BIOLOGICAL OVERVIEW

Migration of border cells during Drosophila oogenesis is a simple and attractive model system in which to address the signaling pathways and mechanisms responsible for guiding cell migration in vivo. Pvr, a receptor tyrosine kinase related to mammalian PDGF and VEGF receptors, acts in border cells to guide them to the oocyte. The oocyte is the source of a ligand for Pvr, PDGF/VEGF factor 1 (Pvf1). Intriguingly, the guidance function of Pvr is largely redundant with that of Egfr. Rac and the Rac activator Myoblast city/DOCK180/CED-5 are implicated as mediators of the guidance signal (Duchek, 2001b).

Border cells are a cluster of 6-10 specialized somatic follicle cells which perform a stereotypic migration during oogenesis. At the beginning of stage 9 of oogenesis, border cells delaminate from the anterior follicular epithelium and initiate their migration between the germline derived nurse cells, toward the oocyte. About 6 hr later, at stage 10, the border cells reach the oocyte, and then migrate a short distance dorsally toward the germinal vesicle (GV). Thus, the migration occurs in two steps: an initial posteriorly directed (oocyte directed) migration, and subsequently, a shorter, dorsally directed migration. The migration of border cells is essential for female fertility (Duchek, 2001b and references).

EGF receptor (Egfr) signaling is responsible for guiding the second part of border cell migration, the dorsal migration (Duchek, 2001a). However, Egfr signaling is not essential for the first phase of migration of border cells toward the oocyte, indicating that an additional cue must direct this migration. A Drosophila ligand of the PDGF/VEGF family (Pvf1) and its receptor, PDGF/VEGF Receptor (Pvr), are required for the first phase of border cell migration. Pvr and Egfr act in a partially redundant manner to guide border cells to the oocyte. Pvr affects actin accumulation in follicle cells through Myoblast city (Mbc) and Rac. In addition, both Mbc and Rac are required for normal border cell migration. Together, these observations suggest that Pvr signaling controls actin accumulation via Mbc and Rac in migrating border cells (Duchek, 2001b).

Cell migration is guided by one or more spatial (guidance) cues. It was reasoned that uniform expression of a key guidance cue, or a rate-limiting component in its production, throughout the target tissue could be expected to confuse the migrating cells and thus cause inefficient migration. To identify candidate guidance molecules for border cell migration, a gain-of-function genetic screen was used. Controlled ectopic expression of random genes in the genome can be obtained using the modular misexpression, or EP element, system. In a screen of 8500 EP insertion lines, two lines were identified that cause inefficient border cell migration when overexpressed in the germline, consistent with the possibility that a guidance cue was being expressed. The first of these lines directed expression of the EGFR ligand Vein. For the second line, EPg11235, sequencing of flanking DNA showed that the EP element was positioned to drive expression of transcripts corresponding to the predicted gene CG7103. The gain-of-function phenotype was reproduced by expressing a corresponding cDNA. This cDNA was sequenced and shown to encode a protein with a signal sequence and a PDGF domain. It shows highest similarity to PDGF and VEGF ligands from vertebrates. It was therefore called PDGF/VEGF 1 (Pvf1). Pvf1 affects border cell migration whether overexpressed uniformly in the germline or in border cells themselves, consistent with production of a secreted molecule. From an existing collection of P element insertions, one line, EP1624, was found to have an insertion in the first intron of the Pvf1 gene. This insertion is a loss-of-function mutant of Pvf1 (Pvf1¹⁶²⁴) with no detectable transcript remaining in the ovary (Duchek, 2001b).

To investigate whether Pvf1 could serve as a guidance cue for the migration of border cells to the oocyte, its expression was analyzed. The Pvf1 transcript was detected in the germline of the ovary at mid-oogenesis, more concentrated toward the oocyte. To look at the protein expression directly, anti-Pvf1 antisera was raised. The anti-Pvf1 sera shows specific staining in the ovary, which is absent from mutant egg chambers. Pvf1 is detected in the oocyte at stage 7 and at stage 8, filling the cytoplasm. At stage 9, when border cells have initiated migration, Pvf1 is still enriched in the oocyte, but now only in the subcortical area of the large oocyte. Thus, the oocyte appears to be the major site of Pvf1 protein production. Pvf1 is expressed both before and during border cells migration, consistent with the possibility that Pvf1 serves as an attractant for border cells (Duchek, 2001b).

The Pvf1¹⁶²⁴ mutant is homozygous viable, and analysis of egg chambers from mutant females reveals minor delays in border cell migration. However, two additional PDGF, VEGF-like ligands appear to exist in Drosophila. To overcome the potential redundancy between PVF ligands, a PVF receptor was sought in order to directly investigate its role in border cells (Duchek, 2001b).

Gene predictions indicate that the Drosophila genome contains a single gene encoding a protein related in sequence and structure to mammalian PDGF and VEGF receptors. The protein has been called Pvr for PDGF/VEGF receptor. It appears to be the only Drosophila member of this family of receptor tyrosine kinases, and thus could be the receptor for all three PVF ligands. Pvr transcripts are detected in mRNA from ovaries and from embryos. Pvr mRNA is detected in embryonic hemocytes and in Schneider cells, the related tissue culture cells. Using a specific antiserum directed against the C-terminal tail of Pvr, endogenous Pvr protein was detected in Schneider cell extracts as an approximately 180 kDa protein, corresponding well to the predicted molecular weight of 170 kDa (Duchek, 2001b).

Does Pvr mediate the effect of Pvf1 on border cell migration and is the effect direct? Immunofluorescence analysis of wild-type ovaries indicates that endogenous Pvr protein is present in all follicle cells and thus might respond to Pvf1. Pvf1 has been identified based on the ability of uniform expression to impede border cell migration. Direct uniform activation of the Pvf1 receptor in border cells should give the same effect or a stronger effect. To test whether Pvr would do this, an activated form of the receptor, lambda-Pvr, was made. This was done by exchanging the normal extracellular ligand binding domain for a constitutive dimerization domain, as has been done for other receptor tyrosine kinases. The Gal4-UAS system and the slboGal4 driver were used to drive expression of lambda-Pvr in border cells, centripetal cells, and a few other follicle cells. Lambda-Pvr is functional in vivo since it stimulates the MAP-kinase pathway (dpERK staining). Expression of lambda-Pvr in border cells also completely blocks their migration. In over 90% of control stage 10 egg chambers, border cells had reached the oocyte, and the rest were only slightly delayed. In contrast, almost none of the border cell clusters expressing lambda-Pvr had moved at all. Thus, uniform activation of Pvr in border cells blocks migration, as expected for a guidance receptor (Duchek, 2001b).

Ectopic expression of the ligand Pvf1 has a detectable but modest effect on migration: all border cell clusters have moved by stage 10, and one-third have arrived at the oocyte. Increased expression of the wild-type Pvr receptor in border cells has, on its own, a negligible effect on migration, but it sensitizes the cells to ectopic expression of Pvf1. Upon coexpression of Pvr and Pvf1, one-fourth of the stage 10 border cell clusters are at the oocyte, but another fourth have not moved at all. The effect is specific to Pvf1, since border cells are not sensitized to ectopic expression of the Egfr-ligand Vein. In fact, Pvr overexpression ameliorates the effect of ectopic Vein expression. The synergy between Pvf1 and Pvr expression supports a specific interaction between the two proteins on border cells (Duchek, 2001b).

To investigate whether Pvr is required for guiding border cell migration to the oocyte, a dominant negative form of the receptor, DN-Pvr, was generated. DN-Pvr was made in the same way as the highly specific dominant negative Egfr. DN-Pvr contains only the extracellular and transmembrane domains of the receptor, allowing it to sequester ligand as well as to form inactive dimers with the endogenous receptor, and thus specifically attenuate signaling from this receptor. When expressed in border cells, DN-Pvr causes some delay of posterior migration. This result was confirmed by quantification of migration at stage 10. Upon expression of DN-Pvr, less than 60% of border cell clusters had reached the oocyte. This phenotype is similar to that seen in Pvf1 homozygous mutant females, indicating that Pvf1 is the major endogenous ligand for Pvr in this context. Thus, Pvr signaling, and lack thereof, affects the efficiency of border cell migration, but it is not essential for the process (Duchek, 2001b).

In addition to Pvr, Egfr also has properties consistent with a role in guiding border cells to the oocyte: both receptor tyrosine kinases are expressed in border cells, and their ligands are found in key locations in the germline. Both give similar gain-of-function effects, and both dominant negative receptors give subtle effects with respect to migration to the oocyte. One possible explanation for the subtle dominant negative effects is that the receptor/ligand pairs are partially redundant. This possibility was first addressed by coexpressing both dominant negative receptors in border cells. This gave a very dramatic effect. Border cells expressing both dominant negative receptors migrate very inefficiently. When quantified at stage 10, 90% of border cell clusters expressing both dominant negative receptors had migrated less than halfway to the oocyte. In 5% of egg chambers, border cell clusters were found off the direct track to the oocyte. This suggests that the cells are motile but poorly guided. This 'off track' phenotype is not observed in wild-type egg chambers or in egg chambers where border cell migration is impaired for another reason (slbo mutant) (Duchek, 2001b).

The effect of expressing dominant negative receptors in Pvf1¹⁶²⁴ mutant egg chambers was also tested. As expected, the Pvf1¹⁶²⁴ mutant phenotype is not made worse by removing activity of its cognate receptor, Pvr. However, reducing activity of the other pathway by expression of dominant negative Egfr has a strong effect. Border cells are not able to reach the oocyte by stage 10, and they also show a low level of 'off track' migration. This confirms the redundancy of function for the two receptors, as well as their ligand specificity. Thus, if either Egfr or Pvr (and corresponding ligand) are left intact, border cells can find the oocyte, but if both receptor functions are severely affected, they cannot. That Egfr is uniquely required for dorsal migration of border cells is explained by the ligand distribution. Only Egfr ligands are expressed differentially on the dorsal side. Gurken is expressed by the dorsally located germinal vesicle, and the protein is found in a gradient originating from there. Spitz and Vein are expressed in dorsal follicle cells (Duchek, 2001b).

These results indicate that Pvr and Egfr are guidance receptors for border cell migration toward the oocyte. A guidance function implies that the critical parameter for proper migration is the differential distribution of signal (ligand) rather than absolute level of signaling. This is supported by the observation that increased expression of Pvr in border cells suppresses the effect of the ectopically expressed Egfr ligand, Vein. The level of Pvr+Egfr signaling in border cells is likely higher upon coexpression, but the signal distribution might be more normal due to increased sensitivity to the spatially graded Pvr ligand relative to the ectopically expressed Egfr ligand. To test the importance of signal distribution versus level more directly, signaling from one receptor was reduced by expression of its dominant negative form and it was asked whether the deleterious effect of ectopic ligand for the other receptor would be enhanced or suppressed. For guidance signaling, the expectation is that cells which can only respond to one type of ligand will require this ligand to be properly distributed and thus be very sensitive to its misexpression. If just the correct level of signal is required, then simultaneously increasing and decreasing signaling should give a less severe phenotype than either alone. The experiment was done for both receptors, and in both cases, a strong enhancement of the migration defect was seen. Ectopic expression of one ligand and the dominant negative form of the other receptor causes a phenotype similar to one expressing both dominant negative receptors: border cells do not reach the oocyte at stage 10. They usually had migrated less than halfway, and sometimes were found off track. As expected, coexpression of a ligand with a dominant negative version of its cognate receptor has little or no additional effect. These results indicate that both receptors receive directional information which guides cell migration. Migration can proceed to some extent if only one receptor receives nonuniform (directional) signaling, consistent with a partially redundant guidance function (Duchek, 2001b).

Pvr affects actin accumulation in follicle cells through Myoblast city (Mbc) and Rac. In addition, both Mbc and Rac are required for normal border cell migration. Together, these observations suggest that Pvr signaling controls actin accumulation via Mbc and Rac in migrating border cells. Other receptor tyrosine kinases may also use this signaling module to guide cell migration in vivo. mbc was first identified in Drosophila based on its requirement in myoblast fusion. Mbc has since been implicated in multiple processes requiring cytoskeletal reorganization, and is intriguingly expressed in early germ cells of the embryo, which undergo a guided migration. Genetic data indicate that mbc, as well as the C. elegans homolog ced-5, acts as an upstream activator of Rac. No GTP-GDP exchange activity has been shown for Mbc/DOCK180/CED-5, but Mbc/DOCK180/CED-5 interacts with nucleotide-free Rac, indicating that it plays a role in activation or localization of Rac. The small adaptor protein Crk interacts specifically with Mbc/DOCK180/CED-5 in all three systems. In mammalian cells, Crk and another adaptor protein, p130-CAS, have been shown to regulate cell migration in a Rac-dependent manner. Crk, CAS, and DOCK180 regulate membrane ruffling in a Rac-dependent manner. In C. elegans, CED-5, CED-2 (Crk), and CED-10 (Rac) are required for normal distal tip cell migration as well as cell engulfment, but the receptors regulating this behavior are not known. Cell engulfment (phagocytosis) by mammalian 293T cells involves the alpha_vß₅ integrin receptor which, in an unknown manner, can stimulate the formation of a p130-CAS-Csk-Dock180 complex and also activation of Rac1. Thus, Mbc/DOCK180/CED-5 and Rac are linked in a well conserved signaling module that affects cell behavior, including migration. With the Drosophila Pvr receptor identified, it should now be possible to determine how this guidance receptor affects Mbc and Rac (Duchek, 2001b and references therein).

Signaling through Mbc and Rac is unlikely to be the only effect of the guidance receptors in border cells. mbc null clones give a phenotype which is stronger than loss of signaling from either receptor alone, but not as severe as loss of both Pvr and Egfr activities. Egfr acts partially redundant with Pvr in guiding border cells, but preliminary evidence suggests that Egfr may act differently from Pvr. Pvr may also have additional effects, given that the dominant effect of activated Pvr on the actin cytoskeleton is strongly attenuated but not abolished in mbc null clones. Thus, the receptor pathways may be only partially overlapping, and other effectors are likely to contribute to the complicated task of guiding cell migration in vivo. Many candidate signaling molecules have been tested for their requirement in border cell migration: MAPK pathway, PI3K, PLC-gamma, as well as RTK adaptors, DOCK, Trio, and Pak. None of these is (individually) required; thus, Mbc and Rac remain the only identified downstream signaling effectors in this context. A number of other genes have been shown to be important for border cell migration, but these are either transcription factors and modulators thereof (which are likely to affect cell fate) or components of the basic cellular machinery for movement/adhesion (Duchek, 2001b and references therein).

Receptor tyrosine kinases serve multiple roles during development. The ability of Pvr to activate the MAP-kinase pathway may be important for control of cell growth and differentiation in other tissues, as is the case for Egfr. Both Egfr and Pvr retain the ability to activate the MAP-kinase pathway when serving the guidance receptor function in border cells, indicating that they can simultaneously display multiple signaling properties. The mammalian PDGF receptors and VEGF receptors also have multiple functions during development and in tissue culture cells, including effects on proliferation and on cell migration. Which pathways downstream of these receptors are critical for which function, and how cell type specific responses are generated, remains an important question. Studies of mice with targeted mutations in specific tyrosines of PDGFR-ß indicate that the requirements for specific docking sites in vivo are not easily predicted from the effect of the same mutations in tissue culture cells (Tallquist, 2000). There may be more compensation and redundancy in vivo, or the importance of different pathways may simply differ in vivo and in tissue culture. In either case, such findings underscore the importance of in vivo analysis. Guidance signaling has been analyzed in a simple and well defined cell migration process in vivo. Based on analyses of mutations in signaling pathway components, the essential involvement of some pathways can be ruled out and another putative pathway (Mbc-Rac) downstream of Pvr (and Egfr) can be implicated. There may be some redundancy in downstream pathways leading to guided border cell migration. In the case of MAP-kinase pathway and PI3K, both loss-of-function and gain-of-function (constitutive activation) mutations were investigated. If a signaling molecule was instructive but redundant, then ubiquitous activation would probably have some effect. However, border cell migration was unaffected, arguing that these signaling molecules do not play instructive roles (Duchek, 2001b).

It is intriguing that, even in this simple cell migration system, there is substantial redundancy between the guidance cues (and between the guidance receptors). It is not a priori obvious that these two different types of receptor tyrosine kinases should show such overlap in function. Redundancy in biological functions of receptor tyrosine kinases is likely to be even more prominent in mammalian systems, which have multiple receptors of each type. Subtle effects of individual factors and genetic redundancy are more the rule than the exception in analysis of axon guidance. It is interesting to speculate that partial reliance on multiple signals is biologically advantageous for continuously and subtly modulated processes such as guidance, as compared to all-or-none cell fate determination switches (Duchek, 2001b).

GENE STRUCTURE

The Pvr cDNA sequences unveiled a total of 21 exons spanning 20 kb genomic DNA. Four different alternatively spliced forms were identified, differing in sequences encoding the three amino-terminal Ig homology domains of the extracellular segment, while the segment encoding the transmembrane and intracellular domains was found to be identical in all clones. RT-PCR experiments allowed the cloning of splice variants #1 (SPV-1) and SPV-4. Several clones of SPV-1 were isolated, while only one clone of SPV-4 was picked. SPV-2 and -3 represented by the Schneider cell line EST cDNA clones could not be detected in embryonic mRNA (Heino, 2001).

cDNA clone length - 5264

Bases in 5' UTR - 304

Exons - 21

Bases in 3' UTR - 430

PROTEIN STRUCTURE

Amino Acids - 1509

Strucural domains

The most common splice variant #1 (SPV-1) encodes a polypeptide of 1509 residues, containing a signal sequence followed by seven Ig domains, a single transmembrane domain and a split tyrosine kinase domain on the cytoplasmic side. This overall domain organization matches the mammalian receptor organization. In the other variants, Ig domains 1-3 are modified by deletion of a peptide segment between domains 2 and 3 (SPV-2) or introduction of a stop codon after domain 1 (SPV-3). In SPV-4, no putative signal sequence could be identified by computer scanning (Heino, 2001).

Evolutionary distances were calculated between several putative or established receptor tyrosine kinases from the complete fly genome and known vertebrate VEGF/PDGF family receptors. This analysis grouped the Drosophila VEGF receptor together with vertebrate VEGF and PDGF receptors. Only one other putative fly kinase sequence (CG3277) clusters with DmVEGFR. However, besides the putative kinase domain, it has no other similarity with the mammalian receptors studied. Analysis of the extracellular domains suggests that DmVEGFR is closer to VEGFR than to PDGFRs. The seven Ig homology domains of the fly receptor also emphasize the closer kinship to the VEGFRs. Taken together, these studies clearly show that there is only one Drosophila gene coding for a VEGF/PDGF receptor (Heino, 2001).

BLAST sequence homology searches of the Drosophila genome and cDNA sequences have identified a single gene (CG8222, cytological region 29A) that encodes a protein with substantial similarity to mammalian VEGFRs. It is called here Vegfr, but it is also known as Pvr (Duchek, 2001b). Three mRNA splice forms have been deduced from multiple overlapping and three full cDNA sequences. The splice forms encode distinct proteins. Isoforms A (147 kDa) and B (150 kDa) are predicted Type I transmembrane proteins with an N-terminal signal peptide, seven Ig-like repeats, a transmembrane domain, and a split tyrosine kinase domain. The tyrosine kinase domain is most similar to those of vertebrate VEGFR1 (Flt-1) and VEGFR2 (KDR/FLK-1) but also shows substantial similarity to other members of the split RTK family including PDGFR-ß. There is additional similarity to the vertebrate receptors in the juxta membrane region and Ig-like repeats. Like vertebrate VEGFRs, the ectodomain of the Drosophila proteins contains seven Ig-like repeats, which distinguishes them from other members of the split RTK family, such as PDGFRs which contain only five. Isoform B differs from A by one amino acid in place of seven between the second and third Ig-like repeats and by 35 additional amino acids in the kinase insert domain. Isoform C (25 kDa) is a truncated protein consisting of the signal peptide and the first two Ig-like repeats. It may be able to bind ligand, because only the second Ig-like repeat of vertebrate VEGFR1 is essential for ligand binding (Cho, 2002).

Identification of VEGF/PDGF-like genes

The VEGFs of vertebrates are characterized by the presence of a so-called VEGF/PDGF homology domain. Using a profile derived from such domains, the Drosophila genomic sequences were searched and a distinct segment on a database entry originating from the X chromosome (cytology 17E1-2) was identified. Three EST clones, LD28763, LD30344 and LD37208 with sequences matching this segment are available from the Berkeley Drosophila Genome Project. They were sequenced and the corresponding gene was named PDGF- and VEGF-related factor 1 (EMBL/GenBank/DDBJ accession number AJ401391). The sequence displays a six exon gene structure spanning about 10 kb of DNA. In Pvf-1 the two clones LD28763 and LD30334 have been shown to be products of alternate splicing affecting the second exon. In the longer version (LD28763), the first ATG (methionine) is 11 triplets (amino acids) more upstream than the first ATG in the shorter version. It was not experimentally established which one is de facto used. Conceptual translation yields a polypeptide containing a putative 22 residue signal sequence in the shorter splice variant and a typical VEGF/PDGF homology domain (named VHD in the Drosophila protein and PDGF in PFAM), characterized by eight similarly spaced cysteine residues. Moreover, the C-terminal sequences following the VHD-domain contain an additional pattern of cysteine residues similar to those of the C-terminal 'silk'-homology domain in VEGF-C (Heino, 2001).

A cDNA clone representing another VEGF/PDGF-like gene was also isolated and sequenced (EMBL AJ312312). This cDNA corresponds to the PDGF- and VEGF-related factor 2 (Pvf2) gene (CG13780). The conceptual translation of this gene yields a 405 amino acids long peptide. A third gene, Pvf3(CG13782), also containing a VEGF/PDGF homology domain was identified on a genomic segment only ~25 kb apart from the CG13780 gene. None of these genes seem to code for all the typical cysteines in the VEGF/PDGF homology domain. Pff1 shows the best similarity to the VEGF/PDGF homology domain and the other two genes were therefore not further studied in this context (Heino, 2001).

An evolutionary analysis comparing the fly and human VEGF/PDGF domains enabled an exhaustive calculation of phylogenetic relationship, based on which a single best tree was proposed. According to this, the fly factor Pvf2 is most closely related to the mouse VEGF-B and to the PDGF A and B chains (Heino, 2001).

date revised: 28 October 2001

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.