TGTCACA Motif Is a Novel cis-Regulatory Enhancer Element Involved in Fruit-specific Expression of thecucumisin Gene*

Cucumisin, a subtilisin-like serine protease, is expressed at high levels in the fruit of melon (Cucumis melo L.) and accumulates in the juice. We investigated roles of the promoter regions and DNA-protein interactions in fruit-specific expression of the cucumisin gene. In transient expression analysis, a chimeric gene construct containing a 1.2-kbcucumisin promoter fused to a β-glucuronidase (GUS) reporter gene was expressed in fruit tissues at high levels, but the promoter activities in leaves and stems were very low. Deletion analysis indicated that a positive regulatory region is located between nucleotides −234 and −214 relative to the transcriptional initiation site. Gain-of-function experiments revealed that this 20-bp sequence conferred fruit specificity and contained a regulatory enhancer. Gel mobility shift experiments demonstrated the presence of fruit nuclear factors that interact with thecucumisin promoter. A typical G-box (GACACGTGTC) present in the 20-bp sequence did not bind fruit protein, but two possiblecis-elements, an I-box-like sequence (AGATATGATAAAA) and an odd base palindromic TGTCACA motif, were identified in the promoter region between positions −254 and −215. The I-box-like sequence bound more tightly to fruit nuclear protein than the TGTCACA motif. The I-box-like sequence functions as a negative regulatory element, and the TGTCACA motif is a novel enhancer element necessary for fruit-specific expression of the cucumisin gene. Specific nucleotides responsible for the binding of fruit nuclear protein in these two elements were also determined.

Timing and levels of gene expression are critical to the proper development of eukaryotic organisms. Regulation of the expression pattern of a particular gene can involve the specific binding of trans-acting factors to the cognate cis-elements, constituting a crucial step in transcriptional initiation and, in turn, on the spatial and temporal expression of genes. Plant genes that show tissue specificity, developmental specificity, and a wide range of expression levels have been characterized, whereas their expression patterns are also influenced by environmental stimuli. A family of genes for fruit proteins provide a model system for the study of the regulatory mechanisms of plant genes, since their expression is restricted to a specific tissue and stage during fruit development (1). A number of fruit-specific genes that are activated during ripening have been isolated from tomato and other fruits, and genes responding to ethylene and nonethylene signals have been identified (1,2). The promoters of fruit-specific genes would also be of great interest for use in strategies to manipulate fruit metabolism and produce valuable proteins such as antibody, biopharmaceuticals, and edible vaccines through methods of genetic engineering (3)(4)(5). However, the detailed mechanisms by which the expression of fruit protein genes are regulated are poorly understood, as many of the essential cis-elements have not been identified.
Melon cucumisin, an extracellular subtilisin-like serine protease, is expressed at high levels in fruit, where it comprises more than 10% of the total protein (6). No expression is detected in other parts of the melon plant. The accumulation of a large amount of the cucumisin protein in juice in the central parts of the fruit at a specific stage during fruit development and the single copy character of the corresponding gene suggest that the cucumisin gene contains a strong promoter, which also controls the fruit-specific expression of cucumisin (7). The cucumisin gene is also characterized by its conspicuous expression pattern in specific tissues such as the placenta, locule, and peripheral tissues around seeds in the fruit and early during fruit development. These specific temporal and spatial expression patterns of the cucumisin promoter may be explained as the result of regulatory assemblies of several transcriptional activators that recognize the cis-elements implicated in fruitspecific expression. The cucumisin promoter may also be useful for genetic engineering of fruit. To understand the mechanisms of fruit-specific gene expression and for the design and application of fruit-specific promoters for the improvement of fruit quality by genetic engineering, detailed analyses of the cucumisin promoter and the transcription factors are required.
Here, we describe cloning of the cucumisin promoter and experiments designed to define cis-acting elements involved in proper expression of the gene. A stretch of only 20 bp of the promoter sequence is required for the promoter activity in young fruit.

EXPERIMENTAL PROCEDURES
Plant Material-Musk melons (Cucumis melo L. var. reticulatus cv. Teresa) were cultivated in a greenhouse at the experimental farm attached to the Faculty of Agriculture, Kobe University, from March to August. Fruits were tagged upon self-pollination, and developing fruits were harvested on the 10th day after pollination (DAP). 1 For the analysis of transient gene expression, developing fruits, leaves, and stems were used immediately after harvest. The central parts of the fruit were separated from the sarcocarp and testae and used for isolation of nuclear proteins, or frozen in liquid nitrogen and stored at Ϫ80°C until use for poly(A) ϩ RNA isolation.
* 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AY055805.
‡ To whom correspondence should be addressed: Laboratory of Biochemistry, Faculty of Agriculture, Kobe University, Rokkodai-cho 1, Nada-ku, Kobe, 657-8501, Japan. Tel./Fax: 81-78-803-5875; E-mail: yamagata@kobe-u.ac.jp. Cloning of the Cucumisin Gene-Melon genomic DNA was prepared from 10 g of young leaves using cetyltrimethylammonium bromide according to the method of Rogers and Bendich (8) as described previously (7). To determine the DNA sequence of the 5Ј portion of the exon/intron region of the cucumisin gene, the genomic DNA (1 g) was used as a template for PCR using a forward primer (HD3, CT-TCAAGCTTTTCTTTTTTAGT) and a reverse primer (SC1, ATAT-GATCGAGCTCCAATAA) that were synthesized on an Applied Biosystems 380A oligonucleotide synthesizer (Foster City, CA) based on the sequence of cucumisin cDNA (7). The PCR conditions were 94°C for 2 min, followed by 30 cycles of 1 min at 94°C, 1.5 min at 55°C, and 2 min at 72°C. The amplified fragment (1.8 kb) was digested with SacI and HindIII, inserted into pBluescript II SK Ϫ (subclone name pH-S1), and sequenced on both strands using a Sequenase version 2.0 sequencing kit (U.S. Biochemical Corp.) as described in the manufacturer's instruction manual.

Reagents-Restriction
The promoter region of the cucumisin gene was amplified by inverse PCR as described by Ochman et al. (9). The genomic DNA (1 g) was digested with AccI, and the DNA fragments were self-ligated with 10 units of T4 DNA ligase (New England Biolabs). This circularized DNA was used as a template for inverse PCR using the forward primer HD3 and a reverse primer (M1, AGAAGACATAGTAGTGCTTTTTGCTAT-CAG). The amplified fragment (0.6 kb) was subcloned into the TA cloning vector (subclone pTA Acc3) prepared from SmaI-digested pBluescript II SK Ϫ by incubation with Amplitaq DNA polymerase (Applied Biosystems Ltd.) and 10 mM dTTP at 70°C for 2 h. Alternatively, the genomic DNA was digested with BamHI and BglII to make the same cohesive ends and circularized, and then inverse PCR was carried out using a forward primer (B2, TGTATACATGGGGAGGAAGCT) and the reverse primer M1. The amplified fragment (1.8 kb) was subcloned into the TA cloning vector (subclone pM1-B2-5) and sequenced. An additional PCR was carried out using the genomic DNA as a template and a forward primer (T3A, CTTAGTGTACTATATCCTTT) and a reverse primer (Z, AAGCGAGTCGATTACTGAAG) to obtain the DNA fragment overlapping pH-S1 and pM1-B2-5, and the amplified fragment (0.4 kb) was digested, subcloned into pBluescript II SK Ϫ (subclone pUK2), and sequenced. Sequence analyses were carried out using the Genetyx program (Software Development Co., Ltd., Tokyo).
Primer Extension Analysis-Primer extension analysis was carried out as described by Sambrook et al. (10). The synthetic oligonucleotide Ex (5Ј-TAGAAGCGAGTCGATTACTGAAGAAAAGACT-3Ј) was 5Ј-endlabeled with [␥-32 P]ATP and T4 polynucleotide kinase (New England Biolabs) and then used as a primer for a reverse transcriptase reaction with poly(A) ϩ RNA prepared from immature melon fruit (ϳ10 DAP) as described previously (7) using avian myeloblastosis virus reverse transcriptase (Invitrogen). The extended product was compared with the sequence ladder obtained from the pUK2 primed with the same synthetic oligonucleotide (Ex).
Particle Bombardment-Immature melon fruits (ϳ10 DAP) cut horizontally into slices (2-mm thickness), young leaves, and stems were put into Petri dishes. Each sample was bombarded twice with gold particles (1.6 m in diameter) coated with plasmids. In all experiments, we cobombarded the samples with two kinds of plasmid, cucumisin promoter fused to ␤-glucuronidase (uidA; Ref. 11) (denoted as GUS) and the firefly luciferase (LUC) reporter gene (pBI⍀FF) constructed as described below as an internal standard. An equimolar amount of each cucumisin::GUS construct (2-2.5 g) and 2.5 g of the pBI⍀FF plasmid were mixed with 3 mg of gold particles and suspended in 100 l of ethanol. Each plant material was bombarded with a 12-l aliquot of the suspension (360 g of gold particles) per shot using a helium-driven Biolistic PDS1000 system (Bio-Rad) with a 28-mm Hg vacuum. The distance between the rupture disc (1,100 psi) and macrocarrier and that between the macrocarrier and sample were 3.0 and 6.0 cm, respectively. After bombardment, the tissues were covered with aluminum foil and incubated in the dark for 24 h at 25°C.
␤-Glucuronidase (GUS) and Luciferase (LUC) Assays and Measurement of Protein Concentration-The central parts of fruit sections, leaves, and stems that were bombarded as described above were ground in liquid N 2 using a chilled mortar and pestle. The powder was dispensed into microcentrifuge tubes and mixed with 300 l of Pica Gene lysis buffer (LC␤/PGC-51; Wako, Osaka, Japan) containing 2 mM diiso-propylfluorophosphate and then centrifuged at 13,000 ϫ g for 5 min at 4°C. The supernatant was frozen at Ϫ80°C until the enzyme assay was conducted. Proteins were measured using a Bio-Rad protein assay kit with bovine serum albumin as a standard. GUS activity was measured by the method of Jefferson et al. (11), using 4-methylumbelliferyl ␤-Dglucuronide (Nacalai Tesque, Kyoto, Japan) as the substrate. The fluorometer was calibrated with 4-methylumbelliferone standards in the same solutions used for sample assays and blanked against the zero time point. LUC assays were performed as described by Miller et al. (12) using a Pica Gene LUC assay kit (Wako). Photon emission derived from LUC activity was counted with a luminometer (model TD-20/20; Turner Designs, Sunnyvale, CA). Background activities from plants bombarded with gold particles only were subtracted from each GUS and LUC value. All GUS values were normalized to the corresponding LUC values. Assays were performed on at least three different tissue samples for each construct, and the average values were expressed as relative activities to that of the positive control (pBI221, CLONTECH).
Construction of Cucumisin: GUS and pBI⍀FF Plasmids-The cucumisin upstream regions were amplified by PCR using two primers containing XbaI and BamHI sites at their 5Ј-ends, respectively. The BamHI site corresponds to the transcriptional start site of the cucumisin gene, and the XbaI site corresponds to the upstream region. The amplified fragments were subcloned into the XbaI-BamHI sites of the pSKGUS3C plasmid harboring GUS and a pea RuBisCO 3C (small subunit of ribulose-1, 5-bisphosphate carboxylase/oxygenase 3C) terminator (13). pBI⍀FF contains the tobacco mosaic virus ⍀ sequence (14) followed by the firefly luciferase coding region from the Cab2::LUC construct (12) in place of GUS between BamHI and SstI sites in pBI221. Both pBI⍀FF and pBI221 have the same cauliflower mosaic virus 35S constitutive promoter and nopaline synthase terminator. For the gainof-function experiments, the 4 ϫ 20-bp tandem repeat sequence was prepared as follows. Two synthesized oligonucleotides, pHISpEX (5Ј-G-GGGGAATTC(GACACGTGTCACAACCTAAT) 4 TCTAGACCCC-3Ј) and pHISmXE (5Ј-GGGGTCTAGA(ATTAGGTTGTGACACGTGTC) 4 GAAT-TCCCCC), were annealed and digested with EcoRI and XbaI to yield a 90-bp fragment denoted as pHIS. Then pHIS was ligated with EcoRI-NotI and XbaI-SacI adapters and digested with the corresponding restriction enzymes. The NotI-XbaI fragment of this sequence in the normal orientation and NotI-SacI fragment in the reverse orientation were placed into NotI/XbaI and SacI/NotI digested p89 plasmids, respectively, to yield pN-X (normal orientation) and pS-N (reverse orientation), and these plasmids were used for particle bombardment. All inserted sequences in each construct were confirmed by DNA sequence analysis.
Isolation of Nuclei and Nuclear Proteins-The nuclear isolation procedure was essentially the same as that described by Foster et al. (15) with some modifications. All procedures were performed at 4°C. The central parts of the immature melon fruit (about 10 DAP) were cut out from the pericarp, and the seeds were removed with a spatula. The tissue including the placenta, columella, and fibrous tissue (500 g) was ground using a commercial blender in 500 ml of nuclear grinding buffer (1 M hexylene glycol, 10 mM PIPES/KOH, pH 7.0, 10 mM MgCl 2 , 0.2% (v/v) Triton X-100, 5 mM 2-mercaptoethanol, 0.8 mM phenylmethanesulfonyl fluoride, and 2 mM diisopropylfluorophosphate). The mixture was filtered through two layers of gauze and 50-m nylon mesh. The filtrate was centrifuged at 2,000 ϫ g for 10 min, and the pellet (crude nuclei) was gently resuspended with a soft brush in 80 ml of nuclear wash buffer (0.5 M hexylene glycol, 10 mM PIPES/KOH, pH 7.0, 10 mM MgCl 2 , 0.2% (v/v) Triton X-100, 5 mM 2-mercaptoethanol, and 1.0 mM diisopropylfluorophosphate). After centrifugation of the mixture at 3,000 ϫ g for 5 min, the pellet was resuspended in 20 ml of nuclei lysis buffer (110 mM KCl, 15 mM HEPES/KOH, pH 7.5, 5 mM MgCl 2 , 1 mM dithiothreitol, 5 mg/ml antipain, 5 g/ml leupeptin, and 5 g/ml chymostatin). Then 2 ml of 4 M ammonium sulfate was added to the mixture in several small aliquots with gentle mixing, and the mixture was ultracentrifuged at 100,000 ϫ g for 90 min using a Beckman W/Ti60 rotor. Then solid ammonium sulfate was added to the supernatant to 0.25 mg/ml, and the protein precipitate was collected by centrifugation at 10,000 ϫ g for 15 min. The pellet was resuspended with 0.5 ml of nuclear extract buffer (70 mM KCl, 25 mM HEPES/KOH, pH 7.5, 0.1 mM EDTA, 20% (v/v) glycerol, 1 mM dithiothreitol, 5 g/ml antipain, 5 g/ml leupeptin, and 5 g/ml chymostatin), and the suspension was dialyzed against dialysis buffer (70 mM KCl, 25 mM HEPES/KOH, pH 7.5, 1 mM 2-mercaptoethanol, 0.1 mM EDTA, 20% (v/v) glycerol) for 2 h. After centrifugation at 12,000 ϫ g for 10 min, the extract was frozen in liquid N 2 and stored at Ϫ80°C until use.

Characterization of the Genomic Clone of the Cucumisin Gene and Determination of the Transcriptional Start Site-The
5Ј part of the cucumisin gene containing the 1.2-kb promoter was obtained by inverse PCR using melon genomic DNA as a template. The combined nucleotide sequence, which was confirmed by several independent rounds of ordinary PCR to correct for possible PCR errors, was 3,049 bp in length and contained five exons encoding Met 1 -Ala 181 of the cucumisin precursor and four introns. Fig. 1 shows the nucleotide sequence of the 1.2-kb promoter and short stretch of the transcribed region. The entire sequence is available from the Gen-Bank TM data base under accession number AY055805. The nucleotide sequence of the coding region was identical to that of the cDNA (7). Introns 1-3 were mapped in the prosequence of the cucumisin precursor, and intron 4 was located in that of the mature protease domain. Primer extension analysis revealed that transcription begins at an adenine residue located 48 bp upstream of the translational start site (Fig. 2). The putative TATA box, TATAAA, and CAAT box, CAAAT, were located 34 and 89 bp upstream, respectively, from this transcriptional start site. In the upstream region, a perfect palindromic G-box (5Ј-GACACGTGTC-3Ј), the target-binding site of plant basic leucine zipper proteins (16), was found at Ϫ234 to Ϫ225 bp from the transcriptional start site. An I-box-like sequence containing a tandem GATA repeat (5Ј-AGATATGATAAAA-3Ј), in that 9 bases were identical to the reported 13-bp I-box element (5Ј-GGATGAGATAAGA-3Ј) (17), was also found at positions Ϫ253 to Ϫ241 (Fig. 1).

5Ј-Deletion Analyses and Transient Assays Showed That the Cucumisin Upstream Region Had cis-Regulatory Regions for
High Level Expression in Fruit-We used horizontal sections (2-3 mm thick) of the immature melon fruit to examine reporter gene expression in transient assays. We fused the DNA sequence extending 1,181 bp upstream from the transcriptional start site (P1181 construct; Fig. 1) to the GUS reporter gene and introduced the construct into the central part of the fruit sections by particle bombardment with the 35S-LUC construct (pBI⍀FF) as the internal standard (Fig. 3). When the relative GUS activity of the control plasmid (pBI221) containing a constitutive cauliflower mosaic virus 35S promoter was taken as 100, the activity of the p1181 was 35, indicating that the 1181-bp 5Ј-upstream region was sufficient for the high level expression of the GUS gene in melon fruit (Fig. 3A).
To broadly determine the positive enhancer region, 5Ј deletion analysis of the 1,181-bp cucumisin promoter was conducted, and the activities of the deletion constructs were measured in young fruits, leaves, and stems. The p234 construct showed comparable levels of activity to the full-length p1181 construct in fruit, indicating that a region sufficient to confer higher expression in fruit is present within the Ϫ234 bp region from the transcriptional start site (Fig. 3A). However, the activity of p228 constructs lacking 6 nucleotides (GACACG) in the G-box element (5Ј-GACACGTGTC-3Ј) was decreased to about half of that of p1181. In the p214 construct, the activity decreased to about one-sixth of that of p1181. No further drop in activity was observed by deleting the promoter to position Ϫ89, and these activities were thought to be the basal level of the cucumisin minimal promoter (p89) containing the TATA box and CAAT box. These results indicated that critical cisacting element(s) responsible for high level cucumisin promoter activity in fruit were localized in the 20-bp fragment from Ϫ234 to Ϫ215, the sequence of which is 5Ј-GACACGTGTCACAAC-CTAAT-3Ј. This sequence contains a perfect palindrome of a typical G-box element (5Ј-GACACGTGTC-3Ј) in the first half. The decrease in activity of the p254 construct suggested the presence of another positive element(s) in the region from Ϫ310 to Ϫ255 and negative element(s) in the region from Ϫ254 to Ϫ235, which contains an I-box-like sequence (5Ј-AGATAT-GATAAAA-3Ј). Since p234 had strong activity comparable with the longest p1181, we did not examine further the possible positive element(s) upstream of Ϫ234.
The Cucumisin 5Ј Upstream Region Can Direct Fruit-specific Expression-All constructs from P1181 to P214 introduced in melon leaves and stems showed very low levels of promoter activity in contrast to the control pBI221 that showed a high level of activity in these organs (Fig. 3, B and C). The activities of cucumisin promoter constructs in stems were slightly higher than those in leaves. These results indicated that the cisregulatory element(s) in the upstream region of the cucumisin promoter is responsible for not only high level expression in fruits but also for fruit-specific expression.
The 20-bp Sequence Is Sufficient to Confer Fruit-specific Expression on a Cucumisin Minimal Promoter-We conducted gain-of-function analysis to determine whether the 20-bp region from Ϫ234 to Ϫ215 is able to confer fruit-specific high level expression on the cucumisin minimal promoter. Four tandem repeats of the 20-bp sequence were fused in the normal or reverse orientation to the truncated (Ϫ89 to Ϫ1 from the transcriptional start site) cucumisin promoter containing the putative TATA and CAAT boxes to yield pN-X and pS-N, respectively (Fig. 4). In the absence of the 20-bp sequence, the construct (p89) showed very low levels of GUS activity in fruit tissues as in leaves and stems. However, fruit bombarded with construct containing four copies of the 20-bp upstream sequence (pN-X) showed an 8-fold induction of GUS activity, and the activity was comparable with that of fruit bombarded with the full-length cucumisin promoter (p1181) (Fig. 4A). It is noteworthy that the tandem repeat of the 20-bp sequence promoted GUS expression in both orientations. However, both pN-X and pS-N did not lead to elevation of basal level of GUS activity derived from p89 in leaf or stem tissues (Fig. 4, B and C). These results indicated that the 20-bp sequence has two functions: to enhance transcription in fruit tissues and to act as a fruitspecific enhancer.
Detection of Nuclear Factors That Interact with the Cucumisin Promoter 5Ј Upstream Region-Transient expression assays suggested that the 20-bp sequence from Ϫ234 to Ϫ215 containing a G-box element (5Ј-GACACGTGTC-3Ј) contains cis-element(s) responsible for high level and fruit-specific expression and suggested that the region from Ϫ254 to Ϫ235 containing the I-box-like sequence (5Ј-AGATATGATAAAA-3Ј) that has a tandem repeat of the GATA sequence has a negative influence on gene expression. Another I-box-related sequence, a GATA element, was also found between Ϫ212 and Ϫ207 (5Ј-AGATAT-3Ј) and between Ϫ195 and Ϫ189 on the opposite DNA strand (5Ј-GATAGAA-3Ј) (Fig. 5). To evaluate whether these sequences interact specifically with proteins in nuclei from melon fruit, we conducted gel electrophoresis mobility shift assays. Nuclear protein extracts were prepared from central parts of immature fruit. First, we divided the Ϫ254 to Ϫ183 region that was thought to contain both positive and negative regulatory elements as described above into two regions, G (Ϫ254 to Ϫ215) and H (Ϫ222 to Ϫ183) as shown in Fig.  5. When these sequences were used as probes in the gel mobility shift assays, binding of nuclear proteins from central parts of young fruit to these sequences was detected as a broad retarded band (Fig. 6, lanes 2 and 4). The signal of the complex of the fruit protein with the G probe was far stronger than that with the H probe, indicating that the nuclear protein(s) in the fruit binds more tightly to the G region than to the H region.
Determination of Protein-binding Region within the Cucumisin Promoter-To localize the cis-element responsible for binding with fruit nuclear proteins more precisely within the G region, we further divided the G and H regions into Ga (Ϫ254 to Ϫ235) and Gb (Ϫ234 to Ϫ215), and Ha (Ϫ222 to Ϫ203) and Hb (Ϫ202 to Ϫ183), respectively (Fig. 5). In addition to these fragments, GH (Ϫ240 to Ϫ201) spanning the G and H regions was synthesized. These oligonucleotides were used as competitors in the gel mobility shift assays with the G, H, and GH fragments as labeled probes (Fig. 7, A-C). Surprisingly, not only the G competitor but also the H competitor completely abolished the retarded band of the G probe, when these competitors were present in a 100-fold molar excess relative to the labeled probe (Fig. 7A, lanes 2-4). Conversely, the binding of the H probe to the protein competed with not only the H fragment but also with the G fragment (Fig. 7B, lanes 2-4). The addition of the GH and Ga competitors to the binding mixture with the G probe significantly diminished the retarded band, whereas the Gb competitor showed weak competition (Fig. 7A,  lanes 5-7). The Hb, but not the Ha, could compete strongly with the G probe (Fig. 7A, lanes 8 and 9). These results confirmed that there are at least two binding regions to the nuclear proteins within the G region (i.e. Ga and Gb), and Ga binds more strongly to the nuclear proteins than Gb. The I-box-like sequence (5Ј-AGATATGATAAAA-3Ј) in the Ga region is a possible cis-element for the observed tight binding of the G region to the nuclear proteins, since a similar I-box-related GATA sequence (5Ј-GATAGAA-3Ј) was present on the opposite DNA strand within the Hb region that also competed well with the G probe. The weak competition of the Ha with the G probe ( Fig.   FIG. 4. Gain-of-  7A, lane 8) can be explained by the presence of another GATA sequence (AGATAT) in the Ha fragment, the same sequence with the first 6 bases in the I-box-like sequence present in the Ga region (Fig. 5). Thus, the I-box-like sequence (5Ј-AGATAT-GATAAAA-3Ј) in the Ga region and 20 bp of Gb sequence are likely to be the target binding sites of the fruit nuclear proteins. When the H fragment was used as a probe, besides H and G fragments, GH, Hb, and Gb fragments were strong competitors, and Ha and Ga fragments were weak competitors (Fig.  7B). These results agreed well with the above suggestion for the estimation of the cis-element in the G region, and I-boxrelated GATA sequence in the opposite strand (5Ј-GATAGAA-3Ј) in the Hb region and another GATA sequence (5Ј-AGATAT-3Ј) in the Ha region seem to be strong and weak binding regions, respectively. When the labeled GH fragment was used as a probe, the signal of the retarded band was much weaker than those seen with G and H probes, and longer exposure of the film was needed, indicating that the nuclear protein binds only weakly to the GH region (Fig. 7C). The G and H fragments as well as the GH fragment itself strongly competed with the GH probe. Also, the competition with the Gb and Ha fragments was stronger than with the Ga and Hb fragments, indicating that the GH region had also at least two elements (Gb and Ha regions) involved in protein binding. Since the DNA sequence downstream of Ϫ214 is thought not to be involved in the high level and fruit-specific expression of the cucumisin promoter ( Fig. 3), the H region was not studied further in the following experiments.
To test the possibility that the nuclear protein binds to the central parts in the G region spanning the Ga and Gb, we synthesized a Gc oligonucleotide (Ϫ244 to Ϫ225; Fig. 5) and used it as a labeled probe for the gel mobility shift assay. As shown in Fig. 8A, the Gc probe did not show any retarded band in contrast to the Ga and Gb probes that showed strong and weak retarded bands, respectively. Therefore, the G-box (5Ј-GACACGTGTC-3Ј) in the Gc region could be excluded from the candidates of cis-elements responsible for binding to nuclear proteins.
I-box-like Sequence and TGTCACA Motif Are cis-Elements Responsible for the Binding of Fruit Nuclear Proteins-When the 40-bp G region was used as a labeled probe and four 10-bp segments derived from the G region (G1 (5Ј-AAGATATGAT-3Ј), G2 (5Ј-AAAAATAGAC-3Ј), G3 (5Ј-GACACGTGTC-3Ј), and G4 (5Ј-ACAACCTAAT-3Ј)) ( Fig. 5) were used as competitors, only G4 alone weakly competed for the binding of nuclear protein to the G probe, whereas G1, G2, and G3 fragments showed little competition (Fig. 8B). The observation that the G3 fragment, a 10-bp palindrome of the G-box, did not compete with the G probe indicated again that the G-box did not significantly contribute to the binding of fruit nuclear protein to the cucumisin promoter but that the G4 fragment was important for the observed protein binding to the Gb probe (Fig. 8A, lane 4). An odd base palindrome, TGTCACA motif, is present in the center of the Gb region and overlaps with the G-box. We suspected this TGTCACA motif to be the cis-element responsible for protein binding. A 13-bp I-box-like sequence (5Ј-AGATAT-GATAAAA-3Ј) inhibited protein binding to the G probe more strongly (Fig. 8B, lane 5), consistent with the above observations for the tight binding of the I-box-like sequence to the nuclear protein. However, since G1 (5Ј-AAGATATGAT-3Ј) and G2 fragments (5Ј-AAAAATAGAC-3Ј) containing the first 9 bases and the last 4 bases of the I-box-like sequence, respectively, hardly inhibited protein binding (Fig. 8B, lanes 4 and 6), the whole I-box-like sequence is probably necessary for protein binding.
These observations from gel mobility shift assays showed that two possible cis-elements, an I-box-like sequence and TGTCACA motif, are present in the G region. The I-box-like sequence can bind the nuclear protein tightly, whereas the TGTCACA motif shows weak binding. Together with the results of transient expression assays (Figs. 3 and 4), we concluded that the I-box-like sequence in the Ga region is a negative regulatory element, and the TGTCACA motif in the Gb region is responsible for the high level and fruit-specific expression of the cucumisin promoter.
Critical Nucleotides Responsible for the Binding of Nuclear Proteins in the I-box-like Sequence and TGTCACA Motif-To further confirm the binding of nuclear protein to the I-box-like sequence and TGTCACA motif, detailed analysis of sequencespecific binding was carried out using mutated derivatives of these sequences in gel mobility shift assays (Fig. 9). Successive single base pair mutations were introduced into these fragments as shown in Fig. 9C (see also "Experimental Procedures"), and the mutated fragments were used as competitors in the competition experiments with the wild-type fragment. As shown in Fig. 9A, the specificity of binding of nuclear protein to the G region was demonstrated by the ability to compete this complex effectively with a 500-fold molar excess of the I-box-like sequence (Fig. 9A, lane 3). However, competitors containing a point mutation located at positions 1, 3, 4, 8, 9, 10, or 11 in the I-box-like sequence failed to fully compete binding of nuclear protein to the wild-type fragment. Especially, the oligonucleotide with a point mutation at position 9 (T) in the I-box-like sequence showed no competition. In contrast, oligonucleotides containing mutations at the middle position (oligonucleotides 5-7) and downstream (oligonucleotides 12-14) of this sequence efficiently competed with the binding of nuclear protein. This indicated that one of the nuclear protein binding sites in the G region is located in a region covering the I-boxlike sequence, and the nucleotides important for the binding are A 1 , A 3 , T 4 , A 8 , T 9 , A 10 , and A 11 , of which T 9 is most important (Fig. 9D).
An analogous experiment was carried out for the TGTCACA motif. A 500-fold molar excess of G5 fragment (5Ј-TGTCA-CAACCTAAT-3Ј) efficiently interfered with the protein binding to labeled Gb fragment (Fig. 9B, lanes 3 and 15). However, competitors containing point mutations located at position 3 or 7 in the G5 sequence failed to fully compete with binding of nuclear protein to the wild-type fragment. These observations indicated again that the TGTCACA motif is responsible for the binding of nuclear protein to the Gb region, and the critical nucleotides are T 3 and A 7 in the G5 fragment (Fig. 9D). Together with transient expression assays (Figs. 3 and 4), these results confirmed the role of the TGTCACA motif as a cisacting enhancer element in the cucumisin promoter.

DISCUSSION
Melon is a typical climacteric fruit, and its ripening is influenced by ethylene, which regulates expression of several ripening-related genes in a variety of climacteric fruits (1,2). The biosynthesis of ethylene in melon fruit occurs at a late stage in fruit development, and the level of 1-aminocyclopropane-1-carboxylate oxidase, a key ethylene biosynthetic enzyme, is very low in unripe melon fruit (18,19). Synthesis of cucumisin, however, was observed at a very early stage (about 5 DAP) of fruit development before the onset of ripening (6,7). Also, ethylene-responsive elements such as the GCC box conserved in the 5Ј upstream region of many ethylene-inducible genes (20,21) could not be found in the cucumisin promoter (Fig. 1). These findings indicated that cucumisin is unlikely to be an ethyleneinducible gene, and the expression of the gene is probably regulated in a developmental and organ (tissue)-specific manner. While ethylene is a dominant hormonal trigger for ripening of climacteric fruit, it has been suggested that both ethylenedependent and ethylene-independent regulatory pathways coexist to coordinate the ripening process in both climacteric and nonclimacteric fruit (1, 2). The observation that both ethylene and additional developmental factors regulate several fruit-specific genes indicates that cis-elements responsible for fruit specificity, in addition to those that mediate ripeningassociated developmental and ethylene-mediated regulation, could be separated.
Techniques for direct DNA transfer and transient transformation are established procedures to study the regulation of gene expression (22). Transient transformation of intact plant organs by particle bombardment has proven to be a powerful technique to study the transcriptional regulation of organspecific genes, since the transformed cells can be monitored in their native organ environment (23,24). Since ballistic transient transformation has the advantage of rapidity of analysis, compared with the stable transformation of plants with the large numbers of promoter-reporter gene constructs necessary for detailed cis-analysis, we applied this technique to analysis of the cucumisin promoter. Transient expression experiments in melon fruit slices with cucumisin promoter-GUS fusion genes have shown that as little as 300 bp of promoter region is sufficient to control the correct spatial expression of GUS activity compared with the activity of the endogenous promoter. This result is in agreement with many other plant promoter studies in which promoter regions in the range of several hundred base pairs to about 1 kb have been found to reproduce faithful expression patterns of reporter genes in vivo. Therefore, introns, downstream sequences, or extended domains in the chromatin at the chromosomal location of the cucumisin gene are unlikely to contribute significantly, if at all, to the regulation of its transcription.
A perfect palindromic G-box sequence, 5Ј-GACACGTGTC-3Ј, was found in the 5Ј upstream region of the cucumisin promoter. G-box elements are currently the best-characterized plant cisregulatory element. A family of plant basic leucine zipper proteins has been identified that interacts with G-box elements to confer high promoter activity (16). We therefore presumed initially that the G-box in the cucumisin promoter was a target for a transcription factor controlling the observed high level expression and fruit specificity of the cucumisin gene. Transient expression assays with truncated promoters fused to the GUS reporter gene, however, indicated that the G-box is not a single dominant cis-element responsible for fruit expression of the cucumisin gene but that the downstream sequence is necessary for full activity of the cucumisin promoter (Fig. 3). Gain-offunction experiments revealed that the 20-bp sequence (5Ј-GACACGTGTCACAACCTAAT-3Ј) containing the G-box in the first half includes an enhancer that controls fruit-specific ex- pression, since four tandem repeats of the 20-bp sequence in both orientations enhanced the transcription in fruit and were sufficient to confer fruit specificity of the gene expression to the minimal promoter (Fig. 4).
Plant G-box elements contain a CACGTG core sequence that is necessary for efficient binding of basic leucine zipper proteins, and sequences flanking the CACGTG core affect basic leucine zipper protein binding specificity (25) and gene expression in vivo (26). The G-box in the cucumisin promoter, a 10-bp perfect palindrome (5Ј-GACACGTGTC-3Ј), can be classified as a high affinity class I G-box element that binds type A G-box binding factor (GBF) (16). Gel mobility shift assays, however, showed that the G-box is not responsible for the binding of fruit nuclear protein to the cucumisin promoter (Fig. 8, A and B). Instead, an odd base palindromic sequence, TGTCACA, localized in the center of the 20-bp sequence and overlapping with the G-box (Fig. 5), was thought to be an essential element for the binding of fruit protein to the Gb region of the cucumisin promoter. Furthermore, competition experiments in the gel mobility shift assay clearly showed that T 3 and A 7 in the TGTCACA motif are critical nucleotides for the binding of nuclear protein in fruit (Fig. 9). Since the G-box element in the cucumisin promoter contains the complete structure in itself for binding of GBF (i.e. further flanking sequence is not necessary for the protein binding (16)), the TGTCACA motif is not likely to be a part of the G-box. Most in vivo expression studies have indicated that G-box elements cannot act on their own but require the presence of additional cis-acting sequences for their functions (16). However, the TGTCACA motif in the cucumisin promoter is also unlikely to be such a cooperative element for the G-box, because it overlaps with the G-box. Taken together, we concluded that the TGTCACA motif is the primary element responsible for the binding of nuclear protein to the cucumisin promoter in fruits. To our knowledge, the TGTCACA motif is a novel cis-element in the promoters of plant genes. For the full activity of the TGTCACA motif, however, part of the G-box is thought to be necessary, since promoter activity was decreased distinctly without 6 bp of the G-box (Fig. 3). One explanation for this observation is that the TGTCACA motif is a possible core sequence of a longer cis-element, and the sequence-flanking TGTCACA may affect its activity as a transcriptional activator. To understand the precise function of the TGTCACA motif, identification of a TGTCACA motif-binding protein is required, and we are now cloning TGTCACA motif-binding protein using the yeast one-hybrid system.
Three members of the tomato gene family for RuBisCO (RBCS1, RBCS2, and RBCS3A) contain a conserved pair of I-box (5Ј-GGATGAGATAAGA-3Ј) and G-box (5Ј-CACGTG-3Ј) elements in an identical spatial arrangement (27). The promoters of these genes are equally active in all green tissues, with the exception of young fruit, where the RBCS3A promoter has strongly reduced activity compared with the RBCS1 and RBCS2 promoters (17). Comparative analysis of the RBCS2 promoter in tomato leaves and young fruits showed that the 37-bp domain containing the conserved I-box and G-box elements is required for high level promoter activity in leaves, while only the G-box is required in young fruits (28). They speculated that these RBCS genes are activated by the coordinate binding of an I-box binding factor (IBF) and GBF in leaves, while the promoter is activated by GBF alone in fruits. cDNAs encoding GBFs were cloned from young tomato fruit, and these GBFs were shown to bind the G-box sequence in the RBCS1, RBCS2, and RBCS3A promoters in vitro (29). However, the differences in the expression patterns of RBCS1, RBCS2, and RBCS3A in tomato young fruit cannot be explained by the G-box alone, since all of these genes have a G-box element in their promoters. Although a negative regulator, the F-box binding factor, acting on the RBCS3A promoter in developing fruit was found (17), it is uncertain whether G-box sequences in RBCS1 and RBCS2 function as genuine FIG. 9. Binding of fruit nuclear proteins to the I-box-like sequence and TGTCACA motif. A and B, determination of the binding site of the nuclear protein in gel mobility shift experiments. G (A) and Gb fragments (B) were used as DNA probes. Competitors, the sequences of which are listed in C, were added in 500-fold molar excess to the corresponding probe. I, I-box-like sequence. C, sequences of mutated derivatives of I-box-like sequence and G5 oligonucleotides. Only the mutated nucleotides in the upper strands are shown below the sequences of wild-type I-box-like sequence and G5. Dashes indicate the same nucleotides as in the wild-type DNA. D, summary of the results of the competition experiments. The I-box-like sequence and G5 sequence are shown. The box indicates the TGTCACA motif. Asterisks indicate nucleotides critical for protein binding. The large asterisk on T 9 in the I-box-like sequence indicates the most critical nucleotide for the binding of nuclear protein.
single dominant positive cis-element in young tomato fruit. Despite the presence of the same kinds of cis-elements and their similar arrangements in promoters of tomato RBCS and melon cucumisin, the functions of these cis-elements seem to be quite different between RBCS and cucumisin genes.
The I-box is less well characterized than the G-box and seems to be involved in light-regulated and/or circadian clockregulated gene expression of photosynthetic genes (30,31). Related GATA motifs are found in many other promoters, some of which are light-regulated but others of which are not (30,32). A tobacco protein, ASF-2, binds not only to the as-2 element within the cauliflower mosaic virus 35S promoter but also to the conserved tandem GATA motifs within the cab (chlorophyll a/b-binding protein) genes (32). The tandem GATA motifs, which are also present in the I-box-like sequence of the cucumisin promoter, have been reported to mediate cell type specificity but not light responsiveness (32). In tomato, the I-box has been shown to be an activating cis-element of the RBCS in leaves (27). Recently, tomato I-box binding factor LeMYB1, a Myb-like protein, has been cloned and demonstrated to have transcriptional activation activity (33). The binding of LeMYB1 in the RBCS3A promoter was mapped to the whole sequence of the I-box (5Ј-GGATGAGATAAGA-3Ј). This would be consistent with our observation that the binding site of nuclear protein in the I-box-like sequence in the cucumisin promoter extended throughout the sequence, and most of the critical nucleotides for protein binding were mapped within the repeats of the core GATA motif (Fig. 9). Thus, we supposed that the cis-element responsible for protein binding in the Ga region is only this I-box-like sequence, and the Myb-like protein probably binds to it. Our finding that the I-box-like sequence is a down-regulating cis-element in gene expression in melon fruit suggests a new function of the I-box-like element as a negative regulator in fruit. The I-box-like sequence and its binding protein may down-regulate the expression of the cucumisin gene developmentally and/or cell type specifically in fruit. In cells in the central part of developing fruit where the cucumisin gene is highly expressed, the binding protein to the I-box-like sequence may be inactive through interaction with other cis-regulatory elements and/or transcription factors.
In conclusion, we have demonstrated that the 20-bp promoter fragment between positions Ϫ234 and Ϫ214 from the transcriptional start site of the melon cucumisin gene is sufficient to mediate the fruit-specific expression pattern observed for this gene. This pattern is most likely established via the action of fruit-specific transcriptional regulators interacting with cis-acting DNA sequences within the analyzed promoter fragment, in that both positive and negative regulators were found. Candidates for such fruit-specific positive and negative elements were identified as a TGTCACA motif and an I-boxlike sequence, respectively. Further studies of their functions in the regulation of gene expression in melon fruit and their potential interactions with transcription factors might provide new insights into the mechanisms of organ-specific transcription in plants.