The Maf factor Traffic jam both enables and inhibits collective cell migration in Drosophila oogenesis

Cell migration is indispensable for metazoan development. The border cell cluster (BCC) in the Drosophila ovarian follicle is an important model system to study the regulation of collective cell migration (Rørth, 2009; Montell et al., 2012). The BCC is a small group of cells that segregate from an epithelium and collectively undergo guided migration on the surface of other cells. BCC migration has been extensively used to study the molecular mechanisms that control cell motility (Rørth, 2009; Montell et al., 2012).

In a Drosophila ovarian follicle, a layer of follicle cells surrounds 16 germline cells that include a posteriorly located oocyte, and 15 anteriorly located nurse cells. Each follicle goes through 14 stages of oogenesis (King, 1970). At stage 8, the anteriormost follicle cells develop into the BCC. The BCC consists of two nonmotile polar cells, surrounded by four to eight migratory border cells, which we refer to as rosette cells (Montell et al., 2012; Niewiadomska et al., 1999). At stage 9, the BCC segregates from the follicular epithelium and migrates between the nurse cells toward the oocyte. After 6 hours, at stage 10a, when most follicle cells, called main-body follicle cells (MBFCs) are arranged around the oocyte, the BCC completes migration. At stage 10b, another group of follicle cells, called centripetal cells, moves interiorly along the oocyte-nurse cell boundary and contacts the BCC (King, 1970).

A number of factors have been identified to regulate various aspects of BCC migration (Rørth, 2009; Montell et al., 2012). Among those, the Jak/Stat signaling pathway, transcription factor Slow border cells (Slbo), and adhesion molecule DE-cadherin (DEcad) form a core pathway that controls the motility of the BCC. The polar cells signal through the cytokine Unpaired (Upd) to the neighboring follicle cells, leading to phosphorylation of the Jak protein Hopscotch (Hop) and subsequently the Hop target Stat92E (Silver and Montell, 2001; Beccari et al., 2002; Ghiglione et al., 2002). Phosphorylated Stat92E activates expression of Slbo, the Drosophila homolog of CCAAT-enhancer binding proteins (C/EBP) (Montell et al., 1992). Slbo is crucial for regulating the expression of various factors involved in migration, including adhesion molecules, cytoskeletal regulators and endocytic factors (Borghese et al., 2006; Wang et al., 2006). An essential target of Slbo is shotgun (shg), which encodes DEcad (Niewiadomska et al., 1999; Mathieu et al., 2007). DEcad is enriched in the BCC compared with other follicle cells, and provides the necessary traction for rosette cells to migrate over the nurse cells (Niewiadomska et al., 1999).

Achieving a balanced expression level of these factors is important for proper motility of the BCC, because over- or under-expression of components of the Jak/Stat pathway, Slbo or DEcad negatively impacts migration. Reduction of Stat92E activity during the migration process (Silver et al., 2005), as well as its overactivation, interferes with BCC migration (Yoon et al., 2011). Loss of slbo function in rosette cells blocks the motility of the BCC (Montell et al., 1992; Rørth et al., 2000), whereas overexpression of slbo causes a delay in migration and frequently a fragmentation of the BCC (Rørth et al., 2000; Starz-Gaiano et al., 2008). Similarly, not only does lack of DEcad abolish BCC motility (Niewiadomska et al., 1999), but also higher-than-normal expression of DEcad in rosette cells delays migration (Schober et al., 2005).

It is not surprising therefore, that molecular checkpoints keep these factors in balance during BCC migration. A Stat92E target, Apontic, negatively regulates Jak/Stat signaling in rosette cells by activating a microRNA that causes Stat92E transcript degradation (Starz-Gaiano et al., 2008; Yoon et al., 2011) and upregulating Socs36E (Monahan and Starz-Gaiano, 2013). Socs36E, which is also a target of Stat92E, acts in a negative feedback loop to limit Jak/Stat signaling in many cell types (Callus and Mathey-Prevot, 2002; Rawlings et al., 2004; Baeg et al., 2005; Issigonis et al., 2009; Singh et al., 2010; Tarayrah et al., 2013), and its overexpression has been shown to impair BCC migration (Silver et al., 2005; Monahan and Starz-Gaiano, 2013). Although the molecular mechanism by which Socs36E regulates the Jak/Stat pathway is not well understood, it has been shown that loss of Socs36E leads to elevated Stat92E expression in the testis stem cell niche (Issigonis et al., 2009). The lifetime of Slbo is tightly controlled by antagonistic actions of two factors: the kinase Tribbles, which removes Slbo through ubiquitin-mediated proteolysis (Rørth et al., 2000; Masoner et al., 2013), and the ubiquitin hydrolase Ubp64E, which stabilizes Slbo by removing ubiquitin from the protein (Rørth et al., 2000). The transcription factor Yan (Anterior open) functions in the endocytosis of DEcad, regulating the local concentration of this adhesion molecule in rosette cells (Schober et al., 2005).

Here, we show that the large Maf transcription factor Traffic jam (Tj) is crucial for establishing the correct balance of the core components within the regulatory cascade that controls BCC motility. We previously reported that Tj is needed for Drosophila oogenesis (Li et al., 2003). Follicle cells that lack Tj display changes in the expression of adhesion molecules, undergo abnormal shape changes, and often leave the follicular epithelium (Li et al., 2003). To better understand the function of Tj in controlling cell behavior, we studied its function in the rosette cells, which naturally separate from an epithelium and undergo migration.

Our analysis revealed that knockdown of tj in rosette cells leads to a significant delay in BCC migration. This defect seems to be mediated by a reduction in Slbo and consequently DEcad. Interestingly, a natural decrease in Tj concentration, which seems to depend on Slbo, was observed in migrating BCCs. We also found that overexpression of Tj severely impedes migration. This suggests that too little or too much Tj has a negative impact on BCC motility. Tj overexpression caused a significant decrease in Slbo, which appears to be mediated by an increase in expression of the Jak/Stat inhibitor Socs36E. It also inhibited DEcad expression even in the presence of Slbo. Our data indicate that Tj is required at low, tightly regulated concentrations to allow proper expression levels of all three components within the core cascade, providing rosette cells with the necessary balance of factors that confers proper cell motility.

Unless noted otherwise, Drosophila crosses were done at 25°C and female flies were fed yeast for 2 days before dissection. We used slbo-Gal4 [P{Gal4-slbo2.6}1206 (Rørth et al., 1998)] to drive expression in rosette cells, and c306-Gal4 [P{GawB}c306; Bloomington Drosophila Stock Center (BDSC)] for expression in anterior follicle cells, including the BCC. The strains UAS-tj^RNAi (10034R-2; NIG-Fly Stock Center), Df(2L)E55 (deletion of tj; BDSC), tj^eo2 (null allele; Schüpbach and Wieschaus, 1991; Li et al., 2003), tj³⁹ (hypomorphic tj allele; T. Panchal, E. Alchits and D.G., unpublished), slbo¹ (Montell et al., 1992), Stat92E^F [Stat92E^ts (Baksa et al., 2002)], Stat92E⁰⁶³⁴⁶ [Stat92E-lacZ (Hou et al., 1996)], Socs36E^PZ1647 [Socs36E-lacZ (Issigonis et al., 2009)] and Df(2L)Exel7070 (deletion of Socs36E; BDSC) were used for loss-of-function analysis, and UAS-tj⁶⁽³⁾ or UAS-tj¹⁽²⁾ (see below), UAS-slbo (Starz-Gaiano et al., 2008), UAS-shg (Sarpal et al., 2012) and UAS-hop3 (Harrison et al., 1995) for overexpression studies. slbo2.6-lacZ [slbo-lacZ (Rørth et al., 1998)], shg1.3-lacZ [shg-lacZ (Mathieu et al., 2007)], 10XStat-GFP [Socs36E-GFP (Bach et al., 2007)], Socs36E-lacZ and Stat92E-lacZ were used as enhancer reporters. Stat92E^ts/Stat92E⁰⁶³⁴⁶ flies, used for temperature-dependent reduction of Stat92E activity, were grown for 12 hours at restrictive temperature (29°C) before dissection. To generate Tj-overexpressing cell clones, hsFlp1/Act5C>CD2>Gal4; UAS-tj¹⁽²⁾/UAS-lacZ flies were heat shocked at 37°C for 15 minutes and dissected 24-48 hours later. Tj-overexpressing clones were identified by expression of lacZ. Unless noted otherwise, slbo-Gal4/+ served as the control genotype, and slbo-Gal4 induced expression of UAS constructs.

A tj cDNA with the complete open reading frame and 3′ untranslated region, including a poly(A), and a partial 5′ untranslated region, was generated from two overlapping tj cDNAs (Li et al., 2003) and subcloned into pUAST vector to generate transgenic UAS-tj flies following standard procedures. UAS-tj¹⁽²⁾ and UAS-tj⁶⁽³⁾ are inserted on the second and third chromosome, respectively. Exogenous Tj properly localized to the cell nucleus and rescued the tj mutant embryonic gonad phenotype (data not shown).

The following primary antibodies were used: polyclonal guinea-pig anti-Tj (G5 1:5000; the same Tj peptide used to immunize rats (Li et al., 2003) was injected into guinea pigs), rat anti-Slbo (1:4000) (Mathieu et al., 2007), rat anti-DEcad (DCad2; 1:25), mouse anti-Fasciclin III (FasIII, 7G10, 1:50), mouse anti-Armadillo (Arm, 7A1, 1:100) (Developmental Studies Hybridoma Bank), rabbit anti-β-galactosidase (β-gal) (1:100,000 for slbo-lacZ, 1:10,000 for shg-lacZ; 1:1500 for Stat92E-lacZ; 1:15,000 for Socs36E-lacZ; MP Biomedicals) and rabbit anti-Stat92E (1:1000) (Amoyel et al., 2013). Secondary antibodies (1:400) were conjugated to Cy3 (Jackson ImmunoResearch Laboratories) or Alexa Fluor 488, Alexa Fluor 555 or Alexa Fluor 647 (Molecular Probes, Life Technologies).

Tissue in situ hybridization was done as previously described (Niewiadomska et al., 1999) using cDNAs of tj (Li et al., 2003) and slbo (RE37385; Drosophila Genomics Resource Center) to generate digoxigenin-labeled (Roche) DNA probes.

All imaging was done with a 40×/1.4 Plan-Apo objective using microscopes from Carl Zeiss MicroImaging. Confocal fluorescence images were acquired with an LSM510 microscope. Regular fluorescence images (not shown), used to calculate migration indices, were acquired with an Axioscope-2 microscope and Axiocam camera, and Axiovision 4.3 was used for length measurements. In situ hybridization images were generated with an Axioplan-2 microscope and a Canon EOS Rebel Digital SLR camera. Images were processed with Adobe Photoshop and Illustrator CS2 (Adobe Software).

Immunofluorescence signals of Tj and Slbo in confocal images were quantified using ImageJ (NIH), measuring a 7.16 μm² circular area per cell nucleus. For each BCC, a z-stack of confocal images was recorded, and for each nucleus, the focal plane with the brightest signal was chosen for measurement.

The BCC migration index [100% denotes complete migration; 0% no migration (Melani et al., 2008)] was determined for stages 9, 10a and 10b of oogenesis (King, 1970). We sub-categorized stage 9 follicles according to the position of the anteriormost MBFCs relative to the distance between the anterior follicle tip (0%) and the oocyte-nurse cell border (100%): 0-30% defined early stage 9, 30-60% mid stage 9 and 60-100% late stage 9. Unpaired, two-tailed Student’s t-tests (for unequal sample sizes and equal variance) were used for all statistical analyses (Microsoft Excel), and graphs were created with Prism4 (GraphPad Software).

BCCs were isolated from follicles as previously described (Wang et al., 2006). Green fluorescent protein (GFP)-positive cells, composed of BCCs and centripetal cells, constituted 70% of the total pool of isolated cells. Total RNA extraction using TRIzol (Life Technologies) typically yielded 150 ng/μl per 150 ovary pairs. Microarray analysis was conducted by the Canadian Drosophila Microarray Center using Drosophila NimbleGen 4-plex (4x72k) arrays (Roche). Raw intensities were obtained using Nimblescan software, and were normalized and converted into fold changes using ArrayStar (DNASTAR).

Our previous work showed that Tj is expressed in the BCC (Li et al., 2003). We examined the expression of Tj in more detail in stage 8 to 10 follicles, from the time of BCC formation to the end of its migration. At early stage 9 when the BCC segregates from the follicular epithelium, similar amounts of Tj protein were detected in the BCC and the neighboring follicle cells (Fig. 1A-A′). When the BCC had detached from the follicular epithelium and started to migrate toward the oocyte, Tj protein appeared significantly reduced in BCCs in comparison with MBFCs (Fig. 1B-C′). Within the BCC, the reduction of Tj protein seemed more prominent in the migratory rosette cells compared with the polar cells. Quantification of the Tj signal intensity at four different stages of oogenesis (early, mid and late stage 9, and stage 10) revealed a gradual reduction in rosette and polar cells during BCC migration (in comparison with MBFCs) (Fig. 1D; supplementary material Fig. S1F). When the BCC reached the oocyte at stage 10, the relative amount of Tj protein had dropped to ∼30% of the original level (Fig. 1D). To ensure that the observed reduction of Tj in the migrating BCC was not caused by inaccessibility to the Tj antibody, we confirmed this result by immunostaining cryosections of ovarian follicles (supplementary material Fig. S1A-E). In contrast to the gradual reduction of Tj protein, tj mRNA appeared already strongly reduced in BCCs at early stage 9, but was maintained at a low level throughout migration (Fig. 1E,F), suggesting that the downregulation of tj expression occurs at the pre-translational level.

The presence of Tj throughout migration prompted us to ask whether it has a function in BCC migration. Generation of tj null mutant cells in the anterior region of the follicular epithelium prevented the recruitment of rosette cells, suggesting that tj is required for BCC formation (J. D. Alls and D.G., unpublished data). To determine whether Tj is needed for the actual migration process, we induced expression of tj double-stranded RNA (UAS-tj^RNAi) during migration using the slbo-Gal4 line, which is strongly active in rosette cells (Rørth et al., 1998). tj RNAi was effective, as Tj protein was reduced to background levels even in early stage 9 rosette cells (compare Fig. 2A-A′ with 2B-B′). To examine the effect of tj knockdown on BCC migration, we determined the BCC migration index at four different stages (mid and late stage 9, stages 10a and 10b) (see Materials and methods) (Fig. 2G). The migration index was significantly lower after tj knockdown compared with the control at all four stages (Fig. 2E,G). In comparison with wild-type BCCs that would have reached the oocyte by stage 10a, the majority of BCCs with tj knockdown had reached the oocyte only by stage 10b. However, although delayed by several hours, most BCCs (66.3%, n=116) were eventually able to reach the oocyte. Removing one wild-type copy of tj only slightly enhanced the migration defect in response to tj RNAi (Fig. 2G), indicating the robustness of the RNAi effect. A significant delay in BCC migration was also observed in a hypomorphic tj mutant (tj^eo2/tj³⁹), in which BCCs could form (Fig. 2G). Our data indicate that a knockdown of tj in rosette cells causes a significant delay in BCC migration.

To determine whether the natural reduction of Tj protein is necessary for BCC migration, we overexpressed full-length Tj [UAS-tj⁶⁽³⁾] in rosette cells (Fig. 2C-C′). Analysis of the migration index indicates that tj overexpression (Fig. 2F,H) impairs migration more severely than tj knockdown (Fig. 2E,G). At mid or late stage 9 when most control BCCs were undergoing migration, almost all tj-overexpressing BCCs failed to migrate (Fig. 2H). Several BCCs appeared to have moved away from the anterior tip of the follicle but were unable to detach from the epithelium. Even at stage 10b, very few BCCs (5.3%, n=113) had reached the oocyte. Moreover, in mosaic BCCs, Tj-overexpressing rosette cells were usually found at the trailing end of the BCC (94%, n=32; Fig. 3G-H; Fig. 5G-H′). Our data suggest that the endogenous downregulation of tj in rosette cells is necessary for BCC migration.

Based on the importance of Slbo (Montell et al., 1992) and Tj as regulators of BCC migration, we asked whether they have a functional relationship. In control BCCs, Slbo is expressed at similar levels in rosette and polar cells throughout migration (Fig. 3A,A′,D,D′,J). Both knockdown and overexpression of tj (using slbo-Gal4) caused a significant reduction of Slbo in rosette cells even by early stage 9 (Fig. 3B-C′,E-F′). Signal quantification indicates a reduction of Slbo by 50% in both genotypes compared with the control (Fig. 3J). Inducing Tj overexpression in cell clones before BCC formation caused Slbo expression to be undetectable in rosette cells (Fig. 3G-I). In situ hybridization suggested that altered tj expression affects slbo at the mRNA level (supplementary material Fig. S2A-C). This was corroborated by microarray analysis using RNA isolated from border cells, showing a reduction of slbo mRNA by 28% when tj was knocked down and 46% when tj was overexpressed (Fig. 3K, data not shown). These data suggest that the endogenous low level of Tj is important for normal transcriptional activity of slbo.

The slbo enhancer reporter slbo2.6-lacZ is active in rosette and centripetal cells (Rørth et al., 1998) (supplementary material Fig. S2D,D′). Interestingly, slbo-lacZ activity in rosette cells seemed noticeably reduced when tj was knocked down (supplementary material Fig. S2E,E′), but did not appear to change compared to the control when tj was overexpressed (supplementary material Fig. S2F,F′). This suggests that Tj may activate slbo expression through this 2.6 kb enhancer element but that Tj inhibits slbo expression through a different molecular mechanism.

To confirm that Tj acts upstream of Slbo, we tested whether slbo overexpression can rescue the BCC migration defect caused by altered tj expression. Expression of exogenous Slbo (UAS-slbo) in rosette cells fully rescued the migration defect caused by tj knockdown (Fig. 4A,E) and partially restored the ability of Tj-overexpressing BCCs to migrate (Fig. 4B,F). Our results indicate that the regulation of slbo by Tj in rosette cells is important for BCC migration. Strikingly, combined tj knockdown and slbo overexpression could induce precocious BCC migration. This was not observed for BCCs that only overexpress Slbo (Jang et al., 2009; Rørth et al., 2000; Starz-Gaiano et al., 2008) (our own observations). Some of those BCCs with reduced Tj and elevated Slbo were able to complete migration at early-to-mid stage 9, when BCCs have normally only begun migration (Fig. 4G,H). This suggests that a combination of tj downregulation and slbo activation can induce BCCs to migrate earlier and/or faster than normal.

To test whether tj downregulation can enable BCC migration independently of Slbo, we examined whether altering tj expression modified the BCC migration defect observed in a hypomorphic slbo mutant. slbo¹ mutant BCCs display a severe delay in BCC migration (Montell et al., 1992), but 30% and 70% of BCCs had moved away from the anterior tip of the follicle by stage 10b and stage 11/12, respectively (supplementary material Fig. S3A,D). When tj was knocked down in slbo¹ mutant BCCs, the percentage of BCCs that showed signs of movement at stage 10b and stage 11/12 was reduced to 10% and 30%, respectively (supplementary material Fig. S3B,D). Overexpression of tj in the slbo¹ mutant background caused complete blockage of BCC migration, even by stage 12 (supplementary material Fig. S3C,D). Thus, both changes in tj expression exacerbated the BCC migration delay in the slbo mutant, possibly due to further reduction of the already low amount of Slbo. This result shows that reduction of Tj expression is not sufficient to drive BCC migration in the absence of Slbo.

Slbo upregulates DEcad expression in the BCC, which was found to be essential for migration (Niewiadomska et al., 1999; Mathieu et al., 2007). As Tj modulates slbo expression, one would expect Tj also to have an effect on DEcad. We examined the activation of a shg enhancer reporter (shg1.3-lacZ that contains putative Slbo-binding sites; Mathieu et al., 2007) in the background of tj knockdown and overexpression. In control follicles, this shg-lacZ is strongly active in rosette cells, weakly in centripetal cells, and not detected in MBFCs (Mathieu et al., 2007) (Fig. 5D,D′, Fig. 6E,E′). shg-lacZ expression was reduced moderately when tj was knocked down and more strongly when tj was overexpressed (Fig. 5E-F′). Consistent with these data, DEcad protein levels were visibly reduced under both conditions (Fig. 5A-C′). This reduction is particularly evident in Tj-overexpressing rosette cells of mosaic BCCs (Fig. 5G-H′). These findings are consistent with a Slbo-dependent function of Tj in the regulation of DEcad.

To confirm that the Slbo-dependent increase in DEcad is an important downstream process of Tj activity, we asked whether an induced increase of DEcad could rescue the tj mutant effects on BCC migration. Exogenous DEcad expression (using UAS-shg) fully rescued the moderate migration delay caused by tj knockdown (Fig. 4C,E) and partially rescued the severe delay induced by tj overexpression (Fig. 4D,F). Although DEcad upregulation completely rescued the tj knockdown phenotype, it did not induce precocious BCC migration, in contrast to slbo overexpression. In addition, DEcad upregulation was not as effective as slbo overexpression in rescuing the tj overexpression phenotype (Fig. 4F). Our data suggest that the upregulation of DEcad is not the only motility-enabling process that is dependent on Slbo, and in turn on Tj.

We previously showed that Tj modulates DEcad expression in MBFCs that do not contain Slbo (Li et al., 2003), raising the question of whether Tj might also regulate shg expression independently of Slbo in the BCC. We tested whether slbo overexpression can rescue the Tj-mediated effect on shg. We first looked at shg-lacZ expression in slbo-overexpressing MBFCs and BCCs (excluding rosette cells that were left behind) without changing Tj expression. Surprisingly, although Slbo expression is crucial for shg upregulation, excessive amounts of Slbo attenuated shg-lacZ expression in the BCC (Fig. 6F,F′). However, MBFCs that normally contain minimal Slbo began to express shg-lacZ in response to slbo overexpression (Fig. 6F,F′). This indicates that Slbo needs to be present at the proper concentration to activate the shg enhancer. Interestingly, shg-lacZ expression was substantially stronger in both the BCC and MBFCs when tj was knocked down in the background of slbo overexpression (Fig. 6G-G′). By contrast, shg-lacZ expression was strongly reduced in the BCC (Fig. 6H,H′) and undetectable in MBFCs (Fig. 6H,H′) when both tj and slbo were co-overexpressed. Analysis of DEcad protein levels yielded similar results (Fig. 6A-D′). Our data suggest that high amounts of Tj suppress activation of the shg enhancer. In summary, the analysis of DEcad expression in BCCs with varying Tj and/or Slbo concentrations indicates that Tj has both Slbo-dependent and -independent effects on DEcad.

Our analysis indicates that the downregulation of Tj starts at the onset of BCC migration when Slbo becomes highly active, raising the question of whether Tj downregulation depends on Slbo. In contrast to the control, Tj protein was found not to be properly reduced in slbo¹ mutant rosette cells, when BCCs were analyzed either according to stage (stage 10b; Fig. 7A-B′,E) or migration index (Fig. 7F). A similar effect was observed for tj mRNA (data not shown). slbo overexpression in rosette cells had the opposite effect, causing a stronger than normal reduction of Tj at early stage 9 (Fig. 7C-D′,G). These observations suggest that Slbo is needed for the downregulation of tj at the onset of BCC migration.

Because the initial reduction of Tj seems to also coincide with the activation of Jak/Stat signaling in BCCs, we studied whether this pathway played a role in regulating Tj expression either through or independently of Slbo. We induced ectopic activity of Hop in anterior follicle cells (c306-Gal4 UAS-hop3), which causes the formation and migration of additional border cells (Silver and Montell, 2001). Similar to endogenous border cells, these extra migratory cells showed progressive decrease in the amount of Tj protein (supplementary material Fig. S4A). This suggests that the Jak/Stat pathway at least indirectly acts upstream of Tj. To test whether it downregulates Tj independently of Slbo, we overexpressed Hop in slbo¹ mutant rosette cells and examined Tj expression levels. Tj protein did not appear reduced in these cells (supplementary material Fig. S4B-B′), suggesting that overactivation of the Jak/Stat pathway is not sufficient to cause a decrease of Tj in the absence of Slbo. We conclude that the Jak/Stat pathway regulates tj through Slbo.

As the Jak/Stat pathway is important not only during BCC formation but also migration (Silver et al., 2005), we examined whether changing Tj expression affected the Jak/Stat pathway. We utilized a Socs36E-GFP enhancer reporter that had previously been used to monitor Jak/Stat activity (Bach et al., 2007). In control follicles (stages 9 to 10), Socs36E-GFP expression was strong in rosette cells and weak in centripetal cells and posterior MBFCs (Fig. 8A,A′; supplementary material Fig. S5A,G,G′). Inducing tj overexpression in these three cell types led to a striking overactivation of Socs36E-GFP (Fig. 8B,B′; supplementary material Fig. S5B). The increase in Socs36E expression seen in Tj-overexpressing rosette cells was confirmed with another Socs36E enhancer reporter line (Socs36E-lacZ) (supplementary material Fig. S5E,F). Although Socs36E-GFP expression did not noticeably change when tj was knocked down during migration (with slbo-Gal4; Fig. 8C,C′), it did appear to be weaker in rosette cells when tj was reduced in follicle cells before stage 8 (with c306-Gal4; supplementary material Fig. S5G-H′). Taken together, these observations suggest that Tj can enhance Socs36E expression.

To test whether Tj might increase Socs36E levels by enhancing Stat92E expression, we examined the activity of the Socs36E enhancer reporter in a Stat92E mutant background (Stat92E^ts/Stat92E⁰⁶³⁴⁶ at the restrictive temperature). As expected, when Tj was unaltered, the expression of Socs36E-GFP was weak in the Stat92E mutant BCC (Fig. 8D,D′). Interestingly, when Tj was overexpressed in the Stat92E mutant, Socs36E-GFP expression appeared considerably enhanced in rosette and centripetal cells, and posterior MBFCs (Fig. 8E,E′), similar to the effect of Tj overexpression in a wild-type background (Fig. 8B,B′; supplementary material Fig. S5B). However, we did not observe an obvious increase in the activity of a Stat92E enhancer reporter (Fig. 8D′,E′) in any of these cell types. In addition, Tj overexpression did not cause a detectable increase in Stat92E protein level in the rosette or centripetal cells (supplementary material Fig. S5C-D′). Taken together, these results suggest that Tj can enhance Socs36E expression independently of Stat92E.

As Socs36E antagonizes the Jak/Stat pathway (Rawlings et al., 2004; Baeg et al., 2005), which normally activates slbo (Silver and Montell, 2001), we asked whether Tj overexpression might reduce Slbo levels through a Socs36E-mediated inhibition of Jak/Stat activity. If increased Socs36E expression is responsible for the observed decrease in Slbo, a reduction of Socs36E function [Socs36E^PZ1647/Df(2L)Exel7070] in the background of Tj overexpression would be expected to restore Slbo. Indeed, Slbo expression returned to control levels under this condition (Fig. 8F-I), and the motility of the BCCs was rescued to a degree similar to that seen when Slbo was directly co-overexpressed with Tj (compare Fig. 8J with Fig. 4F). Our data suggest that high amounts of Tj interfere with BCC migration by enhancing Socs36E expression, which in turn reduces Slbo.

Our findings indicate that Tj is an important component of the molecular network that controls BCC migration. Tj interacts with at least three regulators of BCC motility to coordinate proper migratory behavior of rosette cells (Fig. 9). Tj functions in: (1) maintaining proper Slbo expression; (2) limiting expression of DEcad; and (3) enhancing expression of Socs36E, which in turn presumably restricts the Jak/Stat pathway. In addition to regulating cell motility, Tj influences the correct temporal initiation of BCC migration.

The bZip transcription factors Tj and Slbo seem to act in a feedback loop to keep each other’s expression in balance during migration. The endogenous low amount of Tj is needed to sustain Slbo, as too much or too little Tj leads to Slbo reduction. Tj seems to activate slbo transcription either directly or indirectly, and to reduce Slbo by enhancing expression of the Jak/Stat antagonist Socs36E. Previous analysis revealed that Slbo is regulated at the post-translational level through ubiquitin-mediated proteolysis (Rørth et al., 2000). Our results here indicate an additional layer of Slbo regulation at the transcriptional level, emphasizing the importance of having the appropriate amount of Slbo to enable normal BCC migration. Interestingly, Slbo expression appears not to be completely dependent on Jak/Stat activity once the BCC has formed and initiated migration (Beccari et al., 2002; Silver et al., 2005). Our results suggest that Tj is a key factor that maintains Slbo expression during BCC migration, probably in conjunction with the Jak/Stat signaling pathway.

Similar to Slbo, Tj is needed at a particular expression level to enable cell motility. Not only loss but also excess of Tj has a negative impact on BCC migration. This suggests that limiting Tj expression is necessary for migration, a process that appears to be mediated by Slbo. Whether Slbo inhibits Tj expression through transcriptional repression or a more indirect mechanism awaits further investigation. We conclude that the mutual regulation of Tj and Slbo is essential to confer normal BCC motility.

Our study here uncovers a feedback loop between a large Maf (Tj) and a C/EBP factor (Slbo), expanding our insight into the interactions between these types of transcription factors. It was previously shown that C/EBPβ regulates the expression of MafB in mouse osteoclasts (Smink et al., 2009). In several mammalian tissues, such as ovaries (Pall et al., 1997), liver (Sakai et al., 1997; Akira et al., 1990) and kidney (Sadl et al., 2002; Alam et al., 1992), C/EBP and large Maf factors have similar expression patterns, raising the possibility that the observed cross-regulatory interactions between Tj and Slbo are conserved.

A substantial increase in DEcad in rosette cells is essential for BCC migration (Niewiadomska et al., 1999). Our findings indicate that Tj has both positive and negative effects on DEcad expression. The rise in DEcad levels is dependent on Slbo (Niewiadomska et al., 1999; Mathieu et al., 2007), and therefore indirectly on Tj. The Tj-mediated negative impact on DEcad became apparent when high Tj expression still prevented the upregulation of shg (which encodes DEcad), even in the presence of exogenous Slbo. Given that both Tj and Slbo are bZip proteins, Tj may dimerize with Slbo, preventing it from binding to the shg enhancer. Notably, Tj homologs c-Maf and MafB have been shown to prevent other transcription factors from activating their targets by physically binding to them (Sieweke et al., 1996; Hegde et al., 1998). Alternatively, Tj may inhibit shg upregulation through transcriptional repression. We previously found that loss of Tj function leads to stronger expression of shg in follicle cells (Li et al., 2003). The absence of Slbo in these cells suggests that Tj can repress shg independently of Slbo.

Our findings further indicate that Tj enhances the expression of the Jak/Stat antagonist Socs36E, which in turn reduces Slbo. This effect on Slbo is probably mediated through the Jak/Stat pathway, as Socs36E is a well-known inhibitor of Jak/Stat signaling in multiple tissues, including the follicular epithelium (Callus and Mathey-Prevot, 2002; Rawlings et al., 2004; Issigonis et al., 2009; Singh et al., 2010; Tarayrah et al., 2013). Socs36E acts not only upstream but also downstream of Stat92E (Rawlings et al., 2004; Bach et al., 2007). Our data suggest that Tj affects Socs36E expression in a Stat92E-independent manner. In addition to Tj and Stat92E, the histone demethylase dUTX and the transcriptional regulator Apontic were recently found to upregulate Socs36E (Tarayrah et al., 2013; Monahan and Starz-Gaiano, 2013). It seems that Socs36E has various upstream activators but operates consistently to limit Jak/Stat pathway activity.

We propose the following model for achieving a balance between the factors that regulate BCC motility (Fig. 9). Before BCC formation, follicle cells have a high Tj expression level. This could prevent Slbo expression, block an increase of DEcad, and inhibit Stat92E activity by enhancing Socs36E expression (Fig. 9C). Through a not-yet-fully-understood mechanism, which involves the activation of ecdysone signaling (Jang et al., 2009), the Jak/Stat pathway becomes active, and induces the expression of Slbo in the developing BCC. We speculate that this initiation of Slbo expression triggers a shift in the balance of the factors that leads to a homeostatic feedback loop between Tj and Slbo. This feedback loop also keeps Socs36E activity and DEcad expression at levels that enable migration (Fig. 9A). Tj appears to function as a mediator that balances the activities within the motility regulating core pathway. In addition, Tj is involved in the temporal control of BCC migration, as too much Tj blocks initiation of migration, whereas too little Tj, in the presence of Slbo, drives precocious migration. We propose that the natural reduction of Tj is one of the important temporal cues needed to initiate BCC migration.

The BCC is a system that displays both the pro- and anti-migratory attributes of the large Maf transcription factor Tj, a contradiction that renders it necessary for Tj expression to be maintained at a balanced level. In vertebrates, large Maf factors have not been directly implicated in regulating cell migration during normal development. However, they were shown to be involved in oncogenesis (Eychène et al., 2008). Overactivation of large Mafs can lead to increase in metastasis-inducing factors and tumorous growth in certain tissues (Nishizawa et al., 1989; Hurt et al., 2004; Pouponnot et al., 2006; Morito et al., 2006). However, in other tissues, large Mafs have been shown to repress invasive cell behavior (Pouponnot et al., 2006; Watson et al., 2004). These findings support the view that large Mafs could have pro- and anti-metastatic properties. The role of Tj in fine-tuning cell motility might therefore be a conserved function of large Maf transcription factors.

We thank P. Rørth, D. Montell, D. Harrison, E. Matunis, E. Bach, U. Tepass, the Developmental Studies Hybridoma Bank, the Bloomington Drosophila Stock Center, and the NIG-Fly Stock Center for reagents; A. Soltyk and T. Westwood from the CDMC for advice; X. Chen, Y. Park, L. Dang and H. Hong for excellent technical support; J. Alls for preliminary observations; and U. Tepass for helpful comments on the manuscript.