Role of the ABC transporter Mdr49 in Hedgehog signaling and germ cell migration

Proper morphogenesis of complex tissues and organs depends upon the carefully orchestrated migratory behavior of different cell types (Rørth, 2009; Friedl and Gilmour, 2009; Pocha and Montell, 2014). For example, directed cell migration plays a crucial role in the coalescence of the embryonic gonad in Drosophila melanogaster (Kunwar et al., 2006). The embryonic gonad consists of two different cell types: primordial germ cells (PGCs) and somatic gonadal precursor cells (SGPs) (Boyle and Dinardo, 1995; Boyle et al., 1997). SGPs and PGCs are formed at different locations and times in the fly embryo and their identity depends upon completely different mechanisms. SGPs are mesodermal in origin and are specified by zygotically active segmentation genes during mid-embryogenesis in parasegments 10-13. By contrast, PGCs are formed at the posterior pole on the outside surface of the embryo at the blastoderm stage and their specification depends upon maternal determinants deposited at the pole during oogenesis (Ephrussi and Lehmann, 1992; Smith et al., 1992).

At the onset of gastrulation, PGCs commence their journey towards the SGPs. This is a multistep process (reviewed by Kunwar et al., 2006; Pocha and Montell, 2014). PGCs are first carried into the interior of the embryo by the midgut invagination. The PGCs then undertake an invasive trans-epithelial migratory step across the mid-gut primordium, and subsequently move along the dorsal surface of the midgut until they split into two groups. Germ cells in each group migrate laterally until they reach the gonadal mesoderm on either side of the embryo and subsequently align with the SGPs in parasegments 10-13. The juxtaposed germline and somatic cells eventually coalesce into the primitive embryonic gonad. A combination of repulsive and attractive cues guides PGC migration through the midgut and toward the SGPs. Once the PGCs exit the midgut, repulsive clues, controlled by wunen (wun) and wunen-2 (wun2), are thought to direct their bilateral movement away from midgut surface (reviewed by Santos and Lehmann, 2004a). Subsequently, attractive cues produced by the SGPs direct the PGCs towards the lateral mesoderm and then help to promote and maintain their association with the SGPs (Jaglarz and Howard, 1994; Van Doren et al., 1998).

Several lines of evidence indicate that the signaling molecule Hedgehog (Hh) (for a detailed review of Hh signaling, see Ingham and McMahon, 2001; Jiang and Hui, 2008; Gallet, 2011; Briscoe and Thérond, 2013) is one of the key attractants generated by the SGPs to orchestrate PGC migration (Deshpande et al., 2001, 2007, 2009; Deshpande and Schedl, 2005). Ectopic expression of Hh in other tissues like the nervous system induces a subset of PGCs to migrate away from the SGPs towards the source of the misexpressed Hh. Supporting the idea that Hh functions directly as a guidance cue for migrating PGCs is the finding that two cell-autonomous Hh ‘receptors’, Patched (Ptc) and Smoothened (Smo), are required in PGCs for proper migration. In the absence of the Hh ligand, the transmembrane receptor Ptc inhibits the seven-pass transmembrane protein Smo from mediating signal transduction. Hh binding to Ptc is thought to relieve the negative influence of Ptc, resulting in the relocalization of Smo to cell membranes, and activation of the signal transduction cascade downstream of the Hh signal. Consistent with their reciprocal functions in hh signaling, PGCs compromised for maternally derived ptc or smo activity behave differently. For ptc, PGCs clump prematurely on the external surface of the midgut presumably because of ligand-independent signal transduction. For smo, PGCs behave as if they are ‘signal-blind’ and scatter in the posterior of the embryo.

Since hh is expressed in a wide range of embryonic tissues and functions in patterning throughout much of development, a crucial question is what mechanism(s) distinguishes Hh produced by the SGPs from Hh produced by other cell types in the mesoderm and ectoderm. One important solution to this puzzle is the pivotal role played by the isoprenoid biosynthesis pathway in Hh signaling. The isoprenoid geranylgeranyl-pyrophosphate (GGPP) is used for geranylation of the Gγ1 subunit of the heterotrimeric G-protein complex GαGβGγ1 (Yi et al., 2006). In order for this complex to function it must be tethered to the membrane by the geranylation of the Gγ1 subunit. In Gγ1 mutants Hh is not properly released from the sending cell and instead accumulates in large puncta distributed along the basolateral membrane (Deshpande et al., 2009). Because activity of the GαGβGγ1 complex depends upon geranylation of Gγ1, the same phenotype is observed in embryos mutant for enzymes in the isoprenoid biosynthetic pathway.

The rate-limiting step in this biosynthetic pathway (and as a consequence the extent of Gγ1 geranylation) is the conversion of 3-hydroxy-3-methylglutaryl-CoA to mevalonate by HMG-CoA reductase (Hmgcr). The activity of Hmgcr is regulated at many different levels, including transcription, translation, post-translational modification, and degradation (Ness, 2014; Sharpe et al., 2014). In mammals, these different regulatory inputs on Hmgcr activity are essential for setting the levels of cholesterol and isoprenoids (Edwards and Ericsson, 1999). Although flies lack enzymes needed for de novo cholesterol biosynthesis, regulating Hmgcr activity is important in flies for controlling isoprenoid production and, as a consequence, in modulating the long-distance basolateral signaling activity of the Hh ligand. That the Hmgcr activity normally functions to potentiate the basolateral transmission of Hh, is most clearly illustrated in experiments in which the Hh-GFP fusion protein was ectopically expressed in the embryonic CNS during mid-embryogenesis with or without co-expression of Hmgcr. When co-expressed with Hmgcr, Hh-GFP spreads into and through the mesoderm, making direct contact with the migrating PGCs. By contrast, in the absence of Hmgcr, the transmission of Hh-GFP is largely restricted to the CNS and immediately neighboring cells in the ectoderm (Deshpande et al., 2013). One reason that ectopic Hmgcr activity in the CNS is limiting at this stage of development is that transcription of the Hmgcr gene is subject to stage- and tissue-specific regulation. During early embryogenesis, Hmgcr is broadly expressed throughout the embryonic mesoderm and there is also a substantial maternal contribution. However, by mid-embryogenesis Hmgcr transcripts are restricted to the SGPs. Consistent with this developmental restriction in Hmgcr expression, there are severe PGC migration defects in Hmgcr mutant embryos (Van Doren et al., 1998; Santos and Lehmann, 2004b).

Although our experiments argue that the Hmgcr to isoprenoid biosynthetic pathway is responsible for potentiating the transmission of Hh expressed in SGPs, other mechanisms must also contribute to enhancing the production of this ligand and its long-distance signaling activity. In this regard, cholesterol and cholesterol-derived products deserve special attention. First, cholesterol is required for auto processing of the Hh ligand (Porter et al., 1996). Second, the resulting covalent cholesterol modification then plays a crucial role in the intracellular sorting of the processed Hh peptide, and its subsequent transmission and release from basolateral membranes (Gallet et al., 2003; Callejo et al., 2011). Since flies are cholesterol auxotrophs, they must obtain this sterol through their diet (Clayton, 1964; Santos and Lehmann, 2004a,b). For this reason hh signaling would be expected to be sensitive to levels of dietary cholesterol. It should also depend upon the factors that are responsible for uptake of cholesterol stored in the yolk by cells in the intestine, for movement of cholesterol from the intestine to tissues throughout the animal, and finally, for translocation of cholesterol from the surface of the hh signaling cell to the subcellular compartment where the Hh precursor protein is processed.

A tantalizing clue in this regard came in the form of a discovery of a new component of germ cell migration – the ABC transporter Mdr49. Ricardo and Lehmann (2009) screened embryos mutant in a small set of ABC transporters for PGC migration defects. Of the ABC genes tested, Mdr49 was the only transporter that had a significant effect on PGC migration. They found that mutations in Mdr49 resulted in PGC migration defects, whereas restoration of Mdr49 activity by ectopic expression in the mesoderm (where it is normally expressed) was sufficient to rescue the migration defects. ABC transporters are conserved from bacteria to humans and they transport a wide range of compounds, including hydrophobic lipophilic molecules (reviewed by Higgins, 1992; Chen and Tiwari, 2011). Mdr49 shares sequence similarity with Ste6p, another member of the ABC transporter family in budding yeast that is required for the export of the farnesylated a-mating type pheromone. Based on this connection and the known involvement of the Hmgcr-isoprenoid biosynthetic pathway in PGC migration, Ricardo and Lehmann proposed that migrating PGCs might be attracted by an equivalent isoprenoid-modified peptide that is exported from SGPs by Mdr49.

However, arguing against this simple model, Mdr49 is not the most closely related ABC transporter to Ste6p in flies (30% identity). The closest fly relative is CG1824, which has 40% identity to Ste6p. There are also many other fly ABC transporters, including MRP (33%), CG4562 (32%), CG31789 (32%) and White (32%) that share more homology with Ste6p than Mdr49. The only known PGC attractant, Hh, is probably far too large to be translocated through the SGP membrane via an ABC transporter (Germann and Chambers, 1998). Moreover, the two known lipid modifications of Hh, namely cholesterol and palmitate, are not isoprenoids. For this reason, we decided to explore other less direct functions for Mdr49 in PGC migration.

One of the closest mammalian relatives of Mdr49 is Mdr1/P-cg1 (42% identity). Like other ABC transporters it is relatively promiscuous and can translocate many seemingly unrelated small molecules such as glutamate, cortisol, aldosterone, methotrexate, colchicine and vinblastine (Choi, 2005; Orlowski et al., 2006; Chen and Tiwari, 2011). Mdr1 has also been reported to function in the translocation of cholesterol from the cytosyolic to the exoplasmic bilayer and from the plasma membrane to the endoplasmic reticulum (Garrigues et al., 2002; Metherall and Waugh, 1996). Moreover, Tessner and Stenson (2000) have shown that overexpression of Mdr1 in intestinal cells upregulates the import of cholesterol. Given the importance of cholesterol in the production of a functional Hh ligand, a compelling hypothesis is that during germ cell migration Mdr49 functions at some step in cholesterol translocation. In this model, this translocation function would be essential for ensuring that the levels of cholesterol in SGPs are high enough so that these cells can produce sufficient amounts of processed Hh to effectively signal the migrating PGCs. Here, we have directly tested this model.

We first determined the frequency of PGC migration defects induced by three independent loss-of-function alleles of Mdr49, Mdr49^MB04959, Mdr49^KG08611 and Mdr49^Δ3.16 generated by transposable element insertions. Homozygous mutant embryos were screened for PGC migration defects by counting the number of mismigrated or ‘lost’ germ cells. As can be seen in Fig. S1, there were only minor effects on PGC migration in Mdr49^MB04959, with most embryos having only 0-3 lost germ cells. An intermediate phenotype was observed for Mdr49^KG08611. As reported previously, Mdr49^Δ3.16 displayed the most severe defects of the three mutants, with a substantial number of embryos (>50%) showing 5-7 lost PGCs (Ricardo and Lehmann, 2009). As the frequency of PGC migration defects was even greater in embryos homozygous for a deficiency, Df(2F)Mdr49, which uncovers Mdr49 (not shown), it seems likely that none of these alleles is a null. However, Df(2F)Mdr49 has an ∼120 kb deletion that removes not only Mdr49, but also several other genes. For this reason it is difficult to attribute the strong phenotype solely to loss of Mdr49. Consequently, we used either Mdr49^Δ3.16 or Mdr49^Δ3.16/Df(2R)Mdr49 embryos for most of our experiments.

Since the production of mature Hh proceeds by an internal proteolytic cleavage and the concomitant addition of cholesterol to the C-terminal amino acid of the cleaved peptide, a plausible explanation for the PGC migration defects in Mdr49 mutants is that hh signaling from SGPs is disrupted because there is an insufficient amount of cholesterol to fully process the Hh precursor peptide. Unlike mammals, flies are not able to synthesize cholesterol de novo (Clayton, 1964). Instead they depend upon exogenous sources of this lipid. All of the cholesterol in embryos is of maternal origin and is deposited in the yolk during oogenesis (reviewed by Welte, 2015). Following the cellular blastoderm stage, the yolk is incorporated into the midgut by the process of gastrulation. Thereafter, any requirements for exogenous cholesterol must be met by transporting it from the intestine into the mesoderm and through the mesoderm to the ectoderm. The cholesterol must then be imported into the cell and distributed to the appropriate subcellular compartment(s) for processing of the Hh precursor.

Because of the reported involvement of the mammalian Mdr1/P-cg1 transporter in cholesterol translocation, we wondered whether Mdr49 might be required to ensure that there are sufficient levels of cholesterol in the SGPs for Hh processing. If so, we reasoned that it might be possible to augment the production of a functional Hh ligand in Mdr49 mutants by increasing cholesterol levels in the embryo. For this purpose, we raised Mdr49 flies from the embryonic to the adult stage on a cholesterol-rich diet (see Materials and Methods) and then collected progeny from the cholesterol-fed adults. Fig. 1 shows that raising flies on a cholesterol-rich diet partially suppresses the PGC migration defects evident in Mdr49 embryos. The number of embryos with five or more lost PGCs was reduced by 50% in the progeny of Mdr49 flies raised on a high-cholesterol diet. We also tested the effects of feeding adults on a cholesterol-rich yeast paste before collecting their progeny for analysis. In this instance, we observed a weaker amelioration (20% reduction in number of embryos with five or more lost PGCs; data not shown) of the PGC migration defects compared with the progeny of control adults that were fed yeast lacking the cholesterol supplement. Interestingly, in this experimental design the extent of rescue improved as the duration of feeding increased.

The dietary supplement experiments in the previous section support the idea that Mdr49 functions to ensure that SGPs have sufficient levels of cholesterol for the processing of the Hh ligand. To explore this point further we tested for genetic interactions in embryos trans-heterozygous for the strongest Mdr49 mutation, Mdr49^Δ3.16, and Gγ1¹, a gene involved in Hh signaling and germ cell migration. While heterozygosity for either Mdr49^Δ3.16 or Gγ1¹ alone had little effect on PGC migration (>90% have 0-3 lost PGCs), the frequency of lost PGCs increases in the trans-heterozygous combination. 20% of the trans-heterozygotes had 4-6 lost PGCs, whereas just over 10% had 7 or more lost PGCs (Fig. 2). Since Gγ1 activity depends upon the Hmgcr-isoprenoid biosynthetic pathway, we also tested for genetic interactions between Mdr49^Δ3.16 and Hmgcr. Unlike either Mdr49^Δ3.16 or Gγ1¹, embryos heterozygous for a strong Hmgcr allele had a significant frequency of PGC migration defects (nearly equivalent to that seen in Mdr49^Δ3.16/Gγ1¹ trans-heterozygotes). Perhaps, for this reason there was only a modest increase in the number of lost PGCs in the Mdr49/Hmgcr trans-heterozygotes compared with the Hmgcr/+ control (not shown).

In Drosophila wing discs, Hh is expressed exclusively in the posterior compartment, and functions as a long-distance morphogen to organize patterning of the wing blade (Basler and Struhl, 1994). Hh influences wing development by inducing the expression of downstream targets such as patched (ptc) and decapentaplegic (dpp) in the anterior compartment. In the absence of hh signaling, these target genes are not activated and there are patterning and growth defects along the anterior-posterior (A-P) axis. Conversely, when hh is inappropriately expressed in the anterior compartment in the dominant gain-of-function allele hh^Moonrat (hh^MRT) dpp is ectopically activated causing overgrowth of anterior tissues and partial duplication of distal wing structures (Felsenfeld and Kennison, 1995; Haines and Van den Heuvel, 2000). The severity of the wing patterning abnormalities in heterozygous hh^MRT flies was variable (see examples in Fig. 3A,B) and was classified into five categories, where class 1 is wild type and class 5 has the most extreme wing duplication/deformation. Since the severity of the wing phenotype depends upon excess Hh signaling in the anterior compartment, it can be dominantly suppressed by mutations in genes that promote hh signaling in either the sending or responding compartments.

To test whether Mdr49 activity is needed to potentiate Hh signaling in the wing disc, we generated hh^MRT trans-heterozygous combinations with the strongest Mdr49 allele, Mdr49^Δ3.16 and with Mdr49 deficiency Df(2R)Mdr49. Slightly less than 60% of the control hh^MRT/+ flies had class 2 wings, 30% had class 3 wings and the remaining flies belonged to either class 4 or class 5 (Fig. 3). Although none of the control hh^MRT/+ flies had wild-type class 1 wings, about 20% of the trans-heterozygous hh^MRT/Mdr49^Δ3.16 flies had class 1 wings. There was also a clear shift towards a less severe phenotype in the remaining trans-heterozygotes. There were no class 5 trans-heterozygotes and the frequency of class 3 and class 4 flies was reduced from 30% to 7% in the case of class 3 wings and from 6% to 1% in the case of class 4 wings. An even greater suppression was observed in hh^MRT/Df(2R)Mdr49. In this combination, about 80% of the flies had wild-type wings, whereas the remaining flies were class 2 (Fig. 3). In other experiments, we compared the effectiveness of three Mdr49 alleles. Consistent with their PGC migration phenotypes, Mdr49^MB04959 was the weakest suppressor, Mdr49^KG08611 was intermediate and Mdr49^Δ3.16 was the strongest (data not shown).

While the dominant suppression of hh^MRT argues that Mdr49 is needed to potentiate hh signaling, it is also possible that reducing Mdr49 activity suppresses the wing defects in this gain-of-function allele because it functions in the downstream dpp signaling pathway. To address this issue, we used a wing pouch-specific driver nubbin-Gal4 to overexpress Mdr49 in the wing and then assayed the expression of Ptc, which is a direct target of the Hh signaling pathway. As evident from a comparison of Ptc expression in the control and the UAS-Mdr49/nub-Gal4 wing discs, the total level Ptc protein accumulation increases in the presence of ectopic Mdr49. This increase in the level of Ptc protein is quantitated in Fig. 4C, which also shows that in addition to increasing the total amount of Ptc protein, Ptc expression is induced in cells in the anterior compartment that are located farther from the compartment border than in the wild type.

The experiments in the previous section indicate that ectopic Mdr49 enhances hh signaling in wing discs. To test whether hh signaling is sensitive to a reduction in Mdr49 activity we used nubbin-Gal4 to drive the expression of a UAS-Mdr49RNAi. Fig. 5 shows that the Mdr49 RNAi knockdown indeed led to a reduction in the level of Ptc protein accumulation (compare panels A and C with B and D, respectively). However, the effect of the knockdown of Mdr49 on Ptc levels (Fig. 5E) was modest compared with the upregulation of Ptc observed when Mdr49 is overexpressed.

To assess whether this reduction in Ptc levels leads to phenotypic consequences, we examined the wings of adult flies. Indeed, we found that the area between wing veins L3 and L4 was decreased in both female and male knockdown files (nubbin-Gal4×UAS Mdr49RNAi). In several instances, the anterior crossvein was completely missing (Fig. 5G). In addition, the total area of the adult wing was also reduced in the knockdown flies of both sexes (Fig. 5, compare F and G). To confirm the effects of the knockdown, we measured the wing area of d.f.(2R)Mdr49/Mdr49^Δ3.16 trans-heterozygous flies. As in the case of the knockdown, Mdr49 mutant flies also displayed reduced wing size and area compared with the wild type (not shown). These wing phenotypes are characteristic of mutations in Hh pathway components such as smo, fused (fu) and ci.

The results of the loss- and gain-of-function analysis presented in the previous sections show that Mdr49 functions to potentiate the hh signaling pathway in the wing disc. However, it could be argued that it does not play a similar role in hh signaling during embryonic development in the time frame when PGCs are migrating towards the SGPs. To address this possibility, we examined expression of two segment polarity genes, wingless (wg) and engrailed (en), that are ‘downstream’ of hh in germband extended embryos. During early embryogenesis, the expression of wg, en and also hh are controlled by gap and pair-rule genes and are independent of each other. However, later in development, during germband extension, their regulation switches to a positive autoregulatory loop. In this loop, the Hh ligand expressed by the anterior-most cells in each parasegment signals to the posterior-most cells in the neighboring parasegment, inducing wg expression. In turn, the Wg ligand signals back to the hh-expressing cells, promoting the expression of the En transcription factor. En then activates hh expression. The effects of this autoregulatory loop first become evident only by mid-embryogenesis (stage 11-12). As a result, in mutants such as dispatched or Hmgcr that disrupt the hh signaling pathway, the pattern of wg and en expression is relatively normal during earlier stages of embryonic development. However, their expression begins to decay during germband extension and clear differences from the wild type become evident in stages 11-13 embryos.

To determine whether Mdr49 is required for optimal hh signaling during mid-embryogenesis, we probed stage 11-15 Mdr49^Δ3.16/Df(2R)Mdr49 mutant embryos using antibodies against En and Wg. Fig. 6 shows that both En and Wg protein levels were reduced in stage 11 Mdr49^Δ3.16/Df(2R)Mdr49 mutant embryos compared with the control. Similar results were observed in embryos homozygous for Mdr49^Δ3.16 and Df(2R)Mdr49 mutant embryos (not shown). However, as was observed for PGC migration, the reduction in En and Wg expression was less pronounced in embryos homozygous for the Mdr49^Δ3.16 allele compared with the trans-heterozygous combination. Likewise, the reduction in En and Wg was greater in embryos homozygous for Df(2R)Mdr49.

The results in the previous section support the idea that reducing Mdr49 activity during embryogenesis compromises the range and/or strength of hh signaling. To confirm this suggestion, we examined the pattern of Hh protein accumulation in the ectoderm of germband extended Mdr49^Δ3.16/Df(2R)Mdr49 and Df(2R)Mdr49 embryos. Since the cholesterol modification is crucial for the processing and subsequent release/transmission of the Hh ligand, our hypothesis would predict that the distribution of Hh protein would be abnormal when Mdr49 activity is reduced. The confocal images shown in Fig. 7A-D show that this is indeed the case. There are two striking differences in the pattern of Hh accumulation in the Mdr49 mutant embryos. First, hh-expressing cells accumulate higher levels of Hh protein. Second, Hh protein is not properly transmitted from the expressing cells to the neighboring cells in the parasegment. To quantitate the differences in Hh protein accumulation, we generated a series of sequential scans across the parasegment (see Fig. S2). In Mdr49 mutant embryos, Hh protein was retained in Hh-expressing cells, whereas only relatively low levels were transmitted into the parasegment (Fig. S2, right-hand side).

In the studies reported here, we investigated a possible connection between Mdr49, a recently discovered component of the germ cell migration pathway, and Hh function. The first indication that there is a close connection between Mdr49 and hh signaling came from the discovery that the PGC migration defects evident in embryos compromised for Mdr49 activity can be rescued by feeding their parents a high-cholesterol diet. Since the production of a functional Hh ligand requires the cleavage of the precursor polypeptide that is catalyzed in part by a cholesterol-dependent nucleophilic attack, the most plausible explanation for this finding is that the function of Mdr49 in PGC migration is to ensure that the SGPs are able to generate sufficient levels of functional cholesterol-modified Hh. Consistent with this hypothesis, our experiments demonstrate that Mdr49 is a component of the Hh signaling pathway in two independent contexts – the wing disc and the embryonic ectoderm. In the wing disc, ectopic expression of Mdr49 potentiates Hh signaling, most notably by increasing not only the amplitude of the response but also the distance that Hh signal is able to travel effectively. Mdr49 mutations dominantly suppress the phenotypic effects of the gain-of-function hh allele, hh^MRT, whereas hh signaling in a wild-type background is weakened by RNAi knockdown. In the embryo, Mdr49 mutations disrupt the en→hh→wg→en positive autoregulatory loop during mid-embryogenesis. Finally, we have shown that the release of Hh from hh-expressing cells and its transmission to the neighboring cells in the ectoderm is compromised in Mdr49 mutant embryos.

A question that remains to be answered is the precise function of Mdr49 in cholesterol transport. Is it needed in the SGPs, in which case its function will be cell autonomous? Or does it function in the mesoderm, in which case its germ cell migration function would likely be cell non-autonomous? Based on what is known about the activity of the Mdr49 homolog in mammals, Mdr1/P-cg, a cell-autonomous function, for example in the transport of cholesterol from the plasma membrane of SGPs to the ER, would be plausible. In either instance, yet other ABC transporters (e.g. ABCG1 and ABCA1) will probably be needed to ensure that the SGPs have a sufficient supply of cholesterol for the processing of the Hh precursor (Phillips, 2014). In this context, it will also be of interest to determine whether other proteins involved in the transport or sequestration of cholesterol, such as the fly Niemann-Pick type C proteins, Npc1a and Npc1b (Huang et al., 2005; Fluegel et al., 2006; Voght et al., 2007) are also required for both hh signaling and PGC migration.

Attractive and repulsive signals generated by somatic tissues guide the embryonic PGCs to their ultimate destination, the SGPs. One of the PGC attractants produced by the SGPs is the signaling molecule Hh (Deshpande et al., 2001). While it is known that Hh (or Sonic Hh) can direct cell migration in other biological contexts (Charron et al., 2003; Yam et al., 2009), significant questions remain about how this ligand is able to function as a guidance cue.

For example, although Hh is expressed in the SGPs, this is not the only source of Hh in the embryo that could potentially signal to the migrating PGCs during mid-embryogenesis. In the mesoderm, the fat body precursor cells, which are located in more anterior parasegments (PS 5-9) than the SGPs, also express Hh (Deshpande et al., 2001; Renault et al., 2009). Moreover, in the overlying ectoderm, the cells that define the anterior border of each parasegment synthesize Hh (Tabata and Kornberg, 1994; Struhl et al., 1997). Unless there are mechanisms that effectively differentiate the Hh produced by the SGPs from these other sources, one would expect that that Hh expressed by the fat body precursor cells and cells in the ectoderm could potentially signal to the PGCs and direct their migration away from the SGPs. The same problem of distinguishing the bona fide source of Hh/Sonic Hh from sources that could provide misleading guidance cues exists in other contexts where this signaling molecule is thought to function in directed cell migration.

One mechanism that could help distinguish Hh produced by the SGPs from other sources in the embryo would be a ‘combinatorial’ migration signal consisting of Hh (or Sonic Hh) and one or more additional signaling molecules (Ricardo and Lehmann, 2009). A mechanism of this type for PGC migration was suggested by the discovery that the ABC transporter Mdr49 is required for PGC migration. While a combinatorial code specific for SGP-PGC communication remains a very attractive possibility, no other PGC attractant besides Hh has been identified despite nearly two decades of research. However, several factors that are expected to specifically potentiate Hh signal emanating from the SGPs but not other hh-expressing cells in the embryo have been discovered.

As described above, the release of the lipidated Hh ligand from basolateral membranes is limited by geranylation of the Gγ1 subunit of the heterotrimeric G protein complex, GαGβGγ1. The extent of Gγ1 geranylation is, in turn, determined by the rate-limiting step in the isoprenoid biosynthetic pathway – the synthesis of mevalonate by Hmgcr. During mid-embryogenesis, when the SGPs are sending Hh signals to the migrating PGCs, the only cells in the embryo expressing Hmgcr are the SGPs. Though much less precise, a transcription-based mechanism for potentiating Hh signals produced by the SGPs is also in place for the shifted (shf) and Mdr49 genes. shf encodes an extracellular protein that functions in Hh transmission through the basolateral heparan sulfate proteoglycan. While the transcription of shf is not restricted to SGPs, its expression is controlled by Abdominal-B and it is specifically upregulated in mesodermal cells in the posterior parasegments of the embryo during mid-embryogenesis (Zhai et al., 2010; Deshpande et al., 2013).

Although we can not formally exclude the possibility that Mdr49 has a novel cholesterol-dependent but Hh-independent function in the mesodermal SGPs, the experiments that we have presented here would be most consistent with the idea that Mdr49 functions to potentiate the Hh signals emanating from the SGPs. Because Ricardo and Lehmann (2009) have shown that Mdr49 transcription is restricted to mesodermal cells at this stage as well, this would place Mdr49 in the same category as two other components of the hh signaling pathway – Hmgcr and shf. A plausible inference from these observations is that the potentiation of the Hh signals produced by the SGPs depends primarily on the restricted expression of Hmgcr; however, the effects of Hmgcr are further augmented by the mesoderm-limited expression of shf and Mdr49. In this context, it seems likely there will be yet other genes that help to augment Hh signals emanating from the SGPs.

Since most of the genes that have been implicated in SGP-dependent migration of PGCs have turned out to function in the Hh signaling pathway, it is possible that the only guidance molecule produced by the SGPs is Hh. In this case, instead of a ‘combinatorial’ migration signal, guidance specificity would be achieved by combining multiple mechanisms for enhancing the Hh signals emanating from the SGPs. An important question is whether a similar combinatorial enhancement strategy is also used in other systems in which Hh/Sonic Hh has been implicated as a guidance cue. In mammals, at least some of the molecules that potentiate Hh signals from the SGPs in fly embryos have functions in other signaling pathways (e.g. Shifted: Glise et al., 2005; Gorfinkiel et al., 2005). This means that even if the same strategy is used to generate specificity in mammals, the precise mechanisms for enhancing Sonic Hh signaling from the bona fide source can not be the same as that used in flies. Another interesting question that has remained unresolved is how the Hh ligand influences the motility of PGCs. The transcription-based canonical signal transduction pathway is likely to be inactive in the migrating PGCs. Future experiments will thus explore the non-canonical pathway(s), downstream of Hh, that are specifically involved in cell migration.

Embryo collection and staining were performed essentially as described previously (Deshpande et al., 2013). Vasa (from Paul Lasko, Department of Biology, McGill University, Canada) and Hh (from Tom Kornberg, University of California, San Francisco, USA) antibodies are rabbit polyclonal antibodies. Both were used at 1:500 dilution. Engrailed and Wingless antibodies (Developmental Studies Hybridoma Bank) are mouse monoclonal antibodies and were used at 1∶10 dilution. β-galactosidase antibody was either a rabbit polyclonal [MP Biomedicals (previously Cappel), 55976; 1:1000] or a mouse monoclonal antibody (Developmental Studies Hybridoma Bank; 1:10). GFP antibody is a rabbit polyclonal (Torrey Pines Biolabs, ABIN2562810; 1:1000). Secondary antibodies were purchased from Molecular Probes and used at 1:500. For confocal analysis 40× magnification was used in almost all the instances, and images were collected using identical settings for the control and experimental samples on a Nikon A1 microscope with GaAsP detectors. Multiple pairs of wild-type (sibs) and mutant embryos were imaged in each case and representative examples are presented.

Imaginal discs from third instar larvae were fixed and stained by standard techniques. The specific primary antibodies used were: mouse anti-ptc [1:100; DSHB Drosophila Ptc (Apa 1)-s]. Secondary antibodies were conjugated with Rhodamine Red-X or Cy5 (1:400; Jackson Labs). Images were taken on a TE2000-E confocal microscope (Nikon). Figures were edited using Adobe Photoshop 7.0.

Gγ1 mutant stocks, Gγ1^N159 and Gγ1^k0817, were obtained from Fumio Matsuzaki (RIKEN Center for Developmental Biology, Kobe, Japan). The Mdr49 mutants and overexpression stocks were obtained from Ruth Lehmann (Skirball Biomedical Institute, NYU, USA). The UAS-RNAi lines specific for Mdr49 (Bloomington #32405) and Gal4 stocks used for the misexpression studies were from the Bloomington Stock Center (patched-Gal4, UAS-β-galactosidase, hh-Gal4/TM6 Ubx-lacZ). In most experiments, males carrying two copies of the UAS transgene were mated with virgin females carrying two copies of the Gal4 transgene. Embryos derived from the cross were fixed and stained for subsequent analysis. The genotypes of the embryos were unambiguously determined by using the balancer chromosomes marked with either GFP or β-galactosidase.

The following stocks were used for analysis of wing discs: yw; nub-Gal4; UAS-GFP, UAS-dicer-2; nub-Gal4 (Bloomington #25754), UAS-Mdr49 (a gift from Ruth Lehmann), UAS-RNAi-Mdr49 (VDRC#108237, #VDRC42513, Bloomington #32405). Transgenes were expressed using the Gal4-UAS binary system.

Drosophila food containing cholesterol was prepared as described previously (Fluegel et al., 2006). Cholesterol (Sigma) was used to prepare a stock solution (30 mg/ml) in ethanol. The final concentration was 200 ng/ml. To achieve uniform dispersion of cholesterol, food was mixed thoroughly after cooling it down to 65°C before pouring into either vials or bottles.

For antibodies and stocks, we acknowledge the Developmental Studies Hybridoma Bank, Drs P. Lasko, T. Kornberg, I. Guerrero, R. Lehmann, F. Matsuzaki, E. Olson and the Bloomington Stock Center. We would like to thank Tom Kornberg and Trudi Schupbach for helpful discussions and overall encouragement. We thank Dr Gary Laevsky and the Molecular Biology Confocal Microscopy Facility, which is a Nikon Center of Excellence. Gordon Grey provided fly media.