Site-specific incorporation of ¹⁹F-nuclei at protein C-terminus to probe allosteric conformational transitions of metalloproteins

Abstract

Allosteric conformational change is an important paradigm in the regulation of protein function, which is typically triggered by the binding of small cofactors, metal ions or protein partners. Here, we found those conformational transitions can be effectively monitored by ¹⁹F NMR, facilitated by a site-specific ¹⁹F incorporation strategy at the protein C-terminus using asparaginyl endopeptidase (AEP). Three case studies show that C-terminal ¹⁹F-nuclei can reveal protein dynamics not only adjacent but also distal to C-terminus, including those occurring in a hemoprotein neuroglobin (Ngb), calmodulin (CaM), and a cobalt metalloregulator (CoaR) responding to both cobalt and tetrapyrrole. In Ngb, the heme orientation disorder is affected by missense mutations that perturb backbone rigidity or surface charges close to the heme axial ligands. In CaM, the C-terminal ¹⁹F-nuclei is an ideal probe for detecting the binding states of Ca²⁺, peptides and inhibitors. Furthermore, multiple ¹⁹F-moieties were incorporated into the two domains of CoaR, revealing the intrinsically disordered C-terminal metal binding tail might be an allosteric conformational switch to maintain cobalt homeostasis and balance corrinoid biosynthesis. This study demonstrates that the AEP-based ¹⁹F-modification strategy can be applied to various targets to study allosteric regulation, especially for those biological processes modulated by the protein C-terminus.

The site-specific ¹⁹F labeling at protein C-terminus catalyzed by OaAEP1^C247A is an efficient approach to utilize ¹⁹F NMR, which is compatible with the genetic code expansion technology. Three cases studies show this approach is suitable for probing allosteric conformational transitions in metalloproteins.

Introduction

The molecular mechanisms underlying protein allostery provide fundamental insights into the understanding of protein function and the design of new functional protein scaffolds^1–5. Allosteric modulation is typically triggered by specific effectors such as the heme group in hemoproteins, transition metal ions in metalloregulators, and motifs in protein partners^6–8. Among the experimental methods used to monitor the allosteric conformational changes, nuclear magnetic resonance (NMR) is particularly suitable for highly dynamic targets that are difficult to crystallize or for small proteins that are invisible in cryo-electronic microscopy^9–12. Incorporation of ¹⁹F nuclei into protein provides a simple and efficient way to perform protein NMR spectroscopy, since ¹⁹F is highly sensitive to chemical perturbations and has no background signal in biological samples^13,14.

In the last decades, various isotope labeling approaches have been established to incorporate ¹⁹F nuclei into proteins, with two of the most popular methods being genetic code expansion (GCE) technology and modification of cysteine residues^15–18. A number of non-canonical amino acids or small molecules, such as 4-trifluoromethylphenylalanine (tfmF)¹⁹, 3,5-difluorotyrosine (F2Y)²⁰, and 2,2,2-trifluoroethanethiol²¹, have been incorporated into various proteins to study protein dynamics, including the allosteric modulation of G protein-coupled receptors²². However, these methods have several limitations. For example, the yield is sometimes very low when using the GCE approach, especially for the C-terminus ¹⁹F-labeling²³. The cysteine chemistry approach is unsuitable for modifying metalloproteins that use free cysteine to bind heavy metal ions.

In this study, we utilized an enzymatic reaction catalyzed by asparaginyl endopeptidase (AEP), a cysteine protease with dual protease-ligase activities at Asn/Asp sites, to incorporate ¹⁹F-nuclei site-specifically into the protein C-terminus. The biological function of AEP is well-established in the mammalian immune system, where it degrades protein antigens into small peptides recognized by the major histocompatibility complex²⁴. Compared to its mammalian homologs, recent studies have shown that plant AEP variants, such as OaAEP1^C247A, exhibit biased enzymatic activity²⁵. These variants function primarily as peptide ligases or transpeptidases that recognize a C-terminal NGL motif. This property has been applied in various biological fields, including protein engineering, chemical modification of proteins, cyclic peptide production, and atomic force microscopy-based single-molecule force spectroscopy^26–29.

Here, the compatibility of AEP with ¹⁹F-substituted amines was established in three metalloproteins that undergo distinct conformational transitions, including a hemoprotein neuroglobin (Ngb), a Ca²⁺-binding protein calmodulin (CaM), and a MerR family member, Co(II)-response metalloregulator (CoaR) (Fig. 1). We showed that the AEP-based approach is not only convenient and efficient for site-specific ¹⁹F-incorporation in metalloproteins, but also can provide valuable insights into the allosteric regulation of protein functions. Therefore, this approach has great potential to be applied to other protein targets, especially for those with a functional carboxyl terminus.

Fig. 1. Site-specific incorporation of ¹⁹F-nulcei at protein C-terminus by OaAEP1^C247A to study allosteric regulation of metalloproteins.

A AEP recognizes an NGL motif at protein C-terminus. Four fluorinated amines were tested for protein modification. B Three proteins of interest (POI) were selected for study, including neuroglobin that exhibits heme orientation disorder, calmodulin that undergoes an “extended” to “compact” conformational change, and a MerR family member, metalloregulator CoaR, that binds to cobalt ions and tetrapyrroles allosterically. The ¹⁹F-modification positions were indicated by red stars or triangles. The structures of human Ngb (PDB_IDs: 4MPM and 1OJ6) were shown in surface style. The models of Co(II)-CoaR complex and MerR-like DNA binding domain were generated by AlphaFold3 web server. The structure of tetrapyrrole binding domain was resolved by X-ray crystallography (PDB_ID: 9IPL, this work). The MerR-like DNA binding domain and the tetrapyrrole binding domain were color in green and marine, respectively.

Results and Discussion

To test whether AEP can utilize ¹⁹F-substituted amines as substrates, we first used the small globular protein SUMO as a model. The SUMO protein was engineered with an additional -NGLH₆ sequence at the C-terminus to enable AEP-based modification (Fig. 1A and Supplementary Fig. 1)²⁶. Three commercially available amines, 4-F-benzylamine (pfba), 4-trifluoromethyl-benzylamine (ptfba), and 3,3,3-trifluoro-propylamine (tfpa), were then selected for studies. A mixture of 100 μM protein, 10 mM amine and 1.5 μM AEP was incubated overnight at 25 °C, and the formation of products was monitored by ESI/MS. The observed molecular masses were 11352.0, 11340.5, and 11402.5 Da for the protein modified by pfba, tfpa and ptfba, respectively, which were in agreement with the calculated masses (11351.7, 11339.7 and 11401.7 Da). These results indicate that AEP can incorporate ¹⁹F-substituded amines into the protein C-terminus. Although the highest yield was ~40% for SUMO modification by ptfba, these results inspired us to apply this approach to study other proteins of interest (Fig. 1B).

Case study I: Neuroglobin

Ngb is a monomeric hemoprotein (17 kD) predominantly expressed in the neuron system of vertebrate, which plays a key role in neuroprotection (Fig. 2A)^30–33. It can bind O₂ with a high affinity and thus may function as an O₂-storage protein in retinal cells^34,35. Ngb can also scavenge reactive oxygen and nitrogen species (ROS/RNS) to help cell survival under oxidative stress^36,37. The biological function of Ngb depends on the thermodynamics and kinetics of ligand (O₂, CO, NO) binding to the heme iron. This process is regulated by an intramolecular disulfide bond and the heme orientation disorder^38–40. The latter property has been observed in 1D ¹H NMR, where two heme isomers undergo a slow chemical exchange (Supplementary Fig. 2)⁴⁰. In this case study, we aimed to modify the C-terminus of Ngb to enable efficient monitoring of its conformational states using ¹⁹F NMR, without the complex resonance assignment procedure of heme species in ¹H NMR.

Fig. 2. Conformational states of human Ngb monitored by 1D ¹⁹F NMR.

A X-ray crystal structures of the wild-type human Ngb (PDB_ID: 4MPM). The ptfba moiety attached to C-terminus of Ngb was shown in red stars. The dashed circles show the positions where the mutations are located, including C120S, A15C, R66E, and R94E. B Purification of human Ngb engineered with a short NGL motif at C-terminus. C Modification efficiency of Ngb by ptfba as shown by ESI/MS data (R, reactant; P, product). D–G The ¹⁹F NMR spectra of 100 μM WT, C120S, A15C, R94E/C120S, and R66E/C120S Ngb-ptfba samples in 50 mM potassium phosphate buffer at pH 7.0. The disulfide bond in A15C Ngb (PDB_ID: 7VQG) was shown in sticks. The polar interaction between R94 and D149 and the hydrogen bond between D63 and R66 were shown in orange dash lines as measured in the chain B of WT Ngb structure.

The human Ngb was purified with a short NGL sequence at the end of the C-terminus (Fig. 2B). The ESI/MS data show that AEP can modify >95% Ngb with the ptfba moiety (Fig. 2C and Supplementary Fig. 3). The WT Ngb-ptfba sample exhibits a major signal at δ ~ –61.8 ppm and a minor signal at δ ~ –62.0 ppm in ¹⁹F NMR spectrum (Fig. 2D). The minor peak was decreased significantly upon the addition of 2 mM reductant TCEP. It suggested the minor signal was contributed by some multimeric Ngb species due to the formation of intermolecular disulfide bonds. In addition, both C120S and A15C samples without free C120 exhibit a strong signal at δ ~ –61.8 ppm, confirming that the major signal arises from monomeric Ngb species (Fig. 2E).

Meanwhile, we observed the signals of WT and C120S Ngb monomers were composed of two conformational states (S1 and S2) with a ~ 5 Hz chemical shift difference (Δδ). It indicates that Ngb undergoes a chemical exchange process with a very slow exchange rate. As shown in Supplementary Fig. 2, the two heme orientational isomers result in a subtle chemical environment difference for the protein C-terminus. Exchange between these two orientations requires the breakage of Fe-His bonds with a reported rate of k ~ 1 s⁻¹ for ferric Ngb³⁵. As a consequence, the exchange rate of heme isomers should be slower than 1 s⁻¹, which is in agreement with our finding (k < 5 s⁻¹).

In the case of A15C Ngb, one broad ¹⁹F resonance peak was detected with a doubled linewidth (Fig. 2E and Supplementary Fig. 4). In a previous study, we found that the A15C Ngb mutant is highly thermal-stable (T_m > 100 °C) and has a 1.4-fold higher O₂ release rate, suggesting altered backbone dynamics upon cross-linking of the N-terminus and C-terminus by an engineered disulfide bond of Cys15-Cys120^41,42. This evidence indicates that the heme orientational disorder in Ngb is perturbed by the A15C mutation, presumably due to a rigid backbone conformation. The ¹H NMR spectra of C120S and A15C Ngb show that the peaks contributed by the heme group (12 to 38 ppm) differ significantly, thus confirming the perturbation of heme micro-environment (Supplementary Fig. 5).

Moreover, the resonance of C-terminal ¹⁹F-nuclei is sensitive to surface charge perturbations adjacent to the C-terminus or heme axial ligand (Fig. 2F, G). Two Arg-to-Glu missense mutation sites (R94 and R66) were selected for C120S Ngb. The residue R94 in the distal helix α_F interacts with the negatively charged D149 in helix α_H, whereas R66 in the proximal helix α_E has no direct interaction with the C-terminus region. The results showed that these two Arg-to-Glu mutants exhibited distinct behaviors in the ¹⁹F NMR spectra. The R94E mutation abolishes the R94-D149 polar interaction, causing the ¹⁹F signal to shift downfield by ~0.15 ppm compared to that of the C120S Ngb-ptfba sample. Meanwhile, a more heterogeneous chemical environment was detected. The linewidth of the conformer S1 increased from 5.6 Hz to 9.3 Hz and the Δδ between the conformers S1 and S2 increased from 5.6 Hz to 9.9 Hz (Supplementary Fig. 4). In the case of the R64E/C120S mutant, the chemical shift retained at –61.81 ppm. However, the population ratio of S1:S2 changed from ~1:1 to ~1.6:1, indicating that the R64E mutation in the proximal helix makes one conformer more preferred. The change from positive charge to negative charge at position 66 abolishes the D63-R66 hydrogen bond, thereby perturbing the dynamics of helix α_E and the equilibrium of heme isomers. Therefore, these observations show that the C-terminus ¹⁹F-modification of Ngb provides a convenient method to probe the heme micro-environment perturbations induced by missense mutations not only close to but also distal to the C-terminus.

Case study II: Calmodulin

The second target is calmodulin (CaM), a vital protein responsible for the cellular calcium signaling in eukaryotes^43,44. CaM is composed of an N-lobe and a C-lobe connected by a flexible linker⁴⁵. The binding of two Ca²⁺ ions to each lobe converts CaM to an extended, dumbbell-shaped conformation (Fig. 3A)⁴⁶. In the presence of target peptides or small inhibitors, Ca-CaM transfers from the extended conformation to the compact conformation harboring a large hydrophobic cavity⁴⁷. We found that the C-terminus is close to the pocket entrance, which suggests that the ¹⁹F moiety at C-terminus could be an ideal probe to monitor the allosteric conformational transition in CaM.

Fig. 3. Conformational transitions of CaM induced by calcium ion, peptide and inhibitor.

A The ¹⁹F NMR signal of 200 μM btfba-modified CaM sample (black line) in buffer containing 20 mM Tris (pH 7.4) and 100 mM NaCl. The spectra upon the addition of 10 mM Ca²⁺ (cyan line) or 1 mM EGTA (red line) were shown for comparison. The structures of apo-CaM (PDB_ID: 1DMO) and Ca-CaM (PDB_ID: 1CLL) were shown in cartoon style. B The binding of the RS20 peptide to CaM monitored by ¹⁹F NMR. The N-lobe and C-lobe of CaM were shown in yellow and cyan surface, respectively. The helical structure of RS20 peptide was shown in violet cartoon style (PDB_ID: 1VRK). C Titration of the inhibitor KN93 into 100 μM btfba-modified sample in buffer containing 2% DMSO. D Structural comparison of the CaM-KN93 complexes in 1:1 (PDB_ID: 9JQI, this work) and 1:3 (PDB_ID: 6M7H) stoichiometry. The KN93 molecules were depicted as stick models. The calcium ions were shown as green spheres. The 2Fo-Fc electron density maps within 5 Å of the chlorobenzene ring in KN93 contoured at 1σ were shown as blue meshes. The rotation of helix α_D was indicated by a black arrow.

To test our idea, the C-terminus of human CaM was modified with ¹⁹F-substituted amines by AEP. In this case, 3,5-Bis(trifluoromethyl)benzylamine (btfba) was used to increase signal intensity, since it contains six ¹⁹F atoms instead of three in ptfba. However, we found that the modification efficiency of CaM-NGL with either btfba or ptfba was both less than 50% at room temperature (Supplementary Fig. 6A). Half of the reactant converted into the hydrolysis products. To overcome this problem, we optimized the reaction with a lower starting CaM concentration (50 μM) at a lower temperature (in an FPLC refrigerator). The efficiency was increased to ~70% and no hydrolysis product was detected in ESI/MS (Supplementary Fig. 6B). We also tried an alternative modification strategy using a C-terminal NGLH₆ fusion tag (Supplementary Fig. 7). The unmodified CaM-NGLH₆ was separated by a cobalt resin column. Pure ¹⁹F-modified CaM was obtained using this strategy, as confirmed by SDS-PAGE and ESI/MS data. However, the overall yield was decreased from ~70 to 20%. We propose that the extended hexa-histidine tag may reduce the accessibility of the NGL motif to the catalytic center of AEP and thus decrease the modification efficiency.

The ¹⁹F NMR showed that the modification strategy successfully detected the allosteric transition upon calcium ion binding (Fig. 3A), although slight secondary structure perturbations were observed by the ¹⁹F-modification (Supplementary Fig. 8). The btfba-modified CaM produced a major peak at δ = –62.1 ppm and a minor peak at δ = –62.3 ppm in the ¹⁹F NMR spectra, which can be assigned to the apo-state (apo-CaM) and the calcium-bound state (Ca-CaM) upon the addition of excess EGTA and Ca²⁺, respectively. In addition, the CaM-btfba sample responded to a target peptide (Fig. 3B). The RS20 peptide (RRKWQKTGHAVRAIGRLSSS) is derived from the CaM binding domain of smooth-muscle myosin light chain kinase. It binds to CaM with 1:1 stoichiometry in calcium-dependent manner^48,49. Our ¹⁹F NMR data are consistent with previous findings. The addition of an equal amount of RS20 caused a complete loss of Ca-CaM species. Simultaneously, two weak signals appeared at δ = –61.92 ppm and δ = –62.05 ppm. To investigate whether these two signals resulted from a hindered benzene ring flip or a slow conformational change, we used ptfba-modified CaM to repeat the experiment. As shown in Supplementary Fig. 9, the signal of Ca-CaM-ptfba at δ = –60.27 ppm gradually disappeared upon the titration of RS20 peptide, resembling the behavior of Ca-CaM-btfba (Fig. 3B). Simultaneously, a major peak at δ = –60.61 ppm and a minor peak at δ = –60.55 ppm appeared in the spectra upon RS20 binding. Thus, the data suggest the presence of two peptide-bound CaM species that undergo a slow conformational change.

We further applied this strategy to monitor the binding of an inhibitor to CaM. KN93, a widely used CaMKII inhibitor, has been shown in recent studies to bind directly to CaM, thereby suppressing CaMKII activity^50–52. A previous protein crystallography study showed that CaM can bind up to three KN93 molecules (Supplementary Fig. 10A)⁵². Meanwhile, the isothermal titration calorimetry (ITC) and surface plasmon resonance (SPR) experiments suggested a binding stoichiometry of n = 2 and 4, respectively⁵¹. In addition, the stoichiometry was difficult to assess by the conventional ¹H-¹⁵N HSQC NMR data due to the two different exchange regimes⁵¹. Here, we observed that the ¹⁹F signal of Ca-CaM-btfba at –62.38 ppm gradually disappeared by titrating KN93 in 1:1 stoichiometry (Fig. 3C). Meanwhile, a broad peak appeared at –62.20 ppm, suggesting the formation of CaM-KN93 complex. Titration of two equivalents of KN93 made few changes in the ¹⁹F spectrum. Hence, our data indicate that the binding of one KN93 molecule is sufficient to transform the CaM from the extended conformational state to the compact state, although the large hydrophobic pocket may accommodate additional inhibitors. This finding was further supported by the structure prediction of the RoseTTAFold All-ATOM program and the molecular dynamics simulations, both of which showed that a single KN93 molecule fits well into the pocket (Supplementary Fig. 10B, C).

The NMR and simulation results prompted us to co-crystalize CaM and KN93 under 1:1 stoichiometry (Table 1, Supplementary Fig. 10D). The crystal structure confirmed that in the 1:1 complex, the protein backbone of CaM adopts a compact conformation similar to that in the 1:3 complex (Fig. 3D). But the helix α_D rotates ~15° due to a different ligand conformation. The KN93 adopts two distinct orientations in the 1:1 complex. A clear electron density map of the planar chlorobenzene ring in the KN93 molecule was observed, which was surrounded by several hydrophobic residues forming a cavity in either the N-lobe or the C-lobe. The chlorobenzene ring fitted well into these two cavities (Supplementary Fig. 10E). However, the electron density of the remaining part of KN93 is remarkably low, indicating a flexible ligand conformation within the large binding pocket of CaM.

Table 1.

Data collection and refinement statistics

	TBD (PDB_ID: 9IPL)	CaM-KN93 (PDB_ID: 9JQI)
Data collection
Space group	P61	C121
Cell dimensions
a, b, c (Å)	96.039, 96.039, 148.723	69.336, 40.264, 173.41
α, β, γ (°)	90.0, 90.0, 120.0	90.0, 92.5, 90.0
Resolution (Å)	49.57, −2.28 (2.34–2.28)	57.75–2.10 (2.21–2.10)
Rmerge	0.095 (1.755)	0.158 (0.689)
I /σ (I)	20.5 (2.3)	7.6 (2.8)
Completeness (%)	100.0 (100.0)	99.9 (100.0)
Redundancy	20.1 (19.8)	6.2 (6.5)
Refinement
Resolution (Å)	45.74–2.28 (2.34–2.28)	57.75–2.10 (2.15–2.10)
No. reflections	33586 (2463)	28410 (1824)
Rwork / Rfree	0.2088/0.2297 (0.315/0.274)	0.2484/0.2882 (0.298/0.343)
No. atoms
Protein	2943	3303
Waters	108	91
Ligand/ion	40	114
B-factors
Protein	55.17	39.37
Waters	52.19	31.88
Ligand/ion	90.48	70.45
R.m.s. deviations
Bond lengths (Å)	0.005	0.002
Bond angles (°)	1.270	0.504

Statistics for the highest-resolution shell are shown in parentheses.

Taken together, the case study of CaM shows that the C-terminus ¹⁹F-modification can be used to monitor the allosteric binding process of inorganic ions, small organic molecules, and peptides. It also provides a valuable tool for accessing the binding stoichiometry of POIs and small effectors.

Case study III: A metalloregulator responds to both cobalt and tetrapyrrole

The third POI is CoaR, a MerR family metalloregulator found only in cyanobacteria, a photosynthetic organism that plays an important role in global carbon dioxide fixation^53,54. CoaR has a unique feature among the cobalt metalloregulators. It can enhance the metabolic cross-talk between the cobalamin biosynthesis pathway and the cobalt homeostasis regulation pathway by sensing both cobalt ions and corrinoid. This dual sensing capability helps balance the metabolic flux of corrins and cobalt ions (Fig. 4A)^53,55. The sequence analysis showed that CoaR is composed of a MerR-like DNA binding domain (MLD) and a CobH-like tetrapyrrole binding domain (TBD) connected by a flexible linker (Fig. 4B). The C-terminus region (referred as C-tail) is highly conserved and contains a CHC metal binding motif. Three cysteines (C121, C363 and C365) have been identified as cobalt-binding ligands^53,56. However, the other ligands forming cobalt coordination shell are unclear and the allosteric regulation mechanism triggered by cobalt and corrins remains elusive. In this case study, we applied ¹⁹F NMR based on AEP-catalyzed C-terminus modification and other biophysical techniques to investigate the cobalt ion and tetrapyrrole induced protein allostery in CoaR.

Fig. 4. Assembly of the cobalt binding site in CoaR monitored by ¹⁹F NMR and UV-vis spectroscopy.

A Proposed mechanism of CoaR mediated cross-talk between corrin biosynthesis pathway and cobalt homeostasis regulation pathway. The side chains in tetrapyrroles are omitted for clarity. B The domain organization and ¹⁹F-modification strategy of CoaR. The model structure was generated by Alphafold3 web server. The C-terminus tail (C-tail) with a strictly conserved CHC metal binding motif was shown in purple. C Monitoring metal binding state based on the paramagnetic relaxation enhancement (PRE) effect of Co(II) ion. The ¹⁹F NMR spectra of the mixture sample containing 50 μM MLD-btfba and TBD-tfpa were recorded in the presence or absence of Co(II) ion. D The potential cobalt binding site predicted by Alphafold3. The orientation of C-tail in Co(II) “on” conformation was illustrated as purple dash lines. E The UV-vis spectra of MLD + TBD (red line) or MLD + TBD^C363S/C365S (black line) mixture sample upon loading 50 μM Co(II). F ¹⁹F NMR spectra of the mixture samples containing 50 μM MLD-btfba and unlabeled TBD^C363S/C365S in the absence (black lines) or presence (red lines) of one equivalent Co(II).

The full-length CoaR is insoluble when over-expressed in E. coli., hence we expressed the two isolated domains by protein engineering (Supplementary Fig. 11)⁵⁷. The C-terminus of MLD and TBD were both engineered with a short NGL sequence at non-conserved positions to enable AEP modification. In this case, we applied two different ¹⁹F-substituted amines as substrates. The MLD was modified with btfba (δ ~ −62.2 ppm), while the TBD was modified with tfpa (δ ~ −64.8 ppm). The different chemical shifts of btfba and tfpa make the assignment of resonances much easier. The ESI/MS results confirmed that AEP successfully incorporated these amines into MLD (Supplementary Fig. 12) or TBD (Supplementary Fig. 13).

The metal binding process was monitored by ¹⁹F NMR. Signal intensity decreases due to the strong paramagnetic relaxation enhancement (PRE) effect when a Co(II) ion occupies an adjacent binding site^58,59. As shown in Fig. 4C, we found that the ¹⁹F resonance of MLD-btfba changed slightly in the presence of Co(II) ion. It indicates that MLD cannot bind Co(II) alone, despite having a potential cobalt-binding ligand C121. Meanwhile, for TBD-tfpa sample, the signal intensity decreased by ~50% after adding one equivalent Co(II) ion, which is due to cobalt binding to the CHC motif at C-terminus. Upon mixing MLD-btfba and TBD-tfpa with 1:1 stoichiometry, the chemical shifts of btfba and tfpa were unperturbed. This observation indicates that the interaction between MLD and TBD does not affect the micro-environment of those ¹⁹F-labled regions in the apo-form. By further titrating Co(II) into the mixture sample, we observed a dramatic change of the signal contributed by MLD-btfba. It disappeared completely in the presence of one equivalent Co(II) ion, which strongly suggests the formation of a Co(II) site at the interface between MLD and TBD.

We verified this result using the GCE approach (Supplementary Fig. 14)¹⁹. The codon of Phe65 was replaced by the amber codon TAG to enable the incorporation of the unnatural amino acid tfmF. The ¹⁹F-labeled MLD^tfmF65 sample was purified successfully with high purity, albeit with a lower yield. As expected, the signal of tfmF65 disappeared when one equivalent Co(II) ion was added to the mixture of MLD^tfmF65 + TBD. This is due to the strong PRE effect of Co(II), since the tfmF65 is close to C121 where Co(II) binds. These data show that the modification strategy based on AEP is in agreement with those popular methods.

As predicted by Alphafold3⁶⁰, the first coordination shell is composed of D118 and C121 in MLD and H160 and H318’ in TBD (Fig. 4D). These four residues are arranged in a tetrahedral geometry suitable for Co(II) binding. Although C363 and C365 in the C-tail are distant from the cobalt binding site in the predicted structure, UV-Vis spectroscopy studies indicate that the C-tail binds to cobalt directly. As shown in Fig. 4E, the titration of Co(II) into the MLD + TBD mixture resulted in two specific bands at ~305 nm and ~575 nm. These bands are contributed by the ligand to metal charge transfer (LMCT) of Co-S bond and the d–d transition of Co(II) ion, respectively. Meanwhile, control studies showed that titrating Co(II) into MLD or TBD alone did not produce specific absorption peaks in the visible region (Supplementary Fig. 15).

We also found that the Cys-to-Ser mutation at positions 363 and 365 abolished the cobalt binding capacity of the C-tail. Consequently, the absorbance of the Co-S LMCT bands at ~305 nm decreased by ~2/3 in the Co(II)-loaded MLD + TBD^C363S/C365S sample (Fig. 4E), consistent with the loss of two cysteine ligands. In contrast, the d–d transition band at ~575 nm changed slightly, which suggests that a specific cobalt binding site may still exist without the coordination of C363/C365. Based on these observations, we propose that CoaR can fold into a C-tail Co(II) “on” or “off” conformation. In addition, a subtle structural perturbation in the “off” state was observed in the ¹⁹F NMR spectra (Fig. 4F). The signal of MLD-btfba shifted slightly from −62.25 ppm to −62.14 ppm and the intensity was less reduced upon the addition of one equivalent Co(II) ion into MLD-btfba+TBD^C363S/C365S mixture.

Furthermore, we showed the AEP-based protein engineering approach can be applied to prepare functional full-length CoaR and incorporate multiple ¹⁹F-labels in combination with the GCE approach. The purified MLD and TBD were successfully ligated by AEP, yielding a product with a target molecular mass of ~40 kD (Fig. 5A). ESMA assay showed that the full-length CoaR binds to the cognate coaT promoter with an enhanced DNA binding affinity compared to MLD (Fig. 5B and Supplementary Fig. 12B). It suggests an optimized conformation for the DNA binding domain upon ligation with the tetrapyrrole binding domain, making it more suitable for promoter recognition.

Fig. 5. CoaR binds to cobalt ion in nanomolar affinity and responses to metallo-tetrapyrrole.

A Purification of full-length CoaR monitored by SDS-PAGE. B ESMA assay of 10-50 pmol CoaR dimer binding to 10 pmol 26 bp or 36 bp coaT promoter. C NMR spectra of 40 μM dual ¹⁹F-labeled CoaR upon titration of cobalt(II) ion. The sample was prepared by using MLD^tfmF65 and TBD-tfpa as reactants. D Titration of Co(II) ion to 8 μM full-length CoaR monitored by the intrinsic Trp fluorescence excited at 285 nm in the presence and absence of 36 μM EGTA. The normalized intensity (I/I₀) at 340 nm was fitted to a one site competitive binding model using DynaFit program. E UV-vis spectra of 70 μM CoaR and 90 μM EGTA mixture upon titration of 20–200 μM Co(II) ion. F The difference spectra of 8 μM TBD upon titration of 1–10 μM hemin in 50 mM potassium phosphate buffer at pH 7.0. G ¹⁹F NMR spectra of 50 μM TBD-tfpa in the absence or presence of a non-metal tetrapyrrole HBA or a metallo-tetrapyrrole heme with 1:1 stoichiometry in 50 mM potassium phosphate buffer at pH 7.0. H Proposed model of Co(II) ion and cobalt corrinoids regulated protein allosteric conformational transition in CoaR based on the ¹⁹F NMR and spectroscopic data. The intrinsic disordered C-tail switches “on” or “off” to the cobalt binding site depending on the concentration of Co(II) ion and cobalt corrinoids which further regulate the transcriptional level of the Co(II) efflux pump CoaT.

The cobalt binding process was also monitored by ¹⁹F NMR, fluorescence and UV-Vis spectroscopy. The dual ¹⁹F-labeled CoaR produces two separate peaks at −62.1 and −64.8 ppm in the spectrum (Fig. 5C). The signal of tfmF65 decreases in intensity with a broader linewidth in dual ¹⁹F-CoaR compared to MLD^tfmF65, indicating a more anisotropic chemical environment. In contrast to tfmF65, the signal of tfpa in the C-terminus is unchanged due to its flexible conformation. Upon titration of Co(II) ions, both these signals were decreased as a result of the PRE effect. The dissociation constant was fitted to K_d = 1.2( ± 0.1) × 10⁻¹⁰ M by using fluorescence titration in the presence of the competitive ligand EGTA at pH 8.0 (Fig. 5D and Supplementary Fig. 16). The intrinsic fluorescence intensity of tryptophan decreases upon binding of Co(II) ions. The UV-vis spectra showed that two specific peaks appeared at 535 and 579 nm which contributed by the Co(II) d–d transition (Fig. 5E).

In addition to the above spectroscopic data, we obtained a single crystal of TBD in the apo-form (Table 1, Supplementary Fig. 17). The structure of residues 161-358 was resolved. The C-tail region is intrinsically disordered, which suggests it undergoes a large conformational change upon cobalt ion binding. The overall architecture of apo-TBD is similar to that of precorrin-8x isomerase, CobH⁶¹. It is a homodimer and each protomer is composed of eleven α-helices and six β-sheets. However, we found a distinct structural feature for TBD, where the tetrapyrrole binding pocket is “closed” in contrary to CobH. The residues from H253 to F264 are buried in the pocket, folding into helix α7 and a short loop. A series of hydrogen bonds and hydrophobic contacts are formed to stabilize this conformation. The interaction interface consists of several highly conserved residues, including A191, H253, I254, A260, F264, W270 and S290. The structure also suggests that to “open” the pocket, the helix α7 should move out of the tetrapyrrole binding pocket and stack with helix α8, as observed in hydrogenobyrinic acid (HBA) bound CobH (Supplementary Fig. 17D). This structural information indicates that helix α7 may mediate self-inhibition in TBD, preventing the binding of certain corrinoids such as HBA. As shown in Supplementary Fig. 18, the spectrum of HBA was not affected by TBD, whereas it was shifted upon the binding of CobH and CobE⁶².

Although it is still unclear to which corrinoid CoaR responses, we found that it binds weakly to a natural metallo-tetrapyrroles cofactor, heme, as suggested by the typical UV-Vis spectroscopic feature of hemoproteins (Fig. 5F)⁶³. The ferric form of hemin-TBD complex exhibits a Soret band at 411 nm and a Q-band at 530 nm, while the spectrum of the ferrous form shifts to 422 nm for the Soret band and 558 nm for the Q band, respectively. Moreover, the ¹⁹F NMR results suggest that the intrinsically disordered C-tail contributes to metallo-tetrapyrrole binding (Fig. 5G). The signal of the ¹⁹F-modified TBD-tfpa sample barely changed upon the addition of HBA, but decreased by half when heme bound to the protein. The structural comparison also shows that the C-tail is spatially accessible to tetrapyrrole, which may fold into the binding pocket on the solvent-exposed side (Supplementary Fig. 19A). Another axial ligand might be the residue Y287 located at the axial position of tetrapyrrole molecule. This position in CobH is replaced by a conserved Asn residue, which cannot coordinate to the iron in heme or cobalt in cobalt-corrinoids. We designed a Y287H mutant to test whether Y287 contributes to heme binding, since histidine is a typical axial ligand for heme peroxidases (Supplementary Fig. 19B, C). By using ABTS as a substrate, the stopped-flow spectrometry data showed that the peroxidase activity of heme-TBD^Y287H increased by ~5-fold, as suggested by the absorbance of ABTS^·+ radical cation at 646 nm.

Based on the above results, we proposed an allosteric regulation model for CoaR (Fig. 5H). When cobalt ions are excess in cells, CoaR binds to them with nanomolar affinity. The cobalt ligands distributed in MLD, TBD and C-tail move close together to form a specific cobalt binding site, which induces an allosteric conformational change of the DNA binding domain and coaT operator-promoter⁶⁴. The RNA polymerase is recruited and subsequently transcripts the downstream P-type ATPase CoaT to efflux excess cobalt ions. However, this process is negatively regulated by metallo-tetrapyrroles, such as cobalt-corrinoid, because of the demand for cobalt ions in cobalamin biosynthesis. When cobalt-corrinoid binds to the pocket, the “closed-open” conformational change could cause the C-tail “off” to the cobalt binding site, thereby reducing the transcription activity. During this process, the complex loses two Co(II)-S bonds at the cobalt binding site, whereas two additional coordination bonds are formed at the axial positions of metallo-tetrapyrroles.

Taken together, our data suggest the potential role of the highly conserved C-tail in CoaR, which switches between the cobalt binding site and tetrapyrrole binding pocket, depending on the concentration of Co(II) ions and cobalt corrinoids. The balance between cobalt homeostasis and cobalamin biosynthesis in cyanobacteria also provides new insights into how nature controls the biosynthesis of other metallo-tetrapyrroles, including iron-protoporphyrin, magnesium-chlorophyll, and nickel-F430⁶⁵.

Conclusion

In this study, we report a site-specific ¹⁹F-incorporation approach at the protein C-terminus to probe the allosteric conformational transitions, based on the AEP-catalyzed reaction. This approach is efficient and particularly suitable for modifying metalloproteins with functional cysteines, with modification efficiencies ranging from ~20 to 95% for different protein scaffolds. We also found that lowering the reaction temperature can reduce hydrolysis for some POIs. The AEP-based strategy is also compatible with the GCE technology. Multiple ¹⁹F moieties with a large chemical shift can be incorporated into a protein simultaneously to study protein dynamics using advanced NMR techniques, such as ¹⁹F − ¹⁹F TOCSY⁶⁶. Therefore, this approach has great potential for application to other POIs or functional peptides⁶⁷, especially for those targets with a functional C-terminus.

The applications to three metalloproteins (Ngb, CaM and CoaR) also provide valuable insights that would otherwise be difficult to obtain using other conventional techniques such as X-ray crystallography. (1) The application to human Ngb revealed that the dynamics of protein C-terminus is related to the heme orientation disorder. It can be applied to monitor the heme micro-environment perturbation by missense mutations both close (R94E) and distal (A15C and R66E) to the C-terminus without assignments of ¹H NMR resonances. (2) In the case of CaM, the binding states of calcium ions, target peptides or inhibitors can be effectively detected by the C-terminus labeled ¹⁹F-nuclei. The stoichiometry of CaM and an inhibitor was re-evaluated. Our data showed that the binding of one KN93 molecule is sufficient to transform the conformational state of CaM, although the previous crystal structure shows that CaM can accommodate up to three KN93 molecules. It suggests that this method may have potential application in small-molecule drug discovery^68,69. (3) Moreover, the assembly of the cobalt-binding site can be monitored by introducing ¹⁹F-nuclei in the two domains of CoaR. A strong Co(II)-based PRE was detected in the presence of both domains. The data further suggest that the conserved metal-binding tail at the C-terminus may be an allosteric conformational switch that transits between the cobalt-binding site and the tetrapyrrole-binding pocket. The novel regulation mechanism of CoaR could deepen the understanding of the structural dynamics and functional relationships of proteins not only for metal homeostasis but also for the tight control of tetrapyrrole biosynthesis, a class of molecules essential for all life on earth.

Methods

Vector construction, expression and purification of proteins

The codon-optimized genes encoding protein of interests (POIs) were synthesized by Genescript company, including Saccharomyces cerevisiae Ubiquitin-like protein SMT3 (SUMO), human neuroglobin (Ngb), human calmodulin (CaM), the MerR-like domain (MLD) and tetrapyrrole binding domain (TBD) of Synechocystis sp. PCC 6803 CoaR. These genes were subcloned into pET vector (Novagen) between the NcoI and XhoI sites (Ngb, MLD, TBD) or the NdeI and XhoI sites (SUMO, CaM) using standard molecular cloning techniques. The site-directed mutations were generated using Mut Express II Fast Mutagenesis Kit (Vazyme). The amino acid and nucleotide sequences are listed in the Supplementary text.

These vectors were transformed to E. coli BL21(DE3) strain and selected by 50 μg/mL kanamycin (for SUMO, CaM, TBD) or 100 μg/mL ampicillin (for Ngb, MLD). The clones were grown in lysogeny broth (LB) media containing the appropriate antibiotics at 37 °C overnight. For Ngb expression, 0.5 mL overnight culture was transferred to 500 mL LB supplemented with 1 mM 5-aminolevulinic acid and 100 μg/mL ampicillin. The culture was grown at 37 °C and shaken at 220 rpm for 12 h. Protein expression was induced with 0.4 mM IPTG for additional 12 h at 28 °C. For SUMO and CaM expression, 5 mL overnight culture was transferred to 500 mL LB supplemented with 50 μg/mL kanamycin. The culture was grown at 37 °C until OD_{600 nm} reached 0.6–0.8. Protein expression was induced with 0.4 mM IPTG for additional 6 h at 25 °C. For MLD and TBD expression, 5 mL overnight culture was transferred to 500 mL LB supplemented with antibiotics. The culture was grown at 37 °C until OD_600 nm reached 0.6–0.8 and then cooled to 16 °C for an hour. Protein expression was induced with 0.4 mM IPTG at 16 °C overnight. The 4-(trifluoromethyl)-L-phenylalanine (tfmF) incorporated MLD was expressed following reference²³. Cells co-transferred with pET-MLD^TAG65 and pDule-tfmF plasmids were grown in M9 minimal media containing 1 mM tfmF and 0.2% L-arabinose at 37 °C until OD_600 nm reached 0.4. Protein expression was induced with 0.4 mM IPTG overnight.

The SUMO, Ngb and CaM were purified following a similar scheme. The cell pellet from 1 L culture was resuspended in 20 mL Ni-NTA buffer A (10 mM Tris-HCl, pH 8.0, 100 mM NaCl) and then lysed by sonication on ice. The supernatant was loaded onto a Ni-NTA column (Cytiva) equilibrated with buffer A and then washed by 5% buffer B (buffer A + 0.5 M imidazole). The POIs were eluted by a linear gradient from 5 to 50% buffer B. For Ngb and CaM, HRV3C protease was added to cleavage the N-terminus His-tag. The sample was further loaded onto a Q-HP (Cytiva) column equilibrated with buffer C (10 mM Tris-HCl, pH 8.0). POIs were eluted by a linear gradient from 10 to 50% buffer D (buffer C + 1 M NaCl). For Ngb, fractions with A_{413 nm}/A_{280 nm} > 2.8 were pooled. The purity of other colorless POIs was estimated by SDS-PAGE.

Purifications of MLD and TBD were following a five steps scheme (Ni-NTA, Q-HP, protease cleavage, Ni-NTA, Q-HP). The first two steps were the same as the above-mentioned procedure. A 0.5 mM tris(2-carboxyethyl)phosphine (TCEP) was added into buffer to prevent cysteine oxidization. After protease cleavage, a final concentration of 50 mM imidazole was added into the sample containing MLD. The sample was loaded onto a Ni-NTA column equilibrated with 10% buffer B to remove the enzyme and His-SUMO-tag. The fractions pass through the column were collected and then diluted 2-fold by adding buffer C. The sample was further loaded onto a Q-HP column equilibrated by 15% buffer D. MLD was eluted by a linear gradient from 15 to 30% buffer D. For TBD purification, the sample after protease cleavage was loaded onto a Ni-NTA column equilibrated with buffer A. TBD was eluted by a linear gradient from 0 to 15% buffer B and further purified by a Q-HP column by a linear gradient from 20 to 40% buffer D.

Preparation of activated asparaginyl endopeptidase (AEP)

The gene of OaAEP1^C247A (residues 24-474) was synthesized with a N-terminal 6×His-Ub fusion tag, as described in reference²⁵. Enzyme expression was conducted in E. coli T7 Shuffle strain by inducing with 0.1 mM IPTG at 16 °C overnight. Cell pellets were lysed in buffer containing 10 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.05% (w/v) CHAPS, 10% (v/v) glycerol, 0.5 mM TCEP. The supernatant was loaded to a Ni-NTA column and then washed by 5% buffer B. Fractions containing the proenzyme were eluted by a linear gradient from 5 to 50% buffer B. The sample was further purified using a Q-HP column (Cytiva) and eluted by a linear gradient from 10 to 40% buffer D. The purified proenzyme was concentrated to at least 40 μM using ultracentrifugation.

To activate AEP, the concentrated proenzyme was diluted to 2 μM in buffer composed of 50 mM sodium acetate (pH 4.0), 50 mM NaCl, 1 mM TCEP, and 1 mM EDTA, followed by incubation at 16 °C overnight. The activated enzyme was concentrated and stored at −80 °C with the addition of 10% (v/v) glycerol.

Preparation of C-terminal ¹⁹F-modified protein samples

The purified POIs were mixed with amines (10 mM for tfpa and ptfba, 2.5 mM for btfba) and 1.5 μM activated AEP in 100 mM HEPE buffer at pH 7.5. The reactants were incubated in a 25 °C water bash or a 4 °C FPLC refrigerator overnight. The ¹⁹F-modified POIs were separated from the enzyme and excess amines by a Q-HP anion exchange column. The ¹⁹F labeled proteins were concentrated to 50–200 μM concentration depending on the sample yield.

Preparation of full-length CoaR

Purified MLD and TBD were diluted to 50 μM and then mixed with 1.5 μM AEP. The sample were incubated in a 25 °C water bash for 2 h. The mixture was loaded onto a 5 mL Heparin column (Cytiva) equilibrated with 20% buffer D. Full-length CoaR was eluted by a linear gradient from 20 to 60% buffer D. Fractions containing CoaR was concentrated and further purified using a Superdex 200 10/300 GL column (Cytiva) equilibrated with buffer E containing 10 mM Tris (pH 8.0), 200 mM NaCl, 0.5 mM TCEP.

Nuclear magnetic resonance spectroscopy

The NMR spectra were acquired on a Bruker Avance 500 MHz spectrometer equipped with a 5 mm broad-band tunable probe. The protein samples were added 10% D₂O before data collection. The ¹⁹F NMR data were recorded with 1 s relaxation delay and 1024 scans. Chemical shifts were referenced to external trifluoroacetic acid (−76.5 ppm). The FID was multiplied with an exponential function with line-broadening factor of 10 Hz prior to Fourier transformation unless otherwise noted. The 1D ¹H NMR data of C120S and A15C Ngb-NGL were acquired using the zgespg pulse program to suppress the water signal.

Electrospray ionization mass spectrometry

Protein mass spectra were carried out on the G2-XS QToF mass spectrometer (Waters). The desalted samples were mixed with 1% formic acid and then transferred into the mass spectrometer chamber under positive mode. The multiple m/z peaks were transformed to molecular weight by using MaxEnt1 software.

Crystallizations

The sitting drop vapor diffusion method was employed for the crystallization. The apo-TBD sample was concentrated to ~10 mg/mL concentration. The crystal of TBD^C363S/C365S (residues 154-369) grew rapidly in reservoir solution containing 1.0 M (NH₄)₂SO₄, 0.1 M Bis-Tris (pH 5.5), 2% w/v PEG3350 at 289 K. The selenomethionine substituted CaM-NGL was desalted to buffer containing 20 mM Tris-HCl (pH 7.4), 100 mM NaCl and 5 mM CaCl₂. The sample was further mixed with KN93 in 1:1 stoichiometry and then concentrated to ~20 mg/mL. The crystal of CaM-KN93 complex grew in 0.2 M NaAc, 0.1 M sodium cacodylate (pH 6.5), 28% PEG 8000 at 277 K. Crystals were cryo-protected by soaking in reservoir solution supplemented with 20% v/v glycerol and flash-frozen in liquid nitrogen.

Data collection and structural determination

The X-ray diffraction data were collected in the Shanghai Synchrotron Radiation Facility (SSRF) beamlines BL02U1 and BL19U1 under 100 K with a wavelength of 0.9788 Å^70,71. The data were integrated and scaled using xia2/DIALS^72,73. The crystal structures were resolved by molecular replacement method using Phaser in CCP4 suite^74,75. The alphafold2 predicted structure and the crystal structure of CaM-KN93 (1:3 complex, PDB_ID 6M7H) were used as the searching model for TBD and CaM-KN93 (1:1 complex), respectively. Model refinement was carried out by using refmac5, Phenix and COOT^76–78. The data collection and refinement statistics are shown in the Table 1.

Electrophoretic mobility shift assay

The oligonucleotides were synthesized by Genescript Company. DNA annealing was performed in a buffer of 10 mM Tris-HCl at pH 7.4, 50 mM NaCl, and 5 mM MgCl₂ by heating the mixture at 95 °C for 5 min and slowly cooling to room temperature overnight. A total of 10 pmol DNA duplexes was mixed with various concentrations of protein in a 20 μL system containing EMSA buffer (10 mM Tris-HCl at pH 8.0, 50 mM KCl, 5 mM MgCl₂). The samples were incubated at room temperature for 20 min and then loaded onto an 8% native polyacrylamide gel containing EMSA buffer and 10% v/v glycerol. The gel was electrophoresed in 0.5×TB buffer (50 mM Tris, 41.5 mM borate at pH 8.0) at 85 V at room temperature and stained by Gel-Red.

UV-Vis spectroscopy

The UV-Vis spectra were collected on PerkinElmer Lambda 365 UV/Vis spectrophotometer at room temperature. Purified MLD and TBD were mixed up with 1:1 stereometry and then titrated by CoCl₂ at room temperature. The heme titration experiments to TBD were performed in 50 mM potassium phosphate buffer at pH 7.0. The reduced heme-protein samples were prepared by adding a few of sodium bisulfate power.

Fluorescence spectroscopy

Purified full-length CoaR was diluted to 8 μM with buffer E in a 1 mL cuvette. The intrinsic tryptophan fluorescence spectra were excited at 285 nm and recorded from 300 nm to 440 nm at room temperature using PerkinElmer LS45 fluorescence spectrometer. The sample was titrated with various concentration of CoCl₂ in the absence or presence of EGTA. The data was fitted to a competitive model with one cobalt binding site by using Dynafit⁷⁹. The conditional binding constant was calculated to K’ = 1.1 × 10¹⁰M^-1 for Co(II) and EGTA (pK_a1 = 1.9, pK_a2 = 2.70, pK_a3 = 8.79, pK_a4 = 9.40, lgK_Co(II)-EGTA = 12.3) at pH 8.0⁸⁰.

Stopped-flow spectrometry

The peroxidase activities of heme bound TBD were measured using a dual mixing stopped-flow spectrophotometer (SF-61 DX2 Hi-Tech KinetAsyst^TM) at room temperature. A 100 μM ABTS and 100 mM H₂O₂ were used as the substrate and oxidant, respectively. The protein sample was desalted to 50 mM potassium phosphate buffer at pH 7.0 prior to use. The kinetic data were collected upon mixing an equal volume of the oxidant and the protein/substrate mixture. The data were collected from 350 to 700 nm in 10 s.

Protein structure prediction

The KN93-CaM complex model was generated by RoseTTAFold All-Atom program⁸¹. The Alphafold2 (version 2.3.1) and Alphafold3 programs were applied to predict the model of CoaR and its domains^60,82. The most confident models of target sequences were visualized and analyzed by using PyMOL. The model of heme-TBD^Y287H model was obtained by using the Rosetta program and the protocol in reference⁸³. The open conformation model of TBD was used as starting model which was generated by the SWISS-MODEL web server⁸⁴.

Molecular dynamic simulations

The starting model of CaM-KN93 complex was modified from the crystal structure (PDB_ID: 6M7H) by deleting two KN93 molecules less buried in the pocket. The system was simulated by using Desmond 2022-4 academic version under OPLS_2005 force field with default parameters⁸⁵. An orthorhombic box was built using explicit solvation with TIP3P waters. The salt concentration was set to 0.15 M NaCl. Then it was followed by a 100 ps minimization. The simulation was run in NPT ensemble at 300 K for 200 ns with a 2 fs time step. The coordinates were sampled for every 100 ps. The RMSD and radius of gyration of the system were analyzed using Desmond.

Circular dichroism spectroscopy

The circular dichroism spectra in the far-UV (190–250 nm) region were recorded using a JASCO 1500 instrument. Samples were desalted in 5 mM NH₄Ac buffer prior to use. A 1.0 cm cuvette was used for data collection. The scanning speed was set to 100 nm/min and the number of accumulations was 3.

Reporting summary

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

Author contributions

X. Liu, P. Guo, and Q. Yu preformed the most of the studies; S.-Q. Gao performed ESI-MS studies, X. Liu and H. Yuan performed X-ray crystallographic studies; X. Liu and Y.-W. Lin wrote the manuscript. X. Tan. and Y.-W. Lin, supervising, reviewing, and editing the manuscript. All authors discussed the results.

Peer review

Peer review information

Communications Biology thanks Robert Prosser, Christopher MacRaild and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Tuan Anh Nguyen and Mengtan Xing.

Data availability

The atomic coordinates and structure factors have been deposited to the Protein Data Bank (PDB) with accession codes 9IPL and 9JQI.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-024-07331-x.