Stress-relaxation of fibroblasts in collagen matrices triggers ectocytosis of plasma membrane vesicles containing actin, annexins II and VI, and β₁ integrin receptors

The importance of mechanical force for cell function per-tains to a wide range of biological organisms including bac-teria and plants. Recent interest in this subject is indicated by reviews on ‘tensegrity’ (Ingber and Folkman, 1989), ‘mechanogenetic’ (Erdos et al., 1991) and ‘mechanogenic’ (Vandenburgh, 1992) models of cell growth regulation. Even the well-known dependence of cell growth on cell shape (Folkman and Moscona, 1978) may turn out to be an effect of cell tension (Curtis and Seehar, 1978; O’Neill et al., 1990).

We have been studying the effects of mechanical force on fibroblasts in three-dimensional collagen matrix cultures. In these matrices, fibroblasts bind to individual collagen fib-rils (Bellows et al., 1982; Allen and Schor, 1983) and sur-round clusters of fibrils with cell surface extensions (Grin-nell and Lamke, 1984). Binding interactions are mediated, at least in part, by 0’₂[3₁ integrin receptors (Schiro et al., 1991; Klein et al., 1991).

During culture, fibroblasts reorganize and contract the collagen matrix in which they are embedded (Bell et al., 1979). If the matrix is mechanically anchored during con-traction, then mechanical stress develops in response to ten-sion generated by the cells. Actin stress fibers in the cell cytoplasm become oriented along the long axis of the cells (Farsi and Aubin, 1984; Unemori and Werb, 1986), and cells and collagen fibrils become aligned in the same plane (Nakagawa et al., 1989a). Strain gauge measurements show that the force generated by fibroblasts under these con-ditions is comparable to that generated in contracting skin wounds or during tooth eruption (Kasugai et al., 1990; Delvoye et al., 1991; Kolodney and Wysolmerski, 1992).

Several studies have shown that the proliferative capac-ity of fibroblasts in collagen matrices depends in part on mechanical stress. Cells in contracted collagen matrices that are under mechanical stress divide in response to growth factors, but DNA synthesis stops once stress dissipates (Nakagawa et al., 1989a,b; Fukamizu and Grinnell, 1990). When fibroblasts contract floating collagen matrices, mechanical stress does not develop. Under these conditions the cells show low levels of DNA synthesis (Sarber et al., 1981; Van Bockxmeer et al., 1984; Yoshizato et al., 1985) and become arrested in G₀G₁ (Kono et al., 1990). On the other hand, when external mechanical stress is applied to collagen matrices containing fibroblasts, DNA synthesis increases (Jain et al., 1990).

One approach to understanding how mechanical stress regulates fibroblast function is to characterize the initial events that occur when fibroblasts under stress are allowed to relax. Our laboratory and Tomasek’s laboratory inde-pendently introduced a culture model designed to permit this characterization (Mochitate et al., 1991; Tomasek et al., 1992). In this model, fibroblasts in collagen matrices anchored on culture dishes were cultured for 2-5 days, during which time mechanical stress developed. Subse-quently, the contracted matrices were dislodged mechani-cally, thereby initiating stress-relaxation. We report here on a novel secretory process triggered by stress-relaxation. Small (approx. 200 nm diameter), right-side out, plasma membrane vesicles, which contained actin, annexins II and VI, and β₁ integrin receptors, were released from the cells into the collagen matrix. Details of these findings are reported here.

Maintenance of human foreskin fibroblast monolayer cultures and preparation of hydrated collagen matrices from Vitrogen ‘100’ collagen (Celtrix Labs, Palo Alto, CA) have been described pre-viously (Nakagawa et al., 1989a). Briefly, fibroblasts were added to the neutralized collagen solutions (1.5 mg/ml) at a concentration of 5 ×105 cells/ml. Samples (0.2 ml) of the cell/collagen mix-tures were prewarmed to 37°C for 3-4 min and then placed in Costar 24-well culture plates. Each sample occupied an area out-lined by a 12 mm diameter circular score within a well. Gelation required 60 min at 37°C, after which 1.0 ml of Dulbecco’s mod-ified Eagle’s medium (DMEM, GIBCO) supplemented with 10% fetal bovine serum (FBS, Intergen Co, Purchase, NY) and 50 μg/ml ascorbic acid was added to each well. It was important not to add serum until after gel polymerization, otherwise the matri-ces spontaneously dislodged from the underlying culture surface during subsequent culture.

Collagen matrices containing fibroblasts were incubated in a humidified CO₂ incubator for 2 days, during which time the fibrob-lasts contracted the matrix and mechanical stress developed. To initiate stress-relaxation, matrices were gently dislodged from the underlying tissue culture substratum with a spatula.

Contracted matrices (anchored or dislodged) were rinsed twice for 10 min at 22°C with DPBS (see below) minus divalent cations and then incubated for 10 min at 37°C with 0.2 ml of 0.05% trypsin/0.53 mM EDTA solution (GIBCO) followed by 20-30 min with 0.25 ml of collagenase solution (5.0 mg/ml Sigma type I col-lagenase in 130 mM NaCl, 10 mM calcium acetate, 20 mM Hepes, pH 7.2). After cells dispersed completely, enzymatic activity was blocked by the addition of 0.05 ml FBS. Fibroblasts were col-lected by centrifugation at 200 g for 10 min at 22°C (Sorvall GLC-1 centrifuge) and resuspended in DPBS (150 mM NaCl, 3 mM KCl, 1 mM CaCl₂, 0.5 mM MgCl₂, 1 mM KH₂PO₄, 6 Mm Nα₂HPO₄, pH 7.2). Greater than 95% of the cells recovered from anchored or dislodged matrices were viable, as shown by trypan blue exclusion. Moreover, fibroblasts harvested from either matrix were able to attach and spread on culture dishes within several hours and subsequently proliferated at a similar rate (doubling time approx. 24 h).

Alternatively, after dispersing cells from the matrices, enzy-matic activity could be blocked by the addition of proteinase inhibitors leupeptin and pepstatin A (1 μg/ml each), and AEBSF (aminoethylbenzenesulfonyl fluoride, 1 mM; Calbiochem) in DPBS. This resulted in less contaminating serum proteins being present in the vesicle preparations.

After centrifugation, proteinase inhibitors (see above) were also added to the cell-free supernatants. Samples were then centrifuged for 1 h at 4°C at 100,000 g (Beckman L8-M ultracentrifuge), and the high-speed pellets (vesicle fraction) were resuspended in DPBS.

To prepare Triton X-100-soluble and-insoluble fractions of cells and vesicles, samples were mixed with equal volumes of extraction buffer containing 2% Triton X-100, 160 mM KCl, 40 mM imidazole-HCl, 1 μg/ml leupeptin and pepstatin A, 1 mM AEBSF, 20 mM EGTA, and 8 mM sodium azide, pH 7.0, and incubated for 10 min at 4°C (White et al., 1983). The Triton-insol-uble fraction was separated from the soluble fraction by centrifu-gation for 10 min at 4°C at 12,000 g in an Eppendorf microfuge.

Collagen matrices to be studied by transmission or scanning elec-tron microscopy were prepared on solvent-resistent Thermanox plastic coverslips (Nunc Inc., Naperville, IL). Samples for trans-mission electron microscopy were fixed for 1 h at 22°C with 2% glutaraldehyde, 4% paraformaldehyde and 1% tannic acid in 0.1 M sucrose, 0.1 M sodium cacodylate-HCl buffer, pH 7.4, and then washed and post-fixed for 1 h at 22°C in 1% osmium tetroxide with 1.5% potassium ferrocyanide in the same buffer. After thor-ough washing with distilled water, samples were dehydrated with 30% ethanol for 10 min at 22°C and stained en bloc with 2% uranyl acetate in 50% ethanol for 1 h. Subsequently, the samples were dehydrated, impregnated with pure propylene oxide, and embedded in pure Epon 812/Araldite mixture. Thin sections (80 nm) were cut with a Reichert-Jung Ultracut ultramicrotome, col-lected on copper grids without supporting film, and stained with 2% aqueous uranyl acetate for 12 min and Reynold’s lead citrate for 5 min before observation. All observations and photographs were made with a JEOL 1200EX transmission electron micro-scope at 80 kV.

Samples for SEM were fixed as described for TEM and post-fixed for 90 min at room temperature in 1% osmium tetroxide and then passed through two cycles of saturated aqueous thio-carbohydrazide and 1% osmium tetroxide (15 min each step), with thorough washing with distilled water in between steps. Dehydration was accomplished by a series of 5-min ethanol washes: 50%, 70%, 90%, 95% and 100%, followed by two 15-min changes of absolute ethanol. Dehydrated samples were immersed in hexamethyldisilazane (HMDS) for 5 min and air-dried directly from HMDS. Excess HMDS was blotted dry by gently touching the samples with a filter paper. To expose the underside of collagen matrices, samples were dislodged from the coverslips using a spatula. To expose the interior of the colla-gen matrices, portions of collagen were peeled away from anchored matrices using double adhesive Scotch tape. Air-dried samples were mounted on aluminum SEM stubs with colloidal graphite and observed without additional metal coating. Speci-mens were photographed with a JEOL 840 scanning electron microscope at 20 kV.

Collagen matrices to be studied by indirect immunofluorescence microscopy were prepared on LabTek 2-well chamber slides (Nunc). Samples were fixed for 15 min at 22°C with 3% paraformaldehyde in DPBS. Fixed samples were washed twice for 15 min at 22°C with DPBS containing 1% glycine, 1% crystalline bovine serum albumin (BSA) (Sigma) and then permeabilized with 2% Nonidet P-40 in DPBS containing 1% BSA for 10 min at 22°C. To observe the distribution of actin, samples were stained for 2 h at 37°C with either FITC-phalloidin (Molecular Probes Inc., Eugene, OR) or mouse anti-actin monoclonal antibody (Amersham, Arlington Heights, IL), followed by 1 h at 37°C with FITC-conjugated goat anti-mouse IgM (Cappel, Durham, NC). To observe the distribution of tubulin or vimentin, samples were stained for 2 h at 37°C with mouse anti-[3-tubulin monoclonal antibody (a gift from Dr George Bloom, UT Southwestern Med-ical School) or mouse anti-vimentin monoclonal antibody (Amer-sham) followed by 1 h at 37°C with FITC-conjugated goat anti-mouse IgG (Cappel). In some experiments, samples were counter-stained for 30 min at 37°C with TRITC-conjugated wheat germ agglutinin (TRITC-WGA, Molecular Probes Inc.), which permitted identification of cell contours. At the end of the incu-bations, preparations were washed in DPBS, mounted with Mowiol, and observed and photographed with a Zeiss Photomi-croscope III.

Samples of cells and vesicles were dissolved in an SDS-contain-ing buffer (62.5 mM Tris, 2% SDS, 10% glycerol, 0.01% bro-mophenol blue, 5% mercaptoethanol, pH 6.8), and subjected to SDS-polyacrylamide gel electrophoresis (Laemmli, 1970) on a 4% stacking/10% resolving gel using a mini-gel apparatus (BioRad, Richmond, CA). To identify polypeptides by immunoblotting (Towbin et al., 1979), SDS-gels were washed for 30 min at 22°C with 25 mM Tris-HCl, 192 mM glycine, 20% methanol, and polypeptides were transferred onto nitrocellulose filters with a BioRad mini-transfer apparatus using 100 volts for 1 h at 4°C. Nitrocellulose filter strips with immobilized proteins were incu-bated with blocking solution (150 mM NaCl, 20 mM Tris con-taining 0.2% Tween-20 and 1% bovine serum albumin, fraction V) for 1 h at 22°C. The strips were then incubated 2 h at 22°C with mouse monoclonal antibodies against actin, β-tubulin, vimentin (Amersham) or vinculin (Sigma), rat monoclonal anti-β₁ integrin subunit (mAb 13, a generous gift from Dr Ken Yamada, NIH), or mouse monoclonal antibodies against annexins I, II or VI (Zymed Labs, San Francisco, CA). Subsequently, the samples were incubated with alkaline phosphatase-conjugated goat anti-mouse IgM or IgG (BioRad) or goat anti-rat IgG (BioRad) for 1 h at 22°C, and then developed with nitro blue tetra-zolium according to BioRad specifications.

Fibroblasts were metabolically radiolabeled for 2 h with 20-25 μCi/ml [35S]methionine (ICN, specific activity approx. 1300 Ci/mmole) in methionine-free DMEM (GIBCo) supplemented with 2% FBS and 50 μg/ml of ascorbic acid, and then further incubated for 2 h with non-radioactive medium (2 medium changes). Cells and vesicles were prepared from collagen matri-ces and extracted with Triton X-100 as described above. The sam-ples were dissolved in sample buffer and subjected to SDS-PAGE using a 10% resolving gel and a 4% stacking gel. After elec-trophoresis, gels were fixed with 10% TCA, washed with deion-ized water, and impregnated for 30 min at 22°C with 1 M sodium salicylate containing 2% glycerol. Gel films were dried using a BioRad 583 gel drier and exposed to Kodak XAR-2 film.

Fig. 1 shows inside (A,B) and underside views (C,D) of collagen matrices immediately before (A,C) and 5 min after (B,D) the anchored matrices were dislodged from the plas-tic substratum. During the initial 48 h contraction period, fibroblasts had developed an extended, bipolar morphology (A,C). Within 5 min after dislodging the contracted matri-ces from the substratum (B,D), the uniform appearance of the matrix was lost. Large furrows developed as a result of collagen fibril compression when the interconnected net-work of separate collagen fibrils became more densely packed. Fibroblasts (arrowheads) switched from a stretched to a wavy appearance and decreased in length.

The timing of these changes and their uniform appear-ance, both inside and at the surface of the matrix, suggested that mechanical release of tension resulted in a synchro-nous response of the fibroblast population within the matrix. No longer restrained, the cells retracted their pseudopodia, thereby causing rapid reorganization of collagen fibrils to which the pseudopodia were attached. In this case, matrix contraction appeared to result from a cell-shortening process clearly different from the previously described reor-ganization and contraction of anchored collagen matrices that occurs as fibroblasts extend their pseudopodia (Stopak and Harris, 1982; Grinnell and Lamke, 1984; Nishiyama et al., 1988).

Before stress-relaxation, fibroblasts in contracted colla-gen matrices contained prominent actin stress fibers paral-lel to the long axis of the cells (Fig. 2A) as well as micro-tubule and intermediate filament networks (Fig. 2C and E). Previously we reported that actin stress fibers collapsed within 1 h after stress-relaxation (Mochitate et al., 1991). Fig. 2B shows that their disappearance required no more than 5 min, and at this time we observed actin clusters along the cell margins and in the matrix near the cells (Fig. 2B). Comparable clusters of tubulin or vimentin were not seen (Fig. 2D and F). Thirty minutes later, the clusters could still be seen in the matrix but were no longer visible at the cell surface (not shown), suggesting that in response to stress-relaxation, the cells transiently released a portion of their actin.

Further characterization of actin clusters in fibroblasts observed after stress-relaxation was accomplished by study-ing immunostaining of permeabilized and non-permeabi-lized samples with anti-actin antibodies (Fig. 3A,C). The cells were counter-stained with TRITC-WGA to reveal cell contours (Fig. 3B,D). Unless the cells were permeabilized, little actin immunostaining was observed (A vs C), indi-cating that the cells were intact and that the actin clusters were membrane-enclosed. In addition, by comparing actin and WGA staining, it was evident that actin in the matrix near the cells occurred in vesicle-like structures (A,B, arrows). Most vesicles were intact since, without perme-abilization, vesicle-associated actin could not be detected (C,D, boxed area), and the vesicles were right-side-out,since they could be stained with TRITC-WGA with or without permeabilization.

Release of actin-containing vesicles from the cell surface would not be expected to occur by routine exocytosis, but might occur by ectocytosis, i.e. triggered release of right-side-out membrane vesicles (Stein and Luzio, 1991). Ultra-structural observations confirmed this possibility. Fig. 4A shows the typical appearance of mechanically stressed fibroblasts in contracted matrices: elongated morphology, relatively smooth surface, and collagen fibrils (col) closely apposed to the plasma membrane. After the matrices were dislodged, the cell surfaces became more convoluted (Fig. 4B), and small plasma membrane vesicles containing amor-phous material appeared to be budding from the plasma membrane. Some of these approx. 200 nm diameter vesi-cles were still attached to the cells by thin stalks (Fig. 4B, arrows), and some vesicles appeared to be associated with collagen fibrils.

In studies on fibroblast motility, Chen (1981) published electron micrographs showing the formation of similar vesicular structures at the cell surface during retraction of the cell’s trailing edge. Perhaps, because the cells were in monolayer culture, release of the vesicles went unnoticed.

The above studies indicated that stress-relaxation triggered ectocytosis of plasma membrane vesicles. Since the fibrob-lasts were embedded in an extracellular matrix, the vesicles had the potential to accumulate. This possibility was con-firmed, and vesicles were recovered from the collagen matrix by a three-step procedure: enzymatic digestion using a combination of trypsin and collagenase, low-speed cen-trifugation to remove cells, and high-speed centrifugation to separate vesicles from soluble components. Fig. 5 shows that the high-speed pellet obtained by this procedure con-tained a mixture of membrane vesicles and fragments. The vesicles were similar in size and content to those seen bud-ding from the plasma membrane. Membrane fragments, on the other hand, may have been derived from collapsed vesicles.

Cytoskeletal components and β₁ integrins in cell and vesicle fractions were studied by immunoblotting with monoclonal antibodies against actin, tubulin, vimentin, vinculin and β₁ integrin subunit. Fig. 6 shows that the vesicles contained actin and β₁ integrins but excluded tubu-lin, vimentin and vinculin. These results are consistent with the immunofluorescence findings and show that only a selected subset of cellular components sorted into the vesi-cles, and were probably derived from the cell cortex. The presence of β₁ integrin receptors on the vesicles could account for their association with collagen fibrils (see above).

Control experiments (Fig. 7) showed that cells contained tubulin primarily in the Triton-soluble cell fraction and vimentin mostly in the Triton-insoluble cell fraction. Actin was found about equally partitioned in Triton-soluble and-insoluble fractions of cells and vesicles, which indicated that the vesicles contained a mixture of G- and F-actin. The identity of two bands detected with antibodies against β₁ integrin is unknown but they may result from the enzyme treatments used during the harvesting procedure. Immunoblotting of cell extracts prepared from fibroblasts in collagen gels without enzymatic treatment revealed only a single β₁ integrin subunit. Similarly, fibroblasts harvested from monolayer cultures showed primarily the 53 kDa vimentin band (data not shown).

To learn more about the vesicles, studies were carried out with radiolabeled fibroblasts. Cells in mechanically stressed matrices were incubated in medium containing [³⁵S]methio-nine for 2 h and then chased for 2 h in fresh medium. Sub-sequently, the matrices were dislodged. Table 1 shows the results from three separate experiments. At the end of the 2 h chase period, the vesicle fraction averaged 2.0% of the total radioactivity. Five minutes after stress-relaxation, radioactivity found in the vesicle fraction increased to 8.6% of the total. Therefore, taking into account the different incubation times (2 h vs 5 min), the rate of release of radioactivity from the cells increased almost 100-fold after stress-relaxation.

Fig. 8 shows analyses of the radiolabeled cell and vesi-cle fractions by SDS-PAGE and autoradiography. The polypeptide profile of cells from anchored matrices (AC) was essentially identical to that of cells from dislodged matrices (DC), indicating that no cellular proteins were selectively lost as a consequence of stress-relaxation. In vesicles isolated after stress-relaxation (DV), several [³⁵S]methionine-labeled polypeptides appeared to be enriched compared to the cells, in particular those migrat-ing at 24 kDa, 36 kDa (doublet), 45 kDa and 135 kDa. The 24 kDa and 136 kDa components have yet to be identified, but the 36 kDa polypeptide was tentatively identified as annexin II (see below). Except for actin, all of these components partitioned primarily in the Triton-soluble fraction. In the Triton-insoluble fraction of the vesicles actin appeared to be the major component, whereas actin and a 53 kDa polypeptide (vimentin) were the major components of the Triton-insoluble fraction prepared from cells.

Fig. 8 also shows that the polypeptide profile of vesicles from anchored matrices (AV) was essentially identical to that of vesicles from dislodged matrices (DV). This result suggests that ectocytotic vesicles were released at a slow rate from fibroblasts in anchored matrices, perhaps as a byproduct of cell motility, which would be consistent with the quantitative data (Table 1).

Several laboratories have reported release of approx. 200 nm diameter plasma membrane vesicles by chondrocytes (Anderson, 1984). These so-called ‘matrix vesicles’, which induce biomineralization of cartilage matrix, contain prominent 36 kDa (doublet) and 45 kDa components. The 45 kDa component has been identified as actin (Muhlrad et al., 1982; Morris et al., 1992) and the 36 kDa (doublet) com-ponent has been identified as a mixture of annexins (Genge et al., 1992). Annexin VI also has been found in chondro-cyte matrix vesicles (Wu et al., 1992). We considered the possibility, therefore, that the 36 kDa (doublet) polypeptide in plasma membrane vesicles released from fibroblasts might correspond to one of the annexins. Fig. 9 shows sam-ples of cells and vesicles isolated from the collagen matrix and immunoblotted with antibodies against annexins I, II and VI. All three annexins were detected in the cell lysates, and annexins II and VI were found in the vesicles.

To learn about the effects of tension on fibroblast function, we have been studying a stress-relaxation model. Within 5 min after initiating stress-relaxation, fibroblasts retracted their pseudopodia and pulled collagen fibrils together. Although the contractile mechanism accounting for stress-relaxation is unknown, contraction required the presence of a serum factor and could be inhibited by cytochalasin D (Tomasek et al., 1992) or EGTA (Lee and Grinnell, unpub-lished observation), indicating that it is an active cell-medi-ated process rather than a consequence of passive elastic recoil. Stress-relaxation resulted in disappearance rather than shortening of stress fibers, and therefore did not appear to result from a myofibril-like contraction (Isenberg et al., 1976; Kreis and Birchmeier, 1980; Burridge, 1981).

At the same time as the disappearance of stress fibers, clusters of actin occurred along the cell margins and in the matrix near the cells. The actin did not appear to be released from dying cells, but rather was found inside vesicles that were budding from the cell surface. These regularly shaped, approx. 200 nm diameter vesicles were membrane-enclosed, right-side-out and contained roughly similar pro-portions of G- and F-actin. After 5 min, some plasma mem-brane vesicles were connected to the plasma membrane by thin stalks; others could be found in the collagen matrix, often bound to collagen fibrils. By 30 min, vesicles were no longer observed on the cell surface but could still be found in the matrix, suggesting that the budding process triggered by stress-relaxation was transient.

Taken together, these features indicate that budding trig-gered by stress-relaxation is an example of what has been called cellular ‘ectocytosis’ (Stein and Luzio, 1991). Notic-ing fibroblast ectocytosis would have been difficult had the cells been in monolayer culture because the vesicles would have been released into the medium. As already mentioned, Chen (1981) published electron micrographs showing the formation of small, vesicular structures on the cell surface during retraction of the cell’s trailing edge, but release of vesicles went unnoticed. Having the fibroblasts in a colla-gen matrix, however, allowed the vesicles to accumulate. Although the rate of vesicle released increased dramatically during stress-relaxation, a slow release of vesicles also appeared to occur in anchored matrices, perhaps as a byproduct of cell migratory activity.

Analysis of ectocytotic vesicles isolated from the matrix after stress-relaxation showed that they had a very specific cytoskeletal composition, i.e. they contained actin but not tubulin, vimentin or vinculin. These results suggested that the vesicles were selectively derived from the cell cortex. The vesicles also contained annexins II and annexin VI, and annexins normally are localized in the cell cortical region (Semich et al., 1989; Nakata et al., 1990). Annexin II has been implicated in the membrane fusion event during exo-cytosis (Ali et al., 1989; Zaks and Creutz, 1990) and annexin VI has been implicated in coated vesicle formation during receptor-mediated endocytosis (Lin et al., 1992), so the annexins may play a role in the membrane reorganiza-tion required for re-sealing ectocytotic vesicles after they form.

Previous studies indicated that ectocytosis may result from disruption of the cell’s cortical cytoskeleton (reviewed by Gores et al., 1990). One possibility, suggested for ago-nist-induced ectocytosis of platelet membrane vesicles, is that activation of the cytoplasmic proteinase calpain results in degradation of the cortical cytoskeleton components that stabilize the plasma membrane (Fox et al., 1990), but this theory has been controversial (Wiedmer et al., 1990). In our studies we did not observe differences in high molecular mass components in cells harvested from collagen matrices before and after stress-relaxation.

Another possibility is that the vesicles represent sites at which cells were attached to collagen fibrils before stress-relaxation, sites that were released as the cells retracted, and which then re-sealed to form vesicles. Association of β₁ integrins with the vesicles could account for their adhe-sion to collagen (Schiro et al., 1991; Klein et al., 1991). Also, both annexins II and VI have been reported to be col-lagen-binding proteins (Wirl et al., 1990; Wu et al., 1992). Regardless of whether the vesicles are released and then bind to collagen fibrils or are sites of collagen fibril attach-ment that re-seal after release, the ectocytotic mechanism provides a means by which cells can export cytoplasmic proteins that lack a signal sequence. An analogous mecha-nism has been suggested to occur during the normal course of myoblast development (Cooper and Barondes, 1990).

In the case of chondrocytes, ectocytosis results in release of approx. 200 nm ‘matrix vesicles’ that are involved in matrix remodeling, i.e. these vesicles initiate the biominer-alization process (Anderson, 1984). Release of matrix vesi-cles depends on actin depolymerization (Hale and Wuthier, 1987), and the isolated vesicles have been found to contain prominent 45 kDa and 36 kDa (doublet) components iden-tified as actin (Muhlrad et al., 1982; Morris et al., 1992) and a mixture of annexins (Genge et al., 1992). Matrix vesi-cles also contain annexin VI (Wu et al., 1992).

The physiological function of fibroblast matrix vesicles is probably unrelated to mineralization because, unlike chondrocyte matrix vesicles, fibroblast vesicles do not con-tain detectable alkaline phosphatase (Lee and Grinnell, unpublished observation). One possibility is that fibroblast matrix vesicles play a role in tissue remodeling that begins after the contraction phase of wound repair (Peacock, 1984; Clark, 1985). Vesicles released by platelets have been shown to regulate hemostasis by activating the enzyme pro-thrombinase (Fox et al., 1990; Wiedmer et al., 1990). Vesi-cles released from fibroblasts may play an analogous role in activation of metalloproteinases (Mignatti et al., 1988; He et al., 1989). Interestingly, fibroblasts show activation of procollagenase in vitro after contraction of collagen matrices (Mauch et al., 1989; Nakagawa et al., 1989b). Future studies with isolated vesicles should provide insights into this possibility.

These studies were supported by NIH grant GM31321. We are indebted to Dr William Snell for his advice during the course of this research and to Drs Richard Anderson and Ellis Golub for their help during preparation of the manuscript.