The caenorhabditis elegans CDT-2 ubiquitin ligase is required for attenuation of EGFR signalling in vulva precursor cells

Background Attenuation of the EGFR (Epidermal Growth Factor Receptor) signalling cascade is crucial to control cell fate during development. A candidate-based RNAi approach in C. elegans identified CDT-2 as an attenuator of LET-23 (EGFR) signalling. Human CDT2 is a component of the conserved CDT2/CUL4/DDB1 ubiquitin ligase complex that plays a critical role in DNA replication and G2/M checkpoint. Within this complex, CDT2 is responsible for substrate recognition. This ubiquitin ligase complex has been shown in various organisms, including C. elegans, to target the replication-licensing factor CDT1, and the CDK inhibitor p21. However, no previous link to EGFR signalling has been identified. Results We have characterised CDT-2's role during vulva development and found that it is a novel attenuator of LET-23 signalling. CDT-2 acts redundantly with negative modulators of LET-23 signalling and CDT-2 or CUL-4 downregulation causes persistent expression of the egl-17::cfp transgene, a marker of LET-23 signalling during vulva development. In addition, we show that CDT-2 physically interacts with SEM-5 (GRB2), a known negative modulator of LET-23 signalling that directly binds LET-23, and provide genetic evidence consistent with CDT-2 functioning at or downstream of LET-23. Interestingly, both SEM-5 and CDT-2 were identified independently in a screen for genes involved in receptor-mediated endocytosis in oocytes, suggesting that attenuation of LET-23 by CDT-2 might be through regulation of endocytosis. Conclusions In this study, we have shown that CDT-2 and CUL-4, members of the CUL-4/DDB-1/CDT-2 E3 ubiquitin ligase complex attenuate LET-23 signalling in vulval precursor cells. In future, it will be interesting to investigate the potential link to endocytosis and to determine whether other signalling pathways dependent on endocytosis, e.g. LIN-12 (Notch) could be regulated by this ubiquitin ligase complex. This work has uncovered a novel function for the CUL-4/DDB-1/CDT-2 E3 ligase that may be relevant for its mammalian oncogenic activity.


Background
C. elegans vulva development has been instrumental in the characterisation of numerous major signalling pathways such as EGFR (Epidermal Growth Factor Receptor), and Notch [1,2]. Even though most of the components of these core signalling pathways have been identified, the modulatory mechanisms remain difficult to decipher because of the intricate network formed by negative and positive feedback loops. In an attempt to identify novel players in attenuation of LET-23 (EGFR) signalling, we used a candidate-based approach to screen, by RNAi, for genetic interactors of gap-1 (see materials and methods). Deletion of gap-1 creates a hypersensitive background for the LET-23(EGFR)/LET-60(RAS)/LIN-45(RAF) signalling cascade, in which the loss of an additional attenuator can cause the Multivulvae (Muv) phenotype [2][3][4]. This strategy identified CDT-2, an evolutionary conserved homologue of human CDT2 also called DTL or DCAF2.
Human CDT2 was first discovered as a transcript down regulated following retinoic acid-induced neuronal differentiation in pluripotent NT2 cells [5], suggesting a role in maintenance of self-renewal capacity. CDT2, a WD40 domain containing protein, was later found associated to the CUL4/DDB1 E3 ubiquitin ligase complex [6][7][8]. Within this complex, CDT2 acts as a substrate recognition subunit. Two substrates have been well characterised: the license to replicate, CDT1, and the CDK inhibitor, p21 [7,[9][10][11][12][13]. Degradation of CDT1 and p21 are essential to prevent rereplication or firing of origin of replication following DNA damage-induced stress. Therefore, CDT2, as part of the CUL4/DDB1 E3 ubiquitin ligase complex, plays a critical role in regulation of DNA replication.
Mouse CDT2 activity is essential for viability, which has precluded the study of its role during development [14]. RNAi in C. elegans can sometime create knock down conditions that allow the identification of novel late onset activities. Here, we show that C. elegans CDT-2 and CUL-4 attenuate LET-23 signalling during vulva development. We found that SEM-5 (GRB2) physically interacts with CDT-2 and genetic studies are consistent with CDT-2 acting at the level of the LET-23 receptor. Finally, we confirmed that CDT-2 and SEM-5 are required for receptor-mediated endocytosis in oocytes [15]. We propose a model by which the CUL-4/ DDB-1/CDT-2 ubiquitin ligase complex associates with SEM-5 to target LET-23 and regulate its endocytosis.

RNAi procedure
Briefly, worms for RNAi exposure were synchronised using standard bleaching to isolate embryos. These were grown to the L3 stage and then transferred to RNAi plates. The mothers were transferred onto a new plate after three days and the F1 s laid on this plate were analysed, as previously described [17]. For lin-3 rf , let-23 rf , and lin-45 rf (and doubles with gap-1), F2 s were analysed instead of F1 s since the Vul phenotype leads to small broods. RNAi clones used in this study were all confirmed by sequencing.

Scoring of the Multivulvae (Muv) phenotype
Induction of vulval cells was scored by lineage analysis of vulval precursor cells (VPCs). Briefly, L4 animals were mounted on agarose pads and the descendants of the six Vulval Precursors Cells (VPCs) analysed to assign either the vulval fate or the non-vulval fate (fusion to hyp 7) (Figure 1, assay done as in POULINet al. 2005). Each fully induced VPC is given a score of 1; a wild-type vulva therefore has a score of three because three VPCs adopt the vulval fate. A score of 0.5 is given if only one VPC daughter produces vulval tissue and one fuses with hyp7. A score of 0 is given if the VPC or both of its daughters fuse with hyp7.
Statistical analysis was performed mostly using the Mann-Whitney U-test with the calculated number of VPCs induced, except for the lin-3 rf analysis on which the Fisher's exact test was performed to analyse the proportionality of control worms with wild type vulva compared to cdt-2(RNAi) worms. In this particular case, it is likely that the Mann-Whitney test introduced a type II error (acceptance of the null hypothesis).

egl-17::cfp assay
Briefly, L3 animals were mounted on agarose pads and examined for persistent expression of egl-17::cfp in secondary cells (light grey cells in Figure 1). These cells do not normally express egl-17::cfp at this stage (after second or third division) [18]. Importantly, this assay must be performed prior to L4, since egl-17 expression disappears from the primary cells and appears in secondary cells at mid-L4 [19]. For the analysis of the cul-4 mutants, heterozygous and homozygous animals were analysed in parallel, and were from the same mothers. Therefore the analysed animals were at roughly the same age in absolute time.

vit-2::gfp assay
Briefly, the vit-2::gfp assay was performed as described [20], and RNAi perform as indicated above. Young adults were analysed and animals with gross gonadal defects were not analysed as they could bias the assay.

In vitro pull down assay
Briefly, CDT-2 was produced using an in vitro transcription/translation reaction according to the manufacturer (Promega, Reticulocyte Lysate System, Non-Radioactive labelling). SEM-5 and SLI-1 were fused to GST and purified on column according to manufacturer. The pull down was performed as previously described [21]. Equivalent amount of GST fusion proteins (~200 ng) were used per pull down, the size of the proteins visualised on gels stained by Coomassie, and protein concentrations measured by Bradford assay.

cdt-2 genetically interacts with gap-1
We previously identified cdt-2 as having a potential role in vulva development because knockdown caused a weak synMuv (synthetic Multivulvae) phenotype [17]. RNAi caused a low penetrance synMuv phenotype in a lin-15A background, but it did not pass the penetrance  [47,48]. The anterior of the animal is on the left (P3.p) and posterior on the right (P8.p). In wild type animals only three VPCs (P5.p -P7.p) will adopt the vulval fate, the other VPCs will fuse to hyp7 after first division (white circles). Sometime, P3.p fuses before the first cell division. The VPCs that are committed towards the vulval fate will undergo three rounds of cell division. The descendants after first division are named P5/6/7.px, and after second division P5/6/7.pxx, and after third divison P5/6/7.pxxx (x signify either anterior cell (a) or posterior cell (p)). Of note, two cells will not undergo the third division (see threshold and therefore was not further analysed at the time. It was subsequently shown that lin-15A could act redundantly with gap-1 to prevent erroneous adoption of vulval fate [24]. GAP-1 is a GTPase Activating Protein that acts as an attenuator of LET-60 RAS signalling, and the gap-1 mutant has been shown to be a sensitized background for identifying attenuators [18]. The interaction with lin-15A suggested that some of the weak candidate synMuv genes we previously identified might genetically interact with gap-1. We therefore tested whether cdt-2 could interact with gap-1 to cause a Muv phenotype. We found that cdt-2 (RNAi) in the gap-1 background causes 43% of animals to present a Muv phenotype (Table 1). We also confirmed that cdt-2 only marginally interacts with lin-15A or lin-15B (Table 1). In addition, RNAi of cdt-2 slightly increases penetrance of the Muv phenotype observed in a lin-15AB mutant (Table 1), which is consistent with an atypical synMuv activity.

CUL-4 prevents excessive LET-23 EGFR signalling during vulva development
Mammalian CDT2 has been found associated with the CUL4/DDB1 ubiquitin ligase complex [6][7][8], which prompted us to test whether the C. elegans homologues of the complex would possess an activity similar to CDT-2. RNAi of cul-4, ddb-1, or rbx-1 (also found in this complex [27,28]) did not produce a Muv phenotype in the gap-1 background (data not shown), but the rereplication phenotype [13,29] could be detected in these experiments (data not shown). Because RNAi knockdown animals might retain residual activity, we also investigated the phenotype of a cul-4 deletion mutant.

CDT-2 is expressed in dividing vulval precursor cells
Since CDT-2 plays an important role during vulva development, we analysed its expression using a translational GFP fusion. The fusion protein is predominantly nuclear, as has been seen for other CDT2 homologs. CDT-2::GFP is not detected in P cells at larval stage L1 (data not shown), but is expressed early in all Vulval Precursor Cells (VPCs, Pn.p cells) prior to their first division ( Figure 3A and 3B). The frequency of   expression is lowest in P3.p cells, and highest in P6.p ( Figure 3K). After first division, the cells that adopted the vulval fate all express CDT-2::GFP, but the nonvulval cells generally do not ( Figure 3K, two-cell stage). However, sometimes low expression can be observed in the descendants of P3.p, P4.p and P8.p ( Figure 3K, twocell stage). Interestingly, after second division (four-cell stage) CDT-2::GFP expression disappears from two secondary cells (P5.ppp and P7.paa, see Figure 1); these are the only vulval cells that will not undergo a third cell division ( Figure 3C, D and 3K). Later, at L4 stage no expression is detected ( Figure 3E and 3F). We also observed CDT-2:GFP expression in the cytoplasm during the first mitotic division of P6.p, which quickly relocalised to the nuclei as the nuclear envelope reforms ( Figure 3G-I). The early CDT-2 pattern of expression is consistent with a role during vulval fate adoption, and its down regulation in cells that cease cell division is consistent with a role in DNA replication.
CDT-2 is active at the level of the LET-23 receptor and physically interacts with SEM-5 To try to understand how CDT-2 attenuates the LET-23 signalling cascade during vulva development, we analysed the type of epistatic interactions produced between cdt-2(RNAi) and reduced function (rf) alleles of lin-3/ Egf, let-23/Egfr, and lin-45/Raf. We first tested whether depletion of cdt-2 could rescue the Vul (vulvaless) phenotype produced by lin-3 rf , let-23 rf , or lin-45 rf . Depletion of cdt-2 by RNAi did not affect the penetrance of the Vul phenotype produced by lin-45 rf , but did partially suppress the Vul phenotype of let-23 rf ( Table 3). RNAi of cdt-2 in lin-3 rf also affected the penetrance of the Vul phenotype (2% of control animals with a wild type vulva versus 19% of cdt-2(RNAi) animals, (Fisher's test, p = 0.002)) ( Table 3), indicating that the Vul phenotype caused by a reduction of ligand can be rescued. Of note, the lin-3 n378 allele used here is a reduced function allele that was shown to still retain ligand activity [30]. We obtained similar results performing epistasis experiments in a sensitized gap-1 mutant background ( Table 3). Depletion of cdt-2 did not rescue the Vul phenotype of the lin-45 rf ;gap-1 double but did increase the penetrance of the Muv phenotype of let-23 rf ;gap-1 double mutants, as well as the number of VPCs induced (Table 3). A similar trend was seen with lin-3 rf ;gap-1, though not statistically significant ( Table 3). Depletion of cdt-2 also enhanced the penetrance of the Muv phenotype and the number of VPCs induced in a let-60 gain of function allele (Table 3). Taken together, these results are consistent with cdt-2 acting upstream of lin-45, but downstream or at the level of let-23 to attenuate this signalling cascade. We further analysed the capacity of cdt-2 to genetically interact with other negative modulators of the LET-23 signalling pathway that are known to act at the level of the receptor. Previous work showed that UNC-101 and DPY-23 are adaptins orthologous to the mu1 and mu2 subunits of adaptor protein complex 1 and 2 (AP1 and AP2, respectively), and that they both can act as negative modulators of LET-23 signalling [18,31]. Similarly, SLI-1 is orthologous to CBL, an E3 ubiquitin ligase targeting LET-23 for degradation [4,32] and SEM-5 is GRB2, an adaptor molecule that physically interact with EGFR [2,33,34]. To address whether these genes could interact with cdt-2, we used loss of function (lf) alleles of dpy-23/AP2, unc-101/AP1, sli-1/CBL, and sem-5/GRB2 and performed cdt-2(RNAi). We found that cdt-2 genetically interacts with dpy-23 lf and unc-101 lf , as cdt-2 RNAi induces a Muv phenotype in these backgrounds. In contrast, no interaction was seen with sli-1 lf or sem-5 lf (Table 4).
Since an absence of genetic interaction can sometimes suggest a physical interaction [35,36], we tested whether    A and B), but sometimes (30%) P3.p does not (K), perhaps reflecting fusion to hyp7 prior to first cell division (see Figure 1). At four-cell stage, two cells that will not proceed to the third division stop expressing CDT-2 (C, D and K). At Christmas-tree stage CDT-2 is not detected (E and F). Sequential pictures of P6.p undergoing first division showing that during mitosis CDT-2 can be detected in cytoplasm (G and H), and later that CDT-2 is relocalised at the nuclear membrane (I and J). Frequency of expression observed in vulval cells at different stages (K). Black or dark grey circles represent high frequency of expression (100% and 92%). Light grey circles represent a low to medium frequency of expression (18, 30, 36 or 77%). White circles represent undetected or very low frequency of expression (0 or 1%).
CDT-2 could physically interact with either SLI-1 or SEM-5. We produced in vitro labelled CDT-2 and purified SLI-1 and SEM-5 from bacteria. We found that CDT-2 could physically associate with SEM-5, but not with SLI-1 ( Figure 4). Together, the genetic and physical interaction data suggest that CDT-2 may prevent excessive signalling regulating LET-23 through SEM-5.

Depletion of CDT-2 or SEM-5 causes similar receptormediated endocytosis defect
The association between CDT-2 and SEM-5 suggests that they function together in a common process. Interestingly, both sem-5 and cdt-2 have been identified in an RNAi screen designed to identify genes required for receptor-mediated endocytosis in oocytes [15]. The assay used in this screen is based on the accumulation of VIT-2::GFP (yolk) in body cavities. VIT-2 is secreted into the body cavities by the intestine and is endocytosed by oocytes via the yolk receptor, RME-2 ([20] and Figure 5F). By fusing VIT-2 to GFP, it is possible to assess whether receptor-mediated endocytosis is functional, because if not VIT-2::GFP accumulates in body cavities of young hermaphrodites [20]. We confirmed that reduction of cdt-2 or sem-5 causes body cavity accumulation of the vit-2::gfp reporter ( Figure 5A-C).
Because correct cortical localization of the RME-2 yolk receptor is required for endocytosis [20], we next examined receptor localization in cdt-2 RNAi animals to test whether the accumulation of vit-2::gfp might be caused indirectly by improper localization of the receptor. We found that the expression and localization of an rme-2::gfp reporter is normal in cdt-2(RNAi) animals ( Figure 5D and 5E). The correct localization of RME-2:: GFP combined with the defect in uptake of VIT-2::GFP suggests that CDT-2 plays a role (direct or indirect) in the process of receptor-mediated endocytosis.

Discussion
CDT2 is a recognition subunit of the CUL4/DDB1 E3 ubiquitin ligase complex important for DNA replication and G2/M checkpoint [9,10,27,28]. Previous work has shown that these functions are conserved in C. elegans [11][12][13]29]. We have uncovered a novel role for CUL-4 and CDT-2 in preventing excess LET-23 signalling. Because CDT-2 and CUL-4 normally work as part of the CDT-2/CUL-4/DDB-1 ubiquitin ligase complex, it seems likely that DDB-1 also participates in attenuation of LET-23 signalling. However, RNAi of ddb-1 did not cause obvious LET-23 signalling defects. This might be due to incomplete knockdown, or alternatively, CDT-2 and CUL-4 could act independently of DDB-1 in this context. We also provided in vitro evidence that CDT-2 can associate with SEM-5 directly. CDT-2 and SEM-5 share two functions: they attenuate LET-23 signalling during vulva development and are required for receptormediated endocytosis during oogenesis. Linking these two functions together, we hypothesise that the CUL-4/

The rereplication defect and LET-23 signalling
The rereplication defect caused by depletion of CDT-2 or CUL-4 has been previously characterised as well as the cell cycle arrest phenotype [11,13,29]. However, it is difficult to explain that these defects could cause excessive LET-23 signalling during vulva development.
Indeed, experiments using hydroxyurea to arrest the VPC cell cycle have shown that egl-17 expression remains restricted to P6.p [37]. Therefore, a replication block after first division as in the case of cul-4 deletion mutants is unlikely to cause persistent expression of egl-17::cfp. Furthermore, we observed increased LET-23 signalling in cdt-2 RNAi animals, and an increase in vulval fate adoption in gap-1; cdt-2(RNAi) animals, under conditions where the cell cycle proceeds normally. Therefore, the role of CDT-2 in preventing rereplication is likely to be independent of its function in preventing excess LET-23 signalling.
CDT-2 may attenuate LET-23 signalling as a component of the CUL-4/DDB-1 E3 ligase complex RNAi by feeding in C. elegans has significant false negative rate, but false positives are rare [38]. Hence, the finding that a deletion of cul-4 can cause the same vulval phenotype (egl-17::cfp assay,  Figure 5 Depletion of cdt-2 causes a similar endocytotic defect than observed in a loss of function allele of sem-5. We analysed the uptake of VIT-2, yolk, using a vit-2::gfp expressing strain (A) and observed accumulation in body cavities in cdt-2(RNAi) (B) and in sem-5 loss of function (C). Arrows point to sites that accumulated VIT-2::GFP. We verified that the receptor, RME-2, is properly expressed using an rme-2::gfp expressing strain. No difference in oocytes expressing RME-2 could be observed when the control (empty vector (RNAi)) was compared to cdt-2 (RNAi) animals (D and E). Arrows indicate oocytes.
SLI-1, the CBL homologue [4,32]. SLI-1 is an E3 ubiquitin ligase that can associate with SEM-5 to target RTKs and promote lysosomal degradation [39]. Here we show that CDT-2 can physically interact with SEM-5 and genetic analyses are consistent with action at the level of the LET-23 receptor. This suggests that SEM-5 might regulate attenuation of LET-23 signalling through ubiquitination and subsequent endocytosis using two different ubiquitin ligases: SLI-1 (CBL) and CUL-4/DDB-1/ CDT-2. Unfortunately, we have been as yet unable to directly assess LET-23 receptor localisation or endocytosis during vulva development; immunostaining experiments are inconsistent and current let-23::gfp transgenics are not fully functional. Tests of these models will require better reagents to investigate regulation of the LET-23 receptor.

Ubiquitination and regulation of Notch signalling
Receptor-mediated endocytosis is important to terminate or attenuate signalling [40], not only for EGFR but also for other signalling pathways, e.g. Notch. During vulva development, LIN-12 (Notch) signalling is required for establishment of the secondary cell fate [41] and for the production of the anchor cell, which produces LIN-3 (EGF) [42]. Interestingly, SEL-10, an F-box and a WD40 containing protein that belongs to the CDC4/CUL-1 family of ubiquitin ligase [43], has been shown to play an important role in attenuation of LIN-12 signalling [44]. SEL-10 was also shown to physically interact with LIN-12, implying that it regulates signalling via ubiquitination of LIN-12. Herein we have not investigated the relationship between the CUL-4/DDB-1/CDT-2 ubiquitin ligase complex and LIN-12 signalling. We did not observe any defects in anchor cell development (unpublished data), a process dependent on LIN-12, however, it has been previously shown for SEL-10 that a sensitised background is required to reveal its activity as an attenuator of LIN-12 signalling [44]. Therefore, we may not have detected a potential role for CDT-2 in attenuation of LIN-12 signalling. There is also an intimate link between LIN-12 and LET-23 signalling during vulva development [44]. Indeed, high level of LET-23 signalling triggers expression of LIN-12 ligands in the primary P6.p cell [45]. This activates LIN-12 signalling in the flanking secondary cells and ensures down regulation of LET-23 signalling in P5.p and P7.p cells [18,25,46]. It is not impossible that the depletion of CDT-2 or CUL-4 impairs LIN-12 signalling and thereby prevents appropriate down regulation of LET-23 signalling in secondary cells, which would cause persistent expression of egl-17::cfp in secondary cells.

Localisation of CDT-2
The localisation of CDT-2 fused to GFP is predominantly nuclear in interphase and cytoplasmic during mitosis, which seems contrary with a function in endocytosis. However, we cannot exclude that a proportion of CDT-2::GFP below our limit of detection is cytoplasmic during interphase. Interestingly, early studies showed that human CDT-2 can be detected in the cytoplasm [5], which would be consistent with a role in ubiquitination of cytoplasmic targets. Alternatively, the CUL-4/DDB-1/CDT-2 E3 ubiquitin ligase complex may be active in the cytoplasm only after nuclear breakdown. Further experiments will be required to establish when and whether the complex is active in the cytoplasm, in particular during mitosis.

Conclusion
We have identified CDT-2 and CUL-4 as novel attenuators of LET-23 signalling during vulva development. Both of these proteins are known components of a conserved CUL-4/DDB-1/CDT-2 E3 ubiquitin ligase complex, suggesting a novel function for this complex in signalling during C. elegans development. A similar function in other organisms might have been missed because of its requirement for cell cycle progression. Studying this complex in mammalian cells using knock down conditions that bypass the cell cycle defect might reveal a conserved role in signalling.