Domains in Tropoelastin That Mediate Elastin Depositionin Vitro and in Vivo *

Elastic fiber assembly is a complicated process involving multiple different proteins and enzyme activities. However, the specific protein-protein interactions that facilitate elastin polymerization have not been defined. To identify domains in the tropoelastin molecule important for the assembly process, we utilized an in vitro assembly model to map sequences within tropoelastin that facilitate its association with fibrillin-containing microfibrils in the extracellular matrix. Our results show that an essential assembly domain is located in the C-terminal region of the molecule, encoded by exons 29–36. Fine mapping studies using an exon deletion strategy and synthetic peptides identified the hydrophobic sequence in exon 30 as a major functional element in this region and suggested that the assembly process is driven by the propensity of this sequence to form β-sheet structure. Tropoelastin molecules lacking the C-terminal assembly domain expressed as transgenes in mice did not assemble nor did they interfere with assembly of full-length normal mouse elastin. In addition to providing important information about elastin assembly in general, the results of this study suggest how removal or alteration of the C terminus through stop or frameshift mutations might contribute to the elastin-related diseases supravalvular aortic stenosis and cutis laxa.

The inherent complexity of the elastic fiber, combined with the unique physical properties of its component proteins, has made understanding elastin structure and assembly one of the most difficult problems in matrix biology. Although technical advances in cell and molecular biology have given us new information about elastin gene expression and elastin synthesis, surprisingly little is known about how the elastic fiber is assembled at the molecular level. The major component of the mature fiber is a covalently cross-linked polymer of the protein elastin. Elastin is secreted from the cell as a soluble monomer called tropoelastin. In the extracellular space, lysine residues within tropoelastin are specifically modified to form covalent cross-links between tropoelastin chains. This cross-linked polymer has a high degree of reversible distensibility, including the ability to deform to large extensions with small forces.
The tropoelastin molecule is characterized by a series of tandem repeats, each including a lysine-containing cross-linking region followed by a hydrophobic motif (1). Cross-linking is initiated by the extracellular enzyme(s) lysyl oxidase, which catalyzes the oxidative deamination of lysyl ⑀-amino groups. Under normal conditions, the cross-linking process is extremely efficient, with all but ϳ5 of the ϳ40 lysine residues in tropoelastin participating in covalent linkages that form the functional polymer. How the cross-linking sites within the monomer are aligned prior to cross-linking is unclear. It has long been assumed that microfibrils provide a scaffold or template for elastin assembly by binding and aligning tropoelastin monomers so that lysine-containing regions are in register for cross-linking. This idea evolved from electron microscopic images showing that the appearance of microfibrils is the first ultrastructural indication of the elastic fiber (2)(3)(4) and that microfibrils are associated with elastin throughout the elastogenic period.
Recently, two inherited diseases, autosomal dominant cutis laxa and supravalvular aortic stenosis (SVAS), 1 have been linked to mutations within the elastin gene that may alter the ability of the elastin precursor to undergo normal assembly (5,6). Each of these diseases has a distinct clinical phenotype and may be caused by fundamentally different molecular mechanisms. Autosomal dominant cutis laxa is characterized by redundant, loose, sagging, and inelastic skin with variable systemic organ involvement. The known elastin mutations in autosomal dominant cutis laxa are single nucleotide deletions in exons 30 and 32 that, depending on exon splicing, give rise to a missense peptide sequence extending into the 3Ј-untranslated region (6,7). Although the exact pathomechanism of this disease is not clear, it is thought that alterations at the C terminus of the secreted mutant protein interfere with the deposition of normal elastin in a dominant-negative fashion (5,6). SVAS is characterized by thickening of the arterial wall and either focal or diffuse narrowing of the aorta and, frequently, of other major arteries. Mutational analysis has identified a wide spectrum of mutations associated with SVAS, including large deletions, translocations, inversions, and, most frequently, point mutations (8 -12). Accumulating evidence suggests that the pathomechanism of SVAS is haploinsufficiency (13), which can result from the deletion of one complete copy of the elastin gene (as occurs in Williams syndrome) or through functional hemizygosity arising from loss-of-function mutations in one elastin allele.
The objective of this study was to better characterize elastic fiber assembly and to identify the domains on the tropoelastin molecule that mediate this process. We were also interested in determining whether proteins from transcripts with SVAS-like mutations are capable of incorporating into elastic fibers and, if so, whether they interfere with normal fiber function. Using an in vitro model system of fiber assembly together with expression of tropoelastin constructs in transgenic mice, we provide evidence for a critical assembly domain located in the region between exons 29 and 36 that mediates the interaction of tropoelastin with microfibrils in the extracellular matrix. Studies with synthetic peptides and tropoelastin deletion constructs suggest that the hydrophobic sequence encoded by exon 30 is a major functional element in this region. These findings also provide an explanation for how SVAS mutations could lead to haploinsufficiency when the truncated product of the mutant allele is secreted without this critical assembly domain.

EXPERIMENTAL PROCEDURES
Materials-All reagents were purchased from Sigma unless otherwise noted.
Antibodies-The antibody used to detect bovine elastin in in vitro experiments with pigmented epithelial (PE) cells is a polyclonal antibody generated against recombinantly produced human tropoelastin. The only exception occurred in experiments in which matrices were probed concurrently for tropoelastin and fibrillin-1. In those experiments, BA4, a monoclonal antibody reactive against bovine tropoelastin (14), and Fib-1 CT, a polyclonal antibody generated against the recombinant FBN1 C-terminal domain (15), were used. Tissue immunofluorescence studies made use of BA4 and anti-mouse recombinant tropoelastin antibody N6 -17, a polyclonal antibody generated against recombinantly produced murine tropoelastin exons 6 -17. BA4 will detect bovine and human elastin, but does not interact with mouse tropoelastin. In addition, CT e , a polyclonal antibody directed against a folded peptide encoded by exon 36 of bovine elastin (16), was used in Western blotting to probe for proteolytic fragments containing the tropoelastin C terminus.
Transfection Constructs-The transfection constructs used for this study were derived from a full-length bovine cDNA that has been previously characterized (17). It carries a deletion of exons 13 and 14. When attempts were made to amplify the missing exons from cultured fetal bovine chondrocytes using RT-PCR, cDNA containing the exons could not be identified, suggesting that the ⌬13-14 form is the dominant transcript in these elastogenic cells. Numbering associated with deletion constructs refers to sequences encoded by the exons that make up the bovine tropoelastin cDNA.
To generate the full-length transfection construct, bovine tropoelastin cDNA was cloned into the MluI-XbaI site of the pCIneo vector (Promega). Construction of the 1-28 truncation was achieved by PCR amplification of the appropriate exons from the full-length cDNA. The forward primer contains an MluI site and the ATG from exon 1, whereas the reverse primer contains a stop site and the XbaI restriction site for cloning. Primer sequences are as follows: exon 1F, 5Ј-GCACGCGTATGCGGAGTCTGACGGCTCGG-3Ј; and exon 28R, 5Ј-GCTCTAGAGTCATGCCACACCACCTGGAATGCC-3Ј. Generation of the remaining constructs required a pCIneo vector lacking the manufacturer-provided NotI site (pCIneo*). This was generated by cutting the vector with NotI and filling the resultant linearized vector with Klenow fragment, followed by self-ligation. Full-length tropoelastin was then cloned into the pCIneo* vector at the XbaI site (FL*). Only one NotI site is present in FL*, located at the beginning of exon 29. To generate the truncation constructs 1-29 and ⌬36, PCR products were amplified using the exon 27F primer (5Ј-CCCTGGCCGCAGCTAAAG-CAGCCAAGTTCG-3Ј) and the exon 29R primer (5Ј-GCTCTAGAGTC-AAAATTGGGCTTTGGCGGCA-3Ј) or the exon 35R primer (5Ј-GCTC-TAGAGTCAAAATTGGGCTTTGGCGGCA-3Ј). The reverse primers contain both the stop site for the molecule and an XbaI site for cloning purposes. After amplification, FL* and the PCR products were cut with NotI and XbaI and ligated to one another. Similarly, the ⌬RKRK construct (representing the deletion of the last 4 amino acids from the molecule) was generated by amplification using the above exon 27F primer and the ⌬RKRK-R primer that deletes these C-terminal amino acids (5Ј-GCTCTAGAGAAAATTGGGCTTTGGCGGCA-3Ј). This product was also cut with NotI and XbaI and inserted into the similarly cleaved FL* vector. The internal deletion ⌬33 was constructed by RT-PCR amplification of total RNA isolated from fetal bovine ligament cells. Fetal bovine ligament cells are known to express an alternatively spliced form of tropoelastin in which exon 33 has been spliced out (1). The PCR primers used for the amplifications include the exon 27F primer as well as the exon 36R primer that leads to amplification of the entire 36th exon as well as its stop codon (5Ј-GCTCTAGAGTCACTT-TCTCTTCCGGCCACA-3Ј). The product of the amplification was then cleaved with NotI and XbaI and ligated into the FL* vector cut with the same enzymes. To generate the Cys-to-Ala mutations, oligonucleotidedirected mutagenesis was performed using a QuikChange site-directed mutagenesis kit (Stratagene) with oligonucleotides C755A-F (5Ј-TGG-GGAAATCCGCTGGCCGGAAGAG-3Ј) and C755A-R (5Ј-CTCTTCCG-GCCAGCGGATTTCCCCA-3Ј) and oligonucleotides C751A-F (5Ј-CCA-GGTGGGGCCGCCCTGG-3Ј) and C751A-R (5Ј-ATTCCCCACCGCGG-CCCCACCTGG-3Ј) according to the manufacturer's recommendations.
Several steps were required to generate construct ⌬30. First, exons 28 -36 were amplified by PCR. The primers used were exon 28F (5Ј-GGAATTCAGATCTTGGTGGAGCCG-3Ј) and the exon 36R primer listed above. The exon 28F primer contains an EcoRI site such that the product can be digested with EcoRI and XbaI and ligated into a similarly digested pUC-19 shuttle vector (pUC- 28 -36). To generate the ⌬30 insert, exon 30 was deleted by PCR amplification of a 113-bp fragment using the exon 28F primer and the exon 29/31R primer (5Ј-AACTGC-AGCTGGAGACACACCAAATTGGGCGGCTTTGGCGGC-3Ј) encoding a PstI site in exon 31. The EcoRI-PstI-restricted fragment was ligated into the similarly digested pUC-28 -36 plasmid, resulting in the pUC⌬30 plasmid. pUC⌬30 was then digested with NotI and XbaI, and the bTE⌬30 insert was ligated into the NotI-XbaI-restricted FL* plasmid, resulting in pCIneo-bTE⌬30. All alterations were verified by nucleotide sequencing.
Transfection-Stable transfection of PE cells was performed using Lipofectin (Invitrogen) according to the manufacturer's instructions. Briefly, 2 ϫ 10 5 cells were plated in a six-well plate. When cells reached 60 -80% confluence, 1 g of the chosen construct and 1 l of reagent complex were added to the cells in Opti-MEM serum-free medium (Invitrogen). Sixteen hours later, the transfection reagents were removed, and fresh Dulbecco's modified Eagle's medium containing 10% cosmic calf serum was added. After 48 h, cells were then placed under selection with 500 g/ml active Geneticin (G418 sulfate, Invitrogen).
Immunofluorescence-To detect bovine tropoelastin in the matrix of stable pools of G418-selected PE cells, the cells were plated on glass coverslips in six-well plates and allowed to reach confluence. Seven days after visual confluence, the medium was removed, and the cell layer was washed with phosphate-buffered saline (PBS) three times to remove all non-cell layer-associated proteins. Cells were then fixed with ice-cold methanol for 60 s and washed with PBS to remove any remaining alcohol. The cells were treated with blocking solution consisting of PBS containing 1% gelatin and 0.1% Tween 20. The anti-human recombinant tropoelastin antibody was used at a dilution of 1:250 in blocking solution. Fluorescent secondary antibody (goat anti-rabbit 488, Molecular Probes, Inc.) was used at a concentration of 1:2000 in blocking solution. After secondary antibody incubation, the cells were washed before removal from the six-well plate and inversion of the coverslip onto a glass slide. Fluorescence was protected using anti-fade mounting medium (Gelmount, Biomedia Corp.). A Zeiss Axioscope microscope was used for fluorescence microscopy, and the images were captured on an Axiocam digital camera using Axiovision software. All images are shown at magnification ϫ40 except where noted.
Metabolic Labeling and Immunoprecipitation-Conditioned media from confluent monolayers of fetal bovine chondrocytes, untransfected PE cells, and stably transfected PE cells that were metabolically labeled with L-[4,5-3 H]leucine (1 mCi/ml; ICN Pharmaceuticals, Irvine, CA) were immunoprecipitated for tropoelastin as described previously (19). Immune complexes were pelleted, washed, and resuspended in 35 l of Laemmli sample buffer containing 100 mM dithiothreitol. Samples were electrophoresed on SDS-polyacrylamide gels, fixed, treated with EN 3 HANCE (PerkinElmer Life Sciences) for 1 h, dried, and exposed to X-Omat AR film (Eastman Kodak Co.).
Transgenic Animals-Animals were generated expressing either full-length bTE or bTE-(1-28). To generate the full-length bTE and bTE-(1-28) vectors, the pCIneo constructs described above were linearized, yielding two constructs that contained all of the pCIneo vector components in addition to the appropriate tropoelastin sequence. To generate construct ⌬30, the pCIneo construct was linearized, and extraneous plasmid sequences (neo cassette, etc.) were removed by restriction digestion with ClaI. In all cases, the appropriate fragment was purified using a QIAGEN gel extraction kit. Each linearized product contained, at a minimum, the tropoelastin sequence flanked by the cytomegalovirus immediate-early promoter and the SV40-derived 3Јuntranslated region and poly(A) signal. Each fragment was resuspended at 2 g/ml in injection buffer composed of 5 mM Tris-HCl (pH 7.4), 0.25 mM EDTA (pH 8.0), and 5 mM NaCl and was injected into B6C3/F1 mouse fertilized eggs, which were implanted into the uteruses of pseudopregnant foster mothers. After birth, potential founders were screened for the presence of the transgene using PCR and bovine tropoelastin-specific forward (5Ј-GGGGTACCAGGAGCTGTTCC-3Ј) and reverse (5Ј-CCTTGGGCTTGACTCCTGCTC-3Ј) primers. Once detected, animals positive for the transgene were mated to wild-type animals to stabilize the line.
RT-PCR-RNA was isolated from wild-type and transgenic animals using RNAzol (Tissue Tek) according to the manufacturer's instructions. Heart, lung, kidney, and liver total RNAs were screened in all founders tested. RT-PCR was performed on 1 g of each resultant RNA under the following conditions. Residual DNA was removed by treating the RNA with RQ1 RNase-free DNase in first-strand RT reaction buffer (Invitrogen) containing 2 l of 0.1 M dithiothreitol for 30 min at 37°C. The enzyme was heat-inactivated at 72°C for 10 min. Reverse transcription was performed with Superscript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions in the presence of dNP6 random primers (Roche Applied Science) and dNTPs (Promega). PCR was then performed on the resultant cDNA with bovine tropoelastin-specific forward (5Ј-GGAATTGGAGCCATTCCCACATTT-GGG-3Ј) and reverse (5Ј-GAGCCACGCCGACTCCAGG-3Ј) primers. Thirty-five cycles of amplification were performed with a 67°C annealing temperature and a 30-s extension time. The products of this reaction were then analyzed by gel electrophoresis.
Immunofluorescence Analysis-Tissues from transgenic and wildtype animals were dissected and washed with sterile PBS before being embedded in blocks containing OCT freezing medium (Tissue Tek). The tissues were then frozen over dry ice, and 10-m sections were cut using a microtome.
For immunofluorescence analysis, sections were thawed to room temperature, fixed in cold ethanol, and etched to enhance antigenicity with 1 mg/ml hyaluronidase in 0.1 M sodium acetate buffer (pH 5.5) containing 0.85% NaCl for 30 min at 37°C. After hyaluronidase treatment, the tissues were treated as described above for tissue culture cells. The antibodies used were BA4 and N6 -17. Both were used at a 1:500 dilution. Goat anti-mouse 594 and goat anti-rabbit 488 secondary antibodies (Molecular Probes, Inc.) were used for primary antibody detection.
Quick-freeze Deep-etch Electron Microscopy and Congo Red Staining of Exon 30 Peptides-A peptide encoding exon 30 of bovine tropoelastin (GLGGVGGLGVGGLGAVPGAVGLG) was synthesized by solid-phase peptide synthesis using a 431 A peptide synthesizer (ABI) running Fast-Moc chemistry and dissolved at 10 mg/ml in 7 M guanidine HCl. A second, scrambled version (LVGGGGGGLPVLGGAGALGGVGV) was also synthesized as a control for this assay. The peptide stock solution (10 l) was diluted with 200 l of PBS and 200 l of water and subjected to slow rotation overnight at room temperature. Precipitated peptide was pelleted by centrifugation in a microcentrifuge, smeared on a glass slide, fixed in 95% ethanol, and stained with 0.05% Congo red in 50% glycerol. Evaluation of Congo red staining by polarization microscopy was performed using a Zeiss Axioscope equipped with optimally aligned crosspolarizers. For electron microscopy, freshly prepared mica flakes were added to solutions containing peptide filaments, followed by freeze-drying and platinum replication according to established procedures (18,19).
Purification of Recombinant Tropoelastin-Full-length bovine tropoelastin was cloned into the pQE vector (QIAGEN Inc.), expressed in bacteria, and purified with nickel-nitrilotriacetic acid resin (QIAGEN Inc.) according to the manufacturer's instructions. Bound protein was then eluted with 6 M urea buffer (pH 4.0) and dialyzed in 0.1% acetic acid. Protein concentration was quantified by amino acid analysis, and aliquots of 200 g were lyophilized.
Protease Treatment and Western Blotting-Aliquots (200 g) of recombinant tropoelastin were resuspended in 350 l of PBS containing the protease inhibitors EDTA (1 mM) and pepstatin (1 mM) to block proteolysis by non-trypsin-type proteases. A sample was taken prior to the addition of plasmin at time 0. Then, 1 l of 0.2 g/ml plasmin was added, and 50-l samples were taken at 30, 60, and 120 min. Plasmin activity was stopped by adding 1ϫ SDS-PAGE sample buffer containing 0.1 M dithiothreitol and boiling for 5 min. The sample was electrophoresed on SDS-polyacrylamide gels and transferred to nitrocellulose (Schleicher & Schü ll). Nitrocellulose blots were blocked in 5% (w/v) nonfat milk in 50 mM Tris (pH 7.5), 171 mM NaCl, and 0.05% (v/v) Tween 20 (TTBS). The CT e primary antibody (18) was diluted 1:250 in TTBS containing 2.5% (w/v) nonfat milk and incubated for 1 h at room temperature. The blot was then washed and incubated with peroxidaseconjugated donkey anti-rabbit IgG secondary antibody (Amersham Biosciences) diluted 1:2000. Chemiluminescence detection was performed using ECL Western blot detection reagents (Amersham Biosciences) and exposed to XAR-5 x-ray film (Eastman Kodak Co.).

The Presence of the Tropoelastin C Terminus Is Required for Its Deposition into the Extracellular Matrix of PE Cells-To
determine which regions of tropoelastin are necessary for its association with the extracellular matrix, we generated expression constructs consisting of full-length elastin as well as mutant and deletion forms of the molecule. These constructs were then transfected into PE cells, and the ability of the expressed transgene to associate with microfibrils in the PE cell matrix was determined by immunofluorescence microscopy. PE cells are a cell line derived from the pigmented epithelial cells in the ciliary body of the bovine eye (20). They are known to lay down an elaborate fibrillar matrix composed of fibrillin-1, fibrillin-2, and MAGP1 (microfibril-associated glycoprotein-1), but do not produce endogenous elastin (17). Because these cells produce all of the components necessary to form the scaffolding for elastic matrices, but not tropoelastin itself, they provide a useful system for studying the early stages of elastic fiber assembly.
After stable transfection of PE cells with the full-length bovine tropoelastin construct, microfibrils in the PE cell matrix became decorated with elastin expressed from the transgene (Fig. 1). In this experiment, elastin was detected using the anti-human recombinant tropoelastin antibody, which has been shown to cross-react with the bovine protein. As expected, untransfected cells did not contain elastin in their extracellular matrix ( Fig. 2A). Because earlier studies had suggested that the C terminus of tropoelastin contains a critical assembly site (16), our truncation and mutation constructions focused on this region. Exon 36 encodes a conserved sequence that contains the molecule's only 2 Cys residues and a terminal Arg-Lys-Arg-Lys sequence. The cysteine residues have been shown to form a disulfide-bonded loop structure that creates a highly charged "pocket" at the end of the molecule believed to facilitate interactions between tropoelastin and highly acidic microfibrils. It was an antibody to this exon 36 sequence that disrupted elastin fiber assembly in studies by Brown-Augsburger et al. (16). To assess the importance of this region of the protein to elastin assembly, we generated tropoelastin constructs in which first one and then both of the cysteines residues were mutated to alanines, as well as one construct in which the Arg-Lys-Arg-Lys sequence was deleted. A fourth construct in which exon 36 was deleted in its entirety (⌬36) was also generated. Interestingly, protein from all of the mutant constructs localized to the fibrillar matrix of PE cells (Fig. 2C), indicating that the sequence encoded by exon 36 is not required for the initial association of tropoelastin with microfibrils.
We next assayed a series of C-terminal truncations that systematically deleted individual exons beginning with exon 36. Like construct ⌬36, protein from constructs lacking exons 31-36 (Fig. 2D) was deposited onto microfibrils. When exon 30 was then deleted (i.e. the expressed protein contained only exons 1-29), the association of protein with microfibrils was substantially decreased (Fig. 2E) relative to the full-length control. Deletion of exon 29 (resulting in construct 1-28) completely abolished elastin assembly into the matrix (Fig. 2F). Protein from all of the constructs was detected at high levels in PE cell-conditioned medium (Fig. 3), confirming their synthesis and secretion.
Expression of Full-length and Truncated Tropoelastin as Transgenes in Mice-To determine how the absence of the C-terminal region of tropoelastin alters its incorporation into existing elastic fibers, we generated transgenic animals expressing either full-length bTE or bTE-(1-28). Using speciesspecific antibodies, it was possible to determine whether either of the transgenes was able to incorporate into elastic fibers made by the mouse and whether either acted through a dominant-negative mechanism to disrupt normal mouse elastic fiber assembly. The constructs consist of linearized forms of the transfection constructs used in the in vitro experiments. They contain the immediate-early components of the cytomegalovirus promoter to guide expression of the transgene and a SV40-derived 3Ј-untranslated region and poly(A) signal.
After founder lines were stabilized, animals were tested for expression of the transgene. RT-PCR performed on RNA isolated from full-length and exon 1-28 transgenic animals showed that message from the transgene was being transcribed and was stable in these animals (lung tissue shown in Fig. 4). Using an antibody that detects human and bovine elastin, but not mouse elastin, and one that detects only mouse elastin, we assayed frozen sections from wild-type (non-transgenic) and transgenic lines. As expected, wild-type non-transgenic lungs showed no staining with the bovine-specific antibody (Fig. 4A), but significant elastic fiber staining with the mouse antibody was observed (Fig. 4D). Analysis of the four founders expressing the full-length bovine tropoelastin transgene found uniform deposition of bovine elastic fibers in the heart and developing aorta (data not shown), consistent with past studies showing strong expression of the cytomegalovirus promoter in these tissues (21)(22)(23). In other tissues, deposition of bovine elastic fibers differed among founders, most likely resulting from transcriptional differences associated with positional effects surrounding the location of transgene insertion. Commonly positive tissues included lung (Fig. 4B), kidney, bladder, and small blood vessels of the liver (data not shown). In all tissues in which the full-length transgene was expressed in a given animal, the bovine protein was found to incorporate into existing mouse elastic fibers with no obvious alteration of fiber structure (Fig. 4E). In contrast to what was found with the full- length bovine transgene, the product of the exon 1-28 transgene did not associate with elastic fibers in any of the mRNA-positive tissues when assayed in multiple founders (Fig. 4C). The tissue expression pattern for the exon 1-28 transgene, based on mRNA analysis, was essentially the same as that observed for the full-length construct.
The Amino Acid Sequence Encoded by Exon 30 Aggregates to Form Amyloid-like Fibers with ␤-Sheet Structure-To better understand the mechanism of C-terminal based tropoelastin deposition, we used synthetic peptides to investigate whether particular sequences in this region contribute to tropoelastin aggregation. One sequence in particular was suggested by earlier studies of Robson et al. (24,25), who pointed out similarities between structural motifs in exon 30 of tropoelastin and other proteins that aggregate via ␤-sheet/␤-turn structures. To determine whether the exon 30 sequence might form similar structures, synthetic exon 30 peptide was dissolved in water and analyzed at 6-h intervals for the formation of precipitates. By 12 h, fine rod-like fibers were visible in the solution; and by 24 h, a substantial precipitate had formed. Staining of the precipitate with Congo red and analysis under polarized light showed a characteristic apple green birefringence associated with intercalation of the dye into regions of ␤-structure (26). Electron microscopy of the exon 30 aggregate (Fig. 5) revealed that the peptide formed filamentous aggregates with a diameter of ϳ7-10 nm, similar to those seen for lamprin and other amyloid-forming sequences (27). Duplicate experiments were performed in buffers containing physiologic salt (PBS). Although fibers were visible by eye in these experiments, the salt crystals that formed during the experiment inhibited the visualization of precipitated peptide by the various microscopy methods. Additionally, a change in incubation temperature from room temperature to 37°C did not substantially change the quantity or quality of fibers formed (data not shown). No fibers were ever observed with the exon 30 scrambled peptide.
Deletion of Exon 30 Decreases, but Does Not Ablate, Elastic Fiber Formation-The propensity of the exon 30 peptide to form amyloid-like aggregates and the failure of tropoelastin truncation constructs lacking exon 30 to associate with the PE cell matrix suggested that the exon 30 sequence might facilitate deposition of tropoelastin into the extracellular matrix and hence influence elastin fibrillogenesis. This possibility was tested by expressing, in the PE cell system, a tropoelastin construct with only exon 30 deleted (⌬30). Relative to the full-length construct (Fig. 6A), deposition of the ⌬30 protein into matrix fibers was greatly reduced (Fig. 6B). Rare fields were present, however, in which deposition of ⌬30 did take place. Expression of a construct containing an internal deletion of a different hydrophobic exon in the C-terminal domain, exon 33 (⌬33), yielded an elastin product that was deposited into fibers in the extracellular space. Although the fibers may be qualitatively different from those produced by the full-length construct (Fig. 6C), the fact that they were deposited shows that the result found for ⌬30 was not simply a consequence of exon deletion in general.
Expression of construct ⌬30 as a transgene in mice gave results similar to those observed with PE cells. The ⌬30 protein was detected in several tissues and organs of the transgenic mice, but the amount of incorporated protein was much less than what was found for the full-length transgene (Fig. 7).
Identification of a Hypersensitive Protease Site That May Influence Assembly-In analyzing tropoelastin secreted into the culture medium of transfected PE cells, we noticed several proteolytic breakdown products associated with the full-length molecule and mutation forms that were not evident in samples of the bTE-(1-28) protein (Fig. 3). Further repetitions of the experiment revealed varying degrees of degradation for all of the constructs from experiment to experiment. The only construct for which fragmentation was reproducibly decreased was bTE-(1-28). The molecular mass of the primary breakdown product of the full-length molecule was similar in size to the intact bTE-(1-28) protein (i.e. ϳ55 kDa), suggesting that a cleavage site may be present near the exon 28 border that, when cleaved, leads to removal of the C-terminal assembly sequence.
To investigate susceptible sites for proteolytic cleavage, we treated recombinant tropoelastin with plasmin for times ranging from 30 min to 2 h. Earlier studies determined that the degradation of tropoelastin isolated from developing chick aorta could be largely prevented by the inclusion of inhibitors of trypsin-like proteases (28,29). Given that plasminogen, the pro form of the enzyme plasmin, is a trypsin-like protease present in high concentrations in serum and that blood vessels are a key location for elastin assembly, the choice of this protease seemed to be a reasonable one. When recombinant tropoelastin was treated with plasmin at 37°C, a ladder of fragments similar to the naturally occurring breakdown products was detected (data not shown). At early time points, many fragments of various sizes were present, but by 2 h of treatment, only one resistant C-terminal fragment remained (Fig. 8), as evidenced by Western blotting of the proteolytic fragments probed with an antibody specific for the extreme C terminus of tropoelastin (CT e ) (16).
N-terminal sequencing of the various breakdown products revealed two major tropoelastin cleavage sites: one in exon 26 (R2AAAGLPAGVGP) corresponding to a relatively stable 55-kDa fragment and one in exon 6 (YK2AAAKAGAAG), where the arrow is the site of cleavage by plasmin. Once cleavage is initiated at these sites, complete degradation of the remaining molecule is rapid in the in vitro assay. DISCUSSION Using an in vitro assembly assay together with expression studies in transgenic mice, we have shown that the C-terminal region of elastin, encoded by exons 29 -36, contains an important assembly domain that directs the association of tropoelastin with microfibrils. Our focus on the C-terminal region of tropoelastin was guided by previous studies from our labora-tory suggesting that the C terminus of the molecule is important for its assembly into fibers. In these experiments, accumulation of elastin in the extracellular matrix of fetal bovine chondrocyte cells was inhibited when an antibody to the 17 amino acids encoded by the final exon of the elastin gene (exon 36) was added to the cultured cells. An antibody to a sequence in the N-terminal end of the molecule (exon 4/5) had no effect (16). Our initial mutation and deletion constructs expressed in PE cells were focused on exon 36, but we found that these amino acids are not required for elastin secretion or for its deposition into the extracellular matrix. In fact, mutation of one or both of the cysteine residues to alanine, deletion of the terminal RKRK sequence, or deletion of the entire exon 36 had no observable effect on the ability of the mutant protein to associate with microfibrils. Rather, a large portion of the Cterminal region consisting of exons 29 -36 had to be deleted before elastin accumulation in the extracellular matrix could be entirely inhibited. Because the truncated molecules were effectively secreted and were stable in the culture medium, these findings suggest that sequence 29 -36 contains epitopes, exclusive of exon 36, necessary for mediating the association of tropoelastin with microfibrils in the extracellular matrix.
Although the PE cell assay provides an excellent means to assess whether a given form of tropoelastin can associate with microfibrils in the extracellular matrix, its reliance on immunofluorescence co-localization provides only limited information about the quality of the resultant elastic fiber. We were not able to determine by this method, for example, whether any of the deletion constructs found to co-localize with microfibrils associated less efficiently than the full-length protein or bound in a way that precluded later assembly steps. However, that there may be quantitative differences is suggested by preliminary experiments in which desmosine levels were found to be lower in constructs lacking exon 36. 2 This result is consistent The blots were probed with antibody CT e , which is reactive against exon 36 and therefore detects fragments with an intact C terminus. Multiple fragments were initially generated by cleavage with plasmin; but by 2 h, only a single C-terminal fragment remained. with data reported by Hsiao et al. (30) showing decreased cross-linked elastin in the matrices of cells to which tropoelastin lacking exon 36 had been added.
The importance of the C-terminal sequence to tropoelastin assembly was confirmed using transgenic mice, which also allowed us to determine whether tropoelastin molecules bearing C-terminal mutations would lead to defects in elastic tissues through dominant-negative effects. Characterization of the different mouse lines showed that animals expressing wildtype bovine tropoelastin transgenes successfully incorporated the protein into the mouse elastic fiber. In contrast, no bTE-(1-28) protein could be detected in any of the tissues in which mRNA for the transgene was readily identified. Incorporation of a transgenic construct lacking exon 30 into elastic fibers was observed, but at levels significantly lower than those of the full-length protein. Each of these findings is consistent with results found in PE cells, where construct ⌬30 was found to associate with microfibrils at low levels relative to the fulllength transgene product, and construct 1-28 was found only in the culture medium. Because phenotypic and histopathological assessment of all the transgenic lines found no adverse effects of transgene expression, our results demonstrate that incorporation of the full-length bovine protein into the mouse fiber did not disrupt the assembly or function of the endogenous mouse elastin. Similarly, expression of assembly-incompetent forms of the protein (1-28 and ⌬30) did not interfere in a dominantnegative fashion with deposition of normal elastin. Similar results were obtained by Sechler et al. (31), who showed that transgenic mice expressing rat constructs (full-length and the naturally occurring splice variant ⌬13-15) as transgenes yield healthy animals with no observable harmful health effects.
Fine mapping of the matrix-binding activity of tropoelastin using synthetic peptides or expression constructs with single exon deletions localized sequences in and around exon 30 as being the major interactive site. The sequence encoded by exon 30 contains a tandem repeat (GGLG(V/A)) that resembles sequences found in other proteins that aggregate via ␤-sheet/␤turn structures. Examples include lamprin (a matrix protein of the lamprey annular cartilage) (24), spidroin (a spider dragline silk protein) (32), and various matrix proteins of the chorion or eggshell membrane of insects (33)(34)(35). These proteins assemble through the interdigitation of side chains belonging to residues present in short stretches of cross-␤-conformation (␤-turn/␤sheet) (25,36). Based on these structural similarities, Robson et al. (24) have suggested that sequences of this type contribute to self-aggregation of elastin monomers and alignment of polypeptide chains for cross-linking. Results presented in this report confirm that exon 30 does indeed contribute to tropoelastin assembly, but most likely in the context of interaction with microfibrils. Previous studies in our laboratory have demonstrated interactions between tropoelastin and small expression constructs containing the Pro-and Gly-rich regions of fibrillin-1 and fibrillin-2, respectively (37). Although we did identify the amino acid sequence responsible for tropoelastin binding in the fibrillin fragments, it is interesting to note that the glycinerich region of fibrillin-2 contains several repeats of the GGXGX sequence that could interact with exon 30 of tropoelastin via ␤-sheet/␤-turn structures.
The presence of an assembly site centered on exon 30 has important implications for understanding both normal elastin assembly and the molecular mechanisms of diseases arising from mutations within the elastin gene. For example, the characteristic mutations reported for autosomal dominant cutis laxa are single nucleotide deletions in exons 30 and 32 that result in out-of-frame sequence extending into the 3Ј-untranslated region with notable loss of the cysteine residues in exon 36. It has been speculated that this abnormal sequence might disrupt assembly of normal elastin in a dominant-negative fashion. In contrast, the majority of mutations in isolated SVAS are point mutations that produce premature termination sites. Urbá n et al. (13,38,39) have shown that many of these mutations produce an unstable mRNA transcript that is rapidly degraded through nonsense-mediated decay, resulting in elastin haploinsufficiency at the RNA level. In one case, however, mRNA and protein from a mutant elastin gene were identified in cells from an SVAS individual, although at reduced levels compared with the wild-type allele (13). If mRNA from a gene with a truncation mutation escapes nonsensemediated decay, the absence of a C-terminal assembly domain may preclude its incorporation into the growing fiber. The result would be haploinsufficiency at the protein level.
Removal of the C-terminal domain by proteolytic events outside the cell also appears to be a mechanism for regulating elastin assembly in instances of normal tissue growth and remodeling. For example, closure of the ductus arteriosis shortly after birth involves dissolution of the vessel's elastic laminae and ingrowth of intimal cushions. At the same time, tropoelastin secreted by ductus arteriosis smooth muscle cells is inhibited from forming new elastic fibers by proteolytic removal of the C-terminal domain (40). Although the exact cleavage site within the tropoelastin molecule was not characterized in the ductus arteriosis study, the size of the truncated protein (52 kDa) is similar to that of the fragment expected when tropoelastin is cleaved with plasmin at exon 26. This site in exon 26 has previously been described as being susceptible to cleavage by kallikrein and thrombin (41) and clearly defines a location on the surface of the tropoelastin molecule that is accessible to trypsin-like proteases. The identification of a hypersensitive protease site and the finding in the ductus arteriosis that the C-terminal region of tropoelastin can be specifically removed by proteolysis to inhibit assembly provide evidence for a possible mechanism for extracellular control of elastic fiber formation.
It has long been assumed that coacervation of tropoelastin is a crucial step in assembly of the elastic fiber (42)(43)(44). Our data suggest, however, that coacervation may not be involved in the initial steps of elastin assembly and that coacervation by itself cannot drive elastin assembly in tissues. Coacervation, an entropically driven, inverse temperature transition caused by the interaction of the hydrophobic domains in the molecule, occurs as tropoelastin monomers associate to form large aggregates. Several laboratories have shown that the large hydrophobic sequences in the middle of the molecule play a dominant role in the intermolecular interactions that occur during coacervation (43). If coacervation were the only requirement for assembly, then all of our deletion constructs would form fibers (or at least aggregates) because they all contain the critical middle hydrophobic sequences. This was clearly not the case, as constructs bTE(1-28) and ⌬30 did not undergo assembly when expressed in either PE cells or transgenic mice. Instead, our results argue for a model of nucleated assembly in which the tropoelastin monomer interacts with microfibrils in a process mediated by the exon 30 assembly domain. This process is initiated by ␤-structure interactions between exon 30 of tropoelastin and ␤-structure-containing regions on microfibrillar proteins such as the glycine-rich portion of fibrillin-2. Consequently, we believe that microfibrils are required to initiate or greatly enhance the rate of the assembly process through an interaction with the C-terminal region of tropoelastin. We cannot rule out a role for coacervation in directing tropoelastin self-interaction in later stages of fiber assembly; however, we do not know at what point microfibrils are no longer required.