Migrating cells control morphogenesis of substratum serving as track to promote directional movement of the collective

In order to form organs and other discrete biological structures, groups of cells must undergo directional migration that requires the coordination of cell polarity, cytoskeletal dynamics, and the integration of extrinsic signals within the migrating cohort and with the extracellular environment, which includes other tissues (Lecaudey and Gilmour, 2006; Rørth, 2007). The Drosophila embryonic midgut musculature is formed via the fusion of longitudinal muscle precursors, also known as caudal visceral mesoderm (CVM) and circular muscle precursors, also known as trunk visceral mesoderm (TVM) (Sink, 2007).

The CVM and TVM are derived from exquisitely regulated gene networks that begin with the coordinated migration and spreading of mesoderm cells subsequent to gastrulation and invagination (McMahon et al., 2008). The CVM lineage is specified at the ventroposterior border of mesodermal snail expression, at which a stripe of cells expresses the gene encoding the basic helix-loop-helix transcription factor HLH54F (Georgias et al., 1997; Ismat et al., 2010). Following germband extension (GBE), these cells proliferate and eventually migrate anteriorward in two synchronous cohorts (Fig. 1A-E). The TVM is derived from a subpopulation of mesodermal cells that differentiate in response to segmental expression of bagpipe (bap) (Azpiazu and Frasch, 1993; Azpiazu et al., 1996). Signaling via receptor Anaplastic lymphoma kinase (Alk) within these bap-expressing cells gives rise to a subpopulation of TVM muscle founder cells, which begins as a discrete groups of cells that merge just as the CVM cells make contact at stage 11 (Fig. 1B,F,M; Englund et al., 2003; Lee et al., 2003). Over the course of development, the CVM cells migrate along the TVM as a stream of cells before undergoing cell division at the posterior turn (Fig. 1C,H,P-P″). At the conclusion of stage 13 (Fig. 1D), the CVM cells fuse with fusion-competent myoblasts in the TVM to form the complete midgut musculature (Fig. 1E).

Studies that investigated the effect of ablating FGF signaling, by observing either loss-of-function mutants for the genes encoding the FGF receptor Heartless (Htl) or its ligands Pyramus (Pyr) and Thisbe (Ths), revealed crucial roles for this pathway in ensuring proper directional movement and survival of the CVM (Kadam et al., 2012; Reim et al., 2012). At stage 10, pyr is initially expressed in the gut endoderm before being expressed in the TVM muscle founder cells together with ths at stage 11, whereas the htl is expressed in the CVM. Mutants for either htl or its ligands present significant midline crossing of the CVM and subsequent cell death. Tissue-specific rescue experiments indicated a requirement for these ligands in promoting synchronous, directional CVM migration along the TVM. Surprisingly, ligand expression specifically in the CVM, which normally only expresses FGF receptor, also attenuates the midline crossing defects and supports CVM directional movement. However, expression of a dominant-negative version of Htl or htl RNAi construct in the CVM presents less severe phenotypes than a global loss of function: although asynchronous migration of the CVM groups is observed, there are significantly fewer midline crossing defects. Although the tissue-specific knockdown results may be ascribed to incomplete ablation of htl, when coupled with the observation that forward migration can be restored with CVM-specific expression of FGF ligand, these results suggest that FGF signaling does not play a simple chemoattractive role in guiding the CVM.

In addition, Htl protein localization changes dynamically over the course of development; although initially expressed in all mesodermal cells from invagination at stage 6-7 through subsequent monolayer formation (Sun and Stathopoulos, 2018), expression of Htl protein refines to the hindgut visceral mesoderm (HVM), the CVM, and subsets of the cardiac mesoderm, and is excluded from the TVM by stage 10 (Fig. S1). This brings about the intriguing possibility that FGF not only regulates interdependent interactions between neighboring cell types, but also supports temporally regulated discrete steps in multiple tissues to mediate organ formation.

In order to assess the requirement of htl in multiple tissues, we immunostained htl^AB42 mutant embryos using anti-Fasciclin III (FasIII) antibody to label the TVM muscle founders, and either anti-GFP/anti-RFP (to label CVM-specific HLH54F-gap-Venus or HLH54F-H2A-mCherry reporter constructs crossed into genetic backgrounds, respectively) or anti-Teyrha-Meyrha (Tey) antibodies to label the CVM. FasIII protein expression first becomes apparent at stage 11 in a subpopulation of TVM muscle founder cells that are specified in response to Jeb/Alk signaling (Fig. 1A,B,G,K,I,P,S; Lorén et al., 2003), whereas Tey is expressed in the CVM from stage 10 onward (Inaki et al., 2010; Mandal et al., 2004). Although we observed the same significant CVM midline crossing defects and ipsilateral gaps along each row of TVM cells in htl mutants as previously described, we were surprised to observe contralateral merging of TVM that correlated with CVM midline crossing defects (Fig. 1F,I,L,Q-Q″,T). This TVM merging phenotype was also observed in transverse cross-sections of embryos demonstrating the relative positions of the CVM and TVM (Fig. 1X-X′ compare with 1W,W′). Although previous studies have described TVM morphology defects in FGF mutants, these studies only demonstrated the ipsilateral gaps, which were observed in sagittal sections of embryos (Bradley et al., 2003; Kadam et al., 2012; Reim et al., 2012).

Previously, CVM cells were found to have a reciprocally interdependent relationship with primordial germ cells (PGCs), in that improper PGC migration was observed in the absence of CVM, and CVM migration was delayed and more cohesive in the absence of PGCs (Stepanik et al., 2016). However, combining the germ cell-less (gcl) allele to eliminate the PGCs with htl mutation did not attenuate the CVM midline crossing and TVM contralateral merging defects, demonstrating that these defects are not caused by abnormal PGC migration (Fig. S2).

The observation of contralateral TVM merging defects in htl mutants led us to investigate a possible interdependent relationship between the CVM and TVM beyond the previously described paracrine interaction, with two likely explanations for this phenomenon. The first hypothesis is that the CVM midline crossing defect is a consequence of abnormal TVM morphogenesis, which is likely due to an earlier defect in mesodermal migration observed in htl mutants. Before stage 10, FGF signaling ensures proper mesoderm spreading, which is crucial for the correct positioning of differentiated mesodermal lineages including the TVM (e.g. Beiman et al., 1996). An alternative hypothesis is that contralateral TVM merging is instead driven by the CVM. In this model, abnormal CVM cell migration influences the orientation of the TVM muscle founder clusters, thus leading to the formation of a mesodermal ‘bridge’ between the contralateral tracks.

To determine whether the midline phenotypes are caused by abnormalities in either CVM or TVM, we first scored contralateral TVM merging defects in htl mutant embryos and compared with HLH54F mutants and wild-type controls (Fig. 1F). HLH54F mutants, which completely lack CVM (Ismat et al., 2010), do not exhibit TVM contralateral merging defects or ipsilateral gaps (Fig. 1M,U). However, embryos that are double mutants for both HLH54F and htl present ipsilateral gaps, but had significantly reduced incidence of contralateral TVM merging defects when compared with htl mutants (Fig. 1F,N,V). The ipsilateral gap defects likely relate to an earlier, separate role for FGF signaling that acts in these tissues prior to and distinct from CVM migration that remains intact in the HLH54F mutant (i.e. mesoderm spreading at gastrulation); on the other hand, the contralateral merging phenotypes are mostly dependent on CVM. It is important to note that the limited TVM contralateral merging defects observed in the HLH54F^Δ598;htl^AB42 double mutant embryos occurred at the posteriormost end, at which the contralateral TVM tracks are oriented most closely to each other (Fig. 1L, orange asterisk), as opposed to the more anterior long-distance bridges that occur between TVM segments that are more proximal to the midgut in htl mutants (Fig. 1I,L, orange arrowhead, Q-Q″). Previous studies have demonstrated that the CVM initially migrates along the gut before migrating along the TVM (Reim et al., 2012; Stepanik et al., 2016). Therefore, in htl mutants, these gut-proximal TVM bridges are likely not formed before the initial contact of CVM, but are instead formed as a consequence of abnormal CVM migration along the gut (Fig. 1J,R-R″).

We next examined whether tissue-specific restoration of FGF signaling attenuates the TVM and CVM defects associated with htl mutants. htl^AB42 LOF allele was combined with either the CVM-specific driver G447-GAL4 (Georgias et al., 1997) or the mesoderm-specific driver twi-GAL4 (Carmena et al., 1998; Greig and Akam, 1993), with the latter allowing for expression throughout the trunk mesoderm (including the TVM muscle founders). Although the twi-GAL4 driver supports initial expression in the CVM at low levels, this is not maintained through the onset of migration, mirroring the loss of twi expression in these tissues (Kusch and Reuter, 1999). These drivers were crossed to flies that allow for UAS-mediated expression of either a wild-type or a constitutively active form of htl (Michelson et al., 1998), or a constitutively active form of Ras85D (Gisselbrecht et al., 1996; Lu et al., 1993). Ras85D is a known component of the mitogen-activated protein kinase (MAPK) signaling pathway, and has previously been characterized as a downstream effector of htl (Gisselbrecht et al., 1996); as such, expression of constitutively active Ras85D would mimic activation of MAPK signaling. CVM-specific expression of htl^WT in an htl mutant background attenuates both the CVM midline migration and the contralateral TVM merging phenotype (Fig. 2A-A″). Similarly, both CVM midline migration and TVM merging phenotypes were rescued when either constitutively active Htl or Ras85D were specifically expressed in the CVM (Fig. 2B-C″).

In contrast, TVM-specific rescue with constitutively active Ras85D failed to attenuate the TVM contralateral merging phenotype (Fig. 2E-E″); the merging phenotype is only rescued via the addition of htl^WT, with CVM cells still exhibiting the midline crossing defect (Fig. 2D-D″). This further supports the idea that abnormal CVM migration primarily causes most contralateral TVM merging defects in FGF mutants, and that FGF has an additional role that possibly confers stability to the TVM. In addition, because CVM midline migration defects were rescued via tissue-specific localization of constitutively active Htl (which lacks an extracellular ligand-binding domain) and Ras85D, these data also support the view that forward movement of the CVM cohorts is not entirely dependent on ligands acting as chemoattractive cues from the TVM (Bae et al., 2012; Kadam et al., 2012; Reim et al., 2012).

In addition, we observed a striking correlation between attenuation of the ipsilateral gap phenotype in the TVM and tissue-specific rescue of FGF activity in the CVM: as the CVM cohorts migrate over the TVM, gaps appear to close, forming a more continuous track of TVM muscle founder cells (Fig. S3). Consequently, the incidence of ipsilateral gaps becomes restricted to proximally situated TVM muscle founders upon which the CVM cells have not yet migrated; this is in contrast to htl mutants, in which the CVM cells instead traverse the ipsilateral gaps (Fig. 1O-O″; Kadam et al., 2012).

The correlative relationship between abnormal CVM migration and TVM contralateral merging defects led us to explore potential FGF-dependent signaling pathways that could mediate this interaction. Previous research has demonstrated crucial roles for position specific (PS) integrin signaling in mediating heterotypic cell interactions in an FGF-dependent manner. Integrins are heterodimeric cell adhesion molecules that consist of ɑ and β subunits (Bökel and Brown, 2002). The gene myospheroid (mys) encodes the βPS1 integrin subunit, and is expressed in both the TVM and CVM (Fig. 3A,B,C-C‴,E; Devenport, 2004; MacKrell et al., 1988). In stage 9/10 embryos, FGF signaling is required to support βPS1 protein localization at the interface between mesoderm and ectoderm (McMahon et al., 2010). In addition, FGF was found to regulate basal localization of βPS1 in order to support a mesenchymal-to-epithelial transition (MET) to form a mesoderm monolayer subsequent to GBE (Sun and Stathopoulos, 2018). A role for integrin in mediating proper CVM migration has also been previously demonstrated, as loss of βPS1 results in significantly delayed migration of CVM cells (Urbano et al., 2011).

In stage 11 htl mutant embryos, βPS1 signal was significantly reduced compared with the wild type (Fig. 3D-D‴,F). This relationship extends to a reduction in basally localized βPS1 expression at both the CVM-TVM and trunk mesoderm-ectoderm interfaces. This loss of signal is likely due to a role for FGF in regulating protein expression as opposed to gene expression, as previous studies have demonstrated that mRNA expression of mys and the CVM-specific alpha subunit multiple edematous wings (mew) were unaffected in htl mutants (Bae et al., 2017; Sun and Stathopoulos, 2018). Tissue-specific expression of mys throughout the trunk mesoderm via the twi-GAL4 driver in an htl mutant background attenuates TVM contralateral merging defects, but fails to eliminate ipsilateral gap defects and abnormal CVM migration into the midline (Fig. 2G-G″,H). In contrast, CVM-specific expression of mys using the G447-GAL4 driver failed to attenuate either the contralateral TVM merging or CVM midline migration defects, and ipsilateral gaps were still observed along the TVM (Fig. 2F-F″,H). These results suggest that the role for FGF in regulating integrin expression is required earlier during TVM development, before CVM migration. When a wild-type copy of htl is expressed in the trunk mesoderm via the twi-GAL4 driver in an htl mutant background, we observed restoration of basally polarized βPS1 (Fig. 3G-G‴,H). Constitutively active htl partially restored βPS1 protein expression but failed to restore polarized expression of βPS1 (Fig. 3I,J-J‴), which is consistent with the view that directional signaling by FGF ligands to the Htl FGF receptor is required to support polarized localization of βPS1 (Sun and Stathopoulos, 2018). These insights were verified by quantifying the mean fluorescence intensities of immunofluorescence stainings against βPS1 (Fig. 3L). In HLH54F mutants, some βPS1 signal was observed at what should have been the interface between the CVM and TVM (Fig. 3K-K′,M-M′) and likely relates to new association of gut with TVM (Fig. S4B). Normally, both tissues, CVM and TVM, contribute to integrin adhesion and signaling as a significant reduction of βPS1 fluorescence intensity in HLH54F mutants, and a significantly higher signal in embryos expressing htl^CA via twi-GAL4 in a htl mutant background, likely due to dysregulated integrin expression (Fig. 3L, compare with twi>htl^WT rescue).

These results suggest that restoring βPS1 expression in the TVM before the onset of CVM migration confers a structural stabilizing role in ensuring that the TVM muscle founders are refractory to morphogenetic influence by aberrantly migrating CVM cells in htl mutants, representing an earlier, essential role for Htl well before CVM migration. This earlier role is demonstrated by restoring either wild-type htl or βPS1 expression specifically in the TVM before CVM differentiation, as aberrantly migrating CVM cells become unable to redirect the TVM substrate. When FGF signaling is restored specifically in the CVM, the migrating cohorts are able to rescue TVM defects to a significant degree, including attenuation of ipsilateral gap defects between TVM muscle founder clusters; however, restoring βPS1 expression specifically in the CVM fails to rescue TVM merging defects, suggesting that an FGF signaling component other than integrin is required to support directional migration of the CVM (Fig. 4K,L).

Although our observations have demonstrated the ability of the CVM to influence their TVM substrate in a sensitized htl mutant background, we wanted to determine whether this interaction is relevant to the wild-type situation. Larvae of HLH54F mutants present midgut constriction defects, with a subset lacking constrictions altogether (Ismat et al., 2010). To investigate whether these later defects are potentially due to a loss of morphogenetic influence by the CVM on the TVM, we compared late stage 12 wild-type and HLH54F mutants immunostained against FasIII and found that embryos lacking CVM present abnormal TVM tracks (Fig. S5). Compared with the relatively linear TVM observed in the wild-type TVM, HLH54F-mutant TVM often presents a convoluted, looping morphology (Fig. S5A-A′,C-C′,B).

Furthermore, transverse cross-sections of stage 11 wild-type embryos reveal more columnar, epithelial-like cell shapes in the TVM that appear to be more uniformly arranged when compared with HLH54F mutants (Fig. 4A,F; Fig. S6). Quantification of the long axis orientations of the TVM cells reveals trends that differ in the presence and absence of CVM, suggesting that TVM cell orientation is dynamically influenced by migrating CVM cells (Fig. 4B,G). To gain further insight, we immunostained embryos against phosphorylated focal adhesion kinase (pFAK), which is a component of the complex tethering the actin cytoskeletal network to integrin (Delon and Brown, 2007; Fox et al., 1999). Although FAK is ubiquitously expressed, wild-type embryos present an accumulation of pFAK in both the CVM and TVM at their interface, whereas HLH54F mutants present a more random distribution of pFAK signal (Fig. 4C,D,H,I). This further supports the idea that CVM migration has effects on cytoskeletal components of their TVM substrate, potentially influencing their polarity and morphogenesis (Fig. 4E,J).

Our study, which describes how FGF promotes both directional migration of CVM cells and supports integrin expression in the TVM, highlights the ability of collectively migrating cells not only to respond to extrinsic molecular cues and surrounding topology, but also to physically modify the morphogenesis of their tissue substrate.

All fly stocks were kept at 25°C in standard medium. The following stocks were used in this study: htl^AB42/TM3,ftz-lacZ [Bloomington Drosophila Stock Center (BDSC) #5370]; HLH54F^Δ598/CyO (gift from M. Frasch; Ismat et al., 2010); Df(2R)Exel7150/CyO (BDSC #7891); gcl^Δ/CyO (Robertson et al., 1999); G447-GAL4 (Georgias et al., 1997); twi-GAL4 (BDSC #58804; Carmena et al., 1998; Greig and Akam, 1993); UAS-htl^WT, UAS-htl^CA and UAS-Ras1a* (gifts from A. Michelson; Carmena et al., 1998; Gisselbrecht et al., 1996). The CVM reporters HLH54F-gap-Venus and HLH54F-H2A-mCherry have been described previously (Kadam et al., 2012; Stepanik et al., 2016) and were combined with htl^AB42 mutant and GAL4/UAS lines via standard genetic crosses. Wild type refers to yw/+; HLH54F-gap-Venus/+.

Embryos were collected, fixed and stained using previously described methods (Frasch, 1995; Jiang et al., 1991). The primary antibodies used in this study were goat anti-GFP (1:5000; Rockland, 600-103-215), rabbit anti-RFP (1:1000; MBL International, PM005), mouse anti-FasIII [1:200; Developmental Studies Hybridoma Bank (DSHB), 7G10], mouse anti-βPS1 (1:40; DSHB, CF.6G11), guinea pig anti-Htl (1:1000; Sun and Stathopoulos, 2018), rat anti-Twi (1:200; Ozdemir et al., 2011), rabbit anti-Zfh1 (1:2000; gift from R. Lehmann; Broihier et al., 1998), guinea pig anti-Hb9 (1:1000; gift from J. Skeath; Broihier and Skeath, 2002), rabbit anti-phospho-FAK (1:1000; Thermo Fisher Scientific, 44-624G) and rabbit anti-Tey (1:500; this study). Anti-Wg antibody (1:10; DSHB, 4D4) was used to determine embryo stage to compare wild-type and HLH54F mutant embryos.

Chromogenic staining was performed using the Vectastain Elite ABC system in conjunction with DAB substrate for signal development (Vector Laboratories). Immunofluorescence staining was performed using Alexa Fluor 488, 555 and 647 secondary antibodies (Molecular Probes). Embryos were mounted in Permount (Fisher Scientific) for whole-mount studies of chromogenically stained embryos or in 70% glycerol in 1× PBS buffer for whole-mount or transverse cross-section studies of immunofluorescence-stained embryos. To obtain transverse cross-sections of chromogenically stained embryos, an acetone-araldite solution (Electron Microscopy Sciences) was used as a mounting medium and allowed to harden overnight in mold at 75°C. Using a microtome (LKB Bromma 2218 Historange), 8 μm plastic sections were obtained and mounted in 1:1 acetone-araldite solution. Fluorescent images were obtained using an LSM 800 confocal microscope (Carl Zeiss).

Transverse cross-sections of immunofluorescence-stained embryos were obtained and imaged under standardized conditions using an LSM 800 confocal microscope. Imaging was performed using Zen 2.3 software, with at least four replicates and 10 stacks per sample. Raw .czi files were processed using ImageJ software, with z-projections generated using the ‘Sum Slices’ option. Regions of interest (ROIs) encompassing the CVM–Trunk Mesoderm interface were obtained using the CVM signal as a guide, and mean fluorescence intensity values within the ROI were obtained via the Analyze→Measure→Mean Value function. The same ROI was then shifted towards a region representative of the Trunk Mesoderm–Ectoderm interface, and corresponding mean fluorescence intensity values were obtained and tabulated. Background signal/noise measurements were subsequently measured, and the resulting fluorescence intensity values were subtracted from both interface values to obtain corrected mean values for each region. One-way ANOVA was applied to assess statistical significance.

Immunostained embryos were embedded, cross-sectioned and imaged. A line was drawn connecting two points at the yolk-mesoderm interface flanking the TVM, and individual cells were outlined in order to determine the long axis. Long axis orientation angles of individual cells were determined relative to the line drawn through the TVM using ImageJ, and rose diagrams of binned cell axis measurements were generated using MATLAB scripts.

The anti-Tey antibody was made by expressing T438-N667 with a C-terminal 6×His tag in a modified pET vector (Novagen) in BL21 (DE3) bacteria with the placI-RARE-2 plasmid (Novagen) and protein purified using standard conditions. Purified protein was used to immunize rabbits (Pocono Rabbit Farms and Lab). Anti-Tey antibodies were affinity purified from the resulting serum by conjugating the Tey protein fragment above to beads following the manufacturer's protocol (Aminolink coupling resin, Thermo Fisher Scientific). The specificity of the anti-Tey antibody was confirmed by immunostaining HLH54F mutant embryos lacking CVM.

We thank M. Frasch, R. Lehmann, A. Michelson and J. Skeath for sharing fly stocks and antibodies, and are grateful to V. Stepanik, Z. Ákos, J. Sun, H. L. Curtis and J. Irizarry for assistance and helpful discussions.