Collagen

The collagen family is the most abundant class of proteins in the human body. It is also one of the most versatile. Collagens form a wide range of different structures with remarkable mechanical properties. Weight for weight, fibrous collagens are as strong as steel. Their name, which comes from the Greek words for “glue” and “birth,” reflects the long-known adhesive properties of denatured collagen extracted from animal tissues.

The defining feature of collagens is a rod-shaped domain composed of a triple helix of polypeptides (Fig. 29-1). Each polypeptide folds into a left-handed helix that repeats every third residue with the side chains on the outside. Three of these helices associate to form a triple helix that may be up to 420 nm long. The triple helical domains have a repeating amino acid sequence: glycine-X-Y, where X is most often proline and Y is most often hydroxyproline. The small glycine residues allow tight contact between the polypeptides in the core of the triple helix. Larger residues, even alanine, interfere with packing. Poly-l-proline has a strong tendency to form a left-handed helix like individual collagen chains but does not form a triple helix, owing to steric interference. The triple helix is most stable if all X residues are proline and all Y residues are hydroxyproline, but other residues at some of these positions are essential for collagen to assemble higher-order structures. (Despite their name, α-chains, the collagen polypeptides do not form α-helices.)

Figure 29-1 collagen triple helix. A., End-on view of three left-handed polyproline type II helices with glycines (G) in the core. B, Longitudinal view of the strands of a triple helix. C, Space-filling model of the structure of a short collagen triple helix.

(A, Redrawn from van der Rest M, Garrone A: Collagen family of proteins. FASEB J 5:2814–2823, 1991. C, PDB file: 1BKV.)

The collagen family is remarkably diverse. Humans have about 100 genes with collagen triple repeats, and more than 20 specialized collagen proteins have been characterized (Fig. 29-2 and Appendix 29-1).

Figure 29-2 comparison of major collagen families. Scale drawings and micrographs of collagen molecules and their assembly into higher-order structures. AF, anchoring fibrils; BM, basement membrane.

(Redrawn from van der Rest M, Garrone A: Collagen family of proteins. FASEB J 5:2814–2823, 1991.)

Other proteins, including the extracellular enzyme acetylcholine esterase (see Fig. 11-8) and some cell surface receptors, have similar triple helical domains but are not classified as collagens. To be a collagen, a protein must also form fibrils or other assemblies in the extracellular matrix. Nematodes, which lack connective tissue, seem to have lost the genes for fibrillar collagens but have elaborated a family of 160 genes for collagens that form their cuticle.

Collagen biochemistry is challenging because many tissue collagens are insoluble, owing to covalent cross-linking between proteins. Historic purification protocols began with proteolytic digestion to liberate protease-resistant triple helical fragments. A newer approach has been to isolate intact collagens produced by cells in tissue culture.

The size and shape of collagens vary according to function. Collagens are named numerically (type I, type II, etc.) in the order of their discovery, a nomenclature that bears no relationship to their function. Appendix 29-1 groups collagens according to function. Polypeptides are called α-chains, and Roman numerals in their names correspond to their type number. Some collagens are homotrimers of three identical α-chains. Others are heterotrimers of two or three different α-chains. Some chains (e.g., [a1(II)]) are used in more than one type of collagen.

Fibrillar Collagens

Triple helical rod-shaped collagen molecules about 300 nm long self-associate to form banded fibrils (Fig. 29-2). Collagen fibrils provide tensile strength to tendons, ligaments, bones, and dense connective tissue, thus reinforcing most organs. They also form the scaffolding for cartilage and the vitreous body in the eye. Fibrillar collagens are widespread in nature and have been highly conserved during evolution, so the homologs from sponges to vertebrates are similar. Each fibrillar collagen can form homopolymers in vitro; but in vivo, most form heteropolymers with at least one other type of fibrillar collagen (Appendix 29-1). This mix of the fibrillar collagen subunits is one factor that regulates the size of collagen fibers. Proteoglycans also participate (Appendix 29-2).

The biosynthesis and assembly of fibrillar collagens involve a remarkable number of posttranslational modifications, including several rounds of precise proteolytic cleavage, glycosylation, catalyzed folding, and chemical cross-linking (Fig. 29-4). The final product is a smooth fibril with staggered molecules that are cross-linked to their neighbors. These strong but flexible collagen fibrils reinforce all the tissues of the body, where they form a variety of higher-order structures. Loose connective tissue (see Fig. 32-1A) has an open network of individual fibrils or small bundles of fibrils that support the cells. In many tissues, the fibrils of type I and associated collagens aggregate to form the so-called collagen fibers that are visible by light microscopy (Fig. 29-3A). In extreme cases, such as in tendons, the extracellular matrix consists almost exclusively of tightly packed, parallel bundles of collagen fibers (see Fig. 32-1B). Layers of orthogonal collagen fibers make the transparent cornea through which one sees (Fig. 29-3C). In bone, type I collagen fibrils form regular layers reinforced by calcium phosphate crystals (see Fig. 32-5). In cartilage and the vitreous body of the eye, type II collagen fibrils trap glycosaminoglycans and proteoglycans, which retain enough water for the matrix to resist compression (see Fig. 32-3) and, in the case of the eye, to provide an optically clear path for light.

Figure 29-4 biosynthesis and assembly of fibrillar collagen illustrating details covered in the text. A, Translation of α-chains, chain registration, and folding. B, Secretion, assembly, and cross-linking.

(Redrawn from Prokop DJ: Mutations in collagen genes as a cause of connective tissue diseases. N Engl J Med 326:540–546, 1992. Copyright © 1992 Massachusetts Medical Society. All rights reserved.)

Figure 29-3 micrographs of collagen fibrils in connective tissues. A., Collagen fibrils (pink) in the dense connective tissue of the dermis. B, Electron micrograph of a thin section of a fibroblast, collagen fibrils, and elastic fibers. C, Orthogonal layers of collagen fibrils in the cornea of the eye.

(A, Courtesy of D. W. Fawcett, Harvard Medical School, Boston, Massachusetts. B, Courtesy of J. Rosenbloom, University of Pennsylvania, Philadelphia. C, Courtesy of E. D. Hay, Harvard Medical School, Boston, Massachusetts.)

Biosynthesis and Assembly of Fibrillar Collagens

All fibrillar collagens are most likely to be produced by similar mechanisms, but type I collagen has been studied the most extensively. Type I collagen is synthesized and secreted by fibroblasts, using the exocytic pathway that is employed for other secretory proteins (see Chapter 21), but the biosynthesis of collagen is noteworthy for the extensive number of processing steps required to prepare the protein for assembly in the extracellular matrix.

Large genes with 42 exons encode the α-chains of type I collagen. All the exons for the triple helical domain are derived by duplication and divergence from a primordial exon of 54 base pair (bp) coding for 18 amino acids or six turns of polyproline helix. About half of the exons consist of 54 bp; a few with 45 bp have lost one Gly-X-Y; and the rest are 108 (2 × 54) or 162 (3 × 54) bp. Distinctive exons encode the N- and C-terminal globular domains.

The initial transcript, referred to as preprocollagen, translocates into the lumen of the rough endoplasmic reticulum, where intracellular processing begins (Fig. 29-4). First, removal of the N-terminal signal sequence yields procollagen with unfolded α-chains with N- and C-terminal nonhelical propeptides. Second, enzymes hydroxylate some prolines and lysines. Third, enzymes add sugars (gal-glu or gal) to the delta-carbon of some lysines, by a mechanism distinct from the typical glycosylation of asparagine or serine.

A novel mechanism initiates the folding of collagen in the endoplasmic reticulum: the C-terminal propeptides of three α-chains form a globular structure stabilized by cysteines linked in disulfide bonds. An enzyme, protein disulfide isomerase, catalyzes the formation of these disulfides. Formation of this globular domain has three important consequences. First, it ensures the correct selection of α-chains (two a1-chains and one a2-chain in the case of type I collagen). Second, it aligns the three polypeptides with their C-terminal Gly-X-Y repeats in register, ensuring that the triple helix forms with all three chains in phase. Third, the globular propeptides prevent assembly of procollagen into fibrils during transit through the secretory pathway. Given their repeating Gly-X-Y structure, separated collagen chains without propeptides associate indiscriminately and out of register with other chains. For example, gelatin is simply a mixture of collagen chains without propeptides. Boiling dissociates the chains from each other. When cooled, the chains randomly associate out of register at random positions along their lengths, forming a branching network that solidifies into the gel that is used in food preparation.

Following selection and registration of the three α-chains, the helical rod domains zip together, beginning at the C-terminus. Correct folding of the triple helix requires all-trans peptide bonds. Because proline forms cis and trans peptide bonds randomly, the slow isomerization of cis prolyl-peptide bonds to trans limits the rate of triple helix folding in vitro. The enzyme prolyl-peptide isomerase catalyzes the interconversion of these prolyl-peptide bonds and rapid folding of the triple helix in vivo. The resulting rod-shaped, triple-helix glycoprotein is called procollagen.

Procollagen passes through the Golgi apparatus and moves in vesicles to the cell surface, where it is secreted. Some cells have specialized collagen assembly sites (Fig. 29-4). Like ships laying down communication cables on the ocean floor, fibroblasts help to determine the arrangement of collagen fibrils as they move through tissues (Fig. 29-3C).

Outside the cell, proteolytic enzymes—procollagen proteases—cleave the propeptides from the triple helical domain, forming the mature collagen molecule (formerly called tropocollagen). Relieved of its inhibitory propeptides, collagen self-assembles into fibrils by a classical entropy-driven process (Fig. 29-5). Adjacent collagen molecules are staggered by 67 nm, so a 35-nm gap is required between the ends of the collagen molecules (five staggers at 67 nm = 335 nm = one molecular length of 300 nm + a 35-nm gap).

Figure 29-5 structure of collagen fibrils. Electron micrographs and drawing of molecular packing.

(Micrographs courtesy of Alan Hodges, Marine Biological Laboratory, Woods Hole, Massa-chusetts.)

Weak, noncovalent bonds between collagen molecules specify the self-assembly of fibrils but provide little tensile strength, so covalent cross-linking is required for reinforcement. For most fibrillar collagens, the enzyme lysyl oxidase catalyzes the formation of covalent bonds between the ends of collagen molecules (Figs. 29-4 and 29-6). The enzyme oxidizes the e amino groups of selected lysines and hydroxylysines to aldehydes. These aldehydes react spontaneously with nearby lysine and hydroxylysine side chains to form a variety of covalent cross-links between two or three polypeptides. Disulfide bonds, rather than modified lysine side chains, cross-link type III collagen fibrils. Covalent bonds between the inextensible triple helices give mature collagen fibrils their great tensile strength.

Figure 29-6 covalent cross-linking of collagen molecules. After lysyl oxidase oxidizes hydroxylysine side chains, the aldehydes condense with each other and a lysine to form two- and three-membered (shown) cross-links between adjacent collagen mole-cules.

Point mutations or deletions in collagen genes or lack of function of one of the enzymes that processes collagen (lysyl hydroxylase, lysyl oxidase, or procollagen proteases) can each cause defective collagen fibrils (Appendix 29-1). These defects cause a remarkable variety of deforming and even lethal human diseases: brittle bones (osteogenesis imperfecta), fragile cartilage (several forms of dwarfism), and weak connective tissue (Ehlers-Danlos syndrome). Chapter 34 covers these diseases in more detail.

Sheet-Forming Collagens

A second group of collagens polymerizes into sheets rather than fibrils (Fig. 29-2). These sheets surround organs, epithelia, or even whole animals. Six different human genes for type IV collagen encode proteins that form net-like polymers that assemble into the basal lamina beneath epithelia (Fig. 29-7) and around muscle and nerve cells. The concluding section of this chapter provides details about basal lamina structure, function, and diseases. Hexagonal nets of type VIII collagen form a special basement membrane (Descemet’s membrane) under the endothelium of the cornea. Related collagens form the cuticle of earthworms and the organic skeleton of sponges.

Figure 29-7 anchoring fibrils of type vii collagen. Electron micrograph of a thin section of human skin reacted with a gold-labeled antibody to the C-terminal domain of type VII collagen. Top to bottom, Basal epithelial cell with keratin intermediate filaments (IFs) attached to hemidesmosomes, which link to the basal lamina. Short fibrils of type VII collagen link the basal lamina to plaques in the dermis. Both ends of these bipolar fibrils (Fig. 29-2) are labeled with gold. Bar is 0.1 mm.

(Courtesy of D. R. Keene, Portland Shriners Hospital, Oregon.)

Elastic Fibers

In contrast to inextensible collagen fibrils, elastic fibers are similar to rubber. They are found throughout the body but are prominent in the connective tissue of skin, the walls of arteries (Fig. 29-8), and the lung. They recoil passively after tissues are stretched. Every time the heart beats, pressurized blood flows into and stretches the large arteries. Energy stored in elastic fibers pushes blood through the circulation between heartbeats.

Figure 29-8 elastic fibers in the wall of a small artery. A., Light micrograph of a cross section prepared with a special red stain to visualize elastic fibers. The boxed area includes the internal elastic lamina between the endothelial cells lining the lumen and the underlying smooth muscle cells. B, Electron micrograph of a longitudinal thin section illustrating the internal elastic lamina. In such standard preparations, elastic fibers stain poorly and appear amorphous except for occasional 10-nm microfibrils on the surface.

(Courtesy of Don W. Fawcett, Harvard Medical School, Boston, Massachusetts.)

Elastic fibers are a composite material: A network of fibrillin microfibrils is embedded in an amorphous core of cross-linked elastin, which makes up 90% of the organic mass (Fig. 29-9). Fibroblasts produce both components. Loose bundles of microfibrils initiate assembly. A third protein, called fibulin, is required for elastin subunits to assemble between the micro-fibrils.

Figure 29-9 electron micrographs of developing elastic fibers from a fetal calf. A., Longitudinal section. B, Cross section. Fibrillin microfibrils form a scaffolding for elastin, which stains darkly in this preparation.

(Courtesy of J. Rosenbloom, University of Pennsylvania, Phila-delphia.)

Fibrillin is the primordial component of elastic fibers, having arisen in Cnidarians (see Fig. 2-9). It is a long, floppy protein consisting of a tandem array of domains (Fig. 29-10). Humans have two fibrillin genes, and both fibrillin-1 and fibrillin-2 are components of 10-nm microfibrils, along with several glycoproteins. In microfibrils, fibrillin molecules interact head to tail with a reinforcing disulfide bond, but their arrangement is still being investigated. Microfibrils are about 100 times stiffer than elastin, and they stretch by rearrangement of molecules and domains rather than unfolding.

Figure 29-10 domain organization of human fibrillin. A tandem array of independently folded domains, including 47 epidermal growth factor–like (EGF-like) domains, forms a linear molecule.

(Redrawn from Rosenbloom J, Abrams WR, Mecham R: Extracellular matrix 4: The elastic fiber. FASEB J 7:1208–1218, 1993.)

Elastin subunits are a family of closely related 60-kD proteins called tropoelastins, the products of alternative splicing from a single elastin gene. They have long sequences that are rich in hydrophobic residues interrupted by short sequences with pairs of lysines separated by two or three small amino acids (Fig. 29-11). Lysine-rich sequences are thought to form α-helices with pairs of lysines adjacent on the surface.

Figure 29-11 elastin polypeptides and cross-linking reactions. A., Lysine-rich helical domains separate random chains rich in hydrophobic residues. B–C, Lysyl oxidase converts lysine amino groups to aldehydes, which react with other lysines to form simple linear cross-links or six-membered rings linking two polypeptides. If the peptide bonds are hydrolyzed experimentally (not shown here), the linear cross-link is released as leucyl-norleucine (LNL) and the six-membered cross-link is released as the amino acid desmosine.

(Redrawn from Rosenbloom J, Abrams WR, Mecham R: Extracellular matrix 4: The elastic fiber. FASEB J 7:1208–1218, 1993.)

As tropoelastin assembles on the surface of elastic fibers, lysyl oxidase oxidizes paired lysines of tropoelastin to aldehydes. Oxidized lysines condense into a desmosine ring that covalently cross-links tropoelastin molecules to each other (Fig. 29-11). The four-way cross-links, involving pairs of lysines from two tropoelastin molecules, are unique to elastin. The same enzyme catalyzes the cross-linking of collagen, but it forms only two- and three-way cross-links.

Elastic fibers are similar to rubber except that elastic fibers require water as a lubricant. Hydrophobic segments between the cross-links are thought to form extensible random coils that account for the elastic properties of the fibril (Fig. 29-12). A difference in entropy of the polypeptide in the contracted and stretched states is thought to be the physical basis for the elasticity. The birefringence of elastic fibers increases when they are stretched, presumably as a result of alignment of polypeptide chains. Stretched fibers store energy, owing to ordering (low entropy) of the polypeptide chains. Fibers shorten when the resistance is reduced, because the polypeptide chains return to their disordered, lower-energy, higher-entropy state. Unfolding of fibrillin domains may contribute to the elasticity similar to titin in muscle cells (see Fig. 39-7), but this has not been studied.

Figure 29-12 physical model of elastin elasticity. A., Contracted state with low-energy disordered chains having high entropy. B, Stretched state with high-energy ordered chains having low entropy. Elastin polypeptides form a continuous, covalently bonded network. Application of force stretches the chains between the cross-links. This is a low-entropy, high-energy state. Reduced force allows the chains to contract into a more disordered, higher-entropy state with lower energy.

(Redrawn from Rosenbloom J, Abrams WR, Mecham R: Extracellular matrix 4: The elastic fiber. FASEB J 7:1208–1218, 1993.)

Only embryonic and juvenile fibroblasts synthesize elastic fibers, which turn over slowly, if at all, in adults. Consequently, adults must make do with the elastic fibers that are formed during adolescence. Fortunately, these fibers are amazingly resilient. Arterial elastic fibers withstand more than 2 billion cycles of stretching and recoil during a human life. Many tissues become less elastic with age, particularly the skin, which is subjected to damage from ultraviolet irradiation. Compare, for example, how readily the skin of a baby recoils from stretching compared with that of an aged person. The loss of elastic fibers in skin is responsible for wrinkles.

Collagens are found across the phylogenetic tree, but only vertebrates are known to produce elastin. Invertebrates evolved two completely different elastic proteins. Mollusks have elastic fibers composed of the protein abductin. Insects use another protein, called resilin, to make elastic fibers.

Marfan syndrome, the disease caused by dominant mutations in the fibrillin-1 gene, illustrates the physiological functions of elastic fibers. Elastic fibers of patients with Marfan syndrome are poorly formed, accounting for most of the pathological changes that are observed. Most dangerously, weakness of elastic fibers in the aorta leads to an enlargement of the vessel, called an aneurysm, which is prone to rupture, with fatal consequences. Prophylactic replacement of the aorta with a synthetic graft and medical treatment with drugs that block β-adrenergic receptors (see Fig. 27-3) allow patients a nearly normal life span. In some patients, a floppy mitral valve in the heart causes regurgitation of blood from the left ventricle back into the left atrium. Weak elastic fibers that suspend the lens of the eye result in dislocation of the lens and impaired vision. Weak elastic fibers result in lax joints and curvature of the spine. Most affected patients are tall, with long limbs and fingers, but the connection of these features to fibrillin is not known. The manifestations of the disease are quite variable, even within one family, for reasons that are not understood. Mutations in fibrillin-2 cause congenital contractural arachnodactyly, a disease characterized by joint stiffness. New fibrillin mutations arise spontaneously, and most families that are tested have different mutations, including both point mutations and deletions. All patients are heterozygotes. Most of the known fibrillin-1 mutations make the protein unstable and susceptible to proteolysis. Other point mutations interfere with folding.

Dominant mutations in the elastin gene cause a human disease called cutis laxa. The skin and other tissues of patients with this disease lack resilience.

Glycosaminoglycans and Proteoglycans

Glycosaminoglycans (GAGs, formerly called mucopolysaccharides) are long polysaccharides made up of repeating disaccharide units, usually a hexuronic acid and a hexosamine (Fig. 29-13). With one important exception—hyaluronan—GAGs are synthesized as covalent, posttranslational modifications of a large family of proteins called proteoglycans. These proteins vary in structure and function, but their associated GAGs confer some common features.

Figure 29-13 synthesis of glycosaminoglycans. a–b, Three short oligosaccharides link GAGs (left) to proteoglycan core proteins (right). A, A tetrasaccharide anchors chondroitin sulfate (CS), dermatan sulfate, and heparan sulfate (HS) to serine residues. B, Two different, branched oligosaccharides link keratan sulfate (KS) to either serine or asparagine. C–F, Four parent polymers and postsynthetic modifications. C, Hyaluronan [D-glucuronic acid β (1 →3) D-N-acetylglucosamine β (1 →4)]_n (n ≥ 25,000) is not modified postsynthetically. D, Chondroitin sulfate and dermatan sulfate are synthesized as [d-glucuronic acid β (1 →3) D-N-acetylgalactosamine β (1 → 4)]_n (n usually <250) and then modified. Some N-acetylgalactosamines are sulfated. In dermatan sulfate, D-glucuronic acids are epimerized to L-iduronic acid. E, Keratan sulfate is synthesized as [d-galactose β (1 → 4) D-N-acetylglucosamine β (1 → 3)]_n (n usually = 20–40) and then modified by sulfation. F, Heparan sulfate/heparin is synthesized as [d-glucuronic acid β (1 → 4) D-N-acetylglucosamine α (1 → 4)]_n (n usually <100) and then modified by sulfation and by epimerization of D-glucuronic acid to L-iduronic acid. galNAc, N-acetylgalactosamine; glcNAc, N-acetylglucosamine.

(Redrawn from Wright TN, Heinegard DK, Hascall VC: Proteogly-cans, structure and function. In Hay ED [ed]: Cell Biology of the Extracellular Matrix, 2nd ed. New York, Plenum Press, 1991, pp 45–78. With the kind permission of Springer Science and Business Media.)

Many cells, including all vertebrate cells, synthesize proteoglycans. Most are secreted into the extracellular matrix, where they are major constituents of cartilage, loose connective tissue, and basement membranes. Mast cells package the proteoglycan serglycin, along with other molecules in secretory granules. A few proteoglycans, including syndecan and CD44, are plasma membrane proteins with their GAGs exposed on the cell surface.

Of the known GAGs, hyaluronan (formerly called hyaluronic acid) is exceptional in two regards. First, enzymes on the cell surface synthesize the alternating polymer of [d-glucuronic acid β (1 → 3) D-N-acetyl glucosamine β (1 → 4)]_n (Fig. 29-13). Other GAGs are synthesized as posttranslational modifications of a core protein. Second, hyaluronan is not modified postsynthetically, as are all other GAGs. The linear polymer, often exceeding 20,000 disaccharide repeats (a length >20 mm) is released into the extracellular space.

In contrast to proteins, nucleic acids, and even N-linked oligosaccharides, which are precisely determined macromolecular structures, the GAG chains of proteoglycans appear to vary both in length and the sequence of the sugar groups. The four-step synthesis of GAGs (Fig. 29-13) explains this variability:

The nomenclature for proteoglycans is in flux, so specific proteoglycans may have multiple names. The old nomenclature was based on the identity of the GAGs that are bound to the protein. For example, the major proteoglycan of basement membranes was called heparan sulfate proteoglycan. This nomenclature is imprecise, as more than one type of proteoglycan carries heparan sulfate. Once the core proteins were characterized, it was reasonable to develop a nomenclature based on these core proteins. Consequently, the basement membrane proteoglycan is now known as perlecan, the name of its core protein. The weakness of this system is that the protein name reveals nothing about the associated GAGs. This information is important because various cells add different GAGs to the same core protein or can modify the same GAG in different ways.

Cells secrete many proteoglycans into the extracellular matrix, but they retain some types on the plasma membrane through transmembrane polypeptides or a glycosylphosphatidylinositol anchor (Appendix 29-2 and Fig. 29-14). The core proteins vary in size from 100 to 4000 amino acids. Many are modular, consisting of familiar structural domains: EGF, complement regulatory protein, leucine-rich repeats, or lectin. Three collagens carry GAG side chains: Types IX and XII have chondroitin sulfate chains, and type XVII has heparin sulfate chains.

Figure 29-14 scale drawings of a variety of proteoglycans. Core proteins are purple and pink, and the glycosaminoglycans are color-coded. Proteins were named according to the following: aggrecan aggregates along hyaluronan; decorin decorates collagen fibrils; perlecan resembles a string of pearls; serglycin has 24 ser-gly repeats; syndecan (syndein = link) links cells to the matrix; glypican has a glycosylphosphatidylinositol (GPI) membrane anchor. CS, chondroitin sulfate; DS, dermatan sulfate; Hep, hep-arin; HS, heparan sulfate; KS, keratan sulfate.

(Redrawn from Wright TN, Heinegard DK, Hascall VC: Proteoglycans, structure and function. In Hay ED [ed]: Cell Biology of the Extracellular Matrix, 2nd ed. New York, Plenum Press, 1991, pp 45–78. With the kind permission of Springer Science and Business Media.)

The number of GAGs attached to the core protein varies from one (decorin) to more than 200 (aggrecan) (Fig. 29-14). A particular core protein can have identical (fibroglycan, glypican, versican) or different (aggrecan, serglycin, syndecan) types of GAGs. Some cell types can add different GAGs to the same core protein or secrete a core protein without GAGs.

Given their physical properties and distribution among the fibrous elements of the extracellular matrix, proteoglycans and hyaluronan are thought to be elastic space-fillers. Each hydrophilic disaccharide unit bears a carboxyl or sulfate group or both, so GAGs are highly charged polyanions that extend themselves by electrostatic repulsion in solution and attract up to 50 g of water per gram of proteoglycan. Hyaluronan, the largest GAG, occupies a vast volume. A single hydrated molecule of 25,000 kD occupies a volume similar to that of a small organelle with a diameter of 200 nm. Retention of water by hyaluronan and aggrecan-keratan sulfate/chondroitin sulfate proteoglycan is essential in cartilage (see Fig. 32-3). In the extracellular matrix of other tissues, networks of densely charged hyaluronan restrict water flow, limit diffusion of solutes (especially macromolecules), and impede the passage of microorganisms. Hyaluronan and proteoglycans also act as lubricants in joint cavities and as an optically transparent, space-filling medium in the vitreous body of the eye.

Beyond these mechanical functions, proteoglycans influence cellular behavior such as adhesion or motility. Transmembrane proteoglycans can link cells to fibronectin and connective tissue collagens. Syndecan provides a particularly clear example. Lymphocytes express syndecan twice: early in their maturation, when they adhere to matrix fibers in the bone marrow, and later, when, as mature plasma cells, they adhere to the matrix of lymph nodes. In between, syndecan expression is lower while the lymphocytes circulate in the blood.

Carefully regulated expression allows proteoglycans to influence embryonic development and wound heal-ing in at least three different ways. First, both decorin and fibromodulin regulate assembly of collagen fibrils. Second, membrane-bound proteoglycans, including syndecan and glypican, act as coreceptors for growth factors. Third, many polypeptide growth factors (including platelet-derived growth factor and transforming growth factor-b) bind to proteoglycans in the extracellular matrix. This allows the matrix to concentrate circulating growth factors at specific locations and to release them locally over a period of time.

The well-known anticoagulant effects of heparin and heparan sulfate are attributable to their ability to bind both thrombin (the proteolytic enzyme that converts fibrinogen to fibrin) and a thrombin inhibitory pro-tein. This promotes interaction of the inhibitor with thrombin and inactivates the clotting cascade. A short sequence of five modified sugars has the anticoagulant activity.

Adhesive Glycoproteins

In principle, the macromolecules of the extracellular matrix and the constituent cells might interact relatively nonspecifically, but the evidence suggests that specific molecular interactions mediate virtually all of the interactions that organize the matrix and the associated cells. Most interactions are between proteins. Some are between proteins and sugars. Although some of these interactions are direct (with some cell surface receptors binding collagen directly), adapters called adhesive glycoproteins mediate many of the interactions (Appendix 29-3).

Adhesive glycoproteins were discovered by using biochemical assays for factors that favor particular interactions, such as adherence of cells to a matrix component. Further work revealed that adhesive glycoproteins are more than molecular glue; they also provide cells with signals required for the development and repair of tissues. Cells receive these signals when they bind to the matrix components. Chapter 30 focuses on their receptors.

Adhesive glycoproteins provide specific molecular interactions in the matrix by binding to cells, matrix macromolecules, or both. Adhesive proteins with multiple binding sites for cell surface receptors link cells together. For example, fibrinogen aggregates platelets during blood clotting (see Fig. 30-14). Other adhesive proteins link cells to the extracellular matrix. For instance, fibronectin mediates the attachment of cells to fibrin and collagen (Fig. 29-15). A third group of adhesive proteins link matrix macromolecules together. For example, nidogen attaches laminin to collagen and link protein attaches aggrecan-proteoglycan to hyaluronan.

Figure 29-15 domain organization of fibronectin. A linear array of FN-I (45 residues), FN-II (45 residues), and FN-III (90 residues) domains forms a rod-shaped subunit. Disulfide bonds near the C-termini covalently link two identical subunits in the dimeric molecule. Ligand-binding sites are indicated. The FN-III domain 10 contains the RGD sequence that binds cell surface integrins. Binding sites for fibrin, collagen, and glycosaminoglycans are indicated. Alternative splicing at sites ASRB, ASRA, and ASRC creates different iso-forms.

(Reference: Potts JR, Campbell ID: Fibronectin structure and assembly. Curr Opin Cell Biol 6:648–655, 1994. PDB files: FN7, 1PDC, 1FNA.)

The variety of tasks requires numerous adhesive proteins. In fact, the diversity exists beyond the named proteins (Appendix 29-3), as multiple genes or, more commonly, alternative splicing of the product of a single gene (see Fig. 16-6), generate multiple, functionally distinct isoforms of most of the named proteins. Particular isoforms are often expressed in specific tissues at predictable times during development.

Most adhesive glycoproteins are constructed of a series of compact modules (see Fig. 3-13 and Appendix 29-3). During evolution, duplication and recombination of the coding sequences for the domains produced the genes for these large proteins. In addition to the domains, each of these proteins also contains a significant fraction of unique sequences.

Most adhesive glycoproteins that interact with cells bind to heterodimeric transmembrane receptors called integrins (see Fig. 30-9). Remarkably, the integrin-binding sites of many adhesive proteins include the simple tripeptide arginine-glycine-aspartic acid (RGD [Fig. 29-15]).

Establishing the biological functions of adhesive glycoproteins is challenging because of overlapping functions and the large size of matrix macromolecules. Initial hypotheses were based on the identification of binding partners and the time and place of expression of each protein. Later, antibodies or peptides were used to disrupt specific molecular interactions in live organisms. Disruption of the gene for each protein or its receptors provides the most definitive data, and the consequences can be surprising. Some phenotypes are milder than expected from earlier studies. These results argue that the adhesive glycoproteins function as a complementary system with partially overlapping functions. Two examples illustrate what we know about adhesive glycoproteins.

Fibronectin

Fibronectins are large proteins that consist of two polypeptides of about 235 kD linked by disulfide bonds near their C-termini (Fig. 29-15). In electron micrographs, fibronectin appears as a V-shaped pair of long, flexible rods connected at one end. In solution, the molecule is probably more compact. Each polypeptide is a linear array of three types of domains called FN-I, FN-II, and FN-III. All three types of fibronectin domains consist of antiparallel b strands with conserved residues in their hydrophobic cores. Two disulfide bonds stabilize FN-I and FN-II domains, whereas FN-III domains have no disulfide bonds. FN-I and FN-II domains consist of about 45 residues; FN-III domains are twice as large. FN-I and FN-II domains are present in a few other proteins, whereas the human genome contains about 170 genes with FN-III domains, including proteins in the extracellular matrix (Appendix 29-3), on the cell surface (human growth hormone receptor; see Fig. 24-6), and inside cells (titin; see Fig. 39-7).

Fibronectin binds a variety of ligands, including cell surface receptors, collagen, proteoglycans, and fibrin (another adhesive protein). Thus, it contributes to adhesion of cells to the extracellular matrix and may also cross-link matrix molecules. Ligand-binding sites in the various domains were identified by isolating proteolytic fragments and by expressing fibronectin fragments (Fig. 29-15). Some of these binding sites are cryptic and are exposed only when the protein is stretched. The RGD sequence that contributes to the integrin-binding site of fibronectin is located on an exposed loop of FN-III domain 10. The variably spliced V domain included in plasma fibronectin has a second integrin-binding site. Chapter 30 provides additional details on integrins.

Two pools of fibronectin have different distributions and solubility properties. Tissue fibronectin forms insoluble fibrils in connective tissues throughout the body, especially in embryos and healing wounds. Fibroblasts use an integrin-dependent process to assemble fibronectin dimers into fibrillar aggregates large enough to visualize by light microscopy (Fig. 29-16). The structure of these microscopic fibrils is not known. Denaturing agents and disulfide reduction are required to solubilize these fibrils. Disulfide bonds between the two subunits of fibronectin are also essential to form this continuous protein network, so fibronectin with deletions from the C-terminus cannot assemble into fibrils. Although difficult to study because of their large size and insolubility, fibronectin fibrils seem to bind cells more efficiently than soluble fibronectin, and may have additional activities important for biological functions.

Figure 29-16 fluorescence micrographs of fibronectin networks in tissue culture. A., This fibroblast expressed fibronectin-YFP (fibronectin fused to yellow fluorescent protein, appearing yellow-green) and moesin-CFP (moesin fused to cyan fluorescent protein, appearing red). The fibronectin assembled an extracellular network. Moesin is associated with actin filaments in stress fibers. B, Lower magnification of a fibronectin network.

(Courtesy of T. Ohashi and H. P. Erickson, Duke University, Durham, North Carolina.)

Soluble plasma fibronectin dimers circulate in the body fluids. The protein differs from tissue fibronectin as a result of alternate splicing of the mRNA. In blood clots, the enzyme transglutaminase covalently couples plasma fibronectin to fibrin, forming a provisional matrix for wound repair (see Fig. 32-11).

Given its ligand-binding and assembly properties, together with its expression in vertebrate embryos even before their implantation in the uterus, fibronectin appears to contribute to the extracellular matrix in early embryos. A collagen-based matrix replaces this primordial matrix as the embryo matures. Furthermore, embryonic cells, such as neural crest cells (precursors of pigment cells, sympathetic neurons, and adrenal medullary cells), migrate along tracks in the extracellular, fibronectin-rich matrix. Antibodies or fibronectin fragments that interfere with the adhesion of cells to fibronectin inhibit neural crest cell migration, gastrulation, and the formation of many embryonic structures derived from mesenchymal cells. Consequently, it was thought that fibronectin might provide cellular adhesion sites required for these movements.

Thus, it was predicted that deletion of the single fibronectin gene in mice would be lethal, owing to devastating effects very early in embryogenesis. It is true that homozygous null mutant mice die during embryogenesis as a result of failure to form mesodermal structures, including the notochord, muscles, heart, and blood vessels. However, the surprise was how far the embryos developed without fibronectin. In fact, up to about day 8 (when the basic body plan is already determined), the embryos appeared to be almost normal. (Mice with null mutations in the main fibronectin receptor, integrin a5, have similar but slightly milder defects.) One interpretation is that fibronectin and its receptor are less important for early development than was anticipated. Changes in the expression of other adhesive glycoproteins or receptors may compensate for the fibronectin defects during the first few days of devel-opment.

Tenascin

Tenascins are a family of giant proteins with six arms (Fig. 29-17), found in the extracellular matrix of many embryonic tissues, wounds, and tumors. The N-terminal ends of three subunits self-associate through a triple helical coiled-coil. Disulfides covalently link two of these three-chain units to make the hexameric molecule. The arms of the four isoforms consist of different numbers of EGF and FN-III domains, terminated by three similar FN-III domains and the fibrinogen-like domains.

Figure 29-17 domain organization of the three isoforms of tenascin. TNFbg, fibrinogen-like domain. A, Tenascin-X. B, Electron micrograph of tenascin-C. C, Tenascin-R. D, Tenascin-C. Each of these tenascin molecules has six identical chains. One is shown in its entirety. Five chains are represented only by two of their n-terminal EGF domains (A and C) or just two of the other five chains (D).

(Courtesy of H. P. Erickson, Duke University, Durham, North Carolina.)

All vertebrates express tenascin, but it has yet to be found in an invertebrate. Vertebrates have maintained the tenascin genes over hundreds of millions of years, and each isoform is expressed selectively in particular embryonic tissues, so it was surprising that mice with a disrupted tenascin-C gene appear normal. The expression of tenascin-R, tenascin-X, and tenascin-Y hardly overlaps with tenascin-C, but one of these isoforms may compensate for the loss of tenascin-C. On the other hand, genetic deficiency of tenascin-X is one cause of Ehlers-Danlos syndrome, a human condition with hyperextensible skin and lax joints, most often caused by mutations in collagen type V gene.

Tenascins bind to integrins, proteoglycans, and immunoglobulin-superfamily receptors on the cell sur-face, but the significance of these interactions is un-clear. Depending on the cell and the experimental situa-tion, tenascin can promote or inhibit adhesion of cells to culture dishes. Further experimentation is required to understand the selective advantage provided by tenascins.

The Basal Lamina

The basal lamina, a thin, planar assembly of extracellular matrix proteins, supports all epithelia, muscle cells, and nerve cells outside the central nervous system (Fig. 29-18). This two-dimensional network of protein polymers forms a continuous rug under epithelia and a sleeve around muscle and nerve cells. In addition, basal laminae are semipermeable filters for macromolecules, a particularly important role that they play in the conversion of blood plasma into urine in the kidney. The genes for basal lamina components are very ancient, having arisen in early metazoans.

Figure 29-18 micrographs of the basal lamina. A. and C, Fluorescence micrographs of tissue sections stained with fluorescent antibodies to type IV collagen, a major component of basal laminae. A, Kidney with basal laminae around the tubules and blood vessels, including those of the glomerulus in the center. C, Skeletal muscle with basal laminae around the muscle cells. B, D, and E, Electron micrographs of thin sections showing basal laminae (colored pink). B, Endothelial cell lining a blood vessel with a platelet in the lumen. D, Neuromuscular junction. E, Unmyelinated nerve with numerous axons (yellow) surrounded by invaginations of Schwann cells (blue).

(A and C, From Odermatt BF, Lang AB, Ruttner JR, et al: Monoclonal antibody to human type IV collagen. Proc Natl Acad Sci U S A 81:7343–7347, 1984. B and E, Courtesy of Don W. Fawcett, Harvard Medical School, Boston, Massachusetts. D, Courtesy of J. Heuser, Washington University, St. Louis, Missouri.)

In electron micrographs of thin sections of tissues prepared by chemical fixation, the basal lamina is a homogenous, finely fibrillar material that is separated from the adjacent cell by a clear gap (Fig. 29-18D). This gap is not present when the tissue is prepared by rapid freezing, so it might be an artifact. This would reconcile biochemical evidence that plasma membrane proteins connect cells directly to the basal lamina. In some tissues, fine type VII collagen fibrils connect the lamina to underlying connective tissue. The basal lamina and associated collagen fibrils form the “basement membrane” that is observed in histologic preparations of epithelia. A basal lamina alone cannot be seen by light microscopy without special labels, such as those used in Figure 29-18A and C.

Although many proteins contribute to the stability of the basal lamina (Fig. 29-19), only the adhesive glycoprotein laminin is essential for the initial assembly of basal laminae during embryogenesis. The C-terminal end of the cross-shaped laminin molecule binds to cell surface receptors (integrins, dystroglycan; see Fig. 39-7). Laminins self-assemble into continuous, two-dimensional networks through noncovalent interactions of their short arms. Mouse embryos that lack dystroglycan or laminin die early in development, owing to failure to make basal laminae.

Figure 29-19 molecular model of the basal lamina. The drawing shows the sizes and shapes of the component molecules and their postulated three-dimensional arrangement in the basal lamina.

(Reproduced from Yurchenco P, Cheng YS, Colognato H: Laminin forms an independent network in basement membranes. J Cell Biol 117:1119–1133, 1992, by copyright permission of The Rockefeller University Press.)

The subsequent addition of other proteins reinforces the laminin network. A two-dimensional network of collagen IV self-assembles through head-to-head interactions of the N-termini of four molecules and tail-to-tail interactions of the C-terminal NC1 domains of two molecules (Fig. 29-2). Mouse embryos that lack collagen IV make nascent basal laminae composed of laminin but eventually die from defects in basal lamina.

Other proteins reinforce the collagen IV and laminin in basal laminae. The rod-shaped protein nidogen cross-links laminin to type IV collagen. Perlecan, a heparan sulfate proteoglycan, provides additional cross-links, as it binds to itself in addition to laminin, nidogen, and collagen IV. These cross-links help to determine the porosity of basal lamina and thus the size of molecules that can filter through it. Fibrillin and an associated protein, fibulin, are also present.

Epithelial and muscle cells secrete laminin and the other components of basal lamina. Two different cells can cooperate to produce a basal lamina between two tissues. For example, epithelial cells make laminin, and mesenchymal cells make nidogen for the same basal lamina.

The interwoven network of protein fibers provides the physical basis for the two main functions of the basal lamina: physical support and selective permeability. The basal lamina is a physical scaffold to anchor epithelial, muscle, and nerve cells. In epithelia, all of the basal cells attach to the underlying basal lamina. The basal lamina, in turn, is anchored to the underlying connective tissue. Thus, force applied to an exposed epithelial surface, such as skin, is transmitted through the basal lamina to the connective tissue. Similarly, all epithelial cells in tubular structures, such as blood vessels and glands, adhere to a cylindrical basal lamina that contributes to the integrity of the tube. In muscle, the basal lamina around each cell transmits the contractile forces between cells and to tendons.

The fibrous network in the basal lamina also acts as a filter for macromolecules and a permeability barrier for cellular migration. In kidney, a basal lamina sandwiched between two epithelial cells filters the blood plasma to initiate the formation of urine. The molecular weight threshold for the filter is about 60 kD, so most serum proteins are retained in the blood, whereas salt and water pass into the excretory tubules. The high charge of basal lamina proteoglycans contributes to filtering by electrostatic repulsion. Basal laminae also confine epithelial cells to their natural compartment. If neoplastic transformation occurs in an epithelium, the basal lamina prevents the spread of the tumor until matrix metalloproteinases (see the next section) break down the basal lamina.

The major basement membrane type IV collagen consists of two a1(IV) chains and one a2(IV) chain. No human mutations in the two major type IV collagen genes have been observed, presumably because they have a dominant lethal phenotype, as observed in Drosophila.

Restricted human tissues express four additional type IV collagens; remarkably, each has been implicated in human disease (Table 29-1). Patients with Alport’s X-linked familial nephritis all have mutations in the a5(IV) collagen gene. More than 200 different point mutations and deletions are known. Other patients with autosomally inherited Alport’s syndrome have mutations in their a3(IV) or a4(IV) collagen genes. These mutations generally interfere with folding of the collagen molecule and disrupt the basement membranes that form the blood filtration barrier in the glomerulus of the kidney, causing progressive kidney failure that is usually fatal in males. These mutations also cause defects in the eye and ear, other places where the a5(IV) collagen gene is expressed. In Goodpasture’s syndrome, the immune system produces autoantibodies to the C-terminal NC1 domain of a3(IV) collagen. The protein sequences that elicit autoantibody production are buried in the NC1 domain, so they may be exposed by bacterial infections or organic solvents, predisposing events in the syndrome. Antibodies bound to basement membranes in the kidney and lung cause inflammation that leads to kidney failure and bleeding in the lungs.

Table 29-1 INHERITED DISEASES OR MUTANT PHENOTYPES OF BASAL LAMINA COMPONENTS

Matrix Metalloproteinases

Many physiological processes depend on the controlled degradation of the extracellular matrix. Examples include tissue remodeling during embryogenesis (e.g., resorption of a tadpole tail), wound healing, involution (massive shrinkage secondary to loss of cells and extracellular matrix) of the uterus after childbirth, shedding of the uterine endometrium during menstruation, and invasion of the uterine wall by the embryonic trophoblast during implantation. Conversely, uncontrolled destruction of extracellular matrix contributes to degenerative diseases, such as emphysema and arthritis. In addition to their roles in remodeling, many of these enzymes cleave and release biologically active fragments from matrix or membrane proteins. Three classes of Zn-depended proteases account for both the physiological and pathological degradation of diverse extracellular matrix and cell surface proteins.

The first class is called matrix metalloproteinases (MMPs). These 24 homologous enzymes share a zinc-protease domain (Fig. 29-20) similar to bacterial thermolysin. Gelatinases have three FN-II domains inserted into the sequence of the catalytic domain. All have an N-terminal signal sequence and are processed through the secretory pathway. Between the signal sequence and catalytic domain, all MMPs have an autoinhibitory propeptide, including a conserved cysteine that binds to the zinc ion in the catalytic site. A C-terminal transmembrane domain anchors several MMPs to the plasma membrane. All other MMPs are secreted. Most MMPs have a C-terminal regulatory domain that influences the substrate specificity of the catalytic domain. After secretion, inactive pro-MMPs bind directly or indirectly to cell surface receptors.

Figure 29-20 matrix metalloproteinase structures. A., Domain organization of MMP isoforms. All have an N-terminal cleaved signal sequence, a propeptide that binds to the active site and inhibits the protease activity, and a catalytic domain. Gelatinases have FN-II domains inserted in the catalytic domain. Matrilysin lacks the C-terminal domain. MMP-14 has a transmembrane segment near its C-terminus. B, Atomic structure of MMP-2 showing the arrangement of the domains and the propeptide occupying the active site.

(PDB file: 1CK7.)

MMP activity is carefully regulated at three levels, normally restricting proteolysis to sites of tissue remodeling or physiological breakdown, such as the involut-ing uterus. First, only particular connective tissue, inflammatory, and epithelial cells are genetically programmed to express MMP genes and to respond to growth factors and cytokines to increase production under appropriate circumstances. Second, autoinhibited MMPs on the cell surface require propeptide cleavage for activation. Proteolytic cleavage and dissociation of the propeptide activate the enzyme. Then the cellu-lar movements deliver the active protease to specific substrates. For example, membrane-anchored MMP-14 directly activates MMP-2, which then degrades basement membrane collagen and other substrates. Third, secreted proteins called tissue inhibitors of metalloproteinases (TIMPs) and the plasma protein a2-macroglobulin bind to the active site of MMPs, keeping their activity in check.

Each MMP is selective for targets in the extracellular matrix. In some cases, proteolysis disrupts the mechanical integrity of the matrix. In others, cleavage of a collagen isoform or other matrix protein releases a fragment that favors or inhibits the formation of blood vessels. For example, the N-terminal domain cleaved from collagen XVIII is an inhibitor of angiogenesis that has been called endostatin. Mice survive null mutations in any one of several MMPs tested, but the loss of an MMP may alter the susceptibility to disease dramatically. Mice without MMP-12 (macrophage elastase) are resistant to cigarette smoke, which causes emphysema in normal mice. Without MMP-12, smoke fails to stimulate elastin degeneration, which weakens lung tissue and mediates inflammation. MMPs contribute to the spread of tumors in mice, but disappointingly small-molecule inhibitors of MMPs have not proven useful for treatment of advanced tumors in humans.

The second class of Zn-dependent proteases consists of about 35 proteases called ADAMs (a disintegrin and metalloproteinase). These enzymes are anchored to the plasma membrane by a single transmembrane sequence. Like other metalloproteinases, they are inhibited by TIMPs. ADAMs cleave and release extracellular domains of cell surface proteins, some of which are important informational molecules (e.g., tumor necrosis factor [TNF]-a; transforming growth factor [TGF]-a). ADAM-17 null mutations are lethal during embryogenesis, owing to a lack of TGF-a or other ligands for EGF receptors. A polymorphism in the ADAM-33 gene is strongly associated with human asthma, although the mechanism is not yet understood.

A third class of Zn-dependent proteases is called ADAMTS, ADAMs with a thrombospondin domain. These secreted proteases cleave specific matrix substrates, such as the cartilage proteoglycan aggrecan. Experiments with mice show that inactivation of the protease domain of ADAMTS5 reduces the development of a common joint disease called osteoarthritis.

ACKNOWLEDGMENTS

Thanks go to Robert Linhardt for his suggestions on revisions to this chapter.

APPENDIX 29-1 Collagen Families

APPENDIX 29-2 Proteoglycans

APPENDIX 29-3 Adhesive Glycoproteins