The KRAB Zinc Finger Gene ZNF74 Encodes an RNA-binding Protein Tightly Associated with the Nuclear Matrix*

We previously cloned ZNF74 , a developmentally ex- pressed zinc finger gene commonly deleted in DiGeorge syndrome. Here, the intron/exon organization of the hu- man gene and the functional properties of the expressed protein are presented. This zinc finger gene from the transcription factor IIIA/Kruppel family contains three exons. A truncated Kruppel-associated box (KRAB) lo- cated at the N terminus of the predicted 64-kDa zinc finger protein is encoded by exon 2. The remainder of the protein including the zinc finger domain as well as the 3 (cid:42) -untranslated region (UTR) is encoded by exon 3. Both 5 (cid:42) -UTR (exon 1) and 3 (cid:42) -UTR contain repetitive Alu elements. In vitro translation of a cDNA encoding the entire ZNF74 coding region produced a 63-kDa protein as determined on sodium dodecyl sulfate-polyacrylam- ide gel. A bacterially expressed fusion protein shown to bind tightly to 65 zinc was used to test the nucleic acid binding properties of ZNF74. By RNA binding assays, ZNF74 was found to bind specifically to poly(U) and poly(G) RNA homopolymers. The restricted binding to these homopolymers and not to poly(A) and poly(C) suggested that ZNF74 displays RNA sequence preferences. RNA binding was mediated by the zinc finger domain. Immunofluorescence studies on transfected cells re- vealed ZNF74 nuclear localization. The labeling pattern observed in the nuclei clearly excluded the nucleoli. The zinc finger region lacks a classical nuclear localization signal but was found to be responsible immunofluorescence pattern scattered throughout the nucleus (excluding the nucleoli). Identical results were obtained in human epithelial-like HeLa cells transfected with HA-ZNF74-(aa 1–572); the background staining obtained with the anti-HA antibody in HeLa cells was, however, significantly higher than in COS-7 cells as clearly evidenced in untransfected control HeLa cells (not shown).

We previously reported the cloning of ZNF74, a zinc finger gene found deleted in most DGS patients tested (9). Furthermore, we determined that the coding region of ZNF74 gene lies a few kilobases proximal to a polymorphic marker (D22S264) (10) recently shown to be a distal marker for a large 22q11.2 region of deletion associated with increased susceptibility to schizophrenia (11). Since the initial isolation of ZNF74, six genes have been found to be part of the common DGS region of deletion (Ͼ1.5 megabases) (12)(13)(14)(15)(16), and one gene has been cloned at the t(2:22) translocation breakpoint of a patient presenting mild signs of DGS (17,18). It is not yet known if DGS is a contiguous gene syndrome involving several neighboring genes on the chromosome or a single gene. The involvement of the 22q11.2-deleted genes in DGS or other related disorders remains to be more directly addressed through genetic studies and functional characterization of their protein products.
ZNF74 is a member of the Cys 2 /His 2 (C 2 H 2 ) class of zinc finger genes, which constitutes the largest family of genes encoding potential transcription factors. The C 2 H 2 zinc finger motif corresponds to the consensus sequence: Tyr/Phe-X-Cys-X 2,4 -Cys-X 3 -Phe-X 5 -Leu-X 2 -His-X 3,4 -His-Thr-Gly-Glu-Lys-Pro, where X n,m denotes the presence of n or m amino acids between conserved residues; the cysteines and histidines coordinating the zinc are in bold and the conserved amino acids forming the link between consecutive fingers are underlined. The first described C 2 H 2 zinc finger-containing protein was an abundant oocyte protein from the frog Xenopus called TFIIIA (19,20). The zinc finger motif was subsequently identified in Drosophila genes such as Kruppel, and a search for homologues was undertaken in mammals (21)(22)(23). It is now estimated that between 300 and 700 human genes encode zinc finger proteins of the C 2 H 2 TFIIIA/Kruppel type (24).
Most studied C 2 H 2 zinc finger proteins were found to bind to specific DNA sequences and to be involved in the transcriptional regulation of gene expression (25)(26)(27)(28)(29). However, a small subset of multifinger proteins from Xenopus bind also RNA such as TFIIIA, while others bind only RNA such as p43, Xfin, and some members of the FAR subfamily of zinc finger proteins (e.g. XFG 5-1) (30 -36). Noticeably, TFIIIA is the only known characterized zinc finger protein with a dual role of DNA-and RNA-binding protein (30). It is a specific transcription factor that binds to and activates transcription of genes coding for the 5 S ribosomal RNA. By also binding to the 5 S RNA itself, TFIIIA participates in the transport of this RNA molecule from the nucleus to the cytoplasm and in its storage prior to ribosome assembly (30,37). While base-specific recognition by individual zinc fingers is the driving force in DNA binding, it is thought that it is more the RNA secondary/tertiary structure than its primary structure that determines RNA-protein interactions (30,38).
Zinc finger proteins of the C 2 H 2 family can be subdivided into at least four subfamilies based on the presence and identity of evolutionary conserved domains located at the amino terminus of the coding region. These different domains are known as: FAX (finger-associated Boxes) (39), FAR (fingerassociated repeats) (35), POZ ( pox virus and zinc fingers, also known as ZiN) (40,41), and KRAB (Kruppel-associated box) (42). About one-third of the hundred zinc finger genes of the mammalian genome are estimated by hybridization to contain a KRAB domain, but very few of their corresponding encoded proteins have been studied. The KRAB domain (about 75 amino acids) is a conserved motif found in all eukaryotes from yeast to humans. It has been recently demonstrated that the KRAB domain can confer distance-independent transcriptional repression when fused to the DNA binding domain of the yeast GAL4 transcription factor (43)(44)(45)(46)(47).
In this study, we report that ZNF74 is a human zinc finger gene of the KRAB subfamily whose protein product is a nuclear matrix-attached protein exhibiting RNA binding activity. Furthermore, we show that the zinc finger domain exerts at least a dual function, being responsible for the RNA binding properties and for the nuclear targeting of ZNF74 protein.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-The previously isolated 2.4-kb ZNF74 -1 cDNA cloned in pBluescript SK (Stratagene, La Jolla, CA) was used for preparing the various ZNF74 constructs (9). When convenient restriction sites were not available, selected fragments where amplified by PCR using Vent R DNA polymerase (New England Biolabs, Inc., Beverly, MA) and cloned in the appropriate vectors (48). To eliminate sequence errors potentially introduced by the polymerase, large regions of the PCRamplified inserts were substituted by cloning with corresponding restriction fragments from ZNF74 -1 cDNA. Furthermore, PCR-derived regions remaining unsubstituted as well as cloning junctions were sequenced in all constructs.
Sequencing and Exon/Intron Mapping-ZNF74 -1 cDNA as well as two newly isolated cDNAs ZNF74 -2a and ZNF74 -2b, which formed a composite 3.8-kb cDNA called ZNF74 -2 were obtained from a 23-week human fetal brain library as described before (9). Their sequences were determined by the dideoxynucleotide chain termination method using the Sequenase kit (U.S. Biochemistry Corp.) The use of a terminal deoxynucleotidyl transferase step to extend prematurely interrupted sequencing products (49) and/or the resolution of sequencing products on highly denaturing urea/formamide gels (50) were in some cases required to eliminate sequencing artefacts observed in some strong GC compression areas. To deduce the exon/intron structure of the ZNF74 gene, exon fragments or oligonucleotide primer sequences derived from ZNF74 cDNAs were positioned on previously isolated cosmid clones containing ZNF74 genomic DNA (9). For this purpose, the various cDNA probes and primers were hybridized to a Southern blot of ZNF74 cosmid clones digested with various restriction enzymes. To confirm the position of exons, subcloned EcoRI fragments derived from ZNF74 cosmid clones were sequenced along their exonic regions and at exon/ intron boundaries.
In Vitro Transcription/Translation-For in vitro transcription, the 3.8-kb ZNF74 -2 cDNA and the 2.4-kb ZNF74 -1 cDNA or its various truncated versions designed to eliminate from one to three of the first in frame methionines (ATG at positions 163-165, 259 -261, and 364 -366) were cloned into the HindIII or SmaI site of the pSP64 poly(A) transcription vector (Promega, Madison, WI). For such cloning, ZNF74 -2 cDNA or sequences derived from ZNF74 -1 cDNA and encoding the first three (nt 1-2144 or nt 125-2144), the second and third (nt 222-2144), the third (nt 285-2144), or none of the first three methionines (nt 367-2144) were obtained by digestion with restriction enzymes or by PCR amplification. [ 35 S]methionine-labeled proteins were synthesized from pSP64 poly(A) constructs by using a TNT coupled transcription/ translation rabbit reticulocyte lysate system (Promega, Madison, WI) as indicated by the manufacturer.
Bacterial Expression and Purification of Maltose-binding Protein (MBP) Fusion Proteins-To generate bacterially expressed full-length and truncated ZNF74 as fusion proteins with MBP, we used pMAL-c vector (New England Biolabs). PCR fragments were amplified from ZNF74 -1 cDNA and subcloned in frame in the XbaI site of pMAL-c using primers containing an XbaI site. The various MBP fusion proteins described below include ZNF74 amino acid domains identical to the ones illustrated for hemagglutinin (HA)-tagged proteins in Fig. 10. The various constructs obtained correspond to the full-length 572-amino acid coding region (nt 163-1881; MBP-ZNF74-(aa 1-572)), the zinc finger region (nt 685-1689; MBP-Zn-(aa 175-509)) and the C-terminal non-zinc finger region (nt 1696 -1881; MBP-C-(aa 512-572)). A 1.9-kb EcoRI restriction fragment (nt 478-2416) corresponding to the complete coding region less the first 105 amino acids was subcloned in the EcoRI site of pMAL-c vector (MBP-ZNF74-(aa 106 -572)). When no inserts are cloned in the polylinker site of pMAL-c, this vector generates an MBP-␤-gal-␣ fusion (51 kDa). Escherichia coli DH5␣ (Life Technologies, Inc.) transformed with pMAL-c fusion constructs were grown to an A 600 of 0.4 -0.6 and induced with 0.3 mM isopropyl-␤-D-thiogalactopyranoside for 3 h. Crude soluble extracts containing the soluble MBP fusion proteins were prepared from bacterial cells resuspended in buffer A (20 mM Tris, pH 7.5, 0.2 M NaCl, 1 mM EDTA, 10 M ZnCl 2 , and 10 mM ␤-mercaptoethanol) (51). When required, MBP fusion proteins from these extracts were immobilized on amylose resin (New England Biolabs) (51). Other ZNF74 hybrid proteins were initially made by fusion with glutathione S-transferase in pGEX vectors (Pharmacia, Uppsala, Sweden), but the resulting proteins were all recovered in inclusion bodies as insoluble proteins.
Zinc Blotting-Crude bacterial extracts containing 0.5-2 g of MBP fusion proteins were resolved on a 10% SDS-PAGE gel. One half of the gel was stained with Coomassie Brilliant Blue, and the other half was electrophoretically transferred to nitrocellulose. The washed membranes were probed for 30 -45 min at 25°C with 5-10 Ci of 65 ZnCl 2 (41 Ci/g; Du Pont Canada, Inc., Mississauga, Ontario) in 10 ml of 0.1 M Tris-HCl, pH 7.0, 50 mM NaCl, 5 mM CaCl 2 essentially as described by Makowski (52). In competition experiments, 0.01 mM ZnCl 2 was included during the probing and the subsequent washing steps. The washed and dried membranes were exposed at Ϫ70°C with intensifying screens for 2-10 h. At last, the membranes were stained with 0.1% Ponceau S in 1% acetic acid and destained with water to confirm the efficiency of the protein transfer.
Nucleic Acid Binding Assay-The MBP fusion proteins were affinitypurified on an amylose resin (New England Biolabs) as described by the manufacturer, and the proteins immobilized on the resin were used for RNA binding analysis. Control experiments were done using MBP-␤gal-␣ fusion or a fusion protein corresponding to MBP linked to the C-terminal non-zinc finger portion of ZNF74 (MBP-C-(aa512-aa572)). RNA homopolymers (Sigma) were 5Ј-end-labeled as described by Koster et al. (36). As determined by the manufacturer, the average molecular weights of RNA homopolymers vary from lot to lot. For the lots used, the average molecular weights were 102,000 for poly(U) and 384,000 for poly(G), and the molecular weights varied from 260,000 to 1,200,000 for poly(C) and from 100,000 to 1,700,000 for poly(A). The amylose resin (50 l/assay) with immobilized fusion protein (0.5-2 g) was incubated with 100 ng of radioactive homopolymer RNA ((2 ϫ 10 3 ) Ϫ (1 ϫ 10 4 ) cpm/ng) in a total volume 400 l of binding buffer (10 mM Tris, pH 7.5, 100 mM NaCl, 2 mM MgCl 2 , 50 M ZnCl 2 , 10 mM ␤-mercaptoethanol, 10% glycerol) in the presence or absence of competitor RNA or DNA (sonicated salmon sperm) for 30 min at room temperature. Bovine serum albumin (10 g/assay) was added to the binding buffer to reduce nonspecific binding. After RNA binding, the amylose beads were washed 3 times with 0.75 ml of binding buffer at 4°C. Radioactivity was quantified in a scintillation counter. RNA homopolymers binding to the MBP-␤-gal-␣or MBP-C-(aa512-572) controls was similar to the background binding observed on the amylose resin alone. In some experiments, the ionic strength was varied by increasing the NaCl concentration of the binding buffer from 50 mM to 1 M; these salt conditions did not affect the binding of the MBP fusion protein to the amylose resin (not shown). For all the experiments presented in this paper, the amount of labeled poly(U) and poly(G) used in the RNA binding assay (100 ng) was near the concentration required to have half of the sites saturated as determined in saturation experiments using from 10 ng to 1 g of labeled homopolymers. The binding of 100 ng of homopolymer RNA to increasing concentrations of fusion protein was quantified (up to 40 -60% of input RNA could be retained after washing), and the amount of MBP-ZNF74-(aa 106 -572) protein used in the described experiments was fixed to have always less than 10 -20% of the input labeled RNA bound. MBP fusion protein extracts can be stored at Ϫ20°C over a period of several weeks without loss of specific binding activities.
Cellular Transfections and Immunocytochemistry-Various coding fragments of ZNF74 -1 cDNA were cloned into the eukaryotic expression vector pCGN (53). This plasmid encodes an N-terminal HA epitope tag under the control of the cytomegalovirus promotor. The HA epitope can be recognized by the monoclonal antibody 12CA5 (54). Subconfluent African green monkey COS-7 or human HeLa cells plated in 1 ϫ 2-cm four-well Lab-Tek TM (Nunc, Naperville, IL) were transfected with the various plasmid constructs (2 g) (56). About 30 -36 h after transfection, intact cells or monolayers of cells at different stages of nuclear matrix preparation (see below) were fixed with 4% formaldehyde, permeabilized with 0.2% Tween-20 for 1 h, treated sequentially with 50 mM NH 4 Cl and with 0.2% cold fish gelatin for 30 min, and then processed for indirect immunofluorescence microscopy (57). Primary and secondary antibody incubation were performed essentially as described (57). Ascites fluid containing anti-HA monoclonal antibody 12CA5 was used at 1:500 dilution. The secondary fluorescent isothiocyanate rabbit anti-mouse IgG (Sigma) was used at a 1:100 dilution. The DNA was stained by a 5-min treatment with 2.5 g/ml Hoescht 33258 fluorochrome (Sigma). After rinsing, slides were mounted with 1 mg/ml p-phenylenediamine (Aldrich) in 90% glycerol. For the above immunocytochemistry steps, all solutions were prepared in phosphatebuffered saline, and manipulations were done at room temperature. Preparations were examined under a Leika photomicroscope equipped for epifluorescence and photographed using T-MAX 400 films. Each construct was analyzed in a minimum of three independent transfections. Immunofluorescence analysis showed that transfection efficiency for the cells varied from 5 to 20%.
Cellular Fractionation and Nuclear Matrix Isolation-For electrophoretic analysis, monkey COS-7 cells or mouse L cells were plated at a cell density of 4 -8 ϫ 10 5 /100-mm plate and transfected 24 h later with various plasmid constructs (40 g of DNA) (56). Nuclei were prepared essentially as described by Cockerill and Garrard (58) from cells harvested 48 h after transfection. In brief, cells were washed in phosphate-buffered saline and subjected to hypotonic lysis in RSB buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.5 mM phenylmethylsulfonyl fluoride). Cells were incubated on ice for 10 min, homogenized with a Dounce homogenizer, and centrifuged at 750 ϫ g for 10 min. The supernatant corresponding to the cytoplasm-containing fraction was designated fraction number 1 (or "C" in some experiments). The nuclei recovered in the pellet were washed twice in RSB buffer (the washed pellet was designated "N" in some experiments). The subsequent steps allowing subnuclear fractionation and nuclear matrix isolation were performed essentially as described by He et al. (59). The washed nuclei were freed of the chromatin by digestion with 20 units of RNase-free DNase-1 (Boehringer Mannheim, Laval, Canada) per 10 6 cells at 30°C for 50 min in digestion buffer (10 mM Pipes, pH 6.8, 50 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 1 mM EGTA, 4 mM vanadyl riboside complex, 0.5% (v/v) Triton X-100, 1.2 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 2 g/ml leupeptin, and 1 g/ml pepstatin A). The digested nuclei were then extracted by the addition of ammonium sulfate from a 1 M stock to a final concentration of 0.25 M. The 750 ϫ g supernatant containing the digested chromatin was designated fraction number 2. The pellet corresponds to the complete nuclear matrix containing ribonucleoprotein complexes (60). The nuclear matrix was fur-ther fractionated after resuspension of the pellet in digestion buffer and extraction by the addition of NaCl to a final concentration of 2 M from a 4 M stock in digestion buffer. The supernatant obtained after a 750 ϫ g centrifugation was designated fraction number 3. The pellet was resuspended in digestion buffer (without vanadyl riboside complex if for RNase treatment) and incubated for 1 h at room temperature with or without RNase A (Qiagen, Chatsworth, CA) at 100 g/ml and RNase T (Boehringer Mannheim) at 40 units/ml. The fractions were then centrifuged at 750 ϫ g. The supernatant and pellet untreated with RNases were designated fraction numbers 4 and 5, respectively. The supernatant and pellet treated with the RNases were designated fraction numbers 6 and 7, respectively. The fractions were subjected to Western blot analysis with either mouse 12CA5 anti-hemagglutinin monoclonal antibody (54) or mouse 131C3 anti-lamin A and C monoclonal antibody (kindly provided by Dr. Yves Raymond). All steps used for cell fractionation and nuclear matrix isolation were performed at 4°C unless otherwise specified.
For in situ fractionation, monolayers of cells were extracted sequentially as described in Bisotto et al. (61) using the method of He et al. (59). In brief, the cells were first extracted in cytoskeleton buffer (10 mM Pipes, pH 6.8, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl 2 , 1 mM EGTA, 0.5% Triton X-100, 4 mM vanadyl riboside complex, 1 mM phenylmethylsulfonyl fluoride), digested with DNase I and washed with 0.25 M ammonium sulfate for chromatin removal. Then the cells were washed in high salt conditions with 2 M NaCl. Finally, the extracted cells were treated with RNases A and T. Monolayers of extracted cells at different steps of the nuclear matrix preparation were fixed and treated for immunofluorescence as described above.

Identification of a KRAB Box in ZNF74
Gene-We have previously isolated a 2.4-kb cDNA (ZNF74 -1) corresponding to ZNF74 gene (9). As reported before, this cDNA is predicted to encode 12 Kruppel C 2 H 2 zinc finger motifs bordered by unique N-and C-terminal regions (9). A detailed comparison of various regions of the cDNA with DNA data bases allowed us to identify a sequence, highly homologous to the so-called KRAB box, lying 5Ј to the putative ATG initiation codon (nt 364 -366). A KRAB motif was then confirmed to be in frame at the Nterminal extremity of ZNF74 protein owing to the use of various modifications of traditional sequencing methods to track banding artefacts in extremely GC-rich regions of ZNF74 -1 cDNA (nucleotides 270 -290 and 305-355) (49, 50) (Fig. 1). The KRAB box is a conserved module only found in association with genes containing C 2 H 2 zinc finger motifs and is described as divided into two domains termed A and B boxes (Fig. 2). Indeed, KRAB A and KRAB B boxes are generally encoded by separate exons and in some cases subjected to alternative splicing (62,63).
To determine the exon/intron structure of ZNF74 gene and of its KRAB domain, sizes and sequences of ZNF74 cDNA restriction fragments (or PCR fragments) were compared with those of ZNF74 genomic DNA as described under "Experimental Procedures." To derive cDNA probes and PCR oligonucleotide primers, we used a 3.8-kb composite cDNA (ZNF74 -2), corresponding to two newly isolated and overlapping cDNAs (ZNF74 -2a and ZNF74 -2b), which includes the previously isolated ZNF74 -1 cDNA (Fig. 3). We found that the A and B KRAB boxes from ZNF74 were encoded by a 223-base pair exon (exon 2) (Fig. 1). The remainder of the coding sequence and 3Ј-untranslated region was encoded by a long 2299-base pair exon (exon 3) as often found for many C 2 H 2 zinc finger genes. The first exon corresponds to the long 5Ј-untranslated region.
In Vitro Translation of a 63-kDa Protein from ZNF74 cDNA-As deduced from the 3.8-kb composite ZNF74 -2 cDNA, the first three methionines found in frame with the longest open reading frame are in a context compatible with the Kozak consensus sequence for initiation of translation (64). Initiation at the first methionine should produce a 572-amino acid zinc finger protein with an N-terminal truncated KRAB A box as observed for a few KRAB zinc finger proteins (34) (Figs. 1 and  2). To confirm this longest open reading frame, ZNF74 cDNAs were cloned into a plasmid vector designed for in vitro transcription. As generally found with cDNAs presenting a long 5Ј-untranslated region, the 3.8-kb cDNA could not serve for efficient in vitro translation (not shown). By contrast, when the 2.4-kb ZNF74 -1 cDNA (which includes the whole predicted coding region) was used for in vitro transcription/translation in a reticulocyte lysate, a major protein product of approximately 63 kDa was obtained (Fig. 4, lane 1). The size of this translation product is in good agreement with the calculated molecular mass for the protein with the longest open reading frame (64 kDa, pI 8.84). Two minor and slightly shorter products of 61 and 59 kDa were also seen (Fig. 4, lane 1), suggesting that alternative usage of the second or third in frame ATG for initiation of translation could occur. Consistent with this, shorter translation products of 61 and 59 kDa (Fig. 4, lane 2) or 59 kDa only (Fig. 4, lane 3) were obtained with clones that were truncated for removal of one or two of the first in frame ATGs, respectively. No translation product was observed when the third methionine was removed (Fig. 4, lane 4).
RNA Binding Activity of the Bacterially Expressed ZNF74 -While most characterized zinc finger proteins bind to DNA, a few zinc finger proteins from Xenopus were previously reported to bind RNA (30 -36). In order to assess the potential RNAbinding activity of human ZNF74 and thus to approach its function, full-length and various MBP-ZNF74 fusion proteins were expressed in E. coli for large quantity production of soluble proteins and were used for an in vitro RNA binding assay. Soluble proteins from crude bacterial extracts were separated by SDS-PAGE and stained with Coomassie Blue (Fig. 5A). The

FIG. 2. Comparison of the KRAB region of various zinc finger proteins.
A period indicates that a gap has been introduced for optimal alignment. A consensus sequence is presented at the bottom. In that consensus, capital letters indicate at least 78% identity and boldface capital letters at least 94% identity among the various proteins. When two different amino acids are frequent and are found in more than 78% of the cases, the more common is indicated on the first line and the less frequent on the second line. Superscript numbers at the beginning of the sequences indicate the position of the first shown amino acid as compared with the N-terminal methionine at position 1.
MBP fusion proteins represented the major protein product of the bacterial extract except for the full-length MBP-ZNF74-(aa 1-572) protein, which was found to be unstable and subjected most of the time to extensive protein degradation (Fig. 5A, lane  2). We first confirmed that the bacterially expressed MBP fusion proteins encode zinc-binding proteins by probing equivalent Western blots with 65 ZnCl 2 . Autoradiography of the membrane indicated that each of the ZNF74 fusion proteins encoding the zinc finger region was able to bind 65 Zn(II) (Fig.  5B, lanes 2-4 and 7). As expected, the MBP-C-(aa 512-572) fusion protein containing the C-terminal region of ZNF74 as well as the MBP-␤-gal-␣ fusion did not bind any radiolabeled zinc (Fig. 5B, lanes 5 and 6). The specificity of the zinc binding was demonstrated by the absence of signal when competing nonradioactive ZnCl 2 was included in the assay (not shown). Since MBP-ZNF74-(aa 1-572) fusion protein encoding the fulllength ZNF74 was unstable in the bacterial extract, we decided to use the shorter MBP-ZNF74-(aa 106 -572) fusion protein (Fig. 5A, lane 3) for the RNA binding assays.
For RNA binding analysis, a modification of an in vitro nucleic acid binding assay described by Theunissen et al. (30) was used. MBP-ZNF74-(aa 106 -572) fusion protein immobilized on an amylose affinity resin was incubated with different 5Ј-radiolabeled RNA homopolymers in the presence of increasing amounts of various competitor RNAs. As illustrated in Table I and Fig. 6, ZNF74 fusion protein was able to bind specifically to poly(U) and poly(G) homopolymers but not to poly(A) and poly(C). The binding specificity was evaluated by competition with increasing amounts of various unlabeled RNAs and DNAs (Table II, Fig. 6). Poly(G) and poly(U) binding were only slightly competed by large excesses of tRNA and DNA. By contrast, poly(U) binding was efficiently competed by  the addition of increasing amounts of poly(U) and poly(G), whereas Poly(G) binding was competed by the addition of poly(G) and poly(U). By virtue of RNA duplex formation most probably, poly(U) binding was also significantly reduced by the addition of poly(A) in the same manner that poly(G) binding was reduced by poly(C). The poly(U) and poly(G) specific binding observed on MBP-ZNF74-(aa 106 -572) was equivalent to the binding exhibited by the same molar amount of MBP-Zn-(aa 175-509) fusion protein containing only the zinc finger domain of ZNF74 (Table I). This was suggesting that the zinc finger domain of ZNF74 was sufficient for RNA binding. The influence of the salt concentration on homopolymer binding was also analyzed by varying the concentration of NaCl from 0.05 to 0.8 M. Maximal binding of poly(U) and poly(G) occurred at 100 -200 mM sodium chloride in the range of physiological ionic strengths (Fig. 7). The homopolymer RNA binding properties of ZNF74 protein were comparable with the binding characteristics of the few previously described zinc finger RNAbinding proteins (34 -36).
Nuclear Localization of ZNF74 as Mediated by the Zinc Finger Domain-The subcellular localization of the few zinc finger proteins with a recognized RNA binding domain appears to vary from one protein to the other (32,34,35,37). For example Xenopus TFIIIA is shuttling between the cytoplasmic and nu-clear compartment as a ribonucleoprotein complex (37), while the Xenopus Xfin is only expressed in the cytoplasm (34). To determine the subcellular localization of ZNF74 protein, we performed indirect immunofluorescence studies (Figs. 8 and  10). Monkey fibroblast-like COS-7 cells were transfected with an expression plasmid encoding the full-length ZNF74 protein linked in frame with the influenza HA epitope tag (HA-ZNF74-(aa 1-572)). Comparison of the anti-HA immunofluorescence signal with the DNA-binding Hoescht coloration revealed that ZNF74-tagged protein was localized in the nucleus of transfected cells (Fig. 8, panels a and b). Furthermore, examination of the same field of cells by the phase contrast microscopy to clearly identify nuclei and nucleoli suggested that ZNF74tagged protein was excluded from the nucleoli (Fig. 8, panels b  and c). Most of the cells presented a punctuated granular   24 19 a Numbers correspond to percentage of MBP-ZNF74-(aa 106 -572) RNA binding activity for each homopolymer RNA in the absence or presence of 10 g competitor. In each assay, the binding of 100 ng of radiolabeled homopolymer was tested as described under "Experimental Procedures." These values were obtained from an average of four independent experiments. immunofluorescence pattern scattered throughout the nucleus (excluding the nucleoli). Identical results were obtained in human epithelial-like HeLa cells transfected with HA-ZNF74-(aa 1-572); the background staining obtained with the anti-HA antibody in HeLa cells was, however, significantly higher than in COS-7 cells as clearly evidenced in untransfected control HeLa cells (not shown).
Nuclear proteins are synthesized in the cytoplasm and transported in the nucleus across the nuclear pore complex (65,66). For most nuclear proteins, this transport is thought to be dependent on the existence of nuclear targeting signals present in the transported protein or protein complex. Some nuclear proteins contain a short basic targeting sequence rich in arginine and lysine resembling the nuclear localization signal of the SV-40 large T-antigen, while others encode a bipartite basic nuclear localization signal as first identified in nucleoplasmin (67,68). Since no classical nuclear localization signal could be detected in the ZNF74 protein sequence, we obtained a first delimitation of the region required for nuclear localization. For this purpose, we assessed the subcellular localization of truncated versions of the ZNF74-tagged protein when expressed in transfected COS-7 cells (Figs. 8 -10). As seen in Fig. 9, subcellular fractionation and immunoblot detection indicated that proteins expressed from all the constructs containing at least the zinc finger region were present in the nuclear fraction but not in the cytoplasm-containing fraction (the constructs used are illustrated in Fig. 10). Proteins of the expected size were obtained in each case as detailed in the legend to Fig. 9. The exclusive nuclear localization of these ZNF74 truncated proteins was confirmed by immunofluorescence staining as shown for the construct HA-Zn-(aa 175-509) expressing the zinc finger region alone (Fig. 8, panels d and e). This result allowed us to conclude that the zinc finger region of ZNF74 alone (aa 175-509) was sufficient for nuclear localization. By contrast, truncated proteins corresponding to either the KRAB-containing N-terminal region (HA-N-(aa 1-190)) or the C-terminal region (HA-C-(aa 512-572)) of ZNF74 were clearly seen in the cytoplasm by immunofluorescence staining (Fig. 8, panels f and  g and Fig. 10). Some staining was also observed in the nucleus with these constructs. While this nuclear staining was difficult to quantify by immunofluorescence considering the three-di-  1-572) (a, b, and c), HA-Zn-(aa 175-509) (d and e), and HA-N- (aa 1-190) (f and g). Hoescht stain (a and d), anti-HA immunofluorescence (b, e, and f), and phase contrast image (c and g) are shown. A representation of the same field of cells is shown for HA-ZNF74- (aa 1-572) (a, b, c) as well as for HA-Zn-(aa 175-509) (d and e) and HA-N-(aa 1-190) (f and g) transfected cells. Identical immunolocalization results were obtained with these HAtagged ZNF74 constructs and with ZNF74 constructs tagged with a FLAG epitope (not shown). mensional structure of the cell, subcellular fractionations and Western blot staining revealed that the N-terminal region ( Fig.  11C; 18 kDa) and the C-terminal region (not shown; 8 kDa) of ZNF74 per se were mostly associated with the cytoplasm-containing fraction.
Proteins expressed from constructs containing the zinc finger region were never detected in the cytoplasm of transfected cells, even in the ones exhibiting very low levels of expression. This argues against the possibility that the nuclear targeting results from overexpression in transfected cells. To ascertain that ZNF74 nuclear accumulation resulted from active transport and not from potential passive diffusion and retention, we assessed the localization of ZNF74 full-length protein and of ZNF74 zinc finger region alone (40 kDa) when fused to the large and well characterized reporter protein ␤-galactosidase (116 kDa) (66). While HA-␤-galactosidase-expressed protein was found mostly in the cytoplasm, as seen by indirect immunofluorescence, both HA-␤-galactosidase-ZNF74-(aa 1-572) and HA-␤-galactosidase-Zn-(aa 175-509) fusion proteins were completely localized to the nucleus (not shown). This suggested that active nuclear import of these two fusion proteins did occur, since their size was well beyond the reported diffusion limit across the nuclear pores.
Association of ZNF74 Protein with the Nuclear Matrix-In most of the cells transfected with HA-ZNF74-(aa 1-572), the immunodetected ZNF74 protein presented a nonhomogeneous punctuated pattern. This pattern was reminiscent of some immunovisualized discrete sites of the nucleus recently identified as functional subnuclear domains or domains of unknown function (69 -77). The underlying structure allowing such a nuclear organization has been proposed to be the nuclear matrix, an insoluble subnuclear fraction obtained after nuclease, detergent, and high salt treatment of the nuclei (78). Since some of the few proteins immunodetected at discrete sites of the nucleus have been shown to segregate with the nuclear matrix fraction during biochemical subcellular fractionation (61,70,79,80), we investigated the possible association of ZNF74 with the nuclear matrix. Following subcellular fractionation and Western blot analysis, ZNF74 protein ( Fig. 11A) was indeed found in the same fractions as specific markers of the nuclear matrix, lamins A and C (80) (Fig. 11B). As illustrated in Fig.  11A, ZNF74 protein was exclusively found in the nuclear matrix fraction before (lane 5) or after (lane 7) treatment with RNases. Similar results were obtained when monolayers of COS-7 cells transfected with HA-ZNF74-(aa 1-572) were submitted to in situ sequential fractionation (59,61). As seen in Fig. 12, ZNF74 protein remained associated with the nuclear matrix structure after extraction with detergents, DNase treatment and low salt wash (panels c and d), 2 M NaCl high salt wash (panels e and f), and RNase treatment (panels g and h). The results obtained by these in situ fractionations as well as by the above biochemical subcellular fractionations suggested that ZNF74 was tightly bound to the nuclear matrix and that its association was independent of the RNA component of the matrix that was accessible to RNase digestion.

DISCUSSION
ZNF74 gene was isolated from a human chromosome 22 library and found commonly deleted in patients with DGS developmental disorder (9). In order to address the function of that gene and its potential involvement in the DGS phenotype, we pursued its biochemical and cellular characterization. We report that ZNF74 gene encodes a KRAB zinc finger protein with RNA binding properties. ZNF74 protein is also tightly associated with the nuclear matrix, an important subnuclear structure organizing the nucleus and associated with nucleic acid metabolism. The zinc finger domain, which is composed of 12 C 2 H 2 motifs, is responsible for RNA binding activity and for nuclear targeting of this potential developmental regulator of nucleic acid metabolism.
By translation initiation at a first in frame methionine, a zinc finger protein containing an N-terminal truncated KRAB domain is obtained. Truncated KRAB domains have also been found in other KRAB zinc finger proteins as exemplified by Xfin (Fig. 2) (34). We cannot exclude the possibility that slightly shorter proteins devoid of the KRAB domain may also be produced in certain physiological conditions by initiation at the second or third methionine. These methionines are indeed in a good Kozak context for initiation. Two messenger RNAs of 3.6 and 4.4 kb were previously found to be expressed in various embryonic tissues, while no message was detected in corresponding adult tissues. We are presently trying to determine if these two messages could originate from the use of different Each construct was cloned in the PCGN vector, which encodes an N-terminal HA epitope, and used to transfect COS-7 or HeLa cells. The localization of the full-length and truncated ZNF74 proteins was determined by indirect immunofluorescence using an anti-HA antibody. ZNF74 protein and deletion mutants are represented by rectangular boxes from the N terminus to the C terminus; the various domains are represented and named. The first three in frame methionines (M) are numbered. The name of the construct is indicated on the left. For each construct, the first and last amino acids of ZNF74 included in the HA fusion proteins are indicated. Nuc, nuclear localization; Cyt, cytoplasmic localization.
FIG. 11. Nuclear matrix localization of ZNF74. The various protein fractions, as numbered and described under "Experimental Procedures," were isolated from COS-7 cells transfected with HA-ZNF74-(aa 1-572) (panels A and B) or HA-N-(aa 1-190) (panel C). For each fraction (lanes 1-7), proteins from an equal number of cells (3 ϫ 10 6 cells) were subjected to SDS-PAGE and Western blot analysis. Probing was realized with anti-HA antibody to detect ZNF74 protein (panel A) and truncated version (panel C) or with 131C3 anti-lamin antibody to detect the two nuclear matrix markers, lamins A and C (panel B). Nuclear matrix proteins recovered after a 2 M NaCl wash and treated without or with RNases are found in fractions 5 and 7, respectively. The size of the protein bands is indicated on the left in kilodaltons.
promotors, alternative splicing, or the alternative usage of different polyadenylation signals.
Searches through sequence data bases reveal that zinc finger genes from the large KRAB subfamily always encode multifinger proteins (with more than five finger motifs in tandem). This is contrast to the best characterized C 2 H 2 DNA binding factors such as SP1, the product of an early-immediate gene Zif268 (Egr-1), or the WT Wilm's tumor protein, which all contain fewer than five zinc finger motifs (24 -29). The interactions of the KRAB/C 2 H 2 multifinger proteins with nucleic acid remains to be studied in detail as they represent a class of protein recently discovered. A few Xenopus multifinger proteins containing more that five fingers have been previously described as RNA-binding proteins, and among them TFIIIA was found to be a DNA-binding protein as well (30). We now report that KRAB/C 2 H 2 ZNF74 protein is an RNA-binding protein. Like ZNF74 protein, Xenopus FAR/C 2 H 2 XFG 5-1 binds to poly(U) and poly(G) (36). Other Xenopus multifinger proteins, Xfin and TFIIIA, have a different specificity and bind more specifically to poly(G) (34 -36). To our knowledge, ZNF74 is the first mammalian C 2 H 2 zinc finger protein reported to have a specific RNA binding activity.
A significant portion of the posttranscriptional regulation of gene expression is mediated by RNA-binding proteins. Members of this growing family of proteins are required for multiple steps during mRNA metabolism, including pre-mRNA processing (e.g. splicing, capping, polyadenylation, and transport of pre-mRNA) and mRNA localization, translation, and stability. According to its nuclear localization demonstrated in transfected cells, ZNF74 protein may participate in nuclear RNA metabolism. We do not, however, exclude the possibility that ZNF74 may shuttle between the nucleus and the cytoplasm in certain physiological conditions. Since it is expressed in several embryonic tissues but is undetectable in the corresponding adult tissues (9), it is unlikely to be a constitutive factor but could represent a developmental regulatory factor. The biological function of the numerous and very recently isolated zinc finger genes of the KRAB subfamily is presently unknown. The finding that ZNF74 KRAB zinc finger gene is able to bind RNA supports the hypothesis that this type of multifinger protein may have an additional role as well as (or a different role from) the control of transcription through DNA binding as reported for most of the studied C 2 H 2 zinc finger proteins. To follow up on the biological function of ZNF74, it will now be essential to identify its in vivo RNA targets. Furthermore, studies are also under way to determine if ZNF74 zinc finger protein could function strictly in vivo as an RNA-binding protein such as Xfin (34) or exert a dual function like TFIIIA (30) by binding specifically to DNA and RNA sequences.
We have shown in this paper that the nuclear targeting of ZNF74 is mediated by its zinc finger domain. We suggest that the zinc finger domain of ZNF74 either contains an unusual nuclear localization signal or associates with a protein or ribonucleoprotein complex providing a nuclear targeting signal (81)(82)(83). The Xenopus FAR/C 2 H 2 XFG5-1 multifinger protein was previously reported to have an RNA binding domain and a nuclear transport domain encoded in the multifinger sequence (84). ZNF74 as a mammalian KRAB/C 2 H 2 protein further documents such a dual function of C 2 H 2 multifinger region. Following translocation, ZNF74 retention in the nucleus may be facilitated directly by its nucleic acid binding properties and/or indirectly as a ribonucleoprotein complex or a protein interacting with nuclear components (85,86). In agreement with this last possibility, we found that ZNF74 protein was a tightly associated component of the nuclear matrix, since detergent, DNase, and high salt treatment of the nucleus could not release it from the matrix. Binding to the nuclear matrix also persisted after RNase treatment, indicating that attachment to the matrix was probably not dependent on ZNF74 RNA binding activity. Identification and characterization of nuclear matrix-associated proteins are just beginning. Indeed, much attention has been given recently to this subnuclear structure and its associated proteins, since various studies have lead to the conclusion that the nuclear matrix is involved in chromatin organization, DNA replication, and RNA transcription, processing, and transport and that it is associated with cell type-specific proteins and markers of malignancy (reviewed in Ref. 78). It is increasingly recognized that the interphase nucleus is organized into subcompartments (some of which are of characterized functional significance) as recently pointed out by the immunodetection of punctuated staining with several nuclear antigens (69 -77). It will be interesting to determine if the observed ZNF74 punctuated staining could correspond to de- Other cells were sequentially extracted with 0.5% Triton X-100 in cytoskeletal buffer, digested with DNase I, and extracted with 0.25 M ammonium sulfate to remove chromatin (c and d), further extracted with 2 M NaCl to uncover the network of core filaments of the nuclear matrix (e and f) (59), and digested with RNases (g and h). All the immunofluorescence photographs were taken and printed at the same exposure to facilitate quantitative comparisons. The intensity of the anti-HA immunofluorescence was not altered significantly during the extraction procedures. Residual Hoescht staining was observed after the DNase I step, but no staining remained after the NaCl extraction. The punctuated nuclear staining is very clearly evident on the extracted cells shown (d, f, and h).
fined functional subnuclear domains such as those corresponding to sites of DNA replication, sites of premessenger RNA synthesis, sites of RNA processing, or domains of unknown function. Because of its tight attachment to the nuclear matrix, the possibility that ZNF74 associates with the core filaments of the nuclear matrix will need to be investigated. To date, nuclear matrix core filaments have been visualized by electron microscopy (59,78), but their actual protein composition is unknown. Future studies will also allow us to determine if other members of the large family of KRAB genes may encode constituents of the nuclear matrix.