Transcriptome analysis reveals temporally regulated genetic networks during Drosophila border cell collective migration

Abstract

Background: Collective cell migration underlies many essential processes, including sculpting organs during embryogenesis, wound healing in the adult, and metastasis of cancer cells. At mid-oogenesis, Drosophila border cells undergo collective migration. Border cells round up into a small group at the pre-migration stage, detach from the epithelium and undergo a dynamic and highly regulated migration at the mid-migration stage, and stop at the oocyte, their final destination, at the post-migration stage. While specific genes that promote cell signaling, polarization of the cluster, formation of protrusions, and cell-cell adhesion are known to regulate border cell migration, there may be additional genes that promote these distinct active phases of border cell migration. Therefore, we sought to identify genes whose expression patterns changed during border cell migration.

Results: We performed RNA-sequencing on border cells isolated at pre-, mid-, and post-migration stages. We report that 1,729 transcripts, in nine co-expression gene clusters, are temporally and differentially expressed across the three migration stages. Gene ontology analyses and constructed protein-protein interaction networks identified genes expected to function in collective migration, such as regulators of the cytoskeleton, adhesion, and tissue morphogenesis, but also uncovered a notable enrichment of genes involved in immune signaling, ribosome biogenesis, and stress responses. Finally, we validated the in vivo expression and function of a subset of identified genes in border cells.

Conclusions: Overall, our results identified differentially and temporally expressed genetic networks that may facilitate the efficient development and migration of border cells. The genes identified here represent a wealth of new candidates to investigate the molecular nature of dynamic collective cell migrations in developing tissues.

Keywords: Adhesion; Collective cell migration; Morphogenesis; Oogenesis; RNA-seq; Ribosome.

Fig. 1

RNA sequencing of isolated populations of border cells reveals temporal changes in gene expression. (A-C) Representative egg chambers showing the patterns of slbo-mCD8:GFP expression in the border cell cluster prior to (A), during (B), and after migration (C). (D) Representative migrating border cell cluster showing the pattern of slbo-mCD8:GFP expression in individual border cells (expressing Eyes absent, Eya) and polar cells (FasIII-positive, marked with asterisks). (E) Schematic of how the mCD8:GFP-positive border cells were sorted and selected using mCD8-bound magnetic beads. RNA from isolated border cells was then prepared for RNA sequencing and downstream analyses. (F) Heatmap of all sequencing results, identifying significant differential expression for 1,729 transcripts from isolated border cells during their migration (EBSeq-HMM FDR < 0.05). Several border cell migration-related genes, along with their transcript number (“FBtr”), are highlighted. Scale bars represent 10 μm (A-C) or 5 μm (D). All significantly differentially expressed transcripts are shown in Supplemental Data 2. HiSeq 2500 graphic courtesy of Illumina, Inc

Fig. 2

Migration and development-related genes are temporally expressed during border cell migration. (A-H) Heatmaps of significantly differentially expressed transcripts (EBSeq-HMM FDR < 0.05), focused on subsets of migration-related (A-F) or developmentally related (G, H) gene categories. The primary literature or other databases were used to identify genes and GO terms (see Methods for details). Heatmaps represent differential expression in border cells at pre-, mid-, or post-migration. A z-score of 1 (shown in red) signifies up-regulation; a z-score of -1 (shown in blue) signifies down-regulation. (A-C) Differentially expressed transcripts known to be required or expressed in border cells during their migration (A, 496 transcripts, representing 136 unique genes), or shown to be significantly upregulated in border cells versus non-migratory follicle cells from two microarray studies, Borghese et al. [18] (B, 1145 transcripts representing 475 unique genes) and Wang et al. [19] (C, 853 transcripts, 297 unique genes). (D-H) Transcripts for migration-related (D-F) or developmentally-related (G, H) genes, including the actin cytoskeleton (D, 154 transcripts, 38 unique genes), adhesion (E, 147 transcripts, 46 unique genes), epithelial-to-mesenchymal transition (EMT; F, 191 transcripts, 66 unique genes), oogenesis (G, 625 transcripts, 184 unique genes), or transcription factors (H, 701 transcripts, 263 unique genes). All data are shown in Supplemental Data 3

Fig. 3

Differentially co-expressed transcripts are up-regulated during border cell migration and enriched for shared biological functions. (A-C, left graphs) Significantly differentially expressed transcripts sorted by clust into shared co-expression patterns in pre-, mid-, and post-migration stages. (A-C, right graphs) Metascape pathway and process enrichment analysis results for each co-expression cluster, showing the most significantly enriched terms. N, number of genes enriched for a given annotation term; bars show significance of annotation terms, sorted by p values (-log10P; darker color, more significant values). Note that genes can be found in multiple Metascape annotation categories. Clusters C0 (A), C1 (B), and C8 (C) show patterns of increased expression during border cell migration. All data are shown in Supplemental Data 4 and Supplemental Data 5

Fig. 4

Differentially co-expressed transcripts are down-regulated during border cell migration and enriched for shared biological function. (A-C, left graphs) Significantly differentially expressed transcripts sorted by clust into shared co-expression patterns in pre-, mid-, and post-migration stages. (A-C, right graphs) Metascape pathway and process enrichment analysis results for each co-expression cluster, showing the mostly significantly enriched terms. N, number of genes enriched for a given annotation term; bars show significance of annotation terms, sorted by p values (-log10P; darker color, more significant values). Note that genes can be found in multiple Metascape annotation categories. Cluster C4 (A), C5 (B), and C6 (C) show patterns of decreased expression during border cell migration. All data are shown in Supplemental Data 4 and Supplemental Data 5

Fig. 5

Differentially co-expressed genes form protein-protein interaction networks. Physical protein interaction (PPI) network analysis of gene products from selected co-expression clusters. Functional annotation keywords were used to assign color to proteins in the networks. (A) Co-expression cluster C0, encompassing genes that increase expression during migration, contains PPI nodes with migration-related functions (pink), including functions known to regulate border cell migration such as septate junction regulation (dark blue) and ecdysone response function (mustard). Individual networks consist of proteins with biosynthesis/metabolism (cyan, upper right) and migration-related (pink, lower right) functions. (B) Co-expression cluster C8, encompassing genes that increase expression, forms one major protein interaction network with migration-related (pink), multiple signaling pathways (light green), and gene expression regulation (light blue) functions. One individual network consists of proteins with migration-related and additional categories (upper left). (C) Co-expression cluster C6, encompassing genes that decrease during migration, forms large nodes for ribosome function (magenta), biosynthesis/metabolism (cyan), and regulation of gene expression (light blue). Individual networks consist of proteins with biosynthesis/metabolism (cyan, lower center). For each PPI network, the “other” category (gray) either denotes genes that do not have available FlyBase annotations/data or for which three or fewer genes were annotated. All data, including annotations and keywords, are shown in Supplemental Data 6

Fig. 6

Networks of immune signaling and ribosome biogenesis genes differentially expressed during border cell migration. Graphical representation of the networks of differentially co-expressed genes in border cells annotated with immune (A) or ribosome-related (B) functions. (A) Co-expressed genes enriched for immune functions were sorted into Toll Signaling, JNK Signaling/Cell Death, Imd Signaling, or Immune Response categories. Genes involved in defense against fungus, virus, bacteria, or parasitoid wasp, as well as genes implicated more broadly in immune cell or cell death functions, but not linked to a signaling pathway, comprise the “immune response” category. (B) Co-expressed genes enriched for ribosome function were sorted into ribosome biogenesis in the nucleolus, nuclear export, large and small ribosomal subunit, translation/modification and other categories as indicated. For simplicity, the total number of genes enriched for small ribonuclear protein component functions is shown in the cytoplasm; a subset of these genes have predicted or demonstrated nuclear localization. All data are shown in Supplemental Data 7 (A, immune functions) and Supplemental Data 8 (B, ribosome functions)

Fig. 7

Expression patterns of genes in migrating border cells in vivo. Fluorescence RNA in situ hybridization patterns of differentially expressed genes in migrating border cells from the transcriptomics analysis in migrating border cells using data and images from the Dresden Ovary Table (DOT) [62, 63]. Representative images of stage 9 and/or stage 10 egg chambers were chosen for given genes of interest. RNA signal is in green and DAPI labels the nuclei in magenta. Anterior is to the left in all images. (A-F) RNA signal (green) is shown in pre-, mid-, and post- migratory border cell clusters. (G-L) RNA signal (green) is shown at one stage of migration. (A, B) Expression of slbo (A) and sn (B), two genes known to regulate border cell migration. (C-L) Expression of multiple genes identified from the transcriptome analyses with previously uncharacterized roles in border cell migration. (M) Table of genes found in the DOT, along with their annotated expression patterns in the ovary, based on all available images, and location within one of the significantly differentially co-expressed gene clusters (SigDE Cluster). Arrowheads mark the position of border cells. All data are shown in Supplemental Data 9

Fig. 8

Functional assessment of temporally expressed genes in border cell migration. Validation of differentially expressed genes in border cell migration using RNAi. (A-C) Pattern of c306-GAL4, driving UAS-LacZ (cyan), in stages 9–10 pre-, mid-, and post-migratory border cells and a few additional follicle cells. DAPI (gray) labels all nuclei in the egg chambers. (D-K) Knockdown of candidate genes driven by c306-Gal4; genotypes are c306-Gal4; tsGAL80 > UAS-RNAi. (D) Migration of border cells was scored based on the percentage of border cells that completed their migration (last quartile of the migration distance; 76–100%, blue) by stage 10 of oogenesis, by which time migration should be complete, or if the border cells stalled along the migration pathway, shown as the quartile distance migrated away from the anterior tip of the egg chamber (0–25%, yellow; 26–50%, pink; 51–75%, green) as shown in the schematic. (E) Quantification of migration defects for genes from the transcriptome analyses, along with the negative control (mCherry RNAi) and positive controls bazooka (Baz) and Rap1 RNAi, two genes known to regulate border cell migration. Shown are results for knockdown of 8 genes that resulted in significant migration defects at stage 10 (line); Ches-1-like (asterisk) RNAi had a strong migration defect but may partly be due to off-target effects. (F-K) Images of border cell migration at stage 10 for control (F, G) and experimental RNAi (H-K) egg chambers. Border cells (arrowheads) are labeled with E-cadherin (magenta); DAPI (gray) labels the nuclei of all cells. (F) Complete migration in a stage 10 mCherry RNAi control egg chamber. (G) RNAi knockdown of a positive control, Baz, to show representative migration defects. (H-K) Migration defects in Arf51F (H), serrano (sano; I), Bsg (J), and Cip4 (K) RNAi egg chambers. Anterior is to the left in all images. Arrowheads mark the position of border cells. All scale bars represent 10 μm. RNAi knockdown for each gene/line was performed in triplicate, with an average of 111 egg chambers per replicate. Exact N values, raw data and reagents used are available in Supplemental Data 10