Functional Interaction between the DNA Binding Subunit Trimerization Domain of NF-Y and the High Mobility Group Protein HMG-I(Y)*

The mammalian transcription factor, NF-Y(CBF), contains three known subunit components, NF-YA (CBF-B), NF-YB(CBF-A), and NF-YC(CBF-C), which are all required to reconstitute specific CCAAT box DNA binding activity. In this study, the high mobility group chromosomal protein, HMG-I(Y), has been shown to activate NF-Y in transient transfections in vivo using the natural murine α2(I) collagen promoter and a multimerized version of the proximal NF-Y(CBF) CCAAT box element. In vitro analysis of the α2(I) collagen promoter region inclusive of the NF-Y(CBF) binding site (−106 to −65 base pairs) failed to identify any high affinity HMG-I(Y) DNA-binding sites. However, the heterotrimeric NF-Y complex, as well as the NF-YA subunit alone, was shown to stably interactin vitro with both HMG-I(Y) and phosphorylated HMG-I, as modified by casein kinase II, using far Western and protein-protein interaction solution assays in the absence of CCAAT box DNA. Furthermore, the interaction between HMG-I(Y) and NF-Y was mapped to the highly conserved DNA binding-subunit interaction domain (DBD) of the NF-YA subunit and to a single AT-hook motif in HMG-I(Y). Recombinant HMG-I was also found to stabilize the CCAAT box DNA binding activity of recombinant NF-Y, as well as the native NF-Y complex, in vitro. Together, these results suggest a functional HMG-I(Y) protein binding site has been identified in the NF-Y complex and mapped to the conserved DBD and AT-hook regions of NF-YA and HMG-I(Y), respectively. This protein-protein interaction site may function to modulate NF-Y activity through stabilization of NF-Y binding to its CCAAT box DNA-binding site.

Nuclear factor-Y (NF-Y) 1 (1), also referred to as the CCAATbinding factor, CBF (2), utilizes the highly conserved regions located in three heterologous subunits to create a DNA binding-subunit interaction domain (DBD) that specifically recog-nizes a CCAAT box motif found in the promoter and enhancer regions of many eukaryotic genes (3). The Y box element of all murine major histocompatibility complex (MHC) class II promoters, for example, contains a highly conserved inverted CCAAT DNA sequence whose conservation extends into the flanking regions, as well as its overall physical location in relation to the highly conserved 5Ј X box element, and plays a critical role in MHC class II gene transcription (4). The human and murine NF-YABC DBD elements have been strongly conserved in evolution (5), are highly related to the corresponding regions in the Saccharomyces cerevisiae HAP2/3/5 CCAAT box factor complex, and can be functionally interchanged with these yeast subunits (6,7). Recently, the DBD regions of the NF-YB and NF-YC subunits have been suggested to interact through a protein-protein histone-fold "handshake" motif (8,9) in a manner analogous to the histone proteins, H2B and H2A, respectively (10). In addition, the NF-YA(CBF-B) subunit has been shown to interact with the NF-YB(CBF-A):NF-YC(CBF-C) heterodimer and neither of these subunits alone suggesting that a unique interaction structure for the NF-YA(CBF-B) subunit is created through conformational changes associated with dimerization of NF-YB(CBF-A):NF-YC(CBF-C) (7).
HMG-I(Y) belongs to a group of abundant low molecular mass non-histone chromosomal proteins that possess three copies of a reiterated 9-amino acid motif (the A⅐T hook) that interacts with the minor groove of many AT-rich DNA sequences (11). These proteins are soluble in 5% perchloric acid, resistant to heat denaturation, can alter DNA structure, and have been shown to be important regulators of gene transcription as well as involved in chromatin structure. HMG-I(Y) proteins are substrates for p34 cdc2 /cyclin B kinase (12) and casein kinase II (CKII) (13,14), and in both cases phosphorylation significantly decreases specific DNA binding activity. HMG-I(Y) DNA-binding sites have been identified in functional regions of many gene promoters which include interleukin-4 (15), interleukin-2 receptor ␣-chain (16), lymphotoxin (17), the human papovavirus JC (18), HLA DR␣ (19), and the CD28 response elements within granulocyte-macrophage colony-stimulating factor and interleukin-2 (20). These sites are often found in close proximity to the DNA-binding sites of known transcription factors (e.g. NF-B (21), Tst-1/Oct-6 (18)), and in particular are critical for viral induction of the human IFN-␤ gene (22)(23)(24). In addition, HMG-I(Y) has been shown to bind to the basic leucine zipper region of activating transcription factor-2 (ATF-2 195 ) which promotes the dimerization of this factor and stimulates its binding to the IFN-␤ promoter (25). These observations suggest that HMG-I(Y) plays a critical role in IFN-␤ gene regulation through a combination of effects on transcription factor subunit interactions, factor binding stabilization, DNA bending, exclusion of specific factor DNA binding (e.g. ATF -2 192 ) (25), and assembly of a multi-component enhanceosome (24). HMG-I(Y) binding to the highly conserved regions of Elf-1 (Ets domain) (16) and Tst-1/Oct-6 (POU domain) (18) through protein-protein interactions have been identified. These studies suggest these DNA-independent HMG-I(Y) protein binding sites also play important overall functional roles in modulating specific transcription factor transactivation potential.
Type I collagen protein is composed of a heterotrimer of two ␣1 and one ␣2 subunits and is the predominant fibrillar collagen protein present in bones and tendons. The transcriptional regulatory elements of the murine ␣2(I) collagen promoter have been extensively investigated using both transient transfection, transgenic animal, and in vitro cell-free transcription approaches, and the proximal CCAAT box motif at Ϫ87 nucleotides has been shown to play an important functional role in its cellular regulation (26 -28). In this report, the murine ␣2(I) collagen promoter and multimerized version of the ␣2(I) collagen NF-Y(CBF) binding site have been used to investigate the role of HMG-I(Y) as a protein cofactor in modulating NF-Y function. HMG-I has been shown to activate NF-Y in vivo and to interact with the highly conserved DBD region of NF-Y in vitro. These results represent the first report of an additional functional protein component associated with the NF-Y complex besides its core YA/B/C subunits and suggest that HMG-I(Y) may modulate NF-Y activity through direct association with the DBD in the NF-YA subunit.

MATERIALS AND METHODS
Recombinant Plasmids-The ␣2(I) collagen promoter, pH6 (26), was used as template to generate a NF-Y(CBF) CCAAT box site-directed mutation with the Quick-Change mutagenesis technique as described by the manufacturer (Stratagene, La Jolla, CA). Both the wild-type and NF-Y mutant HindIII fragments, which contain the ␣2(I) collagen promoter derived from pH6, were isolated and cloned into the HindIII site of the pGL3 Luciferase vector (Promega) to generate pH6 GL3 and pH6m GL3, respectively.
Cloning of full-length human and murine NF-YA, YB, and YC subunits into the pGEX2T vector (Pharmacia Biotech Inc.) has been described previously (5). BamHI-EcoRI fragments of each subunit were cloned into the pGEX2TK vector to generate pYA 2TK, pYB 2TK, and pYC 2TK. An NF-YA deletion mutant that lacks the C-terminal DBD was created by partially digesting pYA 2TK with BanII and EcoRI and sequentially treating the digestion products with the large subunit of DNA polymerase I (Klenow) and T4 DNA polymerase I under standard conditions to produce blunt ends (29). The purified ϳ5.6-kilobase pair pYA 2TK (BanII/EcoRI) fragment was ligated to itself, and recombinant plasmids that lacked the YA DBD element were identified to yield pYA 2TK(⌬DBD) which lacks the C-terminal 85 amino acids.
The murine YA (DBD) and YB (DBD) elements were derived by polymerase chain reactions from pYA 2TK and pYB 2TK, respectively, as described previously (5) and cloned in-frame with glutathione Stransferase (GST) into the pGEX2TK BamHI site to generate pYA (DBD) 2TK and pYB(DBD) 2TK. The murine YA DBD element contains 69 amino acids and corresponds to nucleotides 873-1081, and the YB DBD element contains 90 amino acids and corresponds to nucleotides 366 -635 as defined previously (30). To generate pYC(DBD) 2TK, human NF-YC (pYC 2T) was digested with BamHI and ScaI, and the 430-bp YC(DBD) fragment was ligated in-frame with GST into the BamHI-SmaI sites of pGEX2TK. The human YC(DBD) plasmid contains the N-terminal 143 amino acids, nucleotides 1-430 as defined previously (5,7). Cloning of NF-YB(His-YB) and NF-YC(His-YC) into the pET15b vector (Novagen, Madison, WI) has been described previously (5). NF-YA(His-YA) was cloned into the XhoI-BamHI sites of pET15b using polymerase chain reaction. Human HMG-I(His-HMG-I) was cloned by polymerase chain reaction using the GST-HMG-I plasmid (23) as template into the BamHI site of pET15b. GST-HMG-I-  was prepared by digesting GST-HMG-I with SmaI and EcoRI and ligating the blunt-end vector following treatment with T4 DNA polymerase I and contains the N-terminal 37 amino acids. GST-PC4 was digested with BamHI, and the 275-bp PC4 fragment was cloned inframe with GST into the BamHI site of pGEX 2TK to generate pPC4(⌬C103-127) 2TK which lacks the C-terminal 35 amino acids, nucleotides 277-384 as defined previously (31). Plasmid constructions were prepared according to standard techniques and verified by restriction endonuclease and DNA sequence analyses (29).
Nuclear extracts were prepared by the method of Dignam et al. (37), as described previously (37). EMSA binding reactions (25 l) were performed in 10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, and with or without 2 g of the alternating copolymer, poly(dI-dC) (Pharmacia). Nuclear extract protein (ϳ2-10 g) was added following addition of competitor DNA but prior to addition of ϳ0.2 ng of specific 32 P-labeled DNA probe. Binding reactions were performed at room temperature for a total of 30 min. Affinity purified ␣-YB and ␣-Oct-1 polyclonal antibodies were prepared as described previously (5). Anti-HMG-I polyclonal antibodies were kindly provided by D. Thanos (23), and anti-PC4 antibodies were kindly provided by R. Roeder (31). Antisera were added to EMSA reactions either 15 min before or 15 min after the addition of 32 P-DNA probes depending on the experiment. Protein-DNA complexes were resolved using a 5% native polyacrylamide gel (30:1 acrylamide/bisacrylamide ratio) containing 0.5 ϫ TBE (50 mM Tris-HCl, 50 mM boric acid, 1 mM EDTA). Gels were electrophoresed at ϳ150 V at room temperature for 1.5 h. Dried gels were exposed to Kodak XAR-5 film with an intensifying screen at Ϫ80°C.
HeLa nuclear extracts were depleted of HMG-I activity by first passing nuclear extracts over DEAE-Sepharose Fast Flow (Sigma) which had been equilibrated in BC-420 (50 mM Tris-HCl (pH 7.9), 0.42 M KCl, 20% glycerol, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride). The flow-through fraction was dialyzed extensively against several changes of BC-420 using a Spectra/Por 6 (25-kDa cutoff) membrane (Spectrum, Houston, TX). Insoluble material was removed by centrifugation; the supernatant was diluted to 0.1 M KCl with BC buffer and then applied to a heparin-agarose column. The column was developed using a step KCl gradient, and NF-Y was eluted in the BC-420 fraction (0.42 M KCl). This fraction was further dialyzed against BC-100 and stored in aliquots at Ϫ80°C. Depleted HeLa nuclear extracts retain NF-Y but lack HMG-I(Y) DNA binding activity as measured by EMSA using the IFN-␤ probe. To prepare the HeLa YAF fraction HeLa nuclear extracts were passed over DEAE in 0.42 M KCl as described above, and then the flow-through fractions were centrifuged through Centricon 30 filtration devices (Amicon, Beverly, MA). The filtrate fraction was diluted to 0.1 M KCl and the buffer exchanged to BC-100 using Centricon 3 filtration devices (Amicon). The HeLa YAF fraction is devoid of NF-YA, ϪYB, and ϪYC activities as measured by recombination experiments using complementing combinations of recombinant NF-Y subunits in EMSA. 2 An NF-Y CCAAT box DNA affinity column was prepared by coupling a monomeric murine ␣2(I) collagen CCAAT DNA oligomer to CNBractivated Sepharose 4B (Pharmacia) according to the manufacturer. NF-Y was further purified by passing HeLa-depleted nuclear extracts over this CCAAT box DNA affinity column in the presence of the nonspecific DNA competitor, poly(dI-dC) (39). The column was developed using a step KCl (BC buffer) gradient, and NF-Y was eluted in the BC-420 fraction. This material was concentrated and the buffer exchanged to BC-100 using Centricon 10 filtration devices (Amicon). Protein concentrations were determined using the Bradford dye-binding assay (Bio-Rad) with bovine serum albumin (BSA) as the standard (40).
Expression and Purification of Recombinant Proteins-GST-YA, GST-YA(⌬DBD), GST-YA(DBD), GST-YC(DBD), and GST 2TK were expressed and purified from the soluble fraction of Escherichia coli DH5␣ lysates as described by Smith and Johnson (41). GST-YC was induced for 20 min at 37°C using 1 mM isopropyl-␤-D-thiogalactopyranoside as described previously (7), lysed in 1 ϫ in phosphate-buffered saline (PBS), 0.1% Triton X-100, and purified from the DH5␣-soluble fraction. GST-YB, His-YB, and His-YC proteins were expressed and purified as described previously (5). His-YA was purified from the soluble fraction of E. coli BL21(DE3) after induction with 1 mM isopropyl-␤-D-thiogalactopyranoside for 3 h at 37°C using Ni 2ϩ ion affinity chromatography (Novagen). His-HMG-I was purified from the soluble fraction of BL21(DE3) after induction with 1 mM isopropyl-␤-D-thiogalactopyranoside for 1 h at 37°C with the following modifications. Crude bacterial lysates were first heated to 65°C for 10 min, cooled on ice for 20 min, then centrifuged at 10,000 rpm in a SS34 rotor (Dupont, Wilmington, DE) for 20 min at 4°C. The supernatant fraction was made 5% in perchloric acid, left at room temperature for 10 min, and then centrifuged at 10,000 rpm for 10 min at 4°C. The pH of the supernatant was adjusted to 7.5 with 10 N NaOH and then dialyzed against 1 ϫ PBS at 4°C for 3 h using a Spectra Por 6 (8-kDa cutoff) membrane (Spectrum). His-HMG-I was further purified by Ni 2ϩ ion affinity chromatography (Novagen) and passage through a Centricon 30 filtration device (Amicon, Beverly, MA) in BC-420. The filtrate fraction was concentrated using a Centricon 10 device and the buffer exchanged to BC-100. Recombinant protein purity was assessed using Coomassie Blue staining of SDS-PAGE gels, and purified fractions were stored at Ϫ80°C.
Far Western Analysis-pGEX2TK fusion proteins were labeled using heart muscle creatine kinase (HMK) (Sigma) and [␥-32 P]ATP (DuPont) as described for GST expressed from pGEX 2TK (Pharmacia). Unincorporated 32 P label was removed by passing the reaction mixture over a CentriSep spin column according to the manufacturer (Princeton Separations, Adelphia, NJ). Recombinant proteins and pre-stained molecular mass protein standards (New England Biolabs, Inc., Beverly, MA) were electrophoresed through SDS-PAGE gels and transferred to 0.45-m nitrocellulose filters (Schleicher & Schuell) at 4°C using a Mini Trans-Blot transfer apparatus according to the manufacturer (Bio-Rad). Filters were sequentially treated for 15 min each at room temperature with 6, 3, 1.5, and 0 M guanidine HCl solutions that contained 20 mM HEPES (pH 7.9), 50 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, and 10% glycerol (buffer A). Filters were blocked for 2 h at room temperature in buffer A (5% non-fat dry milk) and then incubated overnight at 4°C with the labeled protein probe in buffer A (0.1% non-fat dry milk, 0.1% Nonidet P-40). Filters were washed 5 ϫ for 5 min each using 1 ϫ PBS, 0.2% Triton X-100 at room temperature and then exposed to XAR-5 film with an intensifying screen at Ϫ80°C.
In Vitro Protein-Protein Interaction Assays-His-HMG-I was purified as described above and 32 P-labeled using casein kinase II (CK II) and [␥-32 P]ATP as described for CKII by the manufacturer (Calbiochem). His-HMG-I (ϳ200 ng) was incubated with 10 units of CKII in a 30-l reaction that contained 20 mM Tris-HCl (pH 7.5), 50 mM KCl, and 10 mM MgCl 2 at 30°C for 1 h and then diluted to 100 l using BC-100 that contained 50 mM EDTA. Unincorporated [ 32 P]ATP was removed by passage of the reaction mixture over a CentriSep column. GST-YA(DBD) was cleaved with thrombin (Novagen) as described previously (31) and 32 P-labeled with HMK. Equal amounts of purified GST, GST fusion proteins, and NF-Y subunits (ϳ100 ng each) were immobilized on glutathione-agarose beads (ϳ20 l packed volume) (Sigma). Proteins were incubated with glutathione beads for 30 min at room temperature in 100 l of BC-100 and then washed with 3 ml of BC100. 32 P-HMG-I (10 l) was added to each reaction, and incubations in BC100 (100 l) were continued at 4°C for 1 h with intermittent mixing. Glutathione beads were then washed five times with 1 ml of BC-100 each which contained 0.5% Nonidet P-40 at 4°C. SDS-PAGE sample buffer (100 l) was added, and the samples were heated to 95°C. Aliquots of each bead reaction (20 l) were separated on SDS-PAGE gels. Dried gels were exposed to XAR-5 film at room temperature. In experiments involving interaction of affinity purified NF-Y and GST fusion proteins, glutathione-agarose beads were washed five times with 1 ml of BC-100 each following the 1-h binding reaction at 4°C. Proteins were eluted using BC-420, concentrated, and the buffer exchanged to BC-100 using Centricon 10 filtration devices. Eluted fractions were analyzed using EMSA.

The High Mobility Group Protein HMG-I Activates NF-Y in
Vivo-Biochemical characterization of the native NF-Y transcription factor complex in a variety of human and murine cell lines that represent distinct stages of B lymphocyte development and adipocyte differentiation has suggested that the NF-YABC complex can associate with a group of newly identified protein cofactors (YAFs, NF-Y Associated Factors). 2 The partially purified YAF fraction derived from mature B cells, for example, has been shown to confer the biochemical property of high affinity anion exchange binding to NF-Y derived from plasma B cells and several other MHC class II Ϫ cell lines (e.g. HeLa, 3T3-L1). 2 Further biochemical analyses have suggested the YAF fraction derived from a variety of cell types (e.g. HeLa) contains an HMG-I(Y)-like activity as defined by specific interaction with a known IFN-␤ HMG-I(Y) DNA-binding site, solubility in 5% perchloric acid, resistance to denaturation and loss of DNA binding activity following exposure to elevated temperature (e.g. 65°C), and competition of IFN-␤ DNA binding activity by poly(dI-dC) DNA and specific ␣-HMG-I antibodies (Fig. 1). 2 These observations have led to consideration of the possibility that HMG-I(Y) is in fact one of the YAF proteins and is involved in regulating NF-Y function through direct proteinprotein interactions.
The well-characterized murine ␣2(I) collagen promoter (2, 26 -28) was chosen to assess the functional effect of HMG-I on NF-Y-mediated transcription in vivo. A multimerized NF-Y(CBF) reporter construct derived from the murine ␣2(I) collagen promoter, pFC1, was also tested since a number of studies have documented the relatively low in vivo activity of single and multimerized NF-Y binding site reporters derived from MHC class II promoter Y box elements in lymphocyte cell lines (21,42), and pFC1 has been shown to exhibit moderate relative reporter activity in a variety of cell types (26 -28). As shown in Fig. 2 expression of HMG-I in HeLa cells resulted in an ϳ6-fold activation of the natural ␣2(I) collagen promoter, and an ϳ2.7fold activation of the multimerized NF-Y site reporter, pFC1. In contrast, this level of HMG-I expression did not activate the NF-Y site-directed mutant reporter, pH6m GL3, or the mutant multimerized NF-Y site reporter, pFC2. These results suggest that HMG-I modulates the functional activation of the ␣2(I) collagen promoter directly through NF-Y and its binding to this CCAAT box motif since disruption of the NF-Y(CBF) binding site in both pH6m GL3 and pFC2 resulted in no activation by CMV-HMG-I.
In an attempt to identify potential HMG-I(Y) DNA-binding sites in the murine ␣2(I) collagen promoter comparative EMSA analyses were performed using regions contained in pFC1 (Fig.  3). The ␣2(I) collagen DNA probes, L-collagen (Ϫ101 bp to Ϫ65 bp) and the S-collagen probe (Ϫ98 to Ϫ72 bp), were both shown to specifically bind HeLa cell NF-Y (panels A and B respectively, lanes 1-4) as compared with a control MHC class II E␣ NF-Y binding site (lanes 5-8). L-and S-collagen probes were also compared with a well characterized HMG-I(Y) DNA-binding site from the IFN-␤ PRDII element (23) and to the AT-rich MHC E␣ probe (36). Recombinant GST-HMG-I was observed to bind to the L-collagen probe with very low affinity (panel A, lane 9) in comparison to the IFN-␤ (lane 11) and E␣ (lane 13) probes, and GST-HMG-I was not observed to bind to the Scollagen probe (panel B, lane 9), whereas high affinity binding of GST-HMG-I to the control IFN-␤ probe was observed (lane 13). To extend these observations further highly purified recombinant HMG-I expressed from the pET15b vector was also tested, and HMG-I binding to either the L-collagen (panel C, lane 2) or S-collagen (panel C, lane 5) probes was not observed. Together these results suggest that the ␣2(I) collagen promoter region in pFC1(Ϫ101 to Ϫ65 bp) which contains the functional NF-Y(CBF) site does not contain high affinity HMG-I(Y) DNAbinding sites.
Physical Interaction between NF-Y and HMG-I(Y)-The previous analyses suggested that HMG-I(Y) may function to activate NF-Y through direct protein-protein interactions. This hypothesis was examined using a far Western blot analysis in Figs. 4 and 5. The full-length NF-YA subunit was first 32 Plabeled, recombined with excess purified NF-YB and NF-YC subunits in vitro, and the 32 P-NF-Y complex was used to probe a nitrocellulose filter which contained GST, HMG-I and several other GST-fusion transcription factors (Fig. 4, panel A). Strong interaction of the 32 4, and 6). Full-length 32 P-NF-YB was recombined with excess purified NF-YA and NF-YC subunits in a similar manner and tested as described above; again, a strong interaction was observed between this 32 P-NF-Y complex and both HMG-I and PC4/p15 (panel B). In an attempt to determine if all three NF-Y subunits are required for these in vitro interactions, individual subunits were labeled and tested, alone or in specific NF-Y subunit combinations in far Western analyses. Full-length 32 P-YA was observed to specifically interact with both HMG-I and PC4/p15 (panel C), whereas the individually labeled subunits ( 32 P-YB, 32 P-YB(DBD), 32 P-YC, 32 P-YC(DBD)) or the heterodimeric complexes ( 32 P-YB⅐YC complex and 32 P-YB (DBD)⅐YC(DBD)) interacted in a strong nonspecific manner with the control GST protein, all GST fusion transcription factors, and low abundant bacterial proteins present in these preparations. 2 Both NF-YB and NF-YC subunits contain several hydrophobic regions (7), are tightly associated with each other (44), and may form many nonspecific protein interactions when analyzed in this manner using far Western assays. Together these results suggest that the intact NF-Y complex is capable of stable interaction with HMG-I and PC4; however, the NF-YA subunit directly mediates these interactions, and its contact with HMG-I and PC4 does not depend on additional interactions with NF-YB and NF-YC. As an additional control for these experiments, 32 P-GST was tested against the panel of proteins described in panel A and failed to interact with HMG-I, PC4, or any of the other GST fusion proteins. 2 The ability of full-length NF-YA to interact with HMG-I in vitro suggested that the highly conserved DNA binding-subunit interaction domain (DBD) of YA (5, 7) could contain the region which stably bound both HMG-I and PC4. To test this possibility, the YA(⌬DBD) deletion mutant was first used to probe a panel of GST fusion proteins (panel D). The N-terminal activation domain of NF-YA, YA(⌬DBD), did not form a stable complex with HMG-I (lane 2), PC4 (lane 3), or any other transcription factor tested (lanes 4 -6) and suggested that the DBD element of NF-YA was sufficient for the physical interaction between HMG-I and PC4 as measured by far Western analysis.

Mapping of Regions in NF-Y and HMG-I(Y) That Are Necessary for Protein-Protein Interaction-HMG-Y is identical to
HMG-I in composition except HMG-Y has an 11-amino acid region deleted as a result of an alternative mRNA splicing event (11,23). A previous study has identified a small 10-amino acid region in HMG-Y that is responsible for physical interaction with the Tst-1/Oct-6 POU domain (18). To more accurately map the region in HMG-I(Y) that interacts with NF-Y wildtype HMG-Y, this series of HMG-Y mutants was analyzed The HeLa YAF fraction was tested for HMG-I(Y) DNA binding activity using the IFN-␤ PRDII element (23) (lanes 2-5) and specific antisera (lanes 6 -9) in EMSA. In lanes 6 -9 serum and antiserum were added first before addition of the 32 P-IFN probe ("Materials and Methods"). HeLa-depleted nuclear extracts (ϳ2 g protein) were also incubated with HeLa YAF proteins, and NF-Y DNA binding activity was measured using the ␣2(I) collagen (S-collagen) CCAAT box DNA probe (lanes 10 -12). Lane 1, 32

FIG. 2. NF-Y is activated by HMG-I in vivo.
Luciferase reporter plasmids were transfected into HeLa cells along with an internal ␤galactosidase control, pCMV␤, and either the CMV-HMG-I expression plasmid (23), or the empty CMV vector, pcDNA3, as described under "Materials and Methods." The results of four independent experiments are expressed as the fold induction of the means and standard errors of the means (filled bars) above the luciferase activity observed with the reporter plasmid alone (stippled bars), which was set to a value of 1.0. using far Western analysis (Fig. 5). Both full-length NF-YA in the complete NF-Y complex (panel A) and the full-length NF-YA subunit alone (panel B) stably interacted with HMG-Y and PC4. HMG-Y mutants H20 -56 and H32-46 which essentially contain AT-hook motifs 1 and 2 and AT-hook motif 2, respectively, formed stable complexes with these 32 P-NF-Y and 32 P-YA probes. However, neither of these probes recognized the HMG-Y mutant (H46 -56) that interacts with the Tst-1/Oct-6 POU domain and contains only a 10-amino acid portion of AT-hook motif 2. These results suggest that at least one intact HMG-I(Y) AT-hook motif is required for stable interaction with either the NF-Y complex or the NF-YA subunit and more specifically suggest that amino acids 32-46 of HMG-Y are suffi-cient for interaction with NF-Y. In addition, these probes failed to recognize the PC4 mutant (⌬C103-127) that lacks its Cterminal 35 amino acids (panels A and B, lane 7, respectively). To provide further evidence that the DBD element of NF-YA interacts with HMG-I(Y), the NF-YA(DBD) subunit was 32 Plabeled and tested together with YB(DBD)/YC(DBD) in the NF-Y(DBD) complex (panel C) and also independently (panel D). Both of these probes recognized full-length HMG-Y; however, interaction with all other HMG-Y mutants was severely impaired. These results suggest that the YA(DBD) element is sufficient for stable interaction with the full-length HMG-I(Y) and that an intact AT-hook is sufficient for stable interaction with the NF-Y complex composed of its full-length subunits.

HMG-I Stabilizes NF-Y Interaction with CCAAT Box DNA
Elements-Previous studies in a variety of systems have demonstrated that HMG-I(Y) is capable of stimulating the DNA binding activity of NF-B, Tst-1/Oct-6, ATF-2 195 , and Oct-2A generally under conditions of dilute protein concentration (18,19,23,25). In the case of ATF-2 195 HMG-I has been shown to also stimulate ATF subunit dimerization (25). To determine if HMG-I is capable of stimulating NF-Y binding by a similar mechanism, the effect of recombinant HMG-I on NF-Y DNA binding activity was analyzed (Fig. 6). Under conditions of low relative NF-Y subunit concentration, HMG-I was observed to specifically stimulate the DNA binding activity of both the His fusion (panel A) and GST fusion (panels B and C) NF-Y complexes using either the L-or S-collagen probes. The effect of HMG-I on native NF-Y derived from HMG-I(Y) depleted HeLa nuclear extracts was also tested (panel D). In all cases HMG-I was found to dramatically stimulate NF-Y CCAAT box DNA binding activity with no apparent effect on the relative mobility of the DNA-protein complex.
To determine if the native NF-Y complex is capable of inter-action with recombinant HMG-I, a glutathione-agarose pulldown assay was performed (Fig. 7). NF-Y was derived from depleted HeLa extracts, affinity purified, and its CCAAT box DNA binding activity was observed to be stimulated specifically by His-HMG-I (lanes 12-18). NF-Y was incubated with glutathione-agarose beads alone and with GST, GST-HMG-I, and GST-Dr1 bound to beads. NF-Y binding to GST-HMG-I was observed (lane 4), and its CCAAT box activity was likewise stimulated by exogenous His-HMG-I (lane 8). These results suggest that the native NF-Y complex can stably interact with HMG-I in solution and further suggest that HMG-I may represent one of several nuclear factors that contribute to the apparent overall high level of NF-Y CCAAT box DNA binding activity which is observed in unfractionated nuclear extracts. Phosphorylated HMG-I Interacts with NF-YA in Vitro-Human HMG-I has been shown to be phosphorylated by CKII in its C-terminal region at serine residues 102 and 103 (11), and the DNA binding activity of CKII-treated HMG-I to a known HMG-I(Y) site in the murine G⑀ promoter has been shown to be reduced ϳ5-fold (14). In addition, p34 cdc2 /cyclin B kinase phosphorylates human HMG-I at threonine residues 53 and 78 and has been shown to reduce its DNA binding activity ϳ20-fold under physiological salt conditions (12). To determine if phosphorylated HMG-I stably interacts with NF-Y in solution recombinant HMG-I was 32 P-labeled using CKII and tested using the NF-Y subunits and several additional fusion proteins in a glutathione bead pull-down assay (Fig. 8). 32  . Strong interaction between 32 P-HMG-I and the general transcription initiation factor, TFIIB, was also observed (lane 6); however, the functional relevance of this interaction remains unknown at present. p34 cdc2 /cyclin B-phosphorylated HMG-I was observed to interact in an identical manner to CKII-treated HMG-I in glutathione bead pull-down assays. 2 These results further support the conclusion that HMG-I(Y) interacts with the NF-Y complex through a protein-protein binding site located in the 69 amino acid YA(DBD) region and that phosphorylation of HMG-I by CKII or p34 cdc2 /cyclin B does not block this interaction.
To support further the conclusion that NF-YA and HMG-I physically interact and that phosphorylation of HMG-I does not promote this interaction, YA(DBD) was cleaved from GST-YA(DBD), 32 P-labeled using HMK, and used to probe a set of GST-HMG-Y(I) mutants (Fig. 9). 32 P-YA(DBD) was observed to interact with full-length HMG-Y, mutants which contain an intact AT-hook motif, and the single N-terminal AT-hook derived from human HMG-I (lane 13), whereas no interaction was observed with the HMG-Y-(46 -56) mutant that contains a disrupted AT-hook motif. In conjunction with far Western analyses, these results suggest that a single AT-hook motif in either HMG-I or HMG-Y is necessary and sufficient to support stable interaction with NF-Y. DISCUSSION Ion exchange analysis of the native NF-Y complex in specific states of cellular differentiation has suggested that NF-Y is associated with a group of previously unknown protein cofactors (YAFs). 2 Further characterization of this operationally defined cellular YAF fraction suggested that an HMG-I(Y)-like activity was present and possibly associated with native NF-Y in the absence of CCAAT box binding sites (Fig. 1). 2 Based on these initial biochemical observations, the functional effect of human HMG-I on a well characterized NF-Y(CBF) reporter plasmid derived from the murine ␣2(I) collagen promoter was tested in vivo (Fig. 2). The murine ␣2(I) collagen gene has been shown to contain a number of functional DNA elements with associated transcription factors that are involved in regulating its expression in a tissue-specific manner. In particular, the proximal CCAAT box DNA element (Ϫ84 bp to Ϫ80 bp) has been shown to bind NF-Y(CBF) and to function in a multimerized configuration together with its TATA box and initiation region (Ϫ41 bp to ϩ54 bp) in a number of cell types (2,27,28). These NF-Y-containing promoter constructs were chosen to examine the possible direct functional effects of HMG-I(Y) on NF-Y both in the context of other transcription factors and in the absence of additional transcription factors as well as the effect of other potential AT-rich HMG-I(Y) DNA-binding sites. Cotransfection of HMG-I into HeLa cells resulted in activation of the collagen reporters, pH6 GL3 and pFC1, which were dependent on an intact wild-type CCAAT box DNA sequence (Fig. 2). These results suggested that HMG-I(Y) could play a functional role in modulating the transactivation potential of NF-Y in vivo through direct interaction with NF-Y. To assess the possibility that HMG-I was interacting with AT-rich DNA sequences in the ␣2(I) collagen promoter, a series of EMSA experiments were performed. Using highly purified recombinant HMG-I high affinity HMG-I(Y) DNA-binding sites were not detected (Fig. 3). These observations suggested that activation of the NF-Y reporter by HMG-I could be mediated through direct protein-protein interactions.
To test this possibility far Western assays were performed using both individually labeled NF-Y subunits and NF-YABC complexes that contained one 32 P-labeled subunit. The NF-Y complex was consistently observed to specifically interact with the two general transcriptional coactivators, HMG-I(Y) and PC4/p15 (Fig. 4). In addition, the NF-YA subunit alone was capable of stable interaction with HMG-I and PC4, whereas a truncated version of YA that contained its N-terminal activation domains (26) but lacked the highly conserved DBD element was incapable of stable interaction with these two coactivators (Fig. 4, panel D). The interaction of NF-Y with either HMG-I(Y) or PC4 did not depend on additional interactions with the NF-YB or NF-YC subunits. Together these results suggested that the YA DBD element, itself, was necessary and sufficient for these observed protein-protein interactions. Of particular note was the observation that NF-YA and the NF-Y complex each stably interacted with full-length PC4, while deletion of the C-terminal 35 amino acids from PC4 completely disrupted this interaction. PC4 has been shown to be a general accessory factor that is involved in the response of RNA polymerase II to transcriptional activators, both in in vivo and in vitro reconstituted systems (31,43). In addition, phosphorylated PC4 has been shown to be functionally inactive in reconstituted cell-free in vitro transcription assays, whereas the purified native non-phosphorylated and E. coli-derived form of PC4 are potent activators in vitro (43,45). Mapping of the PC4-NF-Y interaction site to a region in PC4 that is not hyperphosphorylated in vivo (i.e. the extreme N terminus) suggests that PC4 may interact with NF-Y both in its phosphorylated and nonphosphorylated states, and thereby serve to function as both activator and repressor of NF-Y-mediated RNA polymerase II transcription. Further analyses will be directed at uncovering the functional role of PC4 in modulating NF-Y activity in vivo, determining if these phosphorylated forms of PC4 exhibit differential functional effects on NF-Y in specific promoter contexts, and if the coactivators, HMG-I(Y) and PC4, influence each other's interaction with the YA DBD. 2 To begin to map the functionally relevant regions in both the NF-Y complex and in the HMG-I(Y) molecule responsible for these interactions, far Western assays were performed using truncated versions of NF-Y subunits and a series of HMG-Y deletion mutants (18) (Fig. 5). An HMG-Y molecule that contained 1 or 2 copies of its AT-hook motif (11) interacted stably with either the full-length NF-YA subunit alone or in the NF-Y complex; however, partial deletion of a single HMG-Y AT-hook motif 2 completely disrupted these interactions. These results suggest that at least one intact copy of the reiterated AT-hook motif is necessary and sufficient for stable interaction with either the YA subunit or the NF-Y complex. To further map the interaction domain in YA the YA(DBD) probe was tested alone and in the NF-Y complex. Stable interaction with full-length HMG-I(Y) was observed in both cases; however, this interaction was severely weakened in mutants that contained 1 or 2 copies but not 3 copies of the AT-hook motif when the YA(DBD) probe was tested alone. In contrast, the HMG-Y mutant with 2 intact AT-hook motifs interacted stably with the intact NF-Y(DBD) complex. The conformation of YA(DBD) in a complete NF-Y complex may be slightly altered in comparison to YA(DBD) alone, and its interaction with these HMG-Y mutants may be facilitated to a greater extent. Analysis of the HMG-I(Y) mutants using the YA (DBD) as probe further supported the conclusion that a single intact AT-hook motif was required for stable interaction (Fig. 9) and suggested that the reiterated AT-hook motifs in HMG-I(Y) may function both in DNA bending and in specific protein-protein interactions. NF-YA was not observed to discriminate between particular AT-hook motifs, as human AT-hook (motif 1) and murine AT-hook (motif 2) each were sufficient to support stable interaction. In particular gene contexts a single HMG-I(Y) molecule may function to stabilize transcription factor-DNA binding through a combination of its own binding to a proximal minor groove AT-rich sequence and through direct interaction with an adjacent transcription factor through one of its additional AT-hook motifs or through direct effects on the conformation of multi-subunit proteins, such as NF-Y. These results suggest that HMG-I(Y) interacts with the highly conserved DBD element in the NF-YA subunit, and this interaction possibly maps to a face of the YA ␣-helix which is not involved in the primary subunit interactions with the YB/YC heterodimer (29), since inclusion of the YB/YC subunits in interaction assays neither prevented nor were required for the observed binding. Further mutational analyses will be directed at determining the precise amino acids in the 69-amino acid YA(DBD) element which are critical for HMG-I(Y) interaction and their relationship to the yeast homolog, HAP2(DBD).
In an attempt to identify the biochemical mechanisms that underlie the observed functional effect of HMG-I on NF-Y in vivo, in vitro experiments were performed to examine NF-Y DNA binding activity in the presence of HMG-I. Under conditions of low NF-Y subunit concentration, recombinant HMG-I was observed to stimulate NF-Y DNA binding activity to the ␣2(I) collagen CCAAT box. This effect was specific to HMG-I and also was observed using an HMG-I-depleted HeLa cell nuclear fraction that contained native NF-Y. Co-immunoprecipitation experiments using anti-NF-YB antibodies did not conclusively identify in vivo association of NF-Y and HMG-I. 2 In the case of Tst-1/Oct-6 (18), HMG-I has not been shown to alter its EMSA mobility position, whereas slight changes in mobility with NF-B (23) and ATF-2 (22) have been observed.
In these examples HMG-I(Y) has been suggested to stabilize transcription factor-DNA interactions. HMG-I has not been shown to be present in a ternary complex in association with NF-Y or any of the transcription factors identified in these systems in vitro or to be associated with these factors in vivo. HMG-I(Y) may promote transcription factor conformational changes that in turn stabilize overall interaction with their respective DNA-binding sites and only be transiently associated in vivo. In a related system the Tax protein of human T cell leukemia virus type-1 has been shown to stabilize ATF-DNA binding interactions; however, stable complex interaction was not observed using a variety of EMSA techniques but only using a DNA-co-immunoprecipitation assay (38). Together these studies suggest that ternary complexes in general are sensitive to analysis under EMSA conditions in vitro but in particular instances ternary complexes can be observed in solution. In conclusion, this study provides the first report of an additional protein cofactor functionally associated with the NF-Y complex, maps the interaction site between HMG-I(Y) and NF-Y to a single AT-hook motif and the highly conserved DBD element of the NF-YA subunit, respectively, and suggests that HMG-I can stabilize NF-Y-CCAAT box interactions through direct protein-protein interaction.