Calcium Binding to the Innate Immune Protein Human Calprotectin Revealed by Integrated Mass Spectrometry

Abstract

Although knowledge of the coordination chemistry and metal-withholding function of the innate immune protein human calprotectin (hCP) has broadened in recent years, understanding of its Ca²⁺-binding properties in solution remains incomplete. In particular, the molecular basis by which Ca²⁺ binding affects structure and enhances the functional properties of this remarkable transition-metal-sequestering protein has remained enigmatic. To achieve a molecular picture of how Ca²⁺ binding triggers hCP oligomerization, increases protease stability, and enhances antimicrobial activity, we implemented a new integrated mass spectrometry (MS)-based approach that can be readily generalized to study other protein-metal and protein-ligand interactions. Three MS-based methods, HDX-MS, PLIMSTEX and native MS, provided a comprehensive analysis of Ca²⁺ binding and oligomerization to hCP without modifying the protein in any way. Integration of these methods allowed us to (i) observe the four regions of hCP that serve as Ca²⁺ binding sites, (ii) determine the binding stoichiometry to be four Ca²⁺ per CP heterodimer and eight Ca²⁺ per CP heterotetramer, (iii) establish the protein-to-Ca²⁺ molar ratio that causes the dimer-to-tetramer transition, and (iv) calculate the binding affinities associated with the four Ca²⁺-binding sites per heterodimer. These quantitative results support a model in which hCP exists in its heterodimeric form and is at most half-bound to Ca²⁺ in the cytoplasm of resting cells. With release into the extracellular space, hCP encounters elevated Ca²⁺ concentrations and binds more Ca²⁺ ions, forming a heterotetramer that is poised to compete with microbial pathogens for essential metal nutrients.

Graphical Abstract

Introduction

Calprotectin (CP, S100A8/S100A9 oligomer, MRP8/MRP14 oligomer) is a Ca²⁺- and transition-metal-ion-binding protein and an abundant component of the innate immune system.¹ CP functions as a high-affinity metal-ion chelator to compete with microbial pathogens for available transition metal ions that are essential nutrients, including Mn²⁺, Fe²⁺, Ni²⁺ and Zn²⁺.^1–8 This nutrient-starvation strategy is one aspect of innate immunity, a front-line, non-specific defense strategy against invading pathogens to prevent their growth and ultimately infection.

In the current model for the metal-sequestering function of CP, white blood cells (i.e., neutrophils) and epithelial cells produce and release CP into the extracellular space in response to microbial invasion.¹ In this milieu, the Ca²⁺ ions (≈ 2 mM)⁹ modulate the metal-withholding function of the protein. CP binds Ca²⁺ ions and transition metal ions at independent sites (vide infra). Many previous studies demonstrated that the Ca²⁺-bound form of CP exhibits enhanced transition metal ion affinities,^{3, 10–12} antimicrobial activity,^{2–3, 10} and resistance to serine proteases¹³ compared to the apo protein. Although the metal-sequestering function of this fascinating and complex system depends on an interplay of metal binding, conformational change, and self-association, the overall picture for these processes has remained enigmatic. We do not have a molecular model for how CP binds Ca²⁺ ions and how this binding affects the protein’s structure, dynamics, and biological function in metal-withholding and beyond. To address this lack of understanding, we implemented a novel, integrated mass spectrometry (MS)-based approach to unravel the complexities of Ca²⁺ binding to human CP (hCP).

We chose Ca²⁺ binding to hCP because it is important in disease, and the majority of reported coordination chemistry studies of CP focused on this ortholog. hCP is a heterooligomer of two S100 proteins, S100A8 (α subunit, 10.8 kDa) and S100A9 (β subunit, 13.2 kDa).¹⁴ Each S100 subunit has two Ca²⁺-coordination sites: a C-terminal “canonical” EF-hand and an N-terminal “non-canonical” EF-hand.¹⁵ Apo hCP is a S100A8/S100A9 heterodimer, and Ca²⁺ binding to the EF-hands causes self-association of two heterodimers to form a heterotetramer^{10, 16–17} by a process that is not understood. Each S100A8/S100A9 heterodimer harbors two sites for coordinating transition metal ions (Figure 1A). These sites, at the heterodimer interface, include two motifs: His₃Asp (site 1) and a His₆ (site 2).^{3,10, 18–19} The His₃Asp site sequesters Zn²⁺, and the functionally versatile His₆ site withholds Mn²⁺, Fe²⁺, Ni²⁺ and Zn²⁺ from microbes.¹ CP can also sequester copper, but the details of this coordination chemistry are unclear.²⁰

Figure 1.

hCP structure and sequence coverage map of pepsin-digested CP-Ser. (A) Crystal structure of Mn²⁺-, Ca²⁺-, and Na⁺-bound CP-Ser (PDB 4XJK).¹⁸ A heterodimer unit taken from the heterotetramer is shown. S100A8 is green and S100A is blue. (B) Sequence coverage map of pepsin-digested S100A8(C42S) (top), and S100A9(C3S) (bottom) of CP-Ser. Each blue bar represents a peptide identified in a tandem mass spectrometry mapping experiment. Secondary structure elements are indicated by blue lines on top of the sequence. The Ca²⁺-binding loops are indicated by text only.

Our early studies of CP-Ser, a S100A8(C42S)/S100A9(C3S) variant of hCP, showed that Ca²⁺ ions enhance the Zn²⁺ affinities of both the His₃Asp and His₆ sites,¹⁰ and our subsequent work demonstrated that Ca²⁺ also enhances the binding affinities for Mn²⁺, Fe²⁺ and Ni²⁺.^{3, 11–12} Moreover, studies of Mn²⁺- and Fe²⁺-bound to the His₆ site of CP-Ser indicated that Ca²⁺ binding does not affect the coordination geometry or the electronic structure of the His₆ site.^{18, 21} In contrast, studies of Fe²⁺ and Ni²⁺ bound to the His₃Asp site of CP-Ser showed that the coordination environment of this site is perturbed by Ca²⁺ binding.^{12, 21} Magnetic circular dichroism spectroscopy demonstrated that the Fe²⁺-His₃Asp site is 5-coordinate in the absence and presence of bound Ca²⁺, but that the ligand-field strength decreases upon Ca²⁺ binding.²¹ Crystallographic evaluation of Ni²⁺-bound CP-Ser in the absence and presence of Ca²⁺ revealed that the coordination number of Ni²⁺ changed from 6 to 4 at the His₃Asp site as a result of Ca²⁺ binding.¹² Although these observations are informative, they do little to advance our understanding of how Ca²⁺ ions modulate the conformation and transition-metal binding of CP.

Two recent X-ray structures of metal-bound CP-Ser, obtained with crystallization conditions of 10 equivalents of Ca²⁺ per homodimer,^{5, 18} show a Ca²⁺ stoichiometry of < 4 per heterodimer. In the case of Mn²⁺-, Ca²⁺- and Na⁺-bound CP-Ser, the protein crystallized as a heterotetramer.¹⁸ Although anomalous electron density corresponding to bound Ca²⁺ was observed in the four C-terminal EF-hands of the heterotetramer, no evidence for bound Ca²⁺ was found in any of the four N-terminal EF-hands (Figure 1A).¹⁸ Na⁺ ions, abundant in the crystallization solution, were modeled into the N-terminal EF-hands. These results are consistent with the notion that the C-terminal EF-hands bind Ca²⁺ with higher affinity than the N-terminal EF-hands and demonstrate that 10 equivalents of Ca²⁺ are insufficient for Ca²⁺ binding at the N-terminal EF-hands, at least under the crystallization conditions. The results, however, do not provide an answer to the Ca²⁺ binding stoichiometry of CP-Ser or its behavior in solution, motivating the research described here.

Compared to investigating transition-metal binding to hCP, molecular-level insight into how Ca²⁺ interacts with hCP is difficult to achieve. Ca²⁺ is spectroscopically silent, each heterodimer has four different Ca²⁺-binding sites, and Ca²⁺-dependent self-association of heterodimers to form heterotetramers presents an additional variable that must be considered. As a result, methods routinely applied to study transition-metal binding to hCP – ranging from competition titrations for apparent binding affinities to advanced spectroscopic methods for the electronic structure determination of transition-metal-bound species – are poorly suited for evaluating Ca²⁺ binding.

Therefore, we implemented a combination of MS-based structural proteomics approaches that allow not only direct observation of Ca²⁺ binding to hCP but also assessment of oligomerization, and determination of apparent binding affinities. The approaches are (i) hydrogen/deuterium exchange MS (HDX-MS), (ii) protein ligand interactions in solution by mass spectrometry, titration and hydrogen/deuterium exchange (PLIMSTEX), and (iii) native mass spectrometry (native-MS). Importantly, the approaches apply to the native protein and do not require a surrogate for Ca²⁺.

HDX-MS measures relative rates of exchange of backbone amide hydrogen atoms with deuterium, permitting measurement of the changes in hydrogen-bonding and solvent accessibility accompanying both Ca²⁺-binding and self-association in solution. This approach facilitates identification of Ca²⁺-binding sites and protein-protein interfaces.^22–25

PLIMSTEX provides an indirect readout of ligand binding in solution whereby the protein is titrated with a ligand (Ca²⁺) and its conformation is monitored by HDX at each titration point. These measurements can pertain to the entire protein or to regional (peptide) levels by following fragments formed by proteolytic digestion. In either case, the resulting data are fit to models to obtain apparent binding affinities (i.e., association (K_a) or dissociation (K_d) constants). This approach is particularly appropriate for proteins like CP that coordinate ligands with several stoichiometries.^26–28

Lastly, employing native-MS to monitor ligand-binding titrations allows observation of protein oligomers resulting from ligand complexation. The titration is performed in solution, and the use of a volatile supporting electrolyte (ammonium acetate, NH₄OAc) enables direct MS detection of the various oligomers with and without a bound ligand (in this case, Ca²⁺). This titration affords information on molecular weight, stoichiometry, and relative stabilities of protein-ligand complexes.^29–30

Prior MS studies using HDX kinetics and PLIMSTEX for the Ca²⁺-binding proteins calmodulin³¹ and using HDX kinetics for troponin C²⁷ and DREAM (Downstream Regulatory Element Antagonist Modulator)³² provided insight on conformational changes upon Ca²⁺ binding for systems that are less complex than CP. Unlike these prior studies, here we integrate data from HDX kinetics, PLIMSTEX, and native MS. This new, combined approach provides details about the apo heterodimer and the Ca²⁺-bound heterotetramer, uncovers the Ca²⁺-binding stoichiometry, and enables expanded modeling to give stepwise binding affinities. The combination, more than any single approach, delivers a detailed picture of how Ca²⁺ binding affects the structure and function of hCP. In particular, the experimental results accompanied by equilibrium modeling show pathways for Ca²⁺ binding and the tetramerization and provide equilibrium constants for the various steps. The approach should be useful for future investigations of CP and its role in infectious disease^{2, 6, 33} and other pathologies that include Alzheimer’s disease³⁴ and cancer,^35–36 and for other intricate metal-binding proteins.

Results and Discussion

HDX Kinetics Identify the Ca²⁺-binding Sites

We studied the hCP variant CP-Ser, a S100A8(C42S)/S100A9(C3S) oligomer, obtained from recombinant protein expression.¹⁰ CP-Ser allows experiments to be conducted without the complications associated with disulfide bonding (e.g., no need for a reducing agent in the HDX or other measurement). This Cys-null variant has been employed extensively in metal-binding and antimicrobial activity studies, and all available data indicate that it displays the same Ca²⁺-dependent heterotetramerization and transition-metal-binding properties as hCP.^{10, 37}

We first measured HDX kinetics for CP-Ser at the peptide level in the absence and presence of excess Ca²⁺ ions. By examining the S100A8 and S100A9 peptide fragments obtained following pepsin digestion, we identified peptide fragments corresponding to each of the four EF-hand domains and the tetramer interface. A total of 65 (40 unique – duplicates with different charge states are not counted) and 85 (55 unique) peptides were identified in a mapping experiment for the S100A8 and S100A9 subunits, respectively, with nearly complete coverage (≈ 95%) for both the subunits (Figure 1B). We identified multiple overlapping peptides containing the Ca²⁺-binding sites of each subunit. A comparison of HDX for Ca²⁺-free and Ca²⁺-bound CP-Ser revealed protection (reduced HDX) upon Ca²⁺ binding for nearly the entire S100A8 subunit, except for regions represented by peptides 22–26 and 40–52. Likewise, protection of nearly the entire S100A9 subunit was observed, except regions 48–60 and 83–114 of the Ca²⁺-bound state. Protection increases in the presence of Ca²⁺ for both the S100A8 and S100A9 N- and C-terminal EF-hands and for the neighboring regions in each protein subunit (Figures S1 and S2).

Upon Ca²⁺ binding, greater reduction in HDX is observed at the C-terminal EF-hands than at the N-terminal EF-hands; the extent of HDX decreases by as much as 59% at the C-terminal EF-hand of each subunit, indicating strong protection and a clear conformational change (Figures 2, S1 and S2). For example, peptides 16–26 and 27–34 (Figure 2A, B) from the N-terminal EF-hand of S100A8 exhibit smaller differences in HDX between bound and unbound than does peptide 55–68 from the C-terminal EF-hand of S100A8 (Figure 2C). Similarly, for S100A9, the C-terminal EF-hand region 65–77 (Figure 2G) shows greater protection to HDX compared to the N-terminal EF-hand peptide 22–36 (Figure 2E). These results are consistent with stronger Ca²⁺ binding at the canonical C-terminal EF-hands. These sites have greater negative charge and more oxygen donors compared to the non-canonical N-terminal EF-hands.^{15, 18} We also observed reduction in deuterium uptake upon Ca²⁺ binding for peptides adjoining the C-terminal Ca²⁺-binding site for S100A8 (Figure 2D), whereas regions of CP-Ser not involved in Ca²⁺ binding showed nearly invariant HDX behavior (Figure 2F).

Figure 2.

HDX studies of CP-Ser (5 μM) in the absence and presence of excess Ca²⁺ ions (250 μM). (A-H) Representative HDX kinetic plots for peptides from the CP-Ser subunits S100A8(C42S) (A-D) and S100A9(C3S) (E-H) in the absence (red) and presence (blue) of Ca²⁺ ions. Kinetic plots for peptides from EF-hands (A, B, C, E, and G) and other regions (D, F, H) are shown. Peptides sequences and associated charge states are: (A) YHKYSLIKGNF (S100A8(C42S) 16–26 (+2)), (B) HAVYRDDL (S100A8(C42S) 27–34 (+2)), (C) FKELDINTDGAVNF (S100A8(C24S) 55–68 (+2)), (D) VIKMGVAAHKKSHEESHKE (S100A8(C42S) 75–93 (+3)), (E) SVKLGHPDTLNQGEF (S100A9(C3S) 23–37 (+2)), (F) FLKKENKNEKVIE (S100A9 (C3S) 48–60 (+2)), (G) LDTNADKQLSFEE (S100A9(C3S) 66–78 (+2)), and (H) TWASHEKMHEGDEGPGHHHKPGLGEGTP (S100A9(C3S) 87–114 (+3), respectively. (I) Differential HDX uptake in % is mapped onto the crystal structure of the Ca²⁺-bound CP-Ser tetramer (PDB: 4XJK). Different views of the differential HDX map rotated 90° around z-axis from and 180° around the y-axis from are shown. Colored bars represent averaged HDX differences between apo and holo CP-Ser across all seven time points in duplicate. The extent of HDX is represented by colored regions in the map, gray regions mark no or little change with green, cyan and blue representing small, medium and large changes in deuterium uptake in the holo-state, respectively.

HDX indicates that Ca²⁺ binding has a negligible effect on the C-terminal tail of S100A9, defined as residues 96–114; this region is essential for metal sequestration by CP. In addition to containing two histidine residues that complete the His₆ coordination sphere at site 2, the tail changes conformation upon coordination of a divalent metal ion at the His₆ site and thereby effectively encapsulates the bound metal ion.¹⁸ We identified multiple peptides from the tail region of S100A9 (Figures 1B, S4), one of which covers residues 87–114. This peptide and others from this region of S100A9, display comparable HDX in the absence and presence of Ca²⁺ (Figure 2H), indicating that the enhanced transition-metal affinities that occur at the His₆ site upon Ca²⁺ binding do not result from a Ca²⁺-dependent conformational change in the tail region. Because short, overlapping peptides from the tail region of S100A9 were not identified (lost during chromatography or lack of proteolysis), it remains possible that large changes in HDX with the addition of Ca²⁺ occur within this region but were not observed. In such a scenario, signals from residues strongly affected by Ca²⁺ would be diluted by those that change little in this 28-residue long peptide and would be missed.

To obtain a global picture of how Ca²⁺ binding affects HDX across the protein, we calculated the averaged deuterium uptake differences between CP-Ser with and without Ca²⁺ bound. We mapped these differences onto the crystal structure of the Ca²⁺-, Na⁺- and Mn²⁺-bound CP-Ser heterotetramer (PDB: 4XJK)¹⁸ as a model (Figure 2I). That structure contains Ca²⁺ ions bound at the C-terminal EF-hand domains, Na⁺ ions bound in the N-terminal EF-hand domains, and a Mn²⁺ ion bound at the His₆ site, but only the Ca²⁺ ions are shown in the HDX model (Figure 2I). The structural map presented here shows that the CP-Ser heterotetramer exhibits significant protection in HDX at the Ca²⁺-binding sites, the heterotetramer interface, and at several regions remote from the Ca²⁺-binding sites for the S100A8 and S100A9 subunits. Because Ca²⁺ binding causes apo CP heterodimers to self-associate and form heterotetramers,^{10, 16–17} these results indicate that heterotetramer formation observed following Ca²⁺ addition decreases the extent of HDX.

The crystal structure shows the tetramer interface is primarily comprised of hydrophobic residues (Ile60, Ile73 and Ile76) from S100A8 and tryptophan (W88) from the S100A9 subunit. HDX interrogates the tetramer interface in solution and shows a gain in protection for regions represented by the S100A8 peptides containing Ile 60 (also in a Ca²⁺-binding site), Ile73, and Ile76 (Figures 2C, 2D, S1) but not for the tryptophan (W88)-containing peptide at the N-terminus of S100A9 (Figures 2H, S2). Any effect of W88 is presumably diluted by the large number of unaffected residues contained in that peptide, as discussed above.

PLIMSTEX Reports on the CP-Ser Heterodimer

The HDX kinetics provided a foundation for the application of PLIMSTEX to further interrogate the four Ca²⁺-binding events and determine the Ca²⁺-binding affinities of the EF-hand domains in CP-Ser (Figures 3 and 4). In a PLIMSTEX experiment, an appropriate exchange time from the HDX kinetics is chosen, and exchange at specific regions of the protein (e.g., the Ca²⁺-binding sites) is monitored in a titration format (see Experimental Methods), permitting determination of the binding affinity at each site. We titrated CP-Ser with increasing amounts of Ca²⁺ and identified the peptides that underwent significantly decreasing HDX with increasing [Ca²⁺], particularly peptides corresponding to the Ca²⁺-binding sites (Figure 3A,C,E,G) and peptides from the S100A8 that form the tetramer interface (Figure 3C,D). For instance, the peptide (55–68) from S100A8 contains both a Ca²⁺-binding site and residues that contribute to the tetramer interface (Figure 3C). In contrast, one peptide (87–113) from the S100A9 subunit that contains interfacial W88 does not change (Figure 3H), because of its large size (vide supra). We also found peptides corresponding to other protein regions that showed negligible change in HDX (Figure 3B,F); these peptides can be considered to be controls.

Figure 3.

Representative plots of model-free PLIMSTEX data for the titration of 4 μM CP-Ser with Ca²⁺ ions. Data for S100A8(C42S) (A-D) and S100A9(C3S) (E-H) are shown. Peptide sequences and associated charge states are: (A) HAVYRDDL (S100A8(C42S) 27–34 (+2)), (B) TESPQYIRKKGADVW (S100A8(C42S) 40–54 (+2)), (C) FKELDINTDGAVNF (S100A8(C24S) 55–68 (+2)), (D) VIKMGVAAHKKSHEESHKE (S100A8(C42S) 75–93 (+3)), (E) SVKLGHPDTLNQGEF (S100A9(C3S) 23–37 (+2)), (F) FLKKENKNEKVIE (S100A9(C3S) 48–60 (+2)),(G)LDTNADKQLSFEE(S100A9(C3S)66–78(+2)), and(H) TWASHEKMHEGDEGPGHHHKPGLGEGTP (S100A9(C3S) 87–114 (+3), respectively. PLIMSTEX data for peptides from the EF-hands (A, C, E, and G) and other regions (B, D, F, H) are shown.

Figure 4.

PLIMSTEX data for the titration of 4 μM CP-Ser with Ca²⁺ ions fit to the NLLS algorithm using a 1:4 sequential binding model with no consideration of tetramerization. The data were acquired for four EF-hand peptides in the following charge states: HAVYRDDL (S100A8(C42S) 27–34 (+2)) (blue circles), FKELDINTDGAVNF (S100A8(C42S) 55–68 (+2)) (green squares), SVKLGHPDTLNQGEF (S100A9(C3S) 23–37 (+2)) (red diamonds), LDTNADKQLSFEE (S100A9(C3S) 66–78 (+2)) (black triangles), respectively.

For modeling Ca²⁺ binding to CP-Ser, we considered the four different Ca²⁺ sites in the heterodimer and the Ca²⁺-dependent tetramerization event. To begin, we utilized PLIMSTEX data from four peptides that originate from the four Ca²⁺-binding EF-hands of the S100A8 and S100A9 subunits (Figure 3A,C,E,G) and fit the data to a 1:4 sequential binding model that did not consider tetramerization (see Experimental Methods and Supporting Discussion in the SI). The calculated curves for the four peptides fit well the experimental data as judged from both visual inspection and the statistical parameters (resRMS of 0.92 and R² of 0.997) (Figure 4).

Next, using the PLIMSTEX data, we tested four different models incorporating tetramerization in the model-fitting (Figure S3). Because a number of other Ca²⁺-binding proteins (e.g., calmodulin) show cooperative Ca²⁺ binding, we incorporated a combination of tetramerization and cooperativity for up to eight Ca²⁺ ions bound to the CP-Ser heterotetramer into these four models. All four models fit the data well, resulting in our inability to discriminate between models and suggesting that PLIMSTEX modeling at this stage is insufficiently constrained to determine the affinities for Ca²⁺ binding and tetramerization (Figure S3, Table S1, and Supporting Discussion in the SI). Moreover, for each of the four models, the inflection in the titration plot occurs at approximately two equivalents of Ca²⁺ per heterodimer (Figure 4). This result contrasts with results of prior biophysical studies indicating that, in solution, full tetramerization requires approximately 20 equivalents of Ca^{2+.11, 13>} Thus, the PLIMSTEX data must primarily pertain to the CP-Ser heterodimer. Also, the observations from PLIMSTEX suggest that changes in protein conformation are largely complete when two (rather than four) equivalents of Ca²⁺ are added, reinforcing the notion that the information obtained from PLIMSTEX may be incomplete. It is possible, however, that binding of the second Ca²⁺ ion already induces conformational propagation in other EF-hands that bind the third and fourth Ca²⁺ ions (i.e., minimal conformational changes occur upon additional Ca²⁺ binding).³² To draw a comprehensive picture of Ca²⁺ binding to CP-Ser, we reasoned that the PLIMSTEX modeling needed to be constrained by the binding stoichiometry and the equivalents of Ca²⁺ that causes the heterodimer-to-heterotetramer transition.

Sharp-break PLIMSTEX Curves Show that the CP-Ser Heterodimer Binds Four Ca²⁺ Ions

To constrain the PLIMSTEX fitting by the Ca²⁺-binding stoichiometry, we obtained sharp-break PLIMSTEX²⁶ curves by titrating CP-Ser at concentrations that likely exceed the K_d values (Figure 5). We titrated 40 μM CP-Ser – a concentration 10-fold greater than the CP-Ser concentration used in the PLIMSTEX experiment described above – with up to 12 equivalents (480 μM) of Ca²⁺. Almost 90% of the peptides directly from or adjoining the Ca²⁺-binding sites exhibited a near-sharp break at a 4:1 Ca²⁺:CP-Ser heterodimer molar ratio (Figure 5). In contrast, peptides corresponding to the N-terminal region ( residues 1–7), and middle region (residues 48–60) of the S100A8 subunit, and C-terminus (83–114) of the S100A9 subunit showed negligible change in HDX (Figure S4), consistent with the behavior expected for regions that do not bind Ca²⁺. At least two peptides, WFKELDINT (52–62 (+2)) from S100A8 and IINTFHQY (14–21 (+2)) from A9 showed breaks at lower Ca²⁺:CP-Ser stoichiometry, indicating that a few regions of the protein complete their conformational changes when binding less than four Ca²⁺ ions (Figure S4B). With these exceptions, most of the results indicate that the heterodimer binds up to four Ca²⁺ ions and, by corollary, the tetramer binds up to eight Ca²⁺ ions. This analysis is consistent with previous MS studies that reported an 8:1 stoichiometry for Ca²⁺ binding to hCP.^16–17

Figure 5.

Sharp-break PLIMSTEX curves obtained for the S100A8(C42S) (A, B) and S100A9(C3S) (C, D) subunits, respectively. The data were acquired for the four peptides in the following charge states: (A) HAVYRDDLKKLLE (S100A8(C42S) 27–39 (+2)), (B) KKLLETESPQY (S100A8(C42S) 35–45 (+2)), (C) ERNIE (S100A9(C3S) 9–13 (+1)), (D) HIMED (S100A9(C3S) 61–65 (+1)).

Native-MS Reveals the Ca²⁺ Equivalents Required for Dimer-to-Tetramer Transition

To add a further constraint for the modeling of Ca²⁺ binding to CP-Ser, we chose a direct method that simultaneously monitors the heterodimer and heterotetramer and the numbers of Ca²⁺ ions bound to each (Figure 6). For this, we acquired native mass spectra of 15 μM CP-Ser in the presence of 0–500 μM Ca²⁺. We observed that the abundance of the heterodimer decreased and the abundance of the heterotetramer increased as more Ca²⁺ was added (Figure 6A). Moreover, peaks representing up to eight Ca²⁺ ions bound to the heterotetramer occurred at >200 μM Ca²⁺, in agreement with the 8:1 binding stoichiometry for the CP-Ser heterotetramer deduced from the sharp-break PLIMSTEX titration (Figure 5). Importantly, the native-MS experiments also demonstrate that the transition of CP-Ser from the heterodimer to the heterotetramer occurs between 13–20 equivalents (200–300 μM) of added Ca²⁺, and the results suggest that the heterodimer bearing four Ca²⁺ prefers to be in the tetramer state (Figure 6A and Supporting Information). The observation of incomplete Ca²⁺ binding in the heterotetramer (i.e., 6 and 7 Ca²⁺ species) is likely due to loss of the metal ion in the ionization process. A calculation (Supporting Information) starting with 15 μM CP-Ser shows that the probability of having one molecule of CP-Ser heterodimer in a 10-nm ESI droplet is ≈ 0.04 and an upper bound on the fraction of two or more unassociated heterodimers in the droplet being counted in the heterotetramer signal is at most 4%. This analysis supports that the overall signals for heterodimer and heterotetramer in native MS represent the species in solution and are not artifacts of solvent evaporation in the ESI process. In this calculation, the droplet radius in nanospray is taken to be approximately 10 nm after a single droplet fission.^38–40

Figure 6.

Native MS of CP-Ser in the absence and presence of Ca²⁺ ions. (A) Native mass spectra of 15 μM CP-Ser in the presence of 0–500 μM Ca²⁺ ions, zoomed in to show the most abundant charge states of dimer (9+) and tetramer (12+) species. The apo CP-Ser peak is highlighted in light blue and peaks for Ca²⁺-bound species are highlighted in pink. (B) Plots of CP-Ser species distribution at different [Ca²⁺] represented for dimer (blue circles) and tetramer (red squares). Signal intensities for each CP-Ser species are corrected for the fraction of the total dimer subunit for each [Ca²⁺].

An accurate estimate of the equivalents of Ca²⁺ ions required for the heterodimer-to-heterotetramer transition comes from the relative abundance of the CP-Ser heterodimer and heterotetramer species, determined from the corresponding native-MS peak intensities and plotted vs. the [Ca²⁺]/[CP-Ser] ratio ( Figure 6B). A 1:1 ratio of heterodimer and heterotetramer occurs for 15 equivalents of Ca²⁺ (225 μM). This value is reminiscent of the number of Ca²⁺ equivalents required to fully enhance the transition-metal-ion affinities of hCP.^{3, 11} We also note that both the CP-Ser and Ca²⁺ ion concentrations used in this experiment are physiologically relevant. Levels of CP in tissue abscess fluid can exceed 1 mg/mL (≈ 40 μM heterodimer), and the total extracellular Ca²⁺ concentration is ≈ 2000 μM. ^{9, 41}

Combination of PLIMSTEX, Native-MS, and Sharp-break PLIMSTEX Allows Assessment of Ca²⁺-binding Affinities

The results from HDX and PLIMSTEX (at 4 μM CP-Ser) delineate the Ca²⁺-binding behavior of the CP-Ser heterodimer, whereas those from native MS at a higher concentration (15 μM CP-Ser) and from sharp-break PLIMSTEX (40 μM CP-Ser) report the Ca²⁺ equivalents required for the dimer-to-tetramer transition. Together, these data allow us to constrain the modeling and incorporate up to four independently binding Ca²⁺ ions for the heterodimer and eight for the heterotetramer (Figure 7 and Supporting Discussion in SI). The best fit of the data with the model (referred to as Model K in SI and described above) provides the Ca²⁺-binding affinities associated with the four EF-hands of CP-Ser, a tetramerization constant, and an overall Adair binding constant (β) (Table 1). The overall Adair binding constant, β = 2.7 × 10⁴⁵ M⁻⁹, is the product of the binding constants for the four Ca²⁺ ions for each heterodimer and the tetramerization constant (see Supporting Discussion in SI for details).

Figure 7.

Data fitted with NLLS to a model incorporating four independent Ca²⁺ (1–4 Ca²⁺) in the dimer and 8 Ca²⁺ in the tetramer. (A) The data are PLIMSTEX (4 μM CP-Ser) and native MS (15 μM CP-Ser) and represent the binding sites A9(C3S) 23–37 (+2) (red diamonds), A8(C42S) 27–34 (+2) (blue circles), A9(C3S) 66–78 (+2) (black triangles), A8(C42S) 55–68 (+2) (green squares), and native-MS data (purple diamonds). (B) The data are sharp-break PLIMSTEX (40 μM CP-Ser) and native-MS (15 μM CP-Ser). The data are from the Ca²⁺-binding sites A9(C3S) 9–13 (+1) (red diamonds), A8(C42S) 27–39 (+2) (blue circles), A9(C3S) 61–65 (+1) (black triangles), A8(C42S) 35–45 (+2) (green squares), and native-MS data (purple diamonds).

Table 1.

Calculated dissociation constant values from experimental data and model fit.^a

Dissociation Constant	Value (M−1)	Propagated Errorb
K1	9.4 × 10−7	0.74 × 10−7
K2	9.4 × 10−7	0.74 × 10−7
K3	4.9 × 10−5	0.28 × 10−5
K4	4.3 × 10−1	0.06 × 10−1
Ktetramer	1.1 × 10−12 (β = 2.7 × 1045M−9)	0.006 × 10−12

^aDissociation constants were determined from the data presented in Figure 7.

^bFirst order approximation of the uncertainty in the fit parameters as propagated through the least squares fitting from the standard deviations of the replicate HDX measurements.

Because the PLIMSTEX data are similar for the four EF-hands and do not allow clear assignment of the order of Ca²⁺ binding, we did not use those data to assign the four dissociation constants (K₁-K₄) to specific EF-hands. Our modeling shows that two EF-hand domains coordinate Ca²⁺ micromolar affinity whereas the other two EF-hands coordinate Ca²⁺ with substantially lower affinity. We associate K₁ and K₂ with the C-terminal EF-hands and K₃ and K₄ with the N-terminal EF-hands, in accord with the X-ray crystallography¹⁸ and with the HDX kinetics (vide supra). Further, we speculate that the S100A9 and S100A8 non-canonical EF-hands bind the third and the fourth Ca²⁺ ions, respectively. The S100A8 noncanonical EF-hand (LIKGNFHAVY), lacking amino acids that tightly bind Ca²⁺, shows asharp change in conformation when CP-Ser binds two Ca²⁺ ions and then shows no further change with increasing Ca²⁺ equivalents (Figure 4). Once this EF-hand binds the fourth Ca²⁺ ion, the protein transitions to the heterotetramer. The S100A9 noncanonical EF-hand (GHPDTLNQGE) changes its conformation more slowly (Figure 4), and we assign this site as the third to bind Ca²⁺. Indeed, this EF-hand has more side chains that strongly bind Ca²⁺ than does the S100A8 noncanonical EF-hand. The character of the PLIMSTEX curves make it difficult to judge whether the binding is cooperative; thus, the four binding events were modeled to be independent (see Supporting Discussion in SI).

To test whether the model provides equivalent results for a more limited data set, we modeled the PLIMSTEX data at 4 μM CP-Ser in combination with the native-MS titration at 15 μM CP-Ser. These results (Table S4, Figures S5–S8, and Supporting Discussion) provide comparable estimates of K₁ and K₂ and of K_tetramer but underestimate the role of the heterodimer with three bound Ca²⁺ ions by yielding a smaller K₃ (compare Table 1 and Table S4) and a larger K₄. These comparisons underscore the importance of incorporating measurements at a higher protein concentration than 4 μM and show that a more complete picture is provided by modeling the combination of three titrations at 4, 15, and 40 μM, as described above.

Although this work provides the first examination of the Ca²⁺-binding affinities of CP, Ca²⁺ affinities of other S100 proteins have been reported. Those prior studies indicate that C-terminal canonical EF-hands generally have K_d values in the range of 1–50 μM.^{15, 42–44} The K_d values determined here are ≈ 1 μM, at the strong-binding end of this range. Reported K_d values for N-terminal EF-hands fall in the range of 100–500 μM.^{15, 42–44} Our modeling indicates that one of the N-terminal hands (assigned as S100A9, see above) has a K_d value of 50 μM whereas the other (S100A8) is remarkably low at 0.4 M. We speculate that the low binding affinity for the fourth Ca²⁺ ion is likely offset by the high propensity to tetramerize once the heterodimer is loaded, even at a trace level, with a fourth Ca²⁺ ion.

Overall, the K_d values for Ca²⁺ binding determined in this work support the working model postulated for the CP-Ser.¹ In particular, these results indicate that hCP is only partially Ca²⁺-bound at the C-terminal EF-hands in the cytoplasm where Ca²⁺ concentrations are low (nanomolar range). It is likely that a mixture of species (apo, 1 and 2 Ca²⁺ dimers) exists in this milieu. With release into the extracellular space, three EF-hands readily bind Ca²⁺, but the fourth site is only weakly populated. The two low-affinity K_d values are consistent with the non-canonical N-terminal EF-hands of S100A8 and S100A9. The PLIMSTEX data indicate substantial conformational changes upon binding of two Ca²⁺. This behavior suggests that Ca²⁺ binding at the C-terminal EF-hands induces conformational changes at the N-terminal EF-hands to poise the protein to bind additional Ca²⁺ ions once the Ca²⁺ concentration becomes high, as is the case for Ca²⁺ binding to DREAM.³²

Given the weak binding affinity of the fourth Ca²⁺ ion, the binding interface between the two heterodimers is likely to be the deciding factor in the tetramerization of CP-Ser. The best model fit gives a tetramerization K_d value of 1.1 x 10⁻¹² M (Table 1), indicative of a strong binding interface between the two heterodimers. When the heterodimer has bound one or two Ca²⁺ ions, which occurs at low Ca²⁺-to-protein ratios, no tetramerization occurs, and the protein is in a “resting state”. Once the protein is released to extracellular space, it binds a third Ca²⁺ and begins to bind a fourth Ca²⁺ ion. The driving force for tetramerization of the dimer with 4 Ca²⁺ is so favorable that it offsets the weak affinity for a fourth Ca²⁺ ion and “pulls” the equilibrium toward product to form the heterotetramer binding 8 Ca²⁺ ions, preparing the protein to play its role in innate immunity by sequestering transition metal ions.

The working model described above indicates that the Ca²⁺ affinities of hCP are tuned such that the heterodimeric protein becomes heterotetrameric when it encounters higher extracellular Ca²⁺ levels. An informative way to view this behavior is the variation in fractional species from the model (Figures 8 and S10). The fractional-species variation shows clearly that the concentration of the heterodimers containing two and three bound Ca²⁺ ions decreases as the heterotetramer appears and becomes ultimately the dominant species. This working model has support from prior evaluations of the protein under conditions of low and high Ca²⁺ concentrations;^{3, 10–12} however, until now, this view has lacked convincing quantitative support.

Figure 8.

Fractional-species variations as a function of the relative concentration of Ca²⁺ showing the concomitant fit of the 4 μM protein PLIMSTEX data for A8 27–34. The model does show a dimer plus 4Ca²⁺, but its abundance is too low to be seen on this graph. The fractional species also for the apo protein does not start at zero because there was a small amount of adventitious Ca²⁺ in the sample solution, and that is an input to the model. This is also seen in Figures S7 and S10.

Summary and Outlook

The present study provides comprehensive, quantitative molecular information of the Ca²⁺-binding properties of CP and its metal-sequestering function in solution. In particular, the use of an integrated MS approach allowed us to bring new understanding to a complicated system that involves multiple Ca²⁺-binding events and self-association. The resulting quantitative insights support and guide our development of the working model for how CP sequesters transition metal ions in the extracellular space. To the best of our knowledge, this work constitutes the first report of the affinity between S100 oligomers.

The current working model of how Ca²⁺ ions modulate the structure and function of CP starts with the protein existing primarily as a Ca²⁺-free heterodimer when stored in the cytoplasm of immune and epithelial cells where the resting Ca²⁺ concentration is low (≈ 100 nM).⁹ When CP is released into the extracellular space, it encounters higher Ca²⁺ concentrations (≈ 2 mM)⁹ and binds Ca²⁺, facilitating production of a heterotetramer, increasing the transition-metal-ion affinities of the His₃Asp and His₆ sites, enhancing antimicrobial activity, and affording protease resistance.^{1, 13} This model has been largely based on transition-metal-binding studies demonstrating that excess Ca²⁺ ions enhance the transition-metal affinities of the protein and has been corroborated by functional studies.^{1, 3, 10–13} Although there is compelling evidence from prior studies that Ca²⁺-induced changes in CP are essential for function and also impact its fate in the biological milieu, the molecular basis for these effects has remained elusive. Moreover, there is no reported crystallographic or solution structure of the apo CP heterodimer. As a result, we have no ability to link structural snapshots to achieve depictions of conformational change.

The integration of three MS-based approaches enables a further evaluation of the hypothesis that Ca²⁺ ions modulate hCP function in vivo and to (i) confirm the four regions that serve as Ca²⁺ binding sites, (ii) determine the binding stoichiometry for the CP-Ser to be four Ca²⁺ per dimer and eight Ca²⁺ per tetramer, (iii) establish the protein-to-Ca²⁺ molar ratio that causes the dimer-to-tetramer transition, and (iv) determine the Ca²⁺-binding affinities associated with four binding sites (Table 1) and the self-association constant for the Ca²⁺-bound protein. Taken together, the results are consistent with the view that local Ca²⁺ levels in the biological milieu modulate hCP structure, metal-sequestering function, and fate. In particular, our modeling indicates that the third and fourth EF-hands of CP will be populated with Ca²⁺ only at high Ca²⁺ concentrations when the heterotetramer has formed, supporting the idea that intracellular CP is a heterodimer in resting cells. With regard to the fate of the protein, prior work from some of us indicates that self-association modulates the lifetime of CP in biological milieu.^{13, 45} The decrease in HDX for the Ca²⁺-bound heterotetramer relative to the apo heterodimer demonstrates that the protein adopts a less dynamic structure upon Ca²⁺ binding, providing a new structural insight into the marked protease resistance observed for the former species. Furthermore, the number of bound Ca²⁺ ions required for tetramerization suggests that serine proteases degrade apo CP as well as CP heterodimers that have several Ca²⁺-binding stoichiometries.

Despite the success of the integrated workflow developed here, one limitation is the inability to ascertain whether cooperative Ca²⁺ binding occurs (Supporting Discussion in SI). Given that cooperativity is a hallmark of EF-hand proteins, it is possible that the current experimental design precluded observation of cooperativity, and further studies along these lines are certainly warranted.

In closing, we expect that this integrated MS methodology to be applicable to other complicated protein-ligand systems where other approaches afford incomplete pictures. We look forward to extending this approach to evaluate transition-metal binding by hCP and other S100 proteins. We anticipate that the resulting insights from such endeavors will enhance our appreciation of the molecular mechanisms of S100 proteins at the host-pathogen interface and in associated pathologies such as Alzheimer’s disease and cancer.

Experimental Methods

Materials

All chemicals, proteases and solvents were purchased from Sigma Aldrich (St. Louis, MO) unless otherwise stated. Deuterium oxide was from Cambridge Isotope Laboratories Inc. (Andover, MA). All buffers and metal solutions were prepared using Milli-Q water (18.2 MΩ•cm, 0.22-μm filter, Millipore). All mobile chromatography phases for protein purification were prepared with Ultrol grade HEPES (free acid, Calbiochem) and TraceSELECT NaCl (Fluka), and aqueous TraceSELECT NaOH (Sigma) was used for pH adjustments. All buffers were stored in polypropylene bottles. The highest available purity of calcium chloride (99.99%) was purchased from Alfa Aesar. Stock solutions of Ca²⁺ (1 M, 100 mL) were prepared by using Milli-Q water and acid-washed volumetric glassware, and the solutions were transferred to and stored in polypropylene tubes. Protein concentrations were determined by using their optical absorbance at 280 nm (ε280 = 18,450 cm⁻¹M⁻¹ for the CP-Ser heterodimer¹⁰) by using a Nanodrop (Thermo Scientific). All concentrations of CP-Ser and equivalents of Ca²⁺ are reported in terms of the heterodimer.

Protein Preparation

The hCP variant CP-Ser, a S100A8(C42S)/(S100A9(C3S) oligomer, was used for all the experiments in this study. CP-Ser was prepared as the apo heterodimer and handled as previously described.^{10, 37}

Hydrogen-Deuterium Exchange (HDX)

The HDX experiment and data analysis procedures were similar to previously described protocols.³² Briefly, CP-Ser (50 μM) was equilibrated in the absence or presence of 2.5 mM Ca²⁺ for at least 1 h at 25°C in 11 mM HEPES, 11 mM NaCl, pH 7.0 in H₂O. A 2-μL aliquot of the solution was then diluted ten-fold with buffer prepared in D₂O (11 mM HEPES, 11 mM NaCl, pH = 7.0) (or H₂O buffer for an undeuterated control sample) to initiate the HDX reaction. After various exchange times (10, 30, 60, 120, 900, 1800, 3600 and 10800 s), the HDX was quenched by adding 30 μL of ice-cold quench solution (3 M urea in 1% (v/v) aqueous trifluoroacetic acid), which lowered the solution pH to 2.5. Peptide identification was accomplished through a combination of accurate mass analysis and LC-MS/MS by using MassMatrix (version 2.4.2) for identification.⁴⁶ Raw HDX mass spectra were submitted to HDX Workbench for calculation and data visualization.⁴⁷ The HDX data were acquired in LTQ-FTICR mass spectrometer (Thermo Scientific) and presented here as an average of duplicate measurements for each exchange time.

PLIMSTEX

CP-Ser was incubated with increasing concentrations of Ca²⁺ for 1 h at 25°C. The samples were then diluted with an excess (ten-fold) amount of buffer containing D₂O. The final protein and ligand concentrations for the PLIMSTEX experiments were 4 μM CP-Ser and 0–500 μM Ca²⁺ for the normal (not sharp-break) PLIMSTEX titration, and 40 μM CP-Ser and 0–480 μM Ca²⁺ for the sharp-break PLIMSTEX experiment. A total of 21 (0, 0.2, 1, 2, 5, 7, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 150, 200, 300, 400, 500 μM) and 16 (0, 1, 2, 10, 16, 20, 40, 50, 60, 70, 80, 160, 240, 320, 400, 480 μM) Ca²⁺ concentrations were used for the normal PLIMSTEX titrations and the sharp-break PLIMSTEX titrations, respectively. CP-Ser underwent HDX for 30 min at 25 °C and was submitted to on-line pepsin digestion and desalting prior to analysis with an LTQ-FTICR mass spectrometer (Thermo Scientific, Santa Clara, CA). The buffer conditions and instrumental conditions were similar to those described in the HDX experimental section except for the sharp-break PLIMSTEX where an aliquot of 500 μM stock protein in 75 mM HEPES, 100 mM NaCl, pH 7.5 was incubated with the ligand for at least 1 h at 25 °C. From this stock, 2 μL protein stock was diluted 1:10 in D₂O buffer for HDX reactions. The final concentration of CP-Ser injected onto the mass spectrometer was kept at low levels by diluting each sample to a final volume of 500 μL (25-fold dilution) volume with quench solution (1.9 M urea prepared in 1% (v/v) aqueous TFA) for the sharp-break experiment. The mass spectra were searched by using HDX Workbench⁴⁷ to calculate deuterium uptake percentages. Each PLIMSTEX experiment was performed in duplicate, and the data presented here were an average of two measurements.

Native Mass Spectrometry

CP-Ser was buffer exchanged a total of five times in 200 mM NH₄OAc solution, pH 7, by using a 10K MWCO filter (15,000 × g, 10 min, 4 °C). After the fifth round of buffer exchange, the concentration of CP-Ser was determined and the protein solution was diluted to 250 μM with 200 mM ammonium acetate solution and aliquots of this solution were stored at −80 °C until the day of the experiment. At that time, protein solutions (3 μL, 250 μM) were diluted to a final volume of 50 μL by using 200 mM ammonium acetate, pH 7.0. To the protein solutions were added different volumes of 1 mM Ca²⁺ to give final concentrations of 0, 100, 200, 300, 400 and 500 μM. The solutions were then incubated at room temperature for at least 1 h. Each mixture was then loaded onto nano-spray emitter tips ES480 (Thermo Scientific, Waltham, MA) and were analyzed by using the Exactive plus EMR instrument (Thermo Scientific, Waltham, MA). The native mass spectra were acquired from 500–20,000 m/z at a source temperature of 55 °C, CID 100 volts, CE 100 volts, and spray voltage of 1.2–1.5 kV at a mass resolving power of 17,500 (8,750 for the apo-sample with no Ca²⁺). From the native mass spectra obtained for each Ca²⁺ concentration, the fraction heterodimer and heterotetramer were calculated from signal intensities of each species at each Ca²⁺ concentration, and the relative abundances (%) were plotted as a function of the Ca²⁺ concentration to obtain a titration curve. These native-MS titration data for the heterodimer were incorporated in the model fitting dataset as a separate peptide (see Supporting Discussion in the SI) together with PLIMSTEX data for four peptides from the EF-hands. The data presented is the signal averaged over 5–10 minutes in a Thermo Exactive EMR instrument for each sample in an experiment.

PLIMSTEX Data Analysis and Binding Affinity Calculation

The HDX kinetic experiment described above provided a guide for the PLIMSTEX experiment.⁴⁸ From the HDX kinetics, the exchange time at which the difference in deuterium uptake between the apo- and the holo sample was relatively large and constant was chosen to be the time (30 min) for PLIMSTEX experiment. Titration of CP-Ser with increasing concentrations of Ca²⁺ (as shown in Figure 4 and 7) was followed by measuring the extent of deuterium uptake as a function of the Ca²⁺ concentration. The change in deuterium uptake was plotted as a function of the ligand-to-protein ratio to obtain a PLIMSTEX titration curve. The data for four peptides from the EF-hands were then fit to a mathematical model to obtain the binding affinities for Ca²⁺ binding to CP-Ser by using a nonlinear least squares (NLLS) regression analysis, as described previously.^48–49 Several different mathematical models were fit to the PLIMSTEX and native-MS data to arrive at a best-fit model (see Supporting Discussion of the SI).