UBL1, a Human Ubiquitin-like Protein Associating with Human RAD51RAD52 Proteins

RAD55, and RAD57 (Adzuma et al., 1984; Alani et al., The RAD51/RAD52-dependent DNA repair pathway 1989; Basile et al., 1992; Emery et al., 1991; Kans and is involved in DNA recombination and DNA doubleMortimer, 1991; Lovett, 1994; Schild et al., 1983; Shistrand break repair in yeast. Although many proteins nohara et al., 1992; Zheng et al., 1993). It is known in the RAD51/RAD52-dependent DNA repair pathway that RAD51 is a RecA-like protein that has DNA-bindhave been identified in yeast, a novel protein(s) that ing, ATP-binding, and DNA strand-exchange activities, functions with RAD51/RAD52 may also exist in huwhile RAD54 contains regions homologous to DNA helmans. Using a yeast two-hybrid system, we have identiicase. Interactions among RAD52, RAD51, RAD55, and fied a 12-kDa protein that associates with the human RAD57 have been identified (Firmenich et al., 1995; RAD51 and RAD52 proteins. This protein shares sigHays et al., 1995; Shinohara et al., 1992; Smith and nificant amino acid homology with the yeast protein Rothstein, 1995). RAD52 and RAD54 are highly exSMT3, which functionally associates with the yeast mipressed during yeast meiosis (Cole et al., 1989). Yeast tosis fidelity protein MIF2. It also shares moderate hoRAD51 and RAD53 also have cell-cycle-dependent exmology with ubiquitin and several other proteins, inpression (Basile et al., 1992; Zheng et al., 1993), and cluding the N-terminus of the RAD23 protein and a RAD53 also participates in cell cycle control (Sanchez ubiquitin cross reacting protein. Therefore, the gene is et al., 1996). tentatively designated UBL1 for ubiquitin-like 1. The In humans, only RAD51 and RAD52 have been reUBL1 mRNA is expressed in many human tissues, most highly in testis. The UBL1 gene is mapped to chromoported (Muris et al., 1994; Shen et al., 1995; Shinohara some 2q32.2–q33, and a related sequence may be loet al., 1993). The human RAD51 protein has activities cated on chromosome 1q23–q25. q 1996 Academic Press, Inc. similar to those of yeast RAD51, including DNA binding and a specific interaction with RAD52 protein (Benson et al., 1994; Shen et al., 1996a). Also, we have found INTRODUCTION that human RAD52 protein self-associates (Shen et al., 1996b). More recently, RAD51 has been shown to accuDNA double-strand break (DSB) is one of the most mulate in the synaptonemal complex, indicating important forms of DNA damage caused by ionizing involvement of RAD51 in meiosis and chromosome reradiation. Efficient repair of DSB is essential for the combination (Haaf et al., 1995). Human RAD52 overexcell to recover from radiation damage. In yeast, the pression in monkey cells enhances their resistance to RAD52 epistasis group genes encode proteins involved radiation and increases the frequency of homologous in DSB repair and recombination (Friedberg et al., recombination (Park, 1995). However, little is known 1991). Several yeast genes in the RAD52-dependent about how these interactions might fit into the context repair pathway have been cloned. These include of DSB repair. RAD50, RAD51, RAD52, RAD53 (SPK1), RAD54, The importance of DSB repair factors is not limited to their roles in modulating radiation sensitivity, but includes their roles in DNA recombination, in repair of Sequence data from this article have been deposited with the GenBank/EMBL Data Libraries under Accession No. U38784. alkylating and cross-linking agent-induced DNA dam1 To whom correspondence should be addressed at LS-6, MS M888, age, and in some physiological processes. For example, Life Sciences Division, Los Alamos National Lab, Los Alamos, NM mutations in the yeast RAD52 group genes result in 87545. Telephone: (505) 667-2789. Fax: (505) 665-3024. E-mail: dchen@telomere.lanl.gov. increased sensitivity to the alkylating agent methyl 2 Abbreviations used: DSB, DNA double-strand break; UBL1, a methanesulfonate (MMS). Indeed, the yeast RAD52 ubiquitin-like protein 1, 12 kDa; Gal4-DA, the DNA activation dogene was cloned by its ability to complement MMS senmain (amino acids 768–881) of Gal-4 protein; Gal4-DB, the DNA sitivity (Adzuma et al., 1984; Schild et al., 1983). The binding domain (amino acids 1–147) of Gal-4 protein; LacZ, b-galactosidase gene; MMS, methyl methanesulfonate. RAD52 gene has been found to function in plasmid


INTRODUCTION
that human RAD52 protein self-associates (Shen et al., 1996b). More recently, RAD51 has been shown to accu-DNA double-strand break (DSB) 2 is one of the most mulate in the synaptonemal complex, indicating important forms of DNA damage caused by ionizing involvement of RAD51 in meiosis and chromosome reradiation. Efficient repair of DSB is essential for the combination (Haaf et al., 1995). Human RAD52 overexcell to recover from radiation damage. In yeast, the pression in monkey cells enhances their resistance to RAD52 epistasis group genes encode proteins involved radiation and increases the frequency of homologous in DSB repair and recombination (Friedberg et al., recombination (Park, 1995). However, little is known 1991). Several yeast genes in the RAD52-dependent about how these interactions might fit into the context repair pathway have been cloned. These include of DSB repair. RAD50, RAD51, RAD52, RAD53 (SPK1), RAD54, The importance of DSB repair factors is not limited to their roles in modulating radiation sensitivity, but includes their roles in DNA recombination, in repair of recombination induced by the cross-linking agent psoralen (Han and Saffran, 1992;Saffran et al., 1992). In mammalian immune systems, various types of antibody genes or antigen receptor genes are generated by V(D)J rejoining, where a DNA strand-break is introduced, and the DSB repair machinery complies. In other cases, integration of viral DNA into the host genome may require DSB repair-associated mechanisms. During meiosis, mechanisms of chromosome exchange (recombination) may overlap with DSB repair. However, the mechanism of the RAD52-associated repair pathway is poorly understood, especially in human cells.
To elucidate the DSB repair mechanism in humans, one of the first steps is to identify human proteins involved in DSB repair. It is assumed that some proteins involved in the same repair pathway may associate with each other in the cells. Therefore, one strategy to identify novel proteins participating in the RAD51/ RAD52-dependent repair pathway is to identify proteins that actually interact with known proteins, such as RAD51 and RAD52. We have initiated experiments to identify a RAD51/RAD52 interacting protein(s) by using the yeast two-hybrid approach. This approach would also identify a gene(s) involved in meiosis. In this article, we report the cDNA cloning of a ubiquitinlike gene by using the two-hybrid approach. This gene is designated UBL1 for ubiquitin-like 1, as recommended by the Human Gene Nomenclature Commit- Tissue-specific mRNA expression shows that UBL1 hybrid screen. See text for details. expresses highest in testis. By FISH analysis using a cDNA probe and PCR analysis using a panel of mouse The second case is a false positive and was to be eliminated. To do hybrid cells that contain a single human chromosome, this, the procedure illustrated in Fig. 1 was used. A clone grown on the UBL1 gene was mapped to chromosome 2q32.2 -SD/Trp 0 /Leu 0 /His 0 was grown in SD/Leu 0 medium for 5 days and q33; a related sequence may be located on chromosome then plated onto SD/Leu 0 agar plates. Due to Try auxotrophs, some of the yeast grown on SD/Leu 0 agar plates may have lost plasmid 1q23-q25. pGBT9/RAD51(Trp / ). Then, 25 individual colonies from SD/Leu 0 plates were duplicately transferred onto SD/Try 0 /Leu 0 and SD/Leu 0 /

MATERIALS AND METHODS
His 0 agar plates in an organized manner. When the original clone on SD/Trp 0 /Leu 0 /His 0 is a true positive, these clones from SD/Leu 0 plates, which do not grow in SD/Trp 0 /Leu 0 (i.e., have lost pGBT9/ Materials. The yeast strains SFY526 (MATa;112;can r ; RAD51), will also not grow in SD/Leu 0 /His 0 . If the original clone on SD/Trp 0 /Leu 0 /His 0 plate is a false positive, a colony will grow in SD/ URA3::GAL1-LacZ) and HF7c (MATa,112;LYS2::GAL1-Leu 0 /His 0 even if it has lost pGBT9/RAD51 (i.e., not grown in SD/ Trp 0 /Leu 0 ). Therefore, true positive clones were identified. The pACT HIS3; URA3 ::(GAL4 17mers)3-CYC1-LacZ) were purchased from Clontech Laboratories (Palo Alto, CA). SFY526 has a LacZ reporter plasmids from positive yeast clones were isolated and electroporated into HB101 cells, from which a large-scale purification of pACT plas-gene fused downstream of the Gal-1 promoter. HF7c has a His reporter gene controlled by a Gal-1 promoter and a LacZ reporter gene mids was performed. The isolated plasmids were then subjected to automated DNA sequencing, as will be described later. controlled by a CYC1 promoter. The Gal-4 DNA activation domain (Gal4-DA) fused cDNA library in pACT vector was also purchased In vivo assay of protein interaction using the yeast two-hybrid sysfrom Clontech. Two hybrid vectors for RAD51 and RAD52 proteins tem. A description of the protein interaction assay using the yeast have been previously reported (Shen et al., 1996a,b). SD and YPD two-hybrid system can be found in the manufacturer's Matchmaker media/plates were prepared as described in the two-hybrid system manual (Clontech Laboratories) and previous publications (Shen et manual (Clontech Laboratories Inc.). al., 1996a,b). Briefly, plasmids for the two fusion constructs (one fused with the Gal4-DB, the other fused with Gal4-DA) were cotrans-Library screen using the yeast two-hybrid system. The library screening for this yeast two-hybrid system was performed according fected into the genetically constructed yeast cells SFY526. Transformed yeast cells were grown on Trp 0 /Leu 0 synthetic agar plates to the Matchmaker Kit manual (Clontech Laboratory Inc.). Briefly, HF7c yeast was first transfected with pGBT9/RAD51 expressing the for 3 days to select yeast clones bearing both fusion vectors. To measure the expression of the b-galactosidase (LacZ) reporter gene, Gal4-DB/RAD51 fusion protein using the polyethylene glycol/lithium acetate method. Yeast HF7c with pGBT9/RAD51 was subsequently which correlates with the interaction of two fusion proteins expressed from these two vectors, LacZ activity in three independent trans-transfected with cDNAs isolated from the pACT library. These cotransfected yeast were grown in SD/Try 0 /Leu 0 /His 0 agar plates. A formants was measured by filter assay (see Matchmaker Manual; and Shen et al., 1996a,b). Quantitative LacZ activity in Miller's unit positive clone in SD/Trp 0 /Leu 0 /His 0 should contain a pACT plasmid coding for a Gal4-DB/RAD51 interacting protein or a protein that is (Miller, 1972) was assayed according to the Matchmaker kit manual (Clontech Laboratories). Briefly, yeast from a single clone were grown able to activate the reporter gene (His) without the RAD51 protein.
overnight in synthetic media lacking Trp/Leu. The density of yeast was determined by measuring the absorbance at 600 nm. Then, 0.1 ml of culture was mixed with 0.7 ml of Z-buffer (16.1 g/liter Na2H-PO4r7H2O; 5.5 g/liter NaH2PO4rH2O; 0.75 g/liter KCl; 0.246 g/liter MgSO4r7H2O; pH 7.0), 50 ml of CHCl3, and 50 ml of 0.1% SDS. O-Nitrophenylgalactoside (4 mg/ml) was used as substrate for LacZ. After 2 h of 30ЊC incubation, the reaction mix was centrifuged, and the absorbance of the supernatant was read at 420 nm. The LacZ in Miller's unit was calculated as 1000 1 [OD 420/(t 1 V 1 OD 600], where t is time of incubation, V is volume of yeast culture, and OD 600 is the absorbance of yeast culture at 600 nm. Chromosome localization by cDNA FISH mapping. To map the chromosome, the protocol described by Heng et al. was used (Heng et al., 1992;Tsui, 1993, 1994). Briefly, human lymphocytes were synchronized with phytohemagglutinin, and metaphase spreads were made by standard procedures. cDNA probe was biotinylated with dATP using the BRL BioNick labeling kit and hybridized to the chromosome slides. The FISH and DAPI signals were photographed separately. The DAPI chromosome pictures were superimposed on the FISH signal pictures to localize the region better.
DNA sequencing. A primer-walking strategy was used to sequence cloned plasmid cDNA inserts. Sequencing was performed with the Taq DyeDeoxy Terminator Cycle Sequencing Kit and the ABI 373A automatic DNA sequencer (Applied Biosystems Inc., Foster City, CA) according to the manufacturer's protocols. Both cDNA strands were sequenced at least once. Sequence editing was performed with the SeqEd software (Applied Biosystems) on a Macintosh computer. Further sequence analysis was performed using the GCG sequence analysis software package (University of Wisconsin, Madison, WI).
Northern hybridization. A multiple-tissue Northern blot containing mRNA from leukocyte, colon, small intestine, ovary, testis, FIG. 3. Alignment of the UBL1 amino acid sequence with those of prostate, thymus, and spleen was purchased from Clontech Laborathe yeast SMT3 protein (GenBank Accession No. U33057), ubiquitin, tories. This membrane was sequentially hybridized with UBL1 human RAD23A (HHR23A), human RAD23B (HHR23B), and a ubiquitin cross reacting protein (UCRP). The alignment was performed with the BESTFIT program in the GCG sequence analysis software package (University of Wisconsin, Madison).
cDNA, human RAD51 cDNA, and human b-actin cDNA (provided with the blot), with stripping between each hybridization according to the protocol provided with the membrane. Hybridization probes were labeled with a random labeling kit (Amersham Corp., Arlington Heights, IL). Other molecular methods. Primers were synthesized with an Applied Biosystems Model 394 synthesizer. pGAD424/UBL1 was constructed by fusing bases 67 -375 (Fig. 2) to the Gal4 DNA activation domain in vector pGAD424, and pGBT424/hRAD12 was constructed by fusing bases 67 -375 (Fig. 2) to the Gal4 DNA binding domain in pGBT9 vector by PCR using BamHI-and SalI-tagged primers. Detailed procedures and construction of other two-hybrid vectors have been described previously (Shen et al., 1996a).  Comparison of the amino acid sequence with nonre-

dundant protein databases (including GenBank, Swiss-cDNA Cloning and Sequence Analysis
Port, PIR, etc.) showed no direct match. However, a yeast protein, SMT3 (GenBank Accession No. U33057), UBL1 was identified from a pACT vector-based lihas 72% similarity and 52% identity with UBL1 (Fig. brary through its interaction with the human RAD51 3). Also, many proteins in the ubiquitin family showed (hRAD51) protein in a two-hybrid system utilizing His moderate homology (40-55% similarity, Ç20% idenand LacZ as selection genes in the HF7c yeast strain tity) to UBL1. Figure 3 shows the alignment of the (Clontech Laboratories). Subsequently, a 1017-bp UBL1 protein sequence to human ubiquitin (Callis et cDNA was isolated from 0.3 1 10 6 independent clones al., 1989) and several ubiquitin-like proteins, including (Fig. 2). The cDNA contains an open reading frame the N-terminal 85 amino acids of the human RAD23 starting at base 67 and ending at base 369 (Fig. 2).
proteins  Spek et al., Because of the characteristic base A at position 64 (Ko-1994) and an interferon-inducible protein (UCRP) zak, 1984), bases 67-69 were assigned as the transla-(Loeb and Haas, 1992, 1994). Table 1 summarizes the tion start codon. This open reading frame codes for a homology between the above-mentioned ubiquitin-like protein of 101 amino acids with a molecular mass of 12 kDa. proteins. . The procedure is as described previously (Shen et al., 1995). examined the UBL1 mRNA level in several tissues, UBL1 Interacts with Human RAD52 as well including testis, by Northern blot (Fig. 5A). It is evident as with RAD51 that, although UBL1 mRNA is expressed in all the To confirm its interaction with hRAD51, only the codtissues tested, testis exhibits the highest level of ing region of UBL1 was fused to the Gal4 DNA activa-mRNA, which is consistent with human RAD51. tion domain in vector pGAD424, which contains a Northern hybridization to human and mouse total weaker promoter than the pACT vector (Clontech Lab-RNA showed a single mRNA species of 1.3 -1.4 kb (Fig.  oratories). This fusion also eliminated the noncoding 5B), indicating that transcripts of this gene exist in region at the 5-end of UBL1 cDNA. UBL1's interaction human and mouse cells. Since the average poly(A) tail with hRAD51 was further tested in another yeast in an mRNA is about 250 bases (Birnstiel et al., 1985) strain, SFY526 (Clontech Laboratories), using LacZ as and a putative poly(A) addition site (AATAAA) is identhe reporter gene. As shown in Fig. 4, neither the vector tified in the 1017-bp cDNA clone (Fig. 2), this cDNA alone nor UBL1 itself activated the expression of LacZ. clone (Fig. 2) is at least close to full length. Cotransfection of pVD3 (amino acids 72-390 of p53 fused to the Gal4 DNA binding domain in the pGBT9 Chromosome Localization of UBL1 vector) or pGBT9/UBL1 with pGAD424/UBL1 did not activate the LacZ gene, indicating no association of The chromosome localization of UBL1 was deter-UBL1 with the truncated p53 nor UBL1 itself. Howmined by FISH analysis using the full-length cDNA as ever, when pGBT9/RAD51 or pGBT9/hRAD52 was coa probe (Heng et al., 1992;Tsui, 1993, 1994). transfected with pGAD424/UBL1, the LacZ gene was Among 100 cells examined, 50 showed paired chromaactivated, indicating an association of UBL1 with tid signals from chromosome 2q only (Figs. 6A and 6B), hRAD51 as well as hRAD52.
7 showed paired chromatid signals from chromosome 1q only, 32 showed signals from both chromosomes 1q Expression of UBL1 and Human RAD51 mRNA and 2q (Figs. 6C and 6D), and 4 showed paired chromain Testis tid signals from chromosome 5 (data not shown). Therefore, 82% of the cells showed signals from chromosome Since it has been shown that the human RAD51 mRNA is highly expressed in testis where meiosis and 2q, and 39% of the cells showed signals from chromosome 1q. To define the regional localization further, 10 mitosis are active (Shinohara et al., 1993), we further partial human chromosome 2  were used for regional mapping by PCR analysis (Fig. 8B). As shown in Fig. 8B, while DNA from cell lines containing the region 2q32.3 -q33 (cell lines 6CS-5, XHB-78, 6x(neo2)-18, and XRV15b(neo2)-11) showed a positive PCR result, cell lines not containing this region (cell lines 6CS-7 and XHB-104) showed negative PCR results. These data confirmed the localization to 2q32.3 -q33.

DISCUSSION
Similar to ubiquitin, ubiquitin-like proteins seem to be involved in many cellular processes. For example, a UCRP has been shown to be conjugated to many cellular proteins that are distributed in a cytoskeleton pattern. The N-terminal amino acids homologous to ubiquitin in RAD23 proteins are also essential for RAD23's DNA repair function (Pejovic, 1995).
The location of UBL1 to 2q32.3 -q33 and a related gene on 1q23-q24 is worthy of discussion. The 2q33 region has been identified as an aphidicolin-inducible fragile site (Tedeschi et al., 1992), and several cancer cells, such as human small-cell lung carcinoma (Kohno et al., 1994) and ependymomas (Rogatto et al., 1993), have chromosome changes within the region 2q32- mapped to 1q23-q24 (Horikawa et al., 1995). Genetic changes on 1q23-q25 have been observed in ovarian individual cells were analyzed by comparing the locacancers and in Burkitt lymphoma-derived cell lines tions of FISH signals with chromosome banding (Fig. (Polito et al., 1995). 7). Based on these data, we concluded that UBL1 is RAD51 shares a moderate sequence homology and located on 2q32.3 -q33 and that a closely related se-some functional similarity (such as DNA binding and quence is located on 1q23-q24. filament formation along DNA strands) with the bacte-To confirm the chromosome localization, DNAs from rial recombination protein RecA (Benson et al., 1994; a panel of rodent hybrids, each containing only a single Shinohara et al., 1992;. The human chromosome, were used for PCR analysis as human RAD52 protein shares significant homology described previously (Shen et al., 1995). Figure 8A with the yeast RAD52 protein only at the N-terminus shows that a signal identical to total human DNA was (Muris et al., 1994;Shen et al., 1995). Two independent amplified from human chromosomes 1, 2, and 5 and functional domains involved in self-interaction and ina weaker signal from chromosome 13, indicating the teraction with RAD51 have been identified (Shen et al., presence of sequences closely related to UBL1 on these 1996a,b). In addition, expression of human RAD52 in chromosomes. monkey cells enhances radiation resistance (Park, We have also searched GenBank for homologous 1995). UBL1's yeast homolog, SMT3, is a suppressor DNA sequences and found that several expressed se-of the yeast gene MIF2 mutation (see GenBank Accesquence tags (ESTs) in GenBank have significant nu-sion No. U33057). MIF2 protein is a yeast centromere cleic acid identity with UBL1 cDNA. Two of them (Ac-protein with homology to the mammalian centromere cession Nos. R17443 and T16960) have 99% identity protein CENP-C (Brown, 1995; Meluh and Koshland, with UBL1 in a region of Ç400 bp, presumably because 1995). MIF2 is also involved in yeast chromosome segthey are ESTs from UBL1 directly. One (GenBank Acregation (Brown et al., 1993). These results may therecession No. T08096) has 62% identity with UBL1 cDNA fore suggest that UBL1 is also involved in mitosis. Coin a region of 234 bp. Another (GenBank Accession No. incidentally, human RAD51 is highly expressed during T08856) has 57% identity in a region of 184 bp. The the S/G2/M phase of the cell cycle in mammalian cells, last two entries have 81% of the nucleic acids identical and human RAD52 in the G2/M phase (unpublished in a region of 212 bp. Therefore, a family of UBL1 hodata). We have cloned a human homolog of yeast ubimologous genes may exist in the mammalian genome. quitin-conjugation enzyme UBC9, which is involved in This may explain why signals from other chromosomes S-and M-phase cyclin degradation (Seufert et al., 1995) are detectable. and mitosis control (Al-Khodairy et al., 1995). This To confirm the regional localization on chromosome 2 hUBC9-like protein also interacts with UBL1, hRAD52, hRAD51, and p53 proteins (Shen et al., further, a panel of human radiation hybrids containing FIG. 8. PCR analysis of human chromosome hybrids in mouse cells as described before. (A) PCR from chromosomes 1, 2, 3, 4, 5, 11, 12, 13, and 14 are shown. ''Buffer'' denotes negative control using buffer alone as the PCR template, ''mouse A9'' indicates DNA from mouse A9 cells from which the hybrid panel was constructed, and ''HSF'' denotes total DNA from human skin fibroblast. Other lanes are molecular size markers. PCR results from other human chromosomes are negative and not shown. The primers used are 5GGTGATCAAGCCTCAGTC (positions 552-569 in Fig. 2) and 5CCACAGTTCAGTTCTCTG (positions 791-774 in Fig. 2). (B) A few X-ray hybrids, containing partial human chromosome 2, were further analyzed by PCR. Labels at the top denote the cell lines used; other labels are the same as in A. While DNA from cell lines containing the region 2q32.3 -q33 (cell lines 6CS-5, XHB-78, 6x(neo2)-18, and XRV15b(neo2)-11) showed a positive PCR result, cell lines not containing this region (cell lines 6CS-7 and XHB-104) showed negative PCR results. For detailed information about these cell lines, please refer to .
termination and 3 processing: The end is in site. Cell 41: 349-1996c). Therefore, we believe the association of UBL1 359. with RAD51/RAD52 to be functionally relevant. The Brown, M. T. (1995). Sequence similarities between the yeast chroassociation of UBL1 and hRAD51/hRAD52 also sugmosome segregation protein Mif2 and the mammalian centromere gests that DNA repair/recombination pathways and protein CENP-C. Gene 160: 111-116. cell cycle/mitosis control pathways may interface with Brown, M. T., Goetsch, L., and Hartwell, L. H. (1993). MIF2 is reeach other.  (1994). Regional assignment of a human DNA discussions and critical reading of the manuscript.