Both DNA and histone fold sequences contribute to archaeal nucleosome stability.

The roles and interdependence of DNA sequence and archaeal histone fold structure in determining archaeal nucleosome stability and positioning have been determined and quantitated. The presence of four tandem copies of TTTAAAGCCG in the polylinker region of pLITMUS28 resulted in a DNA molecule with increased affinity (DeltaDeltaG of approximately 700 cal mol(-1)) for the archaeal histone HMfB relative to the polylinker sequence, and the dominant, quantitative contribution of the helical repeats of the dinucleotide TA to this increased affinity has been established. The rotational and translational positioning of archaeal nucleosomes assembled on the (TTTAAAGCCG)(4) sequence and on DNA molecules selectively incorporated into archaeal nucleosomes by HMfB have been determined. Alternating A/T- and G/C-rich regions were located where the minor and major grooves, respectively, sequentially faced the archaeal nucleosome core, and identical positioning results were obtained using HMfA, a closely related archaeal histone also from Methanothermus fervidus. However, HMfA did not have similarly high affinities for the HMfB-selected DNA molecules, and domain-swap experiments have shown that this difference in affinity is determined by residue differences in the C-terminal region of alpha-helix 3 of the histone fold, a region that is not expected to directly interact with DNA. Rather this region is thought to participate in forming the histone dimer:dimer interface at the center of an archaeal nucleosome histone tetramer core. If differences in this interface do result in archaeal histone cores with different sequence preferences, then the assembly of alternative archaeal nucleosome tetramer cores could provide an unanticipated and novel structural mechanism to regulate gene expression.

Nucleosomes are highly conserved structures that form the primary level of genome compaction and archiving into chromatin in almost all eukaryotes. Regardless of the species, the nucleosome contains ϳ146 bp of DNA wrapped ϳ1.65 times around a histone octamer core formed by a central histone (H3ϩH4) 2 tetramer flanked by two histone (H2AϩH2B) dimers (1)(2)(3). Although the role of the nucleosome as a structural element has long been recognized, it is now also clear that positioned nucleosomes play regulatory roles in gene expres-sion (4), and this has led to considerable experimental interest in designing, isolating, and characterizing DNA sequences that are preferentially incorporated into nucleosomes and therefore act as nucleosome positioning elements (5). DNA molecules with this property were designed logically, with sequences that conform to tandem repeats of (A/T) 3 NN(G/C) 3 NN, based on the argument that such molecules would be anisotropically flexible and would most readily accommodate the double-helix distortions needed for tight wrapping around the nucleosome core (6,7). Repetition of this simple 10-bp motif results in a DNA molecule with A/T-and G/C-rich regions repeated sequentially approximately once per helical turn. A/T-and G/C-rich sequences were known to accommodate minor and major groove compressions, respectively (8,9). As expected, when such molecules were incorporated into a nucleosome, the alternating A/T-and G/C-rich regions were positioned where the minor and major grooves were compressed, respectively, as they sequentially faced the histone core (7). However, now with high resolution structures available, it is known that the nucleosome is only quasi-symmetric and that the wrapped DNA is also distorted into regions of under-and overwinding and has sites with non-ideal base pair stacking (1)(2)(3). As might be anticipated, to accommodate these additional distortions considerably more complex "rules" have recently been established for nucleosome positioning sequences, based primarily on the sequences of both natural and synthetic DNA molecules selectively incorporated into nucleosomes in vitro by octamer binding (10 -12). Consistent with the designed molecules, these selected molecules also have A/T-and G/C-rich regions alternating in phase with the helical repeat as a dominant feature, but most also have multiple short A-tracts (A 3-4 ) on one DNA strand and an overabundance of TA dinucleotides at ϳ10-bp intervals, features that introduce elements of sequence complexity. For example, the mouse genomic DNA fragments isolated with the highest histone octamer affinity have sequences that contain 10-bp repeats that conform to a consensus TATA-AACGCC motif (11). Subsequently, these molecules have been shown to exhibit exceptional flexibility, both in bending and in accommodating changes to the helical twist (13,14). In this regard, it is noteworthy that TA is the least stable dinucleotide, and TATA is one of the most flexible tetranucleotides (15,16). Consistent with exploiting these sequence-determined structural properties, TA steps have been localized Ϯ15 bp from the central dyad of positioned nucleosomes, specifically at the sites at which the base pair stacking is necessarily most distorted (17). Presumably, this use of TA dinucleotides minimizes the energy costs of nucleosome assembly.
A complicating feature in extending these analyses is that the eukaryotic histone octamer contains four different histones, H2A, H2B, H3, and H4. This makes it difficult to interpret such nucleosome positioning sequences in terms of individual histone-DNA interactions. In addition, because the octamer as-sembly of DNA into nucleosomes in vitro is very salt-dependent, the assembly technology used has been raised as an issue in quantitative data interpretation (12). Fortunately, in contrast, archaeal nucleosome assembly in vitro occurs with only one archaeal histone and occurs spontaneously under almost all solution conditions, providing a much simpler alternative experimental system for parallel and comparative studies (18 -20). The histone folds of archaeal histones and the eukaryotic nucleosome core histones are almost identical, but unlike their eukaryotic counterparts, archaeal histones do not have N-and C-terminal amino acid sequences extending from this fold (21,22). Archaeal nucleosomes resemble the structure formed by the (H3ϩH4) 2 tetramer at the center of the eukaryotic nucleosome. Both contain a histone tetramer circumscribed by ϳ85 bp of DNA but which makes direct histone fold contacts with only ϳ60 bp of sequential DNA (1,23). In both structures, the DNA can be wrapped in either a negative or positive toroidal supercoil (24 -26), and both tetramers recognize and respond to nucleosome positioning sequences (27)(28)(29). Consistent with this homology, the DNA molecules selectively incorporated into archaeal nucleosomes by HMfB, the most extensively studied archaeal histone that originated from Methanothermus fervidus, have short A-tracts and an overabundance of AA (ϭTT), TA, and GC dinucleotides repeated at ϳ10-bp intervals. The GC harmonic is displaced by ϳ5 bp from the AA and TA harmonics, and therefore these molecules also have sequences that conform to repeats of the generic (A/T) 3 NN(G/C) 3 NN nucleosome positioning motif (30). However, they also exhibit sequence complexity. In contrast to the overabundance of TA and GC repeats, repetitions of AT and CG were not highly selected, and almost all of the short A-tracts are located on the same DNA strand.
The experiments reported here were undertaken to define and quantitate how HMfB homotetramers interact with DNA molecules designed to conform to the (A/T) 3 NN(G/C) 3 NN nucleosome positioning motif and with molecules selected experimentally by their preferential incorporation into archaeal nucleosomes. As predicted, the results obtained are consistent with these molecules being assembled into archaeal nucleosomes (29,30) with alternating A/T-and G/C-rich sequences positioned where the minor and major grooves, respectively, face the histone core. The importance of the TA dinucleotide has been directly established. In addition, we have uncovered an unanticipated interdependence of archaeal histone fold-DNA sequences in facilitating archaeal nucleosome assembly. M. fervidus also contains HMfA, a second archaeal histone with a primary sequence 84% identical to that of HMfB, and based on high resolution crystal structures HMfA and HMfB dimers have almost identical histone folds (22,31). However, HMfA does not bind similarly with high affinity to the DNA molecules selectively incorporated into archaeal nucleosomes in vitro by HMfB. Even more surprising, this difference in affinity is apparently determined by differences in the C-terminal region of ␣-helix 3, a region of the histone fold predicted to be buried within the nucleosome core with no direct contacts to the DNA (1-3, 20, 21, 32).

Archaeal Histone Purification and ␣-Helix 3 Exchange
Recombinant HMfA and HMfB were synthesized in Escherichia coli JM105, purified, and quantitated as previously described (18,32,33). A QuikChange kit (Stratagene, La Jolla, CA) was used to change the sequences of codons 61-63 of hmfA and hmfB, cloned in the expression plasmids pKS395 and pKS323, from GAATTAGCT and GAACTAGCA to GAGCTAGCT and GAGCTAGCA, introducing an NheI site (GCTA-GC) into both genes without changing the encoded amino acid sequence (ELA) and generating plasmids pKS597 and pKS600. Oligonucleotides with the sequences 5Ј-CTAGCACGAAAAATGTTCAAATAAGATCTA and 3Ј-GTGCTTTTTACAAGTTTATTCTAGATTCGA and 5Ј-CTAGCA-GTTCGAAGATTTAAGAAATA and 3Ј-GTCAAGCTTCTAAATTCTTT-ATTCGA, were annealed to obtain ds 1 DNA molecules with singlestranded NheI and HindIII half-site extensions that encoded the residues of HMfA and HMfB C-terminal to residue 62. Aliquots of pKS597 and pKS600 plasmid DNA were digested with NheI plus Hin-dIII, and ligation with the DNA molecule encoding the appropriate C-terminal sequence resulted in plasmids pKS603 and pKS604. These plasmids were transformed into E. coli JM105, and isopropyl-␤-D-thiogalactopyranoside (400 M) was added to exponentially growing cultures of the transformants to induce the synthesis of HMfA/B-␣3 and HMfB/A-␣3. These proteins were purified, quantitated, and their CD spectra measured using an AVIV model 62A-DA spectropolarimeter (Aviv, Lakewood, NJ) as previously described for HMfA and HMfB (33)(34)(35). Archaeal histone molar solution concentrations are cited in terms of dimers, and the apparent K d values reported for archaeal nucleosome assembly are given in terms of histone tetramers.

DNA Constructions and Labeling
MM201 and MM301-Plasmid pMM201 and pMM301 preparations (36) were purified from 1-liter cultures of E. coli DH5␣ using Qiagen maxiprep kits (Valencia, CA) and digested with EcoRI plus PstI. The products were labeled at the EcoRI half-sites using [␥-32 P]ATP in exchange reactions catalyzed by T4 polynucleotide kinase (Invitrogen) (37) and separated by electrophoresis through 8% T, 0.11% C polyacryamide gels run at 8 V/cm in TBE (90 mM Tris borate, 2 mM EDTA, pH 8). EcoRI plus PstI digestion released 159-bp and 173-bp molecules, designated MM201 and MM301, from pMM201 and pMM301, respectively, which were located in the wet gel by autoradiography. The regions of the gel that contained these molecules were excised, crushed, and incubated overnight at 37°C in 0.5 M ammonium acetate, 2 mM EDTA, 0.1% (w/v) SDS. The gel fragments were removed by centrifugation, and the eluted DNA was precipitated from the supernatant by the addition of 2 volumes of 95% ethanol, ethanol-washed and redissolved in TE (10 mM Tris/HCl, 1 mM EDTA, pH 7.5).
Litmus28, (TTTAAAGCCG) 2 , (TTTAAAGCCG) 4 , and (TTTAAAGC CG) 6 -Preparations of plasmid pLITMUS28 (New England Biolabs, Beverly, MA) were purified from E. coli DH5␣ using Qiagen maxiprep kits and digested with BssHII plus Acc65I. This released a molecule with a 93-bp ds region from the polylinker region (designated Litmus28) that was 32 P-labeled at the BssHII half-site using [␥-32 P]ATP and T4 polynucleotide kinase, gel-purified, and used as a control DNA that lacked TTTAAAGCCG sequences. Aliquots (1 M) of oligonucleotides with the sequences 5Ј-CTTCGTGGATCCC(TTTAAAGCCG) n TGCAGG-AATTCCTCG, where n was 2, 4, and 6 (Integrated DNA Technologies, Coralville, IA), were converted to ds DNA by primer extension in reaction mixtures that contained 13 M of a primer (5Ј-CGAGGAATTC-CTGCAG), 20 M dNTPs, 50 mM NaCl, 10 mM MgCl 2 , 50 mM Tris/HCl (pH 8). These reaction mixtures were placed at 95°C for 2 min, 47°C for 2 min, and 37°C for 5 min before 3 units of the Klenow fragment of DNA polymerase (Invitrogen) were added, and incubation was continued for 30 min at 37°C. The ds DNA molecules generated were digested with BamHI plus EcoRI, and after gel purification, the products were ligated with BamHI plus EcoRI-digested pLITMUS28 and transformed into E. coli DH5␣. Plasmids designated pTA2, pTA4, and pTA6 were isolated from ampicillin-resistant transformants, sequenced to confirm the constructions, and then digested with BssHII plus AvrII, BssHII plus AatII, and BssHII plus BamHI, respectively, to generate molecules with ds regions of 93 bp, 93 bp, and 89 bp that contained (TTTAAAGCCG) 2 , (TTTAAAGCCG) 4 , and (TTTAAAGCCG) 6 located at the same position within the pLITMUS28 polylinker (see Fig. 2A). These molecules were 32 P-labeled at the BssHII half-site using [␥-32 P]ATP and T4 polynucleotide kinase and gel-purified.

Agarose Gel-shift Assays
Mixtures (15 l) of 100 ng of EcoRI-linearized pBR322 DNA and increasing amounts of the archaeal histone were incubated for 25 min at 25°C in 100 mM KCl, 50 mM Tris/HCl. Gel loading buffer (5 l of 40% sucrose, 0.25% bromphenol blue) was added, and the products were separated by electrophoresis at 0.7 V/cm through an 0.8% agarose gel (Agarose I, Amresco, Euclid, OH) run in 40 mM Tris acetate, 2 mM EDTA, and visualized by ethidium bromide staining (18,33).

Polyacrylamide Gel-shift Assays
Mixtures (10 l) of 0.1 ng of the 32 P-labeled experimental DNA and 1 or 25 ng of sss DNA were incubated with increasing amounts of the archaeal histone for 25 min at 25°C in 100 mM KCl, 50 mM Tris/HCl (pH 8). Gel loading buffer (5 l) was added, and the products were separated by electrophoresis through 8% T, 0.11% C polyacrylamide gels run in TBE buffer at 8 V/cm (33). The gels were dried and used to generate autoradiograms. Radioactively labeled products were quantitated by laser-scanning densitometry of the autoradiogram (Biomed Instruments Inc., Chicago, IL) and directly by ␤-decay measurements using a Packard Instant Imager.

MN Digestion and Translational Positioning
Archaeal histone and the 32 P-labeled (TTTAAAGCCG) 4 -containing DNA were mixed at a molar ratio of ϳ5:1 and incubated in 100 mM KCl, 1 mM CaCl 2 , 50 mM Tris acetate (pH 8.8) for 25 min at 37°C. MN (Sigma) was added (final concentration of 0.005 unit/g DNA), and incubation continued at 37°C (33). Aliquots (1 l) were removed at increasing time intervals and mixed with 2 l of 100 mM EDTA to terminate the MN digestion. DNA molecules that remained were purified by phenol-chloroform extraction and ethanol precipitation, dissolved in TE, and separated by electrophoresis through 12% T, 0.16% C polyacrylamide gels run at 9 V/cm in TBE. The region of the gel that contained the ϳ60-bp MN-protected DNA molecule was located by autoradiography and excised, and the DNA was eluted and purified as described above. Aliquots were subjected to BamHI, BssHII, EcoRI, HindIII, NcoI, and XbaI digestions. The digestion products were mixed with an equal volume of formamide gel loading buffer (80% formamide, 10 mM EDTA, 0.1% xylene cyanol, 0.1% bromphenol blue, pH 8) and separated by electrophoresis at 50 V/cm through 10% T, 0.52% C polyacrylamide gels that contained 8.3 M urea run in 0.5ϫ TBE. The gels were fixed by immersion in a solution containing 10% (w/v) acetic acid and 10% (v/v) methanol, dried, and used to expose x-ray films. The percentage of the population of the ϳ60-bp molecules cleaved by each restriction enzyme was determined by laser-scanning densitometry of the resulting autoradiograms.

Hydroxyl Radical Footprinting
Archaeal nucleosomes (assembled by incubation of 250 ng of HMfA or HMfB with 100 ng of the (TTTAAAGCCG) 4 -containing DNA, clone 1 or 20 DNA at 25°C for 20 min) were exposed to 10 M Fe(NH 4 ) 2 (SO 4 ) 2 , 20 M EDTA, 0.9 mM ascorbate, 0.03% H 2 O 2 for 2 min at 25°C. One strand of the DNA molecule was labeled by PCR amplification using one 32 P-5Ј-end-labeled primer, and each experiment was repeated with the second DNA strand similarly end-labeled by 32 P-primer incorporation. Hydroxyl cleavage reactions were terminated by the addition of thiourea (200 mM final concentration), and after electrophoresis through 8% T, 0.11% C polyacrylamide gels run at 8 V/cm in TBE, the hydroxyl radical-treated complexes were located by autoradiography. The regions of the gel that contained these complexes were excised, crushed, and incubated overnight at 37°C in 0.5 M ammonium acetate, 2 mM EDTA, 0.1% (w/v) SDS. Gel fragments were removed by centrifugation, and the eluted DNA molecules were precipitated from the supernatant by the addition of 2 volumes of 95% ethanol, ethanol-washed and re-dissolved in formamide gel loading buffer. After heating to 95°C for 1 min, these molecules were separated by electrophoresis at 50 V/cm through 10% T, 0.52% C polyacrylamide gels that contained 8.3 M urea run in TBE. The gels were fixed by immersion in 10% (w/v) acetic acid, 10% (v/v) methanol, dried, and used to expose x-ray films resulting in autoradiograms from which band intensity profiles were obtained by laser-scanning densitometry.

RESULTS
Archaeal Nucleosome Assembly on Curved versus Non-curved DNA Molecules-Archaeal nucleosomes, like eukaryotic nucleosomes, were shown previously by electron microscopy to assemble preferentially at sites on DNA molecules that were intrinsically curved (38). To pursue this observation, a comparison was made of archaeal nucleosome assembly on two DNA molecules, designated MM201 and MM301 (36), which contained similar repetitive sequences, (AAATTTGCCG) 12 and (TTTAAAGCCG) 11 , respectively, but have very different intrinsic curvatures. The sequences of both molecules conform to the [(A/T) 3 NN(G/C) 3 NN] n nucleosome positioning motif, but MM201 is very highly curved whereas MM301 is only slightly curved (36). As expected for a highly curved molecule, MM201 had a very anomalous electrophoretic mobility, whereas MM301 had only a slightly anomalous mobility (Fig. 1), but HMfB assembled preferentially into archaeal nucleosome on MM301. Under HMfB-limited conditions, only MM301 molecules were incorporated into archaeal nucleosomes in reaction mixtures that contained an equimolar mixture of MM201 and MM301 (Fig. 1, lane 3). Intrinsic DNA curvature alone did not therefore equate to preferential assembly into an archaeal nucleosome, and as observed for eukaryotic nucleosomes (7), this result also demonstrated that not all DNA molecules with sequences that conform to the [(A/T) 3 NN(G/C) 3 1. Comparison of archaeal nucleosome assembly on intrinsically curved and non-curved DNA. Reaction mixtures contained 0 (Ϫ), 6.5, 32, 64, 130, 160, 325, or 650 nM HMfB, 0.9 nM 32 P-labeled MM201, and 0.9 nM 32 P-labeled MM301. After incubation for 25 min at 25°C, the reaction products were separated by electrophoresis and visualized by autoradiography (33). MM201 is actually 159 bp (R acc ) but has an electrophoretic mobility indicating an apparent length (R app ) of 356 bp and therefore has an R L ratio (R app /R acc ) of 2.24. In contrast, MM301 has an R L of 1.24, consistent with only slight intrinsic curvature (36). MM201 and MM301 were obtained by PstI plus EcoRI digestion of pMM201 and pMM301, respectively (36). As illustrated, they contained the repetitive sequences shown flanked by vector sequences with BamHI (B), KpnI (K), SalI (S), SacI (Sa), and XbaI (X) sites.
Therefore, DNA molecules of 89 -93 bp in length were generated with two, four, or six tandem repeats of the MM301 motif, TTTAAAGCCG, identically positioned within the pLITMUS28 polylinker ( Fig. 2A). The presence of two copies of this motif did not result in increased affinity for HMfB relative to the 93-bp control Litmus28 DNA generated from the same polylinker region without an insert. However, the presence of four and six copies resulted in DNA molecules with much higher affinities for HMfB ( Fig. 2A). In addition to intrinsic shape, MM201 differs from MM301 in lacking TA dinucleotides repeated at 10-bp intervals. To evaluate this feature as a facilitator of archaeal nucleosome assembly, DNA molecules were generated with four copies of variants of the MM301 motif that lacked TA steps. All these molecules had reduced affinities for HMfB relative to the molecule containing four copies of the MM301 motif (Fig. 2B). Simply replacing the A at position 4 with G reduced the affinity for HMfB more than 3-fold, decreasing the apparent K d for HMfB tetramers from 35 Ϯ 5 to 104 Ϯ 6 nM ( Fig. 2C and Table I). For comparison, four copies of the mouse genomic high affinity motif, TATAAACGCC (13,14), were also identically positioned in the pLITMUS28 polylinker, and the 93-bp DNA molecule amplified from this construct did have increased affinity for HMfB (apparent K d of 72 Ϯ 6 nM) relative to the Litmus28 control and the MM301 TA-less variants. However, the affinity of this molecule for HMfB was still ϳ2fold lower than that conferred by four copies of TTTAAAGCCG, the MM301 motif (Fig. 2, B and C and Table I).
Archaeal Nucleosome Positioning by (TTTAAAGCCG) 4 -Hydroxyl radical footprinting was used to determine the rotational positioning of the (TTTAAAGCCG) 4 sequence when assembled into an archaeal nucleosome. By comparing the hydroxyl radical cleavage patterns of this DNA in solution and when assembled into an archaeal nucleosome, it was apparent that the four A/T-rich regions, and specifically the four TA steps, were protected from hydroxyl radical cleavage when the molecule was assembled in an archaeal nucleosome (Fig. 3A). Virtually identical patterns of protection and cleavage were observed with archaeal nucleosomes assembled by using HMfA or HMfB, and as reported for the mouse genomic (TATA-AACGCC)-repeat (13), some sites within the (TTTA-AAGCCG) 4 -sequence, most notably the CC dinucleotides, were cleaved preferentially by the hydroxyl radical both in the presence and absence of histones. These molecules must therefore adopt similar structures, in terms of preferred sites for hydroxyl radical cleavage, both in solution and when assembled into a nucleosome (13). However, specific protection of the TA steps occurred only when the (TTTAAAGCCG) 4 sequence was assembled into an archaeal nucleosome, consistent with the minor groove facing the histone core at each TA-containing sequence.
To determine whether archaeal nucleosomes assembled using HMfB and the (TTTAAAGCCG) 4 -containing DNA were also translationally positioned, such complexes were exposed to MN digestion, which as previously reported (29,30), resulted in the transient appearance of ϳ90-bp DNA molecules followed by the FIG. 2. DNA sequences and archaeal nucleosome assembly. A, 2, 4, and 6 copies of TTTAAAGCCG were cloned into the pLITMUS28 polylinker as illustrated, and digestion of these constructs with AvrII plus BssHII, AatII plus BssHII, or BamHI plus BssHII generated molecules with 93-bp, 93-bp, and 89-bp ds regions that contained 2, 4, and 6 copies of TTTAAAGCCG, respectively. Digestion of the pLITMUS28 polylinker with Acc65I plus BssHII generated the 93-bp control Lit-mus28 DNA that lacked TTTAAAGCCG inserts. Aliquots of these DNA (0.1 ng) were 32 P-labeled and incubated in the absence (ϪHMfB) and presence (ϩHMfB) of HMfB (32 nM) for 25 min at 25°C. The products were separated by electrophoresis and visualized by autoradiography (33). B, the 93-bp Litmus28 control (lane 1) and 93-bp molecules containing four copies of the sequences listed to the left of the figure (lanes  2-6), generated by AatII plus BssHII digestion of the appropriate pLIT-MUS28 construct as illustrated in A, were 32 P-end-labeled. Aliquots (0.1 ng) were incubated with HMfB (32 nM) plus 25 ng of sss DNA, and the reaction products were separated by electrophoresis and visualized by autoradiography. C, 32 P-labeled DNA molecules containing (TTTA-AAGCCG) 4 , (TATAAACGCC) 4 , and (TTTGAAGCCG) 4 were PCR-amplified from pLITMUS28 constructs (A), and aliquots (0.1 ng) were mixed with 1 ng of sss DNA and incubated without (Ϫ) and with 0.6, 3.2, 6.4, 32, 64, 160, and 320 nM HMfB for 25 min at 25°C. The reaction products were separated by electrophoresis, visualized by autoradiography, and quantitated (Table I)   accumulation of ϳ60-bp MN-protected molecules (Fig. 3B). A population of these ϳ60-bp molecules was isolated and aliquots subjected to digestion by different restriction enzymes. Electrophoresis through denaturing gels revealed that the ϳ60-bp molecules actually had single strands ranging in length from 55 to 61 nt, with the majority being 57 Ϯ 1 nt. Approximately 60% of the ϳ60-bp molecules were cut by BamHI and ϳ3% by EcoRI, but there was no detectable cleavage by any of the other restriction enzymes for which restriction sites were originally present in the DNA that flanked the (TTTAAAGCCG) 4 sequence (Fig. 3A). BamHI digestion generated predominantly one large product with single strands of 49 and 44 nt (Fig. 3C), and therefore the majority, but not all, of the archaeal nucleosomes assembled on the (TTTAAAGCCG) 4 -containing DNA positioned such that they protected the same 55-61-bp region from MN digestion (Fig. 3A). This encompassed all four TTTA-AAGCCG sequences and the BamHI restriction site. Because the single strands of the BamHI restriction product differed in length by 5 nt (49 and 44 nt), the ϳ60-bp MN-protected substrate molecules must have had one 3Ј-terminal unpaired nucleotide. Archaeal Nucleosome Positioning by in Vitro Selected Molecules-DNA molecules selected previously by their preferential incorporation into archaeal nucleosomes by HMfB were cloned and sequenced (30). They all had fixed 25-bp sequences (25R and 25L) flanking different 60-bp sequences, and by aligning these variable 60-bp sequences, a consensus sequence was generated that conformed to the [(A/T) 3 NN(G/C) 3 NN] n nucleosome positioning motif (Ref. 30 and Fig. 4). Many of these molecules have TA dinucleotides repeated at ϳ10-bp intervals. In this regard they resemble the (TTTAAAGCCG) 4 -and (TATA-AACGCC) 4 -repeats, but they differ in not having reiterated sequences or inverted repeat symmetry. This lack of symmetry seems noteworthy because these are molecules that were selectively incorporated by HMfB homodimers into archaeal nucleosomes that seem likely to have a symmetric homotetramer histone core. Based on conforming to the [(A/T) 3 NN(G/C) 3 NN] n motif, it was expected that these molecules would assemble into archaeal nucleosomes with the alternating A/T-and G/Crich regions located where the minor and major grooves faced the nucleosome core (30). Hydroxyl radical footprinting of archaeal nucleosomes assembled using clones 1 and 20 as representative DNA molecules, and either HMfB or HMfA, generated results entirely consistent with this prediction. When assembled into archaeal nucleosomes, the A/T-rich regions present in these DNA molecules at ϳ10-bp intervals, which in many cases included a TA dinucleotide, were protected from hydroxyl radical cleavage (Fig. 4). As illustrated by the consensus sequence shown in Fig. 4, the majority of the 60-bp HMfBselected sequences have a poly(A) tract centered around position 20, and this region of both clone 1 and 20 DNAs was almost completely protected from hydroxyl radical cleavage when assembled into an archaeal nucleosome.
HMfA Assembly of HMfB-selected DNA Molecules into Archaeal Nucleosomes-Although most of the HMfB-selected molecules have 60-bp central sequences that conform generically to the [(A/T) 3 NN(G/C) 3 NN] n motif, each has a different sequence and different increased affinity for HMfB (30,39). They are apparently molecules with different levels of optimization for incorporation into archaeal nucleosomes by HMfB, and it was therefore of interest to determine whether these molecules were similarly optimized for archaeal nucleosome assembly by HMfA. Surprisingly, this was not the case (Fig. 5). Although HMfA and HMfB have very similar high resolution structures (22), and all the DNA-binding residues identified by mutagenesis in HMfB (32) are conserved in HMfA, HMfA had affinities for clone 1 and 20 DNAs that were Ͼ2-fold and Ͼ20-fold lower than the affinities of HMfB for these molecules (Table II). However, as noted above, HMfA did assemble these DNAs into archaeal nucleosomes with the same rotational positioning as HMfB (Fig. 4).
Role of ␣-Helix 3 in Archaeal Nucleosome Stability-The sequences of HMfA and HMfB are 84% identical but do differ in three consecutive positions, RKM versus VRR, very near the C terminus (Fig. 6).When the crystal structures of the HMfA and HMfB dimers were superimposed, these differences apparently caused a slight difference in the histone fold orientation of ␣-helix 3 relative to ␣-helix 2 (22). However, based on the structure of the (H3ϩH4) 4 tetramer within the eukaryotic nucleosome core (1,19,21), these C-terminal residues were predicted to be buried inside the archaeal nucleosome histone tetramer core. They were expected to contribute to a four-helix bundle that forms the dimer:dimer interface at a site that has no direct contact with the DNA (Fig. 6). Nevertheless, a difference in the dimer:dimer interface might result in HMfA and HMfB tetramers with structures and/or shapes sufficiently different to require different DNA sequences for optimum nucleosome assembly. To test this hypothesis, HMfA/B-␣3 and HMfB/ A-␣3 variants were constructed. In these variants, the C-terminal residues of ␣-helix 3 of HMfA were replaced by their A, archaeal nucleosomes were assembled on the (TTTAAAGCCG) 4containing DNA, using either HMfA or HMfB, and exposed to hydroxyl radical cleavage. The DNA products were separated by electrophoresis under denaturing conditions, and the dried gel was used to generate the autoradiogram from which band intensity profiles were obtained by laser-scanning densitometry. The control lanes contained aliquots of the DNA that were (ϩ) or were not (Ϫ) exposed to the hydroxyl radical in the absence of the archaeal histone. Regions protected from hydroxyl radical cleavage by assembly into an archaeal nucleosome are connected to the corresponding hydroxyl radical footprint. B, archaeal nucleosomes assembled containing HMfB and 32 P-labeled-(TTTA-AAGCCG) 4 DNA were exposed to MN, and samples were removed at increasing times from 1 to 120 min. The MN digestion products were separated by polyacrylamide gel electrophoresis and visualized by autoradiography. Control lanes contained aliquots of the DNA incubated without (Ϫ) or with (ϩ) MN for 2 min in the absence of HMfB. C, aliquots of a population of ϳ60-bp MN-protected DNA molecules, generated as illustrated in B, were exposed to BssHII (Bss), EcoRI (E), BamHI (B), or NcoI (N) digestion. The products were separated by electrophoresis under denaturing conditions, visualized by autoradiography, and quantitated by scanning densitometry of the autoradiogram. Control lanes contained an aliquot of the ϳ60-bp population (Ϫ) and single-stranded size standards. Quantitation of the BamHI digestion products revealed that Ͼ60% of the ϳ60-bp MN-protected molecules contained the region between the two triangles positioned above the sequence in A.
counterparts from HMfB, and vice versa. Based on ellipticity measurements at 222 nm (⑀ 222 ), under identical solution conditions at 25°C, HMfA/B-␣3 and HMfB/A-␣3 had ϳ90% and ϳ75% of the ␣-helical content of HMfA and HMfB, respectively, and both formed complexes with DNA with features characteristic of the ␣-helix 3 donor (Fig. 7). It has been well established that archaeal nucleosome assembly on DNA molecules of Ͼ2 kbp results in complexes that migrate faster during agarose gel electrophoresis than the DNA molecule alone and that, under saturating conditions, such complexes formed by HMfB migrate faster than those formed by HMfA (33). This difference in mobility was transferred with the ␣-helix 3 residues. Complexes formed by HMfA/B-␣3 with linear pBR322 DNA migrated faster than those formed by HMfA, and complexes FIG. 4. Archaeal nucleosome positioning on clone 1 and 20 DNAs. The consensus sequence identifies the nucleotide that occurred most frequently at each site in the 60-bp variable regions of 89 molecules selected by HMfB incorporation into archaeal nucleosomes. As illustrated, the font size is directly proportional to the frequency of occurrence of the nucleotide at that location (30). The fixed 25-bp sequences (25R and 25L) that flanked each of the different 60-bp selected sequences are shown at the top of the figure. Clone 1 and 20 DNAs were 32 P-end-labeled and assembled into archaeal nucleosomes using either HMfA or HMfB. After exposure to the hydroxyl radical, the products generated were separated by electrophoresis and visualized by autoradiography, and the intensity profiles shown were obtained from the autoradiograms by scanning densitometry. Control tracks contained aliquots of the DNA exposed (ϩ) or not exposed (Ϫ) to hydroxyl radical cleavage in the absence of the archaeal histone. The sequences of 60-bp central regions of clones 1 and 20 DNA are shown above the autoradiograms with the protected regions connected by boxes to the corresponding footprints. The internal AluI site indicated in clone 1 DNA was used to establish that HMfB assembled archaeal nucleosomes predominantly at one translational position on clone 1 DNA (30). These nucleosomes protected the region between arrows shown below the clone 1 sequence from MN digestion. formed by HMfB/A-␣3 migrated slower than those formed by HMfB (Fig. 7). Although the underlying molecular basis for this difference in the mobilities of large archaeal histone-DNA complexes remains unclear, it was found to correlate with a difference in assembly at the single archaeal nucleosome level. The affinities of HMfA/B-␣3 and HMfB/A-␣3 for clone 1 and 20 DNAs also changed to correspond with the affinities of the ␣-helix 3 donor (Table II). As shown for clone 20 DNA (Fig. 5), a molecule with one of the highest affinities for HMfB (30,39), the affinity of HMfA/B-␣3 was ϳ10-fold higher than that of HMfA, whereas the affinity of HMfB/A-␣3 was ϳ35-fold lower than that of HMfB (Table II). DISCUSSION The simplest explanation for the widespread conservation of the eukaryotic nucleosome is that this complex, in essentially its current configuration, existed in the last common ancestor of all eukaryotes. A likely prokaryotic origin for the nucleosome became apparent with the discovery of archaeal histones. All the evidence accumulated since supports that archaeal histones are ancestral and structural homologs of their eukaryotic nucleosome counterparts and that the archaeal nucleosome resembles the structure formed at the center of the eukaryotic nucleosome by DNA wrapped around the histone (H3ϩH4) 4 tetramer (19,20,23,30). Both structures have a histone tetramer core that recognizes positioning signals, directly contacts ϳ60 bp, and wraps ϳ85 bp of DNA alternatively in either a positive or negative toroidal supercoil (24,25,26). However, unlike the eukaryotic histones, archaeal histones form soluble homodimers in solution that assemble spontaneously in vitro into archaeal nucleosomes. The experiments reported here were undertaken to investigate and quantitate single archaeal nucleosome assembly and positioning in reaction mixtures that contained the minimal components, one histone and a defined DNA molecule only ϳ100 bp in length. The results obtained are, overall, fully consistent with the results from previous eukaryotic nucleosome assembly studies (6,7,10,11,12,17). However, we have also obtained results that reveal an unexpected sequence specificity in archaeal histone-DNA interactions, results that hint at a previously unsuspected structural basis for nucleosome positioning that could be exploited to regulate gene expression.
The (A/T) 3 NN(G/C) 3 NN Motif-Neither intrinsic curvature nor a sequence that conformed to the generic (A/T) 3 NN(G/ C) 3 NN nucleosome positioning motif (6, 7) guaranteed preferential incorporation into an archaeal nucleosome. Incubation of HMfB with an equimolar mixture of MM201 and MM301 resulted in the preferential incorporation of MM301 into archaeal nucleosomes, even though MM301 has much less intrinsic curvature than MM201 (Fig. 1). The difference in the affinities of MM201 and MM301 for HMfB could therefore reflect a difference in intrinsic writhe, or the structure of MM301 may more readily accommodate the distortions needed for archaeal nucleosome assembly. In this regard, the MM301 motif (TTTA-AAGCCG) but not the MM201 (AAATTTGCCG) contains TA, the most easily distorted dinucleotide (15), and as few as 40 bp of a TA-containing repetitive sequence were sufficient to result in preferential incorporation of a DNA molecule into an eukaryotic nucleosome (6). Consistent with this, the presence of four copies of the 10-bp MM301 motif resulted in a molecule with increased affinity for HMfB relative to the control Litmus28 DNA ( Fig. 2A and Table I). Based on differences in apparent K d values calculated from polyacrylamide gel-shift data, the difference in free energy (⌬⌬G) of HMfB assembly into an archaeal nucleosome containing the MM301 motif versus assembly using Litmus28 polylinker sequence was 702 Ϯ 150 cal mol Ϫ1 . In contrast, when the TA dinucleotides in this sequence were changed to TGs, there was only a marginal difference in the free energy of assembly of HMfB into an archaeal nucleosome (⌬⌬G ϭ 59 Ϯ 88 cal mol Ϫ1 ) relative to assembly using the Litmus28 control. Given these values, each of the four 10-bp MM301 motifs contributed ϳ175 Ϯ 38 cal mol Ϫ1 (702 Ϯ 150 Ϭ 4) and each of the TA steps contributed ϳ161 Ϯ 59 cal mol Ϫ1 (702 Ϯ 150Ϫ59 Ϯ 88 ϭ 643 Ϯ 238 Ϭ 4) to the increased stability of an archaeal nucleosome containing (TTTAAAGCCG) 4 , relative to an archaeal nucleosome containing Litmus28. Similarly, for comparison, each helical turn of a high affinity TA-containing repetitive molecule was calculated to contribute ϳ200 cal mol Ϫ1 to eukaryotic nucleosome stability (6), and substitution of G for A at the critical TA step located at position Ϫ15 relative to the central dyad decreased eukaryotic nucleosome stability (⌬⌬G) by ϳ480 cal mol Ϫ1 (17). As anticipated, the TA dinucleotides in this sequence were positioned where the minor groove faced the archaeal nucleosome core (Fig. 2), and although they contributed predominantly to the archaeal nucleosome stability, the flanking sequences apparently also played a role. The presence of four copies of the mouse genomic TATAAACGCC motif also resulted in increased affinity for HMfB relative to the control Litmus28 DNA, but archaeal nucleosomes assembled using this TA-containing repeat were less stable than those assembled using (TTTAAAGCCG) 4 , the MM301 repeat. The presence of (TATAAACGCC) 4 resulted in a ⌬⌬G of 276 Ϯ 109 cal mol Ϫ1 (Table I)   TA-containing 10-bp repeat relative to archaeal nucleosome assembly on the Litmus28 DNA.
HMfB-selected Sequences-Most of the DNA molecules selectively incorporated by HMfB into archaeal nucleosomes in vitro have sequences that conform to the (A/T) 3 NN(G/C) 3 NN motif (30). Hydroxyl radical footprinting of clone 1 and 20 DNAs demonstrated that A/T-rich regions, and specifically TA steps, that were centered around positions 20, 30, 40, and 50 within the 60-bp selected sequences were protected from cleavage by incorporation into archaeal nucleosomes (Fig. 4). As in eukaryotic nucleosome footprinting experiments (13,40), the sequences maximally protected from hydroxyl radical cleavage were displaced by ϳ2 bp on the two DNA strands, and as expected, identical footprinting patterns were obtained using either HMfA or HMfB. However, in view of their very similar structures and the lack of N-or C-terminal extensions, it was also expected that HMfA and HMfB would have very similar, high affinities for these HMfB-selected DNA molecules, but this was not the case. HMfA had consistently lower affinities for these molecules than HMfB (Table II). HMfB assembly into an archaeal nucleosome using clone 1 and 20 DNAs resulted in ⌬⌬G values of 1032 Ϯ 117 and 2213 Ϯ 161 cal mol Ϫ1 , respectively, relative to assembly on the Litmus28 DNA, whereas HMfA assembly using these DNA resulted in ⌬⌬G values of only 461 Ϯ 176 and 461 Ϯ 99 cal mol Ϫ1 , respectively, relative to archaeal nucleosome assembly using the control DNA.
Clone 1 and 20 DNAs were isolated after eight rounds of selection by HMfB incorporation into archaeal nucleosomes. The increases in stability conferred by these DNAs on archaeal nucleosomes assembled using HMfB are similar to those reported for eukaryotic nucleosome stabilization (average ⌬⌬G value of ϳ1700 cal mol Ϫ1 ) by DNA molecules isolated after nine rounds of selection by octamer incorporation into nucleosomes (10). However, the highest affinity complex investigated, formed by HMfB with clone 20 DNA, had an apparent K d of 2.7 Ϯ 0.5 nM in terms of HMfB tetramers, whereas subnanomolar K d values have been reported for salt-dependent dissociation constants of eukaryotic nucleosome core particles (41).
The Role of ␣-Helix 3 and Tetramer Formation in Archaeal Nucleosome Positioning-It was previously noted that the orientation of ␣3 relative to ␣2 differed slightly in the histone folds of HMfA and HMfB (22) and that this apparently resulted from differences in the C-terminal residues of the two histones (Fig. 6). These C-terminal residues were therefore exchanged, and the resulting variants, HMfA/B-␣3 and HMfB/A-␣3, formed complexes with DNA with properties typical of the ␣3 donor (Figs. 5 and 7 and Table II). The C-terminal regions of these archaeal histones apparently therefore participate in determining how the archaeal nucleosome tetramer interacts with DNA. The decreased affinity of HMfB/A-␣3 relative to HMfB for clones 1 and 20 DNA might be explained trivially by the decreased solution stability of this variant, but this would FIG. 7. CD spectra and agarose gel-shift assays of complexes formed by archaeal histones. CD spectra were obtained from 10 M solutions of the four archaeal histones, as indicated, dissolved in 25 mM MES, 50 mM K 2 SO 4 (pH 7) at 25°C. Aliquots of linear pBR322 DNA (100 ng) were incubated for 25 min at 25°C with 0 (Ϫ), 25, 50, 75, 100, 150, and 200 ng of each histone. As shown, the products were then separated by agarose gel electrophoresis and visualized by ethidium bromide staining.
FIG. 6. Sequence alignment and structure of an archaeal nucleosome containing clone 1 DNA. The sequence of clone 1 DNA protected from MN digestion by HMfB incorporation into an archaeal nucleosome is shown on the DNA strands assembled in the predominant rotational and translational positions established for clone 1 DNA. The sequences of HMfA and HMfB are shown aligned with the sequence of the histone fold of Xenopus H4 with regions that form the ␣-helices 1, 2, and 3 (␣1, ␣2, and ␣3) and loops 1 and 2 (L1 and L2) of the histone folds identified. The heavy overline and blue shading indicate the C-terminal residues that were exchanged to generate HMfA/ B-␣3 and HMfB/A-␣3. Conserved DNAbinding residues are boxed (32), and the side chains of these residues (Arg-10, Arg-19, and Lys-53) are shown in the figure in stick format on the surface of each HMfB monomer positioned appropriately for DNA binding. not readily explain the changes observed in the agarose gel mobility of the complexes formed by these two proteins with pBR322 DNA. Similarly, it seems very unlikely that a reduced stability argument could explain the increased affinity of HMfA/B-␣3 relative to HMfA for the HMfB-selected molecules. Given homology with the (H3ϩH4) 2 tetramer (1-3), it is most likely that a four-helix bundle involving two ␣2 and two ␣3 helices forms the dimer:dimer interface at the center of an archaeal nucleosome (Fig. 6) and that this interaction stabilizes and establishes the overall shape of the archaeal histone tetramer core (19,20). It has been argued that only a slight change at this interface generates histone tetramers that wrap DNA alternatively in either a positive or negative supercoil (20,25,26). It is certainly therefore conceivable that differences at this interface could also result in histone tetramers with structures/shapes sufficiently different to require different DNA sequences for optimum nucleosome assembly. If this is correct, then formation of alternative archaeal nucleosome tetramer cores could be used as a mechanism to regulate gene expression. All archaeal histones investigated to date form homodimers as well as heterodimers with other archaeal histones, including histones from other archaeal species (18,35). If this promiscuity in partnerships extends to tetramer formation, then there is an obvious opportunity to regulate gene expression by assembling different archaeal nucleosome cores with different shapes and, therefore, different sequence and positioning preferences. This could be an especially useful system for regulating gene expression in a species such as M. jannaschii with six different archaeal histones (20,35). It might also have provided the ancestral foundation for the evolution of the now widespread use of different histone fold partnerships to assemble multisubunit, regulatory complexes (42).