Archaeal Histone Tetramerization Determines DNA Affinity and the Direction of DNA Supercoiling *

DNA binding and the topology of DNA have been determined in complexes formed by >20 archaeal histone variants and archaeal histone dimer fusions with residue replacements at sites responsible for histone fold dimer:dimer interactions. Almost all of these variants have decreased affinity for DNA. They have also lost the flexibility of the wild type archaeal histones to wrap DNA into a negative or positive supercoil depending on the salt environment; they wrap DNA into positive supercoils under all salt conditions. The histone folds of the archaeal histones, HMfA and HMfB, from Methanothermus fervidus are almost identical, but (HMfA)2 and (HMfB)2 homodimers assemble into tetramers with sequence-dependent differences in DNA affinity. By construction and mutagenesis of HMfA HMfB and HMfB HMfA histone dimer fusions, the structure formed at the histone dimer:dimer interface within an archaeal histone tetramer has been shown to determine this difference in DNA affinity. Therefore, by regulating the assembly of different archaeal histone dimers into tetramers that have different sequence affinities, the assembly of archaeal histone-DNA complexes could be localized and used to regulate gene expression.

The histone fold, three ␣-helices (␣1, ␣2, and ␣3) 1 separated by two ␤-strand loops (L1 and L2), was first identified in the four nucleosome core histones and shown to direct their assembly into (H2AϩH2B) and (H3ϩH4) heterodimers (1). Histone folds have since been found to direct the pairwise assembly of subunits in many transcription regulators, including TFIID, CBF, CHRAC, PCAF, SAGA, STAGA, and TFTC, and considerable effort has been focused on establishing the determinants of specific histone fold partnerships (2)(3)(4)(5)(6)(7)(8). In contrast, relatively few studies have investigated how such histone fold dimers are further assembled and whether alternative higher order structures might be formed with different functions or properties. It has been shown that the interface between the two (H3ϩH4) histone dimers in an (H3ϩH4) 2 tetramer is flex-ible and that alternative tetramer structures can result that wrap DNA in either a negative or positive DNA supercoil (9,10). Nucleosomes containing the histone H3 variant alternatively designated CENP-A, CSE4p, CID, HCP-3, or SpCENP-A (11) assemble only at centromeric DNA. Mutagenesis and domain-swap experiments have revealed that this localization is similarly dependent on a difference at the histone dimer:dimer interface in (CENPϪAϩH4) 2 and (CSE4pϩH4) 2 versus (H3ϩH4) 2 tetramers (12)(13)(14)(15)(16)(17). Based on their crystal structures, nucleosomes assembled with histones from different species or histone variants do exhibit higher order structural differences, again most notably at the sites of histone dimer:dimer interactions (18)(19)(20)(21).
Archaeal histones are only histone folds with structures almost identical to the histone folds in eukaryotic histones and transcription factors (22). They are dimers in solution but must assemble into tetramers to bind DNA (23,24). By mutagenesis, the residues have been identified (25) that are directly responsible for DNA binding by HMfB, one of the two archaeal histones present in Methanothermus fervidus. By using a SELEX protocol, DNA molecules with intrinsically high affinity for HMfB have been isolated and characterized (26). Surprisingly, HMfA, the second archaeal histone in M. fervidus, does not bind with equally high affinity to these DNAs (27) even though HMfA and HMfB have 84% identical sequences, completely conserved DNA binding residues, and almost identical high resolution structures (22,25). Domain-swap experiments showed that this difference in DNA affinity results from a difference in the C-terminal ␣3 (27). This region of the histone fold was not previously thought to play any role in DNA binding but rather to be part of a 4HB that formed the interface between histone dimers when assembled into a tetramer. Such 4HBs, stabilized by both hydrophobic and ionic intermolecular interactions, form at the interfaces between the (H3ϩH4) histone dimers in the nucleosome and between the (dTAF II 42ϩdTAF II 62) histone fold dimers in Drosophila TFIID (Refs. 2, 18; see Fig. 1A). Specifically, conserved salt bridges are formed between a histidine in ␣2 (His-113 in histone H3, His-66 in dTAF II 42) and an aspartate in ␣3 (Asp-123 in histone H3, Asp-76 in dTAF II 42). Structural homologs (His-49 and Asp-59) are present in the histone folds of both HMfA and HMfB and, in fact, in all ϳ30 members of the HMfB family for which sequences are available (28). As part of the earlier mutagenesis study (25), HMfB variants with D59A, D59E, and D59N substitutions were generated, but these failed to accumulate as soluble proteins when synthesized in Escherichia coli, consistent with Asp-59 also being required to interact with Arg-52 to form a native monomer histone fold (18,22). Soluble His-49 variants were obtained, and these bound DNA, but the complexes they formed had reduced electrophoretic mobilities when compared with the complexes formed by wild type HMfB (25).
Given the domain-swap evidence that ␣3 participates in determining DNA affinity (27) and the gel-shift indication that His-49 variants form less compact complexes (25), we have constructed and report here the results of a detailed investigation of DNA binding and complex formation by a much wider spectrum of HMfB variants with residue substitutions introduced specifically at the histone fold sites predicted to participate in HMfB tetramer stabilization. Single and multiple substitutions have been made for His-49 and/or Leu-46 and Leu-62, the latter two residues being the HMfB structural homologs in ␣2 and ␣3, respectively, of Cys-110 and Leu-126 in H3 and Tyr-63 and Leu-79 in dTAF II 42 (see Fig. 1A). These hydrophobic residues also interact to contribute stability to the 4HBs at the centers of (H3ϩH4) 2 and (dTAF II 42ϩdTAF II 62) 2 tetramers, respectively (2,18). The results obtained are consistent with residues at these locations in HMfB contributing to DNA affinity through tetramer stabilization. Unexpectedly, the results have also revealed that these residues determine whether the resulting HMfB tetramer has the ability to wrap DNA in either a negative or positive supercoil. By construction and mutagenesis of HMfAϩHMfB and HMfBϩHMfA histone heterodimer fusions, we have also demonstrated directly that it is the structure formed at the site of histone fold dimer:dimer interaction that establishes the difference in DNA affinity exhibited by HMfA versus HMfB tetramers.

Site-directed Mutagenesis, Construction of Archaeal Histone Fusions, and Purification of Recombinant Archaeal Histones
Mutations were introduced into hmfB as previously described (25) using the QuikChange procedure (Stratagene, La Jolla, CA) with mutagenic oligonucleotide primers (sequences available on request) purchased from Ransom Hill Biosciences (Ramona, CA). The presence of the desired mutation(s) was confirmed by DNA sequencing. The mutated gene was introduced into the pKK223-3-based expression system developed for recombinant HMfB synthesis (29), and E. coli JM105 was transformed with the expression plasmids. To construct in-frame gene fusions, the hmfA and hmfB genes were first separately amplified in PCR in which two of the primers had common sequences that encoded the amino acid sequence (TAPDRRG) that links two histone folds in a protein encoded in the Halobacterium NRC-1 genome (30). These PCR products were mixed as the template, amplified, and joined in a PCR as described by Ho et al. (31) to obtain hmfA-hmfA, hmfB-hmfB, hmfA-hmfB, and hmfB-hmfA gene fusions. The encoded HMfA-ATAPDRRG-HMfA, HMfB-TAPDRRG-HMfA, HMfB-TAPDRRGA-HMfB, and HMfA-ATAPDRRGA-HMfB histone fusions were subsequently designated HMfAA, HMfBA, HMfBB, and HMfAB, respectively. For correct folding, the lengths of the linkers had to be adjusted, as shown, to accommodate the unique presence of Met-1 in HMfA and Lys-69 in HMfB (see Fig.  1A). The QuikChange procedure was also used to introduce mutations into the hmfA-hmfB and hmfB-hmfA gene fusions to obtain histone fold fusions with an H49E substitution in either the HMfA or HMfB component. The gene fusions were cloned into the pKK223-3-based expression system and transformed into E. coli XL1-Blue (Stratagene). Synthesis of the recombinant archaeal histones, archaeal histone variants, and archaeal histone fusions was induced by addition of 400 M isopropyl-␤-D-thiogalactopyranoside to growing E. coli cultures. The proteins synthesized were purified, quantitated, and their CD spectra measured using an AVIV 62A-DA spectropolarimeter (AVIV, Lakewood, NJ) as previously described (32). The CD spectra obtained for HMfA, HMfB, the HMfB variants, and the histone fusions were almost identical (available on request), consistent with very similar folded native structures.
Polyacrylamide Gel-shift Assays-Mixtures (10 l) containing 0.1 ng of a 32 P-labeled DNA molecule containing the 60-bp clone 20 sequence amplified from clone 20 plasmid DNA (26) with 10-bp or 25-bp primerderived flanking sequences, 1 ng of sss DNA, and increasing amounts of an archaeal histone were incubated 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 Tris borate/EDTA at 8 V/cm (33). The gels were dried and used to generate autoradiograms. From direct ␤-decay measurements (Packard Instant Imager, Meriden, CT), apparent K d values were calculated in terms of archaeal histone tetramer binding to DNA (27).

DNA Topology Assays
Relaxed circular pUC18 DNA molecules were generated from negatively supercoiled molecules by incubation with a DNA nicking-closing extract (NCE) prepared from chicken blood (34). Increasing amounts of an archaeal histone were incubated with aliquots (400 ng) of this DNA under low salt (10 mM Tris-HCl (pH 8), 2 mM Na 2 HPO 4 , 1 mM EDTA, 50 mM NaCl, or 50 mM potassium glutamate) or high salt (10 mM Tris-HCl (pH 8), 1 mM EDTA, 350 mM potassium glutamate) conditions for 30 min at 37°C. The complexes formed were exposed to the NCE and proteinase K, and the topologies of the resulting deproteinized DNA molecules were determined by one-and two-dimensional agarose gel electrophoresis as previously described (35).

Electron Microscopy
Glutaraldehyde-fixed archaeal histone-DNA complexes were prepared and visualized by EM as previously described (36,37). Using a 1313-bp DNA molecule designated fragment M (38), length measurements were made of histone-free DNA molecules and of a minimum of 50 DNA molecules containing from one to four clearly distinct archaeal histone-DNA complexes. Based on the apparent differences in length, the length of DNA assembled per complex was calculated.

RESULTS
Gel-shift Assays-Archaeal histone binding results in DNA compaction, and complexes formed with DNA molecules Ͼ2 kbp migrate faster during agarose gel electrophoresis than the histone-free DNA molecule (29,33). Using this assay, HMfB variants unable to bind DNA gave no gel shift. However, some variants, including L46I, L46V, H49A, L62I, and L62V, gave reduced gel shifts, suggesting the formation of less compact structures than those formed by the wild type HMfB (25). To pursue this apparent correlation with tetramer formation, HMfB variants were generated with a range of substitutions at one or more of these sites. All of these variants, L46A, L46C, H49D, H49E, H49N, H49K, L62A, L62C, L62M, L62Q, L62R, L62W, L46AϩH49A, L46AϩL62A, H49AϩL62A, and L46AϩH49AϩL62A, formed complexes with linear pBR322 DNA, but in all cases the complexes formed migrated more slowly through agarose gels than complexes formed by wild type HMfB (Fig. 1B). In some cases, the magnitude of the gel shift decreased at higher histone to DNA ratios, consistent with additional histone binding but no additional compaction. Three variants (L46F, H49D, and L62Y) were reported previously (25) to give wild type gel shifts. This was confirmed for the L46F (Fig. 1B) and L62Y variants but not for the H49D variant, which in repeated experiments gave a slightly reduced gel shift. Wild type gel shifts were also observed with several additional variants constructed with residue substitutions introduced at adjacent locations (K45A, K45E, A50G, R66I, R66M).
EM visualization has demonstrated that the agarose gel shift assay reports the net effect on electrophoretic mobility of complexes formed at multiple locations along a long DNA (36,37,39). However, complex formation by a single archaeal histone tetramer can be detected and apparent K d values measured by using a conventional polyacrylamide gel retardation assay with a shorter DNA (24,25,27). All of the Leu-46, His-49, and Leu-62 variants were so assayed using an 80-bp DNA molecule that contained the 60-bp clone 20 sequence, which has inherently high affinity for HMfB (26). Examples of the results obtained are shown in Fig. 1C. All of the variants formed complexes with this DNA but, in many cases, only at much FIG. 1. Histone fold tetramer-DNA complex, histone fold sequences, and gel-shift assays. A, increasingly detailed views of the 4HB formed by the H3 histone folds in an (H3ϩH4) 2 tetramer at the center of a nucleosome (18). The 4HB is viewed from above the central dyad of the nucleosome in the lower section of the left panel and in the right panel. To gain this perspective, the complex illustrated in the upper section of the left panel was rotated by ϳ90°toward the reader. Below the illustrations, the histone fold sequences of HMfB, histone H3, and dTAF II 42 are aligned with the regions forming ␣-helices (␣1, ␣2, ␣3), loops (L1, L2), and the residues in HMfA that differ from those in HMfB identified above the HMfB sequence. The C-terminal residue of HMfA is Lys-68, and the absence of an equivalent of Lys-69 in HMfB is indicated by the symbol #. The locations of five residues that are present in the eukaryotic histone folds that have no archaeal histone counterparts are indicated by hyphens. Conserved residues are identified by asterisks. The side chains of the H3 residues boxed in the sequence alignment are shown in the illustrations above, positioned and interacting as predicted to contribute to tetramer stabilization. The histone fold homologs in dTAF II 42 (2) and HMfB (22)  higher histone to DNA ratios than required for complex formation by wild type HMfB. At the highest ratios tested (Ͼ500 histone variant tetramers/DNA molecule), only a small percentage of the DNA was assembled into complexes by the H49D and H49E variants (Fig. 1C), and these variants failed to form detectable complexes when incubated with other ϳ80-bp DNA molecules that lacked inherently high affinity for HMfB.
Topology Assays-Archaeal histones form complexes in which the DNA is constrained alternatively in a negative or positive supercoil (35,40). Under low salt conditions (ϳ50 mM K ϩ ) and at low histone to DNA ratios, complex formation on relaxed circular pUC18 molecules results in negative DNA supercoiling, but this changes to positive DNA supercoiling with additional histone binding ( Fig. 2A). This change from negative to positive supercoiling occurs only under low salt conditions. Under higher salt conditions (Ͼ300 mM K ϩ ), more similar to the high salt content of the M. fervidus cytoplasm (41), HMfB binding to relaxed pUC18 DNA molecules results in only negative supercoiling of the DNA at all histone to DNA ratios (40). The topology of the pUC18 DNA in complexes formed by the Leu-46, His-49, and Leu-62 variants was determined under both low and high salt assembly conditions. Only the L46F, L62V, and L62Y variants retained the ability to form complexes in which the DNA was either negatively or positively supercoiled. In contrast, the L46A, L46C, L46I, L46Q, L46V, H49A, H49D, H49E, H49K, L62A, L62C, L62I, L62Q, L62R, L62W, L46AϩL62A, L46CϩL62C, L46AϩH49A, H49AϩL62A, and L46AϩH49AϩL62A variants were never observed to form complexes in which the DNA was negatively supercoiled. At all histone to DNA ratios and under both low and high salt conditions, these variants only formed complexes in which the pUC18 DNA was positively supercoiled (Fig. 2).
HMfA versus HMfB Histone Fold Tetramer Formation and DNA Affinity-Both the ␣3 domain-swap evidence (27) and the Leu-46, His-49, and Leu-62 variant DNA affinity results (Fig.  1C) argued that a difference at the dimer interface in (HMfA) 2 : (HMfA) 2 versus (HMfB) 2 :(HMfB) 2 tetramers might be responsible for the difference in HMfA versus HMfB affinity for the clone 20 sequence. To further investigate this, the HMfAB and HMfBA histone heterodimer fusions were constructed and H49E substitutions introduced into either the HMfA or HMfB component. (HMfB) 2 homodimers with an H49E substitution in both monomers were known to be essentially incapable of forming complexes with this DNA (Fig. 1C). However, with only one H49E substitution, it was reasoned that the histone dimer fusion variants HMfAB(H49E), HMfA(H49E)B, HMfBA(H49E), and HMfB(H49E)A might still bind DNA. To do so, two such molecules would interact to form a histone fold tetramer stabilized by a 4HB formed by the histone folds that retained the wild type His-49 residues. If the assembly of this structure determined the overall affinity of the resulting histone fold tetramer for DNA, then the HMfAB(H49E) and HMfB(H49E)A variants would form complexes with HMfA-like affinity. The HMfA(H49E)B and HMfBA(H49E) variants, in contrast, would have higher HMfBlike affinity for the clone 20 DNA. This experiment is illustrated in Fig. 3. The results obtained with both the wild type heterodimer fusion controls, the heterodimer fusion variants, and the HMfAA and HMfBB homodimer fusions are shown in Fig. 4. As anticipated, the HMfAA and HMfBB homodimer fusions formed complexes with DNA affinities, electrophoretic mobilities, and topologies essentially the same as those formed by (HMfA) 2 and (HMfB) 2 homodimers, respectively. The wild type HMfAB and HMfBA heterodimer fusions also had DNA affinities and formed complexes with mobilities and topologies essentially the   Table in Fig. 4B). The HMfAB(H49E) and HMfB(H49E)A fusion variants also formed complexes with pBR322, which migrated through agarose gels with mobilities typical of HMfAcontaining complexes (Fig. 4A). In contrast, when an H49E substitution was introduced into the HMfA component, the resulting HMfA(H49E)B and HMfBA(H49E) fusion variants retained essentially HMfB affinity for clone 20 DNA (see Table, Fig. 4B). They also still formed complexes with pBR322 with mobilities typical of complexes formed by (HMfB) 2 and HMfBB (Fig. 4A). In all cases, the complexes formed by the histone fusion variants with the clone 20 DNA migrated to form somewhat more diffuse bands than those formed by the wild type fusions. The HMfB(H49E)A variant apparently formed two complexes with different mobilities (Fig. 4B), suggesting assembly at two predominant translational positions.
Because the initial histone fold dimer:dimer assembly determined DNA affinity (Fig. 4) and apparently also directed the higher order assembly events assayed by agarose gel electrophoretic mobility, it was expected that the HMfAB(H49E) and HMfB(H49E)A variants would have HMfA-like and the HMfA(H49E)B and HMfBA(H49E) variants would have HMfBlike DNA supercoiling properties. This was the case in terms of the number of supercoils introduced into a pUC18 DNA molecule at different histone to DNA ratios. However, unlike the wild type histones and histone fusions, binding by these his-  Fig. 4), the HMfBA(H49E) and HMfB(H49E)A fusion variants assembled 4HBs as illustrated, stabilized by interactions (see BϭB and AϭA) that involved the histone folds that retained the wild type His-49 residues. The wild type HMfBA fusion apparently assembled a 4HB involving the ␣2 and ␣3 helices of the HMfB histone folds (BϭB, blue), but as illustrated in the box three additional 4HB assemblies are theoretically possible involving the ␣2 and ␣3 helices of only the HMfA histone folds (AϭA) or involving helices from both the HMfA and HMfB histone folds (AϭB, BϭA). With the addition of a third HMfBA(H49E) or HMfB(H49E)A molecule, several alternative 4HB assemblies are possible (AϭA, AϭB, BϭB, and/or BϭA) that would involve either one His-49 and one Glu-49, or two Glu-49 residues as illustrated. To facilitate visualization, the added third molecule is shown displaced from the real circular configuration. In all cases, the resulting tetramer wraps the DNA molecule in a positive supercoil, whereas the further polymerization of wild type HMfBA molecules under physiologically relevant high salt conditions (41) results in negative supercoiling (Fig. 4C). tone variants resulted in complexes in which the pUC18 DNA was only positively supercoiled at all histone to DNA ratios and under both low and high salt assembly conditions (Fig. 4C).

Histone Fold Tetramer Formation and DNA Affinity-Based
on homology with eukaryotic histone folds (see Fig. 1A), Leu-46, His-49, Asp-59, and Leu-62 were identified as residues likely to stabilize a 4HB that formed the dimer:dimer interface in an 2(HMfB) 2 tetramer. To bind DNA, (HMfB) 2 dimers must form a tetramer (23,24), and consistent with reduced ability to form a stable tetramer, almost all of the Leu-46, His-49, and Leu-62 variants investigated had reduced affinity for an 80-bp DNA (Fig. 1C). The three variants with wild type affinity had large conservative side chain substitutions (L46F and L62Y) or a cysteine (L46C) residue at precisely the location of Cys-110 in histone H3 (42). Presumably, the side chains of these residues must still provide fully effective hydrophobic stabilizing interactions within the 4HB formed between the two (HMfB) 2 dimers. The His-49 variants were consistently the most defective in DNA binding, suggesting that the histidine-aspartate interactions lost by these variants are of predominant importance in stabilizing a histone fold tetramer. If this is correct, the presence of a histidine near the C terminus of ␣2 could predict the likelihood of that histone fold participating in stabilizing a histone fold tetramer. The hTAF II 31 human homolog of dTAF II 42 does, for example, have such a histidine that could stabilize (hTAF II 31ϩhTAF II 80) 2 tetramers, whereas the predicted yeast homolog, yTAF II 17, lacks an appropriately positioned histidine, arguing against (yTAF II 17ϩyTAF II 60) 2 tetramer stabilization (3, 6 -8). The aspartate in ␣3 is conserved in virtually all histone folds (42), but because this is also required to stabilize the histone fold monomer (18,22), the presence of this residue is more a predictor of a histone fold than of tetramerization.
Each histone fold dimer binds the DNA at three locations in the nucleosome over an ϳ28-bp region (18 -21) and, as this predicts, complexes containing an archaeal histone tetramer protect ϳ60 bp from micrococcal nuclease digestion (43). Based on EM measurements, complexes in which the DNA is fully circularized involve a minimum of ϳ90 bp (Fig. 3), but with additional archaeal histone polymerization, complexes are formed that protect not only ϳ90 but also ϳ120 and ϳ180 bp (44 -46). Based on the agarose gel-shift results, all the Leu-46, His-49, and Leu-62 variants (Fig. 1B) retained some ability to assemble such larger complexes, but the complexes formed were less compact than those formed by wild type HMfB. In some cases, at relatively high histone to DNA ratios, additional histone binding occurred but resulted in no additional compaction (Fig. 1B).
The affinities of the HMfAB(H49E), HMfA(H49E)B, HMfB(H49E)A, and HMfBA(H49E) fusion variants for clone 20 DNA confirmed that the difference in affinity of HMfA versus HMfB for this DNA sequence reflected a difference at the site of histone fold tetramerization (Fig. 3). Because the HMfB components of the wild type HMfAB and HMfBA fusions would be expected to form histone fold tetramers with higher affinity for the clone 20 sequence than those formed by the HMfA components, it was anticipated and observed (Fig. 4B) that the wild type fusions had HMfB-like affinities for the clone 20 DNA. Their assembly to form complexes with pBR322 DNA with HMfB-like properties was less predictable but is consistent with EM observations. HMfB has been seen to assemble more complexes than HMfA at the same histone to DNA ratios on all DNA molecules so far investigated from ϳ1 to 5 kbp (36, 39). 2 HMfB assembly apparently has fewer sequence or structural demands, which results in the assembly of more complexes and therefore in more compaction and in complexes that migrate faster through agarose gels (Fig. 4A).
Histone Fold Tetramerization and the Direction of DNA Supercoiling-The DNA appears sharply kinked in complexes formed at low archaeal histone to DNA ratios (Ref. 36 and Fig.  3), consistent with partial circularization around a histone tetramer. The assembly of such complexes by the wild type proteins introduces negative superhelicity into a circular DNA (35), presumably because the DNA is held across the dimer: dimer interface in an orientation that, when extended to the whole circular DNA, results in net negative supercoiling of the molecule (Fig. 2A). The regions of the DNA entering and exiting such a structure do not necessarily interact, but with additional histone binding, the entering and exiting DNA must be forced into close proximity as the DNA is fully circularized. Under low salt conditions, this must overcome DNA-DNA repulsive forces which, based on the experimental results ( Fig.  2A), is accomplished by the DNA being wrapped in a positive supercoil. Under higher salt conditions, cation ion binding must shield these repulsive forces, and the negative supercoiling initiated by the initial tetramer-DNA interaction can continue and be extended with additional histone polymerization and DNA circularization (Fig. 2B). Surprisingly, the majority of the Leu-46, His-49, and Leu-62 variants formed complexes in which the pUC18 DNA was only positively supercoiled, regardless of the assembly conditions (Fig. 2, A and B). This result resembles the observations made with (H3ϩH4) 2 tetramers that had large chemical adducts attached to the Cys-110 residues of the H3 histone folds (10), the homolog of Leu-46 in HMfB. These tetramers could wrap DNA in only a negative or positive supercoil, and it was argued that the bulky chemical adducts blocked the dimer:dimer interface movement needed to switch from negative to positive supercoiling (10,47). Possibly, the Leu-46, His-49, and Leu-62 variants have similarly lost structural flexibility and now form HMfB tetramers locked in structures that can only direct positive supercoiling. However, with a weakened dimer:dimer interaction, it seems more likely that these variants form tetramers that lack the strength to hold the DNA in a more demanding negative toroidal supercoil. The DNA molecule in an archaeal histone-DNA complex is overwound at ϳ10 bp/turn, in effect positively supercoiled (27,44). Positive toroidal supercoiling should therefore result in the absence of a strong countering histone dimer:dimer interaction.
The above explanation for the DNA supercoiling phenomena observed with the wild type proteins predicts that both the wild type HMfAB and HMfBA fusions and variant histone fusions should form complexes at low histone to DNA ratios that introduce negative superhelicity into circular pUC18 DNA. This was observed for the wild type fusions, but the HMfAB(H49E), HMfA(H49E)B, HMfB(H49E)A, and HMfBA(H49E) fusion variants only formed complexes in which the DNA was positively supercoiled at all histone to DNA ratios and under both low and high salt conditions (Fig. 4C). The DNA affinity results (Fig. 4B) argue that the first histone fold tetramer formed by the assembly of two of these fusion variant molecules is stabilized by a wild type 4HB, but apparently alone this is insufficient to hold the DNA in negative supercoil. Because the complexes formed by the wild type HMfAB and HMfBA fusions do have this ability, the additional His-49 residues present in these complexes located ϳ2.5 helical turns on either side of the central 4HB (locations occupied by Glu-49 residues in complexes formed by the fusion variants (see Fig. 3)) must also contribute to holding the DNA in a negative supercoil.
Conclusions-That very similar archaeal histones might form complexes with different properties was suggested by the observation that the ratio of HMfA to HMfB changed in M. fervidus cells under different growth conditions (37). The discovery that HMfA had much lower affinity than HMfB for the clone 20 sequence (27) added strong support to this idea, and the widely distributed and conserved presence of histone variants in many Eukaryotes (42) is also consistent with this concept. It is already well established that incorporation of the CENP-A family of histone H3 variants results in nucleosomes that assemble specifically at the centromere (11)(12)(13)(14)(15)(16)(17). Although unique functions have not yet been assigned to these nucleosomes, this positioning is entirely consistent with the demonstration here that very similar histone fold dimers can assemble into tetramers with structures so different that they have different DNA affinities. Regulating the assembly of alternative histone fold tetramers could therefore provide a mechanism to regulate gene expression.