Integrins and morphogenesis

The process of morphogenesis involves a variety of interactions between cells and their environment. The individual cells within the embryo adhere to each other to form distinct tissues, and these tissues adhere to each other to form the organism. Whilst cellular behavior is regulated at many levels, it is clear that cell surface proteins must play an essential role in morphogenesis. Three major classes of proteins involved in embryonic cell interactions have been characterised. Members of two of the classes, the immunoglobulin-like and cadherin families, function as monomers and generally bind to another molecule of the same protein on adjacent cells (homotypic adhesion). The integrins, which compose the third large family of adhesion molecules, possess several unusual features (Hynes, 1992). They are composed of structurally distinct a and β subunits and bind in a divalent cation dependent fashion to a variety of heterotypic ligands, including extracellular matrix molecules and other cell surface proteins. Individual β subunits form heterodimers with different a subunits to generate distinct receptors with unique specificities. To a lesser extent the reverse is also true, where individual a subunits associate with multiple β subunits. Integrins appear to be the link between the cytoskeleton and extracellular matrix proteins, in particular connecting actin associated proteins found at focal contacts to the matrix. At least two lines of evidence confirm that integrins play an essential role in morphogenesis. Antibodies against integrins block gastrulation and neural crest migration when injected into amphibian and avian embryos respectively (Bronner-Fraser, 1985; Darribere et al., 1988) Secondly, as discussed in more detail below, embryos that are mutant for the Drosophila Position-Specific (PS) integrins have morphogenetic defects due to the failure of adhesion between different cell layers.

The Drosophila position-specific (PS) integrins were initially discovered as cell surface antigens that have a restricted distribution in the imaginai discs (the sacs of cells present in the larva that give rise to much of the adult epidermis; Wilcox et al., 1981; Brower et al., 1984). Their name refers to the fact that their expression is not determined by cell type but rather by the position of the cell within the imaginai disc. For example in the late third instar imaginai disc the PS1 integrin (αPS1βPS) is expressed in the cells that give rise to the dorsal surface of the wing while the PS2 integrin (αPS2βPS) is expressed in the complementary cells which make the ventral wing cells. All three PS integrins subunits have now been cloned and sequenced (Bogaert et al., 1987: MacKrell et al., 1988; Werhli et al., 1993). They share many structural features that are common to integrins. They cross the membrane once and are predominantly extracellular with correspondingly short cytoplasmic tails (see Fig. 1). The βPS subunit is very cysteine rich, with all 56 of its cysteines absolutely conserved with those of βl, the most probable orthologue of β PS since it is the vertebrate β that has the highest level of sequence similarity (46% identity) with βPS (Yee and Hynes, 1993). The α subunits both have 7 repeals, (he last three (αPS1) or four (αPS2) of which contain a core with residues that arc thought to bind to divalent cations. Each of the α subunits is cleaved into two fragments that arc linked by disulfide bonds: a completely extracellular heavy chain and a transmembrane light chain. So far there are no vertebrate orthologues of the Drosophila a subunits; αPS1 is approximately equally similar to α3, α6 and α7 (30% identity), and αPS2 is approximately equally similar to α5, α8, αIIb and α_v (30% identity). The αPS2 subunit sequence is modulated by developmentally regulated alternative splicing; one of the 12 exons, exon 8. is a cassette exon, which is regulated to add 25 amino acids to the protein (Brown et al., 1989). This modulation occurs adjacent to the putative ligand binding site and so may alter the PS2 integrin’s affinity or specificity for its presently uncharacleriscd ligand(s). The βps subunit is also alternatively spliced, this time by the alternate choice from two mutually exclusive exons which encode a portion of the protein that is also adjacent to the putative ligand binding site (Hynes, 1992). No alternative splicing of αps1 has been observed (Werhli et al., 1993). but it has not yet been explored exhaustively.

The PS integrins have complex and dynamic patterns of expression (Wilcox et al., 1981; Brower et al., 1984, 1985; Bogaert et al., 1987; Leptin et al., 1989; Zusman et al., 1990; Werhli et al., 1993). At cellular blastoderm αps1 and αps2 transcripts are first detected in mutually exclusive patterns. The mRNA for αps2 is expressed in the presumptive mesoderm while that for αPS is in the ectoderm and endoderm. The proteins arc delectable approximately I hour later and arc concentrated on the basal sides of these cell layers as the mesoderm spreads over the epidermis. At later embryonic stages a complex pattern emerges with high levels of the integrins detected al muscle attachments and al the interface between the visceral mesoderm and the endoderm (see Figs 2 and 3). Thus the two integrins are found on complementary sides of the sites of adhesion between the mesodermal layers and others cell layers. This “refinement” of the integrin pattern of expression as embryogenesis proceeds, for example as seen in the increased level of the PS1 and PS2 integrins visible al the muscle attachment sites (the stripes of βps expression seen in the epidermis of the embryos in Fig. 2). may arise by at least two mechanisms. The increase in the level of PS1 in the one cell wide row of epidermal cells al the segment border, relative to other cells in the epidermis, must arise through a relative increase in the expression of PS1 in these cells (see particularly the dorsal view in Fig. 2A). However, the PS2 integrin may become concentrated at the muscle attachments by lateral mobility within the membrane of the multi nucleate muscles to become “capped” at the muscle termini.

Due to the fact that integrins arc heterodimers. the amount of the integrin on the cell surface could be regulated by regulating just one of the two subunits, since the surface expression of an α subunit has been found to require the expression of a β, and visa versa (e.g. Cheresh and Spiro, 1987; Leptin et al., 1989). The βps subunit mRNA is expressed uniformly in the early embryo (Zusman et al., 1990), and may continue to be so. However, the level of βPS protein in the late stage embryo clearly varies from cell to cell, as shown in Figs 2 and 3. The more complex pattern of βPS integrin protein expression observed at this stage could arise either through the direct transcriptional regulation of the βPS subunit mRNA, or just by regulating the expression of the α subunits, assuming that βPS protein is only stable in the presence of the α subunits. Consistent with the latter model, the αPS2 gene has a 30 kb primary transcript with essential regulatory sequences within the large introns (N.H.B., A. Dokadis and F. C. Kafatos unpublished results), and the αps1 gene is even larger, while the βPS primary transcript is only 8.5 kb. To resolve this question, the expression of the mRNAs for each of the subunits will have to be examined at this late stage.

The tissue-specific complementary expression of αps1 and αps2 in the early embryo is not retained in the developing imaginai discs. Although αps2 may continue to be expressed in the mesoderm, in the larva it is also expressed in the epidermal cells that make up the imaginai disc epithelia, along with αps1. The expression pattern of the two integrins are dynamic but each show’ restricted expression in specific areas of a given disc (Wilcox et al., 1981; Brower et al., 1984; Brower et al., 1985). One of the most interesting patterns is found in the late third instar wing disc, where the cells that will give rise to the dorsal and ventral halves of the wing blade have complementary patterns of PS integrin expression. At metamorphosis the disc everts and the dorsal and ventral surfaces of the wing come into contact. The two layers of the wing adhere to each other, arc separated by expansion of the wing and then adhere once again. The integrins are found al the sites of adhesion between the two surfaces (Fristrom et al., 1993) and. as discussed below, are required to hold the two layers together. Thus the wing in the developing adult is analogous to the embryonic muscle attachments in that two distinct cell layers show complementary expression of the two PS integrins.

Analysis of mutations in the genes encoding the PS integrin subunits has shown that the integrins arc required at the sites where the different cell layers attach to each other and integrin expression is high. All three genes are on the X chromosome. The βps subunit is encoded by the myospheroid locus (Leptin et al., 1989; Bunch et al., 1992), the αps1 has not yet been assigned to a complementation group and the αps2 subunit is encoded by the inflated locus (Brower and Jaffe, 1989; Wilcox et al., 1989; and N.H.B. unpublished data). The phenotype of the myospheroid locus has been extensively characterised by Wright (Wright. 1960). Loss of function myospheroid alleles arc embryonic lethal due lo three major defects. (1) Although the muscles initially attach to the epidermis and endoderm normally, once the muscles begin lo contract the attachments fail, causing the somatic muscles to detach (see Figs 3 and 4) and the midgut elongation lo fail. (2) The central nervous system does not fully condense. (3) The dorsal edges of the epidermis meet each other normally al dorsal closure, but shortly afterward the adhesion al the dorsal midline fails causing a dorsal hole in the epidermis. The first of these phenotypes is clearly correlated with the relatively high concentration of the PS integrins at the interfaces between the muscles and the other cells layers. The failure of the adhesion between the different cell layers illuminates one of the major functions of the PS integrins in morphogenesis: lo keep these different cell layers attached to each other to hold together the organism. In contrast, the other characteristics of the myospheroid phenotype do not correlate with high levels of βps; the levels of protein at the dorsal midline and within the nervous system are just detectable above background. Either not very much PS integrin protein is required to mediate these events or these phenotypes may be secondary consequences of the defect in mesodermal attachment. Null alleles at the inflated locus, which encodes the αps2 subunit, are also embryonic lethal, and show a subset of the defects found in myospheroid mutant embryos (N. H. B. unpublished data). The muscle phenotypes arc similar, although the defects appear later in development, but the epidermis is completely normal. Thus the phenotypes of mutants in two different loci that encode PS integrin subunits are satisfactorily similar, albeit distinct.

The role of the PS integrins in the adult has been examined by making mitotic clones homozygous for embryonic lethal alleles and by examining viable mutations at these loci (Brower and Jaffe, 1989; Wilcox et al., 1989; Zusman et al., 1990). Loss of PS integrin function in the wing causes the separation of the two surfaces of the wing. Some viable alleles of the inflated and myospheroid loci produce small blisters in the wings, while other “stronger” viable inflated mutations result in a ballon shaped wing due to a complete absence of adhesion between the two surfaces (N. II. B. unpublished results). Clones of cells homozygous for myospheroid lethal mutations, fail to attach to the opposite cell layer, generating blisters that extend beyond the boundaries of the clone. As might be predicted, clones mutant for the inflated locus otdy cause blisters when ventral cells arc mutant. Dorsal cells, which do not normally express the PS2 integrin in the late third instar disc, can be mutant without effect (Bahrain and Brower. 1993).

As in the embryo, in the development of the adult there is not a complete correlation between the expression of the PS integrins and the phenotypes resulting from the absence of PS integrin subunits. However the reverse situation is found in the adult since phenotypes cannot be identified that correspond to some of the expression patterns. For example, in the eye imaginai disc. PS1 is found ahead of the morphogenetic furrow and PS2 is expressed behind it. While the PS inlegins arc required for the adhesion of the pigment cells of the retina to the basal fenestrated membrane (Zusman et al., 1990). this requirement does not have a simple relationship lo the observed pattern of expression. The PS integrins arc expressed in the other imaginai discs, and may well be expressed in other adult tissues. It seems likely that the integrins also play a role in the attachment of the adult muscles but this has not yet been examined.

While at the present time the ligands of the PS integrin are unknown, several lines of evidence suggest that they may be extracellular matrix proteins. The majority of vertebrate integrins bind to extracellular matrix ligands such as fibronectin, laminin and collagen. Cells that express the PS2 integrin bind and spread out on the vertebrate extracellular matrix proteins fibronectin and vitronectin (Hirano et al., 1991; Bunch et al., 1992). One of the defects in embryos mutant for the βps subunit is that there is a delay in the appearance of basement membranes (Wright, 1960). This is consistent with other experiments which show that the integrins that bind to extracellular matrix proteins aid in the organisation of these proteins into the observed fibrillar meshwork (Akiyama et al., 1989; Darribere et al., 1990) and would suggest that the PS integrins have a similar role in organising the basement membrane by binding to extracellular matrix proteins.

The sites of integrin adhesion, muscle attachment sites and the sites of adhesion between dorsal and ventral surfaces of the developing wing, look very similar in electron microscope thin sections (see Brown, 1993). The basal membranes are highly interdigitated to increase the surface area of the contact between the cell layers. Multiple electron dense junctions are observed linking the basal surfaces of the cells. It seems likely that the PS integrins are in these electron dense junctions, but it has not yet been shown directly. A thin layer of electron dense material is visible between the two membranes which may be composed of extracellular matrix proteins. However, the distance separating the two membranes, 300-500Å, is close enough to permit direct contact of the integrins, which extend 200-210Â from the membrane (Nermut et al., 1988). A very similar situation is found at the myotendinous junction of vertebrates, where the muscle membrane interdigitates with the collagen rich tendon (Tidball et al., 1986), and integrins are concentrated at this site (Bozyczko et al., 1989). Thus the integrins are recruited to these specialised adhesion sites between different cell layers that are subject to the strong forces produced by muscle contraction and by pumping in of hemolymph (which results in the normal expansion of the the wing after metamorphosis).

The evidence currently points to the PS integrins being receptors for extracellular matrix proteins. However, the analysis of the mutant phenotypes shows that the integrins are essential for the adhesion of different cell layers to each other: muscle to epidermis and endoderm and dorsal to ventral wing blade. So why should the adhesion of these different cell layers to each other be through the extracellular matrix rather than by direct interaction? One advantage of adhesion through the matrix is that cells can stick tightly to each other, while at the same time remaining separated by the physical barrier that is provided by the crosslinked meshwork of proteins that compose the matrix. Thus different cell layers can adhere to each other without the danger of intermingling if integrins are used as the adhesion molecules.

The examination of mutations in the PS integrins has shown us that integrins play a vital role in the adhesion of the different parts of the organism to each other. What relationship does this type of cell adhesion have with other cell interactions occurring during development? We can consider cell adhesion proteins that adhere like cells together, such as cadherins and the product of the crumbs gene in Drosophila (Tepass et al., 1990), to be 1° cell adhesion molecules, while the integrins, whose role is to adhere different cell layers to each other, would then be 2° adhesion molecules. From the PS integrin example it appears that 2° cell adhesion may occur via the extracellular matrix, consistent with the observation that tissues are surrounded by and kept separate by basement membranes.

There are several outstanding questions to be answered before we fully understand PS integrin mediated adhesion, which can be separated into questions about the PS integrin molecules themselves and questions about components that mediate the adhesion both inside and outside the cell. One important point is whether there are other integrin subunits that form functional heterodimers with any of the three PS integrin subunits that have been identified so far. The phenotype of inflated null mutant embryos, which lack the αps2 subunit, is clearly milder than that of myospheroid mutant embryos, which lack the βps subunit (N. H. B. unpublished data). The muscles remain attached longer and there is no defect in the dorsal epidermis. This observation rules out a possible model where the PS integrins function solely by direct binding of PS1 to PS2, since in this case one would expect the phenotypes of myospheroid and inflated to be identical. Possible explanations include postulating the existence of other a subunits that can partially complement the loss of βps2, or residual adhesive activity mediated by the PS1 integrin and other cell surface proteins. Since the integrins are required for the normal rate of extracellular matrix assembly in the embryo, the PS1 integrin might be sufficient to assemble the matrix, and other matrix binding proteins expressed on the surface of the muscle cell might provide the adhesion that keeps the muscle attached for that bit longer.

While we can explain the observation that null mutations in the αps2 subunit have a weaker phenotype than mutations in the βPS subunit, if we have identified all the PS integrin subunits then we would expect that a double mutant for both αps1 and αps2 should have an identical phenotype to a βPS mutant. Recent experiments have shown that this is not the case (N.H.B. unpublished results) suggesting either that there are additional PS integrin α subunits or the more unlikely possibility that the β subunit can get to the surface without an associated α subunit and is functional as a monomer or homodimer. Recently a new β subunit has been cloned from Drosophila (Yee and Hynes, 1993), which may be the first of many additional integrin subunits found in this organism. With 14 α subunits and 8 β subunits identified in vertebrates, there is room for several more, even considering the smaller genome size of Drosophila.

Since we envisage the PS integrins as links between the extracellular matrix and the cytoskeleton it will be important to identify the molecules with which the PS integrins directly interact with. Of the candidate molecules that we might expect to perform these functions, based on studies in vertebrates, collagen, laminin and a-actinin have been cloned from Drosophila. The analysis of mutants for oc-actinin and laminin suggests that they are not essential components of PS integrin mediated adhesion because they do not show similar phenotypes (Fyrberg et al., 1990; Hench-cliffe et al., 1993). It will be advantageous to expoit the genetics of Drosophila to identify these components through genetic screens for additional loci that have similar phenotypes to the PS integrin mutants or through screens for enhancers of PS integrin mutant combinations. It will be interesting to see if the extracellular ligands and cytoplasmic proteins that link the PS integrins to the cytoskeleton are the same in all cells and throughout the life cycle or whether there are specific forms of these proteins.

One final important point to consider is how well the most dramatic aspects of the phentoypes of integrin mutations reflect the function of integrins. It is clear that the muscle attachments, for example, have a specialised structure composed of interdigitating membranes and adhesive junctions that keep the different cell layers attached to each other tightly enough to withstand the strong forces of muscle contraction. The PS integrins clearly play an important role in forming this structure and are likely to be a major structural component of it. The detachment of the muscles in myospheroid mutant embryos is a very dramatic aspect of the phenotype, and we have extrapolated from this observation to a general role for integrins in adhesion between cell layers. This generalisation is supported by the similar failure of the attachment of the visceral muscles to the midgut, the two layers of the wing blade to each other and the pigment cells to the fenestrated membrane in the eye. However, more subtle defects in the differentiation of cells or in cell shape or behaviour may have been missed. While the phenotype of embryos lacking zygotic PS integrin function has been examined reasonably thoroughly, the phenotype of embryos that lack both maternal and zygotic components has only been examined by looking at the late embryonic cuticle, which shows that there is a much more severe phenotype. This phenotype, where germ band shortening does not occur normally, suggests that the PS integrins may also be involved in cell migration and the cell shape changes that accompany the shortening of the germ band. The relatively small amount of the βPS subunit that is deposited in the egg as mRNA and protein must be sufficient to perform these tasks. With the new methods for generating germ line clones more efficiently it should be possible to perform a more thorough analysis of cellular behaviour and differentiation in embryos that completely lack the PS integrins. Further analysis of the PS integin mutations and the phenotypes of other components in integrin mediated events will hopefully clarify whether the integrins play an important role in the fate of cells in addition to their role in holding the organism together.