Cleavage Preferences of the Apoptotic Endonuclease DFF40 (Caspase-activated DNase or Nuclease) on Naked DNA and Chromatin Substrates*

Here we report the co-factor requirements for DNA fragmentation factor (DFF) endonuclease and charac-terize its cleavage sites on naked DNA and chromatin substrates. The endonuclease exhibits a pH optimum of 7.5, requires Mg 2 1 , not Ca 2 1 , and is inhibited by Zn 2 1 . The enzyme generates blunt ends or ends with 1-base 5 * -overhangs possessing 5 * -phosphate and 3 * -hydroxyl groups and is specific for double- and not single-stranded DNA or RNA. DFF endonuclease has a moderately greater sequence preference than micrococcal nuclease or DNase I, and the sites attacked possess a dyad axis of symmetry with respect to purine and pyrimidine content. Using HeLa cell nuclei or chromatin reconstituted on a 5 SrRNA gene tandem array, we prove that the enzyme attacks chromatin in the internucleosomal linker, generating oligonucleosomal DNA ladders sharper than those created by micrococcal nuclease. Histone H1, high mobility group-1, and topoisomerase II activate DFF endonuclease activity on naked DNA substrates but much less so on chromatin substrates. We conclude that DFF is a useful reagent for chromatin research.

Here we report the co-factor requirements for DNA fragmentation factor (DFF) endonuclease and characterize its cleavage sites on naked DNA and chromatin substrates. The endonuclease exhibits a pH optimum of 7.5, requires Mg 2؉ , not Ca 2؉ , and is inhibited by Zn 2؉ . The enzyme generates blunt ends or ends with 1-base 5-overhangs possessing 5-phosphate and 3-hydroxyl groups and is specific for double-and not singlestranded DNA or RNA. DFF endonuclease has a moderately greater sequence preference than micrococcal nuclease or DNase I, and the sites attacked possess a dyad axis of symmetry with respect to purine and pyrimidine content. Using HeLa cell nuclei or chromatin reconstituted on a 5 S rRNA gene tandem array, we prove that the enzyme attacks chromatin in the internucleosomal linker, generating oligonucleosomal DNA ladders sharper than those created by micrococcal nuclease. Histone H1, high mobility group-1, and topoisomerase II activate DFF endonuclease activity on naked DNA substrates but much less so on chromatin substrates. We conclude that DFF is a useful reagent for chromatin research.
Apoptosis, or programmed cell death, plays an important role in the development of an organism and in the maintenance of tissue homeostasis (reviewed in Refs. 1 and 2). Hallmarks of the terminal stages of apoptosis are nucleosomal DNA fragmentation (also termed here "DNA laddering") and chromatin condensation (3)(4)(5). The endonuclease primarily responsible for mediating DNA laddering is activated by caspase-3 treatment of DFF. 1 In its inactive form, DFF is a heterodimer composed of a 45-kDa chaperone and inhibitor subunit (DFF45/ICAD) and a 40-kDa latent endonuclease subunit (DFF40/CAD/CPAN) (6 -10). This protein complex resides in the cell nucleus (7,11). Caspase-3 cleavage of DFF specifically cuts only DFF45, which results in the dissociation of cleaved DFF45 from DFF40 (6 -10). Interestingly, the released endonuclease forms homo-oligomers that are the enzymatically active form of DFF40 (12). In addition, its activity on naked DNA substrates can be further activated by specific chromosomal proteins, such as histone H1 or HMG-1/2 (7,12,13), and topoisomerase II (this report). Furthermore, chromatin condensation can be initiated in isolated nuclei by the addition of recombinant activated DFF40, purified free of caspase-3 and DFF45 breakdown products (7).
It should be noted that the physiological significance of DFF in triggering DNA laddering and chromatin condensation during apoptosis has been unequivocally proven. A homozygous deletion of the single copy gene encoding DFF45 has been created in the mouse germ line (14). Significantly, thymocytes and splenocytes from these knockout mice exhibit greatly reduced DNA laddering or chromatin condensation when exposed to apoptotic stimuli, both in vivo and in vitro, proving that DFF is a principal player for the induction of these events (14,15). Furthermore, such mice still express DFF40, indicating that the chaperone function of DFF45 is required in vivo for creating a potentially functional DFF40 endonuclease. 2 Other gene products also appear to participate in the DFF pathway. Two proteins encoded by genes called cell death-inducing DFF45like effectors, or CIDEs, which exhibit homology to the Nterminal domain of DFF45, can activate apoptosis in a DFF45inhibitable fashion, but their precise mechanism of action remains to be elucidated (16). In addition, an isoform of DFF45, termed DFF35 (also termed ICAD-S (9)), is incapable of acting as a chaperone and acts as an inhibitor of the latent endonuclease (17,18). In conclusion, it is well established experimentally that DFF plays a major and regulated role in apoptotic DNA fragmentation and chromatin condensation, although other pathways have also been identified (19,20).
Nucleases have proven to be valuable reagents for the analysis of chromatin structure (21). Because apoptotic nucleases are known to efficiently create DNA ladders composed of total genomic sequences, if DFF40 is the major player in such laddering, then its sequence specificity for DNA cleavage would appear to be broad enough to be a useful additional tool for chromatin research. Here we report the cleavage site preferences for DFF40 on naked DNA and chromatin substrates and demonstrate that DFF40 is indeed a useful reagent for generating sharp internucleosomal DNA cleavage.

EXPERIMENTAL PROCEDURES
Nuclease Chromatin Substrates-HeLa cell nuclei were purified as described elsewhere (6). Nuclei were successively washed at 4°C in 3 mM MgCl 2 , 10 mM KCl, 1 mM dithiothreitol, 20 mM Tris, pH 7.6, 0.2 M sucrose (buffer A) plus 1% Triton X-100 and then in buffer A alone, suspended at 1 g/l as DNA in buffer A plus 20%, glycerol and aliquots were stored at Ϫ80°C. Chromatin was reconstituted on a tandem 18-mer array of the Lytechinus variegatus 5 S rRNA gene (22), recloned into the vector pGEM3A to yield the recombinant plasmid pGH 207-18 (gift of Dr. Joe Gatewood, Los Alamos National Laboratory). The 18-mer was excised from pGH 207-18 by digestion with HhaI and purified from agarose gels after electrophoresis. Chromatin was reconstituted using a salt step dialysis procedure (23,24). Core histones were purified from HeLa nuclei. Briefly, nuclei were incubated with MNase (Worthington) and washed with buffer containing 1 mM EGTA, and chromatin was eluted with 2 mM EDTA. Chromatin was equilibrated with 0.4 M NaCl and loaded onto a hydroxyapatite column. After washing the column extensively with 0.4 M NaCl, histone H1 was eluted with 0.6 M NaCl, and core histones were subsequently eluted with 2 M NaCl (25); no other protein bands were visible on overloaded SDS-polyacrylamide gels. Core histones were mixed with DNA at a histone:DNA weight ratio of 1.4 in 2.5 M NaCl, 20 mM Tris, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, pH 7.6. The mixture was incubated for 15 min at 37°C, added to a dialysis tube (SpectraPor; M r cut-off of 6000 -8000) and dialyzed at 4°C as follows: 2 h with 1 M NaCl, 3 h with 0.55 M NaCl, 4 h with 0.25 M NaCl, and finally 12 h with buffer without NaCl. If histone H1 was also to be assembled into chromatin, histone H1 purified from HeLa cell nuclei as described above was added at the 0.55 M NaCl step at a concentration of 0.28 g/g of DNA. Chromatin was concentrated using Millipore Ultrafree-MC spin filters to about 0.25 g of DNA/l and stored at Ϫ80°C after glycerol was added to a final concentration of 10%.
Assay for Endonuclease Activity-Recombinant caspase-3 was prepared as described (12). DFF was either purified from HeLa cells or from an Escherichia coli expression system as reported previously (6,12). Similar results were obtained with either source of the protein. However, the recombinant protein had a significantly lower specific activity, possibly due to a lack of the appropriate post-translational modifications. One g of naked DNA, nuclei, or chromatin (as DNA) was incubated at 37°C with 1 l of caspase-3 (0.2 g) and 1 l of DFF (1 unit) in buffer consisting of 10 mM KCl, 3 mM MgCl 2 , 0.5 mM dithiothreitol, 10 mM Hepes, pH 7.5 (final volume 15 l) for varying times, in the absence or presence of supplemented proteins as indicated in the legends to Figs. 3, 4, and 6. When comparisons were made between DFF and MNase, unless otherwise indicated, 3 mM CaCl 2 was added to the above buffer for the MNase reactions. As controls, we routinely preincubated caspase-3 with 10 M Ac-DEAD-CHO (a tetrapeptide aldehyde inhibitor of caspase-3) to demonstrate caspase-3 dependence on DFF activation or similarly postincubated with the inhibitor to block any further action of the protease before the addition of activated DFF to any of the above substrates. Although we have found that digestion of naked DNA by DFF was markedly more active under conditions of higher monovalent cations (e.g. 50 -100 mM), this is not true for digestion of chromatin substrates; thus, to reduce potential protein exchange and redistribution during digestion of chromatin, we chose to employ buffers containing 10 -25 mM monovalent cations. Aliquots of the endonuclease reaction were mixed with 1 ⁄2 volume of stop solution (0.6% SDS, 50 mM EDTA, and 6 mg/ml proteinase K) and incubated for 1 h at 50°C. Gel loading dye buffer was added, and samples were then run on 1.5% SeaKem agarose gels using 1ϫ TAE as the running buffer. After electrophoresis, DNA was stained with ethidium bromide, and gels were scanned with a FluorImager (Molecular Dynamics Inc., Sunnyvale, CA). Images were analyzed using ImageQuant software (Molecular Dynamics) and have been represented as negatives. When radioactive DNA was used as the substrate, gels were dried and imaged with a PhosphorImager (Molecular Dynamics), and images were analyzed as above.
Analysis of Cleavage Sites-For a relative comparison of DNA cleavage site sequence preferences between different endonucleases, pUC19 DNA was digested with EcoRI, 32 P-5Ј-end-labeled with T4 polynucleotide kinase after dephosphorylation, and then DNA circles were created in a T4 DNA ligase-catalyzed reaction as described previously (26). Resulting DNA circles (500 ng) were incubated with 0.25 units of MNase (Worthington) or with 0.0025 units of DNase I (Worthington) for 1, 2, 4, and 6 min, or with 0.5 units of DFF with 0.2 g of caspase-3 for 3, 9, 30, and 60 min at 37°C. Reaction mixtures were 10 l each and contained MgCl 2 and CaCl 2 at 1.5 mM each. DNA was purified, digested with AvaI, and separated on 5% polyacrylamide sequencing gels. For detailed analyses of sequences at cleavage sites, a 177-bp fragment of the HIV-1 5Ј-LTR DNA was excised from plasmid pWLTR11 (26) with BglII and AvaI, purified from an agarose gel after electrophoresis, and 5Ј-end-labeled with T4 polynucleotide kinase. Labeled DNA was incubated with caspase-3 and DFF for 10 min at 37°C, and then purified by phenol/chloroform extraction and ethanol precipitation. To analyze cleavage sites on the coding strand, DNA was digested with BsaI (releasing a 12-bp fragment labeled at the BglII site and allowing direct analysis from the AvaI site). To analyze cleavage sites on the noncoding strand, DNA was digested with ScaI (releasing a 20-bp fragment labeled at the AvaI site and allowing direct analysis from the BglII site). Digestion products were resolved on 6% polyacrylamide sequencing gels together with the appropriate Sanger sequencing reactions (DNA was sequenced with the Amersham T7 Sequenase version 2.0 sequencing kit according to the vendor's protocol).
Analysis of DNA Ends-Reconstituted chromatin was incubated with caspase-3 and DFF, and then DNA was purified. Mononucleosomal DNA was isolated after electrophoresis on a low melting agarose gel, 5Ј-end-dephosphorylated with shrimp alkaline phosphatase (Roche Molecular Biochemicals), and 32 P-5Ј-end labeled with T4 polynucleotide kinase. Aliquots of labeled DNA were digested with EcoRI or 3Ј-end modified with either T4 DNA polymerase (U. S. Biochemical Corp.) or terminal deoxynucleotidyl transferase (U. S. Biochemical Corp.) according to the vendor's protocols. DNA samples were then resolved on 6% polyacrylamide sequencing gels.

Caspase-3-activated DFF Requires Mg 2ϩ , not Ca 2ϩ , and is
Inhibited by Zn 2ϩ -A variety of endonucleases have been implicated in apoptotic DNA laddering, including non-metal iondependent (27)(28)(29), Ca 2ϩ -and Mg 2ϩ -dependent (30 -37), Mg 2ϩdependent (38,39), and either Ca 2ϩ -or Mg 2ϩ -dependent endonucleases (40,41). To rigorously establish the divalent ion requirements for DFF, we first dialyzed the protein against a buffer containing EDTA. We found that caspase-3-activated DFF endonuclease had a pH optimum of 7.5 and only required Mg 2ϩ and not Ca 2ϩ as its divalent cation, even in the presence of EGTA, thereby eliminating the possibility that traces of contaminating Ca 2ϩ in Mg 2ϩ -containing solutions may be required for its activity (data not shown) (see Ref. 42). Under the optimal reaction conditions, the enzyme does not digest either single-stranded DNA or RNA (data not shown). Because Zn 2ϩ has been reported to block apoptotic DNA laddering in certain systems (43,44), we also tested this ion and found it to be a strong inhibitor of DFF endonuclease activity (data not shown).
Caspase-3-activated DFF Possesses Moderately Greater Sequence Preferences Than MNase and DNase I-DFF potentially may be a useful reagent for chromatin structural analyses. We therefore evaluated DFF's sequence preferences for naked DNA cleavage in comparison with those of MNase and DNase I, which have both been well characterized previously (45)(46)(47)(48). Separation of the corresponding cleavage products of 32 P-5Јlabeled pUC19 DNA on a sequencing gel reveals that caspase-3-activated DFF endonuclease possesses moderately more sequence selectivity than either MNase or DNase I (Fig. 1). Nevertheless, DNA products are eventually processed to fragments Յ20 bp after more extensive digestion (data not shown), indicating a quite broad sequence specificity, consistent with the fact that DFF endonuclease is capable of converting purified HeLa cell DNA quantitatively into very small DNA fragments (see below).
Caspase-3-activated DFF Generates Blunt Ends or Ends with a 1-Base 5Ј-Overhang-We determined the nucleotide sequences at various cleavage sites by using 32 P-end-labeled DNA restriction fragments as substrates and resolution of the Watson or Crick strands on sequencing gels. As shown in Fig.  2, the endonuclease generates blunt ends or ends with 1-base 5Ј-overhangs. We analyzed 57 cleavage sites at the nucleotide level within the HIV-1 LTR, the mouse immunoglobulin gene, and pUC19 DNA. While these analyses revealed no simple consensus cleavage sequence, there clearly was a preference with respect to purines and pyrimidines. The frequencies of the 4 bases found on each side of these cleavage sites were as follows: 5Ј-R (72%), R (74%), R (66%), Y (61%) 2 R (65%), Y (67%), Y (75%), Y (75%)-3Ј. It is significant that this purine/ pyrimidine preference exhibits a rotational (dyad) symmetry. This observation, together with the double-stranded cleavage activity of this enzyme, is consistent with our observation that only homo-oligomeric forms of DFF40 are enzymatically active (12).
Oligonucleosomal Ladders Generated by Caspase-3-activated DFF Are Sharper than Those Created by MNase-To determine the utility of DFF as a reagent for chromatin research, we compared its action with that of MNase for generating oligonucleosomal DNA ladders. As shown in Fig. 3, digestion of chromatin in isolated HeLa cell nuclei was dependent on the presence of either enzyme (compare lanes 1 and 2 with lanes  4 -7 and 9 -12) and for DFF required caspase-3 cleavage (compare lanes 14 - 16 with lanes 9 -12). The average nucleosomal DNA repeat length estimated after digestion by either enzyme was approximately 180 bp, in reasonable agreement with previous reports (21). Quantitation by PhosphorImager analysis revealed that 70% of the total genomic DNA could be processed to mononucleosomes by DFF after extensive digestion (Fig. 3A,   lane 9). This incompleteness was not due to DFF-resistant sequences in the human genome, because high molecular weight naked HeLa cell DNA could be quantitatively converted to very small DNA fragments upon digestion by the enzyme (Fig. 3B, lanes 2-6). The incomplete conversion to mononucleosomes may be due to a subfraction of the genome being organized into DFF-resistant chromatin structures, and/or to nuclear heterogeneity in DFF permeability; DFF forms very large oligomeric complexes upon activation (12), and its nuclear access may be partially limited. Consistent with this view is that the addition of 1 mM GTP to reaction mixtures stimulated the digestion kinetics about 2-fold on the nuclear substrate but had no affect on naked DNA digestion kinetics (data not shown), presumably by enhancing the nuclear import and/or retention of DFF by phosphorylation. Digestion of nuclei with caspase-3-activated DFF resulted in oligonucleosomal DNA ladders somewhat sharper than those generated by MNase, as judged by oligonucleosomal multimer band sharpness and the ability to visualize bands of longer oligomers. This is because the interband background between successive oligonucleosomal multimers was higher for MNase digestion products; unlike MNase, DDF lacks exonuclease activity. Furthermore, cleavage within nucleosome core particles was undetectable for DFF digestion products, in contrast to MNase digestion products, which exhibited subnucleosomal DNA fragments (Fig. 3A, compare lanes 6 and 7 with lanes 9 and 10; Fig. 3C, compare lanes  1 and 3). We conclude that DFF is another reagent useful for chromatin research.
In order to establish an in vitro system to study further the action of DFF endonuclease on chromatin, we chose as a model substrate for in vitro chromatin assembly an 18-mer of a 207-bp 5 S rRNA gene sequence, which has been previously shown by Simpson and co-workers (22) to position nucleosomes on each tandem repeat. We then used a salt step gradient procedure with purified HeLa cell core histones to reconstitute nucleosomes onto this 5 S gene tandem array (see "Experimental Procedures"). As shown in Fig. 4, both DFF endonuclease and MNase generated nucleosomal DNA ladders upon digestion of the reconstituted chromatin. However, just as in the nuclear chromatin digests (Fig. 3), it is significant that the nucleosomal DNA ladders generated by DFF endonuclease are sharper than those generated by MNase. Therefore, DFF may be better than MNase as a reagent for some types of chromatin structural

FIG. 2. Analysis of caspase-3-activated DFF DNA cleavage sites.
A 177-bp fragment corresponding to 32 P-end-labeled HIV-1 5Ј-LTR DNA was incubated with caspase-3 and DFF. Fragments generated were cut with either BsaI or ScaI to reveal the coding or noncoding strand patterns after being resolved on a sequencing gel (DFF lanes) in parallel with sequencing reaction products as indicated. The arrowheads depict cleavage products and the positions of DNA strand cleavage within the sequence.
analyses. As expected, digestion of the naked DNA control only generated a nondiscrete smear of DNA fragmentation products (Fig. 4). Interestingly, the kinetics of DNA fragmentation by DFF endonuclease on naked DNA are only severalfold faster than on the chromatin substrate (as judged by the intensities of the high molecular weight doublet bands) (see also Fig. 6), whereas we note that MNase or DNase I attack naked DNA much more rapidly than chromatin (data not shown). We interpret this result as follows. The assembly of DNA into nucleosome core particles, which would be expected to reduce DNA cleavage accessibility, is compensated by an activation of DFF endonuclease by nucleosomal DNA wrapping (see below). In summary, we can generate nucleosomal DNA ladders by digestion of isolated nuclei or in vitro assembled chromatin with activated DFF very much like those generated in vivo during the terminal stages of apoptosis.

Caspase-3-activated DFF Cleaves Chromatin in the Internucleosomal Linkers Leaving 5Ј-Phosphate and 3Ј-Hydroxyl
Groups-To prove that DFF cuts in the linker region between nucleosomes, we isolated mononucleosomal-length DNA fragments generated after digestion of the in vitro reconstituted chromatin with DFF endonuclease and mapped the positions of DNA cleavage with respect to the 5 S rRNA gene repeat. For this purpose, we cut the mononucleosomal DNA with EcoRI, whose pair of closely spaced recognition sequences are known to be located in the internucleosomal linker region after nucleosomal assembly (22). We found that the majority of the DNA fragments were reduced in length by Յ20 bp (Fig. 5,  compare lanes 1 and 3), proving that the cutting sites for DFF and EcoRI are close together in the internucleosomal linker region (Fig. 5, bottom diagram). Quantitation by PhosphorImager analysis revealed that 80% of these DNA molecules were between 200 and 150 bp in length. The 20% below 150 bp in length may be in part accounted for by a background of nonpositioned and/or partially assembled nucleosomes generated under the conditions of chromatin reconstitution that we employed. In addition, we found that the DNA ends generated by DFF endonuclease possessed 3Ј-hydroxyl groups, because they could be extended by terminal deoxynucleotidyl transferase (Fig. 5, compare lanes 3 and 5), and radioactive nucleotides could be incorporated by either terminal deoxynucleotidyl transferase, T4 or T7 DNA polymerase (data not shown). In addition, the 5Ј-ends of DFF-generated mononucleosomal DNA could not be phosphorylated by T4 polynucleotide kinase unless they were first dephosphorylated by treatment with alkaline phosphatase (Fig. 5, compare lanes 2 and 3), indicating the initial presence of 5Ј-phosphate groups. Finally, an extension reaction with T4 DNA polymerase barely shifted the DNA lengths of the products (Fig. 5, compare lanes 3 and 4), indi-

FIG. 3. Comparison of DNA cleavage patterns of MNase and caspase-3activated DFF on chromatin of HeLa cell nuclei.
A, HeLa cell nuclei were incubated as described under "Experimental Procedures" in the presence or absence of the indicated proteins for the specified times in a buffer (Mg/Ca) containing 3 mM MgCl 2 plus 3 mM CaCl 2 for MNase digestion at 25°C or containing 3 mM MgCl 2 for caspase-3-activated DFF digestion at 33°C. DNA was purified, resolved on an agarose gel, and visualized after staining with ethidium bromide. M, molecular weight markers. B, purified HeLa cell DNA was cleaved at 37°C with caspase-3-activated DFF for the indicated times as described under "Experimental Procedures." DNA digestion products were purified and resolved on an agarose gel, and DNA was visualized after staining with ethidium bromide. M, molecular weight markers. C, the DNA samples from A (lanes 7 and 9) were resolved on a 5% acrylamide gel and visualized after staining with ethidium bromide. M, molecular weight markers.

FIG. 4. Cleavage of reconstituted chromatin by caspase-3-activated DFF.
An 18-mer of the 5 S rRNA gene, as naked DNA or in vitro reconstituted chromatin, was incubated in the presence of caspase-3 and DFF for 5, 10, 20, and 40 min at 37°C. As a control, the reconstituted chromatin was also separately incubated with MNase (0.1 units/ l, with 3 mM CaCl 2 ) for 1, 3, 10, and 30 min at 25°C. DNA digestion products were purified and resolved on an agarose gel, and DNA was visualized after staining with ethidium bromide. M, molecular weight markers.
cating the nearly blunt-end nature of the cleavage products. Controls with restriction fragments bearing 5Ј-overhangs proved that the polymerase reaction was functional (data not shown).
Differential Activation of DFF Cleavage by Chromosomal Proteins on Naked DNA and Chromatin Substrates-We have previously reported that DFF endonuclease cleavage of naked DNA substrates can be activated approximately 20-fold by either histone H1 or HMG-1/2 (7,12,13). We have also noted that assembly of naked DNA into chromatin does not dramatically inhibit DFF digestion kinetics as opposed to those of MNase (Fig. 4, data not shown). Histone H1 and HMG-1/2 each prefer to bind to supercoiled DNA and DNA crossovers (49 -51); such stimulation may be in part due to trapping DNA substrates in a conformation more favorably attacked by DFF, which may resemble more closely the DNA wrapping around the histone octamer. To test this hypothesis, we decided to determine whether topoisomerase II, another protein that prefers to bind to supercoiled DNA and DNA crossovers (52), could also stimulate DFF endonuclease cleavage on a naked DNA substrate. Topoisomerase II has also been shown to mediate the initial stages of apoptotic DNA cleavage into 50 -100-kb fragments under conditions of oxidative stress (53), making it an even more attractive candidate protein for further study. As shown in Fig. 6A, topoisomerase II proved to also be an activator of DFF endonuclease activity to a level roughly equivalent to that of histone H1 or HMG-1 (compare lane 4 with lanes 1-3); including higher levels of HMG-1 or topoisomerase II resulted in a 20-fold activation for either protein equivalent to that seen for histone H1 (data not shown). On the other hand, when core histones were added under conditions that did not lead to nucleosome formation, only inhibition of DFF activity was observed (data not shown).
In order to determine whether DFF activity could also be stimulated by the addition of histone H1, HMG-1, or topoisomerase II to a chromatin substrate, we pair-wise compared the effects of adding the same amounts of these proteins on the cleavage kinetics of naked DNA and reconstituted chromatin (Fig. 6A). We estimate that the histone H1 addition stimulated naked DNA cleavage 20-fold (Fig. 6A, compare lanes 1 and 2) but inhibited cleavage of chromatin about 2-fold (Fig. 6A, compare lanes 5 and 6). HMG-1 stimulated naked DNA cleavage about 10-fold (Fig. 6A, compare lanes 1 and 3) but only about 2-fold when added to reconstituted chromatin (Fig. 6A, compare lanes 5 and 7). Topoisomerase II stimulated naked DNA cleavage about 5-fold (Fig. 6A, compare lanes 1 and 4) but only about 2-fold when added to reconstituted chromatin (Fig. 6A,  compare lanes 5 and 8). Thus, the wrapping of DNA around histone octamers may partially mimic the activation effects of histone H1, HMG-1, and topoisomerase II on altering the conformation of naked DNA, which would tend to reduce the degree of activation by these proteins on the chromatin substrate. Finally, because the simple addition of histone H1 to reconstituted core histone-containing chromatin may not correctly assemble the protein onto nucleosomes, we employed step gradient dialysis to correctly bind the protein to generate histone H1-containing reconstituted chromatin (see "Experimental Procedures"). As shown in Fig. 6B, upon assembly of histone H1 into chromatin, MNase processing of oligonucleosomal to mononucleosomal DNA fragments was inhibited 5-10fold (Fig. 6B, compare lanes 1-4 with lanes 5-8), whereas such processing by DFF was only inhibited approximately 3-fold by histone H1 assembly (Fig. 6B, compare lanes 9 -12 with lanes [13][14][15][16]. Thus, although histone H1 condenses chromatin, this conformational alteration is less inhibitory to cleavage mediated by DFF as compared with MNase, perhaps because reduced cleavage accessibility is compensated by an activation of DFF endonuclease by further nucleosomal DNA wrapping. DISCUSSION It should be appreciated that the in vitro properties of DFF endonuclease cleavage fit the phenotype of the DNA products generated by apoptosis in vivo, namely nucleosomal DNA ladders, with fragments bearing 3Ј-hydroxyl groups; these features have been routinely used in bioassays for cells undergoing apoptosis. It is interesting that DFF endonuclease cleavage of a naked DNA substrate is markedly stimulated by topoisomerase II, a protein recently shown to be responsible for the creation of 50 -100-kb DNA cleavage products during the initial stages of oxidative stress-induced apoptosis (53). This suggests a coordination or linkage between topoisomerase II and DFF in triggering efficient DNA breakdown during the terminal stages of apoptosis. Interestingly, the kinetics of DNA fragmentation by DFF endonuclease on naked DNA and chromatin substrates are nearly equivalent, in marked contrast to those of MNase or DNase I, which attack naked DNA orders of magnitude more rapidly than chromatin. Cleavage of naked DNA by DFF is activated by histone H1, HMG-1/2, and topoisomerase II, proteins that trap supercoils and DNA crossovers, which induce DNA conformations that partially mimic nucleosomal DNA wrapping. DFF activation for chromatin cleavage by these proteins, however, is much less than for naked DNA digestion. Thus, the assembly of DNA into nucleosome core particles, which would be expected to reduce DNA cleavage accessibility, is apparently compensated by an activation of DFF endonuclease by the resulting nucleosomal DNA wrapping. The observed activation of DFF endonuclease for naked DNA cleavage may also in part be due to direct interactions with these chromosomal proteins, which may recruit the enzyme to the DNA substrates (12).
DFF has a place as a reagent for chromatin research. In the current study, we have proven that the endonuclease cuts in the internucleosomal linker regions of chromatin. Although the possibility was remote, it was formally possible that the presumptive nucleosomal DNA ladders observed upon apoptotic cell death by researchers in the field could have been a reflection of cleavage of chromatin at, for example, the dyad axis of symmetry of the nucleosomal repeat. The addition of caspase-3 alone or DFF alone to isolated nuclei did not result in detectable DNA breakdown (Fig. 3, lanes 14 and 15). Although DFF exists in the nucleus of living cells (7,11), upon cellular fractionation during nuclei isolation, the protein is found in the cytoplasmic fraction (6). This nuclear leakage of DFF upon cell fractionation means that internucleosomal DNA laddering in isolated nuclei is dependent on the addition of exogenous caspase-3-activated DFF (6). This feature is ideal for experimentally controlling the extent of endonuclease digestion. The DNA sequence preferences for DFF are more selective than even MNase, however, indicating that the apoptotic endonuclease may not be the reagent of choice for precise mapping of in vivo nucleosome positions by first hit kinetics footprinting procedures. However, the lack of exonuclease activity, the marked preference for cleavage in the internucleosomal linker region, and the lack of intranucleosomal DNA cleavage after production of mononucleosomes make DFF an attractive choice for the generation and analysis of mononucleosomes. The extent of DFF endonuclease recognition of "nucleosome-free" regions known as DNase I-hypersensitive sites in chromatin, the nature of putative DFF-resistant chromatin structures, and how DFF attacks nucleosomal arrays in transcriptionally repressed and active genes are interesting questions that remain to be addressed in the future.