Identification of a novel SCAN box-related protein that interacts with MZF1B. The leucine-rich SCAN box mediates hetero- and homoprotein associations.

The SCAN box or leucine-rich (LeR) domain is a conserved motif found within a subfamily of C(2)H(2) zinc finger proteins. The function of a SCAN box is unknown, but it is predicted to form alpha-helices that may be involved in protein-protein interactions. Myeloid zinc finger gene-1B (MZF1B) is an alternatively spliced human cDNA isoform of the zinc finger transcription factor, MZF1. MZF1 and MZF1B contain 13 C(2)H(2) zinc finger motifs, but only MZF1B contains an amino-terminal SCAN box. A bone marrow cDNA library was screened for proteins interacting with the MZF1B SCAN box domain and RAZ1 (SCAN-related protein associated with MZF1B) was identified. RAZ1 is a novel cDNA that encodes a SCAN-related domain and arginine-rich region but no zinc finger motifs. Co-immunoprecipitation assays demonstrate that the SCAN box domain of MZF1B is necessary for association with RAZ1. By yeast two-hybrid analysis, the carboxyl terminus of RAZ1 is sufficient for interaction with the MZF1B SCAN box. Furthermore, MZF1B and RAZ1 each self-associate in vitro via a SCAN box-dependent mechanism. These data provide evidence that the SCAN box is a protein interaction domain that mediates both hetero- and homoprotein associations.

The SCAN box or leucine-rich (LeR) domain is a conserved motif found within a subfamily of C 2 H 2 zinc finger proteins. The function of a SCAN box is unknown, but it is predicted to form ␣-helices that may be involved in protein-protein interactions. Myeloid zinc finger gene-1B (MZF1B) is an alternatively spliced human cDNA isoform of the zinc finger transcription factor, MZF1. MZF1 and MZF1B contain 13 C 2 H 2 zinc finger motifs, but only MZF1B contains an amino-terminal SCAN box. A bone marrow cDNA library was screened for proteins interacting with the MZF1B SCAN box domain and RAZ1 (SCAN-related protein associated with MZF1B) was identified. RAZ1 is a novel cDNA that encodes a SCAN-related domain and arginine-rich region but no zinc finger motifs. Co-immunoprecipitation assays demonstrate that the SCAN box domain of MZF1B is necessary for association with RAZ1. By yeast twohybrid analysis, the carboxyl terminus of RAZ1 is sufficient for interaction with the MZF1B SCAN box. Furthermore, MZF1B and RAZ1 each self-associate in vitro via a SCAN box-dependent mechanism. These data provide evidence that the SCAN box is a protein interaction domain that mediates both hetero-and homoprotein associations.
Zinc finger genes encode an abundant class of DNA-and RNA-binding proteins that represent an estimated 5% of the genes in the human genome. Many C 2 H 2 zinc finger genes have been demonstrated to function as transcriptional regulators and frequently, zinc finger genes are targeted for disruption in a variety of human diseases and cancers. The Krü ppel-like subclass of mammalian C 2 H 2 zinc finger proteins, first identified in the zinc finger transcription factor TFIIIA, share a conserved link between the last histidine of the preceding finger motif with the first cysteine of the next finger (H-C link) (1). Krü ppel-like proteins often contain conserved modular do-mains outside of their zinc finger motifs. These identified domains include the KRAB (Krü ppel-associated box) domains A and B, FAX (finger-associated box) domain of Xenopus, BTB/ POZ (broad complex, tramtrack, and bric-a-brac/poxvirus and zinc finger) or ZiN (zinc finger N-terminal) domain, and the SCAN box or leucine-rich domain.
To date, the functions of the KRAB and BTB/POZ domains have been best characterized. The KRAB domain is a conserved stretch of 75 amino acids found in an estimated one-third of Krü ppel-like zinc finger proteins (2). The KRAB domain, further subdivided into domains A and B, functions as a potent transcriptional repressor (3)(4)(5) and is predicted to fold into two amphipathic helices (2). The KRAB domain from KOX1 interacts with human TIF1␤ (also named KAP-1, KRAB-associated protein-1) (6, 7) and appears to exert its transcriptional repression activity through this interaction (6,7). In addition, the KRAB-A domain of Kid-1 interacts with KRIP-1 (KRAB-A interacting protein), which is likely to be the murine homologue of TIF1␤ and KAP-1 (8). The POZ domain defines a conserved region of approximately 120 amino acids and is found in 5-10% of zinc finger proteins. The POZ domain is a protein interaction motif (9) that mediates both homo- (10,11) and heterodimerization (12). Several POZ proteins are transcriptional repressors, including the oncoproteins PLZF (13) and BCL-6 (14), and the POZ domain has been shown to function as an autonomous transcriptional inhibitory domain (15). The POZ domain also has been demonstrated to interact with the co-repressors N-CoR, SMRT, Sin3, and histone deacetylase (16 -19), suggesting that POZ-containing proteins mediate transcriptional repression by recruiting histone deacetylase through a co-repressor complex. However, this may not be a general mechanism for POZ-containing transcription factors (20).
The leucine-rich (LeR) domain is a conserved motif present in the amino terminus of a subfamily of C 2 H 2 zinc finger proteins (4). This motif has also been designated the SCAN box, which was derived from the first four proteins found to contain this domain (SRE-ZBP, CT-fin-51, AW-1, number 18 cDNA) (21). To date, the SCAN box has been identified in approximately 20 zinc finger proteins from human, mouse, and rat including ZNF174 (21), RLZF-Y (22), FPM315 (23), ZNF213 (24), MZF1B 1 and its murine homologue, MZF-2 (25). SCAN box domains are about 80 residues in length, and approximately two-thirds of the amino acids are highly conserved with 80 -100% sequence identity. The function of the SCAN box is unknown. Based on protein sequence analysis, the SCAN box is predicted to form two or three amphipathic helices that may be involved in protein-protein interactions (4,21 (4,21), suggesting that the SCAN box is not an independent transcriptional regulatory domain.
To gain a better understanding of the SCAN box function, we directed our attention to the SCAN box-containing zinc finger protein, MZF1B. 1,2 MZF1B is an alternatively spliced isoform of the zinc finger transcription factor, MZF1 (26) and the human homologue of murine MZF-2 (25). MZF1 is a 485-amino acid protein that contains 13 C 2 H 2 zinc finger motifs arranged in a bipartite DNA binding domain. The consensus DNA binding sites have been identified (27), and MZF may regulate the expression of specific genes in a tissue-specific manner (28,29). MZF1 expression is both necessary for hematopoietic cell differentiation (30) and critical to the regulation of cell proliferation and apoptosis (31)(32)(33). MZF1B cDNA encodes a 734-amino acid protein that shares identity to the carboxyl terminus of MZF1, including the 13 C 2 H 2 zinc finger motifs. However, MZF1B encodes an additional 257 residues at its amino terminus, which contains a SCAN box domain. Therefore, the aminoterminal domains unique to each isoform may define the distinct functions of MZF1 and MZF1B proteins. We hypothesized that the MZF1B SCAN box is a protein interaction domain. A human bone marrow cDNA library was screened for proteins interacting with the MZF1B SCAN box domain, and RAZ1 (SCAN-related protein associated with MZF1B) was identified.
Full-length MZF1B cDNA (amino acids 1-734) and the unique amino terminus of MZF1B (amino acids 1-257) were amplified by polymerase chain reaction (PCR) 3 methods with oligonucleotides 1 and 2 or oligonucleotides 1 and 3, respectively, and subcloned into the EcoRI and HindIII sites of pcDNA3.1 (A) to produce plasmids MZF1B-mh/pcDNA and MZF1B N-term-mh/pcDNA, where "mh" indicates fusion of the protein to the carboxyl-terminal Myc-His epitope tag.
The unique amino terminus of MZF1B (amino acids 1-257) lacking the Myc/His tag was constructed by subcloning the same PCR fragment of MZF1B N-term-mh/pcDNA into pcDNA3.1 (C) to create MZF1B N-term*/pcDNA, which encodes an extra three amino acids (Lys, Ala, Thr) at the carboxyl-terminal end.
Full-length MZF1 cDNA (amino acids 1-485) was PCR-amplified with oligonucleotides 9 and 2. The PCR fragment, flanked by BglII and HindIII, was subcloned into the BamHI and HindIII sites of pcDNA3.1 (A) to create MZF1-mh/pcDNA. The cDNA isolated from yeast two-hybrid screening was removed from the pACT2 expression vector by digestion with BglII (including the hemagglutinin (HA) tag) and subcloned into the BamHI site of pcDNA3.1 (A) to generate ha-RAZ1/pcDNA. ␤-gal-mh/pcDNA, an expression plasmid encoding Myc-His epitopetagged ␤-galactosidase, was purchased from Invitrogen (Carlsbad, CA).
Yeast Two-hybrid Expression Plasmids-MZF1B and RAZ1 cDNA were subcloned into yeast expression cloning vectors, pAS2-1 and pACT2 (CLONTECH; Palo Alto, CA), to generate fusion proteins with the GAL4 DNA binding domain (BD; amino acids 1-147) or GAL4 To generate the bait plasmid, SCAN/pAS2-1, the cDNA encoding the MZF1B SCAN box (amino acids 47-123) was amplified by PCR using oligonucleotides 10 and 11 and subcloned into the EcoRI and HindIII sites of the mammalian expression vector CB6ϩ (34) to create SCAN/ CB6ϩ. SCAN/CB6ϩ was digested with EcoRI and BamHI and the MZF1B SCAN box insert was subcloned into pAS2-1.
To generate fusions of the MZF1B SCAN box with the GAL4 AD and HA tag, the cDNA encoding the MZF1B SCAN box (amino acids 47-123) was amplified by PCR using oligonucleotides 12 and 13 and subcloned into the EcoRI and XhoI sites of pACT2 to create SCAN/pACT2. RAZ1/pACT2 is the library clone isolated from the yeast two-hybrid screen as described below. This clone contains the RAZ1 cDNA (amino acids 1-217) library insert subcloned into the EcoRI and XhoI sites of pACT2 to generate a fusion protein containing the GAL4 AD and HA tag.
GAL4 Yeast Two-hybrid Analysis-The MATCHMAKER Two-Hy-brid System 2 and GAL4 Human Bone Marrow MATCHMAKER cDNA library were purchased from CLONTECH. A large scale, sequential transformation of the GAL4-SCAN bait fusion (SCAN/pAS2-1) and bone marrow cDNA library was carried out according to the manufacturer's directions. Briefly, the bone marrow cDNA library (50 g) was TABLE II Identification of RAZ1, a novel cDNA library clone that interacts with the MZF1B SCAN box a Number of yeast colony-forming units/g of plasmid DNA (cfu/g) growing on leucine and tryptophan-depleted medium: cfu ϫ total suspension vol (l)/vol plated (l) ϫ dilution factor ϫ amount DNA used (g) ϭ cfu/g of DNA.
Data Base Searching-Computer searches were done using the FASTA, BLAST, and MOTIFS algorithms through the Wisconsin Package software (64) or BLAST version 2.0 through the World Wide Web interface. Nucleotide sequences were compared with entries in the GenBank TM or expressed sequence tag (EST) data bases, while peptide sequences were searched against the Protein Information Resources (PIR) or Swiss-Prot data base.
Chromosome 20 Sequence-The Homo sapiens clone RP5-1121G12 from the RPCI5 library maps to chromosome 20q11.1-11.23 and has been assigned the EMBL/GenBank TM accession number AL109965. These data were produced by the Human Chromosome 20 Mapping and Sequencing Groups at the Sanger Center. Mapping and sequence data can be obtained on the World Wide Web.

RESULTS
The MZF1B SCAN Box Interacts with RAZ1-A human bone marrow cDNA library was screened by yeast two-hybrid analysis for potential MZF1B SCAN box interacting proteins, and one clone was identified (Table II). To confirm the positive protein interaction, we performed two-hybrid assays in the presence and absence of the MZF1B SCAN box domain. The library plasmid did not autonomously activate reporter gene expression, and a positive protein interaction was only ob- served when both the MZF1B SCAN box and library plasmid were co-transformed into yeast (Table II). To determine whether the interaction was an artifact of the fusion protein partner, we switched the AD and the DNA BD fusion partners for both the MZF1B SCAN box and interacting clone. The interaction between the MZF1B SCAN box and the isolated library clone was not dependent upon the fusion protein partner (Table II). These control experiments confirm that we identified a cDNA library insert positive for MZF1B SCAN box protein interaction. We have named this protein, RAZ1, a SCAN-related protein associated with MZF1B.
RAZ1 Is a Novel SCAN Box-Related Protein-The cDNA and amino acid sequences of RAZ1 are shown in Fig. 1A. The sequence is not found in the GenBank TM data base and appears to be a novel clone. The open reading frame for RAZ1 is defined by fusion with the upstream GAL4 activation domain and encodes 217 amino acids, starting with a glycine and ending with a stop codon at nucleotide position 652. The first methionine is at nucleotide 115 and contains a weak Kozak consensus sequence for translation initiation (37). Thus, it is probable that we have isolated a partial cDNA clone from the GAL4 fusion library that is incomplete at the 5Ј-end. The predicted sequence of RAZ1 encodes a SCAN-related domain at its car-boxyl terminus (amino acids 140 -200) but no zinc finger motifs. We refer to the domain as "SCAN-related" because the alignment with SCAN domains in other zinc finger proteins is conserved at the amino terminus and truncated at the carboxyl terminus (Table III). Approximately 20 SCAN box-containing proteins have been reported and/or deposited into the Gen-Bank TM data base that contain zinc finger motifs and/or KRAB domains (Table III). Thus, the SCAN box appears to be frequently associated with zinc finger motifs and sometimes with KRAB domains. It is possible that RAZ1 is a member of a novel gene family of non-zinc finger SCAN proteins. Immediately following the SCAN-related domain is an arginine-rich region (amino acids 201-217). The RAZ1 open reading frame also contains putative sites for post-translational modification: two casein kinase II phosphorylation sites at amino acid positions Ser 67 and Ser 77 , two protein kinase C phosphorylation sites at positions Thr 101 and Thr 144 , one cAMP-and cGMP-dependent protein kinase phosphorylation site at position Thr 211 , and two N-myristoylation sites at positions Gly 50 and Gly 62 (Fig. 1A).
RAZ1 Maps to Chromosome 20 -During the course of our studies, we identified a GenBank TM -deposited human chromosome 20 sequence at 20q11.1-11.23 with identity to RAZ1 at nucleotides 1-59 and 60 -775. This sequence also contains an additional 108-bp insert between nucleotides 59 and 60 of RAZ1. An illustration of the chromosome clone is shown in Fig.  1B. To further examine the RAZ1 gene, 5 ϫ 10 5 clones from a K562 cDNA library were screened with a RAZ1 cDNA probe. Nine unique clones were isolated, and all are identical to RAZ1, extending from nucleotide 21 and through the poly(A) tail (data not shown). In addition, 3Ј-RACE analysis identified products identical to the 3Ј-end of RAZ1 that correspond to the contiguous genomic DNA sequence on chromosome 20 including the stop codon, polyadenylation signal, and poly(A) tail. We obtained six 5Ј-RACE products identical to RAZ1 at the 5Ј-end, of which one extends from nucleotide 17. Interestingly, one 5Ј-RACE product contains the 108-bp insert between nucleotides 59 and 60 of RAZ1 (data not shown). These data provide independent confirmation that two RAZ1 transcripts may exist with divergence at the 5Ј-end, one that does not contain an additional 108 bp of sequence and one that does.
RAZ1 mRNA Is Expressed in Various Cell Lines-Northern blot analysis detects a RAZ1 transcript of ϳ1 kilobase in total RNA isolated from both human hematopoietic and nonhematopoietic cell lines. The highest levels of RNA expression were detected in the cell lines HEL (erythroleukemia) and Caco-2 (colon adenocarcinoma) (Fig. 2). The blot was reprobed with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) to verify equal loading of RNA (Fig. 2). Comparison of RAZ1 with the EST data base identified 100 ESTs between 300 and 600 bp in length that are identical to RAZ1. The ESTs were isolated from various human tissues including brain, breast, fetal heart, kidney, melanocyte, ovarian tumor, and placenta (data not shown). Six of these EST sequences contain the additional 108-bp insert between nucleotides 59 and 60 of RAZ1. Northern blots and reported ESTs suggest that RAZ1 may be widely expressed.
The MZF1B SCAN Box Domain Is Necessary for Interaction with RAZ1-To confirm that full-length MZF1B associates with RAZ1 and to identify MZF1B domains necessary and sufficient for interaction, MZF and RAZ1 proteins were coexpressed in vitro and co-immunoprecipitated with immunospecific antibodies.
As a first step, we demonstrated the immunospecificity of the nonspecific IgG, ␣-His, ␣-HA, ␣-ZF, and ␣-Myc antibodies by immunoprecipitating in vitro expressed MZF and RAZ1 proteins that contain either the amino-terminal HA or carboxylterminal mh epitope tag. The pcDNA expression plasmids used for immunoprecipitations are shown in Fig. 3. The antibodies were immunospecific, and no cross-reactivity was observed (Fig. 4A). Full-length MZF1B migrates as an 80-kDa protein, while the MZF1B NH 2 terminus migrates at approximately 42 kDa (Fig. 5). MZF1B ⌬SCAN and MZF1 migrate at approximately 72 and 50 kDa, respectively. RAZ1 migrates as a doublet of approximately 35 kDa. We consistently observe that the upper band of RAZ1 is more efficiently immunoprecipitated with ␣-HA. Therefore, the upper band of RAZ1 might be a result of translation initiation at the methionine upstream of the HA epitope tag, while the lower migrating band may be due to internal translation initiation at methionine 38 of RAZ1, thus producing a protein that lacks the HA epitope tag.
Protein association was demonstrated by co-expressing both MZF1B and RAZ1 epitope-tagged proteins in vitro and coimmunoprecipitating with the immunospecific antibodies. MZF1B is detected when the lysate is immunoprecipitated with ␣-HA, and RAZ1 is detected with ␣-His (Fig. 4B). This suggests that full-length MZF1B and RAZ1 are being pulled down in the same immunocomplex and are interacting in vitro. In addition, neither RAZ1 nor MZF1B ⌬SCAN proteins were detected in the same immunocomplex, suggesting that the MZF1B SCAN box is necessary for heteroassociation with RAZ1 (Fig. 4C). Furthermore, MZF1, which lacks the SCAN box domain, does not interact with RAZ1, verifying that the association with RAZ1 is unique to the SCAN box-containing amino-terminal region of The carboxyl terminus of RAZ1 is sufficient for MZF1B SCAN box interaction and RAZ1 self-association a Number of yeast colony-forming units/g of plasmid DNA (cfu/g) growing on leucine and tryptophan-depleted medium: cfu ϫ total suspension vol (l)/vol plated (l) ϫ dilution factor ϫ amount DNA used (g) ϭ cfu/g of DNA.
b The percentage of yeast positive for both HIS3 and lacZ reporter gene expression. MZF1B (Fig. 4D).
The Carboxyl Terminus of RAZ1 Is Sufficient for MZF1B SCAN Box Interaction and RAZ1 Self-association-To identify RAZ1 domains sufficient for MZF1B SCAN box association, we performed two-hybrid assays with the MZF1B SCAN bait plasmid and either the amino terminus or carboxyl terminus of RAZ1 that contains the SCAN-related domain. As a control, we demonstrated that the individual constructs did not autonomously activate reporter gene expression (Tables II and IV). Positive protein interactions occurred when the MZF1B SCAN box domain was co-transformed with full-length RAZ1 or the carboxyl terminus of RAZ1 but not with the amino terminus of RAZ1 (Tables II and IV). This demonstrates that the carboxyl terminus of RAZ1 is sufficient for MZF1B SCAN box interaction, suggesting that the SCAN box domains from both proteins are mediating heteroprotein association.
To test whether RAZ1 self-associates via a SCAN-dependent mechanism, we performed two-hybrid assays using RAZ1 fused to the GAL4 DNA BD and RAZ1 fused to the GAL4 AD. The individual plasmids did not autonomously activate reporter gene expression (Tables II and IV). Co-transformation of both RAZ1 fusion proteins resulted in colonies positive for protein interaction, suggesting that RAZ1 self-associates in vitro (Table IV). While amino acids 1-37 do not appear to be necessary for RAZ1 self-association, the carboxyl terminus of RAZ1 is necessary for self-association (Table IV). This demonstrates that the SCAN-related domain of RAZ1 mediates homo-as well as heteroprotein association. The MZF1B SCAN Box Is Necessary for Self-association-In demonstrating that the SCAN box mediates heteroassociation between MZF1B and RAZ1 proteins as well as RAZ1 homoassociation, we reasoned that the SCAN box might also mediate MZF1B homoassociation. To first determine if MZF1B could self-associate, we performed co-immunoprecipitation assays FIG. 5. The MZF1B SCAN box is necessary for MZF1B self-association. A, full-length MZF1B self-associates. Lanes 1-8, controls for antibody specificity; lanes 9 -12, co-expression and immunoprecipitation of tagged and nontagged MZF1B. The IVT reactions for A and B were scaled down to a final volume of 10 l, and 4.5 l of the lysate was used for immunoprecipitations in a final volume of 250 l. B, the MZF1B NH 2 terminus is sufficient for self-association. Lanes 1-6, controls for antibody specificity; lanes 7-9, co-expression and immunoprecipitation of tagged and nontagged MZF1B NH 2 terminus. C, the MZF1B SCAN box is necessary for self-association. Lanes 1-4, controls for antibody specificity; lanes 5-8, co-expression and immunoprecipitation of nontagged MZF1B NH 2 terminus and tagged MZF1B NH 2 terminus ⌬SCAN. The dotted arrows indicate higher order protein complexes. with epitope-tagged MZF1B and nontagged MZF1B. The Myc/ His epitope tag adds ϳ2 kDa, and the two proteins are distinguishable by size as well as immunoreactivity with the epitope tag-specific antibodies, ␣-Myc or ␣-His. The MZF1B proteins were co-expressed in vitro and co-immunoprecipitated with control IgG, ␣-post-SCAN, or ␣-Myc. In lysates expressing both forms of MZF1B, we detected nontagged MZF1B when epitopetagged MZF1B was immunoprecipitated with ␣-Myc, suggesting that MZF1B self-associates in vitro (Fig. 5A). To test for the possibility of nonspecific binding, we repeated the assays by co-expressing both MZF1B and epitope-tagged ␤-galactosidase. MZF1B did not associate with ␤-galactosidase, supporting our observation that MZF1B self-association is not an artifact of our co-immunoprecipitation conditions 4 (data not shown). In addition, the amino terminus of MZF1B is sufficient for selfassociation, since both epitope-tagged and nontagged MZF1B NH 2 terminus proteins were detected in the same immunocomplex (Fig. 5B). Finally, the MZF1B SCAN box is necessary for MZF1B self-association because nontagged MZF1B NH 2 terminus did not co-immunoprecipitate with MZF1B NH 2 terminus ⌬SCAN (Fig. 5C). It should be noted that we consistently observe higher molecular mass bands of Ͼ200 and ϳ80 kDa in immunoprecipitated lysates expressing MZF1B and MZF1B NH 2 terminus, respectively (Fig. 5, A and B). While the identification of these bands has not been confirmed, they may represent higher order complexes of the 80-kDa MZF1B and 42-kDa MZF1B amino terminus. DISCUSSION RAZ1 Protein Structural Motifs-We have described the identification of RAZ1, a novel human cDNA clone isolated from a yeast two-hybrid screen based on interaction with the MZF1B SCAN box. The function of RAZ1 in unknown, but the predicted sequence contains conserved motifs that provide insight into RAZ1's potential role in regulating transcription factor function. RAZ1 cDNA contains an open reading frame of 217 amino acids with a carboxyl-terminal region homologous to the SCAN box domain conserved in zinc finger proteins. Interestingly, the RAZ1 SCAN box is truncated and lacks the predicted third ␣-helix present in other SCAN box proteins. Thus, we have designated this as a SCAN-related domain. In contrast to other SCAN box proteins, RAZ1 does not appear to encode zinc finger motifs based on the sequence that we have obtained.
The sequences for approximately 20 SCAN box-containing proteins have been reported and/or deposited into the Gen-Bank TM data base (Table III). Of these, seven also encode a KRAB A and/or B domain, and 19 contain carboxyl-terminal zinc finger motifs. This suggests that the SCAN box is frequently associated with zinc finger motifs and sometimes with KRAB domains. The remaining SCAN box proteins that do not contain zinc finger motifs include p18, TRFA, PGC-2, and RAZ1. p18 and TRFA are partial clones that do not contain sufficient sequence to determine the presence or absence of zinc finger motifs. PGC-2 (peroxisome proliferator-activated receptor ␥ coactivator-2) is a murine adipogenic cofactor bound by the differentiation domain of the peroxisome proliferator-activated receptor ␥ (38). PGC-2 encodes a 142-amino acid protein with a carboxyl-terminal SCAN-related domain but no zinc finger motifs. The PGC-2 protein shares 76% identity to RAZ1 (49% at the NH 2 terminus; 97% at the COOH terminus), suggesting that PGC-2 may be the murine homologue of RAZ1.
It is possible that RAZ1, PGC-2, and potentially p18 and TRFA represent a novel gene family that contain SCAN box domains but lack zinc fingers. Similarly, the SSX gene family contains KRAB domains without zinc fingers. In addition to the KRAB domain, the SSX gene family encodes a novel transcription repression domain at the carboxyl terminus, SSXRD. Interestingly, this SSXRD domain exerts stronger repression than the KOX1 KRAB domain, and the KRAB-related domain fails to interact with the co-repressor TIF1␤ (KAP-1) (50 -52). Therefore, the protein binding and repression function of the SSX genes that contain KRAB domains and lack zinc fingers appears to be different and distinct from KRAB proteins that contain zinc fingers. Thus, SCAN proteins that lack zinc fingers may contain other conserved domains that modify or define their function. RAZ1 contains an arginine-rich region of 16 amino acids at the carboxyl terminus, immediately following the SCAN-related domain. Short (10 -20-amino acid) arginine-rich sequences have been shown to mediate DNA and RNA binding as well as nuclear localization. The arginine-rich domain found in the amino terminus of the recombination activating gene, RAG-1, exhibits DNA binding activity (53). In addition, the arginine-rich motifs of HIV Rev and Tat proteins, bacteriophage N, 21 N, and P22 N mediate RNA binding (reviewed in Ref. 54). Specifically, the human immunodeficiency virus (HIV) Rev binds to the Rev response element of HIV-1 as an ␣-helix and facilitates the nuclear export of unspliced HIV pre-mRNAs (reviewed in Ref. 54). There is an increasing amount of evidence that the arginine-rich domains present in HIV Rev, Tat, and human retroviruses T-cell leukemia virus type 1 also function as direct importin ␤-dependent nuclear localization signals (55,56). Thus, the arginine-rich domain of RAZ1 may mediate DNA-RNA binding and/or function as a nuclear localization signal. Localization to the nucleus would place RAZ1 in the same cellular environment as zinc finger transcription factors, and nucleotide binding activity may allow RAZ1 to compete with other zinc finger SCAN proteins for DNA-RNA binding sites. In addition, several putative phosphorylation and N-myristoylation sites reside within RAZ1, suggesting that the function of RAZ1 may be regulated by posttranslational modifications.
RAZ1 Gene Structure and mRNA Expression-The isolated RAZ1 cDNA clone is 775 bp in length and contains a putative translation initiation start site, stop codon, and polyadenylation signal (5Ј-AAU GAA AAA-3Ј). Several ESTs, cDNA library clones, and RACE products share identity to RAZ1, suggesting that we have identified a bona fide transcript. In addition, some of the EST and 5Ј-RACE sequences contain an additional 108-bp insert between nucleotides 59 and 60 of RAZ1. It remains to be determined if both transcripts are expressed in vivo. Northern blot analysis detects an ϳ1-kilobase RAZ1 transcript in both hematopoietic and nonhematopoietic cells, and ESTs from various tissues share identity to RAZ1, suggesting that the RAZ1 gene is expressed in a variety of tissues. Based on RACE, cDNA library clones, ESTs, and chromosome 20 sequence, it is likely that we have obtained the complete 3Ј-end of the RAZ1 transcript and are within a few hundred nucleotides of obtaining the entire 5Ј-end. We scanned the chromosome 20 sequence and found that the open reading frame upstream of RAZ1 continues for 128 amino acids and contains a methionine with a weak Kozak consensus (Fig. 1B). However, further analysis is needed to confirm the complete 5Ј-end of the transcript.
Interestingly, the RAZ1 gene is localized to chromosome 20q11.1-11.23. Deletion of the long arm of chromosome 20, most often 20q11.2-13, is associated with myeloid disorders, particularly myeloproliferative disorders, myelodysplastic syn-drome, acute lymphocytic leukemia, and acute myelogenous leukemia (57)(58)(59). This suggests that the genetic loss on chromosome 20q may provide a proliferative advantage to myeloid cells, possibly through the loss of a tumor suppressor gene. In addition, an increased copy number of DNA sequences from chromosome 20q has been observed in pancreatic cancers (60) and breast carcinomas (61), and trisomy of chromosome 20 is associated with the progression of papillary renal cell carcinomas (62). This indicates that a gain of chromosome 20q may facilitate uncontrolled cellular proliferation, possibly through the aberrant expression of an oncogene.
The RAZ1 chromosome 20q11.1-11.23 location raises the question as to whether loss or gain of RAZ1 contributes to any disorders associated with the locus. The myeloid proliferative disorders associated with loss of chromosome 20q are particularly interesting because MZF1/1B appears to be an important regulator of hematopoietic differentiation and proliferation (30 -33). Therefore, RAZ1 or MZF1B may be a tumor suppressor gene, and the interaction between RAZ1 and MZF1B may be necessary to elicit a tumor suppressor function. Thus, a genetic loss of RAZ1 might block tumor suppressor activity, thereby providing a proliferative advantage to hematopoietic cells.
The SCAN Box Function-Co-immunoprecipitation and yeast two-hybrid analyses demonstrate that MZF1B and RAZ1 associate in vitro via a SCAN box-dependent mechanism. In addition, the SCAN box domains are necessary for MZF1B and RAZ1 self-association. Therefore, we have demonstrated that the SCAN box is a protein interaction domain that mediates both hetero-and homoprotein associations. These findings suggest a novel cascade mediated by unique SCAN box protein complexes. To define cellular cascades regulated by SCAN box protein interactions, it will be necessary to identify in vivo SCAN box oligomers and the unique functions elicited by each complex. The identification of mechanisms regulated by SCAN box protein complexes will significantly impact our understanding of the transcriptional role of SCAN box zinc finger proteins and their associated factors.
The transcriptional activity of the SCAN box-containing zinc finger proteins MZF-2 and ZNF174 has been examined. Fulllength murine MZF-2 does not activate reporter gene expression, but a truncated form of MZF-2 markedly enhances transcription (29). The transcriptionally active form of MZF-2 contains the SCAN box domain, but the SCAN box is not necessary for transcriptional activity. ZNF174 is a transcriptional repressor of reporter genes driven by the human tumor growth factor-␤ and platelet-derived growth factor-B promoters (21). The complete amino terminus of ZNF174, including the SCAN box domain, transcriptionally represses reporter gene expression when fused to a heterologous DNA binding domain, but the SCAN box domain is not sufficient for transcriptional repression. The ZNF174 repression domain is probably present within the remaining amino-terminal portion, and the SCAN box may or may not modify this function. Thus, the SCAN box does not appear to function as a transactivation or repression domain. These conclusions are supported by our personal observations 5 and reports by Pengue et al. (4) and Williams et al. (21), which show that the SCAN box does not confer transactivation or repression function onto a heterologous DNA binding domain. While the SCAN box is not an independent transactivation or repression domain, the SCAN box may function to recruit co-repressors and transactivators necessary for transcriptional regulation.
MZF1B Function-We identified RAZ1 as a potential in vivo protein interaction partner with the SCAN box-containing zinc finger protein MZF1B, the human isoform of MZF1 previously identified by Peterson and Morris 1,2 . MZF1 was initially identified as a zinc finger transcription factor necessary for granulocytic differentiation and critical to the regulation of cell proliferation and apoptosis (30 -33). In contrast to previous reports, both MZF1B and MZF1 mRNA transcripts are expressed in numerous tissues. 2 Therefore, previous reports addressing MZF1 function may have been indirect measurements of MZF1B function. Thus, MZF1B may function as an important regulator of granulocytic differentiation, cell proliferation, and apoptosis. The interaction between MZF1B and RAZ1 might be necessary for mediating MZF1B function, or RAZ1 may modify intrinsic MZF1B function. It is also possible that other SCAN box proteins compete with MZF1B for binding to the same protein, thereby providing a transcriptional regulatory mechanism based on the sequestration of specific factors and availability of protein partners. Furthermore, MZF1B and RAZ1 each self-associate in vitro, suggesting that each protein may participate in the formation of unique complexes with distinct functions. Thus, the transcriptional activity of MZF1B is probably mediated by specific protein-protein interactions with RAZ1, MZF1B, and other SCAN box proteins. Identifying the in vivo MZF1B protein partners and their effect on MZF1B activity will provide insight into the possible mechanisms by which MZF1B functions to transcriptionally regulate cell development.