Multiple alignment of membrane proteins for measuring residual dipolar couplings using lanthanide ions bound to a small metal chelator

Abstract

NMR structure determination of helical integral membrane proteins is one of the most challenging undertakings in modern structural biology. The solubilizing detergent micelle or phospholipid bicelle adds significantly to the overall size of the system, often requiring perdeuteration to obtain useable spectra. However, perdeuteration prevents structure determination using traditional NOE analysis. Residual dipolar coupling constraints are an attractive complement to NOE distance constraints, but alignment methods are limited to strained polyacrylamide gels due to the incompatibility of the solubilizing detergent or lipid with alignment media such as bicelles or phage particles. We demonstrate the use of lanthanide ions bound to the protein through a small thiol-linked metal chelator as a robust method for partial alignment of membrane proteins. This method provides multiple alignment orientations depending on the ion bound, and permits RDC measurement of multiple bond vectors. We demonstrate that using this method a large number of RDC's can be measured using 3-dimensional NMR methods where alignment using strained polyacrylamide gels results in fewer peaks due to drastic line-broadening.

While the field of structural biology is advancing rapidly with over 40,000 structures currently in the Protein Data Bank, progress on membrane proteins, which comprise nearly 30% of the human genome, has been much slower; only about 110 of solved structures represent unique membrane proteins. NMR is an attractive alternative to x-ray crystallography for these proteins that are difficult crystallize, and has the added advantage of studying their structure and dynamics in solution.

For membrane protein NMR, the solubilizing detergent micelle or phospholipid bicelle adds significantly to the size of the system, and hence to the rotational correlation time. Thus, the protein behaves as a much larger particle, resulting in weak or missing signals in multidimensional spectra due to rapid R₂ relaxation. Deuteration and TROSY¹ pulse sequences reduce the experimental relaxation rates permitting resonance assignments, but fully deuterated proteins pose a problem for structure determination using traditional NOE analysis. Residual dipolar coupling orientational constraints² have seen extensive use in the determination and refinement of deuterated globular protein structures³. However, there are few reports of RDC's being used to define the structures of any membrane proteins by NMR⁴. Most alignment media – phage, phospholipid bicelles, and organic liquid crystals⁵ – are incompatible with the detergents or lipids used to solubilize these proteins. Strained polyacrylamide gels are inert and provide different alignment orientations by varying the composition of copolymers⁶ or the method of straining⁷. In practice, however, strained gels performed poorly with polytopic helical membrane proteins. With two- and four-helix membrane proteins we have seen poor quality spectra in all tested gels, with weak or missing cross-peaks resulting from the lower achievable protein concentration and restricted tumbling in uncharged gels, and significant interactions between proteins and the matrix in charged gels. Furthermore, neutral acrylamide gels lose alignment over the course of a few days (Figure S1), making the measurement of RDC's by 3D NMR methods impractical. Hence an alternative alignment method was sought that ideally would not reduce the achievable protein concentration, would not significantly increase resonance linewidths, and would yield samples with significant alignment and high quality NMR spectra that were stable for multiple days.

Several reports have described the partial alignment of proteins induced by the binding of lanthanide ions⁸. Chimeric proteins have been constructed by fusing the target protein to a calcium-binding protein or peptide, such as calmodulin⁹ or an EF-hand¹⁰. A similar approach employed small metal-chelating compounds with thiol-reactive groups, linked to a unique cysteine sidechain in the protein¹¹. Several compounds of varying linker lengths have been synthesized and some are commercially available¹². Here we show that the thiol-directed approach overcomes the shortcomings of other alignment media with polytopic helical membrane proteins, providing a robust method for multiple, stable alignments.

Alignment methods were tested using the two transmembrane helix subunit c of the E. coli F₁F_o ATP synthase in micelles. An A79C mutation introduced a unique cysteine at the C-terminus of the protein, which should be the most innocuous location for introducing a chemical modification. Modification was achieved by adding 30 μl of 100 mM N-[S-(2-Pyridylthio) cysteaminyl]EDTA (Toronto Research Chemicals, Inc.) to 500 μl of 600 μM subunit c in 5% LPPG (1-palmitoyl-2-hydroxy-sn-glycero-3-[phospho-RAC-(1-glycerol)], Avanti Polar Lipids, Alabaster, Al), 100 mM Tris pH 8. The reaction proceeded overnight at 42° C. To incorporate metals, it was necessary to preload the N-[S-(2-Pyridylthio)cysteaminyl]EDTA with the lanthanide ion prior to protein modification, as the free ions caused excessive detergent precipitation. The lanthanides were incorporated by adding a three-fold molar excess of LnCl₃ solution to the N-[S-(2-Pyridylthio)cysteaminyl]EDTA sample in the Tris buffer without detergent. After 30 minutes, sufficient EDTA was added to leave a slight (∼5-10%) LnCl₃ excess, resulting in negligible precipitation of the NMR sample.

NMR spectra of high quality were obtained for the modified and metal-bound subunit c (Figures 1, 2, & S2). The spectra of the mutant and EDTA-modified mutant were largely unchanged with respect to the wild-type subunit c spectrum. Upon addition of lanthanide ions such as TbCl₃, significant pseudocontact shifts were observed for residues 66-78, with smaller shifts for residues 61-65. Some perturbation of the resonances for residues 2, 6, 7, and 8, which are located on the N-terminal helix directly opposite Cys79, were also observed.

Figure 1.

Selected upfield and downfield components of a 3D-IPAP HNCO experiment measuring (A) ¹H¹⁵N splittings for Gly18 or (B) ¹³C′¹³C^α splittings for Phe54. The top panel indicates the lanthanide ion used for alignment or no alignment. Tm⁺³ and Yb⁺³ spectra were measured at 800 MHz and Tb was measured at 900 MHz. The observed splitting in Hz is shown.

Figure 2.

(A) ¹H¹⁵N and (B) ¹³C′¹³C^α splittings for Val15 in an unaligned sample, polyacrylamide gel, or Tm³⁺ alignment. All spectra were measured at 800 MHz. The observed splittings in Hz are indicated.

Figure 1 shows selected upfield and downfield ¹H¹⁵N and ¹³C′¹³C^α components from a 3D in-phase anti-phase HNCO experiment, with changes in the observed splittings demonstrating lanthanide-induced alignment of the protein in the magnetic field. The alignment tensor depends on the metal used¹³. Here for example, the Gly18 ¹H¹⁵N RDC was +3.7 Hz with Tm³⁺ and –2.0 Hz with Tb³⁺ (Figure 1A), with the opposite sign clearly demonstrating the different alignments with the different metals. Similarly the ¹³C′-¹³C^α RDC's for Phe54 were –2.2 Hz with Tm³⁺ and 1.4 Hz with Tb³⁺ (Figure 1B). The degree of lanthanide-induced alignment increases with increasing magnetic field strength¹³; the observed range of ¹H¹⁵N RDC's for Yb⁺³ was -6.6 to 5.3 at 800 MHz, and −8.1 to 5.9 at 900 MHz.

Multiple sets of orientational constraints from RDC's will likely be essential for high resolution NMR structures of polytopic helical membrane proteins. Unlike β-barrels, long-range backbone NOE's are vanishingly scarce, and long-range side chain NOE's are difficult to identify unambiguously for these proteins due to extensive chemical shift degeneracy in the side chains. Alignment using strained polyacrylamide gels has yielded useful RDC's for several membrane proteins⁴, but for helical proteins only measurements for ¹H¹⁵N vectors using 2D experiments have been reported. The 3D experiments typically required to fully resolve the ¹H¹⁵N, ¹³C¹⁵N, and ¹³C′¹³C^α cross-peaks appear to be impractical in polyacrylamide gels, due to the reduced sample concentrations and restricted rotation in the gels, as well as the lower sensitivity of the 3D experiments themselves. Additionally, relaxation of uncharged gels reduces the usefulness of any measurement that cannot be completed within one or two days. As shown in the middle panels of Figure 2A and 2B, many of the ¹H¹⁵N and ¹³C′-¹³C^α are simply missing from the 3D data sets. Using lanthanide ions to induce alignment, however, we were able to use 3D experiments to readily measure nearly all of the ¹H¹⁵N and ¹³C′¹³C^α RDC's in subunit c (Figure 2A and B, right panels, and Table S1) with multiple orientations induced by varying the identity of the metal.

Protein alignment via lanthanide metals may be crucial for the application of RDC's to solving the NMR structures of polytopic helical membrane proteins such as subunit c. For the partial alignment of such membrane proteins, lanthanide ions proved to be far superior to acrylamide gels. For proteins that naturally contain a cysteine residue, this method is quite simple and easily yields RDC's for full sets of the ¹H¹⁵N, ¹³C¹⁵N, and ¹³C′¹³C^α vectors, with partial alignment that does not decay over time. Multiple alignments are easily obtained by using different metals or different mutations.

Supplementary Material

1si20061227_02. Supporting Information Available.

¹H¹⁵N spectra of modified and unmodified subunit c in detergent micelles, and tables of RDCs from alignment with Tm³⁺, Tb³⁺, and Yb³⁺. This material is available free of charge at http://pubs.acs.org.