TC

M

*M

BsetpaihatldmatcsbTlbiaidstgmitqa

Tatitipcd

GASTROENTEROLOGY 2008;135:591–600

ranscription Factor Foxq1 Controls Mucin Gene Expression and Granuleontent in Mouse Stomach Surface Mucous Cells

ICHAEL P. VERZI,*,‡ ABDUL H. KHAN,*,‡ SUSUMU ITO,§ and RAMESH A. SHIVDASANI*,‡

Dana-Farber Cancer Institute and Department of Medicine, ‡Brigham & Women’s Hospital and Harvard Medical School, and §Department of Cell Biology, Harvard

edical School, Boston, Massachusetts

dgticf

tgotupafpcf(MtRii

da(sFiuracgtafi

o

BA

SIC–

ALI

MEN

TARY

TRA

CT

ackground & Aims: The gastric mucosa provides atringent epithelial barrier and produces acid andnzymes that initiate digestion. In this regeneratingissue, progenitors differentiate continually into 4rincipal specialized cell types, yet underlying mech-nisms of differentiation are poorly understood. Wedentified stomach-restricted expression of the fork-ead transcription factor FOXQ1. Methods: We usedcombination of genetic, histochemical, ultrastruc-

ural, and molecular analysis to study gastric cellineages with respect to FOXQ1. Results: Within theeveloping and adult gastrointestinal tract, Foxq1essenger RNA (mRNA) is restricted to the stomach

nd expressed predominantly in foveolar (pit) cells,he abundant mucin-producing cells that line the mu-osal surface. Mice carrying Foxq1 coding mutationshow virtual absence of mRNA and protein for theackbone of the major stomach mucin MUC5AC.hese observations correspond to a paucity of foveo-

ar cell secretory vesicles and notable loss of stomachut not intestinal mucus. Transcriptional profiling

dentified a surprisingly restricted set of genes withltered expression in Foxq1 mutant stomachs. MUC5ACs a highly tissue-restricted product that similarlyepends on FOXQ1 in its other major site of expres-ion, conjunctival goblet cells. Conclusions: Takenogether, these observations imply that promotion ofastric MUC5AC synthesis is a primary, cell-autono-ous function of FOXQ1. This study is the first to

mplicate a transcription factor in terminal differen-iation of foveolar cells and begins to define the re-uirements to assemble highly specialized organellesnd cells in the gastric mucosa.

he gastric mucosa contains 4 highly specialized celltypes that differentiate from a common progenitor

nd help execute the stomach’s digestive functions. Al-hough much is known about the morphology and phys-ology of these cell types, few transcriptional regulatorshat govern their differentiation have been character-zed.1–3 Surface mucous cells, also known as foveolar orit cells, line the lumen of the glandular stomach in theorpus and antrum, including mucosal pits that vary in

epth in different areas.4 Acid-secreting parietal cells

ominate the corpus mucosa, which also houses zymo-enic chief cells. These 2 cell lineages are excluded fromhe gastric antrum, where the proportion of foveolar cellss accordingly increased. Foveolar cells constitutively se-rete a viscous mucus that protects the gastric epitheliumrom damage.

Secretory epithelial cells in different organs utilize dis-inct mucin polypeptides as the backbone for extensivelycosylation and mucus synthesis. Although more thanne mucin may be present in a cell lineage, single mucinypes tend to predominate and represent the major prod-ct of specialized secretory cells. Intestinal goblet cellsroduce only MUC2; in the stomach, mucous neck cellsnd basal cells in the antrum produce MUC6, whereasoveolar cells produce MUC5AC.5 Whereas production ofarticular mucin peptides is a hallmark of gastric mu-ous cells, other markers distinguish these cell lineagesurther: pit cells express gastrokine-1 and Trefoil factorTff) 1, whereas mucous neck cells produce Tff2.6,7

UC5AC is also produced by goblet cells in the conjunc-iva, the stratified columnar epithelial lining of the eye.8

egulation of individual mucin genes can thus providemportant clues about lineage-specific cell differentiationn surface epithelia.

In a screen for transcription factor (TF) genes that areifferentially expressed in developing mouse stomachnd intestine,9 we observed that Foxq1 messenger RNAmRNA) is excluded from the intestine and expressedelectively in the stomach. Previous studies have notedoxq1 expression in the stomach of various species, but

ts exact function in this organ is unknown.10 –13 Tonderstand these functions, we studied Satin mice, aadiation-induced mutant strain that is homozygous for

null Foxq1 allele.14 We observed a specific and signifi-ant defect in gastric mucin production and secretoryranule biogenesis in gastric foveolar cells and tracedhese defects to virtually complete absence of MUC5ACnd its glycosylated end products. We confirmed thendings in independent strains of Foxq1 mutant mice

Abbreviations used in this paper: ENU, ethylnitrosourea; PAS, peri-dic acid-Schiff; TF, transcription factor.

© 2008 by the AGA Institute0016-5085/08/$34.00

doi:10.1053/j.gastro.2008.04.019

agaatd

asLTmaicsGc

nttw

Ske4Cas

pupeottavri

sc1tsamdgaspsvt

awoomCSnthmidTts

mgusddcfcssswc(gnit

BA

SIC–

ALIM

ENTA

RY

TRA

CT

592 VERZI ET AL GASTROENTEROLOGY Vol. 135, No. 2

nd also found absence of MUC5AC in conjunctivaloblet cells. A combination of studies in vitro and innimals indicated that the forkhead protein FOXQ1 haslimited but essential function in Muc5ac gene regula-

ion. FOXQ1 is the first TF to be implicated in terminalifferentiation of stomach foveolar cells.

Materials and MethodsMiceMice were housed under pathogen-free conditions

nd handled according to protocols approved by an in-titutional Animal Care and Use Committee. Satin (SB/eJ), Beige, 129/Sv, and C57/BL6 mice were obtained fromhe Jackson Laboratories (Bar Harbor, ME) and CD1ice from Charles River Laboratories (Wilmington, MA),

nd mixed genetic background mice were generated bynterbreeding. Foxq1ENU/� mice were resurrected fromryopreserved sperm at the Mutant Mouse Regional Re-ource Center at the University of California, Davis.enotyping for Foxq1 alleles was done by polymerase

hain reaction (PCR) as described previously.14

Expression Analyses, Histology,Immunohistochemistry, Electron Microscopy,and Transcriptional Reporter AssaysDetails on expression analyses, histology, immu-

ohistochemistry, electron microscopy, and transcrip-ional reporter assays are all included in the Supplemen-ary Materials (see Supplementary Materials online atww.gastrojournal.org).

Microarray Expression AnalysisStomachs from age-matched C57BL/6, Beige, and

B/LeJ mice were harvested immediately after mice wereilled, then washed, and the antrum isolated. RNA wasxtracted with Trizol reagent, labeled, and hybridized to30A2.0 Mouse Expression Arrays (Affymetrix, Santalara, CA). Data were analyzed using dChip software15

nd deposited in the GEO public database, with acces-ion number GSE8943.

ResultsStomach-Restricted Foxq1 ExpressionTranscriptional regulation of cell-specific gene ex-

ression in the gastrointestinal (GI) tract is not wellnderstood. To take steps toward identifying relevantathways, we recently surveyed the temporal and spatialxpression of all known and predicted TF genes in devel-ping mouse gut.9 Among mRNA that are restricted tohe stomach, we identified the forkhead TF Foxq1. Foxq1ranscripts first appear on embryonic day 13, and stom-ch-restricted expression is maintained throughout de-elopment (Figure 1A). Similarly, Foxq1 transcripts areestricted to the adult stomach and absent from adult

ntestine (Figure 1B). Further inspection of Foxq1 expres- t

ion in the stomach corpus (where foveolar, parietal, andhief cells are evident by histochemical staining; FigureC) by in situ hybridization indicated that Foxq1ranscripts in the adult stomach are restricted largely tourface mucous (pit or foveolar) cells (Figure 1D), whichre characterized by robust expression of the lineagearker Muc5AC (Figure 1F). Foxq1 transcripts were also

etected in pepsinogenic chief cells at the base of gastricland units (Figure 1D), although signals were consider-bly weaker than in surface mucous cells and may repre-ent nonspecific background staining. Control (sense)robes typically gave no staining (Figure 1E). Previoustudies using laser capture microdissection found fo-eolar cells enriched and chief cells lacking in Foxq1ranscripts, consistent with our results.3,16

Delineation of Foxq1sa/sa StomachAbnormalitiesTo study FOXQ1 functions, we took advantage of

n existing recessive mutant mouse strain, Satin (Sa),hich was generated by radiation mutagenesis17 and rec-gnized originally by a shiny pelage resulting from dis-rganized hair-shaft medullae.14,18 The genetic defectaps to a Foxq1 nonsense mutation that eliminates the-terminal 112 amino acids of a 400-residue protein.14

erial backcrosses isolated the mutation on a homoge-eous genetic background with tight linkage to an addi-ional mutation, beige; the resulting strain, SB/LeJ, is thusomozygous for both Foxq1sa and Lystbg alleles.17 Beige, autant allele of the Lyst lysosomal transport gene,19 –23

ncreases susceptibility to infection owing to immuneysfunction but has no known role in stomach mucosa.he hair follicle defect in the Satin strain is well charac-

erized,14,24 but the animals seem otherwise normal andtomach defects have not been investigated.

SB/LeJ mice (which we designate Foxq1sb/sb) show nor-al activity, feeding, growth, fertility, and life span; and

ross stomach morphology is intact, without mucosallceration or tumors. Histologic examination of Foxq1sb/sb

tomachs revealed a normal mucosa (Figure 2A), withifferentiated cell types present in normal numbers andistribution, judging by the following immunohisto-hemical markers: H/K-ATPase for parietal cells, gastrinor antral G-cells, and pepsinogen and intrinsic factor forhief cells (Figure 2I, and data not shown). Alcian bluetaining for acidic mucins also showed the typical weakignal in basal mucous cells in the antrum (data nothown). By contrast, periodic acid-Schiff (PAS) staining,hich identifies the neutral mucins secreted by foveolar

ells, revealed a dramatic and completely penetrant defectFigure 2B and C vs Figure 2F and G; n � 10): somelands retained a faint rim of extracellular signal, butone displayed the prominent intracellular apical stain-

ng characteristic of control animals. The full scope ofhe defect can be appreciated in low-magnification pho-

omicrographs (see Supplementary Figure 1A–D online

aggfig

t(msob(gdTas

FasfcbsMtga(Pkm�

ag

4F(rEtbsumMssahp

BA

SIC–

ALI

MEN

TARY

TRA

CT

August 2008 FORKHEAD REGULATION OF TISSUE-SPECIFIC MUC5AC 593

t www.gastrojournal.org) and quantitation of stainedland units (see Supplementary Figure 2A online at www.astrojournal.org). Defective mucin expression was con-ned to the stomach surface; PAS staining of intestinaloblet cells was unaffected (Figure 2D and H).

High-resolution light microscopy revealed absence ofhe apical zone of PAS staining in Foxq1sb/sb pit cellsFigure 3A and B), and we used transmission electron

icroscopy to characterize the defect further. We ob-erved normal size, shape, and polarity of Foxq1sb/sb fove-lar cells (Figure 3E), but apical mucous granule num-ers were markedly reduced compared with controlsFigure 3C and D). In mutant animals, a few of theseranules showed the typical morphology and electronensity, but most were reduced in both size and density.he prominent reduction in PAS staining can thus bettributed to a significant defect in the organelles thattore and release neutral gastric mucin.

Selective Loss of MUC5AC in the Absence ofFoxq1 FunctionTo distinguish whether paucity of apical granules in

oxq1sb/sb pit cells reflects absence of MUC5AC synthesis orfailure to glycosylate and store the protein, we used a

pecific antibody. MUC5AC protein was completely absentrom Foxq1sa/sa stomach samples (Figure 4A, n � 5), indi-ating that FOXQ1 is required to produce the polypeptideackbone for neutral stomach mucin. Judging by othertains, including Alcian blue, synthesis of other mucins,

uc2 and Muc6, is intact in the absence of FOXQ1 func-ion (data not shown). Gastrokine-1, another secreted andranule-bound pit cell product of unknown function,7 islso expressed normally in Foxq1sb/sb surface mucous cellsFigure 4A). Finally, quantitative reverse transcription (RT)-CR analysis of Foxq1sb/sb stomach revealed normal gastro-ine-1, Mucin 1, and stomach trefoil-family factor TFF1RNA levels, whereas Muc5ac transcripts were reduced to3% of levels observed in control samples (Figure 4B).These results reveal virtual absence of Muc5ac mRNA

nd hint at selective loss of MUC5AC among apicalranule contents. To determine the potentially broader

™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™igure 1. Restricted Foxq1 expression in fetal and adult stomach.

A) RT-PCR results from the GIFT database.9 RNA from wild-type mu-ine stomach (St) and intestine (In) at 4 different fetal stages (E11, E13,15, E17) probed using Foxq1-specific primers. Foxq1 is expressed inhe stomach with increasing abundance as the embryo approachesirth but not detected in intestine. (B) RT-PCR in adult mouse stomachhows robust Foxq1 mRNA levels relative to the intestine; GAPDH wassed as a loading control. (D–F) In situ hybridization on adult wild-typeurine stomach using Foxq1 antisense probe (D), sense control (E), oruc5AC antisense probe (F). The Foxq1 probe gave strong signal in

urface epithelium (arrows) throughout the glandular stomach (imageshown here are taken from the corpus, with an H&E-stained section asreference [C]) and considerably weaker signal in chief cells (arrow-

eads); Muc5ac is specific to surface mucous cells (arrows in F).

, Parietal cells.

sntmop

ldluwolC�otpea

Facmtua

F(paoPscci(tGSw

BA

SIC–

ALIM

ENTA

RY

TRA

CT

594 VERZI ET AL GASTROENTEROLOGY Vol. 135, No. 2

cope of Foxq1 function in foveolar cells, we used oligo-ucleotide microarrays to profile mRNA expression inhe gastric antrum. The epithelium of the antrum, the

ost distal portion of the stomach, carries no zymogenicr parietal cells and a correspondingly high fraction of

igure 2. Foxq1sb/sb stomachs are deficient in surface mucin.A–H) Adult mouse stomach or duodenal tissue (Foxq1sb/sb in top 4anels and controls in lower 4 panels) stained with H&E or periodiccid-Schiff (PAS), as indicated. H&E staining reveals normal characterf gastric glandular mucosa; images are taken from the gastric corpus.AS stain reveals marked attenuation of surface epithelial mucoustaining in Foxq1sb/sb stomach (B and C) but normal duodenal goblet-ell staining (D). (I) Immunohistochemistry of adult murine gastric mu-osa (Foxq1sb/sb in top 3 panels and wild type in lower 3 panels), show-

ng normal distribution of markers for chief (pepsinogen) and parietalH/K-ATPase) cells in the corpus and enteroendocrine G (gastrin) cells inhe antrum. Scale bars, 60 �m for PAS-stained cross sections (C and), 120 �m for all others. Low-power photomicrographs are shown inupplementary Figure 1A–D (see Supplementary Figure 1A–D online atww.gastrojournal.org).

it cells4; accordingly, differences in gene expression re- 6

ated to absence of Foxq1 function should be most evi-ent in this region. Because Foxq1sb/sb mice carry tightly

inked mutations in the Foxq1 and Lyst genes, we alsosed expression profiling to compare Foxq1sb/sb stomachith that from the Beige strain, which carries a mutationnly in the Lyst gene and is congenic with the C57BL/6

ine.25 Analysis of antral RNA from Foxq1sb/sb, Beige, and57BL/6 control mice disclosed fewer than 20 genes with4-fold reduction in Foxq1sb/sb samples, and Muc5ac was 1

f only 4 transcripts reduced �10-fold in Foxq1sb/sb an-rum (Figure 4C). RT-PCR on independent stomach sam-les validated the change in Fabp1 mRNA recorded inxpression profiling (see Supplementary Figure 2B onlinet www.gastrojournal.org). These results together impli-

igure 3. Foveolar-cell apical granule deficiency in Foxq1sb/sb stom-ch. (A and B) High-magnification view of PAS-stained gastric foveolarells from adult wild-type (A) and Foxq1sb/sb (B) mice. (C and D) Trans-ission electron microscopy of adult gastric foveolar cells, emphasizing

he apical granular zone in wild-type (C) and Foxq1sb/sb (D) cells. (E) Fullltrastructural profile of Foxq1sa/sa gastric foveolar cells, indicating intactrchitecture except for apical granule deficiency. Scale bars, A and B,

0 �m; C and D, 1 �m; E, 2 �m.

camps

m

imbotsptbeIBsdfmmhcfFddr

sgtpFtciFF

4F(tristaeremaFesptmap

BA

SIC–

ALI

MEN

TARY

TRA

CT

August 2008 FORKHEAD REGULATION OF TISSUE-SPECIFIC MUC5AC 595

ate stomach Foxq1 function in a restricted range ofctivities; Foxq1sb/sb surface mucous cells seem intact inost respects but deficient in apical granules and their

rincipal protein product, MUC5AC, likely reflectingubstantially reduced Muc5ac gene transcription.

Validation of the Role of Foxq1 in FoveolarCell FunctionBecause the parental strain on which the Foxq1sb/sb

utations appeared is no longer available, in the forego-

c

ng histochemical analyses we used a panel of laboratoryouse strains (CD1, C57BL/6, and 129/Sv-C57BL/6 hy-

rids) as controls. In contrast to Foxq1sb/sb stomach, webserved abundant apical mucus staining in every con-rol (Figure 5A, and data not shown), suggesting thattrain background is unlikely to account for the Foxq1sb/sb

henotype. The only other mutation in this strain mapso the closely linked Lyst gene,19 –21,25 which regulatesiogenesis and transport of membranous organelles andnables lysosome-mediated plasma membrane repair.26,27

n hierarchical analysis of mRNA expression profiles,eige and C57BL/6 antra clustered together, whereas Satinamples clustered separately (Figure 4C; the cluster den-rogram reflects the complete array data set). To excludeurther the possibility that pit cell defects in Foxq1sb/sb

ice might reflect Lyst gene inactivity, we examined bg/bgice more closely. PAS staining and MUC5AC immuno-

istochemistry of adult bg/bg stomach were similar toongenic C57BL/6 controls, with abundant signal in sur-ace mucous cells (Figure 5A and see Supplementaryigure 3 online at www.gastrojournal.org). Thus, LYSTeficiency alone cannot account for the foveolar cellefect, which likely results from the Foxq1sa mutationather than the Lystbg allele.

These results do not, however, exclude the formal pos-ibility that combined mutation of the Foxq1 and Lystenes is required to produce the stomach pit cell pheno-ype. Because tight linkage of the satin and beige locirohibits their separation, we asked whether anotheroxq1 mutant strain, generated independently by ethylni-rosourea (ENU) mutagenesis,14 carries the same foveolarell defect. In a previous study, Foxq1enu mice phenocop-ed the hair defect when crossed to Satin mice, butoxq1enu/sb stomachs were not examined.12 Becauseoxq1enu/enu mice were not viable at weaning in our crosses

™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™igure 4. Foxq1sb/sb fail to produce MUC5AC protein or mRNA.

A) Immunohistochemistry of adult Foxq1sa/sa (top) and wild-type (bot-om) mouse gastric mucosa for MUC5AC (left) and gastrokine-1 (Gkn1,ight), indicating absence of MUC5AC but preserved Gkn1 expressionn Foxq1sb/sb stomach; scale bars, 120 �m. (B) qRT-PCR analysis ofelected mucins and secreted foveolar cell products: Muc5AC, Gkn1,refoil factor 1 (Tff1), and Muc1. mRNA expression was first normalizedgainst Gapdh, and the levels in Foxq1sb/sb stomach (yellow bars) arexpressed as a percentage of wild-type levels (blue bars, 100%). Theesults reveal marked attenuation of Muc5ac transcripts, with intactxpression of all other tested markers. (C) Graphic depiction of relativeRNA levels revealed in expression microarray analysis of stomach

ntra from age-matched wild-type C57BL/6 (BL6), Beige, andoxq1sb/sb (Satin) adult mice. Most transcripts showed nearly identicalxpression in the 3 strains; data for all mRNA with �4-fold decrease arehown in expression-heat maps (red, high expression; green, low ex-ression) and as a bar graph. Muc5ac was the second most differen-ially expressed transcript (red bar). The dendrogram below the heatap depicts results of hierarchical clustering analysis of the samples for

ll probes and indicates greatest similarity first between duplicate sam-les and secondly between BL6 and Beige; Foxq1sb/sb (Satin) samples

luster separately.

(Fg5gppsitt

ttawrg

rermbSpmksFdcmttfaCMtlmloT6n

ughhagptodwgtf

Fm(tscSmrsMpFSg

BA

SIC–

ALIM

ENTA

RY

TRA

CT

596 VERZI ET AL GASTROENTEROLOGY Vol. 135, No. 2

P � .008), we generated Foxq1enu/sb mice. Unlikeoxq1enu/� or Foxq1sb/� littermates, compound heterozy-ous Foxq1enu/sb mice showed loss of PAS staining (FigureB and see Supplementary Figure 4 online at www.astrojournal.org, n � 3), similar to SB/LeJ, and hencerovide genetic proof that Foxq1 is responsible for thehenotype. Very weak MUC5AC expression in Foxq1enu/sb

tomachs, which we detected by in situ hybridization andmmunohistochemistry (Figure 5B and see Supplemen-ary Figure 4 online at www.gastrojournal.org), implies

igure 5. Foxq1sb/sb stomach defects specifically reflect Foxq1 geneutations. (A) Staining of gastric mucosa from C57BL/6 (left) and Beige

right 2 panels) mice for PAS (left 2 panels) and MUC5AC immunohis-ochemistry (IHC, far right panel), showing normal expression of neutraltomach mucins and MUC5AC. Images are taken from the gastricorpus, and additional data are shown in Supplementary Figure 3 (seeupplementary Figure 3 online at www.gastrojournal.org). (B) Comple-entation analysis of the Foxq1 gene: examination of Foxq1sb/� (top

ow, control) and Foxq1sb/enu (bottom row) stomach corpus for PAStain (left), stomach antrum for MUC5AC immunostain (middle), anduc5ac in situ hybridization (right). The Foxq1sb/enu strain effectively

henocopies Satin (Foxq1sb/sb) mice, and the results hint that theoxq1enu allele may be hypomorphic. Additional data are shown inupplementary Figure 4 (see Supplementary Figure 4 online at www.astrojournal.org). Scale bars, 60 �m.

hat the Foxq1enu allele is hypomorphic for stomach func- p

ion. Such a hypomorphic phenotype may result fromhe relatively subtle missense mutation in the Foxq1enu

llele (Ile128Ser in the winged helix domain) comparedith the nonsense mutation in the SB/LeJ strain.14 These

esults collectively indicate that Muc5AC production inastric pit cells depends on Foxq1.

The Role of FOXQ1 in Muc5ac GeneRegulation

Because FOXQ1 is a forkhead protein with a likelyole in transcriptional regulation and Muc5ac mRNA lev-ls are low in its absence, we asked whether FOXQ1 mayegulate the Muc5ac gene directly. Previous study of the

ouse Muc5ac promoter found that it could be activatedy transforming growth factor-� signaling, along withmad- and Sp1-family TFs.28 Whereas FOXQ1 is re-orted to bind an AT-rich sequence in the telokin pro-oter, its consensus DNA recognition sequence is un-

nown.13 However, most forkhead TFs recognize theequence RYMAAYA,29 and the binding preference forOXF2, which is closely related to FOXQ1,30 has beenetermined empirically.31 We identified an evolutionarilyonserved forkhead consensus binding sequence in theouse Muc5ac promoter, 100 base pairs upstream of the

ranscriptional start site (Figure 6A). To assess the func-ion of this site, we cloned the Muc5ac promoter sequencerom �199 to �3 upstream of the firefly luciferase genend tested its ability to activate reporter gene expression.ompared with a promoterless reporter construct, thisuc5ac promoter fragment induced robust expression of

he reporter gene in CMT-93 colonic epithelial cells, a celline chosen on the basis of demonstrated Muc5ac pro-

oter activity,27 as well as in the human gastric cancer celline Kato-III (Figure 6B). However, deletion of 2 core nucle-tides in the putative forkhead element (TGTTTAC ¡G–TAC) had little effect on promoter activity (FigureB), and cotransfection of a Foxq1 expression plasmid didot increase it (data not shown).To identify potential Muc5ac enhancers that may fall

nder FOXQ1 control, we searched for conserved inter-enic sequences. Only 2 conserved regions (75% and 72%omology between mouse and human) contained fork-ead consensus sequences (Figure 6C); no other regionsre conserved in either direction until the next structuralenes. We cloned these regions upstream of the Muc5acromoter-reporter and tested FOXQ1-dependent

ranscriptional activation but observed no enhancementver promoter activity alone (data not shown). Thus,espite identification of putative FOXQ1 cis-elements,e gathered no conclusive evidence for direct Muc5acene regulation by FOXQ1. It is, however, importanto note that Fox proteins may remodel chromatin, aunction not accurately reflected in plasmid-based re-

orter assays.32,33

olbsncmcgtsIpmbdsdidM7F

fo

rFsaqsctBTuernFst(me

FtccwapiTh

BA

SIC–

ALI

MEN

TARY

TRA

CT

August 2008 FORKHEAD REGULATION OF TISSUE-SPECIFIC MUC5AC 597

Muc5ac Expression Defects in Satin MiceAre Not Restricted to the StomachMore than half of Foxq1sb/sb mice over 9 months

ld developed ocular surface abnormalities and accumu-ated surface debris, which impaired opening of one oroth eyes (Figure 7A). Other animals maintained in theame colony never showed the same pathology, and weoted that Muc5ac is also present in conjunctival gobletells.8 Most conjunctival epithelial cells produce theembrane-spanning mucins Muc1 and Muc4.8 Goblet

ells occupy the conjunctival fornix; resemble intestinaloblet cells in morphology; and, like mucus in the GIract, secrete products that function to lubricate theurface, clear debris, and provide antimicrobial defense.34

nitial inspection revealed the typical frequency and ap-earance of goblet cells in Foxq1sb/sb conjunctivae, andutant goblet cells stained with both PAS and Alcian

lue, similar to controls (Figure 7B–G); these findingsiffer from the dramatic loss of PAS signal in Foxq1sb/sb

tomach (Figure 2A). However, conjunctival goblet cellsiffer from gastric pit cells, which express only MUC5AC,

n that they express multiple mucin glycoproteins.34 In-eed, immunostaining revealed striking absence ofUC5AC in Foxq1sb/sb conjunctival goblet cells (Figure

H). Muc5ac gene expression thus appears to depend on

igure 6. Conserved regions in the mouse Foxq1 locus contain conranscription directly. (A) Comparison of Muc5ac promoter sequencesonsensus element (blue type) within a region previously shown to be eontrol of the native or FOX site-mutant murine Muc5ac promoter was trith pGL3 promoterless control. Both constructs activated the reportectivity in reporter assays on this promoter. (C) Sequence conservatiipeline.lbl.gov/cgi-bin/gateway2). Conservation was highest in the c

dentified in the promoter and 4 distant regions, 1 that contains rep(A/c)AA(c/t)A. Sequence from these 2 regions is shown below, with aighlighting of putative forkhead-binding elements.

oxq1 function in more than one tissue, and its absence o

rom the conjunctiva is sufficient to produce an overtcular phenotype.

DiscussionSurface GI epithelia engage in continuous self-

enewal and differentiation of highly specialized cells.ew genetic studies have identified TFs that are respon-ible for particular cell features, especially in the stom-ch.1–3 Here, we report the characterization of a TF re-uired for normal gastric foveolar cell differentiation. Wehow that Foxq1 mutations in mice severely limit foveolarells’ ability to synthesize MUC5AC and to fill the secre-ory granules that characterize this unique cell lineage.esides MUC5AC, other foveolar cell products such asFF1 and Gkn1, which are also stored in mucous gran-les,7,35–37 are unaffected by loss of Foxq1 function, andxpression profiling revealed a narrow spectrum of dys-egulated genes. In this light, we expected to find normalumbers and appearance of pit cell mucous granules, butoxq1sb/sb stomach ultrastructure disclosed fewer andmaller granules. This suggests that absence of MUC5AC,he backbone for the major content of these granulesneutral stomach-specific mucin), may preclude normal

ucous granule formation or stability, similar to theffect of MUC2 loss on intestinal goblet cells.38 On the

us forkhead-binding sites, but these may not activate Muc5ac genemouse (top, shaded) and human (bottom) reveals a conserved Foxf2ial for Muc5ac promoter activity.28 (B) A firefly luciferase reporter undercted into CMT-93 or Kato-III cells, and luciferase activity was compared

equally, indicating lack of a requirement for the putative FOX-bindingofile of the murine Foxq1 locus, adapted from Vista Browser (http://

region, but narrow peaks (�70% identity, pink shading) were alsoe elements and 2 that carry consensus forkhead-binding elements,ent of mouse (top, shaded) and human (bottom) sequences and bold

sensfromssentansfer geneon prodingetitivlignm

ther hand, mutant pit cells are not devoid of granules,

atcMmthFSotbc

MmptMffoFtdTsOte

nmmtkvsQngmpdg

t

4F(ay(sM

BA

SIC–

ALIM

ENTA

RY

TRA

CT

598 VERZI ET AL GASTROENTEROLOGY Vol. 135, No. 2

nd gastrokine-1 immunostaining localizes correctly tohe cell apex. The sum of these findings is explained mostonservatively by failure of granule filling when cellular

UC5AC levels are limiting, which suggests that Foxq1ay have evolved to fulfill a limited but essential func-

ion in pit cells. RNA microarray analysis revealed aandful of additional genes that are dysregulated inoxq1 mutant stomach, including Rpgrip, Sec22, andtk25, which have known roles in assembly and functionf membranous organelles.39 – 41 These data raise the par-icular possibility that FOXQ1 makes additional contri-utions toward assembly, stability, or the structure of pitell apical granules.

Although our data establish that Foxq1 is required foruc5ac expression in diverse tissues, the underlyingechanisms remain unresolved. Our analysis of a limited

romoter region and putative enhancers do not supporthe elementary possibility that FOXQ1 directly activates

uc5ac gene transcription. On the one hand, cell trans-ection assays may limit the ability to determine FOXQ1unctions if, for example, an essential cofactor is missingr the epigenetic state is nonpermissive. On the other hand,OXQ1 may regulate Muc5ac transcription through a dis-ant enhancer that eluded our detection or via interme-iary effectors; FOXQ1 may also be just one of severalFs that regulates Muc5ac. These possibilities will require

pecific antibodies and other tools to resolve definitively.ne group previously suggested that FOXQ1 may repress

ranscription,13 in which case, its effects on Muc5ac genexpression could be complex.

The gastric phenotype of Foxq1sb/sb and Foxq1sb/enu miceotably resembles that reported in mice with Slp2a geneutations,42 specifically in the reduced number of pit cellucous granules. Slp2a is a synaptotagmin-like protein

hat interacts selectively with Rab27, a small GTPasenown to regulate intracellular vesicle transport in di-erse cells,43 and Slp2a-Rab interaction seems to be es-ential for mucous granule formation and exocytosis.42

uantitative RT-PCR analysis of Foxq1sb/sb stomach didot uncover altered RNA levels of Slp- or Rab-familyenes (data not shown). Stable synthesis of foveolar cellucous granules thus requires at least 2 independent

rocesses: intact Slp2a-Rab function and Foxq1-depen-ent synthesis of MUC5AC and a small number of otherene products.

Elaboration of mucus is an essential function of cer-ain epithelia, and properties of the mucus secreted in

™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™igure 7. Absence of MUC5AC expression in Foxq1sb/sb conjunctiva.

A) Ocular phenotype of �50% of aged (�6 months) Foxq1sb/sb mice:ccumulation of debris (arrow) around the eye. (B–I) Histochemical anal-sis of serial tissue sections of adult murine conjunctiva from Foxq1sb/sb

B, D, F, H) and age-matched wild-type (C, E, G, I) mice. B and C: H&Etained; D and E: PAS stained; F and G: Alcian blue stained; H and I:

UC5AC immunostain. Scale bars, 60 �m.

dppgepipimaFtsSlmp

ct1

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

BA

SIC–

ALI

MEN

TARY

TRA

CT

August 2008 FORKHEAD REGULATION OF TISSUE-SPECIFIC MUC5AC 599

ifferent tissues are dictated in part by the mucinolypeptide. MUC5AC is by far the predominant mucinroduced in gastric pit cells; secretion of the mature,lycosylated product protects the epithelium from acid,nzymatic, or physical damage. Foxq1 is selectively ex-ressed in surface mucous cells, and our data argue for

ts requirement in Muc5ac gene regulation. Nearly com-lete absence of MUC5AC does not produce overt disease

n the stomach, including ulcers, inflammation, or tu-ors, which may reflect the artificial environment of

nimals maintained in the laboratory; by contrast, agingoxq1sb/sb mice develop ocular symptoms that can be at-ributed in principle to chronic MUC5AC deficiency. Inome human dry-eye (keratitis sicca) conditions such asjögren’s syndrome, reduced Muc5ac mRNA and protein

evels correlate with disease severity.44 Future studiesight thus apply Foxq1 mutant mice to investigate

athophysiology of gastric and ocular mucus deficiency.

Supplementary Data

Note: To access the supplementary material ac-ompanying this article, visit the online version of Gas-roenterology at www.gastrojournal.org, and at doi:0.1053/j.gastro.2008.014.019.

References

1. Jacobsen CM, Narita N, Bielinska M, et al. Genetic mosaic anal-ysis reveals that GATA-4 is required for proper differentiation ofmouse gastric epithelium. Dev Biol 2002;241:34–46.

2. Lee CS, Perreault N, Brestelli JE, et al. Neurogenin 3 is essentialfor the proper specification of gastric enteroendocrine cells andthe maintenance of gastric epithelial cell identity. Genes Dev2002;16:1488–1497.

3. Ramsey VG, Doherty JM, Chen CC, et al. The maturation ofmucus-secreting gastric epithelial progenitors into digestive-en-zyme secreting zymogenic cells requires Mist1. Development2007;134:211–222.

4. Karam SM. Lineage commitment and maturation of epithelialcells in the gut. Front Biosci 1999;4:D286–D298.

5. Lee HS, Lee HK, Kim HS, et al. MUC1, MUC2, MUC5AC, andMUC6 expressions in gastric carcinomas: their roles as prognos-tic indicators. Cancer 2001;92:1427–1434.

6. Longman RJ, Douthwaite J, Sylvester PA, et al. Coordinated lo-calisation of mucins and trefoil peptides in the ulcer associatedcell lineage and the gastrointestinal mucosa. Gut 2000;47:792–800.

7. Oien KA, McGregor F, Butler S, et al. Gastrokine 1 is abundantlyand specifically expressed in superficial gastric epithelium, down-regulated in gastric carcinoma, and shows high evolutionary con-servation. J Pathol 2004;203:789–797.

8. Gipson IK, Argueso P. Role of mucins in the function of thecorneal and conjunctival epithelia. Int Rev Cytol 2003;231:1–49.

9. Choi MY, Romer AI, Hu M, et al. A dynamic expression surveyidentifies transcription factors relevant in mouse digestive tractdevelopment. Development 2006;133:4119–4129.

0. Bieller A, Pasche B, Frank S, et al. Isolation and characterizationof the human forkhead gene FOXQ1. DNA Cell Biol 2001;20:555–561.

1. Choi VM, Harland RM, Khokha MK. Developmental expression ofFoxJ1.2, FoxJ2, and FoxQ1 in Xenopus tropicalis. Gene Expr

Patterns 2006;6:443–447.

2. Frank S, Zoll B. Mouse HNF-3/fork head homolog-1-like gene:structure, chromosomal location, and expression in adult andembryonic kidney. DNA Cell Biol 1998;17:679–688.

3. Hoggatt AM, Kriegel AM, Smith AF, et al. Hepatocyte nuclearfactor-3 homologue 1 (HFH-1) represses transcription of smoothmuscle-specific genes. J Biol Chem 2000;275:31162–31170.

4. Hong HK, Noveroske JK, Headon DJ, et al. The winged helix/forkhead transcription factor Foxq1 regulates differentiation ofhair in satin mice. Genesis 2001;29:163–171.

5. Li C, Hung Wong W. Model-based analysis of oligonucleotidearrays: model validation, design issues and standard error appli-cation. Genome Biol 2001;2(8):RESEARCH0032.

6. Mueller A, Merrell DS, Grimm J, et al. Profiling of microdissectedgastric epithelial cells reveals a cell type-specific response toHelicobacter pylori infection. Gastroenterology 2004;127:1446–1462.

7. Lane PW, Murphy ED. Susceptibility to spontaneous pneumonitisin an inbred strain of beige and satin mice. Genetics 1972;72:451–460.

8. Sundberg JP, Hogan ME. Handbook of mouse mutations with skinand hair abnormalities. Boca Raton, FL: CRC Press, Inc, 1994.

9. Nagle DL, Karim MA, Woolf EA, et al. Identification and mutationanalysis of the complete gene for Chediak-Higashi syndrome. NatGenet 1996;14:307–311.

0. Perou CM, Moore KJ, Nagle DL, et al. Identification of the murinebeige gene by YAC complementation and positional cloning. NatGenet 1996;13:303–308.

1. Barbosa MD, Nguyen QA, Tchernev VT, et al. Identification of thehomologous beige and Chediak-Higashi syndrome genes. Nature1996;382:262–265.

2. Ramsay M. Protein trafficking violations. Nat Genet 1996;14:242–245.

3. Spritz RA. Genetic defects in Chediak-Higashi syndrome and thebeige mouse. J Clin Immunol 1998;18:97–105.

4. Potter CS, Peterson RL, Barth JL, et al. Evidence that the satinhair mutant gene Foxq1 is among multiple and functionally di-verse regulatory targets for Hoxc13 during hair follicle differenti-ation. J Biol Chem 2006;281:29245–29255.

5. Perou CM, Pryor RJ, Naas TP, et al. The bg allele mutation is dueto a LINE1 element retrotransposition. Genomics 1997;42:366–368.

6. Harris E, Wang N, Wu WL, et al. Dictyostelium LvsB mutantsmodel the lysosomal defects associated with Chediak-Higashisyndrome. Mol Biol Cell 2002;13:656–669.

7. Huynh C, Roth D, Ward DM, et al. Defective lysosomal exocytosisand plasma membrane repair in Chediak-Higashi/beige cells.Proc Natl Acad Sci U S A 2004;101:16795–16800.

8. Jonckheere N, Van Der Sluis M, Velghe A, et al. Transcriptionalactivation of the murine Muc5ac mucin gene in epithelial cancercells by TGF-�/Smad4 signalling pathway is potentiated by Sp1.Biochem J 2004;377:797–808.

9. Carlsson P, Mahlapuu M. Forkhead transcription factors: key playersin development and metabolism. Dev Biol 2002;250:1–23.

0. Kaestner KH, Knochel W, Martinez DE. Unified nomenclature forthe winged helix/forkhead transcription factors. Genes Dev2000;14:142–146.

1. Pierrou S, Enerback S, Carlsson P. Selection of high-affinity bindingsites for sequence-specific, DNA binding proteins from random se-quence oligonucleotides. Anal Biochem 1995;229:99–105.

2. Cirillo LA, Zaret KS. An early developmental transcription factorcomplex that is more stable on nucleosome core particles thanon free DNA. Mol Cell 1999;4:961–969.

3. Friedman JR, Kaestner KH. The Foxa family of transcription fac-tors in development and metabolism. Cell Mol Life Sci 2006;63:2317–2328.

4. Dartt DA. Control of mucin production by ocular surface epithelial

cells. Exp Eye Res 2004;78:173–185.

3

3

3

3

3

4

4

4

4

4

Ds5

gtF

lkfl

BA

SIC–

ALIM

ENTA

RY

TRA

CT

600 VERZI ET AL GASTROENTEROLOGY Vol. 135, No. 2

5. Beck S, Sommer P, dos Santos Silva E, et al. Hepatocyte nuclearfactor 3 (winged helix domain) activates trefoil factor gene TFF1through a binding motif adjacent to the TATAA box. DNA Cell Biol1999;18:157–164.

6. Hanby AM, Poulsom R, Singh S, et al. Spasmolytic polypeptide isa major antral peptide: distribution of the trefoil peptides humanspasmolytic polypeptide and pS2 in the stomach. Gastroenterol-ogy 1993;105:1110–1116.

7. Karam SM, Tomasetto C, Rio MC. Trefoil factor 1 is required forthe commitment programme of mouse oxyntic epithelial progen-itors. Gut 2004;53:1408–1415.

8. Velcich A, Yang W, Heyer J, et al. Colorectal cancer in mice geneti-cally deficient in the mucin Muc2. Science 2002;295:1726–1729.

9. Kaiser CA, Schekman R. Distinct sets of SEC genes governtransport vesicle formation and fusion early in the secretorypathway. Cell 1990;61:723–733.

0. Zhao Y, Hong DH, Pawlyk B, et al. The retinitis pigmentosaGTPase regulator (RPGR)-interacting protein: subserving RPGRfunction and participating in disk morphogenesis. Proc Natl AcadSci U S A 2003;100:3965–3970.

1. Preisinger C, Short B, De Corte V, et al. YSK1 is activated by theGolgi matrix protein GM130 and plays a role in cell migrationthrough its substrate 14-3-3�. J Cell Biol 2004;164:1009–1020.

2. Saegusa C, Tanaka T, Tani S, et al. Decreased basal mucussecretion by Slp2-a-deficient gastric surface mucous cells. Genes

Cells 2006;11:623–631.

3. Fukuda M. Rab27 and its effectors in secretory granule exocyto-sis: a novel docking machinery composed of a Rab27 effectorcomplex. Biochem Soc Trans 2006;34:691–695.

4. Argueso P, Balaram M, Spurr-Michaud S, et al. Decreasedlevels of the goblet cell mucin MUC5AC in tears of patientswith Sjogren syndrome. Invest Ophthalmol Vis Sci 2002;43:1004–1011.

Received September 4, 2007. Accepted April 13, 2008.Address requests for reprints to: Ramesh A. Shivdasani, MD, PhD,

ana-Farber Cancer Institute, 44 Binney Street, Boston, Massachu-etts 02115. e-mail: ramesh_shivdasani@dfci.harvard.edu; fax: (617)82-8490.Supported by grant R01DK61139 (to R.A.S.) and fellowship training

rants T32DK007533 (to A.H.K.) and T32DK07477 (to M.P.V.) fromhe National Institutes of Health and a Crohn’s and Colitis Foundationellowship (to M.P.V.).The authors thank Luanne Peters, Qiang Liu, and members of our

aboratory for critical review of the manuscript; Karin Oien for gastro-ine-1 antiserum; David Alpers for antibody against gastric intrinsicactor; and Adam Bass and Dan Podolsky for gifts of gastric cancer cellines.

Conflicts of interest: No conflicts of interest exist.

M.P.V. and A.H.K. contributed equally to this study.