Germline soma communication mediated by gap junction proteins regulates epithelial morphogenesis

Gap junction (GJ) proteins, the primary constituents of GJ channels, are conserved determinants of patterning. Canonically, a GJ channel, made up of two hemi-channels contributed by the neighboring cells, facilitates transport of metabolites/ions. Here we demonstrate the involvement of GJ proteins during cuboidal to squamous epithelial transition displayed by the anterior follicle cells (AFCs) from Drosophila ovaries. Somatically derived AFCs stretch and flatten when the adjacent germline cells start increasing in size. GJ proteins, Innexin2 (Inx2) and Innexin4 (Inx4), functioning in the AFCs and germline respectively, promote the shape transformation by modulating calcium levels in the AFCs. Our observations suggest that alterations in calcium flux potentiate STAT activity to influence actomyosin-based cytoskeleton, possibly resulting in disassembly of adherens junctions. Our data have uncovered sequential molecular events underlying the cuboidal to squamous shape transition and offer unique insight into how GJ proteins expressed in the neighboring cells contribute to morphogenetic processes.


Introduction
In the metazoans, several temporally coordinated interconnected morphogenetic events are instrumental for the transformation of a young embryo into an adult. Such events are spatiotemporally restricted and molecular processes underlying these events are causally as well as mechanistically linked. Among these, cell shape changes are especially crucial for the proper execution of several important developmental landmarks that lead to cell fate specification and patterning [1]. Thus, causes as well as consequences of cell shape changes, are being analyzed extensively in different biological contexts [2]. In this regard, epithelial morphogenesis has served as an extremely informative model.
Epithelial cells are classified into three basic categories based on the aspect ratio, namely cuboidal, squamous and columnar [3,4]. As the names indicate, cuboidal cells have similar height and width while the squamous cells are wider with reduced height. By contrast, columnar cells are less wide but of greater height. Depending on the individual cellular and/or developmental context, epithelial cells display distinct cell shape changes ranging from relatively simple ones like invagination, in-folding, and intercalation (convergent extension) to complex transformations including epiboly and branching morphogenesis [2]. One of the less explored aspects of epithelial morphogenesis is how the local intercellular interactions mediate these distinct shape changes during development. How these transitions are globally modulated, at a tissue level, to ultimately evolve a complex pattern is also relatively unclear. It is also apparent that in several instances different types of epithelial cells come in close proximity and interact with one another leading to distinct morphogenetic outcomes. For instance, close physical interactions between the columnar cells of wing imaginal disc proper with the overlying squamous peripodial cells plays an integral role in mediating wing morphogenesis in flies [5]. In fact, all organ assemblies including skin, lung, trachea, cervix and cornea are critically dependent on proper interactions and inter-convertibility between the epithelial subtypes [6][7][8][9].
In this regard, transformation of cuboidal cells into squamous cells is especially critical for aiding metazoan development [10]. This is important as squamous epithelial cells constitute a protective barrier around the tissues and organs, which helps to regulate the exchange of substances, ions, gases etc. between a given tissue and its environment [11]. Previous studies have indicated that most, if not all, of the shape transitions are modulated via the alterations in the intercellular adhesions between the neighboring cells [2]. Moreover, the cytoskeletal network plays an autonomous role in aiding the shape change by inducing alterations in adherens junctions [12,13]. Nonetheless, despite the functional significance, the precise mechanisms underlying acquisition of the squamous epithelial fate have remained elusive so far.
Here we have focused our attention on Drosophila ovary, which is an excellent model system to study cell shape transitions during development [14,15].Typically, a Drosophila ovary consists of approximately 100-150 oval structures or the egg chambers. An egg chamber is a reproductive unit of a female fly, which consists of 16 germline cells (cyst) enveloped by a layer of somatic follicle cells that are epithelial in nature and cuboidal in shape to begin with [15]. It develops through 14 stages, with the first seven stages being categorized as previtellogenic while the subsequent seven, as vitellogenic. As the oogenesis progresses, proliferating follicle cells accommodate the increasing size of the germline cells size till stage 6. Eventual expansion in the volume of the germline cells is accompanied by the shape change of the cuboidal follicle cells. At the anterior end,~50 follicle cells of cuboidal fate referred to as anterior follicle cells (AFCs) in the text henceforth, reduce their lateral membrane (flatten) and expand their surface area (stretch) to cover the expanded germline cyst by acquiring squamous cell fate also denoted as stretched cell fate. In contrast, the posterior follicle cells elongate their lateral membranes enveloping the oocyte completely, thus procuring a columnar morphology. Despite the contrasting nature of the cell shape changes at the two termini, both these morphogenetic transformations assist in accommodating growth of the increasing size of the germ-line cyst [16].
Follicle cells, somatic in origin, are specified under the combined control of a number of signaling pathways including Notch, Transforming Growth Factor -beta, JAK-STAT, Epidermal Growth Factor and Hedgehog [17][18][19][20][21][22]. Furthermore, some of these signaling events also contribute to the generation of different epithelial subtypes. For instance, Notch signaling in the follicle cells modulates the adherens junctions remodeling to facilitate the follicle cell flattening [18]. In addition, the transcription factor, Hindsight presumably aids in remodeling adhesion between stretching AFCs [23]. While the Transforming Growth Factor signaling is thought to provide the temporal cue to initiate the cuboidal to squamous fate transition [17].
We recently demonstrated that Inx2, a gap junction protein mediates specification of a subset of follicle cells that are anteriorly localized [24]. Our results suggested that Inx2 controls the gradient of JAK-STAT signaling to ultimately influence Border Cell (BC) fate determination within the AFCs. Interestingly, AFCs, adjacent to the BC progenitors, are the ones that undergo cuboidal to squamous cell shape transition. Here we have investigated function of Inx2 in this shape transition during egg chamber morphogenesis. We report a series of molecular events initiated by the calcium flux passing through a heteromeric GJ channel at the germ-line/soma interface. Furthermore, we also demonstrate how these events may ultimately lead to a stereotypical cell shape change experienced by the AFCs. Taken together, our data have uncovered a novel molecular mechanism downstream of two GJ proteins that contributes to the cuboidal to squamous cell shape transition.

Inx2 assists the cuboidal to squamous cell shape change observed in the AFCs
Graded activation of JAK-STAT signaling is essential for the acquisition of distinct follicle cell fates along the anterior-posterior axis of developing Drosophila egg chamber [19]. As we have demonstrated that gap junction protein Innexin2 (Inx2) potentiates the JAK-STAT signaling in the AFCs, we were curious to investigate the role of Inx2 in mediating the cell shape transition in developing follicle cells [24].
We thus decided to assess whether altering Inx2 function can influence the canonical cell shape changes AFCs undergo. In stage 8 egg chambers (c306-Gal4; UAS: mCD8: GFP) cuboidal AFCs display conspicuous, intercellular junctions with distinct lateral membranes and evenly distributed nuclei ( Fig 1A). As oogenesis proceeds, the lateral membranes of the AFCs begin to shrink in size to initiate flattening coupled with significant expansion in their surface area (Fig 1A-1F). The surface area expansion is facilitated by the remodeling of the adherens junctions which eventually aids in the spreading of AFCs to acquire the squamous fate ( Fig 1E) [18]. By stage 10, the squamous cells nuclei are spread out and intercellular junctions are inconspicuous due to considerably reduced lateral membrane (Fig 1C and 1F). To investigate the role of Inx2, we down regulated Inx2 in the follicle cells using RNA interference approach. Since c306-Gal4 expression spans the AFCs of stage 8 egg chambers, we employed it to overexpress inx2RNAi in the developing eggs (S1A-S1C Fig). Upon downregulation, 38.5 ± 1.2% of stage 10 egg chambers exhibited clustering of nuclei within the AFCs (Fig 1G-1L). This clustering of AFC nuclei in stage 10 egg chambers has been referred to as stretching defect henceforth in the text. It could arise either due to flattening defect where the shrinking of the lateral membrane is delayed or the follicle cells are unable to expand their surfaces (Figs 1G-1H and S1G-S1L). This suggests that Inx2 depletion affects the transitioning of cuboidal cells to squamous fate. Next, we tested if the RNAi construct indeed targets Inx2. We immunostained random overexpressing flip-out clones of inx2RNAi and observed lower levels of Inx2 protein compared to their wild-type counterparts suggesting that inx2RNAi indeed targets Inx2 (S1D-S1F Fig). The presence of some residual Inx2 protein in the flip-out clones may be responsible for the modest penetrance. We also observed that smaller clones spanning the AFCs exhibited cell autonomous defects (S1G-S1I Fig). Curiously, in some instances, a few Inx2-depleted follicle cells were completely flattened however they did not exhibit the stretched morphology (S1J-S1L Fig). This particular phenotype suggested that the shape transition of the wild type AFCs likely proceeds via two stages, flattening being an intermediate stage that precedes stretching although we don't understand the mechanistic underpinnings or nature of triggers that allow this progressive transition (S1J-S1L Fig).
To extend these observations, we performed clonal analysis using Mosaic Analysis with a Repressible Cell Marker (MARCM) with two different Inx2 mutant alleles, inx2 G0173a , and inx2 G0059 [24]. MARCM technique allows labeling of homozygous mutant cells by activation of fluorescent reporter in an otherwise heterozygous genetic background of unlabeled cells. We analyzed homozygous mutant clones for both the alleles in developing egg chambers, which span over more than 7-8 follicle cells. Consistent with the earlier observations, homozygous mutant clones of inx2 (inx2 G0173a and inx2 G0059 alleles) in the AFCs exhibited defects in cuboidal to squamous cell shape transition in more than 20 egg chambers analyzed for each inx2 mutant (Figs 1M-1O and S2A-S2C). The mutant follicle cells were smaller with intact lateral membrane suggesting incomplete squamous cell morphogenesis ( Fig 1M' and 1O'). In addition, homozygous mutant follicle cell clones of inx2 G0173a or inx2 G0059 exhibit severe down regulation of Inx2 protein (S2D- S2G Fig). Overall, our data suggest that Inx2 is involved in mediating the cuboidal to squamous transition of AFCs in developing egg chambers. The severity of defects is probably correlated with the extent of downregulation of Inx2 resulting in either severe flattening defects or moderate stretching defects.
Failure to expand the surface area resulted in inability of inx2 mutant cells to cover nurse cells completely thus inhibiting the posterior movement of the main body follicle cells over the oocyte. As a result, the egg chambers with inx2 mutant AFCs exhibited readily discernible morphological defects. (Fig 1M-1O). To assess if the effect of inx2 on the cell shape change is spatially restricted and is confined to cuboidal to squamous cell shape transition, we also analyzed the inx2 mutant clones in posterior follicle cells (PFCs), which acquire columnar shape in wild type egg chambers. Though inx2 is also expressed in the PFCs, unlike the anteriorly localized clones, inx2 clones spanning PFCs displayed normal cuboidal to columnar cell shape transition (S2B- S2C Fig). This observation suggests that Inx2 does not influence cuboidal to columnar shape transition observed in the PFCs. Thus, Inx2 activity is unlikely to be essential for the other cell shape changes during follicular morphogenesis. Furthermore, we did not observe any morphogenesis defects when the germline nurse cells were mutant for Inx2 (inx2 G0173 ) further confirming a specific requirement of Inx2 in the AFCs. (S2H- S2J Fig) To ascertain the specificity of the inx2 function, we overexpressed Inx2cDNA in the mutant cells using actin-Gal4 driver to assess if the shape transition defects observed in the inx2depleted follicle cells are rescued. This experiment was done using MARCM analysis where actin-Gal4 is only active in follicle cells that are homozygous mutant for inx2. Satisfyingly, over-expression of Inx2cDNA in inx2 mutant clones reversed the mutant phenotype completely as the mutant cells overexpressing Inx2 displayed normal flattening and stretching behaviors (Figs 1P-1R and S2K-S2M). It should be noted that we specifically considered mutant egg chambers where the inx2 mutant clones spanned at least 8 contiguous follicle cells (n �12). Together our data suggested that GJ protein Inx2 contributes to ovarian morphogenesis by regulating cuboidal to squamous cell shape transition seen specifically in the AFCs. It is important to note that overexpression of inx2RNAi with the centripetal-posterior follicle cell specific Gal4 (c289b11-Gal4) or main body follicle cell Gal4 (cy2-Gal4) didn't affect the shape transition of AFCs (S2N- S2R Fig). Taken together these data suggest that Inx2 primarily functions in the AFCs that are undergoing cuboidal to squamous shape transition.
Next, we sought to assess if the fate of the follicle cells that do not undergo flattening was affected when inx2 levels are compromised. Wild type follicle cells which transition from cuboidal to squamous fate, respond to Decapentaplegic (Dpp) signaling [17,25]. As Dpp signaling activates transcription of downstream target gene dad (Daughters against Dpp) [26,27], expression of dad reporter i.e., dad-lacz has been used as a readout of the Dpp signaling in the squamous follicle cells [17,28]. In wild type stage 8-9 egg chambers, dad expression initiates in the AFCs that would transition to squamous fate. In stage 10 egg chambers, dad expression is restricted to the squamous follicle cells (approximately 50 cells) that cover the nurse cells and the centripetal cells [29]. Dad expression is neither detected in the precursor cuboidal cells before stage 6 nor is it detected in the main body follicle cells that would envelope the growing oocyte (S3A Fig) [29]. We observed that the Inx2 depleted follicle cells that failed to acquire squamous shape, continue to express dad-lacz reporter (S3B-  [30]. Consistent with our expectation, we observed that the Inx2 depleted follicle cells that failed to acquire squamous shape are devoid of BB127-lacz reporter expression. This confirmed that phenotypes induced by inx2 loss are not due to the retention of the cuboidal fate nor are these cells constituents of main body follicle cells or the centripetal cells (S3E- S3F Fig). Moreover, we also detected the transcription factor, eyes absent (eya), a marker for the AFCs even upon Inx2 depletion (S3G- S3I Fig). Altogether these results strongly suggest that Inx2-depletion specifically affects the AFCs that would have transformed to squamous FCs under normal circumstances.
Earlier reports have indicated possible involvement of Notch pathway components during the cell shape transition event under consideration. These studies also suggested that Notch signaling similarly influences the shape change in the AFCs without affecting their fate [18]. Thus, to examine if Inx2 functions via Notch to modulate the shape change in the AFCs, we analyzed the status of Notch signaling in the FCs compromised for Inx2 (flip-out clones). We used two different Notch pathway readouts, Hindsight, a known target of Notch in the FCs, and NRE-GFP, a transcriptional reporter which is highly responsive to Notch activity [31]. Neither Hindsight nor the NRE-GFP levels appeared to significantly differ between the control and Inx2-depleted follicle cells suggesting that Inx2 likely modulates the shape transition independent of Notch (S3J- S3Q Fig).
Altogether these observations indicated that morphological defects associated with inx2 depletion are not due to inappropriate specification of AFCs. Both the Notch and the BMP pathway targets are relatively unaffected in inx2 depleted follicle cells which suggests that Inx2 may function independent of Dpp and Notch signaling to mediate shape change of the AFCs. Next, we were curious as to how the gap junction protein, Inx2 engineers the cuboidal to squamous cell shape transition in the AFCs.

Inx2 mediated calcium flux plays a crucial role in cell shape transition
Inx2 is a gap junction protein known to function as a channel in combination with other members of the same family. To examine if Inx2 may function as a component of a channel during epithelial shape transition, we employed a fusion transgene of Inx2 where coding portion of Inx2 has been tagged with RFP at the N-terminal end (RFP:Inx2) [32]. The resulting fusion protein (RFP:Inx2) is thought to function in a dominant negative manner that specifically affects the channel activity of Inx2 [32,33]. Using the flip-out technique, we overexpressed the RFP:Inx2 in a random subset of follicle cells. AFCs over-expressing RFP:Inx2 construct failed to acquire squamous shape. (Fig 2A-2C). This phenotype resembled the follicle cellstretching defect in Inx2 depleted AFCs (Fig 1) Moreover, the fusion protein, RFP:Inx2, was localized to the follicle cell membrane resembling the wild type Inx2 (Fig 2D-2F). Next, we assessed if over expression of RFP:Inx2 influences channel independent functions of Inx2 in developing eggs. Recently it has been reported that channel independent activity of Inx2 mediates the process of neolamination which involves fusion of the apical membranes of the border cell cluster with that of migrating centripetal cells in late egg chambers [34] To examine if the RFP: Inx2 construct can affect neolamination, we independently expressed both RFP:inx2 and inx2RNAi transgenes using the slbo-Gal4 driver in the border cells. Consistent with the previous report, expression of inx2RNAi induced neolamination in 48±0.63% of egg chambers. By contrast, overexpression of RFP:Inx2 construct was unable to do so in 6.6%±2.3% egg chambers (S4A- S4D Fig). Altogether, these data suggest that over expression of RFP:Inx2 doesn't perturb the channel independent activity of Inx2 that is linked with neolamination and argue in favour of the possibility that the channel activity of Inx2 contributes to the shape transition of cuboidal to squamous epithelium.
We have previously shown that channel function of Inx2 mediates the passage of free calcium in somatic follicle cells, which plays a crucial role during specification of border cell fate [24]. Therefore, we were curious to check if Inx2 mediated calcium flux influences the cuboidal to squamous shape transition in the AFCs. As reducing Inx2 function quenches the calcium flux, we decided to test if simply increasing the level of calcium can rescue the defects induced by inx2RNAi expression in the follicle cells.
The level of calcium in the cytoplasm can be elevated either by allowing the extracellular calcium to enter the cells via store operated calcium channels or by releasing the calcium from the intracellular stores such as endoplasmic reticulum. Orai is a membrane associated store operated calcium channel that increases the intracellular level of calcium by increasing the influx of calcium from the extracellular milieu [35,36]. We overexpressed Orai in the Inx2-depleted AFCs and quantified the flattening defect. Depletion of Inx2 in AFCs exhibited flattening defect in 39±2.4% of egg chambers as compared to 7±0.8% observed in the Inx2 depleted

PLOS GENETICS
AFCs that also overexpressed Orai (Fig 2G-2H and 2K). In addition, over expression of Inositol 1,4,5-tris-phosphate receptor (Itpr) also ameliorated the stretching defect in Inx2-depleted AFCs (13±3% versus 39±2.4% observed in control) (Fig 2I-2J and 2L). As Itpr mediates the release of calcium from intracellular stores, these data strongly argue that function of Inx2 during cell shape change is dependent on the availability of calcium. Furthermore, its ability to function as a calcium channel is likely critical to mediate the cell shape changes.

Inx2 functions via JAK-STAT signaling to regulate follicle cell stretching
As calcium influx seems to play an important role during the process, we decided to investigate how Inx2 mediated calcium flux modulates the particular shape transition. We have previously shown that Inx2 mediates STAT activity to influence BC specification [24]. We thus wondered if Inx2 functions through the JAK-STAT pathway in this context. To test if the components of the JAK-STAT pathway interact genetically with inx2 in the AFCs undergoing flattening and stretching, we down regulated Inx2 function in genetic backgrounds where JAK-STAT pathway was compromised. To this end, we used flies heterozygous for hopscotch (hop 27 : null allele of Drosophila JAK) and stat (stat92E P1681 : a null allele of Drosophila Stat92E). We quantified the stretching defects displayed by the AFCs that were simultaneously compromised for inx2 and JAK-STAT signaling. This experiment was done at 25˚C to specifically induce a modest phenotype by inx2RNAi overexpression. We have employed this experimental condition as a 'sensitized background' to assess if JAK-STAT pathway components can modify (either suppress or enhance) the phenotypes induced by Inx2RNAi. Compared to the 19 ± 0.6% stretching defect exhibited by the Inx2 depleted AFCs (c306-Gal4; UAS-inx2RNAi), we observed 53% ± 2.3% stretching defect in hop 27 heterozygous (c306-Gal4/ hop 27 ; UAS-inx2RNAi/+) and 37 ± 1.9% stretching defect in stat92E P1681 heterozygous (c306-Gal4/+; UAS-inx2RNAi/stat92E P1681 ) background at 25˚C (Fig 3A-3E). This result shows that simultaneously compromising JAK-STAT pathway components can enhance the stretching defects induced by down regulation of Inx2. These data raised the possibility that Inx2 may influence the JAK-STAT pathway to achieve the cell shape change in the AFCs.
Next, we were curious if the JAK-STAT signaling per se affects the shape change of AFCs. Mutant AFCs clones of hop 27 or stat92E (P1681) retained higher levels of adherens junction protein Armadillo and failed to stretch like the Inx2 mutants (Figs 3F-3H and S5A-S5C). This suggests that depletion of JAK-STAT signaling can inhibit cuboidal to squamous transition of AFCs.
We also tested the status of JAK-STAT signaling in Inx2 depleted AFCs in early stage 8 egg chambers prior to cell shape transition. To monitor the extent of JAK-STAT pathway activity in AFCs, we employed the 10XSTAT-GFP reporter. The extent of STAT activation correlates with the intensity of GFP-specific signal [37]. As previously reported, in control egg chambers (10XSTAT-GFP/+) a decreasing gradient of the STAT reporter was observed from the terminal anterior end to the center of the egg chamber (Fig 3I-3J) [24]. Here we subsequently quantified and compared the GFP intensity (as a proxy of STAT activity) in both, the control and the Inx2 depleted AFCs that would have presumably acquired the squamous fate. To do this, we evaluated the GFP intensity in the AFCs that are 3 cells away from the anterior polar cells to avoid the border cells. In control, the average reporter intensity (STAT activity) was observed as 562.78 ± 28.69 a.u., while Inx2 depleted AFCs exhibited lower STAT activity with average reporter intensity of 324.74±18.14 a.u. (Fig 3I-3M). Thus, down regulation of Inx2 decreases the activity of JAK-STAT signaling in the AFCs.
If the ability of Inx2 to directly modulate the JAK-STAT pathway is crucial for the specific shape transition, elevating the JAK-STAT signaling in Inx2 depleted follicle cells should To circumvent this problem, we sought to use an indirect approach to elevate STAT activity. Asrij/OCIAD1 (ovarian carcinoma immune-reactive antigen (OCIA) domain family) is an endosomal protein that promotes STAT phosphorylation both in embryonic stem cells and Drosophila hematopoietic stem cells [38]. Supporting the conclusion that Asrij can potentiate STAT activity in follicle cells, we confirmed that, overexpression of Asrij in the wild type follicle cells increased the levels of 10XSTAT-GFP (S5G- S5K Fig). We subsequently co-expressed Drosophila Asrij in the Inx2 compromised follicle cells. Satisfyingly overexpression of Asrij in the Inx2 depleted follicle cells rescued the stretching defect (39.5±2.3% to 9.8±1.3%) (Fig 4N-4P). These data suggested that Inx2 may regulate STAT activity to mediate epithelial cell shape transition in the AFCs. Rescue of stretching defects in the Inx2 depleted AFCs by over expression of endosomal protein Asrij suggested that STAT activity could be one of the mediators of shape change of AFC. However, we cannot rule out the contribution of other mediators or signaling cascades affected by the changes in Asrij levels/activity.

Inx2 modulates myosin through WASp to facilitate follicle cell shape transition
In a number of developmental contexts, cell shape changes depend on alterations in actin/ myosin cytoskeleton. We decided to test if similar cytoskeletal changes accompany cuboidal to squamous cell shape transition in the AFCs. We observed higher levels of F-actin in the Inx2 depleted follicle cells (Fig 4A-4C). In a gastrulating embryo, JAK-STAT signaling represses actin nucleator, WASp to spatially regulate myosin activity to facilitate apical constriction [39]. As STAT activity is downstream of Inx2 in this context, we wondered if Inx2 suppresses actin nucleator WASp to influence actin levels. To test this, we compared cell-stretching defects in Inx2-depleted follicle cells (c306-Gal4; UAS-inx2RNAi) in the wild type and WASp 3 heterozygous background respectively. If Inx2 employs WASp repression to mediate follicle cell stretching, reducing WASp level in Inx2 deficient background should mitigate follicle cell flattening defect. Consistently, we observed quantifiable improvement in the stretching defects. Inx2 depleted follicle cells in WASp 3 heterozygous background exhibited 19±0.7% stretching defect compared to 43±0.7% stretching defect to that observed in wild type background at 29˚C (Fig 4D-4F). This result suggested that Inx2 represses WASp activity to assist the flattening of the AFCs as the oogenesis progresses.
To better understand the cytoskeletal changes, we compared the distribution of Zipper i.e., Myosin heavy chain between the wild type and Inx2-depleted follicle cells (Flip out overexpressing clones). In the wild type egg chambers, a basal level of Zipper was observed in the cuboidal follicle cells (S6A- S6B Fig). While, in the Inx2-depleted follicle cells, we observed higher levels defect of the indicated genotypes. Please note the increase in stretching defect exhibited by Inx2-depleted follicle cells in hop 27  Together these data suggest that the function of Inx2 and STAT affects the actin/myosin cytoskeleton to facilitate the stretching observed in the AFCs.
In Drosophila, non-muscle myosin activity depends upon the phosphorylation of its regulatory component, myosin regulatory light chain (MRLC) at threonine-20 and serine-21 amino acid residue [40]. To assess if the myosin activity was different between the wild type and Inx2-depleted follicle cells, we stained egg chambers with anti-pMyosin (phospho-myosin regulatory light chain) antibody, which specifically detects the phosphorylated form of myosin regulatory light chain [41]. In control egg chambers, the cuboidal epithelial cells display lower levels of junctional phospho-myosin (S6I- S6J Fig). As the oogenesis progressed and the AFCs initiated stretching, loss of phospho-myosin was observed from the junctions. In the Inx2 mutant follicle cells, we observed a high level of phospho-myosin accumulation in the membrane and the junctions of the unstretched follicle cells (Fig 5D-5E). This suggests that Inx2 likely affects both the activity and levels of myosin in Drosophila follicle cells that are undergoing cuboidal to squamous fate transition, and loss of Inx2 leads to inappropriate stabilization of Myosin levels/activity. Similar accumulation of phospho-myosin was observed in STAT depleted AFCs (S6K-S6N Fig).
To further validate that the higher activity of myosin can contribute to the cell shape transition, we overexpressed a constitutively active form of Myosin Light Chain Kinase (MLCK) in

PLOS GENETICS
random subset of AFCs using the flip-out technique. MLCK phosphorylates the myosin regulatory light chain and is thought to increase the non-muscle myosin activity [42]. We observed that follicle cells expressing constitutively active MLCK exhibit a delay in the stretching process, which also correlated with the retention of their adherens junction component DE-Cadherin (Fig 5F-5G) (see below for further details).
Reciprocally, to assess if temporally coordinated reduction in myosin activity can promote follicle cell stretching, we down regulated the function of regulatory light chain of the nonmuscle type 2 myosin encoded by spaghetti squash (sqh) in the AFCs by RNA interference. Interestingly, we observed that follicle cells over expressing sqhRNAi (Flip out clones) exhibited larger surface area (117.4±8.6 μm 2 ) compared to the nearby wild type cells (51±3.6 μm 2 ) ( Fig  5H-5J). This observation suggested that precocious reduction in myosin activity can promote (and hasten) stretching of the AFCs. Altogether our data suggest that Inx2 function attenuates myosin activity to facilitate the flattening of AFCs.

Inx2 permits adherens junction disassembly
Myosin accumulation affects the disassembly of adherens junction (AJ) and has been shown to be crucial for modulating cell shape changes in epithelial cells [18,43]. Intriguingly, Inx2 depleted unstretched follicle cells also exhibit accumulation of myosin at their cell junctions. Furthermore, follicle cells expressing constitutively active MLCK also displayed retention of Armadillo, a component of adherens junction complex (Fig 5F-5G). We thus wondered if Inx2 permits the adherens junction disassembly via myosin distribution in the AFCs that are undergoing shape transition.
To check this, we examined the status of adherens junction component Drosophila Epithelial-Cadherin (DE-Cadherin) in the inx2 mutant follicle cells. We stained egg chambers with anti-DE-Cadherin antibody, a major component of adherens junctions [44]. In wild type stage 8 egg chambers, we observed distinct, continuous DE-Cadherin localization at the junctions of the cuboidal follicle cells (Fig 6A). By contrast, the DE-Cadherin accumulation appeared to be discontinuous with several gaps at the interface of clone-non clone follicle cells in late stage 8 or early stage 9 egg chambers (yellow arrow heads in Fig 6D). Upon completion of stretching, the DE-Cadherin specific signal was nearly absent in the AFCs suggesting that disassembly of adherens junction components is concomitant with the flattening process ( Fig 6B). In contrast, the Inx2 mutant follicle cells in stage 10 egg chambers continued to exhibit high levels of DE-Cadherin suggesting that disassembly and the removal of DE-Cadherin was affected in Inx2 mutants (Fig 6C-6D).
To examine if the retention of DE-Cadherin contributes to the follicle cell stretching defect observed in Inx2 depleted egg chambers (c306-Gal4; UAS-inx2RNAi), we down-regulated Inx2 function in wild type and in Shotgun (Shg) (Drosophila cadherin; shg p34-1 , shg 6114 ) mutant background at 29˚C. Indeed, stretching defects induced by the loss of inx2 were partially rescued by reduction in DE-Cadherin levels/activity from 42.3±2.8% in wild type background to 24.3±2.9% and 20.6±2.2% in shg p34-1 and shg 6115 heterozygous background respectively ( Fig  6E-6H). Inx2 thus regulates the levels of DE-Cadherin in the AFCs to enable follicle cell stretching.
Altogether our data suggests that Inx2 mediates JAK STAT signaling to mediate the flattening of the AFCs in the Drosophila egg chambers. Our genetic and cell biological analysis argues that Inx2 represses WASp to modulate myosin activity critical for the disassembly of adherens junction components during flattening of AFCs. These data are consistent with earlier reports, which documented a similar influence of STAT on the cell shape changes in a different cell biological context [39].

Germline specific GJ protein Inx4 (Zpg) influences the AFCs similarly as Inx2
As increase in the size of the germline cells coincides with the shape transition in the follicular epithelium, we were curious to know if the nurse cells participate in the flattening of the follicle cells [45,46]. A possible mode of germline-soma communication in this context could be via formation of a gap junction channel.
Inx2 is a gap junction component and a member of a calcium channel family. Typically, proteins of this family work in combination to form a hetero-dimeric or a multi-component channel(s) that facilitate transport of small molecules including ions or metabolites. Of note, Inx2 has been shown to genetically interact with Inx4 (hereafter referred to as Zpg), a germline-specific gap junction protein to assist the formation of cyst and egg chamber [47]. Importantly, nurse cell specific expression domain of Zpg seems to be in close proximity of the follicle cell specific expression domain of Inx2 [48]. Based on these two observations we wondered if Inx2 from the follicle cells works in conjunction with Zpg at the FC-NC junction.
To test the idea, we co-stained wild type egg chambers with the antibodies against Inx2 and Zpg respectively. Curiously in addition to their subapical localization, Inx2 puncta were also detected at the apical membrane of follicle cells (Fig 7A). While, Zpg localized to the nurse cell membranes and was also detected at the NC-FC junction (Fig 7B) [48]. We also observed a staining overlap between Inx2 and Zpg at the NC-FC junctions suggesting that the two proteins co-localize partially (Fig 7C). Since the overlap is partial, we sought to test if localization of the two proteins is co-dependent. To assess this possibility, we generated homozygous mutant clones of inx2 and stained the egg chambers with the anti-Zpg antibody. Interestingly, we observed loss of Zpg protein at the NC-FC junction abutting the inx2 mutant follicle cell (Fig 7D-7F). This observation suggested that partial co-localization of these two proteins could be functionally relevant.
We repeated this experiment with RFP:Inx2 to ascertain if its overexpression in the wild type and inx2 mutant follicle cells can affect the localization of Zpg. Consistent with our expectation, we observed normal localization of Zpg protein in both the cases with partial co-localization between both the proteins at the follicle cell and nurse cell interface (Fig 7G-7K and S1 Video). These data further support that the structural properties of RFP:Inx2 protein is relatively intact.
To further examine a possible functional connection between Inx2 and Zpg, we decided to evaluate the stretching of Inx2-depleted AFCs (c306-Gal4; UAS-inx2RNAi), in wild type (+/+) and zpg heterozygous background (zpg [z-5352] /+ or zpg [z-2533] /+). Consistent with our expectation, simultaneous depletion of Inx2 and Zpg resulted in higher incidence of stretching defects (38±2.5% & 40.3±3.0% of stretching defect compared to 21.3±0.9% observed in wild type background). This experiment was performed at 25˚C for moderate expression of RNAi, for details see materials and methods (Fig 7L-7O). Together these data suggest that Inx2 and Zpg interact genetically and might function together to modulate shape change of AFCs.
Based on these observations we wondered if Inx2 and Zpg may work together by forming a heterotypic gap junction channel between the NC-FC cell surfaces. Furthermore, such an interaction could facilitate the flattening of the AFCs. We have previously shown that channel activity of Inx2 mediates calcium flux in the follicle cells to modulate JAK-STAT activity during border cell fate specification. Intriguingly, down-regulation of Inx2 by employing the same Gal4 driver (c306-Gal4) not only affected the border cell fate and calcium flux but also the cuboidal to squamous fate transition in the AFCs (Fig 1G-1L). We were therefore curious to check if Inx2 and Zpg function in the somatic cells and nurse cells respectively to modulate the calcium flux in the flattening follicle cells.

Coordination of calcium flux between the germline and soma
By employing green fluorescent calcium sensor, GCaMP6, we have previously documented that communication between follicle cells is mediated by calcium [24]. Our data showed initiation of a periodic wave of free calcium in a random subset of outer follicle cells, which spread to neighboring follicle cells over time [24]. We also showed that the intensity and transmission of the calcium waves were sensitive to levels/activity of Inx2. However, these studies did not address if the calcium flux is also present in the germline cells and if so, whether the follicle cells and germ-line cells communicate via calcium flux. In the following we have first examined if nurse cells display calcium flux and how nurse cell specific activity of Zpg can influence the calcium flux in the nurse cells and the overlying follicle cells.
To probe these possibilities systematically we expressed genetically encoded GCaMP6 reporter both in the wild type nurse cells (T331-Gal4) and AFCs (c306-Gal4). Expression of GCaMP6 reporter using T331-Gal4 exhibited readily detectable level of fluorescence in the nurse cells (Figs 8A and 8B and S2 Video and S3 Video). As the oogenesis progressed, we observed stochastic increase in the flux intensity in a few nurse cells over the basal level suggesting a presence of nurse cell specific calcium flux.
Importantly, in several instances, a conspicuous increase of flux intensity in a given nurse cell also strongly correlated with a higher level of fluorescence in the neighboring follicle cells (Fig 8A 4 -8A 5 ). In addition, we also observed movement of calcium flux between the nurse cells ( Fig 8B 1 -8B 5 ). This observation suggested that calcium flux that originates in the nurse cells might induce a similar flux in the neighboring follicle cells (Fig 8B 2 -8B 4 ). To rule out the possibility of direct physical transfer of the reporter molecule, we overexpressed GCaMP6 reporter only in the nurse cells. In the absence of GCaMP6 in the FCs, calcium flux was never detected in the overlying follicle cells arguing against a direct physical movement of the GCaMP6 reporter across the germline-soma boundary (S4 Video).
Altogether these data suggested that NCs actively communicate with the FCs at the later stages of oogenesis. We have previously shown that depletion of Inx2 function in the follicle cells inhibits the calcium flux in the follicle cells [24].
Next, we examined if the germline specific calcium flux is sensitive to presence of Zpg. Overexpression of zpgRNAi in the germline nurse cells and anterior follicle cells (c306-Gal4; T331-Gal4) completely obliterated calcium flux in all the nurse cells including the basal germline fluorescence but exhibited flux in the follicle cells (Fig 8C 1 -8C 5 and S5 Video). This data suggested that Zpg plays a significant role in the initiation and modulation of calcium flux in the germline nurse cells.  (Fig 8D). While the intensity of the flux in the overlying follicle cells was significantly reduced (821.24±98.0 a.u.) unlike that observed in the control follicle cells adjacent to wild type nurse cells (1415.91 ±126.6 a.u.) (Fig 8E). So, compromising zpg function in the nurse cells results in dampened yet longer lasting calcium flux in the overlying follicle cells. As Zpg seems to affect the dynamics of calcium flux in the follicle cells, we were curious to know if it also influences the follicle cell flattening. Over expression of zpg RNAi using late driver (T331-Gal4; tub-Gal80 ts ) resulted in follicle cells that were unable to stretch as opposed to the control egg chambers (T331-Gal4; tub-Gal80 ts ; UAS-dicer2) (Fig 8F and 8G). In these egg chambers, we also observed a wavy oocyte cortex that has been recently reported suggesting that our zpg RNAi construct was indeed working on the expected lines [34]. Importantly, the total number of border cells in migrating cluster was not significantly altered when Zpg was down regulated in the nurse cells (S7A-S7C Fig), nor was the follicle cell morphogenesis affected when Zpg function was down regulated in the follicle cells (S7D-S7G Fig). Altogether these data suggest that Zpg and Inx2 function in the germline and soma respectively to mediate the cuboidal to squamous cell shape change of AFCs during Drosophila oogenesis (Fig 9).

Discussion
Morphogenesis involves complex cellular interactions, regulated via short-or long-range influences, which operate both at the local as well as global level. Interactions between different cell types have acquired extra-ordinary significance during proper execution of complex morphogenetic processes including organogenesis. Interacting cellular partners of heterogenous origins participate in generation of a number of coordinated cues. Such cues either biochemical or mechanical in nature, operate both in localized and widespread manner to ultimately orchestrate the formation of simple as well as complex multicellular assemblies. Variety of signals that control the process, and underlying mechanisms that synthesize as well as propagate such signals are being investigated in different developmental contexts.
Here we demonstrate that Inx2, a gap junction protein, controls cuboidal to squamous cell shape transition in the follicular epithelium. Effect of Inx2 is mediated by its ability to regulate STAT activity, which in turn can influence the cytoskeleton. Supporting the conclusion, changes in STAT or alterations in the cytoskeletal machinery components, can either enhance or mitigate the phenotypic consequences induced by the loss of Inx2. Importantly, Inx2 engineers this shape transition likely by acting as a component of a calcium channel. It seems to achieve this by partnering with Zpg, another gap junction protein, belonging to the same family.
Gap junction proteins have been shown to be required in a number of different biological processes including establishment of epithelial integrity and morphogenesis. For example, Inx2 has been shown to control formation of proventriculus, during foregut formation in the flies [49]. Previous reports have also suggested that Inx2 can form homomeric as well as heteromeric gap junction channels [32,49,50]. However, the molecular mechanism of Inx2 mediated epithelial morphogenesis has remained elusive.
This study advances our mechanistic understanding of how the gap junction channels contribute to morphological changes in the epithelial cells. Though calcium has been implicated in regulation of cell shape transition in various developmental contexts, the precise molecular mechanism underlying calcium dependent cell shape transition has not been documented. Our findings demonstrate that calcium dependent JAK-STAT signaling regulates Myosin distribution to engineer cell shape transition. Interestingly we have previously shown that the same molecular cassette is used for proper specification of BCs. Together these data suggest that Inx2-zpg mediates the calcium flux in the follicle cells, which probably regulates JAK--STAT signaling in a context specific manner to achieve diverse outcomes.

Communication between the germline and soma assists epithelial morphogenesis
A heterotypic combination of Inx2-Zpg gap junction channel has been reported to regulate germline-soma interaction during early Drosophila spermatogenesis [51]. In addition, heterotypic combination between Inx2 with other somatic gap junction protein, Inx3 has been reported in developing Drosophila embryo during the regulation of epithelial morphogenesis [52]. Here we have assayed the ability of a putative Inx2-Zpg heterotypic GJ channel responsible for cell shape transition underlying egg chamber morphogenesis. Our results suggest that via Zpg, nurse cells induce the follicle cells to elevate free calcium levels as germline specific depletion of Zpg affects the absolute intensity and duration of calcium flux in the follicle cells. We thus believe that optimum amount of calcium is critical for mediating proper shape transition of AFCs from cuboidal to squamous fate. Although, the nurse cells exhibit high level of GCaMP6 fluorescence, signal transmission to the adjacent follicle cells is relatively infrequent. This observation suggests that the transmission of the signal via Inx2-Zpg is spatially and temporally calibrated during oogenesis. Nurse cells are relatively of a larger size and, as such a direct physical contact may be established between a single nurse cell with a number of follicle cells via formation of a heterotypic GJ channel(s). Contact dependent assembly of Inx2-Zpg channel likely confers selectivity on the release and strength of the calcium flux. Taken together our data are consistent with the calcium flux being a crucial player during this process while the precise identity of the signal, remains to be determined. Whether calcium itself is the elusive 'signal' needs to be investigated further. It is possible that the 'signal' is more complex and consists of several components with calcium being one of the pieces of the puzzle. In such a scenario, future experiments ought to focus on uncovering the identity of the other components of the signal. It will also be important to determine as to how the non-autonomous 'signal' secreted by the nurse cells, is conveyed to the somatic neighbors to modulate the calcium flux. Follicle cells express many gap junction proteins including Inx2 on their cell surface and are competent to communicate with one another. Thus, possible contribution of neighboring follicle cells, not in direct contact with the nurse cells, may also be an important determinant.

Endocytosis of membrane component might be crucial for stretching process
Recently product of tao gene has been shown to regulate follicle cell flattening by modulating the removal of lateral membrane component Fas2 by endocytosis [53]. On similar lines, our data shows that Inx2 permits the disassembly, and perhaps also the disappearance of adherens junction component DE-cadherin from cuboidal cells to facilitate the follicle cell stretching.
Since calcium is known to regulate the rate of endocytosis, we propose that calcium may regulate the levels of DE-Cadherin via endocytosis to mediate the shape change in the AFCs. Partial rescue of inx2 phenotype in the heterozygous shg mutant background suggests that the accumulation of DE-Cadherin is indeed functionally connected to the phenotype. Further experiments will help reveal mechanistic connection between calcium flux and adherens junction assembly and its influence on cell shape transitions.

Relevance for skin development: A perspective
Epidermal stratification during development of mammalian skin depends on cell division, collective migration and controlled differentiation. Together these three processes are responsible for the basal to squamous fate transformation [54]. It is reported that this fate change is also accompanied by increase in the levels of GJ proteins and inter-as well as intracellular calcium [55,56]. Our results have implicated combined involvement of GJ proteins and calcium flux in mediating cell shape changes. It will be worthwhile to examine if similar mechanisms are employed by the differentiating basal cells. Intriguingly, several skin disorders have been linked to mutations in connexins (vertebrate counterparts of GJ proteins). Our results thus provide a readily accessible genetic platform to compare and contrast the molecular mechanisms underlying epithelial cell shape transitions in diverse cellular contexts.  For flip-out clonal analysis, flies of genotype hsFLP UAS-mCD8: GFP; actin<flipout>GAL4 was crossed with UAS-inx2RNAi, UAS-RFP:Inx2, UAS-MLCK.CA, UAS-sqhRNAi. F1 flies of appropriate genotype were subjected to heat-shock for 1 hour duration thrice a day for 3 consecutive days. The flies were dissected on the 6 th day after the last heat shock and ovaries were immunostained with Inx2 or Zipper antibody.

Drosophila stocks and genetics
For experiments in which the F1 genotype was lethal (zpgRNAi with T331-Gal4), temperature-sensitive Gal80 was used to downregulate Gal4 activity during early development. Crosses were kept at 16˚C till progenies emerged. F1 flies were shifted to 31˚C (restrictive temperature) for 18 hours to inactivate GAL80 before dissection.
In all of the RNAi experiments, desired F1 genotype flies were fattened at 29˚C for 18 hrs. UAS-Moesin:GFP or UAS-mCD8:GFP were used to normalize the number of UAS constructs in the cross. All imaging were performed with CZI Apotome & LSM 710 confocal microscope.

Quantification of Stretching defect
To quantify the stretching defect, only stage 10 egg chambers, where all anterior cells are supposed to exhibit complete stretching were considered. Egg chambers in which anterior end of egg chambers retain cuboidal cells or smaller cells which lose their cuboidal shape but failed to enlarge as stretched cells were considered as defective in stretching. The percentage stretching defect represents the number of the stage 10 chambers exhibiting stretching defect of AFCs compared to the total number of stage 10 egg chambers analyzed for that particular genotype.

Quantification of Neolamination
To quantify the Neolamination defect, late stage 10-early stage 11 egg chambers were considered. At these stages, the apical side of the border cell clusters physically attaches with the migrating centripetal cells. Egg chambers where the border cell clusters failed to attach to the centripetal cells were referred to as defective in neolamination. We only considered those egg chambers where the border cell clusters were in the vicinity of the oocyte.

10XSTAT-GFP/ NRE-GFP intensity calculation
For the quantification of 10XSTAT-GFP/ NRE-GFP in the follicle cells anterior end of stage 8 egg chambers were captured keeping polar cells at the center plane. Excluding the first 3 cells from the polar cells, intensity calculation was restricted to the next group of 3 cells on either flank outlined on the basis of Armadillo staining/ Phalloidin and the mean intensity of 10XSTAT-GFP/ NRE-GFP was extracted.

Cell area/size measurement
For measuring cell size (area), clones (GFP+) and non-clone cells (GFP-) were compared keeping equal number of cells per group in each egg chamber. Cell number was counted with help of DAPI staining and cell membrane was outlined by marking Arm staining. Further analysis was performed in Image J software where cell area was extracted out from the outlined total area divided by number of cells to finally calculate area/cell (cell size). At least 10 cells were considered for each group (GFP+/GFP-) per egg chamber.

Calcium imaging
P{UAS-GCaMP6m} (BDSC-42748) line was over expressed by T331-Gal4 & c306-Gal4 for detecting free Ca 2+ in the nurse cells & follicle cells. 12 bit images of frame size 512X512 were captured with CLSM 710 (Zeiss) with Plan-Apochromat 63X oil immersion objectives, N.A 1.4. Argon laser line 488 nm was used for GCaMP6 reporter. Fast continuous live imaging with 6.3μm pixel dwell at frame interval 5 seconds for 5-6 minutes was used for recording the calcium flux. Sequential imaging was performed using Argon 488 laser with time interval 7 second per frame for recording calcium flux. c306-Gal4/UAS-dicer2; UAS-GCaMP; T331-Gal4 was used to visualize the activity of calcium simultaneous in the germline Nurse cells and Anterior follicle cells. The experimental genotypes expressed UAS-zpgRNAi instead of UAS-dicer2 in the control. The flies were fattened at 29C for 19.5 hours. Live imaging was carried out as reported previously [24].
12 bit images of frame size 512X512 were captured with CLSM 710 (Zeiss) with 40X oil immersion objectives. Argon laser (488 nm) was used for GCaMP6m reporter (pinhole diameter 3.5 AU). Time lapse imaging was carried out at intervals of 5 secs for at least 10 minutes. For flux intensity quantification, we used Image J. The individual follicle cells were marked on the basis of DIC image. The mean basal intensity and the maximum flux intensity was quantified for each cell showing flux. The maximum flux intensity was subtracted from the basal intensity for each cell, and was compared across the two genotypes. The number of timepoints for which the flux could be seen was used to plot the persistence graph.