Calcium carbonate (CaCO₃) is an abundant bio- and geomineral on Earth and is highly relevant to the environment, being a significant repository of carbon dioxide (CO₂).¹ Knowledge of the mechanism of CaCO₃ nucleation is of central importance for the fabrication of bioinspired materials and CO₂ sequestration. So far, much progress has been made in our understanding of the mechanism for inorganic CaCO₃ generation.²⁻⁵ It has recently been argued that the nucleation of CaCO₃ involves stable prenucleation species, i.e., prenucleation ion clusters (PNCs)^2,6,7 or a metastable liquid condensed phase.^8,9 Water plays a key role in the mineralization: the release of water molecules from hydrated mineral ions results in the formation and stabilization of PNCs and, later, in their dehydration and transformation into metastable dense liquid and amorphous intermediates, thereby kinetically directing the mineralization pathway.^6,9−12 This nonclassical nucleation pathway remains debated^3,4 but has not been disproven.⁵ Rather, the nonclassical nucleation theory unifies various experimental observations, for instance, the liquid mineral intermediates (polymer-induced liquid precursors⁸) for bioinspired material fabrication.¹³

Different from inorganic CaCO₃ mineralization,^2,5−8 CaCO₃biomineralization in nature is more complex, taking place in the presence of proteins and other biomolecules. The process relies on the concerted interfacial interactions among protein, mineral, and water.^14,15 Despite the available studies,^10,16 it remains unclear whether CaCO₃ mineralization by proteins involves prenucleation species (e.g., PNCs) and what is the role of solvent water.

To shed light on the role of water for CaCO₃ mineralization by proteins, we focus on one particularly remarkable system in nature—the sea urchin spine. Its hierarchical supramolecular structure and fascinating material properties have intrigued material scientists for decades.¹⁷⁻¹⁹ We probe the water at the interface of a spine-associated protein during CaCO₃ mineralization in situ using surface-specific vibrational sum frequency generation (SFG) spectroscopy. SFG is a second-order nonlinear spectroscopy, where the signal is only generated at the surface and interface where centrosymmetry is broken.²⁰⁻²² Thus, this method provides the vibrational response of the water molecules only at the spine protein interface for CaCO₃ mineralization.

We select one specific spine-associated protein SpSM50, which has unique structural motifs: an N-terminal C-type lectin domain and a residual glycine- and proline-rich repeat region.²³ We study the N-terminal C-type lectin domain (abbreviated as SM50-CTL below); this domain has been recognized as being functional in regulating early nucleation stages and subsequent phase transitions of the CaCO₃ mineral phase, in this and other proteins.^23,24

To improve our understanding of CaCO₃ mineralization by SM50-CTL proteins, we first perform a potentiometric titration experiment. In this experiment, Ca²⁺ ions are gradually dosed into a 5 μM SM50-CTL solution in 10 mM carbonate buffer (pH 9.0) to initiate the CaCO₃ mineralization process.^2,6 The concentration of free Ca²⁺ ions in solution is quantified by a Ca²⁺ ion-selective electrode (see Supporting Information for details). The obtained titration curve (black), as shown in Figure 1A, allows different mineralization regimes to be distinguished: I–III correspond to under-supersaturation at prenucleation (I), supersaturation at prenucleation (II), and nucleation into (amorphous) minerals (III).^10,19,25,26 As a comparison, the titration curve for Ca²⁺ into Tris buffer (green), which cannot trigger mineralization, is also shown and nearly overlaps the curve for dosed Ca²⁺ (red).

Figure 1

(A) Titration curve of Ca²⁺ into 5 μM SM50-CTL protein in 10 mM carbonate buffer. Different mineralization regimes are marked: under-supersaturation (I), supersaturation (II), and nucleation into (amorphous) minerals (III).^10,19,25,26 The curves for dosed Ca²⁺ (red) and Ca²⁺ titrated into Tris buffer (green) are also shown. (B) Experimental scheme illustrating the interfacial mineralization and SFG spectroscopy. Spectra reported here were recorded with S-, S-, and P-polarized SFG, visible (VIS), and infrared (IR) light, respectively.

Figure 1B illustrates our experimental scheme to investigate the interface of SM50-CTL proteins during CaCO₃ mineralization. The proteins were first allowed to adsorb to the air–carbonate buffer interface. Surface pressure measurement shows the proteins adsorb rapidly to the interface, reaching an equilibrium surface pressure of ∼30 mN/m. The gradual injection of Ca²⁺ has been reported to generate prenucleation CaCO₃ species with the size of several nanometers^6,10 and hydrated CaCO₃ nuclei at early nucleation.¹⁰ These mineral species, together with Ca²⁺ ions, will screen the protein charge but apparently do not change the surface pressure (Figure S1).^27,28 SFG measurements were performed in parallel with the titration experiment (Figure 1A), allowing the different mineralization stages in the SFG measurement to be probed.

Figure 2A shows the SFG spectra in the amide I region for the SM50-CTL protein assembled at the air–carbonate buffer interface with various Ca²⁺ concentrations. The spectra for the protein in the absence of Ca²⁺ show a distinct band at ∼1665 cm^–1, suggesting a well-ordered secondary folding structure at the interface. The response can be adequately described by two sub-bands: one at 1642 cm^–1, consistent with a combination of α-helical and β-turn structures,^29,30 and one at 1679 cm^–1, pointing to extended β-strands.^29,31 When the Ca²⁺ concentration is varied, no noticeable spectral changes are observed. Accordingly, the inferred amplitudes of the two bands (Figure 2B) also show negligible changes. While a full structural characterization would require elaborate NMR experiments,^32,33 the insensitivity of the amide vibrations points toward negligible conformational changes of the SM50-CTL proteins in the early stage of nucleation, which is further supported by the constant surface pressure (Figure S1) and circular dichroism spectra (Figure S3). The structural components inferred from the CD spectra are consistent with those inferred from SFG.

Figure 2

(A) Amide I SFG spectra for SM50-CTL protein at the air–carbonate buffer interface, with various amounts of Ca²⁺ ions indicated in the legend. The fits are shown as black lines. (B) Fitting amplitudes for two amide I bands with different amounts of Ca²⁺. Error bars represent the standard deviations of amplitudes.

Figure 3A shows the SFG spectra in the CH/OH region for the SM50-CTL protein at the air–carbonate buffer with various Ca²⁺ concentrations. The spectra are dominated by intense OH bands, with peaks at ∼3243 and ∼3432 cm^–1, which reflect a distribution of strongly and weakly hydrogen-bonded (H-bonded) OH groups.^22,34 The contribution from protein N–H groups is not discernible in the spectra.³⁵ The negligible N–H contribution is further supported by the weak chiral PSP SFG signal (Figure S4), to which protein N–H groups and chiral water can contribute.^36,37 The fitted spectra reveal main C–H bands at 2876, 2938, and 3065 cm^–1, attributed to symmetric CH₃ stretch, a combination of CH₃ Fermi resonance and CH₂ stretch from different amino acid side chains,^38,39 and ν₂ modes from aromatic sites,⁴⁰ respectively. The observed intensity change of the CH bands upon adding Ca²⁺ ions is caused by interference effects with the changing OH response,^40,41 as confirmed by spectra recorded in isotopically diluted D₂O/H₂O buffer solutions. The CH bands of the protein are unaffected by adding Ca²⁺ ions (Figure S5), in line with the amide I results (Figure 2).

Figure 3

SFG spectra in the CH/OH region for SM50-CTL proteins at the interfaces of air with (A) 10 mM carbonate buffer and (B) 10 mM Tris buffer, at different, indicated Ca²⁺ concentrations. The spectra for pure H₂O are shown for comparison.

In contrast to the protein amide/CH response, the water response changes substantially with increasing Ca²⁺ concentration. For charged interfaces, the SFG signal of water stems from the interfacial layer (χ⁽²⁾ contribution) and the diffuse layer (χ⁽³⁾).⁴²⁻⁴⁵ For the moderate ionic strengths used in the present work, the contributions of the diffuse layer dominate, and as such, the SFG intensity of water reports on the surface potential (ϕ₀). Figure 3 shows that, upon addition of Ca²⁺ ions, the water SFG intensity is reduced. Given the constant pH in our experiment (Figure S2), the reduced SFG intensity can stem from neutralization of the negative protein charge or more effective screening of the surface charge due to the increased ionic strength.^27,28,46

As a reference, we also report SFG spectra for proteins in Tris buffer with the same pH (pH 9) and ionic strength (10 mM) (Figure 3B). Unlike carbonate buffer, the addition of Ca²⁺ ions into Tris buffer cannot induce CaCO₃ mineralization. Therefore, this control experiment examines the effect of “pure” charge screening upon adding Ca²⁺ ions. The different spectral lineshapes for proteins in two buffers without Ca²⁺ can be attributed to the different interference between various (χ⁽²⁾, χ⁽³⁾, and nonresonant) signal sources and/or differences in coupling between OH oscillators.⁴⁷ Comparing spectra for pure and isotopically diluted buffers (Figure S6 and S7) reveals such coupling effects but confirms the trends of the changing OH intensity with varying [Ca²⁺] (Figure S8). As clearly seen in Figure 3, upon adding Ca²⁺ ions, the OH bands decrease in intensity and shift toward higher frequency. The change is more pronounced for the Tris (Figure 3B) than for the carbonate (Figure 3A) buffer.

To quantify the above changes, Figure 4A shows the integrated intensity for the OH bands: ∑∫₃₁₀₀³⁶⁰⁰I(ω)dω, reflecting the degree of water alignment. Upon adding Ca²⁺ ions to Tris buffer, Σ decreases rapidly; for the carbonate buffer, Σ decreases less dramatically. The weaker response of H-bonded water with Ca²⁺ for carbonate buffer is further supported by quantifying the first moment of the spectral distribution: μ = ∫₃₁₀₀³⁶⁰⁰ωI(ω)d(ω)/∫₃₁₀₀³⁶⁰⁰I(ω)d(ω). Figure 4B shows the change of μ versus [Ca²⁺] for the two buffers: μ increases much less for the carbonate buffer than for the Tris buffer. After nucleation ([Ca²⁺] = 1000–1200 μM), the absolute SFG intensity is higher, and μ has changed less for the carbonate buffer than for the Tris buffer. Both observations point to the interfacial water retaining a more ordered structure at the protein interface with CaCO₃ than that with only Ca²⁺ cations.

Figure 4

Integrated SFG intensity (A) and change of the first moment (B) of the two OH bands with adding different Ca²⁺ concentrations into carbonate (navy) and Tris buffer (red). The data were all fit to a single-exponential function (shown as lines). The background colors correspond to the different mineralization regimes in accordance with Figure 1A. Error bars show the standard deviations of the integrated intensities from multiple data sets.

The perturbation of interfacial water alignment, as mainly caused by the screening of protein charges, is illustrated in Figure 5. The interfacial water molecules are well-aligned by the charged proteins; this alignment can be readily perturbed by the screening of the protein charge with Ca²⁺ cations. In the mineralization environment (i.e., carbonate buffer), the charge screening appears less effective, and the first moment indicates that water molecules retain their bulk-like response. The observed water behavior can be rationalized by the co-condensation of other mineralization species, together with Ca²⁺ ions at the interface. Following previous studies on inorganic CaCO₃ mineralization,^6,9−12 these species are likely charge-neutral CaCO₃ PNCs, which show great protein binding affinity. The presence of neutral PNCs at the protein interface would perturb the water structure, but not so much the alignment, and would explain the relatively rapid change of μ with increasing [Ca²⁺] while Σ changes more gradually.

Figure 5

Sketch illustrating decreasing water alignment according to decreasing net charge at the SM50-CTL interface, with (center, carbonate buffer) and without (right, Tris buffer) prenucleation.

Previous experimental^6,10,11 and simulation^9,12 studies state that the stable PNCs further aggregate into nanodroplets in a later prenucleation stage, and the process goes along with water release from the hydration shell and has been related to microscopic liquid–liquid phase separation. This is possibly also relevant for the CaCO₃ biomineralization in the presence of SM50-CTL proteins; the largest water perturbations occur in regime II ([Ca²⁺] = 300–500 μM, Figure 4A), at the point that liquid phase separation may also occur.

In conclusion, surface-specific SFG spectroscopy was applied to probe the SM50-CTL protein and, in particular, water at the interface for mineralization of the sea urchin spine. While the protein structure is unchanged, the structure of interfacial water is perturbed already at prenucleation. The perturbation is caused by the screening of protein charge. However, this screening appears less effective in the presence of carbonate for mineralization than the screening caused by only Ca²⁺ ions, which can be explained by the simultaneous condensation of both CaCO₃ and Ca²⁺ to the protein, in line with the formation of PNCs at the protein interface. Our findings provide evidence for the presence of water in the nonclassical nucleation pathway for mineralization of sea urchin spine. The inferred nonclassical mineralization pathway is possibly relevant for other CaCO₃ biominerals.