The behaviour of PGCs, SGPs and msSGPs was followed by labelling them separately with fluorescent markers: an eGFP-Vasa fusion protein for PGCs 20, and an enhancer from the third intron of Six4 21 to express nls-eGFP in SGPs and msSGPs. Wide field deconvolution microscopy enabled the capture of high resolution, three dimensional image stacks at frequent intervals over an extended period of embryonic development. Wild-type embryos that have been imaged survive to become fertile adults, indicating that our imaging protocols do not significantly perturb gonadogenesis.

For imaging PGC migration, we capture 30 z sections, separated by 1 μm, at 90-second intervals using a 20× objective lens. After mild desiccation of the embryos to reduce their thickness, this is sufficient to observe the path of PGC migration through the mesoderm to the gonads (Additional file 1, Figure 1). Previous observations in fixed tissue suggested that individual PGCs migrate via different paths 22, and our live imaging data confirm this.

To resolve the SGP and msSGP nuclei it is necessary to image at higher magnification and to use longer exposure times to achieve a sufficient range of pixel intensities. We capture 26 sections, separated by 1 μm, at 90s intervals using a 40× objective lens. During germ band retraction in wild-type embryos, the three groups of SGPs coalesce with each other to form a single cluster intermingled with germ cells by the start of stage 13 (Additional file 2, Figure 2a). The movement of the individual SGP clusters relative to each other is difficult to interpret due to the presence of intermingling germ cells. To clarify this we imaged embryos homozygous for a strong allele of Hmgcr, in which few PGCs migrate to the mesoderm 23. In this case it is clear that the three SGP clusters become closely associated by mid stage 12 (Additional file 3, Figure 2b). From stage 13, SGP movements drive the compaction of the gonad. This occurs symmetrically, with SGPs moving both anteriorly and posteriorly towards a central focus (Additional file 2). In Hmgcr mutants it is clear that the movements of gonad compaction follow the initial association of the three parasegmental SGP clusters (Additional file 3), implying that these two processes represent distinct cell behaviours at different developmental stages.

The msSGPs are seen clearly in males as they express nls-GFP to a high level. They migrate anteriorly and dorsally as a tight cluster to join the coalescing gonad at a mean rate of 12.3 microns/hr. This is slower than the rates observed for many cells that migrate individually, but is comparable to the migration rate of border cells during oogenesis, which also move as cohesive cluster 3. In females, a small number of strongly fluorescing cells migrate from parasegment 13 in a similar manner at this time (Additional file 4, Figure 2c). Observations of fixed tissue 14 suggest that a larger number of msSGPs migrate at this stage in wild type females than we observe, but that most of these undergo apoptosis during this migration. Our data suggest that most of the msSGPs in females do not express nls-eGFP, implying that the Six4 enhancer is not active in these cells. The cells that we do observe may be a subset of msSGPs in which the enhancer is active or a distinct cell type. We favour the latter hypothesis because a proportion of cells from the msSGP cluster in males also detach from the gonad shortly after their migration to it (Additional file 2). This suggests that these cells represent a small, distinct population of unknown function that is present in both sexes.

Although the genetic control of SGP specification is well studied, very little is known about their subsequent development and behaviour. The homeodomain protein Six4 is a candidate regulator of SGP function during gonad formation as it is expressed in the SGPs from their first appearance until at least the end of embryogenesis 2425. Although Six4 is required for the specification of several mesodermal lineages 21, and for the correct development of the head (unpublished data), embryos homozygous for the hypomorphic mutation Six4¹³¹ hatch normally and have only mild musculature defects 24. These embryos do show a severe defect in gonadogenesis, however. SGPs are present and appear to be correctly specified as they express the markers 412 24, Eyes absent (Eya, Figure 3b) and Zfh-1 (not shown) but most PGCs do not associate with them and become scattered.

We have determined the number of SGPs present in embryos stained for Eya and Sox100B expression. At the end of germ band retraction, wild-type male embryos contain 68.5 ± 4.84 Eya-expressing SGPs in each gonadal cluster (n = 4). This number excludes the Sox100B-expressing msSGPs 14. The number of SGPs is greater than the early estimate of 26–37 that is frequently cited 26. At a similar stage, male Six4¹³¹ homozygotes have 49.6 ± 2.36 (n = 5) Eya-expressing SGPs in this location indicating that a substantial number of SGPs is specified in the Six4¹³¹ hypomorph.

The failure of PGC migration in Six4¹³¹ embryos is unlikely to be explained by the small reduction in the number of SGPs in this mutant. An alternative explanation is that these SGPs cannot attract the PGCs. Time lapse imaging revealed that PGC migration in Six4¹³¹homozygotes begins normally. The cells initially move correctly through the gut epithelium and migrate to the mesoderm but most fail to coalesce into the gonads following germ band retraction (Additional file 5, Figure 3c). The mutant phenotype becomes overt in late stage 12 when most PGCs in wild type embryos undergo a dorsal migration to the final location of the gonad. In Six4¹³¹ mutants most PGCs remain in a more ventral and posterior position before dispersing, apparently randomly. This suggests that PGCs move correctly to the mesoderm, but subsequently fail to navigate to the SGPs. Both of these migrations require a chemo-attractant that depends on Hmgcr function, firstly in the mesoderm and then the SGPs 23. In wild type embryos, Hmgcr is expressed broadly in the mesoderm at the time that the PGCs detach from the midgut but is then restricted to the SGPs during gonadogenesis. In Six4¹³¹ mutant embryos the broad mesodermal expression of Hmgcr is still seen but there is no expression in SGPs at the time that they would normally attract the PGCs (Figure 3f). This may explain why PGCs detach correctly from the midgut but many do not enter the gonads in Six4¹³¹ mutant embryos. It seems likely therefore that Hmgcr is one of the target genes that must be regulated by Six4 for normal SGP function.

Although SGPs do associate in clusters in Six4¹³¹ homozygotes 24, time-lapse imaging reveals that there are abnormalities in SGP movement. This indicates that regulation of Hmgcr is unlikely to be the only function of Six4 in the SGPs, since gonad formation takes place in Hmgcr mutant embryos. In Six4¹³¹ mutant embryos the three SGP clusters do not merge to form a single gonadal structure. Of ten mutant embryos that we imaged only one formed a single gonad that incorporated the three parasegmental groups of SGPs. In four embryos the anterior (ps10) cluster remained isolated from the others, while in another four the posterior (ps 12) cluster remained isolated (Additional file 6, Figure 4). In one embryo all three of the SGP clusters remained isolated.

Despite the failure to form a unified gonad, SGPs within these clusters display aspects of their normal behaviour in Six4¹³¹ homozygotes. In all of the embryos we examined, a proportion of the SGPs moved towards each other to form a tight cluster, in a process resembling the compaction movements of wild type gonads. In some embryos, isolated clusters of SGPs also compacted so that two gonad-like structures were formed instead of a unified gonad (Additional file 6, Figure 4). Interestingly, several PGCs could be "captured" by these small gonad-like structures, suggesting that short-range PGC-SGP interactions are still possible even though long-range attraction requiring Hmgcr is defective. The associating PGCs may be those that randomly move close to the SGPs during their migration. The retention of some SGP behavioural characteristics in Six4¹³¹ mutant embryos supports our conclusion that the SGPs are correctly specified. Our interpretation is that most mutant SGPs retain the ability to associate with each other and with PGCs over short distances, but there is a defect in their ability to coalesce over longer distances.

The observation that compaction can occur where SGPs fail to coalesce into a single structure suggests that the two processes differ mechanistically perhaps dependent on distinct short- or long-range interactions respectively the former being Six4-independent. The long range interaction requires Six4 but not E-cadherin as extended clusters of SGPs and PGCs persist in E-Cadherin mutant embryos even though they do not compact into a spherical gonad 15. As expected, E-Cadherin expression and localisation in SGPs appears normal in Six4¹³¹ mutants (data not shown). We suggest that communication between SGP clusters during coalescence requires a Six4-dependent signalling mechanism operating between SGPs. This may direct the migration of the three clusters towards each other prior to compaction, or it may influence the polarity of the compaction process ensuring that a single gonadal cluster is formed. Six4 would positively regulate a component of this signalling pathway, either in the signalling or in the receiving SGPs, or both. When this signalling is disrupted, as in Six4¹³¹, compaction may occur via stochastic Six4-independent local cell contacts. This would account for the variability in whether the anterior or posterior SGP cluster fails to be incorporated into the gonad-like structure. Any such signal is unlikely to be Hmgcr-dependent, since mutation of Hmgcr does not prevent clusters of SGPs from associating, even though they do not attract germ cells (Additional file 3, Figure 2b). An alternative explanation for the SGP coalescence phenotype of Six4¹³¹ is that fully functional Six4 is required in SGPs for the compaction process to operate consistently over the distance between SGP clusters, perhaps by regulating the length of productive cellular protrusions that may be required for SGP-SGP contacts.

The msSGPs must migrate a substantial distance to reach the developing gonad. Given the defects in SGP movements, Six4¹³¹ embryos were examined for defects in msSGP migration. In time-lapse experiments, cells originating in parasegment 13 can be identified as msSGPs by the accumulation of high levels of nls-eGFP during stage 13. Unlike wild-type msSGPs, these cells do not migrate but remain in a posterior position (data not shown). Nevertheless, they appear to be correctly specified as they express the msSGP markers Eya and Sox100B 14 and are maintained as a cell cluster in males (Figure 5b) and cells co-expressing both Eya and an apoptosis marker are observed in mutant females as in wild type embryos (Figure 5d). Because Six4 is expressed strongly in msSGPs, as well as SGPs, we cannot yet determine if the failure of msSGPs to migrate may reflect a defect in the msSGPs themselves or defective signalling by the SGPs. The mechanism attracting msSGPs to the gonads is distinct from that attracting the PGCs as they associate correctly with SGPs in Hmgcr mutant male embryos (data not shown). It is plausible that msSGP migration and the mutual association of the SGP clusters are regulated by a common mechanism.

A specific role or roles for Six4 in cell migration may not be restricted to the embryonic gonad. A recent study found that Six4 is upregulated in the migratory ovarian border cells with respect to their stationary neighbours, and that Six4 over expression in these cells disrupts their migration 27. There are similarities in the behaviour of border cells, SGPs and msSGPs, all of which move in cohesive clusters, suggesting that Six4 may influence migratory cell behaviour through regulating some of the same targets at these different stages of reproductive development. Many aspects of Six protein function appear conserved in divergent tissues and organisms 28. In this context it is interesting to note that cell movements and shape changes during morphogenetic furrow progression in the Drosophila eye require the function of Sine oculis, another Six gene, while the C. elegans Six4 homologue, Unc39, is required both for the specification and the motility of cells migrating anteriorly during embryogenesis 2930. The combination of live imaging with the extensive genetic tools available makes the Drosophila gonad a productive system to investigate the roles of Six4 transcriptional targets in different modes of cell migration.

Wild-type flies were of the Oregon R stock. Flies expressing eGFP-Vasa in the primordial germ cells were provided by Satoru Kobayashi 20. The Six4 alleles and the nls-eGFP reporter driven by the Six4 third intron enhancer were described previously 2124.

Live embryos were collected, dechorionated, washed and mounted on coverslips in halocarbon oil as described 31. Images were collected using a DeltaVision wide field fluorescence microscope (Applied Precision) using 20×/0.75NA (PGC Videos) and 40×/1.35NA (SGP/msSGP Videos) objective lenses and a custom GFP filter cube. Up to 4 embryos were imaged in parallel by repeat visiting using a highly accurate XYZ motorised stage. Images were collected using a Roper Coolsnap HQ camera and binned (2 × 2) to increase the range of intensities. Out of focus light was reassigned by iterative deconvolution 31. Images and Videos were prepared for presentation using Softworx (Applied Precision), Image J (NIH), Photoshop (Adobe) and Quicktime (Apple) software. The mean rate of msSGP migration was calculated from two male embryos from the positions of the centroids of shapes drawn around the msSGP cluster at different time points using Image Pro Plus (Media Cybernetics).

Antibody staining was performed on whole mount embryos by standard methods. Detection was by secondary antibodies conjugated to Alexa 488 or 568 fluorochromes (Molecular Probes). The primary antibodies were Eya (mouse 1/50, Developmental Studies Hybridoma Bank, developed by N. Bonnini), Sox100B (gift of Steven Russell), Zfh-1 (1/500 mouse, provided by Z-C Lai), DCAD2 (rat 1/50, Developmental Studies Hybridoma Bank) and cleaved human caspase-3 (Asp715, 1/100, Cell Signalling Technology). For in situ hybridisation an Hmgcr probe was made by transcription of plasmid pNB36C 23 using Roche reagents. Hybridisation was carried out by standard methods and detected using a secondary antibody conjugated to alkaline phosphatase (Roche). Still images were collected using a Provis (Olympus) microscope or a Pascal (Zeiss) confocal.