Promoter Bashing—Mitochondrial Ones Damaged in AD Brain

Recently, scientists reported that certain human promoter sequences are more susceptible to age-related mutagenesis, suggesting that these somatic mutations may underlie many diseases of the elderly, including neurodegenerative disorders such as Alzheimer’s (see ARF related news story). Well, nuclear promoters may not be alone. In this week’s PNAS online, Douglas Wallace and Pinar Coskun at the University of California, Irvine, together with Flint Beal at Cornell University, New York, report that mutations in promoter regions of mitochondrial DNA are more common in people with AD.

The control region (CR) of the mitochondrial genome is the business end of this relatively small piece of DNA. Here, in just over 1,000 bases, reside several promoters, transcription factor binding sites, conserved sequence blocks (CSBs), and origins of replication. When Coskun tested postmortem frontal cortex samples, he found a higher incidence of CR mutations in samples taken from AD donors as compared to those from controls. This difference was statistically significant and was greater in older patients. Among those 80 and above, CR mutations were nearly 2.5 times higher in AD samples.

Could these mutations contribute to AD pathology—and how? Hints come from the location of the mutations. Not only are there more of them in AD samples, but they are also in more critical regions of the CR. In control samples, for example, most of the mutations occur between nucleotides 1 and 200, while most of the known functional sequences lie between nucleotides 200 and 570, which is where most of the AD sample mutations were. In CSB I, for example, Coskun found seven mutations in AD samples but none in control samples. Similar results were found for other key binding sites. The mutations may also affect mitochondrial number, and hence overall cellular fitness, because many of the mutations occurred in proximity to the PL region, a section of the genome that is necessary for transcription of the mitochondrial L-strand. As L-strand transcription is thought to be essential for replication of the genome, these mutations may prevent mitochondrial reproduction. In support of this, the authors found a 50 percent reduction in mtDNA copy number in AD samples.

These results also seem to point the finger, yet again, at reactive oxygen species (ROS) and mitochondrial damage as key players in neurodegeneration. “Once [CR] mutations arise,” write the authors, “they would become enriched in the postmitotic cells of the brain and ultimately result in mitochondrial oxidative phosphorylation deficiency, increased ROS production, activation of the mitochondrial permeability transition pore in the synaptic mitochondria, and loss of synaptic connections through apoptosis.—Tom Fagan

Comments

  1. This is a remarkable paper. Coskun, Beal and Wallace demonstrate that many mutations to mtDNA occur near the PL region, an important regulatory region for replication of the genome. Therefore, this damage could ultimately block mitochondrial reproduction. In support of this possibility, they found a 50 percent reduction in mtDNA copy number in AD samples. Thus, their results partially explain the mitochondrial deficiencies that occur in AD.

    Mitochondria constantly undergo fission and fusion, are replicated throughout the life of the cell, and autophagocytosed when their time has come. However, unlike their host, mitochondria possess multiple copies of their DNA, perhaps four or five. So, damage to one copy does not necessarily compromise the organelle, if it can be removed and disposed. That’s the catch-if the mitochondrion does not get rid of this DNA, eventually it may accrue more damaged copies.

    The mitochondrial theory of aging states that enlarged and functionally disabled mitochondria gradually displace normal ones during senescence. These changes are most pronounced in postmitotic cells, such as neurons. The large size hints that fission is impaired, limiting the effectiveness of mitochondrial turnover. These mitochondria may incur further oxidative damage, resulting in decreased energy production and increased generation of reactive oxygen specie. Disabled mitochondria gradually displace normal ones, and the neuron eventually dies. The Wallace paper did not determine the cause of the damage to mtDNA. Oxidative stress is the most likely culprit, since all patients suffered from late-onset AD, the disease probably results from damage that builds up slowly over time.

    Could a similar scenario explain the etiology of early onset, familial forms of AD? Mitochondria have fewer DNA repair mechanisms than the nucleus, thus cells-especially post-mitotic neurons-may have evolved an alternate response to this damage. It may be that neurons deal with intractable mtDNA damage by deleting the entire copy. Perhaps individual copies of mtDNA are targeted for disposal through association with the mitochondrial permeability transition pore. The oxidative stress that follows mtDNA damage would trigger the opening of the pore, and this complex could subsequently be removed by fission. The APP is involved in intracellular trafficking, and found in the mitochondrial membrane associated with the mitochondrial permeability transition pore, intriguingly, in a non-glycosylated form (Anandatheerthavarada et al. 2003). Could APP and the presenilins direct transport vesicles containing damaged mtDNA (and other mitochondrial components) to lysosomes? Thus, mutations that affect APP processing could affect vesicular transport or targeting, or perhaps even the initiation of this process. The damaged mtDNA would gradually increase over time to disable the neuron as previously described.

    The disposal or recycling of fragments of this organelle has not yet been described, although autophagocytosis of mitochondria has been reported for many models of neurodegeneration. However, we have unpublished evidence that supports the existence of such a process. In our unpublished study, cerebellar granule cells were transfected with wild-type or the Swedish mutant APP, as well as β-lactosidase (Lac) as a control. Cultures were then treated with kainate to stimulate AMPA receptors and fixed for electron microscopy at 0-30 min. All treated cerebellar granule cells showed evidence for mitochondrial membrane breakdown, however, it was more robust for the wild-type, and less so for the Swedish mutant transfected-neurons. These were usually observed near to the microtubule organizing center (see the RAB27-/- frame in Figure 2 of Stinchcombe et al. (2004) for an example of this). Curiously, we observed several lysosomes undergoing exocytosis, as well as numerous extracellular lysosomes and multivesicular bodies among the fibers. This process appears to peak 2 min following kainate treatment, which could explain why it has not been previously observed, as no study has examined neuronal ultrastructure at similar time intervals following glutamate receptor stimulation.

    There are reports that present data which support this hypothesis. One such study demonstrated that neurons showing increased oxidative damage in AD have a significant increase in mtDNA and cytochrome oxidase (Hirai et al. 2001). In this study, much of the mtDNA and cytochrome oxidase was found in the neuronal cytoplasm. Curiously, vacuoles associated with lipofuscin immunostained for mtDNA, while morphometric analysis showed that mitochondria were significantly reduced in AD, consistent with perturbed mitochondrial turnover. In our study, kainate treatment generated vacuoles that originated from mitochondria and appeared to contain lipofuscin, and were virtually identical in size and appearance to those that immunostained for mtDNA in the study by Smith’s group.

    The APP is not functionally constrained to the mitochondria in this model. APP targeted to the plasma membrane is fully glycosylated, while as noted above, the protein found in the mitochondria is not. One function of APP may be to retrieve DNA, lipids and proteins damaged following oxidative stress, and direct these to the Golgi and lysosomes for recycling or disposal. Further studies addressing the localization of mtDNA as well as vesicular targeting following glutamate receptor stimulation are in order.

    References:

    Anandatheerthavarada HK, Biswas G, Robin MA, Avadhani NG. Mitochondrial targeting and a novel transmembrane arrest of Alzheimer's amyloid precursor protein impairs mitochondrial function in neuronal cells. J Cell Biol. 2003 Apr 14;161(1):41-54. PubMed.

    Coskun PE, Beal MF, Wallace DC. Alzheimer's brains harbor somatic mtDNA control-region mutations that suppress mitochondrial transcription and replication. Proc Natl Acad Sci U S A. 2004 Jul 20;101(29):10726-31. PubMed.

    Hirai K, Aliev G, Nunomura A, Fujioka H, Russell RL, Atwood CS, Johnson AB, Kress Y, Vinters HV, Tabaton M, Shimohama S, Cash AD, Siedlak SL, Harris PL, Jones PK, Petersen RB, Perry G, Smith MA. Mitochondrial abnormalities in Alzheimer's disease. J Neurosci. 2001 May 1;21(9):3017-23. PubMed.

    Stinchcombe J, Bossi G, Griffiths GM. Linking albinism and immunity: the secrets of secretory lysosomes. Science. 2004 Jul 2;305(5680):55-9. PubMed.

    View all comments by Michael Leski

  2. A possible role for abnormal mitochondrial biology in the pathogenesis of Alzheimer’s disease (AD) has been an omnipresent theme over the last decades. Regional hypometabolism, documented by positron emission tomography and abnormalities in mitochondrial respiratory chain complexes—most consistently diminished activity of cytochrome c oxidase (COX), have been noted for some time (1-2). Furthermore, an important role for reactive oxygen species (ROS), which originate, at least in part, from mitochondria, in amyloid-β peptide (Aβ)-induced cellular perturbation has been indicated by multiple in vitro studies. In this regard, an attractive concept to consider is that of “retrograde signaling,” whereby mitochondrial generation of ROS diminishes calcium uptake by this organelle, resulting in increased oxidant stress and cytosolic-free calcium ([Ca2+]c), and subsequent activation of kinases, and activation of nuclear genes etc (3). Another close link between cellular perturbation, apoptosis and mitochondrial function is that a 75 kDa subunit of respiratory complex I (NDUFS1) is a functionally significant caspase substrate (4).

    What are the mechanisms through which Aβ targets mitochondria? On the one hand, there could be specific mitochondrial components/macromolecules which are rendered dysfunctional either directly and/or indirectly by Aβ. On the other hand, there might be a genetic propensity towards mitochondrial dysfunction (due to mutations in either nuclear or mitochondrial encoded genes) which renders cells especially vulnerable to Aβ-induced cell stress. Of course, these are not mutually exclusive possibilities.

    The paper from the laboratory of Dr. Wallace provides intriguing insights into these issues as pertains to genetic determinants of mitochondrial function. First, they show that there are certain mutations in the mitochondrial (mt) DNA control region (CR; a region of mtDNA that encodes essential regulatory binding sites) in AD brains that are not found in nondemented controls. Second, there were more mutations in the mtDNA CR of AD patients than controls. Interestingly, these mutations occurred at sites of known mtDNA regulatory elements, and would be expected to be associated with functional consequences for mitochondrial biology. Third, there was a decrease in mtDNA (along with ND6 mRNA) that would be likely be associated with reduced brain oxidative phosphorylation. Two distinct patterns of the mutations were especially notable. In one pattern, AD brains from patients of 74-83 years of age had a lower number of CR mutations, but they were present in a high percentage of the brain mtDNAs (70-80%). In the other pattern, AD patients over 83 years of age had many more CR mutations in frontal cortex mitochondria, but these were present in a lower percentage (We would like to suggest an expansion of this hypothesis and make a prediction. Certain patients have mtDNA mutations (presumably in, but not limited to, the control region), which render neurons vulnerable to oxidant stress. This might be considered a “first hit,” in a “two hit” model of cellular perturbation (see schematic figure below). However, the genetic defect encoding this “first hit” (grey shading) is not associated with a deleterious phenotype until a superimposed “second hit,” (red shading and red arrows) the latter associated with oxidant stress, occurs. The “second hit” could be an excitotoxic stimulus, regional cerebral ischemia or Aβ-induced pathology. If this concept is valid, one might expect that such mtDNA mutations (constituting the “first hit”) were distributed not only in frontal cortex (and, presumably, other AD-affected brain subregions), but also in areas of the brain spared in AD (for example, cerebellum). The occurrence of neuronal stress would require a localized oxidant stress (the “second hit,” due to an Aβ-rich environment, excitotoxic stimulus or ischemic insult), to bring out the phenotype caused by the mtDNA mutation(s). Alternatively, if Aβ-induced oxidant stress were the primary driver of mitochondrial dysfunction (in this case, the “first hit” would not be an important factor), one might expect mtDNA mutations to be focused only in AD-affected regions, to be multiple and locally (not widely) distributed. Of course, a combination of these hypotheses would be possible in which underlying mtDNA mutations occurred early in the disease, rendering neurons vulnerable to Aβ-induced cell stress, and this would provide a basis for additional mtDNA mutations at later times.

     

    image

     

    Another level of Aβ-induced mitochondrial dysfunction relates to a role for specific mitochondrial molecular targets of the amyloid peptide. In this context, our group has identified ABAD, a member of the short-chain dehydrogenase reductase family present in mitochondrial matrix, as a molecule capable of binding Aβ (5). ABAD-Aβ complex appears to promote mitochondrial generation of ROS and dysfunction. Presumably, ABAD-Aβ-induced mitochondrial dysfunction could promote mutagenesis of mtDNA, thereby leading to further mitochondrial dysfunction.

    In summary, the paper by Coskun and colleagues highlights an increasingly recognized potential role of mitochondrial dysfunction in the pathogenesis of AD. It provides testable hypotheses to further investigate the contribution of mutations in the mtDNA CR to neuronal dysfunction.

    References:

    Silverman DH, Cummings JL, Small GW, Gambhir SS, Chen W, Czernin J, Phelps ME. Added clinical benefit of incorporating 2-deoxy-2-[18F]fluoro-D-glucose with positron emission tomography into the clinical evaluation of patients with cognitive impairment. Mol Imaging Biol. 2002 Jul;4(4):283-93. PubMed.

    Bosetti F, Brizzi F, Barogi S, Mancuso M, Siciliano G, Tendi EA, Murri L, Rapoport SI, Solaini G. Cytochrome c oxidase and mitochondrial F1F0-ATPase (ATP synthase) activities in platelets and brain from patients with Alzheimer's disease. Neurobiol Aging. 2002 May-Jun;23(3):371-6. PubMed.

    Butow RA, Avadhani NG. Mitochondrial signaling: the retrograde response. Mol Cell. 2004 Apr 9;14(1):1-15. PubMed.

    Ricci JE, Muñoz-Pinedo C, Fitzgerald P, Bailly-Maitre B, Perkins GA, Yadava N, Scheffler IE, Ellisman MH, Green DR. Disruption of mitochondrial function during apoptosis is mediated by caspase cleavage of the p75 subunit of complex I of the electron transport chain. Cell. 2004 Jun 11;117(6):773-86. PubMed.

    Lustbader JW, Cirilli M, Lin C, Xu HW, Takuma K, Wang N, Caspersen C, Chen X, Pollak S, Chaney M, Trinchese F, Liu S, Gunn-Moore F, Lue LF, Walker DG, Kuppusamy P, Zewier ZL, Arancio O, Stern D, Yan SS, Wu H. ABAD directly links Abeta to mitochondrial toxicity in Alzheimer's disease. Science. 2004 Apr 16;304(5669):448-52. PubMed.

    View all comments by David Stern

  3. Etiology of late-onset Alzheimer’s disease (AD) is unknown; the paper of Coskun et al. [1] lends further support to an exciting hypothesis that it may be caused by somatic mutations in mtDNA. Mitochondria have long been known as a major source of reactive oxygen species (ROS) in actively metabolic cells; AD appears to most strongly affect brain regions with the highest metabolic rate and highest expression of mitochondrial enzymes. Thus, defects in mitochondrial oxidative phosphorylation and increased generation of ROS by defective mitochondria might mediate the relationship between late-onset AD and mtDNA mutations.

    At a first glance, this mechanism is only relevant to the development of late-onset AD and seems to be unrelated to early onset forms of the disease which are caused by point mutations in APP and presenilins; such mutations result in accelerated production of amyloid-β (Aβ). Thus, the question of major importance remains, how can we combine late-onset and early onset AD within such mitochondrial hypothesis? Increased production of Aβ is a probable answer.

    Several years ago, we showed that at low-nanomolar concentrations (i.e., those circulating in CSF and plasma), Aβ is monomeric and functions as an antioxidant [2, 3]. Mechanistically, the antioxidative activity of Aβ is related to its strong capacity to bind transition metal ions [4]; Aβ may therefore function as a metal-chelating, preventive antioxidant under normal physiologic conditions. This conclusion is consistent with results of other studies demonstrating that at low-nanomolar concentrations, Aβ has beneficial effects on neuron survival, axonal length, and neurite outgrowth (see [5, 6] for review).

    Various stress conditions are known to increase Aβ production. Importantly, generation of Aβ increases under oxidative stress induced by different mechanisms (H2O2, UV irradiation, etc.) [5, 6]. Available data strongly suggest that Aβ behaves as a positive acute-phase reactant; antioxidant metal-chelating properties of Aβ may provide a rationale for this phenomenon. Indeed, an increase in Aβ production may be aimed at chelating potentially harmful transition metal ions which can be released, e.g., from metal-binding proteins, during abnormal cellular metabolism and otherwise catalyze adverse oxidation of biomolecules. This mechanism implies that transition metal ions become abnormally sequestrated and need to be chelated in a redox-inactive form by Aβ; indeed, metabolism of transition metals is heavily impaired in AD brains [7]. Oxidative damage to neurons is one of the earliest pathological events in AD [8]; accelerated lipid peroxidation precedes accumulation of Aβ in AD transgenic mice [9]. These data suggest that increased ROS production by defective mitochondria in AD brains may lead to increased generation of Aβ as a compensatory protective response.

    Historically, Aβ has been long considered as a key pro-oxidant in AD [10]. However, to accelerate oxidation, Aβ must be present in concentrations greatly exceeding those normally measured in biological fluids (i.e., micromolar vs. nanomolar; see [2, 5]). In addition, Aβ must be aggregated to fibrils by transition metals; fibrillated Aβ is highly toxic for neurons and other cells [11]. We have therefore hypothesized that Aβ may become a pro-oxidant from an antioxidant, if its concentration increases enough to induce its substantial aggregation and if transition metal ions are available to catalyze this process [5]. This is exactly what Coskun et al. propose in their excellent paper [1]; unfortunately, it does not contain a reference to our work. Furthermore, we have postulated that increased production of Aβ as a result of elevated production of ROS by defective mitochondria with subsequent chelation of transition metal ions, accumulation of toxic Aβ-metal complexes, production of ROS, and neurotoxicity form the temporal sequence of events in the development of late-onset AD [5].

    According to this view, development of AD equally occurs along the pathway of increased Aβ production in early onset forms of the disease; the major difference between the early and late-onset forms lies in the elevated rate of Aβ production in early onset AD due to the presence of genetic defects in APP and presenilins, which should accelerate formation of toxic Aβ-metal complexes and the development of the disease. Targeting a pathologically increased formation of such complexes (i.e. by metal chelators [12]) therefore remains a promising therapeutic strategy in both early and late-onset AD; since Aβ appears to represent an important protective molecule, great caution must be however exercised about this approach, which may not target monomeric Aβ (cf. studies of AD vaccine).

    References:

    Coskun PE, Beal MF, Wallace DC. Alzheimer's brains harbor somatic mtDNA control-region mutations that suppress mitochondrial transcription and replication. Proc Natl Acad Sci U S A. 2004 Jul 20;101(29):10726-31. PubMed.

    Kontush A, Berndt C, Weber W, Akopyan V, Arlt S, Schippling S, Beisiegel U. Amyloid-beta is an antioxidant for lipoproteins in cerebrospinal fluid and plasma. Free Radic Biol Med. 2001 Jan 1;30(1):119-28. PubMed.

    Kontush A, Atwood CS. Amyloid-beta: phylogenesis of a chameleon. Brain Res Brain Res Rev. 2004 Aug;46(1):118-20. PubMed.

    Atwood CS, Scarpa RC, Huang X, Moir RD, Jones WD, Fairlie DP, Tanzi RE, Bush AI. Characterization of copper interactions with alzheimer amyloid beta peptides: identification of an attomolar-affinity copper binding site on amyloid beta1-42. J Neurochem. 2000 Sep;75(3):1219-33. PubMed.

    Kontush A. Amyloid-beta: an antioxidant that becomes a pro-oxidant and critically contributes to Alzheimer's disease. Free Radic Biol Med. 2001 Nov 1;31(9):1120-31. PubMed.

    Atwood CS, Obrenovich ME, Liu T, Chan H, Perry G, Smith MA, Martins RN. Amyloid-beta: a chameleon walking in two worlds: a review of the trophic and toxic properties of amyloid-beta. Brain Res Brain Res Rev. 2003 Sep;43(1):1-16. PubMed.

    Bush AI. The metallobiology of Alzheimer's disease. Trends Neurosci. 2003 Apr;26(4):207-14. PubMed.

    Chen Z, Zhong C. Oxidative stress in Alzheimer's disease. Neurosci Bull. 2014 Apr;30(2):271-81. Epub 2014 Mar 24 PubMed.

    Praticò D, Uryu K, Leight S, Trojanoswki JQ, Lee VM. Increased lipid peroxidation precedes amyloid plaque formation in an animal model of Alzheimer amyloidosis. J Neurosci. 2001 Jun 15;21(12):4183-7. PubMed.

    Markesbery WR. Oxidative stress hypothesis in Alzheimer's disease. Free Radic Biol Med. 1997;23(1):134-47. PubMed.

    Iversen LL, Mortishire-Smith RJ, Pollack SJ, Shearman MS. The toxicity in vitro of beta-amyloid protein. Biochem J. 1995 Oct 1;311 ( Pt 1):1-16. PubMed.

    Ritchie CW, Bush AI, Mackinnon A, Macfarlane S, Mastwyk M, MacGregor L, Kiers L, Cherny R, Li QX, Tammer A, Carrington D, Mavros C, Volitakis I, Xilinas M, Ames D, Davis S, Beyreuther K, Tanzi RE, Masters CL. Metal-protein attenuation with iodochlorhydroxyquin (clioquinol) targeting Abeta amyloid deposition and toxicity in Alzheimer disease: a pilot phase 2 clinical trial. Arch Neurol. 2003 Dec;60(12):1685-91. PubMed.

    View all comments by Anatol Kontush

  4. Mitochondria and Alzheimer’s Disease: A Complex Interrelationship
    Recently, Coskun and colleagues (2004) reported that brains from patients with Alzheimer's disease (AD) present somatic mtDNA control-region mutations, especially in individuals older than 80 years, which supports the mitochondrial hypothesis for AD pathophysiology. However, the doubt remains whether these somatic mtDNA mutations are the cause or consequence of AD pathophysiology. If somatic mutations of mtDNA are a cause of AD, these mutations should appear in the vulnerable brain regions affected by the disease. However, this same pattern will be also observed if the somatic mutation of mtDNA is a consequence of AD pathophysiology. So, how and when do the somatic mutations of mtDNA accumulate in specific regions of brain?

    A recent study suggests that DNA damage, recognized by the formation of 8-hydroxyguanosine (8OHG), a marker of nucleic acid oxidation, is markedly increased in the promoters of genes whose expression is decreased in the aged human cortex (Lu et al., 2004). Since modifications of nucleic acid bases cause mutations, this study also suggests that oxidative stress can cause mutations in gene promoters even in aged control brains.

    Coskun et al. (2004) also observed that the number of mtDNA is decreased in brains of AD patients. It is known that the nucleus regulates the replication of mitochondria according to the metabolic needs of the cell. When cells possess a large amount of mitochondria harboring damaged mtDNA, they cannot function properly to maintain normal energetic metabolism. In that case, dysfunctional mitochondria are responsible for an increased leakage of free electrons that causes the production of harmful reactive oxygen species (ROS). In such situations, the suppression of the replication of abnormal mitochondria seems to be effective against excess ROS production.

    A previous study from our laboratory showed the opposite results with mtDNA (Hirai et al., 2001). Using in-situ hybridization, we observed that brains from AD patients present a striking increase in mtDNA. When we analyzed mtDNA changes by PCR, we found only small differences between AD and age-matched controls, even in cases where the in-situ hybridization technique showed a fourfold mtDNA increase. Judging from our observations that damaged mtDNA is restricted to neurons vulnerable in AD, we should be cautious in the interpretation of the PCR results, since they include all populations of cells that reside in the brain and may include highly damaged DNA.

    Although controversial findings exist, this study emphasizes the importance of the mitochondrial hypothesis in AD pathophysiology. More studies must be done to clarify the role and the place of mitochondria in the complex scenario of AD.

    References:

    Coskun PE, Beal MF, Wallace DC. Alzheimer's brains harbor somatic mtDNA control-region mutations that suppress mitochondrial transcription and replication. Proc Natl Acad Sci U S A. 2004 Jul 20;101(29):10726-31. PubMed.

    Hirai K, Aliev G, Nunomura A, Fujioka H, Russell RL, Atwood CS, Johnson AB, Kress Y, Vinters HV, Tabaton M, Shimohama S, Cash AD, Siedlak SL, Harris PL, Jones PK, Petersen RB, Perry G, Smith MA. Mitochondrial abnormalities in Alzheimer's disease. J Neurosci. 2001 May 1;21(9):3017-23. PubMed.

    Lu T, Pan Y, Kao SY, Li C, Kohane I, Chan J, Yankner BA. Gene regulation and DNA damage in the ageing human brain. Nature. 2004 Jun 24;429(6994):883-91. PubMed.

    View all comments by Kazuhiro Honda

Make a Comment

To make a comment you must login or register.

References

News Citations

  1. After 40, DNA Damage Accrues in Genes, Hampering Expression

Further Reading

Primary Papers

  1. Coskun PE, Beal MF, Wallace DC. Alzheimer's brains harbor somatic mtDNA control-region mutations that suppress mitochondrial transcription and replication. Proc Natl Acad Sci U S A. 2004 Jul 20;101(29):10726-31. PubMed.

Annotate

To make an annotation you must Login or Register.