Mutations in progranulin cause tau-negative
frontotemporal dementia linked to chromosome 17
Matt Baker1*, Ian R. Mackenzie2*, Stuart M. Pickering-Brown5,6*, Jennifer Gass1, Rosa Rademakers1,
Caroline Lindholm3, Julie Snowden6, Jennifer Adamson1, A. Dessa Sadovnick3,4, Sara Rollinson5, Ashley Cannon1,
Emily Dwosh4, David Neary6, Stacey Melquist1, Anna Richardson6, Dennis Dickson1, Zdenek Berger1,
Jason Eriksen1, Todd Robinson1, Cynthia Zehr1, Chad A. Dickey1, Richard Crook1, Eileen McGowan1,
David Mann6, Bradley Boeve7, Howard Feldman3& Mike Hutton1
Frontotemporal dementia (FTD) is the second most common
cause of dementia in people under the age of 65 years1. A large
proportion of FTD patients (35–50%) have a family history of
dementia, consistent with a strong genetic component to the
disease2. In 1998, mutations in the gene encoding the microtu-
bule-associated protein tau (MAPT) were shown to cause familial
The neuropathology of patients with defined MAPT mutations is
characterizedbycytoplasmic neurofibrillary inclusionscomposed
of hyperphosphorylated tau3,4. However, in multiple FTD families
with significant evidence for linkage to the same region on
chromosome 17q21 (D17S1787–D17S806), mutations in MAPT
have not been found and the patients consistently lack tau-
immunoreactive inclusion pathology5–12. In contrast, these
patients have ubiquitin (ub)-immunoreactive neuronal cyto-
plasmic inclusions and characteristic lentiform ub-immuno-
demonstrate that in these families, FTD is caused by mutations
in progranulin (PGRN) that are likely to create null alleles. PGRN
is located 1.7Mb centromeric of MAPTon chromosome 17q21.31
and encodes a 68.5-kDa secreted growth factor involved in the
regulation of multiple processes including development, wound
repair and inflammation14. PGRN has also been strongly linked to
tumorigenesis14. Moreover, PGRN expression is increased in
activated microglia in many neurodegenerative diseases including
Creutzfeldt–Jakob disease, motor neuron disease and Alzheimer’s
disease15,16. Our results identify mutations in PGRN as a cause of
neurodegenerative disease and indicate the importance of PGRN
function for neuronal survival.
FTD is characterized by abnormalities in behaviour, personality
and language with relative preservation of perception and memory,
and may also be associated with motor dysfunction, including both
motor neuron disease (MND) and parkinsonism1,17. The micro-
scopic pathology of FTD varies markedly18; although some cases
based histopathology and show ub-immunoreactive neurites and
neuronal cytoplasmic inclusions (NCI). The ub-immunoreactive
inclusions are characteristically found in layer II of the frontal and
temporal neocortex and in the dentate fascia of the hippocampus
(frontotemporal lobar degeneration with ubiquitin inclusions,
FTLD-U)1,17. This same pattern of ub-immunoreactive pathology
inclusions11–13. Here we
ubiquitinated, the molecular composition of the NCI is presently
Previous genetic linkage studies in FTD families revealed a locus
on chromosome 17q2120(Fig. 1a). Subsequently, over 30 mutations
in MAPT (encoding tau) were identified that account for a pro-
portion of these cases (FTDP-17)3,4(http://www.molgen.ua.ac.be/
FTDmutations). FTDP-17 patients with MAPTmutations inevitably
develop tau neurofibrillary inclusion pathology4. However, since the
identification of MAPT mutations in 1998, nine FTD families have
been conclusively linked to chr17q21 but lack defined MAPT
lack tau inclusions but develop ub-immunoreactive pathology typical
of FTLD-U5–9,11–13. Moreover, several of these families also have ub-
immunoreactive lentiform neuronal intranuclear inclusions (NII)
with a similar distribution to the NCI (Fig. 2a–c)5,6,11–13. We pre-
viously found NII in other families—too small for genetic linkage
analysis—with tau-negative FTD however these lesions wereuncom-
mon in sporadic FTD21. These findings led us to propose that NIIare
characteristic of FTD associated with this genetic locus12,21.
The absence of obvious MAPTmutations in tau-negative FTD-17
families suggested that this form of FTD is caused either by novel
MAPT mutations that have eluded detection or by mutations in a
different gene. However, extensivesequencing of MAPTintronicand
regulatory regionsintwolinkedFTD-17families (UBC17and1083),
and analysis of soluble tau from patient brain extracts (UBC17),
Therefore, we examined other candidate genes within the 3.53-
centimorgan (6.19-Mb) critical region defined by haplotype analysis
in reported families (D17S1787–D17S806)12. The coding exons of
candidate genes were first sequenced in affected and unaffected
members of UBC17 (Table 1), a large Canadian tau-negative FTD
family with highly significant evidence for linkage to chromosome
of ,165 in the region) failed to identify a pathogenic mutation.
However, when progranulin (PGRN) was sequenced, an insertion
mutation of 4 base pairs (bp) was detected in exon 1
(c.90_91insCTGC) causing a frameshift at codon 31 that introduces
a premature termination codon after a read through of 34 residues
(C31LfsX34). PGRN is a 593-amino-acid (68.5-kDa) multifunc-
tional growth factor that is composed of seven-and-a-half tandem
1Department of Neuroscience, Mayo Clinic College of Medicine, 4500 San Pablo Road, Jacksonville, Florida 32224, USA.2Department of Pathology,3Division of Neurology, and
4Department of Medical Genetics, University of British Columbia, 2211 Wesbrook Mall, Vancouver V6T 2B5, British Columbia, Canada.5Division of Laboratory and Regenerative
Medicine, Department of Medicine, University of Manchester, Oxford Road, Manchester M13 9PT, UK.6Centre for Clinical Neurosciences, University of Manchester, Greater
Manchester Neurosciences Centre, Hope Hospital, Salford M6 8HD, UK.7Department of Neurology, Mayo Clinic College of Medicine, Rochester, Minnesota 55905, USA.
*These authors contributed equally to this work.
Vol 442|24 August 2006|doi:10.1038/nature05016
© 2006 Nature Publishing Group
repeats of a 12-cysteine granulin motif14. The mutant (C31LfsX34)
the signal peptide, and does not contain a single intact cysteinyl
repeat (Fig. 1). The C31LfsX34 mutation segregated with disease in
American control individuals.
PGRN was then sequenced in affected individuals from an
the USA (8 families), the UK (17 families), the Netherlands (1 family)
and Scandinavia (8 families). This analysis identified an additional
seven PGRN mutations in eight families, each predicted to cause
premature termination of the coding sequence (Table 1, Fig. 1c). The
mutations include four nonsense mutations (Q125X, W386X, R418X
and Q468X), two frameshift mutations (Q130SfsX124, T382SfsX29)
The latter is likely to lead to skipping of exon 8 from the PGRN
messenger RNA, resulting in a frameshift (V279GfsX4); however, this
could not be confirmed as a source of RNA was not available. The
Q130SfsX124 mutation was found in two FTD families ascertained
independently in Canada (UBC11) and the UK (F53). All seven
mutations segregated with disease in the relevant families (Table 1,
Supplementary Fig. 1) and were absent in 200 North American and
95 UK controls.
The only two FTD families with significant evidence for linkage to
17q21 (UBC1712and 10835,23) were both found to have mutations in
PGRN (Fig. 1, Supplementary Fig. 2). Notably, patients from all nine
signs of MND, and all had neuropathological findings that included
ub-immunoreactive NII, as predicted by our hypothesis that these
lesions are a characteristic feature of FTD linked to this locus12,21.
Analysis of a further series of FTD families from Belgium also
identified mutations in PGRN, including another family (DR8)
with published evidence for linkage to chromosome 17q2111,23.
To investigate the pathogenic mechanism of these truncating
mutations, we determined whether premature termination of the
PGRN coding sequence resulted in nonsense-mediated decay
(NMD) of the mutant mRNAs24. Quantitative PCR with reverse
transcription (qRT–PCR) on RNA extracted from patient lympho-
blasts carrying the R418X (UBC15) and C31LfsX34 (UBC17)
mutations showed that both were associated with a ,50% reduction
in total PGRN mRNA relative to lymphoblasts from unaffected
individuals (Fig. 3a). In addition, PGRN mRNA from both families
consisted almost entirely of wild-type mRNAwith little of the mutant
mRNAs detected (Fig. 3c, Supplementary Fig. 3b). Treatment of
patient lymphoblasts with cycloheximide—a known inhibitor of
NMD24—resulted in an increase in levels of total PGRN mRNA
(Fig. 3b) that was associated with a selective increase in the
C31LfsX34 and R418X mutant mRNAs (Fig. 3c, Supplementary
Fig. 3c). Furthermore, western blot analysis showed that wild-type
PGRN protein was reduced in extracts from R418X and C31LfsX34
lymphoblasts (mean reduction 34%, P ¼ 0.01, t-test) relative to
extracts from unaffected relatives (Fig. 3d). We were also unable to
detect the predicted mutant proteins in extracts from patient brain
tissue (not shown) and lymphoblastoid cells (Fig. 3d). These data
cause disease by creating null alleles, with the mutant mRNAs being
degraded by NMD. This results in loss of functional PGRN (hap-
between the location of each mutation and clinical phenotype, as
each mutation has the same effect—creation of a null allele.
The proposed haploinsufficiency mechanism is also consistent
Figure 2 | Immunohistochemistry in FTD with
PRGN mutations. a–c, Ubiquitin
immunohistochemistry demonstrates neurites
and neuronal cytoplasmic inclusions (NCI)
(arrows) in layer II of the frontal neocortex (a),
NCI in hippocampal dentate granule cells (b),
and neuronal intranuclear inclusions (NII) in the
superficial neocortex (c). d–f, PGRN
immunohistochemistry was positive in some
neocortical neurons (d), but did not stain NCI or
NII in layer II cortex (e). Activated microglia
(arrows) showed strong PGRN expression in
affected areas of FTD (e) and around senile
plaques in Alzheimer’s disease (f). Scale bars,
Figure 1 | Null mutations in PGRN cause tau-negative FTD linked to
chromosome 17. a, Schematic representation of chromosome 17. PGRN is
located 1.7Mb from MAPT, which is mutated in FTDP-17. b, Schematic
representation of the PGRN gene and mRNA encoding the PGRN protein,
with positions of tau-negative FTD-17 mutations indicated. Lettered boxes
refer to individual granulin repeats. c, Location of the mutation altering the
Met1 codon, and the premature termination codons created by the
truncating mutations, in PGRN mRNA. The mutant mRNAs are subject to
nonsense-mediated decay resulting in null alleles. The IVS8þ1G.A
mutation (indicated by an asterisk) is predicted to cause skipping of exon 8
NATURE|Vol 442|24 August 2006
© 2006 Nature Publishing Group
the normal Kozak sequence by altering the Met1 codon (c.2T.C,
M1?) in another tau-negative FTD case (F161) with ub-immuno-
reactive pathology and NII (Table 1, Fig. 1, Supplementary Fig. 4).
This mutation was absent in 200 North American controls and
reduction in mutant mRNA, providing evidence of a null allele
(Supplementary Fig. 4). A second mutation that alters the Met1
codon (c.3G.A) was also observed in a Belgian FTD patient series23.
Next, immunohistochemistry was performed on post-mortem
brain tissue using a panel of antibodies that recognize all regions of
the PGRN protein. Consistent with previously published data on the
distribution of PGRN15,16,25, immunoreactivity was observed in a
subset of cortical neurons and intensely in activated microglial cells,
both in patients from FTD families with PGRN mutations (UBC17,
UBC15 and F129) and in normal aged and Alzheimer’s disease
subjects (Fig. 2d–f). However, the ub-immunoreactive NCI and
NII in the FTD cases with mutations showed no staining for
PGRN (Fig. 2e). Although this finding indicates that the disease
mechanism does not cause the accumulation of PGRN in these
lesions, it leaves the identity of the ubiquitinated protein species in
NCI and NII unknown.
Although the function of PGRN in neurons has not yet been
determined14,our findingsimply thatPGRNisessentialfor neuronal
survival and even partial loss of PGRN eventually leads to neuro-
degeneration. This supports the more general hypothesis that loss of
growth factor support can cause neurodegenerative disease26,27.
PGRN is expressed in many tissues and mediates its role in develop-
ment, wound repair and inflammation by activating signalling
cascades that control cell-cycle progression and cell motility14.
These include the mitogen-activated protein kinase (MAPK) and
phosphatidylinositol-3-OH kinase (PI(3)K) cascades14,28, both of
which regulate crucial functions in neurons. PGRN also stimulates
the induction of other growth factors including vascular endothelial
growth factor (VEGF)29. It is also of interest, that although partial
loss of PGRN apparently results in adult-onset neurodegenerative
The identification of mutations in PGRN fully resolves a ten-year-
old conundrum, namely the genetic basis for FTD linkedto chromo-
some 1720, and explains why multiple families linked to this region
lack MAPTmutations. The fact that PGRN is located within 2Mb of
MAPT and mutations in both genes independently yield indistin-
guishable clinical phenotypes is presumably an extraordinarycoinci-
dence. Our findings are also highly reminiscent of the recent
identification of probable loss-of-function mutations affecting
another secreted factor—angiogenin (ANG)—in patients with
MND30. Tau-negative FTD and MND share overlapping clinical
spectrums and have similar ub-immunoreactive pathology19. Fur-
thermore, PGRNand ANG are both potent inducers of angiogenesis
and are linked with tumorigenesis14,30,31. Although their precise role
Table 1 | Families with premature termination mutations in PGRN
Family Origin Affecteds*Mean age onsetMutation (nucleotide)Mutation (protein)
*Number of affecteds with NII pathology confirmed is in parentheses.
†Families with previously published linkage to chr17.
‡Predicted effect of IVS8þ1G.A mutation.
Figure 3 | Mutant PGRN mRNAs with premature termination codons are
degraded by nonsense-mediated decay. a, qRT–PCR analysis shows a
,50% reduction in PGRN mRNA in lymphoblastoid cells from individuals
carrying the C31LfsX34 (UBC17, n ¼ 3) and R418X (UBC15, n ¼ 1)
mutations. PGRN mRNA levels are shown as a percentage of levels in cells
from unaffected individuals (n ¼ 6). Error bars indicate s.e.m. b, Treatment
of a C31LfsX34 cell line (n ¼ 1) with the NMD inhibitor cycloheximide
(CHX; 500mM) for 2h and 8h results in a progressive increase in total
PGRN mRNA. PGRN mRNA levels are mean values from three replicates.
Error bars denote s.d. c, RT–PCR fragment analysis in lymphoblastoid cells
(trace 2) and brain (trace 6) from patients with the C31LfsX34 mutation
shows that the mutant mRNA (196bp) is virtually absent. Treatment with
CHX (traces 3–5) results in the selective increase in C31LfsX34 mutant
mRNA. d, Western blot analysis of lymphoblast extracts shows that
wild-type (WT) PGRN protein is reduced in lymphoblasts from mutation
carriers (C31LfsX34, n ¼ 2 and R418X, n ¼ 1) relative to unaffected
relatives (n ¼ 4) (right panel; mean reduction 34%, P ¼ 0.01, t-test). In
not detected. Arrows indicate wild-type PGRNand the expected position of
the R418X mutant protein. R418X PGRN generated from an intronless
cDNAconstruct(not subjectto NMD) expressedin HeLacells (left panel)is
included to demonstrate that the mutant protein (if made) is stable.
NATURE|Vol 442|24 August 2006
© 2006 Nature Publishing Group
in neurons has yet to be fully established, it appears that reduced Download full-text
levels of these functionally related factors represents a common
mechanism of neurodegeneration in these two diseases. Moreover,
novel therapeutic strategy in both conditions.
See Supplementary Information for further details on the methods used in this
PGRN genetic analysis. All PGRN exons were amplified from genomic DNA by
PCR using the primers listed in Supplementary Table 1, and then sequenced
using Applied Biosystems protocol. Following initial identification, the
C31LfsX34 mutation was verified with a fluorescent fragment analysis assay,
presence in 550 North American controls. Sequence analysis of all PGRN exons
was used to screen 200 aged North American and 95 UK control individuals for
all remaining mutations. Segregation of all other mutations in relevant families
was checked by sequencing of relevant exons.
Immunohistochemical methods. Immunohistochemistry was performed on
post-mortemtissuesectionsoffrontalcortexandhippocampus from two cases of
disease, and one age-matched control individual. The primary antibodies used
recognize ubiquitin and all regions of the PGRN protein including the amino
terminus, carboxy terminus, and the full-length recombinant human PGRN
Analysis of PGRN protein and PGRN RNA. PGRN protein levels were
quantified by western blot analysis in extracts from lymphoblastoid cells from
patients and unaffected relatives using primary antibodies to the N terminus of
human PGRN, and to GAPDH for normalization. PGRN mRNA levels were
analysed in patients and unaffected relatives by qRT–PCR using SYBR green on
PGRN mRNA were normalized to 28S ribosomal RNA and divided by GAPDH
wild-type PGRN mRNA in C31LfsX34 lymphoblasts and brain tissue were
analysed by RT–PCR fragment analysis using primers spanning exons 1 and 2.
Sequence analysis of RT–PCR products was further used to analyse levels of
C31LfsX34 and R418X mutant mRNAs compared with wild-type in brain tissue
and lymphoblasts. To study the effect of NMD, lymphoblast cells were treated
with cycloheximide for 2–8h.
Received 12 May; accepted 29 June 2006.
Published online 16 July 2006.
1. Brun, A. et al. Clinical and neuropathological criteria for frontotemporal
dementia. The Lund and Manchester groups. J. Neurol. Neurosurg. Psychiatry 57,
416– -418 (1994).
Chow, T. W., Miller, B. L., Hayashi, V. N. & Geschwind, D. H. Inheritance of
frontotemporal dementia. Arch. Neurol. 56, 817– -822 (1999).
Hutton, M. et al. Association of missense and 5
with the inherited dementia FTDP-17. Nature 393, 702– -705 (1998).
Ingram, E. M. & Spillantini, M. G. Tau gene mutations: dissecting the
pathogenesis of FTDP-17. Trends Mol. Med. 8, 555– -562 (2002).
Rademakers, R. et al. Tau negative frontal lobe dementia at 17q21: significant
finemapping of the candidate region to a 4.8cM interval. Mol. Psychiatry 7,
1064– -1074 (2002).
Rosso, S. M. et al. Familial frontotemporal dementia with ubiquitin-positive
inclusions is linked to chromosome 17q21– -22. Brain 124, 1948– -1957 (2001).
Lendon, C. L. et al. Hereditary dysphasic disinhibition dementia: a
frontotemporal dementia linked to 17q21– -22. Neurology 50, 1546– -1555
Kertesz, A. et al. Familial frontotemporal dementia with ubiquitin-positive,
tau-negative inclusions. Neurology 54, 818– -827 (2000).
Froelich, S. et al. Mapping of a disease locus for familial rapidly progressive
frontotemporal dementia to chromosome 17q12– -21. Am. J. Med. Genet. 74,
380– -385 (1997).
10. Bird, T. D. et al. Chromosome 17 and hereditary dementia: linkage studies in
three non-Alzheimer families and kindreds with late-onset FAD. Neurology 48,
949– -954 (1997).
11.van der Zee, J. et al. A Belgian ancestral haplotype harbours a highly
prevalent mutation for 17q21-linked tau-negative FTLD. Brain 129, 841– -852
12. Mackenzie, I. R. et al. A family with tau-negative frontotemporal dementia and
neuronal intranuclear inclusions linked to chromosome 17. Brain 129, 853– -867
0-splice-site mutations in tau
13. Rademakers, R. et al. In IPSEN Meeting Research and Perspectives in Alzheimer’s
Disease: Genotype– -Proteotype– -Phenotype Relationships in Neurodegenerative
Diseases (ed. Cummings, J.) 117– -137 (Springer, Paris, 2005).
14. He, Z. & Bateman, A. Progranulin (granulin– -epithelin precursor, PC-cell-
derived growth factor, acrogranin) mediates tissue repair and tumorigenesis.
J. Mol. Med. 81, 600– -612 (2003).
15. Malaspina, A., Kaushik, N. & de Belleroche, J. Differential expression of 14
genes in amyotrophic lateral sclerosis spinal cord detected using gridded
cDNA arrays. J. Neurochem. 77, 132– -145 (2001).
16. Baker, C. A. & Manuelidis, L. Unique inflammatory RNA profiles of microglia
in Creutzfeldt– -Jakob disease. Proc. Natl Acad. Sci. USA 100, 675– -679
17. Neary, D., Snowden, J. S. & Mann, D. M. Classification and description of
frontotemporal dementias. Ann. NY Acad. Sci. 920, 46– -51 (2000).
18. Trojanowski, J. Q. & Dickson, D. Update on the neuropathological diagnosis
of frontotemporal dementias. J. Neuropathol. Exp. Neurol. 60, 1123– -1126
19. Mackenzie, I. R. & Feldman, H. H. Ubiquitin immunohistochemistry suggests
classic motor neuron disease, motor neuron disease with dementia, and
frontotemporal dementia of the motor neuron disease type represent a
clinicopathologic spectrum. J. Neuropathol. Exp. Neurol. 64, 730– -739 (2005).
20. Foster, N. L. et al. Frontotemporal dementia and parkinsonism linked to
chromosome 17: a consensus conference. Ann. Neurol. 41, 706– -715 (1997).
21. Mackenzie, I. R. & Feldman, H. Neuronal intranuclear inclusions distinguish
familial FTD– -MND type from sporadic cases. Acta Neuropathol. (Berl.) 105,
543– -548 (2003).
22. Cruts, M. et al. Genomic architecture of human 17q21 linked to frontotemporal
dementia uncovers a highly homologous family of low-copy repeats in the tau
region. Hum. Mol. Genet. 14, 1753– -1762 (2005).
23. Cruts, M. et al. Null mutations in progranulin cause ubiquitin-positive
frontotemporal dementia linked to chromosome 17q21. Nature advance online
publication, doi:10.1038/nature05017 (16 July 2006).
24. Maquat, L. E. Nonsense-mediated mRNA decay: splicing, translation and
mRNP dynamics. Nature Rev. Mol. Cell Biol. 5, 89– -99 (2004).
25. Daniel, R., He, Z., Carmichael, K. P., Halper, J. & Bateman, A. Cellular
localization of gene expression for progranulin. J. Histochem. Cytochem. 48,
999– -1009 (2000).
26. Capsoni, S. et al. Alzheimer-like neurodegeneration in aged antinerve growth
factor transgenic mice. Proc. Natl Acad. Sci. USA 97, 6826– -6831 (2000).
27. Salehi, A., Delcroix, J. D. & Swaab, D. F. Alzheimer’s disease and NGF signaling.
J. Neural Transm. 111, 323– -345 (2004).
28. Lu, R. & Serrero, G. Mediation of estrogen mitogenic effect in human breast
cancer MCF-7 cells by PC-cell-derived growth factor (PCDGF/granulin
precursor). Proc. Natl Acad. Sci. USA 98, 142– -147 (2001).
29. Tangkeangsirisin, W. & Serrero, G. PC cell-derived growth factor (PCDGF/GP88,
progranulin) stimulates migration, invasiveness and VEGF expression in breast
cancer cells. Carcinogenesis 25, 1587– -1592 (2004).
30. Greenway, M. J. et al. ANG mutations segregate with familial and ‘sporadic’
amyotrophic lateral sclerosis. Nature Genet. 38, 411– -413 (2006).
31. He, Z., Ong, C. H., Halper, J. & Bateman, A. Progranulin is a mediator of the
wound response. Nature Med. 9, 225– -229 (2003).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank the FTD research team at Vancouver Coastal
Health and the University of British Columbia, and particularly G. Y. R. Hsiung,
for identification and follow-up of FTD families; D. Warden, P. Whitbread and
E. King (OPTIMA project, Oxford, UK) for assisting with collection of UBC17
family samples; J. Chow (Department of Pathology, University of British
Columbia) for help in performing the PGRN immunohistochemistry; and M. Yue,
J. Gonzales (Mayo Clinic), T. de Pooter and M. Van den Broeck (University of
Antwerp) for technical support. This research was funded as part of the Mayo
Clinic ADRC grant from the National Institute on Aging (to M.H.), the Mayo
Foundation (M.H.), and the Robert and Clarice Smith Fellowship program (to
S.M.). I.R.M. and H.F. were funded by the Canadian Institutes of Health research
operating grant. S.M.P.-B. received grants from the Medical Research Council
(UK) and the Motor Neuron Disease Association. R.R. is a postdoctoral fellow of
the Fund for Scientific Research Flanders and a visiting scientist from the
Neurodegenerative Brain Diseases Group of the Department of Molecular
Genetics, VIB, University of Antwerp, Belgium. Finally, we acknowledge and
thank the families who contributed samples, as without them this study would
not have been possible.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Correspondence and requests for materials should be addressed to M.H.
(firstname.lastname@example.org) or I.R.M. (email@example.com).
NATURE|Vol 442|24 August 2006
© 2006 Nature Publishing Group