A Thr/Ser-centered phosphate-binding triad of phosphate-binding protein SphX from Synechocystis

Abstract

Coordinating phosphate acquisition with overall nutritional status is critical for cyanobacterial survival in oligotrophic environments, yet the molecular basis for substrate selectivity in phosphate-binding proteins (PBPs) remains incompletely understood. Here we reveal two evolutionarily distinct PBP architectures in Synechocystis sp. PCC 6803 through integrative structural and biochemical analyses. High-resolution crystal structures (1.76–1.9 Å) of PstS1 and SphX bound to inorganic phosphate (Pi) demonstrate that D-type PBPs utilize a conserved aspartate-mediated low-barrier hydrogen bond (LBHB) for high-affinity Pi binding (nanomolar Kd), whereas S-type SphX employs a previously uncharacterized Asp-Arg-Thr catalytic triad with reduced affinity (micromolar Kd). Despite forming comparable hydrogen bonding networks (14 versus 15 bonds), SphX exhibits 26-fold lower Pi affinity due to threonine-mediated proton acceptance replacing the LBHB mechanism. Systematic mutagenesis confirms the functional importance of each triad residue and indicates partial compensation by neighboring residues. Phylogenetic analysis demonstrates that these architectural variants are conserved across cyanobacterial lineages, representing adaptive solutions to diverse nutrient-limited environments. Our findings establish a mechanistic framework for understanding PBP functional diversification and provide molecular insights relevant to optimizing cyanobacterial productivity in biotechnology applications.

Cyanobacterial phosphate-binding protein PstS1 (D-type) achieves high nanomolar Pi affinity via a conserved aspartate, whereas SphX features a Asp-Arg-Thr catalytic triad with lower micromolar affinity and Thr as proton acceptor.

Introduction

Phosphorus (P) is an essential element for the growth of all organisms, playing a critical role in the synthesis of key biomolecules, including nucleic acids, proteins, and cell membrane lipids¹. Although phosphorus is abundant on Earth, it is primarily sequestered in rocks², limiting its bioavailability. As a result, living organisms have evolved efficient mechanisms to acquire inorganic phosphate (Pi) from their environment³. Phosphate transport in bacteria and archaea is mediated by two distinct systems: the phosphate inorganic transport (Pit) system, a low-affinity, low-selectivity transporter that functions at high phosphate concentrations, and the phosphate-specific transport (Pst) system, which exhibits high affinity and selectivity for Pi⁴. The Pst system, a member of the ATP-binding cassette (ABC) transporter family, comprises three essential components: PstC, PstA, and PstB⁵. These components are encoded by the pstSCAB gene cluster, where pstS encodes a periplasmic phosphate-binding protein (PBP) that delivers Pi to the transmembrane proteins PstC and PstA, which form the transport channel, while PstB is an ATP-binding protein⁶. The freshwater cyanobacterium Synechocystis sp. PCC 6803 (Syn6803 for abbreviation) possesses two Pst systems, pst1 and pst2 gene cluster⁷. Pst1 functions as the primary low-affinity transporter, while pst2 serves as a high-affinity transporter that is upregulated under Pi limitation to enhance Pi acquisition. The two Pst systems in Syn6803 are associated with three PBPs (or PstS proteins): SphX (sll0679), PstS1 (sll0680), PstS2 (slr1247)^8,9. A fourth PBP, Sll0540, is encoded by sll0540 in the Syn6803 genome but does not colocalize with either pst1 or pst2. SphX and PstS1 are colocalized within the pst1 gene cluster, with PstS1 serving as a major component of the Pi sensing machinery⁸. SphX primarily responds to nitrogen stress, while PstS1 acts as a high-affinity, low-velocity system that handles Pi stress⁸. Sll0540 is not colocalized with either pst1 or pst2, and its function remains unknown. The presence of multiple phosphate uptake systems in Syn6803 provides several selective advantages, including adaptation to fluctuating Pi concentrations and functional redundancy to compensate for potential loss of function¹⁰.

In phosphate acquisition research, two pivotal questions focus on the binding mechanisms of PstS proteins and their capacity to differentiate phosphate from its close analogue. Crystal structures of PstS proteins from diverse sources have been elucidated^11–18, offering comprehensive insights into the molecular interactions that enable these proteins to selectively bind phosphate and distinguish it from structurally analogous compounds. PstS proteins adopt a venus-flytrap topology, composed of two large lobes comprising a central β-sheet core flanked by an α-helix and, linked together with a flexible hinge¹⁹. The phosphate-binding pocket is located within a deep cleft between the two large lobes of the PstS protein. Pi binds tightly to PstS proteins through 12-14 strong hydrogen bonds involving 8-9 residues^12,18,20. Specifically, a defining characteristic of PstS proteins is the presence of a critical aspartate residue that forms a short, low-barrier hydrogen bond (LBHB) with the bound phosphate anion¹². This interaction confers high selectivity, enabling PstS proteins to preferentially bind phosphate over competing anions, such as arsenate, with exceptional specificity (arsenate is discriminated against by approximately 3 orders of magnitude, sulfate by approximately 5 orders of magnitude)^11,12. Later research determined the crystal structures of new type of PstS proteins from bacterial that form 14 hydrogen bonds, but no LBHB, with the bound phosphate anion^21,22 (two additional structures with PDB codes 1TWY and 4GD5 exist but lack published papers). Sequence alignment shows that the key aspartate residue is substituted by a serine residue in this type of PstS. Although two additional classical hydrogen bonds are formed, the capability of this type of PstS protein to discriminate phosphate from the closest competing anions, such as arsenate, is greatly reduced. It is hypothesized that the reduced discrimination capacity of this type of PstS results from its unique phosphate-binding configuration²². However, this hypothesis still requires validation.

A previous report separated cyanobacterial PBPs into either the archetypal PstS type or the SphX type to investigate the functional significance of individual PstS proteins⁸. Sequence alignment of PstS homologs, integrated with structural analysis, categorizes them into two distinct types: D-type, characterized by the presence of a key aspartate residue, and S-type, distinguished by a serine residue substitution (Fig. 1). D-type PstS proteins exhibit high selectivity and high affinity with a LBHB of a critical aspartate residue, while S-type PstS proteins exhibit reduced selectivity and affinity with no LBHB. The freshwater cyanobacterium Syn6803 possesses both types of PstS proteins, with PstS1 and PstS2 belonging to the D-type and SphX belonging to the S-type (Fig. 1). The fourth PstS protein, encoded by sll0540, is situated outside of the two Pst systems and shows little response under phosphate-limiting conditions in physiological experiments⁸. The Sll0540 protein exhibits relatively low phosphate-binding affinity, and its crystal structure has been determined²³, which may indicate it possesses chimeric features of both D-type and S-type PstS proteins. Site-directed mutagenesis significantly enhances its phosphate-binding affinity²³. These findings suggest that Sll0540 may represent a vestigial component derived from an ancestral gene, potentially serving as a backup mechanism for adaptation to phosphate scarcity.

Fig. 1. Alignment of the amino acid sequences of PstS with known structures and three PstSs in Syn6803.

Residues in magenta, red, cyan, and green are key residues that interact with phosphate. The red box indicates the main discriminator between D-type and S-type PstSs. The black arrow indicates conserved residues in S-type PstSs. The red triangular arrow indicates conserved residues across both types that interact with phosphate. Three PstSs in Syn6803 are classified into D-type and S-type (SphX), with two belonging to D-type (PstS1 and PstS2).

SphX and PstS1 colocalize within the same functional gene cluster yet are classified as distinct protein types at the sequence/structural level. Here, we assessed the binding specificities of both these two proteins toward various phosphorus species and sulfate in cyanobacteria, confirming Pi as the endogenous ligand. Binding affinity assays revealed that PstS1 exhibits the higher Pi binding affinity, whereas SphX displays significantly reduced affinity. Crystal structures of PstS1 (D-type) and SphX (S-type) were determined at resolutions of 1.9 Å and 1.76 Å, respectively. PstS1, possessing the LBHB aspartate, achieves high Pi selectivity through direct proton acceptance by this critical residue, whereas SphX employs a Asp89-Arg231-Thr17 catalytic triad where threonine serves as the proton acceptor. Phylogenetic analysis suggests the common feature that the divergence of D-type and S-type PstS proteins reflects evolutionary adaptations to diverse oligotrophic environments, enabling cyanobacteria to sense and respond to different environmental stresses.

Results

Characterization of two types of PstS proteins in Syn6803

To investigate the sequence features of PstS proteins, 20 bacterial PstS proteins from Uniprot (https://www.uniprot.org/) with determined crystal structures were analyzed. After removing redundant protein sequences, 14 unique PstS protein sequences were extracted for further study. Sequence alignment was performed on three functionally characterized PstS proteins from Syn6803 and 14 unique PstS proteins from UniProt (Fig. 1). Structural analysis revealed that the LBHB aspartate residue is highly conserved in some PstS proteins, while in others, this aspartate residue is substituted by a serine (marked in the red box in Fig. 1). Based on this distinction, we designated this variant as D-type and S-type, respectively. Another common feature of S-type PstS proteins is that an aspartate residue in the middle of the protein sequence, a threonine at the N-terminus and an arginine at the C-terminus are highly conserved (marked with black arrow in Fig. 1), indicating that they may participate in Pi binding for this protein type.

Sequence alignment revealed that the three PstS proteins from Syn6803 fall into D-type and S-type. PstS1 and SphX, in the same functional gene cluster, representing the D-type and S-type classes of PstS proteins, respectively, were selected for detailed structural and functional characterization (Fig. 2). Differential Scanning Fluorimetry (DSF) is a powerful technique for analyzing protein stability, ligand binding, and conformational changes. Ligand binding can increase or decrease the melting temperature (Tm, the temperature at which half the protein is unfolded), providing a measure of protein thermal stability. We used DSF to characterize the binding of two PstS proteins from Syn6803 to phosphate in different forms and to sulfate (Fig. S1). We calculated the ΔTm for each protein by comparing the melting temperatures before and after incubated with phosphate, phosphite, β-glycerophosphate, tripolyphosphate, or sulfate, which are physiologically relevant anions in cyanobacterial cells that may compete with phosphate for binding to PBPs The ΔTm values of PstS1 and SphX increased significantly upon Pi binding (Figs. 2A, S1 and Table S1). PstS1 also bound phosphite, while SphX bound to β-glycerophosphate, both producing substantial but smaller ΔTm increases compared to Pi (Fig. 2A and Table S1). Neither PstS1 nor SphX responded to sulfate (Fig. 2A and Table S1). The DSF data suggests that Pi is the most likely physiological ligand of Syn6803 PstS proteins.

Fig. 2. Characterization of two types of PstS proteins in Syn6803.

A The ΔTm of PstS1 and SphX with different ligands. PO₄: phosphate; PO₃: phosphite; GP: β-glycerophosphate; TP: tripolyphosphate; SO₄: sulfate. B ITC titration data for PstS1 upon titration with phosphate at pH 5.0 and 8.0. C ITC titration data for SphX upon titration with phosphate at pH 5.0 and 8.0. Data are presented as mean ± SD, with n = 3 biologically independent experiments per condition. The effect of different ligands on ΔTm was assessed by two-way ANOVA followed by Tukey’s multiple comparisons test. All pairwise comparisons were performed within each protein. ****p < 0.0001 (GraphPad Prism does not report more precise values below this threshold). Representative effect sizes (mean differences) and 95% confidence intervals are shown here: For PstS1, the mean differences (95% CI) for PO₄ versus PO₃, GP, TP, and SO₄ were 4.49 °C [3.06–5.92], 5.84 °C [4.41–7.26], 5.59 °C [4.16–7.02], and 7.07 °C [5.64–8.50], respectively (all P < 0.0001). For SphX, the mean differences (95% CI) for PO₄ versus PO₃, GP, TP, and SO₄ were 7.45 °C [6.02–8.88], 7.23 °C [5.80–8.66], 9.73 °C [8.30–11.16], and 8.85 °C [7.42–10.28], respectively (all p < 0.0001).

The bindingxof PstS proteins to Pi is pH-dependent, as phosphate exists in either dibasic (HPO₄²⁻) or monobasic (H₂PO₄⁻) forms at different pH values (e.g., pH 5.0 and 8.0)¹². The binding affinity varies at different ionic strengths and pH values. Isothermal titration calorimetry (ITC) was used to measure the binding affinities of Syn6803 PstS proteins for Pi at two pH values to assess binding to different Pi forms under physiological ionic strength. Prior to measuring phosphate-binding affinity, we removed residual phosphate substrates from the purified PstS proteins by protein denaturation and renaturation²⁴. The equilibrium dissociation constants (Kd) for PstS1-Pi and SphX-Pi binding were determined at pH 5.0 and 8.0 (Figs. 2B, C and S2). Both proteins exhibited high Pi binding affinities at pH 8.0, with affinities decreasing approximately two-fold at pH 5.0 (Figs. 2B, C and S2). This pH-dependent reduction in binding affinity was more pronounced in another S-type PstS protein (PBP-1) from Clostridium perfringens (Fig. S3). These results suggest that PstS proteins preferentially bind the dibasic phosphate species (HPO₄²⁻) that predominates at pH 8.0. Beyond this shared pH sensitivity, the two Syn6803 proteins displayed strikingly different absolute binding affinities. At pH 8.0, PstS1 bound Pi with 26-fold higher affinity than SphX (Fig. 2B, C). Given that PstS1 contains the LBHB aspartate residue while SphX harbors a serine substitution at this position (Fig. 1), we hypothesized that this single residue difference might not fully explain the dramatic affinity gap. To test this hypothesis and identify additional structural determinants, we solved the crystal structures of both PstS1 and SphX in complex with phosphate.

Crystal structures of PstS1 and SphX

While phosphate-binding proteins are conserved across bacteria, cyanobacteria uniquely evolved multiple functional variants. The S-type SphX integrates phosphate, nitrogen, and light signals, enabling nutrient coordination unavailable to heterotrophic bacteria. To elucidate the structural basis for this functional discrimination, we determined the crystal structures of PstS1 and SphX. Prior to crystallization, lysine residues in the proteins were methylated to reduce protein flexibility and enhance crystallization probability. Notably, lysine methylation was confined to solvent-exposed surface residues and did not involve any residues within the phosphate-binding pocket, minimizing its potential impact on pocket hydration or ligand coordination (Fig. S4). The crystal structure of PstS1 complexed with Pi was determined at 1.9 Å resolution in space group P21 using crystallization conditions of 0.1 M Tris-HCl pH 9.0, 0.2 M MgCl₂, and 30% PEG3000 (Fig. 3A and Table 1). The PstS1 crystal contains two protomers in the asymmetric unit, with the final model consisting of 660 amino acid residues distributed across both subunits, along with two Pi molecules, two Mg²⁺ ions, and 965 water molecules. The PstS1 structure harbors a typical Venus flytrap fold that is composed of two globular α/β domains held together by two β-strands (Fig. 3A). The two domains are connected by a flexible hinge region, and each α/β domain contains a central mixed β-sheet of four or five strands surrounded by five α-helices. Likewise, a single Pi molecule binds within the cleft between the two α/β domains (Fig. 3A).

Fig. 3. Crystal structures of PstS1 and SphX with bound phosphate.

A Structure of PstS1. Domains are colored in TV-blue and bright orange, with 11 α-helices and 11 β-sheets labeled. Phosphate is represented as red stick. B Structure of SphX. Domains are colored in lemon and salmon, with 10 α-helices and 11 β-sheets labeled. Phosphate is represented as red stick. C Superimposed ribbon diagram of the structures of PstS1 (TV-blue) and SphX (lemon) with an RMSD of 2.15 Å. The main distinction between these two structures is the loop region, as marked.

Table 1.

Crystallographic statistics

Parameter	PstS1	SphX
X-ray source	BL10U2, SSRF	BL10U2, SSRF
Wavelength (Å)	0.97918	0.97918
Space group	P 1 21 1	P 31 2 1
Unit-cell dimensions (Å or °)	a = 37.718, b = 64.816, c = 129.058, α = γ = 90, β = 91.27	a = 90.426, b = 90.426, c = 198.307, α = β = 90, γ = 120
Unique reflections	48885 (4805)	93274 (9191)
Completeness (%)	99.30 (98.46)	99.87 (99.37)
Mean I/sigma (I)	12.21 (6.22)	14.68 (1.56)
Multiplicity	1.0 (1.0)	2.0 (2.0)
Resolution range (Å)	43.01–1.9 (1.968–1.9)	26.64–1.763 (1.826–1.763)
Rwork	0.1761 (0.2081)	0.1792 (0.2937)
Rfree	0.2131 (0.2764)	0.1952 (0.3053)
Ligands	12	94
Water molecules	896	523
Protein residues	659	607
Bond lengths (Å)	0.007	0.008
Bond angles (°)	0.91	1.03
Ramachandran favored (%)	98.42	98.75
Ramachandran allowed (%)	1.58	1.25
Ramachandran outliers (%)	0.00	0.00
PDB code	9WV8	9WP7

SphX was processed using the same method, with lysine residues methylated prior to crystallization. Crystals of SphX bound to Pi were obtained in a solution containing 0.5 M Li₂SO₄ and 20% PEG8000. The SphX complex crystal (space group P3₁2₁) contains two protomers in the asymmetric unit, with the final model consisting of 596 amino acid residues distributed across both subunits, along with two Pi molecules, two sulfate ions, and 523 water molecules, yielding an R-factor of 17.81% at 1.76 Å resolution (Fig. 3B and Table 1). SphX adopts a similar Venus flytrap fold topology, with almost identical secondary structure (Fig. 3B).

SphX contains fewer amino acids (307 residues) than PstS1 (333 residues), a difference that is clearly visible when the structures are superimposed, yielding an average Cα RMSD of 2.15 Å (Fig. 3C). This shorter length does not appear to be a characteristic feature of S-type PstS proteins^21,22 (Figs. 3B, C and S5). The central α/β domain and the phosphate-binding pocket are relatively conserved. However, the loop regions, particularly after β4 in domain II, show poor structural alignment. These include loops Asn95-Gln106, Leu114-Thr140, and Ile169-Lys183 in PstS1, corresponding to loops Val94-Ser104, Ala118-Asn138, and Val157-Leu168 in SphX, respectively (Fig. 3C). In contrast to the S-type PBP-1 and Sll0540 (PDB code: 4Q8R, 9L8J), where significant loop and C-terminal fragments are missing²², SphX retains these important structural elements and adopts an ingenious Venus flytrap folding arrangement to maintain 3D structural conservation (Figs. S6A and S7A). Despite these structural differences, all S-type PstS proteins retain the same Pi-binding mode (Figs. S6B and S7B).

The phosphate-binding site

The crystallization of both PstS proteins was performed under neutral or alkaline conditions. Given that both PstS proteins bind dibasic phosphate species (HPO₄²⁻) with high affinity, we propose that both protein complex structures contain Pi in the dibasic phosphate form. The electron density maps clearly reveal the phosphate-binding sites in both structures (Fig. S8). Hydrogen bonds involved in phosphate coordination are detailed in Tables S2 and S3.

For the PstS1 protein, the Pi anion is fully sequestered within the interdomain cleft, creating a completely dehydrated binding environment. Pi is bound by 15 hydrogen bonds formed with eight residues distributed between the two domains (Fig. 4A and Table S2). The Pi anion forms hydrogen bonds with 6 backbone amide groups (Ser18, Phe19, Ser47, Gly149, and Thr150), 6 side chain hydroxyl groups (Ser18, Ser47, Asp65, Ser148, Thr150), and 3 arginine guanidinium groups (Arg144) (Fig. 4A, S9 and Table S2). Two key residues play important roles in Pi binding and stabilization. Asp65, which forms an LBHB, has been demonstrated to distinguish between phosphate and other analogs¹³. Arg144 may play an important role in balancing the negative charge of Pi¹⁴. A crucial scientific question is how the PstS protein interacts with the hydrogen atoms in Pi. While the electron density maps do not allow accurate determination of hydrogen atom orientations at this resolution, careful analysis of the interacting residue positions and the LBHB (Asp65 at approximately 2.5 Å) supports a hydrogen bonding network consistent with dibasic phosphate binding (HPO₄²⁻)¹³. In this configuration, the single proton on the dibasic phosphate (covalently bound to O1 in our structure) is donated to the Asp65 side chain carboxylate group, which appears to be the only acceptor group that is not already engaged in hydrogen bonding with other protein residues. This LBHB is characteristic of D-type PstS proteins and confers the exquisite specificity observed for the receptor and its transport system^6,12–14. In addition to the LBHB, the O1 of Pi forms hydrogen bonds with backbone amide groups from Ser18 and Phe19 (Fig. 4A and Table S2). These 2 additional hydrogen bonds may modulate the strength of the LBHB and stabilize Pi binding within the pocket.

Fig. 4. Phosphate binding by PstS1 and SphX and comparison of their binding pockets.

A Hydrogen bonds between PstS1 and phosphate. Phosphate is represented as red stick. B Hydrogen bonds between SphX and phosphate. Phosphate is represented as red stick. C Superposition of the phosphate binding clefts of both proteins. PstS1 residues are colored TV-blue (domain I) and bright orange (domain II). SphX residues are colored lemon (domain I) and salmon (domain II).

Like PstS1, Pi in the SphX complex binds within the interdomain cleft, occupying nearly the same position as in PstS1 (Fig. 4B). However, unlike the completely dehydrated PstS1 binding site, the SphX binding pocket retains at least one water molecule (one water in one protomer, two in the other one, Fig. 4B). Pi is bound by 14 hydrogen bonds formed via eight residues distributed between the two domains and two water molecules (Fig. 4B and Table S3). The Pi anion forms hydrogen bonds with 5 backbone amide groups (Ser16, Ser17, Thr45, Gly147, and Thr148), 7 side chain hydroxyl groups (Ser15, Thr17, Thr45, Ser63, Ser146, and Thr148), and two water molecules (Fig. 4B and Table S3). Two key residues that are crucial for Pi binding in D-type PstS1 are absent in the SphX structure. The corresponding sites of Asp65 and Arg144 in PstS1 are substituted by Ser63 and Pro142 in SphX, respectively (Fig. 4B), which indicates a different binding mechanism. Analysis of the positions of the two water molecules reveals that they occupy the space vacated by the bulkier side chain of Arg144 in PstS1. Similar to arginine’s role in providing positive charge for Pi stabilization, water50 may become positively charged (forming H₃O⁺) via proton transfer from Asp173, thereby providing a counterbalancing positive charge to neutralize the negative charge of the bound phosphate (Fig. 4B). Interestingly, Pi binding in other S-type proteins, such as C. perfringens PBP-1 and Sll0540, occurs in a completely dehydrated environment (Figs. S6B and S7B). The crystal structure indicates that an Arg residue stabilizes the negative charge of Pi^22,23 (Figs. S6B and S7B).

SphX adopts a unique binding mechanism

Since Pi binds to the PstS protein as the dibasic species (HPO₄²⁻), identifying the residue capable of accepting a proton from this species represents a fundamental scientific question. In D-type PstS proteins, the LBHB aspartate residue serves dual functions as both proton acceptor and discriminator. In contrast, S-type PstS proteins lack this key residue, leaving the identity of the proton-accepting residue unclear. The crystal structure of C. perfringens PBP-1 suggests that the single proton carried by the phosphate moiety may be accepted by the Ser13 side chain²². Nevertheless, the hydroxyl hydrogen of serine is not readily dissociable (the pKa of serine -OH side chain is very high), making it challenging to accept an additional proton for hydrogen bond formation.

Comparison of SphX with C. perfringens PBP-1 suggests that Thr17, corresponding to Ser13 in C. perfringens PBP-1, is a likely proton acceptor of Pi (Fig. 5A, B). We analyzed the structural environment around Thr17 to understand this mechanism. This residue interacts with Pi via a conventional hydrogen bond at a distance of 2.7 Å (Fig. 5A). Two key residues in the vicinity, Asp89 and Arg231, appear crucial for Thr17’s function as a proton acceptor (Fig. 5A). We propose the following mechanism: The negatively charged Asp89 forms a salt bridge with Arg231, polarizing the arginine side chain. This interaction positions Arg231 within 3.2 Å of Thr17, enabling Arg231 to abstract a proton from Thr17 (Fig. 5A). The deprotonated Thr17, which directly interacts with Pi, can then accept a proton from the bound phosphate. This Asp-Arg-mediated proton relay mechanism resembles those observed in other protein systems, including haloalcohol dehalogenases, where catalytic triads such as Ser-Tyr-Arg²⁵, or Ser-Tyr-Lys modulate proton transfer during catalysis^26–28. The amino acid triad explains how S-type PstS proteins can effectively bind dibasic Pi despite lacking the LBHB aspartate residue found in D-type PstS proteins. Structural comparison demonstrates that the Asp-Arg pair is structurally conserved across multiple S-type PstS proteins, including C. perfringens PBP-1, Sll0540, V. cholerae, and P. aeruginosa, consistent with our proposed proton relay mechanism (Figs. 5B and S10). All primary residues of SphX that form direct hydrogen bonds with Pi oxygens are further stabilized through secondary hydrogen bonds with other residues (Fig. 5C). Within this hydrogen bonding network, the Thr/Ser-centered Asp-Arg triad appears to be particularly important for accepting protons from the bound Pi.

Fig. 5. Triad binding site of SphX, comparison with S-type PBPs, and hydrogen bonding with dibasic phosphate.

A Triad binding site of SphX. Phosphate is represented as red stick. Water is represented as red dot. B Superposition of the phosphate triad binding clefts of SphX and S-type PBPs from C. perfringens. Phosphate is represented as red stick. C Schematic diagram of the hydrogen bonds between SphX and dibasic phosphate and the networks of hydrogen bonds involving other groups in the binding site. D Proposed mechanism of HPO₄²⁻ binding by SphX.

To validate this hypothesis, site-directed mutagenesis was performed to assess the roles of specific residues in Pi binding (Fig. S11). Mutation of Thr17 to alanine completely abolished SphX binding to Pi (Fig. 6A and Table S4), whereas mutation to serine reduced binding affinity from the nanomolar to micromolar range (Fig. 6B and Table S4). Similarly, the Arg231Ala mutation decreased binding affinity from nanomolar to millimolar levels while the Arg231Lys mutation retains the binding affinity to micromolar (Fig. 6C, D). The retention of residual Pi-binding capacity suggests the presence of an alternative binding mode after removal of the positively charged Arg231 residue. Under these conditions, the interaction may involve a different protonation state of phosphate, although the precise protonation state cannot be definitively assigned based on the current data. (Fig. 6D and Table S4). Asp89 functions as one of the triad residues. We were unable to obtain binding affinity data for the Asp89Ala mutant by ITC, whereas the Asp89Asn mutation preserved Pi-binding capacity but reduced the affinity to the micromolar range (Fig. 6E and Table S4). Collectively, these results demonstrate the essential roles of Thr17, Arg231 and Asp89 in Pi recognition.

Fig. 6. Binding affinity of SphX triad binding site mutants.

ITC titration data for SphX mutants upon titration with phosphate at pH 8.0. A Thr17Ala; B Thr17Ser; C Arg231Lys; D Arg231Ala; E Asp89Asn; F Thr148Ala.

Thr148 is structurally equivalent to Thr17 in SphX and similarly interacts with Asp89 and Arg231, raising the possibility that it also functions as a proton acceptor. As a residue in direct contact with Pi, mutation of Thr148 to alanine did not completely abolish binding but substantially reduced affinity, establishing it as a key contributor to Pi coordination (Fig. 6F and Table S4). The TSA assay was consistent with our ITC results (Fig. S12 and Table S5). However, the Tm of the Thr148Ala mutant showed no significant change upon addition of 2 mM Pi (Fig. S12 and Table S5). This observation suggests altered or weakened stabilization upon phosphate binding. The PBP-1 from C. perfringens displays a conserved residue arrangement, with binding affinities for the wild-type protein and analogous mutants aligning closely with those observed for SphX (Fig. S13 and Table S4). Unexpectedly, substituting Ser13 with alanine in C. perfringens (corresponding to Thr17 in SphX) decreases binding affinity by approximately 30-fold while maintaining residual activity. This residual activity suggests the possible involvement of the structural equivalent threonine as a proton acceptor. We therefore predict that Thr131 in the PBP-1 from C. perfringens (corresponding to Thr148 in SphX), which occupies a structurally equivalent position, also serves as a proton acceptor analogous to Thr17 in SphX.

We therefore refine our mechanistic model as follows: the negatively charged aspartate forms a salt bridge with arginine, increasing the basicity of the arginine guanidinium group. This polarized arginine is strategically positioned to abstract protons from the hydroxyl groups of both threonine residues (Fig. 5D). Arginine abstracts protons from both threonine residues, generating anionic sites capable of accepting a proton from the bound Pi. The positions corresponding to Thr17 and Thr148 in SphX are occupied by functionally interchangeable threonine or serine residues in cyanobacterial PstS proteins. Among S-type proteins, some contain threonine at both positions, others have a threonine/serine mixture, and still others have serine at both positions (Fig. S10). Our model does not exclude the possibility that when one threonine/serine is mutated, the other can still function as a proton acceptor. For example, in C. perfringens PBP-1, the Ser13Ala mutant lost binding affinity significantly but retained most of its binding capacity, suggesting functional compensation by the remaining serine/threonine residue (Fig. S13). Interestingly, although the binding affinity varies, the secondary structure of SphX and its mutants, and the PBP-1 in C. perfringens did not change significantly (Fig. S14).

Phylogenetic insights into evolutionary adaptations of phosphate acquisition in Syn6803

To investigate whether these two types of PstS proteins are conserved among cyanobacterial PstS proteins, we constructed a maximum-likelihood phylogenetic tree using PstS protein sequences from currently available cyanobacterial type strain genomes, together with all Synechocystis sp. PCC 6803 PstS protein sequences (PstS1, PstS2, SphX, and Sll0540). Based on the tree topology, we grouped the PstS sequences into the two major clades corresponding to the D-type and S-type classifications (Fig. 7). The two Synechocystis sp. PCC 6803 D-type PstS proteins clustered together in a single clade, while the S-type PstS proteins each formed separate clades (Fig. 7). This distinct clustering pattern suggests different evolutionary trajectories for these protein types. The tight clustering of D-type proteins indicates strong evolutionary pressure for their maintenance as they are the major component of the Pi-sensing machinery⁸. Given the highly streamlined genomes of cyanobacteria, one of these D-type genes was likely vertically inherited while the other arose from a recent duplication event, creating paralogs that have been retained due to their distinct physiological roles⁸. In contrast, the S-type proteins exhibit a more divergent evolutionary pattern. SphX and Sll0540, which cluster into separate clades, likely perform similar general functions but under different physiological conditions, explaining their sequence divergence while maintaining functional relevance. SphX has been reported to respond to nitrogen stress in addition to phosphate limitation. Our biochemical characterization reveals a functional distinction between the two PstS protein types: D-type proteins such as PstS1 exhibit high-affinity Pi binding, while S-type proteins such as SphX display lower affinity. This divergence may suggest divergent physiological roles under varying nutrient conditions. The high-affinity D-type system is likely to contribute to phosphate acquisition under Pi limitation, while the lower-affinity S-type system, which responds to multiple nutrient signals, may participate in adjusting Pi uptake according to broader environmental cues. Such differential binding properties could potentially facilitate coordination between phosphate uptake and the overall nutritional status of the cell, although this interpretation remains to be experimentally validated in vivo.

Fig. 7. Phylogenetic tree of PstS sequences from cyanobacterial type strains.

A Midpoint rooted phylogeny of the PstS protein sequences built using the maximum likelihood method. The PstS sequences of cyanobacterial type strains were extracted from the NCBI GeneBank. In the tree, PstS protein sequences form two groups: D-type and S-type. B. Sequence alignments of the phosphate binding region of D-type and S-type proteins. The numbers above the top alignments indicate the conservation of the triad in S-type PstS proteins.

Discussion

Through integrated sequence alignment and crystal structure analysis, we identified two functionally distinct classes of PstS proteins in Syn6803. We designated these as D-type and S-type PstS proteins based on their characteristic structural features and sequence motifs. PstS1 and SphX were selected as representative examples of D-type and S-type PstS proteins, respectively, for detailed biochemical and structural characterization. Both PstS1 and SphX share a common ligand specificity, binding Pi in its dibasic form (HPO₄²⁻), yet they differ markedly in their binding affinities. PstS1 demonstrates high affinity and high selectivity for Pi, whereas SphX shows reduced affinity and lower selectivity. Crystal structures of PstS1 and SphX complexed with Pi have been determined at resolutions of 1.9 Å and 1.76 Å, respectively. Both proteins adopt the typical Venus flytrap fold, with Pi binding in the cleft between the two α/β domains via an extensive hydrogen bond network. However, they exhibit distinct hydration patterns in their Pi-binding pockets: PstS1 shows complete dehydration of the binding site, whereas SphX retains one or two water molecules within the Pi-binding pocket. PstS1 contains a typical LBHB aspartate residue that is crucial to the protein’s high affinity and high selectivity^6,12,13, while SphX lacks the LBHB aspartate residue adopting a different Pi binding mechanism. Crystal structure analysis of PstS1 and SphX enabled us to characterize their respective Pi-binding mechanisms and identify a previously undescribed binding mode in SphX. This unique mechanism involves an amino acid triad that serves as a coordinated proton acceptor system. Subsequent mutagenesis and binding affinity studies confirmed the essential role of this triad in mediating high-affinity Pi binding.

SphX demonstrated reduced Pi-binding affinity compared to PstS1, with Kd values differing by more than 26-fold (Fig. 2B, C). This reduced affinity correlates with the absence of the conserved LBHB aspartate residue in SphX, which is essential for high-affinity Pi binding in classical D-type PstS proteins. From a thermodynamic perspective, ligand binding typically involves displacement of structured water molecules from the binding interface, resulting in a favorable entropy gain. In SphX, the retention of one to two ordered water molecules within the phosphate-binding pocket suggests incomplete dehydration upon ligand binding. This may reduce the entropy gain associated with water release and potentially weaken direct protein–phosphate hydrogen bonding, thereby diminishing the overall binding free energy. In contrast, the fully dehydrated pocket observed in PstS1 likely permits more optimal direct interactions and a larger dehydration-driven entropy contribution, consistent with its ~26-fold higher affinity. The lack of this key residue indicates that SphX has evolved an alternative proton acceptance mechanism for Pi recognition. Our crystal structures reveal that this high affinity is achieved through a sophisticated hydrogen bonding network centered on a conserved Asp-Arg-Thr/Ser catalytic triad. Specifically, Asp89 forms a salt bridge with Arg231, which may influence the electronic environment of the arginine guanidinium group. Based on the observed geometry in the crystal structure, Arg231 is positioned in proximity to Thr17 and Thr148, raising the possibility that this arrangement could facilitate proton transfer between these residues and the bound phosphate. However, given that hydrogen atoms are not directly resolved at the current resolution, this proposed proton-relay pathway should be considered an inference derived from static structural information. We note that alternative proton-transfer routes, including pathways mediated by ordered water molecules present in the active site, cannot be excluded. The observed Asp-Arg interaction is reminiscent of salt bridge-mediated proton transfer networks described in other enzyme systems^25,28, although direct of proton movement will require further investigation by DFT calculations. The critical importance of this triad is demonstrated by our mutagenesis studies. We also reveal that other S-type PstS proteins have the same structural arrangement (Figs. 5B and S10). However, our ITC data demonstrate that both Thr17 and Thr148 are critical proton acceptors in SphX, as mutations at either position substantially reduce binding affinity. In contrast, the corresponding mutation in C. perfringens PBP-1 retained significant binding affinity, raising the possibility that when one key residue is mutated, the remaining Thr/Ser can compensate by functioning as the proton acceptor. This proton relay mechanism may provide a basis for substrate promiscuity in phosphate binding. However, further structural studies at higher resolution are needed to directly visualize proton positions and hydrogen bonding geometries, which would provide definitive evidence for the proposed proton acceptance mechanism. Additionally, neutron crystallography could reveal the protonation states of key residues in the phosphate-bound complex.

Cyanobacteria represent an outstanding model organism with vast potential in sustainable biomanufacturing and photoelectric applications, leveraging their photosynthetic autotrophic capabilities, manipulable genomes, and rapid growth rates^29,30. However, fundamental aspects of their growth processes, such as phosphate uptake mechanisms with two different types of PstS proteins, remain incompletely understood. As periplasmic Pi-binding proteins, PstS proteins serve as the primary sensors for detecting external phosphate availability^31,32. When loaded with Pi, PstS proteins interact with the membrane-bound components of the Pst transport system^8,33, triggering conformational changes that activate the transport machinery and facilitate Pi uptake into the cytoplasm⁸. The contrasting binding affinities of these two PstS protein types suggest potential functional differentiation within the Syn6803 phosphate uptake systems. Although PstS1 and SphX are colocalized within the same pst1 gene cluster, previous studies have indicated differential regulatory responses⁸. PstS1 is thought to function primarily in phosphate sensing and acquisition, whereas SphX has been reported to respond to nitrogen stress and high-light conditions⁸, suggesting possible divergence within the phosphate-binding protein family. The high-affinity D-type PstS proteins are likely to contribute to phosphate acquisition under Pi limitation and may play a central role in the phosphate uptake machinery. In contrast, the lower-affinity S-type PstS proteins, which respond to multiple environmental signals, including nitrogen availability and light intensity, may participate in modulating phosphate uptake in coordination with other cellular processes. Such differential properties could facilitate integration of phosphate uptake with overall cellular nutrient status. However, this proposed regulatory framework is currently based on biochemical characteristics and previously reported expression patterns, and direct in vivo functional analyses will be required to validate these roles.

From an evolutionary perspective, the dense clustering of D-type PBPs suggests strong sequence conservation, consistent with a conserved housekeeping role in phosphate acquisition. In contrast, S-type PBPs, including SphX and Sll0540, display greater phylogenetic dispersion, indicative of accelerated divergence. This broader sequence variability may reflect adaptive evolution under fluctuating environmental conditions. Notably, S-type PBPs have been associated with nitrogen limitation and high-light stress responses, suggesting that their diversification may facilitate functional specialization in response to nutrient imbalance or photo-oxidative stress. Although a direct causal relationship cannot be established from sequence analysis alone, the observed phylogenetic pattern is consistent with environment-driven evolutionary tuning of phosphate-binding proteins.

Materials and methods

Statistics and reproducibility

Thermal shift assay data are presented as means ± standard deviations (SDs) from at least three independent biological replicates (n ≥ 3). Each biological replicate represents an independent protein preparation and experiment performed on different days. For statistical significance analysis, two-way ANOVA followed by Tukey’s multiple comparisons test was applied in Fig. 2A, while two-way ANOVA followed by Sidak’s multiple comparisons test was used in Fig. S12 to calculate p values. All data calculations and statistical analyses were performed using GraphPad Prism 8.

Sequence alignment

To analyze conserved phosphate-binding motifs, we obtained sequences of 14 phosphate-binding proteins with reported structures from UniProt (https://www.uniprot.org/). Multiple sequence alignment was performed using Clustal Omega via the Jalview web service. We then annotated key amino acids of the catalytic triad and other functionally relevant residues using Jalview’s feature annotation tools. The final sequence alignment was exported as a PNG image for analysis.

Constructs

Signal peptides of PstS1, SphX and PBP-1 were predicted using the SignalP-6.0 web server³⁴. DNA fragments encoding truncated forms of PstS1 (GenBank accession no. BAA10341.1; residues 51-383) and SphX (GenBank accession no. BAA10342.1; residues 24-336) were amplified from Syn6803 genomic DNA by PCR. The resulting fragments were cloned into the pET28a vector via homologous recombination, introducing an N-terminal 6×His tag and a Small Ubiquitin-like Modifier (SUMO) tag. The DNA fragment encoding truncated PBP-1 (GenBank accession no. WP_011590281.1; residues 31–276), together with the pET28b vector backbone, was codon-optimized and synthesized by BGI Genomics (Shenzhen, China). The resulting construct contained an N-terminal 6×His tag followed by a thrombin cleavage site. Site-directed mutagenesis was carried out using the QuikChange® method. All constructs were verified by DNA sequencing performed by Tsingke Biotechnology. Oligonucleotide primers used for plasmid construction and site-directed mutagenesis are listed in Table S6.

Protein expression and purification

Transformed E. coli BL21(DE3) bacteria were cultured at 37 °C in Luria-Bertani medium until OD₆₀₀ reached ~0.6. Protein expression was induced by adding Isopropyl β‑D‑1‑thiogalactopyranoside (IPTG, Macklin) to a final concentration of 0.1 mM and incubating at 16 °C for 18 h. Cells were harvested by centrifugation (6500 × g, 25 min). Cell pellets were lysed under high pressure in lysis buffer (50 mM Tris-HCl, 300 mM NaCl, pH 8.0). The clarified supernatant was loaded onto a 5 mL Ni-NTA pre-packed gravity column (Sangon Biotech) for initial purification, followed by washing with 200 mL of wash buffer (50 mM Tris-HCl, 300 mM NaCl, 20 mM imidazole, pH 8.0) to remove non-specifically bound proteins. The His-tagged protein was eluted using elution buffer (50 mM Tris-HCl, 300 mM NaCl, 250 mM imidazole, pH 8.0). The eluted protein was buffer-exchanged to an imidazole-free system (50 mM Tris-HCl, 150 mM NaCl, pH 8.0) using PD-10 columns (GE Healthcare). The 6×His-SUMO tag was removed by Ulp1 protease digestion at 8 °C for 16 h, while the 6×His tag of PBP-1 was cleaved using commercial thrombin (Macklin, Cat. No. T832140) at 20 °C for 12 h. The digested sample was reloaded onto the Ni-NTA column to separate the untagged target protein from the cleaved 6×His-SUMO tag and Ulp1 protease. Finally, the purified untagged protein was concentrated for other process and buffer-exchanged to 1×PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄ (pH 7.4)) for crystallization.

Protein refolding

Following cell disruption, the PstS proteins bind to endogenous phosphate, which interferes with subsequent functional assays. To eliminate bound phosphate, protein denaturation was performed on the Ni-NTA column prior to elution. The protein immobilized on the nickel affinity column was subjected to a stepwise urea wash using 50 mL volumes at decreasing concentrations (8 M, 6.4 M, 4.8 M, 3.2 M, 1.6 M) to unfold and refold the protein while removing bound phosphate. The column was then washed with binding buffer (50 mM Tris-HCl, 300 mM NaCl, pH 8.0) to remove residual urea. Finally, the target protein was collected using the elution buffer.

Differential scanning fluorimetry assay (DSF)

DSF assays were performed to determine the change in denaturation temperature (ΔTm) of PstS1, SphX, and their respective mutants. Two buffer systems were employed. Buffer A (20 mM Tris-HCl, pH 8.0, 150 mM NaCl) was used for ligand screening of the two phosphate-binding proteins, whereas Buffer B (0.1 M HEPES, pH 8.0) was used to assess the ΔTm induced by phosphate in refolded SphX and its mutants.

Five ligands were included in screening assays: Sodium phosphate dibasic (Sinopharm Chemical Reagent Co., Ltd, Cat. No. 10020318), Sodium phosphite dibasic (Macklin, Cat. No. S817555), β-Glycerol phosphate disodium salt (Aladdin, Cat. No. D106347), Sodium tripolyphosphate (Macklin, Cat. No. S817361), and Sodium sulfate (Sinopharm Chemical Reagent Co., Ltd, Cat. No. 10020518).

Each reaction was conducted in a final volume of 25 μL, containing either 0.2 mg/mL SphX or 0.4 mg/mL PstS1, 2 mM of the indicated ligand, and 10× SYPRO Orange dye (Sigma) in the appropriate buffer. Fluorescence signals were recorded using a real-time PCR instrument (LightCycler® 480, Roche), with the temperature increased from 20 to 95 °C at a rate of 0.01 °C/s. The resulting fluorescence data were exported in “.ixo” format and analyzed using the LightCycler® Thermal Shift Analysis software.

Protein alkylation

Since sequence analysis revealed numerous surface lysine residues, protein alkylation of lysine residues was performed to reduce protein flexibility and increase crystallization probability. The purified untagged protein was buffer-exchanged into 1×PBS (pH 7.4) and diluted to a concentration below 1 mg/mL. Freshly prepared 20 mM dimethylamine borane (DMAB, CAS: 74-94-2) and 40 mM formaldehyde (CAS: 50-00-0) were added to the protein solution, mixed thoroughly, and incubated at 4 °C for 2 h. This step was repeated once. Subsequently, 10 mM DMAB was added to the protein solution and incubated at 4 °C overnight. Finally, the protein was buffer-exchanged into 1×PBS (pH 7.4) using PD-10 columns to remove residual DMAB and formaldehyde, then concentrated for crystallization trials.

Crystallization and data collection

Following protein alkylation, PstS1 and SphX proteins were concentrated to 7.5 mg/ml using ultrafiltration. Prior to crystallization, 10 mM phosphate buffer was added, followed by a 10-min incubation. Protein crystallization was performed using sitting drop vapor diffusion at 20 °C with a commercial high-throughput screening kit and a nanoliter Mosquito robot. Optimal crystallization conditions were identified as follows: for PstS1, 0.1 M Tris-HCl pH 9.0, 0.2 M MgCl₂, and 30% PEG 3000; for SphX, 0.5 M Li₂SO₄ and 20% PEG 8000. To obtain larger crystals, further optimization of these conditions was performed for both proteins using the hanging drop method with a 1:1 protein-to-reservoir ratio. Prior to X-ray data collection, crystals were flash-cooled in liquid nitrogen using 30% (v/v) glycerol as a cryoprotectant. Data collection was performed at the BL10U2 beamline of SSRF, with diffraction data for both protein crystals collected at slightly different wavelengths as detailed in Table 1. Initial datasets were processed using XDS for further data processing³⁵.

Structure determination

Molecular replacement from the CCP4 suite³⁶ was employed to solve the phase problem using an AlphaFold model³⁷. Low-confidence residues were removed based on residue-by-residue uncertainty estimates. A preliminary model was generated using Phenix and Coot^38,39, with successive refinement performed using the Phenix refinement program and iterative manual adjustments in Coot to achieve the R-factor and R-free values shown in Table 1.

Isothermal titration calorimetry assay (ITC)

Binding affinities of PstS1, SphX, and its mutants for phosphate were measured using a MicroCal™ PEAQ-ITC (Malvern Panalytical) at 25 °C. All experiments were performed in either 0.1 M HEPES, pH 8.0 or 0.1 M sodium citrate, pH 5.0. Instrument parameters were set as follows: reference power, 10 µcal s⁻¹; stirring speed, 750 rpm; initial delay, 60 s; injection spacing, 150 s; injection duration, 4.0 s. Titrations consisted of 13 or 19 injections. Proteins were refolded from urea, dialyzed exhaustively into the respective experimental buffer, and concentrated to 10–100 µM (optimized per construct based on preliminary binding strength). Approximately 300 µL of protein solution was loaded into the sample cell. The titrant, a 1–5 mM sodium phosphate solution (matched buffer), was loaded into the syringe (~80 µL total volume). Control experiments were performed by titrating phosphate into buffer alone (single replicate per condition) to subtract dilution heats. All titrations were conducted in duplicate or triplicate.

Data were processed using Malvern PEAQ-ITC Analysis Software. Heat signals from the first injection were discarded. Integrated heats were fit to a single-set-of-sites binding model with fixed stoichiometry N = 1.0, consistent with crystallographic evidence of 1:1 protein: phosphate binding. The ligand (phosphate) concentration in the syringe was fixed, while the protein concentration in the cell was allowed to float to account for potential loss of active fraction during refolding. Thermodynamic parameters (Kₐ, ΔH, ΔS, Kₑ) are reported as mean ± standard error of the fit. Raw and fitted data were exported in CSV format for archiving and further analysis.

Circular dichroism (CD)

The secondary structure content of refolded SphX and its mutants was assessed by far-UV circular dichroism (CD) spectroscopy using a Chirascan spectrometer (Applied Photophysics). Proteins were first dialyzed into 10 mM HEPES, pH 8.0, to ensure consistent buffer composition, then diluted to 0.1 mg mL⁻¹ in the same buffer for measurement. CD spectra were recorded from 190 to 260 nm at 25 °C using a 1 mm path-length quartz cuvette. Data were collected with a step size of 1.0 nm, a time-per-point of 0.5 s, and a bandwidth of 1 nm. To improve signal quality, each sample and corresponding buffer blank was scanned three times. The replicate scans were averaged and buffer-subtracted to remove background contributions. The resulting spectra were smoothed using a Savitzky-Golay filter with a 7-point window to reduce noise while preserving spectral features. Finally, the data were converted to mean residue ellipticity ([θ], deg cm² dmol⁻¹) using the protein concentration and mean residue weight to allow direct comparison between proteins of different molecular weights. The processed spectra were exported as CSV files for further analysis.

Phylogenetic analysis

All sequences were retrieved from the NCBI GenBank database. Multiple sequence alignment was performed using MAFFT v7.525⁴⁰ with the parameters “--auto --thread 50 --anysymbol”. Maximum likelihood phylogeny was constructed using IQ-TREE v2.4.0⁴¹ with the parameters “-T 80 -m MFP -b 1000”, and the resulting tree was visualized using iTOL⁴².

Reporting summary

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

Author contributions

H.M. and X.L. conceived and designed research. Y.L. and C.W. conducted experiments. H.M., Y.L., C.W., K.Z., and Z.Z. analyzed data. D.W. and K.Z. validate the data. H.M. and Y.L. wrote the original manuscript. X.L. and H.X. revised the manuscript. All authors read and approved the manuscript.

Peer review

Peer review information

Communications Biology thanks Samujjal Bhattacharjee and the other anonymous reviewers for their contribution to the peer review of this work. Primary handling editors: Ingrid Span and Tobias Goris.

Funding

This work was supported by National Key Research and Development Program of China (Grant number 2021YFA0909700, to X.L.), National Natural Science Foundation of China (Grant number 32271358, to H.M.), the Shandong Taishan Scholarship (to H.M. and X.L.); Shandong Provincial Laboratory Project (Grant number LZR202211060073); Natural Science Foundation of Shandong Province (ZR2024QC065, to K.Z.).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. The coordinates and structure factors of PstS1 and SphX structures have been deposited in Protein Data Bank with PDB IDs 9WV8 and 9WP7, respectively. Source data for the CD, TSA, and ITC raw data are provided in Supplementary Data 1.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10481-9.