Two forms of mitochondrial DNA ligase III are produced in Xenopus laevis oocytes.

Full-length cDNAs for DNA ligase IV and the alpha and beta isoforms of DNA ligase III were cloned from Xenopus laevis to permit study of the genes encoding mitochondrial DNA ligase. DNA ligase III alpha and III beta share a common NH(2) terminus that encodes a mitochondrial localization signal capable of targeting green fluorescent protein to mitochondria while the NH(2) terminus of DNA ligase IV does not. Reverse transcriptase-polymerase chain reaction analyses with adult frog tissues demonstrate that while DNA ligase III alpha and DNA ligase IV are ubiquitously expressed, DNA ligase III beta expression is restricted to testis and ovary. Mitochondrial lysates from X. laevis oocytes contain both DNA ligase III alpha and III beta but no detectable DNA ligase IV. Gel filtration, sedimentation, native gel electrophoresis, and in vitro cross-linking experiments demonstrate that mtDNA ligase III alpha exists as a high molecular weight complex. We discuss the possibility that DNA ligase III alpha exists in mitochondria in association with novel mitochondrial protein partners or as a homodimer.

Most eukaryotic cells rely on mitochondria for many vital cellular processes such as the generation of adenosine triphosphate (ATP), the regulation of the cytoplasmic redox state, heat production, apoptosis, and the synthesis of amino acids, pyrimidines, heme, aldosterone, cortisol, sex steroids, and many other key metabolites. In animals, the closed circular mtDNA genome encodes only 13 polypeptides, leaving the nuclear genome to provide the vast majority of mitochondrial proteins. All proteins involved in expression, replication, and repair of mtDNA are encoded in nuclear DNA.
One of the key proteins involved in mtDNA maintenance is DNA ligase, required in the final stages of both DNA replication and repair. Initially it was thought that mitochondria lacked repair mechanisms, as pyrimidine dimers induced by UV light were not removed (1). However, in recent years it has become evident that mitochondria possess the ability to remove damage introduced into the mitochondrial genome by alkylation, oxidation, spontaneous base loss, A:G mismatches, and misincorporation of uracil (2,3). Characterization of enzymes involved in mtDNA repair is still at an early stage.
Our purification of a DNA ligase activity from the mitochondria of Xenopus laevis oocytes provided the first characterization of this important protein (4). The 100 -110-kDa ATP-dependent mtDNA ligase was able to ligate the synthetic homopolymers oligo(dT)⅐poly(dA) and oligo(dT)⅐poly(rA). The size and properties of mtDNA ligase invite comparison with nuclear DNA ligases. Five biochemically different DNA ligase activities have been characterized in the nuclei of mammals, although genes have been identified only for DNA ligases I, III, and IV (5). At least one form, DNA ligase II, is a proteolytic fragment of another activity, DNA ligase III (6). Currently it is believed that DNA ligase I is the major enzyme that ligates Okazaki fragments during lagging strand replication (7)(8)(9)(10)(11). This enzyme is also thought to function in long patch base excision repair (12). DNA ligase III has been implicated in short patch base excision repair (13) and meiotic recombination (14). DNA ligase IV appears to be involved in double strand break repair associated with V(D)J recombination (15) and nonhomologous end-joining (16). All of these functions have been studied in more detail in the nucleus than in the mitochondrion.
The mtDNA ligase activity purified from Xenopus oocyte mitochondria has a size and substrate specificity similar to those of human DNA ligase III or IV. Preliminary Western blotting results showed that the mtDNA ligase was immunologically related to human DNA ligase III (4). Since no Xenopus DNA ligase sequences other than DNA ligase I were represented in public data bases, we analyzed mammalian DNA ligase cDNA sequences for potential NH 2 -terminal mitochondrial localization signals. The sequences of human DNA ligase III and IV were originally reported with the prediction that translation would start at the second in-frame AUG since, in each case, the first AUG was not found within an optimal context for efficient translation initiation (17). The full-length cDNA sequences permitted us to predict the polypeptides that would result if translation did initiate at the upstream AUG residues. For both human DNA ligase III and IV, computer analysis of the NH 2 termini of these predicted polypeptides revealed features characteristic of a mitochondrial localization sequence (4). Experimental data supporting the role of human DNA ligase III in mtDNA maintenance was later obtained with green fluorescent protein targeting experiments (18). Lakshmipathy and Campbell (19) also demonstrated that expression of human DNA ligase III antisense mRNA in tissue culture cells decreased the amount of adenylation activity in mitochondria and reduced mtDNA integrity.
Very little characterization of mtDNA ligase has been reported on the protein level. Alternative splicing of the mouse DNA ligase III mRNA produces two transcripts that are predicted to encode an ␣ isoform that contains a BRCA 1 COOH terminus (BRCT) 1 domain and a ␤ isoform that does not (14). BRCT domains have been shown to mediate protein-protein interactions (20,21), including the stable interaction between nuclear DNA ligase III␣ and XRCC1 (22). It remains unclear whether mtDNA ligase is the product of the DNA ligase III␣ or III␤ mRNA. Lakshmipathy and Campbell (23) were not able to detect XRCC1 in immunoblots of crude mitochondrial protein.
Although the detection limit of this assay is uncertain, it appears that XRCC1 might not be directed to mitochondria, and that association with XRCC1 is not required for the stability of mtDNA ligase.
In this paper we report the cloning of X. laevis DNA ligase III␣, DNA ligase III␤, and DNA ligase IV cDNAs as well as additional characterization of the mtDNA ligase protein. RT-PCR revealed that while DNA ligase III␣ and DNA ligase IV are expressed in all tissues, expression of DNA ligase III␤ was only observed in testes and ovary. We demonstrate that both DNA ligase III␣ and III␤, but not DNA ligase IV are detected in mitochondrial lysates from X. laevis oocytes. Purified mtDNA ligase III␣ is selectively contained in a higher molecular weight form. In contrast, mtDNA ligase III␤ behaves as a monomer in gel filtration and sedimentation experiments. We discuss the possibility that DNA ligase III␣ exists in mitochondria in the presence of novel mitochondrial protein partners or as a homodimer.
Isolation of Full-length cDNA Clones and DNA Sequencing and Analysis-The predicted translation products of human and mouse DNA ligase III were aligned using the CLUSTAL program included in the PCGene Software package. Highly conserved sequences, corresponding to the peptides HIKCMFE and DCIYFND, were chosen in order to design degenerate oligonucleotides to amplify an internal portion of X. laevis DNA ligase III cDNA by PCR on first strand oocyte cDNA. The resulting PCR product was used in successive rounds of 5Ј and 3Ј thermal rapid amplification of cDNA ends (24) with the same oocyte cDNA pool in order to generate a composite sequence representing the full-length DNA ligase III␤ message. The initial partial DNA ligase III cDNA was radiolabeled by random priming with a Prime-It II kit and used to screen several cDNA libraries including a gt11 X. laevis stage 24 embryo cDNA library (25).
A protein alignment of human and yeast DNA ligase IV sequences revealed conserved peptides CILDGEM and DKEWHECM used to design degenerate oligonucleotides that were used in PCR reactions with X. laevis oocyte first strand cDNA to obtain a partial clone which served as a probe in subsequent screens with a ZAP X. laevis oocyte cDNA library (kindly provided by M. Roth).
DNA sequencing of all clones was carried out using either the Sequenase chain terminator method (Amersham Pharmacia Biotech) or an automated ABI-PRISM Model 373 DNA Analysis System (PE Applied Biosystems Inc.). cDNA sequences were analyzed for NH 2 -terminal mitochondrial localization signals using Mitoprot (26) (mips.biochem.mpg.de/proj/medgen/mitop) and for other sequence motifs using the Pfam data base (27) (sanger.ac.uk/cgi_bin/pfam).
Semi-quantitative RT-PCR Analysis-Two micrograms of total RNA from adult X. laevis tissues were reverse transcribed with Superscript II RT primed with oligo(dT) following the conditions listed by the manufacturer. One-microliter aliquots corresponding to 40 ng of total RNA were then used as templates for PCR amplification of sequences unique to X. laevis DNA ligase III␣ cDNA or to DNA ligase III␤ cDNA. This was done by using a primer common to both sequences, 5Ј-CTG-GAAAACCGCCACAAC-3Ј, and another primer specific for either DNA ligase III␣, 5Ј-ATGCCTTCTCCTGGTGTG-3Ј, or for DNA ligase III␤, 5Ј-TGTTGTCGTCTGCTCAGC-3Ј. The same templates were also used in order to determine the tissue-specific expression pattern of the DNA ligase IV gene by PCR amplification with the primers 5Ј-CAATACCA-GGCAGATTCC-3Ј and 5Ј-TTCCCCAGGACCTGAAGC-3Ј. PCR amplification of the constitutively expressed ␤ actin cDNA, utilizing the primers 5Ј-TGGAGAAGAGCTATGAGCTGCCTG-3Ј and 5Ј-GTGCCAC-CAGACAGCACTGTGTTG-3Ј, was performed as a control for template quantity in each reaction. PCR amplification was carried out for 30 cycles at 94°C for 1 min, 56°C for 1 min, and 72°C for 1 min. The 30 cycles of amplification used was determined to be within the log-linear range of amplification by quantitating PCR products as a function of the number of cycles.
The PCR products were subjected to electrophoresis on 1.5% agarose TBE gels and detected by staining with ethidium bromide. The expected sizes of PCR products were 691 bp for DNA ligase III␣, 558 bp for DNA ligase III␤, 431 bp for DNA ligase IV, and 201 bp for ␤ actin. Relative amounts of each PCR product were determined by fluorimetric analysis with a FluorImager 595 (Molecular Dynamics). Molar amounts of each PCR product were calculated, adjusted for differences in product size, and used to derive ratios of DNA ligase III␣, DNA ligase III␤, and DNA ligase IV cDNA with respect to ␤ actin cDNA.
Preparation of GFP Fusion Constructs and Localization Studies-NH 2 -terminal sequences suspected to serve as mitochondrial targeting signals were cloned into pEGFP-N1 in order to test their ability to localize enhanced green fluorescent protein (EGFP) to mitochondria (28,29). Amino acids 1-56 of Xenopus DNA ligase III, 1-88 of human DNA ligase III, 1-73 of X. laevis DNA ligase IV, and 1-68 of human DNA ligase IV were amplified by PCR using primers designed to permit cloning into the polylinker of pEGFP-N1 as EcoRI-BamHI fragments. The localization sequence of a known mitochondrial protein, the accessory subunit of DNA polymerase ␥ (DNA Pol ␥B), was also amplified by PCR and subcloned into pDsRed (30) to create a ␥B-DsRed construct for use as a mitochondrial marker. All constructs were sequenced to verify that the inserts were in-frame with the start codon of EGFP. For transfection, the expression vector plasmid DNA was prepared by double-banded CsCl gradient centrifugation. Each EGFP fusion construct was co-transfected with ␥B-DsRed transiently into HeLa cells following standard calcium phosphate transfection protocols (25). Cells were examined 16 h post-transfection with a Nikon PCM 2000 confocal microscope system using a 60XA/1.40 oil immersion objective.
Partial Purification of X. laevis mtDNA Ligase-Mitochondria were isolated from 270 g of X. laevis ovarian tissue by several cycles of differential centrifugation (31). A cleared Triton X-100 lysate of ovary mitochondria was prepared as described (32)  Two Forms of Mitochondrial DNA Ligase III in X. laevis pepstatin A, and 2 g/ml each of leupeptin, antipain, E64, and aprotinin. The dialyzed flow-through was then applied to a 20-ml SP-Sepharose column. At this and subsequent chromatography steps, fractions containing DNA ligase were assayed for adenylation activity as described below. The peak fractions from SP-Sepharose were pooled, diluted 2.5 times with buffer lacking KCl, and concentrated on a Poros HS/F column by step elution prior to chromatography on a HiLoad Superdex 200 (16/60) gel filtration column equilibrated with a buffer containing 0.3 M KCl, 5% glycerol, 20 mM Hepes, pH 7.5, 2 mM DTT, 0.5 mM EDTA, 0.02% Triton X-100, 40 g/ml acetylated gelatin and the standard protease inhibitor mixture described above. Two peaks of adenylation activity were identified, pooled separately, and subjected to chromatography on a 1-ml heparin-Sepharose HiTrap column. Bound proteins were eluted with a gradient of KCl increasing linearly from 50 to 800 mM. Throughout the purification, chromatography fractions were stored at Ϫ80°C after addition of one-half volume of storage buffer containing 75% glycerol, 20 mM Hepes, pH 7.5, 5 mM DTT, 0.02% Triton X-100, 5 g/ml leupeptin, 5 g/ml aprotinin. Analytical glycerol gradient sedimentation was performed as described (33).
Partial Purification of Human and Xenopus Nuclear DNA Ligase III-Human DNA ligase III was prepared from a HeLa cell nuclear extract using phosphocellulose and mono-S column chromatography (34). DNA ligase was detected in each column eluate by assaying enzyme adenylation. DNA ligase III was purified from the post-mitochondrial supernatant of Xenopus ovary homogenates. The fraction was centrifuged at 100,000 ϫ g and the resulting supernatant was treated with 62% saturated ammonium sulfate to precipitate proteins (35). DNA ligase III was purified by SP-Sepharose and hydroxylapatite chromatography and column fractions were assayed for enzyme adenylation.
Formation of DNA Ligase Adenylate-The mechanism-based adenylation of enzyme was used to detect DNA ligase activity at all purification steps. Enzyme fractions were incubated in 10-l reaction mixtures containing 20 mM Tris, pH 8.0, 10 mM MgCl 2 , 5 mM DTT, 8% glycerol, 0.02% Triton X-100, and 0.2 Ci of [␣-32 P]ATP. Reaction mixtures were incubated at 25°C for 15 min and stopped by addition of an equal volume of sodium dodecyl sulfate sample loading buffer. Proteins were resolved on 8% SDS-PAGE (36). Gels were fixed in 30% methanol, 10% acetic acid, dried, and detected by autoradiography or direct Phospho-rImager analysis of the dried gel.
Production of DNA Ligase III and DNA Ligase IV-specific Antibodies-Primers 5Ј-ACTGGCCCCCATATGATGGCAGAG-3Ј and 5Ј-CTGC-TCATCCTCGAGGGTGAGCTT-3Ј were used to amplify a 0.95-kilobase PCR product, corresponding to amino acids 88 to 353 of DNA ligase III, from mouse testis cDNA. Primers 5Ј-CAAATACTCCCATGGTCATGC-AAAAA-3Ј and 5Ј-ATCAAATAATTCTGCGGCCGCGTAACCA-GA-3Ј were used to amplify a 514-bp DNA fragment of X. laevis DNA ligase IV corresponding to amino acids 347 to 515. PCR reactions were performed with Pfu Turbo DNA polymerase to reduce errors and both products were cloned into the polylinker of pET22b(ϩ). Expression was achieved using the BL21(DE3) Escherichia coli strain after induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside. Induced cells were sonicated in lysis buffer (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 20% glycerol, 1 mM ␤-mercaptoethanol, 0.5 mM benzamidine-HCl, 1 M pepstatin, 5 g/ml leupeptin, and 0.2 mM PMSF). A similar buffer was prepared replacing glycerol with 8 M urea in order to solubilize the antigens, which were present in insoluble fractions. Purification of antigens was accomplished by preparative SDS-PAGE on a Bio-Rad Model 491 Prep Cell. Rabbits were immunized using standard protocols.
Cloning, Expression, and Purification of Recombinant Human XRCC1 in E. coli-Recombinant human XRCC1 protein was expressed in E. coli strain BL21(DE3) using pET22b(ϩ). To prepare these constructs, primers 5Ј-CCAGGCGAAGAATTCGACGTTGAC-3Ј and 5Ј-GT-GTATAGCACAGATCTCAGGCTTGC-3Ј were used to amplify a 1.9kilobase PCR product from a human testis first strand cDNA pool, using Pfu Turbo DNA Polymerase. This product was cloned into the TOPOII vector to produce the plasmid pXRCC1TOPO. For bacterial expression, PCR on pXRCC1TOPO DNA with primers 5Ј-GAATTCGA-CCATATGCCGGAG-3Ј and 5Ј-GTATAGCACAGAAGATCTGCTTGC-3Ј were used to produce a fragment capable of insertion into the NdeI and BamHI sites of pET22b(ϩ) to introduce a polyhistidine tag at the carboxyl terminus of XRCC1. Fusion constructs were confirmed by automated DNA sequencing with an ABI-PRISM Model 373 DNA Analysis System (PE Applied Biosystems Inc.). Expression was achieved using the BL21(DE3) E. coli strain after induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside and purified by Ni-NTA chromatography as described (37). Purified protein was dialyzed to remove imidazole and stored at Ϫ80°C in aliquots after freezing in liquid nitrogen.
Affinity Precipitation of Mitochondrial DNA Ligase III-An XRCC1 affinity column was prepared by loading pure recombinant protein onto Ni-NTA-agarose beads to capacity (4 g of protein/l bead). Beads were then washed twice with 4 column volumes of W20 buffer (500 mM sodium phosphate, pH 8.0, 300 mM NaCl, 3 mM ␤-mercaptoethanol, 0.2 mM PMSF, 0.5% Triton X-100, 20 mM imidazole) and resuspended in an equal volume of W20 buffer. Partially purified DNA ligase III samples, containing both the 110-and the 100-kDa forms of the enzyme, were adenylated and used to set up affinity precipitation reactions as follows. Two hundred-microliter reactions were prepared with partially purified DNA ligase III (7 g of total protein), 10 l wet volume of XRCC1 beads, and W20. Reactions were incubated on ice for 1 h with frequent gentle mixing. 35 l were removed, imidazole added to 250 mM final concentration, and stored on ice as a "load" sample. The remaining reaction mixture was gently centrifuged for 1 s and the supernatant containing nonadsorbed material was removed and saved as an "un-bound" fraction. The Ni-NTA pellet was washed three times with 190 l of W20 buffer. Adsorbed material was eluted from the Ni-NTA beads with two 190-l washes of EB250 buffer (500 mM sodium phosphate, pH 8.0, 300 mM NaCl, 3 mM ␤-mercaptoethanol, 0.2 mM PMSF, 0.5% Triton X-100, 250 mM imidazole). Twenty-five l of load, unbound, wash, and elution samples were analyzed by electrophoresis on 8% SDS-PAGE. Gels were fixed in 30% methanol, 10% acetic acid, dried, and detected by direct PhosphorImager analysis of the dried gel.
In Vitro Cross-linking-One-microliter samples of partially purified DNA ligase III␣ and III␤ activities from X. laevis oocyte mitochondria and human nuclear DNA ligase III were adenylated as described above. Labeled material was incubated with 0, 0.05, 0.25, and 0.5 mM disuccinimidyl suberate for 30 min at room temperature or with 0.001, 0.01, and 0.03% glutaraldehyde for 15 min at 37°C. Following incubation with cross-linker, an equal volume of sodium dodecyl sulfate sample loading buffer was added and proteins were resolved by electrophoresis on 6% SDS-PAGE. Gels were stained with Coomassie Blue to visualize molecular weight markers and dried following a brief period of destaining. Cross-linked products were detected by PhosphorImager analysis of the dried gel.
Blue Native Polyacrylamide Gel Electrophoresis-Partially purified DNA ligase III␣ and III␤ activities from X. laevis oocyte mitochondria and human nuclear DNA ligase III were analyzed by blue native electrophoresis (BN-PAGE) to compare molecular masses and analyze the polypeptide composition of protein complexes (38). A 6 -13% polyacrylamide gradient gel (with a 4% stacking gel) was prepared with 50 mM BisTris, pH 7.0, 500 mM 6-aminocaproic acid using a Hoeffer Tall Mighty Small gel electrophoresis system. Samples were labeled by adenylation and loaded in 0.5% Serva Blue G, 50 mM aminocaproic acid, pH 7.0. The gel was run at 4°C with cathode buffer containing 15 mM BisTris, pH 7.0, 50 mM Tricine, 0.02% Serva Blue G, and anode buffer containing 50 mM BisTris, pH 7.0. Electrophoresis was performed for 1 h at 100 V before increasing the voltage to 500 V for 20 min. Electrophoresis was stopped to change the cathode buffer for fresh buffer without Serva Blue G and resumed at 500 V until the dye front exited the bottom of the gel.

RESULTS
Full-length cDNA Clones Encoding X. laevis DNA Ligase III␣ and III␤-An internal portion of X. laevis DNA ligase III was generated by standard PCR methods with degenerate primers designed from peptide sequences conserved between human DNA ligase III and yeast DNA ligase IV using first strand ovarian cDNA as template. A resulting 1435-bp DNA fragment revealed extensive homology to the central region of human DNA ligase III. 5Ј and 3Ј rapid amplification of cDNA ends were performed as described to complete the clone (24). Five PCR products were isolated and used to construct a 3.0-kilobase composite sequence of a X. laevis cDNA with homology to human and mouse DNA ligase III␤ (14,39). The internal PCR fragment initially obtained with degenerate primers was used as a probe to screen several cDNA libraries, resulting in identification of a 4.8-kilobase cDNA with a COOH terminus similar to human DNA ligase III␣.
The open reading frame of the 110-kDa Xenopus DNA ligase III␣ exhibits 63% sequence identity with human and mouse DNA ligase III␣ and 91% identity with the 100-kDa Xenopus DNA ligase III␤. The predicted translation products beginning with the first AUG codons encountered in the frog cDNAs contain potential amino-terminal mitochondrial localization signals, as described for human DNA ligase III cDNAs in the Introduction. The 5Ј-and 3Ј-untranslated regions flanking both ORFs contain several in-frame stop codons. The 3Ј-untranslated region of the DNA ligase III␣ cDNA extends past the ORF for 2663 bp but does not contain a poly(A) tail. The 3Ј-untranslated region of the DNA ligase III␤ cDNA extends past the ORF for 213 bp before two consensus polyadenylation signals (40) 18 and 8 bp before the poly(A) tail.
The ORFs were analyzed for sequence motifs by searching the Pfam data base and by comparing with previous alignments of ATP-dependent DNA ligases and RNA capping en-zymes (41,61). This analysis revealed a set of characteristic motifs that occur in the same order with similar spacing in ATP-dependent DNA ligases from yeast to man. The DNA ligase III␣ cDNA possesses a COOH-terminal BRCT domain, which is missing from DNA ligase III␤ cDNA (Fig. 1). The predicted translation products of X. laevis DNA ligase III␣ and III␤ also differ at seven amino acid positions. As several specimens were used to generate the various cDNA pools utilized, these differences could be attributed to polymorphisms within the outbred Xenopus population used. In mammals, the two DNA ligase III variants are generated by alternative splicing (14), a process which may also occur in frogs. The Xenopus genome has a high proportion of duplicate genes displaying less than 10% sequence divergence (42), making it difficult to rule out that the differences noted between the DNA ligase III␣ and III␤ clones are produced by two distinct genes.
Full-length cDNA Clone Encoding X. laevis DNA Ligase IV-Screening of a X. laevis oocyte cDNA library resulted in a full-length cDNA clone of 3.1 kilobases. Northern blotting confirmed the presence of a single DNA ligase IV mRNA in Xenopus ovarian tissue with a size corresponding to the cDNA clone The open reading frames for human DNA ligase III␣ (hsDNL3A, GenBank TM accession number NM002311) and mouse DNA ligase III␣ (mmDNL3A; GenBank TM accession number MMU66058) were aligned with X. laevis DNA ligase III␣ (AF393654) and ␤ (AF393655), xlDNL3A and -B, respectively, using the AlignX program included in the Vector NTI software package. Identical amino acid residues are in white type on a black background; residues that are identical in at least two of three sequences are in black on a gray background. Less conserved residues are in black on a white background. Motifs identified by searching the Pfam data base, by utilizing the Mitoprot algorithm and by comparing to previously identified conserved sequences in DNA ligases (41,61), are indicated with bars and named above the sequences. isolated (data not shown). The 5Ј-untranslated region of the cDNA clone does not contain an in-frame stop codon preceding the first ATG. Nevertheless, we suggest that this is the correct translation start site since the amino-terminal sequence is highly conserved (ϳ70%) from frog to man, indicating that it may be important for metazoan DNA ligase IV function (Fig. 2). The hypothetical DNA ligase IV would include sequences upstream of the start site initially identified by Wei et al. (17). No NH 2 -terminal sequence data of DNA ligase IV protein has been reported in any organism to date, and the low abundance of this enzyme is not inconsistent with translation initiation in an unfavorable sequence context.
Xenopus DNA ligase IV shows ϳ80% sequence identity with its human counterpart. The ORF was analyzed for motifs of interest by searching the Pfam data base and by comparing with previous alignments of ATP-dependent DNA ligases and RNA capping enzymes (41). The occurrence of two carboxylterminal tandem BRCT domains is conserved in yeast, frog, and human DNA ligase IV (Fig. 2).
Expression of DNA Ligase III and IV mRNAs in Adult X. laevis Tissues-The tissue-specific expression of X. laevis DNA ligase III and IV genes was investigated for several adult frog tissues using RT-PCR. Specific primers were used to amplify sequences of DNA ligase III␣, DNA ligase III␤, DNA ligase IV, or the constitutively expressed ␤ actin gene, which was used for normalization of samples. The exon-specific primers used for the amplification of the DNA ligase III mRNAs were designed to be of the same length and GC content, so that the relative intensity of PCR products would reflect relative mRNA abundance.
The steady-state level of DNA ligase III␣ mRNA is fairly uniform in most somatic tissues and testes but is increased 1.5-2-fold in muscle and ovary (Fig. 3A). The DNA ligase III␤ mRNA is only detected in ovarian and testicular tissues (Fig.  3A). DNA ligase IV is also expressed at higher levels in skeletal muscle and germ line tissues than in heart, kidney, and liver (Fig. 3B). In general, the highest level of DNA ligase III␣, DNA ligase III␤, and DNA ligase IV mRNA detected is in ovarian tissue.
Functional Analysis of the Putative DNA Ligase III and IV MLS-Most proteins destined for mitochondrial import contain 20 -50-residue NH 2 -terminal sequences capable of forming positively charged amphiphilic helices, which are removed during translocation into the organelle. Due to the lack of a specific consensus mitochondrial localization signal (MLS) sequence, several algorithms have been developed to help identify such signals (26,43). Fig. 4A shows probability scores for MLS sequences for known eukaryotic DNA ligases calculated using the Mitoprot program. DNA ligase III homologs from human and Xenopus are predicted to localize to mitochondria with high probabilities while DNA ligase IV is not predicted to contain an MLS sequence. These prediction methods provided a guide for more definitive experimental tests of mitochondrial localization of these gene products. Constructs were prepared to express the NH 2 -terminal sequences found in the longer translation products of human and Xenopus DNA ligase III and IV as fusions with the red-shifted variant of green fluorescent protein, EGFP. As a control, the localization sequence of a well established mitochondrial protein, DNA polymerase ␥B (44), was fused to the Discosoma red fluorescent protein (DsRed). HeLa cells were co-transfected with DNA ligase-EGFP fusions in addition to ␥B-DsRed and verification of mitochondrial targeting was determined by co-localization. Cells transfected with human and Xenopus DNA ligase III display patterns similar to ␥B-DsRed, confirming their ability to target proteins to mitochondria. Some nuclear distribution of GFP was ob- The open reading frames of DNA ligase IV from human (hsDNL4; GenBank TM accession number NM002312), Saccharomyces cerevisiae (scDNL4; GenBank TM accession number CAA99193.1), and X. laevis (xlDNL4; AF393656) were aligned using the AlignX program included in Vector NTI software. Amino acid conservation is indicated with different typefaces as described in the legend to Fig. 1. Similarly, protein sequence motifs are indicated as in Fig. 1. The asterisk indicates the putative downstream translation start site identified in the original report of the human DNA ligase IV sequence (17). served in these transfected cells in addition to the mitochondrial staining. Neither the human nor Xenopus DNA ligase IV-EGFP constructs displayed a mitochondrial distribution (Fig. 4B). These experiments served to show that the mitochondrial targeting of DNA ligase III is evolutionarily conserved from Xenopus to humans.
Two Forms of DNA Ligase III Are Detected in X. laevis Oocyte Mitochondria-Xenopus oocytes contain two forms of DNA ligase III mRNA that differ with respect to a COOH-terminal BRCT domain that is present in DNA ligase III␣, but not in DNA ligase III␤ cDNAs (Fig. 1). Ovary mitochondrial lysates also contain two forms of DNA ligase protein that differ by 10,000 on SDS-PAGE, with the higher molecular weight form eluting from heparin-Sepharose slightly ahead of the lower molecular weight form (Fig. 5A). Both polypeptides are recognized by DNA ligase III-specific polyclonal antisera generated against a highly conserved NH 2 -terminal region of the mouse homolog (Fig. 5B). DNA ligase IV was not detected (Fig. 5C).
The 10,000 difference in molecular weight is the approximate size of the BRCT domain that is present in DNA ligase III␣, but absent in DNA ligase III␤ cDNAs. In nuclei, this domain mediates binding between DNA ligase III␣ and XRCC1 (22). Since DNA ligase III␤ does not interact with XRCC1 (45), we utilized this differential binding to determine whether mtDNA ligase III activities were translated from III␣ or III␤ mRNA or both. An affinity matrix was generated with polyhistidine-tagged recombinant human XRCC1 as described (45). Fractions containing both the 110-and 100-kDa forms of mtDNA ligase III were adenylated, incubated with the XRCC1 affinity column, and bound proteins were eluted with imidazole. Following resolution on SDS-PAGE, the enzyme was detected by autoradiography. As predicted, only the 110-kDa form of mtDNA ligase III bound specifically to XRCC1, confirming that it possesses a BRCT domain (Fig. 6). A large fraction of the mtDNA ligase III␣ did not bind the XRCC1 column, suggesting that it may be associated with a protein that blocks this interaction. The 100-kDa form of mtDNA ligase III did not bind XRCC1, indicating that it lacks a BRCT domain. This is consistent with the hypothesis that the 110-and 100-kDa Xenopus mtDNA ligase isoforms are products of the DNA ligase III␣ and III␤ messages, respectively. were aligned with the Align X feature of Vector NTI Software and used to construct a phylogenetic tree. NH 2 -terminal sequences were analyzed for their probability of encoding a mitochondrial localization signal using Mitoprot (P score, shown in bold for sequences with high scores). Asterisks denote mitochondrial localization sequences which have already been proven functional in this or other studies (18,52). Panel B, EGFP fusion constructs containing the NH 2 -terminal sequences beginning with the first in-frame AUG in DNA ligase III and IV cDNAs from X. laevis (XL) and H. sapiens (HS) demonstrate that DNA ligase III, but not DNA ligase IV, is capable of producing a mitochondrial product. HeLa cells were co-transfected with ligase-EGFP fusions along with a mitochondrial marker (␥B-DsRed). Cells were examined by confocal microscopy with standard fluorescein isothiocyanate and rhodamine filters 16 h post-transfection.
The two forms of mtDNA ligase were found to reside in complexes of very different native size. A preparation of mtDNA ligase extensively purified by column chromatography was analyzed by gel filtration resulting in two well resolved species, a larger form containing mtDNA ligase III␣ and a smaller mtDNA ligase III␤ that eluted near a 158-kDa protein marker, aldolase (Fig. 7A). A plot of (-logK av ) 1/2 versus Stokes radius for a series of reference standards permitted calculation of the Stokes radii of mtDNA ligase III␣ and III␤ as 68.7 and 54.3 Å, respectively. The mtDNA ligase III␣ and III␤ were pooled separately, concentrated by chromatography on 1-ml heparin-Sepharose columns, and further analyzed by glycerol gradient sedimentation as shown in Fig. 7B. The ␣ and ␤ isoforms sedimented at 6.2 and 5.3 S, respectively. Analysis of the gel filtration and sedimentation data by the method of Siegel and Monty (46) provided estimates of the native mass of mtDNA ligase III␣ and III␤ of 220 and 150 kDa, respectively. These hydrodynamic methods clearly indicate that both forms have an elongated native conformation, so that the calculated native masses are not expected to be highly accurate. Interestingly, DNA ligase I has also been reported to have an elongated conformation (47).
We also attempted to visualize the difference in native size of DNA ligase III␣ and III␤ fractions purified by gel filtration using native polyacrylamide gel electrophoresis. We employed the blue native-PAGE method in which proteins are mixed with Coomassie Blue prior to electrophoresis to impart a relatively uniform negative charge without denaturing the protein. This gel system has been used extensively to characterize oligomeric enzyme complexes (38,48). DNA ligase III␤ migrated as a homogeneous smaller species. DNA ligase III␣ migrated in BN-PAGE as two species, one form in a larger complex migrating slightly faster than a ferritin marker, and a second species with mobility similar to DNA ligase III␤ (Fig. 7C). We conclude that this smaller species is produced by dissociation of the larger form during electrophoresis since the enzyme loaded on this gel was purified by gel filtration within a larger complex. These results are consistent with the hypothesis that the BRCT domain present in mtDNA ligase III␣ mediates the formation of a protein com- plex and that mtDNA ligase III␤ is excluded from this complex. The larger size of mtDNA ligase III␣ is not due to an association with mtDNA pol ␥, since this enzyme elutes from the gel filtration column between the peaks of ligase III␣ and ligase III␤ (see arrow in Fig. 7A) and has a higher sedimentation coefficient (33).
The nature of mtDNA ligase III␣ and ␤ was further examined by chemical cross-linking experiments with glutaraldehyde and disuccinimidyl suberate, a homobifunctional N-hydroxysuccinimide ester that targets primary amines found in proteins at physiological pH (49). Adenylated Xenopus mtDNA ligase III␣, mtDNA ligase III␤, and human nuclear DNA ligase III were incubated with increasing concentrations of cross-linker followed by resolution on SDS-PAGE. Unlike mtDNA ligase III␤, the radiolabeled mtDNA ligase III␣ and human nuclear DNA ligase III␣ clearly form cross-linked species that migrate in the range of 250 -270 kDa (Fig. 8). Higher concentrations of cross-linker caused DNA ligase III␣ to behave with a slightly more rapid electrophoretic mobility, presumably because increasing intramolecular cross-links result in a more compact structure. These results confirm that the BRCT domain found in mtDNA ligase III␣ is utilized for complex formation. MtDNA ligase III␤ was not identified within a discrete complex, although some glutaraldehyde cross-linked material was retained in the gel well. The disuccinimidyl suberate crosslinked mtDNA ligase III␤ migrated near 100,000 with no cross-linked species identified in the range of ϳ150,000, the native molecular weight calculated from gel filtration chromatography and sedimentation. The larger apparent size determined by gel filtration may reflect an elongated shape for the native protein, as noted above.

DISCUSSION
Two MtDNA Ligase Polypeptides Are Forms of DNA Ligase III-The initial suggestion that Xenopus mtDNA ligase is a form of DNA ligase III (4) was considered preliminary since DNA ligase III had not been cloned in Xenopus at that time. This paper provides the sequences of both Xenopus DNA ligase III and IV along with more definitive immunological evidence for the mitochondrial localization of DNA ligase III. Other genetic studies have provided evidence that in humans and rodents the mtDNA ligase is also a form of DNA ligase III (18,19,23). The present work provides the most detailed analysis to date of mtDNA ligase at the protein level demonstrating that both ␣ and ␤ isoforms of DNA ligase III can be isolated from mitochondria of Xenopus oocytes. We have found no differences between these two isoforms of mtDNA ligase in functional ligation assays.
The sequences reported here confirm that the structures of DNA ligase III and IV are highly conserved from amphibians to humans. Both the human and Xenopus DNA ligase III genes contain functional NH 2 -terminal mitochondrial localization signals encoded by sequences following the first in-frame translation start site. Even though these start sites are not found within typical translation initiation consensus sequences, inefficient initiation at these sites may produce a quantity of enzyme sufficient to meet the needs for mtDNA maintenance. Our results do not rule out the possibility that translation initiation at the downstream AUG provides nuclear DNA ligase III as suggested (17). We have found no evidence to support a role for DNA ligase IV as a mtDNA ligase in Xenopus or human cells. Interestingly, budding and fission yeasts only possess DNA ligase I and IV (50,51). In S. cerevisiae, DNA ligase I provides enzyme to both the nuclear and mitochondrial compartments (52). This suggests that the DNA ligase III gene is either a specialized variant that arose later in evolution or was lost when the yeast DNA ligase I acquired a mitochondrial localization signal.
The cloning of Xenopus DNA ligase III and IV has produced additional insights into the biology of these proteins. The RT-PCR data shown in Fig. 3 provides the first indication that, at least in Xenopus, the expression of DNA ligase III␤ is not restricted to testes as reported in mouse (14), but occurs in ovary tissue as well. Since both germ tissues are actively involved in recombination, it will be interesting to determine whether DNA ligase III␤ plays a special role in this process. In addition, the observation that the high homology between DNA ligase IV in Xenopus and humans begins shortly after the first in-frame AUG is consistent with the likelihood that these sequences are expressed in both organisms and may contribute to the function of the enzyme.
DNA Ligase III as an Example of a Dual-function Mitochondrial Nuclear Gene-Proteins involved in mtDNA maintenance can be divided into two classes: those that function exclusively in the mitochondrion, like DNA polymerase ␥, and those that function in both the nuclear and mitochondrial compartments. Mitochondrial genes have been described with shared locations in a variety of cell compartments and in many biochemical pathways (53,54). However, the mtDNA repair pathways appear to be particularly rich in such dual-function genes. The genes for uracil DNA glycosylase, 8-oxoguanine DNA glycosylase, and the human MUTY and NTH1 homologs generate activities for nuclear and mitochondrial compartments by alternative splicing of nascent mRNAs (55)(56)(57). Recently, human AP endonuclease 2 has also been detected in mitochondria (58). In many cases, dual-function genes provide enzyme isoforms to different compartments using alternative splicing. DNA ligase III provides an unusual example of a bifunctional gene that appears to generate multiple activities by alternative translation initiation. A broadly based proteomic analysis will be required to determine the extent to which this occurs for other gene products.
What Is the Role of DNA Ligase III␤?-To our knowledge, our purification of a mitochondrial DNA ligase III␤ represents the first documentation that the ␤-isoform is produced as a stable form in any organism. For the most part, previous papers dealing with the ␤-isoform of DNA ligase III have been limited to studies of the mRNA and have not studied the cellular distribution or stability of the endogenous protein. Since the stability of nuclear DNA ligase III␣ depends on its association with XRCC1 (22), it will be important to determine whether DNA ligase III␤ is stable outside of the mitochondrion. While it is tempting to speculate that mtDNA ligase III␤ plays a special role in mtDNA maintenance, it is important to recall that its mRNA has only been detected in germline tissues ( Fig. 3; Ref. 14). Since somatic cells appear to rely only on mtDNA ligase III␣ for mtDNA maintenance, and since we know of no unique function for mtDNA ligase III␤ in germline cells, at the present we cannot suggest any specific role for mtDNA ligase III␤ that cannot be carried out equally well by mtDNA ligase III␣.
What Protein Binds to MtDNA Ligase III␣?-The BRCT domain found in DNA ligase III␣ has no known enzymatic function but has been implicated in interaction with the BRCT domain of XRCC1. The similar size of Xenopus mtDNA ligase III␣ and human nuclear DNA ligase III␣ observed in Figs. 7 and 8 would appear to support the notion that both forms of DNA ligase III␣ are associated with XRCC1. We have not been able to detect XRCC1 in either mtDNA ligase III␣ or nuclear DNA ligase III from Xenopus ovaries using commercial antibodies generated against recombinant human XRCC1 (data not shown). Since the epitopes recognized by these antisera may not be conserved in Xenopus XRCC1, this negative data does not rule out the possibility that XRCC1 associates with mtDNA ligase III␣. However, mtDNA ligase levels were not reported to decrease in XRCC1-deficient Chinese hamster ovary cells (23). Our observation of a monomeric mtDNA ligase III␤ in Xenopus oocytes provides further evidence that XRCC1 is not required for stability of DNA ligase III in mitochondria.
If XRCC1 is not a partner for mitochondrial DNA ligase III␣, how is this protein assembled into a higher molecular weight form? It is clear that this assembly depends on the presence of the BRCT domain. It appears that mtDNA ligase III␣ must either associate with a novel protein partner or exist as a homodimer. We cannot rule out the possibility that a non-BRCT domain protein binds the BRCT domain of mtDNA ligase III␣, particularly since a non-BRCT domain protein has been identified as a binding partner for BRCA1 (59). Nevertheless, the simplest hypothesis is that a potential binding partner should be another BRCT domain protein. Analysis of the human genome for BRCT domain proteins using the SMART program (smart.embl-heidelberg.de/) did not reveal any potential partners with a BRCT domain, an apparent mitochondrial localization signal and a size consistent with the complex we observed. It is possible that mtDNA ligase III␣ may exist as a homodimer in which the BRCT domains provide a dimer interface. The tertiary structure of the BRCT domain found at the carboxyl terminus of human XRCC1 was recently elucidated by x-ray crystallography (21). The XRCC1 crystal contains two BRCT domains in the asymmetric unit that make contacts through residues found at the amino terminus of the BRCT motif. Dulic et al. (60) used the crystal structure of the XRCC1 homodimer to generate a model of the BRCT interface between the COOH-terminal 96 amino acids of human XRCC1 and the COOH-terminal 77 amino acids of human DNA ligase III␣. Since XRCC1 is capable of binding both to DNA ligase III␣ and to itself, we suggest that the BRCT domain of DNA ligase III␣ might mediate dimerization. A model based on this suggestion does not show any obvious steric interactions or side chain clashes that would prevent dimerization of mtDNA ligase III␣ through BRCT domains. Thus, we suggest as a testable working hypothesis for future research, that mtDNA ligase III␣ may exist as a homodimer.