PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
 
PLoS One. 2010; 5(11): e15382.
Published online 2010 November 8. doi:  10.1371/journal.pone.0015382
PMCID: PMC2975661

Dioxin Toxicity In Vivo Results from an Increase in the Dioxin-Independent Transcriptional Activity of the Aryl Hydrocarbon Receptor

Steven R. Ellis, Editor

Abstract

The Aryl hydrocarbon receptor (Ahr) is the nuclear receptor mediating the toxicity of dioxins -widespread and persistent pollutants whose toxic effects include tumor promotion, teratogenesis, wasting syndrome and chloracne. Elimination of Ahr in mice eliminates dioxin toxicity but also produces adverse effects, some seemingly unrelated to dioxin. Thus the relationship between the toxic and dioxin-independent functions of Ahr is not clear, which hampers understanding and treatment of dioxin toxicity. Here we develop a Drosophila model to show that dioxin actually increases the in vivo dioxin-independent activity of Ahr. This hyperactivation resembles the effects caused by an increase in the amount of its dimerisation partner Ahr nuclear translocator (Arnt) and entails an increased transcriptional potency of Ahr, in addition to the previously described effect on nuclear translocation. Thus the two apparently different functions of Ahr, dioxin-mediated and dioxin-independent, are in fact two different levels (hyperactivated and basal, respectively) of a single function.

Introduction

The Aryl hydrocarbon receptor (Ahr) is the key component in the metabolic response to the extremely widespread, toxic and persistent pollutants dioxins [1], [2], [3], [4]. Dioxins produce epidermal and hepatic toxicity, teratogenesis, autoimmunity and carcinogenesis [5], [6]. Ahr is a cytoplasmic bHLH-PAS transcription factor that, upon binding of dioxin, translocates to the nucleus where it forms a complex with the Aryl hydrocarbon receptor nuclear translocator (Arnt), another bHLH-PAS protein [7], and binds to an eight-nucleotide motif (Xenobiotic Response Element; XRE) to control the expression of specific target genes [8], [9] (Figure 1A). Previous work with Ahr knock-out mice revealed the existence of dioxin-independent functions of Ahr, including developmental functions and ageing-related detoxification [1], [2], [3]. However, the relationship between these two functions of Ahr, toxic and dioxin-independent, remains unclear [5], [10], in particular whether dioxin toxicity entails loss or excess of the dioxin-independent Ahr function, or a new independent function. We have tackled this question, which is central to the treatment of the effects of dioxins, using a Drosophila model for Ahr function and dioxin toxicity. Drosophila, like other invertebrates, does not suffer dioxin toxicity because its Ahr homologue does not bind dioxins [11], [12], but offers an extensive suite of genetic techniques. Hence, we have introduced the mouse dioxin receptor in flies, and studied its activity in vivo in the presence or absence of dioxin. A similar strategy (‘humanising’ or introduction of human genes in animal model systems) has allowed the study of the molecular and genetic basis of other human medical conditions in Drosophila (i.e. Alzheimer's, Parkinson's and Huntington's diseases) [13].

Figure 1
Ahr mediates the cellular response to dioxin.

Results and Discussion

Ahr is able to fulfil dioxin-independent functions in Drosophila

Mouse Ahr and its Drosophila homologous protein, Spineless (Ss), are highly similar in the bHLH and PAS A domains (Figure 1B). The spineless (ss) gene has well-characterised dioxin-independent functions during development [14], [15]. For instance, ss expression controls leg development by repressing the dachshund (dac) and Bar genes [16], [17] (Figure 2A, ,3C).3C). Ss protein locates to the nucleus [18] (Figure 2B) and its presence promotes the nuclear localisation of the fly homologue of Arnt, Tango (Tgo) which is otherwise distributed evenly across the cell [19] (Figure 2B). Ss::Tgo heterodimers target gene expression (both activation and repression) through the XRE motif both in vitro and during development [16], [19]. In Drosophila, the development of the legs and the eyes are sensitive to the precise dosage of Ss and Tgo [14], [15], [19] (see below), and thus provide an accurate read-out of ss-related activity and function.

Figure 2
Ahr fulfils ss functions in the absence of dioxin.
Figure 3
Ahr activity is enhanced by dioxin or addition of Arnt or extra Tgo.

Ss is not able to bind dioxins [11] probably because its PAS-B domain, which contains the dioxin binding domain [20], is highly divergent from vertebrates (Figure 1B). For this reason, we introduced mouse Ahr under an inducible promoter (UAS-Ahr or a tagged UAS-Ahr-GFP) in transgenic flies, and expressed it in the domain of ss using the rnGal4 driver, which mimics ss expression [17], [21]. Once in flies, and in the absence of dioxin, Ahr is able to fulfil the dioxin-independent functions of Ss. Thus Ahr: 1) rescues almost completely the loss of the ss gene during development (Figure 2A, ,3E);3E); 2) localises to the nucleus (Figure 2B); and 3) binds Tgo and promotes its nuclear localisation (Figure 2B and C). We conclude that the dioxin-independent activities of Ahr and Ss are functionally equivalent, and mediated through binding to Tgo/Arnt.

Fruit flies expressing the murine Ahr respond to dioxins

We next explored the effects of adding dioxin on the activity of both Ss and Ahr, in order to discriminate whether dioxin induces in vivo either a reduction or an increase in the activity of Ahr. Exposure of wild-type flies to dioxin produces no effects (Figure 3A), as expected from the inability of Ss to bind dioxin [11], [19]. However, an excess of ss function can be generated in the absence of dioxin by ectopic overexpression of the UAS-ss transgene with Gal4 drivers. This ectopic overexpression interferes with normal development and produces leg deformities, tarsal segment loss and roughened eyes (Figure 3A, ,4A),4A), and ectopic repression of ss's developmental targets Bar and dac [16], [17] (Figure 3B, C). Altogether these results show that the Ss protein acts both endogenously and ectopically in a dioxin-independent manner. In contrast, dioxin produces a marked increase on the activity of Ahr. Ectopic overexpression of UAS-Ahr or UAS-Ahr-GFP (both driven across the leg imaginal disc by dpp-Gal4) produces no obvious effect on either fly morphology or expression of Bar and dac (Figure 3A, B). However, when these flies overexpressing UAS-Ahr are exposed to dioxin, the effects mimic those obtained with UAS-ss: abnormal legs and eyes, and ectopic repression of dac and Bar (Figure 3A–D, 4A–B). Moreover, the presence of dioxin also allows more complete rescue of the loss of ss by Ahr than in its absence (Figure 2A, 3A, 3E, Table S1), a rescue now comparable to that exerted by the UAS-ss transgene itself. Altogether our results suggest that in vivo dioxin produces a hyperactivation of the Ahr protein. Thus, Ahr shows two levels of activity in the context of the whole organism: 1) basal low activity without dioxin, allowing only partial rescue of the endogenous ss gene function, and 2) high activity in the presence of dioxin, allowing Ahr to mimic the full effects of Ss, both endogenous and ectopic.

Figure 4
Hyperactivated Ahr produces abnormal eye development.

Dioxin enhances the transcriptional potency of Ahr::Tgo complexes

In vertebrates, binding of dioxin to Ahr triggers its translocation from the cytoplasm to the nucleus, followed by binding to nuclear Arnt and subsequently to XRE sequences [8] (Figure 1A). However, in Drosophila, the nuclear localisation of Ahr and Ss does not depend on dioxin (Figure 2B), probably due to differences in the proteins that retain Ahr in the vertebrate cytoplasm [22], [23]. Thus, our system suggests that nuclear translocation is not the only limiting step in Ahr function and that the observed hyperactivation of Ahr by dioxin must rely on another mechanism. In genetic experiments, the function of Ss is very sensitive to the dosage of Tgo. Removal of a copy of tgo enhances the loss of ss [19] and, reciprocally, the phenotype caused by ectopic Ss in the eyes can be suppressed almost completely by reducing the dosage of Tgo (unpublished observations). Correspondingly, an increase of Tgo enhances the function of Ahr, in fact, increasing the dosage of Tgo or Arnt mimics Ahr's hyperactivation by dioxin. Thus, flies overexpressing Ahr plus Tgo or Arnt display the same phenotypes as those overexpressing either Ss, or Ahr plus dioxin (Figure 3A, 3E, ,4).4). This result is particularly noteworthy since overexpression of Tgo, Arnt or Ahr on their own produces no discernible alteration of eye or leg development (Figure 3 and S1). The similarity between the effects of dioxin and extra amounts of Tgo or Arnt indicates that firstly, our observed hyperactivation of Ahr by dioxin is bona fide, and secondly, that the endogenous amount of Tgo protein is limiting and dioxin allows Ahr to use Tgo more efficiently in the process involved. One possibility is that dioxin increases or stabilises the Ahr::Tgo complexes, but we do not observe an increase in binding between Ahr and Tgo or Arnt in Co-immunoprecipitation experiments in the presence of dioxin (Figure 5A, B). A second possibility is that dioxin increases the potency of the Ahr::Tgo complexes, in particular their transcriptional efficiency. To test this, we used a reporter gene containing the XRE enhancer sequence from the Bar gene that Ss::Tgo complexes bind to [16] (Figure 5D). Our results show that in the absence of dioxin Ahr is not able to repress this reporter (Figure 5E), despite forming complexes with Tgo as efficiently as Ss (Figure 1C, 5A, 5C). On the contrary, and as expected, transcriptional repression is observed in the presence of dioxin (Figure 5E).

Figure 5
Dioxin enhances Ahr transcriptional potency.

In summary, our results show that dioxin increases in vivo the dioxin-independent activity of Ahr, and hence that dioxin toxicity is best understood as an excess of function of Ahr. The two postulated functions of Ahr (dioxin-dependent and dioxin-independent) in fact seem to correspond to two different levels of a single Ahr activity, basal and hyperactivated. This model envisages that hyperactivated Ahr will not only regulate more strongly the same genes as basal Ahr, but also that some genes could only be bound and regulated by hyperactive Ahr or Ss proteins [24]. In either case, in absence of dioxin, Ahr would not be able to disrupt normal development or fully rescue ss phenotypes as Ss and hyperactive Ahr do. Furthermore, our results also suggest that the dioxin-mediated toxicity of Ahr is not solely due to inducing its nuclear translocation or to changes in affinity for its dimerisation partners, but also entails an increase of the transcriptional regulatory activity of Ahr (Figure 5F). More generally, our results corroborate that reduction, but not elimination, of endogenous Ahr and Arnt function may be the best approach to treat dioxin toxicity, and introduce Drosophila as a new system to study in vivo the genetic and molecular basis of such toxicity.

Materials and Methods

Reagents and fly strains

UAS-Ahr, UAS-Ahr-GFP, UAS-ss, UAS-ss-GFP and UAS-Arnt transgenic lines were generated by cloning the appropriate full cDNAs [14], [25] into a pUASt vector, followed by injection of purified plasmid into embryos by Vanedis and recovery of w+ lines. The ss-GFP cDNA was generated by cloning the GFP coding sequence 3′ and in frame with the ss coding sequence. rn-Gal4 [17], [21], ss mutants [14], and UAS-tgo [19] have been previously described. dpp-Gal4, hs-Gal4, GMR-Gal4 and AB1-Gal4 were obtained from Bloomington Stock Center.

The dioxin 2,3,5,7-tetrachlorodibenzo-p-dioxin (TCDD) was fed to parental flies and larval offspring, diluted in the food at a final concentration of 200 ngr TCDD/gr food [26]. To minimize exposure to dioxin-contaminated food and fly particles, cultures were left for nine days after first eclosions, then filled with SH medium (ethanol 100%[ratio]glycerol, 3[ratio]1; v/v), and the whole contents sieved for fly corpses and pupae, from which legs were dissected.

Anti-Tgo and anti-Dac were obtained from the Developmental Studies Hybridoma Bank (DSHB). Anti-Arnt, anti-GFP and anti- β-galactosidase were from Affinity BioReagents, Roche and Cappel, respectively. Anti-Bar and anti-Ss were provided by T. Kojima [16] and Yuh-Nung Jan [18], respectively. Leg discs and salivary glands were stained as described in the online methods.

Immunohistochemistry

Third instar larvae were collected and dissected in cold PBS and fixed at RT for 10 min. in 4% paraformaldehyde. Larvae were washed in PBTx (0.03% Triton X-100 in PBS) blocked in PBTA (0.4% Bovine Albumin in PBTx). Samples were incubated overnight at 4°C in the primary antibody and in the secondary for 2hr at RT. Samples were mounted in Vectashield. Fluorescence pictures were taken in a Zeiss Axiovert confocal microscope equipped with a LSM520 Meta.

Cuticle and eye preparations

Pharates and eclosed adults were kept in SH medium (ethanol 100%[ratio]glycerol, 3[ratio]1; v/v) and dissected in 70% ethanol. Samples were incubated for 5 minutes at 60°C in NaOH 1M, and then cooked in Hoyer's medium (Hoyers: lactic acid, 1[ratio]1; v/v) for 20 min at 60°C. Finally the legs were mounted in Hoyer's medium. Adult cuticle preparations were observed under a Leica DMRB optical microscope, and the pictures were taken with one of the following softwares: Q-win or Simple PCI 6.6.

Adult flies were dehydrated by washing for 10 min in 70%, 80%, 90%, and 100% ethanol, and finally washed twice for 10 min in hexamethyldisilazane. Then samples were allowed to dry before coating them with a layer of gold/palladium alloy. Eyes were scanned in a Leo420 Steroscan SEM.

Estimation of bristle size

SEM pictures taken at 1000× were analyzed by the imaging program Image J. Sections of the 500×200 pixels from the dorsal part of the eye were cropped. The resulting image was binarized using an appropriate threshold of brightness to isolate the bristles from the rest of the ommatidia. The background was removed by eroding once. The effect of the erosion on bristle size was corrected by dilating once. The size of each bristle was estimated by the number of covered pixels. The output data below 80 pixels or above 300 pixels were discarded to remove any artefacts caused by the background (Figure S1B). These limits were determined empirically. Between five and ten eyes were measured per sample.

Cloning

ss coding sequence was extracted from pGEX-2T-ssA1 (provided by Dr. I. Duncan) by PCR (forward primer GAGAGATCTAGAATCCGCCCTAGCAATGAGCCA, and reverse primer CCGGGAGCTGCATGTGTCAGAGG). An XbaI restriction site was inserted upstream of ss in a tail in the forward primer. ss was then sub-cloned in an XbaI restriction site in pBK-eGFP, leaving the eGFP ORF in frame with the ss N-terminal side. eGFP-ss was then cloned in the pUASt vector into the NotI and SpeI restriction sites.

Ahr and Ahr-eGFP coding sequences were extracted from the construct pEGFP-Ahr (provided by Dr. P. Fernandez-Salguero) and cloned into XbaI in pUASt.

Arnt cDNA was extracted from pCMV-sport6-Arnt (obtained from Gene Service), identified as clone AV-21A3. Arnt was cloned in pUASt in the EcoRI and XhoI restriction sites.

Co-immunoprecipitation assays

For in vivo assays larvae hs-Gal4 UAS-Ahr-GFP, hs-Gal4 UAS-ss-GFP and hs-Gal4 UAS-GFP were heat shocked for 2 hours at 37°C and lysated. Mouse anti-GFP bound to Protein-G Sepharose was used to precipitate Ahr-GFP, Ss-GFP and GFP. Presence of Tgo and GFP proteins was detected by Western-blotting. For in vitro assays Ahr, Ss, Tgo and Arnt were expressed in rabbit reticulocyte lysates (Promega) and labelled with 35S as required. Binding was performed in HEPES 20 mM pH 7.9, NaCl 50 mM, 1% Np40. Anti-Tgo and anti-Arnt bound to Protein-G Sepharose were used to precipitate either Ahr-35S or Ss-35S. When described, dioxin was added during the binding reaction and washes. The range of concentrations was choosen following published data [27]. Presence of 35S-labelled proteins was detected with a Typhoon TRIO scanner.

Supporting Information

Table S1

Quantification of leg phenotypes presented in Fig. 3E. Genotypes: ss mutant (w; CyO/+; rn-Gal4 ssabr/sssta), ss rescue (w; UAS-ss/+; rn-Gal4 ssabr/sssta), Ahr rescue (w; UAS-Ahr/+; rn-Gal4 ssabr/sssta), and Ahr + dioxin rescue (w; UAS-ss/+; rn-Gal4 ssabr/sssta+200 ngr TCDD/gr. food). Total number of legs and total number of flies (in brackets) are indicated. A number of legs is lost during dissection or due to sticking to the food or the pupal cases (see material and methods). Legs were classified according to the number of tarsi present (from one to five). The number of ss mutant flies is higher as they were obtained as internal control segregates in every experiment.

(DOC)

Figure S1

Ectopic expression of either Tgo or Arnt does not affect leg development. (A) First thoracic leg of a male (top panels) and expression of Dac (green) and Bar (red) in leg imaginal disc during third instar (bottom panels). (B) SEM pictures of adult eyes taken at 500× (top row) and 1000× (bottom row). Genotypes are indicated above panels. UAS-Tgo does not affect the wild-type eye phenotype. The binarization and measurement process is shown.

(TIF)

Acknowledgments

We thank Jose I. Pueyo for unpublished results and comments to the manuscript. Rose Phillips and Sarah Bishop for technical assistance, Pedro Fernandez-Salguero, Tetsuya Kojima, Yuh-Nung Jan, Ian and Dianne Duncan and Steve Crews for molecular reagents and Drosophila strains, and Claudio Alonso and Rob Ray for comments to the manuscript.

Footnotes

Competing Interests: The authors have declared that no competing interests exist.

Funding: This work was funded by a Wellcome Trust Senior fellowship to JPC (ref. 087516) and a University of Sussex GTA studentship to MAC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

1. Schmidt JV, Bradfield CA. AH receptor signaling pathways. Annu Rev Cell Dev Biol. 1996;12:55–89. [PubMed]
2. Fernandez-Salguero P, Pineau T, Hilbert DM, McPhail T, Lee SS, et al. Immune system impairment and hepatic fibrosis in mice lacking the dioxin-binding Ah receptor. Science. 1995;268:722–726. [PubMed]
3. Mimura J, Yamashita K, Nakamura K, Morita M, Takagi TN, et al. Loss of teratogenic response to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in mice lacking the Ah (dioxin) receptor. Genes Cells. 1997;2:645–654. [PubMed]
4. Meharg AA, Killham K. Environment: A pre-industrial source of dioxins and furans. Nature. 2003;421:909–910. [PubMed]
5. Bock KW, Kohle C. Ah receptor: Dioxin-mediated toxic responses as hints to deregulated physiologic functions. Biochemical Pharmacology. 2006;72:393–404. [PubMed]
6. Steenland K, Bertazzi P, Baccarelli A, Kogevinas M. Dioxin revisited: developments since the 1997 IARC classification of dioxin as a human carcinogen. Environ Health Perspect. 2004;112:1265–1268. [PMC free article] [PubMed]
7. Hoffman EC, Reyes H, Chu FF, Sander F, Conley LH, et al. Cloning of a factor required for activity of the Ah (dioxin) receptor. Science. 1991;252:954–958. [PubMed]
8. Swanson HI, Chan WK, Bradfield CA. DNA-Binding Specificities and Pairing Rules of the Ah Receptor, Arnt, and Sim Proteins. Journal of Biological Chemistry. 1995;270:26292–26302. [PubMed]
9. Whitlock JP., Jr Genetic and molecular aspects of 2,3,7,8-tetrachlorodibenzo-p-dioxin action. Annu Rev Pharmacol Toxicol. 1990;30:251–277. [PubMed]
10. McMillan BJ, Bradfield CA. The aryl hydrocarbon receptor sans xenobiotics: endogenous function in genetic model systems. Mol Pharmacol. 2007;72:487–498. [PubMed]
11. Butler RA, Kelley ML, Powell WH, Hahn ME, Van Beneden RJ. An aryl hydrocarbon receptor (AHR) homologue from the soft-shell clam, Mya arenaria: evidence that invertebrate AHR homologues lack 2,3,7,8-tetrachlorodibenzo-p-dioxin and beta-naphthoflavone binding. Gene. 2001;278:223–234. [PubMed]
12. Hahn M. Aryl hydrocabon receptors: diversity and evolution. Chemico-Biological interactions. 2002;141:131–160. [PubMed]
13. Fortini ME, Bonini NM. Modeling human neurodegenerative diseases in Drosophila: on a wing and a prayer. Trends Genet. 2000;16:161–167. [PubMed]
14. Duncan DM, Burgess EA, Duncan I. Control of distal antennal identity and tarsal development in Drosophila by spineless-aristapedia, a homolog of the mammalian dioxin receptor. Genes Dev. 1998;12:1290–1303. [PubMed]
15. Wernet MF, Mazzoni EO, Celik A, Duncan DM, Duncan I, et al. Stochastic spineless expression creates the retinal mosaic for colour vision. Nature. 2006;440:174–180. [PubMed]
16. Kozu S, Tajiri R, Tsuji T, Michiue T, Saigo K, et al. Temporal regulation of late expression of Bar homeobox genes during Drosophila leg development by Spineless, a homolog of the mammalian dioxin receptor. Developmental Biology. 2006;294:497–508. [PubMed]
17. Pueyo JI, Couso JP. The 11-aminoacid long Tarsal-less peptides trigger a cell signal in Drosophila leg development. Developmental Biology. 2008;324:192–201. [PubMed]
18. Kim MD, Jan LY, Jan YN. The bHLH-PAS protein Spineless is necessary for the diversification of dendrite morphology of Drosophila dendritic arborization neurons. Genes Dev. 2006;20:2806–2819. [PubMed]
19. Emmons RB, Duncan D, Estes PA, Kiefel P, Mosher JT, et al. The spineless-aristapedia and tango bHLH-PAS proteins interact to control antennal and tarsal development in Drosophila. Development. 1999;126:3937–3945. [PubMed]
20. Coumailleau P, Poellinger L, Gustafsson JA, Whitelaw ML. Definition of a minimal domain of the dioxin receptor that is associated with Hsp90 and maintains wild type ligand binding affinity and specificity. J Biol Chem. 1995;270:25291–25300. [PubMed]
21. St Pierre SE, Galindo MI, Couso JP, Thor S. Control of Drosophila imaginal disc development by rotund and roughened eye: differentialy expressed transcripts of the same gene encoding functionally distinct zinc finger proteins. Development. 2002;129:1273–1281. [PubMed]
22. Bell DR, Poland A. Binding of Aryl Hydrocarbon Receptor (AhR) to AhR-interacting Protein. Journal of Biological Chemistry. 2000;275:36407–36414. [PubMed]
23. Qin H, Zhai Z, Powell-Coffman JA. The Caenorhabditis elegans AHR-1 transcription complex controls expression of soluble guanylate cyclase genes in the URX neurons and regulates aggregation behavior. Developmental Biology. 2006;298:606–615. [PubMed]
24. Tijet N, Boutros PC, Moffat ID, Okey AB, Tuomisto J, et al. Aryl hydrocarbon receptor regulates distinct dioxin-dependent and dioxin-independent gene batteries. Mol Pharmacol. 2006;69:140–153. [PubMed]
25. Santiago-Josefat B, Pozo-Guisado E, Mulero-Navarro S, Fernandez-Salguero PM. Proteasome inhibition induces nuclear translocation and transcriptional activation of the dioxin receptor in mouse embryo primary fibroblasts in the absence of xenobiotics. Mol Cell Biol. 2001;21:1700–1709. [PMC free article] [PubMed]
26. Fernandez-Salguero PM, Hilbert DM, Rudikoff S, Ward JM, Gonzalez FJ. Aryl-hydrocarbon receptor-deficient mice are resistant to 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced toxicity. Toxicol Appl Pharmacol. 1996;140:173–179. [PubMed]
27. Chan WK, Yao G, Gu YZ, Bradfield CA. Cross-talk between the aryl hydrocarbon receptor and hypoxia inducible factor signaling pathways - Demonstration of competition and compensation. Journal of Biological Chemistry. 1999;274:12115–12123. [PubMed]
28. Martinez AM, Schuettengruber B, Sakr S, Janic A, Gonzalez C, et al. Polyhomeotic has a tumor suppressor activity mediated by repression of Notch signaling. Nature Genetics. 2009;41:1076–U1047. [PubMed]
29. Ikuta T, Eguchi H, Tachibana T, Yoneda Y, Kawajiri K. Nuclear localization and export signals of the human aryl hydrocarbon receptor. J Biol Chem. 1998;273:2895–2904. [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science