Normal cell cycle progression requires negative regulation of E2F1 by Groucho during S phase and its relief at G2 phase

The cell cycle comprises a programmed sequence of events, including DNA synthesis, chromosome separation and cytokinesis. Progression through the cycle is highly regulated by protective checkpoints at which intrinsic or extrinsic conditions are monitored. The cycle arrests if suboptimal conditions are sensed, and recommences only once these are resolved through appropriate cellular response mechanisms (Clarke and Giménez-Abián, 2000; Kastan and Bartek, 2004; Musacchio and Salmon, 2007).

Cell cycle-related proteins are regulated at different levels. In the early Drosophila embryo, before the maternal-to-zygotic transition, rapid nuclear divisions are mainly controlled via translational and post-translational regulation of maternally-deposited determinants (Groisman et al., 2002; Becker et al., 2018). As zygotic transcription is induced, additional mechanisms are incorporated. For example, a surge of string (cdc25) expression is responsible for introducing the Gap 2 (G2) phase in otherwise Synthesis (S)-to-Mitosis (M) cycling embryonic cells (Edgar et al., 1994a; Budirahardja and Gönczy, 2009). Other genes that play a part in the cell cycle are also transcriptionally regulated, e.g., those involved in processes such as DNA replication and chromosome segregation, many of which are highly expressed in human tumours (Whitfield et al., 2002). Thus, transcriptional regulation in the context of the cell cycle is an important path to explore.

E2F1, a key cell cycle transcription factor, activates expression of cyclin E (cycE) and other target genes, the protein products of which are required at the initiation of S phase, and at the G2 and M phases (Bennett et al., 1996; Dimova et al., 2003; Dimova and Dyson, 2005; Herr et al., 2012). The E2F1 protein is detectable during all stages of the cell cycle except for S phase, when it is degraded (Reis and Edgar, 2004; Shibutani et al., 2008; Davidson and Duronio, 2012). Preventing E2F1 protein degradation during S phase leads to acceleration of the cell cycle and/or to apoptosis, highlighting the requirement for its removal at this stage (Shibutani et al., 2008). The activator functions of E2F1 and other cell cycle factors have been studied in depth; less is known, however, about transcriptional repressors in this process.

Transducin-like enhancer of split 1 (TLE1), a human orthologue of the Drosophila developmental corepressor Groucho (Gro), acts as an anti-proliferative factor (Zahavi et al., 2017), and TLE family members have been linked to human cancers (Kokabu et al., 2017; Dali et al., 2018). Furthermore, Gro and TLE1 are both phosphorylated in a cell cycle-regulated manner in vitro and in cultured cells (Nuthall et al., 2002), a modification previously shown to mitigate their corepressor activity downstream of receptor tyrosine kinase (RTK) pathways (Paroush et al., 1997; Hasson et al., 2005; Cinnamon et al., 2008; Helman et al., 2011, 2012; Johnston et al., 2016; Zahavi et al., 2017). Herein, we explore the possibility that Gro fulfils an in vivo regulatory function in the cell cycle.

We now demonstrate that Gro-mediated repression at S phase, and its relief at G2 phase, are both crucial for proper cell cycle progression. Specifically, we find that Gro is unphosphorylated and therefore active as a repressor only at S phase. We show that Gro binds within and around the e2f1 gene locus, and that it is a negative regulator of its expression. gro-deficient cells display accelerated cell cycles and accumulate at Gap 1 (G1) phase, phenotypes resembling those caused by E2F1 overexpression. In addition, we find that overexpression of Gro causes cells to accumulate at G2 phase, and that Cyclin-dependent kinase 1 (Cdk1) normally phosphorylates Gro at this stage in vivo, suggesting that this mechanism attenuates the repressor activity of Gro and permits entry into mitosis. Together, our results reveal a novel role for Gro in the cell cycle, showing that it switches between active and inactive states and restricts gene expression in a phase-specific manner. We propose that a similar level of regulation underlies the involvement of TLE corepressors in cancer.

Gro is uniformly expressed throughout Drosophila development (Delidakis et al., 1991). In any given nucleus, however, it is either primarily phosphorylated or mainly unmodified (Fig. S1) (Cinnamon et al., 2008; Helman et al., 2011; Johnston et al., 2016). Previous work demonstrated that phosphorylation of Gro, particularly by extracellular signal-regulated kinase (Erk) in response to RTK signalling, downregulates its repressor function (Hasson et al., 2005; Cinnamon et al., 2008; Helman et al., 2011, 2012). Thus, Gro has a regulated ability to switch between two modes: an active (unphosphorylated) or an inactive (phosphorylated) corepressor (Fig. S2).

Double staining of cycling cells in imaginal discs, with antibodies that are largely specific for Gro and for phosphorylated Gro (pGro; Materials and Methods), revealed that they display limited overlap in their respective signals. In both wing and eye imaginal discs, only 10-20% of cells showed overlapping nuclear signals for both antibodies (Fig. 1A,A′; Fig. S1) (Cinnamon et al., 2008; Helman et al., 2011; Johnston et al., 2016). Using these two antisera, we show below that phosphorylation of Gro fluctuates dynamically in a cell cycle phase-dependent manner, and that it is predominantly modified at all stages of the cell cycle except for S phase.

Specifically, mitotic cells, distinguishable by anti-phospho-Histone 3 at Serine 10 (pH3) staining, were generally positive for pGro, both in embryos (Fig. 1B-C′) as well as in wing imaginal discs (Fig. 1F-G′,J,J′). In contrast, the pH3 signal complemented that of the unphosphorylated active form of Gro in both tissues (Fig. 1D-E′,H-I′,J,J′), indicating that Gro is largely phosphorylated, and therefore inactive as a repressor, at mitosis.

Not all pGro-positive nuclei were mitotically active, however (Fig. 1C′,G′); Gro was also phosphorylated at the G1 and G2 phases as determined using Fly-FUCCI, an in vivo fluorescent, ubiquitination-based indicator system (Fig. 2) (Zielke et al., 2014). In Fly-FUCCI wing imaginal discs, individual cells appear in different colours according to their cell cycle stage (Fig. 2A-B′,E-E′). Staining with anti-pGro and anti-Gro antibodies demonstrated that Gro is mostly phosphorylated at the G2/M and G1 phases in asynchronously-dividing cells throughout the wing imaginal disc. This was best seen in cells along the prospective dorsoventral boundary of the wing pouch (Fig. 2A), where whole populations of cells reduce their proliferation rate and are found either at the G1 or G2 phase of the cycle (Fig. 2B-G′,H). We surmise that Gro is extensively phosphorylated and, consequently, mostly inactive at M, G1 and G2 phases.

The Fly-FUCCI system also indicated that Gro is unphosphorylated at S phase. To confirm this, we carried out two additional experiments. First, wing imaginal discs of PCNA-GFP flies, in which expression of cytoplasmic GFP is a reliable marker for early S phase (Thacker et al., 2003; Strzalka and Ziemienowicz, 2011), were stained for pGro. As Fig. 3A-C shows, the pattern of GFP largely complemented that of pGro. Second, we stained wing imaginal discs with the nucleoside thymidine analogue 5-ethynyl-2′-deoxyuridine (EdU), which labels cells at late S phase (Buck et al., 2008). Here, too, we found that EdU-positive cells mainly co-stain for unphosphorylated Gro but not for pGro (Fig. 3D-F; J-J′).

A similar result is also observed in eye imaginal discs, which allow the analysis of a relatively synchronised S phase cell population. In this tissue, differentiation proceeds as a wave across the disc such that a stereotypic stripe of undifferentiated cells, located posteriorly to the morphogenetic furrow, synchronously enter S phase (arrows; Fig. 3G-I) (Wolff and Ready, 1991). The vast majority of these EdU-positive S-phase cells stained for Gro but not for pGro (Fig. 3G-H′,J,J′), leading us to conclude that, in cycling cells, Gro is unphosphorylated only at S phase. Its repressive activity is, therefore, restricted to this specific stage of the cell cycle.

The roughly non-overlapping patterns of anti-pH3 and anti-Gro staining (Fig. 1) suggest that Gro is primarily phosphorylated in mitotic cells, and therefore inactive at this stage. To determine the significance of Gro phosphorylation to mitosis, we assessed the effects of expressing a non-phosphorylatable Gro mutant on the number of mitotic cells in the rapidly dividing wing imaginal disc. Towards this end, we employed a non-phosphorylatable Gro variant mutated in its two phosphoacceptor sites (Gro^AA), in addition to a phosphomimetic Gro mutant derivative (Gro^DD) (Hasson et al., 2005). If phosphorylation of Gro is a precondition for mitosis, we expect the expression of the constitutively active Gro^AA repressor to dominantly reduce the pH3 signal, but not expression of Gro^DD, the repressor activity of which is attenuated (although recognised by anti-panTLE antibodies, we cannot formally rule out the possibility that Gro^DD is partially inactive due to misfolding; Fig. S3) (Hasson et al., 2005; Cinnamon et al., 2008; Helman et al., 2011, 2012; Johnston et al., 2016).

The two variants were ectopically expressed in wing imaginal discs under Gal4/UAS control (Brand and Perrimon, 1993). Expression of Gro^AA using an early driver (nubbin-Gal4) resulted in small dysmorphic discs, preventing their subsequent analysis. Instead, we used MS1096-Gal4, which drives non-uniform expression relatively late and mainly in the dorsal wing compartment (Fig. S3) (Milán et al., 1998). In this case, expression of Gro^AA, but not of the phosphomimetic Gro^DD variant, reduced the number of pH3-positive cells compared with control lacZ-expressing wings (Fig. 4A-D,G,H), and the ensuing adult wings were markedly smaller (Fig. S3). These results suggested that phosphorylation of Gro correlates with progression to mitosis [other, unrelated functions of Gro in imaginal disc development probably also contribute to the severe adult wing phenotypes brought about by its overexpression (Cavallo et al., 1998; Hasson et al., 2001, 2005)]. There was also a significantly greater overlay between the pH3 and Gro^DD signals than between those of pH3 and Gro^AA (Fig. S3), indicating that cells expressing Gro^DD, but not those expressing Gro^AA, more readily enter mitosis. Phosphorylated Gro is, therefore, compatible with M phase.

Induced expression of native Gro also caused a significant reduction in the number of pH3-positive mitotic cells, both in wing imaginal discs (Fig. 4E-H) as well as in the embryo, which represents a different developmental tissue [using the Krüppel (Kr)-Gal4 driver; Fig. 4I-J′] (Chu et al., 1998). To quantify Gro's effects on mitosis, while controlling for changes in cell size exerted by its expression, we calculated the ratio between the number of pH3-positive cells and the total number of cells in the wing pouch region (mitotic index). We found that the mitotic indices of wing imaginal discs overexpressing Gro (44.4%) and Gro^AA (48.8%), but not Gro^DD (104%), were significantly lower than the mitotic index observed in control lacZ-expressing discs (related to as 100%; Fig. 4K). Notably, expression of native transcriptional regulators and their non-phosphorylatable derivatives often exerts similar effects in other biological settings [e.g. figure 4 in Cinnamon et al. (2008); figures 3D-F and 4C in Li et al. (2016) and figure 5A,B in Lee et al. (2010)].

Strikingly, irregular Gal4-driven expression (Fig. S3) (Ward et al., 2002; Barth et al., 2012) revealed that the negative effect of Gro^AA and Gro on mitosis is largely cell-autonomous: most of the remaining pH3-positive cells in the MS1096- and Kr-Gal4 expression domains were consistently those that expressed null or low levels of induced Gro^AA and Gro (Fig. 4B,F,L; Movie 1). The discordance between the anti-Gro and -pH3 signals, which is also largely apparent in wing discs overexpressing Gro^DD and lacZ (Fig. 4D,H), is consistent with the idea that entry into M phase necessitates the previous attenuation of Gro repressive activity by phosphorylation.

Given that phosphorylation of Gro fluctuates dynamically in a cell cycle phase-dependent manner, Gro must be phosphorylated by a kinase that is active at G2 phase but inactive at S phase. Although Erk activity impinges on cell cycle regulation at different levels (Prober and Edgar, 2000; Mogila et al., 2006; Mirzoyan et al., 2019), there is no evidence that it exhibits similar activation and inactivation dynamics. Instead, we considered Cdk1 as the prime candidate for phosphorylating Gro in this context (Vidwans and Su, 2001). Cdk1 activity shows the appropriate dynamics for regulating Gro activity, being essential at G2 and M phases but inactive at S phase (Bettencourt-Dias et al., 2004). In addition, Cdk1 phosphorylates Gro and its human TLE1 orthologue in vitro and in cell culture (Nuthall et al., 2002). Finally, Cdk1 is a proline-directed serine/threonine kinase (Malumbres, 2014) that, like Erk, is expected to target the Gro proline-glycine-threonine-proline motif. Indeed, the Gro Cdk1 phosphorylation site was previously mapped to a fragment encompassing the Erk phosphorylation motif recognised by the anti-pGro antibodies (Materials and Methods) (Nuthall et al., 2002).

We first confirmed, biochemically, that a purified, activated Cdk1/CyclinB1 complex can phosphorylate recombinant Gro protein in vitro. We found that Cdk1 phosphorylates threonine 308, the same phosphoacceptor amino acid that is modified by Erk and recognised by anti-pGro antisera (Fig. 5A). To test whether Gro also undergoes phosphorylation by Cdk1 in vivo, RNA interference (RNAi) was used to deplete it. Expression of cdk1 RNAi in wing imaginal discs resulted in fewer and larger cells, a known phenotype caused by Cdk1 deficiency (Fig. S4) (Weigmann et al., 1997; Johnston, 1998; Bettencourt-Dias et al., 2004). As Fig. 5 shows, the amount of unphosphorylated Gro increases in such MS1096>cdk1 RNAi wing imaginal discs, predominantly in the dorsal compartment (Fig. 5B-F) (the possibility that phosphorylation by Cdk1 also affects Gro protein stability cannot be ruled out) (Milán et al., 1998). Conversely, the pGro signal decreases in cdk1 RNAi wing imaginal discs, although, due to high variability, this reduction is statistically non-significant (Fig. S4). A comparison between the anterior and posterior compartments of en>cdk1 RNAi-expressing wing imaginal discs, however, disclosed a significant reduction in the pGro signal upon Cdk1 knockdown (Fig. S5).

The observed increase in unphosphorylated Gro is not simply due to a G2/M phase arrest caused by cdk1 depletion (Fig. 5G) (Bettencourt-Dias et al., 2004), as Gro is normally phosphorylated at these stages of the cell cycle (Fig. 2). Our results are, therefore, consistent with Cdk1 being the kinase responsible for inactivating Gro during G2 phase in vivo. Noteworthy, Gro is predominantly phosphorylated at G1 phase when Cdk1 is inactive, possibly due to the persistence of its phosphorylation state (Cinnamon et al., 2008; Helman et al., 2011) or via some other kinase(s).

The Cdk1-mediated phosphorylation of Gro at the exit from S phase appears to be required for normal cell cycle progression. When Gro is overexpressed in Fly-FUCCI wing imaginal discs, a substantial enrichment in yellow-stained, G2/M-phase cells relative to control discs occurs (Fig. 5H-J). As similar Gro expression leads to a cell-autonomous reduction in pH3 staining (Fig. 4), we concluded that the yellow staining reflects a significant increase in the number of cells at G2 phase.

To confirm this result, we conducted flow cytometric analysis of dissociated wing imaginal disc cells. As Fig. 5K shows, a higher proportion of cells ectopically expressing Gro are in G2/M phases compared with cells from control wings. Gro-expressing cells are also larger in size than control cells (Fig. 5K′) (Bettencourt-Dias et al., 2004). These results, together with the largely overlapping expression patterns of pH3 and pGro (but not Gro) staining (Fig. 1) and the Gro-induced decline in pH3-positive mitotic cells (Fig. 4), indicate that cells overexpressing Gro accumulate at G2 phase, before mitosis, leading us to conclude that unphosphorylated, repressive Gro hinders normal cell cycle progression.

An in silico approach was used to identify potential cell cycle targets of Gro-mediated repression. A list of genes was first compiled based on their association with the terms ‘cell cycle’ and/or ‘cell proliferation’. This list was then intersected with lists of genes that are located in proximity to Gro peaks, drawn from various genome-wide studies that profiled Gro binding to chromatin (Orian et al., 2007; Shibutani et al., 2008; Roy et al., 2010; Kaul et al., 2014; Chambers et al., 2017). Candidate Gro-regulated target genes were then ranked based on the number of lists they appear in (Table S1).

From this gene set we chose, as case in point, the e2f1 gene for further investigation. e2f1 stands out as an attractive Gro target, as it encodes a key cell cycle regulator (Asano et al., 1996; Neufeld et al., 1998). The E2F1 protein is detectable at all phases of the cell cycle except for S phase, when it is degraded (Bennett et al., 1996; Dimova et al., 2003; Davidson and Duronio, 2012; Herr et al., 2012). The inappropriate presence of E2F1 at S phase can lead to accelerated cell cycle and/or apoptosis (Bennett et al., 1996; Dimova et al., 2003; Davidson and Duronio, 2012; Herr et al., 2012). Although post-translational regulation of E2F1 has been extensively studied, regulation of e2f1 gene expression is, as yet, largely unexplored (Johnson et al., 1994; Goulev et al., 2008; Øvrebø and Bradley-Gill, 2022).

We confirmed Gro binding to the e2f1 locus by performing chromatin immunoprecipitation assays with sequencing (ChIP-seq) using ML-DmBG3-c2 (BG3) cells, which originate from the larval central nervous system (Cherbas et al., 2011). We also re-examined ChIP-seq data previously generated from two embryonic plasmatocyte-derived cell lines, Kc167 and S2R+ (Kaul et al., 2014). In all three cell lines, we found a shared cluster of Gro peaks within the e2f1 gene that maps to the first intron of the e2f1-RA and e2f1-RB transcripts, and is upstream of the transcription start sites of four other e2f1 transcripts (RC, RD, RE and RF) (Fig. 6A; boxed). Other common peaks of Gro binding are found just downstream of the e2f1 gene (Fig. 6A). The widespread recruitment of Gro in and around the e2f1 locus in varied cell types, derived from different developmental stages, is consistent with e2f1 being a general, rather than a cell-type specific, Gro target.

Several analyses support the notion that Gro represses e2f1 gene expression during S phase, and that phosphorylation of Gro attenuates this repression. Immunofluorescence staining of wild-type wing imaginal discs shows that signals of E2F1 and non-phosphorylated repressive Gro do not greatly overlap (Fig. 6B,C). In addition, ectopic Gro^AA expression significantly reduces anti-E2F1 staining, whereas Gro^DD expression does not affect E2F1 levels or distribution (Fig. S6). Moreover, real-time polymerase chain reaction (RT-PCR) assays revealed that overexpression of Gro in wing imaginal discs decreased the relative transcript levels of e2f1 and several of its targets (Fig. 6D). Importantly, E2F1 is derepressed in homozygous gro^E48 mutant clones, induced in either wing (Fig. 6E-G) or eye imaginal discs (Fig. 6H-J). Finally, we used an e2f1-lacZ enhancer-trap line to distinguish between a role for Gro in the transcriptional regulation of the e2f1 gene or in the post-translational control of E2F1 protein levels (Duronio et al., 1995; Brook et al., 1996). As shown in Fig. 6K-N, Gro negatively regulates reporter expression derived from this enhancer-trap in eye imaginal discs, indicating that, directly or indirectly, it represses e2f1 transcription.

We reasoned that if the cell accumulation at G2 phase, induced by overexpression of Gro, involves the inappropriate repression of e2f1 at G2 phase, a stage when it should be re-transcribed, then the concomitant expression of e2f1 will rescue this phenotype. As Fig. 5L,L′ shows, the co-expression of E2F1 and Dimerization partner (Dp) suppresses the G2-phase accumulation prompted by Gro overexpression, and cell size reverts to normal. This functional link between Gro and E2F1 reinforces the notion that relief of Gro-mediated e2f1 repression at G2 phase, via Cdk1-dependent phosphorylation, is crucial.

The derepression of E2F1 does not induce apoptosis of gro-deficient cells (Fig. S7), perhaps because the level of upregulated E2F1 in gro mutant cells is still below the threshold required to trigger cell death that is attained by E2F1 overexpression. To establish the significance of Gro-dependent repression at S phase, we determined how genetic depletion of gro affects the fate of eye imaginal disc cells posterior to the morphogenetic furrow. These cells are normally in S phase and, therefore, are EdU-positive (Fig. 3G-I) (Wolff and Ready, 1991). When homozygous mutant gro^E48 clones intersect this domain, cells in which E2F1 is derepressed and consequently upregulated are consistently EdU-negative, indicating that they are no longer in S phase (Fig. 7A,A′). The minority of gro-deficient cells that still stain for EdU, usually found at the margins of the clones, are typically E2F1-negative (Fig. 7A,A′; see Discussion).

We hypothesised that if Gro-mediated repression at S phase is crucial, then its absence might perturb proper cell cycle progression. To address this point, we profiled the cell cycle distribution of GFP-marked eye clones of cells heterozygous for the gro^MB36 null allele, in which gro levels were further knocked-down (gro RNAi) (Fig. S1). Flow cytometry analyses revealed that a greater proportion of GFP-positive cells with reduced gro levels have a 2N DNA content (Fig. 7B). Moreover, such eye imaginal disc clones show reduced staining for S, G2 and M phase markers (Fig. S7). They, therefore, accumulate predominantly at G1 phase.

We considered two possibilities that can explain this result: gro depletion either halts the cell cycle before DNA synthesis or accelerates the cell cycle such that cells pass through G2/M phases faster. To distinguish between these options, we determined the doubling times of clones of cells anterior to the morphogenetic furrow in eye imaginal discs overexpressing lacZ or, similarly, of cells in which RNAi was used to knock down gro, by quantifying the number of GFP-positive cells per clone (Fig. 7C). We found that the respective doubling time of lacZ-expressing control clones was, on average, 15.9 h as previously reported (Neufeld et al., 1998; Tseng and Hariharan, 2002). Strikingly, gro knockdown clones exhibit a significantly faster doubling time of 12 h. The accelerated cell cycle phenotype of cells with lower levels of gro is similar to that caused by E2F1 overexpression (Shibutani et al., 2008; Davidson and Duronio, 2012). A comparison between the average lengths of the different cell cycle stages in gro knockdown clones, relative to controls, revealed that the acceleration in the doubling time of the former stems from a shorter G2/M phase duration. Specifically, G1 phase in lacZ-expressing cells took 8 h, the duration of S phase was 2.44 h and G2/M phase was 5.3 h long. In gro RNAi cells, the corresponding times were 8.36 h, 2.15 h and 1.45 h, respectively. The higher proportion of cells at G1 phase in gro knockdown clones (Fig. 7B) resembled the accumulation of cells overexpressing E2F1 at G1 phase (Shibutani et al., 2008; Davidson and Duronio, 2012), supporting the notion that Gro is a negative regulator of e2f1 at S phase.

In summary, our data indicate that Gro represses e2f1 and possibly other genes at S phase. In the absence of Gro, the cycle accelerates and cells ultimately accumulate at G1 phase. Conversely, relief of Gro-mediated repression by Cdk1 activity is required for the correct progression through the G2 phase and for entry into mitosis (Fig. 7D). In light of the phenotypes caused by both depletion and overexpression of Gro, we conclude that Gro is a modulator of cell cycle regulation.

Our results uncover a previously unrecognised tier of cell cycle regulation, and identify e2f1 as a key target of Gro-mediated repression at S phase (Fig. 7D). Several observations point to a direct effect of Gro on e2f1. First, Gro is bound to chromatin within the e2f1 gene locus in multiple cell lines and in Drosophila embryos (Fig. 6) (Orian et al., 2007; Roy et al., 2010; Kaul et al., 2014; Chambers et al., 2017). Second, Gro represses reporter gene expression derived from an e2f1-lacZ enhancer-trap line, which includes the sequences to which Gro binds (Fig. 6K-N) (Duronio et al., 1995; Brook et al., 1996). Third, E2F1 is repressed in cells that accumulate at G2 phase in response to ectopic Gro expression (Figs 5 and 6), despite the fact that it should be normally expressed at this stage (Shibutani et al., 2008; Davidson and Duronio, 2012). Thus, Gro negatively regulates E2F1, a transcription factor that functions at the heart of the cell cycle by activating multiple genes required for the initiation of the cycle, as well as for the G2 and M phases.

Previous reports have shown that preventing E2F1 proteolysis at S phase brings about accelerated cell cycles and/or apoptosis (Shibutani et al., 2008; Davidson and Duronio, 2012). Repression of e2f1 by Gro does not appear to provide a robust, backup mechanism for preventing E2F1-induced apoptosis, because gro mutant cells are viable despite the upregulation of E2F1 (Figs 6, 7; Fig. S7). We surmise that elevated E2F1 protein levels in gro-deficient cells, in which the E2F1 degradation apparatus is presumably still operational, are not high enough to instigate cell death. Instead, E2F1 upregulation associated with the loss of Gro leads to accelerated cell cycles.

Surprisingly, a minority of gro-deficient cells do not upregulate E2F1 and still stain for EdU (Fig. 7A,A′). It is conceivable that these EdU-positive cells, usually found at the periphery of gro clones, are exposed to non-cell autonomous signals from surrounding cells that drive them into S phase, when E2F1 is robustly degraded. Their ability to continue dividing could explain why gro mutant clones are not eliminated.

Surprisingly, cells overexpressing Gro accumulate at G2 phase, and not at G1 phase as expected from depletion of e2f1 (Dobrowolski et al., 1994; Ishizaki et al., 1996; Wu et al., 1996; Fan and Bertino, 1997; Dyson, 1998). We propose that overexpressed Gro blocks de novo e2f1 transcription at G2 phase (following its repression by Gro at S phase) and, consequently, Gro aborts proper progression of the cycle at this stage. In contrast, the G1-to-S phase transition is refractory to Gro overexpression, as the E2F1 protein has already accumulated by then, and is post-transcriptionally regulated. Consistent with this interpretation, co-expression of E2F1 and Dp suppresses the G2-phase phenotype induced by Gro overexpression (Fig. 5L,L′).

Negative regulation by Gro is one of multiple layers of E2F1 control. Hence, the role of Gro appears to be modest in comparison with that of integral cell cycle regulators like Cdk1, which control diverse targets and processes during G2 phase and mitosis. Yet, E2F1 derepression alone cannot account for the overall outcome brought about by depletion of gro. Knockdown of gro, for example, leads to a shortened duration of the G2/M phases without affecting the length of the G1 phase, whereas overexpression of E2F1 causes shortening of both G1 and G2 phases (Neufeld et al., 1998). It is therefore conceivable that Gro also represses additional genes that must be silenced during S phase, in proximity to which it binds. One such prospective gene is transforming acidic coiled coil (TACC), the protein product of which is required to maintain spindle bipolarity and microtubule stability during mitosis (Trivedi, 2013). Silencing of TACC and/or other genes, besides E2F1, could also contribute to the predominant G2 phase arrest exerted by Gro expression (Fig. 5).

The full scope of Gro regulatory functions in the framework of the cell cycle will come to light once the complete repertoire of its targets is revealed and its DNA-binding partner proteins in this framework identified. How phosphorylated Gro is replaced by unphosphorylated Gro at S phase also remains an unanswered question. Future studies will further determine whether the accelerated cycles in gro-depleted cells lead to DNA damage sensitivity and/or to chromosomal aberrations.

The majority of pGro-positive cells are negative for pH3 staining and are, therefore, not mitotic (Fig. 1). These could be non-dividing cells or, if cycling, could be in one of the two Gap phases at which Gro is also phosphorylated (Fig. 2). Regardless, we propose that for cell division to occur, the attenuation of Gro repression must be accompanied by specific pro-proliferative cues. Such positive inputs, exclusive to this particular cellular process, will ensure that the downregulation of Gro by phosphorylation results in induction of a restricted sets of genes.

To date, Gro has been mainly implicated in transcriptional repression events controlling cell fate specification and differentiation (Hasson and Paroush, 2006; Cinnamon and Paroush, 2008). The new role we have uncovered for Gro and its phosphorylation in the cell cycle raises the possibility that it also functions in additional basic cellular processes. Accordingly, other physiological and/or metabolic processes, each employing a specific effector kinase(s), may induce their unique arrays of downstream target genes via phosphorylation of Gro.

Emerging evidence supports the notion that the Gro human TLE orthologues may act similarly to Gro in the context of the cell cycle. TLE1 and TLE3 are both implicated in cancer; TLE1, for example, promotes glioblastoma propagation (Dali et al., 2018) and TLE3 stimulates cell division by suppressing myogenic differentiation via transcriptional repression of the master regulator MyoD (Kokabu et al., 2017). Moreover, a large-scale analysis found that TLE1 is a mitotic bookmarking factor in development and in stem cells (Festuccia et al., 2017), and TLE3 was identified in a phosphoproteomics analysis of full phosphorylation site occupancy during mitosis (Olsen et al., 2010). Finding that TLE corepressors function comparably to Gro in cell cycle regulation may ultimately offer new therapeutic strategies for preventing uncontrolled cell division in cancerous settings.

Flies were cultured and crossed on standard yeast-cornmeal-molasses-malt extract-agar medium at 25°C. cDNAs for Gro, Gro^AA and Gro^DD were cloned into the pUAST-attB vector, and all constructs were subsequently integrated at the attp40 site (BestGene) to generate transgenic lines with comparable expression levels (Fig. S3) (Markstein et al., 2008).

The following GAL4 drivers and responders were used: kr-Gal4 [Bloomington Drosophila Stock Center (BDSC), #58800]; MS1096-Gal4; gro^MB36, UAS-gro RNAi/TM6B (generously provided by Gerardo Jiménez, Institut de Biologia Molecular de Barcelona-CSIC, Barcelona, Spain) (Jennings et al., 2008); UAS-cdk1 RNAi (BDSC, #28368 and #36117); UAS-cdk2 RNAi (BDSC, #28952 and #34856); UAS-cdk4 RNAi (BDSC, #36060 and #27714); UAS-LacZ (Brand and Perrimon, 1994); UAS-GFP (Yeh et al., 1995); UAS-E2F1, UAS-Dp/CyO (BDSC, #4774); PCNA-GFP (kind gift of Robert Duronio, University of North Carolina, Chapel Hill, USA); P[rm729] e2f1-lacZ (BDSC, #34054); and Fly-FUCCI (BDSC, #55124 and #55123) (Zielke et al., 2014). yellow white flies served as wild-type controls.

Mutant clones lacking functional Gro (gro^E48) were generated using FLP-mediated mitotic recombination in the progeny of the following cross: hsflp; P[FRT82B] ubi-GFP/TM6B virgin females and P[FRT82B] gro^E48/TM6B males. Clones were induced 48-72 h after egg laying by heat-shock (60 min at 37°C) and were identified by the loss of the GFP marker and the concurrent appearance of a twin spot clone. Knockdown of gro was attained by crossing hsflp; actin>CD2>Gal4; UAS-nlsGFP/TM6B virgins to gro^MB36, UAS-gro RNAi/TM6B males (Jennings et al., 2008). Clones, induced 48-72 h after egg laying by heat-shock (10 min at 37°C), were distinguishable via the GFP marker.

Wing imaginal disc lysates were prepared for immunoblotting as previously described (Kushnir et al., 2020). Western blotting was carried out using a standard protocol. Briefly, samples were separated on SDS-PAGE (10 cm × 10 cm) and proteins were electro-transferred at 100 mV for 90 min to 0.45 µm Nitrocellulose membranes (Whatman). Membranes were washed with Tris-buffered saline containing 0.1% Tween 20 detergent (TBST) for 5 min at room temperature and incubated with blocking buffer [1× TBS, 0.1% Tween-20 with 5% w/v nonfat dry milk or 5% bovine serum albumin (BSA)] for 1 h at room temperature. Membranes were subsequently washed three times for 5 min in TBST, and incubated in buffer containing the primary antibody in buffer (1× TBS, 0.1% Tween-20 with 5% BSA) overnight at 4°C. Membranes were then washed three times for 5 min with TBST, and incubated with the appropriate HRP-conjugated secondary antibody in buffer (1× TBS, 0.1% Tween-20 with 1% BSA). After three washes for 5 min with TBST, proteins were detected using Pierce ECL Western Blotting Substrate (ThermoFisher Scientific) in accordance with the manufacturer's instructions.

Primary antibodies used in this study were: rabbit anti-phospho-Histone 3 (1:100; Cell Signaling Technology, #9701); mouse anti-phospho-Histone 3 (1:100; Cell Signaling Technology, #9706); rabbit anti-pGro (1:100; Hasson et al., 2005; Cinnamon et al., 2008); mouse anti-Gro (diluted 1:1000 for immunofluorescence and 1:5000 for western blot analysis; generously contributed by Christos Delidakis, Institute of Molecular Biology and Biotechnology, Crete, Greece) (Delidakis et al., 1991); rat anti-total Gro (1:1000 for western blot analysis; Santa Cruz Biotechnology, sc-15786); rabbit anti-panTLE (1:100; Cell Signaling Technology, #4681); mouse anti-GFP [1:100; Developmental Studies Hybridoma Bank (DSHB), #8H11]; mouse anti-Cyclin A (1:20; DSHB, #A12); mouse anti-Cyclin B (1:20; DSHB, #F2F4); rabbit anti-Dcp1 (1:100; Cell Signaling Technology, #9578); rat anti-E2F1 (1:100; generously contributed by Stefan Thor, University of Queensland, Australia); and rabbit anti-HA (1:100; Cell Signaling Technology, #3724). Secondary antibodies were Alexa Fluor 488 AffiniPure Donkey Anti-Mouse IgG (1:400; Jackson ImmunoResearch 715-545-150), Rhodamine Red-X (RRX) AffiniPure Donkey Anti-Mouse IgG (1:400; Jackson ImmunoResearch 715-295-150), Cy5 AffiniPure Donkey Anti-Mouse IgG (1:400; Jackson ImmunoResearch 715-175-151) or Cy5 AffiniPure Goat Anti-Rabbit IgG (1:400; Jackson ImmunoResearch 111-175-144).

Nuclei were labelled using 4′,6-diamidino-2-phenylindole (DAPI) (1:1000; Sigma-Aldrich), and embryos and wing imaginal discs were mounted using Vectashield medium (Vector Laboratories).

In all cases, 90-100 imaginal discs for each genetic background were subjected to immunofluorescent antibody staining. Each staining was repeated at least three independent times.

Rabbit anti-phosphorylated-Gro (pGro) polyclonal antibodies were raised using a synthetic phosphopeptide containing one of two Erk consensus sites (phospho-Threonine 308), which is highly conserved in different Drosophila species as well as in other insects (Fig. S2). Subsequently, the antibodies were affinity-purified on a column with a corresponding non-phosphorylated peptide, and the flow-through was later bound on a column with the phosphorylated peptide. These anti-pGro antibodies detect Gro in its phosphorylated state in vivo, particularly in domains of ongoing and earlier RTK pathway activity (given the persistence of Gro phosphorylation) (Hasson et al., 2005; Cinnamon et al., 2008; Helman et al., 2011; Johnston et al., 2016).

When diluted up to 1:100, the monoclonal mouse anti-Gro antibody, raised against amino acids 120-380 (a region containing the epitope used to generate the rabbit poly-clonal anti-pGro antisera) (Delidakis et al., 1991), detects Gro whether phosphorylated or not (Delidakis et al., 1991). When used at a 1:1000 dilution, however, this antibody primarily recognises non-phosphorylated Gro, generating a signal that largely complements the domain of pGro throughout embryonic and adult development (Fig. S1) (Cinnamon et al., 2008; Johnston et al., 2016). Note that the anti-Gro antibody detects signals across a range of Gro protein levels (i.e. downregulation as well as overexpression; Fig. S1 and Fig. 4). The general lack of anti-Gro staining in anti-pGro-positive nuclei (Fig. 1A,A′) implies that the protein is phosphorylated in these nuclei. Importantly, the mutually exclusive recognition by the anti-pGro and anti-Gro antibodies is also observed in vitro even under denaturing conditions (Cinnamon et al., 2008), suggesting that both antibodies are probably directed against the same epitope and that phosphorylation is enough to mask detection by the anti-Gro antibody.

The specific recognition of nonphosphorylated Gro by the anti-Gro antibody is further illustrated by its failure to detect phosphomimetic Gro^DD, while strongly recognising the nonphosphorylatable Gro^AA variant when similarly overexpressed (see insets in Fig. 4C and A, respectively). Finally, commercially available polyclonal anti-total Gro antibodies (Santa Cruz Biotechnology) were used to detect total Gro levels (i.e. Gro, Gro^AA and Gro^DD) in immunoblots (Fig. S3), and commercially available anti-panTLE antibodies (Cell Signaling Technology) were used to follow transgenic expression in vivo (Fig. S3).

Third instar larval wing and eye imaginal discs were submerged in 1× PBS in the presence of 1:1000 EdU for 1 h with gentle rolling at room temperature. EdU was detected using the Click-iT EdU Alexa Fluor 555 Imaging Kit (Life Technologies).

A GST-tagged, full-length Gro fusion protein was expressed in Escherichia coli and purified on a Glutathione affinity column. Approximately 2 µg of purified protein were incubated with (or without) an activated Cdk1/CyclinB1 complex (Sigma-Aldrich, SRP5009) as per the manufacturer's instructions, with the following changes: the final reaction conditions were 10 mM MOPS (pH 7.2), 5 mM glycerol 2-phosphate, 10 mM MgCl₂, 2 mM EGTA, 0.8 mM EDTA, 0.1 mM dithiothreitol (DTT) and 50 μM ATP. The reactions were proceeded for 30 min at 30°C and terminated by adding sample buffer. Reaction mixtures were separated on SDS-PAGE and analysed by western blotting with anti-pGro and anti-Gro antibodies.

Total RNA was extracted using Aurum Total RNA Mini Kit (Bio-Rad). RT-PCR was performed using iTaq Universal SYBR Green Master Supermix (Bio-Rad) and CFX384 Touch Real-Time PCR Detection System (Bio-Rad). Expression was normalised to GAPDH transcripts in all cases. Each experiment was carried out in biological triplicates, with three technical replicates measured each time (one representative experiment is shown). Data analysis was preformed using Bio-Rad CFX manager 3.1 (Bio-Rad).

Samples were prepared as previously described (Davidson and Duronio, 2012), and DNA content was determined using Hoechst 33342 (Sigma-Aldrich). Cells were subjected to flow cytometry using LSR-Fortessa Analyzer (BD Biosciences), and results were analysed by FCS express 4 software (De-Novo Software). Each experiment was carried out at least three independent times, of which a single representative experiment is presented. In all experiments, larvae were identically staged based on hours after egg laying; the GFP population gated separately and afterwards overlaid; and an equal number of cells per each population are presented.

Mitotic indices, which take into account the effect of Gro on mitosis while controlling for the changes its expression exerts on cell size, were calculated by counting pH3-positive cells in a known, constant area and dividing them by the number of nuclei.

For doubling time measurement, clones were induced by heat shock at 72 h and fixed at 124 h. Doubling time was calculated using the following formula: (Log 2/Log N)h, where N is the median cell number per clone and h is the age of the clones in hours (Tseng and Hariharan, 2002).

We used the ImageJ ROI manager to semi-automatically calculate the percentage of overlapping area (in pixels) between two different signals. Each colour threshold was automatically determined, and individual mask selections were constructed. The ‘AND’ tool in the ROI manager was then used to determine the area of overlap between the two masks, and the ‘OR’ tool to determine the area covered by both masks. The percentages presented in the Figures represent the division of the ‘AND’/‘OR’ values.

The doubling time in each genotype was first determined, and the distribution (i.e. percentage) of cells at each phase of the cell cycle was then established using flow cytometry (Neufeld et al., 1998).

ChIP-seq assays were carried out for Gro binding in Kc167 and S2R+ cells as previously described (Kaul et al., 2014). Two biological ChIP-seq replicates were used to obtain a high confidence set of peaks for Gro in BG3 cells. Peaks were called using the MACS2 v2.1.1.2 software providing input and immunoprecipitated samples simultaneously with default parameters. Peaks present in both biological replicates, with FDR<10% in at least one sample and P-value<0.001 in the other, were selected (Gaspar, 2018 preprint).

Sequences were aligned to the dm6.26 reference genome using bowtie2 v2.3.3 according to default parameters (Langmead and Salzberg, 2012). Unmapped reads were removed and adapters were trimmed. Multimapping reads were removed using the ‘view’ program of SAMtools v1.7 (Li et al., 2009) with the parameter ‘-q 20’. The program ‘markdup’ was used to remove PCR duplicates from mapped reads, with the parameter ‘-r’.

BigWig files were generated from BAM files using bamCoverage from deepTools v3.4.3, and reads normalised to counts per million (CPM) were mapped. BigwigCompare was then used to normalise each file against its input using a bin size of 10 and operation subtract as parameters. Visualisation of the genome tracks of ChIP-seq signals was attained using the Integrate genomics viewer (Thorvaldsdóttir et al., 2013).

ChIP-seq datasets for Gro in Kc167 and S2R+ cells have been previously published (Kaul et al., 2014) and are available from ArrayExpress (E-MTAB-2316). The accession number for the Illumina Sequencing data for Gro ChIP-seq in BG3 cells from this study is ArrayExpress (E-MTAB-12108).

Adult wings were mounted as previously described in Kushnir et al. (2020). Confocal images were attained using a Zeiss LSM710 confocal microscopy. Images were processed using Adobe Photoshop software.

Statistical analyses (binomial, two-tailed and Mann-Whitney U tests) were conducted using GraphPad Prism 8 software. ImageJ was used to measure the intensity of yellow-stained cells and for semi-automated quantifications, and Zen software was used to quantify the numbers of DAPI, pH3-, EdU-, Gro- or pGro-positive cells.

We thank Tatyana Kushnir and Rotem Lange for their continuous help and encouragement. We are grateful to Einat Cinnamon, Russell Finley Jr., Offer Gerlitz, David Ish-Horowicz, Gerardo Jiménez, Michael Klutstein, Benny Shilo and Talila Volk for their thoughtful comments on the manuscript; to Aamna Kaul for generating the ChIP-seq data for BG3 cells; to David Engelberg for his invaluable assistance with the in vitro kinase assays; and to Christos Delidakis, Robert Duronio, Bruce Edgar, Doron Ginsberg, Gerardo Jiménez, Christian Lehner, Marco Milán, Stefan Thor, Hongyan Wang, the Developmental Studies Hybridoma Bank and the Bloomington Drosophila Stock Centre for DNA constructs, antibodies, reagents and fly stocks. Finally, we thank Yael Feinstein-Rotkopf from the Faculty of Medicine's Core Research Facility for her technical assistance with microscopy.