A 3-D Structural Model of Solid Self-Assembled Chlorophyll a/H2O from Multispin Labeling and MAS NMR 2-D Dipolar Correlation Spectroscopy in High Magnetic Field

Abstract

Magic angle spinning (MAS) NMR with Lee–Goldburg cross-polarization (LG-CP) is used to promote long-range heteronuclear transfer of magnetization and to constrain a structural model for uniformly labeled chlorophyll a/H2O. An effective maximum transfer range dmax can be determined experimentally from the detection of a gradually decreasing series of intramolecular correlations with the 13C along the molecular skeleton. To probe intermolecular contacts, dmax can be set to ∼4.2 Å by choosing an LG-CP contact time of 2 ms. Long-range 1H–13C correlations are used in conjunction with carbon and proton aggregation shifts to establish the stacking of the chlorophyll a (Chl a) molecules. First, high-field (14.1 T) 2-D MAS NMR homonuclear (13C–13C) dipolar correlation spectra provide a complete assignment of the carbon chemical shifts. Second, proton chemical shifts are obtained from 1H–13C heteronuclear dipolar correlation spectroscopy in high magnetic field. The shift constraints and long-range 1H–13C intermolecular correlations reveal a 2-D stacking homologous to the molecular arrangement in crystalline solid ethyl-chlorophyllide a. A doubling of a small subset of the carbon resonances, in the 7-methyl region of the molecule, provides evidence for two marginally different well-defined molecular environments. Evidence is found for the presence of neutral structural water molecules forming a hydrogen-bonded network to stabilize Chl a sheets. In line with the microcrystalline order observed for the rings, the long T1's, and absence of conformational shifts for the 13C in the phytyl tails, it is proposed that the Chl a form a rigid 3-D space-filling structure. Probably the only way this can be realized with the sheets is by forming bilayers with interpenetration of elongated tails. Such a 3-D space-filling organization of the aggregated Chl a from MAS NMR would match existing models inferred from electron microscopy and low-resolution X-ray powder diffraction, while a micellar model based on neutron diffraction and antiparallel stacking observed in solution can be discarded.