TTR (transthyretin) was found recently to possess proteolytic competency besides its well-known transport capabilities. It was described as a cryptic serine peptidase cleaving multiple natural substrates (including β-amyloid and apolipoprotein A-I) involved in diseases such as Alzheimer's disease and atherosclerosis. In the present study, we aimed to elucidate the catalytic machinery of TTR. All attempts to identify a catalytic serine residue were unsuccessful. However, metal chelators abolished TTR activity. Proteolytic inhibition by EDTA or 1,10-phenanthroline could be reversed with Zn2+ and Mn2+. These observations, supported by analysis of three-dimensional structures of TTR complexed with Zn2+, led to the hypothesis that TTR is a metallopeptidase. Site-directed mutagenesis of selected amino acids unambiguously confirmed this hypothesis. The TTR active site is inducible and constituted via a protein rearrangement resulting in ~7% of proteolytically active TTR at pH 7.4. The side chain of His88 is shifted near His90 and Glu92 establishing a Zn2+-chelating pattern HXHXE not found previously in any metallopeptidase and only conserved in TTR of humans and some other primates. Point mutations of these three residues yielded proteins devoid of proteolytic activity. Glu72 was identified as the general base involved in activation of the catalytic water. Our results unveil TTR as a metallopeptidase and define its catalytic machinery.
TTR (transthyretin) is a plasma protein existing mainly as a homotetramer . TTR is well characterized, since point mutations can cause FAP (familial amyloid polyneuropathy) , a lethal neurodegenerative disorder hallmarked by deposition of TTR amyloid fibrils . A myriad of crystal structures of human wtTTR (wild-type TTR) are available from the PDB . Backbone hydrogen bonds between TTR monomers favour dimer formation, whereas dimers interact predominantly via amino acid side chains to form a tetramer. TTR stability depends on intact quaternary and tertiary structure, with partial unfolding initiating misassembled aggregates . The presence of metal ions , covalent modification of Cys10 , lowered pH  or point mutations  can destabilize the tetrameric TTR structure and favour protein rearrangement.
TTR is the transporter of T4 (thyroxine) and retinol owing to its association with RBP (retinol-binding protein)  in a 1:1 molar ratio in vivo , an interaction that stabilizes TTR more than 15-fold . Besides interacting with T4 and RBP, approximately 1–2% of total plasma TTR is associated with HDL (high-density lipoproteins), through binding to apoA-I (apolipoprotein A-I) . In-depth analysis of the interaction of TTR with apoA-I revealed the latter to be a peptidase substrate of TTR . TTR is currently described as a peptidase of unknown type (U9G.071) in the MEROPS database . Initial studies suggested TTR to be a serine peptidase  given that it cleaves apoA-I after a phenylalanine residue at an optimal pH of 7 and that it is apparently inhibited by very high concentrations of serine peptidase inhibitors . Although binding of small compounds such as T4 has no effect on the peptidase activity of TTR, interaction with RBP completely abolishes substrate cleavage . To date, a significant number of natural substrates of TTR have been identified such as apoA-I, NPY (neuropeptide Y) and Aβ (amyloid β-peptide) [14,16,17].
ApoA-I is the main protein component of HDL and is crucial for reverse cholesterol transport . Limited proteolysis has demonstrated the vulnerability of the C-terminus of apoA-I in lipid-poor particles, whereas the lipid-bound protein is protected from cleavage . The apoA-I N-terminus appears to be important for stabilizing the lipid-free monomeric structure, whereas the C-terminus is required for interactions with other proteins and lipids . After apoA-I cleavage by TTR, HDL particles display a reduced capacity to promote cholesterol efflux, and truncated apoA-I displays increased amyloidogenicity , suggesting that TTR might have an impact on the development of atherosclerosis. In the case of Aβ, cleavage by TTR has only been demonstrated in vitro , and suggested to be a protective mechanism preventing Alzheimer's disease .
Despite the availability of a large number of wtTTR and mutated TTR X-ray structures crystallized at pH 7–8, the catalytic mechanism of TTR is still unknown. In the present study, we establish TTR as a metallopeptidase and uncover multiple catalytic residues indispensable for proteolytic activity.
Recombinant wtTTR and TTR mutants were produced in BL-21 pLys Escherichia coli cells transformed with pETF1 carrying TTR cDNA . Proteins were isolated and purified as described in . Briefly, after bacterial lysis, protein extracts were run on DEAE-cellulose (Whatman) ion-exchange chromatography, dialysed, freeze-dried and isolated in native Prosieve® agarose (FMC) gels. After electrophoresis, the TTR band was excised and electroeluted in a Elutrap® system (Schleicher and Schuell) in 38 mM glycine and 5 mM Tris/HCl (pH 8.3) overnight at 50 V (4°C). Finally, the protein was dialysed in 50 mM Tris/HCl (pH 7.5). Protein quantification was performed using Lowry-based DC Protein Assay (Bio-Rad Laboratories), following the manufacturer's protocol. The TTR molar concentration was always calculated assuming the exclusive presence of the tetrameric species.
The TTR cDNA cloned into the pETF1 vector was used to produce TTR mutants (using the QuikChange® kit; Stratagene). Mutants were constructed using two mismatched primers, introducing a base substitution in the original sequence. Minipreps of plasmid DNA were tested by sequencing at SeqLab (Göttingen, Germany). After protein production, mutations were confirmed by MALDI (matrix-assisted laser-desorption ionization)-MS of either the full-length protein or of the trypsin-digested TTR peptides (Supplementary Table S1 at http://www.BiochemJ.org/bj/443/bj4430769add.htm), at the Proteomics Unit, Institute of Molecular Pathology and Immunology at the University of Porto, Porto, Portugal.
TTR proteolytic activity was tested with the fluorigenic peptide Abz-YGGRASDQ-EDDnp [where Abz is o-aminobenzoyl and EDDnp is N-(2,4-dinitrophenyl)-ethylenediamine]. This substrate was found when screening a library of fluorigenic peptides . Turnover of Abz-YGGRASDQ-EDDnp by TTR is approximately 20-fold higher (kcat/Km=408.8 s−1·M−1) when compared with Abz-ESFKVS-EDDnp (kcat/Km=17.5 s−1·M−1), an apoA-I sequence encompassing the TTR cleavage site. Hydrolysis of Abz-YGGRASDQ-EDDnp at 37°C in 50 mM Tris/HCl (pH 7.5) was followed by measuring the fluorescence at λem 420 nm and λex 320 nm in an Fmax plate reader (Molecular Devices) for 30 min. Specificity rate constants (kcat/Km) were determined under pseudo-first-order conditions . TTR (5 μM total protein) was added to the substrate (tested at 5 μM and 10 μM) in a final volume of 100 ml of reaction buffer (50 mM Tris/HCl, pH 7.5). Reactions were monitored for 30 min and pseudo-first-order rate constants were obtained from linear plots where the y-axis corresponds to ln[(Fmax−Ftime)/Fmax], where Fmax is the fluorescence corresponding to total degradation of 5 μM or 10 μM substrate and Ftime is the fluorescence measured at each time point, and the x-axis corresponds to the time of reaction. The slope of the linear plots, corresponding to the first-order rate constants, was divided by the total protein concentration to provide kcat/Km. Determination of the concentration of active enzyme in the TTR preparation was performed by titration with an irreversible phosphonate inhibitor that completely blocks TTR activity at 1 μM . Active-site titration was carried out with inhibitor concentrations ranging from 0.01 to 1 μM. After enzyme/inhibitor incubation for 30 min at 37°C, activity was monitored as described above. Concentration of active enzyme was determined from the x-axis intercept in the linear range of the plot of residual activity as a function of inhibitor concentration .
Chemical modification of TTR with serine peptidase inhibitors
TTR (5 μM) was incubated with each of the following peptidase inhibitors: Pefabloc SC (4 mM; Roche), DFP (di-isopropylfluorophosphate) (100 μM; Calbiochem), TPCK (tosylphenylalanylchloromethane) (100 μM; Roche) and PMSF (1 mM; Sigma). After 30 min of incubation at 37°C, reaction mixtures were separated by SDS/PAGE (15% gel) and stained with Coomassie Blue. MALDI-MS of trypsin-digested gel bands corresponding to modified TTR was performed at the Protein Core Facility, Columbia University, New York, NY, U.S.A., as described in . For chemical modification of apoA-I (1.7 μM) with PMSF, the protein was incubated with 1 mM inhibitor for 30 min at 37°C. After SDS/PAGE, MALDI-MS was performed at the Proteomics Unit, Institute of Molecular Pathology and Immunology at the University of Porto, Porto, Portugal.
Inhibition of TTR cleavage of fluorigenic peptides
For analysis of inhibition of TTR proteolytic activity using Abz-YGGRASDQ-EDDnp as substrate, TTR (5 μM) was pre-incubated for 30 min at 37°C in reaction buffer with the following inhibitors: 1–10 mM EDTA (Merck), 10 μM E-64 [trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane] (Roche), 10 μM phosphoramidon (Roche), 10 μM bestatin (Sigma), 0.3 μM aprotinin (Sigma), 100 μM TPCK (Roche), 100 μM TLCK (tosyl-lysylchloromethane) (Roche), 100 μM chymostatin (Sigma), 100 μM leupeptin (Sigma), 1 mM PMSF (Sigma), 1 μM pepstatin (Sigma), 1 mM 1,10-phenanthroline (Roche) and 1 mM 1,7-phenanthroline (Alfa Aesar). After enzyme/inhibitor incubation, 5 μM Abz-YGGRASDQ-EDDnp was added and activity was monitored as described above. Relative fluorescence was converted into percentages of residual activity relative to uninhibited controls. To analyse the effect of addition of divalent metals to TTR inhibited with either 1,10-phenanthroline or EDTA, TTR was pre-incubated with the inhibitor (as described above) and then increasing concentrations of metals were added to the reaction and proteolytic activity was determined. In the experiments with CaCl2 addition, reactivation of TTR with Zn2+ was tested in the presence of 1 mM CaCl2. The apo-enzyme form of TTR was obtained by dialysis of TTR pre-incubated with 10 mM EDTA. Control measurements of the reactions where no TTR was added were performed, and the results obtained were subtracted from the results from reactions containing the peptidase.
TTR proteolysis assay using full-length apoA-I as substrate
ApoA-I digestion by TTR was performed as described previously . Briefly, 2 mg of TTR was incubated with 1 mg of apoA-I in 50 mM Tris/HCl (pH 7.5) at 37°C overnight. The inhibitory effect of EDTA was tested by the addition of 10 mM or 50 mM EDTA to TTR, before addition of apoA-I. After incubation for 30 min at 37°C, apoA-I was added and reactions proceeded at 37°C overnight. Reaction mixtures were subsequently separated by SDS/PAGE (15% gel) and visualized by silver staining.
Determination of Zn2+ levels in TTR
The Zn2+ concentration in TTR samples was measured at the Chemical Speciation and Bioavailability Laboratory, Center of Marine and Environmental Research, University of Porto, Porto, Portugal, as described previously . Briefly, 3.2 mg of TTR was diluted in Zn2+-free double-distilled water and total Zn2+ content was determined by atomic absorption spectrophotometry with flame atomization (PU 9200X, Philips). For Zn2+ quantification, a calibration curve was created using Zn2+ standards. For Zn2+ quantification in TTR treated with EDTA, the protein was incubated with 10 mM EDTA for 30 min at 37°C and thereafter the chelator was removed by dialysis against 50 mM Tris/HCl (pH 7.5). For measurement of Mn2+, Fe2+ and Co2+ concentrations, the same protocol was followed.
For the thioflavin-T-binding assay, 100 μg of purified TTR was incubated in 50 mM sodium acetate (pH 4) for 72 h at room temperature (22°C) in a final volume of 200 μl. For aggregation assays with 10 mM EDTA, 1 mM 1,10-phenanthroline or Zn2+, TTR was incubated with one of these compounds for 30 min at 37°C and subsequently thioflavin-T was added at a concentration of 30 μM in 50 mM glycine (pH 9). The excitation spectra from 400 to 500 nm were recorded on a Horiba Fluoromax-4 spectrofluorimeter at 19°C. The intensity of fluorescence at 451 nm, which is the characteristic maximum for thioflavin-T bound to aggregated fibrils, was subtracted from the intensity of fluorescence at 451 nm of the control without TTR. Results are presented as a percentage relative to the value obtained for wtTTR. Each assay was performed in triplicate.
Size-exclusion chromatography experiments were performed with a pre-packed Superose 12 10/300 GL column (GE Healthcare). The column was equilibrated with 150 mM NaCl and 50 mM Tris/HCl (pH 7.5) and a 0.5 ml/min flow rate was used throughout the experiments. A total of 150 μg of purified TTR was used in each run. Protein elution was monitored by measuring the absorbance at 280 nm. Calibration was carried out using the following protein standards (Stokes radius, elution volume): catalase (5.22 nm, 11.03 ml), aldolase (4.81 nm, 11.26 ml), albumin (3.55 nm, 11.76 ml), ovalbumin (3.05 nm, 12.66 ml), chymotrypsinogen (2.09 nm, 14.66 ml) and ribonuclease (1.64 nm, 15.14 ml). The void volume was determined to be 7.21 ml using Blue Dextran 2000. The Kav parameter was determined according to the equation Kav=(Ve−V0)/(Vt−V0), where Ve represents the elution volume, V0 is the void volume of the column, and Vt is the total bed volume. The Stokes radius (Rs) for the experimental data was calculated by interpolation using: (−logKav)1/2=f(Rs).
All assays were performed at least twice with each dataset in duplicate. Results are means±S.D.
TTR is not a serine peptidase
TTR has been suggested to be a cryptic serine peptidase . Serine peptidases are hallmarked by an activated nucleophilic serine , an oxyanion hole and a substrate-binding cleft. After an in-depth analysis of multiple X-ray structures of monomeric, dimeric and tetrameric TTR, Ser46, Ser77 and Ser85 were selected as putative candidates that could act as activated nucleophiles. We mutated each of these three serine residues to glycine or alanine and confirmed the respective mutations by tryptic MS (results not shown). All three mutants retained proteolytic activity (results not shown), suggesting that TTR is not a serine-like peptidase. Ser8, Ser23, Ser50, Ser52, Ser100 and Ser117 of TTR, although they were very unlikely to be active-site candidates, were also mutated. All of these mutants retained proteolytic activity (results not shown). The only TTR serine residues that were not mutated were Ser112 and Ser115 as these are located in the hydrophobic central channel of TTR, described previously to be irrelevant for TTR activity .
As TTR was suggested to be a serine-like peptidase on the basis of the inhibition of apoA-I cleavage by high concentrations of serine peptidase inhibitors , the effect of these compounds was re-examined in detail using full-length apoA-I as substrate. MS analysis of TTR incubated with the covalent binding serine peptidase inhibitor Pefabloc SC revealed the presence of multiple modified TTR peptides (Supplementary Table S2 at http://www.BiochemJ.org/bj/443/bj4430769add.htm), indicating that highly concentrated Pefabloc SC modifies TTR unspecifically, therefore affecting peptidase activity in a non-specific manner. For TTR incubated with PMSF, DFP and TPCK, no modification of TTR was found (results not shown). However, PMSF was able to modify the substrate (apoA-I) and generate an aberrant PMSF adduct of a peptide consisting of amino acids 227–238 of apoA-I (1540.86 Da) (Supplementary Table S3 at http://www.BiochemJ.org/bj/443/bj4430769add.htm). PMSF modification of the substrate is most likely to be the reason for reduced apoA-I cleavage by TTR. Besides inhibition of TTR activity by serine peptidase inhibitors, we previously observed inhibition of TTR activity with peptide phosphonate inhibitors when using the fluorigenic apoA-I peptide substrate . Phosphonate inhibitors are able to block serine peptidases, but some members of this class of compound are known to be effective inhibitors of metallopeptidases . Combined, the absence of catalytic serine residues in TTR and unspecific modifications of TTR and apoA-I by peptidase inhibitors reveal that TTR is not a serine peptidase.
TTR proteolytic activity depends on metal ions and is inhibited by metal chelators
A highly reproducible assay using fluorigenic peptide substrates was set up to investigate the action of multiple prototypic peptidase inhibitors on TTR activity (Figure 1a). As we previously observed interference with proteolytic activity by post-translational modifications of Cys10 in TTR , all assays were performed with a C10A mutant which was equipotent to wtTTR (Table 1). Only two well known metallopeptidase inhibitors, EDTA and 1,10-phenanthroline, could abolish TTR proteolytic activity (Figure 1a). Interestingly, 1,7-phenanthroline, an isomeric non-chelating analogue of 1,10-phenanthroline, achieved only minor inhibition (Figure 1a). These results were corroborated with a second substrate mimicking the TTR cleavage site in apoA-I  (Figure 1b). Moreover, when using full-length apoA-I as substrate, when increasing amounts of EDTA are used, full inhibition is also achieved (Supplementary Figure S1 at http://www.BiochemJ.org/bj/443/bj4430769add.htm). The same inhibition profile was observed when wtTTR was used, proving that inhibition by EDTA and 1,10-phenanthroline is not dependent on the point mutation C10A (Supplementary Figure S2 at http://www.BiochemJ.org/bj/443/bj4430769add.htm).
TTR cleavage of fluorigenic peptides is inhibited by EDTA and 1,10-phenanthroline
EDTA inhibited TTR in a dose-dependent manner (Figure 2a). Excess amounts of EDTA had to be scavenged with 1 mM Ca2+ ions before addition of Zn2+ could surmount inhibition and restore full TTR proteolytic activity (Figure 2b). Alternatively, excessive EDTA could be removed via dialysis, yielding an inactive apo-enzyme that can be reactivated by addition of Zn2+ (Figure 2c). This indicates that the Zn2+-scavenging effect of EDTA leads to proteolytic inactivation of TTR. Similar results were obtained with wtTTR (Supplementary Figure S3 at http://www.BiochemJ.org/bj/443/bj4430769add.htm).
TTR inhibition by EDTA and 1,10-phenanthroline is reversed by addition of ZnCl2
Complete inhibition of TTR with 1 mM 1,10-phenanthroline could be partially reversed by adding Zn2+ (Figure 2d). Zn2+ at 250 μM restored 36% of proteolytic activity in the presence of inhibitor. This finding is similar to the results obtained with EDTA where the addition of 250 μM Zn2+ restored 43% activity (Figure 2b). Zn2+ concentrations exceeding an optimal value (250 μM for TTR) inhibited TTR (results not shown), as described previously for other metallopeptidases . Nevertheless, after inhibition of TTR with 1,10-phenanthroline, dialysis yielded an active peptidase instead of an inactive apo-enzyme (results not shown). This effect was described previously for carboxypeptidase A where inhibition of the enzyme with 1,10-phenanthroline was reversed by Zn2+ addition, but dialysis of the enzyme treated with the metal chelator restored the active enzyme rather than producing the apo-enzyme. In that case, it was suggested that dialysis leads to dissociation of the Zn2+–1,10-phenanthroline complex permitting the re-incorporation of the Zn2+ ion in the enzyme . Addition of 1 mM CaCl2 did not sequester 1,10-phenanthroline as observed previously for EDTA, since binding of 1,10-phenanthroline to Ca2+ is significantly less strong  (Figure 2e). In summary, these results indicate that Zn2+ plays a crucial role in the proteolytic activity of TTR and thereby suggest that TTR is a Zn2+-dependent metallopeptidase.
TTR is reactivated by other divalent metal ions
The active site necessary for proteolytic activity has never been observed in any X-ray structure of wtTTR obtained at physiologically relevant pH (pH 7–8). This triggered the question of whether this might be due to an incorrect metal being present in crystallization assays. Zn2+ is typically found in the active site of metallopeptidases, but Co2+, Cu2+, Ni2+, Mn2+ or Fe2+ can also produce active metallopeptidases as described for metalloenzymes such as thermolysin  or peptide deformylase . Interestingly, TTR complexed to Mn2+ was even slightly more active than its Zn2+ analogue (Figure 3a), indicating that TTR can be active with metals other than Zn2+in vitro. Further metals tested (Fe2+ and Co2+) also yielded active enzymes (Figures 3b and 3c), whereas in the presence of Ni2+ or Cu2+, TTR displayed very little proteolytic activity (Supplementary Figures S4a and S4b respectively at http://www.BiochemJ.org/bj/443/bj4430769add.htm).
TTR inhibited by 1,10-phenanthroline is reactivated by divalent metal ions
Only a fraction of TTR contains Zn2+ and is proteolytically competent
Metallopeptidases require a metal ion to correctly locate and activate a water molecule . The metal ion is often Zn2+ co-ordinated by three residues of the enzyme and a water molecule. Such a structural feature could not be found in any wtTTR X-ray structure in the PDB crystallized in the pH range 7–8. Recently, several structures of TTR double mutants (F87M/L110M) crystallized in the pH range 4.5–7.5 comprising multiple Zn2+-binding sites were described . In one of these structures, a large reorganization of amino acids 74–90 occurs, shifting the side chain of His88 by almost 9 Å (1 Å=0.1 nm) such that this residue together with His90 and Glu92 co-ordinates a Zn2+ (e.g. PDB code 3GRG) . These three relocated amino acids, comprising two neutral (His88 and His90) and one negatively charged amino acid (Glu92), are reminiscent of active sites in metallopeptidases.
To determine whether recombinant TTR contains Zn2+, the concentration of this ion in wtTTR preparations was measured using atomic absorption spectrophotometry with flame atomization. In 5 μM wtTTR, a Zn2+ concentration of approximately 1 μM could be measured. Moreover, when TTR was pre-treated with EDTA before atomic absorption spectrophotometry, no Zn2+ was detected. As described above, metals other than Zn2+ were shown to interfere with TTR proteolytic activity, as is the case for Mn2+, Fe2+ and Co2+; however, these metals were not detected in the wtTTR preparations (results not shown).
We analysed the effect of adding Zn2+ directly to recombinant TTR. Increasing Zn2+ concentrations did not increase proteolytic activity further, but inhibited it (Figure 4a). Such an effect can be explained by Zn2+ binding near the active site, as observed previously in X-ray structures of other metallopeptidases . Such a second Zn2+-binding site is observed in one of the analysed TTR X-ray structures . One of its residues, Asp74, was mutated, but this mutation did not abolish the inhibitory effect of excessive Zn2+ (Figure 4a). The presence of metal ions can destabilize the TTR tetrameric structure . To check whether the inhibitory effect of increasing Zn2+ concentrations on TTR proteolytic activity was related to an increased TTR aggregation induced by the metal, we performed thioflavin-T-binding assays of TTR in the presence of 100 and 250 μM Zn2+. We observed that Zn2+ addition did not increase TTR aggregation (Figure 4b).
Only a small fraction of TTR comprises Zn2+ and is proteolytically competent
The molar ratio of TTR and Zn2+ determined in our preparations (~5:1 TTR/Zn2+) suggested that not all recombinant TTR is in a proteolytically active state. We therefore determined the percentage of proteolytically active wtTTR by active-site titration of a 5 μM TTR solution at pH 7.5 with a strong phosphonate inhibitor identified previously  (Supplementary Figure S5 at http://www.BiochemJ.org/bj/443/bj4430769add.htm), which labels the enzyme active site by co-ordinating the catalytic Zn2+ . Surprisingly, only a minor fraction of 340 nM wtTTR (~7% of total wtTTR) was in a proteolytically active state (Figure 4c). This percentage was lower than what was expected given the TTR/Zn2+ molar ratio. This result indicates that only a minor amount of wtTTR has proteolytic competence at physiological pH.
TTR is a metallopeptidase with an inducible Zn2+-binding site
On the basis of the above results, we hypothesized that TTR is a metallopeptidase having His88, His90 and Glu92 as the three relevant residues binding the catalytic Zn2+. To confirm this hypothesis, mutants of each of the above residues to alanine were generated, and their kinetic parameters were determined (Table 1). Mutating any of these three residues led to inactive TTR (Table 1), demonstrating that His88, His90 and Glu92 are the active-site residues.
In metallopeptidases, a reactive hydroxyl ion is generated by transferring a proton from the catalytic water to a neighbouring residue which acts as a general base. From structural analysis (Figure 5), Glu72 was favoured to be the general base and Glu89 was a less likely candidate. Supporting this hypothesis, mutating Glu72 abrogated the proteolytic activity of TTR, whereas mutating Glu89 resulted in a partial reduction in TTR activity. This latter residue receives two hydrogen bonds from (i) its own backbone amide, and (ii) the side chain of Thr96′ of a neighbouring second monomer which itself receives a hydrogen bond from His88 of the first monomer (Figure 5). As such, Glu89 fulfils a triple task in (i) stabilizing the new conformation of rearranged amino acids 74–90, (ii) strengthening the interaction between two monomers forming a dimer, and (iii) freezing the required conformation of Thr96′ which keeps His88 in the orientation required to bind Zn2+ (Figure 5).
Structure of TTR active site (PDB code 3GRB)
Lys70 is a potential hydrogen-bond partner of the general base Glu72 and could, as such, influence the pKa and conformational flexibility of this residue. Modelling studies indicated that Lys70 is also in a position to contact a backbone carbonyl oxygen of the substrate as described previously for Arg203 in thermolysin and Arg717 in neprilysin . His31 is a second neighbour of Glu72 and also a putative hydrogen-bond partner. To determine the influence of these two vicinal amino acids of Glu72 on catalytic activity, we mutated Lys70 to alanine and His31 to glycine. The results for both mutants confirmed their importance for catalytic activity with L70A and H31G having only 1.1% and 1.3% residual activity (Table 1).
A second Zn2+-binding site in double mutants of TTR comprising Glu72, His31 and Asp74  is present in close vicinity to the TTR catalytic site. Its relevance for proteolysis in general and the role of Asp74 on proteolytic activity in particular were not obvious which prompted us to generate D74A. Surprisingly, this mutant suffered only a 2-fold decrease in peptidase activity (Table 1), thereby indicating that this second Zn2+-binding site is not required for proteolytic activity.
In summary, TTR is a mononuclear metallopeptidase with an inducible active site with His88, His90 and Glu92 serving as Zn2+-complexing ligands.
TTR proteolytic activity is unrelated to protein aggregation and oligomeric state
Several factors, such as the presence of metal ions , covalent modification of Cys10 , lowered pH  or point mutations , can destabilize the tetrameric TTR structure. To determinate whether TTR aggregation propensity is increased in the presence of 1,10-phenanthroline and EDTA or by the TTR mutants of the active-site residues His88, His90, Glu92 and Glu72, we performed thioflavin-T-binding assays after protein acidification. In the case of 1,10-phenanthroline and EDTA, no major effects on wtTTR aggregation potential were observed (Figure 6a). Regarding the TTR mutants, no increased aggregation was detected after the thioflavin-T-binding assay in any of the mutants tested, if compared with wtTTR (Figure 6b). In accordance with the similar catalytic efficiency and inhibition profile, TTR C10A presented an aggregation potential similar to that of wtTTR (Figure 6b). Interestingly, TTR C10A/H88A, TTR C10A/H90A and TTR C10A/E92A had a decreased aggregation profile if compared with either wtTTR or TTR C10A. For TTR C10A/E72A, no differences in aggregation were detected (Figure 6b). Previous studies with wtTTR in an acidic environment revealed the importance of the EF helix–loop region of residues 75–90 in conformational changes leading to disassembly and subsequent aggregation of TTR monomers . Interestingly, mutation of these residues may lead to a decrease in the protein aggregation potential under acidic conditions. These findings demonstrate that TTR catalytic activity and protein aggregation are independent protein properties.
Correlation between TTR proteolytic activity and protein assembly and aggregation
To further investigate possible alterations in the protein oligomerization state introduced by the mutants described in the present paper, we subjected wtTTR and all TTR variants used in the present study to size-exclusion chromatography. All TTR species eluted in a single highly symmetrical peak (Figure 6c) with an elution volume of 12.25 ml, corresponding to an estimated Stokes radius of 3.75 nm. This size is in agreement with the size (3.20–3.40 nm) reported in the literature for the wtTTR tetramer [35,36]. No species of greater size were detected in any of the TTR variants elution profiles.
TTR is well known as a transporter of T4 and retinol. In the present study, we have shown that a change in the secondary structure of TTR can provide this protein with a complete new functionality. If it undergoes a partial conformational change, it acquires the ability to bind Zn2+ and to obtain proteolytic competence. Despite decades of research on TTR, its proteolytic activity remained elusive, a fact only explained by (i) the low activity of TTR, and (ii) the unfavourable equilibrium between proteolytic inactive and active conformation of wtTTR present at neutral pH. The structural lability of TTR in the region of amino acids 74–90 is clearly observable in wtTTR X-ray structures obtained at low pH (4.0 and 3.5) (PDB codes 3D7P and 3CBR) . These structures revealed significant changes in this region, proving the existence of multiple conformations in wtTTR. Recent NMR measurements  performed on wtTTR at pH 7.5 and increasing amounts of Zn2+ left the vast majority of the protein unaffected in the region 74–90 with only a small part assuming a modified conformation. The authors concluded that multiple conformations of wtTTR exist in a dynamic equilibrium with only a rather small fraction being able to complex Zn2+. These observations coincide well with our finding that only about 7% of total tetrameric wtTTR is in a proteolytic competent conformation at neutral pH.
Interestingly, the very short HXHXE Zn2+-binding pattern is novel and was not found in any metallopeptidase known to date. Since there was no perfect match detected in the MEROPS database , we looked for close analogues and found a permutated analogue HXEXH in family M22 (O-sialoglycoprotein peptidase) . Coincidentally, both enzymes are inhibited by EDTA and 1,10-phenanthroline, but are insensitive towards phosphoramidon, a well-known inhibitor of metallopeptidases.
A search of Pfam  through 597 TTR sequences revealed that only in humans and in some non-human primates are all three amino acids (His88, His90 and Glu92) conserved, whereas most other organisms lack His90. These findings imply that the proteolytic activity in TTR was introduced only very late in evolution possibly to modulate and fine-tune concentrations of specific substrates.
Interestingly, inhibition of proteolytic activity of TTR by metal chelators in vitro can be reversed by addition of various divalent metal ions as already observed for thermolysin . In TTR, Zn2+ and Mn2+ restored full proteolytic activity, whereas other metals such as Fe2+ and Co2+ only partially reactivated the enzyme. In vivo, the plasma concentration of a given metal and the stability of a metal–TTR complex will define which metal is preferentially bound to the TTR active site. The apparent dissociation constant [Kd (app)] of Zn2+–TTR is 1 μM , which suggests that plasma TTR may circulate as a complex with Zn2+. However, we cannot fully exclude that Mn2+ plays a (structural) role different from the one assumed for Zn2+ (catalytic), but other metallopeptidases such as thermolysin  or peptide deformylase  also tolerate multiple metal ions in their active site. Nevertheless, Mn2+, Fe2+ or Co2+ could not be detected in our TTR preparations.
We determined a kcat/Km value of 408.8 s−1·M−1 for the active fraction of TTR (7%), which translates to an effective kcat/Km value of 6.0×103 s−1·M−1. In comparison with other metallopeptidases, TTR has a lower (although comparable in some cases) activity than thermolysin (3×104 s−1·M−1) , ACE (angiotensin-converting enzyme) (2.9×105 s−1·M−1)  and neprilysin (3.5×106 s−1·M−1) . Nevertheless, taking into account the higher plasma concentration of TTR (approximately 5 μM, which corresponds to 350 nM of proteolytically active TTR) if compared with other typical plasma metallopeptidases such as ACE (205 pM) , one can conclude that TTR proteolytic activity and its physiological impact should not be neglected. Moreover, cleavage of prominent substrates such as apoA-I and Aβ, which are both involved in late-onset diseases of prime interest such as atherosclerosis and Alzheimer's disease, drastically increase the significance of TTR proteolysis. In fact, we have shown previously that cleavage of apoA-I by TTR might affect the development of atherosclerosis . In addition, TTR cleavage of Aβ peptide, identified in vitro, might influence Aβ deposition .
In summary, we have determined that TTR is a metallopeptidase and identified the residues important for catalytic activity. The identification of TTR catalytic residues enables the assessment of the physiological relevance of TTR proteolysis by the generation of transgenic mice carrying either human wtTTR or proteolytically inactive human TTR. Furthermore, knowing the TTR catalytic machinery may be useful for the design and screening of compounds modulating diseases that are dependent on TTR-mediated proteolysis such as Alzheimer's disease and atherosclerosis.
matrix-assisted laser-desorption ionization
Ana Damas, Daniel Bur and Mónica Sousa designed the study. Márcia Liz and Sérgio Leite carried out research. Maria Saraiva and Luiz Juliano contributed new reagents and analytical tools. Márcia Liz, Sérgio Leite, Maria Saraiva, Daniel Bur and Mónica Sousa analysed data. Márcia Liz, Daniel Bur and Mónica Sousa wrote the paper.
We thank Dr Matthew Bogyo (Stanford University, Palo Alto, CA, U.S.A.) for the phosphonate compound, Dr Marisa Almeida (Chemical Speciation and Bioavailability Laboratory, Center of Marine and Environmental Research, University of Porto) for the determination of metal levels in TTR, Dr Rita Costa (Instituto de Biologia Molecular e Celular, University of Porto) for the help with the thioflavin-T-binding assays and Dr Frederico Silva (Instituto de Biologia Molecular e Celular, University of Porto) for the analytical gel-filtration assays.
This project was supported by the Fundação para a Ciência e Tecnologia (FCT) [grant numbers PTDC/SAU-GMG/111761/2009 and PTDC/SAU-ORG/118863/2010] under the programs FEDER and COMPETE, and Association Française contre les Myopathies, France. In Brazil, the support was from Fundação de Amparo Pesquisa do Estado de São Paulo (FAPESP) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq). M.A.L. and S.C.L. are Fundação para a Ciência e a Tecnologia (FCT) fellows [grant numbers SFRH/BPD/34811/2007 and SFRH/BD/72240/2010].
These authors contributed equally to this work.
These authors contributed equally to this work.