Integration of single and multicellular wound responses

Results

Single and multicellular wound repair are generally considered separately, with the former being viewed in terms of plasma membrane resealing (11), and the latter being viewed in terms of cell migration (12). However, there is one area of striking overlap: wounds (3-10, 13) or apoptosis (14) in simple epithelia of a variety of systems trigger formation and closure of rings of actin filaments (F-actin) and myosin-2 (actomyosin “purse strings”) that close the hole in the epithelium, similar to the formation and closure of a ring of F-actin and myosin-2 over wounds made in Xenopus oocytes (1,15, 16). Whether the similarity of the two responses is anything more than superficial is unknown, although in both cases the small GTPases Rho and Cdc42 are essential for the wound response (2, 5, 8, 9, 13,14).

To explore the relationship between the single and the multicellular purse string wound response, we utilized Xenopus embryos, which progress from large, relatively undifferentiated cells (blastomeres) to much smaller, relatively differentiated cells (embryonic epithelia) within several hours. Embryos were labeled with eGFP-rGBD, a probe for active Rho (2) and then laser wounded. When blastomeres were wounded near the cell center, a zone of active Rho rapidly formed and closed around the wounds (Fig. 1A), consistent with results from oocytes (2). Strikingly, when wounds were made near cell-cell boundaries, Rho was activated not only around the wound itself, but also at the nearest cell-cell contacts (Fig. 1B). Following Rho activation, the cell-cell contact moved inward, eventually merging with the zone of active Rho around the wound (Fig. 1B). To confirm that the apparent Rho activation in neighboring cells did not simply reflect high levels of signal concentrating at the site of cell-cell contact in wounded cells, embryos were mosaically labeled eGFP-rGBD and lightly labeled cells adjacent to heavily labeled cells were wounded. Neighbor cells clearly activated Rho at the nearest cell-cell contact and, again, the region of high Rho activity moved inward toward the zone of Rho activity around the wound (Fig. 1C; movie 1). By varying the position or size of the wound, it was possible to elicit Rho activation at two (Fig. 1D; movie 2), three (Fig. S1), four (Fig. S2) or more cell-cell contacts. Rho activation at cell-cell contacts appeared to occur at least as quickly as activation around the wounds themselves (Fig. 1B), suggesting that cell-cell contact Rho activation works in parallel with the wound edge response rather than being downstream of wound edge signaling. Consistent with this notion, the onset of Rho activity at contacts began at 12.4 ± 1.4 s after wounding, while the rise at the wound edge began at 20.5 ± 1.7 s (mean ± SEM; p < 0.001, n = 14). The relative autonomy of cell-cell contact Rho activation was underscored by experiments in which the contacts themselves were subjected to laser wounding: Rho activation clearly ensued at the contact remnants beyond the wound before the rest of the cortex responded (Fig. S3).

The ingression of the cell-cell contacts toward the wound implied the involvement of F-actin, an essential component of contractile arrays. To test this point, embryos were microinjected with both eGFP-rGBD and mCherry-UtrCH, a probe for F-actin (12). In mosaically labeled embryos, F-actin accumulated at cell-cell contacts shortly after Rho accumulation and moved inward (Fig. 2A; movie 3), confirming the involvement of F-actin. Double label z-movies revealed that Rho activation and F-actin accumulation occur first at the adherens junctions and then spread basally as the response develops (Fig. S4). More detailed analysis of the movements of cell-cell junctions revealed that initially responding junctions ingressed quickly toward the wound edge purse string and then slowed as the two start to overlap, forming the hybrid purse string (Fig. S5). From that point onward, the two display mechanical continuity, such that the edges of the wound purse string move toward the cell-cell contacts, resulting in a final center point of the purse string that is shifted relative to the original center point (Fig. S5). In contrast, in the absence of a junctional response, the closing purse string remains focused on the original wound site (Fig. S5).

The above observations suggested that the junctional wound response could maintain epithelial integrity following wounding and that it could integrate the single and multicellular wound responses. We initially sought to test the first idea by microinjecting blastomeres with C3 exotransferase to inhibit Rho. However, upon wounding, junctions in C3 injected cells dissolved (not shown) presumably because Rho activity is needed for proper junction formation (18). Instead, we compared the Rho activation response of different embryos to wounding. In 44/48 cases, the neighbor cell responded to wounding via Rho activation, and in all of these, tight contact between the neighbor and wounded cell was maintained throughout healing (e.g. Figure 1B-D; 2A). In the remaining 4 cases, wherein the neighbor cell failed to activate Rho, the edge of the wounded cell was pulled away from the neighbor cell (Fig. S6). Thus, there is a tight correlation between the junctional response and maintenance of cell-cell contacts during healing. The second idea was tested by completely ablating cells rather than wounding them. This resulted in active Rho and F-actin accumulating at cell-cell junctions around the hole in the epithelium created by ablation, forming a multicellular purse string that then closed over the hole, resealing the epithelium (Fig. 2B; movie 4). Thus, the single cell purse string, the hybrid purse string, and the multicellular purse string represent points on a continuum that are integrated by the junctional response.

Single cell wounds also elicit local Cdc42 activation (2), and multicellular purse string healing in embryos is Cdc42-dependent (8,9). We therefore examined Cdc42 activity using a probe for active Cdc42 (eGFP-wGBD; 2). Active Cdc42 concentrated around both the wounded cell and at cell-cell junctions (Fig. 2C,D). Because myosin-2 is the motor that drives closure of single cell (15) and multicellular (13) wound purse strings, the distribution of phosphorylated myosin-2 regulatory light chain (P-RMLC) was also analyzed using a phosphospecific antibody. In wounds made distal to neighbor cells, active myosin-2 is confined to the wound purse string, while in those made near neighbor cells, active myosin-2 accumulated not only around the wound but also at nearby cell-cell contacts (Fig. 2E). Curiously, comparison of the distribution of F-actin and active myosin-2 at the ingressing junctions showed that while a large degree of overlap between the two was evident, active myosin-2 was more abundant at the leading edge of the array than the trailing edge (Fig. 2F). Similarly, active Rho is enriched at the leading edge, while Cdc42 is enriched at the trailing edge (Fig. 2F), patterns which mimic those around the wound itself (not shown but see ref. 2). Analysis of fixed samples also confirmed that F-actin and active myosin-2 accumulate at the adherens junction and spread basally (Fig. S7).

The results suggested that the major determinant of neighbor cell response was distance between the wound and the nearest neighbor. Consistent with this hypothesis, quantification of the frequency of neighbor cell Rho activation relative to time after fertilization and distance to the wound showed that there is little correlation between time after fertilization and the likelihood of neighbor cell response, but there is a clear distance dependence (Fig. 3A). Specifically, wounds made within 50-60μm of a cell-cell boundary often result in Rho activation in the neighboring cell at that boundary, while wounds made more distal to cell-cell boundaries typically fail to elicit Rho activation at the boundary. Further, when individual cells were subjected to multiple wounds, either near or far from cell-cell boundaries, wounds placed near cell-cell boundaries elicit Rho activation in neighbors, while wounds in the same cell distal from cell-cell boundaries do not (Fig. 3B). Finally, as development proceeded, and cells become progressively smaller due to cleavage, cells not immediately bordering the wounded cell displayed Rho activation at their junctions (Fig. 3C).

The single cell resealing (19, 20) and purse string (1,2) repair responses are dependent on the influx of calcium into wounded cells from the extracellular medium. To determine whether calcium inrush is also required for the healing response in embryos, blastomeres were wounded in calcium-free medium. This manipulation prevented Rho activation around the wound, and reduced but did not prevent Rho activation at cell-cell borders (Fig. 4A). However, these results should be interpreted with caution, as extracellular calcium is required for proper junction function. To determine whether changes in intracellular free calcium might be associated with junctional signaling after wounding, embryos were injected with Fluo-3, a calcium reporter (21). Wounding triggered an increase in intracellular free calcium around the wound site as well as in neighboring cells as quickly as could be imaged (within 1s of wounding; Fig. 4B-C; movies 5-6). In contrast to Rho and Cdc42, the pattern of increased calcium detected with Fluo-3 was relatively variable: in some cases, calcium elevation in the neighbor cells was not obviously focused at the cell-cell contacts (Fig. 4B,C, S8), while in other cases it was focused at cell-contacts (Fig. S9, S10; see also below). In those cases where a clear increase in calcium was evident at cell-cell contacts (e.g. Fig. S9, S10), it occurred just as quickly as around the wound itself, consistent with the junctions acting as local platforms for calcium signaling.

Much of the difficulty in interpretation of the experiments performed with Fluo-3 stemmed from some of the features of this probe in this system. Specifically, the extremely intense cytoplasmic signal increase following wounding often obscured any potential details of increased calcium in the region of the plasma membrane and cell-cell junctions (e.g. Fig. 4B). In addition, the dye progressively accumulated in organelles, which prevented imaging after several hours and even prior to that resulted in the artifactual appearance of calcium movement as organelles were swept to the wound via cortical flow (e.g. movie 5). Finally, the emission properties of Fluo-3 prevented its use in double labeling experiments with eGFP-rGBD or eGFP-wGBD.

To circumvent these problems, we expressed the C2 domain of Xenopus protein kinase C-β fused to mRFP (mRFP-C2). This C2 domain binds to the plasma membrane phospholipid, phosphatidylserine, but only in the presence of elevated calcium and has thus been employed as a protein-based probe for calcium elevation in cultured mammalian cells in vivo(22). These results have been confirmed in the Xenopus system, where mRFP-C2 was shown to target to the plasma membrane in response to elevation of intracellular calcium (23).

Consistent with the results obtained with Fluo-3, mRFP-C2 was recruited as quickly as could be imaged to the plasma membrane immediately around the wound (Fig. 4D). More importantly, mRFP-C2 consistently displayed recruitment to nearby cell-cell contacts (Fig. 4D,E, movie 7) at least as quickly as the recruitment to the wound edge itself (wound: 0.51s ± 1.7s; mean ± SD; n = 35; junctions: 0.26s ± 1.45s, mean ± SD, n=113; p = .19). Further, as with active Rho and Cdc42, mRFP-C2 accumulated at cell-cell contacts that did not immediately border the wounded cell (Fig. 4F; S11). Unlike the accumulation of active Rho and Cdc42, which occurs ∼10 s after wounding, the appearance of mRFP-C2 at distal cell-cell contacts was detectable immediately upon wounding.

The accumulation of mRFP-C2 along cell-cell contacts was typically very precise, spreading only a limited distance along the cell contacts (Fig. 4D,E). The potential importance of this observation was revealed by the fact that the regions of cell-cell contacts that ingressed toward the wound site were heavily labeled with mRFP-C2 (Fig. 4D,E; movie 7; 73/74 ingressing cell contacts showed elevated mRFP-C2 recruitment prior to ingression), as observed above for active Rho and Cdc42. Indeed, direct comparison between the distribution of mRFP-C2 and the active GTPases revealed a striking correlation: the lateral pattern of accumulation of mRFP-C2 at cell-cell contacts almost exactly predicted the subsequent pattern of active Cdc42 and Rho accumulation, based on both inspection (Fig. 4G, S12) and line-scan intensity measurements (Fig. S13, S14).

Discussion

The results show that the response of single cells to damage is linked to the epithelial response to damage via actomyosin purse strings as a continuum, going from a strict single cell purse string, to a hybrid purse string comprised of both the wounded cell and junctions contributed by one or more neighbors, to a multicellular purse string closing a hole in the epithelium. The results also show that cell-cell junctions serve as signaling platforms for Rho, Cdc42, and calcium and apparently do so in parallel with the signaling machinery around the wound, rather than being directly dependent on it. This assertion is based on the fact that elevation of calcium and activation of Rho and Cdc42 can occur at junctions that do not directly border the wound, and the fact that stimulation of these signaling players at the junctions occurs as fast or faster than it does around the wound. The immediate implication of these findings is that evolution could have very simply converted the single cell wound repair response to the multicellular response by endowing the cell-cell junctions with the ability to generate the same signals triggered by plasma membrane damage: calcium elevation, Rho activation, and Cdc42 activation. We assume that some of the intermediate players will differ, but it is nevertheless striking that the same temporal pattern holds: an immediate calcium increase both around the wound and at the junctions, followed by activation of Rho and Cdc42 10-20 s later. It is also striking that the same spatial pattern holds: Rho and active myosin-2 concentrate on the leading edge of the wound and the ingressing junction; active Cdc42 on the trailing edge (see also 1,2, 15).

While this represents, to the best of our knowledge, the first demonstration of Rho and Cdc42 activation at cell-cell junctions in response to a specific signal, there is good reason to believe that the Rho GTPases are subject to local regulation at adherens junctions in a variety of systems (24); we therefore think it likely that the wound induced signaling observed here will prove to be a conserved feature of epithelia. Indeed, given the evidence that Rho activation is a feature of junction assembly (18), it is plausible that further elevation of Rho activity “beefs up” not only F-actin and myosin-2 at the junction, but also other components, allowing it to resist the forces exerted upon it by the healing process.

It remains to be determined how the signal is sent from the site of damage to the neighbor cells and beyond. One can imagine three general mechanisms: transmission of the signal from the wound site via gap junctions; paracellular signaling via release of cytoplasm from the wound; activation of membrane channels via tension changes. We think it unlikely that gap junctions are essential (although they could contribute) since it is hard to imagine how a signal (calcium elevation) could travel ∼40 um from the wound site, through gap junctions, and then another 40 um into a neighbor cell in the space of a second or less (e.g. Figs. 4A,F, S11). Furthermore, gap junctional transmission cannot explain why junctional Rho activation is observed when wounds are made in the absence of external calcium. In addition, treatment with two different gap junction inhibitors failed to prevent either calcium elevation or Rho activation in neighbor cells, except at concentrations that were obviously toxic (unpublished results). Paracellular signaling also seems unlikely (although again, it could contribute), since the pattern of calcium elevation frequently appears noncircular (e.g. Fig. 4F, S9,S10), which is not what would be expected for a signal diffusing from the wound outside the embryo. Thus, a mechanism based on wound-induced tension changes appears the most plausible. It has the virtue of explaining the remarkable speed of the initial response, the distance dependence, and the fact that the junctions can apparently function autonomously of the signals around the wound. Testing this hypothesis will require development of the means to specifically prevent tension changes during wounding and/or the means to generate a local change in tension without wounding.

Experimental Procedures

Supplementary Material