Lanthanide binding and IgG affinity construct: Potential applications in solution NMR, MRI, and luminescence microscopy

Abstract

Paramagnetic lanthanide ions when bound to proteins offer great potential for structural investigations that utilize solution nuclear magnetic resonance spectroscopy, magnetic resonance imaging, or optical microscopy. However, many proteins do not have native metal ion binding sites and engineering a chimeric protein to bind an ion while retaining affinity for a protein of interest represents a significant challenge. Here we report the characterization of an immunoglobulin G-binding protein redesigned to include a lanthanide binding motif in place of a loop between two helices (Z-L2LBT). It was shown to bind Tb³⁺ with 130 nM affinity. Ions such as Dy³⁺, Yb³⁺, and Ce³⁺ produce paramagnetic effects on NMR spectra and the utility of these effects is illustrated by their use in determining a structural model of the metal-complexed Z-L2LBT protein and a preliminary characterization of the dynamic distribution of IgG Fc glycan positions. Furthermore, this designed protein is demonstrated to be a novel IgG-binding reagent for magnetic resonance imaging (Z-L2LBT:Gd³⁺ complex) and luminescence microscopy (Z-L2LBT: Tb³⁺ complex).

Introduction

The development of techniques to harness the unique properties of lanthanide series metals can have substantial impact on the field of structural and molecular biology. These properties have been exploited in the past using techniques as diverse as nuclear magnetic resonance (NMR) spectroscopy,¹ electron paramagnetic resonance (EPR) spectroscopy,^2,3 magnetic resonance imaging (MRI)⁴, X-ray crystallography,⁵ and luminescence imaging.^6,7 In NMR the relaxation enhancing and chemical shift perturbing properties of paramagnetic lanthanide ions give rise to ion-nucleus distance measurements in the 10 to 50 Å range. In EPR, more specifically DEER applications, this range can approach 100 Å.⁸ In MRI these same paramagnetic properties give rise to image contrast enhancement, and, in microscopy applications, long-lived Tb³⁺ and Eu³⁺ luminescence (>2 ms)^7,9 offers potentially unique advantages in background suppression and multiple photon excitation.⁶ The luminescent properties are also useful in the determination of ion dissociation constants or distance-dependent energy transfer to a nearby fluorophore.⁹ To fully capitalize on these properties, the capacity to bind a lanthanide ion into a molecule of interest must be developed. Here we present the redesign of a small protein domain known to bind the immunoglobulin G (IgG) fragment crystallizable (Fc) to include a lanthanide binding motif. The use of this particular domain amplifies the potential impact because of the widespread use of IgG-based antibodies in cell biology and biomedical research.

The actual choice of IgG, a small binding protein, and a lanthanide binding motif were strongly influenced by our interest in carbohydrate–protein interactions and how covalently-attached glycans in glycoproteins impact protein function.^10,11 IgG Fc and its attached glycans are of particular interest as these glycans modulate the interaction of Fc with the various Fcγ receptors.¹² Furthermore, alteration in glycan composition can shift the normally proinflammatory Fc into a potent anti-inflammatory mediator.^13,14 An elucidation of the structural and dynamic characteristics of the glycans by methods such as solution NMR would do much to provide an understanding of these effects. IgG Fc is a dimer that contains two Asn-linked glycans, but structural studies of these glycans are limited by dynamic motions that prevent observable glycan-polypeptide distance measurements in the form of NOE crosspeaks in isotope edited NMR experiments.¹¹ Also, preparing the properly glycosylated and isotope labeled Fc protein is challenging, due to the lack of N-glycosylation machinery in Escherichia coli and inappropriate glycosylation found when using many eukaryotic expression techniques.¹⁵ Even if the protein amino acids could not be isotopically labeled, characterization of the IgG glycans would be facilitated by attachment of a lanthanide ion to the Fc polypeptide, as would the structural elucidation of molecules in complex with Fc.

Hence, we have resorted to the incorporation of a lanthanide binding capacity into the Z-domain of the Fc-binding protein, Protein A. This new chimeric protein can be expressed at will and used with even native, non isotope enriched, isolations of IgG Fc fragments. The paramagnetic properties of lanthanide ions can then provide valuable long-range orientation and distance information which can supplant measurements of short-range nuclear dipole-dipole interactions (NOEs, <5–6 Å),^16,17 the traditional foundation of structure determination by solution-state NMR spectroscopy.¹⁸

Many strategies to label molecules of interest with lanthanide binding motifs have been presented previously, including: covalent modification with a metal chelate,^5,17 integration of an unnatural amino acid carrying a chelate moiety,¹⁹ and the incorporation of an internal lanthanide binding polypeptide sequence into the protein expression construct.^20–22 Many of these strategies provide limited benefits for structure-based investigations due to the conformationally labile nature of the lanthanide-binding motifs. Steric restriction has been achieved by using rigid chelates,²³ chelates with multiple protein attachment sites,²⁴ and integration of a lanthanide binding peptide into the middle, as opposed to the terminus of a protein sequence.²⁵ In general we prefer the lanthanide binding peptide approach for its convenience of production by expression in a bacterial host. With our choice of a small protein with affinity for IgG Fc this is entirely feasible. We also prefer the introduction in midsequence since our long-term objectives are to use the chimeric protein in structural characterization, and restricting internal mobility is important.

The lanthanide binding peptide we chose is a short ∼20 amino acid sequence that can be added to the termini of protein sequences or in place of loop structures already in the protein to be modified.^21,24,25 Substitution for native loop structures in a manner that preserves both lanthanide binding sequences and affinity for a target protein, IgG Fc in our case, is not always straight forward. We have therefore used generous linkers between the peptide and protein at some sacrifice to rigidity for this initial application. In this way, we have succeeded in producing a modified Z-domain with a lanthanide binding motif (or tag, LBT) inserted between helices two and three of the Z-domain three helix bundle. The construct has been demonstrated to preserve binding properties for both lanthanides and IgG Fc, and we have illustrated the utility of paramagnetic perturbations by determining the position of the lanthanide in a model of the modified Z-domain, as well as assessing the distribution of glycan conformers in IgG Fc. In addition, we have made preliminary applications to demonstrate the potential of this construct as a reagent to enhance contrast for MRI and confocal luminescence microscopy.

Results

Design of the lanthanide binding Z-domain

The Z-domain, a designed protein based on the B domain of the Staphylococcus aureus Protein A,²⁶ was chosen as a target for insertion of a lanthanide binding motif because of its high affinity for IgG Fc,²⁷ relative structural simplicity,^28,29 and thermal stability.³⁰ In an effort to reduce the conformational heterogeneity experienced by a motif attached at only the N or C terminus (data not shown), the motif was inserted in place of the loop between helices 2 and 3. The sequence of the lanthanide binding motif (Fig. 1) is based upon that of Nitz et al.²⁰ as modified by Su et al.²² to remove the two C-terminal residues (Leu and Ala) and replace the Trp with Ala. The entire construct, which also contains a 6xHis tag at the N-terminus, is shown in Figure 1. This design, Z-L2LBT, was expressed at a high level in E. coli cells (>20 mg/mL) and was of high purity (>95%) after a single immobilized Ni²⁺ affinity chromatography step (Supporting Information Fig. S1). Binding of Tb³⁺ to purified Z-L2LBT resulted in the appearance of a peak at ∼280 nm in the luminescence absorbency profile, suggesting Tb³⁺ sensitization occurs through proximal Tyr sidechains in the protein [Fig. 1(B)]. A titration of Z-L2LBT with Tb³⁺ revealed a dissociation constant of 130 ± 30 nM [Fig. 1(C)], a value consistent with expectations based upon a similar polypeptide sequence.²⁰ Furthermore, Z-L2LBT was shown to bind lanthanide ions and IgG Fc simultaneously using NMR titrations at 25°C and 50°C as well as SDS-PAGE-detected binding to an IgG2a-Sepharose resin in the presence of Tb³⁺ at 25°C (data not shown).

Figure 1.

The chimeric Z-L2LBT protein (A) is built upon a IgG-binding Z-domain scaffold and contains a lanthanide binding motif in place of the loop between helices 2 and 3. (B) Luminescence intensity changes with the Tb³⁺coordination state. The shift of the Tb³⁺:Z-L2LBT profile to the UV region likely reflects sensitization of the Tb³⁺ ion through Tyr residues in the protein. A sharp instrument artifact observed at ∼272 nm in all samples including a water blank was removed from these plots. (C) Luminescence intensity saturates at higher concentrations of Tb³⁺ when titrated into a 0.85 μM solution of Z-L2LBT. These data are fit with Eq. (1).

Determining the location of the lanthanide ion using paramagnetic effects

The designed Z-L2LBT protein was highly soluble and a heteronuclear single quantum coherence (HSQC) spectrum of a ¹⁵N-labeled version showed ∼80 amide crosspeaks of similar intensity. Roughly 30 crosspeaks disappeared and reappeared in a different location during a titration of Z-L2LBT with diamagnetic Lu³⁺ ions. A spectrum for the Lu³⁺-saturated Z-L2LBT is shown in Figure 2. The dissociation of the Lu³⁺:Z-L2LBT complex was slow on the NMR timescale (data not shown) and agreed with expectations based upon the relatively small Tb³⁺ dissociation constant reported above. The crosspeaks were assigned using a suite of standard triple resonance experiments.

Figure 2.

¹⁵N heteronuclear single quantum coherence spectrum of Z-L2LBT bound to a diamagnetic Lu³⁺ ion (black spectrum) or a paramagnetic Dy³⁺ ion (grey spectrum). The PCS and PRE effects of the Dy³⁺ ion appear as frequency shifts (diagonal lines), broadened lines and reduced intensity (inset), respectively, of ¹H-¹⁵N cross peaks. The numbers indicate residue numbers in the amino acid sequence. For protein preparation and purity see also Supporting Information Figure S1.

An ¹⁵N-HSQC spectrum of a complex between Z:L2LBT and a paramagnetic Dy³⁺ ion showed clear paramagnetic effects including resonance frequency perturbations (pseudocontact shifts (PCS)), reduced crosspeak intensities, and increased ¹H linewidths (paramagnetic relaxation enhancement (PRE)) as shown in Figure 2. The characteristic shift of ¹⁵N-¹H crosspeaks along a diagonal line, resulting from a near equal perturbation of ¹⁵N and ¹H resonance positions on a parts-per-million (ppm) scale, provides a useful means of correlating peaks observed in a paramagnetic complex with those already assigned for the diamagnetic (Lu³⁺) complex. The mere presence of PCSs is suggestive of reduced LBT dynamic freedom that was not observed when using a Z-domain modified with a C-terminal LBT (data not shown). The magnitudes of the induced shifts were determined for Z-L2LBT complexed with Ce³⁺, Yb³⁺, and Dy³⁺ ions, and ¹H PRE effects were determined from ¹H R₂ relaxation rates. In total, 68 PCSs and 46 PREs were measured (see Table I and Supporting Information).

Table I.

Summary of Experimental NMR Constraints and Structural Statistics

Dihedral-angle constraints	114
Residual dipolar coupling constraints
H-N	54
N-C(O)	34
Paramagnetic relaxation enhancement constraints	#	R2a
Yb3+	23	0.70
Dy3+	23	0.71
Pseudocontact shift constraints	#	R2a
Ce3+	11	0.61
Yb3+	28	0.46
Dy3+	29	0.86
Total number of restricting constraints	313
Rmsd
Rmsd of backbone  atoms (19–87)	0.4 Å
Bond length	0.04 Å
Bond angle	2.2°
Ramachandran analysis
Most favored region	88.4%
Allowed region	11.6%
Generously allowed	0.0%
Disallowed	0.0%
PSVS Z-scores31
Verify 3D32	−2.4
Prosall33	1.1
Procheck (phi-psi)34	−0.6
MolProbity35	−3.8

^aCorrelation between observed values and values back-calculated from the models.

Structures of the independent Z-domain and the LBT have been determined previously by NMR (pdb-1q2n²⁹) and X-ray crystallography (pdb-1tjb³⁶), respectively, and assuming induced structural changes upon substituting the LBT for a Z-domain loop were minimal, it is appropriate to use these in devising a model for the designed Z-L2LBT construct. To validate this assumption, residual dipolar coupling (RDC) data were collected on a Lu³⁺ bound version of the construct; these are also reported in Table I. RDCs provide high-resolution orientation information on residue-specific ¹H-¹⁵N bond vectors. Retention of previously determined structures can be assessed based upon a Q value that compares experimental RDCs and RDCs back-calculated from a proposed structure.³⁷ Q values <0.4 are suggestive of a high degree of structural similarity between the actual structure and a proposed model. Q values of Z-L2LBT ¹⁵N-¹H and ¹⁵N-¹³C(O) RDCs measured in PEG bicelles are 0.201 and 0.369, respectively, when compared with the X-ray crystallography structure of the unmodified Z-domain and 0.338 and 0.363, respectively, when compared with the X-ray crystallography structure of the LBT. In addition, the resonance frequencies for residues in the three helices were quite similar to those previously determined for the isolated Z-domain,²⁹ and these were used to confirm backbone torsional constraints. The RDC Q values and differences between the resonance frequencies are sufficiently low to suggest a high degree of similarity between the previously determined structures of the individual domains and those domains in Z-L2LBT. Thus, the structure determination was restricted to finding the relative orientation and position of the LBT and Z portions.

The approximate location of the lanthanide ion relative to the three helix bundle was calculated initially using PCS and PRE constraints (Table I) in a grid search. The LBT and linking loop segments were then docked to satisfy this position without violating covalent geometry and van der Waals contact restrictions. Next, the structure and fine position of the ion were refined using a simulated annealing routine in XPLOR-NIH.³⁸ Agreement between the experimental paramagnetic restraints and values calculated using the final set of structures is shown in Supporting Information Figure S2. In the lowest-energy structures, shown in Figure 3, the LBT is oriented perpendicular to the axis of the three helices in the direction of the surface cleft formed between helices 1 and 2.

Figure 3.

Models of Z-L2LBT calculated with restraints from paramagnetic ion perturbations. Overlay of the backbone is shown for the ensemble of 10 models (A, and as rotated 90°, B). A ribbon diagram of the lowest energy model shows the structural elements (C and as rotated 90°, D). (E) A model of the (Z-L2LBT:Ln³⁺)₂–Fc complex. This figure was generated using PyMol.

Calculation of PRE values was based on an assumed Curie mechanism in which only auto correlation terms from nuclear spin—dipolar shielding anisotropy (DSA) interactions were considered. It has been reported that chemical shift anisotropy (CSA)—DSA cross correlation can contribute to observed paramagnetic lanthanide PRE rates, and unlike other nuclear contributions, these effects cannot be canceled by subtracting the relaxation rates for the protein bound to a diamagnetic ion.³⁹ This becomes an issue particularly at large metal—nucleus internuclear vector correlation times (>20 ns) or for small PRE values (<15–20 s⁻¹). These effects were neglected here due to the low Z-L2LBT correlation time (∼6 ns), and relatively large PREs used in our calculations (see Supporting Information). Maximum errors in the 13 Å distance computed from a PRE of 50 s⁻¹ for an idealized Dy³⁺ complex without internal motions and with a 5 ns correlation time at 14T are approximately −1 and +2 Å at the limiting angles of relative CSA and DSA tensor orientations.³⁸

Z-L2LBT dynamics

Independent analysis of RDCs for the LBT and Z-domain suggest Z-L2LBT exhibits substantial internal motional dynamics. Fitting RDCs for the two moieties independently produced different principle alignment parameters indicating motion of the LBT relative to the Z-domain. Thus, RDC data were not used to refine the final Z-L2LBT structure. The presence of dynamic differences in the Z and LBT portions was further probed by measuring R₁ and R₂ relaxation rates for ¹⁵N nuclei as shown in Figure 4. Resonances from the LBT are characterized by slightly greater R₁ rates and a mixture of R₂ rates that are larger and smaller than those of the helices. This result is consistent with an LBT that experiences dynamic motions that are in general faster than the global Lu³⁺:Z-L2LBT tumbling correlation time of ∼6 ns, with a few residues experiencing R₂ relaxation enhancement by conformational exchange on a slower timescale. The number of measures, particularly for the LBT, are not adequate to provide an accurate amplitude of interdomain motion, however, principal alignment parameters from RDC data (Da = −5.74 × 10⁻⁴, 2.17 × 10⁻⁴ and R = 0.60, 0.19 for Z-domain helices and the LBT, respectively) would suggest a motional amplitude of approximately 40° if a simple diffusion-on-a-cone model pertained. PCSs, due in part to the r⁻³ distance-dependency, and PREs, are less sensitive to interdomain conformational heterogeneity than RDCs, particularly those for sites at longer distances. The good, but imperfect, fits of experimental and back-calculated PRE and PCS data (Supporting Information Fig. S2) are explained by a combination of LBT mobility and cross-correlation effects. However, the fits are more than adequate to support the use of paramagnetic-derived restraints, particularly PCSs, in building structural models such as the ones presented here.

Figure 4.

The ¹⁵N relaxation rates R₁ (A) and R₂ (B) of Z-L2LBT show the Z domain and LBT regions have different dynamic characteristics.

Assessment of Fc N-glycan conformational distribution

The spatial orientations of the N-glycan termini relative to the Fc polypeptide were probed using the distance-dependent line-broadening properties of lanthanide ions. A spectrum of Fc bound by two Z-L2LBT molecules, each coordinating a Gd³⁺ ion, showed considerable broadening of ¹³C-HSQC crosspeaks from the N-glycan's terminal galactose (Gal) residue when compared to a spectrum without a lanthanide ion (Fig. 5). A simple analysis indicated the degree of broadening (∼60 Hz) due to the lanthanide is inconsistent with that expected (1–3 Hz) based upon internuclear Gal–lanthanide distances (30–35 Å) in models based on the placement of Z-domain helices and glycans in crystallography studies⁴⁰ [Fig. 3(E)]. This supports the suggestion that in solution the glycan termini deviate from the position observed in solid-state structures of the Fc determined by X-ray crystallography¹¹ and make excursions that bring the Gal residues closer to the lanthanide ion. More complete NMR studies combined with long molecular dynamics simulations may allow a description of the conformational ensemble of exposed glycan structures and the locations of the Gal residues.

Figure 5.

Fc N-glycan resonances are perturbed by lanthanides bound to Z-L2LBT. (A) A portion of the ¹³C-HSQC spectrum at 14.1 T shows Gal resonances of Fc with a Gal-terminated N-glycan. (B) A slice through the C2-H2 crosspeak in (A) shows the H2 lineshape. (C) A portion of the ¹³C HSQC spectrum showing Gal resonances of (Gd³⁺:Z-L2LBT)₂:(¹³C_U-Gal Fc) complex; (D) a slice through the C2-H2 crosspeak in (C). The intense peak at 3.5 ppm in (C) is folded from another part of the spectrum and does not represent a new peak in the viewing window.

Magnetic resonance imaging of IgG:Z-L2LBT complexes

Z-L2LBT has a significant potential as an imaging probe as well as an NMR probe. IgG antibodies prepared against specific antigens recognize target epitopes with incredible specificity through antigen-binding domains, and Z-L2LBT bound to the Fc of these antibodies is capable of specifically illuminating a tissue or other structures of interest by concentrating Gd³⁺ ions for magnetic resonance imaging. The invariant nature of the Fc fragment in these antibodies allows Z-L2LBT to bind a large variety of IgG1 and IgG2 monoclonal antibodies.²⁷

As a surrogate for a tissue sample, a 3 mm phantom was created by sandwiching a layer of IgG2a-conjugated Sepharose® beads between two layers of unconjugated beads in an NMR tube as shown in Figure 6. The paramagnetic properties of lanthanide ions, particularly Gd³⁺, enhance the contrast in magnetic resonance images.⁷ The IgG2a beads concentrated Gd³⁺:Z-L2LBT from the bulk solution of dilute Gd³⁺:Z-L2LBT and were easily detected in a 1D ¹H image of the phantom (Fig. 6). While this illustration employed just a 1D projection of an image, the possibilities of extension to three-dimensional imaging are clear.

Figure 6.

A gradient profile at 14.1 T of IgG2a-coated sepharose® beads sandwiched between two layers of unconjugated Sepharose® beads. This solution contains Gd³⁺:Z-L2LBT. Reduced ¹H T₁ times resulting from the concentration of Gd³⁺ by the IgG2a-coated beads give rise to greater intensity in the gradient profile which corresponds to the location of IgG2a-coated beads in a 3 mm NMR tube. A photograph of the phantom is shown.

Luminescence microscopy

The presence of Z-L2LBT:IgG may also be detected with luminescence microscopy when it carries a Tb³⁺ ion. IgG2a on the surface of a Sepharose® bead was detected by observing Tb³⁺ luminescence with a standard laser-scanning confocal microscope as shown in Figure 7. In the absence of a laser suitable for excitation at the 1-photon absorption band, a two-photon pathway was employed with irradiation at 560 nm. This reduces efficiency, but points to utility in combination with the longer wavelength irradiation typically used in tissue samples. The full advantage of background suppression using time-resolved methods was not explored.

Figure 7.

Z-L2LBT bound to Tb³⁺ is a reagent for luminescence microscopy. Images of control Sepharose® beads (A) or mouse IgG2a-coated Sepharose® beads were obtained in the presence of Tb³⁺:Z-L2LBT. The beads are 45 to 165 μm Sepharose® 4B resin (Sigma). The average signal above background for ∼50 beads from each trial is shown ± standard error (C).

Discussion

Z-L2LBT, even as currently designed, is well suited for many types of structural studies employing Fc-containing and Fc-binding molecules. The internal mobility of the current lanthanide-binding motif is only a minor inconvenience and can likely be ignored in studies in which constraints are long range. However, an improved design with reduced LBT motion would be useful both in terms of simplifying analysis and generating an alternative lanthanide ion position. We anticipate that the paramagnetic properties of improved lanthanide ion coordinating motifs will be crucial to defining the ensemble of dynamic glycan structures and providing information as to the location of the glycan termini in the context of the Fc glycoprotein.¹¹ Furthermore, effects of CSA × DSA cross-correlation on PRE measurements are also expected to be negligible for application to glycans. The CSA of aliphatic carbohydrate nuclei are expected to be much smaller (¹H ∼1–2 ppm; ¹³C ∼20–40 ppm)^41,42 than the protein amides (¹H ∼10 ppm; ¹⁵N ∼140–170 ppm),⁴³ and thus would minimally impact PRE measurements. These data will be important to defining the mechanism by which the Fc N-glycan modulates immune system activation.⁴⁴

Applying Z-L2LBT as an in vivo MRI contrast reagent will likely require other types of optimization, including redesign to permit greater water access to the ion. Gd³⁺-based reagents also face scrutiny due to toxicity in certain patients with reduced renal function that causes slow clearance and accumulation of released ion.⁴⁵ This limitation would be mitigated by optimizing the LBT sequence for enhanced Gd³⁺ affinity, which could be accomplished either by directed mutagenesis based on a high-resolution crystal structure of Z-L2LBT or random mutagenesis of the ion binding motif. Similar optimization of the Z-domain could in principle introduce specificity for immunoglobulin subtypes, or even introduce affinities for unrelated molecules. It is noteworthy that the Z-domain has been the structural basis for designed affinity reagents such as affibodies.^46,47

Using Z-L2LBT as an optical imaging probe will likely benefit from similar modifications; however, LBT modifications and instrumentation for optimal detection of the probe will also be needed to enhance sensitivity. This version of the LBT removed a Trp residue that presumably would increase the excitation efficiency. The microscopy described here is suboptimal due to the available excitation wavelength (559 nm); the optimal excitation wavelength for this complex would be ∼280 nm [Fig. 1(B)]. The long wavelength excitation achieved here may reflect two-photon excitation⁶ of Tb³⁺ which is conventionally achieved with high-intensity pulsed laser excitation of fluorophores with longer wavelength absorption profiles (one-photon 500–600 nm; two-photon 1000–1200 nm).⁴⁸ Despite the limitation, clear enhancement of the Tb³⁺ signal could be attributed to the IgG2a bead surface and future signal enhancement will come by exciting at a shorter wavelength (280 nm) or using a pulsed, high-intensity laser tunable to 560 nm. Pulsed laser operation with delayed detection of luminescence would be of considerable value in suppressing background fluorescence in biological samples, and may even allow access to time course measurements on the ms timescale.

Conclusion

The design of Z-L2LBT combined high affinity for IgG Fc and lanthanide ions into a single polypeptide that can be efficiently expressed in E. coli cells. This chimera opens new avenues of structural and molecular biology research that utilize lanthanide properties, including paramagnetic relaxation enhancement, pseudocontact shifts of resonance frequencies and luminescence. Applications are broadened by the occurrence of Fc domains in the wide variety of monoclonal antibodies available for biomedical research.

Materials and Methods

Materials

All materials, unless otherwise noted, were obtained from Sigma-Aldrich. Stable isotope-enriched materials were purchased from Cambridge Isotopes. Human serum IgG Fc was from Athens Research & Technology. IgG Fc with ¹³C-Gal terminated N-glycans was prepared as previously described.¹¹

Cloning, expression, and purification of Z-L2LBT

Constructs of the Z-domain were synthesized (Genscript) that contained the LBT at the C-terminus or between helices 2 and 3 (Z-L2LBT). Pseudocontact shifts were not observed when using the C-terminal design and this construct was not pursued further. These constructs were also designed to have an N-terminal 6-residue poly His tag.

The Z-L2LBT coding region was cloned into a pET29 expression vector (Novagen) and transformed into E. coli BL21(DE3) cells for expression. Protein was expressed in M9 minimal medium that included either ¹⁵N-enriched ammonium chloride or a combination of ¹⁵N-enriched ammonium chloride, ¹³C-enriched glucose and >95% ²H₂O. These cultures were allowed to grow to an OD₆₀₀ = 0.5 at which point expression was induced with 0.5 mM IPTG. The cells were then incubated at 18°C with shaking for 15 to 18 h. After harvesting, cells were lysed using the French press method. Protein was purified using a Ni²⁺-NTA resin according to the instructions of the manufacturer (Qiagen). Protein was treated with a 10× molar concentration of EDTA and dialyzed against a buffer containing 25 mM MOPS, 100 mM KCl, pH 7.2 to remove imidazole and EDTA. All steps of expression and purification were monitored by SDS PAGE (see Supporting Information Fig. S1).

Determination of the Tb³⁺ dissociation constant

Titration of 1 μM Z-L2LBT with Tb³⁺ was monitored using a Shimadzu RF-5301PC spectrofluorophotometer by exciting at 280 nm and reading the intensity change at 490 and 545 nm. The intensity data were fitted to an equation describing a binding isotherm:

where F_B is the fraction bound, [P] is the protein concentration, [Tb³⁺] is the terbium ion concentration, and K_D is the dissociation constant.

NMR spectroscopy

NMR spectroscopy was performed on Varian instruments with a VNMRS (21.1 T and 14.1 T) or INOVA (a different 14.1 T) console and equipped with a 3 mm (14.1 T) or 5 mm (21.1 T and 14.1 T) cryogenically cooled probe. NMR samples were between 100 and 900 μM and in a buffer containing 25 mM MOPS, 100 mM KCl, and 10% ²H₂O, pH 7.2. Two-dimensional ¹⁵N-HSQC experiments for assessing paramagnetic resonance frequency shifts and measuring ¹⁵N relaxation rates were from the Varian BioPack distribution. Relaxation rates (¹H) were measured at 21.1 T by placing a simple Carr-Purcell R₂ relaxation element⁴⁹ with a single protein refocusing pulse at the beginning of a standard ¹⁵N-HSQC pulse sequence. The effect of field induced RDCs and subsequent modulation of three-bond proton couplings were neglected due to the weak observed alignment of a Dy³⁺-bound form. Three-dimensional HNCO, HNCOCA, HNCACB, and HNCOCACB backbone assignment experiments were collected in the presence of a 1.1 molar excess of Lu³⁺. Spectra were processed using NMRPipe.⁵⁰ Backbone resonances were assigned using the tools in the NMRViewJ software package.⁵¹ Residues 1–12 were not observed in these experiments, but 98.6% (HN), 96.1% (N), 100% (Cα), 100% (Cβ), and 94.7% (C) of the resonances from residues 13–88 were assigned and are deposited in the BMRB (accession 18,126). Crosspeaks from ¹⁵N-HSQC spectra collected in the presence of Dy³⁺, Yb³⁺, or Ce³⁺ were assigned by inspection using the assigned shifts from the Lu³⁺ complex according to the method of Otting.¹

Lu³⁺:Z-L2LBT was partially aligned in 4.2% (v/v) C12E5 alkyl-polyethylene glycol (PEG) bicelle using previously published protocols.⁵² Specifically, 16% (v/v) C12E5 PEG stock was prepared by mixing 50 μL of PEG, 50 μL of ²H₂O, 200 μL of protein buffer (25 mM MOPS, 100 mM KCl, 10% ²H₂O, pH 7.2), and 16 μL of hexanol. The alignment medium was then prepared by mixing 55 μL of 16% PEG stock with 145 μL of protein solution and 20 μL of D2O to reach 4.2%. RDCs of ¹H-¹⁵N amide pairs were measured using a J-modulated experiment.⁵³ N-C(O) residual dipolar couplings were measured according to Liu and Prestegard.⁵⁴

In preparation of a phantom for magnetic resonance imaging, Sepharose® 4B (Sigma #4B-200) beads and Sepharose® 4B [anti-(human placental) alkaline-phosphatase specific monoclonal mouse IgG2a]-conjugated beads (∼12 μL; Sigma #A2080) were washed separately and thoroughly in a buffer containing 25 mM MOPS, 100 mM KCl, pH 7.2. The beads placed in 3 mm tube were resuspended in a buffer containing 15 μM Z-L2LBT, 17 μM Gd³⁺, 25 mM MOPS, 100 mM KCl, pH 7.2. A T₁-weighted, one-dimensional ¹H magnetic resonance image was obtained using the “gmapz” gradient profile experiment supplied with Varian instruments.⁵⁵ The recycling delay was set to 0.5 s with a gradient strength and duration of 9 G/cm and 3.2 ms, respectively.

Structure calculation

Backbone dihedral angles were predicted using TALOS based on the assigned chemical shifts of the HN, CA, CB, CO, and N nuclei.⁵⁶ Starting Da and R values for the partially aligning PEG medium were calculated from principle order parameters determined in PALES.⁵⁷ Starting Da and R values as well as x, y, z coordinates for the ion were calculated using NUMBAT.⁵⁸ Enhanced relaxation due to the presence of a paramagnetic metal ion (PRE) was determined by subtracting amide ¹H R₂ values measured with Lu³⁺:L2LBT from R₂ values measured with Z-L2LBT in a complex with each of the paramagnetic ions. An initial PRE scaling factor was found by assuming a τ_c value and optimizing the fit during the structure calculation. The structural refinement was performed using PRE rate,⁵⁹ dihedral angle, and RDC orientation constraint (for PCS refinement) modules available in XPLOR-NIH.³⁸ The structure of residues 19 to 31, 37 to 48, and 73 to 86 were fixed based upon their location in the pdb structure (1q2n); likewise the structure of the lanthanide and coordinating residues 54 to 64 was held constant based upon (pdb)1tjb structure. The top 10 structures with the lowest total energy of the 50 calculated structures were selected for final structure deposition. The results of these refinements are summarized in Table I. The quality of the structural models was verified using the Protein Structure Validation Suite (PSVS) software (available at: http://psvs.nesg.org).³¹ Structural models of the Z-L2LBT were deposited into the PDB (2LR2). Models of the Z-L2LBT–IgG Fc complex were obtained by fitting the structure of Z-L2LBT to a crystal structure of a complex between IgG Fc and the first two helices of the Z-domain (1L6X). The observed mobility of the LBT likely permits close contact of the Z-domain to the Fc without greatly perturbing affinity.

Luminescence microscopy

Sepharose® 4B (∼10 μL) beads and IgG2a-sepharose® 4B beads (∼10 μL) were washed separately and thoroughly in a buffer containing 25 mM MOPS, 100 mM KCl, pH 7.2. Beads are between 45 and 165 μm in diameter. Beads were then resuspended in 40 μL of a solution containing ∼300 μM Z-L2LBT, 330 μM Tb³⁺, 25 mM MOPS, 100 mM KCl, pH 7.2, and placed on a slide. Bead fields were imaged with a 40× (N.A.1.3) oil-immersion objective on an Olympus FV-1000 laser scanning confocal microscope. Multiple images were acquired at 0.5 μM intervals through the z-plane. Fluorescence was excited with a 559 nM laser and emission spectra detected with a 575 to 620 nM filter. Images of control and experimental samples were processed identically using ImageJ⁶⁰ and pseudocolored with Adobe Photoshop. Images represent maximum intensity projections.

Glossary

Abbreviations

CSA

chemical shift anisotropy

DSA

dipolar shielding anisotropy

EPR

electron paramagnetic resonance

Fc

fragment crystallizable

Gal

galactose

HSQC

heteronuclear single quantum coherence

IgG

immunoglobulin G

LBT

lanthanide binding tag

MRI

magnetic resonance imaging

NMR

nuclear magnetic resonance

PCS

pseudocontact shift

PRE

paramagnetic relaxation enhancement

RDC

residual dipolar coupling

Supplementary material

Additional Supporting Information may be found in the online version of this article.