI-TP's) catalyze the transfer of phosphatidylinositol and phosphatidylcholine between membranes in vitro. However the in vivo function of these proteins is unknown. In this thesis we have used a combined biochemical and genetic approach to determine the importance of PI-TP in vivo. An oligonucleotide based on the amino terminal sequence of the PI-TP from Saccharomyces cerevisiae, was used to screen a yeast genomic library for the gene encoding PI-TP (PIT1 gene). Yeast strains transformed with the positive clones showed overproduction of transfer activities and transfer protein in the 100,000 x g supernatants. The 5$\sp\prime$ terminus of the PIT1 gene correlates with the predicted codons for residues 3-30 of the determined protein sequence. Tetrad analysis of a heterozygous diploid (PIT1/pit1::LEU2) revealed that the PIT1 gene is essential for cell growth. Non-viable spores could be rescued by transformation of the above diploid prior to sporulation, with a plasmid borne copy of the wild type gene. Sequencing of the entire PIT1 gene has revealed that the PIT1 gene is identical to the SEC14 gene. The sec14 ts mutant which exhibits conditional defects at the Golgi stage of protein secretion, is also temperature sensitive for PI-TP activity in vitro. These findings represent the first instance in which a physiological function has been assigned to any phospholipid transfer protein. ^

The characterization of an epithelial chloride channel and purification of a channel component

A membrane fraction (M$\sb{\rm PS}$), enriched in Cl$\sp-$ channels, has been isolated from bovine tracheal epithelia and renal cortex homogenates by hydrophobic chromatography. The tracheal fraction shows a 37 fold enrichment of Cl$\sp-$ channels over crude tracheal homogenates by net Cl$\sp-$ measurements in membrane vesicles. Alkaline phosphatase and (Na$\sp+$ + K$\sp+$)-ATPase are not found in these membranes, suggesting that they are not apical or basolateral plasma membranes. The M$\sb{\rm PS}$ fraction exhibits a protein profile unlike that of other membrane fractions with major proteins of 200 kDa and 42 kDa, proteins of 30 to 35 kDa, and lesser amounts of other proteins. Reconstitution of M$\sb{\rm PS}$ fractions from both trachea and kidney into planar lipid bilayers demonstrates the presence of a single type of anion channel. The current-voltage relationship of this channel is linear with a slope conductance of 84 pS in symmetrical 400 mM KCl, and is identical to that of the predominant anion channel observed in tracheal apical membranes under similar conditions (Valdivia, Dubinsky, and Coronado. Science, 1988). In addition, the voltage dependence, selectivity sequence of Cl$\sp- >$ Br$\sp- \ge$ I$\sp-$, and inhibition by low concentrations of the Cl$\sp-$ channel blocker, DIDS, correspond to those of the predominant apical membrane channel. Thus, although the M$\sb{\rm PS}$ fraction appears to be of subcellular origin, it may be functionally related to an apical membrane Cl$\sp-$ permeability. When renal M$\sb{\rm PS}$ membranes were treated with the detergent octyl-glucoside (OG, 2%) and centrifuged, the supernatant, sM$\sb{\rm PS}$, showed a 2 to 7-fold enrichment in specific Cl$\sp-$ flux activity compared with the detergent treated M$\sb{\rm PS}$. These solubilized proteins were then size fractionated on a Superose 12 HPLC gel filtration column, followed by fractionation on a Mono Q HPLC anion exchange column. Fractions that eluted in high salt consistently exhibited significant Cl$\sp-$ flux activity. These fractions had protein profiles consisting of a major band at 34 kDa, a band at 66 kDa, and variable faint bands. Fractions eluting in lower salt had protein profiles consisting of a single band at 34 kDa, and often had little or no Cl$\sp-$ flux activity. However, co-reconstitution of the low salt, solely 34 kDa protein-containing Mono Q fractions with sM$\sb{\rm PS}$ resulted in an enhancement of flux activity compared to that of sM$\sb{\rm PS}$ reconstituted alone. Flux assays of active Mono Q fractions showed that the channel retained its DIDS sensitivity. Applying sM$\sb{\rm PS}$ to a DIDS-affinity column and eluting with salt resulted in fractions with protein profiles again consisting of at least one major band at 34 kDa, a band at 66 kDa, and variable faint bands. Co-reconstitution with sM$\sb{\rm PS}$ again resulted in an enhancement of activity. Thus, the 34 kDa protein appears to be a component of the M$\sb{\rm PS}$ Cl$\sp-$ channel. ^

Glutamate receptors expressed in {\it Xenopus\/} oocytes: Studies in function and regulation

The amino acid glutamate is the primary excitatory neurotransmitter for the CNS and is responsible for the majority of fast synaptic transmission. Glutamate receptors have been shown to be involved in multiple forms of synaptic plasticity such as LTP, LTD, and the formation of specific synaptic connections during development. In addition to contributing to the plasticity of the CNS, glutamate receptors also are involved in, at least in part, various pathological conditions such as epilepsy, ischemic damage due to stroke, and Huntington's chorea. The regulation of glutamate receptors, particularly the ionotropic NMDA and AMPA/KA receptors is therefore of great interest. In this body of work, glutamate receptor function and regulation by kinase activity was examined using the Xenopus oocyte which is a convenient and faithful expression system for exogenous proteins. Glutamate receptor responses were measured using the two-electrode voltage clamp technique in oocytes injected with rat total forebrain RNA. NMDA elicited currents that were glycine-dependent, subject to block by Mg$\sp{2+}$ in a voltage-dependent manner and sensitive to the specific NMDA antagonist APV in a manner consistent with those types of responses found in neural tissue. Similarly, KA-evoked currents were sensitive to the specific AMPA/KA antagonist CNQX and exhibited current voltage relationships consistent with the calcium permeable type II KA receptors found in the hippocampus. There is evidence to indicate that NMDA and AMPA/KA receptors are regulated by protein kinase A (PKA). We explored this by examining the effects of activators of PKA (forskolin, 1-isobutyl-3-methylxanthine (IBMX) and 8-Br-cAMP) on NMDA and KA currents in the oocyte. In buffer where Ca$\sp{2+}$ was replaced by 2 mM Ba$\sp{2+},$ forskolin plus IBMX and 8-Br-cAMP augmented currents due to NMDA application but not KA. This augmentation was abolished by pretreating the oocytes in the kinase inhibitor K252A. The use of chloride channel blockers resulted in attenuation of this effect indicating that Ba$\sp{2+}$ influx through the NMDA channel was activating the endogenous calcium-activated chloride current and that the cAMP mediated augmentation was at the level of the chloride channel and not the NMDA channel. This was confirmed by (1) the finding that 8-Br-cAMP increased chloride currents elicited via calcium channel activation while having no effect on the calcium channels themselves and (2) the fact that lowering the Ba$\sp{2+}$ concentration to 200 $\mu$M abolished the augmentation NMDA currents by 8-Br-cAMP. Thus PKA does not appear to modulate ionotropic glutamate receptors in our preparation. Another kinase also implicated in the regulation of NMDA receptors, calcium/phospholipid-dependent protein kinase (PKC), was examined for its effects on the NMDA receptor under low Ba$\sp{2+}$ (200 $\mu$M) conditions. Phorbol esters, activators of PKC, induced a robust potentiation of NMDA currents that was blockable by the kinase inhibitor K252A. Furthermore activation of metabotropic receptors by the selective agonist trans-ACPD, also potentiated NMDA albeit more modestly. These results indicate that neither NMDA nor KA-activated glutamate receptors are modulated by PKA in Xenopus oocytes whereas NMDA receptors appear to be augmented by PKC. Furthermore, the endogenous chloride current of the oocyte was found to be responsive to Ba$\sp{2+}$ and in addition is enhanced by PKA. Both of these latter findings are novel. In conclusion, the Xenopus oocyte is a useful expression system for the analysis of ligand-gated channel activity and the regulation of those channels by phosphorylation. ^

Functional analysis of anionic phospholipids in cell metabolism in {\it Escherichia coli\/}

The major goal of this work was to understand the function of anionic phospholipid in E. coli cell metabolism. One important finding from this work is the requirement of anionic phospholipid for the DnaA protein-dependent initiation of DNA replication. An rnhA mutation, which bypasses the need for the DnaA protein through induction of constitutive stable DNA replication, suppressed the growth arrest phenotype of a $pgsA$ mutant in which the synthesis of anionic phospholipid was blocked. The maintenance of plasmids dependent on an $oriC$ site for replication, and therefore DnaA protein, was also compromised under conditions of limiting anionic phospholipid synthesis. These results provide support for the involvement of anionic phospholipids in normal initiation of DNA replication at oriC in vivo by the DnaA protein. In addition, structural and functional requirements of two major anionic phospholipids, phosphatidylglycerol and cardiolipin, were examined. Introduction into cells of the ability to make phosphatidylinositol did not suppress the need for the naturally occurring phosphatidylglycerol. The requirement for phosphatidylglycerol was concluded to be more than maintenance of the proper membrane surface charge. Examination of the role of cardiolipin revealed its ability to replace the zwitterionic phospholipid, phosphatidylethanolamine, in maintaining an optimal membrane lipid organization. This work also reported the DNA sequence of the cls gene, which encodes the CL synthase responsible for the synthesis of cardiolipin. ^