Probing Biological Molecules: Theory and Experiment

FLOW ORIENTED LINEAR DICHROISM TO PROBE PROTEIN ORIENTATION IN MEMBRANE ENVIRONMENTS

Alison Rodger, Jascindra Rajendra, Rhoderick Mortimer, Terrence Andrews, Jonathan D. Hirst, Andrew T.B. Gilbert, Rachel Marrington, Timothy R. Dafforn, David J. Halsall, Malin Ardhammar, Bengt Nordén, Cheryl A. Woolhead, Colin Robinson, Teresa J.T. Pinheiro, Jurate Kazlauskaite, Mark Seymour, Niuvis Perez, Michael J. Hannon

1 INTRODUCTION

Processes occurring on or in membranes are essential in most biological systems, and the study of these processes has been engendering an increasing interest for a long time, as has the creation of artificial lipid membrane systems. Studies of, for example, membrane transport and membrane protein function call for a thorough knowledge of molecular interactions within the membrane, between the lipids themselves and between lipids and other species (proteins, drugs, and ions). To this end, the locations and orientations of molecules bound to the membrane can give important information. However, to date no simple experimental method has been established to achieve this for membrane bound proteins. In this work we report the first flow linear dichroism (LD) study of proteins bound to liposomes. Flow LD of molecules bound to the bilayer of shear-deformed liposomes is one of the few direct methods potentially available for the study of the orientation of membrane guest molecules, provided that the molecules of interest have significant absorption in the visible and near-UV regions.

Linear dichroism is the difference in absorption of light polarised parallel to an orientation direction and light polarised perpendicular to that direction. The LD signal is related to the oscillator strength of a transition (its absorbance intensity) and the polarisation of the transition relative to the orientation axis. It is thus the ideal technique to use to probe the orientation of an analyte and it is widely used, for example, for determining the orientation of drugs bound to flow oriented DNA. The most effective method for achieving flow orientation has proved to be Couette flow where two concentric cylinders with a small annular gap (usually 500 µm) are aligned and one of them rotates. The light is incident radially on the cell so the stationary cylinder needs to have two windows and the rotating cylinder needs to be transparent to the intended radiation.

Liposomes can be considered models of cell membranes, and be used for studying transport and signal mechanisms of membrane proteins in situ. They are also used for drug delivery and as transfecting agents in gene therapy. Ardhammar, Mikati, Lincoln and Norden have shown that aromatic moieties or the aromatic 'arm' of ruthenium dipyridophenazine can be oriented in liposomes and their orientation detected when the liposomes are subjected to shear flow. In this work we probed the orientation achievable by flow with different size model membranes and then the application of LD to determine the orientation of proteins bound on or in liposomes. The proteins studied include gramicidin, cytochrome c, pre-PsbW (a thylakoid membrane protein precursor) and a monoclonal antibody.

2 LINEAR DICHROISM OF ANALYTES IN SHEAR DISTORTED LIPOSOMES

As noted above there are two literature precedents for flow orienting liposomes to assess the orientation of solutes in their bilayer. Ardhammar et al. found that pyrene and anthracene in soybean liposomes (produced by extrusion) had a negative LD for their long axes, in accord with their expectations that the molecules would be oriented with their long axis parallel to the lipid hydrocarbon chains and thus on average perpendicular to the elongation and orientation axis of the liposome. With perylene, however, both the long axis and the short axis had negative LD signals, which the authors took to mean that the diagonal of the molecule was oriented parallel to the lipid chains. For the later ruthenium work the authors modelled a deformed liposome as a cylinder with hemi-spherical caps, assumed that the lipids redistributed evenly in the deformed liposome, and determined that the reduced LD, LDr = LD/isotropic absorbance, to be:

LDr = 3S/4 (1 - 3cos2 αi (1)

where αi is the angle the transition moment of interest makes with the normal to the cylinder surface (i.e. to the lipid long axis), S is the orientation factor that denotes the fraction of the liposome that is oriented as a cylinder perfectly parallel to the flow direction.

This equation can be formally derived as follows. Consider an ellipsoidal liposome under the flow conditions to be approximated by a cylindrical bilayer, whose lipids are perpendicular to the long axis of the cylinder, capped by two hemispheres. Analytes oriented within the hemispherical caps, assuming an average uniform distribution of analytes, will have no net LD, and can therefore be ignored.

The LD experiment defines one axis system {x, y, z}, where z is the orientation axis (the long axis of the cylinder) and the x/y plane is perpendicular to this (Figure 1). In order to determine the LD, it is useful to use another axis system {X, Y, Z} defined by the bilayer and an analyte. Z is the long axis of the cylinder (so z = Z) and X is the normal to the cylinder surface that passes through the origin of the analyte.

The analyte orientation will not be affected by the shear flow (the forces are too small), so on average any analyte transition moment, μi, will be uniformly distributed about the X axis. Let αi be defined as above. Let βi be the angle between the projection of μi onto the Y/Z plane and the Z axis. Thus in the {X, Y, Z} coordinate system

μi = μ(cos ai, sin βi αi, cos βi, sin αi)XYZ (2)

where μ is the magnitude of the transition moment and β is the angle the YZ projection of μ makes with the Z axis.

The LD by definition:

LD = (Az - Ay) = S(μz2 - μy2 (3)

where μz is the z component of the transition dipole in the {x, y, z} coordinate system and, in this case, also in the {X, Y, Z} coordinate system. may be written as the dot product of the transition moment vector and the vector for the y axis in the {X, Y, Z} coordinate system. Thus

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (4)

where γ can take any value from 0 to 2π. Since the isotropic absorbance is μ2/3, we can write

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

Both βi and γi can take any value from 0 to 2π so upon averaging over them:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)

as given by Ardhammar et al. Two special cases of interest are when a transition moment is parallel to the cylinder normal (the lipids), for which αITL = 0 and LDr/3S = -1/2, and when it is perpendicular to that, e.g. on the surface of the cylinder, for which α = 90° and LDr/3S = +1/4.

3 MATERIALS AND METHODS

The solvents were all analytical grade obtained from BDH laboratory supplies and were dried where necessary. Other chemicals including soybean lipid (L-α-Lecithin, Type IV-S: from soybean approx. 40% (TLC)), Gramicidin D from Bacillus Brevis and cytochrome c (Horse heart muscle) were obtained from Sigma Chemical Company and were used as received. Lyso-l-myristoyl-sn-glycero-3-phosphocholine (LMPC) and 1-palmitoyl-2-oleoyl-phosphatidylcholine were obtained from Avanti Polar Lipids. Re-PsbW was made according to the methods in reference. The humanised monoclonal antibody was prepared as described in reference.

The methods were used to create the membrane mimicking environments for these experiments depended on the sample to be studied and the size of membrane structure required. The smallest structures were produced by sonication as the final stage; the largest by vortexing the lipid preparations (this results in large multi-lamellar liposomes as shown by microscopy; unilamellar liposomes were produced from multi-lamellar liposomes by putting them through a LiposoFast Basic Extruder (with a polycarbonate membrane) from 12 and 25 times following reference). In general lipids and analytes were in molar ratio approximately 50:1 for pyrene and ~5 mg/mL lipid to ~1 mg/mL protein for the proteins, though in some cases further dilution was required to avoid excessive absorbance. If the analyte was not water soluble, it was generally simplest to mix lipid and analyte in chloroform. The solvent was then evaporated and kept under vacuum for 24 hours. To the dry lipid mixture was added 5 mM phosphate buffer (PH 7) to a final concentration of lipid ~5 mg/mL. Then the sample was sonicated or vortexed or extruded. If the analytes were water soluble they were added with the buffer or to the liposome solution after this had been prepared. Although Gramicidin is not very water soluble, in this case it was introduced to the membrane by treating as if it were water soluble and extruding the sample which seemed to facilitate its solubilisation by the lipids. Alternatively, Gramicidin was added to a liposome preparation from an ethanol solution. The LD experiments were performed on newly prepared samples.

UV-visible absorbance spectra were recorded using a Cary 1E spectrophotometer. Flow linear dichroism spectra were recorded using a Jasco J-715 circular dichroism spectropolarimeter, which was adapted for flow LD measurements. The photomultiplier tube was moved into the sample compartment next to the LD cell to reduce artefacts due to scattered light using a housing fabricated for the instrument by European Chirality Services. The cell used was developed for this work and is described below. The rotation speed used in the experiments was ~1000 rpm. In each case the speed chosen was the maximum value that avoided significant bubble formation in the cell. An LD baseline was measured on the same sample but without any rotation.

With light scattering samples, this method gave a slightly better baseline correction than a water baseline. For non-scattering samples the same baseline was recorded for the LD sample in a non-spinning cell and for a water or other non-orientatable solution in a spinning cell.

The geometries for the determination of protein structural motif transition moments were constructed using the CHARMM program. Two α-helices were considered: the ideal geometry of Pauling and Corey with (φ = 48°, ψ = 57°) and the mean values observed in experimentally determined structures of (φ = -63°, ψ = -41°). For the p-strand a single geometry with (φ = -135°, ψ = 135°) was chosen. Two different geometries of gramicidin were also considered. For the α-helices and β-strand, constrained minimisations of the blocked peptides Amn-Ala200cbx were performed in the absence of solvent effects. A length of 200 amino acids was chosen to minimise end effects, and thus mimic an infinite helix or strand.

The polarisation calculations were performed using the Matrix Method in which the protein is considered to be a set of n chromophores, each with a characteristic set of excitation bands. These effectively form a basis for the transitions for the entire protein, and are allowed to couple through diagonalisation of the Frenkel Hamiltonian. This yields a set of energies (eigenvalues) and transition couplings (eigenvectors) for the protein transitions. In our calculations we consider only the n [right arrow] π* and π [right arrow] π* transitions for each peptide link in the protein backbone. To date, this approach is the most accurate for the calculation of protein circular dichroism from first principles. The polarisation tensor can be calculated using a sum over states

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)

where [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the transition dipole moment from state a to state b in the λ direction. This can be obtained from the eigenvectors. This approach allows the contribution to the polarisation from each individual transition to be considered and, in particular, allows the polarisation due to the n [right arrow] π* bands to be isolated.