Topology and enzymatic properties of a canonical Polycomb repressive complex 1 isoform

Polycomb repressive complex 1 (PRC1) catalyses monoubiquitination of histone H2A on Lys119, promoting gene silencing. Cells at different developmental stages and in different tissues express different PRC1 isoforms. The topology, subunit composition, structural architecture and molecular mechanism of most of these isoforms are still poorly characterized. Here, we have purified a PRC1 isoform comprising subunits RING1B, PCGF2, CBX2 and PHC2, two stable subcomplexes (RING1B‐PCGF2 and RING1B‐PHC2) and the catalytic subunit RING1B in isolation. By crosslinking mass spectrometry, we identified novel interactions between RING1B and the three non‐catalytic subunits. Biochemical, biophysical, and enzymatic data suggest that CBX2 and PHC2 play a structural role, whereas PCGF2 also modulates catalysis. Our data offer insights into the molecular architecture of PRC1 and its histone ubiquitination activity.

Human PRC1 is classified into canonical (cPRC1) and variant (vPRC1) isoforms. Canonical PRC1 monoubiquitinates H2AK119 at genomic loci where Polycomb repressive complex 2 (PRC2) has deposited H3K27 trimethylation marks, while vPRC1 modifies chromatin independent of PRC2 [4]. Here, cPRC1 comprises different isoforms, which are expressed at various stages of cell differentiation and in different tissues [5]. All cPRC1 complexes are formed by a heterodimer composed of subunits RING1B and PCGF2 (aka Mel18) or PCGF4 (aka BMI1). This heterodimer associates with one CBX subunit orthologue, which contains a chromodomain recognising the H3K27me3 mark posed by PRC2 [6,7], and with one PHC subunit orthologue, which promotes PRC1 self-association or interaction with other protein partners through its SAM domain [8][9][10]. The RING domain of RING1B is responsible for the E3 ligase activity and it is stimulated by the RING domain of PCGF2/4 [9,11]. Instead, CBX is responsible for inducing chromatin compaction via a non-enzymatic mechanism [6,7,12]. The Pc box domain of CBX7, conserved among CBX subunits, has been crystallised in complex with the RAWUL domain of RING1B [13]. Moreover, the HD1 motif of PHC2 has been crystallised in complex with the RAWUL domain of PCGF4 [14]. The structure of the heterodimer of the RING domains of RING1B and PCGF4 in complex with mononucleosomes is also available [15]. However, the molecular mechanism of RING1B stimulation is not well understood, partly because there is a lack of biochemical, biophysical and structural evidence on the molecular organisation and topology of PRC1 (Fig. 1). Specifically, the role of CBX and PHC subunits in complex assembly and in catalysis is unclear.
To improve understanding of the molecular properties of cPRC1, here we have purified a cPRC1 isoform, two stable subcomplexes, and the catalytic subunit RING1B in isolation. We report the topological map of this cPRC1 isoform obtained by crosslinking mass spectrometry (XL-MS). Based on enzymatic, biochemical and SAXS data on this isoform and comparison to its subcomplexes, we discuss the role played by each noncatalytic subunit. We attribute a structural, but not an enzymatic role to CBX and PHC. Moreover, we observe that PCGF stimulates RING1B catalysis and we suggest that such stimulation is partly due to the fact that PCGF increases the affinity of RING1B for the nucleosomes and reduces the affinity of RING1B for the E2 enzyme thus increasing the E2 enzyme turnover.

Materials and methods
Cloning, expression and purification  [16]. Coding sequences of individual subunits were cloned into acceptor (p-ACEBAC-1) and donor (pIDC, pIDS and pIDK) vectors of the MultiBac system by sequence-and-ligation-independent cloning and fused by in vitro Cre-loxP recombination to yield a single plasmid with multiple expression cassettes (primers are listed in Table 1) [17]. The presence of the gene encoding each subunit in the corresponding construct was verified by restriction enzyme digestion. The presence of an in-frame insert was verified by DNA sequencing. Recombinant baculovirus was produced as previously described [18] and used to infect Sf21 insect cells at a cell density of 1.0 9 10 6 per mL in SF900 medium. Cells were collected 72-96 h after proliferation arrest by centrifugation at 1000 g for 15 min and stored at À20°C. Each PRC1 complex from 2 to 3 L pellet was resuspended in 200 mL lysis buffer (HEPES 50 mM pH 8, NaCl 150 mM, 1 mM DTT, 0.1% NP40, 1 mM Leupeptine and 1 mM Pepstatine) by vortexing. The mixture was sonicated on ice for 8 min at 35% intensity using a Sonics VCX-750 Vibra Cell Sonicator (Sonics & Materials Inc., Newtown, CT, USA). The lysate was centrifuged 1 h at 38 500 g at 4°C. The supernatant was loaded onto 7 mL pre-equilibrated beads of strep-tactin resin and the flow-through was collected by gravity flow. The resin was washed with 13 CV of washing buffer (HEPES 50 mM pH 8, NaCl 150 mM, 1 mM DTT). The PRC1 complexes were eluted using 6-7 CV of elution buffer (HEPES 50 mM pH 8, NaCl 150 mM, 1 mM DTT, 5 mM desthiobiotin). PRC1 complexes were loaded onto an S200 16/600 column (GE Healthcare Europe GmbH, Velizy-Villacoublay, France) pre-equilibrated in washing buffer. Fractions containing the targets were pooled and concentrated using a 15 mL Amicon 50 kDa molecular weight cut off and injected onto an S200 10/300 column (GE Healthcare). Fractions containing the targets after this second size exclusion chromatographic step were pooled and used for subsequent experiments.
Peptide mass fingerprinting mass spectrometry DPRC1.2 and RING1B-DPHC2 complexes were separated by SDS/PAGE following staining with Coomassie Brilliant Blue G250 [0.4% (w/v), 10% (w/v) citric acid, 8% (w/v) ammonium sulphate, 20% (v/v) methanol]. Coomassiestained bands were excised, chopped into small pieces and Where structures are available for homologous domains but not for the specific subunit used in this work, the PDB id number is indicated in italic. Based on the available structures, 70% of the cPRC1 structure is still unknown. transferred to 0.5 mL Eppendorf tubes. For all following steps, buffers were exchanged by two consecutive 15 min incubation steps of the gel pieces with 200 lL of acetonitrile (ACN) whereby the ACN was removed after each step. Proteins were reduced by the addition of 200 lL of a 10 mM DTT solution in 100 mM ammonium bicarbonate (AmBic) and incubated at 56°C for 30 min. Proteins were alkylated by the addition of 200 lL of a 55 mM iodoacetamide solution in 100 mM AmBic and incubated for 20 min in the dark. Fifty microlitre of trypsin at 1 ngÁlL À1 were added to the gel pieces, incubated for 30 min on ice and then overnight at 37°C. Gel pieces were sonicated for 15 min, spun down and the supernatant was transferred into a glass vial. Remaining gel pieces were washed with 50 lL of an aqueous solution of 50% ACN and 1% formic acid and sonicated for 15 min. The combined supernatants were dried in a Speedvac rotary evaporator and reconstituted in 10 lL of an aqueous solution of 0.1% (v/v) formic acid. Peptides were separated using the nanoAcquity UPLC system with a nanoAcquity trapping and analytical column, which was coupled to an LTQ Orbitrap Velos (Thermo Fisher Scientific, Waltham, MA, USA) using the Proxeon nanospray source. Full scan MS spectra with a mass range of 300-1700 m/Z were acquired in profile mode with a resolution of 30.000 and a filling time of 500 ms applying a limit of 10 6 ions. The 15 most intense ions were fragmented in the LTQ using a normalised collision energy of 40%. 3 9 10 4 ions were selected within 100 ms and their fragmentation was achieved upon accumulation of selected precursor ions. MS/MS data were acquired in centroid mode of multiple charged (2+, 3+, 4+) precursor ions. The dynamic exclusion list was restricted to 500 entries with a maximum retention period of 30 s and relative mass window of 10 p.p.m. In order to improve the mass accuracy, a lock mass correction using a background ion (m/Z 445.12003) was applied. Acquired data were processed using ISOBAR-QUANT [19] and MASCOT (v2.2.07) (Matrix Science, Boston, MA, USA) using a reversed Uniprot Homo sapiens database (UP000005640) including common contaminants. The following modifications were taken into account: carbamidomethyl (C) (fixed modification), acetyl (N-term) and oxidation (M) (variable modifications). The mass error tolerance for full scan MS spectra was set to 10 p.p.m. and for MS/MS spectra to 0.02 Da. A maximum of two missed cleavages were allowed. A minimum of two unique peptides with a peptide length of at least seven amino acids and a false discovery rate below 0.01 were required on the peptide and protein level to consider the result significant.

Crosslinking mass spectrometry
About 50 lg of purified DPRC1.2, RING1B-PCGF2 and RING1B-DPHC2 complexes were individually crosslinked by addition of 5 lL at 50 mM of an iso-stoichiometric Table 1. Primers used to produce the constructs of this study.

Subunit or vector
Forward primer

Microscale thermophoresis
Microscale thermophoresis was used to determine the K d between UbcH5c E2 enzyme and RING1B/RING1B-PCGF2. UbcH5c was expressed in Escherichia coli Rosetta (DE3) overnight at 18°C. Cells were harvested by centrifugation at 5000 g for 15 min at 4°C. Two litre cell culture was resuspended in TrisHCl 50 mM pH 7.5 at 4°C, NaCl 150 mM, 1 mM DTT, and anti-protease Complete EDTA-free tablets (Roche, Basel, Switzerland). UbcH5c was sonicated 5 min on ice and centrifuged at 27 500 g for 1 h.

Size exclusion chromatography-small angle X-ray scattering
Size exclusion chromatography (SEC) small angle X-ray scattering (SAXS) experiments were performed on the BM29 beamline at the European Synchrotron Radiation Facility (ESRF, Grenoble, France). An online HPLC system was attached directly to the sample-inlet valve of the beamline sample changer. Fifty microlitre of DPRC1.2 at 2.9 mgÁmL À1 and 50 lL at 3.6 mgÁmL À1 of RING1B-PCGF2 were manually injected on an S200 15/150 column, respectively. The column was pre-equilibrated with buffer HEPES 50 mM pH 8, NaCl 150 mM, DTT 1 mM. Buffers were degassed and a flow rate of 0.2 mLÁmin À1 at 4°C was used for all sample runs. Prior to each run, the column was equilibrated with 2 CV of buffer and the baseline was monitored. All data from the run were collected at a wavelength k = 0.99 A using a sample-to-detector (PILATUS 1M; Dectris AG, Baden, Switzerland) distance of 2.87 m corresponding to a q-range of 0.0035-0.167 A À1 where q is the momentum transfer (q = 4pk sinh) and 2h the scattering angle. Approximately 900 frames with an exposure time of 1 s per frame were collected per sample run. 100 initial frames were averaged to create the reference buffer and the frames collected from each elution peak (40 frames/peak for both DPRC1.2 and RING1B-PCGF2), corresponding to the scattering of an individual purified species, were also averaged and subtracted from the reference buffer using the program PRIMUS [24]. Radii of gyration (R g ) and pairwise distance distribution functions [P(r)] were extracted based on the Guinier approximation.  [25]. Membranes were washed with TBST. We incubated the membranes 2 h with the secondary antibodies anti-mouse (1 : 5000, A11002; Thermo Fisher Scientific, for anti-H2AUb) conjugated with Alexafluor dye 532 nm and anti-rabbit (1 : 5000, A32731; Thermofisher, for anti-H2A) conjugated with Alexafluor dye 488 nm. The membranes were washed and the fluorescent signal from the membranes was recorded with a Typhoon trio scanner. The bands were quantified with Quantity One (BioRad). The H2A monoubiquitination activity was calculated as the ratio between ubiquitinated and total H2A and plotted over the concentrations of PRC1 complexes. Data were analysed using GRAPHPAD (GraphPad Software).

H2A monoubiquitination activity assay
E2-discharging assay (single turnover monoubiquitination assay) E2-discharging assays were performed as described [26,27]. Briefly, a 39-concentrated master mix containing Na-HEPES 50 mM pH 7.7, 90 nM E1 enzyme, 1.2 lM E2 enzyme, 6 mM ATP, 30 lM ZnSO 4 , 15 mM MgCl 2 , and 23 lM methylated ubiquitin was pre-heated at 37°C for 1 h to charge the E2 enzyme with ubiquitin (Fig. S6A). The mix was then supplemented with 4.5 UÁmL À1 apyrase (NEB # M0398S) to deplete ATP and incubated at 37°C for another hour. RING1B and RING1B-PCGF2 were then prepared at 3 lM in Na-HEPES 50 mM pH 8 and NaCl 75 mM, and Cy5-labelled nucleosomes were prepared at 0.3 lM in TE buffer with 150 mM NaCl. Apyrase-treated master mix, Cy5-labelled nucleosomes, and the relevant PRC1 subcomplex were then mixed in a 1 : 1 : 1 ratio in a total volume of 165 lL. The reaction was incubated at 37°C. 15 lL were harvested at the indicated time points (see Fig. 5D and Fig. S6) and quenched by addition of 5 lL SDS-containing gel loading buffer and boiling at 95°C for 5 min. Samples were then analysed by SDS/PAGE and visualised with a ChemiDoc Imaging System (BioRad), before Coomassie blue staining ( Fig. 5D and Fig. S6). The bands were quantified with Quantity One (BioRad). The H2A monoubiquitination activity was calculated as the ratio between ubiquitinated and total H2A. Data points were analysed using GRAPHPAD (GraphPad Software).

Biochemical characterisation of human cPRC1 complexes
Based on reported proteomic studies [9,28], we assembled canonical PRC1.2 (RING1B, PCGF2, CBX2 and PHC2) and PRC1.4 (RING1B, PCGF4, CBX2 and PHC2) complexes to check if their subunits can be coexpressed as complexes in heterologous expression systems and purified. We reconstituted both complexes using the MultiBac technology and we expressed them in Sf21 insect cells. We used strep-tagged RING1B as bait for purification. We observed that all the subunits were pulled-down, indicating that CBX2 and PHC2 can interact with the heterodimer formed by RING1B-PCGF2/4, in agreement with proteomic data (Fig. S1) [9,28]. However, subsequent purification steps of PRC1 complexes by gel filtration or ion exchange chromatography showed that PHC2 and CBX2 tend to induce aggregation of the complexes into higher order oligomers.
In human and mammalian cells, a shorter version of PHC2 (DPHC2), missing the first 535 amino acids, is preferentially expressed over full length PHC2 [29,30] and it associates with other PRC1 subunits [28]. Moreover, the L307R mutation in the SAM domain of DPHC2 reduces its polymerisation propensity [10]. Thus, we assembled L307R-DPHC2 with RING1B, PCGF2/4 and CBX2 in MultiBac. We could purify the complex formed by RING1B, PCGF2, DPHC2 and CBX2 (DPRC1.2) in a homogeneous form ( Fig. 2 and Table 2). Besides DPRC1.2, we could also produce and purify to homogeneity two stable subcomplexes of this isoform, namely the RING1B-PCGF2 and  Table 3).

Full-length PRC1 complex is more compact than RING1B-PCGF2 heterodimer
To obtain information on the size and shape of the DPRC1.2 complex we performed SEC-SAXS measurements. From the Guinier plot we calculated an R g of 4.4 nm, while the pair distribution function [P(r)] indicates a D max of 18 nm. The normalised Kratky plot shows a shift from the theoretical peak value expected for a globular protein, suggesting that DPRC1.2 displays regions of flexibility (Fig. 3), as expected from secondary structure prediction of its subunits (Fig. 1).
To understand how DPHC2 and CBX2 affect the shape and flexibility of DPRC1.2, we collected an SEC-SAXS dataset for the RING1B-PCGF2 subcomplex. Interestingly, data analysis shows that RING1B-PCGF2 has R g of 4.7 nm and a D max of 17 nm. Thus, the heterodimer RING1B-PCGF2 has similar R g and D max compared to DPRC1.2, suggesting that RING1B-PCGF2 adopts a more relaxed conformation in the absence of CBX2 and DPHC2 (Fig. 3).

The FCS zinc finger domain of DPHC2 interacts with RING domain of RING1B
Having established that CBX2 and DPHC2 are important for compaction of DPRC1.2, we mapped the inter-subunits interactions of DPRC1.2 by crosslinking mass spectrometry. We used DSS as a crosslinker agent for lysine residues at a C a -C a maximum distance of 27 Å. Our data recapitulate inter-subunit interactions known from available crystal structures of PRC1 domains [11,13,14]. For instance, our data show that RING1B and PCGF2 interact through their RING and RAWUL domains. Interestingly, we could also   (Fig. 4). We could confirm this novel interaction between the FCS domain of DPHC2 and the RING domain of RING1B by performing XL-MS also on the purified RING1B-DPHC2 heterodimer (Fig. 4).
To understand if any rearrangement of RING1B and PCGF2 occurs when DPHC2 and CBX2 are absent, we also performed crosslinking mass spectrometry on the isolated RING1B-PCGF2 subcomplex. In the isolated heterodimer, RING1B is forming interactions with the RAWUL domain of PCGF2 similar to DPRC1.2 complex. In DPRC1.2, the RAWUL domain of PCGF2 is also interacting with DPHC2 (Fig. 4). This observation is in agreement with the reported structure of the RAWUL domain of PCGF4 and the partial HD1 domain of DPHC2 [14].

DPHC2 and CBX2 subunits do not affect H2A ubiquitination on mononucleosomes
Having established that the catalytic RING domain of RING1B interacts with or is in close proximity to motifs of DPHC2 and CBX2, we asked if these latter two subunits can modulate the enzymatic activity of DPRC1.2. To address this question, we measured the E3-ligase activity of DPRC1.2 and compared it with the activity of its subcomplexes. We measured the E3 ligase activity of DPRC1.2 and its subcomplexes on mononucleosomes by quantification of western blot bands using specific antibodies against free histone H2A and ubiquitinated histone H2A. We observed that RING1B is poorly active, as previously reported [8]. Coupling RING1B to DPHC2 (RING1B-DPHC2 heterodimer) does not improve activity, whereas PCGF2 (RING1B-PCGF2 heterodimer) boosts RING1B activity substantially (Fig. 5A,B, Figs S2 and S5). Finally, coupling DPHC2 and CBX2 to the RING1B-PCGF2 heterodimer (DPRC1.2 complex) does not improve activity further, i.e. the DPRC1.2 complex and the RING1B-PCGF2 heterodimer display similar activity (Fig. 5A,B, Figs S2 and S5). These data suggest that CBX2 and DPHC2 do not stimulate the enzymatic activity of PRC1.

PCGF2 activates RING1B by reducing its affinity for the E2 enzyme
Having established a structural rather than functional role for CBX2 and DPHC2 within DPRC1.2, we set out to address the mechanism by which PCGF2 increases DPRC1.2 ubiquitination activity. PCGF2 may affect binding of RING1B to its two substrates, namely the nucleosome, as previously proposed [15], or the E2 enzyme, as previously proposed for PRC1 [31,32] and for other ubiquitin ligases such as APC/C [33]. To test these hypotheses, we performed two sets of assays. First, we measured the affinity of RING1B and RING1B-PCGF2 for UbcH5c by microscale thermophoresis (MST, Fig. 5C). We expressed and purified UbcH5c (Fig. S4), labelled it on Cys85 with the fluorescent dye NT-547 and measured its affinity to RING1B and RING1B-PCGF2. The K d for RING1B-PCGF2 is 4.1 AE 1 lM, similar to the 7 lM value reported for a minimal RING1B-PCGF2 complex encompassing only the two RING domains [32]. By contrast, RING1B exhibits a higher affinity for UbcH5c (K d = 0.23 AE 0.08 lM). Second, we performed E2-discharging assays to compare single-turnover kinetic parameters of RING1B and RING1B-PCGF2 ( Fig. 5D and Fig. S6). In this assay, RING1B-PCGF2 is active (v max = 0.007 AE 0.001 min À1 ), whereas RING1B is inactive.

Discussion
In this work, we report the topological mapping by XL-MS of a canonical PRC1 isoform and of two subcomplexes, and we explore the role of the non-catalytic subunits of this complex in regulating the biochemical and enzymatic properties of PRC1.
Our PRC1 inter-subunit interaction map shows a novel interaction between the catalytic RING domain of RING1B and the FCS zinc finger domain of DPHC2, a region currently not covered by available crystal structures (Fig. 1). Our biochemical data also show that RING1B and DPHC2 can associate independently from the other subunits (Fig. 2). The FCS zinc finger domain of DPHC2 is conserved in PHC1 and PHC3, suggesting that these PHC orthologues may interact with RING1B in a similar manner as PHC2.
Additionally, our XL-MS map of DPRC1.2 shows a new interaction between the RAWUL domain of PCGF2 and the SAM domain of DPHC2 (Fig. 4).
Recently, a complex between the HD1 domain of DPHC2 and the RAWUL domain of PCGF4 was crystallised [14]. The residues of PCGF4 interacting with DPHC2 are conserved in PCGF2, suggesting that such interaction is maintained between DPHC2 and PCGF2. The HD1 domain of DPHC2 is close to its FCS zinc finger domain (Fig. 1). This latter domain is interacting with the RING domain of RING1B. By combining our XL-MS data with available crystal structures, it emerges that RING1B, DPHC2 and PCGF2 are spatially close to each other. Interestingly, CBX2 is less tightly connected to the rest of the complex, but it does come in close proximity to the catalytic module via its Pc box domain, which interacts with all the other subunits. An interaction between the Pc box of an orthologue of CBX2 (CBX7) with the RAWUL domain of RING1B had been captured previously [13]. Such close proximity of DPHC2 and the Pc box of CBX2 to the catalytic RING1B-PCGF2 heterodimer, along with our SAXS data showing similar dimensions for DPRC1.2 and RING1B-PCGF2, suggest a structural role for the DPHC2 and CBX2 subunits in supporting the architectural organisation of PRC1.
By contrast, despite their important structural role, DPHC2 and CBX2 do not affect the H2AK119 monoubiquitination activity of RING1B-PCGF2 (Fig. 5A,B, Figs S2 and S5). The limited impact of CBX2 on the cPRC1 activity is in agreement with previous data [25,34], while the role of DPHC2 in catalysis was unknown. Moreover, it is worth noticing that vPRC1, in which the RYBP/YAF2 subunits replace CBX and PHC, have greatly enhanced catalytic activity with respect to the RING1B-PCGF heterodimer, as demonstrated by ChIP-seq experiments showing a correlation between high levels of H2AK119Ub in gene loci and RYBP localisation [9]. The RYBP stimulation of the monoubiquitination activity of RING1B-PCGF was observed also in vitro [4,9,25,34]. Such lack of effect on catalysis does not necessarily mean that DPHC2 and CBX2 do not play any functional role in PRC1. The positively charged low complexity region of CBX2 is known to be responsible for inducing chromatin compaction via a non-enzymatic mechanism [7,12]. In this respect, it is interesting to notice that such region does not interact with the other subunits of our PRC1 isoform. Moreover, it has been reported that DPHC2 induces clustering of cPRC1 at specific chromatin loci [10], an activity that could be carried out via modulation of self-association by the SAM domain of PHC2 [10,35].
Finally, our enzymatic and biochemical data provide insights into the role played by the PCGF subunit in modulating nucleosome binding and ubiquitination activity of the PRC1 catalytic subunit RING1B. Such role of PCGF remains still largely unclear, despite the structure of the RING domains of RING1B and PCGF4 bound to mononucleosomes being solved [15]. Previous reports suggest that PCGFs may enhance RING1B activity at various steps of catalysis, i.e. by modulating recognition of the substrates or inducing allosteric modulation of the RING1B active site. For instance, certain PCGF4 mutations at the nucleosome interface (i.e. R64A) cause both a 10-fold decrease in nucleosome affinity and a 2-fold decrease in activity. However, some mutations at the same interface (i.e. K62A) have a more limited effect on both affinity and activity (K d-K62A = 0.37 lM vs K d-WT = 0.23 lM, and 80% of wild type activity preserved), while others even abolish activity but increase nucleosome affinity (i.e. E33A, K d-E33A = 0.09 lM) [15]. Additionally, key residues far from the nucleosome interface but close to the active site, namely K73 and D77, which are conserved both in PCGF2/4 (canonical PRC1) and in PCGF1/3/5/6 (non-canonical PRC1), ensure the correct orientation of ubiquitin for the reaction and their mutation results in a lower intrinsic E3 ligase activity [32]. Yet, all these studies have been performed using minimal catalytic PRC1 modules formed by the RING domains of RING1B and PCGFs. For full-length complexes, which are more difficult to produce in large quantities and homogeneous conformations, data is sparser. Here, we analysed two aspects of the PCGF-RING1B interaction using full length PCGF2 and RING1B. First, by single turnover E2-discharging assays we found that PCGF2 activates RING1B, likely by increasing its affinity to the substrate nucleosomes (Fig. 5D). Second, by microscale thermophoresis, we found that PCGF2 decreases the affinity of RING1B to the E2 enzyme [K d (RING1B) = 0.23 AE 0.08 lM, K d (RING1B-PCGF2) = 4.1 AE 1.0 lM], probably preventing that a too tight RING1B-E2 interaction inhibits catalytic turnover (Fig. 5C). At the molecular level, this difference in affinity could possibly be explained with the fact that in RING1B, the E2 enzyme may interact with residues outside the RING domain (i.e. the Cterminal RAWUL domain), which would become inaccessible when PCGF2 binds. Independent of how the interaction actually takes place, our results, along with previous work showing that fusion of E2 to the catalytic PRC1 core increases nucleosome affinity [15], suggest that a well-regulated interplay between the catalytic module of PRC1 and the two substrates, E2 and nucleosomes, are essential to regulate catalysis.
In summary, our analysis of the canonical PRC1 isoform DPRC1.2 provides a topological map of this important chromatin remodelling complex, revealing novel interactions between regions of currently unavailable high-resolution 3D structures, and it suggests specific structural and functional roles for the non-catalytic subunits PCGF, CBX and PHC. EMBL) with support from FRISBI (ANR-10-INSB-05-02) and GRAL (ANR-10-LABX-49-01) within the Grenoble Partnership for Structural Biology (PSB). MC was funded by the EI3POD postdoctoral programme (EMBL/EU Marie Curie Actions Cofund).
Author contributions MM designed the study; MC and OP performed the experiments; MC, OP and MM acquired and analysed the data; MM and MC wrote the manuscript; MM obtained funding and supervised the research.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.    Fig. S4. Purification of the UbcH5c E2 enzyme. Fig. S5. Representative SDS/PAGE gels used for quantification of the H2A monoubiquitination activity of DPRC1.2 (A) and RING1B-PCGF2 (B) using Cy5labelled nucleosomes (quantification reported in Fig. 5B). Fig. S6. E2-discharging assays.