Rb and FZR1/Cdh1 determine CDK4/6-cyclin D requirement in C. elegans and human cancer cells

Cyclin-dependent kinases 4 and 6 (CDK4/6) in complex with D-type cyclins promote cell cycle entry. Most human cancers contain overactive CDK4/6-cyclin D, and CDK4/6-specific inhibitors are promising anti-cancer therapeutics. Here, we investigate the critical functions of CDK4/6-cyclin D kinases, starting from an unbiased screen in the nematode Caenorhabditis elegans. We found that simultaneous mutation of lin-35, a retinoblastoma (Rb)-related gene, and fzr-1, an orthologue to the APC/C co-activator Cdh1, completely eliminates the essential requirement of CDK4/6-cyclin D (CDK-4/CYD-1) in C. elegans. CDK-4/CYD-1 phosphorylates specific residues in the LIN-35 Rb spacer domain and FZR-1 amino terminus, resembling inactivating phosphorylations of the human proteins. In human breast cancer cells, simultaneous knockdown of Rb and FZR1 synergistically bypasses cell division arrest induced by the CDK4/6-specific inhibitor PD-0332991. Our data identify FZR1 as a candidate CDK4/6-cyclin D substrate and point to an APC/CFZR1 activity as an important determinant in response to CDK4/6-inhibitors.

C yclin-dependent kinases (CDKs) in association with cyclin subunits are the key regulators of cell division in eukaryotes 1 . Active CDK-cyclin complexes promote progression through the cell division cycle by phosphorylating critical substrates. The decision to initiate cell division occurs in the G1 phase and involves expression of D-type cyclins in response to mitogenic signalling 2 . These cyclins assemble with CDK4 or CDK6 (CDK4/6), to form kinase complexes that are best known for their phosphorylation of the retinoblastoma (Rb) tumour suppressor protein. Rb family proteins associate with E2F transcription factors and act as transcriptional repressors of cyclin E and other S phase genes. The initial phosphorylation of Rb by CDK4/6-cyclin D is thought to weaken its repressor function, while subsequent phosphorylation by CDK2-cyclin E completely inactivates Rb function and allows full commitment to S phase entry.
Strong support for the in vivo importance of this cell cycle control pathway has come from human cancer studies. Overexpression of CDK4/6 or D-type cyclins and inactivation of the CDK4/6 antagonist p16 INK4A /CDKN2A or Rb tumour suppressor are common in human cancer 3 . These events are largely mutually exclusive, in support of p16 INK4A , CDK4/6, D-type cyclins and Rb acting in a single regulatory pathway. At the same time, various studies have indicated that phosphorylation of Rb is not the only catalytic activity of CDK4/6-cyclin D kinases 4,5 . Additional substrates of cyclin D kinases have been described, the best characterized of which are the Rb-related proteins p107 and p130, and transcription factors SMAD3 and FOXM1 (refs 2,4,6).
To what extent phosphorylation of these targets contributes to carcinogenesis is currently unknown.
Results from studies in mice have caused doubt on whether the functions of CDK4/6-cyclin D kinases are essential for proliferation. Knockout of a single D-type cyclin gene causes limited defects and mice that lack all three D-type cyclins still develop until mid-to-late gestation 7 . Similarly, CDK4/CDK6 double knockout mice complete organogenesis and extensive cell proliferation, with death due to anaemia occurring only in the late stages of embryogenesis 8 . In contrast to normal development, cancer formation in various mouse models depends strongly on CDK4/6-cyclin D kinase activity [9][10][11][12] . This difference in requirement appears to provide a window of opportunity for therapeutics that block cancer growth, while sparing normal cells. Small molecule inhibitors with high specificity for CDK4/6 have been identified, with PD-0332991 as the leading example 13,14 . PD-0332991 induces proliferation arrest in a substantial subset of human cancer cell lines and inhibits cancer formation in mouse models 10,11,13,15 . Based on these results and recent Phase II and Phase III clinical trials, CDK4/6 inhibitors currently receive much attention as promising anti-cancer therapeutics [16][17][18] . Although there are substantially increased progression-free survival rates of cancer patient populations in several studies, biomarkers that predict a positive response to CDK4/6 inhibitor treatment are currently not known. It will be of great clinical importance to reveal which cancer genotypes correspond to cell cycle arrest, or even senescence and apoptosis, in response to inhibitor treatment, and which bypass routes may be used by cancer cells to acquire resistance to CDK4/6-specific inhibitors.
In this study, we examine the critical functions of the CDK4/6 cyclin D kinase, making use of the evolutionary conserved regulation of cell cycle entry in metazoans. Our observations in the nematode C. elegans support that Rb-mediated transcriptional repression and APC FZR1 -mediated protein degradation act in parallel to inhibit G1/S progression, and that phosphorylation by the CDK-4/CYD-1 cyclin D kinase counteracts these inhibitory functions. Importantly, we also observed synergy between Rb and FZR1 knockdown in bypassing the proliferation arrest induced by treatment of human breast cancer cells with the CDK4/6 inhibitor PD-0332991. Our results indicate that the level of APC/C FZR1 activity is an important contributing factor in response of cancer cells to CDK4/6 inhibitor treatment.

Results
C. elegans CDK-4/CYD-1 has multiple critical substrates. We followed a genetic approach to reveal critical functions of CDK4/6 kinases. Cell cycle entry in C. elegans involves a CDK4/6-Rb pathway with limited redundancies (Fig. 1a) 19 . Single genes encode for a C. elegans CDK4/6 kinase CDK-4, a D-type cyclin CYD-1 and a member of the Rb protein family, LIN-35. Candidate null mutations in cyd-1 or cdk-4 result in a general arrest of cell division in the G1 phase during larval development, slow growth and complete sterility (Fig. 1b) 20 . Inactivation of lin-35 Rb by RNA interference (RNAi) or putative null mutation (n745 and n2239 alleles) suppresses the cdk-4 CDK4/6 and cyd-1 cyclin D mutant phenotype in part. Although lin-35 Rb loss allows post-embryonic cell division in cyd-1 and cdk-4 mutants, double mutant animals that lack lin-35 and cyd-1, or lin-35 and cdk-4, remain small and sterile (Fig. 1b) 20 . In addition, quantification of cell numbers in the intestine and ventral nerve cord showed that the division of post-embryonic precursor cells remains limited in such double mutants (Fig. 1c,d) 20,21 . These data indicate that the CDK-4/CYD-1 kinase promotes cell cycle entry not only by inhibiting LIN-35 Rb but also through additional critical activities.

LIN-35 Rb and FZR-1 loss eliminates CDK-4/CYD-1 requirement.
To identify critical substrates, we performed a genetic screen for mutations that suppress the cell division arrest and sterility of lin-35, cyd-1 double mutant animals. In a large-size screen (410,000 haploid genomes), we identified only a single mutant with complete rescue of the lin-35(n745);cyd-1(he112) phenotype. The mutant animals with this he121 suppressor mutation appeared normal and produced progeny numbers similar to lin-35 single mutants (Fig. 1b). Quantification of cell numbers further supported that postembryonic cell divisions in cyd-1(he112) mutants were fully restored by simultaneous lin-35 inactivation and he121 mutation (Fig. 1c,d). In fact, the triplemutant larvae showed slightly higher than normal numbers of intestinal nuclei, again resembling the lin-35 single-mutant phenotype (Fig. 1c, note: the normal intestine never has more than 34 nuclei). Combining lin-35 RNAi with the he121 mutation also restored cell division and fertility in cdk-4(gv3)-null mutants (Fig. 1b). Thus, rescue of the cell cycle arrest phenotype does not result from allele-specific suppression of cyd-1(he112). Instead, the combined lin-35 and he121 mutations appear to abolish cdk-4/cyd-1 requirement. Interestingly, the homozygous he121 mutation also caused some suppression of the cell division arrest of cyd-1 mutants with normal lin-35 Rb function (Fig. 1c,d; note: cyd-1(he112) larvae always have 16 intestinal nuclei). These data indicate that the he121 mutation disrupts a negative regulator of cell cycle entry, which acts in parallel to lin-35 Rb.
We mapped the he121 mutation to chromosome 2, to the right of dpy-10 and close to map position 1. Whole-genome sequencing of the lin-35(n745);he121 cyd-1(he112) strain revealed a nonsense mutation in the fourth codon of the fzr-1 FZR1 predicted open reading frame (CAT to TAG; Glu to Stop). C. elegans fzr-1 encodes a co-activator and substrate specificity factor of the anaphase-promoting complex (APC/C) that is closely related to human FZR1 (also known as Cdh1), Drosophila fizzy related (fzr) and budding yeast Cdh1/Hct1 (refs 22-24). The APC/C FZR1 E3 ubiquitin ligase targets specific proteins for degradation in late mitosis and early G1. APC/C FZR1 is known to inhibit G1/S progression, which appears to agree with he121 being an fzr-1 allele. Additional observations support this conclusion: fosmid and plasmid clones carrying genomic fzr-1 sequences complemented the he121 phenotype when introduced as transgenes (Supplementary Table 1). Further, RNAi of fzr-1, obtained by soaking first stage (L1) larvae, rescued the cell cycle defects in lin-35, cyd-1 double mutants and allowed production of viable offspring (Supplementary Table 2). However, fzr-1 RNAi and the fzr-1(ku298) strong loss-of-function mutation predominantly induced lethality and sterility when combined with lin-35 loss of function, as previously reported 24 . In combination with cyd-1 loss, or double lin-35, cyd-1 mutations, strong fzr-1 RNAi also resulted in a higher number of nuclear divisions in the intestine, compared with the fzr-1(he121) mutation (Fig. 1c, note: the P-derived neuroblasts examined in Fig. 1d are somewhat RNAi insensitive). This combination of results indicates that he121 is a partial loss-of-function allele of fzr-1. The effect of the he121 early termination codon may be weakened by ribosomal read-through, translation initiation at a downstream ATG, or alternative usage of promoter sequences and splice sites. The requirement for an fzr-1 allele to simultaneously suppress cyd-1 loss and avoid synthetic lethality with lin-35 may explain the identification of a rare partial loss-offunction fzr-1 mutation in a large-size screen.
Arrest of cell cycle entry by FZR-1 and LIN-35 Rb. To examine cooperation between fzr-1 and lin-35 in cell cycle arrest, we tested how inactivation of these genes affects expression of mcm-4 S phase reporter constructs. MCM-4 is a subunit of the replicative helicase complex, which we previously observed to accumulate in G1/S, and to be induced by cyclin D and cyclin E CDK activity 25,26 . We used a Pmcm-4::MCM-4::mCherry reporter to Total broods were counted of at least two animals of each genotype and total cell numbers in the indicated tissues of at least ten animals of each genotype. Graphs show mean±s.e.m., each dot (c,d) represents a single animal. detect mcm-4 transcription and protein expression. Remarkably, loss of fzr-1 function by RNAi or use of the ku298 allele resulted in MCM-4::mCherry protein expression in post-mitotic neuronal cells (Fig. 2a,b). This indicates that the strict G0/G1 phase arrest of differentiated neurons involves fzr-1 function (Fig. 2b). Neurons did not initiate DNA synthesis in fzr-1 mutants, based on the absence of EdU (5-ethynyl-2 0 -deoxyuridine) incorporation, and MCM-4::mCherry was not detected in other differentiated cells, such as muscle cells ( Supplementary Fig. 2). lin-35 Rb mutants did not display deregulated expression of MCM-4::mCherry in any cell type (Table 1). To specifically examine loss of transcription regulation, we designed a splityellow fluorescent protein (YFP) reporter combination for detection of S phase gene activation in differentiated muscle cells. We combined expression of an N-terminal YFP fragment (Pmcm-4::nzYFP) under the control of a cell cycle promoter, with expression of a carboxy-terminal YFP fragment from a musclespecific promoter (Pmyo-3::czYFP). In contrast to WT or fzr-1 mutant animals (Fig. 2e), muscle-specific YFP fluorescence was observed in lin-35 mutants, indicating that the mcm-4 promoter is activated or de-repressed in the mutant muscle cells (Fig. 2f). Combining lin-35 Rb and fzr-1 mutations in Pmcm-4::MCM-4::mCherry transgenic animals caused broad expression of the MCM-4::mCherry fusion protein, including body wall muscle expression (MCM-4::mCherry expression in bwm of 152/186 animals; Fig. 2d and Table 1). Thus, LIN-35 Rb-mediated transcriptional repression and APC/C FZR1 -mediated protein degradation act in parallel to prevent S phase gene expression, not only in cells that are temporarily quiescent but also in differentiated postmitotic cells (Fig. 2g).

Posttranslational regulation of FZR-1 and LIN-35 function.
We wondered whether the FZR-1 protein levels or localization are cell cycle regulated and reflect cell cycle stage-specific functions. We created transgenic animals that express FZR-1 with an N-terminal green fluorescent protein (GFP) tag, to be able to follow the FZR-1 protein throughout development in living animals. The GFP::FZR-1 fusion protein is functional, as it rescues the lethal phenotype of lin-35(n745);fzr-1(ku298) double mutants. Transgenic animals showed expression of GFP::FZR-1 in many somatic cells during larval development, at levels that varied in a cell type-and developmental stage-dependent manner (Fig. 3, left panels). In developmentally arrested L1 larvae, GFP::FZR-1 expression was detectable in the cytoplasm and nucleus of postembryonic precursor cells, with the highest levels in nuclei of the syncytial epidermis (hyp7) and intestine, and lowest levels in epidermal seam cells ( Fig. 3a    ARTICLE temporally regulated by CDK activity [27][28][29] . Activation of APC/ C FZR1 follows mitotic cyclin destruction and CDK1 inactivation in anaphase. In the next cell cycle, phosphorylation of FZR1 at the G1/S transition and association with the Emi1 pseudo-substrate inhibitor are thought to switch off APC/C FZR1 activity 29,30 . Although in particular CDK2-cyclin A has been implicated in this phosphorylation 31 , it appears likely to be that CDK4/6-cyclin D contributes to overcoming APC/C FZR1 -mediated G1 inhibition when cell cycle entry depends on this kinase. To test this possibility, we performed kinase assays in vitro, thereby avoiding indirect phosphorylation events by other CDKs and making use of the well-established substrate specificity of CDK4/6-cyclin D kinases in vitro. In these assays, we used C. elegans CDK-4/ CYD-1 expressed and immunopurified from HEK 293T cells in combination with the bacterially expressed substrates GST-LIN-35 and GST-FZR-1. Autoradiography and mass spectrometric analysis revealed efficient and specific phosphorylation of both substrates (Fig. 4a). LIN-35 Rb was phosphorylated by CDK-4/ CYD-1 at Ser714 and Thr719, located between the pocket A and B regions (Fig. 4c). These phosphorylations resemble Ser608 and Ser612 phosphorylation of human Rb, which reduce the affinity for E2F by promoting self-association between the Rb pocket loop and E2F-binding site of Rb 32 . Although human CDK4/6-cyclin D can phosphorylate Ser608 and Ser612, the specificity of this phosphorylation remains disputed 33 . Mass spectrometric analysis of in vitro phosphorylated FZR-1 revealed eight CDK-4/CYD-1-phosphorylated residues in the N-terminal half of FZR-1 (Fig. 4d). Phosphopeptides containing these residues were abundant in the WT CDK-4 kinase reactions and were either not detectable or present at very low levels in the kinase dead CDK-4 control (Supplementary Table 3). All eight phosphorylated residues are followed by a proline residue and four are part of a T/SPxR/K consensus CDK phosphorylation site. For comparison, we also expressed human CDK4-cyclin D2 and found efficient phosphorylation of a human FZR1 N-terminal fragment (Fig. 4b). Mass spectrometric analysis identified four residues that were differentially phosphorylated by CDK4-cyclin D2 compared with the KD CDK4 control (Fig. 4e). These four phosphorylated residues of human FZR1 are also part of prolinecontaining sites. Moreover, their location in the FZR1 N-terminal domain is similar to the location of the CDK-4/CYD-1phosphorylated residues of C. elegans FZR-1 and at least two of the phospho sites are evolutionarily conserved (C.e. Ser45, Thr49 versus H.s. Ser36, Ser40).
A previous study showed that mutation to alanine of four conserved CDK consensus sites prevented inactivation of human  FZR1 (Cdh1) 31 . Two of these sites, Ser40 and Ser151, were phosphorylated by CDK4 in our in vitro kinase assays. Moreover, previous quantitative phosphorylation analysis showed cell cycle-dependent in vivo phosphorylation of FZR1 at five N-terminal serine residues, including Ser32, 36 and 151 (ref. 34). Together, our data indicate evolutionarily conserved inhibitory phosphorylation of Rb and FZR1 proteins, and are consistent with LIN-35 and FZR-1 inactivation in G1 through phosphorylation by CDK-4/CYD-1.

LIN-35
Rb and FZR-1 probably are CDK-4/CYD-1 targets in vivo. We considered alternative or complementary mechanisms for the elimination of CDK-4/CYD-1 requirement following Rb and FZR-1 reduction. Based on results in other systems, loss of FZR-1 might allow persistence of mitotic cyclins into the next G1 phase, or lead to premature CDK-2/CYE-1 cyclin E activation. The inappropriate activity of cyclin B or cyclin E kinases could bypass the CDK-4/CYD-1 kinase requirement. Immunohistochemical staining of fzr-1(he121) and fzr-1(ku298) mutants indicated normal degradation of CYB-1 cyclin B1 and CYB-3 cyclin B3 in mitosis ( Supplementary Fig. 4a-c). This is in agreement with mitotic cyclin degradation by APC/C in association with the Cdc20/Fizzy substrate specificity factor. Immunohistochemical staining of CYE-1 cyclin E also did not reveal apparent abnormalities, as CYE-1 expression was only detected in cells expected to be in S phase. Western blotting experiments were not conclusive but at most indicated slightly increased CYE-1 levels in fzr-1(ku298) and lin-35(n745); fzr-1(he121) larvae ( Supplementary Fig. 4). Human APC/C FZR1 regulates SCF Skp2 and thereby p21 Cip1 and p27 Kip1 degradation 35,36 . In analogy, activation of CDK-2/CYE-1 could also indirectly result from CKI-1 CIP/KIP depletion following fzr-1 loss. However, CKI-1::GFP levels were increased, rather than decreased, following RNAi of fzr-1, indicating that this control is not conserved in C. elegans ( Supplementary Fig. 5). Thus, although CDK-2/cyclin E may replace CDK-4/CYD-1 function, our data favour a model in which CDK-4/CYD-1 kinase activity is required to counter LIN-35 Rb-and APC/C FZR1 -mediated G1 inhibition.
CDK4/6 inhibitor-induced arrest of human cancer cells. Given our results in C. elegans, we wondered whether human tumour cells require CDK4/6-cyclin D activity to counteract APC/C FZR1 . To test this possibility, we combined treatment of human breast cancer cells with the CDK4/6-specific inhibitor PD-0332991 with knockdown of Rb and FZR1. As expected, treatment with 500 nM PD-0332991 induced cell division arrest in CAMA-1 luminaltype breast cancer cells and T47D ductal mammary carcinoma cells 15 . We examined incorporation of the thymidine analog EdU, which confirmed that the arrested cells lack DNA synthesis, and thus S phase (Fig. 5a,b). Transfection of CAMA-1 and T47D cells with pools of small interfering RNAs (siRNAs) targeting Rb or FZR1 messenger RNA caused substantial downregulation of the corresponding proteins (Fig. 5g,h). FZR1 knockdown resulted in limited rescue of the cell cycle arrest induced by PD-0332991 in CAMA-1 and T47D cells (Fig. 5e,f). The effect of Rb knockdown was somewhat stronger, but still limited. In contrast, double inhibition of FZR1 and Rb strongly rescued the cell cycle arrest induced by PD-0332991 in both cell types (Fig. 5c-f). These synergistic interactions between Rb and FZR1 closely resemble the observed interactions between cdk-4 CDK4/6, cyd-1 cyclin D, lin-35 Rb and fzr-1 FZR1 in C. elegans (Fig. 1c,d). Together, our data indicate that not only the mutant or WT status of Rb but also the expression levels of FZR1 and functional activity of the APC/C FZR1 E3 ligase will determine how cancer cells respond to CDK4/6 inhibitor treatment.

Discussion
We found that combined inhibition of the Rb transcriptional repressor and the FZR1 APC/C co-activator eliminates the normal requirement for CDK4/6-cyclin D kinase activity in C. elegans and human breast cancer cells. In addition, we showed that CDK4-cyclin D phosphorylates FZR1 and Rb at multiple target sites in vitro. These phosphorylated residues are present at very similar locations in the C. elegans and human Rb and FZR1 proteins. Previous studies have shown that phosphorylation at these sites corresponds to inhibition of human Rb or FZR1 (refs 31,32). As CDK4/6-cyclin D kinases are notoriously selective even in vitro, these combined observations make FZR1 a candidate novel target for regulation by cyclin D kinase phosphorylation. FZR1 is a well-known CDK substrate, but previous studies have largely focused on inactivating phosphorylations that occur following the G1/S transition, in particular phosphorylation by CDK2-cyclin A and mitotic CDK1 (refs 27-29,31). Our findings imply that regulation of APC/C FZR1 by CDK4/6-cyclin D forms part of the molecular network controlling passage through the G1 restriction point. At least one study in yeast implicated G1-CDKs in the inactivation of APC Cdh1 (ref. 37). However, other experiments showed that degradation of at least some of the substrates of APC Cdh1 continues into S phase, and that S phase cyclins are required for full APC Cdh1 inactivation 38 . Based on these data and our own results, we hypothesize that the inactivating phosphorylation of FZR1 may follow a pattern similar to Rb phosphorylation. As such, initial cyclin D-kinase phosphorylation might weaken the G1 inhibitory function of APC/C FZR1 . Combined with a simultaneous reduction in Rb-mediated transcriptional repression by CDK4/6-cyclin D phosphorylation, this may allow some induction of CDK2cyclin E activity. Phosphorylation by CDK2-cyclin E fully inactivates Rb and possibly APC/C FZR1 , triggering a positive feedback loop and commitment to S phase entry. Further experimentation will be needed to test this model, and to explain how association with pseudo-substrate inhibitors, which provide an alternative mechanism for APC/C FZR1 inactivation, contributes to this switch.
The observed contribution of FZR1 loss in eliminating CDK4/ 6-cyclin D requirement could involve a more indirect mechanism. Mutation of Cdh1/Hct1 in yeast or fzr in Drosophila allows persistence of mitotic cyclins beyond M phase 22,39 . This ectopic cyclin expression triggered S phase in pheromone-exposed yeast cells and Drosophila cells that undergo a final mitosis 22,40 . Our EdU incorporation experiments in C. elegans did not reveal ectopic S phases, even when fzr-1 was strongly inactivated. Moreover, lin-35(0); fzr-1(he121) double mutants and fzr-1 strong loss of function mutants showed normal cyclin B1 (CYB-1) and cyclin B3 (CYB-3) degradation in mitosis. Therefore, we do not expect that persistence of mitotic cyclins is responsible for the fzr-1 phenotype. The APC/C in association with its other coactivator, FZY-1 (Cdc20), probably retains prolonged activity in the absence of FZR-1 and suffices for mitotic cyclin degradation, as has been observed in human cells 41,42 .
As an alternative possibility, premature activation of cyclin E-kinase activity could bypass cyclin D-kinase requirement in lin-35 Rb, fzr-1 double mutants. Although loss of Rb is likely to lead to de-repression of cye-1 cyclin E transcription, reduced APC/C FZR-1 activity has been reported to alleviate CDK2-cyclin E kinase inhibition in human cells 35,36 . The latter effect involves the F-box factor SKP2, which acts as the substrate specificity factor of the E3 ubiquitin ligase SCF. SKP2 is one of the best-characterized targets for APC/C FZR1 in human cells. Hence, downregulation of APC/C FZR1 activity leads to increased SCF SKP2 function, and thereby to degradation of its substrates, the p21 Cip1 and p27 Kip1 CDK inhibitors 35,36 . The protein expression level of CKI-1, the C. elegans homologue of Cip1/Kip1, was increased in fzr-1 mutants, rather than reduced, compared to WT. This indicates that the link between APC/C FZR1 and Cip1/Kip1 regulation is probably not conserved in C. elegans. Therefore, we favour the model that LIN-35 Rb and FZR-1 are the two critical targets for inhibitory phosphorylation by the CDK-4/CYD-1 kinase, which when removed allow cyclin D-kinase-independent cell cycle entry.
We cannot exclude that the above-mentioned indirect mechanisms for elimination of CDK4/6-cyclin D requirement contribute to the escape from PD-0332991-induced arrest of human cancer cells following Rb and FZR1 knockdown. Independent of the mechanism, however, it is important to realize that reduced FZR1 levels correspond to reduced sensitivity for CDK4/6-specific inhibitors. At least three different CDK4/6specific inhibitors are in various phases of clinical trials with human cancer patients [16][17][18] . Given the promise of these inhibitors in cancer treatment, it will be important to identify which genetic makeup corresponds with sensitivity to, or allows escape from, drug treatment. Although mutations in FZR1 are not common, the level of FZR1 protein expression and functional activity can be compromised in human cancer 42 . Probably reflecting this reduced activity, overexpression of the APC/C FZR1 substrate SKP2 has been observed in a subset of human cancers with poor prognosis 43,44 . Future studies should determine whether expression levels of FZR1 or its proteolytic targets such as SKP2 provide prognostic markers for the response of cancer cells to CDK4/6 inhibitor treatment.

Methods
C. elegans culture. Strains used are described in the Supplementary Table 4. Strains were cultured on Nematode Growth Medium (NGM) plates seeded with Escherichia coli OP50 according to standard protocol. All experiments were conducted at 20°C, unless indicated otherwise. Feeding RNAi was performed on NGM plates supplied with 50 mg ml À 1 Ampicillin and 2 mM IPTG. Soaking RNAi was performed as previously described 45 , using synchronized L1 larvae obtained by hypochlorite treatment and hatching of eggs in M9 medium with 0.05% Tween-20. L1 larvae were soaked in 10 ml dsRNA solution for 23 h at 20°C, transferred to NGM plates with OP50 and allowed to develop for the appropriate amount of time.
DNA plasmid construction and transgenics. For primers used, see Supplementary Table 5.
cdk-4::flag and cdk-4kd::flag. To create the cdk-4wt::flag and cdk-4kd::flag microparticle bombardment constructs, we cloned a genomic EcoRV fragment containing the entire cdk-4a coding region as well as 3,466 bp of upstream and 770 bp of downstream sequences into vector pBluescript SK( þ ). To add a flag tag, site-directed mutagenesis was used to create an MfeI site at the stop codon (changing CAAGTGA to CAATTGA). A linker containing the flag tag, created by annealing oligonucleotides FLAG tag linker F and FLAG tag linker R was inserted into this MfeI site. For the KD construct, QuickChange site-directed mutagenesis was used to create an Asp/Asn substitution at amino-acid position 187 using primers Asp187Asn F and Asp187Asn R. To generate the final microparticle bombardment constructs, both the WT and KD cdk-4 constructs were cloned into the unc-119(wt) vector pDP#MM016b 46 . These constructs were subsequently transformed into unc-119(ed3) worms by microparticle bombardment 47 .
lin-35::GFP. Fosmid WRM063cF08 (Source Bioscience) containing 20 kb upstream and 9.5 kb downstream sequences surrounding lin-35 was used to generate lin-35::GFP with a recombination-based approach described previously 50 . This homologous genetic engineering technique in bacteria allowed us to integrate the GFP tag into a large genomic region of almost 40 kb, probably containing all the native regulatory sequences required for proper lin-35 expression.
GST fusion proteins. Fragments encoding the N or C terminus of C. elegans FZR-1 were PCR amplified from a C. elegans cDNA library with fzr-1_1 SalI-F and fzr-1_1197 Not1-R for the N terminus and for the C terminus, fzr-1_988 EcoRI-F and fzr-1_2107 NotI-R. Subsequently, the PCR fragments were cloned into pGEX 4T-3. For GST-LIN-35, a lin-35 fragment encoding the extended C-terminal region with the A and B pocket domains was amplified by PCR from a cDNA library with lin-35_1486 SmaI-F and lin-35_2883_NotIR, and cloned into pGEX 4T-2.
Human GST-FZR1 proteins. To generate human GST-FZR1 proteins, fzr1 cDNA was PCR amplified from a Homo sapiens cDNA library with fzr1_Hs_EcoRI_1 F and fzr1_Hs_NotI_598 R, and cloned into the EcoRI/NotI sites of pGEX4T-3, resulting in an in-frame GST fusion protein. The constructs were introduced into E. coli BL21 and expression was induced with 2 mM IPTG for 3 h. The GST fusion proteins were purified by binding to Glutathione Sepharose beads (GE Healthcare).
Microscopy analysis. Microscopy was performed at 20°C. Two microscope setups were used: a wide-field immunofluorescence microscope and a spinning disc confocal microscope. The wide-field fluorescence microscope is a Zeis Axioplan2 upright microscope equipped with a HAL 100 halogen visible light source controlled by an internal shutter, an HXP 120 C metal halide fluorescence light source controlled by a Uniblitz VMM-D1 shutter, a Â 25/0.8 NA imm Corr objective with DIC slider, Â 63 and Â 100 Plan-APOCHROMAT 1.4 NA objectives with DIC sliders, a DIC polarizer and an Axiocam MRm CCD monochrome camera. The filter turret is equipped with the following Zeiss filter sets: filter set 34 for DAPI (excitation: 390/22 nm band pass, dichroic mirror: 420 nm, emission: 460/50 nm band pass), filter set 13 for GFP (excitation: 470/20 nm band pass, dichroic mirror: 495 nm, emission: 505-530 nm band pass), filter set 46 for YFP (excitation: 500/ 20 nm band pass, dichroic mirror: 515 nm, emission: 535/30 nm band pass), filter set 31 for mCherry (excitation: 565/30 nm band pass, dichroic mirror: 585 nm, emission: 620/60 nm band pass) and a DIC analyser cube. The microscope and camera are controlled by Zeiss Axiovision 4.x imaging software. Confocal images were obtained with a Zeiss LSM 700 confocal microscope.
A similar experiment was performed with human CDK4 and KD CDK4 D158N , co-transfected and expressed with cyclin D2. Kinase assays were performed with GST only or GST fused to the H. sapiens FZR1 N terminus (a.a. . The CDK4/6 inhibitor PD-0332991 was added to a final concentration of 0.9 mM. Mass spectrometry. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium 55 via the PRIDE partner repository (see accession codes section). Gel lanes (WT and KD) were cut into different bands, which were treated with 6.5 mM dithiothreitol for 1 h at 60°C for reduction and 54 mM iodoacetamide for 30 min for alkylation. The proteins were digested overnight with trypsin at 37°C. The resulting peptides were extracted with 10% formic acid.
The samples were analysed using a Proxeon Easy-nLC 1000 (Thermo Scientific) connected to an QExactive mass spectrometer (Thermo Scientific). The injected samples were first trapped (Dr Maisch Reprosil C18, 3 mm, 2 cm, 100 mm) at a maximum pressure of 600 bar with solvent A (0.1% formic acid in water) before being separated on an analytical column (Agilent Poroshell EC-C18, 2.7 mm, 40 cm, 50 mm) at a stable temperature of 40°C. Peptides were chromatographically separated by a 90-min gradient from 7% to 30% solvent B (0.1% formic acid in acetonitrile) at a flow rate of 100 nl min À 1 . The column eluent was directly introduced into the elctrospray source of the mass spectrometer. The electrospray voltage was set to 1.7 kV using a fused silica capillary (360 mm o.d., 20 mm i.d., 10 mm tip i.d., constructed in-house). The mass spectrometer was used in a datadependent mode, which automatically switched between MS and MS/MS using a Top10 method (higher-energy collision dissociation fragmentation). Alternativley, samples were also analysed in an Agilent HPLC 1200 series (Agilent) connected to an LTQ Orbitrap Velos (Thermo Scientific) using a data-dependent decision tree as before 56 .
Proteome data analysis. Raw files were processed using Proteome Discoverer 1.4 (version 1.4.1.14, Thermo Scientific). Mascot (Matrix Science, version 2.4) was used as a search engine against a combined database of the human and E. coli proteomes in which GST-Frz1 and GST-LIN-35 sequences were added. Carbamidomethylation (C) was set as a fixed modification, and oxidation (M) and phosphorylation (STY) were set as variable modifications. The PhosphoRS 3.0 node was used for the phosphosite localization. Trypsin was specified as enzyme and up to two missed cleavages were allowed. The following parameters were used: 50 p.p.m. for the precursor mass tolerance and 0.05 Da for the fragment ion tolerance (0.5 Da for ion trap data). Data filtering was performed using Percolator, resulting in 1% false discovery rate. Additional filters were applied as follows: search engine rank 1 peptides, minimum of 7 residues per peptide and Mascot ion score above 20.
Extracted ion chromatograms were obtained for each phosphopeptide in both the KD and WT kinase (CDK4) experiments. The intensities of all the phosphopeptides containing the same phosphosite(s) were summed (for example, missed cleavages and methionine oxidation) to obtain a total signal in each experiment. The intensities for the different charge states of the same phosphopeptide (for example, 2 þ , 3 þ , 4 þ and so on) were also summed. When multiple phosphopeptides were identified from the same chromatographic peak, only one value was used for the calculation. Normal peptides (that is, non-phosphopeptides) from FZR1 were used to normalize the data between the two samples.
Cell culture and inhibitor assays. HEK293T cells were cultured in 50% DMEM/50% F10 with L-glutamine (Lonza) supplemented with 10% FCS and antibiotics. Cell transfections were performed with 25 kD linear polyethyleninmine (Polysciences) 57 . CAMA-1 and T47D cells were cultured in RPMI-1640 medium (Life Technololgies), supplemented with 10% FCS. Cells were seeded on coverslips and transfected for 3 days with 10 nM of ON-TARGET plus SMARTpool siRNA (Thermo Scientific) targeting FZR1 and/or RB1, using HiPerFect as previously described 57 , and compared with cells transfected with luciferase-targeting siRNA 58 . The CDK4/6 inhibitor PD 0332991 was added in a concentration of 500 nM in dimethyl sulfoxide for 24 h. The Click-iT EdU Imaging Kit (Life Technologies) assays were performed according to the manufacturer's instructions. Briefly, cells transfected for 3days with siRNA were incubated 7 h with 20 mM EdU, fixed, stained using Click-iT EdU Alexa Fluor 594 Imaging Kit and counterstained with DAPI. For quantification, ten independent fields spanning duplicate experiments were imaged for Alexa Fluor 594 and DAPI fluorescence, and counted for EdU nuclear positivity relative to total number of nuclei per field.