Tt

CMa*fB

Bdttmipt(tdvnfedtseaAmlowpacfiomtafic

HkT

GASTROENTEROLOGY 2006;130:532–541

he Transcription Factor Hepatocyte Nuclear Factor-6 Controlshe Development of Pancreatic Ducts in the Mouse

HRISTOPHE E. PIERREUX,* AURÉLIE V. POLL,* CAROLINE R. KEMP,‡ FRÉDÉRIC CLOTMAN,*IGUEL A. MAESTRO,§ SABINE CORDI,* JORGE FERRER,§ LUC LEYNS,‡ GUY G. ROUSSEAU,*nd FRÉDÉRIC P. LEMAIGRE*Hormone and Metabolic Research Unit, Université Catholique de Louvain and Institute of Cellular Pathology, Brussels, Belgium; ‡Laboratoryor Cell Genetics, Vrije Universiteit Brussel, Brussels, Belgium; and §Endocrinology, Hospital Clinic de Barcelona, Institut d’Investigacions

iomèdiques August Pi i Sunyer, Barcelona, Spain

bdcemair

cfidpeiaLmiitdpbt

mttOtt

np

ackground & Aims: A number of hereditary polycysticiseases are associated with formation of cysts withinhe pancreatic ducts. The cysts result from abnormalubulogenesis, but how normal pancreatic duct develop-ent is controlled remains poorly understood. Here, we

nvestigate the transcriptional mechanisms that controlancreatic duct development by addressing the role ofhe transcription factor hepatocyte nuclear factorHNF)-6. Methods: Using immunostaining, we have de-ermined the expression pattern of HNF-6 in pancreaticucts during mouse development. Hnf6 null mice atarious stages of development were studied by immu-olocalization methods to assess the morphology, dif-erentiation, and proliferation status of ductal cells. Thexpression of genes involved in hereditary polycysticiseases was determined by real-time, reverse-transcrip-ion polymerase chain reaction (RT-PCR). Results: Wehow that HNF-6 is expressed in the pancreatic ductpithelium throughout development and that, in thebsence of HNF-6, duct morphogenesis is perturbed.lthough development of the intercalated ducts is nor-al, cysts appear within the interlobular and intralobu-

ar ducts. This is associated with abnormal developmentf primary cilia at the apical pole of the duct cells andith reduced expression of a set of genes involved inolycystic diseases, namely those coding for HNF-1�nd for the cilium-associated proteins polyductin/fibro-ystin and cystin. Conclusions: We identify HNF-6 as therst transcriptional regulator of pancreatic duct devel-pment and reveal the existence of different regulatoryechanisms in distinct duct compartments. HNF-6 con-

rols a network of genes involved in cilium formationnd in hereditary polycystic diseases. Finally, HNF-6 de-ciency represents a genetically defined model of pan-reatic cystic disease.

ereditary polycystic diseases (PD) are a group ofdisorders in which patients present with cysts in

idneys and at lower frequency in liver and pancreas.1–3

he pathophysiology of cyst formation in kidneys has

een extensively studied. It has been suggested thatefects in the assembly and/or function of the primaryilium, an organelle found on the luminal face of tubulepithelia, cause the appearance of cysts.4,5 Indeed, mostutations associated with cystic kidneys in humans or in

nimal models affect genes that code for proteins local-zed to the primary cilium. These genes are commonlyeferred to as cystic disease genes.1,2,4–9

Less attention has been paid to the development ofysts in the pancreas, most likely as a result from insuf-cient knowledge about the mechanisms that governevelopment of the pancreatic ducts. In mice, when theancreas starts to form, it is composed of a pluripotentpithelium surrounded by mesenchymal tissue. This ep-thelium contains precursors of the ductal, endocrine,nd exocrine cells and later gives rise to the islets ofangerhans and to a highly branched ductal compart-ent at the extremity of which cells become organized

nto acini. The mature ductal compartment consists ofntercalated ducts that connect the acini with the in-ralobular ducts, which in turn connect with interlobularucts that drain the exocrine secretions into the mainancreatic duct. Terminally differentiated ducts are linedy epithelial cells, which secrete bicarbonate that neu-ralizes the acidic chyme from the stomach.10–12

The molecular analysis of duct development in theouse indicates that pancreatic precursor cells expressing

he transcription factor Pdx-1 enter the ductal differen-iation program around embryonic day (e) 9.5–e11.5.13

ur earlier data suggest that a subset of the Pdx-1� cellshen becomes enriched in the transcription factor hepa-ocyte nuclear factor (HNF)-1�, giving rise to the

Abbreviations used in this paper: e, embryonic day; HNF, hepatocyteuclear factor; Muc-1, mucin-1; PD, polycystic disease; pHH3, phos-hohistone H3.

© 2006 by the American Gastroenterological Association0016-5085/06/$32.00

doi:10.1053/j.gastro.2005.12.005

bmpHs

otpeHohHgHtacsmPk

fepaataHtpi

sT0aU

(tvH

taPacsscG5CA5GT5GTCP

dii4bPnnE1tao(1nCcTUbAsImGe

February 2006 HNF-6 AND PANCREATIC DUCTS IN THE MOUSE 533

ranched ductal network in which HNF-1� expression isaintained throughout adulthood, as opposed to more

eripherally located HNF-1�-negative acinar cells.NF-1� is thus a marker for the embryonic ductal

tructures of the pancreas.14

HNF-1� is the product of the maturity onset diabetesf the young type 5 (MODY5) gene. In MODY5 pa-ients, the presence of renal cysts is a common andenetrant trait (hypoplastic glomerulocystic kidney dis-ase),15 and, in mice, kidney-specific inactivation of thenf1� gene leads to a polycystic phenotype reminiscent

f human polycystic kidney disease.16 Several groupsave recently shown that the transcription factorNF-1� directly controls the expression of cystic disease

enes such as Pkhd1 and Pkd2,16–18 suggesting thatNF-1� regulates a gene cascade essential for differen-

iation of epithelial cells lining ducts, for cilia function,nd for tubulogenesis. Furthermore, the involvement ofilia in pancreatic duct formation was recently demon-trated by the appearance of pancreatic cysts in the orpkice, which are deficient in the cilium-associated proteinolaris and which also represent a model of polycysticidney disease.19,20

Our previous work has shown that the transcriptionactor HNF-6 (also called Onecut-1) regulates HNF-1�xpression in the liver and pancreas.14,21 Moreover, thehenotype of Hnf6�/� mice shows biliary cysts in the liver21

nd as yet uncharacterized ductal anomalies in the pancre-s.22 In this study, we have addressed the role of HNF-6 inhe development of the pancreatic duct epithelium by an-lyzing the phenotype of Hnf6�/� mice. We show thatNF-6 regulates the differentiation of specific segments of

he pancreatic ducts, that it is required for the formation ofrimary cilia, and that it controls a network of genesnvolved in primary cilium function.

Materials and Methods

Animals

Hnf6 and Oc2 knockout mice were obtained as de-cribed.22,23 TOPGAL transgenic mice were obtained fromhe Jackson Laboratory (Bar Harbor, MA; Stock number04623).24 They were raised in our animal facility and treatedccording to the principles of laboratory animal care of theniversity Animal Welfare Committee.

RNA Purification and Reverse-Transcription-Coupled PCR

Embryos were dissected in phosphate-buffered salinePBS), and the pancreata were isolated. Total RNA was ex-racted with the Tripure RNA Isolation reagent (Roche, Vil-oorde, Belgium) and treated with TURBO DNase I (Ambion,

untingdon, UK). Total RNA (up to 1 �g) was reverse t

ranscribed as described,21 and quantitative real-time polymer-se chain reaction (PCR) was performed with the SYBR GreenCR Core kit or master mix (Eurogentec, Liège, Belgium) onMyIQ thermal cycler (Bio-Rad, Hercules, CA).25 Threshold

ycles were transformed to copy number according to thetandard calibration curve. Absolute copy number for each mes-enger RNA (mRNA) was normalized to absolute �-actin mRNAopy number. Primer sequences were 5=-AAAGGCAACCCTGAA-ACAG-3= and 5=-GCCATGAGCTCCTCTTCTGA-3= for Cys1,=-GCTCTTCTGGAGACCAAGTCA-3= and5=-AAAGCCACCA-CACTCTTGT-3= for sec63, 5=-CCACAGAGGATGAGAAG-TGC-3= and 5=-TTTCAAGGACCGTTCGACTT-3= for Prkcsh,=-TACATCATCGCCCAGTGTGTG-3= and 5=-AGCTGCAGA-TGCCAATGATC-3= for Aqp1, 5=-TGGAGCAGGCATCCTG-ACT-3= and 5=-TTGCCTGGTGTTGTGTTGTT-3= for Aqp5,=-CAGCATGACCTCAGCTTCCT-3= and 5=-TCGTCATT-TCGCTATCCAA-3= for Slc4a4, and 5=-ACAAAACCTGGA-CCCAATG-3= and 5=-GAGCTGTCCAGGAAACTGCT-3= forftr. Primer sequences for �-actin, Hnf1�,14 Pkd1, Pkd2, Pkhd1, andolaris16 were as described.

Immunodetection and �-GalactosidaseStaining

For whole-mount immunohistochemistry, pancreataissected from embryos were fixed in 4% paraformaldehyden PBS for 2 hours at 4°C and treated as described.26 Formmunodetection on sections, embryos were fixed at 4°C in% paraformaldehyde in PBS for 16 hours and then em-edded in paraffin. Immunofluorescence was as described.27

rimary antibodies and dilutions were as follows: monoclo-al mouse anti-E-cadherin at 1:50 and mouse anti-�-cate-in at 1:1000 (Transduction Laboratories, BD Biosciences,rembodegem, Belgium), mouse anti-acetylated tubulin at:5000 (Sigma, Bornem, Belgium), rabbit anti-carboxypep-idase A at 1:1000 (Biogenesis, Poole, England), rabbitnti-phospho histone H3 at 1:50 (Cell Signalling Technol-gy, Bioké, Leiden, The Netherlands), rabbit anti-HNF-6sc13050; Santa Cruz Biotechnology, Santa Cruz, CA) at:50, goat anti-HNF-1� at 1:100 (Santa Cruz), and Arme-ian hamster anti-Muc-1 at 1:200 (Neomarkers, Fremont,A). Primary antibodies were detected by immunofluores-ence using anti-mouse or anti-rabbit antibody coupled toexas Red (Jackson Immunoresearch Laboratories, Soham,K) or Alexa594 (Invitrogen, Molecular Probes, Merel-eke, Belgium) or using biotinylated anti-rabbit or anti-rmenian hamster immunoglobulin G (Roche) followed by

treptavidin-Alexa Fluor 488 conjugate (Molecular Probes).mmunodetection of HNF-1� was carried out with Tyra-ide signal amplification. To assess Wnt activity, TOP-AL transgenic embryos were examined for �-galactosidase

xpression at e16.5, as described.28

Statistical Methods

All values are expressed as the mean � SEM. Sta-

istical analysis was performed by the Student t test, and

r�

oseifameoMedMietsa

Adc(6lfueoedd

vmeutpi

Fdedi

534 PIERREUX ET AL GASTROENTEROLOGY Vol. 130, No. 2

esults were considered statistically significant at a P value.05 (*), � .01 (**), and � .001 (***).

Results

Expression of HNF-6 in DevelopingPancreas

To investigate the role of HNF-6 in developmentf the pancreatic ducts, we first characterized its expres-ion pattern. Earlier data have shown that HNF-6 isxpressed in the pancreatic endocrine precursors but notn mature endocrine cells14,22 and that HNF-6 mRNA isound in Pdx-1� precursor cells, in mature exocrinecini, and in undefined ductal cells.14,22,29,30 Using im-unofluorescence, we now examined HNF-6 protein

xpression and its colocalization with mucin-1 (Muc-1)n pancreatic sections from e13.5 to e17.5 (Figure 1A).ucin-1 is an O-glycosylated transmembrane protein

xpressed at the apical pole of epithelial cells that line theucts of various organs, and, in the pancreas, the anti-uc-1 antibody is known to mark the cells lining the

ntercalated, intralobular, and interlobular ducts.19 At13.5, HNF-6 protein was detected throughout most ofhe pancreatic epithelium (Figure 1A). At that stage, aubset of HNF-6-positive cells expressed Muc-1 at the

e13.5 e

dada

HNF-6/Muc

A

B e13.5

Ac Tub/HNF-6

e

Ac Tub/HN

HNF-6/Muc-1

igure 1. HNF-6 is expressed in the epithelium of all segments of thevelopment were examined by immunofluorescence using antibodiexpressed throughout the pancreatic epithelium and then became resuctal structures. Acetylated tubulin was only found in HNF-6-positi

ntralobular duct; ic, intercalated duct; iel, interlobular duct.

pical pole and delineated immature ductal structures. s

t e15.5, HNF-6 protein was more abundant in theifferentiating ductal cells than in the developing acinarells (Figure 1A). Endocrine cells were devoid of HNF-6Figure 1A, arrows). Muc-1 staining was found in HNF--positive cells, irrespective of the HNF-6 expressionevel. Toward the end of gestation (e17.5), HNF-6 wasound in the cuboidal epithelium that lines the intralob-lar and interlobular ducts, as well as in the flattenedpithelium of the intercalated ducts that delineates partf the lumen of the acini. All the HNF-6-positive cellsxpressed Muc-1 at their apical pole. Altogether, theseata show that HNF-6 is expressed in all segments of theuctal compartment, at all stages of duct development.

Duct Morphogenesis Is Affected in Hnf6�/�

Embryos

We next investigated the role of HNF-6 in de-elopment of the pancreatic ducts by comparing ductorphology in wild-type and Hnf6�/� pancreata. To this

nd, we performed whole-mount immunohistochemistrysing the anti-mucin-1 antibody (Figure 2).19 In wild-ype embryos at e11.5, Muc-1 staining of the dorsalancreatic bud identified the pancreatic lumen, which isn continuity with the gut tube lumen. Several exten-

5

dada

dddd

e17.5

ieliel ialial

HNF-6/Muc-1

e17.5

Ac Tub/HNF-6

icic

ncreatic ducts. Wild-type pancreata at different stages of embryoniccted against HNF-6, Muc-1, and acetylated tubulin. HNF-6 was firstd to the ductal compartment. (A) Muc-1 was found in the developingctal cells. (B) da, differentiating acini; dd, differentiating duct; ial,

15.

-1

15.5

F-6

e pas diretricteve du

ions of the pancreatic lumen were also detected. At that

FpHaowmaewdeffceiasvdedpHwipnvtaawcmadtdsfeiil

February 2006 HNF-6 AND PANCREATIC DUCTS IN THE MOUSE 535

e11.5

Hnf6-/-Control

e12.5

e15.5

e17.5

*

*

e17.5

ic

ialiel

ialic

da

dd

ial

iel

ic

icial

iel

iel

CPA/Muc-1

st

ll

st

d

CPA/Muc-1

**

igure 2. Pancreatic duct mor-hogenesis is affected innf6�/� embryos. Wild-typend Hnf6�/� pancreata at vari-us stages of developmentere processed for whole-ount immunochemistry usingn anti-Muc-1 antibody. At11.5, the pancreatic lumen (l)as enlarged in the Hnf6�/�

orsal pancreatic bud (delin-ated by a dotted line), andewer Muc-1-positive cells wereound around the main duct, asompared with controls. At12.5, Muc-1-positive staining

n Hnf6�/� pancreas displayedn irregular ductal network in-tead of an arborescence of de-eloping ducts around a centraluct, as seen in controls. At15.5, differentiating acini anducts were identified in controlancreas, whereas, in thenf6�/� pancreas, the aciniere connected to cystic (aster-

sk) ducts. At e17.5 in Hnf6�/�

ancreas, the acini were con-ected to cystic ducts (asterisk)ia morphologically normal in-ercalated ducts. The differenti-tion boundary between thecini and intercalated ductsas preserved in Hnf6�/� pan-reas, as evidenced by the nor-al expression of CPA (red) incini and lack of CPA in theucts. The intercalated ductshat connect the acini to theuctal compartment (Muc-1taining, green) are not af-ected. d, duodenum; da, differ-ntiating acini; dd, differentiat-ng duct; ial, intralobular duct;c, intercalated duct; iel, inter-

obular duct; st, stomach.

stnaunIvIofieoueanmmaHM(rHfitnmHt

HteSfaEcTw0Htf

sp

WsfiasPdfs

c�H(dtapssemlMtwirta

ttsd�nt

sct6ic

536 PIERREUX ET AL GASTROENTEROLOGY Vol. 130, No. 2

tage in Hnf6�/� embryos, duct morphology was per-urbed because the pancreatic lumen was enlarged, andearly no Muc-1� extensions were found. One day later,t e12.5, Muc-1 staining in wild-type embryos showedp as a regular arborescence around a central duct, run-ing from the head to the tail of the growing pancreas.n contrast, in Hnf6�/� pancreas, Muc-1 staining re-ealed an irregular and nonorganized network of ducts.n wild-type pancreas at e15.5, differentiating acini werebserved at the periphery of the ductal network, and therst signs of ductal segmentation were detected. At17.5, these segments had adopted a mature morphol-gy, allowing the identification of intercalated, intralob-lar, and interlobular ducts. In Hnf6�/� pancreas at15.5, acini were connected to cystic ducts. At e17.5, thecini located at the periphery of the pancreas were con-ected to cystic intralobular and interlobular ducts viaorphologically normal intercalated ducts. To deter-ine whether the differentiation boundary between the

cini and the ducts was preserved in the absence ofNF-6, we costained control and Hnf6�/� sections foruc-1 and for the exocrine marker carboxypeptidase A

CPA). This was the case, because CPA expression wasestricted to the acinar cells, both in control and innf6�/� pancreata. In addition, these observations con-rmed that intercalated ducts are not dilated. Takenogether, these data indicate that HNF-6 is required forormal development of the inter- and intralobular seg-ents of the pancreatic ducts and that, in the absence ofNF-6, these specific segments give rise to cystic struc-

ures.

Cells Lining the Cysts in Hnf6�/� PancreasDo Not Over Proliferate

To investigate whether formation of cysts innf6�/� pancreas is associated with over-proliferation of

he epithelial cells,31 we assessed the number of prolif-rating cells at e15.5, ie, when the cysts started to form.ections of control and Hnf6�/� pancreata were costainedor phosphohistone H3 (pHH3), which marks prolifer-ting cells, and E-cadherin, which marks the epithelium.-cadherin� cell counts were similar on 7 slides from 2ontrol (3331 cells) and 2 Hnf6�/� (3446 cells) embryos.he percentage of E-cadherin� cells that express pHH3as 8.0% � 0.2% in wild-type pancreata and 5.4% �.4% in Hnf6�/� pancreata. Therefore, the absence ofNF-6 is not associated with increased proliferation of

he epithelial cells, and cyst formation does not resultrom aberrant proliferation.

Cyst formation may also result from uncontrolled fluidecretion from the ductal cells as a consequence from

erturbed expression of ion transporters and aquaporins. 1

e therefore measured the expression of the pancreaticodium-bicarbonate transporter (Slc4a4), and the cysticbrosis transmembrane conductance regulator (CFTR),s well as of aquaporins (AQP) 1 and 5, which aretrongly expressed in the pancreatic ducts.32 Real-timeCR was performed at the time of cyst formation. Noifference in the expression of these transporters wasound between control and Hnf6�/� pancreata (data nothown).

The Cysts in Hnf6�/� Pancreas AreDelineated by a Dysmorphic Epithelium

To analyze the epithelial characteristics of theells lining the cysts, we examined the localization of-catenin and Muc-1 on sections of wild-type andnf6�/� pancreata analyzed by immunofluorescence

Figure 3). In control embryos at e17.5, and at postnatalays (P) 1 and 10, �-catenin was found predominantly athe lateral membrane of the epithelium and Muc-1 at thepical pole (Figure 3, insets) as expected.19 In Hnf6�/�

ancreata, cells lining the cysts displayed a heterogenousize and shape and formed an irregular epithelium. Inome instances, cysts were delineated by a multilayeredpithelium. �-Catenin was not restricted to the lateralembrane domain but was distributed to the basal,

ateral, and apical membrane of the cells (Figure 3).uc-1 staining was discontinuous around the lumen of

he cysts, indicating that a number of epithelial cellsere devoid of Muc-1 at their apical pole (Figure 3,

nsets). From these data, we conclude that HNF-6 isequired for normal polarization and morphogenesis ofhe pancreatic duct epithelium and that cysts lined withn aberrant epithelium develop in the absence of HNF-6.

It was recently proposed that Wnt signaling controlsubulogenesis.33 Our data do not support this proposal inhe context of pancreatic duct formation. Indeed, no Wntignaling activity is detectable in normal pancreaticucts,34 and, in Hnf6�/� pancreas, mislocalization of-catenin was not associated with aberrant Wnt/�-cate-in signaling as detected by the TOPGAL reporter sys-em24 (data not shown).

Defective Assembly of Primary Cilia inHnf6�/� Pancreas

Cyst formation in tubular structures is often as-ociated with an absence or dysfunction of the primaryilia.1,2,5 To verify whether HNF-6 controls cilia forma-ion, we first studied whether cilia are found in HNF--expressing cells in developing pancreas. Cilia weredentified by immunodetection of acetylated tubulin, aonstituent of the ciliary axoneme. As shown in Figure

B, at e13.5, e15.5, and e17.5, cilia were detected in

cdfsbaacd

p(pOStaOno

tkHewe4flw

tdcmr

HwatbmarcaefaPcPwmcr

Ftiwaflsfctcid(�flm

February 2006 HNF-6 AND PANCREATIC DUCTS IN THE MOUSE 537

ells expressing high levels of HNF-6, ie, in developingucts. We then compared cilia formation on sectionsrom wild-type and Hnf6�/� pancreata (Figure 4A). Theections at e14.5 were costained with Muc-1 to visualizeetter the ductal lumina. In contrast to control pancre-ta, Hnf6�/� pancreata showed no primary cilia at e14.5nd e15.5. However, at e17.5, cilia remained absent inells lining the cysts but became detectable in the ductsisplaying a normal morphology.We considered the possibility that cilia eventually ap-

eared as a result of the compensatory activity of Onecut-2OC-2), a paralog of HNF-6 that is also expressed in theancreas.35 We therefore analyzed the pancreas from Hnf6/c2�/� double knockout embryos for the presence of cilia.ingle Oc2�/� knockout embryos did not show cyst forma-ion (data not shown). As in Hnf6�/� pancreata, cilia werebsent from cells that line the cysts in late gestation Hnf6/c2�/� embryos but were detectable in smaller ducts (dataot shown). Therefore, OC-2 is not involved in the late-nset appearance of cilia in the absence of HNF-6.

The transcription factor HNF-1� is known to controlhe expression of genes involved in cilium function inidney.16 We have shown earlier that the expression ofNF-1� in the ducts of Hnf6�/� pancreata is reduced at

14.5 and reaches near normal levels at e17.5.14 In lineith these results, we now confirm that the ducts at

17.5 expressed near normal levels of HNF-1� (FigureB, arrowheads) but that most cells lining the cystsailed to express this factor. The few HNF-1�� cellsining the cysts were located close to or were connected

e17.5

Control

Hnf6-/- *

igure 3. The epithelium lininghe cysts in Hnf6�/� pancreass dysmorphic. Sections fromild-type and Hnf6�/� pancre-ta were examined by immuno-uorescence using antibodiespecific for �-catenin (red) andor Muc-1 (green in insets). Asompared with the ductal epi-helium in control pancreas, theells lining the cysts (asterisk)n Hnf6�/� pancreas showediscontinuous Muc-1 stainingarrowheads) and mislocalized-catenin expression, and they

ormed a disorganized epithe-ium, which was sometimesultilayered.

ith a duct (Figure 4B, open arrowheads), suggesting e

hat HNF-1� is absent specifically from cells that lineilated ductal structures. From these experiments, weonclude that HNF-6 is required for timely cilium for-ation, as well as for normal expression of HNF-1�, a

egulator of cilium function.

Expression of Cystic Disease Genes IsAffected in Hnf6�/� Pancreas

The lack of cilia and the reduced expression ofNF-1� in Hnf6�/� cysts prompted us to investigatehether the expression of so-called cystic disease genes is

ffected in Hnf6�/� pancreata. These genes include a sethat is regulated by HNF-1�16–18 as well as those that haveeen associated with PD in humans and in various animalodels.1,2,36,37 Real-time PCR was performed on control

nd Hnf6�/� pancreata (Figure 5). In line with our previouseport,14 expression of Hnf1� was reduced in Hnf6�/� pan-reata from e12.5 to e16.5. This expression was normalizedt e17.5, despite the fact that cells lining cysts do notxpress HNF-1�. The latter result is compatible with theact that cysts, which lack HNF-1� and cilia, represent onlysmall fraction of the pancreas. Expression of Pkd1, Pkd2,olaris, Sec63, and Prkcsh was not affected in Hnf6�/� pan-reas at any stage (Figure 5). In contrast, expression ofkhd1, which codes for the fibrocystin/polyductin protein,as down-regulated from e12.5 to e15.5 and reached nor-al levels at e16.5. Expression of Cys1, which codes for

ystin, was also reduced in Hnf6�/� pancreas, but theeduction persisted throughout development. From these

P1 P10

* *

xperiments, we conclude that HNF-6 controls the expres-

sb

eaptccc

tasteedm

dHibirp

ueWdcdomrdHd

C

538 PIERREUX ET AL GASTROENTEROLOGY Vol. 130, No. 2

ion of cystic disease genes and that the need for HNF-6 cane either transient or permanent.

Discussion

In this study, we have demonstrated that HNF-6 isxpressed in pancreatic ductal cells during organogenesisnd that, in its absence, development of the ductal com-artment is affected, and cysts develop in inter- and in-ralobular ducts. This is associated with a lack of primaryilium at the apical pole of the epithelial cells lining theysts and with a reduced expression of genes required forilium function and/or formation.

Little is known about the molecular mechanismshat control duct development in the pancreas. Usingn inducible Cre-ER-LoxP system, Gu et al demon-trated that the pancreatic ducts arise from a popula-ion of Pdx1-expressing cells that separates from thexocrine and endocrine lineages between e9.5 and11.5.13 We show here that HNF-6 is expressed in theuctal compartment throughout pancreas develop-

A

B

e17.5

Control Hnf6-/-

*

ontrol

Hnf6-/-

e14.5 e15.5

Ac Tub

*

Ac Tub/Muc-1

Ac Tub/Muc-1 Ac Tub

HNF-1βHNF-1β

ent and that it is required for cilia formation and i

uct morphogenesis in a segment-specific way. Innf6�/� mice, cysts developed in intralobular and

nterlobular ducts, but intercalated ducts appeared toe normal. To our knowledge, this is the first reportdentifying a transcription factor that is criticallyequired for development of specific segments of theancreatic ducts.In Hnf6�/� pancreata, cilia were absent from e14.5

ntil e16.5. Later, cilia remained absent from cystpithelia but became detectable in nondilated ducts.

hy do cilia eventually appear in the normal-sizeucts? One possible explanation is that other factorsompensate for the absence of HNF-6 at late stages ofuct development. A first candidate is OC-2, a paralogf HNF-6 whose expression increases during develop-ent of the pancreas.35 However, this hypothesis was

uled out because double Hnf6/Oc2 knockout embryosisplayed the same anomalies in cilia formation as innf6�/� embryos, with lack of cilia in cysts andelayed appearance of cilia in smaller ducts. HNF-1�

Hnf6-/-

*

e17.5

*Ac Tub

Ac Tub

HNF-1β

Figure 4. Cells lining the cysts inHnf6�/� pancreas do not assem-ble primary cilia. (A) Sectionsfrom wild-type and Hnf6�/� pan-creata were examined by immu-nofluorescence using antibodiesspecific for acetylated tubulin(red) and for Muc-1 (green). Ascompared with the ciliated ductalepithelium in control pancreata,ductal cells (e14.5 and e15.5)and cells lining the cysts inHnf6�/� pancreata did not as-semble primary cilia. At e17.5,cilia were observed in morpho-logically normal ducts (arrow-heads) but not in cystic ducts(asterisk). (B) Sections fromwild-type and Hnf6�/� pancre-ata were examined by immuno-fluorescence using an HNF-1�-specific antibody. In wild-typepancreas, HNF-1� (red) was ex-pressed in the ductal epithe-lium. In Hnf6�/� pancreas,HNF-1� was absent from mostcells lining the cysts (asterisk)but was expressed at normallevels in normal-size ducts (ar-rowheads) and in some cellsthat line the cysts and that arelocated close to nonaffectedducts (open arrowheads).

s another candidate because its expression profile in

Htsbc

adAwwn

cdPctoettttcuifdccmo

acoWmHpgkpBgPdkPfcMwcaPmdHti

P

Fiw(s�EfCoP

February 2006 HNF-6 AND PANCREATIC DUCTS IN THE MOUSE 539

nf6�/� pancreas closely parallels that of cilia forma-ion. However, it is unlikely that HNF-1� aloneuffices to induce cilia formation in Hnf6�/� pancreasecause, in kidney, HNF-1� is required for normalilium function but not for cilium formation.16

An alternative explanation to compensatory mech-nisms would be that HNF-6 is required for ductevelopment only during a restricted time window.ccording to this model, the transcription factor net-ork regulating duct development would be dynamic,ith HNF-6 being required for the early phases but

Hnf1β

khd1

5

10

25

20

15

Pkd1

e12.5 e16.5e15.5e14.5 e17.5

40

0

160

120

80

**

4

0

8

16

12 *

*

***

*

*

0

Cys1

50

10

252015

30

***

****

*

ControlHnf6-/-

Rel

ativ

e ex

pre

ssio

nR

elat

ive

exp

ress

ion

Rel

ativ

e ex

pre

ssio

nR

elat

ive

exp

ress

ion

igure 5. Expression of Hnf1� and of cystic disease genes is affectedn Hnf6�/� pancreas. Real-time PCR experiments were performed onild-type and Hnf6�/� pancreata before the occurrence of cysts

e12.5–e14.5), at the onset of cyst formation (e15.5), or at latertages (e16.5–e17.5). Expression levels were normalized to that of-actin � 1000 for Hnf1� and � 10,000 for Pkhd1, Cys1, and Pkd1.xpression of Hnf1� and of Pkhd1 was reduced in Hnf6�/� pancreasrom e12.5 to e16.5 but was normalized at e17.5, whereas that ofys1 was persistently reduced throughout development. Expressionf Pkd1 was unaffected in Hnf6�/� pancreas, like that of Pkd2,olaris, Sec63, and Prkcsh (see Figure 9).

ot for the late phases of duct development. d

Like Hnf6�/� mice, orpk mice show an absence ofilia associated with cyst formation in pancreaticucts.19,20 Orpk mice are deficient in the expression ofolaris, an intraflagellar transport protein required foriliogenesis. Despite the lack of cilia in pancreashroughout development, cyst formation occurrednly around birth in orpk mice, whereas it startedarlier, at e15.5, in Hnf6�/� mice. This suggests thathe role of HNF-6 in pancreatic duct formation ex-ends beyond the control of cilia formation and func-ion. Furthermore, although the role of cilia in renalubulogenesis has received increasing support, howilia control pancreatic duct formation is not wellnderstood. We may suggest that cilia would functionn pancreatic ducts as fluid sensors, as it is proposedor kidney cells, and that cilia would adjust ductaliameter in the pancreas. In addition, we note thatilia are found in all segments of the mature ductalompartment but not in acinar cells. Therefore, ciliaay play a role in the specification and/or maintenance

f the ductal identity.Lack of HNF-1� in the kidney induces cyst formation

nd decreased expression of cystic disease genes involved inilia function, indicating that HNF-1� regulates a networkf genes involved in tubulogenesis and in cilia function.16

e have now found that, in Hnf6�/� pancreata, cyst for-ation was associated with lack of cilia and lack ofNF-1� expression in cells lining the cysts. In Hnf6�/�

ancreata, expression of Pkhd1, a direct HNF-1� targetene that is associated with autosomal recessive polycysticidney disease,17,18 was reduced according to a temporalrofile that parallels the reduction in HNF-1� expression.ecause binding sites were found for HNF-6 in the Hnf1�ene promoter,21 we propose that a HNF-6 ¡ HNF-1� ¡

khd1 cascade controls tubulogenesis in the pancreas. Re-uced expression of Pkhd1 does not affect cilia formation inidney16 but does so in liver.38 Hence, the involvement ofkhd1 in cilium formation is cell-type dependent. There-

ore, it is possible that the HNF-6 ¡ HNF-1� ¡ Pkhd1ascade controls cilium formation in pancreatic ducts.oreover, we found that the expression of Cys1, a genehose spontaneous mutation in mice (cpk mouse)39 induces

yst formation in kidneys, was also reduced during pancre-tic development in Hnf6�/� mice. Combined defects inkhd1 and Cys1 gene expression may thus affect cyst for-ation. In contrast to the Hnf1� gene, whose expression in

ucts was only transiently reduced in the absence ofNF-6, the reduction of Cys1 expression was persistent

hroughout development, suggesting that Cys1 expressions not strictly dependent on HNF-1�.

In conclusion, we propose that HNF-6 controls

evelopment of specific segments of pancreatic ducts

agramd

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

540 PIERREUX ET AL GASTROENTEROLOGY Vol. 130, No. 2

nd regulates a network of genes involved in tubulo-enesis and hereditary polycystic diseases. Our workeveals the existence of distinct differentiation mech-nisms operating in distinct pancreatic duct compart-ents. Hnf6-deficient mice constitute a genetically

efined model of pancreatic polycystic disease.

References1. Boletta A, Germino GG. Role of polycystins in renal tubulogen-

esis. Trends Cell Biol 2003;13:484–492.2. Zhang Q, Taulman PD, Yoder BK. Cystic kidney diseases: all

roads lead to the cilium. Physiology (Bethesda) 2004;19:225–230.

3. Nicolau C, Torra R, Bianchi L, Vilana R, Gilabert R, Darnell A, BruC. Abdominal sonographic study of autosomal dominant polycys-tic kidney disease. J Clin Ultrasound 2000;28:277–282.

4. Praetorius HA, Spring KR. A physiological view of the primarycilium. Annu Rev Physiol 2005;67:515–529.

5. Watnick T, Germino G. From cilia to cyst. Nat Genet 2003;34:355–356.

6. Guay-Woodford LM. Murine models of polycystic kidney disease:molecular and therapeutic insights. Am J Physiol Renal Physiol2003;285:F1034–F1049.

7. Pazour GJ, Witman GB. The vertebrate primary cilium is a sensoryorganelle. Curr Opin Cell Biol 2003;15:105–110.

8. Delmas P. Polycystins: from mechanosensation to gene regula-tion. Cell 2004;118:145–148.

9. Yoder BK, Hou X, Guay-Woodford LM. The polycystic kidney dis-ease proteins, polycystin-1, polycystin-2, polaris, and cystin, areco-localized in renal cilia. J Am Soc Nephrol 2002;13:2508–2516.

0. Kim SK, MacDonald RJ. Signaling and transcriptional control ofpancreatic organogenesis. Curr Opin Genet Dev 2002;12:540–547.

1. Edlund H. Pancreatic organogenesis—developmental mecha-nisms and implications for therapy. Nat Rev Genet 2002;3:524–532.

2. Murtaugh LC, Melton DA. Genes, signals, and lineages in pan-creas development. Annu Rev Cell Dev Biol 2003;19:71–89.

3. Gu G, Dubauskaite J, Melton DA. Direct evidence for the pancre-atic lineage: NGN3� cells are islet progenitors and are distinctfrom duct progenitors. Development 2002;129:2447–2457.

4. Maestro MA, Boj SF, Luco RF, Pierreux CE, Cabedo J, Servitja JM,German MS, Rousseau GG, Lemaigre FP, Ferrer J. Hnf6 and Tcf2(MODY5) are linked in a gene network operating in a precursorcell domain of the embryonic pancreas. Hum Mol Genet 2003;12:3307–3314.

5. Bingham C, Ellard S, Allen L, Bulman M, Shepherd M, Frayling T,Berry PJ, Clark PM, Lindner T, Bell GI, Ryffel GU, Nicholls AJ,Hattersley AT. Abnormal nephron development associated with aframeshift mutation in the transcription factor hepatocyte nuclearfactor-1 �. Kidney Int 2000;57:898–907.

6. Gresh L, Fischer E, Reimann A, Tanguy M, Garbay S, Shao X,Hiesberger T, Fiette L, Igarashi P, Yaniv M, Pontoglio M. A tran-scriptional network in polycystic kidney disease. EMBO J 2004;23:1657–1668.

7. Hiesberger T, Bai Y, Shao X, McNally BT, Sinclair AM, Tian X,Somlo S, Igarashi P. Mutation of hepatocyte nuclear factor-1�inhibits Pkhd1 gene expression and produces renal cysts in mice.J Clin Invest 2004;113:814–825.

8. Hiesberger T, Shao X, Gourley E, Reimann A, Pontoglio M, Iga-rashi P. Role of the hepatocyte nuclear factor-1 � (HNF-1�)C-terminal domain in Pkhd1 (ARPKD) gene transcription and renal

cystogenesis. J Biol Chem 2005;280:10578–10586.

9. Cano DA, Murcia NS, Pazour GJ, Hebrok M. Orpk mouse modelof polycystic kidney disease reveals essential role of primarycilia in pancreatic tissue organization. Development 2004;131:3457–3467.

0. Zhang Q, Davenport JR, Croyle MJ, Haycraft CJ, Yoder BK. Dis-ruption of IFT results in both exocrine and endocrine abnormali-ties in the pancreas of Tg737(orpk) mutant mice. Lab Invest2005;85:45–64.

1. Clotman F, Lannoy VJ, Reber M, Cereghini S, Cassiman D, Jac-quemin P, Roskams T, Rousseau GG, Lemaigre FP. The onecuttranscription factor HNF6 is required for normal development ofthe biliary tract. Development 2002;129:1819–1828.

2. Jacquemin P, Durviaux SM, Jensen J, Godfraind C, Gradwohl G,Guillemot F, Madsen OD, Carmeliet P, Dewerchin M, Collen D,Rousseau GG, Lemaigre FP. Transcription factor hepatocyte nu-clear factor 6 regulates pancreatic endocrine cell differentiationand controls expression of the proendocrine gene ngn3. Mol CellBiol 2000;20:4445–4454.

3. Clotman F, Jacquemin P, Plumb-Rudewiez N, Pierreux C, Van derSmissen P, Dietz H, Courtoy P, Rousseau GG, Lemaigre F. Con-trol of liver cell fate decision by a gradient of TGFb signalingmodulated by Onecut transcription factors. Genes Dev 2005;19:1849–1854.

4. DasGupta R, Fuchs E. Multiple roles for activated LEF/TCF tran-scription complexes during hair follicle development and differ-entiation. Development 1999;126:4557–4568.

5. Pierreux CE, Vanhorenbeeck V, Jacquemin P, Lemaigre FP, Rous-seau GG. The transcription factor hepatocyte nuclear factor-6/One-cut-1 controls the expression of its paralog Onecut-3 in developingmouse endoderm. J Biol Chem 2004;279:51298–51304.

6. Jacquemin P, Lemaigre FP, Rousseau GG. The Onecut transcrip-tion factor HNF-6 (OC-1) is required for timely specification of thepancreas and acts upstream of Pdx-1 in the specification cas-cade. Dev Biol 2003;258:105–116.

7. van Eyll JM, Pierreux CE, Lemaigre FP, Rousseau GG. Shh-depen-dent differentiation of intestinal tissue from embryonic pancreasby activin A. J Cell Sci 2004;117:2077–2086.

8. Nagy A, Gertsenstein M, Vintersten K, Behringer R. Manipulatingthe mouse embryo: a laboratory manual. 3rd ed. Plainview, NY:Cold Spring Harbor Lab Press, 2003:687–689.

9. Rausa F, Samadani U, Ye H, Lim L, Fletcher CF, Jenkins NA,Copeland NG, Costa RH. The cut-homeodomain transcriptionalactivator HNF-6 is coexpressed with its target gene HNF-3 � in thedeveloping murine liver and pancreas. Dev Biol 1997;192:228–246.

0. Landry C, Clotman F, Hioki T, Oda H, Picard JJ, Lemaigre FP,Rousseau GG. HNF-6 is expressed in endoderm derivatives andnervous system of the mouse embryo and participates to thecross-regulatory network of liver-enriched transcription factors.Dev Biol 1997;192:247–257.

1. Quarmby LM, Parker JD. Cilia and the cell cycle? J Cell Biol2005;169:707–710.

2. Burghardt B, Elkaer ML, Kwon TH, Racz GZ, Varga G, Steward MC,Nielsen S. Distribution of aquaporin water channels AQP1 andAQP5 in the ductal system of the human pancreas. Gut 2003;52:1008–1016.

3. Germino GG. Linking cilia to Wnts. Nat Genet 2005;37:455–457.

4. Dessimoz J, Bonnard C, Huelsken J, Grapin-Botton A. Pancreas-specific deletion of �-catenin reveals Wnt-dependent and Wnt-independent functions during development. Curr Biol 2005;15:1677–1683.

5. Jacquemin P, Pierreux CE, Fierens S, van Eyll JM, Lemaigre FP,Rousseau GG. Cloning and embryonic expression pattern of themouse Onecut transcription factor OC-2. Gene Expr Patterns

2003;3:639–644.

3

3

3

3

Mts

vEgF

February 2006 HNF-6 AND PANCREATIC DUCTS IN THE MOUSE 541

6. Drenth JP, Martina JA, Te Morsche RH, Jansen JB, Bonifacino JS.Molecular characterization of hepatocystin, the protein that isdefective in autosomal dominant polycystic liver disease. Gastro-enterology 2004;126:1819–1827.

7. Davila S, Furu L, Gharavi AG, Tian X, Onoe T, Qian Q, Li A, Cai Y,Kamath PS, King BF, Azurmendi PJ, Tahvanainen P, Kaariainen H,Hockerstedt K, Devuyst O, Pirson Y, Martin RS, Lifton RP, Tah-vanainen E, Torres VE, Somlo S. Mutations in SEC63 causeautosomal dominant polycystic liver disease. Nat Genet 2004;36:575–577.

8. Masyuk TV, Huang BQ, Ward CJ, Masyuk AI, Yuan D, Splinter PL,Punyashthiti R, Ritman EL, Torres VE, Harris PC, LaRusso NF.Defects in cholangiocyte fibrocystin expression and ciliary struc-ture in the PCK rat. Gastroenterology 2003;125:1303–1310.

9. Hou X, Mrug M, Yoder BK, Lefkowitz EJ, Kremmidiotis G,

D’Eustachio P, Beier DR, Guay-Woodford LM. Cystin, a novel a

cilia-associated protein, is disrupted in the cpk mouse model ofpolycystic kidney disease. J Clin Invest 2002;109:533–540.

Received October 24, 2005. Accepted November 2, 2005.Address requests for reprints to: F. P. Lemaigre, MD, Hormone andetabolic Research Unit, Université catholique de Louvain and Insti-

ute of Cellular Pathology, Avenue Hippocrate 75/7529, B-1200 Brus-els, Begium. e-mail: lemaigre@horm.ucl.ac.be; fax (32) 2-764 7507.Supported by grants from the Belgian State Program on Interuni-

ersity Poles of Attraction (to F.P.L. and L.L.), from the D. G. Higherducation and Scientific Research of the French Community of Bel-ium, from the Fund for Scientific Medical Research (to G.G.R. and.P.L.), from the European Union (grant QLRT-2000-01257; to L.L.),

nd from the Instituto de Salud Carlos III (GO3/212; to J.F.).