Chapter 19:  Cell Junctions, Cell Adhesion, and the Extracellular Matrix

Introduction

Specialized cell junctions occur at many points of cell-cell and cell-matrix contact in all tissues, but they are particularly important and plentiful in epithelia. Most of these junctions are too small to be resolved by light microscopy. They can be visualized, however, using either conventional or freeze-fracture electron microscopy, both of which show that the interacting plasma membranes (and often the underlying cytoplasm and the intervening intercellular space as well) are highly specialized in these regions. Cell junctions can be classified into three functional groups: (1) occluding junctions, which can seal cells together in an epithelial cell sheet in a way that prevents even small molecules from leaking from one side of the sheet to the other; (2) anchoring junctions, which mechanically attach cells (and their cytoskeletons) to their neighbors or to the extracellular matrix; and (3) communicating junctions, which mediate the passage of chemical or electrical signals from one interacting cell to its partner.

The major kinds of intercellular junctions within each class are listed in Table 19-1. We shall discuss each of them in turn, except for chemical synapses, which are formed exclusively by nerve cells and are discussed in Chapters 11 and 15.

Tight Junctions Form a Selective Permeability Barrier Across Epithelial Cell Sheets ²

Despite the many structural and biochemical differences among various types of epithelia, all have at least one important function in common: they serve as selective permeability barriers, separating fluids on each side that have different chemical compositions. Tight junctions play two distinct roles in this selective-barrier function, as we shall illustrate by considering the epithelium of the mammalian small intestine, or gut.

The epithelial cells lining the small intestine keep most of the gut contents in the inner cavity (the lumen). At the same time, however, the cells must transport selected nutrients across the cell sheet from the lumen into the extracellular fluid permeating the connective tissue on the other side (see Figure 19-1), from where the nutrients diffuse into small blood vessels. This transcellular transportdepends on two sets of membrane-bound carrier proteins: one is confined to the apical surface of the epithelial cell (the surface facing the lumen) and actively transports selected molecules into the cell from the lumen of the gut; the other, which is confined to the basolateral (basal and lateral) surface, allows the same molecules to leave the cell by facilitated diffusion into the extracellular fluid on the other side. If this directional transport is to be maintained, the apical set of carrier proteins must not be allowed to migrate to the basolateral surface of the cell, and the basolateral set must not be allowed to migrate to the apical surface. Furthermore, the spaces between epithelial cells must be sealed so that the transported molecules cannot diffuse back into the gut lumen through the intercellular space (Figure 19-2).

The tight junctions between the epithelial cells are thought to block both these kinds of diffusion. First, they function as barriers to the diffusion of membrane proteins between apical and basolateral domains of the plasma membrane (see Figure 19-2). This undesirable diffusion of membrane constituents occurs if tight junctions are disrupted, for example, by removing the extracellular Ca²⁺ required for tight-junction integrity. Second, they seal neighboring cells together so that water-soluble molecules cannot leak between the cells: if a low-molecular-weight tracer is added to one side of an epithelial cell sheet, it will usually not pass beyond the tight junction (Figure 19-3). The seal is not absolute or invariable, however. Although all tight junctions are impermeable to macromolecules, their permeability to small molecules varies greatly in different epithelia. Tight junctions in the epithelium lining the small intestine, for example, are 10,000 times more leaky to inorganic ions such as Na⁺ than those in the epithelium lining the urinary bladder. Epithelial cells can transiently alter their tight junctions in order to permit an increased flow of solutes and water through breaches in the junctional barriers. This pathway (called paracellular transport) is especially important in the absorption of amino acids and monosaccharides from the lumen of the intestine (where their concentration is sometimes high enough to drive passive transport in the desired direction).

The molecular structure of tight junctions is still uncertain, but freeze-fracture electron microscopy shows them to be composed of an anastomosing network of strands that completely encircles the apical end of each cell in the epithelial sheet (Figure 19-4A and B). In conventional electron micrographs they are seen as a series of focal connections between the outer leaflets of the two interacting plasma membranes (Figure 19-4C).The ability of tight junctions to restrict the passage of ions through the spaces between cells increases logarithmically with increasing numbers of strands in the network, as if each strand acts as an independent barrier. The strands are thought to be composed of long rows of specific transmembrane proteins in each of the two interacting plasma membranes, which join directly to each other to occlude the intercellular space (Figure 19-5).

Anchoring Junctions Connect the Cytoskeleton of a Cell to Those of Its Neighbors or to the Extracellular Matrix

Anchoring junctions are widely distributed in animal tissues. They enable groups of cells, such as those in an epithelium, to function as robust structural units by connecting the cytoskeletal elements of a cell either to those of another cell or to the extracellular matrix (Figure 19-6). They are most abundant in tissues that are subjected to severe mechanical stress, such as heart muscle and skin epithelium (epidermis). They occur in three structurally and functionally different forms: (1) adherens junctions, (2) desmosomes, and (3) hemidesmosomes. Adherens junctions are connection sites for actin filaments; desmosomes and hemidesmosomes are connection sites for intermediate filaments (see Table 19-1).

Before we discuss the different classes of anchoring junctions, it is worth considering briefly the general principles of their construction. As illustrated in Figure 19-7, these junctions are composed of two classes of proteins: (1) intracellular attachment proteins, which form a distinct plaque on the cytoplasmic face of the plasma membrane and connect the junctional complex to either actin filaments or intermediate filaments; and (2) transmembrane linker proteins, whose cytoplasmic domains bind to one or more intracellular attachment proteins, while their extracellular domains interact either with the extracellular matrix or with the extracellular domains of transmembrane linker proteins on another cell.

Much less is known about septate junctions, which are unique to invertebrates. They are probably best classified as anchoring junctions, for they act as connection sites for actin filaments; but it has been suggested that they can function as permeability barriers in some cases.

Adherens Junctions Connect Bundles of Actin Filaments from Cell to Cell or from Cell to Extracellular Matrix ³

Cell-cell adherens junctions occur in various forms. In many nonepithelial tissues they take the form of small punctate or streaklike attachments that connect actin filaments in the cortical cytoplasm of adjacent cells. In epithelial sheets they often form a continuous adhesion belt (or zonula adherens) around each of the interacting cells in the sheet, located near the apex of each cell just below the tight junction. The adhesion belts in adjacent epithelial cells are directly apposed, and the interacting plasma membranes are held together by transmembrane linker proteins that are members of a large family of Ca²⁺-dependent cell-cell adhesion molecules called cadherins, which we discuss later. At one time an adhesion belt was called a belt desmosome, a misleading name because the adhesion belt is chemically and functionally very different from a real desmosome.

Within each cell a contractile bundle of actin filaments lies adjacent to the adhesion belt, running parallel to the plasma membrane, to which it is attached through a set of intracellular attachment proteins that includes α-, β-,and γ-catenin (discussed later), vinculin, α-actinin,and plakoglobin. The actin bundles in adjacent cells are thus linked, via the cadherins and attachment proteins, into an extensive transcellular network (Figure 19-8). The contraction of this network, which depends on myosin motor proteins, is thought to help mediate a fundamental process in animal morphogenesis - the folding of epithelial cell sheets into tubes and other related structures (Figure 19-9).

Cell-matrix adherens junctions enable cells to get a hold on the extracellular matrix by connecting their actin filaments to the matrix. Cultured fibroblasts migrating on an artificial substratum coated with extracellular matrix molecules, for example, grip the substratum at specialized regions of the plasma membrane called focal contacts, or adhesion plaques, where bundles of actin filaments terminate. Many cells in tissues make analogous focal contacts with the surrounding extracellular matrix. The transmembrane linker proteins that mediate these adhesions and serve as links between the matrix and the actin filament bundles in these plaques are members of a large family of cell-surface matrix receptors called integrins, which we discuss later. The extracellular domain of the integrin at a focal contact binds to a protein component of the extracellular matrix, while its intracellular domain binds indirectly to bundles of actin filaments via a complex of attachment proteins, including talin, α-actinin, and vinculin (Figure 19-10).

Septate junctions are widespread in invertebrate tissues. They share a number of features with adhesion belts, with which they sometimes coexist: (1) they form a continuous band around the apical borders of epithelial cells, (2) they are thought to help hold cells together, and (3) they serve as sites of attachment for actin filaments. They have a highly distinctive morphology, for the interacting plasma membranes are joined by poorly characterized junctional proteins that are arranged in parallel rows with a regular periodicity (Figure 19-11).

Desmosomes Connect Intermediate Filaments from Cell to Cell; Hemidesmosomes Connect Them to the Basal Lamina ⁴

Desmosomes and hemidesmosomes act as rivets to distribute tensile or shearing forces through an epithelium and its underlying connective tissue.

Desmosomes are buttonlike points of intercellular contact that rivet cells together (Figure 19-12A). Inside the cell they serve as anchoring sites for ropelike intermediate filaments, which form a structural framework for the cytoplasm of great tensile strength (Figure 19-12B). Thus, through desmosomes, the intermediate filaments of adjacent cells are connected indirectly to form a continuous network throughout the tissue. The particular type of intermediate filaments attached to the desmosomes depends on the cell type: they are keratin filaments in most epithelial cells, for example, and desmin filaments in heart muscle cells.

The general structure of a desmosome is illustrated in Figure 19-12C. It has a dense cytoplasmic plaque composed of a complex of intracellular attachment proteins responsible for connecting the cytoskeleton to the transmembrane linker proteins, which interact through their extracellular domains to hold the adjacent plasma membranes together. As in adhesion belts, the transmembrane linker proteins belong to the cadherin family of Ca²⁺-dependent cell-cell adhesion molecules. The importance of desmosomes in holding cells together is demonstrated by some forms of the potentially fatal skin disease pemphigus, in which individuals make antibodies against one of their own desmosomal cadherin proteins; these antibodies bind to and disrupt desmosomes between skin epithelial cells (keratinocytes), causing severe blistering as a result of the leakage of body fluids into the loosened epithelium. The antibodies disrupt desmosomes only in skin, suggesting that these desmosomes are biochemically different from those in other tissues.

Hemidesmosomes, or half-desmosomes, resemble desmosomes morphologically but are both functionally and chemically distinct. Instead of joining adjacent epithelial cell membranes, they connect the basal surface of epithelial cells to the underlying basal lamina- a specialized mat of extracellular matrix at the interface between the epithelium and connective tissue. Moreover, whereas the keratin filaments associated with desmosomes make lateral attachments to the desmosomal plaques (see Figure 19-12C), many of those associated with hemidesmosomes have their ends buried in the plaque (Figure 19-13). As in focal contacts, the transmembrane linker proteins in hemidesmosomes belong to the integrin family of extracellular matrix receptors, rather than to the cadherin family of cell-cell adhesion proteins used in desmosomes. The intracellular attachment proteins in hemidesmosomes are also different from those in desmosomes.

Thus, although the terminology for the various anchoring junctions is a muddle, the molecular principles (for vertebrates at least) are simple (Table 19-2). Integrins in the plasma membrane anchor a cell to extracellular matrix molecules; cadherins in the plasma membrane anchor it to cadherins in the membrane of an adjacent cell. In both cases there is an intracellular coupling to cytoskeletal filaments, which can be either actin or intermediate filaments depending on the types of intracellular attachment proteins employed. Moreover, for all these classes of anchoring junctions, the adhesion depends on extracellular divalent cations, although the significance of this dependence is unknown.

Gap Junctions Allow Small Molecules to Pass Directly from Cell to Cell ⁵

Perhaps the most intriguing cell junction of all is the gap junction. It is one of the most widespread, being found in large numbers in most animal tissues and in practically all animal species. It appears in conventional electron micrographs as a patch where the membranes of two adjacent cells are separated by a uniform narrow gap of about 2-4 nm. This gap, however, is spanned by channel-forming protein molecules that allow inorganic ions and other small water-soluble molecules to pass directly from the cytoplasm of one cell to the cytoplasm of the other, thereby coupling the cells both electrically and metabolically. Such cell coupling has important functional implications, many of which are only beginning to be understood.

Cell-cell communication of this type was first demonstrated physiologically in 1958, but it took more than 10 years to show that the physiological coupling correlates with the presence of gap junctions seen in the electron microscope. The initial evidence for cell coupling came from electrophysiological studies of specific pairs of interacting nerve cells in the nerve cord of a crayfish. When a voltage gradient was applied across the junctional membrane through electrodes inserted into each of the two interacting cells, an unexpectedly large current flowed, indicating that inorganic ions (which carry current in living tissues) could pass freely from one cell interior to the other. Later experiments showed that small fluorescent dye molecules injected into one cell can likewise pass readily into adjacent cells without leaking into the extracellular space, provided that the molecules are no bigger than about 1000 daltons (Figure 19-14). This suggests a maximal functional pore size for the connecting channels of about 1.5 nm, implying that coupled cells share their small molecules (such as inorganic ions, sugars, amino acids, nucleotides, and vitamins) but not their macromolecules (proteins, nucleic acids, and polysaccharides).

The evidence that gap junctions mediate electrical and chemical coupling between cells in contact with each other comes from several sources. Gap-junction structures can almost always be found where coupling can be demonstrated by electrical or chemical criteria. Conversely, coupling between vertebrate cells is not found where there are no gap junctions. Moreover, dye and electrical coupling can be blocked by a microinjection of antibodies directed against a major gap-junction protein. More recently, molecular methods have provided direct proof: when a gap-junction protein is reconstituted in synthetic lipid bilayers or when mRNA encoding the protein is injected into either a frog oocyte or a gap-junction-deficient cell line, channels with the properties expected of gap-junction channels can be demonstrated electrophysiologically.

Gap-Junction Connexons Are Composed of Six Subunits ⁶

Gap junctions are constructed from transmembrane proteins that form structures called connexons. When the connexons in the plasma membranes of two cells in contact are aligned, they form a continuous aqueous channel, which connects the two cell interiors (Figure 19-15). The connexons protrude from each cell surface, holding the interacting plasma membranes at a fixed distance from each other - hence the term gap junction, emphasizing the contrast with a tight junction, where the lipid bilayers appear to be in direct contact (compare Figures 19-5 and 19-15). Each connexon is seen as an intramembrane particle in freeze-fracture electron micrographs, and each gap junction can contain a cluster of up to several hundred connexons (19-16).

A connexon is composed of a ring of six identical protein subunits called connexins, each of which contains four putative membrane-spanning α helices. The six subunits are thought to associate to form a connexon with a central aqueous pore that is lined by one transmembrane α helix from each subunit. The six connexins form a larger and more permeable channel than do either the five subunits of the neurotransmitter-gated ion channels or the four subunits (or domains) of the voltage-gated cation channels, which are discussed in Chapter 11 (see Figure 11-33).

Gap junctions in different tissues can have somewhat different properties. The permeability of their individual channels can vary, for example. This is now known to reflect differences in the connexins that form the junctions. In rats, for instance, there are at least 11 distinct connexins, each encoded by a separate gene and each having a distinctive, but sometimes overlapping, tissue distribution. Some cell types express more than one type of connexin, but it is unclear whether different connexin proteins ever assemble into the same connexon. Despite the differences between various connexin proteins, their basic structure and function have been highly conserved in evolution. Thus, in cell culture at least, a cell expressing one type of connexin can often form a functional gap junction with a cell expressing a different connexin, even if the two cells are from different vertebrates.

Most Cells in Early Embryos Are Coupled by Gap Junctions ⁷

In tissues containing electrically excitable cells, coupling via gap junctions serves an obvious function. Electrical coupling between nerve cells, for example, allows action potentials to spread rapidly from cell to cell without the delay that occurs at chemical synapses; this is advantageous where speed and reliability are crucial, as in certain escape responses in fish and insects. Similarly, in higher vertebrates, electrical coupling synchronizes the contractions of heart muscle cells and of smooth muscle cells responsible for the peristaltic movements of the intestine.

It is less obvious why gap junctions occur in tissues that do not contain electrically excitable cells. In principle, the sharing of small metabolites and ions provides a mechanism for coordinating the activities of individual cells in such tissues and for smoothing out random fluctuations from cell to cell. The activities of cells in an epithelial cell sheet, for example, such as the beating of cilia, might be coordinated via gap junctions. More generally, since intracellular mediators such as cyclic AMP and Ca²⁺ can pass through gap junctions, responses of coupled cells to extracellular signaling molecules might be propagated and coordinated in this way.

Cell coupling via gap junctions appears to be important in embryogenesis. In early vertebrate embryos (beginning with the late eight-cell stage in mouse embryos) most cells are electrically coupled to one another. As specific groups of cells in the embryo develop their distinct identities and begin to differentiate, however, they commonly uncouple from surrounding tissue. As the neural plate folds up and pinches off to form the neural tube, for instance (see Figure 19-9), its cells uncouple from the overlying ectoderm. Meanwhile the cells within each group remain coupled with one another and so tend to behave as a cooperative assembly, all following a similar developmental pathway in a coordinated fashion.

It is possible that the coupling of cells in embryos provides a pathway for long-range cell signaling within a developing epithelium. A small molecule, for example, could pass through gap junctions from a region of the tissue where its intracellular concentration is kept high to a region where it is kept low, thereby setting up a smooth concentration gradient. The local concentration could then provide cells with "positional information" to control their differentiation according to their location in the embryo (discussed in Chapter 21). Whether gap junctions actually serve this purpose is not known.

The Permeability of Gap Junctions Is Regulated ⁸

Like conventional ion channels, individual gap-junction channels do not remain continuously open; instead, they flip between open and closed states. Moreover, the permeability of gap junctions is rapidly (within seconds) and reversibly decreased by experimental manipulations that decrease cytosolic pH or increase the cytosolic concentration of free Ca²⁺. These observations indicate that gap-junction channels are dynamic structures that, like conventional ion channels, are gated: they can undergo a reversible conformational change that closes the channel in response to changes in the cell. An attractive model for the type of conformational change that might be involved is shown in Figure 19-17.

The physiological role of pH regulation of gap-junction permeability is unknown. There is one case, however, where the reason for the Ca²⁺ control seems clear. When a cell is damaged, its plasma membrane can become leaky. Ions present at high concentration in the extracellular fluid, such as Ca²⁺ and Na⁺, then move into the cell, and valuable metabolites leak out. If the cell were to remain coupled to its healthy neighbors, these too would suffer a dangerous disturbance of their internal chemistry. But the influx of Ca²⁺ into the sick cell causes its gap-junction channels to close immediately, effectively isolating the cell and preventing damage from spreading in this way.

Figure 19-18 summarizes the various types of junctions formed by vertebrate cells in an epithelium. In the most apical portion of the cell, the relative positions of the junctions are the same in nearly all epithelia: the tight junction occupies the most apical portion of the cell, followed by the adhesion belt and then by a special parallel row of desmosomes; together these form a structure called a junctional complex. Gap junctions and additional desmosomes are less regularly organized.

In Plants, Plasmodesmata Perform Many of the Same Functions as Gap Junctions ⁹

The tissues of a plant are organized on different principles from those of an animal. This is because the plant cells are imprisoned within rigid cell walls,consisting of an extracellular matrix rich in cellulose, as we discuss later. The system of cell walls eliminates the need for anchoring junctions to hold the cells in place, but the need for direct cell-cell communication remains. Thus, in contrast to animal cells, plant cells have only one class of intercellular junctions, plasmodesmata, which, like gap junctions, directly connect the cytoplasms of adjacent cells.

In plants, however, the cell wall between a typical pair of adjacent cells is at least 0.1 microns thick, so a structure very different from a gap junction is required to mediate communication across it. Plasmodesmata (singular, plasmodesma) solve the problem. With a few specialized exceptions, every living cell in a higher plant is connected to its living neighbors by plasmodesmata, which form fine cytoplasmic channels through the intervening cell walls. As shown in Figure 19-19A, the plasma membrane of one cell is continuous with that of its neighbor at each plasmodesma, and the cytoplasms of the two cells are connected by a roughly cylindrical channel with a diameter of 20 to 40 nm. Thus the cells of a plant can be viewed as forming a syncytium in which many cell nuclei share a common cytoplasm. Running through the center of the channel in most plasmodesmata is a narrower cylindrical structure, the desmotubule, which is continuous with elements of the smooth endoplasmic reticulum in each of the connected cells (Figures 19-19Band 19-20). Between the outside of the desmotubule and the inner face of the cylindrical channel formed by plasma membrane is an annulus of cytosol through which small molecules can pass from cell to cell. Plasmodesmata are normally created in all new cell walls as they are assembled during the cytokinesis phase of a cell division; they form around elements of smooth endoplasmic reticulum that become trapped across the developing cell plate (discussed in Chapter 18).

In spite of the radical difference of structure between plasmodesmata and gap junctions, they seem to function in remarkably similar ways. Evidence obtained by injecting tracer molecules of different sizes suggests that plasmodesmata allow the passage of molecules with a molecular weight of less than about 800, which is similar to the molecular-weight cutoff for gap junctions. As with gap junctions, transport through plasmodesmata is regulated. Dye-injection experiments, for example, show that there can be barriers to the movement of even low-molecular-weight molecules between certain cells that are connected by apparently normal plasmodesmata; the mechanisms that restrict communication in these cases are not understood. Conversely, certain plant viruses can enlarge plasmodesmata and use this route to pass from cell to cell, thereby spreading the infection. These viruses produce special proteins that bind to components of the plasmodesmata and dramatically increase the effective pore size of the channel. It is not clear, however, how these proteins work.

Summary

Many cells in tissues are linked to one another and to the extracellular matrix at specialized contact sites called cell junctions. Cell junctions fall into three functional classes: occluding junctions, anchoring junctions, and communicating junctions. Tight junctions are occluding junctions that play a critical part in maintaining the concentration differences of small hydrophilic molecules across epithelial cell sheets by (1) sealing the plasma membranes of adjacent cells together to create a continuous, impermeable, or semipermeable barrier to diffusion across the cell sheet and (2) acting as barriers in the lipid bilayer to restrict the diffusion of membrane transport proteins between the apical and the basolateral domains of the plasma membrane in each epithelial cell.

The main types of anchoring junctions in vertebrate tissues are adherens junctions, desmosomes, and hemidesmosomes. Adherens junctions are connecting sites for bundles of actin filaments, whereas desmosomes and hemidesmosomes are connecting sites for intermediate filaments. Septate junctions also serve as connecting sites for actin filaments, but only in invertebrate tissues. Gap junctions are communicating junctions composed of clusters of channel proteins that allow molecules smaller than about 1000 daltons to pass directly from the inside of one cell to the inside of the other. Cells connected by such junctions share many of their inorganic ions and other small molecules and are therefore chemically and electrically coupled. Gap junctions are important in coordinating the activities of electrically active cells, and they are thought to play a coordinating role in other groups of cells as well. Plasmodesmata are the only intercellular junctions in plants; they function like gap junctions even though their structure is entirely different.

Introduction

To form an anchoring junction, cells must first adhere. A bulky cytoskeletal apparatus must then be assembled around the molecules that directly mediate the adhesion. The result is a well-defined structure - a desmosome, a hemidesmosome, or an adherens or septate junction - that is easily identified in the electron microscope. Indeed, electron microscopy provided the basis for the original classification of cell junctions. In the early stages of development of a cell junction, however, before the cytoskeletal apparatus has assembled, and especially in embryonic tissues, the cells often adhere to one another without clearly displaying these characteristic structures: in the electron microscope one may simply see two plasma membranes separated by a small gap of a definite width. Functional tests may show, nevertheless, that the two cells are sticking to one another, and biochemical analysis can reveal the molecules responsible for the adhesion.

Thus, while cell-cell junctions and cell-cell adhesion might seem to be two names for the same phenomenon, they correspond in practice to two different experimental approaches - one through electron microscopic description, the other through functional tests and biochemistry - and two different emphases - one on mature, adult structure, the other on developmental function. It is only in recent years that these two approaches have begun to converge in a unified view of the molecular basis of cell junctions and cell adhesion. In the previous section we concentrated on the structures of mature cell junctions. In this section we turn to functional and biochemical studies of the cell-cell adhesion mechanisms that have to operate before a full-blown cell-cell anchoring junction can be constructed; later in the chapter we discuss functional and biochemical studies of cell-matrix adhesion mechanisms. We begin with a developmental question: what mechanisms ensure that an embryonic cell will attach to appropriate neighbors at the right time?

There Are Two Basic Ways in Which Animal Cells Assemble into Tissues ¹¹

Many simple tissues, including most epithelia, derive from precursor cells whose progeny are prevented from wandering away by being attached to the extracellular matrix or to other cells or to both (Figure 19-21). But the cells, as they accumulate, do not simply remain passively stuck together as a disorderly pile; instead, as we shall see, the tissue architecture is actively maintained by selective adhesions that the cells make and progressively adjust. Thus, if cells of different embryonic tissues are artificially mingled, they will often spontaneously sort out to restore a more normal arrangement.

Such selective adhesion is even more essential for the development of tissues that have more complex origins involving cell migration, whereby one population of cells invades another and assembles with them, and perhaps with other migrant cells, to form an orderly structure. In vertebrate embryos, for example, cells from the neural crest break away from the epithelial neural tube with which they are initially associated and migrate along specific paths to many other regions. There they assemble with other cells and with one another and differentiate into a variety of tissues, including those of the peripheral nervous system (Figure 19-22). Such a process requires, first, some mechanism for directing the cells to their final destination, such as the secretion of a soluble chemical that attracts migrating cells (by chemotaxis) or the laying down of adhesive molecules in the extracellular matrix or on cell surfaces to guide the migrating cells along the right paths (by pathway guidance). Once a migrating cell reaches its destination, it must recognize and join other cells of the appropriate type in order to assemble into a tissue.

Dissociated Vertebrate Cells Can Reassemble into Organized Tissues Through Selective Cell-Cell Adhesion ¹²

Unlike adult vertebrate tissues, which are difficult to dissociate, embryonic vertebrate tissues are easily dissociated by treatment with low concentrations of a proteolytic enzyme such as trypsin, sometimes combined with the removal of extracellular Ca²⁺ and Mg²⁺ with a divalent-cation chelator (such as EDTA). These reagents disrupt the protein-protein interactions (many of which are divalent-cation-dependent) that hold cells together. Remarkably, such dissociated cells often reassemble in vitro into structures that resemble the original tissue. Such findings suggest that tissue structure is not just a product of history; it is actively maintained and stabilized by the system of affinities that cells have for one another and for the extracellular matrix. Thus, by studying the reassembly of dissociated cells in culture, one can hope to illuminate the role of cell-cell and cell-matrix adhesion in creating and maintaining the organization of tissues in the body.

Experiments on cultured cells from the epidermis (the epithelium of the skin) provide an instructive example. The epidermal cells, known as keratinocytes, adhere tightly to one another and form a multilayered sheet that rests on a basal lamina. The keratinocytes in the basal layer are relatively undifferentiated and proliferate steadily, releasing progeny into the upper layers. There cell division halts and terminal differentiation occurs (see Figure 22-21). Given a suitable substratum, dissociated keratinocytes in culture will likewise proliferate and differentiate. If the concentration of Ca²⁺ in the culture medium is kept abnormally low, however, so that Ca²⁺-dependent cell-cell adhesion systems cannot operate, the keratinocytes grow as a monolayer in which proliferating and differentiating cells are intermingled. If the Ca²⁺ concentration is then raised, the spatial organization of the cells is soon transformed: the monolayer is converted into a multilayered epithelium in which the proliferating cells form the basal layer adherent to the substratum and the differentiating cells are segregated into the upper layers, just as in normal skin. This result suggests that the normal stratified arrangement of keratinocytes, ordered according to their state of differentiation, is maintained by Ca²⁺-dependent cell adhesion mechanisms. One such mechanism involves integrin matrix receptors, which we discuss later; these are absent from differentiated epidermal cells but are present on basal cells, which use the integrins to adhere to the basal lamina. Others involve cadherin cell-cell adhesion molecules, which we discuss below.

A still more striking example of the same phenomenon is seen when dissociated cells from two embryonic vertebrate organs such as liver and retina are mixed together and artificially formed into a pellet: the mixed aggregates gradually sort out according to their organ of origin. Similarly, disaggregated cells are found to adhere more readily to aggregates of their own organ than to aggregates of other organs (Figure 19-23). Evidently there are cell-cell recognition systems that make cells of the same differentiated tissue preferentially adhere to one another; these adhesive preferences are presumably important in stabilizing tissue architecture.

What is the molecular basis of this selective cell-cell adhesion in vertebrates? Two distinct classes of cell-cell adhesion molecules (CAMs) operate in most multicellular animals, one Ca²⁺-dependent and the other Ca²⁺-independent, and it is the Ca²⁺-dependent molecules that seem to be primarily responsible for the tissue-specific cell-cell adhesion seen in early vertebrate embryos. Both classes of adhesion molecules were initially identified by making antibodies against cell-surface molecules and then testing the antibodies for their ability to inhibit cell-cell adhesion in a test tube. Those rare antibodies that inhibit are then used to characterize and isolate the adhesion molecule recognized by the antibodies.

The Cadherins Mediate Ca²⁺-dependent Cell-Cell Adhesion in Vertebrates ¹³

The cadherins are responsible for Ca²⁺-dependent cell-cell adhesion in vertebrate tissues, as mentioned in our account of cell-cell junctions. The first three cadherins that were discovered were named according to the main tissues in which they were found: E-cadherin is present on many types of epithelial cells; N-cadherin on nerve, muscle, and lens cells; and P-cadherin on cells in the placenta and epidermis. All are also found transiently on various other tissues during development. In addition, new types of cadherins are continually being discovered, and at least a dozen are currently known. Virtually all vertebrate cells seem to express one or more cadherins, each encoded by a separate gene, the particular set expressed being characteristic of the cell type. Experiments in vitro and in vivo demonstrate that cadherins are the main adhesion molecules holding cells together in early embryonic tissues. In vitro, the removal of extracellular Ca²⁺ or treatment with anti-cadherin antibodies disrupts the tissue, and if cadherin-mediated adhesion is left intact, antibodies against other adhesion molecules are without effect; in vivo, mutations that inactivate the function of cadherins cause embryos to fall apart early in development.

Most cadherins are single-pass transmembrane glycoproteins composed of about 700-750 amino acid residues. The large extracellular part of the polypeptide chain is usually folded into five domains, each containing about 100 amino acid residues; four of these domains are homologous and contain presumptive Ca²⁺-binding sites (Figure 19-24). In the absence of Ca²⁺, the cadherins undergo a large conformational change and, as a result, are rapidly degraded by proteolytic enzymes. The biological significance of the striking Ca²⁺ dependence of cadherin protein function is unknown.

E-cadherin (also called uvomorulin) is the best-characterized cadherin. We encountered it earlier when we discussed cell junctions, since it is usually concentrated in adhesion belts in mature epithelial cells, where it connects the cortical actin cytoskeletons of the cells it holds together. E-cadherin is also the first cadherin expressed during mammalian development, where it helps to cause compaction, an important morphological change that occurs at the eight-cell stage of mouse embryo development. During compaction the loosely attached cells, called blastomeres,become tightly packed together and joined by intercellular junctions. Antibodies against E-cadherin block blastomere compaction, whereas antibodies that react with various other cell-surface molecules on these cells do not.

It seems likely that cadherins also play crucial roles in later stages of vertebrate development, since their appearance and disappearance correlate with major morphogenetic events in which tissues segregate from one another. As the neural tube forms and pinches off from the overlying ectoderm, for example, the neural tube cells lose E-cadherin and acquire N-cadherin, while the cells in the overlying ectoderm continue to express E-cadherin (Figure 19-25). Moreover, the neural crest cells that form the peripheral nervous system have large amounts of N-cadherin on their surface when they are associated with the neural tube, lose it while they are migrating, and then reexpress it when they aggregate to form a ganglion (see Figure 19-22). Thus three cell groups that originate from one cell-layer exhibit distinct patterns of cadherin expression when separating from one another, suggesting that the switches in cadherin expression are involved in the separation process.

Cadherins Mediate Cell-Cell Adhesion by a Homophilic Mechanism ¹⁴

How do cell-cell adhesion molecules such as the cadherins bind cells together? Three possibilities are illustrated in Figure 19-26: (1) molecules on one cell might bind to other molecules of the same kind on adjacent cells (so-called homophilic binding); (2) molecules on one cell might bind to molecules of a different kind on adjacent cells (so-called heterophilic binding); and (3) cell-surface receptors on adjacent cells might be linked to one another by secreted multivalent linker molecules. All of these mechanisms have been found to operate in animals. Cadherins, however, usually utilize a homophilic mechanism. This has been shown by using a line of cultured fibroblasts called L cells, which do not express cadherins and do not adhere to one another. When L cells are transfected with DNA encoding E-cadherin, the transfected cells now adhere to one another by a Ca²⁺-dependent mechanism and the adhesion is inhibited by anti-E-cadherin antibodies. Since the transfected cells do not bind to untransfected L cells, E-cadherin must bind cells together by the interaction of two E-cadherin molecules on different cells.

If L cells expressing different cadherins are mixed together, they sort out and aggregate separately, indicating that different cadherins preferentially bind to their own type. A similar segregation of cells occurs if L cells expressing different amounts of the same cadherin are mixed together. It seems likely, therefore, that both qualitative and quantitative differences in the expression of cadherins play a crucial part in forming tissues; differences in cadherins probably also explain most of the classical experiments demonstrating organ- and tissue-specific adhesion in a test tube.

Most cadherins, such as E, N, and P-cadherins, function as transmembrane linker proteins that mediate interactions between the actin cytoskeletons of the cells they join together. They are, as we have seen, the adhesion proteins around which cell-cell adherens junctions are constructed. A highly conserved cytoplasmic domain of these cadherins interacts with the actin cortex by means of at least three intracellular attachment proteins called catenins(see Figure 19-24). This interaction is required for cell-cell adhesion: E-cadherin molecules lacking their cytoplasmic domain are unable to hold cells together. Those cadherins that are localized in desmosomes interact with intermediate filaments rather than actin filaments; their cytoplasmic domain is different and binds to a different set of attachment proteins, which in turn bind to intermediate filaments.

Cadherins are not the only proteins that mediate Ca²⁺-dependent cell-cell adhesion: some integrins can also bind cells together through heterophilic interactions with other cell-surface proteins, although most integrins mediate the attachment of cells to the extracellular matrix, as we discuss later. In addition, a family of cell-surface carbohydrate-binding proteins (lectins) called selectins function in a variety of transient cell-cell adhesion interactions in the bloodstream; they enable white blood cells, for example, to bind transiently to endo-thelial cells lining small blood vessels and thereby to migrate out of the blood into tissues at sites of inflammation. Selectins contain a highly conserved lectin domain that, in the presence of Ca²⁺, binds to a specific oligosaccharide on another cell - another example of heterophilic cell-cell adhesion (see Figure 10-42). Since selectins have been discussed in Chapter 10, they will not be considered further here.

We discuss later how cells can regulate the adhesive activity of their integrins. In a similar way it seems likely that some cells, at least, can regulate the adhesive activity of their cadherins, although much less is known about cadherin regulation than integrin regulation. Such regulation may be important for the cellular rearrangements that occur within epithelia when these cell sheets change their shape and organization during animal development.

Ca²⁺-independent Cell-Cell Adhesion Is Mediated Mainly by Members of the Immunoglobulin Superfamily of Proteins ¹⁵

The molecules responsible for Ca²⁺-independent cell-cell adhesion belong mainly to the large and ancient immunoglobulin (Ig) superfamily of proteins, so-called because they contain one or more Ig-like domains that are characteristic of anti-body molecules (discussed in Chapter 23). The best-studied example is the neural cell adhesion molecule (N-CAM), which is expressed by a variety of cell types, including most nerve cells. It is the most prevalent of the Ca²⁺-independent cell-cell adhesion molecules in vertebrates, and, like cadherins, it is thought to bind cells together by a homophilic interaction (between N-CAM molecules on adjacent cells). Some Ig-like cell-cell adhesion proteins, however, use a heterophilic mechanism; some of these, called intercellular adhesion molecules (ICAMs), are expressed on activated endothelial cells, where they bind to integrins on the surface of white blood cells and thereby help to trap these blood cells at sites of inflammation.

There are at least 20 forms of N-CAM. Unlike the cadherins, each of which is encoded by a separate gene, the different N-CAM mRNAs are generated by alternative splicing of an RNA transcript produced from a single gene. The large extracellular part of the polypeptide chain in all forms of N-CAM is folded into five Ig-like domains. Most N-CAMs are single-pass transmembrane proteins with variable-sized intracellular domains, which are thought to be involved in cell signaling or binding to the cytoskeleton. One form does not cross the lipid bilayer and is attached to the plasma membrane by a glycosylphosphatidylinositol (GPI) anchor, while another is secreted and may become incorporated into the extracellular matrix (Figure 19-27). Further variation arises from the glycosylation of N-CAM: some forms carry a large quantity of sialic acid (in the highly unusual form of several chains, each containing hundreds of repeating sialic acid residues), while others carry very much less. By virtue of their negative charge, the long sialic acid chains hinder cell adhesion, thereby modifying the adhesive function of the N-CAM. Indeed, it is possible that N-CAM that is heavily loaded with sialic acid may, in some cases, serve to prevent adhesion rather than cause it. In some neurons, for example, the presence of these polysialic acid chains promotes nerve process outgrowth, presumably by making it easier for the growing tips of the processes to let go of the cells to which they are stuck.

There is substantial evidence that N-CAM and its Ig-like relatives play an important part in vertebrate development. When antibodies against either N-CAM or another Ig-related neural cell-cell adhesion molecule called L1 are injected along the pathway of nerve processes growing from the retina to the brain, they disturb the normal growth pattern of the nerve processes. When used in culture, these antibodies inhibit the tendency of developing nerve cell processes to adhere to one another to form bundles (fascicles). Like N-cadherin, N-CAM is expressed in large amounts on cells of the developing neural tube, but when neural crest cells dissociate from the neural tube and migrate away, they lose N-CAM, only to reexpress it later when they reaggregate to form a neural ganglion (see Figure 19-22). As in the case of cadherins, N-CAM is also expressed transiently during critical stages in the development of many non-neural tissues.

Although cadherins and Ig family members are frequently expressed on the same cells, the adhesions mediated by the cadherins are much stronger, and they almost certainly play the major role in holding cells together, segregating cell collectives into discrete tissues, and maintaining tissue integrity. N-CAM and other members of the Ig family seem to contribute more to the regulation or fine-tuning of these adhesive interactions during development and regeneration. Thus an injection of N-cadherin mRNA into a fertilized frog egg results in the overexpression of N-cadherin in places where it is not normally expressed and leads to a gross disruption of normal tissue architecture. By contrast, the same experiment performed with N-CAM mRNA leads to relatively minor disturbances in development even though N-CAM is overexpressed in many abnormal locations.

The most critical test of the requirement for a protein in a particular biological process is not to overexpress it but instead to inhibit its production by disrupting the gene. While this can now be done in some vertebrates, it is most readily done in genetically tractable invertebrates such as Drosophila and the nematode C. elegans. A number of Ig-like proteins that mediate Ca²⁺-independent cell-cell adhesion have been defined in Drosophila. One of these, fasciclin II, is a close relative of N-CAM: like N-CAM, it has five Ig-like domains and operates by homophilic binding. It is expressed mainly on a subset of nerve cell processes and on some of the glial cells they contact during development. If both copies of the fasciclin II gene are inactivated by mutation, the gross structure of the nervous system is normal. However, at least two of the nerve cell processes that normally express fasciclin II and adhere together now fail to recognize each other and therefore do not form a bundle. This observation is consistent with the view that Ig-like cell-cell adhesion molecules play subtle but important roles in development.

Multiple Types of Cell-Surface Molecules Act in Parallel to Mediate Selective Cell-Cell and Cell-Matrix Adhesion ¹⁶

Morphological, cell biological, and biochemical studies all indicate that even a single cell type utilizes multiple molecular mechanisms in adhering to other cells and to the extracellular matrix. Some of these mechanisms involve organized cell junctions; others do not (Figure 19-28). Just as each cell in a multicellular animal contains an assortment of cell-surface receptors that enables the cell to respond specifically to a complementary assortment of soluble chemical signals such as hormones and growth factors, so each cell in a tissue has a particular combination (and concentration) of cell-surface receptors (cell adhesion molecules) that enables it to bind in its own characteristic way to other cells and to the extracellular matrix. And just as receptors for soluble chemical signals generate intracellular signals that alter the cell's behavior, so too can cell adhesion molecules, although the signaling mechanisms are less well understood for these molecules.

Unlike receptors for soluble chemical signals, which bind their specific ligand with high affinity, the receptors that bind to molecules on cell surfaces or in the extracellular matrix usually do so with relatively low affinity. The latter receptors therefore rely on the enormous increase in binding strength gained through simultaneous binding of multiple receptors to multiple ligands on an opposing cell or in the adjacent matrix. One could call this the "Velcro principle." We have seen, however, that the interaction of the extracellular binding domains of these cell-surface molecules is not enough to ensure cell adhesion: at least in the case of cadherins and, as we shall see, integrins, the adhesion molecules must also attach (via attachment proteins) to the cortical cytoskeleton inside the cell. The cytoskeleton is thought to assist and stabilize the lateral clustering of the adhesion molecules so as to facilitate multipoint binding, and it is also required to enable the adhering cell to exert traction on the adjacent cell or matrix (and vice versa) (Figure 19-29). Thus the mixture of specific types of cell-cell adhesion molecules and matrix receptors present on any two cells, as well as their concentration, cytoskeletal linkages, and distribution on the cell surface, will determine the total affinity with which the two cells bind to each other and to the matrix.

Nonjunctional Contacts May Initiate Tissue-specific Cell-Cell Adhesions That Junctional Contacts Then Orient and Stabilize ¹⁶

Which, if any, of the several types of intercellular junctions discussed earlier in this chapter are involved as cells migrate and recognize one another during the formation of tissues and organs? One way to find out is to use an electron microscope to examine the contacts between adjacent cells when they are migrating over each other in developing embryos or in adult tissues undergoing repair after injury. Such studies show that, with the exception of cells reorganizing within an epithelium, these contacts generally do not involve the formation of organized intercellular junctions. Nevertheless, the interacting plasma membranes often come close together and run parallel, separated by a space of 10-20 nm. As several known transmembrane proteins extend above the plasma membrane by 10-20 nm or more, two cell-surface proteins could readily interact directly with each other across the 10-20-nm gap to mediate the adhesion. This type of nonjunctional contact may be optimal for cell locomotion - close enough to give traction but not tight enough to immobilize the cell.

As anchoring junctions (adherens junctions, desmosomes, hemidesmo-somes, and, in insects, septate junctions) are generally not seen between migrating embryonic cells, the formation of such junctions might be an important mechanism for immobilizing cells within an organized tissue once it has formed. In addition, within epithelia the formation of intercellular junctions is thought to be necessary for mechanical strength and to help polarize and orient the constituent cells. A reasonable hypothesis is that nonjunctional cell-cell adhesion proteins initiate tissue-specific cell-cell adhesions, which are then oriented and stabilized by the assembly of full-blown intercellular junctions. As many of the transmembrane proteins involved can diffuse in the plane of the plasma membrane, they can accumulate at sites of cell-cell (and cell-matrix) contact and therefore be used for junctional as well as nonjunctional adhesions. This has been demonstrated to occur for some integrins and cadherins, which help initiate cell adhesion and then later become integral parts of cell junctions.

As an increasing number of monoclonal antibodies and peptide fragments are characterized, each of which blocks a single type of cell-cell adhesion molecule or matrix receptor - and as the genes that encode these cell-surface proteins become available for manipulation in cells in culture and in experimental animals - it should be possible to inactivate the various types of cell-cell adhesion proteins and matrix receptors individually and in different combinations in order to decipher the rules of recognition and binding used in the morphogenesis of complex tissues.

Summary

Cells dissociated from various tissues of vertebrate embryos preferentially reassociate with cells from the same tissue when they are mixed together. This tissue-specific recognition process in vertebrates is mainly mediated by a family of Ca²⁺-dependent cell-cell adhesion proteins called cadherins, which hold cells together by a homophilic interaction between transmembrane cadherin proteins on adjacent cells. In order to hold cells together, the cadherins must be attached to the cortical cytoskeleton. Most animal cells also have Ca²⁺-independent cell-cell adhesion systems that mainly involve members of the immunoglobulin superfamily, which includes the neural cell adhesion molecule N-CAM. As even a single cell type uses multiple molecular mechanisms in adhering to other cells (and to the extracellular matrix), the specificity of cell-cell adhesion seen in embryonic development must result from the integration of a number of different adhesion systems, some of which are associated with specialized cell junctions while others are not.

Introduction

Tissues are not made up solely of cells. A substantial part of their volume is extracellular space, which is largely filled by an intricate network of macromolecules constituting the extracellular matrix (Figure 19-30). This matrix is composed of a variety of versatile proteins and polysaccharides that are secreted locally and assembled into an organized meshwork in close association with the surface of the cell that produced them. Whereas we discussed cell junctions chiefly in the context of epithelial tissues, our account of extracellular matrix concentrates chiefly on connective tissues (Figure 19-31). In these tissues the matrix is frequently more plentiful than the cells that it surrounds, and it determines the tissue's physical properties. Connective tissues form the architectural framework of the vertebrate body, but the amounts found in different organs vary greatly: from skin and bone, in which they are the major component, to brain and spinal cord, in which they are only minor constituents.

Variations in the relative amounts of the different types of matrix macromolecules and the way they are organized in the extracellular matrix give rise to an amazing diversity of forms, each adapted to the functional requirements of the particular tissue. The matrix can become calcified to form the rock-hard structures of bone or teeth, or it can form the transparent matrix of the cornea, or it can adopt the ropelike organization that gives tendons their enormous tensile strength. At the interface between an epithelium and connective tissue, the matrix forms a basal lamina, a thin but tough mat that plays an important part in controlling cell behavior. We shall focus on the extracellular matrix of vertebrates, but other organisms make many unique and interesting related materials, as in the cell walls of bacteria, the cuticles of worms and insects, the shells of mollusks, and, as we discuss later, the cell walls of plants.

Until recently the vertebrate extracellular matrix was thought to serve mainly as a relatively inert scaffolding to stabilize the physical structure of tissues. But now it is clear that the matrix plays a far more active and complex role in regulating the behavior of the cells that contact it - influencing their development, migration, proliferation, shape, and function. The extracellular matrix has a correspondingly complex molecular composition. Although our understanding of its organization is still fragmentary, there has been rapid progress in characterizing many of its major components.

The Extracellular Matrix Is Made and Oriented by the Cells Within It ¹⁷

The macromolecules that constitute the extracellular matrix are mainly produced locally by cells in the matrix. As we discuss later, these cells also help to pattern the matrix, in that the orientation of their cytoskeleton influences the orientation of the matrix they produce. In most connective tissues the matrix macromolecules are secreted largely by cells called fibroblasts (Figure 19-32). In some specialized connective tissues such as cartilage and bone, however, they are secreted by cells of the fibroblast family that have more specific names: chondroblasts, for example, form cartilage, and osteoblasts form bone. The two main classes of extracellular macromolecules that make up the matrix are (1) polysaccharide chains of the class called glycosaminoglycans (GAGs), which are usually found covalently linked to protein in the form of proteoglycans, and (2) fibrous proteins of two functional types: mainly structural (for example, collagen and elastin) and mainly adhesive (for example, fibronectin and laminin). We shall see (in Figure 19-57) that the members of both classes come in a great variety of shapes and sizes. Glycosaminoglycan and proteoglycan molecules in connective tissue form a highly hydrated, gel-like "ground substance" in which the fibrous proteins are embedded; the polysaccharide gel resists compressive forces on the matrix, and the collagen fibers provide tensile strength. The aqueous phase of the polysaccharide gel permits the rapid diffusion of nutrients, metabolites, and hormones between the blood and the tissue cells; the collagen fibers both strengthen and help to organize the matrix, and rubberlike elastin fibers give it resilience. The adhesive proteins help cells attach to the appropriate part of the extracellular matrix: fibronectin, for example, promotes the attachment of fibroblasts and various other cells to the matrix in connective tissues, while laminin promotes the attachment of epithelial cells to the basal lamina.

Glycosaminoglycan (GAG) Chains Occupy Large Amounts of Space and Form Hydrated Gels ¹⁸

Glycosaminoglycans (GAGs) are unbranched polysaccharide chains composed of repeating disaccharide units. They are called GAGs because one of the two sugar residues in the repeating disaccharide is always an amino sugar (N-acetylglucosamine or N-acetylgalactosamine), which in most cases is sulfated. The second sugar is usually a uronic acid (glucuronic or iduronic). Because there are sulfate or carboxyl groups on most of their sugar residues, GAGs are highly negatively charged (Figure 19-33). Four main groups of GAGs have been distinguished by their sugar residues, the type of linkage between these residues, and the number and location of sulfate groups: (1) hyaluronan, (2) chondroitin sulfate and dermatan sulfate, (3) heparan sulfate and heparin, and (4) keratan sulfate.

Polysaccharide chains are too inflexible to fold up into the compact globular structures that polypeptide chains typically form. Moreover, they are strongly hydrophilic. Thus GAGs tend to adopt highly extended conformations that occupy a huge volume relative to their mass (Figure 19-34), and they form gels even at very low concentrations. Their high density of negative charges attracts a cloud of cations, such as Na⁺, that are osmotically active, causing large amounts of water to be sucked into the matrix. This creates a swelling pressure, or turgor, that enables the matrix to withstand compressive forces (in contrast to collagen fibrils, which resist stretching forces). The cartilage matrix that lines the knee joint, for example, can support pressures of hundreds of atmospheres by this mechanism.

The amount of GAGs in connective tissue is usually less than 10% by weight of the amount of the fibrous proteins. Because they form porous hydrated gels, however, the GAG chains fill most of the extracellular space, providing mechanical support to tissues while still allowing the rapid diffusion of water-soluble molecules and the migration of cells. The importance of GAGs is illustrated by a rare human genetic disease in which there is a severe deficiency in the synthesis of the dermatan sulfate disaccharide shown in Figure 19-33. The affected individuals are dwarves, have a prematurely aged appearance, and have generalized defects in their skin, joints, muscles, and bones.

It should be emphasized, however, that in invertebrates and in plants other types of polysaccharides often dominate the structure of the extracellular matrix. Thus in higher plants, as we discuss later, cellulose (polyglucose) chains are packed tightly together in ribbonlike crystalline arrays to form the microfibrillar component of the cell wall. In insects, crustaceans, and other arthropods, chitin (poly- N-acetylglucosamine) similarly forms the main component of the exoskeleton. Together, cellulose and chitin are the most abundant biopolymers on earth.

Hyaluronan Is Thought to Facilitate Cell Migration During Tissue Morphogenesis and Repair ¹⁹

Hyaluronan (also called hyaluronic acidor hyaluronate) is the simplest of the GAGs. It consists of a regular repeating sequence of up to 25,000 nonsulfated disaccharide units (Figure 19-35). It is found in variable amounts in all tissues and fluids in adult animals and is especially abundant in early embryos. Because of its simplicity, hyaluronan is thought to represent the earliest evolutionary form of GAG, but it is not typical of the majority of GAGs. All of the others (1) contain sulfated sugars, (2) tend to contain a number of different disaccharide units arranged in more complex sequences, (3) have much shorter chains, consisting of fewer than 300 sugar residues, and (4) are covalently linked to protein to form proteoglycans. Moreover, whereas other GAGs are synthesized inside the cell and released by exocytosis, hyaluronan is spun out directly from the cell surface by an enzyme complex that is embedded in the plasma membrane.

Hyaluronan is thought to play a part in resisting compressive forces in tissues and joints. It also has an important role as a space filler during embryonic development, where it can be used to force a change in the shape of a structure. Like styrofoam, it can be quickly and cheaply produced: a small quantity expands with water to occupy a large volume (see Figure 19-34). Hyaluronan synthesized from the basal side of an epithelial sheet, for example, often serves to create a cell-free space into which cells subsequently migrate; this occurs in the formation of the heart, the cornea, and several other organs. When cell migration ends, the excess hyaluronan is generally degraded by the enzyme hyaluronidase. Hyaluronan is also produced in large quantities during wound repair, and it is an important constituent of joint fluid, where it serves as a lubricant.

Many of the functions of hyaluronan depend on specific hyaluronan-binding proteins and proteoglycans, some of which are constituents of the extracellular matrix, while others are integral components of the surface of cells. A number of these molecules (sometimes referred to as hyaladherins) have been shown to have homologous hyaluronan-binding domains containing a characteristic cluster of positively charged amino acid residues.

Proteoglycans Are Composed of GAG Chains Covalently Linked to a Core Protein ²⁰

Except for hyaluronan, all GAGs are found covalently attached to protein in the form of proteoglycans, which are made by most animal cells. As is the case for almost all glycoproteins, the polypeptide chain, or core protein, of a proteoglycan is made on membrane-bound ribosomes and threaded into the lumen of the endoplasmic reticulum. The polysaccharide chains are assembled on the core protein mainly in the Golgi apparatus: first a special link tetrasaccharide is attached to a serine residue on the core protein to serve as a primer for polysaccharide growth; then one sugar residue at a time is added by specific glycosyl transferases (Figure 19-36). While still in the Golgi apparatus, many of the polymerized sugar residues are covalently modified by a sequential and coordinated series of sulfation reactions and epimerization reactions. The epimerizations alter the configuration of the substituents around individual carbon atoms in the sugar molecule; the sulfations greatly increase the negative charge of the proteoglycans.

The proteoglycans are usually easily distinguished from other glycoproteins by the nature, quantity, and arrangement of their sugar side chains: by definition, at least one of the sugar side chains of a proteoglycan must be a GAG. Glycoproteins contain from 1% to 60% carbohydrate by weight in the form of numerous relatively short, branched oligosaccharide chains. The core protein in a proteoglycan is usually a glycoprotein, but it can contain as much as 95% carbohydrate by weight, mostly in the form of long unbranched GAG chains, each typically about 80 sugar residues long. Proteoglycans can thus be much larger than glycoproteins. The aggrecan proteoglycan, for example, which is a major component of cartilage, has a mass of about 3 × 10⁶ daltons; it has over 100 GAG chains, approximately 1 for every 20 amino acid residues. On the other hand, many proteoglycans are much smaller and have only 1 to 10 GAG chains; an example is decorin, which is secreted by fibroblasts and has a single GAG chain (Figure 19-37).

In principle, proteoglycans have the potential for almost limitless heterogeneity. Core proteins range in molecular weight from 10,000 to more than 600,000 daltons and vary greatly in the number and types of their attached GAG chains. Moreover, the underlying repeating pattern of disaccharides in each GAG can be modified by a complex pattern of sulfate groups. The heterogeneity of these GAGs makes it difficult to identify and classify proteoglycans in terms of their sugars. The sequences of many core proteins have been determined with the aid of recombinant DNA techniques, and they too are extremely diverse. Although a few small families have been recognized, there is no common structural feature that clearly distinguishes proteoglycan core proteins from other proteins, and many have one or more domains that are homologous to domains found in other proteins of the extracellular matrix or plasma membrane. Thus it is probably best to regard proteoglycans as a diverse group of highly glycosylated glycoproteins whose functions are mediated by both their core proteins and GAG chains.

Proteoglycans Can Regulate the Activities of Secreted Signaling Molecules ²¹

Given the structural diversity of proteoglycan molecules, it would be surprising if their function in the extracellular matrix were limited to providing hydrated space around and between cells. Their GAG chains, for example, can form gels of varying pore size and charge density, and they could therefore serve as selective sieves to regulate the traffic of molecules and cells according to their size, charge, or both. There is evidence that a heparan sulfate proteoglycan called perlecan has this role in the basal lamina of the kidney glomerulus, which filters molecules passing into the urine from the bloodstream (discussed below).

Proteoglycans are thought to play a major part in chemical signaling between cells. They bind various secreted signaling molecules, such as certain protein growth factors, in a test tube, and it is likely that they do so in tissues. Such binding can enhance or inhibit the activity of the growth factor. Fibroblast growth factor (FGF), for example, which stimulates a variety of cell types to proliferate, binds to heparan sulfate chains of proteoglycans both in vitro and in tissues; for some cells, this binding seems to be a required step for FGF to activate its cell-surface receptor (which is a transmembrane tyrosine kinase, discussed in Chapter 15). Whereas in most cases the signaling molecules bind to the GAG chains of the proteoglycan, this is not always so: the ubiquitous growth regulatory protein transforming growth factor β (TGF-β) binds to the core proteins of several matrix proteoglycans, including decorin; binding to decorin inhibits the activity of the TGF-β.

Proteoglycans also bind and regulate the activities of other types of secreted proteins, such as proteolytic enzymes (proteases) and protease inhibitors. Binding to a proteoglycan could control the activity of a protein in any of the following ways: (1) it could immobilize the protein close to the site where it is produced, thereby restricting its range of action; (2) it could sterically block the activity of the protein; (3) it could provide a reservoir of the protein for delayed release; (4) it could protect the protein from proteolytic degradation, thereby prolonging its action; and (5) it could alter or concentrate the protein for more effective presentation to cell-surface receptors. Proteoglycans are thought to act in all these ways to help regulate the activities of secreted proteins.

GAG Chains May Be Highly Organized in the Extracellular Matrix ^{20, 22}

GAGs and proteoglycans associate to form huge polymeric complexes in the extracellular matrix. Molecules of aggrecan,for example, the major proteoglycan in cartilage, shown in Figure 19-37, assemble with hyaluronan in the extracellular space into aggregates that are as big as a bacterium (Figure 19-38).

Moreover, besides associating with one another, GAGs and proteoglycans associate with fibrous matrix proteins such as collagen and with protein meshworks such as the basal lamina, creating extremely complex structures. The arrangement of proteoglycan molecules in living tissues is generally hard to determine. As they are highly water soluble, they may be washed out of the extracellular matrix when tissue sections are exposed to aqueous solutions during fixation; and changes of pH, ionic, or osmotic conditions can drastically alter their conformation. Thus specialized methods have to be used to visualize them in vivo (Figure 19-39).

Cell-Surface Proteoglycans Act as Co-Receptors

Not all proteoglycans are secreted components of the extracellular matrix. Some, such as serglycin, are constituents of intracellular secretory vesicles, where they help to package and store secretory molecules. Others are integral components of plasma membranes and have their core protein either inserted across the lipid bilayer or attached to the lipid bilayer by a glycosylphosphatidylinositol (GPI) anchor.

Among the best-characterized plasma membrane proteoglycans are the syndecans, which have a membrane-spanning core protein. The extracellular domain of this transmembrane proteoglycan carries a variable number of chondroitin sulfate and heparan sulfate GAG chains, while its intracellular domain is thought to interact with the actin cytoskeleton in the cell cortex. Syndecans are found on the surface of many types of cells, including fibroblasts and epithelial cells, where they serve along with integrins as receptors for collagen, fibronectin, and other matrix proteins to which the syndecans bind. As discussed above, syndecans also bind fibroblast growth factor (FGF) and present it to FGF receptor proteins on the same cell. Similarly, another plasma membrane proteoglycan, called betaglycan, binds transforming growth factor β (TGF-β) and presents it to TGF-β receptors.

Thus plasma membrane proteoglycans act as co-receptors that collaborate with conventional cell-surface receptor proteins, both in binding cells to the extracellular matrix and in initiating the response of cells to some growth factors. The proteoglycans that are discussed in this chapter are summarized in Table 19-3.

Collagens Are the Major Proteins of the Extracellular Matrix ²³

The collagens are a family of highly characteristic fibrous proteins found in all multicellular animals. They are secreted by connective tissue cells, as well as by a variety of other cell types. As a major component of skin and bone, they are the most abundant proteins in mammals, constituting 25% of the total protein mass in these animals. The characteristic feature of a typical collagen molecule is its long, stiff, triple-stranded helical structure, in which three collagen polypeptide chains, called α chains, are wound around one another in a ropelike superhelix. Collagens are extremely rich in proline and glycine, both of which are important in the formation of the triple-stranded helix. Proline, because of its ring structure, stabilizes the helical conformation in each a chain, while glycine is regularly spaced at every third residue throughout the central region of the a chain. Being the smallest amino acid (having only a hydrogen atom as a side chain), glycine allows the three helical α chains to pack tightly together to form the final collagen superhelix (Figure 19-40).

So far, about 25 distinct collagen α chains have been identified, each encoded by a separate gene. Different combinations of these genes are expressed in different tissues. Although in principle more than 10,000 types of triple-stranded collagen molecules could be assembled from various combinations of the 25 or so α chains, only about 15 types of collagen molecules have been found. The main types of collagen found in connective tissues are types I, II, III, V, and XI - type I being the principal collagen of skin and bone and by far the most common. These are the fibrillar collagens and have the ropelike structure we have described for a typical collagen molecule. After being secreted into the extracellular space, these collagen molecules assemble into ordered polymers called collagen fibrils,which are thin (10-300 nm in diameter) structures, many hundreds of micrometers long in mature tissues and clearly visible in electron micrographs (Figure 19-41, and see Figure 19-39). The collagen fibrils often aggregate into larger, cablelike bundles, which can be seen in the light microscope as collagen fibers several micrometers in diameter. Types IX and XII are called fibril-associated collagens as they decorate the surface of collagen fibrils; they are thought to link these fibrils to one another and to other components in the extracellular matrix. Types IV and VII are network-forming collagens: type IV molecules assemble into a feltlike sheet or meshwork that constitutes a major part of mature basal laminae, while type VII molecules form dimers that assemble into specialized structures called anchoring fibrils, which help attach the basal lamina of multilayered epithelia to the underlying connective tissue and therefore are especially abundant in the skin. The collagen types that we discuss are listed in Table 19-4.

Many proteins that contain a repeated pattern of amino acids have evolved by duplications of DNA sequences. The fibrillar collagens apparently arose in this way. Thus the genes that encode the α chains of most of these collagens are very large (up to 44 kilobases in length) and contain about 50 exons. Most of the exons are 54, or multiples of 54, nucleotides long, suggesting that these collagens arose by multiple duplications of a primordial gene containing 54 nucleotides and encoding exactly 6 Gly-X-Y repeats (see Figure 19-40).

Collagens Are Secreted with a Nonhelical Extension at Each End ^{23, 24}

The individual collagen polypeptide chains are synthesized on membrane-bound ribosomes and injected into the lumen of the endoplasmic reticulum (ER) as larger precursors, called pro-α chains. These precursors not only have the short amino-terminal signal peptide required to direct the nascent polypeptide to the ER, they also have additional amino acids, called propeptides, at both their amino- and carboxyl-terminal ends. In the lumen of the ER selected proline and lysine residues are hydroxylated to form hydroxyproline and hydroxylysine, respectively, and some of the hydroxylysine residues are glycosylated. Each pro-α chain then combines with two others to form a hydrogen-bonded, triple-stranded helical molecule known as procollagen. The secreted forms of fibrillar collagens (but not the other types of collagen) are converted to collagen molecules in the extracellular space by the removal of the propeptides (see Figure 19-43).

Hydroxylysineand hydroxyprolineresidues (Figure 19-42) are infrequently found in other animal proteins, although hydroxyproline is abundant in some proteins found in the plant cell wall. In collagen the hydroxyl groups of these amino acids are thought to form interchain hydrogen bonds that help stabilize the triple-stranded helix, and conditions that prevent proline hydroxylation, such as a deficiency of ascorbic acid (vitamin C), have serious consequences. In scurvy, the disease caused by a dietary deficiency of vitamin C that was common in sailors until the last century, the defective pro-α chains that are synthesized fail to form a stable triple helix and are immediately degraded within the cell. Consequently, with the gradual loss of the preexisting normal collagen in the matrix, blood vessels become extremely fragile and teeth become loose in their sockets. This implies that in these particular tissues degradation and replacement of collagen is relatively rapid. In many other adult tissues, however, the turnover of collagen (and other extracellular matrix macromolecules) is thought to be very slow: in bone, to take an extreme example, collagen molecules persist for about 10 years before they are degraded and replaced. By contrast, most cellular proteins have half-lives of hours or days.

After Secretion Fibrillar Procollagen Molecules Are Cleaved to Collagen Molecules, Which Assemble into Fibrils ^{23, 24, 25}

After secretion the propeptides of the fibrillar procollagen molecules are removed by specific proteolytic enzymes outside the cell. This converts the procollagen molecules to collagen molecules, which assemble in the extracellular space to form much larger collagen fibrils. The propeptides have at least two functions: (1) they guide the intracellular formation of the triple-stranded collagen molecules, and (2) because they are removed only after secretion, they prevent the intracellular formation of large collagen fibrils, which could be catastrophic for the cell. The process of fibril formation is driven, in part, by the tendency of the collagen molecules, which are more than 1000-fold less soluble than procollagen molecules, to self-assemble. The fibrils begin to form close to the cell surface, often in deep infoldings of the plasma membrane formed by the tandem fusion of secretory vesicles with the cell surface. The underlying cortical cytoskeleton can therefore influence the sites, rates, and orientation of fibril assembly.

Figure 19-43 summarizes the various steps in the synthesis and assembly of collagen fibrils. Given the large number of enzymatic steps involved in forming a collagen fibril, it is not surprising that there are many human genetic diseases that affect fibril formation. Mutations affecting type I collagen cause osteogenesis imperfecta, characterized by weak bones that easily fracture. Mutations affecting type II collagen cause chondrodysplasias, characterized by abnormal cartilage, which leads to bone and joint deformities. Mutations affecting type III collagen cause Ehlers-Danlos syndrome, characterized by fragile skin and blood vessels and hypermobile joints.

When viewed in an electron microscope, collagen fibrils have characteristic cross-striations every 67 nm, reflecting the regularly staggered packing of the individual collagen molecules in the fibril (Figure 19-44). After the fibrils form in the extracellular space, they are greatly strengthened by the formation of covalent cross-links between lysine residues of the constituent collagen molecules (Figure 19-45). The types of covalent bonds involved are found only in collagen and elastin. If cross-linking is inhibited, the tensile strength of the fibrils is drastically reduced; collagenous tissues become fragile, and structures such as skin, tendons, and blood vessels tend to tear. The extent and type of cross-linking varies from tissue to tissue: collagen is especially highly cross-linked in the Achilles tendon, for example, where tensile strength is crucial.

Fibril-associated Collagens Help Organize the Fibrils ²⁶

In contrast to GAGs, which resist compressive forces, collagen fibrils form structures that resist tensile forces. The fibrils come in a variety of diameters and are organized in different ways in different tissues. In mammalian skin, for example, they are woven in a wickerwork pattern so that they resist tensile stress in multiple directions. In tendons they are organized in parallel bundles aligned along the major axis of tension. And in mature bone and in the cornea they are arranged in orderly plywoodlike layers, with the fibrils in each layer lying parallel to each other but nearly at right angles to the fibrils in the layers on either side. The same arrangement occurs in tadpole skin, which serves to illustrate this organization (Figure 19-46).

The connective tissue cells themselves must determine the size and arrangement of the collagen fibrils. The cells can express one or more of the genes for the different types of fibrillar procollagen molecules. But even fibrils composed of the same mixture of fibrillar collagen molecules have different arrangements in different tissues. How is this achieved? Part of the answer may be that cells can regulate the disposition of the collagen molecules after secretion by guiding collagen fibril formation in close association with the plasma membrane (see Figure 19-43). In addition, as the spatial organization of collagen fibrils at least partly reflects their interactions with other molecules in the matrix, cells can influence this organization by secreting, along with their fibrillar collagens, different kinds and amounts of other matrix macromolecules. The fibril-associated collagens, such as type IX and XII collagen molecules, are thought to be especially important in this regard. They differ from the fibrillar collagens in several ways. (1) Their triple-stranded helical structure is interrupted by one or two short nonhelical domains, which makes the molecules more flexible than fibrillar collagen molecules. (2) They are not cleaved after secretion and so retain their propeptides. (3) They do not aggregate with one another to form fibrils in the extracellular space. Instead, they bind in a periodic manner to the surface of fibrils formed by the fibrillar collagens: type IX molecules bind to type-II-collagen-containing fibrils in cartilage, the cornea, and the vitreous of the eye (Figure 19-47), whereas type XII molecules bind to type-I-collagen-containing fibrils in tendons and various other tissues. The fibril-associated collagens are thought to mediate interactions of collagen fibrils with one another and with other matrix macromolecules. In this way they play a part in determining the organization of the fibrils in the matrix.

Cells Help Organize the Collagen Fibrils They Secrete by Exerting Tension on the Matrix ²⁷

There is yet another way that collagen-secreting cells determine the spatial organization of the collagen matrix they produce. Fibroblasts work on the collagen they have secreted, crawling over it and tugging on it - helping to compact it into sheets and draw it out into cables. This mechanical role of fibroblasts in shaping collagen matrices has been demonstrated dramatically in culture. When fibroblasts are mixed with a meshwork of randomly oriented collagen fibrils that form a gel in a culture dish, the fibroblasts tug on the meshwork, drawing in collagen from their surroundings and causing the gel to contract to a small fraction of its initial volume; by similar activities, a cluster of fibroblasts will surround itself with a capsule of densely packed and circumferentially oriented collagen fibers.

If two small pieces of embryonic tissue containing fibroblasts are placed far apart on a collagen gel, the intervening collagen becomes organized into a compact band of aligned fibers that connect the two explants (Figure 19-48). The fibroblasts subsequently migrate out from the explants along the aligned collagen fibers. Thus the fibroblasts influence the alignment of the collagen fibers, and the collagen fibers in turn affect the distribution of the fibroblasts. Fibroblasts presumably play a similar role in generating long-range order in the extracellular matrix inside the body - in helping to create tendons and ligaments, for example, and the tough, dense layers of connective tissue that ensheathe and bind together most organs.

Elastin Gives Tissues Their Elasticity ²⁸

Many vertebrate tissues, such as skin, blood vessels, and lungs, need to be both strong and elastic in order to function. A network of elastic fibers in the extracellular matrix of these tissues gives them the required resilience so that they can recoil after transient stretch. Elastic fibers are at least five times more extensible than a rubber band of the same cross-sectional area. Long, inelastic collagen fibrils are interwoven with the elastic fibers to limit the extent of stretching and prevent the tissue from tearing.

The main component of elastic fibers is elastin, a highly hydrophobic protein (about 750 amino acid residues long), which, like collagen, is unusually rich in proline and glycine but, unlike collagen, is not glycosylated and contains little hydroxyproline and no hydroxylysine. Elastin molecules are secreted into the extracellular space and assemble into elastic fibers close to the plasma membrane, generally in cell-surface infoldings. After secretion the elastin molecules become highly cross-linked to one another to generate an extensive network of fibers and sheets (Figure 19-49). The cross-links are formed between lysine residues by a mechanism similar to the one that operates in cross-linking collagen molecules.

The elastin protein is composed largely of two types of short segments that alternate along the polypeptide chain - hydrophobic segments, which are responsible for the elastic properties of the molecule, and alanine- and lysine-rich α-helical segments, which form cross-links between adjacent molecules. Each segment is encoded by a separate exon. There is still controversy, however, about the conformation of elastin molecules in elastic fibers and about how the structure of these fibers accounts for their rubberlike properties. In one view the elastin polypeptide chain, like the polymer chains in ordinary rubber, adopts a loose "random coil" conformation, and it is the random-coil structure of the component molecules cross-linked into the elastic fiber network that allows the network to stretch and recoil like a rubber band (Figure 19-50).

Elastic fibers are not composed solely of elastin, however. The elastin core is covered with a sheath of microfibrils, each microfibril having a diameter of about 10 nm. While elastic fibers always contain microfibrils, the same microfibrils can be found in extracellular matrices that do not contain elastin. Microfibrils are composed of a number of distinct glycoproteins, including the large glycoprotein fibrillin, which seems to be essential for the integrity of elastic fibers. Mutations in the fibrillin gene result in Marfan's syndrome, a relatively common human genetic disease that affects connective tissues that are rich in elastic fibers; in the most severely affected individuals, the aorta (whose wall is normally full of elastin - see Figure 19-49) is prone to rupture. Microfibrils are thought to play an important part in the assembly of elastic fibers. They appear before elastin in developing tissues and seem to form a scaffold on which the secreted elastin molecules are deposited. As the elastin is deposited, the microfibrils become displaced to the periphery of the growing fiber.

Fibronectin Is an Extracellular Adhesive Protein That Helps Cells Attach to the Matrix ²⁹

The extracellular matrix contains a number of noncollagen adhesive proteins that typically have multiple domains, each with specific binding sites for other matrix macromolecules and for receptors on the surface of cells. These proteins thus contribute to both organizing the matrix and helping cells attach to it. The first of them to be well characterized was fibronectin, a large glycoprotein found in all vertebrates. Fibronectin is a dimer composed of two very large subunits joined by a pair of disulfide bonds near their carboxyl termini. Each subunit is folded into a series of functionally distinct rodlike domains separated by regions of flexible polypeptide chain (Figure 19-51A and B). The domains in turn consist of smaller modules, each of which is serially repeated and usually encoded by a separate exon, suggesting that the fibronectin gene, like the collagen genes, evolved by multiple exon duplications. The main type of module, called the type III fibronectin repeat, is about 90 amino acid residues in length, and it occurs at least 15 times in each subunit (Figure 19-51C). It is also found in some other matrix proteins, as well as in some plasma membrane and cytoplasmic proteins.

One way to analyze a complex multifunctional protein molecule like fibronectin is to chop it into pieces and determine the function of its individual domains. The protein is treated with low concentrations of a proteolytic enzyme, which cuts the polypeptide chain in the connecting regions between the rodlike domains, leaving the domains themselves intact, so that their binding activity can be tested. In this way it was shown that one domain binds to collagen, another to heparin, another to specific receptors on the surface of various types of cells, and so on (see Figure 19-51). Once a domain with cell-binding activity had been isolated, for example, its amino acid sequence could be determined and synthetic peptides corresponding to different segments of the domain prepared. These peptides were used to localize the main region responsible for cell binding and then to identify a specific tripeptide sequence (Arg-Gly-Asp, or RGD), which is found in one of the type III repeats (see Figure 19-51C), as a central feature of the binding site. Even very short peptides containing this RGD sequence will compete with fibronectin for the binding site on cells and so will inhibit the attachment of the cells to a fibronectin matrix. If these peptides are coupled to a solid surface, they cause cells to adhere to it. The RGD sequence is not confined to fibronectin. It is found in a number of extracellular matrix proteins, and it is recognized by several members of the integrin family of cell-surface matrix receptors that bind these proteins (discussed below). Each receptor, however, specifically recognizes its own small set of matrix molecules, indicating that tight receptor binding requires more than just the RGD sequence.

Multiple Forms of Fibronectin Are Produced by Alternative RNA Splicing ^{29, 30}

There are multiple forms (isoforms) of fibronectin, including one called plasma fibronectin, which is soluble and circulates in the blood and other body fluids, where it is thought to enhance blood clotting, wound healing, and phagocytosis. All of the other forms assemble on the surface of cells and are deposited in the extracellular matrix as highly insoluble fibronectin filaments. In these cell-surface and matrix forms, fibronectin dimers are cross-linked to one another by additional disulfide bonds; unlike fibrillar collagen molecules, which can be made to self-assemble into fibrils in a test tube, fibronectin molecules assemble into filaments only on the surface of certain cells, suggesting that additional proteins are needed for filament formation.

All forms of fibronectin are encoded by a single large gene that is about 50 kilobases long and contains about 50 exons of similar size. Transcription produces a single large RNA molecule that can be alternatively spliced in three regions, depending on the cell type and stage of development. In humans about 20 different messenger RNAs are produced, each encoding a somewhat different fibronectin subunit. Plasma fibronectin, for example, which is secreted mainly by liver cells, lacks two of the type III repeats that are found in cell- and matrix-associated forms of fibronectin. In some cases alternative splicing adds or deletes a cell-type-specific cell binding site: one such site is used by lymphocytes to adhere to fibronectin.

Alternative splicing presumably allows a cell to produce the type of fibronectin that is most suitable for the needs of the tissue. The pattern of fibronectin RNA splicing in the early embryo is different from that seen later in development; but if adult skin is injured, the pattern of fibronectin RNA splicing in the base of the wound switches back to the pattern seen in early development. These observations suggest that the forms of fibronectin produced in the early embryo and in wound healing are especially appropriate for promoting the cell migrations and proliferation required for tissue development and repair.

The crucial importance of fibronectin in animal development has been dramatically demonstrated by gene "knockout" experiments. Mice with both copies of their fibronectin gene inactivated by mutation, for example, die early in embryogenesis. The mutant mice have multiple morphological defects, including abnormalities in the formation of the notochord, somites, heart, blood vessels, neural tube, and extraembryonic membranes.

Glycoproteins in the Matrix Help Define Cell Migration Pathways ³¹

Fibronectin is important not only for cell adhesion to the matrix but also for guiding cell migrations in vertebrate embryos. Large amounts of fibronectin, for example, are found along the pathway followed by migrating prospective mesodermal cells during amphibian gastrulation (discussed in Chapter 21). The migration of these cells can be inhibited by injecting into the developing amphibian embryo various ligands that disrupt the ability of the cells to bind to fibronectin: antibodies against fibronectin, peptides containing the RGD cell-binding tripeptide but lacking the matrix-binding domains of fibronectin, and antibodies against an integrin that serves as a fibronectin receptor on these cells all inhibit the migration. Fibronectin presumably promotes cell migration by helping cells attach to the matrix. The effect must be delicately balanced so that the migrating cells can grip the matrix without becoming immobilized on it.

Many types of adhesive molecules in the matrix are believed to play a part in guiding morphogenetic cell movements, and new ones are continually being discovered. Tenascin, for example, is a large glycoprotein complex of six identical or similar disulfide-linked polypeptide chains, which radiate from a center like the spokes of a wheel (see Figure 19-57). As in fibronectin, each of the polypeptide chains is composed of several types of short amino acid sequences that are repeated many times; a fibronectin type III repeat, for instance, occurs eight or more times in each chain. Each polypeptide chain is folded into a number of functionally distinct domains, one of which binds the cell-surface transmembrane proteoglycan syndecan, while another binds fibronectin. Tenascin has a much more restricted distribution than fibronectin and is most abundant in the extracellular matrix of embryonic tissues. Unlike fibronectin, it can either promote or inhibit cell adhesion, depending on the cell type; the adhesive and anti-adhesive functions are thought to be mediated by different protein domains. There is increasing evidence that anti-adhesive interactions, like adhesive ones, play an important part in guiding cell migration, as we discuss in Chapter 21.

Type IV Collagen Molecules Assemble into a Sheetlike Meshwork to Help Form Basal Laminae ³²

Our discussion of extracellular matrix thus far has focused on the volume-filling material between cells. But in certain places, especially beneath epithelia, extracellular matrix can also be organized as a thin tough sheet - a basal lamina. The construction of basal laminae depends on some specialized types of extracellular matrix molecules, including a specialized variety of collagen, to which we now turn.

Type IV collagen molecules have a more flexible structure than the fibrillar collagens: their triple-stranded helix is interrupted in 26 regions, allowing multiple bends. Like the fibril-associated collagens, they are not cleaved after secretion but retain the terminal regions that hinder side-to-side packing into long fibrils. Instead, they interact via their uncleaved terminal domains to assemble extracellularly into a flexible, sheetlike, multilayered network. Electron microscopic studies of preparations of assembling type IV collagen molecules suggest that these molecules associate by their carboxyl termini to form head-to-head dimers, which then form an extended lattice via amino-terminal associations with three other molecules and the further lateral associations shown in Figure 19-52. Disulfide and other covalent cross-links between the collagen molecules stabilize these associations. The resulting meshwork forms an insoluble scaffolding to which other components of the basal lamina bind via their specific associations with type IV collagen molecules.

Basal Laminae Are Composed Mainly of Type IV Collagen, Heparan Sulfate Proteoglycan, Laminin, and Entactin ³³

Basal laminae are flexible thin (40-120 nm thick) mats of specialized extracellular matrix that underlie all epithelial cell sheets and tubes; they also surround individual muscle cells, fat cells, and Schwann cells (which wrap around peripheral nerve cell axons to form myelin). The basal lamina thus separates these cells and cell sheets from the underlying or surrounding connective tissue. In other locations, such as the kidney glomerulus and lung alveolus, a basal lamina lies between two cell sheets and functions as a highly selective filter (Figure 19-53). Basal laminae serve more than simple structural and filtering roles, however. They are able to determine cell polarity, influence cell metabolism, organize the proteins in adjacent plasma membranes, induce cell differentiation, and serve as specific highways for cell migration.

The basal lamina is largely synthesized by the cells that rest on it (Figure 19-54). As seen in the electron microscope after conventional fixation and staining, most basal laminae consist of two distinct layers: an electron-lucent layer (lamina lucida or rara) adjacent to the basal plasma membrane of the cells that rest on the lamina - typically epithelial cells - and an electron-dense layer (lamina densa) just below. In some cases a third layer containing collagen fibrils (lamina fibroreticularis) connects the basal lamina to the underlying connective tissue. Some cell biologists use the term basement membrane to describe the composite of all three layers, which is usually thick enough to be seen in the light microscope; others use the terms basal lamina and basement membrane interchangeably. In the basal lamina of some multilayered epithelia, such as the stratified squamous epithelium that forms the epidermis of the skin, the lamina densa is tethered to the underlying connective tissue by specialized anchoring fibrils made of type VII collagen molecules. In one type of skin-blistering disease these connections are either absent or destroyed, and the epidermis and its basal lamina become detached from the underlying connective tissue.

Although its precise composition varies from tissue to tissue and even from region to region in the same lamina, most mature basal laminae contain type IV collagen (see Figure 19-52), the large heparan sulfate proteoglycan perlecan, and the glycoproteins laminin and entactin. Laminin is one of the first extracellular matrix proteins synthesized in a developing embryo, and early in development basal laminae contain little or no type IV collagen and consist mainly of a laminin network. Laminin is a large (~850,000 daltons) flexible complex of three very long polypeptide chains arranged in the shape of an asymmetric cross and held together by disulfide bonds (Figure 19-55). Like many other proteins in the extracellular matrix, it consists of a number of functional domains: one binds to type IV collagen, one to heparan sulfate, one to entactin, and two or more to laminin receptor proteins on the surface of cells. Like type IV collagen, laminin molecules can self-assemble in vitro into a feltlike sheet, largely through interactions between the ends of the laminin arms. A single dumbbell-shaped entactin molecule binds tightly to each laminin molecule where the short arms meet the long one; as entactin also binds to type IV collagen, it is thought to act as an additional bridge between the type IV collagen and laminin networks in basal laminae (Figure 19-56).

The shapes and sizes of some of the extracellular matrix molecules discussed in this chapter are compared in Figure 19-57.

Basal Laminae Perform Diverse and Complex Functions ³⁴

In the kidney glomerulus an unusually thick basal lamina acts as a molecular filter, preventing the passage of macromolecules from the blood into the urine as urine is formed (see Figure 19-53). The heparan sulfate proteoglycan seems to be important for this function: when the GAG chains are removed by specific enzymes, the filtering properties of the lamina are destroyed. The basal lamina can also act as a selective barrier to the movement of cells. The lamina beneath an epithelium, for example, usually prevents fibroblasts in the underlying connective tissue from making contact with the epithelial cells. It does not, however, stop macrophages, lymphocytes, or nerve processes from passing through it. The basal lamina plays an important part in tissue regeneration after injury. When tissues such as muscles, nerves, and epithelia are damaged, the basal lamina survives and provides a scaffolding along which regenerating cells can migrate. In this way the original tissue architecture is readily reconstructed. In some cases, as in the skin or cornea, the basal lamina becomes chemically altered following injury - for example, by the addition of fibronectin, which promotes the cell migration required for wound repair.

A particularly striking example of the instructive role of the basal lamina in regeneration comes from studies on the neuromuscular junction, the site where a nerve cell transmits its stimulus to a skeletal muscle cell. At the neuromuscular junction the nerve terminals of a motor neuron form a synapse with a muscle cell. The basal lamina that surrounds the muscle cell separates the nerve and muscle cell plasma membranes at the synapse. The synaptic region of the basal lamina has a distinctive chemical character: special isoforms of type IV collagen and laminin are found there, for example. This junctional basal lamina plays a central role in reconstructing a synapse after nerve or muscle injury. The evidence comes mainly from experiments in frogs. If a frog muscle and its motor nerve are destroyed, the basal lamina around each muscle cell remains and the sites of the old neuromuscular junctions are still recognizable. If the motor nerve but not the muscle is allowed to regenerate, the nerve axons regularly seek out the original synaptic sites on the empty basal laminae and differentiate there to form normal-looking nerve terminals. Thus the junctional basal lamina by itself can guide the regeneration of motor nerve terminals.

Similar experiments show that the basal lamina also controls the localization of the acetylcholine receptors that cluster in the muscle cell plasma membrane at a neuromuscular junction (discussed in Chapter 11). If the muscle and nerve are both destroyed but now the muscle is allowed to regenerate while the nerve is prevented from doing so, the acetylcholine receptors synthesized by the regenerated muscle localize predominantly in the region of the old junctions, even though the nerve is absent (Figure 19-58).

Thus the junctional basal lamina apparently coordinates the local spatial organization of the components in each of the two cells that form a neuromuscular junction. Extracts prepared from junctional basal lamina contain a novel matrix protein called agrin, which, when added to cultured muscle cells, initiates the assembly of synaptic structures in their plasma membrane. Motor neurons have been shown to make agrin, and it is thought that they deposit it (and other specialized macromolecules) in the basal lamina at the developing neuromuscular junction and that these matrix-localized molecules help assemble and stabilize the synaptic connection. Agrin is also made by many other types of neurons, raising the possibility that it directs the assembly of receptors and other postsynaptic macromolecules in synapses throughout the nervous system.

It is likely that basal laminae also play a sophisticated part in guiding cell migrations during embryonic development. Thus, in the nematode worm Caenorhabditis elegans mutation of a gene coding for a lamininlike protein selectively disrupts the paths taken by certain of the mesoderm cells and nerve axons that migrate over the basal lamina underlying the epidermis. Remarkably, only migrations along the dorsoventral axis of the embryo are affected; migrations along the anteroposterior axis of the same basal lamina occur normally. Such findings, and those on the regeneration of the frog neuromuscular junction, suggest that we still have much to learn about the chemistry and functional specializations of basal laminae.

The Degradation of Extracellular Matrix Components Is Tightly Controlled ³⁵

The regulated turnover of extracellular matrix macromolecules is critical to a variety of important biological processes. Rapid degradation occurs, for example, when the uterus involutes following childbirth or when the tadpole tail is resorbed during metamorphosis. A more localized degradation of matrix components is required when cells migrate through a basal lamina, as when white blood cells migrate across the vascular basal lamina into tissues in response to infection or injury, or when cancer cells migrate from their site of origin to distant organs via the bloodstream or lymphatic vessels, a process known as metastasis. Even in the seemingly static extracellular matrix of adult animals there is a slow continual turnover due to degradation and resynthesis.

In each of these cases matrix components are degraded by extracellular proteolytic enzymes that are secreted locally by cells. Most of these proteases belong to one of two general classes: many are metalloproteases, which depend on bound Ca²⁺ or Zn²⁺ for activity, while the others are serine proteases, which have a highly reactive serine residue in their active site. Together, metalloproteases and serine proteases cooperate to degrade matrix proteins such as collagen, laminin, and fibronectin. Some of the metalloproteases, such as the collagenases, are highly specific, cleaving particular proteins at a small number of sites, which are often positioned in such a fashion that the structural integrity of the matrix is destroyed by relatively limited proteolysis; in this way cell migration can be greatly facilitated by a relatively small amount of proteolysis.

An important serine protease involved in matrix degradation is urokinase-type plasminogen activator (U-PA). This acts as the specific trigger in a proteolytic cascade: its immediate target is plasminogen, an inactive serine protease precursor that is abundant in the bloodstream and accumulates at sites of tissue remodeling such as wounds, tumors, and sites of inflammation. U-PA cleaves a single bond in plasminogen to yield the active protease plasmin. In contrast to U-PA, plasmin has a broad specificity, cleaving a variety of proteins, including fibrin (a component of blood clots), fibronectin, and laminin.

Several mechanisms operate to ensure that the degradation of matrix components is tightly controlled. First, like plasminogen, many proteases are secreted as inactive precursors that can be activated locally. Second, the action of proteases is confined to specific areas by various secreted protease inhibitors, such as the tissue inhibitors of metalloproteases (TIMPs) and the serine protease inhibitors known as serpins. These inhibitors are specific for particular proteases and bind tightly to the activated enzyme to block its activity. An attractive idea is that inhibitors are secreted by cells at the margins of areas of active degradation in order to protect uninvolved matrix; they may also protect cell-surface proteins that are required for cell adhesion or migration. Third, many cells have receptors on their surface that bind proteases such as U-PA, thereby confining the enzyme to where it is needed: receptor-bound U-PA is found on nerve growth cones and at the leading edge of migrating white blood cells, for example, where it may serve to clear a pathway for their migration, and it seems to be required for some types of cancer cells to metastasize (Figure 19-59).

Summary

Cells in connective tissues are embedded in an intricate extracellular matrix that not only binds the cells together, but also influences their development, polarity, and behavior. The matrix contains various protein fibers interwoven in a hydrated gel composed of a network of glycosaminoglycan (GAG) chains.

The GAGs are a heterogeneous group of negatively charged polysaccharide chains, which (except for hyaluronan) are covalently linked to protein to form proteoglycan molecules. They occupy a large volume and form hydrated gels in the extracellular space. Proteoglycans are also found on the surface of cells, where they function as co-receptors to help cells bind to the matrix and respond to growth factors.

The fiber-forming proteins can be divided roughly into two functional types: mainly structural (collagens and elastin) and mainly adhesive (such as fibronectin and laminin). The fibrillar collagens (types I, II, III, V, and XI) are ropelike, triple-stranded helical molecules that aggregate into long fibrils in the extracellular space; these in turn can assemble into a variety of highly ordered arrays. Fibril-associated collagen molecules, such as types IX and XII, decorate the surface of collagen fibrils and influence the interactions of the fibrils with one another and with other matrix components. Type IV collagen molecules assemble into a sheetlike meshwork that is a crucial component of all mature basal laminae, which also contain the proteins laminin and entactin, as well as the heparan sulfate proteoglycan perlecan. Elastin molecules form an extensive cross-linked network of fibers and sheets that can stretch and recoil, imparting elasticity to the matrix. Fibronectin and laminin are examples of large, multidomain, adhesive glycoproteins in the matrix; fibronectin is widely distributed in connective tissues, whereas laminin is found mainly in basal laminae. By means of their multiple binding domains, such proteins help organize the extracellular matrix and help cells adhere to it.

Introduction

To understand how the extracellular matrix interacts with cells, one has to identify the cell-surface molecules (matrix receptors) that bind the matrix components as well as the extracellular matrix components themselves. Because of the multiple interactions among matrix macromolecules in the extracellular space, it is largely a matter of semantics where the plasma membrane components end and the extracellular matrix begins. The ultimate link to the cell, however, requires a transmembrane protein that ties the matrix to the cell's cortical cytoskeleton. Although we have seen that some proteoglycans with transmembrane core proteins function as co-receptors for matrix components, the principal receptors on animal cells for binding most extracellular matrix proteins, including collagen, fibronectin, and laminin, are the integrins, a large family of homologous transmembrane linker proteins.

Integrins differ from cell-surface receptors for hormones and for other soluble signaling molecules in that they bind their ligand with relatively low affinity (K_a = 10⁶- 10⁹ liters/mole) and are usually present at about 10- to 100-fold higher concentration on the cell surface. This arrangement makes sense, as binding simultaneously but weakly to large numbers of matrix molecules allows cells to explore their environment without losing all attachment to it. If the binding were too tight, cells would presumably become irreversibly glued to the matrix and be unable to move - a problem that does not arise if attachment depends on multiple weak adhesions. This is an example of the "Velcro principle" mentioned earlier.

Integrins Are Transmembrane Heterodimers ³⁶

Integrins are crucially important receptor proteins because they are the main way that cells both bind to and respond to the extracellular matrix. They are composed of two noncovalently associated transmembrane glycoprotein subunits called α and β, both of which contribute to the binding of the matrix protein (Figure 19-60). Whereas some integrins seem to bind only one matrix macromolecule such as fibronectin or laminin, others bind more than one: an integrin that is present on fibroblasts, for example, binds collagen, fibronectin, and laminin. One subfamily of integrins recognizes the RGD sequence present in these and other matrix proteins, while other integrins recognize various other sequences or domains. Since the same integrin molecule in different cell types can have different ligand-binding activities, it seems that additional cell-type-specific factors can interact with integrins to modulate their binding activity.

The binding of integrins to their ligands depends on extracellular divalent cations (Ca²⁺or Mg²⁺, depending on the integrin), reflecting the presence of three or four divalent-cation-binding domains in the large extracellular part of the a chain. This property can be used to purify integrins: detergent-solubilized plasma membrane proteins are passed over an affinity column that contains an extracellular matrix protein or an RGD-containing peptide, and the bound integrins are then eluted from the column by washing in a divalent-cation-free solution. The type of divalent cation can influence both the affinity and specificity of the binding of an integrin to its ligands, but, as in the case of the cadherins, the physiological significance of this cation regulation is unknown.

Many matrix proteins in vertebrates are recognized by multiple integrins: for example, at least 8 integrins bind fibronectin, and at least 5 bind laminin. About 20 integrin heterodimers, made from 9 types of β subunits and 14 types of α subunits, have been defined, and new ones are still being discovered. This diversity is further increased by the alternative splicing of some integrin RNAs. The β₁ chains, which form dimers with at least 9 distinct α chains, are found on almost all vertebrate cells; α₅β₁, for example, is a fibronectin receptor, and α₆β₁ is a laminin receptor on many types of cells. β₂ chains, by contrast, which form dimers with 3 types of α chains, are expressed exclusively on the surface of white blood cells, and they play an essential role in enabling these cells to fight infection. One of these β₂ integrins (α_Lβ₂) is called LFA1 (for lymphocyte function associated); another (α_Mβ₂) is called Mac1 because it is found mainly on macrophages. The β₂ integrins mainly mediate cell-cell, rather than cell-matrix, interactions by binding to specific ligands on another cell, such as an endothelial cell lining a blood vessel; the ligands, sometimes referred to as counterreceptors, are members of the immunoglobulin superfamily of cell adhesion molecules discussed earlier. The β₂ integrins enable white blood cells, for example, to attach firmly to and cross the endothelial lining of blood vessels at sites of infection. Humans with the genetic disease called leucocyte adhesion deficiency are unable to synthesize the β₂ subunit; as a consequence, their white blood cells lack the entire family of β₂ receptors, and they suffer repeated bacterial infections. β₃ integrins are found on a variety of cells, including blood platelets, and they bind several matrix proteins, including fibrinogen; platelets interact with fibrinogen during blood clotting, and humans with Glanzmann's disease, who are genetically deficient in β₃ integrins, bleed excessively.

Two integrins that share a common β subunit have been described in Drosophila. If both copies of the Drosophila gene encoding this β subunit are mutated, the flies die as embryos; they develop normally until the first muscle contractions begin, at which point the muscles tear away from their extracellular matrix attachment sites.

Integrins Must Interact with the Cytoskeleton in Order to Bind Cells to the Extracellular Matrix ³⁷

Integrins function as transmembrane linkers (or "integrators") mediating the interactions between the cytoskeleton and the extracellular matrix that are required for cells to grip the matrix. Most integrins connect to bundles of actin filaments. (The integrin found in hemidesmosomes -α₆β₄- is an exception in that it connects to intermediate filaments.) Following the binding of a typical integrin to its ligand in the matrix, the cytoplasmic tail of the β chain binds to both talin and α-actinin and thereby initiates the assembly of a complex of intracellular attachment proteins that link the integrin to actin filaments in the cell cortex; this is thought to be how focal contacts form between cells and the extracellular matrix, as discussed earlier. If the cytoplasmic domain of the β chain is deleted using recombinant DNA techniques, the mutant integrins still bind to their ligands but no longer mediate robust cell adhesion or cluster at focal contacts. It seems that integrins must interact with the cytoskeleton in order to bind cells to the matrix, just as cadherins must interact with the cytoskeleton in order to hold cells together. As discussed earlier, a transmembrane attachment to the cytoskeleton appears to be an important general requirement for both cell-matrix and cell-cell adhesions: without such internal anchorage the attachment site is liable to be ripped out of the cell (see Figure 19-29). The cytoskeletal attachments may also help to cluster the integrins together to give a strong aggregate bond (the Velcro principle again).

As we shall discuss next, the interactions that integrins mediate between the extracellular matrix and the cytoskeleton operate in both directions and play an important part in orienting both the cells and the matrix in a tissue.

Integrins Enable the Cytoskeleton and Extracellular Matrix to Communicate Across the Plasma Membrane ^{37, 38}

The matrix can influence the organization of a cell's cytoskeleton. This can be vividly demonstrated with transformed (cancerlike) fibroblasts in culture (discussed in Chapter 24). Transformed cells often make less fibronectin than normal cultured cells and behave differently. They adhere poorly to the substratum, for example, and fail to flatten out or develop the organized intracellular actin filament bundles known as stress fibers. This may contribute to the tendency of cancer cells to break away from the primary tumor and spread to other parts of the body. In some cases the fibronectin deficiency seems to be at least partly responsible for this abnormal morphology: if the cells are grown on a matrix of organized fibronectin filaments, they flatten out and assemble intracellular stress fibers that are aligned with the extracellular fibronectin filaments.

This interaction between the extracellular matrix and the cytoskeleton is reciprocal in that intracellular actin filaments can influence the orientation of secreted fibronectin molecules. Extracellular fibronectin filaments, for example, assemble on or near the surface of cultured fibroblasts in alignment with adjacent intracellular stress fibers (Figure 19-61). If these cells are treated with the drug cytochalasin, which disrupts actin filaments, the fibronectin filaments dissociate from the cell surface (just as they do during mitosis when a cell rounds up). These reciprocal interactions between extracellular fibronectin and intra-cellular actin filaments across the fibroblast plasma membrane are mediated mainly by integrins.

Since the cytoskeletons of cells can exert forces that orient the matrix macromolecules that the cells secrete, and the matrix macromolecules can in turn organize the cytoskeletons of cells that contact them, the extracellular matrix can in principle propagate order from cell to cell (Figure 19-62), creating large-scale oriented structures, as we saw earlier (see p. 984). The integrins serve as adapters in this ordering process, mediating the interactions between cells and the matrix around them.

Cells Can Regulate the Activity of Their Integrins ³⁹

Whereas the matrix-binding integrins of many cells in tissues are constantly in an adhesive-competent state, the integrins on blood cells often have to be activated before they can mediate cell adhesion. Such regulated adhesion presumably allows blood cells to circulate unimpeded until they are activated by an appropriate stimulus; because the integrins do not need to be synthesized de novo, the response can be rapid. Platelets, for example, can be activated by contact with a damaged blood vessel or any one of a number of soluble signaling molecules. The stimulus triggers intracellular signaling pathways, which in turn rapidly and permanently activate a β₃ integrin in the platelet membrane, altering its conformation so that its extracellular domain becomes able to bind the blood-clotting protein fibrinogen with high affinity, thereby promoting platelet aggregation and blood clot formation (Figure 19-63A). Similarly, the weak binding of T lymphocytes, either to their specific antigen on the surface of an antigen-presenting cell or to a virus-infected cell (discussed in Chapter 23), triggers intracellular signaling pathways in the T cells, which leads to the rapid but transient activation of an integrin (LFA1) on the T cells. The activated integrin enables the T lymphocytes to adhere strongly to the target cell, so that they remain in contact long enough to become stimulated; the integrin then returns to an inactive state to allow the T lymphocytes to disengage. The mechanisms by which intracellular signaling events activate the extracellular binding site of an integrin on a blood cell are largely unknown.

Other intracellular events can inactivate integrins. The phosphorylation of a serine residue on the cytoplasmic tail of a β₁ integrin during mitosis in cultured cells, for example, impairs the ability of the integrin to bind fibronectin, which may explain why these cells round up and detach from the substratum during mitosis. Similarly, in some cancer cells the phosphorylation of a tyrosine residue on the cytoplasmic tail of a fibronectin-binding integrin reduces the ability of the integrin to bind to talin, and this is thought to contribute to the relatively poor adhesion of these cells to fibronectin (Figure 19-63B).

Integrins Can Activate Intracellular Signaling Cascades ⁴⁰

Extracellular matrix macromolecules have striking effects on the behavior of cells in culture, influencing their shape, polarity, movement, metabolism, development, and differentiated functions. Many of these effects involve changes in gene expression, and almost all of them are mediated by integrins. We have already discussed how integrins function as transmembrane linkers that connect extracellular matrix molecules to actin filaments in the cell cortex and thereby regulate the shape, orientation, and movement of cells. But there is increasing evidence that the clustering of integrins at the sites of contact with the matrix (or another cell) can also activate several intracellular signaling pathways, including the inositol phospholipid pathway; in addition, several intracellular proteins, including a tyrosine kinase located in focal contacts, become phosphorylated on tyrosine residues. Although the molecular mechanisms are not known, it seems likely that clustered integrins generate intracellular signals by initiating the assembly of a signaling complex at the cytoplasmic face of the plasma membrane, in much the same way that growth factor receptor tyrosine kinases operate (discussed in Chapter 15). Signaling by both integrins and growth factor receptors frequently seems to be required for an optimal cellular response: many cells in culture, for example, will not proliferate in response to growth factors unless the cells are attached via integrins to extracellular matrix molecules. The challenge is to determine how these signaling cascades interact to influence complex cell behaviors such as gene expression and cell proliferation.

The cell adhesion molecules discussed in this chapter are summarized in Table 19-5.

Summary

Integrins are the principal receptors used by animal cells to bind to the extracellular matrix. They are heterodimers that function as transmembrane linkers that mediate bidirectional interactions between the extracellular matrix and the actin cytoskeleton. They also function as signal transducers, activating various intracellular signaling pathways when activated by matrix binding. A cell can regulate the adhesive activity of its integrins by altering either their matrix-binding site or their attachment to actin filaments.

Introduction

The plant cell wall is an elaborate extracellular matrix that encloses each cell in a plant. It was the thick cell walls of cork, visible in a primitive microscope, that in 1663 enabled Robert Hooke to distinguish cells clearly and to name them as such. The walls of neighboring plant cells, cemented together to form the intact plant (Figure 19-64), are generally thicker, stronger, and, most important of all, more rigid than the extracellular matrix produced by animal cells. In evolving relatively rigid walls, which can be up to many micrometers in thickness, early plant cells forfeited the ability to crawl about and adopted a sedentary life-style that has persisted in all present-day plants.

The Composition of the Cell Wall Depends on the Cell Type ⁴¹

Most newly formed cells in a multicellular plant are produced in special regions called meristems, as explained in Chapter 21. These new cells are generally small in comparison to their final size, and to accommodate subsequent cell growth, their walls, called primary cell walls, are thin and only semirigid. Once growth stops and the wall no longer needs to be able to expand, either the primary wall is simply retained or, far more commonly, a rigid, secondary cell wall is produced, either by thickening the primary wall or by depositing new layers with a different composition underneath the old ones. In addition to a structural or "skeletal" role, the cell wall also protects the underlying cell and functions in the transport of fluid within the plant. When plant cells become specialized, they generally produce specially adapted types of walls, according to which the different types of cells in a plant can be recognized and classified.

Although the primary cell walls of higher plants vary greatly in both composition and organization, like all extracellular matrices they are constructed according to a common principle: they derive their tensile strength from long fibers and their resistance to compression from the matrix of protein and polysaccharide in which the fibers are embedded. In the cell walls of higher plants the fibers are generally made from the polysaccharide cellulose, the most abundant organic macromolecule on earth. The rest of the matrix is composed predominantly of two other types of polysaccharide, hemicellulose and pectin, together with structural proteins. All of these molecules are held together by a combination of covalent and noncovalent bonds to form a highly complex structure whose composition depends on the cell type.

The Tensile Strength of the Cell Wall Allows Plant Cells to Develop Turgor Pressure ⁴²

The aqueous extracellular environment of a plant cell consists of the fluid contained in the walls that surround the cell. Although the fluid in the plant cell wall contains more solutes than does the water in the plant's external milieu (for example, soil), it is still hypotonic in comparison to the cell interior. This osmotic imbalance causes the cell to develop a large internal hydrostatic pressure, or turgor pressure, that pushes outward on the cell wall, just as an inner tube pushes outward on a tire. The turgor pressure increases just to the point where the cell is in osmotic equilibrium, with no net influx of water despite the salt imbalance (see Panel 11-1, p. 517). This pressure is vital to plants because it is the main driving force for cell expansion during growth, and it provides much of the mechanical rigidity of living plant tissues. Compare the wilted leaf of a dehydrated plant, for example, with the turgid leaf of a well-watered one. It is the mechanical strength of the cell wall that allows plant cells to sustain this internal pressure.

The Cell Wall Is Built from Cellulose Microfibrils Interwoven with a Network of Polysaccharides and Proteins ⁴³

The tensile strength of the primary cell wall is provided by cellulose. A cellulose molecule consists of a linear chain of at least 500 glucose residues that are covalently linked to one another to form a ribbonlike structure, which is stabilized by hydrogen bonds within the chain. In addition, intermolecular hydrogen bonds between adjacent cellulose molecules cause them to adhere strongly to one another in overlapping parallel arrays, forming a bundle of 60 to 70 cellulose chains, all of which have the same polarity. These highly ordered crystalline aggregates, many micrometers long, are called cellulose microfibrils. Sets of microfibrils are arranged in layers, or lamellae, with each microfibril about 20-40 nm from its neighbors and connected to them by long hemicellulose molecules that are bound by hydrogen bonds to the surface of the microfibrils. The primary cell wall consists of several such lamellae arranged in a plywoodlike network (Figure 19-65).

Hemicelluloses are a heterogeneous group of branched polysaccharides that bind tightly to the surface of each cellulose microfibril as well as to one another and thereby help to cross-link microfibrils into a complex network. Their function is analogous to that of the fibril-associated collagens discussed earlier. There are many classes of hemicelluloses, but they all have a long linear backbone composed of one type of sugar, from which short side chains of other sugars protrude. Both the backbone sugar and the side-chain sugars vary according to the plant species and its stage of development. It is the sugar molecules in the backbone that form hydrogen bonds with cellulose microfibrils.

Coextensive with this network of cellulose microfibrils and hemicelluloses is another cross-linked polysaccharide network based on pectins (see Figure 19-65). These are a heterogeneous group of branched polysaccharides that contain many negatively charged galacturonic acid residues. Because of their negative charge, pectins are highly hydrated and accompanied by a cloud of cations, resembling the glycosaminoglycans of animal cells in the large amount of space they occupy. When Ca²⁺ is added to a solution of pectin molecules, it cross-links them to produce a semirigid gel (it is pectin that is added to fruit juice to make jelly). Certain pectins are particularly abundant in the middle lamella, the specialized central region of the wall that cements together the walls of adjacent cells, and such Ca²⁺ cross-links are thought to help hold cell wall components together. Thus many plant tissues, if treated with a Ca²⁺ chelating agent, dissociate into their constituent cells. Although covalent bonds also play a part in linking the different plant cell-wall components together, very little is known about their nature.

In addition to the two polysaccharide-based networks that are present in all plant primary cell walls, there is a variable contribution from structural proteins. One class of proteins contains high levels of hydroxyproline, like collagen. These proteins are thought to strengthen the wall, and they are produced in greatly increased amounts as a local response to attack by microorganisms. During normal differentiation cells use structural proteins to modify local regions of their walls, as required to create the wide range of functionally specialized secondary walls characteristic of mature cell types (see Panel 1-1, pp. 18-19).

In order for a plant cell to grow or change its shape, the cell wall has to stretch or deform. But because of their crystalline structure, individual cellulose microfibrils are unable to stretch. Thus stretching or deformation of the cell wall must involve either the sliding of microfibrils past one another, or the separation of adjacent microfibrils, or both. As we discuss next, the direction in which the growing cell enlarges depends on the orientation of the strain-resisting cellulose microfibrils in the primary wall, which in turn depends on the orientation of microtubules in the underlying cell cortex at the time the wall was deposited.

Microtubules Orient Cell-Wall Deposition ⁴⁴

The final shape of a growing plant cell, and hence the final form of the plant, is determined by controlled cell expansion. Expansion occurs in response to turgor pressure in a direction that depends on the arrangement of the cellulose microfibrils in the wall. Cells anticipate their future morphology, therefore, by controlling the orientation of microfibrils that they deposit in the wall. Unlike most other matrix macromolecules, which are made in the endoplasmic reticulum and Golgi apparatus and secreted, cellulose, like hyaluronan, is spun out from the surface of the cell by a plasma-membrane-bound enzyme complex (cellulose synthase), which uses sugar nucleotide precursors supplied from the cytosol. As they are being synthesized, the nascent cellulose chains spontaneously assemble into microfibrils that form on the extracellular surface of the plasma membrane - forming a layer, or lamella, in which all the microfibrils have more or less the same alignment (Figure 19-65). Each new lamella forms internally to the previous one, so that the wall consists of concentrically arranged lamellae, with the oldest on the outside. The most recently laid down microfibrils in elongating cells commonly lie perpendicular to the axis of cell elongation (Figure 19-66); although the orientation of the microfibrils in the outer lamellae that were laid down earlier might be different, it is the orientation of these inner lamellae that have a dominant influence on the direction of cell expansion (Figure 19-67).

An important clue to understanding how this orientation is brought about was provided by the discovery that most cytoplasmic microtubules are arranged in the cortex of the plant cell with the same orientation as the cellulose microfibrils that are currently being deposited in that region. These cortical microtubules, forming what is called the cortical array, lie close to the cytoplasmic face of the plasma membrane, held there by poorly characterized proteins (Figure 19-68). The congruent orientation of the cortical array of microtubules (lying just inside the plasma membrane) and cellulose microfibrils (lying just outside) is seen in many types and shapes of plant cells and is present during both primary and secondary cell-wall deposition.

If the entire system of cortical microtubules is disassembled by treating a plant tissue with a microtubule-depolymerizing drug, the consequences for subsequent cellulose deposition are not as straightforward as might be expected. The drug treatment has no effect on the production of new cellulose microfibrils, and in some cases cells can continue to deposit new microfibrils in the preexisting orientation. Any developmental change in the microfibril pattern that would normally occur between successive lamellae, however, is invariably blocked. It seems that a preexisting orientation of microfibrils can be propagated even in the absence of microtubules, but any change in the deposition of cellulose microfibrils requires that intact microtubules be present to determine the new orientation.

These observations are consistent with the following model. The cellulose-synthesizing complexes embedded in the plasma membrane are thought to spin out long cellulose molecules. As the synthesis of cellulose molecules and their self-assembly into microfibrils proceeds, the distal end of each microfibril presumably forms indirect cross-links to the previous layer of wall material. At the growing, proximal end the synthesizing complexes would therefore need to move along the membrane in the direction of synthesis. Since the growing cellulose microfibrils are very stiff, each layer of microfibrils would tend to be spun out from the membrane in the same orientation as the previously laid down layer, with the cellulose synthase complex following along the preexisting tracks of oriented microfibrils outside the cell. Oriented microtubules inside the cell, however, can change this predetermined direction in which the synthase complexes move: they can create boundaries in the plasma membrane that act like the banks of a canal to constrain movement of the synthase complexes to a parallel axis (Figure 19-69). In this view, cellulose synthesis can occur independently of microtubules but is constrained spatially when cortical microtubules are present to define membrane domains within which the enzyme complex can move.

Plant cells change their direction of elongation, and thus their future plane of cell growth and division, by a sudden change in the orientation of their entire cortical array of microtubules. Inasmuch as plant cells cannot move (being constrained by their walls), the entire morphology of a multicellular plant depends on the coordinated, highly patterned control of these cortical microtubule orientations during plant development. It is not known how the organization of these microtubules is controlled, although it has been shown that they can rapidly reorient in response to extracellular stimuli, including low-molecular-weight plant growth factors such as ethylene and gibberellic acid.

Summary

Plant cells are surrounded by a rigid extracellular matrix in the form of a cell wall, which is responsible for many of the unique features of a plant's life-style. The cell wall is composed of tough cellulose microfibrils embedded in a highly cross-linked matrix of polysaccharides (mainly pectins and hemicellulose) and glycoproteins. A cortical array of microtubules can determine the orientation of newly deposited cellulose microfibrils, which in turn determines the manner in which the cell expands and therefore the cell's final shape and cell-division patterns.