Quasi-uncoupled rotational diffusion of phospholipid headgroups from the main molecular frame
Hanne S. Antila, Anika Wurl, O. H. Samuli Ollila, Markus S. Miettinen, Tiago M. Ferreira
aa r X i v : . [ c ond - m a t . s o f t ] S e p Quasi-uncoupled rotational diffusion of phospholipid headgroups from the main molecular frame
Hanne S. Antila, Anika Wurl, O. H. Samuli Ollila, Markus S. Miettinen, and Tiago M. Ferreira ∗ Department of Theory and Bio-Systems, Max Planck Institute of Colloids and Interfaces, 14424 Potsdam, Germany Institute for Physics, Martin-Luther University Halle-Wittenberg Institute of Biotechnology, University of Helsinki, 00014 Helsinki, Finland (Dated: September 16, 2020)Understanding the dynamics of phospholipid headgroups in model and biological membranes is of extremeimportance for an accurate description of the dipolar interactions occuring at membrane interfaces. One funda-mental question is to which extent these dynamics are coupled to an overall molecular frame i.e. if the dipoleheadgroup orientation distribution and time-scales involved depend on the structure and dynamics of the glyc-erol backbone and hydrophobic regions or if this motion is independent from the main molecular frame. Herewe use solid-state nuclear magnetic resonance (NMR) spectroscopy and molecular dynamics (MD) simulationsto show that the orientation and effective correlation times of choline headgroups remain completely unchangedwith 50% mol cholesterol incorporation in a phosphatidylcholine (PC) membrane in contrast to the significantslowdown of the remaining phospholipid segments. Notably, our results indicate that choline headgroups in-teract as quasi-freely rotating dipoles at the interface irrespectively of the structural and dynamical molecularbehavior in the glycerol backbone and hydrophobic regions to the full extent of headgroup rotational dynamics.
Cells rely on rather complex processes to synthetize andmantain specific locations of a myriad of different phospho-lipids in cellular membranes for compartmentalization, sig-naling and transport functions. Membrane composition is aresult of molecular evolution and a variety of phospholipidmembrane compositions are found in nature depending oncell types, organelles and species [1]. Among such diver-sity, one common feature to most if not all biological mem-branes is the ubiquitous dominant presence of di-acyl phos-phatidycholine (PC) and/or di-acyl phosphatidylethanolamine(PE) with nearly identical chemical structures correspondingto a glycerol backbone linked to two acyl chains and a nega-tively charged phosphate group bearing a positively chargedcholine head group, –CH CH N(CH ) +3 , or ethanolaminehead group, –CH CH NH +3 , respectively. Accordingly, thesurface of cell membranes is highly populated of phospho-lipid headgroup dipoles which contribute in a non-trivial wayto the membrane electrostatic potential.The P − –N + dipole orientation and dynamics has been thor-oughly investigated for a number of PC and PE bilayer sys-tems [2–19] . Among the previous investigations, one funda-mental question arises. To which degree is the orientationalbehaviour of the surface dipoles related to the molecular bodyat which they are attached? Two limiting cases may be con-sidered, (a) that the dipoles are nearly or fully uncoupled fromthe rigid-body acting as quasi-Keesom dipoles which are posi-tionally fixed but quasi-free to reorient according to local elec-trostatic interactions or (b) that the orientation and dynamictime-scales for headgroup orientation are largely affected bythe structure and dynamics of the phospholipid molecules as awhole and therefore dependent on e.g. time-scales of motionof the glycerol backbone and hydrophobic acyl chains.From a purely structural standpoint, an independence ofthe orientation of the headgroup from the hydrophobic regionstands out from analysing the previously measured NMR C–H bond order parameters, S CH = h / θ − i , where θ denotes the angle between the C–H bond and the bilayer normal and the angular brackets denote a time-average. NMRorder parameter values are the most accurate observables withatomistic resolution measured from phospholipid bilayers upuntil present. A close look on the large set of previously re-ported S CH values for the α and β positions of the cholineheadgroup, shows that for all the distinct PC bilayers at fullhydration measured in the liquid crystalline phase (L α ), the S β CH values lie on a range between -0.05 and -0.02, while S α CH is equal to +0.05 ± α order parameter fallswithin such extremely narrow range not only between differ-ent systems but remarkably also regardless of temperature. Astructural analysis based on the α and β order parameters isill-defined since a range of semirigid and mobile empiricalmodels can simultaneously fit the set of α and β order pa-rameters [8], however to highlight the stability of the cholineorientation/conformation over the different systems it is rel-evant to note that the wider range of the β order parametersmay be induced by a change in the α - β torsion angle of only2-3 ◦ [7] in a semi-rigid fully empirical model [5].Irrespectively of the molecular model considered, the con-stant value of the α C–H bond order parameter indicates thatthe structure of the headgroup is not affected by the molecularstructure in the acyl chain region which changes considerablyamong the different systems and with temperature. It is knownthough that the headgroup orientation is highly sensitive to adecrease of hydration [21], to the inclusion of charges [14]and molecular dipoles [23], to the presence of salt ions [24]and to the hydrostatic pressure [15], with S α CH ranging from-0.02 to +0.1. This is exemplified also in table I for DMPCat a hydration of approximately 10 water molecules per lipidwith an increase of | S α CH | to 0.07 ± T / ◦ C Phase S α CH − S β CH ref.DMPC 25-35 L α β α α α α α α w /n l =10 57 L α TABLE I: Previously published α and β C–H bond order parametersfrom H NMR spectroscopy for a number of phosphatidylcholinelamellar systems together with values reported here using H- Cdipolar recoupling on DMPC and POPC/cholesterol (1:1). All sys-tems were near to full hydration except for the DMPC samples at awater to lipid molar ratio approximatelly equal to 10. The | S CH | ac-curate values lie within ± fast such configuration space is spanned. To assess the effecton the time-scales of motion, NMR relaxation experimentscan be used to measure relaxation rates, e.g. the spin-latticerelaxation R and the spin-lattice relaxation in the rotatingframe R ρ , which depend on spectral density terms, J ( ω i ) ,where the relevant frequencies ω i depend on the strength ofthe magnetic field, experimental setup and nuclei used, andthe spectral density, J ( ω ) , is the Fourier transform of the ori-entation autocorrelation function of a given molecular-fixedaxis [18, 25–28]. Klauda et al. [18] compared the experi-mental P R dispersion of DPPC vesicles from 0.022 to21.1 T to all-atom molecular dynamics simulations using theCHARMM C27r force-field and found good agreement be-tween simulation and experiments. The motion predicted bythe simulation was then analysed with an often used relax-ation model for fitting relaxation dispersion data, containingwobble and axial rotation of the overall lipid body and fast in-ternal motion. Based on this analysis, Klauda and coworkerssuggested a partial uncoupled motion of the headgroup fromthe overall lipid body since though the model fitted extremelywell the acyl chain and glycerol backbone motions together itdid not extend successfully to the choline segments. The un-coupled motion in the CHARMM C27r model was suggestedto be due to a relatively free rotation around the P–O(-g )bond that connects the phosphate group to the glycerol back-bone based on the potential mean force profile for this torsionangle [18]. A free rotation around the g – g bond as alsobeen previously suggested by Seelig et al.[4] and indeed thedihedral torsion potential for this bond in CHARMM C36 (anupdate of CHARMM C27r for lipids) assumes free rotation.To which extent the motion is uncoupled can be investi-gated by observing how changes in the main molecular framemotional time-scales affect the headgroup dynamics. Roberts and coworkers have later shown experimentally that the phos-phorus dynamic time-scales are largely affected by the in-corporation of cholesterol presumably due to a slower wob-bling motion [29]. A free rotation around the P–O(-g ) bondproposed by Klauda et al. [18] (or around the g – g bond)would partially or even fully decouple the headgroup dynam-ics from the overall rotation of the phospholipid around thebilayer normal, nevertheless a slowdown of the wobble mo-tion timescale could still affect the headgroup motion. How-ever, if the torsion around the P–O(-g ) bond is relatively free,a torsion around the P–O(- α ) bond and possibly around otherbonds to maintain the choline dipole orientation at some pref-erential range of angles may also be considered. Such a setof rotations would potentially decouple the dynamics of theheadgroup both from the overall rotation around the molecu-lar main frame as well as from the wobble motion and could beresponsible for the orientation response of the choline dipoleto the local electrostatic interactions. Such a molecular frame-work would fit in the quasi-Keesom hypothesis.Here we address the effect of the wobble and rotationalslowdown of the main molecular frame on the dipolar head-group using our previously reported methodology to translate S CH , R and R ρ values into C–H bond effective correlationtimes [28]. Figure 1 shows how the C R and R ρ , τ e ob-servables of POPC multilamellar vesicles change with the in-corporation of cholesterol. The experiments were done undera static magnetic field inducing a Larmor frequency equal to500 MHz for H and a spin lock field for R ρ of 50 kHz. Thefits used to determine the presented values are given as supple-mentary information. Note that a value of 50 kHz for the spinlock field ensures that R ρ is only sensitive to motions withinthe fast motion regime and that contributions from possiblecollective motions and from diffusion over vesicles can be ne-glected [28]. The striking observation is that both the R , R ρ and τ e values for the α and β segments remain exactly thesame within experimental uncertainty in contrast to the glyc-erol backbone slowdown, here quantified to be approximatelytwo times slower. This result is remarkable since it shows thatthe headgroup reorientation motion is not only partially un-coupled from the glycerol backbone as previously suggestedbut that the dipole motion of PC headgroups is fully indepen-dent of the overall motion of the main molecular frame andtherefore insensitive to changes in the glycerol backbone mo-tional time-scales and the presence of cholesterol in the bi-layer. Previously, C R ρ measurements have been reportedshowing a slight increase of R ρ for the α and β segments in-duced by cholesterol incorporation in DMPC bilayers. How-ever, it is hard to judge how statistically significant were thosechanges since no error bars have been reported [30].In addition to the experimental demonstration presented,we show that among four widely used MD force fields Slipids,CHARMM C36, MacRog and Berger lipids, only CHARMMC36 and Slipids predicts this result. By comparing the seg-mental effective correlation times with the experimental ones,nearly perfect quantitative agreement within the experimen-tal and simulation uncertainties is obtained for the Slipids and -16 -14 -12 -100123 R -16 -14 -12 -100246-16 -14 -12 -100246 R -16 -14 -12 -100204060-16 -14 -12 -1000.51 e -16 -14 -12 -100510 γ β α g3 (cid:0)2 (cid:1) γ β α (cid:2)(cid:3) (cid:4)(cid:5) (cid:6) γ β α (cid:7)(cid:8) (cid:9)(cid:10) (cid:11) POPCPOPC+cholOP OOO - CC CHH H HN(CH ) +g C C OHO H HR R g g α β γ R / s - R (cid:12) / s - τ e / n s segments FIG. 1: Effect of cholesterol on the C spin-lattice relaxation R ,spin-lattice relaxation in the rotating frame R ρ , and on the effectivecorrelation times τ e of the different segments in the headgroup andglycerol backbone of POPC. CHARMM-C36 force fields. The MacRog and Berger forcefields predict a slowdown of the headgroup with incorporationof cholesterol and fail to provide good quantitative estimatesof τ e for both the systems with and without cholesterol, in-dicating that these force-fields include an erroneous couplingof the headgroup with the main molecular body. A compari-son of simulated and experimental R and S CH values is alsogiven as SI. We are now planning a number of experimentsand simulations on PC and PE systems to investigate such be-haviour with more detail focusing on detailed conformationalchanges.The experimental results (and the more realistic simula-tions) support the quasi-Keesom hypothesis considered abovewhere the dipole orientation fully depends on the local electro-static interactions while fully uncoupled from the rest of the POPCPOPC+cholest erol E xp . S li p i d s C HA R MM C M ac R og B e r g e r γβ (cid:13) g3g2g1 E xp . S li p i d s C HA R MM C M ac R og B e r g e r τ e / n s τ e / n s τ e / n s FIG. 2: Comparison of the effective correlation time profiles pre-dicted by the Slipids, CHARMM C36, MacRog and Berger force-fields for the headgroup and glycerol backbone with the experimentalvalues measured. lipid molecular body. As noted above, it is well known thatthe headgroup conformation/orientation can react strongly tothe inclusion of charges [20] or dipoles [23] in the membranewith a consequent variation of the α order parameter as wellas to the level of hydration [21, 22], salt concentration [24],and hydrostatic pressure [15]. It is yet to be investigated howsuch changes on the orientation distribution affect the head-group dynamic time-scales.The molecular description here presented has rather strongimplications for membrane biophysics and should motivate anumber of additional experiments and simulations. It impliesthat the dipolar surface of bilayers with lipids having the sameheadgroup but distinct acyl chains will have additionally to thesame headgroup orientation the same dynamic time-scales, aswell in mixtures of these phospholipids with cholesterol. Thedynamics of dipolar headgroups in bilayers consisting of mix-tures of phospholipids with different headgroups may also beindependent of the acyl chains involved and dominated by thedipole-dipole interactions between the distinct dipoles. Thisneeds however to be tested since e.g. PE headgroups havea different orientation than PC headgroups at full hydrationand may therefore have a different behaviour. The resultspresented here do not support the assumption in the umbrellamodel that headgroups reorient when cholesterol is present inthe bilayer in order to shield the cholesterol interfacial crosssection from water molecules[31].In summary, our results suggest that for describingthe dipolar interactions at the surface of membranes, thehydrophobic structure may be neglected to a good approx-imation and that the relevant headgroup physics lie on theelectrostatic interactions, which would be remarkably usefulconsidering the complex molecular arrangement in thehydrophobic region of biological membranes.TMF greatly acknowledges Kay Saalw¨achter and AlexeyKrushelnitsky for invaluable support and discussions. ∗ [email protected][1] D. Marsh, Handbook of Lipid Bilayers (CRC press, New York,2013).[2] R. Griffin, J. Am. Chem. Soc. , 851 (1976), URL https://doi.org/10.1021/ja00419a044 .[3] S. Kohler and M. Klein, Biochemistry , 519 (1977), URL http://dx.doi.org/10.1021/bi00622a028 .[4] H. Gally, W. Niederberger, and J. Seelig,Biochemistry , 3647 (1975), URL https://doi.org/10.1021/bi00687a021 .[5] J. Seelig, H.-U. Gally, and R. Wohlgemuth,BBA - Biomembranes , 109 (1977), URL .[6] R. Griffin, L. Powers, and P. Per-shan, Biochemistry , 2718 (1978), URL https://doi.org/10.1021/bi00607a004 .[7] M. Brown and J. Seelig, Biochemistry , 381 (1978), URL https://doi.org/10.1021/bi00595a029 .[8] R. Skarjune and E. Oldfield, Biochemistry , 5903 (1979),URL https://doi.org/10.1021/bi00593a022 .[9] S. Rajan, S. Kang, H. Gutowsky, and O. E,J Biol Chem. , 1160 (1981), URL .[10] D. Siminovitch, M. Rance, and K. Jeffrey, FEBSLetters , 79 (1980), ISSN 0014-5793, URL .[11] T. Rothgeb and E. Oldfield, J BiolChem. , 6004 (1981), URL .[12] R. Ghosh, Biochemistry , 7750 (1988), URL https://doi.org/10.1021/bi00420a025 .[13] M. Milburn and K. Jeffrey, Biophys J. , 543 (1989), URL https://doi.org/10.1016/S0006-3495(89)82701-8 .[14] P. Macdonald, J. Leisen, and F. Marassi,Biochemistry , 3558 (1991), URL http://dx.doi.org/10.1021/bi00228a029 .[15] B. Bonev and M. Morrow, Biophys J. , 518 (1995), URL https://doi.org/10.1016/S0006-3495(95)79925-8 .[16] M. Roberts and A. Redfield, , 17066 (2004).[17] M. U. Roberts MF, Redfield AG,Biophys J. , 132 (2009), URL https://doi.org/10.1016/j.bpj.2009.03.057 .[18] J. Klauda, M. Roberts, A. Redfield, B. Brooks,and R. Pastor, Biophys J. , 3074 (2008), URL .[19] J. Doux, B. Hall, and J. Killian, Bio-phys J. , 1245 (2012), URL http://dx.doi.org/10.1016/j.bpj.2012.08.031 .[20] P. Macdonald, J. Leisen, and F. Marassi,Biochemistry , 3558 (1991), URL https://doi.org/10.1021/bi00228a029 .[21] A. Ulrich and A. Watts, Biophys. J. , 1441 (1994), URL https://doi.org/10.1016/S0006-3495(94)80934-8 .[22] B. Bechinger and J. Seelig, Chemistryand Physics of Lipids , 1 (1991), URL .[23] B. Bechinger and J. Seelig, Biochemistry , 3923 (1991), URL https://doi.org/10.1021/bi00230a017 .[24] C. Altenbach and J. Seelig, Biochemistry , 3913 (1984), URL https://doi.org/10.1021/bi00312a019 .[25] M. Brown, J. Seelig, and U. H¨aberlen, J. Chem. Phys. , 5045(1979), URL https://doi.org/10.1063/1.437346 .[26] M. Brown, A. Ribeiro, and G. Williams, Proc.Natl. Acad. Sci. USA , 4325 (1983), URL https://doi.org/10.1073/pnas.80.14.4325 .[27] C. Morrison and M. Bloom, J. Chem. Phys. , 749 (1994),URL https://doi.org/10.1063/1.468491 .[28] T. Ferreira, O. Samuli, R. Pigliapochi, A. Dabkowska, andD. Topgaard, J. Chem. Phys. , 044905 (2015), URL https://doi.org/10.1063/1.4906274 .[29] V. Sivanandam, J. Cai, A. Redfield, and M. Roberts,J. Am. Chem. Soc. , 3420 (2009), URL https://doi.org/10.1021/ja808431h .[30] C. Le Guernev´e and M. Auger, Bio-phys J. , 1952 (1995), URL https://doi.org/10.1016/S0006-3495(95)80372-3 .[31] J. Dai, M. Alwarawrah, and J. Huang, J.Phys. Chem. B , 840 (2010), URL https://doi.org/10.1021/jp909061h .[32] L. L¨oser, K. Saalw¨achter, and T. Ferreira, Phys.Chem. Chem. Phys. , 9751 (2018), URL http://dx.doi.org/10.1039/C8CP01012A .[33] T. Ferreira, F. Coreta-Gomes, O. Ollila, M. Moreno, W. L. Vaz,and D. Topgaard, Phys. Chem. Chem. Phys. , 1976 (2013),URL http://dx.doi.org/10.1039/C2CP42738A . SUPPLEMENTARY INFORMATIONMETHODSSample Preparation
The phospholipids 1-palmitoyl,2-oleoyl- sn -glycero-3-phosphocholine (POPC) and 1,2-dimyristoyl- sn -glycero-3-phosphocholine (DMPC), cholesterol and chlorophormwere purchased from Sigma-Aldrich. The samples wereprepared by mixing the lipids with chlorophorm and rapidlyevaporating the organic solvent under a nitrogen gas flow andsubsequently drying the lipid film under vacuum overnight.The film was then hydrated in a 0.5 ml EPPENDORF tubeby adding 40 %wt of water and manually mixing with athin metal rod multiple times alternated by sample centrifu-gation until a homogeneous mixture was visually attained.The resulting mixture was then centrifuged into a KEL-FBruker insert with a sample volume of approximatelly 25 µ lspecifically designed for solid-state NMR 4mm rotors.To obtain the low hydration sample, a glass tube containinga DMPC film of 20 mg was left for 1 day in a desiccator (vol-ume of approx. 1l) together with a glass tube containing 2 mlof water under reduced pressure. The water content was thendetermined from integrating the water and γ peaks in the Hspectra.
NMR Experiments
The R and R ρ experiments were performed on a a BrukerAvance II-500 NMR spectrometer operating at a C Larmorfrequency of 125.78 MHz equipped with a a E-free CP-MAS4 mm (13C/31P/1H). The R-PDLF measurements were per-formed on a Bruker Avance III 400 spectrometer operating at a H Larmor frequency of 400.03 MHz equiped with a standard4 mm CP-MAS HXY probe. All experiments were performedunder magic-angle spinning conditions at a rate of 5 kHz. Theprocessing of all NMR data was done with MATLAB 2018b.The R-PDLF, R and R ρ experiments were performed as pre- viously described in references [28, 32].The parameters used were the following. R-PDLF experi-ments : a total of 32 points in the indirect dimension with in-crements equal to two R18 blocks; SPINAL64 was used forproton decoupling during C acquisition, with a nutation fre-quency of approximately 50 kHz, a total acquisition time of0.07 s and a spectral width of 200 ppm; the rINEPT pulseswere set at a nutation frequency of 78.12 kHz. R and R ρ experiments : RF π/ and π pulses were set to a nutation fre-quency of 63.45 kHz. TPPM was used for proton decouplingduring C acquisition, with a nutation frequency of approxi-matelly 50 kHz, a total acquisition time of 0.1 s, recycle delayof 10 s and a spectral width of 140 ppm. The spin-lock fre-quency for R ρ was 50 kHz.For determining R and R ρ for a given carbon segment,we determined the decay over the indirect dimension by fit-ting gaussian lineshapes in the direct dimension and using theanalytic areas of the fitted functions. The decay was then fit-ted with a single exponential decay and the error bounds forboth the R and R ρ values presented are the 95 % confidencebounds from these fits. For estimating τ e we used [28], τ e = 5 R ρ − . R π d N (1 − S ) (1)where the rigid coupling constant d CH value used was 20 kHzand the S CH values used were taken from reference [33]. Theerror, ǫ , for the effective correlation times calculated then be-comes, ǫ ( τ e ) = 5 ǫ ( R ρ ) + 3 . ǫ ( R )4 π d N (1 − S ) (2)For determining the coupling constants with R-PDLF spec-troscopy presented in table I, a fit of the time domain data wasdone by using time domain profiles from numerical simula-tions of the R-PDLF pulse sequence that take into account the B inhomogeneity of the used CP-MAS probe. This proce-dure gives an accuracy which is about ten times higher thanusing frequency domain data and will be described elsewhere.
27 °C S CH = 0.05 0 0.005 0.01 0.015-0.500.51
27 °C S CH = 0.0660 0.005 0.01 0.015-0.500.51
60 °C S CH = 0.032 0 0.005 0.01 0.015-0.500.51
60 °C S CH = 0.066 t (cid:14) s (cid:15) (cid:16) s (cid:17) (cid:18) s (cid:19) (cid:20) s ββ αα M ( (cid:21) ) /(cid:22)((cid:23)(cid:24)(cid:25)(cid:26)(cid:27))(cid:28)(cid:29)(cid:30)(cid:31) FIG. 3: R-PDLF time domain profiles from a sample of DMPC at low hydration at two different temperatures fitted with R-PDLF numericalsimulations taking into account the B inhomogeneity of the CP MAS probe used. The uncertainty of | S CH | is below ± -1 R =2.29 0.05R =2.02 0.06 -1 R =2.33 0.17R =2.23 0.19 -1 R =2.47 0.12R =2.38 0.18 -1 R =3.63 0.58g R =3.07 1.52 -1 R =4.43 0.68g R =4.99 0.77 -1 R =2.67 0.26g R =2.25 0.29 -1 R =2.52 0.18R =2.36 0.2 -1 R =5.23 0.5R =4.95 0.5 -1 R =25.18 2.26R =45.19 10.54 -1 R =14.93 1.49R =34.31 4.68 -1 R =4.34 0.25R =4.39 0.42 -1 R =11.31 0.68R =18.16 2.03 ! " s R ρ e R FIG. 4: The R and R ρ decays measured for POPC (blue) and POPC/cholesterol (red) systems. Each point corresponds to the integraldetermined from a gaussian fit of the corresponding C peak in the high resolution chemical shift spectrum acquired under MAS of 5 kHz.The spin lock field for the R ρ measurement was 50 kHz and the13