Identification of Novel Pancreas-specific Regulatory Sequences in the Promoter Region of Human Pancreatic Secretory Trypsin Inhibitor Gene*

The human pancreatic secretory trypsin inhibitor (PSTI) genes introduced into mice are specifically expressed in pancreas. The 1.0 kilobase pairs of PSTI 5 * flanking sequence directed preferential expression of a linked reporter chloramphenicol acetyltransferase, which was active in a PSTI-expressing pancreatic cell line (AR42j) but not in a PSTI-nonexpressing fibroblast cell line (XC). Two positively acting elements were found, Region I ( 2 161/ 2 116) and Region II ( 2 103/ 2 74), as defined by transfection and binding assays with AR42j cells. Region II is sufficient for the pancreas-specific expression, but the presence of both Regions I and II is needed for the maximum activity. Sequence studies also revealed that these two elements differ from the previously identified recognition sequence for pancreas transcription factor 1 (PTF1). When the same set of experiments was done with XC cells, one negatively acting element was identified, Region IV ( 2 154/ 2 137). Interest-ingly, Regions I and IV kbp, 2 2 exonuclease III diges- tion Acc I site PSTI( Acc I)CAT. 5 9 -deletion clones, PSTI(245)CAT, and PSTI(98)CAT, by inserting polymerase chain reaction products from 2 245, 2 159, 2 141, 2 115, and 2 98 to 2 1, respectively, into pBS0CAT as follows. Sac I and Sma I restriction sites were linked at the 5 9 termini of forward and reverse primers for polym- erase chain reaction, respectively. Each fragment amplified with these primers was inserted into the Sac I- Sma I restriction site of pBS0CAT after digestion with Sac I and Sma I. All constructs were confirmed by sequencing. The polymerase chain reaction forward primer sequences were 5 9 -CGAGCTCGCCTGACAGAATCTTTGCCTTGC for PSTI(245)-CAT,

The differentiation and maintenance of phenotype in eukaryotic cells depends on the regulated expression of selected genes, most of which relies on interaction of specific regulatory protein(s) (transcription factor) and its cognate cis-acting DNA elements including a promoter and enhancer(s).
A cis-acting element has been shown to be involved in the control of the expression of pancreas-specific genes, such as mouse ␣-amylase 2, elastase 2, trypsin, rat elastase I, and chymotrypsin B. This element sequence lies 300 bp 1 upstream from the cap site in the 5Ј-flanking region of these genes (1)(2)(3)(4), has a consensus sequence GXCX(A/C)TGGGAAAX n CTX-CAG(G/C)TGTG(C/A)TX, and consists of bipartite binding regions (Box A and Box B) separated by one or two DNA turns.
This sequence interacts with a pancreas-specific transcription factor 1 (PTF1) (5), which consists of a 64-kDa protein subunit that binds to Box A and a 48-kDa protein subunit that binds to Box B (6). This dimer complex is further associated with a 75-kDa protein that does not directly bind to DNA (7) but acts in transporting the complex into the nucleus. Besides PTF1, several other trans-acting factors have been discovered, such as exocrine pancreas transcription factor 1 (8), which interacts with the (CACCTGX n TTTCCC) motif and Pan/E12, E47, and AP4, which interact with the CAGCTG motif in Box B (9 -11). Thus, the expression of pancreas-specific secretory proteins in acinar cells has been thought to employ essentially the same cis-acting regulatory elements.
Pancreatic secretory trypsin inhibitor (PSTI), consisting of 56 amino acids (molecular mass 6.2 kDa) (12), was first isolated from bovine pancreas in 1948 by Kazal et al. (13) and has been thought to be the cognate inactivation factor for preventing intra-pancreatic trypsin activity. In mammals, it is secreted from pancreatic acinar cells into the pancreatic juice, and naturally, this gene expression control was expected to involve the pancreas-specific common cis-acting regulatory sequence. However, the structure of the genomic PSTI gene, as determined by Horii et al. (14), has neither the consensus sequence nor the typical TATA nor CAAT promoter sequence. Therefore, we initiated an analysis of the 5Ј-sequences flanking the human PSTI gene for identifying the cis-acting elements involved in this pancreas-specific expression through deletion studies and protein binding studies. We show here that the PSTI gene expression control involves two novel 5Ј cis-acting elements defined by their effect on transcription and their interaction with a protein(s) from pancreatic cells. In addition, competition binding assays showed that one of the core sequences of the two elements is CAATCAATAAC, which is unique and can interact with another protein(s) from nonpancreatic cells; this interaction is likely to repress the expression of PSTI gene in nonpancreatic cells. These results suggest that this unique sequence (CAATCAATAAC) functions as a positively acting element in pancreatic cells and a negatively acting element in nonpancreatic cells in the control of the expression of the human PSTI gene.

EXPERIMENTAL PROCEDURES
Identification of the Integrated PSTI Gene-A 27-kbp human PSTI genomic DNA that contains 8.3 kbp of the 5Ј-flanking sequence has been cloned into pWE16 by Horii et al. (14). The DNA for transgenic mice was extracted from this cosmid clone, named cosTI4, by NotI and KpnI digestion and followed by agarose gel electrophoresis.
Production and Identification of Transgenic Mice-Human PSTI genomic DNA in cosTI4 (a NotI-KpnI fragment, see Fig. 1) was microinjected into the male pronucleus of fertilized BDF1 mouse eggs that * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
were allowed to develop to term. Mouse pups were screened for the presence of the injected gene by genomic Southern blotting using 5 g of tail DNA digested with BamHI and hybridization with a 32 P-labeled human PSTI cDNA probe (15,16).
Isolation of Total and Poly(A) ϩ RNA and Northern Blot Analysis-Total cellular RNA was extracted from source cells using acid guanidinium/thiocyanate/phenol/chloroform (17). Poly(A) ϩ RNA was purified from total RNA by repeated passage through an oligo(dT) cellulose (type 7, Amersham Pharmacia Biotech) column. Total RNA or poly(A) ϩ RNA was denatured by heating at 65°C for 15 min in 50% (v/v) formamide and resolved by electrophoresis in a 1% agarose, 2.2 M formaldehyde gel then Northern blotted as described by Tomita et al. (16).
Cell Lines and Tissue Culture-AR42j is a rat pancreatic exocrine cell line that expresses PSTI (18), and XC (19) and RAT2 (20) are rat fibroblast cell lines that do not express PSTI. These were obtained from Dainippon Pharmaceutical Co. (Suita, Japan). The cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 500 units/ml penicillin, and 100 g/ml streptomycin (Life Technologies, Inc.) at 37°C under a 5% CO 2 atmosphere. The XC cells were similarly cultured, except for supplementation with 1% nonessential amino acids.
PSTIdel(103-78)CAT was constructed as follows. The Ϫ159/Ϫ104 and the Ϫ77/Ϫ1 fragments of the 5Ј-flanking region of human PSTI gene were produced by polymerase chain reaction. A BamHI restriction site was linked at the 5Ј terminus of the reverse primer for the Ϫ159/ Ϫ104 fragment and the forward primer for the Ϫ77/Ϫ1 fragment. These fragments were ligated at BamHI sites. In addition, XbaI and SmaI restriction sites were linked at the 5Ј terminus of the forward primer for the Ϫ159/Ϫ104 fragment and the reverse primer for the Ϫ77/Ϫ1 fragment. The fragment ligated at BamHI was inserted in between XbaI and SmaI of pBS0CAT. The forward primer sequence for the Ϫ159/ Ϫ104 fragment was the same as that for PSTI(159)CAT, and the reverse primer was 5Ј-CGGGATCCCGGCCCCACCTACTGGGCTATA. The forward primer sequence for the Ϫ77/Ϫ1 fragment was 5Ј-CGG-GATCCCGTCCAGTCCCAGGCTTCTGAA, and the reverse primer was the same as that for PSTI(159)CAT.
RSV CAT was constructed by inserting 524 bp of the Rous sarcoma virus 3Ј long terminal repeat, substituting for the fragment between the AccI and HindIII sites of PSV0CAT (23).
DNA Transfection and CAT Assay-AR42j cells (2.5 ml of 5 ϫ 10 5 cells/ml) were plated in 6-cm diameter dishes 24 h before transfection. Four g of CAT reporter plasmid and 2.5 g of pCH110, a lacZ expression plasmid (Amersham Pharmacia Biotech) were cotransfected by liposome-mediated transfection (24) using 19.5 l of Lipofectin (Life Technologies, Inc.) in triplicate. Eighteen h later, the cells were washed twice and incubated in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Sixty h later, the cells were collected, and CAT activity was assayed according to the procedure of Gorman et al. (21). The reaction mixture containing 70 l of 100 l of AR42j cell extract, 10 l of acetyl CoA, and 1 l of [ 14 C]chloramphenicol was incubated at 37°C for 2 h, then another 10 l of acetyl CoA was added, and the mixture was incubated for a further 10 h.
XC and RAT2 cells (2.5 ml of 1 ϫ 10 5 cells/ml) were plated in 6-cm diameter dishes 12 h before transfection. One and a half g of CAT reporter plasmid and 2.5 g of pCH110 (25) were cotransfected in triplicate by 5 and 10 l of Lipofectin for 18 h in XC and RAT2 cells, respectively. Forty h after transfection, the cells were similarly assayed, except that 60 l of cell extracts was incubated with 4 l of acetyl CoA and 2 l of [ 14 C]chloramphenicol at 37°C. Another 1 l of acetyl CoA was added after 2 h, and the mixture was incubated for an additional 2 h. The conversion ratio of [ 14 C]chloramphenicol was measured with an image analyzer (BAS2000; Fujix, Tokyo, Japan) and normalized by ␤-galactosidase activity (25).
Preparation of Nuclear Extracts-Crude nuclear extracts from AR42j and XC cells were prepared essentially as described by Dignam et al. (26). All steps were performed at 4°C. The cells (3 to 5 ϫ 10 8 ) were collected from monolayer cultures, washed with phosphate-buffered saline, and pelleted by centrifugation for 5 min at 700 ϫ g. The cell pellets were resuspended in washing buffer (10 mM Tris-HCl, pH 7.5, 130 mM NaCl, 5 mM KCl, and 8 mM MgCl 2 ) to a final volume of 5 times that of the packed cells. The cells were precipitated by centrifugation for 5 min at 700 ϫ g and resuspended in hypotonic buffer (20 mM HEPES-KOH, pH 7.9, 5 mM KCl, 0.5 mM MgCl 2 , 0.5 mM dithiothreitol (DTT), and 0.5 mM phenylmethylsulfonyl fluoride (PMSF), to a final volume of 3 times that of the original packed cells and were allowed to swell for 10 min on ice. The swollen cells were transferred to a glass Dounce homogenizer and disrupted slowly by 20 strokes. The nuclei were collected by centrifugation for 10 min at 700 ϫ g. The nuclear pellets were resuspended in an equal volume of extraction buffer (20 mM HEPES-KOH, pH 7.9, 25% glycerol, 500 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, pH 8.0, 0.5 mM DTT, 0.5 mM PMSF, 0.5 g/ml pepstatin A, and 1.3 g/ml spermidine), stirred gently for 1 h, then precipitated by centrifugation at 45,000 ϫ g for 30 min in a swing-out rotor. The supernatant was collected as the nuclear extract and dialyzed against binding buffer (20 mM HEPES-KOH, pH 7.9, 0.5 mM EDTA, 100 mM KCl, 10% glycerol, 0.5 mM DTT, and 0.5 mM PMSF) for 12 h. After the pellets were removed by centrifugation at 11,000 ϫ g for 30 min, the supernatant was collected as the nuclear extract, concentrated with polyethylene glycol 20000 to 5 mg/ml, and stored at Ϫ80°C.
DNase I Footprinting Assay-A probe containing the nucleotides Ϫ179 to Ϫ1 of the human PSTI gene was excised from PSTI(179)CAT by digestion with HindIII (for a coding strand) or BssHII (a noncoding strand), end-labeled with [␣ 32 P]dCTP by the Klenow fragment, then digested again with BssHII or HindIII. The single end-labeled DNA probe was run in a 5% polyacrylamide gel, and the band was eluted by the crush and soak method described by Maxam and Gilbert (27). The end-labeled fragment (3 fmol; specific activity, 3 ϫ 10 5 cpm) was incubated with 20 ng of nuclear protein for 30 min at room temperature in a volume of 50 l in the presence of 12 mM HEPES-KOH, pH 7.9, 60 mM KCl, 0.12 mM EDTA, 1 mM MgCl 2 , 12% glycerol, 1 mM DTT, and 1 mM PMSF with 1 or 2 g of poly(dI-dC). The MgCl 2 concentration was raised to 3 mM, DNase I was added to a final concentration of 5 g/ml, and the mixture was digested for 1 min at room temperature. Digestion was terminated by stop buffer (100 mM Tris-Cl, pH 7.9, 100 mM NaCl, 1% SDS, 10 mM EDTA, and 25 g/ml calf thymus DNA). Samples were digested with 50 g/ml proteinase K for 30 min at 37°C and extracted with phenol/chloroform; nucleic acids were then precipitated with ethanol. DNase I-resistant material was dissolved in 6 l of sequencing dye and analyzed on 6% sequencing gels after heating at 90°C for 5 min. The dried gel was autoradiographed at Ϫ70°C. Nucleotide sequence markers (G ϩ A) were prepared by the Maxam-Gilbert reaction (27).
Preparation of DNA Probes for Gel Mobility Shift Assay-Wild-type and individual 10-bp substitution mutant probes (Ϫ161/Ϫ116) were prepared as follows. Oligonucleotides (33-mer) were synthesized from the 5Ј-end of coding and noncoding strands, using a DNA synthesizer (Applied Biosystems, Foster, CA). Labeled double-stranded DNA probes were formed by annealing these synthesized oligonucleotides in 10 mM Tris-HCl, pH 7.9, and 5 mM MgCl 2 , then filling in the 3Ј-recessed termini using Sequenase (Sequenase® Ver 2.0, U. S. Biochemical Corp.) in the presence of 1 mM dNTP and [␣ 32 P]dCTP (3000 Ci/mmol). The products were purified by electrophoresis on an 8% polyacrylamide gel. Wild-type and individual 10-bp substitution mutant probes (Ϫ103/Ϫ74) were prepared as follows. Oligonucleotides of coding and noncoding strands were synthesized and then end-labeled with [␥-32 P]ATP (3000 Ci/mmol) using bacteriophage T4 polynucleotide kinase. Doublestranded DNA probes were formed by annealing as described above. The PTF1 binding sequence probes of mouse ␣-amylase (Ϫ158/Ϫ122) and trypsin a (-128/-96) genes (5) were prepared by annealing each pair of synthesized complementary oligonucleotides.
Gel Mobility Shift Assay-The labeled probe (specific activity: 15,000 cpm) and 10 g of crude nuclear extract containing 4 g of poly(dI-dC) were incubated for 30 min at room temperature in 22 l of binding buffer (12 mM HEPES-KOH, pH 7.9, 60 mM KCl, 0.12 mM EDTA, 1 mM MgCl 2 , 12% glycerol, 1 mM DTT, and 1 mM PMSF) in the presence of 4 g of poly(dI-dC). The DNA-protein complex was resolved by electrophoresis in a 4% polyacrylamide gel using 0.5 ϫ Tris borate/EDTA buffer at constant current of 25 mA at 4°C after a pre-run at a constant voltage of 150 V for 90 min. The dried gel was autoradiographed at Ϫ70°C.

RESULTS
The Human PSTI Gene Is Specifically Expressed in the Pancreas of Transgenic Mice-Transgenic mice were prepared by introducing a human PSTI gene containing the 8.3 kbp of the 5Ј-flanking sequence (Fig. 1A). One line that contains the fulllength of the introduced DNA was named 11-4L. Along with this line, we obtained lines 12 and 24 that had integrated truncated sequences, 4.5 and 1.0 kbp respectively, in the 5Јflanking region. The DNA integrated in these animals are shown in Fig. 1B. Aside from the truncation in the distal portion of the 5Ј region, the other regions of the gene were intact. The human PSTI gene, as examined by Northern blots, was expressed in the pancreas of the three transgenic mouse lines despite the differences in the 5Ј-flanking region. The 11-4L mice, which had integrated the intact DNA, expressed the PSTI gene at an extremely high level in the pancreas (Fig.  1C). The low level expression detected in the stomach, kidney, and lung reflected that seen in human orgens (28) (data not shown). With the mouse lines 12 and 24, the expression of this gene, whose levels were lower than that of 11-4L, was detected only in the pancreas ( Fig. 1C and 1D). These data show that the human PSTI gene expression is regulated in the mouse pancreas at the transcriptional level and that its control element(s) is conserved even in the line 24 mouse, which contains only 1.0 kbp of the 5Ј-flanking sequence hooked to the 7.5 kbp of the coding sequence followed by 11.2 kbp of the 3Ј-flanking sequence.
The 5Ј-Flanking Sequence of the Human PSTI Gene Directs the Preferential Expression in the Pancreatic Tumor Cell Line-The genes of pancreas secretory proteins such as amylase, trypsin, elastase I, and chymotrypsin have a similar cis-acting DNA element in the 5Ј-flanking region, and this sequence has been known to be responsible for their pancreas-specific expressions (1)(2)(3)(4). To see whether the 1.0-kbp 5Ј sequence of human PSTI gene were able to drive expression of a reporter gene, as shown in the transgenic mouse line 24 (Fig. 1D), a construct of PSTI(1.0)CAT was made in promoterless plasmid pBS0CAT vector with 1.0 kbp of the 5Ј sequence from genomic fragment of the PSTI gene. The AR42j cell line (a rat pancreatic acinar cell carcinoma) expressing amylase, trypsin, and chymotrypsin genes (29) was confirmed to be able to express RNAs for rat PSTI gene before transfection studies (data not shown). PSTI(1.0)CAT showed preferential expression in AR42j cells, as shown in Table I. The promoterless pBS0CAT was used as a negative control, and the CAT activities were normalized to the activity of Rous sarcoma virus CAT (RSVCAT) directed by the Rous sarcoma virus promoter to correct the differences in transfection efficiency. As controls, the same set of experiments was done with rat fibroblast cell lines XC and RAT2 not expressing the PSTI gene. As expected, Rous sarcoma virus CAT was highly expressed in all of these cell lines, and pBS0CAT showed the lack of expression in all of them. PSTI(1.0)CAT was active in AR42j cells but was inactive in XC and RAT2 cells. The cell-type specificity was 27-fold in AR42j. Thus, both the studies with transgenic mouse line 24 and this transfection study demonstrated the presence of the control region for the pancreas-specific expression of the human PSTI gene in the 1.0-kbp region of the 5Ј-flanking sequence. With human pancreas, only 0.1 g of mRNA was used as a positive control. In both of the panels, blots were probed with 32 P-labeled human PSTI cDNA (15,16). The 5Ј-Sequence Contains the Cell Specific Cis-acting Elements-To examine the functionally significant sequences in the 5Ј-flanking region of the human PSTI gene, we constructed a set of deletion derivatives from PSTI(1.0)CAT by stepwise elimination of the 5Ј border and transfected them into AR42j cells. The results are shown in Fig. 2. PSTI(320)CAT, which possesses 320 bp of the 5Ј-flanking sequence, showed activity identical to that of the starting plasmid (data not shown), demonstrating that the specific regulatory element(s) must be located downstream of the Ϫ320 site. Deletions extending to Ϫ159 showed no measurable effect (activity of around 80-fold that of the vector plasmid), whereas deletions that extended to Ϫ141 and Ϫ134 resulted in a significant reduction of activity. Thus, the 5Ј border of the positively acting cis-element lies between Ϫ159 and Ϫ141. Upon further deletion to Ϫ115, the activity was restored (activity of about 70-fold), but this augmenting activity was negated when a region covering between Ϫ134 and Ϫ115 was added or a deletion extended further to Ϫ98. However, some weak activities (around 10-fold), which seem to be basal level activities, remained. These observations suggest that there is a negatively acting element within Ϫ134 and Ϫ115 and another positively acting element downstream of Ϫ115, the effect of which is eliminated with the Ϫ98 construct. Deletion to Ϫ77 resulted in a further loss of activity from 10-to 3-fold, and with the deletion extended to Ϫ34, the activity was completely eliminated. Thus, the loss of activities with Ϫ77 and Ϫ34 constructs can be regarded as the loss of an element regulating basal level activity and the promoter for transcription initiation, respectively, because the human PSTI gene lacks a CAAT box or TATA box and has multiple transcription start points (14), but the major site for transcription is at Ϫ60 (16). Altogether, these results suggest that the 5Ј sequence downstream of Ϫ179 is crucial for human PSTI gene expression in AR42j cells and contains an upstream positively acting element followed (or overlapped) by a negatively acting element and then a downstream positively acting element and a basal activity-regulating element just in front of the promoter for transcription initiator.
Multiple Sites for Protein Interaction in the 5Ј-Flanking Region of the Human PSTI Gene-To examine trans-acting factors in the 5Ј region of the human PSTI CAT genes, we carried out in vitro DNase 1 footprinting assays using nuclear extracts from the PSTI-expressing AR42j cells or nonexpressing XC cells. The assays were done with both coding and noncoding strands. With the coding strand, three protected regions appeared in a dose-dependent fashion in AR42j nuclear extracts (Fig. 3A) and one protected region appeared in XC nuclear extracts (Fig. 3B): Region I (Ϫ161/Ϫ116), Region II (Ϫ103/ Ϫ74), Region III (Ϫ65/Ϫ42) in AR42j, and Region IV (Ϫ154/ Ϫ137) in XC. With the noncoding strand, the binding profile was similar with Regions I and IV but weak with Region II and none with Region III (Fig. 3B). Comparing the protection profiles of AR42j and XC, both Regions I and II are AR42j-specific, and Region IV is XC-specific, because they show the cell-typespecific protection. Region III is also AR42j-specific but is likely to be the promoter region, judging from the results of the transfection assays and the transcription start point of the PSTI gene.
Region I is divided into two subregions, Region IA (Ϫ161/ Ϫ136) and Region IB (Ϫ131/Ϫ116), separated by a 5-bp nonprotective gap sequence. The left end of Region IA (Ϫ161) is almost identical to the 5Ј border of the upstream positively acting element (Ϫ159/Ϫ141) in Fig. 2, and the left end of Region IB (Ϫ131) lies within the Ϫ134 to Ϫ115 region, where the negatively acting element is located. The left end of Region II (Ϫ103) lies within the Ϫ115 to Ϫ98 region, where the 5Ј border of the downstream positively acting element exists.
To examine the roles of Regions I and II, we checked the influences of the presence or absence of Regions I and/or II on CAT activities in AR42j cells (Fig. 4). PSTI(159)CAT that contains both Regions I and II was used as a reference; although its 5Ј region is 2 bp inside Region I, this did not affect the results. PSTI(115)CAT without Region I showed about half of the activity. PSTIdel(104-78)CAT without Region II showed only marginal activity, although it retained Region I. PSTI(77)CAT without Regions I and II showed no activity at all. Thus, Region II is sufficient for the pancreas-specific expression of the PSTI gene, but for the maximum activity, the presence of the Region I is needed. Region I in combination with the promoter is inactive.
To check the interaction of Regions I and II, the deletion set was then used for DNase 1 footprinting assays (Fig. 5). Deletions from Ϫ179 to Ϫ134 did not affect the protections. However, deletion extending to Ϫ115, completely eliminating Region I, weakened the protection of the 5Ј portion (Ϫ103/Ϫ95) of Region II. These results suggest, although they do not prove, that the binding proteins for Regions I and II are likely to have some mutual influence. But they also indicate that Region II seems to consist of two subregions, Region IIA (Ϫ103/Ϫ95) and Region IIB (Ϫ94/Ϫ74). Thus, the 5Ј border of the downstream positively acting element is found to be involved in the Region IIA, and also the left end of Region IIB lies within the Ϫ98 to Ϫ77 region, where there is the 5Ј border of the basal activityregulating element.
Region IV protection (Ϫ154/Ϫ137) was observed with coding and noncoding strands and was found to be specific to the PSTI-nonexpressing XC cell nuclear extract. Region IV seemed to be overlapping with the negatively acting element. In accordance with this idea, the transfection assays in XC cells using the same set of deletion mutants (Fig. 2) showed essen-

FIG. 2. CAT assays with deletion mutants with progressive loss of the 5 border.
Serial deletion mutants containing various 5Ј-flanking sequences of PSTI gene with the 5Ј border deletion ends at Ϫ320, Ϫ245, Ϫ179, Ϫ159, Ϫ141, Ϫ134, Ϫ115, Ϫ98, Ϫ77, Ϫ34, and 0 bp were transfected into the PSTI-expressing AR42j cells and the nonexpressing XC cells by lipofection (24), and CAT activities were measured. Each activity was normalized by ␤-galactosidase activity (25) and is shown as the fold increase against that of the vector. tially no activity with Ϫ320, Ϫ240, Ϫ179, and Ϫ159 constructs, which have Region IV, whereas there was a weak but nonnegligible 8-to 17-fold activity with Ϫ141 through Ϫ77 constructs, which do not have it.
The Binding Sequences Are Different from That for PTF1-To further examine the nature of the footprint sequences, we performed gel mobility shift and competition assays using 32 P end-labeled synthetic oligonucleotides that correspond to Region I (Ϫ161/Ϫ116) or Region II (Ϫ103/Ϫ74). The Regions I and II oligonucleotides yielded two and three complexes in the binding reaction with and AR42j nuclear extracts, respectively (A 1 and A 2 for Region I and A 3 , A 4 , and A 5 for Region II in Fig. 6). Among these, the A 1 , A 3 , and A 4 complexes were affected in the presence of excess unlabeled homologous oligonucleotides, resulting in a disappearance of A 1 complexes and a dose-dependent inhibition of the formation of the A 3 and A 4 complexes, whereas the A 2 and A 5 complexes showed no significant change. A similar set of binding reactions with XC nuclear extracts and Region IV oligonucleotides was performed. Four complexes were yielded, X 1 -X 4 (right panel of Fig.   6), none of which matched A 1 -A 3 complexes, and which are likely to involve different protein factors from those of AR42j nuclear extracts. The X 1 , X 2 , and X 3 complexes were competed with excess unlabeled homologous oligonucleotides, whereas X 4 complex did not. Furthermore, XC nuclear extracts formed three complexes in the binding reaction with Region II oligonucleotides, although protein binding in Region II was not detected in footprinting assay (Fig. 3). As these are indistinguishable from A 3 -A 5 complexes with AR42j nuclear extracts, the data are omitted from Fig. 6 for clarity. The binding proteins for Region II are likely to be cell-type-nonspecific proteins and to be repressed in the footprinting assay with the XC nuclear extracts. Thus, A 1 for Region I, A 3 and A 4 for Region II, and X 1 , X 2 , and X 3 for Region IV are the sequence-specific protein-binding complexes, whereas A 2 , A 5 , and X 4 are not.
To analyze whether these sequence-specific-binding proteins involve the previously identified PTF1 (5), which is the major pancreas-specific transcription factor, competition gel mobility shift assays were performed in the presence of the PTF1 binding sequences for mouse ␣-amylase (Ϫ158/Ϫ122) and trypsin a (Ϫ128/Ϫ96) genes. The PTF1 oligonucleotides formed two complexes in AR42j nuclear extracts as reported by Cockell et al. (5) (Fig. 7A). However, the A 1 , A 3 , A 4 , and X 1 , X 2 , and X 3 complexes were different in mobilities from those PTF1 complexes and did not compete with a 200-fold molar excess of the PTF1 oligonucleotides (Fig. 7B). These results demonstrate that PTF1 is not the factor involved in the binding with Regions I, II, and IV of the PSTI gene.
The Core Binding Sequence CAATCAATAAC Plays the Most Important Role in the Cell-specific Expression of Human PSTI Gene-To further analyze the recognition sequences in Regions I and II, we synthesized a set of mutant competitor oligonucleotides (mut probes) in which a stretch of 10-bp sequence was substituted by a target sequence for XbaI or EcoRV (Fig. 8A). The results are demonstrated in Fig. 8B. The binding assays with AR42j nuclear extracts showed that the formation of A 1 complex was inhibited with three of the mutant competitors and not with mut (148 -139). The formation of A 3 and A 4 complexes was competed with one of the mutant competitors, but FIG. 3. DNase I footprinting analysis of the 5-flanking region of the human PSTI gene. The Ϫ179/Ϫ1 fragments were labeled with 32 P at the 3Ј-end of coding (A) and at the 5Ј-end of noncoding (B) strands, respectively, and were incubated in the presence of increasing concentrations (0, 5, 10, 20 g) of AR42j nuclear extracts or 20 g of XC nuclear extract. Protein-DNA complexes were partially digested with DNase I, and the digestion products were separated by electrophoresis in a 6% sequencing gel. Maxam and Gilbert (27) G ϩ A chemical cleavage reactions served as size markers. DNA regions that were protected by nuclear protein(s) are indicated by open boxes on the right side of each autoradiograph with nucleotide positions representing the footprint boundaries and have been termed Regions I, II, III, and IV. Region I can be divided into two subregions (termed IA and IB) as described in the text. the mut(93-84) and mut(83-74) did not show the competition. Thus, the most important recognition sequences, as reflected by these noncompeting oligonucleotides, lie in the region between Ϫ148 and Ϫ139 in Region I, which is involved in Region IA (Ϫ161/Ϫ136), and the region between Ϫ93 and Ϫ74 in Region II, which is identical to the Region IIB (Ϫ93/Ϫ74) (Fig. 8C). In these studies, a recognition sequence was not detected in Regions IB and IIA, because these protections were weaker than those of Regions IA and IIB in the DNase 1 footprinting assay. In the binding studies using XC nuclear extracts, the formation of X 1 , X 2 , and X 3 complexes were competed weakly with mut(158 -149) and completely with mut(138 -129) and mut(128 -119), but the mut(148 -139) did not show competing ability. The competition of complex formation in Region II is the same as that in AR42j. Thus, the sequence between Ϫ148 and Ϫ139 is likely the most important target in Region IV. Interestingly, this core sequence overlaps the crucial binding sequence in Region I in the AR42j nuclear extract, although the binding proteins are different. The region between Ϫ149 and Ϫ139 containing this core sequence has the characteristic sequence CAATCAATAAC (Fig. 8C). This unique sequence is important in both PSTI-expressing and nonexpressing cells and is likely to function as a positively and negatively acting element in AR42j and XC cells, respectively.

DISCUSSION
The present study revealed that the PSTI gene, which is expressed selectively in the pancreas, involves a novel recognition sequence, CAATCAATAAC, and possibly novel proteins in the regulation of its expression. The role of the PSTI gene is to inhibit the activity of trypsin when the activation of trypsinogen occurs in the pancreatic duct. Thus, to protect the pancreas from the risk of autodigestion, it may be important to have a system in which excess PSTI is expressed and secreted to the pancreatic duct, especially because the PSTI is not a very potent inhibitor of trypsin (30). It is not surprising, therefore, that the regulation of PSTI gene expression is different from that of pancreatic exocrine enzymes involving trypsin.
Analyses of transgenic mice that have integrated the human PSTI gene showed that the expression of human PSTI gene is maintained in the pancreas of transgenic mice that contains the 5Ј-truncated flanking sequence. The relevant cis-acting control element was shown to reside in the 1.0-kbp genomic region 5Ј to the transcription initiation site. The conclusion obtained with transgenic animals has been supported by an independent systems using CAT assay system with a rat pancreas-derived cell line (AR42j) but not with fibroblast-derived cell lines. The cis-acting control elements are located in a relatively short region immediately upstream of the PSTI gene, consistent with the other known pancreas-specific genes (5, 8 -11).
The present detailed analyses demonstrated that the control region consists of Region I (Ϫ161/Ϫ116), Region II (Ϫ103/Ϫ74), Region III (Ϫ65/Ϫ42), and Region IV (Ϫ154/Ϫ137) and further-  7. Gel mobility shift assays using PTF1 binding sequence as a probe or a competitor. A, ten g of AR42j nuclear extract was incubated with radiolabeled synthetic oligonucleotides corresponding to the PTF1 binding sequence. Lanes 1 and 2, incubation of AR42j nuclear extracts with PTF1 binding sequences for mouse ␣-amylase (Amy) (Ϫ158/Ϫ122) and mouse trypsin a (Trp) (Ϫ128/Ϫ96) genes (5), respectively. Closed triangles indicate nonspecific complexes, and an open triangle indicates unbound DNA. B, competition binding analyses using the PTF1 binding sequences, which were the same as those in A. Region I probe was used as the Region IV probe in the XC assay (see legend to Fig.  6). End-labeled Regions I or II probe was incubated with 10 g of crude nuclear extracts from AR42j or XC in the presence or absence of a 200-fold molar excess of the PTF1 binding sequence as a competitor probe. Symbols are the same as those in A. more, that Regions I and II are composed of two subregions, a positively active Region IA (Ϫ161/Ϫ136) and a negatively active Region IB (Ϫ131/Ϫ116), and a positively active Region IIA (Ϫ103/Ϫ95) and a basal activity-regulating Region IIB (Ϫ93/Ϫ74), respectively, as summarized in Fig. 9. Region III is the promoter region, because the major transcription start point of the human PSTI gene is at Ϫ60 bp. Transfection assays revealed that Region II is an indispensable element responsible for the pancreatic cell-specific expression of the human PSTI gene, whereas Region I, especially Region IA, is an element augmenting the promoter activity only in combination with Region II. So both Regions I and II are needed to show the maximal activation. A similar additive effect has been observed with immunoglobulin heavy chain enhancers that contain four motifs (31) as well as with the simian virus 40 enhancer that contains three segments (32). This functional additivity may involve an interaction of the multiple elements with a cognate protein(s), culminating in the production of an active enhancer, as discussed by Schaffner et al. (33).
The gel mobility shift competition assays using AR42j nuclear extracts showed that several nuclear factors bind to these elements. In Region I, the A 1 complex, which binds with positively active Region IA, was revealed, but a binding protein for negatively acting Region IB could not be detected (Figs. 8 and  9). The A 1 complex was expected to involve the positive transcription factor, for example PTF1, similar to other pancreatic genes. However, this complex was revealed not to involve PTF1 by following three points. 1) The mobility of this protein is different from that of PTF1, 2) this protein does not compete with PTF1 for the recognition sequence, and 3) the core binding sequence is CAATCAATAAC (Ϫ149/Ϫ139), which is different from PTF1 binding sequence (Figs. 7 and 8). We do not know at present the nature of the protein involved in the formation of A 1 complex. There is a possibility that a novel transcription factor may participate. However, because the core recognition sequence is related to CCAAT, which is recognized by C/EBP, NF1, CP1, and CP2 or ␣CBF, the possibility of one or several of these factors being involved cannot be ruled out.
Region II is the element that is sufficient for pancreasspecific expression of the human PSTI gene. The gel mobility shift competition assays revealed two complexes, A 3 and A 4 , which bind with Region IIB but could not show that there was a binding factor for positively acting Region IIA (Figs. 8 and  9). The A 3 and A 4 complexes do not involve PTF1 (Fig. 7) and may be the same as two sequence-specific complexes formed by the binding with Region II and XC nuclear extract, judging from the mobility, affinity, and core binding sequence of the complexes. Therefore, these two complexes are likely to contain constitutive factors that are needed for basal level activation, for example Sp1 protein, because the core sequence in Region II carries the GACCC motif (Ϫ83/Ϫ79), which is related to the CACCC motif for Sp1 binding (34,35). However, the question of whether the Sp1 protein is involved awaits further analyses.
With the extract of PSTI nonexpressing XC cells, we noted a protein binding sequence lying between Ϫ154 bp and Ϫ137 bp (Region IV). The transfection assays in XC cells showed no activity with deletion mutants containing intact Region IV, but deletion extended from Ϫ159 bp to Ϫ141 bp, impairing Region IV, resulted in the appearance of low level activity (around 10-fold), which was similar to the basal activity that PSTI(98)CAT in AR42j cells showed (Fig. 2). This finding raises the possibility that Region IV may play a role in a negatively regulating process of the expression of PSTI gene in nonpancreatic cells. The gel mobility shift assays using XC nuclear extracts showed that Region IV yields complexes X 1 , X 2 , and X 3 (Fig. 6), which are likely to involve the negative factors, although such factors have never been detected so far in other pancreatic genes. The core sequence of Region IV binding complexes is the sequence CAATCAATAAC (-149/-139), which is superimposable to that of A 1 complex in a positively active Region IA (Figs. 8 and 9). Positive and negative regulations for tissue-specific gene expression have been reported in several cases, including albumin (36), ␤-globin of chick (37), pancreatitis-associated protein I (PAP I) of rat (38), p12 of mouse (39), and MUC1 of humans (40); however, none of these carries a shared recognition sequence(s). This sequence CAATCAATAAC that we revealed in the promoter region of human PSTI gene is likely to function to activate PSTI gene expression in pancreatic cells by the interaction of the binding factors for Regions I and II and to repress the same expression in nonpancreatic fibroblast cells by the interaction of binding factors for Region IV, refining the pattern of the pancreas-specific expression. The sequence CAATCAATAAC is a novel element, whose function may alter positively and negatively depending on cell type. The analysis of the putative positively acting and negatively acting factor(s) participating in the human PSTI gene expression is of utmost importance for future study. FIG. 9. Summary of transcriptional regulatory elements in the promoter region of the human PSTI gene. The human PSTI promoter region can be divided into several elements that influence transcription in vitro. The upper panel indicates the regulatory elements, binding complexes, and function of each element in AR42j cells, and the lower panel indicates those in XC cells. Open boxes indicate the regulatory elements including the promoter, and shaded boxes indicate the core binding regions, which were revealed in this paper. Open circles indicate the binding factors that were detected, and the sequence in the center of this panel that is surrounded by an open box indicates the common core sequence of Regions I and IV. The arrow shows the major transcription start site: Ϫ60.