A consensus binding motif for the PP4 protein phosphatase

SUMMARY

Dynamic protein phosphorylation constitutes a fundamental regulatory mechanism in all organisms. Phosphoprotein phosphatase 4 (PP4) is a conserved and essential nuclear serine/threonine phosphatase. Despite the importance of PP4, general principles of substrate selection are unknown, hampering the study of signal regulation by this phosphatase. Here, we identify and thoroughly characterize a general PP4 consensus binding motif, the FxxP motif. X-ray crystallography studies reveal that FxxP motifs bind to a conserved pocket in the PP4 regulatory subunit PPP4R3. Systems-wide in silico searches integrated with proteomic analysis of PP4 interacting proteins allow us to identify numerous FxxP motifs in proteins controlling a range of fundamental cellular processes. We identify an FxxP motif in the cohesin release factor WAPL and show that this regulates WAPL phosphorylation status and is required for efficient cohesin release. Collectively our work uncovers basic principles of PP4 specificity with broad implications for understanding phosphorylation-mediated signaling in cells.

In Brief

The mechanism of substrate recognition by the nuclear serine/threonine protein phosphatase 4 (PP4) is unknown. Ueki, Kruse et al. identify and validate a consensus PP4 binding motif, the FxxP motif, that regulates fundamental nuclear processes.

Graphical Abstract

INTRODUCTION

Phosphoprotein phosphatases (PPPs) mediate serine and threonine dephosphorylation of the majority of proteins to regulate signaling processes (Heroes et al., 2013; Nasa and Kettenbach, 2018; Nilsson, 2019; Shi, 2009). Despite their fundamental importance the principles of how PPPs selects their substrates are in many instances poorly understood. For a number of PPPs consensus-binding motifs have been identified that mediate specificity by recruiting the PPP directly to its substrate or to a substrate specifying binding partner (Cundell et al., 2016; Hendrickx et al., 2009; Hertz et al., 2016; Roy et al., 2007; X. Wang et al., 2016; Wu et al., 2017). The identified PPP consensus-binding motifs represent examples of short linear motifs (SLiMs) that in general are characterized by 2–3 core determinants within a stretch of 10 amino acids present in intrinsically disordered regions of the interacting proteins (Davey et al., 2015; Tompa et al., 2014). The identification of these motifs allow for precise manipulation of PPP interactions to dissect their function in specific pathways.

Phosphoprotein phosphatase 4 (PP4) is a nuclear and chromatin-associated phosphatase that has been implicated in regulating numerous nuclear processes (Chowdhury et al., 2008; Isono et al., 2017; Lee et al., 2012; Nakada et al., 2008; Oler and Cairns, 2012; Su et al., 2017; Toyo-oka et al., 2008; S. Wang et al., 2019; Youn et al., 2018). How PP4 controls signaling in these diverse processes is unknown as we do not know how PP4 selects its substrates. The conserved holoenzyme is a trimeric complex consisting of the catalytic subunit (PP4C) and the regulatory subunits 2 and 3 (PPP4R2 and PPP4R3)(Cohen et al., 2005; Gingras et al., 2005)(Figure 1A–B). In humans, there are two major isoforms of PPP4R3 (PPP4R3A/B also known as SMEK1/2).

Figure 1. Identification and characterization of a PP4 binding motif.

A-B) Schematic of the PP4 holoenzyme complex composed of PP4C-PPP4R2-PPP4R3. C) Venn diagram summarizing ProP-PD selections. D-E) Average peptide sequences selected by ProP-PD using PPP4R3A (D), PPP4R3B (E) as baits and a human disorderome library. F) Purification of YFP-4X(FxxP) and YFP-4X(AxxA) from stable inducible HeLa cells and western blot analysis of the purifications with the indicated antibodies. G) Purifications as in F) but analyzed by label free quantitative mass spectrometry. Each identified protein is indicated by a dot and PP4 holoenzyme components are highlighted in red and proteasome subunits in green. H) Purifications of YFP-PPP4R3B and its interaction to a FxxP containing protein (MELK) and MxPP containing protein (MCE1) analyzed in the presence of RFP-tagged 4X(FxxP) or 4X(AxxA). I) Coomassie stained SDS-PAGE of the purified PP4 holoenzyme. J) Dephosphorylation of a substrate containing a FxxP motif or a AxxA motif by the PP4 holoenzyme complex. An engineered substrate containing 3TP sites was phosphorylated with radioactive ATP and incubated with the PP4 holoenzyme. Removal of radioactive phosphate was monitored over time. Representative of 2 independent experiments with data points from both experiments shown. K) Analysis of the phosphothreonine and phosphoserine preference of the PP4 holoenzyme using synthetic model peptides and monitoring phosphate release. The mean and SD are indicated by red bars. Significance was determined using an unpaired t-test. See also Figure S1, Table S1 and Table S3.

The PPP4R3 regulatory subunits contain a conserved N-terminal Enabled/VASP Homology 1 (EVH1) domain. EVH1 domains are found in numerous proteins and bind to proline-rich sequences that adopt a lefthanded PPII polyproline helix (Ball et al., 2002; Zimmermann et al., 2003). As an example, the EVH1 domain in Ena/VASP proteins bind to FPPPP amino acid sequences present in several interactors (Ball et al., 2000). There are four different families of EVH1 domains: WASP, SPRED, Homer and Ena/VASP that binds to distinct proline-rich sequences (Peterson and Volkman, 2009). Common to all EVH1 domains is a conserved tryptophan residue interacting with a proline residue in the ligand. Four key residues surrounding the conserved tryptophan residue confer directionality and specificity for ligands and these residues varies in the different families. The EVH1 domain in the Falafel protein from Drosophila melanogaster (equivalent to PPP4R3) binds to a DVVFKKPLA sequence in the centromeric CENP-C protein where the F and P residues are key determinants of specificity (Lipinszki et al., 2015). Whether the CENP-C sequence represents an example of a general binding motif is not known.

Here we identify and extensively characterize a consensus-binding motif for the conserved PP4 holoenzyme allowing us to understand its direct binding partners in multiple biological pathways and uncover novel PP4 regulated pathways.

RESULTS

Identification of a consensus binding motif for the PP4 holoenzyme

Given the limited understanding of PP4 substrate selection we decided to focus on how the conserved PP4C-PPP4R2-PPP4R3 holoenzyme (hereafter referred to as PP4) (Figure 1 A–B) selects its substrates. We employed proteomic phage display (ProP-PD) to identify a possible consensus binding motif for PP4 (Wu et al., 2017). ProP-PD employs a phage library that displays the unstructured part of the human proteome as 16mer peptides on the M13 phage coat protein P8 (Davey et al., 2016; Sundell and Ivarsson, 2014). Purified full length PPP4R3A, PPP4R3B, PPP4R2 and the EVH1 domain of PPP4R3A was immobilized on a surface and phages binding to these proteins where enriched (Figure 1C, S1A, Table S1). Selected phages were barcoded, pooled and analyzed by next generation sequencing (NGS). Enriched peptides above a certain sequencing count were analyzed using the SLiMFinder algorithm to identify a consensus motif (Edwards et al., 2007). While PPP4R2 did not enrich any specific phages, both full length PPP4R3A and PPP4R3B baits bound peptides with a FxxP consensus signature (Fig. 1D–E). The isolated EVH1 domain from PPP4R3A enriched for peptides with a similar consensus motif suggesting that the binding site is present within this domain. To go beyond the limited search space of the ProP-PD library, we used a randomized peptide phage library and the PPP4R3A EVH1 domain as bait, which resulted in enrichment of F/WxxP containing peptides (Figure S1B, Table S1). In addition, a few enriched peptides contained an MxPP motif that likely represents a variant of the FxxP motif (Figure S1C, Table S1).

To analyze the FxxP motif in more detail we generated a synthetic model peptide, SLPFTFKVPAPPPSLPPS (positions numbered according to FKVP being residues 1–4) based on the ProP-PD results that constituted the predominant selected amino acid residues at each position. Similarly, peptides with alanine substitutions through the central part of the model peptide were synthesized to probe the importance of each position (Table 1, Table S2 and Data S1). Using isothermal titration calorimetry (ITC), we measured the affinities of these peptides for the PPP4R3A EVH1 domain, which bound the parental model peptide with an affinity of 0.9 μM. The mutation of the F at position 1 and P at position 4 resulted in a large increase in K_D that essentially abolished binding consistent with the strong enrichment of these residues in the ProP-PD screen. The alanine scan also revealed the contribution of residues surrounding the core FxxP motif and flexibility of position 2 and 3.

Table 1. ITC measurements of model and MCE1 peptides.

Affinities measured with indicated peptides and the EVH1 domain from PPP4R3A. See also Table S2 and Data S1.

Model Peptides
Variant	Sequence	KD (μM)
FxxP WT	PFTFKVPAPPPS	0.9
P-3: Ala	AFTFKVPAPPPS	1.1
P-2: Ala	PATFKVPAPPPS	2.7
P-1: Ala	PFAFKVPAPPPS	1.1
P1: Ala	PFTAKVPAPPPS	53.0
P2: Ala	PFTFAVPAPPPS	1.7
P3: Ala	PFTFKAPAPPPS	2.5
P4: Ala	PFTFKVAAPPPS	53.0
P6: Ala	PFTFKVPAAPPS	1.2
P7: Ala	PFTFKVPAPAPS	3.3
P8: Ala	PFTFKVPAPPAS	1.2
P1: Trp	PFTWKVPAPPPS	2.8
P1: Tyr	PFTYKVPAPPPS	12.9
P1: Met	PFTMKVPAPPPS	40.4
P1: Leu	PFTLKVPAPPPS	47.4
P1: Glu	PFTEKVPAPPPS	63.6
P1: Pro	PFTPKVPAPPPS	50.9
P1: His	PFTHKVPAPPPS	51.2
P2: Arg	PFTFRVPAPPPS	1.6
P2: Met	PFTFMVPAPPPS	3.3
P2: Glu	PFTFEVPAPPPS	12.0
P3: Glu	PFTFKEPAPPPS	2.9
P3: Lys	PFTFKKPAPPPS	1.1
P3: Met	PFTFKMPAPPPS	1.7
P5: Glu	PFTFKVPEPPPS	1.7
P6: Glu	PFTFKVPAEPPS	2.7
P6: Gly	PFTFKVPAGPPS	3.3
P7: Gly	PFTFKVPAPGPS	5.7
P8: Gly	PFTFKVPAPPGS	1.1
P9: Gly	PFTFKVPAPPPG	1.2
MCE1 variants
Variant	Sequence	KD (μM)
MCE1	DPDTELMPPPPPKRPHPL	0.98
MCE1-L(-1)A	DPDTEAMPPPPPKRPHPL	3.78
MCE1-M1A	DPDTELAPPPPPKRPHPL	85.6
MCE1-P2A	DPDTELMAPPPPKRPHPL	7.93
MCE1-P3A	DPDTELMPAPPPKRPHPL	7.58
MCE1-P4A	DPDTELMPPAPPKRPHPL	119.0
MCE1-P5A	DPDTELMPPPAPKRPHPL	1.49
MCE1-P6A	DPDTELMPPPPAKRPHPL	1.40

To determine which hydrophobic residues were allowed in position 1 we substituted the F to a number of hydrophobic residues. Mutation to W increased the K_D to 2.8 μM while the Y substitution resulted in a K_D of 12.9 μM. All other substitutions resulted in a K_D above 40 μM (Table 1). This suggests that W and Y can likely substitute for F in position 1 while other amino acid residues are poorly tolerated.

We also performed an alanine scan of the MxPP type of motif using the sequence from MCE1 (Table 1, Table S2, Data S1). This analysis revealed that M in position 1 and P in position 4 are critical and thus the motif resembles the FxxP motif. However the ability to accommodate M in position 1 is unique to the MxPP type of motif.

Since several proteins in the human proteome contain EVH1 domains that bind proline-rich sequences resembling the PP4 FxxP motif, it was of importance to determine if the identified motif was specific for PP4. We fused 4 copies of the model FxxP peptide to YFP with a spacer between each motif (referred to as 4X(FxxP)) and expressed it in HeLa cells. Purification of this construct showed that it bound PP4 components while a control construct with the FxxP motif mutated to AxxA (4X(AxxA)) did not (Figure 1F). Quantitative label-free comparison of 4X(FxxP) and 4X(AxxA) purifications by mass spectrometry revealed that 4X(FxxP) specifically bound all PP4 holoenzyme components (PP4C, PPP4R2, PPP4R3A/B) but not other reported PP4C regulatory subunits (PPP4R1, PPP4R4) (Figure 1G, Table S3). We did not detect a strong enrichment of other EVH1 domain containing proteins or other protein phosphatases arguing that the model peptide is specifically binding to PP4. Another large group of proteins purified by 4X(FxxP) were components of the proteasome, which is likely indirectly co-purified through FxxP unrelated PP4 holoenzyme interactions (see below and Table S3). Thus, the FxxP motif is specific for the EVH1 domain in PP4.

To determine if the FxxP binding pocket on PP4 constitutes a major contact site for PP4 binding partners we investigated if 4X(FxxP) could act as a competitive inhibitor for binding of MELK and MCE1 that binds either through FxxP or MxPP motif, respectively (see below). We analyzed YFP-PPP4R3A binding to FLAG-tagged MELK and MCE1 in the presence of RFP-tagged 4X(FxxP) or 4X(AxxA). 4X(FxxP) displaced both MELK and MCE1 from PPP4R3A purifications, arguing that at least for these binding partners the FxxP binding pocket on PP4 is required for interaction (Figure 1H).

To determine if the FxxP motif can confer PP4 substrate specificity we purified the PP4 holoenzyme by co-expressing its three subunits in HEK293 cells (Figure 1I). We then fused the FxxP model peptide sequence to a peptide sequence from Cdc20 containing 3 TP phosphorylation sites, which is normally a PP2-B55 substrate (Hein et al., 2017), to determine if this would result in PP4 mediated dephosphorylation. Indeed, fusing the FxxP motif but not the AxxA motif to this substrate increased its dephosphorylation rate substantially (Figure 1J). An important characteristic of the PP2A catalytic subunit is its preference for phosphothreonine over phosphoserine, which controls the temporal order of dephosphorylation in biological pathways (Hein et al., 2017). To determine if the PP4 holoenzyme displays a similar preference, we used two model peptides, KRpXIRR and RRApXVA, with either phosphorylated threonine or serine. It was evident that PP4 showed a clear preference for the phosphothreonine substrate (Figure 1K).

Collectively, we have defined a consensus-binding motif for PP4 that in combination with the phosphothreonine selectivity of PP4 is likely key determinants of substrate specificity and site-specific dephosphorylation.

Structure of FxxP peptide bound to PPP4R3A

To gain further insight into how PP4 binds to FxxP motifs, we crystalized our model peptide with the EVH1 domain from PPP4R3A and determined the structure to 1,5 Å (Figure 2A and Figure S1D, Table S4). We compared this structure to the previously reported structure of Falafel bound to a peptide from CENP-C, which revealed an overall similar fold (Figure S1E). In both structures the FxxP containing peptides adopt a PPII type of polyproline helix and bind to a fully conserved hydrophobic pocket on the EVH1 domain with the F and P residues constituting the main points of interaction. Indeed, the conformation of the CENP-C peptide and our model peptide deviate substantially outside of the FxxP motif. The binding pocket for the proline residue is very tight and involves an interaction with W19 (Figure 2B–E). PPII polyproline helixes are pseudosymmetric and can be recognized in both directions by EVH1 domains, but in our structure the hydrophobic binding pocket for the F residue in the FxxP motif confers directionality. Indeed, the peptide binding directionality observed is the same as in EVH1 domains from the WASP family, but opposite to Ena/Vasp-, Homer- and SPRED-class EVH1 domains (Peterson and Volkman, 2009). This appears to correlate with the position of a conserved tyrosine residue (Y11 in PPP4R3A) relative to the conserved tryptophan (W19 in PPP4R3A). Indeed the W19A or Y11A mutations in PPP4R3 blocked binding to FxxP peptides in ITC experiments and to Mce1 in cells (Figure 2F–G). Furthermore, in contrast to PP4 WT the recombinant PP4 W19A holoenzyme was unable to dephosphorylate substrates containing the FxxP motif in vitro confirming the structural observations (Figure 2H and Figure S1F). In PPP4R3 and WASP EVH1 domains this tyrosine residue is located on the side facing the F-binding pocket, whereas in Ena/Vasp-type domains it localizes to the opposite side with respect to the conserved tryptophan, thus likely acting as a determinant of directionality (Figure 2E). Despite of this similarity, the PPP4R3 EVH1 domain cannot be assigned to the WASP class of EVH1 domains. This is because our structure deviates from the WASP class at the other key positions involved in peptide binding used to classify this family. Furthermore, the peptide sequences recognized by WASP EVH1 domains are longer than the ones we identified for PPP4R3 as they have residues that bind to a third pocket on the EVH1 domain.

Figure 2. Structure of the EVH1-FxxP complex.

A) Overall structure of the PPP4R3A EVH1 domain in complex with the FxxP peptide with directionality of peptide indicated. B) Details of the interaction with the binding pocket for the proline and phenylalanine residues indicated. A zoom in on the FxxP binding pocket with the position of W19 indicated as well as the position of the four key residues used to define EVH1 families is shown. C-D) Details of the interaction between the FxxP motif and the EVH1 domain and the key determinants of binding shown as a closeup of the 3D structural model (C) or in a 2D-projection plot (D) E) Comparison of the different EVH1 domains and their interaction with proline rich peptides. The nature of the residues in positions 1–4 used to define the different EVH1 families is shown and compared to the residues in PPP4R3A. F) Affinity between the indicated PPP4R3A variants and a model peptide as determined by ITC. G) Analysis of the interaction between MCE1 and the PP4 holoenzyme containing either WT or W19A PPP4R3B by immunoprecipitation and Western Blotting. H) Dephosphorylation of an in vitro phosphorylated substrate containing an FxxP motif by the PP4 holoenzyme containing either WT or W19A PPP4R3A. Representative of 2 independent experiments with data points from both experiments shown. See also Figure S1 and Table S4.

In conclusion, the binding of FxxP motifs to the PPP4R3A EVH1 domain can be viewed as a minimized version of the WASP type of interactions with two out of three binding modules. Therefore, the PPP4R3 EVH1 domain could constitute a new family of EVH1 domains.

Systems-wide identification of PP4 binding motifs

Although our ProP-PD screen identified multiple FxxP motifs, we wanted to validate the importance of these motifs in a cellular context. To this end, we first established cellular PP4 interactomes using mass spectrometry analysis of YFP-tagged PP4 components (Figure 3A, Figure S2A, Table S3). This resulted in the identification of 174, 131 or 286 specific interactors for PPP4R2, PPP4R3A and PPP4R3B respectively (Figure S2B). PPP4R3A and PPP4R3B shared 56 interactors, revealing both common and unique binding partners for these regulatory subunits.

Figure 3. PP4 interactions mediated by FxxP and MxPP motifs.

A) Quantitative label free mass spectrometry analysis of PP4 holoenzyme components. PP4 components are indicated in red and proteins interacting through FxxP or MxPP motifs indicated in blue. B) Summary of in silico prediction of FxxP motifs in human proteins and in interactors. A PWM P Value cut off of 10E-5 and 10-E4 was used for the proteome and interactors, respectively C) Validated motifs by ITC or co-immunoprecipitation assays with K_D for peptides measured by ITC indicated D) Binding of myc-tagged PRP16 wild type (WT) or FxxP mutant (AxxA) (n=3), FLAG-tagged MCE1 full length or MCE1Δ585–603 (n= 5) or FLAG-tagged CDC25B WT or AxxA to YFP-tagged PPP4R3A or PPP4R3B in HeLa cells (n=3). PP4 interactors were transfected into stable cell lines expressing YFP-tagged PP4 components and purifications analyzed by western blot. Below is indicated the conservation of the FxxP and MxPP motifs analyzed. Inputs for purifications are in Supplemental Figure 2C. E) Schematic of CCDC6 and conservation of FxxP motif. F) Binding affinity for indicated peptides measured by ITC and using the EVH1 domain of PPP4R3A. G) Interaction of FLAG-tagged CCDC6 constructs with PPP4R3A-YFP in cells. The indicated FLAG-tagged CCDC6 constructs were transfected into a HeLa cell line stably expressing inducible PPP4R3A-YFP and complexes were affinity purified using a YFP affinity resin (n=3). H) The indicated YFP-CCDC6 constructs were transfected into HeLa cells and analyzed by live cell microscopy. Bar is 10 μM. The intensity of YFP signal in nucleus and cytoplasm was measured and their ratio was calculated. Representative images are shown and each dot in the plot represents a single cell analyzed. Median distribution of YFP signal is indicated by red line. Significance was determined using an unpaired t-test. See also Figure S2, Figure S3, Table S2, Table S3, Table S5 and Data S1.

To identify FxxP and MxPP type of motifs in our PP4 interactomes, we performed an in silico analysis where three filters were applied to enrich for high confidence motif instances in the human proteome (Figure 3B). First, we generated a position-scoring matrix (PSSM) based on the peptides selected in the ProP-PD screen to identify FxxP and MxPP motifs in proteomes that resemble these peptides (Table S5). For the FxxP motif we allowed F, W and Y in position 1. Secondly, we applied an accessibility filter to ensure that the motifs were in unstructured parts of proteins and, thirdly, a conservation filter to rank motifs. The in silico analysis revealed 289 FxxP and 139 MxPP high confidence motifs (Figure 3B, Table S5). Similarly, we identified 38 high confidence FxxP motifs in our interactomes and additional motifs with lower confidence. Predicted motifs in MCE1, PRP16, Centrobin, MELK and CCDC6 were validated by ITC and by co-immunoprecipitation experiments (Figure 3C–D (see Figure 3G for CCDC6) and Figure S2C–D, S3A–C). For a number of interactors, we tested binding to both PPP4R3A and PPP4R3B and observed instances with preferential binding to one of the isoforms.

Our proteome-wide bioinformatic analysis and ProP-PD selections identified potential PP4 binding motifs in many proteins absent from the PP4 MS interaction screens, which are possibly lost during purifications due to the micromolar affinity of PP4 for its binding partners. For instance, an MxPP motif in the TAF1 protein and conserved FxxP motifs in CDC25A and CDC25B were validated in ITC and Co-IP assays (Figure 3C–D, Figure S2D and S3D, Table S2 and Data S1, for CDC25A a triple mutant was analyzed). These results strengthen the validity of our in silico approach to identify PP4 docking motifs. The conservation of the PPP4R3 EVH1 domain all the way to yeast, prompted us to extend the search to include the proteomes of the model organisms M. musculus, D. melanogaster, C. elegans, S. cerevisiae, and S. pombe (Table S5). A web page for searching for PP4 binding motifs in proteins based on the methods outlined here is available at http://slim.ucd.ie/motifs/pp4/. These motif predictions can be integrated with additional experimental evidence of interaction to obtain high confident motifs.

Collectively, our ProP-PD screen and computational prediction identify numerous FxxP and MxPP motifs providing a rich resource for future studies of this phosphatase.

Regulation of PP4 interaction with FxxP motifs by phosphorylation

Previous analysis of PP1 and PP2A-B56 binding motifs have revealed that phosphorylation of residues in or surrounding the motifs can modulate affinities and constitute an important regulatory element (Hertz et al., 2016; Nasa et al., 2018). To determine if this type of regulation also applies to the FxxP motif, we analyzed the motif from CCDC6 as this contained three reported phosphorylation sites close to the motif (Figure 3E). We first measured the affinities of CCDC6 phosphopeptides for the EVH1 domain, which revealed that phosphorylation of T427 increased K_D from 6.7 μM to 39.8 μM while phosphorylation of S419 and S431 had a minor effect (Figure 3F, Table S2, Data S1). From our structure, it is unclear how phosphorylation at this position blocks binding as it would not sterically clash with the EVH1 domain, but it might interfere with the formation of a polyproline helix.

To determine if phosphorylation of T427 could affect binding of CCDC6 to PP4 in cells, we mutated this residue to either alanine to mimic the unphosphorylated form or to aspartic acid to mimic the phosphorylated form. Strikingly CCDC6 T427A but not CCDC6 T427D showed a strong increase in binding to PPP4R3A, consistent with this residue being phosphorylated in cells hereby preventing binding to PP4 (Figure 3G). To further substantiate this, we treated cell lysate with lambda phosphatase prior to immunopurification to determine if this would promote CCDC6-PP4 interaction. Indeed, CCDC6 bound much stronger to PPP4R3A upon lambda phosphatase treatment (Figure S3E). Finally, to determine if the binding of PP4 to CCDC6 could impact on CCDC6 function, we analyzed the nuclear to cytoplasmic ratio of YFP-CCDC6 constructs as CCDC6 has been reported to shuttle between nucleus and cytoplasm (Celetti et al., 2004). Under our conditions, CCDC6 is mainly in the cytoplasm but CCDC6 T427A showed an increased nuclear localization, suggesting that phosphorylation-mediated regulation of the CCDC6-PP4 complex controls the cellular distribution of CCDC6 (Figure 3H). Indeed we noticed faster migration of CCDC6 T427A on SDS-PAGE. To determine if the FxxP motif could accelerate CCDC6 dephosphorylation in vitro we used a recombinant fragment of CCDC6 harboring multiple reported proline-directed phosphorylation sites that were in proximity of the FxxP motif. Dephosphorylation of this fragment was faster in the presence of an intact FxxP motif (Figure S3F).

In conclusion, we find that phosphorylation of FxxP motifs can modulate binding to PP4 providing a mechanism for regulating substrate selection by PP4.

Regulation of cohesin release through direct WAPL-PP4 interaction

Interestingly, the in silico analysis identified a putative FxxP motif that is highly conserved in the cohesin release factor WAPL (Tedeschi et al., 2013) (Figure 4A). PP4 has been suggested to regulate cohesin release in fission yeast but whether this is through a direct interaction with cohesin regulators is unclear (Birot et al., 2017).

Figure 4. Direct binding of PP4 to WAPL is required for cohesin release.

A) Schematic of WAPL and conservation of FxxP motif. Residues dephosphorylated in a FxxP-dependent manner are indicated in red. B) Coomassie stained SDS-PAGE gel of GST-WAPL 375–625 WT or AxxA used for ITC experiments with the EVH1 domain from PPP4R3A. ITC measurements (n=2) with black curve indicating data and fitting for GST-WAPL 375–625 and red fitting for GST WAPL 375–625 AxxA. C) Interaction between PPP4R3B and RFP-tagged WAPL WT or AxxA mutant. RFP-tagged WAPL constructs were transfected into HeLa cells stably expressing inducible YFP or YFP-PPP4R3B and then an affinity YFP resin was used to purify complexes. A representative blot from 4 independent experiments is shown. D) WAPL WT and mutant (AxxA) phosphorylation was analyzed by Phos-tag or regular SDS-PAGE in asynchronous, S-phase, and mitotic cells. A representative blot from 6 independent experiments is shown. E) Representative images of chromosome spreads from the indicated conditions. Chromosomes were scored based on four categories as indicated and fraction of cells with type I or II morphology calculated. At least 100 cells per condition per experiment were scored. Error bars represent SD. Significance was determined using an unpaired t-test. F) Analysis of anaphase bridges by immunofluorescence in the indicated conditions and using cells stably expressing YFP tagged WAPL WT or AxxA. The fraction of cells showing anaphase bridges was calculated for each condition. At least 100 cells per condition per experiment were scored. Error bars represent SD. Significance was determined using an unpaired t-test. G) Purification of YFP-tagged WAPL proteins from asynchronous cells using an YFP affinity resin and analyzing the binding to cohesin components, RAD21 and SMC3 was analyzed by using quantitative western blotting. The level of RAD21 and SMC3 was set to 1 for WAPL WT purifications. Error bars represent SEM. Significance was determined using an unpaired t-test. See also Figure S4 and Table S3.

To explore if PP4 is binding to WAPL through the predicted FxxP motif, we used ITC measurements with recombinant fragments of WAPL and co-IP from cells using full length WAPL (Figure 4B–C, Figure S4A). Both approaches confirmed that the conserved FxxP motif in WAPL mediates binding to PP4 and that mutating the FxxP motif to AxxA can disrupt the interaction. Consistently, WAPL binding to PPP4R3B was disrupted by the W19A mutation (Figure S4B). Analysis of WAPL WT and the AxxA mutant on phos-tag gels suggested that WAPL WT was dephosphorylated by PP4 in both asynchronous and mitotic cells (Figure 4D). Quantitative mass spectrometry of purified YFP-WAPL WT and AxxA revealed multiple phosphorylation sites in WAPL that was increased when the PP4 binding site was mutated (Figure 4A, Table S3). This analysis also identified sites in Pds5B regulated by PP4 bound to WAPL suggesting additional substrates of PP4 in the cohesion release complex. The fact that WAPL is a direct substrate of PP4 was supported by in vitro assays using purified WAPL 375–625 phosphorylated by Aurora B, which was dephosphorylated by PP4 in a manner dependent on an intact FxxP motif (Figure S4C).

To explore if PP4 binding to WAPL was required for the ability of WAPL to release cohesion, we established a WAPL RNAi rescue system in HeLa cells. We generated stable inducible FLAG-tagged WAPL cell lines that were resistant to a penetrant WAPL RNAi oligo and expressed WAPL at close to endogenous levels (Figure S4D). The ability of WAPL to release cohesin can be monitored in chromosome spreads from mitotic cells as WAPL removal results in unresolved sister chromatids rather than the usual X-shaped chromosomes. Expression of WAPL WT fully rescued the unresolved sister chromatid phenotype, while WAPL AxxA only gave a partial rescue (Figure 4E). Furthermore, we monitored the presence of anaphase bridges during mitotic exit in the different conditions as WAPL removal results in an increase in these due to persistent cohesin (Tedeschi et al., 2013). In cells treated with WAPL RNAi, approximately 20% of cells displayed anaphase bridges, which was partially suppressed by WAPL WT but not WAPL AxxA (Figure 4F). Thus, two independent assays for WAPL functionality reveal that the FxxP motif in WAPL is required for its function. To determine if PP4 binding to WAPL might regulate binding to cohesin components, we purified YFP-tagged WAPL proteins from asynchronous cells and analyzed binding to the cohesin components RAD21 and SMC3 (Figure 4G). We reproducibly noticed a decrease in binding of WAPL AxxA to these cohesin components, suggesting that PP4 binding to WAPL is required for efficient cohesin binding. We note that cohesin components Scc1-SA2 have been reported to bind to WAPL 500–580 and that this region contains the conserved S549 phosphorylation site strongly regulated by PP4 in our mass spec analysis (Ouyang et al., 2013).

Collectively these data show that PP4 is an important regulator of cohesin release through direct binding to WAPL.

DISCUSSION

Here we provide important insight into how an evolutionary conserved PP4 holoenzyme selectively recognizes binding partners through the FxxP and MxPP motifs. The binding pocket for the FxxP motif is fully conserved from yeast to human, suggesting that FxxP motifs likely confer specificity to PP4 throughout eukaryotes, and indeed human PPP4R3 can complement a Psy2 (budding yeast PPP4R3) deletion in cisplatin sensitivity assays (Gingras et al., 2005). Although FxxP motifs constitute an important contact between PP4 and interactors, it is possible that additional contacts between the PP4 holoenzyme and interaction partners can contribute to binding as seen with the PP1 phosphatase (Peti et al., 2013). This might explain why certain proteins selectively binds to either PPP4R3A or PPP4R3B as revealed by our interactome analysis. However the fact that 4X(FxxP) purification lacks most of the interactors we see in PP4 subunit purifications would argue that the FxxP motif is a major binding determinant.

One striking observation regarding the FxxP motif is that the F and P residues are the main residues contacting the EVH1 domain. This is consistent with our affinity measurements of model peptides that revealed a large degree of flexibility of the surrounding residues. This raises the question on how specificity for PP4 is obtained. Clearly, numerous FxxP sequences exists that will not bind to PP4 and likely a complicated interplay between the surrounding residues and the FxxP motif determines its overall affinity to PP4. The fact that PP4 is mainly localized in the nucleus would prevent it from binding to FxxP motifs in cytoplasmic proteins. Despite the fact that EVH1 domains from other proteins bind to similar proline-rich sequences, our data show that our model peptide only binds PP4 and not other EVH1 domain-containing proteins in cells. This would suggest that despite similarity in the peptide sequences recognized by EVH1 domains, they are very selective and thus it should be possible to identify small molecules that can specifically bind to a given EVH1 domain. Given that PP4 inhibition has proven clinical relevant for targeting ovarian cancer, our work provides an important foundation for developing precise PP4 inhibitors (Coscia et al., 2018).

From our validation of motifs, it is clear that PP4 is an important regulator of fundamental nuclear processes and our work opens up for understanding how PP4 regulates these fundamental cellular processes.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Jakob Nilsson (jakob.nilsson@cpr.ku.dk).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell Culture and RNAi mediated depletion

HeLa cells (ATCC) were cultured in DMEM GlutaMax containing 100 U/ml penicillin, 100 mg/ml streptomycin and 10% FBS (all from Thermo Fisher Scientific) at 37°C in a humidified incubator with 5% CO₂. The parental HeLa FRT T-REx cell line was maintained in the above medium containing 5 ug/mL Blasticidin and 50 ug/mL Zeocin. Stable HeLa FRT T-REx cell lines were generated according to the manufacturers guidelines (Invitrogen) and cultured in the medium containing 5 ug/mL Blasticidin and 100 ug/mL Hygromycin B. Cell authentication was not carried out. E. coli DH5α strain was maintained and propagated using standard microbiological procedures. Silencer Select siRNA oligoes against target genes were purchased from life technologies. For nucleotide sequences see the Key Resources Table. siRNA oligoes (10nM final concentration of each) and Lipofectamine RNAi MAX (Invitrogen) were mixed and added to cells for 6 hours to overnight.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-FLAG	Sigma-Aldrich	F3165; RRID: AB_259529
Rabbit anti-RFP	MBL	PM005; RRID: AB_591279
Mouse anti-GFP	Roche	11814460001; RRID: AB_390913
Rabbit anti-GFP	Moravian	RRID: N/A
Rabbit anti-PPP4R3B	Bethyl	A300–842A; RRID: AB_597905
Rabbit anti-PP4C	Bethyl	A300–835A; RRID: AB_597901
Rabbit anti-PPP4R2	Bethyl	A300–838A; RRID: AB_2168750
GFP booster_ATTO488	Chromotek	gba488–10; RRID N/A
GFP-Trap_A	Chromotek	Gta-100; RRID: N/A
Mouse anti-RAD21	Millipore	05–908; RRID: AB_417383
Rabbit anti-SMC3	Behyl	A300–060A; RRID: AB_67579
Rabbit anti-GAPDH	Santa Cruz	sc-25778; RRID: AB_10167668
Rabbit anti-WAPL	Tedeschi et al, 2013	RRID: N/A
Mouse anti-Myc	Santa Cruz	sc-40; RRID: AB_627268
Goat anti-mouse IRDye 680RD	LI-COR	926–68070; RRID: AB_10956588
Goat anti-mouse IRDye 800CW	LI-COR	926–32210; RRID: AB_621842
Goat anti-rabbit IRDye 680RD	LI-COR	926–68071; RRID: AB_10956166
Goat anti-rabbit IRDye 800CW	LI-COR	926–32211; RRID: AB_621843
Bacterial and Virus Strains
E. coli DH5α	N/A	N/A
Phage-resistant BL21(DE3)-R3	SGS consortium	N/A
Chemicals, Peptides, and Recombinant Proteins
Lipofectamine2000	Thermo Fisher	11668019
Lipofectamine RNAiMAX	Thermo Fisher	13778150
Protease inhibitor cocktail	Roche	04 693 159 001
Phosphatase inhibitor cocktail	Roche	04 906 837 001
Phos-tag AAL-107	WAKO	304–93525
Peptide sequences are listed in Table S2	Peptide 2.0	N/A
Critical Commercial Assays
NucleoSpin Plasmid EasyPure	Macherey-Nagel	740727.250
QIAquick PCR purification Kit	QIAGEN	28106
Illustra DNA gel band purification kit	GE Healthcare	28–9034–71
Rapid DNA Ligation kit	Sigma-Aldrich	000000011635379001
QuikChange Multi Site-Directed Mutagenesis Kit	Agilent	200514
PiColorLock Phosphate Detection System	Expedon	303–0030
Deposited Data
MS data	ProteomeXchange	PXD012695
Dataset	Mendeley	http://dx.doi.org/10.17632/dsyhj68hsf.1
PDB dataset	PDB database	PDB 6R8I
Experimental Models: Cell Lines
HeLa	ATCC	ATCC CCL-2
HEK293 6E	Gift Yves Durocher	N/A
Oligonucleotides
Oligonucleotide sequences deposited in Mendeley as “List of oligionucleotides”	Mendeley
Recombinant DNA
pcDNA 5/FRT/TO	Invitrogen	V6520–20
pGEX 4T1	GE Healthcare	28–9545–49
pET30a	Novagen	69909
Software and Algorithms
GraphPad Prism 8 for Mac OS X	GraphPad Software	N/A
Softworx	GE Healthcare	N/A
Image Studio Software v 5.2.5	LI-COR	N/A
ImageJ	NIH	N/A
MicroCal PEAQ-ITC analysis software	Malvern Panalytical	N/A
XDS/XSCALE	Kabsch (2010)	N/A
AIMLESS	Evans and Murshudov (2013)	N/A
autoPROC	Vonrhein et al. (2011)	N/A
phenix.xtriage	Adams et al. (2010)	N/A
PHASER	McCoy et al. (2007)	N/A
PHENIX.AUTOBUILD	Adams et al. (2010)	N/A
COOT	Emsley et al. (2010)	N/A
PHENIX.REFINE	Adams et al. (2010)	N/A
The PyMOL Molecular Graphics System, Version 2.0	Schrödinger	N/A
UCSF Chimera	Pettersen et al. (2004)	N/A
LigPlot+	Laskowski and Swindells (2011)	N/A
ABPS	Baker et al. (2001)	N/A
Clustal Omega	Chojnacki et al. (2017)	N/A
Chimera	Sanner et al. (1996)	N/A
PDB redo	Joosten et al. (2009)	N/A
Molprobity	Chen et al. (2010)	N/A

METHODS DETAILS

Cloning

Constructs for mammalian expression was cloned into pcDNA/FRT/TO vectors with the corresponding affinity tag encoded. cDNAs were amplified by PCR and inserted into these vectors by restriction digest and ligation, and subsequently confirmed by sequencing. Mutations or generation of RNAi resistant constructs was done by PCR based whole plasmid mutagenesis using two complementary primers encoding the mutations. Expression constructs for bacterial expression was pGEX 4T-1 WAPL 375–625, pGEX 4T-1 CDC20(48–78)-FxxP/AxxA, pET30a CDC25A full length, pET30a MCE1 full length, pET30a PPP4R3A/B and PPP4R2 and mutations of truncations of these. These constructs were generated by PCR or gene synthesis of the corresponding cDNA fragment and cloning into the corresponding vectors using appropriate restriction enzymes followed by sequencing of the entire insert. The expression construct for the EVH1 domain of PPP4R3A was generated by ligation independent cloning into pCPR0009 (pNIC28-HIS-StrepII-HA) by amplifying a cDNA fragment encoding amino acids 2–117 of PPP4R3A. Point mutations where introduced by whole plasmid PCR followed by DpnI digest. Full details of cloning will be made available upon request.

Protein expression and purification

Bacterial expression constructs were expressed in BL21(DE3) cells at 18 degrees by induction with 0.5 mM IPTG (isopropyl β-D-1-thiogalactopyranoside) overnight. Cells were suspended in lysis buffer (50 mM Tris-HCl pH=7.5 (for His-tagged proteins phosphate buffer was used), 300 mM NaCl, 10% glycerol, 0.5 mM TCEP 1XComplete EDTA-Free tables (Roche) (for His-tag purifications this buffer contained 10 mM immidazole)) and lysed using a high-pressure homogeniser (Avestin). Cell lysate was clarified by centrifugation and applied to either a GST or a His affinity column (GE Healthcare). Columns were washed extensively with lysis buffer and GST-tagged proteins eluted with lysis buffer containing 20 mM reducted glutathione and His-tagged proteins eluted with 500 mM immidazole. Peak fractions were pooled and applied to appropriate size exclusion columns and peak fractions pooled and aliquoted and flash frozen. For CDC25A recombinant proteins, an additional MonoQ purification step was included. For purification of untagged EVH1 domain, the His-tag was removed by TEV cleavage and then applied to a His affinity column allowing collection of untagged EVH1 domain in the flow through. This untagged EVH1 domain was applied to a size exclusion chromatography column and peak fractions pooled. All proteins were confirmed by mass spectrometry.

For expression and purification of the PP4 holoenzyme, the following 3 plasmids were transfected into HEK293 E6 cells: pcDNA5/FRT/TO-Strep-tag II-PPP4R3A, pcDNA5/FRT/TO 3xFLAG PPP4R2 and pcDNA5/FRT/TO 3XFLAG PP4C. After 3 days of expression in Freestyle F17 medium, the cells were harvested and resuspended in buffer W (100 mM Tris-HCl pH=8.0, 150 mM NaCl, 1 mM EDTA, Complete EDTA-free tablet (Roche), 0,5 mM TCEP) and lysed. Following clarification by centrifugation, the lysate was loaded on a StrepTactin column (IBA) and washed with buffer W extensively. Proteins were eluted with buffer W containing 2,5 mM desthiobiotin and peak fractions were flash frozen.

ProP-PD

The bait protein (15 μg in 100 μl PBS) and GST for pre-selection were coated in separate wells of a flat-bottom Immuno Maxisorp plate (Nunc, Roskilde, Denmark) over night at 8°C. Unbound proteins were removed and the wells w ere blocked with 200 μl 0.5 % (w/v) BSA in PBS for 1h at 8°C. GST coated pre-selection wells were washed four times with 200 μl PT buffer (PBS, 0.05% Tween-20) and incubated with 100 μl naïve phage library in PBS for 1 h (~ 10¹² phage particles per well to remove non-specifically binding phages. Libraries have been described elsewhere (Davey et al., 2016; Ilari et al., 2015)). Wells coated with bait proteins were washed four times with PT buffer and the 100 μl phage library solution was transferred from the pre-selection wells to the target protein wells and allowed to bind to the bait protein for 2h at 8°C under gentle agitation. Unbound phages were removed by washing with cold PT buffer four times. Bound phages were eluted by direct infection into E.coli Ominmax by the addition of 100 μl of log phase (A₆₀₀ = 0.6–0.8) bacteria in 2TY (1% (w/v) yeast extract, 1.6% (w/v) tryptone and 0.5% (w/v) NaCl) and incubation at 37°C with shaking for 30 minutes. Bac teria were hyperinfected with helper phage to enable phage production by the addition of 10 μl 1 × 10¹¹ p.f.u./ml M13KO7 (NEB, Ipswich, MA, USA) followed by incubation for 45 minutes at 37 °C with shaking. The bacterial cultures were transferred to 10 ml 2YT supplemented with 300 μM IPTG, 50 μg/ml carbenicillin and 25 μg/ml kanamycin, and incubated overnight at 37°C with shaking. Bacteria were pelleted by centrifugation (10 min, 4,000 g). The supernatant was transferred to a new tube containing 1/5 of the final volume PEG/NaCl (20% PEG-8000 (w/v), 2.5 M NaCl), and incubated on ice for 10 minutes. Phages were pelleted by centrifugation at 16,000 g for 15 minutes and the supernatant was discarded. The phage pellet was resuspended in 1 ml PBS and used in the next round of selection. Four rounds of selections were carried out. The enrichment of binding phages was followed by pooled phage enzyme-linked immunosorbent assays as previously described (Huang and Sidhu, 2011). The binding enriched phage pools were barcoded and analyzed through next-generation sequencing as described (Davey et al., 2016; Wu et al., 2017)

Immunoprecipitation

In order to analyze the interaction of PPP4R3A and PPP4R3B with putative protein interactors, stable cell lines expressing the inducible YFP (control) or YFP-tagged PP4 regulatory subunits were transfected with constructs encoding the RFP-tagged, FLAG-tagged or myc-tagged proteins of interest 48 h prior to harvesting cells. The expression of PP4 regulatory subunits was induced with doxycycline 24 hours before harvesting. For analyzing WAPL phosphorylation sites (see below) and the interaction between the cohesin complex and WAPL through the FxxP motif, YFP (control), YFP-tagged WAPL WT and AxxA stable cell lines were used. Cells were lysed in low or high salt lysis buffer (50 mM Tris-HCl pH 7.4, 50 (low) or 150 mM NaCl (high), 1 mM EDTA, 0.1% NP40, 1 mM DDT, protease- and phosphatase inhibitors) on ice for 30 min. For WAPL and TAF1 IPs, the lysate was treated with benzonase nuclease (Millipore). Immunoprecipitation was performed at 4 degrees in lysis buffer with GFP-Trap beads (ChromoTek) and precipitated protein complexes were washed four times in lysis buffer and eluted in 2x SDS sample buffer. For lambda phosphatase treatment of extract prior to immunopurification, the cell extract was incubated with 0.25 Units/ul lysis buffer for 45 minutes.

ITC

Peptides were purchased from Peptide 2.0 Inc. Prior to ITC experiments both the recombinant PPP4RA EVH1 protein and the peptides or recombinant interaction partners were extensively dialyzed against 50 mM sodium phosphate pH 7.5, 150 mM NaCl, 0.5 mM TCEP. All ITC experiments were performed on an Auto-iTC200 instrument at 25 °C. Concentrations of peptides and proteins used in each experiment are listed in Supplementary Table 2. The peptides were loaded into the syringe and titrated into the calorimetric cell containing the PPP4R3A EVH1 domain. ITC experiments with the full length proteins, CDC25A, MCE1 and WAPL were performed with the full length proteins in the calorimetric cell and the PPP4R3A EVH1 domain in the syringe. Control experiments with the peptides or the PPP4R3A EVH1 domain injected in the sample cell filled with buffer were carried out under the same experimental conditions. These control experiments showed heats of dilution negligible in all cases. In all assays, the titration sequence consisted of a single 0.4 μl injection followed by 19 injections, 2 μl each, with 150 s spacing between injections to ensure that the thermal power returns to the baseline before the next injection. The stirring speed was 750 rpm. The heats per injection normalized per mole of injectant versus the titrant to titrate molar ratio were fitted to a single-site model. Data were analysed with MicroCal PEAQ-ITC analysis software.

In vitro phosphatase assays

15 μg GST- fusion protein GST-CDC20 49–78, engeneered with wt FxxP or mutant AxxA PP4 motif or 6,6 μg GST-CCDC6 346–474 wt or F243A was incubated with 20U CDK1-CyclinB1 (New England Biolabs) or 1 μL recombinant, human CDK1-CyclinB1 (Sigma). GST-WAPL 325–611 was phosphorylated with ~ 0,58 μg purified AuroraB-INCENP. Kinase reactions were performed in 50 μL and in buffer K (50 mM Tris-HCl pH 7,5, 10 mM MgCl₂, 0,1 mM EDTA, 2 mM DTT, 0,01% Brij 35) with 500 μM ATP and 1μCi (γ−³²P)-ATP (Perkin Elmer) at 30°C for 60 minutes. Reactions were stop ped by the addition of 10 μM RO-3306 (Calbiochem) or 0,5 μM ZM447439 (Tocris Bioscience) for the inhibition of CDK1-CyclinB1 or AuroraB, respectively. PD Spin Trap G25 columns (GE Healthcare) were used to exchange the buffer to buffer P (50 mM Tris pH 7,4, 1 mM MnCl₂, 1mM DTT, 0,1% vol/vol IGEPAL, 150 mM NaCl).

Dephosphorylation reactions were prepared in non-stick tubes treated with blocking buffer (50 mM Tris pH 7,4, 0,1 mM MnCl₂, 1 mM MgCl₂, 1 mM DTT, 0,1% vol/vol IGEPAL, 300 mM NaCl, 2 mg/ml BSA) on ice. For the GST-CDC20 49–78 model substrate dephosphorylation assays, 6 μg of PP4 holoenzyme was added to 155 μl of phosphorylated substrate (~13 μg). For the GST-WAPL 325–611 dephosphorylation assays, 4,8 μg PP4 holoenzyme was added to 84 μl phosphorylated protein (7 μg). For the CCDC6 dephosphorylation assays, 96 μl phosphorylated GST-CCDC6 (~4,9 μg) was incubated with PP4 holoenzyme. Samples of ~1,75 μg were taken out at the indicated time-points, added to 4x SDS loading buffer, boiled for 5 minutes and separated by SDS-PAGE. Gels were dried, exposed for 3–4 days and imaged on Typhoon FL 950 (GE Healthcare). Analyses and quantifications were carried out in ImageJ.

Threonine over serine preference was determined for the purified PP4-PPP4R2-PPP4R3A holoenzyme using synthetic pT (KRpTIRR and RRApTVA) or pS (KRpSIRR and RRApSVA) phosphorylated peptides manufactured by Peptide 2.0 and used at a final concentration of 240mM. Phosphatase reactions were performed in phosphatase buffer (50 mM Tris pH 7.4, 1 mM MnCl₂, 1 mM DTT, 0.1% (vol/vol) IGEPAL, 150 mM NaCl) for 30 minutes at 30 °C. Release of inorganic phosphate was measured usi ng the PiColorLock Phosphate Detection System (Expedon).

Crystallization and structure determination

Crystals of the PP4R3A EVH1 domain in complex with the synthetic model (N’-SLPFTFKVPAPPPSLPPSW-C’) peptide were grown in sitting drops by vapour diffusion. PP4R3A EVH1 domain at 18 mg ml⁻¹ in 20 mM Tris pH 8.0, 150 mM NaCl, 0.25 mM TCEP was incubated with the model peptide in a 1:1.2 ratio. Crystals of the complex were grown at 20 °C by mixing 0.1 μl of PP4R3A EVH1/model with a 1:1 dilution of mother liquid (1 M tri-sodium citrate, 0.1 M sodium cacodylate pH 6.5). Crystals were grown to full size (0.15 × 0.1 × 0.05 mm) after one week and soaked in reservoir liquid supplemented with 20% glycerol before flash frozen in liquid nitrogen for data collection.

Datasets from frozen native crystals were collected at the beamline PXI of the Swiss Light Source (SLS) at the Paul Scherrer Institute (PSI) in Villigen, using a EIGER 16M detector and a temperature of 100K at a wavelength of 1.0 Å as well as an oscillation range of 0.1° over a rotation range of 360°.

The data were integrated and scaled using XDS/XSCALE (Kabsch, 2010) and AIMLESS (Evans and Murshudov, 2013) as integrated in autoPROC (Vonrhein et al., 2011) and analysed in phenix.xtriage (Adams et al., 2010) Crystals diffracted to 1.5 Å resolution and the crystal space group was determined as P 4₃ 2₁ 2 with unit cell dimensions of a = 60.58 Å, b = 60.58 Å, c = 91.90 Å and α = β = γ 90°. Statistics for the crystallographic data and structure solution are summarized in Supplementary Table S4.

The PP4R3A EVH1 domain structure in complex with the model peptide was solved by molecular replacement, as implemented in the program PHASER (McCoy et al., 2007) using the Falafel structure (PDB 4WSF) as a searching model (Lipinszki et al., 2015). An initial atomic model was obtained using PHENIX.AUTOBUILD (Adams et al., 2010) and the model was completed by iterative model building in COOT (Emsley et al., 2010) and refined to good geometry using PHENIX.REFINE (Adams et al., 2010) resulting in final working and free R-factors of 19.83% and 22.2%. (Table S4). Protein structures and interactions were visualized with PYMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.), UCSF Chimera (Pettersen et al., 2004) and LigPlot⁺ (Laskowski and Swindells, 2011). Electrostatic surface potentials were calculated using ABPS (Baker et al., 2001) and conservation surface depictions were created using Clustal Omega (Chojnacki et al., 2017) and Chimera (Pettersen et al., 2004; Sanner et al., 1996). PDB coordinates were validated using PDB redo (Joosten et al., 2009) and Molprobity (Chen et al., 2010) and are available on the PDB database (PDB code 6R8I).

Mass spectrometry analysis

YFP, YFP-PPP4R1, YFP-PPP4R3A, YFP-PPP4R3B, PPP4R3B-YFP, YFP-4X(FxxP), YFP-4X(AxxA), and YFP-WAPL pulldowns (see purification protocol) were performed as independent biological triplicates and analyzed by label-free LC-MS/MS on a Q-Exactive Plus quadrupole Orbitrap mass spectrometer (ThermoScientific) equipped with an Easy-nLC 1000 (ThermoScientific) and nanospray source (ThermoScientific). Peptides were resuspended in 5% methanol / 1.5% formic acid and loaded on to a trap column (1 cm length, 100 μm inner diameter trap packed with ReproSil C₁₈ AQ 5 μm 120 Å pore beads (Dr. Maisch, Ammerbuch, Germany)) vented to waste via a micro-tee and eluted across a fritless analytical resolving column (35 cm length, 100 μm inner diameter fused silica packed with ReproSil C₁₈ AQ 3 μm 120 Å pore beads) pulled in-house (Sutter P-2000, Sutter Instruments, San Francisco, CA) with a 60 minute gradient of 5–30% LC-MS buffer B (LC-MS buffer A: 0.0625% formic acid, 3% ACN; LC-MS buffer B: 0.0625% formic acid, 95% ACN). The Q-Exactive Plus was set to perform an Orbitrap MS scan (R=70K; AGC target = 3e6) from 350 – 1500 Thomson, followed by HCD MS² spectra on the 10 most abundant precursor ions detected by Orbitrap scanning (R=17.5K; AGC target = 1e5; max ion time = 75ms) before repeating the cycle. Precursor ions were isolated for HCD by quadrupole isolation at width = 0.8 Thomson and HCD fragmentation at 26 normalized collision energy (NCE). Charge state 2, 3 and 4 ions were selected for MS². Precursor ions were added to a dynamic exclusion list +/− 20ppm for 20 seconds. Raw data were searched using COMET in high resolution mode (Eng et al., 2013) against a target-decoy (reversed) (Elias and Gygi, 2007)version of the human (UniProt; downloaded 2/2013, 40482 entries of forward and reverse protein sequences) with a precursor mass tolerance of +/− 1 Da and a fragment ion mass tolerance of 0.02 Da, and requiring fully tryptic peptides (K, R; not preceding P) with up to three mis-cleavages. Static modifications included carbamidomethyl cysteine and variable modifications included: oxidized methionine. Searches were filtered using orthogonal measures including mass measurement accuracy (+/− 3ppm), Xcorr for charges from +2 through +4, and dCn targeting a <1% FDR at the peptide level. Peptide quantification of LC-MS/MS spectra was performed using MassChroQ (Valot et al., 2011). For WAPL phosphorylation site analysis retention time alignment for smart quantification for performed using MassChroQ (Valot et al., 2011). Phosphorylation site abundances were normalized to the respective protein abundances in the WAPL pulldown. For protein quantification the iBAQ method was employed (Schwanhäusser et al., 2011).

IBAQ and peptide quantifications were imported into Perseus (Tyanova et al., 2016), and log₂ transformed. For Venus-only control and WAPL samples, missing values were imputed from a normal distribution to enable statistical analysis and visualization by volcano plot. Statistical analysis of protein quantification was carried out in Perseus by two-tailed Student’s t-test.

In silico prediction of motifs

Peptides returned from the phage display screens were aligned by their FxxP-like motifs. For motif specificity determinant visualization (Figure 1), sequence logos of the FxxP-like motifs were constructed using the PSSMSearch tool (Krystkowiak et al., 2018) as relative binomial logos where each amino acid was scored as log₁₀ of the binomial probability (prob^aa = binomial(k,n,p) where k is the observed residue count at each position for a residue, n is the number of the instances of motifs and p is the background frequency of the residue in the disordered regions of the human proteome). For the FxxP motif discovery searches, the consensus motif [FWY]xxP was searched against the human UniProt reviewed proteins and annotated with motif features and attributes using the SLiMSearch framework (Krystkowiak and Davey, 2017). Matches were filtered using (i) extracellular localization GO terms, (ii) intrinsic disorder predictions (retaining only peptides found in disordered regions as defined by an IUPred score < 0.4 (Dosztányi et al., 2005)) and (iii) UniProt annotation of topologically inaccessible regions (e.g. transmembrane and extracellular regions) (UniProt Consortium, 2015). Next, matches were filtered based on the taxonomic range of the motif based on the presence of the consensus at the same position in an orthologue alignment outside the mammalian clade. Finally, a position-specific scoring matrix (PSSM) was constructed using the PSI BLAST IC scoring scheme (Altschul et al., 1997; Krystkowiak et al., 2018). The PSSM was constructed using an alignment of the FxxP-containing peptides returned in the ProP-PD screens for both PP4A and PP4B. The consensus matches were scored using the PSSM as defined in the PSSMSearch tool (Krystkowiak et al., 2018) and were filtered using the PSSM p-value, with a cut-off of 0.0001 for the human dataset and 0.001 for the MS interactions dataset. FxxP discovery analyses were also performed for mouse, worm, fly, budding yeast and fission yeast proteomes. Human MPPP motif discovery searches were performed similarly with the exception of PSSM construction. Due to the limited number of validated instances, the MPPP PSSM was created using the validated human instances and their conserved instances in orthologues of the Quest for Orthologues species.

Phos-tag SDS-PAGE

Mn²⁺-Phos-tag SDS-PAGE gels (6% acrylamide) were casted with 20μM Phos-tag AAL-107 (WAKO). FLAG-WAPL WT and AxxA cells were collected at indicated cell cycle stages and lysates were prepared using an EDTA-free cell lysis buffer. Samples were mixed with the Laemmli sample buffer and boiled before loading. Electrophoresis and transfer were performed according to the manufacturer’s instructions.

Chromosome Spreads and anaphase bridges

Chromosome spread assay was performed according to (Zheng et al., 2017). Briefly, Parental HeLa FRT (control), FLAG-WAPL WT and AxxA cells were seeded in a 6-well plate, WAPL knockdown was performed using RNAi (s22948; Life Technologies), and cells were synchronized using 2mM thymidine followed by 5μM nocodazole treatment. 72 h after WAPL RNAi transfection, mitotic cells were collected by shake-off. After hypotonic treatment, cells were spun onto microscopy slides with a Shandon Cytospin centrifuge (Thermo Fischer), fixed with 4% paraformaldehyde, and then stained with DAPI. Representative images were taken with a 100x objective on a DeltaVision fluorescent microscope under the same condition. Cells were categorized in four types based on their chromosome morphology, and the percentage of cells with Type I and II morphology were calculated and plotted using the GraphPad Prism 8. At least 100 cells per experimental condition were scored. To observe anaphase bridges, control, YFP-WAPL WT and AxxA cells were seeded on coverslips, transfected with WAPL RNAi, and synchronized with overnight 2mM thymidine followed by overnight 10μM RO-3306 treatment. 80 min after the RO-3306 release, cells were fixed with 4% paraformaldehyde and stained with GFP booster (Chromotek) and DAPI. All images were taken with a DeltaVision fluorescent microscope under the same condition. At least 100 cells per experimental group were scored whether they contain anaphase bridge or not, and the percentage of anaphase cells with a bridge were calculated and plotted using the GraphPad Prism 8.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analysis was performed using GraphPad Prism 8 for Mac OS X. Statistical details and definition of parameters are stated in figure legends, and p < 0.05 was considered statistically significant. Statistical methods were not employed to determine sample size or to determine whether the data met the assumptions of the statistical approach.

DATA AND SOFTWARE AVAILABILITY

Mendeley dataset: http://dx.doi.org/10.17632/dsyhj68hsf.1

The mass spectrometry data have been deposited in the ProteomeXchange under accession number PXD012695.

The structural data is available in the PDB database under PDB 6R8I.

All software used in the in silico motif discovery screens are previously published and freely accessible as web servers available at http://slim.ucd.ie/pssmsearch/.

Supplementary Material

2

Table S1: ProP-PD screen data. Related to Figure 1

3

Table S2: ITC data. Related to Figure 3 and Table 1

4

Table S3: Mass spectrometry data. Related to Figures 1, 3 and 4

5

Table S5: Computational motif predictions. Related to Figure 3

Highlights.

• The conserved PP4 holoenzyme binds to FxxP motifs that provide specificity

• FxxP motifs bind to a conserved binding pocket on PP4 regulatory subunit

• Binding to FxxP motifs can be regulated through phosphorylation

• PP4 binding to an FxxP motif in WAPL regulates its cohesin release activity