Solubility-dependent structural formation of a 25-residue, natively unfolded protein, induced by addition of a seven-residue peptide fragment Mitsugu Araki and Atsuo Tamura Graduate School of Science, Kobe University, Nada, Japan In order to elucidate the architectural principle of pro- tein structure, the relationship between protein sequence and tertiary structure has been studied from various perspectives. Factors determining the mecha- nism of structural stability have long been explored, mostly by studying the stability and folding kinetics of natural and mutated proteins [1–3]. Recently, compu- tational protein designs have become advanced and provide new insights into the factors determining pro- tein structure, stability and folding [4], i.e. the redesign of naturally occurring proteins [5–8] and the de novo design of novel structures [9,10]. These studies suggest that each well-packed structure is stabilized by a num- ber of intra- or intermolecular interactions, invoked by the appropriate alignment of amino acid residues, and the number of proteins with well-packed structures seems extremely small compared with all possible sequences (primary sequence space). In such cases, how have the existing proteins attained their well- packed structures? If the majority of the possible Keywords folding; intrinsically unstructured protein; protein stability; self-association; solubility Correspondence A. Tamura, Graduate School of Science, Kobe University, Nada, Kobe 657-8501, Japan Fax ⁄ Tel: +81 78 803 5692 E-mail: tamuatsu@kobe-u.ac.jp Database The atomic coordinates have been deposited in the Protein Data Bank (PDB ID code 2KFQ for FP1) (Received 26 September 2008, revised 10 February 2009, accepted 12 February 2009) doi:10.1111/j.1742-4658.2009.06961.x To elucidate the architectural principle of protein structure, we focused on sequestration from solvent, which is a common characteristic of folding and self-associative precipitation. Because protein solubility can be regarded as a basis for the potential ability to sequester from solvent, we assume that poorly soluble proteins tend not only to precipitate, but also to form solution structures. To examine this, the solubility of a 25-residue, natively unfolded protein, modified from a zinc-finger domain of transcrip- tion factor Sp1, was disturbed by adding a seven-residue hydrophobic pep- tide fragment to the C-terminus. NMR and ultracentrifuge measurements of the resulting sequence showed that a dissolved species forms an a-helical structure in a 15–20 molecule oligomer. To elucidate the mechanism by which the structure forms, we prepared two variants in which the added fragments are less hydrophobic; the structural stabilities were then mea- sured at various pH values. A fairly good correlation was observed between stability and hydration potential, whereas a much stronger correla- tion was observed between stability and solubility, indicating that the stability is more strongly dependent on the ability to precipitate than on dehydration. These results show that, among poorly soluble protein mole- cules, dissolved species can be transformed from the solvent-exposed unfolded state into a loosely packed structure via intermolecular inter- actions. Because decreasing the protein solubility does not require the primary sequence to have a sophisticated design, such a protein structure might form readily and frequently, compared with the well-packed structure found in native proteins. Abbreviations FP, final protein; IP, initial protein; DG dissol , dissolution free energy; DG f , folding free energy; DG hyd , hydration potential. 2336 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS sequences result in the unfolded state, the probability of natural proteins exploring and evolving well-packed structures would be extremely limited. It is thus assumed that moderately structured states needed for decent biological function might arise frequently. This assumption is supported by protein-folding studies, which shed light on the early evolution of natural pro- teins. First, in 74-residue library based on a binary patterning of polar or nonpolar residues, most proteins formed fluctuating structures reminiscent of molten globule intermediates [11]. Second, using the lattice model, folding simulations implied that a randomly chosen sequence of amino acids frequently encodes a globular conformation [12]. This simulation is based on the concept that hydrophobicity is a driving force in protein folding, in which a protein excludes part of the molecule from the solvent water in a geometry- specific manner [13]. In this study, we attempted to identify a crucial fac- tor in the formation of a moderately structured state by focusing on the fact that sequestering from a solvent is a common characteristic of folding and self- associative precipitation or aggregation, the latter frequently occurring during the handling of natural and artificial proteins. Because the tendency of a pro- tein molecule to precipitate is represented by its solu- bility, this physicochemical property can be regarded as a criterion for the potential ability to squeeze out solvent. In the case of a poorly soluble protein, which disfavors exposure to solvent, it is likely that the dis- solved species tend to form solution structures seques- tered from the solvent. In order to confirm this, we attempted to transform a 25-residue, soluble unfolded protein (the initial protein; IP) into a structured pro- tein (the final protein; FP) by altering the solubility. First, the amino acid sequence of the IP was deter- mined on the basis of a zinc-finger domain of tran- scription factor Sp1, which folds into a well-defined structure consisting of a b hairpin and an a helix upon binding to Zn 2+ ; it is unfolded in the absence of the metal [14]. Next, the solubility of the IP was decreased by the addition of seven hydrophobic amino acids to the C-terminus; we anticipated that the added frag- ment would induce long-range interactions between amino acids separated in the primary sequence [15]. As a result, NMR, CD and ultracentrifuge measurements showed that a dissolved species of the resulting sequence, FP1, takes the form of an a-helical structure in a 15–20 molecule oligomer, without addition of Zn 2+ . We thus scrutinized the dependence of protein solubility or hydration potential on the structural stability using two variants of FP1 by varying the pH. A strong correlation between the stability and solubility elucidated mechanism of formation of the loosely packed structure, induced by the association with other copies of the same chain. Results Sequence and structural property of IP Among several candidates for IP, which needs to be unstructured in the native condition, we chose the third zinc-finger domain Sp1f3 of transcription factor Sp1, with two histidines (His21, His25) and two cyste- ines (Cys5, Cys8), to bind coordinately to Zn 2+ [14]. Sp1f3 is known to fold into a well-defined structure upon binding to Zn 2+ , and is unfolded in the absence of the metal. To suppress the excessively high solubil- ity of Sp1f3, which is caused by the high frequency of ionizable amino acid residues (Table 1), residues 26–29 (Gln-Asn-Lys-Lys) in the C-terminal region were removed and Lys1, Lys2, Glu7 and His17 were replaced by alanine or tyrosine. In addition, His25, which binds to Zn 2+ in Sp1f3, was replaced by alanine to suppress any possible interactions with trace amounts of metal ions in solution, resulting in the sequence of IP given in Table 1. In NOESY and TOC- SY spectra of 3 mm IP at pH 3.0, most of the NOE peaks overlapped with TOCSY cross-peaks, indicating that these NOE peaks are intraresidual. The remaining NOE peaks, identified as non-intraresidual, came from the sequential signals, C a H-NH, C b H-NH, NH-NH and those related to C d H of prolines. In addition, far- UV CD spectra of IP, which is typical of unfolded proteins, did not change in the concentration range 0.4–2.9 mm (Fig. 1). All of these NMR and CD ana- lyses show that IP remains unfolded up to a con- centration of 3 mm. Sequence and structural properties of FP We attempted to add a peptide fragment to the C-ter- minus of the IP, anticipating that a solution structure might be formed throughout the length of the mole- cule. The number of additional residues was limited to six, because contiguous hydrophobic residues in a pro- tein resulted in a reduced yield in peptide synthesis. Table 1. Sequences of Sp1f3, IP and FPs. Sp1f3 KKFACPECPK 10 RFMRSDHLSK 20 HIKTHQNKK IP YAFACPACPK RFMRSDALSK HIKTA FP1 YAFACPACPK RFMRSDALSK HIKTAFIVVA 30 LG FP2 YAFACPACPK RFMRSDALSK HIKTAYIVVA LG FP3 YAFACPACPK RFMRSDALSK HIKTAYISVA LG M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2337 Among the various hydrophobic scales reported, we used the hydration potential from the gaseous to the aqueous phase (DG hyd ) of amino acids [16,17], because it is quantitatively related to each side-chain and main- chain component. We chose Gly, Pro, Leu, Ile, Val, Ala, Phe, Cys and Met, which have notably larger hydration potentials (greater than  )2 kcalÆmol )1 ) than those of any other amino acids (less than  )5 kcalÆmol )1 ), as candidates for amino acid resi- dues in the fragment. Next, the sequence (X 1 ,X 2 , , X 6 ) of the extra region was chosen to be complemen- tary [18] to P6, C5, A4, F3, A2 and Y1 in the N-termi- nal region, assuming that the interactions between the N- and C-terminal regions are important for folding of the whole molecule. The resulting sequence of the extra region in the final protein, FP1, became FIVVAL (Table 1). In a NOESY spectrum of 3 mm FP1 at pH 3.0, a number of NOE peaks, including long-range NOEs, i.e. Y1C d H–I27C c H and A2C b H–V29C b H, and medium-range NOEs, i.e. cross-peaks of d aM (i,i+2) and d ab (i,i+3), were observed in addition to intraresi- due and sequential NOEs. However, in the case of 0.9 mm FP1, intraresidue and sequential NOE peaks, detected mostly in the case of IP, were observed. In a NOESY spectrum of 1.5 mm FP1, long- and medium- range NOEs observed at 3.0 mm FP1 were partially observed. In addition to the NMR analyses, far-UV CD spectra of FP1 showed that the shape is dependent on the protein concentration (Fig. 1A). The [Q] value at 222 nm for 0.4 mm FP1 was approximately )2000, which is close to that for 0.4 mm IP, whereas it became more negative, to approximately )4000, with an increase in the protein concentration to 3 mm (Fig. 1B). All of these NMR and CD analyses show that the amount of secondary and tertiary structure in FP1 increases with the protein concentration. Self-association of FP1 We examined the degree of protein association at vari- ous concentrations by measuring the sedimentation equilibrium (Fig. 2A) and sedimentation velocity (Fig. 2B). At 0.9 mm FP1, sedimentation equilibrium measurements showed that most plots of the apparent molecular mass are distributed over the range 0–10 000 Da, which corresponds to a molecular mass for a monomeric or dimeric form of FP1 of 3500 or 7000 Da, respectively, although a few plots are > 10 000. At 1.5 mm FP1, although a large number of plots are distributed over the range 0–10 000 Da, the number of plots above 10 000 becomes noticeably higher. At 3.0 mm, most plots are in the range 50 000–70 000, which corresponds to a molecular mass for an oligomer of 15–20 molecules. Sedimentation velocity measurements were also performed by chang- ing the FP1 concentration. At 3.0 mm, the distribution of the sedimentation coefficients showed two main peaks, one at 4.4 and the other at 4.6 (Fig. 2B). These peaks could be assigned to oligomers of 15–20 mole- cules, because it was shown that most FP1 molecules form this type of oligomers at 3 mm, according to the equilibrium measurements. At 2.3 mm, the distribution showed a distinct peak at 1, which is the smallest observed sedimentation coefficient, in addition to the main peak at 4.5. The sedimentation coefficient of monomeric FP1 can be estimated by using the equation correlating the sedimentation coefficient (S) and the A B Fig. 1. Protein concentration dependence of far-UV CD spectra of IP and FP1. (A) Spectra of FP1. (B) [Q] values at 222 nm for IP and FP1. [Q] is molar ellipticity per residue. Solubility-dependent structure formation M. Araki and A. Tamura 2338 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS molecular mass (M): S=M(1)qv s )D ⁄ RT, where q is the density of a solution, v s is the partial specific vol- ume of the solute, D is the diffusion constant of the sol- ute, R is gas constant and T is absolute temperature. S was calculated to be 0.4 by using M = 3500 gÆmol )1 , q =1gÆcm )3 , v s = 0.7 cm 3 Æg )1 , which was the general value of native proteins [19], R = 8.3 JÆK )1 mol )1 , T = 293 K and D = 9.3 · 10 )11 m 2 Æs )1 , obtained using pulsed-field gradient NMR spectroscopy, as described previously [20,21]. Therefore, the peak at 1 can be assigned to the monomer or small oligomer. These sedimentation equilibrium and velocity measure- ments indicate that the amount of monomeric FP1 decreases with an increase in the concentration, whereas the number of 15–20 molecule oligomers increases. pH dependence of solubility and stability The peptide fragment added to the C-terminus of IP in FP1 consists of hydrophobic amino acid residues, which have notably large hydration potentials, DG hyd . In order to identify a determining factor in the struc- tural formation, we prepared two variants whose hydrophobicity is lower than that of FP1: FP2 (Phe26- Tyr) and FP3 (Phe26Tyr and Val28Ser) (Table 1). The hydration potential of FP2 or FP3 can be calculated to be 5.4 or 12.5 kcalÆmol )1 , respectively; lower than that of FP1, based on DG hyd derived from model com- pounds [16]. The solubility of these mutants was mea- sured at various pH values (Fig. 3A). The individual solubility of FPs increases gradually with a decrease in pH in the range 6.5–7.3, presumably because of an increment in the net charges caused by protonation of the imidazole group in His21 and anionic sulfhydryls in Cys5 and Cys8 (Fig. 3B). At all these pH values, the solubility of FP2 is as high as that of FP1, and that of FP3 is clearly the highest. Plots of the solubility at pH < 6.4 were omitted because they were severely scat- tered owing to the steep slope. Experimentally obtained plots for each FP were fitted using Eqn (5), where r p values for the FPs were set to 384 A ˚ , because the amino acid compositions of the FPs were almost identi- cal. This r p value was obtained by fitting FP3, for which the error in the determination of r p is smallest among FPs. Values of l pðsÞ  l o 0 pðsolÞ  =RT obtained by fitting FP1, FP2 and FP3 were )22.8 ± 0.1, )22.5 ± 0.1 and )20.5 ± 0.0, respectively. The pH dependence of the structural stabilities represented by NOE intensities for 3 mm FP1 is given in Fig. 4A,B. With an increase in pH, integrated inten- sities of long-range NOE peaks (Fig. 4A), as well as short- and medium-range NOE peaks (Fig. 4B), increased. The increment was also confirmed for FP2 and FP3 at 3 mm (data not shown). Discussion Solution structures of FPs Complete assignment of the proton chemical-shift resonances was achieved for FPs, excluding the amide A B Fig. 2. Sedimentation equilibrium and sedimentation velocity mea- surements of FP1. (A) The distribution of apparent molecular mass (Mw app ) against the location in the cell, obtained from sedimenta- tion equilibrium measurements at protein concentrations of 0.9 (black), 1.5 (blue) and 3.0 m M (red). Apparent molecular mass was calculated at respective points in the cell, i.e. the higher the A 250 , the closer to the bottom of the cell. (B) Distribution of sedimenta- tion coefficients obtained from sedimentation velocity measure- ments at protein concentrations of 2.3 (blue) and 3.0 m M (red). M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2339 protons of the N-termini, which were not detected because of rapid exchange with solvent. Some of the NOE peaks in the NOESY spectra of 3 mm FPs, how- ever, overlapped and could not be identified separately. Torsion angle restraints, obtained from DQF-COSY, and distance restraints, obtained from clearly separated NOE peaks at a mixing time of 200 ms, are given in Table 2 for 3 mm FP1 at pH 3.0. Because the sedimen- tation equilibrium and velocity measurements of FP1 showed that most of the dissolved species form oligo- mers consisting of 15–20 monomers at a protein con- centration of 3 mm, it is likely that some distance restraints are because of intermolecular interactions caused by the added fragment. However, the ratio of long-range NOEs related to the added fragment to the total of long-range NOEs is 73%, which is notably higher than that of intraresidue (21%), short-range (30%) and medium-range (28%) NOEs, showing that long-range interactions are generated mainly by the added fragment. We deduced that long-range distance restraints are caused by intermolecular interactions, whereas other distance restraints are produced intra- molecularly. The final structural calculation of a FP1 molecule was performed using a total of 342 intraresi- due, short-range and medium-range distance restraints and 25 backbone / dihedral angle restraints (Table 2). The resulting r.m.s.d. from the mean structure for the backbone atoms is 2.79 ± 0.71 A ˚ , which is worse than that derived from typical native proteins (< 0.5 A ˚ ) because long-range distance restraints were not included in the structural calculation. A stereoview of the 10 best FP1 structures (Fig. 5A) shows that the backbone residues Phe12–Ile22 adopt an a helix. By contrast, in the Sp1f3–Zn 2+ complex, the backbone residues Asp16–Gln26 form an a helix or a 3 10 helix, whereas Phe12–Ser15 forms a turn between the second b strand and the helix. Intermolecular interactions were drawn on the lowest energy structures of FP1, assuming that long-range restraints excluded in the A B Fig. 3. pH dependence of the solubilities of FPs. (A) Experimen- tally obtained plots of the natural logarithm of the solubility, S (molÆL )1 ), in the pH range 6.4–7.4. (B) Solubilities calculated using r p = 384 A ˚ and l pðsÞ  l o 0 pðsolÞ  =RT values of )22.8 ± 0.1, )22.5 ± 0.1 and )20.5 ± 0.0 for FP1, FP2 and FP3, respectively, in the pH range 2.5–6.0. Errors in ln S calc for FPs are < 0.1. (Inset) Net charge curves of FPs calculated using pK values for the amino acid side chains, a-COOH and a-NH 3 + termini: Tyr = 10.9, Cys = 8.3, Lys = 10.8, Arg = 12.5, Asp = 3.9, His = 6.0; NH 3 + of the N-terminus = 9.1, and COOH of the C-terminus = 2.4 [22]. Table 2. Structural statistics for the 10 lowest energy structures of FP1. Number of distance restraints Intraresidue 150 Short-range (|i)j | = 1 residues) 118 Medium-range (|i)j | = 2–4 residues) 74 Long-range (|i)j | > 4 residues) (52) a Number of torsion angle restraints / 25 Geometric statistics r.m.s.d. from the mean structure (A ˚ ) Backbone atoms (residues 3–30) 2.79 ± 0.71 All heavy atoms (residues 3–30) 3.51 ± 0.81 Ramachandran analysis Most favored regions (%) 42.1 Additional allowed regions (%) 37.5 Generously allowed regions (%) 18.9 Disallowed regions (%) 1.4 a Long-range distance restraints were not used in the structure calculation (see text). Solubility-dependent structure formation M. Araki and A. Tamura 2340 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS structural calculation are induced intermolecularly (Fig. 5B). The long-range interactions can be divided into three categories: (a) interactions between the N-terminus (Tyr1–Cys8) and the C-terminus (His21– Ala30), (b) interactions between the N-terminus and the middle (Arg11–Arg14), and (c) interactions between the middle and the C-terminus. The N- and C-termini contain mainly hydrophobic amino acids, and the middle also includes the hydrophobic amino acids, Phe12 and Met13. Because most of the long- range NOEs were assigned to these hydrophobic resi- dues, the intermolecular interactions are presumably hydrophobic. CD analysis and structural determination showed that the amount of a helix increases with an increase in FP1 concentration. Therefore, it is likely that the a helix, whose constituent interactions are mainly intramolecular, is induced by the hydrophobic interactions between FP1 molecules. It should also be noted that the structural specificity is apparently low and the structure is therefore loosely packed because the proton chemical shifts do not disperse compared with those observed in typical native proteins, despite the appearance of a number of NOE peaks. Physicochemical factors that determine the structural stability Here, we scrutinize the dependence of protein solubil- ity or hydration potential on structural stability to elu- cidate the determining factor in structural formation. First, the stability was derived from the fraction of the structured molecules at each pH for 3 mm FPs by using six, clearly separated, short- and medium-range NOE peaks, in which distances related to NOE inten- sities were set to the best structure of FP1 [15]. The fractions obtained are shown in Fig. 4C. Fractions of FP1 and FP2, which are close to 0 at pH 2.5, increase to  0.4 as pH is increased to 4.2. The structural stabilities could not be calculated at pH values > 4.2 because of an increase in the aggregation rate. By con- trast, the FP3 fraction is close to 0 at pH 3.8, and increases to  0.4 as pH increases to 5.6. The folding free energies (DG f ) can be calculated using the fractions (Table 3) according to the following simple scheme: 15D $ N 15 ; because ultracentrifuge measure- A B C Fig. 4. pH dependence of the structural stabilities of FPs. (A) Inte- grated intensities of long-range NOE peaks of FP1. (B) Integrated intensities of short- and medium-range NOE peaks of FP1. (C) Frac- tions of structured molecules of FP1 (red), FP2 (blue) and FP3 (green). M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2341 ments suggested that, up to 3 mm, the dissolved spe- cies contain mainly the monomer and 15–20 molecule oligomer. Second, the hydration potential (DG hyd ) was evaluated at each pH value as follows: DG hyd values for the FPs were calculated using the hydration poten- tials of the amino acid side chains and the backbone [16,17]. In addition, because the individual hydration potential for ionizable side chains, a-COOH and a-NH 3 + termini depends on the pH of the solution, i.e. a protonated cation or deprotonated anion is stabilized or destabilized, respectively, with a decrease in pH [16], pH dependence was taken into account using pK values (Table 3) [22]. Third, solubility was represented as the dissolution free energy of solute p (DG dissol ), which can be calculated using DG dissol = )RT ln S p , where R is gas constant, T is absolute temperature and S p is the solubility of solute p. DG dissol reflects the free energy of transfer of solute p from the solid phase to aqueous solution. For a protein as a solute, solubility is known to depend on polarity, hydrophobicity and net charge [23–25], the latter of which increases positively with a decrease in pH. The solubility at each pH value for the FPs was calculated by taking these factors into account and using Eqn (5) (Fig. 3B and Table 3). Plots of DG f , obtained using NOE peaks as described above, against DG hyd show a fairly good correlation (r = )0.70; Fig. 6A), i.e. the less hydrated the protein, the more stable the solution structure. However, plots of DG f against DG dissol show a much stronger correlation (r = )0.86; Fig. 6B), i.e. the more insoluble the protein, the more stable the structure. These results indicate that the structural stability is more strongly dependent on the precipitation capabil- ity than on the dehydration capability. This means that, even if it becomes more hydrated, a protein that prefers precipitation retains the stable structure, as shown in the case of Phe26Tyr replacement. These pre- cipitable proteins might tend to form the solution structure because both structure formation and precipi- tation require the self-association of protein molecules to sequester from solvent. Mode of formation of loosely packed structure through intermolecular interactions When the hydrophobic peptide fragment of FIV- VALG was added to the C-terminus of unstructured IP consisting of 25 residues, formation of the overall protein structure was induced in FP1, with a drastic AB Fig. 5. NMR structures of FP1. (A) Backbone traces of the 10 best structures. Backbones of residues 12–22, which adopt an a helix, are drawn in red. (B) Schematic diagram of long-range interactions between FP1 molecules. Long-range interactions are represented by blue lines between the lowest energy structures of FP1. The side chains of Tyr1, Phe3, Ala4, Cys5, Pro6, Ala7, Cys8, Arg11, Phe12, Met13, Arg14, His21, Ile22, Lys23, Ala25, Phe26, Ile27, Val28, Val29 and Ala30, which are related to the long-range interactions, are indicated in green. In addition, NOE peaks including aromatic protons of Phe3 and Phe12 in FP1 are not clearly separated because chemical shifts of aromatic protons of Phe3 and Phe12 are close to those of Phe26. Therefore, long-range interactions including aromatic protons of Phe3 and Phe12 in FP2, which is the variant Phe26Tyr of FP1, are added. Solubility-dependent structure formation M. Araki and A. Tamura 2342 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS decrease in solubility, i.e. the solubility of IP was > 2.2 mm at pH 7.1 (data not shown), whereas that of FP1 was  10 lm (Fig. 3A). This loosely packed structure is maintained by intermolecular interactions, indicating that the added peptide fragment confers the ability to form the protein structure by having low-specificity interactions. How much hydrophobicity is needed in the added fragment to form this struc- ture? Phe26Tyr replacement in the added fragment resulted in a similar stability in FP2, while keeping the same solubility. However, the additional Val28Ser replacement led to a drastic decrease in the stability of FP3, which showed higher solubility than FP2, indicating that hydrophilic replacement of two of seven hydrophobic residues in the added fragment deprives the whole protein of the ability to form structures. In fact, a decrease of  1.5 kcalÆmol )1 in the dissolution free energy by the replacements Phe26Tyr and Val28Ser resulted in a decrease of  5 kcalÆmol )1 in structural stability (Fig. 6B). These results demonstrate that, among poorly soluble pro- teins, dissolved species tend to be transformed from the solvent-exposed unfolded state into a loosely packed solution structure through intermolecular interactions. The mechanism of structural formation through intermolecular interactions is similar to that of intrinsi- cally unstructured proteins, which are devoid of the well-defined secondary and tertiary structure in A B Fig. 6. Relationships between structural stability and hydrophobic indices. The correlation between the folding free energy (DG f ) and (A) the hydration potential (DG hyd ), defined as the free energy of transfer of a protein solute molecule from the gaseous phase into water, or (B) the dissolution free energy (DG dissol = )RT ln S), for FP1 (red), FP2 (blue) and FP3 (green). Lines represent linear fits with correlation coefficients of )0.70 and )0.86, respectively. To calculate DG hyd of FPs, hydration potentials of 18 amino acid side chains, excluding Pro and Arg, were taken from the values measured by Wolfenden et al. [16]. Those of Pro and Arg side chains and the backbone are taken from the values measured by Privalov et al. [17]. pK values of the amino acid side chains, a-COOH and a-NH 3 + termini are taken from the values in the legend to Fig. 3 [22]. Table 3. pH dependence of folding free energy (DG f ), dissolution free energy (DG dissol ) and hydration potential (DG hyd ) of FPs. Errors in pH and DG dissol are < 0.02 and 0.1, respectively. The folding free energy was calculated according to DG f = )RT ln ([N 15 ] ⁄ [D] 15 ), where [N 15 ] and [D] are the concentrations of the structured oligo- mer and unfolded monomer, respectively. A error in the DG f value for FP3 at pH 5.65 could not be obtained because only the spec- trum at a mixing time of 300 ms was analyzed owing to an increase in the aggregation rate. pH DG f (kcalÆmol )1 ) DG dissol (kcalÆmol )1 ) DG hyd (kcalÆmol )1 ) FP1 2.67 )44.4 ± 0.2 )3.86 )500.4 2.85 )46.5 ± 0.2 )3.24 )498.9 3.05 )49.6 ± 0.3 )2.59 )497.2 3.27 )55.1 ± 0.4 )1.90 )495.4 3.52 )54.1 ± 0.3 )1.10 )493.4 3.91 )67.4 ± 2.7 0.169 )490.4 4.03 )59.2 ± 0.8 0.545 )489.5 FP2 3.00 )47.9 ± 0.3 )2.93 )502.9 3.17 )48.1 ± 0.3 )2.40 )501.5 3.31 )50.6 ± 0.4 )1.96 )500.4 3.86 )49.5 ± 0.3 )0.180 )496.1 4.16 )58.6 ± 1.0 0.740 )493.9 FP3 3.87 )49.0 ± 0.4 )1.34 )503.1 4.20 )62.0 ± 0.6 )0.342 )500.6 4.47 )57.3 ± 0.4 0.315 )498.7 4.82 )59.6 ± 1.3 0.958 )496.2 5.65 )67.0 2.43 )490.7 M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2343 isolation, but adopt relatively rigid conformations upon binding a specific molecular partner of ligands or substrates [26–29]. In the case of FPs, it is unstruc- tured in isolation, as in the case of intrinsically unstructured proteins, whereas a helix (Phe12-Ile22) is induced with a gain in the concentration. One reason that a local conformation consistent with a helix is formed could be attributed to the potential ability [30,31] and ⁄ or local structural preference [32] in the sequence, because the zinc-finger domain Sp1f3, ini- tially chosen as a basis for the FPs, also forms a ter- tiary structure containing a helix (Asp16–Gln26) upon binding to Zn 2+ [14]. We show that the local structure, stabilized originally by coordinate bonds to the metal ion, could be induced by interactions with other copies of the same chain, after the addition of a proper hydrophobic segment responsible for the decrease in solubility. Furthermore, this inductive mechanism indi- cates that hydrophobicity could be regarded as a driv- ing force in the structural formation of FPs, as observed in protein folding in general [13]. By using a simple model of short, self-avoiding flexible chains on lattices, in which the only energetic feature of the sequence is the hydrophobic interaction, protein fold- ing simulations imply a significant probability that a random sequence of amino acids will encode a globu- lar conformation, in general, and a particular native structure, in specific [12]. The globular conformation is interpreted as being like a molten globule, stabilized by intramolecular hydrophobic interactions [12,33]. However, our results show that the unfolded protein can be transformed into the structured assembly by altering the solubility. Because decreasing the protein solubility does not require a sophisticated design for the primary sequence, it is implied that the loosely packed structures with intermolecular interactions shown in FPs may arise readily and frequently com- pared with well-packed structures. The existence of this moderately structured state might serve as an interme- diate stage in the search for the well-packed structures of natural proteins in the vast primary sequence space. In addition, the high occurrence of a primary sequence that prefers self-association seems closely connected to the inherent tendency of natural proteins to aggregate and form potentially harmful deposits such as amyloid fibrils [34–38]. Materials and methods Protein synthesis and purification Proteins were synthesized using the Pioneer Peptide Synthe- sis System (PerSeptive Biosystems, Foster City, CA, USA) with Fmoc solid-phase chemistry, and were cleaved from the resin with a solution containing 82.5% trifluoroacetic acid, 5% H 2 O, 5% thioanisole, 2.5% 1,2-ethanedithiol and 0.8 m phenol (v ⁄ v ⁄ v ⁄ v ⁄ v). Individual proteins were purified by reverse-phase HPLC (acetonitrile ⁄ H 2 O ⁄ 0.1% trifluoro- acetic acid). Protein identity was confirmed by a laser desorption ToF ⁄ MS apparatus, AXIMA-CFR (SHIM- ADZU, Kyoto, Japan). Protein samples for all studies were lyophilized and stored under anaerobic conditions. These purified proteins were handled under a N 2 atmosphere in buffers deoxygenated with N 2 to prevent cysteine oxidation. CD measurements Spectra were acquired at 20 °C on a Jasco J-720 CD spec- tropolarimeter with 0.1, 0.2, 1, 5 and 10 mm pathlength cuvettes on 0.004 to 3 mm protein samples. After each pro- tein was dissolved in a buffer containing 25 mm acetic acid, 2–4 mm NaOH and 50 mm NaCl in 90% H 2 O and 10% D 2 O, the solution was adjusted to pH 3.0 (± 0.1) with NaOH or HCl. Protein concentrations were determined spectroscopically by measuring the amount of protein sulfhydryls with Ellman’s reagent [39]. Ultracentrifuge measurements Each protein sample was prepared as described above. Sedi- mentation velocity and sedimentation equilibrium measure- ments were performed using a Beckman-Coulter Optima XL-1 analytical ultracentrifuge (Fullerton, CA, USA) with an An-60 rotor and two-channel charcoal-filled Epon cells at 20 °C and pH 3.0 (± 0.1). Sedimentation equilibrium was measured at protein concentrations of 0.9, 1.5 and 3.0 mm; sedimentation velocity was measured at 2.3 and 3.0 mm. Data were analyzed using ultrascan 6.01 (http:// www.ultrascan.uthscsa.edu/). NMR spectroscopy NMR measurements were performed on a Bruker DMX- 750 spectrometer at 20 °C on 0.9 to 3.0 mm protein sam- ples. After each protein was dissolved in a buffer containing 25 mm acetic acid, 0–4 mm NaOH and 50 mm NaCl in 90% H 2 O and 10% D 2 O, the solution was adjusted to objective pH with NaOH or HCl. Pulsed-field gradient NMR spectra were acquired at 20 °C and pH 3.0 contain- ing 40 mm 1,4-dioxane at 0.5 mm FP1, at which concentra- tion, CD and sedimentation equilibrium measurements suggested that FP1 was in the monomer state. All chemical shifts were referenced to the sodium salt of trimethylsilyl- propionate. Pulsed-field gradient NMR spectroscopy, 2QF COSY, TOCSY (mixing time of 30 and 80 ms) and NOE spectroscopy experiments were performed and water suppression was achieved by selective presaturation or field- Solubility-dependent structure formation M. Araki and A. Tamura 2344 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS gradient pulses [40]. Proton resonances were assigned using the sequential assignment procedure [41]. Fractions of structured molecules were obtained by analyzing NOESY spectra at mixing times of 100, 150, 200, 250 and 300 ms for the 3.0 mm protein solution. Structure calculations Distance restraints were obtained by converting integrated NOE peak intensities into distance upper limits, using the macro CALIBA in dyana [42]. Standard pseudo atom distances were used when they were needed. Torsion angle constraints for / were determined from 3 J Na . They were then classified into three categories: )120 ± 70, )120 ± 50 and )120 ± 40° corresponding to 3 J Na < 7.5, 7.5–8.5 and > 8.5 Hz, respectively. With a cut-off of 0.2 A ˚ for upper bound NOE violations, 50 structures were generated by using dyana and the 10 lowest energy structures were selected to represent 3D structures. Ramachandran analysis was evaluated by using procheck [43]. Solubility measurements Solubility was estimated using saturated protein solutions as follows. Samples of 200–400 lL protein suspensions in buffers containing 25 mm phosphate and 50 mm NaCl in 90% H 2 O and 10% D 2 O were mixed thoroughly by pipett- ing, then incubated for 20 min at 25 °C. After the incu- bated samples were centrifuged at 17 000 g for 20 min at 25 °C to remove the precipitate, the pH and concentration of the individual supernatant solutions were measured. Pro- tein concentrations were determined spectroscopically by measuring the amount of sulfhydryls with Ellman’s reagent [39]. Analysis of protein solubility The chemical potential of a solute p in a real solution (l p(sol) ) is generally expressed by l pðsolÞ ¼ l 0 pðsolÞ þ RT ln c p S p ð1Þ where l 0 pðsolÞ is the chemical potential in the ideal solution at a standard concentration of p, R is gas constant, T is absolute temperature, c p is the activity coefficient of p and S p is the concentration of p. As a first approximation, if p is present as p z+ ion impenetrable to the solvent, for com- pact protein ions, l 0 pðsolÞ could be divided into the free energy of solvation (DG 0 solv ) that depends on the valence of the ion, Z, and a term independent on the charge of the ion l 0 0 pðsolÞ  : l 0 pðsolÞ ¼ l 0 0 pðsolÞ þ DG 0 solv;p ð2Þ DG 0 solv in Eqn (2) is expressed by Born equation: DG 0 solv;p ¼ Z 2 e 2 N a 8pe 0 r p 1  1 e r  ð3Þ where e is charge of an electron, N a is Avogadro’s number, e 0 is electric constant, e r is relative permittivity, and r p is the radius of the ion. In addition, c p in Eqn (1) is expressed by extended the Debye–Hu ¨ ckel law: log c p ¼ AZ 2 ffiffi I p 1 þ Br p ffiffi I p ð4Þ where A and B are constants, and I is the ionic strength in the solution. In saturated solution, because l p(sol) is equal to the chemical potential of p in the solid (l p(s) ), we can obtain an equation for the solubility of p by using Eqns (1–4): ln S p ¼ ln 10A ffiffi I p 1 þ Br p ffiffi I p þ C r p ! Z 2 þ l pðsÞ  l 0 0 pðsolÞ RT ð5Þ where C ¼ e 2 N a 8pe 0 RT 1  1 e r  ð6Þ Assuming that the dissolved protein is a spherical ion with a net charge of Z and radius r p impenetrable to solvent [24], experimentally obtained plots of individual protein solubility were fitted using Eqn (5) with A = 0.512 (L 1 ⁄ 2 Æmol )1 ⁄ 2 ), B = 0.329 (A ˚ )1 ÆL 1 ⁄ 2 Æmol )1 ⁄ 2 ) and C = 281 (A ˚ )at25°C. Acknowledgements We thank Miyo Sakai (Institute for Protein Research, Osaka University) for the ultracentrifuge measurements. This work was supported in part by Grants-in-Aid for Science from the Ministry of Education, Culture, Sports, Science and Technology of Japan. References 1 Carlsson U & Jonsson BH (1995) Folding of beta-sheet proteins. Curr Opin Struct Biol 5, 482–487. 2 Chakrabartty A & Baldwin RL (1995) Stability of alpha-helices. Adv Protein Chem 46, 141–176. 3 Jackson SE (1998) How do small single-domain proteins fold? Fold Des 3, R81–R91. 4 Kuhlman B & Baker D (2004) Exploring folding free energy landscapes using computational protein design. Curr Opin Struct Biol 14, 89–95. 5 Ponder JW & Richards FM (1987) Tertiary templates for proteins. Use of packing criteria in the enumeration M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2345 [...]... & Caradonna JP (1997) Structures of zinc finger domains from transcription factor Sp1 Insights into sequence-specific protein–DNA recognition J Biol Chem 272, 7801–7809 Araki M & Tamura A (2007) Transformation of an alpha-helix peptide into a beta-hairpin induced by addition of a fragment results in creation of a coexisting state Proteins 66, 860–868 Wolfenden RV, Cullis PM & Southgate CC (1979) Water,... recognition by intrinsically unstructured proteins J Mol Biol 338, 1015–1026 Hoang TX, Marsella L, Trovato A, Seno F, Banavar JR & Maritan A (2006) Common attributes of nativestate structures of proteins, disordered proteins, and amyloid Proc Natl Acad Sci USA 103, 6883–6888 Chikenji G, Fujitsuka Y & Takada S (2006) Shaping up the protein folding funnel by local interaction: lesson from a structure... with backbone freedom Science 282, 1462–1467 Roy S & Hecht MH (2000) Cooperative thermal denaturation of proteins designed by binary patterning of polar and nonpolar amino acids Biochemistry 39, 4603– 4607 Lau KF & Dill KA (1990) Theory for protein mutability & biogenesis Proc Natl Acad Sci USA 87, 638–642 Dill KA (1990) Dominant forces in protein folding Biochemistry 29, 7133–7155 Narayan VA, Kriwacki... protein folding, and the genetic code Science 206, 575–577 Privalov PL & Makhatadze GI (1993) Contribution of hydration to protein folding thermodynamics II The entropy and Gibbs’ energy of hydration J Mol Biol 232, 660–679 Wouters MA & Curmi PM (1995) An analysis of side chain interactions and pair correlations within antiparallel beta-sheets: the differences between backbone hydrogen-bonded and non-hydrogen-bonded... Protein Sci 4, 2006– 2018 Isogai Y, Ito Y, Ikeya T, Shiro Y & Ota M (2005) Design of lambda Cro fold: solution structure of a monomeric variant of the de novo protein J Mol Biol 354, 801–814 Kuhlman B, Dantas G, Ireton GC, Varani G, Stoddard BL & Baker D (2003) Design of a novel globular protein fold with atomic-level accuracy Science 302, 1364–1368 Harbury PB, Plecs JJ, Tidor B, Alber T & Kim PS (1998)... for NMR structure calculation Solubility-dependent structure formation with the new program DYANA J Mol Biol 273, 283– 298 43 Laskowski RA, Rullmannn JA, MacArthur MW, Kaptein R & Thornton JM (1996) AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR J Biomol NMR 8, 477–486 FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS 2347... Tanford C (1961) Physical Chemistry of Macromolecules Wiley, New York, NY Shaw KL, Grimsley GR, Yakovlev GI, Makarov AA & Pace CN (2001) The effect of net charge on the solubility, activity, and stability of ribonuclease Sa Protein Sci 10, 1206–1215 Wright PE & Dyson HJ (1999) Intrinsically unstructured proteins: re-assessing the protein structure–function paradigm J Mol Biol 293, 321–331 Dunker AK,.. .Solubility-dependent structure formation 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 M Araki and A Tamura of allowed sequences for different structural classes J Mol Biol 193, 775–791 Dahiyat BI & Mayo SL (1997) De novo protein design: fully automated sequence selection Science 278, 82–87 Desjarlais JR & Handel TM (1995) De novo design of the hydrophobic cores of proteins Protein... FEBS Journal 276 (2009) 2336–2347 ª 2009 The Authors Journal compilation ª 2009 FEBS M Araki and A Tamura 40 Piotto M, Saudek V & Sklenar V (1992) Gradient-tailored excitation for single-quantum NMR spectroscopy of aqueous solutions J Biomol NMR 2, 661–665 41 Wuthrich K (1986) NMR of Proteins and Nucleic Acids Wiley, New York, NY 42 Guntert P, Mumenthaler C & Wuthrich K (1997) Torsion angle dynamics for... residue pairs Proteins 22, 119–131 Gohon Y, Pavlov G, Timmins P, Tribet C, Popot JL & Ebel C (2004) Partial specific volume and solvent interactions of amphipol A8 -35 Anal Biochem 334, 318–334 Stejskal EO & Tanner JE (1965) Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient J Chem Phys 42, 288–292 Dingley AJ, Mackay JP, Chapman BE, Morris MB, Kuchel PW, Hambly BD . Solubility-dependent structural formation of a 25-residue, natively unfolded protein, induced by addition of a seven-residue peptide fragment Mitsugu Araki and Atsuo Tamura Graduate School of. 7801–7809. 15 Araki M & Tamura A (2007) Transformation of an alpha-helix peptide into a beta-hairpin induced by addi- tion of a fragment results in creation of a coexisting state. Proteins. YAFACPACPK RFMRSDALSK HIKTAFIVVA 30 LG FP2 YAFACPACPK RFMRSDALSK HIKTAYIVVA LG FP3 YAFACPACPK RFMRSDALSK HIKTAYISVA LG M. Araki and A. Tamura Solubility-dependent structure formation FEBS Journal 276 (2009)