A Suppressor/Enhancer Screen in Drosophila Reveals a Role for Wnt-Mediated Lipid Metabolism in Primordial Germ Cell Migration

Wnt proteins comprise a large family of secreted ligands implicated in a wide variety of biological roles. WntD has previously been shown to inhibit the nuclear accumulation of Dorsal/NF-κB protein during embryonic dorsal/ventral patterning and the adult innate immune response, independent of the well-studied Armadillo/β-catenin pathway. In this paper, we present a novel phenotype for WntD mutant embryos, suggesting that this gene is involved in migration of primordial germ cells (PGC) to the embryonic gonad. Additionally, we describe a genetic suppressor/enhancer screen aimed at identifying genes required for WntD signal transduction, based on the previous observation that maternal overexpression of WntD results in lethally dorsalized embryos. Using an algorithm to narrow down our hits from the screen, we found two novel WntD signaling components: Fz4, a member of the Frizzled family, and the Drosophila Ceramide Kinase homolog, Dcerk. We show here that Dcerk and Dmulk (Drosophila Multi-substrate lipid kinase) redundantly mediate PGC migration. Our data are consistent with a model in which the activity of lipid phosphate phosphatases shapes a concentration gradient of ceramide-1-phosphate (C1P), the product of Dcerk, allowing proper PGC migration.


Introduction
Signaling pathways such as those activated by the Wnt family of secreted ligands control a wide variety of biological processes. WntD (Wnt inhibitor of Dorsal, CG8458 and formerly annotated Wnt8) is a particularly interesting member of this family that is likely specific to Drosophila [1,2]. WntD inhibits nuclear accumulation of the NF-kB homolog Dorsal during dorsal/ventral patterning and during the adult innate immune response. Maternal overexpression of WntD leads to lethal dorsalization of embryos, while wntD KO1 mutants display a mild expansion of nuclear Dorsal protein into the anterior and posterior poles, suggesting that, in vivo, WntD restricts the region of Dorsal protein activity [3,4]. Furthermore, wntD KO1 mutant adults exhibit an aberrant innate immune response to bacterial infection: a subset of Dorsal target genes encoding antimicrobial peptides is more highly expressed than in WT, before and after infection. wntD KO1 mutants also die earlier upon infection with pathogenic bacteria, suggesting that WntD signaling is critical in modulating the innate immune response to microbial infection [3]. This increased lethality may be due to upregulation of the TNF homolog Eiger and a novel gene called Edin in the wntD KO1 mutants. Overexpression of Edin is sufficient to induce lethality, and eiger mutant flies live longer upon Salmonella infection, suggesting that higher expression of either of these genes is detrimental [5,6].
In follow-up studies on wntD KO1 mutants we discovered another role for WntD in development: regulating migration of the embryonic primordial germ cells (PGC) to the gonad. PGC, also known as pole cells, are the first cells in the Drosophila embryo to be specified. They begin their journey at the posterior pole, and remain outside the embryo proper upon cellularization of the embryo blastoderm [7]. During gastrulation, the PGC are carried into the embryo with the invaginating midgut epithelium. Around stage 9-10, the PGC begin to migrate across the midgut epithelium, initially dorsally along the midgut epithelium, and then sort bilaterally. As bilateral sorting proceeds, the PGC begin to be associated with mesoderm and eventually align with mesodermal somatic gonadal precursors (SGP) around stage 13. Around stage 14, the PGC and SGP coalesce to form the embryonic gonad [7,8].
Although much is known about the cell movements and tissue interactions along the PGC migratory path, and many genes required for proper PGC guidance have been discovered, the identity of specific guidance molecules for the germ cells is yet unknown. For example, several Drosophila mutants in the 3hydroxy-3-methyl-glutaryl-CoA reductase (HMGCR) pathway display defects in PGC migration, and misexpression of some of these genes is sufficient to attract PGC to ectopic tissues. These data suggest that a geranylgeranylated molecule originating in the mesoderm is either required for generation of a chemoattractant, or the geranylgeranylated molecule is secreted and acts as the attractant itself. However, this attractant has not been identified [9,10].
Another major mechanism controlling PGC migration is through the lipid phosphate phosphatases encoded by the Wun and Wun2 genes. PGC avoid tissues with endogenous or ectopic Wun or Wun2 expression, and loss of these genes results in highly disorganized PGC movement [11,12,13]. These studies suggest a model in which a phospholipid gradient shaped by Wun/Wun2 activity controls PGC migration, either through attraction of PGC by a phospholipid or repulsion by a non-phosphorylated lipid. It is yet unclear whether attraction or repulsion is acting upon the PGC in this model. Furthermore, Wun and Wun2 encode broad-specificity lipid phosphate phosphatases, and it is unclear what the relevant substrate is in relation to PGC migration [14,15].
Given the wntD KO1 mutant phenotype in germ cell migration, we reasoned that uncovering WntD signal transduction pathway components might lend insight into the mechanism by which PGC are attracted to the gonad. Interestingly, WntD likely signals independently of the canonical Armadillo (Arm)/b-catenin pathway. Overexpression of WntD in tissues sensitive to increased levels of Arm signaling fails to produce detectable phenotypes, and germline mutations in Daxin, a negative regulator of Arm, do not result in embryonic dorsalization [3] [16]. In order to identify components of the WntD signal transduction pathway, we undertook a genetic screen for suppressors and enhancers of WntD-mediated embryonic dorsalization. Here, we report the results from the screen, as well as our analysis of two suppressors of WntD overexpression: Df(1)Sxl-bt, which removes Fz4 (CG4626), a strong candidate for the WntD receptor; and Df(3R)e1025-14, which removes Drosophila Ceramide Kinase (Dcerk, CG16708), a likely downstream effector.
Although our suppressor/enhancer screen is based on the ability of maternal WntD overexpression to lethally dorsalize embryos through inhibition of Dorsal protein activity, there are many examples of a single signaling pathway regulating multiple biological processes throughout development and homeostasis [17,18]. It would therefore be unsurprising to find that a single pathway transducing the WntD signal can affect both dorsal/ ventral patterning and PGC migration. Here, we provide evidence for ceramide kinase activity in WntD signaling and present a model in which the WntD pathway controls the production of ceramide-1-phosphate (C1P), which might be a substrate of Wun and Wun2 to attract PGC to the gonad.

wntD KO1 mutants display defects in primordial germ cell migration
We detected expression of WntD protein in tissues known to influence PGC migration, such as the invaginating midgut in stage 8 embryos ( Figure S1A), and the midgut and gonad in stage 14 embryos ( Figure S1B). We therefore wondered whether wntD KO1 mutant embryos may exhibit a phenotype in migration of primordial germ cells (PGC) to the gonad. WT and wntD KO1 mutant (maternal and zygotic) embryos were stained with antibody directed against the PGC-specific protein Vasa, and PGC that failed to migrate to the gonad by stage 14-16 were quantified ( Figure 1). Indeed, we observed significantly more mislocalized PGC in wntD KO1 mutants than in WT embryos. We also found a higher frequency of wntD KO1 embryos with more than 2 mislocalized PGC than in WT embryos ( Figure 1D).
A suppressor/enhancer screen to identify WntD pathway components It is of interest to determine how the WntD ligand signals to influence PGC migration and other biological processes. To identify components of the WntD signaling pathway, we performed a screen for suppressors and enhancers of the lethality caused by maternal overexpression of WntD [3]. Briefly, we generated Act-Gal4/+; UASp-WntD/+ females carrying deficiencies (heterozygous) from the Bloomington stock center deficiency kit and mated them to WT males ( Figure S2). Progeny of Act-Gal4/+; UASp-WntD/+ mothers have a low rate of survival to pupal stage, while introduction of various deficiencies to this genetic background can enhance or suppress this lethality ( Figure 2). Criteria for categorizing deficiencies as enhancers or suppressors are described in materials and methods. Final categorization of deficiencies is shown in Figure 2, in which results organized by chromosome are shown graphically. Black columns indicate control crosses with WT or WntD overexpressor mothers. Red columns indicate suppression, orange columns represent weak suppression, blue columns and regions with no columns (no survival to pupal stage) represent no effect, and regions with no columns and a green dot represent enhancement. The full data set in Microsoft Excel format is available upon request.
As the lethality phenotype is a readout of signaling by the Dorsal transcription factor, suppressors and enhancers revealed by the screen might be due to deletion of genes in the Toll pathway as well as the WntD pathway. If deficiencies removing positive regulators of the Toll pathway were categorized as enhancers, this enhancing effect may be due to modification of Toll signaling and not specific to WntD signaling. However, we found that deficiencies removing the positive regulators of Toll signalling Toll, Tube, Pelle and Myd88 do not enhance WntD-mediated lethality, suggesting that the modifying deficiencies revealed in the screen are likely specific to the WntD pathway ( Figure 2F).
As mentioned above, previous experiments suggested no role for the b-catenin pathway in WntD signaling. To confirm this result using independent methodology, we examined deficiencies from the screen that remove regulators of Wnt-b-catenin signaling. If WntD signals via a b-catenin-dependent mechanism, deficiencies removing positive regulators of Wnt-b-catenin signaling would suppress WntD-mediated lethality and deficiencies removing inhibitors of Wnt-b-catenin signaling would enhance WntDmediated lethality. However, we found that the deficiency Df(3R)L127, which removes the inhibitor of b-catenin signaling Daxin (as well as dozens of other genes), was categorized as a suppressor ( Figure 2F). Furthermore, the dsh 3 null allele failed to suppress WntD overexpression (data not shown), suggesting that lethal embryonic dorsalization by WntD signaling is not mediated by Dsh or Daxin in their Wnt-b-catenin roles.
To identify genes removed by suppressor and enhancer deficiencies that are strong candidates to encode WntD pathway components, an algorithm was written that assigned a score to each gene based on several criteria. Genes with Gene Ontology (GO) terms consistent with a role in WntD signal transduction or control of Dorsal activity such ''signal,'' ''wnt,'' ''dorsal,'' ''toll,'' ''immune,'' ''kinase,'' and ''phosphatase'' received one point per GO term (up to three points total). Additionally, genes expressed at 2-fold different levels in wntD KO1 mutant adults infected or uninfected, versus WT adults infected or uninfected received one point (''WntD/yw'' and ''Listeria WntD/yw'' columns) [5]. Lastly, for a gene to suppress or enhance WntD signaling in this assay it must be provided maternally to the embryo. Genes expressed maternally received two points; genes not expressed maternally were penalized two points. Strong candidates identified using this method are shown in Table S1. In this paper, we further characterize two deficiencies that strongly suppress WntD overexpression: Df(1)Sxl-bt and Df(3R)e1025-14 ( Figure 2F).

WntD signals through Fz4
The deficiency Df(1)Sxl-bt is predicted to remove Fz4 (a likely WntD signaling component) and at least 29 other genes (Flybase). In order to confirm that Fz4, rather than another gene removed by the deficiency, suppresses WntD overexpression, we generated Df(1)Sxl-bt/+; Act-Gal4/+; UASp-WntD/UASp-Fz4 females and mated them to WT males, comparing survival of progeny to a WT cross, as well as WntD overexpressors with and without one copy of the Df(1)Sxl-bt deficiency chromosome as before. We found that Fz4 expression restores WntD-mediated lethality, suggesting that loss of Fz4 specifically suppresses WntD overexpression. Overexpression of Fz4 alone shows no lethality ( Figure 3A).
To further examine the role for Fz4 in WntD signal transduction, we wished to generate Fz4 null alleles. We therefore performed transposase-mediated excision of P{GSV2}GS7412, a P-element inserted 217 base pairs upstream of the Fz4 start codon ( Figure S3B). We recovered 55 excisions and screened for large deletions by Southern blot (a subset are shown in Figure S3C). The locations of the EcoRI sites at the Fz4 locus used for Southern blot screening are shown in Figure S6.
Two homozygous viable and fertile excision lines, fz4 3-1 and fz4 25-1 , were analyzed further. Sequence analysis revealed that both alleles are large deletions ( Figure S3B). Quantitative RTPCR showed that both alleles are null for mRNA ( Figure 3C).
In order to further confirm the role for Fz4 in WntD signal transduction, we tested the novel fz4 3-1 and fz4 25-1 null alleles for the ability to dominantly suppress WntD overexpression. We compared survival of progeny from WT and WntD overexpressor females with and without one copy of each of the Fz4 mutant alleles Df(1)Sxl-bt, fz4 P{GSV2} , fz4 3-1 , and fz4 25-1 . We found that heterozygosity for each of these Fz4 mutant alleles tested suppress WntD overexpression ( Figure 3B).
Based on the genetic interactions between Fz4 and WntD, we tested for direct binding of purified, soluble proteins by immunoprecipitation of protein G beads. We found that protein G beads highly depleted the purified Fz4::IgG fusion protein and IgG control protein from the supernatant ( Figure 3D, lanes 2 and 4 of lower blot), and the Fz4::IgG bead precipitation was highly enriched for purified WntD over the IgG bead precipitation ( Figure 3D, compare lanes 5 and 6 of upper blot), showing that WntD and the cysteine rich domain (CRD) of Fz4 specifically bind, in agreement with previous data showing that a membranebound form of WntD binds to a soluble Alkaline-Phosphatase-Fz4 fusion protein [19].

WntD signals through Dcerk and Dmulk to control PGC migration
A second intriguing suppressor from the screen is Df(3R)e1025-14. This deficiency is predicted to remove CG16708, encoding the Drosophila homolog of Ceramide Kinase (hereafter referred to as Dcerk) [20,21,22], as well as 24 other genes (Flybase). Human Ceramide kinase (Cerk) catalyzes the phosphorylation of the sphingolipid ceramide to form ceramide-1-phosphate [22]. Ceramide and ceramide phosphate signaling have been implicated in a variety of biological processes, including apoptosis, inflammation, and cell proliferation [23,24]. Because our algorithm for ranking genes identified in the screen assigned Dcerk a score of 3 (Table S1), and given the abundance of evidence that phospholipid metabolism controls PGC migration, we thought it likely that mutations in Dcerk could disrupt proper PGC localization to the gonad [11,12,14].
To first confirm that haploinsufficiency of Dcerk itself confers the suppressive activity to Df(3R)e1025-14 ( Figure 2F), we obtained the homozygous viable and fertile P-element P{GT1}BG01100 (hereafter referred to as dcerk P{GT1} ), inserted in the predicted first intron of Dcerk. We mated Act-Gal4/+; UASp-WntD/+ females, Act-Gal4/+; UASp-WntD/dcerk P{GT1} females and WT females to WT males and compared survival of their progeny ( Figure 4). We Figure 2. A genetic screen reveals 75 modifier deficiencies that are likely specific to the WntD pathway. A-D. Graphical representation of pupal counts from primary screen. Colored bars represent counts of progeny from WntD overexpressor females also carrying deficiency chromosomes. Black columns are WT and WntD overexpressor controls. Red columns represent deficiencies characterized as suppressors; orange columns represent deficiencies characterized as weak suppressors; blue columns represent deficiencies having no effect; green dots represent deficiencies characterized as enhancers. Regions with no columns and no green dot are deficiency crosses resulting in no pupae that are characterized as having no effect. A secondary screen described in Materials and Methods contributed to these final characterizations. A. Chromosome 1 results. B. Chromosome 2 results. C. Chromosome 3 results. D. Chromosome 4 results. E. Summary of deficiencies tested and characterizations. F. Selected deficiencies from the screen that remove interesting genes (Fz4, Dcerk); components of the Toll signaling pathway (Toll, Tube, Pelle, Myd88); and Daxin, a component of the Wnt/Armadillo signaling pathway. It is important to note that these deficiencies remove the aforementioned genes as well as dozens of other genes. doi:10.1371/journal.pone.0026993.g002 found that the dcerk P{GT1} allele strongly suppresses WntD overexpression, similar to the original deficiency.
To confirm that the P-element insertion at the Dcerk locus and not a linked mutation causes suppression of WntD overexpression, we performed transposase-mediated P-element excision. By sequencing the region surrounding the P{GT1}BG01100 insertion site we discovered that one excision allele, dcerk X35 , has 38 bases of the P-element remaining at the locus (data not shown). We tested whether dcerk X35 can suppress WntD overexpression by generating Act-Gal4/+; UASp-WntD/dcerk X35 females and comparing survival of their progeny to WT females, Act-Gal4/+; UASp-WntD/+ females, and Act-Gal4/+; UASp-WntD/dcerk P{GT1} females that had been mated to WT males. We found that, while the original dcerk P{GT1} insertion suppresses WntD overexpression, the near-precise excision dcerk X35 failed to suppress, suggesting that dcerk X35 retains wild-type activity, and Dcerk specifically is required for WntD signaling-mediated embryonic lethality ( Figure 4).
When we tested homozygous dcerk P{GT1} mutant embryos for a PGC phenotype as seen in wntD KO1 mutants (Figure 1), we failed to see an increase in mislocalized PGC compared to WT ( Figure 5). Hypothesizing redundancy between several lipid kinases ( Figure 6A), we tested mutants in the lipid kinase genes Diacylglycerol kinase (dgk f03609 ), Sphingosine kinase 2 (sk2 KG05894 ), CG31873 3414 [25], as well as four deficiencies removing Sphingosine kinase 1 (Sk1) -Df(1)BSC287, Df(1)BSC288, Df(1)BSC722, and Df(1)ED7067 -for the ability to dominantly suppress WntD overexpression, as before. We found that heterozygosity for either the dcerk P{GT1} allele or the CG31873 3414 allele strongly suppressed WntD-mediated lethality, in contrast to every other lipid kinase allele tested ( Figures 6B and S4). CG31873 encodes a homolog of mouse Multi-substrate Lipid kinase (MuLK, also known as acylglycerol kinase, Agk) reported to phosphorylate ceramide in vitro [26]. The CG31873 3414 allele is an insertion of a piggyBac transposon containing splice acceptors, and stop codons in all three reading frames, which generates a nonsense allele and is described in [25]. We hereafter refer to CG31873 as Dmulk.
When we labeled PGC in maternal and zygotic double mutant dmulk 3414 ; dcerk P{GT1} embryos using the Vasa antibody, we found a significant increase in the number of mislocalized PGC in stage 14-16 mutant embryos compared to WT (Figure 7, compare A vs. E). Furthermore, neither single mutant alone displayed a significant increase in mislocalized PGC ( Figures 5C and 7B), nor did dmulk 3414 /CyO; dcerk P{GT1} embryos carrying a single wild type, paternally inherited allele of Dmulk ( Figure 7C). dcerk X43 flies express wild-type levels of Dcerk transcript (data not shown); we therefore consider dmulk 3414 ; dcerk X43 flies to be equivalent to mutants carrying the dmulk 3414 allele alone ( Figure 7B). Importantly, we find that wntD KO1 mutant and dmulk 3414 ; dcerk P{GT1} double mutant embryos display decreased numbers of PGC localized at the gonad when compared to WT, while total numbers of PGC within the embryo are indistinguishable from the WT, suggesting that the mislocalized cells are truly failing to migrate properly, rather than simply being ''extra'' cells ( Figure 7H).

Dcerk and Dmulk regulate Ceramide-1-Phosphate levels in vivo
Previous studies have shown that Dcerk is required for wild-type levels of ceramide-1-phosphate (C1P), suggesting that the gene encodes a bona fide ceramide kinase [21]. Our data showing that Dcerk and Dmulk redundantly control PGC migration implies that C1P, the product of ceramide phosphorylation, plays a role in guiding PGC to the gonad. This would require that Dmulk also encodes a protein with ceramide kinase activity. To test this possibility, we collected overnight egg lays from WT, UASp-Dcerk/ CyO; Act-Gal4/TM6B (maternal Dcerk overexpression), and UASp-Dmulk/CyO; Act-Gal4/TM6B (maternal Dmulk overexpression) incrosses and measured C1P levels, normalized to phosphatidylcholine levels. Upon Dcerk and Dmulk overexpression, we detected a 4.2-fold and 3.3-fold increase in embryonic C1P levels over WT levels respectively, showing that both Dcerk and Dmulk are sufficient to increase C1P production ( Figure 8). Full results showing the effects of Dcerk and Dmulk overexpression on C1P molecular species are shown in Figure S5. Given that the mouse homolog of Dmulk is a bona fide ceramide kinase [26], our data strongly suggest that Dmulk likely also functions as a ceramide kinase.  There is an increase in the proportion of embryos with more than 2 mislocalized cells in wntD KO1 compared to WT. dcerk P{GT1} more closely resembles WT than wntD KO1 . doi:10.1371/journal.pone.0026993.g005

Discussion
Here, we report on a suppressor/enhancer screen designed to identify members of a novel signaling pathway transducing the WntD ligand. Using the methods described above, we narrowed our initial list of 75 modifiers down to 43 genomic regions likely to encode members of the WntD pathway. We further investigated the suppressor deficiencies Df(1)Sxl-bt and Df(3R)e1025-14.  Df(1)Sxl-bt is predicted to remove Fz4 and 29 other genes. Fz4 has long been a strong candidate for the WntD receptor, given previous evidence that the two proteins will bind [19]; however, until now evidence of a functional interaction has been lacking. We rescued the suppression by Df(1)Sxl-bt by co-overexpressing Fz4 and confirmed that mutant alleles of Fz4 can suppress WntD overexpression, strengthening the idea that WntD signals through Fz4 in vivo. No other biological role has previously been ascribed to Fz4, suggesting that this may be the first Wnt-Fz interaction dedicated to activating a b-catenin-independent pathway. Indeed, the Fz4 amino acid sequence is quite different from other members of the Fz family (data not shown), suggesting an interesting signaling paradigm involving an interaction between the most divergent Wnt family member and the most divergent Fz family member.
An outstanding question is whether the receptor Fz4 contributes to PGC guidance. When we examined PGC migration in fz4 3-1 mutant embryos, we did not observe an increase in mislocalized cells when compared to WT (data not shown). One possible explanation for this is that, like the ceramide kinases Dcerk and Dmulk, Fz4 acts redundantly with another receptor to mediate PGC guidance. Alternatively, we cannot exclude the possibility that WntD activates different receptors to regulate D/V patterning and PGC migration. Further experimentation is required to more fully understand the biological roles of WntD/receptor interactions.
The second suppressor deficiency examined, Df(3R)e1025-14, is predicted to remove approximately 25 genes. Here, our algorithm allowed us to focus on a gene removed by the deficiency called Dcerk. We found that the P-element insertion mutant dcerk P{GT1} strongly suppressed WntD overexpression and that near-precise excision of the P-element abrogated the suppression, confirming the role of Dcerk in WntD signaling.
Given the role of Dcerk in lipid metabolism [21,22], and the role of lipid metabolism in Drosophila PGC migration [11], we were interested to further explore the role of WntD-Dcerk signaling in this process. While wntD KO1 mutants display a defect in migration of PGC to the gonad, dcerk P{GT1} mutants do not. However, we identified a Drosophila gene homologous to mouse MuLK, CG31873/Dmulk, which we hypothesized would compensate for loss of Dcerk because mouse Mulk can phosphorylate ceramide in vitro [26]. Indeed, embryos carrying strong loss of function alleles of both Dmulk and Dcerk display defective PGC migration. In agreement with this functional redundancy, we showed that expression of either Dcerk or Dmulk is sufficient to increase embryonic C1P levels, strongly suggesting that both Dcerk and Dmulk are ceramide kinases. Further experiments are required to directly test whether WntD signaling through these proteins regulates C1P metabolism.
These findings may shed light on previous observations regarding the influence of phospholipids on germ cell migration. Studies on Wun and Wun2, genes encoding lipid phosphate phosphatases, have generated a model in which PGC travel up a concentration gradient of a phospholipid. It is currently unclear whether PGC are attracted to the phospholipid or repelled by the non-phosphorylated version; the identity of the lipid/phospholipid is also not known [27]. Wun and Wun2 are broad-specificity phosphatases, and no lipid kinase alleles have previously been reported with PGC guidance defects in Drosophila. Our data are the first to indicate a role for a specific phospholipid in PGC migration and we speculate that C1P may be a relevant substrate for Wun activity in this process.
Both Dmulk and Dcerk are expressed in the midgut and hindgut starting around stage 11 of embryogenesis (BDGP in situ homepage http://insitu.fruitfly.org/cgi-bin/ex/insitu.pl). One may expect that a PGC attractant would be synthesized in the PGC target tissue. However, in this model, C1P must be synthesized and secreted from the midgut and hindgut and must diffuse readily throughout the embryo. Wun and Wun2 phosphatase activity in the gut, central nervous system, and epidermis generates a gradient of C1P such that C1P in these tissues is dephosphorylated to form ceramide, leaving the highest concentration of C1P in the lateral mesoderm, where no Wun or Wun2 is expressed. While most studies on C1P have described intracellular signaling roles for the molecule, there is evidence that C1P is readily secreted from bone marrow-derived macrophages (BMDM), providing further evidence that C1P could be used as a secreted guidance molecule in Drosophila [28]. We hypothesize a model in which WntD expression in the embryonic midgut stimulates the kinase activity of Dmulk and Dcerk, thereby providing C1P as a PGC chemoattractant.
One possible weakness in this model is that, if all ceramide kinase activity were removed from the embryo, it might be expected that no C1P would be generated. If C1P were the phospholipid guiding PGC to the gonad, a very strong phenotype would be observed, as is the case in strong wun, wun2 mutants [29]. However, the phenotype in dmulk 3414 ; dcerk P{GT1} mutants is weak in comparison (Figure 7). This may be explained by the fact that the Dcerk allele used in this study is not a complete null; adult dcerk P{GT1} flies express approximately 20% of wild-type levels of Dcerk mRNA (data not shown). Furthermore, in dmulk 3414 ; dcerk P{GT1} double mutant adults, C1P levels are reduced only 2fold compared to WT (data not shown). If this is also true in embryos, it likely explains why this comparatively weak phenotype is observed. It is possible that true dmulk; dcerk double null embryos would have completely disrupted PGC migration.
It is also possible that C1P can be generated via a mechanism besides ceramide phosphorylation, and true dmulk; dcerk double null embryos would still have significant levels of C1P. Indeed, when Cerk null mice were generated, it was found that, while cerk 2/2 BMDM have no detectable ceramide kinase activity, cellular levels of C1P were approximately 25% those of wild-type cells [28]. These data suggest there is indeed a ceramide kinaseindependent source of C1P in mice; it is likely that this is also the case in other organisms. There are currently no data suggesting a mechanism for this in mouse or Drosophila; however, some arachnid, bacteria and fungus species express a Sphingomyelinase D (SmaseD) protein capable of metabolizing sphingomyelin directly into C1P [30]. While a BLAST search of the Drosophila genome reveals no clear SmaseD homologs, we believe it possible that this type of Smase activity may be encoded by genes not apparently related evolutionarily to SmaseD. In this case, even in the complete absence of ceramide phosphorylating activity, a significant amount of C1P may be available for the generation of a gradient by Wun and Wun2 activity. This hypothetical ceramidekinase-independent source of C1P would also explain the comparatively weak phenotype observed in wntD KO1 embryos.
It is interesting to note that studies have shown that Wun2 can dephosphorylate phosphatidic acid (PA) and lysophosphatidic acid (LPA), but not C1P [14,15]. However, mammalian homologs of Wun can dephosphorylate C1P, as well as S1P, PA and LPA, raising the possibility that Wun and Wun2 can as well [15,31]. If this is true, it is unclear why studies have failed to show Wun or Wun2 activity against C1P. However, given the fact that the majority of Drosophila sphingolipids contain a 14-carbon backbone, while mammalian and commercially-synthesized sphingolipids contain an 18-carbon backbone, we believe it to be likely that Drosophila proteins have decreased affinity for commonly available sphingolipid reagents [32]. Furthermore, we cannot exclude the possibility that Wun activity generates a gradient of multiple different phospholipids used in PGC guidance. It will be interesting to see in the future if further experimentation can address this question.
While vertebrates do not express a WntD ortholog, studies of Drosophila germ cell migration may lend insight into mechanisms of vertebrate cell migration. Granado et al. suggest that C1P induces macrophage migration, although the effect appears to be much weaker than that induced by S1P [33]. Further experimentation in both mammalian and Drosophila systems may reveal similarities in how a gradient of C1P regulates migration of a variety of cell types.
We have presented here the results of an unbiased genetic screen aimed at identifying genes in a novel signaling pathway that transduces the WntD signal. We present the first evidence of a functional interaction between WntD and its receptor Fz4, and identify a novel downstream effector, Dcerk. We further hypothesized that the same transduction mechanism is used to influence dorsal/ventral patterning, the adult innate immune response, and PGC migration. Indeed, we present evidence that two ceramide kinases redundantly mediate the effect of WntD on both dorsal/ventral patterning and PGC migration. Further experimentation will determine whether this pathway also impacts the immune response. Interestingly, our unbiased screen led us to a line of experimentation suggesting a role for ceramide-1phosphate in PGC migration, the first evidence for a specific phospholipid in this process. Further studies should directly test whether C1P represents the long-unknown substrate for Wun and Wun2 in PGC migration.

Screen crosses and general husbandry
Flies were maintained under standard conditions unless otherwise noted. WT flies are yw.
For the suppressor/enhancer screen, Act-Gal4/CyO; UASp-WntD/Tm6B males were mated to virgin females heterozygous for a single deficiency from the Bloomington deficiency kit (or WT virgins as control). After 5 days at 25uC the crosses were shifted to 28uC. Upon eclosion, 5-6 Act-Gal4; UASp-WntD virgin females carrying one deficiency chromosome were mated to WT males and progeny surviving to pupae were quantified after 10-11 days.
For the first pass of analysis, when crosses gave rise to more than zero but fewer than 25 progeny that survive to pupal stage the deficiencies were categorized as having no effect. When crosses gave rise to more than 25 progeny that reached pupal stage, the deficiencies were initially categorized as suppressors. Exceptions were made for several crosses resulting in fewer than 25 pupae when many larvae were observed; these deficiencies were also categorized as suppressors. When crosses resulted in no churning of the media, indicating zero survival to larval stage, the deficiencies were categorized as enhancers. Full results are available in Microsoft Excel format.
Putative suppressors were subjected to a secondary, more quantitative screen in which the above crosses were repeated and embryonic hatching to first-instar larvae was quantified. Crosses giving rise to embryos with a hatch rate from 1-5% were categorized as having no effect. Crosses with hatch rates from 6-19% were categorized as weak suppressors. Crosses with hatch rates greater than or equal to 20% were categorized as strong suppressors.
To rule out the possibility that significant lethality indicating enhancement of WntD overexpression could be due to reduced viability associated with the deficiency itself, rather than enhancement of WntD signaling, females carrying the deficiency plus either the Act-Gal4 or UASp-WntD transgenes, but not both (and therefore not overexpressing WntD), were mated to WT males and survival of progeny was assayed. Deficiencies conferring significant lethality were categorized as ''non-specific enhancers'' and were excluded from further analysis.

WntD suppression by hatch assay
To assay suppression of WntD overexpression by embryonic hatch assay, crosses were performed as described above, but matings between Act-Gal4; UASp-WntD virgins carrying a deficiency and WT males were placed in crossing cages. After 24 hours of embryo collection, collection caps were removed and embryos were aged for 24 hours. Embryos hatching to first-instar larvae were then quantified.

Generation of Dcerk excision alleles
Flies carrying the dcerk P{GT1} allele and a D2-3 transposase transgene were mated to TM3 balancer flies and the progeny were screened for loss of the w + marker.

Southern Blot
For genomic DNA isolation, 15 adult females of each genotype (WT, P{GSV2}fz4 GS7412 , P{GSV2}fz4 GS7412 /Binsinscy, and homozygous or Binsinscy-balanced P{GSV2}fz4 GS7412 excisions) were crushed with a Kontes pestle in 75 mL lysis buffer (100 mM Tris pH 8.8, 100 mM EDTA, 1% SDS). An additional 225 mL buffer was added and samples were further homogenized. Samples were then incubated at 72uC for 30 min. and 66 mL KOAc (5 M) was added. Samples were incubated for 30 min. on ice, followed by centrifugation at 20,8006g for 15 min. at 4uC. The supernatant was transferred to another tube and centrifuged again. The supernatant was again transferred and one volume isopropanol was added. The samples were centrifuged again for 5 min at room temperature to precipitate genomic DNA. Pellets were washed with 70% ethanol and centrifuged once more. The pellets were briefly allowed to air-dry and resuspended in 105 mL TE, and incubated for several minutes at 37uC to dissolve.
The entire genomic DNA preparation was digested overnight at 37uC with EcoRI and subjected to electrophoresis through duplicate 0.7% agarose gels. DNA was transferred to Hybond XL membranes (GE Healthcare) overnight by alkaline transfer in denaturation buffer (0.5 M NaOH, 1.5 M NaCl). Following transfer, membranes were incubated for 30 min. in neutralization buffer (0.5 M Tris pH 6.0, 1.5 M NaCl), washed twice in 2xSSC, and pre-hybridized in Rapid-Hyb buffer (GE Healthcare) for 30 min at 65uC.
Locations of EcoRI sites with respect to the P{GSV2} insertion and Fz4 start codon are indicated in Figure 6.
To purify mRNA, 6 adult females of each genotype were homogenized in 150 mL Trizol (Invitrogen) using a Kontes pestle. An additional 150 mL Trizol was added and flies were homogenized some more. Thereafter, the standard Trizol protocol was followed.
cDNA was synthesized using a Thermoscript kit (Invitrogen) using included oligo-dT primer and standard protocols. Quantitative PCR was performed using Sybr Green PCR master mix (Applied Biosystems) on a Step One Plus thermocycler (Applied Biosystems). Fz4 mRNA levels were normalized against RpS15A mRNA levels.

Ceramide-1-phosphate quantifications
Prior to lipid extractions, Drosophila embryos were collected on agar caps using standard methods, dechorionated in 50% bleach, and washed several times in methanol. Embryos were stored at 220uC in methanol until lipid extraction.

Co-immunoprecipitation
WntD protein was purified as previously described; however, the blue Sepharose fractions contained only WntD and bovine serum albumin and were considered pure [37]. Further purification steps were omitted.
IgG and Fz4::IgG proteins were purified as previously described [38]. The Fz4::IgG construct is an Ig-tagged version of the cysteine rich domain (CRD) of Fz4, which is the high-affinity Wnt-binding cassette of Fz family members.
Protein G-coupled sepharose beads were incubated with equal masses of purified IgG or Fz4::IgG protein overnight at 4uC. Beads were washed with cold PBS and incubated with equal amounts of purified WntD protein overnight at 4uC. Beads were again washed with cold PBS. After washing, sample buffer was added to beads and incubated 5 min at room temperature. All samples were then subjected to SDS-PAGE, followed by western blot by standard protocols. Antibody dilutions: Rabbit anti-WntD 1:1000, AP-conjugated goat anti-Rabbit 1:20000 (Santa Cruz Biotechnology, Santa Cruz, CA), AP-conjugated goat anti-human 1:20000 (Bio-Rad, Hercules, CA). Figure S1 Tissue-specific WntD antibody staining. WT embryos stained with WntD antibody. A. Stage 8 embryo. Expression is observed in invaginating midgut. B. Stage 14 embryo. Expression is observed in midgut (arrow) and gonads (arrowheads). (PDF) Figure S2 Genetic crossing scheme for WntD suppressor/enhancer screen. A-C. Deficiency stock virgin females were mated to Act-Gal4/CyO; UASp-WntD/TM6B males and virgin female progeny with no balancers were selected. These females, overexpressing WntD in the germline and carrying a deficiency, were mated to WT males and survival of progeny was assayed. Band is shifted above expected size of 1.7 kb for intact P-element, and below expected size of 7.8 kb for WT locus or a precise excision, indicating a partial deletion of Fz4. Lane 42: fz4 25-1 . Band (visible above background in right-hand side of lane) appears to be shifted slightly below expected size for WT locus, indicating partial deletion of Fz4. All remaining lanes are uncharacterized excision alleles and are homozygous, except for lanes 13 and 51, which are balanced over Binsinscy.  Figure S5 Dcerk and Dmulk display differential specificity for ceramide molecular species. A. Maternal Dcerk and Dmulk overexpression results in increased embryonic C1P levels compared to WT. Error bars represent SEM. Asterisks represent P,0.05 by Student's T-test compared to WT. Dcerk overexpressor embryos are from a UASp-Dcerk/CyO; Act-Gal4/ TM6B incross. Dmulk overexpressor embryos are from a UASp-Dmulk/CyO; Act-Gal4/TM6B incross. B. Measurements of embryonic C1P level normalized to phosphatidylcholine levels, +/2 standard deviation (SD). (PDF) Figure S6 Locations of EcoRI sites at the Fz4 locus. EcoRI sites within the P{GSV2} P-element and flanking the Southern blot probe (indicated by blue bar) at the Fz4 locus imply predicted bands of 7.8 kb for a wild-type locus or precise Pelement excision; 1.7 kb for a locus with intact P-element, or in which part of the P-element has been excised while the internal EcoRI site remains inserted; 7.8-12.5 kb for a partial P-element excision in which the internal EcoRI has been excised; or ,7.8 kb for a deletion of the entire P-element plus surrounding Fz4 genomic DNA. (PDF)

Supporting Information
Table S1 Strong WntD pathway candidates from screen. Highscoring candidates were narrowed by effects of overlapping deficiencies, by virtue of association with GO terms consistent with signaling pathways, significant change in expression by microarray [5], and by maternal expression. ''Total genes'' indicates the number of genes in the region encompassed by the deficiency, while ''Remaining candidate genes'' indicates the number of genes after taking into account the phenotypes of other overlapping deficiencies from the kit as well as smaller overlapping deficiencies assayed after identification of the candidate regions. The ''WntD/yw'' column indicates the fold difference in the gene's expression level in uninfected wntD KO1 mutant adults compared to uninfected yw adults, and ''Listeria WntD/yw'' indicates fold difference in the gene's expression level in wntD KO1 mutant adults compared to yw adults after injection with Listeria monocytogenes [5]. Genes further characterized in this article are indicated in bold. (PDF)