Structural insights into the dual Ca²⁺-sensor-mediated activation of the PPEF phosphatase family

Abstract

Serine/threonine-protein phosphatases with EF-hands (PPEFs) are a family of highly conserved proteins implicated in cancer and neuronal degeneration. The initially characterized member, Drosophila melanogaster retinal degeneration C (RDGC) contains a calmodulin (CaM)-interacting extended-IQ motif and a Ca²⁺-binding EF-like/EF-hand tandem. However, the molecular regulation of PPEF is poorly understood. In this study, we use cryogenic-electron microscopy to delineate the structures of the RDGC/CaM holoenzyme. In the absence of Ca²⁺, CaM and the EF-like/EF-hand tandem allow the extended-IQ motif to block substrate access to the catalytic sites, constituting an auto-inhibitory mechanism. Upon Ca²⁺ binding, CaM and the EF-like/EF-hand tandem drive drastic conformational changes in the extended-IQ motif to unlock the catalytic sites. This dual Ca²⁺-sensor-mediated activation is evolutionarily conserved in mammals. This study provides mechanistic insight into the molecular activation of PPEFs, paving the way for the development of therapeutic strategies for PPEF-related human diseases.

PPEFs are highly conserved phosphatases. Here, authors use cryo-EM to solve the holoenzyme structures of the initially characterized PPEF family member, Drosophila RDGC, with/without Ca²⁺, uncovering a dual Ca²⁺-sensor-mediated activation mechanism.

Introduction

Reversible protein phosphorylation is a fundamental regulatory mechanism governing various biological processes¹. In contrast to the numerous protein serine/threonine kinases in the human genome, there are relatively few protein serine/threonine phosphatases (PSPs)^2–4. PSPs are categorized into three structurally distinct families, including phosphoprotein phosphatases (PPPs), which are primary serine/threonine phosphatases involved in a variety of cellular processes and diseases^5–7. PPPs are characterized by three main features: a conserved catalytic domain⁸, abundant regulatory subunits⁹, and reversible auto-inhibitory regulatory mechanisms^10,11. PPPs can be classified into protein phosphatase 1 (PP1), PP2A, PP2B, PP4, PP5, PP6, and protein phosphatase with EF-hands (PPEF) subfamilies¹².

Despite the well-characterized functions of the other six PPP members, the structural and mechanical understanding of PPEF¹³ remains unclear, limiting our comprehension of its biological functions and associated diseases. PPEF1 and PPEF2 are two subtypes of PPEF in mammals. PPEF1 displays high expression levels in sensory neurons, including trigeminal ganglion, dorsal root ganglia, inner ear and neural crest-derived cranial sensory ganglia^14,15. PPEF1 is also expressed in the testis¹⁶, uterus, spinal cord, bone marrow, and lymphoblasts^13,17. Moreover, PPEF1 plays a role in breast cancer tumorigenesis, serving as a promising biomarker for prognosis and diagnosis¹⁸. PPEF1 is also a potential mRNA vaccine target for stomach adenocarcinoma¹⁹. PPEF2, which is expressed in the testis, oocytes, skeleton, and cardiomyocytes¹³, is a novel interacting partner and a negative regulator of apoptosis signal-regulating kinase-1, which is implicated in cancer²⁰. Missense mutations in PPEF2 are associated with inborn genetic diseases (ClinVar https://www.ncbi.nlm.nih.gov/clinvar).

Drosophila melanogaster retinal degeneration C (RDGC), the prototype PPEF family member, is essential for the dephosphorylation of retinal rhodopsin^21,22 and transient receptor potential (TRP) channels at S936²³. RDGC is prominently expressed in the retina, ocelli, and mushroom bodies of the Drosophila melanogaster brain²⁴. Deletion of RDGC results in hyperphosphorylation of rhodopsin²¹ and TRP²³, delayed light responses and retinal degeneration^21,25. RDGC consists of an N-terminal extended-IQ motif, a highly conserved catalytic domain, and a C-terminal EF-like/EF-hand tandem. Three splicing variants of RDGC have been identified; the N-terminal domain and solubility of the splicing variants differ²⁶. The extended-IQ motif includes an N-terminal classical IQ motif (aa 1–25), which interacts with calmodulin (CaM)²⁷, and a C-terminal extended α-helix (aa 26–61) (Fig. 1a). A mutation in the classical IQ motif impedes rhodopsin dephosphorylation in vivo and hinders photoreceptor signal termination²⁷. The Ca²⁺ binding capacity of the EF-hand domain has been confirmed in C. elegans, but the role of Ca²⁺ binding in the regulation of phosphatase activity is unclear²⁸. Sequence alignment analyses revealed remarkable conservation between RDGC and its mammal orthologs. Therefore, understanding the structural foundations of RDGC could provide valuable insights into the mechanisms regulating the PPEF subfamily.

Fig. 1. Overall structure of the auto-inhibited RDGC/CaM holoenzyme in the absence of Ca²⁺.

a Schematic representation of the domain organization of RDGC. Dro, Drosophila. b Cryogenic-electron microscopy (Cryo-EM) density map of the RDGC/CaM holoenzyme in the absence of Ca²⁺. Unless otherwise indicated, individual domains are displayed in distinct colors throughout the manuscript, as follows: CaM, dark orange; extended-IQ motif, lime green; catalytic domain, medium purple; EF-like domain, Indian red; and EF-hand domain, dodger blue. The map contour level is 0.16. c Schematic representation of the RDGC/CaM holoenzyme in the absence of Ca²⁺. A close-up view of the catalytic domain, encompassing the extended-IQ motif, EF-like domain, EF-hand domain, and CaM, is shown. The catalytic sites are labeled. d Atomic model and cryo-EM densities of the RDGC catalytic domain and catalytic sites, with metal ions 1 (M1) and 2 (M2) shown as spheres. The map contour level is 0.214. e Extended-IQ motif of RDGC, with side chains displayed as sticks within the cryo-EM densities. The map contour level is 0.1. f Cryo-EM densities and atom representation for the EF-like domain and EF-hand domain of RDGC. The map contour level is 0.13. g Electron density covering the CaM atomic model. Superposition of CaM from the cryo-EM structure (PDB ID: 8JFW) onto the isolated Ca²⁺-free CaM N and C lobes from the crystal structure (PDB ID: 8ZLX) is shown. Ca²⁺-binding pockets are shown in the EM-density maps. The map contour level is 0.1.

In this work, we purify the RDGC/CaM holoenzyme and use cryogenic-electron microscopy (cryo-EM) to elucidate its structural dynamics and regulatory mechanisms. In the absence of Ca²⁺, CaM and the EF-like/EF-hand tandem allow the extended-IQ motif to block substrate access to the catalytic sites, constituting an auto-inhibitory mechanism. Upon Ca²⁺ binding, CaM and the EF-like/EF-hand tandem act as two Ca²⁺-sensors, inducing conformational changes in the extended-IQ motif to unlock catalytic sites. This competent state represents a “ready-to-go” configuration, facilitating efficient substrate dephosphorylation. This dual Ca²⁺-sensor-mediated activation is conserved during evolution, providing structural and mechanistic information about the functions of PPEF family members.

Results

Ca²⁺-dependent phosphatase activity of the RDGC/CaM holoenzyme

An SF9 insect cell co-expression system was used to express and purify full-length wild type Drosophila melanogaster RDGC in complex with CaM (Supplementary Fig. 1a). The phosphatase activity of the RDGC/CaM holoenzyme was significantly enhanced in the presence of 1 mM CaCl₂ compared with buffer containing 1 mM EDTA, which chelates Ca²⁺ (Supplementary Fig. 1b). Thus, the RDGC/CaM holoenzyme is inactive in the absence of Ca²⁺, highlighting the crucial role of Ca²⁺ as a modulator of RDGC/CaM phosphatase activity. To explore protein conformational changes in response to Ca²⁺, we used size-exclusion chromatography coupled with multi-angle light scattering (SEC-MALS). The elution volume of the RDGC/CaM holoenzyme was reduced in the presence of Ca²⁺, indicating extensive conformational changes in response to Ca²⁺ (Supplementary Fig. 1c).

Structure of auto-inhibited RDGC/CaM in the absence of Ca²⁺

To elucidate the structural basis and underlying auto-inhibitory mechanism of the RDGC/CaM holoenzyme, we prepared the RDGC/CaM holoenzyme sample in the absence of Ca²⁺. A density map at an overall resolution of 3.6 Å was developed from 6,822 cryo-EM micrographs. The holoenzyme formed a heterodimer of ~78 Å in height and 101 Å in width (Fig. 1b, c; Supplementary Fig. 2 and Table 1).

Table 1.

Cryo-EM data collection, refinement and validation statistics

	RDGC/apo-CaM(EMDB-36218)(PDB 8JFW)	RDGC/Ca2+-CaM(EMDB-36219)(PDB 8JFY)
Data collection and processing
Magnification	105,000 x	105,000 x
Voltage (kV)	300	300
Electron exposure (e–/Å2)	50	50
Defocus range (μm)	1.0–1.6	1.0–1.6
Pixel size (Å)	0.855	0.855
Symmetry imposed	C1	C1
Initial particle images (no.)	2090955	815266
Final particle images (no.)	202878	56344
Map resolution (Å)	3.64	2.79
FSC threshold	0.143	0.143
Map resolution range (Å)
Refinement
Map sharpening B factor (Å2)	-202.8	-67.9
Model composition
Non-hydrogen atoms	6079	11870
Protein residues	758	1472
Ligands	Zn:1 Fe:1	PO4:2 Ca:1 2 Zn:2 Fe:2
R.m.s. deviations
Bond lengths (Å)	0.003	0.003
Bond angles (°)	0.601	0.609
Validation
MolProbity score	1.42	1.43
Clashscore	5.49	7.01
Poor rotamers (%)
Ramachandran plot
Favored (%)	97.33	97.81
Allowed (%)	2.67	1.92
Disallowed (%)	0	0.27

The central structure framework of the globular catalytic domain of the enzyme was a β-sandwich motif. The catalytic domain consisted of two mixed β-sheets; sheet 1 contained six β-strands (sequence: β1-β5-β6-β9-β11-β10), and sheet 2 contained five β-strands (sequence: β4-β3-β2-β12-β13). The two β-sheets were flanked by α-helical structures (Fig. 1d, upper panel). The catalytic sites, which consisted of Asp158, His160, Asp187, Asn219, His271, and His360 positioned above the closed end of the β-sheet and delineated by several loops (Fig. 1d, lower panel), were obscured by the extended-IQ motif. The extended-IQ motif consisted of a lever-shaped structure and an elongated predominantly nonpolar α-helix that formed a complementary surface with the hydrophobic groove created by the catalytic domain, EF-like/EF-hand tandem and CaM (Fig. 1e). The EF-like and EF-hand domains were both characterized by a pair of helix-loop-helix motifs arranged in an anti-parallel manner. Notably, residues Asp539, Asp541, Glu545, Glu550, Asp579, Asn581, and Asp583 within the EF-hand domain are crucial for Ca²⁺-binding. These residues located in the central loop spanning from α20–α21 to α22–α23 (Fig. 1f). CaM consisted of N and C lobes that formed a C-shaped structure. Each lobe contained a pair of EF-hand motifs connected by a flexible central loop; one EF-hand motif could bind one Ca²⁺ ion (Fig. 1g).

The crystal structure of the mPPEF2 IQ motif (aa 22–39)/apo-CaM complex (PDB ID: 8ZLX) at 2.5 Å resolution was also solved to validate the cryo-EM structure. The crystal structure revealed that neither the CaM N nor C lobes bind Ca²⁺, with only the CaM C lobe interacting with the classical IQ motif (Supplementary Figs. 3a, 3b, 4a, 4b, 6a, 6c; Table 2). The CaM structure obtained from cryo-EM (PDB ID: 8JFW) was aligned with the isolated Ca²⁺-free CaM N and C lobes from the crystal structure (PDB ID: 8ZLX), yielding root-mean-square deviation (RMSD) values of 1.044 Å and 0.930 Å, respectively. This alignment suggests that the cryo-EM structure was captured in a Ca²⁺-free state (Fig. 1g). In summary, the CaM and EF-like/EF-hand tandem synergistically stabilized the extended-IQ motif and blocked the catalytic sites via an auto-inhibited conformation.

Table 2.

Data collection and refinement statistics (molecular replacement)

	RDGC IQ (aa 1−25)/CaM(PDB 8ZLW)	mPPEF2 IQ (aa 22–39)/CaM(PDB 8ZLX)
Data collection
Space group	P 1 21 1	P 43
Cell dimensions
a, b, c (Å)	52.66, 61.59, 88.00	134.45, 134.45, 35.94
α, β, γ (°)	90.00, 94.17, 90.00	90, 90, 90
Resolution (Å)	29.26-2.20 (2.279-2.200) *	19.83 -2.50 (2.589-2.500)
Rsym or Rmerge		0.15
I / σI	2.27 (at 2.20 Å)	2.26 (at 2.50 Å)
Completeness (%)	98.7 (99.08)	99.7 (100.0)
Redundancy	5.5 (5.7)	12.5 (12.3)
Refinement
Resolution (Å)	29.26-2.20 (2.279-2.200)	19.83 -2.50 (2.85-2.50)
No. reflections	28328	22846
Rwork / Rfree	21.61/25.37	21.97/27.78
No. atoms
Protein	539	662
Ligand/ion	CA:4	0
Water	60	16
B-factors	44.71	55.80
R.m.s. deviations
Bond lengths (Å)	0.008	0.009
Bond angles (°)	0.949	1.122

*Values in parentheses are for highest-resolution shell.

Structural basis of the auto-inhibited RDGC/CaM holoenzyme

Structural analysis revealed that the RDGC/CaM holoenzyme was captured in an auto-inhibited form through a network of interactions. CaM bound to the N-terminus (aa 1–25) of the extended-IQ motif. Notably, Lys14, Arg17, and Arg18 in the extended-IQ motif engaged in charge-charge interactions and hydrogen bonds with Glu48, Glu46, and Asn43, respectively, in the N-lobe of CaM (interface 1) (Fig. 2a). The hydrophobic residues Ile12, Trp15, and Tyr16 of the extended-IQ motif exhibited a high affinity for Ile86, Phe90, Val122, Met125, and Val143 in the hydrophobic pocket of the C-lobe of CaM (interface 2) (Fig. 2b). Substitution of the hydrophobic residues Ile12, Trp15, and Tyr16 with alanine disrupted the interaction between the RDGC classical IQ motif and CaM, confirming the pivotal role of these residues in mediating interaction (Fig. 2f; Supplementary Table 1).

Fig. 2. Structural basis for stabilizing the auto-inhibited state of the RDGC/apo-CaM holoenzyme.

a–e Five interfaces involved in forming the auto-inhibited RDGC/CaM holoenzyme. The key residues are displayed in stick representation, with salt bridges and hydrogen bonds indicated by dashed lines. Interface 1 (a) involves the extended-IQ motif and CaM N lobe interaction. Interface 2 (b) is formed by the extended-IQ motif and the CaM C lobe. Interface 3 (c) comprises the extended-IQ motif and the EF-like domain. Interface 4 (d) is formed by the extended-IQ motif and the EF-hand domain. Interface 5 (e) involves the interaction between the extended-IQ motif and the catalytic domain in the blocking conformation. The electrostatic surfaces between CaM N lobe and extended-IQ motif are shown in (a). The hydrophobic surfaces between extended-IQ motif and CaM C lobe, extended-IQ motif and EF-like domain, extended-IQ motif and EF-hand domain and the catalytic domain are shown in (b–e), respectively. aa, amino acids. f Summary of the dissociation constants demonstrating that mutations of critical residues in the interaction interface can abolish the interactions between the classical IQ motif and CaM. Results from three independent isothermal titration calorimetry (ITC) experiments are further presented in Supplementary Table 1. WT, wild type. N.d., Not detectable. g Phosphatase activity assays showing that the catalytic domain alone, without the auto-inhibitory components (CaM, extended-IQ motif and EF-like/EF-hand tandem), exhibits higher phosphatase activity in the absence of Ca²⁺. The freshly purified RDGC full-length/CaM wild type holoenzyme activity in Ca²⁺-free conditions was set to 100%, with the relative activities of the mutants presented as percentages of the RDGC full-length/CaM holoenzyme activity. Mean ± SEM for three different experiments is shown. Source data are provided as a Source Data file.

The EF-like/EF-hand tandem interacted with the extended-IQ motif C-terminus (aa 26–59). Cys29, Ile33, and Leu37 in the extended-IQ motif were situated within a hydrophobic pocket composed of Leu446, Met470, Leu476, Leu478, Pro479, Leu482, Leu483, and Leu487 in the EF-like domain (interface 3) (Fig. 2c). Additionally, Leu48, Phe51, Phe52, Leu55, and Ile56 from the extended-IQ motif possessed large, nonpolar side chains that significantly enhanced the hydrophobic interactions with Leu524, Ala533, Ile537, Ile538, Ala554, Leu557, Leu558, and Met562 in the EF-hand domain (interface 4) (Fig. 2d).

The extended-IQ motif (aa 24–61) played a pivotal role in blocking the catalytic sites by covering the catalytic sites like a lid. Notably, Arg333 at the catalytic sites interacted with Tyr39 of the extended-IQ motif via cation-π interactions. The guanidinium group of Arg333 formed hydrogen bonds with the backbone carbonyl oxygen of Tyr39 and Glu42 to block the catalytic sites. Hydrophobic interactions at both ends of the extended-IQ motif involve Trp31, Phe34, Ala46, and Tyr49, which are nestled in hydrophobic pockets formed by Cys362, Pro364, Tyr385, Ala386, Ile387, and Trp310, Phe314, Trp318, respectively. This auto-inhibition was reinforced by hydrogen bonds between Asn53 and Tyr49 of the extended-IQ motif and Lys307 and Gln311 of the catalytic domain (interface 5) (Fig. 2e). The phosphatase activity of the catalytic domain alone was higher than that of the holoenzyme, confirming the auto-inhibitory functions of CaM, the EF-like/EF-hand tandem and the extended-IQ motif (Fig. 2g).

Structure of RDGC/CaM in the presence of Ca²⁺

The holoenzyme was purified in a buffer containing 10 mM CaCl₂, yielding a uniformly high-quality protein sample to elucidate the activation mechanism of RDGC/CaM. Subsequent processing of 815,266 particles via cryoSPARC software generated a high-resolution map with an overall resolution of 2.8 Å, that aligned well with the protein backbone (Fig. 3a; Supplementary Fig. 7 and Table 1).

Fig. 3. Overall structure of the RDGC/CaM holoenzyme in the presence of Ca²⁺.

a Cryo-EM density map of the heterotetramer RDGC/CaM holoenzyme in the presence of Ca²⁺. The two catalytic domain protomers are indicated in medium purple and silver. The map contour level is 0.079. b Schematic representations of the heterotetramer RDGC/CaM holoenzyme in the presence of Ca²⁺ following 180° vertical rotation. c The top view of (b) reveals the dimer interface in the catalytic domains. d Schematic representation of an RDGC/CaM protomer viewed with 90° vertical rotation demonstrates the substrate-accessible catalytic sites. e Superposition of CaM from the cryo-EM structure (PDB ID: 8JFY) onto the Ca²⁺-CaM N and C lobes from the crystal structure (PDB ID: 8ZLW). Ca²⁺ binding sites are shown in the EM-density maps. Ca²⁺ ions are depicted as spheres. The map contour level is 0.123. f Extended-IQ motif of RDGC, with side chains displayed as sticks within the cryo-EM densities. The map contour level is 0.079. g Representative atomic model and cryo-EM densities of the EF-like domain and EF-hand domain in the presence of Ca²⁺. Ca²⁺ binding sites are shown in the EM-density maps, with Ca²⁺ ions depicted as spheres. The map contour level is 0.092. h Cryo-EM densities and atom representation for the catalytic domain and catalytic sites, coupled with phosphate, showing the dinuclear metal center of RDGC. Metal ion 1 (M1) and metal ion 2 (M2) are depicted as spheres. The map contour level is 0.211.

In the presence of Ca²⁺, RDGC and CaM formed a butterfly-shaped heterotetramer with a stoichiometric ratio of 2:2; the subunits were denoted as RDGCa, RDGCb, CaMa, and CaMb. Interactions between the two catalytic domains of RDGC stabilized the 70 Å by 186 Å heterotetramer (Fig. 3a, b). Arg191 and Arg233 of RDGCa formed a hydrogen bond network with Thr331 and Leu295 of RDGCb, and Arg228 and Arg233 formed cation-π interactions with Trp318 and Trp310 (Fig. 3c). Each RDGC subunit also formed a heterodimer with one CaM, and the catalytic sites were located at the interface formed by the two heterodimers. Protein substrates were sterically prevented from accessing the catalytic sites of RDGC. However, the strong phosphatase activity of the RDGC/CaM holoenzyme in Ca²⁺ buffer and the monomeric behavior of RDGC/CaM holoenzyme in Ca²⁺ buffer suggest that the heterotetrameric structure of the RDGC/CaM holoenzyme may represent a pre-competent state before full activation (Fig. 3d).

CaM in the Ca²⁺-bound RDGC/CaM holoenzyme adopted a conformation in which Ca²⁺ occupied each EF-hand loop, consistent with the high-resolution (2.2 Å) crystallographic study of the RDGC IQ motif (aa 1–25)/Ca²⁺-CaM complex (PDB ID: 8ZLW, RMSD: 0.751 Å) (Fig. 3e; Supplementary Fig. 3c, 3d, 5a, 5b, 6b, and Table 2). Both structures bound four Ca²⁺ ions, one in each EF-hand loop of CaM (Fig. 3e). The extended-IQ motif adopted an elongated lever-arm configuration in which the rod-like C-terminal region was twisted by the Ca²⁺ bound EF-like/EF-hand tandem (Fig. 3f). Although both the EF-like and EF-hand domains folded into helix-loop-helix structures, only the EF-hand domain coordinated Ca²⁺ via the conserved residues (Asp539, Asp541, Glu550, Asn581, Asp583, and Lys585) in the loop²⁹ (Fig. 3g). In the presence of Ca²⁺, the catalytic domain exhibited a similar structural configuration to the structure in absence of Ca²⁺. The catalytic sites coordinated with two metal ions (M1 and M2) and a bound phosphate ion (Fig. 3h). Phosphatase activity assays demonstrate that the inclusion of Fe³⁺, Fe²⁺, Mn²⁺, or Zn²⁺ maintains basal enzymatic activity, with Zn²⁺ exhibiting the highest activity. These results suggest that these metal ions are potential candidates for coordinating the active site. Furthermore, the presence of Ca²⁺ significantly enhances the enzyme’s catalytic activity (Supplementary Fig. 8)³⁰. Besides, residues 606–661 were invisible in the final model due to their inherent flexibility.

Conformational changes in the extended-IQ motif induced by Ca²⁺ binding

CaM and the EF-like/EF-hand tandem induced a flipping and bend of the extended-IQ motif in Ca²⁺ buffer. CaM was a critical force point within the extended-IQ motif. Ile6 and Ile10 from the extended-IQ motif inserted into the deep pocket of the CaM N lobe, which included residues Phe13, Ala16, Leu19, Phe20, Leu33, and Met72 (interface 1) (Fig. 4a). The hydrophobic residues Phe11, Ile12, Trp15, and Tyr16 from the extended-IQ motif were anchored within the hydrophobic pocket of the CaM C lobe, surrounded by Ile86, Phe93, Met110, Met125, Ala129, Val143, and Met145 (interface 2) (Fig. 4b). Additionally, the CaM C lobe formed salt bridges with Lys172 and Arg396 from the catalytic domain via Glu124 and Asp130, respectively, and the carboxyl group of Ile131 formed a hydrogen bond with the backbone of Ile409 in the catalytic domain (interface 3) (Fig. 4c). Replacing Ile6, Ile10, Ile12, and Trp15 with glutamic acid decreased the binding affinity of the classical IQ motif and CaM (Fig. 4g; Supplementary Table 1). Furthermore, simultaneously substituting Ile6, Ile10, Phe11, Ile12, Trp15 and Tyr16 with glutamic acid disrupted the RDGC/CaM interaction with/without Ca²⁺ (Fig. 4h), resulting in a significant reduction in phosphatase activity (Fig. 4j). Under Ca²⁺ condition, the I12E mutant exhibited a 236-fold decrease in affinity and dramatically reduced phosphatase activity, consistent with previous studies²⁷ (Fig. 4g, j).

Fig. 4. Activation of the RDGC/CaM holoenzyme by dual Ca²⁺-sensors.

a–f Five interfaces are involved in forming the activated RDGC/CaM holoenzyme. Interface 1 (a) involves the extended-IQ motif and the CaM N lobe interaction. Interface 2 (b) is formed by the extended-IQ motif and CaM C lobe. Interface 3 (c) comprises the catalytic domain and CaM C lobe. The bending of the extended-IQ motif (d) is induced by the pivotal force of the EF-like/EF-hand tandem, leading to the formation of interface 4 (e). Interface 5 (f) involves the interaction between the catalytic and EF-like domains. The hydrophobic surfaces on CaM, the EF-like domain, and the EF-hand domain are shown in (a), (b), and (e), respectively. aa, amino acids. g ITC assays showing that mutations of critical residues significantly weaken the interaction between the classical IQ motif and CaM. WT, wild type. h GST pull-down assays showing that mutations in the extended-IQ motif abolish the interaction between full-length RDGC and CaM. GST-CaM is the probe protein, GST is the control, and RDGC is the target protein. After pull-down with GST-CaM and GST, proteins were subjected to SDS-PAGE, transferred to a nitrocellulose membrane, and immunoblotted (n = 3). i Summary of ITC assays showing that the EF-hand domain is responsible for Ca²⁺ sensing in EF-like/EF-hand tandem. The EF-hand domain Ca²⁺ binding deficient mutant: D539A, D541A, E545A, E550A, N581A, D583A, and K585A. All ITC titrations in (g) and (i) were performed in triplicate and presented in Supplementary Table 1. N.d., Not detectable. j Phosphatase activity results of different mutants. The CaM binding deficient mutant: I6E, I10E, F11E, I12E, W15E, and Y16E. The Ca²⁺ binding deficient mutant: D539A, D541A, E545A, E550A, N581A, D583A, and K585A. The lever arm deficient mutant: CNWQ29/30/31/32GGGG. Freshly purified RDGC full-length/CaM holoenzyme activity in Ca²⁺ conditions was set to 100%, with the relative activities of the mutants presented as percentages of the RDGC full-length/CaM holoenzyme activity. Mean ± SEM for three different experiments is shown. Source data are provided as a Source Data file.

In addition to CaM, the EF-like/EF-hand tandem functioned as a second Ca²⁺-sensor and exerted a pivotal force on the extended-IQ motif. In the presence of Ca²⁺, the region spanning from Asn30 to Gln43 twisted ~28°, and Gln43 moved ~14 Å compared with the structure of the extended-IQ motif in the absence of Ca²⁺ (Fig. 4d). In addition, the C-terminus of the extended-IQ motif was enveloped by the EF-like/EF-hand tandem in the presence of Ca²⁺. Residues Ile33, Phe34, Leu37, Ala40, Ala46, Tyr49, and Phe52 of the extended-IQ motif were ensconced within a cleft formed by the hydrophobic residues Val427 and Ala431 from the EF-like domain, and Leu588, Leu592, Leu597, Leu600, and His601 from the EF-hand domain (interface 4) (Fig. 4e). Additionally, residues Thr415, Lys416, Leu418, Ser419, Lys421, Gln422, Arg423, and Asp516 in the EF-like domain created a network of hydrogen bonds with Arg113, His171, Lys172, Leu175, and Lys102 from the catalytic domain (interface 5) (Fig. 4f). Substituting residues Asp539, Asp541, Glu545, Glu550, Asn581, Asp583, and Lys585 with alanines completely abolished Ca²⁺ binding of the EF-hand domain (Fig. 4i; Supplementary Table 1) and significantly decreased phosphatase activity (Fig. 4j).

Ca²⁺ stimulation triggered conformational adjustments in CaM and the EF-like/EF-hand tandem, synergistically exerting mechanical force along the extended-IQ motif and exposing the catalytic sites. The extended-IQ motif acted as a critical lever-arm structure, linking the dual Ca²⁺-sensors and facilitating RDGC activation. RDGC Cys29, Asn30, Trp31, and Gln32 in the extended-IQ motif were replaced with glycines, which is supposed to impair the transmission of mechanical torque and disrupt the flipping of the extended-IQ motif induced by Ca²⁺ binding to CaM and the EF-like/EF-hand tandem. The GST pull-down results showed that RDGC CNWQ 29/30/31/32 GGGG mutant bound CaM in the absence of Ca²⁺ but not in the presence of Ca²⁺ (Fig. 4h) and significantly reduced the phosphatase activity compared with the wild type (Fig. 4j). The impairment of flipping function and potential steric-hindrance effects of the mutated extended-IQ motif might both contribute to the inability of the RDGC CNWQ 29/30/31/32 GGGG mutant to interact with CaM in the presence of Ca²⁺ (Supplementary Fig. 9).

Structural insights into the regulation of the RDGC/CaM holoenzyme by Ca²⁺

To investigate the regulatory role of the dual Ca²⁺-sensors in the RDGC/CaM holoenzyme, both local and global conformational changes during the transition from the auto-inhibitory to pre-competent states were determined. In the absence of Ca²⁺, CaM and the EF-like/EF-hand tandem bind to the extended-IQ motif and restrict access to the catalytic sites, assuming the auto-inhibitory conformation. In this state, CaM adopts an open C-shaped conformation with a longer distance between the N and C lobes. However, upon Ca²⁺ binding, the C-shaped conformation becomes more compact, holding the extended-IQ motif more tightly (Fig. 5a). The EF-like/EF-hand tandem also becomes more compact upon Ca²⁺ binding (Fig. 5b). The mechanical forces generated by this conformational rearrangement flip the extended-IQ motif to expose the catalytic sites. Despite these structural changes, the catalytic domain of RDGC is largely unchanged (RMSD: 0.545 Å), except the C-terminal β strand became a disordered loop (Fig. 5c). The minimum structural changes in the catalytic domain suggest that changes in the enzymatic activity of RDGC are mediated by the regulatory subunits^31,32 (Fig. 5d) rather than changes in the shape of the catalytic sites between the inactive and active states³³.

Fig. 5. Structural comparison of the RDGC/CaM holoenzyme in the auto-inhibitory and activation states.

a Comparison of the CaM structure between the apo and Ca²⁺-binding forms. b Comparison of the EF-like/EF-hand tandem structure between the apo and Ca²⁺-binding forms. c Superposition of the catalytic domain between the auto-inhibitory and pre-competent states. The RDGC C-terminal tail formed a β-strand structure before Ca²⁺ binding, but this tail was invisible due to the structural flexibility after Ca²⁺ binding. d Conformational changes of the extended-IQ motif between the apo and Ca²⁺-binding forms.

Conserved dual Ca²⁺-sensor-mediated activation mechanism in mammalian PPEF

Sequence comparisons revealed that PPEF and RDGC share remarkable similarities in primary sequence and secondary structure, particularly within the extended-IQ motif and catalytic and EF-hand domains. Key residues for CaM binding within the extended-IQ motif and Ca²⁺ binding in the EF-hand domain are conserved (Fig. 6a–c; Supplementary Fig. 10). Predictive modeling with AlphaFold demonstrated that PPEF’s structural conformation closely resembled the conformation of RDGC, indicating that PPEF and RDGC share a conserved regulatory mechanism (Fig. 6d). Biochemical assays confirmed that the phosphatase activity of PPEF1 increased in the presence of Ca²⁺ compared with the Ca²⁺-free condition, similar to RDGC (Fig. 6e). Two groups of Ca²⁺-sensing deficient mutants were introduced into Mus musculus PPEF1 (mPPEF1), one in the extended-IQ motif to disrupt its interaction with CaM and another in the EF-hand domain to inhibit Ca²⁺ binding. Notably, mutations in conserved residues within the extended-IQ motif and EF-hand domain significantly reduced phosphatase activity compared with the wild type, highlighting the critical role of the Ca²⁺-sensing residues (Fig. 6e). Overall, our data strongly supports the evolutionary conservation of the dual Ca²⁺-sensor-mediated activation mechanism within the PPEF subfamily.

Fig. 6. Distinct activation mechanism of mammalian PPEF: conserved dual Ca²⁺-sensors, divergent from PP2B.

a Schematic representation and domain organization of the PPEF family. Dro, Drosophila. b Sequence alignments of the extended-IQ motifs from RDGC and hPPEF1/2. Residues corresponding to the interaction interface of CaM are highlighted in red. Fully conserved residues are highlighted in blue. c Sequence alignments of the EF-hand domains from RDGC and hPPEF1/2. Residues corresponding to the interaction with Ca²⁺ are highlighted in red. Fully conserved residues are highlighted in blue. d The AlphaFold Protein Structure Database structures of hPPEF1/2 compared with the RDGC cryo-EM structure. The AlphaFold accession numbers are AF-O14829 (Homo sapiens PPEF1) and AF-O14830 (Homo sapiens PPEF2). e Phosphatase activity measurements of mPPEF1 (aa 10-650) contain extended-IQ motif or EF-hand mutant. Extended-IQ motif mutant contained I17E L21E V22E I23E W26E Y27E. EF-hand mutant contained D580A D582A L586A E591A N622A D624A N626A. The freshly purified mPPEF1 (aa 10-650)/CaM wild type holoenzyme activity in Ca²⁺ conditions was set to 100%, with the relative activities of the mutants presented as percentages of the mPPEF1 (aa 10-650)/CaM holoenzyme activity. Error bars show mean values ± SEM for three different experiments. Source data are provided as a Source Data file. f Structural comparison of the classical IQ and extended-IQ motifs. g Structural comparison of the CaM N and C lobes with the EF-like and EF-hand domains. h Crystal structure of Homo sapiens PP2B (PDB ID: 1AUI) in the auto-inhibited state. The A chain is light cyan, the B chain is thistle, and the auto-inhibitory domain is orange red. CaM bound to PP2B, leading to the removal of the auto-inhibitory domain from the active sites. CnA Calcineurin A. CnB Calcineurin B. i Superposition of RDGC and PP2B catalytic domains in the auto-inhibited states. j Structural comparison of RDGC and the corresponding PP2B in the auto-inhibited states.

Comparison of the extended-IQ motif binding modes of CaM and the EF-like/EF-hand tandem

The extended-IQ motif, a hallmark of RDGC, consisted of an N-terminal classical IQ motif and a C-terminus IQ-like motif, functioning as a lever with both rigidity and flexibility (Fig. 6f). The two Ca²⁺-sensors, CaM and EF-like/EF-hand tandem, at both ends of the extended-IQ motif exhibited both similarities and differences. Structural superimposition revealed a highly similar semi-open conformation in both the CaM N/C lobes and the EF-like/EF-hand tandem, with RMSD values of 1.266 Å and 1.355 Å, respectively (Fig. 6g). Despite this structural similarity, the spatial arrangement of the EF-like/EF-hand tandem on the extended-IQ motif diverged from the spatial arrangement of CaM. In the Ca²⁺-free and bound states, CaM engaged the extended-IQ motif by binding to the opposite sides via a rotational C-shaped wrapping motion. Conversely, the EF-like/EF-hand tandem interaction was characterized by side-by-side contact. In the absence of Ca²⁺, the EF-like/EF-hand tandem was positioned above the extended-IQ motif. However, the extended-IQ motif inserted into the cleft formed by the EF-like/EF-hand tandem, maintaining the side-by-side interaction in the presence of Ca²⁺ (Supplementary Fig. 11).

Discussion

Comparison of the Ca²⁺-sensing activation mechanism between PP2B and PPEF

Before our study, PP2B was the only phosphatase regulated by Ca²⁺ and CaM with a well-characterized structure^34,35. PP2B comprises a catalytic A chain (CnA) and a Ca²⁺-binding regulatory B chain (CnB)³⁶ (Fig. 6h). The catalytic domain of PP2B, similar to other PPP family members (PP1, PP2A, and RDGC), adopts a conserved structural conformation (Fig. 6i; Supplementary Fig. 12). Despite the significant sequence similarity between the catalytic domains of RDGC and PP2B, the activation mechanisms markedly diverge. Apo-CaM does not bind to CnB. Ca²⁺-CaM binds to the PP2B CnA subunit in response to increased intracellular Ca²⁺ concentration. Ca²⁺-CaM binding to CnA induces a conformational change in which the auto-inhibitory domain pulls out of the active site to facilitate transition to a fully active state (Fig. 6j).

Our investigation uncovered notable differences in the activation of RDGC and PP2B. In the inactive PP2B state, CaM does not bind to PP2B. In contrast, CaM still binds to inactive RDGC, priming the enzyme to flip down the auto-inhibitory subunits and swiftly respond to fluctuations in cellular Ca²⁺ levels. In this primed state, CaM associates with RDGC in Ca²⁺-free condition, ensuring the prompt response of the RDGC holoenzyme to Ca²⁺-influx in a “ready-to-go” model. This activation mode resembles CaM-mediated Ca²⁺ regulation in other proteins, including Ca²⁺ channels^37,38, unconventional myosins³⁹, and IQSECs⁴⁰.

The Ca²⁺-sensors in PP2B and RDGC are also different. In active PP2B, the Ca²⁺-sensor CnB disconnects from the catalytic domain. However, the Ca²⁺-loaded EF-hand domain connects to the catalytic domain in RDGC and works with the Ca²⁺-CaM to flip down the extended-IQ motif, although the functional consequences of this linkage require further investigation (Fig. 6j).

Significance of RDGC/CaM holoenzyme heterotetramer formation

Our cryo-EM analysis revealed that RDGC/Ca²⁺-CaM holoenzyme particles are heterotetramers at the concentration of 25 µM. This finding is particularly interesting in light of previous reports demonstrating that Ca²⁺-CaM can bind to PP2B with a 2:2 stoichiometric ratio in crystal structures⁴¹. Formation of the RDGC/CaM holoenzyme heterotetramers at 25 µM concentration are feasible under physiological conditions. In Drosophila melanogaster, the volume of a single microvillus is ~2 × 10^-18 L⁴². Each microvillus contains roughly 25 TRP molecules⁴³. Thus, the TRP concentration can be calculated as 25 µM. Given that RDGC expression levels are reported to be comparable to those of TRP (Protein Abundance Database https://pax-db.org/), suitable concentration for heterotetramer formation does exist within the microvillus.

In our SEC-MALS experiments, we observed that upon injecting 10 µM of the RDGC/Ca²⁺-CaM holoenzyme into the size exclusion column, the holoenzyme predominantly assumed a heterodimeric form. We postulate the existence of a dynamic equilibrium between heterodimers and heterotetramers in solution, a hypothesis corroborated by dynamic light scattering (DLS) results. In the absence of Ca²⁺, the average hydrodynamic radius of the RDGC/CaM complex remained invariant. While in the presence of Ca²⁺, the average hydrodynamic radius of the 20 µM RDGC/Ca²⁺-CaM complex was twice that of the 10 µM complex (Supplementary Fig. 13).

Furthermore, we conducted the mass photometry experiment using Two^MP mass photometer. We found that at concentrations of 25, 50, and 100 nM, the RDGC/CaM complex mainly existed as heterodimers with or without Ca²⁺. Only trace amounts of heterotetramers were detected at these concentrations in the presence of Ca²⁺ (Supplementary Fig. 14). Given that the current Two^MP mass photometer has a sample concentration limit of <100 nM (https://www.refeyn.com/twomp-mass-photometer), future investigations will necessitate the use of the MassFluidix High Concentration (HC) microfluidic system, which can accommodate sample concentrations up to 50 µM.

Another intriguing observation is the presence of a group of unidentified electron densities beneath the heterotetramer holoenzyme of RDGC/Ca²⁺-CaM. We attempted to add several components to the densities, but the distributions did not match. Further investigation is required to identify these electron densities.

Proposed working model for the PPEF family

In this study, a working model for the conserved PPEF family is proposed using the prototype Drosophila melanogaster RDGC. In the absence of Ca²⁺, CaM and the EF-like/EF-hand tandem bind to the extended-IQ motif and obstruct the catalytic sites, resulting in auto-inhibition of the catalytic sites. Ca²⁺ influx induces conformational changes in the two Ca²⁺-sensors (CaM and the EF-like/EF-hand tandem), relieving the auto-inhibition by bending the extended-IQ motif to expose of the catalytic sites. In addition, the formation of heterotetrameric holoenzyme may represent a pre-competent state before full activation. Subsequent substrate binding may induce monomer formation of the RDGC/CaM holoenzyme, leading to its full activation (Fig. 7).

Fig. 7. Proposed activation model of the RDGC/CaM holoenzyme.

The schematic representation illustrates the dual Ca²⁺-sensor-mediated regulatory mechanism in RDGC. In the absence of Ca²⁺, the extended-IQ motif occludes the catalytic sites, inhibiting phosphatase activity through interaction with the catalytic domain. This inhibition is further stabilized by CaM and the EF-like/EF-hand tandem. Upon Ca²⁺ influx, Ca²⁺ binds to CaM and the EF-like/EF-hand tandem, inducing a conformational change in the extended-IQ motif and exposure of the catalytic sites. In addition, the RDGC/CaM holoenzyme forms a heterotetramer composed of two protomers, with a stoichiometric ratio of 2:2, representing a pre-competent state. Substrate binding may induce monomer formation of the RDGC/CaM holoenzyme, leading to full activation.

Methods

Plasmids, protein expression, and purification

The coding sequences of Drosophila melanogaster RDGC (Accession Number: NP_788546.1, aa 1–661), Drosophila melanogaster CaM (Accession Number: NP_53710.1, aa 1–149), Mus musculus PPEF1 (Accession Number: NP_035277.1, aa 1–650), Mus musculus PPEF2 (Accession Number: NP_035278.1, aa 1–757), and Mus musculus CaM (Accession Number: NP_031615.1, aa 1–149) were amplified by PCR from cDNA libraries of Drosophila melanogaster and Mus musculus. Full-length RDGC was cloned into a modified pFAST vector with an N-terminal 10×His-2×Strep-TEV tag using the ClonExpress Ultra One Step Cloning Kit (Vazyme). Constructs for GST and GST-CaM were cloned into a modified pGEX 6P-1 vector. For heterologous expression in cells, the full-length RDGC and the mutants were cloned into a modified EGFP vector. Point mutations in RDGC and CaM were introduced via PCR-based mutagenesis and confirmed by DNA sequencing.

Spodoptera frugiperda (SF9) cells (Thermo Fisher Scientific) were used for expressing proteins. The cells were cultured at 28 °C to a density of 1.5–2.4 × 10⁶ cells/mL. SF9 cells were transfected with recombinant bacmids using Cellfection (Thermo Fisher Scientific). After transfection, the cells were incubated at 28 °C for 5 h in insect medium (Thermo Fisher Scientific) and then further cultivated. The P1 viral stock was used for large-scale expression. P2 and P3 viral stocks were also generated. Appropriate virus infection titters were determined by analyzing cell lysates using western blot.

The cultured cells were centrifuged at 129 × g and resuspended in lysis buffer (100 mM Tris-HCl pH 8.0, 150 mM Na₂SO₄, 1 mM EDTA, and 1 mM protease inhibitor Thermo Fisher Scientific). Cell lysates were generated using a high-pressure homogenizer (Union-Biotech) and subsequently centrifuged at 48,384 × g. Supernatants were added to a gravity flow column containing Strep-Tactin® XT resin (IBA) and incubated at 4 °C for 30 min. After washing the beads twice with 20 column volumes of lysis buffer, the beads were treated with elution buffer (50 mM biotin Sigma, 100 mM Tris-HCl, 150 mM Na₂SO₄, and 1 mM EDTA pH 8.1), followed by a 30-min incubation. The eluted protein was loaded onto a HiLoad^TM 16/600 Superose^TM 6 PG column (Cytiva) in a buffer containing 50 mM HEPES (pH 7.0), 150 mM Na₂SO₄, 1 mM EDTA, and 1 mM DTT for Ca²⁺-free state, and 50 mM HEPES (pH 7.0), 150 mM Na₂SO₄, 10 mM CaCl₂, 1 mM DTT for the Ca²⁺-bound state. The target protein peaks were identified using UV absorption at 280 nm and confirmed using SDS-PAGE gel analysis.

The RDGC (aa 1–25) mutant constructs, dCaM constructs, and mPPEF2 IQ (aa 22–39) and mCaM constructs were cloned into a modified pET32M.3C vector. The RDGC IQ (aa 1–25) was cloned into a V28E6 vector⁴⁴. The expression of target proteins was performed using Escherichia coli BL21 (DE3) host cells maintained at 16 °C. Purification of the His-tagged fusion proteins was initially achieved via Ni²⁺-NTA agarose (Cytiva) affinity chromatography, followed by separation using a HiLoad^TM 26/600 Superdex^TM 200 PG column (Cytiva) in a size-exclusion chromatography step. Subsequently, the N-terminal His-tag was cleaved using 3C protease. The resulting proteins underwent a final purification step involving a second round of size-exclusion chromatography in a HiLoad^TM 26/600 Superdex^TM 75 PG column (Cytiva). Plasmids and primers used in this study are provided in Supplementary Data 1 and Supplementary Data 2.

Size-exclusion chromatography coupled with multi-angle light scattering

Molar mass measurements of protein samples were performed using an AKTA FPLC system (Cytiva) equipped with a static light scattering detector (Dawn, Wyatt) and a differential refractive index detector (Optilab, Wyatt). Protein samples were prepared at a concentration of 10 μM and subsequently applied to a Superose^TM 200 Increase 10/300 GL column (Cytiva), which had been pre-equilibrated with a buffer which consisted of either 50 mM HEPES (pH 7.0), 150 mM Na₂SO₄, 1 mM EDTA, and 1 mM DTT or 50 mM HEPES (pH 7.0), 150 mM Na₂SO₄, 10 mM CaCl₂, and 1 mM DTT.

Protein crystallography and structure determination

Crystals of the mPPEF2 IQ (aa 22–39)/apo-CaM complex (6 mg/mL protein concentrated in 50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM EDTA, and 1 mM DTT) were produced by employing sitting-drop vapor diffusion techniques at a temperature of 16°C. Crystals of the mPPEF2 IQ (aa 22–39)/apo-CaM complex were grown in a 1.4 M sodium citrate tribasic dihydrate buffer (pH 6.1). RDGC IQ (aa 1–25)/Ca²⁺-CaM complex crystals (22 mg/mL protein concentrated in 50 mM Tris-HCl pH 7.5, 100 mM NaCl, 1 mM CaCl₂, and 2 mM β-mercaptoethanol) were produced by employing sitting-drop vapor diffusion techniques at a temperature of 16°C. The RDGC IQ (aa 1–25)/Ca²⁺-CaM complex crystals were grown in buffer containing 2% v/v tacsimate pH 4.0, 0.1 M sodium acetate trihydrate pH 4.6, and 16% w/v PEG 3350. Diffraction data were collected at 100 K at the Shanghai Synchrotron Radiation Facility BL17U and processed and scaled using HKL3000⁴⁵.

The mPPEF2 IQ (aa 22–39)/apo-CaM complex structure was determined through molecular replacement, using the apo-CaM model (PDB ID: 6MBA) in the PHASER software⁴⁶. The structure of the RDGC IQ (aa 1–25)/Ca²⁺-CaM complex was solved by molecular replacement, employing the MBP model (PDB ID: 2VGQ) alongside the separate N and C lobes of CaM (PDB ID: 6U3B). Manual model adjustments and iterative refinements were then performed using COOT⁴⁷ and PHENIX⁴⁸. The final model was validated with MolProbity⁴⁹, which indicated residues in the Ramachandran plot with 97.54%, 2.46%, and 0% in the preferred, allowed, and disallowed regions, respectively, for the RDGC IQ (aa 1–25)/Ca²⁺-CaM structure and 98.42%, 1.42%, and 0.16% in the preferred, allowed, and disallowed regions, respectively, for mPPEF2 IQ (aa 22–39)/apo-CaM. The final refinement statistics are summarized in Table 2. All structure figures were prepared using ChimeraX (https://www.cgl.ucsf.edu/chimerax). The coordination structures have been deposited in PDB under the access codes 8ZLX and 8ZLW for mPPEF2 IQ (aa 22–39)/apo-CaM and RDGC IQ (aa 1–25)/Ca²⁺-CaM respectively.

Isothermal titration calorimetry assay

ITC experiments were conducted at 25 °C using a MicroCal ITC200 calorimeter (Malvern). The protein concentration in the syringe is 500 μM, while the protein concentration in the cell is prepared at 50 μM. Sequential injections were delivered into the sample cell at intervals of 120 sec, beginning with an initial injection of 0.5 μL, followed by 19 subsequent injections of 2 μL each. Titration data were processed in Origin 7.0 software, employing a one-site binding model for curve fitting.

GST pull-down assays

In vitro binding of RDGC wild type/mutants with CaM was analyzed using GST pull-down assays with Glutathione Sepharose^TM 4B beads (Cytiva). GST-CaM or GST proteins were purified with GST affinity columns followed by size-exclusion chromatography using HiLoad^TM 26/600 Superdex^TM 200 PG columns (Cytiva) in a buffer containing 50 mM Tris-HCl (pH 7.5), 150 mM Na₂SO₄, 1 mM EDTA, and 1 mM DTT. The GST beads (20 μL) were pre-cleared in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM Na₂SO₄, 1 mM EDTA, 10% glycerol, 1% Triton X-100, and 1 mM protease inhibitor). Then, 500 μmoL of GST-CaM or GST alone was added to the pre-cleared GST beads and incubated for 30 min at 4 °C. The GST-CaM or GST-loaded beads were mixed with 500 μL of HEK293T cell lysates expressing GFP-tagged RDGC proteins in lysis buffer with 1 mM EDTA or 10 mM CaCl₂, respectively, and incubated at 4 °C for another 30 min. After two washes, the beads were boiled in 1× SDS-PAGE loading buffer to elute the bound proteins, and the supernatant was analyzed by western blot using an anti-GFP antibody (Cell Signaling Technology) at 1:3000 dilution.

Cell culture and transfection

HEK293T (CVCL_0063) cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM, Gibco), enriched with penicillin, streptomycin, and 10% fetal bovine serum (FBS, Gibco). Cultures were kept at 37 °C within a 5% CO₂ incubator. For transfection, ~30,000–50,000 cells were plated in each 10 cm dish and allowed to adhere overnight. After 24 h, transient transfection was performed using Lipofectamine 8000 (Beyotime Biotechnology), following the protocol provided by the manufacturer. The cells were collected 20–30 h post-transfection.

Phosphatase activity assay

The phosphatase activity of RDGC and its mutants was assessed using Invitrogen™ DiFMUP (Thermo Fisher Scientific) based on fluorescence produced through the hydrolysis of the DiFMUP substrate by RDGC. To prepare the assay, DiFMUP (5 mg) was dissolved in 1.71 mL of DMSO to prepare a 10 mM stock solution, stored at –20 °C. The freshly purified RDGC/CaM holoenzyme was transferred into the buffer containing 50 mM HEPES (pH 7.0) and 150 mM Na₂SO₄. The reaction mixture (total volume 100 μL) containing 10 μM substrate, 100 nM enzyme, and 1 mM Ca²⁺ or 1 mM EDTA was added to a 96-well plate and incubated for 60 min in the dark to avoid photodegradation. Fluorescence was measured using a multifunctional enzyme marker at excitation/emission wavelengths of 358/455 nm. We set the RDGC full-length/CaM holoenzyme activity in Ca²⁺ conditions to be 100%, with the relative activities of the mutants presented as percentages of the RDGC full-length/CaM holoenzyme activity. In Fig. 2g, a different approach was applied: to compare the relative phosphatase activities between RDGC full-length/CaM and RDGC (aa 79-412) under Ca²⁺-free conditions, we used the RDGC full-length/CaM holoenzyme activity measured in EDTA as the reference, setting it to 100%. Similarly, in Fig. 6e, the mPPEF1 (aa 10-650)/CaM holoenzyme activity in Ca²⁺ conditions was standardized to 100%, with all other groups represented as percentages of the mPPEF1 (aa 10-650)/CaM holoenzyme activity. All reactions were performed in triplicate, with statistical analysis conducted using GraphPad Prism. Data are presented as mean ± SEM.

Dynamic light scattering

The hydrodynamic radius (RH) of the RDGC/CaM complex, both in the presence and absence of Ca²⁺, was determined at concentrations of 10 µM and 20 µM using a Dynapro Nanostar (Wyatt Technology, USA) with a laser operating at a wavelength of 658 nm. All buffers and samples were filtered and centrifuged at 20,817 × g for 15 min prior to measurement. Triplicate measurements were taken for each concentration. The measurements were conducted at 25 °C, and the RH values were calculated using the integrated Dynamics software.

Mass photometry

Mass photometry assays were conducted using the Two^MP mass photometer (Refeyn, UK). Microscope coverslips and CultureWell gaskets were purchased from Refeyn. The gaskets were placed on cleaned coverslips on the sample stage of the mass photometer, following the manufacturer’s instructions. Prior to measurement, protein samples were freshly purified, centrifuged at 20,817 × g, and the concentration of the supernatant was measured. Both protein samples and standard proteins were diluted in PBS buffer (2.7 mM KCl, 140 mM NaCl, 1.8 mM KH₂PO₄, 10 mM Na₂HPO₄, pH 7.4), which was freshly filtered through a 0.22 µm filter prior to measurement. After dilution to concentrations of 25 nM, 50 nM, and 100 nM, 10 µL protein samples were placed in sealed wells on glass slides, and light scattering was recorded for 1 min. Data calibration was performed using bovine serum albumin (66 kDa, Sigma), beta-amylase (224 kDa, Sigma), and thyroglobulin (660 kDa, Sigma). Data were acquired at a sampling rate of 1 kHz for 100 sec using AcquireMP (Refeyn). Subsequent data analyses were performed using DiscoverMP (Refeyn), followed by Gaussian fitting to the standard protein.

Cryo-EM sample preparation and data collection

For the preparation of specimens suitable for cryo-EM and subsequent data acquisition, the RDGC/CaM proteins were first concentrated to 6 mg/mL (Ca²⁺-free condition) and 2.4 mg/mL (Ca²⁺-bound condition). The glow-discharged holey carbon grids (Quantifoil R1,2/1.3, Au, 300 mesh) were prepared under controlled conditions of 0.39 mBar air pressure, a current of 15 mA, and a duration of 50 sec. The grids with 4 μL sample were then subjected to vitrification using a FEI Vitrobot Mark IV device, operated at a temperature of 4 °C and under conditions of 100% relative humidity. The grids were blotted with Whatman filter paper for a duration of 4 sec under the same temperature and humidity conditions before being rapidly plunged into liquid ethane. These grids were subsequently stored under cryogenic conditions in liquid nitrogen.

For data collection, the grids were transferred to a 300 kV Titan Krios G3i electron microscope (Thermo Fisher Scientific), which is equipped with an EPU2 software package for automated data acquisition. The electron beam was conditioned to a GIF slit width of 20 eV, and the images were recorded using a Gatan K3 Summit direct electron detector in super-resolution mode. The specimens were imaged at a nominal magnification of 105,000×, which corresponded to a pixel size of 0.855 Å/pixel. The exposure time for each image was set to 1.83 sec with a dose rate of 20 e^–/pixel/s, accumulating to a total dose of 50 e^–/Ų. The exposure was fractionated into 32 individual frames to mitigate the effects of beam-induced shift and to enhance the signal-to-noise ratio in the final reconstructed images.

Data processing

All stacks were aligned with the Patch motion correction integrated in cryoSPARC (v3.3)⁵⁰, and the contrast transfer function (CTF) parameters were estimated using patch CTF estimation in cryoSPARC and micrographs were manually curated. A thousand micrographs were selected for automatic particle-picking using the Blob picker and subjected to 2D classification analysis. Particles of good classes were selected using Topaz training⁵¹. All micrographs were selected using Topaz extract and extracted with a box of 256 pixels (RDGC/apo-CaM) or 384 pixels (RDGC/Ca²⁺-CaM). Without 2D classifications, after 3 rounds of ab-initio reconstruction and heterogeneous refinement in cryoSPARC, particles of the best class were used to generate the final map by using non-uniform refinement with globe CTF refinement in cryoSPARC. For RDGC/Ca²⁺-CaM, one round of 2D classification was performed to eliminate junk particles, followed by an additional round of particle subtraction to further optimize the map density. The overall resolution of the final map was determined by the 0.143 criterion of the gold-standard Fourier shell correlation (FSC)⁵². The local resolution maps were evaluated by Local Resolution Estimation in cryoSPARC. The single-particle analysis procedures of the two states are summarized in Table 1.

Model building and refinement

An initial atomic model for RDGC/CaM was obtained from the AlphaFold Protein Structure Database⁵³. The model was manually fitted into the cryo-EM density using UCSF Chimera⁵⁴. Further model optimization and rebuilding were performed in Coot⁴⁷ with ligands docked. The models were then refined against the cryo-EM data using Phenix⁵⁵, applying geometric and secondary structure constraints. Structural visualization was carried out in ChimeraX⁵⁶. Refinement statistics are detailed in Table 1.

Reporting summary

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

Author contributions

J.L. Methodology, Validation, Investigation, Data analysis, Software, Writing-original draft, Visualization. C.W. Methodology, Validation, Writing-editing. Y.-Y.L. Methodology, Validation. Q.-G.C. Methodology, Validation. Y.-Z.D. Methodology, Resources. Z.-Q.L. Methodology. L.-F.P. Methodology, Resources, Funding acquisition. K.X. Writing-review and editing, Supervision. J.-C.L. Conceptualization, Supervision, Writing-review and editing, Funding acquisition. Z.-M.L. Conceptualization, Supervision, Writing-review and editing. W.L. Conceptualization, Supervision, Writing-review and editing, Funding acquisition.

Peer review

Peer review information

Nature Communications thanks Armin Huber and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The cryo-EM structures of RDGC/apo-CaM and RDGC/Ca²⁺-CaM and the crystal structures of mPPEF2 IQ (aa 22–39)/CaM and RDGC IQ (aa 1-25)/CaM have been deposited in the Protein Data Bank under the codes 8JFW, 8JFY, 8ZLX, and 8ZLW, respectively. The cryo-EM density maps of the RDGC/apo-CaM and RDGC/Ca²⁺-CaM have been deposited in the Electron Microscopy Data Bank under the codes EMDB-36218 [https://www.ebi.ac.uk/emdb/EMD-36218] and EMDB-36219 [https://www.ebi.ac.uk/emdb/EMD-36219], respectively. Publicly available PDB entries used in this study included 6MBA, 1AUI, 2VGQ, 6U3B, 8DWL, 8SO0, AlphaFold AF-O14829 [https://alphafold.com/entry/O14829], AF-O14830 [https://alphafold.com/entry/O14830], and AF-P40421 [https://alphafold.com/entry/P40421]. Sequences used in the alignment for Supplementary Fig. 10 included H. sapiens PPEF1, H. sapiens PPEF2, M. musculus PPEF1, M. musculus PPEF2, R. norvegicus PPEF1, C. elegans PPEF, and D. melanogaster RDGC (Uniprot accession codes: O14829, O14830, O35655, O35385, Q3SWT6, G5EBX9, and P40421, respectively). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-58261-z.