Structural Analysis of Nonapeptides Derived from Elastin

Abstract

Elastin-derived peptides are released from the extracellular matrix remodeling by numerous proteases and seem to regulate many biological processes, notably cancer progression. The canonical elastin peptide is VGVAPG, which harbors the XGXXPG consensus pattern, allowing interaction with the elastin receptor complex located at the surface of cells. Besides these elastokines, another class of peptides has been identified. This group of bioactive elastin peptides presents the XGXPGXGXG consensus sequence, but the reason for their bioactivity remains unexplained. To better understand their nature and structure-function relationships, herein we searched the current databases for this nonapeptide motif and observed that the XGXPGXGXG elastin peptides define a specific group of tandemly repeated patterns. Further, we focused on four tandemly repeated human elastin nonapeptides, i.e., AGIPGLGVG, VGVPGLGVG, AGVPGLGVG, and AGVPGFGAG. These peptides were analyzed by means of optical spectroscopies and molecular dynamics. Ultraviolet-circular dichroism and Raman spectra are consistent with a mixture of β-turn, β-strand, and random-chain secondary elements in aqueous media. Quantitative analysis of their conformations suggested that turns corresponded to half of the total population of structural elements, whereas the remaining half were equally distributed between β-strand and unordered chains. These distributions were confirmed by molecular dynamics simulations. Altogether, our data suggest that these highly dynamic peptides harbor a type II β-turn located in their central part. We hypothesize that this structural element could explain their specific bioactivity.

Significance

Elastin fragmentation products, the so-called elastin peptides, may exhibit a bioactivity toward normal and tumor cells. This phenomenon depends on the sequence motif they harbor. Although XGXXPG sequence bioactivity is explained by the presence of a type VIII β-turn allowing interaction with the elastin receptor complex, the structural reasons for XGXPGXGXG-specific activity remain unexplained. Using data mining, we show that elastin nonapeptides define a specific class of tandemly repeated features. Further, spectroscopic and numerical simulations methods suggest the presence of a type II β-turn in their conformation. This structural element could explain their bioactivity.

Introduction

Elastin is the extracellular matrix protein responsible for the structural integrity and function of tissues undergoing reversible extensibility or deformability (1). This protein is extremely stable and resistant and undergoes virtually no turnover. Nevertheless, during aging, mechanical stress and elastase activities contribute to the fragmentation of this macropolymer into elastin-derived peptides (EDPs) (2).

Elastin is characterized by its elasticity and seems devoid of any other biological activity (1). In contrast, EDPs have been shown to regulate numerous biological processes and are thought to be involved in several aging-related pathologies such as atherosclerosis (3,4) and cancer (5, 6, 7, 8). Most biologically active EDPs, i.e., elastokines, possess a GXXPG consensus sequence adopting a type VIII β-turn structure (9) involved in their binding to the elastin receptor complex (10). The interaction of GXXPG-containing elastokines with the elastin receptor complex and the consequences of their interaction have been considerably documented during the last decades. For a comprehensive review on this topic, the reader is referred to (11).

Besides the classical GXXPG-containing sequences, another class of elastokines has also been reported. In 1988, Long and colleagues reported that nonapeptide sequences from elastin were chemoattractant for fibroblasts (12). In 1989, they further demonstrated that this biological activity could be extended to endothelial cells (13). Strikingly, since then, this peculiar class of EDP has been mostly ignored until 2007, when Maeda and colleagues demonstrated that these peptides could promote macrophages migration via a specific, but unknown, receptor (14). More recently, their biological activity has been further linked to lung carcinoma progression (8) and to promotion of invasion in triple-negative cancer cells via MMP-14 and MMP-2 (15).

The relatively low interest encountered by these nonapeptide EDPs can be explained by the fact that considerable advances have been made on both GXXPG-containing EDPs and their receptor biology. As a consequence, few groups considered investigating on these peculiar peptides.

The nonapeptide AGVPGLGVG and other EDPs share similar structural features: random-coil and β-turn conformations. However, AGVPGLGVG induces angiogenesis, tumor progression, and secretion of proteases higher than other EDPs in carcinomas. Moreover, in vitro, it behaves like amyloid-like peptides harboring the XGGXG motif by forming cross β-sheets at the supramolecular level (16). Very recently, Brassart and colleagues have shown that the receptor mediating AGVPGLGVG biological effects and identified as lactose-insensitive is the ribosomal protein SA (RPSA) (17).

The aim of this work is to analyze the structure and conformational behavior of these peptides. Our structural analysis was conducted using sequence analysis, spectroscopic methods, and molecular dynamics simulations.

Our results show that elastin nonapeptides define a peculiar class among peptides harboring the consensus XGXPGXGXG sequence. In elastin, these peptides are mostly tandemly repeated. Furthermore, our data show that they are engaged in a conformational equilibrium between random-coil and turn conformations at the considered concentrations and temperatures. Molecular dynamics simulations suggest that the dominant conformation of these peptides is a type II β-turn. The functional significance of this conformation is discussed.

Materials and Methods

Sequence analysis

Sequence retrieval

Sequences harboring the XGXPGXGXG consensus motif were searched using Python scripts on UniProtKB (Swiss-Prot and TrEMBL) database Fasta files in which spliced variants were ignored. Only exact matches were retrieved and further cleaned informatically and manually so as to ensure that a sequence belonging to a given species (identified by its unique accession number) was present only once in our results. For each sequence hit, the collected data were its accession number, the originating organism, the length of the protein, and the position(s) of the nonapeptide sequence(s).

Pattern analysis

Nonapeptide exact matches were classified and distributed as a function of the taxonomic hierarchical classification lineage of the source organism of the corresponding parent protein. Hits were either individual hits, i.e., only one occurrence was found in the sequence, or multiple ones. In the latter case, we distinguished three possibilities. Multiple hits could be found in different regions of the protein, and each occurrence was isolated from the others. Another possibility was that the repetitions could overlap, i.e., the end of the first occurrence (-X-G) was the beginning of the following one (X-G-). Finally, we observed that nonapeptide sequences could also be tandemly repeated, one after the other. The tree figure has been made with the Python module ETE3 (18) and the Logo sequences with WebLogo (https://github.com/weblogo/weblogo) (19).

Optical spectroscopy

Molecular compounds

Powder samples of amino acids were purchased from Sigma-Aldrich (Saint Quentin Fallavier, France). Lyophilized samples of nonapeptides were obtained from Genecust (Luxemburg, Luxemburg) as zwitterionic peptides in acetate salts. Their purity was at least 95% as determined by mass spectrometry.

Solution samples

Lyophilized powder of amino acids and peptides was dissolved in pure water from a Millipore filtration system (Guyancourt, France). The ionic strength of all peptide samples was increased by adding 150 mM NaCl to stock solutions. Upon dissolution, the pH was between 4 and 4.5. Raman spectra were recorded between 10 and 80°C. Circular dichroism (CD) spectra were obtained at room temperature between 100 and 200 μM.

CD spectroscopy

Room-temperature ultraviolet-CD spectra were analyzed on a JASCO J-810 spectrophotometer (Lisses, France) within the 190–300 nm spectral region. A pathlength of 1 mm and a spectral resolution of 0.2 nm were selected. Each spectrum corresponding to an average of five scans was recorded with a speed of 100 nm per minute. The measured ellipticity for each sample, referred to as [ϕ]_obs, was further normalized to obtain the so-called mean residue ellipticity, [ϕ], by using the expression [ϕ] = [ϕ]_obs/ncl, where n, c, and l are the length of the peptide, its molar concentration, and the optical pathlength, respectively. The normalized ellipticity was expressed in deg cm² dmol⁻¹.

Raman spectroscopy

Room-temperature Stokes Raman spectra were analyzed in bulk samples at right angle on a Jobin-Yvon T64000 spectrometer (Longjumeau, France) at single spectrograph configuration, 1200 grooves/mm holographic grating, and a holographic notch filter. Raman data corresponding to 1200 s acquisition time for each spectrum were collected on a liquid-nitrogen-cooled CCD detection system (Spectrum One, Jobin-Yvon). The effective slit width was set to 5 cm⁻¹. Solution samples were excited by the 488 nm line of an Ar⁺ laser (Spectra Physics, Evry, France), with 200 mW power at the sample.

Post recording spectroscopic data treatment

Buffer subtraction and smoothing of the observed spectra were performed using the GRAMS/AI Z.00 package (Thermo Galactic, Waltham, MA). Final presentation of Raman spectra was done by means of SigmaPlot package 6.10 (SPSS, Chicago, IL).

Secondary structure analysis

Quantification of protein secondary structures using amide I profile decomposition is a method used commonly, and its intrinsic limits are readily known (20,21). For elastin material, its overall accuracy is ±5% as far as the final structural estimates are concerned (22). The initial step of the reconstruction is to guess how many underlying bands are present in the complex amide I profile to be analyzed. This is achieved using a second derivative analysis of the smoothed spectrum. For our data, four components were suggested. This information was provided to the curve fitting software to reconstruct the experimental spectrum using Gaussian/Lorentzian mixture profiles. The curve fitting procedure was successful when the fit was good (low χ²-value), the number of components was unchanged, and their width at half height was consistent. The secondary structure estimates were obtained by adding the fractional areas of amide I structural components belonging to the same secondary structure group.

Atomistic molecular dynamics simulations

Ten independent simulations of 500 ns were performed for each of the four human elastin nonapeptides using the amber99SB-ILDN force field (23) and Gromacs 2016.5 software (24). Peptides were built in an extended conformation with PyMOL (25) and solvated with TIP3P water in a 5 nm cubic box. All studied systems were first minimized by steepest descent for 5000 steps. Then, a 1 ns simulation with the peptides under position restraints was run before the production simulations were performed. Different seeds were used for each simulation during the velocity assignments. Periodic boundary conditions were used with a 2 fs time step. The dynamics were carried out under NPT conditions (310 K and 1 bar). The temperature was maintained using the v-rescale method (26), with τT = 0.1 ps, and an isotropic pressure was maintained using the Parrinello-Rahman barostat (27) with a compressibility of 4.5 × 10⁻⁵ bar⁻¹ and τP = 2 ps. Short-range nonbonded interactions were treated with a cutoff of 1.0 nm, and long-range interactions were calculated with the particle mesh Ewald method (28) with a grid spacing of 0.16 nm. Bond lengths were maintained with the LINCS algorithm (29) and long-range dispersion corrections for energy and pressure were applied. The 3D structures were analyzed with both PyMOL and VMD (30) softwares.

The secondary structures were computed with DSSP (31), and the different types of β-turns have been assigned based on the ΦΨ angles of two residues following each other as defined by Lewis et al. (32) and Hutchinson et al. (33). Distances between each i and i + 3 Cα have been computed and, if this distance was less than 0.7 nm and the amino acids were not part of a helix conformation, the ΦΨ angles of residues i + 1 and i + 2 were considered. To assign a peculiar type of β-turn, Hutchinson and colleagues (33) considered a range of ±30 or ±40° around the standard angles with a deviation of 50° allowed for one of the angles.

The distribution of ΦΨ angles of residues i + 1 and i + 2 we observed during the simulation permitted us to easily identify type I, I′, II, II′, and VIII β-turn populations as broad plot regions (Fig. S1). Because these regions were larger than those proposed by Hutchinson and colleagues (33), we adapted the considered ΦΨ angle ranges to encompass the ranges we observed. The main effect of this change was the reduction of unassigned turn contribution, i.e., type IV β-turns. The corresponding figure and ΦΨ angle ranges for type I, I′, II, II′, and VIII β-turns are provided as Supporting Material (Fig. S1; Table S1). As in Hutchinson, other β-turn types are considered with a range of ±40° around the standard angles, and type IV β-turns correspond to otherwise unassigned β-turns. Clustering was performed using the gromos method (34) with a cutoff of 0.1 nm on Cα atoms of residues 2–6.

To assess the quality of our simulations, CD spectra have been computed from the peptide trajectories by using the SESCA software (35) with the basis set HBSS-3SC1.

Results

Sequence analysis

The retrieval of sequences possessing XGXPGXGXG yielded 96,293 unique sequences. These sequences (Table S2) were classed as a function of their taxonomy (Fig. 1). Most sequences were from Bacteria (71,504 sequences for 75,718 occurrences with 2129 overlaps and 297 tandem repeats), then Eukariota (22,919 sequences for 24,828 occurrences with 811 overlaps and 297 tandem repeats), Archae (1630 sequences for 1694 occurrences with 16 overlaps and six tandem repeats), and Virus (240 sequences for 245 occurrences with no overlaps and no tandem repeats). Strikingly, when these results are observed, it becomes clear that elastins define a specific group. They exhibit numerous occurrences per sequences (171 for 38) with no overlap and a very high ratio of tandem repeats. This specificity of elastin nonapeptides is even more evident when the most frequent residues occurring at X positions are considered. Although other groups mostly favor the presence of G, A, K, P, or L at these positions, elastins allow almost none of those, with the AGVPGFGVG being the dominating motif. In fact, out of the 140 million sequences searched, this motif was always found in elastin sequences except for one bacterial sequence.

Figure 1.

Nonapeptide occurrences and distribution. The XGXPGXGXG consensus was found in 96,293 unique sequences, among which 22,919 were from Eukariota, 71,504 from Bacteria, 240 from Virus, and 1630 from Archae. The Eukariota branch distribution is dominated by Animals, in which Vertebrates are predominant. For each branch of the tree, the total number of hits is reported as well as the number of observed overlapping (O) sequences and of tandem (T) repetitions. The consensus motif is also presented with the most frequent residues found at the four X positions.

The nonapeptide sequences found in elastin entries are reported in Table 1, in which they are compared. The sequences are reported as a function of their alphabetical order. In human elastin, nonapeptides can be observed in exon domains 18, 20, or 26. Single occurrences are observed in the two first exons, whereas they appear almost exclusively as tandem repeats in exon 26. Strikingly, the human repeats can also be observed in other species. For instance, the human elastin motif AGVPGFGVG is also present in mouse exon 26. Overall, human exon 26 (or its equivalent in other species) is a sequence domain characterized by tandem repeats of these nonapeptides. As a consequence, those are supposedly important for the biological function of the polymer, and we decided to further analyze the structural behavior of the four human exon 26 nonapeptides, i.e., AGIPGLGVG (hN3), VGVPGLGVG (hN4), AGVPGLGVG (hN5), and AGVPGFGAG (hN6).

Table 1.

Nonapeptide Stack of Elastin Uniprot Entries

Species	Accession Number	#	Sequence	Position	Exon
Oreochromis niloticus	I3JQ46	8	AGGPGSGIG	657–665	18
Odobenus rosmarus divergens	A0A2U3ZFC3	5	AGIPGFGAG	541–549	–
Canis lupus familiaris	J9NW15	5	AGIPGFGVG	571–579	23
Homo sapiens∗	P15502	3	AGIPGLGVG	576–584	26
Rhinopithecus roxellana	A0A2K6NWE2	3	AGIPGLGVG	562–570	26
Pan troglodytes	A0A2I3SVF0	3	AGIPGLGVG	580–588	27
Macaca mulatta	A0A1D5R663	2	AGIPGVGPG	312–320	15
Papio anubis	A0A2I3MW49	3	AGVPGFGAG	472–480	16
Cercocebus atys	A0A2K5LP54	3	AGVPGFGAG	512–520	18
Felis catus	A0A337S1Z3	4	AGVPGFGAG	404–412	18
Ailuropoda melanoleuca	G1MA94	4	AGVPGFGAG	499–507	23
Canis lupus familiaris	J9NW15	6	AGVPGFGAG	580–588	23
Myotis lucifugus	G1QCG1	2	AGVPGFGAG	483–491	24
Myotis lucifugus	G1QCG1	3	AGVPGFGAG	492–500	24
Myotis lucifugus	G1QCG1	4	AGVPGFGAG	501–509	24
Sus scrofa	A0A287AXU0	5	AGVPGFGAG	599–607	25
Homo sapiens∗	P15502	6	AGVPGFGAG	603–611	26
Rattus norvegicus	Q99372	3	AGVPGFGAG	642–650	26
Rattus norvegicus	Q99372	4	AGVPGFGAG	658–666	26
Mus musculus	P54320	2	AGVPGFGAG	613–621	27
Mus musculus	P54320	3	AGVPGFGAG	622–630	27
Mus musculus	P54320	4	AGVPGFGAG	631–639	27
Mus musculus	P54320	5	AGVPGFGAG	640–648	27
Mus musculus	P54320	6	AGVPGFGAG	649–657	27
Mustela putorius furo	M3YXB8	4	AGVPGFGAG	608–616	27
Pan troglodytes	A0A2I3SVF0	6	AGVPGFGAG	607–615	27
Erinaceus europaeus	A0A1S3W547	3	AGVPGFGAG	584–592	–
Erinaceus europaeus	A0A1S3W547	4	AGVPGFGAG	593–601	–
Enhydra lutris kenyoni	A0A2Y9J5B3	4	AGVPGFGAG	615–623	–
Tursiops truncatus	A0A2U4A8K1	1	AGVPGFGAG	87–95	–
Ursus maritimus	A0A384CQV9	6	AGVPGFGAG	583–591	–
Mesocricetus auratus	A0A1U8CCM2	2	AGVPGFGAG	643–651	–
Mesocricetus auratus	A0A1U8CCM2	3	AGVPGFGAG	652–660	–
Mesocricetus auratus	A0A1U8CCM2	4	AGVPGFGAG	661–669	–
Mesocricetus auratus	A0A1U8CCM2	5	AGVPGFGAG	670–678	–
Dipodomys ordii	A0A1S3GKU4	6	AGVPGFGAG	587–595	–
Erinaceus europaeus	A0A1S3W547	5	AGVPGFGGG	602–610	–
Balaenoptera acutorostrata scammoni	A0A384AI97	3	AGVPGFGPG	584–592	–
Felis catus	A0A337S1Z3	2	AGVPGFGVG	386–394	18
Felis catus	A0A337S1Z3	3	AGVPGFGVG	395–403	18
Canis lupus familiaris	J9NW15	3	AGVPGFGVG	553–561	23
Canis lupus familiaris	J9NW15	4	AGVPGFGVG	562–570	23
Saimiri boliviensis boliviensis	A0A2K6UXV4	1	AGVPGFGVG	491–499	25
Sus scrofa	A0A287AXU0	3	AGVPGFGVG	581–589	25
Sus scrofa	A0A287AXU0	4	AGVPGFGVG	590–598	25
Callithrix jacchus	A0A2R8P8B8	3	AGVPGFGVG	523–531	26
Mustela putorius furo	M3YXB8	2	AGVPGFGVG	590–598	27
Mustela putorius furo	M3YXB8	3	AGVPGFGVG	599–607	27
Odobenus rosmarus divergens	A0A2U3ZFC3	4	AGVPGFGVG	532–540	–
Enhydra lutris kenyoni	A0A2Y9J5B3	2	AGVPGFGVG	597–605	–
Enhydra lutris kenyoni	A0A2Y9J5B3	3	AGVPGFGVG	606–614	–
Ursus maritimus	A0A384CQV9	4	AGVPGFGVG	565–573	–
Ursus maritimus	A0A384CQV9	5	AGVPGFGVG	574–582	–
Dipodomys ordii	A0A1S3GKU4	4	AGVPGLGAG	569–577	–
Dipodomys ordii	A0A1S3GKU4	5	AGVPGLGAG	578–586	–
Danio rerio	F8W4J7	4	AGVPGLGGG	568–576	18
Ailuropoda melanoleuca	G1MA94	3	AGVPGLGVG	490–498	23
Saimiri boliviensis boliviensis	A0A2K6UXV4	2	AGVPGLGVG	500–508	25
Rhinopithecus bieti	A0A2K6MP70	3	AGVPGLGVG	522–530	25
Bos taurus	P04985	1	AGVPGLGVG	562–570	26
Bos taurus	P04985	2	AGVPGLGVG	571–579	26
Bos taurus	P04985	3	AGVPGLGVG	580–588	26
Homo sapiens∗	P15502	5	AGVPGLGVG	594–602	26
Rattus norvegicus	Q99372	2	AGVPGLGVG	633–641	26
Callithrix jacchus	A0A2R8P8B8	4	AGVPGLGVG	532–540	26
Callithrix jacchus	A0A2R8P8B8	5	AGVPGLGVG	541–549	26
Callithrix jacchus	A0A2R8P8B8	6	AGVPGLGVG	550–558	26
Colobus angolensis palliatus	A0A2K5JW24	3	AGVPGLGVG	548–556	26
Bos mutus	L8IPH2	1	AGVPGLGVG	564–572	26
Macaca nemestrina	A0A2K6BHK1	4	AGVPGLGVG	563–571	26
Pan troglodytes	A0A2I3SVF0	5	AGVPGLGVG	598–606	27
Ursus maritimus	A0A384CQV9	3	AGVPGLGVG	556–564	–
Ficedula albicollis	U3KH66	1	AGVPGVGVG	359–367	15
Equus caballus	F7BWV3	1	FGVPGYGVG	418–426	17
Oryzias latipes	A0A3B3HSK9	1	GGAPGFGPG	383–391	14
Tetraodon nigroviridis	H3CXZ1	1	GGAPGSGPG	373–381	8
Sus scrofa	A0A287AXU0	2	GGFPGFGVG	429–437	19
Homo sapiens∗	P15502	2	GGFPGFGVG	401–409	20
Pan troglodytes	A0A2I3SVF0	2	GGFPGFGVG	405–413	20
Callithrix jacchus	A0A2R8P8B8	2	GGFPGFGVG	408–416	21
Balaenoptera acutorostrata scammoni	A0A384AI97	2	GGFPGFGVG	440–448	–
Erinaceus europaeus	A0A1S3W547	2	GGFPGYGAG	426–434	–
Canis lupus familiaris	J9NW15	2	GGFPGYGIG	399–407	17
Ailuropoda melanoleuca	G1MA94	2	GGFPGYGIG	395–403	19
Odobenus rosmarus divergens	A0A2U3ZFC3	2	GGFPGYGIG	396–404
Enhydra lutris kenyoni	A0A2Y9J5B3	1	GGFPGYGIG	456–464
Ursus maritimus	A0A384CQV9	2	GGFPGYGIG	426–434
Felis catus	A0A337S1Z3	1	GGFPGYGVG	289–297	14
Papio anubis	A0A2I3MW49	1	GGFPGYGVG	338–346	14
Cercocebus atys	A0A2K5LP54	1	GGFPGYGVG	354–362	15
Mandrillus leucophaeus	A0A2K5XQM6	2	GGFPGYGVG	284–292	15
Rattus norvegicus	Q99372	1	GGFPGYGVG	457–465	18
Rhinopithecus bieti	A0A2K6MP70	2	GGFPGYGVG	377–385	20
Mus musculus	P54320	1	GGFPGYGVG	454–462	21
Colobus angolensis palliatus	A0A2K5JW24	2	GGFPGYGVG	405–413	21
Mustela putorius furo	M3YXB8	1	GGFPGYGVG	445–453	21
Rhinopithecus roxellana	A0A2K6NWE2	2	GGFPGYGVG	405–413	21
Macaca nemestrina	A0A2K6BHK1	2	GGFPGYGVG	405–413	21
Mesocricetus auratus	A0A1U8CCM2	1	GGFPGYGVG	466–474	–
Dipodomys ordii	A0A1S3GKU4	3	GGFPGYGVG	405–413	–
Canis lupus familiaris	J9NW15	1	GGGPGAGLG	54–62	2
Danio rerio	F8W4J7	1	GGGPGAGLG	33–41	2
Odobenus rosmarus divergens	A0A2U3ZFC3	1	GGGPGAGLG	71–79	–
Dipodomys ordii	A0A1S3GKU4	1	GGGPGAGLG	81–89	–
Oreochromis niloticus	I3JQ46	1	GGGPGFGGG	541–549	16
Oreochromis niloticus	I3JQ46	2	GGGPGFGGG	559–567	16
Oreochromis niloticus	I3JQ46	3	GGGPGFGGG	577–585	16
Oreochromis niloticus	I3JQ46	4	GGGPGFGGG	589–597	16
Oreochromis niloticus	I3JQ46	9	GGGPGLGLG	900–908	23
Oreochromis niloticus	I3JQ46	7	GGIPGLGYG	632–640	16
Danio rerio	F8W4J7	28	GGIPGVGYG	1190–1198	30
Danio rerio	F8W4J7	13	GGLPGGGAG	821–829	22
Danio rerio	F8W4J7	22	GGLPGGGAG	1033–1041	26
Danio rerio	F8W4J7	23	GGLPGGGAG	1041–1049	26
Danio rerio	F8W4J7	26	GGLPGGGAG	1117–1125	28
Danio rerio	F8W4J7	18	GGLPGGGIG	956–964	24
Danio rerio	F8W4J7	16	GGLPGGGLG	895–903	24
Danio rerio	F8W4J7	20	GGLPGGGLG	1005–1013	26
Danio rerio	F8W4J7	21	GGLPGGGLG	1025–1033	26
Danio rerio	F8W4J7	25	GGLPGGGLG	1090–1098	28
Oreochromis niloticus	I3JQ46	11	GGLPGGGPG	1844–1852	49
Oreochromis niloticus	I3JQ46	10	GGLPGGGTG	932–940	23
Danio rerio	F8W4J7	17	GGLPGGGVG	948–956	24
Danio rerio	F8W4J7	2	GGLPGIGAG	144–152	5
Danio rerio	F8W4J7	19	GGLPGSGIG	964–972	24
Danio rerio	F8W4J7	24	GGLPGSGIG	1049–1057	26
Danio rerio	F8W4J7	9	GGLPGSGLG	721–729	20
Danio rerio	F8W4J7	14	GGLPGSGTG	837–845	22
Danio rerio	F8W4J7	5	GGLPGSGVG	661–669	20
Danio rerio	F8W4J7	6	GGLPGSGVG	676–684	20
Danio rerio	F8W4J7	7	GGLPGSGVG	691–699	20
Danio rerio	F8W4J7	8	GGLPGSGVG	706–714	20
Danio rerio	F8W4J7	10	GGLPGSGVG	736–744	20
Danio rerio	F8W4J7	27	GGLPGSGVG	1180–1188	30
Oreochromis niloticus	I3JQ46	5	GGVPGFGGG	613–621	16
Myotis lucifugus	G1QCG1	1	GGVPGLGIG	442–450	22
Oreochromis niloticus	I3JQ46	6	GGVPGVGGG	625–633	16
Macaca nemestrina	A0A2K6BHK1	3	GGVPGVGVG	421–429	21
Sus scrofa	A0A287AXU0	1	PGAPGFGPG	330–338	17
Balaenoptera acutorostrata scammoni	A0A384AI97	1	PGAPGFGPG	315–323	–
Dipodomys ordii	A0A1S3GKU4	2	PGFPGVGAG	303–311	–
Mandrillus leucophaeus	A0A2K5XQM6	1	PGGPGFGPG	203–211	12
Ailuropoda melanoleuca	G1MA94	1	PGGPGFGPG	315–323	17
Homo sapiens∗	P15502	1	PGGPGFGPG	323–331	18
Pan troglodytes	A0A2I3SVF0	1	PGGPGFGPG	327–335	18
Rhinopithecus bieti	A0A2K6MP70	1	PGGPGFGPG	294–302	18
Callithrix jacchus	A0A2R8P8B8	1	PGGPGFGPG	339–347	19
Colobus angolensis palliatus	A0A2K5JW24	1	PGGPGFGPG	322–330	19
Rhinopithecus roxellana	A0A2K6NWE2	1	PGGPGFGPG	322–330	19
Macaca nemestrina	A0A2K6BHK1	1	PGGPGFGPG	322–330	19
Ursus maritimus	A0A384CQV9	1	PGGPGFGPG	333–341	–
Erinaceus europaeus	A0A1S3W547	1	PGVPGFGRG	343–351	–
Callorhinchus milii	V9KAK8	1	QGEPGLGGG	243–251	–
Danio rerio	F8W4J7	3	TGLPGIGPG	463–471	12
Danio rerio	F8W4C7	1	TGRPGNGRG	137–145	3
Macaca mulatta	A0A1D5R663	1	VGAPGGGLG	302–310	13
Danio rerio	F8W4J7	11	VGLPGGGLG	792–800	22
Danio rerio	F8W4J7	15	VGLPGGGLG	875–883	24
Balaenoptera acutorostrata scammoni	A0A384AI97	5	VGVPGFGAG	602–610	–
Odobenus rosmarus divergens	A0A2U3ZFC3	3	VGVPGFGAG	523–531	–
Trichechus manatus latirostris	A0A2Y9RHH3	1	VGVPGFGAG	569–577	–
Heterocephalus glaber	G5BG87	1	VGVPGFGAG	589–597	–
Papio anubis	A0A2I3MW49	2	VGVPGLGVG	463–471	16
Cercocebus atys	A0A2K5LP54	2	VGVPGLGVG	503–511	18
Mandrillus leucophaeus	A0A2K5XQM6	3	VGVPGLGVG	423–431	19
Macaca mulatta	A0A1D5R663	3	VGVPGLGVG	535–543	23
Macaca mulatta	A0A1D5R663	4	VGVPGLGVG	544–552	23
Homo sapiens∗	P15502	4	VGVPGLGVG	585–593	26
Bos mutus	L8IPH2	2	VGVPGLGVG	573–581	26
Bos mutus	L8IPH2	3	VGVPGLGVG	582–590	26
Macaca nemestrina	A0A2K6BHK1	5	VGVPGLGVG	572–580	26
Pan troglodytes	A0A2I3SVF0	4	VGVPGLGVG	589–597	27
Balaenoptera acutorostrata scammoni	A0A384AI97	4	VGVPGLGVG	593–601	–
Tursiops truncatus	A0A2U4A8K1	2	VGVPGLGVG	96–104	–
Danio rerio	F8W4J7	12	YGLPGGGLG	813–821	22

Accession number, accession number of the origin sequence. #, number of the hit within the sequence. Position, position of the hit. Exon, exon number (when available).

^∗Human sequences.

CD spectroscopy

To describe the structural behavior of the nonapeptides in solution, their CD spectra were recorded in different (polar versus low dielectric constant) environments. Routinely, methanol is considered as a primary step in studying the peptides conformational behavior in hydrophobic environments such as the membrane interior or hydrophobic pockets of proteins (36).

The analysis of the CD spectra in water (150 mM NaCl), in a water/methanol mixture (50/50, v/v), and in methanol (Fig. 2) reveal a progressive conformational change of the nonapeptides in going from water to methanol solutions.

Figure 2.

Ultraviolet-CD spectra of the four nonapeptides observed in water, water/methanol mixture (50:50), and methanol. The peptides are hN3 (A), hN4 (B), hN5 (C) and hN6 (D).

The CD spectra recorded in water all present a double negative band shape, i.e., composed of a deep minimum at ∼198 nm, followed by a weaker and broader negative band centered at ∼225 nm. It is a matter of fact that random chains are generally identified by a deep negative band at ∼198 nm (37,38). However, a negative double band, like that observed in the case of the nonapeptides (Fig. 2), is an indicator of turns. For instance, certain β-turns are known to give rise to a negative double band. It is to be mentioned that β-turns are recognized as the most frequent ones in proteins (39, 40, 41). Particularly, the CD fingerprints of four categories of β-turns, i.e., type I (37), II, II′ (42, 43, 44), and VIII (45) β-turns, are identified by a double negative band with unequal ellipticities. It is interesting to note that among all the mentioned β-turns, that corresponding to type VIII is formed by a deep minimum at ∼198 nm, followed by a broad shoulder at ∼220 nm (45), i.e., strikingly similar to that observed in the case of the presently analyzed nonapeptides. However, this overwhelming resemblance between CD markers does not permit rejection of possible unordered chains, of which the CD marker can be overlapped with the negative band at ∼198 nm. In the same manner, a probable overlap may occur between the CD fingerprint of β-strands, generally characterized by a unique negative band at ∼215 nm, and the ∼225 nm negative band from β-turns. Nevertheless, any uncertainty about the presence of polyproline-II-type chains can be discarded because of the recent CD studies highlighting the fact that the polyproline II fingerprint is composed of a deep negative band at ∼198 nm, followed by a positive one centered at ∼220 nm. The presence of the latter band cannot be suspected in the CD spectra displayed in Fig. 2.

In methanol (Fig. 2), the CD signals of the nonapeptides are composed of a broad positive band peaking at ∼205 nm, followed by a weak and broad negative band at ∼225 nm. It is worth noting that the CD spectra observed in a water/methanol mixture (1:1, v/v) are somehow an overlap of those observed in water and methanol. To provide a reliable assignment for the CD signal observed in methanol, we can recall that although a type II β-turn gives rise to a positive band located between 190 and 210 nm (37), the observed negative band at ∼225 nm in methanol can be ascribed to β-strands and/or to type I′ β-turn (37). Based on the observed CD spectra, no unordered chain can be expected in the presently studied nonapeptides.

Raman spectroscopy

Raman spectra of the nonapeptides were recorded in the middle wavenumber region (1800–550 cm⁻¹). They presented a low water contribution enabling an accurate analysis of the observed bands (Fig. 3, A and B). Whatever the nonapeptide, the spectra were identical for the two tested concentrations, i.e., 2.5 and 20 mM (data not shown). Indeed, no substantial change, assignable to either a possible conformational transition or a detectable molecular aggregation, could be observed. Because the spectral features were more pronounced at 20 mM (higher Raman intensities due to higher concentration), we will limit our discussion to the Raman spectra recorded at that concentration.

Figure 3.

Room temperature Stokes Raman spectra in the middle wavenumber region for (A) hN3, hN4, hN5 and (B) hN6, compared to F. Spectra were recorded at 20 mM. See Table 2 for tentative assignments.

The tentative assignments of the Raman spectra of the four studied nonapeptides are reported in Table 2. These assignments have been performed in keeping with the features observed in the Raman spectra of their constitutive amino acids (Fig. S2). The bands observed in the 1500–1320 cm⁻¹ range, mainly arising from the bending modes of CH₂ and CH₃ side-chain moieties, as well as those located below 1200 cm⁻¹ in amino acid spectra, were taken as references for assigning the observed peaks in the spectra of peptides.

Table 2.

Tentative Assignment of the Raman Bands Observed in the Middle Wavenumber Region

AGIPGLGVG hN3	VGVPGLGVG hN4	AGVPGLGVG hN5	AGVPGFGAG hN6	Tentative Assignment
1700-1620 (br)	1700-1620 (br)	1700-1620 (br)	1700-1620 (br)	Amide I
–	–	–	1605 (s)	F1
–	–	–	1586 (m)	F2
1465 (s)	1468 (s)	1466 (s)	–	P, I, L, V, A
1451 (s)	1452 (s)	1452 (s)	1454 (s)	V, L, I
1421 (s)	1421 (s)	1420 (s)	1421 (s)	G, A, L, I, V, P
1391 (m)	1394 (m)	1395 (m)	1392 (m)	G, A, L, I, V, P
1343 (s)	1345 (s)	1343 (s)	1341 (s)	A, L, I, V, P
1315 (s)	1313 (s)	1317 (s)	1320 (s)	A, L, I
1300-1240 (br)	1300-1240 (br)	1300-1240 (br)	1300-1240 (br)	Amide III
–	–	–	1207 (s)	F3
–	–	–	1183 (m)
1161 (m)	1175 (m)	1161 (m)	1159 (m)	V, L, I, P
1128 (s)	1127 (s)	1126 (s)	1126 (m)	V, L, I
–	–	–	1101 (m)	P
–	–	–	1031 (s)	F4
1031 (m)	1034 (m)	1031 (m)	–	G, I, P
1011 (m)	1014 (m)	1014 (m)		A, V
–	–	–	1004 (vs)	F5
960 (s)	962 (s)	960 (s)	–	V, L, I
937 (s)	938 (s)	937 (s)	937 (br)	A, V, L, I
891 (s)	891 (s)	891 (s)	–	V
857 (m)	860 (m)	856 (sh)	851 (br)	L, P
845 (m)	840 (m)	843 (m)	–	A, V, L, I, P
–	–	–	758 (br)	P
736 (br)	735 (br)	737 (br)	–	V, L, I, P
–	–	–	621 (m)	F6

(vs) very strong; (s) strong; (m) medium; (br) broad Raman bands. F1 to F6 refer to the six characteristic Raman bands of phenylalanine. Amide I and amide III vibrations give rise to a broad and strong Raman band peaking at ∼1685 and 1255 cm⁻¹, respectively.

The spectrum of hN6 is characterized by the occurrence of the six Phe characteristic Raman bands, referred to as F1 to F6 (46,47). Being all of the in-plane type and localized in the Phe phenyl ring, the wavenumbers of these six characteristic markers remained very close to those collected from the free amino acid F (Fig. 3 B). The Raman spectra of hN3, hN4, and hN5 (Fig. 3 A) do not present such narrow and resolved aromatic markers, but their striking spectral shape similarity is to be emphasized (Fig. 3). Two regions, corresponding to amide I (1700–1640 cm⁻¹) and amide III (1300–1230 cm⁻¹) vibrations, provide valuable structural information through their decomposition into band components (42, 43, 44). It is noteworthy that whereas amide I vibrations result from the backbone C=O bond stretch motion, more or less coupled to its adjacent N-H angular bending, amide III vibrations mainly arise from the backbone N-H angular bending. The Raman spectra obtained for the free amino acids (Fig. S2) are devoid of amide (I and III) vibrations. However, the presence of low-intensity bands arising from the side chains of Val, Leu, Ile, and Pro falling within the amide III range should be stressed. These bands can be naturally superimposed to those relative to amide III vibrations in peptides. Nevertheless, because of their weakness, one cannot expect a considerable distortion in the structural analysis on the basis of amide III vibrations. In all nonapeptides, amide I and amide III vibrations are characterized by two strong, broad, and incompletely resolved bands peaking at ∼1685 and ∼1265 cm⁻¹, respectively. To go farther in extracting the structural information from amide vibrations, the decomposition of the amide I region for four nonapeptides is displayed in Fig. 4. Through a systematic investigation of the structural analysis of the peptide chains (42, 43, 44,48,49) in an aqueous environment, we have been led to select four band components located at ∼1690 ± 5 (assignable to random chain), ∼1660 ± 5 (assignable to β-strand), and ∼1675 ± 5 and ∼1650 ± 5 (both assignable to turns) to decompose the amide I region in nonapeptides. The populations corresponding to different secondary structural elements, as estimated by the normalized areas of the band components (expressed in percent), are reported in Table 3. As can be seen, between 50 and 60% of the total population is formed by turn structures, and the rest is equally distributed between β-strands and random chains.

Figure 4.

Band decomposition in the amide I region of the nonapeptides observed at 20 mM. The peptides are hN3 (A), hN4 (B), hN5 (C) and hN6 (D). Observed spectra are drawn with black lines and component bands with dark gray lines. The light gray trace corresponds to the sum of component bands (fit). The maximal band wavenumber of each used component is reported. In parentheses: (R) random, (T) turn, (S) β-strand. For the estimated populations of different secondary structural elements, see Table 3.

Table 3.

Underlying Secondary Structure Elements in the Amide I Profile of the Nonapeptides and Fractional Areas

Position (Assignment)	hN3	hN4	hN5	hN6
∼1690 (random)	20	20	24	25
∼1680 (turn)	41	40	37	38
∼1660 (β-strand)	26	23	21	25
∼1640 (turn)	13	17	18	12
Fractional Areas
Random	20	20	24	25
Turn	54	57	55	50
β-strand	26	23	21	25

Component positions are ±5 cm⁻¹. During the computation, their widths were kept between 15 and 25 cm⁻¹. The figures correspond to the normalized areas of the components expressed in percent of the whole amide I profile. Fractional areas estimate the global secondary structures with an accuracy of ±5% (22).

Solution samples containing nonapeptides at 20 mM were heated up to 80°C. Few changes appeared in their Raman spectra, but slight changes could be observed in the amide III region. Fig. 5 exemplifies this behavior for hN5 and hN6. The main effect is a wavenumber downshift of the spectral profile, consistent with a progressive shift from turn to β-strand upon increasing temperature. Albeit the turns remain the major population of secondary structures in the 10–80°C temperature range (data not shown), they show a minor tendency to be transformed into β-strands, presumably because of the weakening or breakdown of the interturn stabilizing hydrogen bonds occurring upon thermal annealing.

Figure 5.

Effect of temperature on the amide III vibrations of nonapeptides. The spectra were recorded for a concentration of 20 mM. Gray scales are used to display the spectra obtained as a function of temperature. The considered peptides are hN5 (A) and hN6 (B).

Atomistic molecular dynamics simulations

To get a more detailed representation of the structures that can be adopted by the four nonapeptides, 10 molecular dynamic simulations of 500 ns have been performed for each peptide. The adopted conformations, analyzed with DSSP (31), are presented in Fig. 6 A and Fig. S3. The figures show that the peptides mainly adopt random-coil conformations, with some residues favoring turns, but no stable structure can be observed. The formation of β-turns is evidenced by distances between Cα atoms of residues i and i + 3 being less than 0.7 nm with residues i + 1 and i + 2 being nonhelical, which is mainly observed in the middle of the peptides between residues 3 and 6 and notably for the hN3 peptide (Fig. S4 A). The type of β-turns can be identified by analyzing the ΦΨ angles of two amino-acid residues following each other as determined by Lewis et al. (32) and Hutchinson et al. (33). Fig. 6 B shows that type II β-turns are formed during a third to a half of the simulation time (52.9 ± 4.1% for hN3, 37.8 ± 4.7% for hN4, 33.9 ± 2.8% for hN5, and 33.5 ± 3.8% for hN6, mean ± SD) for the PG residues being at positions 4 and 5. PG residues are strongly favored at i + 1 and i + 2 positions in experimental structure of type II β-turns, and they are known to promote this type of β-turn. The population of type II β-turns is greater for the hN3 peptide than for the hN4, hN5, and hN6 peptides. Fig. S4, B–E show that type I, I′, II, II′, and VIII β-turns can also be formed in smaller amounts at other positions of the peptides and that type VIII β-turns are mainly observed at the start of the peptides (up to 3% of the simulation time). A proline at position 4 has been shown to promote the formation of this type of turn (33). Analysis of the simulation underlines that the structures are rapidly changing on the nanosecond timescale from extended to more packed conformations, which mainly corresponds to type II β-turns (Fig. S4 F). The population of β-turns has also been analyzed through clustering methods, and the central structures of the dominant cluster population of each peptide is presented as an inset in Fig. 6 B. Altogether, our analysis of molecular dynamics data strongly suggests that the PG motif of the nonapeptides could mainly adopt a type II β-turn conformation.

Figure 6.

Conformational analysis of the molecular dynamics simulations. (A) Secondary structures adopted during the simulation calculated with DSSP (31) are shown. (B) The occurrence of β-turn types during each frame of the simulation trajectories is shown. β-turns are computed considering the ΦΨ angles of the PG residues at positions 4 and 5. Inset shows an overlay of the conformation representative of the main cluster computed for each of the four nonapeptides. Residues 3–6 are represented with a thicker line. Carbon atoms have the same colors as those defined for each peptide in (B). Error bars are standard deviations between the 10 simulations.

To test the agreement of these results with the measured CD spectra, CD spectra were computed from our simulated trajectories with the SESCA software (35). The results (Fig. S5) show a very good agreement between the measured and computed spectra despite noticeable differences around ∼220 nm. This suggests that the computed profiles might have failed to detect several experimental signals. Nevertheless, they reinforce our proposal that β-turns, and notably type II β-turns, are the prevailing structures adopted by the nonapeptides.

Discussion

Among elastin peptides, elastin nonapeptides have been identified initially as chemoattractants for fibroblasts (12). Now, there are growing evidences that these elastin peptides could define a peculiar class of matrikines.

Brassart and colleagues have recently shown that hN5 (i.e., AGVPGLGVG) plays a key role in tumor progression by promoting tumor cell blebbing and extracellular vesicle shedding after its interaction with RPSA (17). The structure of hN5 has been analyzed previously by CD, FT-IR, and NMR spectroscopies (16). This report concluded that the structure of the peptide was a mixture of random coils and β-turns. Our data fully agree with these findings, but they also expand our knowledge of the nature and structural behavior of these elastin peptides.

The occurrence of the XGXPGXGXG consensus sequence was analyzed in a large sequence data bank. Our results show that among the observed hits, the group of elastin sequences defines a specific group (Fig. 1). Further, our analysis underlines the fact that these peptides mostly occur in tandem repeats in human exon 26 (and corresponding sequences in other species). Finally, as far as human sequences are concerned, we observed that similar sequences were also present in numerous species (Table 1). The most striking observation is that hN6 is present in numerous species and is tandemly repeated five times in the mouse sequence. In contrast, hN3 is only found in three species. These observations suggest that, besides their matrikine activity, these sequences could play a key and yet undescribed role in elastin biology, possibly elastin assembly. Indeed, among elastin sequences, these sequences appear much more conserved than the canonical elastin hexapeptide sequence (say, VGVAPG), the occurrence of which is mostly limited to human exon 24 sequence. Further works are needed to test this hypothesis.

Our data (Fig. 1) show that the XGXPGXGXG consensus sequence can be found in most phyla and/or species. To the best of our knowledge, sequences harboring this consensus have caught attention only in elastin because these elastin peptides are bioactive (8,12, 13, 14, 15,17). Up until now, neither a biological activity nor a dedicated receptor has been described for the nonelastin sequences.

As evidenced by the consensus profile reported for each of them (Fig. 1), the nonelastin sequences (96,259 sequences) present a sequence pattern that is mostly dominated by the occurrence of G, A, or K at X positions. These residues are not favored in elastin sequences in which AGVPGFGVG is the prevailing motif. This is remarkably pronounced when nonelastin animal sequences are considered. Their consensus is GGGPGGGGG. These important sequence differences suggest that nonapeptides found in nonelastin sequences would most probably be structured and behave differently than elastin nonapeptides. Nevertheless, a comprehensive structural analysis is required to address this point.

In our study, we focused our analysis on the nonapeptide sequences found in human exon 26 aiming at understanding their structure and structural behavior. First, this investigation was performed using optical spectroscopies. The results gathered by both CD and Raman spectra are consistent with previous reports underlining a conformational equilibrium between random, strand, and turn conformations within elastin sequences (50). The Raman data underline that the conformation of these peptides is dominated by β-turns, as deduced from the quantitative analysis of their amide I profiles. These conformations appear very favored because increasing the temperature up to 80°C did not drastically change the spectra and hence their structure, although a small increase of the β-strand content could be observed upon heating. Altogether, our experimental data suggest that the structures of the nonapeptides are dominated by β-turns and that these conformations could be engaged in a conformational equilibrium with extended and random structures.

To have a better understanding of the structural behavior of these peptides, molecular dynamics simulations were undertaken in explicit water. 10 independent 500-ns simulations were performed for each of the four nonapeptides. The results obtained are extremely consistent with our experimental data. Indeed, they show that the most common structure observed during the simulation is a type II β-turn occurring at the X-P-G-X motif. Quantitatively, this structure seems more abundant in hN3 (∼50%) than hN4, hN5, or hN6 (25–35%). This finding is in very good agreement with the conclusions of Lessing and colleagues (51), who observed that the preference for the X-Pro-Gly sequence to form a β-turn increased with the complexity and hydrophobicity of the side chain at position X (Gly < Ala < Val). In contrast to hN4, hN5, and hN6, which harbor a valyl residue before PG, hN3 possesses an isoleucyl residue at that position. Thus, the higher propensity of hN3 to adopt a β-turn when compared to the other three nonapeptides is consistent with the work of Lessing and colleagues.

Since the first report of their bioactivity on bovine fibroblasts (12), nonapeptide sequences have been tested on various cellular systems: bovine endothelial cells (13), rat macrophages (14), and human tumor cell lines (8,9). Nevertheless, among those publications, few used more than one peptide sequence, thereby preventing the comparison of their respective effects. In the work of Maeda and colleagues, hN5 and hN6 were identified as chemoattractants for rat macrophages (14). In this work, the authors also analyzed the effect of the canonical VGVAPG sequence. Their conclusion was that the two nonapeptides could compete for receptor binding on the elastin binding protein (EBP). Indeed, although VGVAPG appeared as an exclusive ligand for EBP, they underlined that hN5 and hN6 could also bind EBP and a lactose-insensitive receptor, later identified as RPSA. Their work therefore suggest that the EBP-binding motif can be harbored by both hexa- and nonapeptides, whereas the RPSA binding motif can be only adopted by nonapeptides. It has been established that VGVAPG bioactivity relies on its ability to adopt mainly a type VIII β-turn conformation (10). Our molecular dynamics simulations show that the major conformation of nonapeptides could be the type II β-turn but also that type VIII turns can be present. Our experimental data also suggest that both type VIII and type II β-turns can be present in the conformation of the nonapeptides. Thus, at the difference of VGVAPG, nonapeptides could be favorably structured to bind EBP and RPSA. However, it has to be emphasized that the sequences of the hexa- and nonapeptides are different. That way, binding affinity could also be supported by favorable residue side-chain interactions in addition to the locally adopted conformation. Sequence differences could also contribute to the differential bioactivity exhibited by these peptides.

Altogether, our data suggest that the structural signature of elastin nonapeptide is a type II β-turn. Consequently, it is reasonable to propose that this conformation could be relevant of the observed bioactivity of these peptides. This proposal is supported by the observations made by Toupance and co-workers, in which they measured a lesser biological effect of hN6 toward human tumor cells than hN5 (8). In our work, hN6 is predicted to present fewer β-turn conformations than hN5. Therefore, it seems plausible that the bioactivity of hN5 is greater than that of hN6, supporting the proposal that the turn motif is the conformation sustaining bioactivity of these peptides. If our prediction is correct, then one can anticipate that hN3 would be a more bioactive peptide than others.

Despite decades of efforts, the structure-function relationship of RPSA is still an open question (52). Notably, the interaction site of elastin nonapeptides on RSPA is unknown. In contrast, RPSA laminin binding site was described as a secondary function acquired during evolution (53). This site corresponds to the 161–180 region of the human sequence. However, the crystal structure of the protein (54) locates this binding site mostly inside the fold. Therefore, it is thought that RPSA could possess multiple sites and/or associate with other compounds to exert its functions. Consequently, our current knowledge of RPSA structure cannot explain where and how it binds elastin nonapeptides. Further structural explorations are required to address this question.

Conclusions

Our data throw new light on the nature, structure, and conformational behavior of elastin nonapeptides. These peptide sequences exhibit a peculiar composition as demonstrated by our bioinformatic analysis of current databases. Furthermore, experimental (CD and Raman spectroscopies) and theoretical data (molecular dynamics) strongly suggest that the dynamic structure of these peptides is dominated by the type II β-turn conformation. We propose that the observed bioactivity of these peptides toward stromal and tumor cells could be driven by the recognition of this structural feature by their cognate receptor, RPSA. Molecular modeling of their interaction is underway.

Author Contributions

B.H. and S.G.K. performed the Raman and CD measurements and analyzed the spectra. S.B. conceived and ran the simulations. J.M.C., J.T., and N.B. performed the sequence analyses. S.B. and M.D. funded the project and edited the manuscript. L.D. supervised the work and wrote the manuscript.

Editor: Margaret Cheung.

Supporting Material

Table S2

Sequences Presenting the XGXPGXGXG Motif