Template-driven scaffolding of SCFFBXO42 regulates PP2A degradation Authors: Sebastien Coassolo, Nairie Michaelian, Timurs Maculins, Caleigh M. Azumaya, Tommy K. Cheung, Jianping Yin, Inna Zilberleyb, Kanika Bajaj Pahuja, Thomas Garner, Ted Lau, Davis Mau, Matthew Grimmer, Jean-Philippe Fortin, Mike Costa, Yoana N. Dimitrova, Christopher M. Rose, Peter L. Hsu, Robert L. Yauch Published: 2026 ABSTRACT Protein phosphatase 2A (PP2A) is a Ser/Thr phosphatase that regulates the phosphorylation of almost all cellular processes, including cell division and proliferation1,2. PP2A forms heterotrimeric holoenzyme complexes comprising a catalytic subunit (PP2Ac), a scaffolding subunit (PP2Aa) and variable B regulatory subunits that exert precise control over enzyme substrate specificity and prevent indiscriminate dephosphorylation of phosphoproteins3. However, the mechanisms that control the activity of uncomplexed catalytic subunits have remained relatively unclear. Here we find that the E3 ligase SKP1–CUL1–F-box (SCF) complex containing F-box other protein 42 (FBXO42, also known as JFK; hereafter, SCFFBXO42) degrades holoenzyme-free PP2Ac in a complex with the coiled-coil protein CCDC6 to maintain cancer cell fitness. The cryo-electron microscopy structure of the FBXO42–CCDC6–PP2Ac assembly reveals a pseudosymmetric architecture in which CCDC6 forms a central dimeric template that recruits multiple copies of PP2Ac and creates a substrate for FBXO42. Both the quaternary structure of this CCDC6–PP2Ac heterodimer and the post-translationally methylated tail of PP2Ac are recognized by FBXO42 for ubiquitination. The multivalent structure facilitated by CCDC6 enables the assembly of multiple degradation complexes along a single coiled coil, leading to the turnover of free phosphatases and downregulation of catalytic activity. Together, our findings define a mechanism for PP2A control through the ubiquitin-proteosome system and establish a paradigm for cullin-RING ligase–substrate interactions. Main PP2A is a Ser/Thr phosphatase and a master regulator of the phosphoproteome that has an important role in various cellular processes, including cell cycle regulation, by dephosphorylation of key proteins4. Structurally, PP2A is a heterotrimeric complex comprising scaffolding (PP2Aa), catalytic (PP2Ac) and various B regulatory subunits. While the A and C subunits assemble to form the canonical core enzyme, the B subunits are exchanged to define phosphatase substrate specificity3. Given their key role in numerous cellular functions, PP2A complexes are tightly regulated through multiple mechanisms, including direct inhibition by other proteins and post-translational modification that signal for the assembly or disassembly of holocomplexes5–8. After disassembly, the activity of the PP2Ac catalytic subunit in isolation from the scaffold and B subunits must be kept in check to prevent unregulated and uncontrolled dephosphorylation of target proteins. While there have been links to proteasomal degradation as a means to tune PP2Ac levels5,9,10, the molecular mechanisms of PP2A regulation by the ubiquitin system are poorly understood. FBXO42 is a substrate-recognition component of the multi-subunit SCF complex, which has been previously reported to regulate the stability of both the wild-type (WT) and hotspot mutant forms of p5311–13 and several other putative substrates11,14–18. FBXO42 has recently been shown to be essential in a broad range of cancer cells and its depletion led to chromosome misalignments, cell cycle arrest and death in a glioblastoma model19. However, no known FBXO42 substrates have been connected to this phenotype to date, suggesting that uncharacterized substrates drive cancer cell survival. In an effort to understand FBXO42’s role in cancer cell fitness, we show that FBXO42 targets the catalytic subunit of PP2A for ubiquitination and degradation through a recognition mechanism that requires the assembly of a degradosome complex consisting of multiple PP2Ac monomers bound to the coiled-coil protein CCDC6. Notably, this assembly can be used to target the related PP4c, revealing a common mechanism that the cell uses to clear out unneeded phosphatases. FBXO42 regulates cancer cell fitness Whole-genome CRISPR loss-of-function screens previously identified FBXO42 as essential for a broad array of cancer cell lines; however, the underlying substrate necessary for maintaining cell fitness has not been identified. TP53 represents a putative candidate substrate given the previously defined role of FBXO42 in regulating both WT and hotspot point mutant p5311–13. However, we observed that the cellular dependency on FBXO42 can also occur in subsets of cancer cell lines with mutational loss of TP53 (Extended Data Fig. 1a). To validate this observation, we selected a sample of cell lines encoding WT or mutationally inactivated TP53 and evaluated the effect of small interfering RNA (siRNA)-mediated knockdown (KD) on clonogenic growth. FBXO42 KD was sufficient to substantially reduce clonogenic growth of cell lines with no functional TP53 and had no effect on growth of the selected TP53-WT cells (Extended Data Fig. 1b). FBXO42 KD resulted in mitotic arrest, as determined by immunofluorescence (IF) staining of a marker of mitotic chromosome segregation (phosphorylated histone 3 Ser10) (Extended Data Fig. 1c), accompanied by a specific blockade in prophase and enrichment of abnormal and polynucleated cells (Extended Data Fig. 1d–f). FBXO42 contains an N-terminal F-box motif that binds to the SKP1–CUL1–RBX1 scaffold, and a C-terminal Kelch β-propeller to recognize putative substrate proteins for ubiquitination11. To determine whether the loss of FBXO42 E3 ligase function is responsible for the growth defect, we derived stable clones overexpressing exogenous WT or truncated FBXO42 from TP53-mutated HOP-62 cells (Extended Data Fig. 1g). Cells were transfected with siControl or siFBXO42 targeting the endogenous gene, and confluency was monitored. As expected, the expression of WT FBXO42, but not an empty vector, rescued the FBXO42-KD phenotype (Extended Data Fig. 1h). Loss of either the SKP1-binding F-box motif or the putative substrate-binding Kelch domain did not rescue cell growth. This suggests that the phenotype is due to loss of FBXO42 E3 ligase function, probably owing to substrate stabilization. Together, these data demonstrate the essentiality of FBXO42 in the context in which TP53 is lost, suggesting that a critical and missing substrate protein(s) is regulated by FBXO42 in cancer cell growth. Identification of FBXO42 activity modifiers As SCF complexes are generally acknowledged to initiate degradative ubiquitination20, we hypothesized that FBXO42 loss may result in stabilization of its substrate(s), thereby causing the resultant phenotype. To identify modifiers of FBXO42 function, we performed a genome-wide, pooled CRISPR knockout (KO) screen to identify gene KOs capable of overcoming the FBXO42-depletion phenotype. To this end, we generated a doxycycline (DOX)-inducible Cas9 clone of HOP-62 cells expressing a single guide RNA (sgRNA) targeting the FBXO42 gene (sgFBXO42), which resulted in protein loss and arrested cell growth (Fig. 1a,b). After introduction of a whole-genome sgRNA library, cells were treated with DOX for 10 days, and sgRNAs were sequenced to identify differential abundance (Fig. 1c and Methods). We detected significant enrichment for sgRNAs targeting the protein phosphatase 2 catalytic subunit alpha gene (PPP2CA), suggesting that depletion of PP2Ac protein could overcome the effect of FBXO42 KO (Fig. 1d and Supplementary Tables 1 and 2). Notably, our survival screen did not demonstrate any significant enrichment for sgRNAs targeting genes of the scaffolding or regulatory subunits of PP2A (Extended Data Fig. 2a). Nonetheless, sgRNAs targeting LCMT1, a positive regulator of PP2Ac activity6, were also found to be enriched, further suggesting that PP2Ac deregulation has a role in the observed phenotype (Fig. 1d and Extended Data Fig. 2b).Fig. 1FBXO42 targets the catalytic subunit of PP2A for ubiquitination to maintain cancer cell fitness.a, WB verification of FBXO42-KO efficiency in DOX-inducible CRISPR–Cas9 FBXO42-KO HOP-62 cells. b, Live-cell imaging analysis of cell confluency after DOX treatment in cells from a. Data are mean ± s.e.m. n = 3 biologically independent experiments. c, CRISPR KO survival screen workflow. The diagram was created using BioRender; Coassolo, S. https://BioRender.com/8flz312 (2026). d, CRISPR KO survival screen showing gene-level log2[fold change (FC)] versus −log[false-discovery rate (FDR)] of DOX-treated cells after 10 days versus the baseline. Puro, puromycin. e, Cell confluency was monitored using live-cell imaging after DOX-inducible FBXO42 KO in the presence of okadaic acid (OA). Data are mean ± s.e.m. n = 3 independent experiments. f, Volcano plot of the ubiquitinome detected by KGG-MS showing –log2[FC] (DOX versus DMSO) against significance (log10[P]) after DOX-inducible FBXO42 expression in HOP-62 cells (left). The red dots represent adjusted P < 0.05. n = 3 biologically independent samples. Right, the structure of PP2Ac with the ubiquitinated Lys34 highlighted by yellow spheres. g, WB analysis of the expression of PP2Ac and total ubiquitin from TUBE enrichment of DOX-inducible FBXO42 ORF HOP-62 nuclear lysates in the presence of MG132. h, WB of the V5 tag in nuclear lysates of HEK293T cells expressing exogenous V5-PP2Ac after siCtrl or siFBXO42 KD in the presence of MG132. i, WB analysis of the expression of Flag and PP2Ac in Flag–FBXO42–SKP1 recombinant protein IPs from HOP-62 nuclear lysates treated with DMSO, MG132 or MLN4924. j, WB of FBXO42 and PP2Ac from a panel of lung cancer cell lines after siCtrl or siFBXO42 KD for 72 h.Source data FBXO42 targets the catalytic subunit of PP2A for ubiquitination to maintain cancer cell fitness. a, WB verification of FBXO42-KO efficiency in DOX-inducible CRISPR–Cas9 FBXO42-KO HOP-62 cells. b, Live-cell imaging analysis of cell confluency after DOX treatment in cells from a. Data are mean ± s.e.m. n = 3 biologically independent experiments. c, CRISPR KO survival screen workflow. The diagram was created using BioRender; Coassolo, S. https://BioRender.com/8flz312 (2026). d, CRISPR KO survival screen showing gene-level log2[fold change (FC)] versus −log[false-discovery rate (FDR)] of DOX-treated cells after 10 days versus the baseline. Puro, puromycin. e, Cell confluency was monitored using live-cell imaging after DOX-inducible FBXO42 KO in the presence of okadaic acid (OA). Data are mean ± s.e.m. n = 3 independent experiments. f, Volcano plot of the ubiquitinome detected by KGG-MS showing –log2[FC] (DOX versus DMSO) against significance (log10[P]) after DOX-inducible FBXO42 expression in HOP-62 cells (left). The red dots represent adjusted P < 0.05. n = 3 biologically independent samples. Right, the structure of PP2Ac with the ubiquitinated Lys34 highlighted by yellow spheres. g, WB analysis of the expression of PP2Ac and total ubiquitin from TUBE enrichment of DOX-inducible FBXO42 ORF HOP-62 nuclear lysates in the presence of MG132. h, WB of the V5 tag in nuclear lysates of HEK293T cells expressing exogenous V5-PP2Ac after siCtrl or siFBXO42 KD in the presence of MG132. i, WB analysis of the expression of Flag and PP2Ac in Flag–FBXO42–SKP1 recombinant protein IPs from HOP-62 nuclear lysates treated with DMSO, MG132 or MLN4924. j, WB of FBXO42 and PP2Ac from a panel of lung cancer cell lines after siCtrl or siFBXO42 KD for 72 h. Source data To test this hypothesis, we subjected DOX-inducible FBXO42 KO cells to treatment with okadaic acid—a broad-acting protein serine/threonine phosphatase (PPP) inhibitor. Owing to the cellular toxicity of this compound, we determined a concentration (10 nM) that had a minimal impact on HOP-62 cell proliferation but significantly reduced phosphatase activity in cells (Extended Data Fig. 2c,d). Okadaic acid treatment partially reversed the phenotype mediated by FBXO42 KO (Fig. 1e). Similarly, in a separate model (NCI-H1373 cells) in which a degradation tag (DTAG) was genetically knocked into the endogenous FBXO42 locus (DTAG–FBXO42), low dose okadaic acid treatment could further overcome the fitness effect caused by FBXO42 degradation mediated by the DTAGv1 ligand21 (Extended Data Fig. 2e,f). These results further indicate that the deregulation of protein phosphatases, in particular PP2Ac, is a key driver of the FBXO42-depletion phenotype. FBXO42 targets PP2Ac for degradation To identify putative substrates of FBXO42 using an unbiased approach, we used KGG mass spectrometry (KGG-MS), which captures lysine residues carrying a diglycine (K-ε-GG) remnant after trypsin digestion, in engineered HOP-62 cells overexpressing DOX-inducible Flag–FBXO42 treated with MG132 to prevent rapid turnover of ubiquitinated proteins (Extended Data Fig. 3a,b). In summary, 26 proteins exhibited significant KGG peptide enrichment after FBXO42 overexpression. Importantly, the most significantly enriched ubiquitinated peptide aligned to PP2Ac (KQLSESQVKSLCEK34AK) (Fig. 1f and Supplementary Table 3). A few peptides that mapped to other proteins were also enriched (Extended Data Fig. 3c), but were not further investigated owing to their absence in our previous survival screen (Fig. 1d). To confirm this result, we used tandem ubiquitin-binding entities (TUBEs) to enrich for polyubiquitinated proteins in cell lysates after DOX induction of FBXO42. Consistent with the KGG-MS data, PP2Ac was enriched in TUBE pull-downs after FBXO42 overexpression (Fig. 1g), while siFBXO42 in cells expressing exogenous V5-PP2Ac demonstrated reduced ubiquitination (Fig. 1h), therefore confirming PP2Ac as a putative cellular substrate of FBXO42. To determine whether FBXO42 physically associates with PP2Ac, we incubated recombinant Flag–FBXO42–SKP1 protein in HOP-62 cell lysates, and performed a western blot (WB) analysis of PP2Ac after Flag immunoprecipitation (IP). We observed PP2Ac association only under conditions in which cells were pretreated with MG132 or a Neddylation inhibitor (MLN-4924) to hinder cullin-RING ligase activity (Fig. 1i), suggesting that PP2Ac is targeted by FBXO42 for proteasomal degradation. We next evaluated the PP2Ac half-life using a cycloheximide (CHX) chase assay to determine whether the PP2Ac half-life could be altered by FBXO42 depletion. Under normal conditions in HOP-62 cells, the PP2Ac half-life is approximately 36 h; FBXO42 depletion extends the PP2Ac half-life to nearly 60 h (Extended Data Fig. 3d). Furthermore, siRNA-mediated KD of FBXO42 in a small panel of lung cancer cell lines resulted in stabilization of the PP2Ac protein (Fig. 1j), suggesting that FBXO42-mediated cancer cell survival across these lines may involve a common mechanism. Notably, visualization of PP2Ac stabilization after depletion is most apparent in nuclear fractions, implying that FBXO42 may more strongly regulate the nuclear pool of PP2Ac (Extended Data Fig. 3e). For these reasons, certain experiments required nuclear fractionation for improved resolution, which is indicated in the corresponding figure legends. CCDC6 associates with FBXO42 We were unable to biochemically isolate a stable complex between SCFFBXO42 and PP2Ac using recombinant proteins (Extended Data Fig. 4a); we therefore hypothesized that another protein may be necessary to facilitate stable complex formation between E3 and substrate. To explore this possibility, we immunoprecipitated a ligase-inactive Flag–FBXO42(ΔF-box) mutant in HOP-62 cells followed by global proteomic analysis using IP coupled with MS (IP–MS) to identify FBXO42 interactors. Relative to the control IP, we found strong enrichment for peptides corresponding to RBPJ, a previously identified FBXO42-interacting protein14, and CCDC6, a poorly studied coiled-coil protein (Fig. 2a and Supplementary Table 4). Notably, the effects of CCDC6 KO displayed an exceptionally strong correlation with FBXO42 KO in the Cancer Dependency Map (r = 0.75), suggesting a functional association between these proteins in addition to their physical interaction (Extended Data Fig. 4b). Using siCCDC6 in the same panel of FBXO42-sensitive cancer cell lines, we found that CCDC6 depletion phenocopies the FBXO42 KD viability effect, as well as mitotic defects (Extended Data Fig. 4c,d). While our data and the literature do not suggest that CCDC6 is a substrate of FBXO4219 (Fig. 1f), CCDC6 can co-IP with FBXO42–SKP1 only when cells are pretreated with either MG132 or MLN-4924, suggesting that this interaction is labile when ubiquitination and/or the proteasome is active (Fig. 2b). We found that PP2Ac could co-IP with Flag–CCDC6 in the absence of MG132 or MLN-4924 (Extended Data Fig. 4e), raising the possibility that FBXO42 is engaging PP2Ac bound to CCDC6. To investigate this, we immunoprecipitated FBXO42 from HEK293T cells ectopically expressing Flag–FBXO42 after siRNA-mediated KD of CCDC6. Notably, loss of CCDC6 protein completely abrogated the ability of FBXO42 to interact with PP2Ac (Fig. 2c). Similar results were obtained in HOP-62 cells expressing the ligase-dead FBXO42(ΔF-box) protein after CCDC6 siRNA KD (Fig. 2d). These results suggest that CCDC6 is required for the ability of FBXO42 to recognize PP2Ac in a broad range of cell types. To test this at the biochemical level, we set up an in vitro ubiquitination assay using purified components. We found that FBXO42 could only weakly catalyse polyubiquitin chains on PP2Ac; however, addition of CCDC6 into the reaction substantially increased the efficiency at which PP2Ac was ubiquitinated (Fig. 2e).Fig. 2CCDC6 bridges PP2Ac to FBXO42 for ubiquitination.a, Volcano plot of the proteome detected by MS in HOP-62 Flag–FBXO42(ΔF-box) immunoprecipitates, shown as log2[FC] versus significance (log10[P]) relative to control IPs. n = 3 biologically independent samples. b, WB of Flag or CCDC6 in Flag–FBXO42–SKP1 recombinant protein IPs from HOP-62 nuclear lysates treated with DMSO, MG132 or MLN4924. c, WB of PP2Ac in anti-Flag IPs from HEK293T cells expressing Flag–FBXO42 after siCtrl or siCCDC6 KD in the presence of MLN4924. WB with inputs for Flag, CCDC6 and HDAC1. d, WB of Flag, CCDC6 and PP2Ac in anti-Flag IPs from lysates of HOP-62 cells expressing Flag–FBXO42 WT or Flag–FBXO42(ΔF-box) that were previously transfected with siCtrl or siCCDC6 for 72 h. Input WB for Flag, CCDC6 and PP2Ac. e, PP2Ac in vitro ubiquitination assay incubated for 5 min at 30 °C in the presence of increasing concentrations of CCDC6. Ubiquitination was detected using an antibody against PP2Ac, generating a laddering pattern at higher molecular masses, indicative of polyubiquitination. f, A representative HTRF assay to determine association of biotin–FBXO42–SKP1 with methylated or unmethylated PP2Ac. The samples were incubated for 24 h with LCMT-1 in the presence or absence of S-adenosyl methionine (SAM). Each dot represents a technical replicate (n = 8).Source data CCDC6 bridges PP2Ac to FBXO42 for ubiquitination. a, Volcano plot of the proteome detected by MS in HOP-62 Flag–FBXO42(ΔF-box) immunoprecipitates, shown as log2[FC] versus significance (log10[P]) relative to control IPs. n = 3 biologically independent samples. b, WB of Flag or CCDC6 in Flag–FBXO42–SKP1 recombinant protein IPs from HOP-62 nuclear lysates treated with DMSO, MG132 or MLN4924. c, WB of PP2Ac in anti-Flag IPs from HEK293T cells expressing Flag–FBXO42 after siCtrl or siCCDC6 KD in the presence of MLN4924. WB with inputs for Flag, CCDC6 and HDAC1. d, WB of Flag, CCDC6 and PP2Ac in anti-Flag IPs from lysates of HOP-62 cells expressing Flag–FBXO42 WT or Flag–FBXO42(ΔF-box) that were previously transfected with siCtrl or siCCDC6 for 72 h. Input WB for Flag, CCDC6 and PP2Ac. e, PP2Ac in vitro ubiquitination assay incubated for 5 min at 30 °C in the presence of increasing concentrations of CCDC6. Ubiquitination was detected using an antibody against PP2Ac, generating a laddering pattern at higher molecular masses, indicative of polyubiquitination. f, A representative HTRF assay to determine association of biotin–FBXO42–SKP1 with methylated or unmethylated PP2Ac. The samples were incubated for 24 h with LCMT-1 in the presence or absence of S-adenosyl methionine (SAM). Each dot represents a technical replicate (n = 8). Source data Evidence in the literature has shown not only that PP2Ac is heavily C-terminally carboxy-methylated, but also that the methylation state of PP2Ac influences its proteolytic stability in cells6,22. To test this, we methylated PP2Ac (mPP2Ac) using recombinantly purified LCMT-1 (Extended Data Fig. 4f) and tested mPP2Ac binding to FBXO42 in the presence of CCDC6 using an in vitro homogenous time-resolved fluorescence (HTRF) assay. C-terminal carboxy-methylation of PP2Ac resulted in a considerable increase in HTRF signal, indicating that the methylation state of the PP2Ac C-terminal tail contributes to FBXO42 recognition (Fig. 2f). Structure of the PP2Ac degradosome To understand the structural basis for CCDC6-dependent mPP2Ac ubiquitination by FBXO42, we biochemically reconstituted a complex between SCFFBXO42, CCDC6 and mPP2Ac for structural determination (Extended Data Fig. 4g). We solved the structure of an SCFFBXO42–CCDC6–mPP2Ac (FCP) complex to an overall resolution of 3.2 Å using cryo-electron microscopy (cryo-EM) (Fig. 3a and Extended Data Table 1). Focused refinements resulted in substantial map improvements, enabling de novo model building throughout the structure (Extended Data Figs. 5 and 6a–f). From the same dataset, we were also able to determine a structure of a CCDC6–PP2Ac (CP) subcomplex to an overall resolution of 3.5 Å (Extended Data Fig. 5).Fig. 3The overall structure of the FBXO42–CCDC6–PP2Ac degradosome.a, A composite cryo-EM map of the FBXO42–SKP1–CCDC6–mPP2Ac complex at an overall resolution of 3.2 Å. CCDC6 is shown in red, FBXO42 in slate, SKP1 in wheat and PP2Ac in grey. b, The structure of the FBXO42–SKP1–CCDC6–mPP2Ac complex shown from the same angle as in a (left) and rotated 90° to look down CCDC6 (right). The colouring of the subunits is identical to a. Unmodelled loops are represented as dashed lines. c, Schematic of the HTRF competition assay. The TR-FRET ratio was generated owing to the interaction between labelled Flag–HA–mPP2Ac (grey) and biotinylated FBXO42–SKP1 (slate and wheat) in the presence of untagged CCDC6 (red). Unlabelled FBXO42–SKP1 (faded slate and wheat) was titrated into the reaction inhibiting the TR-FRET ratio. The diagram was created using BioRender; Maculins, T. https://BioRender.com/zk6pu21 (2025). d, The nonlinear fit of the HTRF competition assay results performed at different mPP2Ac:CCDC6 ratios. The HTRF signal was normalized to the total FBXO42 (normalized HTRF signal = (observed HTRF signal/[FBXO42labelled]) × [FBXO42total]). n = 3 biologically independent experiments. Data are means ± s.d. Calculated KD values are reported in Supplementary Table 5. e, Cryo-EM map of 3×FBXO42–SKP1 bound to CCDC6–mPP2Ac at an overall resolution of 6.9 Å (left). Subunits are coloured as in a. Right, the structure of 3×FBXO42–SKP1-bound CCDC6–mPP2Ac rotated 90° relative to left panel with modelled CUL1–RBX1 (green and orange; Protein Data Bank (PDB): 1LDK). f, Superposition of a PP2a+c (PDB: 3FGA) heterodimer into the FCP complex. PP2Aa is shown in cyan as cartoon cylinders, and the FCP complex is shown as a transparent surface and coloured as in a. g, In vitro ubiquitination assay using recombinantly purified Flag–FBXO42–SKP1, CCDC6, PP2Ac and PP2Aa+c. Ubiquitination was detected using an antibody against PP2Ac.Source data The overall structure of the FBXO42–CCDC6–PP2Ac degradosome. a, A composite cryo-EM map of the FBXO42–SKP1–CCDC6–mPP2Ac complex at an overall resolution of 3.2 Å. CCDC6 is shown in red, FBXO42 in slate, SKP1 in wheat and PP2Ac in grey. b, The structure of the FBXO42–SKP1–CCDC6–mPP2Ac complex shown from the same angle as in a (left) and rotated 90° to look down CCDC6 (right). The colouring of the subunits is identical to a. Unmodelled loops are represented as dashed lines. c, Schematic of the HTRF competition assay. The TR-FRET ratio was generated owing to the interaction between labelled Flag–HA–mPP2Ac (grey) and biotinylated FBXO42–SKP1 (slate and wheat) in the presence of untagged CCDC6 (red). Unlabelled FBXO42–SKP1 (faded slate and wheat) was titrated into the reaction inhibiting the TR-FRET ratio. The diagram was created using BioRender; Maculins, T. https://BioRender.com/zk6pu21 (2025). d, The nonlinear fit of the HTRF competition assay results performed at different mPP2Ac:CCDC6 ratios. The HTRF signal was normalized to the total FBXO42 (normalized HTRF signal = (observed HTRF signal/[FBXO42labelled]) × [FBXO42total]). n = 3 biologically independent experiments. Data are means ± s.d. Calculated KD values are reported in Supplementary Table 5. e, Cryo-EM map of 3×FBXO42–SKP1 bound to CCDC6–mPP2Ac at an overall resolution of 6.9 Å (left). Subunits are coloured as in a. Right, the structure of 3×FBXO42–SKP1-bound CCDC6–mPP2Ac rotated 90° relative to left panel with modelled CUL1–RBX1 (green and orange; Protein Data Bank (PDB): 1LDK). f, Superposition of a PP2a+c (PDB: 3FGA) heterodimer into the FCP complex. PP2Aa is shown in cyan as cartoon cylinders, and the FCP complex is shown as a transparent surface and coloured as in a. g, In vitro ubiquitination assay using recombinantly purified Flag–FBXO42–SKP1, CCDC6, PP2Ac and PP2Aa+c. Ubiquitination was detected using an antibody against PP2Ac. Source data Notably, we found that CCDC6 forms a homodimeric parallel coiled-coil that runs down the centre of the complex with four pairs of PP2Ac bound to the sides of the coil (labelled motifs 1–4) (Fig. 3a,b). Two copies of FBXO42–SKP1 were observed to flank the PP2Ac motif 3 pair, with each copy of the E3 making contacts with both phosphatase protomers in the motif simultaneously. Comparison of the FCP complex structure with the CP subcomplex shows that binding of FBXO42 does not change the conformation of the CP complex, suggesting that CCDC6 assists in pre-arranging PP2Ac into a state that can be read by the E3 ligase (Extended Data Fig. 6g). To estimate the affinity of FBXO42 to CCDC6–mPP2Ac, we designed a competition HTRF experiment in which the assay signal is generated between fluorophore-labelled FBXO42 and mPP2Ac (Fig. 3c). In the absence of CCDC6, mPP2Ac shows low assay signal and weak binding to FBXO42. By contrast, the competition experiment in the presence of CCDC6 shows around 100 nM binding affinity between FBXO42 and mPP2Ac (Fig. 3d and Supplementary Table 5). Notably, altering the ratio between CCDC6 and mPP2Ac does not appear to change the affinity of FBXO42 to the phosphatase, suggesting no cooperative binding (Fig. 3d). Although PP2Ac is repeated multiple times throughout the structure, we observed no contacts between each pair of phosphatases per motif or across motif interactions (~15 Å spacing for motif 1–3) (Fig. 3b (left)). In addition to the spacing, each PP2Ac pair in motifs 1–3 is rotated approximately 20° to each other, while the distal motif 4 is rotated nearly 70° relative to its motif 3 neighbour (Fig. 3b (right)). From another dataset, we were able to determine a 6.9 Å structure of three copies of FBXO42–SKP1 bound to CCDC6–PP2Ac across motifs 2 and 3 (Fig. 3e (left) and Extended Data Fig. 7). Modelling in CUL1, we find that two CUL1 molecules can exist side by side, with minor clashes at the C terminus (Fig. 3e (right)). We suspect that the flexible CUL1 C-terminal domain can adjust positions to avoid steric clashes and facilitate ubiquitination. PP2Ac is generally thought to be bound constitutively to the PP2Aa scaffolding subunit before loading of a variable B subunit to define the PP2A holoenzyme3. To understand whether FBXO42 and CCDC6 could target PP2Ac in the context of a PP2Aa+c complex, we modelled a PP2Aa+c heterodimer into the structure. We observed severe clashes between PP2Aa, FBXO42 and a neighbouring copy of PP2Ac (Fig. 3f), suggesting that FBXO42 and CCDC6 target the catalytic subunit alone for ubiquitination. In an in vitro ubiquitination assay, we found that the PP2Aa+c complex is a poor substrate for FBXO42 compared with PP2Ac, probably owing to the competitive interactions between CCDC6–PP2Ac and PP2Aa+c (Fig. 3g). Overall, the structure points to a mechanism for recruitment of numerous PP2Ac monomers to a central CCDC6 scaffold for efficient ubiquitination and degradation by FBXO42, forming a large multi-SCF degradosome for PP2A catalytic subunits. FBXO42 recognition of PP2Ac FBXO42 contains an approximately 50-amino acid helical F-box domain that interacts with the SKP1 protein, while the substrate-recognition domain of FBXO42 forms a classical six-bladed Kelch β-propeller, where each blade comprises four anti-parallel β-strands (labelled A–D from the innermost strand)23 (Extended Data Fig. 8a,b). FBXO42 sandwiches a (CCDC6–mPP2Ac)2 heterotetramer from both sides (Fig. 4a), whereby one PP2Ac copy binds to the backside of the FBXO42 Kelch domain to function as an adapter, and a second copy inserts its tail into the hole of the β-propeller and is oriented towards RBX1 for ubiquitination (substrate) (Extended Data Fig. 8c). Owing to the symmetry of the assembly, the adapter–substrate pairs are reversed for the mirrored FBXO42 (Fig. 4a (faded copy)), allowing simultaneous ubiquitination of both phosphatases. An extended interface is formed between the backside of the Kelch domain with the PP2Ac adapter copy, centred on the 1BC and 1D–2A loops of FBXO42, and helix α10 of the phosphatase (Fig. 4b and Extended Data Fig. 8b). Moreover, FBXO42 inserts the 2BC loop into a crevice formed at the junction between CCDC6 and both PP2Ac monomers (Fig. 4c and Extended Data Fig. 8b), specifically interacting with the β8–β9 hairpin of the adapter PP2Ac and the α5 helix of the substrate copy. Notably, all of these loops are among the most well conserved elements of FBXO42 (Extended Data Fig. 8a).Fig. 4FBXO42 engages two PP2Ac monomers simultaneously.a, Overview of the (FBXO42–CCDC6–mPP2Ac)2 assembly, looking down CCDC6 near motif 3 (red cylinders). FBXO42 and SKP1 are coloured in slate and wheat, respectively, and PP2Ac is coloured in grey. The PP2Ac C-terminal YFLme peptide in the central hole of FBXO42 is shown as sticks. The symmetry copy of FBXO42–SKP1 is shown as a transparent cartoon. Unmodelled loops are represented by dashed lines. b, Magnified view of the FBXO42 backside (cartoon and surface) and PP2Ac adapter (grey) interface. Key residues are shown as sticks. c, Magnified view of the FBXO42 2BC loop and CCDC6–mPP2Ac interface. Key residues are shown as sticks. d, IP of Flag–FBXO42 loop deletions and WB of endogenous PP2Ac from HEK293T nuclear lysates in the presence of MLN-4924. e, In vitro ubiquitination assay using recombinant FBXO42 loop-deletion proteins. Ubiquitination was detected by blotting against the HA-tag of PP2Ac. f, Top view of the FBXO42 Kelch domain. Density for the PP2Ac YFLme peptide is shown as a red mesh, and the peptide is shown as grey sticks. g, Magnified view of the interface between PP2Ac YFLme tripeptide (grey sticks) and FBXO42 (slate). Density for the peptide is shown as a red mesh. h, IP analysis of V5-tagged WT or ΔC-ter PP2Ac with endogenous FBXO42 from HEK293T nuclear lysates in the presence of MLN-4924. Blots were probed with an antibody against FBXO42. i, In cell ubiquitination of ectopically expressed V5-PP2Ac WT and ΔC-ter with Flag–FBXO42 in nuclear lysates of HEK293T cells. WB was performed to detect V5-tagged PP2Ac and Flag-tagged FBXO42. j, In vitro ubiquitination assay using recombinantly purified His–PP2Ac with or without the C-terminal tail. Ubiquitination was detected by blotting against PP2Ac. k, The cell confluency at 7 days in HOP62 cells expressing Flag–FBXO42 WT or Kelch mutants W192A and G284F after endogenous FBXO42 KD (5′-UTR-targeting siRNA). n = 3 biologically independent experiments. Each bar represents the mean value. Statistical analysis was performed using one-way analysis of variance.Source data FBXO42 engages two PP2Ac monomers simultaneously. a, Overview of the (FBXO42–CCDC6–mPP2Ac)2 assembly, looking down CCDC6 near motif 3 (red cylinders). FBXO42 and SKP1 are coloured in slate and wheat, respectively, and PP2Ac is coloured in grey. The PP2Ac C-terminal YFLme peptide in the central hole of FBXO42 is shown as sticks. The symmetry copy of FBXO42–SKP1 is shown as a transparent cartoon. Unmodelled loops are represented by dashed lines. b, Magnified view of the FBXO42 backside (cartoon and surface) and PP2Ac adapter (grey) interface. Key residues are shown as sticks. c, Magnified view of the FBXO42 2BC loop and CCDC6–mPP2Ac interface. Key residues are shown as sticks. d, IP of Flag–FBXO42 loop deletions and WB of endogenous PP2Ac from HEK293T nuclear lysates in the presence of MLN-4924. e, In vitro ubiquitination assay using recombinant FBXO42 loop-deletion proteins. Ubiquitination was detected by blotting against the HA-tag of PP2Ac. f, Top view of the FBXO42 Kelch domain. Density for the PP2Ac YFLme peptide is shown as a red mesh, and the peptide is shown as grey sticks. g, Magnified view of the interface between PP2Ac YFLme tripeptide (grey sticks) and FBXO42 (slate). Density for the peptide is shown as a red mesh. h, IP analysis of V5-tagged WT or ΔC-ter PP2Ac with endogenous FBXO42 from HEK293T nuclear lysates in the presence of MLN-4924. Blots were probed with an antibody against FBXO42. i, In cell ubiquitination of ectopically expressed V5-PP2Ac WT and ΔC-ter with Flag–FBXO42 in nuclear lysates of HEK293T cells. WB was performed to detect V5-tagged PP2Ac and Flag-tagged FBXO42. j, In vitro ubiquitination assay using recombinantly purified His–PP2Ac with or without the C-terminal tail. Ubiquitination was detected by blotting against PP2Ac. k, The cell confluency at 7 days in HOP62 cells expressing Flag–FBXO42 WT or Kelch mutants W192A and G284F after endogenous FBXO42 KD (5′-UTR-targeting siRNA). n = 3 biologically independent experiments. Each bar represents the mean value. Statistical analysis was performed using one-way analysis of variance. Source data To understand whether the FBXO42 Kelch backside and 2BC loop interfaces are essential for binding and ubiquitination, we performed co-IP experiments with Flag-tagged loop deletions to pull-down endogenous PP2Ac. Deletion of the FBXO42 1BC loop on the backside ablated over 90% of the interaction, while loss of either the 2BC loop or a combination of deletions completely eliminated binding in cells (Fig. 4d). In an in vitro ubiquitination reaction, we observed that the FBXO42(Δ1BC) mutant can partially ubiquitinate PP2Ac, while the FBXO42(Δ2BC) and deletion combination fully abrogated activity (Fig. 4e). Together, these results indicate that the unique quaternary structure recognition mode of FBXO42 and the (CCDC6–PP2Ac)2 heterodimer is key for both in-cell target engagement and efficient enzymatic activity. FBXO42 reads the PP2Ac carboxy-methyl tail In addition to the loops mediating PP2Ac recognition, we found density corresponding to the terminal YFLme tripeptide from the substrate copy of PP2Ac inserted into the central hole of the Kelch domain (Fig. 4f). Close examination reveals a claw-like conformation for the three PP2Ac residues (denoted by PP) in the pocket, with Tyr307PP and Phe308PP forming tight hydrophobic interactions with FBXO42 (denoted by F42) Trp192F42 and Leu678F42. The methylated Leu309PP (mLeu309) side chain is clamped between Pro305F42 and Trp340F42. Notably, the C-terminal carboxy-methyl is buried against a greasy glycine-rich cavity in the Kelch hole (Fig. 4g). Given our biochemical HTRF data demonstrating the boost in affinity between FBXO42 and PP2Ac after LCMT-1-mediated methylation (Fig. 2f), we examined whether the flexible C-terminal tail of PP2Ac was essential for recruitment to FBXO42. We generated V5-tagged PP2Ac with the C-terminal 14 amino acids removed, and transfected it into HEK293T cells to assess interaction with endogenous FBXO42. Despite only having a short motif inserted into the hole of the Kelch domain, the tailless PP2Ac was unable to IP FBXO42 from cell lysates (Fig. 4h). Cellular ubiquitination experiments also showed a substantial loss of PP2Ac ubiquitination with the tail deletion when overexpressed with Flag–FBXO42 (Fig. 4i). We also observed a total loss of ubiquitination when we used a tailless PP2Ac protein in an in vitro ubiquitination assay (Fig. 4j). Notably, a 14-amino acid carboxy-methylated peptide displayed weak (around 25 µM) affinity to FBXO42 in a surface plasmon resonance (SPR) experiment (Extended Data Fig. 8d and Supplementary Table 5). To test the importance of PP2Ac tail binding to the FBXO42 Kelch domain, we mutated the critical residues Trp192 and Gly284, which directly make contact with the PP2Ac tail, and overexpressed these mutants to assess their ability to compensate for the phenotype induced by siFBXO42. Consistent with our structural data, both mutants failed to rescue the phenotype caused by FBXO42 KD (Fig. 4k and Extended Data Fig. 8e). Together, our analysis indicates that FBXO42 uses its surface loops to recognize the quaternary assembly of a (CCDC6–PP2Ac)2 complex, and its central pocket to read and interpret the methylation state of the PP2Ac substrate. CCDC6 binds to PP2Ac through repeated motifs CCDC6 forms a coiled-coil that runs down the centre of the assembly and scaffolds multiple PP2Ac monomers along its length (Fig. 5a). Detailed examination of each bound copy shows a highly similar binding mode across all four motifs (Fig. 5b–e and Extended Data Fig. 9a (left)), where a set of CCDC6 acidic residues are engaging the active site of PP2Ac. Surface electrostatics of PP2Ac shows a highly basic cleft at the active site where two glutamates and an asparagine from CCDC6 are in direct contact with the deepest parts of the phosphatase (Fig. 5f). A sequence alignment of each motif in CCDC6 reveals a highly conserved stretch of residues defined by a EXXXEXXXN motif, whereby the strictly conserved Glu/Asn residues make specific interactions with PP2Ac, and flanking residues bolster the interface (Fig. 5f and Extended Data Fig. 9a (right) and 9b). This binding mode is highly reminiscent of several recently described structures of PP2A inhibitor proteins24,25, suggesting that CCDC6 also performs an inhibitory function, albeit through the degradation of the phosphatase (Extended Data Fig. 9c). Although we observed only four bound motifs in our structures, sequence analysis shows a fifth motif that is spaced C-terminally much further in the CCDC6 polypeptide (Extended Data Fig. 9a (right) and 9b) that is probably too flexibly tethered to the core of the complex for us to resolve using cryo-EM. In agreement with the structure, using HTRF, we found that CCDC6 robustly interacted with PP2Ac even in the absence of FBXO42 (Extended Data Fig. 9d).Fig. 5CCDC6 nucleates the complex assembly using conserved motifs.a, Overview of the interactions between a CCDC6 dimer (red) and multiple PP2Ac monomers (grey) shown as a cartoon. Each repeating motif with its PP2Ac partner is boxed and highlighted in b–e. b–e, Magnified detailed interactions between CCDC6 motifs 1 (b), 2 (c), 3 (d) and 4 (e) (red) and PP2Ac (grey). The backbone is shown as a cartoon, the side chains are shown as sticks and the manganese ions in the PP2Ac active site are shown as purple spheres. f, Electrostatic complementarity between CCDC6 motif 3 (transparent cartoon, red) and PP2Ac (electrostatic surface). Key side chains in direct contact with PP2Ac are highlighted as green sticks. Electrostatics are coloured from blue (positively charged) to red (negatively charged). g, In vitro ubiquitination assay using recombinantly purified Flag–FBXO42–SKP1, His–PP2Ac and mutant CCDC6 proteins. Ubiquitination was detected by using an antibody against PP2Ac. h, HTRF experiment showing the disassembly of the PP2A–B56 holoenzyme by CCDC6. The TR-FRET ratio was generated due to the interaction between labelled Flag–HA–mPP2Ac and His–PP2Aa, in the presence of untagged B56. Unlabelled CCDC6 was titrated into the reaction inhibiting the TR-FRET ratio. HTRF signal is normalized to the ‘no CCDC6’ control. n = 3 biologically independent experiment. Data are mean ± s.d. i, Immunodepletion of knock-in (KI) HA-tagged CCDC6 in NCI-H1373 lysates followed by WB analysis of CCDC6 and PP2Ac; the signals were quantified and normalized to the loading control vinculin. j, WB of Flag, CCDC6, PP2Ac, PP4c and PP6c in Flag IPs from MLN4924-treated HEK293T-Flag-FBXO42 cells after siCCDC6 KD. Actin was used as the loading control for inputs.Source data CCDC6 nucleates the complex assembly using conserved motifs. a, Overview of the interactions between a CCDC6 dimer (red) and multiple PP2Ac monomers (grey) shown as a cartoon. Each repeating motif with its PP2Ac partner is boxed and highlighted in b–e. b–e, Magnified detailed interactions between CCDC6 motifs 1 (b), 2 (c), 3 (d) and 4 (e) (red) and PP2Ac (grey). The backbone is shown as a cartoon, the side chains are shown as sticks and the manganese ions in the PP2Ac active site are shown as purple spheres. f, Electrostatic complementarity between CCDC6 motif 3 (transparent cartoon, red) and PP2Ac (electrostatic surface). Key side chains in direct contact with PP2Ac are highlighted as green sticks. Electrostatics are coloured from blue (positively charged) to red (negatively charged). g, In vitro ubiquitination assay using recombinantly purified Flag–FBXO42–SKP1, His–PP2Ac and mutant CCDC6 proteins. Ubiquitination was detected by using an antibody against PP2Ac. h, HTRF experiment showing the disassembly of the PP2A–B56 holoenzyme by CCDC6. The TR-FRET ratio was generated due to the interaction between labelled Flag–HA–mPP2Ac and His–PP2Aa, in the presence of untagged B56. Unlabelled CCDC6 was titrated into the reaction inhibiting the TR-FRET ratio. HTRF signal is normalized to the ‘no CCDC6’ control. n = 3 biologically independent experiment. Data are mean ± s.d. i, Immunodepletion of knock-in (KI) HA-tagged CCDC6 in NCI-H1373 lysates followed by WB analysis of CCDC6 and PP2Ac; the signals were quantified and normalized to the loading control vinculin. j, WB of Flag, CCDC6, PP2Ac, PP4c and PP6c in Flag IPs from MLN4924-treated HEK293T-Flag-FBXO42 cells after siCCDC6 KD. Actin was used as the loading control for inputs. Source data To assess the importance of each motif in the ability of FBXO42 to recognize and ubiquitinate the CCDC6–PP2Ac multimer assembly, we made combination alanine mutants of each motif at positions 1, 4, 5 and 9 and tested their ability to stimulate PP2Ac ubiquitination. Relative to WT CCDC6, all five mutants disrupted ubiquitination of the phosphatase significantly (Fig. 5g). Although we did not observe FBXO42 bound at motif 4 or 5 in our structure (Fig. 3b (left) and Extended Data Fig. 5), our mutational data suggest that all five CCDC6 motifs are functional for supporting the ubiquitination of PP2Ac. Given that FBXO42 appeared to preferentially ubiquitinate PP2Ac outside the context of an assembled PP2Aa+c core enzyme (Fig. 3f,g), we examined whether CCDC6 could disrupt PP2A holoenzymes. First, we separately purified PP2Aa and B56, and established a binding assay for the complex by HTRF. When we titrated an increasing amount of CCDC6, we observed a loss of binding between PP2Ac and PP2Aa, indicating that CCDC6 can compete for catalytic subunits in the context of holoenzymes (Fig. 5h). Following up this observation, we wanted to investigate how much of PP2Ac is regulated by CCDC6 in cells. To estimate the fraction of PP2Ac bound by CCDC6, we immunodepleted CCDC6 using an HA-tagged CCDC6 homozygous knock-in model derived from NCI-H1373 cells (Fig. 5i). Immunodepletion was nearly complete and resulted in an approximately 15% reduction in total PP2Ac levels, indicating that around 15% of the cellular PP2Ac pool is bound to CCDC6 and available for regulation by SCFFBXO42 (Fig. 5i (right)). Our data collectively indicate that CCDC6 uses multiple PP2Ac-binding motifs to downregulate a select pool of phosphatases. FBXO42 ubiquitinates both PP2Ac and PP4c PP2A, PP4 and PP6 constitute a subfamily of closely related phosphatases that all assemble into heteromeric holoenzymes for substrate recruitment and dephosphorylation3,26–28. Similar to PP2Ac, both PP4c and PP6c catalytic subunits are C-terminally carboxy-methylated by LCMT122. Given the high sequence similarity between the enzymes (Extended Data Fig. 10a) and their regulation by carboxy-methylation, we examined whether PP4c and PP6c could similarly be recognized and ubiquitinated by FBXO42 in a CCDC6-dependent manner. In NEDDi-treated HEK293T cells ectopically expressing Flag–FBXO42, endogenous PP4c could be co-immunoprecipitated with Flag–FBXO42 similarly to PP2Ac, while PP6c did not appear to be significantly enriched over background. Furthermore, loss of CCDC6 by siRNA KD resulted in abrogation of both the PP2Ac and PP4c interactions (Fig. 5j). Finally, PP4c was stabilized similarly to PP2Ac after FBXO42 KD across two different FBXO42-sensitive cancer cell lines (Extended Data Fig. 10b). To rule out the possibility of technical challenges with antibody detection of PP6c, we purified both PP4c and PP6c, and assayed them in an HTRF experiment side by side with PP2Ac for their ability to bind to FBXO42 in a CCDC6-dependent manner. We chose to use unmethylated PP4c and PP6c in these biochemical experiments as we had also observed binding with unmethylated PP2Ac by HTRF (Fig. 2f). PP4c bound to FBXO42, and this interaction was dependent on CCDC6 concentration, similar to PP2Ac (Extended Data Fig. 10c–e). By contrast, PP6c displayed little to no interaction with FBXO42 under similar conditions (Extended Data Fig. 10c,f). Using a cell-based NanoBiT assay, we were able to detect interaction between a LgBiT–PP2Ac/PP4c and SmBiT–FBXO42; however, LgBiT–PP6c showed little interaction (Extended Data Fig. 10g). Lastly, in an in vitro ubiquitination assay, PP6c was a relatively poor ubiquitination substrate, in contrast to PP2Ac and PP4c (Extended Data Fig. 10h). To understand the structural basis for PP2Ac and PP4c selectivity by FBXO42, we generated AlphaFold229 models for both PP4c and PP6c and modelled both phosphatases into the complex. Examining the adapter–backside interface (Fig. 4a) formed by PP4c with FBXO42, we found that the overall surface is highly similar to that seen with PP2Ac (Extended Data Fig. 10i,j). Nearly all of the residues between PP2Ac and PP4c at the adapter interface are conserved (Extended Data Fig. 10a), suggesting that FBXO42 would be unable to distinguish between the two phosphatases for binding. Modelling with PP6c showed multiple points of collision between the phosphatase and FBXO42 (Extended Data Fig. 10k). The residues in contact with FBXO42 are far less conserved between PP2Ac and PP6c (Extended Data Fig. 10a), suggesting that the divergence at the sequence level contributes to the selectivity seen for binding and ubiquitination. Together, our biochemical data and structural comparison suggest a mechanism for selective regulation of the PP2A and PP4 phosphatases by FBXO42. Discussion PP2A, through its diverse holoenzymes, profoundly impacts all aspects of cell biology, necessitating rigorous regulation. This regulation occurs on several levels, including strictly controlling holoenzyme assembly30,31 and chaperoning unpartnered catalytic subunits to prevent aberrant dephosphorylation32–34. Here we show a key role for the ubiquitin-proteasome system in regulating PP2Ac through the SCFFBXO42 complex. While human FBXO42 regulates a broad set of substrates11,14–17, and previous work in Drosophila implicated regulation of a specific PP2A complex through a different mechanism18, our research demonstrates that FBXO42 collaborates with CCDC6 to target the catalytic subunit of PP2Ac for ubiquitination and degradation. Perturbation of FBXO42 results in a mitotic defect and subsequent cell growth inhibition. This observation aligns with recent literature implicating FBXO42 in successful mitosis and revealing sensitivity to mitosis inhibitors after FBXO42 loss35. Our work reveals that the stabilization of PP2Ac resulting from the loss of FBXO42 function drives the observed phenotype. PP2A is a master regulator of the cell cycle, counteracting key mitotic kinases4,36. We hypothesize that FBOX42 loss and subsequent PP2Ac stabilization cause increased and abnormal dephosphorylation of its substrates, including proteins critical for cell mitosis. The selective sensitivity of FBXO42 loss in different cellular models is potentially driven by differential expression levels of the dysregulated PP2Ac substrate proteins, although this hypothesis warrants future investigation. Earlier research has shown that methylated PP2Ac is degraded in a proteasome-dependent manner. Loss of the demethylase PME-1 substantially reduces the PP2Ac half-life while increasing the overall ubiquitination of the phosphatase37. Our structural and biochemical results suggest that FBXO42 is the primary driver of this carboxy-methyl-dependent degradation mechanism. Based on our PP2Ac degradosome structure, we suspect that FBXO42 targets a mature pool of PP2Ac that has served its purpose and is ready to be decommissioned in preparation for entry into mitosis. In contrast to α4 and TIPRL, which target a demethylated form of PP2Ac representing states of disassembly or latency33,38, FBXO42 specifically recognizes mLeu309 and binds to a PP2Ac conformation that reflects that of an active enzyme. Our structural and biochemical analysis implies that FBXO42 and CCDC6 function to specifically target a pool of PP2Ac that has been disassembled from its holoenzyme (Figs. 3f,g and 5h), and has not yet undergone demethylation by PME-1. Although we observe an overall stabilization of the nuclear PP2Ac protein pool after FBXO42 depletion (Extended Data Fig. 3e), loss of the E3 ligase is not broadly toxic in cells, in contrast to the loss of α4, which is lethal39. This suggests that FBXO42 and CCDC6 probably regulate a very specific subset of PP2Ac subunits, in contrast to α4, which governs PP2Ac shuffling between disassembled and assembled states. Future work will be needed to understand the mechanisms governing which PP2A holoenzymes are targeted for disassembly and turnover. In addition to preferentially binding carboxy-methylated PP2Ac, FBXO42 recognizes a cryptic degron that is found in the quaternary structure between the complex of CCDC6–PP2Ac. In contrast to several other multicomponent E3 recognition mechanisms in which there is affinity and interfaces between all components of the assembly40–44, CCDC6 has no interface/affinity for FBXO42 and functions only to rearrange PP2Ac into a structure that can be recognized by the E3. To our knowledge, this is the first example of a non-substrate protein binding to a substrate protein, and rearranging it to allow for E3 interaction. We propose to term this mode of E3-substrate interaction template-driven protein ubiquitination, in which the starting materials (substrate, E3) are unable to form stable complexes until the substrate protein encounters a non-substrate template protein. The template protein remodels the substrate into a form that can then be recognized by an E3 to form a productive complex. This concept is distinct from an adapter model in which E3 and the adapter form a composite interface to recruit substrates, and distinct from the binding of obligate protein complexes45,46. Our model of template-driven interaction adds an additional layer of regulation whereby expression and protein levels of the template protein will influence the stability of the substrate. Future work will be needed to understand how CCDC6 levels are controlled spatially and temporally to regulate PP2Ac levels in cells. Our results not only highlight a unique mode of E3-substrate recognition, but also raise the possibility that other substrate interactions are governed through a template-driven mechanism. Methods Cancer dependency map analysis Genome-wide gene dependency scores were obtained from the Cancer Dependency Map (DepMap) 2024 Q2 CRISPRGeneEffect.csv public data release47. Cell line generation, exogenous expression, siRNA and cell viability assays Cell lines were grown in RPMI 1640 medium supplemented with 10% FCS with the exception of HEK293T cells, which were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% FCS; CALU1 cells, which were grown in McCoy’s 5A, 10% FBS, 2 mM l-glutamine; and RERF-LC-AI cells, which were grown in MEM, 10% FBS, 2 mM l-glutamine. Cell line identity was verified by high-throughput single-nucleotide polymorphism (SNP) genotyping using Illumina Golden Gate multiplexed assays. SNP profiles were compared to SNP calls from internal and external databases to determine or confirm ancestry. All cell lines tested negative for mycoplasma contamination before storage/use at our institute. siRNA KD experiments were performed using ON-TARGET-plus SMARTpools (siFBXO42 L-022191-00-0010, siCCDC6 L-010551-00-0010) or single SiGenome (siFBXO42 D-022191-02-0005) purchased from Horizon Discovery and were reverse transfected using Lipofectamine RNAiMax (Invitrogen) for 72 h according to the manufacturer’s instructions. All constructs were generated through gene synthesis at Genscript with codon optimization for mammalian cell expression. Flag, HA or His tags were cloned at the N terminus of FBXO42 or CCDC6. FBXO42 deletion mutant amino acid boundaries were as follows: FBXO42(ΔF-box) (100–717), FBXO42(ΔKelch) (1–100), FBXO42(Δ1BC) (140–146), FBXO42(Δ)2BC (192–202) and FBXO42(ΔC-Ter) (2–295). Generation of stable cell lines for expression of Flag–FBXO42 WT or mutants, sgRNAs, DOX-inducible CRISPR–cas9 or Flag–FBXO42 was carried out by the piggybac transposon system. Transfections were performed using Lipofectamine 3000 (Thermo Fisher Scientific) and cells selected for resistance to blasticidin (DOX-inducible expression) or puromycin (constitutive expression). To assess clonogenic capacity, cells were seeded in 6- or 12-well plates at low confluency for 7 to 15 days. Cells were then fixed by 4% PFA and stained using 0.005% Crystal Violet. Live-cell imaging was performed in a 96-well-plate format (×10 magnification) using the Incucyte Live-Cell Analysis System (Sartorius). Cell lysates, IP, immunodepletion and WB Whole-cell extracts were prepared in Pierce IP lysis buffer (87788) including protease inhibitor cocktail. Nuclear fractions were prepared using the Nuclear Extract Kit (Actif Motif 40010) according to the manufacturer’s instructions. For IPs, cells were pretreated with MG132 (10 µM, Cell Signaling, 2194S) or MLN4924 (5 µM, Cell Signaling, 85923S) overnight. IPs of exogenous tagged-proteins were achieved using 1–5 mg of cell lysates incubated overnight at 4 °C in 1× IP High Buffer (Active Motif, 37510) with 50 µl of Magnetic Agarose beads corresponding to the respective bait tag (Thermo Fisher Scientific, Flag, A36797; HA, 88836). IPs of recombinant proteins were done using 1 μg incubated for 4 h at 4 °C in 1× IP High Buffer (Active Motif, 37510) with 50 µl of magnetic agarose beads corresponding to baits tag. Beads were washed four times with the IP buffer and incubated overnight at 4 °C with 1–5 mg of cell lysates. The next day, IPs were washed four times in IP buffer, resuspended in a loading denaturing buffer and boiled for 15 min before WB analysis. For immunodepletion experiments, 2 mg of lysates from HA-CCDC6 knock-in NCI-H1373 cells were incubated overnight in Active Motif High Buffer (37510) with either anti-HA or control IgG magnetic beads. The beads were then removed, replaced with fresh beads and incubation was continued for an additional 4 h. After bead removal, lysates were analysed by WB. For routine WB, the samples were mixed in a denaturing buffer, boiled for 5 min and loaded to Novex Tris-Glycine protein gels at variable percentages. Gels were transferred onto an iBlot 2 Transfer Stacks nitrocellulose membrane by wet transfer (Thermo Fisher Scientific). Membranes were incubated with the following primary antibodies diluted in 5% whole milk and 0.01% Tween overnight at 4 °C: FBXO42 (Genentech, 96), CCDC6 (Sigma-Aldrich, HPA019051), PP2Ac (Cell Signaling, 2038), PP4c (Fortis, A300-835A), PP6c (Cell Signaling, 64221), HA (Cell Signaling, 3724), Flag (Sigma-Aldrich, M8823), HIS (Cell Signaling, 12698), UB (Cell Signaling, 43124), actin (Cell Signaling, 4970), HDAC1 (Cell Signaling, 2062), α-tubulin (Abcam, ab52866), vinculin (Cell Signaling, 4650) and RBPJ (Cell Signaling, 5442). Membranes were incubated with 1/5,000 IRDye 800CW donkey anti-rabbit or anti-mouse for 1 h at room temperature and read by LI-COR ODYSSEY DLx. The signal was quantified using ImageStudioLite. IF analysis Cells were plated in 96-well plates and treated with the corresponding siRNAs for 48 h. Cells were fixed with PFA 4% for 10 min and incubated in a blocking and permeabilization buffer (10% FBS, 1% BSA, 0.1% Triton X-100 and 0.01% NaN3) for 30 min at room temperature. Cells were incubated with the following primary antibodies diluted in the same buffer overnight at 4 °C: anti-phospho-histone H3 (S10) (Cell Signaling, 3377), CENPB (Abcam ab25734) and α-tubulin (Proteintech, 66031-1-Ig). Cells were washed with PBS and incubated with 1/2,000 of ALEXA 488 or 647 secondary antibody for 1 h at room temperature. Cells were counterstained with 1/5,000 HOECHST 33342 solution. Images were taken either using a ×20 objective and analysed using the OPERA PHENIX confocal screening system or a ×60 objective on the Nikon Crest spinning-disc confocal microscope. In vitro ubiquitination assay Reactions (total, 40 µl) were assembled in assay buffer (1×, 30 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM MgCl2, 5% glycerol) containing 2 mM ATP and 2 mM DTT. Final component concentrations were: E1 (UBA1), 0.1 µM; E2 (CDC34), 2 µM; ubiquitin, 20 µM; CUL1–RBX1, 0.15 µM; SKP1–FBXO42, 0.15 µM; CCDC6, 0.15 µM; and the substrate, 0.15 µM. After a 10 min incubation at room temperature, the reactions were incubated at 30 °C for 5–30 min, quenched in SDS sample buffer and analysed using immunoblotting. Phosphatase assay Phosphatase activity was measured using the EnzChek Phosphatase Assay Kit according to the manufacturer’s instructions. Then, 5 µg of cell lysates was incubated with the substrate in the assay buffer containing either DMSO or 10 nM of okadaic acid (Cell Signaling, 5934) for 30 min before measurement. CRISPR–Cas9 survival screen HOP-62 cells were engineered by transfection of a piggyBac transposon vector to express S. pyogenes Cas9 driven by the Tet-On 3 G DOX-inducible system and FBXO42 sgRNA from the U6 promoter. A single-cell clone was isolated that demonstrated DOX-inducible growth arrest, as well as reduced CD81 protein expression after transduction with lentiviral CD81 sgRNA. An sgRNA library targeting all protein-coding genes in the human genome at a complexity of 8 sgRNAs per gene was designed and produced in 8 plasmid pools, as described previously48, for an analogous mouse sgRNA library. The sgRNA sequences are provided in Supplementary Table 1. Library sgRNA plasmid pools were packaged into lentivirus and titred in the screening cell line as described previously48. For the screen, lentiviral stocks of the 8 sublibrary pools were combined in proportion to their viral titres. HOP-62 DOX–Cas9 sgFBXO42 cells were transduced in independent triplicate samples of 9.88 × 107 cells each with the lentiviral whole-genome sgRNA library at a multiplicity of infection of 0.46–0.59, as measured by puromycin resistance, to achieve a representation of 242–292 cells per sgRNA. All cell culture and passaging was performed on the CompacT SelecT (Sartorius Stedim Biotech) automation system in barcoded, triple-layer T500 flasks (132925, Thermo Fisher Scientific). Then, 2 days after transduction (day 2), 9.00 × 107cells were collected for a reference timepoint from each replicate, and 1.07 × 108 cells per replicate were passaged for selection with 2 μg ml−1 puromycin starting on day 3. Next, 5.20 × 107 cells per replicate were passaged without puromycin on days 6 and 9. For each replicate on day 13, 9.00 × 107 cells were collected for a baseline timepoint, and 4.00 × 107 cells of each replicate were plated and treated with 1 μg ml−1 DOX (treatment day 0). At this time, cells from each replicate were also seeded at 250 cells per well into 96-well cell culture plates and treated with and without 1 μg ml−1 DOX, and a mean of 21.5-fold reduction in cell number after 10 days of DOX treatment was measured using CellTiter-Glo (Promega) according to the manufacturer’s instructions. Medium with fresh 1 μg ml−1 DOX was replenished in all flasks on treatment day 2, 4.00 × 107 cells per replicate were passaged into DOX medium on treatment day 4 and DOX medium was replenished again on treatment day 7. On treatment day 10, 3.80 × 107 cells from each replicate were collected. For all cell collections, genomic DNA isolation, PCR amplification of the sgRNA sequences, amplicon purification and next-generation paired-end sequencing were performed as described previously48. Sufficient genomic DNA was used for PCR amplification to maintain a representation of 250 cells per sgRNA for all baseline and DOX-treated cell samples. For each read pair, both read sequences were searched for an exact match of any one of the sgRNA barcode sequences, using the ScreenCounter R package available from the CRISPRVerse49. The number of reads matched to each guide in each sample was used to obtain a guide-by-sample count matrix for further analysis. The raw counts data were stored in a standard Bioconductor SummarizedExperiment object50. Normalization is needed to adjust for the difference in sequencing depth between samples and to account for potential compositional biases due to the competitive nature of pooled genetic screening. We estimated normalization factors for each sample by applying the TMM method51 on the count data for sgRNAs targeting a gold-standard set of non-essential genes52. The sequence reads across all samples achieved a mean sgRNA count of 350–459, with 0.04–0.12% of sgRNAs showing complete depletion (Supplementary Table 1). We performed a differential abundance analysis at the sgRNA level using the popular limma-voom approach53. Specifically, we fitted a linear model to the log-transformed counts per million values for each sgRNA, using voom-derived observation and quality weights. We performed robust empirical Bayes shrinkage to obtain shrunken variance estimates for each sgRNA, and we used moderated F-tests to compute P values for each of the two-group comparisons of interest. To control the FDR in each comparison, we applied the standard Benjamini–Hochberg method to obtain an adjusted P value for each sgRNA (Supplementary Table 1). We used sgRNA-level statistics to obtain gene-level summaries using two complementary statistical approaches. The first approach was to aggregate sgRNA statistics using the minimum Holm approach54, implemented in the metapod R package55, to obtain a combined P value for each gene based on the P values for all associated sgRNAs (Supplementary Table 2). Gene-level P values were corrected for multiple comparisons using the Benjamini–Hochberg method. In the second approach, DrugZ (v.1.1.0.2)56 was run using the default parameters on the raw sgRNA read counts for each comparison of interest across the three available replicates. DrugZ computes z scores of each sgRNA for the fold change between the replicate-matched baseline and DOX-treated samples. z scores are then aggregated at the gene level into a normalized z score56 (normZ; Supplementary Table 2). Both the Holm-min method P values and normZ scores identify PPP2CA as the top screen hit for sgRNA enrichment, with considerably greater statistical significance than any other gene. LCMT1 is the fourth most significant enrichment hit by both methods, with only 3 genes other than PPP2CA with slightly greater significance scores, each mapping near PPP2CA on chromosome 5. Proteomics IP–MS Protein lysate (3 mg) in denaturing buffer (8 M urea, 20 mM HEPES, pH 8.0) was reduced (5 mM dithiothreitol (DTT), 45 min at 37 °C), alkylated (15 mM iodoacetamide (IAA), 20 min at room temperature in the dark) and quenched (5 mM DTT, 15 min at room temperature in the dark). Proteins were pelleted by chloroform–methanol precipitation. The resulting pellet was resuspended in denaturing buffer, diluted to 4 M urea and digested for 4 h at 37 °C with lysyl-endopeptidase (Wako) at an enzyme to protein ratio of 1:100. The sample was further diluted to 1.3 M urea and subjected to overnight enzymatic digestion at 37 °C with sequencing-grade trypsin (Promega; enzyme:protein ratio, 1:50). The resultant peptides were acidified with 20% trifluoroacetic acid (TFA, 1% final concentration), centrifuged at 18,000g for 15 min and desalted using the Sep-Pak C18 column (Waters). The eluted peptides from each treatment group were lyophilized and reserved for global proteome abundance. For global proteome samples, 100 μg of peptides from each sample was dissolved in 100 mM HEPES pH 8.0 (1 mg ml−1). Isobaric labelling was performed using TMTPro16-plex reagents (Thermo Fisher Scientific). Each unit (0.5 mg) of TMT reagent was allowed to reach room temperature immediately before use, spun down on a benchtop centrifuge and resuspended with occasional vortexing in 20 μl anhydrous acetonitrile (ACN) before mixing with peptides (18% final ACN concentration). After incubation at room temperature for 1 h, the reaction was quenched for 15 min with 20 μl of 5% hydroxylamine. Labelled peptides were combined in equimolar ratios and dried. The TMTpro-labelled sample was redissolved in 80 μl 0.1% TFA, centrifuged at 16,000g and the supernatant was processed further. Offline high-pH reversed-phase fractionation was performed on the 1100 HPLC system (Agilent) using an ammonium-formate-based buffer system. Peptides (400 μg) were loaded onto a 2.1 × 150 mm 3.5 μm 300 Extend-C18 Zorbax column (Agilent) and separated over a 75-min gradient from 5% to 85% ACN into 96 fractions (flowrate = 200 μl min−1). The fractions were concatenated into 24 fractions, mixing different parts of the gradient to produce samples that would be orthogonal to downstream low-pH reversed phase LC–MS/MS. The fractions were dried and desalted using C18 stage-tips as previously described57. Peptides were lyophilized and resuspended in 10 μl buffer A (2% ACN, 0.1% formic acid) for LC–MS/MS analysis. For global proteome peptides, LC–MS/MS analysis was performed by injecting 1 μl of each fraction on an Orbitrap Eclipse mass spectrometer (Thermo Fisher Scientific) coupled to the Dionex Ultimate 3000 RSLC (Thermo Fisher Scientific) system using the 25 cm IonOpticks Aurora Series column (IonOpticks) with a gradient of 2% to 30% buffer B (98% ACN, 2% H2O with 0.1% formic acid; flow rate, 300 nl min−1). Global proteome samples were analysed with a total run time of 95 min. For all samples, the Orbitrap Eclipse with FAIMS Pro DUO of −40, −60CV (proteome) collected FTMS1 scans at 120,000 resolution with an AGC target of 1 × 106 and a maximum injection time of 50 ms. FTMS2 scans on precursors with charge states of 3–6 were collected at 15,000 resolution with collision-induced dissociation (CID) fragmentation at a normalized collision energy of 30%, an AGC target of 2 × 104 and a maximum injection time of 100 ms. Real-time database search was performed before acquisition of MS3 spectra using ThermoRTS. The following parameters were used for global proteome analysis: UniProt human database August 2021 version, including 218,136 SwissProt sequences of canonical and protein isoforms, plus common contaminants and decoys; static modifications included Cys carbamidomethylation (+57.0215), Lys and N-term TMTPro (+304.207146); variable modifications included Met oxidation (+15.9949) and Tyr TMTPro (+304.207146). Offline search was performed using comet v.2019.01 with parameters matched to the RTS search. The peptide FDR was filtered to <1% using the linear discriminator algorithm. TMT reporter ions produced by the TMT tags were quantified with Mojave in-house software package by calculating the highest peak within 20 ppm of theoretical reporter mass windows and correcting for isotope purities. Quantification and statistical testing of global proteome proteomics data were performed by MSstatsTMT_2.0.1 R package58. Multiple fractions from the same TMT mixture were combined in MSstatsTMT (v.2.0.1)58. In particular, if the same peptide ion was identified in multiple fractions, only the single fraction with the highest maximal reporter ion intensity was retained. Global median normalization was performed to reduce the systematic bias between channels. KGG-MS: TMTPro sample preparation Protein lysate (700 μg) in denaturing buffer (8 M urea, 20 mM HEPES, pH 8.0) was reduced (5 mM DTT, 45 min at 37 °C), alkylated (15 mM IAA, 20 min at room temperature in the dark), and quenched (5 mM DTT, 15 min at room temperature in the dark). Proteins were pelleted by chloroform–methanol precipitation. The resulting pellet was resuspended in denaturing buffer, diluted to 4 M urea and digested for 4 h at 37 °C with lysyl-endopeptidase (Wako) at an enzyme to protein ratio of 1:100. The sample was further diluted to 1.3 M urea and subjected to overnight enzymatic digestion at 37 °C with sequencing grade trypsin (Promega; enzyme:protein ratio, 1:50). The resultant peptides were acidified with 20% TFA (1% final concentration), centrifuged at 18,000g for 15 min and desalted using a Sep-Pak C18 column (Waters). Around 200 μg of eluted peptides from each treatment group was lyophilized and reserved for global proteome abundance. The remaining eluted peptides were lyophilized and used for di-glycine (KGG) analysis. For global proteome samples, 100 μg of peptides from each sample was dissolved in 100 mM HEPES pH 8.0 (1 mg ml−1). Isobaric labelling was performed using TMTPro16-plex reagents (Thermo Fisher Scientific). Each unit (0.5 mg) of TMT reagent was allowed to reach room temperature immediately before use, spun down on a benchtop centrifuge and resuspended with occasional vortexing in 20 μl anhydrous ACN before mixing with peptides (18% final ACN concentration). After incubation at room temperature for 1 h, the reaction was quenched for 15 min with 20 μl of 5% hydroxylamine. Labelled peptides were combined in equimolar ratios and dried. The TMTpro-labelled sample was redissolved in 80 μl 0.1% TFA, centrifuged at 16,000g, and the supernatant was processed further. Offline high-pH reversed-phase fractionation was performed on the 1100 HPLC system (Agilent) using an ammonium-formate-based buffer system. Peptides (400 μg) were loaded onto a 2.1 × 150 mm 3.5 μm 300 Extend-C18 Zorbax column (Agilent) and separated over a 75-min gradient from 5% to 85% ACN into 96 fractions (flowrate, 200 μl min−1). The fractions were concatenated into 24 fractions, mixing different parts of the gradient to produce samples that would be orthogonal to downstream low-pH reversed-phase LC–MS/MS. The fractions were dried and desalted using C18 stage-tips as previously described59. Peptides were lyophilized and resuspended in 10 μl buffer A (2% ACN, 0.1% formic acid) for LC–MS/MS analysis. For ubiquitylome quantitation of KGG peptides, lyophilized peptides were reconstituted in 1× detergent containing IAP buffer (Cell Signaling Technology) for immunoaffinity enrichment. KGG peptide enrichment was performed at 4 °C on the MEA2 automated purification system (Phynexus) using 1 ml Phytips (Phynexus) packed with 20 μl ProPlus resin coupled to 200 μg of anti-KGG (Cell Signaling Technology) antibody. Phytip columns were equilibrated for 2 cycles (1 cycle, aspiration and dispensing, 0.9 ml, 0.5 ml min−1) with 1 ml 1× IAP buffer before contact with peptides. Phytip columns were incubated with peptides for 16 cycles of capture, followed by 6 cycles of washing, twice with 1 ml 1× IAP buffer and 4× with 1 ml water. Captured peptides were eluted with 60 μl 0.15% TFA in 8 cycles in which the volume aspirated/dispensed was adjusted to 60 μl. Enriched ubiquitinated peptides were prepared as previously described60. Labelled peptides were combined, dried and resolubilized in 0.15% TFA for high-pH reversed-phase fractionation using a commercially available kit (Thermo Fisher Scientific). Fractionation was performed according to the manufacturer’s protocol with a modified elution scheme where 11 fractions were collected (F1, 13.5% ACN; F2, 15% ACN; F3, 16.25% ACN; F4, 17.5 ACN; F5, 20% ACN; F6, 21.5% ACN; F7, 22.5% ACN; F8, 23.75% ACN; F9, 25% ACN; F10, 27.5% ACN; and F11, 30% ACN) and then combined into 6 fractions (F1 + F6, F2 + F7, F8, F3 + F9, F4 + F10, F5 + F11). Peptides were lyophilized and resuspended in 10 μl buffer A for LC–MS/MS analysis. For global proteome and ubiquitylome quantitation of KGG peptides, LC–MS/MS analysis was performed by injecting 1 μl (Proteome) or 2.5 μl (KGG) of each fraction on an Orbitrap Eclipse mass spectrometer (Thermo Fisher Scientific) coupled to the Dionex Ultimate 3000 RSLC (Thermo Fisher Scientific) system using a 25 cm IonOpticks Aurora Series column (IonOpticks) with a gradient of 2% to 30% buffer B (98% ACN, 2% H2O with 0.1% formic acid; flow rate, 300 nl min−1). Global proteome samples were analysed with a total run time of 95 min and ubiquitylome samples were analysed with a total run time of 180 min. For all samples, the Orbitrap Eclipse with FAIMS Pro DUO of −40, −60CV (proteome) or −40, −60 and −75CV (KGG) collected FTMS1 scans at a resolution of 120,000 with an AGC target of 1 × 106 and a maximum injection time of 50 ms. FTMS2 scans on precursors with charge states of 3–6 were collected at 15,000 resolution with CID fragmentation at a normalized collision energy of 30%, an AGC target of 2 × 104 (proteome) or 2 × 105 (KGG), and a maximum injection time of 100 ms (proteome) or 200 ms (KGG). RTS was performed before acquisition of MS3 spectra using InSeqAPI software61, which operated similarly to previously published approaches62,63. The following RTS parameters were used for global proteome analysis: UniProt human database August 2021 version, including 218,136 SwissProt sequences of canonical and protein isoforms, plus common contaminants and decoys; static modifications included Cys carbamidomethylation (+57.0215), Lys and N-term TMTPro (+304.207146); variable modifications included Met oxidation (+15.9949) and Tyr TMTPro (+304.207146). Offline search was performed using comet v.2019.01 with parameters matched to the RTS search. The peptide FDR was filtered to <1% using the linear discriminator algorithm. TMT reporter ions produced by the TMT tags were quantified using Mojave custom software package by calculating the highest peak within 20 ppm of theoretical reporter mass windows and correcting for isotope purities. Quantification and statistical testing of global proteome proteomics data were performed by MSstatsTMT_2.0.1 R package58. Multiple fractions from the same TMT mixture were combined in MSstatsTMT (v.2.0.1)58. In particular, if the same peptide ion was identified in multiple fractions, only the single fraction with the highest maximal reporter ion intensity was retained. Global median normalization was carried out to reduce the systematic bias between channels. RAW spectra files can be found on MassIVE with the dataset identifier: MassIVE MSV000094889. Molecular biology and protein expression All constructs were generated through gene synthesis at Genscript with codon optimization for insect cell expression (except for LCMT-1, which was expressed in Escherichia coli). DNA encoding the genes of interest was cloned into modified versions of the baculovirus transfer vector pAcGP67A or pAcSG2 (BD Biosciences), adapted for high-throughput cloning. The expression cassette is flanked by segments of the baculovirus genome to facilitate transfer into the linearized baculovirus DNA by homologous recombination in insect cells64. Baculovirus was generated using homologous recombination between BEVS transfer vector and BestBac 2.0 Linearized Bacmid (Expression Systems, 91-002) in a high-throughput manner as described earlier for generating P2 virus65. P3 stocks of baculovirus for large-scale production were amplified for 4 days in a flask at 200 ml followed by collection and used for infecting 8–10 l of Sf9 or Tni cells for protein production. Cells were seeded at a density of 2 × 106 cells per ml and infected at multiplicity of infection of 0.5 for Sf9 cells and multiplicity of infection of 1 for Tni cells. Sf9 infected cells were collected after 3 days and Tni cells were collected after 2 days. Pellets were frozen, followed by purification. Protein purification Flag–FBXO42–SKP1 Human FBXO42 (UniProt: Q6P3S6) with an N-terminal truncation (Δ2–44) and an N-terminal Flag tag, and Human SKP1 (UniProt: P63208) with deletions of residues 37–42 and 70–81 (70-81 were replaced with a GGSG linker), were separately cloned (pAcGP67A) and co-expressed in Sf9 cells using the protocol described above. The frozen cell pellet of expressed Flag–FBXO42–SKP1 was resuspended and thawed in CytoBuster Protein Extraction Reagent (EMD Millipore, 50 ml reagent per 1 l biomass), 1 mM TCEP pH 7.5, benzonase (1:1,000 dilution of 0.5 mg ml−1 benzonase stock:lysate), 5 mM MgCl2, 5 mM ATP pH 7.3 and cOmplete EDTA-free Protease Inhibitor Cocktail (Roche, 1 tablet per 50 ml of lysate). The thawed lysate was homogenized with a sonicator until the cell mixture was evenly distributed. The homogenized cell lysates were incubated at room temperature for 5 min and then incubated with anti-DYKDDDDK magnetic beads (GenScript, 0.8–1 ml settled beads per 1 l biomass) on a shaker (Stuart Orbital Shaker SSL1) at 170 rpm for 1 h at 4 °C. The lysates were separated from the magnetic beads using an AmMag Block (GenScript). Batch purification methods were used to purify Flag–FBXO42–SKP1. Anti-DYKDDDDK Magnetic Beads underwent two washes with 5× settled bead volume of wash buffer (50 mM Tris pH 7.5, 150 mM NaCl, 5 mM MgCl2, 5 mM ATP pH 7.3, 10% glycerol, 1 mM TCEP pH 7.5), followed by 3 elutions with 3× settled bead volume of elution buffer (50 mM Tris pH 7.5, 150 mM NaCl, 5 mM MgCl2, 10% glycerol, 1 mM TCEP pH 7.5, 0.3 mg ml−1 3×Flag peptide). The AmMag MR magnetic rack (GenScript) was used to bind beads, decant washes and decant eluates. For each wash, the mixture of buffer and beads was gently inverted 5–6 times before decanting. For each elution, the mixture of buffer and beads was incubated for 1 h at 4 °C before decanting eluates. Incubation of Anti-DYKDDDDK magnetic beads and batch purification methods were carried out twice on the same lysate. All washes and eluate fractions that contained Flag–FBXO42–SKP1 were combined, concentrated and passed over the HiLoad 16/600 Superdex 200 pg (Cytiva) column equilibrated in the following SEC buffer: 20 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol and 1 mM TCEP pH 7.5. The protein concentration was measured using the absorbance at 280 nm (NanoDrop). The final SEC fractions containing Flag–FBXO42–SKP1 were combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. SCFFBXO42 Human CUL1 (UniProt: Q13616) with an N-terminal His-tag and human RBX1 (UniProt: P62877) were cloned into a bicistronic pAcGP67A vector in which each target was downstream of a polyhedrin promoter. Human CUL1, human RBX1, human FBXO42 and human SKP1 were co-expressed in Sf9 cells using the protocol described above. Human FBXO42 and human SKP1 were the same constructs as those used in the Flag–FBXO42–SKP1 purification method above. The frozen cell pellet of expressed SCFFBXO42 was resuspended and thawed in lysis buffer (40 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol; 50 ml buffer per l of biomass), benzonase (1:1,000 dilution of 0.5 mg ml−1 benzonase stock:lysate), 5 mM MgCl2, 5 mM ATP pH 7.3 and cOmplete EDTA-free protease inhibitor cocktail (Roche, 1 tablet per 50 ml of lysate). The thawed cell lysates were passed through a microfluidizer three times. Whole-cell lysates were clarified by ultracentrifugation at 186,000g and filtered by a 0.8-µm filter followed by a 0.45-µm filter. The clarified lysates were gently rocked on a nutator for 1 h at 4 °C with anti-Flag M2 affinity gel resin (Sigma-Aldrich, 1.25 ml settled resin per 1 l of biomass). Resin was separated from lysate by centrifugation (700g). Resin was applied to a gravity column (Bio-Rad) and the sample was washed with 20 column volumes of lysis buffer. SCFFBXO42 was eluted with lysis buffer containing 0.3 mg ml−1 3× Flag peptide. Eluates were supplemented with 3 mM TCEP pH 7.5. Incubation of anti-Flag M2 affinity gel resin and gravity purification methods were performed twice on the same lysate. The eluate fractions that contained SCFFBXO42 were combined, concentrated and passed over the HiLoad 16/600 Superdex 200 pg (Cytiva) column in the following SEC buffer: 20 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol, 3 mM TCEP pH 7.5. The protein concentration was measured using the absorbance at 280 nm (NanoDrop). The final SEC fractions containing SCFFBXO42 were combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. HisAvi(biotin)–FBXO42–SKP1 Full-length human FBXO42 (UniProt: Q6P3S6), with N-terminal His and Avi tags, and SKP1 (the same construct used for Flag–FBXO42–SKP1 above) were separately cloned (pAcGP67A) and co-expressed in Sf9 cells using the protocol described above. The frozen cell pellet of HisAvi(biotin)–FBXO42–SKP1 was resuspended and thawed in 50 mM Tris pH 8.0, 150 mM NaCl, 1 mM TCEP and 0.05% Triton X-100 with protease inhibitor tablets. The cell lysates were homogenized and stirred on ice for 30 min, then centrifuged (186,000g) for 1 h. Next, 5 ml of NiNTA agarose resin was added to the supernatant, and rotated at 4 °C for 90 min. Resin was centrifuged (700g) for 5 min, then washed twice with 50 mM Tris pH 8.0, 150 mM NaCl and 1 mM TCEP. The resin was further washed with 25 ml of 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM TCEP and 7.5 mM imidazole, then 25 ml of 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM TCEP and 12.5 mM imidazole. Protein was eluted with 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM TCEP and 500 mM imidazole. The fractions were analysed using an SDS–PAGE gel. The fractions containing FBXO42–SKP1 were pooled together and diluted 3× in 25 mM Tris pH 8.0 and 1 mM TCEP. The sample was loaded onto the 5 ml Hitrap Q HP (Cytiva) column, then eluted with a 0–500 mM NaCl gradient over 20 column volumes (100 ml, 0–50% B). The fractions containing FBXO42–SKP1 were pooled together and concentrated to 80–100 μM, then diluted to 40–50 μM with 25 mM Tris pH 8.0 and 1 mM TCEP; the NaCl concentration was 150 mM. Avidity’s biotin reaction kit was used, and the sample was incubated at room temperature overnight. Details of the protocol are available online (https://www.avidity.com/commerce/product.asp?NUMBER=1). After biotinylation, the protein sample was concentrated and loaded onto the HiLoad 16/600 Superdex 200 pg (Cytiva) column pre-equilibrated in 25 mM Tris pH 8.0, 150 mM NaCl, 1 mM TCEP. The protein concentration was measured using the absorbance at 280 nm (NanoDrop). The final SEC fractions containing HisAvi(biotin)–FBXO42–SKP1 were combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. PP2Ac, PP4c and PP6c Human PP2Ac (UniProt: P67775-1), human PP4c (UniProt: P60510) and human PP6c (UniProt: O00743-1), each with both N-terminal Flag and HA (YPYDVPDYA) tags, were individually cloned (pAcGP67A) and expressed in Tni cells using the protocol described above. Purification was carried out using the same protocol as Flag–FBXO42–SKP1 above, with the following exceptions: 5 mM MgCl2 was replaced with 1 mM MnCl2; 0.1–0.7 ml of settled anti-DYKDDDDK magnetic beads (Genscript) were used per 1 l biomass; the HiLoad 16/600 Superdex 75 pg (Cytiva) column was used instead of the Superdex 200. The final SEC buffer was as follows: 20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM MnCl2, 10% glycerol, and 1 or 3 mM TCEP pH 7.5. The protein concentration was measured using the absorbance at 280 nm (NanoDrop). The final SEC fractions containing the PP2Ac, PP4c or PP6c were separately combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. Human PP2Ac constructs with a N-terminal His tag, both with and without the C-terminal tail (Δ296-309), were purified using the same method with the following exceptions: Ni-charged Magbeads (GensScript, 1 ml settled resin per 1 l biomass) were used instead of anti-DYKDDDK magnetic beads, and the samples were further processed by a 5 ml HisTrap HP (Cytiva) column using a gradient elution (similar to PP2Aa+c dimer described below) and a 1 ml HiTrap Q (Cytiva) column (similar to His-CCDC6 described below) before being passed over the Superdex 75 column. PP2Aa+c dimer Human PP2Ac (UniProt: P67775-1) with an N-terminal Flag tag and human PP2Aa (UniProt: P30153) with an N-terminal His tag were individually cloned (pAcGP67A) and co-expressed in Tni cells using the protocol described above. Cell lysis and Flag affinity purification were performed using the same methods as for Flag–FBXO42–SKP1 above, with the following exceptions: 2 ml of settled anti-DYKDDDDK magnetic beads (GenScript) were used per 1 l biomass; incubation of anti-DYKDDDDK magnetic beads and batch purification methods were carried out three times on the same lysate. Flag eluates were combined and imidazole was added to 20 mM concentration. The sample was further processed by a 5 ml HisTrap HP (Cytiva) column using a gradient elution (starting buffer: 25 mM Tris pH 7.5, 150 mM NaCl, 20 mM imidazole, 10% glycerol, 5 mM ATP pH 7.3, 5 mM MgCl2, 1 mM TCEP pH 7.5; final elution buffer: starting buffer with 300 mM imidazole and without ATP). Eluates containing dimer were passed over the HiLoad 16/600 Superdex 200 pg (Cytiva) column equilibrated in the following SEC buffer: 20 mM HEPES pH 7.5, 300 mM NaCl, 10% glycerol, 1 mM TCEP pH 7.5. The protein concentration was measured using absorbance at 280 nm (NanoDrop). The final SEC fractions containing the PP2a+c dimer were combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. His–CCDC6 Human CCDC6 (UniProt: Q16204) with an N-terminal His tag was cloned (pAcGP67A) and expressed in Sf9 cells using the protocol described above. The frozen cell pellet of expressed His–CCDC6 was lysed using the same method as for Flag–FBXO42–SKP1 with the following exceptions: Ni-charged MagBeads (GenScript, 1 ml settled beads per 1 l biomass) were used instead of anti-DYKDDDDK magnetic beads. Batch purification methods were used to purify His-CCDC6. Ni-charged MagBeads underwent three washes with 5× settled bead volume of wash buffer (25 mM Tris pH 7.5, 150 mM NaCl, 20 mM imidazole, 10% glycerol, 2 mM MgCl2, 1 mM TCEP pH 7.5), followed by 5–8 washes with 1.5× settled bead volume of elution buffer (wash buffer with 300 mM imidazole pH 8.0). The AmMag MR magnetic rack (GenScript) was used to bind beads, decant washes and decant eluates. For each wash and elution, the mixture of buffer and beads was gently inverted 5–6 times before decanting. Incubation of Ni-charged MagBeads and batch purification methods were carried out four times on the same lysate. All washes and eluate fractions that contained His–CCDC6 were combined, diluted to a final NaCl concentration of 20–50 mM, and processed by a 5 ml HiTrap Q HP (Cytiva) column using a gradient elution (starting buffer: 25 mM Tris pH 8.5, 20 mM NaCl, 1 mM TCEP pH 7.5; elution buffer: starting buffer with 1 M NaCl). Fractions containing His–CCDC6 were combined, concentrated and passed over the HiLoad 16/600 Superdex 200 pg (Cytiva) column equilibrated in the following SEC buffer: 20 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol, 1 mM TCEP pH 7.5. A portion of His-tagged CCDC6 was incubated overnight at 4 °C with TEV protease (in-house) to cleave the N-terminal His tag before injection onto the SEC column. Intact LC–MS was used to confirm His tag removal. The protein concentration was measured using absorbance at 280 nm (NanoDrop). The final SEC fractions containing CCDC6 were combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. PP2Aa/B56 Human PP2Aa (UniProt: P30153) and residues Ile30-Ala437 of B56 (UniProt: Q13362-1), each with an N-terminal His tag, were individually cloned (pAcGP67A) and expressed in Tni cells using the protocol described above. Purification was carried out using the same method as in His–CCDC6 above, with the following exception: samples purified by Ni-charged MagBeads were further processed by a 5 ml HisTrap HP (Cytiva) column using gradient elution (buffers were the same as the batch purification) before being processed by SEC. To obtain a sample without His tag, purified protein was incubated overnight at 4 °C with hyperTEV6066 protease (in-house). His tag removal was confirmed by intact LC–MS and the sample was further processed by reverse nickel using the 1 ml HisTrap HP (Cytiva) column. The flow through from reverse nickel was passed over a HiLoad 16/600 Superdex 200 pg (Cytiva) column equilibrated in the following SEC buffer: 25 mM Tris pH 7.5, 150 mM NaCl, 10% glycerol, 1 mM TCEP pH 7.5. The protein concentration was measured using absorbance at 280 nm (NanoDrop). The final SEC fractions containing PP2Aa or B56 were separately combined, concentrated, flash-frozen in liquid nitrogen and stored at −80 °C until use. Mutant proteins All mutant FBXO42 and CCDC6 proteins were expressed at a 200 ml scale in Sf9 cells and purified using a semi-automated parallel affinity purification method, as described previously65. In brief, the proteins were lysed and purified in a buffer of 25 mM HEPES pH 7.5, 150 mM NaCl, 1 mM TCEP using Integrated Micro Chromatography Systems tips on the Hamilton STAR liquid handler. The elutions were performed using 150 μg ml−1 of 3×Flag peptide for Flag-tagged proteins and 250 mM imidazole for His-tagged proteins. The protein concentration was measured using absorbance at 280 nm (NanoDrop). Proteins were flash-frozen in liquid nitrogen and stored at −80 °C until use. LCMT-1 LCMT-1 (UniProt: Q9UIC8-1) with an N-terminal His tag was cloned (pAcSG2) and expressed in BL21 DE3 cells at 16 °C overnight after induction with 0.4 mM IPTG. Cells were lysed in 20 mM HEPES pH 7.5, 200 mM NaCl, 20 mM imidazole, 0.5 mM TCEP (lysis buffer), and Roche EDTA-free protease inhibitors using a microfluidizer. Whole-cell lysates were clarified by ultracentrifugation (186,000g), and then filtered by a 0.8 μm filter. The sample was then applied to a 5 ml HisTrap column (Cytiva) pre-equilibrated in the lysis buffer. Protein was eluted using lysis buffer with 300 mM imidazole. The elution fractions containing LCMT-1 were processed by anion exchange using a 5 ml Hi Trap Q (Cytiva) column with a gradient elution (starting buffer: 25 mM Tris pH 8.0, 0.5 mM TCEP; elution buffer: starting buffer with 0.5 M NaCl). The fractions containing LCMT-1 were combined, incubated with TEV protease (in-house) to cleave the N-terminal His tag and processed again by anion exchange. The fractions containing LCMT-1 were passed over a HiLoad 16/600 Superdex 200 pg (Cytiva) column equilibrated in 50 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP. The protein concentration was measured using the absorbance at 280 nm (NanoDrop). The final SEC fractions containing LCMT-1 were concentrated, flash-frozen in liquid nitrogen and stored at −80C until use. PP2Ac methylation A 20:2 µg ratio of PP2Ac:LCMT-1 was combined with 200 μM SAM (New England Biolabs) and the mixture was incubated at room temperature. The reaction was monitored by intact LC–MS at the following timepoints: 1 h, 2 h, 5 h and 24 h. Analysis was performed on the Agilent 6230 LC/TOF mass spectrometer coupled to an Agilent 1260 Infinity HPLC system with an AdvanceBio RP-mAb Diphenyl column (2.1 × 50 mm). The sample was eluted by gradient elution with 0.1% trifluoroacetic acid in ACN. Data were analysed using Agilent MassHunter Qualitative Analysis software (Agilent v.B.07.00) and plotted using GraphPad Prism v.10.0.3 for macOS, GraphPad Software (www.graphpad.com). Intact LC–MS spectra were evaluated for a 14 Da increase from the molecular mass of PP2Ac (38,841 Da). At the 24 h timepoint, nearly all PP2Ac had been methylated (Extended Data Fig. 4f). Cryo-EM sample preparation, data collection, image processing and model building Cryo-EM sample preparation Cryo-EM samples were prepared by incubating 1.5 nmol each of mPP2Ac and CCDC6 for 2 h at 4 °C (mPP2Ac was prepared using the PP2Ac methylation method described above with a 88:13 µg ratio of PP2Ac:LCMT1; the reaction was done at a final concentration of 2 mg ml−1 PP2Ac from which 1.5 nmol was used to prepare the cryo-EM sample). Then, 0.8 nmol of SCFFBXO42 was added to mPP2Ac–CCDC6 and the entire sample was allowed to incubate overnight at 4 °C. The entire sample was passed over the Superose 6 Increase 3.2/300 (Cytiva) column equilibrated in the following SEC buffer: 20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5. The two centre fractions on the peak indicated in Extended Data Fig. 4g (asterisks) were combined (around 0.2 mg ml−1 as measured by NanoDrop, absorbance at 280 nm) and vitrified on grids for cryo-EM analysis. A similar protocol was used for the 3×FBXO42-CCDC6–mPP2Ac structure, with the exception that the sample was mixed with in-house developed cryoprotective peptides immediately before vitrification. Data for all structures were collected from Ultraufoil 300 mesh 1.2/1.3 (Quantifoil) grids coated in self-assembled monolayers (4 mM monothiolalkane(C11)PEG6-OH (11-mercaptoundecyl) hexaethyleneglycol (SPT-0011P6, SensoPath Technologies))67 and blotted using the Vitrobot MARK IV (Thermo Fisher Scientific) for 3 s at 7 force. Additional data collection for the 3×FBXO42–SKP1-bound CCDC6–mPP2Ac structure was obtained from Ultraufoil 300 mesh 1.2/1.3 (Quantifoil) grids subjected to glow discharge using a Solarus Model 950 Advanced Plasma System (Gatan). All grids were clipped for use with a Thermo Fisher Scientific autoloader (Nanosoft). Cryo-EM data collection For the high-resolution FBXO42 degradasome, 21,534 videos were collected on the Krios G3i microscope (Thermo Fisher Scientific) with a BioQuantum energy filter and K3 camera (Gatan). The K3 detector was operated in non-CDS counting mode and the energy filter slit width was set to 20 eV. Videos were collected at a nominal magnification of ×105,000, physical pixel size 0.838Å px−1, with a 50 µm C2 aperture at a dose rate of 15 e− px−1 s−1. A total dose of 64 e− Å−2 was collected as a 59-frame video, resulting in a 3 s video with 1.08 e− per frame. Data were collected semi-automatically using SerialEM68 image shift pattern of three images per hole in a 3 × 3 pattern. The nominal defocus was set to −0.8 to −2.4 µm. Beam tilt was corrected for using SerialEM’s image shift vs coma calibration. For the low-resolution 3×FBXO42 complex, 4,412 videos were collected on the Glacios1 microscope (Thermo Fisher Scientific) with a K3 camera (Gatan). The K3 detector was operated in non-CDS counting mode. Videos were collected at a nominal magnification of ×36,000, physical pixel size 1.114Å px−1, with a 50 µm C2 aperture and 100 µm at a dose rate of 15 e− px−1 s−1. A total dose of 36 e− Å−2 was collected as a 30-frame video, resulting in a 3 s video with 1.2 e− per frame. Data were collected automatically using Smartscope69 and SerialEM68 image shift pattern of three images per hole in a 3 × 3 pattern. Nominal defocus was set to −1.0 to −2.5 µm. Beam tilt was corrected for using SerialEM’s image shift vs coma calibration. Cryo-EM image processing For the high-resolution FBXO42 degradasome, cryo-EM data were transferred, patch motion-corrected, patch CTF-estimated, template picked and extracted in a 450 px box binned to 128 px. In total, 7,261,626 putative particles were picked and after two rounds of 2D classification and selection, 2,309,685 particles were further classified in 3D. In total, 551,049 of particles from the classes with the largest assemblies were used to generate 5 ab initio references. The 2.3 million particles were subjected to four rounds of subsequent heterogeneous refinement using these five classes as input references. Then, 578,587 particles from two large assembly classes were subjected to a new round of ab initio reconstruction to generate five new 3D references. The same 578,587 particles were run through a round of heterogenous refinement in parallel, the best FBXO42-bound class from each selected, and the duplicates were removed before performing a NU-refinement with dynamic masking. Global CTF refinement and another round of heterogenous refinement were performed on the remaining 147,335 particles. After the last round of heterogeneous refinement, 98,992 particles were extracted from the original micrographs at a pixel size of 1.24 Å px−1. Concurrently the original 2.3 million particles were subjected to three rounds of heterogenous refinement with a good FBXO42-bound class from the original ab initio references and three junk classes from a previous, unrelated dataset. Then, 408,123 particles were recovered in the good class and subjected to a new round of ab initio reconstruction to generate four new volumes. After a final round of heterogenous refinement, 135,171 particles were extracted at 1.24Å px−1, combined with the original track particles, and duplicates within 50 Å of each other were removed. The remaining 155,244 particles were run through reference-based motion correction at pixel size 0.838 Å px−1, NU refinement and multiple local refinements to generate maps that ranged in resolution from 3.0 to 4.2 Å. From the second track of classification, two classes were selected from the final heterogeneous reconstruction representing 8× occupied PP2Ac. These particles were extracted at 1.24 Å px−1, duplicates within 50 Å were removed and a NU-refinement yielded a map at 3.5 Å. Local refinements were run with multiple masks to generate maps ranging from 3.2 to 4.5 Å resolution. All data were processed in cryoSPARC (v.4.4.0)70. For the low-resolution 3×FBXO42 complex, cryo-EM data were transferred, patch motion-corrected and patch CTF-estimated in cryoSPARC live. Images with CTF fits higher than 8 Å were processed. Initial 2D averages for template picking were generated from 8,889 particle blobs picked with a diameter of 140–170 Å. Better templates were generated by template picking 510,891 particles with a diameter of 180 Å. These templates were used to pick particles from 2609 micrographs. Then, 32,536 particles from the best 2D averages were used to generate ab initio models and 2,672,451 particles were subjected to iterative heterogeneous refinement. After a final stack of 209,666 particles was refined using NU-refinement, these particles were subjected to 3D classification without alignment at different resolution filters. Classes with any indication of more than two FBXO42 densities bound were subjected to another round of 3D classification and selection. The final particle stack of 57,743 particles was used to train a Topaz model for picking 3,111 micrographs. These 2,468,634 particles were subjected to heterogeneous refinement using the original ab initio models. A final stack of 158,066 particles was subjected to focused 3D classification and the classes with strongest density for the third FBXO42 density were selected. Two classes from the initial 2D classification that also showed extra FBXO42 density around the central core were also selected. These stacks of particles were merged, and duplicates were removed to give a final stack of 28,195 particles. These particles yielded a 6.9 Å map after local refinement around the central PP2A and FBXO42 densities. The local resolution in the map ranged from 5 to 15 Å. Model building and refinement Model building for the FBXO42-bound degradosome (FCP) and CCDC6–PP2Ac (CP) complexes was done by combining all the local refinements in a single composite map using the Combine Focused Maps job in Phenix71. A PP2Ac model was derived from PDB 3FGA and rigid-body fit into the maps using ChimeraX72. We generated a model of FBXO42–SKP1 using AlphaFold229 and rigid body fit into the maps using ChimeraX. The resolution in the map was sufficiently high enough for us to manually build CCDC6 in Coot73. Models were then subjected to local and global real-space refinement in Coot and Phenix. For the 3×FBXO42 structure, we first docked in the high-resolution FCP structure, and used the coordinates of FBXO24–SKP1 derived from the FCP structure to dock in the third copy using ChimeraX. Coordinates of this low-resolution model have been uploaded to Zenodo (10.5281/zenodo.17260372). Final statistics of the maps and models are reported in Extended Data Table 1. NanoBit Construct generation The NanoBiT constructs and the respective amino acid boundaries of the LgBiT-tagged proteins are annotated as follows: LgBiT-PP2Ac (aa. 2-309); LgBiT-PP4c (aa. 2-307); LgBiT-PP6c (aa. 2-305); FBXO42-SmBiT-IRES-mNeon (aa. 95-717). The cDNAs were synthesized, codon optimized for mammalian expressions, and then cloned into a pRK5 vector, which is adapted for high-throughput cloning. Recombinant genes are expressed as fusion proteins when cloned into one of the available restriction enzyme sites under the control of the strong CMV promoter. Stable cell line generation and maintenance The FBXO42∆F-box-SmBiT_IRES_mNeon construct was used for stable cell line generation. A549 FBXO42 KO cells were seeded at 2.5 × 106 per 15 cm2 TC-treated dish in RPMI medium 24 h before transfection. For transfection, 7 μg of the DNA construct and 3 μg of the PiggyBac Transposase Expression Vector were diluted in Opti-MEM (Gibco, 31985-062). FuGene HD Transfection Reagent (Promega, E2312) was used according to the manufacturer’s instructions at a ratio of 3 μg (FuGene):1 μg (DNA). After a brief incubation, diluted DNA and transfection reagents were mixed and incubated for 5 min. The mixture was added to the cells drop by drop. Cells were incubated at 37 °C TC incubator with 5% CO2 for 48 h before medium exchange containing 200 μg ml−1 of Zeocin (Thermo Fisher Scientific, R25001) for 2 weeks. Transfection, cell lysis and NanoBiT assay The protocol was adopted from a previous study74. A549 FBXO42 KO cells stably expressing FBXO42∆F-box-SmBiT fusion were maintained in a tissue culture incubator at 37 °C and 5% CO2 in RPMI medium supplemented with FBS (v/v, 10%), l-glutamine (3 mM) and penicillin–streptomycin antibiotics (100 U ml−1). Before transfection, cells were seeded at 2.5 × 106 into 15 cm2 TC-treated dishes for 24 h. Individual transfection of LgBiT constructs was carried out according to the manufacturer’s instructions using FuGene HD Transfection Reagent (Promega, E2311) diluted with Opti-MEM (Gibco, 31985-062) at a ratio of 3 μg (FuGene):1 μg (DNA). For individual transfection, 7 μg DNA was used per 15 cm2 dish. After 48 h of transfection, the medium was removed, and the cells were washed once with PBS buffer. The cells were then collected and lysed with 1× passive buffer (Promega, E1941) supplemented with protease (Roche, 05892970001) and phosphatase (Roche, 04906837001) inhibitors according to the manufacturer’s instructions. The cell lysates were sonicated for 10 s and centrifuged in a tabletop centrifuge at 4 °C for 10 min at maximum speed. The cell lysates were then quantified using a BCA protein assay kit (Thermo Fisher Scientific, 23227), snap-frozen in liquid nitrogen and then stored at −80 °C. Thawed cell lysates were diluted using titration buffer (25 mM Tris pH8, 100 mM NaCl) to 1 μg μl−1 and dispensed to white Proxiplates (PerkinElmer, 6008280) (7 μl per well). The Nano-Glo Live Cell Substrate (Promega, N2012) was prepared according to the manufacturer’s instructions, and 3 μl of the diluted substrate was added per well. Plates were then centrifuged at 360 rpm for 1 min and incubated for 10 min on a shaker at 300 rpm, and the luminescence signal was recorded using the Envision plate reader. HTRF assays FBXO42, CCDC6 and phosphatase binding HTRF First, CCDC6 was prepared at concentrations of 120 nM, 60 nM, 30 nM and 0 nM (control) in the assay buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5, 0.1% BGG). This was done in the presence of 2 nM Europium-anti-Flag donor dye (PerkinElmer, AD0274). Next, twofold serial dilutions of 240 nM Flag–HA–PP2Ac, Flag–HA–PP4c or Flag–HA–PP6c were performed at the fixed CCDC6 concentrations described above. These mixtures were then incubated for 30 min on a rotor with mild agitation. Third, the acceptor mix was prepared by labelling either 120 nM, 60 nM or 0 nM (control) biotinylated Avi–FBXO42–SKP1 with 0.01 mg ml−1 Streptavidin-XL665 acceptor (PerkinElmer, 610SAXLB). This solution was incubated for 30 min. Finally, an equal amount (4 µl) of each donor and acceptor mix was pipetted into individual wells of assay plates (ProxiPlate-384 F Plus, Revvity 6008260) and incubated on a shaker with mild agitation for 1 h. Data were subsequently collected using the PheraStar plate reader (BMG Labtech). FBXO42, CCDC6 and mPP2Ac binding HTRF Methylated or non-methylated Flag–HA–PP2Ac, diluted to a 2× concentration (120 nM) in assay buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5, 0.1% BGG), was incubated with 2 nM Europium-anti-Flag donor fluorophore (PerkinElmer, AD0274) and 2× CCDC6 (120 nM) for 30 min at room temperature. Concurrently, 2× biotinylated Avi–FBXO42–SKP1 (120 nM) was incubated with 0.01 mg ml−1 Streptavidin-XL665 acceptor (PerkinElmer, 610SAXLB) for 30 min. An equal volume (4 µl) of both the donor and acceptor mixes was then pipetted into each well of ProxiPlate-384 F Plus assay plates (Revvity 6008260), resulting in equimolar final protein concentrations (60 nM). The plates were incubated on a shaker with mild agitation for 1 h at room temperature, after which data were collected using the PheraStar plate reader (BMG Labtech). For peptide competition assays, the procedure was amended as follows. Proteins were labelled with either donor or acceptor fluorophores at 4x concentrations (240 nM), combined in equal volume and incubated for 1 h at room temperature. Subsequently, 4 µl of this combined mix was pipetted into each well of ProxiPlate-384 F Plus assay plates (Revvity 6008260), followed by the addition of 4 µl of serially diluted peptides at 2× concentrations in assay buffer. The HTRF ratio obtained at the 3 h timepoint is presented on the graph. PP2Ac and CCDC6 binding HTRF assays Non-methylated Flag–HA–PP2Ac was serially diluted in assay buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5, 0.1% BGG, 1 mM DTT) containing 2 nM Europium-anti-Flag donor fluorophore (PerkinElmer, AD0274) at concentrations of 480 nM, 240 nM, 120 nM, 60 nM and 0 nM (control). Biotinylated CCDC6 was diluted in the assay buffer in the presence of 0.01 mg ml−1 of SA-d2 acceptor fluorophore (Revvity, 610SADLB). The proteins were then labelled for 30 min at room temperature. An equal amount (4 µl) of each donor and acceptor mix was pipetted into individual wells of assay plates (ProxiPlate-384 F Plus, Revvity 6008260). The HTRF ratio was subsequently quantified using the PheraStar plate reader (BMG Labtech) in a time-course manner. The 3 h timepoint is presented on the graph. FBXO42 competition HTRF assays To generate the donor mix, methylated Flag–HA–PP2Ac was diluted to a 4× concentration (240 nM) in the assay buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5, 0.1% BGG) containing 4 nM Europium-anti-Flag donor fluorophore (PerkinElmer, AD0274), and incubated for 30 min at room temperature. CCDC6 was then serially diluted at 4× concentration using labelled PP2Ac mix to achieve various PP2Ac:CCDC6 ratios (0.5:1, 1:1, 2:1 and 4:1), as well as a sample without CCDC6. The acceptor mix was prepared by incubating biotinylated Avi–FBXO42–SKP1 at 4× concentration (240 nM) with 0.02 mg ml−1 Streptavidin-XL665 fluorophore (PerkinElmer, 610SAXLB) for 30 min at room temperature. Next, the donor and acceptor mixes were then combined in equal volumes and incubated for one hour. Following this, 4 µl of the combined mix was added to each well of the assay plate (ProxiPlate-384 F Plus, Revvity, 6008260). Finally, the competitor unlabelled non-biotinylated Avi–FBXO42–SKP1 was serially diluted in the assay buffer and 4 µl of the competitor was added to wells. A time-course quantification was then performed using the PheraStar plate reader (BMG Labtech); the 3 h timepoint was used for quantification of the apparent binding affinity. The apparent binding affinity (KD) was calculated using the nonlinear fit of the HTRF competition assay results. The HTRF signal was normalized to the total FBXO42 (normalized HTRF signal = (observed HTRF signal/[FBXO42labelled]) × [FBXO42total]). Holoenzyme HTRF assays To assess whether CCDC6 can induce the disassembly of the PP2A holoenzyme complex, 4× concentrated non-methylated Flag–HA–PP2Ac (20 nM) was first labelled in the assay buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 1 mM TCEP pH 7.5, 0.1% BGG) at 4 nM Europium-anti-Flag donor fluorophore (PerkinElmer, AD0274). Concurrently, 4× concentrated His-tagged PP2Aa was labelled at 200 nM anti-His-d2 acceptor fluorophore in the presence of 200 nM untagged B56. Both donor and acceptor mixes were combined in equal volumes and allowed to incubate for an additional 2 h. CCDC6 was serially diluted at 2× concentrations (concentration range, 1,000–0 nM), and 4 µl of the assembled holoenzyme and CCDC6 was added to wells in the assay plate ProxiPlate−384 F Plus assay plates (Revvity, 6008260). Data were collected using the PheraStar plate reader (BMG Labtech). The 3 h timepoint is presented in the figure. SPR SPR experiments in this paper were recorded on a Biacore 8K instrument. Protein was captured at 20 °C with data collection at 5 °C. A Xantec SAHC 200M (Streptavidin) chip was inserted into the instrument and primed into running buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 0.002% Tween-20, 0.25 mM TCEP and 2% DMSO). Avi-tag-biotinylated FBXO42 protein in a complex with SKP1 was prepared at 400 nM in capture buffer (50 mM HEPES pH 6.5, 150 mM NaCl, 0.002% Tween-20, 0.25 mM TCEP and 2% DMSO) and captured on FC2 to 3,500–4,000 RU. Affinity was measured using a titration series, injected from low to high with a six-point, twofold dilution dose–response with a top concentration of 50 µM in multicycle kinetics format with a flow rate of 50 μl min−1, and a contact and dissociation time of 30 s and 300 s, respectively. All data were analysed using the Insight software (Cytiva) using a 1-to-1 binding affinity model. Statistics and reproducibility Representative WB experiments The numbers (n) of independent repeats for Figs. 1g,h, 2c,d, 3g, 4d,e,i and 5g,i,j and Extended Data Figs. 3d, 8e and 10h were n ≧ 3. The numbers (n) of independent repeats for Figs. 1a,i,j, 2b,e and 4h,j and Extended Data Figs. 2e, 3a,b,e, 4e and 10b were n ≧ 5. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-026-10368-z. Supplementary information Supplementary Figure 1Raw images with at least one molecular mass marker (kDa) of all WB in the Article. Reporting Summary Supplementary Table 1CRISPR KO cell survival screen 1 (guide level). Supplementary Table 2CRISPR KO cell survival screen 2 (gene level). Supplementary Table 3KGG-MS in HOP-62 DOX-inducible FBXO42 cells. Supplementary Table 4IP–MS in HOP-62 Flag–FBXO42(ΔF-box) cells. Supplementary Table 5Affinities and IC50 values measured by HTRF and SPR. Peer Review File Raw images with at least one molecular mass marker (kDa) of all WB in the Article. Reporting Summary CRISPR KO cell survival screen 1 (guide level). CRISPR KO cell survival screen 2 (gene level). KGG-MS in HOP-62 DOX-inducible FBXO42 cells. IP–MS in HOP-62 Flag–FBXO42(ΔF-box) cells. Affinities and IC50 values measured by HTRF and SPR. Peer Review File Source data Source Data Figs. 1–5 and Source Data Extended Data Figs. 1, 2, 4, 9 and 10. Source Data Figs. 1–5 and Source Data Extended Data Figs. 1, 2, 4, 9 and 10.