ABSTRACT
During development, many cell types migrate along stereotyped routes determined through deployment of cell surface or secreted guidance molecules. Although we know the identity of many of these molecules, the distances over which they natively operate can be difficult to determine. Here, we have quantified the range of an attractive signal for the migration of Drosophila germ cells. Their migration is guided by an attractive signal generated by the expression of genes in the 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (Hmgcr) pathway, and by a repulsive signal generated by the expression of Wunens. We demonstrate that the attractive signal downstream of Hmgcr is cell-contact independent and acts at long range, the extent of which depends on Hmgcr levels. This range would be sufficient to reach all of the germ cells for their entire migration. Furthermore, Hmgcr-mediated attraction does not require Wunens but can operate simultaneously with Wunen-mediated repulsion. Finally, several papers posit Hedgehog (Hh) as being the germ cell attractant downstream of Hmgcr. Here, we provide evidence that this is not the case.
INTRODUCTION
Cells are often on the move. Microorganisms migrate to find nutrients or a suitable host. Cells in developing embryos can be swept around via large morphogenetic movements, or move either individually or as small collectives of cells pushing through and between tissues. Cells find their way by detecting secreted or cell surface molecules that act as either chemoattractants or chemorepellants. Chemoattractants may be secreted by destination tissues and also by cells along the migratory route that act as intermediate targets. Localised destruction or uptake of chemoattractants are often important for shaping these gradients, as well as encouraging cells to leave the intermediate staging points (Yu et al., 2009; Boldajipour et al., 2008). Cells may also use multiple chemoattractants simultaneously, such as in the case of border cells in the Drosophila ovary (Duchek and Rorth, 2001; Duchek et al., 2001).
One cell type whose migration has been studied extensively is the primordial germ cells, the cells that give rise to the gametes in adults. They are formed early in development and migrate during embryogenesis to the gonad in many model organisms (Barton et al., 2016). Their prominence as a model for cell migration arises from their importance for species continuation, ease of identification by morphology, position and gene expression profile, and highly stereotyped migratory routes.
Drosophila primordial germ cells initially moved owing to gastrulation rearrangements from their site of formation at the posterior pole into the posterior midgut pocket. Migration begins with the germ cells pulling away from each other and traversing the posterior midgut (Seifert and Lehmann, 2012) (Fig. 1A, stage 10). They move towards the dorsal side of the midgut epithelium and enter the overlying mesoderm, partitioning bilaterally (Sano et al., 2005; Fig. 1A, stage 11). In the mesoderm they associate with the somatic gonadal precursors (SGPs) (Fig. 1A, stage 12), at which point their migration ceases, and together they coalesce to form the embryonic gonad (Boyle and DiNardo, 1995; Fig. 1A, stage 14).
Genetic screens in Drosophila have identified two important enzymatic pathways for germ cell migration. The first comprises enzymes of the 3-hydroxy-3-methylglutaryl coenzyme A (Hmgcr) pathway which catalyses the conversion of acetyl groups to the isoprenoids farnesyl- and geranyl geranyl-pyrophosphate, which are used for protein prenylation, as well as being precursors for other lipids (Bellés et al., 2005). Mutations in the Drosophila Hmgcr gene [also known as columbus (clb)] cause germ cells to scatter over the posterior of the embryo (Van Doren et al., 1998). Hmgcr is expressed broadly in the mesoderm before becoming enriched in just the mesodermally derived target tissue, the SGPs (Van Doren et al., 1998). Ectopic expression of Hmgcr, in tissues such as the CNS or the ectoderm, is sufficient to attract a small number of germ cells into the tissue of ectopic expression (Van Doren et al., 1998; Ricardo and Lehmann, 2009). These data suggest that the Hmgcr pathway produces a chemoattractant that attracts the germ cells to the SGPs (Fig. S1).
Some studies report that Hedgehog (Hh) is the Hmgcr-dependent germ cell attractant (Deshpande and Schedl, 2005; Deshpande et al., 2001, 2013). However, Hh itself is not prenylated (Eaton, 2008) and the ability of Hh to attract germ cells has not proven reproducible (Renault et al., 2009). Therefore, the identity of the chemoattractant molecule downstream of Hmgcr remains controversial.
The second pathway involved in Drosophila germ cell migration comprises two enzymes, Wunen and Wunen2 (encoded by wun and wun2), hereafter collectively referred to as Wunens. The Wunens are lipid phosphate phosphatases (LPPs), integral membrane enzymes that can dephosphorylate and internalise extracellular lipid phosphates (Sigal et al., 2005). The Wunens are expressed in somatic regions that germ cells do not normally enter and, in their absence, germ cells scatter over the posterior of the embryo (Starz-Gaiano et al., 2001; Zhang et al., 1997). Overexpression of Wunens blocks germ cell entry into the ectopic tissue and induces death of many germ cells (Starz-Gaiano et al., 2001). In a purely phenomenal (but not necessarily a mechanistic) sense, Wunen expression can be thought of as repelling germ cells (Fig. S1). Wunens are also expressed on germ cells themselves (Hanyu-Nakamura et al., 2004; Renault et al., 2004) and this leads to germ-cell–germ-cell repulsion that may be responsible for their initial dispersal out of the posterior midgut (Renault et al., 2010; Fig. S1).
The prevailing idea is that Wunens act to remove an extracellular lipid phosphate attractant (Renault et al., 2004). Although this molecule has not been identified for Drosophila, in the ascidian Botryllus schlosseri, sphingosine 1-phosphate (S1P) can direct germ cell migration (Kassmer et al., 2015). S1P is an in vitro substrate for LPPs (Roberts et al., 1998) raising the possibility that S1P, or a related molecule, acts as an attractant in Drosophila. Recent work has demonstrated that the signal downstream of Wunens is likely perceived by germ cells using Tre1, a G-protein-coupled receptor (GPCR) (LeBlanc and Lehmann, 2017).
The most recently proposed model of germ cell migration supposes that the Hmgcr and Wunen pathways work independently of each other (Barton et al., 2016). SGPs produce a prenylated germ cell attractant via the action of the Hmgcr pathway. This prenylated attractant is perceived by germ cells via an unidentified receptor and acts as an attractant. Wunens act on a different molecule, which also acts as a germ cell attractant, such as extracellular S1P or a related lipid, creating a gradient through its localised destruction (Barton et al., 2016). The Tre1 GPCR on germ cells is responsible for sensing the substrate of the Wunens (LeBlanc and Lehmann, 2017) leaving the identity of the germ cell receptor for the Hmgcr-dependent chemoattractant unknown.
Such a model leaves several open questions. Firstly, do the two chemoattractants operate with similar or different characteristics? Perhaps one is long range to get the germ cells initially moving in the right direction from the midgut while the other acts over a short range to finesse the later migration to the SGPs. Secondly, how do germ cells integrate these two signals? For example, how would germ cells respond when given conflicting guidance information by these two pathways? Perhaps, in this scenario, one pathway is dominant over the other.
Previously, we have shown that Wunens expressed in somatic cells repel germ cells without the need for cell-to-cell contact over at least a distance of 33 µm, implying they regulate a long-range diffusible signal (Mukherjee et al., 2013). In this paper, we have used germ cell response to ectopic Hmgcr expression to obtain quantitative information on the range of the Hmgcr-dependent signal. We show that, like Wunens, the Hmgcr-dependent signal also acts at long range and can attract germ cells at distances of up to 51 µm. We have used epistatic analyses to investigate the relationship between the hmgcr pathway, wun and hh. We find that hh does not act downstream of Hmgcr in attracting germ cells and that Wunens are not essential for Hmgcr-mediated attraction. Finally, we discuss these data in relation to models of germ cell migration that posit one versus two chemoattractants.
RESULTS
Ectopic Hmgcr expression is sufficient to attract germ cells into the ectopic domain
To address the question of whether Hmgcr produces a short- or long-range signal, we wanted to examine the distances that germ cells migrate when entering domains of ectopic Hmgcr expression (hereafter termed the ectopic domain). We constructed a tagged UAS Hmgcr overexpression construct allowing us to simultaneously attract germ cells and visualise the region of misexpression. Ectopic expression of HmgcrGFP was as effective at disrupting germ cell migration as previously described untagged Hmgcr constructs, indicating that the HmgcrGFP fusion protein was functional (Fig. S2).
We next wanted to ascertain whether Hmgcr expression could attract germ cells into ectopic domains as was suggested previously using CNS and ectodermal Gal4 lines (Van Doren et al., 1998). We used the Gal4 driver line NP5141 previously used to measure the repulsive forces exerted by the Wunens (Mukherjee et al., 2013). This driver expresses in parasegments 2 and 14. The former is far enough anterior that it is unlikely to affect the germ cells, whereas the latter, at stages 10–11, lies dorsally, but posterior to, where the germ cells would normally migrate (Fig. S2A). We found that HmgcrGFP expression in the NP5141 domain is sufficient to attract germ cells away from their normal migration route and for them to enter the ectopic domain (Fig. 1I,E).
To determine the time during which the germ cells were attracted, we examined the number of germ cells in the ectopic domain at different stages. Germ cells were inside the ectopic domain from stage 10 when germ cells have just crossed the posterior midgut and are starting to enter the mesoderm (Fig. 1B,F). Between stages 10 and 12 there were significant increases in the number of germ cells in the ectopic domain (Fig. 1C–E,G–I,R), indicating that germ cell attraction occurs continually rather than at a discrete time point. However, between stages 12 to 14 there was no significant increase in the number of germ cells in the ectopic domain (Fig. 1R). It is at these stages that germ cells contact the SGPs suggesting that this may curb attraction to the ectopic domain.
Ectopic and endogenous domains of Hmgcr compete to attract germ cells
In the above experiment, the SGPs (which naturally express Hmgcr) and the ectopic Hmgcr are likely competing to attract germ cells. This may lead us to underestimate the attractive range of the Hmgcr-mediated signal because potentially more germ cells would be attracted to the ectopic domain were it not for endogenous SGP Hmgcr expression.
To test this hypothesis we expressed Hmgcr using the NP5141 driver in a columbus (clb)-null background (Hmgcr loss-of-function alleles are termed clb). The number of germ cells in the ectopic domain was significantly increased compared to the wild-type background at all stages (Fig. 1F–I,N–R). Furthermore, the increase in germ cell number inside the ectopic domain continued past stage 12, unlike in the wild-type background (Fig. 1R). Therefore, germ cells can continue to migrate and be attracted to the ectopic domain even late into embryogenesis in the absence of SGP Hmgcr expression.
We conclude, firstly, that ectopic Hmgcr does compete with endogenous Hmgcr in germ cell attraction and, secondly, that the temporal limit of germ cell attraction in wild-type embryos is due to interaction with SGPs rather than a stage-dependent shut down of the germ cell migratory programme.
Hmgcr-mediated attraction can occur in the absence of somatic Wunens
Given that endogenous Hmgcr restricts the number of germ cells that can be ectopically attracted, we wanted to test whether other regulators of germ cell migration also have this effect. We therefore examined germ cell attraction in the background of a deficiency that removes somatic wun and wun2 (hereafter referred to as a wun mutant background). In some of the genetic backgrounds used in these experiments a sizeable number of germ cells failed to exit the posterior midgut properly at stage 10 (and presumably could not be attracted to the ectopic domain); therefore, we scored the number of germ cells in the ectopic domain as a percentage of the total of all the germ cells that were outside the midgut or hindgut. The latter number was however sufficiently large enough for all genotypes (Fig. 2G) to accurately assess the attractive capacity of ectopic Hmgcr.
We found that in a wun mutant background, germ cells were still attracted to ectopic Hmgcr (Fig. 2D,G). However, unlike what was seen in a clb mutant background, which drastically increased the percentage of germ cells that were attracted, there was no significant increase in a wun mutant background (Fig. 2B,D,G). This is despite the fact that in wun mutants just expressing ectopic GFP some germ cells stray into the posterior of the embryo due to random mismigration (Fig. 2A,G) as previously observed (Mukherjee et al., 2013). We conclude, firstly, that attraction to Hmgcr does not require the Wunens but, secondly, that Wunen expression does not limit attraction by ectopic Hmgcr in a wild-type background. The latter conclusion is supported by the fact that wun2 is not expressed in the region between the posterior midgut and parasegment 14 (curly bracket in Fig. S3D).
To determine whether Wunens might limit attraction in a clb background, we examined germ cell attraction in the triple mutant (loss of function for wun, wun2 and clb). We found a significant reduction in the percentage of germ cells in the ectopic domain in the triple mutant compared to in the clb mutant alone (Fig. 2C,E,G). This reduction was not due to a reduced volume of the ectopic domain in a wun background (Fig. S3C). We conclude that repulsion by Wunens does not also limit attraction by ectopic Hmgcr in a clb background; however, there is some beneficial effect of Wunen expression. This could be due to an effect of Wunen expression either directly on the gradient of the Hmgcr-mediated signal or on the positioning of germ cells to ensure they exit the posterior midgut correctly.
In support of the latter hypothesis, we found a large number of germ cells that were in the hindgut at stage 13–14 in the triple mutant background (Fig. 2E). This correlated with a large cluster of germ cells that failed to cross the posterior midgut at stage 10 (Fig. 2E′). Germ cells failing to cross the posterior midgut would normally be found in the midgut at later stages (for example when dominant versions of Rho family GTPases are expressed in germ cells rendering them unable to migrate; Renault et al., 2010). In this case, however, the germ cell cluster was further posterior towards the hindgut (arrow in Fig. 2E′) leading to the germ cells ending up there at later stages. We speculate that without Wunen expression in the posterior midgut (Fig. S1 and arrow in S3D) and without the normal early pan-mesodermal Hmgcr expression (Van Doren et al., 1998 and see also Fig. 6A) many germ cells are attracted to ectopic Hmgcr before they leave the posterior midgut. These germ cells move towards the hindgut and get trapped there, presumably as they are unable to cross the hindgut epithelium. This is in line with the observation that in srp mutants, in which the posterior midgut cells resemble the hindgut, the germ cells become stuck in the midgut (Jaglarz and Howard, 1994; Renault et al., 2010; Reuter, 1994). These data imply that germ cells are able to sense the Hmgcr-mediated signal while they are still inside the posterior midgut.
The Hmgcr-mediated signal is long range
We next wanted to make a quantitative assessment of the effective range over which the Hmgcr-mediated signal attracts germ cells. Our rationale was to determine how far germ cells are from the ectopic domain when labelled with just GFP because it is at those distances that some germ cells would be attracted when the ectopic domain expresses HmgcrGFP.
We focused on stage 10 embryos when the germ cells are first attracted to the ectopic domain. The median germ cell distance from the ectopic domain remains fairly constant between stages 10 to 13 (Fig. S4), therefore we are not overestimating the effective range by focusing on a stage in which the germ cells are particularly close. Firstly, we asked how many germ cells are in the ectopic domain in experimental embryos in which the ectopic domain expresses HmgcrGFP. In a wild-type background, this was on average two germ cells (Fig. 1R). Secondly, we took control embryos in which the ectopic domain expressed just GFP and measured the distance of every germ cell to the nearest surface of the ectopic domain, which had been computationally segmented (Fig. 3A,B). We recorded the distance of the second closest germ cell based on the assumption that, because Drosophila germ cells migrate with a high degree of directionally (figure 1D in Sano et al., 2005; Fig. 4), it would be the two closest germ cells that would have been attracted to ectopic domain if it were expressing HmgcrGFP.
The distances of the second closest germ cells were then averaged for the 10 embryos examined (Fig. 3D). In a wild-type background, the second closest germ cell was on average 31 µm (s.e.m. of 2.9 µm) from the ectopic domain, leading us to conclude that the Hmgcr-mediated signal is able to attract germ cells over at least this distance.
We next considered whether we might be underestimating the effective range of the Hmgcr-mediated signal, because germ cells in wild-type embryos are still subject to competition from endogenous Hmgcr in SGPs (Fig. 1). Therefore, we applied the same methodology to clb mutant embryos. In this case, there were on average 13 germ cells in the ectopic domain expressing HmgcrGFP in clb mutant embryos (Fig. 3C). In a clb mutant embryo with the ectopic domain expressing GFP alone, the thirteenth closest germ cell was on average 51 µm (s.e.m. 3.4 µm) from this domain (Fig. 3E). We conclude that the Hmgcr-mediated signal is able to attract germ cells over at least 51 µm.
To distinguish whether the long-range signal is a diffusible molecule (and therefore cell-contact independent) versus acts via cytonemes (and therefore cell-contact dependent) we tested whether Hmgcr overexpressing cells send out long protrusions. Upon examining embryos with ectopic co-expression of Hmgcr and the membrane marker CD2 using the NP5141 driver, we were unable to visualise any protrusions from cells of the ectopic domain (Fig. 3F,F′, image representative of 10 embryos). Taken together, we conclude that Hmgcr acts to produce a long-range, cell-contact independent (and therefore paracrine) signal in Drosophila embryos, which attracts germ cells.
The range of the Hmgcr-mediated signal is concentration dependent
To test whether the range of Hmgcr-mediated signal is dependent on the levels of Hmgcr expression we tested whether increasing the ectopic Hmgcr expression level could increase the number of germ cells attracted to the ectopic domain. We found that while having an extra copy of the UASHmgcrGFP transgene had no significant effect, having two copies of both the NP5141 Gal4 driver and UASHmgcrGFP transgenes significantly increased the number of germ cells in the ectopic domain (Fig. 4A–D).
To test whether this effect also occurred in a clb mutant we increased the ectopic Hmgcr expression level by having two copies of the NP5141 Gal4 driver in a clb mutant background. In such embryos, it was possible to attract virtually all of the germ cells into the ectopic domain (Fig. 4E, compare the number of germ cells outside the ectopic domain to Fig. 1F). The very furthest germ cell was on average 92 µm (n=10, s.e.m. 4.2 µm) from the ectopic domain upon crossing of the posterior midgut at stage 10 in clb mutant embryos expressing ectopic GFP and remained at that same distance in stage 11 and 12 embryos (data not shown). Therefore, with higher levels of Hmgcr expression, the range of the Hmgcr-mediated signal is increased. Taken together, we conclude that the range of attraction of the Hmgcr-mediated signal is dependent on the level of Hmgcr expression and these data support that it acts at a long range.
Live imaging of ectopic germ cell attraction supports the long-range nature of the signal
So far, we have estimated the range of the Hmgcr-mediated signal by analysing germ cells in fixed embryos. To see whether we could observe germ cells being ectopically attracted over such distances in living embryos, we used light-sheet microscopy and a nanos>moeGFP construct to visualise the germ cells (Sano et al., 2005).
In a control embryo, in which the amnioserosa was labelled with GFP using Krüppel-Gal4 and UAS GFP constructs, to visualise the germ band movements of the embryo, the germ cells moved from the posterior midgut pocket to the gonad over a period of ∼6 h. The path of migration and the lack of noticeable germ cell death indicates that the embryos were not adversely affected under the imaging conditions used (Fig. 5A).
In an experimental embryo in which the ectopic domain expressed untagged Hmgcr to avoid interference with the germ cell labelling, we observed germ cells migrating to the ectopic domain. We tracked the majority of germ cells and colour coded them according to whether they migrated to the gonad (Fig. 5B, blue/cyan tracks), to the ectopic domain (Fig. 5B, pink/purple tracks) or that remained at the midline (Fig. 5B, yellow tracks). We saw that germ cells entered the ectopic domain from late stage 10 and continued to enter until late stage 12 (Fig. 5B,C). Once associated with the gonad at stage 13, germ cells remained there and did not exit and migrate to the ectopic domain. These observations are in agreement with those from the fixed embryo analysis. Once in the ectopic domain, germ cells remained there and stopped migration, indicating that high levels of Hmgcr even in non-SGP somatic cells are sufficient to stop the migratory programme of the germ cells (Fig. 5B).
We then focused on the portions of migratory movements of germ cells entering the ectopic domain in which the germ cells broke away from the normal migratory path. We measured the distance over which this abnormal migration took place (Fig. 5C). At stage 11, we observed two germ cells each migrating for ∼47 µm to enter the ectopic domain and at late stage 12 we observed two germ cells migrating 39 µm and 41 µm to enter the ectopic domain (Fig. 5C). These distances are in strong agreement with our estimates of a range of 51 µm from fixed embryos (Fig. 3D) and support the notion that Hmgcr is mediating a long-range signal.
Germ cells are within range of the Hmgcr-mediated signal throughout their migratory journey
To see how our estimate for the range of the Hmgcr-mediated signal compares to the distance of germ cells to Hmgcr-expressing SGPs in wild-type embryos, we measured such distances in stage 10 and 11 embryos (Fig. 6A–C). We found that germ cells were located between 5 and 58 µm from their closest SGP at stage 10 and ranged from 0 to 30 µm at stage 11. Therefore, for stages 10 and 11, 98% and 100% of germ cells respectively would be within our estimate of 51 µm for the range of the Hmgcr-mediated signal. We conclude that germ cells are potentially under the influence of the Hmgcr-mediated signal for their entire migratory journey.
Hmgcr and wun operate simultaneously
We next wanted to know which of the two pathways, Hmgcr or Wunen, is dominant. To do this we gave germ cells conflicting guidance cues by simultaneously attracting them to the ectopic domain using Hmgcr expression and repelling them by co-expressing wun. When wun is expressed using the NP5141 Gal4 driver there is no effect on overall germ cell migration (Mukherjee et al., 2013; Fig. 7A–D). When wun and Hmgcr are co-expressed, germ cells are still attracted towards the ectopic domain to a similar degree to when Hmgcr was expressed alone (Fig. 7E–H). However, despite some germ cells arriving at the ectopic domain as early as stage 11 (Fig. 7F), germ cells are not subsequently found within the ectopic domain but instead remain at its boundary (Fig. 7G,H).
This positioning of the germ cells could result from attraction to the ectopic domain by Hmgcr and then wunen activity either repelling germ cells from entering the domain or killing those germ cells that do enter, as happens for example when Wunens are ectopically expressed in the mesoderm (Starz-Gaiano et al., 2001). To distinguish between these two hypotheses, we tested what would happen if larger numbers of germ cells were to be attracted to the ectopic domain. We performed the same experiment in a clb mutant background in which competition for attraction by the SGPs is eliminated. In this scenario, the number of germ cells attracted to the ectopic domain was indeed increased and significantly more germ cells accumulated at the ectopic domain boundary (Fig. 7M–Q).
Despite the large number of germ cells being attracted, we did not observe germ cells or remnants of dying germ cells inside the ectopic domain, making it unlikely that germ cells were entering the ectopic domain and dying. This interpretation is supported by two further pieces of evidence. The first is the change in the number of germ cells at the ectopic domain border between stages 12 and 14. In the case of Hmgcr expression in both wild-type and clb mutant backgrounds, this number decreases as germ cells move past the border and enter the domain. When wun is co-expressed, however, this number increases as more germ cells arrive and those already present fail to move past the border (Fig. 7Q). If germ cells were dying then we would predict that this number would fall as germ cells enter the domain and then die. The second is the overall germ cell survival rate between stage 10 and stages 13–14, which is similar in embryos ectopically expressing Hmgcr in a clb mutant background compared to those ectopically expressing both wun and Hmgcr in a clb mutant background (Fig. 7R). This suggests that co-expression of wun is not causing extensive germ cell death.
These data show that Wunens can repel germ cells and prevent them from entering an Hmgcr-expressing ectopic domain. Taken together, we conclude that neither the wun nor the Hmgcr pathway is dominant and germ cells position themselves using the information provided by both pathways simultaneously.
hh is not required downstream of Hmgcr for germ cell attraction
We wanted to test whether hh is required downstream of Hmgcr for the attraction of germ cells. We therefore asked whether germ cells could be attracted to ectopic Hmgcr in an hh-null background. If hh is the attractant downstream of Hmgcr, we would predict that germ cells would not be attracted to Hmgcr in a hh background. On the other hand, if hh is not the downstream attractant, we would predict that germ cells would still be attracted to ectopic Hmgcr in a hh mutant.
We used the null allele, hhAC, which, when homozygous, causes embryos to have cuticles with a characteristic strong hh phenotype consisting of a continuous lawn of denticles identical to that published in Lee et al. (1992) (Fig. S6A,B). hhAC embryos have very severe patterning defects that are evident from stage 13, which causes germ cells to scatter over the poorly patterned posterior of the embryo. Therefore, we examined earlier hhAC embryos, at stage 12, when germ cells are mostly on track and none have mismigrated into a control ectopic domain that expresses just GFP (Fig. 8A,B). We founnd that ectopic expression of HmgcrGFP can attract germ cells in hh homozygous mutant embryos similar to in sibling heterozygous controls (Fig. 8C–E). We conclude that zygotic hh is not required downstream of Hmgcr for attracting germ cells.
One caveat is a potential for maternally provided hh message to be a source of Hh that acts downstream of Hmgcr. Although maternal hh message was not previously detected by northern blot analysis (Lee et al., 1992), we checked for potential perdurance of maternal hh mRNA in the ectopic domain by in situ hybridisation. We do not see hh mRNA at stage 10 in hhAC mutant embryos (Fig. S6E,F), therefore we find no evidence of a role for maternally provided hh downstream of Hmgcr.
To test whether hh could be acting as a germ cell attractant independently of Hmgcr, we also tested whether ectopic hh expression from the NP5141 Gal4 driver was sufficient to attract germ cells. We found no mis-migrated germ cells in this domain under these conditions (Fig. 8E) despite the UAS hh construct being able to induce patterning defects identical to that seen in Fietz et al. (1995) when expressed using a patched Gal4 driver (Fig. S6C,D). Taken together, these data support the conclusion that hh is not the germ cell attractant downstream of Hmgcr.
DISCUSSION
Here, we have examined the range of influence of a signal downstream of Hmgcr that attracts germ cells in Drosophila embryos. We have found that this signal can act at distances of at least 51 µm and is dependent on the levels of Hmgcr overexpression. This distance is greater than the distance of virtually all of the germ cells from the target SGPs at stages 10 and 11 and therefore, distance-wise at least, should be sufficient to attract germ cells to the gonad. Furthermore, the signal can operate at the same time as a second pathway, namely that mediated by the Wunens. This is most strikingly demonstrated by finding that the simultaneous overexpression of both components in the same ectopic domain produces a phenotype different from that seen upon overexpression of either component alone, in that we see both simultaneous attraction and repulsion as the germ cells line up at the edge of the expression zone. Finally, we provide evidence that the extracellular signalling molecule Hh is not the chemoattractant downstream of Hmgcr.
Our 51 µm estimate of the range of the Hmgcr-mediated signal represents approximately six germ cell diameters (Drosophila germ cells being 8–9 µm in diameter, Fig. S5) or nine mesodermal cell diameters (Drosophila stage 12 mesodermal cells being ∼5–7 µm in diameter, Fig. S5), which would make it a long-range signal. This range is broadly in line with other long-distance signalling molecules in Drosophila and other species. For instance, in Drosophila imaginal wing discs the TGF-β family member Dpp acts at long range, influencing cells up to 20 cell diameters away (Nellen et al., 1996). In Xenopus embryos, TGF-β ligands can be detected 7–10 cell diameters away from their source (McDowell et al., 2001; Williams et al., 2004), while in zebrafish embryos, cells can respond to endogenous TGF-β (nodal) signalling at distances up to 200 µm (Harvey and Smith, 2009). The mean total length of the tracks of successfully migrating germ cells in our live imaging movies (from early stage 10 to stage 13) is 381 µm (s.e.m. 18.4 µm) (Fig. 5). Although this is much longer than our estimate of the range of the Hmgcr-mediated signal, much germ cell movement is non-cell autonomous and comes from the bulk embryonic movements of germband retraction.
Wnt ligands on the other hand can act at either short or long range. Wingless acts as a short-range inducer in Drosophila embryos, being secreted by stripes of ectodermal cells and being received only by their neighbours (van den Heuvel et al., 1989). In mouse organoids, Wnt3 also acts at short range being visualised only 1–2 cells away from synthesising cells (Farin et al., 2016). On the other hand, Wingless in Drosophila imaginal wing discs acts at long range, influencing cells 20 or more cell diameters away (Zecca et al., 1996), and EGL-20 in C. elegans can be seen in a gradient up to 50 µm from its source (Coudreuse et al., 2006).
In these examples, the ligands are providing positional information to static cells by inducing concentration-dependent transcriptional responses. In the case of the Hmgcr and Wunen, however, the responding germ cells are motile, and a transcriptional response seems unlikely given the speed and the need for signal directionally not just strength. Ligands acting in a similar fashion include chemokines such as stromal cell-derived factor 1 (SDF1; also known as CXCL12), which acts as a long-range attractant for several cell types. SDF-1a-expressing cells transplanted into zebrafish embryos can attract germ cells over distances of at least 250 µm (Blaser et al., 2005), and SDF-1-soaked beads can attract interneurons in mouse brain slice cultures over similar distances (Li et al., 2008).
We have estimated the distance over which Hmgcr is potentially able to operate via overexpression studies and shown that the range is influenced by the degree of overexpression. To ascertain the relevance of our distance estimations to the wild-type situation, we have compared the number of cells expressing Hmgcr ectopically to those normally expressing Hmgcr. We estimate there are just over 1000 HmgcrGFP ectopically expressing cells in parasegment 14 at stage 10 when driven by NP5141 Gal4 (Fig. S3A). At stages 9–10, Hmgcr is expressed broadly in the mesoderm (Van Doren et al., 1998). We estimate there are ∼250 Hmgcr-expressing mesodermal cells at stage 10 that lie dorsally to the germ cells, and to which the germ cells will migrate (Fig. S3B). By stage 12, Hmgcr is highly expressed in the SGPs (Van Doren et al., 1998) of which there are only 25–35 cells in total per gonad (Sonnenblick, 1941). Therefore, the number of cells ectopically versus endogenously expressing Hmgcr is comparable, at least at early stages.
We have examined whether the wunen and Hmgcr pathways act simultaneously rather than consecutively. We found that the pathways can act simultaneously when wunen and Hmgcr are overexpressed (Fig. 6M–P). We believe that the behaviour of the germ cells in these ectopic expression experiments is relevant to the wild-type scenario because wun2 and Hmgcr are normally expressed at the same time, although in different parts of the embryo, throughout the period when germ cells are migrating (Starz-Gaiano et al., 2001; Van Doren et al., 1998 and Fig. S3B,D).
There are two possible models of the interactions between Hmgcr and Wunen (Fig. 8F,G). The prevailing view is a two-signal model (Fig. 8G; Barton et al., 2016). One chemoattractant results from Hmgcr expression in the mesoderm and is perceived by germ cells via an unidentified receptor. The second chemoattractant is perceived by germ cells using the Tre1 GPCR (LeBlanc and Lehmann, 2017). It is also a substrate for the Wunens and is dephosphorylated and thereby destroyed by Wunen-expressing cells, including the germ cells, which collectively act as a chemoattractant sink. In this model, the spatial information provided by Hmgcr and Wunens is integrated at the level of the germ cells, which use the information provided by both chemoattractants.
In a one-signal model, Hmgcr expression would result in secretion of a chemoattractant from the mesoderm that is also the substrate for the Wunens and is detected on germ cells by the Tre1 GPCR (Fig. 8F). In this model, the spatial information provided by Hmgcr and Wunens is integrated at the level of the chemoattractant gradient, which depends on the combined actions of both of these enzymes.
Both models have precedents from other extracellular gradients both in Drosophila and other organisms. The one-signal model (Fig. 8F) resembles classical source–sink models for both chemoattractant and morphogen gradients (Cai and Montell, 2014). The use of simultaneous attraction and repulsion, as per the two-signal model (Fig. 8G), is seen in Drosophila axonal pathfinding where commissural axons are attracted and repelled by the ligands Netrin and Slit, respectively (Dickson and Gilestro, 2006). The migration of vertebrate trunk neural crest cells is controlled by both positive and negative regulators including ligand–receptor pairs such as ephrin–Eph, and Sdf1–Cxcr4 (Shellard and Mayor, 2016).
Our data do not definitively discriminate between these two models. In support of the one-signal model (Fig. 8F) we note, firstly, that the signals downstream of both Wunens and Hmgcr operate over similar long ranges, which means they are potentially the same molecule. Secondly, zygotic loss of function mutants of wun and clb both exhibit similar very strong mis-migration phenotypes with few germ cells reaching the gonad in either mutant alone (Van Doren et al., 1998; Zhang et al., 1997) similar to what is seen in the double mutant (Fig. 2E). If each pathway influenced their own independent signal, then one might expect that removal of either pathway alone would result in partial germ cell mis-migration (with only some germ cells mis-migrating) as the other would still be active and also acting over a long range.
However, some of our data are difficult to reconcile with a single signal. We would have expected that if ectopic Wunens are degrading a signal generated by ectopic Hmgcr, co-expression would decrease the range of the signal and delay the time at which cells mis-migrate to the border. However, we see almost as many germ cells at the ectopic domain border (Fig. 7P) when both HmgcrGFP and Wun are expressed in the ectopic domain in a clb mutant background, as we see germ cells inside the ectopic domain when just HmgcrGFP is expressed there in a clb mutant background (Fig. 1Q). Germ cells also reach the border at similar stages in these two conditions (compare clb NP5141>HmgcrGFP with clb NP5141>wun2myc HmgcrGFP in Fig. 7Q). Therefore, either any decrease in range is minimal or there are two signals. The alignment of germ cells at the ectopic domain border is surprisingly precise given that the Wunen-dependent signal can be contact independent and long range (Mukherjee et al., 2013). This opens up the possibility that there are several in vivo substrates of the Wunens, like there are in vitro (Renault et al., 2004), some of which could act at much shorter range.
The ultimate confirmation of which model is correct will require identification of the chemoattractant(s). It is interesting to note that germ cell migration in other species such as chicken and zebrafish seems to require only a single chemoattractant (SDF-1) in spite of the much longer migratory journeys, both in terms of distance and time, in these species (Barton et al., 2016). It is clear that Drosophila germ cells cannot be responding to SDF-1 as no SDF-1 homologue exists in flies. What is less clear is whether the signals downstream of wunens and Hmgcr exist in vertebrates, perhaps playing a more subtle role. Tantalising evidence from zebrafish suggests this might be the case, with simultaneous knockdown of all the Wunen homologues causing some germ cells to mis-migrate (Paksa et al., 2016). Therefore, the cues that regulate Drosophila germ cell migration might actually be more conserved than we first thought. In addition, Hmgcr overexpression in stomal cells acts in a paracrine fashion to promote prostate cancer cell growth (Ashida et al., 2017) suggesting that Hmgcr-mediated signals are also relevant in humans to tumour progression and metastasis.
MATERIALS AND METHODS
Fly stocks
The following Drosophila lines were described previously: Df(2R)wunGL, a deficiency removing wun and wun2 (Zhang et al., 1996); clb11.54, a loss of function allele of Hmgcr (Van Doren et al., 1998); hhAC, an amorphic allele resulting from a 8.6 kb deletion removing the promoter and part of the coding region (Lee et al., 1992); UAS wunGFP (Burnett and Howard, 2003); UAS wun2myc (Starz-Gaiano et al., 2001) [the wunGFP and wun2myc constructs behave indistinguishably – both can rescue the tracheal phenotypes caused by wun loss of function and both cause identical amounts of germ cell death when ubiquitously embryonically overexpressed (Ile et al., 2012)]; UAS lazGFP (Garcia-Murillas et al., 2006), expression of which does not affect germ cells (Mukherjee et al., 2013); UAS Hmgcr (Van Doren et al., 1998); UAS CD2 (Dunin-Borkowski and Brown, 1995); HmgcrEY04833, a UAS-containing insertion 5′ of the Hmgcr gene (stock 16619, Bloomington Stock Center); p(GawB)NP5141, a Gal4-containing insertion 5′ of the gene ken (Drosophila Genetic Resource Center); y M{vas-int.Dm}ZH-2A w; PBac{y+-attP-3B}VK00033 used as a landing site for the UASHmgcrGFP transgene. nanos>moeGFP was used to label the germ cells for live imaging (Sano et al., 2005). The following labelled balancer chromosomes were used: TM3 P{w[+mC]=GAL4-Kr.C}DC2, P{w[+mC]=UAS-GFP.S65T}DC10, Sb1 and TM3 P{ftz-lacZ.ry+}TM3, Sb1 ry*.
Immunohistochemistry and imaging
Embryos were laid at room temperature, dechorionated in 50% bleach for 3 min, fixed for 20 min in 4% formaldehyde (37% for in situ hybridisation) in PBS–heptane, devitellinised using heptane–methanol, and stained using standard protocols. Primary antibodies were as follows: polyclonal rabbit anti-Vasa (courtesy of Ruth Lehmann, Skirball Institute of Biomolecular Medicine, New York, USA, 1:10,000), rabbit anti-LacZ (MP Biomedicals 559761, lot 06680, 1:10,000), chicken anti-GFP (Abcam ab13970, lot GR89472-6, 1:1000), rabbit anti-MYC (Abcam ab9106, lot GR41743-1, 1:1000), mouse α-spectrin (DSHB 3A9, 1:10) and mouse anti-CD2 (Bio-Rad MCA154GA, lot 0515, 1:2000). Secondary antibodies conjugated to Alexa Fluor 488 or 648 (Invitrogen) and Cy3 (Jackson ImmunoResearch) were used at 1:500.
To visualise Hmgcr or wun2 expression, full-length Hmgcr or wun2 cDNA clones in pNB40 and pBSK vectors, respectively, were linearised and used to make a digoxygenin-labelled RNA probe by in vitro transcription with T7 RNA polymerase, and hybridisation and fluorescent detection was carried out as described previously (Lécuyer et al., 2008). To visualise hh expression, an 800 bp fragment of hh coding sequence was amplified by PCR from cDNA using the primers 5′-GATCGTCTTGCCGATGGTCT-3′ and 5′-CACAAACGTGAGCTTCTGGC-3′ and cloned into pGEM T-easy vector (Promega). The vector was linearised and used to make a digoxygenin-labelled RNA probe by in vitro transcription, and hybridisation and colourimetric detection was carried out as described previously (Lehmann and Tautz, 1994).
Fluorescently stained embryos were either mounted in aquamount (Polysciences) or dehydrated in methanol and mounted in benzylbenzoate–benzyl alcohol (2:1). Images were acquired using an LSM 880 confocal microscope with a 20× NA 0.5 air or 40× NA 1.3 oil objective and Zeiss Zen2 acquisition software. Live imaging was performed on a Zeiss Z1 light-sheet microscope. Embryos were aged until approximately stage 9, dechorionated, transferred into cooled, but still liquid, 1% low-melt agarose dissolved in distilled water, and drawn into a glass capillary. Once the agarose had set, the capillary tube was transferred to the light-sheet microscope and embryos imaged with a 20× NA 0.5 air objective using the 488 nm laser until the end of embryogenesis. Such embryos were able to hatch into larvae, indicating that the conditions used did not noticeably impair development. Germ cells were tracked using Imaris software (Bitplane).
Image analysis
3D reconstructions, segmentations and distance measurements were made with Imaris software (Bitplane). For germ cell distance measurements, late stage 10 embryos were chosen in which the germ cells had exited the posterior midgut. Germ cell positions were detected automatically (using the spots tool) and manually edited for accuracy. The ectopic domain was segmented using the surfaces tool, and the distance of the edge of each spot (using the Imaris minimum intensity statistic) to the nearest ectopic domain surface was measured using the MeasurementPro extension. Germ cells that were ‘stuck’ in the midgut or hindgut were identified by both being in a tight round or elongated cluster and inside a tubular structure visible as background fluorescence by increasing the channel brightness. For the scoring of germ cells inside the ectopic domain in Fig. 2G, the ectopic domain was labelled using GFP or HmgcrGFP, with the exception of wun wun2 clb mutant embryos where instead the number of germ cells within 50 µm of the embryo posterior (this being the mean length of the ectopic domain in NP5141>GFP embryos) was used.
The volumes of the NP5141>HmgcrGFP domain in stage 14 embryos was determined using the segmentation function of Imaris (Bitplane). The number of cells in the NP5141 domain and in the wild-type Hmgcr-expressing domain were scored manually in ImageJ using the cell counter plugin. The diameters of germ cells and mesodermal cells were measured in ImageJ.
Germ cell survival rates were calculated as the average total number of germ cells in stage 13–14 embryos divided by the average total number of germ cells in stage 10 embryos (when the germ cells leave the tight cluster in the posterior midgut and become easily scorable).
Generation of UASHmgcrGFP flies
The Hmgcr coding sequence was amplified from cDNA clone in pNB40 using the primers 5′-CACCATGAGGACGTTTGTTTCGC-3′ and 5′-GCTGATGGGCTGCAGCTGG-3′ and cloned into the pENTR/D-TOPO vector (Invitrogen). The sequence was verified and moved into the destination vector pUAST-attB-WG (a gift from Saverio Brogna, School of Biosciences, University of Birmingham, UK, producing C-terminal GFP fusions) with the use of the Gateway reaction. This resulting expression vector pUAST-attB-Hmgcr-WG was microinjected into embryos containing phiC31 integrase and an attP site on the third chromosome.
Acknowledgements
We thank Tom Starkey for recombinant fly stocks, the University of Nottingham School of Life Sciences Microscope facility for training on and access to the Zeiss 880 confocal microscope, and Malcolm Bennett and Antony Bishopp (School of Biosciences, University of Nottingham) for use of the Zeiss Z1 light-sheet microscope (funded through BBSRC award BB/M012212/1). We thank Christian Feldhaus (Max Planck Institute for Developmental Biology) for assistance with distance measurements in Imaris and Markus Owen, Andrew Johnson and Fred Sablitzky (University of Nottingham) for manuscript comments. We acknowledge the Flybase consortium for gene information and the Bloomington Stock Center at Indiana University for stocks. We thank the Saverio Brogna laboratory for reagents.
Footnotes
Author contributions
Conceptualization: K.K., A.D.R.; Methodology: K.K., A.D.R.; Formal analysis: K.K., A.D.R.; Investigation: K.K., A.M., A.D.R.; Writing - original draft: A.D.R.; Writing - review & editing: A.M., A.D.R.; Supervision: A.D.R.; Project administration: A.D.R.; Funding acquisition: A.D.R.
Funding
K.K. was supported by a University of Nottingham studentship.
References
Competing interests
The authors declare no competing or financial interests.