Introduction

Metal ions exhibit important properties that assist characterization of macromolecular structure and motion. The high electron density of many metals provides unambiguous signals in X-ray diffraction data that can be used to determine phases (Silvaggi et al. 2007). Paramagnetic metals can be observed directly, as contrast reagent, or reveal a wealth of information regarding the structural and dynamic features of macromolecules using electron paramagnetic resonance, magnetic resonance imaging or nuclear magnetic resonance (NMR) spectroscopy (Hiruma et al. 2013; Iwahara and Clore 2006; Pilla et al. 2015; Shulman et al. 1969; Skinner et al. 2015; Tolman et al. 1995). Long distance restraints (5–50 Å) from measurements of paramagnetic effects are invaluable to refining molecular models built from short-range nuclear Overhauser effects (NOEs, <6 Å) and complementary solution NMR spectroscopy data (Bertini et al. 2002).

Paramagnetic metals can occur naturally in a macromolecule (Butler 1998; Sjodt et al. 2015) or be introduced by replacing an existing metal (Keniry et al. 2006), covalently or non-covalently added as a chelate (Huang et al. 2013; Lee et al. 2015; Loh et al. 2015; Wei et al. 2013; Yang et al. 2015), bound by an unnatural amino acid (Loh et al. 2013; Park et al. 2015), or lanthanide-binding peptide sequence tag (LBT) engineered into the protein sequence (Barthelmes et al. 2011; Feeney et al. 2001; Gaponenko et al. 2000; Martin and Imperiali 2015; Nitz et al. 2003; Su et al. 2008), to reference a few key examples. One common limitation to artificial tags is the presence of metal ion motion. Modeling motion as an additional variable in structure calculations has proven effective with paramagnetic probes including the non-metallic nitroxide spin label and EDTA-Mn(II) complexes (Iwahara et al. 2004a, b). For incorporating metals, polypeptide-based tags have the advantage of ease-of-use by simple encoding into a protein expression vector but predicting tag motion for structure calculations is challenging due to an increase in tag size (~1800 Da) and number of rotatable bonds when compared to the nitroxide labels (~170 Da). Engineering proteins to attach LBTs within the protein rather than at the N or C termini has proven successful, either through incorporation into a loop connecting two defined secondary structural elements or through secondary attachment to a cysteine residue (Barb et al. 2012; Barthelmes et al. 2011; Saio et al. 2009). Such alterations likely will reduce, but not eliminate, peptide-based tag motion. We sought to improve existing LBT sequences by eliminating unnecessary amino acids that contribute to neither conformational stability nor lanthanide binding.

Paramagnetism from unpaired electrons introduces a strong local magnetic field that increases relaxation rates of nearby nuclei with an r −6 distance dependence (paramagnetic relaxation enhancement, or PRE) or perturbs resonance frequencies with an orientation and r −3 distance dependence (pseudocontact shifts, or PCSs) in NMR spectra (reviewed in Koehler and Meiler 2011; Otting 2008). One advantage of these phenomena is that they are measurable using only 1d or 2d NMR experiments.

Paramagnetic lanthanide ions are becoming increasingly utilized as structural probes because of their limited chemical reactivity, bioorthogonal nature, similar coordination chemistry across the group, and strong magnetic properties. The magnetic susceptibility (χ)-tensor for most paramagnetic lanthanide ions is characterized by significant anisotropy that leads to partial alignment of lanthanide chelates in a magnetic field and, as an independent phenomenon, PCSs of resonances (Otting 2008). Anisotropy is greatest for the Tb3+, Dy3+ and Tm3+ ions. Though paramagnetic, Gd3+ has an isotropic χ-tensor and does not cause alignment or PCSs. The diamagnetic lanthanide ions La3+ and Lu3+ are useful in control experiments because they neither align nor cause PCSs and PREs.

Partial alignment of a paramagnetic molecule reintroduces residual dipolar couplings (RDCs) that can be used as a powerful mechanism to measure molecule alignment. RDCs are not normally observed in solution NMR experiments due to rapid isotropic tumbling that causes dipolar couplings to average to zero. The coupling strength is related to the distance between two nuclei and their average orientation with respect to the static magnetic field, thus it is possible to measure the angles between different internuclear vectors in a molecule (Prestegard et al. 2004). The strength of molecular alignment is reflected in the strength of residual dipolar couplings (RDCs) (Koehler and Meiler 2011; Otting 2008 and references therein). Larger RDC values and a larger principle alignment tensor (Da) of the alignment indicate a greater degree of alignment. Furthermore, RDCs contribute high-resolution information relating relative domain orientations of multi-domain macromolecules.

Paramagnetic probes are particularly applicable to structural investigations of large systems by solution NMR spectroscopy where molecular mass combined with limited labeling options or conformational exchange favors sparse amino acid labeling and prohibits many modern 3d and 4d experiments to measure long-range through-space and through-bond nuclear interactions (Otting 2008; Yagi et al. 2013). One such system of interest is the Fragment crystallizable (Fc) region of immunoglobulin G1 (IgG1). IgG1 Fc is a homodimer with a single N-glycan at Asn297 that is critical to Fc γ receptor mediated immune cell activation (Hanson and Barb 2015) and glycan composition is strongly correlated with rheumatoid arthritis disease (Parekh et al. 1985). As a result of the post-translational modification, active IgG1 Fc can only be produced in mammalian cells that do not permit deuterium labeling during expression. To probe the IgG1 Fc structure in solution, a chimeric Fc and lanthanide-binding protein was designed based upon three helices of the Fc-binding Z-domain (Nilsson et al. 1987; Tashiro et al. 1997) and a lanthanide-binding tag (LBT) (Nitz et al. 2003; Su et al. 2006), termed Z-l2LBT (Barb et al. 2012). Despite replacing loop 2, connecting helices 2 and 3 of the Z-domain, with the LBT, considerable tag motion was observed (Fig. 1).

Fig. 1
figure 1

Features of chimeric IgG and lanthanide-binding Z domain designs. a A green ribbon diagram depicts the Z domain with a lanthanide binding tag in loop 2 (Z-l2LBT) (Barb et al. 2012). b Z-l2LBT binds to the same surface of IgG1 Fc as the parent Z-domain. c Variant A and B differ from the parent Z-l2LBT protein by removal of 5 and 2 residues from the LBT/Z linker, respectively. Arrows indicate LBT residues that directly interact with Ln3+ ions. Underlining of the variant A sequence denotes assigned amide resonances

Full size image

Here we present the improvement of one polypeptide-based LBT to reduce motion and preserve affinity towards lanthanide ions. Though LBT sequences are applicable to numerous types of techniques, we utilized solution NMR spectroscopy to provide a high-resolution and quantifiable characterization of protein structure and motion. Two variants with reduced loops linking the Z domain to the LBT were prepared and characterized. We also investigated the utility of one design to probe the structure of a large 72.2 kDa complex with IgG1 Fc.

Experimental conditions

Materials

All materials, unless otherwise noted, were from Sigma-Aldrich. Stable isotope-enriched compounds were purchased from Cambridge Isotopes.

Protein expression and characterization

Z-l2LBT expression using E. coli BL21* cells was described previously (Barb et al. 2012). Open reading frames for variants A and B were synthesized (Genscript) and cloned into pET29 as described for the Z-l2LBT construct (Barb et al. 2012). Expression, stable isotope labeling, purification, and Tb3+ binding measurements of variants A and B were performed as previously described (Barb et al. 2012). A modified construct of variant A was prepared that permitted tobacco etch virus protease (TEV)-catalyzed cleavage of the His tag (liberating G13V14D15…K83) for studies of the Fc:A complex. The TEV reaction was performed at 25 °C for 16 h in the dark in a buffer containing 50 mM trisaminomethane, 100 mM sodium chloride, 0.5 mM ethylenediaminotetraacetic acid and 2 mM beta-mercaptoethanol, pH 8.2. The cleaved variant A product was isolated by passing the reaction mixture over a Ni–NTA column (Qiagen) and collecting the flow-through fraction. Cleaved variant A (1 mL) was then subject to dialysis for 4 h against 1 l of 25 mM 3-(N-morpholino)propanesulfonic acid (MOPS), 100 mM potassium chloride at 25 °C using a 3000 molecular weight-cutoff dialysis tubing (Spectrum Labs). The dialysis was repeated against a fresh 1 l of the same MOPS buffer.

Expression and purification of the human IgG1 Fc was conducted using HEK293F cells as described (Subedi et al. 2015).

NMR spectroscopy

NMR spectrometers operating at 21.1 T, 14.1 T (both Varian VNMRS) or 18.8 T, 16.4 T (Bruker Avance 3, Avance 2, respectively) were equipped with cryogenically-cooled 5 mm probes. 1H resonance frequencies were internally referenced to DSS; 13C and 15N frequencies were indirectly referenced using the spectrometer’s 1H frequency at 0 ppm (Markley et al. 1998). NMR data were analyzed using Topspin (Bruker; v.2.1), NMRPipe (Delaglio et al. 1995), NMRViewJ (Johnson and Blevins 1994) and Sparky (version 3.115; Goddard and Kneller, University of California, San Francisco).

Assignment of 1H–15N correlation peaks collected using the variant A:Lu3+ chelate was completed as previously described (Barb et al. 2012) using triple resonance HNCO, HNCA, HNCACB and CBCACONH experiments collected at 14.1T and 10 °C (Cavanagh 2007). Assignments for variant A were deposited in the BioMagResBank (BMRB) as accession number 19769. Assignment of residues from the paramagnetic complexes (+Dy3+ or + Tb3+) were made by comparison to a spectrum of the corresponding diamagnetic complex (+Lu3+) and using available backbone resonance assignment data of diamagnetic proteins where applicable [Z-l2LBT (BMRB 18126); variant A (19769); IgG1 Fc (25224)].

NMR spectra to probe the binding of variant A to IgG1 Fc were collected at 16.4 T and 50 °C. 1H–15N heteronuclear single quantum coherence (HSQC) spectra of isotopically enriched IgG1-Fc (150 μM: dimer) with selective [15N]-Tyr amino acid residues were collected with and without variant A (330 μM) plus Lu3+ (330 μM). NMR samples were prepared in 25 mM MOPS, 100 mM sodium chloride, 0.5 mM DSS and 5 % D2O, pH 7.2.

Measurements of A and B alignments were made with samples containing 150 μM protein with one molar equivalent of either Lu3+ or Dy 3+ at 25 °C and 21.1 T. NMR spectra to record pseudo-contact shifts (PCS) and residual dipolar coupling measurements (RDCs) on the IgG1 Fc:(variant A:Ln3+)2 complex were collected at 18.8 T and 50 °C using 125 μM IgG1 Fc (dimer), 275 μM TEV-cleaved variant A, 290 μM Tb3+(or Lu3+) in a buffer containing 10 mM MOPS, 100 mM KCl, 0.5 mM DSS and 5 % D2O, pH 7.2. RDCs were measured with a J-modulation based strategy as described (Barb et al. 2012; Liu and Prestegard 2009) by subtracting 1H–15N couplings measured with diamagnetic samples (containing the diamagnetic Lu3+ ion) from couplings measured with paramagnetic samples (containing the Tb3+) and analyzed as described using PALES (Zweckstetter and Bax 2000). PCS were analyzed using NUMBAT (Schmitz et al. 2008).

Structure figures were prepared using PyMol (The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC).

Results and discussion

Lanthanide-binding properties of truncated Z-l2LBT variants

Z-l2LBT was targeted for optimization to reduce LBT motion relative to the three Z-domain helices. We designed two variants by eliminating amino acid residues in the two segments connecting the LBT and Z moieties that were believed to facilitate LBT mobility (Barb et al. 2012). Variant A (5 residues removed) and variant B (2 residues removed) expressed as soluble proteins at high yields using E. coli (>100 mg protein/L culture) and were purified (Fig. 1).

Variants A and B bound Tb3+ with high affinity in a spectrofluorophotometric assay that excites only polypeptide-bound ions at 280 nm through a Tyr residue in the LBT moiety. Titrating variant A or B with Tb3+ led to saturation of luminescence at 490 and 540 nm and is shown in Fig. 2. Variants A (700 ± 50 nM) and B (130 ± 50 nM) bound with dissociation constants (K d) that are comparable to that of the original Z-l2LBT protein [130 ± 30 nM (Barb et al. 2012)]. As a result of the lanthanide ion binding affinity, either variant will be saturated at a ratio of 1:1 in NMR experiments that typically use >50 μM protein.

Fig. 2
figure 2

Tb3+-binding affinity of variants A and B

Full size image

Domain motion of the lanthanide-binding Z-l2LBT variants

We estimated field-induced alignments of the Z domain moiety for the A:Dy3+ and B:Dy3+ complexes by measuring RDCs. If the Dy3+:LBT moieties of each variant are assumed to align in an identical manner, a reasonable assumption due to complete conservation of the lanthanide-binding residues, principle alignment tensors (Da) fitted by analyzing RDCs of the underlying Z domains may be compared between variants as a relative measurement of Z domain alignment and thus interdomain LBT/Z mobility. This approach does not require the underlying Z-domains to align in the same orientation and was utilized here because the LBT residues were broadened beyond detection due to the paramagnetic ion, eliminating the possibility of measuring the amplitude of interdomain motion directly using RDCs.

HSQC spectra of 15N-labeled variant A, variant B and Z-l2LBT with a 1 molar equivalent of Dy3+ were compared to spectra of diamagnetic Lu3+ complexes as shown in Fig. 3. The spectrum of variant A was affected substantially by PREs in the Dy3+ complex with only 14 amide peaks visible. A spectrum of variant B + Dy3+, like the parent Z-l2LBT + Dy3+ complex, contained more than twice as many peaks as A. All three spectra showed the effect of PCSs (>0.5 ppm) (the Z-l2LBT spectrum is shown in Barb et al. 2012).

Fig. 3
figure 3

RDCs of the designed proteins. a The distribution of one-bond 1H–15N coupling values for variants A, B, and Z-l2LBT was measured using unassigned peaks in the presence of Dy3+ at 21.1 T. b, c 15N HSQC spectra of variants A and B, respectively, in complex with Dy3+. Pseudo-contact shifts (PCSs) of assigned resonances are shown with a grey line with the origin corresponding to the position in the spectrum collected in the presence of Lu3+. Errors bars represent the residue-specific sum of errors for individual fits of the Lu3+ and Dy3+ 15N–H coupling observations. d, e Plots showing the observed RDC values of the A:Dy3+ and B:Dy3+ complexes versus the calculated values. The line represents y = x

Full size image

A comparison of the one-bond scalar 15N–H coupling (1 J NH) from the unassigned spectra of the designed proteins in a Dy3+ complex provided insight into the different molecular properties. Variant A displayed the largest range of values (81–114 Hz) while the range of B (85–109 Hz) was less than A but greater than Z-l2LBT (88–103 Hz) as shown in Fig. 3a. The greater range of 1 J NH values and fewer peaks resulting from PREs observed for A is consistent with an LBT that is closer to the three Z domain helices and experiences a reduced amplitude of motion.

Measurements of residual 1H–15N dipolar couplings for variant A, variant B and Z-l2LBT with one molar equivalent of Dy3+ provided an estimate of LBT motion for the three proteins (Fig. 3a). Peaks from A + Dy3+ (11) and B + Dy3+ (14) were assigned with high confidence and used for further analysis as shown in Fig. 3b, c. Alignment tensors (Da) fitted to the RDCs using a model of the Z-domain [pdb 1q2n (Zheng et al. 2004)] were qualitatively consistent with the ranges of 1H–15N coupling values (Table 1; Fig. 3a, d, e). Though only a limited number of assigned residues are visible in spectra collected in the presence of Dy3+, these data do provide a means to estimate the motion of each LBT relative to the Z domain helices. The LBT of A and B experience a range of motion that is estimated to be 0.39 and 0.68 times that of Z-l2LBT, respectively, based on the tensor magnitude. A previous analysis found Z-l2LBT exhibited 40° of interdomain motion using a diffusion-in-a-cone model (Barb et al. 2012; Barbato et al. 1992). By comparison variants A and B exhibited ~16° and 27° of motion, respectively. This analysis, though a simplification that assumes motion is equally distributed within a cone, is informative as a singular estimate of relative mobility between the LBT and Z moieties.

Table 1 RDC alignment parameters for Z-l2LBT, and variants A and B in complex with Dy3+ at 21.1T
Full size table

Structural assessment of Z-l2LBT variants

The structural features of these proteins were probed by solution NMR spectroscopy. Both variants A and B appeared folded and experienced conformational rearrangements upon binding a diamagnetic Lu3+ ion based on two-dimensional 1H–15N HSQC spectra of variants A and B, before and after titrating Lu3+ (Fig. 4). Previous results indicate a significant change in Z-l2LBT conformation upon lanthanide binding is likely limited to the LBT moiety (Barb et al. 2012). Due to the large proportion of the structured protein formed by the LBT (21–23 %) and central location it is anticipated that lanthanide binding will induce noticeable chemical shift changes in a majority of the protein as observed (Fig. 4). The designed proteins exchanged between lanthanide bound and unbound states in the slow regime of the NMR timescale (data not shown).

Fig. 4
figure 4

2d NMR spectroscopy of variants A and B. 15N-HSQC spectra of variants A (a) and B (b) before and after addition of Lu3+ at 25 °C reveal large changes in protein conformation. c Backbone amide crosspeaks of the variant A:Lu3+ complex were assigned with standard triple-resonance experiments collected at 10 °C. Sidechain amides are identified with horizontal grey lines

Full size image

A complete backbone resonance assignment of variant A assigned all major peaks observed in an 1H–15N HSQC spectrum (88 % of HN assigned, 89 % N, 79 % CA, 76 % CB, and 73 % CO from residues 13–83; Fig. 4c) and was consistent with a high degree of structural similarity to the parent Z-l2LBT protein. The high degree of structural similarity is supported by identical functional roles that are dependent upon structure, namely effective Tb3+ binding (discussed above) and antibody binding (discussed below). A few residues from loops connecting LBT to helices 2 & 3 plus a few residues from the middle of helix 3 were not observed (Fig. 1c).

NMR characterization of the A:Fc complex

We probed binding of variant A to IgG1 Fc in order to characterize the effects of two paramagnetic ions in the 72.2 kDa IgG1 Fc:(variant A:Tb3+)2 complex (IgG1 Fc is a homodimer with two Z domain binding sites, see Fig. 1b). To reduce the rotational correlation time of the five-component complex and increase sensitivity as much as possible, a variant A form was prepared with a TEV-cleavable 6xHis tag. This variant was shown to bind Tb3+ with the same high affinity as the variant A construct containing a 6xHis tag and the TEV-cleaved form was utilized for subsequent experiments (data not shown).

A comparison of IgG1 Fc spectra before and after the addition of the variant A:Lu3+ chelate revealed the formation of a tightly bound complex as expected based on affinity of the Z domain for IgG1 Fc (Fig. 5) (Barb et al. 2012; Starovasnik et al. 1997). Two IgG1 Fc Tyr crosspeaks shifted as a result of complex formation; Tyr436 is proximal to the Z/Fc interface and experienced a large displacement (Deisenhofer 1981; DeLano et al. 2000). The Tyr373 crosspeak shifted slightly. The remaining six crosspeaks remained in the same spectral position but were slightly less intense likely due to an increase in the rotational correlation time upon variant A binding.

Fig. 5
figure 5

15N-HSQC spectra of [15N-Tyr]-IgG1 Fc collected at 16.4 T before and after addition of two molar equivalents of variant A and Lu3+. Amino acid assignments are based on a published report (Matsumiya et al. 2007)

Full size image

Paramagnetic effects of variant A in complex with Fc

Strong paramagnetic effects were observed in the IgG1 Fc:(variant A:Tb3+)2 complex. Though significant PREs broadened many signals including K248, K288 and K320 as shown in Fig. 6, we focused on PCSs and field-induced RDCs of the Fc that indicate structural restriction of the LBT. The presence of PREs is not a good measure of structural restriction because these can be strong even if the tag experiences a high degree of mobility relative to the underlying protein.

Fig. 6
figure 6

Pseudo-contact shifts (PCSs) were detected among the [15N-Tyr/Lys]-IgG1 Fc amide resonances at 18.8 T. a An overlay of 15N-HSQC-TROSY spectra collected using the IgG1 Fc(Z-l2LBT:Lu3+)2 and IgG1 Fc(Z-l2LBT:Tb3+)2 complexes are shown in black and orange contours, respectively. Significant differences in resonance frequencies are denoted with white and grey arrows. b The observed 1H PCSs correlate strongly to 1H PCSs calculated by the NUMBAT program (Schmitz et al. 2008), using a model of the Fc protein determined by X-ray crystallography (DeLano et al. 2000). A line corresponding to the function y = x is shown in black

Full size image

Substantial 1H PCSs were observed in an 15N-HSQC-TROSY spectrum of the IgG1 Fc:(variant A:Tb3+)2 complex when compared to the Lu3+-bound complex (Fig. 6). Of 21 assignable cross peaks from the Tb3+-bound complex, 20 showed measureable PCSs between −0.221 and 0.081 ppm indicating the LBT tag was stabilized in a position to influence IgG1 Fc resonances. A small degree of heterogeneity is observed with some peaks, including Y436. It is unclear whether the LBT, lanthanide, or Fc contributes to this heterogeneity. A comparison of all observed PCSs with PCSs calculated using a structural model of the Fc monomer [pdb- 1L6X (DeLano et al. 2000)] revealed a high degree of similarity (Fig. 6b, the coefficient of determination squared (R 2) = 0.978).

That the PCSs fit the known structure with a high degree of similarity is important for three reasons. First, the Tb3+ ion is coordinated near the Fc protein and experiences limited motion, if any, relative to the Fc. Second, the ion is positioned in a manner that preferentially impacts resonances on the nearest monomer, and minimally affects the more distant Fc monomer in the Fc dimer. Third, these data confirm that the 1L6X structural model is a good representation of the Fc conformation in complex with the Z domain and supports recent solution measurements of Fc quaternary structure in solution (Subedi and Barb 2015).

We also pursued RDCs as a measurement of field-induced IgG1 Fc:(variant A:Tb3+)2 alignment. We observed six RDCs with a range of −7.6 to 14.3 Hz and error <2.3 Hz (Table 2; Fig. 7). The smaller number of RDCs with low errors (6) with respect to PCS observations (20) reflects the effect of PREs on signal intensity in the J-modulation experiments that led to increased errors with the Tb3+ sample. These presence of RDCs is further strong evidence for conformational stability of the variant A LBT tag in the IgG1 Fc complex.

Table 2 Field-induced RDCs from IgG1 Fc residues in the IgG1 Fc:(variant A:Tb3+)2 complex
Full size table
Fig. 7
figure 7

Residual dipolar couplings (RDCs) of the Fc amide resonances were identified using the IgG1 Fc(Var. A:Tb3+)2 complex. Spectra collected using isotropic (black line with black circles; complex with Lu3+) and partially aligned (orange line with open circles; complex with Tb3+) conditions at 18.8 T revealed significant positive and negative RDCs using a J-modulation method to measure 1H–15N couplings (Liu and Prestegard 2009)

Full size image

Conclusions

Here we detailed the modification of a chimeric lanthanide and IgG1 binding protein to dramatically reduce the size and mobility of the metal binding moiety. This redesigned LBT tag is widely applicable to other structure-based investigations and has been utilized to study the two-domain SpA-N protein (Prof. T. Oas, Duke University), a two-domain roundabout (ROBO) family construct and a fragment of the extracellular domains of a leukocyte common antigen-related receptor protein tyrosine phosphatase (Prof. K. Moremen, UGA). We demonstrated the utility of one new design to investigate the structure of a large 72.2 kDa complex. These shortened LBTs provide one approach to reducing the motion of LBTs in many other protein systems that require limited LBT motion.