Maize chromosomal HMGc. Two closely related structure-specific DNA-binding proteins specify a second type of plant high mobility group box protein.

The chromosomal high mobility group (HMG) proteins are small and abundant non-histone proteins common to eukaryotes. We have purified the maize HMGc protein from immature kernels and characterized it by mass spectrometry and amino acid sequence analysis. HMGc could be resolved into two similar proteins by reversed phase chromatography. Cloning and characterization of the corresponding cDNAs revealed that they encode two closely related maize HMGc proteins, now termed HMGc1 and HMGc2. Their theoretical masses of 15,316 and 15,007 Da are >300 Da lower than the masses determined for the proteins purified from maize, indicating post-translational modifications of the proteins. Despite sequence similarity to maize HMGa (and previously described homologous proteins of other species) amino acid sequence alignments reveal that HMGc is in several conserved regions distinct from these proteins. Consequently, we have identified a novel type of plant protein containing an HMG box DNA binding domain and belonging to the HMG1 protein family. HMGc1 and HMGc2 were expressed in Escherichia coli, purified to homogeneity, and analyzed for their DNA binding properties. They proved to bind to DNA structure-specifically since they formed complexes with DNA minicircles at concentrations ~100-fold lower than the concentrations required to form complexes with linear fragments of identical sequence. Furthermore, HMGc1 and HMGc2 can constrain negative superhelical turns in plasmid DNA.

High mobility group (HMG) 1 proteins represent a heterogeneous family of small chromosomal proteins common to eukaryotic organisms. They are among the most abundant and ubiquitous non-histone proteins of the nucleus. The HMG proteins have an unusual amino acid composition, containing a high proportion of acidic and basic amino acid residues, and are soluble in 2% trichloroacetic acid or 5% perchloric acid (1). In vertebrates, HMG proteins have been divided into three structurally unrelated subgroups (2): HMG1/2 (M r around 25,000), which contain two HMG box DNA binding domains and bind structure-specifically to DNA recognizing distortions in doublestranded DNA; HMG14/17 (M r around 10,000) which contain a conserved nucleosome binding domain and have a higher affinity for nucleosome core particles than for naked DNA; and HMGI/Y (M r around 10,000), which contain three copies of the A/T hook DNA binding motif and bind to A/T stretches of double-stranded DNA.
Chromosomal HMG1/2-like proteins containing a common DNA binding motif known as the HMG box domain have been identified from various eukaryotes (3)(4)(5)(6). In addition to the abundant chromosomal non-histone proteins of the HMG1/2 class, this domain is found in less abundant sequence-specific DNA-binding proteins such as the polymerase I transcription factor UBF (3), the polymerase II transcription factor LEF-1 (7), and the testis-determining factor SRY (8). Members of both groups of HMG box proteins recognize distorted DNA structures (9 -14) and induce bends in linear duplex DNA (15)(16)(17). A number of HMG1/2-like proteins were reported to constrain negative superhelical turns in plasmid DNA (18 -23). In addition, they can act as architectural elements promoting the formation of complex nucleoprotein structures (15,16,24,25).
In plants, four major HMG proteins, as in vertebrate tissues (HMG1/2, HMG14/17), have been detected, displaying a slightly higher mobility in SDS-PAGE than their animal counterparts (26). Analysis of these proteins by peptide mapping, amino acid analysis, and amino-terminal amino acid sequencing revealed significant dissimilarities to the animal proteins (26,27). Nevertheless, several members of the HMG1/2 and the HMGI/Y families have been identified by cDNA cloning from various plant species (28).
The well conserved HMG1/2-like plant proteins (M r around 17,000) contain only one HMG box DNA binding domain (as invertebrate HMG1/2-like proteins), in contrast to vertebrate HMG1/2, which have two. In addition, the plant proteins have an approximately 35-amino acid residue basic region at the amino terminus, and within their acidic carboxyl-terminal region the glutamic and aspartic acid residues are interspersed with some other residues, which is not the case in HMG1/2 (29 -31). The homologous HMG box proteins, wheat HMGb and maize HMGa, bind in vitro with some preference to diverse A/T-rich promoter regions of various plant genes (32,33). Similar to vertebrate HMG1 (11,17), the maize HMGa protein can recognize distorted DNA structures such as four-way junctions and DNA minicircles (34). Furthermore, in a ligase-mediated circularization assay, HMGa displayed a strong DNA bending activity (34), as observed with HMG1 (16,17). * This work was supported by a grant from the Deutsche Forschungsgemeinschaft (to K. D. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) Y08297 and Y08298.
Up to now no plant HMG14/17-like protein has been identified. The most likely candidates are proteins such as wheat HMGc and HMGd (26), which bind to nucleosomes, although in the electrophoretic retardation assay used the interaction appeared to be rather unspecific (35). However, HMGc and HMGd were released preferentially from nuclei by ethidium bromide (36), a characteristic shared by animal HMG14/17 (37,38). Amino-terminal amino acid sequences of wheat HMGc and HMGd show no similarity to their putative animal counterparts (27), but there is also no amino-terminal amino acid sequence homology between animal and plant HMG1/2 and HMGI/Y proteins (28).
To solve the so far unknown primary structure of plant HMGc proteins and to analyze their DNA binding properties, we first purified HMGc from maize which was detected previously in nuclear extracts from immature kernels (39). Taking advantage of the sequence information we obtained from sequencing amino-terminal and internal peptides, we could isolate two closely related cDNAs encoding maize HMGc. Sequence analysis revealed that HMGc has no similarity to HMG14/17 and belongs to the HMG1/2 family of chromosomal proteins containing an HMG box DNA binding domain, but it is distinct from the previously identified plant HMG1/2-like proteins, thus representing a novel type of plant HMG box protein.

EXPERIMENTAL PROCEDURES
Purification of HMGc from Maize Tissue-Frozen maize tissue (immature kernels or black Mexican sweet suspension culture cells, typically 100 g/extraction) was homogenized in liquid nitrogen with pestle and mortar. All of the following procedures were carried out at 4°C. The frozen homogenate was taken up in 10 volumes of buffer X (20 mM Tris-HCl pH 7.0, 10 mM 2-mercaptoethanol, 1 mM EDTA, 0.5 mM PMSF) containing a final concentration of 1.5-2.5% (w/v) trichloroacetic acid or 5% (w/v) perchloric acid. After stirring for 15 min, the suspension was centrifuged at 15,000 ϫ g for 15 min. Trichloroacetic acid was added to the supernatant (filtered through Miracloth) to a final concentration of 25% (w/v), and the mixture was centrifuged at 25,000 ϫ g for 20 min. The pellet was washed twice with cold acetone, vacuum dried, and resuspended in 5 ml of buffer D (10 mM sodium phosphate, pH 7.0, 1 mM EDTA, 1 mM dithiothreitol, 0.5 mM PMSF). The protein solution was centrifuged at 15,000 ϫ g for 10 min and the supernatant applied to a ResourceQ FPLC column (Pharmacia) equilibrated with buffer D. The column was washed with this buffer and then eluted with a linear gradient of 0 -1 M NaCl in buffer D. HMGc-containing fractions were identified by SDS-PAGE, pooled, and the buffer was changed to buffer D using PD-10 columns (Pharmacia). Proteins were characterized by SDS-PAGE in 18% polyacrylamide gels and Coomassie staining, aminoterminal sequencing, and MALDI-TOF/MS.
Proteinase Digests of HMGc-HMGc was digested with 2% (w/w) trypsin or subtilisin overnight at 37°C. The peptides were separated by reversed phase HPLC and characterized by MALDI-TOF/MS and peptide sequencing.
Separation of HMGc and Peptides by Reversed Phase HPLC-Maize HMGc eluted from the ResourceQ column and peptide digests were separated on a Vydac C-18 column (1.0 ϫ 250 mm) using the HP 1100 HPLC system (Hewlett Packard). Peptides were eluted from the column at a flow rate of 70 l/min applying the following water (solvent A)/ acetonitrile (solvent B) gradient: at 0 min, 1% B; at 70 min, 35% B; at 90 min, 70% B. Fractions were collected manually. HMGc was separated in the same way except that the gradient was as follows: at 0 min, 2% B; at 60 min, 60% B.
MALDI-TOF/MS Analysis-All mass spectrometric analyses were performed on the Hewlett Packard G2025A MALDI-TOF mass spectrometer equipped with a nitrogen laser (337 nm). As matrix, sinapic acid was used for protein analysis and cyano-4-hydroxycinnamic acid for peptide analysis.
Sequencing of Peptides and Proteins-Automated Edman degradation of peptides and proteins was performed using the Hewlett Packard G1005A protein sequencing system.
Library Screening and Nucleotide Sequence Analysis-PCR was performed with 3 g of reverse transcribed rice seedling leaf RNA (provided by R. Freyer), 1 M primers (5Ј-ATGAAGGGCAAGGCCGACGCCTC and 5Ј-GCTCAGACATAGCTCGCCACTTC, according to the sequence D22537 of the expressed sequence tag (EST) data base), 0.2 mM deoxyri-bonucloside triphosphates, and 1 unit of Taq DNA polymerase (Pharmacia) according to the manufacturer. The obtained PCR fragment partially encoding a putative rice HMGc was cloned into pGEM-T (Promega) and sequenced. After digestion of this plasmid with SacII and SpeI the fragment was purified by gel electrophoresis and labeled with digoxigenin (Boehringer Mannheim) by the random priming method. The labeled fragment was denatured by heating to 100°C and used as a hybridization probe to screen ϳ300,000 plaque-forming units of a maize Uni-ZAP cDNA library (Stratagene). Hybridizations were performed overnight at 58°C in 5 ϫ SSC, 0.1% (v/v) N-lauroylsarcosine, 0.02% (w/v) SDS, 2% (w/v) blocking powder (Boehringer), before the nitrocellulose membranes (Schleicher & Schü ll) were washed twice in 2 ϫ SSC, 0.1% (w/v) SDS at room temperature and twice in 0.1 ϫ SSC, 0.1% (w/v) SDS at 58°C. Hybridizing plaques were detected using an alkaline phosphatase-conjugated anti-digoxigenin antiserum and the standard alkaline phosphatase color reaction (Boehringer). Positive phage clones were purified by rescreening at low plaque density. The recombinant pBluescript plasmids were isolated from purified phage clones by in vivo excision with helper phage ExAssist according to the manufacturer (Stratagene). Nucleotide sequences of the cDNA inserts were determined with an automated sequencer (ALF, Pharmacia) and analyzed using the GCG V8.0 sequence analysis system. Screening of the library in search for cDNAs encoding the second type of maize HMGc was performed using the coding region of the first maize HMGc cDNA clone as hybridization probe as described above, but hybridization and stringent wash were performed at 65°C.
Expression and Purification of Recombinant HMGc1 and HMGc2-The coding regions of HMGc1 and HMGc2 were amplified by PCR with primers (5Ј-AAGGATCCATGAAGGGCAAGGCTGACACCT, 5Ј-AATT-AAGCTTACTCATCGTCATCGTTTTCCT and 5Ј-AAGGATCCATGAA-GGGCAAGGCCAACGCCT, 5Ј-AATTAAGCTTACTCGTCGTCGTCGT-TATCCTCGT) as described above, except HotTub polymerase (Amersham) was used. The obtained PCR fragments were digested with BamHI and HindIII, purified using PCR Preps (Promega), and cloned into the expression vector pQE9cm (created by replacing the ampicillin resistance of pQE9 (Qiagen) by the chloramphenicol resistance of pBC-SK (Stratagene)) giving pQE9cm-HMGc1 and pQE9cm-HMGc2. These plasmids provide an amino-terminal 6 ϫ His-tag suitable for Ni-NTA affinity purification (Qiagen) of the fusion proteins. The plasmids were checked by DNA sequencing. Escherichia coli M15 cells were transformed to chloramphenicol resistance with the plasmids pQE9cm-HMGc1 and pQE9cm-HMGc2. Cultures were grown at 37°C in 2 ϫ YT medium (containing 50 g/ml chloramphenicol and 30 g/ml kanamycin), induced at A 600 ϭ 0.8 with 2 mM isopropyl-1-thio-␤-D-galactopyranoside and grown for 2 h. All of the following procedures were carried out at 4°C. Cells were collected by centrifugation, resuspended in buffer A (50 mM Tris-HCl pH 7.5, 1 M NaCl, 10 mM 2-mercaptoethanol, 0.5 mM PMSF, 100 g/ml benzamidine), and lysed with a French press. After centrifugation at 30,000 ϫ g for 15 min the supernatant was applied to an Ni-NTA column (Qiagen). The column was washed with buffer B (50 mM Tris-HCl, pH 7.5, 1 M NaCl, 10% (v/v) glycerol, 10 mM 2-mercaptoethanol, 0.5 mM PMSF, 100 g/ml benzamidine) and eluted with buffer C (50 mM Tris-HCl, pH 7.5, 350 mM imidazole, 10 mM 2-mercaptoethanol, 0.5 mM PMSF, 100 g/ml benzamidine). The eluted proteins were then applied to an S-Sepharose Fast Flow FPLC column (Pharmacia) equilibrated with buffer D (see above), and the column was washed with this buffer. Bound proteins were eluted with a linear gradient of 0 -1 M NaCl in buffer D. Fractions containing HMGc were pooled, adjusted to 100 mM NaCl, and applied to a ResourceQ FPLC column as described above. Purified HMGc1 and HMGc2 were desalted and characterized as described above by amino-terminal sequencing, MALDI-TOF/MS, and SDS-PAGE. Protein concentrations were determined by amino-terminal sequencing.
Electrophoretic Mobility Shift Assay (EMSA)-EMSAs were performed as described previously (34), using a 78-bp KpnI/XbaI fragment of pBluescript (Stratagene) terminally 32 P-labeled with the Klenow fragment of DNA polymerase I. The 78-bp minicircles were produced by intramolecular ligation of the 32 P-labeled 78-bp fragment (after creating blunt ends with the Klenow fragment of DNA polymerase I) using T4 DNA ligase in the presence of recombinant maize HMGa/M1-K123 protein as described before (34).

RESULTS AND DISCUSSION
Purification and Characterization of Maize HMGc-HMGc proteins have been detected previously in wheat embryo chromatin (26) and in nuclear extracts from immature maize kernels (39), but their primary structure has not yet been solved.
To purify and characterize maize HMGc, we first fractionated extracts from immature kernels (14 days after pollination) with 1.5-2.5% trichloroacetic acid and 5% perchloric acid, taking advantage of the acid solubility of HMG proteins (1). HMGc, along with HMGa, HMGd, and several other proteins, is strongly enriched in the acid-soluble fractions (Fig. 1A, lanes  1-4). One prominent protein band of the 1.5% trichloroacetic acid extract migrating immediately below HMGa (lane 1) was hardly soluble at higher acid concentrations. This protein, previously designated HMGb, proved by peptide sequencing after V8-digestion to be histone H2B1. The 2% trichloroacetic acid fraction was selected for the further purification of HMGc, since it combined a good yield with a reasonable enrichment of the HMG proteins. The proteins of the 2% trichloroacetic acidsoluble fraction of a kernel extract were separated by ion exchange chromatography resulting in essentially pure HMGa and HMGc (Fig. 1A, lanes 5 and 6). Amino-terminal sequencing confirmed the identity of both proteins, demonstrating that the amino-terminal sequence of HMGa is in complete agreement with the sequence deduced from the maize HMGa cDNA (29). The amino-terminal sequence of the maize HMGc protein (MKGKAD/NT/ASKKDEG/ARLRA) is ϳ75% identical to the amino-terminal sequence determined for HMGc purified from wheat embryos (27). To gain more sequence information of maize HMGc to facilitate cloning of the corresponding cDNA, the protein was tryptically digested, the resulting peptides separated by reversed phase HPLC, and some of the peptides sequenced (Fig. 1B). Similarly, a subtilisin digest was performed, resulting in an additional HMGc peptide sequence (KAAGEKWRA/SM). The ambiguities within the amino-terminal sequence of HMGc at positions 5, 6, and 13 suggested that maize HMGc might be composed of two very similar proteins that were copurified by the ion exchange chromatography. Therefore, HMGc was analyzed by reversed phase HPLC resolving it into two prominent peaks (Fig. 2). MALDI-TOF/MS and amino-terminal sequencing of the proteins corresponding to both peaks demonstrated that HMGc could be separated into two distinct proteins of similar mass. In agreement with the amino acid sequence determined from the "pre-HPLC" HMGc protein preparation, the amino-terminal sequences of both proteins differ only at positions 5, 6, and 13 (Fig. 2). Accordingly, we termed the two proteins HMGc1 and HMGc2.
Isolation and Characterization of cDNAs Encoding HMGc1 and HMGc2-Searching of the EMBL and GenBank data bases using the determined HMGc peptide sequences did not show significant similarities to known sequences, although the sequences of the subtilisin peptide (KAAGEKWRA/SM) and one of the tryptic peptides (VFMSEFR) exhibit similarities to HMG box DNA binding domains. A search of the EST data base, however, revealed a partial rice cDNA sequence of unknown function (accession number D22537) which was compatible with the HMGc peptide sequences. Four out of the five maize HMGc peptide sequences displayed striking similarity to the rice EST sequence. Therefore, we used reverse transcribed rice RNA and primers matching the rice HMGc-like EST sequence to amplify a 230-bp fragment by PCR. DNA sequencing of the fragment demonstrated that it represented as expected the 5Ј part of the coding region of the putative rice HMGc sequence. This fragment was used as a hybridization probe to screen a maize cDNA library for clones encoding maize HMGc. The DNA sequences of the inserts of two cDNA clones we isolated in the first screen were compatible with all five HMGc peptide sequences. Since the amino acid sequences deduced from these cDNAs were in full agreement with the amino-terminal HMGc1 peptide sequence we had determined, the two isolated cDNAs encoded maize HMGc1 (Fig. 3). To isolate cDNAs encoding HMGc2, we screened the maize cDNA library again, this time using the coding region of the maize HMGc1 cDNA as a hybridization probe. Two of the isolated cDNAs were compatible with the HMGc2 amino-terminal peptide sequence and the internal peptides, thus encoding maize HMGc2 (Fig. 3). In addition to the coding region, the longest isolated HMGc1 cDNA of 797 bp contained 5Ј-and 3Ј-flanking regions, whereas in the longest HMGc2 cDNA of 816 bp the coding region started with the guanine residue of the start codon and also contained the 3Ј-flanking region. The HMGc1 cDNA encodes a 139-amino acid protein with a theoretical mass of 15,316 Da, and the HMGc2 cDNA a 138-amino acid protein with a theoretical mass of 15,007 Da. Since the experimentally determined masses of HMGc1 and HMGc2 purified from maize kernels (and a black Mexican sweet cell suspension culture) were Ͼ300 Da higher than the theoretical values (see Fig. 2), both HMGc proteins might be post-translationally modified, which is a common feature of HMG proteins (40). Both proteins are rich in basic (ϳ20%) and acidic (ϳ20%) amino acid residues, typical for HMG proteins (1,2). The amino acid sequences of HMGc1 and HMGc2 share ϳ89% identical residues.
HMGc1 and HMGc2 Specify a Second Type of Plant Chromosomal HMG Box Protein-The conservation of a number of critical amino acid residues (primarily basic and hydrophobic) between amino acids Lys-32 (Lys-33 in HMGc2) and Phe-101 (Ile-102 in HMGc2) of HMGc1/2 and HMG box proteins strongly suggests that HMGc1 and HMGc2 are members of the protein family containing a HMG box DNA binding domain  (44). The blocks of amino acid residues that are conserved between the HMGa type of plant HMG box proteins (KLAV and KDPN, as mentioned in the text) and which are different in HMGc1/2 are underlined. The HMG box domain that has a high overall positive charge and a distinctive positioning of acidic, hydrophobic, and proline residues is the only region common to all proteins of the alignment and is indicated by darker lettering. Basic, acidic, and hydrophobic residues conserved within the HMG box domains of these proteins (in at least 8 of the 11 sequences shown) are indicated by (ϩ), (Ϫ), and (ૺ), respectively. Regions underlined below the sequences correspond to the three potential ␣-helical regions (I, II, III) of the HMG box, as predicted by comparison with the structures of the HMG boxes of HMG1 and HMGD determined by NMR spectroscopy (45)(46)(47)(48). (4 -6). In addition, HMGc1 and HMGc2 have a highly acidic carboxyl-terminal region, typical of (animal and plant) HMG1/ 2-like proteins. Accordingly, a search of the EMBL data base with the complete cDNA-derived protein sequences of HMGc1/2 revealed significant similarities to proteins containing HMG box DNA binding domains, such as the chromosomal HMG1/2-like proteins of various organisms, whereas no similarity to HMG14/17 and HMGI/Y proteins was detected. The most similar proteins are the previously reported plant HMG box proteins such as maize HMGa. For example, HMGc1 and HMGc2 are, in 50 and 54% of their amino acid residues, respectively, identical to maize HMGa (vertebrate HMG1 and HMG2 are ϳ80% identical). An alignment of maize HMGc1/2 and HMG1/2-like proteins from mono-and dicotyledonous plants (29 -31), Drosophila (41,42), yeast (43), and Tetrahymena (44) also containing only one HMG box domain demonstrates the significant similarities between these proteins (Fig.  4). Nevertheless, the alignment also shows that HMGc1 and HMGc2 differ from the previously identified plant HMG box proteins since they are distinct in various residues highly conserved between the "HMGa type" of the plant HMG1/2-like proteins. Consequently, some blocks of amino acid residues entirely conserved between the previously identified plant HMG box proteins are different in HMGc1/2, for instance the sequences KLAV and KDPN within the amino-terminal region of HMGa-like proteins (see Fig. 4). Moreover, the size of HMGc1/2 is smaller (ϳ15,000 Da) compared with the HMGa type proteins (ϳ17,000 Da). The differences in amino acid sequence and in protein size indicate that HMGc1 and HMGc2 specify a so far unidentified, second type of plant chromosomal HMG box protein.
The three-dimensional structure of various HMG box DNA binding domains has been solved recently by NMR spectroscopy, revealing an L-shaped molecule consisting of three ␣-helices with an angle of ϳ80°between the arms (45)(46)(47)(48). Since this protein fold appears to be conserved and because of the high degree of sequence similarity between the HMG box domains of HMGc1/2 and other HMG box proteins, it seems reasonable to assume a related structure for the HMG box domains of HMGc1/2. Biochemical analyses (49 -51) and the solution of two structures of HMG box-DNA complexes (of the sequencespecific proteins SRY and LEF-1) by NMR spectroscopy (52,53) emphasized the importance of the concave surface and the amino-terminal extended strand of the HMG box for interactions with DNA which occur primarily through the minor groove. Differences in the amino acid sequence of HMGc1/2 close to this extended strand of the HMG box (AASG in HMGc1/2, KDPN in the HMGa type of plant HMG box proteins, see Fig. 4) could possibly contribute to somewhat different DNA binding properties of HMGc1/2 compared with the previously identified plant HMG1/2-like proteins. The aspartic acid residue within this KDPN region (e.g. Asp-37 in maize HMGa) corresponds to residue Asp-4 in rat HMG1, which is implicated in DNA binding (48). In addition, the basic amino-terminal and acidic carboxyl-terminal regions of HMGc1/2 which are significantly different from the other plant HMG box proteins probably modulate their interactions with DNA, since analogous regions of other HMG box proteins have been found to influence DNA binding (19,20,23,54). HMGc from wheat (36) and maize (KDG, data not shown) can be released preferentially from nuclei by ethidium bromide, a characteristic shared by animal HMG14/17 (37,38). Since the other plant HMG box proteins (e.g. wheat HMGb and the homologous maize HMGa) are not released under these conditions, the two types of plant HMG box proteins, the HMGa type and the HMGc type, could interact (despite their structural similarity) differentially with plant chromatin.
Interactions of Recombinant HMGc1 and HMGc2 with DNA-HMG1/2-like proteins from various organisms have been shown to bind DNA non-sequence specifically, or with only low sequence specificity, rather recognizing (to various degrees) structural features of DNA such as four-way junctions, cisplatin-modified DNA, and DNA minicircles (11,13,23,34,54,55). Therefore, we compared the binding of HMGc1 and HMGc2 to linear DNA and to DNA minicircles of identical sequence. First, HMGc1 and HMGc2 were expressed in E. coli as 6 ϫ His-tagged fusion proteins and purified to homogeneity using three-step column chromatography. The proteins were characterized by SDS-PAGE (Fig. 5A), MALDI-TOF/MS, and amino-terminal amino acid sequencing (data not shown). Increasing concentrations of each recombinant protein were incubated in the presence of both a linear 78-bp plasmid frag-  1 and 11) are indicated by l and c, respectively. ment and the corresponding 78-bp minicircle, and the formation of protein-DNA complexes was analyzed by EMSAs. HMGc1 and HMGc2 displayed similar DNA binding properties in this assay, binding with strong preference for the DNA minicircle (Fig. 5B). Both proteins start in this assay at ϳ100fold lower concentrations (ϳ10 nM, lanes 4 and 14) to form complexes with the 78-bp minicircles, when compared with the formation of complexes with the linear 78-bp fragment (ϳ1 M, lanes 9 and 19), indicating structure-specific DNA recognition by HMGc1 and HMGc2. At increasing protein input, HMGc1 and HMGc2 form three distinct complexes with the DNA minicircles, whereas the signals corresponding to the complexes with the linear fragments are more diffuse, probably reflecting the varying stoichiometry of the bound HMGc1/2. Rat HMG1 and maize HMGa also display structure-specific binding to DNA minicircles in EMSAs when compared with linear DNA (17,34). The clear preference of vertebrate HMG1 for the structure of four-way junction DNA over linear duplex DNA (11) seems to be significantly less prominent with insect and plant HMG1/2-like proteins (34,54,55), possibly (at least partially) due to a higher affinity of the insect and plant proteins for linear (A/T-rich) double-stranded DNA. Thus, the HMG box domain appears to have been adapted in many different proteins to a wide variety of DNA binding modes: non-sequencespecific, sequence-specific, and structure-specific. The recognition of bent and deformable DNA sequences might be an important determinant of the target site selection by these proteins (55,56).
Since various HMG1/2-like proteins have been reported to modulate DNA structure by bending, looping, and supercoiling (16 -22, 25, 34, 55), we performed supercoiling assays with recombinant HMGc1 and HMGc2. Both proteins were incubated at increasing concentrations with relaxed, closed circular plasmid DNA in the presence of topoisomerase I. Analysis of the deproteinized DNA in agarose gels revealed that HMGc1 and HMGc2 have the capability to insert superhelical turns into relaxed plasmid DNA. The efficiency of both proteins is similar in this assay (Fig. 6), and the superhelical turns inserted by HMGc1/2 proved to be negative, as evident from electrophoresis of the DNA in the presence of chloroquine (data not shown).
HMG1/2-like proteins that bind to DNA structure specifically and are capable of modulating DNA structure are considered to be architectural elements facilitating the formation of complex nucleoprotein structures (57,58). In this context, it appears conceivable that maize HMGc1/2 and related proteins of other species could play a similar role in plant chromatin. Further work will be needed to solve the problem of the biological function of HMG1/2-like proteins, but this might prove to be difficult since there is a significant interchangeability of different architectural elements in (at least some) nucleoprotein complexes (15,16,24,25).