Dynamics of protamine 1 binding to single DNA molecules.

Protamine molecules bind to and condense DNA in the sperm of most vertebrates, packaging the sperm genome in an inactive state until it can be reactivated following fertilization. By using methods that enable the analysis of protamine binding to individual DNA molecules, we have monitored the kinetics of DNA condensation and decondensation by protamine 1 (P1) and synthetic peptides corresponding to specific segments of the bull P1 DNA binding domain. Our results show that the number of clustered arginine residues present in the DNA binding domain is the most important factor affecting the condensation and stability of the DNA-protamine complex prior to the formation of inter-protamine disulfide cross-links. The high affinity of P1 for DNA is achieved by the coordinated binding of three anchoring domains, which together in bull P1 contain 19 Arg residues. The single DNA molecule experiments show that sequences containing two or more anchoring domains have an off-rate that is at least 3 orders of magnitude slower than those containing a single domain. The use of Arg, rather than Lys residues, and the inclusion of Tyr or Phe residues in the hinge regions between anchoring domains provide additional stability to the complex.

Protamine molecules bind to and condense DNA in the sperm of most vertebrates, packaging the sperm genome in an inactive state until it can be reactivated following fertilization. By using methods that enable the analysis of protamine binding to individual DNA molecules, we have monitored the kinetics of DNA condensation and decondensation by protamine 1 (P1) and synthetic peptides corresponding to specific segments of the bull P1 DNA binding domain. Our results show that the number of clustered arginine residues present in the DNA binding domain is the most important factor affecting the condensation and stability of the DNA-protamine complex prior to the formation of inter-protamine disulfide cross-links. The high affinity of P1 for DNA is achieved by the coordinated binding of three anchoring domains, which together in bull P1 contain 19 Arg residues. The single DNA molecule experiments show that sequences containing two or more anchoring domains have an offrate that is at least 3 orders of magnitude slower than those containing a single domain. The use of Arg, rather than Lys residues, and the inclusion of Tyr or Phe residues in the hinge regions between anchoring domains provide additional stability to the complex.
Several different mechanisms have been identified for packaging DNA in eucaryotic cells. The best understood is the packaging scheme used to organize DNA inside the nucleus of all somatic cells, in which the histones coil the DNA into 11-nm nucleosomal subunits that are further organized into 30-nm chromatin fibers. This type of packaging allows individual genes to be activated or repressed depending on the needs of the cell. In contrast, small basic nuclear proteins called protamines bind to the entire length of the haploid genome in sperm cells and coil 50-kb segments of DNA into ϳ60-nm toroidal subunits (1)(2)(3)(4). Protamine binding maintains the DNA in this highly compact and transcriptionally inactive state for extended periods of time (5)(6)(7)(8) until the sperm fertilizes an oocyte and its genome is reactivated.
Although a variety of protamine-like proteins have been isolated from different vertebrate sperm, only two distinct types of protamines, protamine 1 (P1) and protamine 2 (P2), package DNA in mammalian sperm. P1 has been found to be present in the sperm of every species examined (9,10) and is the most abundant protein present in the sperm of many mam-mals (10,11). Both P1 and P2 package DNA in rodent and primate sperm.
Sequence analyses of numerous P1 molecules and their genes have revealed that all mammalian P1 molecules appear to have several sequence-related features in common. The Arg residues that comprise the DNA binding domain of P1 are usually clustered in the center of the molecule (1, 10) in a series of DNA "anchoring domains." These anchoring domains, which typically contain 3-7 Arg residues (1), are usually separated by one or more uncharged amino acids. The interspersed amino acids linking the anchoring domains appear to function as a hinge, enabling the anchoring domains to wrap around the major groove of the DNA helix (12)(13)(14) and optimize the interactions that form between each Arg residue in the DNA binding domain of P1 and the negatively charged phosphates in the phosphodiester backbone of the two DNA strands (14). One P1 molecule appears to be bound to each turn of DNA (15,16) such that the charge associated with each phosphate group in both strands of DNA is neutralized by an Arg residue of the bound P1. Ser and Thr residues located in the amino-and carboxylterminal peptide segments flanking the DNA binding domain have been shown to be phosphorylation sites that are believed to be modified prior to the deposition of P1 on DNA so that it binds correctly (17)(18)(19).
The protamines of eutherian mammals also contain numerous cys residues that are used to form intra- (20) and interprotamine disulfide cross-links (10) during the final stages of sperm chromatin maturation. Intramolecular disulfide crosslinks begin forming in late-step spermatids, shortly after the protamines bind to DNA. As the spermatids leave the testis and begin moving through the epididymis, a final set of disulfide bonds is formed that covalently link each protamine to its neighbor. This creates a chromatin complex that is highly compact and transcriptionally inert.
Previous studies using salmon protamine (salmine) (21), bull and boar P1 (8,22), chicken protamine (23), and various oligoarginine or oligolysine peptides (24 -28) have shown that the large number of basic amino acids present in the DNA binding domain of protamine enables it to condense DNA. Although the exact structure of the complex has not yet been determined, Raman spectroscopy and other studies have suggested that P1 must bind to DNA in an extended conformation with a footprint that covers ϳ11 bp of sequence (14,16,29).
In an effort to gain additional insight into how the protamines interact with DNA and inactivate the genes of the developing spermatid, we have examined the functions of specific subsets of the P1 DNA binding domain that enable protamine to bind to DNA and condense it into toroids. These experiments extend previous studies that have been conducted by others using various Arg-and Lys-containing oligopeptides (24 -28) and focus on elucidating how three specific features of the P1 sequence affect the ability of P1 to bind to DNA and maintain it in a condensed state as follows: 1) the number of poly(Arg) clusters or anchoring domains, 2) the relative importance of Arg and Lys residues in these domains, and 3) the presence of aromatic amino acids interspersed between anchoring domains. Peptides corresponding to subdomains of the bull P1 sequence were synthesized, and kinetics of their binding to individual phage DNA molecules were examined under conditions that preclude intermolecular aggregation (21).

EXPERIMENTAL PROCEDURES
Peptide Synthesis and Purification-Bull P1 was isolated from pooled bull semen (American Breeders Service, DeForest, WI) and purified using high performance liquid chromatography as described previously (11). The peptide sequences synthesized for this study used bull P1 as the primary reference sequence, ARYRCCLTHSGSRCRRR-RRRRCRRRRRRFGRRRRRRVCCRRYTVIRCTRQ. The bull P1 sequence shares a strong consensus with the sequences for P1 molecules found in other mammals (10).
Peptides corresponding to various subsets of the bull P1 sequence were synthesized by SynPep Corp. (Dublin, CA) and were purified by high performance liquid chromatography. Electrospray mass spectrometry was used to confirm their sequence and purity. Peptides located in the DNA binding domain of P1 were synthesized with an acetylated amino terminus and an amide on the carboxyl terminus to minimize terminal charge effects and mimic the internal peptide sequences present in P1. The amino-terminal protamine peptide ARYRSSLTHSGSRS was synthesized with a carboxyl-terminal amide group, and the carboxyl-terminal protamine peptide VSSRRYTVIRSTRQ was synthesized with an amino-terminal acetyl group.
Protein Concentration Verification-To obtain accurate concentrations of the nuclear proteins used in the condensation experiments, the purified peptides and bull P1 were dissolved in 0.1% trifluoroacetic acid, known aliquots were distributed into glass tubes, and the samples were frozen and lyophilized. Replicate samples were dissolved in 6 N hydrochloric acid and hydrolyzed at 105°C for 16 h, and the amino acid content of each tube was determined by AAA Service Laboratory (Boring, OR). The protein content of each sample was determined using the known sequence of bull P1 and the synthetic peptides. Tube to tube variations in the aliquoted samples used for the experimental studies were found to be less than 2%.
Preparation of Stained DNA Bead Sample-phage DNA (Sigma) was dissolved in buffer (degassed 50% sucrose (w/v), 100 mM sodium bicarbonate, pH 8, 30 mM dithiothreitol), tagged with biotin, and attached to 1-m streptavidin-coated polystyrene spheres as described previously (21). YOYO-1 dye was then added so that the dye molecule to DNA base pair ratio was 1:4. Experiments performed with salmine and various peptides have shown that this concentration of YOYO-1 does not alter the kinetics of binding to DNA (21). The same buffer, minus the YOYO-1 dye, was used to dissolve the protein samples. In experiments involving the binding and dissociation of bull P1 to DNA, dithiothreitol was included to prevent the formation of disulfide bonds between P1 molecules following their binding to DNA.
Condensation/Decondensation Experiments-The experimental apparatus is identical to that used previously to examine the condensation of individual DNA molecules by salmon protamine (21). Separate solutions containing protein and DNA molecules attached to polystyrene beads were introduced into a dual port flow cell at a flow velocity of ϳ50 m/s where they flowed side by side, in a laminar fashion, with little mixing. Beads containing individual DNA molecules were trapped using an infrared laser optical trap ( ϭ 1.047 m). An argon ion laser ( ϭ 488 nm) was used to excite the YOYO-1-stained DNA molecules, and their fluorescence was visualized using an image-intensified CCD camera.
Working at a depth of about 20 m below the coverslip (the flow cell depth was 40 m), the DNA molecule was pulled across the interface separating the two flowing buffer solutions, and the shortening of individual DNA molecules was monitored as the protein or peptide bound to it and induced condensation. The condensed DNA was evident as a bright fluorescent spot at the end of the DNA molecule. The length of the stained phage DNA molecule (48.5 kb) extended by flow was measured to be 19.2 Ϯ 0.82 m at a flow rate of 72 m/s in 50% sucrose. Under these conditions, the DNA was 95% extended. 1 Analyses of Protein and Peptide Binding Kinetics-The condensation rates for 5-20 individual DNA molecules were measured at several different protein or peptide concentrations, and a linear least squares fit (in conjunction with the peptide footprint) was used to determine the binding, or "on-rate", k on , of the peptide or protein to DNA. The binding of peptides and proteins to DNA was shown previously to be the ratelimiting step in condensation, and the length of the DNA molecule during condensation was shown to decrease linearly with time (21, 30) ( Fig. 1, A and B). The decondensation rate was determined by pulling partially condensed DNA molecules back to the DNA side of the flow cell (where there was an absence of peptide or protein) and observing the DNA re-extend as the bound peptide left the DNA. The mean of ϳ10 molecules was used to determine the "off-rate," k off , of each peptide or protein dissociating from the DNA molecule. The dissociation constant K d was determined by dividing the off-rate by the on-rate, K d ϭ k off /k on .
Computational Analysis of Peptide Solution Conformation Using Molecular Dynamics-Molecular dynamics simulations were performed with AMBER (31) using the Cornell force field (32). Structures of the R 6 GGR 6 and R 6 FGR 6 peptides were constructed in extended conformations and solvated in boxes of TIP3P water (33) sufficient in size to have at least 15 Å of water between the protein and the solvent interface (81 ϫ 72 ϫ 45 Å 3 initially). Twelve chloride ions were added to each system to neutralize the charge. The systems typically consisted of 21,400 atoms (ϳ7,000 water molecules). Each system was energy-minimized using 250 steps of steepest descents and 750 steps of conjugate gradients. Constant temperature and pressure dynamics were performed on these energy-minimized systems. Periodic boundary conditions were used, and electrostatic interactions were treated by Particle Mesh Ewald methods (34) with a 9-Å cut-off in direct space, cubic interpolation, and using a 1-Å grid. SHAKE (35) was used to constrain bonds containing a hydrogen, and a time step of 2 fs was used. These systems were coupled to a heat bath at 300 K. Coupling constants of 5 and 2 ps were used for peptide and solvent, respectively. Each simulation was performed for a total of 2.5 ns, and the last 2.0 ns were used for analysis. Coordinates were saved every 200 time steps. The peptide conformers were clustered using the program NMRCORE (36). Solventaccessible surface areas were calculated using the program surfv (37). CLTHSGSRCRRRRRRRCRRRRRRFGRRRRRRVCCRRYTVI-RCTRQ) were determined for a range of protein and peptide concentrations ( Table I). The results show that the rate of DNA condensation is affected by the number of Arg residues in the peptide or the protein's DNA binding domain, the presence of an aromatic residue in the hinge region between anchoring domains, and the substitution of Lys for Arg residues. Arg 6 was the least effective of the Arg-containing peptides in condensing DNA. While R 6 GGR 6 condensed the DNA molecules 42 times faster than Arg 6 , increasing the number of Arg 6 anchoring domains from two to three, as in bull P1 and R 6 GGR 6 GGR 6 , did not increase the rate of condensation. Bull P1 and R 6 GGR 6 condensed DNA with equal rates, while the condensation rate for R 6 GGR 6 GGR 6 was surprisingly smaller. Substituting an aromatic residue for one of the Gly residues in R 6 GGR 6 resulted in condensation rates that were a factor of 10 -25 smaller (R 6 YGR 6 and R 6 FGR 6 ). The substitution of Lys residues for Arg also altered the rate of condensation. Lys 6 was observed to be 10-fold less effective in condensing DNA than Arg 6 , and K 6 GGK 6 was nearly 10-fold less effective than R 6 GGR 6 . Substitutions of individual Arg residues with Lys also decreased the condensation rate (Table I).

Rate of DNA Condensation by P1 and
Peptides corresponding to the amino-terminal (ARYRSS-LTHSGSRS-CONH 2 ) and carboxyl-terminal (CH 3 CO-VSSR-RYTVIRSTRQ) sequences of P1 outside the DNA binding domain were tested for their ability to condense DNA at concentrations similar to those used for the peptides containing anchoring domain sequences. The Ser residues were substituted for Cys to eliminate the possibility that the peptides might form disulfide-linked folded forms (38) or multimers in solution and complicate our interpretation of the results. No DNA condensation was observed for either peptide at a concentration of 2 M. While these results suggest that the amino-and carboxyl-terminal sequences flanking the protamine DNA binding domain do not bind to DNA, we cannot exclude the possibility that the cysteine residues normally present in these sequences may stabilize the formation of some disulfide-dependent conformation in the peptide that enables DNA binding.
Protein and Peptide Off-rates Determined by Decondensation-The DNA decondensation rates and dissociation constants (K d ) measured for bull P1, and the synthetic peptides are shown in Table II. Bull P1 remained attached to DNA for the longest period of time with a dissociation rate (k off ϭ 0.23 molecules/s) that was similar to previous rates obtained for salmine and hamster P1 (8,21,30) The peptide sequences R 6 GGR 6 GGR 6 , R 6 GGR 6 , and RKRKRKGGRKRKRK dissociated from DNA at faster rates (k off ϭ 6 -7 molecules/s), while Arg 6 and Lys 6 dissociated from DNA much more rapidly (k off ϭ 17,000 and 9,000 molecules/s, respectively). K 6 GGK 6 dissociated from DNA at a significantly faster rate (k off ϭ 25 molecules/s) than R 6 GGR 6 .
The inclusion of an aromatic amino acid between two Arg 6 domains also appeared to increase the stability of the peptide-DNA complex. Sequences containing Tyr had substantially slower off rates than those containing only Gly. The inclusion of a single Tyr residue decreased the off-rate of the peptide by 6-fold. Peptides containing Phe exhibited a similar, but less pronounced effect. Replacing the first Gly in R 6 GGR 6 with Phe decreased the off-rate by 2-fold.
Computational Simulation of the Protamine Anchoring Domain by Using Molecular Dynamics-Molecular dynamics runs were performed on the R 6 GGR 6 and R 6 FGR 6 peptides to examine how the presence of the aromatic Phe residue located between two anchoring domains might affect the conformation adopted by the adjacent peptide segments and ultimately impact the binding of these sequences to DNA. Structures of each peptide were constructed in extended conformation and solvated in boxes of water utilizing periodic boundaries. Molecular dynamics simulations were run for a total time of 2.5 ns using a fixed number of particles held at constant pressure and temperature.
Clustering analysis performed on 200 structures selected at 10-ps intervals during the last 2.0 ns of the simulations are shown in Fig. 2, A and B. Comparisons of these structures suggest that both peptides form a significant amount of structure along the backbone of the entire sequence. In the simulation performed with the R 6 GGR 6 sequence, this backbone remains reasonably extended as a result of the extension and solvation of the Arg side chains. The mobility of the carboxylterminal region of this peptide is constrained by the formation of a salt bridge between the terminal carboxyl group and the side chain of residue Arg 11 . The most interesting structure, however, is that developed in the R 6 FGR 6 peptide during the last 1.1 ns of the simulation. In this peptide, the phenyl ring stacks against the peptide backbone and appears to form a stable core that is surprisingly tight (Fig. 2B). In these structures, the aromatic ring of Phe interacts with the backbone amides of Gly 8 and Arg 9 . One consequence of this stacking is that the backbone is bent between the two anchoring (Arg 6 ) domains.
The distances measured from the amide hydrogen to the center of mass of the Phe ring during the last nanosecond of the simulation are 3.88 Ϯ 0.63 and 3.52 Ϯ 0.42 Å for Gly 8 and Arg 9 , respectively. One effect of the interaction between the aromatic ring and backbone is that the amount of motion sampled by the ring is reduced significantly. The positional fluctuation of the Phe side chain is 4.54 Å for the first nanosecond of dynamics but reduces to 2.72 Å when stacked against the peptide backbone. The average solvent-accessible surface area of R 6 FGR 6 (2766.5 Ϯ 87.4 Å 2 ) is lower than for R 6 GGR 6 (2850.4 Ϯ 79.4 Å 2 ) even though R 6 FGR 6 has the greater intrinsic surface area.

DISCUSSION
Importance of Multiple Arg 6 DNA Anchoring Domains-DNA condensation and decondensation experiments were performed using a series of synthetic peptides containing one (Arg 6 ), two (R 6 GGR 6 ), and three (R 6 GG R 6 GGR 6 ) Arg 6 anchoring sequences as model subsets of the bull P1 DNA binding domain and two 14-residue peptides corresponding to the amino-terminal and carboxyl-terminal sequences flanking the bull P1 DNA binding domain. Even though the sequences of the model peptides do not correspond to the exact sequences present in bull P1 (the Cys residues were changed to Ser or Gly to prevent disulfide bond formation and model anchoring domains containing six Arg residues separated by two Gly residues were used to build up a peptide sequence corresponding to a "representative" P1 DNA binding domain), these experiments clearly showed that each of the peptides containing one or more Arg 6 anchoring sequences bound to and condensed DNA. In contrast, neither the synthetic amino-terminal nor carboxyl-terminal P1 peptide sequences condensed DNA.
The number of Arg residues present in the anchoring domain peptides had a significant effect on both the concentration of the peptide required to condense DNA and the stability of the complexes. While even a single Arg 6 anchoring domain condensed DNA, condensation could only be achieved at peptide concentrations that were 44-fold higher than those used for bull P1. Rates of Arg 6 binding to DNA were observed to be 12-fold lower than the rate of bull P1 binding at equivalent concentrations. The ineffectiveness of Arg 6 in condensing DNA was also seen in the instability of the Arg 6 -DNA complex. By increasing the number of Arg 6 anchoring domains present in the peptide sequence from one to two (R 6 GGR 6 ), the rate of dissociation of the peptide from the complex was reduced by 2500-fold.
Surprisingly, increasing the number of anchoring domains in the synthetic peptides from two to three did not increase either the effectiveness of the peptide in condensing DNA or the stability of the complex. R 6 GGR 6 bound to and condensed DNA at rates similar to those observed for bull P1, but this rate was 2.7 times faster than that achieved with R 6 GGR 6 GGR 6 . In addition, the stability of the R 6 GGR 6 and R 6 GGR 6 GGR 6 -DNA complexes, as reflected by the rates of dissociation of the peptides, was identical. These results show that the rate of binding, and the stability of the resulting complex, are not strictly determined by the number of Arg residues in the P1 DNA binding domain but suggest that other factors must be involved. One factor that may contribute to the rate of binding is the increase in flexibility inherent in longer peptides. Once the length of the binding domain increases beyond two anchoring domains, the increased flexibility of the sequence could adversely influence the rate of binding and condensation.
Formation and Relative Stability of Complexes Formed with Lys-containing Anchoring Domains-Previous studies (26,39) have shown that the Arg residues in peptides and proteins bind more tightly to DNA than Lys residues because the guanidinium group in Arg forms both a hydrogen bond and salt bridge to the phosphate groups that comprise the backbone of DNA. Several different groups have examined and compared the ability of peptides containing sequences with multiple Lys or Arg residues to bind and condense DNA (26 -28, 39, 40). The majority of this work focused on small (dimer to tetramer) or high molecular weight Arg or Lys polymers.

FIG. 2.
A, the superposition of the backbone atoms of R 6 GGR 6 from molecular dynamics calculations. Only the N, CA, and C atoms of the peptide are shown. A total of 200 structures, taken from 10-ps intervals, are shown. B, overlay of the backbone atoms for R 6 FGR 6 from the last 1.16 ns of molecular dynamics calculations (116 structures). Only the N, CA, and C atoms of the peptide are shown for clarity. C, the last of the 116 structures shown in B. The Phe and backbone atoms of Gly 8 and Arg 9 are drawn as Corey-Pauling-Koltun space-filling models (49). The arrows in B and C point to the aromatic ring of Phe 7 that stacks against the backbone amides of Gly 8 and Arg 9 . and KXXK peptides containing aromatic amino acids also provided evidence that aromatic amino acids located between two Lys residues did exhibit a detectable but small effect on the base sequence preference for peptides binding to DNA (24,39,41). These previous studies also provided evidence that the insertion of non-basic amino acids between Lys or Arg residues in short homopolymers eliminated the base sequence specificity that has been observed for poly(Lys) and poly(Arg) binding to duplex DNA (24).
In the present study, experiments conducted with one or two Lys 6 anchoring domains showed that they were remarkably ineffective in condensing DNA compared with Arg-containing peptides. At least a 10-fold higher concentration of the Lyscontaining peptides was required to achieve a condensation rate similar to that of Arg-containing peptides. Once bound, however, Lys was as effective as Arg in maintaining DNA in a condensed state. To within experimental error, peptides of the same length, composed of Lys and/or Arg residues, containing GG in the hinge region between domains, and corresponding to one (Arg 6 and Lys 6 ) or two (R 6 GGR 6 , K 6 GGK 6 , and RK-RKRKGGRKRKRK) anchoring domains, had the same off-rate. The DNA condensation and binding rate of K 6 GGK 6 and RKRKRKGGRKRKRK were essentially identical. This suggests that the dominant factor contributing to DNA binding in sequences containing at least two anchoring domains is the number of positively charged side chains available for binding to the phosphodiester backbone of DNA.
Contribution of Aromatic Amino Acids to DNA Binding and Complex Stability-Mammalian and other vertebrate protamines frequently contain Tyr or Phe residues located within the DNA binding domain (11,23,42). Bull P1, used as the model P1 sequence for these experiments, contains a single Phe residue positioned between two Arg 6 anchoring domains. Decondensation experiments conducted with the bull P1 sequence R 6 FGR 6 , and its analogs, R 6 GGR 6 and R 6 YGR 6 , showed that the presence of an aromatic amino acid increased the stability of the DNA-peptide complex by decreasing the off-rate of the peptides. R 6 YGR 6 formed the most stable complex (k off ϭ 1.1 molecules/s), followed by R 6 FGR 6 (k off ϭ 3.3 molecules/s) and R 6 GGR 6 (k off ϭ 6.7 molecules/s), respectively. These results and the difference in off-rates observed for bull P1 and R 6 GGR 6 GGR 6 indicate that these aromatic amino acids probably do contribute to the overall stability of the DNA-protamine complex.
As has been observed for other DNA-binding proteins (1,(43)(44)(45)(46), the aromatic ring of Tyr or Phe may intercalate between base pairs or bind in the groove through hydrophobic interactions or hydrogen bonding. Numerous sites are available inside the major groove for hydrogen bonding to the Tyr hydroxyl group, including the amino groups of cytosine and adenine, the N-7 groups of guanine and adenine, and the carbonyl oxygen of thymine or guanine. Although the current experiments do not provide enough information to enable us to discriminate between these possibilities, preliminary NMR studies of a complex formed between a hairpin DNA and a slightly different but related P1 sequence containing Phe have indicated that the phenyl ring is intercalated (1). The insertion of this ring between base pairs would force the DNA molecule to bend. Bends in the DNA-protamine complex would be expected to facilitate, and may even initiate, toroid formation.
The 10 -30-fold slower on-rate we observed for R 6 XGR 6 sequences containing an aromatic amino acid and the molecular dynamics simulations of the conformation adopted by the peptides in solution are both consistent with binding that involves an intercalation event. The results of the molecular dynamics simulation of the conformation of the R 6 FGR 6 peptide suggest that the aromatic side chain may stack against the backbone of two adjacent amino acids (Gly 8 and Arg 9 ), lowering the hydration state of the peptide and kinking the backbone slightly. Aromatic ring-backbone interactions are not uncommon and have been shown to occur in other proteins (47,48). In contrast, the equivalent peptide containing only Gly residues in the hinge region (R 6 GGR 6 ) remains extended throughout the simulation. These results suggest that anchoring domain sequences separated by an aromatic amino acid should take longer to bind to DNA in a manner that optimizes the interactions of the Arg residues in the anchoring domains with the phosphodiester backbone of the DNA. Once the first anchoring domain binds to DNA, additional time would be required for the aromatic side chain to unstack with the protein backbone and insert into the helix before the second anchoring domain could bind properly to the phosphodiester backbone of DNA.
The stacking of the aromatic ring of Phe or Tyr against the peptide backbone might also be expected to reduce, somewhat, the flexibility of larger sequences, such as bull P1, that contain a third anchoring domain. This could explain the differences we observed for the on-rates of bull P1 and the model DNA binding domain R 6 GGR 6 GGR 6 . Whereas the presence of an aromatic amino acid between two Arg 6 anchoring domains reduced the rates of peptide binding and DNA condensation (R 6 GGR 6 versus R 6 FGR 6 and R 6 YGR 6 ) 10-30-fold, its presence is a sequence containing three anchoring domains (bull P1) increased the rate of peptide binding to DNA (P1 versus R 6 GGR 6 GGR 6 ) by 2-3-fold.
Information About the Site of Protein/Peptide Binding Derived from Kinetics-Analyses of the binding kinetics of salmine obtained in a previous study (21) and bull P1 and Arg 6 in this study (Fig. 1, A and B) have revealed that the length of an individual DNA molecule undergoing condensation decreases linearly with time. In each study we have observed that the binding of these proteins or peptides is the rate-limiting step in the condensation process. It follows that the rate of binding of these molecules to DNA is constant and time-independent. The first order equation describing the protein on-rate is given by Equation  One interpretation of these results is that the fixed binding site (or sites) is located at the point where the DNA bends and is condensed into the toroid, which starts in our experimental system at the free end of the extended DNA molecule. If this interpretation of the data is correct, the constant rate of condensation suggests that the protein/peptide can only bind to the DNA if the DNA is bent to a dimension on the order of its persistence length (toroids are 50 nm (3)). Whereas this hypothesis is intriguing, the best way to verify this theory is by designing experiments that enable the direct visualization of the site of protein binding.
Relationship between DNA Binding Domain Sequence and Protamine Function-The results obtained from these single molecule DNA condensation studies have provided new insight into the features of the sequence of the P1 DNA binding domain that facilitate its function. As these experiments have demonstrated, relatively low concentrations of the protein are required to condense DNA and maintain it in a condensed state for extended periods of time if the DNA binding domain contains multiple Arg 6 anchoring domains. This is particularly important in mammals because the maturation process for the developing sperm can take as long as 2-3 weeks, and the disulfide bonds that cross-link the DNA-bound protamines together (so they cannot dissociate or be displaced from DNA) are not formed until very late in the sperm maturation process (during epididymal transit).
These extremely low off-rates also suggest that large concentrations of unbound protein are not required to maintain the entire genome in an inactive, condensed state. The low off-rate of the protamine molecule is particularly important in those non-mammalian vertebrates that use protamines to package their sperm DNA. Because the protamines in these species do not contain Cys, inter-protamine disulfide bonding cannot be used to stabilize the final structure of the mature sperm chromatin complex.
The results of these studies also appear to provide a plausible "rationale" for why protamines tend to be arginine-rich rather than lysine-rich DNA-binding proteins. Because DNA is condensed most effectively by protamines with DNA binding domains that are composed almost exclusively of Arg, selective pressures applied by the sperm during the evolution of the sequence of the protamine DNA binding domain should select against the use of Lys residues as replacements for Arg. This appears to be generally the case. Occasional Lys residues can be found in the DNA binding domains of several mammalian protamines, but they are never clustered together as the only components of an entire anchoring domain in P1.