Phosphorylation by cdc2 Kinase Modulates DNA Binding Activity of High Mobility Group I Nonhistone Chromatin Protein*

Chromatin high mobility group protein I (HMG-I) is a mammalian nonhistone protein that has been demonstrated both in vitro and in vivo to preferentially bind to A. T-rich sequences of DNA. Recently the DNA- binding domain peptide that specifically mediates the in vitro interaction of high mobility group protein (HMG)-I with the narrow minor groove of A-T-DNA has been experimentally determined. Because of its predicted secondary structure, the binding domain peptide has been called “the A*T hook” motif. Previously we demonstrated that the A*T hook of murine HMG-I protein is specifically phosphorylated by purified mammalian cdc2 kinase in vitro and that the same site(s) are also phosphorylated in vivo in metaphase- arrested cells. We also found that the DNA binding affinity of short synthetic binding domain peptides phosphorylated in vitro by cdc2 kinase was significantly reduced compared with unphosphorylated pep- tides. Here we extend these findings to intact natural and recombinant HMG-I proteins. We report that the affinity of binding of full-length HMG-I proteins to A*T-rich sequences is highly dependent on ionic conditions and that phosphorylation of intact proteins by cdc2 kinase reduces their affinity of in vitro binding to A-T-DNA by about 20-fold when assayed near nor- mal mammalian physiological salt concentrations. To investigate the ionic strength dependence of recombinant hu- man HMG-I binding to substrate DNA, aliquots of a concentrated NaCl solution were added to the usual assay mix containing 50 mM NaCl and 100 nM DNA, and titration with dye was performed as described. NaCl concentration was varied from 50 to 188 mM. In all cases, the concentration of HMG-I or phosphorylated HMG-I was fixed at 10 nM. The final assay volume (before addition of dye) was fixed at 3.00 ml. and Vivo partially or 32P presence

conformation and function of chromatin (1). HMG-I and HMG-Y are the most thoroughly characterized members of an isoform subgroup of these chromatin proteins (collectively called the HMG-I family) that are probably produced from alternatively splicing precursor mRNAs (2,3). HMG-I proteins (not to be confused with the unrelated mammalian HMG-1 chromatin proteins; Ref. 1) are DNA-binding proteins that in vitro preferentially bind to stretches of A. T-rich sequence (4)(5)(6)(7). Recently, members of the HMG-I family have been demonstrated to specifically bind in vitro to the narrow minor groove of A. T-DNA by means of an 11-amino acid peptide binding domain, which, because of its predicted structure is called the "A.T hook motif" (8). I n uiuo the HMG-I proteins have also been immunolocalized to the A-T-rich G/ Q and C bands of mammalian metaphase chromosomes (9) suggesting that these proteins may play a critical role in the changes in chromosome structure that occur during the cell cycle (91. 2 HMG-I proteins are also of considerable biological interest because they are expressed at elevated levels in actively proliferating cells (10, 11) and have been observed to be a characteristic feature of undifferentiated or neoplastically transformed cellular phenotypes (2, 3, [12][13][14][15]. High HMG-I levels have been found to be a consistent feature of rat and mouse malignant tumors (12)(13)(14) and have been suggested to be protein markers for both neoplastic transformation (13) and metastatic potential (16).
Their specific A. T-binding characteristics have led to several postulated functions for the HMG-I proteins including nucleosome phasing (5), the 3'-end processing of mRNA transcripts (17), and possible involvement in amplification of autonomously replicating sequences in mouse cells (18,19). The HMG-I proteins may also be involved in transcriptional regulation of genes containing, or in close proximity to, A. Trich control sequences of DNA. Specific binding of HMG-I or HMG-I-like proteins have been demonstrated for regulatory sequences found in the 3"untranslated regions (3'-UTRs) of many lymphokine and growth factor genes (20,21) and to the promoter/enhancer sequences of several other types of mammalian genes (22,23): It is known from their primary sequences that the HMG-I proteins (3,7,8) have the overall structure of typical Ptashne-type transcriptional activator proteins possessing both a DNA binding domain and a highly acidic COOH terminus (25). Furthermore, in vitro binding of HMG-I has been demonstrated to alter the conformation and stability of A-T-rich regions of DNA (24), properties often associated with DNA-binding gene regulatory proteins. Phosphorylation of HMG-I by cdc2 Kinase Alters DNA Binding the most highly phosphorylated protein species in the nucleus, and the relative degree of phosphorylation is cell cycle-dependent (11,(26)(27)(28)41,42). In mammals the extensive phosphorylation of histone H1 that occurs in proliferating cells (29,30) is catalyzed by an enzyme homolog of the p3PdC2 serine/threonine protein kinase of yeast (formerly called growth-associated histone H1 kinase; Ref. 31), the activity of which is elevated at mitosis (32)(33)(34). In both yeast (35)(36)(37)(38) and mammalian cells (39,40), cdc2 kinase is required at two points in the cell cycle: (i) at a point in late GI (called "start" in yeast), which commits the cells to S phase and a round of DNA synthesis, and (ii) at a point in late G2 just before mitosis that commits the cell to undergo division. Since homologs of cdc2 kinase exist in all types of eukaryotic cells examined (43)(44)(45)(46)(47)(48), it appears that this type of protein kinase participates in universal mechanisms for the control of cell cycle progression (35). Like histone H1, the mammalian HMG-I proteins have been demonstrated to be phosphorylated in vitro by purified mammalian cdc2 kinase and, furthermore, the same modified amino acid residues have also been shown to be phosphorylated in vivo in metaphase-arrested cells (41,42). Recently we demonstrated that in vitro cdc2 kinase phosphorylation of a short synthetic peptide corresponding to the A. T hook binding domain residues of HMG-I significantly reduces the DNA binding affinity of the peptide in vitro (42). In the present report we considerably extend these previous findings for short synthetic peptides to the effects of in vitro cdc2 phosphorylation on intact HMG-I proteins. We demonstrate that the DNA binding affinity of whole HMG-I proteins is highly dependent on ionic conditions and that phosphorylation of HMG-I by purified cdc2 kinase reduces its affinity of in vitro binding to A.T-rich DNA by about 20-fold near normal mammalian physiological salt concentrations. Furthermore, in cell synchronization studies we demonstrate that murine HMG-I proteins in viuo are phosphorylated in a cell cycledependent manner on the same amino acid residues modified by purified cdc2 kinase in vitro. Together these results strongly support the notion that in vivo the HMG-I proteins are natural substrates for mammalian cdc2 kinase and that their cell cycle-dependent phosphorylation by this enzyme(s) significantly modulates their DNA binding affinity thereby possibly altering their biological function(s1.

MATERIALS AND METHODS
Purification of Native Murine HMG-I Protein-crude HMG protein samples were obtained by extracting murine R1.l ascites cells (American Type Culture Collection TIB42) with 5% perchloric acid (11). Proteins were precipitated from the pooled extracts by adding an equal volume of 50% trichloroacetic acid. The precipitated HMG proteins were collected by centrifugation, washed with acetone, dried, and resuspended in water. HMG-I proteins were purified from the crude extract by reverse-phase high performance liquid chromatography (RP-HPLC) on a C-4 column, as previously described (11). Following chromatography, the purity of the protein samples was determined by acid-urea polyacrylamide gel electrophoresis (11). HMG-I protein concentrations were determined spectrophotometrically using the relationship czz0 = 74,000 liters/mol cm (8).
Bacterial Production of Recombinant Human HMG-I Protein-Employing standard recombinant DNA procedures (49,50) sitespecific mutations were introduced into a full-length human HMG-I cDNA (clone 7C; Ref. 3) so that the protein-coding region of the cDNA could be ligated into an appropriate plasmid expression vector for production of recombinant HMG-I in bacteria. Briefly, a commercial in vitro mutagenesis kit (U. S. Biochemical Corp.) based on the method of Vandeyar et al. (51) was used to introduce an NdeI tide 1) using the antisense 25-mer mutagenesis primer 5"CTCGAC restriction enzyme cut site in the HMG-I initiation codon (at nucleo-TCACTCATATGCCCTTCTC-3'. Similarly, a BamHI cut site was one nucleotide past the termination codon of the protein coding introduced into the 3'-untranslated tail region (at nucleotide 325), region of cDNA (31, using the 25-mer primer 5"GGCGGCTCGGGA TCCTCACTGCTCC-3'. The isolated 325-base pair NdeIIBamHI restriction fragment was directionally subcloned into the plasmid expression vector PET-3a (52) between the unique NdeI and BamHI sites so that the HMG-I protein coding sequence was ligated in-frame to reconstitute a proper initiation codon downstream of the powerful phage T7 (~1 0 ) promoter. The resulting recombinant vector, pET7C, was introduced into the double lon/ompT protease mutant B strain of E. coli BLBl(DE3)pLysS. This host strain is lysogenic for the bacteriophage DE3 which carries the T7 RNA polymerase gene under the control of the IPTG-inducible lacUV5 promoter (53). Addition of 0.4 mM IPTG to a growing culture of the BL21(DE3)pLyS lysogen induces large amounts of T7 RNA polymerase which, in turn, efficiently transcribes the target HMG-I DNA in the pET7C plasmid. Induction of bacteria followed the protocols of Studier et al. (53). After 4 h (37 'C) of IPTG treatment the induced cells were collected by centrifugation and immediately frozen and stored at -90 "C until extraction.
Purification of Recombinant Human HMG-I Protein-Pellets of induced bacteria were resuspended and lysed by vigorously vortexing in a phosphate-buffered saline solution containing 0.1% Triton X-100 and 1 mM phenylmethylsulfonyl fluoride. An equal volume of cold 10% perchloric acid was added and the mixture incubated on ice for 15 min with occasional vortexing. Acid-insoluble material was pelleted by centrifugation and the remaining soluble proteins precipitated from the extract by adding 100% (w/v) trichloroacetic acid to 25% final concentration. The recombinant HMG-I protein was purified from this crude precipitate by cation exchange chromatography on a Macro-Prep 505 column (1 X 20 cm; Bio-Rad) employing a potassium salt gradient. Buffer A was 25 mM KH2P04, pH 7.0, and buffer B was buffer A with 1.0 M KC]. The column was initially equilibrated in 5% buffer B (50 mM KCI) and loaded with approximately 2 mg of recombinant HMG dissolved in the same buffer. The KC1 concentration was initially raised to 300 mM in 5 min (5% B/ min) followed by 300-550 mM in 70 min (0.36% B/min) at a flow rate of 1.0 ml/min. One-ml fractions were collected, and recombinant HMG-I protein was recovered by trichloroacetic acid precipitation as previously described.
In Vitro Phosphorylation of HMG-I Proteins by cdc2 Kinase-Mammalian cdc2 kinase was purified from washed chromatin of logphase Novikoff rat hepatoma cells and quantified as previously described (54,55). In vitro radiolabeling of pure proteins with phosphate was carried out with [Y-~'P]ATP (ICN) under the conditions for assay of growth-associated (cdc2) kinase activity previously described (54,55) at a substrate concentration of 0.2 mg/ml. Incubation temperature was 37 "C. Quantitative determinations of the extent and stoichiometry of in vitro phosphorylation of pure HMG-I proteins by cdc2 kinase were carried out at an enzyme concentration of 0.07 units/ml and substrate concentration of 0.5 mg/ml for 4 h at 37 "C. Determinations of the time course of the reactions showed that phosphorylation was complete at that time. Phosphorylated HMG-I was isolated from the reaction mixtures by RP-HPLC.
After binding, 25-rl reaction samples (-lo4 cpm) were loaded onto native, low ionic strength, 4% polyacrylamide gels (80:1, acrylamide:bisacrylamide) and protein-DNA complexes electrophoretically separated. Gels (16 cm X 16 cm X 1.5 mm) were electrophoresed at room temperature for about 3.5 h at 100 V with buffer recirculation. Gels were then dried and exposed to x-ray film for -15 h at room temperature without an intensifying screen.
Determination of Protein Binding Constants-Dissociation constants of HMG-I protein binding to DNA were determined using a fluorescence competition assay (42,67). Solutions containing 10 mM HEPES, pH 7.5, 50 mM NaC1, 1 mM EDTA, 3'-UTR DNA, and various concentrations of HMG-I protein were titrated with the fluorescent dye Hoechst 33258 (Sigma). DNA concentration was fixed in all experiments at 100 nM expressed as phosphate. Final concentrations of Hoechst 33258 ranged from 0 to 50 nM. Following the procedures of Loontiens et al. (73), stock solutions of Hoechst 33258 were prepared daily by dissolving the dye in 5 mM HC1. Stock solution concentration was determined spectrophotometrically following dilution into 10 mM HEPES pH 7.0 using c338 = 42,000 liters/mol cm (at pH 7.0). Working solutions of 1.5 p M were prepared volumetrically by dilution of the high concentration stock solution with 5 mM HC1. All glassware and quartz cuvettes were siliconized to prevent adsorption of the dye. Titrations were monitored using a Shimadzu RF-540 fluorescence spectrophotometer with XEX = 354 nm and XEM = 450 nm. Addition of reagents were made with microburettes, and the cuvette was not removed from the sample compartment during the course of a titration. All titrations were done at 25 "C. The change in fluorescence (AF) of the sample due to binding of the dye by DNA was expressed as the difference of the fluorescence of the test solution and a "blank" containing only buffer and dye.
As previously demonstrated (42,67), binding of Hoechst 33258 to DNA in the presence or absence of competitors may be analyzed To investigate the ionic strength dependence of recombinant human HMG-I binding to substrate DNA, aliquots of a concentrated NaCl solution were added to the usual assay mix containing 50 mM NaCl and 100 nM DNA, and titration with dye was performed as described. NaCl concentration was varied from 50 to 188 mM. In all cases, the concentration of HMG-I or phosphorylated HMG-I was fixed at 10 nM. The final assay volume (before addition of dye) was fixed at 3.00 ml.
Cell Culture, Synchronization, and in Vivo Labeling"NIH3T3 murine cells obtained from the American Type Collection were maintained as attached monolayers in Dulbecco's modified Eagle's medium plus 5% calf serum (complete medium). Non-proliferating and partially synchronous populations of proliferating cells derived by serum starvation methods were obtained as previously described (15). Nonproliferating or exponentially growing cells were labeled with inorganic 32P (Du Pont-New England Nuclear) for 4 h (50 pCi/ml) in the presence of phosphate-free Dulbecco's modified Eagle's medium supplemented with 5% calf serum (dialyzed) and 0.5 mM adenosine (Sigma). Mitotically blocked cells in partially synchronous populations were similarly labeled for 4 h (between 18 and 22 h post-seeding) in the presence of 0.4 pg/ml nocodazole (Sigma) and collected by the "mitotic detachment" method (57). Following 32P labeling, cells were harvested, washed once with phosphate-buffered saline solution, and resuspended in 500 p1 of 5% perchloric acid. Cells were lysed by three cycles of freezing and thawing, and lysates were cleared by centrifugation. Carrier protein (bovine serum albumin, 50 pg) was added to the supernatant, and proteins were extracted and purified as described above.
Tryptic Peptide Analysis of 32P-Labeled HMG-I Proteins-RP-HPLC-purified murine HMG-I proteins, phosphorylated either in vitro by cdc2 kinase or in vivo as a result of metabolic labeling with inorganic [32P]phosphate, were dissolved in 100 mM NH,HC03, pH 8.0, 0.1 mM CaC12, and trypsin was added at 1:lO (w/w) ratlo. Following overnight incubation at room temperature, the pH of the reaction mixture was adjusted to 2 with trifluoroacetic acid and peptides were resolved by C-18 RP-HPLC. Chromatographic fractions were collected and counted for radioactivity and the 32P-labeled peptides in the indicated fractions (see text) were collected and sequenced (Applied Biosystems 470A sequenator).

Cell Cycle-dependent Phosphorylation of Murine HMG-I on cdc2 Kinase
Sites-The results shown in Fig. 1 confirm and extend the earlier findings that both human (41) and murine (42) HMG-I proteins are in vitro substrates for efficient phosphorylation by mammalian cdc2 kinase and that the same in vitro sites of amino acid modification are also found to be phosphorylated in vivo in metaphase cells, by comparing the levels of in vivo phosphorylation of HMG-I in NIH3T3 cells in different parts of the cell cycle. The NIH3T3 cells were partially synchronized by serum starvation and subculture and then labeled with [32P]orthophosphate at various times after subculture/re-seeding (15). The 32P-labeled murine HMG-I proteins were isolated from: (i) 3-day confluent  Fig. 1B;  cf. Ref. 15). The in vivo labeled murine proteins were purified and extensively digested with trypsin and the resulting tryptic peptides separated by RP-HPLC and counted for radioactivity (Fig. 1).
Previously, three major in vivo phosphorylated HMG-I tryptic peptides were identified in metaphase mouse NIH3T3 cells, only one of which corresponded to the site of in vitro cdc2 kinase phosphorylation (42). In agreement with this observation, the phosphopeptide labeled 3 in Fig. 1 (EP-SEVP'TPK) is the site of cdc2 kinase phosphorylation both in vitro and in vivo, with the threonine being the phosphorylated residue as determined by peptide sequencing (42). This phosphorylation site corresponds to threonine residue 53 at the amino-terminal end of the principle DNA-binding domain peptide ('TPKRPRGRPKK) of the murine HMG-I protein (8,42). As is evident in Fig. 1, relative to the two other non-cdc2 kinase substrate peptides (1 and 2), the amount of in vivo phosphorylation on the cdc2 kinase substrate peptide ( 3 ) is extremely low in confluent cells (panel A; Go/GI cells ), somewhat higher in exponentially growing (i.e. random) cells

(panel B ) , and quite high in metaphase cells (panel C; >90%
M phase cells) collected by the mitotic detachment method. Perhaps the most revealing result, however, is shown in Fig.  lD, which shows the profile of tryptic phosphopeptides obtained from the HMG-I proteins isolated from the cells remaining attached to the plastic culture dish after the mitotic cells (shown in Fig. 1C) have been selectively removed. In these remaining "non-mitotic" cells (>go% in G2 and S phase), the amount of radioactivity in the cdc2 kinase substrate peptide is very low compared with the other two major phosphopeptides (1 and 2). These results clearly and unambiguously demonstrate that phosphorylation of threonine 53 at the amino-terminal hook end of the principle DNA-binding domain peptide of murine HMG-I protein is indeed cell cycledependent, as would be expected for an authentic site of in vivo phosphorylation by cdc2 kinase.

Production and Purification of Recombinant Human HMG-
I Protein-An unusual feature of HMG proteins, their solubility in dilute acids (l), facilitated the recovery and purification of bacterially produced recombinant HMG-I proteins. Fig. 2 shows a typical chromatographic profile of recombinant human HMG-I protein(s) extracted from E. coli with 5% perchloric acid. Since most bacterial proteins are insoluble in dilute acid, the chromatographic profile of such extracts (Fig.  2) is relatively simple with the major protein peaks corresponding to either full-length recombinant human HMG-I protein (peak 1, fractions 43-53) or a proteolytic degradation product of HMG-I (peak 2, fractions 62-70). The identity of recombinant HMG-I protein (peak 1 ) has been confirmed by both partial peptide sequence analyses (data not shown) and by specific DNA-binding assays (see below). By amino acid composition analysis, the peak labeled 2 (fractions 62-70) has been identified as a proteolytic cleavage product of the full- length HMG-I that is missing the carboxyl-terminal region of the intact protein (data not shown). Consistent with this identification of peak 2 is the fact that this smaller protein still retains the ability to specifically bind to DNA in an identical manner to the full-length recombinant HMG-I when assayed by gel electrophoretic mobility shift assay^.^ These findings indicate that this COOH-terminal truncated HMG-I protein still retains the A-T hook motifs necessary for specific DNA binding. Nevertheless, the remainder of the experiments described below were conducted using the fulllength (peak 1 ) recombinant human HMG-I protein.
Competition of Recombinant Human HMG-I Proteins with Hoechst 33258 for DNA Binding- and parentheses indicate residues present a t some but not all sites (58,59). In contrast to murine HMG-I, which contains a single such consensus sequence that is phosphorylated by cdc2 kinase in vitro (threonine 53) (42), the human HMG-I protein has two such sites, at threonine residues 53 and 78, both of which have been show to be in vitro phosphorylation sites for cdc2 kinase in HeLa cell proteins (41). Preliminary experiments indicated that purified cdc2 kinase could also quantitatively phosphorylate these same two threonine residues in recombinant human HMG-I protein in vitro, a finding confirmed by the results shown in Fig. 4. When the tryptic peptide fragments from unphosphorylated (Fig. 4A), or cdc2 kinase phosphorylated (Fig. 4B), recombinant HMG-I were separated by RP-HPLC the phosphorylated peptide fragments eluted several minutes earlier than did the corresponding unphosphorylated tryptic peptides. For example, whereas the unphosphorylated tryptic peptide(s) containing threonine 78 (Fig. 4A,peak a ) eluted at about 24-25 min, the corresponding phosphorylated peptide(s) eluted earlier a t about 21-22 min ( Fig. 4B; peak a ' ) . Similarly, phosphorylated peptide(s) containing threonine 53 (Fig. 4 A ; peak b ) eluted several minutes earlier than the unphosphorylatedpeptides ( Fig. 4B;peak  b'). The identity of the indicated peptide fragments was confirmed by amino acid and sequence analysis (data not shown). The doublet nature of the a and a' peptide fragments is the result of alternative cleavage by trypsin at repeated basic amino acid residues flanking the ends of the fragments. From these results we conclude that recombinant human HMG-I (rhHMG-I), like the natural protein, is a good in vitro substrate for cdc2 kinase and was stoichiometrically phosphorylated (2 mol of P/mol of rhHMG-I) under the conditions described.

Ionic Strength Dependence of HMG-Z Binding to DNA-
Ionic environment often plays a significant role, both in vitro and in vivo, in regulating specific protein-or peptide-DNA interactions (71, 72). Such ionic effects are predicted to be particularly important for the specific binding of the HMG-I protein to substrate (8). Each of the three DNA-binding domains within the protein contain five or more basic Arg/ Lys side chains, or laterally projecting cationic "bristles", distributed along the length of the planar binding domain peptide backbone. These cationic bristles are positioned in such a way that they are predicted to interact ionically with the anionic phosphate groups of the two antiparallel DNA strands that define the width of the narrow minor groove of A.T-rich DNA (8). Fig. 5 and Table I Table I, the affinities of binding of all of these ligands to DNA decrease with increasing ionic strength, in agreement with the findings of others for the binding of Hoechst 33258 to substrate DNA (73). This ionic strength-dependent reduction in Kd values is least for Hoechst dye but is quite marked for both the unphosphorylated and phosphorylated HMG-I protein (Table I). These observed differences in affinities due :de2 Kinase Alters DNA Binding to ionic effects are consistent with the fact that Hoechst dye is a monocationic ligand (with a single positive charge on piperazine methyl N) whereas the HMG-I proteins are polycationic ligands owing to the numerous Arg/Lys bristles present on the DNA-binding domains of the protein (cf. Refs. 71 and 72).
Of even greater interest, however, is the fact that with increasing ionic strength, the difference in Kd values between the unphosphorylated and cdc2 kinase phosphorylated HMG-I proteins becomes more accentuated (Table I) At 188 mM NaCl the difference is even greater with about a 20-fold reduction in binding affinity of the phosphorylated protein (Kd -320 nM) compared with the unphosphorylated protein ( K d -16.6 nM). We suggest that these marked differences in substrate binding affinities within the range of mammalian physiological salt concentrations reflect similar changes in DNA binding affinities of the HMG-I proteins i n uiuo as they are phosphorylated by cdc2 kinase in a cell cycledependent manner.

DISCUSSION
In the present report we demonstrate that murine HMG-I proteins are phosphorylated i n uiuo in a cell cycle-dependent manner on the same amino acid residues modified by purified cdc2 kinase i n vitro (Fig. 1). We also present evidence that recombinant human HMG-I protein specifically binds to A a T-rich substrate DNA in a manner indistinguishable from the native protein (Figs. 2 and 3). Additionally, we demonstrate that the recombinant protein is an efficient in uitro substrate for phosphorylation by purified cdc2 kinase (Fig. 4) and that this modification markedly decreases the affinity of binding of the protein to DNA in the normal physiological salt concentration range of mammalian cells ( Fig. 5 and Table I).
Moreno and Nurse have proposed (59) that the following criteria should be met in order to identify in vivo substrates of cdc2 kinase. (i) The purified protein kinase should efficiently phosphorylate the putative substrate i n vitro. (ii) The phosphorylation sites in uitro should be identical with the phosphorylation sites i n viuo, and phosphorylation of these sites should vary in the cell cycle in concert with changes in cdc2 kinase activity. (iii) It should be established that the observed phosphorylation changes elicit a biochemical change in the putative substrate that has appropriate biological consequences for mitosis. Although cdc2 kinase has been shown to phosphorylate a considerable number of proteins in uitro, few of these have been demonstrated to met these stringent criteria for i n vivo substrates (59).
Based on the data reported here, in combination with previous work, we suggest that the mammalian HMG-I proteins have now been demonstrated to fulfill the above suggested criteria as authentic in uivo cdc2 kinase substrates.
Both human (41) and murine (42) HMG-I proteins have previously been demonstrated to be efficient i n vitro substrates for purified cdc2 kinase. Furthermore, the same sites of in uitro phosphorylation of both the human (41) and mouse (42) have also been shown to be phosphorylated in vivo in metaphase blocked cells. Finally, previous work involving synthetic peptides corresponding to the consensus DNA-binding domain of mammalian HMG-I proteins has demonstrated that phosphorylation of these peptides by purified cdc2 kinase substantially reduced their affinity for binding substrate DNA in vitro (42). The present work considerably expands on these earlier findings by demonstrating the following. (i) During different parts of the cell cycle in partially synchronized murine NIH 3T3 cells, the level of in vivo phosphorylation of the amino acid residue previously identified as the primary site of modification by cdc2 kinase in vitro (threonine 53) fluctuates in a manner paralleling known variations of in vivo cdc2 kinase activity during the cycle (Fig. 1). (ii) Compared with unphosphorylated protein, stoichiometric phosphorylation of recombinant human HMG-I by cdc2 kinase significantly reduces (by 5-20-fold) the affinity of binding of the modified protein to substrate DNA when the ionic conditions of binding are within the physiological range (140-188 mM NaC1) of mammalian cells (Fig. 5 and Table I). It can be argued that these latter findings fulfill the requirement (59) that the observed cdc2 kinase phosphorylation changes elicit a biochemical change in the HMG-I substrate, namely significant reduction in DNA binding affinity, that has appropriate biological consequences for mitosis. Specific reductions in HMG-I binding affinity by cdc2 kinase phosphorylation could easily be envisioned participating in various cell cycle-regulated modulations in chromatin structure or perhaps affecting changes in gene regulation. We therefore conclude that the accumulated experimental evidence meets all of the aforementioned criteria (59), strongly suggesting that HMG-I is an authentic in vivo substrate for cdc2 kinase.
Phosphorylation of recognition sites in other DNA-binding cdc2 kinase substrates may also diminish their DNA binding affinity. Sea urchin histones H1 and H2B contain multiple repeats of a consensus cdc2 kinase site (60), which have been shown to be specific binding sites for the minor groove of A.
T-rich DNA (67)(68)(69). These sperm histone sites are also phosphorylated in vivo (61, 62), and such modifications have been suggested to affect the DNA-binding characteristics of the amino-terminal ends of these proteins (61, 62, 67-69). Mammalian somatic histone H1 (58), as well as trout testis histone H1 and chicken erythrocyte histone H5, also contain sequences of this type, which are known to be phosphorylated in vivo (63,64). In several cases phosphorylation of H1 histones has been shown to affect the compactness of chromatin or histone H1-DNA complexes (61,65,66). Thus alterations in the affinity of DNA-binding proteins by site-specific phosphorylations may be an important component of the mechanism by which cdc2 kinase regulates cell cycle progression.