BMP-gated cell cycle progression drives anoikis during mesenchymal collective migration

SUMMARY

Tissue homeostasis involves the elimination of abnormal cells to avoid compromised patterning and function. While quality control through cell competition is well-studied in epithelial tissues, it is unknown if and how homeostasis is regulated in mesenchymal collectives. Here we demonstrate that collectively migrating Drosophila muscle precursors utilize both Fibroblast growth factor (FGF) and Bone morphogenetic protein (BMP) signaling to promote homeostasis via anoikis, a form of cell death in response to substrate de-adhesion. Cell cycle-regulated expression of cell death gene head involution defective is responsible for caudal visceral mesoderm (CVM) anoikis. Secreted BMP ligand drives cell cycle progression via a visceral mesoderm-specific cdc25/string enhancer to synchronize collective proliferation, as well as apoptosis of cells that have lost access to substrate-derived FGF. Perturbation of BMP-dependent cell cycle progression is sufficient to confer anoikis resistance to mismigrating cells, facilitating invasion of other tissues. This BMP-gated cell cycle checkpoint defines a quality control mechanism during mesenchymal collective migration.

Graphical Abstract

eTOC Blurb

Quality control during collective cell migration is essential for proper organ assembly. Macabenta et al. show that Drosophila embryonic muscle precursors utilize secreted BMP ligand to coordinate mitosis along with concomitant expression of the cell death gene Hid, which eliminates cells that lose access to substrate-derived FGF ligand during migration.

INTRODUCTION

The coordinated directional movement of groups of cells, which is known as collective cell migration, is primarily responsible for the highly stereotyped assembly of specific organs and involves an incredible diversity of strategies and heterotypic interactions depending on the organ or tissue type (Macabenta and Stathopoulos, 2019a; Scarpa and Mayor, 2016). While originally used to describe only epithelial or epithelial-like groups of cells that have stable adhesions, the definition of collective cell migration has since expanded to accommodate more transient, dynamic adhesive contacts within groups of mesenchymal cells (Theveneau and Mayor, 2012, 2013). Maintaining homeostasis via the elimination of abnormal or hyperplastic cells in migrating collectives is essential for normal development, as dysregulated collective migration is a hallmark of metastatic cancer and chronic inflammation. While cell competition has been identified as an evolutionarily-conserved strategy for eliminating abnormal cells in epithelial tissues, it is unknown if, and how, mesenchymal collectives exert quality control via the removal of abnormal or mismigrating cells.

In this study, we use the Drosophila embryonic caudal visceral mesoderm (CVM) as a model for understanding how a collectively-migrating cohort of mesenchymal cells maintains integrity via removal of lost cells prior to organ assembly. CVM migration is a multistep process that involves the synchronous bilateral migration of cell cohorts along the trunk visceral mesoderm (TVM), which serves as a substrate track (Figure 1A). Additionally, CVM cells form heterotypic interactions with co-migrating primordial germ cells (PGCs) (Stepanik et al., 2016), which is similar to the chase-and-run mechanism described in collectively-migrating Xenopus cranial neural crest and placodal cells (Theveneau et al., 2013). Our lab and others have previously characterized an essential role for FGF signaling in supporting CVM cell migration and survival (Kadam et al., 2012; Reim et al., 2012). The Drosophila FGFs Pyramus (Pyr) and Thisbe (Ths) are expressed in the TVM, where they interact with the FGF receptor Heartless (Htl) expressed by the migrating CVM cohorts to promote proper pathfinding and integrin-dependent adhesion (Kadam et al., 2012; Macabenta and Stathopoulos, 2019b; Sun and Stathopoulos, 2018). Furthermore, unlike other examples of FGF-dependent collective cell migration in Drosophila [e.g. mesoderm spreading during gastrulation (McMahon et al., 2010; Sun and Stathopoulos, 2018)], CVM cells undergo apoptosis when FGF signaling is ablated. In wildtype embryos, posterior CVM cells that have lost contact with the TVM undergo apoptosis (Kadam et al., 2012; Reim et al., 2012), while mutants for biniou (bin) and bagpipe (bap) result in more widespread CVM cell death due to complete loss of the TVM (Reim et al., 2012). This makes the CVM an ideal system for studying anoikis, which is a form of apoptosis induced by the detachment of cells from the extracellular matrix, therefore preventing adhesion-independent growth. Acquiring anoikis resistance is a prerequisite to cancer progression and metastatic colonization of other tissues (Paoli et al., 2013). Furthermore, we have uncovered the unprecedented ability of CVM cells to influence the morphogenesis of their tissue substrate (Macabenta and Stathopoulos, 2019b). During normal development, migrating CVM cells appear to reorient cells in the TVM as they pass over, such that a complete absence of CVM results in abnormal TVM morphology. In FGF mutants, an even more dramatic phenotype is observed: CVM cells mismigrate and alter the directional growth of the TVM, forming contralateral ‘bridges’ (Macabenta and Stathopoulos, 2019b). There is clearly a significant morphogenetic consequence to allowing mismigrating CVM cells to survive; therefore, one can extrapolate that a mechanism for inducing apoptosis in cells that have lost access to FGF signaling (such as cells that have moved off-track) is critical to establishing normal midgut musculature.

Figure 1. CVM cell death due to loss of FGF signaling is temporally regulated and requires hid.

(A) Schematic of CVM cells (red) migrating bilaterally along the TVM track (cyan). (B-B’) WT embryos expressing HLH54F-Gap-Venus (HGV) reporter immunostained with GFP antibody to mark CVM cell membranes (green), Tey antibody (red) to mark CVM nuclei, and FasIII antibody to mark the TVM (cyan). Loss of Tey nuclear stain was observed in GFP-positive regions that undergo apoptosis (yellow arrowhead). (C-F) Stage 13 embryos immunostained with Tey antibody (red) and Fas-III antibody (cyan). Tey-positive nuclei for each genotype were counted and quantified as a metric for cell survival by stage 13 (G, n=10 per genotype). (H) Schematic demonstrating relative positions of CVM (red), TVM (cyan), and central nervous system (CNS, dark blue) in a stage 13 embryo. (I, J) Stage 13 embryos immunostained with Tey antibody to visualize CVM and ventral muscles (red), GFP to visualize the CVM-specific reporter (green), and BP102 to visualize the CNS (dark blue). In htl hid double mutants, CVM cells occasionally invade the CNS (J). (K) FISH reveals temporal activation of hid expression (green) in the CVM marked with anti-Tey (purple). (L) Temporal dynamics of cell division over the course of CVM migration. Embryos expressing the HLH54F-Gap-Venus (HGV) reporter were immunostained with GFP antibody to visualize cell membranes (red) and PH3 antibody to visualize actively dividing cells (white). (M-O’) In stg mutant embryos, hid expression is highly delayed (N, N’ O, O’), compared to a stg heterozygote (M, M’). Insets show magnified views. p<0.05, Scale bars = 20μm. See also Figure S1.

RESULTS

Temporally regulated cell death in CVM cells requires hid

To investigate programmed cell death over the course of CVM migration, we immunostained stage 11-13 embryos with antibodies to Teyrha-meyrha (Tey) to visualize the CVM and Fasciclin III (FasIII) to visualize the TVM (Figure 1A,B,B’). Tey nuclear staining is completely lost in cells that undergo apoptosis, which is apparent when used in conjunction with the CVM-specific HLH54F-Gap-Venus (HGV) reporter to visualize the cytoplasm and cell membranes (Stepanik et al., 2016); Tey signal is absent in cellular debris and blebs visualized by HGV, presumably due to nuclear envelope breakdown in dying cells (Figure 1B, arrowheads). Consistent with previously reported findings (Kadam et al., 2012; Mandal et al., 2004), CVM cells in htl mutant embryos are almost completely lost to apoptosis by stage 13 as indicated by visualization of terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL)-positive cells and a near-total absence of Tey staining (Figures S1A,B and 1C,D,G). However, apoptosis only occurs during the later stages of migration; CVM cells in htl mutants remain alive through stages 10-12, albeit with highly abnormal migration (Kadam et al., 2012; Reim et al., 2012). Apoptosis is also observed in WT embryos, with mismigrating cells and posterior CVM cells that have lost contact with the TVM going off-track (Movie S1). We therefore wanted to investigate how this temporally-restricted cell death is achieved in migratory CVM.

The gene head involution defective (hid), which is an ortholog of mammalian Smac/DIABLO (Verhagen and Vaux, 2002), is well-known for its function in promoting programmed cell death (Grether et al., 1995; Verhagen and Vaux, 2002). Hid protein targets the cell survival factor Drosophila Inhibitor of Apoptosis 1 (DIAP1) for proteasomal degradation, initiating the downstream signaling cascade that culminates in Caspase-3 cleavage and subsequent apoptosis (Goyal et al., 2000; Holley et al., 2002; Ryoo et al., 2002; Wang et al., 1999; Yoo et al., 2002). Expression of hid was observed in the CVM and verified via fluorescence in situ hybridization (FISH) with hid riboprobes (Figure 1K). To test whether CVM cell apoptosis relates to hid expression, double mutants of htl and hid were generated and CVM cell survival was quantified with anti-Tey staining (Figure 1E,G). We found that hid loss of function attenuated apoptosis of CVM observed in htl single mutants, suggesting that hid is essential for the cell death observed in later stages (Figure 1E, compare with C,D). Intriguingly, this rescue of cell survival also resulted in “undead” htl hid mutant CVM cells occasionally invading other tissues, including the embryonic central nervous system (CNS) (Figure 1H-J), demonstrating that Hid initiates anoikis in aberrantly migrating CVM cells. Consistent with previously published literature, the cell death phenotype was also attenuated by expressing constitutively active Ras (Ras1a) in the CVM, which works downstream of FGF and is known to antagonize hid [Figure S1C-E’; (Bergmann et al., 1998; Gisselbrecht et al., 1996; Kurada and White, 1998; Vincent et al., 1998)].

hid expression in CVM cells is restricted to the later stages of migration, starting at stage 12 (Figure 1K), which coincides with CVM cell division [as detected by phosphorylated histone H3 (PH3) immunostaining to label mitotic cells] at the posterior turn before rearranging into discrete linear cohorts along the TVM prior to myoblast fusion (Figure 1M; Shibata et al., 1990; Su et al., 1998). We investigated whether there is a correlation between timed CVM cell division and the onset of hid expression by first looking at mutants for string (stg), which encodes a Cdc25 phosphatase that promotes expression of the mitotic kinase Cdk1 (Cdc2; Edgar and O’Farrell, 1990). Expression of stg is therefore essential for the G2-M transition, such that loss of function mutants have fewer cells due to delayed mitosis. If hid expression in the CVM is regulated by the cell cycle, stg mutants would likely present a delay or absence of hid mRNA expression. Indeed, FISH assays indicate a highly reduced and delayed onset of hid expression in stg single mutants, with detectable expression starting only at stage 13, well after the CVM have completed their migration (Figure 1N,N’,O,O’, compare with M,M’). Furthermore, double mutants for htl and stg showed that a loss of stg function largely attenuated apoptosis caused by a loss of FGF signaling (Figure 1F,G). By counting the numbers of CVM nuclei at stage 11 prior to cell division, we found that the numbers of CVM cells in WT, htl, and htl hid mutants was not significantly different, with only htl stg mutants presenting a reduced number of cells (Figure S1F-H, K). This suggests that temporal activation of hid in the CVM is regulated by the cell cycle and correlated with mitosis. Additionally, we examined the effect of the E2f1 transcription factor on hid expression dynamics in the CVM. E2f1 plays a central role in cell cycle regulation and has been shown to regulate hid expression in a context-dependent manner (Bilak and Su, 2009; Davidson and Duronio, 2012; Tanaka-Matakatsu et al., 2009). We found that in E2f1 mutant embryos, hid expression is delayed in a manner similar to what is observed in mitosis-defective stg mutants, suggesting that cell cycle regulation as the cells reach the posterior turn at stage 12 plays an important role in hid expression dynamics in the CVM (Figure S1I,J).

A BMP-responsive enhancer controls stg expression in the CVM

To investigate how the timing of cell division is controlled in migratory CVM, we used publicly-available ChIP-seq data for the transcription factors Biniou (Bin) and Myocyte Enhancer Factor 2 (Mef2), which are known to support enhancer activity in the visceral mesoderm (i.e. CVM and TVM) (Junion et al., 2012; Zinzen et al., 2009) in conjunction with published enhancer data (Kvon et al., 2014) to identify candidate stg visceral mesoderm enhancers. Published enhancer constructs VT49281 and VT49282 (Kvon et al., 2014) contain an overlapping region that shows strong occupancy by both Bin and Mef2 from 6-8hr ChIP-seq data, which allowed us to identify a 1.5-kb enhancer that supports expression of stg in both the CVM and TVM (Figure 2A,B). Additionally, phosphorylated Mothers Against Decapentaplegic (pMAD) ChIP-seq data from 6-8hr embryos (Junion et al., 2012) showed strong occupancy at this enhancer, suggesting a direct requirement for BMP signaling in controlling expression (Figure 2A). We verified that this enhancer is expressed in the CVM by checking colocalization of the lacZ reporter mRNA signal with Tey antibody to label CVM and Bin antibody to label both CVM and TVM (Figure S2A-C’). To test whether this stg visceral mesoderm (stgVM) enhancer is BMP-responsive, we crossed the stgVM-lacZ reporter into a background that is mutant for the BMP type I receptor encoded by thickveins (tkv) (Brummel et al., 1994; Penton et al., 1994; Terracol and Lengyel, 1994). Crossing the stgVM reporter construct into a tkv mutant background resulted in a loss of lacZ expression as visualized by FISH (Figure 2D, compare with C, S2F), which was also observed when a dsRNA construct targeting tkv was expressed specifically in the CVM (Figure S2D,F); conversely, overexpressing constitutively-active Tkv specifically in the CVM resulted in increased lacZ mRNA signal (Figure S2E,F), demonstrating that this enhancer is BMP-responsive. Identification of the stgVM enhancer complements previous studies that described a role for BMP signaling in influencing context-dependent stg expression dynamics throughout development, including in the early embryo and the trachea (Djabrayan and Casanova, 2016; Edgar et al., 1994).

Figure 2. A BMP-responsive stg enhancer controls timing of cell division in the CVM in response to destination-derived Dpp and cell-autonomous Tok activity.

(A) Location of ~1.5kb stgVM enhancer relative to stg locus and previously published Stark Lab constructs (Kvon et al., 2014) and ChIP-seq data for Bin, Mef2, and pMAD (Junion et al., 2012; Zinzen et al., 2009). (B) Alkaline phosphatase ISH with lacZ riboprobe in stage 12 and stage 13 embryos expressing the stgVM-lacZ reporter construct. (C, D) FISH with lacZ probe (green) and anti-Tey antibody staining (purple) in stage 12 WT (C) and tkv mutant (D) embryos. Yellow arrowheads denote CVM-specific signal. (E) Hybridization chain reaction (HCR) staining with dpp riboprobes (white) and immunostaining with Tey antibody (purple) in sequentially-staged embryos. Lateral ectoderm expression is denoted by green arrowheads and dorsal epidermis expression is denoted by light blue arrowheads. (F) Embryos expressing HGV reporter immunostained with anti-GFP (red), anti-FasIII (green), and anti-pMAD (blue) antibodies. PS7 expression domain is indicated by orange arrowhead in E and F. (G-H’) HCR staining with tkv (green) and HLH54F (purple) riboprobes in WT stage 11 sagittal (G) and transverse (H) views, with magnified view (H’) showing tkv mRNA expression in the CVM. (I-J) Embryos stained with pMAD antibody, with WT (I) showing normal PS3 (blue border) and PS7 (yellow border) expression domains. In org-1^OJ487 (J) embryos, PS7 expression domain is highly reduced and PS3 expression is missing. In bap>dpp (K) embryos, pMAD expression domain is expanded along the length of the TVM. (L-N) Dorsal-view stage 11 WT, org-1^OJ487, and bap-GAL4>dpp (bap>dpp) embryos immunostained with anti-Tey antibody (white). (O) Quantification of stage 11 dorsal view CVM cell number using Tey antibody. The number of Tey-positive nuclei in both left and right sides was counted and tabulated for each genotype (n=8-9 per genotype). (P-R) Lateral-view stage 13 WT, org-1^OJ487, and bap-GAL4>dpp (bap>dpp) embryos immunostained with anti-Tey antibody (white). (S) Quantification of CVM cell number in stage 13 WT, org-1^OJ487, and bap>dpp embryos (n=10 per genotype). p<0.05, Scale bars 20μm. See also Figure S2.

We therefore investigated how BMP signaling is transduced in the CVM. The secreted BMP2/4-related ligand Decapentaplegic (Dpp) is essential for coordinating growth and proliferation in wing discs, where Dpp is expressed in a thin stripe and laterally diffuses as a gradient (Akiyama and Gibson, 2015). In addition to strong lateral ectoderm and dorsal epidermis expression in the embryo, Dpp is expressed in the TVM, where it eventually refines into a discretely localized stripe of expression at parasegments 3 and 7 (Jackson and Hoffmann, 1994). Immunostaining with pMAD antibody to visualize active BMP signaling demonstrated a correlation between expression of Dpp and pMAD expression domains in the embryo, including PS7 of the TVM (Figures 2F, S3A-A”). Additionally, we verified the expression of tkv mRNA in the CVM (Figure 2G-H’), and found that ectopic expression of constitutively active Tkv using the CVM-specific G447-GAL4 driver was sufficient to promote precocious cell division (Movie S2). Discrete Dpp expression in the TVM, particularly in segments PS3 and PS7, requires inputs from the homeotic genes abdominal-A (abd-A) and Ultrabithorax (Ubx) (Capovilla et al., 1994; Rauskolb and Wieschaus, 1994; Reuter et al., 1990; Thüringer and Bienz, 1993) and is required for normal midgut morphogenesis (Galeone et al., 2017; Panganiban et al., 1990). Additionally, the Drosophila Tbx1 ortholog optomotor-blind-related-gene-1 (org-1) is required for expression of Ubx and dpp in the TVM, while not affecting specification of the midgut musculature (Schaub and Frasch, 2013). Over the course of migration, CVM cells directly contact and pass over PS7 (Figures 2E,F,S3). As Dpp is expressed by a large number of tissues in the embryo in addition to PS7-specific expression in the TVM (Figure 2E,I) over the course of stages 11-13, and due to the difficulty of separating earlier functions of Dpp from potential later functions in the lateral ectoderm and dorsal epidermis expression domains, we investigated whether PS7-derived Dpp serves as a more specific source of ligand for BMP signaling in the CVM. To this end, we used org-1^OJ487 (referred to as org-1 mutant in this study), an allele that results in highly-reduced PS7-specific dpp expression and complete loss of PS3-specific expression [(Schaub and Frasch, 2013); Figure 2J,M,Q compare with I,L,P)]. While CVM cell number between WT and org-1 mutants was not significantly different at stage 11 (Figure 2O), we observed a reduced number of CVM cells in org-1 mutant embryos at the conclusion of stage 13, suggesting that TVM-derived Dpp helps coordinate CVM proliferation (Figure 2S). However, ectopic expression of Dpp throughout the TVM was not sufficient to induce precocious or otherwise dysregulated proliferation of CVM cells (Figure 2K,N,O,R,S), suggesting another layer of regulation limits BMP signaling capacity during migration. Furthermore, tkv mutant embryos exhibit wide-ranging defects due to multiple roles of BMP signaling in supporting earlier patterning, including TVM specification and morphogenesis (Bradley et al., 2003; Frasch, 1995). In order to investigate a more CVM-specific role for BMP signaling and its apparent regulation during collective migration, we used published RNAseq data to identify BMP signaling effectors that are enriched in the CVM (Bae et al., 2017).

Secreted Tok metalloprotease regulates BMP signal transduction in the CVM

The secreted metalloprotease encoded by the gene tolkin (tok), a known effector of BMP signaling, is expressed specifically in the CVM throughout migration (Bae et al., 2017; Finelli et al., 1995) (Figure 3A-B’). Like its closely-linked gene, tolloid (tld), the product of tok is an ortholog of mammalian BMP-1 and participates in BMP signaling by cleaving the inhibitor Short gastrulation (Sog), freeing BMP ligands, including Dpp (Serpe et al., 2005). We hypothesized that CVM-derived Tok works cell-autonomously to process Dpp ligand secreted from tissue substrates, transducing BMP signaling over the course of migration and influencing the timing of proliferation in migratory CVM. At stage 12, in both WT and htl mutants, the CVM contains a subset of cells that are pMAD-positive as visualized by immunostaining (Figure 3C-D’). However, in tok loss of function mutants, we observed an absence of pMAD staining in CVM cells at this same stage, suggesting that Tok functions as a BMP effector in migratory CVM (Figure 3E-E’, S3B-B”).

Figure 3. Loss of BMP signaling attenuates cell death in FGF mutant CVM collectives.

(A-B’) FISH with tok riboprobe (green) and anti-Tey antibody staining (purple) in dorsally-oriented stage 1l (A, A’) and laterally-oriented stage 13 (B, B’) WT embryos. (C-E) Stage 12 embryos expressing HGV reporter immunostained using GFP antibody (red), FasIII (green), and pMAD (blue). In both WT (C, C’) and htl mutant (D, D’) embryos, pMAD expression was observed in CVM cells (white arrowheads), while tok showed a lack of pMAD expression in the CVM (E, E’). (F) Stage 13 WT embryo expressing the HGV reporter immunostained with anti-GFP (red) and anti-FasIII (cyan) antibodies, and assayed with TUNEL (green). Over the course of migration, a proportion of CVM cells will have undergone apoptosis, as demonstrated by presence of TUNEL-positive cells along CVM cell rows and posterior (blue arrowheads). (G) Stage 13 WT embryo expressing the hid5’F-GFP reporter to mark dead/dying cells immunostained with anti-GFP (green), anti-Tey (red), and anti-FasIII (cyan) antibodies. (H-M) Dorsal views of stage 11 embryos immunostained using Tey antibody. (N-S) Lateral views of stage 13 embryos immunostained with Tey antibody. (T-U) Quantification of CVM cell number at stage 11 dorsal view (T) versus lateral view using Tey antibody (U). For stage 11, the number of Tey-positive nuclei in both L and R was counted and tabulated for each genotype (T, n=8-10), while number of nuclei was counted in either Left and Right (lateral view) for stage 13 embryos (U, n=7-10 per genotype). p<0.05, Scale bars=20μM. See also Figure S3.

Furthermore, CVM cell number in tok mutants was significantly reduced compared to WT in a stage-specific manner as quantified via Tey antibody staining (Figure 3H,N,J,P). At stage 11, a similar number of CVM cells is present in tok mutants compared to wildtype (Figure 3T); however at stage 13, subsequent to cell division, wildtype cell number has increased but the increase observed in tok mutants is significantly smaller (Figure 3U). Surprisingly, double mutants for tok and htl presented significantly improved CVM cell survival (Figure 3I,O, compare with K,Q; T,U). This was also observed in double mutants for htl and tkv as well as in htl mutants expressing a dsRNA hairpin targeting tkv specifically in the CVM (Figure 3L,R,M,S,T,U). Based on our data showing how BMP signaling regulates stg expression in the visceral mesoderm, we believe that the reduction in cell number in the tok mutant background compared to WT is due to a disruption in stg-mediated proliferation due to loss of BMP signaling activity. Therefore, the BMP-FGF double mutants phenocopy the htl stg mutant phenotype, suggesting that the role of BMP signaling in coordinating proliferation directly impacts the timing of cell death during CVM migration. Collectively, these results suggest that in this context, 1) BMP signaling promotes timely anoikis via its role in cell cycle progression to facilitate hid expression, and that 2) activating BMP signaling to promote cell cycle progression requires CVM-specific Tok metalloprotease, which in turn allows CVM collectives to transduce BMP signaling in the presence of secreted Dpp ligand in the extracellular milieu.

We further investigated whether spatial constraints relate to CVM cell death. TUNEL staining at stage 13 supports the view that cells not contacting the TVM undergo apoptosis (Figure 3F). Using a reporter of hid expression, hid5’F-GFP (Tanaka-Matakatsu et al., 2009), to assay CVM cells at stage 13, we show that even in wildtype embryos a few CVM cells are associated with elevated hid expression but no Tey staining, which indicates cell death. Furthermore, hid5’F-GFP signal is elevated only in cells located at a distance from the TVM and reduced in the majority of cells that directly contact the TVM (Figure 3G). Increased hid5’F-GFP expression in CVM cells that appear at a distance from the TVM may relate to loss of FGF signaling and subsequent de-adhesion of the CVM, as the TVM expresses Pyramus (Pyr), a membrane-associated FGF ligand (Kadam et al., 2012; Reim et al., 2012; Stepanik et al., 2020).

The G2-M transition temporally positions CVM anoikis during migration

To gain further insight into the mechanism by which BMP signaling promotes both cell proliferation and apoptosis in the CVM, we used the fly Fluorescent Ubiquitin-associated Cell Cycle Indicator (Fly-FUCCI) system (Figure 4A-C). Fly-FUCCI consists of a bicistronic construct under the control of a UAS promoter that allows for expression of fluorophore-conjugated peptides of the cell cycle genes E2f1 and CycB to present a readout of cell cycle stage; the RFP-conjugated CycB degron is expressed through DNA synthesis (S) and growth phase 2 (G2) before being degraded, while the GFP-conjugated E2f1 degron is expressed beginning in G2 before being degraded prior to S phase (Zielke et al., 2014). CVM-specific expression of Fly-FUCCI reveals the transition from G2 through mitosis followed by a transient G1 at the posterior turn at stage 12 in both WT and htl mutant backgrounds (Figure 4A,B); however, in a tok htl double mutant background, CVM cells are arrested at G2 at stages 12-13 (Figure 4C). Furthermore, direct manipulation of cell cycle effectors can recapitulate survival or cell death phenotypes in the CVM in a cell cycle-dependent manner (Figure 4D-M); CVM-specific expression of a stg dsRNA construct using the GAL4 system (G447>stgRNAi) in a htl mutant background results in prolonged survival past stage 12 (Figure 4E,I, compare with D,H), while ectopic expression of stg using this same expression system (G447>stg) has the opposite effect. Expression of stg in a tok htl double mutant ‘un-rescues’ the survival associated with loss of both FGF and BMP signaling components (Figure 4G,K,M compare with Figure 3K,Q,U), suggesting that cell proliferation is sufficient for temporally positioning cell death in migratory CVM. Additionally, precocious cell death was observed when a dsRNA construct targeting tribbles (trbl), which encodes a negative regulator of Stg (Mata et al., 2000), was expressed specifically in the CVM in a htl mutant background (G447>trblRNAi; Figure 4F,J,L,M), further supporting a model in which cell cycle progression positions a quality control checkpoint for the collective (Figure 4N,O).

Figure 4. The role of BMP in cell cycle progression is sufficient for temporal positioning of anoikis in FGF mutant CVM.

(A-C) Sequentially-staged embryos expressing the FlyFUCCI reporter in which RFP-conjugated CycB degron is expressed through S and G2 phases of the cell cycle before being degraded, while the GFP-conjugated E2f1 degron is expressed beginning in G2 before being degraded prior to S phase (Zielke et al., 2014) immunostained using antibodies for RFP (red) and GFP (green). In WT (A) and htl mutant (B) embryos, the G2-M transition occurs concomitant to migration around the posterior turn. In tok htl mutant (C) embryos, CVM cells are arrested at G2. (D-G) Dorsal views of stage 11 embryos immunostained with Tey antibody. (H-K) Lateral views of stage 13 embryos immunostained with Tey antibody. The number of Tey-positive nuclei in both L and R was counted and tabulated for each genotype in dorsally-oriented stage 11 embryos (L, n=8-10), while number of nuclei was counted in either Left and Right (lateral view) for laterally-oriented stage 13 embryos (M, n=9-10). (N) Schematic summarizing CVM cell response to destination-specific cues during collective migration. Exposure to both FGF and Dpp signals ensures survival and on-track migration, while de-adhesion and exposure only to Dpp signal results in anoikis cell death. (O) Model summarizing mechanism of FGF and BMP signal transduction in CVM cells and the survival vs death response via downstream control of stg transcription and antagonism of Hid. p<0.05, Scale bars=20μM. See also Figure S4.

DISCUSSION

In this study, we described a mechanism for correcting errors in mesenchymal collective cell migration prior to organ assembly. First, we described how initiation of CVM cell death due to loss of FGF signaling requires cell cycle-regulated expression of hid (e.g. Figure 1I,K). Next, we demonstrated how BMP signaling transduced by cell-autonomous Tok activity coordinates the timing of cell cycle progression of CVM cells over the course of migration by promoting BMP signaling (e.g. Figure 3C-E and 4C). Third, we showed how coordinated cell cycle progression ensures timely apoptosis of cells that have gone off track, ensuring that cells that remain on-track undergo myoblast fusion (e.g. Figure 3F,G). In this system, fitness is conferred by proximity and ability to transduce both FGF and BMP signals by members of the collective. Although both FGF and BMP receptors are expressed in the CVM, FGF ligand expression is restricted to the TVM track, while Dpp is expressed broadly throughout the embryo, including in a subpopulation of TVM cells anterior to migratory CVM (Figures 2E and 4N). As such, Tok metalloprotease secreted by CVM cells is critical to ‘sensing’ a given source of Dpp, triggering either growth or anoikis (Figure 4O).

Cell proliferation and negative regulation of Hid-mediated apoptosis proceeds when CVM cells are both adhered to an FGF ligand-presenting substrate as well as close to a source of secreted Dpp, which is normally bound by Sog -- necessitating cleavage by Tok to free Dpp ligand for activation of BMP signaling in the CVM (Figure 4O). In cases of CVM de-adhesion or mismigration to a substrate that expresses Dpp but not FGF, secreted Tok processes Sog and frees Dpp for signaling and subsequent cell cycle progression as normal; however, the absence of FGF-dependent Hid antagonism results in rapid apoptosis of the lost cell (Figure S4A). Finally, when both FGF and BMP signaling components (i.e. either Tok ligand-activating protease or Tkv receptor) are lost, CVM cells are rendered refractory to secreted Dpp, resulting in cell cycle arrest and survival in a senescence-like state (Figure S4B). This is a striking example of a developmental context in which Dpp can promote apoptosis, as opposed to its more well-known role in promoting cellular growth, thus complementing studies in epithelial tissues describing a requirement for boundary Dpp in positioning cell death for leg morphogenesis (Manjón et al., 2007) and the extrusion of BMP mutant cells from epithelial wing primordia (Gibson and Perrimon, 2005).

Additionally, Dpp’s function as a cell cycle regulator is sufficient for timing apoptosis in migratory CVM, as direct manipulation of cell cycle effectors can either recapitulate or ablate the survival phenotype presented by double mutants for FGF and BMP signaling components (Figure 4D-M). Positioning of the G2-M transition by Stg is a critical step of the regulatory cascade leading to apoptosis, but likely requires the concomitant activity of the E2f1 transcription factor, which has a known role in promoting apoptosis via regulation of hid (Davidson and Duronio, 2012; Moon et al., 2005). Surprisingly, surviving CVM cells in tok htl double mutants that arrest at G2 in stage 13 (Figure 4C) re-enter the cell cycle by stage 15 (Figure S4C,D); this suggests that the absence of both FGF and BMP signaling first induces an arrested senescence-like state, whereby survival after a prolonged G2 results in cell cycle re-entry, perhaps through the activation of additional signaling pathways. This is reminiscent of the cell cycle stalling induced by JNK signaling in the fly wing disc in response to cellular damage (Cosolo et al., 2019; Ohsawa et al., 2012). Therefore, our study has tantalizing implications regarding the existence of additional cell types in Drosophila that undergo a similar senescence-like state. Furthermore, the role for BMP in promoting apoptosis has potential parallels in higher organisms, as reduced apoptosis in pulmonary artery smooth muscle cells (PASMCs) is associated with BMP signaling, and can lead to hypertrophy of the pulmonary vasculature in primary pulmonary hypertension (PPH) (Zhang et al., 2003).

We have shown that hid is required for eliminating CVM cells that have gone off-track; in hid single mutants, we observed cells that have migrated ventrally that are absent in WT embryos (Figure S4G, compare with E). However, it is also likely that additional pro-apoptotic genes like grim and reaper (rpr) are involved, as the htl hid double mutant background resulted in a significantly lower number of surviving CVM cells when compared to WT (Figure 1G). Indeed, when we compared hid single mutants with a deficiency in which the three pro-apoptotic genes grim, rpr, and hid are deleted [Df(3L)H99, referred to as DfH99 in this study (Chen et al., 1996; Grether et al., 1995; Mackay and Bewley, 1989; White et al., 1996; Zhou et al., 1997)], we not only found a significantly higher number of CVM cells in DfH99 embryos when compared to both WT and hid single mutants (Figure S4F), but also observed differences in cell morphology between the different genotypes; strikingly, a continuous ‘bridge’ of CVM cells was observed in the DfH99 embryo (Figure S4E-H’). This may hint at additional components of the quality control system that not only support apoptosis of different populations of CVM cells, but may potentially regulate non-apoptotic functions, such as directional myoblast fusion.

We expect that this study will open new avenues for understanding how a continuum of cell interactions can contribute to development and cancer. While it is generally accepted that the acquisition of a mesenchymal phenotype is required for metastasis in most cancers, the extent to which cooperative interactions are required between migratory mesenchymal cancer cells and other tissue types has yet to be determined. Furthermore, the invasion of other tissues by CVM that have lost the cell death program resembles pathologies associated with Tuberous Sclerosis Complex (TSC), a developmental disorder that can result in the formation of smooth muscle cell-derived benign tumors in other tissues (Lesma et al., 2005). The developmental basis for how these tumors arise and the role that collective migration may play remain to be investigated, and is a promising avenue for identifying broadly conserved regulatory mechanisms governing tissue homeostasis.

Limitations of the study

Although we focused on the canonical role for Tok metalloprotease in mediating BMP signaling via cleavage of the Dpp inhibitor Sog for this study, we observed migration defects in the tok and BMP mutant backgrounds that were apparent at stage 11, prior to the cells reaching the posterior turn. This suggests that there may be an additional role for BMP signaling in ensuring proper CVM migration, and perhaps additional targets for Tok-dependent cleavage that might contribute to this system, which will require further study. Further work will also be needed to identify which signaling pathways might be responsible for preferential migration of FGF mutant CVM towards the CNS in the htl hid double mutant background.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Angelike Stathopoulos (angelike@caltech.edu).

Materials availability

Drosophila strains and other reagents generated in this study will be available upon request from the lead contact, or the commercial sources listed in the key resources table.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit Anti-Teyrha-meyrha	(Macabenta and Stathopoulos, 2019b)	N/A
Mouse Anti-Fasciclin III	Developmental Studies Hybridoma Bank (DSHB)	7G10
Goat Anti-GFP	Rockland Immunochemicals	Cat# 600-101-215; RRID: AB_218182
Rabbit Anti-RFP	MBL International	Cat# PM005; RRID: AB_591279
Rabbit Anti-Beta-Galactosidase	MP Biomedicals	Cat# 559761; RRID: AB_2687418
Rabbit Anti-Phospho-SMAD1/5 (Ser463/465)	Cell Signaling Technology	Cat# 9516S; RRID: AB_491015/ Lot: 9
Rabbit Anti-Phospho-Histone H3 (Ser10)	EMD Millipore	Cat# 06-570; RRID:AB_310177
Rabbit Anti-Biniou	(Jakobsen et al., 2007)	N/A
Sheep Anti-Digoxigenin Polyclonal Antibody	Thermo Fisher Scientific	Cat# PA1-85378; RRID:AB_930545
Anti-Digoxigenin-AP, Fab Fragments Antibody	Sigma-Aldrich	Cat# 11093274910; RRID:AB_2734716
Alexa Fluor 488 goat anti-guinea pig	Molecular Probes	Cat# A-11073; RRID: AB_2534117
Alexa Fluor 555 donkey anti-rabbit	Molecular Probes	Cat# A-31572; RRID: AB_162543
Alexa Fluor 647 donkey anti-mouse	Molecular Probes	Cat# A-31571; RRID: AB_162542
Chemicals, Peptides, and Recombinant Proteins
Digoxigenin labeled nucleotides	Roche	11277073910
Critical Commercial Assays
In Situ Cell Death Detection Kit, Fluorescein	Roche	11684795910
Hybridization Chain Reaction	Molecular Technologies	N/A
Deposited Data
stg Enhancer - Stark Lab Fly Enhancers	(Kvon et al., 2014)	VT49281
stg Enhancer - Stark Lab Fly Enhancers	(Kvon et al., 2014)	VT49282
Mef2 6-8hr ChIP Data	(Zinzen et al., 2009)	E-TABM-649
Bin 6-8hr ChIP Data	(Zinzen et al., 2009)	E-TABM-652
pMAD 6-8hr ChIP Data	(Junion et al., 2012)	E-TABM-1184
Experimental Models: Organisms/Strains
D. melanogaster: htlAB42/TM3,ftz-lacZ	Bloomington Drosophila Stock Center (BDSC)	#5370
D. melanogaster: tok3/TM3,Sb1,Ser1	BDSC	#4569
D. melanogaster: tkv7/CyO	BDSC	#3242
D. melanogaster: hid1	BDSC	#631
D. melanogaster: org-1OJ487/FM7c, P{ftz-lacC}YH1	BDSC	#91555
D. melanogaster: stg4/CyO	BDSC	#2500
D. melanogaster: P{20XUAS-6XmCherry-HA}attP2	BDSC	#52268
D. melanogaster: P{w[+mC]=UASp-GFP.E2f1.1-230}26 P{w[+mC]=UASp-mRFP1.NLS.CycB.1-266}4/CyO, P{ry[+t7.2]=en1}wgen11; MKRS/TM6B, Tb1	BDSC	#55110
D. melanogaster: UAS-stgN16/CyO	BDSC	#4777
D. melanogaster: UAS-dpp.S	BDSC	#1486
D. melanogaster: UAS-tkv.Q253D.Nb/TM3, Sb1 Ser1	BDSC	#36536
D. melanogaster: UAS-Ras1a	Norbert Perrimon	N/A
D. melanogaster: G447-GAL4	(Georgias et al., 1997)	N/A
D. melanogaster: P{GawB}tey5053A/TM6,Tb	BDSC	#2702
D. melanogaster: bap-GAL4	BDSC	#91540
D. melanogaster: HLH54F-gap-Venus (HGV reporter)	(Kadam et al., 2012)	N/A
D. melanogaster: HLH54F-H2A-mCherry (HC3 reporter)	(Kadam et al., 2012)	N/A
D. melanogaster: P{w[+mC]=hid-EGFP.5'F-WT}2	BDSC	#50750
D. melanogaster: P{TRiP.JF01486}attP2 (tkv RNAi)	BDSC	#31041
D. melanogaster: P{TRiP.GL00513}attP40 (stg RNAi)	BDSC	#36094
D. melanogaster: P{TRiP.HMJ02089}attP40 (trbl RNAi)	BDSC	#22113
D. melanogaster: Df(3L)H99, kniri-1 pp/TM3, Sb1	BDSC	#1576
Oligonucleotides
HLH54F RNA probes (HCR)	Molecular Technologies	N/A
dpp RNA probes (HCR)	Molecular Technologies	4049/E192
tkv RNA probes (HCR)	Molecular Technologies	N/A
GATCCGGGAATTGGGAATTCTATCGAG AAATATAT	This study	StgVM_F
GCAGATCTGTTAACGAATTCCGTGTGC ATTTGCCA	This study	StgVM_R
Recombinant DNA
hid CDNA clone AT13267	Drosophila Genomics Resource Center (DGRC)	Stock #11756
stgVM.evep.lacZ	This study	N/A
Software and Algorithms
Zen 3.0 (Blue edition)	Zeiss	N/A
G*Power	(Faul et al., 2009)	https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower.html
Prism 9	GraphPad Software	https://www.graphpad.com/
Fiji/ImageJ	(Schindelin et al., 2012)	https://imagej.nih.gov/ij/

Data and code availability

• No large-scale datasets have been generated in this study. The raw microscopy data that support the findings of this study are available from the lead contact upon reasonable request.

• This study did not generate any software and code. Any additional information required to reanalyze the data shown in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly stocks and husbandry

All fly stocks were kept at 22-25°C in standard medium. The CVM reporter HLH54F-gap-Venus was described previously (Kadam et al., 2012; Stepanik et al., 2016) and combined with htl^AB42 mutant and GAL4/UAS lines via standard genetic crosses. Wild type (WT) refers to yw unless otherwise noted. For experiments, flies were kept in collection cages with apple juice agar plates supplemented with yeast paste and allowed to lay for at least eight hours before egg collection and dechorionation in 100% bleach.

METHOD DETAILS

Whole-mount In situ hybridization, immunofluorescence staining

Embryos were collected and fixed using previously described methods (Frasch, 1995; Jiang et al., 1991). The antisense Digoxigenin (DIG) labeled RNA probe targeting hid was generated with linearized DGRC CDNA clone AT13267 (Stock #11756), and the lacZ RNA probe was generated from linearized plasmid. For alkaline phosphatase in situ hybridization (AP ISH), an anti-DIG antibody conjugated to alkaline phosphatase (1:500, Sigma, 11093274910) was used in conjunction with NBT/BCIP (Roche) to visualize signal. For fluorescent in situ hybridization (FISH), a sheep anti-DIG polyclonal antibody (1:200, Thermo Fisher, PA1-85378) was used in conjunction with Alexa Fluor secondary antibody (1:500, Molecular Probes).

Antibody stainings were performed as previously described. The primary antibodies and dilutions used in this study were Rabbit anti-Tey (Macabenta and Stathopoulos, 2019b), Mouse anti-FasIII (1:200, DSHB, 7G10), Goat anti-GFP (1:5000, Rockland, 600-101-215), Rabbit anti-RFP (1:500, MBL, PM005), Rabbi anti-Beta-Galactosidase (1:500, MP, 559761), Rabbit anti-Phospho-SMAD1/5 (1:50, Cell Signaling, 9516S), Rabbit anti-PH3 (1:500, EMD, 06-570), Sheep anti-DIG (1:200, Thermo Fisher, PA1-85378), Anti-DIG AP (1:200, Sigma, 11093274910), and Rabbit anti-Biniou (1:500, Jakobsen et al., 2007). Immunofluorescence staining was performed using Alexa Fluor 488, 555 and 647 secondary antibodies (1:500, Molecular Probes). TUNEL was performed by using the In Situ Cell Death Detection Kit (Roche) according to manufacturer protocols. Embryos were mounted in Permount (Fisher Scientific) for whole-mount AP ISH experiments or in 70% glycerol in 1× PBS buffer for whole-mount immunofluorescence-stained experiments. Imaging was performed using an Axio Imager Z2 (Zeiss) for AP ISH embryos, and an LSM 800 (Zeiss) for immunofluorescence-stained embryos. Zen blue edition (Zeiss) was used to process images for subsequent analysis and figures.

Quantification of CVM cell number

Confocal images of embryos immunostained with Tey antibody were obtained for stage 11 (dorsal view and 13 (lateral view) using an LSM800 with 20X objective. Embryos were staged by assessing both FasIII antibody signal intensity and lateral amnioserosa morphology over the course of germband retraction. Tey-positive nuclei were counted manually in Fiji/ImageJ (Schindelin et al., 2012) by planing through the z stacks and using the Cell Counter plug-in. For lateral views, ventral CNS and lateral M12 muscle-specific expression of Tey was excluded from counts. Experiments were repeated 2-3 times and power analysis to determine sample size was performed using the G*power tool (Faul et al., 2009). Graphs and statistical analyses were generated using Prism 9 (GraphPad).

Hybridization chain reaction

Hybridization chain reaction (HCR) was performed as described in (Slaidina et al., 2021) with modifications to initial steps to account for embryonic as opposed to ovarian tissue. The probes used were dpp, tkv, and HLH54F.

Cloning of reporter construct and generation of transgenic flies

The stgVM sequence was PCR amplified using Phusion High Fidelity polymerase and ligated into the eve2promoter-lacZ vector using standard techniques. Site-directed transgenesis was carried out using a D. melanogaster stock containing an attP insertion site at positions ZH-51C (Bloomington stock #24482) and ZH-86Fb (Bloomington stock #23648). The primers used were:

Stg^VM_F: GATCCGGGAATTGGGAATTCTATCGAGAAATATAT

Stg^VM_R: GCAGATCTGTTAACGAATTCCGTGTGCATTTGCCA

Quantification of lacZ mRNA signal

Embryos were collected and FISH was performed using lacZ riboprobes in conjunction with Tey antibody staining. Individual embryos were subsequently staged and imaged using an LSM 800 confocal microscope. Three replicates were used for each genotype, with 23-25 total CVM measurements across three samples. Raw .czi files were processed using ImageJ software, with single planes used for quantification. Regions of interest (ROIs) were demarcated by using Tey-positive CVM nuclei as a guide, and mean fluorescence intensity values for lacZ riboprobe signal within each ROI were obtained via the Analyze→Measure→Mean Value function. Corrected mean values were obtained by measuring background signal/noise in a separate ROI and subtracting the resulting fluorescence intensity value from each CVM-associated lacZ mean fluorescence intensity measurement. One-way ANOVA was applied to assess statistical significance.

Supplementary Material

1

Movie S1. Cell division and anoikis in migratory CVM cells. Related to Figure 1. Embryo expressing hexameric mCherry reporter specifically in the CVM via G447-GAL4 driver. CVM cells divide at the posterior turn, and a number of cells go off-track and undergo anoikis cell death shortly after.

2

Movie S2. Precocious CVM mitosis induced by constitutively-active Tkv. Related to Figure 2. Embryo with HC3 reporter expressing constitutively-active Tkv protein specifically in the CVM via the G447-GAL4 driver. Precocious cell division was observed in a subset of cells.

Highlights.

• Cell death in migrating caudal visceral mesoderm (CVM) cells requires Hid

• Hid levels are correlated with onset of CVM cell division

• A BMP-responsive stg enhancer controls timing of cell division during CVM migration

• Cell cycle-coupled Hid expression eliminates cells that have lost access to FGF