The CRL4 E3 ligase Mahjong/DCAF1 controls cell competition through the transcription factor Xrp1, independently of polarity genes

Cell competition, which is the elimination of, in most cases, slower-growing cells by faster-growing cells in mosaics, is important for precise development, regeneration and physiological maintenance (Clavería and Torres, 2016; Nagata and Igaki, 2018; Baker, 2020; Morata, 2021; Parker et al., 2021). Cell competition was first recognized in Drosophila in the case of cells lacking one copy of ribosomal protein genes (Rp/+). These mutants, which are known as ‘Minutes’ because of their thin body bristles, also display slow growth (Bridges and Morgan, 1923; Lambertsson, 1998). Minute mutant cells are eliminated from mosaics with wild-type cells by caspase-dependent cell death (Morata and Ripoll, 1975; Simpson and Morata, 1981; Moreno et al., 2002; Li and Baker, 2007; Kale et al., 2015). Super-competition, the name given to the process of eliminating wild-type cells, happens in mosaics with faster-growing Myc- or Yorkie (Yki)-expressing cells (de la Cova et al., 2004; Moreno and Basler, 2004; Tyler et al., 2007; Neto-Silva et al., 2010). Considered together, cell competition and super-competition suggest that comparison of cellular fitness leads to cell competition. Because the mammalian homologs of Myc and Yki are proto-oncogenes (Dang, 2012; Moroishi et al., 2015), it has been suggested that super-competition might contribute to tumor expansion, as several recent studies have confirmed (Suijkerbuijk et al., 2016; Di Giacomo et al., 2017; Patel et al., 2017; Liu et al., 2019; Madan et al., 2019; Moya et al., 2019).

Cell competition may also be tumor suppressive. Global loss of apico-basal polarity genes such as lgl [l(2)gl)] or scribble (scrib) leads to polarity-deficient neoplasia of Drosophila imaginal discs (Bilder et al., 2000; Bilder and Perrimon, 2000; Humbert et al., 2003), but clones of lgl or scrib cells are eliminated from mosaics (Igaki et al., 2006; Froldi et al., 2010; Menendez et al., 2010; Tamori et al., 2010; Chen et al., 2012). These lgl or scrib mutant clones do not form tumors unless cell competition is blocked and mutant cells remain in the epithelium (Menendez et al., 2010; Chen et al., 2012; Khan et al., 2013).

Competition of Rp/+ mutant cells might also serve a tumor-surveillance role. Rp genes are spread throughout the genome (Uechi et al., 2001; Marygold et al., 2007), and it has been shown they can serve as sensors for aneuploidy, leading to elimination of aneuploid cells containing monosomies that affect Rp gene dose (Ji et al., 2021). Mutants with disrupted cell competition accumulate aneuploid cells (Ji et al., 2021). These would be expected to be tumorigenic in mammals, where aneuploidy is associated with tumorigenesis (Ben-David and Amon, 2020; Molina et al., 2021; Li and Zhu, 2022). Competition of Rp/+ cells depends on the Drosophila bZip AT-hook domain transcription factor Xrp1 (Lee et al., 2016; Baillon et al., 2018; Lee et al., 2018). The Xrp1 expression induced in Rp/+ imaginal discs is also responsible for most of their altered gene expression, their slow growth and their reduced translation, in addition to their propensity to be eliminated by cell competition (Lee et al., 2018). Xrp1 is also expressed in the DNA damage response, where its transcription is p53 dependent (Brodsky et al., 2004; Akdemir et al., 2007). Xrp1 induction in Rp/+ cells is independent of p53 but dependent on a particular Rp protein, RpS12, which is thought to play a role in signaling the defect in ribosome biogenesis (Kale et al., 2015; Kale et al., 2018; Lee et al., 2018; Ji et al., 2019).

The tumor-suppressive cell competition of polarity gene mutant cells has been proposed to go through Mahjong (Mahj) (Tamori et al., 2010), a CRL4 E3 ubiquitin ligase (Ly et al., 2019). Mahj physically interacts with Lgl, and its overexpression in lgl mutant clones suppresses their elimination from mosaic tissues (Tamori et al., 2010). Interestingly, mahj knockdown in MDCK cell cultures also leads to their elimination by co-cultured normal MDCK cells, suggesting a cell competition mechanism that is conserved between Drosophila and mammalian cells (Tamori et al., 2010). The mammalian homolog of mahj, known as DDB1-Cul4-associated factor 1 (DCAF1) or human immunodeficiency virus type 1 accessory protein Vpr-binding protein (VprBP), is important for G2 cell cycle arrest and virus replication after HIV1 infection (Le Rouzic et al., 2007; Tan et al., 2007). Dcaf1 is required for mouse embryogenesis and its knockdown affects cell proliferation, cell cycle and cell survival in multiple cell types (McCall et al., 2008; Guo et al., 2016). Dcaf1 interacts with the Hippo pathway and its knockdown also stabilizes p53 (Li et al., 2014; Ly et al., 2019; Han et al., 2020), but there has been no report of Dcaf1/VprBP affecting epithelial cell polarity in mammals.

In Drosophila, the overall transcriptional signature of mahj mutant wing discs is unexpectedly similar to that of Rp/+ mutants, including upregulation of Xrp1 mRNA (Kucinski et al., 2017). Because mahj and Rp/+ cells were thought to represent distinct mechanisms of cell competition, this finding suggested a gene expression signature common to cells targeted by cell competition (Kucinski et al., 2017). Besides transcription, other similarities have been reported between mahj and Rp/+ mutant cells, including autophagosome accumulation and evidence of proteotoxic stress (Nagata et al., 2019; Baumgartner et al., 2021).

Here, we show that mahj mutant cells trigger cell competition through an Xrp1-dependent pathway like that in Rp/+ cells, and distinct from cell competition of lgl or scrib clones, which do not express or depend on Xrp1 function for elimination. Xrp1 expression also makes mahj mutant cells phenotypically like Rp/+ cells, that is, results in ‘Minute-like’ thin thoracic bristles, slow growth, reduced translation, altered autophagy and increased JNK signaling. Regulation of Xrp1 by mahj likely requires its E3 ligase activity, depending on DNA Damage Binding Protein 1 (Ddb1) and Cullin 4 (Cul4). These results show that mahj mutant cells suffer cell competition because of a transcriptional response to altered ubiquitinylation mediated by Xrp1 and therefore resembling Rp/+ mutant cells. This seems unrelated to elimination of scrib or lgl mutant cells – the polarity-defective cells. Thus, loss of mahj function is an additional genotype triggering elimination by the Xrp1-dependent pathway that also removes Minute cells, not the mechanism for eliminating tumorigenic polarity-deficient cells.

Because Xrp1 mRNA is elevated in mahj mutant wing discs (Kucinski et al., 2017), we examined Xrp1 protein expression and Xrp1 function in clones of cells undergoing cell competition due to loss of mahj. First, loss-of-function clones of mahj¹ allele were created; these are known to be outcompeted when next to control (mahj/+ or +/+) cells (Tamori et al., 2010). Xrp1 protein expression was examined making use of an allele tagged with HA at the endogenous Xrp1 locus (Blanco et al., 2020). Xrp1-HA protein was undetectable in wild-type cells but clearly expressed in mahj¹ loss-of-function clones in the wing discs (Fig. 1A,A′). We also knocked down mahj through expression of dsRNA in the posterior wing compartment using en-Gal4. mahj knockdown in posterior compartments also resulted in Xrp1-HA protein expression (Fig. 1B,B′). This indicates that cell competition is not required for Xrp1 expression, as all the posterior compartment cells are depleted of mahj. Interestingly, mahj knockdown also reduced the relative size of the posterior compartments, a result that was variable but statistically significant (Fig. S1A-C). The reduction in posterior compartment size suggests mahj knockdown impacts cellular growth regulation. We made multiple attempts to assess whether Xrp1 is required for the reduced compartmental growth that results from mahj depletion in wing disc compartments, but so far it has not been possible to obtain larvae of these mahj Xrp1 co-depletion genotypes.

To determine the functional significance of Xrp1 protein expression in mahj mutant cells, the size of mahj¹ mutant clones was compared with parallel mahj¹ clones expressing Xrp1 RNAi, and we also measured apoptosis in these clones. Xrp1 knockdown rescued mahj¹ clone size significantly and reduced the cell death that was otherwise seen at the boundaries of mahj mutant clones with wild-type areas (Fig. 1C-G). The functional requirement for Xrp1 was further confirmed by making flip-out clones that expressed mahj dsRNA, in comparison with clones expressing both mahj dsRNA and Xrp1 dsRNA. In this case also, clone size and boundary cell death were significantly rescued by Xrp1 knockdown (Fig. S1D-H). Notably, we occasionally found accumulation of some dying cells accumulating within the mahj Xrp1 double knock-down clones (Fig. S1F,F′, cyan arrowhead). It is important to note that, not only do the similar phenotypes of mahj mutant clones and clones expressing mahj dsRNA confirm the specificity for mahj loss of function and argue against any off-target effect or passenger mutation causing the cell competition, but the rescue of mahj phenotypes using distinct Xrp1 dsRNAs also confirms the specificity of the Xrp1 knockdown results.

Finally, Xrp1 loss-of-function clones were made in a global mahj mutant background. Reminiscent of FRT82 control clones and their twin spots (Fig. 1H), Xrp1 loss of function clones grow similarly to twin-spot controls in an otherwise wild-type context (Lee et al., 2018). By contrast, Xrp1 mutant clones were enlarged in mahj mutants and the twin spots almost eliminated (Fig. 1I,J). This suggests that Xrp1 was sufficient to induce cell competition between mahj mutant cells based on Xrp1 expression. Clearly, mahj mutant cells expressing two wild-type copies of Xrp1 were disadvantaged compared with mahj mutant cells that were also mutant for Xrp1.

Because we found that mahj mutant cells resembled Rp/+ cells in their Xrp1-dependent cell competition, and in reduced growth rate (Fig. 1C-J), we wondered whether mahj loss of function would affect bristle size, as Rp mutations do (Bridges and Morgan, 1923). As predicted, expression of mahj dsRNA specifically in bristle primordia resulted in small and thin bristles, similar to Rp/+ mutants (Fig. 1K,L). Although expression of Xrp1 dsRNA by itself had no effect on bristles (Fig. 1N), the mahj knockdown phenotype was partially rescued upon co-expression of Xrp1 ds RNA (Fig. 1M,O), illustrating another similarity between mahj and Rp/+ phenotypes.

mahj mutant cells are reported to exhibit multiple abnormalities that are also seen in Rp/+ mutant cells (Kucinski et al., 2017; Nagata et al., 2019), in which case they are Xrp1 dependent (Lee et al., 2018). First, JNK signaling is activated in and required for the competitive cell death of mahj mutant cells (Tamori et al., 2010). Elevated JNK activity is Xrp1 dependent in Rp/+ mutant cells (Lee et al., 2018). We found that, similarly, JNK hyperactivity disappeared from mahj mutant cells after Xrp1 knockdown (Fig. 2A-B″, Fig. S2A). Second, autophagosomes accumulate in both mahj mutant clones and Rp/+ mutant cells (Nagata et al., 2019; Baumgartner et al., 2021). Autophagosome accumulation is Xrp1 dependent in Rp/+ mutant cells (Langton et al., 2021; Kiparaki et al., 2022). Autophagosomes are believed to accumulate because of reduced autophagic flux, although there is disagreement about whether autophagy is detrimental or protective for Rp/+ cells(Nagata et al., 2019; Baumgartner et al., 2021). Using lysotracker dye accumulation as a marker for autophagosomes (Nagata et al., 2019; Recasens-Alvarez et al., 2021), we confirmed accumulation in mahj mutant discs, which was Xrp1 dependent (Fig. 2C-E, Fig. S2B). Third, some authors have reported lower general translation levels in mahj mutant cells (Nagata et al., 2019), although this has not been observed by others (Baumgartner et al., 2021). Reduced translation is also a feature of Rp/+ mutant cells, where it is Xrp1 dependent (Lee et al., 2018). Using accumulation of OPP (O-propargyl-puromycin), an alkyne analog of puromycin, to measure total cellular translation (Lee et al., 2018), we confirmed that mahj mutant clones displayed lower translation than wild-type cells (Fig. 2F-F″). The difference was abolished by Xrp1 knockdown, indicating that overall translation is reduced by Xrp1 in both Rp/+ and mahj mutant cells (Fig. 2G-G″). Overall, multiple cellular defects in mahj mutant cells were found to be Xrp1 dependent.

The expression of Xrp1 protein, which is very low in wild-type imaginal discs, is known to be induced through multiple distinct mechanisms. First, Xrp1 is the major transcriptional target of p53 after DNA damage (Brodsky et al., 2004; Akdemir et al., 2007). Second, in Rp/+ mutants, Xrp1 expression depends on rpS12, not p53 (Lee et al., 2018; Ji et al., 2019). Finally, Xrp1 protein expression can be induced by eIF2α phosphorylation (Brown et al., 2021; Langton et al., 2021; Ochi et al., 2021; Kiparaki et al., 2022). eIF2α phosphorylation commonly occurs as a result of ER stress and results in the overall reduction of CAP-dependent translation initiation, while favoring the translation of some transcripts (Ryoo and Vasudevan, 2017; Wek, 2018).

To assess the role of p53, we knocked mahj down in posterior wing compartments, and observed the same level of Xrp1-HA expression even when dominant-negative p53 (p53-DN) was co-expressed (Fig. 3A-B′, Fig. S3A). We also compared expression of a P53 reporter, rpr-150 lacZ (Brodsky et al., 2000), upon mahj knockdown and observed no change. This indicates that loss of mahj does not increase p53 activity, which should activate this reporter (Fig. S3B,B′). Finally, p53 dsRNA failed to rescue competitive elimination of mahj mutant clones (Fig. S3E-G), although p53 dsRNA was sufficient to prevent cell death induced by irradiation (Fig. S3C-D).

To assess the role of Rps12, Xrp1-HA protein expression in mahj mutant clones was assessed in the background of the rpS12^G97D, the rpS12 mutation that prevents Xrp1 induction in Rp/+ cells (Kale et al., 2018; Lee et al., 2018). Xrp1-HA protein expression in mahj mutant clones was unaffected (Fig. 3C-D′, Fig. S4A), as was their growth and survival, indicating that rpS12 was not required for mahj mediated cell competition (Fig. S4B-D). Moreover, an Xrp1-LacZ enhancer trap was unaffected by mahj knockdown in posterior wing compartments (Fig. S4E,E′), although RpS12 regulates this enhancer trap in Rp/+ cells (Lee et al., 2018; Ji et al., 2019). Because Xrp1 mRNA levels are elevated in mahj mutants (Kucinski et al., 2017), this suggests the Xrp1-LacZ enhancer trap may not report all aspects of Xrp1 mRNA regulation.

We found that mahj knockdown resulted in higher eIF2α phosphorylation (Fig. S4F-G). eIF2α is also phosphorylated in Rp/+ cells, and is responsible for their reduced overall translation rate (Baumgartner et al., 2021; Langton et al., 2021; Ochi et al., 2021; Recasens-Alvarez et al., 2021; Kiparaki et al., 2022). As for Rp/+ cells (Kiparaki et al., 2022), co-expression of the eIF2α phosphatase PPP1R15 had no effect on Xrp1-HA levels induced by mahj dsRNA, indicating that eIF2 α phosphorylation was not required for Xrp1 protein expression caused by mahj depletion (Fig. 3E, Fig. S4H). Overall, none of the currently known mechanisms explains induction of Xrp1 upon mahj knockdown, suggesting an additional way to induce Xrp1 and launch cell competition in mahj mutant cells.

We explored whether mahj regulates Xrp1 protein expression and cell competition, through its ubiquitin ligase function. Cullin Ring Ubiquitin ligases (CRL) are the largest family of E3 ubiquitin ligases (Hershko and Ciechanover, 1998; Sang et al., 2015). Cullins act as scaffolds to link a Ring-box protein required for interactions with an E2 ubiquitin ligase with an adapter protein that recruits substrates and determines the substrate specificity of ubiquitylation (Petroski and Deshaies, 2005; Angers et al., 2006). Mahj is such a substrate adapter, binding to Cul4 through DDB1 to constitute the CRL4 (Ly et al., 2019). We found that knockdown of either cul4 or ddb1 resulted in Xrp1 protein expression in wing discs (Fig. 4A-B′). Xrp1 protein expression was also observed upon knockdown of Roc1a (Fig. S5A,A′), which is responsible for CRL4 interaction with the E2 ligase (Angers et al., 2006). Cul4 function requires neddylation with Nedd8 (Pan et al., 2004). Mutation of the neddylation sites results in a dominant-negative Cul4 molecule (Lin et al., 2009). Over-expression of dominant-negative Cul4 also resulted in Xrp1 protein expression in wing discs (Fig. 4C,C′). Finally, as expected if perturbation in ubiquitin-mediated protein turnover is responsible for Xrp1 protein expression in mahj mutants, Xrp1 protein expression was also observed upon knockdown of different proteasome subunits (Fig. 4D,D′, Fig. S5B-C′). If mahj mutants lead to cell competition by affecting ubiquitin-dependent protein turnover, we would expect that knocking down the partners of Mahj in the CRL4 complex would also lead to cell competition. Consistent with this notion, knockdown of cul4 is known to result in poor clone survival (Tare et al., 2016). Here, we made ddb1 knockdown clones by expressing ddb1 RNAi in actin Gal4 flip-out clones. 48 h after induction, these clones displayed extensive cell death both at the boundaries with wild-type cells and within the clones themselves (Fig. 5A-B′). No ddb1 knockdown clones could be detected 72 h after clone induction (Fig. 5G). Significantly, simultaneous knockdown of Xrp1 within the clones dramatically reduced cell death both at the clone boundary and within clones (Fig. 5C-E), and allowed these clones to survive for 72 h (Fig. 5F-I). Similar to ddb1 knockdown clones, expression of cul4 RNAi in actin-Gal4 flip-out clones for 48 h also resulted in cell death both within clones as well as at clone borders, dependent on Xrp1 in most wing discs (n=8/10, Fig. S6A-B′,D-E). A few discs did not show rescue of cell death (n=2/10, Fig. S6C,C′). Additionally, like mahj knockdown, expression of cul4 RNAi resulted in phosphorylation of eIF2α in an Xrp1-dependent manner (Fig. 5J-K′).

Overall, these findings show that ddb1 and cul4 loss of function result in Xrp1 protein expression and Xrp1-dependent clone elimination similar to that seen with loss of mahj. This strongly suggests that mahj regulates cell competition through CRL4-dependent ubiquitylation of a protein that would otherwise promote Xrp1 expression. It is interesting that ddb1 mutants also show ‘Minute’-like bristles (He et al., 2006), a further connection between the ubiquitin ligase function of mahj and the Minute phenotype caused by Rp/+ genotypes. It is possible that ddb1 and cul4 loss of function also cause cell-autonomous cell death due to other survival roles of these genes.

In neural stem cells, Mahj recruits Warts for CRL4-dependent ubiquitylation (Ly et al., 2019). Warts (Wts), a serine threonine kinase, is a member of the Hippo pathway that phosphorylates and inhibits the transcription co-activator Yorkie (Yki), thereby inhibiting growth (Justice et al., 1995; Wu et al., 2003; Huang et al., 2005; Dong et al., 2007; Hao et al., 2008). Because Yki affects cellular growth and differences in Yki activity can trigger competition between cells (Tyler et al., 2007; Neto-Silva et al., 2010), we wondered whether Warts could be the Mahj target regulating Xrp1 expression and cell competition in mahj mutant cells. To check how Warts affects mahj clones in wing discs, we first examined Hippo signaling reporters after knockdown of mahj. LacZ reporters of ex, fj and diap, which are sensitive to Hippo signaling (Wu et al., 2003; Cho et al., 2006; Hamaratoglu et al., 2006; Tyler and Baker, 2007; Wang and Baker, 2015), were each downregulated in posterior compartments upon mahj knockdown, consistent with reduced Yki activity (Fig. 6A-C′), and as expected based on the neural stem cell findings (Ly et al., 2019). However, no changes in the reporters were observed upon mahj overexpression (Fig. S7A-B′), suggesting that mahj is required but not sufficient to regulate Wts in the wing disc. Interestingly, wts mutant clones also displayed higher translation (Fig. 6D,D′), as did overexpression of Yki or of its miRNA target Bantam [Fig. S7C-D′, also recently reported by another group (Nagata et al., 2022)], which would be consistent with the Hippo pathway affecting Xrp1. To test whether Warts stabilization is the mechanism whereby mahj regulates Xrp1 and translation (Ly et al., 2019), Wts was overexpressed in the wing discs.

Wts overexpression led to the expected reduction in size of the wing pouch (Lai et al., 2005), but no Xrp1-HA protein expression was observed (Fig. 6E-F′). Similar results were obtained after overexpression of hippo (hpo), which encodes a serine threonine kinase that phosphorylates Warts and positively regulates Warts activity (Pantalacci et al., 2003; Wu et al., 2003). Hpo overexpression also greatly reduces growth of wing disc cells (Huang et al., 2005; Tyler and Baker, 2007), and led to a more severe reduction in size of the wing pouch than did Wts overexpression (Fig. 6F, Fig. S7F). Xrp1-HA protein expression was still not observed (Fig. S7E-F′). To investigate whether Hippo signaling might be necessary for Xrp1 expression, although not sufficient to induce it alone, we co-expressed wts RNAi and mahj RNAi, but observed no change in Xrp1 protein (Fig. 6G-H′, Fig. S7I). In the same way, Xrp1 protein expression was not changed in cells with knockdown for mahj and simultaneous overexpression of Yki (Fig. S7G-H′). These findings show that Hippo signaling and Yki activity levels do not contribute to Xrp1 expression in mahj mutant cells. Thus, although mahj did regulate Hippo signaling in wing discs, consistent with regulation of Warts stability, this did not seem to be the mechanism of Xrp1 protein expression in mahj mutant wing disc cells.

Mahj came to the attention of the cell competition field as a binding partner of Lgl, proposed to mediate the competitive elimination of cells with lgl mutations and perhaps mutations for related cell polarity genes, including scrib (Tamori et al., 2010; Baker, 2011; Levayer and Moreno, 2013; Claveria and Torres, 2016). As we have found that mahj affects cell competition through Xrp1 and therefore through a pathway broadly similar to that seen in Rp/+ cells, we expected that competitive elimination of lgl and scrib mutant clones would also be mediated by Xrp1. Contrary to this expectation, no Xrp1-HA expression was detected in lgl loss-of-function clones (Fig. 7A,A′). Moreover, there was no rescue in clone size or boundary cell death when Xrp1 was knocked down in lgl mutant cells (Fig. 7B-D, Fig. S8A). We also found that scrib mutant clones were still eliminated in an Xrp1 mutant background (Fig. 7E-G, Fig. S8B). Thus, Xrp1 was not required for cell competition in these two polarity-defective mutant genotypes.

In many of our experiments where mahj was knocked down in posterior wing disc compartments, the A/P compartment boundary became markedly irregular, and sporadic loss of GFP expression was observed within the compartment. Double-labeling of every cell with DAPI confirmed that posterior compartments depleted for mahj often contain cells lacking GFP expression, something never seen in en-Gal4 UAS-GFP controls(Fig. 7H-I″). Such sporadic loss of marker expression is thought to reflect loss of heterozygosity reflecting genomic instability(Dekanty et al., 2012). Thus, mahj may have functions unrelated to epithelial cell polarity, including a contribution to genome stability yet to be fully elucidated in Drosophila.

In this research article, we explore the cell competition mechanisms of Mahj, a CRL4 ubiquitin ligase (Ly et al., 2019), the mutation of which triggers similar cellular effects to Rp/+ mutations, including similar changes in gene expression, global translation rates, JNK activity and autophagy, leading mahj cells to be eliminated by competition with wild-type cells, as Rp/+ cells are (Fig. 7J) (Tamori et al., 2010; Kucinski et al., 2017; Nagata et al., 2019). The basis of the similarity is that mahj and Rp loss of function both activate expression of Xrp1, the transcription factor that coordinates these effects (Figs 1,2). Unlike Rp/+ genotypes, which activate Xrp1 through a rpS12-dependent mechanism (Lee et al., 2018; Ji et al., 2019), Mahj regulates Xrp1 most likely through its ubiquitin ligase activity, which depends on DDB1, Cul4 and Roc1a (Fig. 4, Fig. S5A,A′), although the specific ubiquitylated target has not yet been identified. We suggest that Xrp1 is likely to be activated by a protein, or proteins, that are normally degraded by Mahj-dependent ubiquitylation, because Xrp1 is also activated by inhibition of the proteasome (Fig. 4D,D′, Fig. S5B-C′), which is expected to affect the degradation of ubiquitylated proteins, but not other functional consequences of ubiquitylation. The relevant target does not seem to be Warts, despite the fact that levels of Warts and Hippo pathway activity also control cellular growth (Pantalacci et al., 2003; Udan et al., 2003; Wu et al., 2003; Huang et al., 2005) and global translation levels (Fig. 6D,D′, Fig. S7C-D′; Nagata et al., 2022), and can stimulate cell competition (Tyler et al., 2007; Neto-Silva et al., 2010). These studies support the notion that Xrp1 is a sensor of multiple cellular defects that cause cell competition, rather than that of a ‘loser signature’ common to distinct cell competition mechanisms. Another group has also reported that Xrp1 is required for cell competition of mahj mutant clones, but without the further analysis described here (Langton et al., 2021).

Mahj was previously thought to be responsible for the cell competition of cells mutated for lgl (Tamori et al., 2010; Baker, 2011; Levayer and Moreno, 2013), a gene that controls apical basal cell polarity (Bilder et al., 2000; Humbert et al., 2003). Mahj was originally linked to apical-basal polarity because of a physical interaction with Lgl, and because Mahj overexpression can rescue lgl mutant clones for elimination, suggesting that Mahj behaves as an intracellular signal transducer of lgl activity in cell competition (Tamori et al., 2010). As such, it was surprising when similar gene expression changes were observed in mahj mutant and Rp mutant wing discs, because these were assumed to reflect distinct cell competition pathways and suggested a common gene expression signature associated with competed cells (Kucinski et al., 2017). We show here, however, that neither lgl nor scrib, another related cell polarity gene (Bilder and Perrimon, 2000), affects cell competition by the same mechanism as mahj, because neither lgl nor scrib mutant cells express or require Xrp1 (Fig. 7, Fig. S8). Interestingly, several distinct pathways have recently been described to mediate the elimination of scrib mutant cells in competition with wild-type cells, and none of these pathways are shown to be required for the elimination of Rp/+ mutant cells (Vaughen and Igaki, 2016; Yamamoto et al., 2017). In addition, mahj loss by itself does not result in apical-basal polarity defects (Tamori et al., 2010), and its mammalian homolog is implicated in cell cycle regulation, genome integrity and p53 activity (Hrecka et al., 2007; Cooper and Giancotti, 2014; Lubow and Collins, 2020). Drosophila mahj, which is an essential gene, regulates neural stem cell reactivation (Ly et al., 2019) and may have other roles in non-neuronal tissues, as suggested by defects observed when mahj is depleted in posterior wing compartments (Fig. 7H-I″). Accordingly, we conclude that mahj mutants affect cellular growth and cell competition in a manner unrelated to lgl and scrib, and that the functional relationship of mahj to apical-basal polarity pathways, should any exist, is unclear (Fig. 7J). The functional importance of physical interaction between Mahj and Lgl remains to be explored. It is known that lgl clones are rescued by reduced Hippo signaling (Menendez et al., 2010; Khan et al., 2013), although we did not detect reduced Hippo signaling after mahj overexpression in the absence of lgl mutations (Fig. S7A-B′).

Our studies provide further evidence for Xrp1 as an integrator of multiple seemingly independent cellular defects that each result in a common spectrum of cellular responses and predispose cells to competitive elimination by wild-type neighbors (Kiparaki et al., 2022). These functional roles for Xrp1 first became apparent through its role in the slow growth, reduced translation and competitive elimination of Rp/+ cells, in which Xrp1 expression is induced in an rpS12-dependent manner (Lee et al., 2018; Ji et al., 2019; Kiparaki et al., 2022). In the case of mahj, Xrp1 protein expression is induced to confer a very similar spectrum of cellular effects, but independently of rpS12 and perhaps depending on stabilization of a protein normally targeted for proteasomal turnover by mahj-dependent ubiquitylation. Xrp1 expression was first found as a p53-regulated gene, perhaps part of the DNA damage response (Brodsky et al., 2004; Akdemir et al., 2007). Recently, Xrp1 induction has also been found as a response to ER stress, possibly through translational regulation downstream of eIF2α phosphorylation (Brown et al., 2021; Ochi et al., 2021; Kiparaki et al., 2022). It has been suggested that eIF2α phosphorylation, and Xrp1 expression, can also be triggered by a global, cytoplasmic proteotoxic stress, which is suggested to occur as a consequence of deficient ribosome assembly in Rp mutant cells (Baumgartner et al., 2021; Langton et al., 2021; Recasens-Alvarez et al., 2021). Xrp1 expression in response to proteasome inhibition is one piece of evidence for this model (Fig. 4D,D′, Fig. S5B-C′). We show here, however, that Xrp1 is induced, and cell competition results after loss of mahj, a single E3-ligase adapter protein that probably targets only a moderate number of proteins for degradation. Thus, an alternative explanation of Xrp1 induction after proteasome inhibition is that this could reflect stabilization of one or a few specific proteins. Overall, a picture is emerging of Xrp1 as a stress-responsive transcription factor whose expression can be initiated by multiple distinct pathways, then leading to a common cellular response, including the elimination of the stressed cells by competition with nearby wild-type cells, when such cells are available (Kiparaki et al., 2022).

Importantly, cells depleted for DCAF1/VprBP, the mammalian homolog of Mahj, are eliminated by competition with wild-type cells in mammalian cell culture (Tamori et al., 2010). Thus, cell competition of mahj mutant cells may be a conserved process. Conservation of cell competition has not yet been demonstrated for Rp/+ cells in mammals, although it may very well occur (Oliver et al., 2004; Baker, 2020). In mammals, knockdown of either mahj or its binding partner ddb1 results in P53 activation, which is functionally required for the resulting phenotypes (Cang et al., 2006; Han et al., 2020). Differences in p53 activity lead to cell competition in many mammalian systems (Baker, 2020). p53 is not required for mahj-mediated cell competition in Drosophila (Fig. 3A-B′, Fig. S3A-G), but because Xrp1 is a target of Drosophila p53 in irradiated cells, it is possible Xrp1 is a p53 target that has replaced the cell competition role of p53 in Drosophila, as has already been suggested for the competition of Rp/+ cells, which is also p53 independent in Drosophila, although Rp mutations activate p53 in mammals (Baker et al., 2019). Thus, mahj-mediated cell competition may provide another example where Xrp1 mediates a process in Drosophila that is dependent on p53 in mammals.

All fly stocks and crosses were maintained at 25°C unless otherwise mentioned. The following fly stocks were used in this study: FRT42 mahj¹ (Tamori et al., 2010), hsflp UAS GFP; FRT 42 tubgal80; tub Gal4 (a gift from D. J. Pan, UT Southwestern Medical Center, Dallas, TX, USA), hsflp; FRT40tubgal80; tubGal4UAS-GFP (a gift from J. Secombe, Albert Einstein College of Medicine, New York, USA), Mahj RNAi (BL:34912), Xrp1 RNAi [Vienna Drosophila Resource Center (VDRC): 107860], Xrp1 RNAi (Bloomington Drosophila Stock Center: 34521), Xrp1⁰²⁵¹⁵ (Spradling et al., 1999), Xrp1-HA (Blanco et al., 2020), En-Gal4-UAS GFP, White RNAi (Bloomington Drosophila Stock Center: 33623), Mahj^Df (Bloomington Drosophila Stock Center: 5764), 109-68-GAL4 (Bloomington Drosophila Stock Center: 6479), lgl⁴ (Tamori et al., 2010), Scrib¹ (Bilder et al., 2000), P53 RNAi (Bloomington Drosophila Stock Center: 41720), P53DN (Bloomington Drosophila Stock Center: 8420), rpr150-LacZ (Brodsky et al., 2000), Xrp1^{attp FLOX} (Blanco et al., 2020), rps12^G97D (Kale et al., 2018), diap-lacZ (Wu et al., 2008), ex-LacZ (Blaumueller and Mlodzik, 2000), fj-LacZ (Villano and Katz, 1995), UAS-hpo (Udan et al., 2003), UAS-yki, UAS-ddb1 RNAi (Bloomington Drosophila Stock Center: 41997), cul4 RNAi (Bloomington Drosophila Stock Center: 50614), UAS-FLAG-Cul4^KD (Lin et al., 2009), FRT82B Ubi-mRFP (Bloomington Drosophila Stock Center: 30555), UAS-ban (Thompson and Cohen, 2006), FRT82 Xrp1^M2-73 (Lee et al., 2018), nub-Gal4 UAS RFP (Bloomington Drosophila Stock Center: 63148), prosalpha5 RNAi (Bloomington Drosophila Stock Center: 34786), prosbeta5 (Bloomington Drosophila Stock Center: 34810), prosbeta6 (Bloomington Drosophila Stock Center: 34801), UAS-PPP1R15 (Bloomington Drosophila Stock Center: 76250), Wts RNAi (VDRC: 106174), UAS-Wts-myc (Ch-II, a gift from Kenneth Irvine's lab, Waksman Institute, Piscataway, NJ, USA), Roc1a RNAi (VDRC: 32399) and P{GAL4-Act5C(FRT-CD2).P}S (Bloomington Drosophila Stock Center: 51308).

Loss- or gain-of-function somatic clones were generated using FLP/FRT-mediated mitotic recombination. To induce somatic loss- or gain-of-function clones using heat shock flippase (hsflp), heat shock was given for either 15 min or 30 min for cis or trans-chromosomal recombination, respectively, at 60±12 h after egg laying, and dissection was carried out 72±12 h after heat shock. To control hs-FLP copy number in different genotypes, only male larvae were selected for experiments where clone areas in wing discs of different genotypes were compared.

Wandering third instar were dissected in 1×PBS buffer and fixation was carried out in 4% paraformaldehyde (in 1×PBS buffer). Fixed imaginal discs were washed in PBT (0.3% Triton X-100, 1×PBS) three times for 10 min each. Incubation of imaginal discs with primary antibody was carried out overnight at 4°C. After overnight incubation, washing was carried out thrice with PBT for 10 min each. Secondary antibody incubation was carried out at room temperature followed by three washes with PBT for 10 min each. Primary antibodies used in this study were mouse anti-βGal (1:100, Developmental Studies Hybridoma Bank, mAb40-1a) (Ghattas et al., 1991), rabbit anti-active-Dcp1 (1:100, Cell Signaling technology, 9578), mouse anti-HA (1:100, Cell Signaling Technology, 2367) and rabbit pJNK (1:200, Promega, V793B). Secondary antibodies used were Cy2 and Cy5 conjugates (1:200, Jackson ImmunoResearch, 115-225-166 and 711-175-152, respectively), and goat anti-mouse and anti-rabbit Alexa Fluor 555 (1:400, Thermo Fisher Scientific, A11001 and A21429, respectively). To measure global translation, a Click-iT Plus OPP Alexa Fluor 594 Protein Synthesis Assay Kit (Thermo Fisher Scientific, C10457) was used as described previously (Lee et al., 2018). For lysosomal activity detection, Lysotracker red DND-99 (Thermo Fisher Scientific) was used at the concentration of 4 μM and staining was carried out in Schneider's Drosophila Medium with 10% FBS for 40 min. To measure cell death after irradiation, larvae within food vials were exposed to 4000 rad γ-rays at 72±12 h after egg laying while dissection was carried out 72 h post irradiation. Images were acquired using SP8 confocal microscopes (Leica) followed by processing of images with NIH ImageJ and Adobe Photoshop software. ImageJ was used to measure clone area and clone perimeter, and to quantify fluorescence signal intensity of a wing disc. To make z projections for the wing disc with mutant or control clones, equal numbers of sections were combined while avoiding basal sections. Dying cells marked by anti-Dcp1 on the clone perimeter were counted as competitive cell death. To calculate dying cells per unit of clone perimeter, the total number of dying cells at the clone boundary were divided by the length of the clone perimeter of a wing disc. Analysis was not carried out blind. All samples acquired were analyzed and no method of randomization used. Number of wing discs analyzed are reported as n values in figure legends. Sample sizes were based on previous experience with similar experiments and not determined by statistical methods. All experiments assessing cell competition and all but one experiment overall were independently performed on at least two occasions. The exception was p53 RNAi after irradiation (Fig. S3C). Data presented graphically represent mean values, e.g. mean clone area or mean cell death. Error bars represent ±1 s.d. Statistical comparisons were generally made using t-test assuming normal distribution. When n<10, or for ratios between clone and twin-spot sizes, the Mann–Whitney test was used. The Wilcoxon matched-pairs test was used to compare fluorescence signal of anterior and posterior signal (Fig. S4G). No adjustments were made for multiple comparisons. P-values are provided in the figure legends.

We thank Drs W.-M. Deng, C.-T. Chien, K. Irvine, D. Pan, J. Secombe and H. Wang for fly stocks. Other fly stocks were obtained from the Bloomington Drosophila Stock Center (supported by NIH P40OD018537) and the Vienna Drosophila Resource Center. We thank Drs A. Jenny, C. Khan, M. Kiparaki and S. Nair for comments on manuscript, Dr Kiparaki for sharing unpublished data, and Dr Khan for sharing unpublished fly stocks. Confocal microscopy was performed in the Analytical Imaging Facility of the Albert Einstein College of Medicine (supported by the NCI P30CA013330) using the Leica SP8 microscope acquired through NIH SIG 1S10 OD023591. The monoclonal antibody mAb40-1a developed by J. R. Sanes was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD and maintained at the University of Iowa, Department of Biology, Iowa City, IA 52242.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200795.reviewer-comments.pdf.