Competition and determination of leading and trailing cells during collective cell migration is a widespread phenomenon in development, wound healing and tumour invasion. Here, we analyse this issue during in vivo ganglionic branch cell migration in the Drosophila tracheal system. We identify Sequoia (Seq) as a negative transcriptional regulator of Branchless (Bnl), a Drosophila FGF homologue, and observe that modulation of Bnl levels determines how many cells will lead this migrating cluster, regardless of Notch lateral inhibition. Our results show that becoming a tip cell does not prevent others in the branch taking the same position, suggesting that leader choice does not depend only on sensing relative amounts of FGF receptor activity.
The Drosophila tracheal system develops through cell migration and cell rearrangements to form a complex network of interconnected epithelial tubes that allow gas exchange. The Drosophila FGF, Branchless (Bnl), binds to and activates the Drosophila FGF receptor, Breathless (Btl), which is expressed in all tracheal cells, and triggers cell migration (Sutherland et al., 1996). Relatively little is known about how this system is finely tuned at the genetic level and how bnl expression is regulated in order to achieve the three-dimensional branching pattern of the larval tracheal system. In a forward genetic screen to isolate new genes involved in tracheal development, we identified new alleles of sequoia (seq), which encodes a nuclear protein with two putative zinc-finger domains (Brenman et al., 2001). Previously, seq has been reported to have a role in dendrite development (Gao et al., 1999; Brenman et al., 2001), axonal targeting of photoreceptor cells (Petrovic and Hummel, 2008) and external sensory organ formation (Andrews et al., 2009). Here, we report that Seq is a negative regulator of bnl expression and analyse its role in modulating the number of tip cells in the ganglionic branches (GBs) of Drosophila trachea.
Results and Discussion
sequoia (seq) mutants show bifurcations at the tip of the ganglionic branches
In wild-type embryos, GBs form at the ventral side and grow towards the central nervous system (CNS) (Fig. 1A) (Manning and Krasnow, 1993). The cell at the tip of each GB (the GB tip cell) behaves as a leading cell and later differentiates as a terminal cell (Englund et al., 1999). Thus, in GBs, tip-cell behaviour is associated with terminal cell development, in contrast with dorsal branches (DBs), that possess two tip cells that differentiate into a terminal cell and a fusion cell (Samakovlis et al., 1996a; Samakovlis et al., 1996b). The process of terminal cell differentiation requires the activity of the gene DSRF (officially known as blistered), the Drosophila homologue of the serum response factor (SRF) (Guillemin et al., 1996; Gervais and Casanova, 2010).
In an ethyl methanesulfonate (EMS) mutagenesis screen, we identified three mutants that displayed branch bifurcations at the tip of the GBs (Fig. 1B,C and data not shown). In these mutants, we observed lumenal bifurcations in 4–7 GBs per embryo at stage 15 (mean 4.5, n=25 embryos) and in 7–12 GBs per embryo at stage 16 (mean 8.2, n=46 embryos). These cells adopted a terminal fate, as judged by DSRF expression, and developed their own terminal lumen, causing the observed luminal bifurcations (Fig. 1D–G). The three identified mutants belonged to a single complementation group and failed to complement the previously reported sequoia alleles seqU5 and seq22. These mutant alleles, and all trans-heterozygous combinations, also displayed the same branch bifurcation phenotype (Fig. 1C and data not shown). Thus, seq is required for the proper morphogenesis of GBs in the tracheal system.
Seq regulates GB patterning non-autonomously
seq is expressed exclusively in the developing and mature nervous system (Brenman et al., 2001). Accordingly, we did not detect expression of seq in the tracheal cells (Fig. 1K,L). As such, seq expression in the developing CNS might be responsible for proper patterning of the GBs. To test this hypothesis, we expressed wild-type seq in the ectoderm of seq mutants, using a 69BGAL4 driver, and achieved full rescue of the GB bifurcation phenotype (n=37). In addition, we drove the expression of wild-type seq in midline precursors and glia of seq mutants, using a simGAL4 driver and accomplished ~50% rescue (n=51), with those embryos that were not fully rescued showing a considerable attenuation of the phenotype. These results suggest a non-autonomous effect of seq in GB patterning mediated by its expression in cells of the CNS. Correspondingly, overexpression of seq in the CNS of wild-type embryos resulted in the abrogation of GB branching and DSRF expression in the tracheal system, owing to a reduction in bnl expression in the CNS (Fig. 2E and data not shown). Taken together, these results suggest a non-autonomous effect of seq in GB patterning by its expression in cells of the CNS.
seq acts an upstream regulator of bnl expression
Btl activation in tracheal cells, by its ligand Bnl, is thought to stimulate and guide tracheal migration and to induce DSRF expression at the tip cells of the main branches, probably through the transcription factor Pointed (Pnt) (Samakovlis et al., 1996a; Sutherland et al., 1996). To examine the relationships between seq and bnl we analysed the induction of extra terminal branches upon ectopic expression of bnl. In the CNS, bnl expression in the midline, using simGAL4, phenocopied the seq mutant (Fig. 3F), whereas higher levels, induced by inscGAL4 in the whole CNS, caused a more extreme phenotype with many cells adopting a terminal fate (Fig. 2F; Fig. 3D). Importantly, simultaneous overexpression of both seq and bnl in the CNS had the same effect as just bnl overexpression (Fig. 3E), clearly indicating that forced bnl expression overcomes the repressor effect of seq on GB terminal branch formation. This result leads us to propose that, in wild-type embryonic development, Seq transcriptionally regulates bnl (Fig. 3E). In this scenario, seq, by fine-tuning bnl expression, could restrict sprouting of terminal branching to a single cell per GB. Accordingly, we found that removal of one copy of bnl or one copy of pnt rescues the GB bifurcation phenotype of loss of seq conditions (Fig. 2B,D).
Throughout development, bnl is expressed very dynamically and at low levels around the clusters of tracheal cells at positions in which new branches will form and grow (Sutherland et al., 1996) (Fig. 2I). In addition, owing to the extreme effects observed when bnl is overexpressed in the CNS (Fig. 2F; Fig. 3D), we expect mild variations in bnl expression to account for the differences in GB patterning between seq and wild-type embryos. Accordingly, we could only detect mild differences in bnl expression, as determined by in situ hybridization, in seq mutants (Fig. 2, compare L with K). In order to more clearly assess the role of Seq as a putative repressor of bnl expression, we analysed the levels of bnl transcripts by quantitative real-time PCR (qRT-PCR) in whole embryos. In seq mutants, we detected a 2.4-fold increase in bnl expression. Consistently, we also observed a 50% reduction upon ectopic expression of seq using nulloGAL4, a driver active from blastoderm stages (Lécuyer et al., 2007). As a positive control, we detected a 7.2-fold increase of bnl mRNA levels upon ectopic expression of bnl using the same GAL4 line (supplementary material Table S1).
To confirm the role of seq as a bnl repressor we performed in situ hybridisation and phenotypical analysis after seq overexpression at the onset of tracheal development. At early developmental stages, individual branches outgrow towards nearby specific groups of cells expressing bnl; in bnl mutants, such branches fail to form (Sutherland et al., 1996). We induced ectopic seq expression at early developmental stages and found the same tracheal phenotype as in bnl mutants (Fig. 2G). Accordingly, in such embryos we failed to detect bnl expression around the tracheal cells (Fig. 2J), indicating that seq can act as a bnl repressor. Thus, Seq regulates bnl expression, directly or indirectly, acting in the tracheal system in a non-autonomous manner.
More tracheal cells adopt tip-cell behaviour in seq mutants
The observation that the additional terminal cells in seq mutants are closer to the CNS than their non-terminal counterparts in the wild-type, suggests that these cells migrate over a longer distance than in wild-type embryos. Indeed, these extra DSRF-positive cells are not placed behind the leading terminal cell but, from early stages, migrated in parallel to it (Fig. 1M–O; supplementary material Movies 2, 4). This was also observed upon overexpression of bnl in the CNS (Fig. 3B–D,F). From very early stages, the induced extra terminal cells migrated side by side in the direction of the CNS, shifting their positions from trailing to leading (Fig. 3H). bnl signalling was active in these extra tip cells, as judged by the concomitant DSRF expression.
In vivo time-lapse analysis of seq mutants confirmed these results. As the GB cluster migrated towards the CNS, more than one cell could be seen moving to a forward tip-cell position (Fig. 4B; supplementary material Movies 3, 4). Cells that were previously trailing could be seen moving to a tip-cell position as they started expressing DSRF (Fig. 4A; supplementary material Movies 1, 2). These results suggest that GB tip cell induction is due to higher non-autonomous levels of Bnl and concomitant FGF pathway activation in tracheal cells. In seq embryos, upregulation of bnl expression increases the number of leading cells during GB migration. This seems to be a finely tuned mechanism because small variations in bnl expression, the result of seq loss-of-function mutations, are sufficient to increase the number of cells at the tip of the branches.
Although this fine-tuning mechanism works by adjusting the levels of the FGF, another regulatory mechanism, mediated by sprouty (sty), has been reported to act as a negative-feedback regulator of the FGF pathway itself. Sty is a known regulator of FGF signalling, is expressed at the tips of tracheal branches and is induced by high levels of Bnl (Hacohen et al., 1998). Loss of sty function results in ectopic branching of the stalk cells at the tip of tracheal branches, as in the seq mutant GBs. We tested whether there was any interaction between seq and sty, by means of expression and epistatic analysis. As expected, the sty domain of expression was expanded in both seq mutant embryos and when bnl is overexpressed in the CNS (supplementary material Fig. S1C–F). However, whereas higher levels of Bnl induced higher levels of sty expression, these higher levels of sty repression were not enough to overcome the inductive effects of the higher levels of FGF signalling. In addition, double mutants for seq and sty showed increased numbers of GB tip cells compared with upon seq and sty mutation alone (supplementary material Fig. S1A,B). These results indicate that absence of Sty allows for more GB cells to become tip cells when subjected to the same levels of Bnl, as expected from the function of Sty as a negative feedback of the FGF pathway.
Notch-dependent lateral inhibition is not active in modulating tip-cell numbers in the GBs
It has been previously shown that in each tracheal DB, the cell with a highest level of Btl receptor activity becomes the tip cell. For this to happen, it has been proposed that the abovementioned tip cell generates a Notch-dependent lateral inhibitory signal, implementing a follower cell fate upon its neighbours. As a result, there can only be one tip cell per branch, whereas the other cells become followers (Ghabrial and Krasnow, 2006). We examined whether a similar mechanism could operate in the GBs but, in contrast to what has been reported for the DBs, tracheal expression of constitutively active Notch prevented tracheal fusion, as was previously reported (Ikeya and Hayashi, 1999; Llimargas, 1999; Steneberg et al., 1999), but did not arrest outgrowth or stall cells near the base the GB (Fig. 3I). Furthermore, when this activated form of Notch was expressed in the tracheal cells of seq mutant embryos, this was not sufficient to rescue the GB bifurcation phenotype (Fig. 3J,K). Activation of Notch in tracheal cells does not inhibit migration of GB cells and is not able to overcome the effects of higher FGF in the GB surrounding tissues of seq embryos. Thus, whereas a Notch inhibitory mechanism could operate in the DB cells, it is not at work in all tracheal migratory clusters and does not act to select single tip cells in the GBs.
Taken together, although high levels of FGF can induce the terminal cell fate in all tracheal cells (Sutherland et al., 1996), small variations in FGF levels can also establish how many cells in the migrating cluster behave as leading cells irrespective of Notch inhibition. This argues for different mechanism being responsible for the branching morphogenesis of GBs and DBs. In the case of the DBs, it has been proposed that tip cell choice requires Notch-driven selection of a leader cell (Ghabrial and Krasnow, 2006). Here, we report that GB tip-cell selection is not dependent on a Notch lateral inhibitory mechanism. DB migration is in many ways very different from GB migration. For most of their migration, DBs maintain two leading cells, the one that will become the terminal cell and the one that will take the fusion cell fate (Samakovlis et al., 1996a; Ribeiro et al., 2002; Caussinus et al., 2008). Notch lateral inhibition plays a crucial role in singling out the fusion cell (Ikeya and Hayashi, 1999; Llimargas, 1999). Thus, Notch-mediated effects in DB migration might be associated with this fate choice rather than with tip-cell selection. Fusion cells are not present at the tip of GBs, and therefore their migration is not affected autonomously by Notch. In addition, Notch signalling has a non-autonomous effect in tracheal development by negatively regulating bnl expression (Ikeya and Hayashi, 1999), which might mask its real autonomous effects in DB migration. In this scenario, we propose that the Notch-independent mechanism, observed here for GB migration, is likely to be the norm for most tracheal clusters and other migratory cell groups where fate choices are not an issue.
In conclusion, FGF signals received by tracheal cells, associated with FGF receptor activation, can induce more than one tip cell, irrespective of the migratory behaviour of the neighbouring tracheal cells, as has been previously observed (Reichman-Fried et al., 1994; Lee et al., 1996; Sutherland et al., 1996). Our results show that one cell becoming the tip cell does not inhibit others in the migrating cluster taking up the same position. This suggests that the distinction between leading and trailing cells could depend not only on a competition mechanism sensing the relative amounts of FGFR activity, but also on a level of FRGR activity above a critical threshold, induced by a variation in levels of Bnl.
Materials and Methods
Drosophila stocks and genetics
The following stocks are as described in FlyBase (http://flybase.bio.indiana.edu): UAS-srcGFP, UAS-tauGFP, UAS-seq simGAL4, inscGAL4, elavGAL4 and scaGAL4, nulloGAL4 shg2, shgE17B, bnlP1 (also known as bnl00857) and pntΔ88. sequoiaU5 has been described previously (Andrews et al., 2009). We used UAS-bnl (from Mark Krasnow, Stanford University, Palo Alto, CA) btlRFPmoe (from Markus Affolter, University of Basel, Basel, Switzerland), UAS-Ni (from Gary Struhl, Columbia University, New York, NY) and nulloGAL4 (from Walter Gehring, University of Basel, Basel, Switzerland). We used the GAL4 system (Brand and Perrimon, 1993) for over- or mis-expression experiments. We used btlGAL4 (from Markus Affolter) as a pan-tracheal driver and DSRFGAL4 (from Mark Krasnow) as a specific terminal cell driver. To ectopically express bnl and seq, we used simGAL4, inscGAL4, nulloGAL4 and the double driver stock: elavGAL4; scaGAL4.
Immunohistochemistry and image acquisition
We used antibodies that recognised Sequoia (Brenman et al., 2001), DSRF (CSHL), GFP (Molecular Probes and Roche), β-galactosidase (Cappel and Promega), mAb2A12 (DSHB) and conjugated secondary antibodies (Jackson ImmunoResearch). Chitin was visualised with both CBP (NEB) and WGA (Molecular Probes). In situ hybridization was performed according to standard protocols. Images were acquired and processed according to (Araujo et al., 2007).
Real-time quantitative PCR
qRT-PCR was performed with a Roche LightCycler 480 using the SYBR Green I method according to the manufacturer's instructions. The two sets of primers for bnl and for the rp49 and alpha-Tubulin-84B controls were designed over exon–exon junctions to avoid genomic amplification, using the Primer Express software (Applied Biosystems). To assess levels of expression we extracted total RNA from embryos with TRIzol reagent (Invitrogen) according to the manufacturer's instructions. sequoia homozygous mutant embryos were separated by means of a GFP balancer. We prepared cDNAs with the RevertAid H Minus First Strand cDNA Synthesis Kit (Fermentas) using random hexamer primers. The nature of the PCR products was confirmed by melting curve analysis. All analyses were performed using the Relative Expression Software Tool (REST) according to the user manual and references. The main steps of the automatic REST workflow are as follows: PCR efficiencies were calculated for every pair of primers by generating standard curves at increasing dilutions of cDNA (1:1, 1:5, 1:25, 1:125, 1:625), and used to correct raw data. rp49 and alpha-tubulin-84B were assumed to be equally expressed in wild-type and mutant embryos and were used as a reference to normalize data. A ratio between the normalized signals of tested genes in mutant and wild-type embryos was calculated and expressed as a fold change, and statistically tested by a bootstrap test (10,000 randomizations). We used a sample size of six repeats per plate, for each of the tested embryonic genotypes, using two different sets of primers for bnl, and each experiment was repeated independently three times.
We are grateful to M. Llimargas and E. Martin-Blanco for critically reading the manuscript. We thank M. Affolter, H. Bellen, M. Krasnow and Y. N. Jan and the Bloomington stock center for fly stocks and reagents. We thank L. Bardia for assistance and advice with confocal microscopy; N. Molist for technical assistance; M. Grillo for help and advice with qRT-PCR and J. Photopolous for help with mapping the sequoia mutation. S.J.A. is a Ramon y Cajal Researcher previously supported by an I3P postdoctoral contract from the Consejo Superior de Investigaciones Cientificas (CSIC) and a Beatriu de Pinós fellowship (AGAUR). This work is supported by grants from the Generalitat de Catalunya and the Spanish Ministerio de Ciencia e Innovación (BFU2009-07629 and Consolider CSD2007-00008).