Acetylation of insulin receptor substrate-1 is permissive for tyrosine phosphorylation
© Kaiser and James. 2004
Received: 07 July 2004
Accepted: 02 November 2004
Published: 02 November 2004
Skip to main content
© Kaiser and James. 2004
Received: 07 July 2004
Accepted: 02 November 2004
Published: 02 November 2004
Insulin receptor substrate (IRS) proteins are key moderators of insulin action. Their specific regulation determines downstream protein-protein interactions and confers specificity on growth factor signalling. Regulatory mechanisms that have been identified include phosphorylation of IRS proteins on tyrosine and serine residues and ubiquitination of lysine residues. This study investigated other potential molecular mechanisms of IRS-1 regulation.
Using the sos recruitment yeast two-hybrid system we found that IRS-1 and histone deacetylase 2 (HDAC2) interact in the cytoplasmic compartment of yeast cells. The interaction mapped to the C-terminus of IRS-1 and was confirmed through co-immunoprecipitation in vitro of recombinant IRS-1 and HDAC2. HDAC2 bound to IRS-1 in mammalian cells treated with phorbol ester or after prolonged treatment with insulin/IGF-1 and also in the livers of ob/ob mice but not PTP1B knockout mice. Thus, the association occurs under conditions of compromised insulin signalling. We found that IRS-1 is an acetylated protein, of which the acetylation is increased by treatment of cells with Trichostatin A (TSA), an inhibitor of HDAC activity. TSA-induced increases in acetylation of IRS-1 were concomitant with increases in tyrosine phosphorylation in response to insulin. These effects were confirmed using RNA interference against HDAC2, indicating that HDAC2 specifically prevents phosphorylation of IRS-1 by the insulin receptor.
Our results show that IRS-1 is an acetylated protein, a post-translational modification that has not been previously described. Acetylation of IRS-1 is permissive for tyrosine phosphorylation and facilitates insulin-stimulated signal transduction. Specific inhibition of HDAC2 may increase insulin sensitivity in otherwise insulin resistant conditions.
The insulin receptor substrate (IRS) proteins represent key elements in insulin and insulin-like growth factor (IGF) actions, transducing pleiotropic effects on cellular function and regulating processes such as metabolism, growth, cell differentiation and survival . At least four members (IRS 1–4) have been identified that differ with regard to tissue distribution, subcellular localization, developmental expression, binding to the insulin receptor, and interaction with Src homology 2 (SH2) domains. They are all structurally characterised by N-terminal pleckstrin-homology (PH) and phosphotyrosine-binding (PTB) domains, which are required for coupling to the activated insulin/IGF receptors, and a C-terminal region with multiple sites for tyrosine phosphorylation by the receptors. IRS proteins thus act as molecular adapters in recruiting, inter alia, a number of SH2-containing proteins binding to specific phosphorylated tyrosine residues. This leads to activation of different intracellular cascades , one of which is the PI 3-kinase signalling cascade implicated in mediating the metabolic effects of insulin .
The best-substantiated post-translational modification of IRS proteins, in addition to tyrosine phosphorylation, is phosphorylation of specific serine residues. Phosphorylation on these residues is associated both with inhibition of insulin-induced tyrosine phosphorylation of IRS proteins and with facilitation of the effects of insulin . Phosphorylation catalysed by protein kinase C (PKC) isozymes [5, 6], c-Jun N-terminal kinase (JNK) , inhibitor κB kinase (IKK) isozymes , mitogen activated protein kinases (MAPK)  and the mammalian target of rapamycin (mTOR)  are all associated with reducing the ability of insulin to stimulate tyrosine phosphorylation of IRS proteins and therefore may be part of the physiological and pathophysiological negative regulation of insulin signalling through the IRS pathway. Specific mechanisms explaining why serine phosphorylation leads to reduced tyrosine phosphorylation have not been completely identified, but candidates for this are reduced interaction of IRS proteins with the insulin receptor  and increased degradation of IRS [12, 13]. Furthermore, phosphorylation of different residues can lead to different effects. Thus, phosphorylation of serine 307 in rat IRS-1 (serine 312 in human IRS-1) is associated with reduced insulin signalling  whilst phosphorylation of serine 302 has recently been suggested to facilitate insulin signalling , although this has been contested .
In addition to phosphorylation of different amino acid residues, insulin signalling through IRS proteins has been shown to be regulated by at least two other mechanisms. Prolonged signal transduction via phosphoinositide 3-kinase (PI3K), which generates the lipid second messenger phosphatidylinositol 3,4,5-trisphosphate, has been shown to induce a state of insulin resistance in cells , in part through degradation of IRS-1 . Thus, insulin signalling can be negatively regulated through modulation of IRS concentrations in cells, via degradation of the proteins in the proteasomal pathway [18–20]. The mechanism by which IRS proteins are degraded by the proteasome is not completely understood, but the N-terminal PH and PTB domains are required . In addition, the sub-cellular localisation of IRS proteins may be important for appropriate insulin signalling. The sub-cellular localisation is not absolutely defined, with various lines of evidence pointing to potential places in the cell where the proteins can be found. In addition to the plasma membrane, IRS proteins have been associated with high-density pellets  implicating association with the cytoskeleton and recently also with the nucleus [23, 24]. Thus, IRS proteins may be located to different parts of the cell where they carry out different functions.
Multiple histone acetyltransferases (HATs) and histone deacetylases (HDACs) control the state of histone acetylation and hence play a regulatory role in modulating the structure and function of chromatin . About 20 HATs have been detected to date, grouped in three different classes on the basis of structural properties. They all have one structural motif in common, the so-called A-motif responsible for acetyl CoA recognition . Several HATs also have non-histone substrates but it is not yet possible to identify putative acetylation sites within a protein simply by sequence analysis. Generally, acetylation affects DNA-binding, protein-protein interactions, protein stability, and protein localization . The acetyl-mediated signals are reversed by HDACs that counteract the effects of HATs by deacetylating lysine residues on histone tails. In higher eukaryotes, HDACs can be subdivided into three distinct groups known as classes I, II, III, according to similarities of their sequences to those of yeast founding members . To date, four enzymes, HDAC1, 2, 3 and 8, are the known members of class I deacetylases [28, 29]. HDAC1 and 2 are the best characterised, and are chief constituents of the multiprotein transcriptional-repression complex Sin3/HDAC and the nucleosome remodelling deacetylase NuRD/Mi2/NRD complex . Complexes that contain class I HDACs bind to numerous transcription factors, either directly, or indirectly through the nuclear-hormone corepressors NCOR and SMRT (silencing mediator for retinoid and thyroid hormone receptors). Although all class I and II HDACs can deacetylate histone tails, other cellular proteins can be specifically targeted by different HDACs as well, such as α-tubulin and importin-α . Recent developments have shown that the class I enzymes are regulated by phosphorylation, by casein kinase II amongst others, which increases activity [32–34]. The fact that class II enzymes are phosphorylated has been known for longer, a reaction which is associated with re-localization of the enzymes to the cytoplasm through interactions with 14-3-3 proteins .
We now demonstrate that HDAC2 interacts with IRS-1 under conditions when the ability of cells to respond to insulin is compromised. As such, this interaction may constitute a new component of the negative regulation of IRS protein function. We also show that IRS-1 is acetylated, and that augmenting the acetylation level by treating cells with Trichostatin A (TSA, a non-specific inhibitor of HDACs) or with short inhibitory RNA oligonucleotides against HDAC2 partially restores normal responsiveness to insulin.
To map the interaction site of HDAC2 on IRS-1 we used a GAL4-based yeast two-hybrid system, where interactions take place in the nucleus of the yeast cell. Cells were transformed with vectors encoding full length HDAC2 and different truncation mutants of IRS-1. The truncations of IRS-1 that were used were the PH domain (residues 1–155), the PH-PTB domains (residues 1–578) and the PH-PTB-pre-C-terminal domains (residues 1–895). Using growth of yeast cells on selective medium as a readout for interaction between HDAC2 and IRS-1 showed that the interaction requires the C-terminal portion of the IRS-1 protein (Figure 1B). In order to confirm the interaction further in vitro, we used a coupled in vitro transcription/translation system in which full length IRS-1 and the HDAC2 N-terminal portion from the initial yeast two-hybrid screen were transcribed and translated in the presence of S35 methionine. IRS-1 was subsequently immunoprecipitated from the mixture and the proteins were resolved by SDS-PAGE. Gels were then subjected to autoradiography. Results showed that two radioactive protein bands were visible in the IRS-1 immunoprecipitates (Figure 1C) and their molecular weights corresponded to those of full length IRS-1 (approx 160 kD) and truncated HDAC2 (approx 35 kD). When the IRS-1 antibody was boiled prior to immunoprecipitation (Figure 1C lane 2) or omitted (Figure 1C lane 3), no radioactive proteins were observed, indicating that the interaction between the two proteins is specific and not due to non-specific interactions with immunoglobulins or beads. Thus, IRS-1 and HDAC2 proteins are able to interact with each other in cell-free systems.
To assess whether the interaction measured between IRS-1 and HDAC2 in vitro as described above occurs in vivo, we prepared lysates of liver tissues prepared from different mouse lines. The ob/ob mouse, which lacks functional leptin, was chosen as an insulin resistant animal model, and C57/bl6 was used as its genotype control. A PTP1B knockout mouse  was used as an insulin-sensitised animal model and balb/cJJ was used as its genotype control. IRS-1 was immunoprecipitated from liver lysates and western blotted for co-immunoprecipitation of HDAC2. The data showed that whilst a clear interaction between IRS-1 and HDAC2 was seen in livers from ob/ob mice (Figure 2C), no interaction was evident in the C57/bl6 control. In contrast, no interaction was evident in livers of PTP1B knockout animals, whilst the balb/cJJ genotype control demonstrated a measurable interaction. Taken together with the in vitro data, these results showed that IRS-1 and HDAC2 are able to interact with each other in the cytoplasmic compartment of cells and that the interaction occurs under conditions of reduced insulin sensitivity, both in mammalian cells and in animals. The cytoplasmic location of the interaction is interesting in view of the fact that HDAC2 is considered to be largely a nuclear protein. In our work with cells and tissues, we have utilised lysis methods that are designed to retain nuclei intact and thereby minimise cross-contamination of compartments [39, 40]. Whilst we have not formally excluded the possibility of contamination of cytoplasmic extracts with nuclear lysate, thereby leading to the presence of HDAC2, we feel that the body of evidence indicates that cytoplasmic HDAC2 is interacting with cytoplasmic IRS-1 in our experiments. The yeast two hybid "Sos recruitment system" is built on the rescue of cell growth through the interaction of proteins in the cytoplasm, which is how we detected this interaction. Interestingly, it has recently been shown that histone deacetylase 1, another class I histone deacetylase, which was considered to be exclusively nuclear, is present in a cytoplasmic protein complex by virtue of interaction with a cellular phosphatase complex .
The regulation and function of proteins such as sterol regulatory element binding protein 1c (SREBP1c)  and p53  has been shown to be altered by changes in acetylation. The alterations in lysine acetylation in IRS-1 induced by TSA raised the possibility that insulin signal transduction may be altered in cells after treatment with this compound. To assess the effects of changes in IRS-1 acetylation on insulin signalling, MCF-7 cells were treated with PMA, TSA and insulin in different combinations and immunoprecipitated IRS-1 protein was immunoblotted for the presence of phosphotyrosine. PMA alone and in combination with TSA did not increase tyrosine phosphorylation of IRS1 above basal, as expected (Fig 4 lanes 4–6). Furthermore, the ability of insulin to induce tyrosine phosphorylation of IRS-1 was reduced by 60% in cells pre-treated with PMA (Fig 4 lane 3) consistent with a state of insulin unresponsiveness. However, pre-treatment with TSA in the presence of PMA reduced this unresponsiveness, increasing insulin-stimulated tyrosine phosphorylation to 70% of control (Fig 4 lane 2). Thus, increases in IRS1 acetylation via TSA-mediated HDAC inhibition were able to restore insulin signalling significantly. This restoration occurred without reducing PMA-induced serine 312 phosphorylation of IRS-1 (Fig 2B lane 2), indicating that acetylation of IRS1 overcomes the inhibitory effects of phosphorylation of serine 312.
A major functional response downstream of the PI3K arm of insulin signal transduction is increased glucose transport mediated by the GLUT4 transporter. We sought to examine the effects of TSA on glucose transport in rat L6 myotubes to see if the enhanced insulin signalling mediated by TSA treatment of cells translated into increased glucose uptake. We found that treatment of L6 myotubes with PMA resulted in increased basal glucose transport and had no effect on insulin-stimulated glucose transport (Kaiser & James, unpublished). Such effects are in line with data presented for rat epitrochlearis muscle  and indicated that L6 cells do not exhibit a clear insulin-resistance phenotype after PMA treatment, at the level of glucose transport. We also have similar observations in the human neuroblastoma cell line SHSY-5Y, which demonstrates insulin-stimulated glucose uptake . Phorbol ester treatment of these cells increased basal glucose transport but in contrast to data in L6 cells, also inhibited insulin-stimulated glucose transport (Kaiser and James, unpublished). We have therefore not been able to distinguish an effect of TSA on GLUT4-mediated glucose transport owing to the large PMA-stimulated increases in insulin-independent glucose transport (presumably mediated by GLUT1), and are at present analysing other cells for their response to phorbol ester treatment. Interestingly, Takigawa-Imamura et al.  recently showed that several HDAC inhibitors increase glucose transport in muscle cells in culture. Although the treatment regimens with these inhibitors in these experiments were chronic, the data show that inhibition of HDAC activity enhances glucose transport. Molecular mechanisms behind this effect could be several, including enhanced insulin signalling through increases in intracellular protein acetylation.
The data we present here imply that treatment of insulin-resistant or diabetic animals with inhibitors of HDAC2 should increase insulin responsiveness. We attempted to assess the effects of TSA on insulin sensitivity in ob/ob mice. The animals were divided into two groups: vehicle (DMSO) and TSA (0.1 mg/kg) and treated subcutaneously for three days. At the same time as drug injection, all food was withdrawn from the animals and 4 hours later, blood was collected from the tail vein for blood glucose and plasma insulin analysis. On the third day, an insulin tolerance test (ITT) was performed 4 hours after administration of the drug. After 24 hours, fasting blood glucose tended to be lower in treated animals than vehicle controls, but after three days no difference was evident. Furthermore, we were unable to detect a change in insulin sensitivity after drug treatment during the ITT on day 3 (Kaiser, Warpman & James, unpublished). In addition, no changes in lysine acetylation of IRS-1 were observed, indicating that the lack of effect on insulin sensitivity could be due to the inability of TSA to work through the molecular mechanism of increasing IRS-1 acetylation. TSA is rapidly metabolised by liver cells in culture in two stages, initially by reduction to the imide followed by demethylation, leading to inactive metabolites . It is therefore probable that the compound was rapidly metabolised by hepatic phase I metabolic processes in these experiments so that it was unable to exert pharmacodynamic effects on the animals. The poor bioavailability of TSA  has led to its discontinuation as a clinical candidate for the treatment of human disease and the possibility of testing the insulin sensitizing effects of HDAC inhibition must await the availability of a drug with better pharmacokinetics. Furthermore, HDACs are not redundant, but have specific expression patterns and functions. Therefore, it is of great importance to develop specific HDAC-inhibitors to be able to assess their respective contributions to increases in insulin sensitivity in vivo.
In this study, we have identified a previously undescribed interaction between IRS-1 and HDAC2 in the cytosolic compartment of cells. The interaction is observed both in vitro and in vivo during conditions of compromised insulin signalling, as seen by reductions in insulin-stimulated IRS-1 tyrosine phosphorylation and PKB activation and increased phosphorylation of the negative regulatory phosphorylation site, serine 312. Our data indicate that it is the interaction with HDAC2 itself rather than its catalytic activity that is integral to the insulin unresponsiveness that ensues. Furthermore, our data show that IRS-1 is a lysine-acetylated protein, a previously unidentified post-translational modification of IRS-1, and that increases in the level lysine acetylation of IRS-1 result in improved insulin signal transduction. Increases in IRS-1 acetylation can be achieved pharmacologically (with TSA) or by ablation of HDAC2 specifically by use of RNAi. Out data therefore indicate that a new dimension to the physiology and pathophysiology of insulin sensitivity and insulin resistance involves changes in the degree of lysine acetylation of IRS-1 and that specific small molecule inhibitors of HDAC2 activity could represent novel therapeutics for the treatment of diseases that centre around insulin resistance, such as type 2 diabetes and obesity.
The CytoTrap™ (Stratagene) yeast two-hybrid system was used to discover protein-protein interactions in the cytoplasm of yeast cells. Interactions were detected by recruitment to the cell membrane of the human Sos (hSos) gene product, which activates the Ras pathway. The yeast strain used (cdc25H) harbours a temperature sensitive mutation in the cdc25 gene, the yeast homologue for hSos, which means that the cells can grow at 25°C but not at 37°C unless rescued with a protein-protein interaction. A human foetal brain plasmid cDNA library (Stratagene), harboured in the pMyr vector (with a myristylation signal to direct and anchor proteins in the membrane), was used as "prey" and the sub-cloned full length IRS-1 gene in the pSos vector was used as "bait". When prey and bait proteins interact the hSos is brought into close proximity to Ras and subsequently the yeast survive and are selected by growth at 37°C. The IRS-1/HDAC2 interaction rescued growth at 37°C in this way. The corresponding pMyr yeast plasmid was isolated and co-transformed with the pSos bait construct to perform false positive tests.
HDAC2 was full length cloned using RACE cDNA obtained from human heart tissue together with gene specific primers and the Advantage 2 polymerase mix (Clontech). With the purpose of mapping the interaction site of HDAC2 on IRS-1 we used the Matchmaker 3 yeast two-hybrid system (Clontech). This is a GAL4-based two-hybrid system that provides a transcriptional assay for detecting specific protein-protein interactions in yeast. Two nutritional markers and one enzymatic reporter gene were used to detect interactions. Different domains of IRS-1 (PH domain, residues 1–155, the PH-PTB domains, residues 1–578 and the PH-PTB-pre-C-terminal domains, residues 1–895, obtained by PCR) were sub-cloned into a "bait" vector (pGBKT7), fused to the DNA-binding domain of GAL4. Full length HDAC2 was sub-cloned into the "prey" vector (pGADT7), fused to the activation domain of GAL4. Cell growth on medium lacking the two nutritional markers was used as a readout of the interaction between the predator and prey.
In order to confirm the IRS-1/HDAC2 interaction in vitro, we used a coupled transcription/translation system (Promega) comprising a rabbit reticulocyte lysate solution with RNA polymerase, nucleotides, salts, a ribonucleoside inhibitor, and [35S]-methionine (Amersham Biosciences) to allow detection of translated proteins. Since the prey vector pMyr already contains a T7 promoter, this was used directly in the system. However, the bait vector pSos lacks a T7 promoter and thus the IRS-1 gene was subcloned into a T7-containing vector (pGBKT7; Clontech) to permit transcription. The individually transcribed and translated proteins were mixed and co-immunoprecipitated with anti-IRS-1 antibodies (Upstate Biotechnologies) and subsequently analysed by polyacrylamide gel electrophoresis (4–12%). The gel was dried analysed by phosphorimagery.
The human breast adenocarcinoma cell line MCF-7 was cultured in a mixed medium of Dulbecco's Modified Eagle Medium with nutrient mixture F12 (Invitrogen) lacking phenol red with 10% Foetal Bovine Serum (Gibco). At near confluency, cells were starved of serum for 16 h and subsequently treated with IGF-1, insulin, PMA (phorbol myristic acid; Sigma) or TSA (Trichostatin A; Sigma), or combinations thereof, for different lengths of time as indicated in individual figures. Cells were harvested in hypotonic cell lysis buffer comprising 20 mM Hepes, pH 7.6, 20% glycerol, 10 mM NaCl, 1.5 mM MgCl2, 0.2 mM EDTA, 0.1% NP40, 25 mM NaF, 25 mM β-glycerophosphate, 1 mM DTT, 1 mM Na-orthovanadate and protease inhibitors.
Cell lysates were cleared by centrifugation at 16000 g for 10 min at 4°C, and protein content was determined using the Bradford method (BioRad). For immunoprecipitations, matched amounts of protein were incubated with primary antibody (amount used as recommended by the manufacturer or empirically determined) for 2 h at 4°C followed by addition of 20μl of protein A/G agarose suspension (Santa Cruz) for 1 h at 4°C with rotating tube. After washing (3 times with high salt (500 mM NaCl) and twice with isotonic buffer), beads were heated with SDS-PAGE sample buffer for 10 minutes at 70°C and proteins were resolved by 4–12% gradient SDS-PAGE. After blotting, membranes were blocked in 5% non-fat dried milk in Tris-buffered saline containing 0.1% Tween 20 for 1 h prior to addition of the primary antibody. After incubation with secondary horseradish peroxidase-conjugated antibody, protein bands were visualised using enhanced chemiluminescence (ECL-plus detection kit, Amersham Biosciences).
Antibodies used were anti-IRS-1 (Upstate, cat. no 06–248); anti-HDAC2 (Santa Cruz, cat. no. sc-9959 and sc-6296); anti-phosphotyrosine (Santa Cruz, cat. no. sc-7020); anti-acetyl lysine (Cell Signalling, cat. no. 9681); anti-ubiquitin (Santa Cruz, cat. no. sc-6085 and sc-9133); anti-phospho-serine 307 IRS-1 (Upstate, cat. no. 07–247), HRP-conjugated anti-mouse IgG (Amersham Biosciences, cat. no. NA931V); HRP-conjugated anti-goat IgG (Dako cat. no. PO449) and HRP-conjugated anti-rabbit IgG (Upstate, cat. no. 12–348).
Double stranded RNA duplexes corresponding to amino acids from the C-terminal part of human HDAC2 (5'CAGCUCAGCAACCCCUGAAtt3') were annealed and transfected into human MCF-7 cells (Lipofectamine 2000 from Invitrogen was used as transfection agent): The effect of RNAi on HDAC2 expression and on insulin dependent IRS-1 tyrosine phosphorylation was measured after 48 hours. A second oligonucleotide (5'GGAGCAAAGAAAGCUAGAAtt3') was found to be non-silencing at a dose of 80 pmol, in contrast to the silencing oligonucleotide above, and was used in control experiments showing that no effect on IRS-1 phosphorylation or acetylation was observed (data not shown).
Male 8-week old ob/ob mice were obtained from Bomhultsgard, Denmark and housed according to standard procedures. C57/bl6 genotype control mice were obtained from Scanbur BK AB (Sollentuna, Sweden). PTP1B knockout animals on a balb/cJJ background were purchased from McGill University, Montreal, Canada. Balb/cJJ genotype controls were obtained from Scanbur BK AB. In our hands, balb/cJJ mice are generally a healthy mouse strain that breeds well. In side-by-side experiments, the mice are more insulin sensitive than C57/bl6 mice whilst being less insulin sensitive than the PTP1B knockout animals on the same genetic background. The animals are somewhat smaller than C57/bl6 mice and have a relatively high body fat content.
For compound treatment experiments and insulin tolerance tests, animals were divided into two groups: vehicle (1% DMSO sub-cutaneous injection (s.c.), n = 15 and TSA 0.1 mg/kg s.c., n = 15) and subsequently treated s.c. at 09.00 h for three days. At the same time as injection, all food was withdrawn from the animals. Four hours later, blood was collected from the tail vein for blood glucose and plasma insulin analysis. On the third day, an insulin tolerance test (ITT) was performed 4 h after administration of the drug. ITT: insulin (actrapid 0.5 U/kg) was administered i.p. Following insulin administration, blood samples were collected after 15, 30, 60, 120 and 180 min from the tail vein for glucose analysis. Animals were then sacrificed and livers were dissected and immediately frozen in liquid nitrogen and stored at -70°C. All experiments were performed in accordance with permission from the local Swedish ethics committee and the company Pharmacology ethics review team.
For western blotting of liver proteins, frozen liver was powdered finely under liquid nitrogen using a pestle and mortar pre-cooled to -70°C. Powdered liver (1 g) was homogenized at 4°C using a Polytron in 3 ml of homogenisation buffer (4 mM EDTA, 50 mM NaF pH 8.0, 1 mM Na-orthovanadate, 1μM okadaic acid, 0.1% (v/v) 2-mercaptoethanol, with protease inhibitor cocktail). The homogenates were centrifuged at 13000 × g for 10 minutes at 4°C and the supernatant removed and used immediately for Western blot analysis or snap frozen in aliquots at -70°C until needed.
desulfo coenzyme A
histone acetyl transferase
insulin-like growth factor
insulin receptor substrate
protein kinase B
Phorbol myristic acid
phosphotyrosine binding domain
We are grateful to Ullrika Warpman for performing the studies examining the effects of TSA in vivo and to Ing-Marie Renström for performing glucose uptake analyses.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.