Fibrillar Collagens

The collagen molecule, the basic unit of collagen fibers, has an asymmetrical, rodlike structure composed of three polypeptide chains called α chains. Because of the Gly-X-Y repeating units and their stereochemistry, each α chain forms a minor helix (Fig. 4-2). Three α chains wind around a common axis to form a right-handed triple helix. In some collagens, all three α chains are identical, while in others two or three unique α chains form the triple-helical molecule. Type I and type III collagens are the most abundant collagens in the blood vessel and together form the striated fibrils. With the exception of types XXV and XXVII, fibrillar collagens form an uninterrupted triple-helical domain of approximately 300 nm. The type I collagen α chains contain 338 Gly-X-Y repeats, and there are 341 such triplets in type III α chains. At both the NH₂ and COOH ends of each α chain are short segments of nonhelical sequences of approximately 15 to 20 amino acid residues, referred to as telopeptides.

Figure 4-2 An overview of the main steps involved in the synthesis of fibril-forming collagens.

The α-polypeptide chains are synthesized on membrane-bound ribosomes and secreted into the lumen of the endoplasmic reticulum (ER). The main steps in collagen biosynthesis are (i) cleavage of the signal peptide (not shown), (ii) hydroxylation of specific proline and lysine residues, (iii) glycosylation of certain asparagine residues in the C-peptide, and (iv) formation of intramolecular and intermolecular disulfide bonds. A nucleus for the assembly of the triple helix is formed in the C-terminal region after the C propeptides of three α-chains become registered with each other and ~ 100 proline residues in each α-chain have been hydroxylated to 4-hydroxyproline. The triple helix formation proceeds toward the N-terminus in a zipper-like fashion. Procollagen molecules are transported from the ER to Golgi, where they begin to associate laterally and exit the cell via secretory vesicles. This is followed by cleavage of N and C propeptides, spontaneous self-assembly of the collagen molecules into fibrils, and formation of cross-links.

(From Myllyharju J, Kivirikko KI: Collagens, modifying enzymes and their mutations in humans, flies and worms. Trends Genet 20:33–43, 2004.)

Because of their similarities, type I and type III collagens are discussed together here. The type I collagen molecule is a heterotrimer of two identical α chains, α1(I), and a different α chain, α2(I), and has the chain structure [α1(I)]₂α2(I)]. The type III collagen molecule is formed by three identical α chains and has the chain structure [αI(III)]₃. The helical domain of the α chain contains a repeating triplet sequence of [Gly-X-Y]_n, where X and Y may be any amino acid but are most frequently proline or hydroxyproline. The amino acid residues in the Y position are nearly always hydroxylated (4-hydroxyproline). The configuration of the amino acids forces the α chain to assume a left-handed helix, thus allowing α chains to form a right-handed supercoil with a one-amino-acid stagger between adjacent chains. The presence of glycine (without a bulky side chain) as every third amino acid is critical because it will occupy the center position within the triple helix. Substitution of any other amino acid for glycine in the Gly-X-Y leads to disruption of the triple helix.

The collagen triple helix is further stabilized by interchain hydrogen bonds contributed by hydroxyproline residues. Thus, the collagen molecule is a long cylindrical rod with dimensions of 1.5 nm  ×  300 nm. Under physiological conditions of ionic strength, pH, and temperature, collagen molecules spontaneously aggregate into striated fibrils. Fibril formation occurs by lateral aggregation of collagen molecules, in which each neighboring row of molecules is displaced along its long axis by a distance of 68 nm. In addition, within the same row, there is a gap of approximately 40 nm between the end of one molecule and the beginning of the next (see Figs. 4-1 and 4-2). The short nonhelical telopeptides at the NH₂ and COOH ends of each α chain are located in the gap or hole zone of the fibril and are therefore accessible to enzymes that regulate collagen cross-linking.

Network-Forming Collagens

As shown in Figure 4-1, collagen types IV (α1-α6 chains), VI (α1-α5 chains), VIII (α1-α2 chains) and X are known to form networks in the ECM of basement membranes. The supramolecular organization and function of type IV collagen has been extensively characterized. Six different α polypeptide chains of collagen IV are each encoded by an evolutionary conserved gene. The amino and carboxyl propeptides of type IV collagen remain as integral parts of the molecules when they are deposited in the basement membrane. As a result, rather than forming a quarter-stagger, side-by-side alignment of individual molecules, as seen in types I, II, and III collagens, type IV collagen α chains form chicken-wire structures by end-to-end associations stabilized by lysine-derived cross-linking and interchain disulfide bonds (Fig. 4-3). The α1(IV) and α2(IV) collagen chains are more closely related to each other than to α3(IV)1, α4(IV), α5(IV), and α6(VI); the latter share a high degree of sequence homology with each other. The amino terminal domains of α1(IV) and α2(IV) collagen chains are 143 and 167 amino acids, respectively; the NH₂-termini of the other four α chains are much smaller (ranging in size from 13 to 19 amino acids). Theoretically, all six α chains of type IV collagen may combine randomly to generate 56 unique triple-helical permutations. However, as shown in Figure 4-4, in vascular basement membranes the most common composition of triple-helical fibrils is [α1(IV)1]₂ α2(IV). The [α3(IV)1]₂ α4(IV) and [α5(IV)1]₂ α6(IV) are also present in basement membrane.^9,10

Figure 4-3 A, Linear structures of human collagen IV α-chains. Six different genes encode collagen IV α-chains. Each polypeptide is composed of three distinct domains: a cysteine-rich N-terminal 7 S domain, a central triple-helical domain with multiple small interruptions (boxes), and a globular C-terminal noncollagenous NCl domain. The NCl and central triple-helical domains are of an equivalent size, whereas 7 S domains are shorter in the cases of α3, α4, α5, and α6 compared with α1 and α2. On the basis of sequence homology, type IV collagen α-chains can be divided in two groups: the α1-like (α1, α2, α5) and the α2-like (α2, α4, α6). B, Assembly of collagen IV α chains. Assembly of trimers is dependent on the association of NCl domains, followed by formation of triple-helical structure and 7 S domains in a spider-shaped structure; the two trimers interact head-to-head through their NCl domains, forming a sheet structure. Several trimers can also lace together along their triple-helical domains, thickening the structure.

(Adapted from Company of Biologists Ltd., Ortega N, Werb Z: New functional roles for noncollagenous domains of basement membrane collagens. J Cell Sci 115:4201, 2002.)

Figure 4-4 Localization of the α1•α2 and α1•α2•α5•α6 networks of type IV collagen in vascular basement membranes (BMs).

Schematic diagram of a large artery (aorta) depicts its multilayered structure (right). Endothelial cells (En) rest on a subendothelial BM, which contains the α1•α2(IV) collagen network (right). Smooth muscle cells (SMCs) in the media are surrounded by smooth muscle BM and are sandwiched between an internal and external elastic lamina (IEL and EEL, respectively). The α1•α2 and α1•α2•α5•α6 networks of type IV collagen coexist in smooth muscle BM (right).

(Adapted from Borza DB, Bondar O, Ninomya Y, et al: The NCl domain of collagen IV encodes a novel network composed of the alpha-1, alpha-2, alpha-5, and alpha-6 chains in smooth muscle basement membranes. J Biol Chem 276:28532, 2001.)

Organization of the type IV collagen genes is unusual. The COLA4A1 and COLA4A2 genes are paired head-to-head on the same chromosome and are transcribed in opposite directions. The pairs of COLA3A4 and COLA4A4 and COLA4A5 and COLA4A6 genes are similarly arranged, except each pair is located on a different chromosome. Type IV collagen genes are very large, as exemplified by COLA4A1 and COLA4A5 genes that exceed 100 kb in size.

Type VI collagen, another network-forming molecule, is represented by six distinct α chains in the mouse and five α chains in humans; the gene encoding the putative α4(VI) collagen chain is not functional in humans. Heterotrimers of different α chains, encoded by unique genes, form the basic unit of type VI collagen. Alternate splicing of messenger ribonucleic acids (mRNAs) generates additional variants of α2 (VI) and α3 (VI) chains.^7,9 The Gly-X-Y domains of α chains of type VI collagen microfibrils are rather short (about 330 amino acid residues) and are flanked by a number of von Willebrand factor (vWF) A domains.

Type VI collagen forms relatively unusual aggregates by a stepwise assembly into the triple-helical monomeric units that form dimers in an antiparallel fashion. The dimers in turn form tetramers, held together by disulfide bonds, to create scissors-like structures. The supramolecular assemblies of type VI collagen, formed by end-to-end associations of tetramers, appear as beads on a string, as revealed by electron microscopy.⁹ These characteristic structures have been observed in vascular subendothelium and skeletal muscle basement membranes. Type VI collagen microfibrils exhibit unique adhesive properties to other ECM components, such as other collagens, heparin, and vWF, and may be involved in the adhesion of platelets and SMCs. In the medial layer, type VI collagen facilitates interaction between SMCs and elastin by bridging the elastin fibers and cells.¹¹

As illustrated in Figure 4-5 (see discussion in “Metalloproteinases”), types VIII and X collagens comprise a unique subfamily of collagens that form hexagonal networks. These relatively short collagens, containing noncollagenous domains on their NH₂ and COOH termini, are collectively known as the multiplexin family of collagens. Type VIII collagen is expressed in many tissues, especially in the endothelium, while type X is exclusively associated with hypertrophic chondrocytes during cartilage and bone development. The preponderance of evidence to date indicates that the two α chains of collagen VIII, encoded by COL8A1 and COL8A2, assemble into homotrimers of α1(VIII) and α2(VIII) (Fig. 4-6). Hexagonal aggregates of type VIII collagen have been observed both in vivo (e. g., Descemet’s membrane of the cornea) and in vitro with purified protein. It is believed that type VIII collagen is capable of assuming other forms of macromolecular aggregates, since hexagonal lattices have yet to be observed in the subendothelial ECM.⁹

Figure 4-6 A, Linear structure of human collagen XV and XVIII α1 chains. The α1 chains of collagen XV and XVIII are structurally homologous; they comprise the multiplexin family on the basis of their central triple-helical domain with multiple long interruptions. They are also characterized by a long noncollagenous N-terminal domain–containing thrombospondin sequence motif, with two splicing variants in human collagen XVIII and long, noncollagenous, globular C-terminal domain or NCl domain. B, Functional subdomains of human NCl (XVIII) and protease cleavage sites. The NCl domain contains three functionally different subdomains: these domains consist of an N-terminal noncovalent domain involved in trimerization, a hinge domain containing multiple sites that are sensitive to different proteases, and an endostatin globular domain covering a fragment of 20 kD with antiangiogenic and antivessel sprouting activities. Numerous enzymes can generate fragments containing endostatin. Cathepsin L and elastase are the most efficient, but in contrast to matrix metalloproteinase (MMP) cleavage, which leads to accumulation of endostatin, cathepsins L and B degrade the molecule.

(Adapted from Company of Biologists Ltd., Ortega N, Werb Z: New functional roles for non-collagenous domains of basement collagens. J Cell Sci 115:4201, 2002.)

Figure 4-5 Domain structure of matrix metalloproteinases (MMPs).

The MMPs are multidomain enzymes that have a pro-domain, an enzymatic domain, a zinc-binding domain, and a hemopexin/vitronectin (VN)-like domain (except in MMP-7 and MMP-26). Additionally, membrane-type MMPs contain membrane anchor, with some membrane type (MT)-MMPs also possessing a cytoplasmic domain and a carboxyl terminus. Gelatinases contain a gelatin-binding domain with three fibronectin (FN)-like repeats. In particular, MMP-9 also contains a serine- threonine- and O-glycosylated domain. N-glycosylated sites, one of which is conserved in most MMPs, are denoted with a Y symbol. Part of the propeptide, which contains the chelating cysteine, and part of the zinc-binding domain with three histidines are indicated with one letter code for amino acids.

(Adapted from Hu J, et al: Matrix metalloproteinase inhibitors as therapy for inflammatory and vascular disease. Nat Rev Drug Discov 6:480–498, 2007.)

Fibril-Associated Collagens with Interrupted Triple Helices

As the name suggests, the FACIT collagens (types IX, XII, XIV, XVI, XIX, XX, XXI, XXII, and XXVI) do not form fibrils themselves but associate with other fibril-forming collagens.⁹ Type IX collagen, the prototype of this group, is cross-linked to the surface of type II collagen fibrils in cartilage (see Fig. 4-1); type XII and type XIV collagens are found in both cartilage and noncartilaginous tissues, where they are involved in controlling the diameter of collagen fibrils (see Fig. 4-1). The other FACIT-like collagens (e.g., types XVI, XIX, and XXII) are localized in specialized basement membranes. For instance, XVI collagen is associated with fibrillin 1 near the epidermal basement membrane. Collagen types XXI and XXII are closely related to each other in structure and are involved in formation of supramolecular aggregates in the basement membranes of myotendinous junctions.^12,13 As a key constituent of cutaneous basement membranes, anchoring fibrils of type VII collagen form a structural continuum between the dermis and epidermis of normal human skin. The vWF A–like domain in collagen VII binds to fibrils of type I collagen in vitro.¹⁴

Minor Collagen Types with Unique Structures

As illustrated in Figure 4-1, the transmembrane collagens (types XIII, XVII, XXIII, and XXV) contain a cytoplasmic domain, a membrane-spanning hydrophobic domain, and extracellular triple-helical domains interspersed with noncollagenous domains; these collagens may also exist in a soluble form. Type XVII collagen is a unique member of this group that is expressed on the basal surface of keratinocytes that bind to laminin found in the basement membrane; compared with the other three members of this group, type XVII has a rather large intracellular domain whose function remains unknown. Collagen types XIII, XXIII, and XXV are similar to each other in their primary structure, but the patterns of their expression appear to be unique. Type XXV collagen is enriched in the senile plaques of Alzheimer’s disease brains.^9,12,13,15 High expression of full-length collagen XXIII is found in the lungs, whereas its shed form is enriched in brain, suggesting that shedding of XXIII collagen occurs in a tissue-specific manner.

Collagen types XV and XVIII are highly pertinent to the EC biology in several ways.^16,17 The full-length types XV and XVIII collagen are basement membrane components; their triple-helical domains share a high degree of homology. Collagen types XV and XVIII were initially identified as PG core proteins containing chondroitin sulfate and heparan sulfate (HS) side chains, respectively. The COOH-terminal domains of XV and XVIII collagens can be cleaved to generate biologically active peptides, endostatin and restin, respectively; these peptides inhibit migration of ECs and thus potently block angiogenesis. In vitro, recombinant collagen XV binds to fibronectin (FN), laminin, and vitronectin (VN) but not to fibrillar collagens, fibril-associated collagens, or decorin.¹⁸

Finally, collagens XXVI and XXVIII are newly discovered collagens that are unique both with regard to their structures and tissue-specific distributions. The triple-helical domain of type XXVI is rather small, with only 146 Gly-X-Y repeats. Expression of type XXVI collagen occurs predominantly in testis and ovary. The von Willebrand factor–A domains flank the triple-helical structure of type XXVIII collagen that is almost exclusively expressed in the peripheral nerves.^19,20

Regulation of Collagen Biosynthesis

Collagen chains are synthesized as prepro-α chains from which the hydrophobic leader sequence is removed prior to secretion, and the pro-α chains are secreted into the extracellular space (see Fig. 4-2). The pro-α1(I) chain contains an NH₂ propeptide (N-peptide) and a COOH propeptide (C-peptide). The N-peptide consists of a 139-residue sequence that precedes a 17-residue sequence of nonhelical telopeptide. This is followed by a 1014 amino acids-long Gly-X-Y helical sequence attached sequentially to a 26 residues-long COOH telopeptide and a 262-residues-long nonhelical C-peptide. The domain organization of pro-α2(I) and pro-α1(III) chains are similar except for minor variations in the number of amino acid residues.^6,7,21

The genomic organization and chromosomal locations for genes that encode collagens have been studied. In humans, the genes encoding 43 distinct α chains are dispersed on at least 15 chromosomes. Unlike majority of the collagen-encoding genes, the six homologous α-chains of type IV collagen are encoded by genes that are located in pairs with head-to-head orientation on chromosomes 13 (COL4A1 and COL4A2), 2 (COL4A3 and COL4A4), and the X chromosome (COL4A5 and COL4A6). Interestingly, the promoters of these pairs of type IV collagens overlap, suggesting a coordinate regulation of the gene pairs. The precise molecular mechanisms of this regulation, however, remain incompletely known.^6,7,9,21,22

The molecular events involved in procollagen biosynthesis, from transcription and splicing of mRNA to its transport and translation in the cytoplasm, are nearly identical to most other proteins synthesized by eukaryotic cells. Regulation at the level of transcription and mRNA turnover appears to be involved in the coordinated synthesis of two pro-α1(I) chains for every one of pro-α2(I) chain. Most cells that produce type I collagen also produce type III collagen in variable amounts, depending on the specific type of tissue, its age, and the physiological and pathological situations.

The molecular mechanisms of regulation of biosynthesis of a number of collagens have been studied to varying degrees, both in physiological and pathological settings. Regulation of genes that encode α chains of type I collagen has been studied extensively and is briefly summarized. Transcriptional regulation of genes that generate fibrillar (COL1A1, COL1A2, COL3A1) and basement membrane (e.g., COL4A1-6) collagens evidently involves both genomic and epigenomic (deoxyribonucleic acid [DNA] methylation and posttranslational modification of histones) mechanisms. Although collagen genes are predominantly regulated at the level of transcription, a number of reports indicate that posttranscriptional regulation is also exerted under some conditions.

The cis-acting elements of COLA1 and COLA2 genes are modularly organized on either side of the transcription start point (TSP). The regulatory elements are distributed over a distance of 100 to 150 kb of genomic DNA, depending on the specific gene and the assays used to study their transcriptional and posttranscriptional regulation. The tissue-specific and inducible activation of collagen genes involves complex interactions among the cis-acting modules of their promoters and enhancers. Promoters of COLA1 and COLA2 genes contain TATA boxes located 25 to 35 bp upstream of the TSP. Existence of a number of enhancer and repressor cis-elements around the TSP and in the first intron of COLA1 gene has been demonstrated. A key role for CAAT-binding factor, Sp1, Sp3, Ap1, nuclear factor (NF)-κB, and SMADs has been reported for several collagen genes; a number of orientation-dependent enhancer-like elements have also been documented.^23,24

Fibrillar and nonfibrillar collagens found in subendothelial ECM are regulated by many cytokines and growth factors; collagen gene expression in response to cytokines (e.g., transforming growth factor [TGF]-β, tumor necrosis factor [TNF]-α, interleukins [ILs]), glucocorticoids, estrogen, androgen, and retinoids has been reported. The signaling cascades initiated by intrinsic and exogenous regulators impinge on a distinct set of cis-acting elements that bind to constitutive and inducible transcription factors. The emerging theme from these studies is that various cis– and trans-acting factors interact to recruit selective transcriptional coactivators and co-repressors in response to specific stimuli.^23,24 However, the precise mechanisms that determine combinatorial interactions under physiological and inflammatory conditions remain to be elucidated.

Following translation, pre-procollagen α chains are chaperoned from the endoplasmic reticulum (ER) to the Golgi. It has been reported that the heat shock protein-47 (Hsp47) functions as a collagen-specific chaperone; thus, hsp47 is presumed to provide a quality control mechanism needed for proper maturation of newly synthesized procollagen chains. To demonstrate a role of hsp47 in vivo Nagai and coworkers²⁵ inactivated Hsp47 gene by homologous recombination. The mutant embryos died in utero before 11.5 days of postcoitus development as a result of severely reduced levels of mature type I collagen in their tissues.

As shown in Figure 4-2, fibrillar and nonfibrillar collagens also undergo a number of posttranslational modifications for proper maturation; these include proteolysis of signal peptides, hydroxylation of key proline and lysine residues, glycosylation, and formation of interchain and intrachain disulfide bridges.^6,7,21 Thus, optimal biosynthesis and assembly of collagens depends on a number of key enzymes. These include three hydroxylases, two collagen-specific glycosyl transferases, two unique proteinases that cleave the NH₂– and COOH-termini, and a collagen-specific oxidase that is needed for cross-link formation. The posttranslational processing of the procollagen molecules also needs a peptidyl proline cis-trans isomerase and a protein disulfide isomerase (PDI).

Vitamin C–dependent 4-prolyl hydroxylase, an α₂β₂-tetramer located in the ER, plays a central role in collagen synthesis because 4-proline hydroxylation is obligatory for cross-link formation. In humans, there are three known isozymes of 4-prolyl hydroxylases, each with a distinct α subunit, but all contain PDI as their β subunit. Hydroxylation of lysine is carried out by lysyl hydroxylase, which also uses the same cofactors as prolyl hydroxylase and reacts only with a lysine residue in the Y position of the Gly-X-Y triplets. There are three known isozymes of lysyl hydroxylase in humans. The under-hydroxylation of procollagen leads to reduced secretion and rapid degradation. Deficiency of lysine hydroxylase is associated with skeletal deformities, tissue fragility, and vascular malformations.^6,7,21

Several collagens undergo glycosylation; both galactose and glucose residues are attached to some hydroxylysine residues during pre-procollagen biosynthesis. The enzyme UDP galactose:hydroxylysine galactosyltransferase adds a galactose residue to the hydroxyl group of hydroxylysine. The UDP glucose galactosyl:hydroxylysine glucosyltransferase then transfers a glucose residue to the hydroxylysine-linked galactose. The two enzymes act in sequence so that galactose is added first, with glucose added only to galactose. Glycosylation occurs during nascent chain synthesis and before the formation of triple helices. Only two of seven hydroxylysine residues of α1(I), α2(I), and α1(III) contain the disaccharide; most of the hydroxylysine residues are glycosylated in other collagens. Glycosylation of some hydroxylysine residues imparts stability to the cross-link.

Assembly of procollagen chains into triple-helical molecules is directed by the COOH-terminal propeptide, with formation of interchain disulfide bonds (see Fig. 4-2). There is a high degree of structural conservation within the propeptide of fibrillar collagens across species. Following its triple-helical assembly, the procollagen molecule is secreted into the extracellular space. Once secreted, however, the NH₂ and COOH propeptides are removed by the actions of N- and C-specific peptidases to yield the collagen molecule. The two proteinases that remove the NH₂ and COOH propeptides from the newly synthesized collagen are represented by three isozymes each. The C-specific peptidases, members of the tolloid family, also cleave a number of other ECM proteins, and fragments of the propeptides can inhibit procollagen synthesis by a feedback mechanism.^26,27

Extracellular Maturation of Collagens

During collagen fibril formation, lysyl oxidase catalyzes the oxidative deamination of specific lysine or hydroxylysine residues in the NH₂– or COOH-terminal telopeptides to yield allysine and hydroxyallysine, respectively.²⁸ These reactive aldehydes, being located in the hole zone of the fibril, are free to react with the ε-amino group of lysine or hydroxylysine residues on adjacent chains to form a Schiff base, which undergoes Amadori rearrangement to form ketoimine. With time, two ketoimine structures condense to form a trivalent cross-link, 4-hydroxy-pyridinium. All three types of cross-link may coexist in different fibrils.

A second type of cross-link seen in collagen originates from the condensation of two aldehydes in allysine or hydroxyallysine on adjacent chains. The resulting aldol condensate has a free aldehyde that reacts with other ε-amino groups of lysine or histidine, thus potentially linking three or four collagen chains.²⁹ Once the aldehydes of allysine and hydroxyallysine are formed, subsequent aldamine and aldol condensation reactions proceed spontaneously. Thus, inter- and intramolecular cross-linking of fibrillar collagens results in formation of insoluble macromolecular aggregates that possess high tensile strength.

Turnover of Collagen

Metabolic turnover of collagens in intact tissues during adulthood is extremely low. In contrast, a very rapid breakdown and synthesis of collagen takes place during tissue remodeling. In their native fibrillar state, collagens are quite resistant to the action of proteases, yet once their helical structure is disrupted, they are readily degraded by a number of proteases. The FACITs such as types IX, XII, and XIV and other collagens containing noncollagenous domains (e.g., type VI collagen) are relatively more susceptible to proteases. After cleavage of the nonhelical segments, the triple-helical domains of collagens denature at 37 °C and become susceptible to nonspecific proteases. Additionally, a specific class of proteinases, the matrix metalloproteinases (MMPs), degrades collagens in vivo and in vitro (see later discussion). For example, MMPs cleave the native type I collagen molecule at a single position within its triple helix, between amino acid residues 775 and 776, and the resulting collagen fragments denature spontaneously at body temperature and pH and become highly susceptible to the actions of many other proteases.

Elastin Metabolism and Vascular Homeostasis

After deposition, tropoelastin production is strikingly reduced; the half-life of elastin in normal humans has been estimated in years. In the event of injury, production of elastin can be quickly initiated. A number of growth factors and cytokines induce biosynthesis of tropoelastin. Under these conditions, a very specific set of proteinases named elastases are responsible for elastin remodeling. Elastin fibers may be degraded by a number of MMPs, particularly MMP-2, -3, -9 and -12, that are present as latent enzymes under physiological conditions but are activated following vessel wall injury.⁴⁷ The MMPs from neutrophils or macrophages are believed to degrade the elastin-rich ECM found in inflamed tissues. A hereditary defect in circulating elastase inhibitors is associated with a progressive destruction of the elastin-rich alveolar wall, resulting in premature emphysema. Furthermore, experimental instillation of elastase into the lungs of animals causes destruction of the lung similar to that seen in patients with α1-proteinase inhibitor deficiency.

Mice with a disrupted elastin gene have provided important insights into the function of elastin protein. Heterozygous (elastin^+/− mice) had decreased arterial compliance and were hypertensive. The homozygous elastin-null mice died young due to arterial obstruction caused by uncontrolled proliferation of smooth muscle cells (SMCs).^39,48 A direct link between occlusive vascular diseases and perturbation in the organization of the elastic fibers in the vessels has also been established.⁴⁹ Mutations in the elastin gene are associated with supravalvular aortic stenosis (SVAS) and Williams-Beuren’s syndrome (WBS), pediatric disorders characterized by hemodynamic stress and loss of elasticity.⁴⁹ Furthermore, haploinsufficiency of elastin resulting from aberrant degradation of mutated protein in humans or ablation of the elastin gene in transgenic mice caused intimal hyperplasia and thickened arteries.^{40,45,50–52} Apparently, VSMCs, the primary producers of elastin, organized more cell layers to compensate for lost elasticity and biomechanical support in developing blood vessels of elastin haploinsufficient patients and transgenic mice.

Vascular smooth muscle cells from SVAS patients, WBS patients, and elastin^−/− mice show increased rates of proliferation and chemotactic migration, and reduced rates of elastin synthesis in vitro.^50–52 Exogenous supplementation of recombinant tropoelastin and α-elastin to these cultures reversed their phenotype. The elastin-rich ECM serves as an autocrine regulator of VSMC; Karnik et al.⁵³ inserted elastin-coated stents in a porcine coronary injury model of restenosis and found that intimal thickness and arterial stenosis were significantly reduced. Although the identities of the specific receptors mediating elastin VSMC interactions and the signaling mechanisms underlying vascular remodeling remain obscure,⁵⁴ restoring elastin to an injured arterial wall is known to reduce obstructive vascular pathology.^55,56 Inhibitors of MMPs have been shown to prevent degradation of elastic fibers after vascular injury and ameliorate neointimal thickening.⁴⁷

Fibronectin Structure

In blood plasma, FN exists in a soluble state, synthesized and secreted by the liver, and is converted into an insoluble supramolecular complex in the ECM. The soluble FN is made of two disulfide-linked monomers of similar or identical mass (220-255 kD). As shown in Figure 4-10, each FN monomer is a mosaic of repeating modules termed type I, II, and III repeats that are 40, 60, and 90 amino acids long, respectively.^67,68 A cluster of 15 to 17 type III repeats (depending on alternative splicing) located in the middle of the molecule represents 90% of the FN monomer. In addition, there are 12 type I and 2 type II repeats in each monomer of FN. The type III repeats fold into nearly identical shapes despite having only 20 to 40 amino acid sequence identity (see Fig. 4-10). The striking modular organization of the repeated peptide sequences in FN is reflected in the organization of its gene. The FN gene consists of 47 exons spanning nearly 100 kb in the human genome and generates multiple alternatively spliced mRNAs.^69–72 A single gene thus generates about 20 variants of FN protein that may be preferentially synthesized under various physiological and pathological situations.⁶⁸ Fibronectin monomers containing or lacking the extra domain A (EDA) or B (EDB) are particularly significant with regard to their biological functions. Plasma FN (soluble) lacks both EDA and EDB domains; in contrast, FN assembled into ECM contains variable mixtures of cellular FN with or without EDA and EDB domains (see Fig. 4-10).

Figure 4-9 A, Domain organization of fibrillin.

All three fibrillins have a similar modular organization, with strong homologies with each other. Fibrillin-1 has a proline-rich region, whereas fibrillin-2 has a glycine-rich sequence; in contrast, fibrillin-3 has a region that is both proline- and glycine-rich. This region lies between the first 8-cysteine module and the fourth epidermal growth factor (EGF)-like domain. B, Schematic of various ligands that bind to fibrillin to assemble microfibrils. Putative binding sites for various ligands are based on in vitro observations; association of various molecules with fibrillins is most likely regulated dynamically by physiological and pathological stimuli. BMP, bone morphogenetic protein; LTBP, latent transforming growth factor β-binding protein; MAGP, microfibril-associated glycoprotein.

(Adapted from Ramirez F, Sakai LY: Biogenesis and function of fibrillin assemblies. Cell Tissue Res 339:71–82, 2010.)

Figure 4-10 Primary structure of fibronectin (FN) and its modular organization.

Hypothetical scheme represents an FN dimer with various sequence modules. A, Different types of homologous domains (12 type I, 2 type II, and 15 type III) are shown. Numbering of type III homologies excludes extra domain (ED)A and EDB domains. Types I, II, and III domains are made of 40, 60, and 90 amino acids, respectively. Constitutively expressed (RGD), alternatively spliced (LDV), synergy (PHSRN) and EDA (EDGIHEL) cell-binding sites are indicated, together with integrin receptors to which they bind. EDA and EDB splicing is similar in all species, whereas the IIICS region is spliced in a species-specific (five variants in humans, three in rodents, and two in chickens) manner. Type III homologies are organized in seven antiparallel β strands. Spatial and planar representations of type III module are shown in B and C, respectively.

(Adapted from White ES, Baralle FE, Muro AF: New insights into form and function of fibronectin splice variants. J Pathol 216:1–14, 2008.)

Functional Domains of Fibronectin

Fibronectin is a multifunctional molecule with a series of specialized modules.^67,68 Proteolysis of FN generates a number of fragments that bind to specific ligands. The NH₂-terminal 70-kD fragment of FN binds to a surprisingly large number of ECM ligands that include collagen, gelatin, fibrin, and heparin. The 70-kD FN also binds to some gram-positive bacteria (e.g., Staphylococcus aureus, Streptococcus pyogenes, Streptococcus pneumoniae) via the so-called microbial surface components recognizing adhesive matrix molecules (MSCRAMM). Thus lipoteichoic acid, M proteins, and several other bacterial adhesins anchored in the cell wall bind to FN and enhance opsonization and phagocytosis of bacteria.⁷³ Gram-negative bacteria do not bind to FN.

Modular Proteoglycans

Modular PGs consist of a heterogeneous group of highly glycosylated PGs characterized by a tripartite structure of their core proteins. This group is further divided into two subfamilies represented by HA- and lectin-binding PGs, hyalectans and non-HA-binding PGs.

Hyaluronan- and Lectin-Binding Proteoglycans

Four distinct PGs—versican, aggrecan, neurocan, and brevican—constitute the hyalectan family of PGs.^105–109 The tripartite organization of their core proteins consists of the NH₂– and COOH-terminal domains separated by a distinct central domain where GAGs are attached. The amino terminal domain of these PGs binds to HA and their carboxyl termini contain lectin-like domains, thus the name hyalectans (see Table 4-4). The central domain of hyalectans contains variable numbers of GAG chains, ranging from 3 seen in brevican to around 100 found in aggrecan. Versican, the largest member of the hyalectan family, may be considered a prototype^109,110 (see Fig. 4-14 and Table 4-4). Versican and other hyalectans are believed to connect lectin-containing proteins on the cell surface with HA in the intercellular space. Versican has an immunoglobulin (Ig)-like motif and two tandem HA-binding domains near the NH₂-terminus; an EGF-like domain and a lectin- like motif are located near the COOH-terminus of versican. Four alternatively spliced versican isoforms (V0, V1, V2, and V3) are preferentially expressed in different tissues. Versican binds to numerous cell-associated and ECM molecules that include type I collagen, tenascin, fibulins, fibrillin-1, FN, selectins, chemokines, CD-44, integrin β1, EGF, and Toll-like receptors.^109–113

Aggrecan typically contains around 100 CS-enriched and 30 KS-enriched GAGs that are covalently linked to about a 220-kD core protein (see Table 4-4). As the most abundant constituent of cartilage ECM, aggrecan is found in giant aggregates with link proteins and HA, and occupies a large hydrodynamic volume (2  ×  10^− 12 cm³) that may be equivalent to a bacterium.¹⁰⁸ The lectin modules of both versican and aggrecan can interact with simple sugars found in glycoproteins; this binding is calcium dependent.¹¹⁴ Defective cartilage and shortened limb development have been demonstrated in mice, chickens, and humans that contain mutated aggrecan genes.^115,116

Neurocan and brevican, with tripartite organization of core proteins characteristic of hyalectans, are the most abundant PGs of this class in the central nervous system.¹¹⁷ Brevican is synthesized in the brain as a secreted full-length molecule, as well as a truncated form that lacks the COOH-terminal domain. The short form of brevican is attached to the plasma membrane via a GPI anchor. Neurocan and brevican promote neuronal attachment and outgrowth of neurites in developing neurons. Brevican activates EGF receptor (EGFR) signaling that results in enhanced expression of cell adhesion molecules such as FN. Highly aggressive central nervous system tumors elicit accelerated synthesis and proteolysis of brevican and thus may promote tumor metastasis.

Non-Hyaluronan-Binding (Basement Membrane) Proteoglycans

Perlecan, agrin, and bamacan are usually present in the vascular and epithelial basement membranes of mammalian tissues.^97,109 Whereas the three GAG chains of perlecan and agrin consist primarily of HS and CS, the three bamacan GAG chains contain only CS. The core protein of the human perlecan is about 470 kD in size and contains five well-defined domains (I through V) that are mosaics of sequences found in other proteins. Thus, domain I of perlecan consists of three serine-glycine-asparagine triplets to which HS side chains are covalently attached, and an SEA module (containing sperm protein, enterokinase, and agrin homology sequence). Domain II contains four low-density lipoprotein (LDL) receptor class-A repeats located next to IgG-like motifs. Three laminin IV globular domains interspersed with nine laminin EGF-like motifs comprise the domain III of perlecan. Domain IV contains 21 IgG-like motifs that share homology with neural cell adhesion molecules (N-CAM). Finally, domain V is made of three laminin G motifs separated by two sets of EGF-like repeats; the 85-kD COOH-terminal fragment of perlecan called endorepellin is a potent antiangiogenic molecule.^97,109,118

Agrin is a major PG of basement membranes of the renal glomerulus and nerve-muscle junctional synapses. Although the four-domain structure of agrin, with three HS–rich GAG chains, resembles perlecan, there are critical structural differences between perlecan and agrin. The amino terminus of agrin is required for binding to laminin-111 as well as for secretion of the newly synthesized agrin. Agrin is essential to aggregate acetylcholine receptors at the neuromuscular synapse and facilitates synaptogenesis during neuromuscular junction development. Although originally isolated from the Reichert membrane, CS-rich bamacan is found in variable amounts in most basement membranes.

In addition to imparting structural integrity to the basement membranes, PGs modulate cellular behavior because of their ability to interact with a large number of molecules, as exemplified by perlecan.^{97,109,110,118} Although perlecan is embedded within the subendothelial basement membranes, it binds to fibroblast growth factor 2 (FGF-2), vascular endothelial growth factor (VEGF), platelet-derived growth factor (PDGF), several cell surface molecules, and ECM proteins. Heparan sulfate GAGs of perlecan associate with FGF-2 and serve as its reservoir in blood vessel walls. During aortic morphogenesis, there is an inverse correlation between perlecan expression and smooth muscle proliferation in the rat. Perlecan interacts with α₁β₂ integrin, an integrin that also binds to fibrillar collagens. Dynamic interactions of perlecan and fibrillar collagens with integrins potentiate atherosclerosis, angiogenesis, and carcinogenesis. Perlecan regulates the motility of ECs and transformed cells and promotes metastasis.¹¹⁹ Perlecan-null mice die in utero or shortly after birth.

Small Leucine-Rich Proteoglycans

Nine known members of this family of PGs, characterized by central leucine-rich domains and DS/KS GAG chains, are currently known.¹²⁰ Based on their primary structures and evolutionary relationships, SLRPs may be further divided into three subclasses (see Table 4-4). Members of the SLRP family have been¹²¹ implicated in diverse functions that include regulation of growth factor accessibility (e.g., TGF-β) and control of collagen fibrillogenesis.¹¹⁹ Presence of decorin, fibromodulin, or lumican retards collagen fibrillogenesis in vitro.^122–125 A reciprocal relationship between the amount of decorin and rate of collagen fibril growth in the developing tendon of the chicken was also demonstrated.¹²⁶ The abnormal collagen fibril formation and reduced tensile strength of the skin seen in decorin knockout mice support a functional role for decorin in proper collagen fibrillogenesis.¹²²

Decorin, a prototype of the SLRP, is organized into four discernible domains.¹⁰⁹ Core protein of decorin binds TGF-β1, -2, and -3 with high affinity. Because the TGF-β/decorin complex is incapable of intracellular signaling, decorin is believed to facilitate deposition of inactive TGF-βs at specific tissue locations. Perturbation of this interaction and activation of TGF-βs may occur in response to inflammatory reactions. Decorin itself has been shown to regulate cell proliferation, and ectopic expression of decorin could suppress the growth of cancer cells. Decorin directly interacts with EGFR with a 1:1 stoichiometry; this interaction inactivates EGFR and its downstream signaling.¹²⁷ Vascular ECs undergoing cord formation that precedes angiogenesis in vitro synthesize decorin, whereas proliferating ECs showed enhanced synthesis of biglycan; these two structurally similar SLRPs appear to regulate EC phenotype in an opposite manner.

Cell Surface Proteoglycans

Two main classes of cell surface PGs are membrane-spanning syndecans and GPI-linked PGs represented by glypicans.^109,128,129 There are four members of the syndecan subfamily. Syndecan-4 is distributed ubiquitously, but syndecan-1, -2, and -3 have more restricted tissue- or development-specific expression.^128–130 For example, syndecan-1 is highly expressed in the developing embryo, and syndecan-3 is primarily enriched the neural tissue. Syndecans are involved in multiple signaling pathways to regulate cell proliferation, adhesion, motility, and differentiation. The cell surface HS-enriched syndecans serve as co-receptors for FGF and EGF to facilitate their binding and signal transduction. Syndecan-1 is cleaved by MMPs, and the soluble ectodomain of syndecan promotes tumor growth and invasiveness in vitro.^121,131,132 Exogenous treatment of microvascular ECs with EGF and FGF leads to shedding of syndecan-2 that in turn affects EC behavior.¹³³

The subfamily of GPI-linked cell surface PGs includes six member glypicans and a splicing variant of brevican that lacks the COOH-terminal lectin binding and EGF motifs. While most tissues express glypican-1, expression of glypican-3, -4, and -5 is restricted to the central nervous system. In contrast, glipican-2 is expressed abundantly in the embryo, whereas glypican-6 is found mainly in the heart, kidney, and intestine. Glypican-3, which inhibits hedgehog (Hh) signaling by competing for its receptor Patched, is up-regulated in neuroblastoma and Wilms tumor. Glypican regulates binding and signaling of a number of other morphogens and growth factors that include Wnts, slit, FGFs, insulin-like growth factors (IGFs), and BMPs. The regulatory actions of glypicans on proliferation and differentiation of cells appear to be context dependent.^129,134,135

Biosynthesis of Proteoglycans

Biosynthesis of all PGs involves similar steps, the rate-limiting step being translation of the core protein.^119,136,137 Following its synthesis, core protein undergoes covalent modification with GAG chains that begins with the linkage of xylose to a specific serine(s). Proteoglycan synthesis occurs in late ER and the Golgi with attachment of a xylose residue to the OH group of serine in the protein core.¹³⁸ Linking of galactose to xylose is carried out by galactosyl transferase. A second galactose is then transferred by a distinct galactosyl transferase that is followed by addition of the first glucuronic acid by UDP–glucuronic acid transferase. Growth of the GAG chain then occurs by alternating transfer of hexosamine and uronic acid residues. Thus, the UDP derivatives of N-acetylglucosamine and glucuronic acid are precursors for HA, heparin, and HS; whereas N-acetylgalactosamine and glucuronic acid are precursors for CS and DS. After addition of the first sugar, elongation occurs by the same N-acetyl-hexosaminyltransferase and glucuronosyltransferase, regardless of which chain is being synthesized. The respective chains are variably modified by pathway-specific epimerization and sulfation reactions to yield iduronic acid and sulfation.^107,127

Nonspecific sulfotransferases transfer a sulfate group from 3-phosphoribosyl phosphoadenosine 5-phosphoribosyl phosphosulfate to the appropriate site on the GAG. Because no partial sulfation occurs, it is believed the GAG may become attached to the particulate-bound sulfotransferases and glycosyl transferases and are completely sulfated before release. This N-sulfation is unique to heparin and HS; all other PGs contain an O-linked sulfate group. The synthesis of this N-sulfate linkage proceeds through the N-acetylglucosamine addition to the GAG chain, deacetylation, and replacement of the acetyl group by sulfate. Iduronic acid formation in heparin, HS, and DS takes place after polysaccharide synthesis by epimerization of glucuronic acid. Proteoglycan size is extremely heterogenous and mainly reflects Gag chain length.

Degradation of Proteoglycans

Compared to collagen and elastin, PGs have more rapid rates of turnover, with turnover of 2 to 10 days in younger animals.^{97,109,110,119} Degradation of the PGs involves proteolysis of the core protein by MMPs, breakdown of the sugar chain, and desulfation of sugars. The dramatic loss of cartilage matrix that results from experimental intravenous injection of papain illustrates the importance of the protein core to the structural integrity of PGs. Fibroblasts, macrophages, and neutrophils produce a variety of enzymes that can degrade PGs at neutral pH. Degradation of sugar chains occurs mainly in lysosomes. Perhaps the best characterized GAG-degrading enzyme is testicular hyaluronidase, which degrades HA, CS, and DS to oligosaccharides. Other glycosidases are required to complete the breakdown of oligosaccharides to monosaccharides. Lysosomes contain glucuronidase and N-acetylhexosaminidases that remove the terminal glucuronic acid and hexosamine residues, respectively. Lysosomes also contain β-xylosidase, β-galactosidase, and α-iduronidase, which complete the breakdown. Lysosomal sulfatases are responsible for removal of sulfate groups from oligosaccharides. Inherited defects in the activity of various GAG-degrading enzymes cause mucopolysaccharidoses, characterized by faulty catabolism of one or another type of GAGs.^{97,109,110,119}

Extracellular Matrix–Integrin Bi-Directional Signaling

Each of the many molecules found in the vascular ECM recognizes a variety of cell surface proteins predominated by integrins, a superfamily of heterodimeric transmembrane receptors (Fig. 4-15). Integrins are assembled by selective pairings of 18 individual α and 8 unique β chains. Twenty-four integrin receptors with distinct ligand selectivity, cell-specific expression, and signaling properties have been described in mammals.^147,148 The extracellular segment of the α subunit of integrin consists of a β-propeller domain, an I-domain, and three Ig-like domains. The ectodomain of the integrin β subunit also has a modular organization, with two tandem I-domains, an Ig hybrid motif, a plexin-semaphorin-integrin (PSI) domain, and four EGF-like domains. Integrins specifically bind to several ECM molecules, their subfragments, and divalent cations. With respect to vasculature, ECs express integrins that specifically bind to collagen (α₁β₁, α₂β₁, α₁₀β₁, and α₁₁β₁), FN (α₄β₁ and α₄β₁), and laminin (α₃β₁, α₆β₁, and α₆β₄). Based on a number of in vitro and in vivo observations, at least eight integrins (α₁1₇, α₂β₁, α₃β₁, α₄β₁, α₅β₁, α₆β₁, α_vβ₃, and α_vβ₅) have been implicated in the process of angiogenesis.^149,150 In contrast, leukocytes express a number of unique integrins on their surface that include α₄β₇, α_Lβ₂, and α_Mβ₂ integrins. We should note that the ligand selectivity of integrins is far from absolute, as exemplified by α_vβ₃, which binds to several RGD sequence–containing proteins and peptides.^147,148 The following discussion summarizes numerous observations that underscore the fact that integrins lie at a unique crossroads of extracellular microenvironment, cytoskeleton mechanics, and intracellular signaling networks to alter the behavior of vascular walls in health and disease (see Fig. 4-15).

Figure 4-15 Illustration of various components of extracellular matrix (ECM), their cell surface receptors, and major intracellular signaling molecules.

Potential steps that may be exploited for pharmacotherapy include: (1) Blocking synthesis of ECM by blocking specific growth factors such as transforming growth factor beta (TGF-β), their receptor molecules, or intracellular signal transduction. (2) Blocking degradation of ECM by interfering with enzymes involved in its remodeling (e.g., matrix metalloproteinases (MMPs), ADAMTS, cathepsins) and/or their inhibitors (tissue inhibitor of MMP [TIMPs]). (3) Interfering with ECM signaling pathways (e.g., via integrins) either by blocking ECM and integrin interactions or subsequent signal transduction cascades. (4) Influencing transcription of specific ECM molecules (e.g., by siRNA). Please note that that purple rods in cytoplasm, marked as microfibrils, are in fact actin cytoskeleton intimately involved in intracellular signing by ECM. BM, basement membrane; PM, plasma membrane.

(From Jarvelainnen H, Sainio A, Koulu M, et al: Extracellular matrix molecules: potential targets in pharmacotherapy. Pharmacol Rev 61:198–223, 2009.)

Integrins are unusual proteins among the transmembrane receptors, with an ability to relay signals in both directions.^147,148,151 The intracellular changes induced by ECM-liganded integrins are referred to as outside-in signaling. Conversely, inside-out signaling occurs when intracellular biochemical changes trigger reorganization of the cytoskeleton, which alters the shape of the ectodomain of integrin and its affinity for the ligand. The ECM-liganded integrins are clustered as dot-like foci that sequentially evolve into focal adhesions, fibrillar adhesions, and finally into supramolecular three-dimensional adhesions. Such mass clustering of integrins into focal adhesions results in summation of numerous weak-affinity interactions of individual integrins into an adhesive unit with high affinity and high avidity.

Clustering and activation of integrins induce a number of characteristic biochemical and physical changes in the cells that are collectively referred to as outside-in signal transduction.^143,144,152 Since integrins themselves lack catalytic activity, they signal indirectly via a host of accessory proteins that assemble multi-protein platforms to recruit bona fide signaling catalysts into the focal adhesions. The assembly of bi-directional signaling complexes depends on interactions among a large number of integrin-binding proteins (e.g., talin), adapter molecules (e.g., vinculin, paxillin), and signaling enzymes (e.g., FAK, RhoA-kinase [ROCK], myosin phosphatase). It has been estimated that the “integrin adhesome” consists of more than 150 unique proteins; therefore it is conceivable that recruitment of unique sets of adapter and signaling molecules to focal adhesions might be different under differing physiological and pathological conditions.^153,154

The inside-out signaling of integrins is best exemplified by their own activation, particularly on the surface of leukocytes, where integrins are normally present as inactive receptors.^143,144,152 This mechanism enables the immune cells and platelets to circulate through the bloodstream without undesirable adherence to the vessels or causing premature thrombosis, respectively. The activation of T cells, neutrophils, and platelets can occur by integrin-independent pathways (i.e., occupancy of the T-cell receptor by MHC-loaded antigenic peptides, ligation of selectins on the surface of neutrophils, and interaction between platelet glycoprotein IV and collagen. Such activation initiates a series of intracellular reactions that lead to binding of kindlin and talin to intracellular tails of β-integrin and its conformation-dependent activation. As a consequence, the affinity of integrin for its ligands is greatly enhanced, as is its signaling strength.

Integrin signaling is mainly propagated by kinases and phosphatases that, by dynamic phosphorylation and dephosphorylation, alter protein-protein interactions of their substrates and catalytic activities of signaling enzymes in focal adhesions.^{143,144,152,155} The major enzymes involved in this process include tyrosine kinase, FAK, protein kinase C (PKC; a serine/threonine kinase), a lipid kinase (PI3-kinase), and the receptor protein tyrosine phosphatase α (PTPα). Activation of integrin immediately initiates tyrosine phosphorylation of specific substrates that include the integrin β tails. Concomitantly, there is a surge in intracellular concentration of lipid second messengers, phosphatidylinositol-4,5-bisphosphates and phosphatidylinositol-3,4,5-trisphosphates, and a reorganization of the cytoskeleton.

As shown in Figure 4-16, the characteristic physical link between integrins and cytoskeleton is initiated and maintained by integrin-bound proteins that also bind to actin (e.g., talin, filamin) in collaboration with other proteins that regulate the structure of cytoskeleton indirectly (e.g., paxillin, FAK, kindlin). Additionally, the integrin-cytoskeleton linkage and signal propagation is critically dependent on actin-binding proteins (e.g., vinculin) and several signaling adapters (e.g., RhoA, Rac1, cdc42). Numerous biochemical and biophysical analyses have revealed that talin, vinculin, α-actinin, and integrin-linked kinase (ILK) are indispensable for bidirectional signaling of focal adhesions.^{143,144,152,155}

Figure 4-16 General model of cell–extracellular matrix (ECM) adhesions and downstream signaling pathways.

Cell-ECM adhesions containing clusters of integrins recruit cytoplasmic proteins (e.g., talin, paxillin, vinculin) and kinases (focal adhesion kinase [FAK], Src, and c-jun NH 2-terminal kinase [JNK]), which in cooperation with other cell surface receptors (growth factor receptors; see Fig. 4-17) control diverse cellular processes, functions, and phenotypes. Details of these interactions are described in text. ERK, extracellular signal regulated kinase; MEK, mitogen extracellular signal regulated kinase.

(Adapted from Berrier AL, Yamada KM: Cell-matrix adhesion. J Cell Physiol 213:565–573, 2007.)

Integrin activation leads to a sequence of biochemical and biophysical reactions that may be divided into three temporal phases. First, high-affinity ligand-integrin association leads to immediate activation of phosphatidylinositol 3-kinase (PI3K) and concomitant phosphorylation of several proteins. The second stage of signaling, observable within half an hour, proceeds to activation of the Rho family of GTPases that are crucial for the architectural reshaping of the cytoskeleton. The final phase of signaling—executed over many hours—culminates in the nucleus, with reprogramming of gene expression carried out by transcription factors and their coactivators.

The unique molecular composition of focal adhesions at different locations in the vasculature is likely to affect signal strength and quality.^77,78,154 The striking diversity of signal transduction cascades documented in various cell types also involves crosstalk among the canonical and alternative signaling pathways (see Fig. 4-16). Finally, the growth factor microenvironment (see later discussion) greatly influences the mechanisms that control attachment, polarity, and directional motility of vascular cells. Aberration of these signals leads to dramatic changes in the ability of cells to attach, polarize, and migrate, as well as their growth factor– and anchorage-independent proliferation.

Signaling Triad of Extracellular Matrix, Integrin, and Growth Factors

Although the ECM-integrin signaling axis has been the focus of most investigations, it has become evident in recent years that subendothelial connective tissues make additional contributions to the bidirectional signaling of vascular and nonvascular cells. A number of growth factors and developmental morphogens are sequestered in the fibrillar and nonfibrillar constituents of ECM. Presumably, such ECM-bound factors have the potential to be released in a highly regulated manner in response to developmental cues or tissue injury and inflammation.^78,145,146 Furthermore, a number of ECM molecules serve as coreceptors for growth factors, as is the case for FGF and VEGF; both bind to heparin and HS chains of PGs that are needed for optimal ligand presentation and signal transduction. Proteoglycans modulate TGF-β signaling in a number of ways. Transforming growth factor beta binds to decorin in the ECM and in this state cannot bind to its cell surface receptors. Additionally, the binding of TGF-β to its signaling receptors is dependent on a plasma membrane-bound PG (β-glycan) and the cell surface glycoprotein endoglin, both of which serve as co-receptors. Finally, some growth factors can independently activate a number of downstream effectors of integrin signaling (FAK, Src, PI3-K, MAPK); this is exemplified by synergistic modulation of ECM-integrin signaling by IGF-1 and PDGF (Fig. 4-17).

Figure 4-17 Regulation of growth factor receptor (GFR) signaling by integrins.

Several mechanisms by which integrins might control GFR activity have been proposed. A, The first potential mechanism by which integrins control GFR activity involves recruitment of specific adapters (Shp2 and Gab1) to plasma membrane (left). Integrins are thought to recruit adapters to plasma membrane and concentrate them in close proximity to GFRs (right), thus enhancing their signaling. B, The second proposed mechanism is that integrins, upon close association with GFRs, change their subcellular localization of these receptors. Thus, coaggregation of integrins and GFRs in focal adhesions may alter the quality and/or strength of downstream signaling cascades that culminate in altered gene expression. There are compelling data to suggest that an association between GFRs and integrins may also facilitate crosstalk between extracellular matrix (ECM) liganded integrin and GFRs. DNA, deoxyribonucleic acid; FAK, focal adhesion kinase; MAPK, mitogen-activated protein kinase; MEK, mitogen extracellular signal regulated kinase; TF, transcription factor.

(Adapted from Alam N, Goel HL, Zarif MJ, et al: The integrin–growth factor receptor duet. J Cell Physiol 213:649–653, 2007.)

Comodulation of integrin and growth factor signaling has been documented for EGF, hepatocyte growth factor (HGF), IGF-1, and PDGF, both during early development of model organisms and in cell lines.^156–162 Evidently, a trimolecular interaction among the BMP homolog DPP, its cell surface receptor, and type IV collagen is crucial for optimal relay of signals involved in morphogenesis of the dorso-ventral axis in Drosophila. A similar mechanistic interaction of HGF with two other proteins has been reported. Apparently, fibronectin- or VN-bound HGF forms a trimolecular complex that contains (FN or VN)-HGF, c-Met (HGF receptor), and α₃β₄ integrin. A functional role of these interactions was corroborated by showing that ligand-independent activation of c-Met by cellular adhesion plays a crucial role in the growth and invasive potential of epithelial cells in response to HGF. Finally, it has been reported that the downstream signaling evoked by the tyrosine kinase transmembrane glycoprotein EGFR was complemented by integrins; thus, EGFR was shown to laterally coaggregate with α_vβ₃, α₂β₁, and α₆β₄ integrins.

As reviewed in a number of excellent publications,^78,144 the crosstalk between growth factors and integrins involves a number of mechanisms (see Fig. 4-17). First, it involves integrin’s ability to concentrate signaling adapters in close proximity of growth factor receptors to enhance their intracellular signaling ability. An additional mechanism of crosstalk may be dependent on lateral mass aggregation of integrins, a process posited to alter the location and/or concentration of growth factor receptors on the cell surface. The observation that α₁β₂ integrin and EGFRs were coaggregated at the foci of cell-cell contact bolsters such a mechanism. Furthermore, integrin-mediated adhesive interactions may alter the rate of internalization and degradation of some receptors (e.g., PDGFR). Conversely, integrin activation may lead to enhanced expression of some growth factors; thus, it was shown that increased biosynthesis of VEGF and IGF-2 occurred via the actions of α₆β₄ and α₁β_1A integrins, respectively. These observations and many others are consistent with the notion that bidirectional crosstalk between integrins and growth factors is highly relevant to cellular behavior and phenotype. However, the relative contribution of these two processes to bidirectional signaling remains controversial. Similarly, the hierarchical relationship between integrin and growth factor signaling pathways remains to be sorted out.^78,143,145

Vascular Extracellular Matrix and Platelet Interaction

Circulating platelets act as sentinels of vascular integrity and do not attach to vessel walls under normal conditions. However, upon injury of blood vessel walls, platelets rapidly adhere to ECM of the subendothelium and neointima (formed after repeated damage to vessel wall) and aggregate with each other at the site of injury. The adhesive interactions between subendothelial ECM and platelets and thrombus formation have been extensively studied.^86,163–165 Binding and activation of platelets by the subendothelial ECM is modulated by hemodynamic stress, composition of the vascular tissue, and extent and depth of the lesion.

The fibrillar collagens, of which types I, III, IV, V, and VI are widely distributed in the vascular ECM, are key initiators of thrombosis in injured vessels.^7,9,166 The hemostatic cascade begins when platelets adhere to the exposed subendothelial collagens via vWF, a large glycoprotein that is constitutively synthesized by ECs.^86,87 The vWF both circulates in plasma as a soluble molecule and is stored in the α granules of platelets. Endothelial or plasma vWF, which forms multimeric aggregates, binds to collagen. It is believed that high hemodynamic stress causes local uncoiling of multimeric vWF, further facilitating its binding to platelet GPIbα and collagen.^167,168 Additional interactions of collagen with α₂β₁ integrin and glycoprotein VI (GPVI) synergize the binding of platelets to vascular ECM. Although a pivotal role of collagens in thrombosis is borne out by numerous observations, the relative contributions of individual fibrillar collagens to this process appears to be highly context dependent.^86,87,169 This view is supported by the observation that platelets adhere to and are activated by types I-V collagens under relatively high and moderate hemodynamic stress. In contrast, binding and activation of platelets by type V-VIII collagens occurs at much lower shear stress, and type V collagen binds to platelets only under static conditions.

Over the years, many platelet-associated proteins have been claimed as candidate collagen receptors.^170–178 However, a careful review of the literature would suggest that some of these claims reflect vagaries of different experimental models (in vivo vs. ex vivo), while others may have emanated from a lack of technical or interpretational rigor.^87,166,169 Historically, the search for collagen receptors on platelets began with studies using purified type I and III collagens or their cyanogen bromide (CB) subfragments that were tested for their ability to bind and activate platelets. These studies led to the discovery that purified native type I/III collagens and some of their CB fragments interacted with platelet integrin α₂β₁. More recently, a series of overlapping triple-helical peptides spanning human collagen types I and III have been systematically tested for their ability to bind and activate platelets.^179–181 These elegant studies have unraveled discrete sequence motifs on types I and III collagen that interact with receptors on the surface of platelets and vWF. Based on these analyses, it has been surmised that type I collagen contains four α₂β₁ integrin-binding sites of varying affinities; apparently, similar sequence motifs are also conserved in type III collagen. The molecular interactions between high-affinity type I collagen triple-helical peptide and α₂β₁ integrin have been elucidated with precision. These and related studies have shown that fibrillar collagens also interact with α_IIbβ₃, albeit indirectly. Finally, the use of well-defined triple-helical collagen peptides has led to the discovery of amino acid sequence motifs in types I and III collagen that specifically bind to the platelet surface GPVI, as well as to vWF.^86,87,180

How collagen-platelet interactions culminate in the formation of a thrombus has been investigated in considerable detail.^86,87,180 Under high shear stress, the obligatory first step is co-engagement of α₂β₁ integrin and vWF by collagens. Platelet proteins GPVI and GPIV (CD36) strengthen the initial encounter with collagen. Glycoprotein VI has a modular organization that is composed of 2 Ig domains, an O-linked glycosylated stalk, an intramembrane domain, and an intracellular tail. In platelets, GPVI exists in association with the FcR γ-chain, a transmembrane protein containing an immunoreceptor tyrosine activation motif (ITAM) motif. Interaction of platelets with collagen triggers phosphorylation of the FcR γ-chain and signaling that mimics signaling cascades associated with T-cell activation. The intracellular signaling is mainly driven by Ca⁺⁺ mobilization that also occurs in response to other platelet-activating stimuli (e.g., thrombin, adenosine diphosphate [ADP]). The role of CD36 in thrombosis is somewhat debatable in light of the observation that nearly 5% of Japanese lack functional CD36 protein in their platelets without overt complications of hemostasis.

Platelets also adhere to other components of ECM, particularly to FN and laminin, both of which are essential ECM components of the subendothelial connective tissues.^{86,87,182,183} Since the original 1978 report of Hynes et al. supporting a role of FN in platelet adhesion, a number of apparently conflicting data have muddled this issue.^{86,87,182,183} There is little doubt that platelets adhere to FN; platelets possess two integrin receptors that bind FN (i.e., α₅β₁, α_IIbβ₃). Depletion of FN from plasma reduced the ability of platelets for thrombus formation on fibrillar substratum or subendothelium, and if FN was restored, interplatelet aggregation and activation was restored. The NH₂-terminal 70-kD fragment of FN could be incorporated in the clot and was cross-linked to fibrin. Further bolstering a positive role of FN in thrombosis is the phenotype of mice with various gene knockouts. Thus, FN^−/− mice were shown to elicit defective thrombus formation. Similarly, mice lacking both fibrinogen and vWF elicited thrombosis primarily driven by FN and other proteins at the sites of blood vessel injury. However, it was noted that FN-platelet interactions were apparently too weak to withstand the normal shear stress. Based on these observations, it has been posited that FN plays a complementary role in thrombus formation in the presence of fibrinogen and vWF.^184,185

These paradoxical observations on the role of FN in platelet activation may be reconciled in a model that suggests that the ultimate response of platelets may be determined by their initial encounter with different sets of ECM molecules in the endothelium. It is known that platelets can assemble a fibrillar network of FN from soluble FN. However, when platelets attach to vWF, it suppresses their ability to deposit FN. Under these conditions, aggregation and activation of platelets may be driven mainly by fibrinogen-α_IIbβ₃ integrin interaction. In contrast, if fibrin or collagen initiates adhesion of platelets, they acquire a greater ability to convert plasma FN into supramolecular FN that gets incorporated it into thrombi. It follows then that aggregation and clustering of platelets via fibrinogen or assembled FN may be regulated in a mutually exclusive manner.^86,87

In the injured vessel wall, platelets are exposed to various forms of laminin; thus, laminin-411 and laminin-511 are present not only in the vascular ECM but also in platelets that contain laminin-522. Laminins bind to platelets and thus may play a role in their adhesion and activation. Despite the presence of laminin-binding integrin receptors on the platelet surface, the relative role of laminin-mediated adhesion in thrombus formation remains to be more precisely delineated. Analogous to more nuanced regulation of FN-platelet interaction, where interplay among fibrinogen, vWF, and VN negatively regulate FN assembly, interactions among laminin, fibrin, and collagen elicit an opposite effect on FN. Thus, it is not unreasonable to conclude that the precise outcome of platelets’ encounter with the endothelium is highly context dependent.^186,187

Finally, we should note that nonfibrillar constituents of the subendothelium might also impinge on the function of platelets in a number of ways. For instance, heparin binds to serine protease inhibitors (SERPINS), antithrombin III, and heparin cofactor II and thus accelerates formation of the thrombin-antithrombin complex. Thrombin is a pivotal serine protease of the coagulation pathway, and it also exerts a positive feedback effect on its own biogenesis by activating factors V and VIII. Inhibition of thrombin by GAGs is a key mechanism in the regulation of blood clotting.^188–191 Platelets contain platelet factor 4 (PF4) in α granules, where it is stored in a complex with chondroitin-4-sulfate PG. When released, PF4 has the ability to neutralize the anticoagulant effects of heparin and heparan. The PG-PF4 complex is dissociated by conditions of high ionic strength and by GAGs. Based on these interactions, it is conceivable that any HS– and DS–containing PGs present in blood vessel walls could serve to control deposition of PF4 locally by competitively dissociating the PG-PF4 complex.