AUF1 Is Expressed in the Developing Brain, Binds to AT-rich Double-stranded DNA, and Regulates Enkephalin Gene Expression*

During our search for transcriptional regulators that control the developmentally regulated expression of the enkephalin (ENK) gene, we identified AUF1. ENK, a peptide neurotransmitter, displays precise cell-specific expression in the adult brain. AUF1 (also known as heterogeneous nuclear ribonucleoprotein D) has been known to regulate gene expression through altering the stability of AU-rich mRNAs. We show here that in the developing brain AUF1 proteins are expressed in a spatiotemporally defined manner, and p37 and p40/42 isoforms bind to an AT-rich double-stranded (ds) DNA element of the rat ENK (rENK) gene. This AT-rich dsDNA sequence acts as a cis-regulatory DNA element and is involved in regulating the cell-specific expression of the ENK gene in primary neuronal cultures. The AT-rich dsDNA elements are present at ∼2.5 kb 5′upstream of the rat, human, and mouse ENK genes. AUF1 proteins are shown here to provide direct interaction between these upstream AT-rich DNA sequences and the TATA region of the rENK gene. Double immunohistochemistry demonstrated that in the developing brain AUF1 proteins are expressed by proliferating neural progenitors and by differentiating neurons populating brain regions, which will not express the ENK gene in the adult, suggesting a repressor role for AUF1 proteins during enkephalinergic differentiation. Their subnuclear distribution and interactions with AT-rich DNA suggest that in the developing brain they can be involved in complex nuclear regulatory mechanisms controlling the development- and cell-specific expression of the ENK gene.

. Although this region alone can control the rate of ENK gene transcription (7,10), it is insufficient to govern the precise developmental-and cell-specific expression of the ENK gene. However, an ϳ2700-bp-long part of the rat ENK (rENK) gene, which includes the PRC, has been shown to be necessary and sufficient to govern correct cell-specific expression (3,13,14).
Regulation of the half-life of mRNAs has been emerging as an important control mechanism during development (15). Frequently, such regulation is mediated via AU-rich sequences, located at the 3Ј-untranslated region of mRNAs (16). Proteins that bind these sequences include the AU-binding factor 1 (AUF1), also called hnRNP D (17,18). AUF1 has the characteristic domain structure that includes two RNA binding domains, oligomerization domain and a glycine-arginine-rich domain (19,20). This domain structure is shared by many other hnRNP proteins (18,21). It has four isoforms, p37, p40, p42, and p45, that are the result of alternative splicing of exon 2 and exon 7 (20,22). AUF1 has also been described to participate in other regulatory mechanisms. For example, p40 has been shown to interact with the TATA-binding protein and can activate reporter gene expression in a cell culture model (23).
AT-rich dsDNA elements, which constitute the core sequence of matrix-or scaffold attachment regions (MAR/SAR) (24) and their binding proteins such as nucleolin (25) and SATB1 (26), have been shown as key regulators of gene expression during cellular differentiation. Their molecular functions involve complex interactions with other nuclear regulatory proteins such as chromatin remodelers (27,28).
Our systematic search for transcription factors that control rENK gene expression during neural differentiation (3,14) has indicated that AUF1 is participating in this process. Here we show that AUF1 proteins are expressed by progenitors and differentiating neurons and regulate ENK gene expression in an AT-rich dsDNAdependent manner.

EXPERIMENTAL PROCEDURES
Isolation of Nuclear Proteins, DNA Probes-Microdissection of brain regions, isolation of nuclear proteins, and isolating and labeling DNA were carried out as published previously (29). For the magnetic bead-based protein-DNA binding assay, the 5Ј upstream region of the rENK gene from the pRESS1 vector (a kind gift of Dr. Steve Sabol) was either cleaved by restriction endonucleases or DNA fragments were generated by PCR using specific primers (Fig. 1). Fragments were radioactively labeled by [␣-32 P]dATP incorporation using Klenow polymerase and purified as described previously (29). The original ϳ700-bp-long fragment h was subcloned, and the resulting fragments (h 2-3 ) were radioactively labeled as above and tested for binding by EMSA (see below).
Protein-mediated DNA-DNA Interaction (PMDDI) Assay-The assay is based on biomagnetic separation technology using streptavidin-coated superparamagnetic beads (Dynal Corp., Oslo, Norway). The procedure is illustrated by the flowchart in Fig. 2. The isolated (proximal) DNA fragment a, representing the TATA region of the rENK gene, was filled in at the EcoRI site with biotinylated dATP and dNTP. The biotinylated fragment was then conjugated to M-280 streptavidin-coupled paramagnetic beads according to the manufacturer's instructions (Dynal Corp.). Conjugated beads were washed three times with 200 l of buffer containing 12 mM HEPES, pH 7.9, 100 mM KCl, 10 mM NaCl, 1 mM MgCl 2 , and 5% glycerol and stored at 4°C. In our experience, conjugated beads can be stored up to 4 weeks at 4°C. Distal DNA fragments representing the entire 5Ј regulatory region of the rENK gene were generated and labeled as above. For the PMDDI assay, 0.1 pmol of the magnetic beadconjugated fragment a was mixed with 2 g of nuclear extracts isolated from various brain regions at various ages (14) and 5 fmol of radioactively labeled (distal) DNA fragments (a through k) in 12 mM HEPES, pH 7.9, 100 mM KCl, 10 mM NaCl, 1 mM MgCl 2 , and 5% glycerol. To block nonspecific binding, the mixture contained 0.2 g per l of poly[d(I-C)] and 0.1% BSA. Protease activities were blocked by including 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 10 g per l of leupeptin, 10 g per l of aprotinin in the binding mixture. The binding FIGURE 1. The upstream regulatory region of the rENK gene and the names, locations, and sizes of DNA fragments generated for PMDDI and other assays. DNA fragments were either generated by restriction digest or by PCR amplification as indicated. The sizes of fragments are proportional.
reaction was incubated at room temperature for 30 min under constant gentle mixing. The binding was terminated by magnetic separation, and the beads were washed three times in 100 l of 12 mM HEPES, pH 7.9, 100 mM KCl, 10 mM NaCl, 1 mM MgCl 2 , 0.5 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 10 mg/ml leupeptin, 10 mg/ml aprotinin. The washed beads were resuspended in 10 l of buffer and transferred to 3MM Whatman filter paper and dried, and radioactivity was quantified by using a PhosphorImager.
EMSA-The 5Ј upstream region of the rENK gene from the pRESS1 vector (a kind gift of Dr. Steve Sabol) was either cleaved by restriction endonucleases or DNA fragments were generated by PCR using specific primers as above. The fragments were radioactively labeled as described earlier (29). Fragment h (Ϫ2,067; Ϫ2,823 BssHII-BamHI ) was subcloned, and the resulting 500-and 300-nucleotide-long fragments (h 2 and h 3 , respectively) were radioactively labeled as above and tested for binding by EMSA. In the competition assay, unlabeled DNA fragments were used as above in 100ϫ excess prior to the addition of the nuclear extract ("head-to-head" competition). The ATrich regions of mouse (mAT ENK , Ϫ3150 -30622; GenBank TM accession number U20894) and human ENK (hAT ENK , Ϫ3980 -3901 sequence information was provided by Dr. Hyman) genes were chemically synthesized, radioactively labeled, and annealed as described (29). For supershift assay, 1 l of polyclonal rabbit anti-AUF1 antibodies (kind gifts of Drs. Brewer and Tolnay) or rabbit preimmune serum and a control antibody (rabbit anti-EGFR;, Santa Cruz Biotechnology) were used. Antibodies were mixed with 2 g of nuclear extracts or with 1 ng of recombinant AUF1 protein, in the presence of 0.1 g per l of poly[d(I-C)] in the binding reaction buffer described for EMSA. The reaction mixture was incubated on ice for 30 min, and radioactively labeled double-stranded rAT ENK or rAT ENK mut probes were then added, and the mixture was further incubated for 10 min on ice and analyzed using an agarose thin layer gel system (30).
UV and Chemical Cross-linking-The rAT ENK DNA sequence was labeled by filling in radioactive nucleotides. For the binding reaction 10 fmol of radioactively labeled probe was mixed with 2 g of nuclear extracts in the presence or absence of distamycin and 0.1 g per l of poly[d(I-C)] in 12 mM HEPES, pH 7.9, 100 mM KCl, 10 mM NaCl, 1 mM MgCl 2 , and 5% glycerol in a total volume of 10 l, and the binding reaction was incubated at 0°C for 10 min. Then the mixture was exposed to UV irradiation in a UV Stratalinker 1800 cross-linker, using the AutoCross Link program according to the manufacturer's instructions (Stratagene, La Jolla, CA). After UV irradiation, 2 units of DNase I (Worthington) was added to the reaction mixture, and the cleavage reaction was allowed to proceed for 10 min at 22°C. The proteins were separated on NuPAGE 4 -12% MOPS SDSpolyacrylamide gel along with molecular weight markers according to the manufacturer's instructions (Invitrogen). The FIGURE 2. Flowchart of the PMDDI assay. Nuclear proteins (they were isolated from different brain regions at various developmental ages for this paper) are incubated with one of the DNA fragments of interest (the PRC of the rENK gene) that is conjugated to paramagnetic beads and with the second DNA fragment of interest that is radioactively labeled (from the 5Ј upstream regulatory region of the rENK gene for this paper). After magnetic separation and several washing steps, the protein-DNA-DNA complexes are dotted on filter paper, and radioactivity indicating protein mediated DNA-DNA interactions between the two DNA fragments is quantified by using PhosphorImaging.  -R  ATG GTT AAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TAT TTT TGG TTC  rAT ENK  Mut -F  GAA CTG TTC TTA CTA TCG CCA TCA CCG TTA TTG CTC AAT TCG TAC TAG TTC ATG AAG TCT TTG G  rAT ENK  mut -R  CCA AAG ACT TCA TGA ACT AGT ACG AAT TGA GCA ATA ACG GTG ATG GCG ATA GTA AGA ACA GTT C  oATT-F  TTT AGC TCA GTG GTA GAG CGC  oATT-R  GTT TCT GCC CTT TCC AAC  gel was transferred by vacuum onto 3MM filter paper and was subjected to autoradiography. DNase I Footprinting-Fragment h was cleaved by HindHIII digest, and the resulting fragment (h 2 ) was further digested by AflII and StuI. The ϳ300-bp-long subfragment (h 3 ) was subcloned into pUC19 plasmid. The insert containing the rAT ENK sequence was asymmetrically labeled by filling in the AflII site with radioactive nucleotides. Footprinting was performed as described earlier (14). For the footprint reaction, 10,000 cpm/l of radioactively labeled probe was mixed with 1 g of poly[d(I-C)] and 3 g of nuclear extracts. The mixture was incubated on ice for 20 min, and then the reaction mixture was transferred to 22°C for 2 min. After the incubation, 0.5 units of DNase I enzyme (Worthington) was added to the mixture in 1 l, and the cleavage was allowed to proceed for 1 min at room temperature. In DNase I titration reactions 2.0, 1.0, and 0.5 units of DNase I were added to the DNA without the addition of nuclear extracts (naked DNA). After the 1-min incubation 10 g of counter-transcribed RNA was added to the reaction mixture in 1 l immediately followed by the addition of 50 l of DNase I stop solution (Stratagene). The reaction mixture was precipitated, washed by ethanol, dried, and dissolved in water to 3000 cpm per l concentration. Nine thousand cpm of the DNase I-digested fragments were separated on each lane of a 6% denaturing polyacrylamide sequencing gel, and the gel was subjected to autoradiography . As marker, G-reaction ladder was generated by using a Maxam-Gilbert sequencing kit (PerkinElmer Life Sciences) (data not shown).
DAPSTER Assay-The assay was performed as described previously with minor modification (31). Nuclear extracts were prepared as published previously (29). Fifty g of nuclear extract was precleared with 30 l of streptavidin-agarose beads (Pierce) washed in buffer Z (25 mM HEPES, pH 7.9, 20% glycerol, 0.1% Igepal, 0.1 M KCl, 12.5 mM MgCl 2 , 1 mM dithiothreitol, 0.1 M ZnCl 2 ). After separation by centrifuging, 75 pmol of competitor ds-or ssDNA or an equal volume of 1ϫ kinase buffer (50 mM Tris/HCl, pH 7.5, 10 mM MgCl 2 ) was added in the presence of 0.1 g/l poly[d(I-C)], 1 mM sodium orthovanadate, and Complete Mini protease inhibitors (Roche Applied Science) and incubated under gentle rotation for 15 min at 4°C. 5 pmol of biotinylated ds-or ssDNA, rAT ENK or rAT ENK mut , were added to the recovered nuclear extract and mixed by gentle rotation for 2 h at 4°C. (See sequences in Table 1.) Thirty l of washed streptavidin-agarose bead slurry was added, and the suspension was further incubated by gently rotating for 2 h at 4°C. The mixture was centrifuged briefly, and the pelleted beads were washed three times in buffer Z, and beads were boiled in the presence of lithium dodecyl sulfate sample buffer (Invitrogen) under reducing conditions. The bound proteins were separated on NuPAGE 4 -12% BisTris gels, transferred onto polyvinylidene difluoride membranes by electroblotting (Invitrogen) followed by incubating the membranes with various primary antibodies overnight at 4°C. Immunoreactive protein bands were visualized using SuperSignal West Pico chemiluminescent kit (Pierce). The membranes were stripped using the Restore Western blot stripping mixture (Pierce) and reprobed with anti-actin antibody (Sigma) as described above.
Chromatin Immunoprecipitation-The assay was performed by optimizing existing protocols (32) for embryonic and neonatal brain tissues. Briefly, frozen microdissected brain regions were placed into 0.5 ml of ice-cold PBS containing 1% formaldehyde and incubated first on ice for 10 min and then at 37°C for 30 min. Cross-linking reaction was stopped by the addition of 0.75 ml of cold PBS, followed by 3 washes with 1 ml of cold PBS each. After the last wash, 0.9 ml of RT lysis buffer (50 mM Tris/HCl, pH 8.1, 10 mM EDTA, 1% SDS, and protease inhibitors) was added, and the mixture was incubated on ice for 10 min with occasional mixing by gentle pipetting. The cells were disrupted by sonication on ice (four times for 10 s at level 4, separated by 30-s breaks). Debris was removed by centrifuging, and the supernatant was diluted 10-fold with dilution buffer (16.7 mM Tris/HCl, pH 8.1, 167 mM NaCl, 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, and protease inhibitors) and filtered through a 45-m pore size filter. Aliquots (500 l) were precleared with DNA-and BSA-blocked protein A/G-coupled Sepharose beads (Amersham Biosciences) and with preimmune serum and beads. The supernatant was recovered and was incubated with 1 g of the following antibodies at 4°C overnight: a pan-specific anti-AUF1 antibody (a kind gift from Dr. Tolnay), anti-hnRNPA2/B1 antibody (Santa Cruz Biotechnology), or EGFR antibody (Santa Cruz Biotechnology). Sixty l of DNA-and BSA-blocked, protein A/G-coupled Sepharose beads were added to each sample, and they were rotated for 1 h at 4°C. Subsequently, the samples were washed five times in ice-cold wash buffer 1 (20 mM Tris/HCl, pH 8.1, 150 mM NaCl, 2 mM EDTA, 0.1% SDS, 1% Triton X-100), once in RT wash buffer 2 (Tris/HCl, pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% Nonidet P-40, 0.5% sodium deoxycholate), and twice in RT wash buffer 3 (Tris/HCl, pH 8.0, 1 mM EDTA). After the final wash, beads were resuspended in 100 l of wash buffer 3 in the presence of 0.5% SDS and 0.5 mg/ml proteinase K and incubated at 37°C for 12 h and then 65°C for 12 h. DNA was purified by phenol/chloroform/isoamyl alcohol (25:24:1) extraction and was ethanol-precipitated. After washing and drying, DNA was resuspended in 10 l of 10 mM Tris/HCl, pH 7.5, and was size separated on a 1.5% TBE agarose gel. DNA in the 300 -500-bp size range was isolated from each sample and was extracted from the gel by using the agarose gel DNA extraction kit (Roche Applied Science). DNA was eluted in 50 l of Tris/ HCl, pH 8.3, and 3 l was used as template in PCR using the qPCR TM core kit (Eurogentec, San Diego) according to the manufacturer's instructions. The first PCR was carried out with 25 cycles, using primers specific to the rAT ENK , the proximal cassette, and control upstream region, respectively. The 100 times diluted product of the first reaction was used as template in the (semi-) nested reaction, performed with 25 or 30 cycles. The following PCR primers were used. The rAT ENK region was amplified with the primer pair of 5ЈoATT and 3ЈoATT (Ϫ2654; Ϫ2395) in the first reaction, whereas the nested primer pair was 5ЈnATT and 3ЈnATT (Ϫ2612; Ϫ2471). The ENK TATA region was first amplified with the sense primer 5ЈoTATA and the antisense 3ЈoTATA (Ϫ103; ϩ73). In the semi-nested reaction, the sense primer was used in the pair with the antisense primer 3ЈsnTATA (Ϫ103; ϩ25). A distinct regulatory region containing the septamer site (74) located between Ϫ569 and Ϫ247 of the rENK gene was used as control and was amplified with primers 5ЈoSept and 3ЈoSept. The nested primers were 5ЈnSept and 3ЈnSept (Ϫ560; 287). (See sequences in Table 1.) The PCR products were separated on 2.5% agarose gel by agarose gel electrophoresis along with a 100-bp molecular weight ladder; the bands were visualized with ethidium bromide.
Expression Cloning-Neonatal rat brain pEAK8 cDNA library (Edge BioSystems, Gaithersburg, MD) of 500,000 colony formation units was divided into 960 pools, representing ϳ500 clones per pool complexity. Plasmids were isolated from the pools, and PEAK Rapid embryonic epithelial cells were transfected with each plasmid pools using the calcium phosphate transfection protocol (Edge BioSystems, Gaithersburg, MD). After protein expression the cells were subjected to lysis, and each pool was tested by EMSA using radioactively labeled rAT ENK probe, and protein-DNA complexes were separated by agarose thin layer gel electrophoresis (30). A single pool that contained the rAT ENK binding activity was further divided into 96 sub-pools, and the screening procedure was repeated but this time only with the sub-pools. A single positive pool was identified, and a subsequent single clone sorted by PCR resulted in a single clone. The clone was sequenced and analyzed using the NCBI nucleotide data base.
To obtain expression vectors of the various AUF1 isoforms, the expression library pools were screened and sorted by PCR using isoform-specific primers. The expressed proteins were analyzed by using the NuPAGE 4 -12%, MOPS SDS-polyacrylamide gel system (Invitrogen), and the various cDNA fragments were cloned into the mammalian expression vector pCI (Promega, Madison, WI) according to the manufacturer's instructions.
Cell Cultures, DNA Molecular Decoy, and Reporter Assays-Animals were handled according to protocols approved by the Committee on Animal Research at Uniformed Services University of the Health Sciences. Following pentobarbital anesthesia, rat embryos were obtained from timed pregnancies; the brains were removed, and striatal and cortical primary embryonic neuronal cultures were prepared and maintained as published previously (29). C6 cells were obtained from the ATCC and cultured as described previously (33). For DNA molecular decoy assays, primary neuronal cultures were cotransfected with a DNA mixture containing the following: 0.5 g of the reporter plasmid 2700/ 703ENK-Luc, 2 g of competitor dsDNA corresponding to the rAT ENK region, or a control dsDNA (rAT ENK mut ) as described for EMSA above and 0.1 g of pCMV-Renilla luciferase. Cells were transfected by using the polyethyleneimine gene delivery system (34). Forty eight hours after transfection, cultures were harvested, and firefly and Renilla luciferase activities were measured using the dual luciferase assay system (Promega, Madison, WI). The control luciferase (Renilla or Firefly) activities were used to normalize the activities of the reporter gene for transfection efficiency.
Immunohistochemistry-Following pentobarbital anesthesia, rat embryos were obtained from timed pregnancies, and the brains were removed and fixed with paraformaldehyde by immersion. Frozen coronal sections containing the hypothalamus, cerebral cortex, striatum, and septum (or their primordia) were cut and processed for immunohistochemistry using anti-AUF1 antibodies from Dr. Schneider (1:1000). Sections were processed for immunofluorescent detection using fluorescein isothiocyanate-conjugated secondary antibodies (1:500, Jackson ImmunoResearch, West Grove, PA). Sections were viewed and photographed using an Olympos 70 IX microscope equipped with a Spot digital camera. Images were processed, and inserts were created by using Adobe Photoshop.

RESULTS
Identification of the AT-rich DNA Region as a cis-Regulatory DNA Element of the rENK Gene-As a first step toward identifying the required cis-elements for the development-and cellspecific expression of the rENK gene in the rodent brain, we generated a reporter plasmid carrying the firefly luciferase gene under the control of a 2.7-kb upstream regulatory region of the rENK gene, inclusive of the first intron and most of the second exon ϭ prENK2700 -703FIR (Fig. 3). Previous studies have implicated that this DNA region contains the cis-regulatory DNA elements necessary for the cell-specific expression of the gene (3,13). Because the PRC of the rENK gene contains the inducible cis-regulatory DNA elements as mentioned above, we also constructed a plasmid in which the firefly luciferase reporter gene was placed under the control of the PRC only (prENK190 -53FIR, same as fragment a in PMDDI assay, see Fig. 3). We then transiently transfected differentiating striatal and cortical primary neuronal cultures (29) with either reporter construct and compared the obtained reporter gene activities. When the primary cultures were transfected with prENK2700 -703FIR, the observed reporter gene activities were consistent with the levels of endogenous ENK mRNA, which are high in the striatum and very low in the cortex (1) (Fig. 3). In contrast, we observed no significant difference in reporter activities in striatal versus cortical cultures when prENK190 -53FIR reporter plasmid containing only the PRC was used. Consistent with previous observations using different systems, we concluded that the 2.7-kb upstream region contains the cis-element(s) that regulate cell-specific expression of the rENK gene. Based on these findings, we hypothesized that the difference in the observed reporter gene activities is the result of transcription factor(s) selectively expressed in striatal versus cortical neurons that directly or indirectly interact with the PRC of the rENK gene.
To identify these proteins and their DNA-binding sites, we screened for protein-DNA interactions that fulfilled the following criteria: 1) the protein-DNA interaction should show either negative or positive correlation with the anatomical distribution of enkephalinergic neurons in the adult rat brain (i.e. striatum ϭ high and cortex ϭ low); 2) the DNA-protein interactions should occur in the developing (ϳE16 to ϳP2) but not in the adult rat brain, and 3) the proteins should mediate interaction between the cis-element and the TATA region.
To identify the protein(s) and their cis-regulatory DNA elements, we have designed a magnetic bead-based assay enabling the detection of protein-mediated DNA-DNA interactions (Fig. 2). We immobilized biotinylated nucleic acids with the sequence of PRC of the rENK gene (fragment a) to streptavidincoated magnetic beads. DNA fragments (b through k) representing different portions of the 5-kb upstream regulatory region of the rENK gene were radioactively labeled (Fig. 1). Each of these fragments was separately incubated with the immobilized fragment a and nuclear proteins isolated from the developing cortex and the striatum, respectively, at selected developmental ages (14). PMDDI showed protein-dependent interaction between fragment a and fragment h, representing nucleotides Ϫ2,067 and Ϫ2,823 of the rENK gene in the presence of nuclear proteins isolated from the developing cortex between E18 and P2 (Fig. 4).
Nuclear proteins isolated from the striatum at any ages failed to mediate this interaction. No interaction was detected in the absence of proteins or in the presence of excess (50-fold) unlabeled fragment h (specific competitor). The other DNA fragments used as competitors had no significant effect on the binding. Because of its relatively large size, a portion of fragment h was subcloned and tested (see Fig. 1). The resulting fragment h 2 retained full activity as tested by PMDDI. For the identification of the DNA-binding site, we performed DNase I footprinting FIGURE 4. The upstream region of the rENK gene contains DNA-binding sites for proteins that are expressed in a brain region-and developmentspecific fashion as shown by the PMDDI assay. Nuclear extracts isolated from the cerebral cortex at ages E18 and P2 contained nuclear protein(s) that bind to both the immobilized proximal fragment a and the labeled distal fragment h. The relative binding activities of the various distal DNA fragments representing the entire 6-kb 5Ј upstream regulatory region of the rENK gene are expressed as cpm per g of nuclear protein as a function of developmental age per DNA fragment. The position of fragment h is approximately Ϫ2.5 kb relative to the transcriptional start site. on subfragment h 3 in the presence of nuclear proteins isolated from the neonatal cortex and the globus pallidus, another brain region lacking enkephalinergic neurons (1). Nuclear extracts of the neonatal cortex protected an AT-rich DNA repeat region (ATT) 19 of the rENK gene (Fig. 5A).
Nuclear extracts isolated from the neonatal globus pallidus showed a weaker but identical protection of the same AT-rich region. For further studies, the protected AT-rich region with an additional 12 nucleotides flanking at both the 5Ј and 3Ј ends was chemically synthesized, and the complementary strands were annealed. The resulting ds(ATT) 19 oligonucleotide (rAT ENK ) fully retained the behavior of fragment h 3 in PMDDI assays and was used in subsequent experiments.
To further characterize the spatiotemporal distribution of the AT-rich dsDNA binding activity, we performed EMSA. EMSA showed a single, low mobility protein-DNA complex in the presence of cortical nuclear extracts isolated from rats between ages E18 and P2. No complex was detected in the presence of striatal nuclear extracts (Fig. 5B).
These results were consistent with those observed with PMDDI. Screening four additional major brain regions (globus pallidus, thalamus, spinal cord, and cerebellum; data not shown) has demonstrated that AT-rich dsDNA binding activity and the differentiation into the enkephalinergic phenotype are negatively correlated.
In sum, the observed protein-DNA interaction displayed a strong negative correlation with the spatial expression pattern of the ENK gene and thus fulfilled criteria 1; it was detectable only during the period of phenotypic differentiation and thus fulfilled criteria 2; and the nuclear protein(s) mediated the interaction between the AT-rich DNA and the TATA region and thus fulfilled criteria 3.
A series of competitive EMSAs using the full-length of (ATT) 19 as probe with nuclear extracts derived from the neonatal cortex and truncated versions of (ATT) 19 as competitors showed that a minimum of 12 repeats ((ATT) 12 or 36 nucleotides) is required for binding. For high affinity binding (50% competition when 50-fold excess cold competitor is used), 16 repeats ((ATT) 16 ) or a 48-nucleotide-long A-and T-rich nucleotide sequence was necessary (data not shown).
We searched the 5Ј regulatory regions of the mouse and the human ENK genes (mENK and hENK) but found no (ATT) 19 repeats. However, both genes contained AT-rich sequences at far upstream locations (Fig. 5D).
To test if these AT-rich DNA regions can serve as binding sites for nuclear protein(s) and show thus they are orthologous sequences to that of the (ATT) 19 of the rat ENK gene, we synthesized these AT-rich regions of mENK and hENK genes along with their respective 5Ј and 3Ј flanks (mAT ENK , and hAT ENK , respectively). Their binding properties were compared using nuclear proteins isolated from the developing mouse and rat cortex by EMSA. Testing different combinations of DNA probes and nuclear extracts showed that there is a single protein-DNA complex, irrespective of the combination (Fig. 5C).
Moreover, competitive EMSA indicated comparable binding affinities (data not shown). These results suggested that the rat, mouse, and human AT-rich regions are orthologous sequences and that the regularity of A and T nucleotides present in the rat ENK gene is not critical for the binding.
To test if (ATT) 19 dsDNA acts as a cis-regulatory DNA element for the rENK gene, we co-transfected embryonic differentiating cortical and striatal neurons with the reporter plasmid prENK2700 -703FIR and with either the specific dsDNA com- FIGURE 5. An AT-rich DNA region of the rENK gene is the binding site for protein(s) expressed in the P2 rat brain. DNase I footprinting showing the protection at the ds rAT ENK region of the fragment h 3 . Slope indicates increasing amounts of the free probe; the nuclear extracts used were GP, globus pallidus; Cx, cortex (A). EMSA shows a single, low mobility protein-DNA complex in the presence of nuclear protein(s) isolated from E18 and P2 rat cortex. Nuclear proteins isolated from the striatum do not show binding at any other ages tested. B, 2 g of rat or mouse cortical nuclear proteins isolated from P2 and P3 brains, respectively, were probed with synthetic dsDNA rAT ENK containing the AT-rich DNA region of the rat (R), mouse (M), or human (H) ENK gene in single-stranded (ss) or double-stranded (ds) form and separated by EMSA (C). Orthologous AT-rich DNA regions are present on the rat (location, Ϫ2450), the mouse (Ϫ3150), and the human (Ϫ3980) ENK genes and serve as binding sites for nuclear protein(s) present in the developing cortex (D).
petitor rAT ENK or the control rAT ENK mut that failed to bind the protein in vitro. The reporter gene activity was substantially higher in cortical than in striatal cultures when both cultures were co-transfected with the specific competitor rAT ENK dsDNA (Fig. 6). When rAT ENK mut dsDNA was co-transfected as competitor molecule, no significant changes were observed in reporter gene expression in either neuronal culture. These results implicated the (ATT) 19 DNA element as a cis-regulatory DNA element for the rENK gene in differentiating neurons.
Because distamycin can compete with proteins that bind the minor groove of AT-rich dsDNA (39), we tested the effect of distamycin treatment on reporter gene activity using the same cortical and striatal primary neuronal cultures. Differentiating striatal and cortical neuronal cultures were transfected with the reporter gene prENK2700 -703FIR and 24 h after transfection were treated with 500 nM distamycin or with equal volume of vehicle (Me 2 SO) for 48 h. Distamycin treatment of cortical neuronal cultures resulted in a very substantial, ϳ7-fold increase in reporter gene activity (Fig. 6). Consistent with the experiment above, treating striatal neuronal cultures with distamycin resulted only in a moderate (ϳ40%) increase in reporter gene activity. These experiments suggest that the protein-DNA interaction involves the minor groove of the AT-rich dsDNA element.
Identification of AUF1 as the AT-rich dsDNA-binding Protein-As a first step toward identification of the double-stranded (ATT) 19 -binding protein(s), we have performed UV cross-linking to assess the molecular mass of the protein(s). Nuclear proteins isolated from developing cortex, thalamus, and globus pallidus, each containing high AT-rich dsDNA binding activities as detected by EMSA, were UV cross-linked to rAT ENK dsDNA, and the complexes were separated on SDS-polyacrylamide gels. Because of the significant effect of distamycin treat-ment seen in vivo, we preincubated half of the nuclear extracts with distamycin prior to UV cross-linking. UV cross-linking showed a broad protein band with the average molecular mass at around 45 kDa (Fig. 7A).
The broad appearance of the band suggested that there may be multiple proteins of slightly different molecular weights present in the complex. We also detected a sharper second band with an apparent molecular mass of ϳ100 kDa. We tentatively named the ϳ45-kDa protein as dAT1 for developmentand AT-specific protein 1 (dAT1) and the 100-kDa protein as dAT2. Preincubation of the nuclear extracts with distamycin FIGURE 6. The AT-rich element of the rENK gene (rAT ENK ) acts as cis-regulatory DNA element in differentiating neurons. Primary embryonic cortical (black bar) and striatal (white bar) neuronal cultures were co-transfected with the prENK2700 -703FIR reporter gene and either ds (ATT) 19 or control ds(AT)mut dsDNA competitor molecules or were treated with 500 nm distamycin dissolved in ethanol or with ethanol only as a vehicle control. pRSV-REN was used as control plasmid, and dual luciferase activities were measured, normalized, and expressed as in Fig. 3 above. n ϭ 3. RLU, relative light units. FIGURE 7. Identification of the ds rATENK-binding proteins in the developing rat brain. UV cross-linking of nuclear proteins derived from neonatal striatum (Str), thalamus (Th), cortex (Cx), and globus pallidus (GP) to fragment h 3 show an ϳ45-kDa (dAT1, labeled with a gray arrowhead) and a less abundant ϳ100-kDa (dAT2, labeled with a black arrowhead) binding protein (A). The presence of distamycin in the reaction mixture prevented the binding of both proteins. dAT1 was isolated by expression cloning and identified as AUF1 p37 (B). Clone candidates for fragment h 3 binding were sorted and expressed in pools marked 1-6. The host cell extract was used as negative control (Ϫ). As the positive control (ϩ), recombinant SATB1 protein was used because it binds to AT-rich dsDNA (26). Arrow shows the low mobility protein-DNA complex, whereas the * indicates a nonspecific band. The AT-rich dsDNA-binding activity in expression pool 1 is marked by an arrowhead. The AT-rich dsDNA-protein complexes formed by using recombinant AUF1 and nuclear extracts isolated from P2 cortex are super-shifted in the presence of a pan-specific AUF1 antibody (C). Two g of nuclear extract or 0.1 g of recombinant protein were incubated with radioactively labeled dsDNA (fragment h 3 ) containing the ds rAT ENK site and preimmune serum as control, or the anti-dAT1/hnRNP D antibody, or an antibody raised against HMG-I/Y. FP, free probe. almost completely prevented the formation of both ϳ45and 100-kDa complexes with every nuclear extract tested, further implicating the AT-rich minor gro0ve as a binding site for both proteins. To isolate the cDNA encoding the AT-rich dsDNAbinding proteins, we screened an embryonic rat brain expression library using radioactively labeled ds(ATT) 19 as probe (Fig.  7B) (30). Sequence analysis of the single positive clone showed that it is identical to p37, the smallest isoform of AUF1. The identity of the obtained cDNA clone and that of the corresponding protein was further verified by supershift assays (Fig. 7C).
Recombinant p37 protein and nuclear extracts isolated from the neonatal rat cortex were incubated with radioactively labeled rAT ENK dsDNA in the presence of either preimmuneserum, a pan-specific AUF1 antibody, or an antibody raised against HMG-I/Y protein. HMG-I/Y antibody was selected because HMG-I/Y protein can also bind certain AT-rich DNA elements (40). In the presence of the AUF1 antibody, the complexes formed with the recombinant p37 protein and also with the nuclear extract were supershifted. Neither the preimmune serum used as negative control nor the presence of anti-HMG-I/Y antibody altered the mobility of the complexes. We therefore concluded that the recombinant AUF1 and dAT1 present in the developing rat cortex are identical proteins. However, the complex of nuclear proteins showed a different mobility. This suggests that in addition to the p37 protein, other isoforms may also bind the AT-rich DNA or that there are proteins present in the nuclear extract that interact with AUF1 protein(s). Because AUF1 proteins have been primarily known to bind single-stranded nucleic acids (15,41), we performed competitive EMSAs in the presence of specific and nonspecific ds-and ssDNA competitors (Fig. 8).
The presence of an excess amount of rAT ENK dsDNA used as a specific competitor in the reaction mixture completely eliminated the binding of recombinant p37. Similarly, when an excess amount of single-stranded top or bottom strand of the rAT ENK sequence was present as competitor DNA, no complexes were detected. The rAT ENK mut , either in singlestranded or double-stranded form, had no effect on the binding. These experiments further demonstrated that in addition to the previously described AU-rich sequences, p37 can also bind AT-rich dsDNA.

Cell-specific Expression of AUF1 in the Developing Rat Brain-
To obtain information about the expression of AUF1 proteins at the cellular level, we have mapped the spatial and temporal distribution of AUF1ϩ cells by immunohistochemistry (Fig.  9A).
At embryonic day 14 (E14), practically all cells populating the ventricular zone (VZ) were immunoreactive for AUF1. In addition, many cells in the pallidal subventricular zone and in the telencephalic wall were positive for AUF1 suggesting its involvement in the proliferation and/or differentiation of neural progenitors. Interestingly, however, there is a difference in the intensity of AUF1ϩ cells between the lateral and medial VZs. Cells located in the lateral VZ showed much higher staining intensity than cells located in the medial VZ.
At E18, practically all cells in the developing cortex were AUF1-immunoreactive. Similarly, in the developing septum and rhinencephalon, most cells were also AUF1ϩ. In the striatum, however, only a subset of cells was immunoreactive. The proportion of AUF1ϩ cells in the striatum seemed propor- tional to the population of neurons that are expressing other neurotransmitter genes but not ENK in the adult striatum. At P2, the overwhelming majority of AUF1 immunoreactive cells was found in regions that contain very few or no enkephalinergic neurons in the adult, including the cerebral cortex, septum, thalamus, etc. These findings were consistent with our previous results obtained by using EMSA (see Fig. 5B).
Previous studies have shown that AUF1 proteins can be both cytoplasmic and also nuclear in their subcellular localization (42). By using higher power imaging, we found that AUF1 immunoreactivity in the developing rat brain is exclusively nuclear (Fig. 9B), and we have found no immunohistochemical evidence for cytoplasmic localization. Moreover, within the nucleus, AUF1 immunoreactivity was not distributed evenly. Subnuclear speckle-like "hot spots" (43) were highly immunoreactive for AUF1, although most of the nucleoplasm lacked immunoreactivity suggesting that AUF1 is involved in the regulation of many other genes in addition to ENK (Fig. 9B, inset). AUF1 Binds to AT-rich dsDNA Element in an Isoform-specific Manner-Previous reports have shown distinct regulatory functions associated with the various isoforms (20,42). To test which isoforms are expressed in the developing rat brain and which one of them binds the AT-rich dsDNA in vitro, we performed the DAPSTER assay followed by Western analysis (see "Experimental Procedures") ( Fig. 10A).
We have found that all of the four isoforms, p37, p40/42, and p45 are expressed in the developing rat brain albeit at different abundance. Quantification of band intensities following DAP-STER and Western analysis showed that the approximate ratios of the isoforms were as follows: p37:p40/p42:p45, ϳ20:60:20. DAPSTER assay also showed that only p37 and p40/42 bind AT-rich dsDNA. The presence of excess free specific competitor DNA rAT EMC 19 practically blocked the binding of p37 and largely reduced the binding of p40/42. No effect on binding was detected when an excess amount of mutant dsDNA competitor was added to the reaction.
To test if AUF1 proteins do bind the AT-rich dsDNA of the rENK gene in vivo, we performed chromatin immunoprecipitation assay using microdissected neonatal brain tissue (14,44). The cross-linked protein-DNA complexes were immunoprecipitated by anti-AUF1 or by a control antibody against hnRNPA2/B1. The DNA fragment containing the (ATT) 19 region of the rENK gene was successfully amplified from the various AUF1 antibody-precipitated chromatins with primers flanking the (ATT) 19 region (Fig.  10B) indicating in vivo interaction in the nucleus. Because our PMDDI assay showed a protein-dependent interaction between the AT-rich and the TATA region of the ENK gene, we also used primers flanking the TATA region of the rENK gene to test binding. Indeed, the DNA fragment containing the TATA region was also amplified following immunoprecipitation confirming the in vitro data. As a control, we immunoprecipitated protein-DNA complexes by using an antibody against hnRNPA2/B1 that is expressed in the developing cortex and binds to another cis-regulatory DNA element called septamer present on the rENK gene (74). 4 The AUF1 antibody-precipitated samples were negative with the septamer-specific primer set even though the DNA region containing the other cis-regulatory elements could be amplified both from the corresponding antibody-precipitated chromatin and from the input sample. These experiments confirmed the results of in vitro studies and demonstrated that in the developing brain AUF1 protein binds the AT-rich dsDNA element and also the TATA region of the rENK gene in vivo.

DISCUSSION
The aim of the present study was to identify transcription factors that regulate the developmental expression of the rENK gene. Unexpectedly, we have found that one of these transcription factors is AUF1, a known regulatory molecule of mRNA stability (17,45). As we showed above, AUF1 is expressed in a developmentand cell-specific manner in the rat brain and upon binding to the AT-rich DNA element of the rENK gene regulates its expression in differentiating neurons.
We found AUF1 based on its specific binding to the (ATT) 19 cis-regulatory DNA element present on the 5Ј (upstream) region of the rENK gene. We found that the mouse and human genes lack the regularity/repetitive nature of the rat AT-rich DNA sequence; however, they have orthologous regions at similar upstream locations. This further illustrates that in contrast to the typical short DNA-binding sites, AT-rich DNA elements do not have a consensus sequence in a classical sense (24). The binding site for proteins is a "pocket" formed on the minor groove of the AT-rich DNA (46). DNA sequences that are composed of more than 70% A or T nucleotides and span the length of three or more helical turns (Ͼ36 nucleotides) can form such "pockets." The precise dimensions of the pocket are determined by the nucleotide composition and by the frequency of A and T nucleotides. These sequences are prone to unwinding/base un-pairing under altered superhelical stress, which is believed to contribute to their regulatory functions (47). AT-rich DNA elements have been shown to regulate the expression of a wide variety of genes by modifying the chromatin structure (48,FIGURE 10. AUF1 binding to the ds rAT ENK site in vitro and in vivo. DAPSTER assay. Nuclear proteins isolated from P1 rat cortex were incubated with biotinylated ds rAT ENK in the presence or absence or various competitor DNAs. The bound (B) and free (F ) proteins were separated on a 10% acrylamide gel and analyzed by Western using a pan-specific AUF1 antibody. The isoforms are shown with an arrow by their respective position (A). For chromatin immunoprecipitation, pan-specific AUF1 antibody (A) was used. As a control (C ), we used an antibody against hnRNPA2/B1 and amplified its septamer cis-regulatory DNA element (29). 5 Input, I. Positive controls (ϩ) were amplifications from plasmids coding the respective regulatory regions. EGFR antibody was used as a negative control (Ϫ). Horizontal lines show the different regions of the rENK gene amplified after immunoprecipitation: TATA ENK , the TATA region of the rENK gene; rsept ENK , the binding site for hnRNPA2/B1 (septamer). For the location of the regions, see "Experimental Procedures" (B). 49). Several proteins that specifically bind to AT-rich MAR/ SAR DNA regions have been identified. The list includes nucleolin (25), histone H1 (50), HMGI/Y (40), and the thymus-specific developmental regulator SATB1 (26). However, we have found that SATB1 is identical to the 100-kDa AT-rich dsDNA-binding protein (tentatively named as dAT2). SATB1 (and also its related protein SATB2) are also expressed in development-and cell-specific fashion in the rat brain and are involved in regulating gene expression in differentiating neurons (33). 5 Many AT-binding proteins have a specific domain, called the AT-hook, thought to be critical to bind to AT-rich dsDNA sequences (51). However, several proteins, including SATB1, lack the AT hook and have a distinct domain for binding AT-rich DNA (52)(53)(54). AUF1 also lacks the AT-hook and shows little or no structural homology to SATB1 and to other AT-binding proteins. In summary, how AUF1 proteins bind to AT-rich dsDNA elements is currently not known. One can only speculate that dimerization or multimerization may play a role in the binding. In addition to their ability to bind AT-rich dsDNA as we have shown above, previous studies demonstrate that AUF1 proteins can directly interact with nucleolin that binds MAR/SAR DNA elements (55) and with the MAR-associated factor SAF-B (56) further implicating AUF1 in MAR/SAR-related regulation.
We found that AUF1 proteins interact with the TATA complex. Our studies showed AUF1-dependent direct or indirect interaction between the AT-rich DNA element and the TATA region of the rENK gene in vitro. These results are consistent with previous findings showing interaction between p40 and the TATA-binding protein and the transactivation of the complement 2 receptor gene in a heterologous reporter system (36). In vivo, however, the AT-rich DNA elements of the rat, mouse, and human ENK genes are all located at ϳ2.5 kb 5Ј upstream of the TATA region suggesting that other structural DNA elements of the ENK gene are required for the observed direct interaction. Indeed, we have previously identified a structural DNA element called septamer that can undergo protein-dependent DNA bending (29). We have also shown earlier that another DNA element, the TG/AC repeat of the ENK gene, can act as a "hinge" through sliding of the DNA strands in calciumdependent manner (57). Based on our observations, we have predicted that polymorphic alterations at this TG/AC repeat can have functional consequences on ENK gene expression that may affect neuropsychiatric function. Recently, a positive correlation has indeed been found between polymorphic changes at the TG/AC repeat and opiate dependence in human subjects (58,59).
We have found that in the rat brain, AUF1 proteins are expressed in a development-and cell-specific manner in the rat brain. Their spatial and temporal pattern of expression and the cell culture experiments implicate AUF1 proteins as developmental repressors of ENK gene expression. Transcription factors exert their activator or repressor functions depending on the cellular context, i.e. on the presence of interacting proteins (5). It has been suggested that the presence of multiple cis-elements are required to govern the cell-specific expression of the ENK gene (3,13).
Previous experiments using adult tissues indicated that AUF1 expression is quite ubiquitous; however, there is also evidence that distinct isoforms are expressed in a tissue-and cellspecific manner (38). We show here that AUF1 proteins are present both in proliferating progenitors and in differentiating neurons suggesting that AUF1 proteins may be directly involved in regulating the expression of multiple, developmentally relevant genes during neuronal development. The difference in AUF1 expression between the medial and lateral VZ can be potentially important given that cells originating from the lateral versus medial VZ can have a different phenotypic fate in the mature brain (66). Additional studies focusing on the precise spatial and temporal pattern of expression of AUF1 in the developing brain will be needed to identify the cell type-specific expression of AUF1.
AT-rich DNA elements are frequent in the genome (24). Accordingly, AUF1 and other AT-rich DNA-binding proteins have large numbers of binding sites on a large numbers of genes (60,61). This notion is supported by the dotted pattern of AUF1 immunoreactivity in the nuclei of developing neurons, further suggesting multiple targets for AUF1 regulation in differentiating neurons. The appearance of these AUF1 hot spots is similar to the distribution of other nuclear protein complexes thought to control gene expression by altering subnuclear architecture and/or chromatin dynamics (43,56,(62)(63)(64)(65).
How AUF1 fits into the transcriptional regulatory network controlling the differentiation of progenitors is currently not clear. Previously identified transcriptional regulators of neuronal differentiation control complex developmental events, such as lineage decision, proliferation, and migration of progenitors (66). However, the identity of transcription factors that directly control the expression of the ENK gene in developing neurons is currently not known.
Studies have shown that another AT-rich dsDNA-binding protein, SATB1, exerts its regulatory role by regulating chromatin structure. We have recently identified another AT-rich dsDNA-binding protein in the developing rat brain, called SATB2 (50). Multiple studies have shown that in the lymphoid lineage, SATB1 and also SATB2 interact with chromatin remodelers, and it is involved in organizing the chromatin structure thereby controlling the accessibility of multiple transcription factors (28). These data implicate SATB1 and SATB2 as "global" coordinators of cellular differentiation in the hematopoietic and neuronal lineage and chromatin remodeling as a critical mechanism of coordinating the expression of a large numbers of genes. Future studies will need to address how AUF1 fits in the complex nuclear regulatory network that controls cell-specific gene expression in the developing brain.