Different substrate specificities of the two DNA ligases of mammalian cells.

Mammalian cells contain the DNA ligases I and II. These enzymes show different molecular weights and heat labilities, and antibodies against ligase I do not inhibit ligase II. Here, the nonidentical substrate specificities of the enzymes are described. Under standard reaction conditions DNA ligase I, but not ligase II, catalyzes blunt-end joining of DNA, while ligase II is the only activity that joins oligo(dT) molecules hydrogen-bonded to poly(rA). These differences facilitate the distinction between the two enzymes and should permit further analysis of their functions.

DNA ligases are required for replication and repair processes. Such enzymes have been found in extracts of mammalian cells (for reviews see Refs. l and 2). However, the mammalian DNA ligases have not been fully characterized, apparently because of the ready availability and ease of use of microbial enzymes for in vitro ligation during molecular cloning. We have previously described two distinct DNA ligase activities in the nuclei of mammalian cells (3, 4), as opposed to the single species found in bacteria and yeast. The two mammalian DNA ligases both require ATP as a cofactor, but they do not cross-react serologically, and only one of them is induced on cell proliferation (4,5).
The major ligase activity in proliferating mammalian cells is due to a high molecular weight protein termed DNA ligase I. This enzyme is induced up to 15-fold together with DNA polymerase CY during rat liver regeneration (5,6), suggesting a role for ligase I in DNA replication. The enzyme exhibits a short half-life (about 30 min) in vivo (7). Ligase I resembles the well-known phage T4 DNA ligase with regard to its biochemical mechanism of action, and covalent enzyme-AMP and DNA-AMP reaction intermediates have been shown to occur (1). The native mammalian enzyme has a molecular weight of 180,000-200,000, as determined by sedimentation velocity and gel filtration measurements (3,6, 8-10), and the same size is observed for the protein by sodium dodecyl sulfate-gel electrophoresis (7). Size heterogeneity has been reported for ligase I, and the enzyme appears to be present both in monomeric form and as a 400-kDa dimer in cell extracts (6,11). In addition, ligase I is susceptible to proteolysis and is easily degraded during extraction and purification * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aclvertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. to large active fragments of molecular weight 90,000-140,000.
The activities of such truncated forms are completely inhibited by rabbit antibodies against ligase I (4,7).
The mammalian DNA ligase I1 is a smaller protein of molecular weight -80,000. The enzyme is very heat-labile, exhibits a sharp pH optimum at 7.8, and is not affected by antibodies against ligase I (3, 4). Unlike ligase I, ligase I1 is not induced on cell proliferation (4-6). Furthermore, on subcellular fractionation DNA ligase 11 appears to be more firmly associated with the chromatin than does ligase I (4,6).

EXPERIMENTAL PROCEDURES
Enzyme Purification-DNA ligases I and 11 were purified from calf thymus as described previously (3,4) with minor modifications.
Briefly, 600 g of fresh thymus glands were disrupted at 0 "C in a Waring blender with 3 liters of a buffer containing 0.1 M NaCl, 50 mM Tris-HC1 (pH 7 3 , 10 mM 2-mercaptoethanol, 1 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 pg ml" each of the protease inhibitors pepstatin, leupeptin, and chymostatin. After centrifugation, nucleic acids were removed by precipitation with 0.5% Polymin P in the presence of 0.2 M NaCI. The activities were purified by ammonium sulfate fractionation (42-67% saturation), dialyzed, applied to phosphocellulose (two parallel columns of 4 X 30 cm each) in 50 mM NaCl, 50 mM Tris-HC1 (pH 7.2), 10 mM 2-mercaptoethanol, 1 mM KZHP04, 1 mM EDTA, and eluted step-wise with the same buffer containing 0.5 M NaCl and no EDTA. Both ligase I and ligase I1 were present in this fraction, which was applied directly to a hydroxyapatite (HA-Ultrogel, LKB Products) column (2.5 X 18 cm) and eluted with a linear gradient (800 ml) of 1 to 200 mM K2HP0, in 0.5 M NaCI, 50 mM Tris-HC1 (pH 7.5), 10 mM 2-mercaptoethanol. This procedure separates ligases I and I1 (see below, Fig. 2). The two peaks of activity were further purified separately as described previously (41, ligase I by gradient chromatography on phosphocellulose and gel filtration on Sephadex G-150, and ligase 11 by gel filtration only. Active fractions were pooled, %fold concentrated by dialysis against column buffers containing 50% glycerol, and stored at -20 "C. In agreement with previous results (4), DNA ligase I preparations by this procedure were approximately 1000-fold purified in 10% yield and had specific activities of 0.01-0.02 unit (12)/mg protein, while DNA ligase I1 preparations were about 200-fold purified in 15% yield and had specific activities of 0.003-0.005 unit/mg protein. Phage T4 DNA ligase was purchased from New England Biolabs.
Nicked circular DNA was prepared by EcoRI cleavage of plasmid pAT153 in the presence of ethidium bromide (14); 0.2 pg of nicked DNA was included in each 20-p1 reaction mixture with 1.5 x loA6 units (12) of T4 DNA ligase, or mammalian ligase I or 11. Reactions were allowed to proceed for 1 h at 14 "C and were terminated by extraction with buffered pheno1:chloroform (1:1), The DNA solutions were subsequently characterized by agarose gel electrophoresis.
Blunt-ended DNA substrates were prepared by cleavage of +X174 replicative form DNA with Hpal. DNA (0.2 pg) was incubated in a 9079 40-pI standard reaction mixture supplemented with polyethylene glycol 6000 (17.5% when not otherwise stated) for 1 h at 37 "C with 1.5 X 10-" units (when not otherwise stated) of T 4 DNA ligase or mammalian DNA ligase I or 11.
Inhibition of Ligases-Ligase I antibody (0.5 pg, purified from serum by precipitation with ammonium sulfate (4)) was preincubated for -5 min at 4 "C with T4 ligase or mammalian ligase I or I1 in reaction mixtures. The conversion of nicked circular DNA to a covalently closed form by mammalian ligase I was inhibited, while ligase I1 and T4 ligase were not affected (Fig. 1). Heat inactivation experiments with the mammalian ligases I and I1 were performed as described previously (3).

RESULTS AND DISCUSSION
Joining of Hybrid Substrate"T4 DNA ligase effectively joins oligodeoxyribonucleotides hydrogen-bonded to a complementary polyribonucleotide, while Escherichia coli DNA ligase is unable t.o catalyze this reaction (2). Similarly to the E. coli enzyme, the ATP-dependent mammalian DNA ligase I does not show detectable activity with a hybrid oligo(dT).
poly(rA) substrate (9), as also observed here (Fig. 2)  amounts of T4 ligase were as active as the T4 enzyme by itself (data not shown). In contrast to the results obtained with ligase I, the mammalian ligase I1 joined the hybrid substrate at about 50% of the rate observed with a nicked double-stranded polydeoxyribonucleotide (Fig. 2). The reaction of DNA ligase I1 with either substrate was directly proportional to time (up to 2 h at 16 "C) and to enzyme concentration (up to 3 X lo-' units). Moreover, the joining activities for oligo(dT) bound to either a poly(dA) or a poly(rA) complementary chain cochromatographed during hydroxyapatite chromatography (Fig. 2) as well a s on further purification of ligase I1 by gel filtration. Furthermore, the two activities exhibited the same heat lability (50% inactivation in 5 min at 42 "C) and may be ascribed to the same enzyme.
As shown earlier, ligase I1 is unable to join strand interruptions in double-stranded polyribonucleotides, or singlestranded oligo(dT) molecules (1). By comparison, ligase I1 must be more than 100-fold more active than ligase I with the poly(rA) .oligo(dT) substrate under our standard assay conditions and a range of related conditions. The presence of a distinct catalytic activity of mammalian ligase 11, not found in ligase I, would appear to confirm that the two enzymes represent different gene products.
Formation of 3'-5'-Phosphodiester Ronda-In order to demonstrate that the alkaline phosphatase resistance of radioactive phosphate residues in the polymer substrates treated with mammalian DNA ligase I or ligase I1 was due to the formation of phosphodiester bonds, the substrates were degraded to mononucleotides with micrococcal nuclease and spleen phosphodiesterase (12). Generation of 3'-5'-phosphodiester bonds between oligo([5'-:"P]dT) moieties hydrogen-bonded to polV(dA) or poly(rA) would allow the recovery of [3'-:i2P] d T M P , while this would not be possible if no ligat.ion had occurred. After incubation of poly(dA) .oligo( [5-:"P]dT) with DNA ligase I (1.5 X IO-' units) under standard reaction conditions, more than 90% of the phosphatase-resistant radioactive material was recovered as 3'-dTMP after analysis by high pressure liquid chromatography. Similarly, incubation ofpoly(rA). oligo( [5'-:'"P]dT) with mammalian DNA ligase I1 ( 3 x lo-' units) allowed the isolation of >90% of the phosphatase-resistant radioact,ive material as 3'-dTMP. These data show that both ligase I and ligase I1 act as DNA ligases, and that ligase I1 generates phosphodiester bonds with the poly(rA). oligo(dT) hybrid substrate. Blunt-cnd Joining of DNA-The T 4 D N A ligase can join blunt-ended DNA fragments (2), and this activity has often been employed for the construction of recombinant. DNA molecules. This function is less efficient than the sealing of single-strand interruptions in DNA, but blunt-end joining can be promoted in reaction mixtures by macromolecular crowding conditions, e.g. by the addition of polyethylene glycol (15). When the mammalian DNA ligases I and I1 were assayed wit.h a blunt-ended DNA substrate in the presence of 17.5% polyethylene glycol 6000 (Fig. 3 ) , ligase I was able to perform this joining reaction (lane 4, while ligase I1 showed no detectable activity (lane 6). The activity of ligase I was blocked by antibodies against the enzyme (lane .5) while T 4 D N A ligase was not similarly inhibited (lanes 2 and 3 ) . Further, the activity of ligase I was proportional to enzyme concentration (up to 1.5 X lo-" units) and approximately linearly dependent on time up to 60 min. Increasing the enzyme concentration 5-fold, decreasing the temperature to 16 "C while increasing the time of incubation, or supplementing the reaction mixture with KC1 (50-200 mM) failed to reveal any detectable activity of ligase I1 with the blunt-ended DNA substrate.
To determine the efficiencies of blunt-end joining at different concentrations of polyethylene glycol 6000, the concentrations ofthis reagent were varied in reaction mixtures (Fig.   4). Mammalian DNA ligase I showed optimal blunt-end joining at 17.5% polyethylene glycol 6000 (lane 5), with lower but still detectable joining occurring at 12.5% (lane 4 ) and 22.5% (lane 7) polyethylene glycol. Blunt-end joining by ligase I was not observed after 1 h a t 37 "C in the absence of added polyethylene glycol 6000 (lane 2 ) but could be detected after incubation at 16 "C for 72 h (-30% of maximal joining). We estimate that in the absence of polyethylene glycol, DNA ligase I performs blunt-end joining of DNA 20-50 times less efficiently than sealing of single-strand interruptions. No detectable blunt-end joining catalyzed by DNA ligase I1 was observed a t a n y polyethylene glycol concentrations (lanes 9-1 4 , either at 37 or 16 "C. We conclude that mammalian DNA ligase I, but not ligase 11, is able to join blunt-ended DNA molecules under our standard assav conditions. Zimmerman and Pfeiffer (15) have reported that a partly purified DNA ligase preparation from rat liver nuclei could seal doublestrand breaks in DNA, and this activity may now tentatively be assigned t,o DNA ligase I.
Ligase Heterogeneity-The number of mammalian DNA ligases has been a controversial matter. The existence of ligases I and I1 as two separate activities (4) has been confirmed by Creissen and Shall (16) and by Chan and Becker (6,11). However, Teraoka and Tsukada (10) have reported that mammalian cells only seem to contain one DNA ligase.
corresponding to ligase I, and these authors and Mezzina et al. (17) have further suggested that different size classes of mammalian ligase might perhaps be ascribed to proteolysis. In contrast, in a more recent publication ( 7 ) , Teraoka and Tsukada have confirmed our finding that antibodies against calf thymus DNA ligase I do not inhibit ligase 11. Yeast cells appear to have a single DNA ligase (18) but this does not necessarily imply a similar situation for mammalian cells; for example, DNA polymerase has not been found in yeast. although it. is widely distributed among higher eukaryotes (19). The availability of separate and specific assay procedures for each of the two mammalian DNA ligase activities. as described here, should help to further clarify the situation. Since only one of the two enzymes may be required for essential replication events, it becomes an interesting possibility that one of the inherited human syndromes associated with retarded strand joining of damaged DNA could be associated with a molecular defect in a DNA ligase (20,21). Acknowledgment-We thank Dr. S. Soderhall for carrying out a preliminary experiment on the joining of hybrid substrates by mammalian DNA ligases.