DEVELOPMENTAL DYNAMICS 215:179–189 (1999) Role of Actin Stress Fibres in the Development of the Intervertebral Disc: Cytoskeletal Control of Extracellular Matrix Assembly ANTHONY J. HAYES, MICHAEL BENJAMIN, AND JAMES R. RALPHS* Connective Tissue Biology Laboratory, Cardiff School of Biosciences, Cardiff University, Cardiff, United Kingdom ABSTRACT Orientation of collagen fibrils is a key event in the development of many tissues. In the intervertebral disc, the outer annulus fibrosus comprises lamellae of parallel collagen fibres, the direction of orientation of the long axis of which alternates in angle between lamellae. In development, this organisation is preceded by the formation of sheets of oriented fibroblasts, which then deposit the oriented lamellae. Here, using fluorescent labelling, confocal and electron microscopic techniques on developmental series, we show that the orientation of cells in lamellae is associated with the formation of adherens junctions intercellularly, involving cadherins and vinculin, and longitudinal stress fibres (label for filamentous actin and tropomyosin) intracellularly. The stress fibres direct the initial elongation of cells and control the deposition of oriented extracellular matrix via junctional complexes with the matrix involving vinculin and ␣ 5 ␤ 1 integrins, which in turn promote the formation of oriented fibronectin at the cell surface; oriented collagen is deposited between cells at the same stages. Shortly after birth, the stress fibres disappear, probably because cells now gain orientational cues from the matrix, and are undergoing differentiation-related changes to form fibrocartilage cells. Dev Dyn 1999;215:179–189. r 1999 Wiley-Liss, Inc. Key words: intervertebral disc; development; stress fibres; cell–cell interactions; cell–matrix interactions INTRODUCTION Orientation of collagen fibres during their deposition by fibroblasts is a key event in development of many connective tissues. For example, fibrils must be more-orless parallel and run along the long axis of tensile structures (e.g., tendon and ligament) or are interwoven in tissues experiencing load from various directions (e.g., dermis). Some tissues consist of a series of collagenous lamellae, with parallel fibrils within lamellae but with orientation being different in successive lamellae. The most highly ordered and tightly controlled example is probably in the cornea, but such organisation also occurs in bone and in the intervertebral disc. Here, we describe the sequence of events occurring in the deposir 1999 WILEY-LISS, INC. tion of the lamellae of the annulus fibrosus of the intervertebral disc. The intervertebral disc is a composite structure, consisting of the outer annulus fibrosus, of sclerotomal origin, and inner nucleus pulposus, initially derived from the notochord. Two distinct regions of the annulus can be distinguished: the outer annulus, which is fibrous in origin, and the inner annulus, which is highly fibrocartilaginous and is initially derived from cartilage (Peacock, 1951a,b; Walmsley, 1953; Humzah and Soames, 1988; Rufai et al., 1995). The outer annulus is constructed of lamellae of ordered collagen fibril bundles, which are arranged concentrically around the nucleus, and anchored firmly into the cartilage endplates of the vertebral bodies (Humzah and Soames, 1988). Within individual lamellae, collagen fibres run parallel to each other and are inclined at a precise angle to the long axis of the spine (Hickey and Hukins, 1980). The inclination of this angle alternates in adjacent lamellae, conferring cross-ply strength to the annulus fibrosus. Functionally, the annulus allows twisting and bending of the spine, and the nucleus resists compressive loading; some of that compressive loading is transferred to the inner, fibrocartilaginous part of the annulus (Bogduk, 1997). During development, the lamellar organisation of oriented collagen fibres in the annulus is preceded by a similar organisation of sheets of oriented fibroblasts arranged concentrically around the early nucleus (Peacock,1951 a,b; Walmsley, 1953; Rufai et al., 1995). The oriented sheets of fibroblasts clearly lay down the organised lamellae, but how cells become oriented into cell sheets, and how the cell sheets lay down an oriented extracellular matrix (ECM) is unclear. Cells lay down oriented collagen in association with a series of cellular compartments (Birk and Zycband, 1994) and, at least in tendon, appear to cooperate with one another in depositing the oriented matrix (McNeilly et al., 1996). However, as the developing annulus fibrosus shows, there has to be a cellular orientation phase Grant sponsor: Arthritis and Rheumatism Council; Grant number: R0521. *Correspondence to: James R. Ralphs, Connective Tissue Biology Laboratory, Cardiff School of Biosciences, Cardiff University, Preclinical Buildings, Museum Avenue, P.O. Box 911, Cardiff CF1 3US, United Kingdom. Received 19 January 1999; Accepted 26 March 1999 180 HAYES ET AL. before cells can deposit their collagen in the correct lamellar orientation. Whilst considerable attention has been paid to segmentation and patterning along the axial skeleton (e.g., Wallin et al., 1994; Burke et al., 1995; Christ and Ordahl, 1995; Smith and Tuan, 1995), little is known about development downstream of patterning events, and how these highly ordered structures actually form: what causes cells to orient and deposit oriented ECM? Cell shape and movement is mediated by the cytoskeleton, under the control of various external influences including cell–cell and cell–matrix interactions (e.g., Zanetti and Solursh, 1986; Hay, 1989; Ingber et al., 1994; Mooney et al., 1995). In the present study, we examine the role of the cytoskeletal elements actin and vimentin and their associated cell–cell and cell–matrix junctions in the formation of oriented cell sheets and the subsequent deposition of collagen lamellae. We show important roles for actin-associated cell–cell junctions in early formation of annulus fibrosus blastemas and actin stress fibres and cell–matrix junctions in early orientation of cells and the ECM. RESULTS Development of the Rat Intervertebral Disc: Histological Overview At embryonic day (E)15, the spine consisted of repeated perinotochordal condensations of mesenchyme, the intervertebral disc anlagen, which were separated by the early cartilage of the developing vertebral bodies (Fig. 1). At this stage, the cells making up the disc anlage were closely packed together, with no histologically detectable ECM (see also next paragraph). The outermost cells were flattened and the innermost were more rounded or polygonal. By E16, the notochord had bulged at regular intervals along its length, clearly defining the position of the future nuclei pulposi, and the annulus had formed distinct inner and outer parts (Figs. 2 and 3). The inner annulus was cartilaginous, continuous with the vertebral bodies, and contained round-to-oval cells that were separated by small amounts of highly Alcian-blue-positive cartilage matrix. The outer annulus consisted of compact sheets of elongate, parallel fibroblasts, again with no histologically discernible ECM (Figs. 3 and 4). These cells were precisely oriented at a constant angle to the long axis of the vertebral column. The inclination of this angle alternated in neighbouring sheets, producing a complex lattice-like arrangement of cells when seen in parasagittal sections (Fig. 4). With further development, ECM accumulated between cell sheets, separating them from one another. In the neonate, the lamellar arrangement of cells and ECM was fully apparent, and by 2 weeks the annulus consisted of rows of rounded cells separated by the prominent oriented collagenous lamellae (Fig. 5). Ultrastructurally, at E15 outer annulus cells were somewhat flattened, with many cell–cell contacts (Fig. 6). Many of these contacts contained distinct cell junc- tions, probably adherens junctions (Fig. 7; see below). Spaces between cells contained an extremely sparse ECM. By E16, cells had become very elongated and the alternating orientation of cells in successive cell sheets could easily be appreciated (Fig. 8). There was still very little ECM, although small bundles of banded collagen fibrils could be seen between the closely packed cells, intimately associated with the cell surface (Fig. 9). The cells themselves contained prominent longitudinal fibrillar rods, consistent in appearance with stress fibres (Figs. 8–10) and positioned just beneath the plasmalemma (Figs. 9 and 10). Dense plaques were sometimes observed immediately beneath the plasmalemma that could have represented cell–matrix junctions (probably focal adhesions; see below). In the neonate, cells were much more rounded in outline, embedded in a densely collagenous, highly ordered ECM and had no obvious stress fibres in the cytoplasm (Fig. 11). Cytoskeletal Organisation Actin organisation was intimately associated with the stage of development of the outer annulus. At E15, actin label was diffuse throughout cells of the disc anlagen (Fig.12). Starting at late E15, and fully by E16, actin label was strong in the oriented fibroblasts of the outer annulus (Fig. 13). This pattern of actin distribution persisted in the outer annulus until birth (Figs. 14–18). In sections tangential to the outer annulus, parallel cables of actin ran along the long axes of the elongated fibroblasts, with the change in cell orientation in neighbouring cell sheets giving the appearance of a lattice-like arrangement of actin filaments (Figs. 14, 16, and 18). In dual labels, tropomyosin colocalised with actin cables in the outer annulus in all embryonic stages investigated, showing them to be actin/myosin stress fibres (Fig. 16). In transverse sections of the disc, the stress fibres were clearly seen to be restricted to the closely packed fibroblasts of the outer annulus (Fig. 15). In the neonate, the complex lattice of actin cables had disappeared (Fig. 19), although, as at all stages, the notochordal cells of the nucleus pulposus labelled strongly. Actin label in the annulus at this stage often appeared punctate, with few traces of the fibre organisation observed earlier, and there was no label for tropomyosin. Vimentin was absent from the annulus up to E18/19, although it was strongly labelled in the notochord and the notochordally derived nucleus pulposus. At E20, weak label was detected in the outer annulus, along with strong label in the nucleus pulposus and in chondrocytes of the vertebral body, notably hypertrophic chondrocytes (Fig. 20). In neonates, vimentin label was more prominent, with longitudinally running filaments being present in the annular fibroblasts (Fig. 21). Cell Junctions Cadherins were detected at the E15 stage, in a region of cells with bright label on their surfaces in the position of the future annulus (Fig. 22) and in the ACTIN AND MATRIX IN ANNULUS FIBROSUS DEVELOPMENT Figs. 1–5. Histology of the developing intervertebral disc. Fig. 1. Resin section of E15 spine, midsagittal plane. Dense cellular anlagen of the intervertebral discs (d) are positioned between the cartilage of the developing vertebral bodies (vb), with the notochord (n), which is of uniform diameter, running longitudinally through the centre of all. The outer part of the disc anlagen appears dark and consists of close-packed, flattened cells (see Fig. 6). The inner part consists of larger, paler cells of more polygonal shape. (Toluidine blue.) 181 Fig. 3. Higher magnification view of E16 disc shows nucleus pulposus (np) and differentiation of annulus into cartilaginous inner (ia) and fibrous outer (oa) parts. Line indicates approximate position of section shown in Fig. 4. (Haematoxylin, eosin and Alcian blue.) Fig. 4. Section tangential to outer annulus of E16 disc, in the approximate position indicated in Fig. 3. The section shows a lattice-like arrangement of cells oriented into parallel sheets linking the vertebral bodies (vb), with successive sheets lying at an angle to one another. (Haematoxylin, eosin and Alcian blue.) Fig. 2. Paraffin section of E16 spine, midsagittal plane. The notochord has bulged at regular intervals within the developing discs to form the precursors of the nuclei pulposi (np), and overt tissue differentiation has occurred in the annulus (See Fig. 3). (Haematoxylin, eosin and Alcian blue.) Fig. 5. Midsagittal section of outer annulus of 2-week disc. Layers of cells have laid down oriented lamellae of collagen, with successive lamellae showing alternating orientation of parallel collagen fibrils. The inner annulus (ia) is at the lower edge of the figure. (Haematoxylin, eosin and Alcian blue.) notochord. By E16, label was clearly restricted to the region of the future outer annulus (Fig. 23); at this and all subsequent stages, label could be detected in the notochordally derived nucleus pulposus. By E17, label could no longer be detected in the annulus (Fig. 24). Vinculin could be detected in the annulus fibrosus at all 182 HAYES ET AL. Figures 6–11. ACTIN AND MATRIX IN ANNULUS FIBROSUS DEVELOPMENT stages examined. At the late E15/early E16 stage when cell orientation had begun, vinculin had distribution similar to that of cadherins (Fig. 25). Later, dual labels showed clusters of small, bright foci of vinculin label, some forming streaks along the actin stress fibres (Fig 26), others as more rounded clusters that were on the cell surface in the region of the cell around the nucleus. Desmosomal antigen could not be detected at all in the annulus at the stages examined, although it was prominent between notochordal cells of the nucleus pulposus at all stages (Fig. 27). ␣5 and ␤1 integrin subunits were present at all stages. At E15, they were associated with cells but showed no obvious orientation. From E16 onwards, the integrins labelled as elongate streaks on the surfaces of oriented annular fibroblasts, giving a lattice-like appearance in glancing sections of outer annulus (Figs. 28 and 29). Cultured annular fibroblasts contained prominent stress fibres (Fig. 30) that were associated with rods of fibronectin on the cell surface (Fig. 31). Extracellular Matrix In vivo, fibronectin label was present pericellularly throughout the disc tissue in all stages investigated. Before E16, it had no obvious organisation; at E16 and beyond, in the outer annulus it formed a bright, prominent lattice-like pattern, with long streaks of label being associated with the cell surface and co-localising with the submembranous actin bundles (Fig. 32). Type I collagen label occurred only in the outer annulus. At E16 it was weakly present, but labelling intensity increased with age. It formed a lattice-like arrangement of fibres in the embryonic stages. By the neonate stage, a well-developed lamellar arrangement of type I collagen fibre bundles had developed (Fig. 33). DISCUSSION The developmental sequence for the annulus fibrosus matches that previously described in human intervertebral discs (Peacock, 1951a,b; Walmsley, 1953) and extends our previous observations on rat disc develop- Figs. 6–11. Ultrastructure of the developing annulus fibrosus. Fig. 6. Low-power view of future outer annulus fibrosus cells at E15. Cells are flattened and have many cell processes and cell–cell contacts. There is a very sparse extracellular matrix between cells, with no visible collagen fibrils. Fig. 7. High-power view of cell–cell contacts at E15. Adherens junctions (arrows) are prominent at points of contact between cells. Fig. 8. Low-power micrograph of the E16 outer annulus. Cells are elongated and spindle shaped, with the alternating orientation of cells in successive sheets clearly visible by the differing cell outlines—highly elongate in longitudinal section (white asterisks) and more rounded in oblique section (black asterisks). Cells contain prominent fibrillar longitudinal stress fibres (arrowheads). Fig. 9. High-power view of two cells from adjacent cell sheets at E16; the cell on the left is cut in longitudinal section, on the right in transverse section. The image is oriented at 90° to Fig. 8. The section passes close to 183 ment (Rufai et al., 1995). There is clearly the requirement for orientation of fibroblasts to deposit an oriented ECM structure. The results presented here give indications of processes that may be important in initial orientation of fibroblasts, and subsequently in the deposition of the oriented collagenous lamellae. We provide evidence that cell–cell and cell–matrix interactions involving the actin cytoskeleton are important in cell organisation and subsequent matrix organisation. Stress fibres are often regarded as artefacts of culture, and indeed for many cell types this is probably the case; for example, chondrocytes do not contain them in vivo (Durrant et al., in press), but contain them in abundance if grown on tissue culture plastic surfaces (e.g., Benya et al., 1988). However, stress fibres do occur in vivo at some sites and under some circumstances, e.g., in endothelial cells (White and Fujiwara, 1986; Sugimoto et al., 1995) and repairing epithelia (Nodder and Martin, 1997). There are also reports of contractile actin filaments in some connective tissue cells in vivo: prominent actin filaments have been associated with matrix deposition in fibroblasts/myofibroblasts of granulation tissue (Gabbiani, 1994; Horiba and Fukuda, 1994; Desmouliere, 1995; Berry et al., 1998), including that of healing and remodelling ligament scar (Faryniarz et al., 1996) and healing cornea (Petroll et al., 1993), and stress fibres occur in tendon cells (Ralphs et al., 1998). In the present study, the appearance of stress fibres, as shown by the codistribution of filamentous actin and tropomyosin, and actin-associated cell–cell and cell–matrix junctions are intimately associated with the onset of development and growth of the outer part of the annulus fibrosus. At the earliest stages, E15 and E16, the presence of cadherins (important in cell–cell interactions and components of adherens junctions, e.g., Yonemura et al., 1995) on annular fibroblasts suggests that cell–cell attachment is of key importance in establishing cellular orientation. The codistribution of vinculin (component of adherens junctions and focal adhesions with the ECM) with cadherins at these stages, along with ultra- the midregion of both cells, and the nuclei (nu) occupy most of the cytoplasm. A longitudinal stress fibre (sf) is clear on the left, between the nucleus and the plasmalemma. A transverse section of a stress fibre (white asterisk) is present on the right, again between the nucleus and plasmalemma. Obliquely sectioned collagen fibrils are present in the narrow space between the cells (black asterisk). Fig. 10. Similar view of two more adjacent cells. The longitudinally sectioned cell on the left contains a longitudinal stress fibre (sf). The cell on the right has prominent focal densities on the cytoplasmic side of the plasmalemma (arrows). Longitudinally running collagen fibrils are present between the cells (asterisk). m, mitochondrion; er, rough endoplasmic reticulum. Fig. 11. Low-power micrograph of neonatal outer annulus. The annulus now contains prominent oriented lamellae of collagen fibrils with alternating orientation—obliquely sectioned at the white asterisk, transversely at the black asterisk—with cells positioned between them. Cells are more rounded with no stress fibres in the cytoplasm. 184 HAYES ET AL. Figures 12–31. ACTIN AND MATRIX IN ANNULUS FIBROSUS DEVELOPMENT 185 structural observations, suggests that cells are linked together via adherens junctions. The integrins are also present, indicating the presence of cell–matrix associations as well. Adherens junctions and focal adhesions have been shown to mediate assembly of stress fibres from the junctional structures (e.g., Chi-Rosso et al., 1997; Mackay et al., 1997; Turnacioglu et al., 1998). Therefore, it may be that the initial assembly of stress fibres at the E15/16 stage is initiated by the formation of cell–cell and/or cell–matrix junctions. The fact that cadherin expression is lost after E16 suggests that the cell–cell junctions are of key importance in establishing the initial cellular pattern. Stress fibre formation probably drives the elongation of outer annulus cells within the plane of their cell sheets after E15; thus, the orientation of successive sheets of fibroblasts seen clearly at E16 would be determined by the arrangement of junctional contacts in the E15 disc. The continued expression of vinculin, along with the organisation of integrin subunits into streaks associated with stress fibres, suggests that cell–matrix contacts are important in subsequent organisation and deposition of the ECM. Ultrastructural evidence supports this: cells at stages where there are prominent stress fibres have collagen fibrils at the cell surface, along with evidence of dark plaques at the plasma membrane, suggestive of focal contacts. To summarise, initial stages of annulus organisation require cell–cell interactions involving cadherins, resulting in the formation of adherens junctions and stress fibres, but when cellular patterns are established and matrix synthesis begins, cells switch to having their major interactions with the ECM. The notochord is of obvious importance to disc development. It plays a key role in the onset of annulus development; there is no overt differentiation until the notochord starts to bulge in the regions of the future nuclei pulposi. Recent work (Aszodi et al., 1998) indicates that the bulging of the notochord is driven by Figs. 12–31. Confocal microscope images show cytoskeletal organisation and distribution of cell junctional components in the developing annulus fibrosus and in vitro; all but Figs. 16 and 26 are counterstained with propidium iodide to give cellular context. chondrocytes of the vertebral bodies, especially hypertrophic chondrocytes (h). Weak label is present in the outer annulus; vimentin label was not detected in the annulus before this stage. Fig. 12. E15 disc anlagen labelled with phalloidin–fluorescein isothiocyanate. Label is prominent in cells of the disc (d), notochord (n) and vertebral bodies (vb). It appears brighter in the disc anlagen because of the high cell density. Fig. 13. E16 disc labelled with phalloidin–fluorescein isothiocyanate. Label is of particularly high intensity in the outer annulus, although still present in other cell types, including the bulging notochord (n). The white line indicates the approximate position of the section shown in Fig. 14. Fig. 14. E16 disc labelled with phalloidin–fluorescein isothiocyanate. High-power view of a tangential section, position indicated in Fig. 13. Cells contain prominent longitudinal actin cables, consistent in position and orientation with those seen ultrastructurally. At least two layers of cells are present in this 20-µm cryosection, their alternating orientations giving rise to the lattice pattern in the confocal reconstruction. Fig. 15. Transverse section of E17 annulus, labelled with phalloidin– fluorescein isothiocyanate. The section shows that actin cables are tightly restricted to the outer annulus (oa): neither the cartilaginous inner annulus (ia) nor the surrounding tissue (upper left of figure) contain them. Fig. 16. Tangential sections of E18 outer annulus dual labelled with phalloidin–fluorescein isothiocyanate (green; top) and indirect immunofluorescence for tropomyosin with Texas red second antibody (red; bottom). Distributions in the outer annulus are identical, confirming identification of actin cables as stress fibres. Fig. 21. Tangential section of neonatal outer annulus, immunolabelled for vimentin. Outer annular fibroblasts contain longitudinal vimentin filaments. Fig. 22. Sagittal (slightly oblique) section of E15 disc anlagen immunolabelled with pan-cadherin antibody. The notochord (n) passes centrally through the vertebral body (vb) to the right and the disc anlagen (d) but then leaves the section plane. Prominent label is present between cells in the notochord and disc anlagen; there is none in the vertebral bodies. Fig. 23. Sagittal section of E16 annulus immunolabelled with pancadherin antibody. Punctate label is prominent between cells of the organising cell sheets in the outer annulus (oa). Fig. 24. Sagittal section of E17 disc immunolabelled with pancadherin antibody. Label is restricted to the notochordal nucleus pulposus; none can be detected in the outer annulus (oa). Fig. 25. Sagittal section of late E15/early E16 outer annulus immunolabelled for vinculin. Numerous small foci of label are present between cells of the developing outer annulus (oa). ia, inner annulus. Fig. 26. Tangential section of E19 outer annulus, dual labelled for actin (phalloidin: green) and vinculin (immunolabel, Texas red: red). Vinculin label occurs as clusters of foci, some of which line up along stress fibres (presumably where these meet the cell surface), and others form rounded clusters. These are associated with the cell surface in the vicinity of the nucleus (not visible in this preparation). Fig. 17. E19 disc labelled with phalloidin–fluorescein isothiocyanate. Low-power view of developing disc. Actin label remains intense in the outer annulus, and in the notochordal cells of the nucleus pulposus (np). Fig. 27. Sagittal section of neonatal disc, immunolabelled for desmosomal antigen. No desmosomes are detected in the annulus; they are present between the notochordal cells of the nucleus pulposus (n) and in association with blood vessels at the periphery of the disc (arrow). Fig. 18. Slightly oblique tangential section of E19 outer annulus, labelled with phalloidin–fluorescein isothiocyanate. Successive layers of cells with alternating orientation are clearly demonstrated by nuclear angle and actin cable direction. Figs. 28 and 29. Tangential sections of E18 outer annulus immunolabelled for ␣5 and ␤1 integrins, respectively. Both are distributed as longitudinal streaks associated with the cells. Fig. 19. Tangential section of neonatal outer annulus, labelled with phalloidin–fluorescein isothiocyanate. Actin cables, prominent until the day of birth, have dispersed, leaving a more punctate actin distribution. Fig. 20. Midsagittal section of E20 disc immunolabelled for vimentin. Label is prominent in the notochordal nucleus pulposus (np) and in Figs. 30 and 31. Cultured outer annular fibroblasts labelled with phalloidin for actin and immunolabelled for fibronectin, respectively. Cells contain longitudinally oriented actin stress fibres, and longitudinally oriented streaks of fibronectin at the cell surface. Fibronectin also can appear as clusters of bright foci on the cell surface in the vicinity of the nucleus, usually on the undersurface of the cell (Fig. 31, arrow). 186 HAYES ET AL. Figs. 32 and 33. Extracellular matrix organisation in the developing annulus fibrosus. Confocal reconstructions of immunolabelled 20-µm cryosections. studies (not shown) indicate that these run longitudinally in association with outer annular fibroblasts. Fig. 32. Tangential section of E21 foetal outer annulus immunolabelled for fibronectin. Label is in the form of thin, bright rods; dual label Fig. 33. Tangential section of neonatal outer annulus immunolabelled for type I collagen. Collagen fibres appear as a lattice in this reconstruction, which contains several alternately oriented lamellae. pressure from the developing cartilage in the vertebral bodies. The notochord narrows as it is squeezed circumferentially and stretched longitudinally by the developing vertebral bodies, its cells being pushed into the notochordal bulges in the weaker, nonpressure-producing disc anlagen. This mechanism of notochordal movements explains the prominence of cytoskeletal elements and cell–cell junctions in notochordal cells: they must remain attached firmly to one another so that all end up in the early nucleus pulposus. If cartilage development is prevented (by type II collagen knockout), the notochord does not bulge, and the intervertebral disc does not differentiate (Aszodi et al., 1998). It would appear, therefore, that provided the axial pattern of repeated disc anlagen/vertebral body anlagen is established, then disc differentiation will proceed, provided cartilage differentiation occurs in the vertebral bodies. Within the disc anlagen, some changes are evident before notochordal bulging, because there are differences between the E15 inner and outer annulus; this could potentially be mediated by signalling molecules from the notochord. The stimulus to promote disc differentiation could simply be mechanical: the notochordal bulges will exert compressive loads on the inner annulus precursors, enclosed as they are by outer annulus precursors, which could lead to the chondrogenic differentiation of the inner annulus. Cells of the outer annulus may experience different loads, because they are attached to one another and are wrapped around an enlarging structure; they would thus experience some tensile forces, which could trigger stress fibre formation and deposition of fibrous matrix. The organisation of ECM (fibronectin at the cell surface and collagen fibrils of the developing lamellae) mirrors that of the actin stress fibres in the cell sheets. Indeed, we show that all of the elements are in place at the appropriate developmental stages for there to be a transmembrane link from the stress fibres to the ECM—in sequence, stress fibre, vinculin, integrin, cell surface fibronectin, and collagen. As further support for this suggestion, the ultrastructural studies showed intimate association between actin stress fibres with the inside of the plasmalemma and collagen fibres with the outside. The integrins examined together form the fibronectin receptor, ␣5␤1 integrin (Cheresh and Mecham, 1994). We suggest that this receptor is organised from the interior of the cell by stress fibres via adhesion complexes, and thus organises fibronectin on the cell surface. Similar events seem to occur in early stages of bone formation (Winnard et al., 1995). Recent in vitro studies also show close association of stress fibres, integrins, and fibronectin, as seen with annulus cells. Fibronectin accumulates on cell surfaces in vitro via the ␣5␤1 integrin (Sechler et al., 1997; Bordoulous et al., 1998) and then self-assembles into filaments under the influence of other receptors, or ‘‘matrix assembly sites’’ (Sechler et al., 1997; Hocking et al., 1998). Fibronectin assembly at the cell surface leads to stress fibre reorganisation and also modulates junction formation and outside-in signalling processes (Sechler and Schwarzbauer, 1997; Hocking et al., 1998). In the developing disc, it seems likely that stress fibres organise the positions of the receptors and direction of assembly of cell surface fibronectin. The organisation of the receptor may be dependent on ligand binding; in formation of typical focal contacts, ligand binding leads to clustering of the receptor and interaction with the cytoskeleton (e.g., Pfaff et al., 1998). It is unclear whether unbound receptors can link with the cytoskeleton (Hughes and Pfaff, 1998). In the developing annulus, therefore, fibronectin may be ‘‘captured’’ by the integrin and then organised along stress fibre–associ- ACTIN AND MATRIX IN ANNULUS FIBROSUS DEVELOPMENT ated focal contacts. This seems likely to be the sequence of events because fibronectin is present throughout the period studied, but only organises when stress fibres are present. It is of crucial importance to tissue function that cells and matrix orient correctly, and thus it seems probable that a cell organisational event—stress fibre formation based on cell–cell contact —precedes a matrix organisational event, i.e., fibronectin at the cell surface. Once the pattern is established, integrinmediated signals will almost certainly feed back into the cell, and can have profound effects on cell behaviour (Bourdoulous et al., 1998; Sechler and Schwarzbauer, 1998). The individual integrin subunits examined can also take part in forming receptors for other ECM molecules (Cheresh and Mecham, 1994), and it seems likely that other integrins will be present that will also take part in matrix organisation. The intermediate filament vimentin does not seem to be involved in the cell and matrix orientation process, because it appears too late. It certainly does not appear to be involved in the cell–cell interactions because no desmosomal contacts can be detected. At the later stages, cells do label for vimentin, and here it could thus be involved in cell–matrix interactions. Vimentin could also be associated with annulus differentiation. In tendons, fibrocartilage occurs at sites where the tendon experiences compressive loads; at these sites, fibrocartilage cells accumulate large quantities of vimentin (Ralphs et al., 1991, 1992). Intermediate filaments are of clear importance in adaptation of tissues to mechanical stress (Galou et al., 1997). It may be, therefore, that the accumulation of vimentin in annulus cells relates to the onset of fibrocartilaginous differentiation, with increased compression being transferred from the nucleus pulposus at late foetal stages and after birth. The loss of stress fibres from annular fibroblasts after birth suggests that they are no longer required for cell orientation and matrix organisation. It may be that with a well-ordered matrix already laid down, cells now orient to the ECM, rather than requiring their own internal organisers. Bone cells may behave similarly—at the earliest stages of membrane bone formation, the osteoblasts contain prominent actin fibres that organise fibronectin at the cell surface (Winnard et al., 1995). Fibronectin then loses its association with the cell surface, presumably when the osteoblasts have established their initial oriented ECM. There may be further reasons for the disappearance of stress fibres from annulus cells. If, as discussed above, the rapid bulging of the notochord causes tensile stress in the outer annulus and triggers stress fibre formation, then the more static nature of the structure once it has become established may lead to a reduction in tension and thus disassembly of the stress fibres. Also, it should be noted that cell shape changes occur subsequent to birth, with cells becoming progressively more rounded as they differentiate into fibrocartilage cells (Rufai et al., 1995). The stress fibres would have to be dismantled 187 to allow cells to round up, but also possession of stress fibres appears to preclude a chondrogenic phenotype. Chondrocytes cultured on plastic flatten, form stress fibres, and dedifferentiate to a fibroblastic phenotype; if stress fibres are eliminated by treatment with cytochalasin, they redifferentiate to the chondrocyte phenotype (e.g., Zanetti and Solursh, 1984; Benya et al., 1988). The loss of stress fibres from the annulus fibroblasts may thus be an important early stage in fibrocartilaginous differentiation. EXPERIMENTAL PROCEDURES Animals Embryonic and young postnatal white Wistar rats were used throughout. Pregnant females were killed by cervical dislocation, and embryos were removed from the uterus, decapitated, and processed for histology or immunohistochemistry. For routine histology, the torsos were processed whole; for electron microscopy, regions containing individual discs or their primordia were dissected out; for immunohistochemistry, the whole lumbar spine was dissected out before processing. At least five animals were used for each procedure at each of the following ages: Embryonic day (E) 15,16,17,18, 19, 20, 21; neonates; and 1 week after birth. Histology Specimens were fixed in 10% neutral buffered formol saline for 5 days; those from animals older than E19 were in addition decalcified in 2% nitric acid, and then all were dehydrated, cleared, and embedded in paraffin wax. Serial 8-µm sections were cut in the sagittal plane and stained with Alcian blue and haematoxylin and eosin. Micrographs were taken using a Leica DM12 photomicroscope. For electron microscopy, specimens were fixed in 2% glutaraldehyde/2% paraformaldehyde, postfixed in osmium tetroxide, dehydrated, and embedded in araldite. Semithin sections were stained with toluidine blue for light microscopy; ultrathin silver-gold sections were stained with uranyl acetate and lead citrate and examined with a Phillips EM 208 transmission electron microscope. Immunohistochemistry Specimens were frozen unfixed with dry ice, and serial sagittal cryosections were cut at a thickness of 15 µm. Sections passing through intervertebral discs were fluorescently labelled as follows. Cytoskeletal components. Actin was directly labelled using phalloidin/fluorescein isothiocyanate (FITC) (1µg/ ml; Sigma, Poole, UK), the actin stress fibre component tropomyosin indirectly labelled using monoclonal antibody TM311 (1:400, Sigma; Das et al., 1993), and the intermediate, filament vimentin using monoclonal antibody Vim13.2 (1:200, Sigma; Adams and Watt, 1988). Junctional complexes. Cell junctions associated with actin were labelled with antibodies to vinculin (associated with cell–cell adherens junctions and cell–matrix focal contacts; hVin1, 1:200, Sigma; Goncharova et al., 188 HAYES ET AL. 1992), and cadherins (pan-cadherin; monoclonal CH19, 1:100, Sigma; Takeichi, 1988), associated with cell–cell adherens junctions. Desmosomes, associated with cell–cell junctions involving intermediate filaments, were labelled using monoclonal antibody ZK31 (1:400, Sigma; Lang et al., 1986). Matrix receptors. The components of the fibronectin receptor ␣5␤1 integrin were labelled using rabbit polyclonal anti ␣5 integrin, (1:20, Chemicon, London, UK; Rossino et al., 1990; Pena et al., 1994) and rabbit polyclonal anti ␤1 integrin, 1:20, gift from Prof. R.O. Hynes; Marcantonio and Hynes, 1988). ECM components. Fibronectin and type I collagen were labelled using rabbit polyclonal antihuman fibronectin (1:200; Sigma) and monoclonal type I collagen (COL1, 1:2,000; Sigma), respectively. Antibody labelling was performed using standard procedures for indirect immunofluorescence, with labelling for collagen and fibronectin requiring enzyme pretreatment of hyaluronidase (0.25 U/ml) and chondroitinase ABC (0.1 U/ml at 37°C for 30 min to unmask epitopes. Controls were incubated with nonimmune mouse immunoglobulins (10 µg/ml), rabbit serum (1:100), or the primary antibody incubation step was omitted. Primary antibody binding was detected using FITC-conjugated goat anti-mouse IgG (1:100; Sigma). Dual labels for actin and vinculin, and actin and tropomyosin, were performed by first immunolabelling sections with the primary antibodies and a Texas-red conjugated second antibody (10 µg/ml; Vector Laboratories, Peterborough, UK) followed by direct labelling for actin as above. Where appropriate, nuclei were counterstained with propidium iodide (0.5 µg/ml; Molecular Probes, Leiden, The Netherlands). All sections were mounted in Vectashield fade-retarding mountant (Vector Laboratories) and examined using a Molecular Dynamics Sarastro 2000 confocal laser scanning microscope, set up either for single- or dual-channel fluorescence recordings. Specimens were scanned with a 25-mW argon ion laser with appropriate excitation and emission filters for fluorescein (excitation 494 nm/ emission 518 nm), fluorescein/propidium iodide or fluorescein/Texas red (excitation 488 nm, primary beamsplitter 510 nm, secondary beamsplitter 595 nm, detector filters 600 nm and 530 nm). Optical section series were taken through the whole cryosection thickness using ⫻40 and ⫻60 objectives, and three-dimensional projections were prepared from the section series using Molecular Dynamics ImageSpacey software running on a Silicon Graphics XS24 4000 workstation. Cell Culture Intervertebral discs were dissected from E19 embryos by cutting the endplates off the vertebral bodies. The endplate was then punctured and the nucleus pulposus squeezed out whole with gentle pressure. The remaining annulus fibrosus and endplates were incubated for 30 min in 1% collagenase (Sigma) to remove the cartilage of the inner annulus and the endplate. This treatment left just the fibrous annulus from which cells were disaggregated by further collagenase treatment. Cells were washed and plated into 30-mm tissue culture dishes in high glucose Dulbecco’s Modified Eagle’s Medium (DMEM) containing 10% foetal bovine serum (all tissue culture reagents from Gibco, Paisley, UK). After reaching confluence they were passaged and then 3 hr after replating were fixed in 90% methanol at 4°C and labelled for actin and fibronectin as above. REFERENCES Adams JC, Watt FM. 1988. An unusual strain of human keratinocytes which do not stratify or undergo terminal differentiation in culture. J. Cell Biol 107:1927–1938. Aszodi A, Chan D, Hunziker E, Bateman JF, Fassler R. 1998. Collagen II is essential for the removal of the notochord and the formation of the intervertebral discs. J Cell Biol 143:1399–1412. Benya P, Brown PD, Padilla SR. 1988. Microfilament modification by dihydrocytochalasin B causes retinoic acid-modulated chondrocytes to re-express the chondrogenic phenotype without a change in shape. J Cell Biol 106:161–170. Berry DP, Harding KG, Stanton MR, Jasani B, Ehrlich HP. 1998. Human wound contraction: collagen organisation, fibroblasts, and myofibroblasts. Plast Reconstr Surg 102:124–131. Birk DE, Zycband E. 1994. Assembly of the tendon extracellularmatrix during development. J Anat 184:457–463. Bogduk N. 1997. Clinical anatomy of the lumbar spine and sacrum, 3rd ed. New York: Churchill Livingstone. Bordoulous S, Orend G, Mackenna DA, Pasqualini R, Ruoslahti E. 1998. Fibronectin matrix regulates activation of RHO and CDC42 GTPases and cell cycle progression. J Cell Biol 143:267–276. Burke AC, Nelson CE, Morgan BA, Tabin C. 1995. Hox genes and the evolution of vertebrate axial morphology. Development 121:333– 346. Cheresh DA, Mecham RP. Integrins: 1994. molecular and biological responses to the extracellular matrix. San Diego: Academic Press, 1994. Chi-Rosso G, Gotwals PJ, Yang J, Ling L, Jiang K, Chao B, Baker DP, Burkly LC, Faswell SE, Koteliansky VE. 1997. Fibronectin type III repeats mediate RGD-independent adhesion and signalling through activated beta1 integrins. J Biol Chem 272:31447–31452. Christ B, Ordahl CP. 1995. Early stages of chick somite development. Anat Embryol 191: 381–396. Das KM, Dasgupta A, Mandal A, Geng X. 1993. Autoimmunity to cytoskeletal protein tropomyosin: a clue to the pathogenetic mechanism for ulcerative colitis. J Immunol 150:2487–2493. Desmouliere A. 1995. Factors influencing myofibroblast differentiation during wound healing and fibrosis. Cell Biol Int 19:471–476. Durrant LA, Archer CW, Benjamin M, Ralphs JR. In press. Organisation of the chondrocyte cytoskeleton and its response to changing mechanical conditions in organ culture. J Anat. Faryniarz DA, Charponnier C, Gabbiani G, Yannas IV, Spector M. 1996. Myofibroblasts in the healing lapine medial collateral ligament: possible mechanisms of contraction. J Orthop Res 14:228– 237. Gabbiani G. 1994. Modulation of fibroblastic cytoskeletal features during wound healing and fibrosis. Pathol Res Pract 190:851–853. Galou M, Gao J, Humbert J, Mericskay M, Li Z, Paulin D, Vicart P. 1997. The importance of intermediate filaments in the adaptation of tissues to mechanical stress: evidence from gene knockout studies. Biol Cell 89:85–97. Goncharova EJ, Kam Z, Geiger B. 1992. Involvement of adherens junction components in myofibrillogenesis in cultured cardiac myocytes. Development 114:173–183. Hay EB. 1989. Extracellular matrix, cell skeletons, and embryonic development. Am J Med Gen 34:14–29. Hickey DS, Hukins DWL. 1980. X-ray diffraction studies on the arrangement of collagenous fibres in human foetal intervertebral disc. J Anat 131:81–90. ACTIN AND MATRIX IN ANNULUS FIBROSUS DEVELOPMENT Hocking DC, Sottile J, Mckeown-Longo PJ. 1998. Activation of distinct alpha5beta1-mediated signalling pathways by fibronectin’s cell adhesion and matrix assembly domains. J Cell Biol 141:241–253. Horiba K, Fukuda Y. 1994. Synchronous appearance of fibronectin, intergrin alpha 5 beta 1, vinculin and actin in epithelial cells and fibroblasts during rat tracheal wound healing. Virchows Arch 425: 425–434. Hughes PE, Pfaff M. Integrin affinity modulation. 1998. Trends Cell Biol 8:359–364. Humzah MD, Soames RW. 1988. Human intervertebral disc: structure and function. Anat Rec 220:337–356. Ingber DE, Dike L, Hansen L, Karp S, Liley H, Maniotis A, McNamee H, Mooney D, Plopper G, Sims J, Wang N. 1994. Cellular tensegrity: exploring how mechanical changes in the cytoskeleton regulate cell growth, migration, and tissue pattern during morphogenesis. Int Rev Cytol 150:173–224. Lang AB, Odermatt BF, Ruettner JR. 1986. Monoclonal antibodies to human cytokeratins: application to various epithelial and mesothelial cells. Exp Cell Biol 54:61–72. Mackay DJG, Esch F, Furthmayr H, Hall A. 1997. Rho- and Racdependent assembly of focal adhesion complexes and actin filaments in permeabilized fibroblasts: an essential role for ezrin/radixin/ moesin proteins. J Cell Biol 138:927–938. Marcantonio EE, Hynes RO. 1988. Antibodies to the conserved cytoplasmic domain of the integrin beta 1 subunit react with proteins in vertebrates, invertebrates and fungi. J Cell Biol 106:1765–1772. McNeilly CM, Banes AJ, Benjamin M, Ralphs JR. 1996. Tendon cells in vivo form a three dimensional network of cell processes linked by gap junctions. J Anat 189:593–600. Mooney DJ, Langer R, Ingber DE. 1995. Cytoskeletal filament assembly and control of cell spreading and function by extracellular matrix. J Cell Sci 108:2311–2320. Nodder S, Martin P. 1997. Wound healing in embryos: a review. Anat Embryol 195:215–228. Peacock A. 1951a. Observations on the pre-natal development of the intervertebral disc in man. J Anat 85:260–274. Peacock A. 1951b. Observations on the postnatal development of the intervertebral disc in man. J Anat 86:162–179. Pena L, Nieto A, Perez Alenza MD, Rodriguez A, Sanchez MA, Castano M. 1994. Expression of fibronectin and its integrin receptor alpha 5 beta 1 in canine mammary tumours. Res Vet Sci 57:358–364. Petroll WM, Cavanagh HD, Barry P, Andrews P, Jester JV. 1993. Quantitative analysis of stress fiber orientation during corneal wound contraction. J Cell Sci 104:353–363. Pfaff M, Liu S, Erle DJ, Ginsberg MH. 1998. Integrin beta cytoplasmic domains differentially bind to cytoskeletal proteins. J Biol Chem 273:6104–6109. Ralphs JR, Benjamin M, Thornett A. 1991. Cell and matrix biology in the rat: a structural and immunocytochemical study of fibrocartilage in a tendon subject to compression. Anat Rec 231:167–177. Ralphs JR, Tyers RNS, Benjamin M. 1992. Development of functionally distinct fibrocartilages at two sites in the quadriceps tendon of the rat: the suprapatella and the attachment of the tendon to the patella. Anat Embryol 185:181–187. 189 Ralphs JR, Banes AJ, Benjamin M. 1998. Tendon cells in vivo contain prominent longitudinal actin stress fibres, linked cell to cell via anchoring junctions. Trans Orthop Res Soc 23:630. Rossino P, Gavazzi I, Timpl R, Aumailly M, Giancotti F, Silengo L, Marchisio PC, Tarone G. 1990. Nerve growth factor induces increased expression of a laminin-binding integrin in rat pheochromocytoma PC12 cells. Exp Cell Res 189:100–108. Rufai A, Benjamin M, Ralphs JR. 1995. The development of the rat intervertebral disc. Anat Embryol 192:53–62. Sechler JL, Schwarzbauer JE. 1997. Coordinated regulation of fibronectin fibril assembly and actin stress fiber formation. Cell Adhes Commun 4:413–424. Sechler JL, Schwarzbauer JE. 1998. Control of cell cycle progression by fibronectin matrix architecture. J Biol Chem 273:25533–25536. Sechler JL, Corbett SA, Schwarzbauer JE. Modulatory roles for integrin activation and the synergy site of fibronectin during matrix assembly. Mol Cell Biol 8:2563–2573. Smith CA, Tuan RS. 1995. Functional involvement of Pax-1 in somite development: somite dysmorphogenesis in chick embryos treated with Pax-1 paired-box antisense oligonucleotide. Teratology 52:333– 345. Sugimoto K, Yoshida K, Fujii S, Takemasa T, Sago H, Yamashita K. 1995. Heterogeneous responsiveness of the in situ rat vascular endothelial cells to mechanical stretching in vitro. Eur J Cell Biol 68:70–77. Takeichi M. 1988. The cadherins: cell–cell adhesion molecules controlling animal morphogenesis. Development 102:639–655. Turnacioglu KK, Sanger JW, Sanger JM. 1998. Sites of monomeric actin incorporation in living PtK2 and REF-52 cells. Cell Motil Cytoskel 40:59–70. Wallin J, Wilting J, Koseki H, Fritsch R, Christ B, Balling R. 1994. The role of Pax-1 in axial skeleton development. Development 120:1109– 1121. Walmsley R. 1953. The development and growth of the intervertebral disc. Edinburgh Med J 60:341–364. White GE, Fujiwara K. 1986. Expression and intracellular distribution of stress fibres in aortic endothelium. J Cell Biol 103:63–70. Winnard RG, Gerstenfeld LC, Toma CD, Franceschi RT. 1995. Fibronectin gene expression, synthesis and accumulation during in vitro differentiation of chicken osteoblasts. J Bone Min Res 10:1969– 1977. Yonemura S, Itoh M, Nagafuchi A, Tsukita S. 1995. Cell-to-cell adherens junction formation and actin filament organisation: similarities between non-polarized fibroblasts and polarized epithelial cells. J Cell Sci 108:127–142. Zanetti NC, Solursh M. 1984. Induction of chondrogenesis in limb mesenchymal cultures by disruption of the actin cytoskeleton. J Cell Biol 99:115–123. Zanetti NC, Solursh M. 1986. Epithelial effects on limb chondrogenesis involve extracellular matrix and cell shape. Dev Biol 113:110– 118.