Structural Basis for Stable DNA Complex Formation by the Caspase-activated DNase*

We describe a structural model for DNA binding by the caspase-activated DNase (CAD). Results of a mutational analysis and computational modeling suggest that DNA is bound via a positively charged surface with two functionally distinct regions, one being the active site facing the DNA minor groove and the other comprising distal residues close to or directly from helix α4, which binds DNA in the major groove. This bipartite protein-DNA interaction is present once in the CAD/inhibitor of CAD heterodimer and repeated twice in the active CAD dimer.


Expression and Purification of Recombinant Proteins-Recombinant
DFF was produced by co-expressing GST-tagged CAD variants together with human ICAD-L (DFF45) in E. coli BL21Gold(DE3) cells harboring the two compatible expression vectors pGEX-2T-CAD and pACET-DFF45. For each DFF variant, recombinant protein from 2 liters of culture was purified via glutathione affinity chromatography. GST-tagged DFF was bound to glutathione-Sepharose 4-B beads in buffer A (20 mM HEPES-NaOH, pH 7.4, 100 mM NaCl, 5 mM dithiothreitol, 1 mM EDTA, 10% glycerol, 0.01% Triton X-100) supplemented with 2.5 mM MgCl 2 , treated with DNase I (20 units) to digest contaminating DNA, and subsequently washed intensively with buffer A containing 750 mM NaCl. Purified complex was eluted from the glutathione-Sepharose 4-B beads using buffer A supplemented with reduced glutathione to 20 mM. Initially, protein was purified via GST affinity chromatography without DNase I and high salt treatment; it was further purified by anion-exchange chromatography over a Mono-Q HR5/5 column by applying a linear gradient in 15 ml of 100 -400 mM NaCl in buffer A at a flow rate of 1 ml/min. Eluted DFF was concentrated using Centriplus-30 ultrafiltration units. Caspase-3 was expressed and purified as described previously (27).
Generation of Arginine and Phenylalanine Mutants-Arginine mutants of CAD were prepared as described previously (28). The following primers were used to introduce the desired mutations: R151A (AGGTTTG- FIGURE 1. Conserved arginine residues contribute to DNA binding and cleavage by CAD. A, alignment of the C-terminal region of CAD proteins from Mus musculus (mCAD, GenBank TM accession number AB009377), Rattus norvegicus (rCAD, GenBank TM accession number AF136598), Homo sapiens (hCAD, GenBank TM accession numbers AF064019, AF039210, AB013918), Gallus Gallus (chCAD, GenBank TM accession number AF406761), Danio rerio (fCAD, GenBank TM accession number AF286179), and Drosophila melanogaster (dCAD, GenBank TM accession number AF149797, AB036773). The black and gray vertical bars highlight conserved arginine, phenylalanine, and lysine residues investigated in this study. The black horizontal bar represents helix ␣4. The two ␤-strands of the ␤␤␣-Me-finger active site motif are shown as black arrows. Important active site residues (Asp 262 , His 263 , and His 308 ) are highlighted by asterisks. B, preparation of arginine variants of the caspase-activated DNase. Recombinant proteins (GST-CAD/DFF45) were expressed in E. coli and purified by glutathione affinity chromatography. Analysis by SDS-PAGE of the free nuclease variants released from the DFF complex through activation by caspase-3. C, DNase activity of the arginine variants of CAD as determined in plasmid DNA cleavage assays. GGATCGCCGGCTACCTAAGAGAGGT), R212A (CAGCTATTTCGA-CGCAGGCGCCGAGGCCAGCAG), R250A (CCCTATGGCAACGCAG-AGTCGCGAATCCTCTTCAGTAC), R269A (TATAATAGAGAAGAA-GGCAACCGTTGTACCCACGCTG), F152A (AAGGTTTGGAGTCGA-GAGCGCGCAATAAGTCGGGCTA). The codon for alanine is shown in bold type, and restriction sites co-introduced as markers are underlined.
Enzyme Activity Assays-To measure enzymatic activity of CAD variants, DFF was treated with recombinant caspase-3 and the free nuclease incubated at 37°C with plasmid DNA as substrate in a buffer consisting of 20 mM HEPES-KOH, pH 7.4, 100 mM NaCl, 2 mM EDTA, 10% glycerol, 0.01% CHAPS, and 5 mM MgCl 2 . After certain time points, aliquots were taken from the reaction mixture and DNA cleavage analyzed by agarose gel electrophoresis. To quantitate enzymatic activity the intensity of the band representing supercoiled plasmid DNA normalized to the linear and open circular forms was measured with a BioDocAnalyze gel analysis system (Biometra) using the implemented software. The disappearance of supercoiled plasmid DNA over time was used to determine relative cleavage activities from plots of band intensity versus time.
Electrophoretic Mobility Shift Assays (EMSA)-EMSA with plasmid DNA and recombinant DFF was carried out as described previously (17). EMSA with PCR-derived DNA was performed by incubating a 273-bp PCR product at a concentration of 139 nM for 10 min with indicated amounts (Fig. 6) of DFF variants at 23°C in shift buffer (20 mM HEPES-KOH, pH 7.4, 100 mM NaCl, 2 mM EDTA, 10% glycerol, 0.01% CHAPS). Bound complexes were analyzed by electrophoresis on 4% native polyacrylamide gels followed by ethidium bromide staining and visualization.
Poly(ADP-ribose) Binding Assay-To analyze the influence of poly-(ADP-ribose) (PAR) on the DNA binding properties of CAD, we performed EMSA in the presence of PAR with a 44-bp double-stranded DNA fragment 32 P end labeled on one strand. DFF or caspase-3-activated DFF at a concentration of 1 M was incubated in shift buffer with the indicated amount ( Fig. 2) of PAR for 10 min prior to adding DNA to a final concentration of 10 nM (ϳ30,000 cpm). Aliquots of the binding reactions were separated by electrophoresis on 4% native polyacrylamide gels and analyzed by autoradiography.

Conserved Arginine Residues Contribute to DNA Cleavage by CAD-
Arginine residues are frequently involved in substrate binding by proteins interacting with nucleic acids (29 -31). Primary sequence analysis of CAD proteins from diverse species suggests the presence of six conserved arginine residues (Arg 151 , Arg 166 , Arg 168 , Arg 212 , Arg 250 , and Arg 269 ) in the C-terminal catalytic domain of CAD (Fig. 1A). To investigate the contribution of these residues to DNA binding and cleavage by CAD, we produced CAD variants with substitution for alanine of these conserved residues, expanding previously performed mutational analyses of amino acid residues critical for activity of the enzyme (11,27,32). Activity assays with a plasmid DNA substrate and recombinant proteins purified from E. coli by glutathione affinity chromatography revealed that all arginine variants of CAD tested here exhibit reduced cleavage activities, albeit to different extents, and thus contribute directly or indirectly to DNA binding and/or cleavage by this enzyme (Fig. 1, B and C). One of these conserved arginine residues (Arg 269 ) is located at the ␤␤␣-Me-finger active site motif of CAD, two residues (Arg 250 and Arg 212 ) are involved in homodimerization of CAD or located in spatial proximity to the N-terminal end of helix ␣4 (residues Ser 156 -Ser 179 ), respectively, and three of these residues (Arg 151 , Arg 166 , and Arg 168 ) belong to either a loop region preceding helix ␣4 or helix ␣4 itself (Fig. 1A), suggesting that this structural element plays an important role in DNA binding and/or cleavage by CAD (10,17).
PAR Interferes with DNA Binding by CAD and the CAD⅐ICAD Complex-We have shown previously that DFF (CAD⅐ICAD) binds to DNA, though in the heterodimeric DFF complex the nuclease is inactivated by the inhibitory subunit (17). Given that a single amino acid substitution in the nuclease subunit (Lys 155 3 Gln) abolishes DNA binding by DFF and the free nuclease itself, we proposed that the nuclease subunit alone is responsible for DNA binding by DFF (17). In addition to the catalytic center of CAD, which consists of a ␤␤␣-Me-finger motif and very likely contacts DNA via the minor groove, a structural analysis of the active CAD homodimer suggested that helix ␣4 is in a favored position to bind DNA via the major groove (10). As shown recently, this helix ␣4 overlaps with a conserved poly(ADP-ribose) binding motif (residues Arg 151 -Arg 172 in CAD) that allows modulation of CAD activity by non-covalent interaction with PAR ( Fig. 2A) (33). PAR is known to bind to specific domains in DNA damage checkpoint proteins (34 -36). We have used this interaction of CAD with PAR to determine whether the DNA binding ability of not only the free nuclease, as shown earlier, but also DFF (CAD⅐ICAD) can be modulated by PAR. As a result, competition assays in which DFF and caspase-3-treated DFF were incubated with PAR prior to DNA binding revealed that PAR interferes with the ability of free CAD and CAD in the DFF complex to induce a mobility shift in the DNA (Fig. 2B). To interfere with DNA binding by DFF as compared with the free nuclease a higher concentration of PAR was needed, indicating that CAD in complex with the inhibitory subunit ICAD-L (DFF45) probably binds PAR more weakly than free CAD. These results suggest that helix ␣4 is indeed part of a DNA binding site accessible in free CAD as well as CAD bound to its inhibitor in the DFF complex.
Models of CAD⅐DNA Complexes-To gain insight into the structural basis for stable DNA complex formation by CAD, we modeled CAD⅐DNA complexes by superimposing the active site motifs of ColE7 and Vvn nuclease from DNA co-crystal structures with the active site of CAD (Fig. 3). The structural analysis yielded two models of CAD binding to DNA, with differences arising from dissimilar conformational changes induced in the DNA by the two bacterial nucleases. Although ColE7 does not produce major distortions in the DNA, Vvn nuclease widens the minor groove and bends the DNA toward the major groove at an angle of ϳ20° (21,22). We also modeled the CAD⅐DNA complex with DNA from the ColE9⅐DNA⅐Mg 2ϩ ternary complex (23). Given the high similarity between ColE7 and E9 this yielded a model that, as expected, is highly similar to the one obtained with DNA from the ColE7⅐DNA complex (data not shown). The electrostatic potential distribution on the DNA binding surface of CAD suggests that DNA might indeed be bent toward the major groove, as seen in the Vvn⅐DNA complex (data not shown). As expected, in both models the active site motif of CAD faces the DNA minor groove, and helix-␣4 nicely aligns with the DNA major groove.
Selection, Purification, and Stability of CAD Variants-To support our predicted model for CAD⅐DNA complexes and to analyze the CAD-DNA interaction in more detail we purified from E. coli an assortment of CAD variants with substitution of the conserved arginine residues described above and of selected variants (K155Q, K310Q, and H313N) described previously (11,32). In addition we exchanged Phe 152 of CAD to alanine, because we suspected this highly conserved amino acid residue forms contacts to the DNA together with its neighboring residue Arg 151 from a loop region preceding helix ␣4 (Fig. 4A). Together, these variants carry amino acid substitutions at conserved residues that are either part of the ␤␤␣-Me-finger active site motif of CAD or located DNA Binding by CAD DECEMBER 16, 2005 • VOLUME 280 • NUMBER 50 FIGURE 2. Interference of DNA binding by DFF and free CAD with PAR. A, caspase-activated DNases contain a PAR binding consensus sequence (for sequence descriptions refer to Fig.  1). The residues comprising the PAR binding consensus sequence belong to the N-terminal part of helix ␣4. They overlap with a surface area that is involved in DNA complex formation by CAD. B, DFF and free CAD were incubated with indicated amounts of poly(ADP-ribose) (PAR) prior to binding to a 44-bp radiolabeled double-stranded oligodeoxyribonucleotide. Binding reactions were separated on 4% polyacrylamide gels and subsequently analyzed by autoradiography. PAR interferes with DNA complex formation by free CAD and the DFF (CAD⅐ICAD) complex. This suggests that helix ␣4, which overlaps with the PAR binding motif of CAD (33,36), is crucial for stable DNA complex formation.
distal from the active site in the region around helix ␣4 that largely overlaps with the PAR binding region (compare Figs. 2A and 4A).
Because some of the DFF complexes with certain CAD variants when purified by ion-exchange chromatography over a Mono-Q column as described previously turned out to be unstable, we established a different protocol for DFF purification: DFF bound to glutathione-4B-beads was treated with DNase I and then extensively washed with high salt buffer to get rid of contaminating DNA that would otherwise interfere with the gel retardation analysis (17). Using this protocol the majority of desired DFF complexes could be produced, with the exception of the DFF N-F152A variant (carrying the nuclease subunit with Phe 152 exchanged to Ala), which could not be obtained as a DNA-free complex irrespective of the purification protocol applied (Fig. 4B). Intriguingly, most CAD variants that turned out to be proficient in forming stable DNA complexes (see below) could be readily purified by ion-exchange chromatography over a Mono-Q column, whereas those variants that turned out to be deficient in stable DNA complex formation caused dissociation of the DFF complex on the Mono-Q column, suggesting a correlation between the stability of the variant DFF complexes during the purification procedure and their ability to form stable DNA complexes. All of the nuclease variants produced showed little if any plasmid cleavage activity, with the notable exception of the R250A variant (Fig. 4C).
Variants of CAD in Complex with ICAD (DFF) and as the Free Enzyme Behave Similarly in the EMSA-To test the DNA binding ability of the selected CAD variants in complex with ICAD and as the free enzyme, we first carried out EMSA with DFF (CAD⅐ICAD) and with the free nuclease released from DFF by caspase-3 treatment. To this end, DFF with mutated nuclease subunits at a final concentration of 5 M was incubated with plasmid DNA and the binding reactions analyzed by agarose gel electrophoresis (Fig. 5A, upper panel). In parallel, DFF complexes with mutated nuclease subunits were treated with recombinant caspase-3 to release the nuclease, and DNA binding reactions were analyzed as above (Fig. 5A, lower panel). Generally, in the assay DFF complex and the free nuclease behave similarly with only slight differences in the appearance of the protein⅐DNA complexes, suggesting that ICAD does not interfere with the function of residues involved in stable DNA complex formation by CAD.
As seen from Fig. 5A, only one of the selected variants (H313N) displays wild type-like DNA binding activity in the EMSA with plasmid DNA, whereas all other mutants exhibit a reduced ability to induce a mobility shift in the DNA, albeit to different extents. Although the variants R166A, K310Q, and R269A are still able to shift plasmid DNA in the assay, the variants R151A, K155Q, R168A, and R212A do not induce a detectable mobility shift in the DNA under the conditions applied, with the exception of R250A, which leads to a slight mobility shift in the DNA. As shown above (Figs. 1C and 4C), most of the nuclease variants, as expected, show little if any plasmid cleavage activity, with the exception of the R250A variant. This variant cleaves plasmid DNA with reasonable activity, even at nanomolar enzyme concentration. Because the R250A variant is the only variant tested so far that readily cleaves DNA but fails to induce a strong mobility shift at the concentrations used above, we prepared larger amounts of this variant and tested its DNA binding activity at concentrations exceeding 5 M. In these experiments the variant R250A induces a moderate mobility shift in plasmid DNA at relatively high concentrations compared with the wild type enzyme (Fig. 5B).
Relative DNA Binding Activities of CAD Variants in Complex with ICAD-All the nuclease variants analyzed so far show diminished DNA cleavage activity, and some also lost the ability to induce a mobility shift in a plasmid DNA substrate. To gain information on the relative binding activities of the CAD variants, we used an assay that allowed us to obtain more quantitative data on DNA complex formation by CAD. We carried out EMSA with DFF variants and a 273-bp PCR-derived doublestranded DNA substrate (Fig. 6A). Analysis of the DNA binding data in conjunction with the localization of the amino acid residues exchanged revealed that the variants fall into three groups. The first group comprises variants that are deficient in DNA cleavage yet proficient in inducing a mobility shift in the DNA (variants with amino acid substitution of Arg 166 , Arg 296 , Lys 310 , and His 313 ). This behavior can be con-  DECEMBER 16, 2005 • VOLUME 280 • NUMBER 50 sidered typical for variants mainly affected in catalysis. The second group consists of variants that are deficient in DNA binding as well as DNA cleavage (variants with amino acid substitution of Arg 151 and Arg 168 ), which can be considered typical for mutants mainly affected in substrate binding. Phe 152 most likely also contacts DNA directly; however, exchanging this residue in CAD leads to a highly unstable variant that cannot be analyzed by EMSA. The third group comprises variants with substitution of residues probably contributing to the stability of CAD, thus most likely indirectly influencing its DNA binding and cleavage abilities (Lys 155 , Arg 212 , and Arg 250 ) (Fig. 6B). Together these residues form a bipartite DNA binding surface and a structurally important region responsible for stable DNA complex formation as well as DNA cleavage by CAD (Fig. 7).

DISCUSSION
In the present study we have reported the modeling of CAD⅐DNA complexes accompanied by a mutational and biochemical analysis of the DNA binding activities of selected CAD variants. On the basis of their DNA binding properties and available structural data, the variants investigated here can be sorted into three groups, those with substitution of active site residues having no influence on stable DNA binding and those with substitution of residues close to or directly from helix ␣4, being directly or indirectly important for stable DNA complex formation, respectively. Indeed, these different categories of amino acid residues reflect distinct functional regions of a bipartite DNA binding surface important for stable DNA complex formation and DNA cleavage by CAD as well as a structurally important region that contributes indirectly to stable DNA complex formation and DNA cleavage by the enzyme.
One part of the bipartite DNA binding surface encompasses the active site of CAD with a ␤␤␣-Me-finger motif that faces the DNA minor groove (Figs. 3 and 7). In contrast to other ␤␤␣-Me-finger nucleases, CAD aligns a loop rather than an ␣-helix with the DNA minor groove. This loop contains the active site residues His 308 , His 313 , and Lys 310 involved in cofactor binding and substrate positioning (10,11,32). Also included in this part of the bipartite DNA binding surface is residue Arg 269 , conserved in all CAD proteins (Fig. 1A). Variants with substitution of amino acid residues from this region, K310Q, H313N, and R269A (and also H263N characterized previously, Ref. 17), retain the ability to form stable DNA complexes yet show little DNA cleavage activity, suggesting that these surface-exposed residues from the active site are not required for stable DNA complex formation by CAD. However, different from ColE7 and Vvn nuclease, CAD has established pronounced interactions with the DNA major groove via a second DNA binding region. This region, which locates distal to the active site, represents the second part of the bipartite DNA binding surface and largely overlaps with the previously identified PAR binding site (Figs. 2A and 7B) (33). It includes residues nearby or directly from helix ␣4 of CAD that nicely align with the DNA major groove (Figs. 3 and 7). Conserved arginine residues from helix ␣4 (Arg 166 and Arg 168 ) span the major groove, with the side chain of Arg 166 being very close to the  active site and contacting the phosphodiester backbone with the scissile bond and the side chain of Arg 168 pointing away from the active site, contacting the phosphodiester backbone of the opposite strand (Fig. 7A). Upon amino acid substitution of Arg 166 or Arg 168 , CAD variants display drastically reduced DNA cleavage activities. However, variant R166A, in contrast to variant R168A, retains the ability to form stable DNA complexes, suggesting that Arg 166 falls into the class of an active site residue, whereas Arg 168 is indeed required for stable DNA complex formation. Two other conserved surface-exposed amino acid residues (Arg 151 and Phe 152 ), located farthest away from the active site in a loop and a short ␤-strand directly preceding helix ␣4, are also part of this second DNA binding region and thus critical for stable DNA complex formation by CAD. Unfortunately, the likely role of Phe 152 in DNA binding could not be analyzed because the F152A variant was unstable. Considering the correlation we observed between the stability of CAD variants during the purification procedure and their ability to form stable DNA complexes, it seems very likely that Phe 152 together with its neighboring residue Arg 151 directly contacts the DNA. Finally, the resi-dues Lys 155 , Arg 212 , and Arg 250 represent a group of residues that define a structurally important region in CAD. Whereas Lys 155 is buried in the protein and located at the N-terminal end of helix ␣4, Arg 212 and Arg 250 are surface-exposed residues but no part of the DNA binding surface of CAD. These residues are most likely indirectly involved in DNA binding and seem to play an important structural role in stabilizing helix ␣4 (Lys 155 and Arg 212 ) as well as supporting homodimerization of CAD (Arg 250 ) (Fig. 7) (10).
Interestingly, exchange of Arg 250 that is located at the dimerization interface of CAD influences the DNA binding and cleavage activity of CAD. The effect seen upon substitution of this conserved arginine residue is likely to be indirect because it is too far away from the DNA substrate. It could arise from structural perturbation of the DNA binding surface and/or the dimerization interface of CAD. Provided that disturbing the homodimerization of the enzyme is responsible for the drop in DNA binding and cleavage by CAD, our finding would support the hypothesis that preventing homodimerization of CAD by ICAD is the principle mechanism of inhibition of this nuclease (10). It would also mean that dimerization of CAD already occurs at pico-to nanomolar Variants fall into three groups, those with active site residues having no influence on stable DNA binding and those with residues close to or directly from helix ␣4, being either directly involved in stable DNA binding or being structurally important and thus indirectly influencing DNA complex formation.
concentrations because recombinant CAD is reasonably active at these concentrations.
With its dimeric structure and its ability to form stable DNA complexes, CAD on the one hand significantly differs from other nonspecific nucleases and instead displays similarities to sequence-specific nucleases such as restriction enzymes or homing endonucleases (16,17,37). Whereas the latter mostly act as dimers or pseudo-dimers and exhibit complex protein⅐DNA interfaces, the nonspecific nucleases typically act as monomers or dimers with independent active sites. They usually form only a minimum of contacts sufficient to bring about DNA cleavage, at the same time avoiding specific DNA complex formation (14,19,38). On the other hand, different from restriction enzymes or homing endonucleases, the stable DNA complexes formed by CAD are nonspecific and thus enable the enzyme to degrade DNA, still without selectivity, a prerequisite for the physiological function of the enzyme, which is to cut any accessible linker DNA it engages. Because CAD forms stable DNA complexes in the presence of its inhibitor ICAD, it is tempting to speculate that the biological function of stable DNA complex formation by the CAD⅐ICAD complex is to bind the DNA in apoptotic cells prior to nuclease activation by caspase-3 in order to transform the proximity to its substrate into highly efficient DNA fragmentation (17). It could well be that in non-apoptotic cells the CAD⅐ICAD complex forms transient interactions with chromatin, employing direct contacts to the DNA as well as indirect contacts to its substrate mediated by chromatin-associated proteins such as histone H1 and HMGB1 and 2, which have been shown to interact with CAD and stimulate its activity (18, 39 -41). Interaction of CAD with these proteins might be important for the enzyme to gain access to the linker DNA between nucleosomes.
In summary, our results demonstrate that CAD possesses, in addition to the active site, a second DNA binding region responsible for stable DNA complex formation by the free enzyme and the DFF (CAD⅐ICAD) complex. This binding site is formed by conserved surface-exposed amino acid residues close to and directly from helix ␣4, which is distal to the ␤␤␣-Me-finger active site motif of this nuclease and binds in the DNA major groove. In addition, we identified a region important for stabilization of helix ␣4 that comprises amino acid residues in spatial proximity to the N-terminal end of this helix. The bipartite protein DNA interaction is found once in the CAD⅐ICAD heterodimer and is repeated twice in the active CAD homodimer (Fig. 7B). Our model explains the unusual capacity of CAD to form stable DNA complexes as a nonspecific nuclease on the structural level.