Bovine elastin and kappa-elastin secondary structure determination by optical spectroscopies.

Elastin is the macromolecular polymer of tropoelastin molecules responsible for the elastic properties of tissues. The understanding of its specific elasticity is uncertain because its structure is still unknown. Here, we report the first experimental quantitative determination of bovine elastin secondary structures as well as those of its corresponding soluble κ-elastin. Using circular dichroism and Fourier transform infrared and near infrared Fourier transform Raman spectroscopic data, we estimated the secondary structure contents of elastin to be ∼10% α-helices, ∼45% β-sheets, and ∼45% undefined conformations. These values were very close to those we had previously determined for the free monomeric tropoelastin molecule, suggesting thus that elastin would be constituted of a closely packed assembly of globular β structural class tropoelastin molecules cross-linked to form the elastic network (liquid drop model of elastin architecture). The presence of a strong hydration shell is demonstrated for elastin, and its possible contribution to elasticity is discussed.

Elastin is the macromolecular polymer of tropoelastin molecules responsible for the elastic properties of tissues. The understanding of its specific elasticity is uncertain because its structure is still unknown. Here, we report the first experimental quantitative determination of bovine elastin secondary structures as well as those of its corresponding soluble -elastin. Using circular dichroism and Fourier transform infrared and near infrared Fourier transform Raman spectroscopic data, we estimated the secondary structure contents of elastin to be ϳ10% ␣-helices, ϳ45% ␤-sheets, and ϳ45% undefined conformations. These values were very close to those we had previously determined for the free monomeric tropoelastin molecule, suggesting thus that elastin would be constituted of a closely packed assembly of globular ␤ structural class tropoelastin molecules crosslinked to form the elastic network (liquid drop model of elastin architecture). The presence of a strong hydration shell is demonstrated for elastin, and its possible contribution to elasticity is discussed.
The elasticity required for the appropriate functioning of skin, lung, and large blood vessels is due to the presence of elastic fibers within their extracellular matrix (1). The predominant component of these complex structures is the elastin protein, which endows them with its characteristic property of elastic recoil. Elastin is a macropolymeric protein synthesized by mesenchymal cells as a soluble precursor, tropoelastin, whose primary transcript undergoes alternative splicing resulting in the translation of several protein isoforms (2,3). After release in the extracellular space, most of the lysyl residues of tropoelastin are enzymatically deaminated. Following a series of non-enzymatic reactions, the activated residues condense to form specific tetrafunctional cross-links, named desmosines, which appearance allows the spreading of the elastic network within the microfibrillar component of the fiber (for a review, see Ref. 1).
The primary structure of BTE 1 (3,4) consists of an alternance of cross-linking regions, where the lysyl residues are located, and of large hydrophobic domains responsible for elastin elasticity. The highly hydrophobic BTE molecule possesses a very basic C-terminal sequence where its only two Cys residues are located. Those were recently shown to form an intrachain disulfide bridge stabilizing an hydrophilic pocket (5). This C-terminal feature seems involved in elastin fiber assembly (6).
The presence of numerous cross-links and the extreme hydrophobicity of BTE chains are responsible for the great resistance of polymeric BE as well as its total insolubility in any solvent (1). BE-K is the heterogeneous mixture of peptides obtained from BE when it is solubilized by KOH (7). It is a form more suitable for biological tests than BE, as it is soluble. BE-K is thought to be a good model of insoluble BE because of its ability to form a matrix (coacervate) akin to hydrated insoluble BE, the elastic form of BE, at physiological temperatures and high concentrations (7).
The elasticity of BE has an entropic nature (8). However, its exact origin remains uncertain, as the results gathered about BE and BE-K structures are few. Indeed, the very peculiar physico-chemical properties of these molecules do not allow significant structural results using the classical physical investigation methods. This lack of structural data explains why BE molecular models forward an explanation of the elastic mechanism without knowledge of its structures. A description of BE conformation is urgently needed.
Among the various models proposed (see Ref. 1 for a review), only three are still discussed. (a) In the globular liquid drop model of Weis-Fogh and Andersen (9, 10), BE is described as an aggregate of tropoelastin globules. Elasticity originates from hydrophobic interactions at the protein solvent interface of the globules as they deform during stretching. (b) The random network (8) regards BE as a protein devoid of any organization, much as rubber. It is connected to a classical elasticity theory. This model is supported by the works of Tamburro and coworker (11,12) who have established that peptides found in BTE sequence form transient ␤-turns, which stability is influenced by both the surrounding water (13,14) and the length of the peptide (15). (c) In the fibrillar ␤-spiral model of Urry (16,17), BE is considered as a regular arrangement of consecutive ␤-turns (␤-spiral). In this context, elasticity arises from librational motions at the level of the spiral ␤-turns.
Following predictive (18) and experimental (19) evidences, we have proposed a ␤-class molecular model for BTE. The present work reports the structural investigation of the polymeric forms of this molecule, BE and BE-K, using CD, FT-IR, and NIR FT-R spectroscopies. The first estimations of BE and BE-K secondary structure contents are presented. The numerical values obtained for the polymers are compared to those formerly determined for monomeric BTE. Their consequences toward BE existing models and the possible elasticity mechanisms are discussed.

MATERIALS AND METHODS
BE Purification-A fresh bovine ligamentum nuchae was collected at the local slaughterhouse. BE was purified according to the sequential method previously described (7,20). Prior to collagen removal, the Clostridium histolyticum collagenase (Sigma, type VII, clostridiopeptidase A, EC 3.3.24.3) was purified by affinity chromatography on a column of BE, as suggested earlier (20).
Amino Acid Analysis-Samples were hydrolyzed, in vacuo, at 110°C for 24 h in 6 N HCl and analyzed by high performance liquid chromatography using a Waters PICO TAG TM amino acid analysis system equipped with a reverse-phase C 18 PICO TAG column.
BE-K Preparation-Solubilization of elastin was achieved using 1 M KOH in 80% aqueous ethanol as described previously (7).
CD Spectroscopy-BE-K was used at 0.5 mg/ml in distilled water to avoid coacervation. The spectrum was measured at 21°C in 0.1-cm path length cells from 260 to 190 nm with a Mark III dichrograph (Jobin Yvon). Data are expressed as mean ellipticity per residue [] r . The residue mean molecular mass used was 85.3 Da as derived from the BTE sequence (4). The secondary structure contents were determined according to the method of Provencher and Glöckner (21) using their CONTIN program.
FT-IR Spectroscopy-KBr pellets of BE and BE-K samples were prepared with 1 mg of protein per 100 mg of KBr. The spectra were recorded on a BRUCKER IFS 48 spectrometer by the accumulation of 200 interferograms with a 4-cm Ϫ1 resolution. The conformation-sensitive amide I, II, and III bands were checked for secondary structure variations, while amide A bands were used to investigate the N-H shieldings of the peptidic bonds (22).
NIR FT-R Spectroscopy-The spectra of BE and BE-K in powder were recorded at room temperature on a BRUCKER FRA 106 system coupled to a IFS 88 spectrometer in the frequency range 200-4000 cm Ϫ1 with 4-cm Ϫ1 resolution. The infrared laser excitation line was 1.06 m. Its power was 300 mW. The signal-to-noise ratio was improved by the accumulation of 200 interferograms. Secondary structures were determined by decomposition of the amide I band (CϭO stretching mode of the peptidic bond, 1630 -1700 cm Ϫ1 ) into individual components assigned to substructures, as the Raman sensitivity of that band to conformation is well known (23)(24)(25). First, Fourier self-deconvolution (26), second derivative (27), and maximum entropy (28) methods were independently applied to the original amide I bands. Second, among the components yielded by the resolution enhancement methods, only the positions of the most conserved and prominent ones were used as input parameters for a least square curve-fit procedure. No parameters were fixed during the calculation except the nature of the underlying profiles, which were assumed to be 80% gaussian and 20% lorentzian. The structural assignments of the computed components were made according to both their positions before and after reconstruction (23)(24)(25). The cumulated fractional area contribution assigned to a given substructure represented its relative total content in the protein conformation. The enhanced profiles were computed by the SPOV program (developed in Ovtchinnikov and Shemyakin Institute in Moscow). The decompositions were made using the CURVEFIT module of the LabCalc package (Galactics Industries).
For side chains such as alanine, valine, leucine, and isoleucine, the most prominent frequencies are those associated with the CH 2 bending mode found at 1465 Ϯ 20 cm Ϫ1 (29) and with the CH 3 antisymmetric deformation mode found at 1450 Ϯ 20 cm Ϫ1 (22). The behavior of the band centered around 940 cm Ϫ1 has been studied, as it is characteristic of the ordered ␣-helices as shown for poly-L-lysine used as a model (30,31). Bands arising from aromatic (tyrosine, phenylalanine) and sulfur (cysteine, methionine) residues and from the polypeptide backbone were readily identified (32,33).

RESULTS AND DISCUSSION
The amino acid composition of BE (Table I) was in good agreement with those obtained by others (7,20). The preparation was free of collagen as the level of Hyp residues was low, and no hydroxylysine was detected. Likewise, the presence of Asp and Glu residues with values that compared to BTE Asp and Gln ones, respectively, demonstrated the absence of microfibrillar proteins. Fundamentally, no Trp nor His was detected, and the estimated quantities of Gly, Ala, Pro ϩ Hyp, Val, Leu, and Ile residues compared well with those of BTE composition (Table I). The main discrepancy between elastin and BTE compositions was the estimated number of lysyl residues. This arose from the great difficulty in detection of all elastin crosslinking amino acids. With the occurrence of one Met residue being below the technique precision, the composition of BE (Table I) indicated a high level of purity, allowing its solubilization and the use of optical spectroscopic methods to analyze its structures.
The main features observed in the FT-IR spectra of proteins are those associated with the planar peptidic bond vibrational modes, the so-called amide bands, which positions, widths, and intensities are characteristic of the vibrational modes associated and thus of the local geometry of the peptidic chain. They are the amide I (CϭO stretching), amide II (mainly C-N  stretching), amide III (N-H in plane deformation), and amide A (N-H stretching) bands; the first three are very sensitive to conformational changes (34,35), while the last one brings information about the hydrogen bondings undergone by the peptidic N-H groups (22). The FT-IR spectra of our samples in KBr pellets (Fig. 1) compared well with the data obtained by others for insoluble elastin (36) and solubilized elastin (37). The two molecules shared close global conformations as their structure-sensitive amide I, II, and III bands were found at comparable positions (amide I at 1659 and 1657 cm Ϫ1 , amide II at 1538 and 1542 cm Ϫ1 , and amide III at 1237 and 1238 cm Ϫ1 , for BE and BE-K, respectively). However, BE amide I band was much broader than that of BE-K, underlining some structural differences. The occurrence of amide A bands at different positions (3322 and 3307 cm Ϫ1 , respectively) also demonstrated that their peptidic N-H groups were involved in different types of hydrogen bondings.
Laser-visible Raman spectroscopy is a very powerful technique to investigate the conformation of biological molecules, as it can provide information about the secondary structures, the microenvironment of the residues, and the polypeptidic backbone geometry. However, BE is so highly fluorescent that convenient standard Raman data could not be reached (38,39). Thus, we have preferred NIR FT-R spectroscopy instead of normal visible Raman, as the use of an infrared source was less likely to excite the intense protein autofluorescence.
The characteristic bands of a protein Raman spectrum were observed in the BE spectrum (Fig. 2) as follows: 1) the conformationally sensitive amide I (CϭO stretch, 1630 -1700 cm Ϫ1 ) and amide III bands (N-H in plane deformation, 1230 -1310 cm Ϫ1 ), which arise from the Raman-active vibrational modes of the planar CONH peptidic bond; 2) bands assigned to residue side chains like aromatic cycles (only Phe and Tyr in the present case), CH, CH 2 , and CH 3 groups, and stretching of bonds containing sulfur atoms; and 3) bands corresponding to the C ␣ -C and C ␣ -N stretches of the polypeptidic backbone. The existence of disulfide bridges in BE was clearly demonstrated by the occurrence of two bands assigned to S-S (527 cm Ϫ1 ) and C-S (665 cm Ϫ1 ) stretching modes, respectively. These should mainly possess a local gauche-gauche-trans-geometry, as the S-S mode was observed at 525 Ϯ 10 cm Ϫ1 (32). They certainly arose from BTE intrachain bondings. The NIR FT-R spectrum of BE-K, in contrast to the BE spectrum, showed a relatively poor signal-to-noise ratio reflecting the very heterogenous nature of the solubilized elastin (data not shown). Nevertheless, its amide I band was clear enough for structural analysis.
Our analysis of the Raman data has mainly been focused on secondary structure quantitation. The amide I band originates from the CϭO stretching modes of all the peptidic bonds of the protein. Depending upon the particular secondary structure, a CϭO group is involved in a given type of hydrogen bonding, whose characteristics influence its frequency of vibration. That way, all the CϭO occurring in ␣-helices will vibrate similarly but differently from those encountered in ␤-sheets. Likewise, regular ␤-sheets and irregular ones will yield different signals. The vibrational frequencies from one substructure to another are not very different but sufficient to be distinguished (22). They all occur in the same characteristic spectral range (1630 -1700 cm Ϫ1 ), and their respective Raman signals overlap to yield the complex amide I band. Decomposition methods aim at directly accessing those structural contributions that overlap. In this manner, it is thereafter possible to determine the secondary  Table II and were used for quantitation (Table III). structure contents of the molecule by the standard assumption that their respective areas correspond to their conformational contributions (23)(24)(25). The mathematical solution to a given decomposition problem is never unique, and it is always very difficult to tell which solution is the best, as one has no idea of how many contributions really exist and where they are located. But, fortunately, the use of more than only one resolution enhancement method permits assessment of these parameters. Here, we have used the three major ones. Each of them proposed several underlying contributions in our amide I profiles (data not shown). By comparison and correlation between their results, we were able to choose a small representative number of components whose positions could be accurately estimated. The calculated components centered in 1630 -1700-cm Ϫ1 range were assigned to ␣-helices, ␤-strands, and undefined (turns ϩ coils) secondary structure elements according to both theoretical and experimental results (22)(23)(24)(25).
A component near 1640 cm Ϫ1 corresponding to the hydration water bending mode (32) was resolved in both decompositions (Fig. 3, Table II) as the molecules were in the solid state. This result was in good agreement with the paradoxical demonstration that the very hydrophobic and insoluble BE formed very tight hydrogen bondings with water (40). One helix, two ␤-strands, and two undefined components were evidenced for either BE (Fig. 3a) or BE-K (Fig. 3b). The occurrence of an ␣-helical component correlated well with the observation of a band centered around 934 cm Ϫ1 in the BE spectrum (Fig. 2), as that feature is characteristic of ordered helices (30,31). Moreover, the presence of ␤-strands was confirmed by the position of BE amide III band (1248 cm Ϫ1 ) since it fell within the typical Raman amide III domain of ␤-structures (24).
The quantitative results compiled in Table III showed that BE and BE-K possessed similar global conformations, which were consistent with high levels of both extended (43 and 46%, respectively) and unordered (48 and 41%, respectively) structures. However, the analysis of their respective NIR FT-R amide I components (Fig. 3, Table II) indicated strong variations in their local structures. For example, the first undefined structure component (1661 cm Ϫ1 for BE and 1656 cm Ϫ1 for BE-K; see Table II) accounted for 45% of the global structure before solubilization and 12% after. Meanwhile, the second undefined structure component (1672 cm Ϫ1 for BE and 1666 cm Ϫ1 for BE-K; see Table II) varied from 3 to 29%. So, an inversion of population between the modes giving rise to those components had occurred, and the conditions of hydrogen bondings were different between the two molecules as revealed by the FT-IR data analysis. Insoluble and soluble elastins had quantitatively but not qualitatively identical conformations. This observation strongly suggests that BE-K is probably not a good model for BE, as far as local conformations are concerned.
The CD spectrum of BE-K in water (Fig. 4) was in good agreement with those recorded previously by others for soluble elastins (41,42). The broad negative band observed at 200 nm tended to support the view that BE-K was disordered. However, this spectral feature could be assigned to short and distorted ␤-sheets (43) as was the case for the BTE spectrum (19). The CD quantitation (Table III) confirmed this possibility, as a high level of ␤-structures was estimated for BE-K disolved in water. The value (47%) compared well with that estimated for BE-K in the solid state (46%). Nevertheless, the structural contents of BE-K seemed to change when it was dissolved (Table III). The most striking feature was the apparent disappearance of ␣-helical structures. This was quite surprising, as BTE helices (ϳ5%) were preserved in solution (19). A hydration effect upon BE-K helices remained possible but uncertain all the more so since the precision of the quantitative methods used was Ϯ5%.
The conformation contents of BE (Table III) were in very good agreement with those of our free BTE ␤-class molecular model (19). For BTE, they were 5% of helices (cross-linking domains), 50% of ␤-strands, and 45% of undefined conformations (rest of the molecule or elastic regions). The finding that TABLE II Spectral parameters and assignments of elastins NIR FT-R amide I components i is the estimated initial frequency in cm Ϫ1 , f the calculated frequency in cm Ϫ1 , and is the computed width at half-height in cm Ϫ1 . Assignments were made according to i and f values following experimental and theoretical results (22)(23)(24)(25). For components assigned to secondary structure elements, S represents the respective fractional area contribution in percent rounded to the nearest integer. For a given substructure, the sum of the S values was assumed to represent its total content in the molecule conformation (see Table III the global structures of free and cross-linked BTE were significantly identical strongly suggested that our monomeric model could apply to the elastin polymer, meaning that BE would be constituted of a three-dimensional arrangement of globular BTE molecules connected by cross-links. This structural definition typically corresponded to the liquid drop model of BE (9,10). This model was also greatly supported by recent scanning tunneling microscopy observations of reconstituted BE (44) and human recombinant tropoelastin coacervates. 2 Our structural results clearly contradicted the random network (8) and ␤-spiral (16, 17) models of BE architecture, as BE did possess high levels of ordered structures. The structure of BE could thus be described as a threedimensional repetition of our molecular model of BTE (19), which is to say that ␤-class BTE molecules are closely packed together and cross-linked by helical domains (ϳ10% ␣-helices), while the entropic "elastic" regions would consist mainly of buried short and/or distorted antiparallel ␤-strands (ϳ45%), which are probably packed in ␤-barrels and alternate with external turns and coil substructures (ϳ45% for the sum). In addition, we would like to underline that the hydrophobic domains of BE are highly mobile (45,46) and that coil-turn (11,12,15) or sheet-coil-turn (18) conformational transitions are possible. These transitions are most certainly mediated by the hydration water molecules (13,14,18).
The present work reports the first experimental estimation of the secondary structures of insoluble bovine elastin. Conclusions about the tertiary and quaternary structures of the elastomer have also been reached. Our results provide valuable information for understanding the elastic function of BE as they demonstrate that the structure-elasticity relationships must be envisaged in a liquid drop architecture context (9, 10).
Nevertheless, we point out that our results do not mean that the elasticity mechanism (hydrophobic interactions) proposed in 1970 by Weis-Fogh and Andersen (9, 10) is correct. We only agree with the architecture they have proposed for the molecule. Indeed, we suggest that their explanation of elasticity is incorrect, as it is based on a diphasic description of the swollen polymer (protein chains ϩ water) and neglects the hydration water of the molecule.
Recently, water solvent was shown to act as a plasticizer for elastin (47), that is to say it enhances its mobility. Moreover, the action of solutes on the structure of elastin is indirect and seems to be mediated through its hydration shell (48). That way, if solvent water molecules are considered as particular solutes, their plasticizing effect should be processed through the hydration shell of the molecule. We thus feel that the strong hydration shell demonstrated for elastin could have some functional significance. Swollen elastin would then be better described as a triphasic system, protein ϩ hydration water ϩ solvent water, and, in this view, the conformational transitions we and others have proposed for BE hydrophobic domains (11,12,15,18) should have a functional role. The elasticity theory connected to this proposal and the molecular events occurring during stretching or relaxation need now to be completely described. Further experiments in this way are in hand as well as molecular modelings of isolated and/or cross-linked tropoelastins.