The 10–12 nm diameter microfibrils of the extracellular matrix (ECM) impart both structural and regulatory properties to load-bearing connective tissues. The main protein component is the calcium-dependent glycoprotein fibrillin, which assembles into microfibrils at the cell surface in a highly regulated process involving specific proteolysis, multimerization and glycosaminoglycan interactions. In higher metazoans, microfibrils act as a framework for elastin deposition and modification, resulting in the formation of elastic fibres, but they can also occur in elastin-free tissues where they perform structural roles. Fibrillin microfibrils are further engaged in a number of cell matrix interactions such as with integrins, bone morphogenetic proteins (BMPs) and the large latent complex of transforming growth factor-β (TGFβ). Fibrillin-1 (FBN1) mutations are associated with a range of heritable connective disorders, including Marfan syndrome (MFS) and the acromelic dysplasias, suggesting that the roles of 10–12 nm diameter microfibrils are pleiotropic. In recent years the use of molecular, cellular and whole-organism studies has revealed that the microfibril is not just a structural component of the ECM, but through its network of cell and matrix interactions it can exert profound regulatory effects on cell function. In this review we assess what is known about the molecular properties of fibrillin that enable it to assemble into the 10–12 nm diameter microfibril and perform such diverse roles.
Connective tissue consists of specialized cell types present at low cell density surrounded by abundant extracellular matrix (ECM). It has evolved to withstand mechanical force, but, in addition to being tough and resilient in tissues such as bone, it may also need to be flexible and extensible, as in elastic arteries, skin and lungs. The precise combination of fibrous proteins and glycosaminoglycans/proteoglycans determines the structural properties of the ECM in each case. The 10–12 nm diameter microfibrils of connective tissue, of which fibrillin is the major component , are important highly conserved macromolecular assemblies found in both vertebrates and invertebrates (Figure 1) . In simple invertebrate metazoans such as jellyfish, the fibrillin microfibrils appear to confer elasticity and extensibility on the ECM . In vertebrate tissues such as the lung, blood vessels and skin, microfibrils form the periphery of the elastic fibre, acting as a scaffold for the deposition of elastin [4–6]. In addition, microfibrils can occur as elastin-independent networks in tissues such as the ciliary zonule, tendon, cornea and glomerulus where they provide tensile strength and have anchoring roles (Figure 1) [1,7,8]. More recently it has become clear that microfibrils play a role in tissue homoeostasis through specific interactions with growth factors, such as the bone morphogenetic proteins (BMPs) [9,10], growth and differentiation factors (GDFs)  and latent transforming growth factor-β-binding proteins (LTBPs) [11,12], cell-surface integrins (such as α5β1, αvβ3 and αvβ6) [13–17] and other ECM protein and proteoglycan components [18–28]. Several heritable connective tissue diseases are caused by mutations in the human fibrillin-1 (FBN1) gene including Marfan syndrome (MFS) , stiff skin syndrome  and the acromelic dysplasias , which give rise to very different patient phenotypes. These data, together with various mouse models [32–34], have highlighted the pleiotropic roles of fibrillin microfibrils in tissues, and have provided clues to the molecular mechanisms underlying their functional properties, which appear to lead to dysregulation of transforming growth factor-β (TGFβ) . This review highlights the recent molecular and cellular studies, which have provided new insights into the structural and regulatory roles of this important connective tissue fibril.
Microfibril tissue localization and composition
The simplest organisms in which fibrillin can be found are the cnidarians, which include jellyfish. Inspection of available genomes demonstrates the presence of fibrillin in many higher metazoan species, although they are absent from the model organisms Caenorhabditis elegans and Drosophila melanogaster [36,37]. The number of different fibrillin genes varies from organism to organism. Mammals generally have three fibrillin genes, although there are some exceptions such as rodents, where only two fibrillin isoforms are expressed. Simple metazoans such as cnidarians, arthropods and molluscs have one fibrillin gene, as do primitive vertebrates such as lampreys and hagfish. As is typical of fibrous proteins, the fibrillins exhibit a characteristic domain organization where, in the most part, individual exons encode single protein domains. The conservation of protein domain architecture is striking from simple to complex metazoans. Furthermore, substantial sequence conservation is observed at the amino acid level (Figure 1). Rotary shadowing electron microscopy of isolated microfibrils extracted from jellyfish  and human tissue  demonstrates that they both have a similar ultrastructure, forming beaded structures with a characteristic untensioned periodicity of 50–60 nm (Figure 1). It is therefore possible that conservation of domain architecture is required for the higher-order assembly of fibrillins into microfibrils and/or for its inherent biophysical properties, once assembled. It is interesting to note that connective tissue disease phenotypes associated with human genetic mutations in FBN1 that lead to in-frame exon skipping are usually quite severe . In contrast, deletion of the hyb1 domain in a mouse model was found to result in the assembly of apparently normal microfibril networks when analysed by immunofluorescence microscopy . This suggests that deletions in different parts of the molecule may have various effects on the register of fibrillin within the microfibrils.
STRUCTURE OF FIBRILLIN DOMAINS
Fibrillin-1 was identified in 1986 as a 350 kDa glycoprotein component of 10–12 nm diameter microfibrils . Cloning of FBN1 [41–43] and highly homologous FBN2  showed them to encode proteins belonging to a small superfamily characterized by the presence of an 8-cysteine, or TGFβ-binding protein-like (TB), domain (Figure 2). There are a total of three isoforms of fibrillin now identified in humans (fibrillin-1, -2 and -3) [42–45], all having an evolutionarily conserved domain organization that includes the fibrillin unique N-terminus (FUN), epidermal growth factor (EGF)-like, calcium-binding epidermal growth factor (cbEGF), TB and hybrid (hyb) domains, together with a small 2-Cys motif. The main difference between isoforms at the sequence level is the amino acid composition of a distinct region between domains TB1 and EGF4, which is proline-rich in fibrillin-1, glycine-rich in fibrillin-2 and proline/glycine-rich in fibrillin-3. There are further differences in amino acid sequence suggesting that their protein–protein/glycosaminoglycan interactions may be specific to each isoform . High-resolution structures are available for examples of each of the domain types from fibrillin-1, except the proline-rich, 2-Cys and propeptide regions, in pairwise or triple domain arrangements, giving insight into the native architecture of fibrillin-1 (Figure 2). All fibrillin-1 domains, with the exception of the proline-rich region and the N- and C-terminal propeptides, are disulfide-rich, with little hydrophobic core, and many of them make extensive hydrophobic interfaces with their adjacent domains. Thus, fibrillin-1 has evolved to withstand mechanical forces applied to connective tissues, and, as a consequence of its domain structure, is highly protease-resistant .
Fibrillin structure and microfibril organization
There are seven TB domains within the domain organization of fibrillin-1 and they are found solely within the fibrillin/LTBP family of proteins. Each TB domain has four disulfide bonds in a 1–3, 2–6, 4–7, 5–8 arrangement, with a conserved aromatic residue located within its core [15,48]. Apart from TB1, which has the proline-rich domain at its C-terminus, all other TB domains are flanked by cbEGF domains (Figure 2). In the cbEGF22–TB4–cbEGF23 fragment of human fibrillin-1 solved by X-ray crystallography , the TB domain forms extensive interfaces with the cbEGF domain either side of its globular core suggesting that, in the absence of any tensional force, this region forms a near-linear unit with N- and C-termini protruding at each end (Figure 2). The TB4 domain contains a single functional ‘RGD’ integrin-binding site on a solvent-exposed loop, which mediates binding with different affinities to integrins α5β1, αvβ3 and αvβ6 [13–17]. Intriguingly, the C-terminal linker sequence that connects TB4 to cbEGF23 forms a β-hairpin. This has been postulated to act as an elastic spring under appropriate conditions, able to absorb some of the changes in tension experienced by dynamic connective tissues . Calcium-binding measurements of all TB–cbEGF pairs in fibrillin-1, with the exception of TB6–cbEGF32, show them to exhibit high/moderate-affinity calcium binding (Kd in nanomolar or micromolar range) [49–51], suggesting that they will adopt a similar structural arrangement to cbEGF22–TB4–cbEGF23. TB6–cbEGF32, however, has very different properties. Although the fold of TB6 is conserved, cbEGF32 lacks the high/moderate -affinity calcium binding usually observed within TB–cbEGF pairs. This is due to the absence of an extensive hydrophobic interface between the two domains [49,52]. The significance of this site of flexibility is unknown; however, it is important for function since a missense mutation that impairs the native calcium-binding properties of cbEGF32 by substitution of serine for an asparagine side chain ligand (N2144S) causes MFS . It is possible that the weak Ca2+ affinity (Kd ∼1.6 mM) normally observed at this site imparts specific biophysical properties to the assembled microfibril, or provides limited flexibility to fibrillin to aid protein–protein and/or protein–proteoglycan interactions.
There are two hybrid (hyb) domains within fibrillin and each has a 1–3, 2–5, 4–6, 7–8 disulfide arrangement. The structure of the triple domain fragment cbEGF9–hyb2–cbEGF10 from fibrillin-1 (Figure 2) was solved by X-ray crystallography and showed the hybrid domain to have a combination of the structural properties of the TB domain and a cbEGF domain , consistent with its suggested evolution from the fusion of a TB–cbEGF domain pair [42,43]. Similar to the TB4 domain, hyb2 forms extensive interfaces with its flanking cbEGF domains and its C-terminal linker is also arranged in a β-hairpin. The first hybrid domain, hyb1, has an absolutely conserved additional cysteine residue (Figure 1) that may be solvent-exposed and has been proposed to participate in higher-order intermolecular interactions in the microfibril assembly process . It should be noted, however, that a hyb1-deletion mouse model showed normal microfibril assembly .
EGF/cbEGF domains are common extracellular protein domains, which usually share a 1–3, 2–4, 5–6 disulfide bond arrangement and a central β-hairpin secondary structure element [56–59]. They are found much more widely in other extracellular/cell-surface proteins than the TB/hybrid domains . In human fibrillins, 43 out of the 47 EGF domains contain a calcium-binding consensus sequence–D/N-X-E/Q-Xm-D/N*-Xn-Y/F–, where m and n indicate a variable number of amino acid residues and * denotes possible β-hydroxylation of the aspartate/asparagine side chain. EGF domains with this consensus bind Ca2+ in pentagonal bipyramid geometry via oxygen atoms from side-chain residues, backbone carbonyl atoms or water. In addition a backbone carbonyl group of a conserved serine residue may also contribute a ligand. EGF domains often occur as multiple tandem repeats.
Ca2+ has an important role in maintaining the architecture of fibrillin and microfibrils: it binds at the N-terminus of each cbEGF domain within a cbEGF pair and, in conjunction with hydrophobic packing interactions between a conserved aromatic residue and the side chain of an ‘XG’ dipeptide on the central β-hairpin, usually confers rigidity on the domain interface and an overall near-linear arrangement of the domain pair (Figure 2). The single amino acid linker between Cys6 of an N-terminal cbEGF domain and the first calcium-binding residue of the C-terminal cbEGF domain of a pair restricts conformational flexibility at the interface, conferring specific twist and tilt angles. Rotary shadowing electron microscopy of microfibrils has demonstrated that their characteristic extended architecture is dependent on the presence of Ca2+ , consistent with this metal ion conferring a rod-like arrangement on tandem repeats of cbEGF domains . The importance of Ca2+ in stabilizing the extended conformation of microfibrils has also been demonstrated in situ by X-ray diffraction studies of bovine zonular filaments .
Most calcium-binding sites within homologous cbEGF domain pairs of fibrillin-1 exhibit affinities in the 10–200 μM range but there are some very-high-affinity sites (Kd, in nanomolar) observed with TB–cbEGF and hyb–cbEGF pairs of fibrillin-1 [49,54]. The significance of such sites is unknown; they may aid folding of fibrillin in the ER or restrict/regulate the extensibility of this region of the polypeptide when subjected to tensional force (as occurs in elastic tissues such as aorta/lung). Many of the intermolecular interactions involving fibrillin are also Ca2+- dependent, including those with MAGP-1 [22,25], fibulins-2, -4 and -5 [19,24], and aggrecan and versican . Calcium is also required for homotypic and heterotypic interactions between fibrillin-1 and fibrillin-2 [64,65]. This suggests that the Ca2+- stabilized interface, conferred by covalent linkage of the cbEGF domain to its N-terminal neighbouring domain, is important for molecular recognition. As a consequence of its Ca2+-stabilized and disulfide-rich molecular structure, fibrillin is resistant to proteolysis . This makes it well adapted to an ECM environment, which may undergo little turnover in many tissues of the adult organism under normal physiological conditions. The crucial role played by Ca2+ in fibrillin and microfibril function is illustrated by the large number of missense mutations identified in MFS patients, which result in substitution of calcium-binding residues within a single cbEGF domain.
The FUN domain adopts a compact conformation with little secondary structure . It comprises two loops stabilized by 1–3 and 2–4 disulfide bonds, which pack against an N-terminal segment. A conserved PGW sequence in the FUN domain (Figure 1) is involved in packing interactions with the adjacent EGF1 domain, rigidifying the structure of this region. Located at the start of the FUN domain is a sequence of glycine residues which confer flexibility; this region has been used as a site of insertion of a GFP tag such that the resulting GFP–fibrillin fusion protein can be utilized in cellular assays of microfibril assembly . The FUN domain forms an interface with the adjacent EGF1; likewise EGF2 and EGF3 also form a stable interface, but the linker between EGF1 and EGF 2 confers flexibility on this region (Figure 2) . This region contains both a recognition site for the C-terminus of fibrillin-1 [19,68] and a binding site for heparan sulfate proteoglycans (HSPGs) [18,66,69].
N- and C-terminal regions
Exon 1 of human fibrillin-1 encodes the signal peptide and the N-terminal propeptide of 17 residues, which contains a furin/PACE (paired basic amino acid cleaving enzyme) protease- cleavage site. Although the N-terminal propeptide sequence is not conserved, the furin/PACE-cleavage site R-X-K/R-R is conserved evolutionarily across the fibrillins [36,37]. The sequence beyond cbEGF43 at the C-terminus is unique at the amino acid level, and its structure remains to be solved. Between cbEGF43 and the propeptide there is a highly conserved 2-Cys motif of unknown function followed by another furin/PACE recognition site , cleavage of which is known to be essential for microfibril assembly at the cell surface [71–73]. Mutation of this cleavage site, either as a result of a disease-causing substitution or by protein engineering of a recombinant fibrillin-1 to block cleavage of the ∼140 residue C-terminal propeptide in the ECM, prevents incorporation of fibrillin into the microfibril [67,74]. The importance of this region for fibrillin-1 function can be seen by the number of missense and nonsense mutations associated with MFS or related disorders [75–78]. These alter the propeptide sequence beyond the furin-cleavage site, and suggest that this region plays an important role in regulating microfibril assembly.
The main distinguishing feature of the three mammalian isoforms of fibrillin is a region between TB1 and EGF4 (Figure 2), which is proline-rich in fibrillin-1, glycine-rich in fibrillin-2 and proline/glycine-rich in fibrillin-3. It is the only region within the mature processed form of fibrillin that is not disulfide-rich. The proline-rich region is unlikely to be structured , but could impart flexibility to the N-terminal region of fibrillin-1. It is postulated to interact with tropoelastin  during elastic fibre assembly, which is consistent with the absence of conservation of this region in organisms which lack elastin.
FIBRILLIN-1 ORGANIZATION IN MICROFIBRILS
The central problem here is determining how fibrillin monomers, which have a length of ∼150 nm, are arranged in microfibrils that have an average bead-to-bead periodicity of 55 nm [38,80]. At present, two broad classes of models of microfibril organization have been proposed: a pleated model in which individual monomers are folded to fit within one interbead repeat [81,82], and extended models in which individual monomers span two or more interbead repeats [15,83]. As microfibrils are dynamic structures, models of microfibril organization need to provide not only an explanation of the arrangement of fibrillin monomers, but also a mechanism for their mechanical properties.
The high-resolution structures available for fibrillin-1 domains in their native context, together with calcium-binding analyses of heterologous and homologous cbEGF pairs, facilitate homology modelling studies of fibrillin-1 (Figure 2). At least two flexible domain interfaces, EGF1–EGF2 and TB6–cbEGF32, have been identified experimentally [52,66]. The proline-rich region is also thought to be a flexible hinge-like region based on its amino acid sequence and variability through evolution, although structural studies on this region are still required. The bulk of the molecule consists of tandem repeats of cbEGF domains that would form a near-linear rigid structure in the presence of calcium. Based on the available structural data, the TB and hybrid domains that interrupt the tandem cbEGF repeats are expected to introduce slight ‘kinks’ into the structure as a result of pairwise interactions (Figure 2). Extensive hydrophobic interfaces, stabilized by calcium binding to the cbEGF domain, are formed at most TB–cbEGF and hyb–cbEGF interfaces [15,49,54].
Based on the observation of specific N- to C-terminal interactions [64,65] and immunolocalization studies [84,85], all current models of microfibril organization agree on the positioning of the fibrillin N- and C-termini, which overlap in the ‘bead’ structure (Figure 2). The pleated model of microfibril organization is based on data from scanning transmission electron microscopy mass mapping, antibody mapping, SAXS data derived from recombinant fibrillin-1 fragments and electron tomography [81,85]. In this model, the region between domains TB4 and TB6 is highly compacted and the more globular TB and hybrid domains are predicted to correspond with the globular features of the microfibril observed by electron microscopy. Although this model provides a mechanism for microfibril extensibility, based on the rearrangement of domain interactions that lead to a more extended conformation under tension , the mechanism of structural restoration on relaxation is unclear. The model also requires either specific interactions between distantly spaced regions of the molecule or long-range structural features that are not currently supported by high-resolution data.
Linear models of microfibril organization are largely based on X-ray crystallography and NMR data, and on the mapping of antibody-binding and protease-cleavage sites within fibrillin and the microfibrils [15,52,57,59,83]. In the most current version fibrillin monomers are organized so that the N- and C-termini are located within the beaded structure of the microfibril. The N-terminus overlays a core formed by the C-terminal domains, consistent with the observation that the C-terminus is likely to initiate microfibril assembly by associating into bead-like structures [68,84]. Monomers are also staggered so that the ‘neonatal’ region, where missense mutations are often associated with a severe neonatal form of MFS, is also positioned near the bead. This model is supported by all the current high-resolution structural data and data from isolated fibrillin-1 molecules viewed by electron microscopy, which appear to show molecules with an extended conformation and a length of 140–150 nm [64,87,88]. The mechanism of extension and recoil in this model may involve transient localized unfolding and refolding at domain interfaces driven by the elastic β-hairpin linkers, extensive hydrophobic interactions and high calcium affinities seen at heterologous TB–cbEGF domain interfaces [15,89]. Microfibrils have been shown to be fully extended, without additional tension under physiological conditions, under conditions of saturating calcium , so cbEGF–cbEGF domain interfaces are unlikely to contribute to increases in interbead distance on extension. TB–cbEGF domain interfaces may contribute up to 5 nm of extension each, or 30 nm from all six domain pairs and the two hyb–cbEGF interfaces may contribute an additional 7 nm on extension [15,54]. This would be consistent with data suggesting that reversible extensibility of microfibrils is limited to interbead periodicities up to ∼100 nm [90,91] rather than the 140–150 nm observed for isolated fibrillin molecules.
ASSEMBLY OF FIBRILLIN-1
Progress utilizing a wide range of molecular and cellular assays has been made in understanding the pathway by which fibrillin assembles into the 10–12 nm microfibril. Recombinant protein studies indicated a specific interaction between N- and C-terminal regions of fibrillin-1 [64,65]. These results were recapitulated in pull-down experiments using limited N- and C-terminal fragments, FUN–EGF3 and cbEGF41–cbEGF43 . In addition to this specific interaction, Hubmacher et al.  demonstrated that a C-terminal fragment of fibrillin-1, from domain cbEGF22 to the C-terminus, was able to multimerize. When visualized by rotary shadowing electron microscopy, the ultrastructure of this material resembled a product of protease digestion obtained from microfibrils extracted from human amnion, suggesting that the multimer may represent an intermediate of fibrillin assembly. The assembly process could be inhibited by the addition of heparin sulfate. It is interesting to note that, in the high-resolution structural analysis of the FUN–EGF3 region, the C-terminal fibrillin-1-binding site was observed to overlap with a heparin-binding site . This suggested that HSPG interactions might regulate the assembly of the microfibril at the cell surface thus providing a molecular basis for at least some of the observed inhibitory effect of heparin sulfate on microfibril assembly. Recent work has shown that cell-surface HSPGs are likely to control microfibril assembly by preferentially binding to C-terminally multimerized fibrillins and inhibiting N- to C-terminal interactions .
Other studies further showed that mammalian fibrillins in mesenchymal tissues are dependent on the presence of a fibronectin matrix for efficient microfibril assembly [93–95], although microfibril assembly by epithelial cells does not appear to be fibronectin-dependent . Interestingly, organisms such as jellyfish  and sea cucumbers , which produce microfibrils, do not express fibronectin homologues. This absence of fibronectin homologues may reflect a requirement for a greater degree of organization in the assembly of microfibrillar tissues in the more complex extracellular matrices of higher metazoans.
Independent experiments, utilizing co-culture assays between human embryonic kidney (HEK)-293 cells expressing recombinant fibrillin-1 variants and fibroblasts, have been used to examine the requirements for microfibril assembly [67,98]. The HEK-293 cells are unable to assemble fibrillin into microfibrils, but the secreted recombinant protein is incorporated into fibroblast-assembled microfibrils instead, and can be distinguished by the presence of the GFP epitope  or through the use of species-specific fibrillin-1 antibodies . Variants of the GFP–fibrillin-1 fusion construct (Figure 3) were used to demonstrate that covalent linkage of the C-terminal propeptide was required for secretion of full -length fibrillin from cells, but that subsequent cleavage of the propeptide by furin was essential in order for fibrillin to be incorporated into the nascent microfibril. For this experiment, the dibasic ‘RR’ sequence that forms part of the fibrillin furin/PACE recognition sequence  was replaced by ‘AA’ and shown to result in the secretion of a higher molecular weight form similar to that of unprocessed fibrillin-1. Immunofluorescence analysis of co-cultures demonstrated that while wild type (WT) GFP–fibrillin-1 co-localized with fibroblast-assembled microfibrils, the RRAA double mutant did not (Figure 3). This is consistent with earlier studies showing that mutations affecting propeptide cleavage influence microfibril assembly and that propeptide cleavage is a prerequisite for microfibril assembly [72,74].
In addition, it was further shown that domains cbEGF41–cbEGF43, together with the native propeptide, were required in order for full-length recombinant fibrillin-1 to be secreted out of the cell, suggesting an inter-dependence of the two regions. Both regions may be required for co-operative folding and/or intramolecular interactions, which have evolved to prevent premature assembly of fibrillin-1 into microfibrils within the cell. Endogenous secretion of native fibrillin is unaffected by the co-expression of a recombinant fibrillin-1 with either a C-terminal propeptide deletion or cbEGF41–cbEGF43 deletion, which causes ER retention. This suggests that full-length profibrillin monomers within the secretory pathway are prevented from premature assembly by their own intramolecular structure. Interestingly, shorter recombinant C-terminal fibrillin-1 fragments lacking the propeptide domain (rF6H and rF6H ΔC) have been purified from the culture medium of stably transfected HEK-293 cells [22,68], in contrast with what would be expected from the data obtained from the GFP–fibrillin system . The reason for this apparent difference is unknown, but may relate to differences in expression levels or the presence of the N-terminus in the GFP–fibrillin construct.
Collectively, these findings suggest a model for ordered assembly of microfibrils dependent on the presence of cell-surface HSPGs, regulated cleavage of the propeptide and C-terminal multimerization (Figure 3). The dependency of fibril assembly on the fibronectin network in higher metazoans probably reflects the complexity of ECM in these species, and has also been observed for other fibrous proteins such as collagen type I .
In addition to the homotypic interactions between different fibrillin molecules within the microfibril, heterotypic interactions have been demonstrated with a variety of matrix/cell-surface components (Figure 4). From the perspective of the known molecular architecture of fibrillin-1, it is interesting to consider how such a large collection of diverse ligands can bind specifically. The elongated architecture of the protein clearly provides a large surface area for binding, and the sequence variation observed in the surface loops of the various domain types allows selectivity of binding. Furthermore, the observed regions of flexibility (EGF1–EGF2 and TB6–cbEGF32) may facilitate the ability of different combinations of accessory proteins and proteoglycans to bind to assembled fibrillin in a spatio-temporal manner. Interestingly, many of the interactions identified involve the N-terminal region of fibrillin-1, which contains both a heparin-binding site and a site of flexibility (Figure 4).
Cell–matrix interactions involving microfibrils
Fibronectin has been demonstrated to be essential for the assembly of microfibrils by mesenchymal cells in higher metazoans [93–95]. In contrast, retinal pigmented epithelial cells that express high levels of epithelial cadherin (E-cadherin) were recently reported to assemble fibrillin-1 independently of fibronectin, relying instead on syndecan-4 . The major interaction site was identified within the C-terminal regions of fibrillin-1, -2 and -3. Within fibronectin, the binding site was localized to FNI6–FNI9 . Fibronectin appears to act as a facilitator of microfibril assembly, but does not always remain bound to fibrillin, since the two networks can also be seen independently of each other when observed by immunoelectron microscopy of mature human skin fibroblast cultures .
A number of integrins, including αvβ3, α5β1 and αvβ6, have been shown to bind with various affinities to recombinant fibrillin-1 fragments containing TB4 (Figure 4) [13–17]. Interactions between α8 integrins and fibrillin-1 have also been suggested by experiments using mesangial cells of the kidney glomerulus . The full repertoire of integrins has not been exhaustively tested due to the difficulty of obtaining purified reagents, and it is possible that other integrins may be added to this list. Integrin binding to fibrillin is of known physiological significance since a number of FBN1 mutations that cause stiff skin syndrome all map to the RGD-containing integrin-binding domain TB4 of fibrillin-1, and a mouse model of systemic scleroderma was created by an RGE knockin within TB4 . In both cases a skin fibrosis results. Prior to these studies, the tight skin phenotype of the tsk mouse had been attributed to a large in-frame duplication of the TB4 region that results in expression of a fibrillin-1 polypeptide containing two TB4 domains with RGD sequences in each [102,103]. Thus maintenance of the appropriate level of fibrillin/integrin engagement appears to play a role in regulating normal ECM production in a tissue-specific manner and, when disrupted, leads to ECM overproduction or a fibrosis response in specific tissues such as the skin. Although it is clear from these studies that defective integrin/fibrillin interactions play a role in pathogenesis, it is not clear which integrin(s) and signalling pathway(s) are involved in regulating matrix production, or why these are tissue-specific. Furthermore, maintenance of the ratio of fibrillin/integrin interaction appears to be important, since mutations leading to a loss of microfibrils in tissues (such as in MFS) do not trigger the fibrotic response in skin, unlike stiff skin syndrome or acromelic dysplasias where microfibril assembly is not obviously affected .
Microfibril assembly by dermal fibroblasts or in the HEK-293/fibroblast co-culture system can be inhibited by the addition of heparin or heparin sulfate analogues [26,105]. Although many ECM molecules including fibronectin bind to such glycosaminoglycan chains, it is notable that fibrillin-1 itself has seven heparin-binding sites of variable affinity within the FUN domain, cbEGF38 C-terminal furin-cleavage site, cbEGF12–cbEGF14 and TB5–cbEGF25 (Figure 4) [18,26,66,69,92,105]. Fibrillin could potentially interact with heparan sulfate chains presented by cell-surface proteoglycans such as syndecans or secreted proteoglycans such as perlecan, both of which are evolutionarily conserved from humans to cnidarians, and competition between cell- surface proteoglycans and the C-terminal domain of fibrillin-1 for the FUN domain is thought to play a role in regulating fibrillin assembly .
The LTBPs (four isoforms in humans) are closely related to the fibrillins since they contain TB, hybrid and EGF/cbEGF domains . LTBP-1, -2 and -4 bind fibrillin-1 via their C-terminal region (Figure 4) [11,12]. This interaction may be important in regulating activation of latent TGFβ in the ECM, since TGFβ dysregulation is observed in all diseases associated with FBN1 mutations [31,35,107]. Domain TB2 of LTBP-1, -3 and -4 forms a covalently linked complex with the latency-associated propeptide (LAP) [108–111], which itself binds TGFβ in a non-covalent manner . TGFβ activation can be achieved by a range of physiological activators, but integrin binding to RGD motifs in the LAP complex appears to be the most important, generating the force required to pull the small latent complex apart . The mechanisms by which FBN1 mutations result in TGFβ dysregulation are unclear and probably reflect changes in the biophysical properties of the matrix and cellular interactions. It is also possible, although still to be demonstrated, that fibrillin microfibrils contribute to the necessary anchorage to the large latent complex required for the integrin to pull against during the activation process.
Interactions with other growth factors
A number of other growth factors including BMP-2, BMP-4, BMP-5, BMP-7, BMP-10 and GDF-5 have been observed to bind to recombinant fibrillin-1 fragments in the N-terminal region with various affinities (Figure 4) [9,10,114]. At present, it is not clear whether all possible interactions at the N-terminus of fibrillin-1 take place simultaneously considering the relative sizes of the FUN–EGF3 fragment of fibrillin-1 and a pro-BMP molecule (Figure 4B). As a polymeric structure, however, assembled microfibrils provide multiple binding sites so simultaneous interactions would be possible across the structure. Like TGFβ, BMP-10 growth factor/prodomain complexes are latent and unable to signal until they are activated by cleavage by BMP-1. BMP-4 and BMP-5, in contrast, can activate their receptors with their propeptides still intact . Binding to fibrillin may be another mechanism of conferring latency on these proteins, until displaced by protease cleavage or competing interactions. In a recent study using fibrillin-2 and BMP-7 mouse models, Sengle et al.  showed that the microfibril environment controls BMP signalling during muscle development, with interactions between fibrillin-2 and the prodomain of the BMP-7 complex conferring latency on the growth factor.
Fibrillin-1 has been shown recently to bind members of the ADAMTS(L) [a disintegrin and metalloproteinase with thrombospondin motifs(-like)] superfamily of extracellular proteases and proteins [23,28,31,114,117,118], suggesting a role for microfibrils in the regulation of tissue development and remodelling. Mutations affecting these proteins cause recessive forms of microfibril-associated disorders, including Weill–Marchesani syndrome (ADAMTS10) , Weill–Marchesani-like syndrome (ADAMTS17) , geleophysic dysplasia (ADAMTSL2)  and isolated ectopia lentis (ADAMTSL4)  and whose symptoms overlap with the disease spectrum caused by dominant FBN1 mutations. Conditioned medium from fibroblasts isolated from geleophysic dysplasia patients contains increased levels of both total and active TGFβ relative to unaffected controls . The mechanism linking a loss of ADAMTSL2 function with increased TGFβ signalling is not clear; however, ADAMTSL2 has been shown to interact with LTBP-1 . This suggests that ADAMTSL2 is part of a TGFβ regulatory network involving both fibrillin-1 and LTBP-1. Although the roles of members of the ADAMTS(L) family in microfibril biology are not yet clear, they may play a role in ‘functionalizing’ microfibrils with specialized activities in different tissues . In addition, some members of this family of proteins, including ADAMTSL4 , ADAMTSL6  and ADAMTS10 , are able to modulate microfibril assembly.
The ECM is an insoluble polymer with many components rapidly assembling into fibrous networks, which then form numerous cell–matrix interactions. The 10–12 nm microfibrils, which are largely composed of macromolecular assemblies of fibrillins are no exception. Despite this, many aspects of fibrillin-1 and microfibril structure have been solved, and the assembly process is starting to be understood. Identification of the high-resolution structures of fibrillin domains is almost complete, with just the C-terminal propeptide region remaining as a target. Reasonable models for microfibril structures are available that allow us to understand how these fibrils have evolved to facilitate cell-matrix interactions, act as docking stations for a variety of growth factors and perform structural roles in a range of tissues. Our next challenge is to unravel the signalling pathways that are influenced directly or indirectly by microfibrils, and which become perturbed in different disease states.
The advent of new genome editing tools, used in combination with other molecular, cellular and whole-organism techniques, should facilitate dissection of the cellular effects of microfibrils. The next few years should prove to be exciting times in ECM research, and allow us to gain new fundamental mechanistic insight into matrix regulation as well as understanding pathogenesis of connective tissue disorders.
This work was supported by the Arthritis Research UK [grant number 20785 (to S.A.J. and P.A.H.)]; and the Medical Research Council [grant number MR/M009831/1 (to Christina Redfield and P.A.H.)].
a disintegrin and metalloproteinase with thrombospondin motifs(-like)
bone morphogenetic protein
calcium-binding epidermal growth factor
epidermal growth factor
fibrillin unique N-terminus
growth and differentiation factor
human embryonic kidney
heparan sulfate proteoglycan
latent transforming growth factor-β-binding protein
paired basic amino acid cleaving enzyme
transforming growth factor-β