The DNA fragment encoding isoform 2 of papp5 from Arabidopsis thaliana (UniProtKB: Q84XU2-2; 1-484) was codon-optimized and flanked by NdeI and Sall restriction sites (Geneart, Invitrogen Life Technologies). Afterward, the fragment was subcloned into the pET-28a vector (Novagen) using the corresponding restriction sites.

The plasmid carrying the coding sequence for the photosensory module of Arabidopsis thaliana phytochrome B (UniProtKB: P14713-1; 1-651) was kindly provided by Andreas Zurbriggen (University Düsseldorf) and subcloned into the pCDF Duet-1 vector (Novagen) by polymerase chain reaction using primers, which introduced PciI and PstI restriction sites. S84D and S86D mutants were generated by QuickChange site-directed mutagenesis (Stratagene). All primers are listed in Supplementary Table 1. Sequences of the resulting plasmids were controlled by dideoxy sequencing (GATC).

Expression was performed in the E. coli BL21 Gold (DE3) strain (Novagen). Cultures for AtPAPP5 overproduction were grown in TB medium containing 35 mg/L kanamycin at 37°C to an OD₅₉₅ of 0.6, the temperature was decreased to 18°C and expression was induced with 50 μM IPTG. After 22 h the cells were harvested by centrifugation (8,200 g, 15 min, 4°C), resuspended in Mn-buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 4 mM MnCl₂, 0.1 mM β-mercaptoethanol), frozen in liquid nitrogen and stored at −80°C. Bacterial cells were lysed by a French Press cell (AMINCO) and the supernatant was separated by centrifugation (40,000 g, 30 min, 4°C), followed by Ni²⁺-affinity chromatography (HisTrap^TM HP column, GE Healthcare) and eluted with TIS buffer (50 mM Tris, 250 mM imidazole pH 7.8, 100 mM NaCl, 4 mM MnCl₂, 1 mM β-mercaptoethanol). A final purification step was done by size exclusion chromatography (Superdex 200 26/60, GE Healthcare) using 50 mM Tris pH 7.8, 1 mM EDTA, 100 mM NaCl, 4 mM MnCl₂, 1 mM β-mercaptoethanol as buffer. Before usage of AtPAPP5 for assays the affinity tag was cleaved off by thrombin (overnight at 4°C) and removed by reversed Ni-NTA.

Overproduction and purification of wild-type AtPhyB(1-651) and the phosphomimic mutants harboring the PCB chromophore was performed as described before (Von Horsten et al., 2016). Prior to usage samples were centrifuged for 10 min at 4°C and 13,000 rpm.

The assay was performed as described by Ryu et al. (2005). In 100 μL reaction mixture, 1 μg AtPAPP5 in kinase/phosphatase buffer (KP: 25 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 20 mM Mg(CH₃COO)₂, 0.2 mM EDTA, 0.2 EGTA) was used in the absence or presence of arachidonic acid/6% ethanol. The concentration of p-nitrophenyl phosphate (pNPP) was varied. After an incubation time of 15 min at 30°C the reaction was terminated by the addition of 900 μL 0.25 M NaOH and the absorbance at 410 nm was measured.

Microscale thermophoresis was conducted to determine the AtPAPP5 dissociation constant toward the AtPhyB(1-651) mutants S84D and S86D by using a Monolith NT.115 instrument (NanoTemper). Freshly prepared AtPAPP5 (50 μM in labeling buffer) was labeled at 8°C overnight using the cysteine-reactive Monolith NT.115 Protein Labeling Kit GREEN MALEIMIDE (NanoTemper) to achieve a 1:1 molar ratio of labled protein to dye. A serial titration from 4 nM to 70 μM of S86D and 2 nM to 29 μM of S84D was prepared and labeled. AtPAPP5 was diluted to each sample until a final concentration of 50 nM. Afterward 100 μM AA was added and the mixture was incubated for 15 min before measurement. The assay was then transferred into premium coated capillaries (NanoTemper) and measured at 25°C. The LED power for each thermophoresis measurement was set to 80% and the laser power to 20%, the heating time was set to 30 s, followed by 5 s of cooling. Binding curves were obtained from the thermophoresis hot to cold ratio. Negative controls were performed without the addition of AA and with AtPhyB WT under the same conditions mentioned above. Each experiment was repeated in triplicate, analyzed with the NT Analysis software (NanoTemper) and the K_D value was calculated with Origin 9.0 (OriginLab) using a 1:1 binding model.

A 25 μM of AtPAPP5 in KP buffer (25 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 20 mM Mg(CH₃COO)₂, 0.2 mM EDTA, 0.2 EGTA) buffer was injected onto a Superdex 200 10/300 column using an Äkta purifier (GE Healthcare). The setup was repeated after adding 30 μM arachidonic acid. The runs were monitored at 280 nm. The apparent molecular masses of the eluted species were calculated by using a molecular weight standard curve based on a column calibration (LMW kit, GE Healthcare) performed previously.

A solution of 40 μM AtPAPP5 in HEPES buffer (10 mM HEPES pH 8.0, 100 mM NaCl, 4 mM MgCl₂) with 0 μ, 25, 50, 75, 100, and 150 μM AA, using a 3.2 mM AA stock solution in EtOH were incubated at room temperature for 15 min and measured with a spectropolarimeter J-810 (Jasco) using a 1 mm cuvette. After a baseline correction with the corresponding buffer, which contained an equal concentration of ethanol, the data was averaged over three scans.

AtPAPP5 was transferred into crystallization buffer (10 mM Tris pH 7.5, 100 mM NaCl, 5 mM MnCl₂), concentrated to 15 mg/mL and crystallization attempts were pipetted with a Honeybee robot system (Zinsser Analytic) using commercially available screens (Qiagen) in a 96-well format. Crystals were grown at 18°C using the sitting-drop method and vapor-diffusion. After 2 weeks, small crystals were found growing in condition C10 from crystal screen Core I (Qiagen). To obtain bigger crystals optimizations were done using streak seeding with crushed crystals from the original hit. The crystallization drop for the optimizations consists of 1 μL protein solution and 1 μL of reservoir solution. For optimization, the hanging-drop vapor diffusion method was used. After 4 weeks, crystals appeared in a condition containing 0.1 M HEPES pH 7.0 and 50% (v/v) PEG4000. These were picked and incubated in cryoprotection buffer (reservoir buffer supplemented with 30% glycerol) prior to freezing in liquid nitrogen.

An AtPAPP5 dataset was collected at beamline 14.2 of BESSY-II (Berliner Elektronenspeicherring-Gesellschaft für Synchrotronstrahlung, Berlin, Germany) and processed with the XDS (Kabsch, 2010) and CCP4 (Collaborative Computational Project, 1994) package. The structure for the dataset was solved by molecular replacement using PHENIX autosolve (Adams et al., 2010) and 1WAO as search model (Yang et al., 2005). Manual model building was conducted with COOT (Emsley and Cowtan, 2004) and refinement by PHENIX refine (Adams et al., 2010). Figures were created using PYMOL 1.6 (DeLano Scientific).

HDX-mass spectrometric analysis AtPAPP5 (60 μM) with and without arachidonic acid (75 μM) was carried out using a commercial HDX-automation setup (SYNAPT G2-Si, Waters) including a two-arm robotic autosampler (LEAP Technologies), an ACQUITY UPLC M-Class system (Waters) and HDX manager (Waters). AtPAPP5 samples were transferred by a PD-10 column into a low salt buffer (10 mM Tris pH 8.0, 100 mM NaCl, 4 mM MgCl₂, 1 mM β-mercaptoethanol) and centrifuged for 10 min at 16,100 g and 4°C. Exchange reactions and analysis were performed as described before (Von Horsten et al., 2016).

Final assignment of deuterium incorporation was done with DynamX 3.0 (Waters). The minimum peak intensity was set to 10³ counts and a peptide length between four and 15 was chosen. Moreover, tolerances of 0.5 min for the retention time and 25 ppm for m/z values were applied for the peptide assignment, generating an overall sequence coverage of 80.0%. 125 peptides were analyzed with an overall redundancy of 3.0 per amino acid. A standard deviation of 4 σ was used to quantify the amount of variation between the repetitions. A coverage map of all peptides is presented in Figure 1D. The results of the analysis were mapped onto the structure of AtPAPP5.

Molecular dynamics analyses of AtPAPP5 were done with the GPU-accelerated Amber18 suite (Lee et al., 2018) using the ff19sb force field for the protein (Tian et al., 2020) and the SPC/E water model. pKa values of ionizable residues of AtPAPP5 at pH 6.5 were estimated by H++. Using B3LYP/6-31G^∗ model chemistry we calculated parameters for the μ-hydroxyl bridged di-Mn²⁺ center either in its free state or complexed to methyl-phosphate or phosphoserine-comprising phyB peptides by gaussian16 (Frisch et al., 2016) and MCPB.py (Li and Merz, 2016). As start geometry for these QM-calculations we used the di-Mn²⁺ center of human PP5 complexed to the Cdc37 substrate peptide (Oberoi et al., 2016). For MD simulations free AtPAPP5 was neutralized by 57 chloride and 58 sodium ions (∼0.15 M) and placed in a box with 31,718 water molecules, which equilibrated to box dimensions of 100.6 × 100.8 × 100.5 ų (Supplementary Table 2). AtPAPP5 ensembles were minimized, pre-equilibrated as NVT ensembles followed up for 5.5 ns as NpT ensembles at 300 K using 2 fs steps, 10 Å cutoff and a Monte-Carlo barostat. The systems were further equilibrated for 100 ns before generating projection trajectories with lengths of 180–1,000 ns. AtPAPP5 trajectories were processed and evaluated in jupyter notebooks using pytraj 2.0.5 (Roe and Cheatham, 2013),¹ AmberTools20 and NGLview 2.7.7 (Nguyen et al., 2018). The total length of four independently generated production trajectories for AtPAPP5 in the unbound, closed and the phyB S80-T89 phosphorylated peptide states amount to 4 μs for each, for the complex between AtPAPP5 and the phyB D82-S86 peptide phosphorylated at S84 to 2.88 μs and the complex with phosphoserine to 3.50 μs (Supplementary Table 2). AtPAPP5-bound peptides were capped at their N- and C-termini by acetyl and methylamine groups, respectively.

For structural analyses, we used recombinant AtPAPP5 that was produced as histidine-tagged monomer by heterologous expression in E. coli. In contrast to previous reports (Ryu et al., 2005) we removed affinity tags for activity measurements.

Earlier reports showed that AA activates AtPAPP5 as well as the human PP5 ortholog (Chen and Cohen, 1997; Skinner et al., 1997; Ryu et al., 2005). To confirm the functionality of recombinant AtPAPP5, we measured the phosphatase activity by an assay based on para-nitrophenyl phosphate (pNPP) as chromogen. The amount of released phosphate (P_i) corresponds to the formation of para-nitrophenolate that is monitored at 405 nm. The assay was initially conducted for AtPAPP5 in the presence of 70 μM AA that was dissolved in ethanol. Control reactions were carried out with an equivalent volume of ethanol. Surprisingly, ethanol alone already activates AtPAPP5 by releasing the blocked active site (Figure 2A). In the presence of AA AtPAPP5 shows a Michaelis-Menten kinetics for varying pNPP concentrations with a K_M-value of 92.2 mM and a V_max value of 15 ± 1.1 μmol/mg/mL. These values resemble kinetic constants obtained by Ryu et al. (2005) for a GST-AtPAPP5 fusion with 100 μM AA (K_M 160 mM; V_max 22 μmol/mg/mL). Interestingly, addition of arachidonic acid caused only a doubled turnover rate relative to the ethanol-activated state (V_max = 8.9 ± 1.9 μmol/min/mg) and a fourfold increase compared to wildtype without any AA and EtOH (V_max = 3.6 ± 0.7 μmol/min/mg).

We also checked PAPP5 activities in the presence of different fatty acids using saturating substrate concentrations, i.e., 250 mM pNPP (Figure 2B). None of the other fatty acids activated PAPP5 like AA. To further investigate the influence of ethanol on the phosphatase activity, ethanol concentrations were varied between 0 and 90% (v/v). Interestingly, these data show an up to 12-fold boost of PAPP5 activity with 75% ethanol (Supplementary Figure 1), which may reflect either partial unfolding of the TPR region or a disruption of TPR-PP2A domain interface.

Structural studies on the catalytic domain of human PP5 complexed to a Cdc37-derived peptide showed that phosphomimetic S→D mutations serve as useful tools for analyzing phosphatase-phosphopeptide interactions (Oberoi et al., 2016). Accordingly, we proved direct interaction between AtPAPP5 and the NTE of AtPhyB by generating two phosphomimetic variants of AtPhyB(1-651), S84D, and S86D. Phosphorylation/dephosphorylation of S84 and S86 in the NTE phyB was shown to be strongly involved in the negative in vivo regulation of phyB (Medzihradszky et al., 2013; Nito et al., 2013). We performed sensitive interaction assays by microscale thermophoresis (MST) using AtPAPP5 that was labeled before with a green-fluorescent dye. After preincubation with AA for 15 min AtPAPP5 binding was measured against a concentration series with either AtPhyB/S84D or AtPhyB/S86D. Since our MST setup uses a green laser for fluorescence detection (λ_exc: 515–525 nm, λ_em: 650–685 nm), the phytochrome variants were kept in a photodynamical equilibrium with a prevalence of their activated P_fr state. AtPAPP5 showed a high affinity interaction to AtPhyB phosphomimic variants with an apparent K_D of 1.3 ± 0.3 μM for S84D and 1.8 ± 0.5 μM for S86D, respectively (Figures 2C,D). To rule out non-specific binding, the measurements were repeated using recombinant wild-type AtPhyB(1–651) and resulted in signals not exceeding signal-to-noise ratios. Accordingly, non-specific binding to the photosensory module of AtPhyB in its non-phosphorylated P_fr state can be excluded. Furthermore, measurements in the absence of AA resulted in baselines showing the necessity of AtPAPP5 activation by AA (Figure 2C).

We analyzed the photochemical behavior of our phosphomimic mimics by UV/Vis spectroscopy. Compared to wild-type AtPhyB (Von Horsten et al., 2016) absorption maxima are bathochromically shifted, especially in the P_fr form with shifts of 8 nm for the S84D and 5 nm for the S86D variant (Supplementary Figure 2). Furthermore, both phytochrome B variants exhibit an accelerated dark conversion rate (t_1/2 of S84D: 98 min; S86D: 40 min) compared to the photosensory module of wildtype phyB (t_1/2: 270 min; Figures 2E,F; Von Horsten et al., 2016). Addition of AtPAPP5 and AA induced no further shift of the maxima according to UV/Vis spectra. However, there is a slightly reduced dark conversion rate with half-lives of 105 min for S84D and 49 min for S86D.

We crystallized AtPAPP5 in a monoclinic crystal form that resulted in a dataset at 3.0 Å resolution. Unfortunately, we could not treat these AtPAPP5 crystals with AA because they rapidly dissolved upon incubation. Molecular replacement using protein phosphatase 5 from Homo sapiens (HsPP5; PDB Code 1WAO) as search model (sequence identity: 56% for 484 amino acids) allowed structure determination with two molecules per asymmetric unit. Table 1 shows data collection and refinement statistics of our AtPAPP5 structure.

The AtPAPP5 structure shows the characteristic bilobal shape of a type 5 serine/threonine phosphatase with two manganese ions bound in the active site. Three TPR repeats (TPR1: V10-L42, TPR2: W50-V76 and TPR3: S81-L110) are linked via a long helical linker (P115-T173) to the PP2A domain (L174-N484). The TPR repeats of the N-terminal domain consist each of two α-helices as part of a helix-turn-helix motif; these motifs stack to each other causing formation of a superhelical tertiary structure. The last 14 amino acids of the PP2A domain form a separate motif, the C-terminal inhibitory motif that consists of a linker (K471-A476) and a short α-helical segment (αJ, N477-M482; Figure 3A). The strictly conserved PP2A domain adopts a compact α/β fold consisting of eleven α-helices and eleven β-strands, which form together the central β-sandwich (Figure 3E). This sandwich is flanked on one side by three short helices (α9-α11) and on the other side by six helices (α12-α17). The active site itself is embedded in the PP2A domain and covered by the TPR region and the C-terminal inhibitory motif. In its center two catalytic manganese ions are observed (Figures 3A,B).

A superposition of AtPAPP5 with HsPP5 shows, that AtPAPP5 mimics the overall structure of the human phosphatase with an r.m.s.d. of 0.70 Å for 365 C_α-atoms (0.99 Å for 390 C_α -atoms). These structures represent the closed state of type 5 serine/threonine protein phosphatases, because the TPR region covers major parts of the active site region. Minor deviations are shown in the arrangement of the TPR region and the helix of the linker region (Figures 3C,D). In contrast to mammalian PP5 structures (Yang et al., 2005; Haslbeck et al., 2015) in our AtPAPP5 structure the linker region is completely defined by electron density.

The local environment of the active site is conserved between HsPP5 and AtPAPP5 as exemplified by residues D228, H230, D257, N289, H338 and H413 (numbering of amino acids corresponds to AtPAPP5), which are the coordination partners of the manganese ions. Here, Mn²⁺ ion 1 is tetra-valently coordinated by amino acids D257, N289, H338 and H413, whereas Mn²⁺ ion 2 interacts with amino acids D228, H230 and D257 (Figure 4A). For comparison, a crystal structure of the isolated human PP2A domain (PDB Code: 1S95) (Swingle et al., 2004) shows two additional water molecules. One of them acts like the oxygen atom of the opposing D257 as μ-bridge between the two Mn²⁺ ions thus forming slightly distorted octahedral coordination spheres around the catalytic Mn²⁺ ions (Figure 4B). These μ-bridging water ligands are missing in the AtPAPP5 structure due to low resolution. Instead, we observe a coordinated chloride ion in our AtPAPP5 structure (Figures 4A,C). This partly occupied chloride (molecule A: 59%, B: 65%) replaces a phosphate anion that was observed before in the active site of the isolated PP2A domain of HsPP5. Its five interacting ion partners are preserved in AtPAPP5 (R261, N289, H290, R386, and Y437) and require only some swiveling of the sidechains of R261 and R386 for phosphate binding.

The AtPAPP5 structure represents the autoinhibited state, in which both the TPR region as well as the C-terminal inhibitory motif (K471-N484) block the active site and deny access for putative substrate-like phosphopeptides (Figure 5). This mode of self-inhibition was already demonstrated by activity measurements of PP5-type deletion variants (Sinclair et al., 1999; Ryu et al., 2005).

An analysis of the interactions between the TPR and PP2A domain shows that all three TPR repeats form interactions with the PP2A domain (Figure 5A). Here, the prominent hydrogen bonds between the carboxylate group of E61 to Y437 and R261 are of particular interest, since both amino acids are involved in the coordination of the substrate’s phosphate group. Mutation of E61 in HsPP5 to alanine unblocked the active site and raised 10-fold the intrinsic phosphatase activity (Kang et al., 2001). Further stabilization of the inhibited state may be also provided by the interaction between the hydroxyl group of Y63 with R261. These interactions with the phosphate binding site as well as those between the amide oxygen of A93 and the hydroxyl group of Y299 are also conserved in the human and rat PP5 phosphatases. Finally, H27 from TPR1 and K59 from TPR2 form H-bonds with the main chain carbonyl group and the sidechain of Q440 in the PP2A domain, respectively.

The αJ-helix of the C-terminal inhibitory motif is indirectly involved in shielding the active site by stabilizing the closed-state interactions of the PP2A domain with the TPR region (Yang et al., 2005; Haslbeck et al., 2015). In AtPAPP5, the thioether group of M482 forms interactions with the carbonyl oxygen of Y85 as well as van-der-Waals interactions with L92 (Figure 5B). Unlike mammalian PP5, not only non-polar interactions are found in this region of AtPAPP5. One prominent interaction is formed by N484, which interacts via H-bonds with the carbonyl oxygen of Y63 on TPR1 and Y299 of the PP2A domain. The position of this amino acid is further stabilized by interactions with the amide group of L480. Overall, the polar amino acid N484 is found to bridge both the TPR and the PP2A domains.

The positioning of the αJ-helix is also supported by an interaction with the TPR-PP2A linker. Interestingly, in the two molecules of the asymmetric unit the αJ-helix interacts with different amino acids. Heterogeneity of this helix was also observed in the two PP5 structures from Rattus norvegicus (Haslbeck et al., 2015). While the hydrophobic interaction between A127 and the aromatic side chain of F479 is present in both molecules, the polar interactions occur only in molecule A (E123 and R481) and not in molecule B. In the latter, the side chains of R481 and N478 are in H-bonding distance (Figures 5C,D). This interaction pattern might be unique to plant phytochromes since mammalian PP5 lack the involved residues and show no comparable interactions for the C-terminal five amino acids in its structure.

Heterogeneous interactions of the C-terminal inhibitory motif may indicate that the C-terminus only partly inhibits phosphatase activity and a dynamic equilibrium is formed between active and auto-inhibited AtPAPP5 forms. This might explain the small, but still measurable activity of the AtPAPP5 wildtype due to reversible breakage and reformation of interactions between the TPR and the PP2A domains. However, to achieve full phosphatase activity in vitro the addition of AA is needed. Size exclusion chromatography (SEC) shows an earlier elution of AtPAPP5 upon addition of AA, which indicates a larger hydrodynamic radius than the closed, non-activated state of AtPAPP5 (Figure 1C). To address this issue, we performed CD measurements in the presence of different AA concentrations (Figures 1A,B). These data reveal that the relative α-helical fraction decreases in the presence of AA thus implying partial unfolding of α-helical regions in AtPAPP5. In order to localize corresponding regions, we applied HDX-MS measurements to AtPAPP5 in the non-actived, closed state as well as in the AA-activated state. Generally, less folded regions are usually characterized by enhanced hydrogen deuterium exchange rates (Figure 1D).

A priori one may predict higher exchange rates for the TPR repeats in the presence of AA due to a disruption of the TPR-PP2A domain interface for opening up the active site of the PP2A domain. Interestingly, the HDX-MS data show only for TPR2 and the beginning of the TPR3 repeat increased exchange rates (Figure 6A). In the AA-activated state H27 of TPR1 experiences surprisingly a reduced H/D exchange rate, whereas its interaction partner within the closed state of AtPAPP5, Q440, shows increased H/D exchange. Y437 located on the same loop as Q440 shows an increased exchange rate as well. Furthermore, R261 experiences increased exchange so that it can be assumed that its interactions with E61 and Y63 are no longer formed in the AA-actived state. K385 shows a slightly reduced and K59 a slightly increased exchange. Interestingly, some of the interactions between the TPR3 repeat and the PP2A domain apparently remain intact, since Y299 and A93 show no significant changes. It is striking, that only D183 located on the α8-helix has an increased exchange rate. Accordingly, the interactions that D183 had formed to S163 of the linker and K188 of the PP2A domain in the closed state of AtPAPP5 seem to be broken by addition of AA (Figure 6A).

Two further regions (P160-A165 and H467-A474) stand out due to their significantly reduced uptake rates in the AA-activated state (Figures 6B,C). The first one (P160-A165) is part of the linker between the TPR and PP2A domains. The second region with a reduced exchange rate is part of the C-terminal inhibitory motif (H467-A474). The αJ-helix of the C-terminal motif shows no significant changes in the presence of AA. The last three amino acids of the C-terminus that form both interactions with the TPR and the PP2A domain as well as the corresponding interaction partners are not defined by our HDX-MS analysis (Figure 6C). A comparison of the amino acid sequences of the PP5 phosphatases showed that the linker region contains only three conserved amino acids (Y162, G164, and P165). In contrast, the beginning of the C-terminal inhibitory motif is strictly conserved (Figure 6C).

In order to check for transferability of our AtPAPP5activation mechanism to other members of the PP5 family, sequence identities of the PP5 phosphatases from different organisms were aligned and analyzed for conservation by the Consurf server (Glaser et al., 2003; Landau et al., 2005). The active site and the interaction interface of the PP2A domain with the TPR region are strongly conserved, compared to the rest of the structure (Supplementary Figure 3). Interestingly, two of the three amino acids of the TPR region that interact with the C-terminal inhibitory motif (Y85 and L92) are conserved in all PP5 phosphatases, as well as E61, which binds in the active site. A complete conservation of the helices that unfold upon addition of AA might not be necessary, as only specific bonds are broken. Furthermore, both regions, which showed a decrease in HDX rate in the presence of AA, are conserved within the PP5 family indicating the general applicability of our activation mechanism. However, the nuclear localization signal is also localized in C-terminal inhibitory motif and could explain the high degree of conservation.

Although the HDX-MS analysis failed to identify unambiguously the AA binding of AtPAPP5, we can summarize (1) that AA binding induces a loss of structure only in the TPR2 repeat of the TPR region and (2) that the C-terminal inhibitory motif stays associated with the TPR region in the active state of AtPAPP5. One may tentatively suggest that H27 at the tip of TPR1 may represent an interaction partner of AA due to ionic interaction with the AA’s carboxylate group, as several hydrophobic, surface exposed side chains are located near H27 (F24, Y29, F55).

As the rather local effects of AA-binding to PAPP5 are sufficient for activation of AtPAPP5, we wondered to what extent the TPR region has to unblock the active site for allowing substrate binding and catalysis. For that purpose, we performed a series of molecular-dynamics simulations, where we sequentially increased the size of the substrate (analog) bound to the di-manganese site in the active center. In the first step, we added a capped phosphoserine residue to the active site, equilibrated this system as NpT ensemble for 80 ns before extending the bound phosphoserine to the phyB D82-S86 peptide phosphorylated at S84. The orientation of the phyB was taken from the complex of human PP5 with a Cdc37 peptide (Oberoi et al., 2016). After further equilibration for 100 ns the AtPAPP5•peptide complex was elongated to the phyB S80-T89 peptide.

MD trajectories for the closed state of AtPAPP5 show only minor differences for the association of the TPR region and the PP2A domain when comparing to the crystal structure of AtPAPP5 and other PP5-type phosphatases (Figure 7). Only the interactions of Q440 with H27 and K59, which were observed in the crystal structure, are apparently weaker in the MD ensembles (Supplementary Table 4) due to intrusion of bridging water molecules. Docking of the short phosphoserine peptide N-acetyl-SEP-NMe into the active site of AtPAPP5 causes apparently no major structural adaptions of the closed state. The center-of-mass (COM) distances between the TPR region and the PP2A domain increases only slightly from ∼28 Å to 29–30 Å (Supplementary Figure 4). Nevertheless, the H-bonding interactions Y437/E61, Q440/K59 and Y299/Y63 of the substrate-free state are broken in the TPR-PP2A interface. Elongating the phosphopeptide to the S84-phosphorylated phyB peptide D82-S85 severely distorts the TPR-PP2A interface by triggering a rotation of the TPR-region by ∼40° relative to the PP2A domain and the linking helix (Figure 7) and an increase of the COM-COM distance of both domains to 32–34 Å. As a consequence, main chain interactions between G64-S65 from TPR2 and K385-R386, which are a feature of the closed states, are disrupted. Despite the large-scale conformational changes the C-terminal inhibitory motif remains associated during simulation time with the TPR region (Figure 8A). Given the mostly hydrophobic interactions of this motif with TPR3 residues, it is notable that its only salt bridge in the closed state between the C-terminal carboxyl group of N484 and K59, which is apparently only transiently formed during MD simulations of the closed state (Supplementary Table 4) and not defined at all for molecule B of the AtPAPP5 crystals, forms a new salt bridge to the side chain of K120 from the linker helix. Finally, in the complex with the capped phyB decapeptide S80-T89 there is a slight increase of the TPR-region rotation to ∼50° without a change of the COM-COM distance. The MD trajectories of this AtPAPP5 complex tend to form again more homogeneous ensembles in the now open state than the AtPAPP5•phyB/D82-S85 complexes, because the former exert a lesser degree of conformational variation for the TPR-PP2A domain arrangement (Figure 7 and Supplementary Figure 4).

The phosphoserine S84 of the phyB peptides is coordinated to the dimanganese center and forms ionic interactions with the sidechain of R261; the protonated H290 forms a hydrogen bond to the OG atom of the phosphoserine and apparently acts as general acid during catalysis (Figure 8B). Despite being sandwiched between the TPR and PP2A domains, the conformation of the bound phosphopeptides follows the in-out-in-out rule for peptide binding by PP5-type phosphatases as previously suggested (Oberoi et al., 2016), where the phosphoserine adopts the inward orientation toward the active site.