Specific and Potent Inhibition of NAD+-dependent DNA Ligase by Pyridochromanones*

Pyridochromanones were identified by high throughput screening as potent inhibitors of NAD+-dependent DNA ligase from Escherichia coli. Further characterization revealed that eubacterial DNA ligases from Gramnegative and Gram-positive sources were inhibited at nanomolar concentrations. In contrast, purified human DNA ligase I was not affected (IC50 > 75 μm), demonstrating remarkable specificity for the prokaryotic target. The binding mode is competitive with the eubacteria-specific cofactor NAD+, and no intercalation into DNA was detected. Accordingly, the compounds were bactericidal for the prominent human pathogen Staphylococcus aureus in the low μg/ml range, whereas eukaryotic cells were not affected up to 60 μg/ml. The hypothesis that inhibition of DNA ligase is the antibacterial principle was proven in studies with a temperature-sensitive ligase-deficient E. coli strain. This mutant was highly susceptible for pyridochromanones at elevated temperatures but was rescued by heterologous expression of human DNA ligase I. A physiological consequence of ligase inhibition in bacteria was massive DNA degradation, as visualized by fluorescence microscopy of labeled DNA. In summary, the pyridochromanones demonstrate that diverse eubacterial DNA ligases can be addressed by a single inhibitor without affecting eukaryotic ligases or other DNA-binding enzymes, which proves the value of DNA ligase as a novel target in antibacterial therapy.

Multiple drug resistance among bacterial pathogens is spreading even in developed countries and has made many currently available antibiotics ineffective (1). As a consequence the number of reports on therapy failures increases and treatment costs rise, causing a growing public health problem. Thus, the search for novel antibacterial classes with innovative mechanisms of action is crucial to keep pace with the innate adaptability of the bacterial population. From the information revealed by sequencing more than 80 bacterial genomes, many novel target ideas have emerged in the last decade. However, even classical target areas such as cell wall, protein, or DNA synthesis contain many vital reactions not exploited in antibacterial therapy so far.
DNA ligases are promising target candidates because they are indispensable for many fundamental processes in DNA metabolism including the linkage of Okazaki fragments during replication, recombination processes, and repair pathways requiring resynthesis of DNA (2,3). Their crucial function is emphasized by the fact that eukaryotic cells contain several isoenzymes and that viruses encode their own ligases (3,4). The reaction catalyzed by the DNA ligases, the joining of nicked DNA strands, proceeds in three sequential nucleotidyl transfer reactions (2). The first step is the nucleophilic attack by the active site lysine on the AMP moiety of a cofactor, resulting in a covalent enzyme-AMP intermediate. The AMP is then transferred to the 5Ј-phosphate end of the nicked duplex DNA and finally released when the ligase catalyzes the attack by the adjacent 3Ј-hydroxyl group during the formation of a phosphodiester bond.
One aspect in considering DNA ligase as a potential antibacterial target is the distribution and homology among major bacterial pathogens. Eubacterial DNA ligases are extensively conserved over the entire length of the polypeptides (5). In contrast, if eubacterial and eukaryotic representatives are compared, there is little sequence homology apart from a short KXDG motif around the active site lysine. However, recent investigations reveal structural similarities especially in the AMP-binding region (3,5).
Different cofactor requirements between the distinct ligase families raise the chance to find specific inhibitors directed exclusively against the eubacterial enzymes. Although the cofactor of eukaryotic, archaeal, and viral DNA ligases is ATP, which is degraded to AMP and pyrophosphate in the course of ligation, all eubacteria possess a ligase which uses NAD ϩ for this purpose, the final products being AMP and NMN (3,5). In studies with temperature-sensitive or deletion mutants NAD ϩdependent DNA ligases were shown to be essential for survival in several bacterial species (e.g. LigA in Escherichia coli (6), YerG in Bacillus subtilis (7), and Lig in Staphylococcus aureus (8)).
For human DNA ligase I several natural product inhibitors have been described which mostly intercalate into DNA and which were evaluated for their potential as antitumor agents (9 -11). In contrast, there is only one report on compounds targeting a eubacterial DNA ligase with some specificity. Ciarrocchi et al. (12) demonstrated that derivatives of the antimalaria drug chloroquine inhibited the E. coli DNA ligase LigA with IC 50 values down to the single digit micromolar range. However, human DNA ligase I and the ligase of T 4 -bacteriophage were also affected at 10-fold higher concentrations (12).
Here we present the pyridochromanones as a novel class of potent DNA ligase inhibitors. The compounds inhibited the purified NAD ϩ -dependent enzymes from both E. coli and Streptococcus pneumoniae in the nanomolar concentration range. This is to our knowledge the first demonstration of an inhibitory activity against a DNA ligase from a Gram-positive bacterium. Although the pyridochromanones interfere broadly with eubacterial DNA ligases, human DNA ligase I is not inhibited up to 75 M demonstrating that it is possible to discriminate between NAD ϩ -and ATP-dependent representatives with high specificity.

MATERIALS AND METHODS
Antibacterial Compounds-Pyridochromanones were synthesized at Bayer according to a previously described procedure (13). Ciprofloxacin was also prepared at Bayer, and novobiocin was obtained from Sigma-Aldrich.
Measurement of E. coli DNA Ligase Activity via Detection of AMP Release-In this assay format we detected the AMP released by E. coli DNA ligase LigA upon ligation of oligonucleotides. Adenylate kinase and pyruvate kinase were used to convert the AMP to ATP, which was quantified by the luminescence generated by ATP-dependent firefly luciferase. The nicked duplex template for DNA ligase was obtained by mixing 100 M of each of the 3 oligonucleotides 5Ј-CGAATTCACTGG-CCGTCGTTTTACAACGTCGTGACTGGG-3Ј, 5Ј-ACGGCCAGTGAAT-TCG-3Ј, and 5Ј-CCCAGTCACGACGTTGTAAAACG-3Ј in 10 mM Tris/ HCl, pH 8.0. The mixture was boiled for 10 min and chilled to room temperature. The ligase reaction was performed for 30 min at room temperature in a volume of 30 l containing 0.05 unit of E. coli ligase (Roche), 0.62 M ligase template, and 5 M NAD ϩ in 25 mM Tris/HCl, pH 8.0, 6 mM MgCl 2 , 10 mM (NH 4 ) 2 SO 4 , 0.5 mM EDTA, 0.37 mM dithiothreitol, and 0.067% (w/v) BSA. 1 Subsequently, a conversion mixture was added containing 0.3 unit of adenylate kinase (Sigma-Aldrich), 0.5 unit of pyruvate kinase (Sigma-Aldrich), 750 M CTP, 2.55 mM Penolpyruvate, and 1 mM luciferin (Promega) in 20 l of buffer (25 mM Tris/HCl, pH 8.0, 6 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, and 0.05% (w/v) BSA) and incubation was continued for an additional 7 min. The luminescence reaction was started by adding 71 fmol of luciferase (Promega) in 20 l of the same buffer and was immediately measured in a luminescence reader.
Measurement of E. coli DNA Ligase Activity via Detection of Linked Oligonucleotides-This alternative ligase assay format designated here as "sealed duplex assay" is based on ligation of a fluorescein labeled oligonucleotide to a biotinylated oligonucleotide within a double strand duplex. After denaturation successful ligation was quantified by the amount of fluorescence captured on a streptavidin-coated microtiter plate. 200 l/well of the following oligonucleotides (Eurogentec) were hybridized on StreptaWell HighBind plates (Roche) in H-buffer (5ϫ SSC, 0.1% BSA, and 0.1% Triton X-100) for 90 min at 20°C: 5Ј-fluorescein-AAAATGACCCCC-3Ј (320 pmol), 5Ј-CCCAGACAACGTCG-biotin-3Ј (32 pmol), and 5Ј-CGACGTTGTCTGGGGGGGGTCATTTT-3Ј (32 pmol). Wells were washed with 300 l of ligase buffer (30 mM Tris/HCl, pH 8.0, 10 mM (NH 4 ) 2 SO 4 , 10 mM MgCl 2 , 1.2 mM EDTA, and 1 mM dithioerythritol). Ligation was performed in 150 l of ligase buffer with 0.5 unit of E. coli ligase and 28 M NAD ϩ for 45 min at 20°C followed by denaturation in 0.5% SDS, 0.5 M NaOH for 1 min. After washing the wells twice with 300 l of washing buffer (TBS, 0.05% Tween 20, and 0.1% BSA) fluorescence was measured in 100 l of the same buffer at an excitation wavelength of 485 nm and an emission wavelength of 525 nm.
Measurement of S. pneumoniae DNA Ligase Activity-The pneumococcal DNA ligase gene was amplified by PCR and cloned into NdeI/ BamHI-digested pET16b vector (Novagen). After expression in E. coli BL21DE3 (Stratagene) the N-terminally His-tagged protein was purified according to standard procedures. The AMP release assay was conducted with 0.01 g of the purified protein.
Measurement of the Activity of Human DNA Ligase I-The catalytic domain of human DNA ligase I (comprising amino acids 250 -919) was excised by SacI and KpnI from the recombinant plasmid pBluescript SK Ϫ ::hlig 250 -919 (14), which was provided by D. E. Barnes (Clare Hall Laboratories, Herfordshire, UK). The fragment was cloned into pQE31 (Qiagen) and expressed in E. coli M15 (Qiagen). The N-terminally His-tagged protein was purified by nickel-nitrilotriacetic acid affinity chromatography and eluted with 0.3 mM imidazole according to standard protocols provided by the manufacturer (Qiagen). For measurement of the ligase activity the sealed duplex assay was used. The procedure was the same as described above for E. coli ligase except that ATP (1 mM) was used as a cofactor instead of NAD ϩ . 2.5 g of the purified human ligase I were employed per assay. Measurement of the Activity of DNA Topoisomerase II-DNA gyrase activity was determined by supercoiling of relaxed pUC18 DNA. For preparation of the template pUC18 was relaxed by E. coli topoisomerase I expressed as a N-terminally His-tagged protein from pET28a (Novagen) in E. coli BL21DE3pLysS (Stratagene). The relaxation mixture contained 40 g of pUC18 and 15 g of topoisomerase I in 300 l of buffer (0.5ϫ TBE buffer (pH 8) and 1 M MgCl 2 ) and was incubated for 1 h at 37°C. The protein was removed by phenol/chloroform extraction. The relaxed plasmid was concentrated by ethanol precipitation and solubilized in the same buffer. For the supercoiling assay 100 ng of relaxed pUC18 DNA was incubated with 2 unit of Micrococcus luteus DNA gyrase (Invitrogen) in 35 mM Tris/HCl (pH 7.5), 20 mM KCl, 0.1 mM EDTA, 10 mM mercaptoethanol, 10% (v/v) glycerol, 2 mM spermidine (Sigma-Aldrich), 1 mM ATP, and 0.004% (w/v) BSA for 60 min at 37°C and analyzed in a 1% agarose gel.
Gel Shift Assay-200 ng of plasmid DNA (pBluescript SK ϩ , Stratagene) was incubated with increasing inhibitor concentrations in TE buffer for 30 min at room temperature. Subsequently, the DNA was analyzed in a 1% agarose gel. DAPER dye (Pierce) was used as a positive control.
Ethidium Bromide Displacement Assay-This assay measured the DNA intercalating properties of a compound by its ability to compete with ethidium bromide for DNA binding. Detection of ethidium bromide displacement from DNA is based on the strong loss in fluorescence that occurs upon its detachment from the double helix (15). The assay mixture contained in a volume of 100 l, 6.6 g of salmon sperm DNA (Invitrogen), 5 M ethidium bromide, 35 mM Tris/HCl, pH 8.0, 100 mM NaCl, and 1 mM EDTA. Upon addition of the inhibitor in increasing concentrations the fluorescence of ethidium bromide was immediately detected at an excitation wavelength of 485 nm and an emission wavelength of 612 nm.
Heterogeneous Expression of Human DNA Ligase I and S. aureus DNA Ligase in an E. coli ligA ts Mutant-The recombinant plasmid pBluescript SK Ϫ ::hlig 250 -919 containing the catalytic domain of human DNA ligase I (14) was transformed into E. coli GR501 carrying the chromosomal temperature-sensitive ligA mutation lig251 ts (6) The S. aureus lig gene was amplified from the genome of S. aureus 133 (deposited at the DSM strain collection, Braunschweig, Germany, under the number DSM 11832). Primers were derived from the sequence determined by the Institute for Genomic Research for S. aureus COL (tigr.org). The PCR product was cloned into SacII-and XhoI-digested pBluescript SK Ϫ (Stratagene) and transformed into E. coli GR501. As GR501 encoded no lac repressor, expression of the proteins from pBluescript SK Ϫ occurred constitutively. In growth experiments the strains expressing the human or S. aureus DNA ligases were compared with a control GR501 strain carrying empty pBluescript SK Ϫ without gene insertion. All strains were grown in Isosensitest broth (Oxoid) at 28 or 37°C. For the control strain GR501 28°C represents the permissive and 39°C the restrictive temperature. At 37°C growth was already strongly delayed. Propagation of the plasmids was ensured by the addition of 50 g/ml ampicillin to the overnight culture. In the actual growth experiments the antibiotic was omitted.
Determination of Antimicrobial Activity-MIC values were determined in broth microdilution assays in microtiter plates in a volume of 200 l. Serial 2-fold dilutions of antibacterial compounds were seeded with a final inoculum of 10 4 CFU/ml in the case of S. aureus 133 and B. subtilis 168 trpC2 (16), of 10 7 CFU/ml in the case of E. coli GR501, and of 10 3 CFU/ml in the case of Candida albicans ATCC 200498. Bacteria were grown in Isosensitest broth and C. albicans in yeast nitrogen base broth (Difco), the latter being supplemented with 7.2 g of Na 2 HPO 4 , 3.55 g of KH 2 PO 4 , and 10 g of glucose/liter of medium. After incubation for 18 h for bacteria or 24 h for C. albicans at 37°C in ambient air, MICs were read as the lowest concentrations of compounds that prevented visible microbial growth. In the case of E. coli the medium contained 25 g/ml polymyxin B nonapeptide (Sigma-Aldrich) to facilitate permeation of the inhibitors across the outer membrane.
Time-kill Studies-An exponentially growing culture of S. aureus 133 in Isosensitest broth was treated at an A 600 of 0.2 with increasing inhibitor concentrations. The effect on growth and viability was investigated by monitoring the A 600 and the number of CFU for 5 h after addition of the antibacterial compound. For CFU quantification culture aliquots were serially diluted in phosphate-buffered saline and plated on Isosensitest agar. After incubation for 18 h at 37°C visible colonies were counted.
Determination of the Susceptibility of Chinese Hamster Ovary Cells-In this assay the fluorescent DNA-specific dye DAPI (Sigma-Aldrich) was used as viability marker. Chinese hamster ovary cells were cultivated in RPMI medium (Invitrogen) containing 9% fetal calf serum, 1.8 mM L-glutamine, and penicillin (100 IU)/streptomycin (100 g/ml) (Roche). A number of 5 ϫ 10 4 cells/ml was incubated with increasing concentrations of test compound in a volume of 1 ml in 24-well cell culture plates (Costar) for 24 h at 37°C and 5% CO 2 . After cultivation for an additional 48 h in fresh medium in the absence of inhibitor the culture medium was removed and the cells were stained for 1 h with 0.2 ml of a DAPI solution (5 g/ml in distilled water). Finally, cells were transferred into a white microtiter plate (Dynatech Laboratories), and fluorescence was measured at an excitation wavelength of 350 nm and an emission wavelength of 460 nm.
Microscopy of DAPI-stained B. subtilis-A B. subtilis 168 culture in Isosensitest broth was treated at an A 600 of 0.1 with increasing inhibitor concentrations. At various time points after addition of the inhibitor culture aliquots of 300 l were removed and mixed with 11 l of a DAPI solution (150 g/ml in distilled water). After staining for 15 min on ice the samples were immobilized on an agar-coated microscope slide and analyzed in a fluorescence microscope (Zeiss) at an excitation wavelength of 360 nm and an emission wavelength of 397 nm. For coating the slides were dipped into a solution of 2% agar in distilled water.

Potent Inhibition of Isolated Bacterial DNA Ligases in
Vitro-Based on purified DNA ligase LigA from E. coli we developed an assay to screen for low molecular weight compounds as selective inhibitors of bacterial DNA ligases. The successful ligation reaction was quantified by the amount of AMP released by the enzyme upon joining of two oligonucleotides. The AMP was subsequently converted to ATP by adenylate kinase and pyruvate kinase, and the resulting ATP was finally visualized by luminescence generated by ATP-depend-ent firefly luciferase. High throughput screening and subsequent chemical variation of a promising screening hit led to the pyridochromanones, which inhibited E. coli DNA ligase with IC 50 values in the nanomolar concentration range (Table I).
Similar inhibitory concentrations were obtained in an alternative assay format relying on a totally different readout procedure (Table I). In this second ligation assay, designated here as sealed duplex assay, E. coli LigA was used to ligate a fluorescent oligonucleotide to a second biotinylated oligonucleotide within a double-strand duplex. After disruption of the duplex, successful ligation was quantified by the amount of fluorescence captured via the biotin anchor at a streptavidincoated microtiter plate. The pyridochromanones were equally potent in both assay systems clearly indicating that they are indeed direct inhibitors of the DNA ligase and that the signal reduction observed in the assays is not based on a potential interference with a component of the readout cascade. In addition to using short oligonucleotides as DNA templates we measured the ligation of nicked plasmid DNA. Compound 3, which is depicted as an example, showed beginning inhibition of this reaction at a concentration of 0.1 M (Fig. 1).
All experiments presented so far demonstrate inhibition of E. coli LigA which we chose for our initial screening as it is the prototype of an eubacterial DNA ligase, known for decades and characterized in detail (2). To investigate whether the pyridochromanones interfere also with the activity of a Gram-positive DNA ligase we purified the S. pneumoniae Lig protein after heterologous expression in E. coli and assessed its activity in the "AMP release assay." Submicromolar IC 50 values were observed also for the pneumococcal enzyme (Table I) indicating that the inhibitory activity of the pyridochromanones covers NAD ϩ -dependent DNA ligases from Gram-positive as well as from Gram-negative bacterial genera.
High Specificity for NAD ϩ -dependent Versus ATP-dependent DNA Ligases-With respect to a potential novel antibiotic we were interested only in such inhibitors that targeted specifically the bacterial DNA ligases but not the human counter- The influence of the inhibitors on the activities of NAD ϩ -dependent bacterial DNA ligases from a Gram-positive (S. pneumoniae) and a Gram-negative (E. coli) representative were analyzed and contrasted with the effect on the major ATP-dependent human DNA ligase. Two different assay formats were employed detecting the release of AMP upon successful ligation or the ligated DNA product. n.d., not determined. *, highest concentration tested.
parts. To address this question of specificity we chose the major replicative DNA ligase in humans (DNA ligase I) (17) as an example of a eukaryotic ATP-dependent ligase. Previous studies had shown that the C-terminal region of this enzyme comprising amino acids 250 -919 is sufficient for catalytic activity (14). As this catalytic domain is far more stable in E. coli than the full-length protein, we overexpressed this region in E. coli and employed the purified enzyme in the sealed duplex assay. In contrast to the nanomolar inhibition of the bacterial DNA ligases the pyridochromanones showed IC 50 values for the human enzyme that exceeded the highest concentration tested (75 M, Table I). This finding demonstrates a remarkable specificity for the bacterial NAD ϩ -dependent representatives with a selectivity index of more than 3 orders of magnitude for the most potent compounds.
Pyridochromanones Inhibit DNA Ligase Competitively with Respect to NAD ϩ -To elucidate the molecular mechanism by which the pyridochromanones inhibit DNA ligase, we determined their influence on the ligase reaction employing the AMP release assay. In the absence of the inhibitor we determined a K m of 4 M for NAD ϩ and of 0.1 M for the DNA template, which is in good agreement with previously reported data (18). When in our inhibition studies the amount of NAD ϩ was increased from 0.5 to 64 M in the presence of varying concentrations of compound 3 (0 -400 nM) and saturating DNA amount (0.67 M) the kinetics indicated competitive inhibition as visualized in a double-reciprocal plot (Fig. 2, A and B). The linear regression using the apparent K m value leads to a K i of about 0.094 M (Fig. 2C). In contrast, the binding of the DNA under saturating NAD ϩ concentration (64 M) in the presence of the inhibitor seems to follow a non-competitive binding mode because V max remains reduced even at high DNA concentrations. These findings strongly suggest that the binding site of the pyridochromanones overlaps at least partially with that of the cofactor NAD ϩ .
Studies with a regulated NAD ϩ mutant of B. subtilis 168 support this interpretation. In the strain AL 811 (trpC2, ⌬nadD, ⌬thrC::xylR-nadD-spec R ) the nadD gene coding for nicotinic acid mononucleotide adenylyltransferase, which is essential for NAD ϩ biosynthesis (19), was deleted at the original locus, and a chromosomal copy was expressed under the control of the xylose promoter. When AL 811 was grown in LB medium supplemented with 0.25% xylose ensuring full expression of NadD, the MIC for compound 3 was 6 g/ml, whereas under conditions of promoter catabolite repression in the presence of 0.2% glucose the MIC dropped to 1.5 g/ml. The observation that the lowered cytoplasmic NAD ϩ level enhances antibacterial activity supports the notion of competition with the cofactor also on the cellular level.
Several experiments indicate that there is no general interaction of the pyridochromanones with DNA. In a gel shift assay plasmid DNA was incubated with rising inhibitor concentra-tions followed by analysis of its electrophoretic mobility. The pyridochromanones did not effect the migration behavior of the DNA up to the highest concentration tested (25 M), whereas the intercalating DAPER dye (20) had an IC 50 below 5 M. In addition, the ligase inhibitors were not able to displace ethidium bromide from DNA even at a concentration of 250 M representing a 50-fold molar excess over ethidium bromide.
Consistent with the result that no DNA interaction was observed, the pyridochromanones did not interfere with the activity of DNA gyrase from M. luteus, which we chose as an example of other DNA processing enzymes. The supercoiling activity of DNA gyrase was not affected up to 200 M, whereas an IC 50 of 60 nM was determined for novobiocin, which was employed as reference inhibitor (21). The fact that bacterial gyrase as well as human ligase are not hampered by the pyridochromanones demonstrates that the basis of the activity of this compound class is the targeted inhibition of NAD ϩ -dependent DNA ligases but not general, nonspecific disturbance of the DNA topology.
Inhibition of DNA Ligase Is the Basis for Bacterial Cell Death-The pyridochromanones possess substantial antibacterial activity directed primarily against Gram-positive bacteria. Growth inhibition of B. subtilis was already described above. Another example is the important human pathogen S. aureus for which MIC values in the low g/ml range were obtained (Table II). As exemplified for compound 3 the activity was bactericidal with a concentration-dependent reduction in the number of viable bacteria, e.g. a decrease of CFU by 2 log units over 2 h at a concentration of 2 times the MIC (Fig. 3). Intact E. coli cells were not affected but they became susceptible when the integrity of the outer membrane was disturbed by the permeabilizing agent polymycin B nonapeptide (22). This result implies that once the outermembrane barrier is overcome the target is inhibited also in E. coli. In accordance with the result that no inhibition of human DNA ligase I was detected, the growth of the fungus C. albicans was not impaired (Table  II). Furthermore, no cytotoxicity was observed for Chinese hamster ovary cells even though the staining procedure chosen to determine cell viability relied on the DNA-specific dye DAPI (23) and was thus especially sensitive to DNA damage (Table II).
A novel antibacterial compound under investigation may prevent the activity of a desired isolated enzyme in a biochemical in vitro assay and may also inhibit the growth of intact bacteria. However, it is still conceivable that the two activities are unrelated because the antibacterial effect is based on a rather nonspecific interference with a multitude of essential cellular functions. Such a global disturbance of cell viability may be caused by disruption of membrane integrity, by DNA intercalation, or by elevated chemical reactivity of the compound. To exclude such undesired effects in the case of the pyridochromanones we determined their influence on DNA ligase in the intact bacterial cell. For this purpose we used the E. coli DNA ligase mutant GR501 harboring the temperaturesensitive lig251 mutation in LigA (6). Although this strain grew well at 28°C the increase in cell mass was strongly delayed at 37°C (Fig. 4). This growth behavior reflects the ligase activity previously measured in this mutant. Although the mutated ligase was fully functional in cell extracts at 25°C, no residual activity was detected at 40°C (6). Heterologous expression of plasmid-encoded human DNA ligase I was described previously to restore growth of GR501 at restrictive temperatures (14) and was reproduced by us (Fig. 4). An equal or even slightly better rescue effect was obtained by expressing S. aureus DNA ligase in the E. coli mutant (Fig. 4). Together these three strains form a convenient system to determine the effect of an inhibitor of bacterial ligases in the cellular environment. Although non-complemented GR501 was highly susceptible for the pyridochromanones at elevated temperatures, the antibacterial activity of this compound class was completely abolished in the presence of the human ligase (Table II). That the bactericidal effect was totally overridden by the expression of the human enzyme proves that inhibition of DNA ligase is indeed the cause for bacterial cell death and confirms again that the function of the human ligase is not impaired. In contrast, S. aureus ligase did not rescue the E. coli mutant from the activity of the inhibitors (Table II), demonstrating that the enzyme from S. aureus is also targeted by the pyridochromanones. The observed decrease in susceptibility of the strain supplemented with S. aureus ligase in relation to the noncomplemented GR501 (e.g. 0.1 versus 6 g/ml for compound 3, respectively) could be assigned to the low residual activity of the temperature-sensitive E. coli ligase compared with the high copy number of overexpressed S. aureus ligase.
Also in whole B. subtilis cells we obtained proof that the ligase function is disturbed. Microscopic investigation of a pyridochromanone-treated culture (Fig. 5) showed the formation of filaments in accordance with induction of the SOS cascade and

TABLE II
Antibacterial activity and effect on growth of eukaryotic cells MIC values were determined by broth microdilution for the bacterial strains E. coli and S. aureus as well as well as for the fungus C. albicans. E. coli GR501 is impaired in ligase activity, a defect that is restored by the overexpression of human DNA ligase I or S. aureus ligase. Polymyxin B nonapeptide (25 g/ml) was added to the E. coli growth medium to facilitate passage of the inhibitors across the outer membrane. Proliferation of eukaryotic CHO cells was quantified by the fluorescence of the DNA-specific fluorophor DAPI.
the resulting inhibition of cell division (24). Staining of the DNA in these cells with the fluorescent dye DAPI revealed massive chromosome degradation, consistent with impaired joining of the Okazaki fragments. The phenotype induced by the pyridochromanones in this DNA staining experiment resembled the one observed in the presence of quinolones such as ciprofloxacin (Fig. 5), known to cause DNA strand breaks and degradation after arresting gyrase on the DNA in a covalent ternary complex (25). Induction of the SOS response by pyridochromanones was confirmed by an experiment with B. subtilis YB3001. In this strain the lacZ reporter gene is coupled to the recA promoter and integrated into the chromosomal amyE locus (26). When a X-gal-containing agar plate was overlaid with a suspension of this strain, expression of ␤-galactosidase produced a blue halo from the chromogenic substrate around a drop of a compound 3 solution (data not shown).

DISCUSSION
NAD ϩ -dependent DNA ligase has been discussed as a potential antibacterial target for a couple of years on the basis of its specific cofactor requirement and major sequence discrepancies between prokaryotic and eukaryotic representatives. However, proof had been pending that it is possible to find high affinity inhibitors of bacterial DNA ligases that do not interfere with their eukaryotic counterparts. The pyridochromanones presented here provide that evidence. NAD ϩ -dependent DNA li-gases from Gram-positive and Gram-negative bacterial species are inhibited in the nanomolar concentration range, although the ATP-dependent human DNA ligase I is not affected. Our study demonstrates that such potent inhibition of this essential bacterial enzyme is sufficient to inflict bacterial cell death. A major physiological effect on the bacterial cell is DNA degradation, which in turn triggers induction of the SOS system and cell division blockade. The pyridochromanones inhibit bacterial growth in the low g/ml range and possess bactericidal activity of considerable strength. These results clearly demonstrate the value of NAD ϩ -dependent DNA ligase as a novel target in antibacterial therapy.
Such NAD ϩ -dependent enzymes were long believed to be the only DNA ligases present in eubacteria, but as more bacterial genomes are mapped it becomes obvious that some species encode sequences for additional ATP-dependent forms (5,27). The reason why some bacteria have both ATP-and NAD ϩ -dependent DNA ligases and others do not has yet to be elucidated. Phylogenetic comparison hints to horizontal gene transfer from viruses, eukaryotes, or Archaea (5). With respect to their biological role, involvement in specific cellular processes or growth phases is discussed as it was reported for the eukaryotic isoenzymes (3, 5). One example is B. subtilis where two ATP-dependent homologues co-exist beside the NAD ϩ -dependent YerG (7). Our study shows that the pyridochromanones prevent the growth of B. subtilis and lead to DNA degradation within the cell. Therefore, also in this case, the NAD ϩ -dependent ligase is effectively inhibited, and the ATP-dependent variants cannot compensate for this defect. It is also noteworthy that some bacteria have more than one NAD ϩ -dependent DNA ligase. In the E. coli genome ligB was identified by virtue of its sequence similarity with the primary NAD ϩ -dependent ligA, and ligB homologues were also detected in Yersinia and Salmonella (28,29). For E. coli LigB the NAD ϩ -dependent nick-joining activity was demonstrated, but the specific activity was only 1% that of LigA (28). Thus, the precise function of LigB still has to be elucidated; however, LigB cannot compensate for the loss of LigA.
Different cofactor specificity is a major distinction between eubacterial and eukaryotic DNA ligases and one rationale for why it seems feasible to find specific inhibitors with a preference for one of the two separate ligase families. In our competition studies, the result that the highly specific pyridochromanones bind competitively with respect to NAD ϩ and not with respect to DNA fits into this picture. The identification of such a type of inhibitor was facilitated by the screening format employed, because the AMP release assay was performed in the range of the K m for NAD ϩ (5 M) and at a saturating DNA concentration (0.67 M).
The difference between NAD ϩ and ATP lies in the presence of the nicotinamide ribose monophosphate moiety whereas the AMP portion occurs in both cofactors. So far, no crystal structure has been published for a eubacterial DNA ligase in a complex with a complete NAD ϩ molecule, which would show unequivocally the position of the NMN moiety. However, struc-tural information is available for the adenylation domain of the NAD ϩ -dependent ligase from Bacillus stearothermophilus (30) and the complete NAD ϩ -dependent enzyme from Thermus filiformis with AMP covalently attached (31). With respect to ATP-dependent ligases, crystal structures were obtained for the enzyme from bacteriophage T7 with bound ATP (32) and for the chlorella virus ligase-AMP intermediate (33). These structures show the overall architecture of the different enzymes as well as the location and structure of the AMP-binding pocket. Each of the four DNA ligases have in common that they contain two domains divided by a deep cleft, an N-terminal adenylation domain in which the cofactor-binding site is located and a C-terminal DNA-binding domain. Despite the lack of sequence homology between the two families the AMP-binding pocket shows remarkable structural similarity, and most of the few amino acids that are conserved among the two classes make important contacts with the AMP moiety. In the NAD ϩ -dependent ligase from T. filiformis the AMP is located in a pocket between two ␤-sheets of subdomain 1b and the covalent contact is established to Lys 116 (31). This pocket is spacious enough for the AMP moiety but is too small for the NMN portion, which is believed to extend along the surface of the adenylation domain. In a recent site-directed mutagenesis study with E. coli LigA several conserved residues were identified in subdomain Ia, which are crucial for the interaction with the nicotinamide nucleoside. The authors propose that this subdomain, which occurs only in NAD ϩ -dependent ligases, performs a conformational change upon binding of the cofactor that puts NAD ϩ in the proper orientation for hydrolysis of the phosphoanhydride bond (34).
To obtain a first hint on the molecular-binding mode of the pyridochromanones, we selected for a highly resistant colony (MIC Ն 60 g/ml) of S. aureus 133 on compound 3-containing agar plates. Sequencing of the lig gene revealed a transition of guanine for adenine in position 1117 resulting in an amino acid exchange of Ala 373 for Thr. This alanine residue is widely conserved among Gram-positive and Gram-negative eubacteria, although a few deviations exist such as exchange for Ser in Helicobacter pylori and Clostridium perfringens or even Arg in Borrelia burgdorferi. The corresponding residues in B. stearothermophilus and in T. filiformis are Ala 375 and Ala 378 , respectively, which reside within the so-called oligomer-binding fold, a hinge region that connects the N-terminal and C-terminal domains. In the T. filiformis ligase structure Ala 378 is located directly opposite the entrance of the AMP-binding pocket (Fig.  6). It is well conceivable that an inhibitor which occupies this position hampers the access of the cofactor to its binding site with concomitant maintenance of specificity.
In summary, the example of the pyridochromanones clearly demonstrates that by direct interaction with a region which differentiates NAD ϩ -dependent and ATP-dependent DNA ligases rather than by mere intercalation-like interaction with the DNA substrate, it is possible not only to selectively target the eubacterial DNA ligases with high potency but also to avoid the interference with the activity of other DNA processing enzymes.