Developmental Cell - Flower Forms an Extracellular Code that Reveals the Fitness of a Cell to its Neighbors in Drosophila

Flower Forms an Extracellular Code that Reveals the Fitness of a Cell to its Neighbors in Drosophila

Christa Rhiner, Jesús M. López-Gay, Davide Soldini, Sergio Casas-Tinto, Francisco A. Martín, Luis Lombardía, Eduardo MorenoSee AffiliationsHint: Rollover Authors and Affiliations Spanish National Cancer Centre (CNIO), Melchor Fernández Almagro 3, E-28029 Madrid, Spain


Summary
Cell competition promotes the elimination of weaker cells from a growing population. Here we investigate how cells of Drosophila wing imaginal discs distinguish winners from losers during cell competition. Using genomic and functional assays, we have identified several factors implicated in the process, including Flower (Fwe), a cell membrane protein conserved in multicellular animals. Our results suggest that Fwe is a component of the cell competition response that is required and sufficient to label cells as winners or losers. In Drosophila, the fwe locus produces three isoforms, fweubi, fweLose-A, and fweLose-B. Basal levels of fweubi are constantly produced. During competition, the fweLose isoforms are upregulated in prospective loser cells. Cell-cell comparison of relative fweLose and fweubi levels ultimately determines which cell undergoes apoptosis. This extracellular code may constitute an ancient mechanism to terminate competitive conflicts among cells.

In metazoans, one mechanism by which suboptimal cells are culled is exemplified by the cellular interaction known as cell competition (reviewed in Milàn, 2002; Adachi-Yamada and O'Connor, 2004,Diaz and Moreno, 2005,Gallant, 2005). Cell competition was first described in Drosophila using mutations in ribosomal protein genes called Minutes (Morata and Ripoll, 1975,Simpson, 1979,Simpson and Morata, 1981). Minute homozygous flies are lethal, but heterozygotes are viable and normally sized, although it takes them longer to complete larval development due to lack of a fully active ribosomal machinery (Lambertsson, 1998,Marygold et al., 2007). However, when Minute heterozygous cells (M/+) and wild-type (WT) cells are generated in the same wing disc, (M/+) cells are no longer viable in the Drosophila wing (Moreno et al., 2002a). The fact that M/+ cells are eliminated only when growing next to WT cells was the reason why this phenomenon was termed cell competition (Morata and Ripoll, 1975). Cell competition is now believed to be a method by which weaker cells are eliminated from a population in order to optimize tissue fitness (Moreno et al., 2002a).

Since the discovery of cell competition using Minutes, other genes have been linked to cell competition. Among them dmyc, the Drosophila homolog of the proto-oncogene c-Myc (Johnston et al., 1999), is best studied. Cells with higher levels of dMyc outcompete adjacent lower dMyc-expressing cells, which are eliminated by apoptosis (de la Cova et al., 2004,Moreno and Basler, 2004). dMyc can activate a variety of genes encoding components of protein synthesis pathways (Orian et al., 2003,Grewal et al., 2005) and therefore stimulate protein translation. Cells that overexpress dMyc but simultaneously harbor a mutation in a ribosomal protein gene in heterozygosis lose the ability to outcompete surrounding cells (Moreno and Basler, 2004), suggesting that the apposition of cells with unequal rates of protein synthesis is one of the triggers of cell competition.

Competitive interactions among cells are thought to be short range (Simpson and Morata, 1981) and are classically initiated by an insult, such as mutations in Minute or dmyc genes, which increases or decreases the fitness of a particular cell within the imaginal disc epithelium of Drosophila. This translates into imbalances in morphogen and survival factor signaling, because the cell with reduced fitness is less efficient at endocytosing several extracellular factors (Moreno et al., 2002a,Moreno and Basler, 2004,Diaz and Moreno, 2005). Through a yet-unknown mechanism, cells probably monitor the signaling levels of their neighbors and recognize distortions in the morphogen gradient (Adachi-Yamada et al., 1999,Adachi-Yamada and O'Connor, 2002). Winner cells are then thought to produce a killing signal of unknown identity (Senoo-Matsuda and Johnston, 2007). This ultimately leads to caspase activation in the loser cell (Moreno et al., 2002a), which in turn induces an engulfment response in the winner cell (Li and Baker, 2007). Because the expansion of winners occurs at the expense of losers, total cell numbers do not change and the normal pattern of the organ is preserved (Moreno and Basler, 2004). Therefore, it has been proposed that dmyc-induced supercompetition could play a role in early stages of cancer, when mutant cells overproliferate but still obey the overall organ size control mechanisms imposed by the genome (Moreno, 2008,Rhiner et al., 2009).

In this paper, we investigate how cells of Drosophila wing imaginal discs distinguish winners from losers during cell competition. We took a genomic approach and combined it with functional assays in order to identify genes expressed early during the cell competition response. We found six markers upregulated early specifically in the loser cells: CG9233, CG1084, CG4672, CG6151, CG2198, and CG3305. Five of these genes encode for cell membrane proteins, suggesting that initial stages of cell competition rely heavily on cell-cell communication. We have further characterized CG6151. CG6151 encodes for Flower, a cell membrane protein that is conserved in multicellular animals and proposed to be a Ca2+ channel in neurons (Yao et al., 2009). Our results suggest that differential expression of CG6151 isoforms generate the scaffold and the extracellular epitopes required to mediate lose/win decisions during competitive interactions among cells of varying fitness.

We performed gene expression microarray analysis to identify novel molecular determinants and markers of the multistep process resulting in cell competition. It has been previously shown that cells expressing slightly higher levels of dMyc under the tubulin promoter (tub > dmyc) behave as supercompetitors and can outcompete neighboring WT cells, which in this setting perform as loser cells (de la Cova et al., 2004,Moreno and Basler, 2004). Therefore, a tub > dmyc > Gal4 transgene was used that allowed the Flippase (Flp) recombinase-mediated generation of Gal4-expressing cells that are WT regarding dmyc expression and marked by including UAS-lacz or UAS-GFP reporters (Brand and Perrimon, 1993) (Figure 1A). Such GFP-positive WT cells were outcompeted in a spatial and temporal pattern characteristic for cell competition (Moreno and Basler, 2004). Control clones that expressed similar levels of Gal4 and GFP (tub > cd2 > Gal4), but were not surrounded by tub > dmyc cells (no competition), survived normally. Heat shock was optimized to maximize the amount of boundaries where GFP and non-GFP clones contact each other. mRNA was extracted from wing discs undergoing competition (tub > dmyc > Gal4) and control discs (tub > cd2 > Gal4) at different time points (0, 12, 24, and 48 hr) and the profiles were analyzed using BDGRC microarrays (see Figure S1 available online).

To exclude dMyc downstream targets, we eliminated all those genes whose expression was differentially regulated in discs with homogeneous high levels of dMyc (tub > myc discs). The genes we identified as differentially expressed during cell competition could be roughly divided into four categories: (1) upregulated early, (2) downregulated early, (3) upregulated late, and (4) downregulated late. Among the downregulated genes we found expected controls like Gal-4 and LacZ, because Gal-4 expressing LacZ/GFP-marked WT cells disappear when in competition. The set of upregulated late genes contained several proapoptotic factors. We were mainly interested in early upregulated genes that, according to the microarray data, were expressed as early as 1224 hr after clone induction (ACI), because such genes could play an initiating role in the competition process.

To know whether those early upregulated genes were expressed in the loser or the winner cells, as well as to confirm our microarray data, mRNA fluorescence in situ hybridization (FISH) was performed in combination with antibody staining to visualize the GFP-positive WT cells in a tub > dmyc background (Figure 1A) or GFP-positive WT cells in a WT background (tub > gal4) (Figure 1C). Because GFP is typically lost during in situ hybridization, we developed a novel protocol to do FISH and antibody double staining that works efficiently (see Experimental Procedures). From the top ten early upregulated genes identified in the microarray (Figure S1), we could confirm at least six by in situ hybridization (Figure 1B): (1) CG9233, a zinc finger protein of unknown function (Celniker et al., 2002); (2) membrane GPI-anchored protein Contactin (CG1084) (Faivre-Sarrailh et al., 2004); (3) the transmembrane protein TMS1 (CG4672) (Celniker et al., 2002), a homolog of the serine incorporator family SERINC; (4) amalgam (CG2198), a cell membrane protein member of the immunoglobulin superfamily (Seeger et al., 1988); (5) the transmembrane protein Lamp (CG3305), found both cell membrane and lysosome-associated (Celniker et al., 2002); and (6) the transmembrane protein Flower (CG6151) (Yao et al., 2009). None of them had previously been described as a marker for cell competition. All six candidates were upregulated in the WT loser cells (Figure 1B) after induction of WT clones in a tub > dmyc background (Figure 1A), whereas no upregulation was detected in WT cells in a WT background (no competition) (Figures 1C and 1D).

To functionally analyze the role of those early loser-specific competition markers, we decided to knock-down the genes individually with UAS-RNAi constructs (Dietzl et al., 2007) in WT loser cells and quantify the number and size of remaining loser clones 72 hr ACI (Figure 2A). Again, transgenic flies of genotype tub > dmyc > Gal4 allowed the Flp recombinase-mediated generation of Gal4-expressing cells that are WT regarding dmyc expression and marked by GFP. Clones of such GFP-positive WT cells were eliminated by tub > dmyc supercompetitors after 72 hr. When UAS-RNAi against the genes CG1084 (contactin), CG4672 (TMS1), and CG2198 (amalgam) was specifically expressed in the loser cells, they were still eliminated similar to the controls where the yellow gene was targeted by RNAi or UASlacz was expressed instead (Figures 2A and 2B).

In contrast, when we tested UAS-RNAi against the genes CG9233, CG3305 (Lamp), and CG6151 (Flower) specifically in loser cells, they survived significantly better during cell competition. Seventy-two hours ACI, the rescue of WT clones using UAS-RNAi against CG3305 (Lamp), CG6151 (Flower), and CG9233, was comparable to the rescue obtained when apoptosis was blocked with the caspase inhibitor p35 (Hay et al., 1994), which is a known inhibitor of cell competition-induced apoptosis (Moreno et al., 2002a) (Figures 2A and 2B). Only expression of UASdMyc in the loser WT clones achieved a stronger rescue effect. The ultimate size of the surviving clones was comparable in all settings regardless whether RNAi was used (for example RNAi against Flower) or overexpression of dMyc, p35, or lacz (Figure 2C), but the number of remaining clones differed significantly 72 ACI (Figure 2B). All flies were previously examined at 24 hr to ensure that an equal amount of WT clones had been generated in all genotypes during heat shock (Figures S2A and S2B). Hereafter, we describe in more detail the function of Flower during cell competition. The role of the other genes will be discussed elsewhere.

The CG6151 locus encodes for Flower, a predicted protein with three or four transmembrane domains (Yao et al., 2009) and three isoforms of similar length that differ in their C-terminal part (Celniker et al., 2002) (Figure 3A). In order to verify the prediction of the three splice forms, we have fully sequenced the respective ESTs previously cloned by the BDGP. The protein sequence of Flower is conserved throughout evolution in all animals, from Drosophila to humans (Figure S2C). In silico predictions for the protein encoded by flower (fwe) suggested a conformation with an intracellular N-terminal part and an extracellularly exposed C terminus.
We could confirm this for the three isoforms by transfecting S2 cells with various either C-terminally or N-terminally hemagglutinin (HA)-tagged constructs (Figures 3B3E). S2 cells expressing Fwe isoforms with C-terminal HA tags could be detected with anti-HA antibodies in the absence of detergents (Figures 3B3D), confirming that the C-terminal portion of the protein is exposed to the extracellular space. The same was observed when C-terminally HA-tagged proteins were expressed in the epithelial cells of the wing imaginal discs and stained with an in vivo protocol to detect extracellular epitopes (Strigini and Cohen, 2000; see Experimental Procedures) (Figures 3F and 3G). In contrast, HA tags located at the N terminus were only detected by anti-HA antibodies when transfected S2 cells were treated with detergents that destroy the integrity of the cell membrane and allow the antibodies to enter the intracellular space (not shown), but not in the absence of detergents (Figure 3E). Therefore, the protein products of fwe are most likely three-pass transmembrane proteins with an intracellular N-terminal part and an extracellular C terminus of variable sequence (Figure 3H). All three isoforms share the same scaffold but differ in their extracellularly exposed C-terminal epitopes (Figure 3H).

We next studied the expression patterns of the three isoforms of fwe in the imaginal tissues, both in the absence and presence of cell competition. To this end, we generated several isoform-specific RNA probes or antibodies because the probe that showed upregulation of fwe in the initial in situ recognized all three isoforms (fwe total, Figure 3A). First, we generated a monoclonal antibody that specifically recognizes the C terminus of one of the protein isoforms (Figures 3H3L). This isoform was detected in all imaginal disc cells (Figures 3I3L) and it localized to the apico-lateral membrane in wing imaginal disc (Figure 3N) and salivary gland cells (Figure 3O). Therefore, we named this isoform fweubi because it is expressed ubiquitously in imaginal discs. Fweubi stainings also colocalized with phalloidin, in accordance with membrane localization (Figure S3A). Activation of UAS-fwe RNAi in the dorsal compartment, which targets all fwe isoforms, confirmed efficient downregulation of Fweubi levels (Figures 3A and 3J). Similarly, overexpression of UAS-fweubi driven by the dorsal apterous-Gal4 promoter was easily detectable by the anti-Fweubi-specific antibody (Figure 3K), whereas the same antibody did not cross-react with the two other forms when overexpressed in act > y > gal4 clones or using the apterous-Gal4 promoter (Figure S3B), confirming that it recognizes exclusively the Fweubi form.

The two other splice forms were not expressed at detectable levels in imaginal discs in the absence of cell competition (Figure 3M), as revealed with a high-affinity LNA probe that specifically detects those two isoforms (Figures 3A and 3M). However, when WT cells were exposed to competition by tub-dmyc cells, these isoforms were specifically induced in the loser (WT) cells, as verified with the LNA probe recognizing both non-fweubi forms or probes detecting only one of the two non-fweubi isoforms (Figures 4A4C). Because their expression appeared in loser cells, we termed them fweLose-A and fweLose-B, respectively. The fweLose signal was typically detected throughout loser clones and not just at clonal boundaries (Figures 4A4C) as it might be expected for short-range competitive interactions, suggesting that the triggered Loser state is somehow propagated within the clone.

The expression of fwe Lose isoforms was not induced in cells where apoptosis was triggered by other means than cell competition, e.g., targeted overexpression of eiger in the eye imaginal disc or clonal overexpression of activated hemipterous, the kinase activating Drosophila JNK (Adachi-Yamada et al., 1999) (Figures S4D and S4E). GMR-Gal4 driven expression of UASeiger in the eye leads to eye ablation as a consequence of massive JNK-dependent cell death (Moreno et al., 2002b). Coexpression of UASeiger RNAi or UAShid RNAi significantly rescued the eye ablation phenotype (Figures S4A and S4B; data not shown), whereas coactivation of UASfwe RNAi had no effect (Figure S4C), suggesting that fwe is a dedicated component of cell competition-induced apoptosis.

We next tested whether loser cells in other models of cell competition equally upregulate fweLose-A and fweLose-B, such as Minutes (M/+) (Simpson and Morata, 1981), thickveins (tkv) (Burke and Basler, 1996) or scribble (Brumby and Richardson, 2003) mutant cells. Slowly proliferating M/+ cells in a WT background were generated with a translocation in the Minute RpL19 gene and marked as described in Figure S4F. MARCM clones (Lee et al., 2000) in all three cell competition scenarios, scribble / cells, tkv / cells and M/+ cells in a WT background, induced fweLose isoforms (Figures 4D4F).

To test whether fweLose-A and fweLose-B are sufficient to label cells as losers and trigger their elimination, we decided to activate UAS-fweLose-A and UAS-fweLose-B constructs in WT clones in the absence of (dmyc-induced) competition. Transgenic flies of genotype act > yellow > Gal4, allowed the Flp recombinase-mediated generation of Gal4-expressing clones to overexpress the respective UAS-constructs driven by the ubiquitous actin promoter (act > Gal4; UASfwe). Such cells were marked by the use of UAS-GFP, whereas surrounding WT cells expressed the yellow gene (actin > yellow), which does not induce competition. When UAS-fweLose-A or UAS-fweLose-B were specifically activated in GFP-marked WT cells surrounded by unmarked WT cells, the green FweLose-expressing cells tended to disappear from the tissue over a time course of 72 hr (Figures 5A5C and 5G). Consistent with this, caspase was activated in fweLose-A and fweLose-B-expressing cells (Figure 5C, inset; Figures S5A and S5B). In general, fweLose-A and fweLose-B behaved indistinguishably in all experiments. In contrast, overexpression of Fweubi did not interfere with the growth of such cells and large Fweubi-expressing clones were observed 72 hr ACI (Figures 5D5G), which were negative for Caspase 3 (data not shown). Cells expressing a construct encoding for a truncated fwe form without any C-terminal extracellular epitopes (UASfwedelC) showed intermediate levels of apoptosis (Figures S5CS5F).

To further corroborate the differential effects of fweubi versus fweLose expression in nonepithelial cells, we transfected macrophage-like Drosophila S2 cells with ubi or -Lose forms of fwe together with a GFP construct in a 20:1 ratio. Twenty-four hours after transfection, a 12 hr video was recorded. Cells transfected with GFP (Movie S1) or fweubi (Movie S2) survived and did not show morphological changes. However, overexpression of fweLose forms induced cell death. Interestingly, we observed a correlation between the length of interaction with nontransfected S2 cells and imminent cell death of the transfected cell. Analysis of the videos showed that apoptotic corpses first fragmented and then cellular debris were engulfed by surrounding nontransfected cells (Movies S3 and S4). fweLose-expressing clones (act > y < Gal4; UASfweLoseB) generated in wing discs homozygously mutant for the Drosophila cell corpse engulfment receptor draper (Li and Baker, 2007), equally activated Caspase 3 at the clone border and underwent apoptosis (Figures S5G and S5H). Expression of fwedelC was not sufficient to trigger apoptosis of S2 cells, in contrast to fweLose forms (Movie S5).
From these experiments, we conclude that both fweLose-A and fweLose-B isoforms are sufficient to mark cells as losers in the presence of WT cells and trigger apoptosis in the absence of functional engulfment.

In order to distinguish if neighboring WT cells are required to detect and eliminate fweLose-expressing cells or if fweLose can trigger apoptosis cell autonomously, we performed a series of experiments where fweLose-A or fweLose-B was first expressed in large continuous clones and then ubiquitously in organs or the entire animal. Activation of fweLose forms in large cell populations (polyclones), where most fweLose-overexpressing cells lacked direct contact to WT cells, was achieved by applying long heat shocks to act > y > Gal4 flies. Apoptosis in such fweLose-overexpressing polyclones was diminished in the interior of the clone and a high proportion of apoptotic cells, revealed with anti-Caspase 3, was located specifically at the border of polyclones, where interaction with WT cells occurred (Figures 5H and 5I). Similarly, S2 cells cotransfected with any of the two fweLose isoforms and gfp, but lacking direct cell-to-cell contact to nontransfected cells due to plating at low confluency did not undergo apoptosis (Movie S5). Finally, the importance of the presence of neighboring WT cells (that do not overexpress any fweLose) to induce death of fweLose-expressing cells is strikingly illustrated in situations where all cells of the wing imaginal disc uniformly overexpress UASfweLose under the actin promoter and no increase in apoptosis is observed (Figure 5J). Similarly, overexpression of Lose forms in an entire compartment does not affect disc size/morphology (Figure S3B). In fact, adult flies that ubiquitously overexpress homogeneous levels of the Lose isoforms activated by act > Gal4; UASfweLose do not show any obvious defects (5K), which indicates that cells need to detect a relative difference in fweLose levels in order to recognize and outcompete fweLose-expressing cells.

To further analyze how the fweLose isoforms are specifically activated in the loser cells, we performed epistasis analyses. First, we checked if caspase activation is needed to trigger fweLose-A and fweLose-B induction as it is required, for example, for engulfment (Hoeppner et al., 2001). After blocking caspase activation specifically in loser cells with UAS-p35 (Hay et al., 1994) (Figure 6A), UAS-dIAP1 (the Drosophila inhibitor of apoptosis protein 1 (dIAP1) (Hay et al., 1995) (Figure 6B) or RNAi against the proapoptotic gene hid (Figure 6C) (Grether et al., 1995), fweLose-A and fweLose-B were still present at high levels in loser cells (Figures 6A6C), despite the fact that these cells did not undergo apoptosis, proving that caspase activation is not required for fweLose-A and fweLose-B expression in loser cells.
Next we tried to prevent fweLose-A and fweLose-B expression by interfering with an upstream event during cell competition such as imbalances in survival factor signaling. Because loser cells have been proposed to show a deficit in survival signaling (Moreno et al., 2002a,Moreno and Basler, 2004,Ziv et al., 2009), we overexpressed UAS-dpp (the Drosophila homolog of BMP2/4) in WT loser cells surrounded by supercompetitors. This reduced the levels of fweLose-A and fweLose-B activation as detected by FISH, but small WT clones surrounded by numerous supercompetitors and peripodial membrane cells (Figure 6D, arrowheads) still exhibited detectable fweLose levels. Activating Rab 5 using a UAS-rab5 construct, which stimulates endocytosis and rescues loser cells to a greater extent than Dpp overexpression (Moreno and Basler, 2004), strongly reduced levels of fweLose, even in small clones (Figure 6E). This suggests that fwe functions downstream of events that affect the overall fitness of a cell leading to decreased endocytosis of survival factors. As a control, we performed RNAi in loser cells against the fwe sequence common to all isoforms (see Figure 3A), which efficiently suppressed the fweLoseA signal (Figure 6F).

Unlike the fweLose isoforms, fweubi was not upregulated in loser cells and rather seemed to be slightly downregulated. This was visible especially at late time points (96 hr ACI), probably due to Fweubi protein perdurance (Figure 6G). In an alternative genetic set up, we created clones of WT cells growing in a slowly proliferating M/+ background. Again, such M/+ cells showed upregulation of fweLose and reduced levels of Fweubi. The reduction of Fweubi was subtle, but discernible in quantifications (Figures S6AS6C). This raises the possibility that loser cells undergo a switch during cell competition leading to the upregulation of fweLose-A and fweLose-B, probably at the expense of fweubi.

In order to study the consequences of a decrease of Fweubi, we generated fwe null mutants by inducing deletions via imprecise P-element excision. We obtained several deletions that affected the fwe locus, including two that remove the fwe promoter and 5UTR (fwee70 and fwee122) and one that eliminates the fwe 5UTR and the first two exons (fwee202), including the start ATG in exon 1 (Figure 3A). All three deletions were lethal, probably due to early defects in nervous system formation (Figures S7A and S7B), and failed to complement although they removed different regions of the locus, proving that all of them are mutations in fwe. As expected from the type of lesion, fwee202 clones did not show expression of fweubi 96 hr ACI, confirming that it is a null mutation (Figure 7A). Because fweubi is the only isoform detected in the wing imaginal cells in the absence of cell competition, we could analyze its function using our null mutants. Most experiments were done with fwee202, but fwee122 behaved identically. Clones of cells mutant for fwee202 survived initially but were gradually eliminated from the wing imaginal discs (Figures 7B7E) and were completely absent 120 hr ACI (Figure 7D). Consistent with this, fwee202 mutant clones showed caspase activation 96 hr ACI (Figures 7B and 7C). The pattern of the fwe/ apoptotic cells was particularly striking, because C3-positive cells appeared in rings at the periphery of the clones (Figures 7B and 7C), indicating that fwe/ mutant cells in contact with Fweubi-expressing cells were the first cells to be eliminated.

As expected from the expression pattern of the different fwe isoforms, only reintroducing Fweubi with a UASfweubi transgene, but not expression of UAS Lose forms inside the fwee202/ clones, using the MARCM system (Lee et al., 2000), could rescue cell death of fwe mutant cells revealed by an antibody that recognizes activated Caspase-3 (Figure 7F and data not shown), confirming that Fweubi is the isoform required to restore cell survival. Lack of Fweubi did not generally affect known survival transcription factors such as pMad (Tanimoto et al., 2000), dMyc (Johnston et al., 1999), or Vestigial (Halder et al., 1998), because fwe/ clones showed identical expression levels compared with neighboring WT cells (Figures S7CS7E).
We next sought to determine if relative differences in Fweubi expression are also involved in establishing win/lose decisions. To this end, we generated clones overexpressing the UASfwe RNAi used previously (Figure 2), which targets fwe exon-3 and 4 (Figure 3A), common to all fwe isoforms. However, this time we activated fwe RNAi in WT cells that are not suffering from cell competition (act > yellow > Gal4; UASfwe RNAi) and therefore only express the fweubi isoform. After Flippase induction, such cells were marked by UAS-GFP and overexpressed the RNAi against fwe (act > Gal4;UASfweRNAi). Clones of cells with downregulated fweubi levels activated Caspase 3 and tended to die 96 hr after clone induction when surrounded by cells with WT levels of fweubi (Figure 7G).

Because death of fwe mutant and fwe RNAied cells could be due to cell autonomous growth defects, we next devised an experiment where fwe is removed in cells with a proliferative advantage because they are facing Minute (M/+) cells (Figures 7H7J; see Experimental Procedures).

First, we carefully proved that +/+; fwe/ cells have indeed a proliferative advantage over M/+ cells. In order to test this we took advantage of the fact that such cells can form full posterior compartments homozygously mutant for fwe (Figures 7I and 7J). We found that those compartments reach their final size before the neighboring anterior compartment. For example, when dissected early in development (Martín and Morata, 2006), the size ratio posterior versus anterior compartment (P/A ratio) in fwe/M compartments is 0.73 in contrast to the maximum P/A ratio of 0.65 in a WT/WT disc, and similar to previous results of 0.7 for a WT/M disc P/A ratio, (Martín and Morata, 2006). This confirms that Minute+/+; fwe/ cells indeed have a growth advantage (Figures 7I and 7J).

Despite this growth advantage, the Minute+/+; fwe/ cells are forced to activate Caspase-3 (25 out of 31 clones showed massive signs of apoptosis) when coexisting in the same compartment with M/+ cells expressing fweubi (Figure 7H), but not when all the cells of the compartment lack fwe (Figures 7I and 7J). The death of fwe/ cells is caused solely by the presence of neighboring Fwe-expressing cells, independent of differential growth rates.
Consistent with the results shown in Figures 7I and 7J, homogeneous knockdown of fwe levels in a whole compartment using the UASfwe RNAi used previously and the engrailed-Gal4 driver (Figures 7K and 7L) did not induce increased apoptosis as revealed with anti-Caspase 3 antibody (Figures 7K and 7M) nor affect compartment growth (Figure 7N), despite strongly reduced levels of Fweubi in the posterior compartment (Figure 7L).

This illustrates that lack of Fweubi does only induce apoptosis when fwe/ cells interact with neighbors expressing WT levels of Fweubi, independent of their differential growth rates.
Our initial epistasis experiment showed that knockdown of all three fwe forms exclusively in loser cells can inhibit/delay competitive interactions (Figure 2). We next studied what happens to cell competition if fwe is knocked down in a whole compartment. To this end, we monitored the spread of tub > dmyc supercompetitors in a WT background (tub > cd2), in discs where RNAi of fwe was exclusively activated in the posterior compartment (hs-flp; en-Gal4, UAS-gfp/+; UASfwe RNAi/ tub > cd2 > dmyc) (Figures 7O and 7P). tub > dmyc supercompetitors grew equally in both compartments when UAS-lacz was expressed on the posterior side as a control (clone size in A/clone size in P ratio = 0.99) (Figure 7P) as previously reported (Moreno and Basler, 2004).
In contrast, RNAi of fwe significantly reduced the expansion of dMyc-overexpressing clones on the posterior side (clone size in A/clone size in P ratio = 1.53 (Figures 7O and 7P), without affecting overall growth of the posterior compartment (Figures 7N and 7O). This epistasis experiment shows that fwe reduction of function in both winner and loser cells exclusively affects cell competition mediated growth, whereas it does not interfere with normal tissue growth.

Here we have explored how cells of Drosophila wing imaginal discs distinguish winners from losers during cell competition. Five out of six genes identified in our screen for early competition markers were membrane proteins pointing to an important role of cell-to-cell communication during the initial phase of cell competition. Our results suggest that the membrane protein Fwe (CG6151) is a dedicated component of the cell competition response that is required and sufficient to label cells as winners or losers.

Fwe mediates win/lose decision by means of three differentially expressed isoforms, fweubi, fweLoseA, and fweLoseB. Cells are identified as losers when relative differences of fweubi or fweLose levels are detected (Figure 8). This system bears the advantage that cells are able to survive general stress conditions that uniformly affect the entire population within a compartment. We propose that, in outcompeted cells, the fwe transcript is alternatively spliced and fweLose isoforms are induced at the expense of fweubi. Probably both, the downregulation of fweubi, as well as the upregulation of fweLose contribute to establish the lose/win decision. We do not yet know how the alternative splicing is regulated. The simplest possibility is that when cells competing unsuccessfully for extracellular resources are deprived of survival factors (Diaz and Moreno, 2005), they are also depleted from some crucial splicing factors and default splicing will result in the formation of the normally repressed Lose forms. The observation that fweLose upregulation was usually detected throughout the entire loser clone and not just at clone borders could be the consequence of a mechanism that propagates the loser state in outcompeted clones. We consider two hypotheses as likely: a cell-to-cell signal that efficiently transmits the Lose verdict among outcompeted cells. Alternatively, border cells may transiently increase their uptake of survival factors such as Dpp, for example by generating cytoneme-like extensions (Hsiung et al., 2005), which would further deplete survival factors in the interior of the loser clone.

fwe shares certain features with proapoptotic or growth promoting genes with respect to cell competition, but overall it behaves differently and seems to stand in a class of its own.
Genes mediating apoptosis (hid, reaper) show a similar behavior to fweLose in certain aspects, in that they are triggered in loser cells and their elimination inhibits cell competition-induced apoptosis. Likewise, FweLose can trigger cell death in clones in the absence of cell competition. However, such proapoptotic factors induce apoptosis when overexpressed ubiquitously, whereas overexpression of fweLose (or lack of fweubi) throughout the wing imaginal disc or in the entire fly does not interfere with cell viability nor organ size. This context-dependence implicates that fwe does not work as a simple killing signal or some sort of toxic protein acting cell autonomously.

Fwe also shares features with genes known to affect normal tissue growth like Minutes (M/+) or dmyc such as cell-nonautonomous effects on survival in a heterotypic background. Homozygously mutant fwe cells show normal survival when all cells of one compartment are of the same genotype, but they are forced to undergo apoptosis when surrounded by wt cells, a hallmark of cell competition. However, this death does not depend on growth differences: (a) fwe/ cells are forced to activate caspase-3 in the presence of Minute cells, which have a lower proliferation rate, but do express fweubi; and (b) removal or downregulation of fwe throughout a compartment specifically inhibits cell competition without affecting the growth rate of the whole compartment.

It has been proposed that Fwe is a calcium channel (Yao et al., 2009). However, during cell competition we observe antagonistic functions for the different isoforms, unlike in synaptic vesicles where the two isoforms that we call here Ubi and Lose A seem to be functionally equivalent (Yao et al., 2009).

Finally, our data show that fweLose is not just an eat me signal because fwe Lose forms are able to trigger Caspase-3 activation and cause cell death before and in the absence of functional engulfment (; Figures S5G and S5H).

We propose that within a multicellular organ cells are constantly tagged by extracellularly exposed Fwe epitopes that function as a code. This extracellular code is composed by different Fwe isoforms and allows comparison of relative fitness. During cell competition in Drosophila, the fwe isoforms work as a simple ternary code (fweubi, fweLose-A, and fweLose-B) with a binary output, because fweubi is translated as intact cellular fitness whereas fweLose-A and fweLose-B are redundant and lead to cell elimination (Figure 8). The experiments with the C-terminally truncated Fwe form suggest that the presence of the Lose epitopes aid in the labeling of cells as losers, although the lack of the ubi tail may also help in their elimination.

We expect other molecules to interact with Fwe, which are able to interpret the thresholds or read the extracellular epitopes displayed by the Fwe isoforms. It is likely that the signal recognized by neighboring cells includes not only the variable C-terminal epitopes, but also the constant extracellular loop because its sequence is conserved from flies to humans.
The code composed by the Fwe isoforms may have biomedical implications beyond cell competition because imbalances in cell fitness appear during aging, cancer formation, and metastasis.
For the microarrays, marked WT cells were generated in a dMyc-overexpressing background with the tub > dmyc > Gal4 transgene as described in Moreno and Basler, 2004. For control clones, the tub > cd2 > dmyc transgene was used instead. Heat shock was optimized to maximize the amount of boundaries where GFP and non-GFP clones contact each other, and mRNA was extracted with TRIzol reagent (Invitrogen) and RNeasy (Quiagen) from both genetic setups at 0, 12, 24, and 48 hr and the profiles were analyzed using BDGRC microarrays.

Primary antibodies used were anti-GFP (rabbit, 1:200; Abcam; and 1:300; Invitrogen), anti-β Gal (rabbit, 1:1000, Cappel), anti-C3 antibody (rabbit, 1:300; Cell Signaling), anti-repo (mouse, 1:50; Developmental Hybridoma Bank), anti-Wg (mouse, 1:50; Developmental Hybridoma Bank), anti-HA (rat, 1:500 in imaginal discs, 1:100 in cells; Roche), anti-Vestigial (rabbit, 1:100, a gift of Sean Carroll; and 1:50, guinea pig, a gift of Gines Morata), anti-Brinker (rabbit, 1:300), anti-dMyc (guinea pig, 1:300), and anti-pMad (rabbit, 1:300) (all three gifts of Gines Morata). Phalloidin Alexa-488 was applied 1:100 (Molecular Probes).

To generate specific antibodies against the Fweubi isoform, a synthetic peptide with the C-terminal sequence (NNAQPFSFTGAVGTDSNV) was used to immunize mice and generate monoclonal antibodies. Anti-Fweubi antibodies were used 1:30. Stainings were performed using standard procedures except for the staining of Figure 3F, where the protocol described by Strigini and Cohen, (2000) was used. Embryos were dechorionated in 30% bleach and fixed in 4% PFA according to standard procedures. Images were taken in a SP2 or SP5 Zeiss confocal microscope and analyzed with Adobe Photoshop CS2 and ImageJ 1.41.

Larvae were dissected in sterile-filtered phosphate-buffered saline (PBS) on ice, fixed for 20 min in filtered 4% PFA/0.4% PBS-Triton-X (PBT) on ice and washed in PBT (0.4%). Then, larvae were incubated for 10 min in hybridization buffer (HB)/PBT (1:1) for 10 min, followed by blocking in HB for 12 hr. 500800 ng mRNA probe was diluted in HB and hybridized for 48 hr at 56C. The LNA-probe against fweLose-A/Lose-B (Exiqon) was used at a concentration of 2.5 pM with a hybridization time of 24 hr at 56C.

Larvae were then washed twice in HB (at 56) for 10 min followed by rehydration in prewarmed (56C) HB:PBT (3:1, 1:1, 1:3) at room temperature for 30 min each. After rehydration, samples were washed in PBT and blocked with PBT-BSA (1%). Antibodies used to detect DIG-labeled probes and loser cells were anti-DIG (mouse, 1: 750) (Jackson) and anti-GFP (rabbit, 1:200) (Abcam), respectively. The Dig signal was amplified with biotinylated anti-mouse (1:200) (Jackson) followed by Tyramide amplification (Invitrogen).

Primers used to generate specific CG6151 probes by in vitro transcription (Roche) were:
CG6151:ATTTAGGTGACACTATAGAAGAGTTCGGCCTGTGGAATGTG/TAATACGACTCACTATAGGGAGACACGGAAGTACAAGGGCT
Lose-A: ATTTAGGTGACACTATAGAAGAGGCTTCTCGAGAGGACATGG/TAATACGACTCACTATAGGGAGAGGCGCCAGACATCGG
Lose-B:ATTTAGGTGACACTATAGAAGAGGCCATTCCGCCCATTAT/TAATACGACTCACTATAGGGAGAATAGGTTCCGGTTCCTCT
LNA-probe (recognizing both Lose isoforms): CATTCGGTTAGCTTTCAAATATAGT.

Deletions in the fwe locus were obtained by imprecise excision of transposon P(EPgy2) in the strain CG6151EY08496 (Bloomington). Transposon jumping was induced by crossing flies to a transposase source (delta 2-3). Recovered transposon jumps were balanced and DNA of > 250 established stocks screened for deletions in fwe. Three different deletions were recovered: fwee70, fwee122, and fwee202. E202 is a clean deletion in the fwe locus of 957 bps, which removes exon 1 (including the ATG) and exon 2 of the coding sequence breakpoints: AAGTACAACAGGATTTTTTT/TTTGATAACTTTTTATTTCG. Breakpoints for E122: GCTGATATTTTCGAG/CGTTCCGTGTGACGTG and for E70: ATCTCCATATGCTCGTTTT/TTCCGTGTGACGTGAAAAGT.

All three deletions are lethal and were recombined with FRT80B (Xu and Rubin, 1993) and maintained balanced over Tm6b.

cDNAs of fweubi, fweLose-A, and fweLose-B as well as the fwedelC sequence (see Figure S5) were fully sequenced and subcloned into the pUASp vector using KpnI (BamHI for fweubi) and XbaI restriction sites. In order to generate N- and C-terminal HA-tagged forms, the respective cDNAs were amplified with primers containing the HA sequence and subcloned into KpnI and XbaI sites of pUASp. Primer sequences are available upon request. Multiple transgenic lines with insertion on different chromosomes were generated by Bestgene.

For CG9233, two independent transformants 19804 and 19805 were used that gave similar results (data shown for 19804). For CG1084, two independent transformants, 28294 and 40613, were used that gave similar results. For CG4672, transformant 8380 was used. For CG6151 transformants 39596, 104993 (VDRC) and stock 27323 (Bloomington) were tested for rescue of WT loser cells and gave similar results. Data for 39596 are shown, which targets exon-3 and exon-4, common to all three CG6151 isoforms. For CG2198, two independent transformants, 22944 and 22945, were used that gave similar results. For CG3305 two independent transformants, 7308 and 7309 (data shown), gave similar results. All the data shown in Figure 2 were obtained from female larvae in order to minimize growth variability. For rescue of Eiger-induced cell death in the eye (Figure S4), we used transformants 45252 and 45253.

All error bars represent the standard error of the mean, except in Figure 7, where the standard deviation is shown.

open here to see the full-text:
Developmental Cell - Flower Forms an Extracellular Code that Reveals the Fitness of a Cell to its Neighbors in Drosophila

open here to see the related-information:
http://www.cell.com/developmental-cell/

open here to see the CNIO contents:
http://www.cnio.es/es/index.asp