The small GTPase Rac induces actin polymerization, membrane ruffling, and focal contact formation in cultured single cells1, but can either repress or stimulate motility in epithelial cells depending on the conidtions2, 3. Therefore the role of Rac in collective epithelial cell movements in vivo, which are important for both morphogenesis and metastasis4-7, is difficult to predict. Recently photoactivatable analogs of Rac (PA-Rac) have been developed, allowing rapid and reversible activation or inactivation of Rac using light8. In cultured single cells, light-activated Rac leads to focal membrane ruffling, protrusion, and migration. Here we show that focal activation of Rac is also sufficient to polarize an entire group of cells in vivo, specifically the border cells of the Drosophila ovary. Moreover activation, or inactivation, of Rac in one cell of the cluster caused a dramatic response in the other cells, suggesting that the cells sense direction as a group based on relative levels of Rac activity. Communication between cells of the cluster required Jun N-terminal kinase (JNK) but not guidance receptor signaling. These studies further show that photoactivatable proteins are effective tools in vivo.
Border cells are a group of 6-8 cells that arise from the monolayer of ~650 epithelial follicle cells that surround 15 nurse cells and one oocyte in a structure called an egg chamber (Figure 1a-c). Border cells migrate ~175 μm in between the nurse cells, as an interconnected group of two distinct cell types: 4-8 migratory cells surround two central polar cells (Figure 1d-i, k). Polar cells cannot migrate but secrete a cytokine that activates the JAK/STAT pathway rendering the outer cells motile9. The outer cells carry the polar cells and lose the ability to move in the absence of continuous JAK/STAT activation10. Thus each cell type requires the other. Border cells also require steroid hormone, receptor tyrosine kinase, Notch, and other signaling cascades11-16. Thus border cells experience a rich and complex signaling environment, as do most cells in vivo.
The requirement for Rac in border cell migration was one of the earliest demonstrations of its role in cell motility in vivo17. Expression of either dominant-negative or constitutively active Rac impedes migration13, 17, 18, suggesting that its activity must be spatially and/or temporally controlled. However the precise function of Rac remains unclear.
To evaluate the effect of locally activating Rac in border cells we generated transgenic flies expressing the photoactivatable form of Rac (PA-RacQ61L) tagged with mCherry, under control of the Gal4/UAS system. When expressed in border cells using slbo-Gal4, the protein was distributed throughout the cells, in the cytoplasm, nuclei, and at cell surfaces (Figure 1g-i). In the absence of laser illumination, border cell migration was normal (Figure S1; Movies S1 and S2).
Upon exposure to repeated pulses of laser light, border cell migration could be redirected (Figure 1l-t; Movie S3). In this example, border cells were migrating along the path designated by the solid arrow and the leading cell extended a prominent forward-directed protrusion. The laser was applied to the cell next to the leading cell, which did not exhibit any detectable protrusion at the time. Following illumination the cluster retracted the original forward protrusion, changed direction and began moving to the side, a behavior never observed in wild-type19, 20. Light pulses were delivered once per minute due to the reversibility of PA-Rac8. The border cells reached the side of the egg chamber after ~60 minutes (Figure 1m,n and Movie S3). Although light pulses were continuously delivered, the cluster did not move further down the side of the egg chamber over the next 20 minutes (Figure 1o-q), suggesting there might be a barrier or repellent in this region. When we shifted the site of illumination toward the center of the egg chamber (Figure 1r), the cells responded by moving in that direction (Figure 1s,t; Movie S3). A single amino acid substitution in the LOV domain (C450M) renders the protein light-insensitive8 and this construct could not redirect border cell migration even in the presence of light (Figure 1u-w; Movie S4).
To determine if Rac activity was only required in the lead cell, we co-expressed dominant-negative Rac (RacT17N) together with PA-RacQ61L in all border cells and photoactivated Rac in one cell. RacT17N alone strongly inhibits border cell motility18 and photoactivation of Rac in the front cell failed to promote forward movement of the cluster in this background (Figure 1g-i). However, activating Rac in approximately half of the cells in the cluster caused them to move forward, albeit very slowly (Figure 1j-m). These results suggest that each cell requires some Rac activity for motility, and each cell contributes to the migration speed of the cluster, but the highest level of Rac activation determines the direction of movement.
We then tested whether PA-RacQ61L was sufficient to cause border cells to move in a direction opposite to their normal movement (Figure 2). Border cells expressing PA-RacQ61L were first driven in the normal direction, and prominent lamellipodia-like protrusion was evident at the site of illumination (Movie S5). Then, we illuminated the rear. Front protrusion ceased rapidly (Movie S5) but rearward movement was initially very slow. After a variable delay, clusters moved backwards (Figure 2d-i), sometimes reconnecting with a follicle cell within the epithelium (Movie S5). In contrast, the light-insensitive control protein did not reverse the migration direction (Movie S6). On average, the PA-Rac-induced forward migration speed exceeded the reverse migration speed by 4.5-fold (Figure 2p; PA-Rac Front vs. PA-Rac Back), suggesting an influence of endogenous directional signaling on the behavior induced by PA-Rac.
To explore the interaction between endogenous signals and PA-Rac, we compared the responses of wild-type cells to those of cells with reduced guidance receptor activity. PVR and EGFR are receptor tyrosine kinases that function redundantly to guide migrating border cells13-15. Border cells expressing dominant-negative forms of both guidance receptors, PVRDN and EGFRDN, extend protrusions in all directions and make little forward progress20. PA-Rac rescued both the morphological defect and directional movement in this genotype (Figure 2j-l), consistent with the idea that Rac normally functions downstream of the receptors to determine the direction of movement. When clusters were illuminated at the front, the cells moved forward (Figure 2j-l). When the same clusters were illuminated at the back, rearward movement (Figure 2m-o) resulted. In contrast to the responses of wild-type clusters, average forward and reverse migration speeds were indistinguishable in border cells expressing PVRDN and EGFRDN (Figure 2p), supporting the idea of competition between endogenous guidance receptor signaling and PA-Rac induced directionality.
After stimulating rearward protrusion, we stopped illuminating and observed the recovery (Figure S2). Both wild-type and PVRDN- and EGFRDN-expressing clusters rapidly protruded in response to rear illumination and retracted the rearward protrusion following cessation of the light. However, wild-type cells protruded less and retracted more (Figure S2q). Over longer time courses, wild-type cells typically stalled after cessation of rear illumination but eventually recovered movement in the normal forward direction (Figure S3a-m). In contrast, PVRDN and EGFRDN-expressing clusters failed to recover forward movement (Figure S3m-y). These results also suggest that endogenous PVR and EGFR signals compete with PA-RacQ61L-induced polarization.
The inability of PA-RacQ61L to cause border cells to move down the side of the egg chamber led us to probe the microenvironment further. Within the anterior ~1/3 of their normal travel path, focal Rac activation could steer border cells from the center path all the way to the follicle cells, or along the perimeter of the egg chamber (Figure 3a-c and m). However if we treated cells after they reached the center of the egg chamber, they could not be redirected to the follicle cell layer (Figure 3d-i and m), although PA-RacQ61L could still move them forwards or backwards. Within the posterior 1/3 of their normal path, the cells could again be directed off their normal course, in between the nurse cells (Figure 3j-l and m). A summary of the responses to PA-RacQ61L is shown in Figure 3m. Thus there are regions in the egg chamber that actively repel the border cells or lack important structural or chemical substrates for migration, suggesting that there is additional guidance information besides the ligands for PVR and EGFR.
PA-RacQ61L was also insufficient to cause border cells to migrate earlier than normal, possibly because high levels of JAK/STAT signaling, which are required for the border cells to initiate movement, are not achieved at earlier time points21, 22. Consistent with this, PA-RacQ61L did not cause protrusion or migration in border cells expressing a dominant-negative form of the receptor Domeless, which is required for STAT activation (Figure S4a-f). A key downstream target of STAT is the transcription factor Slow Border Cells (SLBO)23, and slbo mutant border cells cannot extend protrusions. However PA-RacQ61L could not rescue protrusion or migration in slbo mutants (Figure S4g-l). PA-RacQ61L also failed to rescue guidance receptor deficiency after stage 10 (not shown). Together these findings demonstrate that PA-RacQ61L reveals temporal as well as spatial constraints on migrating cells.
To evaluate the effects of locally inhibiting Rac, we generated transgenic flies expressing PA dominant-negative Rac (UAS-PA-RacT17N). Illuminating the leading border cell arrested migration and, strikingly, led to protrusion at the cluster rear (Figure S5a-j; Movie S7). In contrast illumination of the rear of the cluster enhanced forward protrusion (Figure S5k-m) and migration (Figure 2p). The magnitude of the effect was smaller in PVRDN and EGFRDN-expressing cells (Figure S5n-p; Figure 2p).
The non-autonomous effects of PA-RacQ61L and PA-RacT17N were striking so we examined the morphological consequences at higher magnification. Specifically, activation of Rac in one cell of either a wild-type cluster (Figure 4a-c) or a cluster expressing PVRDN and EGFRDN (Figure 4d-f) resulted in retraction of protrusions by the other cells and movement of the cluster in the direction of the light. This was true whether the illumination was provided at the front of the cluster (not shown) or at the back. Strikingly PA-RacT17N had precisely the opposite effect in a polarized wild-type cluster (Figure 4j-l). Focal inhibition of Rac in the protruding lead cell caused a loss of polarization and random protrusion of all the cells in the cluster (Figure 4l).
To quantify these results, we developed an automated method to count the number of protruding cells (Figure S6) and calculated the directionality index, which measures the degree of polarization of the cell cluster20. PA-RacQ61L treatment rescued the PVRDN,EGFRDN polarization and the number of protruding cells nearly to wild-type (Figure 4m,n).
Inhibition of the JNK pathway also reduces the directionality index24 (Figure 4n). The JNK pathway helps to coordinate border cell movement by promoting cohesion between border cells. To test the hypothesis that cluster cohesion is important for the non-autonomous effects of Rac, we monitored the effect of PA-RacQ61L in cells with reduced JNK signaling. Photoactivation of Rac at the back of clusters with impaired JNK signaling did not cause retraction of protrusions that were extended in other directions and resulted in little net movement of the cluster (Figure 4g-i). The same effect was observed whether JNK signaling was reduced by expression of the JNK phosphatase Puckered (UAS-Puc2A) or by expressing a dominant-negative form of the kinase (not shown).
The inability of PA-RacQ61L to rescue the JNK knockdown phenotype could have been because JNK signaling is required autonomously downstream of Rac to generate lamellipodial protrusion. However PA-RacQ61L induced autonomous cell protrusion in the direction of illumination, even in cells over-expressing puc2A (Figure 4g-i). Therefore JNK signaling is not required downstream of Rac to promote protrusion, consistent with the published observation that reduction of JNK signaling does not lead to reduced protrusion24. Together these results suggest that JNK signaling is required for the non-autonomous propagation of directional information from the cell with highest Rac activity to the other cells of the cluster. This could be due to direct mechanical coupling of the cells or via signaling pathways downstream of adhesion receptors or both.
Our results suggested that Rac is normally active in all the cells of the cluster, that the leading cell has a higher level of Rac activity, and this asymmetry is lost in PVRDN- and EGFRDN-expressing cells. To test this we took advantage of a Rac fluorescence resonance energy transfer (FRET) biosensor25. When expressed in Drosophila S2 cells, biosensor activity increased in response to EGF stimulation, and the increase was blocked by co-expression of dominant-negative Rac (Figure S7a-k). We generated transgenic flies expressing the biosensor under the control of Gal4/UAS. When expressed with slbo-Gal4, we consistently observed a FRET signal in border cells (Figure S7l,m), and this was dramatically reduced upon co-expression of dominant-negative Rac (Figure S7n,o). Moreover the signal within the border cell cluster was asymmetric and appeared highest in elongating protrusions, which were most prominent in the leading cell (Figure 5a-e). This FRET signal was inhibited by co-expression of RacT17N (Figure 5l). To quantify the asymmetry we divided the border cell cluster into 30 sectors (where sector 0 represents the front of the cluster and −15 and +15 represent the rearmost sector), and measured the FRET efficiency in each sector for more than 30 clusters (Figure 5f-g). As predicted, the Rac activity was highest at the front (between sectors −5 and +5) and lowest at the back (Figure 5h, i, m). We then measured the Rac activity in more than 30 border cell clusters expressing PVRDN and EGFRDN and found no difference between front and back (Figure j,k,m), consistent with the proposal that asymmetric Rac activation requires guidance receptor input. In the absence of such asymmetry, non-directional signals activate Rac uniformly, stimulating random protrusion.
During normal morphogenesis and in tumor metastasis, many cells move in interconnected groups in a process termed collective cell migration4-7. Border cells represent one model for the study of such movements. We previously found that guidance receptor signaling not only promotes border cell protrusion at the front of the cluster but also polarizes the group so as to inhibit protrusion at the rear20. However it was unclear to what extent each cell sensed direction independently or whether they did so collectively and what intracellular signal(s) downstream of the receptors would be sufficient to polarize the group26. The results presented here demonstrate that a local increase in Rac activity is sufficient not only to stimulate protrusion autonomously in the treated cell but also to cause retraction of side and back cells, resulting in net cluster polarization and movement in the direction of highest Rac activity. Conversely inhibition of Rac in the lead cell caused the other cells to protrude in all directions as if guidance receptor activity were lost. These results suggest that elevated guidance activity at the front of the cluster activates Rac to a higher level in the front cell and this is sufficient to set the direction of migration for the whole group. Despite the fact that receptor tyrosine kinases activate myriad downstream signaling pathways, other pathways do not appear to be necessary, though they may play redundant or overlapping roles. Thus, asymmetric Rac activity is key for direction-sensing in vivo. We also show that JNK signaling is required to transmit the guidance signal between cells of the cluster. This work further suggests that photoactivatable proteins are likely to be a powerful new class of tools for the manipulation of protein activities with fine spatial and temporal control to address a variety of biological questions in animals.