GAGA Regulates Border Cell Migration in Drosophila

Collective cell migration is a complex process that happens during normal development of many multicellular organisms, as well as during oncological transformations. In Drosophila oogenesis, a small set of follicle cells originally located at the anterior tip of each egg chamber become motile and migrate as a cluster through nurse cells toward the oocyte. These specialized cells are referred to as border cells (BCs) and provide a simple and convenient model system to study collective cell migration. The process is known to be complexly regulated at different levels and the product of the slow border cells (slbo) gene, the C/EBP transcription factor, is one of the key elements in this process. However, little is known about the regulation of slbo expression. On the other hand, the ubiquitously expressed transcription factor GAGA, which is encoded by the Trithorax-like (Trl) gene was previously demonstrated to be important for Drosophila oogenesis. Here, we found that Trl mutations cause substantial defects in BC migration. Partially, these defects are explained by the reduced level of slbo expression in BCs. Additionally, a strong genetic interaction between Trl and slbo mutants, along with the presence of putative GAGA binding sites within the slbo promoter and enhancer, suggests the direct regulation of this gene by GAGA. This idea is supported by the reduction in the slbo-Gal4-driven GFP expression within BC clusters in Trl mutant background. However, the inability of slbo overexpression to compensate defects in BC migration caused by Trl mutations suggests that there are other GAGA target genes contributing to this process. Taken together, the results define GAGA as another important regulator of BC migration in Drosophila oogenesis.


Introduction
Collective cell migration was found in many different organisms during embryonic development and wound healing, as well as in some metastatic cancers. In Drosophila oogenesis, a small set of follicle cells (FCs) originally located at the anterior tip of each egg chamber become motile and migrate as a cluster through nurse cells toward the oocyte. These specialized cells are referred to as border cells (BCs) and provide a simple and convenient model system to study the mechanisms that control collective cell migration in vivo [1][2][3][4]. More specifically, BCs arise within an epithelium consisting of more than a thousand FCs that encircle a cluster of 16 germline cells to form an egg chamber [5]. At early stage of oogenesis, two specialized FCs, referred to as polar cells, differentiate both at the anterior and posterior tips of the egg chamber [6][7][8]. The anterior polar cells recruit several (from 4

Molecular Characterization of Chromosomes Carrying Trl 362 and Trl 3609 Mutations
First, we verified fly stocks with Trl mutations chosen for the study, Trl 362 , Trl 3609 , Trl 13C and Trl R85 , by genotyping PCR with allele-specific primers (Table S1). In addition, since according to our previous experience, transgenic Drosophila lines frequently bear extra uncharacterized transposon constructs, sometimes even within genes relevant to the studied process [30], we checked the Trl mutant lines for the presence of additional P element sequences by quantitative PCR. Indeed, this analysis demonstrated that chromosomes carrying Trl 362 and Trl 3609 mutations carry extra P element transgenes, one in each case ( Figure S1A). The subsequent inverse-PCR mapping of P element insertion sites in Trl 362 and Trl 3609 lines revealed previously unknown additional transgenic constructs located within the alan shepard (shep) and couch potato (cpo) genes, respectively ( Figure S1B). The transposon inserted in the intron/promoter of the shep gene (3L: 5,248,444-5,248,451; here and afterwards, coordinates are from Release 6 of the Drosophila melanogaster genome assembly [40]) consists almost exclusively of P element end sequences and has a total length of 913 bp. The transgene inserted in the intron/promoter of the cpo gene (3R: 17,944,070-17,944,077) is much longer (about 9 kb) and its internal composition (DNA sequence between P element ends) was not investigated. The presence of these novel transgenes in Trl 362 and Trl 3609 lines was confirmed by PCR with primers specific to the P element ends and the sequences flanking the insertion sites ( Figure S1B, Table S1).

Decrease in Trl Expression Delays BC Migration
We examined whether the GAGA protein is important for BC migration. For that, we studied this process in hypomorphic Trl mutants, Trl R85 /Trl 362 and Trl 3609 /Trl 13C . A significant delay in BC migration was observed in both mutant combinations compared to the control ( Figure 1A). Particularly, the migration and completion indexes, parameters that characterize the distance covered by BCs by the end of stage 10 [41], were much lower in Trl R85 /Trl 362 and Trl 3609 /Trl 13C mutants than in Trl + /Trl + flies ( Figure 1B). These defects were almost completely rescued by ubiquitous overexpression of GAGA by the means of the hsp83: GAGA-519 transgene [42] in the Trl R85 /Trl 362 and Trl 3609 /Trl 13C mutant backgrounds ( Figure 1B) pointing to the necessity of this transcription factor for BC migration. Notably, the overexpression of GAGA per se had very little effect on migration of BCs ( Figure 1B). To assess the input of BC-specific expression of the Trl gene in the observed phenomenon, we induced RNA interference (RNAi) of this gene using the slbo-Gal4 driver, which is exclusively active in BCs, posterior and centripetal FCs [29,30]. This resulted in BC migration defects similar to those observed in Trl R85 /Trl 362 and Trl 3609 /Trl 13C mutants ( Figure 1B). Thus, we concluded that proper migration of BCs depends on the level of the GAGA protein.  The combination of the slbo-Gal4 (Novo16) driver [30] with the UAS-GFP reporter construct or immunostaining with anti-Fasciclin III antibody (α-Fas III Ab) was used to mark BCs. UAS-Dicer2 was used to increase efficiency of RNAi. "+" indicates wild-type second or third chromosome(s) depending on the genotype. For quantitation, the nurse cell region of egg chambers was divided into 6 groups according to the percentage of the total distance travelled by BCs: 0% (black), 1-25% (dark red), 26-50% (light green), 51-75% (violet), 76-99% (blue) and 100% (brown) [30]. M.I., C.I. and N denote the migration index, the completion index and the number of egg chambers examined, respectively.

GAGA Regulates Transcriptional Activity of the slbo Gene during Migration of BCs
Next, we wondered whether the GAGA protein is expressed in BCs along their migration to the nurse cell-oocyte boundary. To assess that, we used the slbo 1 allele caused by the insertion of LacZ reporter gene within the promoter region of the slbo gene [9]; this mutation is also known as slbo-1 Figure 1. Border cell (BC) migration is delayed in Trithorax-like (Trl) mutant egg chambers. (A) Stage 10 egg chambers from Trl mutant and Trl +/+ flies labelled with slbo-Gal4 > UAS-GFP (yellow), Phalloidin (red) and DAPI (cyan). Note that in the shown examples of Trl mutant egg chambers, BCs had not reached the nurse cell-oocyte boundary. Scale bar is 100 µm. (B) Quantification of the BC migration phenotypes in stage 10 egg chambers of the indicated genotypes. The combination of the slbo-Gal4 (Novo16) driver [30] with the UAS-GFP reporter construct or immunostaining with anti-Fasciclin III antibody (α-Fas III Ab) was used to mark BCs. UAS-Dicer2 was used to increase efficiency of RNAi. "+" indicates wild-type second or third chromosome(s) depending on the genotype. For quantitation, the nurse cell region of egg chambers was divided into 6 groups according to the percentage of the total distance travelled by BCs: 0% (black), 1-25% (dark red), 26-50% (light green), 51-75% (violet), 76-99% (blue) and 100% (brown) [30]. M.I., C.I. and N denote the migration index, the completion index and the number of egg chambers examined, respectively.

GAGA Regulates Transcriptional Activity of the slbo Gene during Migration of BCs
Next, we wondered whether the GAGA protein is expressed in BCs along their migration to the nurse cell-oocyte boundary. To assess that, we used the slbo 1 allele caused by the insertion of LacZ reporter gene within the promoter region of the slbo gene [9]; this mutation is also known as slbo-LacZ. It was previously demonstrated that the β-galactosidase expression pattern driven by slbo 1 is matching that of the endogenous Slbo protein. Importantly, heterozygous slbo 1 egg chambers develop normally and morphologically are indistinguishable from the wild-type counterparts [9]. Immunostaining of slbo 1 /+ stage 9 and stage 10 egg chambers with anti-β-galactosidase and anti-GAGA antibodies revealed that these proteins colocalize in the BC nuclei along the entire process of cell migration; while GAGA was also detected in all FCs (Figure 2A,B). Furthermore, the reduction in the GAGA protein level in slbo 1 /+; Trl R85 /Trl 362 mutants led to a pronounced decrease in β-galactosidase staining at both analyzed stages of egg chamber development, suggesting the regulation of the slbo gene activity by GAGA ( Figure 2C,D).  (F) Downregulation of GAGA enhances BC migration defects observed in slbo mutants. "homo" indicates homozygous state. Quantification of the BC migration phenotypes in stage 10 egg chambers of the indicated genotypes. X-gal staining was used to visualize BCs in all samples except the yw control, which was immunostained with anti-Fasciclin III antibody (α-Fas III Ab). For quantitation, the nurse cell region of egg chambers was divided into 6 groups according to the percentage of the total distance travelled by BCs: 0% (black), 1-25% (dark red), 26-50% (light green), 51-75% (violet), 76-99% (blue) and 100% (brown) [30]. M.I., C.I. and N denote the migration index, the completion index and the number of egg chambers examined, respectively.
To check this hypothesis, we used RT-qPCR to measure the abundance of slbo and Trl transcripts in ovaries of different genotypes ( Figure 2E). Indeed, a strong reduction in slbo transcripts (down to~20%) was detected in Trl R85 /Trl 362 ovaries. At the same time, slbo transcript levels were much less affected in hypomorphic slbo 1 /slbo 1 and slbo 1 /slbo e7b mutants. In addition, overexpression of GAGA had no obvious effect on the slbo gene transcription ( Figure 2E). Taken together, the slbo gene expression in migrating BCs appears to be regulated by the GAGA protein.

Genetic Interaction between Trl and slbo Genes
To further study the possible regulation of the slbo gene expression by GAGA, we employed the genetic approach. Specifically, we compared BC migration defects observed in hypomorphic slbo mutants with those demonstrated by hypomorphic slbo and Trl double mutants. Analysis of stage 10 egg chambers from slbo 1 /slbo 1 and slbo 1 /slbo 1 ; Trl R85 /Trl 362 flies showed that BC migration was completely blocked in 36% and 42% of cases, respectively ( Figure 2F). Similarly, the frequency of such strong defects observed in slbo 1 /slbo ry7 and slbo 1 /slbo e7b egg chambers increased, respectively, from 21% to 85% and from 59% to 75% in Trl R85 /Trl 362 mutant background ( Figure 2F). On the contrary, the addition of one copy of the hsp83: GAGA-519 transgene slightly rescued the complete blockage of BC migration observed in slbo mutants; the defect was observed in 12%, 8% and 44% of slbo 1 , hsp83:GAGA-519/slbo 1 , slbo 1 , hsp83:GAGA-519/slbo ry7 and slbo 1 , hsp83:GAGA-519/slbo e7b egg chambers, respectively ( Figure 2F). Overall, the results indicate that BC migration defects observed in slbo mutants are severely enhanced by mutations in the Trl gene, whereas GAGA overexpression slightly diminishes these defects.

Transcriptional Activity of the slbo-Gal4 Drivers Depends on the GAGA Protein Level
As GAGA is a sequence-specific transcription factor, the most straightforward mechanism of its involvement in the regulation of the slbo gene expression would be the direct protein binding to its recognition site(s) within the target gene regulatory elements followed by local chromatin remodeling [35]. Therefore, we searched for the potential GAGA binding sites (the GAGAG, GAGnnnGAG, GAGnGAG, CTCnnnGAG, GAGnnnnnCTC and (GA) 3 motifs [43]) within the slbo gene locus. Several GAGA motifs were found within the slbo promoter region as well as within the previously described 2.6-kb enhancer element ( Figure 3A), which is present in slbo-Gal4 drivers chic6458, Novo11, Novo16 and Novo22 [29,30]. Due to availability of the drivers, we decided to check whether their functioning depends on the amount of the GAGA protein. To this end, we first compared the intensities of the slbo-Gal4-driven GFP signals within BC clusters of stages 9 and 10 egg chambers in the wild-type and Trl R85 /Trl 362 mutant backgrounds ( Figure 3B,C). This analysis demonstrated that, on average, the GFP fluorescence intensity was about 4.4 times lower in Trl mutants than in the control. In addition, results of RT-qPCR revealed that the GFP expression driven by different slbo-Gal4 drivers was reduced from 1.8-to 5.2-fold in Trl mutant background ( Figure 3D). Thus, we concluded that GAGA appears to regulate the transcriptional activity of the slbo-Gal4 drivers containing the 2.6-kb enhancer element.

Increase in the slbo Expression Level in Trl Mutants Enhances BC Migration Defects
Since slbo seems to be a target gene for GAGA, we wondered whether slbo overexpression can rescue BC migration defects observed in Trl mutants. To check this, we overexpressed exogenous copy of the slbo coding sequence (UAS-slbo) in BCs using the slbo-Gal4 (Novo16) driver. In a wild-type background, this led to about 10.5-fold increase in slbo mRNA level and resulted only in a minimal disruption of the collective cell migration process (Figure 4), which is principally consistent with earlier observations [25]. In Trl 362 /Trl R85 mutants, slbo-Gal4 (Novo16) -driven slbo overexpression increased the expression level of the gene only about 2.0-fold ( Figure 4A). Surprisingly, this was accompanied by stronger defects of BC migration than in Trl 362 /Trl R85 mutants alone ( Figure 4B). The same effect was observed when slbo-Gal4 (Novo22) driver was used ( Figure 4B). Taken together, these results suggest that the decrease in the slbo expression in Trl mutants is not the only reason for the observed defects in BC migration. Most likely, there are some other GAGA target genes, which expression levels are crucial for the studied process.    Figure 2E. Error bars represent standard error of the mean. (B) Quantification of the BC migration phenotypes in stage 10 egg chambers of the indicated genotypes. The combination of the slbo-Gal4 (Novo16) or slbo-Gal4 (Novo22) driver [30] with the UAS-GFP reporter construct was used to mark BCs. For quantitation, the nurse cell region of egg chambers was divided into 6 groups according to the percentage of the total distance travelled by BCs: 0% (black), 1-25% (dark red), 26-50% (light green), 51-75% (violet), 76-99% (blue) and 100% (brown) [30]. M.I., C.I. and N denote the migration index, the completion index and the number of egg chambers examined, respectively.

Trl Expression Does Not Depend on Slbo
Considering strong genetic interaction between slbo and Trl genes and the fact that Slbo is also a sequence-specific transcription factor [9,26], we asked whether the Trl gene could be a target of Slbo. To answer this question, we measured the Trl gene expression in slbo mutant ovaries by two different approaches. First, RT-qPCR measurements showed that the Trl expression was not substantially affected in slbo 1 /slbo 1 and slbo 1 /slbo e7b mutant ovaries ( Figure 2E). However, this result is not very informative since the slbo gene is known to be active only in a minor fraction of Drosophila ovarian cells [29,30]. Second, we immunostained slbo 1 /slbo e7b egg chambers, in which slbo expression is decreased by 2.2-fold ( Figure 2E), with anti-GAGA antibodies to estimate the amount of this protein. This assay also did not detect any obvious change in the amount of the GAGA protein in BCs and in  Figure 2E. Error bars represent standard error of the mean. (B) Quantification of the BC migration phenotypes in stage 10 egg chambers of the indicated genotypes. The combination of the slbo-Gal4 (Novo16) or slbo-Gal4 (Novo22) driver [30] with the UAS-GFP reporter construct was used to mark BCs. For quantitation, the nurse cell region of egg chambers was divided into 6 groups according to the percentage of the total distance travelled by BCs: 0% (black), 1-25% (dark red), 26-50% (light green), 51-75% (violet), 76-99% (blue) and 100% (brown) [30]. M.I., C.I. and N denote the migration index, the completion index and the number of egg chambers examined, respectively.

Trl Expression Does Not Depend on Slbo
Considering strong genetic interaction between slbo and Trl genes and the fact that Slbo is also a sequence-specific transcription factor [9,26], we asked whether the Trl gene could be a target of Slbo. To answer this question, we measured the Trl gene expression in slbo mutant ovaries by two different approaches. First, RT-qPCR measurements showed that the Trl expression was not substantially affected in slbo 1 /slbo 1 and slbo 1 /slbo e7b mutant ovaries ( Figure 2E). However, this result is not very informative since the slbo gene is known to be active only in a minor fraction of Drosophila ovarian cells [29,30]. Second, we immunostained slbo 1 /slbo e7b egg chambers, in which slbo expression is decreased by 2.2-fold ( Figure 2E), with anti-GAGA antibodies to estimate the amount of this protein. This assay also did not detect any obvious change in the amount of the GAGA protein in BCs and in other cell types of slbo 1 /slbo e7b mutants at stage 9 ( Figure 5A) or at stage 10 ( Figure 5B) compared to the appropriate controls (Figure 2A-D).

Discussion
Collective cell migration plays significant roles in normal development and tumorigenesis [44][45][46][47]. Therefore, thorough understanding of regulation of this process is very important. In this study, we found that along with reduced female fecundity due to egg chamber apoptosis prior to oocyte maturation [32], Trl mutants also demonstrate strong defects in BC migration. Particularly, different Trl allele combinations lead to defects in 35-49% of stage 10 egg chambers. The presence of additional P element insertions in the chromosomes carrying Trl 362 and Trl 3609 mutations, which were revealed in this study, most likely does not substantially influence the BC migration process due to the following two reasons. First, the rescue experiments with the hsp83:GAGA-519 transgene clearly show that it is the lowered amount of GAGA that is responsible for BC migration defects observed in Trl mutants. Second, shep and cpo affected by the additional transposon insertions have not been so far identified as BC-or ovary-specific genes. However, the additional molecular features of the Trl 362 -and Trl 3609 -bearing chromosomes should be taken into account in experiments on cells/tissues expressing these genes, such as neurons and/or glial cells in the central nervous system (CNS) [48][49][50][51].
Considering the importance of Slbo [9] as one of the main regulators of BC migration, it was interesting to assess whether Slbo and GAGA may have a functional relationship during this process. Indeed, the decrease in GAGA level in BCs results in substantial decrease in slbo activity, but not vice versa. At the same time, overexpression of GAGA has no effect on slbo mRNA level indicating that GAGA regulates slbo expression only positively. A strong genetic interaction between Trl and slbo mutants, along with the presence of putative GAGA binding sites within the slbo promoter and enhancer sequences, suggests the direct regulation of this gene by GAGA. The reduction in the slbo-Gal4-driven GFP expression within BC clusters in Trl mutant background supports this idea. It is worth noting that no putative GAGA binding sites were found within the minimal hsp70 promoter element present in the slbo-Gal4 construct. The inability of slbo overexpression to compensate defects in BC migration caused by Trl mutations indicates that there are other GAGA target genes

Discussion
Collective cell migration plays significant roles in normal development and tumorigenesis [44][45][46][47]. Therefore, thorough understanding of regulation of this process is very important. In this study, we found that along with reduced female fecundity due to egg chamber apoptosis prior to oocyte maturation [32], Trl mutants also demonstrate strong defects in BC migration. Particularly, different Trl allele combinations lead to defects in 35-49% of stage 10 egg chambers. The presence of additional P element insertions in the chromosomes carrying Trl 362 and Trl 3609 mutations, which were revealed in this study, most likely does not substantially influence the BC migration process due to the following two reasons. First, the rescue experiments with the hsp83:GAGA-519 transgene clearly show that it is the lowered amount of GAGA that is responsible for BC migration defects observed in Trl mutants. Second, shep and cpo affected by the additional transposon insertions have not been so far identified as BCor ovary-specific genes. However, the additional molecular features of the Trl 362 -and Trl 3609 -bearing chromosomes should be taken into account in experiments on cells/tissues expressing these genes, such as neurons and/or glial cells in the central nervous system (CNS) [48][49][50][51].
Considering the importance of Slbo [9] as one of the main regulators of BC migration, it was interesting to assess whether Slbo and GAGA may have a functional relationship during this process. Indeed, the decrease in GAGA level in BCs results in substantial decrease in slbo activity, but not vice versa. At the same time, overexpression of GAGA has no effect on slbo mRNA level indicating that GAGA regulates slbo expression only positively. A strong genetic interaction between Trl and slbo mutants, along with the presence of putative GAGA binding sites within the slbo promoter and enhancer sequences, suggests the direct regulation of this gene by GAGA. The reduction in the slbo-Gal4-driven GFP expression within BC clusters in Trl mutant background supports this idea. It is worth noting that no putative GAGA binding sites were found within the minimal hsp70 promoter element present in the slbo-Gal4 construct. The inability of slbo overexpression to compensate defects in BC migration caused by Trl mutations indicates that there are other GAGA target genes contributing to this process. Alternatively, Slbo could require GAGA and/or some other co-factor(s), which are downregulated in Trl mutants, to properly activate transcription of its target genes. Taken together, the results of this study define GAGA as another important regulator of BC migration in Drosophila oogenesis.

Identification and Verification of P Element Transgene Insertion Sites
Genomic DNA was isolated from 50 flies according to the protocol described previously [55]. Determination of copy number of P element end sequences, mapping and verification of P element-based transgene insertion sites were performed according to [30] with the following modifications. Only primers specific for P element 5 end and the reference Vps36 gene were used for quantitative real-time PCR. Templates for inverse PCR were prepared using MspI (SibEnzyme, Novosibirsk, Russia) and Kzo9I (SibEnzyme) restriction enzymes. Sequences of primers used to verify P element insertion sites by PCR are listed in Table S1.

Generation of UAS-slbo Transgenic Flies
To make pUASTattB-slbo construct for ectopic expression of the Slbo protein, we first PCR-amplified the full-length slbo coding sequence with primers 5 -AAAGAATTCCAAAATGCTGAACATG GAGTCGC-3 and 5 -AAATCTAGACTACAGCGAGTGTTCGTTGG-3 using genomic DNA isolated from yw flies as a template. Next, the amplified DNA fragment was cloned into the pUASTattB plasmid vector [53] by using the unique EcoRI and XbaI sites (underlined in the primer sequences). The pUASTattB-slbo construct was verified by Sanger sequencing that revealed several synonymous nucleotide substitutions within the slbo coding sequence. The plasmid was injected at the concentration of 250 ng/µl into embryos of the BDSC line #24482 as described in [56].

Quantitative Measurement of GFP Signal Intensity
The GFP signals were detected using confocal microscope LSM 710 (Carl Zeiss). Fluorescence intensity quantification was performed for individual confocal images acquired at the same settings using the ZEN 2012 software v 8.1. Experiments were performed in three biological replicates for each genotype and stage of egg chamber development.

Total RNA Extraction, cDNA Synthesis and Quantitative Real-Time PCR
For each genotype, three replicates of 50 ovaries from 1-2-day-old flies were dissected in 1×PBS, preserved in 100 µL of RNAlater solution (#AM7020; Thermo Fisher Scientific, Waltham, MA, USA) and stored at 4 • C. Subsequent isolation of total RNA, reverse transcription and quantitative PCR (RT-qPCR) were carried out as reported previously [30] with primer pairs specific for A. vinelandii GFP coding sequence and Drosophila slbo, Trl, RpL32 and Rap2l genes (for primer sequences, see Table 1). The latter two genes were used as reference genes. The mean Cq values obtained from independent biological replicates are reported in Table S2.  Figure S1. Molecular characterization of chromosomes carrying Trl 362 and Trl 3609 mutations. Table S1. Primers used for PCR verification of the Trl mutations and additional transposon constructs identified within the alan shepard (shep) and couch potato (cpo) genes. Table S2. RT-qPCR data: mean Cq values obtained and expression values calculated from independent biological replicates.