Planar polarity has been linked to tumorigenesis in several organisms, tumor models, and patient tumor samples (Kaucká et al., 2013; Asad et al., 2014; Puvirajesinghe et al., 2016; VanderVorst et al., 2018; Chen et al., 2021b; Li et al., 2021; Chen et al., 2021a). We explored how the core PCP complex might be involved in tumorigenesis. To do so, we used the versatile and widely used Ras^V12 tumor model in Drosophila eye imaginal discs. In this model, eye disc cells are transformed into highly tumorigenic cells by co-expressing the constitutively active Ras^V12 oncoprotein and RNAi against the tumor suppressor Scribble (Scrib) (Pagliarini and Xu, 2003). Cells expressing these genes quickly expand into massive tumors and invade the neighboring brain tissue (Figure 1A–C).

To study the requirement for PCP signaling in tumors, we expressed Ras^V12 and scrib RNAi throughout the eye, and simultaneously downregulated PCP core proteins by also expressing validated RNAis against fmi, fz, dsh, vang, and pk. We then imaged tumor size through the pupal cuticle. We observed that tumor growth was unaffected by knockdown of any of the core PCP genes at third instar larval or at early pupal stages (Figure 1A–H). We then asked whether PCP signaling might be required by tumors when they’re confronted by wildtype cells. To test this hypothesis, we created clonal tumors in the eye disc by expressing Ras^V12 and scrib RNAi under control of ey-Flp. In this model, nearly every cell in the eye disc will perform chromosomal recombination (Figure 1—figure supplement 1), producing an eye disc composed of approximately 50% wildtype and 50% tumor cells. Under these conditions, tumor cells expand aggressively, displacing wildtype cells and causing massive overgrowth of the eye disc and metastasis to the larval brain (Figure 1I–K). We created clonal tumors that were also null for either fmi, fz, dgo, vang, or pk and evaluated them for their ability to grow and metastasize when confronted with wildtype cells (Figure 1I–P). Remarkably, only fmi interfered with tumorigenesis, restraining growth of tumors in the eye disc and preventing them from invading neighboring tissues (Figure 1L). To ensure that the growth inhibitory effect of loss of Fmi in clonal tumors but not in whole eye tumors was not a result from comparing a null allele in clonal tumors vs RNAi in whole eye tumors, we generated fmi RNAi clonal tumors and compared them to control RNAi clonal tumors. While the effect was not as dramatic as with null clones, the size of fmi RNAi clonal tumors was substantially reduced compared to that of clonal tumors expressing w RNAi (Figure 1—figure supplement 1A and B). Therefore, the same partial level of fmi knockdown impairs clonal tumor growth but not growth of whole eye tumors.

We also wished to directly compare growth of fmi-null clonal and fmi-null whole eye tumors. Because fmi alleles are embryonic lethal, we generated fmi-null tumors while eliminating WT cells from the eye disc using the GMR-Hid, l(2)CL system (Stowers and Schwarz, 1999). While fmi-null clonal tumors showed restrained growth (Figure 1L), elimination of WT cells in eye discs removed competition, and fmi-null tumors grew to a size comparable to control whole eye tumors (Figure 1—figure supplement 1C and D). Because the effect of fmi was only apparent when tumors were induced in juxtaposition to wildtype cells, we hypothesized that Fmi may only be required in tumors that undergo cell competition.

Molecular mechanisms used by tumors in cell competition are sometimes shared by developmental cell competition. Malignant cells have been shown to actively eliminate surrounding cells by mechanical cell competition, apoptosis, and engulfment (Levayer et al., 2016; Kohashi et al., 2021), similar to the cell death-mediated elimination of losers during developmental cell competition and EDAC (Maruyama and Fujita, 2017; Parker et al., 2021) Induction of JNK-mediated cell death is also common to developmental competition and tumor cell competition, as shown in intestinal adenomas (Suijkerbuijk et al., 2016). We hypothesized that the requirement of Fmi by competing tumors might also be shared in developmental cell competition. We therefore evaluated the role of Fmi in non-malignant, developmental cell competition models.

Cells expressing high levels of Myc behave as winners when growing among wildtype cells, proliferating faster than wildtype cells and inducing their elimination. Despite Myc being a proto-oncogene, in contrast to Ras^V12 scrib RNAi tumors, Myc overexpressing clones do not become tumors, but instead expand to occupy larger than normal domains of morphologically normal tissue (Moreno and Basler, 2004; de la Cova et al., 2004; Clavería et al., 2013). As for the Ras^V12 scrib RNAi tumor model, we used UAS-myc and ey-Flp to create eye discs comprising initially equal populations of Myc overexpressing (>>Myc) and wildtype cells. As expected, >>Myc cells behaved like supercompetitors, and the majority of cells (~65%) in late third instar discs were >>Myc cells (Figure 2A). However, when >>Myc cells were depleted of fmi, they failed to outcompete wildtype cells and behaved as one would expect of losers; the >>Myc, fmi^-/- cells were outcompeted by wildtype cells such that late third instar eye discs were composed of only ~35% >>Myc, fmi^-/- cells, suggesting that wildtype cells behave as winners against >>Myc, fmi^-/- cells (Figure 2B and D). We then tested whether Fmi had an effect in the losers during >>Myc competition. Removing Fmi from the WT losers had no effect on this competition, as >>Myc clones did not compete more successfully than when confronting wildtype cells (Figure 3C and D). Importantly, loss of Fmi alone did not induce competition, as in eye discs comprising initially equal populations of fmi^-/- and wildtype cells the fmi mutant cells grew equally well as the wildtype cells (Figure 2—figure supplement 1A–C).

If the requirement of Fmi is a general feature of cell competition, it should be observable in other tissues. We therefore assessed whether depleting Fmi in wing disc >>Myc winner clones would also reverse the competition outcome. We used hsp-Flp to generate twin spot clones expressing UAS-Myc, and as in eye discs, those clones behaved like supercompetitors. Myc clones were on average 1.7 times larger than the homozygous RFP+ (hRFP) twin spots (Figure 2E). However, when >>Myc clones lacked Fmi, their ability to compete was severely impaired, producing much smaller clones, being on average half as large as their hRFP twin spots, and thus, like in eye discs, behaved as one would expect of losers when competing against wildtype cells (Figure 2F–H). Interestingly, fmi^E59 Myc clones were also highly fragmented in the wing disc, a phenomenon we did not observe in eye discs. As observed in eye discs, removing Fmi from the wildtype losers had no effect on the ability of >>Myc clones to compete (Figure 3G and H). Similarly, making clones mutant for fmi alone had no effect on their growth in wing discs (Figure 2—figure supplement 1D–F).

In the setting of tumorigenesis, either tumor cells or wildtype cells may emerge as winners of cell competition. Tumors outcompete wildtype cells in the processes of invasion and metastasis, whereas transformed pre-malignant cells are often eliminated by wildtype cells through competition-dependent defense mechanisms such as EDAC (Kon et al., 2017; Kon and Fujita, 2021). We therefore further explored the potential requirement for Fmi in a system in which both outcomes occur at different times. In eye discs, cells lacking Scrib display a proto-oncogenic behavior, losing polarity and overproliferating in the developing eye disc. However, as eye development progresses, they subsequently become losers, and are eliminated via JNK-mediated apoptosis during late third instar and pupal development so that they are virtually absent from the tissue before the fly ecloses (Brumby and Richardson, 2003).

Using the ey-Flp system to activate RNAi against scrib in eye discs, we confirmed that downregulation of scrib generated clones that perdure through larval development (Figure 3D , E, and M) but were eliminated from the tissue by the end of pupal development and were not detected in the adult eye (Figure 3F and F’). Compared to control RFP clones, which represented around 40% of the eye cell population in early third instar larva (Figure 3A, B, and M, Figure 3—figure supplement 1A) and remained at roughly that proportion in the adult eye (Figure 3C and C’), scrib RNAi clones began to be eliminated at or before the time the morphogenetic furrow progresses (Figure 3E and M, Figure 3—figure supplement 1B), and were almost completely absent in adult eyes (Figure 3F and F’). Their elimination left scars in the adult eye, likely because the structure of the eye is established with the passing of the morphogenetic furrow, and the compound eyes were smaller compared to wildtype eyes (Figure 3C and F). Wildtype cells therefore behave as winners when confronted with scrib RNAi clones during late larval and pupal development.

We then evaluated the effect of removing Fmi from the wildtype winner clones in scrib cell competition. Loss of fmi in wildtype cells had little impact on the size of scrib RNAi clones in third instar larval discs (Figure 3G, H, and M, Figure 3—figure supplement 1C), but as development progressed, wildtype winner cells depleted of fmi lost their ability to compete with and eliminate scrib RNAi clones; the scrib RNAi clones survived during pupal development and indeed constituted the majority of the adult eye (Figure 3I and I’), showing a reversal in the competition outcome much as the outcome of >>Myc competition is reversed when the winners lack Fmi. The surviving scrib RNAi cells failed to differentiate, and instead produced large scars in the adult eye (Figure 3I and I’). We also observed considerable lethality in this population of flies, presumably due to the uncontrolled proliferation of scrib RNAi cells that could not be eliminated.

During >>Myc supercompetition, removing Fmi from the loser wildtype cells showed no effect in cell competition in eye and wing discs (Figure 2C and G). We therefore considered what might happen if scrib RNAi loser clones are made to lack Fmi. However, this situation is more complex than the >>Myc competition, since scrib RNAi clones have been previously documented to initially overproliferate, perhaps behaving as winners, before ultimately becoming losers (Brumby and Richardson, 2003). We were unable to quantify this early proliferation of scrib RNAi eye clones before they are eliminated by their wildtype neighbors, as the discs are too small for us to dissect and count. However, assuming that they are behaving as winners in early larval development, one might expect that removing Fmi from scrib RNAi clones would impair their ability to overproliferate early, such that by third instar, they would be smaller than scrib RNAi clones with intact Fmi. Indeed, this is what we observed (Figure 3J, K, and M, Figure 3—figure supplement 1). While some portion of their smaller size relative to control clones is attributed to their elimination by wildtype winners as described above, the remainder may be a result of the loss of their ability to overproliferate earlier in larval development.

The small population of fmi^E59 scrib RNAi clones remaining at third instar was almost completely eliminated by the end of larval development, such that the adult eyes were indistinguishable from wildtype eyes (Figure 3L and L’). We hypothesize that the lack of fmi in scrib clones hinders their ability to overproliferate in early larval stages, resulting in smaller clones that are quickly eliminated by their wildtype neighbors, allowing compensatory differentiation to generate a morphologically normal eye.

Our observations suggest that lack of fmi renders winner clones and tumors incapable of outcompeting the neighboring tissue in multiple cell competition scenarios. Cell competition relies mainly on two mechanisms to allow winner cells to take over the tissue when confronting less fit cells: faster proliferation than losers and elimination of loser cells by induced apoptosis or extrusion (Amoyel and Bach, 2014; Maruyama and Fujita, 2017; Morata, 2021). We therefore explored how the lack of fmi in tumors and winner >>Myc cells during competition affected both mechanisms.

Drosophila control Ras^V12, scrib RNAi eye tumors trigger apoptosis in the neighboring cells, as detected by Dcp1 staining (Figure 4A–D), consistent with previous reports (Karim and Rubin, 1998; Pérez et al., 2017) and with their highly invasive behavior when surrounded by wildtype cells. However, when fmi was removed from the clonal tumors, the outcome of this competition was reversed, as described above, and instead, fmi^-/-, Ras^V12, scrib RNAi tumors displayed excess apoptosis compared to the surrounding wildtype tissue (Figure 4E–H). Moreover, we found cell debris from fmi-deficient tumor cells inside lysosomal vesicles in wildtype cells, suggesting that wildtype cells are clearing neighboring loser tumor cells through engulfment (Figure 4—figure supplement 1). We then tested whether this reversal of apoptosis burden was specific to the tumor model, or if it is a mechanism shared in other cell competition models. We counted apoptotic cells in WT and >>Myc clones in the eye disc only when these cells were located near to the clone boundary (1–3 cells from the boundary, see Materials and methods for details). We observed that control >>Myc clones showed low levels of apoptosis and similar levels in their neighboring WT cells (Figure 4I–L; p-value 0.6049), in line with previous results (de la Cova et al., 2004), whereas, consistent with the tumor model, apoptosis was significantly increased in fmi^E59, >>Myc clones but not in their WT neighbors (Figure 4M–P, p-value 0.0006), likely contributing to the reversal in competition we previously observed (Figure 2A–D).

JNK signaling plays a prominent role during cell competition. Previous reports have shown that activation of JNK mediates engulfment and elimination of scribble clones during cell competition (Ohsawa et al., 2011; Yamamoto et al., 2017). However, JNK signaling is also responsible for increased tumor growth and invasion in Ras-activated cells or scrib mutants (Igaki et al., 2006; Uhlirova and Bohmann, 2006; Leong et al., 2009). We therefore asked whether the role of Fmi in tumorigenesis and cell competition could be related to the regulation of JNK signaling. Confirming previous observations, we detect activation of JNK signaling via the puckered lacZ reporter puc^E69 both in Ras^V12, scrib RNAi tumors (Figure 4A–C) and scrib RNAi clones in eye discs (Figure 4—figure supplement 2A–C). However, Ras^V12 tumors lacking fmi, despite displaying cell death in tumor cells competing with wildtype neighbors, show no change in Puc activation (Figure 4E–G). Similarly, scrib RNAi clones activate JNK signaling independently of fmi (Figure 4—figure supplement 2D–F). We therefore conclude that Fmi acts independently of JNK signaling.

Cell competition does not rely solely on apoptosis to eliminate loser cells. An increased proliferation rate of winners is also a hallmark of cell competition (Morata, 2021; Madan et al., 2022). To evaluate whether Fmi is also required to maintain a higher proliferative ratio in winners, we quantified mitotic cells in winner clones with and without Fmi. The wing disc displays distributed cell divisions throughout larval development, in contrast to eye discs, where passing of the morphogenetic furrow limits proliferation as cells start differentiating into compound eye cells (Wolff and Ready, 1991; Tsachaki and Sprecher, 2012). Therefore, for the quantification of proliferation, the wing is a better model than the eye disc. When wing disc >>Myc clones were depleted of Fmi, cell proliferation, as measured by pHis3 staining, was significantly reduced (Figure 4Q–S). Taken together, these observations directly link the need for Fmi to proliferation and induced apoptosis, the two key events of cell competition, in several models of cell competition.

Previously, some key players involved in cell competition, including Flower (Rhiner et al., 2010) and Xrp-1 (Baillon et al., 2018), were shown to be either up- or downregulated during competition. We evaluated whether Fmi protein levels might also be affected during competition. Fmi is ubiquitously expressed at low levels in Drosophila larva, pupa, and adult (Brown et al., 2014). We stained third instar larval wing discs with >>Myc clones to determine whether >>Myc supercompetition affects the levels of Fmi protein at the membrane, either inside the clone or near the boundary where cell competition occurs. We observed no change in Fmi protein levels (Figure 5—figure supplement 1), suggesting that Fmi is involved in competition through a mechanism that does not rely on altering protein levels.

Cell competition is thought to rely on intercellular communication to compare fitness and determine the outcome between prospective winner and loser cells (Kon et al., 2017; Baker, 2020; Ogawa et al., 2021). If Fmi is involved in these determinative intercellular communication events by signaling as a trans-homodimer, it should be required in both prospective winner and loser cells. Our previous eye and wing disc >>Myc supercompetition results already suggested this is not the case. While the removal of fmi caused winner clones to effectively become losers during >>Myc competition (Figure 2), removing fmi from the losers had no effect on the losers’ outcome, either in eye discs (Figure 2C and D) or wing discs (Figure 2G and H). In both tissues, winner >>Myc clones showed no change in relative size to their twin spot counterparts, nor were loser clones eliminated more effectively (Figure 2D and H).

Our results so far suggested that the effect Fmi exerts on cell competition does not operate through bidirectional intercellular PCP communication. To further consolidate this hypothesis, we examined the effect of removing a dedicated core PCP protein other than Fmi from >>Myc supercompetitors. Generating vang^A3 >>Myc clones in the wing disc, we observed that the >>Myc clones were unaffected and remained winners (Figure 5). vang^A3, >>Myc clones were on average 2.2±0.82 times larger than their hRFP twins, not significantly different from the 1.7±0.45-fold value for >>Myc clones (Figure 2H; Figure 5C).

Our results have so far demonstrated that Fmi is required only in winner cells and that its function in cell competition is independent of PCP. Furthermore, our results rule out a role for Fmi homodimers in directly communicating fitness information between prospective winner and loser cells. However, the possibility remained that Fmi might function through trans-homodimerization between prospective winner cells to sustain winner status via a signal, via adhesion, or both. To ask if Fmi fulfills its role by mediating adhesion, we tested whether we could restore winner status to >>Myc clones that lack Fmi by providing a transgenic fmi that lacks the nine cadherin domains of Fmi (arm-fmiΔCad).

Co-expression of arm-fmiΔCad in fmi^E59, >>Myc clones re-established the ability of these clones to outcompete their neighbors, and they again became supercompetitors (Figure 6). When comparing the ratio of >>Myc clone cell number to twin spot cell number, the presence of arm-fmiΔcad restores competition to a level comparable to >>Myc supercompetitors (1.43±0.48 GFP/hRFP cells vs 1.71±0.45 GFP/hRFP cells respectively), far above the 0.49±0.29 value for fmi^E59 >>Myc clones compared to twin spots.

Supercompetitor clones mutant for fmi consistently show clone fragmentation (Figures 2F, 4R, and 6B). Despite lacking the cadherin repeats, rescued clones not only restored their supercompetitor status, but fully recovered the ability of >>Myc clones to remain cohesive (Figure 6C). This suggests that the clone fragmentation we observed is unlikely due to the ability of Fmi to contribute to adhesion and more likely caused by a feature of the competition between wildtype cells and fmi^-/->>Myc loser clones that causes their elimination.

Taken together, these results demonstrate that, while Fmi is essential in winner cells to eliminate less fit neighbors, this effect is independent of PCP or other homodimer-mediated signaling, and independent of Fmi-mediated cell adhesion, suggesting instead an as yet uncharacterized function for Fmi.

Several arguments support the conclusion that the role for Fmi in cell competition is distinct from its role in PCP signaling. First, Fmi is the only core PCP component among the six that were surveyed to inhibit the ability of Ras^V12 scrib RNAi tumor clones to compete in the eye. If Fmi’s role in cell competition were to signal winner fate to losers and vice versa by mirroring its role in PCP signaling, where it signals the presence of proximal (Vang, Pk) components in one direction and the presence of distal components (Fz, Dsh, Dgo) in the other direction between adjacent cells (Lawrence et al., 2004; Strutt and Strutt, 2007; Strutt and Strutt, 2008; Chen et al., 2008; Struhl et al., 2012), then one might expect either the proximal or distal components to also be required in winner clones. No such requirement was observed. Second, in PCP signaling, Fmi functions as a trans-homodimer to transmit those signals and requires the cadherin repeats and other extracellular domains, implying its function as a trans-homodimer (Kimura et al., 2006). In PCP signaling, removing Fmi from either of two adjacent cells completely blocks PCP signaling. In contrast, in cell competition, while Fmi is required in winners, removing Fmi from losers has no effect on competition. Thus, the model of bidirectional signaling via Fmi is not supported by our results.

These observations, however, do not rule out the possibility that Fmi trans-homodimers contribute to intercellular signaling among winner cells. Nonetheless, as discussed above, adhesion seems not to be a meaningful part of the function for Fmi in winners, as an adhesion-deficient Fmi construct (fmiΔCad) fully rescues both competition and clone adhesion. The potential contributions of adhesion vs other possible signaling mechanisms are discussed at more length below.

Unlike the other core PCP genes, fmi^-/- mutations are lethal due to requirements in the nervous system (Usui et al., 1999). Though not fully characterized, its roles in the nervous system appear to be distinct from PCP signaling. Fmi is required for outgrowth and guidance of the R8 axon of the eye to the M3 layer of the medulla via a mechanism that appears independent of other components of the core PCP signaling pathway (Gao et al., 2000; Lee et al., 2003; Senti et al., 2003), but does interact with Golden goal, a transmembrane phosphoprotein that is not associated with PCP signaling (Takechi et al., 2021). Growth of the dendrites of dorsal da neurons is also regulated by Fmi. During embryogenesis, da dendrites in fmi mutant embryos emerge precociously and overgrow as they approach the dorsal midline, and later, during larval growth, dendrites from opposite sides fail to avoid each other (tile) and instead overlap (Gao et al., 2000; Sweeney et al., 2002; Kimura et al., 2006).

Fmi is classified as an atypical cadherin and a Class-B adhesion G protein-coupled receptor (AGPCR), as it contains in its extracellular domain several conserved functional domains including cadherin repeats, epidermal growth factor-like repeats, laminin A G-type repeats, and a GPCR autoproteolytic inducing (GAIN) domain that contains within it a GPCR proteolytic site (GPS) (Rosa et al., 2021; Einspahr and Tilley, 2022; Sreepada et al., 2022). These extracellular domains are followed by seven transmembrane domains and an intracellular C-terminal domain. Although much remains to be learned about this large subfamily of GPCRs, a general model has emerged in which activation by membrane bound or extracellular protein, peptide, proteoglycan or small molecule ligand, or mechanical force exposes a tethered ligand at the N-terminus of the C-terminal fragment in the GAIN domain that, upon exposure, interacts with the transmembrane portion to activate a G-protein signaling cascade. In many but not all AGPCRs, the tethered ligand is exposed by GAIN domain-mediated autoprotolysis of its GPS. Non-cleaved AGPCRs are hypothesized to expose the tethered ligand by an allosteric conformational change. Some de-orphanized AGPCRs interact with multiple ligands, and ligand binding can result in partial or full activation. In some cases, it appears that engineered truncation of portions of the extracellular domain can produce some level of ligand-independent activation (Rosa et al., 2021).

When expressed in da neurons, an Fmi construct lacking the cadherin, laminin G, and EGF-like repeats partially rescued the embryonic dendritic overgrowth phenotype (Kimura et al., 2006), suggesting a function independent of homodimerization. The G protein Gαq (Gq) has been proposed to function downstream of Fmi to mediate this repressive function (Wang et al., 2016). A recent preprint reports the resolved structure of the CELSR1 extracellular domain, showing that the protein has two distinct domains: an adhesion domain comprised of the first eight cadherin repeats, and a compact domain that extends from the ninth cadherin repeat to the transmembrane domains that is involved in GPCR signaling. Indeed, they demonstrated that a CELSR1 construct lacking only the cadherin repeats 1–8 retains the ability to activate Gαs, which has been predicted to interact with CELSR1 (Bandekar et al., 2024). Notably, our results showed that a similar Fmi construct lacking the cadherin domains substantially rescues the requirement for Fmi in winner cells during cell competition (Figure 6). These observations suggest that supplying adhesion is not the principal function of Fmi in these events, but are consistent with the possibility that homodimeric adhesion, or interaction with a different ligand, normally activates the receptor, and that the truncated FmiΔCad behaves as a constitutively activated receptor capable of binding to either Gα proteins. While no biochemical characterization of Fmi has been reported, the human orthologs CELSR1-3 have been studied in detail. CELSR2 is autoproteolytically cleaved while CELSRs 1 and 3 are not, yet all three couple to GαS (Huong Bui et al., 2023). Additional efforts will be required to determine the functional ligand(s) for Fmi and whether it signals similarly. Furthermore, AGPCRs participate in a wide variety of developmental and physiologic events through diverse effectors (Einspahr and Tilley, 2022; Sreepada et al., 2022). The pathway by which Fmi participates in cell competition remains to be explored.

Figure 1:

Figure 2:

Figure 3:

Figure 4:

Figure 5:

Figure 6:

Figure 1—figure supplement 1:

Figure 1—figure supplement 2:

Figure 2—figure supplement 1:

Figure 4—figure supplement 1:

Figure 4—figure supplement 2:

Figure 5—figure supplement 1: