Abstract
During development, interfaces between cells with distinct genetic identities generate signals organizing local cell behaviors to drive tissue morphogenesis. During Drosophila embryonic axis extension, stripes of pair rule genes position parasegment boundaries and orchestrate polarized cell intercalation driven by planar polarized Myosin-II via patterned expression of several Toll receptors. However, how Toll receptors define polarized contractile interfaces remains unknown. Here we show that differential expression of Toll-8 polarizes Myosin-II through a physical interaction with the adhesion GPCR Cirl/Latrophilin. We report that Toll-8 activates Myosin-II by inducing Cirl asymmetric localization at the boundary of its expression domain. Cirl asymmetry is in turn sufficient to activate Myosin-II. Toll-8 and Cirl exhibit dynamic interdependent interactions to yield robust planar polarity when neighboring cells express different levels of Toll-8. This reveals how planar polarized signaling arises from a scalar gradient of a single Toll receptor interpreted through self-organized feedbacks with a GPCR.
Main Text
Signaling at the interface between cells with distinct genetic identities is an important source of information for developmental patterning, cell growth and division, polarization and morphogenesis.
Interfaces between differently fated cells often lead to cell sorting due to mechanical barriers that block cell mixing in developmental (1) or pathological contexts (2). For instance, at developmental compartment boundaries between cells expressing different transcription factors, actomyosin contractility can be increased due to differential expression of cell surface proteins, such as Eph-Ephrin (1, 3).
Interfaces between cells of different genetic identities are not always stable. In myc-induced cell competition, cells expressing high levels of the proto-oncogene Myc (“winner” cells) eliminate cells expressing lower levels of Myc (“loser” cells) by apoptosis (4, 5). Cells are principally eliminated at the interface between these two groups indicating direct cell-cell communication driven by differential expression of cell surface proteins (6). Cell mixing is locally promoted at winner-loser cell interfaces through cell intercalation increasing the surface of contact (7).
Planar Cell Polarity (PCP) is another phenomenon resulting from interfacial signaling that can generate stable or dynamic cell interfaces. The core PCP pathway, involving surface proteins such as Flamingo/Celsr and Frizzled, translates tissue-scale cues into vectorial cell polarity required for planar polarized actomyosin contractility in certain contexts (8, 9). Gradients of the Fat/Dachsous adhesion molecules induce planar polarized activation of the myosin Dachs at cell interfaces and dynamic cell rearrangements important for tissue growth and morphogenesis (10).
What are the strategies used by cells to encode mechanical polarity at the boundary between cells with different genetic identities? In Drosophila embryos, maternally supplied morphogens gradients along the anteroposterior (AP) axis trigger a transcriptional cascade that culminates in the formation of stripes of pair rule genes, which control planar polarized Myosin-II (Myo-II) enrichment at interfaces between AP neighboring cells (11, 12). Positions of cells along the gradient determine their positional identities and their mechanical behaviors: certain columns of cells become parasegment boundaries that block cell mixing (13–15) while in the rest of the field cells undergo planar polarized cell intercalation required for tissue extension (11, 16, 17). Spatial control over the amplitude and polarity of actomyosin contractility determines the nature of these changes (18, 19). Levels of Myo-II activation at cell interfaces are quantitatively controlled by G protein-coupled receptor (GPCR) signaling (20, 21) but the molecular mechanisms controlling Myo-II planar polarity remain unclear. Interactions between pair rule genes define periodic and partially overlapping stripes of Toll receptors perpendicular to the AP axis (22). Each column of cells expresses a different combination of Toll-2,6,8, which are required for planar polarized Myo-II activation and axis extension (14, 22).
Toll-2,6,8 are part of the “Long Toll” subfamily of Toll receptors and are characterized by a large number of Leucine Rich Repeats (LRRs) in their extracellular domains (23). Toll receptors are well known for their conserved function in developmental patterning and innate immunity through the NFκB pathway (24). They recently emerged as conserved molecules involved in embryonic axis elongation (15, 22, 23), organogenesis (25), cell competition (26, 27) and synaptic partner matching (28). Surprisingly, their canonical role in NFκB pathway is dispensable in some processes (24, 25, 28). Other LRR proteins are also required for boundary formation and cell sorting in Drosophila embryos (15) and wing discs (29), as well as in vertebrate system such as Xenopus embryos (30) and chick limb development (31, 32).
How Toll receptors, and LRR proteins in general, elicit downstream signaling to define mechanical interfaces remains largely unknown. Here we investigated how Toll receptors control Myo-II planar polarity in conjugation with GPCR signaling.
Asymmetric expression of Toll-8 activates Myo-II
It has been hypothesized that trans-interactions across cell-cell interfaces between different Toll receptors signal to activate Myo-II (22). In light of the observation that ectopic expression of Toll-2 or Toll-8 alone can induce Myo-II activation late in embryogenesis (22), we first tested the simplest hypothesis that a single Toll is sufficient to activate Myo-II in the embryonic ectoderm. To this end, we injected embryos with dsRNAs targeting Toll-2,6,7 (toll-2,6,7 RNAi), leaving only endogenous Toll-8 expressed in vertical stripes (fig. S1A) and observed Myo-II with mCherry-tagged Myo-II regulatory light chain (MRLC–Ch). We found Myo-II specifically enriched at the interfaces between Toll-8 expressing and non-expressing cells (fig. S1, A and B), suggesting that Toll-8 alone can activate Myo-II.
To assess the signaling capacity of a single Toll in a more convincing manner independent of the anteroposterior (AP) patterning system, we engineered embryos expressing a single stripe of Toll-8 along the AP axis that runs orthogonal to the endogenous Toll-2,6,8 stripes, using an intermediate neuroblasts defective (ind) enhancer (33) (ind-Toll-8–HA). We monitored Myo-II with GFP-tagged Myo-II regulatory light chain (MRLC–GFP) and detected Myo-II activation at the border between ectopic Toll-8 expressing and wild-type cells (Fig. 1, A and B, Toll-8FL, and fig. S2A). Note that this ectopic Myo-II activation runs along the AP axis and is thus perpendicular to the endogenous planar polarity of Myo-II. When we deleted the Leucine-rich Repeats (LRRs) from the Toll-8 extracellular domain, Toll-8 no longer localized to the plasma membrane and failed to activate Myo-II (Fig. 1, A and B, Toll-8ΔLRR), suggesting that the LRRs of Toll-8 are essential for Myo-II activation. Consistent with an upregulation of cortical tension, the boundary of Toll-8 expressing cells was smoother when full-length Toll-8 (Toll-8FL) was expressed compared to Toll-8ΔLRR (Fig. 1, A to C, and fig. S2A, left).
When we injected ind-Toll-8–HA embryos with dsRNAs targeting only endogenous Toll-2,6,8 (toll-2,6,8 RNAi), Myo-II was still activated at the boundary of ectopic Toll-8 expressing cells and formed a supracellular cable along the AP axis (Fig. 1, D and E). This boundary was also smooth (Fig. 1, D and F). Altogether, we conclude that an interface defined by a single Toll-8 is sufficient to activate Myo-II.
This led us to further dissect the mechanisms underlying the capacity of Toll-8 to elicit Myo-II activation using clonal analysis to generate random interfaces between wild-type and Toll-8 overexpressing cells. We studied Toll-8 signaling in growing wing imaginal discs since clonal analysis is not possible in early embryos. Wing imaginal discs exhibit polarized supracellular cables of Myo-II at the periphery of the pouch region near the hinge (34). Toll overexpression causes wing developmental defects (25) and Toll-8 is expressed in the wing hinge region but not in most of the pouch region (25, 27). We induced clones overexpressing Toll-8–GFP in the larval wing disc epithelium, and monitored Myo-II localization with MRLC–Ch 24 hours after clone induction in the wing pouch. This resulted in striking and robust planar polarized Myo-II activation specifically at the boundary of Toll-8 overexpressing clones (Toll-8FL, Fig. 1, G and H). As in embryos, we found that LRRs were required for Toll-8 plasma membrane localization and Myo-II activation in wing discs (Toll-8ΔLRR, Fig. 1, G and H). Toll-8FL clones were more compact with a smoother boundary compared with Toll-8ΔLRR clones (Fig. 1, G and I, and fig. S2A, right). The cytoplasmic domain of Toll proteins is necessary for canonical Toll signal transduction (24). To test whether the cytoplasmic domain of Toll-8 is required for Myo-II activation, we generated wing disc clones overexpressing a truncated version of Toll-8, removing its cytoplasmic domain (Toll-8ΔCyto). Surprisingly, Myo-II was still activated at the boundary of Toll-8ΔCyto overexpressing clones (Fig. 1, G and H), and the clonal border was still smooth compared to Toll-8 ΔLRR clones (Fig. 1, G and I). We conclude that the cytoplasmic domain of Toll-8 is dispensable for Myo-II activation. This indicates that the signaling events leading to Myo-II activation at the border between Toll-8 expressing and non-expressing cells may require another protein, presumably interacting with Toll-8 at the cell surface.
Toll-8 and Cirl form a molecular and functional complex
To look for binding partners of Toll-8, we performed affinity purification-mass spectrometry using lysates isolated from embryos overexpressing Toll-8–YFP as a bait. Of the specifically bound proteins, Cirl was the most abundant target (Fig. 2A and table S1) and of particular interest, since it is the Drosophila homologue of vertebrate Latrophilin (fig. S3), a member of the adhesion G protein-coupled receptor (aGPCR) subfamily (35, 36). Interestingly, the Toll-8 extracellular domain shares sequence similarities with the extracellular domains of the human LRR proteins FLRTs (37) (fig. S3), which are known to form protein complexes with human Latrophilins (38–41). We also recovered Toll-2 as a significant Toll-8 target (table S1), in agreement with previous findings in vitro (22).
Using a Cirl–RFP knock-in line (42) to visualize endogenous Cirl, we found that Cirl localized to the membrane and was enriched at adherens junctions and at cell-cell interfaces slightly above junctions, both in the embryo and the wing disc (fig. S4, D to G). We next asked whether Cirl is required for Myo-II activation in the embryo. cirl maternal and zygotic mutant embryos showed delayed extension of the embryonic axis (fig. S4, A to C) and a strong reduction of Myo-II at junctions in the ectoderm (Fig. 2, B and C), resembling Myo-II reduction observed in toll-2,6,8 RNAi embryos (Fig. 2D and fig. S1A). We further tested whether Cirl is required for Toll-8-induced polarization of Myo-II. We analyzed the boundary of ind-Toll-8–HA stripe in wild-type and cirl null mutant embryos, and found a strong reduction of Myo-II activation and smoothness of the Toll-8 stripe in the absence of Cirl (Fig. 2, E to G). Moreover, in wing imaginal discs we observed a similar reduction in Myo-II activation at the boundary of clones overexpressing Toll-8–YFP when cirl was null mutant (Fig. 2, H and I). Note that Myo-II activation was not completely abolished, suggesting that other molecules in addition to Cirl mediate Toll-8-dependent polarization of Myo-II at cell interfaces. Taken together, we conclude that Toll-8 and Cirl form a molecular complex and interact functionally to polarize Myo-II at the interface of Toll-8 expressing and non-expressing cells.
Cirl asymmetric interfaces activate Myo-II
Interestingly, we observed that Myo-II was activated on both sides of the interface between Toll-8 expressing and non-expressing cells (fig. S5A, cyan arrowheads). This shows that Myo-II is also activated in cells that do not express Toll-8. We thus tested whether Cirl mediates signaling on both sides of the interface in a bidirectional manner. To this end, we performed Mosaic Analysis with a Repressible Cell Marker (MARCM) in wing discs and induced cirl null mutant clones either adjacent to (Fig. 3A) or coincident with (Fig. 3B) Toll-8 overexpressing clones. We found that Myo-II was activated on both sides at the boundary of Toll-8 overexpressing clones when Cirl was absent in neighboring cells (Fig. 3A, orange arrows, and fig. S5A, orange arrowheads), as at control interfaces (Fig. 3A, cyan arrows, and fig. S5A, cyan arrowheads). When Cirl was absent only within cells of Toll-8 overexpressing clones, Myo-II was still activated on both sides of the clonal boundary (Fig. 3B, magenta arrows and fig. S5B, magenta arrowheads). Thus, Cirl is dispensable either in Toll-8 expressing or in responding/contacting cells (Fig. 3C). This is remarkable since Cirl must be present on at least one side of Toll-8 overexpressing clones, as complete removal of Cirl on both sides reduces significantly Myo-II activation (Fig. 2, H and I). We thus hypothesized that the role of Toll-8 is to induce an asymmetry of Cirl activity at the clonal boundary. To test this, we generated cirl mutant clones in the wing disc without overexpressing Toll-8. We reasoned that at the clonal interface Cirl activity is thereby de facto asymmetric. This is supported by the observation that Cirl is still localized in wild-type cell interfaces in contact with cirl mutant cells (fig S5C). We found that Myo-II was indeed activated at the boundary of cirl mutant clones and that the boundary was smooth compared to control clones (Fig. 3, D to F). This is similar to Toll-8 overexpressing clones, albeit to a lesser extent (Fig. 3, A to E). We conclude that asymmetric Cirl interfaces can activate Myo-II and suggest that this underlies Toll-8-dependent planar polarization of Myo-II. While at the boundary of the cirl mutant clones, Cirl asymmetry is artificial, we hypothesized that Cirl localization is made asymmetric at the boundary of Toll-8 expressing cells.
Toll-8 generates a Cirl interfacial asymmetry
To test this, we assessed the effect of Toll-8 overexpressing clones on endogenous Cirl localization in wing discs. In cells that do not overexpress Toll-8–YFP, Cirl–RFP was concentrated at, and slightly above, adherens junctions (fig. S4, F and G). However, in Toll-8– YFP overexpressing cells, two situations were observed. When Toll-8–YFP was expressed at low levels, Cirl–RFP localization was restricted to cell-cell interfaces around adherens junctions (Fig. 5B). When Toll-8–YFP was expressed at higher levels, its localization was also at adherens junctions and expanded more apically (fig. S6A’). In both cases, Cirl–RFP consistently colocalized with Toll-8–YFP and was strikingly depleted from the orthogonal junctions in wild-type cells in direct contact with Toll-8 overexpressing cells (Fig. 4A, yellow arrows, Fig. 5B, and fig. S6, A and A’). These observations are consistent with Toll-8 recruiting Cirl in trans, as well as in cis but it is unclear how Cirl–RFP localizes on each side of the Toll-8–YFP +/-interface.
To address this directly we used MARCM to observe endogenously tagged Cirl–RFP only inside (i.e in cis, Fig. 4, B and D) or only adjacent to (i.e. in trans, Fig. 4, C and E) Toll-8 overexpressing cells. Note that in Cirl–RFP negative cells, untagged endogenous Cirl is present. When present only in trans, Cirl–RFP was planar polarized in cells in direct contact with the clone boundary, recruited to the clone border and depleted from the junctions orthogonal to the clone border (Fig. 4C and fig. S6C). In these cells, at the clone boundary, Cirl-RFP was strictly colocalized with E-cad at adherens junctions and did not extend apically (Fig. 4E, arrow). When present only in cis, Cirl–RFP was shifted above adherens junctions at the clone boundary (Fig. 4, B and D, arrow, and fig. S6B). Thus, at cell-cell interfaces between wild-type and Toll-8 overexpressing cells, where cells are in direct contact, Cirl is specifically localized to adherens junctions in trans (Fig. 4E, arrow) while Cirl is shifted above adherens junctions in cis (Fig. 4D, arrow). These observations indicate that Toll-8 can induce an interfacial asymmetry of Cirl localization at its expression border (Fig. 4G, middle). The trans-interaction between Toll-8 and Cirl did not require the presence of Cirl in cis, since Cirl–RFP was still planar polarized in trans when cirl was removed inside Toll-8 overexpressing clones (Fig. 4F and fig. S6D). In this context, Cirl is still recruited in trans creating an interfacial asymmetry that activates Myo-II (Fig. 3, B and C, and Fig. 4F). Importantly, we showed that asymmetric Cirl interfaces between wild-type and cirl mutant cells activate Myo-II (Fig. 3, D and E). Taken together, we propose that interfacial asymmetry of Cirl induced by Toll-8 is a key signal leading to Myo-II activation (Fig. 4G).
Toll-8 and Cirl form dynamic self-organized polarity patterns
In previous experiments we monitored Toll-8, Cirl and Myo-II at least 24 hours following clone induction. The levels of Toll-8 expression were, as a result, very high. However, Myo-II planar polarization is an intrinsically dynamic and much faster process. To gain further insight into the process as it develops, we investigated the emergence of Toll-8-induced Myo-II activation in a dynamic context. We performed ex vivo live imaging (43) of nascent Toll-8 overexpressing clones in cultured wing discs. We used a temperature-sensitive GAL80 (GAL80ts) to precisely time the onset of Toll-8 expression by a temperature shift to 30°C. After 2h15 at 30°, we isolated wing discs with Toll-8–YFP clones and MRLC–mCherry and imaged them in a thermostatic chamber at 30°C. This allowed us to analyze Myo-II planar polarization as Toll-8 levels were low and as they increased over time. We took advantage of the fact that in this experimental setup Toll-8 expression was not induced synchronously in all cells of a given clone, likely due to stochasticity in GAL80ts inactivation/GAL4 de-repression (Fig. 5A and movies S1 to S3). This led to the generation of dynamic changes in Toll-8 expression creating quantitative stepwise differences in Toll-8 expression levels between neighboring cells (i.e. scalar gradient). This opened the possibility to correlate stepwise differences in Toll-8 expression and Myo-II planar polarization. We found that Myo-II was not only activated at the boundary of Toll-8–YFP expressing cells facing Toll-8–YFP negative cells consistent with previous experiments, but also enriched at interfaces between cells with different Toll-8–YFP levels (Fig. 5A, 110’, cyan arrow, fig. S7A, and movies S1 and S2). The kinetics of Myo-II polarization at the boundary of cells expressing different levels of Toll-8 is in the range of about 10 min in this assay (Fig. 5A’, between 40 min and 50 min, Myo-II becomes polarized at a Toll-8 interface), which is commensurate with Myo-II polarization in embryos. Moreover, as the levels of Toll-8–YFP increased, and once Toll-8–YFP expression reached the same level between these contacting cells, Myo-II enrichment was no longer sustained at these interfaces (Fig. 5A’, arrowheads), and was stabilized only at the boundary of the clone (Fig. 5A, 330min, and fig. S7A). This argues that stepwise differences in Toll-8 expression are sufficient to activate and polarize Myo-II.
Strikingly, between neighboring cells with stepwise differences in Toll-8 expression levels, Toll-8–YFP was initially planar polarized in cells expressing lower levels and tended to accumulate at junctions facing away from the cells expressing higher levels (Fig. 5, A and A’, arrows, and movies S1 to S3). Moreover, in cells where Toll-8–YFP was planar polarized, Myo-II was specifically activated at Toll-8-enriched interfaces, leading to Myo-II planar polarity across several rows of cells (Fig. 5A, 110min, and movies S1 and S2). Therefore, Toll-8 polarization emerged when neighboring cells express different levels of Toll-8 and caused polarization of Myo-II. We then asked how Cirl localization was affected by these nascent polarized patterns of Toll-8. Since Cirl–RFP was too low to be detected in our live imaging setup, we fixed wing discs treated under the same conditions. We found that Cirl–RFP was recruited to the interfaces where Toll-8 was planar polarized (Fig. 5B, arrows, and fig. S7b). We conclude that Cirl forms planar polarized patterns in response to Toll-8 polarization. Finally, we asked if this transient Toll-8 planar polarity requires Cirl. To test this, we induced nascent Toll-8 overexpression clones in wild-type or cirl mutant wing discs by a temperature shift at 30° for 2h15. We found that Toll-8 planar polarity was strongly reduced between cells expressing different levels of Toll-8 in the absence of Cirl (Fig. 5, C and D), showing that Toll-8 and Cirl mutually attract each other in a positive feedback.
In this experimental setup, Toll-8 expression is not induced synchronously in all cells, which results in juxtaposition of cells with differences in Toll-8 levels. The first cell expressing Toll-8 (cell A) recruits Cirl in trans from neighboring cells by depleting it from their orthogonal junctions (Fig. 5E, dashed lines), resulting in Cirl planar polarity in these neighboring cells (Fig. 5E, panel 1). When one of these cells (cell B) initiates expression of Toll-8, Toll-8 expression levels are lower than in cell A. Due to this stepwise difference in Toll-8 expression levels and a preexisting Cirl planar polarity in cell B (where Cirl is depleted from orthogonal interfaces in contact with cell A), Toll-8 in cell B is attracted to the remaining interfaces containing Cirl in trans, which are facing away from cell A (Fig. 5E, panel 2, green line in cell B). These new Toll-8-enriched interfaces will stabilize even more Cirl in trans at the expense of orthogonal junctions in the cells adjacent to cell B that do not express Toll-8, thus propagating Cirl planar polarity one row of cells further away (Fig. 5E, panel 2).
Polarization of Toll-8/Cirl arises from differences in Toll-8 expression levels due to a sequential onset of expression between neighboring cells. We propose that a similar polarization could arise from a stable gradient of Toll-8 expression such as suggested in embryos (22). In such a scenario, Cirl in any given cell is attracted toward the interfaces shared with the neighboring cell expressing the highest Toll-8 level. Cirl enrichment then stabilizes Toll-8 resulting in Toll-8/Cirl planar polarity along the slope of the gradient. In summary, we propose that stepwise differences in Toll-8 expression levels between cells (i.e scalar gradient) translate into self-organized Toll-8/Cirl/Myo-II planar polarity due to local interactions between Toll-8 and Cirl.
Discussion
Here, we identified mechanisms underlying planar polarization of cell contractility by Toll receptors. We discover a direct molecular and functional link between a Toll receptor and the adhesion GPCR Cirl/Latrophilin. We show that Toll-8 asymmetric interfaces recruit Cirl differentially in trans and in cis, thus creating an interfacial Cirl asymmetry that activates Myosin-II. We further show that sequential expression of Toll-8 generates a scalar gradient (i.e. stepwise differences in Toll-8 expression levels), which drives Toll-8/Cirl planar polarity via feedback interactions across cell interfaces: Toll polarizes Cirl and Cirl is required for Toll polarity at the boundary of cells expressing different levels of Toll-8. Through this feedback mechanism a weak asymmetry can be amplified and Toll-8 and Cirl can self-organize to drive planar polarized cell contractility. This process takes about 10 min (Fig. 5A’). Interestingly, we found that various quantitative and spatial patterns of Toll-8 have different outputs. Groups of cells expressing homogeneously high levels of Toll-8 form strong Myo-II cables around them leading to cell sorting that resembles tissue compartment boundaries, while stepwise differences of low levels of Toll-8 expression lead to planar polarity of Myo-II across several cells. Thus, we propose that boundary formation and cell intercalation can be considered as a continuum, depending on quantitative inputs of LRR protein asymmetries interpreted by adhesion GPCRs. The role of GPCRs and G protein signaling in conveying planar polarized input quantitatively is substantiated by the observation that overexpression of heterotrimeric G protein subunits GβGγ increases Myo-II planar polarization in early Drosophila embryos and converts intercalating cell columns into boundary like structures (21).
We show that Toll-8 polarizes MyoII through the polarized localization of Cirl. How this interfacial asymmetry of Cirl activates Myo-II on both sides of the Toll-8 expression boundary remains to be elucidated. Contrary to the adhesion GPCR Flamingo/CELSR (44), Cirl does not require homophilic interaction in trans to localize at the membrane, but interestingly in C. elegans Latrophilin extracellular domains can form stable dimers (45). If extracellular domains of Cirl dimerize in trans, this could inhibit Cirl signaling. The destabilization and asymmetric localization of Cirl induced at the boundary of Toll-8 expression could release Cirl auto-inhibitory dimers and induce signaling. The fact that signaling happens on both sides of the Toll-8 clonal boundaries when Cirl is absent on one side is intriguing. One possibility is that the asymmetry of Cirl extracellular domain can be sensed by another transmembrane protein present on both sides of the clone boundary, which then signals bidirectionally to activate Myo-II. Alternatively, Cirl can sense its own asymmetry, activate Myo-II through its intracellular domain and this signal is mechanically transmitted across interfaces. Thus, Cirl signaling may involve other surface proteins, in particular other GPCRs such as Smog (20), with which Cirl could interact. Cirl expression asymmetry may also induce Rho1 signaling and Myo-II activation without auxiliary proteins in a manner akin to Cadherin2 at the interface between the neural tube and the epidermis in Ciona (46). Cadherin2 homotypic binding recruits a RhoGAP and inhibits Rho1. The Cadherin2 +/-interfaces cannot recruit RhoGAP leading to Rho1 and Myo-II activation specifically at these interfaces. By analogy, Cirl +/-expression asymmetry (Fig. 3D), or localization and functional asymmetry induced by Toll-8 (Fig. 4 and Fig. 5) may modulate Rho1 activation through regulation of RhoGAPs or RhoGEFs.
It was proposed that heterophilic interactions between Toll receptors were required to recruit Myo-II (22). Here we show that asymmetric expression of a single Toll receptor, Toll-8, activates Myo-II independent of other Toll receptors through feedback interactions with the adhesion GPCR Cirl/Latrophilin. We further show that Toll-6 and Toll-2 activate Myo-II at the boundary of their expression domains independent of their cytoplasmic tails (fig. S8, A and C). Toll-6 interacts with Cirl (fig. S8B) while Toll-2 might interact with another binding partner (fig. S8D). Although clonal analysis is not possible in early Drosophila embryos, preventing the dissection of how a Cirl interfacial asymmetry might be generated in this tissue, we can propose two hypotheses on how several Toll receptors and Cirl may interact to define planar polarized interfaces for Myo-II activation in this tissue. On the one hand, as Toll-2, 6 and 8 all have distinct expression patterns and boundaries in the embryo (22), and since Toll-8 alone is capable of polarizing Myo-II in embryos and discs, local Cirl asymmetry could be induced at each boundary independently, thereby allowing polarization on Myo-II at all interfaces on the basis of what we report here with Toll-8/Cirl (14). On the other hand, interactions between different Toll receptors (22) (table S1) might modulate Toll-8/Cirl binding. This could lead to Cirl asymmetry at vertical interfaces across which different combinations of Toll receptors are expressed.
Recently Toll receptors, including Toll-8, were shown to be involved in cell competition (26, 27) and growth (47). Toll-8 is differentially expressed between Myc cells and wild-type cells (27). The interplay between cell competition or tumor growth and mechanics was recently documented (7, 48–50). It is tantalizing to suggest a role for Cirl and Toll-8 in this context.
The depletion of Cirl observed around Toll-8 overexpressing cells resembles the effect of Flamingo (CELSR in vertebrates) overexpression on Frizzled localization (51, 52), both central components of the conserved core Planar Cell Polarity (PCP) pathway (53). Intriguingly, Flamingo is also an adhesion GPCR and shares sequence homology with Cirl (54). It is also required for junctional Myo-II activation in the chick neural plate (9) and in early Drosophila embryos (data not shown). The C. elegans Cirl homologue LAT-1 aligns cell division planes to the embryonic AP axis (55). The vertebrate Cirl homologues Latrophilins regulate synaptogenesis and neuronal migration via trans-heterophilic binding with FLRTs (38–41), which share sequence similarities with Toll-8 (37). FLRT3 plays an important role in tissue morphogenesis (30–32), which might depend on its interaction with Latrophilins considering the evolutionary conservation of the module we characterized in this study. The self-organized planar polarization of Toll-8/Cirl identified in our study suggests the possibility that GPCRs, such as Cirl and Flamingo, evolved different modalities of symmetry breaking. In the core PCP pathway, Flamingo symmetry breaking is thought to be biased by long range mechanical (56) or chemical gradients of adhesion molecules (Fat, Dachsous) or ligands (Wnt) (8), which align Flamingo polarity across the tissue. In the system described here, Cirl symmetry breaking is determined by stepwise transcriptional level differences in Toll-8 that provide an interfacial input through the interdependent polarization of Toll-8 and Cirl proteins. The direction of the transcriptional scalar gradient of Toll-8 defines the orientation of Toll-8 and Cirl polarity. Thus, our work sheds new light on how positional identity is translated directly into vectorial cell interfacial information required for axis extension.
Funding
This work was supported by grants from the ERC (Biomecamorph no. 323027) and the Ligue Nationale Contre le Cancer (Equipe Labellisée 2018). J.L. was supported by the Fondation Bettencourt Schueller and the Collège de France. S.H. was supported by an EMBO Long-Term Fellowship (EMBO ALTF 217-2017) and by a Centuri Postdoctoral Fellowship (Centuri, France). The Marseille Proteomics (IBiSA) is supported by Institut Paoli-Calmettes (IPC) and Canceropôle PACA. Proteomics analysis was supported by the Institut Paoli-Calmettes and the Centre de Recherche en Cancérologie de Marseille. Proteomic analyses were done using the mass spectrometry facility of Marseille Proteomics (marseille-proteomique.univ-amu.fr) supported by IBISA (Infrastructures Biologie Santé et Agronomie), Plateforme Technologique Aix-Marseille, the Cancéropôle PACA, the Provence-Alpes-Côte d’Azur Région, the Institut Paoli-Calmettes and the Centre de Recherche en Cancérologie de Marseille. We acknowledge the France-BioImaging infrastructure supported by the French National Research Agency (ANR–10–INBS-04-01, Investments for the future).
Author contributions
J.L., Q.M. and T.L. conceived the project. Q.M., J.L. and S.K. performed experiments in embryos, Q.M analyzed them. J.L. performed all the experiments and data analysis in fixed wing discs. S.H. performed and analyzed live experiments in wing discs, and quantified data in Fig. 5D. Q.M. and A.L. prepared the samples for the mass spectrometry which was done and analyzed by S.A. and L.C. J-M.P. designed and generated the molecular constructs. J.L., Q.M. and T.L. wrote the manuscript and all authors made comments.
Competing interests
The authors declare no competing financial interests.
Data and materials availability
The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files. Raw image data are available upon reasonable request. Codes and material are available upon request. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD017895.
Supplementary Materials
Materials and Methods
Fly strains and genetics
The following mutant alleles and insertions were used: ind-Toll-8FL::HA (attP40 on 2L), ind-Toll-8ΔLRR::HA (attP40 on 2L), UASp-Toll-8FL::eGFP (attP40 on 2L), UASp-Toll-8ΔLRR::eGFP (attP40 on 2L), UASp-Toll-8ΔCyto::eGFP (attP40 on 2L), UASt-Toll-8::sYFP2 (VK27 (attP9744) on 3R), UASp-Toll-6FL::eGFP (attP40 on 2L), UASp-Toll-6ΔLRR::eGFP (attP40 on 2L), UASp-Toll-6ΔCyto::eGFP (attP40 on 2L), UASt-Toll-6::sYFP2 (VK27 on 3R), UASp-Toll-2FL::eGFP (attP40 on 2L), UASp-Toll-2ΔLRR::eGFP (attP40 on 2L), UASp-Toll-2ΔCyto::eGFP (attP40 on 2L), UASt-Toll-2::sYFP2 (attP40 on 2L), CirlKO (LAT84 and LAT154, gifts from T. Langenhan) (35), Cirl::RFP knock-in (42) (Cirl::RFPKIN, LAT159, gift from T. Langenhan), sqh-Lifeact::mCherry (VK27 on 3R, gift from P.F. Lenne), E-cad::eGFP knock-in (57) (E-cad::eGFPKIN), eve::sYFP2BAC (BAC construct, S2E.MSE.eve.YFP, FBal0279504, gift from M. Ludwig) (58), hs-FLP (59), Ubx-FLP (Bloomington BL 42718), Act5C>STOP>GAL4 (Bloomington BL 4780), FRT42D (Bloomington BL 1802), FRT42D arm-LacZ (Bloomington BL 7372), FRT42D tub-GAL80 (Bloomington BL 9917), tub-GAL80ts (Bloomington BL 7108) and tub-GAL80ts (Bloomington BL 7017). 67-Gal4 ({matαtub-GAL4-VP16}67) is a ubiquitous, maternally supplied GAL4 driver (gift from E. Wieschaus). MRLC is encoded by the gene spaghetti squash (sqh, Genebank ID: AY122159). Sqh was visualized using sqh-Sqh::mCherry (VK18 (attP9736) on 2R or VK27 (attP9744) on 3R for experiments in the wing disc or a construct on chromosome 2 from A. Martin (60) for live-imaging experiments in the embryo) and sqh-Sqh::eGFP transgenics (gift from R. Karess). All fly constructs and genetics are listed in Table S2.
Constructs and transgenesis
Expression vectors drivers
UASp expression vector driver was generated by inserting a PhiC31 attB sequence downstream from the K10 3’-UTR of pUASp. UASt expression vector driver corresponds to pUASTattB which contains a PhiC31 attB sequence inserted downstream from the SV40 3’-UTR (61). ind is an early horizontal stripe expression vector driver generated by modifying the pbphi-evePr-MS2-yellow (62) (gift from T. Gregor) as follows. First, EVEstr2 enhancer sequence was replaced by the ind1.4 enhancer (63) (gift from Gerardo Jiménez). Second, a part of Hsp70Bb 5’-UTR was added after the eve basal promoter. Third, MS2-yellow sequence was replaced by a small polylinker for further cloning.
Expression vectors constructs and transgenics
Toll-8 (Tollo, CG6890), Toll-6 (CG7250) and Toll-2 (18 wheeler, CG8896) whole ORFs were amplified using specific pACMAN genomic plasmids and cloned inside each expression vectors. UASp driven Tolls ORFs were all tagged Cterminally by mEGFP with a GSAGSAAGSGEF flexible aa linker in between.UASp-Toll-8FL::eGFP is the full length Toll-8 (1346aa). UASp-Toll-8ΔCyto::eGFP is a cytoplasmic truncated version of this vector (deletion from aa H1052 to M1346 last aa). In UASp-Toll-8ΔLRR::eGFP, all LRR repeats were removed (deletion from aa E99 to L917). UASp-Toll-6FL::eGFP is the full length Toll-6 (1514aa). UASp-Toll-6ΔCyto::eGFP is a cytoplasmic truncated version of this vector (deletion from aa H1088 to A1514 last aa). In UASp-Toll-6ΔLRR::eGFP, all LRR repeats were removed (deletion from aa A139 to G964). UASp-Toll-2FL::eGFP is the full length Toll-2 (1385aa). UASp-Toll-2ΔCyto::eGFP is a cytoplasmic truncated version of this vector (deletion from aa F1026 to V1385 last aa). In UASp-Toll-2ΔLRR::eGFP, all LRR repeats were removed (deletion from aa F110 to L900). UASt-Toll-2::sYFP2, UASt-Toll-6::sYFP2 and UASt-Toll-8::sYFP2 are Cter sYFP2 tag construct of full length Toll-2,6 and 8 ORFs cloned into UASt using the same GSAGSAAGSGEF flexible aa linker in between. ind-Toll-8FL::HA is a Cter HA tag construct of full length Toll-8 with no linker in between. In ind-Toll-8ΔLRR::HA, all LRR repeats were removed (deletion from aa E99 to L917).
All recombinant expression vectors were built using “In-Fusion cloning” (Takara Bio), verified by sequencing (Genewiz) and sent to BestGene Incorporate for PhiC31 site specific mediated transgenesis into attP40 (2L, 25C7) or VK27 (attP9744, 3R, 89E11). Fully annotated FASTA sequences of all these vectors are available on request.
Antibodies
The following primary antibodies were used: rat-anti-E-Cad (1:200, DHSB DCAD2 concentrate), mouse-anti-β-catenin (1:400, DHSB N2 7A1 Armadillo concentrate), mouse-anti-LacZ (1:100, DHSB 40-1a concentrate), rat-anti-HA (1:100, Anti-HA High Affinity rat IgG1, Roche ROAHAHA). Sqh::eGFP was detected with rabbit-anti-GFP (1:500, Life Technologies A11122 or 1:1000 Abcam ab6556). Cirl::RFP was detected with rabbit-anti-RFP (1:1000, Rockland 600-401-679). The following secondary antibodies were used: donkey-anti-rabbit Alexa Fluor 488 IgG (Life Technologies A 21206), donkey-anti-rabbit Alexa Fluor 568 IgG (Life Technologies A10042), donkey anti-mouse Alexa Fluor 568 IgG (Life Technologies A10037), donkey-anti-mouse Alexa Fluor 647 IgG (Jackson ImmunoResearch 715 605 151) and donkey-anti-rat Alexa Fluor 647 IgG (Jackson ImmunoResearch 712 605 153). All secondary antibodies were used at 1:500.
Affinity Purification Mass Spectrometry
Protein purification and mass spectrometry
Roughly 600 embryos for each sample were collected from overnight cages kept at 25°C for the following crosses: yw (control) or females; 67-GAL4/+; UASt-Toll-8::sYFP2/+; x males ; 67-GAL4/+; UASt-Toll-8::sYFP2/+; (Toll-8::YFP maternal and zygotic overexpression), dechorionated with bleach, transferred directly to lysate buffer (10 mM Tris/Cl pH 7.5; 150 mM NaCl; 0.5 mM EDTA; 0.5% NP-40, supplemented with protease and phosphatase inhibitors) and crushed manually on ice over 30 minutes. Lysates were centrifuged to clear debris and protein concentrations of post-centrifugation supernatants were determined. The crude protein yield per lysate sample is usually 1000∼3000 µg. In each experiment, lysates of comparable protein concentration were incubated with pre-rinsed GFP nano-trap agarose resin (Chromotek, gta-20) at 4°C for 90 min, rinsed 3 x and resuspended in 2x Laemmli buffer with DTT. Protein extraction and purification was performed 3 times each for each cross and verified with silver staining. Protein samples were further purified on NuPAGE 4-12% Bis-Tris acrylamide gels (Life Technologies) and treated with in-gel trypsin digestion (64) with minor modifications. Peptides were harvested with two extractions, first in 5% formic acid and then in 5% formic acid in 60% acetonitrile. Samples were reconstituted with 0.1% trifluoroacetic acid in 4% acetonitrile and analyzed by liquid chromatography (LC)-tandem mass spectrometry (MS/MS) with an Orbitrap Fusion Lumos Tribrid Mass Spectrometer (Thermo Electron, Bremen, Germany) online with an Ultimate 3000RSLCnano chromatography system (Thermo Fisher Scientific, Sunnyvale, CA). A detailed mass spectrometry protocol is available upon request.
Protein identification and quantification
Relative intensity-based label-free quantification (LFQ) was processed using the MaxLFQ algorithm (65) from the freely available MaxQuant computational proteomics platform (66). Spectra were searched against a Drosophila melanogaster database (UniProt Proteome reference, date 2017.08; 21982 entries). The false discovery rate (FDR) at the peptide and protein levels were set to 1% and determined by searching a reverse database. For protein grouping, all proteins that could not be distinguished based on their identified peptides were assembled into a single entry according to the MaxQuant rules. The statistical analysis was done with Perseus program (version 1.5.1.6) from the MaxQuant environment (www.maxquant.org). Quantifiable proteins were defined as those detected in at least 67% of samples in at least one condition. Protein LFQ normalized intensities were base 2 logarithmized to obtain a normal distribution. Missing values were replaced using data imputation by randomly selecting from a normal distribution centered on the lower edge of the intensity values that simulates signals of low abundant proteins using default parameters (a downshift of 1.8 standard deviation and a width of 0.3 of the original distribution). To determine whether a given detected protein was specifically differential, a two-sample t-test was done using permutation-based false discovery rate (pFDR) with a threshold at 0.1% (5000 permutations). The p-value was adjusted using a scaling factor s0=1 (Table S1). In Figure 2, differential proteins are highlighted by a cut-off for log2|Fold change|>2 and a p-value<0.01. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (67) partner repository with the dataset identifier PXD017895.
Bright-field live imaging in embryos
Images of wild-type or mutant embryos were collected on an inverted microscope (Zeiss, AxioVision software) equipped with a programmable motorized stage to record different positions over time (Mark&Find module from Zeiss). Images were acquired every 2 min for 60 minutes from post dorsal movement of the posterior midgut primordium (0 min). The extent of elongation was measured by dividing the travel distance of the posterior midgut primordium at 40 min and normalized to the maximum travel distance.
RNA interference in embryos
dsRNA probes
dsRNA probes were made using PCR product containing the sequence of theT7 promoter (TAATACGACTCACTATAGG) followed by 18-21 nucleotides specific to the gene. The dsRNA probe against Toll-2 (18w, CG8896) is 393-bp long and located in the 5’UTR region (Forward primer: AGTTTGAATCGAAACGCGAGGC; Reverse primer: ATGCCAGCCACATCTTCCA). The dsRNA probe against Toll-6 (CG7250) is 518-bp long and located in the 5’UTR region (Forward primer: TCGAAAATCAGCCAACGTGC; Reverse primer: CGATTCACGGTTTAGCTGCG). The dsRNA probe against Toll-7 (CG8595) is 749-bp long and located in the coding region (Forward primer: TGGCAACCGTCTGGTTACTC; Reverse primer: CGTTCATGATGCTCTGCGTG). The dsRNA probe against Toll-8 (Tollo, CG6890) is 423-bp long and located in the 5’UTR region (Forward primer: CGTTTGTCGTTCAGCGGATG; Reverse primer: CCCCTCATAACCTCCCCGAT) and does not target the ind-Toll-8::HA transgenes. Gel purified PCR products were subsequently used as a template for the in vitro RNA synthesis with T7 polymerase using Ribomax (Promega, P1300). The dsRNA probes were purified using Sure-Clean (Bioline, BIO-37047). Triple dsRNA probes against Toll-2,6,8 and Toll-2,6,7 were prepared and injected at a final concentration of 5 µM each in RNAse-free water.
Embryo injections
Embryos were collected from fresh agar plates in cages kept at 25°C allowed for 30-min egg laying. Embryos were then dechorionated in 50% bleach, rinsed and aligned on cover slips (#1.5) covered with heptane-glue. After a few minutes of desiccation, embryos were covered with Halocarbon 200 oil and injected with dsRNA or RNase-free water. Post-injection embryos were stored at 25°C until live imaging.
Fluorescence live imaging and image processing in embryos
Embryos were aligned on cover slips (#1.5) with heptane-glue and were covered with Halocarbon 200 oil. Dual channel time-lapse imaging was performed on a Nikon Eclipse Ti inverted spinning disc microscope (Roper) with a 100x/1.4 oil-immersion objective at 22°C, controlled by the Metamorph software. Z stacks (step size: 0.5 µm) of 6∼10 slices were acquired every 30 seconds, for 15∼45 minutes starting from stage 6. Laser power was measured and kept constant across all experiments.
To generate 2D projections in experiments with E-Cad::GFP (Fig. 1D and figs. S1 and S2B), a custom FIJI macro (68) integrating the ‘stack focuser’ plugin from M. Umorin was used to perform maximum intensity projection for all channels with 3 Z planes around the junctional plane (labeled by E-cad::GFP). For Fig. 2B, a single plane at the junction level is manually selected based on maximum junctional sqh::GFP signals. The resulting 2D images were subjected to a background subtraction procedure using the rolling ball tool (radius 50 pixels). The 2D images were segmented on E-cad::GFP or LifeAct::Ch channels semi-automatically with manual corrections in the FIJI plug-in Tissue Analyzer (56). The resulting segmentation masks were then dilated by 5 pixels on either side of the junction and used as masks for subsequent quantifications.
Immunofluorescence and image processing in embryos
Embryos were fixed with 8% formaldehyde for 20 min at room temperature. Embryos were processed and stained according to standard procedures (69). Embryos were mounted in Aqua-Polymount (Polysciences). Images were acquired on a Leica SP8 inverted confocal microscope with a 63x/1.4 NA oil-immersion objective (with exception of fig. S4, D and E, acquired on a Zeiss LSM780 with a 63x/1.4 NA oil-immersion objective). Z stacks with step size of 0.25-0.4 μm were collected.
2D images were generated by maximum intensity projections followed by the same procedure as for live imaging experiments in embryos (except for 3-pixel dilatation in segmentation masks generated from β-catenin stainings).
Clonal analysis in wing discs
Flies were allowed to lay eggs in vials for ∼8h at 25°C and vials were kept at 25°C until heat-shock. For clonal overexpression of Tolls (Fig. 1G, Fig. 2H, Fig. 4A and figs. S6A and S8) 72h AEL (after egg laying) old larvae were heat-shocked at 37° for 12 minutes and dissected after 24h. For GAL80ts experiments (Fig. 5 and fig. S7), 72h AEL larvae were heat-shocked at 37° for 12 minutes, kept at 18° for 48 hours and subsequently incubated at 30°C for 2h15min in order to inactivate GAL80ts and allow expression of Toll-8::YFP.
For MARCM experiments (Mosaic Analysis with a Repressible Cell Marker, Fig. 3, A and B, Fig. 4, B to F, fig. S5 and fig. S6, B to D), 72h AEL larvae were heat-shocked at 37° for 1h, kept at 18° and heat-shocked again 7 hours later at 37° for 1h. Larvae were kept at 18° for 20 hours, shifted to 25° and dissected 24h later. Keeping the larvae at 18° allowed growth of the clones in the presence of no/low levels of Toll-8 expression. Larvae to observe cirl mutant clones (Fig. 3B) were treated the same way.
Immunofluorescence and image processing in wing discs
Staged larvae were dissected in PBS, transferred to 4% PFA in PBS and fixed under agitation for 18 min at room temperature. After fixation, wing discs were first rinsed with PBS, then extensively washed with PBT (PBS plus 0.2% Triton-X100) and blocked in PBT with 5% normal donkey serum (NDS, Jackson Immuno Research Laboratories, 017-000-001) for at least 30 min at room temperature, followed by incubation with primary antibody in 2% NDS overnight at 4 °C. The next day wing discs were washed in PBT and incubated in secondary antibody with 2% NDS for 1h30min at room temperature. After six rounds of washes with PBT, samples were mounted in Mowiol (Sigma-Aldrich, 324590). Larval mouth hooks were used as spacers in the experiments where Myo-II was observed. Images were acquired on a Leica SP8 inverted confocal microscope with a 63x/1.4 NA oil-immersion objective. Toll-8::YFP and Sqh::Ch were visualized with their endogenous fluorescence. Image stacks with step size of 0.25-0.5 μm were collected.
Peripodial signal was masked from the image stacks in ImageJ to avoid interference with signals from the wing disc proper. 2D projections were generated using the aforementioned custom stack focuser macro in ImageJ, projecting two z planes around the junctional plane of each cell (detected by E-cad staining, except Fig. 3, A and B, projected on Sqh::Ch signals). This allows to project the entire wing pouch independently of the shape of the wing disc. The 2D-projected stacks were then segmented on E-cad stainings (except Fig. 3, A and B, segmented on Sqh::Ch signals) using Tissue Analyzer (56).
Ex vivo live imaging and image processing in wing discs
The culture medium used for long-term time lapse imaging of wing imaginal disc explants is described in Dye et.al. (43). In short, Grace’s insect medium (Sigma G9771, without sodium bicarbonate) was buffered with 5mM BisTris and the pH adjusted to 6.6-6.7. Subsequently the medium was sterile filtered (0.2µm pore size) and kept at 4°C for up to 4 weeks. At the day of the experiment the medium was supplemented with 5% fetal bovine serum (FBS), Penicillin-Streptomycin (final 1x from a 100x stock solution, Sigma P4333) and warmed to 30°C in a water bath. Just before dissection of the larvae, 20-Hydroxyecdysone (20E, Sigma, H5142) was added to yield a total concentration of 20nM. 20E was kept as a 1000x stock solution in ethanol at −20°C. For the experiment, 72h AEL larvae were heat-shocked at 37° for 12 minutes, kept at 18° for 48 hours and subsequently incubated at 30°C for 2h15min in a water bath. Subsequently, larvae were floated out of the food using 30% glycerol and washed in sterile water twice. Surface sterilization in 70% Ethanol was followed by another wash in sterile water and then in medium. Larvae were dissected in culture medium, wing discs isolated and mounted on a round cover slip. In order to restrict disc movement during imaging, discs were covered by a porous membrane (Whatman cyclopore polycarbonate membranes; Sigma, WHA70602513) using two stripes of double-sided tape as spacers. Finally, this sandwich was mounted in an Attofluor cell chamber (A7816, Invitrogen) and filled with 1ml of medium and covered with Parafilm M (P7793, Sigma-Aldrich) to avoid evaporation. Discs were imaged on a Nikon Eclipse Ti inverted spinning disc microscope (Roper) equipped with an incubation chamber heated to 30°C. Imaging was done using a 60x/1.2 NA water-immersion objective (Fig. 5A, fig. S7A and movies S1 and S2) or a 100x/1.4 NA oil-immersion objective (Movie 3 and 4). Dual imaging of Toll-8::YFP and Sqh::Ch was performed by simultaneous excitation of fluorophores with 515nm and 561nm laser lines using a dichroic mirror to image on two cameras (Evolve 512, Photometrics). Stacks of 40 slices with 0.7µm spacing were acquired every 10min (60x movies) or every 5min (100x movies).
A maximum projection of the disc proper junctional plane was obtained by masking the peripodial epithelium and the lateral portion of the disc proper manually in ImageJ based on sqh::Ch signals. Background subtraction was done using a rolling ball (50px radius) in ImageJ.
Data analysis
Definition of expression interfaces
When references channels (Toll-8::HA, Toll-8::YFP, Toll-8::GFP, or LacZ staining) were available, expression interfaces were defined from reference channels in Tissue Analyzer.
To define horizontal cell rows in Fig. 1D and fig. S2B, cell rows were counted from the ventral midline, with the 4th cell row (most ventral) being 2 rows away from the ventral midline. The border between the 2nd and the 3rd cell rows is consistent with the position of ind ventral expression border.
To define vertical cell columns in fig. S1, parasegment boundaries were visualized with Eve::YFP, with the anterior boundary of Eve::YFP signal defined as the parasegment boundary between even- and odd-numbered parasegments. Thus, cell columns 1-4 belong to odd-numbered parasegments (Eve::YFP+), while 5-8 belong to even-numbered parasegments (Eve::YFP-).
Quantification of junctional Myo-II intensities
Raw pixel intensities from segmented junctions were measured in Tissue Analyzer. To extract data tables containing raw pixel intensities from Tissue Analyzer, a customized R procedure was developed using the RSQLite package. Adjusted junctional pixel intensities were obtained by subtracting mean cytoplasmic intensity value measured on each image. Enrichment was calculated as ratios of adjusted junctional intensity values between junctions of interest and those in nearby wild-type tissues (fig. S2A).
Quantification of border smoothness
Boundary smoothness for ventral ind expression border in the embryo was calculated as the ratio between distance between two end vertices over total junctional length (fig. S2A, left). Border smoothness value approaches 1 as the border gets smoother.
For clone smoothness in the wing disc, an original method developed by P. Villoutreix (Centuri, France) was implemented in Tissue Analyzer by B. Aigouy (IBDM, France) under the plugin ‘Clone wiggliness’. In brief, the boundary of the clone was extracted, the vertices present at the clone boundary were ordered, and an angle was calculated for each vertex with its two neighboring vertices present at the clone boundary (fig. S2A, right). A mean value per clone was then calculated and this value is getting closer to 180° if the clone is smooth.
Quantification Toll-8 planar-polarity (Fig. 5D)
We restricted our analysis to single cells expressing low Toll-8::YFP levels in the vicinity of a high Toll-8::YFP expressing cell. This ensured that each junction included in our computation only contained Toll-8::YFP originating from the single cell that was quantified. For each single cell, junctional Toll-8::YFP levels were extracted (using the segmented line tool in ImageJ, line width = 6px) along the parallel junctions (Toll-8‖, the junctions not being in contact with the high Toll-8 expressing cell) and along the two junctions being in direct contact with high Toll-8 expressing cell (Toll-8Ʇ). Toll-8::YFP enrichment at parallel junctions was computed by calculating the ratio between mean Toll-8‖ and mean Toll-8Ʇ levels (Fig. 5D, scheme).
Data visualization
With the exception of fig. S7A, data visualization was performed in R with customized scripts. The following custom packages were used: “fields” and “ggplot2”. Intensity profiles shown in fig. S7A were plotted in Phython using the Seaborn library.
Box-plot elements are defined as follows: center line represents the median; box limits represent the first and third quartiles (the 25th and 75th percentiles); upper whisker extends from the box limit to the largest value no further than 1.5x interquartile range from the box limit and lower whisker extends from the box limit to the smallest value at most 1.5x interquartile range from the box limit; all data points are plotted on the graph.
Statistics
All P values were calculated using the Mann-Whitney U test in R. No statistical method was used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.
Repeatability
All measurements were performed in 8–100 independent experiments. Each embryo and clone in the wing disc is considered as an independent experiment.
Table S1. List of proteins associated with Toll-8 (in green) selected using pFDR at 0.1% and s0=1. The proteomes of embryos overexpressing Toll-8–YFP maternally and zygotically (N = 3 protein isolations: T3, T4, T5, each with 3 replicas: T3a, T3b, T3c, …) were compared with yw embryos (N = 3 protein isolations: C3, C4, C5, each with 3 replicas: C3a, C3b, C3c, …) with affinity purification mass spectrometry (PRIDE dataset identifier PXD017895). Protein list is displayed by order of Difference log2(T/C). Rows highlighted in green denote significant hits from a two-sample t-test using permutation-based false discovery rate (pFDR) controlled at 0.1% (5000 permutations).
Table S2. List of genotypes employed in the experiments in the indicated figure panels.
Movie S1. Time lapse movie of a wing disc showing live dynamics of Toll-8–YFP (green) and MRLC–Ch (magenta) in a nascent Toll-8–YFP overexpressing clone. Toll-8–YFP starts being detectable at 0 min and its levels increase during the course of the time lapse. Asterisks indicate a cell that dynamically upregulates expression of Toll-8–YFP during ∼50min. Myo-II is being activated at junctions that show Toll-8–YFP planar polarity (arrow) and being deactivated once Toll-8–YFP expression reaches similar levels between contacting cells (arrowheads). Scalebar: 5 μm.
Movie S2. Time lapse movie of a wing disc showing live dynamics of Toll-8–YFP (green) and MRLC–Ch (magenta) in a nascent Toll-8–YFP overexpressing clone. Scalebar: 5 μm.
Movie S3. Time lapse movie of a wing disc showing live dynamics of Toll-8–YFP (green) and MRLC–Ch (magenta) in Toll-8–YFP overexpressing clones. Scalebar: 5 μm.
Acknowledgments
We thank all members of the Lecuit team, B. Aigouy (IBDM, France) for stimulating discussions during the course of this project, the IBDM imaging facility for microscopy assistance, FlyBase for maintaining curated databases and the Bloomington stock center for providing fly stocks. We thank T. Langenhan and N. Scholz (Leipzig, Germany) for sharing information about Cirl/Latrophilin and fly reagents, and for stimulating discussions. We thank B. Aigouy (IBDM, France) and P. Villoutreix (Centuri, France) for developing the method for quantifying clone smoothness in wing discs. We thank B. Habermann (IBDM, France) for her valuable guidance on performing pair-wise alignment between Toll-8 and FLRTs. We are grateful to M. Ludwig (Birmingham, UK), P.F. Lenne (IBDM, France), E. Wieschaus (Princeton, USA), A. Martin (MIT, USA), R. Karess (IJM, France), G. Jiménez (IBMB, Spain) and T. Gregor (Pasteur Institute, France) for the gift of flies and vectors.
Footnotes
Introduction and discussion detailed; description of Figure 4 and Figure 5 clarified; Some experiments added.