Protein tyrosine phosphatases (PTPs) are a family of enzymes that mediate the reversible dephosphorylation of phosphorylated tyrosine residues on substrate proteins, one of the most common and critical post-translational modifications (Tonks, 2006). They are closely associated with the regulation of various intracellular signaling and cellular processes by controlling protein-protein interaction, protein stability, and enzyme activity (Ardito et al., 2017; Hunter, 1995). The Drosophila genome contains 44 PTPs, which are less than half of 109 PTPs encoded by the human genome (Hatzihristidis et al., 2015). The 44 Drosophila PTP proteins can be divided into four subgroups: 37 class I PTPs that include 16 classical PTPs and 21 dual-specificity phosphatases (DUSPs), four class II low molecular weight-protein tyrosine phosphatases (LMW-PTPs), two class III CDC25 proteins, and one eyes-absent phosphatase (Hatzihristidis et al., 2015). The number of members of each PTP subgroup is higher in humans than in Drosophila, except for LMW-PTPs, of which the human genome contains only one (1/109, 0.9%), whereas the Drosophila genome contains four (4/44, 9.1%), including Primo-1, Primo-2, CG14297, and CG31469 (Hatzihristidis et al., 2015).

LMW-PTPs are found in a broad spectrum of genera from bacteria to higher eukaryotes, including Drosophila and human. While most PTP proteins consist of multiple domains, LMW-PTPs comprise only a single PTP domain. LMW-PTPs act on phosphotyrosine, but not on phosphoserine and phosphothreonine (Caselli et al., 2016); however, a case of atypical substrate preference of LMW-PTP was also reported. Based on structural and biochemical analyses, Fuhrmann et al. (2013) reported that YwlE, an LMW-PTP protein from Bacillus subtilis, is active toward phosphoarginine rather than phosphotyrosine, rendering it a novel type of phosphatase, protein arginine phosphatase. Moreover, the authors conducted a bioinformatics search, identifying CG31469 from D. melanogaster as the first putative eukaryotic protein arginine phosphatase (Fuhrmann et al., 2013).

The four Drosophila LMW-PTP members, including CG31469, have not been intensively investigated thus far. While Primo-2 was verified to have phosphatase activity toward para-nitrophenylphosphate (pNPP) (Miller et al., 2000), Primo-1 and CG31469 were previously classified as “pseudophosphatases” without experimental proof (Hatzihristidis et al., 2015), contradicting the previous finding that CG31469 has arginine phosphatase activity (Fuhrmann et al., 2013). Herein, we present the crystal structures of Primo-1 and CG31469, representing the first three-dimensional structures of Drosophila LMW-PTPs. Structural and biochemical analysis revealed that Primo-1 and CG31469 are catalytically active toward phosphotyrosine and phosphoarginine, respectively. Based on these data, we consider that both the enzymes should be removed from the list of pseudophosphatases. Notably, we named CG31469 as “DARP”, which stands for Drosophila ARginine Phosphatase. To our knowledge, the present crystal structure of DARP is the first as the eukaryotic protein arginine phosphatase structure.

Preparation, crystallization, and structure determination of DARP and Primo-1

The DNA fragments coding for DARP and Primo-1 were amplified by polymerase chain reaction and subcloned into the pET21a plasmid, which were then used as templates for preparing mutant proteins: DARP(C7S), DARP(T11I), DARP(Y127F), DARP(T11I·Y127F), and Primo-1(C9S). The proteins were produced in the Escherichia coli BL21(DE3) RIL strain (Novagen, USA) at 18_°C and purified using a Ni-NTA column (Qiagen, Germany) and a HiLoad^TM 26/600 Superdex^TM 75 pg column (GE Healthcare, USA), equilibrated with 20 mM Tris-HCl (pH 7.5), 200 mM NaCl, and 2 mM dithiothreitol. Crystals were obtained by the sitting-drop vapor diffusion method at 18_°C by mixing and equilibrating 1 μl protein solution (12 mg/ml DARP(C7S) or 6 mg/ml Primo-1) and precipitant solution containing 1.6 M ammonium phosphate monobasic, 100 mM sodium acetate trihydrate (pH 5.2), and 10 mM calcium chloride for DARP(C7S) or 0.1 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)-NaOH (pH 7.47), 1.2 M sodium citrate tribasic, and 9% (w/v) D-sorbitol for Primo-1. Before data collection, the crystals were briefly soaked in a cryoprotectant solution consisting of the precipitant solution supplemented with 20% ethylene glycol in the case of DARP(C7S) or with 5% glycerol in the case of Primo-1. Diffraction data were collected on the beamline 7A at the Pohang Accelerator Laboratory, Korea, and processed using the program HKL 2000 (Otwinowski and Minor, 1997). Both structures were determined by molecular replacement using the program Phaser (McCoy et al., 2007) using the Streptococcus pyogenes SP-PTP structure as a search model (Ku et al., 2016). Coot (Emsley and Cowtan, 2004) and PHENIX (Adams et al., 2010) were used for model building and refinement, respectively. Crystallographic data statistics are summarized in Table 1.

Dephosphorylation assays using Primo-1

The phosphatase activity of Primo-1 toward pNPP was measured by detecting the absorbance at 405 nm for 120 min using a SpectraMax plus 384 spectrophotometer (Molecular Devices, USA). The reaction mixture contained 30 nM recombinant Primo-1, 10 mM pNPP, 100 mM Bis-Tris (pH 6.0), 200 mM NaCl, and 2 mM dithiothreitol. Kinetic parameters were calculated from initial velocity data obtained by varying the pNPP concentration in from 0 to 20 mM. The EnzChek Phosphate Assay Kit (Thermo Fisher Scientific, USA) was used for detecting the phosphatase activity of Primo-1 toward phosphotyrosine by measuring the absorbance at 360 nm for 60 min. Assays were performed at room temperature using 2.5 μM recombinant Primo-1 and 5 mM phosphotyrosine in a 200 μl reaction mixture containing 0.2 mM 2-amino-6-mercapto-7-methylpurine riboside, 0.2 units purine nucleoside phosphorylase, 50 mM Bis-Tris (pH 7.5), and 1 mM MgCl2. The reaction conditions were optimized by varying the amount of enzyme and substrates and pH.

Dephosphorylation assays using DARP

Substrate specificity was measured at room temperature using 1 μM wild-type or C7S mutant recombinant DARP and four types of 600 μM phosphoamino acids in a 200 μl reaction mixture containing 0.2 mM 2-amino-6-mercapto-7-methylpurine riboside, 0.2 units purine nucleoside phosphorylase, 50 mM Tris-HCl, and 1 mM MgCl2. The EnzChek Phosphate Assay Kit was used for determining the phosphatase activity of DARP against the four phosphoamino acids by detecting the absorbance at 360 nm for 60 min using a SpectraMax plus 384 spectrophotometer. Kinetic parameters were calculated from initial velocity data, which were obtained at room temperature using 800 nM recombinant DARP and 0 to 5 mM phosphoarginine in a 200 μl reaction mixture containing 0.2 mM 2-amino-6-mercapto-7-methylpurine riboside, 0.2 units purine nucleoside phosphorylase, 50 mM Tris (pH 7.0), and 1 mM MgCl2. Specific activities of the wild-type and mutant DARP proteins toward phosphoarginine and phosphotyrosine were calculated from initial velocity data (0-10 min), which were determined at room temperature by using the EnzChek Phosphate Assay Kit using a 200 μl reaction mixture containing 500 μM protein and 800 nM phosphoarginine at pH 7.0 or 10 μM protein and 5 mM phosphotyrosine at pH 7.5. The nonenzymatic hydrolysis of substrates was also measured for baseline correction using a control sample. Substrate specificity and kinetic constant measurements were performed in triplicate, and specific activity measurements were performed in duplicate.

Data availability

The atomic coordinates and the structure factors of DARP(C7S) and Primo-1 have been deposited in the Protein Data Bank under the accession codes 7BTG and 7CUY, respectively.

Crystal structures of DARP and Primo-1

Among the four Drosophila LMW-PTPs (Fig. 1A), wild-type and catalytic cysteine-substituted mutant DARP, Primo-1, and Primo-2 were successfully expressed in E. coli and subsequently purified. The purified recombinant proteins were subjected to crystallization. Crystals of DARP(C7S) in the space group P212121 and wild-type Primo-1 in the space group P21 were obtained, and the structures of these two Drosophila proteins were determined at resolutions of 2.2 Å and 2.1 Å, respectively (Table 1). One DARP and two Primo-1 molecule(s) were contained in the asymmetric unit of each protein crystal. Both structures adopt a canonical LMW-PTP fold, which comprises a central β-sheet that is composed of four parallel β-strands and five α-helices and one 3₁₀-helix that surround the β-sheet (Fig. 1B). The two crystal structures superimposed well, with a root-mean-square deviation (RMSD) value of 0.93 Å over 150 aligned C_α atoms (Fig. 1C, left). Only slight structural differences were observed; for example, α4 of DARP is longer than that of Primo-1 (Supplementary Fig. S1A). The structures (Fig. 1C, middle and right), as well as amino acid sequences (Supplementary Fig. S2), of DARP and Primo-1 share high similarity with those of previously defined LMW-PTP-folded proteins, such as human LMW-PTP isoform A (Fonseca et al., 2015), S. pyogenes SP-PTP (Ku et al., 2016), and B. subtilis arginine phosphatase YwlE (Fuhrmann et al., 2013).

Structural analysis of the catalytic loop composition of DARP and Primo-1

Given that DARP and Primo-1 were previously classified as enzymatically inactive pseudophosphatases (Hatzihristidis et al., 2015), we investigated whether the two proteins have any structural feature in their catalytic pocket region that negatively affects their catalytic activity. We first analyzed the P-loop arrangement of the two proteins, which is a key factor restricting phosphatase activity in a number of PTP proteins such as PTPRU, DUSP6, DUSP9, and Tk-PTP (Fjeld et al., 2000; Hay et al., 2020; Hong et al., 2005; Yun et al., 2018; Zhou and Zhang, 1999). The catalytic motifs of DARP and Primo-1 are composed of C⁷IGNTCR¹³ and C⁹LGNICR¹⁵, respectively, and are located at the β1−α1 loop, which is called the phosphate-binding loop or simply P-loop (Fig. 1A). The P-loop motif contains the catalytic cysteine residue (substituted with serine in the DARP structure), which functions as a nucleophile, and the conserved arginine residue (Arg13 in DARP and Arg15 in Primo-1), which anchors the phosphate group of the substrate during dephosphorylation (Fig. 2A). The center of P-loop is occupied by a single phosphate ion in the DARP structure and by the sulfonic group of a HEPES molecule in the Primo-1 structure. Oxygen atoms of these molecules mediate hydrogen bonds with the protein main chain amides and electrostatic interactions with the guanidinium group of the conserved arginine residue, which not only stabilize their positioning in the catalytic pocket but also maintain the active loop conformation of the Drosophila LMW-PTP proteins (Fig. 2A, Supplementary Fig. S1B). These backbone amides are oriented toward the catalytic pocket, and together with the arginine guanidinium group, create a positive electrostatic potential, which is a characteristic feature of catalytically active PTP proteins. This potential is not only complementary to the negatively charged phosphate group of the substitute but also stabilizes the side chain of the catalytic cysteine residue in a thiolate form (Kolmodin and Aqvist, 2001; Zhang et al., 1994). The conserved general acid/base aspartate residues (Asp125 in DARP and Asp124 in Primo-1) that stabilize the phosphoryl-intermediate during the catalytic reaction are also present in both structures at the typical position (Fig, 2A). The DARP and Primo-1 structures also contain two loops surrounding the P-loop region, so-called W-loop and D-loop (Fig. 2B). These loops are conserved among LMW-PTP structures; W-loop carries a tryptophan residue that contributes to substrate accommodation, while D-loop harbors the general acid/base aspartate residue and two substrate recognition-involved hydrophobic/aromatic residues (Fig. 2B). Taken together, we revealed that the P-loops of DARP and Primo-1 are arranged in the canonical form, typically found in enzymatically active PTP proteins, such as human LMW-PTP. Therefore, the proteins are expected to be capable of mediating dephosphorylation reactions.

Structural analysis of the phosphoarginine phosphatase activity of DARP

In contrast to a previous report classifying DARP as a pseudophosphatase (Hatzihristidis et al., 2015), a study by Fuhrmann et al. (2013) identifying B. subtilis YwlE as a bacterial protein arginine phosphatase showed that DARP (called CG31469 in their report) also exhibits a phosphoarginine phosphatase activity. We, therefore, analyzed the structure of DARP to account for its nonconventional substrate preference. In YwlE, the threonine residue (Thr11) located at position 5 of the P-loop, which is occupied by isoleucine in most LMW-PTPs (Fig. 3A), is critical for the enzyme’s protein arginine phosphatase activity. In the YwlE structure, an arginine residue from a crystallographically adjacent molecule protrudes into the catalytic pocket, mimicking a substrate (Fig. 3B, left). The hydroxyl group of YwlE Thr11 points toward the catalytic pocket and forms hydrogen bonds with the arginine guanidinium group through its hydroxyl group, and therefore enhances the accessibility of the substrate phosphoarginine (Fuhrmann et al., 2013). This key threonine residue is also conserved in the P-loop of DARP at the same position as Thr11 (Fig. 3A). In the DARP structure, however, the hydroxyl group of Thr11 points away from the catalytic cleft due to the absence of a protruding and directing arginine residue. Instead, a histidine residue protrudes from a crystallographically adjacent molecule, but its side chain does not interact with the hydroxyl group of Thr11 (Fig. 3B, right). We assume that the hydroxyl group of Thr11 in DARP reorientates toward the catalytic pocket during the enzymatic reaction and thereby participates in phosphoarginine recognition like the corresponding group of the threonine residue in YwlE.

Next, to analyze the accessibility of the catalytic pocket more precisely, we structurally compared the size of the catalytic pocket between eight LMW-PTPs, including DARP and Primo-1, that contain a bound negatively charged ion or a phosphotyrosine-mimetic molecule. Intriguingly, DARP has a characteristically narrow catalytic pocket compared to that of other LMW-PTPs, while the Primo-1 catalytic pocket is comparable to those of canonical enzymes (Fig. 3C). The distance between the Cγ atom of Trp43 and the Cζ atom of Tyr127 is 7.8 Å in DARP, which is shorter than the corresponding distances in human LMW-PTP (9.3 Å), Primo-1 (9.6 Å), and SP-PTP (9.8 Å). Consistently, structural alignments demonstrate that phosphoarginine, rather than phosphotyrosine, might fit into the narrow active pocket of DARP (Fig. 3D). Atomic-level structural analysis revealed several features that presumably contribute to the shape of the DARP catalytic pocket: the bulky side chain of Trp43, which is pushed toward the cleft by the aliphatic chain of Arg41; nonconserved electrostatic interactions between Arg41 and Glu92; and a hydrogen bond between Asn44 and Tyr127 that attracts this tyrosine residue toward the pocket (Fig. 3E). These features are unique to DARP, as they are absent in other LMW-PTP structures, including that of Primo-1, YwlE, and human LMW-PTP (Fig. 3E, Supplementary Fig. S1C). In contrast, the catalytic pocket composition of Primo-1 is highly homologous to that of conventional LMW-PTPs such as human LMW-PTP (Supplementary Fig. S1C). Collectively, these data strongly suggest that the accessibility of the catalytic pocket region, which is controlled by the presence of nonconserved threonine in the P-loop region (Figs. 3A and 3B) and the pocket size (Figs. 3C-3E), determines the substrate preference of DARP.

Biochemical characterization of Primo-1 phosphatase activity

Our structural data implied that DARP and Primo-1 might be enzymatically active phosphatases rather than pseudophosphatases. In vitro biochemical assays were performed to verify the phosphatase activity of the two proteins. First, the dephosphorylation capability of Primo-1, which was assumed to be a canonical LMW-PTP, was examined using pNPP, a widely used phosphatase substrate. Fig. 4A shows that recombinant Primo-1 was active on pNPP, confirming this protein as a catalytically active phosphatase. Kinetic constants were calculated using the Michaelis–Menten curves and the Lineweaver–Burk plots. Accordingly, the k_cat, k_M, and k_cat/k_M values of Primo-1 toward pNPP were determined as 59.8 min–¹, 0.520 mM, and 115 min–¹ mM–¹ (1.92 × 10³ s–¹ M–¹), respectively (Fig. 4B). Next, we measured the dephosphorylation activity of Primo-1 toward phosphotyrosine. The reaction of Primo-1 with phosphotyrosine released free phosphates, as evidenced by the absorbance at 360 nm (Fig. 4C), demonstrating that Primo-1 is able to dephosphorylate phosphotyrosine. The dephosphorylation activity of Primo-1 toward pNPP (Fig. 4A) and phosphotyrosine (Fig. 4C) was abrogated by substituting the catalytic cysteine residue (Cys9) to serine, further supporting that Primo-1 is a cysteine-based phosphatase.

Biochemical characterization of DARP phosphatase activity

Next, we conducted in vitro biochemical assays to verify the phosphatase activity of DARP toward four types of phosphoamino acids: phosphotyrosine, phosphoserine, phosphothreonine, and phosphoarginine. Recombinant DARP was only active on phosphoarginine, as evidenced by the absorbance at 360 nm, clearly demonstrating that DARP is an enzymatically active arginine phosphatase (Fig. 5A). Substitution of the catalytic Cys7 residue to serine abrogated the dephosphorylation activity of DARP (Fig. 5A). Initial velocity data were obtained to calculate the kinetic constants. From the Michaelis–Menten curve and Lineweaver–Burk double reciprocal plots, k_cat, k_M, and k_cat/k_M of DARP toward phosphoarginine were determined as 15.4 min–¹, 1.34 mM, and 11.5 min–¹ mM–¹ (192 s–¹ M–¹), respectively (Fig. 5B). The kinetic constants of DARP toward phosphotyrosine, phosphoserine, and phosphothreonine were not calculable.

The enzymatic property of DARP was further analyzed by comparing the specific activities of wild-type and mutant proteins toward phosphoarginine and phosphotyrosine. Compared to the wild-type, substitution of Thr11 to isoleucine caused a remarkable decrease in the dephosphorylation of phosphoarginine and a noticeable increase in the dephosphorylation of phosphotyrosine (Fig. 5C). These results support the finding of the previous study reporting that the threonine residue at position 5 of the P-loop significantly contributes to the substrate specificity of DARP (Fuhrmann et al., 2013). We also mutated Tyr127 to phenylalanine to abrogate the hydrogen bond between Tyr127 and Asn44 (Fig. 3E), which, according to our crystal structure, is expected to widen the catalytic pocket of DARP and thereby improve phosphotyrosine accessibility (Figs. 3C-3E), without critically altering the active loop arrangement. Activity tests demonstrated that the Y127F mutant was slightly less active on phosphotyrosine than the wild-type enzyme. However, the dephosphorylation activity of the double mutant T11I·Y127F toward phosphotyrosine was noticeably higher than that of the T11I single mutant protein, suggesting that the pocket size of DARP (Fig. 3) is an additional factor controlling the access of phosphotyrosine into the catalytic site (Fig. 5C). Collectively, our biochemical analysis indicates that DARP is a catalytically active arginine phosphatase, and its substrate specificity depends on the amino acid composition of the catalytic pocket region.

In this study, we demonstrated that the two Drosophila proteins DARP and Primo-1 have not only the canonical LMW-PTP protein fold (Fig. 1) but also well-ordered P-loop and active site-surrounding loops (Fig. 2), implying that their catalytic pocket region are suitable for conducting the dephosphorylation reaction. More critically, our biochemical analysis proved that both enzymes exhibited phosphatase activity (Figs. 4 and 5). Taken together, we conclude that DARP and Primo-1 are catalytically active phosphatases rather than pseudophosphatases. One critical difference between the two proteins is that while Primo-1 is assumed to be a conventional phosphotyrosine-targeting phosphatase, DARP has a noncanonical substrate specificity toward phosphoarginine, consistent with the previous report (Fuhrmann et al., 2013). It is known that PTPs belonging to the same subgroup and thus sharing overall folding and P-loop motif sequence can target different amino acid substrates and even distinct molecules. Differences in one or two amino acids, particularly in the signature P-loop motif, can significantly affect the three-dimensional composition of the catalytic pocket region and thereby alter or even abrogate the enzymatic activity of a PTP protein. Representatively, the secondary PTP domain D2 of all receptor-type PTPs are highly homologous to their primary PTP domain D1; however, D2 domains are enzymatically inactive and known as pseudophosphatases, with their various amino acid alterations in the P-loop motif and/or catalytic pocket region (Andersen et al., 2001). A recent study showed that the D1 domain of PTPRU is also a pseudophosphatase that, due to its atypical P-loop conformation, has no phosphatase activity (Hay et al., 2020). Altered P-loop arrangement is also a critical factor severely restricting the phosphatase activity of several DUSP proteins such as DUSP6, DUSP9, and Tk-PTP (Fjeld et al., 2000; Hong et al., 2005; Yun et al., 2018; Zhou and Zhang, 1999). In this study, we confirmed that DARP prefers phosphoarginine over phosphotyrosine as a substrate (Fig. 5), which relies on its nonconserved and hydrophilic threonine residue instead of hydrophobic isoleucine in the P-loop motif (C⁷IGNTCR¹³) and narrow catalytic pocket (Fig. 3). Therefore, the present crystal structure of DARP is the first and only three-dimensional structure of a eukaryotic protein arginine phosphatase.

Arginine, one of the twenty standard amino acids, contains a side chain composed of three-carbon aliphatic straight chain capped by a positively charged guanidinium group where a phosphate moiety can be covalently linked to. To date, phosphorylation and dephosphorylation of this amino acid, which are nonconventional compared to those of serine, threonine, and tyrosine, were mostly investigated in gram-positive bacteria; these reactions are involved in diverse bacterial physiological processes such as stress responses and protein degradation (Elsholz et al., 2012; Fuhrmann et al., 2016; Schmidt et al., 2014; Trentini et al., 2016). Contrastively, even though arginine phosphorylation was first reported in mouse proteins more than 30 years ago (Matthews, 1995; Wakim and Aswad, 1994), its precise biological functionality and associated proteins in eukaryotes remain to be elucidated. Therefore, the physiological role of DARP in Drosophila signaling pathways, and whether protein arginine phosphatases exist and function in other species, including human, are important and challenging issues to be addressed.

In summary, we determined the crystal structures of DARP and Primo-1, which are catalytically active phosphatases with an LMW-PTP fold. Structural and biochemical analyses revealed that DARP and Primo-1 dephosphorylate phosphoarginine and phosphotyrosine, respectively. The results of the present study will aid in finding unknown substrates for these proteins and signaling pathways in which the proteins are involved. This, in turn, might expand our understanding of post-translational phosphorylation and dephosphorylation controlling signaling pathways in Drosophila.

We thank the beamline 7A at the Pohang Accelerator Laboratory in Korea for assistance during the data collection. This work was supported by the National Research Foundation of Korea funded by the Ministry of Science and ICT (MSIT; 2019M3E5D6063955) and by the Korea Research Institute of Bioscience and Biotechnology Research Initiative Program (to B.K.).

H.S.L. and Y.M. conceived and analyzed experiments and wrote the manuscript. H.C.S. provided expertise and feedback. S.J.K. provided supervision. B.K. conceived and analyzed experiments, wrote the manuscript, secured funding, and provided supervision.

The authors have no potential conflicts of interest to disclose.

Data collection and structure refinement statistics