Cryo-EM structures of PP2A:B55 with p107 and Eya3 define substrate recruitment

Abstract

Phosphoprotein phosphatases (PPPs) achieve specificity by binding substrates and regulators using PPP-specific short motifs. Protein phosphatase 2A (PP2A) is a highly conserved phosphatase that regulates cell signaling and is a tumor suppressor. Here, we use cryo-electron microscopy and nuclear magnetic resonance (NMR) spectroscopy to investigate the mechanisms of human p107 substrate and Eya3 regulator recruitment to the PP2A:B55 holoenzyme. We show that, while they associate with B55 using a common set of interaction pockets, the mechanism of substrate and regulator binding differs and is distinct from that observed for PP2A:B56 and other PPPs. We also identify the core B55 recruitment motif in Eya3 proteins, a sequence conserved amongst the Eya family. Lastly, using NMR-based dephosphorylation assays, we demonstrate how B55 recruitment directs PP2A:B55 fidelity through the selective dephosphorylation of specific phosphosites. As PP2A:B55 orchestrates mitosis and DNA damage repair, these data provide a roadmap for pursuing new avenues to therapeutically target this complex by individually blocking a subset of regulators that use different B55 interaction sites.

Main

Serine/threonine phosphorylation-dependent signaling is essential for the control of most biological processes¹. Serine/threonine kinases, the enzymes that phosphorylate serine and threonine residues, typically recognize their substrates using kinase-specific phosphosite recognition sequences, which ensures phosphorylation fidelity and functional signaling events². Serine/threonine phosphoprotein phosphatases (PPPs) counteract the action of kinases^1,3,4. However, a PPP-specific phosphosite recognition sequence has only been identified for protein phosphatase 2B (PP2B; also known as calcineurin)⁵. Indeed, substrate recognition by PPPs, including PP2B, is predominantly achieved through the recruitment of substrates using PPP-specific short linear motifs (SLiMs; short 4–8-aa linear stretches of protein sequences that mediate protein–protein interactions) to sites distal from the active site^6,7,8,9,10. These PPP-specific SLiMs, which are present in either the substrates themselves or the scaffolding proteins that recruit substrates, bind directly to preformed SLiM-binding pockets on their cognate PPPs. Indeed, PPP-specific SLiMs have been identified for PP1 (refs. ^{8,11,12,13,14,15}), PP2B (refs. ^6,16), PP2A:B56 (refs. ^7,10) and PP4 (ref. ⁹). Thus, most PPP-specific inhibitors and substrates use the same interaction pockets for PPP binding (that is, they compete for the same interaction sites)^1,17. Indeed, viruses have exploited this characteristic of PPP regulator and substrate binding by producing SLiM-containing proteins that inhibit PPP-specific dephosphorylation by binding and blocking PPP regulator-binding or substrate-binding sites, rendering them unable to recruit and, in turn, dephosphorylate their specific substrates^6,18,19,20. Notably, for PP2B, blockage of regulator-binding and substrate-binding sites is also used by the immunosuppressant drugs FK-506 and cyclosporin A⁶.

Here, we focus on the interaction of substrates and regulatory or scaffolding proteins with PP2A:B55. PP2A is a trimeric holoenzyme comprising a catalytic subunit, PP2Ac, a scaffolding subunit, PP2Aa, and one of several regulatory subunits, commonly referred to as B subunits, each of which adopts a distinct fold (B55, B56, PR72 and PR93)^21,22,23. The molecular basis of substrate recruitment to B56, which is a heat repeat protein, is well established as B56 binds substrates and regulators that contain LxxIxE SLiMs, with the binding affinities of distinct SLiMs modulated by phosphorylation and/or adjacent dynamic charge–charge interactions^7,10,24. In contrast, much less is known about the mechanism(s) of substrate and regulator recruitment for the most abundant PP2A holoenzyme, PP2A:B55 (B55 is a WD40 domain; Fig. 1a)^25,26. Recently, we determined the structures of two protein inhibitors, phosphorylated ARPP19 and FAM122A, bound to PP2A:B55 (ref. ²⁷). The interactions were extensive, leading to the identification of multiple surfaces on B55, whose shallow pockets collectively mediate tight inhibitor binding, including the B55 platform (pockets P1–P6), the B55 wall (P7–P9), the B55 entry and the B55 hook (P10) (Fig. 1a–c). These structures also showed that ARPP19 and FAM122A bind PP2A:B55 through helices rather than extended interactions, suggesting that B55-based recruitment is fundamentally distinct from that of its fellow PPP members. While some PP1 regulators were previously shown to bind PP1 using helices in addition to canonical PP1-specific SLiM motifs (Inhibitor 2, spinophilin, NIPP1 and MYPT1), it was the SLiM motifs that were essential for PP1 binding. Consistent with this observation, both ARPP19 and FAM122A had shared and distinct interactions with B55 and PP2Ac, highlighting that the interactions between regulators and PP2A:B55 are likely more complex for this holoenzyme²⁷.

Fig. 1: PP2A:B55 and its interactors.

a, PP2A:B55 (B55, lavender; PP2Aa, light gray; PP2Ac, dark gray) bound to inhibitors ARPP19 (dark blue) or FAM122A (orange). Inhibitors bind B55 through the B55 platform (beige; ARPP19/FAM122A), B55 wall (green; ARPP19/FAM122A), B55 entry (pink; ARPP19) or the B55 hook (yellow; ARPP19). PDB 8TTB and 8SO0 were used to create the figure. b, Close-up view of the B55 binding surface, with key interaction pockets (P1–P10) indicated (PDB 8TTB). c, Cartoon illustrating the PP2A subunits and B55_LL; B55_LL is monomeric, as it is unable to assemble with PP2Aa or PP2Ac. d, Domain structure of p107, with key constructs studied herein indicated. e, Domain structure of Eya3 (top) with constructs used in this study (bottom).

Compared to other PPP family members, only limited experimental data are available to fully define the mechanism(s) of PP2A:B55-specific substrate and regulator recruitment. Recently, the structure of the PP2A:B55–IER5 (immediate early response 5) substrate complex was described²⁸, which showed that, while IER5 also uses helices to bind to PP2A:B55, the interaction is distinct from that observed for ARPP19 or FAM122A, increasing the likelihood that there are multiple mechanisms used by B55 to recruit substrates and regulators. Here, we study the interaction of the PP2A:B55 substrate p107 (Fig. 1d) and the scaffold and regulator Eya3 (Fig. 1e) to define the molecular basis of B55-specific substrate recruitment. p107 is a member of the retinoblastoma family of growth suppressor proteins (RB, p107 and p130) that facilitate the coordinated regulation of cell-cycle progression by modulating the E2F family of transcription factors^29,30. Phosphorylation by cyclin-dependent kinases (CDKs) at the intrinsically disordered connector between the two structured pocket domains, especially S640, inactivates p107 (refs. ^31,32). PP2A:B55 activates p107 by reversing this CDK phosphorylation, a process dependent on the B55-specific recruitment of p107^33,34. Eya3 is a member of the Eya protein family (Eya1-4) whose members regulate embryonic development through their ability to participate in transcriptional regulation and phosphorylation signaling^35,36,37. While it is established that Eya proteins contain a conserved C-terminal tyrosine phosphatase domain, recent work demonstrated that Eya isoforms also exhibit serine/threonine phosphatase activity through the ability of their N termini to bind and recruit PP2A:B55 holoenzymes^38,39. Recently, it was shown that Eya3 recruits PP2A:B55 to c-Myc, where it leads to the dephosphorylation of T58 and, thus, regulates c-Myc stability and activity⁴⁰. Here, we use cryo-electron microscopy (cryo-EM), nuclear magnetic resonance (NMR) spectroscopy and binding and activity assays to molecularly define how PP2A:B55 recruits p107 and Eya3 and understand how these binding interactions facilitate substrate dephosphorylation.

Results

Mapping the interaction of p107 and Eya3 with B55 and PP2A:B55

We and others previously showed that p107 binds B55 and PP2A:B55 using its intrinsically disordered interdomain region (aa 594–783). NMR data for p107 aa 589–731 are shown in Extended Data Fig. 1a–c, wherein aa 619–625 are essential for B55 binding; pulldown assays with the p107 interdomain region (aa 612–687; hereafter referred to as p107) confirmed PP2A:B55 binding (Extended Data Fig. 2a,b)³³. The two-dimensional (2D) [¹H,¹⁵N] heteronuclear single quantum coherence (HSQC) spectrum of p107_589–731 and the B55-interacting p107 interdomain (p107_612–687; Fig. 2a) confirmed that it is an intrinsically disordered protein (IDP) with modest preferred secondary structure propensities according to the chemical shift index (CSI) (Extended Data Fig. 2c). Similarly, the 2D [¹H,¹⁵N] HSQC spectrum of the unbound Eya3 PP2A:B55-interacting domain (Eya3_27–155 and Eya3_62–108EE) also confirmed that it is an IDP with a single lowly populated α-helix (Fig. 2b, Extended Data Fig. 3a and Supplementary Fig. 1).

Fig. 2: The p107 and Eya3 interactions with B55_LL and PP2A:B55 triple complex.

a, The 2D [¹H,¹⁵N] HSQC spectrum of ¹⁵N-labeled p107 with (orange) and without (black) B55_LL. Peaks missing in the presence of B55_LL are labeled. b, The 2D [¹H,¹⁵N] HSQC spectrum of ¹⁵N-labeled Eya3 with (orange) and without (black) B55_LL. Peaks missing in the presence of B55_LL are labeled. c, Plot of peak intensity versus p107 protein sequence for the spectrum in a. d, Plot of peak intensity versus Eya3 protein sequence for the spectrum in b. e, The 2D [¹H,¹⁵N] HSQC spectrum of ¹⁵N-labeled p107 with (pink) and without (black) PP2A:B55. f, The 2D [¹H,¹⁵N] HSQC spectrum of ¹⁵N-labeled Eya3 with (pink) and without (black) PP2A:B55. g, Plot of peak intensity versus p107 protein sequence for the spectrum in e. h, Plot of peak intensity versus Eya3 protein sequence for the spectrum in f.

Both p107 and Eya3 were previously shown to bind PP2A:B55 holoenzymes^33,40. To test this, we incubated ¹⁵N-labeled (NMR-active) p107 or Eya3 with or without monomeric B55 (B55_LL; in which the PP2Aa interaction loop (aa 126–164) is replaced with residues NG, rendering it unable to bind the core PP2A enzyme PP2Aa:PP2Ac) and then overlaid the 2D [¹H,¹⁵N] HSQC spectra of p107 or Eya3 to identify peaks (residues) with reduced intensities (peaks with reduced intensities are because of a direct interaction, a dynamic charge–charge interaction or an intermediate timescale conformational exchange; Fig. 2a–d and Extended Data Figs. 1 and 3). For p107, residues 614–630 experienced the strongest reductions in intensity (Fig. 2c). These residues are identical to those previously observed to exhibit reduced intensities with wild-type (WT) B55, confirming not only that p107 binds B55 but also that the B55 PP2Aa binding loop that is missing in B55_LL is dispensable for p107 binding. Similar results were obtained with Eya3. Specifically, in the presence of B55_LL, Eya3 residues 68–83 showed reduced intensities (Fig. 2d), confirming that Eya3 binds directly to B55_LL. Notably, the extent of the interaction for both p107 and Eya3 to B55_LL is ~15 residues.

We then repeated the NMR interaction analysis of p107 and Eya3 with the PP2A:B55 holoenzyme (PP2Aa:B55:PP2Ac; purified, active PP2A:B55 methylated on PP2Ac mL309)²⁷. For both interactors, additional residues showed reductions in intensity with PP2A:B55 (Fig. 2e–h; p107, aa 614–685; Eya3, aa 63–103), compared to B55_LL alone. These data show that residues outside of the B55-interacting domains of p107 and Eya3 interact with PP2Aa and/or PP2Ac. Furthermore, these data show that both p107 and Eya3 bind directly to the full PP2A:B55 holoenzyme with more than ≥20 residues, suggesting that they bind in manners more similar to those observed for ARPP19 and/or FAM122A, rather than the limited interactions observed for substrates and regulators that bind specifically to the B56 subunit in PP2A:B56 holoenzymes using a short B56-specific SLiM (LxxIxE)^7,10.

PP2A:B55–p107 and PP2A:B55–Eya3 cryo-EM structures

We determined the cryo-EM structures of PP2A:B55–p107 and PP2A:B55–Eya3 at average resolutions of 2.6 and 2.7 Å, respectively (Fig. 3a,b, Extended Data Figs. 4–6 and Table 1). The previously solved cryo-EM structure of PP2A:B55 bound to FAM122A (ref. ²⁷) was used to model the PP2Aa, B55 and PP2Ac subunits. Both complexes adopted the same contracted horseshoe-shaped conformation of PP2Aa observed in the inhibitor bound structures of PP2A:B55–ARPP19 and PP2A:B55–FAM122A. Clear density was observed for the methylated PP2Ac C terminus (aa 294–309), which, as observed previously, extends across the PP2Aa central cavity to bind an extended pocket at the B55:PP2Aa interface, positioning mL309_C to bind a hydrophobic pocket in PP2Aa. These structures confirm that the contracted conformation of PP2A:B55 is largely unaffected by substrate, regulator or inhibitor binding and, instead, is dictated by the methylation state of the C-terminal residue of PP2Ac. In both complexes, unaccounted for density was observed at the B55 platform (Fig. 3c,d), which comprises loops aa 177–179, aa 197–199 and aa 222–231 and contains hydrophobic pockets 1–3 and acidic binding pocket 4, while the B55 platform wall comprises loops aa 278–287 and aa 336–345 and contains hydrophobic pockets 3 and 4. This density was attributed to p107 (residues 615–635) or Eya3 (residues 72–84). Although they bind the same surface of B55, they do so in distinct conformations, with p107 forming a helix and Eya3 binding in an extended conformation (Fig. 3e).

Fig. 3: Structures of the PP2A:B55–p107 and PP2A:B55–Eya3 complexes.

a, Cryo-EM map and model of PP2A:B55–p107. Two views of the map (top) are shown with the corresponding view of the molecular model (bottom). b, Cryo-EM map and model of PP2A:B55–Eya3. c, Overlay of p107 (salmon) and Eya3 (dark magenta) bound to PP2A:B55 (PP2A colored as in Fig. 1a). d, Close-up view of p107 and Eya3 bound to B55, with interacting residues shown as sticks. Key interacting pockets on B55 are labeled and colored (pockets in the platform, beige; pockets in the wall, green). e, Same as d, but only p107 and Eya3 are shown with the N-terminal and C-terminal residues labeled.

Table 1 Cryo-EM data collection, refinement and validation statistics

PP2A:B55 recruitment of p107

p107 binds B55 as a helix, with the residues N-terminal to the helix binding the B55 wall and those C-terminal extending toward the PP2Ac catalytic site (Fig. 4a). The interaction of the p107 N terminus (⁶¹⁵SPLMHP⁶²⁰) with the B55 wall is anchored by L617_p107, which binds B55 hydrophobic pocket P7 (F280_B55, F281_B55, I284_B55, Y337_B55 and F343_B55; Fig. 4b), and H619_p107, which binds B55 hydrophobic/polar pocket P8 (S247_B55, E283_B55, I284_B55, S287_B55, F343_B55 and K345_B55; Fig. 4c). At P620_p107, p107 kinks to bind across the base of the B55 platform, forming a three-turn helix (621–632) that extends toward the PP2Ac catalytic pocket, with p107 residues 633–635 abutting PP2Ac (Fig. 4a). While residues 634–687 of p107 were not sufficiently ordered to be modeled, our NMR data suggest that they contribute to binding through a dynamic (fuzzy) charge–charge interaction^5,24,41 (Fig. 2g).

Fig. 4: PP2A:B55 preferentially dephosphorylates p107 CDK2–cyclin A2 phosphosite pS640.

a, PP2A:B55–p107 showing that p107 (salmon, shown as sticks) binds the B55 platform (beige) and wall (green). B55 interaction pockets occupied by p107 are labeled and their positions are indicated by arrows. Key p107 interacting residues are labeled. The interaction sequence is shown below, with residues that mediate intermolecular contacts underlined and those important for intramolecular contacts in italics. b, p107 L617 (salmon) binds the B55 P7 hydrophobic pocket (P5 pocket residues shown as sticks). c, p107 H619 (salmon) binds the B55 P8 hydrophobic/polar pocket (P6 pocket residues are shown as sticks; polar interactions are shown with a black dotted line). d, Overlay of p107 (salmon) and the FAM122A B55-binding helix (orange), with side chains shown as sticks. Structural alignment of p107 and FAM122A interacting residues are also shown. e, p107 R621 (salmon) binds the B55 P4 acidic pocket (P4 pocket and key p107 residues are shown as sticks; the multiple intermolecular and intramolecular interactions that stabilize the conformation are shown as black dotted lines). f, p107 V622 and V625 bind the B55 hydrophobic P2 and P1 pockets (P1 and P2 pocket residues are shown as sticks), respectively, while p107 R626 binds the hydrophobic/polar P3 pocket (P3 pocket residues are shown as sticks, with hydrogen bonds between R626 and the ²²²MEELT²²⁶ carbonyls indicated by dashed lines). g, PP2A:B55 inhibition by p107 and Eya3 (mean ± s.d.; n = 3 experimental replicates). IC₅₀ values are reported in Extended Data Table 1. h, Cartoon illustrating p107 phosphorylation (CDK2 and cyclin A2) and subsequent time course of dephosphorylation (PP2A:B55) assays. i, The 2D [¹H,¹⁵N] HSQC spectrum of ¹⁵N-labeled unphosphorylated p107 (black) overlaid on CDK2–cyclin A2-phosphorylated ¹⁵N-labeled p107 (purple); S615 and S640 are phosphorylated, with shifted peaks labeled. j–n, The 2D [¹H,¹⁵N] HSQC spectra of unphosphorylated ¹⁵N-labeled p107 (black) overlaid with on CDK2–cyclin A2-phosphorylated ¹⁵N-labeled p107 incubated with PP2A:B55 (red) at time points of 45 min (j), 129 min (k), 213 min (l), 409 min (m) and 1,277 min (n). o, Changes in peak intensities of ¹⁵N-labeled p107 residues pS640 (red) and S640 (black).

Source data

Previously, we showed that the PP2A:B55-specific inhibitor FAM122A binds B55 using a helix and that the FAM122A residues that mediate B55 binding (using a short helical motif (SHelM), defined as R-L/V-x-x-I/V-K/R-x-E/D)²⁷ are conserved in the B55-binding domain of p107, suggesting that they bind using similar mechanisms. The structure of p107 bound to PP2A:B55 confirms that this is the case, with p107 ⁶²¹RVKEVRTD⁶²⁸ residues binding the B55 platform in a manner identical to FAM122A ⁸⁴RLHQIKQE⁹¹ (Fig. 4d). Like R84_FAM122A, R621_p107 forms a bidentate salt bridge with D197_B55 (pocket P4), a residue previously shown to be essential for p107 binding³³. The conformation of R621_p107 is stabilized by intramolecular electrostatic interactions with E624_p107 and D628_p107 and an intermolecular π-stacking interaction with H179_B55 (pocket P4) (Fig. 4e). Similarly, V622_p107 and V625_p107 anchor p107 to the B55 platform through the same hydrophobic pockets used by L85_FAM122A and I88_FAM122A (bound to pockets P2 and P1, respectively; P2: L225_B55 and V228_B55; P1: Y178_B55, L198_B55, M222_B55, L225_B55 and V228_B55). Lastly, R626_p107, like K89_FAM122A, forms multiple hydrogen bonds with the carbonyls of B55 loop ²²²MEELT²²⁶ (P3) (Fig. 4f). C-terminal to the B55-binding helix, p107 residues 631–633 extend with less well-defined density toward the PP2Ac active site, suggesting that p107 may transiently hinder active site access. To test whether p107 binding alters PP2Ac catalytic activity, we performed half-maximal inhibitory concentration (IC₅₀) measurements using the active site phosphomimetic DiFMUP as a pseudosubstrate. The data show that, as predicted, p107 inhibited DiFMUP dephosphorylation (IC₅₀: 245 ± 27 nM; Extended Data Table 1 and Fig. 4g).

p107 dephosphorylation by PP2A:B55

While p107 slows the dephosphorylation of DiFMUP, p107 itself is a protein substrate of PP2A:B55. p107 is phosphorylated by multiple kinases, including the proline-directed serine/threonine CDK2–cyclin A2. Of the 14 S and T residues in the p107 interdomain (aa 612–687), S615_p107, S640_p107 and S650_p107 are followed by a proline and all have been observed to be phosphorylated in multiple phosphoproteomics studies (PhosphoSitePlus)⁴²; S640_p107, in particular, is important for determining the activation status of p107, with phosphorylation inactivating p107 and dephosphorylation activating p107 (ref. ³¹). To identify the p107 residues phosphorylated by CDK2–cyclin A2, we incubated ¹⁵N-labeled p107 with purified CDK2–cyclin A2 and then identified the phosphorylated residues using NMR spectroscopy (Fig. 4h, step 1). The 2D [¹H,¹⁵N] HSQC spectrum of phospho-p107 (pp107) showed that CDK2–cyclin A2 fully phosphorylates ⁶¹⁵SP⁶¹⁶ and ⁶⁴⁰SP⁶⁴¹ (Fig. 4i); ⁶⁵⁰SP⁶⁵¹ was not phosphorylated, likely because of its proximity to p107’s cyclin A2-recruiting motif (⁶⁵⁴GSAKRRLFGE⁶⁶³)⁴³. We then monitored the kinetics of pS615_p107 and pS640_p107 dephosphorylation by PP2A:B55 by following the intensity changes in the phosphorylated (pS615_p107 and pS640_p107) or unphosphorylated (S615_p107 and S640_p107) HN/N cross peaks in the 2D [¹H,¹⁵N] HSQC spectrum (Fig. 4h–o, step 2). While pS640_p107 is rapidly dephosphorylated (Fig. 4j–l,o), the dephosphorylation of pS615_p107 is comparatively slow, with little to no dephosphorylation observed until pS640_p107 is almost completely dephosphorylated (Fig. 4l–n and Extended Data Figs. 7 and 8). These results are fully consistent with the cryo-EM structure of the PP2A:B55–p107 substrate complex. Specifically, S615_p107 binds directly to B55 in a pocket more than 40 Å away from the PP2Ac active site, making it unable to be dephosphorylated when bound to B55. In contrast, S640_p107 is positioned next to the PP2Ac active site when bound to B55, making it ideally located for rapid dephosphorylation. Together, these data suggest that pS640_p107 is the primary phosphosite targeted by PP2A:B55.

PP2A:B55 recruitment of Eya3

While p107 bound B55 as predicted, the lack of the B55-specific SHelM in Eya3, coupled with the lack of a preferred secondary structure in the unbound state (Extended Data Fig. 3a), suggested that it may bind B55 through a distinct mechanism. The structure of PP2A:B55–Eya3 confirms that Eya3 binds B55 in an extended rather than helical manner (Fig. 5a). Despite this difference in conformation, the B55 interaction pockets engaged by Eya3 mirror those used by other B55-specific regulators. First, Y72_Eya3, K75_Eya3 and Y77_Eya3 bind hydrophobic pockets P8, P9 and P2, respectively (Fig. 5b), anchoring the N terminus of Eya3 to the center of B55. This positions Eya3 H79_Eya3 (a residue that, when substituted to Ala, was previously shown to abolish B55 binding⁴⁰) to bind the B55 acidic pocket, P4. H79_Eya3 binds D197_B55 from the opposite direction compared to R621_p107 in the p107 complex, allowing it to form a nearly perfect π-stacking interaction with H179_B55 (Fig. 5c). This interaction is further stabilized by I80_Eya3 (bound to P1) and L81_Eya3 (bound to P3) (Fig. 5c). Lastly, as observed for p107, the C-terminal Eya3 residues ⁸²SVP⁸⁴ extend toward the PP2Ac active site. To test whether Eya3 influences the PP2Ac catalytic activity against a small phosphomimic, we again performed IC₅₀ measurements using DiFMUP as a substrate, which showed weak inhibition (IC₅₀: 1102 ± 84 nM; Extended Data Table 1 and Fig. 4g).

Fig. 5: Eya3 binds B55 in an extended conformation through a conserved interaction sequence.

a, PP2A:B55–Eya3 showing that Eya3 (purple, shown as sticks) binds the B55 platform (beige) and wall (green). B55 interaction pockets occupied by Eya3 are labeled and their positions are indicated by arrows. Key Eya3 interacting residues are labeled. The interaction sequence is shown below, with residues that mediate intermolecular contacts underlined. b, Interactions between Eya3 (purple) and B55 (wall, green; platform, beige) at pockets P2, P8 and P9 shown. Interacting residues are shown as sticks and labeled. Hydrogen bonds and salt bridge interactions are shown as black dotted lines. c, Interactions between Eya3 (purple) and B55 (platform; beige) at pockets P1, P3 and P4 are shown. Interacting residues are shown as sticks and labeled. Hydrogen bonds and salt bridge interactions are shown as black dotted lines. The π-stacking interaction between H79_Eya3 and H179_B55 is indicated by gray shading. d, Sequence conservation logo for the Eya3 B55-interacting domain identified using Jackhmmer⁴⁸. The key interaction region is highlighted in gray. e, Sequence alignment of the Eya3 B55-interacting domain with the corresponding residues in Eya1, Eya2 and Eya4. Residues corresponding to H79_Eya3 are in bold. f. Expi293F cell lysates expressing GFP–B55, 3×FLAG–PP2Ac and 6×His–PP2Aa were incubated with or without purified Eya2 (residues 40–84, including the putative B55-interacting domain), immunopurified using GFP trap beads (EGFP nanobodies coupled to agarose resin) and isolated proteins detected by immunoblot. Results are representative of three independent experiments (Extended Data Fig. 2a). g, Same as f, except the lysate was incubated with either purified Eya3 or Eya3_H79R. (residues 62–108). h, Quantification (mean ± s.d.; n = 3 experimental replicates) of Eya3 binding observed in g, with individual data points shown as dots.

Source data

The Eya3 residues that mediate the core interaction with B55 are also highly conserved, especially Y77_Eya3 and H79_Eya3 (Fig. 5d). In particular, the latter residue was previously shown to be essential for PP2A:B55 binding, as the Eya3 H79A variant severely reduces its ability to recruit PP2A:B55 and, in turn, nearly abolishes its threonine-specific phosphatase activity^39,40. Comparing the Eya3 B55-interacting sequence with the remaining Eya isoforms (Eya1, Eya2 and Eya4) showed that this region is conserved among the Eya family, with the greatest differences in sequence observed for Eya2, the isoform previously identified to exhibit the highest threonine phosphatase activity (Fig. 5e)⁴⁰. We first tested whether the Eya2 N-terminal IDP (aa 40–84), which includes the putative B55-interacting residues ⁵⁰FSRSCPRVLPRQ⁶¹, binds directly to PP2A:B55 using pulldown assays. The data show that Eya2 residues 40–84 are sufficient to bind and recruit PP2A:B55 (Fig. 5f). We then generated the Eya3 H79R variant, in which H79_Eya3 was replaced with the equivalent residue in Eya2 (R56_Eya2), to test whether an arginine impacts PP2A:B55 recruitment. The data show that the B55 binding was statistically equivalent (Fig. 5g,h); thus, either residue binds the B55 P4 binding pocket effectively. Consistent with this observation, modeling an arginine in this position in the PP2A:B55–Eya3 structure showed that H79R is, like H79, optimally positioned to form both a π-stacking interaction with H179_B55 and a salt bridge with D197_B55 (Extended Data Fig. 9).

Discussion

A molecular understanding of how PP2A:B55 recruits its substrates and regulators is necessary to identify its endogenous substrates on a proteome-wide scale. The cryo-EM structures of PP2A:B55 in complex with two B55-specific protein inhibitors, FAM122A and ARPP19, showed that they interact with PP2A:B55 in unexpected manners²⁷. First, while most PPP-specific regulators and substrates bind their cognate PPPs in extended conformations through SLiMs, FAM122A and ARPP19 bind PP2A:B55 using helices (SHelMs). Second, the structures also showed that, while FAM122A and ARPP19 bind extensively to both B55 and PP2Ac, they do so in distinct conformations. In addition, the recently determined structure of the PP2A:B55–IER5 complex²⁸ showed that IER5 forms a helix–turn–helix bundle that engages B55 through a third conformation. These data suggest that regulators and substrates use diverse strategies to bind and recruit PP2A:B55. To define the mechanisms used by substrates and regulators to recruit PP2A:B55 and to understand how these interactions direct PP2A:B55 activity, we determined the cryo-EM structures of a B55-specific substrate, p107, and a regulator, Eya3, bound to PP2A:B55 and determined how these interactions direct substrate dephosphorylation.

Our data show that both p107 and Eya3 bind to B55 using B55 interaction surfaces also used by FAM122A, ARPP19 and IER5 (that is, the B55 wall and B55 platform) (Fig. 6a). Thus, while two interaction surfaces of B55 engage multiple regulators (wall and platform), others (thus far) are used exclusively by ARPP19 (entry and hook); future work will be needed to determine whether these latter interaction surfaces are truly specific to ARPP19 or are used by other regulators and/or substrates. In addition, only the inhibitors FAM122A and ARPP19 stably bound the PP2Ac active site (Fig. 6b).

Fig. 6: Combinatorial mechanisms used by B55-specific regulators, substrates and inhibitors contribute to PP2A:B55 phosphosite selectivity.

a, Overlay of PP2A:B55 (B55 in lavender with interaction sites colored (platform, beige; wall, green; entry, pink; hook, yellow); PP2Ac, dark gray; PP2Aa, white) bound to p107 (salmon), eya3 (purple), FAM122A (orange), ARPP19 (dark blue) and IER5 (gray). The location of the PP2Ac catalytic pocket is indicated in red. b, Same as a, but individual complexes, separated by regulators or substrates and inhibitors. c, Location of individual binding pockets on B55, colored as in a. d, B55-interacting residues bound at B55 pockets P1, P2 and P4. B55 pocket residues, beige sticks; B55-interacting residues, sticks colored as in a. e, Same as d, but for B55 pocket P3. f, Same as d, but for B55 pockets P7–P9. g, B55 pockets, with residues defining the pockets indicated and the corresponding B55-interacting residues listed. h, Cartoon illustrating how B55-specific recruitment contributes to PP2A:B55 phosphosite dephosphorylation selectivity.

While p107 and Eya3 bind the same interaction surfaces, they do so in distinct conformations. p107 adopts a helical conformation, overlapping almost perfectly with the B55-interacting helix of FAM122A (this was predicted as their sequences define the only known B55-specific SHelM, R-L/V-x-x-I/V-K/R-x-E/D)²⁷. In contrast, Eya3 binds B55 in an extended manner. Superimposing all five B55 complexes with regulators and substrates showed that only p107 and FAM122A bind B55 in identical conformations; the remaining regulators bind the wall and platform pockets in distinct conformations (Fig. 6c–g). This shows that B55-interacting proteins (substrates and regulators) use different combinatorial strategies to recruit B55 through the wall and platform binding pockets, suggesting that predicting novel B55 substrates and regulators from sequence alone will be challenging.

In addition to defining substrate binding, the PP2A:B55–p107 structure, coupled with atomic-resolution dephosphorylation assays, showed that, while the p107 interdomain is robustly phosphorylated by CDK2–cyclin A2 on two residues, S615_p107 and S640_p107, it is only pS640_p107 and not pS615_p107 that is readily dephosphorylated by PP2A:B55 (Fig. 4). This preference in phosphosite dephosphorylation is readily explained by the PP2A:B55–p107 structure. Specifically, in the bound state, S615_p107 is associated with B55 at the wall interaction surface, more than 40 Å away from the PP2Ac catalytic pocket. In contrast, in the bound state, S640_p107 is positioned directly adjacent to the active site (Fig. 6h, top), making substrate engagement dephosphorylation readily achievable. In this way, B55 binding ensures that only specific phosphosites (that is, pS640_p107, a residue whose phosphorylation status determines p107 activity³¹) are preferentially dephosphorylated (Fig. 6h, bottom).

Lastly, our work also highlights the importance of combining cryo-EM with NMR spectroscopy to gain a comprehensive understanding of how IDP regulators and substrates engage their cognate folded binding partners. By studying the ensemble of IDP regulators in their unbound and bound forms, we showed that, while FAM122A and ARPP19 have a preference for prepopulated secondary structures in their unbound forms, this behavior has not been observed (thus far) for regulators or substrates. The data also show that, while cryo-EM identified the most highly populated conformations of substrates and regulators in the bound state, residues outside this region are also critical for binding (Fig. 2e,f)⁴⁴. Indeed, for both p107 and Eya3, residues beyond the core B55-interacting domains showed a loss of intensity in the presence of PP2A:B55, likely corresponding to a dynamic, charge–charge (fuzzy) interaction with the surface of PP2Ac or PP2Aa (refs. ^45,46,47). These data are fully consistent with previous PP2A:B55 work. For example, in addition to the well-ordered interactions of ARPP19 with the B55 wall, platform and hook, cryo-EM and especially NMR spectroscopy showed that a prepopulated helix in the free state dynamically engages the B55 entry interaction surface, an interaction that is essential for the simultaneous binding of ARPP19 and FAM122A to PP2A:B55 (ref. ²⁷). The importance of dynamic interactions in regulator and substrate recruitment by PPPs is true not only for PP2A:B55 but also for PP2A:B56 (ref. ²⁴), PP2B (ref. ⁵) and PP1 (ref. ⁴¹). NMR spectroscopy is the only technique that captures these dynamic interactions at atomic resolution; thus, combining NMR spectroscopy with other high-resolution methods is essential to obtain a comprehensive understanding of how PPP-specific regulators and substrates recruit their cognate PPP.

Together, these studies provide a molecular understanding of regulator and substrate recruitment of the PP2A:B55 holoenzyme. Because of the key regulatory functions of PP2A:B55 in mitosis and DNA damage repair, these data provide a roadmap for characterizing disease-associated mutations and pursuing novel avenues to therapeutically target this complex, by individually blocking a subset of regulators that use distinct B55-interacting sites.

Methods

Bacterial protein expression

Human PP2Aa_9–589, p107_612–687, EYA2_40–84EEE, Eya3_62–108EE and Eya3_27–155 were subcloned into pTHMT containing an N-terminal His₆-tag followed by maltose-binding protein (MBP) and a tobacco etch virus (TEV) protease cleavage site. Human p107_589–731 was subcloned into RP1B containing an N-terminal His₆-tag followed by a TEV protease cleavage site. For expression, plasmid DNAs were transformed into Escherichia coli BL21(DE3) cells (Agilent). Freshly transformed cells were grown at 37 °C in Luria–Bertani broth containing kanamycin (50 µg ml⁻¹) until they reached an optical density at 600 nm (OD₆₀₀) of ~0.8. Protein expression was induced by addition of 1 mM IPTG to the culture medium and cultures were allowed to grow overnight (18–20 h, 250 rpm shaking) at 18 °C. Cells were harvested by centrifugation (8,000g, 15 min, 4 °C) and stored at −80 °C until purification. Expression of uniformly ¹³C-labeled and/or ¹⁵N-labeled protein was carried out by growing freshly transformed cells in M9 minimal medium containing 4 g L⁻¹ [¹³C]d-glucose and/or 1 g L^{−1 15}NH₄Cl (Cambridge Isotopes Laboratories) as the sole carbon and nitrogen sources, respectively. The Eya3 H79R variant was generated by site-directed mutagenesis, sequence-verified and expressed as described above. For pulldown experiments, p107, EYA2, EYA3 and EYA3_H79R were expressed with a C-terminal StrepII tag for detection using an anti-Strep antibody (Genscript, A01732; 1:1,000).

Cell culture

Expi293F cells were obtained from Thermo Fisher Scientific (A14527) and grown in HEK293 cell complete medium (SMM293-TII, Sino Biological, M293TII). For transient overexpression of B55 and PP2Ac constructs, cells were transfected using polyethyleneimine (PEI) transfection reagent. For western blot and immunoprecipitation studies, whole-cell extracts were prepared by lysing cells in ice-cold lysis buffer (20 mM Tris pH 8.0, 500 mM NaCl, 0.5 mM TCEP, 1 mM MnCl₂, 0.1% Triton X-100 and phosphatase inhibitor cocktail (Thermo Fisher Scientific)), sonicating and clearing the lysate by centrifugation at 15,000g for 20 min at 4 °C. Total protein concentrations were measured using the Pierce 660 protein assay reagent (Thermo Fisher Scientific).

Mammalian protein expression

Full-length B55_1–477 was cloned into pcDNA3.4 including an N-terminal GFP tag followed by a TEV cleavage sequence. Full-length PP2Ac₁–₃₀₉ was cloned into pcDNA3.4 with an N-terminal Strep tag followed by a TEV cleavage sequence. B55_LL (Fig. 1c) was cloned into pcDNA3.4 with an N-terminal GFP tag followed by a TEV cleavage sequence. All plasmids were amplified and purified using the NucleoBond Xtra Maxi Plus EF (Macherey-Nagel). B55_WT and B55_LL were individually expressed in Expi293F cells (Thermo Fisher Scientific). B55_1–477 and PP2Ac_1–309 were coexpressed in Expi293F cells at a 1:2 DNA ratio.

Transfections were performed in 500 ml of medium (SMM293-TII, Sino Biological) in 2-L flasks using PEI (Polysciences) reagent according to the manufacturer’s protocol in an incubator at 37 °C and 8% CO₂ under shaking (125 rpm). On the day of transfection, the cell density was adjusted to 2.8 × 10⁶ cells per ml using fresh SMM293-TII expression medium. DNA of PP2Ac and B55 (2:1 ratio) was diluted in Opti-MEM reduced-serum medium (Thermo Fisher Scientific). Similarly, in a separate tube, PEI at three times the amount of DNA was diluted in the same volume of Opti-MEM reduced-serum medium (Thermo Fisher Scientific). The DNA and PEI mixtures were combined and incubated for 10 min at room temperature, before being added to the cell culture. Valproic acid (2.2 mM final concentration; Sigma) was added to the cells 4 h after transfection; then, 24 h after transfection, sterile-filtered glucose (4.5 ml per 500 ml of cell culture, 45%, glucose stock) was added to the cell culture flasks to boost protein production. Cells were harvested 48 h after transfection by centrifugation (2,000g for 20 min, 4 °C) and stored at −80 °C. Protein concentration was measured using the Pierce 660-nm assay (Thermo Fisher Scientific) and a BMG Labtech plate reader (Clariostar^Plus, Clariostar version 4.20, build 17).

p107, Eya3 and Eya2 purification

All p107, Eya2 and Eya3 constructs followed the same purification protocol. Cell pellets were resuspended in ice-cold lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X-100 and a protease inhibitor tablet (Thermo Fisher Scientific)) and lysed by high-pressure cell homogenization (Avestin Emulsiflex C3). Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C) and the supernatant was filtered with 0.22-µm syringe filters (Millipore). The proteins were loaded onto a HisTrap HP column (Cytiva) pre-equilibrated with buffer A (50 mM Tris pH 8.0, 500 mM NaCl and 5 mM imidazole) and eluted using a linear gradient (0–60%) with buffer B (50 mM Tris pH 8.0, 500 mM NaCl and 500 mM imidazole). Fractions containing the protein were pooled and dialyzed overnight at 4 °C with TEV protease (in-house; His₆-tagged) to cleave the His₆–MBP tag. Following cleavage, the sample was loaded under gravity onto Ni²⁺-NTA beads (Prometheus) pre-equilibrated with buffer A; then, the flowthrough and wash A fractions were collected and heat-purified by incubating the samples at 80 °C for 20 min. Samples were centrifuged at 15,000g for 10 min to remove precipitated protein. Supernatant was concentrated and purified using size-exclusion chromatography (SEC; Superdex 75 16/60, Cytiva) in either p107 NMR buffer (10 mM Na₃PO₄ pH 6.3, 150 mM NaCl and 0.5 mM TCEP), Eya3 NMR buffer (10 mM Na₃PO₄ pH 6.5, 200 mM NaCl and 0.5 mM TCEP) or IC₅₀ assay buffer (20 mM Tris pH 8.0, 150 mM NaCl and 0.5 mM TCEP). Purified samples were again heat-purified (80 °C for 5 min) and were either directly used for NMR data collection or flash-frozen and stored at −80 °C.

PP2Aa purification

Cell pellets expressing PP2Aa_9–589 were resuspended in ice-cold lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X-100 and an EDTA-free protease inhibitor tablet (Thermo Fisher Scientific)) and lysed by high-pressure cell homogenization (Avestin Emulsiflex C3). Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C) and the supernatant was filtered with 0.22-µm syringe filters (Millipore). The proteins were loaded onto a HisTrap HP column (Cytiva) pre-equilibrated with buffer A (50 mM Tris pH 8.0, 500 mM NaCl and 5 mM imidazole) and eluted using a linear gradient (0–40%) with buffer B (50 mM Tris pH 8.0, 500 mM NaCl and 500 mM imidazole). Fractions containing the protein were pooled and dialyzed overnight at 4 °C with TEV protease (in-house; His₆-tagged) to cleave the His₆–MBP tag and loaded under gravity onto Ni²⁺-NTA beads (Prometheus) pre-equilibrated with buffer A. Flowthrough and wash A fractions were collected, concentrated and loaded onto a HiTrap Q HP column (Cytiva) for further purification. The proteins were eluted with a 100 mM–1 M salt gradient (buffer A: 20 mM Tris pH 8.0, 100 mM NaCl and 0.5 mM TCEP; buffer B: 20 mM Tris pH 8.0, 1 M NaCl and 0.5 mM TCEP). PP2Aa fractions were concentrated and further purified using SEC (Superdex 200 26/60 [Cytiva]) in assay buffer (20 mM Tris pH 8.0, 150 mM NaCl and 0.5 mM TCEP). Samples were either directly used or flash-frozen and stored at −80 °C.

CDK2–cyclin A2 kinase expression and purification

CDK2 was subcloned into pGEX3C. Cyclin A2_156–429 was subcloned into pET16b. The plasmids were transformed separately into E. coli BL21(DE3) cells (Agilent). Freshly transformed cells were grown at 37 °C in LB broth containing carbenicillin (50 µg ml⁻¹) until they reached an OD₆₀₀ of ~0.8. Protein expression was induced by addition of 1 mM IPTG to the culture medium and cultures were allowed to grow overnight (18–20 h, 250 rpm shaking) at 18 °C. Cells were harvested by centrifugation (8,000g, 15 min, 4 °C) and stored at −80 °C until purification. Cell pellets expressing CDK2 and cyclin A2 were resuspended in lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X-100 and a protease inhibitor tablet (Thermo Fisher Scientific)) and combined before high-pressure cell homogenization (Avestin Emulsiflex C3). Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C), and the supernatant was filtered with 0.22-µm syringe filters (Millipore). The proteins were loaded onto a HisTrap HP column (Cytiva) pre-equilibrated with buffer A (50 mM Tris pH 8.0, 500 mM NaCl and 5 mM imidazole) and eluted using a linear gradient (0–60%) with buffer B (50 mM Tris pH 8.0, 500 mM NaCl and 500 mM imidazole). Fractions containing both proteins were pooled and dialyzed overnight at 4 °C in SEC buffer (20 mM Tris pH 7.5, 500 mM NaCl and 1 mM DTT). The complex was concentrated and further purified using SEC (Superdex 200 26/60, Cytiva). Fractions containing the proteins were pooled and dialyzed in dialysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1 mM MgCl₂ and 0.005% Tween-20) overnight at 4 °C. The CDK2–cyclin A2 complex was mixed with 50% glycerol and stored at −80 °C.

p107 phosphorylation

Purified ¹⁵N-labeled p107 (25 μM) was incubated with CDK2–cyclin A2 kinase (500:1 ratio) in phosphorylation buffer (100 mM Tris pH 7.5, 2 mM DTT and 10 mM MgCl₂) with 500 µM adenosine triphosphate (Sigma) for phosphorylation. The kinase reaction was incubated at 37 °C for 1 h. pp107 was immediately heat-purified by incubation at 80 °C for 10 min. The sample was centrifuged at 15,000g for 10 min to remove precipitated CDK2–cyclin A2 and either immediately used for experiments or flash-frozen and stored at −80 °C. Complete phosphorylation was confirmed by monitoring the chemical shift changes of the phosphorylated serine residue(s) using NMR spectroscopy (2D [¹H,¹⁵N] HSQC spectrum).

p107, Eya3 and Eya2 interaction with the PP2A:B55 complex

Purified p107 (~25 μg), Eya3, Eya3_H79R or Eya2 (~25 μg) alone were mixed with Expi293F whole-cell extracts expressing B55, PP2Ac and PP2Aa constructs. Input samples were collected before incubation with GFP trap agarose beads (Extended Data Fig. 2a). GFP-tagged B55 and associated proteins were captured by incubating equal amounts of total protein (~500 μg) for each condition with GFP trap agarose beads prepared as described below at 4 °C for 2 h. Following three washes with wash buffer (20 mM Tris pH 8.0, 500 mM NaCl, 0.5 mM TCEP and 1 mM MnCl₂), bound proteins were eluted with 2% SDS sample buffer (90 °C, 10 min), resolved by SDS–PAGE (Bio-Rad) and transferred to PVDF membranes for western blot analysis using indicated antibodies (Nature Portfolio Reporting Summary) anti-B55 (Cell Signaling Technology, 2290 S; 1:1,000), anti-PP2Ac (Millipore, MABE1783; 1:1,000), anti-PP2Aa (Biolegend, 824901; 1:1000), anti-Strep (Genscript, A01732; 1:1,000), goat anti-rat IgG (Thermo Fisher Scientific, SA5-10024; 1:3,000), goat anti-rabbit IgG (Bio-Rad, 12005869; 1:3,000) and goat anti-mouse IgG (Bio-Rad, 12004158; 1:3,000). Antibody fluorescence signals were captured using a ChemiDoc MP Imaging System (Image Lab Touch Software 2.4; Bio-Rad) and band intensities were quantified using ImageJ 1.53t. Uncropped blots are provided in the Source Data.

EGFP–nanobody protein expression, purification and immobilization onto agarose beads

For expression, pOPIN_EGFP_nanobody plasmid DNA (gift from M. Bollen, KU Leuven) was transformed into E. coli BL21(DE3) cells (Agilent). Freshly transformed cells were grown at 37 °C in LB broth containing carbenicillin (50 µg ml⁻¹) until they reached an OD₆₀₀ of ~0.8. Protein expression was induced by addition of 0.5 mM IPTG to the culture medium and cultures were allowed to grow overnight (18–20 h, 250 rpm shaking) at 18 °C. Cells were harvested by centrifugation (8,000g, 15 min, 4 °C) and stored at −80 °C until purification. Cell pellets expressing EGFP–nanobody were resuspended in ice-cold lysis buffer (50 mM Tris pH 8.0, 500 mM NaCl, 5 mM imidazole, 0.1% Triton X-100 and an EDTA-free protease inhibitor tablet (Thermo Fisher Scientific)) and lysed by high-pressure cell homogenization (Avestin Emulsiflex C3). Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C) and the supernatant was filtered with 0.22-µm syringe filters. The proteins were loaded onto a HisTrap HP column (Cytiva) pre-equilibrated with buffer A (50 mM Tris pH 8.0, 500 mM NaCl and 5 mM imidazole) and eluted using a linear gradient (0–60% B) with buffer B (50 mM Tris pH 8.0, 500 mM NaCl and 500 mM imidazole). Fractions containing the protein were pooled, concentrated and further purified at room temperature using SEC (Superdex 75 26/60, Cytiva) in PBS pH 7.5. Purified and concentrated EGFP–nanobody protein was immobilized onto agarose beads (20 mg of protein per column) using an AminoLink Plus immobilization kit (Thermo Fisher Scientific) following the manufacturer’s instructions in PBS pH 7.5 coupling buffer.

B55_LL purification

Expi293F cell pellets expressing EGFP–B55_LL were resuspended in ice-cold lysis buffer (20 mM Tris pH 8.0, 500 mM NaCl, 0.5 mM TCEP, 0.1% Triton X-100 and an EDTA-free protease inhibitor tablet (Thermo Fisher Scientific)) and lysed by high-pressure cell homogenization (Avestin Emulsiflex C3). Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C) and the supernatant was filtered with 0.22-µm syringe filters (Millipore). Lysates were mixed with GFP–nanobody-coupled agarose beads prepared as described above, pre-equilibrated with wash buffer 1 (20 mM Tris pH 8.0, 500 mM NaCl and 0.5 mM TCEP) and slowly rocked at 4 °C for 2 h. After 2 h, the lysate–bead mixture was loaded onto gravity columns, the first flowthrough (FT1) was collected and the column was washed three times with 25 ml of wash buffer 1 (washes 1–3). The GFP–B55 resin was resuspended in wash buffer 2 (20 mM Tris pH 8.0, 250 mM NaCl and 0.5 mM TCEP) and TEV was added for on-column cleavage with rocking overnight at 4 °C. The second flowthrough was collected (FT2) and the resin was washed once with 20 ml of wash buffer 2 (wash 4) and twice with 20 ml of wash buffer 1 (washes 5 and 6). FT2 and washes 4–6 were collected, diluted to ~100 mM salt concentration (with 0 mM NaCl wash buffer) and loaded onto a HiTrap Q HP column (Cytiva) for further purification. The proteins were eluted with a 100 mM–1 M salt gradient (buffer A: 20 mM Tris pH 8.0, 100 mM NaCl and 0.5 mM TCEP; buffer B: 20 mM Tris pH 8.0, 1 M NaCl and 0.5 mM TCEP). B55_LL was concentrated and further purified using SEC (Superdex 200 26/60, Cytiva) in either p107 or EYA3 NMR buffer or assay buffer (20 mM Tris pH 8.0, 150 mM NaCl and 0.5 mM TCEP).

PP2A:B55 complex purification

Expi293F cell pellets expressing StrepII–PP2Ac and EGFP–B55 constructs were resuspended in ice-cold lysis buffer (20 mM Tris pH 8.0, 500 mM NaCl, 0.5 mM TCEP, 1 mM MnCl₂, 0.1% Triton X-100 and an EDTA-free protease inhibitor tablet (Thermo Fisher Scientific)) and lysed by high-pressure cell homogenization (Avestin Emulsiflex C3). Purified PP2Aa was added to the cell lysate. Cell debris was pelleted by centrifugation (42,000g, 45 min, 4 °C) and the supernatant was filtered with 0.22-µm syringe filters (Millipore). Lysates were loaded onto a GFP–nanobody-coupled agarose bead column prepared as described above, pre-equilibrated with wash buffer 1 (20 mM Tris pH 8.0, 500 mM NaCl, 1 mM MnCl₂ and 0.5 mM TCEP) and slowly rocked at 4 °C for 2 h. After 2 h, FT1 was collected and the column was washed three times with 25 ml of wash buffer (washes 1–3). The GFP–B55 resin was resuspended in 20 mM Tris pH 8.0, 250 mM NaCl, 1 mM MnCl₂ and 0.5 mM TCEP and TEV was added for on-column cleavage rocking overnight at 4 °C. FT2 was collected and the resin was washed once with 20 ml of wash buffer 2 (20 mM Tris pH 8.0, 250 mM NaCl, 1 mM MnCl₂ and 0.5 mM TCEP) (wash 4) and twice with 20 ml of wash buffer 1 (washes 5 and 6). FT2 and washes 4–6 were collected, diluted to ~100 mM salt concentration (with 0 mM NaCl wash buffer) and loaded onto a HiTrap Q HP column (Cytiva) for further purification. The proteins were eluted with a 100 mM–1 M salt gradient (buffer A: 20 mM Tris pH 8.0, 100 mM NaCl, 1 mM MnCl₂ and 0.5 mM TCEP; buffer B: 20 mM Tris pH 8.0, 1 M NaCl, 1 mM MnCl₂ and 0.5 mM TCEP). PP2A:B55 complex and B55 fractions were pooled, concentrated and further purified using SEC (Superdex 200 26/60, Cytiva) in either p107 or Eya3 NMR buffer or assay buffer (20 mM Tris pH 8.0, 150 mM NaCl, 1 mM MnCl₂ and 0.5 mM TCEP).

Cryo-EM sample preparation

The PP2A:B55–p107_612–687 complex was prepared by purifying PP2A:B55 and incubating it with a 1.5 molar ratio of PP2A:B55 to p107_612–687 at a total concentration of 2.4 mg ml⁻¹. The PP2A:B55–Eya3_62–108EE complex was prepared by purifying PP2A:B55 and incubating it with a 1.56 molar ratio of PP2A:B55 to Eya3_62–108EE at a total concentration of 2.4 mg ml⁻¹ (Extended Data Fig. 4). Immediately before blotting and vitrification (Thermo Fisher Scientific Vitrobot MK IV, 18 °C, 100% relative humidity), CHAPSO was added to a final concentration of 0.125% (w/v) for PP2A:B55–p107_612–687 and 0.1% (w/v) for PP2A:B55–Eya3_62–108EE. Then, 3.5 μl of the sample was applied to a freshly glow-discharged UltAuFoil 1.2/1.3 300-mesh grid (SPT Labtech), blotted for 5 s and plunged into liquid ethane.

Cryo-EM data acquisition and processing

For PP2A:B55–p107_612–687, imaging was performed at the Pacific Northwest Cryo-EM Center (PNCC) using a Thermo Fisher Scientific Titan Krios G3i (fringe-free) operating at an accelerating voltage of 300 keV equipped with a Gatan BioQuantum energy filter and K3 camera operated in super-resolution counting mode. Acquisition (SerialEM version 4.1) and imaging parameters are given in Table 1. All data processing steps were performed using cryoSPARC version 4.4.1 (hereafter cryoSPARC)⁴⁹ and are summarized in Extended Data Fig. 5. Super-resolution videos were imported into cryoSPARC and motion-corrected using patch motion correction. Initial contrast transfer function (CTF) parameters were then calculated using the patch CTF estimation. Micrographs were filtered to remove outliers in motion correction and/or CTF estimation results and screened manually to remove micrographs with notable nonvitreous ice contamination. The 2D class averages of particles picked by blob picker from 4,687 micrographs were used as templates for the template picker. Picks were subjected to 2D and three-dimensional (3D) classification and a subset used to train a Topaz⁵⁰ model, which was then used to pick 2.8 million particles. Multiple rounds of 2D and 3D classifications in cryoSPARC were used to remove junk particles from the initial picks, resulting in selected particles from classes showing a clear secondary structure and representing the full complex. The resolution in both datasets was then further improved by cycles of CTF parameter refinement and particle polishing. A final round of 3D classification was used to identify particles with well-resolved density for p107. The final map (nonuniform refinement, cryoSPARC) was refined from 195,950 particles to a resolution of 2.58 Å. For PP2A:B55–EYA3_62–108EE, imaging was performed at the National Institutes of Health (NIH) National Center for Cryo-EM Access and Training (NCCAT) using a Thermo Fisher Scientific Titan Krios G4 (fringe-free) operating at an accelerating voltage of 300 keV equipped with a Falcon 4i direct detector. Acquisition (Leginon 3.7) and imaging parameters are given in Table 1. All data processing steps were performed using cryoSPARC and are summarized in Extended Data Fig. 6. Electron event representation videos were imported into cryoSPARC and motion-corrected using patch motion correction. Initial CTF parameters were then calculated using the patch CTF estimation. Micrographs were filtered to remove outliers in motion correction and/or CTF estimation results and screened manually to remove micrographs with notable nonvitreous ice contamination. The 2D class averages of particles picked by blob picker from 4,084 micrographs were used as templates for the template picker. Picks were subjected to 2D and 3D classification and a subset used to train a Topaz model, which was then used to pick 874,057 particles. Multiple rounds of 2D and 3D classifications in cryoSPARC were used to remove junk particles from the initial picks, resulting in selected particles from classes showing clear secondary structure and representing the full complex. The resolution in both datasets was then further improved by cycles of CTF parameter refinement and particle polishing. A final round of 3D classification was used to identify particles with well-resolved density for Eya3. The final map (nonuniform refinement, cryoSPARC) was refined from 107,919 particles to a resolution of 2.7 Å. All global map resolutions reported in this work were calculated using gold-standard half-maps at Fourier shell correlation (FSC) = 0.143 as reported in cryoSPARC. The structures were visualized using PyMOL (Schrödinger) or UCSF ChimeraX (version 1.7.1)⁵¹.

Cryo-EM model building

All models were built and refined by iterating between manual rebuilding and refinement in Coot⁵² and automated global real-space refinement in PHENIX⁵³ using the unsharpened maps. For both complexes, PP2Aa, B55 and PP2Ac from the PP2A:B55–FAM122A complex (Protein Data Bank (PDB) 8SO0) were docked into the maps using ChimeraX. Density corresponding to p107 was clearly helical while density corresponding to Eya3 was extended, with the strongest density at the bottom of the B55 platform showing a clear π-stacking interaction with H179_B55. p107 and Eya3 residues identified to interact with B55 by NMR were readily modeled into the density. Model geometry and map-model validation metrics are given in Table 1.

DiFMUP fluorescence intensity assay for PP2A:B55 IC₅₀ measurements

DiFMUP-based IC₅₀ assays were conducted in 384 well plates (Corning, 4411). For the p107_612–687 and Eya3_62–108EE IC₅₀ assays, PP2A:B55 holoenzyme in enzyme buffer (30 mM HEPES pH 7.0, 150 mM NaCl, 1 mM MnCl₂, 1 mM DTT, 0.01% Triton X-100 and 0.1 mg ml⁻¹ BSA) was preincubated with various concentrations of p107_612–687 and Eya3_62–108EE variants for 30 min at room temperature. The reaction was started by adding DiFMUP (final concentration: 50 μM) into the PP2A:B55–p107_612–687 and PP2A:B55–Eya3_62–108EE enzymatic reaction (final concentration of PP2A:B55 holoenzyme: 1 nM) and then incubated at 30 °C for 30 min. Endpoint reads (excitation, 360 nm; emission, 450 nm) were taken on a CLARIOstarPlus (BMG Labtech) plate reader (using reader control software version 5.7R2) after the reaction was stopped by the addition of 300 mM potassium phosphate (pH 10). The experiments were independently technically repeated at least three times (n = 3–6) and the averaged IC₅₀ and s.d. values were reported. The data were evaluated using GraphPad Prism 10.2.2.

NMR data collection

All NMR data were collected on a Bruker Avance Neo 600-MHz or 800-MHz NMR spectrometer equipped with a TCI HCN z-gradient cryoprobe at 283 K. [¹⁵N,¹³C]-labeled Eya3 (170 µM) was prepared in Eya3 NMR buffer with 5–10% (v/v) D₂O added immediately before data acquisition. The sequence-specific backbone assignment for Eya3 was determined by recording a suite of heteronuclear NMR spectra: 2D [¹H,¹⁵N] HSQC, 3D HNCA, 3D HN(CO)CA, 3D HNCACB, 3D CBCA(CO)NH, 3D HNCO and 3D HN(CA)CO. Spectra were processed in Topspin (Bruker; version 4.1.3/4.4) and referenced to internal DSS.

Sequence-specific backbone assignment, CSI and chemical shift perturbation

Peak picking and sequence-specific backbone assignment for Eya3 were performed using the program CARA version 1.8.4.2 (http://www.cara.nmr.ch) (Supplementary Fig. 1). The p107 chemical shifts were transferred from the Biological Magnetic Resonance Data Bank (28091)³³. CSI calculations of p107 and Eya3 were performed using both Cα and Cβ chemical shifts for each assigned amino acid, omitting glycine, against the RefDB database⁵⁴. Secondary structure propensity (SSP) scores were calculated using a weighted average of seven residues to minimize contributions from chemical shifts of residues that are poor measures of secondary structure. Chemical shift differences (∆δ) were calculated using the following equation:

$$\Delta \delta \left({\rm{ppm}}\right)=\,\sqrt{{(\Delta {\delta }_{H})}^{2}+{\left(\frac{\Delta {\delta }_{N}}{5}\right)}^{2}}$$

NMR interaction studies of p107 and EYA3 with PP2A:B55 and B55_LL

All NMR interaction data of p107_612–687 and Eya3_62–108EE with PP2A:B55 or B55_LL were recorded using a Bruker Neo 600-MHz or 800-MHz NMR spectrometer equipped with a TCI HCN active z-gradient cryoprobe at 283 K. All NMR interaction data of p107_589–731 and Eya3_27–155 with PP2A:B55 were recorded using a Bruker Neo 800-MHz NMR spectrometer equipped with a TCI HCN active z-gradient cryoprobe at 283 K. All NMR measurements of all p107 and Eya3 constructs were recorded using ¹⁵N-labeled protein in p107 or Eya3 NMR buffer and 90% H₂O with 10% D₂O. For each interaction, an excess of unlabeled B55_LL or PP2A:B55 complex (≥25% surplus ratio) was added to ¹⁵N-labeled p107 or Eya3 and incubated on ice for 10 min before the 2D [¹H,¹⁵N] HSQC spectrum was collected. The p107 and Eya3 concentrations ranged from 2 to 6 μM. NMR data were processed using NMRPipe⁵⁵ and the intensity data were analyzed in Poky (build 20230213)⁵⁶. The p107_612–687 intensity data were normalized to the C-terminal residue peak and the difference between the free 2D [¹H,¹⁵N] HSQC spectrum of p107 was compared to its respective peak, if present, on the 2D [¹H,¹⁵N] HSQC spectrum of p107 in complex with B55_LL or PP2A:B55. The Eya3_62–108EE intensity data were normalized to the average of the last two C-terminal residue peaks and the difference between the free 2D [¹H,¹⁵N] HSQC spectrum of Eya3 was compared to its respective peak, if present, on the 2D [¹H,¹⁵N] HSQC spectrum of Eya3 in complex with B55_LL or PP2A:B55 (Microsoft Excel version 2301, b16.0.16026.20002 and GraphPad Prism version 10.2.2). Any overlapping peaks were omitted for this analysis. CCPNMR version 3.2.0 was used for all spectrum overlays and figures⁵⁷.

NMR dephosphorylation

All dephosphorylation data were recorded using a Bruker Neo 600-MHz NMR spectrometer equipped with a TCI HCN active z-gradient cryoprobe at 283 K. All measurements were recorded using ¹⁵N-labeled p107 in p107 NMR buffer and 90% H₂O with 10% D₂O. A reference 2D [¹H,¹⁵N] HSQC was recorded. Unlabeled active PP2A:B55 was added to ¹⁵N-labeled p107 in a 1:500 ratio. A 2D [¹H,¹⁵N] HSQC spectrum was measured every 28 min to monitor dephosphorylation. NMR data were processed using NMRPipe⁵⁵ and the disappearing intensities of the phosphorylated peaks and the reappearing intensity of the unphosphorylated peaks were analyzed with Poky (build 20230213)⁵⁶. CCPNMR version 3.2.0 was used for all spectrum overlays and figures.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.