ATP‐Driven Allosteric Regulation of 14‐3‐3: Positive Modulation of ATP Hydrolysis and Negative Regulation of Peptide Binding

ABSTRACT

Many 14‐3‐3 paralogs, except sigma, could bind and hydrolyze ATP. However, the catalytic residues and the significance of ATP binding or hydrolysis remain unknown. Here we confirm that there are two binding pockets for ATP, one at the peptide binding amphipathic pocket and the other at the dimer interface. As predicted by a new computational method, CLICK, and by limited proteolysis coupled to mass spectroscopy, we identify E131 and E180 as the catalytic residues. We further confirm that ATP hydrolysis is an inherent property of 14‐3‐3, and mutations result in either gain or loss of ATPase activity. The dimeric fold of the protein is mandatory for ATP hydrolysis but not for peptide binding. While ATP at the dimer interface acts as an allosteric activator of ATP hydrolysis, it acts as a selective negative regulator of a nonphosphopeptide, originating from ExoS, a pathogenic Pseudomonas protein. This study for the first time, unveils the hidden allosteric properties of the 14‐3‐3 proteins and its role in excluding specific ligands of disease relevance.

This graphical abstract illustrates dual allosteric modulation by ATP. ATP binding at allosteric site activates the enzyme and hydrolyzes the ATP at catalytic site to ADP and iP. On contrary, ATP binding at allosteric site exhibits the inhibition to the natural ligand peptide of 14‐3‐3.

1. Introduction

Responding to environmental cues precisely and effectively is a prime requisite for a cell to survive and thrive. The scaffolding proteins play a crucial role in various cellular processes by acting as regulators of response/modulators to intrinsic and extrinsic environmental stimuli. These scaffolds provide an appropriate platform for the “action heroes” of the signaling cascades to assemble, enable their site‐specific localization, and coordinate the positive and negative feedback responses so that functions are executed in a precise time‐dependent manner at the correct location. The abundant seven‐member 14‐3‐3 family of proteins of the human (β, ε, η, γ, σ ζ, τ) acts as one such scaffold by binding to a plethora of phosphoproteins and regulates cell cycle progression, centrosome duplication, protein trafficking, and stress response [1, 2]. These proteins mainly function as homo or heterodimers and, on occasion (still controversial), as monomeric proteins [3].

Many crystal structures of 14‐3‐3 homodimers are available, and by and large, the structures are well conserved. Each monomer of 14‐3‐3 consists of nine alpha‐helices arranged in an antiparallel manner (Figure 1A). Helices 1, 2, and 4 form the dimer interface. The salt bridges (R18‐E89, D21‐K85, E22‐K85, and E5‐K74) [4] in this region hold 14‐3‐3 monomers together (all numbering of amino acids is according to ζ paralog) (Figure 1B). Phosphorylation at S58 leads to the disruption of the dimer, hence generating monomeric protein [5].

FIGURE 1.

Structure of 14‐3‐3ζ. (A) Mode 2 phosphopeptide, bound to 14‐3‐3 front (left panel) and top view (right panel) (adapted from PDB ID: 1QJA) (B) Zoomed in view of dimer interface depicting crucial residues responsible for dimerization. (C) Zoomed in view of the amphipathic pocket displaying critical residues that interact with the phosphopeptide.

The main functional unit of the 14‐3‐3 protein is the cup‐shaped, highly conserved amphipathic pocket composed of the 3rd, 5th, 7th, and 9th alpha‐helices present in each monomer [6]. A significant portion of ligand proteins that bind to 14‐3‐3 carry a serine or threonine residue phosphorylation. Phage display and the analysis of the natural ligands that bind to 14‐3‐3 proteins have led to the identification of four binding motifs: (a) RSXpS/TXP or mode 1, (b) RXXXpS/TXP or mode 2, (c) pS/RTX‐COOH or mode 3, and (d) SXpS/pTXPXXφ or mode 4 (φ, hydrophobic and X, any residue) [7, 8, 9]. Residues K49, R56, R127, and Y128 are the major interacting residues that bind to the negatively charged phosphorylated serine or threonine residues of the ligand. N224, N173, and K120 form hydrogen bonds with the backbone residues of the phosphopeptide sequence while E180, V176, Y179, and W228 interact with the peptide side chains [4] (Figure 1C). Besides these motifs that belong to phosphoproteins, 14‐3‐3 also binds to unphosphorylated proteins such as p190RhoGEF [10], exoenzyme S [11], and R18, a phage display‐derived peptide [12].

Although the majority of the literature on 14‐3‐3 proteins is on the characterization of the phosphoprotein binding and regulation of the functions of the interacting partner, a handful of reports have shown that 14‐3‐3 can bind to other small molecule ligands such as ATP, ADP, and AMP [13, 14]. Very early literature in 1997 demonstrated that 14‐3‐3τ has an ATP/ADP exchange activity [13]. In the presence of ATP, Drosophila 14‐3‐3ζ and Hsp70 together solubilize aggregated proteins [14]. Recent reports show that AMP stabilizes 14‐3‐3‐ChREBP peptide‐complex in the cytoplasm preventing nuclear localization of ChREBP. Neither AMP nor the nonphospho ChREBP peptide alone can bind strongly to the 14‐3‐3. However, both ligands were found to mutually enhance interaction and the trimodular complex and formed was the most stable complex [15].

The strongest evidence for the role of 14‐3‐3 in direct nucleotide binding and associated functions has come from Ramteke et al. [16]. This study has shown that many members of the 14‐3‐3 family proteins are ATPases except the sigma paralog. They identified two mutations that enhanced or reduced the ATPase activity. The D124A mutation within the amphipathic pocket led to an increase in ATPase activity, and the R55A mutation at the dimer interface led to a decrease in the ATPase activity. These interesting results not only highlighted the enzymatic function of 14‐3‐3 but also added a new dimension to the structure conservation/or not, and activity [16]. Most of the ATPases have Walker domain A/P‐loop, Walker B domains, a beta‐strand, and a glycine‐rich sequence; however, surprisingly, 14‐3‐3 has none of these. The Walker domain A(G‐x(4)‐GK‐[TS]) binds to ATP/GTP. P loop‐like ATPases carry the conserved sequence C‐x(5)‐R‐[ST]. The A‐loop upstream of the Walker A motif has conserved aromatic amino acids essential for ATP‐binding. Walker B is a motif downstream of the A‐motif with the consensus sequence [RK]‐x(3)‐G‐x(3)‐LφφφD (φ, hydrophobic and X, any residue) [17, 18]. The Rossman fold, made up of alternating beta‐strand and alpha‐helical segments that are hydrogen bonded to create a beta‐sheet, is known to bind to ATP [19]. 14‐3‐3 lacks any similarity with these canonical ATPases, and hence, the amino acid residues driving ATPase activity and the possible role of the ATPase activity of 14‐3‐3 remain unknown.

In this study, we have deciphered the precise ATP binding pocket/s in14‐33‐3ζ. We have identified the residues involved in catalysis and established the influence of ATP binding on peptide interaction and the explicit requirement of a dimeric structure for ATP hydrolysis. We also provide evidence for the dual allosteric role of ATP—(a) as a positive modulator of ATPase activity and (b) as a negative regulator of a subset of nonphosphorylated ligands of a bacterial pathogenic protein.

2. Materials and Methods

2.1. Prediction of ATP Binding Site for Query Protein Using Descriptor Library

To identify the binding site of ATP on 14‐3‐3ζ, we applied an approach that relies on the similarity of the geometrical and chemical characteristics of ATP binding sites. The CLICK [20, 21] and DEPTH [22, 23] programs were utilized for this analysis. CLICK is a 3D structure comparison program that superimposes 3D biomolecular structures regardless of topology. Each protein in the query set was first prepared for the search process by removing all HETATM records and any small molecules or ions present in the structure. The nonprotein components could mask the ligand binding site due to steric clashes with the descriptors from our library, thus interfering with accurate binding site prediction.

As we were testing for the ATPase activity of 14‐3‐3ζ, we checked for hydrogen bonds between ATPΔp (ATP excluding the phosphate tail) and protein in the descriptor library. The phosphate tail was initially excluded from consideration as ATP specificity would be conferred by interaction with the base and not the phosphate tail. Earlier studies have shown that base and ribose specific interactions are not only specificity determining factors but also conserved in proteins [24, 25].

The coordinates of hydrogen bond acceptor (ACC) and donor (DON) pairs and the ATPΔp coordinates were extracted for each identified hydrogen bond. These coordinates were then used to create a molecular descriptor library that captures the geometric and environmental features of ATP binding sites. The extracted coordinates from each structure were saved in separate PDB files (Table S1), resulting in a total of 155 descriptor files in the library.

To ensure that only the most promising binding site candidates were retained, configurations that failed to form at least three hydrogen bonds or were solvent‐exposed without being situated within a cavity (as determined by the DEPTH software) were excluded from consideration. These filtering criteria were applied heuristically, based on empirical rules developed from prior experience with descriptor‐based docking. The goal was to prioritize configurations with favorable geometric and energetic features likely to support ATP binding. The remaining configurations were rank‐ordered based on their clash scores. These scores quantify the steric clashes between the descriptor and the surface. The high‐ranking configurations with a nucleophile (e.g., a residue with a negative charge) situated close to the phosphate tail of ATP were chosen as the predicted ATP binding site.

2.2. Strains and Plasmids

14‐3‐3ζ WT and mutant genes were cloned and expressed using E. coli DH5α and BL21 strains, respectively. 14‐3‐3ζ R55A, Y82A, D124A, E131A, N173V, E180A, Y82AN173V, F117AN173V, E131AN173V, E180AN173V, and E131AE180A mutations were obtained using overlap extension using the polymerase chain reaction (PCR). The list of primers used is provided in (Table S2). The presence of each mutation was confirmed by Sanger sequencing.

2.3. Protein Expression and Purification

14‐3‐3ζ WT and mutant apoproteins were expressed in E. coli BL21 DE3 cells. The hexa‐histidine tagged proteins were purified using Ni‐NTA affinity chromatography followed by size exclusion chromatography. The histidine tag was removed using TEV protease. The untagged protein was subjected to gel filtration on a Hiload 16/600 Superdex 200 column (GE Healthcare, Chicago, IL, USA), and peak fractions were collected. The buffer used for purification is 50 mM Tris/Cl pH 8.0, 400 mM NaCl, and 10% glycerol [26]. For all experiments, the protein was buffer exchanged in 50 mM Na‐HEPES buffer pH 8.0, 150 mM NaCl, and 10% glycerol.

2.4. Limited Proteolysis of 14‐3‐3

All enzymes and reagents used in this study were LC–MS grade, ensuring high purity and compatibility with mass spectrometry applications. The basic protocol of LiP‐MS was adapted from Schopper et al. [27]. Briefly, 35 μg of purified 14‐3‐3ζ WT apoprotein was incubated with ATPγS at a final concentration of 500 μM and a temperature of 25°C for 30 min. In the case of control samples, an equivalent amount of solvent (MQ) was added instead of ATPγS. Following the binding step, limited proteolysis was performed using Proteinase K (PK) at an enzyme: substrate ratio of 1:100 for 1 min at 25°C in a thermocycler machine. After the LiP step, the PK was immediately heat‐inactivated by incubating at 99°C for 5 min. The samples were cooled at 4°C in ice for five minutes and processed by mass spectrometry analysis as described below.

2.5. Sample Processing for LC–MS Analysis

Sodium deoxycholate (DOC) was added to the samples from the previous step to a final concentration of 5%. The disulfide bonds were reduced by adding tris(2‐carboxyethyl) phosphine (TCEP) to a final concentration of 5 mM, followed by incubation at 55°C for one hour [28]. The samples were alkylated with iodoacetamide (IAA) at a final concentration of 40 mM in the dark at 30°C for 30 min. Following alkylation, the samples were diluted with 100 mM triethylammonium bicarbonate (TEAB) buffer to a DOC concentration of 1%. The samples were then digested with trypsin at an enzyme: substrate ratio of 1:50 and incubated at 37°C overnight. The digestion was stopped by adding trifluoracetic acid (TFA) to a final concentration of 2%. The precipitated DOC was removed by centrifugation at 15,000 g for 10 min. The supernatant was carefully transferred to another tube and evaporated using a refrigerated CentriVap centrifugal concentrator (Labconco, Kansas City, USA). The dried samples were reconstituted in 300 μL of 0.1% TFA and desalted using PierceTM C18 peptide desalting spin column (Thermo Fisher Scientific, Waltham, MA, USA). The eluted samples were dried using a refrigerated CentriVap centrifugal concentrator and stored at −80°C until further analysis.

2.6. LC–MS/MS Analysis

The desalted samples were reconstituted in 0.1% formic acid and analyzed on a Triple‐TOF instrument (SCIEX, Framingham, MA) coupled to nanoLC (Eksigent ekspertTM). 150 ng of protein was loaded and first trapped by a 200 μm × 5 cm C18 trap column (Eksigent, Dublin, CA, USA), followed by separation on a C18 column (75 μm × 15 cm, Eksigent). The injection volume was 6 μL. A gradient elution method using buffer A (0.1% formic acid in water) and buffer B (0.1% formic acid in acetonitrile) was used as follows: 4.8% B (0 min), 10% B (12 min), 30% B (92 min), 50% B (112 min), 80% B (113 min), 80% B (126 min), 4.8% B (127 min), and 4.8% B (146 min). The flow rate was 300 nL min⁻¹. The data was acquired in electrospray positive ionization mode using the information‐dependent acquisition (IDA) method, where the top 30 precursor ions were fragmented via rolling collision energy. The m/z scan range was 350–1250 and 200–1800 for MS1 and MS2 scans, respectively. The accumulation time was 250 ms and 50 ms for MS1 and MS2 scans, respectively. Cycle time was 1.7 s.

2.7. Analysis and Representation of LiP‐MS Data

The raw data file in WIFF format was converted to mzML format using MSConvert [29]. Peptide identification was performed using the FragPipe proteomics workflow that utilizes the MSFragger search engine [30]. Modifications specific to LiP‐MS experiments were applied to FragPipe's default parameters [31]. A peptide spectrum match was initially performed using the human FASTA database obtained from UniProt. The 14‐3‐3ζ was confidently identified at a 1% false discovery rate. Label‐free quantification of peptides was carried out using the IonQuant algorithm integrated into the FragPipe pipeline. Interestingly, one peptide of 14‐3‐3ζ (YDDMAACMK, 19–27) was misidentified as a peptide from 14‐3‐3 beta/alpha (YDDMAAAMK, 21–29). Since the purified protein showed no such mutation of cysteine to alanine during sequencing, we investigated the root cause. Desulfurization of cysteine to alanine can occur in the presence of TCEP and heat [32]. To check this possibility, we modified our search parameters to include desulfurization as a variable modification at cysteine residues. Using the 14‐3‐3ζ FASTA from UniProt, we confirmed that desulfurization was indeed occurring. Despite this adjustment, the statistically significant peptides (adjusted p‐value ≤ 0.05 and fold change ≥ 2) remained consistent, irrespective of whether the complete human FASTA or 14‐3‐3ζ FASTA was used, as the Benjamini‐Hochberg (BH) correction was applied independently for each protein [31].

The output data from FragPipe was further utilized for the automated identification and statistical analysis of semi‐tryptic and full‐tryptic peptides using FLiPPR [31]. FLiPPR analysis resulted in the identification of conformotypic peptides responsible for ATPγS binding. Peptide abundance data was binned for histogram analysis using the pandas and numpy modules in Python (version 3.11.7). Visualizations, including histograms and volcano plots, were prepared using the seaborn and matplotlib modules in Python.

2.8. Native PAGE

The oligomeric status of 14‐3‐3ζ WT and mutant apoproteins was determined by running the proteins on a Native PAGE gel. 15 μg of proteins was resolved on 10% Native gel and stained with Coomassie blue.

2.9. Determination of Protein Stability

The thermal stability of 14‐3‐3ζ WT and mutant apoproteins was determined using nano‐differential scanning fluorimetry (DSF) (PrometheusNT.48) from Nanotemper Technologies (Munchen, Germany). The 350 nm/330 nm intensity ratio is plotted as a function of temperature. The first derivative was used to determine the melting temperature (T_m) of proteins. 15 μg of proteins in a 50 mM Na‐HEPES buffer (pH 8.0) containing 150 mM NaCl and 10% glycerol was used for the experiment.

2.10. CD Spectroscopy

The far UV CD spectra of 14‐3‐3ζ WT and mutant apoproteins (0.1 mg/mL in PBS) were recorded using a JASCO J‐815CD spectrometer. Data were collected from wavelength 260‐190 nm at 0.5 nm intervals. The spectra were blank‐subtracted (buffer alone). The molar residual ellipticity was calculated using the formula °*M*100/(L*/C) [33] three independent scans. The spectra were analyzed online by BeStSel, and secondary structures were determined [34].

2.11. ATPase Assay

To check the ATPase activity of 14‐3‐3ζ WT and mutant proteins, we used the ADP‐GloTM max ATPase assay kit (Promega) and followed the manufacturer's instructions. 10 μM 14‐3‐3ζ WT and mutant proteins in 50 mM Na‐HEPES buffer pH 8.0, containing 150 mM NaCl, 10 mM MgCl₂ and 1 mM DTT were incubated with 1 mM ATP at 37°C for 2 h. The luminescence was recorded using Cytation 5, Biotech (Winooski, VT, USA). To estimate the amount of ADP produced (release of inorganic phosphate Pi) in the ATPase reaction, the standard curve was obtained that represents the luminescence corresponding to the conversion of ATP to ADP. The specific activity was calculated using the formula “Conc. of Pi released/time*conc. of protein,” and data were analyzed using GraphPad PRISM 6. The kinetic parameters were calculated using the Michaelis–Menten nonlinear regression fit and kcat, km and Vmax of 14‐3‐3ζ WT and mutant proteins.

2.12. Fluorescence Polarization

The binding of 14‐3‐3ζ WT and mutant proteins to the FAM‐labeled peptides in solution was measured using fluorescence polarization (FP). The N‐terminal FAM‐labeled peptides were commercially synthesized by GL Biochem (Minhang, China). The list of the peptides used is provided in Table S3. The peptide concentration was kept constant at 62.5 nM, and the protein concentration was titrated from 8 to 0.0625 μM concentration. In experiments designed to test the effect of ATP, 1 mM ATP or ATPγS was used in the binding buffer. Prior to measurements, the reaction was incubated at 25°C in the dark for 30 min. The change in anisotropy was measured using Cytation 5, Biotek (Winooski, VT, USA). The K_D values were determined using GraphPad PRISM 6 (GraphPad Software Inc., San Diego, CA, USA).

2.13. Microscale Thermophoresis (MST)

To monitor the effect of ATP on the binding of FAM‐labeled peptides with 14‐3‐3ζ WT, we performed microscale thermophoresis (MST) using the Monolith NT.115 instrument (Nanotemper Technologies, Munchen, Germany). The experiment was standardized for fluorescent counts, and the power to be used was determined per the manufacturer's instructions. 32 μM of 14‐3‐3ζ WT and mutant proteins were serially diluted 2‐fold using a 10 μM FAM‐labeled peptide in the same buffer used for the FP experiments. Different concentrations of ATP were used to check its effect on peptide binding. The samples were loaded onto standard MST glass capillaries, and an experiment was performed. The data were analyzed using the inbuilt MST software and GraphPad PRISM 6.

2.14. Surface Plasmon Resonance (SPR)

To assess the effect of ATP on the kinetics of the interaction between 14‐3‐3 and peptides, Surface Plasmon Resonance (SPR; BiacoreT200; GE Healthcare) was used. 14‐3‐3ζ WT and mutant proteins were covalently immobilized on a CM5 chip by amine coupling reaction at a final concentration of 4000 RUs. After immobilization, 10 μM to 1 mM ATPγS with 10 μM peptide was passed on the chip in HBS‐P running buffer (10 mM Na‐HEPES pH 7.4, 150 mM NaCl, and 0.005% v/v Tween‐20). The change in RU upon binding and dissociation of ATPγS and peptides was used to assess kinetic parameters using the BiaEvaluation software (Biacore, GE Healthcare). Kinetic parameters were established by fitting sensorgrams to a 1:1 binding model. The data is plotted using GraphPad PRISM 6.

3. Results

3.1. Prediction of the ATP Binding Pockets, Identification of Binding and Active Site Residues Using In Silico Methods and LiP‐MS

Our previous report established that 14‐3‐3 is an ATPase and mutations in the protein resulted in a gain or loss of activity [16]. Two putative binding sites were proposed, one at the dimer interface and the other at the amphipathic pocket. However, it was not clear which of the binding sites acted as a catalytic site and what may be the significance of ATP binding to these different sites. To identify the catalytic residues, we utilized an inhouse In silico method (descriptor library and the CLICK program, see Methods and Table S1) and an experimental technique called the LiP‐MS.

3.1.1. In Silico Method

The structure of 14‐3‐3ζ was searched against the molecular descriptor library using CLICK. Search hits on the protein surface that matched the geometry of known ATP binding pockets were considered (Figure 2A). These candidate sites were further filtered using the DEPTH software, which predicts the likelihood of small molecule binding sites on proteins based on residue depth and solvent accessible surface area [22] (Figure 2B,C). The final predicted pose (Figure 2D) of the ATP was selected based on the presence of negatively charged residues near the end of the phosphate tail, which are necessary for facilitating a nucleophilic attack. We examined the dynamics of the protein and the protein‐ligand complex using Molecular Dynamic (MD) simulations. A triplicate of 500 ns simulations was carried out for the predicted ATP‐bound and unbound structures of 14‐3‐3ζ. The Root Mean Square Fluctuations (RMSF) of ATP bound and unbound states were compared for different principal components (Figure S1). Intra‐protein and protein‐ligand hydrogen bonds were also monitored.

FIGURE 2.

Prediction of ATP binding site for 14‐3‐3ζ. (A) Surface representation of the 14‐3‐3ζ protein, highlighting the overall structure with side (left) and top down (right) views. All predicted ATP binding site positions are displayed with the ATP in stick representation without phosphate tails. (B, C) Top scoring model of the predicted ATP‐bound 14‐3‐3ζ monomer shown in surface representation rendered by electrostatic color gradient from blue (positive) to red (negative) and DEPTH color gradient from blue (exposed) to red (buried). (D) A zoomed in view of the predicted ATP binding site of 14‐3‐3ζ. The ATP is shown in stick representation and encased in its surface in mesh representation. The residues of 14‐3‐3ζ predicted to be interacting with the ATP are shown in stick representation.

To investigate the local and global dynamics of the 14‐3‐3ζ protein in its ATP‐bound and unbound forms, we performed two triplicates of 500 ns molecular dynamics simulations of the protein. The essential dynamics, as indicated by Principal Component 1 (PC1) and Principal Component 3 (PC3) (Movies S1), suggest no difference in global motions between the ATP‐bound and ATP‐free forms of 14‐3‐3ζ. However, significant local conformational changes appear in Principal Component 2 (PC2). In the ATP‐bound form, local twisting/compressing motions bring residues E131 and E180 closer (3.85 Å and 3.91 A°) to the ATP phosphate tail, thereby allowing the possibility of a water‐mediated nucleophilic attack for ATP hydrolysis. The method and the knowledge that ATP is hydrolyzed by 14‐3‐3 enabled us to identify the key residue important for orienting the ATP and those important for hydrolysis.

3.1.2. LiP‐MS to Identify the Binding Pockets

The in silico method allowed us to establish that the amphipathic pocket is one of the two ATP binding sites in 14‐3‐3. Most importantly, it not only allowed us to narrow down on the catalytic residues involved, but also established the consensus mode of binding of the adenosine in ATP to phenylalanine (F117) in 14‐3‐3 as seen with other ATP binding proteins [35]. However, the possibility of the second binding site for ATP and the reduction in ATP binding and hydrolysis of the R55A dimer interface mutant [16] could not be established by the computational approach. Therefore, we resorted to LiP‐MS, which in brief, is an advanced structural proteomics technique used to investigate protein dynamics and structural changes induced by small molecule binding or other perturbations [27, 36]. This powerful method provides insights into secondary structure alterations, allosteric effects, protein refolding, aggregation, and protein‐ligand interactions.

We employed the LiP‐MS strategy to map the ATP‐binding regions of the 14‐3‐3ζ protein and create the proteolytic fingerprint of semi and full tryptic peptides. It is anticipated that ATP binding would confer structural protection, reducing the susceptibility of certain regions to proteolytic cleavage by Proteinase K. The SDS PAGE gel (Figure S2A) shows that ATPγS protects the protein against proteolytic cleavage. To pinpoint the regions involved in ATPγS binding, we analyzed the semi‐tryptic peptides and a shift in Proteinase K cleavage sites in the ATPγS bound form of the protein by mass spectrometry. Our analysis revealed a significant reduction in the abundance of semi‐tryptic peptides in the ATPγS‐bound form of the 14‐3‐3ζ protein compared to its unbound form (Figure 3A). As expected, a notable decline in PK cut sites was seen in the ATPγS‐bound state versus the unbound state (Figure 3B). This signifies that ATPγS binding protects the protein rendering 14‐3‐3 more stable to PK in the ATPγS‐bound form.

FIGURE 3.

Analysis of LiP‐MS data of 14‐3‐3ζ. (A) Histogram of peptide abundance with overlaid Kernel Density Estimation (KDE) curve showing that semi‐tryptic peptides are less abundant in ATPγS‐bound protein samples compared to unbound form. The X‐axis denotes the log2 fold change of ATPγS bound versus unbound protein samples, and the Y‐axis denotes the number of peptides corresponding to fold change. (B) Histogram of cut site abundance with overlaid Kernel Density Estimation (KDE) curve showing that the cut sites of Proteinase K are less abundant in ATPγS ‐bound protein samples compared to unbound form. The X‐axis denotes the log2 fold change in cut sites, and the Y‐axis denotes the number of cut sites. (C, D) Volcano plot representing fold change in the peptides that are abundant or cut sites of ATPγS bound versus unbound protein samples. The peptides and cut sites showing significant change (≥ 2) and adjusted p‐value (≤ 0.05) are highlighted in red color. (E) 3D rendered representation of 14‐3‐3ζ (Full, Dimer interface, near amphipathic pocket). Khaki and gray represent two monomers while, faint blue represents the protected peptides and navy blue represents the residues selected for the mutational study.

We further examined the data to identify specific ATPγS‐binding regions within the protein. Figure 3C presents the volcano plot of peptide fold changes, showing statistically significant peptides with altered abundances. It is noteworthy that the abundance of semi‐tryptic peptides decreased while full tryptic peptides increased in the ATPγS‐bound condition. These structure‐specific or conformotypic peptides signify the ATPγS‐binding regions and are listed in Table 1. The significantly altered cut sites in the ATPγS‐bound form align well with the identified peptide sequences (Figure 3D). The MS/MS spectra of conformotypic peptides in presence and absence of ATPγS is given in Figure S2B and the protected regions of 14‐3‐3ζ WT are shown in Figure 3E. It can be seen that E180, one of the nucleophiles predicted from in silico approach resides within 175–187 fragment protected from proteolysis. In addition, 81–91 residues which are part of the dimer interface, were exclusively protected by ATPγS (Table 1). Thus, LiP‐MS not only confirmed the amphipathic pocket as the ATP binding site but clearly demonstrated that there is a second ATP binding site at the dimer interface.

TABLE 1.

The list of significantly altered conformotypic peptides of 14‐3‐3ζ WT upon ATPγS binding. The peptides showing a fold change of ≥ 2 and p‐value ≤ 0.05 from FLiPPR analysis are provided.

Peptide sequence	Start	End	Peptide length	Cleavage type	Adjusted p	Log2 fold change
YDDMAAC	19	25	7	C_SEMI	2.49E‐02	−1.0
YDDMAACMK	19	27	9	FULL_TRP	6.47E‐03	1.8
EYREKIETELR	81	91	11	FULL_TRP	3.14E‐02	1.3
SVFYYEILNSPEK	175	187	13	N_SEMI	1.90E‐02	−1.6
SYKDSTLIMQLLR	210	222	13	N_SEMI	2.14E‐02	−1.3

3.2. Site Directed Mutagenesis of in Silico and LiP‐MS Predicted Residues, Stability of Mutant Proteins and Effect on Peptide Binding

The binding of the ATP to the amphipathic pocket was suggested by the CLICK algorithm. Furthermore, analysis of the ATP docked 14‐3‐3ζ structure indicated that either E131 or E180 or both within the amphipathic pocket could be ATP hydrolyzing residues. The data obtained from the LiP‐MS suggested that ATPγS can bind both to the amphipathic pocket and dimer interface. To validate these predictions, amino acid residues residing in the dimer interface and amphipathic pocket were mutated to alanine or valine. R55A, Y82A, D124A, E131A, N173V, E180A, F117AN173V, E131AN173V, E180AN173V, Y82AN173V mutant proteins were generated, expressed, and purified (Figure 4A). To understand the effect of mutation on the structure and conformation of the proteins, native PAGE, CD and thermal denaturation experiments were performed. 14‐3‐3ζ WT, R55A, D124A, E131A, N173V, E180A, F116AN173V, E131AN173V, and E180AN173V migrated as dimers on a native PAGE gel. Migration of Y82A and Y82AN173V were faster as compared to the other mutant proteins, indicating that they could be monomeric in nature (Figure 4B lane 2 and lane 3). The gel filtration of Y82A mutant protein confirmed the same. The elution volume for 14‐3‐3ζ WT was 68 mL and that of Y82A was 74 mL (Figure 4C). The T_m of ζWT, D124A, and E131A was similar to a previously published report [26]. The CD spectra of WT and all mutant proteins were dominated by double‐minima at 220 and 208 nm and a positive ellipticity at 195 nm, features characteristic of α‐helix. Although double mutants had reduced ellipticity, the overall secondary structure was not altered (Figure 4D; and Table S4). As monitored by nano‐DSF all mutant proteins were generally less stable than the WT protein, with all four double mutants being more unstable than their corresponding single mutants (Figure 4A). Representative thermal stability profile obtained using nano‐DSF for WT is shown in, Figure S3 and T_m values are summarized in Table S5. In particular, the E180AN173V double mutant protein was most unstable and prone to aggregation; therefore, we did not use this mutant protein for further analysis.

FIGURE 4.

Purification and biochemical characterization of WT and mutant 14‐3‐3ζ apoprotein. (A) Purified proteins were resolved on 12% SDS PAGE. (B) Purified proteins were analyzed on a nondenaturing 12% gel. (C) Gel filtration profile of WT, Y82A and Y82AN173V proteins on Hiload 16/600 Superdex 200. (D) Secondary structure of 14‐3‐3ζ WT and the mutant proteins were determined using far UV CD. (E) Thermal stability of the WT and the mutants were monitored using nano‐DSF. The first derivative of tryptophan fluorescence was used to determine the T_m. Data are represented as mean ± S.D. (n = 3). S.D.– standard deviation.

To test whether such instability in structure prevents the main function of the protein, we tested the peptide binding ability of all the mutant proteins using FAM‐labeled peptides. All mutant proteins bound to the peptides (except the D124 and N173 derivatives) with affinity remarkably close to that of the WT protein (Figure 5A,B). N173 which a forms hydrogen bond with D124 and with the backbone of the peptide [37], and the N173V mutant expected to behave similar to D124A [38], was indeed defective in peptide binding assays (Figure 5A). The N173V mutant was however more stable than the D124A. Taken together, these results indicate that the structure of the 14‐3‐3ζ protein is extremely plastic and is capable of regaining the lost conformation upon binding to a ligand.

FIGURE 5.

(A) Binding of mode2 phosphopeptide to 14‐3‐3 ζWT and amphipathic pocket mutant proteins by Fluorescence polarization. Data are represented as mean ± SEM. SEM – standard error of mean. (B) Binding of mode2 phosphopeptide to 14‐3‐3 ζWT and dimer interface mutant proteins were checked by FP. (C) ATPase activity of ζWT and amphipathic pocket mutant proteins were performed using ADPGloMax promega kit. Data represents the nM of inorganic phosphate (Pi) released per minute per micromolar protein (normalized with buffer) of three independent experiments with ± SEM.

3.3. Experimental Validation of Predicted Residues in ATP Hydrolysis and Evaluation of the Second Binding Site

Having established that the mutant proteins were stable or regained their conformation upon peptide binding, we proceeded to check their ATPase activity. The ATPase activity of the D124A mutant was higher than that of the WT protein; K_cat, V_max and K_m values were consistent with previously published studies [16]. Similar to D124A, N173V had increased ATPase activity and has more affinity for ATP (Figure 5C, Table 2). Hence, N173V shows similar trends like D124A, that is, it is better than WT in hydrolyzing ATP and it binds less well to the phosphopeptide ligands.

TABLE 2.

The V _max, K _m and k _cat for 14‐3‐3ζ WT and mutant proteins.

	WT	D124A	N173V	E131A	E180A
k cat	0.0148 ± 0.0005512	0.3179 ± 0.01795	0.1828 ± 0.01094	0.01601 ± 0.0008603	0.01594 ± 0.0009839
K m	54.72 ± 6.219	113.7 ± 15.92	62.82 ± 10.46	87.64 ± 12.62	72.1 ± 12.6
V max	148	3179	1828	160.1	159.4

E131 and/or E180 within the amphipathic grove were predicted as the potential nucleophiles by the computational method, and the region near this residue was protected by bound ATPγS as seen by LiP‐MS. Single mutations of these residues to Alanine made no difference to the ATPase activity, and K_cat values were the same as WT protein. However, the ATPase activity of E131AE180A double mutant decreased dramatically, indicating that these negatively charged residues indeed play a catalytic role and perhaps are assisted by the bound water (Figure 5C, Table 2).

In addition to the E131 and E180, comparative structural analysis also indicated that F117 in 14‐3‐3ζ would be involved in ATP binding. F117 in 14‐3‐3ζ stacks against the ring structure of the Adenine, and this is a conserved interaction seen in other known ATPases [35]. Since the F117A protein was unstable, we made this mutation in the background of the N173V mutant 14‐3‐3ζ, which was a more thermally stable mutant with enhanced ATPase activity. The ATPase activity of this F117A/N173V double mutant was approximately 50% less than that of N173V alone, indicating that F117 is indeed proximal to the bound Adenine and provides the stacking energy (Figure 5C). Taken together, the results suggest that (a) either E131 or E180 is necessary for ATPase activity and (b) that F117 could be a key residue in nucleotide binding. It is clear that one of the binding pockets for ATP is the peptide binding amphipathic pocket and more importantly this pocket carries the catalytic residues important for ATP hydrolysis.

3.4. Mechanistic Rationale of the Enhanced ATPase Activity of the D124 and N 173 Mutant 14‐3‐3ζ

Both D124A and N173V mutations were defective in peptide binding but showed enhanced ATPase activity. These proteins were less stable and had an open conformation compared to the WT protein. While this may enable the “cryptic site” to become accessible to ATP, it also enhances peptide dissociation. To understand the effect of these mutations on the structure, we performed long MD simulations.

To quantify the local interaction pattern of the binding site, we obtained the hydrogen bonding network of the binding site residues across different time points (every one ns) during the simulations (Figure 6A). We observe a consistent inter‐helical hydrogen bonding network between the side chains of the residues D124, N173 and Y149, proximal to the ATP binding site (Figure 6B). Specifically, residue D124 always remains bound to either N173 or Y149 via hydrogen bonds and is never isolated across all snapshots, making it a crucial residue for the binding site (Figure 6C). Similarly, N173 also forms hydrogen bonds, but for only ~80% of the sampled frames with residues D124 and Y149 (Figure 6D). It is not just the residence time that was different in the two cases, the average donor‐acceptor distance for hydrogen bonds were 3.16 Å between D124 and N173, 2.64 Å between D124 and Y149, and 3.56 Å between N173 and Y149. Also, the average inter‐helical distances between apo and the holo‐structures were 8.59 Å and 9.33 Å respectively. These differences in the inter‐helical hydrogen bond network formed by the side chains of these residues may explain the higher stability of N173 mutant compared to the D124 mutant.

FIGURE 6.

Hydrogen bond network of 14‐3‐3ζ during MD simulations. (A) The number of hydrogen bonds (labeled within the square box) formed by residues of the ATP binding site and ATP (x‐axis) with other residues (y‐axis) in over 1500 snapshots extracted from the triplicate 500 ns MD simulations. (B) Hydrogen bonds (cyan dotted lines) formed by residues (stick representation) in the ATP (sphere representation) binding site of 14‐3‐3ζ (ribbon representation). (C, D) Hydrogen bonds (blue dot) formed by residues D124 and N173 across 1500 ns (x‐axis) simulation with other residues‐atoms in their proximity (y‐axis). The blue dot on the row labeled “0:X” on the y‐axis indicates time points at which no hydrogen bonds were observed. (E) ATPase activity of Y82A and Y82AN173V mutant proteins were performed using ADPGloMax promega kit. Data represents the nM of inorganic phosphate (Pi) released per minute per micromolar protein (normalized with buffer) of three independent experiments with ± SEM.

This inter‐helical hydrogen bond network also maintains the ATP binding site geometry by keeping the three helices in proximity. Disrupting this network would open the binding cavity, leading to less efficient and reduced ATP binding. However, this larger cavity would allow ATP to enter freely, leading to a higher probability for E131/E180 to catalyze ATP hydrolysis. This also corroborates the findings that N173 and D124 mutant proteins behave similarly both in the peptide binding assay and in ATPase activity, as the overall effect of the mutations is affecting the same set of helices (Figure 6C). This is also evident by increased binding of Bis‐ANS dye (a dye that binds to nonpolar cavities in proteins) to the Y82AN173V mutant, indicatina g more open pocket (Figure S4).

3.5. Identification of an ATPase Null Mutant, Requirement for the Dimeric Structure for ATP Hydrolysis

Among the many regions that we mapped using LiP‐MS, a stretch of residues from 81 to 91 was unique as it was explicitly protected upon ATPγS (Figure 3C). Y82 is one of the residues in this region, and it tethers many salt bridges that are important for dimerization [3] and it was mutated to an Alanine. As noted before, Y82A is a stable monomer and bound the peptide with comparable affinity to the WT (Figure 5B). However, when tested the ATPase activity of this mutant surprisingly showed near background luminescence readings‐ it is practically an ATPase null mutant (Figure 6E). This was intriguing as the residues E131 and E180 predicted to be the catalytic residues for ATP hydrolysis, and confirmed by mutagenesis, were located in the amphipathic pocket. The Y82AN173V mutant, which is also a monomer, did not lose ATPase activity completely; however, its activity was 20% of the N173V mutant (Figure 6E) due to the catalytic residues present within the amphipathic pocket (E131 and E180). Additionally, ATP binding at the dimer interface is necessary for efficient ATP hydrolysis at the amphipathic pocket. The WT protein is a weaker ATPase as compared to the N173V mutant, and therefore, the Y82A mutation has a more exasperated effect on the ATPase activity of 14‐3‐3ζ WT and less on the N173V protein. In other words, ATP, bound to the dimer interface, acts as a positive allosteric modulator of the ATPase activity at the amphipathic pocket. It is possible that this allosteric effect may facilitate ATP access to the otherwise phosphopeptide binding, amphipathic, active site pocket.

3.6. Effect of ATP on 14‐3‐3ζ Canonical Substrate Binding

Results from previous experiments indicate that ATP at the dimer interface acts as a positive modulator of ATP hydrolysis. The catalytic residues for ATP hydrolysis are located within the amphipathic pocket well‐characterized to bind to small phospho peptides and phosphopeptide regions of bound proteins. To test if ATP at any of these sites influences peptide binding, we monitored the binding of FAM labeled phospho and nonphosphopeptides in the presence of ATP or ATPγS. We found no effect of ATP/ATPγS on the binding of mode 1 or the mode 2 peptides to 14‐3‐3ζ by FP (Figure S5A–C). One would expect that ATP, even though it is a weaker binder than the (phospho) peptides, in large excess concentrations, would competitively prevent the binding of the peptide at the amphipathic pocket [39]. However, we failed to see any competition event even at 1 mM ATP. To check whether this is true of all modes of binding, we tested the binding of one of the nonphosphorylated peptides from the protein ExoS, a known ligand of 14‐3‐3ζ. The binding of this nonphosphorylated peptide (5FAM‐ QGLLDALDLAS) to the WT protein (Figure 7A) as well as the E180A mutant (Figure S5D) was marginally but reproducibly decreased when 1 mM ATP was present in the assay mixture.

FIGURE 7.

Effect of ATP on the binding of nonphosphorylated ExoS peptide to 14‐3‐3ζ. (A) Binding of ExoS peptide to 14‐3‐3 ζWT in presence of ATP by FP. 1 mM ATP was used to check its effect on peptide binding. (B) Binding of ExoS peptide to 14‐3‐3 ζWT. (C) to 14‐3‐3ζ R55A (Dimer interface mutant) and (D) to 14‐3‐3 Y82A (monomer) by MST. Different concentration of ATPγS was used to check its effect on peptide binding. (E) Binding of ExoS peptide to 14‐3‐3 ζWT by SPR. Different concentration of ATP was used to check its effect on peptide binding. (F) The bar diagram represents the decrease in relative response of ExoS peptide with increasing concentration of ATPγS.

It is possible that the technique of fluorescence polarization is not sensitive enough to show the differences in the presence of ATP, which is much smaller in size than the labeled peptides. Microscale thermophoresis (MST) is a better technique to monitor changes due to the binding of small molecules of different sizes as it uses physical properties, i.e., thermal motion and not the size, as a probe of interaction [40]. We first checked the binding of WT 14‐3‐3ζ to standard phosphopeptides (Figure S6A) and ExoS peptide (Figure S6B) using MST. The K_D values obtained using MST and FP experiments were comparable. We then proceeded to check the effect of ATP on the binding of the phosphorylated and nonphosphorylated peptides. Consistent with the FP data, the MST experiment did not show any effect of ATP on the binding of phosphopeptide with ζWT (Figure S6C). However, the binding of nonphosphorylated ExoS peptide decreased in the presence of ATP (Figure 7B). In addition, we observed an ATP concentration‐dependent inhibition of the binding of ExoS peptide with ζWT, Y82A, and R55A mutant proteins (Figure 7B). We analyzed the kinetic parameters of these interactions (Table S6). There was a significant decrease in the B_max of ExoS binding to the WT protein, however, the K_D of the peptide remained approximately constant. This type of inhibition clearly suggested thatATP, prevented peptide binding in the amphipathic pocket of the 14‐3‐3ζ WT, not by direct competition, but as a non‐competitive inhibitor by binding to the dimer interface. In contrast to the 14‐3‐3ζ WT, the B_max of the ExoS peptide for the mutant proteins Y82A (a monomer and null for ATPase activity) and R55A (a dimer with less ATPase activity) showed very little difference (Figure 7C,D respectively). However, a very small but perceptible increase in K_D was observed in the case of Y82A and dimeric R55A proteins, with higher concentration of ATP (Figure S5) indicating some competitive binding at this high concentration. A likely explanation for such a complex nature of ATP dependent inhibition of peptide binding is discussed later.

Surface plasmon resonance, SPR, is a sensitive and most widely used technique to explore the kinetics of biomolecular interactions [41]. We used SPR to follow the binding of the nonlabeled standard mode 2 phosphopeptide (not shown) and ExoS nonphosphorylated peptide with 14‐3‐3ζ WT following published protocols [26]. The K_D value obtained using this orthologous technique was comparable to those obtained by MST and FP experiments for the interaction of FAM‐labeled ExoS peptide with the ζWT (Figure S7A).

We then used SPR to monitor the effect of ATPγS on peptide binding. It is possible that ATPγS although a weak binder compared to the peptide, may show some detectable changes in RU under these conditions and had to be corrected for. For the same reason, this technique could not be used to calculate any of the thermodynamic or kinetic parameters and was restricted to a qualitative estimation of peptide binding in the presence of ATPγS. Again, consistent with the FP and MST data, the SPR experiment did not show any effect of ATPγS on the binding of mode 2 phosphopeptide to 14‐3‐3ζ WT (Figure S7B) but showed a definitive decrease in the binding of the nonphosphopeptide ExoS to ζWT, Y82A, and R55A proteins (Figure 7E). Similar to MST data, we were able to see an ATPγS concentration‐dependent inhibition of the binding of ExoS peptide with 14‐3‐3ζ WT (Figure 7F) and the 82A, and R55A mutant (Data not shown).

4. Discussion

The results presented in this study provide an in‐depth understanding of the structure and activity correlation of ATP and peptide binding of 14‐3‐3ζ proteins, bringing out a nuanced role for the dimeric structure in allosteric regulation. These new results, including intricate mechanisms, were derived from a combination of multiple techniques such as computational modeling and docking, LiP‐MS, detection of the ATP binding site by structural comparisons of known ATP binding proteins, enzyme assays, peptide binding assays, and various structure‐guided mutations. The following are the outcomes:

1. Our computations suggest that the catalytic site for ATP hydrolysis is the amphipathic pocket; and that E131 and E180 are the nucleophiles that are likely to act through bound water to activate hydrolysis. A phenylalanine (F117) seen in other ATP‐binding proteins is possibly essential for stacking the adenine ring and contributes to ATP binding/hydrolysis (Figure 8). These mentioned residues are conserved in all 14‐3‐3 paralogs.

2. LiP‐MS data suggested the second ATP binding site is located at the dimer interface. This is not a catalytic site; however, the binding of ATP at this site is extremely crucial for hydrolysis of ATP at the amphipathic pocket. This is because Y82A is a well‐folded monomer with an intact peptide binding function, but it is an ATPase null mutant. Similarly, R55A, with intact peptide binding function, is an ATPase defective mutant. These observations can be reconciled if we imagine that the ATP bound to the dimer interface is required for allosteric regulation of ATP hydrolysis at the amphipathic pocket (Figure 8).

3. Y82AN173V mutant is less active (80% reduction in ATPase hydrolysis) as compared to the more active N173V peptide null, hyperactive ATPase monomer. This indicates that the loss of the allosteric site contributes to a significant loss in ATP hydrolysis (Figure 8).

FIGURE 8.

Model for the mechanism of action: 14‐3‐3ζ WT interacts with ExoS peptide at amphipathic pocket and ATP at dimer interface (allosteric site). This allosteric ATP inhibits the interaction of ExoS peptide but drives the hydrolysis of ATP at amphipathic pocket. ATP hydrolysis is carried out by either of the available nucleophile (E131A or E180A). D124A or N173V mutation induces conformational changes in the protein resulting in the opening of the amphipathic pocket, leading to increased accessibility and higher ATPase activity. Y82A mutation makes the 14‐3‐3 monomeric and is an ATPase null mutant, while Y82AN173V mutant, because of the more open state of the protein (due to N173V mutation) shows some ATPase activity (less than the WT) even in the absence of the allosteric site.

Besides the above direct outcome, many insights into the mechanism of ligand binding could be discerned. For example, experiments designed to probe whether the binding of ATP would prevent peptide binding brought out some unexpected features: the phosphopeptide binding of 14‐3‐3ζ WT could not be outcompeted by ATP or its analogs. This was unexpected since a weaker binder in excess should compete with the ligand if the binding site is the same (mutually exclusive). The binding of a nonphosphorylated ExoS peptide, part of the pseudomonas full‐length toxin, was indeed inhibited by ATP and its analogs. However, surprisingly, the kinetic parameters indicate that in the WT protein, this inhibition is noncompetitive in nature; i.e., with increasing concentrations of ATP/ATPγS, there was a progressive decrease in B_max with no change in K_D of the bound peptide. This prompted us to ask whether the ATP at the interface was acting as a negative modulator of this peptide interaction. We tested Y82A (monomer) and R55A (dimer interface mutation) mutant 14‐3‐3ζ proteins which were defective in binding to ATP. Both mutant proteins, as briefed before, bind all peptides with the same affinity as the WT. The presence of ATP did not affect the binding of the ExoS peptide to these proteins, indicating that the loss of peptide binding in the WT protein was primarily due to negative allosteric regulation by ATP bound at the dimer interface. A subdued “competitive inhibition” of peptide binding by ATP is also seen with these two mutants. These results once again indicate that the dimer interface has better affinity to ATP as compared to the amphipathic pocket, and ATP binding to the amphipathic pocket is much weaker than that of the peptide. This is the reason why some degree of competitive inhibition could be achieved whenever the allosteric site at the dimer interface was defective in ATP binding upon mutation.

These results were encouraging as they highlighted the possible physiological relevance of ATP binding albeit what seems to be a highly selective function. Among all the peptide ligands with solved crystal structures, the ExoS peptide alone binds in a reverse orientation—i.e., it binds in the C to N‐ orientation in contrast to others, which bind in the N to C direction. Also, the ExoS peptide sits in a relatively extended form, covering the entire length of the binding pocket as compared to the mode1/2 peptides, covering only half of the binding site [42]. It is unclear, however, how this orientation and binding differences may facilitate allosteric communication between the interface ATP and the peptide binding groove. But we confirm that the ATP at the interface can play an allosteric role by drawing some parallels from the analysis of structures of proteins with bound ATP [43]. Two different ATP conformations have been observed: a linear and stretched ATP at the catalytic sites and a bent conformation occupying a lesser volume, which is the predominant conformation of the allosteric form of ATP [43]. Our observations from mutation experiments indicate that the amphipathic pocket is the only likely catalytic site in 14‐3‐3ζ, and in the docked structures, the ATP sits in this pocket in an extended linear form, similar to ATP within the known catalytic sites [43]. In the dimer interface, the volume available for ATP binding is enough to accommodate one ATP in the bent conformation and is likely the allosteric ATP.

In conclusion, this work, which combines computational methods and detailed biochemical and biophysical experiments, confirms that 14‐3‐3 is an ATPase, and the catalytic activity resides within the amphipathic pocket. An allosteric binding site for ATP at the dimer interface acts as a positive modulator of ATP hydrolysis (binding) and a negative regulator of selective peptide ligands at the amphipathic pocket. Despite the apparently nonconserved ATP binding signatures, the adenine binding hydrophobic phenylalanine residue and the catalytic glutamic acid residues are functional conservations seen in other well‐characterized ATPases. The semblance to the conformation of ATP at the catalytic and allosteric sites further indicates that the in vivo significance of this enzymatic role of 14‐3‐3 remains to be explored. Alternatively, there may be other unexplored mechanisms that prevent 14‐3‐3 from functioning as an ATPase in normal physiological conditions, and it may have undergone a directed evolution into a phosphoprotein binding protein.

Author Contributions

P.B. conceptualized and designed experiments, performed mutation, cloning, and purification of proteins, performed all biochemical and biophysical studies, and wrote the manuscript. N.S. and M.S.M. performed computational modeling and docking, wrote, and edited the manuscript. D.J. designed, performed LiP‐MS, analyzed the data, and contributed to writing. S.D. initiated the study with a few experiments and contributed to writing. T.K. performed computational modeling during the revision process. P.V. conceived, conceptualized, designed, and supervised the study, analyzed the data, and edited the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: fsb270985‐sup‐0001‐supinfo.docx.

Figure S1: Root Mean Square Fluctuations (RMSF) of the first three principal components of 14‐3‐3ζ in their ATP bound (A, C, E) and ATP unbound (B, D, F) states. The RMSF was calculated across three principal components using the triplicate 500 ns Molecular dynamics simulations trajectories for both ATP‐bound and ATP‐unbound states. Each replicate is represented in a different color. Predicted binding site residues are marked with asterisks.

Figure S2: (A) SDS‐PAGE shows protection by ATPγS. (B) The MS/MS spectra of conformotypic peptides identified by Lip‐MS of 14‐3‐3ζ in the presence and absence of ATPγS.

Figure S3: Thermal stability of the 14‐3‐3ζ WT monitored using nano‐DSF. The first derivative of tryptophan fluorescence was used to determine the T_m.

Figure S4: Conformational changes of the WT and mutant proteins. (A) binding of fluorescent Bis‐ANS dye to 14‐3‐3ζ WT and mutant proteins.

Figure S5: Effect of ATP on the binding of peptides using FP. (A) Effect of ATP on binding of mode1 and mode2 peptides with 14‐3‐3ζ WT protein. (B) Effect of ATP on binding of mode1 peptide with 14‐3‐3ζ mutant proteins. (C) Effect of ATP on binding of mode 1 and mode 2 with 14‐3‐3ζ E180A mutant protein. (D) Effect of ATP on the binding of ExoS peptide with 14‐3‐3ζ WT and E180A mutant protein.

Figure S6: Peptide binding of 14‐3‐3 measured using MST and the effect of ATP. (A) Interaction 14‐3‐3ζ WT with mode1 peptide. (B) Interaction 14‐3‐3ζ WT with ExoS peptide (C) Effect of ATP on binding of phosphopeptide with 14‐3‐3ζ WT.

Figure S7: Binding of peptides by 14‐3‐3 and effect of ATPγS measured using SPR: (A) The kinetics 14‐3‐3ζ WT with exoS peptide. The sensorgram were fitted to 1: 1 binding model. (B) The effect of 1 mM ATPγS on binding of model 2 phosphopeptide was checked.

Table S1: ATP Bound structures used to create molecular descriptor library.

Table S2: List of primers designed for mutations.

Table S3: Sequence of peptides used for interaction study.

Table S4: Secondary structure estimation of 14‐3‐3 mutant proteins using BESTSEL software.

Table S5: The T_m of the proteins calculated from the first derivative of tryptophan fluorescence using nanoDSF.

Table S6: Table represents the kinetic parameters for the binding of ExoS peptide with 14‐3‐3 mutant proteins.

Video S1: fsb270985‐sup‐0004‐VideoS1.mp4. Simulation of 14‐3‐3 Zeta with ATP.

Video S2: fsb270985‐sup‐0005‐VideoS2.mp4. Simulation of 14‐3‐3 Zeta without ATP.

Video S3: fsb270985‐sup‐0006‐VideoS3.mp4. Principle component 3 (PC3) of 14‐3‐3 Zeta in ATP bound form.

Bagdiya P., Soni N., Jaiswal D., et al., “ ATP‐Driven Allosteric Regulation of 14‐3‐3: Positive Modulation of ATP Hydrolysis and Negative Regulation of Peptide Binding,” The FASEB Journal 39, no. 18 (2025): e70985, 10.1096/fj.202500445R.

Data Availability Statement

The datasets generated and/or analyzed during the current study are available as part of the Supporting Information files. Additional data or materials supporting the findings of this study are available from the corresponding author upon reasonable request.