miR-989 Is Required for Border Cell Migration in the Drosophila Ovary

microRNAs (miRNAs) are small non-coding RNAs that regulate gene expression by destabilizing target transcripts and/or inhibiting their translation. miRNAs are thought to have roles in buffering gene expression to confer robustness. miRNAs have been shown to play important roles during tissue development to control cell proliferation, differentiation and morphogenesis. Many miRNAs are expressed in the germ line of Drosophila, and functions have been reported for a few miRNAs in maintenance of stem cell proliferation during oogenesis. Here, we analyse the function of Drosophila miR-989 in oogenesis. miR-989 is abundant in ovaries. Mutants lacking miR-989 did not display gross abnormalities affecting egg chamber formation or maturation. However, the migration of the border cell cluster was severely delayed in miR-989 mutant egg chambers. We demonstrate that miR-989 function is required in the somatic cells in the egg chamber, not in germ line cells for border cell migration. Loss of miR-989 from a fraction of the border cell cluster was sufficient to impair cluster migration as a whole, suggesting a role in border cells. Gene ontology analysis reveals that many predicted miR-989 target mRNAs are implicated in regulating cell migration, cell projection morphogenesis, cell adhesion as well as receptor tyrosine kinase and ecdysone signalling, consistent with an important regulatory role for miR-989 in border cell migration.

Introduction miRNAs are small non-coding RNAs that function as regulators of gene expression in a wide range of biological contexts [1,2]. miRNAs associate with their target transcripts via partial complementary base pairing to target sites which are usually located in the target 3'UTR or in coding sequences [3,4]. In general, miRNAs act as negative regulators of gene expression at the post-transcriptional level by promoting target transcript destabilization and/or by reducing their translation [1,2].
Border cells serve as a model system for the study of collective cell migration during Drosophila oogenesis [5,6,7]. Drosophila eggs mature in compound entities called egg chambers, which are comprised of 16 interconnected germ-line cells that are encapsulated by a monolayer of somatic follicle cells [8] (Fig. 1). One of the 16 germ-line cells differentiates as the oocyte, while the other 15 become polyploid nurse cells, which produce RNAs, proteins and organelles for incorporation into the oocyte to aid its maturation. The somatic follicle cells undergo a complex developmental and morphogenetic program that is tightly linked to germ line development and ultimately leads to the formation of the egg shell [7]. A subset of follicle cells, called border cells, has a special role during oogenesis, which involves an invasive, directed, cell migration. During stage 8 of oogenesis the border cells are specified at the anterior pole of the follicular epithelium and start to express the C/EBP transcription factor, Slow border cells (Slbo; Fig 1A). The border cells detach from the follicular epithelium and migrate as a cluster toward the oocyte during stage 9 to 10A ( Fig. 1B, C). At stage 10B, the border cell cluster has reached the anterior face of the oocyte and migrates laterally to its anterodorsal position (Fig. 1D). Specification of the border cells and the transition to coordinated cell migration involve several conserved signalling pathways and extensive remodelling of the cytoskeleton and cell adhesion properties [5,6,7]. The JAK/STAT pathway is required for border cell specification and for migration [9,10,11]. Ecdysone signalling regulates the timing of border cell specification [12,13,14]. Within the border cells, the receptor tyrosine kinases EGFR and PVR interpret guidance cues produced by the oocyte to direct anterior migration and later dorsal migration of the cluster [15,16]. Homophilic adhesive interactions between border cells and the nurse cells involving Cadherins are crucial for normal cluster migration [17].
In this report, we identify the miRNA miR-989 as a regulator of border cell migration. We show that border cell migration is delayed in miR-989 mutant egg chambers, and that this phenotype can be rescued by transgenic expression of the miRNA. Moreover, we demonstrate that miR-989 is active in the somatic cells of the egg chamber and required in border cells for efficient migration. Predicted targets encompass most of the pathways known to be involved in regulation of border cell migration.

Results and Discussion
Deep sequencing of an ovarian small RNA library identified miR-989 as the most abundant miRNA species in the Drosophila ovary, constituting 15.9% of all annotated sequencing reads [18].
To test whether miR-989 has an important function during oogenesis, we generated a deletion allele (designated miR-989 KO ) by ends-out homologous recombination [19,20]. Deletion of the miR-989 gene was confirmed by PCR on genomic DNA (not shown). Ovaries derived from young females bearing the miR-989 KO allele in trans to a genomic deficiency (Df(2R)Exel7130) uncovering the miR-989 locus proved to be morphologically normal (not shown).

Delayed border cell migration
We observed that border cell migration was frequently delayed in miR-989 KO / Df(2R)Exel7130 ovaries compared to controls and quantitated this phenotype during two stages of egg chamber development (Fig 2). During late stage S9 and S10A, we measured the distance between the leading border cell and the anterior-most cells in the sheet of follicle cells, as it migrates toward the oocyte ( Fig. 2A). In all control genotypes, border cells kept pace with the advancing sheet of external follicle cells (Fig. 2B). In contrast, the border cell cluster lagged behind the follicular epithelium in homozygous miR-989 KO egg chambers (p,0.001 in comparison to the heterozygous control). Similar results were obtained when the miR-989 KO allele was placed in trans to two independent genomic deficiencies (Df(2R)50C-38 and Df(2R)Exel7130) that uncover the miR-989 locus (p,0.001 compared to all controls, Fig 2B). In wild type egg chambers, border cells typically have reached the oocyte by stage 10B and have begun to migrate toward their final  anterodorsal position. In S10B egg chambers lacking miR-989, border cells were frequently found within the nurse cell compartment. We quantitated this phenotype by scoring whether the border cells had reached the oocyte (''not delayed''), or whether they were still found in the anterior half (''.50% delayed'') or posterior half ('',50% delayed'') of the nurse cell compartment (Fig. 2C). Using this scoring scheme, we found that most border cell clusters in the heterozygous control egg chambers had arrived at the oocyte at stage 10B. Over 1/3 of border cell clusters derived from miR-989 mutant females were delayed. These observations suggested that miR-989 is required for some aspect of border cell migration towards the oocyte.

miR-989 activity in somatic cells
Border cell migration depends on guidance signals from germline cells, signal interpretation by the border cells to produce directed migration and interaction between the two cell types to allow movement of the border cell cluster on and between the nurse cells [5,6,7]. We therefore asked whether miR-989 was acting in the somatic cells or germ line cells of the ovary. As a first step to address this question, we generated transgenic flies that express a miRNA sensor [21] for miR-989. This sensor encodes eGFP expressed ubiquitously under control of a tubulin promoter, followed by a 3' UTR containing two target sites for miR-989. In a wild type background, sensor-directed GFP was present in all the germ line cells, but GFP was not detected in the somatic follicle cells (Fig. 3 A, C). In the miR-989 mutant background, a homogenous GFP signal was also observed in all somatic cells, including the border cell cluster while GFP expression in the germ line cells was unchanged (Fig. 3B, D, arrows). This suggests that miR-989 is predominantly active in the somatic follicle cells.

miR-989 is required in somatic cells for efficient border cell migration
The predominant pattern of follicle cell expression suggested that miR-989 activity may be required in the follicle cells or the border cells to promote normal border cell migration. To test this idea, we generated mosaic egg chambers that were partially wildtype and partially mutant for miR-989 and scored the migratory behaviour of the border cells. We found that wild-type border cells migrated normally when the germ line cells lacked miR-989 (Fig. 4A, left panel). In contrast, migration was delayed when somatic cells including the border cell cluster were mutant for miR-989 but the germ line was wild-type (Fig. 4B, right panel). The quantification of these observations is shown in Figure 4(B, C). Border cell migration was delayed in a statistically significant manner (p,0.001) when all cells in the border cell cluster were mutant for miR-989, but the germ cells were wild-type. We did not observe a statistically significant delay in egg chambers in which the border cells were wild-type but the germline was mutant for miR-989. These results provide evidence that miR-989 is required in somatic cells for normal migration, but dispensable in the germline.
We also observed egg chambers in which the border cell cluster was partially wild type and partially mutant for miR-989 while germ line was wild-type (Fig. 4B, C). Intriguingly, such mixed genotype border cell clusters were also delayed in their migration (p,0.001, compared to control egg chambers). The delay was similar in magnitude to that of completely mutant clusters (p.0.1 between completely and partially mutant border cell clusters). This suggests that lack of miR-989 in just some border cells is sufficient to cause delays affecting the entire cluster. In other words, the presence of wild-type border cells cannot compensate for the lack of miR-989 in some cells. However, we do not exclude the possibility that loss of miR-989 from other somatic cells might contribute to the border cell migration phenotype. The border cell migration defect can be rescued by transgenic miR-989 expression To confirm that lack of miR-989 was responsible for the border cell migration defects described above, we expressed miR-989 from an UAS transgene. Expression with the Slbo-Gal4 driver did not rescue border cell migration in a miR-989 KO mutant background. In contrast, miR-989 expression under the control of a heat-shock inducible actin-flip-out-Gal4 cassette was able to rescue. This technique allowed us to restore miRNA expression in subsets of cells that were positively marked by presence of GFP. GFPnegative border cell clusters lacking miR-989 were delayed in their migration (Fig. 5A, C, D), whereas GFP-positive border cell clusters with transgenic miR-989 expression migrated almost normally (Fig. 5B, C, D; p,0.001). This demonstrates that loss of miR-989 was responsible for the delayed border cell migration phenotype in the miR-989 mutant egg chambers.
miR-989 is expressed in the somatic cells throughout egg chamber development. In the flip-out clonal experiment, Gal4 expression was induced well before the onset of border cell migration. The Slbo gene turns on in stage 8, shortly before border cell migration begins. Expression of a miRNA under Slbo-Gal4 control is expected to take some time to effectively repress its targets. Comparing the Slbo-Gal4 and flip-out clonal rescue results suggests that miR-989 may be required from the onset of border cell migration.
predicted miR-989 targets miR-989 target predictions were obtained from TargetScanFly [22], TargetScanFly ORF [3], MinoTar [3] and miRNA.org [23]. Together, these algorithms identify 724 non-redundant candidate target transcripts. We performed process GO term analysis using the GOrilla interface [24] to identify targets that might be involved in regulating border cell migration ( Figure S1 and Table S1). This analysis revealed 163 enriched GO terms. Among these were potentially informative GO term nodes with lower GO term hierarchy level which were significantly enriched: 'cell migration' (enriched 2.8 fold, p = 1.6610 27 , 31 putative target genes), 'cell adhesion' (enriched 2.6 fold, p = 7.4610 27 , 25 putative targets), 'cell projection morphogenesis' (enriched 2.4 fold, p = 8.6610-5, 23 putative targets), 'transmembrane receptor protein tyrosine kinase signaling pathway' (enriched 2.7 fold, p = 2.9610 24 , 16 putative targets) and 'response to ecdysone' (enriched 4.6 fold, p = 6.3610 24 , 7 putative targets). miRNAs typically downregulate their targets, in part through target transcript destabilization. It is therefore expected that target levels would increase in miRNA mutant cells. The ovary has two main cell types: somatic and germ line. Somatic cells, where miR-989 is expressed, comprise a small fraction of the total tissue. Because the border cells are only a small fraction of the somatic cells, we have not attempted to validate upregulation of candidate targets in the miR-989 mutant border cells by monitoring target RNA levels.
Predicted miR-989 targets associated with the GO terms 'cell migration' and 'regulation of cell projections' are obvious candidates whose de-regulation may cause border cell migration phenotypes. The receptor tyrosine kinase receptors EGFR and Pvr are required in the border cells to receive guidance queues that direct cluster migration towards the oocyte [15,16]. Upregulation of negative regulators of these signalling pathways could delay border cell migration. Ecdysone signaling play a crucial role in regulating the timing of border cell specification, and defects in border cell specification can delay migration [12,13,14].
Predicted miR-989 targets associated with the GO term 'cell adhesion' may be particularly interesting. Border cells migrate collectively in a coordinated fashion, and migration events critically depend on remodelling events that change the adhesive properties of the migrating cells [5,6,7]. In the case of the border cell cluster, border cells have to give up their connection to neighbouring follicular epithelial cells when they initiate migration towards the oocyte. Likewise, while they do need to adhere to the nurse cell membranes, through homophilic interactions of DEcadherin [17] they must not adhere too tightly in order to migrate efficiently. In this light, it is interesting to note that loss of miR-989 from a subset of border cells is sufficient to cause border cell migration delays. Since border cells adhere tightly to each other [5,6,7], it is easily conceivable that few cells that adhere too tightly to the follicular epithelium or the migration substrate would impair migration of the entire cluster. This has been documented for clusters partially mutant for Slbo [25], which affects border cell specification and for clusters partially mutant for DE Cadherin [17]. miR-989 could act on multiple targets including those affecting the processes discussed above to permit normal border cell migration.

Fly stocks
The miR-989 KO allele was generated using pRMCE as described [19,20]. The UAS-miR-989 transgene was cloned into pUAS-T.attB-SLIC [26] and integrated into the landing site 86Fb to generate pT-989@Fb. To generate the miR-989 sensor, oligos encoding two perfect miR-989 binding sites were annealed and cloned into the tub.eGFP transgene [21]. The sensor was then transformed by P-element mediated germ line transformation and individual transgenic lines were established by standard procedures. Oligonucleotide sequences are available on request. The genomic deficiencies Df(2R)Exel7130 and Df(2R)50C-38 and the hsFLP; FRT42A ubiGFPnls strain were obtained from the Bloomington stock centre. Actin-Flipout-Gal4 (AFG) UAS-10xGFP was obtained from the Rørth lab [27]. For mosaic analysis (Fig. 4), males bearing a recombinant FRT42A miR-989 KO chromosome were crossed to hsFLP; FRT42A ubiGFPnls virgins and offspring third instar larvae were heat-shocked for 1-2h at 37uC. For the rescue analysis, third instar larvae of the genotype hsFLP/+; miR-989 KO /miR-989 KO ; pT-989@Fb/AFG,10xGFP were heat-shocked at 37uC for 1-2h and ovaries were then dissected from 2-3d old adult females for analysis.

Immunostainings and Microscopy
1-3d old mated females were collected and placed on wet yeast for 12-24 h prior to dissection. Ovaries were dissected into Schneider's S2 cell medium supplemented with 50 ml/ml FCS and 1ul 10mg/ml Insulin (Sigma) per ml medium. Following dissection in 4% formaldehyde, samples were rinsed and washed in PBST (PBS +0.3% Triton-X), then blocked in PBSTB (PBST +3% BSA) and incubated overnight in primary antibody (rat anti slbo 1:500) at 4uC. The ovaries were then extensively washed with PBSTB at room temperature and incubated overnight in secondary antibodies (AlexaFluor antibodies (Invitrogen), 1:250 of the 1:1 glycerol stocks). If applicable, Alexa633-or Alexa555-conjugated Phalloidin was added at 1:150 (Invitrogen). Samples were then washed extensively and counterstained with DAPI. For sensor analysis, ovaries were fixed as above, extensively washed in PBST and counterstained with DAPI. Images were obtained either on a Zeiss Axio Imager 2.0 or a Zeiss LSM700 confocal microscope. Images were analysed with Zeiss Axiovision software. Graphs and statistical analysis were done with Microsoft Excel.
GO term analysis of predicted miR-989 targets miR-989 target lists were obtained from TargetScanFly [22], TargetScanFly ORF [3], MinoTar [3] and miRNA.org [23], standardized to CG gene name annotation using Flybase [28], pooled and then pruned of redundant entries using Excel. A genome-wide list of CG annotated genes in Drosophila melanogaster was obtained from Flybase as a reference pool. GO term analysis was performed using the GOrilla platform (http://cbl-gorilla.cs. technion.ac.il/), querying all Drosophila ontologies using the miR-989 target predictions as target set and all Drosophila genes as a background set. Figure S1 GO term analysis of predicted miR-989 targets. Figure S1 shows a directed acyclic graph created by the GOrilla interface. It shows GO term enrichment of predicted miR-989 targets (color coded).

(JPG)
Table S1 GO terms enriched among the predicted miR-989 targets. Table S1 shows an annotated list of the GO terms that are enriched among predicted miR-989 targets. (XLSX)