The Ca²⁺ response of a smart forisome protein is dependent on polymerization

2.2.1 CD spectroscopy

CD spectroscopy allows the assignment of secondary structures based on empirical spectral properties, but comparisons with databases of solved protein secondary structures are used to improve accuracy.¹⁷ In the case of forisome proteins such as MtSEO-F1, the secondary structural elements of closely related proteins have not yet been determined, so the CD spectra can only provide estimates of the relative abundance of different secondary structures. Accordingly, CD data were used to detect changes in the secondary structure of MtSEO-F1 dimers in the presence and absence of 0.5 mM Ca²⁺. We first confirmed that the buffer solution did not influence the CD spectra recorded in the range 200–260 nm (Figure S2). The curves representing MtSEO-F1 dimers in the absence and presence of Ca²⁺ were similar, with minima at 208 and 222 nm, which signifies the presence of α-helices (Figures 2 and S3). The predicted secondary structure was composed of 37% disordered regions, 14% β-turns, 25% β-sheets, and 24% α-helices. Overall, these data indicated that the secondary structure of MtSEO-F1 dimers is not affected by the presence of Ca²⁺. This makes it unlikely that changes in secondary structure at the level of individual dimers accumulate to trigger the remarkable Ca²⁺-induced conformational changes observed in whole forisomes. However, this left the possibility that Ca²⁺ may influence the overall fold of the dimers, so we used MST to investigate changes in tertiary and quaternary structure following the addition of Ca²⁺.

2.2.2 Microscale thermophoresis

MST detects changes in the spherical shape, ionic loading or water shell of soluble proteins by measuring their diffusion relative to an interaction partner within a temperature gradient. By gradually changing the concentration of the interaction partner, we can determine the binding constant and affinity of the protein for this partner. We labeled purified MtSEO-F1 dimers with the red fluorescent dye NT-647-NHS. The protein concentration was kept constant at 182.5 nM while we altered the Ca²⁺ concentration in the range 0–501 μM. These concentrations were chosen to meet and exceed the threshold Ca²⁺ concentration (20–50 μM) that induces MtSEO-F1 artificial forisomes to undergo a conformational change.⁴ Before each MST measurement, all capillaries were scanned to confirm a similar fluorescence level and to ensure that adhesion to the capillary material did not hinder the free diffusion of proteins (Figure S4).

The MST results are presented as the fluorescence ratio before and after thermophoretic movement plotted against increasing Ca²⁺ concentrations (Figure 3). The values in the range of Ca²⁺ concentrations that promote forisome conformational changes varied from 886‰ at 27 μM Ca²⁺ to 885‰ at 62 μM Ca²⁺. The binding of Ca²⁺ to MtSEO-F1 dimers should produce a sigmoid curve, but we observed no change in normalized fluorescence as the Ca²⁺ concentration increased, so we were unable to determine the dissociation constant. In addition, we observed no measurable change in the tertiary or quaternary structure of the MtSEO-F1 dimers in response to Ca²⁺.

Taken together, the CD and MST data suggested that MtSEO-F1 dimers do not associate with Ca²⁺. As stated above, a Ca²⁺-binding site has yet to be identified empirically or in silico in the SEO-F protein family despite the presence of a potential thioredoxin fold domain that could function as a platform for low-affinity and high-capacity Ca²⁺ adsorption, as described for calsequestrin in mammalian muscle cells.^{14, 18} However, calsequestrin features three thioredoxin fold domains with 24%–29% negatively charged amino acids in each,¹⁹ whereas MtSEO-F1 has one potential thioredoxin fold domain with 18% negatively charged amino acids. Our results also provided no empirical evidence that Ca²⁺ associates with MtSEO-F1 dimers, almost certainly excluding the presence of an undiscovered Ca²⁺-binding site.

2.3.1 MtSEO-F1 dimers adsorbed on a gold surface

Given that forisomes clearly undergo a striking conformational change triggered by Ca²⁺, cation association must instead occur at a higher level of forisome substructure. Such an interaction could disturb the arrangement of filaments and fibrils within the forisome body, possibly by generating repulsive Coulomb forces.²⁰ To test this idea experimentally, we adsorbed MtSEO-F1 dimers onto a gold surface to generate a continuous protein film, in which large agglomerates of the protein are likely to form. These films were analyzed by PM IRRAS as a model system to investigate Ca²⁺ association at the level of MtSEO-F1 protein aggregates somewhere between the soluble dimer and the giant forisome body. Soluble MtSEO-F1 protein in D₂O electrolyte solution one (pD 5.6) was adsorbed onto a gold surface, favoring intermolecular interactions and thus the inherent tendency of MtSEO-F1 to form aggregates.² This allowed us to analyze MtSEO-F1 polymers that are less complex than whole artificial forisomes, focusing on their response to Ca²⁺ in terms of secondary structure, hydration, and potential effects on functional groups in the agglomerated protein.

The background-corrected PM IRRA spectra of the MtSEO-F1 film over the range 1,800–1,300 cm⁻¹ are shown in Figure 4a. Figure S5a shows the corresponding raw PM IRRA spectra. The most prominent band (centered at 1660 cm⁻¹) corresponded to the amide I′ mode that originates predominantly from ν(C═O) stretches of the peptide bond in the polypeptide backbone.²¹ This is sensitive to the secondary structure of the protein, and appeared broad and asymmetric for MtSEO-F1, reflecting the complex structure of this protein. Fourier self-deconvolution (FSD) and second derivative procedures were used to deconvolute the amide I′ mode (Figure S6). Deconvolution of the amide I′ band revealed six individual modes corresponding to different secondary structures. Two modes were assigned to β-sheets (1,694 and 1,625 cm⁻¹), and one was assigned to α-helices (1,654 cm⁻¹). The assignment of the position of the amide I′ mode to other structural elements (e.g., β-turns or random coils) was recently found to be more complex^{22, 23} than reported previously.²⁴ Thus, the modes at 1,681 and 1,639 cm⁻¹ were assigned to “other structural elements” because they represent nonhelical and non-β-sheet structures. The sixth band (at 1,669 cm⁻¹) was not easy to assign, but uncommon structures such as 3₁₀ helices and π helices absorb infrared radiation in this spectral region,²⁴ and appear in response to hydration changes in α-helices.²⁵ Hydrophobic clusters in the folded polypeptide can upshift the amide I′ mode of α-helices toward higher wavenumbers²⁶ and the aggregation of helices is known to upshift the amide I′ mode (to 1,665–1,670 cm⁻¹).²⁷ According to the predicted structure of the MtSEO-F1 monomer,²⁸ α-helices are abundant in the domain that mediates protein dimerization driven by hydrophobic interactions² and α-helices may therefore aggregate due MtSEO-F1 dimerization. We therefore assigned the 1,669 cm⁻¹ band to hydrophobic α-helices.

The other mode characteristic of proteins is amide II, which represents CN stretches and NH in-plane bends in the peptide bond, and is sensitive to the isotopic substitution of the amide groups. The use of D₂O instead of H₂O therefore caused a downshift of the amide II′ mode (to 1,480–1,410 cm⁻¹).^{24, 29} Deconvolution of the amide II′ mode revealed two components centered at 1,467 and 1,448 cm⁻¹, respectively. Carboxylate groups present in the MtSEO-F1 side chains added two further modes, with ν_s(COO⁻) stretches centering at 1,405 cm⁻¹ and ν_as(COO⁻) stretches forming a broad band centered at 1,540 cm⁻¹. This indicated that carboxylate residues in amino acid side chains were deprotonated.

Several spectral changes were observed when the MtSEO-F1 film was exposed to Ca²⁺. The stretching modes of carboxylate groups were the most significantly affected by the interaction with Ca²⁺ (Figure 4). The ν_as(COO⁻) mode was attenuated and could not be distinguished from the background spectrum (Figure S5a). In contrast, the ν_s(COO⁻) mode became more intense and diverged into two components at 1,420 and 1,386 cm⁻¹. These changes indicated that Ca²⁺ ions interacted directly with the carboxylate groups of aspartic and glutamic acid residues because the position of carboxylate stretching modes partly depends on the coordination of the cation to this group.^{30, 31} The splitting of the ν_s(COO⁻) mode resulted from a concurrent downshift (1,386 cm⁻¹ band) and upshift (1,420 cm⁻¹ band), indicating the monodentate and bidentate coordination of Ca²⁺ by the carboxylate groups.³⁰ The simultaneous decreasing intensity of ν_as(COO⁻) and increasing intensity of ν_s(COO⁻) are likely to reflect a molecular scale order in the self-assembled protein film on the gold surface.³² According to the IRRAS surface selection rule, attenuation at one mode must be connected to enhancement at another because the transition dipole moment vectors of the ν_as(COO⁻) and ν_s(COO⁻) modes are perpendicular.³²

Further spectral changes were observed for the amide I′ mode. The bands assigned to α-helices (1,654 cm⁻¹) and to other secondary structures at 1,639 cm⁻¹ upshifted to 1,658 and 1,643 cm⁻¹, respectively. Exposing the MtSEO-F1 film to Ca²⁺ caused the mode at 1,669 cm⁻¹ to disappear. Ca²⁺ binding therefore appears to influence the hydrophobic environment of the dimers in the adsorbed state. Quantitative analysis of secondary structures in the amide I′ band in the presence and absence of Ca²⁺ (Figure ) revealed that the abundance of β-sheets increased from 20% to 23% when Ca²⁺ was added, in broad agreement with the CD data, whereas the α-helices represented by the mode at 1,654 cm⁻¹ increased from 30% to 38%. The presumably hydrophobic helices (mode 1,669 cm⁻¹) constituted ca. 20% of all structural elements of the adsorbed MtSEO-F1. These values differ from the CD results, but given the IRRAS surface selection rule, the proportions of secondary structures in the anisotropic MtSEO-F1 film may differ from those of the isotropic MtSEO-F1 protein in solution.³² In PM IRRAS, the intensity of the IR absorption bands depends not only on the concentration of the molecules but also on their average orientation in the film, which is determined by the angle between the electric field vector of the IR beam (normal to the surface) and the average orientation of the transition dipole moment of a given vibration mode in the adsorbed molecule. If the orientation is uniform, some modes may be enhanced while others are attenuated. The adsorption of forisome dimers resulted in the formation of ordered aggregates, so the proportions of different secondary structure values determined by PM IRRAS may not reflect the real composition due to the surface selection rule.

Based on the observed spectral changes, the abundance of hydrophobic helices dropped below 5% in response to Ca²⁺ association. This indicated a transition to hydrated α-helices, possibly contributing to the concurrent increasing abundance of hydrated α-helices and other structures. Ca²⁺ is likely to introduce large amounts of water into the system because it coordinates seven water molecules in its first hydration shell.³¹ Furthermore, Ca²⁺ association probably expanded the tertiary structure so that water gained access to formerly dehydrated areas.

To summarize, adsorbed MtSEO-F1 produced a well-ordered anisotropic film on the gold surface, and PM IRRAS data indicated that MtSEO-F1 aggregates interact with Ca²⁺ via the carboxylate groups of aspartic and glutamic acid residues, changing the hydration status of hydrophobic helical regions. Minimal changes in the overall content of the secondary structure elements indicated that the Ca²⁺ association in MtSEO-F1 aggregates does not require the adaptive refolding of secondary structures. Together with the CD and MST analysis of soluble MtSEO-F1 dimers, we have strong evidence that MtSEO-F1 can only associate Ca²⁺ in the form of higher-order aggregates or protein films.

2.3.2 MtSEO-F1 forisomes adsorbed on a gold surface

Finally, we adsorbed complete artificial MtSEO-F1 forisomes to the gold surface, allowing us to compare results obtained from soluble dimers, substructural aggregates and the fully-assembled protein polymer. This also permitted the first high-resolution analysis of potential secondary structural changes in the fully-assembled forisome in response to Ca²⁺. Artificial MtSEO-F1 forisomes expressed in yeast were purified by density gradient centrifugation. We confirmed the dispersion reaction in response to Ca²⁺ and observed the characteristic longitudinal contraction and lateral dispersion, accompanied by an increase in forisome volume⁸ (Figure 5a). We then adsorbed the forisomes onto a gold surface and collected PM IRRA spectra as described above for the adsorbed MtSEO-F1 dimers.

The spectra before and after exposure to Ca²⁺ are shown in Figure 5b, along with the deconvoluted signals (the corresponding raw spectra are shown in Figure S5b). Characteristic amide I′ and amide II′ signals were observed as broad and asymmetric modes centered at 1,659 and 1,445 cm⁻¹, respectively. FSD and second derivative procedures were used to deconvolute the amide I′ and II′ modes (Figure S7 shows the data for amide I′). The deconvoluted amide II′ band diverged into two modes centered at 1,465 and 1,445 cm⁻¹. The deconvoluted amide I′ band diverged into five modes that could be attributed to β-sheets (1,693 and 1,626 cm⁻¹), α-helices (1,661 cm⁻¹), and other structural elements (1,678 cm⁻¹ and 1,643 cm⁻¹). The ν_as(COO⁻) had two components at 1,598 and 1,540 cm⁻¹ in the absence of Ca²⁺. The ν_s(COO⁻) was centered at 1,405 cm⁻¹. The ν_s(COO⁻) signal was more intense than the ν_as(COO⁻) signal, suggesting that forisomes adsorbed on the gold surface are packed in an orderly manner. In the presence of Ca²⁺, new bands at 1,736 and 1,706 cm⁻¹ arose from nondissociated side chains of aspartic and glutamic acid residues^{27, 33} or denatured proteins.³⁴ Furthermore, the 1,598 cm⁻¹ ν_as(COO⁻) component upshifted to 1,610 cm⁻¹ whereas the 1,540 cm⁻¹ signal decreased in intensity. The signal for the ν_s(COO⁻) mode split into two distinct modes (1,425 and 1,380 cm⁻¹). These combined changes in the ν(C═O), ν_as(COO⁻), and ν_s(COO⁻) modes suggested the monodentate and bidentate coordination of Ca²⁺ by the side chains of aspartic and glutamic acid residues, as observed for the MtSEO-F1 dimer film.³⁰

In terms of secondary structural changes induced by Ca²⁺ association, the overall intensity of the amide I′ and amide II′ modes increased, most likely due to rearrangements within the protein film (Figure 5b). Despite this increase in band intensities, only minor changes were observed in the secondary structure composition (Figure 5c). The content of α-helices increased from 35% to 40% and the proportion of β-sheets decreased slightly from 18% to 15%. It is challenging to compare these values directly with the adsorbed MtSEO-F1 dimers because the PM IRRAS results strongly depend on the orientation of the protein relative to the substrate surface according to the IRRAS surface selection rule.³²

Given the available evidence, the association of Ca²⁺ with MtSEO-F1 forisome bodies does not appear to influence the secondary structure of the proteins radically. Accordingly, the contraction triggered by Ca²⁺ may explain the increase in absorbance. Furthermore, PM IRRAS clearly revealed that negatively charged amino acids facilitate the association between forisomes and Ca²⁺. Indeed, the mutation of two aspartate residues in the MtSEO-F1 sequence was recently shown to diminish the Ca²⁺ response of artificial forisomes.¹⁵ However, given the presence of 93 negatively charged amino acids in the MtSEO-F1 protein (14% of the primary sequence) together with the lack of a significant mode indicating the binding of Ca²⁺ in the PM IRRA spectra, it seems likely that additional, yet unidentified, negatively charged amino acids contribute to Ca²⁺ association.

PM IRRAS detected structural changes in both types of MtSEO-F1 aggregates. However, Ca²⁺ association caused different re-folding events in the MtSEO-F1 dimer aggregates and forisomes. The hydration of dehydrated protein areas and α-helices was detected in the dimer aggregates but not the forisomes, suggesting that helical structures are hydrated in forisome bodies. The influx of water into forisome bodies as a consequence of forisome contraction triggered by Ca²⁺ is well documented¹⁰ but water probably fills the space between forisome body substructures such as fibrils and is therefore barely detectable by PM IRRAS.³⁵ Films of MtSEO-F1 dimer aggregates therefore offer more insight into cation-related refolding events than forisome bodies. Nevertheless, we confirmed that Ca²⁺ association occurs only when MtSEO-F1 proteins form larger aggregates, and not in their soluble state. The interaction between forisome proteins and Ca²⁺ was previously attributed to larger patches of negatively charged carboxylate groups that allow several protein monomers to form one platform for Ca²⁺ association.³ Our data support this hypothesis and confirm that aggregation is required to develop responsive smart polymers based on forisome proteins.