Structure of Avian Thymic Hormone, a High-affinity Avian β-parvalbumin, in the Ca²⁺-free and Ca²⁺-bound States

Abstract

Originally isolated on the basis of its capacity to stimulate T-cell maturation and proliferation, avian thymic hormone (ATH) is nevertheless a parvalbumin, one of two β-lineage isoforms expressed in birds. We recently learned that addition of Ca²⁺-free ATH to a solution of 8-anilinonaphthalene-1-sulfonate (ANS) markedly increases ANS emission. This behavior, not observed in the presence of Ca²⁺, suggests that apolar surface area buried in the Ca²⁺-bound state becomes solvent-accessible upon Ca²⁺ removal. In order to elucidate the conformational alterations that accompany Ca²⁺ binding, we have obtained the solution structure of the Ca²⁺-free protein using NMR spectroscopy and compared it to the Ca²⁺-loaded protein, solved by x-ray crystallography. Although the metal ion-binding (CD-EF) domains are largely coincident in the superimposed structures, a major difference is observed in the AB domains. The tight association of helix B with the E and F helices in the Ca²⁺-bound state is lost upon removal of Ca²⁺, producing a deep hydrophobic cavity. The B helix also undergoes substantial rotation, exposing the side-chains of F24, Y26, F29, and F30 to solvent. Presumably, the increase in ANS emission observed in the presence of unliganded ATH reflects the interaction of these hydrophobic residues with the fluorescent probe. The increased solvent exposure of apolar surface area in the Ca²⁺-free protein is consistent with previously collected scanning calorimetry data, which indicated an unusually low change in heat capacity upon thermal denaturation. The Ca²⁺-free structure also provides added insight into the magnitude of ligation-linked conformational alteration compatible with a high-affinity metal ion-binding signature. The exposure of substantial apolar surface area suggests the intriguing possibility that ATH could function as a reverse Ca²⁺ sensor.

Introduction

The second-messenger role of Ca²⁺ in eukaryotic signal transduction pathways is largely mediated by EF-hand proteins^1–4. The human genome, for example, encodes 242 EF-hand family members⁵. Some of these, calmodulin being the archetype, have explicit regulatory activity, modulating the activities of effector proteins in a Ca²⁺-dependent manner. Other EF-hand proteins function as mobile intracellular Ca²⁺ buffers. Regardless of their precise role, all display the hallmark Ca²⁺-binding motif – a central ion-binding loop flanked by short amphipathic helical segments. The term “EF-hand” was inspired by the recognition that the spatial arrangement between these structural elements can be mimicked with the fingers of the right hand⁶.

Despite the general similarity of their metal ion-binding sites, EF-hand proteins exhibit broad variations in divalent ion affinity. We are exploring the physical and structural basis for these differences, using specific parvalbumin (PV) isoforms. Parvalbumins are small (M_r 12,000), vertebrate-specific EF-hand proteins^3,7,8. The PV family includes α- and β sub-lineages, distinguished by isoelectric point (pI > 5 for α) and lineage-specific sequence differences^9,10. Mammals express two isoforms, one from each lineage¹¹, that exhibit 49% sequence identity. Despite the sequence similarity, in 0.15 M NaCl at pH 7.4, rat α-PV binds Ca²⁺ with a standard free energy change 3.5 kcal mol^-1 more favorable than that of rat β-PV¹². Whereas Na⁺ competes, albeit weakly, for vacant EF-hand sites in both proteins, only the β isoform binds K⁺. Thus, when K⁺ replaces Na⁺ as the major solvent cation, the α isoform experiences an apparent increase in divalent ion affinity, binding Ca²⁺ a full 5.5 kcal mol⁻¹ more tightly than rat β-PV. Besides improving our understanding of this biologically important class of protein, an explanation for the disparity in binding affinity could furnish insight into protein-ligand interactions in general.

The PDB contains more than 20 high-resolution structures of Ca²⁺-bound parvalbumins – including carp β (5CPV¹³), leopard-shark α (5PAL)¹⁴, pike β (2PVB)¹⁵, rat α (1RWY)¹⁶, and rat β (1RRO)¹⁷. Their overall structural similarity – average RMSD < 1.0 Å for backbone atoms – suggested that variations in divalent ion-binding affinity might, in fact, reflect differences in the structures of the apo-proteins. Accordingly, solution structures were obtained for the Ca²⁺-free rat α- and β-parvalbumins^18,19. Ca²⁺ removal from the β isoform evidently provokes a substantial conformational rearrangement, implying that the attenuated divalent ion affinity may reflect the energetic cost associated with isomerizing the apo-protein. If correct, then Ca²⁺ binding should be accompanied by more muted structural alterations in high-affinity isoforms. That idea is supported by solution structural data for the Ca²⁺-free rat α isoform, which closely resembles the Ca²⁺-bound protein. Together, these findings implied a direct correlation between binding affinity and conformational similarity of the Ca²⁺-free and Ca²⁺-loaded states. The Ca²⁺-free parvalbumin structure described herein suggests that the correlation is not quite that straightforward.

Avian thymic hormone (ATH) was originally isolated on the basis of, and named for, its capacity to stimulate avian T-cell maturation and proliferation^20,21. Quite unexpectedly, sequence analysis of the purified protein revealed that it was a β-parvalbumin²². In fact, ATH is one of two β-PV isoforms expressed in chicken thymus tissue. The other is known as chicken parvalbumin 3²³, or CPV3, also believed to serve an endocrine role in the avian immune system²⁴. Both EF-hand sites in ATH qualify as high-affinity, or Ca²⁺/Mg²⁺, sites. The Ca²⁺-binding constants in 0.15 M NaCl, at pH 7.4, are 2.4 × 10⁸ and 1.0 × 10⁸ M⁻¹; the corresponding Mg²⁺ constants are 2.2 × 10⁴ and 1.2 × 10⁴ M^{−1 25,26}.

We recently learned that addition of Ca²⁺-free ATH to a solution of the hydrophobic probe 8-anilinonaphthalene-1-sulfonate (ANS) produces a large increase in fluorescence quantum yield and a pronounced blue shift²⁷. Addition of the Ca²⁺-bound protein is without effect. This behavior, which is not observed with the Ca²⁺-free forms of either rat α-PV or CPV3, implies that the conformations of the Ca²⁺-free and Ca²⁺-bound protein differ substantively. This finding conflicts with the proposed correlation between divalent ion affinity and conformational similarity of the Ca²⁺-free and Ca²⁺-loaded states of high-affinity parvalbumin isoforms. In an effort to resolve this issue and to delineate the conformational changes that occur in ATH upon Ca²⁺ binding, we have determined the solution structure of Ca²⁺-free ATH. To permit a more meaningful evaluation of the structural changes provoked by Ca²⁺ removal, we also report the crystal structure of Ca²⁺-bound ATH.

Results

Ca²⁺-bound ATH

The structure of Ca²⁺-bound ATH was solved by X-ray crystallography (Table 1). The crystal morphology (long needles) and space group (P1), its sensitivity to radiation damage, and the unit cell dimensions (containing 8 copies of the protein) made this task challenging. Ultimately, the experimental hurdles were overcome by employing the micro-diffraction beamline at NE-CAT 24-ID-E of the Advanced Photon Source. The structure was solved to a resolution of 1.95 Å, using a wedge-based data collection strategy, featuring a wide oscillation width of 4° per frame, as described in Materials and Methods.

Table 1.

X-ray diffraction data collection and refinement statistics^a

Wavelength (Å)	0.9792
Space group	P1
Unit cell parameters (Å, °)	a = 47.6, b = 51.6, c = 67.6 α = 87.4, β = 85.4, γ = 87.4
No. of molecules in unit cell	8
Diffraction resolution (Å)	50 – 1.95 (2.02 – 1.95)
No. of observations	196, 092
No. of unique reflections	45, 451
Redundancy	4.3 (4.3)
Completeness (%)	98.0 (96.8)
Mean I/σI	16.6 (3.2)
Rmerge	0.086 (0.416)
No. of protein atoms	6528
No. of water molecules	449
No. of glycerol molecules	3
No. of sulfate ions	4
No. of calcium ions	16
Rcryst	0.164 (0.206)
Rfreeb	0.221 (0.283)
Maximum-likelihood coordinate error (Å)	0.27
RMSD bond lengths (Å)c	0.004
RMSD bond angles (deg.)c	0.715
Ramachandran plotd
Favored (%)	99.4
Allowed (%)	0.6
Outliers (%)	0.0
Average B-factors (Å2)
Protein	21
Water	25
Glycerol molecules	34
Sulfate ions	67
Calcium ions	22
PDB accession code	3FS7

^aValues for the outer shell of data are given in parentheses.

^b5% random test set.

^cCompared to the parameters of Engh and Huber.⁶¹

^dThe Ramachandran plot was generated with RAMPAGE.⁶²

Ca²⁺-bound ATH displays the characteristic parvalbumin fold, consisting of six α-helices (labeled A-F in Fig. 1A) organized into two domains. The AB domain, spanning residues 1–38, includes the A and B α-helices and an extended loop. The remaining polypeptide chain forms the CD/EF domain, which includes the two EF-hand metal ion-binding motifs. The eight chains in the asymmetric unit of the crystal have very similar structures. Pairwise root mean square differences (RMSDs) for C^α atoms in the eight chains span the range 0.34 – 0.93 Å, with an average of 0.61 Å.

Fig. 1.

Tertiary structure of ATH. (A) Crystal structure of Ca²⁺-bound ATH. Superposition of the eight Ca²⁺-bound molecules in the asymmetric unit. (B) Solution structure of Ca²⁺-free ATH. An ensemble of the 20 lowest energy structures obtained with CYANA. (C) Surface rendering of the ensemble-average Ca²⁺-bound ATH structure. (D) Surface rendering of the ensemble-average Ca²⁺-free ATH structure, emphasizing the hydrophobic cavity produced by the displacement of the B helix. The views depicted in (C) and (D) were obtained by 90° counterclockwise rotation of the views in (A) and (B) about the vertical axis. This figure and Figs. 4 and 6 were produced with PyMOL ⁶³.

Ca²⁺-free ATH

Resonance assignments

Assignments were made using a series of triple-resonance experiments. CBCACONH and HNCACB spectra yielded tentative assignments for nearly all of the C^α and C^β nuclei. The CBCACONH experiment provides inter-residue correlations between the amide signal for the i^th residue, H^N(i), and the C^α and C^β nuclei in the (i-1)^th residue, C^α/C^β(i-1). The HNCACB spectrum provides the corresponding intra-residue correlation, as well as an inter-residue correlation for the majority of amide signals. The HNCA/HNCOCA and HNCO/HCACOCANH spectral pairs were used to confirm the backbone assignments and to resolve ambiguities. The HNCOCA experiment correlates H^N(i) with C^α(i-1). The HNCA experiment correlates H^N(i) with C^α(i) and, frequently, H^N(i) with C^α(i-1). Due to their narrower spectral windows, these experiments offer greater resolution of the C^α signals. The HNCO experiment correlates H^N(i) with C′(i-1), the backbone carbonyl of the preceding residue. The HCACOCANH correlates H^N(i) with C′(i) and, often, H^N(i) with C′(i-1). Aliphatic carbon assignments beyond C^β were made with the CCONH experiment, which correlates H^N(i) with all of the carbon nuclei in the side-chain of the i–1 residue. The aliphatic side-chain carbon assignments were complete, excluding the 18 carboxylates (Asp, Glu, and C-terminus), the seven carboxamides (5 Gln, 2 Asn), and the R75 guanidinino group.

Aliphatic ¹H assignments were made with HBHACONH, HCCONH, ¹⁵N-edited TOCSY-HSQC, and HCCH-TOCSY experiments. The HBHACONH spectrum correlates H^N(i) with H^α/H^β(i-1); the HCCONH spectrum correlates correlates H^N(i) with all of the carbon-bound protons in the i-1 side-chain; the ¹⁵N-TOCSY-HSQC spectrum provides the corresponding intra-residue correlations; and the HCCH-TOCSY spectrum provides intra-residue correlations between each aliphatic carbon nucleus and all of the carbon-linked protons. Assignments for H^δ and H^ε in Phe and Tyr were obtained from the HBCBCGCDHD and HBCBCGCDCEHE experiments, which correlate those protons with C^β. Proton assignments (side-chain and backbone) were 95% complete.

Fig. 2 displays the ¹H-¹⁵N HSQC spectrum of Ca²⁺-free ATH at 20 °C. Main-chain amide signals are observed for all residues except I2. The following pairs of resonances exhibit substantial overlap: D41/V33, G48/G98, Y26/L77, K44/D90, D92/D94, T32/I43, and D51/K107. The spectrum of the apo-protein is compared to the spectrum of the Ca²⁺-loaded protein, collected under identical conditions, in Supplemental Fig. S1.

Fig. 2.

¹H-¹⁵N HSQC spectrum of Ca²⁺-free chicken ATH in 0.15 M NaCl, 0.01 M Mes, 0.005 M EDTA, pH 6.0, at 20°C.

Solution structure of Ca²⁺-free ATH

The tertiary structure of Ca²⁺-free ATH was calculated with CYANA²⁸, using distance and dihedral angle restraints (Table 2). An ensemble of 20 low-energy conformers is displayed in Fig. 1B. The RMSD, relative to the ensemble average, is 0.75 Å for the backbone atoms (C^β, C^α, C′, O, and N), 1.12 Å for all heavy atoms. Table 2 lists additional structural quality statistics for the ensemble. 99.9% of the φ,Ψ combinations in the 20 conformers reside in allowed regions of the Ramachandran plot. The average NOE restraint violation is 0.026 Å, and there are no violations exceeding 0.5 Å in six or more of the structures. The average dihedral restraint violation is 0.59°, with no violations exceeding 5° in six or more structures.

Table 2.

List of restraints and statistical analysis for the Ca²⁺-free ATH solution structure

Number of experimental restraints
total NOEs	2355
Intraresidue	590
Sequential	588
Medium-range (1 < |i-j| ≤ 4)	587
Long-range (|i-j| > 4)	590
TALOS	74
Residual CYANA target function	3.47 ± 0.25
Restraint violations
NOE restraints (>0.1Å, 6 or more structures)	23
NOE restraints (>0.2Å, 6 or more structures)	12
NOE restraints (>0.3Å, 6 or more structures)	6
NOE restraints (>0.4Å, 6 or more structures)	3
Dihedral restraints (>5°, 6 or more structures)	0
RMSD from experimental restraints
NOE restraints (Å)	0.026
Dihedral restraints (deg.)	0.59
RMSD from idealized covalent geometry
bonds (Å)	0.0035
angles (deg)	0.57
Dihedral angles (deg)	21.7
improper angles (deg)	0.83
Coordinate RMSD from average structure (Å)
backbone (Cβ, Cα, C′, O, N)	0.75 ± 0.16
all heavy atoms	1.12 ± 0.16
Ramachandran plot (ensemble averages)
most favored regions (%)	69.9
allowed regions (%)	27.7
generously allowed (%)	2.3
disallowed (%)	0.1

¹⁵N relaxation data

Data were collected on Ca²⁺-free ATH at 20 °C. The R₁ and R₂ data are well accommodated by a two-parameter exponential decay model. Representative data are presented in Supplemental Fig. S2. R₁ and R₂ values for 90 of 107 amide vectors are plotted in Fig. 3A and Fig. 3B, respectively. The corresponding numerical values are tabulated in Supplemental Table S1.

Fig. 3.

Summary of ATH ¹⁵N relaxation data and internal mobility analysis. For reference, the positions of the six helices are indicated at the top of the figure. (A) R₁ values. (B) R₂ values. (C) Calculated R₂/R₁ ratios. The filled circles (●) represent amide vectors used in the calculating the mean. (D) {¹H}¹⁵N NOE values. (E) Order parameter (S²) determined by model-free analysis, as described in the text. Two of the residues required inclusion of a second order parameter (●), corresponding to motion on a slower timescale, to satisfactory model their relaxation behavior. (F) τ_e (○) and R_ex (●) values for residues displaying motion on a timescale exceeding 20 ps.

The rotational correlation time (τ_c) was estimated from the subset of amide vectors (Fig. 3C, •) exhibiting an R₂/R₁ ratio within one standard deviation of the mean. The data are well accommodated by a spherically symmetric rotational diffusion model, yielding a τ_c value of 6.68 ± 0.04 ns. Axially symmetric and fully symmetric models yielded insignificant reductions in χ².

The {¹H}¹⁵N NOE values (Fig. 3D) exhibit a high degree of uniformity. The mean NOE value is 0.80 ± 0.04.

Internal mobility and model-free analysis

Main-chain flexibility in Ca²⁺-free ATH was examined using the Lipari-Szabo model-free treatment^29,30. Relaxation data for 79 amide vectors were analyzed, employing a spherically symmetric diffusion model. The results are displayed in Fig. 3E and Fig. 3F, and the model-free parameters are listed in Supplemental Table S2.

The majority of signals (61/79) can be modeled with the overall rotational correlation time (τ_c) and a generalized order parameter (S²). Three amide vectors require a τ_e term to describe internal motion on the 20 ps - 10 ns timescale. An additional 14 require an R_ex term to describe internal motion on the μs-ms timescale. One amide (D4) can be accommodated only by inclusion of both τ_e and R_ex terms. Finally, residue 108 exhibits behavior consistent with motion on two timescales shorter than the overall rotational correlation time. The average order parameter for the apo-protein is 0.95, comparable to the value of 0.92 obtained for the Ca²⁺-free forms of rat α-PV and rat β-PV^18,19,31.

The data for eleven of the 90 vectors for which relaxation data were collected are not compatible with any of the five standard models, suggesting that those amides are undergoing more complex motions. The residues in question are A8, K9, S23, S37, F57, L67, K68, S72, S73, S79, and T82. Their positions in the Ca²⁺-free and Ca²⁺-bound states are displayed in Supplemental Fig. S3.

Comparison of the Ca²⁺-bound and Ca²⁺-free ATH structures

The ensemble-averaged Ca²⁺-free ATH structure (pink) has been superimposed on the crystal structure of the Ca²⁺-bound protein (green) in Fig. 4. Although the peptide backbones appear largely coincident, a major difference is observed in the region spanning residues 22 through 32. In the Ca²⁺-loaded structure, helix B packs snugly against the E and F helices. Upon removal of Ca²⁺, however, the B helix undergoes substantial reorientation, resulting in loss of contact with the CD-EF domain and creation of a deep cavity (Fig. 1C). Analysis with the CASTp server indicates that the hydrophobic cavity has a surface area of 554 Ų and volume of 817 ų.

Fig. 4.

Stereoview of the superimposed Ca²⁺-bound (green) and Ca²⁺-free forms (pink) of ATH.

A statistical comparison of the two structures is presented in Fig. 5. This distance difference matrix compares the inter-residue distances, measured between α carbon atoms, in the Ca²⁺-free and Ca²⁺-bound forms of ATH. Substantive structural differences are largely confined to residues 20 – 30. The pairwise RMSDs between the eight chains of the crystal structure and the twenty models of the Ca²⁺-free structure span the range 3.1 – 4.4 Å, with an average of 3.8 ± 0.3 Å (C^α atoms). This range is substantially larger than the RMSDs for the individual chains of either the crystal structure (0.6 Å) or the NMR ensemble (1.1 Å), indicating significant conformational differences between the Ca²⁺-bound and Ca²⁺-free structures. In fact, the main structural differences between the two forms of the protein are limited to the AB domain (Fig. 4, Fig. 5, and Supplemental Fig. S4).

Fig. 5.

Distance difference matrix. The changes in inter-residue distance (in Å) that accompany Ca²⁺ removal were calculated with CNS. The quantity r_ij (Ca²⁺-free) – r_ij (Ca²⁺-bound) is displayed as a contour plot, where r_ij(Ca²⁺-free) is the C^α-C^α distance for residues i and j in the Ca²⁺-free structure, and r_ij(Ca²⁺-bound) is the corresponding distance for the Ca²⁺-bound structure. The differences were averaged over all possible combinations of the eight chains of the X-ray ensemble and the 20 chains of the NMR ensemble.

A detailed stereoview of the Ca²⁺-free protein in the vicinity of the B helix is displayed in Fig. 6. In panel B, the path of the main-chain in the Ca²⁺-bound protein is also depicted (in magenta), which serves to emphasize both the magnitude of the structural changes and their highly localized nature. The largest displacement of the main chain (13 Å) in response to Ca²⁺ removal occurs at Y26, positioned at the N-terminus of helix B. In addition to moving away from the CD-EF domain, the B helix also undergoes a significant rotation. As a consequence, the side-chains of F24, Y26, F29, F30 – normally part of the hydrophobic core – become solvent accessible (Supplemental Fig. S5). In fact, the reorientation of helix B opens a window on the hydrophobic core of the protein. In addition to the residues mentioned above, the following core residues experience some increase in solvent accessibility: I11, L15, F66, L67, F70, F85, and L105.

Fig. 6.

Structural changes accompanying removal of Ca²⁺ from ATH. (A) Stereoview of the hydrophobic cavity created by the movement of helix B. (B) This stereoview of the region of interest compares the path of the polypeptide backbone in the Ca²⁺-free and Ca²⁺-bound (magenta) states. Note that large displacements of the polypeptide chain are confined to the vicinity of the AB loop and the B helix.

A previous analysis of thermal stability indicated that ΔC_p, the increase in heat capacity upon denaturation, is lower for Ca²⁺-free ATH than for Ca²⁺-free rat β-PV²⁷. Because the magnitude of ΔC_p primarily reflects increased exposure of apolar surface area in the unfolded state, the decreased value obtained for ATH suggests that the increase in apolar surface area upon denaturation is atypically small. Exposure of apolar surface area in the native state, as implied by the solution structure, would provide a rationale for the lower ΔC_p value.

As noted above, 11 amide vectors exhibit complex dynamics. Their locations are indicated in Supplemental Fig. S3. Residues S23 and S37 bracket the B helix. Thus, it is reasonable that their conformations might sample alternative conformations on a timescale not compatible with the model-free analysis. In the Ca²⁺-bound state, L67 contacts the side-chains of F24, F29, and F30, helping to anchor the extended D/E loop to the AB domain. The displacement and rotation of the B helix that accompanies Ca²⁺ removal eliminates those contacts. The anticipated increase in flexibility and mobility of the D/E loop may explain why the dynamics of the amide vectors for L67, K68, S72, and S73 cannot be captured by the model-free treatment. Similarly, S79 and T82 are located at the N-terminal end of the E helix, which associates with the B helix in the Ca²⁺-loaded protein. The departure of the B helix upon Ca²⁺ removal and resulting disruption of those contacts could significantly heighten flexibility. A8 and K9 are located at the N-terminal end of the A helix, which is proximal to the C-terminal end of helix B. Their conformations may be sensitive to the pronounced movement of the B helix.

Discussion

Interest in the structure of Ca²⁺-free ATH was originally sparked by the observation that addition of the Ca²⁺-free protein to a solution of ANS produced a marked increase in the quantum yield of the fluorophore. This behavior, not observed with the Ca²⁺-bound protein, suggested that the apo-protein presents a hydrophobic surface with which the fluorescent probe can interact. Presumably, ANS associates with the apolar side-chains that are exposed by the reorientation of the B helix that accompanies Ca²⁺ removal.

Besides ATH, structural data have previously been acquired for the Ca²⁺-free states of rat α-PV and rat β-PV^18,19. The EF-hand sites in the rat α isoform are typical high-affinity (Ca²⁺/Mg²⁺) sites, with average binding constants, in buffered saline at pH 7.4, of 1.2 × 10⁸ and 1 × 10⁴ M⁻¹ for Ca²⁺ and Mg²⁺, respectively. The rat β isoform exhibits substantially lower divalent ion affinity¹². For example, the total free energy change that accompanies Ca²⁺ binding at 25 °C is 3.5 kcal/mol less favorable in buffered saline at pH 7.4. Whereas the solution structure obtained for Ca²⁺-free rat α closely resembles the Ca²⁺-bound form, the unliganded and bound forms of rat β-PV exhibit some substantive differences. In particular, the hydrophobic core of the protein evidently undergoes significant reorganization when Ca²⁺ binds. This observation suggested that the attenuation of divalent affinity observed for the rat β isoform may reflect the energetic cost of the conformational change attendant to metal ion binding. More generally, the structural results for the two rat parvalbumin isoforms imply a direct correlation between divalent ion affinity and the similarity of the unliganded and bound states.

The divalent ion affinity measured for ATH eclipses that of rat α-PV. Thus, we would anticipate that the apo- and Ca²⁺-loaded forms of the protein would be largely indistinguishable. However, as described above, the conformations exhibit a marked difference in the vicinity of the B helix. It would seem that the aforementioned correlation of divalent ion affinity with the conformational similarity of the apo- and Ca²⁺-bound protein requires amendment.

In ATH, the differences between the apo- and Ca²⁺-bound forms are confined to the B helix. The conformation of the CD-EF domain is largely independent of the divalent ion-binding status. In the rat β isoform, by contrast, the conformational change associated with Ca²⁺ binding apparently perturbs the packing of the hydrophobic core in the CD-EF domain as well as the AB/CD-EF interdomain interaction. Thus, the structural rearrangements that accompany Ca²⁺ binding/removal differ qualitatively in ATH and rat β-PV. In the former, they can be reversed by displacement and rotation of the B helix alone, with minimal involvement of the rest of the molecule. The fact that the conformational difference is localized in ATH may explain why the molecule retains its high affinity for divalent ions, even though binding/dissociation is accompanied by a remarkable structural alteration. In rat β-PV, by contrast, reversal of the conformational alterations linked to metal ion binding involves coordinated movement throughout the molecule.

Under resting state conditions in vivo, assuming intracellular Ca²⁺ and Mg²⁺ concentrations of 10⁻⁷ and 10⁻³ M, respectively, ATH would reside predominantly in the Mg²⁺-bound form. Because the Ca²⁺-free state studied here is negligibly populated in the cell, one could argue that the structural difference observed between the unliganded and Ca²⁺-bound forms of ATH has doubtful physiological significance. Interestingly, however, the impact of the Mg²⁺-bound protein on ANS fluorescence is comparable to that of the apo-protein²⁷. This result suggests that the conformation of the B helix in the Mg²⁺ bound protein may resemble that observed in the Ca²⁺-free protein, with comparable exposure of hydrophobic surface area. If correct, then ATH could conceivably function as a reverse Ca²⁺ sensor. In the Mg²⁺-bound (resting) state, exposure of apolar surface would permit interaction with a target molecule. Upon exchange of Mg²⁺ for Ca²⁺ following a transient increase in cytosolic Ca²⁺ concentration, the apolar surface would be hidden, and the target molecule would be released. Clearly, structural data for the Mg²⁺-loaded protein would be highly desirable, and efforts to obtain those data are underway.

Materials and Methods

Protein expression and purification

The chicken ATH coding sequence, optimized for expression in E. coli, was obtained from Genscript (Piscataway, NJ) and cloned between the Nde I and Bam HI sites of pET11a. Bacteria harboring the ATH-pET11 construct were cultured at 37 °C in ¹⁵N- or ¹³C, ¹⁵N-labeled Spectra 9 medium (Cambridge Isotope Laboratory, Andover, MA), supplemented with ampicillin (100 μg/mL). IPTG (0.25 mM) was added when the absorbance of the culture at 600 nm reached 0.6. After an additional 20 h, the bacteria were collected by centrifugation. The protein was purified as described previously²⁶. Each liter yielded 20–25 mg of protein, with purity exceeding 98%.

Crystallization

ATH was crystallized from ammonium sulfate at 21 °C by vapor diffusion. Prior to crystallization, the protein was dialyzed extensively against 0.05 M MES, pH 6.0, containing 10 μM Ca²⁺. Hanging drops were prepared by combining 40 μL of ATH solution (10 mg/mL) with 40 μL of 70% (NH₄)₂SO₄, 0.05 M MES, pH 6.0. Thin rod-like crystals, with a maximum thickness of 20 μm (Fig. 7), formed in 4–8 weeks. The largest of the crystals were prepared for low-temperature data collection by soaking in 15% glycerol. They were then picked up with Hampton loops and plunged into liquid nitrogen.

Fig. 7.

ATH crystal morphology. One of the crystals used for data collection at the NE-CAT 24-ID-E microbeam line of the Advanced Photon Source.

X-ray diffraction data collection and structure determination

Initial X-ray diffraction experiments were performed at beamline 4.2.2 of the Advanced Light Source. The resulting data set, having a high-resolution limit of 2.7 Å, was used for preliminary assessment of diffraction quality and molecular replacement calculations. The space group was P1. The unit cell, which contained eight molecules, had these parameters: a = 48 Å, b = 52 Å, c = 68 Å, α = 87°, β = 85°, γ = 87°. The corresponding V_m is 1.8 ų/Da, and the solvent content is low, just 31 %.

In an attempt to obtain higher resolution data, additional experiments were performed at the NE-CAT 24-ID-E beamline of the Advanced Photon Source. This beamline is optimized for analysis of diminutive crystals. It employs a highly focused 20 × 100 micron X-ray beam, which may be shaped down to 5 microns, for data collection. Additionally, it is equipped with a Maatel MD-2 micro-diffractometer to facilitate visualization and centering of small crystals.

The crystals analyzed at 24-ID-E had typical dimensions of 300 × 20 × 20 μm. The X-ray beam size was therefore reduced to 20 microns by inserting an aperture. This adjustment substantially reduced background scatter, dramatically increasing the signal-to-noise ratio of the resulting data. Small crystals are susceptible to primary X-ray radiation damage from the highly focused X-rays generated by a third generation synchrotron source^32–35. To minimize radiation damage, wedges of data were collected from different parts of the crystal, then merged using HKL2000³⁶ to produce complete data sets. Several data collection strategies were investigated by varying beam attenuation, oscillation angle, and wedge size. The distance of the ADSC Q315 detector was set to 250 mm and the detector 2θ angle was zero.

The best data set was obtained by merging the reflections from five non-overlapping wedges, each wedge consisting of twenty frames collected with an oscillation width of 4 ° per frame, exposure time of 2 seconds per frame, and beam attenuation of 43% transmission at 12662 eV. Thus, the merged data set consisted of a continuous 400° of data. The resulting data set is 98 % complete to 1.95 Å resolution, with an average redundancy of 4.3 (Table 1).

Molecular replacement calculations were performed with MOLREP³⁷ using a search model derived from the coordinates of the 1.05 Å resolution structure of rat α-parvalbumin (PDB code 1RWY¹⁶). A solution having eight molecules in the asymmetric unit was obtained. The model was improved with iterative rounds of model building in COOT³⁸ and refinement in PHENIX³⁹. Initially, NCS restraints and group B-factors were used, but analysis of R_free indicated that NCS restraints could be released, and the use of individual B-factors with TLS was appropriate.

NMR sample preparation

Samples were prepared with uniformly labeled ¹⁵N-labeled, uniformly ¹³C,¹⁵N-labeled ATH, or uniformly labeled ¹⁵N-ATH that was also fractionally labeled (15%) with ¹³C. In each case,18 mg of protein – sufficient to yield a 0.5 mL sample at 3 mM — was concentrated to 5 mL by ultrafiltration, then dialyzed for 48 h, at 4°C, against 4 L of 0.15 M NaCl, 0.025 M Hepes, 5.0 mM EDTA, pH 7.4. Dialysis was continued for 48 h against 0.15 M NaCl, 0.01 M Mes, 5.0 mM EDTA, pH 6.0. Following the addition of buffer prepared in D₂O (0.1 volume) and 10% sodium azide (to a final concentration of 0.1%), the solution was concentrated to 0.5 mL. The resulting sample was loaded into a 5 mm Shigemi microcell (Shigemi, Inc., Allison Park, PA).

NMR spectroscopy

All data were acquired at 20°C on a Varian INOVA 600 MHz spectrometer equipped with a triple-resonance cryoprobe. ¹H chemical shifts were referenced relative to DSS; ¹³C and ¹⁵N shifts were referenced indirectly, employing the ¹H/X frequency ratios. Data were processed with NMRPipe⁴⁰ and analyzed with Sparky⁴¹.

¹H,¹⁵N-HSQC spectra were collected using ¹H and ¹⁵N windows of 8380 Hz (2048 complex points) and 2430 Hz (256 complex points), respectively. ¹H,¹³C-HSQC spectra were collected on the aliphatic carbon region, using ¹H and ¹³C dimensions of 8380 Hz (2048 complex points) and 12060 Hz (256 complex points), respectively. Backbone ¹³C assignments were made with these pairs of 3D experiments: HNCA⁴² and HN(CO)CA⁴³; HNCACB^44,45 and CBCA(CO)NH⁴⁶; and HNCO⁴² and HCACOCANH⁴⁷. In each case, the ¹H and ¹⁵N dimensions spanned 8380 Hz (2048 complex points) and 2430 Hz (80 complex points), respectively. The number of complex points and ¹³C window width were 60 and 2062 Hz (HNCO, HCACOCANH), 160 and 4525 Hz (HNCA, HNCOCA), or 140 and 12065 Hz (HNCACB, CBCACONH). Aliphatic carbon assignments beyond C^β were made with the CCONH⁴⁸ experiment, acquired with the ¹H and ¹⁵N settings just listed and a ¹³C window of 12065 Hz (160 complex points). Stereospecific assignments for the methyl groups in valine and leucine were obtained by analysis of the splitting patterns observed in the ¹H,¹³C-HSQC spectrum of fractionally labeled protein, as described by Neri et al.⁵² Those data were acquired with a ¹H window of 8380 Hz (2048 complex points) and ¹³C window of 3015 Hz (1024 complex points).

Aliphatic ¹H assignments were made with HBHACONH, HCCONH¹⁵N-edited TOCSY-HSQC⁴⁹, and HCCH-TOCSY⁵⁰ experiments. The first three were acquired with an 8380 Hz spectral window in both ¹H dimensions (256 complex points in t₁, 2048 in t₃) and an ¹⁵N window of 2430 Hz (80 complex points). The HCCH-TOCSY spectrum was acquired with identical ¹H settings and a 12065 Hz ¹³C window (108 complex points). The H^δ and H^ε resonances in Phe and Tyr were assigned with the HBCBCGCDHD⁵¹ and HBCBCGCDCEHE⁵¹ experiments. Those spectra were collected with a ¹H window of 8380 Hz (2048 complex points) and a ¹³C window of 4525 Hz (64 complex points).

Solution structure calculations

For the collection of NOE-based distance restraints, 3D ¹⁵N-edited and ¹³C-edited NOESY-HSQC⁵³ data sets were collected on ¹³C, ¹⁵N-labeled protein, employing mixing times of 125 ms and 100 ms, respectively. Cross peaks were picked manually and integrated in Sparky. φ,Ψ dihedral angle restraints were obtained for 77 residues using TALOS⁵⁴. Structure calculations were performed with CYANA²⁸. All NOE assignments were made automatically in CYANA. The quality of the final structures was analyzed using PROCHECK⁵⁵.

¹⁵N relaxation data

R₁, R_2, and {¹H}¹⁵N NOE data were collected on ¹⁵N-labeled protein employing Varian BioPack pulse sequences. R₁ data were acquired with these relaxation delays: 50, 100, 150, 250, 350, 450, 600, 800, 1000, and 1200 ms. R₂ data were acquired with these delays: 10, 30, 50, 70, 90, 110, 130, 150, 170, and 190 ms. Replicate data were collected at three delay values to evaluate experimental uncertainty. To measure the steady-state heteronuclear {1H}15N-NOE, HSQC spectra were collected, with and without proton saturation (3.0 s), employing a total recycle delay period of 5.0 s. Duplicate experiments provided an estimate of the experimental uncertainty.

Signal intensities were measured automatically in Sparky. R₁ and R₂ values were estimated by fitting the intensities to a two-parameter single-exponential decay, in Origin, version 7.5 (OriginLab). The NOE values were calculated from the ratio of the signal intensities in the presence and absence of proton saturation.

The relaxation data were analyzed with Tensor2⁵⁶. An overall rotational correlation time (τ_c) was estimated from the subset of amide vectors having R₂/R₁ values within one standard deviation of the mean value⁵⁷. The isotropic rotational diffusion model yielded a τ_c = 6.68 ± 0.04 ns, which corresponds to a rotational diffusion coefficient of 2.50 × 10⁷ s⁻¹. Axially symmetric and fully anisotropic models did not yield significant improvements in χ².

Internal mobilities were examined with the Lipari-Szabo model-free formalism^29,30. Tensor2 employs the five models suggested by Clore et al.^58,59 and the model selection strategy described by Mandel et al.⁶⁰.

Abbreviations

ANS

8-anilinonaphthalene-1-sulfonate

ATH

avian thymic hormone

CD site

parvalbumin metal ion-binding site flanked by the C and D helices

CPV3

chicken parvalbumin 3

DSS

sodium 2,2-dimethyl-2-silapentane-5-sulfonate

EDTA

ethylenediaminetetraacetic acid

EF site

parvalbumin metal ion-binding site flanked by the E and F helices

Hepes

4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid

HSQC

heteronuclear single-quantum coherence

Mes

2-(N-morpholino)ethanesulfonic acid

NMR

nuclear magnetic resonance

NOE

nuclear Overhauser effect

NOESY

NOE spectroscopy

PV

parvalbumin

R₁

longitudinal relaxation rate (1/T₁)

R₂

transverse relaxation rate (1/T₂)

RMSD

root-mean-square-difference

S²

generalized Lipari-Szabo order parameter

TALOS

torsion angle likelihood obtained from shifts and sequence similarity

τ_c

overall rotational correlation time

τ_e

internal correlation time

R_ex

rate constant for μs/ms motion resulting from chemical or conformational exchange