On the evolutionary significance and metal-binding characteristics of a monolobal transferrin from Ciona intestinalis

Abstract

Transferrins are a family of proteins that bind and transport Fe(III). Modern transferrins are typically bilobal and are believed to have evolved from an ancient gene duplication of a monolobal form. A novel monolobal transferrin, nicatransferrin (nicaTf), was identified in the primitive ascidian species Ciona intestinalis that possesses the characteristic features of the proposed ancestral Tf protein. In this work, nicaTf was expressed in Pichia pastoris. Extensive solution studies were performed on nicaTf, including UV-vis, fluorescence, CD, EPR and NMR spectroscopies, and electrospray time-of-flight mass spectrometry. The expressed protein is nonglycosylated, unlike the protein isolated from the organism. This property does not affect its ability to bind Fe(III). However, Fe(III)-bound nicaTf displays important spectral differences from other Fe(III)-bound transferrins, which are likely the consequence of differences in metal coordination. Coordination differences could also account for the weaker affinity of nicaTf for Fe(III) (log K = 18.5) compared with bilobal human serum transferrin (HsTf) (log K = 22.5 and 21.4). The Fe–nicaTf complex is not labile, as indicated by slow metal removal kinetics by the high-affinity chelator tiron at pH 7.4. The protein alternatively binds up to one equivalent of Ti(IV) or V(V), which suggests that it may transport nonferric metals. These solution studies provide insight into the structure and function of the primitive monolobal transferrin of C. intestinalis for comparison with higher order bilobal transferrins. They suggest that a major advantage for the evolution of modern transferrins, dominantly of bilobal form, is stronger Fe(III) affinity because of cooperativity.

The widespread occurrence and diverse roles of the family of transferrin (Tf) proteins drive investigations into their evolution and function (1–3). Their binding and transport of hydrolysis-prone Fe(III) are well studied, although by no means fully explained. Transferrins are typically glycoproteins with two homologous lobes (N- and C-lobes, ≈40 kDa each). Each lobe usually coordinates one equivalent of Fe(III) with two tyrosinates, an aspartate, and a histidine and an exogenous bidentate synergistic anion, usually carbonate (1, 4), in a pseudooctahedral geometry. The synergistic anion is bound by conserved threonine and arginine residues (3). Conserved residues among transferrins overall are surprisingly modest, however; they exhibit ≈30% sequence identity (2, 3).

The best-studied transferrins include melanotransferrin (mTf), lactoferrin (lTf), ovotransferrin (oTf), and serum transferrin (sTf), which is most abundant in vertebrates (1–3). Among other possible roles, the lTf and oTf proteins promote bacteriostasis by sequestering iron (5, 6), mTf participates in cell proliferation and tumorigenesis (7), and sTf delivers Fe(III) to cells by receptor-mediated endocytosis (1). The Tf structure has been called an “ancient but useful scaffold” for building differently functioning proteins, including neurotoxin inhibitors, protease inhibitors, metal-binding proteins, and enzymes (8).

The transferrin family is a member of an even larger family with a periplasmic binding protein-like II fold (9). It is through this superfamily classification that the transferrins are related to the bacterial ferric binding proteins (Fbp). The Fbp proteins bind one equivalent of Fe(III) with similar but not identical coordination sites to the transferrins (10, 11); there is very low sequence identity (<10%), the metal ligands come from different regions of the fold, and metal binding appears to have evolved independently.

Modern bilobal transferrins are believed to have evolved by gene duplication from an ancestral Tf protein with just a single metal-binding lobe (12), after its divergence from the fold superfamily. The gene duplication event may be correlated with the evolution of a filtration kidney in higher organisms, because single-lobed transferrins are readily excreted by the kidney (13). Transferrin relatives with two or more lobes, however, are present in primitive organisms. Thus far, a true monolobal transferrin has been found only in ascidians, marine invertebrate chordates that are on the evolutionary boundary between vertebrates and invertebrates (14). Ascidians often sequester high concentrations of metals (15–20). They evolved 0.5–1 billion years ago, when the atmosphere was already oxidizing, and so Fe(III) handling mechanisms were already in place (19). This time frame overlaps the one (200–500 million years ago) in which the common ancestor to all of the modern transferrins occurred (21). A monolobal transferrin was first isolated from the ascidian Pyura stolonifera (22), and one from Halocynthia roretzi (23) was used to obtain the gene sequence. Characterization of the proteins, however, was quite limited.

Recently a monolobal transferrin, named nicatransferrin (nicaTf), was isolated from the plasma of the ascidian Ciona intestinalis. The native protein was characterized and its sequence analyzed in detail (R. Uppal, K. V. Lakshmi, and A.M.V., unpublished work). C. intestinalis is a model organism for ascidians: its genomic sequence has been fully determined, and it can be maintained in the laboratory and genetically manipulated (24). The species is more primitive (25–27) than H. roretzi and P. stolonifera; thus, it was predicted that a monolobal transferrin isolated from C. intestinalis would display properties more characteristic of the ancestral monolobal transferrin. The protein sequence of nicaTf [supporting information (SI) Fig. 4] possesses several of the predicted features of an ancestral Tf protein (12). It is a monolobal Tf that has more sequence identity to the N-lobe of modern transferrins and to melanotransferrin, the oldest to diverge of the bilobal transferrins (2, 3). All of the Fe(III)-binding residues are conserved in nicaTf: Asp-52, Tyr-78, Tyr-180, and His-237. The secondary-shell residues important for hydrogen bonding in human sTf (HsTf) (Gly-65, Glu-83, Tyr-85, and Arg-124) are mostly conserved. In nicaTf, the arginine residue, also important for binding of the synergistic anion, is conservatively replaced by lysine. The synergistic anion-binding residue Thr-120 (HsTf) is conserved. All transferrin lobes are believed to have six ancient disulfide bridges in topologically equivalent positions and additional disulfides that later evolved (28). NicaTf has all 12 cysteines comprising the six ancient disulfides but none of the four cysteines of the two modern disulfides present in the N-lobe of bilobal transferrins. Preliminary studies on the native protein have revealed that it is heterogeneously glycosylated and despite a difference in the proposed synergistic anion-binding residues, the protein does bind one equivalent of Fe(III) (R. Uppal, K. V. Lakshmi, and A.M.V., unpublished work).

To explore the biophysicochemical characterization of nicaTf, the protein was cloned and expressed in the yeast Pichia pastoris. NicaTf was isolated in high purity and exists as a monomeric species in solution. Unlike the native protein, the expressed protein is not glycosylated, but this property does not impede binding of one equivalent of Fe(III). Fe(III)-bound nicaTf was evaluated by competitive binding and kinetics, CD, EPR, NMR, fluorescence, and UV-vis studies. These studies elucidate important similarities and differences between the spectroscopic properties of Fe(III)-bound nicaTf and typical transferrins that reflect a modified coordination environment. The ability of nicaTf to bind other hard metals, Ti(IV) and V(V), was also examined. The metal-binding properties of nicaTf are analyzed in context with other transferrin-associated proteins found in C. intestinalis and are compared with those of HsTf. The role played by nicaTf in C. intestinalis and the metal-binding advantage for bilobal transferrin evolution are discussed below.

Results

Isolation and Characterization of NicaTf.

NicaTf was expressed in P. pastoris cells. This system has been used to express other transferrins (29–31). The purification is illustrated in SI Fig. 5. A single strong band near 35 kDa was observed by SDS/PAGE for the purified protein. A typical yield was ≈10 mg of protein per liter of culture.

The isolated protein was subjected to LCMS/MS analysis after trypsin digest. Peptide fragments belonging to the sequence of nicaTf from the C. intestinalis genome were identified (SI Fig. 6), confirming the identity of nicaTf. The protein coverage by total amino acid count was 94%.

NicaTf samples subjected to metal removal dialysis were confirmed to be Fe(III) free by a ferrozine assay and Ti(IV) and V(V) free by a 2,3-dihydroxynaphthalene-6-sulfonate assay. The solution of apo-nicaTf is slightly yellow in color. Its UV-vis spectrum reveals an absorbance at 320 nm different from that of the apo-protein isolated from C. intestinalis plasma (R. Uppal, K. V. Lakshmi, and A.M.V., unpublished work) that could be due to binding of an unidentified chromophore from the media, which is not uncommon (32, 33). The chromophore could not be removed by eluting the protein from a PD-10 desalting column (Amersham Biosciences) or by treating it with activated charcoal (Aldrich) (32).

The mass of nicaTf, as determined by ES TOF MS (35,378.81 ± 1.72 Da) is 16 Da greater than the predicted mass (35,363 Da) (SI Fig. 7). The mass difference is much lower than the mass of a sugar molecule and suggests that the protein is not glycosylated. This finding was confirmed by a phenol-sulfuric acid assay (34, 35). A UV-vis difference (treated–untreated) spectrum (SI Fig. 8) of the protein (17 μM) shows a negligible absorbance at 490 nm, indicating less than one equivalent of either an N- or O-linked sugar molecule (36).

The far- and near-UV CD spectra of apo-nicaTf were collected (SI Fig. 9). The far-UV region suggests lower helical content than apo-HsTf (37, 38), but it is uncertain to what degree the signal is masked by the protein-bound chromophore, which absorbs in this region.

Dynamic light scattering revealed that apo-nicaTf exists as a monomeric species in solution and that its hydrodynamic radius is 3.1 nm (SI Fig. 10). The hydrodynamic radius of apo-HsTf is 4.0 nm (data not shown). The measured hydrodynamic radius of nicaTf or of HsTf is unchanged by the presence of one equivalent of metal ion.

Characterization of Fe(III)-Bound NicaTf.

NicaTf binds up to one equivalent of Fe(III) (0.99 ± 0.07), as determined by the ferrozine assay. The presence of Fe(III) alters some of the spectroscopic properties of nicaTf. The UV-vis difference (holo-apo) spectrum (Fig. 1) of Fe-nicaTf shows a maximum absorbance at 306 nm (ε = 10,200 M⁻¹·cm⁻¹) and a shallow shoulder at 413 nm (Δε = 2,490 M⁻¹·cm⁻¹). Fe(III) binding quenches the fluorescence emission (λ_ex = 295 nm) of nicaTf at 382 nm (SI Fig. 11) by 26%. The EPR spectrum (SI Fig. 12) of Fe-nicaTf compares favorably with that reported for the native protein (R. Uppal, K. V. Lakshmi, and A.M.V., unpublished work). The spectral features are in accord with those for a S = 5/2, high-spin Fe(III) system. Three EPR transitions as expected are observed at g′ = 2.02, 4.31, and 9.16 (39). The EPR spectrum of Fe-nicaTf at g′ = 4.31 is more isotropic than that of Fe₂-HsTf(CO₃)₂ (39).

Fig. 1.

The UV-vis difference (holo – apo) spectrum of 19.8 μM Fe(III)-nicaTf.

The concomitant binding of a synergistic anion when nicaTf binds Fe(III) was examined. A citric acid assay (40) was conducted on nicaTf that was Fe(III)-loaded with iron citrate (metal:ligand ratio of 1:21). The samples have no citrate present, suggesting that nicaTf does not coordinate Fe(III) with citrate serving as a synergistic anion. A proton-decoupled ¹³C NMR experiment was conducted to determine whether carbonate serves as a synergistic anion. Because Fe(III) is paramagnetic, the ¹³C NMR signal of iron-bound (H)CO₃⁻; in the HsTf-binding sites is broadened beyond detection (41, 42). Fig. 2 shows the ¹³C NMR spectra of H¹³CO₃⁻ in the presence of nicaTf with and without one equivalent of iron NTA (ratio of 1:2). The peak at 160.3 ppm is the free or protein-bound H¹³CO₃⁻. The broad peak centered at 174 ppm comes from protein carbonyl carbons. The (H)CO₃⁻ signal disappears upon the addition of one equivalent of iron NTA (metal ligand ratio of 1:2) but reappears (data not shown) when an additional equivalent of iron NTA is supplied. The data indicate that (H)CO₃⁻ does bind to iron in the protein binding site but is competed by NTA when more than one equivalent of iron NTA is available at these concentrations. NTA appears to also serve as a synergistic anion as indicated by an independent equilibrium dialysis experiment in which one equivalent of Fe(III), from iron NTA (1:2), bound to nicaTf in a solution thoroughly purged of bicarbonate.

Fig. 2.

The proton-decoupled ¹³C NMR spectrum of 170 μM nicaTf and 170 μM H¹³CO₃⁻ in the presence (A) and absence (B) of one equivalent of iron NTA (1:2). The peak at 160.3 ppm is the free H¹³CO₃⁻. The broad peak centered at 174 ppm comes from protein carbonyl carbons.

The affinity of nicaTf for Fe(III) was measured by a competition experiment with NTA via equilibrium dialysis at 25°C and pH 7.56 using microdialyzers. The concentration of HCO₃⁻ (20 mM) was maintained significantly higher than that of NTA to favor synergistic binding of carbonate. The concentration of bound and unbound Fe(III) in the reaction of 37.2 μM nicaTf (placed on one side of the dialyzer) with 47.6, 48.8, and 58 μM Fe(III) in the presence of 113 μM HNTA²⁻ (placed on both sides of the dialyzers) are reported in SI Table 1. Apparent affinity constants (K_app) were determined from these data by using the equation (Eq. 1) for a single binding site (43):

where B is the ratio of bound Fe(III) to nicaTf and L is the free concentration of Fe(III). K_app was then corrected for the overall equilibrium (K₁) including the appropriate speciation of Fe(III) NTA (SI Table 2) according to Eq. 2

where x is the total number of Fe(III)-bound NTA, y is the total number of Fe(III) bound hydroxide, and z is the total charge of the Fe(III) species. For comparison with Fe(III) binding to a lobe of HsTf, three protons are assumed to be released, two from the tyrosines and one from HCO₃⁻ when Fe(III) binds to nicaTf. SI Table 3 reports all of the equilibria and corresponding stability constants pertaining to the reaction with 47.6 μM Fe(III). The average Fe-nicaTf(CO₃)²⁻ binding constant (K_eq) was determined to be 2.9 × 10¹⁸ at pH 7.56.

The lability of the Fe-nicaTf(CO₃)²⁻ complex was examined by monitoring the kinetics of metal release with tiron, a high-affinity Fe(III) chelator (44). The growth of the LMCT band of Fe(tironate)₃⁹⁻ was followed by UV-vis under pseudofirst-order conditions (excess tiron) at 25°C and pH 7.4. More than one phase is involved in the reaction of 14.7 μM Fe-nicaTf with 5 mM tiron (Fig. 3). The data were fitted with three exponentials: k_obs1 = 4.2 × 10⁻³ (± 5 × 10⁻⁴) s⁻¹ (A₁ = 0.04064), k_obs2 = 2.8 × 10⁻⁴ (± 3 × 10⁻⁵) s⁻¹ (A₂ = 0.01054), k_obs3 = 1.2 × 10⁻⁵ ± (2 × 10⁻⁶ s⁻¹) (A₃ = 0.0375) and ΔA_total = 0.0887. Fitting with two exponentials yielded a fit with greater residuals (SI Fig. 13). Despite the rapidity of the initial phase of metal release, nicaTf is able to retain a significant amount of Fe(III) for several hours in the presence of a high concentration of tiron. After 24 h of reaction, Fe(III) is fully complexed by tiron as confirmed by the absence of Fe(III) in the final sample.

Fig. 3.

Metal release kinetics of 14.7 μM Fe-nicaTf with 5 mM tiron. The reaction of 14.7 μM Fe-nicaTf with 5 mM tiron after the growth of the LMCT for Fe(tironate)₃⁹⁻. Reaction condition: pH 7.4, 25°C, and I = 0.1 M. Data (solid line) are overlaid with multiple exponential fit (dashed line) with k_obs1 = 4.2 × 10⁻³ (±5 × 10⁻⁴) s⁻¹ (A₁ = 0.04064), k_obs2 = 2.8 × 10⁻⁴ (±3 × 10⁻⁵) s⁻¹ (A₂ = 0.01054), k_obs3 = 1.2 × 10⁻⁵ ± (2 × 10⁻⁶·s⁻¹) (A₃ = 0.0375), and ΔA_total = 0.0887.

NicaTf Binding of Ti(IV) or V(V).

At the metal and citrate concentrations used for studying Ti(IV) or V(V) binding to nicaTf, Ti(IV) is predicted to be present as Ti(C₆H₄O₇)₃⁸⁻ (45), and V(V) is predicted to be present as dimeric monocitrate complexes with varying protonation states (46). When 100 μM citrate is included to prevent potential hydrolysis of unbound metal in the dialysis buffer, one-half to one equivalent of Ti(IV) or V(V) is bound to nicaTf, as indicated by the respective 2,3-dihydroxynaphthalene-6-sulfonate assays. A clearly defined LMCT band for Ti(IV) bound nicaTf was not observed, but a shallow shoulder at 321 nm (Δε = 5,700 M⁻¹·cm⁻¹) occurred in the UV-vis difference spectrum (data not shown). No LMCT band for V(V)-bound nicaTf was observed.

Discussion

Although the protein sequence of nicaTf suggested that this monolobal transferrin possesses characteristics of a proposed single-lobed ancestral Tf protein (12), the biophysicochemical study performed in this work shows that nicaTf has several structural and spectroscopic features distinct from modern bilobal transferrins. The heterologously expressed apo-nicaTf was nonglycosylated. The absence of glycosylation, however, does not impede binding of Fe(III) or other hard metals such as Ti(IV) and V(V), which suggests that the function of glycosylation in the native protein is not related to metal binding. Glycosylation of transferrins in general has little effect on their Fe(III) binding (47). The hydrodynamic radius of apo-nicaTf (3.1 nm) determined by dynamic light scattering is bigger than one lobe of apo-HsTf (≈2.0 nm). The apparent lower helical content, monitored by far UV CD, of the heterologously expressed nicaTf and even of the native protein (data not shown) relative to HsTf (38) could account for the bigger radius.

There are some striking differences in the spectroscopic signals because of metal binding by nicaTf. Fe(III) binding to nicaTf does not cause detectable changes in the near-UV region of the protein CD spectrum. Significant perturbations of the aromatic residues in this region are observed when Fe(III) binds to HsTf (38). The fluorescence emission of nicaTf (λ_ex = 295 nm) is quenched by only 26% on binding Fe(III), whereas for HsTf it is quenched by 75% (48). One possible explanation for the difference in fluorescence quenching is that there may be smaller global conformational changes because of Fe(III) binding, and so the chemical environment surrounding the tryptophan residues is not greatly altered. Another possibility arises from the differences of the Trp residues in nicaTf and HsTf. In the N-lobe of HsTf, Trp-128 and -264 are responsible for almost the entire fluorescence change upon iron binding (49), and nicaTf lacks a tryptophan in one of these positions (SI Fig. 14). NicaTf has additional Trp residues not present in either lobe of HsTf that may contribute to fluorescence, but that may not be much affected by metal binding. The EPR spectrum of Fe-nicaTf is quite different from that of Fe₂-HsTf(CO₃)₂, particularly at g tensor 4.31. EPR simulations of Fe₂-HsTf show that the line shapes of the spectra are quite sensitive to changes in the ligand field (50). The line shapes are dictated by the zero-field splitting (ZFS) parameter ratio, E/D (E = rhombic ZFS and D = axial ZFS) (50), and change dramatically for Fe₂-HsTf depending on the synergistic anion bound.

Differences in the metal-nicaTf coordination environment are quite evident in the UV-vis difference spectrum. For Fe(III) bound to HsTf, the tyrosine ligand to metal charge transfer band occurs at 465 nm with a molar absorptivity of 2,600 M⁻¹·cm⁻¹ per lobe when carbonate serves as a bidentate synergistic anion (4). In contrast, a shallow shoulder appears at 413 nm in the UV-vis difference spectrum of Fe-nicaTf with a comparable molar absorptivity of 2,490 M⁻¹·cm⁻¹. The shift to higher energy of the Fe-nicaTf LMCT band is likely the result of a modified coordination environment. The λ_max for Fe₂-HsTf can greatly shift depending on the coordination (4). The λ_max can range from 415 to 510 nm when the synergistic anions glycolate and thioglycolate bind, respectively (4). NicaTf might bind Fe(III) by coordinating the Fe(III) delivery source intact or with carbonate coordinated to the metal center in a monodentate fashion (R. Uppal, K. V. Lakshmi, A.M.V., unpublished work). Similarly the LMCT band of Ti-nicaTf is not a distinct absorbance in the UV-vis difference spectrum but a very shallow shoulder. The molar absorptivity, however, is comparable with that observed for hydrolyzed Ti(IV) bound to HsTf (51). Although no LMCT band is detectable for V(V)-nicaTf, V(V) is likely coordinated as VO₂⁺, similar to its coordination in HsTf (52), because of the high affinity of the oxo ligands to V(V).

The coordination of Fe(III) by nicaTf was investigated particularly with respect to the synergistic anion. A citric acid assay conducted on Fe-nicaTf prepared with iron citrate (1:21) showed that no citrate was present even though one equivalent of Fe(III) was bound, indicating that Fe(III) is not bound as a citrate complex nor with citrate serving as a synergistic anion. Citrate also does not serve as a synergistic anion for Fe(III) binding to HsTf (4) although it can serve as one for a Fbp (53). A proton-decoupled ¹³C NMR experiment suggested that carbonate is bound as a synergistic anion. The ¹³C NMR signal of bound (H)CO₃⁻ broadened beyond detection upon addition of one equivalent of iron NTA (1:2). This experiment, however, does not distinguish between a monodentate or bidentate coordination by carbonate.

To evaluate the affinity constant of Fe-nicaTf with carbonate as synergistic anion, an assumption was made that three protons are released upon Fe(III) binding as when binding to HsTf (54). [For the N-lobe of transferrins, an additional proton can be liberated from one of the two noncoordinating lysines involved in a proposed Fe(III) releasing “dilysine trigger” (55). This trigger is not conserved in nicaTf.] The affinity constant of Fe-nicaTf (log K = 18.5) is weaker relative to that of HsTf (log K_{C lobe} = 22.5; log K_{N lobe} = 21.4) (56–58) but sufficiently high to protect Fe(III) from hydrolysis. The affinity of nicaTf is also presumed to be reasonably high for Ti(IV) and V(V), considering that the metal loading occurred in the presence of significantly elevated citrate concentration (5 mM). The weaker affinity of nicaTf for Fe(III) may be because of nicaTf being less preordered for binding Fe(III) relative to HsTf. There is an extensive hydrogen bond network for Fe(III) and carbonate binding in HsTf (1, 2). The presence of Lys in place of the Arg important for binding of the synergistic anion in modern transferrins could result in a weaker hydrogen bond network that stabilizes the synergistic anion or alters the way the anion is bound in nicaTf. These differences might result in weaker Fe(III) affinity and distinct spectroscopic signals for Fe-nicaTf.

Despite the weaker affinity of nicaTf for Fe(III) relative to HsTf, the Fe-nicaTf is not a labile complex. As with HsTf and even Fbp, the kinetics of iron release from nicaTf are multiphase. It involves at least three slow kinetic phases that remain to be investigated. Several catechol, hydroxypyridone, and phosphonic acid-based ligands form exceptionally stable Fe(III) complexes (59). One would expect these ligands to easily outcompete the proteins for binding Fe(III) at pH 7.4 kinetically, although these proteins can compete for Fe(III) binding. Many of the Fe(III) high-affinity ligands require a considerable excess to remove bound Fe(III) at rates on the order of hours (60, 61). NicaTf is able to retain a significant amount of bound Fe(III) in the presence of high concentrations of the strong affinity chelator tiron (59). Iron-accumulating ascidians typically have high concentrations of phenolic metabolites (62) and DOPA-containing polypeptides (63), and tiron is a good model of these molecules. That Fe(III)-nicaTf is not readily labile suggests that the protein would be an efficacious delivery agent for Fe(III) to target receptors.

The relatively high affinity of nicaTf for Fe(III), Ti(IV), or V(V) implicates a specific role for nicaTf as a metal transporter. Metal transport is a role played by HsTf for Fe(III) and proposed for Ti(IV) (51, 64, 65) and V(V) (52, 66, 67) and for other metal ions (68). This role is a likely possibility for nicaTf, considering that ascidians accumulate these ions at high concentration, many orders of magnitude higher than their environment, depending on the species (15–20, 69). The remarkable levels of these metal ions would almost certainly be toxic for humans. The transport function suggests the presence of a transferrin receptor or iron acceptor on cell surfaces in C. intestinalis. The genome of C. intestinalis includes several genes that might encode Tf receptors and an apparent bilobal Tf, which is expected to be membrane bound (SI Fig. 4) (R. Uppal, K. V. Lakshmi, and A.M.V., unpublished work). NicaTf may also play a bacteriostatic function like lactoferrin (70) if its affinity for Fe(III) is high enough to maintain the metal bound in its environment. Evidence for this role comes from the observation that nicaTf retains Fe(III) binding in the presence of high concentrations of a strong affinity chelator. It is not expected, however, that nicaTf would play a long-term storage function. C. intestinalis possesses the genes for four postulated iron-storing ferritins.

In conclusion, nicaTf, a monolobal transferrin from C. intestinalis, possesses several important properties characteristic of the ancestor of modern transferrins. Although the heterologously expressed nicaTf is not glycosylated like the protein isolated from the plasma of the organism, this trait does not affect its metal binding properties. NicaTf binds up to one equivalent of Fe(III), Ti(IV), or V(V), which indicates that it may serve a metal transport role like HsTf. The affinity of nicaTf for Fe(III) with bound carbonate is weaker than HsTf, suggesting that nicaTf may be less specific than higher-order transferrins for Fe(III) transport, and that it may serve additional functions. Weaker affinity of Fe-nicaTf may be because of differences in the primary and/or secondary metal coordination sphere, which could account for the similar yet distinct spectroscopy of the holo-protein relative to other transferrins. We predict that the structure of nicaTf will show a lower helical content, a site less preorganized for binding Fe(III) in the apoprotein, and a weaker hydrogen bond network in the holo-protein.

An ancient gene duplication resulted in the formation of bilobal transferrin. It was postulated that this process was evolutionarily connected with the emergence of a filtration kidney (13); however, the presence of bilobal Tf in species lacking a filtration kidney (e.g., insects and ascidians) provides evidence against this argument. Instead, the structural differences and weaker Fe(III) affinity exhibited by primitive nicaTf suggest that the gene duplication is correlated with an increase in Tf Fe(III) affinity to maintain iron homeostasis. The cooperativity (56, 71–74) that exists between the two lobes of modern transferrins increases Fe(III) affinity for each lobe and concomitantly improves protein stability. Cooperativity would explain why modern transferrins are dominantly bilobal.

Materials and Methods

Cloning and Expression of NicaTf.

An Escherichia coli stab culture with the gene for nicaTf (clone cibd013a17) in a pBluescriptII vector was provided by Noriyuki Satoh (Kyoto University, Kyoto, Japan) and Yuji Kohara (National Institute of Genetics, Japan).

Cloning and expression of nicaTf in the yeast P. pastoris was done with a Pichia Expression Kit (Invitrogen K1710–01). A Coomassie-stained SDS/PAGE gel band containing nicaTf was analyzed by the Tufts University Core Facility by liquid chromatography (LC)MS/MS. The MS/MS spectra were searched against the C. intestinalis genome (75). [Genome resources for C. intestinalis can be found at Kyoto University (http://ghost.zool.kyoto-u.ac.jp/indexr1.html) and at the Department of Energy Joint Genome Institute (http://genome.jgi-psf.org/Cioin2/Cioin2.home.html)]. Further details are found in SI Text in addition to information on all reagents and chemicals used.

Preparation of Apo-nicaTf.

Apo-nicaTf was prepared by dialyzing the protein against 4 liters of 0.1 M sodium acetate and 10 mM sodium EDTA at 4°C and pH 5.00 for 2 d. The protein was then dialyzed against 4 liters of 50 mM Hepes and 0.1 M NaCl in 1-liter portions for >2 d. Protein samples were quantified by the DC Assay and metal content was evaluated by using ferrozine and 2,3-dihydroxynaphthalene-6-sulfonate assays (see below). UV-vis spectra were recorded on a Cary 50 spectrophotometer (Varian). Unless otherwise specified, the protein buffer consisted of 50 mM Hepes, 0.1 M NaCl, and 20 mM NaHCO₃.

Characterization of NicaTf.

NicaTf was characterized by electrospray (ES) TOF (Micromass LCT ES TOF mass spectrometer), protein glycosylation (phenol-sulfuric acid) assay, dynamic light scattering (DLS; Protein Solutions DynaPro99) and CD (AVIV model 215). Experimental details are available in SI Text.

Metal NicaTf-Binding Studies.

Following are details of the determination of Fe(III) binding. An iron citrate solution was prepared by dissolving 1.4705 g of Na₃Citrate (5 mmol) in water and adding 0.0764 g of Fe(citrate) (0.25 mmol). The solution was then buffered to pH 7.4 at 4°C. Apo-nicaTf was dialyzed in the iron citrate solution for several days at 4°C. The protein was then dialyzed in four changes of iron citrate free buffer over 4 d. An alternative source of Fe(III) was 1 mM iron NTA in a ratio of 1:2. Fe(III) content was quantified by the ferrozine assay (SI Text).

Fluorescence Study.

Apo-nicaTf (10 μM) and Fe(III)-loaded nicaTf (10 μM) were examined at 25°C by fluorescence emission (λ_ex = 295 nm) (Photon Technology International QM-4 spectrofluorometer). The excitation and emission slit widths were 1 and 10 nm, respectively.

EPR Study.

Fe(III)-nicaTf (500 μM) was prepared with iron citrate and extensively dialyzed in pH 7.4 Hepes buffer. Glycerol (20% total) was added. The solution was Ar purged to remove O₂ and then frozen in liquid N₂.

The EPR measurement was performed on an Elexys 500 spectrometer (Bruker Instruments) operating at a microwave frequency of 9.382 GHz, equipped with a TE₁₀₂ cavity and a helium flow cryostat (Oxford Instruments) at 4.5 K. Three scans were collected at a microwave power of 25 mW. The center field of the EPR scans was set to 2,000 G with a sweep width of 3,400 G.

Citric Acid Assay.

A citric acid assay was performed on 85.2 μM Fe-nicaTf prepared from Fe citrate (details in SI Text) (40).

¹³C NMR Study.

A 170-μM apo-nicaTf sample was prepared in 20 mM (pH 7.4) Tris buffer (90:10 D₂O/H₂O, 170 μM H¹³CO₃⁻, 0.1 M NaCl). A proton decoupled ¹³C NMR spectrum was collected. Then one equivalent of concentrated iron NTA (1:2 metal/ligand) was added, and another scan was obtained. An additional equivalent of iron NTA was added. Approximately 60,000 scans were collected on a Bruker 500-MHz instrument.

Affinity of NicaTf for Fe(III).

NicaTf (37.2 μM) was dialyzed in pH 7.56 buffer. Three Fe(III) NTA solutions were prepared that contained 20 mM NaHCO₃,226 μM NTA, and 95.2, 97.6, and 116 μM Fe(III) in the final dialysis buffer. The protein and chelate solutions were reacted in dialyzers over 4 d at 25 ± 1°C and agitated at 200 rpm to quantify the bound Fe(III) content. The Fe-nicaTf affinity constant was determined with a single binding site model (43) and by correcting for NTA affinity for Fe(III) (76). All experiments were performed in triplicate.

Kinetics of Fe(III) Release from NicaTf.

Fe-nicaTf (14.7 μM) was reacted with 5 mM tiron. The formation of Fe(tironate)₃⁹⁻ (ε = 6,200 M⁻¹·cm⁻¹) was monitored at 481 nm (77). The kinetic profile was fitted to multiple exponentials using Origin 6.0. After 24 h, the solution was rapid-spin dialyzed to determine the iron content.

NicaTf Binding of Titanium(IV) and Vanadium(V).

Ti(IV) citrate solution was prepared by dissolving 0.3572 g of Ti(citrate)₃²⁻ (0.5 mmol) in 500 ml of pH 7.4 buffer. Apo-nicaTf was dialyzed in Ti(citrate)₃⁸⁻ solution (45) for 2 d at 4°C and then dialyzed in four changes of Ti(citrate)₃⁸⁻ free buffer that included 100 μM citrate over 4 d. The titanium content in the final sample was quantified by the 2,3-dihydroxynaphthalene-6-sulfonate chelation assay (51).

V(V)-bound nicaTf was prepared similarly. V(V) citrate solution was made by dissolving 0.0586 g of NH₄VO₃ (0.5 mmol) in 500 ml of buffer. Excess V(V) was removed with 100 μM citrate containing buffer. The vanadium content was quantified by the 2,3-dihydroxynaphthalene-6-sulfonate chelation assay (ε_{500 nm} = 4,600 M⁻¹·cm⁻¹).