INTRODUCTION

Mammalian kidneys are constantly exposed to steep osmotic gradients because of the urine-concentrating mechanism. However, kidney cells, especially the epithelial cells lining the inner renal medulla, are protected from the osmotic effect of concentrated sodium ion and urea in the interstitium by accumulating organic osmolytes such as sorbitol, betaine, inositol, glycerophosphocholine, and taurine (1). These organic osmolytes are involved in maintaining cell volume and electrolyte contents without perturbing the protein structure and function over a wide range of concentrations (2).

Sorbitol is one of the principal organic osmolytes in the inner medulla (3, 4). It is produced via the reduction of D-glucose by the enzyme aldose reductase (AR¹; EC 1.1.1.21). AR is present in a variety of tissues including kidney (5), liver (6), ocular lens, and retina (7) and in erythrocytes (8). Although AR has been implicated in the pathogenesis of various diabetic complications such as diabetic cataracts, retinopathy, and neuropathy (9-11), the physiological role of this enzyme is still unclear, except that in the renal medulla, it produces sorbitol to serve as an osmoprotectant (12), and in the testis, in conjunction with sorbitol dehydrogenase, it synthesizes fructose, the major energy source for sperm (13).

The expression of the AR gene in the renal medulla of mammals has been shown to be osmotically regulated. In this tissue, AR mRNA and sorbitol content increase during dehydration or antidiuresis and decrease during diuresis (14). Moreover, experiments on PAP-H25 cells derived from rabbit renal papillae demonstrated that external hyperosmolality enhances transcription of the AR gene (15), resulting in a rapid increase in AR mRNA (16), followed by an increase in AR activity and sorbitol content (17).

In mammals, several genes other than aldose reductase are known to be regulated by osmolality. They include the myo-inositol, betaine, and taurine transporters (18). However, how the initial signal of external hyperosmolality is relayed to the nucleus to induce the expression of these genes is still unknown. Recently, the sequence of the putative osmotic response element of the canine betaine transporter gene, TonE, was described (19), providing a starting point to understand this interesting mode of gene regulation. In addition, while we were preparing this report, the 11-bp osmotic response element (ORE) essential for osmoregulation of the rabbit AR gene was reported and found to be located 1105 bp upstream of the transcription start site (20). Recently, we identified a putative functional ORE found to be located 3.7 kb upstream of the transcription start site of the aldose reductase gene (21). We hereby show that this putative AR ORE resulted from a cloning artifact occurring at 1066 bp (MboI restriction site) upstream of the initiation transcription site in a human aldose reductase gene isolated from a commercially obtained genomic library constructed by partial digestion with MboI restriction enzyme.

In this study, we have identified an osmotic response region of the human AR gene within a 132-bp region located 1235 bp upstream of the transcription start site of human AR gene. Within this region, there are three sequences homologous to TonE of the canine betaine transporter gene and to the ORE of the rabbit AR gene. The osmotic response function is reduced when any one of these three sequences is mutated, suggesting that these sequences have to work in concert to provide maximum transcription induction.

MATERIALS AND METHODS

Chang liver cells (American Type Culture Collection) were maintained in isosmotic growth medium (Eagle's basal medium supplemented with 10% calf serum and 2 mM L-glutamine; 300 mosmol/kg of H₂O) at 37 °C in the presence of 5% CO₂. To investigate the osmotic induction of AR mRNA, these cells (3 × 10⁶) were seeded onto 100-mm culture dishes and incubated in fresh isosmotic growth medium for 24 h. The medium were removed, and the cells were incubated in either isosmotic or hyperosmotic growth medium (isosmotic medium supplemented with 5 M NaCl to final concentrations of 50 mM (380 mosmol/kg of H₂O) and 100 mM (500 mosmol/kg of H₂O)).

A total of 1 × 10⁷ cells were used to extract RNA in each experiment. Total RNA was extracted by a modification of the protocol of Auffray and Rougeon (22). Chang liver cells on culture dishes were washed twice with ice-cold phosphate-buffered saline; scraped off the dishes; and then transferred into tubes containing 1.5 ml of 3 M LiCl and 6 M urea solution. These cells were kept on ice, mixed, and homogenized by a Polytron homogenizer (Kinematic, Lucerne, Switzerland) for 1-2 min with a 30-s interval. The homogenate was kept on ice overnight. Total RNA was pelleted by centrifugation at 15,000 × g for 30 min at 4 °C. Pellets were washed twice with the 3 M LiCl and 6 M urea solution, drained, and resuspended in 0.4 ml of Tris/EDTA buffer containing 0.5% SDS and 200 µg of proteinase K. The resuspensions were incubated at 37 °C for 30-60 min. RNA was extracted twice with phenol/chloroform (1:1, v/v) and once with chloroform/isoamyl alcohol (24:1, v/v) and precipitated with 0.1 volume of 3 M sodium acetate and 2.5 volumes of absolute ethanol at 20 °C for 2 h. The mixtures were centrifuged for 20 min at 15,000 × g at 4 °C. The RNA was washed in 70% ethanol and resuspended in Tris/EDTA buffer. The concentration of RNA was determined by measuring the absorbance at 260 nm.

20 µg of each total RNA sample from Chang liver cells was loaded on 1.2% agarose gel containing 3% formaldehyde. After electrophoresis, the RNA was transferred onto Hybond-N⁺ membrane (Amersham International, Buckinghamshire, United Kingdom) by capillary blotting in 20 × SSC. The membrane was first incubated with prehybridization buffer (7% SDS and 0.25 M sodium phosphate) at 65 °C for 1 h and then with hybridization buffer containing [-³²P]dCTP-labeled human AR cDNA probe and incubated at 65 °C for 12 h. The AR cDNA probe (23) was isolated from a Bluescript subclone and labeled by random priming of DNA polymerase (Megaprime kit, Amersham International) to a specific activity of 1-2 × 10⁹ cpm/µg. After hybridization, the filter was washed with 0.1 × SSC and 0.5% SDS at 65 °C for 1 h and exposed to X-Omat AR film (Eastman Kodak Co.) at 70 °C with intensifying screens. The filter was then stripped in 0.1 × SSC and 0.5% SDS at 100 °C and rehybridized to an [-³²P]dCTP-labeled human -actin cDNA (CLONTECH, Palo Alto, CA).

To obtain the promoter sequence of AR, a human genomic DNA library in FIXII vector (Stratagene, La Jolla, CA) was screened using two [-³²P]ATP 5-end-labeled oligonucleotide primers derived from the genomic clone we described before (24): ARpr1, 5-GAATCTTAACATGCTCTGAACC-3; and ARpr2, 5-GCCCAGCCCTATACCTAGT-3. Both primers are located 2.1 kb upstream of the transcription start site of the AR gene. An [-³²P]dCTP-labeled AR cDNA was used as a hybridization probe in the secondary screening.

Southern Blot Analysis of AR 5-Flanking Region

Southern blot hybridization using different probes was used to confirm the identity of the genomic clone AR5. As shown in Fig. 2A, probe S spans from 1536 to 1174 of AR5; probe Co spans from 1099 to 468, which is common to both AR5 and the AR gene of Ruepp et al. (21); and probe G spans from 3339 to 2765 of the AR gene identified by Ruepp et al. (21). Probe G was prepared by polymerase chain reaction using a pair of amplification primers, Prpr1 (3339 to 3320, 5-ATGCAACTAGAGTGTCCACC-3) and Prpr2 (2765 to 2785, 5-AGGTTCTACTGGTTCTACTAC-3). The polymerase chain reaction product was cloned into a plasmid vector and verified by DNA sequence determination. Caucasian and Chinese DNAs were digested with KpnI, HindIII, and EcoRV restriction enzymes and hybridized to radioactively labeled probes S, Co, and G, respectively, in a Southern blot.

Two luciferase reporter vectors (pGL3), one controlled by the SV40 promoter (SVLuc) and the other being promoterless (BLuc), were purchased from Promega (Madison, WI). -Galactosidase reporter vector controlled by the adenovirus promoter (ADGal) was purchased from CLONTECH. The various reporter constructs are shown schematically in Fig. 3. 2.8AR.BLuc was constructed by subcloning a 2.8-kb fragment of AR 5-flanking sequence, which spans from 2733 to +41, into the unique HindIII-NcoI site immediately upstream of the luciferase gene of BLuc. A series of progressively shorter AR constructs, namely Del1.BLuc, Del2.BLuc, Del3.BLuc, Del4.BLuc, Del4.1.BLuc, and Del5.BLuc, was made by restriction digestion of 2.8AR.BLuc with SacI and HindIII in the multiple cloning site on the backbone of BLuc, followed by a unidirectional nested deletion in the 2.8-kb human AR 5-flanking region of 2.8AR.BLuc using the exonuclease III/mung bean nuclease deletion kit (Stratagene). The lengths of different deletions were determined by DNA sequence analysis and are indicated in Fig. 3. To confirm the location of the osmotic response region, p132-SVLuc was made by inserting a 132-bp fragment (from 1235 to 1104) into the KpnI-SmaI site immediately upstream of the SV40 promoter of SVLuc.

To better understand if there is more than one putative osmotic response element within the 132-bp fragment, six double-stranded oligonucleotides were made and then engineered to form different reporter constructs as described below (nucleotides that differ from the wild-type sequence are in boldface).

pAB-SVLuc was constructed by cloning fragment AB into the KpnI-MluI site upstream of the SV40 promoter of SVLuc, pC-SVLuc by inserting fragment C into the MluI-XhoI site of the SV40 promoter of SVLuc, pABC-SVLuc by inserting fragment C into the MluI-XhoI site of pAB-SVLuc, and pABCD-SVLuc by inserting fragment D into the BglII site of pABC-SVLuc. pABmCD-SVLuc was constructed by inserting fragment mC into the MluI-XhoI site and fragment D into BglII site of pAB-SVLuc, respectively; and pmABCD-SVLuc and pAmBCD-SVLuc were constructed by inserting fragments mAB and AmB into the KpnI-MluI site of SVLuc, respectively, followed by inserting fragment C into the MluI-XhoI site and fragment D into the BglII site of SVLuc. The sequences of all the constructs were verified by DNA sequence determination.

The sequence of the 5-flanking region of AR5 was obtained by sequencing the clones generated by nested deletions. Sequencing was performed using the fmol^TM DNA sequencing system purchased from Promega and later using the ABI prism dye terminator cycle sequencing ready reaction kit purchased from Perkin-Elmer.

DNA Transfection and Luciferase, -Galactosidase, and Protein Assays

Plasmid DNAs were purified as described (25) with CsCl ultracentrifugation steps. The promoter activities of various constructs were determined in transiently transfected Chang liver cells. In each transfection experiment, Chang liver cells were seeded into six 60-mm culture dishes (6 × 10⁵ cells/dish) and incubated for 24 h in isosmotic medium before transfection. Cells were then cotransfected with 2 µg of luciferase reporter construct and 1 µg of -galactosidase reporter construct using Tfx-50 reagent (Promega) in serum-free medium. After incubation for 1 h at 37 °C, the medium was removed, and the cells were incubated in fresh isosmotic medium (300 mosmol/kg of H₂O) for 12 h. A set of three dishes was changed to fresh isosmotic medium, while the other set was changed to fresh hyperosmotic medium (100 mM NaCl; 500 mosmol/kg of H₂O). After incubation for 12 h, the cells were harvested after washing with phosphate-buffered saline and adding 400 µl of lysis buffer (Promega).

Luciferase activity expressed as relative light units was assayed by mixing 10-20 µl of cell extracts with luciferase assay reagent (Promega) according to manufacturer's protocol, and the light emitted was measured by a Lumat LB 9507 luminometer (EG&G Berthold, Berlin, Germany). -Galactosidase activity was measured in the same way, except that the cell extracts were assayed with reaction buffer containing chemiluminescent substrate (Tropix Inc., Bedford, MA). Total cell protein was determined with a Bio-Rad protein assay kit using -globulin as the standard. Pilot studies showed that the amount of cell extract used falls within the linear range of luciferase and -galactosidase assays.

In each transfection experiment, the luciferase activity in relative light units/µg of total cell protein was normalized by -galactosidase activity in relative light units/µg of total cell protein. A minimum of three independent transfections were done for each construct. Results are expressed as means ± S.D.

Chang liver cells (1 × 10⁷) were initially grown in isosmotic medium (300 mosmol/kg of H₂O) in 100-mm tissue culture dishes. The medium was replaced by either fresh isosmotic or hyperosmotic medium (100 mM NaCl; 500 mosmol/kg of H₂O) for another 24 h. The cells were washed and scraped into 5 ml of cold phosphate-buffered saline and centrifuged for 1 min at 4 °C, and the pellet was resuspended in 1 ml of cold Buffer A (10 mM HEPES-KOH, pH 7.9, at 4 °C, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride). The cells were again pelleted and resuspended in 45 µl of Buffer A with 0.05% Nonidet P-40. Samples were incubated for 5 min and centrifuged for 10 min. The supernatant fraction was discarded, and the pellet was resuspended in 10 µl of cold Buffer C (20 mM HEPES-KOH, pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.5 mM dithiothreitol, and 0.5 mM phenylmethylsulfonyl fluoride) and incubated for 15 min. After vortexing, the cellular debris was removed by centrifugation for 10 min at 4 °C, and the supernatant fraction containing the DNA-binding proteins was diluted with 60 µl of modified Buffer D (20 mM HEPES, pH 7.9, 0.05 M KCl, 0.2 mM EDTA, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, and 20% glycerol), divided into aliquots, and stored at 70 °C. Protein concentration was determined by the Bio-Rad protein assay kit using -globulin as the standard.

The 132-bp fragment (1235 to 1104) containing the osmotic response region was released from p132-SVLuc by digestion with the restriction enzymes KpnI and XhoI. The resulting fragment was then labeled with [-³²P]dCTP using Klenow fragment. Nuclear extracts (1-4 µg) were incubated with 20,000 cpm of ³²P-labeled 132-bp probe in 20 µl of binding reaction mixture containing 12 mM HEPES-KOH, pH 7.9, 60 mM KCl, 5 mM MgCl₂, 0.12 mM EDTA, 0.3 mM phenylmethylsulfonyl fluoride, 0.3 mM dithiothreitol, 2 µg of poly(dI·dC), and 1 µg of pGEM-7zft (Promega). After incubation at room temperature for 10 min, the mixture was electrophoresed on a 4% polyacrylamide gel (79:1 acrylamide/bisacrylamide) in 45 mM Tris borate, 45 mM boric acid, and 2 mM EDTA at 4 °C. The gel was then dried and exposed for 1-2 days at 70 °C with intensifying screens. For competition experiments, 10- and 50-fold molar excesses of unlabeled 132-bp fragment were added to the reaction mixture prior to the addition of radiolabeled probe.

RESULTS

Hyperosmolality has been shown to induce the expression of the AR gene in a number of cell lines tested in addition to renal cell lines (26-28). We found that Chang liver cells, a human liver cell line, also induce AR expression in response to hyperosmotic stress. Fig. 1 shows the time- and osmolality-dependent changes in AR mRNA in Chang liver cells following exposure to medium supplemented with either 50 mM (380 mosmol/kg of H₂O) or 100 mM (500 mosmol/kg of H₂O) NaCl. Control cells showed a relatively constant level of AR mRNA throughout the time course of 36 h. In cells maintained in hyperosmotic medium, there was an induction of AR mRNA expression within 12 h, reaching a maximum within 24 h and remaining at steady state at 36 h. After 24 h, there was a 30-fold induction of AR mRNA in cells maintained in hyperosmotic medium. The maximum hyperosmotic induction was higher in 100 mM NaCl-treated cells than in 50 mM NaCl-treated cells despite a slight decrease of -actin mRNA expression in 100 mM NaCl-treated cells. The above results suggest that the expression of the AR gene in Chang liver cells is regulated by high salt level in the medium; hence, this cell line was used as a host to identify the cis-acting element(s) in the AR gene involved in osmotic regulation of transcription.

Cloning and Confirmation of Human AR 5-Flanking Region

We have previously obtained a human AR genomic DNA clone (1K8) containing sequence up to 2.7 kb upstream of the transcription start site (24). In order not to miss the regulatory sequence farther upstream, we set forth to clone larger fragments of the upstream sequence. Six clones were isolated from primary screening of the human genomic library using oligonucleotides deduced from the 1K8 sequence (see "Materials and Methods") as probes. Two clones were confirmed to contain the AR gene using AR cDNA as a probe in the secondary screening and by comparing part of their sequences with published data (29, 30). One of the two clones (AR5) was found to span 15 to +4 kb relative to the transcription start site of the AR gene. Preliminary experiments showed that a putative ORE is located within a 2.8-kb fragment containing the transcription start site and 41 bp of exon 1 (see below). The sequence of this fragment and exon 1 and 513 bp of intron 1 sequence from AR5 were determined and deposited in the GenBank^TM/EMBL Data Bank. The sequence from positions 650 to +617 is identical to the available published sequence with the exception of 8 bp (30), indicating that the 2.8-kb fragment contains the region upstream of the AR gene, and it is unlikely that it is part of an AR pseudogene. A discrepancy became apparent between this region and a corresponding region of a putative aldose reductase gene promoter containing a functional ORE (clone 6A3a) located ~3.7 kb upstream of the transcription start site (21). The sequence of this 5-flanking upstream region was found to be entirely different from that of AR5 upstream of position 1066 (an MboI site), whereas the sequences downstream of that position are identical in the two clones. This suggested that either AR5 or 6A3a consists of two disjoined fragments linked together during cloning. This was tested by Southern blot hybridization, and the results, shown in Fig. 2, indicate that the 2.8-kb fragment within AR5 is contiguous in the genome, while clone 6A3a probably consists of part of the AR gene linked to a fragment from a different part of the genome. This artifact occurred during the construction of the partially digested MboI human genomic library (21), CLONTECH.

To locate the ORE of the human AR gene, the 2.8-kb fragment of the genomic clone AR5 containing 2694 to +41 of the AR gene was fused to the luciferase reporter gene (2.8AR.BLuc) and transfected into Chang liver cells. The -galactosidase reporter gene under the control of the adenovirus promoter was included in every transfection to monitor the efficiency of transfection. Twelve hours after transfection, cells were switched to either isosmotic or hyperosmotic medium. After 12 h of incubation in these media, cells were harvested for luciferase and -galactosidase activity assays as described under "Materials and Methods." As shown in Fig. 3, transfected cells incubated in hyperosmotic medium have 12.5-fold higher luciferase activity than those incubated in isosmotic medium, suggesting that the 2.8-kb fragment contains an ORE as previously shown for the rabbit aldose reductase gene (31). Inclusion of a 5-kb upstream fragment did not affect the osmotically induced transcription activity (data not shown), and this fragment was not analyzed further.

The location of the ORE within the 2.8-kb fragment was further delimited by 5 to 3 nested deletions and testing the shortened fragments in transient transfection assays as described earlier. The results are summarized in Fig. 3. Deletions from 2694 to 1900 did not appreciably affect the osmotically induced activity of the reporter gene. There was a slight drop in osmotic inducible activity when the sequence was shortened from 1900 to 1500 and another minor drop when shortened to 1235. The AR 5-flanking sequences deleted to 1235 can still be induced by hyperosmotic medium by 7.6-fold over those in isosmotic medium. Further deletion to 1104 totally obliterated the osmotic response. These results indicate that the major ORE is located in the 132-bp sequence between 1235 and 1104 and that there may be an auxiliary ORE between 1900 and 1235 that is required for full osmotic response.

To test if the 132-bp sequence alone is sufficient to direct osmotically induced transcription and to test if this sequence can enhance transcription of the heterologous promoter, it was fused to the SV40 promoter-driven luciferase reporter gene and transfected in Chang liver cells as described above. Hyperosmotic medium induced the luciferase activity of this construct (called p132-SVLuc) to a 10-fold higher level compared with that in isosmotic medium (Fig. 4B), indicating that the osmotic response region within the 132-bp fragment is able to enhance the transcription of the heterologous promoter.

Analysis of the 132-bp fragment revealed three sequences that resemble part of TonE (TGGAAAAGTCCA) of the canine betaine transporter gene (19) and the ORE (CGGAAAATCAC) of the rabbit AR gene (20) (Fig. 4A). OreA (TGGAAAAATAT, 1230 to 1220) and OreC (TGGAAAATCA, 1157 to 1148) are in the same orientation as TonE. OreB (TGGAAAAATTT, 1188 to 1198), on the other hand, is in the opposite orientation. Sequence D (1117 to 1111), highly homologous to the AP-1 consensus sequence, is located adjacent to these sequences. It is noteworthy that an AP-1 consensus sequence was also identified at the 3-end of the betaine tonicity element. To further examine if all three TonE-like sequences and the AP-1-like sequence within the 132-bp fragment are involved in osmotic regulation of transcription, we cloned these sequences into the SVLuc vector and tested them in transfection assays as indicated in Fig. 4B. Fragment AB contained OreA and OreB, fragment C contained OreC, and fragment D contained the AP-1 sequence. Assessment of the ability of these fragments to respond to hyperosmotic induction in transfected cells was performed as described above. The induction ratio (luciferase activity in hyperosmotic medium/luciferase activity in isosmotic medium) of fragments AB, C, and D linked together was 4.9, while that of fragments AB and C was 4.0, indicating that the AP-1-like sequence does not play an important role in the osmoregulation of AR gene transcription. When base changes were introduced into OreA, changing the core sequence from TGGAAAA to TTTCGAA in the pmABCD-SVLuc construct, the induction ratio dropped to 2.4, suggesting that OreA is required for full osmotic response. Similarly, when mutations were introduced into OreB, changing the core sequence from TGGAAAA to TCTGGCA in the pAmBCD-SVLuc construct, the induction ratio dropped to 1.6, indicating that OreB is also required for full osmotic response. Among the three sequences, the sequence of OreC was closest to the rabbit ORE and the canine TonE sequence. When a single base substitution was made in OreC, changing the core sequence from TGGAAAA to TAGAAAA in the pABmCD-SVLuc construct, the induction ratio dropped to 0.6, indicating that OreC is the key component of the osmotic response region. However, with fragment C alone, the induction ratio was only 0.9. Therefore, OreC alone is not sufficient to induce transcription under hyperosmotic stress. Similarly, fragment AB alone also did not respond to hyperosmotic induction. These results suggest that the osmotic response region in human AR consists of at least three TonE-like sequences. Each of these sequences alone is unable to induce transcription.

To see if the sequence within the putative osmotic response region interacts with transcription factors, we performed gel mobility shift assay using the 132-bp fragment (1235 to 1104) as a probe. As shown in Fig. 5, two prominent bands were observed when the probe was incubated with nuclear extracts from Chang liver cells exposed to hyperosmotic medium (lane 3), suggesting that there may be two different types of DNA/protein interaction. Nuclear extracts from Chang liver cells maintained in isosmotic medium resulted in a similar banding pattern, but a much weaker signal (lanes 2 and 4). These bands could be competed out by the addition of excess unlabeled 132-bp DNA (lanes 4-6 and 7-9), indicating the specificity of the DNA/protein interaction. These results suggest that Chang liver cells cultured in isosmotic medium express a very low level of transcription factor(s), whose expression is highly induced when cells are incubated in hyperosmotic medium. Alternatively, in isosmotic medium, an inactive form of the transcription factor(s) is made, and it is activated in response to hyperosmotic stress.

DISCUSSION

The induction of aldose reductase or aldose reductase mRNA by hyperosmolality has been demonstrated in a variety of cells. These include a cell line derived from rabbit renal papillae (15, 16), human renal proximal tubule cells (27), Chinese hamster ovary cells (26), a lens epithelial cell line from a transgenic mouse (32), human retinal pigment epithelial cells (28), human embryonic epithelial cells (33), and, in our study, Chang liver cells. Therefore, it appears that hyperosmotic induction of AR is not unique to renal medullary cells, which are constantly exposed to extreme hyperosmolality. Many other cells also employ this protective response when they are exposed to hyperosmotic stress.

Aldose reductase and the betaine, myo-inositol, and taurine transporters represent a class of protein that is induced by hyperosmotic stress (18). At this point, how the transcription of these genes is regulated and whether they share some common transcription factors are not known. During the preparation of this report, the ORE of the rabbit AR gene was identified (20). This 11-bp element shares sequence homology with TonE of the canine betaine transporter gene. We found that within the 132-bp fragment that contains the osmotic response region of human AR, there are three sequences that resemble these osmotic response elements, indicating that osmotic regulation of gene expression is probably similar in different mammals. The three sequences within the 132-bp fragment of human AR have to be present together for efficient induction of transcription. It is likely that a similar situation occurs in the rabbit AR gene, where a single independently active ORE has been identified. However, the osmotic induction ratio is reduced from 9.4 to 2.0 when the sequence is reduced from 3221 to 11 bp, and the authors suggest the possibility that other cis-elements may potentiate the osmotic response. A similar gradual reduction in the induction ratio upon shortening of test fragments containing TonE was also observed in the canine betaine transporter gene (19). Interestingly, the concatemer of TonE of this gene and the concatemer of the ORE of the rabbit AR gene induced transcription to a much higher level than a single element. Among the three putative OREs that we identified, OreC is most similar to TonE and to the rabbit AR ORE, and it contributes the most to the osmotic induction of transcription. Yet, even this element, when fused to the reporter gene alone, did not induce transcription to any significant level. The TonE element is functional in both orientations. We found that OreB, located between OreA and OreC, is in the opposite orientation to these two sequences. This further supports the idea that these putative enhancer elements are functional in both orientations.

Regulation of gene expression in response to hyperosmolality was better studied in Escherichia coli. In response to hyperosmotic stress, there is an increase in the expression of the Pro-U operon, which encodes a high affinity betaine transporter (34). A DNA-binding protein (H-NS) acts as a transcriptional repressor of the Pro-U operon under isosmotic conditions. The inhibitory effect of H-NS is relieved by extracellular hyperosmolality, resulting in transcription of the operon (35). The regulation of AR by hyperosmolality is quite different from that of the Pro-U operon of E. coli because the results of the gel mobility shift assays (Fig. 5) showed that there is an increase in the transcription factor(s) binding to the putative ORE under hyperosmotic conditions. It is interesting that using the 132-bp fragment as a probe, gel mobility shift assay generated two bands. Although our transfection assays showed that all three putative OREs within this 132-bp fragment are necessary for osmotic response, it is possible that, under our in vitro conditions, one or more of the putative OREs are occupied by transcription factors, resulting in two different types of protein-DNA complex. Alternatively, it is possible that this reflects interactions between OREs. The slower migrating band may contain aggregates of protein-DNA complexes generated through protein/protein interaction.

Although hyperosmolality also activates immediate-early genes like egr-1 and c-fos (36) and genes in the signal transduction pathway such as jnk (37), phospholipase C, Raf-1 kinase, mitogen-activated protein kinase kinase, and mitogen-activated protein kinase (38, 39), there is still little information available regarding the mechanism by which the signal of hyperosmolality mediates gene expression. Identification of the trans-acting factors that interact with the osmotic response region may help us elucidate this osmolality signal transduction pathway of gene regulation.