Structure determination of a Galectin-3-carbohydrate complex using paramagnetism-based NMR constraints

Abstract

The determination of the location and conformation of a natural ligand bound to a protein receptor is often a first step in the rational design of molecules that can modulate receptor function. NMR observables, including NOEs, often provide the basis for these determinations. However, when ligands are carbohydrates, interactions mediated by extensive hydrogen-bonding networks often reduce or eliminate NOEs between ligand and protein protons. In these cases, it is useful to look to other distance- and orientation-dependent observables that can constrain the geometry of ligand-protein complexes. Here we illustrate the use of paramagnetism-based NMR constraints, including pseudo-contact shifts (PCS) and field-induced residual dipolar couplings (RDCs). When a paramagnetic center can be attached to the protein, field-induced RDCs and PCS reflect only bound-state properties of the ligand, even when averages over small fractions of bound states and large fractions of free states are observed. The effects can also be observed over a long range, making it possible to attach a paramagnetic center to a remote part of the protein. The system studied here is a Galectin-3-lactose complex. A lanthanide-binding peptide showing minimal flexibility with respect to the protein was integrated into the C terminus of an expression construct for the Galectin-3-carbohydrate-binding domain. Dysprosium ion, which has a large magnetic susceptibility anisotropy, was complexed to the peptide, making it possible to observe both PCSs and field-induced RDCs for the protein and the ligand. The structure determined from these constraints shows agreement with a crystal structure of a Galectin-3-N-acetyllactosamine complex.

Figure 1.

¹H–¹⁵N HSQC and TROSY overlays for 0.3 mM Galectin-3–LBT at a ¹H frequency of 600 MHz with (A) 0.3 mM Lu³⁺ and (B) 0.3 mM Dy³⁺.

Figure 2.

¹H–¹⁵N HSQC spectra at a ¹H frequency of 600 Hz for 0.3 mM Galectin-3–LBT with 0.3 mM Lu³⁺ (red) or with 0.3 mM Dy³⁺ (black). Many peaks disappear for the protein sample with paramagnetic Dy³⁺. Peak assignments were accomplished by drawing a diagonal line between diamagnetic and paramagnetic peaks as shown in the figure.

Figure 3.

Experimental RDCs and PCSs at a ¹H frequency of 600 Hz vs. back-calculated PCSs and RDCs using the structural coordinates determined by a 2 Å grid search.

Figure 4.

Experimental PCSs and RDCs at a ¹H frequency of 600 MHz vs. back-calculated PCSs and RDCs using structural coordinates determined using XPLOR-NIH.

Figure 5.

RDCs measurement of C2–H2 of galactose and C5–H5 of glucose. RDCs were measured as the splitting difference in the ¹³C dimension between the sample with diamagnetic Lu³⁺ and the sample with paramagnetic Dy³⁺.

Figure 6.

Final ligand–protein complex structure determined by XPLOR-NIH using ligand and protein RDCs, PCSs, and a single ligand–protein NOE.