Email updates

Keep up to date with the latest news and content from Genome Biology and BioMed Central.

Highly Accessed Minireview

Integration of splicing, transport and translation to achieve mRNA quality control by the nonsense-mediated decay pathway

Thomas Schell1, Andreas E Kulozik23 and Matthias W Hentze13*

Author Affiliations

1 EMBL Heidelberg, Gene Expression Programme, Meyerhofstrasse 1, 69117 Heidelberg, Germany

2 Department of Pediatric Oncology, Hematology and Immunology, University of Heidelberg, Im Neuenheimer Feld 150, 69120 Heidelberg, Germany

3 Molecular Medicine Partnership Unit, EMBL/University of Heidelberg, Im Neuenheimer Feld 156, 69120 Heidelberg, Germany

For all author emails, please log on.

Genome Biology 2002, 3:reviews1006-reviews1006.6  doi:10.1186/gb-2002-3-3-reviews1006

The electronic version of this article is the complete one and can be found online at: http://genomebiology.com/2002/3/3/reviews/1006


Published:26 February 2002

© 2002 BioMed Central Ltd

Abstract

When pre-mRNAs are spliced, a multi-component complex is deposited onto them, close to the sites of intron removal. New findings suggest that these exon-exon junction complexes and the complexes that bind mRNA caps are key effectors of the fate of spliced mRNAs and may regulate whether mRNAs containing premature stop codons are degraded.

Minireview

Nonsense-mediated mRNA decay (NMD) is a specific RNA-degradation process by which eukaryotic cells can eliminate mRNAs that contain premature translation-termination codons (PTCs). Such mRNAs encode truncated polypeptides that could, when translated, exert dominant-negative effects; it is therefore advantageous to cells to minimize translation of these mRNAs. NMD has been found in yeast, plants, nematodes, flies and vertebrates [1].

There are several examples that illustrate the medical importance of NMD. Most of the mutations that introduce a PTC into the human β-globin mRNA cause a form of β-thalassemia that is recessive; heterozygous carriers of these mutations are thought to be protected by degradation of the mutant mRNAs via NMD. A mutation that escapes NMD is associated with dominant β-thalassemia, however [2]. Similarly, certain PTCs in the human ROR2 mRNA, which encodes an orphan receptor tyrosine kinase, fail to trigger NMD and cause a dominantly inherited form of brachydactyly type B [3]. Finally, in B and T lymphocytes, the somatic mutation and recombination events of the immune system frequently lead to the production of PTC-containing mRNAs. The NMD pathway is thought to prevent the expression of truncated immunoglobulin and T-cell-receptor polypeptides that would arise from such aberrant mRNAs [4].

Three proteins have been shown to be directly and specifically involved in NMD in Saccharomyces cerevisiae: Upf1p/Nam7p, Upf2p/Nmd2p and Upf3p (reviewed in [1,5]). Functional homologs of Upf1p, Upf2p and Upf3p have been described in humans and in other organisms, and the three Upf proteins can interact with each other [6,7,8,9,10]. Each of the mammalian Upf proteins has a characteristic subcellular localization. Human Upf1p (hUpf1p) is cytoplasmic and associated with polysomes [11]; most hUpf2p localizes to the nuclear periphery; and the two human Upf3p homologs (encoded by two genes, hUpf3a and hUpf3b) are shuttling proteins that are found predominantly in the nucleus [8,9,10]. The interaction of Upf1p with the translation-termination factors eRF1 and eRF3 [12] links NMD biochemically to the termination of translation.

A central question about the mechanism of NMD is how the cell distinguishes a premature from a normal stop codon (Figure 1). Remarkably, the position of introns seems to play a critical role in metazoans: a PTC is functionally defined as an in-frame stop codon that is located at least 50 nucleotides upstream of the last exon-exon junction in a spliced mRNA (see Figure 1; reviewed in [13,14]). Consequently, naturally intronless PTC-containing mRNAs are not usually destroyed by NMD [15,16], and transfection of PTC-containing cDNAs of human β globin yields 'NMD immune' mRNAs [17]. Furthermore, insertion of an intron more than 50 nucleotides downstream of a normal stop codon triggers NMD of an mRNA that in fact has an intact open reading frame [18,19]. As might be expected, it is unusual to find an intron in the 3' untranslated region of an mRNA. When such introns do occur, they are almost always positioned less than 50 nucleotides downstream of the stop codon [20]. These observations collectively support a model in which splicing deposits a 'mark' at or near exon-exon junctions and thus tags the position of excised introns for later 'inspection' by the translation machinery. In yeast, in which only few genes have introns, a downstream sequence element seems to serve as such a positional mark [5,13,21]. When the mRNA first passes through the ribosome, NMD is thought to be triggered when the mark is encountered downstream from the position of translation termination [13,14,21] (Figure 1).

thumbnailFigure 1. Specification of premature (versus normal) stop codons in mammalian cells. (a) In the nucleus, a normal intron-containing transcript is subjected to multiple modifications, such as splicing, addition of a 7mGpppN cap and polyadenylation, to yield a mature mRNA. This mRNA can be translated after export to the cytoplasm. The exon-exon junction complex (EJC) is associated with the mRNA just upstream of each splice site. (b) If an mRNA carries a premature termination codon (PTC), however, it can be degraded by the NMD system. Note that the physiological stop codon has no EJC positioned downstream of it, whereas NMD-sensitive PTCs are usually followed by at least one EJC (curved arrows in (a,b)). The EJC is therefore proposed to play an important role in mRNA surveillance.

Beyond a model: a multi-component exon-exon junction complex

Recent work has identified a multi-subunit complex that could serve as a mark of intron removal (Figure 2). This exon-junction complex (EJC) consists of at least five proteins: the splicing-associated factors SRm160, DEK and RNPS1, the mRNA export factor REF/Aly and the mRNA-shuttling protein Y14. It is deposited approximately 20 nucleotides upstream of exon-exon junctions [22] and at least part of it is exported together with the spliced RNA to the cytoplasm (see below). Using a shortened β-globin mRNA with a first exon of either 38 or 17 nucleotides ('β/38' or 'β/17', respectively) as a reporter, Le Hir et al. [23] showed that the β/38 mRNA but not the β/17 bound the EJC after splicing. Deposition of the EJC onto the mRNA was associated with a substantial increase in RNA export from the nucleus. This increase can be explained by the ability of the EJC to interact with the export factors REF/Aly and TAP/p15, which in turn interact with the nuclear pore complex (reviewed in [24]). As the EJC travels from the nucleus to the cytoplasm, its composition changes (Figure 2): only antibodies against SRm160 (weakly) and Y14 (strongly) co-immunoprecipitated spliced β/38 mRNA from the cytoplasm, suggesting that DEK, RNPS1 and REF/Aly dissociate from the EJC before or soon after export [23].

thumbnailFigure 2. A model of the changes in PTC-containing mRNA-protein complexes (mRNPs) during nuclear export. A PTC-containing transcript, depicted as in Figure 1, is capped, spliced and polyadenylated. Concomitantly, the cap-binding complex (CBC, consisting of CBP80 and CBP20), the exon junction complex (EJC, which consists of the DEK, REF/Aly, RNPS1, SRm160 and Y14 proteins), the NMD protein hUpf3 and the export factor TAP/p15 are all added, to give a fully processed mRNP. Before or soon after export from the nucleus through the nuclear pore complex (NPC), DEK, RNPS1 and REF/Aly dissociate from the EJC and hUpf2 binds. It is not clear whether the first round of translation can be mediated by CBC (left) or whether CBC is first exchanged for eIF4E. After this first round, the PTC-containing mRNA undergoes nonsense-mediated decay. See text for further details. Figure modified from [23,26].

Does the exon-junction complex play a role in NMD?

Although it cannot yet be determined conclusively whether the EJC has a role in NMD, there are some strong clues that it might. As mentioned above, the Y14 and SRm160 proteins escort the spliced mRNA into the cytoplasm. Furthermore, recent reports implicate the EJC component RNPS1 in NMD [25] and show an association between the EJC and the NMD factors hUpf3 and hUpf2 [23,26].

Lykke-Andersen et al. [25] used a tethering approach to examine the role of the different components of the EJC in NMD. Using as a reporter gene a human β-globin mRNA with binding sites for the bacteriophage MS2 coat protein in its 3' untranslated region (UTR), they examined whether EJC proteins fused to the MS2 coat protein could trigger NMD when the MS2 binding sites were located at an appropriate distance (over 50 nucleotides) downstream of the stop codon. Like any of the hUpf proteins [9], when tethered in this manner RNPS1, and to a lesser extent also Y14, induced NMD of the reporter β-globin mRNA in transfected HeLa cells (with RNPS1, 22-34% of the RNA remained compared with control RNA without binding sites; with Y14, 56-72% remained). Importantly, this effect could be partially inhibited by a dominant-negative mutant of hUpf1. None of the other components of the EJC exhibited NMD activity in this assay. RNPS1 is known to shuttle between the nucleus and the cytoplasm of HEK293 human embryonic kidney cells, and thus could contribute to NMD in both subcellular locations.

When the EJC components were tagged individually, only RNPS1 could co-immunoprecipitate hUpf1, hUpf2 and hUpf3 from lysates of transfected HEK293 cells; this result suggests that RNPS1 is the component of the EJC that interacts with the NMD factors [25]. On the other hand, Kim et al. [26] found using a tag-specific antibody that tagged hUpf3a or hUpf3b could co-immunoprecipitate Y14. Furthermore, recombinant Y14 could bind hUpf3a or hUpf3b translated in vitro. Although these experiments suggest a direct interaction between Y14 and hUpf3a/b, the possibility that proteins (such as RNPS1) present in the in vitro translation reactions could mediate the interaction cannot be excluded. In support of the hypothesis that hUpf3 is a bona fide component of the EJC at post-splicing stages of mRNA maturation [9,23], Kim et al. [26] showed that FLAG-tagged hUpf3b stably associated with spliced mRNA at a position similar to that of the EJC. Taken together, the experiments of Lykke-Andersen et al. [25] and Kim et al. [26] provide biochemical support for the idea that the EJC subunits RNPS1 and/or Y14 are involved in NMD by providing a bridge between the splicing 'mark' and the Upf proteins (Figure 2), which in turn are linked to translation termination.

Translation and NMD

For NMD to protect the cell from aberrant proteins, it is crucial to destroy PTC-containing mRNAs as soon as possible and before multiple truncated polypeptides have been translated. Could this destruction happen as early as when the mRNA is still in the nucleus? Although the possibility of 'nuclear translation' has recently received renewed attention [27,28], cytoplasmic translation appears to be required for NMD [13,14,18,21]; the first round of translation, that is, the first time that the mRNA passes through the ribosome, is considered to be particularly important [29]. What distinguishes the initial round from subsequent rounds of translation? Conceivably, NMD could take place while the mRNA is being exported from the nucleus. During this phase, nuclear factors bound to the mRNA could be removed and potentially replaced by cytosolic proteins. Is this initial round of translation initiated by the canonical translation initiation factors or does it involve any special translation initiation factors?

In the nucleus, the 7mGpppN cap structure of mRNAs is bound by a heterodimeric cap-binding complex (CBC, consisting of CBP20 and CBP80), which can affect splicing and 3' end processing [30]. It appears that mRNAs are exported with their 5' ends first, exposing the CBC to the cytoplasm early in the transport process [31]. The initiation factor eIF4E binds the cap structure in the cytoplasm, displacing the CBC, and mediates translation by recruiting eIF4G, other translation factors and ultimately the ribosome. This raises the question of whether the CBC and the cytoplasmic eIF4E exchange before, during or after the initial round of translation (Figure 2). Notably, both yeast and mammalian CBCs can, like eIF4E, interact with eIF4G (via their CBP80 subunits) [32,33], and the yeast CBC has been shown to be able to support cell-free translation, albeit quite inefficiently [32].

Recently, Ishigaki et al. [34] proposed that a "pioneer round of translation" may be initiated while the mRNA is still associated with the CBC, and that this could act to survey the mRNA for the presence of PTCs. Two versions of a human β-globin mRNA served as a reporter system for the experiments of Ishigaki et al.: a β-globin mRNA with a PTC ('β-globin PTC') and one without ('β-globin normal') [34]. Following immunoprecipitation with antibodies against CBP80, the level of co-immunoprecipitated β-globin PTC RNA was reduced to 20% of the level of β-globin normal RNA. This reduction quantitatively reflects NMD of the β-globin PTC RNA in the total cellular RNA fraction, and corresponds to the level of reduction in the β-globin PTC RNA (15%) that was precipitatable with antibodies to eIF4E. Thus, the authors suggest that surveillance and NMD of β-globin PTC take place while the mRNA is associated with CBP80 and before the remaining intact RNA is handed over to eIF4E [34].

Ishigaki et al. [34] also showed that hUpf2, hUpf3 (but not hUpf1) and the nuclear poly(A)-binding protein (PABP2) co-immunoprecipitated with anti-CBP80 antibodies in an RNA-dependent manner, whereas complexes precipitated with the anti-eIF4E antibody did not contain any of these proteins. Given that hUpf2 and hUpf3 are involved in NMD, these results can be interpreted to suggest that NMD occurs while the mRNA is still bound by CBP80, but not detectably when the RNA is bound to eIF4E [34].

Do the majority of mRNAs undergo the initial round of translation without the support of eIF4E, which was thought to be a very important translation initiation factor? This idea offers a very interesting and unexpected twist to our perception of the translation and NMD pathway, and will surely stimulate many interesting future experiments. Before accepting the notion of a 'pioneer round' of mRNA translation under the guidance of CBP80 as a general aspect of mRNA export, however, some additional considerations may be warranted. At least in vitro, the CBC is an ineffective translation-initiation factor compared with eIF4E [32]. Moreover, a yeast strain with the CBP80 gene deleted does not display any overt translational defects nor any lack of NMD [35]; CBP80 is thus not essential either for NMD or for translation in yeast. The function of the CBC in mammalian cells may differ from its functions in yeast, however, and/or it may act as a more efficient translation factor in vivo than in vitro. The finding that the amount of β-globin PTC mRNA associated with CBP80 was reduced to 20% of the level of β-globin normal mRNA [34] certainly suggests that translation and NMD can occur before CBC has been exchanged for eIF4E. Only about 15%-30% of the CBC and eIF4E proteins and their associated RNAs from the extract was accessible to immunoprecipitation [34], however, and we therefore do not know about the mRNAs associated with the other 70-85% of these proteins. Comparisons of mRNA levels were made between PTC-containing and wild-type RNAs each co-immunoprecipitated with either CBP80 or eIF4E, but we do not know the percentage of the β-globin PTC and β-globin normal mRNAs present in the cells that was analyzed [34]. Thus, we do not yet know to what extent the analyzed mRNA-protein complexes (mRNPs) are representative of the majority of mRNAs being exported from the nucleus and being surveyed for NMD. Undoubtedly, future experiments will address this question as well as some of the intriguing implications of this work.

From a broader perspective, the recent progress highlights the dynamic nature of mRNP remodeling during mRNA processing, export, translation and degradation. We are on the way to a better understanding of how the different levels of gene expression are integrated and coordinated, how they affect and influence each other, and what the players and mechanisms are at the biochemical level.

References

  1. Culbertson MR: RNA surveillance. Unforeseen consequences for gene expression, inherited genetic disorders and cancer.

    Trends Genet 1999, 15:74-80. PubMed Abstract | Publisher Full Text OpenURL

  2. Hall GW, Thein S: Nonsense codon mutations in the terminal exon of the beta-globin gene are not associated with a reduction in beta-mRNA accumulation: a mechanism for the phenotype of dominant beta-thalassemia.

    Blood 1994, 83:2031-2037. PubMed Abstract OpenURL

  3. Schwabe GC, Tinschert S, Buschow C, Meinecke P, Wolff G, Gillessen-Kaesbach G, Oldridge M, Wilkie AO, Komec R, Mundlos S: Distinct mutations in the receptor tyrosine kinase gene ROR2 cause brachydactyly type B.

    Am J Hum Genet 2000, 67:822-831. PubMed Abstract | Publisher Full Text OpenURL

  4. Li S, Wilkinson MF: Nonsense surveillance in lymphocytes?

    Immunity 1998, 8:135-141. PubMed Abstract | Publisher Full Text OpenURL

  5. Czaplinski K, Ruiz-Echevarria MJ, Gonzalez CI, Peltz SW: Should we kill the messenger? The role of the surveillance complex in translation termination and mRNA turnover.

    BioEssays 1999, 21:685-696. PubMed Abstract | Publisher Full Text OpenURL

  6. Applequist SE, Selg M, Raman C, Jack HM: Cloning and characterization of HUPF1, a human homolog of the Saccharomyces cerevisiae nonsense mRNA-reducing UPF1 protein.

    Nucleic Acids Res 1997, 25:814-821. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  7. Sun X, Perlick HA, Dietz HC, Maquat LE: A mutated human homologue to yeast Upf1 protein has a dominant-negative effect on the decay of nonsense-containing mRNAs in mammalian cells.

    Proc Natl Acad Sci USA 1998, 95:10009-10014. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  8. Mendell JT, Medghalchi SM, Lake RG, Noensie EN, Dietz HC: Novel Upf2p orthologues suggest a functional link between translation initiation and nonsense surveillance complexes.

    Mol Cell Biol 2000, 20:8944-8957. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  9. Lykke-Andersen J, Shu M, Steitz JA: Human Upf proteins target an mRNA for nonsense-mediated decay when bound downstream of a termination codon.

    Cell 2000, 103:1121-1131. PubMed Abstract | Publisher Full Text OpenURL

  10. Serin G, Gersappe A, Black JD, Aronoff R, Maquat LE: Identification and characterization of human orthologues to Saccharomyces cerevisiae Upf2 protein and Upf3 protein (Caenorhabditis elegans SMG-4).

    Mol Cell Biol 2001, 21:209-223. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  11. Pal M, Ishigaki Y, Nagy E, Maquat LE: Evidence that phosphorylation of human upfl protein varies with intracellular location and is mediated by a wortmannin-sensitive and rapamycin-sensitive PI 3-kinase-related kinase signaling pathway.

    RNA 2001, 7:5-15. PubMed Abstract | Publisher Full Text OpenURL

  12. Czaplinski K, Ruiz-Echevarria MJ, Paushkin SV, Han X, Weng Y, Perlick HA, Dietz HC, Ter-Avanesyan MD, Peltz SW: The surveillance complex interacts with the translation release factors to enhance termination and degrade aberrant mRNAs.

    Genes Dev 1998, 12:1665-1677. PubMed Abstract | Publisher Full Text OpenURL

  13. Hentze MW, Kulozik AE: A perfect message: RNA surveillance and nonsense-mediated decay.

    Cell 1999, 96:307-310. PubMed Abstract | Publisher Full Text OpenURL

  14. Maquat LE: Nonsense-mediated RNA decay in mammalian cells: a splicing-dependent means to down-regulate the levels of mRNAs that prematurely terminate translation. In In Translational control. Edited by Hershey JWB, Mathews MB, Sonenberg N. Cold Spring Harbor: Cold Spring Harbor Laboratory Press,. Edited by Hershey JWB, Mathews MB, Sonenberg N. Cold Spring Harbor: Cold Spring Harbor Laboratory Press; 2000:849-868. OpenURL

  15. Maquat LE, Li X: Mammalian heat shock p70 and histone H4 transcripts, which derive from naturally intronless genes, are immune to nonsense-mediated decay.

    RNA 2001, 7:445-456. PubMed Abstract | Publisher Full Text OpenURL

  16. Brocke KS, Neu-Yilik G, Gehring NH, Hentze MW, Kulozik AE: The human intronless melanocortin 4-receptor gene is NMD insensitive.

    Hum Mol Genet 2002, 11:331-335. PubMed Abstract | Publisher Full Text OpenURL

  17. Neu-Yilik G, Gehring NH, Thermann R, Frede U, Hentze MW, Kulozik AE: Splicing and 3' end formation in the definition of nonsense-mediated decay-competent human β-globin mRNPs.

    EMBO J 2001, 20:532-540. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  18. Thermann R, Neu-Yilik G, Deters A, Frede U, Wehr K, Hagemeier C, Hentze MW, Kulozik AE: Binary specification of nonsense codons by splicing and cytoplasmic translation.

    EMBO J 1998, 17:3484-3494. PubMed Abstract | Publisher Full Text OpenURL

  19. Carter MS, Li S, Wilkinson MF: A splicing-dependent regulatory mechanism that detects translation signals.

    EMBO J 1996, 15:5965-5975. PubMed Abstract OpenURL

  20. Nagy E, Maquat LE: A rule for termination-codon position within intron-containing genes: when nonsense affects RNA abundance.

    Trends Biochem Sci 1998, 23:198-199. PubMed Abstract | Publisher Full Text OpenURL

  21. Hilleren P, Parker R: Mechanisms of mRNA surveillance in eukaryotes.

    Annu Rev Genet 1999, 33:229-260. PubMed Abstract | Publisher Full Text OpenURL

  22. Le Hir H, Moore MJ, Maquat LE: Pre-mRNA splicing alters mRNP composition: evidence for stable association of proteins at exon-exon junctions.

    Genes Dev 2000, 14:1098-1108. PubMed Abstract | Publisher Full Text OpenURL

  23. Le Hir H, Gatfield D, Izaurralde E, Moore MJ: The exon-exon junction complex provides a binding platform for factors involved in mRNA export and nonsense-mediated mRNA decay.

    EMBO J 2001, 20:4987-4997. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  24. Conti E, Izaurralde E: Nucleocytoplasmic transport enters the atomic age.

    Curr Opin Cell Biol 2001, 13:310-319. PubMed Abstract | Publisher Full Text OpenURL

  25. Lykke-Andersen J, Shu MD, Steitz JA: Communication of the position of exon-exon junctions to the mRNA surveillance machinery by the protein RNPS1.

    Science 2001, 293:1836-1839. PubMed Abstract | Publisher Full Text OpenURL

  26. Kim VN, Kataoka N, Dreyfuss G: Role of the nonsense-mediated decay factor hUpf3 in the splicing-dependent exon-exon junction complex.

    Science 2001, 293:1832-1836. PubMed Abstract | Publisher Full Text OpenURL

  27. Iborra FJ, Jackson DA, Cook PR: Coupled transcription and translation within nuclei of mammalian cells.

    Science 2001, 293:1139-1142. PubMed Abstract | Publisher Full Text OpenURL

  28. Hentze MW: Protein synthesis. Believe it or not-translation in the nucleus.

    Science 2001, 293:1058-1059. PubMed Abstract | Publisher Full Text OpenURL

  29. Cheng J, Maquat LE: Nonsense codons can reduce the abundance of nuclear mRNA without affecting the abundance of pre-mRNA or the half-life of cytoplasmic mRNA.

    Mol Cell Biol 1993, 13:1892-1902. PubMed Abstract OpenURL

  30. Lewis JD, Izaurralde E: The role of the cap structure in RNA processing and nuclear export.

    Eur J Biochem 1997, 247:461-469. PubMed Abstract OpenURL

  31. Mehlin H, Daneholt B, Skoglund U: Translocation of a specific premessenger ribonucleoprotein particle through the nuclear pore studied with electron microscope tomography.

    Cell 1992, 69:605-613. PubMed Abstract | Publisher Full Text OpenURL

  32. Fortes P, Inada T, Preiss T, Hentze MW, Mattaj IW, Sachs AB: The yeast nuclear cap binding complex can interact with translation factor eIF4G and mediate translation initiation.

    Mol Cell 2000, 6:191-196. PubMed Abstract | Publisher Full Text OpenURL

  33. McKendrick L, Thompson E, Ferreira J, Morley SJ, Lewis JD: Interaction of eukaryotic translation initiation factor 4G with the nuclear cap-binding complex provides a link between nuclear and cytoplasmic functions of the m(7) guanosine cap.

    Mol Cell Biol 2001, 21:3632-3641. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL

  34. Ishigaki Y, Li X, Serin G, Maquat LE: Evidence for a pioneer round of mRNA translation: mRNAs subject to nonsense-mediated decay in mammalian cells are bound by CBP80 and CBP20.

    Cell 2001, 106:607-617. PubMed Abstract | Publisher Full Text OpenURL

  35. Das B, Guo Z, Russo P, Chartrand P, Sherman F: The role of nuclear cap binding protein Cbc1p of yeast in mRNA termination and degradation.

    Mol Cell Biol 2000, 20:2827-2838. PubMed Abstract | Publisher Full Text | PubMed Central Full Text OpenURL