Interplay between integrins and PI4P5K Sktl is crucial for cell polarization and reepithelialisation during Drosophila wound healing

Phosphatidylinositol(4,5)-bisphosphate [PI(4,5)P2] regulates cell adhesion and actin dynamics during cell migration. PI(4,5)P2 binds various components of the cell adhesion machinery, but how these processes affect migration of the epithelial cell sheet is not well understood. Here, we report that PI(4,5)P2 and Sktl, the kinase that converts PI4P to PI(4,5)P2, are both localized to the rear side of cells during wound healing of the Drosophila larval epidermis. The Sktl localization requires JNK pathway activation and integrins, but not PVR. The sktl knockdown epidermis displays strong defects in would closure, reminiscent of the JNK-depleted epidermis, and shows severe disruption of cell polarity, as determined by myosin II localization. Sktl and βPS integrin colocalize at the rear side of cells forming the trailing edge during wound healing and the two are inter-dependent in that the absence of one severely disrupts the rear localization of the other. These results strongly suggest that the JNK pathway regulates the rear localization of Sktl and integrins and the interplay between Sktl and integrins sets up cell polarity, which is crucial for reepithelialisation during wound healing.


GFP-Sktl and PI(4,5)P2 translocate to the rear side of epidermal cells during wound healing.
To investigate possible roles of PI (4,5)P2 in Drosophila wound healing, we analysed the localization of Sktl protein after epidermal injury generated by pinching the cuticle and abrading approximately 30 epidermal cells. Before or immediately after wounding, the functional fusion protein GFP-Sktl 31 was evenly distributed in the dorsal epidermal cells of the third instar larva, but translocated to the rear side of the cell during wound healing, which was most evident in the first 2-3 rows of cells from the wound margin. The localization was noticeable 4 h after injury and became distinct by 7 h, the time interval when cells changed their shapes dramatically and migrated to close the wound hole (Fig. 1a,b) 21 . We analysed the endogenous Sktl protein and obtained similar results (see Fig. 2c). We also examined the localization of PI(4,5)P2 using the reporter gene PH PLCδ -cerulean 32 . Although the distribution pattern was subtle, the plotting results of the fluorescence intensity of Cerulean across the frontal-rear axis of the cells indicated that the levels of PI (4,5)P2 also increased on the rear side during wound healing (Fig. 1c-e; see Methods for the quantification of rear localization).
The rear localization of Sktl requires JNK and integrins, but not PVR. To identify the regulators for Sktl localization, we examined Sktl localization in different genetic backgrounds where the signalling pathways critical for wound closure were deficient. In the larval epidermis expressing the gene encoding a dominant negative form of the Drosophila JNK (bsk DN ), cells remained largely polygonal in shape after injury, as reported previously 18,21 . GFP-Sktl was also evenly distributed across the cell, which was in drastic contrast to wild type (Fig. 2a,b'). We quantified the rear localization of GFP-Sktl based on the GFP fluorescence intensity (Fig. 2c). In A58-GAL4 is a larval epidermis-specific driver. A58-GAL4 UAS-GFPsktl was abbreviated as A58 > GFP-sktl. Cell boundaries were stained with anti-FasIII antibody in red. The asterisk marks the wound hole and the white dotted line indicates the wound margin. Scale bars: 50 μm. (e) Rear localization index values for PH PLCδ -Cerulean were calculated (see Methods). Mann-Whitney U-test was used for statistical significance. the larval epidermis expressing RNAi for integrin βPS (mys) or talin (rhea), cells changed their shapes moderately, consistent with the delayed closure of the wound hole 26 , but the rear localization of GFP-Sktl was severely disrupted (Fig. 2c,d-f '). In the larval epidermis expressing the gene encoding a dominant negative form of PVR (PVR DN ), GFP-Sktl translocalized to the rear side, which was essentially the same as in wild-type ( Fig. 2c,g,g').
Altogether, we conclude that Sktl localization requires JNK and integrins, but not PVR. sktl is required for wound closure. We examined whether sktl-deficient larvae displayed wound closure defects. Because the null mutations of sktl are embryonic lethal 31,33 , we knocked down the gene via UAS-RNAi transgenes using the larval epidermis-specific A58-GAL4 18 . sktl knockdown using two independent RNAi lines targeting different regions of sktl equally and effectively blocked closure of a wound hole with the size of approximately 30 epidermal cells analysed 16 h after injury, whereas GAL4-only control larvae had completely closed the wound hole by that time (Fig. 3a-c,f). We confirmed the result by rescuing the open-wound phenotype www.nature.com/scientificreports www.nature.com/scientificreports/ by simultaneously overexpressing UAS-GFP-sktl ( Fig. 3d-f; UAS-GFP was added to control groups to balance the UAS copy number). We also confirmed that in sktl-knockdown larval epidermis, Sktl protein was substantially reduced, as examined by anti-Sktl antibody staining and Western blotting ( Fig. 3g-j; Supplemental Fig. 3). Altogether, sktl is required for larval epidermal wound closure.
The sktl requirement might be ascribed, not to the function of PI(4,5)P2 itself, but to PI(4,5)P2 serving as a precursor for the synthesis of PI(3,4,5)P3, so far, the most versatile and dynamically regulated type of phosphoinositides 6,13 . Alternatively, it might be due to the requirement for diacylglycerol (DAG) and inositol 1,4,5 triphosphate (IP3), which are generated by the hydrolysis of PI(4,5)P2 by phospholipase C (PLC). We examined wound closure in various larvae where the activities of the key enzymes catalysing each of these reactions were depleted and found that the wound holes closed normally when examined at 16 h (Supplemental Fig. S1). Thus, we conclude the sktl requirement for wound closure is not due to a lack of PI(3,4,5)P3, DAG, or IP3. sktl knockdown larvae display disrupted cell polarization during wound healing. We examined further Sktl function in cell polarization during wound healing via the localization of nonmuscle myosin II 21,22 . In wild-type, the myosin II heavy chain Zip and the functional fusion protein GFP-Zip translocate to the rear side of cells 4-8 h after injury, which requires JNK pathway activation 21 .
In wild-type larvae, 89.1% of the cells located in the first two rows from the wound margin responded correctly to the wound stimulus, as measured by GFP-Zip translocation 7 h after injury (Fig. 4a,c). In sktl knockdown larvae, in contrast, merely 11.8% of the cells located within the same distance responded correctly (p < 0.01), and the rest of the cells failed to relocate GFP-Zip; the protein remained mainly in the peri-nuclear region as if the cells did not receive the wound signals (Fig. 4b,c). The cells displaying the wrong directionality were very minor; 5.4% in wild-type and 3.9% in sktl knockdown larvae (Fig. 4c).
For epistatic analysis, we performed the reverse experiments, analysing the localization of GFP-Sktl protein in zip knockdown larvae. Knockdown of zip made larval epidermal cells slightly abnormal and unhealthy, as observed previously 21 , and consequently, the localization pattern of GFP-Sktl was not as distinct as in wild-type. Nonetheless, the rear localization of GFP-Sktl protein was observed in many of the first three rows of cells (Fig. 4d,e). Thus, Sktl, and possibly PI(4,5)P2, are critically involved in setting up cell polarity during wound healing and functions upstream of myosin II.
We also examined whether JNK pathway activation was affected by Sktl depletion. Induction of the JNK pathway reporter msn-lacZ, analysed 7 h after wounding, was grossly normal in sktl knockdown larval epidermis (Supplemental Fig. 2). Therefore, we concluded that Sktl functions downstream of JNK and upstream of myosin II during wound closure.
Sktl protein was visualized using anti-Sktl antibody in green, and the cell nuclei were stained with DAPI in blue. In Western blotting, β-Tubulin was used as a loading control. The asterisk indicates a non-specific band (see Supplemental We also assessed cell polarization via immunostaining for βPS integrin, another protein that localizes at the rear side of cells during larval wound healing of the epidermis 26 . In control larvae, βPS integrin localized to the rear side of the cells in the first two rows from the wound margin 7 h after injury, as reported previously (Fig. 5a,a') 26 .
In sktl knockdown epidermis, however, βPS integrin appeared evenly distributed in most cells (Fig. 5b,b'). In GFP-sktl-expressing, knockdown-rescued epidermis, the localization pattern of βPS integrin was also rescued (Fig. 5c,c'). We quantified the results and calculated the rear localization index value (Fig. 5d). Together with the results in Fig. 2e,f we conclude that Sktl and βPS integrin are inter-dependent for their rear localization.  Fig. 1; data not shown) 26 . First, βPS localized on the basal side of the epidermal cells facing the basement membrane, visualized using Vkg-GFP (Fig. 6a), and E-cad was on the lateral side, presumably marking the adherens junction (Fig. 6b,c) 34 . In wounded epidermis, βPS integrin and Sktl colocalized to the rear part of cells that appeared the trailing edge (Fig. 6d,d'). E-cad protein was maintained laterally (Fig. 6d' ,e'), whereas the rear part of the cell overlapped with the front part of the cells in the next row, as if the cells lying behind crawled on the cells ahead (Fig. 6d,e'). Consistently, anti-E-cad staining and anti-βPS staining did not overlap in this region; anti-Sktl and anti-βPS staining visualized the presumptive trailing edge, whereas anti-E-cad staining tended to display pentagonal or hexagonal shapes of the cell (Fig. 6d,e'). Localization of FasIII, a septate junction protein that is more basally located, compared to E-cad (Fig. 6e') similarly marked polygonal shapes of cells (Fig. 6e,e'). Thus, in the rear part of the cell, the region demarcated by E-cad and βPS integrin seems to form the trailing edge, and this was the region with concentrated βPS integrin and Sktl (Fig. 6f).

Discussion
Here, we report that sktl, encoding a Drosophila PI4P5K, is required for wound closure of the larval epidermis. Sktl and βPS integrin help each other colocalize to the rear side of cells during wound healing, and this interaction is crucial for setting up cell polarity, which leads to the rear localization of myosin II and reepithelialisation.
We showed, here and previously, that JNK was upstream of these events, as depletion of JNK activation led to disruption of all the events 21,26 . On the other hand, JNK pathway activation was normal in sktl, βPS, talin, or zip knockdown larvae. sktl or βPS knockdown larvae showed disruption in myosin II localization, but zip knockdown did not affect Sktl localization. Thus, wounding in the larval epidermis generates as-yet unknown signals that activate the Rho-family small GTPases 22 , which may be activated differentially in the front and the rear sides of the cell, similar to that in single cell migration 2,3 ; this leads to the activation of JNK 22 , which is translated to the rear localization of Sktl and βPS integrin 26 .
An interesting question is then the mechanism of how JNK signalling relays the information of cell polarity. In a wounded field, JNK signalling is activated in a graded manner, waning towards the distal area, as analysed by the induction of msn-lacZ or puc-lacZ 18 . It is thus plausible that JNK itself provides the frontal-rear information in individual cells, based on the minute difference in JNK activation levels across the frontal-rear axis of individual cells, and possibly, via mutual inhibition between the front and the rear 2 . Alternatively, it may be that a missing factor provides such information and JNK is merely an essential bystander.
We observed that both Sktl protein and PI(4,5)P2 localized to the rear side, but did not provide data to argue strongly that it is PI(4,5)P2 that executes setting up the cell polarity and possibly integrin regulation. One way to address this issue would be to carefully perturb PI(4,5)P2 levels or to swap the wild-type Sktl with a kinase-dead version. Sktl is involved in various processes including cell viability, cytoskeletal organization of actin and www.nature.com/scientificreports www.nature.com/scientificreports/ microtubule, and polar transport of mRNA, proteins, and vesicles 31,33,[35][36][37] , and the PI(4,5)P2 requirement was nicely shown in some of these cases 35,36 . PI(4,5)P2 interacts with talin, vinculin, moesin, myosin-X, and factors involved in cytoskeletal reorganisation 6 . We, therefore, favour the hypothesis that Sktl and the integrin-containing adhesion complex interact with each other to localize to the rear, and the resulting increase of the local concentration of PI(4,5)P2 affects the behaviour of the adhesion complex and promotes epithelial wound closure.
Lastly, it is worth mentioning some unresolved questions in our study. The first is the question of how PI(4,5) P2 affects the integrin adhesion complex and ultimately reepithelialisation. Interaction between talin and PI4P5K or PI(4,5)P2 38,39 enhances talin binding either to β integrin in mammalian mesenchymal cells 6,8 or to myosin II in Dictyostelium 40 . These results are, however, inconsistent with our observation that focal adhesion-like structures containing talin were not particularly enriched in the posterior of the cells during reepithelialisation (data not shown). Thus, the two systems must work differently, and a more careful investigation of the function of PI(4,5) P2 may be necessary. Second, our results showing that the presumptive trailing edge of the cells underlies the frontal part of the cells behind are intriguingly different from what has been observed in the migration of MDCK cell sheets, in which the forward protrusion of so-called cryptic lamellipodia of submarginal cells contributes to cell-sheet migration 41 . Future studies should explore the function of the rear localization of PI(4,5)P2, PI4P5K, and integrins, as regulation of PI4P5K activity might be intimately related to various diseases including cancer metastasis [42][43][44] .   Quantification of the rear localization of PH pLcδ -Cerulean, Sktl-GFP, Sktl, and βPS integrin.

Methods
The fluorescence intensity of Cerulean, GFP, or other fluorophores from immunohistochemical staining was measured along the frontal-rear axis of a marginal cell in the wounded area using ImageJ 2 A. Upon determining the frontal-rear axis, the longest line that lied within ± 45° from the hypothetical perpendicular line towards the wound centre was chosen. The rear localization index was calculated as follows: (the intensity value of the rear half of the cell − that of the frontal half) divided by the intensity value of the frontal half. At least > 80% of the total cells located in the first and second rows from the wound margin were measured, which removed non-analysable cells due to damage or other abnormal conditions. At least six larvae were examined per genotype. For the case of PH PLCδ -Cerulean, at least > 70% of the total cells located in the first and second rows from the wound margin were examined in four larvae.
Quantification of GFP-Zip localization. The localization of GFP-Zip was examined in the first two rows of cells from the wound margin. A line with an arrowhead was drawn in each of these cells by placing its arrowhead on the nucleus and the endpoint on the middle of GFP-Zip aggregation. If the arrow had a direction that was within ± 45° from the hypothetical wounded centre, the cell was sorted as 'normal' . If the arrow was directing outside of the range, the cell was sorted as 'wrong' . If GFP-Zip remained as if the cell did not receive wound signals (similar to the cells in unwounded samples), the cell was sorted as 'no response' . At least 70% of the total cells located in the first and second rows from the wound margin were in analysable conditions, and eight larvae were examined.