Determinants of DNA binding and bending by the Saccharomyces cerevisiae high mobility group protein NHP6A that are important for its biological activities. Role of the unique N terminus and putative intercalating methionine.

The non-histone proteins 6A/B (NHP6A/B) of Saccharomyces cerevisiae are high mobility group proteins that bind and severely bend DNA of mixed sequence. They exhibit high affinity for linear DNA and even higher affinity for microcircular DNA. The 16-amino acid basic segment located N-terminal to the high mobility group domain is required for stable complex formation on both linear and microcircular DNA. Although mutants lacking the N terminus are able to promote microcircle formation and Hin invertasome assembly at high protein concentrations, they are unable to form stable complexes with DNA, co-activate transcription, and complement the growth defect of Deltanhp6a/b mutants. A basic patch between amino acids 13 and 16 is critical for these activities, and a second basic patch between residues 8 and 10 is required for the formation of monomeric complexes with linear DNA. Mutational analysis suggests that proline 18 may direct the path of the N-terminal arm to facilitate DNA binding, whereas the conserved proline at position 21, tyrosine 28, and phenylalanine 31 function to maintain the tertiary structure of the high mobility group domain. Methionine 29, which may intercalate into DNA, is essential for NHP6A-induced microcircle formation of 75-bp but not 98-bp fragments in vitro, and for full growth complementation of Deltanhp6a/b mutants in vivo.

The high mobility group (HMG) 1 proteins are a family of heterogeneous chromatin-associated DNA-binding proteins in eukaryotic cells. They are the most abundant non-histone proteins found in the nucleus and are divided into three classes: HMG-I/Y, HMG14/17, and HMG1/2 (1,2). These proteins were classically recognized by their high electrophoretic mobility in polyacrylamide gels and acid solubility. The HMG-I/Y class of proteins function as accessory transcription factors, whereas the HMG14/17 class are associated with nucleosomes. The HMG1/2 class of proteins contain the HMG DNA binding domain and are present at a level of about 1 copy/2-3 nucleosomes (3). The HMG domain is a 70 -80-amino acid region consisting mainly of hydrophobic and charged residues with a few highly conserved aromatic residues (4).
Members of the HMG1/2 class can be further divided into two subfamilies based upon the number of HMG domains, their DNA sequence specificity, and their evolutionary relationship (5). The sequence-specific HMG1/2 proteins contain a single HMG domain, which are usually restricted by cell type and interact with relatively high affinity to a specific DNA sequence. These proteins include the human sex-determining factor SRY (6,7), the lymphoid enhancer-binding factor LEF-1 (8), and the T-cell factor TCF-1 (9). The other subfamily of HMG1/2 proteins are ubiquitously expressed and bind to DNA with structural specificity but little or no sequence specificity. One group of the non-sequence-specific DNA-binding proteins contain multiple HMG domains such as the abundant HMG1 and HMG2 proteins (10), human nucleolar transcription factor hUBF (11), mitochondrial transcription factor mTF-1 (12), and yeast ARS-binding protein ABF-2 (13). Other non-sequencespecific HMG1/2 proteins found in yeast (14,15), plants (16), insects (17,18), and protozoa (19,20) contain only a single HMG domain and, often, an accessory basic and/or acidic domain.
The HMG domain of all these proteins probably fold into an L-shaped region of three ␣-helices, as shown by the NMR structures of rat HMG1 domain B, SRY, LEF-1, HMG1 domain A, HMG-D, and SOX4 (21)(22)(23)(24)(25)(26)(27). The domain contains a primary hydrophobic core at the vertex of the L-shaped structure formed by conserved aromatic residues from the three ␣-helices. The NMR structures of the two sequence-specific HMG proteins, SRY and LEF-1, include their DNA recognition sequence (23,24). These structures show the DNA to be greatly distorted in the region of protein contact, with an overall bend of 80°and 120°for SRY and LEF-1, respectively. The DNA is severely underwound, resulting in a widened and shallow minor groove and a highly compressed major groove. In conjunction with the helical underwinding, large positive roll angles are induced by numerous DNA-protein contacts, which include a partial intercalation of an amino acid side chain into the minor groove of the DNA.
The sequence-specific HMG structures provide a basis for understanding how the non-sequence-specific HMG1/2 proteins interact with high specificity to distorted DNA containing bends, cruciforms, or DNA kinked by cisplatin (28 -32). Recently, it has been demonstrated that a subgroup of non-sequence-specific HMG1/2 proteins, which contain only one HMG box, can bind with higher affinity to linear DNA than HMG1/2 but exhibit only a modest preference for cruciform DNA (30,33,34). It has become increasingly clear that the addition of basic amino acids to a single minimal HMG box enhances the bending capacity and affinity for DNA (24,(35)(36)(37). However, despite the fact that some of the non-sequence-specific single HMG box proteins rival the DNA binding affinities of sequence-specific proteins, the exact mode of DNA binding by the non-sequencespecific HMG proteins has not yet been determined.
The biological functions of HMG1/2-like proteins are just beginning to be elucidated. They can function as architectural factors through their ability to strongly distort DNA structures. In this regard, they have been shown to promote assembly of specialized recombination complexes derived from prokaryotic systems (38 -40). HMG1/2 proteins and their homologs can substitute for prokaryotic chromatin proteins to condense DNA in vivo (33) and can facilitate nucleosome assembly and disassembly in vitro (42)(43)(44)(45). Moreover, efficient activated transcription of certain genes requires HMG proteins, which, in part, may be related to their ability to promote assembly of preinitiation complexes (41,46). On the other hand, some in vitro studies have found that HMG proteins inhibit transcription (47)(48)(49).
The yeast system provides a genetic tool to investigate the biological roles of HMG proteins and to correlate in vitro activities with in vivo functions. A number of HMG box proteins have been reported in Saccharomyces cerevisiae including the non-histone proteins 6A/B (NHP6A/B) (14) and more recently the high mobility group proteins HMO1/2 (15). NHP6A/B are closely related 11-kDa proteins that contain a single HMG box. The HMG box of NHP6A is 45% identical to rat HMG1 box B and 80% identical to the HMG box of NHP6B (Fig. 1A). NHP6A/B are unique in that they both contain a highly basic amino acid region that precedes the HMG box. These proteins have a higher affinity for and bend DNA more efficiently than mammalian HMG1/2 (33). In this report, we investigate the importance of the N-terminal basic segment for high affinity DNA binding. Neither NHP6A nor NHP6B is essential since ⌬nhp6A and ⌬nhp6B mutants grow normally, but the double mutant grows slowly at 30°C and is not viable at high temperatures. The ⌬nhp6A/B mutants display a variety of morphological changes and are defective in activated transcription of a subset of genes (41,50). We show in this paper that the Nterminal segment of NHP6A is critical for its DNA binding activities and functional properties. NHP6A mutants containing substitutions in other selected regions have also been analyzed with respect to their in vitro and in vivo properties.

EXPERIMENTAL PROCEDURES
Construction of NHP6A Mutants-The NHP6A mutants were constructed by direct cloning of PCR products using mutant oligonucleotide primers or by site-directed mutagenesis using the method of Kunkel (51). The sequence of the oligonucleotides used to generate the mutations is given in Table I. In several cases (P21A, M29A, and M29D), mutant PCR products were cloned by a three-way ligation using internal restriction sites (Table I). pRJ1228 (pET11a-NHP6A; Ref. 33) was used as the template for PCR and the vector for reconstructing mutant NHP6A genes. pRJ1340 and pRJ1341 were generated by cloning into pBS KS ϩ and pBS KS Ϫ , respectively, a PCR product obtained using a 5Ј NHP6A primer containing an EcoRI/NdeI site and a 3Ј NHP6A primer containing a BamHI site. Single-stranded DNA for site-directed mutagenesis was prepared from pRJ1340 and pRJ1341 using CJ236 (dut ung; Ref. 52). The mutant genes were subsequently transferred using the NdeI and BamHI sites into pET11a for protein overexpression. Each mutant gene was sequenced in its entirety.
The NHP6A mutations were introduced into yeast as follows. The 300-bp control region upstream of the NHP6A gene was obtained by PCR of genomic S. cerevisiae DNA using primers containing XhoI and NdeI sites engineered at the 5Ј and 3Ј ends, respectively. After digestion with XhoI and NdeI, the product was inserted into pRJ1340 to give pRJ1342, which links the NHP6A promoter to the NHP6A gene flanked by NdeI and BamHI restriction sites. This NHP6A region was then subcloned between the XhoI and BamHI sites into pRS314 (TRP1 CEN6 ARSH4; Ref. 53) to create pRJ1364. Different mutant NHP6A genes were substituted in place of the wild-type gene using the unique NdeI and BamHI sites. In addition, HMG1 box B and box BЈ were obtained by PCR of pT7-RNHMG1 (54) with NdeI and BamHI engineered ends. After digestion with NdeI and BamHI, these products were ligated into pRJ1364, positioning them downstream of the NHP6A promoter.
Protein Expression and Purification-Recombinant proteins were expressed from the pET11a-derivatives in RJ1878 (BL21 (DE3) hupA::cm hupB::km; Ref. 33). NHP6A synthesis was induced for 3 h at 37°C in LB when the cells reached an A 600 ϭ 0.6 by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside. Two liters of cells were disrupted by sonication in 1/10 volume of 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM DTT, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride. The extract was clarified by centrifugation at 30,000 ϫ g, and the NaCl concentration was increased to 1 M. Polyethyleneimine (Sigma) was added to 0.3%, and the nucleic acids were removed by centrifugation at 20,000 ϫ g. Residual polyethyleneimine was removed by batch chromatography with 20% (v/v) cellulose phosphate P-11 (Whatman), and the supernatant was dialyzed overnight against 0.05 M buffer A (20 mM Tris-HCl (pH 7.5), 1 mM DTT, 1 mM EDTA, 10% glycerol, plus 0.05 M NaCl). The dialysate was passed through a 4-ml S-Sepharose (Pharmacia) column equilibrated in the same buffer, and the NHP6A protein was eluted in a 50-ml linear gradient from 0.05 M to 1.0 M NaCl in buffer A. Fractions containing the NHP6A protein were pooled and subject to 2% trichloroacetic acid precipitation at 0°C for 30 min to remove contaminating proteins. After centrifugation for 30 min at 30,000 ϫ g, the supernatant was adjusted to 10% trichloroacetic acid, and the homogeneous NHP6A was recovered by centrifugation as before. The precipitate was washed with acetone, dried briefly, resuspended in buffer B (20 mM HEPES (pH 7.5), 0.1 M NaCl, 1 mM DTT, 1 mM EDTA, and 50% glycerol) and dialyzed overnight in the same buffer. Wagner et al. (55) reported that trichloroacetic acid precipitation affected DNA binding by HMG1, but we have been unable to detect any difference between NHP6A purified using trichloroacetic acid precipitation or under entirely native conditions using multiple chromatography steps. Proteins concentrations were quantitated by laser densitometry (Molecular Dynamics) of SDS-polyacrylamide gels stained with Coomassie Blue using a titration of lysozyme as a standard. Values obtained were within 1-3% of the amounts determined by quantitative amino acid analysis (UCLA Protein Microsequencing Facility). Gel Mobility Shift, Ligase-mediated Circularization, and Hin DNA Inversion Assays-Labeled DNA fragments for gel mobility shift and ligation assays were obtained using PCR reactions containing [␣-32 P]dATP and pRJ551-44, pRJ551-66, or pRJ551-76 as described (38,56). After digestion with EcoRI, fragments of lengths 66, 75, and 98 bp containing EcoRI cohesive ends, respectively, were purified in 10% polyacrylamide gels. Circular DNA was formed from reactions containing 8 ng of linear fragment, 40 ng of NHP6A, and 20 units of T 4 ligase that were incubated at 30°C for 1 h under the conditions described previously (33). The products were then digested with 100 units of exonuclease III for 60 min, extracted with phenol/chloroform (1:1, v/v) and precipitated with ethanol resulting in Ͼ90% monomer circle. Gel mobility shift and ligase-mediated circularization assays were performed as described in Ref. 33 and electrophoresed on 6% polyacrylamide gels (29:1 acrylamide:bisacrylamide). Hin-catalyzed DNA inversion reactions were performed as described previously using pMS551-83, which contains 83 bp between the centers of hixL1 and the proximal Fis binding site in the enhancer (56).
Stoichiometry of NHP6A-DNA Complexes-NHP6A-HMK, containing the heart muscle kinase recognition sequence (RRASV) fused to the C-terminal end of NHP6A, behaved identically to the wild-type in DNA binding assays (data not shown). 10 g of NHP6A-HMK was labeled by incubating with 5 units of bovine heart muscle kinase (Sigma) and 50 Ci of [␥-32 P]ATP (Ͼ6000 mCi/mmol; Andotek) in 20 mM HEPES (pH 7.5), 100 mM NaCl, and 12 mM MgCl 2 for 1 h at 37°C. The reaction was quenched with 10 mM EDTA, precipitated with trichloroacetic acid as described above, and the 32 P-labeled NHP6A-HMK was resuspended in buffer B. The specific activity of the protein under potential gel quenching conditions was determined by polymerizing a known amount in a 6% polyacrylamide gel plug (29:1 acrylamide:N,N-bisacryloylcystamine; Sigma). The plug was incubated with 50 mM ␤-mercaptoethanol for 30 min at 50°C to dissolve the disulfide cross-links and counted in scintillation fluid (EcoSint; National Diagnostics). 98-bp DNA fragments were prepared as described above. The DNA concentration was determined by a Hoefer TK100 fluorometer using Hoechst dye 33258 (Sigma) and by quantitation (ImageQuant, Molecular Dynamics) of digital images of ethidium bromide-stained polyacrylamide gels in which an aliquot of the fragment was electrophoresed together with varying concentrations of pBR322 digested with HaeIII. Both methods gave the same results. Gel mobility shift assays were performed as described above, except that electrophoresis was in 6% 29:1 acrylamide: N,N-bisacryloylcystamine gels. The gels were stained with SybrGreen (Molecular Probes) and visualized with a Hitachi FMBIO II Fluorimager. The first and second complex bands were excised, the gel slice dissolved with ␤-mercaptoethanol as described above, and counted in scintillation fluid. The molar amount of protein in each complex was calculated from its specific activity; the molar amount of DNA was determined by comparison to varying amounts of the DNA fragment electrophoresed on the same gel.
For quantitation of NHP6A levels, transformants were grown in 25 ml of liquid selective media until the A 600 was 1.0. The cells were washed with distilled water and resuspended in 5 ml of 50 mM sodium phosphate buffer (pH 7.0) and 1 mM phenylmethylsulfonyl fluoride. The yeast were lysed by passing the mixture three times through a French press. After a brief incubation at 100°C, the extract was cleared by centrifugation at 30,000 ϫ g for 30 min. 200 g of total protein was trichloroacetic acid-precipitated and electrophoresed on a 15% SDSpolyacrylamide gel. The gel was immunoblotted with anti-NHP6A (41) and anti-glucose-6-phosphate dehydrogenase (Sigma) as a loading control.
CUP1 transcription was assayed in RJY6024 (RJY6012 ϩ pL⌬3D241 containing the CUP1-lacZ reporter; Refs. 41 and 58). At least 2-4 transformants were grown in SD media under selective conditions for 2-4 days and then subcultured into fresh selective media and grown for 6 h. 1 mM CuSO 4 was then added, and ␤-galactosidase activity measured after an additional 2 h at 30°C in which the final culture A 600 was 0.5-1.0.
In Vitro Transcription-Transcription reactions (25 l) were performed using 130 g of yeast nuclear extracts in the presence of the acidic activator GAL4-VP4, NHP6A, and 100 ng of G 5 E 4 T as the DNA template. mRNA levels were detected by primer extension. Reaction conditions, GAL4-VP4, and G 5 E 4 T have been described previously (41,59).
Physical Properties-Circular dichroism spectra from 190 to 240 nm was measured with an AVIV 62ADS circular dichroism spectrophotometer with 20 iterations at 25°C. Proteins were dialyzed into 10 mM sodium phosphate buffer (pH 7.4) and protein concentrations determined as above. Secondary structure characteristics were calculated by the self-consistent method using DICROPROT (60). 2 Thermal denaturation curves were obtained under the same conditions at 222 nm. The temperature was varied from 10 -90°C with an incubation time of 5 min for every 5°C.

DNA Binding Properties of Wild-type NHP6A
NHP6A forms stable complexes on DNA of varying sequences as revealed by polyacrylamide gel electrophoresis ( Fig.  2A; Ref. 33). With increasing concentrations of NHP6A, a discrete set of higher order complexes are obtained. 32 P-Labeled NHP6A was used to determine the stoichiometry of binding in the first and second complexes formed on a 98-bp DNA fragment as described under "Experimental Procedures." Quantitation of four to eight individual NHP6A-DNA complexes from separate experiments gave average molar ratios of 0.92 Ϯ 0.2 for the first retarded complex and 2.2 Ϯ 0.4 for the second complex. Therefore, NHP6A binds to DNA as a monomer in a stepwise manner. The binding constant (K D ) for the initial complex is 100 nM, a value that is constant on DNA of different sequences and lengths (23-300 bp).
The ability of NHP6A to bend DNA is most directly demonstrated by ligase-mediated circularization assays where the wild-type protein can form up to 70% monomer circles on DNA substrates as short as 66 bp (33). By measuring the amount of wild-type NHP6A required to generate the half-maximum yield of 98-bp circles, we calculate a K circle of 1.5 ϫ 10 Ϫ8 M (Fig. 3B). NHP6A binds much more tightly to curved DNA of mixed sequence than to linear DNA. This is shown in Fig. 4A where binding to a 98-bp microcircle was assayed by polyacrylamide gel electrophoresis. Two discrete complexes are formed at low concentrations of NHP6A, followed by the formation of higher order species with increasing protein concentrations. Gel mobility shift assays on 75-bp microcircles also generates two high affinity complexes, but three high affinity complexes are formed on 66-bp microcircles (Fig. 4, F and G). The number of high affinity complexes were identical regardless of whether NHP6A was added to purified microcircles or whether the products of the NHP6A ϩ DNA ligase reactions used to generate the microcircles were directly analyzed by native gel electrophoresis (data not shown). The binding constant for these curved DNA molecules is ϳ1.5 nM. Binding to the pre-bent substrates is extremely stable, as revealed by the resistance of NHP6A-microcircle DNA complexes to added competitor DNA (Fig. 4H). The addition of 2.5 mg/ml salmon sperm DNA (corresponding to a 2,500,000:1 w/w ratio of competitor to microcircular DNA) was unable to remove NHP6A from the two high affinity sites on the 98-bp microcircle, but 5.0 g/ml (500:1 w/w ratio of competitor to linear DNA) was sufficient to dissociate most of the prebound NHP6A from a linear substrate.

The N Terminus Is Necessary for Efficient Binding and Bending of DNA
The 94-amino acid NHP6A protein contains a 16-amino acid region located N-terminal to the minimal HMG domain. This segment contains two blocks of basic residues: KKR between residues 8 and 10 and RKKK between 13 and 16 ( Fig. 1). To determine the importance of these amino acids for DNA binding and NHP6A function, three different truncations of NHP6A were constructed: a deletion of the entire N terminus ⌬(2-16), a deletion that removed the first block of basic residues ⌬(2-12), and a deletion that retains both blocks of basic residues ⌬(2-7).
Deletion of the entire N terminus of NHP6A abolished the high affinity of the protein for linear DNA with no distinct DNA-protein complexes being formed (Fig. 2D). We estimate that this mutant binds approximately 600-fold poorer than wild-type protein by gel mobility shift assays. This resembles the low affinity association of HMG1 to linear DNA observed under similar assay conditions (29,33,61). NHP6A ⌬(2-16) also bound poorly to microcircles (Fig. 4C). This unstable DNA association by the minimal HMG box of NHP6A represents a functional HMG interaction and not a nonspecific association since ⌬(2-16) was able to induce microcircle formation. As shown in Fig. 3A, NHP6A ⌬(2-16) converted 40% of a 98-bp DNA fragment into monomer circles; however, the amount of protein required was about 300-fold higher than for the wildtype protein (Table II). These results demonstrate that the basic N-terminal tail of NHP6A performs a critical function in stable DNA binding, although a weak functional association with DNA is possible with only the minimal HMG box.
In contrast to the properties of the deletion of the entire 16-amino acid N terminus, a deletion of the first 12 amino acids resulted in a mutant that retained relatively high affinity binding, although it was not capable of producing discrete DNA complexes on polyacrylamide gels ( Fig. 2C and Table II). Therefore, we conclude that the RKKK motif between residues 13 and 16, which are retained in this mutant, is the critical determinant within the N terminus for mediating high affinity DNA association. Although NHP6A ⌬(2-12) did not form individual complexes on linear DNA, it bound to 98-bp microcircles with similar affinity as the wild-type (Fig. 4B). In addition, the ability of ⌬(2-12) to form 98-bp microcircles was indistinguishable from wild-type (Fig. 3B).
NHP6A ⌬(2-7) retains both blocks of basic amino acids in the N terminus and exhibited completely normal DNA binding, including the formation of stable complexes on linear DNA (Fig. 2B), and the formation and binding of DNA microcircles (Table II). Therefore, no role in DNA binding is detectable for residues 2-6. The difference in binding properties on linear DNA between ⌬(2-7) and ⌬(2-12) suggest that the KKR motif between residues 8 and 10 functions to stabilize binding of individual protomers of NHP6A to DNA, but it is clearly not critical for productive DNA interactions.
Circular dichroism was used to assess the folding of wildtype NHP6A as compared with the N-terminal deletion mutant (Fig. 5A). According to the self-consistent method of Sreerama and Woody (62), wild-type NHP6A was predicted to contain 49% ␣-helix, Ͻ1% ␤-sheet, 22% turn, and 19% other structure. These values are slightly lower than the amount of ␣-helical structure predicted by circular dichroism data on HMG1 box B (31,37,63), although the C-terminal extension of box B was not included in their spectra. The spectrum of the NHP6A ⌬2-16 mutant (Fig. 5A) showed that it was highly structured with approximately 67% ␣-helix, Ͻ1% ␤-sheet, 13% turn, and 11% other structure. These values are similar to the 75% estimate of ␣-helical content derived from NMR data on the minimal HMG1 box B (25, 26) and the HMG-D box (22). The increase in FIG. 1. HMG domain proteins. A, sequence alignment of five non-sequence-specific HMG box domain proteins and two site-specific HMG box proteins based upon homology model building (76). The position of the three ␣-helices that constitute the HMG box B domain is shown above the sequences (25,26). ␣-helical content in the NHP6A N-terminal deletion mutant as compared with wild-type is consistent with the N-terminal 16 amino acids of NHP6A being unstructured in solution. In addition, the T m of wild-type and NHP6A ⌬2-16 were both found to be 39°C (Fig. 5D), further indicating that the N terminus does not play a role in stabilizing the overall structure.

Importance of Prolines at Position 18 and 21 for DNA
Binding by NHP6A NHP6A contains two prolines (positions 18 and 21) that are located within the N-terminal end of the HMG domain. The proline at position 21 of NHP6A, which shows remarkable conservation among all non-sequence-specific HMG1/2 proteins (4,5), stabilizes a secondary hydrophobic core that associates the N-and C-terminal ends of the HMG domain (22). In LEF-1, proline 66, located at the C-terminal end of the HMG domain, appears to be important in directing its C-terminal basic extension that is essential for high affinity LEF-1-DNA binding (24). To determine whether the N-terminal prolines in the NHP6A HMG domain may be performing analogous roles, we created alanine substitutions at these positions and analyzed their DNA binding and folding properties.
NHP6A P18A displayed a 2.5-fold reduction in affinity for linear DNA whereas NHP6A P21A showed 4-fold reduced affinity (Fig. 2, E and F). The behavior of the double mutant P18A/P21A was identical to that of P21A (Fig. 2G). Binding to microcircular DNA was similarly reduced with a K D of 4 nM for P18A and 20 nM for both P21A and P18A/P21A (Fig. 4D). All three mutants were able to convert 70% of 98-bp linear DNA fragments into circles, although P18A required 2 times more protein than wild-type and both P21A and P18A/P21A required 4-fold more protein (Fig. 3B, Table II). Thus, the prolines at position 18 and 21 play a modest role in DNA binding.
CD analysis of the proline mutants indicated that P18A is folded identically to wild-type, whereas the P21A and the double mutant P18A/P21A displays a small (Ͻ10%) reduction in ␣-helical content compared with wild-type (Fig. 5B). Proline 21, therefore, contributes to the folding of the NHP6A HMG domain probably via hydrophobic interactions with residues of ␣-helix III (see Discussion). Proline 18 may facilitate positioning of the basic N-terminal segment upon DNA binding as also elaborated in the Discussion.

Importance of an Intercalating Side Chain for DNA
Binding by NHP6A DNA binding and bending by either LEF-1 or SRY is facilitated by a hydrophobic amino acid located near the N terminus of helix 1 that intercalates between base pairs via the minor groove (23,24). Alignment of amino acid sequences as in Fig.  1A suggests that methionine 29 of NHP6A would be in the correct position to function as an intercalating side chain. To test whether methionine 29 of NHP6A is important for DNA binding, this residue was mutated to alanine. Surprisingly, M29A did not show any difference in its affinity to linear DNA as compared with wild-type (Fig. 2H). Even changing methionine 29 into a negatively charged aspartic acid resulted in less than a two-fold reduction affinity to linear DNA. Both M29A and M29D bound to 98-bp microcircles with a K D of 2.5 nM and formed 98-bp circles with similar efficiency as wild-type NHP6A (Fig. 3C).
The relatively modest effect on DNA binding by substitutions at position 29 led us to test whether adjacent amino acids could function in this capacity. Tyr-28, Phe-30, and Phe-31 were changed to valine or aspartic acid. F30V exhibited very little difference in affinity for linear DNA and 98-bp microcircles compared with wild-type and did not differ in its DNA bending properties. Y28D and F31D were strongly defective in binding to linear and circular DNA, but we show below that they are unfolded in solution. F31V was able to bind poorly to both linear DNA and microcircles (Figs. 2J & 4E); however, the ability to form microcircles was greatly reduced with a maximum of only 4% of the input DNA ligated into 98-bp circles when very high amounts of protein were added (Fig. 3B).
CD analysis demonstrated that Y28D and F31V were largely unfolded in solution, whereas M29A showed only a slight change in CD profile (Fig. 5C). The CD data, combined with the known HMG box structures, suggest that the Phe-31 and Tyr-28 side chains are oriented such that they are contributing to the primary hydrophobic core that stabilizes the HMG fold (22,64). Since F31V forms complexes with DNA microcircles, interaction with the pre-bent DNA ligand is presumably stabilizing its folded structure. The HMG box structure in F31D, however, is probably completely disrupted since it is not capable of forming complexes with curved DNA.
The lack of any clear effect on DNA interactions by M29A, M29D, or F30V revealed by the previous assays led us to test these mutants for their ability to circularize 75-bp fragments. The formation of 75-bp microcircles are predicted to require the greatest amount of DNA bending per bound NHP6A protomer (see "Discussion"). In this assay, a clear difference between wild-type and M29A or M29D was observed (Fig. 3C). Neither M29A nor M29D was able to form 75-bp microcircles, even at very high concentrations of added protein. F30V was fully functional for 75-bp microcircle formation. Taken together, we conclude that Met-29 does play a role in NHP6A-induced DNA bending, but this is only revealed by assays that demand DNA bending near the maximum possible extent possible for wildtype NHP6A.

Ability of the NHP6A Mutants to Function in Biological Reactions
Growth Phenotypes-Using Saccharomyces cerevisiae, we are able to correlate the in vitro properties of DNA binding by NHP6A mutants to their biological functions. The double knockout mutant (⌬nhp6a/b-RJY6012) exhibits a slow growth phenotype forming colonies of heterogeneous size, which are temperature-and cold-sensitive at 38°C and 23°C, respectively (33,50). A yeast CEN shuttle vector containing the endogenous NHP6A promoter was used to express NHP6A and various mutants in the ⌬nhp6a/b cells. Western blotting confirmed that episomal expression of NHP6A and most NHP6A mutants were similar to endogenous levels of chromosomal expression (Table III). The exceptions were Y28D and F31V, which are unfolded in solution and were present at Ͻ10% of wild-type levels. Interestingly, ⌬(2-12) was present at 70% the level of wild-type whereas ⌬(2-16) was present at Ͻ20% of the wild-type level despite normal in vitro folding. The presence of episomal NHP6A and all of the above mutants except for ⌬(2-16), Y28D and F31V were able to reverse the temperature and cold sensitivity of ⌬nhp6a/b cells (data not shown).
The generation time of the ⌬nhp6a/b cells during log phase is approximately 250 min, nearly 2.5 times slower than the NHP6A/B parent (Table III). The expression of normal NHP6A protein in RJY6012 restored the generation time of these cells to a near wild-type 130 min. The presence of ⌬(2-16) had little stimulatory effect on the growth of the ⌬nhp6a/b mutant, but this may be primarily due to its low steady state expression levels. ⌬(2-7), ⌬(2-12), P18A, and F30V restored near normal growth rates, as expected from their in vitro properties. NHP6A M29A only partially rescued growth rates of ⌬nhp6a/b cells (190 min generation time), even though this mutant has normal in vitro DNA binding properties in most regards. ⌬nhp6a/b cells expressing P21A and P18A/P21A, which have a disruption of the secondary hydrophobic core, also grew significantly slower (175-180 min generation time). The addition of Y28D or F31V had no stimulatory effect on ⌬nhp6a/b cell growth, but as noted above, steady state levels of these proteins are very low.
HMG1 box BЈ is able to efficiently complement the growth phenotype of ⌬nhp6a/b mutants (RJY6398 , Table III), demonstrating the in vivo functional relationship between these homologous proteins. Significantly, HMG box B also requires a basic region to be active in vivo since the minimal HMG box B has no stimulatory effect on growth (RJY6271 , Table III).
Activated Transcription at the CUP1 Locus in Vivo-The NHP6A mutants were also tested for their ability to specifically enhance activated transcription of the CUP1 promoter, one of a subset of genes whose activated expression is facilitated by the NHP6A/B proteins (41). Following 2 h of exposure to 1 mM CuSO 4 in minimal media, ␤-galactosidase activity from a CUP1-lacZ reporter construct was induced 40-fold in ⌬nhp6a/b cells expressing NHP6A from pRJ1342, similar to the 47-fold induction measured in wild-type cells (Fig. 6A). The induced level of CUP1-LacZ expression in the ⌬nhp6a/b mutant cells was only 8-fold above basal level. Activated transcription of the CUP1 promoter by the different NHP6A mutants largely paralleled their effect on growth. ⌬(2-16) had no activity, whereas ⌬(2-12) displayed essentially wild-type CUP1 expression levels. CUP1 transcription in the presence of M29A was also notably reduced.
Stimulation of in Vitro Transcription-The G 5 E 4 T template containing five GAL4 binding sites upstream of the adenoviral E 4 T promoter was used as a model substrate to test the ability of the mutants to facilitate activated transcription in vitro. Transcription in yeast nuclear extracts activated by GAL4-VP4 was stimulated 2-5-fold by increasing amounts of exogenous NHP6A (Fig. 6B). A similar increase in transcript levels was observed by the addition of bovine HMG1 or HMG2 (Fig.  6C), but HU, an unrelated Escherichia coli DNA bending protein, had no effect on this reaction (data not shown). NHP6A (Fig. 6B) or HMG1/2 (data not shown) had no effect on basal transcription from this promoter, similar to that observed for the GAL1 promoter (41).
Most of the mutants stimulated transcription at G 5 E 4 T to a similar extent as wild-type ( Fig. 6C and data not shown). NHP6A ⌬(2-16) was completely defective even when large amounts of protein (Ͼ2 g) were added, but ⌬(2-12) behaved indistinguishably from wild-type. Transcription in the presence of M29A approached wild-type levels, but was significantly reduced in the presence of M29D. As expected, the unfolded mutants Y28D and F31V had little or no stimulatory activity.
Hin-catalyzed Site-specific DNA Inversion-When the recombinational enhancer is located within 100 bp of a recombination site, HU or an HMG1/2 protein is needed to assemble  b Ability to form discrete protein-DNA complexes of increasing complexity beginning with a monomer of NHP6A. c Determined by the amount of protein required to shift 50% of the input 98-bp microcircular DNA into slower mobility complex(es) upon polyacrylamide gel electrophoresis. d K circle is defined as the amount of protein needed to make 50% of the maximal amount of monomer circles from input DNA in ligase-mediated circularization reactions. The maximal percentage of monomer circle for each mutant is given in parentheses.
e Up to 25,600:1 molar ratio of protein to DNA was used. f Most of the binding probably represents nonspecific interactions. the catalytically competent invertasome (33,38,56). In this reaction, the auxiliary DNA bending protein is believed to function strictly as a DNA architectural factor to facilitate DNA looping of the enhancer segment. In Fig. 7A, the amount of NHP6A mutant added to the DNA inversion reaction was adjusted to compensate for defects in DNA binding. Under these conditions, all the mutants are able to stimulate inver-tasome assembly with the exception of the unfolded proteins Y28D and F31V, which bind and bend DNA poorly at all tested concentrations. NHP6A ⌬(2-16) displayed surprisingly high activity in this assay (Fig. 7B), particularly in comparison to its complete inactivity in stimulating transcription in vitro. The proficiency of ⌬(2-16) in promoting invertasome formation at moderately high concentrations of protein probably reflects its ability to promote microcircle formation at very high protein concentrations. DISCUSSION DNA Binding and Bending by Wild-type NHP6A-NHP6A is a member of a subclass of HMG1/2 proteins that contain a single HMG box and bind nonspecifically to linear DNA with relatively high affinity. NHP6A binds equivalently to DNA varying in sequence and length and forms distinct complexes upon polyacrylamide electrophoresis with a K D of 100 nM. We show in this report that each complex represents a monomer of NHP6A bound to DNA. Although the complexes formed on linear DNA are stable during long term electrophoresis in a gel matrix, they readily dissociate in the presence of excess DNA. NHP6A binds with Ͼ50-fold higher affinity to microcircular DNA. The complexes formed on microcircular DNA are extremely stable; NHP6A remains bound to the microcircles in the presence of 1000-fold excess linear DNA. Preferential binding to microcircular DNA has also been reported for HMG1 and HMG-D (30,65). Even though NHP6A binding is presumably targeted to the distorted structures present on microcircular DNA, it does not bind preferentially to four-way junctions (data not shown). High affinity binding to four-way junctions is a feature of many HMG proteins, although the related HMG-D and cHMG1a proteins also do not display this property (17,66).
The formation of stable complexes of defined numbers of NHP6A protomers on microcircular DNA of varying lengths can be used to estimate the degree of DNA bending introduced upon NHP6A binding. The measured persistence length of a Yeast strains were wild-type, ⌬nhp6a, ⌬nhp6b, or ⌬nhp6a/b containing a plasmid with no insert, NHP6A, NHP6A mutant, or HMG1 genes. All strains were grown in minimal media under plasmid selection.
b Generation times were measured during the exponential growth phase. c NHP6A/B expression was determined by Western blotting. ϩϩϩ represents wild-type levels, ϩϩ represents about 70% of wild-type, ϩ Ͻ20%, ϩ/Ϫ 10%, and ND is not determined. DNA (67) gives an average intrinsic flexibility of about 2.4°/bp (360°/150 bp). Thus, DNA lengths of 66, 75, and 98 bp can generate 158°, 180°, and 235°of curvature. Based on the number of NHP6A protomers bound to preformed microcircles (Fig.  3) or upon formation of microcircles (data not shown) of lengths 98 bp (two complexes), a 75 bp (two complexes), or 66 bp (three complexes), the minimal amount of protein-induced bending can be calculated. To create a 98-bp microcircle, each of the two NHP6A-induced bends required to complete the DNA circle would be approximately 60°, which corresponds to a 120°DNA bending angle by each NHP6A protomer relative to linear DNA. A 75-bp microcircle requires a 45°bend in the DNA from each of the two bound NHP6A protomers resulting in a 135°b ending angle. Since the 66-bp microcircle has three bound NHP6A protomers, an equilateral triangle can be used to estimate the induced bending of the DNA to be 60°, which corresponds to a bending angle of 120°. Therefore, from these experiments we estimate NHP6A induces bend angles between 120 and 135°. These values are in the range of the angles observed with LEF-1 by NMR (120°; Ref. 24) and cHMG1a by fluorescence resonance energy transfer (150°; Ref. 68). However, they are greater than the 80°observed in the SRY-DNA complex by NMR (23), or the 60°estimated for DNA binding by HMG-D by a ligase-mediated circularization assay (30). High Affinity DNA Binding by NHP6A Requires Its Unique N Terminus-The 16-amino acid segment located N-terminal of the minimal HMG domain of NHP6A is essential for its unusu-ally stable DNA interaction. Removal of this region abolishes the ability of the protein to form discrete complexes on both linear and microcircular DNA and eliminates most of its biological activities. The remaining minimal HMG box is capable of poorly binding to DNA in vitro, as evidenced by its ability to form microcircles, promote Hin invertasome formation, and induce DNA supercoiling in the presence of topoisomerase I (data not shown) at high protein concentrations. These activities provide strong evidence that the minimal HMG box remains capable of an authentic, albeit weak, HMG-DNA interaction. Circular dichroism and thermal stability data indicate that the N-terminal segment is unstructured in solution and has no effect on the integrity of the folded HMG domain. Upon NHP6A-DNA interaction, the basic N-terminal arm presumably associates with the DNA to anchor the complex. We imagine that the N-terminal arm may cross over the phosphate backbone and protrude into the major groove (Fig. 8A), as does the C-terminal arm of LEF-1 (24), or possibly continue along the minor groove as schematically drawn in Fig. 8B.
There are two blocks of basic amino acids within the NHP6A N-terminal segment, block 1 between residues 8 and 10 (KKR) and block 2 between residues 13 and 16 (RKKK). Both play a role in stabilizing DNA binding, but block 2 is of primary importance. A deletion, NHP6A ⌬(2-12), that removes patch 1 but retains patch 2 binds linear DNA and promotes microcircle formation almost as well as wild-type, although it does not form stable monomeric complexes on linear DNA by gel mobility shift assays. The deficiency in stable complex formation on linear, but not microcircular DNA, does not seem to affect any of its biological activities that were measured. The first 7 amino acids of the N-terminal arm play no detectable role in DNA binding or biological properties of NHP6A.
We find it significant that NHP6A ⌬(2-16) is not capable of potentiating transcription in vitro, even in reactions containing a large excess of protein. Moreover, this mutant was unable to form promoter complexes together with TBP and TFIIA, unlike wild-type NHP6A or the NHP6A ⌬(2-12) mutant (Ref. 41; data not shown). By contrast, NHP6A ⌬(2-16) is still able to promote Hin invertasome assembly at moderately high concentrations. These differences may indicate that the NHP6A protein is performing a specific function in transcription rather than functioning merely in an architectural role, as is probably the case for the recombination reaction.
Steady state cellular levels of NHP6A ⌬(2-16) are Ͻ20% of wild-type. This difference suggests that the mutant protein may be rapidly turned over in the yeast cells even though it is well structured. The apparent instability of NHP6A ⌬(2-16) could be due to its weak DNA binding properties that result in poor nuclear retention and thereby rapid degradation. A lack of retention in the nucleus may also result from the removal of a nuclear localization signal (NLS) present in the N-terminal arm. The amino acid sequence of NHP6A between amino acids 8 and 16 (KKRTTRKKK) matches an NLS motif (69). Other NLS sequences have been identified in the HMG domain of mUBF, SRY, SOX9, and LEF-1 (70 -73). Cellular levels of NHP6A ⌬ (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) are also reduced to about 70% of wild-type, but NHP6A ⌬(2-7) levels are normal. This provides further support for the importance of residues 8 -16 for protein stability, and implies that the residues between 13 and 16 are more important than the residues between 7 and 12. Further experiments will be needed to confirm that the basic N terminus contains a nuclear localization signal in addition to being required for high affinity DNA binding.
Other HMG proteins also require a basic region adjacent to the HMG domain. Teo et al. (37) could not form microcircles with HMG1 box A or B but could form them with box BЈ, which contains a patch of basic amino acids at its C terminus. This observation is likely to be related to our finding that HMG1 box BЈ is able to largely complement the growth defect of ⌬nhp6A/B mutants, whereas HMG box B is ineffective (although an NLS in the box BЈ basic region may also contribute to this difference). HMG-D also contains a positively charged patch of amino acids at the C-terminal end of its HMG domain, which are required for efficient DNA binding on linear DNA in vitro (30). However, in contrast to the NHP6A ⌬(2-16) mutant, the minimal HMG-D domain can form stable complexes on microcircles. A short basic patch adjacent to the C terminus of LEF-1 was found to be required for high affinity and bending by this sequence-specific HMG protein (36).
Role of Prolines at the N Terminus of the NHP6A HMG Domain-NHP6A contains two prolines at positions 18 and 21, which are predicted to be located near the top of the L-shaped HMG domain adjacent to the N-terminal arm (see Fig. 8) based upon the structures of box A and B from HMG1 (21,22,25,26). When proline 18 in NHP6A was changed to an alanine, a 2-3-fold decrease in DNA binding to linear and microcircular DNA was measured. CD analysis did not reveal a significant difference between wild-type NHP6A and NHP6A P18A. We postulate that proline 18 may direct the peptide backbone to facilitate positioning of the N-terminal arm. In the absence of this proline, the N terminus is still able to interact with DNA, although not quite as effectively. In all in vitro reactions tested, NHP6A P18A is very active, provided additional protein is added to compensate for the modest effect on binding affinity. This role for proline 18 of NHP6A is analogous to the function of proline 66 of LEF-1, which directs its C-terminal arm into the major groove where extensive DNA contacts are made (24). A proline to alanine substitution at the same relative position in HMG1 box A as our NHP6A P18A mutant has also been analyzed, but it was not shown to have any effect on DNA binding (64).
The proline at position 21 of NHP6A is conserved among all non-sequence-specific HMG1/2 proteins but corresponds to a valine or isoleucine in sequence-specific HMG proteins. The importance of this residue for NHP6A is shown by a 4-to over 10-fold reduction in binding affinity of P21A to linear and microcircular DNA, respectively. NHP6A P21A is partially defective in its ability to complement ⌬nhp6A/B mutants for growth and CUP1 expression, but increased levels of protein largely compensate for reduced activity in reactions in vitro. CD analysis indicates that the P21A mutation causes a 10% loss of ␣-helicity. Based upon the structures of HMG-D, HMG1 box A and B, and evidence from cHMG1a (21,22,25,26,34), the conserved proline at position 21 is probably involved in stabilizing a secondary hydrophobic pocket between helix III and the extended peptide chain between the N terminus and the start of helix 1 (Fig. 8). Thus, the reduction in DNA binding affinity of this mutant is attributed to a structural disruption of the N terminus of the HMG domain and consequently the N-terminal arm.
The Hydrophobic Core of NHP6A-Mutations at Tyr-28 and Phe-31 lead to an unfolded protein, as determined by CD analysis in vitro and an unstable protein in vivo. The aromatic side chains of these conserved amino acids are predicted to be directed into the primary hydrophobic core that stabilizes the three-helix fold (Fig. 8), as observed for other HMG box proteins (5). Interestingly, the presence of DNA appears to stabilize the structure of F31V since it is able to form complexes with microcircular and linear DNA with modest affinity. However, F31D is completely defective in all activities measured.
Methionine 29 Is Required for Maximal NHP6A-induced DNA Bending-The structures of DNA complexes formed with the sequence-specific HMG proteins LEF-1 and SRY show that a hydrophobic amino acid located in the N terminus of helix 1 protrudes into the DNA from the minor groove, leading to a pronounced base pair unstacking. The critical role of the intercalating isoleucine in SRY-DNA interaction is demonstrated by the sex reversing mutation I68V, which causes a severe defect in DNA binding and bending (23,74,75). Methionine 29 is located at the analogous position in NHP6A by sequence alignment (Figs. 1A and 8). However, NHP6A M29A and even M29D binding to linear DNA is nearly indistinguishable from wildtype, and the Met-29 mutants are proficient in Hin inversion, and inducing DNA supercoiling (data not shown) in vitro. On the other hand, NHP6A M29A only restored about 50% of the growth rate defect that results from deleting NHP6A/B. Moreover, activated transcription of CUP1 in the presence of M29A in vivo is significantly reduced, as well as activated transcription of G 5 E 4 T in vitro in the presence of M29D (and M29A to a small extent). Thus, Met-29 appears to be important for some activities, including co-activation of transcription, but not other HMG1-promoted functions. The different behavior of Met-29 mutants in these reactions may be related to their ability to form microcircles of varying lengths. Although M29A and M29D were unaltered in their ability to form 98-bp DNA microcircles, they were completely defective in 75-bp microcircle formation. Thus, Met-29 is not required for productive DNA binding but is essential to induce maximum DNA bending.