Healthy aging relies on proper function of the transcription factor p53. Known as “guardian of the genome,” this protein regulates key physiological events such as the cell cycle, DNA repair, and apoptosis. Research published in this week’s Neuron suggests that a p53 family member could play a similar protective role in the nervous system. Led by David Kaplan and Freda Miller at the University of Toronto, Canadian researchers report that older mice with halved levels of p73, a p53 homolog expressed primarily in the brain, display behavioral deficits characteristic of Alzheimer’s and related diseases. Accompanying these cognitive and motor problems were numerous brain abnormalities. Most notably, neurons degenerated and phospho-tau filaments accumulated throughout relevant brain areas. Furthermore, p73 haploinsufficiency triggered early and robust tau tangle formation and neurodegeneration in a separate mouse model of AD amyloidosis that typically lacks these features. Based on these findings, p73 levels appear to “determine whether your brain is going to be susceptible to injury as you get older,” Kaplan told ARF. But the story may be more complex. P73 isoforms can be pro- or anti-apoptotic, with adult brain primarily expressing the latter. While the new data highlight a protective role for p73, other studies—including several published recently by Alejandra Alvarez and colleagues at Universidad Católica de Chile in Santiago—suggest that Aβ-induced neuronal damage can stabilize pro-apoptotic p73 isoforms and lead to increased cell death.

In the nervous system, p73 is expressed primarily as a dominant-inhibitory isoform (DNp73) that lacks the transactivation domain and thereby opposes the pro-death activity of p53 and full-length p73 (TAp73). Previous work by Kaplan, Miller and colleagues has shown that DNp73 prevents neuronal apoptosis during development (Pozniak et al., 2000; Pozniak et al., 2002) and helps adult neurons resist injury (Walsh et al., 2004). More recent studies have found a weak association between decreased p73 expression and AD susceptibility (Li et al., 2004) and shown that some 10 percent of the human population has lost one copy of a genomic region encompassing the p73 gene (Wong et al., 2007).

First author Monica Wetzel and colleagues put these findings to the test by looking for signs of neurodegeneration in p73-deficient mice (Yang et al., 2000). At three to four months of age, wild-type and p73+/- mice appeared fine behavior-wise. “But when they were old, they started to get all the behavioral and anatomical problems that we see in people who age, or get dementia or AD,” Kaplan said. Specifically, 16- to 18-month-old p73 heterozygotes fared poorly on various neurological and sensory-motor assays including open-field exploration, paw grip endurance, grid walking, and gait analysis. They also did worse than wild-type and young p73+/- animals in the Morris water maze that tests learning and memory.

Using 3D magnetic resonance imaging, the researchers found evidence of neurodegeneration that might underlie these outward abnormalities. Relative to their wild-type counterparts, aged p73+/- mice had 5 to 16 percent reduced volume in the motor cortex, dentate gyrus, posterior cerebellum, and other brain areas important for the behavioral tests. In the motor cortex of old p73+/- mice, neuronal density was lower, and total neuron numbers were down 29 percent compared with aged controls. Sure enough, silver staining revealed dying neurons in aged p73+/- motor cortex but not in young p73+/- or wild-type brains.

As it turns out, the aged p73+/- mice harbored a laundry list of abnormalities typical of neurodegenerative disorders. The list included increased microglial density in the brain, greater numbers of reactive astrocytes in the motor cortex, doubling of the number of neurons aberrantly re-entering the cell cycle, and more brain cells expressing senescence-associated β-galactosidase (SA-β-Gal), a marker of cell aging.

Perhaps the biggest surprise came when the researchers found phospho-tau-containing paired helical filaments (P-PHF-tau) in immunostained sections of aged p73+/- brains—a result established using three P-PHF-tau antibodies and confirmed by the disappearance of the immunostaining upon phosphatase treatment.

What happened when p73-deficient mice were crossed with transgenic CRND8 mice (an AD model that expresses double-mutant amyloid precursor protein (APP) and develops plaques, but not tangles, by three months, and does not show neurodegeneration)? At just 1.5 to two months of age, the researchers saw whopping amounts of P-PHF-tau filaments in CRND8 mice with hemizygous but not wild-type p73 levels. They also observed a 22 percent drop in neuron numbers in the motor cortex of p73+/- CRND8 animals relative to p73+/+ CRND8 littermates.

"These remarkable findings provide an AD model that not only exhibits P-PHF-tau positive filaments and amyloid plaques but, more importantly, demonstrates obvious neurodegeneration that seems to underlie behavioral alterations observed at similar time points,” wrote AD researcher Yong Shen in an e-mail to ARF. Shen is an expert in neuronal cell death signaling at Sun Health Research Institute in Sun City, Arizona.

As for how p73 might regulate neuronal survival, Kaplan and colleagues considered several possibilities—the first being that it antagonizes pro-death family members. This explanation was dismissed, however, when levels of p53 and p63 mRNA and several of their downstream targets were found to be equal in aged p73+/- and p73+/+ cortices. Another possibility is that p73’s effects involve JNK, a death protein that gets upregulated during nerve injury and that has been shown to bind and be inhibited by DNp73 (Lee et al., 2004). JNK also appears associated with AD, as it directly phosphorylates tau on residues important for tangle formation.

To see if p73 might be linked to AD via JNK, the researchers measured JNK activity in cultured p73+/+, p73+/-, and p73-/- cortical neurons. As p73 levels dropped, levels of activated, phosphorylated JNK rose, and neurons became more vulnerable to death in mildly excitotoxic glutamate concentrations. Furthermore, p73-deficient neonatal cortices showed increased tau phosphorylation at two JNK target sites, and treatment with a JNK inhibitor decreased tau phosphorylation in p73+/+ and p73-/- neurons. Further support for the idea that p73 exerts its neuroprotective effects via JNK came from immunocytochemistry studies that revealed in p73+/- CRND8 brains a 20-fold increase in phospho-JNK-positive cells—many of which were also positive for the two phospho-tau residues examined in the neonatal cortices.

All told, Kaplan said, these data suggest that normally “p73 binds to JNK and keeps it from phosphorylating tau.” He said that large-scale genetic analyses could help tease out whether p73 is a susceptibility factor for AD and/or other neurodegenerative diseases—by showing that individuals with lower p73 levels indeed have higher rates of such disorders, as suggested previously in a smaller study (Li et al., 2004). Studies with p73-inducing drugs also could test the claim that p73 prevents age-related neurodegeneration. In other words, Kaplan said, “Can we increase levels of p73 and in doing so protect the brain as we age?”

While the new work points to an overall protective function for p73 in the aging brain, recent reports by Alvarez and others suggest that the protein may have a more nuanced role. Full-length (TAp73) isoforms are regularly removed in healthy cells by the ubiquitin-proteasome system; however, c-Abl kinase activated in response to cell damage has been shown to phosphorylate and stabilize p73 (Tsai and Yuan, 2003). In this month’s issue of the journal Brain, Alvarez, first author Gonzalo Cancino, and colleagues report that the c-Abl/p73 system is activated in cultured neuronal cells exposed to β amyloid, as well as in rats given hippocampal β amyloid injections and AD transgenic mice (APPsw/PSEN1deltaE9).

In these models, intraperitoneal administration of the c-Abl inhibitor STI571 (imatinib mesylate or Gleevec, an FDA-approved treatment for chronic myelogenous leukemia and gastrointestinal stromal tumor) not only reduced Aβ-induced behavioral defects but also apoptosis and tau phosphorylation (Cancino et al., 2008). For previous data on neurodegeneration-relevant effects of Gleevec, see Netzer et al., 2003; Bantscheff et al., 2007. In Alvarez’s hands, Gleevec also reduced neurological symptoms and cerebellar apoptosis in a mouse model for Niemann-Pick type C (NPC) disease, another disorder marked by progressive neuronal loss (Alvarez et al., 2008). Alvarez proposed that cellular damage signals, which appear to stabilize and induce increased expression of full-length p73, might change the balance between anti-apoptotic and pro-apoptotic p53 family members. The new data from Kaplan’s team “highlight the central role of the p53 family in the nervous system—not only in development but also in neurodegeneration—and raise new questions about the complex functions and crosstalk between these proteins,” she wrote to ARF (see full comment below).

Along similar lines, Richard Killick of King's College London, U.K., told ARF via e-mail that the work by Kaplan and colleagues clearly implicates the p53 family in the hyperphosphorylation of tau tangle formation. Though the new data (showing more P-PHF-tau filaments in mice with less p73) would appear at odds with earlier evidence of reduced tau phosphorylation in p53-deficient mice (Ferreira and Kosik, 1996), Killick argues that the findings could be reconciled if “p53 is the main driver of tau phosphorylation and that in brain the dominant-negative forms of p73 repress this (see full comment below).”—Esther Landhuis

Comments

  1. During the course of our research, we have found our attention inexorably drawn to a well-known and very widely studied gene that is mutated in more than 55 percent of all cancers, p53. P53 functions as a sequence-specific transcription factor, which upon activation by a variety of cellular stresses, activates downstream target genes, through which it regulates the cell cycle, DNA repair, and apoptosis. Our own data and that of others, along with recent advances in the p53 field, lead us to believe that the p53 family is a central player in AD pathology.

    We have shown that p73, a homologue of p53 preferentially expressed in brain, is able to phosphorylate tau (Hooper et al., 2006), as does p53 itself (Hooper et al., 2007). In fact, the third member of the family, p63, is also capable of doing this (our own unpublished data). This phenomenon only applies to those forms of the p53 family that contain the transactivation domain (TA). The forms that lack this domain, the ΔN forms, do not exert this effect.

    Tau phosphorylation is reduced in the p53 KO mouse (Ferreira and Kosik, 1996). Now the Kaplan group have shown that in brains of old but not young p73+/- mice there are accumulations of phospho-tau-positive paired helical filaments (p-PHF-tau) reminiscent of tangles. Moreover, when crossed to the TgCRND8 mouse model of AD (a double APP mutant) the resultant mice show early and robust formation of p-PHF-tau tangle-like structures. This would appear to be at odds with our own work and the observations made in the p53 KO mouse. However, it can be explained if, with regard to tau phosphorylation, p53 is the main driver of tau phosphorylation and that in brain the ΔN forms of p73 repress this. The knocking out of one copy of p73 would reduce the expression of both TA and ΔN forms of p73, leading to increased p53 activity and increased tau phosphorylation. Indeed, the authors argue along these lines themselves. The p53 family is now clearly implicated in the hyperphosphorylation of tau tangle formation.

    References:

    . Accelerated neuronal differentiation induced by p53 suppression. J Cell Sci. 1996 Jun;109 ( Pt 6):1509-16. PubMed.

    . p53 is upregulated in Alzheimer's disease and induces tau phosphorylation in HEK293a cells. Neurosci Lett. 2007 May 11;418(1):34-7. PubMed.

    . TAp73 isoforms antagonize Notch signalling in SH-SY5Y neuroblastomas and in primary neurones. J Neurochem. 2006 Nov;99(3):989-99. PubMed.

    View all comments by Richard Killick
  2. This paper reporting the role of p73 in regulating neurodegeneration and tau phosphorylation during aging and Alzheimer disease is very exciting. It provides strong validation for the balance of the p53 family in the control of neuronal fate (survival versus apoptosis).

    This work also offers strong support for p73, the p53 family member with highest expression in the PNS and CNS (Yang et al., 2000; Pozniak et al., 2000), as a key element in the survival versus apoptosis balance in neurons.

    P73 is a structural and functional homolog of p53 (Irwin and Kaelin, 2001). There are many different p73 protein isoforms—resulting from C-termini alternative splicing and the use of two different promoters—which fall into two classes: 1) full-length, or TAp73 proteins that have a N-terminal domain (TA) participate in DNA damage-induced cell-cycle arrest or apoptosis (Agami et al., 1999; Gong et al., 1999; Yuan et al., 1999; Melino et al., 2004), and 2) ΔNp73, or short-form proteins, which lack the TA domain. The ΔNp73 proteins cannot induce apoptosis, but instead appear to block the function of full-length forms showing anti-apoptotic properties.

    In PNS and CNS, the principally expressed p73 isoforms are DNp73, which prevent neuronal apoptosis during development (Walsh et al., 2004). Overexpression of ΔNp73 inhibits sympathetic neuronal apoptosis caused by NGF withdrawal and apoptosis of cortical neurons induced by camptothecin (Pozniak et al., 2000; Pozniak et al., 2002; Walsh et al., 2004).

    The results of this paper emphasize and reinforce the role of p73 as a protective gene in neurons, since a decrease in the expression of the p73 that is normally expressed in healthy neurons (the truncated isoforms) results in a diminished ability of the cells to face damage or insults. Moreover, this work brings the protective role of DNp73 from a developmental neuronal death context to the neurodegeneration and pathology context, because perturbations of its function may accelerate neuronal aging and/or predispose to neurodegenerative disorders.

    However, the role of p73 in neurodegeneration may not be limited to a protective role (performed by DNp73). We have described that p73 increases during neuronal damage induced by β amyloid (Alvarez et al., 2004), and that the TAp73 proapoptotic isoform and its apoptotic function are activated in neurodegeneration models (Cancino et al., 2008; Alvarez et al., 2008).

    TAp73 isoforms are regularly removed in healthy cells through the ubiquitin-proteasome system; however, in response to cell damage, the activated c-Abl kinase regulates p73 protein levels by phosphorylating and stabilizing it (Tsai and Yuan, 2003). In non-neuronal systems c-Abl is the activator of p73 and controls growth arrest and apoptotic response to genotoxic stress (Wang and Ki, 2001).

    We have found that the c-Abl/p73 pathway participates in AD neurodegeneration. The c-Abl/p73 system is activated in cultured neuronal cells exposed to β amyloid and in two Alzheimer disease (AD) in vivo models—rats exposed to β amyloid hippocampal injections and transgenic APPsewPSEN1dE9 mice (Cancino et al., 2008). In these models the intraperitoneal administration of imatinib mesylate (STI571), a c-Abl inhibitor, reduced the behavioral deficit induced by β amyloid as well as apoptosis and tau phosphorylation (Cancino et al., 2008).

    In agreement with a role for TAp73 in AD, levels of p73 have been reported to be increased in the nuclei of AD pyramidal neurons, and p73 is present in dystrophic neurites with cytoskeletal pathology (Wilson et al., 2004). In this context, some p73 gene polymorphisms—probably associated with the relative expression of p73 isoforms—have been linked to AD (Li et al., 2004). In addition, it has been recently shown that p73 is activated in striatal neurons of Huntington disease patients (Hoshino et al., 2006) and in motor neurons of an ALS mouse model (Martin et al., 2007). Therefore, it is possible that the activation of the c-Abl/p73 module is a common pathological mechanism participating in a number of neurodegenerative diseases.

    We have also explored the relationship between Niemann-Pick type C (NPC) disease neurodegeneration and activation of the c-Abl/p73 apoptotic system (Alvarez et al., 2008). We found that both c-Abl and p73 proteins are expressed in the cerebellum, the region most affected in NPC brains. In Purkinje cells of NPC mice, expression of the proapoptotic, phosphorylated p73 colocalized with c-Abl and active caspase 3, and p73-proapoptotic target gene levels were increased in NPC cerebellum. Strikingly, inhibition of c-Abl kinase with imatinib mesylate reduced weight loss, neurological symptoms, and cerebellar apoptosis, increasing Purkinje cell numbers and survival of NPC mice.

    Because TAp73 isoforms are regularly removed in healthy cells through the ubiquitin-proteosome system, it might be difficult to detect TAp73 in total brain. However, there are some neuronal populations that express TAp73, including the hippocampus (Cabrera-Socorro et al., 2006). Damage signals stabilize TAp73 and induce an increase of its levels; therefore, damage can change the balance between the anti-apoptotic and proapoptotic p53 family members.

    Moreover, in sympathetic neurons, p63 (another p53 homolog) is essential for developmental neuronal death (Jacobs et al., 2005).

    The Wetzel et al. paper also shows that JNK signaling is a key element in the p53 family member’s functions, and the balance between survival and death in neurons. And more interesting, JNK signaling regulates P-tau levels—a central issue in neurodegeneration.

    This wonderful paper highlights the central role of the p53 family in the nervous systems—not only in development but also in neurodegeneration—and opens new questions about the complex functions and crosstalk between these proteins. What is the protein expression profile for the different p73 isoforms in neuronal populations or brain areas? Are there other signaling pathways that control p73 function upstream in neurons?

    References:

    . p73-deficient mice have neurological, pheromonal and inflammatory defects but lack spontaneous tumours. Nature. 2000 Mar 2;404(6773):99-103. PubMed.

    . An anti-apoptotic role for the p53 family member, p73, during developmental neuron death. Science. 2000 Jul 14;289(5477):304-6. PubMed.

    . Role of the newer p53 family proteins in malignancy. Apoptosis. 2001 Feb-Apr;6(1-2):17-29. PubMed.

    . Interaction of c-Abl and p73alpha and their collaboration to induce apoptosis. Nature. 1999 Jun 24;399(6738):809-13. PubMed.

    . The tyrosine kinase c-Abl regulates p73 in apoptotic response to cisplatin-induced DNA damage. Nature. 1999 Jun 24;399(6738):806-9. PubMed.

    . p73 is regulated by tyrosine kinase c-Abl in the apoptotic response to DNA damage. Nature. 1999 Jun 24;399(6738):814-7. PubMed.

    . p73 Induces apoptosis via PUMA transactivation and Bax mitochondrial translocation. J Biol Chem. 2004 Feb 27;279(9):8076-83. PubMed.

    . The invulnerability of adult neurons: a critical role for p73. J Neurosci. 2004 Oct 27;24(43):9638-47. PubMed.

    . p73 is required for survival and maintenance of CNS neurons. J Neurosci. 2002 Nov 15;22(22):9800-9. PubMed.

    . Activation of the neuronal c-Abl tyrosine kinase by amyloid-beta-peptide and reactive oxygen species. Neurobiol Dis. 2004 Nov;17(2):326-36. PubMed.

    . STI571 prevents apoptosis, tau phosphorylation and behavioural impairments induced by Alzheimer's beta-amyloid deposits. Brain. 2008 Sep;131(Pt 9):2425-42. PubMed.

    . Imatinib therapy blocks cerebellar apoptosis and improves neurological symptoms in a mouse model of Niemann-Pick type C disease. FASEB J. 2008 Oct;22(10):3617-27. PubMed.

    . c-Abl stabilizes p73 by a phosphorylation-augmented interaction. Cancer Res. 2003 Jun 15;63(12):3418-24. PubMed.

    . Choosing between growth arrest and apoptosis through the retinoblastoma tumour suppressor protein, Abl and p73. Biochem Soc Trans. 2001 Nov;29(Pt 6):666-73. PubMed.

    . STI571 prevents apoptosis, tau phosphorylation and behavioural impairments induced by Alzheimer's beta-amyloid deposits. Brain. 2008 Sep;131(Pt 9):2425-42. PubMed.

    . The p53 homologue p73 accumulates in the nucleus and localizes to neurites and neurofibrillary tangles in Alzheimer disease brain. Neuropathol Appl Neurobiol. 2004 Feb;30(1):19-29. PubMed.

    . TP73 allelic expression in human brain and allele frequencies in Alzheimer's disease. BMC Med Genet. 2004 Jun 2;5:14. PubMed.

    . Transcriptional repression induces a slowly progressive atypical neuronal death associated with changes of YAP isoforms and p73. J Cell Biol. 2006 Feb 13;172(4):589-604. PubMed.

    . Motor neuron degeneration in amyotrophic lateral sclerosis mutant superoxide dismutase-1 transgenic mice: mechanisms of mitochondriopathy and cell death. J Comp Neurol. 2007 Jan 1;500(1):20-46. PubMed.

    . Multiple isoforms of the tumor protein p73 are expressed in the adult human telencephalon and choroid plexus and present in the cerebrospinal fluid. Eur J Neurosci. 2006 Apr;23(8):2109-18. PubMed.

    . p63 is an essential proapoptotic protein during neural development. Neuron. 2005 Dec 8;48(5):743-56. PubMed.

    View all comments by Alejandra Alvarez
  3. This is an exciting paper whose results may trigger a series of follow-up studies. Several questions could be studied further. Being a transcription factor, how does p73 regulate and/or interact with APP to cause early onset of AD-like pathology? Specifically, how could this lead to P-PHF? AD is a chronic neurodegenerative disease, whereas TgCRND8+/- mice exhibit an extremely quick response and develop plaques much earlier than the Tg2576 and APP23 strains. The p73+/-/TgCRND8+/- mice start showing AD-like pathology at 1.5 or two months of age. It would be interesting to see whether p73+/- mice crossed with Tg2576 and APP23 transgenic mice would also generate early onset of AD-like pathology. For microglial activation, it seems that the glial activation may be caused by the p73 knockout. It need not necessarily be relevant to Aβ in this model, as p73+/- mice alone exhibit glial activation. Thus, p73 may have more functions to further explore. For example, does p73 affect APP processing? Does p73 affect Aβ degradation enzymes? Beside JNK, does p73 influence other kinases that may phosphorylate tau?

    The recent paper in Brain (Cancino et al., 2008) demonstrates that p73 protein mainly has two isoforms: 1) the full-length TAp73 protein whose N-terminal domain participates in cell-cycle arrest or apoptosis induced by DNA-damage and 2) Np73 short proteins lacking this domain. These cannot induce apoptosis but instead appear to block the function of full-length forms. Cancino et al. mainly focus on the full-length p73 protein. Using antibodies that recognize the TAp73 or truncated Np73 isoforms, they found that Aβ induced a significant increase in the TAp73 isoform level, whereas Np73 isoforms were moderately decreased. Wetzel et al. propose that the reported neurodegeneration phenotypes are the result of reduced levels of Np73, which is the only detectable p73 isoform in the postnatal mouse CNS. In this sense, the data in both papers fit together.

    References:

    . Electroencephalography as a diagnostic tool in dementia. Dement Geriatr Cogn Disord. 1997 Mar-Apr;8(2):110-6. PubMed.

    View all comments by Yong Shen

Make a Comment

To make a comment you must login or register.

References

Paper Citations

  1. . An anti-apoptotic role for the p53 family member, p73, during developmental neuron death. Science. 2000 Jul 14;289(5477):304-6. PubMed.
  2. . p73 is required for survival and maintenance of CNS neurons. J Neurosci. 2002 Nov 15;22(22):9800-9. PubMed.
  3. . The invulnerability of adult neurons: a critical role for p73. J Neurosci. 2004 Oct 27;24(43):9638-47. PubMed.
  4. . TP73 allelic expression in human brain and allele frequencies in Alzheimer's disease. BMC Med Genet. 2004 Jun 2;5:14. PubMed.
  5. . A comprehensive analysis of common copy-number variations in the human genome. Am J Hum Genet. 2007 Jan;80(1):91-104. PubMed.
  6. . p73-deficient mice have neurological, pheromonal and inflammatory defects but lack spontaneous tumours. Nature. 2000 Mar 2;404(6773):99-103. PubMed.
  7. . Evidence that DeltaNp73 promotes neuronal survival by p53-dependent and p53-independent mechanisms. J Neurosci. 2004 Oct 13;24(41):9174-84. PubMed.
  8. . c-Abl stabilizes p73 by a phosphorylation-augmented interaction. Cancer Res. 2003 Jun 15;63(12):3418-24. PubMed.
  9. . STI571 prevents apoptosis, tau phosphorylation and behavioural impairments induced by Alzheimer's beta-amyloid deposits. Brain. 2008 Sep;131(Pt 9):2425-42. PubMed.
  10. . Gleevec inhibits beta-amyloid production but not Notch cleavage. Proc Natl Acad Sci U S A. 2003 Oct 14;100(21):12444-9. PubMed.
  11. . Quantitative chemical proteomics reveals mechanisms of action of clinical ABL kinase inhibitors. Nat Biotechnol. 2007 Sep;25(9):1035-44. PubMed.
  12. . Imatinib therapy blocks cerebellar apoptosis and improves neurological symptoms in a mouse model of Niemann-Pick type C disease. FASEB J. 2008 Oct;22(10):3617-27. PubMed.
  13. . Accelerated neuronal differentiation induced by p53 suppression. J Cell Sci. 1996 Jun;109 ( Pt 6):1509-16. PubMed.

Other Citations

  1. CRND8 mice

External Citations

  1. p73
  2. APPsw/PSEN1deltaE9

Further Reading

Papers

  1. . Imatinib therapy blocks cerebellar apoptosis and improves neurological symptoms in a mouse model of Niemann-Pick type C disease. FASEB J. 2008 Oct;22(10):3617-27. PubMed.

Primary Papers

  1. . STI571 prevents apoptosis, tau phosphorylation and behavioural impairments induced by Alzheimer's beta-amyloid deposits. Brain. 2008 Sep;131(Pt 9):2425-42. PubMed.
  2. . p73 regulates neurodegeneration and phospho-tau accumulation during aging and Alzheimer's disease. Neuron. 2008 Sep 11;59(5):708-21. PubMed.