Novel Zn2+-binding Sites in Human Transthyretin

Human transthyretin (TTR) is a homotetrameric protein involved in several amyloidoses. Zn2+ enhances TTR aggregation in vitro, and is a component of ex vivo TTR amyloid fibrils. We report the first crystal structure of human TTR in complex with Zn2+ at pH 4.6–7.5. All four structures reveal three tetra-coordinated Zn2+-binding sites (ZBS 1–3) per monomer, plus a fourth site (ZBS 4) involving amino acid residues from a symmetry-related tetramer that is not visible in solution by NMR. Zn2+ binding perturbs loop E-α-helix-loop F, the region involved in holo-retinol-binding protein (holo-RBP) recognition, mainly at acidic pH; TTR affinity for holo-RBP decreases ∼5-fold in the presence of Zn2+. Interestingly, this same region is disrupted in the crystal structure of the amyloidogenic intermediate of TTR formed at acidic pH in the absence of Zn2+. HNCO and HNCA experiments performed in solution at pH 7.5 revealed that upon Zn2+ binding, although the α-helix persists, there are perturbations in the resonances of the residues that flank this region, suggesting an increase in structural flexibility. While stability of the monomer of TTR decreases in the presence of Zn2+, which is consistent with the tertiary structural perturbation provoked by Zn2+ binding, tetramer stability is only marginally affected by Zn2+. These data highlight structural and functional roles of Zn2+ in TTR-related amyloidoses, as well as in holo-RBP recognition and vitamin A homeostasis.

Human transthyretin (TTR) 5 is a 55 kDa, ␤-sheet-rich homotetramer found in serum and cerebrospinal fluid. It par-ticipates in thyroxine (T4) transport (1) and in the binding of holo retinol-binding protein (holo-RBP) and is believed to reduce the glomerular filtration of the relatively small (21 kDa) holo-RBP (2). In humans, TTR and its more than 100 variants are involved in amyloid diseases such as senile systemic amyloidosis (SSA) and familial amyloidotic polyneuropathy (FAP), among others (3). In SSA, deposits composed of the wild-type TTR (WT-TTR) are found in the heart. Autopsies show that at least 25% of all individuals over 80 years have such deposits, which prove fatal in 10% of these cases (4). FAP is caused by the accumulation of amyloid fibrils formed by TTR mutants, mainly around basal membranes of Schwann cells, inside the endoneurium of peripheral nerves (5,6). Onset typically occurs in the third decade, with a life expectancy of ϳ10 years (6). Several drugs are being tested against TTR amyloidosis (3), but so far, the only treatment for FAP is liver transplantation (7).
A widely accepted hypothesis for TTR amyloidogenesis presupposes tetramer dissociation into a misfolded monomer which aggregates, giving rise to amyloid fibrils. In vitro these events can be triggered by mild acidification (pH 5.0 -4.0), although amyloid formation takes several days for completion (8). In vivo these events might occur inside lysosomes during TTR turnover (8). Recently, J. W. Kelly and co-workers (9) described an engineered monomer of TTR (M-TTR) in which the introduction of two bulky amino acid residues in the protein-protein interface (F87M/L110M) avoids oligomerization into dimers or tetramers. M-TTR aggregates in solution in the same pH range as WT-TTR, although for M-TTR this occurs on a much faster time scale because tetramer dissociation into monomers is the rate-limiting step in TTR aggregation (9).
Monomers (A-D) of TTR are composed of eight anti-parallel ␤-strands (A-H) arranged as a ␤-barrel (Greek key) with a short ␣-helix following strand E. The dimer (AB) is held together by hydrogen bonds involving the H and F strands located along the edge of each monomer. Two dimers (AB, CD) associate backto-back via hydrophobic interactions between residues of loops connecting strands AB and GH (10) (Fig. 1, A and B).
Numerous studies with WT and variants of TTR have attempted to unravel the structural rearrangements that precede amyloid formation (10 -13). Only minor global and local changes are observed in most of these structures (10), providing no conclusive structural clues for the possible mechanisms behind TTR structural conversion and amyloidogenesis. Recently, the x-ray structures of WT-TTR and two variants, I84A and I84S, under acidic conditions (and without Zn 2ϩ ) revealed structural disorder in the loop E-␣-helix-loop F region in relation to the structure solved at pH 7.0, indicating a possible participation of this segment of the protein in fibril formation (14,15).
Five decades of evidence support the interaction of TTR with Zn 2ϩ in the biological milieu but nothing is known about the accompanying structural correlates (16,17). The observation that concentrations of Zn 2ϩ in the plasma of healthy individuals are ϳ12-15 M (18), while the Zn 2ϩ :TTR apparent dissociation constant (K dapp ) is 1 M (19), suggests that TTR may circulate as a complex with Zn 2ϩ in plasma. High concentrations of Zn 2ϩ and Cu 2ϩ can trigger TTR amyloid formation in vitro (19,20), and chelating agents disrupt these amyloid structures (21). Zn 2ϩ was found to be the main mineral in ex vivo ocular amyloid deposits from FAP patients bearing the V30M mutation (22). Altogether, these data suggest that binding of Zn 2ϩ (or Cu 2ϩ ) might induce structural changes that lead to TTR amyloidogenesis. The aims of this study are to dissect the structural modifications caused by Zn 2ϩ binding and to explore their implications for TTR-mediated physiological recognition of holo-RBP, for TTR stability and for amyloidogenesis.

EXPERIMENTAL PROCEDURES
Proteins-Recombinant human WT-TTR and M-TTR were expressed and purified as previously described (23), using plasmids kindly provided by Dr. Jeffrey W. Kelly (9) (The Scripps Institute, San Diego, CA). Retinol-binding protein (RBP4) was obtained from Phoenix Pharmaceuticals (Burlingame, CA).
X-ray Crystallography-For crystallographic experiments, we used the double mutant F87M/L110M (M-TTR), a construct which displays a decreased self-association constant compared with WT-TTR (9). M-TTR crystals were obtained by the hanging-drop vapor-diffusion method at 20°C. Crystals were grown in ϳ5 days in 200 mM zinc acetate, 100 mM sodium citrate, 2.0 M ammonium sulfate, adjusted to the desired pH. X-ray diffraction images were collected in steps of 1°of oscillation per frame from single crystals cooled at 100 K. We used a wavelength of 1.458 Å (1.305 Å for the crystal obtained at pH 4.6, PDB ID 3DGD) at the MX2 beam line at LNLS (National Synchrotron Light Laboratory, LNLS, Campinas, Brazil (24)) and diffraction intensities were measured using a MarMosaic 225 detector (MAR Research GmbH, Norderstedt, Germany). Images were indexed, processed and integrated with Mosflm v. 7.0.3 (25). All M-TTR structures solved here belong to space group P2 1 . We scaled experimental intensities, solved, and refined the structure with the CCP4 v. 6.0.2 suite (26). The complex was solved by molecular replacement, using monomer A with only C␣ and C␤ atoms of PDB ID 1F41 (10) as search model, resulting in a solution for four monomers in the asymmetric unit. Model building and refinement were performed with Coot v. 0.5 (27) and the CCP4 v. 6.0.2 suite (26). Water molecules were added using Coot v. 0.5 (27).
Structural validation of the model performed with the CCP4 v. 6.0.2 suite (26) showed that almost all dihedral angles are within the favored region (for 3GRG (pH 7.5), 97.4% of residues are in favored regions, 2.0% are in allowed regions and 0.6% are in outlier regions; for 3GRB (pH 6.5), 98.5% of residues are in favored regions, 1.5% are in allowed regions and 0.0% are in outlier regions; for 3GPS (pH 5.5), 97.3% of residues are in favored regions, 2.0% are in allowed regions and 0.7% are in outlier regions; for 3DGD (pH 4.6), 99.1% of residues are in favored regions, 0.9% are in allowed regions and 0.0% are in outlier regions). Omit maps were generated with the CCP4 v. 6.0.2 suite (26). All crystallographic figures were generated with PyMol (28). Selected crystallographic parameters are found in supplemental Table S1.
Multiple Sequence Alignment and Conservation Analysis-A multiple sequence alignment (MSA) for the TTR family was obtained from the Pfam server (29). After the elimination of redundant and fragmented sequences, a total of 262 sequences remained. Overall site conservation and site-dependent conservation were calculated using Cϩϩ routines designed for statistical coupling analysis (SCA) (30).
NMR Spectroscopy-All NMR experiments were performed at 25°C on an 800 MHz Bruker III Avance spectrometer equipped with a 1 H[ 13 C/ 15 N] pulsed-field gradient probe (Bruker Corp., GmbH). Deposited sequential assignments of WT-TTR described elsewhere (BMRB 5507) (31) were used with adjustments due to differences between pH used in the deposited assignment and pH used in our experiments. To adjust assignment, we collected heteronuclear single-quantum coherence, transverse relaxation optimized spectroscopy (HSQC-TROSY) spectra using samples of WT-TTR 2 H/ 15 N (25 M) at pH 5.7, 6.5 (buffer: MES 25 mM, KCl 50 mM, pH 5.7 or pH 6.5) and 7.5 (buffer: deuterated Tris-HCl 25 mM, KCl 50 mM, pH 7,5). We used the pulse sequence TROSYETF3GPSI, Avance version 12142007 (Bruker Corp., GmbH) (32), with 8 transients in 1 H dimension with 1024 points of resolution and 16 transients in 15 N dimension with 256 points of resolution.
For titration, the same HSQC-TROSY spectra were recorded using WT-TTR 2 H/ 15 N (100 M) in deuterated Tris-HCl 25 mM, KCl 50 mM, pH 7.5. Titration experiments were performed by adding ZnCl 2 to the sample to attain final concentrations of 100, 200, 300, 400, and 900 M Zn 2ϩ . To the sample with 900 M ZnCl 2 , a small volume of a concentrated stock of EDTA was added, reaching 1 mM EDTA. All spectra were processed using the TopSpin suite (Bruker Corp., GmbH) and were analyzed using Computer Aided Resonance Assignment (CARA) (33). CSP values were calculated using assignments of the zinc-free WT-TTR spectrum as reference, and using Equation 1 (34), where ␦X ref is the chemical shift of the reference spectrum and ␦X the chemical shift of the spectrum at each concentration of ZnCl 2 for each amidic proton (NH) or amidic nitrogen (N). We chose 0.025 ppm as a threshold for considering a CSP to be significant.
For TROSY-HNCO we used the pulse sequence TRHNCOGP2H3D, Avance version 07042004 (Bruker Corp., GmbH) (35), with 8 transients in 1  Holo-RBP Interaction with TTR-Titrimetric assays of WT-TTR interaction with RBP bound to retinol (holo-RBP) were performed by fluorescence anisotropy measurements in an ISS-PC1 spectrofluorometer (ISS Inc, Champaign, IL) equipped with Glan-Thompson polarizers in L-format as previously described (36). Excitation and emission wavelengths were set at 330 and 460 nm, respectively. Assays were performed in 10 mM Tris-HCl, 150 mM NaCl, pH 7.2, 400 nM holo-RBP, and at the indicated ZnCl 2 concentrations. Dissociation constants were calculated as previously described (37), according to Equation 2, Ϫ (4*RBP t ϫ n ϫ TTR t )) 0.5 ))͞(2 ϫ RBP t )) (Eq. 2) where ␣ is the fraction of holo-RBP:WT-TTR formed, TTR t is the total WT-TTR concentration, K d is the dissociation constant, n is the number of holo-RBP binding sites present on the WT-TTR molecule and RBP t is the total RBP concentration used in the assay. The fraction of holo-RBP:WT-TTR formed was calculated according to Equation 3, where A obs is the measured anisotropy at a given TTR t , and A i and A f are the limiting initial and final anisotropies.
Thermodynamic Stability in the Presence of Zn 2ϩ -The influence of Zn 2ϩ over M-TTR and WT-TTR stability was probed by both chemical (urea) and physical (pressure) variables. M-TTR or WT-TTR (3.5 M) was incubated in 25 mM Tris-HCl, 50 mM KCl, pH 7.5 (25°C) in the absence or presence of ZnCl 2 (as stated in the legend of each figure). Pressure was raised to the desired value and equilibrated for 30 min before spectroscopic measurement. Tryptophan fluorescence emission was used as a sensor of the tertiary structure changes in the protein. The spectra were recorded by exciting samples at 280 nm and collecting emission from 300 -400 nm. The center of spectral mass values (ϽϾ ϭ ⌺n i F i /⌺F i , where F i is the fluorescence emitted at wavelength i ) were calculated in nm and plotted against pressure applied. From these plots, the extent of reaction (␣) was calculated and converted to a log scale to assess the free energy change of unfolding (⌬G unf ) and the standard volume change of folding (⌬V) according to Equation 4, where ␣ p is the extent of reaction at pressure p, n is the number of subunits (n ϭ 1 for M-TTR; n ϭ 4 for WT-TTR), p is pres-sure, R is the gas constant, C is the molar concentration of protein and T the temperature in K. The ⌬V is calculated from the slope of the curve and K u is the intercept on the ordinate (⌬G unf ϭ ϪRTlnK u ) (38).
M-TTR or WT-TTR (3.5 M) was incubated in 25 mM Tris-HCl, 50 mM KCl, pH 7.5 (25°C) for 72 h at varying concentrations of urea (0 -8 M). ZnCl 2 was added to desired concentration and tryptophan fluorescence emission (280 nm and collecting emission from 300 -400 nm) was immediately measured. Then, the center of spectral mass of tryptophan emission of each spectrum was calculated and converted into the extent of reaction (␣) as described under "Experimental Procedures" and plotted against [urea]. From these experiments, ⌬G unf was calculated according to Equation 5, where ␣ is the extent of reaction, n is the number of subunits, [D] is urea concentration, R is the gas constant, C is the concentration of protein in molar and T the temperature in K. The m unf is calculated from the slope of the curve and K u is the intercept on the ordinate (⌬G unf ϭ ϪRTlnK u ) (39).
Size-exclusion Chromatography (SEC)-SEC was performed in a TSK3000 (Tosoh Bioscience, Tokyo, Japan) column preequilibrated with 50 mM Tris-HCl, 100 mM KCl at pH 7.5 and with the indicated concentration of ZnCl 2 . Samples of 3.5 M WT-TTR diluted in the absence or in the presence of ZnCl 2 were injected into the column. TTR tetramers and monomers eluted at 7.5 and 9 min, respectively, as confirmed by previous column calibrations with RNase and bovine serum albumin as standards.

RESULTS
Crystallographic Determination of Zn 2ϩ -binding Sites in TTR-We solved the crystal structures of M-TTR in complex with Zn 2ϩ at pH 7.5, 6.5, 5.5, and 4.6 at resolutions of 1.9, 1.7, 1.8, and 1.3 Å, respectively, with C␣ r.m.s.d. among them less than 0.4 Å (Fig. 1A). Our attempts to solve the structure of WT-TTR in the presence of Zn 2ϩ failed due to the fragility of the crystals. M-TTR, which at the concentrations used crystallizes as a tetramer (9), forms robust crystals in the presence of Zn 2ϩ .
The Zn 2ϩ :M-TTR complex presented the typical fold of TTR at all pH values (Fig. 1A), and in Fig. 1B the structure at pH 7.5 obtained in the presence of Zn 2ϩ is superposed with that obtained at the same pH in the absence of Zn 2ϩ . However, despite their close resemblance, the structural alignment of Zn 2ϩ :M-TTR with Zn 2ϩ -free M-TTR shows large conformational differences in both secondary and tertiary structures, achieving 5 Å in some segments of the monomer (Fig. 1C). Fluctuating conformations in the loops connecting ␤strands BC, DE, and FG were previously described (10). The more prominent structural modification, apart from local changes in residues involved in Zn 2ϩ coordination, is a perturbation of the loop E-␣-helix-loop F region and residues nearby ( Fig. 1, B and C and Fig. 3A). These modifications occur in all four subunits of the tetramer (Fig. 1C).
Zn 2ϩ ions were coordinated at three Zn 2ϩ -binding sites (ZBS 1 to 3) in each monomer. All three sites displayed coordination spheres based on tight interactions of Zn 2ϩ with a imidazole nitrogen from His, the thiol group from Cys, and carboxylic acid groups from Glu or Asp side chains (Fig. 2). These elements are typical of selective Zn 2ϩ -binding sites in proteins (40). In fact, the structures revealed the presence of a fourth ZBS (ZBS 4), which involves residues from the symmetryrelated tetramer as explained bellow, and thus we assumed it was generated by crystallographic contacts (Fig. 2). NMR experiments confirmed the absence of ZBS 4 in solution (Figs. 4 and 5).
In ZBS 1, Zn 2ϩ is coordinated by the thiol group of Cys-10 (S␥ atom), the imidazole nitrogen from His-56 (N␦1 atom) and two water molecules (Fig. 2, upper left structure). In ZBS 2, Zn 2ϩ is bound to the imidazole nitrogens of His-88 (N⑀2 atom) and His-90 (N␦1 atom), the carboxylic acid group of Glu-92 side chain (O⑀1 atom) and one water molecule (Fig. 2, upper right structure). ZBS 1 and ZBS 2 presented this organization at all pH values. ZBS 3 consists of one water molecule, the imidazole nitrogen of His-31 (N␦1 or N⑀2 atom), the carboxylic acid group of Asp-74 (O␦1 or O␦2 atom) and the carboxylic acid group of Glu-72 ([O⑀1 or O⑀2 atom], Fig. 2, lower left structure). However, in some subunits of structures obtained at pH 7.5, 6.5, 5.5 and in all subunits of the structure obtained at pH 4.6, Glu-62 replaces either Glu-72 or the water molecule (ZBS 3Ј, Fig. 2, lower middle structure). In ZBS 3, His-31, Glu-72, and Asp-74 are from the same asymmetric unit (H31 AU , E72 AU , D74 AU , Fig.  2, lower middle structure). In ZBS 3Ј, His-31 and Asp-74 are both from the same monomer in the asymmetric unit (H31 AU and D74 AU , Fig. 2, lower middle structure), while Glu-62 is from a symmetry-related TTR molecule (E62 symm , Fig. 2, lower middle structure). In solution, ZBS 3 might be composed of His31 AU , Asp72 AU , Glu74 AU , and solvated accordingly. ZBS 4 is composed of two residues from the symmetry-related tetramer (H31 symm and D74 symm , Fig. 2, lower right structure), while the carboxylic acid group of the Glu-61 side chain is from the asymmetric unit (E61 AU , Fig. 2, lower right structure). Because it includes two residues from a symmetry-related tetramer, ZBS 4 is unlikely to exist as a functional Zn 2ϩ -binding  (9)) with Zn 2ϩ -bound M-TTR at pH 7.5 (PDB ID 3GRG). Major Zn 2ϩ -dependent differences are seen in the region of the ␣-helix and F-loop (residues 74 -90).

FIGURE 2. Details of the three Zn 2؉ -binding sites in TTR (ZBS 1-3), and a fourth site (ZBS 4) that is formed by amino acid residues from a symmetry-related tetramer.
Close-up views of Zn 2ϩ -binding sites and schematic diagrams of Zn 2ϩ coordination spheres. ZBS 1 is composed of Cys-10, His-56, and two water molecules (red spheres). ZBS 2 is composed of His-88, His-90, Glu-92, and one water molecule. ZBS 3 involves His31 AU , Asp74 AU and Glu72 AU (at higher pH values) or Glu62 symm (at pH4.6, ZBS 3Ј). ZBS 4 involves His31 symm , Asp74 symm , and Glu61 AU in addition to water molecules (AU, asymmetric unit; symm, symmetry-related unit). Meshes are electron densities from 2 F obs -F calc omit maps contoured at 1. The electron density displayed is limited to within 1.5 Å of the residues. Orange residues in ZBS 3 and ZBS 4 represent residues from neighboring tetramers of symmetrically related molecules in the crystal. The Zn 2ϩ -binding sites described here present the canonical constitution of other selective Zn 2ϩ -binding sites found in proteins (40). site in TTR in solution, since stable Zn 2ϩ coordination typically involves at least two amino acid interactions from the same molecule (40). Moreover, ZBS 4 is occupied in all four monomers exclusively at pH 4.6, being vacant in some of the monomers of the pH 5.5, 6.5, and 7.5 structures.
Analysis of multiple sequence alignments (MSA) (41,42) for all non-redundant, non-fragmented TTR sequences obtained from Protein Families data base (Pfam (29)) shows that the amino acids that we identify within each ZBS are among the least conserved regions of TTR (e.g. Cys in position 10 is present in only 12.9% of sequences, while His occupies position 56 in only 14.8%) (supplemental Table S2).
An analysis of the amino acids occupying two distinct positions in a given protein family, if these residues vary and are not completely conserved, tells us if for some reason, normally related to the structure or function of the protein, these two positions co-evolved (42). Most of the residues that form ZBS 1, ZBS 2, and ZBS 3 tend to appear simultaneously for a subset of the sequences (e.g. all sequences with a Cys in position 10 have His in positions 56 and 88, and most of them have an Asp in position 74) (supplemental Table S2). Interestingly, the TTR sequences that simultaneously exhibit these correlated positions are predominantly from mammals (supplemental Fig. S1).
Structural Details of Zn 2ϩ :TTR Complex-In addition to the local changes in the amino acids that coordinate Zn 2ϩ , the binding of Zn 2ϩ to the M-TTR structure causes striking structural modifications compared with the Zn 2ϩ -free state. The loops connecting ␤-strand E, the ␣-helix, and ␤-strand F, a region involved in the recognition of holo-RBP by TTR (2), are those that are most perturbed by Zn 2ϩ binding (Fig. 3, A and B). ZBS 2 and ZBS 3 flank this region.
Interestingly, in going from neutral to acidic pH, the Zn 2ϩ : M-TTR structures show a complete disruption of the ␣-helix into a disordered loop comprising residues 74 -90, but with a well defined electron-density map (supplemental Fig. S2). Even at pH 7.5, there is a substantial loss in the helix structure (Fig. 3, A and B) caused by Zn 2ϩ binding, but this structural perturbation is very pronounced in the structure solved at pH 4.6, where the ␣-helix disappears in all four subunits (Fig. 3A). In the crystallographic structure of WT-TTR at pH 3.5 in the absence of Zn 2ϩ , the electron density of the amino acids from this region is completely disordered and not visible in the electron density map (15). Here, we could observe precisely the disruption of this specific region of the protein in the presence of Zn 2ϩ upon acidification, a condition that triggers TTR aggregation. At pH 7.5, the side chain of Asp-74 turns toward His-31, forming ZBS 3 (Fig. 3B). The His-90 side chain turns toward the consolidated ZBS 2, accompanied by an almost 180°flipping of the His-88 side chain (Fig. 3B) in all four subunits and at all pH values. Various main-chain hydrogen bonds of the ␣-helix are disrupted (Tyr-78 3 Asp-74, Trp-79 3 Thr-75, Lys-80 3 Lys-76, Ala-81 3 Ser-77, arrows point from donors to acceptors), leading to loss of the ␣-helix stability and structure.
Affinity of WT-TTR for holo-RBP in the Absence and Presence of Zn 2ϩ -The crystal structure of the TTR-holo-RBP complex indicates that surface of loops 63-67 and 92-98 of RBP penetrate into a crevice formed by the arrangement of three monomers in the tetramer of TTR (2). Point mutations in the ␣-helix of TTR lead to a decrease in its affinity for holo-RBP (36). Thus, we envisioned that the structural changes induced by Zn 2ϩ binding in the region delimited by ZBS 2 and ZBS 3 could disrupt this topological complementarity between TTR and holo-RBP. Indeed, the extended segment of TTR resulting from the

. A close-up view of the loop E-loop F region flanked by ZBS 2 and ZBS 3 shows major structural modifications when Zn 2؉ is bound.
A, structures of the four subunits of TTR (ABCD; orange, red, blue, and yellow) obtained in the presence of Zn 2ϩ at different pH are overlapped with the Zn 2ϩ -free structure (green). Note the movement of the ␣-helix residues as the pH decreases from pH 7.5 to 4.6, causing the helix to unwind and form an extended loop. B, close-up view of the ZBS 2 and 3 region, in green, the apo form and in red the structure obtained in the presence of Zn 2ϩ at pH 7.5, where the structural alterations in the ␣-helix region are better seen. Residues in pink are those that undergo major reorientation upon Zn 2ϩ binding. Zn 2ϩ atoms are shown as gray spheres. C, Zn 2ϩ binding to WT-TTR decreases its affinity for holo-RBP. Isothermal holo-RBP:WT-TTR binding assays were performed by measuring anisotropy changes that take place upon TTR binding to holo-RBP in the absence (F) or in the presence of 15 M ZnCl 2 () and 50 M ZnCl 2 (f). Data are normalized to fraction of holo-RBP:TTR complex formed. ␣-helix unwinding upon Zn 2ϩ binding, which is very pronounced at acidic pH but considerable at pH 7.5, clashes with RBP (supplemental Fig. S3), a conformational change that might compromise the affinity of TTR for holo-RBP.
To address this issue, the affinities of WT-TTR for holo-RBP in the absence and presence of Zn 2ϩ were measured using fluorescence anisotropy (Fig. 3C). In the absence of Zn 2ϩ , the dissociation constant of holo-RBP:WT-TTR complex was 227 Ϯ 43 nM, in agreement with previous measurements (36,43). In the presence of Zn 2ϩ , the binding curves shift to higher WT-TTR concentrations (Fig. 3C and inset), indicating a decrease in affinity of WT-TTR for holo-RBP. The dissociation constants obtained in the presence of 15 M and 50 M Zn 2ϩ were 566 Ϯ 105 nM and 1,000 Ϯ 223 nM, respectively (Ϯ S.D.). Size-exclusion chromatography demonstrates that the oligomeric state of WT-TTR is not affected by the interaction with Zn 2ϩ , as indicated by the prevalence of tetramers at Zn 2ϩ concentrations as high as 100 M (Fig. 3D). This rules out the possibility that the decrease in the affinity of TTR for holo-RBP is due to TTR aggregation or changes in tetramer-monomer equilibrium.
Mapping the Zn 2ϩ -binding Sites by TROSY-HSQC NMR Measurements-To confirm that the Zn 2ϩ -binding sites and the structural changes involving the ␣-helix previously observed in the crystal structure of M-TTR are also present in solution and in the tetrameric WT-TTR, a stepwise titration of Zn 2ϩ was performed using TROSY-HSQC experiments ( 2 H 15 N-labeled WT-TTR, pH 7.5) (Fig. 4, supplemental Figs. S4 and S5). These experiments also allowed for the hierarchical identification of each ZBS in WT-TTR and the accompanying structural modifications evoked by Zn 2ϩ binding to each one. Significant chemical-shift perturbations (CSP) were observed in several regions of WT-TTR upon stepwise additions of Zn 2ϩ .
As seen in Fig. 4A, incubating 100 M WT-TTR (equivalent to 400 M monomeric subunits) with 100 M Zn 2ϩ leads to significant changes in Cys-10 and in the adjacent amino acid residues of ␤-strand A, suggesting that ZBS 1 is likely the first Zn 2ϩ -binding site to be filled. Unfortunately, the resonance of His56, the other coordinating residue of ZBS 1, was not localized in the spectrum (red bar) at any concentration of Zn 2ϩ added.
Increasing the concentration of Zn 2ϩ to 200 M (Fig. 4B) eliminates the resonance signal from Gly-57 (blue bar), adjacent to His-56. This might be related to Zn 2ϩ binding to His-56. At this concentration of Zn 2ϩ , a prominent CSP occurs in the resonance of Glu-92, a residue that contributes to ZBS 2 (supplemental Fig. S5). Some of the residues of ␤-strands G and H are also perturbed, reflecting the fact that ␤-strands A, G, and H are anchored to each other by a network of backbone hydrogen bonds (44). With a further increase to 300 M Zn 2ϩ (Fig.  4C), the saturation of ZBS 2 is nearly complete (see residues His-88 and Glu-92 in Fig. 4C and supplemental Fig. S5). His-90 is only slightly perturbed at higher Zn 2ϩ concentrations (Fig. 4,  D and E). These data suggest that ZBS 2 is the second Zn 2ϩbinding site to be occupied in solution.
At 400 M Zn 2ϩ (Fig. 4D), significant changes appear in ZBS 3 and in the residues in its vicinity, resulting in CSPs for Ile-73, Thr-75, Lys-76, and Trp-79, a segment that encompasses the end of ␤-strand E and the subsequent ␣-helix. The resonance signals from the other coordinating residues of ZBS 3, Asp-74 and Glu-72, vanish from the spectrum along with Val-71 (blue bars), indicating the induction of conformational change due to Zn 2ϩ binding. Altogether, these modifications suggest a rearrangement in the vicinity of ZBS 2 and ZBS 3, to allow a proper spatial orientation for Zn 2ϩ coordination. All these binding events lead to a reorganization of the secondary structure elements of TTR, which includes the partial unwinding of the ␣-helix at pH 7.5, resulting in the consolidation of the Zn 2ϩ : TTR complex, as observed in the crystallographic structure (Fig. 3). Interestingly, the resonance from Glu-62, a possible component of ZBS 3, does not undergo CSP even at the higher concentrations of Zn 2ϩ (Fig. 4E). In the crystallographic structures, as mentioned, ZBS 3 was formed by Glu72 AU at higher pH values or by Glu62 symm at pH 4.6 (Fig. 2, ZBS 3 and ZBS 3Ј); the latter resonance is perturbed and vanishes from the spectrum. Thus, we postulate that ZBS 3 in solution comprises residues His-31, Glu-72, Asp-74 and one water molecule.
At 900 M Zn 2ϩ (Fig. 4E), CSP propagate throughout the whole WT-TTR structure. Nevertheless, the resonance of Glu-61, which contributes to ZBS 4, does not undergo perturbation, reinforcing the idea that ZBS 4, which involves residues from a symmetry-related subunit, cannot be formed stably in solution. The reversibility of all structural changes induced by Zn 2ϩ binding was confirmed by the addition of 1 mM EDTA, a condition that was found to restore the original values of chemical shifts of the Zn 2ϩ -free protein (Fig. 4F).
It would be interesting to assess the dissociation constant (K d ) for Zn 2ϩ binding to ZBS 1-3. The apparent K d value already reported in the literature, based on fluorescence measurements, is ϳ1 M at pH 7.4 (19). This apparent K d is not related to a specific Zn 2ϩ -binding site but to all of them. Unfortunately, the high protein concentrations necessary to obtain adequate signals in the NMR experiments make the precise calculation of the dissociation constants very difficult, because most of them are in the nM-M range. Nevertheless, by plotting the CSP as a function of Zn 2ϩ addition for residues located in ZBS 1-3 (supplemental Fig. S5) we were able to extract at least a C 0.5 (Zn 2ϩ concentration that promotes 50% change).
Regarding Cys-10, the addition of 100 M Zn 2ϩ was enough to promote the same CSP for Cys-10 as that observed with 200 -900 M. This suggests saturation of ZBS 1, which includes Cys-10. This behavior is typical of a binding equilibrium in slow-exchange regime, as expected for high-affinity binding sites (K D Ͻ 10 Ϫ7 M). The slow-exchange regime is characterized as an on-off switch: the free and bound states of the protein are observed with their respective chemical shifts, but not the intermediate chemical shifts associated with partially bound species. Thus, we expect that ZBS 1 will have a K d on the order of a few nM.
In contrast to what was observed with ZBS 1, the resonance signals associated with ZBS 2 and ZBS 3 change stepwise with Zn 2ϩ addition (supplemental Fig. S5

Probing the Secondary Structural Changes Elicited by Zn 2ϩ
Binding through the Use of NMR-Because our crystallographic data suggest major perturbations in the loops connecting ␤-strand E, the ␣-helix, and ␤-strand F (Fig. 3, A and B), we triple-labeled WT-TTR ( 2 H, 13 C, 15 N) and acquired triple resonance experiments (HNCA and HNCO) to assign C␣ and carbonyl resonances. The chemical shifts of these resonances correlate well with secondary structural changes (45). The experiments were performed at pH 7.5. As seen in Fig. 5, addition of Zn 2ϩ did not promote any significant CSP in the signals of the residues located in the ␣-helix (residues 74 -82). These data unequivocally show that in solution at neutral pH, the ␣-helix persists when Zn 2ϩ binds. The changes observed in the crystal probably arise from a dynamic equilibrium among different configurations of this ␣-helix. Note that there are pronounced chemical-shift changes in the C␣, carbonyl (Fig. 5), and HN (Fig. 4) for the residues comprising the nearby residues of this ␣-helix, mainly the loop before strand F and the C-terminal of strand E. These changes reflect fluctuations in the position of the ␣-helix. We also measured line broadening in HN resonances of the residues located in the loops close to the ␣-helix (supplemental Fig. S4). The broadening indicates conformational exchange on a time scale of milli-to microseconds (46). Fluctuations of these loops could ultimately lead to fluctuation of the helix that is perceived in the crystallographic structures.
Indeed, analyzing the Ramachandran plots of these residues (supplemental Fig. S6), we see that in the crystal structures obtained in the absence of Zn 2ϩ (1GKO and 1F41), the torsion angles and are located in a restricted, low-energy region of the Ramachandran plot. In the structures obtained in the presence of Zn 2ϩ (3DGD, 3GPS, 3GRB, 3GRG), the angles adopted by these residues tend to be more dispersed within the allowed ␣-helix region, suggesting that in the presence of Zn 2ϩ , the ␣-helix shows conformational diversity, fluctuating between two positions.
Probing the Thermodynamic Stability of TTR in the Presence of Zn 2ϩ -As shown, Zn 2ϩ binding induces broad tertiary structural changes in TTR, with a particular secondary structure change consisting in the perturbation of the ␣-helix and in other regions of the protein. To assess the stability cost of such structural effects, we challenged M-TTR and WT-TTR with urea and high hydrostatic pressure (HHP) denaturation (38,47), using tryptophan fluorescence emission as a sensor of structural integrity.
Both urea and pressure isotherms show a typical denaturation curve for WT-TTR (Fig. 6, A and C, respectively). However, Zn 2ϩ concentrations up to 600 M promote only minor changes in the denaturation profile, indicating marginal effects of Zn 2ϩ on stability of the WT-TTR tetramer. In contrast, with M-TTR there is a progressive change in the unfolding transition as a function of Zn 2ϩ concentration, for both urea (Fig. 6B) and pressure (Fig. 6D).
A quantitative thermodynamic analysis of M-TTR denaturation reveals a steep decrease in ⌬G unf up to 100 M Zn 2ϩ (Fig. 6E). From the urea denaturation experiments, the ⌬G unf in the absence and in the presence of 100 M Zn 2ϩ decrease from 2.54 to 1.66 kcal/mol (⌬⌬G unf ϭ 0.88 kcal/mol) (filled circles in Fig. 6E). Additionally, the ⌬G unf in the absence and in the presence of 100 M Zn 2ϩ calculated from the HHP experiments equal 2.97 and 2.41 kcal/mol, respectively, again showing a decrease in stability of M-TTR by ϳ0.6 kcal/mol (squares in Fig. 6E), a change very similar to that obtained from the urea-induced unfolding. Fig. 6F shows the changes in the m unf (circles) and ⌬V (squares) parameters extracted respectively from the urea and HHP denaturation curves (Equations 4 and 5). Interestingly, in the presence of Zn 2ϩ the volume change of folding (⌬⌬V) decreases by ϳ 20 ml/mol (from 71 to 52 ml/mol), while the m unf parameter decreases from 0.86 to 0.41 in the presence of 100 M Zn 2ϩ . These data corroborate the idea that M-TTR in the presence of Zn 2ϩ loses part of its tertiary structure, resulting in a lower cavity content and a decreased number of intra-chain contacts, changes that render the protein less stable.

DISCUSSION
The first human TTR crystallographic structure (PDB ID 2PAB (48)) revealed metal coordination involving Cys-10 and His-56, residues from ZBS 1, because the thiol group of Cys-10 underwent mercurization due to treatment with heavy metal for isomorphous derivatization. The crystals were then washed with chelating agent, thus the other ZBS were not detected (48). Occupation of ZBS 1 alone does not induce any global conformational changes, nor any modification in the ␣-helix or the EF loops, as presented here (Figs. 1 and 3). ZBS 1, the first site to be filled in low Zn 2ϩ concentrations, might be permanently occupied in circulating TTR. This occupancy could prevent TTR aggregation by committing Cys-10 to the task of Zn 2ϩ coordination and thereby avoiding SH-modifications (such as TTR-Cys, TTR-GSH, and TTR-CysGly), which have been implicated in TTR destabilization and amyloidogenesis in vitro (49).
In solution, Zn 2ϩ binding to ZBS 2 and ZBS 3 results in a pronounced structural rearrangement in the EF loops and the ␣-helix (Figs. 3 and 5), a region that becomes more dynamic and flexible. Interestingly, a recently solved structure of TTR at acidic pH in the absence of metal ions, a condition that triggers TTR aggregation due to the formation of a misfolded amyloidogenic intermediate, displayed ␣-helix unwinding (15). Thus, we suggest that Zn 2ϩ binding to TTR at neutral pH into ZBS 2 and 3 induces structural modifications that resemble those observed under acidic conditions, where TTR forms amyloid fibrils. Zn 2ϩ binding might trigger TTR aggregation by perturbing the ␣-helix region and residues nearby. This structural perturbation is intensified at acidic pH, a condition in which the ␣-helix completely disappears, forming a long loop and exposing a segment of the protein rich in hydrophobic and aromatic residues (78YWKALGISF87, Tyr-78, Trp-79, and Phe-87), which may form hydrophobic andstacking interactions with adjacent protein molecules, as well as hydrogen bonds, leading to aggregation. ZBS 4, which involves amino acids from adjacent subunits, might recruit more Zn 2ϩ , which could then bind and stabilize the quaternary structure of the amyloid fibril.
Interestingly, 2 H-1 H exchange experiments performed with WT-TTR at pH 4.5 (44) and with FAP variants at pH 7 (31) revealed that the CBEF sheet exhibits faster exchange rates than the ADGH sheet, which is the stable core of TTR. Thus, Zn 2ϩ binding, mainly in ZBS 2 and 3, maps precisely to the labile region of the protein: the region around strands E and F.
These structural rearrangements involve the holo-RBP binding region in TTR, leading to a decrease in holo-RBP:WT-TTR affinity (36) (Fig. 3C and supplemental Fig. S3). Most holo-RBP circulates in serum bound to TTR (2), resulting in an interaction with mutual benefits such as the avoidance of RBP clearance, the stabilization of the retinol:RBP complex (2), and prevention of TTR dissociation and amyloid formation (43). Indeed, a 2-fold decrease in affinity observed in the presence of 15 M Zn 2ϩ , the plasmatic concentration of Zn 2ϩ , is rather small. However, this small decrease is to be expected, since at this Zn 2ϩ concentration, the structural perturbations caused by Zn 2ϩ binding to TTR are not all complete and thus its affinity for holo-RBP would be only marginally affected. However, in the presence of 50 M Zn 2ϩ , a concentration that is likely physiologically relevant at certain places, the affinity decreases 4 -5fold. In addition, one has to take into account the fact that TTR in the plasma is bound to other ligands, such as the hormone T4, and these other partners may modulate its affinity for holo-RBP when bound to Zn 2ϩ .
As shown here (Fig. 6, B, D, and E), Zn 2ϩ binding decreases monomer stability, and this effect is probably related to all structural changes induced by Zn 2ϩ binding. However, Zn 2ϩ binding does not alter WT tetramer stability to a great extent (Fig. 6, A and C). This is expected since the tetrameric contacts are not severely affected by the formation of ZBS 1-3.
Aggregation of TTR presupposes the formation of a monomeric amyloidogenic intermediate, which is aggregation-prone (8). The existence of an altered tetramer with amyloidogenic properties has also been postulated (13,50). It may be that a tetramer with ZBS 2 and 3 filled with Zn 2ϩ serves as this tetrameric amyloidogenic intermediate, capable of giving rise to amyloid fibrils either directly or by populating a monomeric amyloidogenic species.
In FAP patients, the amyloid deposits of TTR variants are found mainly around basal membranes of Schwann cells, inside the endoneurium of peripheral nerves (5,6). The myelin sheath accumulates a high concentration of Zn 2ϩ (about 50 M) (51). In addition, mutations predispose tetramer to dissociation and Zn 2ϩ might trigger amyloidosis by occupying its binding sites in TTR, provoking the structural rearrangement described here. Thus, the high concentration of Zn 2ϩ in the region in conjunction with the enhanced tetramer instability caused by the mutations, might explain the tropism of TTR amyloids to FIGURE 6. Zn 2؉ binding decreases monomer stability but is innocuous to the tetramers. WT-TTR (3.5 M tetramers; panels A and C) or M-TTR (3.5 M monomers; panels B and D) were subjected to denaturation by urea (panels A and B) or by high hydrostatic pressure (panels C and D) in Tris-HCl 25 mM, KCl 50 mM, pH 7.5 (25°C) at varying ZnCl 2 concentration (0 -600 M, as stated in the figure inset). The center of spectral mass of tryptophan emission of each spectrum was calculated and converted into the extent of reaction (␣) as described under "Experimental Procedures" and plotted against the variables. From these plots and by using Equations 4 and 5, the ⌬G unf (panel E) or the ⌬V or m unf parameters (panel F) for the unfolding of M-TTR were calculated. Circles represent urea denaturation and squares show HHP. Error bars in panels E and F are standard deviations obtained from non-linear regression analysis (most of them are smaller than the symbols). Details under "Experimental Procedures." this specific site. The same scenario could be envisioned in the vitreous humor of the eye, where the concentration of Zn 2ϩ is also high (2.4 Ϯ 0.95 mg/liter; 36 Ϯ 14 M) (52), and is associated with Zn 2ϩ -rich TTR amyloid deposits (22) that cause vitreous opacity (53). Fig. 7 summarizes the main findings of the present study. The concentration of Zn 2ϩ in the plasma is sufficient to fill ZBS 1, which does not impede TTR-holo-RBP interaction and the efficient transport of retinol. However, when TTR encounters a place with an increased Zn 2ϩ concentration, such as the endoneurium, ZBS 2 and 3 are occupied by Zn 2ϩ , leading to a structural reorganization in the region around the ␣-helix, which fluctuates-displaying structural diversity (Fig. 5). This rearrangement interferes with holo-RBP binding or favors proteinprotein interactions, which are crucial for TTR aggregation.
Evidence for the importance of metal ions in amyloidogenic diseases such as Alzheimer (54) and Parkinson disease (55) is growing (56 -58); it may be that metal binding is a common feature of amyloids. Thus, a better characterization of the role and control of metal homeostasis in such pathologies could lead to a general therapeutic approach to ameliorate amyloid diseases (59).