We have prepared 3T3 cells doubly labeled to visualize simultaneously the extracellular fibronectin (FN) matrix and intracellular actin cytoskeleton in living cell cultures. We used FN-yellow fluorescent protein (FN-yfp) for the FN matrix, and the actin-binding domain of moesin fused to cyan fluorescent protein (cfp-Moe) to stain actin. Actin filament bundles were clearly seen in the protruding lamellae of the cells. FN matrix assembly appeared to be initiated as small spots of FN at the ends of actin filament bundles. The spots then elongated along the actin filament bundle toward the cell center to form FN fibrils. The end of the fibril towards the cell edge appeared immobile, and probably attached to the substrate, whereas the end toward the cell center frequently showed movements, suggesting attachment to the cell. Combining our data with the observations of Pankov et al. we suggest that fibrils grow by stretching this mobile end toward the cell center while adding new FN molecules at the end and along the entire lenght. When the cell culture was treated with cytochalasin to disrupt the actin cytoskeleton, some fibrils contracted substantially, suggesting that the segment attached primarily to the cell surface is stretched.
Fibronectin (FN), an extracellular glycoprotein, is secreted as a disulfide bonded dimer, which can be assembled into matrix fibrils by some cells. These FN fibrils form the primitive extracellular matrix during embryogenesis and wound healing and can be assembled also in cell culture(Hynes, 1990). A gene knockout of FN in mice demonstrated that FN was essential for embryonic development. Embryos without FN died at day 8.5 owing to defects in mesoderm and vasculature (George et al.,1993). More recently, Cre-lox conditional FN knockout mice were generated to study the function of FN in adult tissue(Sakai et al., 2001). That study discovered that plasma FN is not essential for healing of skin wounds but does protect brain cells from ischemic damage.
The FN matrix assembles only on the surface of living cells, in a process that requires integrins and probably other cell surface receptors(Mosher, 1993;Sechler et al., 2000;Wu et al., 1995). Halliday and Tomasek (Halliday and Tomasek,1995) found that cells need to develop tension to assemble a FN matrix, and Zhong et al. (Zhong et al.,1998) provided evidence that stretching exposes a cryptic assembly site near the first FN-III domain. Several groups have demonstrated assembly of FN in vitro into structures resembling matrix fibrils(Baneyx and Vogel, 1999;Brown et al., 1994;Ejim et al., 1993;Mosher and Johnson, 1983;Peters et al., 1998). These assemblies typically require partial denaturing conditions and some shear in the solution.
Our previous study used a FN-green fluorescent protein (gfp) fusion protein to visualize the dynamics of the established FN matrix. We showed that some FN matrix fibrils are highly stretched in living cell culture(Ohashi et al., 1999). However, we were not able to image the cell boundaries and internal structures, and thus we could not easily relate the FN matrix fibrils to the cytoskeleton. In addition, we were not able to reliably image the initial stages of FN fibril assembly.
Hynes and Destree (Hynes and Destree,1978) showed that extracellular FN matrix fibrils colocalized with intracellular actin filament bundles, and Singer(Singer, 1979) visualized the junction of extracellular FN fibrils with cytoplasmic actin by EM. It is now widely recognized that integrins and associated molecules bridge the gap between the extracellular FN matrix and the intracellular actin cytoskeleton(Calderwood et al., 2000). Because of this close association of FN and actin, we wanted to develop the gfp technology to visualize simultaneously the FN matrix and the actin cytoskeleton. Color-shifted gfps are now available for dual labeling, and the combination of yellow and cyan shifted green fluorescent proteins (yfp and cfp) is especially useful (Ellenberg et al., 1999).
There are two ways to visualize the actin cytoskeleton using a gfp tag. One is to prepare a fusion of actin and gfp. However, it has been reported that actin-gfp affects cellular morphology(Ballestrem et al., 1998;Westphal et al., 1997),especially if its expression is more than 30% of the total actin(Westphal et al., 1997). An alternative labeling strategy is to express gfp as a fusion to the actin-binding domain of moesin (gfp-Moe). It was originally expected that gfp-Moe might interfere with actin function, but in previous applications there was no effect on Drosophila development(Edwards et al., 1997), even when gfp-Moe was expressed under the control of a strong, constitutive promoter (Kiehart et al.,2000) or a tissue and temporal specific promoter(Bloor and Kiehart, 2001). Therefore, we chose to use Moe to visualize the actin cytoskeleton. A recent paper has described a similar Moe-gfp construct to visualize actin dynamics,and this also showed no effect on the cells(Litman et al., 2000). In the present study, we created FN-yfp and cfp-Moe expression constructs to visualize simultaneously the FN matrix and actin cytoskeleton. We report here our observations on the initial formation of FN matrix fibrils, how they are associated with actin filament bundles and how they appear to be in a stretched state while attached to the cell.
Materials and Methods
Construction of expression vectors
Diagrams of the FN-yfp and cfp-Moe fusion proteins are shown inFig. 1. The FN-yfp expression vector was constructed by replacing the gfp domain of pAIPFN-gfp with yfp; PCR amplified yfp from the yfp expression vector (EYFP-C1 vector, Clontech) was inserted at a NotI restriction site between FN-III domains 3 and 4(Ohashi et al., 1999). The cfp-Moe expression vector was constructed by inserting the C-terminal segment of Drosophila moesin into the cfp expression vector (ECFP-C1 vector,Clontech). The sequence at the fusion site was (cfp)....ELYKSGLRSRAQASLQDEV...(Moe)...FENM (the cfp and Moe sequences are underlined, and the SGLRSRAQAS sequence is derived from the cloning site). This construct is different from the original gfp-Moe construct, which had the entire coiled-coil domain (Edwards et al.,1997), from the gfp-Moe construct driven by the ubiquitous spaghetti squash promoter (Kiehart et al.,2000) and also different from the construct of Litman et al.(Litman et al., 2000), which had the Moe on the N-terminus of gfp. It is similar to the gfp-Moe designed for tissue-specific expression in fly that is driven by yeast upstream activator sequences (which are expressed in the presence of a yeast GAL4 transgene) (Bloor and Kiehart,2001).
The FN-yfp and cfp-Moe expression vectors (10:1 ratio) were cotransfected with lipofectamine (Gibco-BRL) into NIH 3T3 fibroblasts. The cfp vector has a neo selection cassette, and clones were selected with G418 (0.7 mg/ml,Gibco-BRL). Clones were tested for FN secretion by western blot and for cfp-Moe expression by fluorescence microscopy and then checked for assembly of a FN matrix. Some clones did not assemble a matrix well, but we chose one clone that assembled a substantial FN matrix visible by FN-yfp and showed localization of the cfp-Moe to the actin cytoskeleton.
Western blotting and immunostaining
For western blots, the conditioned media was run on 5% SDS gels and transferred to Immobilon (Millipore) using a semi-dry electroblotter(Multiphor II, Pharmacia LKB). The membrane filters were incubated with polyclonal antibody HB5 at a dilution of 1:1000 (this antibody was generated against human FN and can recognize mouse FN, unpublished observations) or anti-gfp polyclonal antibody (1:1000, Clontech) and horseradish-peroxidase-conjugated secondary antibody at a dilution of 1:1000(BIOSOURCE) and were stained with a diaminobenzedine,H2O2 and NiCl2. The stained membranes were scanned using an Agfa Arcus II scanner, and the images were analyzed by the NIH Image program for quantitative estimation.
For actin staining and immunostaining, cells were fixed with 3.7%formaldehyde in PBS, permeabilized with 0.2% Triton X-100 and then incubated with polyclonal antibody HB5 (1:200) for FN, monoclonal antibody 9EG7 (1:100,Pharmingen) for activated β1 integrin or with rhodamine-phalloidin.(1:2000, Molecular Probes) for actin filaments. The secondary antibodies,rhodamine-labeled antirabbit IgG antibody (1:200, BIOSOURCE) for FN and rhodaminelabeled anti-rat IgG antibody (1:200, PIERCE) for β1 integrin were used.
Cells were maintained in DMEM (Gibco-BRL) with 10% calf serum. Cell suspensions (100 μl; 5×104 cells/ml) were dropped onto 25 mm circular coverslips in 35 mm tissue culture dishes, incubated for 30 to 60 minutes at 37°C, then 2 ml medium (phenol red-free OPTI-MEM including 1%FN-depleted fetal calf serum) was added and the culture maintained for 48-72 hours. The main reason for switching the media was to use a phenol-red-free and HEPES-buffer-based media, which worked very well for monitoring living cell culture in our original study (Ohashi et al., 1999). Coverslips were set into a windowed chamber with 1 ml medium and observed with a Zeiss LSM410 at 37°C(Ohashi et al., 1999). Excitation/emission filter sets 41028 and 31044 (Chroma) were used for yfp and cfp, respectively. The images were recorded with a cooled CCD camera (C4880,HAMAMATSU) and Metamorph software (Universal Imaging Corp.) and were analyzed by Adobe Photoshop and NIH Image computer programs. For superimposition, the digital images were artificially colored green for yfp and red for cfp to obtain higher contrast.
Characterization of the dual labeling
The FN matrix and actin cytoskeleton were visualized by yfp and cfp fluorescence or by the HB5 antibody against FN and by phalloidin-staining for actin (Fig. 2). The yfp fluorescence colocalized with the FN immunostaining, and cfp fluorescence colocalized with the phalloidin-staining actin cytoskeleton. The expression levels of cfp-Moe were variable from cell to cell(Fig. 2b), and the expression was generally lower in confluent culture than in sparse culture. Actin filament bundles were most clearly resolved in the thin protruding regions of the cells. Phalloidin staining of actin correlated very closely with cfp-Moe(Fig. 2d,e) indicating that the cfp-Moe is labeling F-actin polymer, including both actin filament bundles and diffuse actin networks at the leading edge(Fig. 5; strong labeling of the leading edge of pseudopods can be observed). We noted that some cells(∼1%) expressing high levels of cfp-Moe following the initial transfection showed abnormal morphology but after selecting for stable transfectants all cells showed the normal range of actin filament bundle distributions, meaning that the cfp-Moe did not disrupt the cytoskeleton. These results are essentially the same as those previously reported for the distribution of a Moe-gfp construct (Litman et al.,2000).
Unlike our original FN-gfp-transfected CHO cells, which only made a FN matrix after confluence (Ohashi et al.,1999), FN-yfp-transfected 3T3 cells assembled a matrix in sparse cell cultures. Fig. 2f shows a typical pattern of matrix assembly on a spread fibroblast. The FN is arranged as short fibrils about 5 μm long, located over actin filament bundles. The FN fibrils start about 5 μm in from the leading edge. FN fibrils on adjacent actin filament bundles have variable length and are staggered,producing a jagged band of FN parallel to the cell edge.
Western blots of conditioned medium showed that the total amount of FN plus FN-yfp secreted by the selected 3T3 clone was about twice the amount of FN secreted by the wild-type 3T3 cells (Fig. 3a; quantitative analysis of scanned blots also showed this). The FN-yfp (∼275 kDa) was slightly larger than the endogenous FN (∼257 kDa), consistent with the added mass of the yfp domain. The minor degradation band in Fig. 3a is ∼185 kDa. Fig. 3b shows that the FN-yfp secreted by the 3T3 cells is about half the amount of FN-gfp secreted by the transfected CHO cells created previously(Ohashi et al., 1999). Since the CHO cells secrete almost no endogenous FN, the total level of FN plus FN-yfp secreted by the 3T3 cells is about the same as the FN-gfp secreted by the CHO cells. Although only half the FN molecules are labeled in the 3T3 cells, the matrix could easily be imaged by fluorescence(Fig. 2).
Colocalization of FN, integrin and actin
The colocalization of FN matrix, actin cytoskeleton and activated β1 integrin (the antibody we used is specific for the activated integrin) is shown in Fig. 4. FN fibrils are typically localized along actin filament bundles, although much of the actin does not have associated FN. The actin filament bundles are best seen without the superimposed FN fluorescence (Fig. 4c,f). The arrows indicate the same spot in the three panels. Note that the two arrows in 4c indicate the most distal point of the actin filament bundles. There is a small spot of β1 integrin at this point, which is probably at the cell edge. Note in Fig. 4b that FN extends from the integrin patch along the actin filament bundle toward the cell center, but also extends a short distance beyond the cell edge. The integrin patches with FN at the cell edge were not very common. Most of the integrin patches with visible FN were located about 5μm in from the cell edge (Fig. 2f, Fig. 4d,e). Some streaks of activated integrin show no obvious associated FN(Fig. 4b,e), similar to the observations of Pankov et al. (Pankov et al., 2000). There was frequently a diffuse area of integrin staining around the center of the cell.
Initial matrix assembly on cells
The doubly transfected 3T3 cells synthesize FN matrix fibrils on single cells in sparse culture, permitting us to visualize the initial stages of matrix assembly. In addition, the cfp-Moe labeling lets us localize the newly formed FN fibrils relative to the actin cytoskeleton. Five examples of initial matrix assembly are shown in Fig. 5 (the sites are indicated by numbers in the 0:00 frame).Fig. 5 shows six images, and the supplemental movie (jcs.biologists.org/supplemental) shows eight successive images at 30 minutes intervals. In making the movie the eight successive images were carefully aligned using patches of FN that appeared to be fixed to the substrate. Thus all movements observed should represent movements of the cells and associated fibrils. The steps of matrix assembly are most clearly seen by stepping back and forth through the movie one frame at a time. Our description and interpretation are given below.
Site 1 shows the typical initiation of small FN patches over the distal segments of actin filament bundles and their extension along the actin toward the cell center. At 0:00 there is no FN over the actin at site 1. The upper arrow indicates a spot that appears at 0:30. At 1:00 this spot moves toward the cell center, and it is also elongating in the same direction. This movement of the whole FN patch leads to the important conclusion that it is attached to the cell and weakly or not at all to the substrate. After 1:00 this distal segment of the fibril appears fixed and no longer moves, and we conclude that it is now attached to the substrate. The fibril elongates 20μm toward the cell center, for an average growth rate of 8-10 μm/hour. A major growth step occurred from 2:00 to 2:30, when elongation appears to have leap-frogged by initiating a new spot of fibril and leaving a gap between the old and new segments. By 3:00 the gap has been filled and the fibril has increased in brightness, indicating the addition of FN molecules all along its length. Another example of addition of FN along the length is the fibril just above the fibril shown with an arrow. It is barely visible as a small patch at 2:00, but at 2:30 it has extended considerably. At 2:30 it is still quite dim,but at 3:00 and 3:30 it increases in brightness while extending only slightly.
Three other FN fibrils are initiated at site 1 at 0:30 and 1:00. The small FN fibril indicated by the lower arrow at 1:00 again shows movement of the entire fibril toward the cell center from 1:00 to 1:30. From 2:00 the initial segment does not move and appears to be attached to the substrate as the fibril elongates toward the cell center. This fibril also shows a two-part growth: at 2:00 a new segment appears along this same extended fibril, and they subsequently grow together and fuse.
Similar patterns of growth are seen at site 2. The fibril indicated by the upper arrow at 1:00 appears to be fixed at its distal tip, as this does not move in subsequent frames. However the segment of this fibril toward the cell center rotates at 1:30, suggesting that this part of the fibril is attached to the cell and not the substrate. The fibril indicated by the lower arrow is initiated separately as two spots at 30 and 60 minutes. These grow and coalesce into a fibril at 1:30, which shows small movements in subsequent frames. This initiation at two spots, which later grow together and fuse, is similar to the leap-frog growth observed at site 1.
A different pattern of assembly is seen at sites 3, 4 and 5. The first stage is similar to the above, as short FN fibrils are initiated on several adjacent actin filament bundles, forming a zig-zag zone(Fig. 2 shows a similar arrangement). However, in a second stage, these short fibrils appear to coalesce, forming a thick band more or less parallel to the cell edge. The integrity of this band as a fusion of the initial FN segments is indicated by the stretching seen at site 4 from 3:00 to 3:30. Thus the fusion of fibrils initially perpendicular to the cell edge forms a thicker band of FN that is largely parallel to the cell edge.
Finally, we should note the very dramatic movement of two FN fibrils marked by arrows at 0:00. The right hand fibril breaks its attachment at the upper right at 0:30 and contracts. From 1:00 to 3:00 it appears loosely attached to the substrate and shows only small movements. The left hand fibril exhibits substantial movements of its upper part from 0:00 to 2:00, and then from 2:00 to 3:30 the largely invisible cell on the right pulls the middle segment of the fibril about 20 μm to the right, bending and stretching the fibril. The contour length of this fibril increased about 50% from 1:00 to 3:00.
FN fibril movements following cytochalasin treatment
We used cytochalasin B to disrupt the actin cytoskeleton to see how the newly assembled FN fibrils respond when released from the cell.Fig. 6 shows an example in which a matrix fibril is released and contracts to about one third of its original length. The fibrils toward the bottom of the field are apparently attached to the substrate, as they are completely superimposed at the three time points. The vertical fibril indicated by the arrow, and the horizontal fibril to which it is attached, show substantial contraction and movement when the cytoskeleton collapses at 20 minutes.
The cytochalasin experiments were performed 15 times. In every experiment,the majority of the FN fibrils showed no movement, whereas a few fibrils showed a substantial contraction. After cytochalasin treatment, the cells showed a small contraction or shrinkage toward the cell center. On the contrary, the contractions of FN matrix fibrils always occurred in the opposite direction, indicating that the fibril movements are not due to the cell movements.
Fig. 7 shows a cell two hours before, just before and 30 minutes after adding cytochalasin B. The actin cytoskeleton was completely disrupted at the 30 minute time point. Staining for integrins (Fig. 7d) showed that most integrin streaks were dispersed as the actin cytoskeleton was disrupted, but some remained, colocalizing with prominent FN fibrils. FN fibrils displayed two types of behavior. Most fibrils, especially the segments toward the cell edge, appeared rigidly fixed both before and after the addition of cytochalasin, but others showed substantial growth before or contraction after adding cytochalasin. The fibrils that did move showed contractions to a half to a quarter of their initial length. The movements are best seen in the superimpositions of two time frames.Fig. 7a,b shows a superimposition of the FN fibrils at times minus 120 minutes and 0 minutes,showing that some fibrils elongate (seen as a green fibril extended to the right). The segments of these elongating fibrils toward the cell edge are perfectly superimposed at the two time points. Since these segments showed no movement at all, we believe they are attached to the substrate.Fig. 7b,c superimposes FN just before and 30 minutes after adding cytochalasin. Again, most fibril segments,especially those toward the cell edge, are perfectly superimposed, implying attachment to the substrate. However, several fibrils do show substantial contractions, typically of the segment toward the cell center. Examples of contracted fibrils are shown at higher magnification inFig. 7b′, c′. The upper two panels in Fig. 7b′,c′ show little increase in fluorescence for the contracted fibrils, but this may be because the fibril appears to split in two. Most contracted fibrils showed significantly increased fluorescent intensity (the middle and the bottom sets in Fig. 7b′, c′).
A major advance in the present study is the use of dual yfp/cfp labels, so we can visualize simultaneously the extracellular FN matrix and cytoplasmic actin structures that define the cells. Another advance over our previous study is the use of 3T3 cells, which assemble a matrix over individual cells in sparse culture. The CHO cells used in our previous study only assembled a matrix in confluent culture, where it was difficult to locate cell borders. On the 3T3 cells we have been able to see the initial stages of matrix assembly,as FN fibrils appear and grow on the cell surface.
Hynes and Destree (Hynes and Destree,1978) originally used double label immunofluorescence to demonstrate the colocalization of FN and actin filament bundles. Singer(Singer, 1979) visualized the junction of extracellular FN fibrils with actin filament bundles by EM. Both of these studies concluded that the FN was primarily on the bottom (substrate side) of the cells, and our observations agree with this. These previous studies were necessarily static and also could not determine whether the FN fibrils were in early or late stages of development. We have now been able to visualize the initial stages of matrix assembly from the first appearance of a spot of FN to its growth into fibrils and patches.
Our observations suggest general features of FN matrix assembly. Assembly begins as a deposition of a small spot of FN, and this always occurs over the end of an actin filament bundle. These ends stained for activated β1 integrin, in agreement with previous findings(Pankov et al., 2000). The tips of the actin filament bundles that initiate FN fibrils are usually not the ones closest to the cell edge but more mature ones 5-10 μm toward the cell center. However, it is possible that FN assembly begins closer to the leading edge but does not accumulate enough FN-yfp to be visible until cell migration has moved these sites a bit inward. When first initiated the FN patches appear to be attached to the cell and not the substrate, as they can move with the cell (Fig. 5)(Movie at jcs.biologists.org). As the fibril starts to elongate this distal tip typically becomes fixed, as if it were attached to the substrate. The fibril then grows by extending the other end toward the cell center. This segment of FN fibril toward the cell center frequently shows movements suggesting attachment to the cell and not the substrate.
The movements of the initial patch of FN toward the cell center might be due either to movement of the focal contact, as observed by Smilenov et al.(Smilenov et al., 1999) or to translocation of β1 integrins out from a fixed αVβ3 focal contact, as reported by Pankov et al.(Pankov et al., 2000). Since we generally observed only FN and not the β1 integrins we cannot distinguish between these two mechanisms. We observed fibrils to elongate at a rate of 8-10 μm/hour toward the cell center, which is similar to the 6.5-7.2 μm/hour velocity of integrin movements reported previously(Pankov et al., 2000;Smilenov et al., 1999).
The location of the initial patch of FN over the actin filament bundles is clear, but determining where FN molecules are added as the fibril extends is a more difficult question. We can combine our observations with those of Pankov et al. (Pankov et al., 2000)for the following model. Our observations suggest that once the fibril has elongated somewhat, the older segment, toward the cell edge, appears static. It shows no movement and therefore appears to be attached to the substrate. Moreover, it shows no increase in brightness, so there is no indication that these fixed segments are growing. The segment toward the cell center shows both extension and an increase in brightness. This implies that new FN molecules are being added all along the segment. The growth toward the cell center could be due to addition of new molecules at this end of the fibril or to extension of the existing fibril by translocation and stretching. The latter scenario is suggested by Fig. 5D of Pankov et al. (Pankov et al., 2000), where a fibril assembled from labeled FN was allowed to grow for 40 minutes with unlabeled FN. The labeled fibril grew longer and was labeled along its entire new length. This suggests that the existing FN segment was extended by stretching and adding new FN molecules along its length as it was stretched.
Are these growing FN fibrils under tension and stretched? This question is complicated by the tendency of the fibrils to form attachments to the substrate. Thus the majority of fibrils showed no detectable movement in the 30 minutes following cytochalasin treatment. However, a small fraction of fibrils did show movement, usually a segment toward the cell center. Importantly, when a fibril did move following cytochalasin, it always contracted toward the cell edge and usually contracted substantially. This suggests that these fibrils are stretched. By inference all fibrils may be stretched but adhesion to the substrate prevents their contraction.
Movie available on-line
Supported by NIH grants CA47056 to H.P.E.; GM33830 and GM61240 to D.P.K.;and the MFEL Program ONR N00014-94-0818 and AFOSR F49620-00-1-0370.