In silico APC/C substrate discovery reveals cell cycle-dependent degradation of UHRF1 and other chromatin regulators

The anaphase-promoting complex/cyclosome (APC/C) is an E3 ubiquitin ligase and critical regulator of cell cycle progression. Despite its vital role, it has remained challenging to globally map APC/C substrates. By combining orthogonal features of known substrates, we predicted APC/C substrates in silico. This analysis identified many known substrates and suggested numerous candidates. Unexpectedly, chromatin regulatory proteins are enriched among putative substrates, and we show experimentally that several chromatin proteins bind APC/C, oscillate during the cell cycle, and are degraded following APC/C activation, consistent with being direct APC/C substrates. Additional analysis revealed detailed mechanisms of ubiquitylation for UHRF1, a key chromatin regulator involved in histone ubiquitylation and DNA methylation maintenance. Disrupting UHRF1 degradation at mitotic exit accelerates G1-phase cell cycle progression and perturbs global DNA methylation patterning in the genome. We conclude that APC/C coordinates crosstalk between cell cycle and chromatin regulatory proteins. This has potential consequences in normal cell physiology, where the chromatin environment changes depending on proliferative state, as well as in disease.


Introduction
Regulated protein degradation is central to cell and organismal physiology and plays a particularly important role in proliferation. In eukaryotes, protein degradation is controlled largely by a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 suggest that cell cycle-regulated protein degradation helps organize the epigenetic landscape during proliferation. This suggests a potential mechanistic link contributing to changes in the chromatin landscape observed between proliferating and non-proliferating cells [32,33]. We predict that altering APC/C function could promote changes in the histone and DNA modification landscape and that these effects could contribute to the biochemical and phenotypic features of diseases, including cancer and neurological disorders.

Identification of APC/C substrates
To identify human APC/C substrates, we first performed FLAG immunoprecipitations (IP) from asynchronous HEK-293T cells expressing amino-terminal-tagged FLAG-Cdh1 or an empty vector and analyzed precipitated proteins by mass spectrometry (S2 Data). Several APC/C complex components and known substrates, including Rrm2, Kif11, Claspin, and cyclin A, were enriched in Cdh1 pulldowns. Compared with a previously established dataset [34], we identified 15 out of 53 known substrates. However, hundreds of proteins were enriched over controls and many known substrates scored weakly. For example, a single spectral count was observed for the substrate Kif22/KID [35,36]. The prevalence of non-specific interactors and other non-substrate-binding proteins confounded our ability to prioritize these data to identify new substrates.
We considered computationally identifying substrates based on features common among substrates. APC/C binds substrates most often through D-and KEN-box degron motifs. The minimal D-box motif (R-x-x-L) is present in most human proteins and insufficient as a prediction tool. The KEN-motif is found in approximately 10% of human proteins (2,206; S3 Data), and several D-box-regulated substrates also contain a KEN-motif, including Securin and Cdc6 [37,38]. In addition, the gene expression of most APC/C substrates oscillates during the cell cycle [39]. We cross-referenced the KEN-motif containing proteins against a set of 651 proteins whose mRNAs scored in at least 2 cell cycle mRNA profiling studies [40][41][42][43]. Overlapping the 2,206 KEN-motif containing proteins with 651 transcriptionally controlled genes produced a set of 145 proteins, which represent known and putative APC/C substrates ( Fig 1A  and S3 Data).
We compared our in silico analysis with 2 previously curated datasets, 1 containing 53 known APC/C targets [34] and a second containing 33 specifically KEN-dependent APC/C substrates [16]. When compared with these lists of 53 and 33 substrates, our dataset captured 26 and 22 of them, respectively, the latter representing an enrichment of more than 140-fold, compared with what would be expected by chance (Fig 1B). We compared both our Cdh1 IP/ MS dataset and in silico analysis to several other studies that identified APC/C substrates, interactors, proteins degraded at mitotic exit, or proteins ubiquitylated in mitosis (S4 Data) [34,[44][45][46][47][48][49]. Our in silico analysis identified the most KEN-dependent substrates relative to these studies (Fig 1C and S4 Data). When compared to the set of 53 substrates, which includes both D-and KEN-box-dependent substrates, our dataset captured 26 out of 53 known substrates, despite not focusing on D-box substrates. Combining the in silico predictions with our Cdh1-pulldown proteomics data, we captured 31 out of 53 substrates.
Among the 145 computationally identified known and potential substrates, gene ontology (GO) analysis showed a strong enrichment for processes linked to various aspects of cell division ( Fig 1D). Whereas the analysis of cell cycle genes expectedly enriched for GO terms related to cell division (Fig 1D), these same terms were more significantly enriched when the analysis was restricted to those cell cycle genes that encode proteins containing a KEN-motif. Manual curation demonstrated that nearly half of the proteins we identified (70 of 145) have well-established roles in cell cycle. These were subclassified into the following subcategories: cytoskeleton and motors, centromere-kinetochore, APC/C and spindle checkpoint, cytokinesis, mitotic entry, cell cycle transcription, cohesion and condensation, and DNA replication ( Fig 1E). Among these 70 proteins, 50% have literature evidence for regulation by APC/C, highlighting our enrichment for APC/C substrates (Fig 1E; shown in magenta). All 145 proteins, their known function, subcategory, KEN-box sequence motif with flanking sequence, aliases, and citations describing regulation by APC/C are detailed in S3 Data.

Regulated degradation of chromatin factors
Unexpectedly, our dataset revealed several proteins involved in chromatin regulation (Fig 2A) and an enrichment for GO processes related to chromatin (Fig 2B). The dataset includes readers and writers of histone posttranslational modifications, including the lysine acetyltransferases, PCAF/KAT2B and NCOA3/KAT13B, the lysine methyl-transferase MLL2/KMT2D, the chromatin reader and histone Ub ligase UHRF1, and the mitotic histone H3 kinase Aurora B (Figs 2A and 1E). We identified proteins involved in chromatin assembly and structure, including CHAF1B, a component of the CAF-1 nucleosome assembly complex; TTF2, a Swi2/ Snf2 family member and DNA-dependent ATPase; KI-67, which prevents chromosome aggregation in mitosis and regulates histone posttranslational modifications; and proteins associated with cohesion and condensation such as SMC4 and NIPBL (Fig 1E). We also identified proteins involved in DNA damage repair.
To validate potential substrates, we developed an in vivo APC/C activation assay that is amenable to analysis of endogenous or exogenously expressed proteins and which is similar to approaches described elsewhere [50]. U2OS cells were synchronized in mitosis with the microtubule poison nocodazole. After harvesting cells by mitotic shake-off, CDK1 was inactivated with either the CDK1-specific inhibitor RO-3306 or pan-CDK inhibitor Roscovitine, driving cells out of mitosis and triggering APC/C activation and destruction of substrates, including FoxM1, NUSAP1, and cyclin B (Fig 2C and S1 Fig) [51].
Using a combination of exogenous expression and endogenous protein analysis, we examined the levels of chromatin-related proteins not previously shown to be APC/C substrates. Using this assay, we detected a decrease in the levels of several writers of histone modifications, including UHRF1, PCAF, TTF2, and NCOA3 (Fig 2C and S1A and S1B Fig). We observed a decrease in the levels of the chromatin assembly factors NASP and CHAF1B as well as the RNA processing proteins LARP1 and LARP7 (Fig 2C and S1A and S1B Fig). All of these have been previously identified as ubiquitylated in proteomics studies by an unknown E3 ligase [52][53][54][55][56].
Since the role of APC/C in chromatin regulation is not well established, we focused our attention on the potential regulation of chromatin proteins by APC/C. We determined the ability of a subset to bind Cdh1 by co-immunoprecipitation (coIP). CHAF1B, PCAF, NCOA3, and TTF2 interact with Cdh1 by coIP in 293T cells (Fig 2D-2G). Accordingly, the levels of In silico analysis reveals a high confidence set of APC/C substrates involved in mitosis. (A) KEN-box containing human proteins were identified and cross-referenced against a set of 651 genes whose expression is cell cycle regulated based on multiple, independent studies. This revealed a set of 145 KENbox containing proteins whose mRNA expression is cell cycle regulated. (B) Analysis of the enrichment of bona fide KEN-dependent substrates among these 3 datasets (blue, KEN box only set (2,206); black, cell cycle regulated mRNAs (651); red, the overlapping set of 145 proteins compared against a curated set of bona fide, KEN-dependent APC/C substrates [16]). Enrichment was calculated based on the expected number of substrates, which would be captured by chance based on the size of the dataset. (S1 Data) (C) Analysis of putative substrates recovered in the indicated studies. (S1 Data) (D) GO analysis for indicated studies (blue, KEN box only set (2,206); black, cell cycle-regulated mRNAs (651); red, the overlapping set of 145 proteins). (S1 Data) (E) The set of 145 putative substrates was manually curated and analyzed for roles in various aspects of cell cycle progression. Seventy proteins, involved in cell cycle activities, are shown. The ones labeled in magenta signify that there is evidence in the literature of their regulation by APC/C. (Note that AURORA B, a mitotic kinase that phosphorylates histone H3, is listed here and in Fig 2A). APC/C, anaphase-promoting complex/cyclosome; GO, gene ontology.
https://doi.org/10.1371/journal.pbio.3000975.g001 PLOS BIOLOGY endogenous CHAF1B, TTF2, and NCOA3 oscillate during the cell cycle in U2OS, analyzed following a nocodazole-induced block in mitosis and then released into the cell cycle ( Fig 2H). PCAF levels did not decrease at mitotic exit in U2OS (S1C Fig) but do decrease at mitotic exit in HeLa cells (S1C Fig), suggesting a potentially complex regulation. Finally, we purified recombinant TTF2 and found that APC/C could trigger its ubiquitylation in vitro (S2 Fig). A table of all proteins tested in these assays and their validation is shown in S5 Data. Taken together, this analysis uncovered new APC/C substrates and a role for APC/C in controlling chromatin regulators.

UHRF1 regulation by APC/C Cdh1
To further understand the function of APC/C in chromatin biology, we pursued UHRF1, a key chromatin regulator that reads and writes histone modifications. UHRF1 associates with the DNA methyltransferase DNMT1 and is required for DNA methylation [26]. UHRF1 has also been implicated in replisome assembly [57,58] and its phosphorylation oscillates during the cell cycle [59].
We examined UHRF1 protein levels following a mitotic block and release. Immunoblotting for UHRF1 and other cell cycle markers showed that UHRF1 protein levels decrease during mitotic exit in HeLa S3, HeLa, and U2OS cell lines (Fig 3A and S3A and S3B Fig). In each cell line, UHRF1 levels remain low in G1 and then reaccumulate starting around G1/S-phase, based on the expression of other cell cycle markers, such as cyclin E and cyclin A, and then further increasing throughout the subsequent G2/M phase.
We performed several assays to assess whether UHRF1 is regulated by APC/C. We analyzed UHRF1 in the aforementioned in vivo APC/C activation assay. U2OS cells were arrested in mitosis and then treated with RO-3306. We observed a decrease in UHRF1 that was partially mitigated by the proteasome inhibitor, MG-132, indicating that the reduction is dependent on the proteasome (Fig 3B). In addition, transient small interfering RNA (siRNA) depletion of Cdh1 (Fzr1 mRNA transcript) augmented UHRF1 protein levels ( Fig 3C). Conversely, ectopic expression of increasing concentrations of FLAG-Cdh1 led to a dose-dependent decrease in both exogenous and endogenous UHRF1 protein levels ( Fig 3D). Moreover, Cdh1-depleted cells undergoing mitotic exit showed delayed endogenous UHRF1 degradation, comparable with well-established APC/C substrates ( Fig 3E). Additionally, utilizing a cell-free human extract system that recapitulates APC/C substrate degradation, we observed UHRF1 levels The set of 145 known and putative APC/C substrates is enriched for proteins involved in various chromatin-related process. This includes chromatin readers and writers, chaperones, RNA regulation and processing, DNA damage repair, and others. (Note that AURORA B, a mitotic kinase that phosphorylates histone H3, is listed here and in Fig 1E) (B) GO analysis of the overlapping KEN-box containing cell cycle-regulated transcripts. This set is enriched for the indicated biological process, including DNA metabolism, protein-DNA complex assembly, DNA packaging, and DNA conformation. (S1 Data) (C) APC/C activation assay to monitor substrate degradation. Following synchronization in mitosis, cells were washed 1 time and treated with CDK inhibitors to remove inhibitory phosphorylation marks that hinder the formation of APC/C Cdh1 needed for the M/G1 phase transition. Protein degradation was monitored by immunoblot. CHAF1B and PCAF are putative APC/C substrates, and FoxM1 and Cyclin B are known targets. Data representative of n = 3 experiments. (D) coIP of HA-Cdh1 with Myc-CHAF1B in transiently transfected HEK-293T cells treated with proteasome inhibitors prior to harvesting. The underline indicates which protein or tag was blotted for in a particular panel (here and below). Input equal to 1% of IP, here and below. Data representative of n

PLOS BIOLOGY
Chromatin regulation by APC/C decrease similar to Securin, a well-established APC/C substrate, and this degradation depends on the E2-conjugating enzyme UBE2S (Fig 3F). We also examined UHRF1 levels in cells that were first synchronized in G1 by a mitotic block and release and then treated with the pharmacological APC/C inhibitor proTAME for 90 min (S3C Fig). This led to an increase in endogenous UHRF1 levels. Together, these data suggest that APC/C controls UHRF1 in vivo.

UHRF1 ubiquitylation by APC/C Cdh1
UHRF1 is a multi-domain protein ( Fig 4A) that exhibits multivalent binding with chromatin through histone and DNA binding domains [24,60,61]. Additionally, UHRF1 is a RING domain E3 that ubiquitylates histone H3 [22,23,25]. To determine whether UHRF1 is a direct APC/C Cdh1 substrate, we tested its binding to Cdh1. Endogenous Cdh1 protein interacted with endogenous UHRF1 by co-IP in mitotic synchronized cells ( Fig 4B). Next, we examined binding by expressing HA-Cdh1 and Myc-UHRF1 in HEK-293T cells. Cells were treated with the proteasome inhibitor MG-132 prior to harvesting to prevent UHRF1 degradation. Myc-UHRF1 was enriched in the HA-Cdh1 pull-down, and HA-Cdh1 was enriched in the Myc-UHRF1 pull-down (Fig 4C and 4D).
Next, we purified and fluorescently labeled recombinant, bacterially expressed, full-length (FL) UHRF1 (FL-UHRF1 � , where the � denotes fluorescently labeled protein). We found that FL-UHRF1 � was ubiquitylated in an APC/C-and Cdh1-dependent manner using an entirely in vitro recombinant system ( Fig 4E). Multiple, high molecular weight ubiquitylated forms are observed using either wild-type Ub or methylated-Ub, the latter of which cannot form poly-Ub chains. This indicates that APC/C ubiquitylates multiple lysines in UHRF1 (Fig 4E and  S4A and S4B Fig).
Since UHRF1 can auto-ubiquitylate itself through its RING domain, we confirmed that its ubiquitylation is APC/C dependent. First, we purified a version of APC/C selectively missing the APC2 WHB domain and the APC11 RING domain, which are required to recruit its initiating E2 UBE2C (designated ΔRINGΔWHB) [62,63]. This version of APC/C was unable to ubiquitylate UHRF1 ( Fig 4F).
Next, we purified and fluorescently labeled a truncated version of UHRF1 that contains the Linker, PHD, and SRA domains (termed (LPS)), spanning amino acids 287-715 ( Fig 4A). The LPS fragment omits 3 potential APC/C D-box degron motifs, as well as the RING domain, precluding auto-ubiquitylation. A D-box motif remains in the highly structured SRA domain but is unlikely to be accessible as a degron motif [64].

PLOS BIOLOGY
APC/C Cdh1 -UBE2C, while APC/C Cdh1 -UBE2S elongates Ub chains, indicating that UHRF1 ubiquitylation is similar to that of other substrates tested in this in vitro system ( Fig 4G). We conclude that UHRF1 is a bona fide APC/C substrate.
The ubiquitylation of truncated LPS-UHRF1 � (Fig 4E-4G) strongly suggests the importance of the KEN-motif, located in an unstructured region at amino acids 622-624 ( Fig 4A). Alanine substitutions were introduced into the KEN sequence (UHRF1 KEN:AAA ). The KEN mutant version (Myc-UHRF1 KEN:AAA ) showed reduced, although not completely abolished, binding to HA-Cdh1 by coIP, compared with Myc-UHRF1 WT (Fig 4H). Additionally, the KEN mutant versions of FL-UHRF1 � and LPS-UHRF1 � were completely resistant to ubiquitylation by APC/C (Fig 4I). We conclude that UHRF1 ubiquitylation by APC/C Cdh1 is dependent on its KEN-box motif.
APC/C substrates are recruited by Cdc20 and Cdh1, and many substrates can be controlled by both coactivators. To test if UHRF1 is controlled by APC/C Cdc20 , in addition to APC/C Cdh1 , we used a phosphomimetic version of APC/C (termed pE-APC/C) that can utilize either Cdc20 or Cdh1, since Cdc20 cannot bind to unphosphorylated APC/C [62]. Surprisingly, unlike other, well-established APC/C substrates, including cyclin B (CycB NTD , amino acids 1-95) and Securin, the FL-UHRF1 � and LPS-UHRF1 � were ubiquitylated by APC/C Cdh1 but not by APC/C Cdc20 (Fig 4J and S4C Fig).
We transiently expressed FLAG-Cdh1 in HEK-293T cells in combination with either Myc-UHRF1 WT or mutant versions harboring alanine mutations in either the KEN-box (Myc-UHRF1 KEN:AAA ) or the fourth D-box motif (Myc-UHRF1 D4 ). Ectopic FLAG-Cdh1 overexpression triggers the degradation of Myc-UHRF1 WT and Myc-UHRF1 D4 , whereas Myc-UHRF1 KEN:AAA is resistant to degradation (Fig 5A), further supporting the importance of the KEN-motif in UHRF1 degradation.
Next, we generated cell lines constitutively expressing GFP-tagged UHRF1 WT or UHRF1 KEN:AAA using lentiviral transduction and examined UHRF1 stability upon mitotic exit. Exogenous UHRF1 levels were only moderately overexpressed compared to endogenous levels ( Fig 5B). Following synchronization with nocodazole, GFP-UHRF1 WT levels decrease at mitotic exit. Conversely, GFP-UHRF1 KEN:AAA levels remain stable through mitotic exit and G1 phase (Fig 5B). Cells expressing GFP-UHRF1 KEN:AAA exit mitosis normally based on immunoblotting for the APC/C substrates cyclin A, cyclin B, cyclin F, and Aurora A, which are degraded with normal kinetics (Fig 5B). Thus, the KEN-box regulates UHRF1 ubiquitylation in vitro and degradation in vivo. In addition, the mild overexpression of UHRF1 in these cells does not affect overall APC/C activity.

UHRF1 degradation and cell cycle progression
Since many APC/C substrates are linked to proliferative control, we examined the contribution of UHRF1, and its degradation by APC/C, to cell cycle. Consistent with prior reports, UHRF1 depletion using 3 independent siRNAs increased the fraction of cells in G1 phase (S5A Fig; [65]). To further investigate the role of UHRF1 in cell cycle, we examined mitotic

PLOS BIOLOGY
cells following UHRF1 depletion. We observed an approximately 3-fold increase in cells with misaligned chromosomes in metaphase and anaphase in UHRF1 depleted cells using 2 independent siRNA oligonucleotides (S5B Fig). Surprisingly, there was no statistically significant difference in the overall percentage of mitotic cells.
To determine the role of UHRF1 degradation in cell cycle, we examined cell cycle markers in cells expressing UHRF1 WT or UHRF1 KEN:AAA . In HeLa cells traversing the cell cycle after synchronization at G1/S, following a double thymidine block and release, we found that the GFP-UHRF1 KEN:AAA cells contain more of the G1/S regulator cyclin E (S6A Fig). This was also evident in cells that had been synchronized in mitosis and released into G1 (Fig 5B). This suggested that an inability to degrade UHRF1 in G1 alters cyclin E expression, a key driver of S-phase entry. UHRF1 depletion increased the percentage of G1 phase cells and expression of nondegradable mutant accelerated G1 progression. Together, these data suggested that UHRF1 might promote progression into S-phase and that a failure to degrade UHRF1 could shorten the duration of G1. To better address this possibility, we depleted endogenous UHRF1 with an shRNA targeting the UHRF1 3 0 UTR [66]. Cells expressing GFP-UHRF1 WT or GFP-UHRF1 KEN:AAA were synchronized in mitosis, released into the cell cycle, and analyzed by immunoblot. Increased expression of GFP-UHRF1 KEN:AAA is evident and consistent with its increased stability. Some residual degradation is evident in KEN-mutant expressing cells passing through G1-phase, which could be due to unrealized contributions from other degron sequences (see Discussion). Nevertheless, several markers of S-phase entry accumulate early in cells expressing GFP-UHRF1 KEN:AAA compared with GFP-UHRF1 WT . Both cyclin E and the G1/S transcription factor E2F1 are elevated at early time points following release from mitosis ( Fig 6A). Elevated levels of cyclin E and E2F1 are evident in asynchronous RPE1-hTRET cells and to a lesser extent in asynchronous HeLa S3 cells, where cell cycle transcription is perturbed due to HPV oncoproteins (S6B and S6C Fig).
To analyze G1 duration, cells were release from a mitotic block and pulsed with EdU prior to harvesting for flow cytometry to determine the percent of cells that were in S-phase. GFP-UHRF1 KEN:AAA expressing cells begin S-phase earlier than control cells (Fig 6B). Six hours after release into the cell cycle, 3.6% of control cells had entered S-phase, whereas 9.6% of GFP-UHRF1 KEN:AAA expressing cells had started S-phase. Thus, a failure to degrade UHRF1 accelerates G1, indicating a key role for UHRF1 destruction in determining timing between the end of mitosis and start of DNA synthesis.

UHRF1 degradation and DNA methylation homeostasis
UHRF1 is required for DNA methylation maintenance [26]. To determine if stabilizing UHRF1 in G1 affects DNA methylation, we performed base-resolution DNA methylation analysis at approximately 850,000 unique human CpG loci spanning all genomic annotations and regulatory regions using the Infinium MethylationEPIC BeadChip (EPIC arrays, Illumina, Madison, Wisconsin) [67,68]. We compared parental U2OS cells and those expressing GFP-UHRF1 WT or GFP-UHRF1 KEN:AAA . Considering all probes, DNA methylation changes between parental, GFP-UHRF1 WT , and GFP-UHRF1 KEN:AAA were insignificant ( Fig 7A). However, multidimensional scaling (MDS) of the top 50,000 variable CpG probes among all samples/replicates (agnostic of sample group) clustered experimental conditions (Fig 7B), indicating a unique and reproducible profile of methylation patterning.
We queried the GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA samples for differentially methylated CpGs relative to the parental controls. Consistent with a previous report [29], expression of GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA induced a comparable number of hypomethylation events (Fig 7C). Alternatively, GFP-UHRF1 KEN:AAA induced approximately 2-fold more hypermethylated CpGs compared with GFP-UHRF1 WT (Fig 7C). Analysis of differentially methylated CpG probes between GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA revealed a 32% overlap in hypomethylated probes and a 17% overlap in hypermethylated probes ( Fig 7D). Significantly, hypermethylated CpG probes in the GFP-UHRF1 KEN:AAA

PLOS BIOLOGY
expressing cells were 2.5-fold more abundant compared with GFP-UHRF1 WT , despite no significant change in hypomethylated CpG probes. Thus, the nondegradable form of UHRF1 induces site-specific DNA hypermethylation (Fig 7D).
The CpGs that were hypermethylated in GFP-UHRF1 KEN:AAA -expressing cells started with a higher methylation level than other categories and gained methylation due to expression of nondegradable mutant (Fig 7E). Enrichment analysis of the differentially methylated CpGs revealed that gene body annotations, including exons, introns, and transcription termination sites (TTS), were positively enriched for hypermethylation in GFP-UHRF1 KEN:AAA -expressing cells (Fig 7F, left panel). We next queried enrichment of differential methylation events in regions of early and late replication [69]. Hypermethylation events in GFP-UHRF1 KEN:AAA , but not GFP-UHRF1 WT , were positively enriched in early replicating regions of the genome, while hypomethylation events by both GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA (alone or shared in common) were enriched in late replicating DNA (Fig 7F). The enrichment of these hypermethylated features was consistent with known DNA methylation patterns that occur across gene bodies and early replicating DNA (Fig 7E), as CpG loci in these regions typically demonstrate high levels of methylation [70,71]. Taken together, these results demonstrate that expression of nondegradable UHRF1 enhances methylation at gene-rich, early replicating regions of the genome.

Identification of new E3 ligase substrates
APC/C is a core component of the cell cycle oscillator and mounting evidence points to its dysfunction in cancer and neurological disease. Here, we provide an unencumbered, annotated list of known and candidate KEN-dependent APC/C substrates. Our data highlight the importance of APC/C in various aspects of proliferative control and points to its potentially broader impact on unanticipated cellular processes, including chromatin organization.
Identifying E3 substrates remains technically challenging. Since E3-substrate interactions exhibit low stoichiometry, mapping substrates by defining interactors is difficult. In addition, Ub ligase substrates are often in low abundance. APC/C is inhibited throughout the cell cycle by myriad mechanisms [72], and the time when APC/C binds substrates coincides with when targets are being degraded and thus at their lowest abundance. This complicates many proteomics-based approaches. Alternative techniques for identifying E3 ligase substrates, including Global Protein Stability Profiling (GPS) and in vitro expression cloning, circumvent these challenges by measuring changes in substrate stability using fluorescent reporters or metabolic labeling with radioisotopes. These represent powerful tools for mapping E3 substrates [56,73]. However, both approaches are laborious and time intensive, require significant technical

PLOS BIOLOGY
expertise, and depend on gene expression libraries, which are neither complete nor available to most laboratories. We bypass these challenges using a simple in silico approach based on publicly available information, which is simple, inexpensive, and easily repeated with different variables. While our approach shares some similarities with previous approaches, it improves upon those in its simplicity, expanded use of multiple cell cycle mRNA datasets, and inclusion of a degron motif in the search criteria [35,39,74]. Its success stems from the use of orthogonal filtering criteria, that is, unlinked features between mRNA and proteins. While our current approach was limited to substrates whose mRNAs are cell cycle regulated and proteins that contain a KEN-box degron, repeating this analysis with datasets capturing cell cycle protein dynamics, as they come available and more reliable, along with other known degrons, represents an interesting future approach. We predict that similar uses of unrelated properties could be leveraged for mapping targets of other enzymes such as kinases where defining substrates has proven similarly challenging. It is notable that degron sequences remain unknown for most Ub ligases, highlighting the importance of mechanistic studies in enabling systemslevel discoveries.

Involvement of APC/C in chromatin regulation
Determining the enzymes and substrates in kinase signaling cascades has been instrumental in determining proliferative controls in normal cells, their responses to stress and damage, and disease phenotypes and treatments. Relatedly, decoding Ub signaling pathways involved in proliferation is likely to provide insight into enzyme function in normal cell physiology as well as in disease.
A major finding of this work is that numerous chromatin regulators are controlled temporally during proliferation by APC/C. Impairing the degradation of one such substrate, UHRF1, altered the timing of cell cycle events and changed global patterns of DNA methylation. Since numerous chromatin regulators are controlled by APC/C, we anticipate widespread, pleiotropic effects on chromatin in cells where APC/C activity is impaired, either physiologically or pathologically.
Our observations raise the possibility that dysregulation of the cell cycle machinery, as is seen in diseases such as cancer, could alter the chromatin environment. The discovery that many chromatin regulators are mutated in cancer, a disease of uncontrolled proliferation, together with our data, imply a bidirectional relationship between the chromatin landscape and the cell cycle oscillator. Consistent with the notion that dysregulation of APC/C controlled proteins could play important roles in determining the chromatin environment in disease, the mRNA expression of our 145 known and putative substrates strongly predict breast cancer aneuploidies and copy number variations (S7 Fig). This observation is not due solely to the selection of specific breast cancer subtypes since our gene signature is elevated in multiple breast cancer subtypes. Interestingly, the expression of this signature correlates with the CIN70 signature, which was previously developed based on gene expression in chromosomally unstable cancers [75]. We observed an extraordinary correlation between the CIN70 and our 145 gene signature in breast cancer (S7 Fig). This is remarkable since our signature was generated completely independent of gene expression in cancer and was instead derived, in part, by short sequence motifs on proteins.
Multiple lines of in vitro and in vivo evidence support the regulation of UHRF1 by APC/C during the cell cycle. However, UHRF1 mRNA expression is also regulated during the cell cycle. In fact, cell cycle-dependent transcription was an inclusion criterion for our in silico analysis since most known APC/C substrates are dynamically expressed in cycling cells. Accordingly, it remains difficult to assess the relative contributions of mRNA expression and protein degradation to overall protein expression during cell cycle. It is also unknown if other ubiquitin ligases might control UHRF1 during cell cycle, as is the case for some APC/C substrates [76,77]. The overall contributions of mRNA expression or additional E3s to the dynamics of protein expression during cell cycle are unknown for most APC/C substrates. Others have sought to address these concerns for some APC/C substrates with live cell imaging of exogenous, transiently expressed, and fluorescently tagged proteins. For UHRF1, and many other APC/C substrates, these studies represent an area of future investigation. APC/C Cdh1 , but not APC/C Cdc20 , ubiquitylates UHRF1 in vitro. This ubiquitylation is dependent on a KEN-box motif in UHRF1. Interestingly, there are several other potential Dbox motifs in UHRF1 that could also contribute to APC/C binding, particularly since APC/C can bind D-box and KEN-box motifs simultaneously [15,63]. This UHRF1 ubiquitylation is notable because the Cdh1-bound form of APC/C is active both G1 and quiescent cells and is critical for restraining S-phase entry. Our findings suggest that impaired UHRF1 degradation promotes a premature G1/S transition. We propose that the proper degradation of UHRF1, and other chromatin regulators, serves to integrate growth factor-dependent proliferative decisions with the chromatin regulatory environment. This regulation could help explain the complex chromatin rearrangements observed in quiescent cells, where APC/C Cdh1 is active [32,33,78]. Further, APC/C controls key cell cycle transcriptional regulators, including the G2/ M transcription factor FoxM1 and the repressor E2F proteins, E2F7 and E2F8 [77,79]. Thus, our data point to a higher-order role regulatory role for APC/C in gene regulation, by controlling transcription factors (i.e., FoxM1), transcriptional repressors (i.e., E2F7, E2F8,), and chromatin modifiers. A few studies have also linked UHRF1 to DNA damage repair [80], and it is also possible that this role of UHRF1 contributes to altered cell cycles and the expression of cell cycle proteins in our assays.
Aberrant DNA methylation is a hallmark of cancer [81]. UHRF1 promotes DNA methylation maintenance, and too much or too little UHRF1 expression is detrimental to methylation stasis [26,29]. It is interesting to speculate that the redistribution of DNA methylation in disease could be caused, in part, by the aberrant stabilization of UHRF1, resulting from APC/ C Cdh1 inactivation. In the future, it will be important to determine if oncogene activation acts through the APC/C to reorganize the chromatin landscape. Furthermore, determining Ub ligase substrates, like UHRF1, that might be dysregulated in pathological settings via altered degradative mechanisms could suggest therapeutic strategies to reverse their effects.

Computational identification of putative APC/C substrates
Human proteins containing a KEN-box sequence (amino acid sequence K-E-N) were identified using the "Find a Sequence Match" feature on the Scansite web search platform (currently https://scansite4.mit.edu/4.0/#home). Proteins with cell cycle-regulated mRNA were curated from 4 independent cell cycle transcriptional studies [40][41][42][43]. The genes that scored in 2 or more of these screens were previously compiled in the supplemental data of Grant and colleagues [41]. Gene and protein name conversions were performed using the DAVID online tool (https://david.ncifcrf.gov/conversion.jsp). The overlapping set 145 proteins, which contain a KEN sequence and exhibit oscillating cell cycle-regulated mRNA expression, were identified. For all 145 proteins, we manually curated information on their alias, function, sequence flanking the KEN motif, and evidence for regulation by APC/C from various online databases and repositories, including UNIPROT, PubMed, and Genecards.
The set of 33 well-validated, KEN-containing human APC/C substrates was derived from [16]. Our own FLAG-Cdh1 IPs were compared with other APC/C substrate discovery efforts [47,48]. Singh and colleagues identified "clusters" of proteins whose levels changed at mitotic exit. For each cluster, they reported a top percentile, and for the clusters that most accurately revealed APC/C substrates [1][2][3], we compile their data in S4 Data in terms of which KENdependent substrates were identified. Their data from Cluster 1, which identified the most KEN-containing APC/C substrates, are shown in Fig 1C. Lafranchi and colleagues rank ordered proteins based on the degree of change from mitosis to G1, analyzed by proteomics. We curated their data to identify the cut-off point where the last KEN-dependent APC/C substrate was identified among their rank-ordered list. Since they provided no cut-off point, the data comparison in Fig 1C represents the best estimate of their ability to capture APC/C substrates.
GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA stable overexpression cells were generated by transducing HeLa S3, U2OS, and RPE-1-hTERT cell lines with pHAGE-GFP lentivirus that had been produced in HEK-293T cells. Infections were performed in the presence of 8-μg/mL polybrene for 48 h prior to antibiotic selection. Cells were selected for 5 to 7 days with 8 μg/mL (HeLa S3 and U2OS) or 10 μg/mL (RPE-1) Blasticidin. Lentiviral particles were produced by transfecting HEK-293T cells with Tet, VSVg, Gag/pol, and Rev viral packaging vectors together with the pHAGE-GFP lentiviral vectors using TransIT MIRUS (cat no. MIR 2700). Viral particles were collected 48 and 72 h after transfection and stored at −80˚C prior to transduction.
To generate the rescue cell lines, the U2OS and HeLa S3 stable GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA expression cell lines were transduced with previously described and validated pLKO.1 lentiviral vectors encoding either shControl or 3 0 UTR targeting shUHRF1 [66], using 8 μg/mL polybrene to aid infection. After 48 h, cells were selected with 2-μg/mL Puromycin for 3 to 5 days. Viral particles were produced by transfecting HEK-293T cells with the pLKO.1 constructs and psPAX2 and pMD2.G packaging vectors using TransIT MIRUS, collected after 48 and 72 h as mentioned previously.
Mitotic block was induced by treating 25% confluent HeLa S3 cells with 2 mM thymidine for 24 h. After washing the plates 3 to 4 times with warm media and incubating in drug-free media for 3 to 4 h, cells were treated with 100 ng/mL nocodazole for 11 h prior to harvesting by mitotic shake-off. Samples were washed 3 or 4 times with warm media, counted, and replated for indicated time points.
To synchronize cells in G1/S, HeLa S3 were plated at 20% confluency prior to addition of 2 mM thymidine. After 16 h, cells were washed 3 times with warm media and left to incubate for 8 h before the second block in 2-mM thymidine for another 16 h. Cells were washed 3 times in warm media and collected at specific time points as they progressed through the cell cycle.
To transiently inactivate the APC/C, HCT116 or U2OS cells were treated with 15 μM pro-TAME (Thermo Fisher, Waltham, Massachusetts, cat no. I-440-01M), a pan-APC/C inhibitor [82], for 90 min prior to harvest and immunoblotting. Cells had been released from nocodazole-induced mitotic block for 90 min in drug-free media prior to addition of drug.

In vivo APC/C activation assay
A total of 70% to 80% confluent U2OS cells were transfected with the indicated plasmids for 24 h and then exchanged into fresh media. Alternatively, untransfected cells were used to PLOS BIOLOGY analyze endogenous proteins. After an 8-h incubation in fresh media following transfection, cells were treated with 250-ng/mL nocodazole for 16 h. Mitotic cells were isolated by shakeoff, washed once in prewarmed media, counted, and divided equally among 15-mL conical tubes. Cells in suspension were treated with DMSO, RO-3306 (10 μM), Roscovitine (10 μM), or MG-132 (20 μM) for the indicated amount of time at 37˚C. Identical volumes of cells were removed from cell suspensions by pipetting, isolated by centrifugation, and frozen at −20˚C prior to processing for immunoblot.
Cell-free, G1 extract APC/C substrate degradation assay UHRF1 degradation in G1 phase-synchronized HeLa S3 cell extracts was performed as described in [83] Briefly, confluent HeLa S3 cells were seeded to 25% confluence (3 million cells) in 15-cm plates. Next day, cells were transfected with 20 nM of control FF or 2 independent UBE2S siRNAs [18] using Lipofectamine RNAiMAX reagent (Life Technologies) according to the manufacturer's protocol. After 8 h, 2 mM thymidine was added to the cell medium for 24 h, after which cells were washed with warm PBS once, twice with Dulbecco's Modified Eagle Medium (DMEM), and released for 4 h before treatment with 100 ng/mL of nocodazole in DMEM for 11 h. To obtain a G1 phase population, cells were washed as described previously and released for 2 h before collection. Extract preparation was performed exactly as described [84,85]. The resulting G1 extracts were mixed in a 1:1 ratio with SB buffer and supplemented with energy mix and Ub to monitor APC/C substrate degradation. Reactions were incubated at 30˚C for the indicated times, quenched with equal volume of 2× SDS sample buffer, boiled, and analyzed by SDS-PAGE and western blot.

Molecular biology
Plasmid transfection of HEK-293T, U2OS, and HCT116 was performed with either TransIT MIRUS or PolyJet (cat no. SL100688) at 1:3 or 1:4 DNA:plasmid ratio on cells with 50% to 60% confluency. After 24 h, the media was changed, and cells were expanded to larger dishes as needed. Samples were collected 24 to 48 h after siRNA transfection was performed using a 1:3 ratio of RNAi oligonucleotide to RNAiMAX (cat no. 13778-030). UHRF1 was cloned into the indicated lentiviral vectors mentioned previously using standard gateway recombination cloning. Other APC/C substrates tested for binding to Cdh1 or degradation in the APC/C activation assay were obtained from either the ORFeome collection and cloned into the indicated vectors using gateway recombination cloning or from Addgene (S6 Data) [86].

Cell lysis and immunoblotting
Cells were lysed on ice for 20 min in phosphatase lysis buffer (50 mM NaH 2 PO 4 , 150 mM NaCl, 1% Tween-20, 5% Glycerol (pH 8.0) filtered) or NETN (20 mM Tris (pH 8.0), 100 mM NaCl, 0.5 mM EDTA, 0.5% NP40) supplemented with 10 μg/mL each of aprotonin, pepstatin A, and leupeptin; 1 mM sodium orthovanadate; 1 mM NaF; and 1-mM AEBSF (4-[2 aminoethyl] benzenesulfonyl fluoride). Following incubation on ice, cell lysates were centrifuged at (20,000 × g) in a benchtop microcentrifuge at 4˚C for 20 min. Protein concentration was estimated by BCA assay (Thermofisher cat no. PI-23227) according to manufacturer's protocol. Cell extracts were diluted with SDS-PAGE Gel Loading Buffer (Laemmli Buffer) prior to analysis by SDS-PAGE. Typically, 20 to 40 μg of protein were loaded on SDS gels (either BioRad 4% to 12% Bis-Tris or homemade SDS-PAGE gels) and separated at 140 to 200 V for approximately 1 h. Proteins were transferred by wet-transfer methods onto nitrocellulose membrane, typically at 100 V for 1 h or 10 to 17 V overnight at 4˚C. Nitrocellulose membranes were then incubated with TBST (137 mM NaCl, 2.7 mM KCl, 25 mM Tris (pH 7.4), 0.5% Tween-20) supplemented with either 5% bovine serum albumin or non-fat dry milk for at least 1 h or overnight at 4˚C. Blocked membranes were incubated overnight with primary antibodies at 4˚C, washed in TBST, incubated in appropriate secondary antibodies for 1 h at room temperature, and then developed by chemiluminescence using Pierce ECL (ThermoFisher) or Clarity ECL (Bio-Rad, Hercules, California). See reagent list in the Supplement information for detailed primary and secondary antibody information.

Immunoprecipitation
For exogenous coIP experiments, cells were lysed in NETN for 20 min on ice and then centrifuged in a benchtop centrifuge on maximum speed (20,000 × g) for 20 min at 4˚C, prior to determining protein concentration by either Bradford or BCA assay. A master mix of 1 to 2 mg/mL protein concentration was calculated, 10% of which was retained as input while the remaining 90% was used for coIP. Prior to coIP, 50 μL of antibody-coated beads were prewashed with 1 mL of 1× PBST (0.1% Tween-20) 3 times and then preblocked with 1 mL PBS/ 1% BSA for 1 h at 4˚C. Clarified cell lysates were also precleared by incubation with the same volume (50 μL) of empty Protein A/G agarose beads, rotating at 4˚C for 1 h. After preblocking, beads were washed 3 times in lysis buffer, using low-speed centrifugation to collect beads. Buffers were removed using a small orifice, gel-loading tip to limit bead aspiration between washes. After preclearing the lysates, they were centrifuged at low speed to collect Protein A/G beads at the bottom of the tubes. Samples were carefully pipetted from the same tubes without disturbing the resin. Cell lysates were immunoprecipitated for 2 to 4 h at 4˚C with 50 μL of EzView M2-or Myc-antibody beads (F2426-1ML or E6654-1ML). After coIP, beads were pelleted at low-speed centrifugation, washed 3 times with wash buffer (NETN containing additional 1% Triton-X-100, no inhibitors added), and transferred to new microfuge tubes for 1 final wash with lysis buffer to remove unbound/contaminating proteins. Wash buffer was removed from beads using small orifice, gel-loading tip as above. After removal of the final wash, beads were resuspended in 50 μL of 2× SDS-PAGE Gel Loading Buffer (Laemmli Buffer) and boiled 5 to 10 min at 95˚C. Samples were removed from the beads using a 27-gauge needle to avoid bead aspiration after boiling and transferred to new microfuge tubes. Typically, 20 μL of the coIP was loaded alongside 1% of the input volume. Samples were analyzed by immunoblotting as described.
For endogenous coIP, liquid nitrogen, flash-frozen pellets previously stored at −80˚C were resuspended in phosphate lysis buffer containing protease and phosphatase inhibitors (as described above) for 20 min on ice and then centrifuged in a benchtop centrifuge on maximum speed (20,000 × g) for 20 min at 4˚C, prior to determining protein concentration by BCA assay. A master mix of 4.2 mg/mL protein concentration was calculated, 10% of which was retained as input while the remaining 90% was used for coIP. Prior to coIP, approximately 130 μL slurry of SureBeads Protein G magnetic beads/sample (Bio-Rad, #161-4023) was prewashed 3 times with 1 mL of 1× PBST (0.1% Tween-20) and 1 time with lysis buffer before incubating the beads with cell extract for 1 h, rotating at 4˚C. Cell lysates were incubated overnight, rotating at 4˚C, with 2-μg antibody/mg protein (8.4 μg) using either control mouse IgG (Santa Cruz, Dallas, Texas, #sc-2025) or mouse Cdh1 (clone DCS-266) (Santa Cruz, #sc-56312) antibodies. After coIP, magnetic beads were incubated with antibody-containing cell lysates for 1 to 2 h, rotating at 4˚C. Then, beads were centrifuged briefly and pelleted using magnetic rack. Beads were washed by gentle pipetting 3 times with phosphate lysis buffer (without any inhibitors) and then transferred to new microfuge tubes for 1 final wash to remove unbound/contaminating proteins. After aspiration of the final wash, beads were resuspended in 50-μL 2X SDS-PAGE Gel Loading Buffer (Laemmli Buffer) and boiled for 10 min at 70˚C. Samples were removed from the boiled beads with the magnetic rack and transferred to new microfuge tubes. Samples were analyzed by immunoblotting as described.

Flow cytometry cell cycle analysis
HeLa S3 GFP-UHRF1 WT and GFP-UHRF1 KEN:AAA (shUHRF1) cells were synchronized in mitosis by sequential thymidine-nocodazole treatment as described above, using 2 mM thymidine and 100 ng/mL nocodazole. After release, cells were pulsed with 10 μM EdU 30 min prior to collection at specific time points. After counting the cells, 2 million cells were retained for western blotting (WB) analysis, and 1 million cells were fixed for flow cytometry. For WB, cells we pelleted and washed once with cold PBS prior to freezing at −20˚C. For flow cytometry, cells were fixed in 4% formaldehyde/PBS for 15 min at room temperature. Cells were pelleted and resuspended in 1% BSA/PBS and stored overnight at 4˚C. In the next day, cells were pelleted and resuspended in 1% BSA/PBS/0.5% Triton X-100 for 15 min at room temperature. Cells were pelleted, resuspended with labelling solution (100 mM ascorbic acid, 1 mM CuSO 4 , 2 μM Alexa Fluor 488 azide in PBS), and incubated for 30 min in the dark at room temperature. After addition of 1% BSA/PBS/0.5% Triton X-100, cells were pelleted and stained with 1 μg/mL DAPI in 1% BSA/PBS/0.5% Triton X-100 for 1 h in the dark at room temperature. Flow cytometry was performed on an Attune Nxt Flow Cytometer (Thermo Fisher Scientific). Channel BL1 was used for Azide 488 dye. Channel VL1 was used for DAPI dye. Following acquisition, data were analyzed using FlowJo software.
For the siRNA UHRF1 depletion experiments, asynchronous U2OS cells were transfected with 3 independent siRNAs against UHRF1 for 48 h prior to a 30 min EdU pulse (10 μM). Samples were collected for flow cytometry as described above, except for 70% ethanol fixation and 4˚C overnight storage prior to staining and analysis.

Immunofluorescence imaging
HeLa cells were plated on poly-L-lysine-coated #1.5 coverslips. Next day, cells were treated with siRNA (control siFF and siUHRF1) and RNAiMax according to manufacturer's protocol (Invitrogen). After 48 h of siRNA treatments, cells were fixed in 3% paraformaldehyde in PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl 2 (pH 7.0)) for 15 min at 37˚C. Then, cells were washed with PHEM buffer and permeabilized using 0.5% of Nonidet P-40 in PHEM buffer for 15 min at room temperature. Cells were washed and then blocked with 5% BSA in PHEM. Primary antibodies used were α-CENP-C (MBL:1:1000) as a kinetochore marker and α-tubulin (Sigma, St. Louis, Missouri: 1:500). Samples were incubated in primary antibody solution for 1 h at 37˚C. All fluorescently labeled secondary antibodies (anti-mouse Alexa 488, anti-guinea pig 564) were diluted 1:200 dilution, and cells were incubated for 1 h at 37˚C. DNA was counterstained with DAPI for 15 min at room temperature after washing out secondary antibodies. All samples were mounted onto glass slides in Prolong Gold antifade (Invitrogen). For image acquisition, three-dimensional stacked images were obtained sequentially at 200 nm steps along the z axis through the cell using MetaMorph 7.8 software (Molecular Devices, San Jose, California) and a Nikon Ti-inverted microscope equipped with the spinning disc confocal head (Yokogawa, Sugar Land, Texas), the Orca-ER cooled CCD camera (Nikon, Melville, New York), and an ×100/1.4 NA PlanApo objective (Nikon).

Genomic DNA isolation for methylation analysis
Genomic DNA was isolated from Parental U2OS cells and U2OS cells overexpressing either GFP-UHRF1 WT or GFP-UHRF1 KEN:AAA . All samples groups were processed in biological triplicates. Briefly, cells were lysed overnight at 37˚C in 2 mL of TE-SDS buffer (10 mM Tris-HCl (pH 8.0), 0.1 mM EDTA, 0.5% SDS), supplemented with 100 μL of 20 mg/mL proteinase K. DNA was purified by phenol:chloroform extraction in 3 phases: [1] 100% phenol, [2] phenol: chloroform:isoamyl alcohol (25:24:1), and [3] chloroform:isoamyl alcohol (24:1). For each phase, the aqueous layer was combined with the organic layer in a 1:1 ratio. Samples were quickly shaken, allowed to sit on ice for approximately 5 min, and then separated by centrifugation at 1,693 RCF for 5 min at 4˚C. The top aqueous layer was then transferred to a new tube for the next organic phase. Following extraction, DNA was precipitated with 1/10 volume 3 M sodium acetate (pH 4.8) and 2.5 volume 100% ethanol and stored overnight at −20˚C. Precipitated DNA was pelleted by centrifugation at 17,090 RCF for 30 min at 4˚C. The pelleted DNA was washed twice with 70% ethanol, allowed to dry for 15 min, and resuspended in TE buffer (10 mM Tris-HCl (pH 8.0), 0.1 mM EDTA). Samples were then treated with 1 mg/mL RNAse A at 37˚C for 30 min and then repurified by ethanol precipitation as described above.

Infinium methylation EPIC BeadChip (EPIC array)
Genomic DNA was quantified by High Sensitivity Qubit Fluorometric Quantification (Invitrogen), and 1.5 μg of genomic DNA was submitted to the Van Andel Institute Genomics Core for quality control analysis, bisulfite conversion, and DNA methylation quantification using the Infinium Methylation EPIC BeadChIP (Illumina) processed on an Illumina iScan system following the manufacturer's standard protocol [67,68].

EPIC array data processing
All analyses were conducted in the R statistical software (Version 3.6.1) (R Core Team [92]). R script for data processing and analysis is available in S1 Text.
Raw IDAT files for each sample were processed using the Bioconductor package "SeSAMe" (Version 1.2.0) for extraction of probe signal intensity values, normalization of probe signal intensity values, and calculation of β-values from the normalized probe signal intensity values [93][94][95]. The β-value is the measure of DNA methylation for each individual CpG probe, where a minimum value of 0 indicates a fully unmethylated CpG and a maximum value of 1 indicates a fully methylated CpG in the population. CpG probes with a detection p-value of >0.05 in any one sample were excluded from the analysis.

Genomic and replication timing annotation
CpG probes were mapped to their genomic coordinate (hg38) and were then annotated to their genomic annotation relationship (promoter-TSS, exon, etc.) using HOMER (Version 4.10.3) [96].
Repli-seq data for U2OS cells used for determining CpG probe localization relative to replication timing was generated by Dr. David Gilbert's lab (Florida State University) as part of the 4D Nucleome project (Experiment #4DNEXWNB33S2) [69]. Genomic regions were considered early-replicating if the replication timing value was >0 and late-replicating if <0. CpG probes were annotated for replication timing domains by intersecting the Repli-seq genomic coordinates with CpG probe coordinates using BEDTools (Version 2.16.2) [97].

Identification of differentially methylated CpG probes
The Bioconductor package "limma" (Version 3.40.6) was used to determine differential methylation among sample groups and perform MDS analysis [94,95,98]. For statistical testing of significance, β-values were logit transformed to M-values: . M-values were then used for standard limma workflow contrasts to determine differential methylation of U2OS GFP-UHRF1 WT or GFP-UHRF1 KEN:AAA overexpression to Parental U2OS cells [98,99]. CpG probes with an adjusted p-value of �0.05 were considered significant, and log fold-change of the M-value was used to determine hypermethylation (logFC > 0) or hypomethylation (logFC < 0) relative to U2OS parental cells.

Enrichment bias calculation and hypergeometric distribution testing
Enrichment bias calculations were done by first determining the following values for each feature (e.g., genomic annotation, replication timing): q = Number of CpGs that are differentially methylated in feature (e.g., exon) m = Total number of CpGs on the EPIC array that match feature (e.g., exon) n = Total number CpGs on the EPIC array that do not match feature (e.g., everything that is not an exon) k = Total number of all differentially methylated CpGs Next, the expected number of CpGs that would be differentially methylated in that feature by random chance was determined with the following equation: Finally, percentage enrichment bias was calculated with the following equation: where positive or negative enrichment values indicate more or less enrichment for a feature than would be expected by random chance, respectively. Hypergeometric distribution testing for determining significance of enrichment bias was performed using the phyper() function in R with the following values: q, m, n, k.

Data access
EPIC array data can be found under GEO Accession # GSE137913.

Signature evaluation in TCGA BRCA samples
Upper quartile normalized RSEM gene expression data for TCGA BRCA (n = 1201) was downloaded from the GDC legacy archive (https://portal.gdc.cancer.gov). The data was log2 transformed and median centered. To determine the per sample UB signature score, the samples were ranked by the median expression of the 145 UB gene signature. Samples were then divided at the median and grouped as high or low based on rank. Copy number burden, aneuploidy, and homologous recombination deficiency data were extracted from Thorsoon and colleagues [100] and plotted by UB signature group and PAM50 subtype [101]. Significance was calculated by t test. The CIN70 score was determined as previously described in Fan and colleagues [102]. The CIN70 was plotted against the UB, colored by PAM50 subtype, and r 2 and Pearson correlation were calculated. All analysis was performed in R (v3.5.2). 3) and then neutralized with Tris 1M (pH 10.0). The total eluted protein was reduced (5 mM DTT) and alkylated using iodoacetamide (1.25 mM) for 30 min in the dark. The resultant protein was then digested overnight with sequencing grade trypsin (Promega). The trypsin:protein ratio was maintained at 1:100. Total peptides were purified on Pierce C18 spin columns (Cat 89870) using the manufacturer's protocol. Peptides were eluted using 70% acetonitrile and 0.1% TFA solution in 50 μL volumes twice, dried on a SpeedVac at room temperature, and processed by mass spectrometry proteomic analysis.

Mass spectrometry
Peptides were separated by reversed-phase nano-high-performance liquid chromatography using a nanoAquity UPLC system (Waters Corp., Milford, Massachusetts). Peptides were first trapped in a 2-cm trapping column (Acclaim PepMap 100, C18 beads of 3.0 μm particle size, 100 Å pore size) and a 25-cm EASY-spray analytical column (75 μm inner diameter, C18 beads of 2.0 μm particle size, 100 Å pore size) at 35˚C. The flow rate was 250 nL/minute over a gradient of 1% buffer B (0.1% formic acid in acetonitrile) to 30% buffer B in 150 min, and an in-line Orbitrap Elite mass spectrometer (Thermo Scientific) performed mass spectral analysis. The ion source was operated at 2.6 kV with the ion transfer tube temperature set at 300˚C. A full MS scan (300 to 2000 m/z) was acquired in Orbitrap with a 120,000 resolution setting, and data-dependent MS2 spectra were acquired in the linear ion trap by collision-induced dissociation using a 2.0 m/z wide isolation window on the 15 most intense ions. Precursor ions were selected based on charge states (+2, +3) and intensity thresholds (above 1e5) from the full scan; dynamic exclusion (one repeat during 30 s, a 60-s exclusion time window) was also used. The polysiloxane lock mass of 445.120030 was used throughout spectral acquisition.