This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Huggenvik, J. I. Articles by Mowen, K. A. Search for Related Content PubMed PubMed Citation Articles by Huggenvik, J. I. Articles by Mowen, K. A.

Characterization of a Nuclear Deformed Epidermal Autoregulatory Factor-1 (DEAF-1)-Related (NUDR) Transcriptional Regulator Protein

Department of Physiology Southern Illinois University School of Medicine Carbondale, Illinois 62901-6523

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
A monkey kidney cDNA that encodes a nuclear regulatory factor was identified by expression and affinity binding to a synthetic retinoic acid response element (RARE) and was used to isolate human placental and rat germ cell cDNAs by hybridization. The cDNAs encode a 59-kDa protein [nuclear DEAF-1-related (NUDR)] which shows sequence similarity to the Drosophila Deformed epidermal autoregulatory factor-1 (DEAF-1), a nonhomeodomain cofactor of embryonic Deformed gene expression. Similarities to other proteins indicate five functional domains in NUDR including an alanine-rich region prevalent in developmental transcription factors, a domain found in the promyelocytic leukemia-associated SP100 proteins, and a zinc finger homology domain associated with the AML1/MTG8 oncoprotein. Although NUDR mRNA displayed a wide tissue distribution in rats, elevated levels of protein were only observed in testicular germ cells, developing fetus, and transformed cell lines. Nuclear localization of NUDR was demonstrated by immunocytochemistry and by a green fluorescent protein-NUDR fusion protein. Site-directed mutagenesis of a nuclear localization signal resulted in cytoplasmic localization of the protein and eliminated NUDR-dependent transcriptional activation. Recombinant NUDR protein showed affinity for the RARE in mobility shifts; however it was efficiently displaced by retinoic acid receptor (RAR)/retinoid X receptor (RXR) complexes. In transient transfections, NUDR produced up to 26-fold inductions of a human proenkephalin promoter-reporter plasmid, with minimal effects on the promoters for prodynorphin or thymidine kinase. Placement of a RARE on the proenkephalin promoter increased NUDR-dependent activation to 41-fold, but this RARE-dependent increase was not transferable to a thymidine kinase promoter. Recombinant NUDR protein showed minimal binding affinity for proenkephalin promoter sequences, but was able to select DNA sequences from a random oligonucleotide library that had similar core-binding motifs (TTCG) as those recognized by DEAF-1. This motif is also present between the half-sites of several endogenous RAREs. The derived consensus- binding motif recognized by NUDR (TTCGGGNNTTTCCGG) was confirmed by mobility shift and deoxyribonuclease I (DNase I) protection assays; however, the consensus sequence was also unable to confer NUDR-dependent transcriptional activation to the thymidine kinase promoter. Our data suggests that NUDR may activate transcription independently of promoter binding, perhaps through protein-protein interaction with basal transcription factors, or by activation of secondary factors. The sequence and functional similarities between NUDR and DEAF-1 suggest that NUDR may also act as a cofactor to regulate the transcription of genes during fetal development or differentiation of testicular cells.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
The final differentiated state of organs, tissues, and cells that make up an adult animal is the end result of a developmental program of gene expression that is initiated at fertilization and elaborated during embryogenesis. Molecules such as retinoic acid (RA) must sometimes occur in concentration gradients to produce regional gene expression leading to appropriate morphological changes, such as those observed in limb bud development (1). Alternatively, regional expression of transcription factors may allow combinatorial interactions of the factors at target genes, which then result in relevant signaling cascades. For example, in male gonadogenesis, the Wilms’ tumor 1 (WT1) gene product appears to act as a dose-dependent cofactor with steroidogenic factor 1 (SF-1) to produce regulated expression of the Mullerian-inhibiting substance (MIS) gene (2). The synergistic action of WT1 and SF-1 can be antagonized by the X-linked, sex-reversal gene product, Dax-1, most likely through direct interaction with SF-1 (2).

Precise regional expression and interaction of specific transcription factors are also likely to be relevant to the mechanisms by which a master set of ‘selector’ genes termed HOM-C/Hox can determine the final morphological characteristics of an organism (reviewed in Ref. 3). The HOM-C gene cluster in Drosophila encodes a group of proteins that govern the anterior-to-posterior (A-P) segmentation patterning of developing fly embryos, and the homologous Hox gene clusters have been identified in all metazoans that have been examined (4, 5). In vertebrates, there are four Hox gene clusters, which show not only sequence homology to Drosophila HOM-C, but also display the linear chromosomal arrangement and 3'-to-5' temporal expression pattern observed during development. Mutations in these genes lead to the substitution of one body part for another (homeotic transformation). The HOM/Hox genes encode proteins that share a highly conserved 60-amino acid DNA-binding domain called the homeodomain (6) and are thought to exert their broad range of regulatory effects through DNA binding and transcriptional regulation of downstream target genes (7).

The high degree of amino acid similarity in the homeodomains of these proteins also appears to produce similar DNA-binding specificities in both in vitro and in vivo experiments (see reviews in Refs. 8, 9). The dilemma of achieving precise regulatory specificity with homeodomain proteins could be resolved if interactions with additional cofactors supplied the required specificity (3, 10). Indeed, a growing number of cofactors have been shown to provide functional enhancement to homeodomain proteins. For example, the human Oct-1 homeodomain protein will form high-affinity complexes with certain octamer motifs only in the presence of the coactivator VP16 (11, 12). The homeodomain protein extradenticle in Drosophila and its mammalian counterpart Pbx have also been shown to cooperatively bind DNA in the presence of some homeodomain proteins (13, 14, 15). More recent studies suggest that protein cofactors are required to switch some homeodomain proteins into a transcriptionally active state (3, 16).

Deformed epidermal autoregulatory factor-1 (DEAF-1) was identified as a nonhomeodomain protein that interacted as a cofactor with the Deformed protein from the HOM-C gene cluster in Drosophila. Deformed (Dfd) is a homeodomain protein that is expressed in the mandibular and maxillary segments of the embryonic head and is required for the subsequent development of structures derived from these segments (17, 18, 19). A 120-bp region of the Dfd promoter, referred to as module E, is capable of driving embryonic expression of a reporter gene in a pattern similar to endogenous Dfd expression (20). Module E was further dissected into a 24-bp Dfd-binding site and a 51-bp sequence (called region 5–6) that was required for appropriate Dfd expression (20). Thus, Dfd is capable of autoregulating its own expression but requires additional cofactors to provide segment-specific expression. DEAF-1 was identified as a protein cofactor that bound the 5–6 region in gel mobility shift and deoxyribonuclease I (DNase I) protection assays (21). Mutations in region 5–6 that improved DEAF-1 binding in vitro increased expression in transgenic embryos, indicating DEAF-1 or a similar protein is a required cofactor in Dfd expression. In addition, DEAF-1 was shown to bind multiple regions of the human HOXD4 promoter, a homolog of Dfd (21), suggesting a mammalian counterpart of DEAF-1 may exist to assist in the regulation of paralogous group 4 Hox genes.

Retinoic acid receptors (RARs) are nuclear transcription factors that, along with retinoid X-receptors (RXRs), bind to retinoic acid (RA) response elements (RAREs) of target genes to mediate developmental events such as RA-dependent regulation of embyrogenesis and cellular differentiation (22). We had previously shown that the catalytic subunit of cAMP-dependent protein kinase [protein kinase A (PKA)] could phosphorylate recombinant RAR in vitro, and that PKA potentiated RA signaling in transfected cells (23). Rochette-Egly and co-workers (24) went on to show that serine 369 of RAR was phosphorylated upon cotransfection of PKA or by forskolin treatment of F9 cells. Surprisingly, mutation of this serine to alanine or glutamate did not eliminate PKA potentiation of RAR signaling (Ref. 24 and our own unpublished studies have confirmed this result). To support a mechanism for the PKA-dependent regulation of RA signaling, we hypothesized that other PKA-regulated transcription factors were binding to RAREs or interacting with RAR/RXR dimers. In a search for binding proteins that might recognize RAREs, we have identified a protein with significant homology to DEAF-1 which we have designated as NUDR (for nuclear DEAF-1 related protein). We present evidence that NUDR is a nuclear protein that activates transcription from the human proenkephalin promoter, a gene that is expressed in many neuroendocrine and reproductive tissues. NUDR protein is expressed at elevated levels in testicular germ cells and developing fetus and will likely function to regulate gene expression in these tissues. The sequence and functional similarities between NUDR and DEAF-1 suggest that NUDR is a potential mammalian homolog of DEAF-1 and may therefore serve as a transcriptional cofactor of homeodomain proteins in rapidly dividing or differentiating tissues.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
Isolation of Human and Monkey cDNAs for NUDR
In an attempt to identify proteins other than RAR and RXR that may interact at the RARE, we used a ³²P-labeled oligonucleotide containing a RARE to screen a CV-1 monkey kidney cDNA library that was induced to produce ß-galactosidase fusion proteins. The RARE sequence AGGGTTCACCGAAAGTTCA (RARE half-sites are shown in bold and the pentameric nucleotide spacer is underlined) was based upon the DR5-RARE motif that occurs in the transcriptional promoter of the RARß gene (25). Of approximately 700,000 plaques that were screened, a single plaque remained positive through tertiary screening. Sequencing of the 1604-bp cDNA (sNUDR1.6) confirmed that an appropriate fusion-protein to the lacZ gene product was produced. The cDNA was then used to rescreen the CV-1 cDNA library by hybridization, to obtain a second clone of 2405 bp (sNUDR, accession no. AF049461), which contained 395 bases of 5'-sequence before the first methionine codon and a continuous open reading frame of 565 amino acids. We have designated the encoded protein as NUDR (see Discussion). Preliminary Northern blot analysis had indicated the presence of similar transcripts in the human choriocarcinoma cell line JEG-3. A cDNA library was produced from this cell line and screened by hybridization with the monkey cDNA clone. Two human clones of 2065 bp (hNUDR, accession no. AF049459) and 2329 bp (hNUDR8, accession no. AF049460) were chosen for further characterization. Human (h) NUDR contained 50 bases of 5'-sequence before the first methionine codon and an open reading frame that, similar to the monkey clone, encoded a protein of 565 amino acids. The monkey and human conceptual proteins showed 99% sequence similarity when compared with each other, with only five amino acid changes (Fig. 1). hNUDR8 contained 356 bases of 5'-untranslated sequence that was 95% identical to the monkey sequence. The hNUDR8 coding sequence contained a 42-bp deletion relative to hNUDR, which eliminated amino acids 16–29 and converted amino acid 15 from glutamate to aspartate. Presumably, this difference arises from differential splicing of a primary transcript, perhaps to alter the functionality of an alanine-rich motif that occurs in this region of the protein. Directly after the protein-coding region and before polyadenylation, the human and monkey cDNAs displayed 320 and 315 bases of 3'-untranslated sequence, respectively, and were 96% identical. The 3'-untranslated sequence of hNUDR contained a 105-bp region with 95% identity to a sequence-tagged site (dbSTS 40925), strongly suggesting that the NUDR gene maps to human chromosome 11.

View larger version (58K): [in this window] [in a new window]   Figure 1. Amino Acid Sequence Comparison of Human, Monkey, and Rat NUDR with Rat Suppressin The amino acid sequences for human, monkey, and rat NUDR were deduced from the cDNA sequences and compared with rat suppressin (accession no. U59659). The upper line shows the one-letter amino acid sequence of human NUDR (hNUDR, accession no. AF049459). Amino acid identity of monkey NUDR (sNUDR, accession no. AF049461), rat NUDR (rNUDR, accession no. AF055884), and rat suppressin with hNUDR are indicated by dots, and amino acid differences are shown as letters. The initiator methionine indicated for rat suppressin corresponds to methionine at position 69 of NUDR, and dashed lines show missing sequence. A second human NUDR clone (hNUDR8, accession no. AF049460) contains a 42-bp deletion (underlined) that resulted in the deletion of 14 amino acids (between amino acids 16 and 29) and a change of a glutamic acid to aspartic acid at position 15 (denoted by a D above the hNUDR sequence) as the deletion occurred within two codons. A potential bipartite NLS is boxed.

Comparison of Sequences with Similarities to NUDR
Comparison of the NUDR sequences to the nonredundant GenBank databases identified two cDNAs, designated as rat suppressin (1882 bp, accession no. U59659) and human suppressin (1888 bp, accession no. AF007165), with significant nucleotide and amino acid homologies. The rat suppressin clone was identified by screening of a pituitary cDNA library (26) with an antibody to bovine suppressin, a protein isolated and characterized from bovine pituitary (27). The suppressin protein has been shown to be synthesized and secreted by the GH₃ rat pituitary cell line and to inhibit proliferation of immune cells (27). Analysis of the rat suppressin and rat (r) NUDR sequences suggest that suppressin is a partial NUDR cDNA that would produce a protein lacking 68 amino acids (Fig. 1) if the suggested protein initiation site of suppressin were used. Comparison of the conceptual proteins showed they shared 94% amino acid identity while the nucleotide sequences were more than 99% identical. Comparison of human suppressin with hNUDR suggests it is also a partial NUDR cDNA that lacks the 5'-sequence coding for the first 48 amino acids and differs in sequence to produce four additional amino acid changes. As presented in the Discussion, we believe that it is unlikely the protein encoded by NUDR is the equivalent of the secreted protein characterized as suppressin.

The second most similar sequence identified in computer comparisons was the Drosophila DEAF-1. For this reason, we have designated the primate protein as nuclear DEAF-1 related (NUDR) protein. hNUDR shows 29% identity and 46% overall similarity with the 576 amino acids of DEAF-1. Other proteins identified in the database comparison show sequence similarity to specific regions of hNUDR and suggest the presence of five distinct functional domains (Fig. 2). The alanine-rich region near the amino terminus of NUDR is followed by an acidic-rich region, and together these regions (Alanine-Acidic, AA) produce matches primarily with proteins recognized as developmental transcription factors (Fig. 2A). Of the 17 proteins shown, 10 are homeodomain-containing proteins and another four are putative transcription factors closely associated with embryonic development. The three remaining proteins with AA similarity include the transcription factor JUN-D, human progesterone receptor (hPR), and a protein (MLL) that is associated with malignant transformation in t(11;19) leukemias (28).

View larger version (74K): [in this window] [in a new window]   Figure 2. Regional Homology of NUDR with Other Proteins The diagram at the top depicts approximate locations of potential functional domains in NUDR determined by sequence alignments to proteins shown in subsequent panels. The open triangle represents a naturally occurring deletion observed in the hNUDR8 cDNA; the solid triangle represents the NLS, and the arrow indicates the peptide region used for antibody production. Regions of similarity were identified by the BLAST program (National Center for Biotechnical Information, Bethesda, MD) and then aligned using the PILEUP program from the Wisconsin sequence analysis package by Genetics Computer Group, Inc. (GCG, Madison, WI). Shading represents mutationally conserved amino acids with scores above 0.7 in the default GCG scoring matrix. The alignments are: panel A, an alanine-acidic-rich (AA) region; panel B, an ND region; and panel C, a region with potential ZFH. Not shown are the alignments that indicated two conserved proline-rich regions (PR1, PR2). The following are the names and Entrez Protein Accession numbers used: hNUDR; DEAF-1, 1209883; nervy, 790600; cbf1 (core binding factor subunit 1), 735898; BF-2, 603460; shn, 1079151; HOXA13, 1832353; oct1 (octamer binding transcription factor 1), 2135847; POUIII (POU domain protein), 1730449; MLL, 2160396; HME1 (homeobox engrailed-1), 462291; Evx2, 106292; BarH1, 103026; BarH2, 419958; hPR (human progesterone receptor), 130894; Arx, 2317259; HB9, 1082461; JUN-d, 135307; HoxD8 (Hox4.3), 110005; LYSP100-B, 1173654; SP100-B, 1173656; pp41 (phosphoprotein 41), 1362889; RACK7, 1199659; AML1-MTG8, 407727; BS69, 1362759; HNF-4 (homolog), 1708276; t-BOP, 1809322; PDCD2, 998901; celeganF23 (c. elegans conceptual), 2088857; and yeast72KD (hypothetical), 2497149.

The carboxy terminus of hNUDR contains a zinc finger homology (ZFH) region with 56% similarity to the analogous region in DEAF-1 and also shows similarity to a functionally diverse set of proteins through the conserved spacing of cysteine and histidine residues (Fig. 2C). The spacing of these residues in the nervy gene product was previously suggested to resemble proteins that coordinate zinc through zinc-finger domains, while probably lacking the necessary arrangement for DNA binding (37). However, similarities of the NUDR ZFH domain with the DNA binding domains of hPR and HNF-4 [a member of the nuclear hormone receptor superfamily (38)], suggest a potential role of this region in DNA binding. The AML-1/MTG8 fusion protein arises from a t(8:21) translocation occurring in acute myeloid leukemias (AML), and its homology to the ZFH domain suggests a third potential link of hNUDR to oncogenic proteins produced in leukemias. Four other proteins with ZFH similarity have potential roles in cell signaling: RACK7 is a protein kinase C-binding protein (39); t-BOP is a zinc-finger protein expressed in T cells and muscle (40); BS69 is an inhibitor of adenovirus E1A transactivation (41); and PDCD2 is associated with the process of programmed cell death (42). The conservation of the ZFH pattern can also be seen in proteins from lower eukaryotes (celeganF23, yeast72kd in Fig. 2C), suggesting evolutionary conservation of a functional motif. In summary, the database similarities shown in Fig. 2 suggest that NUDR contains functional domains often found in nuclear transcription factors with developmental or oncogenic potential.

Tissue Distribution of NUDR mRNA
Northern blot analysis shows that the predominant NUDR mRNA form in CV-1 cells has a molecular size of 2.4 kb, indicating that the monkey cDNA that has been isolated is likely to be full length (Fig. 3A). Examination of various rat tissues for NUDR RNA expression showed the 2.4-kb mRNA in all tissues, with highest levels of expression in brain, adrenal, and lung (Fig. 3B). A second hybridizing band of RNA, with an estimated size of 6 kb, was observed in most tissues and represented the more abundant form in lung. Whether this longer RNA represents an alternative splice variant of a single NUDR gene or an RNA transcript from a highly related gene is not currently known.

View larger version (79K): [in this window] [in a new window]   Figure 3. Tissue Distribution of NUDR mRNA Expression in CV-1 Cells and Rat Tissues A, Ten micrograms of total RNA (lane 1) and poly A+ RNA (lane 2) from CV-1 cells were electrophoresed in a 1% denaturing agarose gel, transferred to a nylon membrane, and hybridized with radiolabeled sNUDR probe. B, Ten micrograms of poly A+ RNA from various rat tissues (lanes 3–13) were subjected to Northern blot analysis as in panel A except the blot was hybridized with a radiolabeled rat cDNA probe. The blots were exposed to film and the film was scanned with a densitometer. The mobility of NUDR is shown on the left and the RNA size standards (in kilobases) are shown on the right.

View larger version (49K): [in this window] [in a new window]   Figure 4. In Vitro Synthesis of NUDR Proteins 35S-Labeled proteins were produced by incubating 1 µg of the indicated plasmid DNA with 40 µCi of [35S]methionine in an in vitro transcription/translation system (TNT, Promega) and separated on 10% gel by SDS-PAGE. The gel was dried and imaged with a Molecular Dynamics’ PhosphorImager. Lane 1, Monkey NUDR cDNA in pBSSK (sNUDR); lane 2, human NUDR cDNA in pBSSK (hNUDR); lane 3, control plasmid (pBSKS); and lane 4, human RXR cDNA in pBSKS (hRXR). The mobility of NUDR and the prestained protein size markers (in kilodaltons) are shown on the left and right, respectively.

View larger version (41K): [in this window] [in a new window]   Figure 5. Western Blot Analysis of NUDR Proteins A, Four nanograms of recombinant sNUDR with an amino-terminal histidine tag were compared with 100 µg of total protein from untransfected monkey CV-1 cells, human JEG-3 cells, and the indicated adult rat tissues. Proteins were separated by SDS-PAGE on 10% gels, and immunoreactive proteins were detected with a polyclonal antibody to NUDR by Western blot analysis (see Materials and Methods). B, Total proteins (100 µg) from untransfected CV-1 cells were compared with proteins isolated from adult rat testis; cultured Sertoli, myoid, and Leydig cells; total germ cells (TGC); elongating spermatids (ES); round spermatids (RS); and spermatocytes (SPC). Western blot analysis was performed as in panel A. C, Total proteins (50 µg) from untransfected CV-1 cells were compared with total proteins (25 µg) isolated from 14-, 15-, and 17-day-old mouse embryos by Western blot analysis as in panel A using a polyclonal antibody to NUDR (left) or preimmune serum (right) on duplicate blots. The mobility of the prestained protein size markers (in kilodaltons) are shown on the right of each panel.

Since DEAF-1 had been characterized in the developing fly embryo, we sought to determine whether NUDR protein was synthesized during vertebrate development. Total proteins were prepared from 14-, 15-, and 17-day mouse fetuses and subjected to Western blot analysis (Fig. 5C). In each age of fetus tested, the 72-kDa NUDR protein was detected with the immune serum, as well as one or two proteins of higher molecular mass.

Although all tissues had shown the presence of the 2.4 kb NUDR mRNA in Northern blots (Fig. 3), detectable amounts of the 72-kDa NUDR protein were only observed in testis, fetus, and the cell lines. This may indicate that NUDR protein is unstable in most tissues and that significant levels of the protein may occur only in cells that are rapidly dividing or undergoing differentiation.

Cellular Localization of Endogenous and Transfected NUDR
To confirm that NUDR was a protein distinct from the protein characterized as the secreted protein suppressin, we tested whether NUDR was secreted into the media of CV-1 and HeLa cells that were transfected with an expression vector for hNUDR. Cells were incubated with ³⁵S-labeled methionine/cysteine, and the proteins in the culture media and cells were immunoprecipitated (Fig. 6). ³⁵S-labeled NUDR protein was observed in the cell extracts of CV-1 and HeLa cells but was not detected in the culture media. These results indicate that, under our experimental conditions, NUDR is not a secreted protein.

View larger version (84K): [in this window] [in a new window]   Figure 6. In Vivo Labeling of NUDR Proteins CV-1 cells and HeLa cells were transfected with hNUDR and incubated with 35S-labeled methionine/cysteine for 4 h before collecting the culture media and harvesting the cells. Media (lanes 4, 5, 8, and 9) and cell extracts (lanes 6, 7, 10, and 11) were treated with either preimmune serum (PI) or immune serum to full-length NUDR (I) followed by immunoprecipitation with protein A agarose beads. 35S-labeled sNUDR produced by in vitro transcription/translation was loaded directly in lane 1; or subjected to immunoprecipitation with preimmune serum (lane 2) or immune serum (lane 3). Immune complexes were separated on 9% gels by SDS-PAGE, and 35S-labeled proteins were detected with a PhosphorImager. The mobility of the prestained protein size markers (in kilodaltons) are shown on the right.

View larger version (65K): [in this window] [in a new window]   Figure 7. Localization of Endogenous NUDR and Overexpressed NUDR To detect endogenous NUDR protein, CV-1 cells were fixed on slides and incubated with preimmune serum (panel A, control) or serum containing antibodies to full-length NUDR (panel B, endogenous) followed by a biotinylated second antibody and fluorescein-conjugated avidin. CV-1 cells were transfected with pEGFP-N3 (panel C, GFP) and pEGFP-hNUDR (panel D, GFP-hNUDR) and fixed, and cells expressing GFP were visualized by fluorescence microscopy. CV-1 cells were transfected with pCMVhNUDR (panel E, hNUDR) and pCMVhNUDR-R302T/K304T (panel F, hNUDR-R302T/K304T) and incubated with an antibody to NUDR peptide followed by a fluorescein-labeled second antibody. Fluorescence was visualized with fluorescein filters using an Olympus IMT-2 microscope (panels A and B, 40x magnification) or Olympus Fluoview Confocal Imaging System attached to an Olympus IX70 microscope (panels C–F, 400x magnification).

View larger version (43K): [in this window] [in a new window]   Figure 8. Electrophoretic Mobility Shift Assay A, A 32P-labeled DR5 RARE oligo (120 fmol) was incubated with 35 pmol of recombinant hNUDR protein with increasing quantities (0–2000 ng) of nonspecific competitor DNA: either poly dI-dC, poly dA-dT, or salmon sperm DNA (ssDNA). B, Either 32P-labeled DR5 RARE or DR2 RARE oligos (240 fmol) were incubated with 38 pmol of recombinant hNUDR protein and/or 1 pmol of recombinant hRAR and in vitro translated hRXR. Protein-DNA complexes were separated from free probe on a 4% nondenaturing gel. Results were imaged with a PhosphorImager. In panel B, the arrows indicate hNUDR-shifted RARE complex and RAR/RXR-shifted RARE complex.

View larger version (28K): [in this window] [in a new window]   Figure 9. Selective Transcriptional Activation by NUDR CV-1 cells were cotransfected with 2 µg of reporter plasmid (pBLCAT6, pDynCAT3, pBLCAT5, pRARECAT5, pEnk77CAT6, or pRARECAT6) and either 1 µg of CMVNeo (control) or 1 µg of CMVhNUDR (hNUDR) as described in Materials and Methods. Reporter names and schematic representations of the constructs are shown to the left of the graph (not to scale). The fold induction due to cotransfection of NUDR is relative to each reporter, which is set to a value of 1. The results from this experiment are representative of three experiments and are expressed as the average ± SD of triplicate plates.

NUDR activation of the pRARECAT6 reporter was compared with the activation by the ligand-inducible retinoic acid receptor (hRAR) to ascertain the effectiveness of NUDR as a transcriptional activator at the RARE and to investigate potential interactions between the two proteins in vivo. CV-1 cells have low endogenous levels of RARs as evidenced by a 1.8-fold increase in reporter activity with retinoic acid (RA) treatment (Fig. 10). Cotransfection of cells with RAR showed a minimal increase in CAT activity in the absence of ligand (2-fold), which was increased to 7-fold in the presence of RA. In contrast, hNUDR increased transcription 41-fold in the absence of RA, which was further elevated to 50-fold with RA treatment. The RA-dependent increase is most likely due to low levels of endogenous RA-activated factors and not due to hNUDR binding of RA. These observations demonstrate the potency of NUDR as a transcriptional activator relative to RAR.

View larger version (38K): [in this window] [in a new window]   Figure 10. Transactivation of pRARECAT6 by RAR and NUDR CV-1 cells were transfected with 2 µg of the reporter plasmid pRARECAT6 and 1 µg of the indicated expression vector, CMVhRAR (hRAR) and/or CMVhNUDR (hNUDR). The cells were treated in the absence (-) or presence (+) of 1 µM retinoic acid (RA) for 24 h before harvest. Fold inductions are relative to the reporter alone. The results from this experiment are representative of three experiments and are expressed as the average fold induction ± SD of triplicate plates.

Nuclear Import of NUDR Is Essential for Transcriptional Activation from the Proenkephalin Promoter
We compared the transactivation potential of monkey NUDR (sNUDR), rat NUDR (rNUDR), and the human NUDR deletion variant (hNUDR8) to the previously tested hNUDR. As shown in Fig. 11, sNUDR, rNUDR, and hNUDR8 activated transcription from the proenkephalin promoter to a similar extent as hNUDR; however, in some assays the naturally occurring deletion of the alanine-rich region in hNUDR8 showed a trend for reduced transactivation by as much as 33% (not shown).

View larger version (45K): [in this window] [in a new window]   Figure 11. Nuclear Localization Is Critical for NUDR Activity CV-1 cells were transfected with 2 µg of the reporter plasmid, pEnk77CAT6, and 1 µg of a CMV-promoter driven expression plasmid [containing no insert (control) or the cDNAs for the following: human NUDR (hNUDR), hNUDR with a 14-amino acid deletion (hNUDR8), monkey NUDR (sNUDR), rat NUDR (rNUDR), and hNUDR with mutations (R302T, K304T, and R302T/K304T) in the NLS]. The results from this experiment are representative of two experiments and are expressed as the average fold induction over control (set at 1) ± SD of triplicate plates.

NUDR Activation of Proenkephalin Promoter Regions Appears to be Independent of DNA Binding
In an attempt to identify the sequences through which NUDR activated transcription of the pEnk77CAT6 reporter, enkephalin sequences in the 5'-untranslated region (UTR) and 3'-UTR were deleted, and the resulting reporter constructs were tested in CAT assays. Deletion of nucleotides 65–153 of the intron (position 71–157) resulted in a 60.9-fold increase in activation by NUDR (Fig. 12), but much of this fold change can be attributed to a significant decrease in basal activity (relative to pEnk77CAT6), as shown by the normalized CAT activity data. The deletion of additional 5'-UTR sequences (65–213) decreased NUDR activation (32.6-fold), indicating that sequences between 153 and 213 may contribute to activation by NUDR. Additional sequences in the 3'-UTR (1081 bp) may also contribute slightly to activation by NUDR as seen by the decreased fold change with the reporters, pEnk77CAT665–213,3'-UTR and pEnk77CAT63'-UTR. The normalized CAT activities are shown for each reporter to indicate that there are changes in both basal and NUDR-stimulated CAT activities. In all of these constructs, the majority of the NUDR-dependent activation appeared to map to a minimal promoter region (position -77 to +65), suggesting that NUDR may act in close proximity to the basal transcriptional machinery.

View larger version (25K): [in this window] [in a new window]   Figure 12. Analysis of Enkephalin Sequences Involved in NUDR Regulation Schematic representations and names of the reporters used in the transfections are shown to the left of the table. CV-1 cells were cotransfected with 2 µg of indicated reporter plasmid, 5 µg of SV2ßgal, and either 1 µg of CMVNeo (basal) or 1 µg of CMVhNUDR (hNUDR), as described in Materials and Methods. Normalized CAT activity is the ratio of CAT activity to ß-galactosidase activity, and the results are expressed as the average ± SD from triplicate plates. Fold change is the ratio of normalized CAT activity of +hNUDR to basal. The results are representative of two experiments.

Identification of NUDR-Binding Sequences
To identify DNA sequences to which NUDR might bind with high affinity, we used recombinant NUDR protein to select oligonucleotides from a library of double-stranded degenerate oligonucleotides and then amplified the selected sequences based on the method described by Lu et al. (46). Briefly, a set of oligonucleotides were synthesized that contained 30 random bases (1 x 10¹⁸ potential sequences), flanked on each end by different primer-specific sequences. Glutathione-S-transferase-sNUDR (GST-sNUDR1.5) fusion protein was immobilized on glutathione-agarose beads and used to select DNA sequences from the random set of oligonucleotides for which NUDR had affinity. Primers to each of the flanking sequences were then used to amplify the selected internal sequences by PCR, and the processes of affinity binding and amplification were repeated six times to select for DNA sequences that were consistently bound to NUDR protein. The DNA sequences were cloned into pBLCAT5 and sequenced to produce a collection of NUDR-binding sequences or NBSs. The alignment of 23 sequences from the approximately 100 oligonucleotides sequenced is shown in Fig. 13A. Analysis of the sequences showed 52% of the 23 NBSs contained one or more copies of the sequence TTCG, previously identified as the DEAF-1 core-binding sequence (21). Twenty-two percent and 61% of the NBSs contained one or more copies of the consensus motifs TTCGGG and TTTCCG, respectively.

View larger version (39K): [in this window] [in a new window]   Figure 13. Alignment of NBSs A, NBSs were selected from a library of double-stranded oligonucleotides (70 mers, 30 random nucleotides flanked on either side by 20 nucleotides with different primer-specific sequences) by affinity to recombinant NUDR. After six cycles of binding to NUDR and DNA amplification, the oligonucleotides were cloned and sequenced. A consensus was derived from the alignment of 23 of the sequences. B, Recombinant NUDR was used in EMSA to shift a subset of the oligonucleotides from the sixth round of selection. A consensus sequence was derived from the alignment of these eNBSs. The consensus derived from each alignment corresponds to a plurality of greater than 50%.

To confirm the DNA-binding activity of NUDR, EMSA and DNase I protection assays were performed on DNA fragments derived from the NBS consensus sequence (Fig. 14). In both assays, poly dI-dC was maintained at a level that had eliminated NUDR binding to the RARE and proenkephalin sequences (500 ng). Increasing levels of NUDR protein produced increased band intensity of the shifted consensus sequence in an EMSA (Fig. 14A). In DNase I protection assays, increasing levels of NUDR protein provided increased protection to three different radiolabeled DNA fragments (Fig. 14, B–D), verifying NUDR binding to the entire length of the consensus sequence. NUDR also extended its protection into the flanking vector sequence gatccgg, which resulted from the ligation of the consensus sequence into the BamHI site of pBLCAT5. Similar results were obtained for NUDR binding to several of the individual NBS sequences in EMSA and DNase I protection assays (data not shown).

View larger version (87K): [in this window] [in a new window]   Figure 14. EMSA and DNase I Protection Assays of the NBS Consensus A, Radiolabeled DNA containing the NBS consensus (sequence shown in panel E) was incubated alone (no protein) or with 10 pmol and 30 pmol of recombinant hNUDR. Samples were separated on a 4% nondenaturing polyacrylamide gel and analyzed by autoradiography. B–D, Radiolabeled double-stranded DNA containing one copy of the NBS consensus (B, bottom strand labeled and C, top strand labeled) or two copies of the NBS consensus (D, top strand labeled) was left untreated (U), or was treated with DNase I in the absence (O) or presence of increasing amounts of recombinant hNUDR (indicated by the wedge). Samples were separated on 6% denaturing sequencing gels and analyzed by autoradiography. The protected regions in panels A–C are indicated by hatched bars to the right of each panel, and the nucleotide sequences protected are shown by the corresponding hatched bars above or below the sequences shown in panels E and F. The NBS consensus sequences are shown in bold capital letters.

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

Depicted as a secreted protein and a novel inhibitor of cell proliferation, suppressin was initially purified from bovine pituitary and was identified as a monomeric polypeptide with an isoelectric point (pI) of 8.1 and a molecular mass of 63 kDa (27). A polyclonal antibody to bovine suppressin was used to screen a rat pituitary cDNA library, and a partial cDNA (691 bp) of rat suppressin was obtained. The 691-bp clone was subsequently used to isolate a 924-bp cDNA by rescreening the library by hybridization, and additional 5'-sequence was obtained by 5'-RACE (26). The 1882-bp cDNA called rat suppressin in the GenBank (U59659) is most likely a compilation of these partial sequences. Comparison of the conceptual protein encoded by the 1882-bp rat suppressin cDNA (26) and purified bovine suppressin (27) shows little similarity in amino acid composition and pI. Based on the 99% nucleotide homology to rat NUDR (Fig. 1), we suggest that the 1882-bp rat sequence represents a partial cDNA with a downstream methionine selected as the initiator methionine. Furthermore, the amino acids that follow the indicated initiator methionines in either suppressin or NUDR do not conform to the motifs required of signal peptides (47), making it highly unlikely that these proteins would be secreted. Using in vivo labeling, we were able to detect ³⁵S-labeled NUDR protein in CV-1 cell and HeLa cell extracts by immunoprecipitation, but were unable to detect any secreted proteins in the culture media (Fig. 6). As detailed in the experimental results, we have determined that the encoded NUDR proteins localize to the nuclei of cells and behave as transcription factors, making them unlikely candidates for secreted regulatory factors. And although the secreted protein characterized as suppressin has been shown to inhibit cell proliferation, it remains to be demonstrated that the protein encoded by the suppressin rat cDNA produces a secretory product that can inhibit cell proliferation.

Using antibodies to NUDR in Western blot analysis (Fig. 5), we observed three proteins in rat brain extracts that were lower in molecular mass than full-length NUDR but approximated the size of bovine pituitary suppressin (63 kDa). The antibody also detected several proteins in muscle and heart in the 35–45 kDa range that either share similar antigenic determinants with NUDR or are NUDR derivatives. Thus, it is conceivable that NUDR has antigenic determinants that could be recognized by anti-suppressin antibodies, potentially enabling the identification of a rat NUDR clone by antibody screening.

The second most similar sequence to NUDR identified by computer comparisons was the Drosophila DEAF-1. DEAF-1 has been shown to be an important cofactor in Deformed (Dfd) gene expression during embryonic development. A 120-bp region of the Dfd promoter, referred to as module E, is capable of driving embryonic expression of a reporter gene in a pattern similar to endogenous Dfd expression (20) and contains binding sites for DEAF-1 and Dfd, through which Dfd can autoregulate its own expression (21). hNUDR showed 46% similarity overall with DEAF-1 and contained regions of higher homology. DEAF-1 was initially purified from embryonic nuclear extracts by DNA affinity chromatography and migrated as a 85-kDa protein in SDS gels (21). DEAF-1, like NUDR, shows anomalous migration in protein gels, as the calculated molecular mass of DEAF-1 is 62 kDa and the protein produced by in vitro transcription/translation migrates like a 85-kDa protein (noted in Ref. 21).

An alanine-rich region in the amino terminus of NUDR may contribute to its transactivation as suggested by the decrease in CAT activity with hNUDR8, which has a naturally occurring deletion of the alanine-rich region (Fig. 11). A decrease in transactivation was also obtained using an amino-terminal deletion of sNUDR that lacked the first 75 amino acids (data not shown). The sequence similarity of this region to numerous homeodomain factors suggests that it may mediate an important transactivation function during development.

NUDR shows high regional similarity to SP100 proteins that colocalize with PML to subnuclear dot-like structures termed PML nuclear bodies. In APL, the normal pattern of PML localization to nuclear bodies is disrupted by a t(15;17) translocation, which produces a PML-RAR fusion protein. The oncoprotein shows aberrant localization and potentially contributes to APL pathogenesis (33). Treatment of APL-derived cells with RA restores the nuclear bodies (33), and RA can produce clinical remission of APL patients through differentiation of leukemic cells into mature granulocytes with associated decreases in cell proliferation (for reviews see Ref. 48). While a uniform nuclear presence of NUDR is observed in HeLa and CV-1 cells, the similarities to the SP100 proteins suggest interaction at PML nuclear bodies might potentially occur in appropriate cell types. Because NUDR also shows homology to leukemic oncogenes at its AA and ZFH domains, its precise subnuclear localization in normal and transformed lymphoid cells should be investigated.

The ZFH domain located at the carboxy terminus of both NUDR and DEAF-1 contains the sequence Cys-X₂-Cys-X₇-Cys-X₂-Cys-X₅-Cys-X₃-Cys-X₇-His-X₃-Cys(Cys₆HisCys). While cysteine residues frequently form disulfide bonds in extracellular proteins, cysteine and histidine residues are often used to bind metal ions, such as zinc, and stabilize structural folds of intracellular proteins (49). Although the exact spacings between cysteine residues and the histidine in NUDR and DEAF-1 are somewhat unusual, the motif is cognate of forming a zinc-binding domain (49) that may contribute to protein-protein interactions, DNA binding, and transcriptional activation. Gross and McGinnis (21) suggest this domain serves a non-DNA-binding role, since they were unable to obtain DNA binding in EMSA and DNase I protection assays with an amino-terminal truncated DEAF-1 that consisted of only the last 84 amino acids. In addition, they state that altering the second and third cysteines to serines in this domain did not affect the ability of full-length DEAF-1 to bind DNA in EMSA. Similarly, we have found that deletion of the last 60 amino acids of hNUDR resulted in only slight decreases in DNA binding by EMSA and transactivation in CAT assays (data not shown). Unlike many zinc finger domains that are critical for DNA binding, these data suggest that the ZFH region does not appear to be an independent DNA-binding module.

NUDR’s affinity for DNA was used to select specific DNA-binding sequences, and a direct repeat of TTC(G/C)GG was derived as a NUDR-binding motif. Interestingly, the sequence TTCGG is found between the RARE half-sites in hRARß2 (50), mRARß2 (51), and the RA-responsive mHoxa-1 (52) and may contribute to the higher levels of NUDR-induced CAT activity from pRARECAT6 over pEnk77CAT6. However, NUDR may also bind to some DNAs with low affinity and/or recognize additional sequences, since NUDR was able to bind the DR2 RARE (Fig. 8B), which lacks a TTCGG sequence, and because moderate levels of nonspecific competitor DNA readily displaced NUDR binding at the RAREs. NUDR binding of DNA is not synonymous with transactivation since constructs containing one and two copies of a RARE, NBS, or the NBS consensus were unable to confer NUDR-dependent activation to the thymidine kinase promoter in CAT assays. If NUDR DNA-binding activity exists to promote transactivation, then these studies indicate that the promoter context and/or the presence of multiple binding motifs of specific spacing may be important for NUDR transcriptional activation of target genes. However, as already noted, we have been unable to demonstrate high-affinity binding of NUDR to any proenkephalin sequences, which implies that NUDR may regulate proenkephalin transcription through mechanisms other than DNA binding, such as protein-protein interaction.

The TTCG motif has been identified as the core binding sequence of DEAF-1 and is found in region 5–6 of the Dfd promoter (21). Multiple copies of the motif and/or mutations that improved DEAF-1 binding in vitro generally increased the expression of the corresponding transgene in Drosophila embryos (21). Since the Dfd binding site and region 5–6 were required for segment- specific expression of Dfd in the embryo (20), DEAF-1 or a similar protein was postulated to be a cofactor of the homeodomain protein Dfd and required for Dfd expression (21). DEAF-1 is one of a growing number of nonhomeodomain proteins that have been identified as cofactors of homeodomain proteins. Two models have been proposed in which protein cofactors may interact with Hox and homeodomain-containing proteins to achieve greater target gene binding specificity. The coselective binding model envisions cofactors selectively targeting homeodomain proteins to different DNA sites, and the widespread binding model envisions cofactors altering the activity of homeodomain proteins that are already bound to the DNA (3). The close proximity of the binding sites for DEAF-1 and the homeodomain protein Dfd in the Dfd promoter would suggest a likely interaction among these proteins. However, Gross and McGinnis indicated that, at least in mobility shift assays, they had failed to detect cooperative interaction between DEAF-1 and Dfd and postulated that additional factor(s) may be required to form an activating transcription complex (21). The demonstration that DEAF-1 was also able to bind multiple regions in the promoter of human HOXD4 gene, a vertebrate homolog of Dfd (21), implies the existence of a mammalian counterpart of DEAF-1. The expression of NUDR during mouse fetal development and the strong sequence and functional similarities between NUDR and DEAF-1 would indicate that NUDR is a potential vertebrate homolog of DEAF-1 and, by analogy, may serve as a potential cofactor to homeodomain proteins and regulate the expression of the paralogous group 4 Hox genes and other downstream target genes. Future investigations will focus on the possible mediating role of NUDR in these developmental processes.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
Cloning of Monkey and Human NUDR cDNAs
A partial monkey NUDR cDNA was identified during expression screening of a CV-1 cell cDNA library (cDNA library 936111 in ZAP II vector, Stratagene, La Jolla, CA). Protein replica filters were prepared as described (53): 7 x 10⁵ pfu were plated (20 x 150-mm plates) on a lawn of E. coli strain XL-1 Blue MRF'. Plates were incubated 3.5 h at 37 C until pinpoint size plaques appeared and then overlayed with nitrocellulose filters that had been soaked in 10 mM isopropyl-ß-D-thiogalactopyranoside (IPTG). The filters were incubated at 37 C for 5 h to induce ß-galactosidase-fusion protein production before being placed in blocking solution [25 mM KCl, 25 mM HEPES (pH 7.3), 0.05% Triton X-100, 2% nonfat milk, 1 mM dithiothreitol (DTT)] at 4 C. The filters were then screened for proteins capable of binding to the DR5 RARE found in the human RARß promoter (25). The oligo 5'-GATCCAGGGTTCACCGAAAGTTCACTG-3' was hybridized with the complementary oligo 5'-GATCCAGTGAACTTTCGGTGAACCCTG-3' (the RARE-containing sequence is shown in bold, with flanking BamHI sites) and 50 pmol of this sequence were radiolabeled by a fill-in reaction with T4 DNA polymerase and ³²P--dATP (ICN Pharmaceuticals, Costa Mesa, CA), followed by phosphorylation and ligation of the DNA to form multimers. The radiolabeled probe was added to the filters in 75 ml of binding buffer (blocking buffer with nonfat milk reduced to 0.25%, and 10 µg/ml salmon sperm DNA) and incubated overnight with shaking at 4 C. The filters were washed over a 30-min period in 2 x 250 ml binding buffer with salmon sperm DNA and 2 x 500 ml binding buffer without salmon sperm DNA, and then exposed to x-ray film overnight. Initial screening with the RARE sequence showed two positive plaques, but only one continued to show binding through tertiary screening. After excision from the ZAP II vector and EcoRI digestion, the pBluescript plasmid was found to contain a 1.6-kb insert, which, upon further sequence analysis, was identified as a partial clone (sNUDR1.6).

A human choriocarcinoma cell line (JEG-3) cDNA library was constructed from cDNA synthesized from JEG-3 mRNA primed with oligo-dT using the SuperScript Choice System (Life Technologies, Gaithersburg, MD). The cDNA was ligated into Zap II and packaged with Gigapack III Gold packaging extracts (Stratagene). The JEG-3 cDNA library contained 3 x 10⁶ independent clones.

A 1.5-kb EcoRI/SmaI fragment of the monkey NUDR cDNA (sNUDR1.6) was radiolabeled by random priming and used to rescreen the CV-1 cDNA library (10⁶ clones) and the JEG-3 cDNA library (10⁶ clones) by hybridization. DNA from the plaques was lifted onto nitrocellulose filters and hybridized to the radiolabeled probe in 5 x SSPE (750 mM NaCl, 50 mM sodium phosphate, 5 mM EDTA), 0.5% SDS, 10 µg/ml salmon sperm DNA, 0.1% Ficoll, 0.1% polyvinylpyrrolidone, 0.1% BSA, and 50% formamide, overnight with shaking at 50 C. Filters were washed in 2 x SSPE for 45 min at 50 C, dried, and exposed to x-ray film to identify hybridizing clones. A monkey clone (2405 bp, sNUDR) and two human clones (2065 bp, hNUDR; 2328 bp, hNUDR8) were isolated and sequenced using Thermo Sequenase cycle sequencing kit (Amersham Corp., Arlington Heights, IL). Sequence comparison of the 1.6-kb and 2.4-kb monkey clones revealed a single nucleotide difference in the coding region that resulted in the substitution of an aspartic acid in the shorter clone relative to the asparagine (codon 287) in sNUDR.

A Sprague Dawley rat testicular germ cell library was constructed from cDNA synthesized in a similar manner to the JEG-3 cDNA library and contained 1 x 10⁶ independent clones. Oligonucleotide primers corresponding to position 1227–1245 and 1677–1660 of rat suppressin (accession no. U59659) were used to amplify a 434-bp DNA fragment from rat testis cDNA, which was cloned into pBSKS and sequenced to confirm the identity. The DNA fragment was radiolabeled and used to screen the germ cell library for full-length rat NUDR clones using conditions similar to those stated above.

Bacterial Expression Plasmids and Recombinant Protein Production
The cDNAs for sNUDR and hNUDR were subcloned into the pET-16b vector (Novagen, Inc. Madison, WI) for production of recombinant proteins in bacteria. The cDNA fragments containing sNUDR and hNUDR were excised from pBSSK by BspEI and EcoRI digestion, followed by T4 DNA polymerase fill-in, ligation of XhoI linkers, and digestion with XhoI. The 2.0-kb XhoI fragments were subcloned into the XhoI-digested pET-16b vector. DNA sequencing was used to confirm the correct insertion of the cDNAs in the vector. Resulting fusion proteins have an amino-terminal extension of 10 histidines followed by a factor Xa cleavage site.

hNUDR and sNUDR in pET-16b plasmids were introduced into E. coli strain BL21(DE3), and the expression of His-Tag-hNUDR and His-Tag-sNUDR proteins was induced by the addition of 1 mM IPTG during the last hour of bacterial growth. Bacterial pellets were sonicated in 5 ml of buffer A [6 M guanidine HCl, 0.1% IGEPAL CA-630 (Sigma, St. Louis, MO), 100 mM KCl, 20 mM Tris (pH 8.0)] and shaken for 1 h to solubilize proteins. Insoluble material was removed by centrifugation at 15,000 x g at 15 C for 20 min, and the supernatant was loaded onto a column containing 1 ml of His-Bind metal chelation resin (Novagen). The column was washed with buffers and recombinant His-Tag proteins were eluted from the column with buffer D [8 M urea, 100 mM KCl, 20 mM Tris (pH 6.8), 500 mM imidazole]. Proteins were renatured by five successive rounds of dialysis at 4 C in buffers that reduced the urea and increased glycerol to a final buffer of 15 mM Tris (pH 7.5), 50 mM KCl, 50% glycerol, 10 µM ZnCl₂, and 1 mM DTT.

A GST-sNUDR (1.5) fusion protein was produced in bacteria from a plasmid constructed by ligation of a 1.5-kb SmaI fragment of sNUDR1.6 into the SmaI site of pGEX-2T (Pharmacia Biotech, Piscataway, NJ). The expression of the GST-sNUDR1.5 fusion protein in E. coli strain CAG 748 (New England BioLabs, Beverly, MA) was induced by the addition of IPTG, and the recombinant protein was purified by affinity chromatography on glutathione-agarose.

hRAR was excised from the plasmid pGEMhRAR (kindly provided by Dr. R. Evans, Salk Institute) by MscI digestion, followed by ligation of BamHI linkers, digestion with BamHI, and ligation of the DNA fragment into the BamHI site of pGEX-2T. GST-hRAR fusion protein was purified by affinity chromatography and treated with thrombin to cleave the GST moiety from hRAR immediately before use in the EMSA.

Purified recombinant proteins and BSA protein standards were separated on SDS-PAGE, stained with Coomassie blue, and scanned with a Densitometer SI (Molecular Dynamics, Sunnyvale, CA) to determine protein concentrations.

In vitro transcription/translation was used to produce recombinant proteins using the TNT Coupled Reticulocyte Lysate System (Promega, Madison, WI) and 1 µg of the plasmids, sNUDR in pBSSK, hNUDR in pBSSK, and hRXR (the cDNA equivalent to position 76–1866 in Ref. 54) in pBSKS, according to the supplied instructions.

Mammalian Expression Plasmids
The cDNAs for NUDR were subcloned into pCMVNeo to obtain high levels of expression from the human cytomegalovirus immediate early gene promoter (CMV). The cDNAs for hNUDR, sNUDR, rNUDR, and hNUDR8 were excised from pBSSK by EcoRI digestion, followed by T4 DNA polymerase fill-in, ligation of BamHI linkers, and digestion with BamHI. The BamHI fragments were subcloned into the BglII site of eukaryotic expression vector pCMVNeo (55). Sequenc-ing pCMVhNUDR, pCMVsNUDR, pCMVrNUDR, and pCMVhNUDR8 confirmed that the orientation of the 5'-end of the cDNAs was adjacent to the CMV promoter.

To examine subcellular localization of hNUDR by a nonimmunological method, hNUDR was also subcloned into pEGFP-N3 vector (CLONTECH, Palo Alto, CA) to produce a fusion protein with GFP at the amino terminus. Expression is under the control of the CMV promoter. The cDNA fragment containing hNUDR was excised from pBSSK by BspEI and Bsu36I digestion, followed by Klenow fill-in, ligation of BamHI linkers (10 mers, New England BioLabs), and