Mutations in PTF1A cause pancreatic andcerebellar agenesisGabrielle S Sellick1, Karen T Barker1, Irene Stolte-Dijkstra2, Christina Fleischmann1, Richard J Coleman1,Christine Garrett3, Anna L Gloyn4, Emma L Edghill4, Andrew T Hattersley4, Peter K Wellauer5,Graham Goodwin6 & Richard S Houlston1

Individuals with permanent neonatal diabetes mellitus usuallypresent within the first three months of life and require insulintreatment1,2. We recently identified a locus on chromosome10p13–p12.1 involved in permanent neonatal diabetes mellitusassociated with pancreatic and cerebellar agenesis in agenome-wide linkage search of a consanguineous Pakistanifamily3. Here we report the further linkage analysis of thisfamily and a second family of Northern European descentsegregating an identical phenotype. Positional cloningidentified the mutations 705insG and C886T in the genePTF1A, encoding pancreas transcription factor 1a, as disease-causing sequence changes. Both mutations cause truncationof the expressed PTF1A protein C-terminal to thebasic-helix-loop-helix domain. Reporter-gene studies using aminimal PTF1A deletion mutant indicate that the deletedregion defines a new domain that is crucial for the function ofthis protein. PTF1A is known to have a role in mammalianpancreatic development4,5, and the clinical phenotype of theaffected individuals implicated the protein as a key regulator ofcerebellar neurogenesis. The essential role of PTF1A in normalcerebellar development was confirmed by detailedneuropathological analysis of Ptf1a�/�mice.

Linkage to our previously identified locus on 10p13–p12.1 associatedwith permanent neonatal diabetes mellitus (PNDM) defined by thesingle-nucleotide polymorphism (SNP) markers rs1398431 andrs2149639 in the Pakistani family (Family 1; Fig. 1a) was confirmedin both families using microsatellite markers (Fig. 1a,b). Marker dataallowed further positional refinement of the linked region to the 7.7-Mb interval between rs1398431 and D10S586. The combined lod scorefor both families between markers D10S548 and D10S586 was 4.4.

We selected three genes mapping to the region of linkage, PIP5K2A,PTF1A and CACNB2, for candidate gene screening because of theirexpression in human or mouse pancreatic and cerebellar tissue and

implied biological function. We sequenced coding regions includingall exon-intron boundaries of each gene using genomic DNA from allavailable members of both families. In addition to a commonpolymorphism detected in family 1 (787T-C, dbSNP identifierrs7918487, control C allele frequency 0.45), which did not segregatewith the disease phenotype, we identified two different mutations inPTF1A, both of which resulted in truncation of the expressed protein(Fig. 2a). In family 1, affected individuals (V:4, V:8 and V:9; Fig. 1a)were homozygous with respect to a 886C-T transition in exon 2causing a nonsense mutation (R296X). In family 2, the affectedindividual (V:1; Fig. 2a) was homozygous with respect to a single–base pair insertion (705insG) in exon 1, which caused a frameshiftmutation (P236fsX270) and truncation of the PTF1A protein at codon270 (Fig. 2a,b). The mutations segregated with the disease phenotypeand carrier status in both families. Neither mutation was present in190 alleles from healthy, unrelated individuals or had been previouslydocumented by either dbSNP or the Celera Human SNP databases.

The 705insG and 886C-T mutations in PTF1A cause truncation ofthe C-terminal 32 amino acids of the expressed protein, which is afunctional domain a priori. To examine the effect of the deletion ontranscriptional activity, we compared the abilities of full-length andtruncated PTF1A (PTF1ADC) to activate a luciferase reporter con-struct containing Ptf1-E47 recognition motifs.

The availability of a mouse Ptf1a clone and literature describingthe binding motif recognized by murine Ptf1 led us to work withthe mouse clone, constructing a truncated version at the positionequivalent to the most C-terminal deletion in the human sequence.The Ptf1-E47 recognition motif of the rat chymotrypsin gene com-prises two conserved sequences: an A box that is unique to Ptf1binding sites6 and an E box recognized by ordinary basic helix-loop-helix (bHLH) proteins7. E47 is a protein that heterodimerizes withbHLH proteins to stimulate transcription from E box–containingpromoters. Using a promoter containing four repeats of the A/Emotif of the rat chymotrypsinogen B gene, Obata et al.8 showed that

Published online 14 November 2004; doi:10.1038/ng1475

1Section of Cancer Genetics, Institute of Cancer Research, Surrey SM2 5NG, UK. 2Department of Clinical Genetics, University Hospital Groningen, Groningen 9700RB, The Netherlands. 3North West Thames Regional Genetics Service, Kennedy-Galton Centre, North West London Hospitals NHS Trust, Harrow HA1 3UJ, UK.4Diabetes and Vascular Medicine, Peninsula Medical School, Exeter EX2 5AX, UK. 5Swiss Institute for Experimental Cancer Research (ISREC), CH-1066 Epalinges,Switzerland. 6Section of Molecular Carcinogenesis, Institute of Cancer Research, Surrey SM2 5NG, UK. Correspondence should be addressed to R.S.H.(Richard.Houlston@icr.ac.uk).

NATURE GENETICS VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 1 30 1

L E T T E R S©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

coexpression of PTF1A and E47 acts synergistically to transactivate thepromoter in a luciferase reporter assay.

Transfection of Cos-7 cells with the E47, PTF1A and PTF1ADCconstructs alone produced very low activation of the A/E-luciferasereporter pG5luc A/E (Fig. 3a). Transfection with the full-length PTF1Aconstruct in combination with E47 mediated strong transactivation ofthe pG5luc A/E reporter (Fig. 3a). In contrast, the truncated PTF1ADCin combination with E47 produced no detect-able transactivation of the pG5luc A/E repor-ter (Fig. 3a). Western-blot analysis of thetransfected cells showed that the two PTF1Aconstructs had equivalent protein expression(Fig. 3b). These results show that the C-terminal 32 amino acids of PTF1A are neces-sary to enable transcription by PTF1A incombination with E47 from the Ptf1 motif.

The C-terminal 80 amino acids of thePTF1A orthologs comprise highly conserveddomains, which are completely absent in onehuman Ptf1 deletion mutant (P236fsX270)and partially deleted in the PTF1ADC mutant(Fig. 2b). The C-terminal 32 amino acids ofPTF1A could therefore be part of a structured80–amino acid domain that interacts directlyor indirectly with E47. Alternatively, thisregion may contain a smaller subdomainthat interacts directly or indirectly with E47on its own.

PTF1A, a bHLH transcription factor, is the48-kDa DNA-binding subunit of the trimericpancreas transcription factor-1 (PTF1) pro-tein complex, which is expressed in adultpancreatic exocrine cells where it acts as animportant transcriptional activator7,9. PTF1Ais also essential in the specification of pan-creatic endocrine, exocrine and duct cells inmammals4,5,8,10. Studies in mice have shownthat Ptf1a is expressed in the ventral anddorsal pancreatic buds in embryos at embryo-nic day (E) 9 to E10 (ref. 8). Similarly, in thezebrafish embryo, PTF1A is expressed in thedeveloping pancreas10. Knockout studies inmice have also shown that the Ptf1a�/� nullphenotype is postnatally lethal (average sur-vival 2–3 h) with associated below-averagebirth weight and complete failure of normalpancreatic development4, a phenotype closelyparalleling that observed in the affectedindividuals from our families. Detailed post-mortem macroscopic and microscopic exam-ination of individual V:1 from family 2 foundno sign of pancreatic tissue. Moreover, at thepapilla, there was only a ductus choledochuswith no sign of endocrine or exocrine pan-creatic tissue present surrounding the papilla.Circulating insulin and C-peptide levels weremeasurable, albeit at very low levels, in allaffected individuals from both families.Despite exhaustive detailed pathologicalexamination of V:1 from family 2, we couldnot determine the source of the insulin and

C-peptide. The presence of detectable insulin and C-peptide in theabsence of normal pancreatic tissue is suggestive of the existence of apopulation of aberrant pancreatic cells in affected individuals. Thispostulate is supported by PTF1A knockout experiments and sensitivecell lineage–marking studies in mice. These have shown that a smallnumber of pancreatic progenitor cells differentiate in the absence ofPTF1A expression into the four endocrine cell types, misallocating to

PNDM & cerebellar agenesis/hypoplasia

Non-insulin dependent diabetes mellitus

Gestational diabetes mellitus

Family 1

Family 2

1 2I

II

III

IV

V

I

II

III

IV

V

1 2 3 4 5 6

1 2 3 4 5 6 7 8

151413121110987654321D10S548D10S1423D10S595D10S211D10S245D10S586

D10S548D10S1423D10S595D10S211D10S245D10S586

D10S548D10S1423D10S595D10S211D10S245D10S586

D10S548D10S1423D10S595D10S211D10S245D10S586

193230189196166124

193230189196166124

124170198 198191

191 191

191226

124170198191

191226

230

166120

198

191

191226

170124

198

191

191230

166124

198

191

191230

166120

208

191

189234

166124

194

187

183226

160124

193 193230 230189 189196 196166 166124

193230191198166120

124

193230189196166124

193230189196166124

193230189196166124

187226183194160124

193230189196166124

193230189196166124

193230189196166124

193230189196166124

193230189196166124

193230189196166124

193230189196166124

193230189196166124

187226183194160124

1 2 3 4 5 6 7 8 9

191218193198162124

191218193198162124

191218193198162124

189226191194164124

189226191194164124

189226191194164124

191218193198162124

191218193198162124

191222197200164124

189226191194164124

191218193198162124

191218193198162124

1 2 3 4 5 6

2

2

1

1

21

21 3 4

a

b

Figure 1 Pedigrees showing microsatellite marker genotypes in the linked region on chromosome

10p13–p12.1 (a,b). The marker order is telomere, D10S548, D10S586, centromere. The black region

shows the disease-associated haplotype in each family. (a) Family 1. (b) Family 2.

1 30 2 VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 NATURE GENETICS

L E T T E R S©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

splenic mesenchyme and persisting in a functional state until thepostnatal death4,5.

PNDM is both phenotypically and genetically heterogeneous. Theprincipal genetic causes identified to date are homozygous missensemutations in the gene encoding glucokinase (GCK) and activatingmutations in the gene encoding the Kir6.2 subunit of the ATP-sensitive potassium pancreatic beta cell channel (KCNJ11)11–13. Col-lectively, these mutations probably account for up to 50% of PNDMcases. It is possible that allelic variants of PTF1A may cause PNDMoutside the context of the phenotype in the two families we studied.To explore this possibility, we sequenced DNA from 15 probands withPNDM that did not have germline mutations in KCNJ11 or GCK.We identified only two previously documented polymorphisms, a

787T-C transition in 11 of the individuals (C allele frequency 0.56 inPNDM cases and 0.45 in controls) and a G-A SNP in the 3¢untranslated region of the gene (dbSNP identifier rs10828415)found in two individuals (A allele frequency 0.06 in PNDM casesand 0.02 in controls; dbSNP control frequency 0.07). Therefore,germline mutations in PTF1A probably cause PNDM only in thecontext of a restricted phenotype.

PTF1A is expressed in the developing neural tissue of mouseembryos at E9–E10, with expression extending to the cerebellumprimordia later in embryonic development (E12)8. Similarly, in thezebrafish embryo, PTF1A is expressed in the midbrain-hindbrainboundary, which corresponds to the prospective cerebellum10. Post-mortem examination of the brain from individual V:1 (family 2)

Family 1

Family 2

886C → T

705insG

P236fsX270

Mutated allele

Mutated allele

Wild-type

Wild-type

R296X

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

Human PTF1AMouse PTF1ARat PTF1AZebrafish PTF1AX. laevis PTF1A

328324326265121

91233295293296

R296X

bHLH

40182235233236

P236fsX270

128175173176

79118118120

54606060a b

Figure 2 Identification of mutations in PTF1A as the cause of autosomal recessively inherited PNDM. (a) Mutations in PTF1A identified in the two probands.

Sequence analysis of PTF1A identified a homozygous 886C-T transition that causes the amino acid substitution R296X in family 1 and a single-nucleotide

insertion 705insG that causes a frameshift (P236fsX270) and premature protein truncation in family 2. (b) Multiple protein alignment of human PTF1A

with orthologs of mouse, rat, zebrafish and the frog, X. laevis. Black shading shows fully conserved amino acids; gray shading indicates conservation of

residues in at least four of the species. The human protein sequence was used as the basis for amino acid numbering beginning from the initiating

methionine. Above sequences, the bHLH functional domain is shown as a horizontal bar. Homozygous amino acid changes in individuals with PNDM are

indicated by black triangles.

0

10

20

30

40

50

60

70

Rel

ativ

e ac

tiva

tio

n

PTF1APTF1A∆C

E47

50 kDa

FLAG

β-actin

PT

F1A

∆CP

TF

1A∆C

PT

F1A

PT

F1A

36 kDa

64 kDa

50 kDa

−−

−−

−

−−

−

+

+

+

+

+

++

a bFigure 3 Transcriptional effects of deletion of the C-terminal 32 amino

acids of PTF1A. (a) Cos-7 cells were transfected with pg5luc A/E; pSV40

Renilla luciferase; and, where indicated, PTF1A, PTF1ADC, E47 and

pFLAG-CMV2 vector without insert. Luciferase results were normalized with

respect to Renilla readings. Relative activation of the pG5luc A/E reporter by

protein expressors was determined by comparing luciferase readings with

those obtained with empty pFLAG-CMV2 vector or pFLAG-CMV2 plus E47

where appropriate. Values are means of six replicates. Equivalent results

were obtained by three independent determinations. (b) Equal expression

of full-length (PTF1A) and truncated (PTF1ADC) products by western-blotanalysis. The upper panel shows two independent transfections of either

full-length or truncated protein detected using an antibody to the N-terminal

FLAG tag. The lower panel shows equal protein loading using an antibody

to b-actin.

NATURE GENETICS VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 1 30 3

L E T T E R S©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

confirmed the absence of cerebellar tissue documented on computedtomography brain scans from affected individuals in both families. Wetherefore hypothesized that loss of function of the expressed PTF1Aprotein probably affects development of the cerebellum. This led us toexamine development of this organ in Ptf1a�/�mice. Serial sectioningof fixed whole brains from null homozygotes confirmed that Ptf1a hasa crucial role in neurogenesis, as the cerebellar primordium wassubstantially reduced in size in E16.5 embryos (Fig. 4a), leading tocerebellar agenesis at birth (Fig. 4b). Our observations indicate thatPTF1A is vital for normal cerebellar neurogenesis as well as pancreaticdevelopment in mammals and that germline mutations in PTF1A maytherefore be involved in the etiology of cerebellar developmentalanomalies outside the context of PNDM.

METHODSSubjects. The study was carried out with approval from the University Hospital

Groningen, North West Thames Regional Genetics Service and Peninsula

Medical School, and informed consent was obtained from all subjects. Affected

individuals in family 1 (Fig. 1a) were children of consanguineous Pakastani

parents with autosomally inherited neonatal diabetes and cerebellar agenesis/

hypoplasia, which has been previously described in detail14. Individual V:1

(family 2), originating from a Northern European consanguineous mating

between first cousins once removed (Fig. 1b), had an identical phenotype of

severe intrauterine growth retardation, flexion contractures of arms and legs,

very little subcutaneous fat and optic nerve hypoplasia (very small and pale

discs). Very low levels of circulating insulin and C-peptide were detected in the

presence of severe hyperglycemia. A computed tomography scan of the brain

showed cerebellar and vermis agenesis. The karyotype was normal. He died at

the age of 6 weeks. At autopsy there was no pancreas present, and detailed

macroscopic and microscopic examination failed to detect any pancreatic tissue

in the abdominal cavity. Post-mortem examination of the brain showed total

absence of the cerebellum. The roof of the distended fourth ventricle was

membranous.

The fifteen subjects with PNDM were all diagnosed before 6 months of age

(median age at diagnosis was 4 weeks, range 0–20 weeks); six were diagnosed

before 1 week of age and nine before 3 months of age. All remained diabetic

and were treated with insulin at the time of their latest examination, at least

1 year after diagnosis (median 8 years, range 1–27 years). None of these subjects

had cerebellar aplasia, but three had related parents (two marriages between

first cousins and one between second cousins). Mutations in KCNJ11 and GCK

were not detected by direct sequencing.

Linkage analysis. A previous genome-wide linkage screen of family 1 using a

first-generation SNP array (GeneChip Mapping 10K Xba Array, Affymetrix)

identified a shared region of homozygosity on chromosome 10p13–p12.1 (ref.

3). We identified six microsatellite markers (D10S548, D10S1423, D10S595,

D10S211, D10S245 and D10S586) mapping to the 8.5-Mb region of homo-

zygosity by querying the July 2003 freeze of the University of California at Santa

Cruz human genome database. We amplified microsatellite markers (Invitro-

gen) by PCR in all available family members from both families using genomic

DNA isolated using standard procedures. We analyzed PCR products on an ABI

Prism 3100 genetic analyzer and determined allele sizes using a standard

software package (Genotyper Analysis) with size standards provided by the

manufacturer (Applied Biosystems). Information on primer sequences, allele

size ranges and suggested amplification conditions are available on request.

We estimated multipoint lod scores using the HOMOZ/MAPMAKER

program assuming equal allele frequencies for each maker (eight alleles per

marker) and a population frequency of the disease-associated allele of 0.0001

(ref. 15). Variation in frequency of the disease-associated allele did not

significantly alter the lod score analysis.

Gene analysis. We identified candidate genes in the region of shared homo-

zygosity (defined by rs1398431 and D10S586) using the National Center for

Biotechnology Information and University of California at Santa Cruz human

genome databases. We selected genes for mutational analysis on the basis of

biological function and expression in pancreatic and cerebellar tissue by

Unigene clusters in humans, mice or rats (National Center for Biotechnology

Information). We designed PCR primers using standard software (Primer 3)

+/– –/– +/– –/–

1 mm 1 mm

1 mm

0.25 mm

0.5 mm

Cep

Cep Chp

Chp

MbMb

Mb Mb

MbMb

MoMo

CxCx

Mo MoCe Cx Cx

Mo

Mo

Chp

Chp

Ce

a b

Figure 4 Cerebellar development is abnormal in the Ptf1a-deficient mouse. (a) Giemsa-stained sagittal sections from the cranial mid-plane of Ptf1a+/� and

Ptf1a�/� E16.5 embryos show that a cerebellar primordium of reduced size forms initially in the mutant. (b) Microscopic examination of whole brains or

Giemsa-stained sagittal sections from the brain mid-plane show cerebellar aplasia (open arrows) in Ptf1a�/� mice at birth. Rostral is toward the right. Ce,

cerebellum; Cep, cerebellar primordium; Chp, choroid plexus; Cx, cortex; Mb, midbrain; Mo, medulla oblongata.

1 30 4 VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 NATURE GENETICS

L E T T E R S©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s

to amplify across all intronic-exonic boundaries in CACNB2, PIP5K2A

and PTF1A. PCR primers and conditions are available on request. We

purified amplified PCR products (Qiagen) and sequenced them bidirectionally

using BigDye Terminator chemistry implemented on an ABI 3100 sequencer

(Applied Biosystems). We aligned sequences and compared them with con-

sensus sequences obtained from the human databases using the software

package Sequencher (Version 3.1.1, Gene Codes). We isolated control DNA

from 95 individuals of Northern European descent with no medical history of

chronic illness.

Bioinformatics. We identified orthologs of human PTF1A for zebrafish, the

frog, Xenopus laevis, mouse and rat from BLASTP comparison with GenBank

(National Center for Biotechnology Information). We aligned protein

sequences using ClustalW (EMBL). We predicted protein motifs and domains

using the program Pfam.

Plasmids. We constructed PTF1A expression vectors, both full length (PTF1A)

and truncated (PTF1ADC), by PCR amplification using a murine PTF1-p48

cDNA clone. We used a 5¢ primer to amplify both PTF1A and PTF1ADC

sequences. The 3¢ primer for PTF1A used the natural stop codon. The 3¢ primer

for PTF1ADC included a stop codon that terminates the protein at the

equivalent position to mutation R296X found in family 1. Primer sequences

are available on request.

We cloned the products into pFLAG-CMV2 (Sigma) in-frame with the

vector-encoded FLAG tag to allow determination of protein expression. The

pcDNA3-based expression vector encoding Syrian hamster E47 with a

N-terminal myc tag has been previously described16. We constructed a

luciferase reporter plasmid (pG5luc A/E) containing four A/E box repeats of

the Ptf1 motif of the rat chymotrypsin B gene by PCR amplification from

annealed synthetic oligonucleotides and cloning into the EcoR1 and Kpn1 sites

of pG5luc (Promega). The A/E box repeat sequence was based on that

previously described by Obata et al.8. We used pG5luc as a control reporter

and showed that none of the expressors activated this minimal reporter.

Transient transfections and reporter assays. We cultured Cos-7 cells (1.2 �105 cells per well) in six-well plates in Dulbecco’s modified Eagle medium

supplemented with 10% fetal bovine serum at 37 1C in 5% CO2. After 24 h, we

transfected cells, using Polyfect (Qiagen), with 150 ng of reporter plasmid;

75 ng of E47, PTF1A, PTF1ADC and pFLAG-CMV2 (empty vector) in

combinations as shown; and 75 ng of pSV40 Renilla luciferase reporter gene,

to normalize the results for transfection efficiency. We lysed cells 48 h after

transfection and carried out luciferase assays with a Dual-Luciferase Reporter

Assay System (Promega) in accordance with the manufacturer’s instructions.

Readings were measured with a Dynex luminometer.

Western-blot analysis. To assess PTF1A and PTF1ADC protein expression in

the transient transfections, we carried out western blotting. For each transfec-

tion mix, we transfected parallel wells, one for the luciferase assay and the other

for immunoblotting. We separated proteins using standard denaturing SDS-

PAGE and transferred them to nitrocellulose using standard conditions. We

detected PTF1A and PTF1ADC expression using a FLAG M2 antibody (Sigma).

E47 expression was verified by western blotting using an antibody to myc tag

(9E10; Abcam). We used an antibody to b-actin (Abcam) as a loading control.

Histology. We examined brain morphology of three PTF1A-deficient E16.5

embryos and five newborns and compared it with that of heterozygous

littermates. The early postnatal death of mutant mice precluded analysis

beyond the newborn stage. We dissected whole brains of newborn mice under

a stereo lens using micro-scissors and fixed them in 4% paraformaldehyde. We

embedded whole E16.5 embryos or dissected brains in paraffin, made serial

sections (4 mm) as described4, stained them with aqueous 5% Giemsa (Sigma)

and mounted them in DPX (Fluka). Pictures were taken with a photomicro-

scope or stereo lens (both from Zeiss) fitted with digital cameras.

URLs. The human genome databases are available at http://genome.ucsc.

edu/ and http://www.ncbi.nlm.nih.gov/. Primer 3 is available at http://

www.broad.mit.edu/cgi-bin/primer/primer3_www.cgi/. ClustalW is available

at http://www.ebi.ac.uk/clustalw/. SNP databases are available at http://www.

celeradiscoverysystem.com/ and http://www.ncbi.nlm.nih.gov/SNP/. Pfam is

available at http://www.sanger.ac.uk/Software/Pfam/search.shtml.

Accession numbers. GenBank: human PTF1A mRNA sequence, NM_178161.

GenBank protein: human PTF1A transcript sequence, NP_835455; mouse

PTF1A transcript sequence, NP_061279; rat PTF1A transcript sequence,

NP_446416; zebrafish PTF1A transcript sequence, AAO92259; X. laevis PTF1A

transcript sequence, AAQ74877.

ACKNOWLEDGMENTSWe thank members of the families for their collaboration, K. Meyer andM. Sigvardsson for the E47 expression vector, B. Spence-Dene for the murinePtf1a cDNA clone and H. Spendlove for technical assistance. A.T.H. is aWellcome Trust research leave fellow. This work was supported by the Institute ofCancer Research and grants to R.S.H. and G.G. from Cancer Research UK, and agrant from the Swiss National Science Foundation to P.K.W.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Received 3 September; accepted 21 October 2004

Published online at http://www.nature.com/naturegenetics/

1. Gentz, J.C. & Cornblath, M. Transient diabetes of the newborn. Adv. Pediatr. 16,345–363 (1969).

2. Shield, J.P.H. et al. Neonatal diabetes: A study of its aetiopathology and genetic basis.Arch. Dis. Child Fetal Neonatal Ed. 76, F39–F42 (1997).

3. Sellick, G.S., Garrett, C. & Houlston, R.S. A novel gene for neonatal diabetes mellitusmaps to chromosome 10p12.1-p13. Diabetes 52, 2636–2638 (2003).

4. Krapp, A. et al. The bHLH protein PTF1-p48 is essential for the formation ofthe exocrine and the correct spatial organization of the endocrine pancreas. GenesDev. 12, 3752–3763 (1998).

5. Kawaguchi, Y. et al. The role of the transcriptional regulator PTF1A in convertingintestinal to pancreatic progenitors. Nat. Genet. 32, 128–134 (2002).

6. Cockell, M. et al. Identification of a cell-specific DNA-binding activity that interactswith a transcriptional activator of genes expressed in the Acinar pancreas. Mol. Cell.Biol. 9, 2464–2476 (1989).

7. Krapp, A. et al. The p48 DNA-binding subunit of transcription factor PTF1 is a newexocrine pancreas-specific basic helix-loop-helix-protein. EMBO J. 15, 4317–4329(1996).

8. Obata, J. et al. p48 subunit of mouse PTF1 binds to RBP-Jk/CBF-1, the intracellularmediator of Notch signalling, and is expressed in the neural tube of early stageembryos. Genes Cells 6, 345–360 (2001).

9. Rose, S.D. et al. The role of PTF1-P48 in pancreatic acinar gene expression. J. Biol.Chem. 276, 44018–44026 (2001).

10. Zecchin, E. et al. Evolutionary conserved role of PTF1A in the specification of exocrinepancreatic fates. Dev. Biol. 268, 174–184 (2004).

11. Njølstad, P.R. et al. Neonatal diabetes mellitus due to complete glucokinasedeficiency. N. Engl. J. Med. 344, 1588–1592 (2001).

12. Njølstad, P.R. et al. Permanent neonatal diabetes caused by glucokinase deficiency:inborn error of glucose-insulin signalling pathway. Diabetes 52, 2854–2860 (2003).

13. Gloyn, A.L. et al. Activating mutations in the gene encoding the ATP-Sensitivepotassium-channel subunit Kir6.2 and permanent neonatal diabetes. N. Engl. J.Med. 350, 1838–1849 (2004).

14. Hoveyda, N. et al. Neonatal diabetes mellitus and cerebellar hypoplasia/agenesis:report of a new recessive syndrome. J. Med. Genet. 36, 700–704 (1999).

15. Kruglyak, L. et al. Rapid multipoint linkage analysis of recessive traits in nuclearfamilies, including homozygosity mapping. Am. J. Hum. Genet. 56, 519–527 (1995).

16. Sigvardsson, M. Overlapping expression of early B-cell factor and basic helix-loop-helixproteins as a mechanism to dictate B-lineage-specific activity of the lambda5promoter. Mol. Cell. Biol. 20, 3640–3654 (2000).

NATURE GENETICS VOLUME 36 [ NUMBER 12 [ DECEMBER 2004 1 30 5

L E T T E R S©

2004

Nat

ure

Pub

lishi

ng G

roup

ht

tp://

ww

w.n

atur

e.co

m/n

atur

egen

etic

s