Sequence and Structure Determinants for the Self-aggregation of Recombinant Polypeptides Modeled after Human Elastin*

Elastin is a polymeric structural protein that imparts the physical properties of extensibility and elastic recoil to tissues. The mechanism of assembly of the tropoelastin monomer into the elastin polymer probably involves extrinsic protein factors but is also related to an intrinsic capacity of elastin for ordered assembly through a process of hydrophobic self-aggregation or coacervation. Using a series of simple recombinant polypeptides based on elastin sequences and mimicking the unusual alternating domain structure of native elastin, we have investigated the influence of sequence motifs and domain structures on the propensity of these polypeptides for coacervation. The number of hydrophobic domains, their context in the alternating domain structure of elastin, and the specific nature of the hydrophobic domains included in the polypeptides all had major effects on self-aggregation. Surprisingly, in polypeptides with the same number of domains, propensity for coacervation was inversely related to the mean Kyte-Doolittle hydropathy of the polypeptide. Point mutations designed to increase the conformational flexibility of hydrophobic domains had the unexpected effect of sup-pressing coacervation and promoting formation of amyloid-like fibers. Such simple polypeptides provide a useful model system for understanding the relationship between sequence, structure, and mechanism of assembly of polymeric elastin. Elastin is the

Elastin is the extracellular matrix protein that imparts the important properties of extensibility and elastic recoil to tissues such as large blood vessels, lung parenchyma and elastic ligaments. To achieve the remarkable durability and structural integrity of the elastin matrix, monomers of elastin must be appropriately aligned and covalently cross-linked into an extensive polymeric fibrillar structure (1). However, the mechanisms by which this assembly process takes place are not well understood.
Tropoelastin, the monomeric form of elastin, is coded for by 34 exons in humans. With a few exceptions, each exon codes for either a "hydrophobic" or "cross-linking" domain, and these domains generally alternate along the sequence of the protein.
Hydrophobic domains are rich in amino acids such as glycine, proline, valine, and leucine, in many cases present in a variety of tandem repeat sequences. In contrast, cross-linking domains contain the lysine residues that are destined to take part in covalent cross-linking and are often rich in alanine (1).
One of the unusual properties of elastin is its ability to undergo coacervation, a self-aggregation process in which the protein comes out of solution as a second phase on an increase in solution temperature. Coacervation of elastin was first demonstrated with heterogeneous mixtures of polypeptides prepared by hydrolytic solubilization of insoluble, polymeric elastin (2)(3)(4)(5). Subsequently, both synthetic polypeptides cor-responding to hydrophobic domains of elastin (6 -9) and full-length tropoelastin (10 -12) have also been shown to undergo this process. The ability to undergo coacervation is related to the predominantly nonpolar character of these polypeptides, and the kinetics of the transition appears to be that of a nucleation process (3,12). The temperature at which this transition takes place is dependent on several factors including protein concentration, ionic strength, and pH (12,13).
In contrast to heat-induced protein denaturation, which results in disordered structures, coacervation of elastin polypeptides has been shown to promote the formation of well ordered filamentous structures (10 -11, 14 -16). For this reason, it has been proposed that coacervation might play an important physiological role in the ordering and alignment of monomeric elastin for cross-linking into the polymeric elastin matrix (11,16,17). Weiss and colleagues (18 -20) have studied the effect of truncation and deletion of specific domains of human tropoelastin on the ability of the protein to coacervate. Our laboratory has investigated the coacervation behavior of smaller, well defined recombinant polypeptides based on human elastin sequences (13) and has recently shown that coacervation of polypeptides containing as few as five domains of human elastin promotes alignment of lysine residues for cross-link formation and allows self-assembly of these polypeptides into crosslinked fibrillar matrices with physical properties similar to those of native elastin (16,17). Such polypeptides provide an excellent model system for understanding the mechanism of polymeric assembly of elastin and the relationship between structural organization and physical properties. Here, using a series of defined recombinant polypeptides modeled after human elastin, we investigate the role of sequence, polypeptide size and character, domain context, and chain flexibility on the process of hydrophobic self-assembly of elastin. [21][22][23] n for primer sets BamHI-Ex20 and Ex20-EcoRI to generate fragments consisting of exons 20- (21-23-20) n . A 520-bp fragment consisting of exons 20-21-23-20-21-23-20 was purified from the PCR mixture. This fragment was then digested with BamHI and EcoRI and inserted into the BamHI/EcoRI-treated pGEX-2T vector to generate EP20-20-20. Exons 24-21-23-24-21-23-24)-PCR was  carried out using EP20-24-24-24-24 as the template and BamHI-Ex24  and Ex24-EcoRI as the primer set. A 650-bp fragment consisting of exon  24-21-23-24-21-23-24 was purified from the PCR products. EP24-24-24 was generated by ligating this 650-bp fragment into the BamHI/EcoRItreated pGEX-2T vector.

/G)) and EP 20-CRS[P/G]-CRS[P/G] (Human Elastin
All oligonucleotide primers were synthesized, and construct sequences were confirmed using facilities provided by the Centre of Applied Genomics at the Hospital for Sick Children. Synthetic polypeptides consisting of exons 20 and 24 were kindly provided by Dr. A. Tamburro (University of Basilicata, Italy).

Polypeptide Expression and Purification
DNA constructs were transformed into BL21 cells by electroporation for peptide expression. A single colony was inoculated in 2ϫ YT with ampicillin (50 g/ml) and chloramphenicol (34 g/ml) at 37°C overnight. This culture was then reamplified in 2ϫ YT containing 2% glucose and ampicillin for 3.5 h (A 600 0.8 -1.0) before adding isopropyl-␣-D-thiogalactopyranoside (0.1 mM) to induce protein expression. After a 4-h incubation at 37°C, the bacterial culture was harvested by centrifugation for 10 min at 7500 ϫ g. The cell pellet was digested with cyanogen bromide in 70% formic acid at room temperature overnight followed by dialysis (3.5-kDa cut-off; Pierce) against water for 24 -36 h. Elastin polypeptides were purified from this mixture by Sephadex G25 (Amersham Biosciences) chromatography, eluting with 20 mM sodium acetate, followed by chromatography using a Sepharose SP (Amersham Biosciences) ion exchange column eluted with 80 mM NaCl in 20 mM sodium acetate. All of the polypeptides (excluding EP20-24[P/G]-24[P/ G]) were desalted on a Sephadex G25 column and lyophilized before final purification by reverse-phase high pressure liquid chromatography using a Jupiter 10-m C4 300-Å column (Phenomenex, Torrance, CA). Amino acid compositions and concentrations of all polypeptides were determined by amino acid analysis, and molecular weights were confirmed by Q-TOF mass spectrometry using the facilities of the Advanced Protein Technology Centre (Hospital for Sick Children).

Coacervation of Polypeptides
Unless otherwise noted, polypeptides were dissolved in coacervation buffer (50 mM Tris, pH 7, containing 1 mM CaCl 2 and 1.5 M NaCl), and coacervation was performed by increasing solution temperature at a rate of 1°C/min, monitoring absorbance at 440 nm using a Cary 3 spectrophotometer equipped with a temperature controller. Coacervation temperature was determined as the temperature of onset of increased absorption (13).

Circular Dichroism Spectrometry
Polypeptides were dissolved in water at 15 M and stored overnight at room temperature. CD spectra were obtained at 20°C using an AVIV 62DS spectrometer.

Thioflavin-T Assay
The assay for binding of thioflavin-T was performed according to Ref. 22. Briefly, polypeptides were dissolved in water at a concentration of 0.5 g/l and left at room temperature for 4 h. Then 10 l of polypeptide was incrementally added to 1 ml of 3 M thioflavin-T in potassium phosphate buffer (50 mM, pH 6.0). Fluorescence at 482 nm (excitation at 450 nm) was monitored using a Hitachi F-4000 fluorescence spectrophotometer. ␤-amyloid 1-28 (American Peptide Co., Sunnyvale, CA) was used as positive control.

Transmission Electron Microscopy
EP20-24-24(P/G) was dissolved at a concentration of 250 M in water containing 1.5 M NaCl, incubated at 20°C overnight, and stored at 4°C for 1 week. The sample was then applied to a carbon-coated copper grid, negatively stained with 2% aqueous uranyl acetate, and viewed under a Hitachi H600 transmission electron microscope as described elsewhere (17).

RESULTS
Representations of the polypeptides used, along with domain sequences, are shown in Table I. All polypeptides are named on the basis of their hydrophobic exons; intervening cross-linking domains, consisting of exons 21 and 23, are always identical, with the exception of EP20-24-24[21Y/A], in which the tyrosine residues in the two copies of exon 21 were mutated to alanines. Characteristics of the polypeptides, including molecular weight, hydropathy, and coacervation temperature under standard conditions (see "Experimental Procedures") are also included. Mean hydropathies of the polypeptides were calculated as the sum of the Kyte-Doolittle hydropathy values for each amino acid (23) divided by the number of amino acids.
Number of Hydrophobic Domains-The ability of native tropoelastin or elastin polypeptides to coacervate has been attributed to the presence of the hydrophobic domains (19). Indeed, earlier data from our laboratory using recombinant human elastin polypeptides demonstrated that polypeptides containing cross-linking domains alone did not undergo coacervation (13). Using polypeptides EP20-24, EP20-24-24, and EP20-24-24-24-24, which differ in the number but not in the nature of hydrophobic domains, we investigated the effect of hydrophobic domain number on the propensity for self-aggregation in these polypeptides. In all cases, polypeptide concentrations and salt contents of the coacervation buffer were identical (25 M polypeptide and 1.5 M NaCl, respectively). All three polypep-tides underwent coacervation. EP20-24, containing two hydrophobic domains flanking a single cross-linking domain, coacervated at ϳ41°C, whereas EP20-24-24 and EP20-24-24-24-24, containing three and five hydrophobic domains, coacervated at ϳ29 and 21°C, respectively (Fig. 1A). These results indicated a strong relationship between number of hydrophobic domains and propensity for self-aggregation (Fig. 1B). Because of the similarity in composition of these three polypeptides, their mean hydropathies were essentially identical (Table I).
Co-coacervation Behavior of Polypeptide Mixtures-The nucleation-like kinetics of self-aggregation of elastin polypeptides raises the question of whether coacervation of one polypeptide can induce that process for other elastin polypeptides in solution. Availability of elastin polypeptides with similar natures but widely differing coacervation temperatures provided the possibility of investigating this question. Equal concentrations (12.5 M) of EP20-24 and EP20-24-24-24-24 were dissolved in the standard coacervation buffer (see "Experimental Procedures"), and the solution temperature was raised. Under these conditions, EP20-24 and EP20-24-24-24-24 alone coacervate at ϳ52 and 22°C, respectively (Fig. 2). In an equimolar mixture with EP20-24-24-24-24, the coacervation temperature of EP20-24 was unchanged. Moreover, despite the coacervation of EP20-24-24-24-24 at 22°C, EP20-24 remained in solution until the expected coacervation temperature for this polypeptide was reached (Fig. 2). These results indicated that the process of coacervation of each of these two polypeptides was independent of the presence of the second polypeptide in solution.
Effect of Individual Hydrophobic Domains-Exon 24 contains the most striking tandem repeat in human elastin, consisting of a 7-fold repeat of a PGVGVA sequence. In contrast, exon 20 contains a 4-fold repeat of GVGGVP that includes an internal 3-fold repeat of PGVGGV. Based on previous speculations on the role of such repeated sequences in self-aggregation (21), we compared propensities for coacervation of polypeptides containing only exon 20 or 24 as hydrophobic domains (EP20-20-20 and EP24-24-24). Again, coacervation conditions (25 M polypeptide, 1.5 M NaCl) were identical for both polypeptides. Unexpectedly, the polypeptide containing three exon 24 domains showed an increased coacervation temperature relative to EP20-24-24. In contrast, EP20-20-20 coacervated at a temperature more than 10°C below that of EP24-24-24 (Table I).
These results demonstrated clearly that the presence of exon 24 domains in these polypeptides did not increase their propensity for self-aggregation.
Coacervation characteristics of several other polypeptides were also compared under standard conditions (25 M polypeptide, 1.5 M NaCl). All of these polypeptides had approximately the same molecular weight, and all contained three hydrophobic domains separated by two cross-linking domains. In each case, using EP20-24-24 as a reference polypeptide, other hydrophobic sequences were substituted for one or both of the exon 24 domains. These polypeptides included EP20-24-26 (exon 26 contains a 3-fold nonapeptide repeat of PGL-GVGVGV), EP20-24-30 (exon 30 has a relatively low proline content and contains a 3-fold repeat of GGLGV), and EP20-24-LRS. LRS contains the tandemly repeated sequence GGLGY present in lamprin, a self-aggregating matrix protein of lamprey cartilage (21). Similar repeated sequences are present in structural proteins of a variety of invertebrates (21). Characteristics of these polypeptides are summarized in Table I. Substitution of exon 30 for exon 24 had little effect on coacervation temperature. In contrast, substitution of exon 26 for exon 24 resulted in a polypeptide with a coacervation temperature ϳ5°C lower than that of EP20-24-24. Most strikingly, substitution of the LRS for exon 24 lowered the coacervation temper- Coacervation conditions were 25 M polypeptide in 50 mM Tris buffer, pH 7, containing 1 mM CaCl 2 and 1.5 M NaCl. The temperature at onset of coacervation was determined as described previously (13). B, relationship between molecular weight (number of hydrophobic domains) and coacervation temperature for these three elastin polypeptides. ature to ϳ4°C, more than 25°C lower than that of EP20-24-24.
In order to assess the role of the tyrosine in the GGLGY repeat of the LRS in determining its remarkably increased propensity for self-aggregation, EP20-24-LRS[Y/A] was expressed, in which each of the tyrosines in the GGLGY sequences was replaced with an alanine residue. These substitutions resulted in an increase of ϳ20°C in the coacervation temperature of the polypeptide (Table I).
Whereas coacervation appears to require the presence of hydrophobic domains (13), sequences within the cross-linking domains could also have an effect on self-aggregation properties of these polypeptides. Because of the strong effect of the presence of tyrosine in EP20-24-LRS, we also investigated the effect of mutating the tyrosine residue in exon 21 located adjacent to one of the lysine residues. This polypeptide, designated EP20-24-24[21Y/A] and containing tyrosine to alanine mutations in exon 21 sequences present in both cross-linking domains, showed an increase in coacervation temperature of about 7°C compared with the EP20-24-24 (Table I). This result was consistent with the effect of tyrosine mutations in EP20-24-LRS and confirmed that sequences in the cross-linking domains could indeed affect propensity for self-aggregation.
We have proposed that promotion of self-aggregation might be related to the ability of such tandemly repeated sequences to form repeated short ␤-sheet/␤-turn structures displaying large hydrophobic side chains on both sides of the plane of the ␤-sheet in what has been called a "Lego motif" (21). If such conformations are important for coacervation, the nature of the repeat (e.g. pentapeptide versus hexapeptide) may have a significant effect. Indeed, whereas the hexapeptide PGVGVA is the prominent repeated sequence in human elastin, the corresponding region of chicken elastin contains a tandem repeat of the pentapeptide PGVGV. We therefore investigated the effect of replacing the hexameric PGVGVA repeat in one or both exon 24 domains of EP-20-24-24 with a pentameric PGVGV repeat (CRS). These polypeptides were designated EP20-24-CRS and EP20-CRS-CRS. In both cases, coacervation temperatures of the polypeptides were lowered relative to EP20-24-24, with the effect being greater in the polypeptide in which both positions of the exon 24 domain sequences were replaced (Table I).
Although all of these polypeptides contained different sets of hydrophobic domains, their molecular weights were very similar, and there was no apparent relationship between their coacervation temperatures and these small differences in mo-  M) (f) and EP20-24-24-24-24 (12.5 M) (q) were each coacervated alone under standard buffer conditions. The solution temperature of a mixture of 12.5 M each of EP20-24 and EP20-24-24-24-24 (E) was raised to 30°C, and the coacervated material was removed by centrifugation at 30°C. The temperature of the remaining solution (Ⅺ) was then raised to 60°C. The inset shows SDS-polyacrylamide gel electrophoresis of solutions of these polypeptides. Polypeptides were detected by Western blotting using a polyclonal antibody raised to EP20-24 (13). lecular weight (Fig. 3A). In contrast, there was a clear relationship between coacervation temperature and the mean hydropathies of the polypeptides (Fig. 3B).
Effect of Context of Hydrophobic Domains-A variety of hydrophobic polypeptides based on elastin sequences have previously been shown to undergo coacervation (6 -9). However, all of these contained only hydrophobic domains without the presence of the interspersed cross-linking domains present in tropoelastin. Therefore, it was not clear whether a more elastinlike context for the hydrophobic domains, as represented in the polypeptides described here, would also contribute to propensity for self-aggregation. This possibility was investigated using polypeptides corresponding to sequences of human exons 20 and 24, both generous gifts of Dr. A. Tamburro.
Under standard coacervation conditions (25 M polypeptide, 1.5 M NaCl), neither of these polypeptides could be induced to coacervate by increased solution temperature. However, increasing the polypeptide concentrations to 800 M allowed coacervation to take place. Consistent with the differences in coacervation temperatures between EP20-20-20 and EP24-24-24, at 800 M concentrations, the polypeptide consisting of exon 24 coacervated at a temperature ϳ25°higher than that consisting of exon 20 (Fig. 4A). In order to quantitatively compare propensities for coacervation between EP20-20-20 and exon 20 alone, coacervation temperatures for EP20-20-20 at a series of polypeptide concentrations were determined. Earlier results from our laboratory (13), had demonstrated a linear relationship between the natural logarithm of polypeptide concentration and coacervation temperature, allowing interpolation between the data points (Fig. 4B). At a concentration of 2000 M, the coacervation temperature of exon 20, determined experimentally, was ϳ22°C. Based on the relationship shown in Fig.  4B, the concentration of EP20-20-20 required to coacervate at that temperature would be ϳ16 M. Thus, coacervation of EP20-20-20 under the same conditions as exon 20 alone requires 125-fold less polypeptide. Taking into account that EP20-20-20 includes three copies of exon 20/mol, the propensity for coacervation of EP20-20-20 is ϳ40 times that of exon 20. These data indicate that the presence of hydrophobic domains in an elastin-like context has a major effect on coacervation properties.
Effect of Increased Polypeptide Chain Flexibility-With the exception of exon 30, most hydrophobic domains of elastin are relatively rich in proline. Because of the fixed peptide bond angles imposed by proline residues in the polypeptide chain, replacement of these prolines with glycines should confer increased conformational flexibility and might be expected to substantially affect coacervation properties of these polypeptides. In order to investigate the role of polypeptide chain flexibility, proline to glycine substitutions were made for all Neither EP20-24-24[P/G] nor EP20-24[P/G]-24[P/G] could be induced to coacervate by increased solution temperature. EP20-24-24[P/G] was soluble in water but in the presence of 1.5 M NaCl formed a fibrillar precipitate resembling amyloid fibrils (Fig. 5C). EP20-24[P/G]-24[P/G], containing proline to glycine substitutions in both of the exon 24 hydrophobic domains, was soluble in 88% formic acid but formed a precipitate on dialysis into water. Similarly, EP20-CRS[P/G]-CRS[P/G], containing proline to glycine mutations in both pentapeptide repeat domains, did not coacervate but rather formed an amyloid-like precipitate in water. In contrast, EP20-CRS-CRS[P/G] coacervated at 27°C under standard conditions (25 M polypeptide, 1.5 M NaCl). However, storage of this polypeptide at room temperature in 0.5 M NaCl for 1-2 days resulted in the precipitation of amyloid-like fibrils.
Differences in these characteristics of the polypeptides were supported by structural data provided by circular dichroism spectroscopy in water (Fig. 5A). EP20-24-24, used as a reference coacervating polypeptide, showed a CD spectrum typical of elastin-like polypeptides (3, 9, 12, 24 -27), with a small trough at ϳ222 nm and a larger trough at ϳ200 nm. This spectrum is usually interpreted to represent largely random coil structure with some contribution of ␣-helical segments attributed to the cross-linking domains (26,27). The CD spectrum of EP20-24-24[P/G] was similar to that of EP20-24-24. However, the CD spectrum of EP20-24[P/G]-24[P/G], which formed amyloid-like fibrils in water, showed a deep trough at ϳ222 nm and a large positive peak at 195 nm, suggesting the presence of extended ␤-sheet structure.
Formation of amyloid-like fibrils was also assessed by binding of the dye thioflavin-T (Fig. 5B), a well established assay for detecting amyloid fibril assembly (22). For EP20-24-24 and EP20-24-LRS, both of which undergo coacervation, thioflavin-T fluorescence was minimal and was not concentration-dependent. Similar results were seen for EP20-24-24[P/G] in water. However, in 1.5 M NaCl, conditions under which this polypeptide formed an amyloid-like precipitate, thioflavin-T fluorescence became strongly concentration-dependent. EP20-24[P/ G]-24[P/G], the polypeptide that produced an amyloid-like precipitate in water, also showed a similarly strong concentration dependence for dye binding. Ultrastructural evidence (Fig.  5C) confirmed that the fibrillar arrays produced from the amyloid-forming polypeptides were similar to those previously published for amyloid networks (28) but very different from the ordered fibrils produced from the coacervated elastin polypeptides (10, 11, 15, 16). Taken together, these results indicated that proline to glycine mutations in hydrophobic domains of the elastin polypeptides did not promote coacervate formation but rather increased the tendency of these polypeptides to form amyloid-like fibrils. DISCUSSION The assembly of monomeric elastin (tropoelastin) into its polymeric extracellular matrix structure is a process that is not well understood. Whereas extrinsic factors such as the microfibrillar scaffolding (29 -32) and the 67-kDa elastin-binding protein (33,34) undoubtedly play a role in promoting assembly and determining the final architecture of the elastic matrix, it has become clear that tropoelastin holds within its unusual domain structure the intrinsic ability to self-assemble into fibrillar structures. Indeed, recent evidence from our own laboratory has demonstrated that polypeptides representing a relatively small portion of the tropoelastin molecule can undergo such ordered self-assembly (16). The remarkable durability of elastin as a structural protein, together with this intrinsic capacity for self-organized polymerization, has recently attracted much attention from laboratories interested in designing new biomaterials with useful physical properties, and several groups, including our own, have used elastin-like sequences to fabricate materials with properties similar to native elastin (16,31,35,(37)(38)(39). 2 Organized self-assembly of elastin involves the process of coacervation, a reversible phase separation induced by increased solvent temperature. The physical chemistry of coacervation of elastin and elastin-like polypeptides is related to the phenomenon of hydrophobic hydration (41)(42)(43)(44). Generally, it is agreed that the largely nonpolar character of elastin sequences, especially in the hydrophobic domains, and the consequent reduced ability of the protein to form hydrogen bonds with water molecules in the surrounding solvent results in the formation of stable, hydrogen-bonded clathrate water structures surrounding the elastin molecules in solution, shielding them from each other and therefore from the possibility of self-aggregation. However, this clathrate water is destabilized both by increased ionic strength of the solution and by increased temperature. The coacervation temperature is the temperature at which this clathrate water structure breaks down, resulting in exposure of hydrophobic domains and the formation of aggregates. The process is thermodynamically favorable because the increased order induced in the aggregates of elastin proteins is more than offset by the decreased order of the surrounding water. In this study, we have used the coacervation temperature, under standard conditions of pH, ionic strength, and polypeptide concentration, as a measure of the propensity of a polypeptide to self-aggregate.
Coacervation was first reported many years ago for heterogeneous mixtures of hydrolytic fragments of insoluble elastin (2,5). Coacervation of a variety of elastin-like polypeptides as well as both full-length and some deletion and truncation mutations of tropoelastin has been reported (8, 11-13, 18 -20). However, there is still relatively little information on the detailed relationship between sequence, structure, and propensity for coacervation of elastin. Although Urry has reported coacervation characteristics of a number of synthetic polypeptides (7), the sequences of these polypeptides were for the most part based only on hydrophobic domains of elastin. In contrast, in this study we have utilized well defined polypeptides that contain both hydrophobic and cross-linking domains and conform to the alternating domain structure of elastin.
Our results demonstrated that the general hydrophobic nature of a polypeptide (measured by mean Kyte-Doolittle hy-dropathy) is not the sole determinant of coacervation temperature. Polypeptides with essentially identical hydropathies and consisting of the same domain sequences had widely differing propensities for coacervation, depending on the number of domains included. In practice, these experiments cannot distinguish between effects of additional hydrophobic domains and those due to increased overall molecular weight. Weiss and colleagues (12) have shown that even relatively low concentrations of tropoelastin, which has a molecular mass of ϳ70 kDa and contains 34 of these domains, will coacervate at ϳ37°C in physiological solutions and have reported correlations between molecular weight and coacervation temperature in variants of tropoelastin (20). We have previously reported that EP20-24, a polypeptide containing only two hydrophobic domains flanking a single cross-linking domain, although capable of coacervation, was less able to produce compacted, fibrillar, cross-linked structures (13), suggesting that ordered self-assembly may be promoted in polypeptides with larger numbers of hydrophobic domains and therefore greater potential for overlapping of domains in the formation of fibrillar structures.
Despite the apparent nucleation kinetics of the coacervation process, our results showed clearly that self-aggregation of one polypeptide did not induce coacervation of a second polypeptide (i.e. the temperatures at which coacervation took place for each of these polypeptides were independent of the presence of the other in the solution). Such behavior is consistent with observations of Wu and Weiss (20) on tropoelastin and with the view that coacervation takes place when conditions are such that the clathrate water surrounding the polypeptide in solution becomes destabilized. That is, elastin-like polypeptides in solution do not "see" each other until breakdown of the ordered shell of water surrounding them. Furthermore, the stability of this ordered water shell is determined by the nature of the polypeptide, and destabilization of the clathrate water around one polypeptide does not influence the stability of the water shell surrounding a second polypeptide.
Comparing polypeptides that included three hydrophobic domains flanking two cross-linking domains and were closely similar in molecular weights, experimental results demonstrated that the nature of the hydrophobic domains can exert a large influence on propensity for self-assembly. Under standard conditions, coacervation temperatures for polypeptides in this group varied by over 30°C. Unexpectedly, the presence of exon 24 with its prominent tandem repeat of the hexamer PGVGVA did not, as originally speculated, impart to polypeptides any special propensity for coacervation. Indeed, comparisons between polypeptides containing exons 20 and/or 24, either in the alternating domain context of elastin or as exon sequences alone, indicated that exon 20 appeared to confer greater propensity for coacervation. Similarly, hydrophobic domains containing the pentapeptide repeat PGVGV had somewhat lower coacervation temperatures. Although Jensen et al. (18) have reported that exon 26 is important for coacervation of tropoelastin and Kozel et al. (30) have identified exon 30 as a major functional element for elastin deposition into the extracellular matrix, in the context of the polypeptides used in this study, neither exon 26 nor exon 30 was particularly potent for promoting coacervation.
For all polypeptides used in this study, there was a clear relationship between mean Kyte-Doolittle hydropathy of the polypeptides and coacervation temperature, with decreased hydropathy associated with increased propensity for coacervation. Although the nature of this relationship might at first seem unexpected, we suggest that uniformly highly nonpolar domains may, in fact, promote the formation of more stable shells of clathrate water, resulting in higher coacervation temperatures.
In contrast, "imperfections" in clathrates surrounding less hydrophobic domains that allow some hydrogen bonding to the polypeptide side chains may result in less stable shells and lower coacervation temperatures. Consistent with this hypothesis is the observation that the presence of tyrosine residues in the polypeptides, introducing phenolic hydroxyl groups capable of hydrogen-bonding to water, was in all cases associated with substantially decreased coacervation temperatures.
Similar effects of residue substitution have been reported by Urry et al. (7). Using "guest" residues introduced into a series of synthetic polypeptides modeled after hydrophobic domains of elastin, Urry et al. (7) proposed a new hydrophobicity scale using coacervation temperature as a measure of polypeptide hydrophobicity. In Urry's scale, tyrosine, for example, occupies a much more hydrophobic position than it does in the Kyte-Doolittle scale, which is based on partition of amino acids into nonpolar solvents. Indeed, replotting of the data in Fig. 3B using the Urry hydrophobicity scale results in a relationship in which decreased coacervation temperature is now proportional to increased hydropathy (data not shown). On the one hand, this might be expected, since Urry's scale was based on coacervation temperature. However, it was interesting to note that this relationship held despite the fact that Urry's scale was designed using hydrophobic domains only, in comparison with the polypeptides used here that contained both hydrophobic and cross-linking domains.
For all of the polypeptides used in this study, hydrophobic domains were interspersed with a representative cross-linking domain consisting of exons 21 and 23, giving the polypeptides the alternating domain structure of native elastin. Although the cross-linking domain on its own is not capable of coacervation (13), such an alternating domain structure appeared to significantly enhance the ability of the polypeptides to coacervate. Using exon 20 as a common hydrophobic sequence, it was possible to compare coacervation characteristics of exon 20 alone with that of a polypeptide containing three exon 20 domains interspersed with two cross-linking domains. The results demonstrated that, although hydrophobic sequences such as exon 20 are able to coacervate on their own, placing these sequences in the context of the alternating domain structure of native elastin considerably increased their propensity for coacervation. The structural basis of this change in properties is not known, but it may be related to increased opportunities for ordered self-aggregation introduced through spacing of the hydrophobic domains. An effect of domain context on coacervation of tropoelastin has also been proposed by Toonkool et al. (19).
Other than a general role for ␤ structures, there is as yet no current consensus on structural features of elastin-like polypeptides that are related to their ability to coacervate. Urry (46,47) has suggested that the hydrophobic domains of elastin take on a ␤-spiral conformation, whereas Tamburro and others (42,45,48,49) have provided evidence for other structures involving repeated and perhaps sliding ␤-turns. Difficulties in establishing the structure of these hydrophobic domains may in fact be due to their inherent conformational flexibility, a characteristic that has been related to their properties as elastomers (49). Indeed, this flexibility may mean that the structures present after coacervation may be induced during the coacervation process.
To investigate the effect of conformational flexibility on coacervation characteristics, polypeptides were produced in which proline residues, with their fixed peptide bond angles, were replaced with glycine residues. The consequence of this substitution was a change in the nature of the aggregation process from coacervation to formation of amyloid-like fibers. Amyloid formation has been explained by stacking of exposed, extended ␤-sheet structures, often generated by partial denaturation of proteins (28, 50 -53). On the basis of the results presented here, whatever the structures promoting coacervation of elastin-like polypeptides, these structures must not contain regions of extended ␤-sheet. Moreover, the presence of proline residues in hydrophobic domains of elastin may be an important factor in promoting structures that result in coacervation and preventing the formation of structures that allow amyloid-type aggregation. Indeed, it has been reported that introduction of proline residues into amyloidogenic polypeptides inhibits their tendency for amyloid formation (40).
Interestingly, Kozel et al. (30) have recently reported that a polypeptide consisting of exon 30 alone, which contains relatively few proline residues, formed amyloid-like fibrils, although such a tendency was not seen for exon 30 in the context of our polypeptide sequences (EP20- [24][25][26][27][28][29][30]. Although the importance of proline residues in allowing interactions resulting in coacervation rather than amyloid formation seems clear, this is not entirely consistent with our observation that EP20-24-LRS, the polypeptide containing as one of the hydrophobic domains the lamprin repeat sequence (GGLGY), which lacks proline residues, not only did not show a tendency for amyloid formation but also had the greatest propensity for coacervation among all polypeptides studied here. Similarly, although substitution of alanines for tyrosines in the LRS (EP20-24-LRS[Y/A]) raised the coacervation temperature substantially, there was no indication of amyloid formation by this polypeptide. A better understanding of the basis of such inconsistencies must await more detailed structural information on these polypeptides.
Self-aggregation of tropoelastin through coacervation is an ordering process that contributes to the formation of stable fibrillar structures. The fact that such ordered self-assembly can be mimicked by relatively small, well defined polypeptides modeled after the alternating domain structure of native elastin makes these polypeptides a useful tool for understanding the self-assembly process. Insight into the sequence and structural constraints that promote and modulate coacervation will be important for the understanding of elastin assembly itself and may provide information for the design of other self-assembling polymeric proteins with useful physical properties.